Encyclopedia of Electrochemistry 10 Volume Set

1
1
Electrode potentials
Oleg A. Petrii, Galina A. Tsirlina

Department of Electrochemistry, Moscow State University, Moscow, Russia
1.1 Introductory Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

1.2 External and Internal Potentials . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3 Electromotive Force (emf) and Gibbs Energy of Reaction . . . . . . . . 5
1.4 Diffusion Potential Drop . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.5 Classification of Electrodes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.6 Membrane Electrodes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.7 Standard Potentials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.8 Specific Features of Certain Reference Electrodes . . . . . . . . . . . . . 13
1.9 Values of Redox Potentials . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.10 Potential Windows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.11 Experimental Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.12 Effects of Ion Pairing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
1.13 Absolute Electrode Potential . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.14 The Role of Electric Double Layers . . . . . . . . . . . . . . . . . . . . . . . 20
1.15 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3
1.1 electrochemistry. Electrodes can be based

Introductory Remarks on metals, alloys, any type of semicon-
ductors (namely, oxides and salts with
The term ‘‘electrode potential’’ combines electronic or mixed conduction, electron
two basic notions: ‘‘electrode’’ and ‘‘poten- conducting polymers, various covalent
tial’’. The electrical potential ϕ, as known compounds of metals), and also on a
from the physical definition, represents variety of composite materials. The cor-
the electrical energy (work term), which responding ionic conductors are usually
is necessary for transferring a unit test electrolyte solutions or melts, solid elec-
charge from infinity in vacuum into the trolytes (particularly, amorphous and poly-
phase under consideration. We call this meric materials), supercritical fluids, and
charge as ‘‘probe’’ (sometimes also ‘‘test’’ various quasiliquid systems. In certain
or ‘‘imaginary’’) in order to emphasize cases, the terms ‘‘electrode’’ and ‘‘electrode
that it is affected only by the external field potential’’ are applied also to semiperme-
and does not interact with the medium able membranes – the systems separating
via non-Coulombic forces, and, moreover, two solutions of different composition (al-
is sufficiently small to induce any charge though these ‘‘membrane electrodes’’ do
redistribution inside the phase. In actual not exactly satisfy the definition given
practice, the charges exist only in a com- above).
bination with certain species (elementary Real electrodes sometimes represent ex-
particles, particularly, electrons, and ions). tremely complicated systems, which can
Hence, the value of ϕ appears to be beyond include several interfaces, each locating
the reach of experimental determination, a certain potential drop. These inter-
a fact that poses the problems concerned faces, together with the phases in contact,
with interpretation of the electrode po- take part in the equilibria established
tential and brings to existence numerous in the electrochemical system. This is
potential scales. A number of discussions why the electrode potential is one of
of various levels can be found in the litera- the most important notions of electro-
ture [1–20]. chemical thermodynamics (to say nothing
A system that contains two (or more) about its role in electrochemical kinetics),
contacting phases, which includes at least which describes the equilibrium phenom-
one electronic and one ionic conductor, ena in the systems containing charged
is usually considered as an electrode in components. The term ‘‘component’’ as
4 1 Electrode potentials
applied to the charged species does not 1.2

mean that the ions are considered as the External and Internal Potentials
components amenable to Gibbs’ phase
rule. The electrical state of any phase α can
Consideration of equilibria in the sys- be characterized by its internal potential,
tems of this sort requires the applica- which is a sum of the external (or outer)
tion of the notion of electrochemical potential ψ α induced by free electrostatic
potential introduced by Guggengheim charges of the phase and the surface
[21] on the basis of the relationship potential χ α [6, 23–26]:
for the electrochemical Gibbs free en-
ergy Ḡ: ϕα = ψ α + χ α (4)
When the free electrostatic charge in

dḠ = −S dT + V dp + µi dNi phase α turns to zero, ψ α = 0 and
i
ϕ α = χ α . The surface potential of a liquid
+F zi ϕ dNi (1) phase is dictated by a certain interfacial
i orientation of solvent dipoles and other
molecules with inherent and induced
where S is the entropy, V is the volume dipole moments, and also of ions and
of the system, T and p are its temperature surface-active solute molecules. For solid
and pressure, respectively, µi is the phases, it is associated with the electronic
chemical potential of the ith species, gas, which expands beyond the lattice (and
Ni is the number of moles of the also causes the formation of a dipolar
ith component, zi is the charge of the layer); other reasons are also possible.
ith species taking into consideration its A conclusion of fundamental impor-
sign, and ϕ is the electrical potential tance, which follows from Eq. (3), states
in the location point of the ith species, that the electrical potential drop can
which is also called the internal (or be measured only between the points,
inner) potential of the corresponding phase which find themselves in the phases of
[22–26]. one and the same chemical composition.
According to Eq. (1), the electrochemical This conclusion was first formulated by
β
potentials of the components, µ̄i , are Gibbs. Indeed, in this case, µi = µαi and
characteristic values for the ith species β
µ̄ − µ̄αi
β
in any phase α under consideration: α ϕ = ϕ β − ϕ α = i . Otherwise,
zi F
when the points belong to two different
∂ Ḡ phases, the experimental determination of
µ̄αi = µαi + zi F ϕ α = (2)
∂Ni p,T ,Nj =i the potential drop becomes impossible.
Furthermore, these arguments imply that
Only these values or the differences the difference between the internal poten-
β
of electrochemical potentials referred to tials at the interface, α ϕ, cannot be mea-
phases α and β β
sured. The quantity α ϕ is called Galvani
potential and determines the electrostatic
β β
µ̄i − µ̄αi = (µi − µαi ) + zi F (ϕ β − ϕ α ) component of the work term correspond-
(3) ing to the transfer across the interface,
are experimentally available. whereas the ‘‘chemical’’ component is
1.3 Electromotive Force (emf) and Gibbs Energy of Reaction 5
determined by the difference between the real species present in any solution [27]; in
chemical potentials. In the literature, the aqueous and certain other media, their
term Galvani potential is also applied to equilibrium concentration is extremely
the separate value of internal potential. low, which, however, does not prevent us
However, it is not significant, because of from formally using the thermodynamic
the relative nature of this value, which is consideration of the electrode potential in
discussed below. the framework of the concept of electronic
The electrochemical equilibrium re- equilibria. However, in real systems the
quires the equality of electrochemical exchange current density for Mz+ /M pair
potentials for all components in the con- appears to be so small in the limit under
tacting phases. From this condition of elec- consideration that equilibrium cannot be
trochemical equilibrium, the dependence maintained, and Eq. (5) no longer applies.
of Galvani potential on the activities of The equilibrium with participation of
potential-determining ions can be derived, solvated electrons can be expressed by a
which represents the Nernst equation [27] scheme
for a separate Galvani potential. That is,
for an interface formed by a metal (M) M (solid) ⇐⇒ M (in solution)
and solution (S) containing the ions of this ⇐⇒ Mz+ + zēs (6)
metal Mz+ ,
The dissolved metal atoms can be con-
RT
M
S ϕ
M S
= ϕ − ϕ = const + ln aMz+ sidered as an electrolyte, which dissociates
zF producing Mz+ cations and the sim-
(5)
where ϕ M and ϕ S are the internal poten- plest anions, ēs . Actually, the condition
tials of metal and solution, respectively, aMz+ = 0 cannot be achieved, because a
and aMz+ is the activity of metal ions in finite value for the ‘‘solubility product’’
solution. of metal exists, and, simultaneously, the
The impossibility to measure a separate condition of electroneutrality is valid. By
Galvani potential rules out the possibil- using the condition of the equality of elec-
ity of establishing a concentration corre- trochemical potentials of electrons in the
solution and in the metal phases, we ob-
sponding to M S ϕ = 0.
A formal consideration of Eq. (5) pre- tain
dicts the infinitely large limiting value of
ē − µē
µM S
RT
M
S ϕ = = const − ln aē
the Galvani potential (M S ϕ → −∞) when F F
the activity of metal ions approaches zero (7)
(a hypothetical solution containing no ions where µM ē and µ S are the chemical
ē
of this sort, or pure solvent). However, potentials of electrons in metal and
this is not the case, because the limit solution, respectively.
of thermodynamic stability of the solvent
will be exceeded. For an ideal thermo-
dynamically stable (hypothetical) solution, 1.3
this uncertainty can be clarified, if the Electromotive Force (emf) and Gibbs
electronic equilibria between the metal Energy of Reaction
and the solvent are taken into account,
which requires the consideration of sol- To study the electrical properties of the M|S
vated electrons (ēs ) [7–11]. The latter are interface presented schematically in Fig. 1,
Fig. 1 A scheme of potential drops at

∆M
Sψ metal|solution interface:
MS ϕ – metal/solution Galvani
potential; χ S and χ M the
χS χM solution/vacuum and metal/vacuum
surface potentials, respectively;
MS ψ – metal/solution Volta potential.
S M
∆M
Sϕ

it is necessary to construct a correctly M M
E = M11 ϕ = M1 ϕ + SM1 ϕ + MS ϕ
connected circuit (Fig. 2) – a galvanic cell
(9)
satisfying the condition of identical metal
which corresponds to the route of probe-
contacts at both terminals. For the system
charge transfer 1 ⇒ 2 ⇒ 3 and consists
under discussion, it is a circuit of the
of three Galvani potentials. On the other
following type:
hand, when going from M1 to M1 via the
M1 |S|M|M1

(8) route marked by points 1 ⇒ 2 ⇒ 3 ⇒ 4

(Fig. 2), we obtain

in which the electronic conductors M1 and
M M
M1 consist of the same material, differ- E = M11 ϕ = M S
S ψ + M1 ψ + M ψ
1
ing only in their electric states (all the (10)

interfaces are assumed to be in equilib- Hence, the potential discussed can be
rium). It is assumed that both terminals expressed by three values, which are called
of voltmeter used for measurement of the Volta potentials (in some cases, contact
potential consist of the same metal. It is potentials). They represent the potential
evident that the measuring device with ter- differences between the points just outside
minals also made of M1 records the total the phases M and S in vacuum, and are
difference of electric potentials, or emf, measurable quantities. The distances from
M1 M1′'
1′ 4′
3
3′
M
2′
2 Fig. 2 Illustration of two

possible routes of probe-charge
transfer in a correctly connected
1
circuit.
1.3 Electromotive Force (emf) and Gibbs Energy of Reaction 7

these points to the interface should be The interface M1 |M (circuit 1.8) which,
sufficiently large, as compared with the in the simplest case, represents a boundary
characteristic distance of molecular and of two metals, is easy to construct.
image forces (up to 10−4 cm), and, at the However, the M1 |S contact, which includes
same time, not too high to prevent the an additional electrochemical interface, is
weakening of interaction with the charges rarely feasible, being usually unstable and
inside the phases. Volta potential is a dependent on the nature of M1 . Generally,
measurable quantity, because the points in place of M1 , a special electrochemical
between which it is measured are located system, the so-called reference electrode
in the same phase and exclusively in the should be included into cell 1.8. The
field of long-range forces. This means that electrode potential can be determined
the Volta potential does not depend on the as the emf E of a correctly connected
charge of the probe. electrical circuit formed by the electrified
For the metal/metal boundary, these po- interface under discussion and a reference

M
tential differences (M1 ψ) can be defined electrode. According to this definition,
by the difference of work functions [22]. the potential of any reference electrode
The latter can be obtained from the pho- is assumed to be zero.
toelectron emission (or thermoelectron Thus, electrochemists deal with the val-
emission) and also directly by using a cir- ues of potentials, which are, actually, the
cuit differences of potentials. The nature of the
M1 |M|vacuum|M1 emf of a circuit for potential measurement
depends on the type of the electrodes. If
where M1 is the reference metal. The main we ignore a less frequent situation of ide-
experimental problems are the surface pre- ally polarizable electrodes, for the majority
treatment (purification) and the elimina- of systems, two half-reactions take place
tion of potential drop between the samples simultaneously in the cell. For ideally po-
of M1 and M in ultrahigh vacuum. larizable electrodes, their contribution to
Volta potential for the interface metal emf can be considered, in the first approx-
(mercury)/solution can be found by using imation, as the potential drop inside a ca-
the following cell reference electrode|solu-
pacitor formed by the electric double layer.
tion|inert gas|Hg|solution|reference elec-
Hence, this drop depends on the electrode
trode when the solution flows to the system
free charge (see Sects. 1.3 and 1.5):
through the internal walls of a vertical tube,
where mercury flows out via a capillary
Ox1 + n1 ē ⇐⇒ Red1
placed axially in a vertical tube and is bro-
ken into drops. These mercury drops carry (on the electrode under study), (11)
away the free charges, thus eliminating the
Ox2 + n2 ē ⇐⇒ Red2
potential difference in inert gas between
Hg and the solution. If a plate of solid (on the reference electrode). (12)
metal is used instead of liquid metal, it is
necessary to eliminate the potential drop Taken together, Eqs. (11) and (12) corre-
mentioned above. A similar technique can spond to the chemical reaction:
be used for the solution/solution inter-
face. There are also other techniques of ν1 Red1 + ν2 Ox2 ⇐⇒ ν2 Red2 + ν1 Ox1
Volta potential determination [1–6, 28]. (13)
the relationship for stoichiometric coeffi- molecules (salts, acids) in Eq. (17) can be
cients being substituted for partial activities of ions,
which cannot be measured experimentally.
n1 ν1 = n2 ν2 = n (14)
The applications of various types of
where n is the number of electrons electrochemical cells to chemical thermo-
transferred across the electrochemical cell dynamics are considered in Sect. 2.
in a single act of reaction (13). Experimentally, the emf can be mea-
From Eq. (3), the emf of the cell, sured either by compensating the circuit
which corresponds to reaction (13), can be voltage (classical technique which became
expressed as follows: rare nowadays) or by using a voltmeter
1 of very high internal resistance. The accu-
E= (ν1 µRed1 + ν2 µOx2 − ν2 µRed2 racy of emf determination of about 1 µV
nF
can be achieved in precise measurements,
− ν1 µOx1 ) (15)
whereas common devices provide the ac-
The right-hand term in parentheses curacy of about 1 mV. The potential unit
equals the Gibbs energy of reaction (13) named Volt, which is used in the modern
with the opposite sign. Hence, the emf literature (particularly, below), is the so-
of an electrochemical cell (a cell with called absolute (abs) Volt; it differs slightly
eliminated diffusion and thermoelectric from the international (int) Volt value. The
potential drops is implied, see later) corre- ratio abs/int is 1.00033. To determine the
sponds to the reaction Gibbs free energy: sign of emf, a conventional rule is adopted,
G which states that the left electrode should
E=− (16) be considered as the reference one.
nF
For carrying out the experimental mea-
This statement is true also for more surements of electrode potentials, a system
complicated reactions with the participa- chosen as the reference electrode should
tion of N reactants and products, and can be easy to fabricate, and also stable and
be generalized in the form of the Nernst reproducible. This means that any pair of
equation: reference electrodes of the same type fabri-
RT
N cated in any laboratory should demonstrate
E = E0 − νi ln ai (17) stable zero potential difference within the
nF
i=1 limits of experimental error. Additionally,
where the stoichiometric numbers νi < 0 the potential differences between two ref-
for the reactants and νi > 0 for the erence electrodes of different type should
products, ai are their activities, n is the remain constant for a long time.
total number of electrons transferred in Another point is that the transfer of
the coupled electrochemical reactions, and electricity (although of very low quan-
E 0 is the standard potential, which is tity) occurs in the course of emf mea-
expressed by the standard Gibbs free surements. Thus, the reference electrode
energy G0 as should comply with the requirement of
nonpolarizability: when the currents (usu-
G0 ally in the nanoampere range) flow across
E0 = − (18)
nF the system, the potential of the refer-
Under equilibrium conditions, the mean ence electrode should remain constant.
activities or the activities of neutral One of the most important features that
1.4 Diffusion Potential Drop 9
determines this requirement is the ex- In Planck’s model for a sharp boundary,
change current density i0 , which expresses for a 1,1 electrolyte,
the rates of direct and reverse processes
under equilibrium conditions. This quan- RT
ϕdiff = ln ξ (20)
tity determines the rate of establishment F
of equilibrium (the state of the electric where function ξ can be found from a
double layer is established very rapidly as transcendent equation in which ‘‘+’’ and
compared with the total equilibrium). ‘‘−’’ denote the sets, which consist of all
cations and of all anions, respectively, and
λ denotes the limiting conductivity of the
1.4 corresponding ion:
Diffusion Potential Drop
(2) (2) (1) (1)
ξ λ + c+ − λ + c+
The use of reference electrodes frequently + +
poses the problem of an additional poten- (2) (2)
(1) (1)
λ − c− − ξ λ − c−
tial drop between the electrolytes of the
− −
electrode under study and of the reference  (2) 
one. For liquid electrolytes, this drop arises ci
at the solution/solution interface (liquid  i 
. ln 
 (1) 
 − ln ξ
junction). The symbol .. conventionally de- ci
notes an interface of two solutions with i
=  (2) 
a diffusion potential drop in between; if ci
this drop is eliminated (see later), then the  i 
.. ln 

 + ln ξ
(1) 
symbol .... is used. In this case, the equilib- ci
rium is not exact because of the existence i
of a diffusion potential drop ϕdiff . The (2)
(1)
latter has the meaning of Galvani poten- ξ ci − ci
i i
tial and cannot be measured; however, it × (2)
(1)
(21)
can be estimated by adopting a model ap- ci −ξ ci
proach to the concentration distribution i i
of ions in the interfacial region, the mod-
The Henderson equation that has gained
els of Planck [29] and Henderson [30, 31]
wider acceptance can be written as follows
being the most conventional. The general
for concentrations c having the units of
expression for ϕdiff at the interface of
normality:
liquid solutions (1) and (2) is as follows:
λi (2)
RT (2) ti (1)
(ci − ci )
ϕdiff = − d ln ai (19) RT i zi
F (1) zi ϕdiff = (2) (1)
F λi (c − c ) i i
where ti is the transport number of the
i
ith ion, that is, the portion of current (1)
transferred by this ion through the solu- λ i ci
tion. In the first approximation, ϕdiff can i
× ln (2)
(22)
be estimated by substituting correspond- λ i ci
ing concentrations for the partial activities. i
The solution is given for the case of a electrode classification based on the nature
smeared out boundary and linear spatial of species participating in electrochemical
distributions of concentrations. equilibria.
Generally, Eqs. (20)–(22) yield similar Electrodes of the first kind contain elec-
results; however, for junctions with a tronic conductors as the reduced forms,
pronounced difference in ion mobilities and ions (particularly, complex ions) as
.
(such as HCl .. LiCl), the deviation can the oxidized forms. The equilibrium can
reach about 10 mV. A specific feature of be established with respect to cations and
the Planck equation is the existence of anions; in the absence of ligands, the
two solutions, the first being close to that cations are more typical. The examples
of Henderson, and the second one being are: Cu|Cu2+ . . . or . . . Au|[Au(CN)2 ]− . . . .
independent of the solution concentration This group can be supplemented also
and of no physical meaning [32]. by amalgam electrodes (or other liquid
In practice, in place of model calcula- electrodes) and electrodes fabricated from
tions and corresponding corrections, the nonstoichiometric solids capable of chang-
elimination of the diffusion potential is ing their composition reversibly (interca-
conventionally applied. This is achieved lation compounds based on carbons, ox-
by introducing the so-called salt bridges ides, sulfides, and multicomponent salts,
filled with concentrated solutions of salts, particularly, Li-intercalating electrodes of
which contain anions and cations of close batteries).
transport numbers. A widely known ex- Electrodes of the second kind contain a
ample is saturated KCl (4.2 M); in aqueous layer of a poorly soluble compounds (salt,
solutions, potassium and ammonium ni- oxide, hydroxide), which is in contact with
trates are also suitable. However, the a solution of the same anion. The equi-
requirement of equal transport numbers librium is always established with respect
is less important as compared with that of to anions. The most typical examples are
high concentration of electrolyte solution, based on poorly soluble compounds of
which fills the bridge [33, 34]. A suitable mercury and silver (Table 1).
version of the salt bridge can be chosen The redox polymers, both of organic and
for any type of cells, when taking into ac- inorganic origin (such as polyvinylpyridine
count the kind of studies and the features modified by redox-active complexes of met-
of chosen electrodes. als; Prussian blue and related materials),
In melts, an additional problem of can be considered as a version of electrodes
thermo-emf arises, and the emf correction of the second kind; however, the equilib-
can be calculated from thermoelectric rium is usually established with respect
coefficients of phases in contact [35]. to cations. Electron conducting polymers
(polyanyline, polypyrrol, and so forth) also
pertain, in the first approximation, to the
1.5 electrodes of the second kind, which main-
Classification of Electrodes tain equilibrium with respect to anion.
Ion exchange polymer films on electrode
No universally adopted general classifica- surfaces form a subgroup of membrane
tion of electrodes exists; however, when electrodes.
dealing with thermodynamic aspects of the Rather exotic electrodes of the third
electrode potential notion, we dwell on the kind (with simultaneous equilibria with
1.5 Classification of Electrodes 11
Tab. 1 Conventional reference electrodes [36]
Reference Potential versus Analogs Media

electrode SHE (aqueous systems,
recommended values
for 25 ◦ C)
[V]
Calomel electrodes Mercurous Aqueous and

bromide, iodide, mixed (with
iodate, acetate, alcohols or
oxalate electrodes dioxane)
Saturated (SCE) 0.241(2)
Normal (NCE) 0.280(1)
decinormal 0.333(7)
Silver chloride 0.197(6) Silver cyanide, Aqueous, mixed,
electrode oxide, bromate, abs. alcoholic
(saturated KCl) iodate,
perchlorate
Nitrate Aprotic
Mercury/mercurous 0.6151(5) Ag/Ag2 SO4 , Aqueous, mixed
sulfate electrode Pb/Pb2 SO4
Mercury/mercuric 0.098 Aqueous, mixed
oxide electrode
Quinhydrone Chloranil, 1,4-nap- Any with
electrode htoquinhydrone sufficient
solubility of
components
0.01 M HCl 0.586(8)
0.1 M HCl 0.641(4)
Note: NCE: Normal calomel electrodes; SCE: saturated calomel electrode.
respect to anions and cations) were also Au|[Fe(CN)6 ]3− , [Fe(CN)6 ]4− · · ·
proposed for measuring certain special
characteristics, such as the solubility prod- or Pt|[Co(EDTA)]− , [Co(EDTA)]2− · · ·
uct of poorly soluble salts. The rare When equilibrium is established be-
examples are Ag|AgCl|PbCl2 |Pb(NO3 )2 . . . tween ions in solution and the gas phase,
or M|MOx|CaOx|Ca(NO3 )2 . . . where M = we have gaseous electrodes; the inter-
Hg or Pb, and Ox is the oxalate mediates of this equilibrium are usually
anion. adatoms: Pt, H2 |H+ . . ., or Pt, Cl2 |Cl− . . .
When the metal does not take part in Palladium, hydrogen-sorbing alloys, and
the equilibrium and both components of intermetallic compounds pertain to gase-
the redox pair find themselves in solution, ous electrodes. However, at the same time,
the system is called a redox electrode, these systems are example of intercalation
for example, a quinhydrone electrodes processes operating under equilibrium
(Table 1) or electrodes on the basis of conditions, which brings them close to
transition metal complexes: electrodes of the first kind.
The various kinds of electrodes consid- in which charged groups are bound with
ered above can be unified, if we take into the membrane matrix; solid membranes
account the concept of an electronic equi- reversible with respect to certain ions (as
librium at metal/solution interface, that for glass electrode); biomembranes. The
is, bear in mind that a certain activity of simplest model of a membrane electrode
ēs in solution, which corresponds to any is the liquid/liquid interface, which is
equilibrium potential, exists. prepared by contacting two immiscible
solutions and usually considered together
with liquid electrodes in terms of soft
1.6 electrochemical interface (see Sect. 3.4).
Membrane Electrodes
A special comment should be given of 1.7

membrane electrodes. The potential drop Standard Potentials
across a membrane consists of the diffu-
sion potential (which can be calculated Standard potential values are usually those
in a similar manner as that for com- of ideal unimolal solutions at a pres-
pletely permeable systems) and two so- sure of 1 atm (ignoring the deviations of
called Donnan potential drops, named fugacity and activity from pressure and
after F. G. Donnan who was the first to concentration, respectively). A pressure of
consider these systems on the basis of 1 bar = 105 Pa was recommended as the
Gibbs thermodynamics [37]; more detailed standard value to be used in place of
consideration was given latter in Ref. [38]. 1 atm = 101 325 Pa (the difference corre-
The Donnan potential can be expressed sponds to a 0.34-mV shift of potential). If
via the activities of ions capable of per- a component of the gas phase participates
meating through the membrane, and its in the equilibrium, its partial pressure is
value is independent of the activities of taken as the standard value; if not, the
the other ions. This fact forms the basis standard pressure should be that of the
for the widespread experimental technique inert gas over the solution or melt. In a
of measuring the activities in the systems certain case, a standard potential can be
with eliminated diffusion potentials (ion- established in a system with nonunity ac-
selective electrodes [39, 40]). The system tivities, if the combination of the latter
best known among the latter is the pH- substituted in the Nernst equation equals
sensitive glass electrode, for which the unity. For any solid component of re-
modes of operation, the selectivity, and dox systems, the chemical potential does
microscopic aspects were studied inten- not change in the course of the reaction,
sively [41]. Sensors of this sort are known and it remains in its standard state. In
for the vast majority of inorganic cations contrast to the common thermodynamic
and many organic ions. The selectivity of definition of the standard state, the tem-
membranes can be enhanced by construct- perature is ignored, because the potential
ing enzyme-containing membranes. of the standard hydrogen (protium) elec-
The following brief classification of trode is taken to be zero at any temperature
membrane electrodes can be used [42]: in aqueous and protic media. The zero
inert membranes (cellulose, some sorts of temperature coefficient of the SHE corre-
porous glass); ion exchange membranes, sponds to the conventional assumption of
1.8 Specific Features of Certain Reference Electrodes 13
the zero standard entropy of H+ ions. This In general, electrodes of the second
extrathermodynamic assumption induces kind are more convenient as reference
the impossibility of comparing the values electrodes, because they do not require
referred to the hydrogen electrodes, in dif- a source of gaseous hydrogen. Dynamic
ferent solvents. Of the systems regarded as hydrogen reference electrodes, which rep-
reference electrodes, platinum hydrogen resent a wire of platinum metal (or another
electrodes exhibit i0 values ranked among metal with low hydrogen overvoltage) ca-
the greatest known (10−3 –10−2 A cm−2 , thodically polarized up to the hydrogen
at least, in aqueous acidic solutions). This evolution, give a possibility of avoiding
type of electrodes is used as the ref- the use of gaseous hydrogen. Their sta-
erence for tabulating universal potential ble potential values are determined by the
values [43–47]. existence of the current–potential relation-
The universal definition of the standard ship, which is possible for the electrode
potential E 0 of a redox couple Red/Ox is processes with high degree of stability. A
as follows: the standard potential is the special type of stable hydrogen electrode
value of emf of an electrochemical cell, is based on Pd hydride (β-phase). Another
in which diffusion potential and thermo- advantage of the electrodes of the second
emf are eliminated. This cell consists of an kind is their applicability in a wider range
electrode, on which the Red/Ox equilibria of temperatures, and also suitability for a
establish under standard conditions, and wide range of pressures. Operating at high
a SHE. temperatures presents a real challenge in
finding a suitable reference electrode (al-
though there are few examples stable up
1.8 to 250–300 ◦ C). One should never use any
Specific Features of Certain Reference reference electrode containing mercury at
Electrodes elevated temperatures. For other reference
electrodes (e.g., silver/silver chloride), sta-
An important type of reference electrodes bility and reproducibility should be tested
is presented by the so-called reversible carefully before it is used at temperatures
hydrogen electrodes (RHE) in the same so- above 100 ◦ C.
lution, which makes it possible to avoid For aqueous hydrogen electrodes, the
the liquid junction. These electrodes are potentiometric data are now available in
preferentially used in electrochemical ex- both sub- and supercritical regions, up to
periments when investigating the systems 723 K and 275 bar [48]. However, their use
with H+ and H2 involved in the process as the high-temperature reference systems
under study (adsorption and electrocataly- involves numerous complications.
sis on hydrogen-adsorbing surfaces, such Among reliable systems for high-
as platinum group metals). The RHE can temperature measurements, gaseous (first
be produced for a wide range of pH; certain of all, oxygen) electrodes based on solid
special problems associated with neutral electrolytes of the yttria-stabilized zirconia
solutions can be solved by using buffer- (YSZ) type can be recommended, which
ing agents. In media containing organic are always applied in the electrochemical
components, the possible catalytic hydro- cells with solid electrolytes. Gaseous elec-
genation of the latter poses an additional trodes based on Pt or carbon are also widely
limitation for hydrogen electrodes. used in high-temperature melts, especially
those based on oxides (oxygen reference on Tl− and Zn in liquid ammonia and hy-
electrode) or chlorides (chlorine reference drazine, and also Hg/Hg2 F2 electrode in
electrode). pure HF.
There are lots of systems, especially An exhaustive consideration of the
for electroplating and electrosynthesis, in specific features of various reference
which electrodes of the first kind can be electrodes (fabrication, reproducibility and
used, without any liquid junctions (the stability, modes of applicability, effects
example is liquid Al in AlF3 -containing of impurities, necessary corrections) can
melts). More universal systems for melts be found in Refs. [36, 51–54]. Nowadays,
of various kinds are: a chlorine electrode in certain new findings in this field are
equimolar NaCl + KCl melt and Ag/Ag+ possible because of the novel approaches
electrodes with the range of Ag+ con- to immobilization of redox centers on the
centrations (0.01–10 mM) corresponding electrode surfaces.
to usual solubility values. Reference elec- The attempts to interrelate the poten-
trodes of the second kind can hardly be tial scales in aqueous and nonaqueous
used in melts because of the high solubility solutions have been undertaken and are
of the majority of inorganic solids.
still in progress. Such a relationship
When studying nonaqueous systems by
could have been found if the free en-
means of galvanic cells with aqueous or
ergy of transfer was known at least for
mixed reference electrodes, we cannot
one type of charged species common to
avoid liquid/liquid junctions and estimate
the solvents. It is evident that ways of
the corresponding potential drop from
solving this problem can be based on
any realistic model. In protic nonaqueous
the assumptions beyond the scopes of
media (alcohols, dioxane, acetone, etc.),
thermodynamics. Thus, it was mentioned
a hydrogen electrode can be used; it is
in Ref. [18–20] that the free energy of
also suitable for some aqueous/aprotic
mixtures. However, the i0 values for the solvation of Rb+ ion is low and approx-
hydrogen reaction are much lower as imately the same in different solvents
compared with purely aqueous solutions. because of its great size. However, any
When studies are carried out in nonaque- direct application of rubidium electrodes
ous media, in order to avoid liquid/liquid is hampered by their corrosion activity as
junction preference should be given to the well as by the fact that their potential
reference electrodes in the same solvent as lies in the region of electroreduction of
the electrode of interest. a number of solvents. Another proposition
In aprotic (as well as in protic and concerns the fact that the free energies
mixed) media, the two reference sys- of transfer for certain cations and neu-
tems of choice are ferrocene/ferrocenium tral molecules of the redox pairs A/A+
and bis(biphenil)chromium(I/0). The pen- are the same in different solvents, which
tamethylcyclopentadienyl analog of the is also caused by their great sizes [51].
former was recently shown to yield higher Ferrocen/ferrocenium-like systems were
performance [49, 50]. Among other typical taken as the A/A+ pairs. It is inter-
electrodes, Ag/AgNO3 should be men- esting that these scales – rubidium and
tioned. We can also mention special A/A+ – adequately correlate with one an-
reference systems suitable for certain sol- other. Different extrathermodynamic as-
vents, such as amalgam electrodes based sumptions are compared in Ref. [54].
1.9 Values of Redox Potentials 15
Special reference electrodes are intro- It should be mentioned that no standard

duced when considering the thermody- electrode can be experimentally achieved,
namics of surface phenomena on elec- because the standard conditions are hypo-
trodes [55]. That is, when we use a ‘‘virtual’’ thetical (no system is ideal at a unimolal
reference electrode potential, which is re- concentration and the atmospheric pres-
ferred to the electrode reversible with sure). Hence, the majority of tabulated
respect to cation or anion in the same solu- values were recalculated from the data for
tion, changes by a value of (RT /F ) ln a± more dilute solutions, usually after the
(a± is a mean activity coefficient), then preliminary extrapolation to the zero ionic
the thermodynamic relationships can be strength. When dealing with the reverse
generalized for a solution of an arbitrary procedure (recalculation of equilibrium
concentration without introducing any no- potentials for dilute solutions from tabu-
tions on the activity of a separate ion lated standard values), we should take into
(Sect. 5). account that in real systems the time of
equilibrium establishment increases with
decreasing concentration, because of the
1.9 decrease in rates of half-cell reactions.
Values of Redox Potentials Hence, the possibility to prepare a re-
versible electrode is limited to a certain
The values of redox potentials were tabu-
concentration value.
lated in numerous collections [43–47], the
A famous collection of calculated poten-
latest collections being critically selected.
tials was presented by Pourbaix [46] for
The potentials of redox systems with par-
aqueous solutions of simplest compounds
ticipation of radicals and species in excited
of elements with oxygen and hydrogen.
electronic states are discussed in Refs. [56,
Although this collection ignores the non-
57]. There is no need to measure the
stoichiometry phenomena, which are of
potentials for all redox pairs (and, corre-
spondingly, G for all known reactions). great importance for many oxides and hy-
If we obtain the partial values for a number droxides, it remains helpful for making
of ions and compounds, the characteristic rough estimates and also predicting the
values for any other reactions can be com- corrosion peculiarities. It should be noted
puted. The idea of calculation is based on that these predictions cannot be observed
the fact that the emf value only depends in some systems under pronounced ki-
on the initial and final states of the sys- netic limitations.
tem, being independent of the existence The shifts of reversible potentials of ox-
of any intermediate states. This fact is of ide electrodes with composition were first
great importance for systems, for which considered in Ref. [12] using the example
it is impossible, or extremely difficult, to of hydrated MnO2−x , which demonstrates
prepare a reversible electrode (redox cou- a wide region of inhomogeneity. Later, the
ples containing oxygen or active metals). considerations of this sort were carried out
From considerations of the equilibria with also for other systems [58].
participation of a solvated electron, the There are lots of empirical and semiem-
E 0 value that determines the value of the pirical correlations of redox potentials with
constant in Eq. (7) was estimated: – 2.87 V molecular characteristics of substances,
(SHE) [56]. especially for the sequences of related
compounds, which can be used for semi- is usually used, which characterizes the
quantitative estimates of potentials of width of the interval between the poten-
novel systems. The most advanced ap- tials of cathodic and anodic background
proaches known from coordination chem- processes. The narrow potential window
istry take into account σ -donating and of water (about 1.3–1.4 V on platinum
π-accepting abilities of the ligands, which group metals, and up to 2 V on mercury-
can be expressed, for example, in terms like metals, which do not catalyze the
of Hammett or Taft, and also other effec- hydrogen evolution reaction) stimulates
tive parameters [59]. For the sequences of the potentiometric studies in nonaqueous
complexes with different central ions and solvents. Aprotic solvents usually demon-
related ligands, the steric factors (namely, strate windows of about 3–3.5 V, if the
chelating ability) are known to significantly optimal supporting electrolyte is chosen.
affect the redox potentials. An extremely wide window is known, for
A separate field deals with correlating example, for liquid SO2 , which is of highest
the redox potentials with spectroscopic interest for measuring extremely positive
parameters, such as the energy of the redox potentials (up to +6.0 V (SCE)) [62].
first allowed d–d transition in a complex,
Low-temperature haloaluminate melts are
the energy of the metal-to-ligand charge-
highly promising systems [63]. Finally, a
transfer band (i.e., the separation between
number of unique mixed solvents with
the HOMO (highest occupied molecular
extremely wide windows were found in
orbital) on the metal and LUMO (lowest
recent studies of lithium batteries [64].
unoccupied molecular orbital) on the
Among electrode materials, the widest
ligand), and nuclear magnetic resonance
windows are known for transition metal
(NMR) chemical shifts. The problem
of extending these correlations over a oxides, borides, nitrides, and some spe-
wider range of reactions, including those cially fabricated carbon-based materials. It
irreversible (for which the kinetics makes should be mentioned that, if the nature
a substantial contribution to the formal of electrode material can affect the formal
potential value, the latter being evidently potential value by changing the mecha-
nonthermodynamic) was considered a nism and kinetic parameters, the solvent
long time ago [60]. Advanced spectroscopic frequently has a pronounced effect on the
techniques are widely used for solving the equilibrium potential, because of the sol-
reverse problem of determining the redox vation contribution to free Gibbs energy.
potentials of irreversible couples [61].
1.11
1.10 Experimental Techniques
Potential Windows
Direct potentiometry (emf measurements)
The feasibility of determining the redox requires the potential of indicator electrode
potential depends strongly on the solvent, to be determined by the potential of the re-
the electrolyte, and the electrode material: dox pair under study. Sometimes this is
all of them should remain inert in po- impossible, because of the low exchange
tential ranges as wide as possible. In this current densities and/or concentrations
connection, the term ‘‘potential window’’ (particularly, at low solubilities or limited
1.12 Effects of Ion Pairing 17
stabilities of the components), when extra- chemistry and the activity coefficients. The
neous redox pairs contribute significantly latter task covers a wide field of pH-metric
to the measured potential. The sensitivity applications, and also the analytical tech-
of direct potentiometry can be enhanced by niques based on ion-selective electrodes.
using preconcentration procedures, partic-
ularly in polymeric films on the electrode
surfaces [65]. However, it is common to 1.12
use voltammetry, polarography, and other Effects of Ion Pairing
related techniques in place of potentiom-
etry. The formal potential (determined as An important factor affecting redox po-
the potential between anodic and cathodic tentials is also ion pairing, which is
current peaks on voltammograms) does conventionally taken into account in terms
not generally coincide with the equilib- of activity coefficients. For redox potentials
rium redox potential, the accuracy depend- of coordination compounds, the following
ing on the reaction mechanism [66–68]. equation is known [70]:
For reversible and simplest quasireversible
electrode reactions, the exact determina- RT Kred
tion of standard potentials from the data of E = E0 + ln (23)
F Kox
stationary voltammetry and polarography
is possible. Only these dynamic techniques
where Kred and Kox denote the equilib-
are suitable for the studies of equilibria
rium constants of a complex (particularly,
with participation of long-living radicals
ion pair) formation. In as much as outer
and other excited species, which can be
sphere ionic association is highly sensitive
introduced into solution only by in situ
to the ionic charge, Kred appears to be
electrochemical generating.
higher than Kox , and the potential shifts
If voltammetric and related techniques
to more positive values with the ionic
are used, the ohmic drop should be either
compensated (now this is usually done by strength. That is, for usual concentrations
the software or hardware of electrochem- of hexacyanoferrates in aqueous solutions,
ical devices [69]), or reduced by using, for the shift induced by the association with
example, a Luggin (Luggin-Gaber) capil- K+ ions reaches 0.1 V and higher at usual
lary (see in Ref. [12]). Another important concentrations.
technical detail is that the components In electroanalysis, Eq. (23) forms the
of reference redox systems (such as fer- basis of highly sensitive techniques
rocene/ferrocenium) are frequently added of potentiometric titration. The practi-
immediately into the working compart- cal applications are presented by numer-
ment when voltammetry-like techniques ous potentiometric sensors, particularly
are applied. biosensors [71–74]. The potential mea-
Applications of potentiometry are rather surements in microheterogeneous media
widespread, and its efficiency is high (microemulsions; anionic, cationic and
enough when operating with relative po- nonionic micelles) were worked out re-
tential values. A mention should be made, cently [75]. A separate field is the potentio-
first of all, of the determination of ba- metric studies of liquid/liquid junctions
sic thermodynamic quantities, such as directed to the determination of ionic
the equilibrium constants for coordination transport numbers.
1.13 and the problem reduces to the determi-

Absolute Electrode Potential nation of χ S and χ M , because the Volta
potential MS ψ can be measured. The es-
Although it has long been known that only timates of χ H2 O for the water/air interface
relative electrode potentials can be mea- were made by considering the adsorption
sured experimentally, numerous attempts effects that inorganic acids, HClO4 , and
were undertaken to determine the value HBr [87, 88] and aliphatic compounds [89]
of potential of an isolated electrode with- exert on the Volta potential of water/air
out referring it to any reference system interface, and also by measuring the tem-
(‘‘absolute electrode potential’’). Exhaus- perature coefficient of χ H2 O [90]. In this
tive and more particular considerations of connection, the measured real hydration
the problem of ‘‘absolute potential’’ can (real)
energies of ions, Ghydr , were compared
be found in Refs. [1–20, 76–86]. These with the calculated chemical hydration en-
attempts were concentrated around the (chem)
ergies, Ghydr . The real (total or free)
determination of a separate Galvani po-
energies were considered as the energy
tential, MS ϕ (which was named initially changes, which accompany the transfer
as the absolute electrode potential) and also
of ions from air to solution. This value
included the search for a reference elec-
can be divided into the chemical hydration
trode, for which the maximum possible
energy, which is caused by the interac-
work of any imaginary electrode process tion of an ion with surrounding water
equals zero. Sometimes, the problem was molecules, and the change of electric en-
formulated as a search for the hypotheti- ergy that equals zi F χ H2 O (where zi is the
cal reference state determined as reckoned charge of the ion of the corresponding
from the ground state of electron in vac- sign):
uum (a ‘‘physical’’ scale of energy with
the opposite sign). In this connection, the (real)
Ghydr = Ghydr
(chem)
+ zi F χ H2 O (25)
requirement for the reference electrode
under discussion was formulated in the
On the other hand, the real solvation
absence of any additional electrochemical
energy can be written as
interfaces.
Finally, these studies have transformed (real)
Ghydr = −Gsubl − Gion
into a reasonable separation of the mea-
sured emf for a cell, which consists of + zi FM
H2 O ψ − zi F We (26)
an electrode under consideration and a
standard reference electrode, in order to where Gsubl is the free energy of metal
determine two electrode potentials re- sublimation (atomization), Gion is the
ferred to individual interfaces (or to the free energy of ionization of a metal atom,
separate half-reactions) by using only the We is the electron work function of M,
experimentally measurable values. and M H2 O ψ is the Volta potential of the
To illustrate the technique of Galvani interface corresponding to the equilibrium
potential calculation, we return to the M|S potential in a solution, in which the
interface. According to scheme in Fig. 1, concentration of metal ions is unity. The
(real)
most reliable values of Ghydr were
M M S
S ϕ = S ψ − χ + χ
M
(24) reported in Ref. [91].
1.13 Absolute Electrode Potential 19
(chem)
The value of Ghydr cannot be mea- schematically illustrates a ‘‘metal box’’
sured directly; however, it can be estimated corresponding to the jellium model.
According to this figure,
on the basis of information on the liq-
uid structure of water and the structure We = V − εF = e0 χ M + Vex − εF (27)
(chem)
of its molecule. Having Ghydr and
(real) the notations are given in the figure
Ghydr , we can find χ H2 O from Eq. (24).
caption. In the framework of the jellium
The estimates of this value given in the
model, we can calculate χ M by two
literature for water/air interface disagree techniques: directly from the distribution
with one another significantly (from −0.48 of the electronic gas outside the metal, and
up to +0.29 V). A critical comparative con- also by calculating Vex and εF with the
sideration of various studies leads to a subsequent use of Eq. (26). Unfortunately,
conclusion about a low positive value of the jellium model has a serious limitation;
χ H2 O , namely, 0.13 V [1–9]. This value the estimates of χ M will be improved in
corresponds to the infinite dilution and future as the theory of electronic structure
can be changed by changing the solution of metals and their surfaces becomes more
concentration. advanced.
The estimation of χ M is an even As was correctly reasoned in Ref. [10,
more complicated problem. Figure 3 11], the scale of ‘‘absolute potential’’
electron in vacuum
(at infinity)
−e0y M mM
e /NA
electron in vacuum
(near the metal surface)
Fermi level
We
V = Vex + Vel
eF
Fig. 3 A scheme of electron energy levels for a model of ‘‘metal box’’ (according to
C. Kittel, Elementary Solid State Physics, Wiley, New York, 1962): µ̄M
e – electrochemical
potential of electron in metal, NA – Avogadro number, εF – Fermi energy, ψ M – outer
potential of metal, We – electron work function, V – electron potential energy in metal
(V = Vex + Vel , Vex – energy of electron exchange, which corresponds to the electron
interaction with positively charged jellium, Vel = e0 χ M – the surface component of
electron potential energy).
constructed in accordance with Ref. [7] The ‘‘absolute potentials’’ of the hydro-
and recommended in Ref. [9] is, in fact, gen electrode in a number of nonaqueous
equivalent to the EK scale first proposed solvents are reported in Ref. [9].
in Refs. [76–81], which should be named The value of EK can be expressed via
after Kanevskii: metal/vacuum electron work function and
Volta potential [10, 11]:
(real)
[Gsubl + Gion + Ghydr ]
EK = − EK = We + M
S ψ (31)
zi F
(28) The majority of electrochemical prob-
By introducing this scale, we can sep- lems can be solved without separating
arate the total emf of the cell into two the emf into ‘‘absolute potentials’’. How-
quantities, for which the contributions ever, it should be mentioned that the
(real)
Gsubl , Gion , and Ghydr can be ex- problem concerning the structure of the
perimentally determined. For a hydrogen interfacial potential drop becomes the top-
electrode, ical problem for the modern studies of
electrified interfaces on the microscopic
(EK )H+ ,H2 level, particularly, in attempts of test-
ing the electrified interfaces by probe
1 (real)
2 Gdiss + Gion + Ghydr techniques [92–94]. The absolute scales
=− are also of interest for electrochemistry
F
(29) of semiconductors in the context of cal-
which gives the possibility to recommend ibrating the energy levels of materials.
the value −4.44 ± 0.02 V for the ‘‘abso- This problem is related to another general
lute potential’’ of hydrogen electrode (at problem of physical chemistry – the de-
298.15 K). For other temperatures, the cal- termination of activity coefficient of an
culations of this sort are limited by the individual charged species.
lack of information on the temperature
coefficient of work function. It should
be mentioned that the uncertainty of 1.14
absolute potential calculation is substan- The Role of Electric Double Layers
tially higher than the accuracy of direct
potential measurements with respect to Later, a more detailed consideration of the
a reference electrode. The useful com- M|S interface is given, which takes into
ments to Kanevskii’s scale can be found account the formation of electric double
in Refs. [16–20, 84]. In the ‘‘absolute’’ layer. Figure 1 shows the potential drops
scale, the ‘‘absolute potential’’ of any at the metal/solution interface. It follows
aqueous electrode at 298.15 K can be de- from this figure that
termined as
M S M
S ψ = χ + S ϕ − χ
M
(32)
EK = (EK )H+ ,H2 + E ≈ −4.44 + E The value of the Galvani potential M S ϕ

(30) can be represented as the sum of the
where E is the potential referred to potential drop inside the ionic double layer
the SHE. Two scales are combined for ϕ and the surface potentials of solution
comparison in Fig. 4. and metal, which have changed as a result
1.14 The Role of Electric Double Layers 21
Electrochemical Physical
scale scale
[V] [eV]
(Absolute potential scale) = −(Physical scale)
E(abs) / V = E(SHE) / V + 4.44
− 4.44 0 electrons at rest
in a vacuum
−4
−1
E 0(Li+/Li) −3.05 −1.39

−3
Hydrated electron −2.87 −1.57
−2 −2.07
E 0(Mg2+/Mg) −2.37
−2
E 0(Al3+/Al) −1.66 −2.78

−3
−1
E 0(Zn2+/Zn) − 0.76 −3.68
Potential of zero charge of gallium − 0.69 −3.75
−4
Potential of zero charge of mercury − 0.19 − 4.25
Standard hydrogen electrode (SHE) 0 − 4.44
Saturated calomel electrode (SCE) 0.24 − 4.68
−5
E 0(Ag+/Ag) 0.80 −5.24
1
Standard oxygen electrode (aH+ = 1) 1.23 −5.67
–6
E 0(MnO 4−/MnO2) 1.70 − 6.14
Fig. 4 Comparison of SHE and ‘‘absolute’’ scales, as reported in Ref. [9].
of mutual contact: at their common interface, respectively.

Hence,
M
S ϕ = ϕ − χ
S(M)
+ χ M(S) (33)
M S
S ψ =χ −χ
S(M)
+ ϕ + χ M(S) − χ M
where χ S(M) and χ M(S) are the sur-
face potentials of solution and metal = ϕ + δχ M − δχ S (34)
1
2
0.6
In Tl
In
0.4
WeHg − WeMe
Cd Cd
[eV]
Pb
Bi Bi
0.2
Sn Ga Ga
Sn
0
Sb
0 0.2 0.4 0.6

= 0 − Es = 0 , Es = −18 − Es = −18
EsHg Me Hg Me
[V]
Fig. 5 Correlation of work function differences and the differences of
potentials at constant electrode charge σ for σ = 0 (1) and
−18 µC cm−2 (2), as reported in Ref. [5].
By writing such relationships for two Eq. (36) and (37) obtained by Frum-
interfaces, M1 |S and M2 |S and substitut- kin [1–5] can be classified as the solution
ing corresponding M 1 M2
S ψ and S ψ into of the famous Volta problem of the nature
Eq. (10), we obtain of emf of electrochemical circuit. Equa-
tion (37) demonstrates that the difference
E = M
M1 ψ + (ϕ1 − ϕ2 ) of potentials of zero charge (pzc) of two
metals is approximately equal to their Volta
+ (δχ1M − δχ2M + δχ2S − δχ1S ) (35)
potential. In as much as the Volta poten-
If the metals M1 and M2 interact very tial M 2
M1 ψ is equal to the difference of
weakly with the solvent molecules, we can work functions of the metals (WeM2 and
assume that, in the first approximation, WeM1 ), the verification of Eq. (37) can be
the last right-hand term in parentheses is carried out by comparing the differences
zero, and both in pzc and in work functions (Fig. 5).
The data of Fig. 5 confirm Eq. (37) and, si-
E∼
= M
M1 ψ + (ϕ1 − ϕ2 ) (36) multaneously, demonstrate the possibility
of substantial deviations from this rela-
When both metals, M1 and M2 , are tionship for the systems with pronounced
zero charged (their excess surface charge solvent/metal interaction (as, for example,
density σM = 0), and the ionic double for the Ga/water interface).
layers are absent on their surfaces, In the context of concept of zero charge
potential [1–5], a specific scale of electrode
EσM =0 ≈ M2
M1 ψ (37) potentials should be mentioned, a reduced
1.15 Conclusion 23
scale, in which the potential of electrode 2. A. Frumkin, A. Gorodetzkaja, Z. Phys. Chem.

referred to its pzc for given solution 1928, 136, 215–227.
3. A. N. Frumkin, Phys. Z. Sowjetunion 1933, 4,
composition. It was used actively in the
239–261.
studies of the electric double layer as the so- 4. A. N. Frumkin, J. Chem. Phys. 1939, 7,
called Grahame rational scale (see Sect. 5). 552–553.
5. A. N. Frumkin, Potentsialy nulevogo zaryada
(in Russian) (Potentials of Zero Charge),
Nauka, Moscow, 1979.
1.15 6. R. Parsons in Modern Aspects of Electrochem-
Conclusion istry (Eds.: J. O’M. Bockris), Butterworths Sci.
Publ., London, 1954, pp. 47–102, Vol. 1.
As mentioned earlier, the role played by the 7. S. Trasatti, J. Electroanal. Chem. 1975, 66,
electrode potential in electrochemistry and 155–161.
8. S. Trasatti in Comprehensive Treatise of Elec-
surface science is not limited to thermody- trochemistry (Eds.: J. O’M. Bockris, B. E. Con-
namic aspects. The electrode potential is way, E. Yeger), Plenum Press, New York,
the factor that governs (directly or via the London, 1980, pp. 45–81, Vol. 1.
electrode charge) the surface reconstruc- 9. S. Trasatti, Pure Appl. Chem. 1985, 58,
tion and phase transitions in adsorbed 955–966.
10. A. Frumkin, B. Damaskin, J. Electroanal.
layers. It is also the key factor, which de- Chem. 1975, 66, 150–154.
termines the Gibbs free energy of reaction 11. A. N. Frumkin, B. B. Damaskin, Dokl. AN
and, correspondingly, the Frank–Condon SSSR 1975, 221, 395–398.
barrier. In electrochemical kinetics, the 12. K. J. Vetter, Electrochemische Kinetik, Sprin-
ger-Verlag, Berlin, 1961.
characteristic value is not the potential
13. R. G. Compton, G. H. W. Sanders, Electrode
itself, but the overvoltage, which is the de- Potentials, Oxford University Press, New
viation of the potential from its reversible York, 1996.
value in the same solution when the 14. J. R. Runo, D. G. Peters, J. Chem. Educ. 1993,
current flows across the electrochemical 70, 708–713.
15. M. I. Temkin, Bull. Acad. Sci. l’URSS 1946,
interface. The potential plays an impor-
N2, 235–244.
tant role in electrosynthesis, because the 16. B. V. Ershler, Usp. Khim. 1952, 21, 237–249.
controlled potential electrolysis can be per- 17. B. V. Ershler, Zh. Fiz. Khim. 1954, 28,
formed with high selectivity, in the absence 957–960.
of foreign reagents (oxidizers or reduc- 18. V. A. Pleskov, Usp. Khim. 1947, 16, 254–278.
19. V. A. Pleskov, Zh. Fiz. Khim. 1949, 23, 104.
ers). The concept of membrane potential 20. V. A. Pleskov, Zh. Fiz. Khim. 1950, 24, 379.
is one of the basic concepts in biophysics. 21. E. A. Guggenheim, J. Phys. Chem. 1929, 33,
All these wide fields of science demon- 842–849.
strate the interplay with electrochemical 22. W. Schottky, H. Rothe in Handbuch
thermodynamics and provide the basis for der Experimentalphysik. (Eds.: W. Wien,
F. Harms, H. Lenz), Acad. Verlag. m.b.h.,
further interpenetration of chemistry, bi- Leipzig, 1928, pp. 145–150, Vol. 13 (2.Teil).
ology, and physics, which was pioneered 23. E. Lange, K. P. Miscenko, Z. Phys. Chem.
two centuries ago by the studies on the 1930, 149A, 1–41.
problem of electrode potential. 24. E. Lange in Handbuch der Experimentalphysik
(Eds.: W. Wien, F. Harms, H. Lenz), Acad.
Verlag. m.b.h., Leipzig, 1933, pp. 263–322,
References Vol. 12 (2.Teil).
25. E. Lange, Z. Elektrochem. 1951, 55, 76–92.
1. A. N. Frumkin, Ergeb. Exact. Naturwiss. 1928, 26. E. Lange, Z. Elektrochem. 1952, 56, 94–106.
7, 235–273. 27. W. Nernst, Z. Phys. Chem. 1889, 4, 129–181.
28. P. P. Edwards in Advances in Inorganic Chem- 51. H. Strehlow, The Chemistry of Non-Aqueous
istry and Radiochemistry (Eds.: H. J. Emeleus, Solvents, Academic Press, 1967, pp. 129–171,
A. G. Sharpe), Academic Press, New York, Vol. 1.
1982, pp. 135–185, Vol. 25. 52. J. N. Butler in Advances in Electrochem-
29. M. Plank, Ann. Physik. 1890, 40, 561–576. istry and Electrochemical Engineering (Ed.:
30. P. Henderson, Z. Phys. Chem. 1907 59, P. Delahay), Intersci. Publi., New York, 1970,
118–127. pp. 77–175, Vol. 7.
31. P. Henderson, 1908, 63, 325–345. 53. G. Gritzner, J. Kuta, Pure Appl. Chem. 1984,
32. B. B. Damaskin, G. A. Tsirlina, M. I. Borz- 56, 461–466.
enko, Russ. J. Electrochem. 1998, 34, 199–203. 54. Y. Marcus, Pure Appl. Chem. 1986, 58,
33. J. B. Chloupek, V. Z. Danes, B. A. Danesova, 1721–1736.
Collect. Czech. Chem. Commun. 1933, 5, 55. A. N. Frumkin, Zh. Fiz. Khim. 1956, 30,
469–478. 2066–2069.
34. J. B. Chloupek, V. Z. Danes, B. A. Danesova, 56. D. M. Stanbury in Advances in Inorganic
Collect. Czech. Chem. Commun. 1933, 5, Chemistry (Ed.: A.G. Sykes), Academic Press,
527–534. INC, New York, 1989, pp. 69–138, Vol. 33.
35. N. Q. Minh, L. Redey, Molten Salt Techniques, 57. D. D. Wayner, V. D. Parker, Acc. Chem. Res.
Plenum Press, New York, 1987, Vol. 3. 1993, 26, 287–294.
36. D. J. G. Ives, G. J. Janz, (Eds.), Reference Elec- 58. G. A. Tsirlina, O. V. Safonova, O. A. Petrii,
trodes. Theory and Practice Academic Press, Electrochim. Acta 1997, 42, 2943–2946.
New York, London, 1961. 59. J. Chatt, Coord. Chem. Rev. 1982, 43,
37. F. G. Donnan, Z. Elektrochem. 1911, 17, 337–347.
572–581. 60. A. A. Vlcek, Prog. Inorg. Chem. 1963, 5,
38. F. G. Donnan, E. A. Guggenheim, Z. Phys. 211–384.
Chem. 1932, 162A, 346–360. 61. F. Sanchez, P. Perez-Tejeda, F. Perez et al.,
39. W. E. Van der Linden in Comprehensive An- J. Chem. Soc., Dalton Trans. 1999, N 17,
alytical Chemistry. (Ed.: G. Svehla), Elsevier, 3035–3039.
Amsterdam, 1981, Vol. 1. 62. E. Garcia, A. Bard, J. Electrochem. Soc. 1990,
40. J. Vesely, D. Weiss, K. Stulik, Analysis with 137, 2752–2759.
Ion-Selective Electrodes, Ellis Horwood, Chich- 63. C. L. Hussey in Chemistry of Nonaqueous
ester, 1978. Solutions (Eds.: G. Mamantov, A. I. Popov),
41. G. Eisenman, Biophys. J. 1962, 2, 259–323. VCH, New York, 1994, pp. 227–275.
42. J. S. Newman, Electrochemical Systems, 2nd 64. K. Xu, N. D. Day, C. A. Angell, J. Electrochem.
ed., Prentice-Hall, Englewood Cliffs, 1991. Soc. 1996, 143, L209–L211.
43. R. Parsons, Redox Potentials in Aqueous 65. J. Huang, M. S. Wrighton, Anal. Chem. 1993,
Solutions: A Selective and Critical Source Book, 65, 2740–2746.
Marcel Dekker, New York, 1985, pp. 1–34. 66. A. J. Bard, L. R. Faulkner, Electrochemical
44. A. J. Bard, R. Parsons, J. Jordan, Standard Methods. Fundamentals and Applications, Wi-
Potentials in Aqueous Solution, Marcel Dekker, ley, New York, 2001.
New York, 1985. 67. D. K. Gosser, Cyclic Voltammetry: Simulation
45. M. S. Antelman, F. J. Harris, (Eds.), The En- and Analysis of Reaction Mechanisms, VCH
cyclopedia of Chemical Electrode Potentials, Publ., Deerfield Beach, 1993.
Plenum Press, New York, London, 1982. 68. J. F. Cassidy, Compr. Anal. Chem. 1992, 27,
46. M. Pourbaix, Atlas d’Equilibres Electrochemi- 1–69.
ques, Gauthier-Villars, Paris, 1963. 69. R. M. Souto, Electroanalysis 1994, 6, 531–542.
47. S. G. Bratsch, J. Phys. Chem. Ref. Data 1989, 70. J. Heyrovsky, J. Kuta, Zaklady Polarografie,
18, 1–21. Nakl. Ceskoslov. Academ. Ved, Praha, 1962.
48. K. Eklund, S. N. Lvov, D. D. Macdonald, J. 71. J. Janata, Chem. Rev. 1990, 90, 691–703.
Electroanal. Chem. 1997, 437, 99–110. 72. J. F. Coetzee, B. K. Deshmukh, C.-C. Liao,
49. J. T. Hupp, Inorg. Chem. 1990, 29, Chem. Rev. 1990, 90, 827–835.
5010–5120. 73. M. Alvarez-Icaza, U. Bilitowski, Anal. Chem.
50. J. K. Bashkin, P. J. Kinlen, Inorg. Chem. 1993, 65, 525A–533A.
1990, 29, 4507–4509. 74. J. Wang, Anal. Chem. 1995, 67, 487R–492R.
1.15 Conclusion 25
75. S. A. Myers, R. A. Mackey, R. A. Brajter- 85. H. Reiss, A. Heller, J. Phys. Chem. 1985, 89,
Toth, Anal. Chem. 1993, 65, 3447–3453. 4207–4213.
76. E. A. Kanevskii, Zh. Fiz. Khim. 1948, 22, 86. Yu. V. Pleskov, J. Phys. Chem. 1987, 91,
1397–1404. 1691–1692.
77. E. A. Kanevskii, Zh. Fiz. Khim. 1950, 24, 87. A. Frumkin, Z. Phys. Chem. 1924, 111,
1511–1514. 190–210.
78. E. A. Kanevskii, Zh. Fiz. Khim. 1951, 25, 88. A. Frumkin, Z. Phys. Chem. 1924, 109, 34–48.
854–862. 89. A. N. Frumkin, Z. A. Iofa, M. A. Gerovich,
79. E. A. Kanevskii, Zh. Fiz. Khim. 1952, 26, Zh. Fiz. Khim. 1956, 30, 1455–1468.
633–641. 90. J. Randles, D. Schiffrin, J. Electroanal. Chem.
80. E. A. Kanevskii, Zh. Fiz. Khim. 1953, 27, 1965, 10, 480–484.
296–309. 91. J. Randles, Trans. Faraday Soc. 1956, 52,
81. E. A. Kanevskii, Dokl. Akad. Nauk SSSR 1573–1581.
1981, 257, 926–929. 92. A. C. Hillier, S. Kim, A. J. Bard, J. Phys.
82. V. A. Pleskov, B. V. Ershler, Zh. Fiz. Khim. Chem. 1996, 100, 18808–18817.
1949, 23, 101–103. 93. R. C. Thomas, P. Tangyunyong, J. E. Hous-
83. B. V. Ershler, V. A. Pleskov, Zh. Fiz. Khim. ton et al., J. Phys. Chem. 1994, 98, 4493,
1951, 25, 1258–1260. 4494.
84. V. M. Novakovskii, Zh. Fiz. Khim. 1956, 30, 94. Z.-X. Xie, D. M. Kolb, J. Electroanal. Chem.
2820–2822. 2000, 481, 177–182.
33
2.1 The interfaces dielectric–electrolyte are

Electrochemical Interfaces: At the Border less important in classical electrochem-
Line istry, but are central for bioelectro-
chemistry and colloid science. These
Alexei A. Kornyshev
.. .. are, e.g., ionic crystal–electrolyte, lipid-
Research Center Julich, Julich, Germany and
bilayer–electrolyte, and protein–elec-
Imperial College of Science, Technology and
trolyte interfaces. Such interfaces are not
Medicine, London, UK
chargeable, but they may contain charges
Eckhard Spohr attached to surfaces. All these interfaces,
.. ..
Research Center Julich, Julich, Germany chargeable or containing fixed charges,
zwitter-ions or dipoles, are called electrified
Michael A. Vorotyntsev interfaces.
Université de Bourgogne, Dijon, France
Properties of the interface play a crucial
2.1.1 role in electrochemical energy conversion,
Introduction electrolysis, electrocatalysis, and electro-
chemical devices. On the other hand, the
Most of the events in electrochemistry take chargeable interface can and has been
place at an interface, and that is why widely used to probe various surface prop-
interfacial electrochemistry constitutes the erties. Indeed, due to Debye screening in
major part of electrochemical science. Rel- electrolyte, the electric field at the electro-
evant interfaces here are the metal–liquid chemical interface is localized in a narrow
electrolyte (LE), metal–solid electrolyte interfacial region and the interface can
(SE), semiconductor–electrolyte, and the be easily charged to tens of µC cm−2 at
interface between two immiscible elec- quite modest potentials. Physicists called
trolyte solutions (ITIES). These interfaces electrochemistry a surface science with a
are chargeable, that is, when the exter- joystick. The potentiostat is the joystick,
nal potential is applied, charge separation which allows to vary the electric field at
of positive and negative charges on the the interface from zero to 10 V Å−1 . Mea-
two sides of the contact occurs. Such suring the given signal dependence on
an interface can accumulate energy and the potential modulation makes it possi-
be characterized by electric capacitance, ble to distinguish the terms related to the
within the range of ideal polarizability be- interface, since the electric field in equi-
yond which Faraday processes turn on. librium is zero in the bulk. That joystick
34 2 Electrochemical Double Layers
was not available in surface science, and methods, the extensive research area of
thus physicists started to immerse solid semiconductor/electrolyte and metal/SE
surfaces into the electrolyte environment interfaces will not be considered here.
in order to probe the interface response to We will start with phenomenology, and
charging. then move, where possible, to a micro-
Distribution of potential near a charged scopic theory and simulations. Without a
interface is crucial in electrode kinetics, solid phenomenological basis one may risk
since the potential drop plays the role of calculating things from the scratch many
a variable driving force of reaction. The times, whereas the calculation of one pa-
reactions take place at the interface and rameter of the phenomenological theory
with participation of adsorbed reactants. might be all that is actually needed from a
However, not only reactants but other microscopic model! On the other hand, the
species as well can adsorb at the interface. phenomenology often receives its justifica-
Competitive adsorption of these species, tion and validity criteria from microscopic
ions or dipoles can play, depending on the theories.
situation, a catalyzing or inhibiting role on
the reaction rates. 2.1.2
On the whole, understanding the struc- Basic Concepts
ture of the interface and its response to
charging is most essential for electrochem- Electrified interfaces represent the princi-
istry and its implications for colloid science pal object of ‘‘interfacial electrochemistry’’.
and biophysics. Various aspects of this area are described
The electrochemical interface is com- in monographs and textbooks on this
posed of molecules (solvent, adsorbed subject [1–11]. A brief outline of this in-
molecular species) and ions (of electrolyte), formation is given in the following text.
which can be partially discharged when Electrified interfaces represent a spe-
chemisorbed, electrons and skeleton ions cific quasi two-dimensional object in the
in the case of metal electrodes, electrons vicinity of the geometrical boundary of
and holes in the case of semiconductor two phases containing, in general, mo-
electrodes, mobile conducting and im- bile electronic and/or ionic charges, and,
mobile skeleton ions in SEs. Molecules correspondingly, an electric-field distribu-
and ions are classical objects but elec- tion generated by these charges. Practically
trons, holes with small effective mass, at least one of these media in contact is
and protons are quantum objects. Inter- conducting, while the second one may
action between molecules and surfaces be either a conductor of another kind
is quantum-mechanical in nature in the (metal–electrolyte interface, semiconduc-
case of chemisorption. Thus, microscopic tor–electrolyte interface), or of the same
description of the interface requires a kind (interface of two immiscible elec-
combination of quantum and classical trolytes), or an insulator. As important
methods. One can benefit, however, from exceptions, one can point out pure sol-
simple or more involved phenomenologi- vent/air or pure solvent/insulator (e.g.
cal descriptions of the interface. silica-water) boundaries where a signifi-
In this chapter we will focus on charge- cant potential drop may be created by the
able interfaces: metal–LE interface and formation of a dipole layer, without the
ITIES. Although similar in concepts and generation of a space-charge region.
2.1 Electrochemical Interfaces: At the Border Line 35
The simplest example is given by the acquires an extra electronic charge, while
free metal surface. Owing to the Pauli the adjacent layer of another phase con-
principle the metal electrons at the highest tains an excessive positive charge formed
occupation levels possess a significant by the uncompensated ionic skeleton. The
kinetic energy so that their density extends thickness of each charged layer depends
outside the ionic skeleton of the electrode. on the screening properties of the corre-
It results in the formation of a structure sponding medium (their Thomas–Fermi
characteristic for numerous electrified or Debye screening lengths) so that the
interfaces; two oppositely charged spatial whole space-charge region is very thin for
regions (with zero overall charge), this time a contact of two metals (within 0.1 nm),
a positively charged region inside the ion while it may be much more extended in
skeleton due to the depletion of electrons the case of semiconductors, depending on
and a negatively charged region formed their bulk properties.
by ‘‘the electronic tail’’. Because of this In all cases such interfaces are non-
‘‘nanocondenser’’, the bulk metal has got polarizable, since the only way to change
a high positive potential of several Volts the interfacial potential difference is to
with respect to the vacuum. This manifests modify the bulk properties of the phase(s),
itself in the value of the electron work for example, the degree of doping or
function the work needed to withdraw stoichiometry for semiconductors.
an electron from the metal across its Analogous ionic systems are represented
uncharged surface. by a contact of two electrolyte solutions in
A more complicated example is provided immiscible liquids, which contain a com-
by the contact of two electronic conductors, mon ion. Owing to its interfacial exchange
metals or semiconductors. At equilibrium the electrochemical potential of this ion is
the electrochemical potentials of electrons to be constant throughout the system, and
(i.e. their Fermi energies) must be equal it leads again to a relation for the interfa-
in both bulk phases. The Galvani potential cial potential difference similar to Eq. (1)
difference between them is determined by (called Donnan equation this time), that is,
the bulk properties of the media in contact, the interface is nonpolarizable without a
i.e. by the difference between the chemical change of the bulk media properties.
potentials of electrons: The same process of interfacial ion ex-
change determines the potential difference
e − µe
µm1 m2
ϕ m1/m2 = (1) for other ion conductors, like membranes
e or SEs. The existence of this potential dif-
where e is the elementary charge, since ference again leads to the formation of the
the right-hand-side terms are constant, one EDL, but this time it is formed by exces-
cannot change the interfacial Galvani po- sive ionic charges in the surface layers of
tential in equilibrium conditions without both solutions. Since this EDL is created
changing the chemical composition of the by the ion transfer across the interface,
phases. The difference between the bulk without charge transport across the bulk
potentials required by Eq. (1) has to be es- media, the Donnan potential drop corre-
tablished by the formation of an ‘‘electrical sponding to the composition of the phases
double layer’’ (EDL) owing to the electron is established very rapidly.
exchange between the surface layers of the Similar to the contact of two elec-
phases. Finally, one of the surface layers tronic conductors, the imposition of a
nonequilibrium potential difference bet- being primarily dependent on the solvent

ween two ion conductors (having a com- properties.
mon ion) from the external circuit results The principal object of electrochemical
in a dc current passage. Subsequent pro- interest is given by another type of elec-
cesses depend on the type of electrode trified interface, contacts of an electronic
reactions at the reversible electrodes im- (liquid or solid metal, semiconductor) and
mersed into the ion-conducting phases. If an ionic (liquid solution, SEs, membranes,
both reactions generate or consume the etc) conductor. For numerous contacts of
same exchangeable ion, even a long-term this kind, one can ensure such ionic com-
realization of the process does not change position of the latter that there is practically
the bulk composition of the phases. In no dc current across the interface within a
the case of the electrode reactions gener- certain interval of the externally applied
ating or consuming different species, the potential. Within this potential interval
longtime application of this potential dif- the system is close to the model of an
ference will lead to electrolysis, that is, to ideally polarizable interface; the change of
the gradual change of the composition of the potential is accompanied by the relax-
the bulk phases. The latter is accompanied ation current across the external circuit and
by the change of the equilibrium values the bulk media that vanishes after a cer-
of all interfacial potential drops (between tain period. For sufficiently small potential
ion conductors and at both reversible elec- changes, dE, the ratio of the integrated
trodes), with the approach to a different relaxation current, dQ, to dE is indepen-
equilibrium potential difference. dent of the amplitude and it determines the
The contact of solutions that possess principal electrochemical characteristics of
more than one common ion, in particular the interface, its differential capacitance
all contacts of solutions with identical per unit surface area, C:
or miscible solvents, exhibits a more
dQ
complicated phenomenon. At the initial C = S −1 (2)
stage, all exchangeable ions are transferred dE
across the boundary generating a long- S being the surface area. Its value can
living nonequilibrium structure, ‘‘liquid be measured for each potential within the
junction (LJ)’’, which retains its ‘‘diffusion ideal polarizability interval, and it varies
potential difference’’ for an extended time, generally as a function of this potential
despite the absence of the thermodynamic as well as of the nature of the electrode
equilibrium (different compositions) bet- and the solution composition. The EDL
ween the interphase and two bulk phases. for such systems contains an excess
If the solutions in contact have the same (positive or negative) electronic charge
solvent, this potential drop is usually at the electrode surface and the ionic
relatively small, mostly within a few countercharge at the solution side. The
dozen mV or less (but it may sometimes possibility to freely change the interfacial
reach a much larger value for dilute potential difference implies a qualitative
solutions even in the same solvent). analogy to a condenser whose separated
Elaborated procedures for its calculation charge is directly related to the above-
as a function of the solutions’ composition mentioned quantity, dQ.
exist [12]. For nonidentical solvents, this This simple picture gives a reasonable
potential difference may be much larger, idealization of the interface in the case of
when the electronic and ionic charges of components of the solution. As a result
the EDL are separated by a layer of the the EDL includes this ‘‘fixed’’ charge at the
solvent molecules, as it is in the case of a surface and the counterions in the solution
‘‘surface-inactive electrolyte’’ discussed in part. For a further discussion, see Ref. [14].
the following text. The specific adsorption Such charged interfaces are typical for
of the ionic-solution components, that is, various ionic or polar solids, for ion
their significant accumulation in the im- exchange, lipid or biological membranes,
mediate contact with the metal surface, or for any insulating surface in the
is usually accompanied by the strong re- presence of amphiphilic solute species.
distribution of the electronic charge near Even though one cannot polarize this
each adsorbed species. As a result they interface from the external source, the
may possess a much lower charge, com- potential difference and the separated
pared to the one in the bulk solution charge can change as a function of the
(it is termed as adsorption with a par- solution composition, in particular its pH
tial or complete charge transfer). Under and ionic strength.
these conditions the charge dQ supplied In many cases the solvent molecules
via the external circuit is not equal to the and solutes penetrate into the surface
change of the ‘‘free electronic charge’’ dis- layer of the insulator, thus forming a ‘‘gel
tributed along the surface of the metal. layer’’ composed of the components of
Such ‘‘perfectly polarizable electrodes’’ re- both phases. Its thickness is sometimes
quire a more elaborated description; see significant. Then the charge distribution
Chapter 3.1 of this book [13]. across such interphase is quite different
At approaching the limits of the ideal- from the simple ‘‘condenser’’ idealization,
polarizability interval, the rate of faradaic in particular the overall potential differ-
processes at the interface increases rapidly ence between the bulk phases is composed
(discharge of the solvent or solute compo- of two contributions for the insulator–gel
nents), with a dc current passing through layer and gel layer–solution interfaces,
the system. with a plateau between them, if the gel
The existence of this ‘‘ideal-polariz- layer is sufficiently thick [15].
ability’’ range of potentials is of great This two-step profile of the potential
practical importance, since it allows one across the interphase is also characteris-
to study an individual faradaic process of tic for certain types of modified electrodes in
the added electroactive species in the pres- which the metal surface is coated with a
ence of the background electrolyte, whose film, whose thickness significantly exceeds
presence provides important advantages the atomic scale. Such systems represent
(diminished Ohmic potential drop, rapid a much more complicated type of electri-
charging of the EDL, etc). fied interfaces, since the distribution of
A specific type of the electrified interface charged species depends crucially on the
is given by the contact of an insulating specific properties of the film. In most
phase with electrolyte solutions. Since the cases of sufficiently thick films, the pro-
charged species cannot cross the insulator, file of the Galvani potential across the
the EDL formation originates from the ion- interphase possesses a plateau inside the
ization process of surface groups (most fre- bulk film separating two potential drops
quently, the proton-based dissociation or at its interfaces with the electrode and the
association) or/and the adsorption of ionic solution [16].
Traditionally, special attention in This equation enables one to completely

electrochemistry has been paid to the determine the charge–potential relation.
metal–solution interfaces. The interfacial Moreover, the knowledge of the interfa-
potential drop and the corresponding cial tension–potential curves for various
electrode charge within the interval of the electrolyte concentrations enables one to
ideal polarizability can be regulated by the find all other equilibrium characteristics
imposed external voltage. The change of of the interface, without any additional
this voltage requires a supply of extra measurements, see the following text.
electron and ion charges to both sides However, this electrocapillary approach
of the interface across the corresponding to the interfacial properties is actually less
bulk media. This charging process is popular than the one based on capaci-
mostly limited by the transport in the tance curves, because the latter has been
solution (as well as inside the electrode more developed instrumentally and can
in certain cases), since the relaxation also be used in the case of solid elec-
time inside the interfacial region itself trodes. Furthermore, it does not require a
(Debye–Falkenhagen time determined by graphical/numerical differentiation of the
the double-layer thickness and the ion- experimental curve, which is related to an
diffusion coefficient, L2D /D) is extremely increase of the measurement errors.
short. Generally the current distribution Within the latter method, the charge–
along the electrode surface is nonuniform potential relation is obtained by the
due to geometrical reasons (shape and numerical integration of capacitance data
positions of electrodes), hydrodynamic for various potentials, according to Eq. (2).
structure, electrode surface properties, and The difficulty of this treatment is to find
so on. If the effect of these factors is the integration constant, that is, the value
sufficiently small, the charging process of the charge for some potential must be
can be described by a simple RC circuit known.
composed of the Ohmic resistance of the For liquid electrodes, one can directly
solution and the interfacial capacitance measure the potential of zero charge
(multiplied by the surface area, S) defined (p.z.c.), Eσ =0 , with the use of an open-
in Eq. (2). This result is widely used circuit streaming electrode. Then the
for the measurement of this important charge Q for any potential E is ob-
interfacial characteristic, with the use tained by integration of the capacitance
of chronoamperometry or electrochemical curve, C(E), from Eσ =0 to E. For solid
impedance. electrodes, the experimental p.z.c. determi-
For liquid metals (mercury, gallium) or nation represents a serious problem since
their alloys, one can measure another in- only indirect methods are available [18,
terfacial quantity, interfacial tension equal 19]. Moreover, the very definition of this
to the specific energy of the interface important characteristics is ambiguous in
formation, γ , at different values of the elec- this case because of the crystallographic
trode potential, E [1, 17]. At equilibrium inhomogeneity of the solid surfaces (ex-
the Lippmann equation relates it to the cept for perfect single-crystal faces), a
electrode-charge density σ , that is, to the time-variation of their properties and the
charge Q in Eq. (2) per unit surface area: adsorption of the solution components
Q dγ with the partial charge transfer, see [13,
σ = =− (3) 20] for a further discussion. For systems
S dE
in which these complications are assumed anion-reversible electrodes in the case

to be not especially important, the most of a binary electrolyte solution, Mzi A−zi .
popular method of the p.z.c. determina- Then the electrocapillary equation gives
tion is to measure capacitance curves for a a relation of the above-mentioned quanti-
series of dilute concentrations of an elec- ties to the chemical potential of the salt,
trolyte, which is not adsorbed specifically. µ = µ+ + µ− , and the Gibbs adsorptions
If the capacitance minimum is observed of cations or anions, ± :
at about the same potential and the mini-
mum depth increases monotonously with dγ = −σ dE+ − − dµ
diluting the solution, then this potential of ≡ −σ dE− − + dµ (4)
the minimum is identified to Eσ =0 (for
symmetrical electrolytes). A more thor- Hence,
ough analysis includes the treatment of
∂γ ∂γ
the whole capacitance curves in accordance + = − , − = −
∂µ E− ∂µ E+
with the Grahame theory, see the follow-
ing text. As soon as the p.z.c. value has (5)
been determined, one can calculate the It means that the knowledge of the inter-
charge-potential curves similar to the case facial tension as a function of any electrode
of liquid electrodes. The transition to the potential and the chemical potential of the
charge density, σ , requires the knowledge electrolyte, that is, of its concentration,
of the real surface area, A. For liquid elec- enables one to determine all other interfa-
trodes, the latter can be identified with the cial quantities by differentiation (Eqs. (3)
geometrical area, while its determination and (5)).
is much more complicated in the case of a It is often more useful to choose another
solid surface [21, 22], see Sect. 2.1.6. electric variable, electrode charge (density)
σ instead of electrode potentials. Then,
2.1.3 one needs to make a transition to the
Thermodynamics of Electrified Interfaces interfacial pressures (dependent on the
aforementioned choice of the reference
In the aforementioned discussion of electrode):
Eqs. (2) and (3), the solution composition
ξ± = γ + σ E ± (6)
was considered to be fixed. A compari-
son of the data for different electrolyte for which
concentrations opens a way to extract
extensive additional information on the dξ+ = E+ dσ − − dµ,
interfacial properties on the basis of the
dξ− = E− dσ − + dµ (7)
Gibbs thermodynamics. To exploit this
possibility one has to deal with elec- One by one, the coefficients in these
trode potentials (instead of the Galvani relations may be represented as functions
potential). This quantity is defined as the of the same variables, σ and µ:
voltage between the working electrode and
m/s dµ
the reference electrode immersed to the dE+ = C −1 dσ − t− ,
same solution and reversible with re- zi e
spect to one of its ionic components, m/s dµ
for example, E± for such cation- or dE− = C −1 dσ + t+ (8a)
zi e
m/s dσ dµ the others if the charge vs. potential de-

d+ = −t+ + Cµ ,
zi e zi2 e2 pendence has been found for different
electrolyte concentrations [24]. However,
m/s dσ dµ this procedure based on integration of
d− = −t− + Cµ (8b)
zi e zi2 e2 m/s
t+ ≡ zi e∂E+ /∂µ over the charge to de-
termine the Gibbs adsorptions of ions,
Because of the general relations:
± , requires to know an ‘‘integration con-
µ stant’’, that is, the values of ± as a
E+ − E− ≡ − , σ ≡ zi e(− − + )
zi e function of the electrolyte concentration
(9) for some electrode charge, for example,
the differentials (8a, 8b) contain only three σ = 0. The slope of the latter dependence
independent characteristics of the inter- is directly related to the aforementioned
face, each of them providing important parameter Cµ .
information on its structure [23–25]. C is Such information (supplementary to ca-
the interfacial capacitance per unit surface pacitance data) may be obtained from
m/s m/s electrocapillary experiments. According to
area, see Eq. (2); t+ = 1 − t− are the
Eq. (3), the interfacial tension, γ , passes
‘‘interfacial numbers of cations and an-
the maximum at the p.z.c. as a function
ions’’: they show which fraction of the
of the potential (at a fixed concentration).
electrode charge variation, dσ , is com-
In dilute solutions this maximum value
pensated by the corresponding ion (at a
of γ for the Hg–water interface is al-
constant solution composition)
most independent of the concentration
m/s ∂σ+ m/s ∂σ− of the ‘‘surface-inactive electrolyte’’ (see
t+ = , t− = (10) following text). The data for more concen-
∂σ ∂σ
trated solutions recalculated with the use
σ± = ±zi e± are the partial ion charges of Eq. (5) show a linear dependence on the
inside the EDL; alternatively, one can bulk concentration, c:
use partial cation and anion capacitances
related to the contributions of these ± = −zH c (12)
ions into the change of the whole ion
countercharge: where the factor, zH , is within 0.2–
0.4 nm [26, 27].
m/s ∂σ+
m/s For the interpretation of this finding,
C+ ≡ t+ C = ,
∂E one must recollect that the Gibbs adsorp-
m/s m/s ∂σ− tion represents an integrated difference of
C− ≡ t− C = ,
∂E the actual concentration profile for this
(µ = constant, E = E+ or E− ) (11) species across the interface, and the ‘‘un-
perturbed’’ one in which the concentration
Cµ = zi2 e2 ∂± /∂µ at constant electrode is equal to the bulk value up to the Gibbs
charge, see the discussion in the follow- plane and zero beyond it. The position of
ing text. the Gibbs adsorption plane is chosen here
One should emphasize that all these from the condition of the zero adsorption
quantities represent equilibrium charac- of the solvent.
teristics of the interface, and the afore- Since the electrolyte concentration in the
mentioned identities allow one to calculate above-mentioned experiments was not too
high (less than or about a mole per liter), Interfacial thermodynamics has become
the Gibbs plane is sufficiently close to the one of the bases of the theory of the electro-
physical boundary between the metal and chemical phenomena determined by the
the adjacent solvent layer. Then the above charge transport, in particular in mod-
negative values of ion Gibbs adsorptions at eling the electrochemical impedance [25,
the p.z.c. and the linear slope in Eq. (11) 28]. The analysis of the results obtained
points to the existence of an ion-free layer within the framework of this approach
of the solvent near the metal surface, that has shown that for transport processes
is, a compact layer, the existence of which under usual electrochemical conditions,
has been postulated long before these the characteristic timescales of the interfa-
experiments. The factor in this relation, cial charging and of the diffusion process
zH , corresponds to its thickness, that is, are well separated, and one can use a
to the distance of the closest approach of simplified treatment in which the inter-
the centers of nonadsorbing ionic species facial charging effects are characterized
to the metal surface, compared with the by a single parameter, C. In particular,
similar quantity for the solvent (Gibbs the relaxation after a potential step or the
adsorption plane). impedance in the proper frequency range
Surprisingly this important experimen- may be represented as an in-series com-
tal information on the properties of the bination of the solution resistance and
compact layer has been ignored in mod- the interfacial capacitance. On the other
ern microscopic theories of the interfacial hand, the interfacial contribution into the
structure, while the thickness of the ion- impedance arising from the perturbation
free layer at the electrode surface could be of the electrolyte concentration outside of
calculated on the basis of the distribution the EDL becomes essential for thin-layer
functions for the solvent and ions. systems in which the characteristic fre-
Effects of the interfacial charging are of quencies are overlapping or insufficiently
importance in various kinetic phenomena, separated.
in particular in the ion transport. The
simplified treatment assumes that these 2.1.4
effects can be adequately described by Classical Models of Electrified Interfaces
a single parameter, the capacitance C.
However, the passage of a nonstationary 2.1.4.1 Helmholtz Model
current is accompanied by the time- The thermodynamic treatment does not
variation of the electrolyte concentration allow one to extract more detailed infor-
outside the EDL region, and it must mation on the interfacial structure. The
give an additional contribution to the already discussed experimental results for
change of the interfacial potential, Eq. (8a). the ion-free solvent layer may be consid-
Besides, there is generally an imbalance ered as an illustration of the simplest de-
between the partial ion fluxes inside the scription of the interface in the Helmholtz
diffusion layer and the time derivatives of model. The latter postulates two charged
the corresponding partial ionic charges, planes containing the extra electrons and
σ± , Eq. (10). These factors result in the the counterions, which are separated by
dependence of the nonstationary response a solvent (‘‘compact’’) layer having thick-
of the system on all three interfacial ness zH , that is, the system represents
parameters, C, t+ m/s and Cµ . a nanocondenser. Then the ratio of the
electronic charge density at the electrode, In electrochemical kinetics, this model

σ = Q/S, and the interfacial potential dif- corresponds to the Butler–Volmer equa-
ference, ϕ m/s , gives the integral EDL tion widely used for the electrode reaction
(interfacial) capacitance (per unit surface rate. The latter postulates an exponential
area): (Tafel) dependence of both partial faradaic
σ currents, anodic and cathodic, on the over-
KH = (13)
ϕH all interfacial potential difference. This
where assumption can be rationalized if the elec-
tron transfer (ET) takes place between the
ϕH = ϕ m/s − ϕ m/s σ =0 ; electrode and the reactant separated by
the above-mentioned compact layer, that
ϕ m/s = E + ϕ ref (14)
is, across the whole area of the poten-
Here, ϕ ref is the potential drop at the ref- tial variation within the framework of the
erence electrode. Another constant term, Helmholtz model. An additional hypothe-
ϕ m/s σ =0 , corresponds to the potential sis is the absence of a strong variation of the
drop at the p.z.c. In the contemporary in- ‘‘electronic transmission coefficient’’, for
terpretation, its nonzero value is attributed example, in the case of adiabatic reactions.
to the potential variation inside the surface The fundamental drawback of the classi-
layer of the metal (affected by the adjacent cal Helmholtz picture is its inability to
solvent molecules) as well as to the polar- provide any explanation for the depen-
ization of the solvent layer, see Sect. 2.2. dence of the electrode charge (or of the
Equation (13) may also be rewritten as: capacitance) on the electrolyte concentra-
tion, which is observed experimentally in
 −1
sufficiently dilute solutions. Besides, there
σ
d
dσ  KH (σ )  is a large group of electrokinetic phe-
CH = ; CH ≡  

 nomena that demonstrate a much greater
dϕH dσ
extension of the EDL than it is assumed
(15) in the Helmholtz condenser model (see
containing the differential capacitance of Sect. 5.4).
the Helmholtz layer.
Equations (13)–(15) reproduce qualita- 2.1.4.2 Gouy–Chapman–Stern–Grahame
tively the experimental findings for suffi- Model
ciently concentrated solutions of surface- Gouy [29] and Chapman [30] indepen-
inactive electrolytes. In accordance with dently proposed an alternative treatment.
expectations, the capacitance values and It is based on an analytical solution of
their dependence on the interfacial poten- the nonlinear Poisson–Boltzmann equa-
tial are characteristic for each particular tion for the electric potential created by
metal/solvent contact. One also observes a a system of mobile point charges near a
noticeable (sometimes, even strong) varia- charged wall imitating an electrode sur-
tion of the capacitance with the electrode face. However, in the original form of
charge. Equation (13) predicts a distorted this theory, most of its predictions were
(due to the dependence of KH on σ ) at variance with experimental data.
parabolic shape for γ (E), with a maximum The situation changed radically after
at p.z.c., see Eq. (3), also in conformity with the suggestion of Stern [31] to combine
experimental observations. the two approaches by Helmholtz and
Gouy–Chapman (GC), that is, to consider whole interface, K ≡ σ/(E − Eσ =0 ), to

the whole interphase as consisting of those of the compact and diffuse layers,
two layers, a compact (or ‘‘inner’’ or KH in Eq. (13) and KGC ≡ σ/ϕGC (σ, c):
‘‘Helmholtz’’) and a ‘‘diffuse’’ one. The
K −1 = KH −1 + KGC −1 (19)
former corresponds to the aforementioned
hypothesis of an ion-free layer of the Differentiation of Eq. (16) gives a similar
solvent at the metal surface. The diffuse relation for the corresponding differential
layer is located between the compact one capacitances:
and the bulk solution and the whole
C −1 = CH −1 + CGC −1 (20)
counterion charge is distributed inside this
region in accordance with the GC theory. see Eqs. (2) and (15).
The version of this theory, which is ac- dσ ε
tually used in interpretation of all data for CGC (σ, c) ≡ = LB
dϕGC 2e
surface-inactive electrolytes, was proposed
by Grahame [32]. The total potential drop × [σ 2 + (σ ∗ )2 ]1/2 (21a)
across the interface is written as a sum
For aqueous solutions at room tempera-
of the compact- and diffuse-layer contribu-
ture
tions:
CGC (σ, c) = 19.46[σ 2 + 137.8 c]1/2
ϕ m/s ≡ E + ϕ ref = ϕH (σ ) + ϕGC (σ, c)
(21b)
(16) where CGC is in units of µF cm−2 , σ in
where both terms depend on the electrode µC cm−2 , and c in M. In particular, at
charge density, σ , while only the second the p.z.c.
term depends on concentration c. The
CGC (0, c) = 228.5 c1/2 (21c)
latter corresponds to the GC theory, for
example, in the solution of a symmetrical One can rewrite Eqs. (17) and (21a) to
single-charge electrolyte: represent the charge and the capacitance
as a function of the potential drop within
2e
ϕGC (σ, c) = the diffuse layer, ϕGC :
εLB

σ 2 1/2
eϕGC
σ σ = σ ∗ sinh
× ln + 1+ (17) 2kB T
σ∗ σ∗
ε eϕGC
CGC = cosh (22)
Here, 4πLD 2kB T
1/2
∗ e 1 However, these widely quoted formulas
σ = LD =
2πLB LD 8πLB c are useless for real treatment of capaci-
tance data since the potential drop, ϕGC ,
e2
LB = (18) is unknown. Therefore in order to apply
εkB T Eqs. (20) and (21a), one must first find
ε is the bulk solvent dielectric constant, kB the charge–potential relations for all the
the Boltzmann constant, LD is the Debye available solution compositions.
screening length of the electrolyte solution The key feature of Grahame’s approach,
and LB is the Bjerrum length. which determines its success in interpre-
Division of each term in Eq. (16) by tation of experimental data, is its semi-
σ relates the integral capacitance of the phenomenological character. It postulates
that the compact (Helmholtz)-layer capac- layer determines the shape of the capac-
itance, CH , is independent of the concen- itance curves in dilute solutions near the
tration of the ‘‘surface-inactive electrolyte’’ p.z.c. where the theory predicts a pro-
(which does not penetrate noticeably into found minimum whose depth increases
the compact layer) but avoids to accept a with diminishing the electrolyte concen-
model hypothesis on its dependence on tration. The minimum is usually well
the electrode charge, CH (σ ), that is, the developed to about 0.01 M concentrations
latter is to be determined from experimen- while in systems with a high hump of
tal data. On the contrary, the properties of CH (σ ) near σ = 0 (such as for Ag or
the diffuse layer are completely defined by Au single-crystal faces) a deep diffuse-
Eqs. (17) and (21a), without any unknown layer minimum should already appear
parameter. at 0.1 M. All these qualitative predictions
The ratio of the two contributions in of the Gouy–Chapman–Stern–Grahame
Eq. (20) depends on the values of the (GCSG) theory are in line with exten-
electrode charge and the electrolyte con- sive experimental information for numer-
centration, as well as on the particular type ous metal–solution interfaces for surface-
of the media in contact. The diffuse-layer inactive electrolytes (i.e. for systems with-
capacitance, Eqs. (21a) and (21b), is grow- out specific adsorption of solute species,
ing with the increase of the absolute value see following text) [22, 33, 34].
of the charge and with the concentration.
Besides, it becomes almost independent of 2.1.5
the concentration far from p.z.c. Near this Diffuse and Compact Layer Properties
point, σ = 0, the diffuse-layer capacitance
shows a minimum, which is becoming 2.1.5.1 Grahame Model in Comparison
progressively deeper at diluting the solu- with Capacitance Data for Liquid Electrodes
tion: CGC (0, c) = 228 µF cm−2 for c = 1 M Let us now discuss how to verify this
but only 7.2 µF cm−2 for 0.001 M aque- theory quantitatively as well as to extract
ous solution of a 1,−1 electrolyte. The the information on the interfacial structure
compact-layer capacitance, CH , is within with the use of the data for the Hg/H2 O-
20–60 µF cm−2 for most ‘‘mercury-like’’ NaF system [35, 36] as an illustration. The
metals (Hg, Pb, In, Cd, Bi, Sb, etc) in measured capacitance curves for several
aqueous and nonaqueous solutions, with electrolyte concentrations are presented in
much greater values for some other sys- Fig. 1. One can notice the expected diffuse-
tems, such as Ag and Au single-crystal layer minimum for dilute solutions but the
faces, up to 100–150 µF cm−2 near p.z.c. whole shape of the curves is sufficiently
As a result, for sufficiently concentrated complicated, in particular within the range
solutions, about 1 M or higher, the diffuse- of negative potentials.
layer contribution in Eq. (20) represents The streaming electrode measurements
in most cases only a minor correction to show a constancy of the p.z.c. value, Eσ =0 ,
the compact-layer term. Then, one can for different concentrations, which is con-
even use the simple Helmholtz model sidered as the necessary condition for the
for qualitative interpretation of the data. absence of a specific ionic adsorption.
In particular, far from the p.z.c., this In this case one can obtain the charge-
approximation is acceptable for all con- potential relations for all concentrations
centrations. On the contrary, the diffuse by the integration of the corresponding
30
25
[µF cm−2]
C
1
20 2
3
4
15
10
0.3 0 −0.5 −1.0 −1.5
6 E
[V]
Fig. 1 Experimental capacitance curves for the mercury electrode in
aqueous solutions of NaF. Concentrations: 0.916 M (1), 0.1 M (2),
0.01 M (3), 0.001 M (4) [35, 36].
capacitance curve, that is, one plots the on the electrode charge and it may be
capacitance versus the charge for all con- even greater for millimolar solutions). The
centrations as in Fig. 2. These curves retain corresponding ranges of the dispersion are
the diffuse-layer minimum at low concen- shown in Fig. 2 for two dilute solutions.
trations but their behavior at high electrode Even though the initial dispersion in Fig. 1
charges has been strongly simplified, in was independent of the potential, the
particular all curves are practically identi- errors in Fig. 2 turned out to be markedly
cal at high negative charges. varying as a function of the electrode
Any justified comparison of measured charge. They are close to those of the
data with theoretical predictions must take corresponding C(E) curves at the zero
into account the experimental accuracy. charge and near the cathodic minimum.
However, the resulting error bars appear Then the errors are slightly enhanced at
to depend strongly on the choice of the dropping part of the cathodic branch,
the coordinate system, used for this but they are several times smaller at lower
treatment. In order to demonstrate the electrode charges, where the capacitance
latter, let us assume for simplicity that varies very rapidly. The latter effect is due
the measured capacitance data in Fig. 1 to a correlation of errors in the capacitance
have got the precision of ±0.1 µF cm−2 for and the charge within this interval. It
two highest concentrations, ±0.2 µF cm−2 implies that even a considerable deviation
for the 0.01 M solution, and ±0.5 µF cm−2 between the theoretical prediction and
for 0.001 M, within the whole potential experimental data in the initial C(E)
interval (in real measurements at low curves within this potential range may look
frequencies the dispersion also depends rather small in the coordinates of Fig. 2.
C
[µF cm−2]
30
1
2
3
4
20
10
15 10 5 −5 −10 −15 −20

σ
[µC cm−2]
Fig. 2 Capacitance versus electrode charge curves for the mercury
electrode in aqueous NaF solutions recalculated from the data in
Fig. 1. The error bars were calculated from those in the capacitance
versus potential curves (Fig. 1) that were assumed to have the
precision of ±0.1 µF cm−2 for the two highest concentrations,
±0.2 µF cm−2 for the 0.01 M solution and ±0.5 µF cm−2 for 0.001 M,
within the whole potential interval, as it is discussed in the text.
If one relies on the validity of the GC inside the experimental error dispersion
model for the diffuse-layer properties, one over the whole range of the electrode
can use Eq. (20) to calculate the values of charges. This comparison enables one
the compact-layer capacitance, CH , from to conclude that the deviation of the
experimental data for C and theoretical data for the highest concentration at
formula (21a) for CGC . This operation high positive charges is probably due to
performed for all available values of the the violation of some assumption of the
charge and the electrolyte concentration model, for example, dissolution of mercury
results in Fig. 3, in which the effect of the or specific adsorption of cations. The
experimental dispersion is shown again concentration dependence of the diffuse-
for two dilute solutions. layer capacitance is probably too weak to
According to Grahame’s hypothesis, explain this divergence, see in following
these curves for different concentrations text, even if the GC theory may be
should be coincident. One can notice inapplicable quantitatively.
first of all its brilliant confirmation in The curve for 0.01 M also demonstrates
Fig. 3 for the two highest concentrations; a marked increase of the experimental
a deviation is only seen at high positive dispersion near the zero charge. This
charges, above 10 µC cm−2 . The proximity effect is even more pronounced for the
of the compact-layer capacitance curves 0.001 M solution in which the dispersion
is even more impressive for the 0.1 M does not exceed ±0.8 µF cm−2 , except for
and 0.01 M solutions, whose difference is the range of very small charges where the
CH
[µF cm−2]
35
25
20
15 10 5 0 −5 −10 −15 −20

σ
(a) [µC cm−2]
CH
[µF cm−2]
64
35
30
25
20
15 10 5 0 −5 −10 −15 −20

σ
(b) [µC cm−2]
Fig. 3 Compact-layer capacitance versus electrode charge curves,
CH (σ ), for mercury electrode in aqueous NaF solutions calculated
from the data in Fig. 2 with the use of Eqs. (20) and (21a), for each
electrolyte concentration, 0.916 M (∇), 0.1 M (), 0.01 M (solid line in
Fig. 3a) and 0.001 M (solid line in Fig. 3b). Solid lines for lower
concentrations were obtained with the use of the error bars in Fig. 2 for
the corresponding curves. The error bars are small at the two higher
concentrations and are not shown.
uncertainty in the CH value is enormous. It charge, since it is too sensitive even to

means that the capacitance data for dilute small experimental errors. On the other
solutions cannot be used immediately to hand, assuming that the compact-layer
determine the CH value at zero electrode capacitance is a smooth function of the
charge, one can interpolate its values for theory predicts a linear dependence with a
positive and negative charges, to obtain unit slope for each value of the electrode
CH (0) for a dilute solution. Such a charge. Such plots for the same experi-
procedure in Fig. 3 results in the same mental data are shown in Fig. 4, again
value, about 29 µF cm−2 , for this quantity with indication of the error dispersion
as for the other concentrations. of each point. All lines in the figure
The curve for the 0.001 M solution de- have the theoretical slope. Their extrap-
viates downward noticeably for moderate olation to the ordinate axis gives the value
positive and negative charges. The most of CH −1 for the corresponding electrode
probable reason is the increase of exper- charge. The PZ method to check the va-
imental errors for such dilute solution lidity of the Grahame model is more
related to the current passage across the objective (since it suggests verifying the
solution. linearity of the plots and their slope), es-
An alternative and the most commonly pecially if the experimental dispersion is
used way to verify the Grahame model is not taken into consideration. However, in
based on Parsons–Zobel (PZ) plots [37], most publications (especially for solid elec-
in which the inverse values of the exper- trodes) this method has been used in a
imentally measured capacitance, C −1 , are reduced form, where only the data for
shown as a function of the inverse diffuse- the p.z.c. are plotted, thus giving only
layer capacitance given by Eq. (21a). The CH (0). It means that the Grahame theory is
1 2 4
0.07
0.15
1
C −1
C −1
0.10 0.05 5
2
6
0.05 3 0.03
0 0.05 0.10 0.02 0.04 0.06

(a) CGC−1 (b) CGC−1
Fig. 4 Parsons–Zobel plots, C−1

versus CGC −1 , indicated at each graph (in µC cm−2 ): 0 (plot 1),
for the mercury electrode in contact with +1 (plot 2), +2 (plot 3) in Fig. 4(a), −3 (plot 1),
aqueous NaF solutions, at several constant −2 (plot 2), −1 (plot 3), +1 (plot 4), +2 (plot 5),
values of the electrode charge density, σ , that are +3 (plot 6) in Fig. 4(b).
verified incompletely. Besides, it does not Analysis of this figure leads to a

immediately give the properties related to surprising conclusion. It was accepted
individual parts of the interface, like the long ago that the capacitance data could
CH (σ ) curve for the compact layer, or the only provide valuable information on the
characteristics of the diffuse layer. diffuse-layer properties in a close vicinity
The latter information may be ex- of the p.z.c., since this part of the in-
tracted with the use of another coordinate terphase gives a minor contribution to the
system [38], δC −1 versus σ , where δC −1 total interfacial capacitance outside this po-
is the difference of inverse experimen- tential range, even in very dilute solutions.
tal capacitance values for two electrolyte However, this reasoning does not take into
concentrations corresponding to the same account the very high precision of the ca-
electrode charge: pacitance measurements, which ensures a
small dispersion of experimental points in
δC −1 ≡ [C(σ, c1 )]−1 − [C(σ, c2 )]−1
Fig. 5 even for high electrode charges. This
(23a)
is especially true for the positive branch,
These curves can be plotted immediately
where the dispersion is much less than
on the basis of the experimental data
the value itself. The points in Fig. 5 are
for C(σ, c) dependencies, see in the
limited to the range of ±10 µF cm−2 since
preceding text. For example, treating
the positive branch above this interval may
the data for Hg/H2 O-NaF interface in
be perturbed by the process of mercury
Fig. 2 leads to Fig. 5, in which the three
dissolution. One should keep in mind
lower concentrations, 0.1 M, 0.01 M and
that these experimental capacitance curves
0.001 M, were used for c1 , while c2 was
could be used to check any theory of the
taken each time equal to the maximum
diffuse-layer structure within a wide range
concentration. The points thus found are
of electrode charges, since the GC model
shown together with the corresponding
has not been used to obtain experimental
experimental dispersion.
points in these coordinates.
On the one hand, it is sufficient to accept
Figure 5 also demonstrates the results
one of the Grahame’s hypotheses, the one
of this check for the GC model. One
on the independence of the compact-layer
can see that it is able to reproduce ade-
capacitance of the electrolyte concentra-
quately the data at the p.z.c. as well as
tion, in order to relate the δC −1 value
within a wide range of the electrode charge,
to the characteristics of the diffuse layer,
for both 0.1 M and 0.01 M solutions. One
that is, the change of its inverse capac-
itance [Cdiff (σ, c)]−1 , with concentration, can see a noticeable discrepancy for the
at a fixed value of the electrode charge: 0.001 M concentration, even for moder-
ate electrode charges. However, the latter
δC −1 ≡ [Cdiff (σ, c1 )]−1 − [Cdiff (σ, c2 )]−1 has probably a different origin since one
(23b) should hardly expect that the mean-field
On the other hand, if we also accept type Poisson–Boltzmann theory fails at
the second postulate of the Grahame lower concentrations, while being applica-
theory, then Cdiff (σ, c) can be calculated ble for higher ones!
with the use of the GC model, Eq. (21a). These results represent a challenge for
The corresponding theoretical curves are the modern statistical-mechanical theories
also shown in Fig. 5, in comparison with of the interfacial structure, since the
experimental ones. values of δC −1 in Eq. (23b) could be
δ C −1
[µF cm−2]
14
12
10
3 2
2
1
10 8 6 4 2 0 −2 −4 −6 −8
σ
[µC cm−2]
Fig. 5 Experimental and theoretical δC −1 versus σ curves for mercury/aqueous NaF
solution [38], see Fig. 2, and the definitions in Eqs. (23a) and (23b). Error bars for the
experimental curves correspond to those in Fig. 2.
immediately calculated from the theo- theories, no attempts have been made to
retically derived ion concentration pro- subject them to the high precision ex-
files. Although drastic deviations from the perimental verification in accordance with
GC model have been predicted in these Eq. (23b).
One should avoid, however, the mis- range of its applicability to higher ionic
take of interpreting the success of the GC strengths.
passage through this test as its complete
confirmation, in particular the proof of the 2.1.5.2 Compact Layer Properties
validity of Eq. (21a) for the diffuse-layer The exposed variant of the theory has
capacitance at high electrode charges. In been applied to the study of numerous
reality the theoretical lines in Fig. 5 based metal–solution interfaces, in particular for
on this formula demonstrate a weak de- liquid metals, Hg, Ga, and their alloys.
pendence of the diffuse-layer capacitance First of all for a particular metal/solvent
on the concentration (except for the p.z.c.), boundary one has to ensure ‘‘no surface
especially in dilute solutions (it is the ef- activity’’ of the solute components, that
fect, which is correctly reproduced by the is, their absence inside the compact layer.
model). However, the dominant part of the Experimentally the necessary condition of
diffuse-layer capacitance has been cancelled this surface inactivity of an ionic species
(together with the compact-layer contri- in the vicinity of the p.z.c. (which can
bution) by the subtraction in Eq. (23b). be measured directly for liquid electrodes)
Therefore the data in Fig. 5 cannot give is assumed to be the constancy of this
any information on the charge dependence potential, Eσ =0 , for different electrolyte
of the diffuse-layer capacitance far away concentrations. Then the data are com-
from the p.z.c. pared with the predictions of the Grahame
One has to keep in mind this impossi- model, as an additional check of the surface
bility to separate the compact-layer term inactivity.
and the concentration independent part of One should keep in mind that a de-
the diffuse-layer contribution in Eq. (20) tailed analysis has only been carried out for
at significant electrode charges. It results the Hg–H2 O-NaF interface (see previous
text), in which both postulates (concen-
in an uncertainty in the values of the
tration independence of the compact-layer
compact-layer capacitance, which is of pri-
properties and the GC model for the dif-
mary importance in the modeling of the
fuse layer) seem to be confirmed. For some
metal–solvent interfacial structure.
other systems, the Grahame treatment was
Recently the GC model has been mod-
applied, to determine the compact-layer ca-
ified to take approximately into account pacitance curve, CH (σ ), from the data for
the screening effects for ions inside the the highest electrolyte concentration and
diffuse layer [39, 40]. For this goal, the to calculate on its basis the ‘‘theoretical’’
electrolyte concentration in the equations C(E) curves for more dilute solutions, for
of the GC model (see previous text) were their comparison with experimental ones,
replaced by the electrolyte activity in the without taking into account the experimen-
bulk solution. This treatment can be justi- tal dispersion. Moreover, in most cases the
fied from the fundamental principles for a analysis was even reduced to constructing
low charge density inside the diffuse layer the PZ plot for the p.z.c. only. Generally the
(in the vicinity of the p.z.c.). On the other Grahame theory seems to provide at least a
hand, the ion screening in this layer at qualitative description of the results, even
high electrode charges is quite different though certain ‘‘anomalies’’ have been ob-
compared with the bulk solution. Never- served for some systems (e.g. for liquid
theless this modification may extend the gallium [41]), see [38] for review.
Thus-found compact-layer capacitance solvent molecules (see Sects. 2.1.7–2.1.9),

curves turned out to be strongly depen- so that the ‘‘continuous dielectric’’ consid-
dent on the specific type of the metal (or eration alone may become totally mislead-
alloy) and the solvent, be it water or a non- ing. In particular, this simplified treatment
aqueous medium. On the other hand, the identifies two physically different interfa-
curves for different electrodes in contact cial characteristics, the ion-free layer of
with the same solvent are practically merg- the solvent and the region of a lowered
ing at high negative potentials (but not at dielectric permittivity (compared with the
positive ones). This experimental fact is bulk-solvent one). The former manifests
still waiting for its interpretation. itself in Gibbs adsorption measurements
The simplest interpretation of the of surface-inactive ions, Eq. (12); while the
compact-layer capacitance is represented latter is related to the spatial correlation
by the Helmholtz model of the slab filled properties of the solvent polarization, see
with a dielectric continuum and located Sect. 2.1.7.
between a perfect conductor (metal sur- This distinction can be checked ex-
face) and ‘‘the outer Helmholtz plane’’ perimentally if the measurements are
considered as the distance of the closest performed for different surface-inactive
approach of surface-inactive ions. Experi- electrolytes but for the same electrode
mental determination of its thickness, zH , and solvent. According to the general the-
may be based on Eq. (12). Moreover, its ory [42–44] discussed in Sect. 2.1.7, the
dielectric permittivity, εH , is often con- distance of the closest approach, which
sidered as a constant across the whole should be noticeably different for these
compact layer. Then its value can be esti- electrolytes, has got almost no effect in
mated from the values of the compact-layer the compact-layer capacitance if the ions
capacitance, for example, it gives about 6 do not penetrate into the region of the re-
or 10 (depending on the choice of zH ) duced dielectric response near the surface.
for mercury–water interface, that is, a This theoretical prediction turns out to be
value that is much lower than the one in conformity with experimental data [35,
in the bulk water, 80. This diminution 37, 45, 46] for three mercury–aqueous so-
was interpreted as a consequence of the lution interfaces for which the PZ plot
‘‘dielectric saturation’’ of the solvent in at the p.z.c. gives practically identical
contact with the metal surface, its modified values for the compact-layer capacitance,
molecular structure or the effects of spatial CH (0) ∼= 29 µF cm−2 (Fig. 6).
inhomogeneity. The effective ‘‘dielectric One should keep in mind that the
permittivity’’ of the compact layer shows theory predicts some effect of the ion
a complicated dependence on the elec- properties on the compact-layer capaci-
trode charge, which cannot be explained tance, which should manifest itself ex-
by the simple hypothesis of the saturation perimentally for systems with high val-
effects on one hand or by the unperturbed ues of this capacitance. Preliminary data
bulk-solvent nonlocal polarizability on the for single-crystal faces testify in favor
other hand. of this conclusion, but the theoreti-
Later studies revealed a more compli- cal relation has not been checked as
cated structure of the compact layer whose yet. Interpretation of the observed val-
properties are determined by the distribu- ues of the compact-layer capacitance
tion of the metal electrons and discrete and its dependence on the electrode
0.15
C −1
0.09
0.06 0.12
CGC−1
Fig. 6 Parsons–Zobel plot at the p.z.c. for mercury electrode in
contact with various aqueous electrolyte solutions [38]. Capaci-
tances are given in µF cm−2 . Experimental capacitance values:
NaF(◦) [35], NaH2 PO4 () [37] and Na2 SO4 (, ∇) [45, 46].
charge requires a microscopic model of attains significant values, being depen-

the interphase, as discussed in Sects. dent on both the electrode charge and the
2.1.8– 2.1.10. electrolyte concentration, Eq. (17). Since
the ET distance between the reactant
2.1.5.3 Manifestations of the Diffuse Layer and the electrode does not exceed a few
in Kinetics of Electrode Reactions angstroms under usual conditions, the
The existence of the EDL at an interface reacting species must approach the com-
is of crucial importance for all interfacial pact layer or even enter it. Therefore the
electrochemical phenomena, in particu- ET rate is only affected by a fraction of
lar for the kinetics of electrode reactions. the overall potential difference, ϕ m/s in
In most cases the principal factors influ- Eq. (16), equal to ϕ m/s − ψ1 where ψ1
encing their rate are the overall potential corresponds to the average potential at the
difference across the interface (e.g. accord- point of the reactant localization, z∗ . Be-
ing to the Butler–Volmer equation for a sides, its concentration varies across the
simple ET reaction) and the specific ad- diffuse layer in accordance with the Boltz-
sorption of reactants, reaction products or mann law.
intermediates. However, there are impor- Frumkin has combined this reasoning
tant cases in which the diffuse layer plays with his theory of the slow discharge [49]
a dominant role [47, 48]. to derive the expression for the rate of
In a dilute solution the potential an electrochemical reaction [50–52], for
difference across the diffuse layer, ϕGC , example, for the cathodic component of
the faradaic current: from p.z.c., Eqs. (16), (13) and (17). Then
◦ the ratio of these potentials
ic = neSkc c(z∗ )
ϕ m/s ∼ CGC (0, c)
ne =1+ (26)
× exp −α(ϕ m/s − ψ1 ) ψ1 CH (0)
kB T
determined by the ratio of the diffuse and
◦ ◦
= neSkc c × exp −α(ϕ m/s − ψ1 ) compact-layer capacitances at the p.z.c. is
about 1 (or even close to 1), see Eq. (21a).
ne zi e The type of the effect on the polarization
− ψ1 (24)
kB T kB T curve depends also on the sign of the co-
efficient, zi − αn. In most cases this sign
Here, n is the number of transferred
is determined by the charge valence of the
electrons (n > 0), S the surface area, kc◦
reactant, zi . In the case of the electrore-
the cathodic rate constant, α the ca-
duction of a cation (e.g. for the hydrogen
thodic charge-transfer coefficient, and zi
evolution or a metal deposition) zi > 0,
the charge valence of the reactant. This
and the diffuse-layer factor Kψ1 changes
expression should be modified to take into
in the same direction as the Tafel term, exp
account the adsorption energies of the re-
(−αϕ m/s ne/kB T ). Therefore the former
actant and the product if the ET takes place leads to a deviation from the straight line
for a species inside the compact layer [11]. with the 2.3 kB T /αne slope in the Tafel
The value of the coordinate, z∗ , which coordinates, but the current-potential de-
determines the ψ1 potential is a priori un- pendence remains monotonous. Besides,
known. In most studies it was postulated zi and αn have opposite signs and for
that the ET took place at the species located single-charged cations, the coefficients of
at the outer Helmholtz plane (the bound- the diffuse-layer effect represents a rather
ary between the compact and diffuse lay- small correction to the Tafel dependence.
ers). Then, ψ1 coincides with the potential Nevertheless it cannot be ignored, both for
drop within the diffuse layer, ϕGC , Eq. (17). the quantitative treatment of the polariza-
Equation (24) contains an additional tion curves and for the interpretation of the
factor, dependence of the current on the concen-
tration of the indifferent electrolyte [53].
e
Kψ1 ≡ exp −(zi − αn)ψ1 (25) Quite a different situation takes place
kB T
for the reduction of anions, or for the
compared to the conventional Tafel for- oxidation of cations. Then the diffuse-
mula. In sufficiently concentrated solu- layer multiplier, Kψ1 , varies in the opposite
tions, for example, in the presence of direction compared to the Tafel term. For
the background electrolyte, the Frumkin sufficiently dilute solutions, the former
correction may be neglected. On the con- factor dominates and the cathodic current
trary, it may radically change the shape diminishes at the negative shift of the
of the polarization curve in dilute solu- electrode potential in the vicinity of
tions near p.z.c., depending on the value the p.z.c. At higher electrode charges
of zi − αn. This effect originates from the (positive or negative) the variation of
linear dependence of the overall (ϕ m/s ) the diffuse-layer potential becomes very
and diffuse-layer (ψ1 ) potential drops on slow (logarithmic function of the electrode
the electrode charge, σ , at small deviations charge, Eq. (17)) compared with the overall
potential because the compact layer gives then a deep minimum of the current
the principal contribution. Within these within the range of small negative elec-
potential ranges the variation of the trode charges.
diffuse-layer factor only modifies the global Experimentally such nonmonotonous
tendency determined by the Tafel term, curves were observed for numerous re-
that is, the cathodic current increases actions of the anion electroreduction:
for negative potential shift. As a whole, S2 O8 2− , S4 O6 2− , Fe(CN)6 3− , PtCl4 2− , and
Eq. (24) predicts a nonmonotonous current- so on, where all above-mentioned qual-
potential curve in dilute solutions, with itative predictions of the slow-discharge
a dropping branch in the vicinity of the theory were confirmed [54]. For the quan-
p.z.c. The amplitude of this effect depends titative treatment, it was suggested that
strongly on the charge of the reactant, zi , corrected Tafel plots be used in co-
and the electrolyte concentration, and the ordinates: log ic + (zi e/2.3kB T )ψ1 , E −
current of the anion reduction may almost ψ1 [55], where one should observe a
vanish within a wide range of negative straight line with the slope −αne/2.3kB T .
electrode potentials, E − Eσ =0 < 0. Besides, the data for different concentra-
One must keep in mind that all this anal- tions of the background electrolyte should
ysis was performed for systems without a fall on the same line [56]. These predic-
specific adsorption of reacting species or
tions have been confirmed (some cur-
indifferent electrolyte. The violation of this
vature of corrected plots was found in
requirement leads to the necessity to in-
some cases) by experimental studies of
troduce the energies of specific adsorption
the persulfate, tetrathionate, ferricyanide,
of reactants or their interaction with ad-
and some other anions [57].
sorbed species, into Eq. (24). Fortunately
An additional test of the theory is given
this problem is of less importance for the
by the comparison of the data for different
case of anion reduction, since the most
electrode materials. Experimentally the re-
interesting effects take place at the nega-
tively charged electrode surface in which duction of the same anion at different
anion-adsorption effects are minimal. metals leads to quite different patterns
The analysis in Eq. (24) does not take for the polarization curve, in particular, a
into account the transport limitations. drastic variation of the interval in which
Therefore these predictions of the diffuse- the diffuse-layer minimum is observed.
layer effects are only valid for the interval of According to the theory (24), if the react-
the potentials in which the rate of the pro- ing species do not enter the compact layer
cess is determined by the electron-transfer and the possible change of the electron-
step. If this kinetic regime corresponds tunneling factor does not influence the
to high negative electrode charges, the reaction rate (e.g. for adiabatic electro-
Kψ1 factor in Eq. (25) varies slowly and chemical reactions [58]) the corrected Tafel
the deviations from the Tafel behavior are plots must be independent of the electrode
rather weak. On the contrary, the anion material. This prediction has also been
electroreduction wave that starts at posi- confirmed experimentally for the persul-
tive electrode charges may demonstrate a fate reduction at Hg (amalgams), Bi, Sn,
complicated curve: a usual behavior within Pb, and Cd [59]. Similar results have been
this potential range, with approach to the obtained for the persulfate reduction in
limiting current at less positive potentials, some nonaqueous solvents [60, 61].
The exposed version of the model factors [64, 65] has allowed one to provide
disregards an essential feature of this class a quantitative interpretation of experimen-
of reactions, that is, the formation of ion tal data for anion electroreduction at high
associates between the reacting anion and overvoltages and has confirmed expecta-
background cations, which is especially tions of its proximity to the activationless
pronounced for the multicharged anions. region.
The ways to determine the charge of the
discharging species in the bulk solution 2.1.5.4 Electrokinetic Phenomena
(which is generally different from the one Another area in which the existence of the
inside the EDL) are analyzed in [62]. Some diffuse part of the EDL plays a dominant
deviations from the Frumkin theory may role is represented by various electrokinetic
also originate from the difference of the phenomena [67, 68].
closest approaches of anions and cations Some of them are observed at mem-
to the electrode surface [47]. branes separating two identical electrolyte
The Frumkin correction in Eq. (24) re- solutions. For example, in electroosmosis,
duces the EDL structure effect to its single one can impose a difference between
characteristic, ψ1 potential, while real re- the electric potentials in these solutions,
actants represent a distribution of electric by means of immersed electrodes, which
charges. A recent development has taken leads to the generation of an electric field
into account this distribution in the cal- across the membrane because of the cur-
culation of the work term, so that it rent passage. As a response, one observes
depends on the whole profile of the poten- a solution flow across the membrane. For
tial across the EDL. Besides, the presence its qualitative interpretation, one considers
of the reactant perturbs this potential pro- pores of the membrane as a set of parallel
file in the vicinity of the species (a kind cylindrical capillaries. An EDL is formed at
of discreteness-of-charge effects [48] dis- the internal surface of each capillary, due
cussed in Sect. 2.1.11.3), due to both the to the ionization of surface groups or/and
field created by its charges and the cavity ion adsorption/association, accompanied
effects arising from the replacement of sol- by the accumulation of the counterions
vent molecules in the volume occupied by inside the diffuse layer. As soon as the elec-
the solute species [63, 64]. tric field E is imposed along the capillary
The quantitative theory also has to take surface, the force zi e E acts on each ion,
into account a variation of the reaction including those within the diffuse layer.
volume with the electrode charge [65]. An- Owing to the mobility of the latter and the
other weak point of the model (24) is the viscosity of the solvent, this force induces
hypothesis of the Tafel dependence on the movement of the solution along the
the potential difference across the compact capillary. The velocity of this flow (with
layer. Quantum-mechanical estimates [64] respect to the electric field amplitude) in
testify in favor of a considerable variation the stationary conditions is described by
of the charge-transfer coefficient, α, over the Smoluchowski equation:
the wide potential range explored exper-
v
imentally, or even its approach to zero, = (4π)−1 εη−1 ζ (27)
which corresponds to the activationless E
regime of the charge-transfer process [66]. where ε is the relative dielectric permittiv-
Recent analysis which has included these ity of the solvent, η the dynamic viscosity
of the solvent, and ζ (’’zeta potential’’) the to the solution, for example, under the
potential in the ‘‘slip plane’’ inside the influence of the gravity force (sedimen-
diffuse layer separating the hydrodynam- tation). The viscosity force results in a
ically immobile (attached to the surface) displacement of the mobile diffuse-layer
and moving volumes of the solution. This charges with respect to the fixed ones at
potential is often identified to the poten- the surface, that is, to the creation of an
tial of the outer Helmholtz plan, ϕGC , electric dipole. The superposition of the
Eq. (16). fields of such dipoles gives an electric field
An inverse phenomenon (streaming po- oriented along the movement axis (sedi-
tential), generation of an electric field mentation potential).
inside the membrane, takes place if the so- It should be noted that the direction of
lution passes through this porous medium the induced effect depends on the sign
due to an imposed hydrostatic pressure. of the diffuse-layer potential, so that one
This time it is the flow of the fluid inside can determine from such experiments the
the pores that induces the displacement of sign of the ‘‘fixed’’ surface charge or even
the mobile part of the EDL at the surface of estimate its value.
capillaries, with respect to the charges at- The variation of the electrolyte concen-
tached to the surface. These dipoles create tration or the solution composition leads to
an electric field, which, under stationary drastic changes in the properties of the col-
conditions, prevents the farther displace- loidal system, in particular to changes of its
ment of the mobile charges. The resulting stability, which is primarily determined by
potential difference across the membrane, the electrostatic repulsion between the par-
ϕ, is proportional to the excessive hydro- ticles. The latter is due to the formation of
static pressure, P : similar EDLs owing to the identical chem-
ical nature of the surfaces. The repulsion
ϕ
= (4π)−1 εη−1 −1 ζ (28) is only efficient at long distances between
P the particles (at shorter distances the in-
where is the electrolyte conductivity, and teraction between the particles becomes
the remaining notation is identical to that attractive due to van der Waals forces). This
in Eq. (27). is why long-range diffuse layers are neces-
The existence of the extended diffuse sary for the colloid’s stability. The increase
layer in dilute solutions is also of crucial of the electrolyte concentration strongly di-
importance in colloidal systems formed by minishes the Debye screening length, that
particles whose size is comparable to the is, the distance of the particle–particle in-
wavelength of light, that is, about 1 µm. For teraction, with the coalescence and the sub-
example, electrophoresis, the movement of sequent sedimentation. Another scenario
colloidal particles with respect to the so- of the kind takes place upon change of the
lution under the influence of an external solution pH. In most cases the fixed charge
electric field, originates from the displace- of the particle’s surface is formed at least
ment of the particle and the mobile part partially by the proton-exchange equilib-
of its EDL in the opposite directions under rium. The change of pH results in the mod-
the action of this field, since these com- ification of this surface charge, and the sys-
ponents of the system possess charges of tems loses its stability at the moment when
opposite signs. A related phenomenon oc- the surface charge passes zero and the re-
curs if the solid particles move with respect pulsion between the particles disappears.
2.1.6 areas:
S
Solid Electrodes: Effects of Polycrystallinity R≡ , R≥1 (29)
and Surface Roughness Sapp
2.1.6.1 ‘‘Uniform’’ Model of Solid

The real surface area S being unknown
Electrodes. Surface Roughness for solid electrodes, the charge density
A brief outline of this complicated and still and the capacitance in this case were
developing area is given in this section. A defined with the use of the apparent
more detailed description may be found in (or ‘‘geometrical’’) surface area, compare
Eqs. (3) and (2):
reviews [22, 33, 34] as well as Chapter 2.4
of this volume [20]. Q
Historically the interfacial properties of σapp ≡ = Rσ
Sapp
the solid metal–solution interface were
dQ
initially studied for polycrystalline (PC) Capp ≡ Sapp −1 = RC (30)
samples. At that time it was expected that dE
the characteristics of all metals should be In accordance with the Stern–Grahame
rather similar, saying nothing about the model of the EDL structure the values
different single-crystal faces of the same of C are determined by both the diffuse
metal. and compact-layer properties, the latter
The first studies of capacitance curves being dependent on the metal properties.
for numerous PC metals were qualitatively However, in very dilute solutions of a
in accordance with these expectations; for surface-inactive electrolyte the dominant
carefully prepared smooth PC surfaces in contribution to C near the p.z.c. (at the
the absence of the specific adsorption of capacitance minimum) is given by the
ionic or neutral species, a minimum was diffuse layer, C ∼
= CGC (0, c). Therefore the
observed in sufficiently dilute solutions, ratio of capacitances in these conditions
the depth of which increases for lower should be close to the RF for the surface of
electrolyte concentrations. Therefore the the solid metal M:
potential of this minimum was interpreted
Capp M
as the p.z.c., Eσ =0 , of the metal–solution R≡ (31)
interface. C Hg
The capacitance values at this minimum It became clear later that this treatment
for solid metals turned out to be notice- had certain defects. The data for most
ably higher compared with those for the dilute solutions (typically, within mM
mercury electrode at the same electrolyte range) are especially sensitive to various
concentrations. Since the interfacial prop- experimental errors, while the capacitance
erties of different metals were assumed values to be measured are very low.
during that period to be similar, it was nat- Another weak point of the reasoning is
ural to attribute these higher experimental the assumption that the solid surface is
capacitance values for solid electrodes to a uniform, which is hardly justified, as we
greater interfacial area. see in the following text. One should
On this basis, one even tried to estimate also keep in mind that this consideration
the roughness factor (RF) of the solid is inapplicable to the surface roughness,
surface, R, that is, the ratio of the ‘‘real’’ whose scale is smaller or comparable with
(S) and the ‘‘apparent’’ (Sapp ) surface the thickness of the diffuse layer, that is,
with the Debye screening length in the is more complicated). Besides, contrary
solution, which is about 10 nm in mM to the theoretical predictions, there was a
solutions. strong variation of the cathodic branch of
In a modified variant of this approach, the curve as a function of the electrolyte
the surface roughness was determined concentration, which was hardly possible
from the same ratio of capacitances (31) to attribute to the specific adsorption of
measured for the identical high negative cations. The anodic branch was even more
charges [69]. Under these conditions the irregular.
diffuse-layer contribution may be practi- As a result, within the initial period of the
cally neglected while the difference be- solid metal studies, the only quantitative
tween the compact-layer capacitances of treatment was made for the data at the
different metals becomes rather small, com- diffuse-layer minimum. In accordance
pared to the vicinity of the p.z.c. or at to Eq. (32), the corresponding PZ plots
positive electrode charges. The principal (dependence of Capp −1 versus CGC −1 ) for
shortcoming of this estimate is a signifi- various solid metals had lower slopes than
cant variation of the cathodic branch of the 1, which were interpreted as R −1 . The plots
capacitance curve with the electrolyte con- (if the concentration range was sufficiently
centration observed in most publications. broad) showed also a significant curvature
All this reasoning is based on the in most cases. The latter was attributed to
hypothesis that the separation (20) of the various experimental errors.
capacitance to unit area of the real surface, After having the RF determined one
C, into the compact- and diffuse-layer could calculate the charge density per unit
contribution remains valid for solid metals ‘‘real’’ surface area, σ , by the integration of
(see [70] and earlier references therein). It the C(E) ≡ R −1 Capp (E) curve. Determin-
gives for experimentally measurable values ing then CGC in Eq. (32) one may find the
of the capacitance: compact-layer capacitance for all electrode
charges. Finally, the ‘‘theoretical’’ capac-
(Capp )−1 = R −1 CH −1 (σ ) itance curves could be found for lower
+ R −1 CGC −1 (σ, c) (32) electrolyte concentrations. However, such
an analysis has almost never been per-
One must keep in mind that the GC expres- formed, that is, only the values of R and the
sion (21a) for the diffuse-layer capacitance, compact-layer capacitance for the p.z.c.,
CGC , contains the charge density per unit CH (0), were extracted (see Ref. [34] for a
real surface area, so that it can only be review).
calculated (except for its value for σ = 0) if An alternative method to determine
the RF is known. the RF and the compact-layer capacitance
The attempts to apply the Grahame curve from experimental data for solid
treatment in the quantitative manner en- surfaces was proposed in [69]. It is based
countered certain problems. First, the on the same model (32) of a uniformly
impedance data interpreted in terms of charged rough surface. On the basis of this
the in-series combination of the elec- equation the compact-layer capacitance
trolyte resistance and the EDL capac- curve, CH (σ ), was calculated for several
itance, C, demonstrated a significant values of the RF R, see Fig. 7, for a fixed
frequency dispersion of these elements electrolyte concentration. Similar to Fig. 3
(which implies that the equivalent circuit for the mercury electrode (in which the
200
1.1
1.2
1.3
[µF cm−2]
CH
100
10 0 −10
σ
[µC cm−2]
Fig. 7 The CH (σ ) plots for a polycrystalline silver electrode
calculated with the use of Eq. (32) for several values of the RF, R,
indicated at each plot [69].
effect originated from the experimental or (if verified) not satisfied, especially for
errors in the capacitance values), the PC electrodes.
shape of the curve within the interval Equation (32) of this model implies
near the p.z.c. was especially sensitive to that the surface roughness is realized
the choice of R. In particular, a value by the spatial configuration whose scale
of R (R = 1.2 in Fig. 7) was found for is well above the thickness of the EDL,
which the whole curve was passing this in particular the Debye screening length
interval smoothly. It was assumed that of the solution, LD (18), to allow the
this procedure allowed one to determine interfacial structure to follow the curved
both R and the CH (σ ) curve. The value surface without its especial distortion.
of R obtained in such a way was in Sometimes in earlier studies, (see [34] for
accordance with the one found from the review) the treatment of the capacitance
comparison of EDL capacitance values for data for solid metals was based on a
PC samples of this metal (silver) and different relation:
its single-crystal faces at high negative
electrode charges. (Capp )−1 = R −1 CH −1 (σapp )
One should note that one has to verify, + CGC −1 (σapp , c) (33)
whether these characteristics are the same
for different electrolyte concentrations, Physically, it may simulate the situation
like it was done in Fig. 3 for mercury. of a solid electrode with a microroughness
This requirement was mostly not verified whose scale is well below the diffuse-layer
thickness. Then the compact-layer capac- like Pb, Bi, Sb, Sn, and Cd, for which
itance per unit apparent surface area is the difference of the p.z.c. potentials be-
increased by the RF, while the diffuse layer tween the single-crystal faces of the same
is formed outside this microrough inter- metal is sufficiently small, within 100 mV
face and its structure is identical to the one or less. On the contrary, a great dispersion
for liquid metals. of p.z.c. values for different single-crystal
Model (33) predicts a unit slope for the faces (300 mV and greater) was revealed for
PZ plots. It is in contradiction with ex- Ag, Au, Cu, and so forth. Besides, for all
perimental data, at least for PC electrodes. face-centered cubic metals of this group,
However, it should be noted that recent two basic crystal faces, (111) and (100), pos-
theoretical studies of the EDL structure at sess the most positive among those p.z.c.
rough (but uniform) surfaces have shown values, while even a relatively small change
that these two models may be considered of the crystallographic orientation results
as limiting cases of a more general descrip- in a rapid shift of the p.z.c. to the most
tion, Sect. 2.1.6.4. negative values that are characteristic for
One must keep in mind that the increase most higher-order crystal planes.
of the EDL thickness upon a decrease of
This finding led to the development of
the electrolyte concentration is only valid
a new concept emphasizing the effects of
for a close vicinity of the p.z.c. The GC
crystallographic inhomogeneity of the PC
formulas for the ion charge distribution
electrodes by Frumkin [71]. It postulates
show that outside this narrow interval
(in its simplest version) that the PC sur-
of the electrode charges, the diffuse-layer
face represents a combination of various
thickness becomes independent of the
single-crystal faces – each of them having
ionic strength of the solution, see Eq. (21a)
or (21b) as an illustration. This effect their own EDL, the structure of which is
originates from the nonlinear response of identical to the one at the correspond-
the diffuse layer and makes the area of the ing real single-crystal electrode (model
applicability of Eq. (33) quite narrow. of ‘‘independent electrodes’’ [72] or ‘‘inde-
pendent diffuse layers (IDL)’’ [20]). It was
noted [71] that at a fixed value of the PC
2.1.6.2 Crystallographic Inhomogeneity
Effects for Solid Electrode Surfaces electrode potential, E, the charge densities
The analysis of models (32) and (33) of individual faces were different and that
demonstrates the necessity to be very this factor might change the shape of the
careful in comparing the values of the C(E) curve for the PC electrode near its
RF of the same surface found by different diffuse-layer minimum.
experimental techniques, since they may Later the capacitance properties of a
be crucially dependent on the scale probed PC surface were analyzed on the basis
in the corresponding method. of Eq. (34) [69, 73, 74], where the C PC (E)
Significant new information was pro- curve was represented as the sum of the
vided by capacitance measurements at capacitance curves for individual faces,
single-crystal faces of solid metals. It was C (i) (E), multiplied by their fractions of
found that there existed two groups of the total surface area, θ (i) :
solid metals having qualitatively different

properties. One of them includes numer- C PC (E) = θ (i) C (i) (E) (34)
ous ‘‘mercury-like’’ metals or semimetals, i
Though this concept looks almost self- in the vicinity of the zero PC charge,
evident, one should emphasize that it σ PC = 0, because of incorrect subtrac-
results in quite different predictions on tion of the diffuse-layer contribution.
the EDL properties, compared to the Moreover, such calculated ‘‘compact-
aforementioned model of a rough but layer curves’’ are strongly dependent on
uniform surface (32) or (33): the electrolyte concentration and have
no relation to the ones obtained by the
• the diffuse-layer minimum of the averaging of the compact-layer curves
C (i) (E) curve in dilute solutions is for individual faces with the same frac-
widened and not so deep, in compar- tions θ (i) , similar to Eq. (34).
ison with the minimums for individual
faces; These strong ‘‘anomalies’’ were predicted
• for metals with a considerable variation for metals whose PC surfaces were com-
of p.z.c. values among the faces the posed by low-index single-crystal faces (e.g.
minimum of the C (i) (E) curve, Emin PC , (111),(100) and (110) for face centered cu-
is close to the most electronegative bic (fcc) lattices) having a strong difference
among p.z.c. of the faces, for example, between their p.z.c. (several hundred mV).
Emin PC ∼ = Eσ =0 (110) for Ag, Au, Cu; For ‘‘mercury-like’’ PC metals, the pre-
• the formally constructed PZ plot, that dicted effects were less pronounced, even
is, the dependence of (Cmin PC )−1 on though the formal PZ plots were to be
CGC −1 , cannot be used to determine markedly curved.
the electrode surface roughness or the It was also recognized [74] that there
compact-layer capacitance; existed an alternative model of the PC sur-
• this plot is expected to be curved (down- face, ‘‘EDL with a common/united diffuse
ward for lower concentrations) as a layer’’. It means that if the size of the in-
result of crystallographic inhomogene- dividual faces at the surface is sufficiently
ity effects, if the concentration interval small, the diffuse layer is smooth along
is sufficiently wide; this heterogeneous surface, i.e., it is anal-
• the overall charge of the PC surface, ogous to the one for a uniform surface but
σ PC , is nonzero at Emin PC , and its value with the charge density in expression (21a)
depends on the electrolyte concentra- replaced by the average charge density of
tion; the heterogeneous surface:
• at the p.z.c. of the PC surface, σ PC = 0,

the charges of regions (faces) at the σ PC = θ (i) σ (i) (35)
surface are nonzero, σ (i) = 0, that is, i
this point corresponds simply to the
compensation of the contributions of Then the interfacial capacitance of such
opposite signs into the overall PC PC surface should be described by the
surface charge; equation:
• if the compact-layer capacitance curve
of the PC surface, CH PC (σ PC ), is cal- (C PC )−1 = (CH PC )−1 + CGC −1 (σ PC , c)
culated with the use of the traditional (36a)
scheme assuming the uniformity of the where ‘‘the compact-layer capacitance of
PC surface, Eq. (32), then this curve dis- the PC surface’’, CH PC , is given by the
plays rapid nonmonotonous variation average of the compact-layer capacitances
of the individual faces: the individual faces, θ (i) , in Eqs. (34) and
(i) (35) since they have been postulated on
CH PC = θ (i) CH (σ (i) ) (36b) an intuitive basis rather than determined
i from measurements. One should also
The latter depends on the charge density keep in mind the necessity to distinguish
at the corresponding face, σ (i) , rather than between the capacitance values related to
on the average density σ PC . Evidently this the ‘‘real’’ or ‘‘apparent’’ surface areas of
model predicted quite different effects: the PC electrode, Eq. (30). In other words,
one must take into account the RF of the
• the coincidence of the p.z.c. of the PC PC surface, R PC . Then the experimentally
surface with the potential of its C PC (E) measured capacitance value, Capp PC (E),
curve, Emin PC ; is given by Eq. (37) if the ‘‘model of
• the independence of its value on the independent electrodes’’ is used:
electrolyte concentration;
• a straight line with a unit slope for the Capp PC (E) = R PC θ (i) C (i) (E) (37)
i
PZ plots at any fixed value, and so on.
Here, similar to the previous relations, the
In an attempt to choose an adequate sum of the fractions of the individual faces
treatment, the predictions for both mod- at the PC surface, θ (i) , is equal to 1. The
els were compared with experimental data capacitance curves C (i) (E) correspond to
for PC samples of various metals. Qual- perfect faces without any roughness.
itatively the conclusions of the model of To avoid a misunderstanding, one must
‘‘independent electrodes’’ turned out to be keep in mind that all conclusions derived
in agreement with experimental data for on the basis of Eq. (34) remain valid for
such metals as Ag, Au, Cu and so on, see this case, too. In particular, it is impossible
reviews in Ref. [20, 34], contrary to the fail- to obtain the RF of the PC surface, R PC ,
ure of the ‘‘common diffuse layer (CDL)’’ from the slope of the PZ plots. On the
model, Eq. (36a). However, in quantitative other hand, one may determine both the
terms the effects of crystallographic inho- RF and the fractions of the individual
mogeneity were smaller than predicted by faces from the comparison of experimental
the model based on Eq. (34). For example, capacitance curves for the PC surface
the capacitance at the minimum, Cmin PC , and the individual faces (if the latter are
for PC Ag diminishes more rapidly than in known) [75]. These characteristics of the
its theoretical estimates. The experimental PC surface are found by minimizing the
slope of the PZ plots calculated with no difference between the left- and right-
account for the surface heterogeneity ef- hand side terms in Eq. (37) with respect to
fects is close to 1 for Bi, within a very wide the fitting parameters, R PC and θ (i) . This
range of the concentration, in contrast to operation applied to the PC Ag electrode
the model analysis. The value of R found enabled one to reproduce properly its
from the same plot is close to 1 for some experimental capacitance curve on the
PC Ag electrodes in rather concentrated basis of those for the three silver low-index
solutions. faces, (111), (100) and (110). Moreover, it
One of the possible reasons of this was found that the values of R PC and θ (i)
disparity may be an inappropriate choice of obtained from the fitting for each electrolyte
the fractions of the PC surface occupied by concentration turned out to be practically
the same for all concentrations. The value observed for single-crystal faces of many
of R PC obtained in this way was close to metals. Then the response of the EDL is
the ratio of the Capp PC and C (i) at high always nonlinear, since the electron charge
negative potentials, in accordance with density is sufficiently high at least on some
earlier expectations, see previous text. of the surface domains, since the electrode
potential E is far from their p.z.c. val-
2.1.6.3 EDL Structure for NonUniform ues, Eσ =O (i) . For these surface regions,
Electrode Surfaces the dominant contribution to the local ca-
All this modeling was based on the pacitance is given by the compact layer,
hypothesis that the EDL structure is one- even if the electrolyte concentration is very
dimensional, that is, all profiles (concen- low. For some other regions, the poten-
trations, ion charge density, and electric tial E may be close to their p.z.c., the
potential) depend on a single coordinate EDL structure may be described within
normal to the surface, either within the the linear-response model and their overall
whole diffuse layer in model (36a) or capacitance may be essentially determined
within the EDL at each single-crystal face by the diffuse-layer contribution. However,
in model (34) or (37). However, a rig- the latter leads to a strong reduction of
orous theoretical analysis [76, 77] shows the contributions of such regions into the
that these distributions are essentially per- overall capacitance of the heterogeneous
turbed near all contacts of the surface surface, since the contributions of the re-
domains possessing different p.z.c. or/and gions are additive.
compact-layer capacitance values. It results As a result the criterions of models
in a nonuniform distribution of the surface (37) or (36a) depend mostly on the
charge density. The width of this perturbed difference of the partial p.z.c. values and
zone is determined by the screening prop- on the partial compact-layer capacitances.
erties of the electrolyte inside the diffuse In particular, for the most interesting case
layer, in particular it is close to the De- of a strongly heterogeneous surface (i.e.
bye screening length in the bulk solution, having a significant difference between
LD , in the simplest approximation of the the partial p.z.c. values) the principal
linear response. Then the criterion of the condition relates the characteristic size of
aforementioned models is in the compar- the uniform regions, y∗ , and the lengths
ison between the characteristic size of the (i)
LH derived from the values of the partial
individual uniform areas at the PC sur- (i)
compact-layer capacitances, CH , e.g., for
face (e.g. single-crystal faces), y∗ , and the
the ‘‘model of independent electrodes’’
Debye screening length, LD . For example,
(Eqs. (34) and (37)):
for model (34) and (37) it has the form
y∗ L D .
y∗ max{εLH (1) , εLH (2) , . . .},
This approximation of the linear re-
sponse may be justified for heterogeneous LH (i) ≡ (4πCH (i) )−1 (38a)
surfaces with a small difference of the par-
tial p.z.c. values, in particular for some One should emphasize that this condition
mercury-like solid metals. However, the is sufficient even for very dilute solutions,
analysis is much more complicated [72, in which the Debye screening length
78] in the case of a surface with a great dif- in the bulk, LD , may strongly exceed
ference of these p.z.c. values, for example, the geometrical size, y∗ , contrary to
the intuitive expectations based on the structure. The corresponding model was
linear-response approximation. first proposed for the imperfect single-
The criterion is slightly more compli- crystal (111) face of Au in Ref [79], which
cated for systems with a moderate (but was considered as a combination of per-
not small) difference of the partial p.z.c. fect (111) regions and intermediate zones.
values for which, the diffuse-layer min- Then in accordance with the approach
ima for the different uniform regions as of ‘‘independent electrodes’’, the over-
a function of the electrode potential overall capacitance was represented as the
lap partially. Then, condition (38a) must be sum of the perfect C 111 (E) curve and
supplemented by another one; the geomet- the one of the PC Au electrode. Possi-
rical size should also significantly exceed ble influence of various higher-order faces
the shortest among the partial screening and surface defects on the capacitance
lengths inside the diffuse layers at each properties of PC surfaces was discussed
region: in [80]. Later, it was suggested to repre-
sent the overall PC capacitance as the
y∗ max{LD (1) , LD (2) , . . .}, sum of the contributions from all low-
LD (i) ≡ (4πεCGC (i) )−1 (38b) index faces (identical to Eq. (37)) and from
the intermediate regions, the latter be-
determined by the diffuse-layer capac- ing modeled in accordance with Eq. (36).
itances for the corresponding regions, This treatment was extensively used to
CGC (i) . The latter depends on the local analyze the properties of Bi electrodes,
charge density for this region according to see [20].
Eq. (21a). For metals with a great difference be-
The characteristic lengths εLH (i) vary tween the p.z.c. values of the faces, it is
widely as a function of the metal and necessary to keep in mind the specific
solvent properties. For aqueous solutions, dependence of the p.z.c. values of the
they are within 5 nm, or even smaller for single-crystal faces as a function of their
metals with high values of the compact- orientation; for example, for various fcc
layer capacitance for single-crystal faces at metals the low-index faces, (111) and (100),
small electrode charges, such as Ag or Au. possess ‘‘singular’’ p.z.c. properties, while
Therefore this analysis testifies in favor this characteristic rapidly approaches a
of the ‘‘model of independent electrodes’’ much more positive value typical for (110)
for such systems, in conformity with and higher-index faces, at a relatively small
the experimental evidences mentioned change of the crystallographic orientation.
above. Therefore, one cannot exclude an alterna-
However, one should keep in mind that tive interpretation of the above-mentioned
all this analysis is based on the assumption experimental observations for such PC
that the principal contribution to the PC ca- electrodes. Namely, one can envisage that
pacitance is given by perfect single-crystal their surfaces do not contain perfect (111)
faces at its surface, which must have a suf- or (100) faces but only perturbed ones.
ficiently large size, Eq. (38a). One cannot Their p.z.c. values are not much differ-
a priori exclude a considerable contribu- ent from those for the (110) or (210)
tion due to intermediate regions between faces.
well-defined uniform surface fragments, The possibility of getting a substantiated
or even a completely amorphous surface answer might lie in the use of modern
methods of analysis to establish the PC the linearized Poisson–Boltzmann ap-

surface structure more reliably [21, 22]. proximation, was considered in Ref. [83];
On the other hand, one must always keep the solution obtained helps in explain-
in mind that each particular experimental ing a number of experimentally observed
technique often uses different definitions of potential-dependent effects. The latter is,
the term ‘‘surface structure’’, in particular however, more cumbersome for presen-
its ‘‘roughness’’. tation, and we will explain the idea of
the theory for the case of small electrode
2.1.6.4 Surface Roughness Effect on the polarizations [82].
Diffuse-Layer Capacitance As it was shown in Ref. [81] within the
Characterizing roughness by a geometrical linearized Poisson–Boltzmann approxi-
RF, as the ratio of the true surface area and mation the diffuse double-layer capaci-
the apparent cross-section area, Eq. (29), tance takes the form,
one distinguishes typically the cases of
weak, moderate, and strong roughness, de-
C = R̃(κ)CGC (39)
pending on whether R − 1 << 1, R − 1 ≈
1, R − 1 >> 1. Weak roughness is typical
for single-crystalline electrodes, moderate where R̃(κ) is not a roughness factor but a
roughness for polycrystalline electrodes, ‘‘roughness function’’ of the inverse Debye
while the case of strong roughness is met length, κ. This function varies between
for specially fabricated catalysts. However, the obvious limits R̃(0) = 1 and the RF
as we have mentioned, the RF alone can de- R = R̃(∞) > 1. For the case of weak
scribe the effect of surface corrugation only roughness, a relationship was derived
if the characteristic scales of roughness are between the roughness function and the
much greater than the diffuse-layer thick- height-height correlation function of the
ness, while in the opposite limiting case its surface corrugation,
effect will not be seen. In the general case,
one must develop a more general theory.
κh2
Such a theory was developed re- R̃(κ) = 1 + dKg(K)
(2π)2
cently [81–83], however, with no account
for possible ‘‘energetic inhomogeneity’’ of × [ (κ 2 + K 2 ) − κ] (40)
the surface, as discussed in the preced-
ing text. Since, facets of a rough surface Here the ‘‘height’’ h stands for the mean
always expose different elementary faces square departure from flatness and g(K)
of the crystal structure, such a theory is the height–height correlation function
has a limited applicability; one might ex-
pect it to be valid only for metals with a
|ξK |2
minor difference in p.z.c. Since this the- g(K) = (41)
ory is discussed also in Chapters 2.2 and Sapp h2
2.4 of this volume, we only briefly out-
line its basic ideas. We will do it for the where Sapp is the apparent surface
case of relatively small electrode charges, area, ξ(R) the corrugation amplitude
when the linearized Poisson–Boltzmann (< ξ(R) >= 0) as afunction of the lateral
theory could be used [81]. The case of arbi- coordinate R, ξK = dRξ(R) exp{−iKR} is
trary large electrode polarizations, beyond its Fourier transform (FT).
For Gaussian height–height correla- roughness approximation, and an exact

tions, g(K) = πl 2 exp(−l 2 K 2 /4), one ob- numerical solution, obtained for any ratio
tains between the height and the period of grat-
√ 2 2 ing. The comparison of the two has shown
2 πh2 κ κ l
R̃(κ) = 1 + exp that the analytically tractable approxima-
l 4 tion of weak roughness gives practically

κl a universal description of the possible
× 1− (42)
2 roughness effect on the diffuse-layer ca-
pacitance for nonfractal electrode surfaces.
and The situation is more complicated for
2h2
R =1+ (43) a self-affine fractal surface. The latter is
l2 characterized by the self-affine exponent,
where (x) is the probability function [84]. H , which is related to the Haussdorf
For this case, R̃(κ) is plotted in Fig. 8. fractal dimension D = 3 − H [85], and a
As it was shown in Refs. [81, 82], the long-length cutoff, l, which characterizes
particular form of the corrugation has only the scales after which the surface is
a small effect on the roughness function, apprehended as a Euclidian one. As it
which is constrained between 1 and R. was argued in Ref. [81], the roughness
In order to verify how critical the as- function for the Debye length shorter than
sumption of weak corrugation for this the long-length cutoff scales as R̃(κ) − 1 ∝
approach is to the calculation of the rough- (h/ l)2 (lκ)2(1−H ) , at least for moderate
ness function, the case of corrugation fractal dimensions, that correspond to
described by a model of rectangular grat- H < 0.5. It could have been interesting
ing was studied [82]. For this case, there to verify this law.
is an exact analytical expression for the However, it is more important to verify
roughness function, obtained in the weak first the basic predictions of the theory
1.5
1
1.4
Roughness function
1.3
[R]
~
2
1.2
3
1.1
1
0 0.2 0.4 0.6 0.8 1
k
[nm−1]
Fig. 8Roughness function dependence on inverse Debye length for the
random Gaussian surface. ε = 80, h = 5 nm, l = (1)10, (2)15, (3)20 nm.
for nonfractal electrodes. Efforts in this describe the main points of this theory [43,
direction have been made by the Tartu 44, 90].
group [86] that have extracted R̃(κ) =
C/CGC from their data and plotted it 2.1.7.1 Basic Equations
versus κ. It was found that for metals At a given surface charge density, σ ,
with close values of the p.z.c. for different of the electrode, the potential distribu-
crystal faces the trends of the theory are tion along the z-normal to the elec-
basically correct, but even for such metals trode surface, φσ (z), obeys the Pois-
the deviations from the nonequipotential son equation d2 φσ (z)/dz2 = −4π(ρms σ +
character of the outer Helmholtz surface σ σ
ρion ), where ρms is the bound charge
(assumed to be equipotential in Ref. [81]) density of the metal and solvent in con-
are likely to be important for a quantitative tact and ρionσ is the charge density of
comparison. For metals with strong vari- electrolyte ions. Subtracting from this
ance of the p.z.c., the theory in its present equation the same equation at σ = 0, for
form essentially fails. It needs an exten- the so-called rationalized potential, φ(z) =
sion that would combine this theory with φσ (z) − φ0 (z), we get
the models of energetic inhomogeneity de-
scribed in the previous section. Note that d2 φ(z) 0
Lust and coworkers compared the nonlin- = −4π(ρms + ρion
σ
− ρion ) (44)
dz2
ear variant of the theory also [83], that is,
the so-called potential-dependent roughness where ρms = ρms σ − ρ 0 . To close this
ms
function, with the values extracted from equation let us introduce the nonlocal
their data [87], and came essentially to permeability of the metal/solvent system,
the same conclusions. Further system- χ(z, z ), which relates the change in the
atic data would be highly welcome, as bound charge density to the corresponding
well as the mentioned extension of the change of the electric field
theory. ∞
d dφ(z )
ρms = dz χ(z, z ) (45)
dz −∞ dz
2.1.7
Semi-phenomenological Nonlocal Theory Then the equation for φ reads:
of the Double Layer
∞
d dφ(z )
dz ε(z, z )
The dielectric response of the interface dz −∞ dz
can be described in a unified manner in 0
= −4π[ρion
σ
− ρion ] (46)
terms of the nonlocal electrostatic the-
ory [88, 89]. Indeed, it was shown to be where
possible to express the electric proper-
ties of the interface through the dielectric ε(z, z ) = δ(z, z ) + 4πχ(z, z ) (47)
function of the metal/solvent system, not
applying a particular form of this func- is the static nonlocal dielectric function of
tion, for any structure of the interface. the metal/solvent system that keeps all the
Such an approach allows revealing gen- information about its structure [88–91].
eral properties of the double layer and Equation (46) is still not closed, as we
expressing the parameters involved via still need an equation that relates the r.h.s.
the nonlocal dielectric function. We briefly with φ. A particular closure is determined
by a specific statistical-mechanical scheme the nonlocal dielectric function. Referring

for the ionic subsystem, the classical Pois- the reader to Ref. [44], we show here only
son–Boltzmann scheme approximation the result for the capacitance in the limit
(which results in the Gouy–Chapman– of low electrolyte concentrations:
Stern (GCS) model) being the simplest
1 1
one. Here, we will show the equation that = + LH + O(κ) (50)
follows in the case of the linearized Pois- 4πC εκ
son–Boltzmann approximation, valid for Here,
σ → 0 and moderately large concentra- ∞ ∞
tions c: LH = dz dz ε−1 (z, z )
∞ −∞ −∞
d dφ(z )
dz ε(z, z ) = κ02 θ(z − l) θ(z − l)
dz −∞ dz − (51)
(48) ε
√
Here, κ0 = κ ε where ε is the dielectric
where the inverse operator ε−1 (z, z ) is
constant of the bulk solvent and κ −1 is the
determined by
Debye length; θ stands for the Heaviside
∞
unit step function, which reflects here the
dz ε−1 (z, z )ε(z, z ) = δ(z, z ) (52)
distance of closest approach, l, of ions to −∞
the electrode (l is calculated from the edge
of the metal skeleton). To the accuracy of small O(κ)-terms
Because of the general properties of the Eq. (50) coincides with the Grahame
kernel ε(z, z ), Eq. (48) possesses a unique ansatz at σ = 0. The Grahame ansatz is
solution. For the solution obtained, the thus recovered in the low concentration
capacity of the metal electrolyte interface limit, giving for the ‘‘compact-layer capac-
is calculated via the formula, ity’’ the expression
1
1 φ(−∞) − φ(∞) CH = (53)
= lim 4πLH
4πC σ →0 σ
φ(−∞) − φ(∞) Hence, we see that CH is determined
= ∞ (49)
2
primarily by the quantum-statistical prop-
κ0 dzφ(z) erties of the metal-solvent systems rather
l
than by the distance of solvent approach.
Generally in order to find the solution, The primitive local estimate gives an exam-
one must specify the particular form ple when l is, indeed, of minor importance.
of ε(z, z ). Such solutions have been Take a model of a film with low dielectric
repeatedly reported in the literature (see constant that separates an ideal metal and
e.g. [44, 92–96]). However, in the spirit of the bulk solvent:
the semi-phenomenological approach, we ∞ z < 0
are now interested in the solution for a
ε(z, z ) = δ(z, z ) ε∗ 0 > z > z∗
general form of ε(z, z ). Such a solution
ε z > z∗
can be obtained approximately [44] using
the technique of matching asymptotic If l > z∗ , this model gives LH = z∗ (1/ε∗ −
expansions [97], in the case of a dilute 1/ε) + l/ε, and for ε ε∗ and z∗ ∼ l,
electrolyte when the Debye length is much LH ≈ z∗ /ε∗ , that is, it is determined by the
greater than all the characteristic lengths of parameters of dielectric inhomogeneity
but not by the distance of closest approach dipoles in the layer bounding the metal,
of electrolyte ions. ‘‘spin–up’’ (↑) and ‘‘spin–down’’ (↓),
An extension of the nonlocal dielectric with a dipole pointing away or to the
theory to the range of moderate deviations metal, respectively. Some residual energy
from the p.z.c. was reported in Refs. [43, is attributed to each state in the field
90]. The results differ in replacing 1/εκ by of an uncharged electrode, U↑ and U↓ .
1/4πCd , where Cd is the nonlinear differ- Boltzmann statistics relates then the oc-
ential capacitance of the diffuse layer (e.g. cupation numbers of the states, N↑ and
of GC type), and ε(z, z ) by the so-called N↓ : kB T ln(N↑ /N↓ ) = U↓ − U↑ + 2p/d
differential dielectric function ε(z, z , σ ) (for where p is a permanent dipole molecule
details see Ref. [90]). Somehow the Gra- of the molecule, is the potential drop
hame parameterization is again recovered, across the layer of thickness d, and
but the validity criteria now involve the kB is Boltzmann’s constant. Here the
value of σ . Critical is the requirement that dipole moments are assumed to have the
the amount of the electrolyte charge, ac- same magnitude for both orientations,
cumulated within the layer of thickness but the dipole moments are, in general,
z∗ , constitutes a small portion of the total
different from those in the gas phase
charge σ . For aqueous electrolytes, the cri-
or in the bulk of the solvent. The or-
terion limits the results to σ < 10 µC/cm2
der parameter then reads (N↑ − N↓ )/N =
and c < 0.1 M. This is the sufficient crite-
tanh p/kB T d, where = − m ,
rion. In fact the differences between the
m = d/2p(U↑ − U↓ ), and N = N↑ + N↓
values of CH as evaluated from small and
is the total number of molecules in the
high concentrations typically do not exceed
layer per unit surface area. This is not yet
5%. The reason for that phenomenon is
discussed in Ref. [88]. In general Grahame a closed equation; the closure is reached
parameterization may eventually work at by the assumption that is a superposi-
higher charges and concentrations. tion of the potential drop in the absence
of dipoles and the contribution due to
2.1.8
dipoles:
‘‘Dipolar’’ Models of the Compact Layer
1 p
= 4πσ d − 4πpN tanh
Historically, molecular models of the ε∞ kB T d
compact layer were first to rationalize (54)
its nonlinear dielectric response. Starting where σ is the specific charge of the
with Ref. [98] and until the beginning of electrode and ε∞ is the high frequency
the 1980s, this type of models comprised effective dielectric constant, related to
the main direction of research in the the intramolecular polarizability of the
theory of the electrochemical interface (for molecules. The p.z.c., obtained from
detailed review see Refs. [88, 99]). We will Eq. (54) at σ = 0, lies generally between
discuss several of the most important ones 0 and m .
in the following text. Equation (54) implicitly determines the
value of potential for a given charge of
2.1.8.1 Two-State Model the electrode. Its differentiation over σ
This model [98] assumes the existence gives the famous bell-shaped differential
of two admissible orientations of solvent capacitance of the molecular layer with a
maximum centered at m . 2.1.8.2 Multistate Models

A three-state model was suggested in
ε∞ Np 2 1
CM = + Ref. [103] with a third state (→) cor-
4πd kB T d 2 2 p responding to orientation of a dipole
cosh
kB T d parallel to the electrode surface. It was
(55) analyzed on a level similar to Ref. [101]
The parametric dependence of CM on σ and successfully applied for interpreta-
is given by Eq. (55) combined with σ () tion of the capacitance data for aprotic
dependence given by Eq. (54). According solvents.
to Eq. (54) the maximum of capacitance is A cluster model, essentially a four-state
at σ = ε∞ m /4πd. Electrochemists used model, was suggested [104] and devel-
to believe that, if there were a prefer- oped [105, 106] to interpret the capacitance
ential orientation of water molecules on data in associated solvents. According to
the metal surface, the molecule would this model the molecules in the bound-
look at the metal by its negative end. ing layer can exist in free states or in
That means U↓ > U↑ . Hence, m < 0. clusters. Up-and-down orientations are al-
Since the maximum of the compact- lowed for a free molecule or a cluster.
layer capacitance is in the anodic but Owing to constraints on the mutual ori-
not in the cathodic range, historically entations of molecules in a cluster, the
this model was not accepted in electro- effective dipole moment of a cluster is
chemistry. Nowadays we are less certain assumed to be smaller than that of a
in that (see discussion in Sect. 2.1.10). free molecule. The residual energy for a
Furthermore, since we now know that molecule in a cluster state is assumed to
this is only one of the contributions be lower than in a free state. In such
to the inverse capacitance, this limita- a model there will be more clustered
tion may not be that critical. Indeed for than free molecules at the p.z.c. However,
small |m | the maximum of CH may charging the electrode shifts the equilib-
still be shifted to the anodic range by rium towards the occupation of free states
the basic slope of the σ -dependence of due to the gain in polarization energy,
CH (see Sect. 2.1.9.2). Somehow the two- that is, it causes the field-induced de-
state model [98, 100] has resulted in the struction of the clusters. Since the decay
understanding that humps on the capaci- of clusters increases the net polarization
tance curve may be associated with dipolar of the layer, the rising branches of the
reorientations. The two-state model ratio- CM (σ ) curve emerge. Because of the in-
nalized the value of the effective dielectric terplay between the polarization of single
constant of the Helmholtz layer, ε∗ (σ ) = molecules and clusters, and the cluster
ε∞ + 4πNp 2 /kB T d[cosh{p/kB T d}]2 . decay and formation, the CM (σ ) curve
Further modifications of the model [101, may generally have a sophisticated form.
102] were aimed at shifting the maximum Qualitative features of Grahame’s curve
towards the anodic range at the cost of for the compact-layer capacitance of the
an artificial assumption that the dipole Hg–H2 O-NaF interface and its tempera-
moments in the up and down orientations ture dependence were reproduced in the
are different (for a critical discussion of elegant parameterization of Ref. [105] (dis-
other modifications made in Refs. [101, cussed also in Refs. [88, 99]) with three
102] see Ref. [88]). fitting parameters and four parameters,
which can be evaluated from independent surfaces [109]. First calculations were per-
data. A better fit was achieved in a more formed in the so-called jellium model [110],
involved analysis of Ref. [106], but as noted in which the skeleton is treated as a ho-
in Ref. [99] the best fit would have been mogeneous semi-infinite background of
achieved for a dipole moment of water positive charge. Later on the calculations
twice as large as its value in the gas phase, were modified to account for the discrete
which may be a bit too much. Furthermore, structure of the skeleton with the use of
the sign of the temperature derivative [105, ionic pseudopotentials [111].
106] is opposite to the experimentally ob- These predictions have been verified by
served one. experiments on low-energy helium atom
The lesson from this four-state model scattering by metal surfaces [112]. The
is that the interplay between the field- helium atom repels the Bloch electrons,
induced destruction of the clusters and because their wave function must become
reorientations of the clusters and sin- distorted to preserve orthogonality to the
wave functions of the closed shells of
gle molecules may be a source of a
helium. The repulsion potential appears
most peculiar CM (σ )-behavior. The rich-
in practise to be proportional to the local
est curve exhibits four maxima and three
free electron density. Thus a helium atom
minima.
scatters like a ping-pong ball from the
2.1.9
electron cloud of the metal, and this allows
The Role of Metal Electrons to probe the distribution of electrons in the
cloud, that is, the profile of the electronic
We now briefly describe the effects asso- tail and its lateral corrugation that usually
ciated with the surface electronic profile follows the periodicity of the surface crystal
plane [112].
of the metal and its influence on the re-
How does the surface electronic pro-
sponse of the electrochemical interface to
file deform with charging? Although there
charging. For a detailed review, the reader
was no direct observation of the charge-
is addressed to Ref. [107].
induced profile modulation, a lot is known
from the theory and indirect manifes-
2.1.9.1 Surface Electronic Profile and its tation of this effect in field emission.
Response to Charging For the jellium model, after the pioneer-
How close do solvent molecules come to ing papers [110, 113, 114], the problem
the metal? Where is the boundary of the was studied in detail in Ref. [115] within
metal? To answer these questions, we must the trial function version of the den-
go back to the structure of the metal sur- sity functional formalism [116] and in
face. The existence of a tail of quasi-free Ref. [117, 118] within the more general
electrons spilled out of the ionic skele- Kohn–Sham scheme [119]. We summa-
ton of the metal was predicted in the rize here the properties necessary for our
early days of quantum mechanics [108]. further discussion.
In the 1970s, this picture was approved If no discreteness of lattice charge is
and detailed by the electron-density func- taken into account or if we speak about the
tional theory of the inhomogeneous elec- laterally averaged electron density profile,
tron gas (later leading to Nobel fame for n(z), neglecting Friedel oscillations [110],
Walter Kohn in 1998) applied to metal it can be roughly approximated by a trial
function of its center of mass,

nσ (z) 1 ∞
zσ = − dz z[nσ (z) − n0 (z)] (58)
1 − 0.5 exp[β(σ )(z − z̄)], z < z̄ σ −∞
= n+
0.5 exp[β(σ )(z̄ − z)], z > z̄ which for the approximation (56) takes
σ the form
z̄ = − (56)
n+
z̄ n+ 1 1
zσ = + −
where n+ is the volume density of the 2 σ β(0)2 β(σ )2
positive charge of the ionic skeleton that oc-
cupies the half-space z < 0, σ is the net sur- ≈ z0 + pσ + rσ 2 (59)
0
face charge density ( −∞ [nσ (z) − n+ ] +
∞ again with definite expressions for z0 , p,
0 nσ (z) = −σ ). According to Ref. [115], r [115]. Thus-calculated zσ plots appear
the expansion to be close to direct Kohn–Sham calcu-
β(σ ) ≈ β(0) + ãσ + b̃σ 2 + d̃σ 3 (57) lations [117, 118].
n(z) obeys an exact sum rule with no
with the obtained analytical expressions for dependence on the particular approxima-
the coefficients β(0), ã, b̃, d̃ as functions tion for the electron density functional.
of n+ reproduces well the direct numer- For example, the Theopilou sum rule
ical simulation of β(σ ). Figure 9 shows derived for the metal vacuum interface
0
the thus calculated profile of nσ (z) for reads −∞ dz z[nσ (z) − n0 (z)] = σ 2 /2n+ .
n+ corresponding to Hg. The correspond- The substitution of a single parameter trial
ing excess charge distribution nσ (z) = function (56) gives just an equation on β
nσ (z) − n0 (z) is shown in Fig. 10. as a function of σ and n+ . This gives a
The important characteristic of the qualitatively similar charge dependence,
excess charge distribution is the position but quantitatively it is different from the
0.014
0.012
0.010
0.008
[a.u.]
n(z)
0.006
σ = −0.005 a.u.
0.004 (= −28.5 µC cm−2)
0.002 σ=0
0.000
−4 −2 0 2 4
z
[a.u.]
Fig. 9 Surface electronic profile of a neutral and negatively charged jellium
calculated for the bulk electron density of Hg.
σ = − 0.005 a.u.
(= −28.5 µC cm−2)
0.002
∆ n(z)
[a.u.]
0.000
−4 −2 0 2 4
z
[a.u.]
Fig. 10 Distribution of excess negative charge corresponding to the profile
shown in Fig. 9.
4.0
3.5 Hg
3.0
[a.u.]
2.5 Ga
zσ
2.0
1.5
1.0
− 0.004 − 0.002 0.000 0.002 0.004

−σ
[a.u.]
Fig. 11 The center of mass of excess charge distribution as a function of the
net charge on jellium metals. At zero charge, it is located outside the jellium;
with negative charging, it moves further out. 1 a.u. of surface
charge = 5.710−3 µC cm−2 .
density functional results, because neither and Ga are shown in Figs. 11 and 12,
is the trial function exact nor the den- respectively. For all the metals studied,
sity functional. The dramatic history of the z0 > 0, ã > 0, p < 0. This means that
employment of exact sum rules in electro- for negative charges the electronic profile
chemistry is discussed in Ref. [107]. broadens and the excess charge center-
The zσ − and β(σ )-curves for bulk of-mass moves further outside the metal.
electron densities corresponding to Hg These two features are also essential
1.5
Hg
[a.u.]
1/β
1.0
Ga
0.5
−0.004 − 0.002 0.000 0.002 0.004
−σ
[a.u.]
Fig. 12 The same graphs as in Fig. 11, but for the characteristic length of
smearing of the surface electronic profiles. Smearing increases with negative
charging.
for understanding the metal response to The profile of an ideally smooth interface
charging in situ, although the bounding is sketched in Fig. 13. The half-space z < 0
medium may affect the values of β(σ ) is occupied by the ionic skeleton of the
and zσ . However, the qualitative behavior metal. This can be described, roughly, in a
of these functions remains the same, jellium model, as a continuum of positive
unless the solvent molecules chemisorb charge n+ and the effective dielectric
on the metal surface. In that case with constant εb due to the polarizability of
negative charging of the metal the extra the bound electrons (this quantity is,
electrons may have to be localized within with rare exceptions (Hg: εb = 2, Ag:
the skeleton half-space, thereby decreasing εb = 3.5), typically close to 1 [125]). The
zσ and β(σ )−1 . gap 0 < z < a accounts for a nonzero
distance of the closest approach of solvent
2.1.9.2 Response to Charging of the molecules to the skeleton. The region of
Electrochemical Interface and the a < z < a + d stands for the first layer
Compact-Layer Capacitance: Main of solvent molecules, while z > a + d is
Qualitative Effects the diffuse-layer region. n(z) denotes the
The reader interested in the history of profile of the density of free electrons. This
the step-by-step comprehension of the is, of course, an extremely crude picture,
role of metal electrons in the double-layer but it eventually helps to rationalize the
theory and various ideas in this field is results of the various theoretical models
referred to the reviews [107, 120–124]. In and simulations.
the following sections, we will quote only The integral KH or differential CH ca-
the references needed for understanding pacitances of the compact layer, related to
the simple but basic qualitative features each other as 1/CH = (d/dσ )[σ/(KH (σ ))],
discussed in this tutorial overview. are extractable from the overall capacitance
n (z)
n+
εεb ε=1 εε ((σ)

σ)
z
0 a a+d
Fig. 13 A cartoon of a profile of a smooth electrochemical interface. The
half-space z < 0 is occupied by the metal ionic skeleton that, within the
jellium model, is described as a continuum of positive charge density
(n+) and the dielectric constant due to bound electrons (εb ), the value of
which lies typically between 1 and 2. The gap accounts for a finite distance
of closest approach of solvent molecules to the skeleton; the gap is
determined by the balance of forces that attract the molecules to the
metal and the Pauli repulsion of the closed shells of the molecules from
the free electron cloud of the metal of density n(z). The regions
a < z < a + d and z > a + d correspond, respectively, to the first layer of
solvent molecules (which can be roughly characterized by
charge-dependent effective dielectric constant) and the diffuse-layer part.
of the interface via for example, the PZ Here, we just refer to the behavior that
plots [37] or other methods discussed in explains the systematic trend for all s,p-
this chapter. To the accuracy of exponen- metals [126] – the positive slope of the
tially small terms, they can be written as: capacitance near the p.z.c., shown Fig. 14.
Such a behavior was first found in
1 d
a − zσ + (60) Refs. [127, 128] by density functional
4πKH ε(σ ) simulation. Here, for clarity, the degrees
1 da d of freedom of the molecular reorientations
a+σ − z∗ + in the first layer were frozen, that is,
4πCH dσ ε∗ (σ )
ε(σ ) = const. The peculiar a(σ ) behavior
d(σ zσ ) 1 d σ
z∗ = , = is determined by two main effects. The
dσ ε∗ (σ ) dσ ε(σ ) first effect is asymmetric with respect to
(61) p.z.c. That is, the position of the solvent
Here the d/ε(σ )-contribution is due to molecules that are ‘‘sitting on the tail’’
the potential drop across the first layer of free electrons follows the deformation
of solvent molecules, where the effective of the tail. With negative charging the
dielectric constant ε(σ ) takes into account profile swells out of the metal skeleton
the nonlinear dielectric polarizability of and a increases; with positive charging the
the layer. Thus in order to understand profile moves back towards the metal and a
the σ -dependence of the compact-layer decreases. The second effect is symmetric
capacitance, it is sufficient to understand with respect to the p.z.c. Any capacitor
the σ -dependence of a − zσ and d/ε(σ ). with a flexible distance between the plates
Different microscopic models of the will contract upon charging regardless of
interface may be classified with regards the sign of the charge. The combination
to these dependencies (see Ref. [107]). of these two effects results in the negative
KH
zσ
σs 0 σs 0
−σ −σ
Fig. 14 A sketch of the charge dependence of the distance of closest approach
of solvent molecules to the skeleton, a, the center of mass of the excess charge
distribution, zσ , and the corresponding plots for the integral capacitance of the
compact layer, KH . In the case of ‘‘soft landing’’ of the molecule onto the metal
skeleton edge (dashed curves), the spike in the capacitance transforms into a
hump.
slope of a(σ ) near the p.z.c. (which is were attributed to the nonlinear response
typically steeper than the slope of zσ ) and of molecular reorientations that were so
the maximum in the cathodic range. At far frozen in order to make clear the
some positive charge, denoted here as σs , mere effect of a(σ ) and zσ . In terms
a will reach zero and will stop changing of Eq. (60), that would mean that ε(σ ) =
with further increase of σ , because the const. For the simplest two-state molecular
molecules will encounter the edge of models, it has a bell-shaped form, the
the metal skeleton. The stop would not maximum of which is expected in the
be abrupt (solid curve), but soft (dashed cathodic range. The hump is, however,
curve) if the skeleton were not a hard observed in the anodic range. Since
wall for solvent molecules. The simplest the (a(σ ) − zσ )-determined ‘‘background’’
approximation for a(σ ) is curve is properly asymmetric, a bell of ε(σ )
a(σ ) = [a(0) − Aσ − Bσ 2 ], centered near the p.z.c. will be seen as a
hump in the anodic range.
A > 0, B > 0 (62) Is the smeared ‘‘stop-spike’’ for KH
(Fig. 14) what we see for a hump on
The difference between a(σ ) and zσ
(Fig. 14) leads to KH (σ ) plots that are the measured capacitance curve, or is it
in line with the data for simple met- screened by molecular reorientations? We
als (Figs. 15 and 16), (for review and don’t know that. Moreover, the specific
references, see Ref. [88]) in contact with adsorption of ions, which was assumed
surface-inactive electrolytes; the slope at absent, can give rise to a hump too.
the p.z.c. is positive, the minimum lies in Somehow this feature is interesting be-
the cathodic range and a kind of a spike, cause of its quantum-mechanical origin
which might be apprehended as a hump, and direct dependence on the nature of the
takes place in the anodic range. metal. The more densely packed the metal
Of course there could be other reasons surface, the more abrupt the stop, the
for the hump. In the molecular models sharper the spike. Open surfaces should
of the compact layer (see Sect. 2.1.8), they favor ‘‘soft landing’’ and the hump would
Ga
120
100
80
[µF cm−2]
CH
60
InGa
40 Bi
Hg
20
12 8 4 0 −4 −8 −12
σ
[µC cm−2]
Fig. 15 Metal effect on the compact-layer differential capacity: the
series Hg, Bi, InGa alloy, Ga. Data from Refs. [129–132].
130
Ag
100
[µF cm−2]
CH
70
Cd
40 Hg
40 20 0 − 20
σ
[µC cm−2]
Fig. 16 Metal effect on the compact-layer differential capacity: the
series Hg, Cd, polycrystalline Ag. Data from Refs. [35, 133, 134].
be smoother. For single-crystalline sur- metal, the denser the electronic tail, and
faces, the spikes should be sharper than the stronger the repulsion of the molecules
for liquid metals, in which the interface from it. Higher positive charges will be
undergoes fluctuations, and for uneven needed to press the solvent to the skele-
polycrystalline surfaces, in which inhomo- ton and σs will move farther to the anodic
geneous broadening may take place. The range. This is what we see, if we compare
higher the bulk electron density of the Ga and Hg.
In contrast to what was suggested in approximate the response function, spec-

an early ‘‘swing’’ and’’ slide’’ model [135], ify the irreducible integrals, and rationalize
there is no simple relation between the their dependence on the charge of the
slope of KH (σ ) and n+ (for details see metal and frequency (for review see [107,
Refs. [107, 126]). 137–140]).
Models of this ‘‘flexible capacitor’’ came At planar interfaces there are four main
across negative values of the calculated types of linear optical signals that are de-
CH (σ ). The reasons for that exotic effect tected by different techniques. Three of
(not necessarily an artifact!) have been them are related to reflection and one
discussed in detail in reviews [107, 121, to attenuated total reflection. In reflection
136]. methods the basic measurable parameters
are related to rp and rs – the complex am-
2.1.9.3 Optical Response and the plitudes of the reflection coefficients of the
Interfacial Structure light polarized parallel (p) and perpendic-
In the early 1980s, phenomenological ular (s) to the incidence plane, respectively
electrodynamics of smooth interfaces (Fig. 17). These are reflection coefficient
was developed with a capability to pa- Rs(p) = |rs(p) |2 , phase of reflected light
rameterize the main types of electro- δs(p) = (1/2i) ln{rs(p) /rs(p) }, ellipsometric
modulated optical signals. The signals parameters ψ = tan−1 |rp /rs | and =
were expressed through ‘‘irreducible in- δp − δs . In attenuated total reflection, by
tegrals’’ of the nonlocal response func- measuring the frequency position of the
tion, which characterizes the interface dip in reflection coefficient as a function
in the optical frequency range. This of the angle of incidence, one detects the
had opened opportunities for model dependence of the surface plasmon fre-
microscopic theories that could calculate or quency on the wave-vector, ω̄(K ).
θ
Metal Electrolyte
z
O
Fig. 17 The geometry of electromagnetic wave reflection

experiments: (z, x) is the incidence plane, and θ is the angle of
incidence.
The modulation technique is based where C is the capacitance of the

on the application of a potential E = metal–electrolyte interface, ε∞ is the bulk
E0 + δE sin(t), where δE → 0 and optical dielectric constant of the solution,
is smaller than all the characteris- εM (ω) is the bulk optical dielectric constant
tic frequencies at this interface. Mea- of the metal, and εb (ω) − 1 is the high fre-
suring the -harmonics of the sig- quency contribution to it, which comes
nals, for example, rs(p) = rs(p) |E=E0 + from bound electrons of the skeleton.
(drs(p) /dE|E=E0 ) δE sin(t), gives the When (da/dσ ) = 0, this formula reduces
precise value of the derivative over po- to the well-known Aspnes and McIntyre re-
tential (if there is no singularity at sult [142], which tells us that the potential
E = E0 ). The four types of modulation dependence of the s-polarized electrore-
spectroscopy therefore give d ln(Rs(p) )/dE flectance signal should follow the C(E)
(electroreflectance), dδs(p) /dE (electro- plot. However, this is not observed exper-
modulation of reflected phase), dψ/dE imentally. It is known that for a number
and d/dE (electromodulated ellipsom- of systems the signal may even change
etry), and ω̄(K , E)(potential-induced sur- sign with the variation of E. Equation (63)
face-plasmon shift). The potential mod- can easily explain this effect by chang-
ulation allows to separate a contribution ing the sign of (da/dσ ) = 0. Experimental
of the microscopic interfacial layer from data of Ref. [143, 144] were used to extract
da/dσ [141, 145] together with the data for
the signal. Though different equipment is
optical constants of In and Pb from the
needed to register these signals, a unified
literature.
electrodynamic theory exists for their pa-
Since the optical constants were not
rameterization [137–140]. The interested
measured in situ in the same experiments
reader is referred to Ref. [107] for a sum-
the results may be considered as a rough
mary of results of this theory and to see
evaluation only (for the same reason
how far can one proceed with microscopic
the independence of extracted values for
models in the interpretation of the po-
da/dσ on ω was not checked). Somehow
tential dependence of all the mentioned
the results shown in Fig. 18 seem to
signals. Here, we will draw one simple be in line with general expectations of
example – electroreflectance of s-polarized the interfacial relaxation theory sketched
light [141]. in the upper corner of the figure; the
As it was shown in Ref. [141] in a Drude- meaningfulness of the obtained absolute
like model of the metal and the model of values (for In, 1 µC causes a variation of a
the interface sketched in Fig. 13 by 0.26 a.u.) is discussed in Ref. [145].
√ The results for Pb [145], however, caused
d(ln Rs ) ε∞ 1 debates [107]. This electrode is not the best
= C4ω cos θ
dE c en+ candidate for probing the plasma free elec-

da 1 tron properties. Single electron interband
× (ε∞ − 1) + ε∞ − Imεb transitions for Pb lie in the low frequency
dσ en+
part (∼2 eV) of the spectrum. On the other
× ImεM + (ReεM − ε∞ )Imεb hand, the data on potential dependent elec-
troreflectance [144] were reported at one
1 relatively low frequency (1.96 eV) where an
× (63) interference between single electron and
(ε∞ − ReεM )2 + (ImεM )2
a = (aO − Aσ − Bσ 2) θ ( σs − σ)
a
Y ≡ en + dσ
d
σ
10 O σs
da
= (− A − 2Bσ) θ (σs − σ)
dσ
O σ
σs
5
p.z.c.
E In Pb
[V]
−1.5 −1.4 −1.3 −1.2 −1.1 −1.0 − 0.9 − 0.8 − 0.7
−5
−10
Fig. 18 The evaluated interfacial relaxation, extracted from the data on

s–polarized electroreflectance, as a function of electrode potential. The
insert plots show schematically, the shape of the a(σ ) and da(σ )/dσ
curves for parabolic approximation (Eq. (62)).
plasma contributions may be expected. For compression of the surface electronic pro-
more on this interference, see in Ref. [107]. file under the adsorbed anions. The greater
Simple jellium-model concepts also the field-induced adsorption, the stronger
helped to understand the potential de- the compression, and this affects the signal
pendence of a nonlinear optical signal in line with experimental observations.
from the electrochemical interface – the In this section, we have focused on
second harmonic generation [146, 147], the qualitative features of the response of
and of the frequency of surface plasmon the metal–electrolyte interface to charging
resonance [148, 149]. In both cases, the rather than on first principle calculations
effect of field-induced adsorption of an- of the absolute values of the parameters
ions at positive charges of the electrodes that characterize the interface, about which
was rationalized. An application of the one can learn more in Refs. [107, 124].
‘‘inverted’’ Lang model [107] revealed the Such calculations, however, are often
insufficiently paid back, due to the lack The simplest model that goes beyond the
of precise knowledge of the solvent effect Debye–Hückel and GC theories of point
on the metal surface electronic properties. ions and takes into account the finite size
Refs. [150, 151] show what a ‘‘resistance’’ of ions is the so-called primitive model.
the steps towards ‘‘improving’’ the picture Here, ions are described as simple hard
of this interaction encounter. spheres and the solvent is approximated
by a dielectric continuum. An even more
2.1.10 special case is the so-called restricted
Effects of the Molecular Nature of primitive model, in which both ions have
Electrolyte Solutions the same size.
Various attempts have been made in go-
The ‘‘real’’ electrochemical double layer ing beyond the primitive models towards
is not a simple arrangement of capacitors civilized models. In the most simple of
with concomitant potentials and fields, but such extensions, called the solvent-primitive
it consists of ‘‘real’’ atomic and molecu- model [152], the solvent is still treated as
lar ions and of a vast number of solvent a dielectric continuum. In addition, in-
molecules, all of which have a finite size dividual solvent molecules are treated as
and whose arrangement depends as much hard spheres. This takes into account the
on the interatomic and intermolecular in- molecular nature of the solvent in but the
teractions as on external electric fields and crudest fashion. Replacing the description
potentials. In addition, molecular ions pos- of the solvent from dielectric continuum
sess a very specific and asymmetric charge by molecular point dipoles at the cen-
distribution. Owing to their (chemical) ter of soft spheres representing solvent
interactions with the metallic phase the molecules (the Stockmayer model) leads
charge of all ions can also depend to some to the ion-dipole models. Various extensions
extent on its distance from the metal sur- of these models incorporate higher-order
face (see previous text). Furthermore, polar electrostatic moments (up to octupole mo-
solvent molecules possess a permanent ments [153]) and molecular polarizability
dipole moment and higher electrostatic for the description of the solvent bound
moments, that lead to lateral interactions charge density.
between them, which in turn, render the In order to understand ionic distribu-
simple picture of the solvent as a dielectric tions in the EDL a realistic description
continuum or n-state dipole models (see of hydration or, more generally, solva-
Sect. 2.1.8.2) inappropriate; protic solvents tion phenomena is necessary, which in
such as water and alcohols have, in addi- protic liquids implies an adequate descrip-
tion, the capability to form directional yet tion of the hydrogen-bond network. Theory
fluctuating hydrogen bonds that give rise and computer simulation of bulk liquids
to specific structural arrangements of the showed that the most efficient way to in-
solvent molecules relative to each other clude these properties into the models
and in the solvation sphere around ions. is via distributed charge models in which
These arrangements, in turn, can either the intramolecular charge distribution is
enhance or counteract the orientational represented by several point charges. The
influence due to the electronic structure point charges are adjusted to reproduce ex-
and surface charge (or potential) of the perimental dipole and/or quadrupole mo-
metal. ments of the molecule, the bulk structure
of the liquid, or various other properties. interface. Simple models of the metal are
For the ubiquitous solvent water, the sim- hard walls with image interactions that
plest models treat water as a rigid three-site account, in a very simplistic way, for the
molecule in its gas phase geometry with electronic properties of the metal. This
partial charges on oxygen and hydrogen model is nonspecific for the chemical
sites. Ions are described as charged soft identity of the metal. The jellium model
spheres with a point charge at their center. incorporates the chemical identity of the
Molecular and ionic size is maintained metal in the form of the density of free
through Lennard–Jones interactions or electrons and ionic skeleton pseudopoten-
similar potential functions in an attempt to tials; it models the electronic spillover effect
describe short-range exchange correlation into the liquid phase (see Sect. 2.1.9.1).
and long-range dispersion interactions. Several interaction models employed in
Hydrogen bonding and hydration are pro- computer simulations are chemically spe-
duced by these models through the balance cific, that is, they have been developed
between electrostatic and nonelectrostatic for a specific metal and/or surface ge-
interactions. In addition to these simple ometry. These models may or may not
rigid three-site models many more so- be based on quantum-chemical calcula-
phisticated flexible and polarizable water tions of the interaction between water
and ion models have been developed. The molecules or ions and a small cluster
importance of the dipolar nature of the (or an extended periodic slab) of ions. A
solvent and of the interactions between widely used model has been the model of
solvent and electrode were recognized the Platinum(100)–water interface [155].
in the double-layer model by Bockris, It consists of a flexible atomic represen-
Devanathan, and Müller [154]. Water ‘‘hy- tation of the metallic phase, of surface
drates’’ the electrode, which is regarded corrugation through site-dependent water-
as a giant ion, and so contributes to the platinum interactions, and a description
electric fields near the interface. of the orientational anisotropy via specific
Molecular computer simulations are the chemical interactions between the oxygen
method of choice when it comes to in- atom of a water molecule and the plat-
vestigating molecular models like the one inum metal. This model has been extended
discussed earlier, both in the bulk and at to the (111) and other surfaces; a similar
interfaces. Contrary to analytical theories, model has been used for mercury–water
simulation methods can treat a variety of interactions [156, 157].
different models on the same statistical- In spite of the fact that rather sophisti-
mechanical footing, without mathemati- cated models can be treated, several lim-
cally mandated approximations that can- itations of computer simulation methods
not be rigorously justified and/or tested. such as Monte Carlo (MC) or Molecular
In fact, computer simulations have been Dynamics (MD) limit their use in interfa-
used as computer experiments to verify or cial electrochemistry:
falsify theoretical work.
In addition to explicitly accounting for • The spatial and concentration inhomo-
the molecular nature of electrolyte solu- geneities of the EDL are large. This
tions near the EDL, computer simulations requires the simulation of large sys-
can also be used towards realistic model- tems at high electrolyte concentrations.
ing of the metal part of the electrochemical Dynamic processes near the interface
occur on a much longer time scale date, only few calculations have been
than the inherent fluctuations of the performed [158].
hydrogen-bond network and the hydra- • The use of classical simulations on the
tion shell, or they are very rare, which basis of the Born–Oppenheimer ap-
mandates long simulation times. Even proximation is problematic when deal-
with present-day computer resources, ing with metals, even in the absence
the requirements of large system size of electrochemical reactions involving
and long simulation times are hard to the transfer of electrons, which can cer-
reconcile. tainly not be approximated by classical
• Simulations with molecular solvent de- mechanics.
scription can only be performed in • Ab initio simulation methods (like
the concentration range above about the Car–Parrinello method) are cur-
0.5 mol/liter, much higher than the rently so much limited in system size
concentrations of many electrochemi- (typically substantially less than 200
cal experiments. atoms) and simulation time (a few pi-
• Simulations of the EDL structure coseconds), that they have only been
around surface defects and thus the applied to vacuum systems, for ex-
modeling of real interfaces is almost ample, to the investigation of molec-
impossible, both because of the large ular and dissociative adsorption of
system size involved and because of water on a magnesium oxide sur-
insufficient knowledge of the electronic face [159, 160], and not to an en-
structure around these defects, and thus tire metal–solution interface. However,
the lack of suitable interaction poten- these methods hold substantial promise
tials. for the future.
• MC and MD are methods of statistical
In spite of these shortcomings, simulation
mechanics. The quality of the results de-
methods have nevertheless substantially
pends on the extent to which the phase
contributed to our understanding of the
space of the system can be sampled. At
double layer because of their ability to
best the statistical error of a simulation
describe:
decreases only slowly of the order of the
square root of the computational effort; • Packing effects like the formation of
at worst, the phase space contains bot- distinct layers of solvent molecules
tlenecks, making ergodicity impossible and ions;
to achieve. • The formation of a hydrogen-bond
• Reliable and accurate potential energy network in protic solvents such as
surfaces on the basis of quantum water;
chemistry for the interaction between • Orientational structure and its change
molecules and ions with the metal near charged and uncharged surfaces;
surface are still very scarce. Especially, • Hydration of ions;
cluster calculations fail to reach the • Diffusion and reorientational dynam-
limit in which the cluster of metal atoms ics;
begins to exhibit metallic properties. • Solvent effects on the chemical dynam-
Methods based on periodic slabs appear ics of ion transfer, ET and several other
to be more promising for the future; to electrochemical reactions.
One of the goals of applying molec- density profile and ρi is the bulk value,
ular simulation to the electrochemical z the distance from the charged surface
double layer is, of course a molecular and i runs over all ion species, shows
interpretation of capacitance data (see drying at low density and exhibits a maxi-
Sects. 5 and 6). Because of the above- mum in the range of the critical density of
mentioned limitations of the method, the bulk electrolyte [161]. With increasing
progress in this field has been rather lim- surface charge density, ionic adsorption
ited. Henderson and coworkers recently passes from negative to positive.
studied the properties of the electrochemi- MC simulations of the solvent primitive
cal double layer both at high ionic concen- model [163] show, not surprisingly, that
trations (a model of the molten salt–metal the double-layer capacity increases with in-
interface) and at low ionic concentrations creasing electrolyte concentration and also
(modeling the double layer of an elec- with increasing solvent concentration. At
trolyte solution) by MC calculations. They the same time, with increasing solvent con-
showed, with the primitive model, that at centration, the ionic density profiles pass
low temperatures the double-layer capac- from nonoscillatory to oscillatory behav-
itance decreases with decreasing temper- ior, a common feature of dense systems.
atures. This behavior was observed both Furthermore, at low surface charge densi-
for the low-concentration electrolyte solu- ties the contact value of the charge density
tion [161] and for the molten salt case [162]. increases and a passage between partial
In their studies, they were only able to cal- drying (at low solvent concentration) and
culate the integral capacitance at the point partial wetting of the surface by the elec-
of zero charge. GC theory without (GC) trolyte is observed with increasing solvent
and with a Stern layer (GCS) as well as density. The temperature dependence is
the mean spherical approximation (MSA), such that the ion adsorption is nega-
on the other hand, predicts a monotonous tive at low temperature and positive at
decrease of double-layer capacitance with high temperature. The negative ion ad-
temperature. The primitive model MC data sorption at low temperature becomes less
showed that the capacitance versus tem- pronounced with increasing particle den-
perature curve goes, at least at lower ion sity. The behavior can be explained by the
densities, through a maximum and that fact that electrostatic ion–ion interactions
at high temperature, analytical theories and dominate at low temperature; whereas the
MC experiment agree well. The primitive excluded volume effect of the solvent is
model simulations thus indicate that ion more important at higher temperature.
size-dependent effects become important, Extending these studies to ion-dipole
thus limiting the range of applicability of mixtures in which a background dielectric
the analytical theories of sizeless ions to constant is replaced by fluctuating solvent
the high-temperature regime. point dipoles, leads to ‘‘practical nonergodic-
At high electrolyte concentrations, the ity’’ due to the formation of clusters and/or
primitive model simulations clearly ex- strings of particles [164].
hibit the importance of ion size in the Computer simulations of ‘‘realistic’’
form of oscillatory density and charge models of water, ions, and metal surfaces
profiles in the vicinity of charged sur- also suffer from nonergodicity problems;
faces[162].
The MC adsorption isotherm however, they are able to paint a very de-
= i [ρi (z) − ρi ]dz, where ρi (z) is the tailed picture of the microscopic liquid
86
10 0.15
0.1
8 0.05
6 0
ρµ (z)
ρ (z)/ρ b
[Cm−2]
4 − 0.05
2 Electrochemical Double Layers
− 0.1
2
− 0.15
(a) (c)
1
0.5
0.0 1
[V]
χ (z)
ρ c(z)
−0.5 0
[1010 Cm−3]
−1.0
−1
−1.5
−2
−15 −10 −5 0 5 10 15 −15 −10 −5 0 5 10 15
z z
(b) [Å] (d) [Å]
Fig. 19 (a) Oxygen (full) and hydrogen (dashed) density profiles ρ(z), normalized to the bulk density ρb . (b) Charge density ρc (z). (c) Dipole density
ρµ (z). (d) Water contribution to the surface potential χ(z) calculated from the charge density ρc (z) by means of Eq. (64). All data are taken from a 150 ps
simulation of 252 water molecules between two mercury phases with (111) surface structure using Ewald summation in two dimensions for the
long-range interactions.
structure near a metal electrode and near artificial character of these systems should
a liquid–liquid interface. Models of the in- be kept in mind.
terface range from simple image charge The charge density profile and the dipole
models [165] through jellium models [166] density profile can be calculated from the
to complex potential energy functions atomic density profiles and the orienta-
derived from quantum-chemical calcu- tional distributions. Figures 19b and 19c
lations [167, 168] of clusters of metal show the dipole and charge density, respec-
atoms interacting with water molecules. tively, for water near the mercury surface.
These studies have been reviewed re- For both the first and the second water
cently [169–171]. Only some key results layers, a quadrupolar charge profile with al-
are discussed in the following text, and ternating regions of positive, negative, and
the interested reader should consult the again positive charge density is observed.
reviews and the references therein for fur- The form of the charge profiles originates
ther information. from the fact that the centers of mass (or
Figure 19 shows as an example the data the oxygen atom) of the water molecules
from a simulation of TIP4P water that are mostly well localized, while the hydro-
is confined on both sides by a rhombo- gen density distribution is broadened as a
hedral mercury crystal with (111) surface result of librational motions and as a result
structure. Bosio and coworkers [172] de- of the hydrogen bonding between layers.
duced from their x-ray studies that a solid Beyond the second layers the charge den-
α-mercury lattice with a larger lattice con- sity approaches zero within the limits of
stant in z direction may be used as a statistical uncertainty. The dipole density
good structural model for liquid mercury. profile indicates ordered dipoles in the ad-
Thus the mercury phase was modeled as sorbate layer. The orientation is largely due
a rigid crystal, in order to simplify the to the anisotropy of the water–metal inter-
simulations. The surface of such a crystal action potential in this particular model,
shows rather low corrugation. Figure 19a which favors configurations in which the
contains the oxygen and hydrogen den- oxygen atom is closer to the surface. Most
sity profiles. As a result of the significant quantum-chemical calculations of water
adsorption energy of water on transition near metal surfaces to date predict a sig-
metal surfaces (typically of the order of 20 nificant preference of ‘‘oxygen-down’’ con-
to 50 kJ mol−1 ; see, e. g. Ref. [173]), pro- figurations over ‘‘hydrogen-down’’ ones
nounced density oscillations are observed at zero electric field (e. g., Ref. [168,
next to the metal. Between three and four 178–183]). The dipole orientation in the
water layers have also been identified in second layer is only weakly anisotropic (see
most simulations near uncharged metal also Fig. 20). Solving the one-dimensional
surfaces, depending on the model and on Poisson equation with the charge density
statistical accuracy. Beyond about 10–12 Å profile ρc (z) (Fig. 19d), the electrostatic
from the surface the density is typically potential drop near the interface can be
constant and equal to the bulk value. In calculated according to
strong unscreened electric fields, several
z
authors [174–177] report a phase transi-
tion towards a ferroelectric crystalline state χ(z) = −4π ρc (z ) · (z − z ) dz
−∞
in their simulations. However, the rather (64)
p(cos ϑµ ) ρ(z) p(cos ϑµ )

a
σ = 8.8
c
σ = 4.4
d
e
σ = 2.2
f
h σ=0
j σ = − 2.2
l
σ = − 4.4
m
σ = − 8.8
o
−1 − 0.5 0 0.5 1 −1 − 0.6 − 0.2 0.2

cos ϑµ cos ϑµ
Fig. 20 Left side: Orientational distribution of the density profile ρ scaled to its bulk value.
the molecular dipole moment on uncharged Right side: Orientational distributions of
Hg(111). Cos(θµ ) is the cosine of the angle adsorbed water molecules for various
between the water dipole vector and the surface homogeneous surface charge densities, σ , which
normal that points into the water phase. Panels a are given in units of µC cm−2 on the graphs.
to p on the left are sampled from the distance Data are from simulations of 700 TIP4P water
intervals that are indicated by the cuts through molecules between Hg(111) surfaces.
It has been demonstrated in Ref. [184, potential in the center of the lam-
185] that the use of the Ewald sum- ina.
mation is crucial to obtain the field- Lateral density fluctuations are mostly
free bulk region of constant electrostatic confined to the adsorbed water layer. In the
first layer the oxygen distribution shows the basis of the water–metal interaction
the structure of the substrate lattice. In the potential. Within the adsorbate layer, there
second layer the distribution is more or is a transition from the preference for ori-
less isotropic. As a consequence, oxygen entations in which the dipoles point more
motion is predominantly oscillatory rather or less into the solution (a and b) to one
than diffusive in the first layer. Con- where a substantial fraction of the dipoles
sequently, the self-diffusion coefficient point more or less towards the surface (c
in the adsorbate layer is strongly re- and d). The orientational anisotropy ranges
duced compared to the second or third as far into the liquid phase as the density
layer [186–189]. It was concluded that the inhomogeneities do (roughly up to panel
motion in the first layer is characteristic of m), with increasingly less pronounced
a solid phase, while the motion in the sec- features. Slightly beyond the second max-
ond layer has more liquid-like character. imum in the density profile the orienta-
The liquid–liquid water–mercury inter- tional distribution is isotropic, as it has to
face has also been studied [190]. The major be the case for a bulk-like liquid. Thus,
difference of the water structure between like in the case of nonpolar surfaces, the
the liquid–solid and the liquid–liquid in- orientational distribution is governed by
terface is due to the capillary width of water–water interactions. All orientational
the liquid mercury surface, which smears distributions are rather wide and liquid-
out the water-density profiles consider- like, although the fact that the orientational
ably [191, 192]. preference changes within the adsorbate
The orientational structure of water near layer is sometimes attributed to ice-like
a metal surface has obvious consequences structural elements near the interface (see
for the electrostatic potential across an in- Ref. [193] for a more detailed discussion).
terface, since any orientational anisotropy An external electric field changes the ori-
creates an electric field that interacts with entational distribution and consequently
the metal electrons. Hydrogen bonds are the orientational polarization of the water
formed mainly within the adsorbate layer molecules in the interfacial region. The ef-
but also between the adsorbate and the fect of homogeneous [174, 175, 194–197]
second layer. The left side of Fig. 20 shows and inhomogeneous [198] electric fields on
the orientational distribution of the molec- the orientational distribution near smooth
ular dipole moment, relative to the surface, model surfaces has been investigated us-
normal in various distance ranges from the ing lattice models [197] and distributed
Hg(111) surface. Additionally, the oxygen point charge models for water. Also the
density profile is plotted. The baselines be- field-induced changes on the more realis-
tween distribution functions cut through tic Pt(100) surface were studied [176, 199,
the density profile. The distribution func- 200]. Because of the absence of free ions
tion in each panel on the left side is for in these simulations of pure water the
the subset of molecules that are located in electric field is only screened by the wa-
the distance range between these lines on ter dipoles themselves; consequently a net
the right side. Over the first peak in the electric field and a concomitant polariza-
density profiles (panels a to d), there are al- tion persist through the lamina. At large
most no molecules whose dipole moment field strengths, this has been found to
is perpendicular to the surface, as would lead to a field-induced phase transition
be expected for an isolated molecule on to a crystalline water phase [174–176]. The
right part of Fig. 20 shows the orientational are not contact adsorbed, Cl− being a
distribution of the dipole moment vectors borderline case. Calculations on the ba-
in the adsorbed water layer for various sur- sis of quantum-chemical potentials [167,
face charge densities, σ . In the field-free 191] show contact adsorption for almost
case a wide bimodal-orientational distribu- all ions, although a detailed analysis of
tion is observed. The overall preference of the local structures around these ions re-
‘‘oxygen down’’ bonding is manifest in the vealed characteristic differences between
larger probability to find positive values large and small ions.
of cos(θ) (corresponding to angles smaller The first simulation studies of full
than 90◦ between dipole vector and the double layers with molecular models of
surface normal that points into the liquid ions and solvent were performed by
phase) than to find negative ones. For pos- Philpott and coworkers [212, 213] for the
itive surface charges, the average dipole NaCl solution. The authors studied the
moment (the first moment of the distribu- screening of a negative surface charge by
tion) shifts with increasing surface charge free ions in several highly concentrated
density towards larger absolute values and NaCl solutions. A combination of (9-3) LJ
the distribution becomes increasingly nar-
potential and image charges was used to
rower. For low negative surface charge
describe the metal surface. More recently,
densities, the dipole-orientational distri-
Spohr [214, 215] investigated the 2.2 molal
bution becomes more symmetric around
NaCl and CsF solutions in the vicinity of a
the parallel orientation. At larger negative
corrugated surface as a function of surface
surface charge densities, the orientational
charge density. In a similar manner the
distribution changes in such a way that the
interface between a 1 M KCl solution and a
hydrogen atoms point preferentially to the
mercury electrode was studied by Dimitrov
surface. The effect of field-induced crys-
and Raev [216].
tallization was observed at much higher
surface charges [176]. The surface X-ray In the studies of Ref. [214, 215], which
experiments by Toney and coworkers [201] will be discussed in some more detail, wa-
give experimental evidence for voltage deter films consisting of 400 water molecules
pendent ordering of water on a silver solvate 32 ions in the vicinity of the metal
electrode. They observed a shift of the surface, with image charges and a full 2D
silver-oxygen distance with applied poten- Ewald treatment for the Coulomb interac-
tial. However, they also made a debated tions. Like in the studies by Philpott and
conclusion a strong increase in local water coworkers, the solution is not necessarily
density near the surface, while MD data electroneutral; the total charge in the so-
suggest only an increase in correlation but lution is balanced by the image charges,
not in overall packing density. which give rise to a surface charge density
Several simulation studies of the poten- σ equal to the excess image charge divided
tial of mean force of a single ion near by the area of the interface.
a metal surface have been performed, Figure 21 shows the ion density profiles
with image charge models and quantum- near the metal surface for three surface
chemical models of the metal [202–211]. charges. Beyond z = 15 Å all ion density
With the image charge models, only the profiles are identical within the limits
larger halide ions are contact adsorbed, of statistical errors. The oxygen density
whereas small ions such as Li+ and F− profile does not change much with surface
NaCl CsF
ρO(z) ρO(z)
4 4
3 3
2 2
1 1
10 σ = +9.9 10 σ = + 9.9
8 8
6 6
4 4
2 2
10 σ=0 10 σ=0
8 8
6 6
4 4
2 2
10 σ = −9.9 10 σ = − 9.9
8 8
6 6
4 4
2 2
5 10 5 10
z z
[Å] [Å]
Fig. 21 Density profiles ρ(z) and running integrals Nion (z) of the ion densities for cations (full lines) and anions (dashed lines) at three different surface
2.1 Electrochemical Interfaces: At the Border Line
charge densities in units of µC cm−2 as indicated. Left: NaCl solutions; right: CsF solutions. The top graphs show, for reference, the corresponding
oxygen density profile near the uncharged surface. Density profiles are normalized to the bulk densities corresponding to 2.2 molal solutions in each case.
91
charge and is repeated here only to surface charge densities. Obviously, with
provide a geometric reference for the ion the models used in this study, Na+ does
positions. Together with the ion densities not contact-adsorb. The Cl− density profile
the running integrals of the densities, at σ = −10 µC cm−2 is similar to the one
defined as at σ = 0.
z The monotonically increasing curves of
Nion (z) = Bx By ρ(z ) dz (65) the running integrals give the number of
0
ions, whose distance from the electrode
are plotted for cations (full lines) and is smaller than z. The position at which
anions (dashed lines). Bx = By = 18 Å are the difference between the anion and
the box dimensions parallel to the interface cation running integral becomes 0 (for
and ρ is the particle number density. the uncharged surface) or ±2 (for the
Figure 21 (left) shows the ion density charged surfaces) corresponds roughly to
distributions of NaCl in the vicinity of the thickness of the diffuse part of the
the metal electrode. Near the uncharged double layer. This thickness is less than
electrode, there are no pronounced adsorp- 10 Å in all cases.
tion maxima. The density of Na+ (full line) Figure 21 (right) shows the correspond-
is slightly increased in the range around ing density profiles for the CsF solutions.
z = 4.3 Å between the first and second At zero surface charge, there is very little
density maximum of water. The Cl− den- contact adsorption of Cs+ and no contact
sity (dashed line) is significantly reduced adsorption of F− . The preferred position of
up to about z = 5 Å. There is no contact ad- the Cs+ ions is in the second water layer,
sorption of Na+ , since no cations are found while F− ions prefer the region between
for z < 2.8 Å, while water molecules can be two water layers, similar to Na+ . At posi-
found up to z = 2 Å. Cl− ions, on the other tive surface charge densities, the amount
hand, can be found (with low probability) of F− in the interlayer region increases,
at distances below 2.3 Å. but no contact adsorption is observed. No
At positive surface charge density, the Cs+ cations are found in the first layer. At
Cl− density exhibits a large maximum at the negative surface charge density, Cs+
very short distance from the electrode. forms a contact-adsorbed layer, similar to
The position of this maximum is closer Cl− at the positive surface charge. The
to the electrode than that of the first thickness of the diffuse layer is also in
water layer, thus giving a clear indica- the range of 10 Å, judging from the be-
tion of contact adsorption of this anion. havior of the running integral of the ion
The Na+ density near the electrode is density.
slightly reduced as a result of the re- No contact adsorption occurs in the
pulsion between the positively charged simulations on uncharged electrodes. The
cations and the positively charged sur- small ions Na+ and F− do not adsorb
face. The position of the first maximum directly on the electrode surface at mod-
of the Na+ density profile is not shifted erate negative and positive surface charge
very much. densities, respectively. These ions form
At negative surface charge density, the rather rigid hydration shells consisting of
Na+ density exhibits a large maximum six or seven water molecules and favor
at around z = 4 Å. This position is very the interlayer region between the first and
similar to the one at vanishing and positive second water layer, in which they can
∞
form stable hydration shells. However, ◦
i = [ci (z) − ci ] dz (66)
the large ions Cs+ and Cl− exhibit con- zG
tact adsorption when their interaction
with the surface charge density becomes Here, coordinate z is normal to the inter-
attractive. face, and the integration is performed over
From the ion density profiles, it is the whole perturbed region. An important
obvious that the surface charge is screened point in this formula is the lower integra-
within less than 10 Å. Thus the thickness tion limit, zG , corresponding to the choice
of the diffuse layer is of the same order of the ‘‘phase boundary’’, whose attribu-
tion is not self-evident. This ambiguity of
of magnitude as the one derived from the
the definition (66) is removed by the ther-
Debye length (rD = 2.1 Å at 2.2 mol/liter),
modynamics of charged interfaces based
in spite of the invalidity of the GC theory
on the Gibbs electrocapillary equation (67):
at these high concentrations. The results
clearly show the correlation between the
dγ = − j dµj (67)
strength of the hydration forces and j
the preferred positions of the various
ions relative to charged and uncharged where γ is the interfacial energy, µj
electrode surfaces. the electrochemical potential of the com-
Solvent effects on chemical reactions ponent j (the value of µj is identical
are presently being studied intensively by in the bulk medium and within the
molecular computer simulation methods. interphase, in view of the thermody-
Outer sphere ET (e. g. Ref. [217–228]), ion namic equilibrium); summation includes
transfer [229–232], proton transfer [233], all components of the phases in contact.
and bondbreaking [234, 235] are among The variations of different electrochemical
the reactions studied near electrodes and potentials in Eq. (67) are not independent;
liquid–liquid interfaces. Much of this work in the bulk phase they are coupled by
has been reviewed recently [170, 236, 237] the Gibbs–Duhem identity. If the elec-
and will therefore not be discussed here. trochemical potentials of all components
All studies involve the calculation of a free j , except for component i and solvent s,
energy profile as a function of a spatial or are kept constant, Eq. (67) leads to the
a collective solvent coordinate. relation:

(s) Ni ∂γ
2.1.11 i ≡ i − s =− ,
Ns ∂µi
Ionic Adsorption
keeping µj = constant, j = i, s
2.1.11.1 Qualitative Aspects (68)
In the general meaning of this term, The multiplier Ni /Ns in this equation rep-
adsorption of a solution component means resents the ratio of moles of components i
any perturbation of its concentration and s in the bulk phase, or the ratio of their
within the interphase, ci (z), compared molar fractions. The resulting quantity,
with its value in the bulk phase, ci ◦ . Its (s)
i , which is called the (Gibbs) adsorption
intensity is characterized by the integral of component i with respect to component
value of this perturbation, which is called s (or the relative interfacial excess of com-
the adsorption of this component: ponent i), is already independent of the
choice of the ‘‘phase boundary’’, zG . The The estimates show that the charge-
latter is often chosen in the way to make the independent term in Eq. (69) is only
solvent adsorption zero, s = 0; then, both noticeable for not too dilute solutions,
(s)
quantities become identical, i ≡ i . In about 1 M or higher. For lower elec-
dilute solutions the so-defined position of trolyte concentrations, it can be mostly
the separation plane is close to the physi- disregarded. The other terms in Eq. (69)
cal boundary between the electrode and the represent monotonous functions of the
adjacent solvent layer, being practically in- electrode charge, with the anion/cation
dependent of the electrolyte concentration. contribution being dominant for suffi-
Let us consider first a surface-inactive ciently positive/negative values.
solution, whose properties have already The behavior, that the electrolyte is sur-
been outlined in Sects. 2.1.3 to 2.1.6. By face–inactive with respect to an interface
definition, this term means that the com- over a wide range of potentials includ-
pact layer is solely composed of solvent ing the p.z.c., represents a rare exception,
molecules. The distribution of compo- while the majority of solute ion and neu-
nents within the diffuse layer is deter- tral species demonstrate specific adsorption,
mined mostly by the electrostatic forces. If that is, they significantly penetrate the
the image forces (see Sect. 2.1.11.3) can be compact layer. However, the existence of
disregarded the concentrations of solute at least one surface-inactive electrolyte MA
species, ionic or neutral, at the p.z.c. are for a particular metal–solvent interface
identical to their bulk values, and the in- provides very important advantages for
tegral (66) over the diffuse layer vanishes. the further study of the surface activity
Inside the compact layer the concentra- of other species. First, the measurements
tion, ci (z), is zero, and the integration in pure MA solutions allow one to de-
in (66) yields Eq. (12) for the Gibbs ad- termine the background characteristics of
sorption of surface-inactive components, the metal–solvent interface, including the
± = −zH c, which allows one to measure value of the p.z.c. and the compact-layer
capacitance curve (see Sects. 2.1.2–2.1.4).
the thickness of the ion-free layer of the
Then, further studies (e.g. of the adsorp-
solvent, zH .
tion of anion A ) are performed in mixed
For nonzero charges of the electrode,
electrolyte solutions, MA + MA , possess-
σ = 0, the distribution of ions across
ing a constant ionic strength but a variable
the diffuse layer is governed by the self-
fraction of anions A [238, 239]. If the prop-
consistent electric potential, for example,
erties of these two anions inside the diffuse
given by the formulas of the GC theory
layer are sufficiently close, the variation of
given in Sect. 2.1.4.2. In particular, for a
the interfacial energy, γ , with the fraction
binary 1 : −1 electrolyte, this model yields
of anions A enables one to determine the
for the adsorption of surface-inactive ions:
specific adsorption of these anions, that is,
(1)
its amount inside the compact layer, A ,
± = −zH c−(1/2e)
rather than the overall interfacial excess of
× [σ∗ ± σ − (σ 2 + σ∗2 )1/2 ] (69) these species, A , that appears in Eqs. (67)
and (68).
F is the Faraday constant, c the electrolyte Alternatively a study can be realized
concentration, and σ∗ is the characteristic for binary electrolyte solutions, MA , at
charge defined by Eq. (18). varying concentrations [24] that provide,
in favorable cases, the same information. cations. In particular, the alkali cations
However, the specific adsorption of A are almost surface-inactive, except for
must vanish at sufficiently high (but still the interval of sufficiently high negative
experimentally available) negative elec- electrode charges, where the adsorption
trode potentials, so that the capacitance of Cs+ is most noticeable (see also
curves for the MA and MA solutions Sect. 2.1.10). However, even for this case
at identical concentrations coincide. This the adsorption is weak, that is, the charge
condition is necessary for the back- of their specific adsorption, σ1 , is less than
integration operation, thereby allowing the the absolute value of the electrode charge,
calculation of the charge-potential depen- σ , that is, the charge of the diffuse layer
dence for MA solutions. Besides, one has has the opposite sign compared to the
to subtract the diffuse-layer adsorption of electrode charge, and the distribution of
A , A , from its overall adsorption, which
(2)
the average potential across the interphase
is based on the formula (69) in which is given by a monotonous function.
the electrode charge, σ , is replaced by Anion adsorption possesses quite differ-
(1)
the sum, σ + σ1 , σ1 = zA eA being the ent features. For most anions, the specific

charge of A inside the compact layer, and adsorption is only absent for sufficiently
zA the charge valence of anions A . As high negative electrode charges. The
one can conclude from the analysis in diminution of this charge results in a rapid
Sect. 2.1.5.1, the charge dependence pre- increase of the amount of anions inside the
dicted by this formula cannot be verified compact layer, even before the p.z.c. of the
from existing experimental data for high system is reached. For the zero charge of
electrode charges, so that the separation of the electrode, the EDL is formed by the an-
the overall adsorption value into the diffuse ions inside the compact layer and the coun-
and compact-layer contributions becomes tercharge in the diffuse layer. It means a
ambiguous. shift of the corresponding electrode poten-
These measurements enable one to de- tial, that is, the p.z.c., in the negative direc-
termine the (specific) adsorption isotherm, tion, the value being correlated with the in-
that is, the dependence of the interfacial tensity of the anion specific adsorption. For
excess of the species on their concentra- their adsorption at the positively charged
(1)
tion in the bulk phase, A (cA ). For a metal surfaces, one observes the recharg-
constant value of the bulk concentration, ing effect; the absolute value of the anion
its adsorption also depends on the electri- charge, σ1 , exceeds the electrode charge so
cal variable, for example, on the electrode that the diffuse layer should be charged pos-
potential E. In mixed electrolytes, this pro- itively. It leads to a nonmonotonous profile
cess is also affected by the overall ionic of the electric potential across the inter-
strength, in particular via the diffuse-layer phase, with important consequences for
structure. In more complicated cases a the phenomena determined by the diffuse-
combined specific adsorption is observed, layer effects, Sect. 2.1.5.
that is, the compact layer contains (besides Data on ionic adsorption are available for
the solvent) species of different kinds, for numerous systems, see for example, [34,
example, both cations and anions. 240–245]. One can see that the surface
Generally the specific adsorption at the coverage by adsorbed ions (adsorption
surface of s,p-metals from protic solvents values with respect to their maximum
is more pronounced for anions than for level corresponding to the completely
covered surface) is mostly much lower adsorption obeys the Henri isotherm, that
than 1 for mercury and ‘‘mercury-like’’ is, the amount of ions inside the compact
metals, even at high electrode charges. layer is proportional to the bulk con-
This limitation arises from a strong centration. At higher concentrations the
repulsive interaction of the electrostatic adsorption increases much more slowly,
origin between adsorbed ions. On the approximately like the logarithm of the
other hand, the adsorption may approach a concentration. The behavior in both re-
complete coverage in some other systems, gions is influenced by the electrode poten-
for example, for halides adsorption at the tial and the ionic strength of the solution.
silver surface [246], which implies that the In the further discussion of the modeling
ions have practically lost their charge in of the adsorption isotherms, (c), it is
the transition to their adsorbed state, so assumed that the contribution of the
that this process resembles the adsorption diffuse-layer adsorption, (2) , is already
of neutral organic molecules, in which subtracted or simply disregarded, that is,
the complete coverage of the surface is the symbol will be used for the amount
often reached. Similar pronounced charge of specifically adsorbed species (called (1)
transfer from adsorbed ions to the metal in Sect. 2.1.11.1).
takes place for many transition metals. Generally the expression for this depen-
Despite of this charge transfer many dence should be derived from the equality
of these systems may be treated as of the electrochemical potential of the ion
‘‘ideally polarizable electrodes,’’ if the in the adsorbed state and in the bulk
adsorbed species are not transformed solution:
into a different component present inside
the bulk phase. The latter condition is µads () = µsol (c) (70a)
violated, for example, in the hydrogen
adsorption at metals of the platinum group In most cases the solution is sufficiently
in which the adsorbed hydrogen atoms dilute to use the ideal-solution approxima-
can be in equilibrium with protons in tion for the bulk value,
solution and hydrogen molecules in gas
µsol (c) ∼
= kB T ln(mc) + constant (70b)
phase or hydrogen dissolved inside the
metal. The latter system corresponds to m being molar fraction of the surface-
perfectly polarizable electrodes, see Ref. [13] active ion in the mixed electrolyte, and
for further discussion. c the overall concentration (m = 1 for the
Another possible complication may arise binary electrolyte). On the other hand the
from a slow approach to the thermody- approximation for the µads () dependence
namic equilibrium leading to hysteresis must take into account the interactions
phenomena, which are also typical for between the adsorbed ions, that is, the
many transition metals. change of the energy to transfer an ion
from the bulk solution to the adsorbed
2.1.11.2 Ion Adsorption Models state as a function of the adsorption level.
For metals of the mercury group, one The first theory of ion adsorption was
can distinguish different regions in the proposed by Stern [247]. His results can
ion adsorption isotherms. For sufficiently be obtained from Eq. (70a) if the elec-
low bulk concentrations (which may be trochemical potential of the ion in the
too low to be studied experimentally), the adsorbed state is represented as the
sum of the entropy term (with account predicted by the Stern theory. For zero elec-
of the saturation term) and the work to trode charge, the average potential in the
transfer the ion from the bulk solution adsorption plane (ϕGC in Eq. (72)) is equal
into the adsorbed state in the average field to the electrode potential (at the p.z.c.),
of the EDL: Eσ =0 . Then the derivation of the Eq. (72)
leads to an inequality:

µads = kB T ln
(max − ) zi e dEσ =0
<1 (73)
+ zi eϕGC + constant (71) kB T d ln c
zi e is the charge of the species. Stern as- However, the experimental data show
sumed the location of the adsorption ion that the left-hand-site quantity exceeds
at the outer Helmholtz plane separating significantly its upper limit, 1.
the compact and diffuse layer, so that the This ‘‘paradox’’ (which remains unre-
potential in the work term in Eq. (71) solved even if the adsorption plane is
corresponds to the potential difference shifted into the compact layer) represented
across the diffuse layer. The combination a strong stimulus for the development
of Eqs. (70a) and (71) gives Stern’s expres- of the ion adsorption theory. It was
sion for the adsorption isotherm from a shown [249] that it originated from the cal-
binary electrolyte solution for the case of a culation of the work term in Eq. (71) as an
single ionic species being accumulated in average potential in the adsorption plane.
the adsorbed state: The qualitative explanation of experimen-
tal observations can be achieved, if one
−(W̄ads + zi eϕGC ) takes into account that the transfer of the
= c exp
(max − ) kB T ion into the adsorption site is accompanied
(72) by a redistribution of the other adsorbed
where ‘‘the potential of specific adsorp- ions that creates a ‘‘hole’’ for it with the
tion’’ W̄ads originated from the constant diminished repulsive potential. This in-
terms in Eq. (70b) and (71) which re- applicability of the approximation of the
flect the nonelectrostatic contributions to uniform distribution of the adsorbed ion
the energy of transfer into the adsorbed charge along the plane is called discreteness-
state, for example, the partial desolvation of-charge effects, which must be taken into
of the ion. account in the theory of this phenomenon.
This model allows one to explain such Another crucial factor is an ade-
qualitative features of the phenomenon as quate inclusion of these diffuse-layer and
the drastic (exponential) dependence of discreteness-of-charge effects into the
on the electrode charge, as well as the ex- model, which play a marked role if the over-
istence of the aforementioned two regions all electrolyte concentration is well below
in the (c) variation. However, a more 1 M. It is why the adsorption parameters
detailed experimental study has revealed found by the interpretation of experimen-
serious deviations from the predictions tal data with the use of various isotherms
of Eq. (72). For many ions, one can ob- derived mostly for uncharged species
serve the effect of Esin and Markov [248], (Flory–Huggins, Frumkin, Temkin, etc)
a stronger shift of the p.z.c. with the vari- may strongly deviate from their proper
ation of the bulk concentration than it is values.
The aforementioned diffuse-layer and to minimize the total energy of the system
discreteness-of-charge effects have been after the insertion of an additional ion,
taken into consideration in the the term in Eq. (74) dependent on the
model proposed by Grahame and charge of the specific adsorption, σ1 , was
Parsons [26, 250–252]. First, it was diminished by ‘‘the discreteness-of-charge
assumed (unlike in the Stern model) that factor’’, λ, the value of which should
the specifically adsorbed ions were located lie between 0 and 1. In the original
at the distance from the metal surface (in formulation by Grahame, this factor was
‘‘the inner Helmholtz plane’’) ensuring expressed through the ratio of KO and the
their maximum bond strength, owing to capacitance of the whole compact layer,
the combination of forces of electrostatic KH in Eq. (19). Later, its value was treated
and quantum-mechanical origins. It shows as an unknown parameter to be found
the need for the partial or even complete from the fitting of theoretical predictions
desolvation of the adsorbed species and to experimental data.
its deep penetration into the compact The use of the thus modified Eq. (74)
layer. The position of this adsorption plane and the equality of the electrochemical
depends on all components of the system, potentials (70a) lead to the adsorption
metal, solvent, and adsorbed ion. isotherm:
The second modification compared to ϕGC
ln(mc) = ln + 2B + zi e + ln β̃
the Stern theory was the introduction kB T
of ‘‘the discreteness-of-charge factor’’, λ. σ
The electrostatic part of the work of ion ln β̃ = ln β̃o + zi e KO (75)
kB T
transfer from the bulk solution into the λ
adsorption plane across the EDL field in B = zi 2 e2 KO
2kB T
Stern’s approximation may be represented
as the sum of the contributions of the Here the coefficient, B, is called attraction
compact and diffuse layers: constant, in analogy with the correspond-
ing factor in the Frumkin adsorption
WEDL = zi e(ϕ1 − ϕGC ) + zi eϕGC isotherm. However, it has got the oppo-
= zi e(σ + σ1 )KO −1 + zi eϕGC site sign, that is, it reflects the repulsion
between the adsorbed species, the am-
(74) plitude of which is diminished owing to
In the second expression in this formula, the relaxation of the adsorbed ion ensem-
the potential difference across the exter- ble (factor λ). The entropy term for the
nal part of the compact layer is replaced adsorption state is taken without the ‘‘sat-
by the sum of the electrode and specifi- uration term’’, compare Eqs. (71) and (72),
cally adsorbed charges, with the coefficient since the charging degree for s,p-metals is
inversely proportional to the integral ca- mostly very low with respect to the com-
pacitance of ‘‘the external part of the complete coverage.
pact layer’’, KO , that is, between the inner The expression (75) represents a virial
(adsorption plane) and outer (diffuse-layer isotherm for the (c) dependence, with
boundary) Helmholtz planes. a correction for the diffuse-layer effect,
To take into account the adjustment of if the electrical variable is kept constant.
the distribution of the previously adsorbed Unlike the case of the Frumkin isotherm
ions inside the adsorption plane in order in which it was the electrode potential, the
Grahame–Parsons isotherm uses the elec- Popov [253]. Further discussion on this
trode charge, σ , whose value influences development is given in Chapter 3.1 of
the parameter, β̃, related to the standard this volume.
adsorption energy. Equation (75) predicts
that the isotherms for different electrode 2.1.11.3 Image Potential for Ions in the
charges must be congruent. Diffuse or Compact Layer
This isotherm was used successfully to Within the framework of the GCSG treat-
interpret experimental data for adsorp- ment (Sect. 2.1.4) the ion concentration
tion of various simple inorganic ions, across the diffuse layer is governed by
including weak specific adsorption of the Boltzmann distribution in the aver-
cations. At the same time, significant age electric field created by the electrode
deviations have been observed for some charge (plus the charge of the specifically
other systems, especially for low sur- adsorbed ions in the general case), and the
face coverages as well as for the surface ion countercharges in the diffuse layer:
recharge by adsorbed anions. Moreover,
in some cases in which the predictions ◦ −zi eϕ(z)
ci (z) = ci exp (76)
of Eq. (75) were formally in accordance kB T
with experimental data, the values of the
This formula implies that the presence of
‘‘discreteness-of-charge factor’’, λ, were
a particular ion at point z does not disturb
found outside the expected interval, nega-
this potential distribution. In reality, it is
tive or exceeding 1.
only valid as a zero order approximation.
Another shortcoming of this model
A more accurate formula contains an
is the absence of its derivation from
additional factor:
the theory based on the analysis of the

statistical-mechanical properties of the ◦ −zi eϕ(z) Wim (z)
ci (z) = ci exp −
ensemble of adsorbed ions. Progress in kB T kB T
the latter direction is discussed in the next (77)
subsection. where the ‘‘image potential’’ Wim (z) repre-
All these approaches disregard the effect sents the difference between the energies
of the gradual change of the dielectric of the whole system if the additional ion of
properties of the compact layer as a type i is added either at point z or in the
consequence of the replacement of the bulk solution, as a source of the external
solvent molecules in it by the adsorbed field for the rest of the system. Since this
species. This approximation limits the quantity is related to a single ion, expression
applicability of such theories to moderate (77) contains the Boltzmann constant, kB .
adsorption and to adsorbed species of a Within the diffuse layer the domi-
small size. In the opposite case of large nant contribution is given by forces of
ions (especially, those of organic type), electrostatic origin, analogous to the De-
this effect may become predominant even bye–Hückel ion atmosphere around each
for sufficiently low adsorption charges. charge species in the bulk solution. It
In the simplest approximation, one may was assumed frequently on the intuitive
accept a linear dependence of the compact- ground that the image potential in Eq. (77)
layer capacitance on the charge of the related to the location of the ion in the
specific adsorption, such as in the model vicinity of the metal may be modeled by
proposed by Kolotyrkin, Alexeyev, and its expression for a charged species in the
dielectric medium near its boundary with (which already appeared earlier, for exam-
a conductor: ple, in Eq. (38a)):
◦ (zi e)2 LH ≡ (4πCH )−1 (80)

Wim (z) = − (78)
4εz
The value of this length can be found
The negative sign in this formula means
for any metal–solvent interface as soon as
an attraction of the ion to the surface, in
the capacitance data for a surface-inactive
particular an accumulation of the ions in-
electrolyte are available; for example, it is
side the diffuse layer near the uncharged
about 0.3 Å for the Hg–water interface.
metal surface (at the p.z.c.). However,
As in other expressions containing LH ,
a more substantiated analysis has dis-
this length is multiplied by the bulk-solvent
proved this conclusion completely, see re-
dielectric constant, ε, which gives about
views [89, 254, 255] and references therein.
2.4 nm for this interface.
Let us consider first an idealized case
Equation (79) shows that Eq. (78) is
of the uncharged metal in contact with the
only valid at distances from the surface
solvent without any electrolyte. The expres-
much greater than this distance, z εLH ,
sion for the image potential within a wide
where the amplitude of the image potential
range of the ion-metal distances (except for
is negligible compared with the thermal
the close vicinity to the compact-layer sur-
energy, kB T , that is, the concentration
face) can be derived without specifying the
profile in Eq. (77) is not influenced by this
particular form of the dielectric properties
factor. In the other interval of the distances,
of the media in contact, solvent and metal,
z εLH (but not too close to the compact
including their defect structure within the
layer) Eq. (79) may be approximated as:
interphasial region (compact layer). The
only assumption on these properties of the (zi e)2
media and the interface is the existence of Wim (z) = (81)
4εz
a thin layer within the interphase having
a lowered value of the effective dielectric that is, the ion is repelled by the interface.
constant. The analysis of Sect. 2.1.4–2.1.6 This effect is explained by a stronger
shows that the capacitance data for surface- interaction of the ion with the charges
inactive electrolytes prove the existence of induced inside the compact-layer region,
the compact layer, thereby justifying this which have got the same sign as the
assumption. ion, see [254, 255] for a more detailed
Correspondingly, the profile of the discussion. These two branches of Wim (z)
image potential: are separated by a minimum of purely
electrostatic origin, whose depth depends
∼ (zi e)2 2z on these characteristic parameters, ε and
Wim (z) = 1−E ,
4εz εLH LH , and the species charge, zi e.
∞ A similar analysis has been carried out
E(y) ≡ 2y exp(y) z−1 e−z dz (79) for an ion near the contact between a
y
charged metal and an electrolyte solu-
is only dependent on the integral char- tion, in which the image potential pro-
acteristic of this ‘‘defective region’’, the file is also a function of the electrode
compact-layer capacitance, CH , defined by charge and the electrolyte concentration,
Eq. (20), via a characteristic length, LH Wim (z; σ , c) [254–256]. Depending on
these parameters, the profile, Wim (z), may All these results correspond to the case
be a monotonous function (increasing or that the ion is located within the diffuse
decreasing), or a curve with a minimum, layer far away from the outer Helmholtz
see, for example, the numerical illustra- plane. The conclusions change drastically
tions in Ref. [257]. for charged species inside the compact
For example, at the p.z.c. (σ = 0) the layer, in particular, for specifically ad-
shape of the profile depends on the ratio sorbed ions. This interphasial region rep-
of the Debye screening length in the bulk resents a nonuniform dielectric medium
solution (32), LD , and the product εLH whose effective dielectric constant, ε(z),
this ratio being identical to the ratio of the has got the values strongly reduced in
compact- and diffuse-layer capacitances, comparison with those in the bulk solu-
CH and CGC in Eq. (20). tion. For such systems, the interaction
For a sufficiently low electrolyte con- of the species with the metal or the
centration (so that CH CGC ), the pro- diffuse-layer ions represents a minor cor-
file within the interval, z LD , is close rection, while the dominant contribution
to its form at vanishing concentration, to the image potential is given by the
Eq. (79), in particular there are the repul- short-range interaction with the dielectric
sion and attraction branches, Eqs. (81) and medium [258, 259]. For ions of a suffi-
(78). For even larger distances from the ciently small radius, ro , one can use as a
metal surface (comparable with the Debye qualitative estimate a simple formula for
length), the interaction with the interface the electrostatic contribution to the image
is screened: potential:
(zi e)2 −2z/LD
Wim (z) = − e (82) (zi e)2 1 1
4εz Wel−stat (z) ∼
= − (85)
2ro ε(z) ε
More important changes take place in
more concentrated solutions where CH Because of lowered values of the effective
CGC (which means c 0.1 M for the dielectric constant inside the compact
Hg–water interface). The profile repre- layer (independent of the physical reason
sents a monotonically decreasing curve leading to this reduction), the term given
(pure repulsion of the ion from the in- by Eq. (85) represents a strong repulsion
terface at all distances), in accordance with of ions from this interphasial region. In
the formula: reality, one should add the effect of the ion
desolvation to enter this layer.
(zi e)2 −2z/LD
Wim (z) = e (83) This consideration gives an explanation
4εz of the existence of surface-inactive ions
Thus the repulsive behavior at shorter for which the attractive contribution due
distances is a general feature of the image to the short-range (electron exchange)
potential profile at the p.z.c., whereas, the metal surface–ion interaction is absent
large-distance variation depends on the or insufficiently strong to compensate the
ionic strength of the solution: aforementioned repulsive effect. It may
also be responsible for the strong influence
(zi e)2 εLH − LD −2z/LD of the iodide adsorption on the hydrogen
Wim (z) = e ,
4εz εLH + LD evolution, despite the absence of marked
z min(LD , εLH ) (84) specific adsorption of H3 O+ ions [260].
One can also derive analytical expres- and it leads to fundamental consequences
sions for the variation of the image for the correlation properties of this
potential for a charged species inside ensemble as well as for the form of the
the compact layer as a function of the adsorption isotherm (see Sect. 2.1.11.6).
bulk electrolyte concentration [254–256]. Intuitively, it seems to be ‘‘natural’’ to
This contribution is relatively small com- use an analog of Eq. (78) based on the same
pared with the short-range interactions, model of the ‘‘image charge’’ formed by the
but its knowledge is of importance, since metal electrons as a response to the field
it represents the change of the single-ion of the adsorbed ion:
(standard) adsorption energy with varia-
1 1
tion of the ionic strength of the solution in U o (R) = (ze)2 −
the adsorption isotherm, see Sect. 2.1.11.6. εR ε(R 2 + 4a 2 )1/2
2a 2
∼
= (ze)2 3 (86)
2.1.11.4 Interaction Between Adsorbed εR
Ions
The deviation of the adsorption isotherm at Here, a is the distance between the center
a uniform electrode surface from the linear of the adsorbed charged species and the
behavior (Henri isotherm) is related to the metal surface. Equation (86) predicts a
interaction between the adsorbed species. rapid decrease of the interaction energy
A substantiated derivation of its form can as a function of the distance between
only be made on the basis of an analysis of adsorbed ions, R, and it means that the
the statistical-mechanical properties of the ensemble of species with such interaction
whole ensemble of adsorbed ions, which, law should correspond to the one with
in turn, requires the knowledge of the short-range interactions. The uncertain
interaction potential between the ions, U , point within this approach is to choose
as a function of their distance, R, along the value of the dielectric constant of the
the surface. This quantity is defined as medium in Eq. (86), ε.
a difference between the energies of the It is tempting to identify it with the value
system, when these two ions are fixed at typical for the compact layer, which would
distance R or are very far from each other. mean that the interaction propagates along
There are numerous types of short- this interphasial region. However, it can
range interactions between adsorbed ions be easily seen from simple calculations
such as exchange-correlation effects, an that the interaction energy must diminish
overlap of their electronic cores, mutual exponentially (as a function of R), if the
polarization, interaction via the electrons interaction between the charges is realized
of the metal, interaction between their along the lines passing inside the compact
solvation shells (which may partially be layer, for example, in the model in which
retained in the adsorbed state), changing of this layer is surrounded by two conducting
the local structure of the compact layer, and media imitating the metal and the diffuse
so on. All these forces decay rapidly with layer. As a result, this model cannot explain
ion-ion distances exceeding the molecular the very large values of the interaction
scale, and the electrostatic contribution constants observed experimentally for ion
becomes dominant. A crucially important adsorption. The alternative attribution of
feature of this long-range contribution is a ε to the bulk-solvent value, also leads to a
surprisingly slow decay as a function of R, very weak interaction.
It will be seen from the discussion that state, and so on. In the simplest model of
follows that the electrostatic interaction the compact layer as a uniform dielectric
between adsorbed ions follows quite a slab of thickness zH and dielectric con-
different functional form. A detailed anal- stant εH , this parameter is given by the
ysis of this problem may be found in formula:
za ai
reviews [89, 254, 255, 261]. = (88a)
Similar to Sect. 2.1.11.2, let us consider zi zH
first the interaction of two charged species where ai is the distance between the
located at a small distance (atomic scale) adsorbed ion and the metal surface (con-
from the metal–solvent interface. Once sidered in this formula as a point charge),
again, it turns out that the functional form za and zi are the charges of the ion in
of the U (R) dependence is only influenced the adsorbed state and in the bulk solu-
by the same characteristic length, εLH , tion. A somewhat more general model of
which had already appeared in Eq. (79) for the compact layer as a nonuniform dielec-
the image potential, that is the product of tric slab whose local dielectric permittivity
the bulk dielectric constant of the solvent depends on the normal coordinate, ε(z),
and the inverse value of the compact -layer results in the formula (88b) containing the
capacitance, Eq. (80). ratio of the ‘‘effective thickness’’ for the
The binary interaction potential, U (R), overall compact layer, LH ≡ (4πCH )−1 ,
follows indeed the R −3 law within its long- and that of its ‘‘inner part’’ between the
distance limit: bulk metal and the adsorption plane, LI
2εLH 2 (again, the charge of the adsorbed ion
U (R) ∼
= (zi e)2 2 (87) is considered here as the point charge
R3
localized at the distance a from the sur-
but only for very large distances between face) [262]:
the adsorbed ions, R εLH . One should zH
remember that this characteristic distance, za LI dz
= , where LH = ,
εLH (known for each system from inde- z LH −∞ ε(z)
pendent capacitance experiments), is well ai
dz
beyond the atomic scale, for example, it LI = (88b)
−∞ ε(z)
is about 2.4 nm for any ion adsorbed at
the Hg–water interface. The amplitude in Similar expressions can easily be de-
Eq. (87) is several dozen times larger than rived for any particular model of the
the value predicted by Eq. (86). metal–solvent interphasial structure on
The only parameter in the whole U (R) the basis of the general formula for this
expression that is unknown from inde- parameter [254, 261, 263], including a spa-
pendent experiments is the multiplier tial distribution of the electronic charge of
2 , which depends on numerous factors, the ion in the adsorbed state. Generally,
like, location of the adsorbed ion within one may expect that its value should be
the compact layer, the distribution of the between 0 and 1.
dielectric properties across this layer (ana- On the other hand, this parameter, ,
logues of the ratio of the capacitances, can be found for any particular ion from
KH /KO , in the Grahame–Parsons theory, the experimental data on its adsorption,
Sect. 2.1.11.2), partial transfer of the elec- see Sect. 2.1.11.6. It means that the form
tronic charge by the ion in its adsorbed of the binary interaction potential and the
correlation properties of this ion ensemble unchanged, including the universal func-
(Sect. 2.1.11.5) can be described without tional form of U (R) with a single unknown
any unknown parameters. parameter, , a high intensity of the in-
Equation (87) only represents a limiting teraction (compared to its estimates within
behavior at very large distances between oversimplified model (86)), existence of the
adsorbed ions, R. There exists a wide ‘‘Coulombic’’ behavior of the potential (89)
range of intermediate distances in which and so on.
the interaction potential follows quite a The only qualitative modification is the
different functional law: appearance of the exponential decay at
2 distances exceeding the screening length
U (R) ∼
= (zi e)2 2,
εR within the diffuse layer, which was as-
for zH R εLH (89) sumed to be not too short (much longer
than the compact-layer thickness). As a
The crucial difference of this formula in result the interval of intermediate dis-
comparison to Eqs. (86) and (87) is a very tances in which the R −1 behavior is valid,
slow decrease of the interaction potential Eq. (89), is determined by the inequalities,
within this interval of R. The existence of zH R R∗ , where R∗ is the minimum
this intermediate asymptotic behavior of of the distance εLH , and the screening
U (R) leads to crucial consequences for the length inside the diffuse layer for the val-
correlation properties of the ion ensemble, ues of its charge and the ionic strength
Sect. 2.1.11.5. of the solution, in particular, the Debye
One can demonstrate that this large
screening length of the bulk solution for
increase of the intensity of the interaction
the zero charge of the diffuse layer.
is due to a specific distribution of the lines
The intensity of the interaction depends
of the electric field created by an adsorbed
also on parameter , whose experimen-
ion. Because of a great difference between
tally found values for the adsorption at
the effective dielectric constants in the
s,p-metals are not much smaller than 1 (see
compact layer and in the bulk solvent, the
lines do not propagate along the compact Sect. 2.1.11.6). For such systems, one can
layer. A fraction of these lines determined see that the ‘‘Bjerrum length’’, LB , where
by the factor 1 − , connect the ion with the interaction energy is equal to the ther-
the charges at the metal surface. The mal energy belongs to the aforementioned
remaining lines (their fraction being ) interval of the intermediate distances,
quit the compact layer and propagate zH LB R∗ . The importance of such a
outside of it at a distance R. The interaction sequence of characteristic distances for the
given by Eqs. (87) and (89) is determined correlation properties of the system will be
by the latter fraction of lines. seen in the next section.
A similar analysis was carried out for
the case of an electrochemical system 2.1.11.5 Statistical Mechanics of Adsorbed
in which the ion is adsorbed at the Ion Ensemble
charged metal–electrolyte solution inter- During an extended period the structure
face. The derived analytical expressions of the two-dimensional (2D) ensemble
for the interaction potential become a bit formed by adsorbed ions located in the
more complicated [89], but all the above- inner Helmholtz plane was postulated on
formulated qualitative conclusions remain intuitive grounds.
Most often, it was assumed that the ion the form of the U (R) dependence in
distribution represents a regular 2D lattice, Sect. 2.1.11.4 confirm the existence of such
hexagonal or square, owing to the repulsive a ‘‘cut-off-disk’’ but its size does not depend
electrostatic forces between the ions. How- on and is markedly different from the
ever, the theoretical studies (confirmed by postulated values.
experimental data) related to a similar 2D These assumptions on the distribution
one-component system of charged species, of ions in the adsorption plane were com-
electrons at the surface of liquid helium, bined with some expression for the binary
have shown that such a regular lattice interaction potential, U (R) (in most cases,
becomes unstable with respect to certain given by Eq. (86)), to get the contribution
collective motions (dislocation formation, from the interactions into the chemical po-
etc) at extremely low temperatures, that is, tential of an adsorbed ion. However, the
under the conditions in which the average final results of such treatment can hardly
interaction energy per ion is still much be considered as reliable.
larger than the thermal energy (contrary A substantiated analysis must be based
to the standard stability criterion for 3D on the dependence of the binary interac-
lattices). The application of this stability tion energy on the distance between ions,
condition for 2D lattices to the ensemble U (R), discussed in Sect. 2.1.11.4. Certain
of adsorbed ions shows that the hypothet- specific features of this system compared
ical regular lattice never formed in these with usual 3D ones originate immediately
systems by electrostatic repulsion within from its two-dimensional character [264,
the experimental temperature range, since 265], e.g., the absence of a long-range crys-
the 2D density is limited by the repulsion talline order at any finite temperature, slow
between the cores of these ions. decay of the autocorrelation function of the
Another hypothesis assumed that the species velocity in the long time limit, and
binary interaction between ions is weak, so forth.
so that the system could be considered Additional effects arise from the modifi-
as close to a 2D ideal gas. However, cation of the form of the binary interaction
the analysis of the function, U (R), has energy, U (R), for the ions at the interface,
revealed that this energy exceeds the compared to the one in the bulk solution.
thermal one up to the distance about One of the most important effects is the
RB ∼ 1 nm, that is, it is strong. great intensity of this interaction, for ex-
The third popular approach was based ample, Eqs. (87) and (89), which may in
on a cut-off-disk model. According to it certain cases be stronger than the same
the electrostatic repulsion between the characteristic in the bulk phase (dielec-
ions leads to the formation of an area tric or solution). Another crucial feature
around each ion practically free from is a slow decrease of this interaction,
other adsorbed ions. The size of this area U ∼ R −1 , in the intermediate range of
was identified with the average distance the distances, zH R R∗ , with a much
between the ions, so that it was dependent steeper diminution outside at greater dis-
on the 2D density, that is, on the ion tances, Eqs. (89) and (87), correspondingly.
adsorption, (as in the previous Sects. This combination makes it impossible to
2.1.11.2–2.1.11.4, this symbol is used for apply standard approaches of statistical
the amount of specifically adsorbed ions, mechanics derived separately for systems
unlike in Sect. 2.1.11.1). The results on with a strong short-range or weak long-range
binary interaction potential. Besides, one interval are close to those of a 2D one-
must take into account the screening prop- component plasma, despite the absence
erties of the diffuse layer, which depend of the R −1 behavior of U (R) at very long
markedly on the electrode charge and ion distances. It would be of great interest in
force of the solution. the future to perform experimental studies
The statistical-mechanical analysis of of these predictions with the use of small-
correlation properties of such an ion angle scattering techniques, in particular
ensemble at the uniform metal surface of the anomalous behavior at small Fourier
was performed in Ref. [266], see also vectors owing to the plasma correlations.
reviews [254, 255, 261, 267]. There are At even higher system density, ∼ B ,
two characteristic values of the 2D den- the interaction energy between a pair of
sity, = ∗ ∼ (πLB R∗ )−1 and = B ∼ adsorbed ions located at the average dis-
(πL2B )−1 corresponding to the intervals: tance becomes comparable to the thermal
0.05–0.5 nm−2 and 0.5–2 nm−2 , respec- energy, kB T . Then the properties of the
tively. system correspond to those of a 2D ir-
Owing to the rapid decrease of the regular condensed state. Ultimately, when
interaction potential at large distances, the approaching the monolayer coverage de-
system behaves like a 2D dilute gas with termined by the short-range repulsion of
a weak interaction between the species ion cores or/and their solvating shells, a
if the surface coverage is extremely low, typical 2D ‘‘solid’’ is formed, which rep-
∗ . The virial expansion may be used resents a crystal lattice possessing a quasi
to determine all energetic and correlation long-range ordering.
characteristics. However, the coefficients
of this virial expansion turned out to be 2.1.11.6 Ion Adsorption Isotherm
very large (compared with their estimates In a surprising contrast to this complicated
with the use of the Bjerrum length). As picture for the correlation properties of the
a result a noticeable repulsion between ensemble of adsorbed ions (Sect. 2.1.11.5),
adsorbed ions in the intermediate distance the same analysis has resulted in a very
range takes place already at very low ion simple formula for the average interaction
densities, ∼ ∗ . energy, that is, for the contribution to the
Within this transitional interval of den- chemical potential of adsorbed ions, µint ,
sities, ∼ ∗ , the correlation properties which is given as a linear function of the
of the system are subject to a gradual coverage over the whole range of surface
change. For lower coverages, the princi- coverages:
pal contribution to the average interaction
energy is due to pair ‘‘collisions’’ of ions. µads = kB T ln + µint + constant,
On the contrary, at higher densities, µint ∼
= 2vkB T (90)
∗ , new collective plasma-like excitations
arise, leading to the screening of the inter- The coefficient in this linear relation serves
actions at the distance R . This new 2D as an analogue of the attraction con-
screening length depends on the density stant in the Frumkin isotherm (with the
of the system, R ∼ (LB )−1 , that is, the minus sign) or the constant B in the
increase of the coverage results in a more Grahame–Parsons isotherm (75). How-
effective screening. Thus the correlation ever, unlike these models, in which this
properties of this ensemble within this term as well as its dependence on the
electrical variable (see following text) were electrode potential, the adsorption energy
introduced phenomenologically, Eq. (90) and the attraction ‘‘constant’’ as functions
was derived on the basis of the binary of this variable being modeled by the sim-
interaction potential, U (R), so that the plest approximation (linear or quadratic).
constant, v, is related immediately to this In the theory of Grahame and Parsons,
characteristic: Eq. (75), it is the electrode charge.
∞ According to the derivation of isotherm
−U (R)
v=π 1 − exp RdR (93), the ions are transferred into the
LB kT adsorbed state at a fixed value of the
(91)
electrode potential, E, so that the parameters
The introduction of the expression for
of the isotherm, Wads and v, depend on
the binary interaction potential discussed
this variable, as well as on the overall bulk
in Sect. 2.1.11.4 into this equation leads
concentration, c, in the case of the mixed
to the formula (92) for the ‘‘repulsion
electrolyte:
constant’’, v:
◦
Wads = WEDL + Wim c + Wads ,
zi2 e2
v = 2 v̄, v̄ = (92) WEDL = zi e(E − ϕGC )(1 − )
2kB T (CH + CGC )
In this formula the coefficient, v, is + zi eϕGC (94)
represented as a product of the factor
In other words, the single-ion adsorp-
2 introduced in Sect. 2.1.11.4, and a
tion energy is represented as the sum
function of the compact- and diffuse-layer
of the EDL term (compare Eq. (74)),
capacitances (the GC expression for the the image potential as a function of
latter, Eq. (21a), has been used in this the electrolyte concentration, Wim c =
derivation). Wim (z; σ, c) − Wim (z; σ, 0) and the stan-
The combination of Eq. (70a) and (90) dard adsorption energy, Wads ◦ . The an-
results in the adsorption isotherm (93) for alytical expression for Wim c is given in
a mixed electrolyte solution [267, 268]: Ref. [256, 267, 268]. All these expressions
Wads including the capacitances in Eq. (92) and
ln(mc) = + ln + 2v (93) the diffuse-layer potential difference, ϕGC ,
kB T
in Eq. (94) depend on the electrode po-
where Wads is the single-ion adsorption tential; these dependences correspond to
energy corresponding to the energy of ion zero specific adsorption so that they should
transfer to the surface in the absence of be taken from experimental data for the
other adsorbed ions. surface-inactive electrolyte.
Similar to several other models, isotherm After this treatment the isotherm only
(93) has the virial form, with a linear de- contains two microscopic parameters of
pendence in the interaction term. The the adsorbed ion, whose values cannot
essential difference between them is re- be measured without experiments on
lated to another driving parameter of the its specific adsorption, Wads ◦ and ,
system, which is called the electrical vari- which may generally be dependent on
able and which characterizes the state the electrode potential. The theoretical
of the electrode while the bulk concen- expression for (in particular, its simplest
tration of ‘‘active’’ species changes. The form (88a)) shows that its value is expected
Frumkin isotherm is valid for a fixed to be between 0 and 1; it should diminish,
if the ion is located closer to the electrode determination of experimental values of

surface or/and at a greater charge transfer on the basis of the capacitance (or elec-
from the ion to the metal (i.e. if the trode charge, or interfacial tension for liq-
adsorption is stronger). uid electrodes) measurements requires to
Thus the derived isotherm has got a perform the operation of the graphical dif-
very simple final form, with a minimum ferentiation of the curves for the interfacial
number of fitting parameters, less than in energy, γ (E, m, c) (for mixed electrolyte so-
the other approaches used to interpret data lutions), with respect to the solution com-
for ionic adsorption. position, ln m or ln c (see Sect. 2.1.11.1).
If the adsorption is measured as a func- This graphical differentiation can be
tion of the bulk concentration for different avoided, owing to the particular form of
values of the electrode potential, (c, E), the isotherm (93), which has allowed one
then the parameters in Eq. (93), Wads and to derive an analytical relation between γ
v, can be easily found by fitting for each and m [261, 267, 269]. Then the fitting first
potential, then their values may be used to provides the values of Wads and v, and
determine the fundamental characteristics from that Wads ◦ and are obtained.
of the adsorbed ion, Wads ◦ and , without This method has been successfully
any further uncertainties. However, the applied to analyze the ion adsorption
v
[cm2 µC −1]
0.175
0.15
0.1 0 − 0.1
ϕ
[V]
Fig. 22 Experimental values of the ‘‘repulsion constant’’, v, in
isotherm (93) for iodide adsorption at Bi-water interface (points are
shown with their dispersion) as a function of the electrode potential
in the rational scale, ϕ ≡ E − Eσ =0 . The theoretical curve (solid
line) was calculated from Eq. (92) for = 0.71 with the use of
experimental data for the compact- and diffuse-layer capacitances
of the same interface in a surface-inactive electrolyte solution.
from mixed electrolyte solutions in the Grahame–Parsons isotherm to interpret

systems [261, 267, 270, 271], namely, these data was unsuccessful, despite a
Bi/H2 O−I− , InGa/H2 O−I− , InGa/H2 O greater number of fitting parameters. The
−Br− , Hg/ethylene glycole−Cs+ , Hg/ analysis with the use of isotherm (93)
propylene glycole−Cs+ , Hg/H2 O−Cl− , gave the values of the intermediate pa-
and so forth. These systems were chosen rameters, v and Wads , which are shown
to verify the theoretical predictions for the in these figures as a function of the
form of the adsorption isotherm as well electrode potential in the rational scale,
as for the dependence of the values of its ϕ ≡ E − Eσ =0 . The experimental disper-
parameters, Wads ◦ and , on the properties sion in the thus obtained values of v
of the metal, solvent, and adsorbed ion is indicated for each point. It increases
(both cations and anions). strongly at negative potentials since the
The results of this treatment of the data coverages are much lower in this range
for the first of these systems are pre- and the precision of the determination of
sented in Figs. 22 and 23. One should the interaction parameter becomes much
note that the earlier application of the lower.
Wads /kBT Wadso + Wimc/kBT
− 10
−5
0.1 − 0.1 − 0.3
ϕ
[V]
Fig. 23 Dependences of the total single-ion adsorption energy,
Wads /kB T (curve 1), and of the sum of the ‘‘specific’’ and image
contributions, (Wads ◦ + Wim c )/kB T(curve 2) for the system,
Bi/H2 O−I− , as a function of the electrode potential,
ϕ ≡ E − Eσ =0 . The values of Wads were found from the fitting of
isotherm (93) to experimental adsorption data for each electrode
potential. Then the values of Wads ◦ + Wim c were obtained with
the use of Eq. (94) and the data for surface-inactive electrolyte
solutions. The EDL term, WEDL , was calculated for the same value
of the adsorption parameter, = 0.71, see Fig. 22.
Then the value of was determined chemical bonds. It confirms the principal
from the experimental point for the most conclusion of Sect. 2.1.11.4 concerning the
positive potential, with the use of the value binary interaction potential, namely, that
of v̄ found from Eq. (92) and experimental the dominant role is played by the inter-
data for the Bi–water interface in the action along the field lines passing across
background electrolyte solution. This value the diffuse layer, rather than those along the
was used to calculate, already without any compact layer.
unknown parameter, the whole theoretical The values of v for each particular system
curve for v (Fig. 22) as well as for demonstrate a marked variation with the
Wads ◦ + Wim c (Fig. 23) as functions of ϕ. potential, which can be described by
One can notice that a constant value of Eq. (92) with a constant value of , that is,
is sufficient to reach an agreement with all this dependence on the electrode potential
experimental points for v, within their dis- is given by the theoretical expression,
persion, which is very small for the positive v̄(E, c), which does not contain any
potential range. Similarly the single-ion ad- unknown parameter.
sorption energy (after removal of its EDL As another success of the approach, one
component) is practically constant within may notice a large variation between the
the same range of potentials. Thus, in values of v for different systems, up to
accordance with expectations, the micro- a factor of 5, while the estimates of the
scopic characteristics of the adsorbed ion, principal microscopic parameter of the
I− , do not change markedly in this region adsorbed ion, , turned out to vary within
of electrode potentials, despite the change a much more narrow range, from 0.55 to
of the electric field created by the charged 1.0. Moreover, all these values fall inside
metal surface. the expected interval < 1. If one uses
The data at high negative potentials formula (88a) or (88b) to get a simple
show a significant reduction of the ad- qualitative estimate of this parameter, one
sorption energy with the potential. It also may conclude on the basis of these results
corresponds to the qualitative expectations that all these ions are located not too close
concerning the weaker bonding of an an- to the metal surface, and that they retain
ion at negatively charged metal surface. most (if not all) of their electric charges in
Similar results have been obtained for the adsorbed state, za ∼ zi .
the other systems listed in the preceding Within the framework of the hypothe-
text. In particular, the values of the ‘‘resis that no charge transfer takes place in
pulsion constant’’, v, were found within the course of the ion transition into the
the limits of 0.6 and 3.5 nm2 per ad- adsorbed state, the value of for each sys-
sorbed ion. These results represent an tem would allow one to determine the ratio
additional confirmation of the predic- of the capacitance between the metal sur-
tion of strong intensity and long-range face and the adsorption plane to the overall
of the binary interaction potential dis- compact-layer capacitance, Eq. (88b), sim-
cussed in Sect. 2.1.11.4. One should keep ilar to the Grahame–Parsons treatment
in mind that the attempt to interpret the in Sect. 2.1.11.2. (the Helmholtz–Stern
same experimental data with the use of model of the compact layer as a uniform
the ‘‘single-image’’ expression (86) would dielectric slab does not correspond to the
result in the values of the ion-metal dis- modern view of the interfacial structure,
tance, ai , well beyond the usual length of see previous sections, so that it is hardly
possible to use Eq. (88a) to estimate the • for the same electrode, InGa, the
real distance between the adsorbed ion and stronger adsorbed anion, iodide, has
the metal surface.) However, these adsorp- got a smaller value of this parameter,
tion data cannot give any argument in favor than bromide;
of or against such a partial charge transfer • in the case of a weakly adsorbed cation,
in these systems, in other words, the re- cesium, there is no significant differ-
duced value of this adsorption parameter, ence between ethylene and propylene
< 1, represents generally a combined carbonate solutions.
effect of partial charge transfer from the
adsorbed ion to the metal and of the loca- Despite the success in the interpretation
tion of the adsorbed ion inside the compact of these experimental data for small-size
layer, see Eq. (88b) as an illustration. inorganic ions, one should keep in mind
The only exception is the case of the the need for the further development
Hg–H2 O-Cl− interface [267, 271] in which of the theory, in particular to include
the value was close to 1. It allows one the effect of the gradual change of the
to conclude on the absence of a notice- dielectric properties of the compact layer
able charge loss by this ion as well as on with the increase of the surface coverage
close values of the characteristic lengths, by adsorbed ions.
LH and LI in Eq. (88b), that is, of the corre-
sponding capacitances. However, the latter 2.1.12
does not mean that the physical location Electric Field-Induced Modifications of the
Surface Structure
of the adsorbed ion is close to the outer
Helmholtz plane. There are various indi-
Charging the electrode may not only affect
rect indications (in particular, presented in
the polarization of molecular dipoles on
previous sections) that the effective dielec-
the surface, deform the surface electronic
tric permittivity, ε(z), is quite nonuniform
profile, and induce the adsorption of ions.
across the ‘‘compact layer’’, its lower values
It can also induce surface reconstruction
being within the ‘‘tail’’ of the metal elec-
and roughening transitions, that is, modify
trons while the external part of the compact
the atomic structure of the metal surface.
layer possesses a much higher permittivity
In this section, we will briefly describe the
(but still much lower than the one in the
principles of the statistical theory of this
bulk solvent). Then the adsorbed ion can
phenomenon [272–274].
be localized at the distance of 0.1 to 0.2 nm
In ultrahigh vacuum (u.h.v.), clean low-
from the ion skeleton of the metal, that is,
index single-crystal surfaces of gold and
deeply inside the compact layer having a
platinum are known to be reconstructed
thickness of about 0.3 nm, Eq. (12), but
at low temperatures and unreconstructed
can be close to 1.
at temperatures above the critical tem-
The comparison of the values for
perature [275]. Transition from the recon-
different systems have shown expected
structed to the high-temperature unre-
tendencies:
constructed state is called deconstruction.
• this parameter for adsorbed iodide is This term does not necessarily imply that
greater at the Bi electrode than at InGa the surface structure at high temperature
(the adsorption energy is greater for resumes the full 1 × 1 unreconstructed
InGa); symmetry. Indeed the deconstruction is
Fig. 24 Top (upper) and side (lower)

views on differently reconstructed (110)
surfaces. The numbers indicate
corresponding spin assignments of
Ising model.
1 1 1 1 1 1 1 1 1 emersion of the electrode into u.h.v.,

in situ x-ray diffraction (XRD), electrore-
flectance, second harmonic generation,
and scanning tunneling microscopy (for
review see [276]). For instance, the 5 × 20
(1 × 1)
reconstructed Au(100) and the 1 × 23 re-
constructed Au(111) surfaces are stable at
sufficiently negative electrode potentials,
but at positive potentials the reconstruc-
tion is lifted. It was noticed that anion
adsorption definitely lifts the reconstruc-
1 −1 1 −1 1 −1 1 −1 1 tion, but it was hard to distinguish it from
the mere effect of the positive charging, be-
cause ion adsorption is also a field-induced
effect (see Sect. 2.1.11). Somehow at the
metal electrolyte interface there is no need
(1 × 2) to go to high temperatures to deconstruct
the surface. The reconstruction takes place
by negative charging at room temperature.
Surfaces of Au(110) and Pt(110) in u.h.v.
reconstruct to 1 × 2 or a mixture of 1 × 2
and 1 × 3 missing row structures (Fig. 24).
In the electrochemical environment, the
reconstructed state is stable for a range of
electrode potentials smaller than a positive
critical value. For potentials more positive
than the critical value, the reconstruc-
(1 × 3)
tion is lifted. Potential variation reversibly
changes the structure between the recon-
accompanied by the appearance of various structed and unreconstructed states. We
kinds of imperfections, and the transition will dwell on the missing row recon-
itself is not abrupt but passes through struction in the following text, the easiest
the phases of disorder, incommensurate example for tutorial purposes, since it was
superstructures and roughness. rationalized by a relatively simple theory.
Surface deconstruction has been dis-
covered on single-crystalline electrodes 2.1.12.1 Electrochemical Missing Row
in contact with electrolytes with the Reconstruction: Facts and Speculations
help of different experimental methods. Gao and coworkers [277] obtained scan-
These were the combination of cyclic ning tunnelling microscopy (STM) images
voltammetry with ex situ analysis after of Au (110) in 0.1 M HClO4 aqueous
solutions. At potentials more negative than emerges by removing atoms from holes,
−0.3 V versus. s.c.e., they report the ap- which grow into missing rows, and redepo-
pearance of 1 × 2 reconstruction (with sition of these atoms into the added rows.
small impurity of 1 × 3 and 1 × 4 blocks). Thus qualitatively the STM results of
Unreconstructed patterns emerge at po- Behm’s and Weaver’s groups agree with
tentials larger than zero; Hamelin [278] each other. Simultaneously Ocko and
reported a threshold value of 0.1 V. coworkers [280] examined the same sur-
Since the p.z.c. for this system is about faces in 0.1 M perchloric acid and alkali
−0.02 V [278], at zero charges the surface halide salts solutions by a ‘‘more objective’’
is not reconstructed in contrast to the same in situ synchrotron diffraction method. For
surface in u.h.v. at room temperature. The a meticulous discussion of this work, see
authors of Ref. [277] speculate, that this the review [273]; here we just stress that
might be due to some adsorption of cations the results, although less unambiguous in
used to stabilize the reconstructed surface a variety of details, were qualitative similar
in u.h.v., which plays a role similar to neg- to the results of STM studies.
ative charging. Between −0.3 and 0 V the
interpretation of the images is unclear. 2.1.12.2 Driving Force for Reconstruction
The same surfaces were studied by The structure and the shape of the metal
Magnussen and coworkers [279] in 0.1 M crystal are determined by the net bal-
acid solutions: HClO4 , H2 SO4 – pure and ance of forces. In the bulk of the noble
with traces of Cl− . The Cl− -free so- metal, the stabilizing force is the com-
lutions showed no dependence on the pressive stress of s-p electrons balanced
nature of electrolyte, but even small ad- by repulsive interaction of the filled d-
ditions of Cl− lifted reconstruction. The shells. The s-p electrons play the role of
reconstructed state was observed at poten- a ‘‘glue’’, which is as follows. They always
tials from 0.05 down to −0.3 V, although ‘‘like’’ to spread out along the surface.
less ordered than in u.h.v. A window However, at the (110) surface the external
of disordered and roughened states was filed of the skeleton ions is very inho-
found between 0.05 and 0.25 V, while at mogeneous. The electrons will have to
more positive potentials the 1 × 1 unrecon- form rather inhomogeneous distributions,
structed phase prevails. For slow variation which are energetically costly. This forces
of potential, the authors were able to follow surface atoms to rearrange to form tightly-
the initial steps of reconstruction. They re- packed facets that mimic the lower energy
ported the transformation beginning near (111) structure. On (110) surfaces, this is
0.2 V via a slight roughening of the step achieved by missing row (Fig. 24) recon-
edges; further potential decrease induced struction [281, 282]. The force that pushes
roughening and zigzag step structure of the surface to reconstruct is stronger when
the reconstructed surface with a simul- more electrons are present in the surface
taneous growth of anisotropic holes and layer. This happens when the surface is
islands one atom wide but up to 10 nm negatively charged.
long in the 110 direction! When averaged The field-induced reconstruction was
over a big terrace, the concentration of studied in the ground state Kohn–Sham
islands was found to be nearly the same electron density functional simulations for
as the concentration of holes, in accor- fixed atomic structures of the surface lay-
dance with the idea that the reconstruction ers [283, 284]. Their energies are compared
in order to find that configuration which an ideally 1 × 2 reconstructed surface,

has, at a given surface charge, the lower the s = 1 and s = −1 rows, oriented in
energy. It appears that in a range of nega- y-direction (denoted as [11̄0]), alternate in
tive charges the missing row configuration x-direction, the one perpendicular to the
has lower energy at T = 0, and one may ex- rows ([001]). For an ideal 1 × 1 surface,
pect that the surface will be reconstructed. all spins have s = 1 (or s = −1, which
Physically, negative charging increases the is equivalent). Let Jy and Jx and be the
density of electrons at the surface, which ‘‘spin interactions’’ of neighboring sites
produces a tension that is relaxed by re- along the rows and perpendicular to the
construction, because the missing row rows. The Ising–Hamiltonian then takes
structures allow for tighter packing of the form
nearest neighbor atoms. Although no such
calculations have been performed for the H=− (Jx si,j si+1,j + Jy si,j si,j +1 )
metal–electrolyte interface, in u.h.v. the i,j
comparison of 1 × 1 and 1 × 2 structures (97)
is available for (110) surfaces of Ag and where i, j are the indices of the lattice sites
Au. Subject to these calculations, one may determining their coordinates xi and yj .
approximate the difference in the ground Jy > 0 for an anisotropic surface, and we
state surface energies (per 1 × 1 surface will consider it to be constant. As for Jx, it
unit cell) could be positive or negative. If Jx > 0,
the ground state is ‘‘ferromagnetically’’
E(σ ) = E1×2 (σ ) − E1×1 (σ ) (95) ordered; if Jx < 0, it is ‘‘antiferromag-
netically’’ ordered, using the lexicon of
as
magnetic spin analogies. The transition be-
E(σ ) ≈ A(σ − σ0 ) (96)
tween these two structures at T = 0 takes
at least for σ close to the σ0 , the point where place when Jx changes the sign.
E1×2 (σ ) and E1×1 (σ ) curves intersect; for If we know E(σ ), it is easy to calculate
Au(110), σ0 0. A is a parameter, which Jx . Indeed the ground state energies of the
we do not know for metal–electrolyte inter- two configurations calculated per unit cell
face, but taking into account the u.h.v. den- of the reconstructed surface are E1×2 =
sity functional simulation, we may expect E0 + Jx − Jy and E1×1 = E0 − Jx − Jy ,
it to be no greater than 0.01 (eV cm2 ) µC−1 . where E0 is the part of the energy that
does not depend on the spin state. Hence,
2.1.12.3 Ising Model. Order–Disorder E
Transitions Jx = (98)
2
Kornyshev and Vilfan [272–274] suggested
the simplest way of describing the tran- Jy can also be obtained from the
sitions between 1 × 1 and 1 × 2 states, ground state calculations, since Jy = E0 −
based on an anisotropic 2D Ising model, 1/2(E1×2 + E1×1 ). However, for further
and derived a closed expression for the consideration, we will simply treat it as a
temperature-surface charge phase dia- scaling parameter, whose charge depen-
gram. dence is of minor importance.
In this model an occupied topmost At finite temperatures, we may use
site is described by ‘‘spin’’ s = 1 and the well-known equation on the critical
an empty site by s = −1 (Fig. 24). For temperature of the anisotropic 2D Ising
model [285]: The phase diagram corresponding to

this function is shown in Fig. 25. This
|Jx (σ )| Jy
sinh sinh = 1 (99) diagram tells us that moving at a given tem-
kB T kB T perature from negative to positive charges,
Since, Jx is σ -dependent, this determines the system undergoes first a transition
a critical line on the charge-temperature from reconstructed to a disordered state
plane. Solving Eq. (99) for Jx and using and then from a disordered state to the
Eq. (98), we obtain ordered unreconstructed state. The width
of the disordered region follows from
|E(σ )| kB T Eq. (103):
=F (100)
Jy Jy

where kB T Jy
σ = 8 exp − (104)
A kB T
1
F (x) ≡ x ln cosh (101)
x
Taking as an estimate A ∼ 0.01 eV cm2
Using, finally, Eq. (96), we obtain the equa- µC−1 and Jy ∼ 4kB T , we get σ ∼
tion for the critical line that determines the 0.5 µC cm−2 , but depending on the pa-
phase diagram: rameters, this value may be considerably
larger. The origin of the disordered phase
sgn(σ − σ0 ) 2kB T
σ = σ0 + F (102) is clear. In order to have transitions, Jx
a Jy must change sign at σ0 , but then |Jx |
where must be small near σ0 , and there it cannot
A compete with the disordering effect of the
a= (103)
Jy temperature.
1.0
D
k BT Jy−1
0.5
(1 × 2) (1 × 1)
0.0
−1.0 − 0.5 0.0 0.5 1.0
a (σ − σ 0) Jy−1
Fig. 25 Ising model phase diagram. The reconstructed (1 × 2) and
unreconstructed (1 × 1) regions are separated by Ising disordered
phase (D).
2.1.12.4 Roughening Transitions is rough, because there is a random

While the Ising model of field-induced sequence of up-and-down steps. Strong
surface reconstruction is very conve- anisotropy simplifies the calculation, be-
nient for tutorial purposes, it cannot cause the steps and walls are oriented
describe roughening transitions that take predominantly along the top atomic rows.
place in the course of reconstruction- One can imagine a step as a trajectory
deconstruction transition. In order to take of a fermion moving in one-dimensional
them into account, a more involved the- space, our x-axis, while the y-axis plays
ory [272–274] was developed, based on the the role of time. The ground state energy
Vilfan and Villain approach to roughening of the corresponding quantum Hamilto-
transitions [286]. Kornyshev and Vilfan ob- nian is then equal to the free energy of
tained the phase diagram by studying the the steps. In the flat phase the free en-
excitation lines – Ising walls and roughen- ergy is positive and infinitely long steps
ing steps that break the long-range order. cannot be thermally excited. In the rough
Each Ising domain wall is composed of phase the free energy of a single step
two oppositely oriented steps, so that at would be negative and the steps become
both sides of the Ising domain wall the thermodynamically stable. The steps re-
surface heights are the same. As long as pel each other and at the roughening
opposite steps are bound into pairs, the transition the density of steps is infinites-
surface behaves as flat. When the steps imally small. Therefore it is enough to
are no longer bound into pairs, thermal consider only one step, whose free energy
excitations causes the surface to become is vanishing at the roughening transition
rough. Below the roughening transition, temperature.
the surface can be flat but with Ising Kornyshev and Vilfan considered the
disorder whereas above the transition it statistical thermodynamics of such steps
1.0
r
R
D
r
i
k BT Jy−1
0.5
(1 × 2) (1 × 1)
0.0
− 0.50 − 0.25 0.00 0.25 0.50
a (σ − σ 0) W0−1
Fig. 26 Phase diagram of the Kornyshev and Vilfan theory
exhibits a line of an Ising order-disorder transition and two lines
of roughening transitions, with disordered, D, and rough, R,
phases in between.
based on a free Fermion analogy, having • Oil–water interface [289, 290], impor-
again utilized the relationship of the tant for oil extraction and catalysis.
charge dependence of the basic param-
eters of the theory with E. Not going The size of this chapter will not allow us to
any further into the details of the theory consider all these systems. We will there-
(for tutorial presentation see the review fore discuss the last one, well studied in
article [273]), we draw in Fig. 26 the re- electrochemistry, that is the interface be-
sulting phase diagram. While the Ising tween two immiscible electrolytes (ITIES).
transition has a second order, the roughen- This is a biomimetic system and a model
ing transitions presumably are of Koster- system for reaction kinetics, but also an
actual medium for phase transfer catal-
litz–Thouless universality class (see the
ysis. In this system, one of the solvents
discussion in Ref [273]).
is usually water, and the other one is a
In Ref. [273] the role of field-induced
low polar organic liquid (oil), such as, for
adsorption of anions together with a sim-
example, nitrobenzene or dichloroethane.
ple model of its incorporation into the
This interface separates ‘‘hydrophilic’’ and
theory is thoroughly discussed. Together
‘‘hydrophobic ions’’, such as simple alkali-
with the analysis of the possible effects
halides and organic ions, respectively.
due to the presence of water [273], all When two such salts are dissolved in this
this provides a framework for under- system, one composed of hydrophilic and
standing the interesting phenomenon of the other one of hydrophobic ions, they
electrochemically induced missing row form back-to-back EDLs and the interface
reconstruction. can be polarized [289, 290]; standard elec-
trochemical methods can be used for the
2.1.13
characterization of this interface. The re-
Soft Electrochemical Interfaces
sulting electric field across the interface
affects a set of phenomena that occur at
Electrochemistry of soft interfaces is a the interface. A controllable variation of
rapidly progressing interdisciplinary field. the potential drop across this interface can
The most important systems here are as shed extra light on these phenomena.
follows: First models of liquid–liquid interfaces
• Polymer electrolyte–aqueous electro- in electrochemistry treated them as flat
lyte interfaces in the porous interior and sharp [289, 290]. The nonlinear capac-
of proton conducting or ion exchange itance due to two back-to-back ionic double
membranes for separation, desalina- layers at such an interface is given simply
tion, and fuel-cell applications [287]. by [289]
• Interfaces between aqueous phase and 1 1 1
= + (105)
the volumes confined by amphiphilic CGC C1 C2
molecules [288]. In vitro, these refer (1/2)
to lipid vesicles and micelles, lipid εi κi 4π 2 σ 2
Ci = 1+ , i = 1, 2
lamellae, cubic and hexagonal phases, 4π (εi κi )2
Langmuir–Blodgett (LB) films, which (106)
are important in colloid science and in where σ is the surface density of the charge
extraction technology. In vivo, these are accumulated in each double layer per unit
the surfaces of biological membranes. area of a flat interface.
However, these interfaces are known to that the liquid–liquid interface is

be never ideally flat because of the thermal typically corrugated. The corrugation
excitation of capillary waves. Depending originates from thermal fluctuations of
on the interfacial tension between the the interface, known as capillary waves.
two liquids, the interface between two These fluctuations are described through
solvents may occur to be somewhat diffuse. the dependence of the local height of
Stronger effect of diffuseness may come the interface, z,on the lateral coordinate
from the penetration of ‘‘hydrophobic’’ R = (x, y), where the local height z = ξ(R)
ions into water and ‘‘hydrophilic’’ ions is determined relative to the reference
into the organic phase, if the free energies plane, z = 0, where the average value of
of transfer of these ions between the the profile is zero (< ξ(R) >= 0).
two solvents are not much larger than The spectrum of capillary waves can be
kB T . The overlap of the two space charge calculated starting with the functional of
regions influences the interfacial double- the free energy of two contacting liquids
layer capacitance. The two effects, the
diffuseness of the interface and capillary F = F0 + Fcw [ξ(R)] (108)
waves are generally not unrelated. The
where F0 is the free energy of the system
short wave length excitations trigger local
with a flat uncharged interface, while the
protrusions that contribute to the interface
second term stands for the excess energy
diffuseness, but the ion interpenetration
due to surface corrugation:
reduces surface tension and affects long-

range capillary waves. A rough estimate 1
for the ‘‘width of the interface’’, d, is often Fcw [ξ(R)] = dRγ [(∇ξ(R))2
2
used [291–293], 2 2
+ kgr ξ (R)] (109)
2 2 2
d = d0 + dcw (107)
Here, γ is the interfacial surface tension,
2 = ρg/γ is the small wave-vector grav-
kgr
where d0 and dcw are the intrinsic and cap-
illary wave contributions to the interface itational cutoff, where g is the gravitational
diffuseness. There is no unified theory acceleration, and ρ > 0 is the density
that involves both phenomena; it is rather difference between the two liquids. After
a subject for molecular dynamic simula- transformation to Fourier space,
tion [294]. However, separately each effect
has been rationalized [295, 296]. In the ξ(k) = dRξ(R) exp[−ikR] (110)
following text, we summarize the main
results of these papers, in which one can the free energy becomes a quadratic
read more about the history of the question functional of the fluctuating variable ξ(k),
and the previous achievements.
γ dk
Fcw [ξ(k)] = ξ(k)ξ(−k)
2.1.13.1 Capillary Waves at ITIES. Field 2 (2π)2
Effect on Capillary Waves and Capillary × (k 2 + kgr
2
) (111)
Waves Effect on Capacitance
Measurements of optical [297, 298] Note that this approximation is valid,
and neutron reflectivity [299] and MD strictly speaking, only if the amplitudes of
simulations (e.g. Ref. [294, 300]) show capillary waves are small (i.e. when for each
Fourier component of surface fluctuation correlation spectrum than the integral

the inequality |ξ(k)| < 2π/|k| holds). characteristics of fluctuations, dcw . Both
Further treatment is then trivial. For the experiments and MD [281, 284–287, 292],
Fourier transform of the height–height show that the macroscopic capillary wave
correlation function theory works surprisingly well down to the
nanometer range.
g(k) = d2 Rξ(R)ξ(0) exp(−ikR) Time-resolved quasi-elastic laser scatter-
(112) ing (QELS) experiments [306, 307] allow
with the help of the equipartition theorem the measurement of the dispersion law
one obtains [301] (the relation between the frequency of
capillary waves, ω, and their wavelength,
kB T 2π/k). To derive the dispersion law theo-
g(k) = (113)
γ (k 2 + kgr
2) retically, one must consider the dynamics
of the two degrees of freedom – the posi-
The mean square height of roughness, tion of the surface, z = ξ(x, y, t), and the
that is, the capillary wave contribution to fluid velocity. In the limit of small viscosity,
the width of the interface, is expressed such a theory [291] gives
through the correlation function as
ω2 − 2iω(k)k 2 − ωc2 (k) = 0 (116)
1 kmax
2
dcw = dkkg(k) (114) where
2π 0
k
For g(k) given by Eq. (113), ωc2 (k) = γ (k 2 + kgr
2
),
ρ1 + ρ2

2
dcw =
kB T kmax η1 + η2
2πγ
ln (115) (k) = 2k 2 (117)
kgr ρ1 + ρ2
The upper wave-vector cutoff, kmax , is in- ρ1 , ρ2 and η1 , η2 are the densities and
troduced to eliminate the divergence of the viscosities of the two liquids in contact. In
integral, which takes place for g(k) given contrast to neutron and x-ray scattering ex-
by Eq. (113), because in reality g(k) de- periments that allow the study of capillary
creases faster at large k. Various estimates fluctuations down to the nanometer scale,
of kmax −1 (proportional to the root mean the QELS provides information on long
square height [293], to the characteristic wavelength fluctuations only, with wave
smearing of the interface [302], or to the lengths and frequencies of the order of
largest bulk correlation length of the con- 6.6 × 10−3 cm and 10 kHz, respectively.
tacting liquids [303]) were proposed in the Capillary waves at the interface of two
literature, but in any case, π/kmax cannot immiscible electrolytes have been directly
be less than the molecular diameter [304]. observed by time-resolved QELS [306, 307],
Gravity suppresses the long-range fluctu- investigated in MD computer simula-
ations and removes a divergence of the tions [294, 300] and described within
correlation function at small wave vectors. the framework of a phenomenological the-
Both dcw and g(k) can be measured ory [295, 308, 309]. The principles of this
in optical, neutron, and x-ray scattering theory are clear, but the formalism, par-
experiments [297–299, 305], although it is ticularly for polarizations considerably far
more subtle to measure the height–height from the p.z.c. [295], is rather involved.

We refer the reader to Ref. [295], in which + −
these are described in detail. Here, we − (n1 + n1 − 2n01 ) + kB T
only summarize the basic approach and ∞
the main results. × dR dz n2 + (z, R)
The interface is described as before, ξ(R)

using the approximation of smooth corru-
n2 + (z, R)
gations. The two electrolyte solutions are × log + n2 − (z, R)
characterized by the solvent dielectric con- n02

stants, ε1 and ε2 , and Debye lengths, κ1 −1
n2 − (z, R)
and κ2 −1 . In addition to the interfacial ten- × log
sion term, the free-energy functional now n02
has two new terms. These are the electro-
static energy, Fe , and the term responsible − (n2 + + n2 − − 2n02 ) (120)
for the entropy of a dilute electrolyte, Fs ,
so that The notations are as follows: φi (z, R) and
n±i (z, R), the electrostatic potential and
F = F0 + Fcw [ξ(R)] + Fe [ξ(R)] + Fs [ξ(R)] concentrations of positive and negative
(118) ions in the phase (i), respectively; U is the
where overall potential drop across the interface;

Fe [ξ(R)] = −
ε1
dR Q and −Q are the overall charges in the
8π second and first phases, respectively. The
ξ(R) U Q-term reflects the extra work needed to
× dz[∇ϕ1 (z, R)]2 + e dR maintain the overall potential difference U
−∞
if the latter is fixed [310].
ξ(R)
This form of the functional assumes that
× dzϕ1 (z, R)(n1 + (z, R)
−∞ the difference in the free energy of each
ion transfer between water and organic
ε2
− n1 − (z, R)) − dR phase is so large that each ion species
8π
is located either completely in water or
∞
× dz[∇ϕ2 (z, R)] + e2
dR completely in oil. Furthermore, formation
ξ(R) of a spontaneous dipolar potential drop
ξ(R) across the contact of the two solvents
× dzϕ2 (z, R)(n2 + (z, R) and its possible coupling with the ion
−∞ distribution is not part of the model.
− n2 − (z, R)) + U Q (119) The functional depends on three types of
fields, namely, surface fluctuations ξ(R),
and electrostatic potential φi (z, R), and the
ξ(R) ionic concentrations n± i (z, R). Minimizing
Fs [ξ(R)] = kB T dR it with respect to φi and n±
−∞ i at a given
ξ(R), one obtains Poisson–Boltzmann
+ n1 + (z, R) equations that describe the distribution
× dz n1 (z, R) log
n01 of the electrostatic potential and ionic
concentrations. Substituting the results
− n1 − (z, R) into the density functional, one can
+ n1 (z, R) log
n01 minimize that with respect to ξ(R). This
can be easily done in consistence with the the harmonic approximation breaks down,
already used harmonic approximation for as they, essentially, reflect the electric field-
Fcw [ξ(k)] (Eq. 109). Then all characteristics induced destruction of the surface.
of the interface are obtained in a closed Indeed the potential-dependent FT of the
form. For voltage drops smaller than kB T , height–height correlation function now
the results look relatively compact [308, reads
309], but for larger voltages they are very kB T
cumbersome [295]. g(k, V ) = (122)
γ [k 2 + kgr
2 ] − f (k, V )
The result for the electric capacitance
reads where f (k, V ) is a very complicated
but elementary function of its argu-
C = CGC (V )R(κ1 , κ2 , V ) (121)
ments [295], which grows with |V |. For
where V = eU/kB T , CGC (V ) is the GC moderate |V |, it is proportional to V 2 ;
capacitance (Eqs. (105) and (106)) and for instance, for small wave vectors the
R(κ1 , κ2 , V ) is the so-called roughness expression is particularly simple,
function, the factor in which all the 1 2 2 (0)
deviations from GC are accumulated; f (k, V ) ≈ k V CGC (k κ1 , κ2 )
2
when γ → ∞, R → 1. The general (123)
(0)
expression for R̃(κ1 , κ2 , V ) [295] is too where CGC ≡ CGC |σ =0 .
cumbersome to repeat here, but its The harmonic approximation breaks
typical graph is shown in Fig. 27. This down when the denominator in Eq. (122)
graph shows that the roughness function approaches zero, but the tendency towards
increases rapidly with V because the dramatic enhancement of fluctuations is
surface area increases due to a dramatic correct. This translates into the depen-
enhancement of the amplitude of capillary dence of the mean square height of
waves. The steepest wings in Fig. 27 2 , for which there is a
corrugation, dcw
should be ignored, because in this region full expression [295], but here we quote
20
18
16
14
1
12
2
R
~
10 3
8
6
4
2
0
−15 −10 −5 0 5 10 15
eE (kBT )−1
Fig. 27 Roughness function versus potential. ε2 = 10;
γ = 30 dyn cm−1 ; κ2−1 , nm: (1) 1; (2) 2; (3) 5.
only its simplified version for the case of capacitance curves show a strong asym-
moderate V : metry as a function of the potential. The
V 2 kB T (0) discrepancy between experimental results
2 (0) 2
dcw ≈ (dcw ) + C and the GC theory has stimulated theoret-
4πγ 2 GC
ical work that went beyond the classical
κ1 κ2
× ln (124) scheme. The first step in this direction
(κ1 + κ2 )kgr has been done in Refs. [317–319], in
(0) which the ‘‘mixed boundary layer’’ was
where dcw is the root mean square
introduced and the effect of overlapping
height at the p.z.c. Equation (124) shows
of the two space-charge regions on the
unambiguously that charging the interface
double-layer capacitance has been consid-
enhances surface corrugation. The general
ered. The authors used a quasi-chemical
expression deviates from this law for
approximation (QCA) [317, 318] and MC
V > 4 showing a stronger enhancement.
The dispersion law is also modified by simulations [319] in order to reveal the
charging the interface. It now reads [295, density profiles of the ions (the Pois-
308], son–Boltzmann equation has been solved
numerically, assuming, for computational
ω2 = −2iω(k) − ωc2 (k) simplicity, that the dielectric constants of
k the two solvents are equal to each other
+ f (k, V ) (125) and do not vary with the distance from
ρ1 + ρ2
the interface). Numerical studies of the
that is, the charging reduces the frequency capacitance [317–319] reveal the effect of
of capillary waves. the mixed layer. They, however, do not dis-
However, such a simple theory has its close general laws for the capacitance of
own limitations. One of them is that relax- this interface, leaving open the following
ation of the Debye plasma cannot follow questions:
the high frequency capillary waves, and
their contribution to surface corrugation • Which characteristics determine the
will not be ‘‘seen’’ by electrolyte ions. In observed capacitance dependence on
Ref. [295], one can find a discussion of the the nature of ions?
yet unsolved dynamic aspects of the prob- • Which parameters determine the sign
lem and suggestion for a correction, taking of the deviation of the capacitance from
into account the incomplete adiabaticity. the Gouy–Chapman result and the
asymmetry of the capacitance curves
2.1.13.2 Ion Penetration into an as a function of potential?
‘‘Unfriendly Medium’’ • What information on the ionic profiles
The treatment of interfacial capacitance in and interfacial structure in general can
terms of the capacitance of two ‘‘back- we gain from the capacitance data?
to-back’’ GC double layers works fairly
well for some electrolytes, while it fails Furthermore, smearing of the interface
for the majority of others [289, 311–316]. between two solvents should cause a
In contrast to the predictions of the sim- smooth variation of dielectric properties
ple GC theory, it was found that the across the interface (if it were sharp,
capacitance of ITIES depends on the na- similar effects would emerge due to
ture of the ions [311, 313–318]. Often the nonlocal dielectric polarizability [92–94,
91, 89]). It is important to understand how the combined parameter,

this will affect the capacitance curves. +∞
1 1
A step towards answering these ques- L3 = dz exp(−g1 + (z))
tions was made in an analytical theory, re- −∞ 2 ε1
ported in Ref. [296]. The theory is based on 1 1
+ exp(−g1 − (z)) + exp(−g2 + (z))
a modified nonlinear Poisson–Boltzmann ε1 ε2
equation that takes into account the over-
1 1
lap of the two back-to-back double layers + exp(−g2 − (z)) − (128)
ε2 ε(z)
in the interfacial region and a smooth vari-
ation of dielectric properties across the which is also affected by of the effective
interface. Similarly to [44], the analytical dielectric constant profile ε(z), where
solution of this equation is obtained within ε(z → −∞ = ε1 ) and ε(z → ∞) = ε2 .
the perturbation theory, which utilizes the The effect of ion nature on these in-
smallness of the ratio of the ‘‘mixed layer’’ tegral parameters can be clarified on the
thickness to the Gouy lengths in the adja- basis of a microscopic model of the in-
cent solutions. The solution is thus valid terface or by MD simulations, giving
for not too high electrolyte concentrations ±
g1,2 (z) and ε(z). On the other hand,
and voltage drops across the interface. Note these parameters can be extracted from
that the most interesting results have been the capacitance measurements, and their
observed in the range of low ionic concen- dependence on the ion nature can be
trations [311–318]. traced. This suggests a new framework
The basic formula for capacitance, de- for the treatment of the capacitance data
rived in Ref. [296], shows that the capac- and establishes a relationship between ex-
itance dependence on the nature of ions perimental results and the microscopic
is controlled by three integral parameters. structure of ITIEs. Here, we will not
These parameters do not depend on the repeat the bulky expression for the ca-
potential or ionic concentrations. They are pacitance, referring instead the reader to
expressed through the z-dependent pro- Ref. [296]. We show just one example here,
files of the quantities, varying across the a graph (Fig. 28) that demonstrates the
interface: (i) the short-range contributions effect on capacitance of the penetration
to the free energy of ion transfer for all of a hydrophobic ion into the aqueous
double layer ions and (ii) the effective phase. Generally the integral parameters
dielectric constant of the solvent. These L1 , L2, and L3 can take either posi-
parameters are: tive or negative values depending on the
+∞ shape of the free energy profiles and
L1 = dz[exp(−g1 + (z)) dielectric function. As a result the capac-
−∞ itance of ITIES can be either higher or
− exp(−g1 − (z))] (126) lower than the GC capacitance of the two
+∞ back-to-back double layers with a sharp in-
L2 = dz[exp(−g2 + (z)) terface (see Figs. 3–6 in Ref. [296]). Both
−∞ types of behavior have been observed ex-
− exp(−g2 − (z))] (127) perimentally [313–315] and in numerical
simulations [317, 319].
±
where g1,2 (z) are the short-range energy The effect of concentration on the ca-
profiles for the corresponding ions, and pacitance curves has been studied using
3.0
0.3
2.5
C/C 0GC
2.0 0
Gouy-Chapman
1.5
−0.3
1.0
−4 −2 0 2 4
E (k BT )−1
Fig. 28 The effect of penetration of a hydrophobic positive ion into the
aqueous solution on the potential dependence of the normalized
0
capacitance, C/CGC . Curves correspond to the GC result obtained for
L1 = L2 = L3 = 0, and to the indicated values of the parameter
+∞
L+ +
2 = −∞ dz[exp(−g2 (z)) − θ(z)] in nm. The figure is plotted for the
case of no penetration of other ions through the interface, and no
smearing of the dielectric profile. ε1 = 80, ε2 = 10,
κ1 = κ2 = 0.3 nm−1 .
an equally concentrated solution in each Thus, CH can be positive or negative,

phase. For this case, the results of depending on the value of L3 . It should
Ref. [296] suggest that the deviations from be stressed, that there is no reason to expect
the GC capacitance and differences bein this system a true ‘‘compact layer’’, because
tween capacitance measured for different the interface is permeable for both salts. This
ions increase proportionally to the square result thus rationalizes a long-standing
root of the concentration, in line with ex- puzzle – the appearance of a compact-
perimental data [313]. layer contribution in some systems and its
There is a number of other interest- absence in others. On the whole it clarifies
ing findings in Ref. [296], including the the nature of the intercept of the PZ plot.
meaning of the apparent compact-layer
contribution in the PZ plot. The latter is 2.1.14
most transparent for small potential drops From Basic Concepts to Surface
across the interface, where the expressions Electrochemistry
are most simple, that is:
In this chapter, we have tried to give a con-
1 1 1
= 0 + (129) sistent and comprehensive picture of the
C CGC C H principles that describe the electrochemi-
cal interface, moving from long-standing
with the ‘‘compact-layer capacitance’’
but still valuable classical concepts of elec-
−1 trochemistry towards recent modern con-
CH = (130) cepts based on condensed matter physics
(4πL3 )
and statistical mechanics. To some extent of molecular simulation methods is the

the chapter followed the historic devel- ‘‘natural’’ way in which the solvent is
opment of the ideas on the interfacial included, thus avoiding the use of ‘‘ad hoc’’
structure and properties. assumptions concerning the behavior of
Experimental interfacial electrochem- the dielectric constant, to name only one
istry is nowadays dominated by the so- important aspect.
called ‘‘electrochemical surface science’’ or In general interfacial electrochemistry is
‘‘electrochemistry with atomic resolution’’. an area, in which the conceptual physi-
Various methods and results of surface cal models have to interact with chemical,
electrochemistry are thoroughly discussed quantum-mechanical ‘‘first principles’’ ap-
in this and in subsequent volumes of the proaches. The former can guide the
Encyclopedia. In terms of theoretical in- research, formulate questions and help ra-
terpretation, they often ask for insight on tionalizing the results of the latter. Such
the level of first principles calculations of a synthesis may, in a world of avalanche-
electronic structure. like progress in computational resources,
We did not review such calculations in not lie in the distant future. It could be
any depth, the reason being two-fold; first, ‘‘tomorrow’’.
before one performs such calculations, one
should become acquainted with the basic
References
phenomenological concepts and princi-
ples of a global or ‘‘collective’’ description 1. A. J. Bard, L. R. Faulkner, Electrochemical
of the interface. Secondly, while first prin- Methods, Wiley, New York, 1980 (1st ed.),
ciple calculations in surface science have 2001 (2nd ed.).
achieved a high level of sophistication, 2. J. Goodisman, Electrochemistry; Theoretical
Foundations, Wiley, New York, 1987.
their combination with crude models of
3. J. Koryta, Ions, Electrodes and Membranes,
the neighboring solvent is controversial. At Wiley, New York, 1991.
any rate, choosing the proper combination 4. C. M. A. Brett, A. M. O. Brett, Electrochem-
is currently a ‘‘fine art’’. istry: Principles, Methods and Applications,
Using the methods of semi-empirical Oxford University Press, Oxford, 1993.
quantum chemistry can be as instructive 5. J. O’ M. Bockris, S. U. M. Khan, Surface
Electrochemistry. A Molecular Level Approach,
as it can be, quite frequently, illusive,
Plenum Press, New York, 1993.
when it is disconnected from the real 6. E. Gileadi, Electrode Kinetics for Chemists,
dynamics and the global properties of the Chemical Engineers and Material Scientists,
interface. The development of theoretical VCH, Weinheim, 1993.
interfacial electrochemistry in the near 7. K. B. Oldham, J. C. Myland, Fundamentals
future will proceed in the direction of of Electrochemical Science, Academic Press,
San Diego, 1994.
a combination of ab initio quantum 8. R. G. Compton, G. H. W. Sanders, Electrode
chemical methods and MD. Progress and Potentials, Oxford University Press, Oxford,
obstacles in this direction were recently 1996.
reviewed [320]. MD of the solution part 9. Z. Galus, Fundamentals of Electrochemical
of the interface, based on empirical Analysis, Ellis Horwood, Chichester, 1997.
10. C. H. Hamann, A. Hamnett, W. Vielstich,
atom–atom interaction potentials, is often
Electrochemistry, Wiley-VCH, Weinheim,
of great help for the verification of global 1998.
models, rather than direct comparison 11. B. B. Damaskin, O. A. Petrii, G. A. Tsirlina,
with experiments. One obvious benefit Electrochemistry, Khimiya, Moscow, 2001.
12. O. A. Petrii, G. A. Tsirlina, Electrode poten- 31. O. Stern, Z. Elektrochem. 1924, 30, 508.
tials, Thermodynamics and Electrified In- 32. D. C. Grahame, Chem. Rev. 1947, 41, 441.
terfaces, Encyclopedia of Electrochemistry 33. A. Hamelin in Modern Aspects of Electro-
(Eds.: E. Gileadi, M. Urbakh), Wiley-VCH, chemistry (Eds.: B. E. Conway, R. E. White,
Germany, Vol. 1, Chap. 1. forthcoming. J. O’ M. Bockris), Plenum Press, New York,
13. B. B. Damaskin, O. A. Petrii, Specific ad- 1985, p. 1, Vol. 16.
sorption, Thermodynamics and Electrified 34. M. A. Vorotyntsev in Modern Aspects of
Interfaces, Encyclopedia of Electrochemistry Electrochemistry (Eds.: J. O’ M. Bockris,
(Eds.: E. Gileadi, M. Urbakh), Wiley-VCH, B. E. Conway, R. E. White), Plenum Press,
Germany, Vol. 1, Chap. 3. forthcoming. New York, 1986, p. 131, Vol. 17.
14. O. A. Petrii, Electrochim. Acta 1996, 41, 35. D. C. Grahame, J. Am. Chem. Soc. 1954, 76,
2307. 4819.
15. M. A. Vorotyntsev, Yu. A. Ermakov, V. S. 36. D. C. Grahame, Techn. Report no 14 to
Markin et al., Russ. J. Electrochem. 1993, 29, the office of Naval Research of Febr. 18,
513. 1954. Deposited at the Library of Congress,
16. M. A. Vorotyntsev, A. A. Rubashkin, J. P. Washington, D.C.
Badiali, Electrochim. Acta 1996, 41, 2313. 37. R. Parsons, F. G. R. Zobel, J. Electroanal.
17. A. N. Frumkin, Electrocapillary Phenomena Chem. 1965, 9, 333.
and Electrode Potentials, Commerce typogra- 38. M. A. Vorotyntsev, in Advances in Science
phy, Odessa, 1919. & Technology, VINITI, Moscow, 1984, p. 3,
18. A. N. Frumkin, Potentials of Zero Charge, Vol. 21.
Nauka, Moscow, 1979. 39. R. Gonzalez, F. Sanz, Electroanalysis 1997,
19. S. Trasatti in Comprehensive Treatise of 9, 169.
Electrochemistry (Eds.: J. O’ M. Bockris,
40. B. B. Damaskin, Sov. Electrochem. 2001, 37,
B. E. Conway, E. Yeager), Plenum Press,
745.
New York, 1980, p. 45, Vol. 1.
41. A. N. Frumkin, N. B. Grigoryev, Sov. Elec-
20. E. Lust, Double layers at single crystal
trochem. 1968, 4, 533.
and polycrystalline electrodes, Thermody-
42. A. A. Kornyshev, M. A. Vorotyntsev, Surf.
namics and Electrified Interfaces, Encyclo-
Sci. 1980, 101, 23.
pedia of Electrochemistry (Eds.: E. Gileadi,
43. A. A. Kornyshev, M. A. Vorotyntsev, Can. J.
M. Urbakh), Wiley-VCH, Germany, Vol. 1,
Chap. 2.4. forthcoming. Chem. 1981, 59, 2031.
21. S. Trasatti, O. A. Petrii, Pure Appl. Chem. 44. A. A. Kornyshev, W. Schmickler, M. A.
1991, 63, 711. Vorotyntsev, Phys. Rev. B 1982, 25, 5244.
22. S. Trasatti, E. Lust in Modern Aspects 45. B. B. Damaskin, N. V. Nikolaeva-Fedoro-
of Electrochemistry (Eds.: R. E. White, vich, Zh. Fiz. Khim. 1962, 36, 1483.
B. E. Conway, J. O’ M. Bockris), Kluwer 46. R. Payne, J. Electroanal. Chem. 1975, 60, 183.
Academic/Plenum Publishers, New York, 47. W. R. Fawcett, M. Opallo, Angew. Chem.,
London, 1999, p. 1, Vol. 33. Int. Ed. 1994, 33, 2131.
23. D. C. Grahame, J. Chem. Phys. 1948, 16, 48. W. R. Fawcett in Electrocatalysis (Eds.:
1117. J. Lipkowski, P. N. Ross), Wiley-VCH, New
24. D. C. Grahame, B. A. Soderberg, J. Chem. York, 1998, p. 323.
Phys. 1954, 22, 449. 49. A. N. Frumkin, Z. Phys. Chem. 1933, 164,
25. M. A. Vorotyntsev, J. P. Badiali, G. Inzelt, J. 121.
Electroanal. Chem. 1999, 472, 7. 50. A. N. Frumkin, G. M. Florianovich, Dokl.
26. D. C. Grahame, R. Parsons, J. Am. Chem. Akad. Nauk SSSR 1951, 80, 907.
Soc. 1961, 83, 1291. 51. G. M. Florianovich, A. N. Frumkin, Zh. Fiz.
27. A. N. Frumkin, R. V. Ivanova, B. B. Khim. 1955, 29, 1827.
Damaskin, Dokl. Akad. Nauk SSSR 1964, 52. A. N. Frumkin, in Advances in Electrochem-
157, 1202. istry and Electrochemical Engineering (Ed.:
28. B. M. Grafov, V. V. Elkin, Sov. Electrochem. P. Delahay) Intersci. Publ., New York,
1990, 26, 1015. Vol. 1, 1961, p. 65; Vol. 3, 1963, p. 287.
29. G. Gouy, J. Phys. Radium 1910, 9, 457. 53. L. I. Krishtalik, Electrochim. Acta 1965, 13,
30. D. Chapman, Philos. Mag. 1913, 25, 475. 1045.
54. P. Delahay, Double Layer and Electrode Ki- Electrodes, Tartu University, 1981, p. 146,
netics, Intersci. Publ., New York, London, Vol. VI.
Sydney, 1965. 76. M. A. Vorotyntsev, Sov. Electrochem. 1981,
55. K. Asada, P. Delahay, A. Sundaram, J. Am. 17, 472.
Chem. Soc. 1961, 83, 3396. 77. M. A. Vorotyntsev, Sov. Electrochem. 1981,
56. O. A. Petrii, A. N. Frumkin, Dokl. Akad. 17, 835.
Nauk SSSR 1962, 146, 1121. 78. M. A. Vorotyntsev, Sov. Electrochem. 1981,
57. A. N. Frumkin, O. A. Petrii, N. V. Nikolaeva- 17, 162.
Fedorovich, Electrochim. Acta 1953, 8, 177. 79. J. Lecoeur, C. R. Acad. Sci. (Paris) 1976,
58. R. R. Dogonadze, A. M. Kuznetsov, M. A. 283C, 651.
Vorotyntsev, Croat. Chim. Acta 1972, 44, 80. A. Hamelin, Sov. Electrochem. 1982, 18,
257. 1413.
59. A. N. Frumkin, N. V. Nikolaeva-Fedorovich, 81. L. I. Daikhin, A. A. Kornyshev, M. I. Ur-
N. V. Berezina et al., J. Electroanal. Chem. bakh, Phys. Rev. E 1996, 53, 6192.
1975, 58, 189. 82. L. I. Daikhin, A. A. Kornyshev, M. I. Ur-
60. D. I. Dzhaparidze, V. V. Shavgulidze, Sov. bakh, Electrochim. Acta 1997, 19, 2853.
Electrochem. 1973, 9, 1390. 83. L. I. Daikhin, A. A. Kornyshev, M. I. Ur-
61. N. V. Fedorovich, M. D. Levi, A. V. bakh, J. Chem. Phys. 1998, 108, 1715.
Shlepakov, Sov. Electrochem. 1977, 13, 89. 84. M. Abramovitz, I. Stegun, (Eds.), Handbook
62. R. R. Nazmutdinov, I. V. Pobelov, G. A. of Mathematical Functions, Dover, New
Tsirlina et al., J. Electroanal. Chem. 2000, York, 1965.
491, 126. 85. B. B. Mandelbrot, The Fractal Geometry of
63. R. R. Nazmutdinov, G. A. Tsirlina, Yu. I. Nature, Freeman, New York, 1982.
Kharkats et al., J. Phys. Chem. B 1998, 102, 86. E. Lust, A. Jänes, V. Sammelselg et al., Elec-
677. trochim. Acta 1988, 43, 373.
64. R. R. Nazmutdinov, G. A. Tsirlina, O. A. 87. E. Lust, A. Jänes, V. Sammelselg et al., Elec-
Petrii et al., Electrochim. Acta 2000, 45, 3521. trochim. Acta 2000, 46, 373.
65. D. J. Gavaghan, S. W. Feldberg, J. Elec- 88. A. A. Kornyshev in The Chemical Physics
troanal. Chem. 2000, 491, 103. of Solvation (Eds.: R. R. Dogonadze,
66. L. I. Krishtalik, Charge Transfer Reactions E. Kalman, A. A. Kornyshev, J. Ulstrup),
in Electrochemical and Chemical Processes, Elsevier, Amsterdam, 1988, p. 355, Part C
Plenum Press, New York, London, 1986. Solvation Phenomena in Physical, Chemical
67. R. J. Hunter, Zeta Potential in Colloid Sci- and Biological Systems.
ence. Principles and Applications, Academic 89. M. A. Vorotyntsev, A. A. Kornyshev, Electro-
Press, London, New York, 1981. statics of Media with the Spatial Dispersion,
68. J. Lyklema, Fundamentals of Interface and Nauka, Moscow, 1993.
Colloid Science, Academic Press, London, 90. M. A. Vorotyntsev, V. Yu. Izotov, A. A.
Fundamentals, Vol. 1, 1993; Solid-liquid Kornyshev, Sov. Electrochem. 1983, 19, 364.
Interfaces, Vol. 2, 1995. 91. A. A. Kornyshev in The Chemical Physics
69. G. Valette, A. Hamelin, J. Electroanal. of Solvation (Eds.: R. R. Dogonadze,
Chem. 1973, 45, 301. E. Kalman, A. A. Kornyshev, J. Ulstrup),
70. D. Leikis, K. Rybalka, E. Sevastyanov et al., Elsevier, Amsterdam, 1985, p. 77, Part A
J. Electroanal. Chem. 1973, 46, 161. Theory of Solvation.
71. A. N. Frumkin, J. Res. Inst. Catalysis 92. M. A. Vorotyntsev, A. A. Kornyshev, Sov.
Hokkaido Univ. 1967, 15, 61. Electrochem. 1979, 15, 560.
72. M. A. Vorotyntsev, J. Electroanal. Chem. 93. A. A. Kornyshev, M. A. Vorotyntsev, J. Uls-
1981, 123, 379. trup, Thin Solid Films 1981, 75, 105.
73. N. B. Grigoryev, Dokl. Akad. Nauk SSSR 94. A. A. Kornyshev, J. Ulstrup, Chem. Scr.
1976, 229, 647. 1985, 25, 58.
74. I. A. Bagotskaya, B. B. Damaskin, M. D. 95. A. A. Kornyshev, J. Electroanal. Chem. 1986,
Levi, J. Electroanal. Chem. 1980, 115, 189. 204, 79.
75. V. E. Kazarinov, M. A. Vorotyntsev, M. Ya. 96. A. A. Kornyshev, J. Ulstrup, J. Electroanal.
Kats in Double Layer and Adsorption at Solid Chem. 1985, 183, 387.
97. J. D. Cole, Perturbation Methods in Applied I. Prigogine, S. A. Rice), Wiley, New York,
Mathematics, Plaisdell, Waltham, 1968. 2001, p. 337.
98. R. G. Watts-Tobin, Philos. Mag. 1961, 6, 133. 125. M. A. Vorotyntsev, A. A. Kornyshev, Sov.
99. W. R. Fawcett, Isr. J. Chem. 1979, 18, 3. Electrochem. 1984, 20, 1.
100. N. F. Mott, R. G. Watts-Tobin, Electrochim. 126. A. A. Kornyshev, M. A. Vorotyntsev, J. Elec-
Acta 1961, 4, 79. troanal. Chem. 1984, 167, 1.
101. S. Levine, G. M. Bell, A. L. Smith, J. Phys. 127. J. W. Halley, B. Johnson, D. Price et al.,
Chem. 1969, 73, 3534. Phys. Rev. B 1985, 31, 7695.
102. I. L. Cooper, J. A. Harrison, J. Electroanal. 128. V. I. Feldman, A. A. Kornyshev, M. B.
Chem. 1975, 66, 85. Partenskii, Solid State Commun. 1985, 53,
103. W. R. Fawcett, J. Phys. Chem. 1978, 82, 1385. 157.
104. B. B. Damaskin, A. N. Frumkin, Electrochim. 129. A. V. Schlepakov, I. A. Bagotskaya, Elek-
Acta 1974, 19, 173. tokhimija 1982, 18, 26.
105. R. Parsons, J. Electroanal. Chem. 1975, 59, 130. L. M. Doubova, I. A. Bagotskaya, Elek-
229. tokhimija 1977, 13, 64.
106. B. B. Damaskin, J. Electroanal. Chem. 1977, 131. E. K. Patjarv, U. V. Palm, Elektokhimija
75, 359. 1976, 12, 806.
107. A. A. Kornyshev, Electrochim. Acta 1989, 34, 132. R. Payne, J. Am. Chem. Soc. 1967, 89, 459.
1829. 133. G. Vallette, A. Hamelin, J. Electroanal.
108. Ya. I. Frenkel, Z. Phys. 1928, 51, 232. Chem. 1973, 45, 301.
109. M. D. Lang, Solid State Phys. 1973, 28, 225. 134. V. A. Panin, K. V. Rybalka, D. I. Leikis,
110. N. D. Lang, W. Kohn, Phys. Rev. B 1971, 3, Elekrokhimija 1972, 8, 390.
1215. 135. D. Price, J. W. Halley, J. Electroanal. Chem.
1983, 150, 347.
111. I. E. Inglesfield in The Chemical Physics of
136. M. B. Partenskii, V. Dorman, P. C. Jordan,
Solid Surfaces and Heterogeneous Catalysis,
Int. Rev. Phys. Chem. 1996, 15, 153.
Elsevier, Amsterdam, 1981, p. 183, Vol. 1.
137. A. M. Brodsky, M. I. Urbakh, Usp. Fiz.
112. J. Harris, A. Liebsch, Vacuum 1983, 33, 655.
Nauk 1982, 138, 413.
113. A. K. Theophilou, A. Modinos, Phys. Rev. B
138. A. M. Brodsky, L. I. Daikhin, M. I. Urbakh,
1972, 6, 801.
Usp. Khim. 1985, 54, 3.
114. M. B. Partenskii, Ya. G. Smorodinskii, Fiz.
139. A. M. Brodsky, M. I. Urbakh, Prog. Surf. Sci.
Tverd. Tela 1974 16, 644. 1984, 15, 121.
115. P. G. Dzhavakhidze, A. A. Kornyshev, G. I. 140. A. M. Brodsky, M. I. Urbakh, Electrodynam-
Tsitsuashvili, Solid State Commun. 1984, 52, ics of Metal Surfaces, Nauka, Moscow, 1989.
401. 141. A. A. Kornyshev, M. I. Urbakh, J. Elec-
116. J. R. Smith, Phys. Rev. 1969, 181, 522. troanal. Chem. 1987, 235, 11.
117. P. Gies, R. Gerhardts, Phys. Rev. B 1985, 31, 142. J. D.-E. McIntyre in Advances in Electrochem-
6843. istry and Electrochemical Engineering (Eds.:
118. P. Gies, R. Gerhardts, Phys. Rev. B 1986, 33, P. Delahay, C. N. Tobias), Wiley, New York,
982. 1973, p. 61, Vol. 3.
119. W. Kohn, L. J. Sham, Phys. Rev. 1965, 140, 143. A. B. Ershler, A. M. Foontikov, V. N. Alex-
1133. eev et al., Elektrochimiya 1980, 16, 1171.
120. W. Schmickler, D. Henderson, Prog. Surf. 144. A. G. Foontikov, Dissertation, A. N.
Sci. 1986, 22, 323. Frumkin Institute of Electrochemistry of
121. V. I. Feldman, M. B. Partenskii, M. M. the Academy of Sciences of the USSR,
Vorobjev, Prog. Surf. Sci. 1986, 23, 3. Moscow, 1981.
122. J.-P. Badiali, Ber. Bunsen-Ges. Phys. Chem. 145. A. A. Kornyshev, M. I. Urbakh, J. Elec-
1987, 91, 270. troanal. Chem. 1988, 257, 305, 1987, 235,
123. S. Amokrane, J.-P. Badiali in Modern As- 11.
pects of Electrochemistry (Eds.: B. E. Conway, 146. P. G. Dzhavakhidze, A. A. Kornyshev,
J.O’M. Bockris, R. E. White), Plenum Press, A. Liebsch et al., Phys. Rev. B 1992, 45, 9339.
New York, 1992, p. 22, Vol. 1. 147. P. G. Dzhavakhidze, A. A. Kornyshev,
124. J. W. Halley, S. Walbran, D. Price A. Liebsch et al., Electrochim. Acta 1991, 36,
in Advances in Chemical Physics (Eds.: 1835.
148. P. G. Dzhavakhidze, A. A. Kornyshev, 172. L. Bosio, R. Cortes, C. Segaud, J. Chem.

A. Tadjeddine et al., Phys. Rev. B 1989, 39, Phys. 1979, 71, 1979.
13 106. 173. P. A. Thiel, T. E. Madey, Surf. Sci. Rep. 1987,
149. P. G. Dzhavakhidze, A. A. Kornyshev, 7, 211.
A. Tadjeddine et al., Electrochim. Acta 1989, 174. A. M. Brodsky, M. Watanabe, W. P. Rein-
34, 1677. hardt, Electrochim. Acta 1991, 36, 1695.
150. A. A. Kornyshev, A. M. Kuznetsov, 175. M. Watanabe, A. M. Brodsky, W. P. Rein-
G. Makov et al., J. Chem. Soc., Faraday hardt, J. Chem. Phys. 1991, 95, 4593.
Trans. 1996, 92, 3997. 176. X. Xia, M. L. Berkowitz, Phys. Rev. Lett.
151. A. A. Kornyshev, A. M. Kuznetsov, 1995, 74, 3193.
G. Makov et al., J. Chem. Soc., Faraday 177. K. J. Schweighofer, X. Xia, M. L. Berkowitz,
Trans. 1996, 92, 4005. Langmuir 1996, 12, 3747.
152. D. Henderson, M. Lozada-Cassou, J. Colloid 178. R. R. Nazmutdinov, M. M. Probst, K. Hein-
Interface Sci. 1986, 114, 180. zinger, J. Electroanal. Chem. 1994, 369, 227.
153. L. Blum, F. Vericat, D. Bratko, J. Chem. 179. M. W. Ribarsky, W. D. Luedtke, U. Land-
Phys. 1995, 102, 1461. man, Phys. Rev. B 1985, 32, 1430.
154. J. O’ M. Bockris, M. A. Devanathan, 180. A. Ignaczak, J. A. N. F. Gomes, J. Elec-
K. Müller, Proc. R. Soc. 1963, A274, 55. troanal. Chem. 1997, 420, 209.
155. E. Spohr, J. Phys. Chem. 1989, 93, 6171. 181. H. Sellers, P. V. Sudhakar, J. Chem. Phys.
156. J. Böcker, R. R. Nazmutdinov, E. Spohr 1992, 97, 6644.
et al., Surf. Sci. 1995, 335, 372. 182. H. Yang, J. L. Whitten, Surf. Sci. 1989, 223,
157. J. Böcker, E. Spohr, K. Heinzinger, Z. 131.
Naturforsch. 1995, 50a, 611. 183. M. Rosi, C. W. Bauschlicher Jr., J. Chem.
158. T. Kramar, D. Vogtenhuber, R. Podloucky Phys. 1989, 90, 7264.
et al., Electrochim. Acta 1995, 40, 43. 184. S. E. Feller, R. W. Pastor, A. Rojnuckarin
159. W. Langel, M. Parrinello, Phys. Rev. Lett. et al., J. Phys. Chem. 1996, 100, 17 011.
1994, 73, 504. 185. E. Spohr, J. Chem. Phys. 1997, 107, 6342.
160. W. Langel, M. Parrinello, J. Chem. Phys. 186. E. Spohr, Chem. Phys. 1990, 87, 141.
1995, 103, 3240. 187. K. Foster, K. Raghavan, M. L. Berkowitz,
161. D. Boda, D. Henderson, K.-Y. Chan et al., Chem. Phys. Lett. 1989, 162, 32.
Chem. Phys. Lett. 1999, 308, 473. 188. K. Raghavan, K. Foster, M. L. Berkowitz,
162. D. Boda, D. Henderson, K.-Y. Chan, J. Chem. Phys. Lett. 1991, 177, 426.
Chem. Phys. 1999, 110, 5346. 189. K. Raghavan, K. Foster, K. Motakabbir et al.,
163. D. Boda, D. Henderson, J. Chem. Phys. J. Chem. Phys. 1991, 94, 2110.
2000, 112, 8934. 190. J. Böcker, Z. N. Gurskii, K. Heinzinger, J.
164. D. Boda, K.-Y. Chan, D. Henderson, J. Phys. Chem. 1996, 100, 14 969.
Chem. Phys. 1998, 109, 7362. 191. J. Böcker, R. R. Nazmutdinov, E. Spohr
165. R. M. Townsend, J. Gryko, S. A. Rice, J. et al., Surf. Sci. 1995, 335, 372.
Chem. Phys. 1985, 82, 4391. 192. J. Böcker, E. Spohr, K. Heinzinger, Z.
166. W. Schmickler, E. Leiva, Mol. Phys. 1995, Naturforsch. 1995, 50a, 611.
86, 737. 193. M. L. Berkowitz, I.-C. Yeh, E. Spohr in Inter-
167. J. Seitz-Beywl, M. Poxleitner, M. M. Probst facial Electrochemistry (Ed.: A. Wieckowski),
et al., Int. J. Quantum Chem. 1992, 42, 1141. Marcel Dekker, New York, 1999.
168. S. Holloway, K. H. Bennemann, Surf. Sci. 194. J. Hautman, J. W. Halley, Y.-J. Rhee, J.
1980, 101, 327. Chem. Phys. 1989, 91, 467.
169. P. A. Bopp, A. Kohlmeyer, E. Spohr, Elec- 195. S. B. Zhu, G. W. Robinson, J. Chem. Phys.
trochim. Acta 1998, 43, 2911. 1991, 94, 1403.
170. I. Benjamin, Modern Aspects Electrochem. 196. G. Aloisi, M. L. Foresti, R. Guidelli et al., J.
1997, 31, 115. Chem. Phys. 1989, 91, 5592.
171. E. Spohr in Advances in Electrochemical 197. G. Aloisi, R. Guidelli, J. Chem. Phys. 1991,
Science and Technology (Eds.: D. M. Kolb, 95, 3679.
R. Alkire), Wiley-VCH, Weinheim, 1999, 198. A. A. Gardner, J. P. Valleau, J. Chem. Phys.
pp. 1–75, Vol. 6. 1987, 86, 4171.
199. G. Nagy, K. Heinzinger, J. Electroanal. 228. B. B. Smith, A. J. Nozik, J. Phys. Chem.

Chem. 1990, 296, 549. 1997, 101B, 2459.
200. G. Nagy, K. Heinzinger, E. Spohr, Faraday 229. I. Benjamin, J. Chem. Phys. 1992, 96, 577.
Discuss. 1992, 94, 307. 230. I. Benjamin, Science 1993, 95, 6675.
201. M. F. Toney, J. N. Howard, J. Richter et al., 231. O. Pecina, W. Schmickler, E. Spohr, J. Elec-
Nature 1994, 368, 444. troanal. Chem. 1995, 394, 29.
202. D. A. Rose, I. Benjamin, J. Chem. Phys. 232. O. Pecina, W. Schmickler, E. Spohr, J. Elec-
1991, 95, 6956. troanal. Chem. 1995, 405, 239.
203. D. A. Rose, I. Benjamin, J. Chem. Phys. 233. O. Pecina, W. Schmickler, J. Electroanal.
1993, 98, 2283. Chem. 1997, 431, 47.
204. E. Spohr, Chem. Phys. Lett. 1993, 207, 214. 234. A. Calhoun, M. T. M. Koper, G. A. Voth,
205. L. Perera, M. L. Berkowitz, J. Phys. Chem. Chem. Phys. Lett. 1999, 305, 94.
1993, 97, 13 803. 235. A. Calhoun, M. T. M. Koper, G. A. Voth, J.
206. E. Spohr, Acta Chem. Scand. 1995, 49, 189. Phys. Chem. 1999, 103B, 3442.
207. J. N. Glosli, M. R. Philpott, J. Chem. Phys. 236. I. Benjamin, Chem. Rev. 1996, 96, 1449.
1992, 96, 6962. 237. I. Benjamin, Annu. Rev. Phys. Chem. 1997,
208. J. N. Glosli, M. R. Philpott, J. Chem. Phys. 48, 407.
1993, 98, 9995. 238. H. D. Hurwitz, J. Electroanal. Chem. 1965,
209. J. Seitz-Beywl, M. Poxleitner, K. Heinzinger, 10, 35.
Z. Naturforsch. 1991, 46a, 876. 239. E. Dutkiewics, R. Parsons, J. Electroanal.
210. B. Eck, E. Spohr, Electrochim. Acta 1997, 42, Chem. 1966, 11, 100.
2779. 240. R. Payne, Advances in Electrochemistry and
211. G. Toth, K. Heinzinger, Chem. Phys. Lett. Electrochemical Engineering, Intersci. Publ.,
1995, 245, 48. New York, 1970, p. 1, Vol. 7.
212. M. R. Philpott, J. N. Glosli, J. Electrochem. 241. R. Payne, J. Electroanal. Chem. 1973, 41, 277.
Soc. 1995, 142, L25. 242. B. B. Damaskin, R. V. Ivanova, Usp. Khim.
213. M. R. Philpott, J. N. Glosli, S. B. Zhu, Surf. 1979, 48, 1747.
Sci. 1995, 335, 422. 243. R. Parsons, J. Electroanal. Chem. 1983, 150,
214. E. Spohr, J. Electroanal. Chem. 1998, 450, 51.
327. 244. A. Hamelin, T. Vitanov, A. Popov et al., J.
215. E. Spohr, Electrochim. Acta 1999, 44, 1697. Electroanal. Chem. 1983, 145, 225.
216. D. I. Dimitrov, N. D. Raev, J. Electroanal. 245. M. D. Levi, A. V. Shlepakov, B. B. Damaskin
Chem. 2000, 486, 1. et al., J. Electroanal. Chem. 1982, 138, 1.
217. D. A. Rose, I. Benjamin, J. Chem. Phys. 246. G. Valette, A. Hamelin, R. Parsons, Z. Phys.
1994, 100, 3545. Chem. N. F. 1978, 113, 71.
218. X. Xia, M. L. Berkowitz, Chem. Phys. Lett. 247. O. Stern, Z. Electrochem. 1924, 30, 508.
1994, 227, 561. 248. O. Essin, B. Markov, Acta Physicochim.
219. J. B. Straus, G. A. Voth, J. Phys. Chem. 1993, URSS 1939, 10, 353.
97, 7388. 249. B. V. Ershler, Zh. Fiz. Khim. 1946, 20, 679.
220. J. B. Straus, A. Calhoun, G. A. Voth, J. 250. D. C. Grahame, J. Am. Chem. Soc. 1958, 80,
Chem. Phys. 1995, 102, 529. 4201.
221. B. B. Smith, J. W. Halley, J. Chem. Phys. 251. J. M. Parry, R. Parsons, Trans. Faraday Soc.
1994, 101, 10 915. 1963, 59, 241.
222. D. A. Rose, I. Benjamin, Chem. Phys. Lett. 252. C. V. D’Alkaine, E. R. Gonsalez, R. Parsons,
1995, 234, 209. J. Electroanal. Chem. 1971, 32, 57.
223. B. B. Smith, J. W. Halley, A. J. Nozik, Chem. 253. Ya. M. Kolotyrkin, Yu. V. Alexeyev, Yu. A.
Phys. 1996, 205, 245. Popov, J. Electroanal. Chem. 1975, 62, 135.
224. A. Calhoun, G. A. Voth, J. Phys. Chem. 1996, 254. M. A. Vorotyntsev, Advances in Science &
100, 10 746. Technology, VINITI, Moscow, 1988, p. 3,
225. Y. G. Boroda, A. Calhoun, G. A. Voth, J. Vol. 26.
Chem. Phys. 1997, 107, 8940. 255. M. A. Vorotyntsev in The Chemical Physics
226. I. Benjamin, J. Phys. Chem. 1991, 95, 6675. of Solvation (Eds.: R. R. Dogonadze,
227. B. B. Smith, J. T. Hynes, J. Chem. Phys. E. Kalman, A. A. Kornyshev, J. Ulstrup),
1993, 99, 6517. Elsevier, Amsterdam, 1988, p. 401, Part C
Solvation Phenomena in Specific Physical, 277. X. Gao, A. Hamelin, M. J. Weaver, Phys.

Chemical, and Biological Systems. Rev. B 1991, 44, 10 983.
256. M. A. Vorotyntsev, K. Holub, Sov. Elec- 278. A. Hamelin, J. Electroanal. Chem. 1992, 329,
trochem. 1984, 20, 256. 3350.
257. M. A. Vorotyntsev, E. M. Itskovich, Sov. 279. O. M. Magnussen, J. Wiechers, R. J. Behm,
Electrochem. 1979, 15, 114. Surf. Sci. 1993, 289, 139.
258. M. A. Vorotyntsev, Sov. Electrochem. 1978, 280. B. M. Ocko, G. Hegelson, B. Schardt et al.,
14, 781. Phys. Rev. Lett. 1992, 69, 3350.
259. M. A. Vorotyntsev, Sov. Electrochem. 1978, 281. V. Heine, L. D. Marks, Surf. Sci. 1986, 165,
14, 783. 65.
260. M. A. Vorotyntsev, V. S. Krylov, L. I. Krish- 282. K.-M. Ho, K.-P. Bohnen, Phys. Rev. Lett.
talik, Sov. Electrochem. 1979, 15, 636. 1987, 59, 1833.
261. M. A. Vorotyntsev, J. Res. Inst. Catalysis 283. C. L. Fu, K. M. Ho, Phys. Rev. Lett. 1989, 63,
Hokkaido Univ. 1982, 30, 167. 1617.
262. M. A. Vorotyntsev, Sov. Electrochem. 1980, 284. K.-P. Bohnen, K.-M. Ho, Electrochim. Acta
16, 1112. 1995, 40, 129.
263. M. A. Vorotyntsev, Sov. Phys. Surf.: Phys., 285. B. M. McCoy, T. T. Wu, The Two-Dimen-
Chem., Mech. 1989, 39. sional Ising Model, Harvard University Press,
264. L. D. Landau, E. M. Lifshitz, L. P. Pitaevskii, Cambridge, 1973.
Statistical Physics, Pergamon Press, New 286. I. Vilfan, J. Villain, Surf. Sci. 1991, 257, 368.
York, 1980. 287. G. Pourcelli, C. Gavach in Proton Conduc-
265. A. Z. Patashinskii, V. L. Pokrovskii, Fluctu- tors (Ed.: Ph. Colomban), Cambridge Uni-
ation Theory of Phase Transitions, Pergamon versity Press, Cambridge, 1992, p. 294.
Press, Oxford, 1979. 288. M. Daoud, C. E. Williams, (Eds.), Soft Mat-
266. M. A. Vorotyntsev, S. N. Ivanov, Sov. Phys. ter Physics, Springer, Berlin, 1999.
JETP 1985, 61, 1028. 289. H. H. J. Girault, D. H. Schiffrin in Electro-
267. M. A. Vorotyntsev in Condensed Matter analytical Chemistry (Ed.: A. J. Bard), Marcel
Physics Aspects of Electrochemistry (Eds.: Dekker, New York, 1989, p. 1, Vol. 15.
M. P. Tosi, A. A. Kornyshev), World Scien- 290. A. G. Volkov, D. W. Deamer, D. I. Tanelian
tific, Singapore, 1991, p. 229. et al., Liquid Interfaces in Chemistry and
268. M. A. Vorotyntsev, Sov. Electrochem. 1985, Biology, Wiley, New York, 1989.
21, 257. 291. C. A. Croxton, Statistical Mechanics of the
269. M. A. Vorotyntsev, Sov. Electrochem. 1985, Liquid Surface, Wiley, New York, 1980.
21, 323. 292. M. P. Gelfand, M. E. Fisher, Physica A 1990,
270. M. A. Vorotyntsev, Sov. Electrochem. 1985, 166, 1.
21, 470. 293. F. Buff, R. Lovett, F. Stillinger, Phys. Rev.
271. A. V. Shcheglov, M. A. Vorotyntsev, B. B. Lett. 1965, 15, 621.
Damaskin et al., Sov. Electrochem. 1985, 21, 294. I. Benjamin in Modern Methods for Multidi-
1061. mensional Dynamics Computations in Chem-
272. A. A. Kornyshev, I. Vilfan, Phys. Rev. B istry (Ed.: D. L. Thompson), World Scien-
1993, 47, 10 775. tific, Singapore, 1998, p. 101.
273. A. A. Kornyshev, I. Vilfan, Electrochim. Acta 295. L. I. Daikin, A. A. Kornyshev, M. Urbakh, J.
1995, 40, 109. Electroanal. Chem. 2000, 483, 68.
274. A. A. Kornyshev, I. Vilfan in Microscopic 296. L. I. Daikin, A. A. Kornyshev, M. Urbakh, J.
Models of Electrode-Electrolyte Interfaces, Elec- Electroanal. Chem. 2001, 500, 461.
trochem. Soc. Proceedings Pennington, NJ, 297. D. Beaglehole, Phys. Rev. Lett. 1987, 58,
1993 (Eds.: J. W. Halley, L. Blum), p. 205, 1434.
Vol. 93–5. 298. L. T. Lee, D. Langevin, J. Meunier et al.,
275. P. Fery, W. Moritz, D. Wolf, Phys. Rev. B Prog. Colloid Polym. Sci. 1990, 81, 209.
1988, 38, 7275. 299. L. T. Lee, D. Langevin, B. Farnoux, Phys.
276. D. M. Kolb in Structure of Electrified Inter- Rev. Lett. 1991, 67, 2678.
faces (Eds.: J. Lipkowski, Ph. Ross), 1993, 300. K. J. Schweighofer, I. Benjamin, J. Elec-
VCH, New York, p. 65. troanal. Chem. 1995, 391, 1.
301. S. A. Safran, Statistical Thermodynamics of Plenum Press, New York, 1993, p. 1,

Surfaces. Interfaces and Membranes, Addison- Vol. 25.
Wesley, Reading, 1994. 312. Z. Samec, Chem. Rev. 1988, 88, 617.
302. D. Beaglehole, Phys. Rev. Lett. 1987, 58, 313. C. M. Pereira, A. Martins, M. Rochas et al.,
1434. J. Chem. Soc., Faraday Trans. 1994, 90, 143.
303. J. D. Weeks, J. Chem. Phys. 1977, 67, 3106. 314. H. H. Girault, D. J. Schiffrin, J. Electroanal.
304. R. Evans, Adv. Phys. 1979, 28, 143. Chem. 1983, 150, 43.
305. B. M. Ocko, X. Z. Wu, E. B. Sirota et al., 315. D. Homolka, P. Hajkova, V. Marecek et al.,
Phys. Rev. Lett. 1994, 2, 242. J. Electroanal. Chem. 1983, 159, 233.
306. Z.-H. Zhang, I. Tsuyumoto, S. Takashi 316. C. Yufei, V. J. Cunnane, L. Murtomaki
et al., J. Phys. Chem. A 1997, 101, 4163. et al., J. Chem. Soc. Faraday Trans. 1991,
307. Z.-H. Zhang, I. Tsuyumoto, T. Kitamori 87, 107.
et al., J. Phys. Chem. B 1998, 102, 10 284. 317. C. M. Pereira, W. Schmickler, F. Silva et al.,
308. L. I. Daikin, A. A. Kornyshev, M. Urbakh, Chem. Phys. Lett. 1997, 268, 13.
Chem. Phys. Lett. 1999, 309, 137. 318. C. M. Pereira, W. Schmickler, F. Silva et al.,
309. L. I. Daikin, A. A. Kornyshev, M. Urbakh, J. Electroanal. Chem. 1997, 436, 9.
Electrochim. Acta 1999, 45, 685. 319. T. Huber, O. Pecina, W. Schmickler, J. Elec-
310. L. D. Landau, E. M. Lifshitz, Electrodynam- troanal. Chem. 1999, 467, 203.
ics of Continuous Media, Pergamon Press, 320. J. W. Halley, S. Walbran, D. L. Price, Adv.
New York, 1984. Chem. Phys. 2000, 116, 337.
311. H. H. Girault in Modern Aspects of Elec-
trochemistry (Eds.: J.O’ M. Bockris et al.),
2.2 Electrical Double Layers: Theory and Simulations 133
2.2 electrochemical systems, and even within

Electrical Double Layers: Theory and this realm we have to restrict ourselves to
Simulations two important cases: metal–solution and
liquid–liquid interfaces. Of course, semi-
Wolfgang Schmickler conductor–solution interfaces are just as
Abteilung Elektrochemie, University of Ulm,
important in electrochemistry, but they
Ulm, Germany
are of little interest to electrochemical
theory: their double-layer properties are
dominated by the space-charge region
2.2.1
of the semiconductor, which are covered
Introduction
in textbooks on semiconductor physics.
Therefore, in the remainder of this chap-
Electrical double layers exist in many sys-
ter we shall treat the above-mentioned two
tems. They form whenever two conducting
systems in turn, and cover both theory and
phases meet at an interface that is imper- simulations, as suggested by the title. We
meable to the charge carriers. Then the shall assume throughout that there is no
application of a potential difference bet- specific adsorption of ions at the interface.
ween the two phases leads to the buildup While it is doubtful if there is any sys-
of charges of equal magnitude and oppo- tem for which this assumption is strictly
site sign at the interface. The resulting valid, specific adsorption is different for
charge distribution is the same as in a ca- each system, so there is no point in includ-
pacitor, and therefore much research has ing it in a general theory. So the models
been focused on the resulting capacity. presented in this chapter form the basis of
Such double layers may also form in sys- double-layer theory, to which the effects of
tems in which charge carriers can cross specific adsorption must be added for each
the interface, for example, at the contact individual system.
between two metals with different work
functions. In this case, an external voltage 2.2.2
gives rise to a current across the inter- Gouy–Chapman Theory and its Extensions
face, but often when the current is small
and the concentration of charge carriers Statistical double-layer theory started at
is high, the double layer remains undis- the beginning of the last century with the
turbed. Finally, double layers even exist works by Gouy [1] and Chapman [2], two
at an interface in which one of the two of the most influential papers ever written
adjoining phases does not conduct cur- in electrochemistry. Their theory, which
rent, but carries an intrinsic immobile we will briefly review below, explains the
excess charge on its surface – a situation double-layer capacity for metal–solution
that occurs in membranes and in biolog- interfaces at low electrolyte concentrations
ical systems. This excess charge is then quite well. Unfortunately, further progress
compensated by a charge layer in the con- has been slow, and during the last
ducting system. decade, there has been more work in
A general review of double layer the- simulations of the double layer than in
ory covering all these cases would cer- proper theory.
tainly be desirable, but is beyond the The Gouy–Chapman (GC) theory is
scope of this chapter. Here we focus on treated in all good textbooks, so we only
summarize the main points. It considers of the ions in the bulk of the solution, kB
a planar metal electrode in contact with is the Boltzmann constant, and T is the
an ionic solution, and starts from the temperature. The Debye length LD , which
following assumptions: is familiar from the Debye–Hückel theory
for electrolyte solutions, is given by
• the metal is a perfect conductor, and its
excess charge is distributed evenly on 1/2
εε0 kT
the surface; LD = (2)
2(ze0 )2 n0
• the solvent is a dielectric continuum,
characterized by a dielectric constant ε; The GC capacity has a pronounced mini-
• the ions are point particles, whose mum at the pzc, and rises rapidly on both
distribution is determined by the Pois- sides (see Fig. 1). It compares well with
son–Boltzmann equation. experimental data for low electrolyte con-
For the case of a z–z electrolyte, explicit centrations and for low to medium charge
expressions for the differential capacity densities, where the structure of the sys-
and for the profile of the electrostatic tem, the size and chemical nature of the
potential can be derived. We give the solution, and the electronic properties of
expression for differential capacity C per the metal, do not matter.
unit area: Considering the simplicity of the GC
theory, it works surprisingly well–so well
dσM εε0 ze0 (φ − φpzc ) that it still serves as the reference point
CGC = = cosh for all further advances in double layer
dφ LD 2kT
(1) theory, and new results are often stated as
where φ is the electrode potential and corrections to the GC theory.
φpzc is the potential of zero charge (pzc), A cursory review of the experimental
at which the charge density σM on the data shows that the GC theory overes-
electrode vanishes. n0 is the concentration timates the capacity for high electrolyte
3.0
2.0
[F m−2]
10−1 M
C
1.0
10−2 M
10−3 M
0.0
−0.10 −0.06 −0.02 0.02 0.06 0.10
(f − f pzc)
[V]
Fig. 1 The Gouy–Chapman capacity for several ionic
concentrations of a 1 : 1 electrolyte.
concentrations and for high charge densi- can be obtained by measuring the capacity,
ties. In a real solution, the ions and the at a given charge density, for several
solvent molecules have a finite size, which electrolyte concentrations, and plotting
prevents their centers from coming arbi- the inverse of the experimental capacity
trarily close to the metal surface, and which versus the calculated GC capacity; this is
also limits the amount of charge that can known as a Parsons and Zobel plot [4].
be packed into the first few layers. To ac- This should result in a straight line, and
count for these effects, Stern [3] suggested the Helmholtz capacity can be obtained
that the first layer of the solution is free of from the intercept. For some metals, in
ions, and that this inner or compact layer particular for mercury, this procedure
contributes a term to the capacity, which works well, while for others, such as
is in series with the GC term: gold [5], the plot is curved, which may
1 1 1 indicate that there is always at least weak
= + (3) specific adsorption on these metals; of
C CH CGC
course, it may also indicate that Eq. (3)
where the Helmholtz capacity CH is does not hold for these systems.
attributed to the inner layer; CH is The Helmholtz capacity always depends
sometimes called the inner layer or compact strongly on the charge density σM ; dif-
layer capacity, but since Eq. (3) also ferent metals may have quite different
holds in models that do not assume capacities, but typically there is a maxi-
the existence of an inner layer, we shall mum, also known as the hump, near the
not use that terminology. In the absence pzc (see Fig. 2). Since the strength of
of specific adsorption, the Helmholtz the electric field increases on both sides
capacity is expected to be independent of of the pzc, the occurrence of the hump
the electrolyte concentration, and also of can be explained by dielectric saturation.
the nature of the ions. Experimentally it Watts-Tobin [6] elaborated this idea into
80
70
Ag(111)
60
50
[µF cm−2]
CH
40
30
20 Hg
10
0
−12.0 −8.0 −4.0 0.0 4.0 8.0 12.0
s
[µC cm−2]
Fig. 2 The Helmholtz capacity for Ag(111) and Hg in contact with
an aqueous solution; after Ref. [77] with permission from the author.
a quantitative model: he treated the in- in which each point is occupied by one
ner layer as a two-dimensional lattice of water molecule, which can form hydrogen
dipoles, which could be oriented either bonds with its nearest neighbors and
parallel or antiparallel to the surface nor- take up a fairly large number (25) of
mal. In effect, this is equivalent to the Ising orientations. The statistical mechanics of
model, which was evaluated in the mean this ensemble was treated in the quasi-
field approximation (MFA). As expected, chemical approximation, which is much
this model predicts a capacity hump at better than the mean-field approximation
the pzc, but otherwise it does not agree that was employed in the Ising-like models
well with the Helmholtz capacity of mer- discussed above.
cury, on which the early theories were However, all these early monolayer
focused, or of any other metal. In mer- models suffer from two defects: the
cury, the Helmholtz capacity rises again at connection of the inner layer to the rest of
charge densities above the hump. This the solution is not clear and the electronic
increase was explained [7] by the pres- structure of the metal is neglected – it just
ence of two water species in the inner acts as a perfect conductor. Nowadays, it
layer: single water molecules and wa- is generally recognized that there are three
ter clusters composed of two molecules, contributions to the double-layer capacity
which have a smaller dipole moment. on metals:
A high electric field can break up the
1. the capacity of the diffuse layer, which
clusters, and the capacity increases again
is described by the GC theory;
since the average dipole moment becomes
2. a contribution from the boundary layer
larger. Parsons [8] worked out the statisti-
of the solution, which extends over a
cal mechanics of this four-state model (two
distance of several solvent diameters
orientations for each water species), and by
toward the bulk;
an appropriate choice of the system param-
3. the electronic response of the metal
eters he obtained reasonable agreement
surface to the high field at the interface.
with the data for mercury. Later this model
was elaborated by other researchers, who A treatment of these effects requires
added further details and increased the more advanced methods than the GC
number of adjustable parameters, so that theory does; therefore, we shall briefly
the mercury data could be fitted perfectly; review the methods that have been used in
such work has been critically reviewed by the double-layer theory before discussing
Guidelli [9]. However, if this kind of model the latter two contributions in greater
were to represent the data for other met- detail.
als, all parameters would have to be fitted
again, so the physical significance of these 2.2.3
models was dubious from the beginning. Theoretical Methods and Principles of
Modern theories have abandoned the Computer Simulations
idea that the water dipoles can take up only
two orientations, though it occasionally Double-layer theory is complicated by the
still crops up in the literature. A significant fact that it involves the theory of liquids
conceptual progress was achieved in the and of solids, and of the interactions be-
work of Guidelli [10, 11]. He modeled the tween them. While the theory of the bulk
inner layer as a two-dimensional lattice, of these phases is quite well developed,
the treatment of boundary effects is notori- the Ornstein–Zernike relation is often

ously difficult, and the interaction between used as a starting point. It is based on
the two phases is almost intractable. At the following idea: the correlations be-
present, we understand the basic features tween two particles can be decomposed
of the double layer, but quantitative cal- into two sources, that is, the direct in-
culations, even for simple systems, are teraction between the particles and the
beyond our capabilities. interaction mediated by other particles
Below we briefly review the methods that in the vicinity. Therefore a direct cor-
have been used in double-layer studies; relation function cij (1, 2) is introduced
they comprise both theoretical methods through
and simulations. The latter have become ρk
quite popular in recent years, since they hij (1, 2) = cij (1, 2) +
allow the treatment of more complicated ω
k
models than can be handled by pure
theory. × cik (1, 3)hik (3, 2) d3 (4)
2.2.3.1 Integral Equations where ρk is the density of the species k

The structure of a bulk fluid can be charac- and ω is a normalization factor, which
terized by the pair distribution functions is unity for most systems of interest.
gij (1, 2), which give the probability of So far, Eq. (4) is but a definition for
finding a particle of species i at the po- the direct correlation cij (1, 2). However,
sition r1 with orientational coordinates 1 the range of the direct correlation is of
and a particle of species j at the po- the same order of magnitude as that of the
sition r2 with orientational coordinates interaction potentials vij (1, 2) and shorter
2 , while all other particles are dis- than that of the total correlation function,
tributed statistically. The arguments (1, 2) which contains the mediated interactions.
are shorthand for r1 , 1 , r2 , 2 . These dis- This fact has been used as a basis for
tribution functions are normalized such approximations based on diagrammatic
that they tend to unity as the separation expansion schemes. They start from the
between the two particles goes to infin- exact expansion:
ity. The quantity hij (1, 2) = gij (1, 2) − 1
c(1, 2) = h(1, 2) − βv(1, 2)
is called the total correlation function; it
vanishes for infinite separation and char- − log[h(1, 2) + 1] + d(1, 2) (5)
acterizes the interaction between the two
particles. where the subscripts are omitted to sim-
In principle, these correlation functions plify notations. Here, d(1, 2) is the sum
should be calculated from the interaction of all bridge diagrams in the diagrammatic
potentials vij (1, 2) between the particles. expansion and β ≡ 1/kT . Unfortunately,
However, even for an ensemble of hard the combination of Eqns. (4) and (5) is
spheres, this is an intractable problem, intractable because it involves an infi-
let alone for systems with realistic inter- nite series of many-dimensional integrals
actions. Therefore various schemes have of products of h(1, 2). Neglecting the
been devised to calculate the correlation bridge functions results in the hypernet-
functions from approximate relations be- ted chain approximation (HNC). This is
tween these functions. For this purpose, an example of a closure relation, because
together with Eq. (4) it forms a closed several simple models of water [12, 13];
system of equations for the unknown func- the results are all numerical.
tions cij (1, 2) and hij (1, 2), which can be The MSA, HNC, RHNC, and similar
solved at least in principle. In the spe- schemes were first developed to treat bulk
cial case in which the interaction is small, fluids, and then applied to interfaces by
βvij (1, 2) < 1, the total correlation func- using the following trick: one spherical
tions hij are also small, and the HNCs particle from the ensemble is singled out
simplify to the mean spherical approxima- and its radius is taken to be infinitely large
tion (MSA): or, if that limit is not tractable numerically,
much larger than the radii of all other
cij (1, 2) = −βvij (1, 2) (6) molecules. This large particle represents
the surface.
which results in a linear theory. This
is a useful approximation for weakly
2.2.3.2 Computer Simulations
interacting particles in the vicinity of
Even comparatively simple models turn
the pzc, where the Coulomb forces are
out to be difficult to treat by the methods
weak.
of statistical mechanics. In such cases,
A better approximation, which has be-
computer simulations may be used to
come quite popular recently, is the reference
explore the consequences and predictions
hypernetted chain (RHNC) approximation.
of the model, even though they do not
It makes use of a reference system consisting
give the same detailed understanding as
of a fluid of neutral hard spheres, which
analytical methods do. This is not the place
is the only potential for which the bridge
to describe these methods in any detail; the
function d(1, 2) in Eq. (2) is known with
book by Allen and Tildesley [14] provides
sufficient accuracy. If we denote by the sub-
an excellent introduction to the field. There
script 0 a property of the reference system
are three different methods.
and set h = h0 + h, c = c0 + c, v =
v0 + v, and d = d0 + d ≈ d0 , Eq. (5)
can be written as Monte Carlo simulations The basic me-
thod works in the following way: starting
c(1, 2) = h(1, 2) from a given configuration, a new con-
figuration is tentatively chosen at random
− log[g0 + h(1, 2)]
from a certain region of phase space, for
+ log g0 − βv(1, 2) (7) example, by moving one particle a cer-
tain distance. If the energy of the new
where g0 ≡ h0 + 1 is only a function configuration is lower than that of the
of the separation r ≡ |r1 − r2 | between previous one, it is accepted, and a new
the two particles. The pair-interaction Monte Carlo (MC) step is initiated. Other-
potential v0 of the reference system is wise a random number r is chosen from
just the spherically symmetrical part of the interval (0, 1). If r < exp −E/kT ,
v(1, 2). Equation (7) is exact except for the where E is the change in energy, the
replacement of the true bridge function new configuration is accepted, otherwise
d(1, 2) of the system under study by the it is discarded. It can be shown that
bridge function d0 (1, 2) of the reference averaging over a large number of config-
system. The RHNC has been applied to urations that have been produced in this
way leads to correct thermodynamic quan- bring them up again as we discuss specific
tities [15]. There are various variants of this models.
method; they mainly provide information Real systems contain a practically infi-
about the structure and thermodynam- nite number of particles, while the ensem-
ics of an ensemble, but a few have bles that are used in simulations contain
been designed to investigate the dynamics typically a few hundred, at best a few
as well. thousand, particles. In order to mimic an
infinite system, cyclic boundary conditions
Molecular dynamics The orbits of the can be imposed. In electrochemical investi-
particles in an ensemble are obtained gations they are usually imposed in the two
from Newton’s second law: at each step directions parallel to the electrode surface.
the forces on all particles, and hence In the perpendicular direction the dimen-
their accelerations, are calculated, and sions should be so large that a region
from these their movements during a with bulklike properties is obtained, so
small time step, in which the forces that bulk and surface properties can be
can be taken as constant, are obtained. contrasted.
Then the forces are calculated again,
and the system moves another step 2.2.4
forward. During an initialization period, Electronic Structure Calculations
the system must be equilibrized to the
desired temperature, and then it can be In spite of the explosion of computing
sampled to obtain statistical properties. power during the last decade, it is still
Obviously, this method gives both the impossible to calculate the electronic
dynamics and the thermodynamics of the structure of semi-infinite solids. Three
ensemble. different approaches have been taken to
deal with this problem.
Car–Parinello method In ordinary simu-
lations the electronic structure of the par- Simple model systems A good method in
ticles is taken as fixed, and the molecules such situations is a simplification of the
interact through given potentials. In the model so as to make it mathematically
Car–Parinello method [16], at each step of tractable. The art of the theorist consists
the simulation the electronic structure of in keeping the most relevant features of
the system is recalculated, usually through the system and discarding the others.
density-functional methods [17, 18], and Since in the electric double layer the
from these the forces on the particles are electronic properties of the metal surface
obtained. In principle, this is the most play the major role, the jellium model [18],
exact but also the most time-consuming which disregards the details of the atomic
method, so that only small ensembles can structure and focuses on the distribution
be considered. Sometimes this method of the electrons, has been widely applied.
is simplified by restricting the electronic While this model does not give quantitative
structure calculations to a part of the en- results for particular systems, it has
semble. greatly improved our understanding of
All of these methods have been applied the electronic effects; we will discuss
to electrochemical systems, and we shall it later.
Slab calculations At a perfect surface the 2.2.5

atomic structure of a solid is periodic Models for the Solution
in the direction parallel to the surface.
This symmetry makes it possible to The surface properties of a phase generally
calculate the electronic structure of slabs differ from the bulk properties. There-
that are infinite in two directions and fore the GC theory, which is based on
consist of a few layers of atoms. Typically, macroscopic bulk concepts such as the
about 10 layers are sufficient to obtain dielectric constant, requires corrections.
a region with bulklike properties in Indeed, this was the idea behind the mono-
the center and two surface regions. By layer models discussed earlier – their main
their very nature such calculations are fault was that they were too simple: bound-
limited to structures that are periodic ary layers extend beyond the first layer
parallel to the surface; the unit cell of molecules that is in contact with the
must not be too large – with present-day metal. Furthermore, these models com-
computers, it must not contain more than pletely neglected the effect of the metal,
a few atoms – or the calculations become which was treated as a perfect conductor
computationally too expensive. without any electronic structure. In this
section we will discuss a few models for
Cluster calculations The conceptually the solution before we turn to the metal
simplest method consists in replacing surface and then to the metal–solution
the infinite system by a large cluster interface.
of particles. Such calculations can be
performed with several packages, from 2.2.5.1 Field-theoretical Approach
both commercial and public domain, and to the Double Layer
have therefore become quite popular. The GC theory is based on macroscopic
However, even in relatively big clusters electrostatics, while much of the present-
a large fraction of the atoms lie at day work is aimed at a description on a
the surface. Therefore the results of molecular level. There is, however, an el-
such calculations have to be interpreted egant alternative based on the methods of
with care – particularly, if the system is statistical field theory, which was devel-
charged or has an uneven distribution of oped by Badiali and associates [19, 20].
charges, since charges tend to accumulate This leads to a coarse-grained descrip-
on surfaces. In addition, the results tion in terms of the particle-distribution
obtained often depend on the size of functions. By setting up a suitable model
the cluster, so that one has to focus Hamiltonian, a wide variety of phenomena
on relative rather than on absolute can be described.
values. Badiali and coworkers started by apply-
Nowadays such quantum calculations ing this method to a combination of the
are usually based on density functional primitive model, in which the ions are con-
formalism [17], in which the electronic sidered as point charges and the solution
density, and not the wave functions, is is treated as a dielectric continuum, and a
used as the basic variable. This formalism simple model for the metal in which the
results in much faster programs, so that charge distribution is taken as uniform in
larger systems can be treated than that the direction parallel to the surface. In this
with the conventional approaches. model the Hamiltonian depends only on
the charge distributions e0 q(r) in the dou- help in understanding the incorporated
ble layer and e0 q0 (z) in the metal, and can effects.
be written in the form
βe02 [q0 (z) + q(r)] 2.2.5.2 Ensembles of Hard Spheres

[q0 (z ) + q(r )] In the GC theory the size of the molecules
βH [q(r)] = dr dr was disregarded, so it is a natural im-
8πε|r − r |
∞ provement to treat them as hard spheres.
an
+ [q(r)]n dr (8) Actually, there are two different models
ρb n! that use hard spheres: in the primi-
n=2
tive model the ions are considered as
where β = 1/kB T , and ρb is the bulk charged hard spheres, while the solvent
concentration of the ions. The first term is a dielectric continuum; in the civilized
is the electrostatic energy in the systems, or nonprimitive model both the ions and
and all other terms, in particular structural the solvent molecules are considered to
and entropic terms, have been written be spheres, but the latter have a dipole
as an expansion in powers of q(r). moment at their center. The primitive
The coefficients an are dimensionless model is mainly of academic interest:
quantities. ions and solvent molecules are roughly
The simplest way to treat this Hamilto- of the same size and should be treated on
nian is the MFA. The statistical average the same footing. However, one predic-
q(z) is then the value of q(r) for which
tion from the primitive model deserves
the functional derivative δH [q(r)]/δq(r)
attention: it suggests that at low tem-
vanishes. For the Hamiltonian of Eq. (8)
peratures the interfacial capacity should
this gives
increase with temperature, while the GC
∞
theory and most model calculations predict
an
βV (z) + q(z)(n−1) = 0 (9) a decrease. In this context, it is useful to
ρb
n=1 define an effective or reduced temperature
where V (z) is the average electrostatic through T ∗ = kT σi (4πεε0 )/e02 , where σi
potential. If one assumes that the average is the diameter of the ions, which is as-
charge distribution is related to the average sumed to be equal for the two species.
potential via a Boltzmann distribution and Low temperature means T ∗ < 1, so that
that the metal is a classical conductor (i.e. aqueous solutions are high-temperature
q0 (z) ∝ δ(z)), then all the coefficients an systems because of their high dielectric
can be determined, and one obtains the constant. Somewhat surprisingly, molten
nonlinear GC theory. salts are low-temperature systems by
So far, this is nothing but an ingenious this definition. This result is not only
method for obtaining a well-known result. based on simulations within the primi-
However, this approach can be extended tive model [21] but also on more general
in various ways by including other effects. arguments based upon pressure balance
Stafiej and coworkers discuss explicitly at the interface [22]; it may explain why
ion–ion interactions, chemical interac- the capacity of gold electrodes in contact
tions with the wall, and nonlocality. The with a frozen, highly concentrated aque-
main value of this method lies in the fact ous solution has been found to increase
that the calculations are transparent and with temperature [23].
We now focus our attention on the developed into a power series in κ; the first
civilized model. In the electrochemical two terms are
context the basic system consists of hard
1 1 1 1 ε−1
spheres in contact with a charged hard = + σi + σs
C εε0 κ 2 λ
wall; there are three types of hard spheres: (10)
solvent molecules of diameter σs , which The next term is of the order of κ. Within
have a dipole moment µ at their center, the MSA, the dielectric constant ε is related
and ions with a diameter σi and ions to the dipole moment µ and the density ρs
with a charge number ±1 – note that both of the solvent through
kinds of ions have the same diameter.
The integral equations for this model ε−1 4πµ2 9
= ρs (11)
have been solved analytically, but in the ε 3kT λ (λ + 1)2
2
MSA [24, 25] only, so that the results are
valid for small excess-charge densities on and the parameter λ is obtained from the
the electrode, that is, in the vicinity of dielectric constant via
the pzc.
λ2 (1 + λ)4 = ε (12)
The solution is implicit and consists of
a pair of nonlinear algebraic equations, The first term in Eq. (10) is just the GC
which can be solved numerically. In the capacity at the pzc; the second term is
limit of low concentrations, where the independent of the ionic concentration,
Debye inverse length κ = 1/LD tends to and can be identified with the Helmholtz
zero, explicit results can be derived. The capacity. However, in this model the
distribution of the ions gi (x) and of the Helmholtz capacity is not caused by a
dipoles gs (x) as a function of the separa- single monolayer of solvent with special
tion x from the electrode show oscillations, properties, like in the Stern model, but
which are caused by the packing of the par- results from an extended boundary layer.
ticles (see Fig. 3); they die out far from the It depends on the dielectric properties of
surface. the solvent and on the diameters of the
The interfacial capacity C is of particular particles. Since λ ε, the influence of
interest since it is measurable. For low the ions on the capacity is predicted to be
ionic concentrations, its inverse can be small. This is in line with the experimental
2
[kT]
e0f
Fig. 3Oscillations in the electrostatic

potential in an ensemble of hard
0 1 2 3 4 spheres caused by the packing of the
particles against the electrode; the
x straight line is the prediction of the GC
[σ] theory. (Data taken from Ref. [24].)
observation that the Helmholtz capacity is 2.2.6

practically independent of the nature of The Role of the Metal
the ions, provided they are not specifically
adsorbed. So this prediction is one of 2.2.6.1 Qualitative Considerations
the successes of the theory. The second In concentrated solutions the electric field
term in the Helmholtz capacity represents at the electrode surface is so high that it
the effect of the solvent. The parameter distorts the electronic density of the metal,
λ accounts for the fact that the dielectric and thereby changes the surface potential
properties at the interface differ from those χ. It is natural to define a contribution
of the bulk; roughly speaking, the effective of the metal to the interfacial capacity
dielectric constant in this region is much through
reduced. 1 ∂χ
= (13)
For aqueous solutions, with σs ≈ 3 Å Cm ∂σM
and ε ≈ 80, Eq. (10) predicts a Helmholtz
According to Le Chatelier’s principle,
capacity of the order of 16 µF cm−2 at
this metal capacity is negative: a positive
the pzc; this is much smaller than any
charge density increases the potential drop
experimental values. The contribution of
the metal is missing in this model. We will between the metal and the solution, and
see later that the electronic polarizability of the surface electrons react so as to make it
the metal surface increases the Helmholtz smaller. Thus a positive charge density
capacity. on the metal surface gives rise to an
These results for the ensemble of hard electric field directed outward from the
spheres are now more than 20 years old; all surface, which pulls the electrons back
the same, it has not been possible to extend into the metal and thereby decreases
them into the nonlinear region away from the surface dipole moment. Conversely,
the pzc in a rigorous way, though heuristic a negative charge pulls the electrons
extensions have been suggested [26]. Even toward the solution and increases the
computer simulations have met with little surface potential (see Fig. 4). Since the
success: for realistic values of the dipole metal capacity is independent of the
moments and the diameters, the ensem- electrolyte concentration, it contributes to
ble tends to get stuck in metastable states the Helmholtz capacity, which can thus be
away from equilibrium, so that it is impos- decomposed into parts pertaining to the
sible to obtain thermodynamic averages. metal and to the solvent:
Therefore, Henderson and coworkers [27] 1 1 1
have performed MC simulations for an = + (14)
CH Cm Csol
ensemble in which both the dipole mo-
ment of the solvent and the charges on the However, since the metal and the solvent
ions have been scaled down to a fraction interact at the interface, these two con-
of their values for aqueous solutions. Such tributions are not independent. Since the
systems do reach equilibrium; the parti- metal capacity is negative, it makes the
cle profiles show a strong layering of the total capacity, which must be positive in
particles, similar to that observed in the any reasonable model, larger. In essence,
MSA, but it is not possible to extrapolate the metal surface possesses a high elec-
the results to realistic charges and dipole tronic polarizability, which increases the
moments. double-layer capacity.
1.25
1.00 Metal surface
0.75
n n −1b
Metal
0.50
Effective
0.25
image plane
0.00
−3.0 −1.5 0.0 1.5 3.0
x
[Å]
Fig. 4 Normalized electronic density profile at the surface of
jellium with a bulk density of 15.3 atomic units. The upper dashed
curve is for a surface charge density σM = −0.1 C m−2 , the lower
dashed curve for σM = 0.1 C m−2 . The full curve is for an
uncharged surface. The arrow gives the position of the effective
image plane for an uncharged surface.
Trasatti [28] has noticed that in aqueous A major breakthrough in our under-
solutions the Helmholtz capacity CH of standing of the electric double layer
the simple sp metals, taken at the pzc, came with the introduction of the jel-
correlates with their electronic densities lium model, which had already been used
(see Fig. 5). This indicates that the surface extensively in the theory of metal sur-
polarizability of simple metals increases faces [31, 32], into electrochemistry [34,
with their electronic densities, a trend that 36]. In this model the lattice of posi-
can be explained by the jellium model tively charged metal ions is represented
presented below. by a constant positive background charge,
which drops abruptly to zero at the sur-
2.2.6.2 The Jellium Model face. The electrons are modeled as an
As early as 1928, Rice [29] attempted to inhomogeneous electron gas, which inter-
estimate the contribution of the metal to acts with the positive background. Explicit
the interfacial capacity within the frame- calculations for this model are usually
work of the Thomas–Fermi theory. How- based on density-functional theory, which
ever, he obtained a positive contribution of has already been mentioned above. The
the metal to the capacity, which was prac- quantum-mechanical self-interactions of
tically the same for all metals. Only some the electron gas – the exchange and the
50 years later, Kornyshev and cowork- correlation energies – are treated in the lo-
ers [30] showed that Thomas–Fermi-like cal density approximation, which is based
models are unsuitable for double-layer the- on the following idea: these interactions
ory because they do not give a realistic are known for an electron gas of con-
description of the work functions and the stant electronic density. As a first ap-
surface potentials of metals. proximation, one can assume that in an
4
Hg
3 Tl
[m2 F −1]
C H−1
Cd
2
In
Zn
1
Ga
0
8 16 24 32
n × 1000
[a.u.]
Fig. 5 Inverse Helmholtz capacity of aqueous solution in contact
with sp metals; experimental values from Trasatti [28]. (The dashed
line is based on a model calculation by Schmickler and
Henderson [26].)
inhomogeneous gas the exchange and cor- in front of the geometrical surface. Thus
relation energies at each point take on a positive value of xim indicates that the
the same values that they would have effective plate separation is smaller than
in a homogeneous gas with the same the physical separation between the two
density. While this approximation is sim- surfaces. Typical values for xim are of the
ple, it has the advantage that it fulfills a order of 0.5 Å; for a normal plate capacitor
number of sum rules that the exact expres- such a small shift of the apparent plate
sion must also obey. Most importantly, position is quite negligible. However, for
it gives good results for the electronic high ionic concentrations the double layer
structures of surfaces, and only for com- can be considered as a capacitor with a
plicated interactions must one resort to plate separation of a few Ångstrøms, so
better mathematical treatments such as that a shift of this order of magnitude has
the generalized gradient approximation. a noticeable effect.
The electronic polarizability of the jel- The relation between the position of
lium surface, which is the relevant quantity the image plane and the metal capacity
for the interfacial capacity, can be ex- is given by
pressed in terms of the effective position ∂χ 1
of the image plane, which has the meaning xim = −ε0 = −ε0 (15)
∂σM Cm
that its name suggests: a test charge placed
in front of the jellium surface experiences As one might expect, the polarizability
an image force as if the image plane were increases with the electronic density (see
a distance xim in front of the surface (see Fig. 6), which qualitatively explains the
Fig. 4). In a metal capacitor the effective corresponding increase of the Helmholtz
positions of the plates lie at a distance xim capacity, though for jellium in vacuum
2.0
1.5
x im
1.0
−10 [Å]
0.5
−5 25
0 20
sM
[µC cm−2] 5 15 n × 103
10 [au]
10
Fig. 6 Variation of the effective position of the image plane with the
electronic density and with the surface charge density, calculated
within the jellium model.
this dependence is relatively weak. It also reproduce trends and orders of magni-
depends on the surface charge density, tudes for polycrystalline materials. In an
and is much higher when there is a improved version the positive background
considerable excess of electrons on the charge is replaced by a lattice of pseu-
surface. dopotentials so that single crystals can be
As already mentioned, the distribu- represented. Usually, the pseudopotentials
tion of electrons at the jellium surface are averaged parallel to the surface in order
entails a surface dipole moment. There- to keep the calculations one-dimensional.
fore, at the pzc a water molecule situated Such models predict sizable differences
within the electronic tail experiences a for the surface polarizabilities, measured
positive field that tends to orient the in terms of xim , of different single-crystal
molecule with its oxygen end toward surfaces, and may thereby explain why the
the metal surface. Conversely, the water interfacial capacity varies with the crystal
molecules tend to enhance the electronic orientation [40, 42]. However, the calcu-
spillover. As a result, the work func-
lated values depend more strongly on the
tion for jellium in contact with water
orientation than do the experimental val-
is lower than in the vacuum. Model
ues; this indicates that the contributions
calculations for this effect have been per-
of the metal and of the solvent are not
formed by several authors [34, 37, 39–41].
independent.
While the results depend on the details
of the model, it is generally agreed that
the larger the overall change in the 2.2.6.3 Calculations for Metal Clusters
dipole potential, the higher the electronic Modern quantum-chemical methods ac-
density. count for the chemical properties of
Jellium is a simple, structureless model the atoms much better than jellium-type
for a metal, and can therefore only models can, but they are usually limited
1.26
1.25
1.25
1.25
[Å]
x im
1.24
1.24
1.23
−20 −10 0 10 20
sM
[µC cm−2]
Fig. 7 Effective position of the image charge for a mercury cluster as
a function of the surface charge density. (Data taken from Ref. [44].)
to treating a finite number of atoms. It 2.2.7

is therefore natural to model an elec- The Metal–Solution Interface
trode surface by a large cluster of metal
atoms. However, even in a large cluster The models presented in the previous two
most of the atoms lie near the surface, sections dealt with just one side of the
so that the charge distribution differs interface. Here, we discuss models for
considerably from that in a semi-infinite the whole metal–solution interface. They
metal. Therefore an early attempt [43] to can be constructed by combining models
estimate the metal contribution to the ca- for the solution and for the metal, but
pacity by quantum-chemical calculations a few attempts have also been made to
for a lithium cluster met with mixed treat the interface as a whole. We present
success: the values obtained for the sur- several approaches in increasing order of
face polarizability were quite large – in the complexity.
case of Li(111) even excessively large. A
new attempt has recently been under- 2.2.7.1 Jellium in Contact with Model
taken by Nazmutdinov and coworkers [44]. Solutions
These authors suggested that the sur- The simplest molecular model for an
face polarizability should not be calculated electrolyte solution is an ensemble of hard
from the change in this surface dipole spheres treated in the MSA. This can be
potential, but from a renormalization of combined with jellium to obtain a model
the density matrix. They chose mercury for the whole interphase [46–48]. The hard
as a model system and obtained values sphere model has been solved at the PZC
for the effective position of the image only, so the combined model is restricted
plane that increase toward negative charge to this point. It is natural to consider the
densities (see Fig. 7) and are quite sim- jellium surface as a hard wall for the
ilar to those obtained from jellium-type electrolyte and add the contributions of
calculations. the hard-sphere electrolyte and jellium to
Fig. 8 Metal–solvent
separation ‘‘s’’ as a function of
the surface charge density for
two different choices of the
pseudopotentials [52, 53].
2.0
(a)
(b)
[A°]
s
1.5
−10 0 10
sM
[µC cm−2]
the inverse capacity. However, in this way contribution:

one obtains negative interfacial capacities 1 σM ∂s
for metals with higher electronic densities, = (16)
Cshift ε0 ∂σM
which are clearly unphysical. Obviously,
the interaction of the metal with the to the inverse capacity. Halley’s
solution affects the response to an external group [49–51] proposed that this is a
field. A simple remedy was proposed by major contribution, which determines the
Schmickler and Henderson [47, 48]: they form of the capacity charge characteristics.
introduced a barrier that repels the metal This idea was taken up by Amokrane
electrons from the solution and reduces and Badiali [52, 53], who constructed a
the surface polarizability. The height of metal–water potential composed of an
the barrier was chosen such that the model attractive, long-range dispersion force and
a short-range repulsion. They performed
gives reasonable values for the capacities
explicit calculations for Ag(111), modeled
of the sp metals – in this way they obtained
as jellium with pseudopotentials, and
the broken line in Fig. 5. Obviously, this
water, and obtained a surprisingly strong
ad hoc procedure is not quite satisfactory,
variation of the water–metal separation
even though the experimental trend is well
s with the surface charge density σM
reproduced. (see Fig. 8). The separation is smaller for
The main problem is the interaction high charge densities, irrespective of their
between the metal and the solution. sign, because the electrostatic pressure
Several groups have tried to improve pushes the water toward the surface. This
on the simple hard-wall model by con- effect is less marked for negative charge
structing semiempirical interaction po- densities, where the electrons spill out
tentials. A particular issue in these ap- further and exert an outward pressure
proaches has been whether the average on the solvent. The absolute values of
distance s between the metal surface the s − σM characteristics depend on the
and the first layer of solvent molecules pseudopotentials for the jellium–water
changes with the charge density. If it interaction, which determine the repulsive
does change, it makes an additional part of the potential, but the shape of these
Fig. 9 Solvent capacity for aqueous

solutions according to the Experimental
phenomenological theory of Amokrane
100
and Badiali [52, 53]. The full lines are
from the experimental data for a
solution of KPF6 , the dashed curves
for NaF.
[µF cm −2]
C
50
Solvent
Solvent
0
−10 0 10
sM
[µC cm−2]
curves is the same for reasonable choices in the sense that different experimental
of these pseudopotentials. capacity data give almost the same solvent
To arrive at a complete double-layer capacities.
model, Amokrane and Badiali proceeded Bérard and coworkers [54] have applied
in a semiempirical manner. Assuming the RHNC to a simple water model and
that the contributions of the metal and combined this with the jellium model. The
the solvent to the inverse capacity are solution and the jellium were assumed to
additive (see Eq. 14), they obtained the interact electrostatically. This leads to a
solvent capacity from experimental data coupled set of equations for the density
for the Helmholtz capacity of Ag(111). In profiles of the electrons and the parti-
this way they obtained a solvent capacity cles in the solution, which was solved
for water, which shows a pronounced self-consistently through an iterative pro-
maximum near the pzc (see Fig. 9); this cedure. A critical issue in this coupling is
curve can be fitted to a model in which the the distance of closest approach of the
solvent is represented by a layer of dipoles. solvent molecules to the jellium edge,
This procedure can be carried fur- modulated by the repulsion of the inhomo-
ther by taking experimental data for the geneous electron gas from the orbitals of
Helmholtz capacities of other metals and the solvent molecules. Bérard and cowork-
extracting the metal contribution by as- ers [54] allow the solvent spheres to come
suming that the solvent part is not affected into contact with the jellium edge. More-
by the metal. In this way, Amokrane and over, owing to the presence of the solvent,
Badiali arrive at a consistent interpreta- the electron density spills further out of the
tion of all experimental data. However, metal, dropping to 0.1% of the bulk value at
their theory hinges on the calculations for 3 Å. At an uncharged metal, the resulting
Ag(111). In fact, the variation of the cal- strong coupling causes the solvent dipoles
culated metal capacity is so strong that it in the layer contacting the metal to have a
practically determines the solvent capacity small net orientation normal to the metal
Fig. 10 Oxygen (solid line) and

hydrogen (dashed line) densities as a
4 function of the distance from the center
of the simulation box. The densities
Densities
3 have been normalized to unity in the

bulk. (Data taken from Ref. [56].)
2
−10 −5 0 5 10
z
[Å]
surface, with the positive ends out. This such complications increase the computer
is in agreement with the results of several time.
computer simulations that we will discuss The interaction of water with the metal
below. has been modeled in various ways: hard-
wall, Lennard–Jones, and Morse poten-
2.2.7.2 Molecular Dynamics Simulations tials have been used, as well as potentials
Realistic models for solvents such as derived from ab initio cluster calculations.
water are too complicated to be treated Whatever interaction is chosen, the pres-
by anything but computer simulations, ence of the metal always induces a layering
and molecular dynamics has usually been of the water molecules, similar to that seen
the method of choice. Most studies have in the hard-sphere ensemble. As an ex-
been limited to pure water in contact ample, we show the density profiles for
with a metal surface, but recently a the oxygen and hydrogen atoms obtained
few simulations have been performed for by Foster and coworkers [56] in Fig. 10.
concentrated aqueous solutions. We will The corresponding simulations were per-
not attempt to survey this very active field, formed in the absence of an external field,
but give a few representative results. that is, for the pzc, and in the z-direction the
There are a variety of semiempirical water was contained between two identical
models for water; most of them model walls, while periodic boundary conditions
the molecule by a few point charges were applied in the xy directions. As a
with a Lennard–Jones or similar potential result of the layering effect, the density pro-
superimposed. For example, the popular files exhibit a series of oscillations, whose
rigid SPC/E model [55] consists of three amplitude diminishes toward the bulk. In
point charges of −0.8476e0 on the oxygen the first two layers the profiles for oxygen
and 0.4238e0 on the hydrogen sites (e0 is and hydrogen differ somewhat: on an av-
the proton charge) and a Lennard–Jones erage, the oxygen atoms are a little closer
potential centered at the oxygen atom. to the electrode surface than the hydrogen
The resulting dipole moment is somewhat atoms. Therefore the dipole moment of
larger than that of the isolated water the water molecules has a small net ori-
molecule in the gas phase in order to entation along the direction of the surface
account for the polarizability. Other water normal, which produces a surface dipole
models have a built-in polarizability, but potential χ. The magnitude of this effect
[V]
χ
0
−1
−2
−10 −5 0 5 10
z
[Å]
Fig. 11 Surface potential χ as a function of the distance from the center
of the simulation box. (Data taken from Ref. [44].)
depends on the interaction between the the water model employed and on the in-
metal and water; Fig. 11 shows the dipole teraction of the first layer with the metal
potential for a simulation of water in con- surface. Xia and Berkowitz [61] suggest
tact with a mercury electrode. The total that at very high fields water undergoes
potential drop between the electrode sur- a phase transition to an icelike ordered
face and the bulk is of the order of several structure.
tenths of an electron volt, which agrees The simulations reported above have
qualitatively with estimates of Trasatti [57] been performed for pure water. Modeling
based on experimental data. the double layer requires the introduction
The detailed structure of water at the of ions, which are usually represented as
surface is often discussed in terms of charged spheres with a Lennard–Jones
potential. Given the limited size of the
the bilayer model proposed by Doering
ensembles that can be handled, a statisti-
and Madey [58] to explain water adsorp-
cally meaningful number of ions results
tion on metal surfaces in the vacuum at
in a high concentration; typically of the
low temperatures. As the name suggests,
order of 2 M. At present, a few such sim-
this model proposes two layers of water
ulations have been reported [62, 63], and
that are hydrogen-bonded to each other. more can be expected in the future as
Obviously, in the liquid state and at ambi- computing power grows. The results ob-
ent temperatures such a bilayer must be tained so far are still tentative, but generally
strongly disturbed, though a vestige seems seem to be in line with physicochemical
to appear in some simulations [59]. Simi- intuition and the notions developed by
lar, but more icelike, structures have been Grahame [64, 65] and Parsons [66], though
suggested by Krämer and coworkers [60]. a few details are surprising. Thus, the sol-
When the electrode surface is charged vation of the ions plays a central role in
or, equivalently, an electric field is intro- the double-layer structure. In particular,
duced into the simulation box, the water the fact that the anions tend to be less
molecules tend to align with the field; the strongly solvated than the cations induces
extent to which this happens depends on an asymmetry with respect to the pzc. As
152
4
Oxygen density
Ionic densities
3 3
␴=0 ␴=0
2 2
1 1
0 0
Ionic densities Ionic densities

10.0
7.5
7.5 ␴ = 9.9 µC cm−2 ␴ = −9.9 µC cm−2

5.0
5.0
2.5
2.5
0.0 0.0
0 5 10 15 20 25 0 5 10 15 20 25
z z
[Å] [Å]
Fig. 12Distribution of the ions at an idealized metal surface with various surface charge densities. Dotted lines: cations (Na+ ), full lines: anions (Cl− ).
For comparison, the figure (a) shows the distribution of the oxygen atoms at the pzc.
an example, we show in Fig. 12 the den- and interaction potentials derived from
sity profiles of Na+ and Cl− obtained by quantum-chemical calculations.
Spohr [63] at the pzc and at surface charge The system is investigated by molecular
densities of σ = ±9.9 µF cm−2 . At the pos- dynamics, performing a self-consistent
itively charged surface the first peak for calculation for the electronic subsystem
the weakly solvated anions is closer to the at each step. The presence of the solvent
metal (in this case mercury) surface than is found to have a pronounced effect on
the peak for the cation is at the negative the electronic distribution on the metal
surface. However, at the pzc the sodium surface. In particular, the simulations
ions are closer to the surface because their clearly show the formation of image
solvation shell can be accommodated be- charges on the metal surface.
tween the first and second water layers. This and similar studies from this group
The total potential drop at the interface is provide a very detailed description of
determined both by the distribution of the the interface, to which we cannot give
ions and by the orientation of the water; full justice in this limited space. Of
these two contributions tend to cancel each course, such comprehensive simulations
other, so there is a complicated interplay. are presently of a preliminary nature, but
While it is still too early to obtain the they do indicate the important physical
interfacial capacity from such simulations, processes that occur at the interface. We
first estimates give a reasonable order of may expect more work of this type to be
magnitude [63]. performed in the future.
2.2.7.3 Car–Parinello-type Simulations 2.2.8

In the simulations reported above the Rough Electrodes
interactions between the various compo-
nents were mostly based on semiempir- Most of the theoretical work has been re-
ical potentials. In contrast, Price, Halley, stricted to perfect single-crystal electrodes.
and their collaborators attempt to model However, in practical applications such as
the whole interface in the spirit of the fuel cells, in which the electrode acts as a
Car–Parinello method [16]. One of the first catalyst, it is desirable to maximize the sur-
systems investigated was the interface be- face area by using rough electrodes. Very
tween a copper electrode and water [67]. rough electrodes can be modeled as frac-
For this purpose these authors set up tals [68], but a double-layer theory for such
a simulation cell with approximate di- electrodes has remained elusive. The op-
mensions of 42 Å × 15 Å × 15 Å. Each cell posite limit of small roughness has been
contained a slab of copper atoms that treated by Daikhin and coworkers [69] at
were five-layers thick, the two surfaces hav- the GC level.
ing (100) structure. The remaining space These authors start by defining a rough-
was filled with water molecules. Cyclic ness function R̃(κ) through the relation
boundary conditions were applied in all
directions. Obviously, an ab initio, all elec- C = R̃(κ)CGC (17)
trons calculation is quite impossible for
such a system, and may not even be where C is the actual capacity, CGC the GC
desirable. Instead, Halley and collabora- value, and κ is the Debye inverse length.
tors used a mixture of pseudopotentials Effectively, the Debye length 1/κ defines
the yardstick with which the surface area uneven distribution of the surface charge
is measured. Therefore, must also be considered: the charge
accumulates on the surface protrusions.
lim R̃(κ) = 1 and lim R̃(κ) = R Then the roughness function becomes
κ→0 κ→∞
(18) quite involved, and depends also on the
where R is the geometrical roughness electrode potential [70].
of the electrode, which is defined as the
ratio between the true and the apparent 2.2.9
surface areas. If the z-direction is taken Liquid–Liquid Interfaces
as perpendicular to the average surface,
then, for a moderately rough electrode, the In a certain sense, interfaces between two
true surface is given by a single-valued liquids are simpler than those between
function: a metal and a liquid, since they do not
z = ζ (x, y) (19) involve solid-state properties. However, for
a long time the structure of liquid–liquid
When the roughness is small it can interfaces has remained a matter of
be treated as a perturbation, and the controversy. Essentially, there are two
roughness function can be expressed different views: one holds that the interface
through the Fourier transform (FT) of is sharp and contains a compact layer
the surface. This theory is simple only of solvent molecules into which the ions
near the pzc; away from the pzc the cannot penetrate. The other view posits the
Fig. 13 Schematic picture of the interface between two immiscible

liquids.
existence of an extended boundary layer in where φi∞ is the value of the poten-
which the two solutions mix. tial in the indicated phase far from
During the last few years, the latter view the interface. The capacity can be de-
has received more supporting evidence. composed into a series combination of
Already the early experimental work of two GC capacities Ci , one for each
Girault and Schiffrin [71], who determined interface:
the surface excess of water at the interface
εi ε0 zi e0 (φ s − φi∞ )
with 1,2-dichloroethane, had indicated the Ci = cosh (21)
existence of a mixed boundary layer. LiD 2kT
Recent X-ray scattering experiments [72]
where φ s is the potential at the idealized,
indicate an average interfacial width of the
flat interface (cf. Eq. 1). The latter must be
order of 3 to 6 Å. These experiments are
calculated from the Poisson–Boltzmann
in line both with model calculations based
equation [77].
on the density functional formalism [73]
The Verwey–Niessen model gives a rea-
and with computer simulations [74, 75].
sonable first approximation, but at a closer
Accordingly, the interface is best visualized
glance significant deviations are observed
as rough on a molecular scale as indicated
even at low ionic concentrations, where
in Fig. 13.
this theory is expected to hold. Surpris-
So far, there have been no molecular dy-
ingly, at low (≤10−2 M) concentrations
namics simulations with a larger number
the experimental capacity is often higher
of ions for these interfaces, and the double-
than that predicted by Verwey–Niessen
layer theory rests on analytical calculations
theory – this is just the opposite to the be-
and on simulations of the lattice gas model
havior of metal–solution interfaces, where
(LGM).
the capacity is always lower than the
GC value.
2.2.9.1 Verwey–Niessen Theory These deviations were first explained by
The basic system consists of two immisci- the presence of a compact, ion-free layer
ble solvents, each of which contains a salt at the interface; this is known as the mod-
that is badly solvable in the other solvent. ified Verwey–Niessen model. Obviously,
When a potential difference is applied be- the presence of an ion-free layer can only
tween the two solutions, opposing space reduce the capacity, so the theory had to
charge layers form on both sides of the be modified further. For a few systems
interface and give rise to a capacity. The a consistent interpretation of the experi-
simplest model consists of two diffuse dou- mental capacity was achieved [78–80] by
ble layers back to back, which are described combining this model with the so-called
by the GC theory – in the context of liq- modified Poisson–Boltzmann (MPB) the-
uid–liquid interfaces this model is known ory [81], which attempts to correct the GC
as the Verwey–Niessen theory [76]. theory by accounting for the finite size
The interfacial capacity per unit area at of the ions and for image effects, while
the interface between two solutions labeled the solvent is still treated as a dielectric
(1) and (2) is defined as continuum. The combined model has an
adjustable parameter, so it is difficult to
dσ judge whether the agreement with exper-
C= (20)
d(φ2∞ − φ1∞ ) imental data is significant. The existence
of a compact solvent layer is just a postu- which these authors handle by a pertur-
late, and is difficult to reconcile with the bation expansion treating the roughness
extended boundary layer discussed in the as a small parameter. Following Daikhin
previous section. and coworkers [69] (see previous section),
Pecina and Badiali express the deviation
2.2.9.2 Capillary Waves from the Verwey–Niessen theory through
Since liquid–liquid interfaces are rough, a roughness function R̃(LiD , εi ), which here
the ideas for rough electrodes presented depends on the dielectric constants of
the two solutions and on their Debye
above can be applied, though with some
lengths. Explicit expressions for the rough-
modifications. Again, if this roughness
ness function can be given for simple
is not too large, the interface can be
surface corrugations. While Pecina and
represented by a function ζ (x, y) at a
Badiali consider small deviations from the
particular moment of time. The two-
pzc, the nonlinear case has been treated by
dimensional Fourier components A(q) of
Urbakh and coworkers [83].
this surface function are known as capillary
The concept of capillary wave explains
waves; q is a two-dimensional wave vector.
the increase of the interfacial capacity
Thus, a rough surface can be considered as
beyond the Verwey–Niessen limit, and
a superposition of capillary waves, which
thus helps in understanding the structure
are present at all wavelengths. Of course,
of the interface. A quantitative interpre-
these cannot be observed as traveling
tation of experimental data within this
waves, but only as a fluctuating surface
model is, however, difficult since there are
roughness.
other causes for the deviations from Ver-
The energy stored in a capillary wave
wey–Niessen theory besides the surface
is proportional to the surface area it roughness.
creates. This entails a prediction for the
distribution of the amplitudes:
2.2.9.3 Lattice Gas Model
kT The Ising model is one of the most
A(q)A∗ (q) = (22)
γ |q|2 versatile models in physics and chemistry,
and in the equivalent form of the LGM
where the angular brackets denote ther- it has also been applied to liquid–liquid
mal expectation values. This relation has interfaces. In this case it is based upon a
indeed been observed in computer simu- three-dimensional, typically simple cubic
lations of liquid–liquid interfaces [74, 75]. lattice. Each lattice site is occupied by one
The concept of capillary waves can be of a variety of particles. In the simplest
used to explain how the surface roughness case the system contains two kinds of
increases the interfacial capacity beyond solvent molecules, and the interactions
the Verwey–Niessen value. For this pur- are restricted to nearest neighbors. If we
pose, Pecina and Badiali [82] have solved label the two types of solvent molecules
the linear Poisson–Boltzmann equation S1 and S2 , the interaction is specified by
across the interface between two solu- a symmetric 2 × 2 matrix wij , where each
tions with different dielectric constants element specifies the interaction between
and Debye lengths separated by a corru- two neighboring molecules of type Si
gated surface. A major difficulty is the and Sj . Whether the system separates
boundary condition at the rough interface, into two phases or forms a homogeneous
mixture depends on the relative strength and a salt K2 + A2 − that is preferentially

of the cross-interaction w12 with respect dissolved in solvent 2. This can be
to the self-interaction terms w11 and achieved by choosing suitable interaction
w22 , which can be expressed through the parameters between the ions and the two
combination solvents.
In addition to the nearest-neighbor
w = w12 − (w11 + w22 )/2 (23) interaction, each ions experiences the
electrostatic potential generated by the
If w/kT is large, the system separates into other ions. In the literature this has
two phases: a phase 1, which contains generally been equated with the macro-
mainly solvent molecules S1 and has scopic potential φ calculated from the
only a low concentration of S2 , and a Poisson–Boltzmann equation. This cor-
phase 2, which is mainly composed of responds to a mean-field approximation,
S2 . If w is small, entropy wins and the in which correlations between the ions
systems forms one phase. If the system are neglected. The better this approxi-
separates, it can be extended to a model mation, the lower the concentrations of
for the interface between two solutions the ions.
by introducing ions. In the basic case The lattice gas can be treated both
the system contains a salt composed by analytical approximations such as the
of cations K1 + and anions A1 − , which mean-field or the quasi-chemical approx-
is preferentially solvated by the solvent imation, or by MC simulations. Both
S1 , but badly solvable in solution 2, methods complement each other: the
2.0
1.5
Particle density
Anions Cations
1.0
Cations
Anions
0.5
0.0
−4 −2 0 2 4
z L−1D
Fig. 14 Normalized distribution c(x)/c0 of the ions for a potential
drop of e0 φ/kT = 1. The full lines give the distribution of the
majority ions; in the region x < 0 these are the anions, in x > 0
these are the cations. The dashed lines give the distributions of the
counterions. The calculations have been performed in the
quasi-chemical approximation for equal Debye lengths and
dielectric constants in both solutions [84].
simple analytical approximations work 2.2.10

best at low ionic concentrations, while Conclusion
the simulations, because of the finite en-
semble size, are more suitable for higher All electrochemical processes take place in
concentrations. The quasi-chemical ap- the double-layer region. Therefore, elec-
proximation has been employed by Pereira trochemists need to know the distribution
and coworkers [84] and MC simulations of the various particles and of the elec-
have been employed by Huber and cowork- trostatic potential and the charge near the
ers [85]. Both works give similar results. interface in detail. After more than a cen-
The most important effect is that the tury of double-layer theory, and more than
roughness of the interface entails an over- 50 years after the seminal paper by Gra-
lap of the double-layer regions pertaining hame [64], it is pertinent to ask how far
to the two solvents (see Fig. 14). This over- we have succeeded. Of course, the follow-
lap decreases the average distance between ing remarks are the personal view of the
the two opposing charges, and hence it author.
increases the capacity. At higher ionic con- The hard-sphere model, introduced in
centrations the finite size of the particle the 1980s, was a major step forward: it
becomes noticeable, an effect that makes demonstrated the existence of an extended
the capacity smaller. Hence, depending on boundary layer at the solution side of
the concentrations and on the various in- the interface and gave a good estimate
teraction parameters, the capacity can be for the contribution of the solution to
larger or smaller than that predicted by the the interfacial capacity. However, it was
Verwey–Niessen theory; a few examples solved within the rough MSA, which
are shown in Fig. 15. Further details of the holds only for small excess charges.
LGM can be found in a recent article by Within the integral-equations approach,
Schmickler [86]. further progress has been slow. So our
2.0
1.5
(3)
[F m −2]
(2)
1.0
C
(1)
0.5
0.0
−0.25 −0.15 −0.05 0.05 0.15 0.25
∆f
[V]
Fig. 15 Interfacial capacity in the lattice gas model for various
ion–solvent interactions. The data were obtained from an MC
simulation [85]. The crosses denote the Gouy–Chapman capacity.
main knowledge of the distribution of Acknowledgments

particles in the solution rests on computer
simulations. With increasing computing I would like to thank Dr. E. Spohr, Jülich,
power, it has become possible to include for useful discussions on simulations, and
ions in the ensemble. To a large extent, for sending me the data of Fig. 10.
these simulations support the notions
presented in Grahame’s review: small References
cations are less likely to approach the metal
surface than anions, the first layers of 1. G. Gouy, Compt. Rend. 1910, 149, 654.
water have a more rigid structure, and the 2. D. L. Chapman, Philos. Mag. 1913, 25(6), 475.
3. O. Stern, Z. Elektrochem. 1924, 56, 508.
water dipoles are oriented by the electric 4. R. Parsons, F. G. R. Zobel, J. Electroanal.
field. However, the simulations do provide Chem. 1965, 9, 333.
additional insight into the water structure 5. D. Eberhardt, E. Santos, W. Schmickler, J.
and the solvation of ions near the interface. Electroanal. Chem. 1996, 419, 23.
This is true both for metal–solution and 6. R. J. Watts-Tobin, Philos. Mag. 1961, 6, 133.
7. B. B. Damaskin, A. N. Frumkin, Electrochim.
for liquid–liquid interfaces.
Acta 1974, 19, 173.
Another advance in the 1980s was 8. R. Parsons, J. Electroanal. Chem. 1975, 21,
the recognition that the metal surface 229.
contributes to the interfacial capacity. This 9. R. Guidelli in Trends in Interfacial Electro-
explains why the capacity varies so strongly chemistry (Ed.: A. F. Silva), Riedel, Dordrecht,
1986, pp. 387–452.
with the nature of the metal. Calculations
10. R. Guidelli, J. Electroanal. Chem. 1981, 123,
based on the jellium model gave an 59.
estimate for the order of magnitude of the 11. R. Guidelli, J. Electroanal. Chem. 1986, 197,
metal contribution, and reproduced the 77.
main trends. But again, further progress 12. D. R. Bérard, G. N. Patey, J. Chem. Phys.
1992, 97, 4372.
has been slow, and so far the methods of
13. G. M. Torrie, P. G. Kusalik, G. N. Patey, J.
modern computational chemistry have not Chem. Phys. 1988, 88, 7826.
contributed much to our understanding of 14. M. P. Alleen, D. J. Tildesley, Computer Sim-
the double layer. ulations of Liquids, Clarendon Press, Oxford,
So the advances during the last 50 1987.
15. K. Binder, Monte Carlo Methods in Statistical
years have been mainly in our qualitative
Physics, Topics in Current Physics, Springer,
understanding of the double layer. Given Berlin, 1986, Vol. 7.
the complexity of the interface, this is no 16. R. Car, M. Parinello, Phys. Rev. Lett. 1985, 55,
small achievement. However, quantitative 2471.
results are clearly desirable. Most probably 17. N. H. March, Electron Density Theory of Atoms
and Molecules, Academic Press, London,
they will come from further advances
1992.
in simulation techniques and computing 18. S. Lundquist, N. H. March, (Ed.), Theory of
power, while the contribution of proper the Inhomogeneous Electron Gas, Plenum
theory will diminish. This trend toward Press, New York, 1983.
more numerical works may be regrettable, 19. J. Stafiej, J. P. Badiali, Z. Borkowska, J. Elec-
troanal. Chem. 1995, 395, 1.
but is unavoidable. In any case, there
20. J. Stafiej, M. Dymitrowska, J. P. Badiali, Elec-
are many open questions in double-layer trochim. Acta 1996, 41, 2107.
modeling, and it will stay an active field of 21. D. Boda, D. Henderson, J. Chem. Phys. 2000,
research for years to come. 112, 8934.
22. D. Henderson, J. Chem. Phys. 2000, 112, 50. J. W. Halley, B. Johnson, D. Price et al., Phys.
6716. Rev. 1985, B31, 7695.
23. A. Hamelin, S. Röttgermann, W. Schmickler, 51. J. W. Halley, B. Johnson, D. Price et al., Phys.
J. Electroanal. Chem. 1987, 230, 281. Rev. 1988, 38, 9357.
24. S. L. Carnie, D. Y. C. Chan, J. Chem. Phys. 52. S. Amokrane, J. P. Badiali, Electrochim. Acta
1980, 73, 2949. 1989, 34, 39.
25. L. Blum, D. Henderson, J. Chem. Phys. 1981, 53. S. Amokrane, J. P. Badiali, J. Electroanal.
74, 1902. Chem. 1989, 266, 21.
26. W. Schmickler, D. J. Henderson, J. Chem. 54. D. R. Bérard, M. Kinoshita, X. Ye et al., J.
Phys. 1984, 80, 3381. Chem. Phys. 1994, 101, 6271.
27. D. Boda, D. Henderson, J. Chem. Phys. 1998, 55. H. J. C. Berendsen, J. R. Grigeria, T. P.
109, 7362. Straatsma, J. Phys. Chem. 1987, 91, 6269.
28. S. Trasatti, J. Electroanal. Chem. 1981, 123, 56. K. Foster, R. Raghavan, M. Berkowitz, Chem.
2031. Phys. Lett. 1989, 162, 32.
29. O. K. Rice, Phys. Rev. 1928, 31, 105. 57. S. Trasatti in Trends in Interfacial Electro-
30. A. A. Kornyshev, W. Schmickler, M. A. chemistry (Ed.: A. F. Silva), Reidel, Dordrecht,
Vorotyntsev, Phys. Rev. B 1982, 25, 5244. 1986.
31. J. R. Smith, Phys. Rev. 1969, 181, 522. 58. D. L. Doering, T. E. Madey, Surf. Sci. 1982,
32. N. D. Lang, W. Kohn, Phys. Rev. B 1970, 1, 123, 305.
4555. 59. J. P. Valleau, A. A. Gardner, J. Chem. Phys.
33. N. D. Lang, W. Kohn, Phys. Rev. B 1971, 3, 1987, 86, 4162.
1215. 60. A. Krämer, M. Vossen, F. Forstmann, J.
34. J. P. Badiali, M. L. Rosinberg, J. Goodisman, Chem. Phys. 1997, 106, 2792.
J. Electroanal. Chem. 1983, 143, 73.
61. X. Xia, M. L. Berkowitz, Phys. Rev. Lett. 1995,
35. J. P. Badiali, M. L. Rosinberg, J. Goodisman,
74, 3193.
62. M. R. Philpott, J. N. Glosli, S. B. Zu, Surf.
36. W. Schmickler, J. Electroanal. Chem. 1983,
Sci. 1995, 335, 422.
150, 19.
63. E. Spohr, Electrochim. Acta 1999, 44, 1697.
37. W. Schmickler, D. J. Henderson, J. Chem.
64. D. C. Grahame, Chem. Rev. 1947, 41, 441.
Phys. 1984, 80, 3381.
65. D. C. Grahame, R. Parsons, J. Am. Chem.
38. W. Schmickler, D. J. Henderson, Prog. Surf.
Sci. 1986, 22, 323. Soc. 1961, 83, 1291.
39. W. Schmickler, Chem. Phys. Lett. 1983, 99, 66. R. Parsons in Comprehensive Treatise of Elec-
135. trochemistry (Eds.: J. O’M. Bockris, B. E. Con-
40. E. Leiva, W. Schmickler, Proc. Indian Acad. way, E. Yeager), Plenum Press, New York,
Sci. 1986, 97, 267. 1980, Vol. 1.
41. M. I. Rojas, E. P. M. Leiva, J. Electroanal. 67. D. L. Price, J. W. Halley, J. Chem. Phys. 1995,
Chem. 1991, 303, 55. 102, 6603.
42. E. Leiva, W. Schmickler, Surf. Sci. 1993, 291, 68. B. B. Mandelbrot, The Fractal Geometry of
226. Nature, Freeman, San Francisco, 1982.
43. W. H. Mulders, J. H. Sluyters, J. H. van 69. L. I. Daikhin, A. A. Kornyshev, M. Urbakh,
Lenthe, J. Electroanal. Chem. 1989, 261, 273. Phys. Rev. E 1996, 53, 6192.
44. R. R. Nazmutdinov, M. Probst, K. Heinz- 70. A. Kornyshev, private communication.
inger, Chem. Phys. Lett. 1994, 222, 101. 71. H. H. Girault, D. H. Schiffrin, J. Electroanal.
45. J. Böcker, R. E. Nazmutdinov, E. Spohr et al., Chem. 1983, 150, 43.
Surf. Sci. 1995, 333, 372. 72. A. M. Tikhonov, D. M. Mitrinovic, M. Li
46. J. P. Badiali, M. L. Rosinberg, F. Vericat et al., et al., J. Phys. Chem. B 2000, 200, 6336.
J. Electroanal. Chem. 1985, 158, 82. 73. D. Henderson, W. Schmickler, J. Chem. Soc.,
47. W. Schmickler, D. J. Henderson, J. Phys. Faraday Trans. 1996, 92, 3839.
Chem. 1984, 80, 3381. 74. I. Benjamin, Chem. Rev. 1996, 96, 1449.
48. W. Schmickler, D. J. Henderson, J. Phys. 75. I. Benjamin, Annu. Rev. Phys. Chem. 1997,
Chem. 1985, 82, 2925. 48, 407.
49. D. Price, J. W. Halley, J. Electroanal. Chem. 76. E. J. W. Verwey, K. F. Niessen, Philos. Mag.
1983, 150, 347. 1939, 28, 435–446.
77. W. Schmickler, Interfacial Electrochemistry, 83. M. Urbakh, A. A. Kornyshev, L. I. Daikin, J.

Oxford University Press, New York, 1996. Electroanal. Chem. 2000, 483, 68.
78. T. Wandlowski, K. Holub, V. Marecek et al., 84. C. M. Pereira, W. Schmickler, A. F. Silva
Electrochim. Acta 1996, 40, 1897–2887. et al., Chem. Phys. Lett. 1997, 268, 13.
79. Q. Cui, G. Zhu, E. Wang, J. Electroanal. 85. T. Huber, O. Pecina, W. Schmickler, J. Elec-
Chem. 1994, 372, 15. troanal. Chem. 1999, 467, 203.
80. Q. Cui, G. Zhu, E. Wang, J. Electroanal. 86. W. Schmickler The lattice gas and other
Chem. 1995, 383, 7. simple models for liquid-liquid interfaces
81. C. W. Outhwaite, L. B. Bhuyiyan, S. Levine, in Liquid Interfaces in Chemical, Biologi-
J. Chem. Soc., Faraday Trans. 2 1980, 76, cal and Pharmaceutical Applications (Ed.:
1388. A. G. Volkov), Marcel Dekker, New York,
82. O. Pecina, J. P. Badiali, Phys. Rev E 1998, 58, 2000.
6041.
2.3 boundary layer can have any configura-

Electrochemical Double Layers: tion and the tales of this distribution
Liquid–Liquid Interfaces for different particles can even overlap
in the boundary layer (Fig. 1a, c). At a
Alexander G. Volkov completely polarizable interface, charge
Oakwood College, Huntsville, Alabama
transfer between bulk phases becomes im-
Vladislav S. Markin possible.
University of Texas, Dallas, Texas A different concept of an ideally polariz-
able interface was formulated by Koenig [7]
2.3.1 and represents a particular case of a com-
Polarizable and Nonpolarizable Liquid pletely polarizable interface (Fig. 1b, d).
Interfaces Two homogeneous phases α and β are
separated by a transitional layer whose
The electrical double layer (edl) at the properties gradually change from the prop-
oil–water interface is a heterogeneous erties of phase α to those of phase β. A
interfacial region that separates two bulk Koenig’s surface, which is impermeable to
phases of polarized media and maintains a all charged particles (both ions and elec-
spatial separation of charges. EDLs at such trons), is drawn in the transitional layer.
interfaces determine the kinetics of charge This surface represents an infinitely thin
transfer across phase boundaries, stability and infinitely high-energy barrier for all
and electrokinetic properties of lyophobic the charged particles involved. It is this
colloids, mechanisms of phase transfer or assumption that distinguishes an ideally
interfacial catalysis, charge separation in polarizable interface from a completely
natural and artificial photosynthesis, and polarizable surface, which makes it irrel-
heterogeneous enzymatic catalysis [1–5]. evant that charged particles are contained
The interface between two immiscible in each phase. Koenig’s definition cannot
electrolyte solutions (ITIES) can be either be described thermodynamically because
polarizable or nonpolarizable, depending it suggests a definite structure of the
on permeability to charged particles. If interfacial region and employs nonthermo-
the interface is relatively impermeable it dynamic assumptions about the existence
is called polarizable, otherwise it is called of an infinitely high-energy barrier for
nonpolarizable or reversible. Planck [6] in- all charged particles. The term ‘‘ideally
troduced a rigorous thermodynamic con- polarizable interface’’ was introduced by
cept called a completely polarizable interface, Grahame and Whitney [8], whereas Koenig
defined as ‘‘an interface whose state is used Planck’s term. Frumkin demon-
completely determined by the charge that strated that a completely polarized elec-
has passed through it beginning from trode, unlike an ideally polarizable one,
a given instant’’. Planck’s definition as- can include systems with local adsorp-
sumes that direct current cannot flow tion charge transfer. Although local charge
through the interface. There can be only transfer exists at the interface and Koenig’s
the transitive current in the boundary surface cannot be drawn, there is no actual
layer. However, the definition says nothing charge transfer between the phases, so the
about how charged particles from different electrode is completely polarizable accord-
phases are distributed in this transitional ing to Planck, but nonideally polarizable
layer. The distribution of particles in the according to Koenig.
2.3 Electrochemical Double Layers: Liquid–Liquid Interfaces 163
Phase α
+ − + −
−
− −
− + −
− − − − − − − −
− −
Interface Koenig surface
+ + + + + + + +
+
+
+
Phase β −
−
+ −
(a) (b)
ace
surf
nig
C αj C αj Koe
C kβ C kβ
C
Phase α Phase β Phase α Phase β
(c) x (d) x
Fig. 1An arbitrary scheme of the structure of a perfectly polarizable interface between two
immiscible liquids (a, b, c, d) and an ideally polarizable interface (b, c).
If phases α and β contain at least one properties of some interfaces are very close
common ion that can freely pass across the to being ideal.
interface, the interface may be called re-
versible or nonpolarizable (Fig. 2). Although 2.3.2
the latter term is commonly accepted, it Verwey–Niessen Model
does not correctly describe the state of the
system. A reversible (nonpolarizable) in- Verwey and Niessen first described the
terface can be partially polarizable or com- EDL at ITIES as two noninteracting diffuse
pletely nonpolarizable. A completely non- layers, one at each side of the interface [9].
polarizable interface containing at least Both solvents were assumed to be struc-
one common ion can pass a high current tureless media with macroscopic dielectric
in either direction without causing a devi- permittivities, and potential distribution in
ation of the interfacial potential difference the EDL was defined by Gouy–Chapman
from the equilibrium value. Although, theory.
in practice we encounter neither ideally Gavach and coworkers [10] extended
polarizable nor completely nonpolarizable the Verwey–Niessen model (MVN) by
interfaces, under certain conditions the introducing an ion-free transition layer
Fig. 2 Scheme of the structure of a

+ − nonpolarizable interface between two
− +
− − immiscible electrolyte solutions (a) and
− +
+ the distribution of the concentrations of
+ −
α − − β ions (b) capable of passing from one
− + + phase to another.
−
−
+ +
+ +
(a)
α c kβ
c αj
c βj
C
c αk β
(b) X
at the ITIES. This is directly analo- on different sides of an ideal interface is

gous to the compact layer (or the in- usually called the ion-free layer.
ner Helmholtz layer) in classical elec- The maximum electrical potential in the
trochemistry. Stern theory was extended compact layer αβ φi includes a dipolar po-
to ITIES, and the final model is re- tential αβ g, which is shown schematically
ferred to as the modified Verwey–Niessen as a narrow region at the sharp interface. A
model. dipolar layer can be located not only in the
In the MVN model, the EDL consists of compact layer but can also occupy part of
two diffuse ion layers back-to-back, which the diffuse layer. The amplitude and sign
produce a compact inner layer between the of αβ g can differ from the total interfacial
two phases (Fig. 3). Dielectric permittivity potential. Figure 4 illustrates four possi-
of the medium at any point in the diffuse bilities for potential distribution at ITIES.
layer is assumed to be constant and equal Generally, the dipolar potential depends
to the bulk phase value ε. The compact on the total interfacial potential αβ φ.
layer or inner Helmholtz layer is located The interfacial potential difference con-
between −δ α and +δ β . In a more detailed sists of the sum of the potential drops:
analysis, the dielectric permittivities in β
both parts of the compact layer are εhα αβ φ ≡ φ α − φ β = φd − φdα + αβ φh (1)
β
and εh . Charge density at each side of the EDL is
In the MVN model, the boundary of the usually called the interfacial free charge
compact layer (called the outer Helmholtz density, and depends on the model
plane) is at the distance of closest ion chosen for the interphase. In an ideal
approach to the interface. For ions with polarized electrode, the free charge equals
different radii, a few outer Helmholtz the total thermodynamic charge in the
planes can be introduced as necessary. Lippmann equation of electrocapillarity if
In the absence of specific adsorption, all components of an interface have the
the compact layer located between two same charges as they have in the bulk
nearest outer Helmholtz planes disposed phases.
εα ε αh ε β εβ
h
φ αd
∆βαφ h ∆βαφ
∆αβ g
φ dβ
α β
−δ α 0 δβ
Fig. 3 Distribution of the electric potential in the electric
double layer at the interface between two immiscible electrolyte
solutions.
φ φ
X X
(a) (b)
φ φ
Fig. 4 Different profiles of

Galvani potential differences at X X
the interface between two (c) (d)
immiscible electrolyte solutions.
We will denote surface charge density From the Gouy–Chapman theory for a
β
in the diffuse layers as qdα , qd and in the 1 : 1 electrolyte, it follows that:
compact layer as qh . From electroneutrality 2ε0 RT F φd
of the interphase, we have: qd = − εκ sinh (4)
F 2RT
β
qdα + qh + qd = 0 (2) where κ is the Debye length,

At the potential of the free zero charge 2F 2 c0
(fzcp), the compact layer has not adsorbed κ= (5)
εε0 RT
ions and the charges of the diffuse layers
are equal to zero: and c0 is the bulk electrolyte concentration
β
in the corresponding phase.
φdα = φd = 0 and αβ φh = αβ φ f zcp (3) If a Helmholtz layer does not have
charges, potential drops in the two diffuse
The dipolar potential αβ φh also needs
layers have a simple relation:
the index fzcp because of its dependence
on total interfacial potential. Usually, the β
F φdα F φd
dependence is weak and Eq. (3) is used to εα κ α sinh + εβ κ β sinh =0
2RT 2RT
determine the dipolar potential. (6)
The relation between potentials of two RT

φdα =
diffuse layers depends on a parameter: F
(εα κ α /εβ κ β )
ε βκβ εβ c0β + exp[F (αβ φ − αβ φh )/2RT ]
ηα/β = α α ≡ (7)
ε κ εα c0α × ln
(εα κ α /εβ κ β )
which is the reverse ratio of the diffuse- + exp[−F (αβ φ − αβ φh )/2RT ]
layer capacitance. There is a direct pro-
RT
portional dependence between φdα and φd
β
= −(αβ φ − αβ φh ) +
F
with coefficient ηα/β when the potential
drops are small: (εβ κ β /εα κ α )
+ exp[F (αβ φ − αβ φh )/2RT ]
φdα × ln
= −ηα/β (8) (εβ κ β /εα κ α )
β
φd + exp[−F (αβ φ −αβ φh )/2RT ]
From Eqs. (7) and (8), it follows that the (11)
potential drop φd in the diffuse layer will The spatial distribution of potential near
be less if the dielectric permittivity ε or the interface is accompanied by changes
electrolyte concentrations c0 are increased in electrolyte concentrations that produce
or, if the diffuse double-layer capacitance surface excesses of ions. In each phase, the
Cd is increased. At the contact between surface excess i of an ion i can be divided
aqueous and organic phases, virtually the into two components in the inner layer ih
entire drop of the potential occurs in the and in the diffuse layer id :
organic phase. However, with increasing
interfacial potential, the situation changes i = ih + id (12)
dramatically. When potentials are large
enough, Eq. (6) can be presented in a The charge of each contacting phases is
different approximation: equal to:

RT q= zi F i (13)
β β
φdα = −φd − sin φd ž ln ηα/β (9) i
F
Here, the coefficient of proportionality is For a binary electrolyte, the ion excess in
equal to one. the diffuse layer can be written as:
From Eqs. (1) and (6), it is possible to find
the dependence of the potential drop in 2c0 zi F φd
id = exp − −1
both diffuse layers αβ φ − αβ φh on poten- κ 2RT
β 
tial drops φdα and φd in each diffuse layer:
2c0 q2
= α  1−
β RT κ 8ε0 εRT c0
φd =
F 
(εβ κ β /εα κ α ) zi q
+ exp[F (αβ φ − αβ φh )/2RT ] + − 1 (14)
8ε0 εRT c0
× ln
(εβ κ β /εα κ α )
+ exp[−F (αβ φ − αβ φh )/2RT ] where c0 is the electrolyte concentration in
(10) the bulk phase.
The interfacial capacitance C consists ra types of anions with chemical poten-

of the capacitance of the compact layer tials µi . Using Hansen’s reference system
Ch and capacitance of two diffuse layers and choosing the solvents α and β as ref-
β
Cdα and Cd . Differentiating Eq. (13) with erence substances, one can write down
respect to charge q α of phase α, and Hansen’s rendition of Gibbs adsorption
assuming that the drop of the potential equation:
in the compact layer αβ φh does not
depend on an electrolyte concentration, s s
dγ = −s(α,β) dT + τ(α,β) dp
we have:

1 1 1 1 − h(α,β) dµh − i(α,β) dµ̄i
= + +
C(q, c) Ch (q) Cdα (q, c) C β (q, c) h =α,β i
d
(15) (17)
Diffuse layer capacitance depends on s
Here s(α,β) s
, τ(α,β) and h(α,β) are the
potential and electrolyte concentration: surface excesses of the entropy, volume,
and substance h in the reference system,
2ε0 εc F φd
Cd = F cosh when the absolute excesses α and β are
RT 2RT zero. The terms corresponding to different

F components of the system are divided into
= 8ε0 εRT c + q 2 (16)
2RT two summations, one for neutral and the
In classical electrochemistry, the capaci- other for charged particles, the reference
tance usually depends on surface charge q substances α and β being excluded from
(Fig. 6). the sum of neutral particles. Individual
components of the system, both charged
2.3.3 and neutral, can be present either in both
The Electrocapillary Equation phases or in only one. The electrocapillary
equation in its final form, which includes
A comprehensive thermodynamic the- the electromotive force of the measuring
ory of the electrocapillary phenomena cell, can be written as:
at polarizable and nonpolarizable liq-
uid interfaces was developed by Markin
1
and Volkov [11, 12] using Hansen’s
s
dγ = − s(α,β) − Qα sR∗ α
zj F
method [13]. Electrochemical processes oc-
curring at the interface between two im- 1
− Qβ sR∗ β dT
miscible liquids are traditionally described zm F
on the basis of Gibbs thermodynamics of
1
surfaces. However, Hansen’s method, by s
+ τ(α,β) − Qα νR∗ α
extending and generalizing Gibbs method zj F

gives us a better understanding of the 1 β ∗
nature of interfacial phenomena and pro- − Q νRβ dp
zm F
vides us with an improved method for
describing them. Consider two phases in 1
− h(α,β) dµh −
contact, α and β with rh neutral com-
k =k
νj− k
h =α,β
ponents, whose chemical potentials are
designated by µh , rc types of cations and × k(α,β) dµj k

1 or 1,2-dichloroethane as phase β. The
− j (α,β) dµj k interface between phases 1 and 2 is the
j =j
νj+k
polarized oil–water interface serving as
a working electrode (interface). The com-
1
− m(α,β) dµl m mon cation A+ is usually the terabutylam-
m =m
νl− m monium ion, and B− is tetraphenylborate
or dicarbollylcobaltate.
1 1 The potential difference measured in the
− + l(α,β) dµlm −
l =l
νlm zk νj− k cell (I) consists of the sum of two interfacial
  potential drops:
Ag Ag
× zk k (α,β) + zj j (α,β)  −E = αβ φ + α φRE1 − α φRE2 + φD
j =j (20)
φD is the Nernst–Donnan potential dif-
1 ference between the nonaqueous fraction
× dµj k −
zlk νl+ m and the aqueous solution of RE2 with the
  common ion A+ :

× zl l (α,β) + zm m(α,β)  φD = −αβ φA0 + +
RT
ln
a2
(21)
m =m F a3
α(j ) and αβ φA0 + is the standard potential
× dµl m − Qα dEβ(m ) (18)
difference for A+ known from literature
Note the symmetry of the last term with data or calculated theoretically.
respect to the transposition of the α and β By suitable selection of the Cl− con-
indexes: centrations and ionic strengths of the
α(j ) β(m )
electrolytes into which Ag/AgCl− elec-
Qα dEβ(m ) = Qβ dEα(j ) (19) trodes are immersed, we obtain:
Equation (18) contains rh + rc + ra − 1 Ag Ag RT a1
α φRE1 − α φRE2 = ln (22)
independent differentials of intensive vari- F a3
ables, whose number is equal to the num-
Substituting (7.18) and (7.19) into (7.17),
ber of the degrees of freedom of the system.
we obtain:
The impedance or electrocapillary prop-
erties of the ITIES can be measured with RT a3
dαβ φ = −dE − dαβ φD + d ln
the cell mentioned in Sch. 1. F a1
Silver–silver chloride electrodes are usu- (23)
ally used as the reference electrodes (RE), With Eq. (23), the electrocapillarity equa-
water is used as phase α, and nitrobenzene tion at the flat oil–water interface can be
1 2 3
RE1 H2O (α ) A+B− (β) A+C− (α ) RE2
M+Cl− oil H2O
a1 a2 a3
∆Agα φ RE1 ∆αβ φ ∆φD −∆Agα φ RE2
Scheme 1
written as follows: Koenig barrier, which rules out a possible

overlap between the groups (Fig. 1b). In
dγ = − i dµi − Qα dE this case, the free charges qα and qβ of the
i phases are equal to the thermodynamic
RT a2 a1 charges: qα = Qα = −qβ = −Qβ .
+ Qα d ln 2 (24)
F a3
2.3.4
where Q is the charge that must be Zero-Charge Potentials
supplied to each side (α or β, corre-
spondingly) of the polarizable interface Potentials of zero charge of the interface
when expanding it by a unit area in can be found reliably by the same
order to maintain the potential differ- independent methods that are used at
ence between the phases [14]. Figures 5 the metal–water interface. These in-
to 7 illustrate dependencies of interfacial clude finding the differential capacitance
tension and capacitance on the poten- minimum of the electric double layer, from
tial difference between two immiscible electrocapillary curves, with a flowing-
electrolyte solutions. electrolyte electrode, with the vibrat-
At the ideally polarizable interface, the ing boundary method, with radiotracers,
ions of each group are separated by the or by measuring the second harmonic
25
20
[mN m−1]
Br−
γ
15
F−
Cl−
10
Fig. 5 Comparison of electrocapillary
curves for the interface between
nitrobenzene solution of 0.1 M
hexadecyltrimethylammonium
tetraphenylborate and aqueous solu-
tion of 0.05 M LiF (◦), LiCl (∇), and 5
−0.2 0 0.2
LiBr (). (From Ref. [15], reproduced W−
by permission of The Chemical Society E O−− Epzc
of Japan.) [V]
Fig. 6 Electrocapillary curves at 25 ◦ C

25 for the interface between nitrobenzene
solution of 0.1 M tetrapentylammonium
tetraphenylborate and aqueous solution
of 0.05 M LiCl in the presence of x mmol
2
1 dm−3 C12E4: x = 0 (curve 1), 1
3 (curve 2), 2 (curve 3), 5 (curve 4), 7
4 (curve 5), 10 (curve 6), 15 (curve 7), 20
20 5
(curve 8), 30 (curve 9), 40 (curve 10), 50
[mN m−1]
6 (curve 11), 70 (curve 12), 80 (curve 13),

γ
7 and 100 (curve 14). (From Ref. [16],

8
reproduced by permission of The
9 Chemical Society of Japan.)
10
15
11
12
13
14
10
0.1 0.2 0.3 0.4
W−
E O+
[V]
generation. Potentials of the zero free three capacitors in series:

charge at nitrobenzene/water and 1,2-
β
dichloroethane/water interfaces, obtained dαβ φ dαβ φh dφdα dφd
= + + (26)
from the differential capacitance mini- dq dq dq dq
mum of the electric double layer in solu-
tions of a surface-inactive electrolyte do not One can use the following approximation
necessarily correspond to the potentials of when modeling the electric double layer:
thermodynamic zero charge (Fig. 7). They β
d2 αβ φ d2 αβ φh d2 φdα d2 φd
can depend on electrolyte concentration = + +
(Fig. 8) when the capacitance of the com- dq 2 dq 2 dq 2 dq 2
pact layer is affected by surface charge as a (27)
result of nonlinear double-layer properties. where

The potential difference between two Cφ Chφ
immiscible electrolyte solutions can be αβ φ = − 3
; αβ φh = − h
;
C Ch3
written as the sum:
α β
β Cdφ Cdφ
αβ φ = αβ φh + φdα + φd (25) φα = − α
; φβ = −
β
(28)
(Cdα )3 β
(Cd )3
where αβ φ and φd are the potential drops
across the compact and diffuse layers, We find the position of the minimum in
respectively. In the linear approximation the C vs. αβ φ curve as:
and in the absence of specific adsorption,
β
the electric double layer is equivalent to αβ φ = αβ φh + φdα + φd = 0 (29)
27 0.4
26
0.3
[mN m−1]
[F m−2]
C
25
γ
0.2
24
23 0.1
0.15 0.20 0.25 0.30 0.35 0.40 0.45
−E
[V]
Fig. 7 Interfacial tension and differential capacity as functions of applied
potential difference. The medium: 0.1 M LiCl in water, 0.1 M
tetrabutylammonium tetraphenylborate in the nitrobenzene at 25 ◦ C.
(Experimental points are taken from Refs. [17, 18].)
It follows from the Gouy–Chapman concentration (Fig. 8). Therefore, a cor-

diffuse-layer theory that if qα = qβ = rect determination of the zero free-charge
0 then, (φdα ) = (φd ) = 0. Therefore,
β
potential for the nitrobenzene-water sys-
β φ = 0 applies only when αβ φi = 0,
α tem in the presence of a binary surface-
that is, when the compact layer capacitance inactive electrolyte is possible only at
is independent of surface charge. It has base-electrolyte concentrations less than
been shown that Cc generally depends on 0.01 M. For the water-nitrobenzene sys-
potential and charge, and αβ φi = 0. As tem, this quantity w φ
nb pzfc
= −0.29 V and
the value of αβ φi in the region of the for the water-1,2-dichloroethane system
potential of zero free charge increases, w φ
de pzfc
= −0.27 V.
the value of q corresponding to the For many systems, the maximum of the
minimum in the C vs. αβ φ curve will electrocapillary curve is located in the re-
also increase [14]. gion of ideal polarizability of the electrode,
For the water-nitrobenzene system, one and the maximum potential corresponds
observes a strong dependence of ca- to the potential of zero total (thermo-
pacitance of the compact double layer dynamic) and zero free charge. At the
on surface charge, even in the absence mercury–water interface, the region of
of specific adsorption. This leads to a ideal polarization is a few volts when sol-
dependence of the potential minimum uble mercury salts are absent (2.2 V), but
in the capacitance curve on electrolyte at the interface between two immiscible
50
40
30
[mV]
E min
20
10
0
0.000 0.025 0.050 0.075 0.100
C
[M]
Fig. 8 Displacement of the differential-capacity minimum on
the potential axis as a function of electrolyte concentration in
the water-nitrobenzene system. The medium: 0.1 M LiCl in
water and 0.1 M tetrabutylammonium tetraphenylborate in
nitrobenzene. (Constructed from data of Samec and
coworkers [19].)
liquids, the region is about a tenth of methods of capacitance measurements

that value. Under conditions in which the as impedance, galvanostatic, and poten-
current flow does not significantly alter tiodynamic techniques. More recently, it
the compositions of the liquid phases and has become possible to measure elec-
in which the equilibrium at the interface tric double-layer capacitance at the ITIES
is not disturbed by the current, the elec- using a four-electrode potentiostat or two-
trocapillary curves yield the potential of electrode potentiostat. Commonly stud-
zero total charge. However, under realis- ied systems are nitrobenzene/water and
tic conditions in which the region of ideal 1,2-dichloroethane/water, in which the
polarization is narrow, the maximum of organic phase has a relatively high dielec-
the electrocapillary curve corresponds to tric permittivity and a high dissociation
the potential of zero total charge, rather constant.
than zero free charge. This closely resem- In typical investigations, cyclic current-
bles systems consisting of amalgams and potential curves are usually determined
liquid electrolyte (LE) solutions that are before impedance measurements in order
described by Frumkin [20]. to find the potential range in which the con-
The structure of the electric double tribution of faradaic impedance is low. The
layer at ITIES has been investigated electric double-layer capacitance is mea-
using such common electrochemical sured in the potential window between
extremes in the cyclic voltamogram. As a result, the equation for the

Over this range of potentials, the wa- capacitance of the electric double layer is:
ter–dichloroethane interface has proper-
ties close to those of an ideally polarizable qw (κw )2 εw εorg
electrode. C= = −
worg φ worg φ
A ‘‘diffuse’’ picture of the compact xw
layer has been considered [19]. According 2
× (φ − worg φ) dx = κw εw εorg
to this hypothesis, adsorbed ions can −∞
penetrate into the compact layer, which  
consists of solvent dipoles, analogous to εh κh
 1+ exp(κh δ) 
nonlocalized electron gas penetration from  εorg κorg 
 
a metal electrode to an aqueous electrolyte  
 + 1−
εh κh 
 exp(−κh δ) 
solution. Here, a penetrating ion can act as  εorg κorg 
 
a hydrophobic anion in the organic phase × 

and potential distribution φ(x) near to  εw κw εh κh 
 1+ 1+ 
the point of zero charge can be calculated  εh κh εorg κorg 
 
using the Poisson–Boltzmann equation:  
 × exp(κ δ) + 1 − εh κh 
 h 
If x < x2w , ε = εw :  εorg κorg 
 
 εw κw 
φ = (κw )2 (φ − w
org φ) (30) × 1− exp(−κh δ)
εh κh
org (35)
If x2w < x < x2 , ε = εh : org
where δ = x2 − x2 . It should be noted
w
φ = (κh )2 φ (31) that the size of a hydrophobic penetrating
org anion is large in relation to the compact
If x > x2 , ε = εorg :
part of the EDL and it is not clear if the
φ = (κorg )2 (φ − w
org φ) (32) Poisson–Boltzmann equation can be used
in this case.
In each of these three areas, the solution
can be written as 2.3.5
Specific Adsorption at Liquid–Liquid
φ(x) = A+ eκx + A− e−κx (33) Interfaces
where six coefficients A+ and A− can Ions can be adsorbed specifically if the
be determined from the six boundary main contribution to their interaction with
conditions: the interface (ions, dipoles) is caused by
φ(−∞) = w non-coulombic short-range forces. Specific
org φ
adsorption cannot be explained using
φ(∞) = 0 only the theory of diffuse double layer.
Specifically adsorbed ions penetrate to the
φ(x2w − 0) = φ(x2w + 0)
(34) compact layer and form a compact or loose
org org
φ(x2 − 0) = φ(x2 + 0) monolayer. The surface passing through
the centers of specifically adsorbed ions is
εw φ (x2w − 0) = εh φ (x2w + 0)
usually called the inner Helmholtz plane. If
εh φ (x2
org org
− 0) = εorg φ(x2 + 0) several kinds of specifically adsorbed ions
are present, each ion can have its own and the potential dependence of the ca-
inner Helmholtz plane. pacitance is strongly influenced by the
It was found while studying the mech- adsorption isotherms owing to the interfa-
anism of electron transfer (ET) across the cial ionic association.
interface between two immiscible liquids When there is specific adsorption of
that specific anion adsorption occurs at ions dissolved in phase α the condition
the octane–water interface in the presence of electroneutrality can be written as
of metalloporphyrins. This adsorption in-
creases in the order of Cl− , Br− , I− , and q β = −q α = −(qiα + qdα ) (36)
is caused by coordinative bonding of the
anions as ligands of the porphyrin metal where qi is the charge of the in-
atoms. ner Helmholtz plane. The separation of
Specific ion adsorption at the po- qα from qiα and qdα cannot be done
larizable nitrobenzene–water interface without introducing a model of the
containing monolayers of phosphatidyl- interface.
choline, phosphatidylserine, octaethylene Although capacitance of the interface
glycol monodecyl ethers, tetraethylene can be calculated as before using the
glycol monodecyl ether, and hexade- equation
cyltrimethylammonium (HTMA+ ) was β
dαβ φ dαβ φi dφdα dφd
studied in detail. HTMA+ exhibited no = + + (37)
specific adsorption in the potential range dq dq dq dq
in which the aqueous phase was pos- the second term in the right side of
itive [21], whereas a strong adsorption the equation is not the diffuse-layer
occurred in the potential range in which capacitance since the charge of a diffuse
the electric potential in the aqueous phase layer in phase a is equal to
was negative with respect to the nitroben-
α,j
zene (Fig. 9). qdα = −qβ − qi (38)
Another example of specific ion ad- j
sorption was discussed in terms of the
formation of interfacial ion pairs between The diffuse-layer capacitance is equal to
ions in the aqueous and the organic
dqα
phase. The contribution of specific ionic Cd = − (39)
adsorption to the interfacial capacitance dφdα
can be calculated using the Bjerrum the-
and one can write
ory of ion-pair formation. The results  
α,j
show that a phase boundary between two d qi
immiscible electrolyte solutions can be de-  
 j 
scribed as a mixed solvent region with C −1 = (C i )−1 + (Cdα )−1 1 + 
 dq β 
varying penetration of ion pairs into it,
depending on their ionic size. The capac-
+ (Cd )−1
β
itance increases with increasing ionic size (40)
in the order Li+ < Na+ < K+ < Rb+ <
Cs+ . Yufei et al. [22] found that significant The drop of a potential in the compact layer
specific ion adsorption occurs at the inter- depends on the surface charge qα and the
α,j
face between two immiscible electrolytes charge of the inner Helmholtz plane qi .
−
+
−
+ −
+ +
+ − +
− +
−
+ −
+ −
−
+
−
+
−
− +
−
NB W
+
φO
φW
OHP OHP
(NB) (W)
Fig. 9 Schematic representation of the double-layer structure of the interface between
nitrobenzene and aqueous solutions in the presence of the specific adsorption of
hexadecyltrimethylammonium ions [21]. (Reproduced by permission of the Chemical
Society of Japan.)
It shows that or
dφi 1
≡
∂φi dq β Ci
dφi = dq β

∂q β α,j
qi ∂φi α,j
∂φi dqi
= +
∂φi α,j
∂q β q α,j ∂q
α,j
α,k dq
β
i q β ,qi
+ α,j
dqi (41) i
∂qi q β ,qiα,k (42)

α,j α,j consider this monolayer as a 2-d system.

Here the index qi means that all qi are
constant except one. The solvent molecules do not form a mono-
In the absence of specific adsorption layer, but rather a multilayer. Therefore,
the capacitances of the interface, compact the transition from 3-d- to 2-d-geometry
and diffuse layers are always positive. The should be specified. Consider molecules
capacitance of the compact layer in the of both solvents that are substituted by
presence of specific adsorption can be a surfactant (Fig. 10). Suppose that these
either positive or negative. molecules can be assembled into columns
consisting of mo molecules of oil and mw
2.3.6 molecules of water. Suppose that one col-
Adsorption Isotherm umn of oil molecules matches the nw
molecules of water. This match of 1 oil
Traditional models for calculation of ad- column and nw water columns will be con-
sorption isotherms are based on the as- sidered in what follows as a quasi-molecule
sumption that surface-active compounds of solvent Q. These quasi-molecules con-
at the interface can substitute for adsorbed stitute a ‘‘monolayer’’ of solvent. They
molecules of one solvent but cannot pen- consist of mo oil molecules and nw mw
etrate the second phase. Although these water molecules.
models are useful for metal–water inter- Designate the molecules of surfactant in
faces, recent interest has focused on the the bulk as A, and in the monolayer as B.
surface chemistry of amphiphilic com- At the interface, aggregation of surfactant
pounds that can penetrate both phases molecules can take place, B ⇔ rA, such as
and replace adsorbed molecules of both dimerization of porphyrin or pheophytin
solvents, for example, water and oil. Am- molecules at the octane–water interface.
phiphilic molecules consist of two moieties Let the surfactant B replace p quasi-
with opposing properties: a hydrophilic po- molecules at the interface. Therefore, one
lar head and a hydrophobic hydrocarbon can write
tail. We present here, a theoretical analy-
sis of the generalized Frumkin adsorption pQ + rA = B + p(oil) + pnw (water)
isotherm for amphiphilic compounds. (43)
The interface between two immiscible The chemical potentials for (7.76) are:
liquids may be considered to be a surface
solution of surfactant in a special kind pµsQ + rµbA = µsB + pµbo + pnw µbw (44)
of solvent. In order to calculate the
Taking the 2-d solution as ideal, we have:
entropy of such a solution, we will
adopt a simplified lattice model and use µsQ = µ0,s s
Q + RT ln XQ (45)
lattice statistics, a widely used method
for describing surface solutions. The µsB = µ0,s s
B + RT ln XB (46)
transition from three-dimensional (3-d)
to two-dimensional (2-d) geometry may In the bulk phase, we have:
cause errors in statistical formulas, if
some peculiarities of 2-d solutions are µbA = µ0,b b
A + RT ln XA (47)
overlooked.
µbo = µ0,b b
o + RT ln Xo (48)
The main difficulty when dealing with
a monolayer of a surfactant, one can µbw = µ0,b b
w + RT ln Xw (49)
Oil A
A
pQ Q
Oil Oil Oil

B
H2O H2O H2O H2O H2O H2O
nw
Water
Fig. 10Structure of oil–water interface with adsorbed monolayer of
amphiphilic surfactant B.
In all these equations, X designates the for this state with real particles A, O, W
mole ratio of corresponding substances. becomes
Substituting these equations into (44), one
NAs
obtains: XsA = (54)
NAs + NOs + Nws
pµ0,s 0,b 0,s 0,b 0,b
Q + rµA − µB − pµo − pnw µw NOs
XsQ = (55)
(XbA )r Xsb NAs + NOs + Nws
+ RT ln = RT ln
(Xbo )p (Xbw )pnw (XsQ )p Nws
Xsw = (56)
(50) NAs + NOs + Nws
Using the standard Gibbs free energy of
adsorption and, we can obtain:
XsA Xs
sb G0 = µ0,s 0,b 0.b
B − rµA + pµo XsB = ; XsQ = s o s (57)
XsA
+ Xos XA + Xo
0,s
+ pnw µ0,b
w − pµQ (51)
The adsorption isotherm can then be
one obtains the adsorption isotherm: presented in the form

XsB (XbA )r sb G0 XsA
= b p b pn exp − (Xs + Xso )p−1
(XsQ )p (Xo ) (Xw ) w RT (XsQ )p A

(52) (XbA )r sb G0
We considered the 2-d solution of surfac- = exp − (58)
(Xbo )p (Xbw )pnw RT
tant B in the solvent of quasi-particles Q,
in which the mole ratios were defined as
In the past, the adsorption isotherm was
Ns NQs presented in terms of the fraction θ of the
XsB = s B s ; XsQ = s (53) surface actually covered by the adsorbed
NB + NQ NB + NQs
surfactant.
Some authors prefer another set of If we introduce η as the ratio of areas
definitions when real particles in the occupied in the interface by the molecules
interface are considered. The equation of surfactant and oil, the mole fractions
in surface solution can be presented as the number of columns of oil, which

follows: could be supplanted with one molecule
η(1 − ) of surfactant. Therefore, p is a relative size
XsB = ; Xso = of the surfactant molecule in the interfacial
+ η(1 − ) + η(1 − )
layer. It is reasonable to suppose that:
(59)
The adsorption isotherm takes the form η=p (64)

[ + η(1 − )]p−1 If the concentration of surfactant in the
η (1 − )p
p
solution is not high and the mutual
(XbA )r sb G0 solubility of oil and water is low, then we
= b p b pn exp − (60) can use the approximation Xbo = Xbw = 1,
(Xo ) (Xw ) w RT
so that the general Eq. (63) simplifies to:
In this isotherm, the mole fractions XbA , Xbo ,
[p − (p − 1)]p−1
Xbw of the components in the bulk solution exp(−2a)
are presented. In the general case, they p p (1 − )p

must be substituted with activities: b r
sb G0
= (XA ) exp − (65)
RT
[ + η(1 − )]p−1
ηp (1 − )p
This is the final expression for the
(aAb )r sb G0 isotherm that we will call the amphiphilic
= b p b pn exp − (61)
(ao ) (aw ) w RT isotherm [23]. It is straightforward to
derive classical adsorption isotherms from
If the molecules B can interact as pairs the amphiphilic isotherm (65).
in the adsorbed layer and the energy
of each new particle is proportional to 1. The Henry isotherm, when a = 0, r =
its concentration, then their chemical 1, p = 1, 1:
potential, µsB , instead of Eq. (46), should
be presented as: b
sb G0
= Xa exp − (66)
RT
µsB = µs,0
B + RT ln X − 2aRT X (62)
2. The Freundlich isotherm, when a = 0,
where a is so-called attraction constant.
p = 1, 1:
Then, after some algebra we obtain,

instead of Eq. (60), the isotherm: sb G0
b r
= (Xa ) exp − (67)
[η − (η − 1)]p−1 RT
exp(−2a)
ηp (1 − )p
3. The Langmuir isotherm, when a = 0,
(XbA )r sb G0 r = 1, p = 1:
= b p b pn exp − (63)
(Xo ) (Xw ) w RT
b
sb G0
= Xa exp − (68)
Recall that η was introduced as the 1− RT
ratio of areas occupied in the interface
by the molecule of surfactant to the 4. The Frumkin isotherm, when r = 1,
same of oil, and p was introduced as p = 1:

s G0 or chlorophyll molecules. These are sup-
exp(−2a) = Xba exp − b ported by molecular dynamic studies in the
1− RT
systems decane/water, nonane/water, and
(69) hexane/water. The structure of both wa-
ter and octane at the interface is different
Therefore, the amphiphilic isotherm
from the bulk. Adsorbed at the interface,
(65) could be considered as a generaliza-
octane molecules have a lateral orientation
tion of the Frumkin isotherm, taking into
at the interface, as it shown in Fig. 11.
account the replacement of some solvent
molecules with larger molecules of surfac-
2.3.7
tant. Of course, the amphiphilic isotherm
Image Forces
includes all the features of the Frumkin
isotherm and displays some additional
Any charge near phase boundary interacts
ones. To elucidate them, it will be conve-
with it because of the different polarization
nient to change the variable XbA to the rel-
of two phases. The force of this interaction
ative concentration y = XbA /XbA ( = 0.5),
can be found by solution of the Laplace
where XbA (0.5) is the concentration corre-
equation, but it is more convenient to use
sponding to the surface coverage θ = 0.5:
the method of images. This method em-
[p − (p − 1)]p−1 ploys a fictitious charge (‘‘image’’), which
y= exp(a − 2a) together with the given charge creates the
(p + 1)p−1 (1 − )p
(70) right distribution of electric potential in
This equation gives the coverage fraction the phase [24]. All ions interact with all
θ as a function of relative concentration images giving rise to the so-called ‘‘image
y, while a and p are the parameters of forces.’’ The image forces are largely re-
this isotherm, the first being the attraction sponsible for both positive and negative
constant and the second, the size of surfac- adsorption. Important contribution to this
tant. These parameters play an important effect is given by the change of the size
role because their effect on the shape of of solvation shells of ions in the boundary
amphiphilic isotherm is very strong. layer. This has an impact on the planes
Amphiphilic isotherm (65) analysis can of closest approach of ions and dipolar
be used for the determination of the inter- molecules to the phase boundary.
facial structure. An amphiphilic molecule, The electrostatic Gibbs free energy for an
which consists of two moieties with oppos- ion in the vicinity of a boundary between
ing properties such as a hydrophilic polar two liquids with dielectric constants ε1 and
head and a hydrophobic hydrocarbon tail, ε2 (Fig. 12) is determined by the Born ion
should be used as an analytical tool located solvation energy and by the interaction
at the interface. Pheophytin a is a well- with its image charge. In research dealing
known surfactant molecule that contain with the energy of image forces and
a hydrophobic chain (phytol) and a hy- with the interactions of charges at the
drophilic head group. The value of p, less oil–water interface, approximate models
than 1.0, indicates that adsorbed molecules of the interface are often employed, which
of n-octane are parallel to the interface be- are based on the traditional description of
tween octane and water [23] (Fig. 11). Sub- the interface between two local dielectrics.
stitution of one adsorbed octane molecule In the oil–water system, the force of
requires about 4 to 5 adsorbed pheophytin attraction (or repulsion) of charge in the
CH2
CH H CH3
H3C CH2CH3
NH N
H H
H3C N HN
CH3
H
HH
CH3OOC
Water O
O
O
Octane Octane Octane Octane Octane
P
h
y
Octane t
phase o
l
Fig. 11 The amphiphilic compound pheophytin a, which can

substitute one-sixth of the lateral oriented adsorbed molecule of
octane, is a tool for the evaluation of the interfacial structure.
organic (β) phase with its image in the uniform charge distribution and can con-
aqueous phase (α) sitting on the same tinuously pass between the two phases. For
axis perpendicular to the dividing plane is a point charge at long distances from the
given by: interface, electrostatic Gibbs free energy
can be written as
εα − εβ (ze0 )2
F (h) = − (71) (ze0 )2 εα − εβ 2a
εα + εβ 16πε0 εβ h2 G = 4+
32πε0 εα a εα + εβ h
where h is the distance from the interface. (72)
If εα > εβ , the charge in the nonpolar where a is an ionic radius, h is the distance
phase is attracted to its image, but if from the interface and ze is the charge of
εα < εβ there is repulsion between the an ion.
charge in an organic phase and its image. The potential φ for the spherically
From Eq. (71) it follows that charges in symmetrical charge distribution is
the organic phase are attracted to the
 ze0
oil–water interface. Image forces attract 


 4πε 0 εα r1
the diffuse layer on the organic side, 

making it thinner, and repel and thicken  ze0 εα − εβ
φ= + (in region α)
the aqueous diffuse layer.  4πε0 εα r1 εα + εβ


The Kharkats–Ulstrup model [25] incor- 
 ze0 2εβ

 (in region β)
porates the finite radius of an ion a, which 4πε0 εβ r1 εα + εβ
is assumed to have a fixed spherically (73)
Fig. 12 The relative arrangement of a

spherical ion and a planar interface
between two immiscible liquids with
dielectric permittivities εα and εβ : β α β α
(a) the ion-center distance from the
interfacial plane P exceeds the ion r1' 1r '
radius; (b) the ion partially penetrates r1
the interface, h < a. r1
a a
h h h h
εβ εα
εβ εα
P P
(a) (b)

Using standard procedures of calculations, (1 + h/a)(1 − 2 h/a)
×
Kharkats and Ulstrup obtained equations 1 + 2 h/a
for the electrostatic part of the free Gibbs
a 2h
energy of a finite-size ion. + ln 1 +
2h a
When h > a
2 2
(ze0 ) 2εβ h
(ze0 )2 εα − εβ 2a + 1−
G = 4+ 16πε0 εβ a εα + εβ a
32πε0 εα a εα + εβ h (75)
2 If h = 0 and center of an ion is at the
εα − εβ
+ interface, electrostatic Gibbs energy will
εα + εβ
be equal to
2 a 2h + a
× + ln (ze0 )2
1 − (2h/a)2 2h 2h − a G(h = 0) = (76)
(74) 4πε0 (εα + εβ )a
The first term in (74) is the Born solvation
The Figure 13 illustrates the effect of
energy, and the second is the interaction
image forces on the electrostatic Gibbs
with the image. For a charge located in
energy of an ion at the oil–water interface.
region β, the electrostatic free energy is ob-
Image forces play a significant role in
tained from (74) by exchanging εα ⇔ εβ .
electric double-layer effects. The excess
Kharkats and Ulstrup [25] obtained the
surface charge density is
expression for the electrostatic free energy
in region 0 ≤ h ≤ a: hc
g(h)
q = ec0 exp − − 1 dh
ha kB T
(ze0 )2 2h (77)
G = 2+
32πε0 εα a a where

εα −εβ 2h εα −εβ 2 (ze0 )2
+ 4− + g(h) = G(h) − (78)
εα +εβ a εα +εβ 8πε0 εβ a
Fig. 13 Electrostatic Gibbs energy

36 ε 2 = 7.5 profiles for ion transfer across the ITIES
boundary. Solid lines: finite-size ion
profiles in units of (ze)2 /ε1 a, ε1 = 78
ε 2 = 10
and different values of ε2 . Dashed lines:
profile for the point change model in the
28 same units, ε1 = 78, ε2 = 10 [25].
(Reproduced by permission of Elsevier
w/(z 2e 2/8ε 1a)
Sequoia S. A.)
20
ε 2 = 20
12
ε2 = 50 ε1 = 78
−6 −2 0 2 6
h /a
G(h) is the electrostatic contribution to first approach uses the Gouy–Chapman

the Gibbs energy of solvation, and ha and theory and the average electrical field in the
hc are the distances of closest approach diffuse layer is calculated from the Pois-
of the anion and the cation. If image son equation. The structure and physical
forces are taken into account, the diffuse properties of the double layer were then
layer capacitance can be calculated by the calculated from the restricted primitive
equation: electrolyte model. As a result, the modified
Poisson–Boltzmann theory (MPB) was de-
g(h0 ) veloped.
Cd = CdGC exp (79)
2kB T The second statistical approach incor-
porates correlation functions and integral
where CdGC is the Gouy–Chapman diffuse equations from the theory of liquids. In
layer capacitance without image terms. this case, the well-known uniform Orn-
As noted by Kharkats and Ulstrup [25], stein–Zernike equations were modified
Eq. (79) always gives diffuse layer capaci- to calculate the ion distribution function
tance corrections toward higher values. near a charged interface. Modified in this
The first step to statistically correct the way, the Ornstein–Zernike equations be-
Gouy–Chapman theory for the diffuse came nonuniform and were solved using
double layer used a restricted primitive hyperneted chain approximations (HCA)
electrolyte model. This model considers incorporating a mean spherical approxi-
ions to be charged hard balls of identical mation (MSA) for correlation functions
radii in a structureless dielectric contin- (HCA/MSA). A third statistical approach
uum with constant dielectric permittivity. used to describe the EDL was achieved
There are three main approaches to creat- by using the integral Born–Green–Ivon
ing a statistical theory with this model. The equations.
These correlation function approaches function that couples polarization vectors

resulted in the theory of ionic plasma in throughout the solvent.
semi-infinite space. The next advance in
describing the double layer was achieved 2.3.8
by upgrading the restricted primitive elec- Modified Poisson–Boltzmann (MPB) Model
trolytes model to a ‘‘civilized’’ or ‘‘non-
primitive’’ model. This was accomplished The popular Poisson–Boltzmann equa-
by addition of hard spheres with embedded tion considers the mean electrostatic
point dipoles to the restricted ‘‘primitive’’ potential in a continuous dielectric with
model. This theory takes into account the point charges and is therefore, an ap-
presence of long-range Coulomb interac- proximation of the actual potential. An
tions, image effects, and external electrical improved model and mathematical solu-
fields. The new ‘‘nonprimitive’’ model tion resulted in the MPB equation [26].
became the basis for the theory of ion- This equation is based on a restricted prim-
dipole plasma. The serious mathematical itive electrolyte model that considers ions
difficulties encountered in describing the as charged hard spheres with diameter
‘‘nonprimitive’’ models necessitated the d in a continuous uniform structureless
development of less statistically rigorous dielectric medium of constant dielectric
intermediate models. In these intermedi- permittivity ε. The sphere representing an
ate models, the first monolayer of solvent ion has the same permittivity ε. The model
molecules at the charged interface was con- initially was developed for an electrolyte at
sidered as a complex of discrete particles a hard wall with dielectric permittivity εw
and an electrolyte outside this layer was and surface charge density σ . The charge
described by the Gouy–Chapman theory. is distributed over the surface evenly and
This intermediate approach was able to continuously.
provide good theoretical agreement with This theory takes into account the finite
experimental results. size of ions, the fluctuation potential, and
The concept of constant dielectric per- image forces in the electrolyte solution
mittivity in the double layer is also of next to a rigid electrode, but it still an
concern. Different factors such as high approximation. The MPB theory begins
ionic concentration or strong electric fields with the Poisson equation for the mean
can affect the dielectric permittivity. In ad- electrostatic potential ψ in solution:
dition, description of the dielectric proper-
ties of solutions by a single parameter – the d2 ψ 1
2
=− zi e0 n0i g0i (x) (80)
dielectric permittivity – also came under dx ε0 ε
i
criticism and led to the development of
nonlocal electrostatics. where ε0 is the electrical constant, zi
Nonlocal electrostatics takes into con- is the charge number of ion i, e0 is
sideration discrete properties of solvent the elementary charge, n0i is the bulk
molecules. It assumes that fluctuations number density, and g0i (x) is the wall-
of solvent polarization are correlated in ion distribution function representing the
space. This means that the average polar- ratio of the local ion density to the ion
ization at each point is correlated with the density of the bulk electrolyte solution. The
electric displacement at all other points distance x is measured from the electrode,
and therefore, uses the solvent dielectric so that the distance of closest approach
of an ion i is d/2. According to Kirkwood, In Eq. (84), B0 is a constant, and r ∗ is

the distribution function is given by the the distance from the mirror image in
following equation: the interface. Hence Eq. (84) takes into
account both the ionic atmosphere and
1
g0i (x) = ςi (x) exp − zi e0 ψ(x) + ηi image forces.
kT If r < d and x > 0, the equation for the
(81) fluctuation potential should be written as
The factor ζi (x) accounts for the ex-
cluded volume of the ion, ηi is the d2 ψ
∇ 2 φi = − , r < d, x > 0 (87)
fluctuation potential that takes into con- dx 2
sideration the ionic atmosphere around an The term ζ (x) can be presented as a power
ion created by other ions. The fluctuation series expansion in the bulk density
potential can then be presented as
zi e0 4πd 3 0
1 ζi (x) = 1 + B(x) + nj (88)
ηi = − limr→0 (φi∗ − φib
∗
)d(zi e0 ) 3
kT 0 j
(82)
where where
zi e0  x+d
φi∗ = φi − (83) 
4πεε0 r 
 [(x − x)2 − d 2 ]



 d/2
φi is the fluctuation potential created by the 


 ×g0j (x ) dx ,
ion i and its atmosphere, r is the distance 

  d 3d
from the center of the ion. The difference ≤x≤
B(x) = π n0j
φi∗ is the potential created by the ionic 
 x+d
2 2
∗ is the value of this j 

atmosphere only and φib 
 [(x − x)2 − d 2 ]


potential in the bulk electrolyte solution. 



x−d
There are several methods for calculating 

 ×g0j (x ) dx , x ≥ d
φi , and if the image effect is taken into 2
account, it can be presented as: (89)
Now, we can combine these equations and
1 f
φi = B0 exp(−κr) + ∗ exp(−κr ∗ ) , obtain the modified Poisson–Boltzmann
r r
equation for the diffuse layer:
if r > d, x > 0 (84)
d2 ψ 1 0 zi e0
where f is the image factor determined = − z e n
i 0 i iζ (x) exp −
dx 2 ε0 ε kT
by the difference between two dielectric i

constants of the electrolyte solution and zi e0
the electrode: × L1 (ψ) + (F − F0 ) ,
8πε0 εd
ε − εw d
f = (85) x> (90)
ε + εw 2
The local Debye–Huckel parameter at a Since the distance of the maximal ap-
given point is proach of ions to the interface cannot be
less then d/2, the usual linear dependence
e02 2 0 of potential in the compact layer can be
κ2 = zi ni g0i (x) (86)
ε0 εkT introduced:
i
dψ(0) e02 2 0
ψ(x) = ψ(0) + x (91) κ02 = lim κ 2 = zi ni (99)
dx x→∞ ε0 εkT
i
Additional functions in Eq. (173) are
The boundary conditions for these equa-
4πd 3 0 tions are
ζi (x) = 1 + i ni
3 1. ψ(x), ψ (x) → 0, if x → ∞
x+d 2. ψ (0+) = −δ/ε0 ε (100)
+π n0i
d 3. ψ(x), ψ (x) are continuous at x > 0
i max ,x−d
2
This set of equations completes the
× [(x − x)2 − a 2 ]g0i (x ) dx (92) mathematical formulation of the problem,
but the equations are very cumbersome
F
L1 (ψ) = [ψ(x + d) + ψ(x − d)] and can only be solved numerically. The
2 major difficulty is related to the finite ion

F − 1 x+d size. If we consider the limiting case of
− ψ(x ) dx (93) point charges, the equation for potential in
2d x−d
 2(1 + κx) exp(−κd)  the double layer would be:

 



 +2(1 − κd) exp(−κx) 

 d2 ψ 1 0

 
 =− ni zi e0

 a 
 dx 2 ε0 ε

 
 i

 
+2κ dx r exp(−κr) dr
2 −1

zi e0 zi e0
x × exp − ψ(x) +

 
 kT 8πε0 ε

 f 


 + [exp(−κd) 


 
 f


κx 
 × κ0 − κ + exp(−2κx)

 +2κ(d − x) exp(−κx) 


 

2x
− exp(−κ(2x + d))] (101)
F= ,

 2[exp(−κd)(1 + κ(x + d)) 
 In comparison with the Gouy–Chapman

 


 + exp(−κx)] 
 theory, the MPB equation for point
charges contains two improvements in
 f

− [2κx exp(−κx)+(1+κd) 
 that the Debye–Huckel parameter κ de-
 κx
 

  pends on distance, and the ion image is
×(exp(−2κx) − 1) exp(−κd)] screened.
d In solving the MPB equations, differ-
≤x≤d (94)
2 ent estimates of ζi (x) were considered.
Depending on the type of ζi (x), the equa-
(1 + f δ2 )
F= , if x ≤ d (95) tions were called MPB1, MPB2, MPB3 and
(1 + κd)(1 − f δ1 ) so on.
1 The MPB theory overcomes certain lim-
F0 = lim F = (96)
x→∞ 1 + κ0 d itations of the Gouy–Chapman model by
taking into account ionic volume, the ionic
1
δ2 = exp[κ(d − 2x)] sinh κd (97) atmosphere, and image forces. Results
2κx of both theories coincide at low elec-
(κd cosh κd − sinh κd)δ2 trolyte concentrations and small surface
δ1 = (98) charges, but beyond this limiting case,
(1 + κd) sinh κd
there is a very significant qualitative and that MPB4 describes the experimental
quantitative discrepancy. For instance, the results at the nitrobenzene/water and
MPB theory predicts a thinner diffuse dichloroethane/water phase boundaries. It
ion layer and higher capacitance than the reproduces the Monte Carlo (MC) cal-
Gouy–Chapman model. Near a charged culation more accurately than the MVN
interface, MPB predicts considerable ad- theory for 1 : 1 electrolytes over a wide
sorption of co-ions. The major difference range of conditions including variations
is that the decaying potential distribution in the image forces, ion size, the inner-
of the Gouy–Chapman model transforms layer potential distribution, electrolyte con-
into the decaying-oscillating solution of the centration, surface charge density and
MPB theory, with oscillations beginning solvent effects. They used the equa-
at high electrolyte concentrations when tion for electrostatic potential from the
κ0 d > 1.24. This oscillation was also pre- MPB4 model for the inner layer at ITIES
dicted by Stilinger and Kirkwood [27] who when x < a/2:
estimated a different value κ0 d > 1.03.
q aq
Other modern theories predict oscillatory ψ(x) = ψ(0) + xψ (0) − xλc|q aq |1/2
behavior of the electrostatic potential, but εh
under different conditions. In the HCA (102)
theory, the condition is κ0 d > 1.23, which where ψ(x) is the electrostatic potential at
is very close to MPB theory. The Bo- coordinate x, ψ(0) is the electrostatic po-
golyubov–Born–Green–Ivon theory puts tential at the surface, ψ (0) is the derivative
the condition at κ0 d > 1.72. Therefore, os- of ψ(0), c is the electrolyte concentra-
cillatory behavior is a consistent finding of tion, qaq is the surface charge density
all theories. It is worthwhile to mention of the aqueous phase, λ is the parame-
that damped oscillation in the MPB theory ter determined by fitting the experimental
is due to a fluctuating potential and its values of the Galvani potential differences
existence does not depend on the excluded and εh is the dielectric permittivity of
volume. the compact layer. The third term in
One of the major drawbacks of MPB Eq. (102) represents the dependence on
theory is that it does not take into ac- electrolyte concentration, surface charge
count the discreteness of a solvent. In the density, and the solvent effect on the inner-
diffuse layer, the discreteness of solvent layer potential distribution. These effects
molecules strongly influences ion–ion can be ascribed to ionic penetration into
interactions at small distances. Further- the opposite phase and ion–ion correla-
more, an understanding of the structure tions across the interface. Cui et al. [29]
and properties of the compact layer is obtained good agreement between theo-
impossible without an adequate model retical calculations and experimental data
of solvent molecules and their interac- and came to the conclusion that the
tions with the electrode and between structure of the 1,2-dichloroethane–water
themselves. interface is similar to that of the wa-
Application of MPB model to ITIES ter/nitrobenzene interface, except that the
was made first by Torrie and Val- effects of the image force and ion size
leau [28]. Cui and coworkers [29] applied are more pronounced and the inner-layer
the modified Poisson–Boltzmann the- potential drop plays a more important
ory (MPB4) to the ITIES and found role.
References 16. T. Kakiuchi, T. Usui, M. Senda, Bull. Chem.

Soc. Jpn. 1990, 63, 2044–2050.
1. A. G. Volkov, D. W. Deamer, D. I. Tanelian 17. T. Kakiuchi, M. Senda, Bull. Chem. Soc. Jpn.
et al., Liquid Interfaces in Chemistry and 1983, 56, 1322–1326.
Biology, John Wiley & Sons, New York, 1998. 18. T. Osakai, T. Kakutani, M. Senda, Bull.
2. A. G. Volkov, D. W. Deamer, D. I. Tanelian Chem. Soc. Jpn. 1984, 57, 370–376.
et al., Prog. Surf. Sci. 1996, 53, 1–136. 19. Z. Samec, V. Marecek, D. Homolka, J. Elec-
3. O. S. Ksenzhek, A. G. Volkov, Plant Energet- troanal. Chem. 1985, 187, 31–51.
ics, Academic Press, San Diego, 1998. 20. A. N. Frumkin, Zero Charge Potentials, Nauka
4. A. G. Volkov, D. W. Deamer, (Eds.), Liq- Publ., Moscow, 1979.
uid–Liquid Interfaces: Theory and Methods, 21. T. Kakiuchi, M. Kobayashi, M. Senda, Bull.
CRC Press, Boca Raton, London, Tokyo, Chem. Soc. Jpn. 1987, 60, 3109–3115.
1996, p. 421. 22. C. Yufei, V. J. Cunnane, D. J. Schiffrin et al.,
5. A. G. Volkov, (Ed.), Liquid Interfaces in Chem- J. Chem. Soc., Faraday Trans. 1991, 87,
ical, Biological, and Pharmaceutical Applica- 107–114.
tions, Marcel Dekker, New York, 2001. 23. V. S. Markin, A. G. Volkov in Liquid–Liquid
6. M. Planck, Ann. Phys. 1891, 44, 385–428. Interfaces. Theory and Methods (Eds.: A. G.
7. F. O. Koenig, J. Phys. Chem. 1934, 38, Volkov, D. W. Deamer), CRC Press, Boca
111–128. Raton, London, Tokyo, 1996, pp. 63–75.
8. D. C. Grahame, R. W. Whitney, J. Am. Chem. 24. L. D. Landau, E. M. Lifshitz, Electrodynamics
Soc. 1942, 64, 1548–1552. of Continuous Media, 2nd ed., Pergamon
9. E. J. W. Verwey, K. F. Nielsen, Philos. Mag. Press, New York, 1984.
1939, 28, 435–446. 25. Yu. Kharkats, J. Ulstrup, J. Electroanal. Chem.
10. C. Gavach, P. Seta, B. d’Epenoux, J. Elec- 1991, 308, 17–26.
troanal. Chem. 1977, 83, 225–235. 26. C. W. Outhwaite, L. B. Bhuiyan, S. Levine,
11. V. S. Markin, A. G. Volkov, Prog. Surf. Sci. J. Chem. Soc., Faraday Trans. 2 1980, 76,
1989, 30, 233–356. 1388–1408.
12. V. S. Markin, A. G. Volkov, Electrochim. Acta 27. F. H. Stilinger, J. G. Kirkwood, J. Chem.
1989, 34, 93–107. Phys. 1960, 33, 1282–1290.
13. R. S. Hansen, J. Phys. Chem. 1962, 66, 28. G. M. Torrie, J. P. Valleau, J. Electroanal.
410–415. Chem. 1986, 206, 69–79.
14. A. G. Volkov, Langmuir 1996, 12, 3315–3319. 29. Q. Cui, G. Y. Zhu, E. K. Wang, J. Electroanal.
15. T. Kakiuchi, M. Kobayashi, M. Senda, Bull. Chem. 1995, 383, 7–12.
Chem. Soc. Jpn. 1988, 61, 1545–1550.
2.4 metal–electrolyte phase boundary struc-

Electrical Double Layers. Double Layers at ture (see also Vol. 2). A large number of
Single-crystal and Polycrystalline Electrodes techniques have been used for the exper-
imental study of the edl structure and for
Enn Lust
the determination of the potential of zero
University of Tartu, Tartu, Estonia
charge (pzc) Eσ =0 , which is a fundamental
2.4.1 parameter in electrochemistry [1]. How-
General Aspects ever, the greatest success at solid polycrys-
talline (PC) and single-crystal electrodes
Remarkable progress has been made in has been achieved mainly by impedance,
the last few years in electrochemistry on optical, scrape, and surface tension meth-
single-crystal surfaces [1–10]. This par- ods, and therefore in this review more
allels with the progress on the surface attention has been paid to the data obtained
science and it has been partly simulated using these methods [1–5, 12, 13].
by developments in that field especially The charged particles may or may
regarding the preparation and characteri- not cross the interface between two
zation of surfaces. The most widely used phases and in this respect, interphases
techniques for the determination of the may be divided into the two limiting
structure of single-crystal surfaces are the types: nonpolarizable and polarizable,
low-energy electron diffraction (LEED), respectively [12]. The analysis by Frumkin
scanning tunneling microscopy (STM), et al. [14] in 1970 led to the conclusion that
atomic force microscopy (AFM), X-ray scat- it is possible to treat thermodynamically
tering (SXRS), and other methods [4–6, not only the polarizable but also the
8–10]. Auger electron spectroscopy (AES) nonpolarizable electrode cases, and to
is used to characterize the chemical com- define the charge density Q = σ + AH+ ,
position or cleanliness of surfaces. To the where σ is the surface charge density and
first approximation, the low Miller in- AH+ is the fraction of charge that has
dexes reflect the symmetry of the bulk crossed the interface moving from one
structure and they have lower surface en- phase to the other. Thus, it has been
ergy than vicinal surfaces (the high index proposed to call the potential at which
planes). However, LEED, AFM, STM and Q = 0 the potential of zero total charge
SXRS studies have revealed two important (PZTC) and the potential at which Q =
features of atomic surface structure: relax- σ = 0 the potential of zero free charge
ation and reconstruction [4–6, 8–11] (see (PZFC) (see also Chap. 1 in Vol. 1). A
also Chaps. 14–21 in Vol. 3). PZTC has been observed and measured
for Pt-group metals, and this is due to the
2.4.1.1 The Potential of Zero Charge and chemisorption of hydrogen atoms. In all
Electric Double-layer Structure other cases the PZFC is usually observed
The electric double layer (edl) is fun- and for simplicity will be termed as a pzc,
damental for electrochemistry because denoted by Eσ =0 .
the rate and mechanism of the vari-
ous electrochemical reactions (hydrogen 2.4.1.2 Differential Capacitance of
evolution, corrosion and corrosion inhibi- Electrode–Electrolyte Solution Interface
tion by surfactants, metal deposition and For (ideally) polarizable metals (i.e. elec-
dissolution, and so forth) depend on the trodes having a large energy barrier for
2.4 Electrical Double Layers. Double Layers at Single-crystal and Polycrystalline Electrodes 189
charge transfer) with a sufficiently broad (GCSG) model [1], in a surface-inactive

double-layer region (such as Hg, Ag, Au, electrolyte solution, the potential of the
Bi, Sn, Pb, Cd, Tl, and others [1]), Eσ =0 can minimum in C, E curve for cel ≤ 10−2 M
be obtained from measurements of the edl would correspond to the condition σ = 0,
capacitance in dilute solutions, where it is that is, to Eσ =0 value. Exact values of Eσ =0
detected by a pronounced minimum in the in the absence of adsorption can be ob-
capacitance–potential (C, E) curve. The retained by linear extrapolation of Emin as a
lated differential capacitance C, defined as function of the electrolyte concentration at
2 cel ≤ 1 × 10−2 M [5, 15, 16].
∂σ ∂ γ Models put forward in the 1960s to
C= =− (1)
∂E T ,p ∂E 2 T ,P ,µi 1970s by several authors for describing
the metal–solution interface did not take
can be measured by a great variety of into account the metal contribution and
techniques [1, 5, 9] (see also Chaps. 1–9 in consider the solid as a perfect conductor,
Vol. 3). In Eq. (1), γ is the interfacial (for so that the potential drop in the metal
liquid electrodes) or superficial (specific layer is charge-independent and the poten-
surface) work (for solid electrodes). tial drop at the metal–solution interface
It is usually assumed that the edl at is just given by the solvent molecules and
the ideally polarizable metal–electrolyte free ionic charges. Although early theories
interfaces is splitted into two parts: the considered the penetration of the electric
inner layer and the diffuse layer, and field into the metal, this idea was sub-
can be represented by two capacitances sequently dropped for a long time [1, 9,
in series [1, 5, 7, 9]: 17–19]. Penetration of the electric field
into the solid has been found to be impor-
C −1 = Ci−1 + Cd−1 (2) tant for semiconductors, and its possible
relevance to semimetals such as Bi and
Here, Ci is the inner-layer capacitance,
Sb was suggested [1, 5, 9, 20, 21]. In
independent of the surface-inactive elec-
the past 20 years different models have
trolyte concentration and Cd is the diffuse
been proposed, describing the traditional
(Gouy) layer capacitance, expressed ac-
so-called inner-layer capacity as 1/Ci =
cording to the Gouy–Chapman theory for
1/CM + 1/Cdip + 1/CH , where CM , Cdip ,
a z,z-type electrolyte by
and CH are the capacitances of the metal
dσ |z|F 2 phase, the dipole (solvent) layer, and the
Cd = = 4A c + σ 2 (3) Helmholtz layer, respectively. However, it
dφd 2RT
is impossible experimentally to separate
where φd = 2RT (|z|F )−1 arcsh(0.5σ A−1 the overall capacity of inner layer into its
c−1/2 ) is the potential drop in the dif- components [1–5] (see also Chap. 2.3 in
√
fuse layer and A = 2ε0 εd RT ; ε0 is Vol. 1).
the permittivity of vacuum and εd is
the dielectric constant of the diffuse 2.4.1.3 Electrical Double-layer Structure at
layer, usually taken equal to the macro- Solid Electrodes
scopic dielectric permittivity of the sol- During several decades, the quantitative
vent. Thus, at Eσ =0 , the value of Cd analysis of the edl structure, adsorption of
√
decreases linearly with c. According ions and molecules on the ideally polar-
to the Gouy–Chapman–Stern–Grahame izable solid electrodes was carried out on
the basis of relationships, developed for the edl and adsorption parameters (e.g.
the liquid Hg surface without any con- Eσ =0 , interaction parameter in the adsorp-
sideration of the structure of the solid tion (Frumkin’s) isotherm, maximal Gibbs
surface. But the theoretical analysis of the adsorption max , adsorption equilibrium
electrochemical characteristics for the PC constants, and so forth), obtained for the
solid electrodes, even for simple metals, is PC electrodes, often may be apparent [1–5,
considerably complicated in comparison 7, 15, 16, 20–22].
with that for Hg. One of the reasons is It may be noted that for a PC sur-
both the differences between properties of face, there are complications in defining
individual faces of the metal and the in- the PZC because one does not have
fluence of various defects of the surface zero-charge density of all crystal faces si-
structure [1–10]. In the beginning of the multaneously, exposed on PC electrode.
1970s, by a careful experimental study of The work function is different for vari-
the adsorption of pyridine on the solid ous homogeneous regions (planes), but if
bismuth drop electrode of known PC sur- the grains are in contact, the electrochem-
face structure and of tetraalkylammonium ical potential of the electron has a single
ions on PC Zn, the phenomenon of split- value. Thus, the Volta or outer potential
ting the adsorption–desorption maximum is different for various faces, so differ-
was discovered [1, 20, 21] (Fig. 1). Already ent surface charge densities are present.
the first investigations showed that split- Similarly, the surface structure of solid
ting the maxima was due to the PC nature metal makes the electronic structure more
of the surface, and also to the nature of complicated than that for a liquid metal or
the adsorbing particle and the solvent, to for a single-crystal plane. Certainly, the use
the surface charge density σ , and to the so- of the density properties depending only
lution concentration. On account of that, on distance perpendicular to the interface
140 60
120 50
4
100 3 40
[µF cm−2]
[µF cm−2]
80
C
30
60
20
40 2 1
1
20 10 2
3
0 0
−2 −1.8 −1.6 −1.4 −1.8 −1.3 −0.8
E (SCE) E (SCE)
(a) [V] (b) [V]
Fig. 1 Differential capacitance versus electrode potential dependence: (a) for PC Zn in
0.1 M KI containing TBAI (M): (1) 0, (2) 3 × 10−4 , (3) 3 × 10−3 , (4) 3 × 10−2 ; (b) for PC Bi
in 0.1 M KF containing pyridine (M): (1) 0, (2) 0.1, (3) 0.2 (updated from Refs. [1, 20, 21]).
is an oversimplification [1, 4, 5, 7–10, 15, used as a probe particle, is smaller than

16, 20–22]. the smallest surface irregularity, the entire
surface can be evaluated. But each method
2.4.1.4 Surface Roughness of Solid is applicable to a limited number of elec-
Electrodes trochemical systems so that a universal
Surface roughness is an important prop- method of surface area measurements is
erty in electrochemistry of solid electrodes not available for the time being. Thus, it is
as the most of edl and adsorption char- useful to stress again that the value of R
acteristics, as well as kinetic parameters, depends on the method used [1–5, 15, 16,
are extensive quantities and are referred to 22, 23].
the apparent unit (flat cross-section) area
of electrode surface [1–5]. The examina- 2.4.1.5 The Applicability of the
tion of the working area of solid electrodes Gouy–Chapman–Stern–Grahame Model
is a difficult matter owing to the irregu- to Solid Electrodes
larities at a submicroscopic level. For the
determination of the real surface area of 2.4.1.5.1 Shape of C, E Curves Depen-
the solid electrodes, different in situ and dence of the C, E curves for a solid metal
ex situ methods have been proposed and on a method of electrode surface prepara-
used, which are discussed in Refs. [5, 23]. tion was reported long ago [1–5, 7, 20–22,
The in situ methods more commonly used 25, 26]. In addition to the influence of im-
in electrochemistry to obtain the surface purities and faradaic processes, variation
roughness in the surface roughness was pointed out
Sreal as a possible reason for the effect. For
R= (4)
Sgeom the determination of R, it was first pro-
posed to compare the values of C of the
(where Sreal and Sgeom are the real solid metal (M) with that of Hg, that is,
(working) surface and geometrical area, R = C M /C Hg . The data at Eσ =0 for the
respectively) are (1) differential capaci- most dilute solution were typically used
tance measurements in the region of for such a comparison to eliminate the
ideal polarizability, Parsons–Zobel (PZ) influence of possible differences in the
plot method [24], Valette–Hamelin ap- inner-layer capacities. However, Ci of dif-
proach [25, 26], and other similar meth- ferent solid metals and Hg has shown
ods [1, 5, 7]; (2) mass transport under diffu- such a large variation that this approach
sion control with assumption of homoge- can hardly be considered as appropriate. It
neous current distribution; (3) adsorption is to be noted that the error in C, which
of ratioactive organic compounds hy- for solid electrodes is much higher than
drogen, oxygen, or metal monolayers; for liquid electrodes, increases with the
(4) microscopy (optical, electron, STM, and decrease of cel ; further, as shown later
AFM); and (5) quartz microbalance, as well (Sect. 2.4.1.5.2), the effects of the surface
as a number of ex situ methods [23]. It crystallographic inhomogeneity on C val-
is to be noted that depending on the ues also prove especially appreciable in the
irregularity-to-probe size ratio, either the dilute electrolytes [1–5, 7, 15, 16, 22].
entire surface or only a fraction of it is Frumkin was the first to give a qualitative
accessible to a particular measurement. consideration of the electrochemical prop-
Only when the size of the molecule or ion, erties of PC electrodes. He noted that the
charge σj at individual faces j may be dif- with σ = 0. The same conclusions hold
ferent at E = const and this may change for PC Au (σAu PC = −0.03 C m2 ), PC Bi
the form of the capacitance curve near (σBi = −0.01 C m−2 ), and for other PC
PC
the diffuse-layer capacitance minimum [1]. electrodes [5].
Important results were obtained in a pio- Mathematical simulation of CPC , E
neering paper by Valette and Hamelin [25, curves shows that the shape of the diffuse-
26]. They compared experimental capaci- layer capacitance minimum depends on
tance curves for a PC Ag electrode and Ag the difference of Eσ =0 for individual faces,
faces. It was found that the capacitance of exposed on the PC electrode surface, and
a PC Ag electrode can be obtained by the their fractions, as well as on the shape of
superposition of the corresponding Cj , E partial Cj , E curves (Fig. 3). The results of
curves for individual faces exposed at the experimental capacitance studies at two-
PC surface: plane model PC Bi electrodes were in
agreement with these conclusions [5, 15,
CPC (E) = θj Cj (E) (5)
16]. Thus it has been shown that the
j
potential of the diffuse-layer capacitance
where θj is the fraction of the face on minimum for a PC electrode with notice-
the PC electrode surface. A weighed sum able difference in Eσ =0 values for various
of C, E curves for the faces was found planes does not correspond to the PZC
to be similar to the C, E curve for a PC of the whole surface and j θj σj = 0 at
electrode, and thus all main Ag faces are Emin .
exposed on the surface. The diffuse-layer
capacitance minimum potential Emin PC for
2.4.1.5.2 Parsons and Zobel Plot Method
a PC Ag electrode was only slightly less Substantial contributions to the interpre-
negative (30 mV) than the PZC of the tation of the experimental data for solid
Ag(110) face (i.e. for the face with the
electrodes have been made by Leikis
more negative value of Eσ =0 ). The diffuse-
et al. [28] and by Valette and Hamelin [25,
layer capacitance minimum for PC Ag was
26]. Both approaches are based on the
wider and less deeper than for the Ag
same model: the value of Eσ =0 and of the
faces.
inner-layer capacitance per unit ‘‘true’’ sur-
Thereafter, the influence of the crys-
face area, Ci , are assumed to be constant
tallographic inhomogeneity of PC and
over the whole PC surface and the value
monocrystalline electrodes (with various PC was identified by E
of Emin σ =0 . Thus, the
surface defects) has been discussed for
GCSG model is considered as applicable
various metals in many papers [1–5, 7,
to the capacitance characteristics related
15, 16, 22] (Fig. 2). Bagotskaya et al. [27]
to the unit of ‘‘true’’ surface area, which
showed that integration of the partial C,
differ by a factor R −1 from those per unit
E curves from Eσ =0 of each face to
‘‘apparent’’ surface area
the minimum potential of the CPC , E
PC , on the PC electrode, taking
curve, Emin Creal (E) = R −1 Cgeom (E) and
into account the fraction of plane, gives
PC = −0.04 C m−2 at E PC . At E PC , the
σAg min min σreal (E) = R −1 σgeom (E) (6)
Ag(110) plane has a positive σ and other −1 −1
[Ci (σreal )] = [Creal (σreal )]
planes have a negative σ , thus, at the PC
Ag surface there are no surface regions − [Cd (σreal )]−1 (7)
20
1
3
15
[µF cm−2]
C
4 5
10
7
9
10
5
−1.6 −1.1 −0.6
E vs SCE
[V]
Fig. 2 C, E curves (ν = 210 Hz) for (BiDE )R (1, 4, 8), Bi(111)EP (2, 6, 10),
(BiPC )MP CE (3, 7), and Bi(111)CE (5, 9) in aqueous solution of NaF, M: (1,
2) – 0.1; (3–6) – 0.01; (7–10) – 0.002.
where Cd (σ ) is the diffuse-layer capac- shown that the test is somewhat more
itance obtained according to the Gouy complicated [1–5, 7, 29]. The PZ plots
theory [1–5, 7]. Extrapolation of Cgeom −1 for Hg–surface-inactive electrolyte solu-
−1 −1
versus Cd to Cgeom = 0 provides the tion interfaces at σ = 0 and σ < 0, albeit
−1 usually linear, exhibit reciprocal slopes that
inner-layer capacitance as R −1 Ci,geom . The
are somewhat greater than unity. The main
inverse slope of the PZ plot at EminPC is
reason for this is the experimental error in
identified by R. In the absence of ion- measuring C, as well as the hyperbolic
specific adsorption and for ideally smooth form of C in the GCSG model [15, 16,
surfaces, these plots should be linear 22, 29].
with the unit slope. However, data for The error in Ci is the total differential of
Hg and single-crystal face electrodes have Eq. (2):
40 40
[µF cm−2]
[µF cm−2]
C
C
4
20 20
1 2 3 1 2 3
0.5 1.0 0.5 1.0

−E (SCE) −E (SCE)
(a) [V] (b) [V]
15
1
2
3
[µF cm−2]
4
C
10
0.75 1.00
−E (SCE)
(c) [V]
Fig. 3 Theoretical C, E curves (1, 2, and 3) for single-crystal planes and for a model PC
surface (4) calculated by the superposition of the C, E curves at E = const with
θ1 = θ2 = θ3 = 1/3, (1) faces with high hydrophilicity; (2, 3) faces with weak
hydrophilicity. (1, 2) Eσ =0 (max) = 0.4 V and (3) Eσ =0 (max) = 0.09 V.
1 1 with (1) experimental measurement of C,

dy = dCi = · dC (8)
x2 C2 (2) preparation of solutions of an exact cel ,
(3) incomplete dissociation of electrolytes,
where dC, for a given value of σ ,
(4) slight specific adsorption of anions, and
includes the experimental error in the
determination of C and the error in the (5) deviations of diffuse layer from GC
integration of the C, E curves. When theory. In the case of solid electrodes, in
x(x = C −1 − Cd−1 ) is small, dCi is large addition to the above-mentioned reasons,
and tends to infinity; when x is large, dCi sources of inaccuracy are the possible er-
is small and tends to dC. For a given ratic preparation of the electrodes with
positive x, the smaller the Cd , the more the same geometrical surface area and the
an error in C affects Ci . The same error same crystallographic orientation [1–5, 15,
for Bi, Cd, and Ag at fixed cel causes the 16, 22, 29]. Studies with wedge-shaped,
error in Ci to increase in the same order of two-faced Bi electrodes show that with
metals since the value of Ci increases. The increasing Eσ =0 of different faces ex-
same experimental error entails a larger posed at a model PC electrode surface, the
uncertainty in Ci for the lowest cel and deviation of the PZ plot from linearity
σ . At |σ | 0, the uncertainty in C does increases and the value of R also in-
not bear on Ci since x is large [5, 15, 16, creases [15, 16] (Fig. 4). This was explained
22, 29]. by the fact that R is a complex quantity,
In the case of liquid Hg, the un- being R = fR fCR , where fCR is a factor of
certainty in the measurements may be crystallographic inhomogeneity of the PC
induced by possible errors connected electrode surface [15, 16, 22, 25–30] and
0.2
1
2
0.16
3
4
[cm2 µF−1]
0.12
1/C(exp)
0.08
0.04
0
0 0.04 0.08 0.12 0.16
1/Cd(theor)
[cm2 µF−1]
Fig. 4 (Ci )−1 versus (Cd )−1 dependence for wedge-shaped two-plane electrode: (1) 0.5(011̄) +
0.5(1̄01̄); (2) 0.8(011̄) + 0.2(111); (3) 0.3(011̄) + 0.7(111); and (4) 0.5(011̄) + 0.5(111) with
Eσ =0 = 15 mV (1) and Eσ =0 = 70 mV (2–4) (Ci , inner-layer capacitance; Cd , diffuse-layer
capacitance) (updated from Refs. [15, 16]).
fR was assumed to be the actual surface factor R of solid PC surfaces and to test
roughness factor independent of cel and σ . the GCSG theory on the basis of Eqs. (6)
The more dilute the solution and the larger and (7). For each cel , a set of Ci , σ curves
the difference between Eσ =0 of individ- was calculated for various R values and
ual planes (i.e. the homogeneous regions) the optimum value of R was selected on
exposed at a PC surface the higher the the basis of the assumption that near
values of fCR have been established. Thus, Emin the Ci , σ curve must be smooth.
fCR decreases as |σ | increases. A compar- The experimental values of R were found
ison of the data for F− , BF4 − , and ClO4 − to increase as cel decreased from 1.40 to
solutions shows that R increases in the or- 1.80. This was explained by the fact that R
der F− < BF4 − < ClO4 − with increasing is a complex quantity, being R = fR fCR .
weak specific adsorption [5, 15, 16, 22]. Later, the influence of the crystallographic
Thus, the experimental value of R de- inhomogeneity of PC and monocrystalline
pends on the surface charge density and electrodes (with various surface defects)
sometimes on the electrolyte concentra- was discussed in many papers [1–5, 7, 15,
tion cel . The real R cannot depend on 16, 22, 25–29] (Fig. 5).
σ and on cel . Experimental PZ plots for
single-crystal faces in the region of Eσ =0 2.4.1.5.4 Polycrystalline Electrode Models
(−3 < σ < 3 µC cm−2 ) often show slopes The edl models for PC electrodes may
increasing with |σ |, that is, the apparent be classified in two groups [5, 7, 10,
R decreases as |σ | rises [1–5, 7, 15, 16, 27, 30]. The first group considers the
22]. These findings indicate that the ex- PC electrode surface as one consist-
perimental R (the inverse slope value of ing only of relatively large homogeneous
the PZ plot) is not a real measure of the (monocrystalline) regions with linear pa-
actual R. However, the practically unit rameter y ∗ 10 nm, which corresponds
slope of the C −1 , Cd−1 -plots for Hg, Bi, to macropolycrystallinity. Within these
Cd, Sb, Ag, and for other systems with surface regions, both the compact and the
correlation coefficients better than 0.996 diffuse layers at different homogeneous
provides convincing evidence both for the areas can be viewed as independent ones,
validity of the GC theory and for the lack and accordingly
of experimentally detectable deviations of Xj Cij Cdj
app
the roughness factor from unity. Slopes of CPC = R (9)
Cij + Cdj
C −1 , Cd−1 plots much lower than unity very j
near Eσ =0 (−0.5 ≤ σ < 1 µC cm−2 ) can
where Cij and Cdj are the inner- and
be interpreted as deviations from the sim-
diffuse-layer capacities for the plane j ,
ple GC theory caused by the geometrical
respectively. Therefore, this is the so-called
roughness and energetical inhomogeneity
model of independent diffuse layers (IDL)
of the solid electrode surface [2–5, 15, 16,
(Fig. 6).
22, 25–29].
In terms of the second group of mod-
els, the PC surface consists only of very
2.4.1.5.3 Surface Roughness and Shape of small crystallites with linear parameter
Inner-layer Capacitance Curves In 1973, y ∗ , whose sizes are comparable with the
Valette and Hamelin [25, 26] proposed an- edl parameters in moderate electrolyte
other method to determine the roughness concentrations. For such electrodes, the
140
120
100 1
3 2
80
[µF cm−2]
Ci
60
4
40
20
0
−12 −8 −4 0 4
σ
[µC cm−2]
Fig. 5 The inner-layer capacitance versus electrode charge dependence for
the PC Cd electrode at different values of roughness factor: (1) 1; (2) 1.08;
(4) 1.15; (3) extrapolation by an approximate equation determined between
−1.0 and −10 µC cm−2 .
compact layers at different monocrys- app

CPC can be obtained by the equation
talline areas are considered to be inde- app

pendent, but the diffuse layer is common CPC = Cd (σ̄ )R Xj Cij
for the entire surface of a PC electrode j
 −1
and depends on the total medium charge

density σ̄PC = R j Xj σj , averaged over × Cd (σ̄ ) + R Xj Cij  (10)
the PC electrode surface, and capacitance j
CH, 1(σ1) CD, 1(σ1) Fig. 6 Theoretical model of the

electrical double layer at an ideally
CH, 2(σ 2) CD, 2(σ 2) polarizable electrode with a PC surface
structure. (a) Model of independent
CH, n(σ n) CD, n(σ n) diffuse layers (Eq. 9); (b) model of
common diffuse layer (Eq. 10).
(a)
CH, n + 1(σ n + 1)
CH, n + 2(σ n + 2) CD(σ )
CH, m(σ m)
(b)
This model is known as the model of surface regions (y ∗ > 10 nm), and be-
the common diffuse layer (CDL) [27]. tween the large homogeneous areas there
As shown in Ref. [30], both models can are aggregates, which consist of many very
describe only some limiting cases and the small crystallites, whose sizes are probably
exposition for the total capacity of the PC smaller than the effective Debye screening
electrode (equivalent circuit) investigated lengths. Therefore, the structure of edl on
depends on the relationship between a real PC electrode can be expressed by the
the three lengths: (1) the characteristic relations
size of the individual planes at a PC app
CPC
electrode surface y ∗ ; (2) the effective
screening length in the bulk of the
n
Xj Cij Cdj
=R + Xm+1

layer near the face j ldj ≡ ld (σj ) =
diffuse
j =1
Cij + Cdj
ld / 1 + (σj /σ ∗ )2 , where ld is the Debye  
screening length and σ ∗ = εRT /2πF ld ;
m
 Cd (σ̄ )R Xj Cij 
(3) εlij , where ε is the bulk dielectric  
 j =n+1 
constant of the solvent and the length lij = ×  (11)

m 
(1/4)πCij is determined by the capacity of  C (σ̄ ) + R Xj C 
dj ij
the compact layer of the face j . According j =n+1
to the theoretical analysis, the CDL model

n
m
is valid for PC electrodes with very small Xm+1 + Xj = 1, Xj = 1,
grains (y ∗ < 5 ÷ 10 nm) first of all with j =1 j =n+1
a moderate difference in Eσ =0 for the m>n
faces ( Eσ =0 = 0.1 ÷ 0.15 V) and for very
dilute electrolyte solutions (c ≤ 0.01 M) The results of computer simulation of
exp
near the point of total zero charge. Thus, many experimental CPC , E curves for
if the value of y ∗ is much less than two of various bismuth solid drop electrodes
the three lengths: y ∗ εli1 ∼ εli2 , ld1 , ld2 , show that 10 to 30% of the whole surface
the CDL model is valid. For other cases, of bismuth solid drop electrode is covered
the model of IDL would be valid [30]. with small crystallites (y ∗ < 10 nm), and
The real PC electrode usually consists ¯
the standard deviation ( C)
exp
of CPC , E
app
of the comparatively large homogeneous curves from CPC , E curves is smaller if
Eq. (11) is used instead of Eqs. (9) or (10). smallest characteristic length of surface
The data for the wedge-shaped two-plane inhomogeneity, R̃(κ) can be expressed as
electrode show that the share of small
crystallites at the surface of this electrode 1 − H 2
R̃(κ) ≈ R (14)
is not more than 5 to 10% [15, 16, 22]. 2Rκ 2
where H 2 is the mean square curvature

2.4.1.5.5 Debye Length–dependent Sur- of surface [31–33]. Equation (14) shows
face Roughness Model Daikhin et al. [31] that the roughness function R̃(κ) ap-
have shown that the competition between proaches the geometrical roughness factor
the Debye length and the characteristic R at small lD . With the increase of lD it
sizes of roughness would modify the GC decreases with respect to R, and the cor-
result for the diffuse-layer capacitance. The rection is proportional to the square of lD .
limiting value of the capacitance at short (2) In the range of large Debye lengths (low
Debye length should follow the Gouy re- concentrations), R̃(κ) can be expressed as
sult, but with surface area value replaced
by Sreal = RSgeom . Thus, the diffuse layer κh2
simply follows every bump or dip of the R̃(κ) ≈ 1 + −κ 2 h2 (15)
L
electrode, the surface of which looks flat
at the Debye scale. In the limit of long lD where h is the height of the character-
(dilute solutions), the roughness cannot be istic size of roughness in the z-direction
manifested in the capacitance that would (h denotes the root-mean-square depar-
obey the native Gouy expression and the ture of the surface from flatness); L is
analogy with macroscopic capacitance can the length that is in the order of the
be used. The crossover between these two maximal correlation length lmax . At very
limits can be simulated as low electrolyte concentrations (lD > lmax ),
the roughness of surface is not detectable
C = R̃(κ)CGC (12) in the capacitance and the first correc-
tion to the flat surface result is linear
where the roughness function, R̃(κ), varies in κ. Intermediate to these two limiting
between R̃(0) = 1 and R̃(∞) = R > 1; κ cases, the specific form of the roughness
is the Gouy length (the inverse Debye function depends on the morphology of
length). It was demonstrated that the PZ the interface. The various geometrical sur-
plot method is not the most convenient face layer models – sinusoidal corrugation,
tool for the characterization of surface random Gaussian roughness, and periodi-
roughness and the more informative cal system of linear defects and rectangular
would be the plot of R̃(κ) versus κ, where grating – have been analyzed [31–33].
−1
R̃(κ) ≈ (Cexp − Ci−1 )−1 CGC
−1
(13) Experimental results of electrochem-
ically polished, cut, electrochemically
and Ci is evaluated from the measure- etched, and chemically etched Bi, Sb, and
ments at high electrolyte concentration ac- Cd electrodes (Fig. 7) demonstrate that
cording to Valette–Hamelin approach [25, R̃(κ) rises with the increase of surface
26]. In the case of weak nonfractal surface roughness, in agreement with AFM data
roughness the two limiting approxima- and conclusions of Refs. [32, 33].
tions are valid: (1) At high electrolyte The nonlinear version of roughness the-
concentrations, when lD is shorter than the ory applicable for arbitrary large charges
1
2
3
2 4
R (κ )
5
6
~
7
8
9
1
0 0.4 0.8 1.2
κ
[nm−1]
Fig. 7 Roughness function versus inverse Debye length plots of Bi(111)CE (1); Bi(111)ECE
(2); Bi(DE)R (3); and for various theoretical simulation models: sinusoidal corrugation
(SC) – h = 30, lSC = 50 (4); random Gaussian roughness (RGR) – lRGR = 50, h = 30 (5);
SC − h = 5.0, lSC = 2.5 (6); RGR − h = 2, lRGR = 1 (7); SC − h = 0.85, lSC = 2.5 (8); and
RGR − h = 0.9, lRGR = 2.0 (9) (all parameters in nm) (l, characteristic lateral correlation
length; h, root-mean-square height of surface roughness).
has been worked out in Ref. [34]. Far value R̃max > R, and the crossover occurs
from Eσ =0 the effects of nonlinear screen- at ER ≈ 65 mV (T ≈ 298 K).
ing have been found, which lead to the This model has been used for inter-
charge dependence of PZ plot, observed pretation of the experimental capacitance
experimentally in many works [1–5, 7, data of rough Cd electrodes (Fig. 8). In
25–33]. Thus, the roughness factor of PC the region of moderate Er , the rough-
surface depends on the effective diffuseness function rises with the decrease of
layer thickness κeff (Er ) = κ cosh(eβEr /2), the effective Debye length leff . At higher
where eβEr is the dimensionless electrode Er , the surface roughness function lev-
potential, Er = Ex − Eσ =0 is the ratio- els off to the constant value, that is, to
nal electrode potential, β = (kB T )−1 (kB the so-called geometrical surface rough-
is the Boltzmann constant and e is the ness value. In the region of Eσ =0 there
elementary charge); Ex is an electrode are deviations from the theory caused
potential. According to theoretical calcu- by the energetic inhomogeneity of elec-
lations, R̃(κ, E) is minimal close to Eσ =0 trode surface. It is to be noted that the
and rises to the limiting value at mod- new effective Debye length–dependent
erate and high σ . According to Ref. [34], surface roughness theory needs future cor-
for a fixed value of κ, the effect of sur- rections taking into account the role of
face structure on the capacitance increases crystallographic inhomogeneity of the ge-
with Er . However, in contrast to the ometrically rough electrode surface, that
case of low ER , R̃(κ, E) could be a non- is, the dependence of the local surface
monotonous function of κ with a maximal charge density on the crystallographic
1.9
1.6
R (κ , E)
1.3
~
1
2
1.0
3
4 5
0.7
−0.5 −0.4 −0.3 −0.2 −0.1 0.0
E vs SCE
[V]
Fig. 8 Experimental roughness function R̃(κ, E) versus E curves for electrochemically etched
Cd(0001) for the fixed electrolyte (NaF + H2 O) concentrations: (1) 0.05 M; (2) 0.02 M;
(3) 0.007 M; (4) 0.002 M; (5) 0.001 M.
structure of energetically homogeneous of the interface according to the definition

regions exposed on the surface of the of Mandelbrot [35–37]. Such a structural
macropolycrystalline electrode. nonuniformity may result in a nonuniform
distribution of σ at the electrode surface
2.4.1.5.6 Electric Double Layer and Frac- due to the different Eσ =0 of the differ-
tal Structure of Surface Electrochemical ent grains existing at the electrode surface.
impedance spectroscopy (EIS) in a suffi- The fractal carpet model [35, 36] is repre-
ciently broad frequency range is a method sentative of this approach. The impedance
well suited for the determination of equi- of a fractal electrode is
librium and kinetic parameters (faradaic Z = RS + Z0 (j ω)−α (16)
or non-faradaic) at a given applied poten-
tial. The main difficulty in the analysis of where Z0 is a preexponential factor (anal-
impedance spectra of solid electrodes is the ogous to the inverse of the capacitance of
√ ω = 2πν is the angular fre-
‘‘frequency dispersion’’ of the impedance the edl (1/C)),
values, referred to the constant phase quency, j = −1 and α is a dimensionless
or fractal behavior and modeled in the parameter with a value usually between 0.5
equivalent circuit by the so-called con- and 1. The CPE angle ϕ is related to α by
stant phase element (CPE) [5, 15, 16, 22, 35,
π
36]. The frequency dependence is usually ϕ= (1 − α) (17)
attributed to the geometrical nonunifor- 2
mity and the roughness of PC surfaces The value α = 1 corresponds to ideal
having fractal nature with so-called self- capacitive behavior. The so-called fractal di-
similarity or self-affinity of the structure mension D introduced by Mandelbrot [37]
resulting in an unusual fractal dimension is a quantity that attains a value between
2 and 3 for a fractal structure and reduces anion from H2 O at Ag is mainly

to 2 when the surface is flat. According to caused by the squeezing-out effect [4, 5,
Ref. [35, 36], D is related to α by 38].
The inner-layer capacitance Ciσ =0 in-
1
α= (18) creases in the sequence Ag(111) < Ag(100)
(D − 1)
< Ag(110) as the reticular density of
The CPE model has been used to study EP planes decreases, in good agree-
PC Au, Cd, Ag, Bi, Sb and it has ment with the conclusions of Leiva and
been found that for electrochemically Schmickler [18], and with the data of
polished electrodes the surface roughness quantum-chemical calculations of water
is very small compared with mechanically adsorption at metal clusters [40]. For EG
polished surfaces [5, 15, 16, 22, 35, 36] (see electrodes, Ci varies in the reverse order
also Chap. 8 in Vol. 3). and the values of Ciσ =0 for quasi-perfect Ag
planes are remarkably higher than those
2.4.2 for real Ag planes [2–5].
Experimental Aspects. Short Review of Data The contradiction in the data might be
explained by various concentrations of sur-
2.4.2.1 Silver face defects that are claimed to be minimal
Ag crystallizes in the face-centered cu- on EG Ag(111) and Ag(100), and some-
bic (fcc) system. Various procedures for what higher on EP and CP electrodes.
the preparation of the Ag electrode sur- EG Ag(100), Ag(111), and Cd(1000) were
face have been used [5, 7]. Vitanov et al. [2, found to be free of screw dislocations and
3, 38] have used electrodes electrochem- atomically smooth. However, the surface
ically grown (EG) in a Teflon capillary
structure of real Ag single-crystal faces was
(so-called quasi-perfect Ag(100) and Ag(111)
observed by STM to be very smooth and
planes). Valette and Hamelin [25, 26] have
atomically perfect too. For thermally grown
used electrochemically polished (EP) Ag
and thereafter CP Ag surfaces, the correct
single crystals (so-called real planes), and
LEED pattern has been obtained [8]. The
Trasatti et al. [39] have used chemically
large difference of Eσ =0 values existing
polished (CP) Ag crystal faces. Eσ =0 is
found to depend on the crystallographic for the various Ag faces should cause a
orientation of Ag faces (Table 1), increas- marked dependence of Emin on the surface
ing with the atomic density of the faces structure [1–5, 7, 15, 16, 22, 25, 26]. How-
(see also Sect. 2.4.3.1). The specific ad- ever, the Eσ =0 values obtained by different
sorption of anions increases in the order groups demonstrate a surprisingly good
PF6 − ≤ BF4 − ≤ F− < ClO4 − for Ag(100) agreement (Table 1; see also Chaps. 14–20
and Ag(111) and in the order BF4 − ≤ in Vol. 3).
PF6 − < ClO4 − < F− for Ag(110). The ac- The temperature dependence of the
tivity of ClO4 − on Ag electrodes increases edl parameters has been determined for
in the sequence Ag(110) < Ag(100) < real and quasi-perfect Ag planes [4, 5].
Ag(111), but for F− in the reverse or- The electronic structure of Ag single-
der. This was explained by the decrease crystal face H2 O + NaF interfaces has
in hydrophilicity in the order Ag(111) < been studied by Chao et al. by ellipsometry
Ag(100) < Ag(110) because the adsorp- and differential capacitance, which is
tion of the large structure-breaking ClO4 − discussed in Ref. [5].
Tab. 1 Electrical double-layer parameters for Ag, Au, and Cu in aqueous solutions
Electrode Electrolyte Eσ =0 [V] R Atomic density Method References

versus SHE [cm−2 ]
PC AgEP NaF −0.70 to −0.72 1.2 – Impedance 5, 25

’’ ’’ −0.66 ± 0.02 – – Open circuit scrape 5, 13
Ag(111)EG NaF −0.45 ± 0.01 1.02 1.380 × 1015 Impedance 38
Ag(111)EP NaF −0.454 ± 0.010 1.07 ± 0.10 – ’’ 2–5, 26
Ag(111)CP NaF −0.450 ± 0.016 – – ’’ 3, 39
Ag(100)EP,EG,CP NaF −0.621 ± 0.005 1.15 1.195 × 1015 ’’ 2–5, 26, 38, 39
Ag(110)CP NaF, NaBF4 , KPF6 −0.734 ± 0.005 1.10–1.20 0.844 × 1015 ’’ 39
Ag(311)CP NaF, KPF6 −0.664 ± 0.005 – 0.724 × 1015 ’’ ’’
Ag(331)CP NaF, KPF6 −0.670 ± 0.010 – – ’’ ’’
Ag(210)EP NaF −0.750 ± 0.010 – 0.536 × 1015 ’’ 2–5
Au(111) 0.05–0.2 M NaF 0.56 – – ’’ 2
Au(100) ’’ 0.33 – – ’’ 10
Au(110) ’’ 0.19 – – ’’ 2
Au(210) ’’ 0.11 – – ’’ ’’
Au(311) √ ’’ 0.25 – – ’’ ’’
Au(111) (22 × 3) 0.1 M HClO4 0.55 – – Immersion 5
Au(111) (1 × 1) ’’ 0.47 – – Impedance 6
Au(100) ‘‘hex’’ ’’ 0.54 – – ’’ ’’
Au(100) (1 × 1) ’’ 0.32 – – ’’ ’’
PC Cu 0.1 M NaF −0.50 – – Open circuit scrape 1, 13
’’ 0.050 M. Na2 SO4 + TBN+ −0.64 ± 0.05 – – Impedance 45
Cu(111) 0.050 M. Na2 SO4 + TBN+ −0.33 ± 0.02a – 1.772 × 1015 ’’ ’’
’’ 0.001–0.04 M KClO4 −0.20 ± 0.01 1.0 – ’’ 44
Cu(100) 0.050 M. Na2 SO4 + TBN+ −0.46 ± 0.05a – 1.535 × 1015 ’’ 45
’’ 0.001–0.04 M KClO4 −0.54 ± 0.01 1.0 – ’’ 44
Cu(110) 0.05 M Na2 SO4 + TBN+ −0.63 ± 0.05a – 1.084 × 1015 ’’ 45
’’ 0.005 M NaClO4 −0.69 ± 0.01 – – ’’ 46
Note: Abbreviations: TBN+ : tetrabutylammonium ion; SHE: standard hydrogen electrode; PC: polycrystalline; EP: electrochemically polished; CP:
2.4 Electrical Double Layers. Double Layers at Single-crystal and Polycrystalline Electrodes
chemically polished; EG: electrochemically grown.

a By back integration method.
203
The edl at PC Ag–aqueous solution certain precautions are taken with respect
interface has been discussed by Leikis to the applied potential and the electrolyte
et al., Valette and Hamelin, Vorotyntsev, composition. The surface reconstruction
and Trasatti and Lust [2–5, 7, 25–28]. of Au(110) is more rapid than that of
A diffuse-layer minimum was found at Au(111) and Au(100). A comprehensive
Emin = −0.96 ± 0.02 V (SCE), but Emin review of reconstruction phenomena at
and Ci do not correspond to Eσ =0 single-crystal faces of various metals has
and Ciσ =0 (see also Sect. 2.4.1.5). Emin been given by Kolb [6] and Gao et al. [42].
values for PC Ag–nonaqueous electrolyte The edl at a PC Au–solution interface
interface, depending on the solvent nature, has been studied mainly by Clavilier and
have been discussed in Ref. [5]. Nguyen Van Houng, Hamelin, Borkowska
et al., and others [1–7, 43]. A diffuse-layer
2.4.2.2 Gold minimum was found at Emin = −0.04 ±
Au crystallizes in the same system (fcc) 0.02 V (SCE) for PC Au|H2 O (NaF, KClO4 ,
as Ag. Experimental data obtained by HClO4 ), but Emin obtained for PC Au
cyclic voltammetry (CV), electrochemi- does not correspond to σ̄PC = i θi σi = 0
cal impedance, STM, AFM, SXRS, and and Ci at Emin does not correspond to
other techniques (see also Chaps. 14–20 Ciσ =0 [2–5, 7]. The value of Emin de-
in Vol. 3) show that the edl and inter- pends noticeably on the solvent nature
facial properties depend strongly on the and Emin (versus bis-biphenylchromium
crystallographic structure of the electrode (BBCr(I/0)) scale) rises in the order
surface, the surface charge density, as well of solvents MeOH < H2 O < DMSO <
as the chemical composition of the solvent dimethylformamide < propylene carbo-
and the electrolyte [2–11] (Table 1). Most nate < acetonitrile [5, 43].
of the results before 1980 were obtained
with gold faces isolated by a noncontami- 2.4.2.3 Copper
nating resin and cycled before use in dilute Cu crystallizes in the fcc system. The exper-
Na2 SO4 or H2 SO4 solutions until a stable, imental data for single-crystal Cu|H2 O and
reproducible CV was obtained. From 1980 for PC Cu are controversial [2–5, 7, 44, 45].
onward, most of the results were obtained The first studies with Cu(111), Cu(100),
after a flame annealing treatment of Au Cu(110), and PC Cu in surface-inactive
planes, applied in a way similar to that orig- electrolyte solutions show a capacitance
inally used for ordered Pt electrodes [41]. minimum at E less negative than the
The surface cleanliness achieved by this positive limit of ideal polarizability of Cu
method allowed investigations to be made electrodes. More reliable values of Eσ =0 for
in a meaningful way already during the Cu single-crystal faces have been obtained
first or initial scans of the CV. by Lecoeur and Bellier [44] with EP Cu(111)
Over the past 15 years it has been and Cu(100) (Table 1). Foresti et al. [46]
demonstrated by a variety of in situ and have found Eσ =0 = −0.93 ± 0.01 V (SCE)
ex situ techniques that flame-annealed Au for the Cu(110)–aqueous solution inter-
faces are reconstructed in the same way face, and the validity of the GCSG model
as the surfaces of samples prepared in has been verified. As for Zn single-crystal
ultrahigh vacuum (UHV) [2–11] and that electrodes, reliable values of Eσ =0 have
the reconstructed surfaces are stable even been obtained indirectly from the depen-
in contact with an aqueous solution if dence of the adsorption–desorption peak
potential E max of TBN+ on the crystal- studied by impedance. The potential of

lographic orientation [45]. Like other sd the minimum in C, E curves was inde-
metals, σPC = 0 on the PC Cu–H2 O inter- pendent of cNaF , as well as of cNa2 SO4 . The
face at Emin [2–5, 7, 25–27]. dependence of Eσ =0 on the atomic density
In situ AFM studies in aqueous H2 SO4 of the faces is weak, and the value of Eσ =0 ,
and HClO4 solutions show that Cu(111) becoming less negative as the atomic den-
exhibits the correct hexagonal structure. sity decreases (Table 2), is in contradiction
Cu(100) shows a correct structure and to that for other fcc metals [2, 3, 5, 7]. The
spacing at E = −0.63 V (SCE) in 0.1 M inner-layer capacitance at σ = 0 increases
HClO4 , but the Cu(110) surface in dilute in the sequence Pb(100) ≤ Pb(110) ≤
HClO4 (0.003 M) at E = −0.13 V (SCE) Pb(112) < PC Pb < Pb(111), which has
(160 mV more negative than the rest po- been explained by the increasing hy-
tential) exhibits a correct unreconstructed drophilicity in the same order [48].
structure (see also Chaps. 15–17 in Vol. 3). EDL data for PC Pb have been obtained
The strength of the Cu–O bond will be in aqueous, glacial acetic acid, MeOH,
lower on the Cu(111) face than on the EtOH, and dimethylformamide surface-
Cu(100) and Cu(110). These results sug- inactive electrolyte solutions. At Emin
gest that a specific Pourbaix pH–E phase the value of σPC ≈ 0 and thus Emin ≈
diagram is needed to describe the behavior Eσ =0 [1–5, 7, 48].
of each low index face of Cu [47].
2.4.2.5Tin
2.4.2.4 Lead Sn crystallizes in the tetragonal sys-
Pb crystallizes in the fcc system and elec- tem. Eσ =0 increases only slightly with
trochemically polished Pb(111), Pb(100), the atomic density of faces Sn(110) <
Pb(110), and Pb(112) faces have been Sn(100) < Sn(111) [49, 50] (Table 2), in
Tab. 2 Electrical double-layer parameters for Pb [5, 48] and Sn [5, 49, 50] in
various solvents
Electrode Solution Eσ =0 [V] versus SHE R
PC Pb 0.01 M NaF/H2 O −0.60 ± 0.02I 1.10

’’ KF/MeOH −0.43 ± 0.02I 1.16
’’ CH3 COONa/CH3 COOH −0.58 ± 0.02I 1.15
’’ NaClO4 /FM −0.61 ± 0.02IM
Pb(111) 0.01 M NaF/H2 O −0.62 ± 0.01I 1.05
Pb(100) 0.01 M NaF/H2 O −0.59 ± 0.01I 1.10
Pb(110) 0.01 M NaF/H2 O −0.58 ± 0.01I 1.10
Pb(112) 0.01 M NaF/H2 O −0.58 ± 0.01I 1.15
PC Sn NaClO4 + (HClO4 ) −0.43 ± 0.02I –
Sn(001) Na2 SO4 /H2 O −0.37 ± 0.01I 1.1
Sn(1̄10) Na2 SO4 /H2 O −0.37 ± 0.01I 1.1
Sn(110) Na2 SO4 /H2 O −0.38 ± 0.01I 1.1
Note: Abbreviations: I: impedance method; IM: immersion method; PC:

polycrystalline; FM: formamide; MeOH: Methanol.
accordance with the general trend ob- in surface-inactive electrolyte solutions

served for sd and sp metals [2–5, 7]. (NaF, LiClO4 ) the Eσ =0 of Cd (112̄0) is
The applicability of the GCSG theory for slightly shifted to more negative potentials
monocrystalline and PC Sn has been ver- compared with the Eσ =0 of polished PC
ified by Khmelevaya and Damaskin [49, Cd. On a scraped surface of Cd(112̄0),
50]. At Emin , σPC ≈ 0 for PC Sn has been Eσ =0 is more negative than on a polished
obtained. one, and this difference can be related to
the greater inhomogeneity of the scraped
2.4.2.6 Zinc surface. Vitanov, Popov et al. [54, 55] have
Zinc crystallizes in the hexagonal close- studied Cd(0001), EG in a Teflon capillary.
packed (hcp) system. As the zinc dissolu- A linear PZ plot with R = 1.09 has been
found at σ = 0. A slight dependence
tion takes place at potentials very close to
of Ci , σ -curves on cel has been found.
Eσ =0 , the differential capacitance curves in
The temperature dependence of the edl
the region of Eσ =0 in pure surface-inactive
parameters has been studied for Cd(0001).
electrolyte solutions can be determined di-
The positive value of ∂Eσ =0 /∂T (0.33 ±
rectly for the Zn(112̄0) face only, for which
0.03 mV K−1 ) is taken to indicate that
Eσ =0 is more negative than the steady
the H2 O dipoles are oriented with their
state potential [51, 52] (Table 3). The val-
negative end toward the metal surface. Ci
ues of Eσ =0 obtained by the salting-out
has been found to be higher for Cd(0001)
method for Zn(112̄1) and by the depen-
than for PC Cd [54, 55].
dence of Emin on cKCl at cTU = const are
Electrochemically polished and chemi-
in good agreement with the values of Eσ =0
cally treated Cd(0001), Cd(101̄0), Cd(112̄0),
obtained from the C, E curves for pure
Cd(101̄1), and Cd(112̄1) electrodes have
KCl solutions. Eσ =0 values for other faces
been studied by impedance [15, 16, 22],
have been obtained by the dependence
and the main parameters obtained are
of Emin in the C, E curves on cKCl in
given in Table 3. The value of Emin for
solutions in the presence of thiourea at
PC Cd [1–5, 7] lies between the Eσ =0
cTU = 3 × 10−1 M [51, 52]. The maximum
values for single-crystal planes, and at
difference between the Eσ =0 values for
Emin , σPC ≤ 0.01 µC cm−2 . Eσ =0 for a cut
the various faces of Zn is about 90 mV,
Cd(0001)c electrode is in good agreement
and this is in good agreement with earlier
with that for electrochemically polished
estimates [1–5, 7] (see also Sect. 2.4.3.1).
Cd(0001). For chemically treated Cd elec-
The edl at PC Zn–H2 O interfaces has
trodes, a remarkable dependence of Emin
been studied in many works, but the
on cNaF has been observed [5, 22], char-
situation is somewhat ambiguous and acteristic of PC surface. The inner-layer
complex. The values of Emin depend on capacitance of EP Cd faces increases as the
the surface preparation method. The value atomic density decreases, in good agree-
of Eσ =0 for PC Zn lies between Eσ =0 ment with Leiva–Schmickler model and
values for single-crystal planes, and at quantum-chemical calculations [18, 40].
Emin , |σPC | ≤ 0.01 C m−2 [1–5, 7, 51, 52]. It has been suggested that hydrophilic-
ity increases in the order Cd(0001) <
2.4.2.7Cadmium Cd(101̄0) < Cd(112̄0). The same order
Cadmium crystallizes in the hcp system. has been proposed on the basis of
Korotkov et al. [53] have established that data of organic compound adsorption.
Tab. 3 Electrical double-layer parameters of Zn and Cd in various solvents
Electrode Solution Eσ =0 [V] R αa Db Atomic density Reference

versus SHE [cm−2 ]
PC Zn NaClO4 (pH = 7) −0.92 – – – – 51, 52

’’ 0.1 M KCl + TU −0.91 – – – – ’’
Zn(0001) ’’ −0.78 ± 0.02 – – – 1.630 × 1015 ’’
Zn(101̄0) ’’ −0.84 ± 0.02 – – – 1.315 × 1015 ’’
Zn(112̄0) ’’ −0.87 ± 0.02 – – – – ’’
Zn(112̄0) 3 × 10−3 M KCl −0.85 ± 0.01 – – – – ’’
PC Cd KF/H2 O −0.73 ± 0.02 – 0.94 2.06 – 1, 5, 7
’’ LiClO4 /DMF −0.600 ± 0.01 – – – – 5, 7
’’ LiClO4 /i-PrOH −0.75 ± 0.01 – – – – ’’
Cd(0001)EG NaF/H2 O −0.75 ± 0.01 1.09 – – 1.308 × 1015 54, 55
Cd(0001)EP ’’ −0.73 ± 0.01 1.06 0.98 2.01 – 22
Cd(101̄0)EP ’’ −0.77 ± 0.01 1.12 0.97 2.03 1.038 × 1015 ’’
Cd(112̄0)EP ’’ −0.76 ± 0.01 1.1 – – 0.692 × 1015 53
Cd(112̄0)EP ’’ −0.78 ± 0.01 1.22 0.98 2.02 – 22
Cd(101̄1)EP ’’ −0.760 ± 0.015 1.24 0.94 2.07 0.544 × 1015 ’’
Cd(112̄1)EP ’’ −0.750 ± 0.015 1.25 0.95 2.06 0.442 × 1015 ’’
a Fractional
exponent, dimensionless parameter.
b Fractal
dimension, dimensionless parameter.
Note: Abbreviations: PC: polycrystalline; EG: electrochemically grown; EP: electrochemically polished; DMF:
N,N-dimethylformamide; i-PrOH: iso-propanol; TU: thiourea.
2.4 Electrical Double Layers. Double Layers at Single-crystal and Polycrystalline Electrodes
207
The inverse capacitance of metal phase, and this can be explained by the dif-
shown in Fig. 9, obtained according to the ferent surface states of the Bi faces.
Amokrane–Badiali model [15, 16, 19, 22], Emin for PC Bi lies between Eσ =0
depends on the crystallographic structure values for various planes with σP C ≤
as well as on the chemical nature of the 0.01 µC cm−2 .
metal studied. The admittance of Bi single-crystal
faces and PC Bi in acetonitrile and
2.4.2.8 Bismuth various alcohols including LiClO4 solu-
Bi crystallizes in a rhombohedral sys- tions has been studied, and Eσ =0 has
tem. Electrochemically polished Bi single- been found to depend on the nature of
crystal faces were first used for edl the solvent. Maximal Eσ =0 for vari-
studies by Frumkin, Palm and cowork- ous faces decreases in the order of sol-
ers [1, 20, 21]. The Emin (Table 4) was vents: acetonitrile > H2 O > methanol >
observed to be shifted 30 mV toward 2-propanol > ethanol [5, 15, 16].
more negative potentials compared with Ci values increase as the atomic density
Bi solid drop electrode and EP PC Bi. of the face decreases, except for Bi(111),
Later, electrochemically polished Bi(011̄), in good agreement with the theoreti-
Bi(21̄1̄), Bi(001), and Bi(101) faces have cal calculations by Leiva and Schmick-
been studied [5, 15, 16]. The difference ler, predicting the lowest interfacial ca-
between Eσ =0 for the above faces and pacitance for the most densely packed
the Bi(111) face is comparatively big planes [18]. The capacitance of a metal
1 0.03
0.03 2 1
3
0.02 0.02 2
5 6
4 3
[nm]
[nm]
lM
lM
0.01 7
0.01
8 0
−0.01
−0.02 −0.01
−10 −5 0 −10 −5 0
σ σ
[µC cm−2] [µC cm−2]
Fig. 9 The dependence of the ‘‘effective’’ thickness of the potential drop region in the metal surface
layer, calculated by the Amokrane–Badiali model [19] (at constant capacitance of solvent monolayer,
CS ), on the charge density for aqueous solution on different electrodes (a): 1, Sb(111); 2, Sb(001); 3,
Bi(111); 4, Hg; 5, Bi(001); 6, Cd (0001); 7, Cd(112̄0); and 8, Ga; and for Bi(001) in various solvents
(b): 1, methanol; 2, H2 O; 3, acetonitrile (updated from Ref. [15, 16]).
Tab. 4 Electrical double-layer parameters of Bi and Sb [5, 15, 16, 22] electrodes in various solvents
Electrode Solvent Eσ =0 [V] versus R D Atomic

aq. SHE density [cm−2 ]
Bi(111) KF/H2 O −0.434 ± 0.005 1.04 2.03 5.6 × 1015

Bi(1̄01̄) ’’ −0.34 ± 0.01 1.07 2.05 4.5 × 1015
Bi(001) ’’ −0.35 ± 0.01 1.06 2.05 5.3 × 1015
Bi(011̄) ’’ −0.35 ± 0.01 1.05 2.04 6.4 × 1015
Bi(21̄1̄) ’’ −0.33 ± 0.015 1.14 2.06 7.4 × 1015
Bi(111) LiClO4 /AN −0.25 ± 0.01 1.06 2.04 –
Bi(1̄01̄) ’’ −0.18 ± 0.01 1.01 2.04 –
Bi(001) ’’ −0.17 ± 0.01 0.98 2.03 –
Bi(011̄) ’’ −0.18 ± 0.01 1.08 2.06 –
Bi(21̄1̄) ’’ −0.17 ± 0.015 1.03 2.05 –
Bi(111) LiClO4 /MeOH −0.28 ± 0.01 1.05 – –
Bi(001) ’’ −0.21 ± 0.01 1.07 – –
Bi(011̄) ’’ −0.23 ± 0.01 1.10 – –
Bi(111) LiClO4 /EtOH −0.21 ± 0.01 1.03 2.04 –
Bi(001) ’’ −0.18 ± 0.01 1.08 2.07 –
Bi(011̄) ’’ −0.20 ± 0.01 1.07 – –
Bi(21̄1̄) ’’ −0.22 ± 0.01 1.12 – –
Bi(111) LiClO4 /i-PrOH −0.21 ± 0.01 1.02 – –
Bi(001) ’’ −0.170 ± 0.015 1.07 – –
Bi(011̄) ’’ −0.220 ± 0.015 1.03 – –
SbDER KF; NaF/H2 O −0.17 ± 0.01 1.10–1.14 – –
PC Sb NaF/H2 O −0.16 ± 0.02 1.03 2.05 –
SbDER HClO4 /MeOH −0.02 ± 0.025 1.08 – –
Sb(111) KF/H2 O −0.22 ± 0.01 1.06 2.04 –
Sb(001) ’’ −0.13 ± 0.01 1.04 2.04 –
Sb(011̄) ’’ −0.15 ± 0.01 1.18 2.08 –
Sb(21̄1̄) ’’ −0.10 ± 0.015 1.25 2.10 –
Sb(111) LiClO4 /EtOH −0.02 ± 0.01 1.01 – –
Sb(001) ’’ 0.05 ± 0.01 1.04 – –
Note: Abbreviations: AN, acetonitrile; MeOH, methanol; EtOH, ethanol; i–PrOH, isopropanol;
SbDER , an antimony drop electrode with remelted surface; PC, polycrystalline.
phase, CM , has been calculated according The influence of the surface pretreat-
to the Amokrane–Badiali model [19]. The ment of Bi single-crystal faces has been
thickness of a thin metal surface layer studied, and a noticeable dependence of
lM = 1/4πCM as a function of σ decreases Eσ =0 on the surface structure has been
in the order Bi(111) > PC Bi > Bi(101) > established [5, 15, 16, 22].
Bi(21̄1̄) > Bi(011̄) > Bi(001), in agree-
ment with other data. The value of lM max
2.4.2.9 Antimony
rises in the order AN < H2 O < MeOH < Sb crystallizes in the same rhombohedral
EtOH (Fig. 9b), which has been explained system as Bi. Sb(111) exhibits an appre-
by slight deviation of experimental systems ciably more negative value of Eσ =0 com-
from the Amokrane–Badiali model [19]. pared with the other planes (Table 4). The
difference Eσ =0 for Sb(001), Sb(011̄), as well as by the higher thickness of the

and Sb(21̄1̄) is not more than 0.05 V and thin metal layer [5, 15, 16, 22].
Eσ =0 decreases as the atomic density of
the surface increases. Eσ =0 for the Sb solid 2.4.2.10 Iron
drop electrode has an intermediate value, Fe crystallizes in the body-centered cubic
and σPC ≤ 0.01 C m−2 at Emin PC [5, 15, 16,
(bcc) system. Fe(100) and Fe(111) single-
22]. crystal faces were grown at 750 ÷ 780 ◦ C
The Ci values for Sb faces and PC Sb from FeBr2 in pure H2 atmosphere and
are noticeably lower than those for Bi. The reduced for 1 h at E = −0.95 V (SCE)
closest-packed faces show the lowest values in the working solution (pH = 2.5). A
of Ci (except Sb(111)), in good agreement diffuse-layer capacitance minimum was
with the Leiva–Schmickler model [18]. The observed with Emin practically indepen-
anomalous position of Sb(111) and Bi(111) dent of cel (Table 5). The PZ plot was linear
is probably caused by a more pronounced with R somewhat higher than unity. The
influence of the capacitance of the metal inner-layer capacitance decreases from
phase compared with other Sb and Bi faces Fe(111) to Fe(100) as the atomic density
(Fig. 9a). of the face increases [56].
The edl structure of Sb(111)–EtOH The PC Fe electrodes with electrochem-
and Sb(001)–EtOH interfaces has been ically polished (cathodically pretreated for
investigated by impedance. As for H2 O, 1 h) surfaces show a diffuse-layer mini-
Eσ =0 in EtOH for Sb(001) and Sb(111) mum at Emin = −0.94 V (SCE), indepen-
is higher than for Bi(111) and Bi(001). dent of cH2 SO4 [57]. A dependence of the
Ciσ =0 for Sb single-crystal faces in EtOH PZC on pH has been observed [5].
are lower than for Bi, which has been The edl at the renewed Fe–LiClO4
explained by the lower lyophilicity of Sb, interface has been studied in various
Tab. 5 Electrical double-layer parameters of Fe [5, 56–58] and Ni [5, 59] in various solvents
Electrode Solutiona Eσ =0 [V] versus aq. R Atomic density

SHE [cm−2 ]
PC Fe KF(0.01 M)/ −0.70 ± 0.02 – –

’’ NaClO4 /H2 O −0.64 – –
Fe(111) NaClO4 /H2 O −0.69 ± 0.01 1.43 0.707 × 1015
Fe(100) NaClO4 /H2 O −0.72 ± 0.01 1.33 1.222 × 1015
PC Fe LiClO4 /TMU −0.56 5.0 –
’’ LiClO4 /DMF −0.41 – –
’’ LiClO4 /DMAA −0.31 – –
’’ LiClO4 /N-MP −0.61 – –
’’ LiClO4 /HMPA −0.56 4.8 –
PC Ni H2 SO4 /H2 O (pH = 3.0) −0.24 – –
Ni(100) HClO4 /H2 O −0.39 ± 0.02 1.0 1.614 × 1015
Ni(110) HClO4 /H2 O −0.53 ± 0.02 1.0 1.141 × 1015
a TMU:tetramethylurea; DMF: dimethylformamide; DMAA: N,N,-dimethylacetamide; MPF:

N-methyl-N-2-pyridylformamide; HMPA: hexamethylphosphoramide.
nonaqueous aprotic solvents by Sa- by scraped method are in reasonable

fonov et al. [58]. The Emin in C, E agreement with impedance data [5, 13].
curves depends on time; for Fe|1,1,3,3-
tetramethylurea (TMU) + LiClO4 inter- 2.4.2.12 Pt-group Metals
face, Emin = −0.8 V (SCE in H2 O) just Pt and Pt-group metals (poly- and sin-
after surface renewal and −0.35 V after gle crystals) have long been among the
15 min. The PZ plot at Emin was lin- most intensively studied systems in elec-
ear, with R equal to 5.0 in TMU and 4.8 trochemistry; nevertheless, reliable Eσ =0
in hexamethylphosphoramide. The recon- values have been determined only re-
struction of the surface is probable during cently [1, 5, 8, 14, 60–67].
the experiment with a decrease in surface The analysis by Frumkin et al. [1, 14]
roughness with time [58]. in 1970 led to the conclusion that in
A less pronounced dependence of Ci the case of nonpolarizable electrodes, two
on E for an Fe electrode than that for values of zero-charge potential can be
Hg has been observed, and the values defined: Eσ =0 and EQ=0 (see Sect. 1).
of Ci for Fe are remarkably lower than The values of Eσ =0 obtained by the scrape
that for Hg. Thus, the interfacial potential method [13] are in reasonable agreement
drop is substantially higher for Fe than for with the Frumkin et al. data (Table 6).
Hg and, accordingly, strong chemisorption Results for other Pt-group metals are
of solvent molecules, weakly depending due to Frumkin and coworkers [1, 14]
on E, is probable at an Fe–electrolyte (Table 6). An Rh electrode with the surface
interface [58]. renewed in closed circuit has been used
by Lazarova [60]. In 0.005 M Na2 SO4 ,
2.4.2.11 Nickel Eσ =0 = −0.09 ± 0.02 V (SCE), while in
Ni crystallizes in the fcc system. A diffuse- 0.5 M Na2 SO4 , pH 2.5, Eσ =0 = −0.22 V
layer minimum has been observed for has been reported with ∂Eσ =0 /∂pH =
the Ni(100) and Ni(110) single-crystal 55 mV.
faces in H2 O/Na2 SO4 system, with Emin It is to be stressed that in the case of
independent of ν and cel , and R ≈ 1 PC Pt-group metals, the crystallographic
(Table 5). In the case of Ni(111) and structure of the surface probably exerts
PC Ni the capacitance also decreases a very pronounced influence, so that
with dilution, but no deep minimum the experimental PZC values do not
was observed in the C, E curves. Very correspond to the condition σ̄PC = 0.
low Ci values have been obtained. The The surface electrochemistry of Pt
difference between sp and sd metals has single-crystal electrodes has been exhaus-
been explained by a different strength of tively studied using cyclic voltammetry [5,
the interaction between the metal and 8–12, 61–67]. Phenomena of a step re-
solvent molecules [59]. construction and step coalescence have
Eσ =0 of PC Ni is located in the been observed [61]. For Pt(111)–H2 O in-
negative potential range and depends terface, prepared by the flame annealing
strongly on the solution’s pH [1, 5]. Widely method, a double-layer charging has been
scattered Emin values have been reported observed only in a very narrow potential
by different authors at similar solution pH region (0.1 < E < 0.35 V (SCE) in 0.05 M
(Table 5). However, the values obtained H2 SO4 ), which depends on the chemical
212
Tab. 6 Potentials of zero charge of Pt-group metal single-crystal faces in aqueous solutions
Electrode Electrolyte Eσ =0 [V] local [V]

EQ=0 Adsorption state Method Atomic density References
versus SHE versus SHE (Terrace) (cm−2 )
Pt(111) 0.1 M HClO4 – 0.34 (111) Cyclic voltammetry 1.503 × 1015 68

0.1 M HClO4 – 0.29 – ’’ – 63
0.1 M HClO4 – 0.27 – ’’ – 66
Pt(110) 0.1 M HClO4 – 0.16 (110)-(1 × 2)(a) ’’ 0.920 × 1015 68

0.15 M HClO4 – 0.18 (110) ’’ – 66
Pt(100) 0.1 M HClO4 – 0.25 (100)-1D ’’ 1.302 × 1015 68
Rh(111) 0.1 M HClO4 – 0.16–0.17 (111) ’’ 1.599 × 1015 ’’
Ir(111) 0.1 M HClO4 – 0.13 (111) ’’ 1.574 × 1015 ’’
PC Pt 0.3 M HF + 0.12 M KF (pH 2.4) 0.185 0.235 – Titration – 1, 14
’’ NaClO4 (pH = 1.0) 0.24 – – Scrape – 1, 13
’’ NaClO4 (pH = 7.0) 0.02 – – ’’ – ’’
’’ NaClO4 (pH = 11.0) −0.30 – – ’’ – ’’
PC Pd 0.05 M Na2 SO4 + 0.001 M H2 SO4 (pH 3) 0.10 0.26 – Titration – 1, 14
’’ NaClO4 (pH 7.0) 0.00 – – Scrape – 5, 13
PC Rh 0.3 M HF + 0.12 M KF (pH 2.4) −0.005 0.085 – Titration – 1, 14
’’ NaClO4 (pH 1.0) 0.17 – – Scrape – 5, 13
’’ NaClO4 (pH 7.0) −0.02 – – ’’ – ’’
PC Ir 0.3 M HF + 0.12 M KF (pH 2.4) −0.01 – – Titration – 1, 14
’’ NaClO4 (pH 1.0) 0.21 – – Scrape – 5, 13
’’ NaClO4 (pH 7.0) 0.52 – – ’’ – ’’
PC Pt LiClO4 AN 0.14 – – Impedance – 69, 70
PC Pt LiClO4 /DMSO 0.09 – – ’’ – ’’
PC Pd LiClO4 /AN 0.32 – – ’’ – ’’
PC Pd LiClO4 /DMSO 0.24 – – ’’ – ’’
Note: Abbreviations: AN: acetonitrile; DMSO: dimethylsulfoxide.

nature of the anion [5, 41, 61] (see also et al. have studied the stepped Pt(111) sur-
Chaps. 14–17 and 19 in Vol. 3). faces [66]. The dependence of the PZTC
For Pt(111) in the aqueous solutions of for Pt(111) surfaces in acidic aqueous so-
hydrofluoric acid, Wagner and Moylan [62] lution having increasing densities of the
have estimated Eσ =0 = 0.22 V (versus re- ordered monoatomic steps in the (111) to
versible hydrogen electrode) by comparing (110) and (111) to (100) zones is evaluated
voltammetric curves and high-resolution from CO ‘‘charge-displacement’’ measure-
electron energy loss spectroscopic data of ments, with the objective of elucidating the
water + H+ coadsorption from the gas influence of the electrochemical double
phase. Iwasita and Xia [63] prepared Pt layer on the large step-induced changes in
single crystals according to the Clav- surface potential known for the clean un-
ilier et al. method [41]. According to the charged surface in UHV [67]. The PZTC
data of Fourier transform infrared (FTIR) values in 0.1 M HClO4 and 0.5 M H2 SO4
reflection adsorption spectra, the water ori- solutions decrease noticeably upon in-
entation changes from hydrogen down to creasing the (110) step density, whereas
oxygen down at 0.29 V (SHE), and this smaller effects are found for the (100)
has been taken to indicate that EQ=0 for steps. The location of the EQ=0 values
Pt(111) is close to this value. within the so-called hydrogen region, how-
Hamm et al. [64] treated a Pt(111) sur- ever, complicates the interpretation of the
face in a UHV chamber by sputtering and EQ=0 – step density dependences owing
annealing until the surface was clean and to the presence of faradaic charge asso-
well ordered. A very narrow potential re- ciated with potential-dependent hydrogen
gion, 0.13 < E < 0.28 V (SCE), is left to adsorption [66].
bare double-layer charging with a capaci- A detailed study of N2 O reduction on
tance C = 110 ± 13 µF cm−2 . Thus, only a variety of single-crystal Pt-group metal
in this region, a Pt(111) surface is free electrodes shows that N2 O reduction cur-
from surface adlayers. Thereafter the im- rent maxima occurred exactly at Clavilier’s
mersion method (IM) was used, and the Q, EQ=0 . Therefore the N2 O reduction data
E-plot has been found to go through zero have been used for obtaining the EQ=0
at E ≥ 0.80 V (SHE). Assuming a small values (Table 5). The importance of a lo-
dependence of C on E, a somewhat more cal value of EQ=0 has been emphasized,
positive Eσ =0 ∼ 1.1 V (SHE) has been esti- especially with respect to reconstructing
mated. However, it is to be noted that these metal surfaces such as Pt(100) and Pt(110),
values of Eσ =0 correspond to the potential which can be prepared in a variety of crys-
region, where the Pt surface is oxidized [5]. tallographic states [5, 65–68].
Clavilier et al. [65] have studied CO Pt and Pd electrodes with a renewed
adsorption on electrochemically faceted surface obtained by cutting with a ruby
Pt(111) and Pt(110) electrodes, and from knife under the working nonaqueous
the charge transients, these authors have solution have been studied by Petrii
provided a ‘‘definite’’ determination of et al. [69, 70] and by other investigators
EQ=0 . However, electrochemically faceted (Table 6). A more detailed discussion of
Pt(111) electrodes have a PC surface struc- data obtained is presented in Ref. [5].
ture, and thus the value of EQ=0 for
(111) Electrical double layer at solid metal
such electrodes lies between EQ=0 for alloy–electrolyte interface have been stud-
(110)
terraces and EQ=0 for steps [5]. Gomez ied, but one of the features of alloys is that
their bulk and surface compositions are, Zn and Cd both crystallize in the hcp
as a rule, substantially different because system. For both metals, the basal (0001)
one of the components is more surface- face is the most dense, and as for fcc
active than the other. It is to be noted metals, Eσ =0 is more positive than for the
that the surface composition of solid alloys other faces [15, 16, 22, 48] (Fig. 11). For
can change with time as well as through Sn planes the variation of Eσ =0 with the
surface process (selective oxidation or dis- crystal face is practically negligible [49, 50].
solution, surface migration and diffusion Being semimetals, Bi and Sb show
of components). Therefore, for the more anomalies in the correlation of Eσ =0
detailed discussion the future experimen- with the surface atomic density [15, 16,
tal studies are inevitable; however, a short 22], explained in terms of face-specific
review of experimental data is given in space charge effects. However, the definite
Ref. [5]. dependence of Eσ =0 on the reticular
density of plane has been established in
2.4.3 a good agreement with general tendency.
General Correlations The data discussed above indicate that
the difference in Eσ =0 between the most
2.4.3.1 Zero-charge Potential and dense and the most open surface depends
Reticular Density of Planes on the ‘‘softness’’ of the metal surface,
Figure 10 shows the graphical dependence which to a first approximation can be
of the PZC (or PZTC for Pt planes) on the measured by the melting point of the
crystallographic orientation of the surface given metal [2–5]. Eσ =0 is highest for
for fcc metals. The plots exhibit a typical Au (Tf = 1063 ◦ C) and lowest for Sn
pattern with minima and maxima, which (Tf = 231.9 ◦ C).
fall at the same angle for all fcc metals,
and Eσ =0 values are linearly correlated 2.4.3.2 Potential of Zero Charge, Work
with the coordination number of surface Function, Interfacial Parameter, and
metal atoms [2–5]. Inner-layer Capacitance
Eσ =0 values are correlated with the According to the basic concepts of elec-
density of atoms on the given surface trochemistry (see also Chap. 1 in Vol. 1),
and are given in Fig. 11. The PZC is the electrode potential E measured with
more positive for dense surfaces and respect to the constant reference electrode
more negative for open surfaces. Pb can be expressed as
also crystallizes in the fcc system and
therefore the same dependence of Eσ =0 E(vs. ref.) = + δχM + δχS + χion
on the crystallographic orientation should + const (19)
be expected, but Eσ =0 varies exactly in the
(S)
other direction [48]. A possible explanation where δχM = χM − χM is the perturba-
can be sought in the surface mobility of Pb tion of the electron distribution at the
atoms at room temperature, which may surface of the metal (χM is the surface
lead to extensive surface reconstruction potential of the bare metal surface in vac-
(M)
phenomena [5]. The data for Ni(001), for uum), δχS = χS − χS is due to the effect
Ni(110) [59], and for Pt planes (EQ=0 ) [68] of any solvent dipolar layer preferentially
are in good correlation with those for Au, oriented at the phase boundary (χS is the
Ag, and Cu planes. surface potential at the free surface of the
0.8
0.4
Au
0 Ag
[V vs SHE]
Eσ =0; EQ =0
Cu
Pt
−0.4
Ni
−0.8
(110) (331) (111) (311) (100) (210) (110)

−1.2
−20 −10 0 10 20 30 40 50 60 70
Crystallographic orientation
Fig. 10 Dependence of the pzc, Eσ =0 , on the crystallographic orientation for
the metals (Cu, Ag, Au, and Ni), which crystallize in the fcc system. For Pt
electrodes, EQ=0 values have been used. (Updated from Ref. [5].)
Au
0.5
Pt
Eσ =0; EQ =0
Rh
[V (SHE)]
0 Ir
Sb
Bi
Ni Cu
−0.5
Ag
Cd Zn
−1
0 0.5 1 1.5 2
Atomic density∗10−15
[atoms cm−2]
Fig. 11 The dependence of zero-charge potential (PZTC values for Pt metals) on the atomic density
for various single-crystal electrodes (metals noted in figure) (updated from Ref. [5]).
solvent), and χion accounts for any possi- quantities are as a rule obtained in different
ble charge separation across the interface. laboratories with different techniques.
At the PZC in surface-inactive electrolyte More than with Eσ =0 , the problem is with
solutions, the ionic specific adsorption the selection of values for . The values
is absent ( χion = 0) and Eq. (19) mod- of Eσ =0 have been obtained more recently
ifies to and are usually known with accuracy to
±0.01 V or better, the measurements of
Eσ =0 (vs. ref.) = + δχM + δχS + const are rather occasional and have an
= + χ + const (20) accuracy from ±0.05 to ±0.25 eV [4, 5].
Thus, experimental data do not ensure
Since the two perturbation terms (δχM an appropriate accuracy for X and
and δχS ) are specific to the given interface the uncertainty may outweigh the value
and are experimentally inseparable, they itself. The best way to proceed is to
have grouped into a single quantity X plot Eσ =0 versus for many electrodes
and, according to Trasatti, is called the and to derive information about X
interfacial parameter [1, 4, 5, 71, 72]. term from graphical correlations using
Thus, the difference in Eσ =0 for different statistical analysis.
metals can be written as Eσ =0 = + Figure 12 shows a dependence of Eσ =0
X if the same reference electrode on . For PC sp metals, the average
is used. work function values, defined as M =
M
The main problem in Eσ =0 versus j θj j , have been used [5, 73]. Hg is
correlation is that the two experimental taken as a reference surface and a straight
Ir(111) Pt(111)
Pd(111)
5.5
(111)
(100)
(100)
(110)
(110) Ni
Au
5
[eV]
Φ
(100)
(111)
(110) (100) Cu
Fe Sb
(110)
4.5 Ag Sn
Zn (111) Hg
(100) Bi
(110)
Cd Pb
Tl In Ga
4
−1 −0.5 0 0.5
Eσ = 0; EQ = 0
[V (SHE)]
Fig. 12 The dependence of the electron work function (into UHV) on the zero-charge potential for
various PC (filled marks) and monocrystalline (open marks) electrodes (metals and faces noted in
the figure). ( ) Straight line of the unit slope through the point for Hg. For Pt-metals potentials
of zero total charge are shown.
line of unit slope is drawn through its principle, those for singular planes, specif-
point. The data show that all metals lie ically Au(111), for which ‘‘congruent’’ pairs
on the left of the straight line, thus all of data exist for and Eσ =0 [4, 5].
of them have an X term more negative Eσ =0 values for PC d-metals are not very
than that of Hg and there are no metals reliable, except Fe and Ni, and therefore
on whose surface the contact with water the points for PC Fe and Ni only as well
produces a lower than on Hg. The as for single-crystal planes are included
horizontal distance (along the Eσ =0 axis) in Fig. 12 as broadly representative of d-
of each metal point from the line of unit metals. X values for these electrodes are
slope measures X with respect to Hg. higher than those for sp and sd metals.
Thus, X measures the relative value Pt metals have never found a def-
of the contact potential difference at the inite position in Eσ =0 , correlation.
metal–water interface [4, 5]. Recently, with the improvement achieved
All sp metals, except Zn, can be in the preparation and control of sur-
apparently gathered in a single group faces, the PZTC has been estimated
and there is a clear trend within the indirectly (0.22 < EQ=0 < 0.35 V (SHE))
group for Eσ =0 to become more negative close to the PZC of PC Pt. In view of
as decreases. Thus, X increases the heterogeneity of Pt surfaces, this close-
ness may be puzzling [5]. However, if the
as decreases. Zn has much higher
PZTC estimated around 0.30 V (SHE) is
value of X than the other sp metals
taken for Pt(111) as EQ=0 , the point of Pt
with comparable work function. Since
would be located further from the line for
X ≡ δχ M + δχ S , the major contribution
d-metals, with a high value of X that is
is probably in δχ S .
not justified by the known behavior of the
In view of the large heterogeneity effect
Pt surface [5].
on the value of Eσ =0 for sd metals, the
The value of Eσ =0 for Pt(111), obtained
data for single-crystal planes are shown in
directly by the IM [64], is much more
Fig. 12. It is remarkable that each metal positive than others reported [62, 63, 68]. It
forms a separate group in which the is in the right direction with respect to a PC
faces are aligned along apparently parallel surface, even though it is an extrapolated
straight lines with the same slope as for sp value and probably corresponds to the
metals. X varies in the sequence Au < oxidized Pt(111) surface [5]. If this value
Ag < Cu and for the same metal with the is taken as PZTC and 6.1 eV assumed to
crystal face (111) < (100) < (110). These be for Pt(111), then a linear dependence
relations are in good agreement with the seems to be valid for Hg, Au(111), and
work function drop values, caused by the Pt(111), and X increases in the order
adsorption of H2 O molecules, obtained Hg < Au(111) < Pt(111). The value of
from the UHV data [4, 5]. Eσ =0 given by Frumkin et al. [1, 14] would
It is to be noted that in many cases, the imply = 5.4 eV for the PC Pt, which
preparation of a single-crystal face is not is within the usual range of experimental
followed by a check of its perfection by data. However, there remains the puzzling
means of appropriate spectroscopic tech- aspect that while EQ=0 refers to a
niques, and therefore we actually have surface with hydrogen adsorbed on it,
‘‘nominal’’ single-crystal faces. The only values do not include contributions from
values of X of high reliability are, in adsorbed H [5].
Data for single-crystal planes of sp and of a solvent as the two phases are
metals have not been plotted in Fig. 12 brought in contact (see also Chap. 14 in
as there are no correct values in the Vol. 3). The reorientation of a solvent
literature. For that reason the so-called molecule is possible if there are orienting
electrochemical work function values ∗ forces acting on it, which in the absence
have been obtained according to Frumkin of an electric field (σ = 0) can only be
et al. and Trasatti’s conception discussed the short-range chemical interactions. So,
in Refs. [1, 71, 72]. According to the data in X can also be interpreted in terms of
Fig. 13, there is a good linear correlation metal–water interaction, and the concept
between ∗ and Eσ =0 values for single- of ‘‘hydrophilicity’’ has been introduced [1,
crystal planes of sp metals, with the same 4, 5, 71, 72]. However, a quantitative com-
slope as for PC sp metals. The dependence parison (of X with ) with reliable
of X ∗ on the crystallographic orientation recent data is possible only for Ag(110),
of plane is weak, but X ∗ seems to Cu(110), and Pt(111), and the gas phase
increase with the decrease of the reticular data confirm the sequence observed elec-
density of planes, thus in the same trochemically. Thus, varies in the
direction as for fcc sd metals. order Au < Ag < Cu ( for PC Au is
X and measure change in the compared with X for the (110) face);
structure of the surface layers of a metal also, decreases Cu(110) > Pt(111) >
1
4.6 32
4
6
9
11 7 5
12
14
15 8
10
[eV]
30 13
Φ∗
31 19 17
4.2 20 16
32 23
26 24 18
33 21
28 22
27 25
29
3.8
−1.0 −0.7 −0.4 −0.1
Eσ =0
[V (SHE)]
Fig. 13 Dependence of the electrochemical work function ∗ on the pzc Eσ =0 for
various single-crystal and PC electrodes: 1, Sb(21̄1̄); 2, Sb(001); 3, Sb(011̄); 4, PC-Sb; 5,
Hg; 6, Sb(111); 7, Bi(21̄1̄); 8, Bi(1̄01̄); 9, Bi(011̄) and Sn(001); 10, Bi(001); 11, Sn(110);
12, PC-Bi; 13, Sn(11̄0); 14, PC-Sn; 15, Bi(111); 16, Pb(110) and Pb(112); 17, Pb(100); 18,
PC-Pb; 19, Pb(111); 20, In(Ga); 21, Ga; 22, Tl(Ga); 23, Cd(0001); 24, PC-Cd; 25, Cd(101̄0);
26, Cd(112̄0) and Zn(0001); 27, PC-Zn; 28, Zn(101̄0); 29, Zn(112̄0); 30, Zn(0001); 31,
PC-Zn, 32, Zn(101̄0); 33, Zn(112̄0). The values of ∗ 1–29 have been calculated
according to the equation Eσ =0 (versus SHE, V) = 1.27∗ (eV) − 5.91; and 30–33
according to equation Eσ =0 (versus SHE, V) = ∗ (eV)–5.01 (updated from Ref. [71]).
Ag(111) [4, 5]. According to the data of electropolished PC and single-crystal sp

thermal desorption spectroscopy (TDS), and sd metal electrodes in H2 O medium
the Ag–H2 O interactions are weaker than (Fig. 14). The decrease of li with the in-
H2 O–H2 O interactions, and the activity crease of X appears to be a general
of Cu(110) is higher than that of (111) occurrence [5, 71], and for electropolished
and (100) faces. The M–H2 O interactions Ag electrodes, li decreases in the sequence
are stronger on Cu than on Ag. A sur- (111) > (100) > (110) as the hydrophilic-
face of Pt(111) is more reactive than that ity of surface decreases. The point of
of sd metals, but the high-temperature Au(100) is also located near the general
TDS peak for Pt(111) is very close to correlation [5, 37].
that of Cu(110). More complex spectra A linear correlation between li and
are exhibited by Ni(110), which is known electrochemical X ∗ values for sp single-
to be easily oxidized and thus tends to crystal planes (Fig. 14), in correlation to
react strongly with water molecules [5, the general behavior for PC sp metals and
74]. electropolished Ag and Au planes, can be
A high value of X entails an apprecia- observed. For Bi(111) and Sb(111) there
ble modification of the surface structure are small deviations toward higher values
as a consequence of strong interactions of li , caused by the different semimetallic
between the phases and vice versa, and properties of these planes [5, 71].
thus the inner-layer capacitances have to On the other hand, the same values
correlate with X (or ) values. A of Ci cannot fit Fig. 14 if the values of
good correlation between the experimen- X estimated by Valette are used [26].
tal inner-layer thickness li = 4πCi and The same is the case for the values
X has been demonstrated for various of Ci reported by Popov et al. [38] for
0.04
(111)
Sb Bi (111)
(211) (211) (011)
(001)
(011) (001) (112)
0.03
Hg (110) Pb
(110) Cd
(001) (0001)
(110) Sn (100)
[nm]
(1010)
0.02
li
Au
(111) (1120)
(111) (001)
In(Ga)
0.01 Ag (110) Zn
(0001)
Ga
(1010)
(1120)
0
0 0.1 0.2 0.3
∆X; ∆X ∗
[V]
Fig. 14 Dependence of the inner-layer thickness on the interfacial parameter X for various
polycrystalline sp metals (filled marks) and single-crystal sp and sd metal electrodes, noted in
figure (updated from Refs. [5, 71]). For single-crystal sp metals, X ∗ values have been used.
single-crystal faces grown in a Teflon charged surface claimed by Toney et al.

capillary. For electropolished Cd, higher on the basis of surface X-ray scattering
capacitances for the faces with more experiments, which proposed that H2 O
negative PZC have been obtained in good molecules are oriented with H toward
agreement with the data for EP Ag [4, a negatively charged surface [5, 77, 78].
5, 22]. For the Cd electrodes grown in This picture is not confirmed by far-
a Teflon capillary, a reverse order of IR spectroscopy results [79] according to
capacitance values has been reported [54, which, although H2 O molecules change
55]. This indicates that differently prepared the orientation with charge, they always
Ag and Cd single crystals behave not point O atom toward the surface. The
only quantitatively but also qualitatively importance of the presence of other
differently [4, 5]. molecules for the interaction of a water
molecule with a metal surface is seen
2.4.3.3 Data of Quantum-chemical clearly in calculations. The adsorption
Calculations energy calculated for H2 O on Hg is
Quantum-chemical calculations, molecu- −32.2 kJ mol−1 , which is less than that
lar dynamics (MD) simulations, and other for hydrogen bonding, and therefore Hg
model approaches have been used to de- behaves as a hydrophobic surface [80].
scribe the state of water on the surface of According to quantum-chemical calcula-
metals. The decrease in at the surface of tions, the metal–H2 O interaction has been
Ag(110) has been successfully reproduced proposed to be in the sequence Hg <
by a jellium/point dipole model by assum-
Ag(100) < In < Cu(100) [81]. Compared
ing a disordered water structure at the
to the experimental data, it appears
interface [75]. In the case of hydrophilic
that the positions of In and Ag(100)
surfaces, the first layer (or two) of water is
are exchanged. The complete neglect
strongly oriented by the forces emanating
of differential overlap method predicts
from the surface, but the bulk structure is
for any given metal a weaker inter-
soon recovered in a few layers after some
action on the more dense surface [40].
very disorganized layers (see also Chap. 2.3
in Vol. 1). Thus, the predicted sequence is (111) <
MD simulations have been used for (100) < (110) for fcc metals and (0001) <
H2 O at Pt(100) and Pt(111), as well (1100) for hcp metals. However, for the
as at Ag(111). The structure of water most compact surfaces, the calculated
is predicted to conform to a hexagonal sequence is Hg < Ag(111) < Cu(111) ≈
pattern and the metal–H2 O interaction is Zn(0001) < Au(111) < Cd(0001). It is dif-
probably stronger for the (111) than for ficult to accept that Zn can be less hy-
the (100) surface. However, on the basis drophilic than Au or Cd, and also that Au
of the extended Hückel theory, different can be more reactive than Cu. More re-
conclusions in favor of the (100) face have cent calculations gave the other order of
been made [5, 76, 77] (see also Chap. 2.3 metals: Ag < Au < Cu. This confirms the
in Vol. 1). position of Cu, but Au still appears to be
Water reorientation is usually predicted more reactive than Ag [5].
by theoretical models, but for Ag(111), MD Thus, the data obtained are controver-
simulations do not confirm the noticeable sial on the whole, theoretical calculations
increase in water population near a provide only a general insight into the
problem of water–metal interactions prob- const. Under similar circumstances, G0A

ably because not all factors are appropri- is only a function of GM−S , hence it is
ately taken into account [5]. expected to be correlated with X, as well
as Ci values.
2.4.3.4 Adsorption of Organic Compounds Data in Fig. 15 show a nice linear corre-
and Hydrophilicity of Electrode lation for various normal alcohols, with
Adsorption at electrodes is considered to ( G0A ), decreasing as X increases
be a solvent replacement reaction: B(sol) + ( ( G0A ) is the difference between the
nH2 O(ad) ↔ B(ad) + nH2 O(sol), where B Gibbs energy of adsorption of an organic
is an adsorbing substance replacing n wa- compound at a metal–electrolyte interface,
ter molecules on the electrode surface [1, G0A , and the increase of Gibbs energy
4, 5, 71, 72]. Adsorption will affect the of organic compound adsorption, caused
PZC since water dipoles are replaced by by the addition of electrolyte into the so-
adsorbate dipoles. On the other hand, the lution, G0sol ( ( G0A ) = G0A − G0sol ).
Gibbs energy of adsorption for a given Thus the more hydrophilic metals (or
adsorbate B can be divided into several con- faces) adsorb less. However, there are two
tributions: G0A = GM−B –GB−S –GM−S , aspects: (1) The slope of the correlation
where S stands for solvent and G is the depends marginally on the nature of the
bond strength. If the same adsorbate is adsorbate, that is, it is a property of the
studied on different metals in the same sol- adsorbed layer structure. (2) The correla-
vent, then GB−S ≈ const. Furthermore, if tion is valid for both PC and single-crystal
only physical adsorption occurs, GM−B ≈ electrodes, which suggests that common
0
Au(111)
Ag(111)
Ag(100)
2 Hg
−10
Hg
[kJ mol−1]
∆(∆G 0A)
3 Sb
4 Bi
In(Ga)
5 Sn
−20
Pb
6 Ga
Cd
Zn(0001)
−30
−0.1 0.1 0.3 0.5
∆X
[V]
Fig. 15 Dependences of the difference ( G0A ) for various alcohols on the interfacial
parameter X for various metals. The number of each plot is equal to the number of carbon
atoms of given aliphatic compound (updated from Refs. [5, 71]). ( G0A ) = G0A − G0sol
where G0A is the Gibbs energy of adsorption of an organic compound at a metal–electrolyte
interface and G0sol is the increase of Gibbs energy of organic compound adsorption, caused by
the addition of electrolyte into the solution.
−10
22
21
−14
20 16
[kJ mol−1]
∆(∆G 0A)
19
18
17
12 10 2
15 9 4
−18 7 1
14 13 5
11 8 6
3
−22
0 0.01 0.02 0.03 0.04
li
[nm]
Fig. 16 Dependence of the difference ( G0A ) for cyclohexanole adsorption on the thickness of
inner layer at σ = 0 in the solution of surface-inactive electrolyte for various electrodes: 1,
Sb(111); 2, Bi(111); 3, Sb(21̄1̄); 4, PC-Bi; 5, Bi(1̄01̄); 6, Sb(001); 7, Bi(21̄1̄); 8, Bi(011̄); 9, Bi(001);
10, Pb(001); 11, Pb(110); 12, Hg; 13, PC-Pb; 14, Sn(001); 15, Pb(111); 16, PC-Cd; 17, Cd(0001);
18, Cd(101̄0); 19, Cd(112̄0); 20, Zn(0001); 21, Zn(101̄0); 22, Zn(112̄0) (updated from Ref. [71]).
factors are behind this phenomenon. X deviation from the general plot, which is
and water–metal interaction strength are mainly caused by the more pronounced in-
parallel, and these data confirm the hy- fluence of the capacitance of metal phase of
drophilicity sequence Hg < Au < Ag and these electrodes in comparison with other
the crystal-face sequence for fcc metals Bi and Sb single-crystal planes.
(111) < (100) < (110). Therefore the hy-
drophilicity scales of Popov et al., Silva References
et al., and Valette cannot be sustained on
the basis of sound experimental and theo- 1. A. N. Frumkin, Zero Charge Potentials,
retical arguments [5]. Nauka, Moscow, 1979.
This order of metals is in good agree- 2. A. Hamelin in Modern Aspects of Electro-
chemistry (Eds.: B. E. Conway, R. E. White,
ment with the data in Fig. 16, where the
J. O’M. Bockris), Plenum Press, New York,
dependence of ( G0A ) on the inner-layer 1985, pp. 1–102, Vol. 16.
thickness is presented. To the first approx- 3. A. Hamelin, T. Vitanov, E. Sevastyanov et al.,
imation, this plot can be considered linear J. Electroanal. Chem. 1983, 145, 225–264.
and the adsorption activity of organic com- 4. S. Trasatti, L. M. Doubova, J. Chem. Soc.,
pound at the electrode–solution interface Faraday Trans. 1995, 91, 3311–3325.
5. S. Trasatti, E. Lust in Modern Aspects of Elec-
decreases as the thickness of double layer
trochemistry (Eds.: R. E. White, B. E. Conway,
decreases, thus as the hydrophilicity of J. O’M. Bockris), Kluwer Academic/Plenum
electrode increases [71]. Data for Bi(111) Publishers, New York, London, 1999,
and Sb(111) planes show a noticeable pp. 1–216, Vol. 33.
6. D. M. Kolb in Structure of Electrified Interfaces 24. R. Parsons, F. G. R. Zobel, J. Electroanal.

(Eds.: J. Lipkowski, P. N. Ross), VCH, New Chem. 1965, 9, 333–348.
York, 1993, pp. 65–102. 25. G. Valette, A. Hamelin, J. Electroanal. Chem.
7. M. A. Vorotyntsev in Modern Aspects of Elec- 1973, 45, 301–319.
trochemistry (Eds.: J. O’M. Bockris, B. E. Con- 26. G. Valette, J. Electroanal. Chem. 1989, 269,
way, R. E. White), Plenum Press, New York, 191–203.
1986, pp. 131–222, Vol. 17. 27. I. A. Bagotskaya, M. D. Levi, B. B. Dama-
8. R. Adzic in Modern Aspects of Electro- skin, J. Electroanal. Chem. 1980, 115,
chemistry (Eds.: R. E. White, J. O’M. Bockris, 189–209.
B. E. Conway), Plenum Press, New York, 28. D. Leikis, K. Rybalka, E. Sevastyanov et al., J.
1990, pp. 163–236, Vol. 21. Electroanal. Chem. 1973, 46, 161–167.
9. J. O’M. Bockris, S. U. M. Khan, Surface Elec- 29. A. Hamelin, L. Stoicoviciu, J. Electroanal.
trochemistry. A Molecular Level Approach, Chem. 1989, 271, 15–26.
Plenum Press, New York, 1993, pp. 59–210. 30. M. A. Vorotyntsev, J. Electroanal. Chem. 1981,
10. A. Hamelin in Nanoscale Probes of the 123, 379–387.
Solid/Liquid Interface (Eds.: A. A. Gewirth, 31. L. I. Daikhin, A. A. Kornyshev, M. Urbakh,
H. Siegenthaler), Kluwer, Dordrecht, 1995, Phys. Rev. E 1996, 53, 6192–6199.
pp. 285–306. 32. E. Lust, A. Jänes, V. Sammelselg et al., Elec-
11. G. A. Somorjai, Chemistry in Two Dimensions: trochim. Acta 1998, 43, 373–383.
Surfaces, Cornell University Press, Ithaca, 33. E. Lust, A. Jänes, V. Sammelselg et al., Elec-
1991. trochim. Acta 2000, 46, 185–191.
12. R. Parsons in Comprehensive Treatise of Elec- 34. L. I. Daikhin, A. A. Kornyshev, M. Urbakh,
trochemistry (Eds.: J. O’M. Bockris, B. E. Con- J. Chem. Phys. 1998, 108, 1715–1723.
35. L. Nyikos, T. Pajkossy, Electrochim. Acta
way, E. Yeager), Plenum Press, New York,
1985, 30, 1533–1539.
1980, pp. 1–44, Vol. 1.
36. T. Pajkossy, J. Electroanal. Chem. 1994, 364,
13. R. S. Perkins, T. N. Andersen in Modern As-
114–125.
pects of Electrochemistry (Eds.: J. O’.M. Bock-
37. B. B. Mandelbrot, The Fractal Geometry of
ris, B. E. Conway), Plenum Press, New York,
Nature, Freeman, San Francisco, 1982.
1969, pp. 203–282, Vol. 5.
38. T. Vitanov, A. Popov, E. S. Sevastyanov, J.
14. A. Frumkin, O. Petrii, B. Damaskin, J. Elec-
Electroanal. Chem. 1982, 142, 289–297.
troanal. Chem. 1970, 27, 81–100. 39. M. Bacchetta, S. Trasatti, L. Doubova et al., J.
15. E. Lust, K. Lust, A. A.-J. Jänes, Russ. J. Elec- Electroanal. Chem. 1986, 200, 389–396.
trochem. 1995, 31, 807–821. 40. An. M. Kuznetsov, R. R. Nazmutdinov, M. S.
16. E. Lust, A. Jänes, K. Lust et al., Electrochim. Shapnik, Electrochim. Acta 1989, 34,
Acta 1997, 42, 771–783. 1821–1828.
17. M. A. Vorotyntsev, A. A. Kornyshev, Elek- 41. J. Clavilier, R. Faure, G. Gainet et al., J.
trokhimiya 1984, 20, 3–47. Electroanal. Chem. 1980, 107, 205–209.
18. E. Leiva, W. Schmickler, J. Electroanal. Chem. 42. X. Gao, G. J. Edens, A. Hamelin et al., Surf.
1987, 143, 73–88. Sci. 1993, 296, 333–351.
19. S. Amokrane, J. P. Badiali in Modern Aspects 43. Z. Borkowska, G. Jarzabek, J. Electroanal.
of Electrochemistry (Eds.: R. E. White, J. O’M. Chem. 1993, 353, 1–17.
Bockris, B. E. Conway), Plenum Press, New 44. J. Lecoeur, J. P. Bellier, Electrochim. Acta
York, 1991, pp. 1–95, Vol. 22. 1985, 30, 1027–1033.
20. B. B. Damaskin, U. V. Palm in Itogi Nauki i 45. H. Hennig, V. V. Batrakov, Elektrokhimiya
Tekhniki. Elektrokhimiya (Eds.: Yu. M. Polu- 1979, 15, 1833–1837.
karov), VINITI, Moscow, 1977, pp. 99–143, 46. M. L. Foresti, G. Pezzatini, M. Innocenti, J.
Vol. 12. Electroanal. Chem. 1997, 434, 191–200.
21. A. N. Frumkin, V. V. Batrakov, A. I. Sidnin, 47. J. R. LaGraff, A. A. Gewirth in Nanoscale
J. Electroanal. Chem. 1972, 39, 225–228. Probes of the Solid/Liquid Interface (Eds.:
22. E. Lust, A. Jänes, K. Lust et al., Electrochim. A. A. Gewirth, H. Siegenthaler), Kluwer,
Acta 1997, 42, 2861–2878. Dordrecht, 1995, pp. 83–101.
23. S. Trasatti, O. A. Petrii, Pure Appl. Chem. 48. L. P. Khmelevaya, A. V. Chizhov, B. B. Dama-
1991, 63, 711–734. skin, Elektrokhimiya 1978, 14, 1304–1305.
49. L. P. Khmelevaya, B. B. Damaskin, Elektro- 66. R. Gómez, V. Climent, J. M. Feliu et al., J.

khimiya 1981, 17, 1721–1725. Phys. Chem. B 2000, 104, 597–605.
50. L. P. Khmelevaya, Thesis, Moscow State Uni- 67. M. J. Weaver, Langmuir 1998, 14, 3932–3936.
versity, 1982, p. 4. 68. G. A. Attard, A. Ahmadi, J. Electroanal. Chem.
51. V. V. Batrakov, B. B. Damaskin, J. Electro- 1995, 389, 175–190.
anal. Chem. 1975, 65, 361–372. 69. O. A. Petrii, I. G. Khomchenko, J. Electro-
52. Yu. P. Ipatov, V. V. Batrakov, Elektrokhimiya anal. Chem. 1980, 106, 277–286.
1976, 12, 1174–1178. 70. E. Yu. Alakseyeva, V. A. Safonov, O. A. Pet-
53. A. P. Korotkov, E. B. Bezlepkina, B. B. Dam- rii, Elektrokhimiya 1984, 20, 945–950.
askin et al., Elektrokhimiya 1985, 22, 71. E. Lust, A. Jänes, K. Lust et al., J. Electroanal.
1298–1304. Chem. 1997, 431, 183–201.
54. R. Naneva, V. Bostanov, A. Popov et al., J. 72. S. Trasatti, J. Electroanal. Chem. 1971, 33,
Electroanal. Chem. 1989, 274, 179–183. 351–378.
55. I. A. Popov, N. Dimitrev, R. Naneva et al., J. 73. D. R. Lide, (Ed.), CRS Handbook of Chemistry
Electroanal. Chem. 1994, 376, 97–100. and Physics, 76th ed., CRS Press, New York,
56. V. V. Batrakov, N. I. Naumova, Elektrokhi- 1995,1996, 12-122–12-123.
miya 1979, 13, 551–555. 74. B. W. Callen, K. Griffiths, U. Memmert et al.,
57. L. E. Rybalka, D. I. Leikis, A. G. Zelinskii, Surf. Sci. 1990, 230, 159–174.
Elektrokhimiya 1976, 12, 1340–1341. 75. M. I. Rojas, E. P. M. Leiva, Surf. Sci. 1990,
58. V. A. Safonov, L. Yu. Komissarov, O. A. Pet- 227, L121–L124.
rii et al., Electrokhimiya 1987, 23, 1375–1381. 76. K. Heinzinger, Pure Appl. Chem. 1991, 63,
59. J. Arold, J. Tamm, Elektrokhimiya 1989, 25, 1733–1742.
1417–1418. 77. K. J. Scweighofer, X. Xia, M. L. Berkowitz,
60. E. M. Lazarova, Elektrokhimiya 1982, 18, Langmuir 1996, 12, 3747–3752.
1654–1655. 78. J. G. Gordon, O. R. Melroy, M. F. Toney,
61. A. Rodes, J. Clavilier, J. Electroanal. Chem. Electrochim. Acta 1995, 40, 3–8.
1993, 348, 247–264. 79. A. E. Russell, A. S. Lin, W. E. O’Grady, J.
62. F. T. Wagner, T. E. Moylan, Surf. Sci. 1988, Chem. Soc., Faraday Trans. 1993, 89,
206, 187–202. 195–198.
63. T. Iwasita, X. Xia, J. Electroanal. Chem. 1996, 80. R. R. Nazmutdinov, M. Probst, K. Heinzin-
411, 95–102. ger, J. Electroanal. Chem. 1994, 369, 227–231.
64. U. W. Hamm, D. Kramer, R. S. Zhai et al., J. 81. R. R. Nazmutdinov, M. S. Shapnik, O.
Electroanal. Chem. 1996, 414, 85–89. I. Malyucheva, Elektrokhimiya 1991, 27,
65. J. Clavilier, R. Albalat, R. Gómez et al., J. 1275–1278.
2.5 Analyzing Electric Double Layers with the Atomic Force Microscope 225
2.5 piezo translator. This height is controlled

Analyzing Electric Double Layers with the by a feedback loop that maintains a
Atomic Force Microscope constant force between tip and sample.
.. Image contrast arises because the force
Hans-Jurgen Butt
.. between tip and sample is a function
Universitat Siegen, Siegen, Germany
of both tip–sample separation and the
material properties of the tip and the
sample. To date, in most applications
2.5.1
the AFM was used to image surface
Introduction
topography in the contact mode. Then,
image contrast is obtained from the very
Electric double layers are studied by a
short-range repulsion that occurs when
variety of methods. These methods give
the electron orbitals of tip and sample
information about different properties of
overlap (Born repulsion). However, further
the double layer. One method to study interactions between tip and sample can
the properties of electric double layers is be used to investigate material properties
surface force measurements. They directly on a nanometer scale.
inform about the distance dependency Soon after the invention of the AFM,
of the potential and ion concentration. it was realized that by taking force-
In addition, surface force measurements versus-distance measurements, valuable
allow determining the surface charge or information about the surfaces could be
potential. A brief introduction to surface obtained [2, 3]. These measurements are
force measurements is given in the second usually known as ‘‘force measurements.’’
chapter. The technique of force measurements
With the introduction of the atomic force with the AFM is described in detail in
microscope (AFM), the measurement of the third chapter. Force measurements
surface forces was greatly facilitated. The with the AFM were first driven by the
AFM was invented in 1986 by Binnig, need to reduce the total force between
Quate, and Gerber [1]. It has become one tip and sample in order to be able to
of the major tools to analyze surfaces. image fragile, biological structures [4, 5].
Usually the AFM is used to image the Therefore it was obligatory to understand
topography of surface (Fig. 1). This can the different components of the force.
be done in vacuum, gaseous, or liquid In addition, microscopists tried to un-
environment. For imaging, the sample is derstand the contrast mechanism of the
scanned underneath a tip, also called a AFM to interpret images correctly. Nowa-
‘‘probe,’’ which is mounted to a cantilever days most force measurements are done
spring. The sample is scanned by a by surface scientists, electrochemists, or
piezoelectric translator. While scanning, colloidal chemists who are interested in
the force between the tip and the sample surface forces per se. Excellent short [6]
is measured by monitoring the deflection or comprehensive [7] reviews about sur-
of the cantilever. A topographic image face force measurements with the AFM
of the sample is obtained by plotting have appeared. Also an older review about
the deflection of the cantilever versus its surface force measurements in aqueous
position on the sample. Alternatively, it is electrolyte exists [8]. This overview focuses
possible to plot the height position of the on electrostatic double-layer forces.
Photosensor Fig. 1 Schematic drawing of an

Light beam
atomic force microscope.
Sample
z
y
x
Piezoelectric scanner
First measurements of the electrostatic with the DLVO theory, named after Der-
double-layer force with the AFM were jaguin, Landau, Verwey, and Overbeek [12,
done in 1991 [9, 10]. The electrostatic 13]. The DLVO theory uses a similar ap-
double layer depends on the surface charge proach as in the Gouy–Chapman (GC) [14,
density (or the surface potential) and the 15] and Debye–Hückel [16] models of the
ionic strength. A brief introduction to the electric double layer. To explain the coagu-
theory of the electrostatic force is given lation of sols after the addition of salts, the
in Chap. 4. The electrostatic double-layer DLVO theory takes explicitly into account
force is in many cases responsible for the interaction between colloidal particles.
the stabilization of dispersions. An AFM The interaction is assumed to consist of
experiment can be regarded as directly two contributions: the van der Waals at-
probing the interaction between a sample traction and an electrostatic double-layer
surface and a colloidal particle (or the AFM repulsion. At low salt concentration, the
tip). Since the AFM tip is relatively small, repulsion is strong enough to keep the
this interaction can be probed locally. The colloidal particles apart. With increasing
lateral spacial resolution can be of the order salt concentration, the electrostatic repul-
of few nanometers. sion is more and more screened, and at a
certain concentration, the van der Waals
2.5.2 attraction overcomes the repulsive electro-
Surface Forces in Aqueous Electrolyte static barrier and coagulation sets in.
Van der Waals forces arise from cor-
2.5.2.1 DLVO Forces relations between time-dependent dipoles
At the end of the 19th century, it was in atoms or molecules (for a review see
well known that many colloids in aque- Refs. [17–19]). The time dependence oc-
ous medium coagulate after the addition curs either because a permanent dipole is
of salt. It was even known that di- or changing its orientation in space or be-
trivalent ions are much more efficient in cause the dipole itself is changing with
destabilizing dispersions than monovalent time. The dominant term in the van
ions [11]. The explanation for this behavior der Waals force is usually the second
was eventually given in a quantitative way frequency-dependent term, also known as
the dispersion force. Since all materials one of the surface charges is reversed. For a
are polarizable, a van der Waals force oc- more complete description of double-layer
curs between all atoms and all surfaces. forces see Refs. [18, 19].
The force between two atoms is relatively
weak and short-ranged, decreasing pro- 2.5.2.2 Early Surface Force Measurements
portional to D −7 for small separations The development of the DLVO theory
and D −8 for larger separations where the stimulated an interest in measuring forces
force is ‘‘retarded’’ owing to weaker corre- between the surfaces to verify the theory
lations between the atoms. Calculation of (review Ref. [21], see also Ref. [22]). One
the forces in many atom systems is com- of the earliest direct force measurements
plex because the forces between individual were made between two polished glass
atoms are not additive, and the modern bodies [23, 24]. One glass body was fixed,
treatment of van der Waals forces, known the other was mounted to a spring. The
as Lifshitz theory, is a combination of distance between the two glass surfaces
quantum electrodynamic theory and spec- and the deflection of the spring were
troscopic data [20]. Fortunately, assuming measured. Multiplication of the deflection
additivity is a reasonable approximation in with the spring constant gave the force.
many cases, some simple equations can be Using these simple early devices and
derived. after overcoming severe problems, it
The electrostatic double-layer force is the was possible to verify the theoretically
other contribution to the DLVO force. This predicted van der Waals force for glass
force arises because of surface charges interacting with glass across gaseous
at interfaces. Water has a high dielectric medium [25, 26]. Derjaguin, Rabinovich,
constant. Thus, surface dissociation or and Churaev measured the force between
adsorption of a charged species in water two thin metal wires. In this way they
is very common. The surface charge is could determine the van der Waals force
balanced by dissolved counterions that for metals [27].
are attracted back to the surface by the In these early measurements two prob-
electric field, but spread away from the lems became obvious: the minimal dis-
surface to increase the entropy. Together tance accessible (≈20 nm) was rather lim-
the ions and charged surface are known ited and experiments could not be done
as the electric double layer. When another in liquid environment. In more recent
surface approaches, the double layer is experiments these problems could be over-
perturbed, and the resulting force is come and the minimal distance could
known as the double-layer force. When be decreased to ≈2 nm with improved
the approaching surface charges have devices [28, 29]. Historically, however, im-
the same sign, the concentration of ions provements were made by using other
between the surfaces always increases and types of devices.
results in a repulsive force. The origin of One inherent problem of all techniques
the surface charge is dissociation, so for to measure surface forces directly is
finite dissociation constants, the surface surface roughness. The surface rough-
charge is influenced by the interaction ness over the interacting areas limits
with another surface. This phenomenon is the distance resolution and the accu-
called charge regulation, and results in less racy of how well zero distance (contact)
repulsive forces, or even attractive forces, if is defined. Practically, zero distance is
the distance of closest approach because level [30–32]. The SFA contains two
if a protrusion is sticking out of one crossed atomically smooth mica cylin-
surface, the other surface cannot ap- ders of approximately 1 cm radius between
proach any further. Two methods can which the interaction forces are measured.
be used to reduce this problem of sur- One mica cylinder is mounted to a piezo-
face roughness: either atomically smooth electric translator. With this translator, the
surfaces are chosen or the interacting distance is adjusted. The other mica sur-
areas are reduced. The first approach face is mounted to a spring of known and
was chosen with the development of adjustable spring constant. The separation
the surface forces apparatus (SFA) and between the two surfaces is measured by
the measurement and analysis of surface means of an optical technique using mul-
interaction forces (MASIF), while the ad- tiple beam interference fringes. Knowing
vantages of the AFM are small interacting the position of one cylinder and the separa-
areas. tion to the surface of the second cylinder,
the deflection of the spring and the force
2.5.2.3The Surface Forces Apparatus can be calculated.
The development of the SFA (Fig. 2) was With the SFA, the main predictions of
a big step forward because it allowed the DLVO theory were verified. In par-
to measure directly the force law in liq- ticular the electrostatic double-layer force
uids and vapors at ångström resolution was analyzed for different salts and under
Transmitted light
to spectroscope
Microscope
R objective Lid of outer box
DS Glass cell Teflon tube

A
B A drop of PZ
liquid
crystal
S Mica
surfaces
Base of
3 cm outer box
Heat-filtered
white light
Fig. 2 Apparatus to measure surface forces between two mica surfaces (SFA).
(The figure is reproduced with kind permission from J. Klein [33].) This
particular device was built to study the behavior of liquid crystals confined in the
narrow gap between the mica surfaces.
different conditions. Also, limits of the force between lipid bilayers and biological
DLVO theory became obvious. The DLVO macromolecules.
theory treats the intervening medium as
continuous, so it is not surprising that 2.5.2.5 The Bimorph Surface Force
the model breaks down when the liquid Apparatus
medium between two surfaces is only few Despite the new insights achieved with
molecular diameters in width. At high salt the SFA, there are also severe problems
concentrations and small separations, an and limits. It is difficult and slow to oper-
additional monotonic repulsive force of ate, only mica and few other materials [42,
about 1 to 3 nm range was detected in 43] provide suitable substrates, elastic de-
aqueous electrolyte [34]. It has been at- formation has to be taken into account.
tributed to the energy required to remove With the invention of a bimorph surface
the water of hydration from the sur- force apparatus, known as MASIF, some
face, or the surface-adsorbed species (sec- of these limitations could be overcome [44,
ondary hydration), presumably because of 45]. In this case the force between two glass
strong charge–dipole, dipole–dipole or or polymer spheres of typically few mil-
H-bonding interactions. Because of the limeter diameter is measured. One sphere
correlation with the low (or negative) en- is therefore attached to a piezoelectric bi-
ergy of wetting of these solids with water, morph that serves as a force sensor.
it was termed hydration force [35–38]). With the MASIF, the existence of an-
other force component, namely, the hy-
drophobic force [46], which is not de-
2.5.2.4 The Osmotic Stress Method scribed by the DLVO theory and which
In parallel, another important (although was already observed with the SFA [47],
less direct) technique for measuring forces was confirmed. The hydrophobic force is a
between macromolecules or lipid bilay- monotonic attractive force of longer range
ers was developed, namely, the osmotic (100 nm) and greater magnitude than van
stress method [39–41]. A dispersion of der Waals forces. These occur in aqueous
vesicles or macromolecules is equilibrated medium between surfaces that generally
with a reservoir solution containing wa- exhibit high water contact angles (e.g.
ter and other small solutes, which can hydrocarbon and fluorocarbon surfaces).
freely exchange with the dispersion phase. At present there is no clear understand-
The reservoir also contains a polymer ing of the mechanism of these forces. It
that cannot diffuse into the dispersion. is easy to conceive of a negative hydra-
The polymer concentration determines tion force, for example, the disruption of
the osmotic stress acting on the disper- water H-bonding near a noninteracting
sion. The spacing between the macro- surface, with a similar range as the hy-
molecules or vesicles is measured by dration force. However, it is difficult to
X-ray diffraction (XRD). In this way, one conceive of one with the long measured
obtains pressure-versus-distance curves. range. At present the focus of understand-
The osmotic stress method is used to ing these long-ranged forces is on the role
measure interactions between lipid bilay- of dissolved gas.
ers, DNA, polysaccharides, proteins, and Even with the MASIF, a severe limit re-
other macromolecules [36]. It was particu- mained: that of surface roughness. Glass
larly successful in studying the hydration and, in particular, polymer surfaces are not
atomically smooth over areas of millimeter to the piezoelectric translator, onto which
or even 10 µm diameter. This limit could the sample is mounted, while measuring
be drastically reduced with the AFM. In the cantilever deflection. The direct result
the AFM the interacting areas are typically of such a force measurement is the
10 to 50 nm in diameter. Hence, all sam- cantilever deflection, zc , versus position
ples whose surface is smooth, clean, and of the piezo, zp , normal to the surface
homogeneous over such small areas can (Fig. 3). To obtain a force-versus-distance
be quantitatively analyzed. This increased curve, zc and zp have to be converted
the number of accessible materials enor- into force and distance. The force, F , is
mously. obtained by multiplying the deflection of
Small interacting areas provide another the cantilever with its spring constant K:
principal advantage. When two solid sur- F = K · zc . The tip–sample separation
faces come into contact, the surfaces de- D is calculated by adding the deflection to
form. Even under zero load, the adhesion the position.
between two surfaces leads to the forma- The deflection of the cantilever is
tion of a contact area rather than a contact usually measured using the optical lever
point. It also leads to an indentation of technique. Therefore a beam from a laser
the two interacting surfaces. For a sphere diode is focused onto the end of the
and a flat planar surface (or two crossed cantilever and the position of the reflected
cylinders), this central indentation
δ at neg- beam is monitored by a position sensitive
ligible load is proportional to δ ∝ 3 R/E 2 , detector (PSD). Often the back of the
cantilever is covered with a thin gold
where R is the radius of the sphere and E
layer to enhance its reflectivity. When a
is Youngs modulus [48, 49]. Using typical
force is applied to the probe, the cantilever
values for the SFA (R = 1 cm, E of mica
bends and the reflected light beam moves
≈60 GPa [50]), the MASIF (R = 1 mm, E
through an angle equal to twice the change
of silica ≈60 GPa), and the AFM (R =
of the end-slope tan α. For a cantilever
50 nm, E of silicon nitride ≈150 GPa),
with a rectangular cross section of width
the estimated central indentations at zero
w, length L, and thickness h, the change
loads are in the order of 14 nm, 6 nm, and
of the end-slope (Fig. 4) is given by
0.1 nm, respectively. This was calculated
with the theory of Johnson, Kendall and 6F L2
Roberts [48, 49] using a surface energy of tan α = (1)
Ewh3
γ = 0.05 N m−1 in all the cases. Hence,
E is the elastic modulus of the cantilever
in a usual AFM measurement, the elas-
material. F is the force applied to the
tic deformation can be neglected, while in
end of the cantilever in normal direction.
SFA and MASIF experiments, deforma-
The signal detected with the optical lever
tions have to be taken into account.
technique is proportional to the end-slope
2.5.3
of the cantilever. The deflection of the
Force Measurements with the AFM cantilever is given by
4F L3 2
2.5.3.1 The Practice of Force zc = 3
= L tan α (2)
Measurements Ewh 3
In a force measurement the sample is Hence, the deflection is proportional to
moved up and down by applying a voltage the signal. One should, however, keep in
Fig. 3 Schematic deflection-versus- (c) (b)

piezo translator position curve (zc
versus zc ). At large separation, no
force acts between tip and sample (a).
The cantilever is not deflected. When
approaching the surface, it was
assumed that a repulsive force is acting
(b) and the cantilever is bent upwards.
At a certain point, the tip often jumps
onto the sample. This happens when
the gradient of attractive forces, for ∆zc
example, the van der Waals forces,
exceeds the spring constant of the
cantilever. After the jump-in, the tip is in
contact with the sample surface (c).
When retracting the tip, often an ∆zp
adhesion force is observed (d) and the
tip has to be pulled off the surface. The
deflection-versus-piezo translator
position curve has to be transformed as D
described in the text to obtain a
force-versus-distance curve
F = K ⋅ ∆zc
(F versus D). (d) (a)
D = ∆zp +
∆zc
Fig. 4 Schematic diagram of a w

cantilever with rectangular cross L
section.
h α ∆zc
F
mind that these relations only hold under with t being the time for measuring a pixel
equilibrium condition. If the movement of of the force curve in seconds. The result
the cantilever is significantly faster than is in meter. With typically t = 0.1 ms, the
that allowed by its resonance, frequency deflection resolution is 0.1 Å. In practice
Eqs. (1) and (2) are not valid anymore and the deflection sensitivity is often limited
the signal is not necessarily proportional by thermal cantilever vibrations, which are
√
to the deflection. 4kB T /3K [103]. Here, kB and T are the
The resolution of the optical lever
√ tech- Boltzmann constant and the absolute tem-
nique is approximately 10−13 / t [102], perature, respectively. With typical spring
constants between 0.01 to 1 N m−1 , the resonance frequency is also important

amplitude of thermal noise is between 7 to to be able to scan fast because the
0.7 Å at room temperature. resonance frequency limits the time
resolution [104, 105]. In liquids the
resonance frequency is reduced by
2.5.3.2 The Cantilever
a typical factor 3 to 6 because of
An important property is the spring con-
the liquid that is dragged with the
stant of the cantilever K. The spring
cantilever [105–107].
constant can in principal be calculated
• Cantilevers have different top and bot-
from the material properties and dimen-
sions of the cantilever. For a cantilever with tom faces. The top portion is usually
rectangular cross section, it is coated with a layer of gold to in-
crease its reflectivity. Therefore, any
F Ewh3 temperature change leads to a bend-
K= = (3) ing of the cantilever as in a bimetal. In
zc 4L3
addition, adsorption of substances or
The cantilever is in fact a key element of electrochemical reactions in liquid en-
the AFM and its mechanical properties are vironment slightly change the surface
largely responsible for its performance. A stress of the two faces. These changes
good cantilever should have a high sensi- in surface stress are in general not the
tivity. High sensitivity in zc is achieved same on the bottom and the top. Any
with low spring constants or low ratio h/L. difference in surface stress σ will lead
Hence, in order to have a large deflection at to a bending of the cantilever [108, 109]
small force, cantilevers should be long and
thin. In addition, the design of a suitable 4L2 σ
cantilever is influenced by other factors: zc ≈ (5)
Eh2
• With the optical lever technique, deflec- Practically, these changes in surface
tions down to 0.1 Å can be detected. It is stress lead to an unpredictable drift of
useless to make cantilevers so soft that the cantilever deflection that disturbs
thermal fluctuations exceed this value force measurements. To reduce drift,
by more than, say, a factor 10. Hence, the ratio h/L should be high.
the lowest useful spring constant is in
the order of 0.02 to 0.1 N m−1 at room Hence, the optimal design of a cantilever
temperature. is a compromise between different fac-
• External vibrations, such as vibrations tors. Depending on the application, the
of the building, the table, or the appropriate dimensions and materials are
noises that are usually in the low- chosen. Cantilevers for AFMs are usually
frequency regime, are less transmitted ‘‘V’’ shaped to increase their lateral stiff-
to the cantilever, when the resonance ness. They are typically 100 to 200 µm
frequency of the cantilever [104, 105] long (L), each arm is about 20 µm wide
(w) and 0.5 µm thick (h). The spring con-
h E stant of ‘‘V’’-shaped cantilevers is often
νr = 0.163 2 (4) approximated by that of a rectangular bar
L ρ
of twice the width of each arm [104]. A
(ρ is the density of the cantilever more detailed analysis of the mechanical
material) is as high as possible. A high properties of cantilevers and calculations
of spring constants are given in Refs. [105, 2.5.3.3 The Probe: Tip or Colloidal Particle
110–113]. In the AFM, the two interacting surfaces
In principle, the spring constant can are the planar sample surface and the
be calculated. The thickness and modulus surface of the AFM probe. For microscopy
of elasticity of cantilevers are, however, the most commonly used probes are the
not easy to determine, so it is desir- sharp microfabricated silicon nitride or
able to measure spring constants. Several silicon tips that provide a high lateral
methods have been described [105, 114, resolution. These probes also provide high
115] but none appears to be simple, re- resolution when measuring surface forces,
liable, and precise at the same time. but introduce the problem that the surface
In the most common methods, tung- geometry is difficult to determine in
sten or gold beads are attached to the the 10-nm regime [120–122]. This would,
end of the cantilever spring. The spring however, be necessary since typical radii of
constant can be calculated by measur- curvature are from 5 to 60 nm. In addition,
ing a change in the resonant frequency the surface chemistry of silicon nitride tips,
due to the additional mass at the end which are most frequently used for force
of the spring [116]. Torii and coworkers measurements, is rather complex [51, 52,
determine the spring constant of micro- 123, 124].
fabricated cantilevers by pressing them Alternatively, spherical particles of typ-
against a calibrated macroscopic reference ically 2 to 20 µm diameter are attached
cantilever [117]. Sader and coworkers de- to the end of cantilevers. This can be
termined the spring constant of cantilevers achieved either by gluing them with a
with a rectangular cross section from the water-resistant polymer glue [9, 10] or by
resonance frequency and the quality fac- melting [97] or sintering them onto the
tor [118]. Another promising approach is to cantilever (Fig. 5). With particles of defined
use thermal cantilever noise to determine radius, the force can be analyzed more
the spring constant [119]. quantitatively. In particular, the surface
(a) 3 µm (b) 3 µm
Fig. 5 Scanning electron micrographs of a silica microsphere glued with polymer glue to
the end of a silicon nitride cantilever (a). The polystyrene particle shown to the right was
sintered at 120 ◦ C to the cantilever (b).
charge and den Hamaker constant can gold or other materials that can be sput-
be extracted. The number of materials tered or evaporated [125, 126]. First, the
that can be brought into the shape of material is deposited onto mica. Then, a
microspheres with sufficient smoothness steel plate is glued on top. Finally, the
is, however, limited. Most frequently, sil- steel plate with the deposited material is
ica microspheres are used. Silica spheres cleaved off the mica surface. Since mica
can be made with different diameters. is extremely inert, chances are high that
They have a low surface roughness of cleavage really happens at the mica sur-
less than 1 Å (as determined by AFM) face. The surface of the deposited material
and their surface chemistry is well known. now exposed can then be used in a force
Different kinds of polymer particles are experiment.
commercially available. They are, how-
ever, much rougher than silica. Their 2.5.3.4 An Instability: The Jump-in
roughness can be decreased by anneal- Implicit in the measurement of surface
ing roughly at the glass transition tem- forces with a ‘‘spring force’’ is the as-
perature. Still, the surface roughness is sumption that the two are equal and
significantly higher than silica. Zirconia opposite and thus the spring is not in
microspheres were produced by annealing motion. However, close above the sur-
zirconia powder [83]. face, the spring often cannot provide
While the preceding paragraph concen- a sufficient force to counterbalance the
trated on the importance of knowing the surface force. These regimes are charac-
geometry of the tip, it is equally impor- terized by rapid motion of the cantilever,
tant to know the geometry of the sample. and cause difficulty for determining sur-
Again, the most convenient geometry is a face forces. The onset of this unstable
smooth sphere or a planar surface. Prob- regime (‘‘jump-in’’) is characterized by
the point at which the gradient of the
lems due to sample roughness are greatly
attractive force exceeds the spring con-
reduced in the AFM compared to other
stant plus the gradient of the repulsive
surface force techniques since the sample
force [127]. Several techniques have been
needs only to be smooth on a scale com-
employed to access the unstable regime:
parable to the radius of curvature at the
(1) the application of a force feedback
end of the tip. Several materials such as (electric or magnetic) to balance the sur-
mica, oxidized silicon, graphite (HOPG), face force [33, 128–131]; (2) measurement
MoS2 , and so on are available with suffi- of the time-dependent force and dis-
cient smoothness over the required areas placement in the unstable regime, then
(Table 1). In addition, the surface of many calculation of the velocity and accelera-
materials can be smoothed by template tion to yield the surface force according
stripping. Therefore the material, say a to Newton’s equation of motion [132];
polymer, is melted on a smooth sur- (3) use a high approaching speed so that
face [76]. This can be a mica or a silicon the hydrodynamic drag FH , which for a
wafer surface. After cooling and right be- sphere approaching a flat with a veloc-
fore an experiment, the surface is peeled ity v increases with decreasing distance
off and the freshly exposed polymer sur- as FH = 6πηvR 2 /D (η is the viscosity
face is used for the force measurement. of the liquid) [133], compensates for the
The same technique can also be used for attraction [134].
Tab. 1 Measurements of electrostatic double-layer forces with the AFM
Probe Sample Refs.
Silicon nitride tip Silicon nitride 51, 52

Silicon nitride tip Mica 4, 10, 51
Silicon nitride tip SiO2 53
Silicon nitride tip Al2 O3 54
Silicon nitride tip Graphite 55
Silicon nitride tip PEEK 56
Silicon nitride tip LB film 57
Silicon nitride tip Silane monolayer on SiO2 58
Silicon nitride tip Lipid bilayer 59, 60
Silicon nitride tip Purple membrane 54, 61, 62
Silicon nitride tip Oocysts 63
SiO2 sphere SiO2 9, 52, 64–69
SiO2 sphere Mica 68, 70
SiO2 sphere Al2 O3 71, 72
SiO2 sphere TiO2 73, 74
SiO2 sphere Carbon 75
SiO2 sphere Polypropylene 76, 77
SiO2 sphere Ultrafiltration membrane 78
Glass sphere SiO2 10, 79
Amine-modified glass sphere Amine-modified glass 80
LB film on glass sphere LB film on glass 81
Al2 O3 particle Mica 10
TiO2 particle TiO2 82
ZrO2 sphere ZrO2 83–85
ZnS sphere Mica 86
ZnS sphere PbS 87
ZnS sphere ZnS sphere 88
Gold sphere Gold 64, 89–93
Thiol monolayer on gold-coated tip Thiol monolayer on gold sample 94–96 (review)
Polystyrene sphere Polystyrene 97, 98
Polystyrene sphere Mica 98
Protein-coated probe Protein-coated sample 99, 100
Bacteria-coated tip Mica/glass 101
Note: AFM measurements of electrostatic double-layer forces in aqueous electrolyte. Force

measurements between surfactant layers are not included. They are discussed in the text. PEEK:
Poly(etheretherketone); LB film: Langmuir–Blodgett film.
Another instability occurs when tip or jump occurs for nonelastic deformation.
sample surfaces are not perfectly rigid but In this case the jump depends on the rate
are deformed by the tip. For that case of of the plastic deformation [136].
elastic deformation, Pethica and Sutton
showed that at sufficiently small separa- 2.5.3.5 AFM-related Techniques
tions, typically 1 to 2 Å, the tip and the Commercial AFMs are made for imaging.
sample would jump together, irrespective For force measurements, certain features,
of apparatus construction [135]. A similar such as the lateral scanning capability and
the feedback electronic that regulates the fluid cell can be made of one or few mate-
height, are not required. Hence, if one rials (such as Teflon and quartz) that can
plans to use the device solely for force mea- be cleaned thoroughly, for example, with
surements, it is significantly less expensive hot sulfuric or nitric acid.
to build homemade setups [137–139]. In
addition, some features of commercial 2.5.4
AFMs are not perfectly suited for measur- Theory of Measuring Surface Forces with
ing force curves. One such feature is the the AFM in Aqueous Electrolyte
split photodiode, which has an extremely
high sensitivity but a limited dynamic 2.5.4.1 Derjaguins Approximation
range. In addition, the sensitivity depends The force between two surfaces depends
on the precise shape of the reflex and the on both the material properties and the
AFM needs to be calibrated for each ex- geometry of the surfaces. Derjaguin [144]
periment [140]. These disadvantages were approximated the influence of arbitrary
overcome in some homebuilt devices by re- geometry on the interaction potential
placing the spilt photodiode with a linear V (D) by reducing it to the simple geometry
position-sensitive device [132, 141]. With of two flat surfaces:
∞
position-sensitive devices, the accessible
force range is highly increased. This is V (D) = VA (z) dA (6)
D
particularly important when measuring
long-range forces during the approach and V (D) is the interaction potential that
the adhesion force during the retraction depends on the distance between the in-
in one experiment. With microfabricated teracting surfaces D. VA (z) is the potential
tips, long-range forces in aqueous elec- energy per unit area between two flat sur-
trolyte are of the order of 0.1 to 1 nN. faces that are separated by a distance z.
Adhesion forces can be 100 times higher. dA is the increase of the cross-sectional
To have a sufficient sensitivity as well as area of the two surfaces with increasing
the capability to measure the strong ad- separation distance.
hesion force, a position-sensitive device is According to Eq. (6), the fundamental
necessary. property of the material is the interaction
The position of the sample is adjusted by energy, VA , (or force, FA = −dVA /dz) per
the piezoelectric translator. Piezoelectric unit area. The most useful measurement
crystals show creep and hysteresis, which is the one between two surfaces of known
affect the accuracy of the distance determi- geometry, so that VA can be determined
nation [142]. One possibility to overcome and thus the interaction energy of all
this problem is to use piezoelectric trans- geometries can be calculated.
lators with integrated capacitative position The problems of unknown surface
sensors, which are commercially avail- geometry and chemistry can partly be
able [143]. In the same setup, another solved by attaching smooth micrometer-
deficit of commercial AFMs was overcome. sized spheres to the cantilever. For a
Standard fluid cells of commercial AFMs sphere of radius R, and a flat, a simple
are small and manually difficult to access. equation relates the measured force, F , to
In addition, they consist of different ma- the interaction energy per unit area, VA :
terials (glass, steel, silicon, etc.) that are F (D)
difficult to clean. In self-made devices the VA (D) = (7)
2πR
From this relationship, it can also be seen calculations that are also valid at higher
that a larger radius results in a higher potentials and that explicitly take into
force, and thus greater sensitivity in VA . account the geometry, see Refs. [150–154].
Naturally, this increase in VA sensitivity The requirement of low potentials is
comes at the expense of reduced lateral not the only limit of Eq. (8). In ad-
resolution. Another advantage of using dition, Eq. (8) is restricted to distances
spheres as probes is the possibility to make roughly larger than the Debye length. For
a variety of probes by attaching particles D ≤ λD , boundary conditions influence
of different chemical composition to the the force. The force is calculated from
cantilevers. Relation (7) is derived from the potential distribution between the two
Eq. (6) by assuming a parabolic shape surfaces [18, 19]. In continuum theory the
of the probe. It can also be used for potential distribution is determined from
microfabricated tips since they often show the Poisson–Boltzmann equation, which
a roughly parabolic shape. In this case, R is is a second-order differential equation.
the radius of curvature of the tip at its end. To solve this equation, certain bound-
ary conditions have to be assumed. Two
2.5.4.2 The Electrostatic Double-layer boundary conditions are often used: ei-
Force ther it is assumed that upon approach
When two charged surfaces approach each the surface charges remain constant (con-
other at some point, the electric double stant charge) or that the surface poten-
layers start to overlap and a force, also tials remain constant (constant potential).
called disjoining pressure, begins to act [13, These boundary conditions have a strong
145]. This electrostatic double-layer force influence on the electrostatic force at dis-
decays roughly exponentially. When the tances smaller than λD . Two surfaces with
surface potentials of both surfaces are constant charge of equal sign always repel
below approximately 25 to 50 mV, the force each other for D → 0. Two surfaces with
between a spherical tip and a flat sample constant potential are attracted for D → 0,
can be approximated in an analytical even when the surface potentials have the
form [146–149]: same sign (except for the hypothetical case
that the potentials are precisely equal in
4πRσ1 σ2 λD −D/λD
Fel = e (8) magnitude and sign) [155].
εε0 To demonstrate the effect of boundary
with the Debye length conditions, Fig. 6 shows the force between
a microsphere of 3-µm radius interacting
εε0 kB T with a flat surface versus distance. The
λD = force was calculated with the so-called
2n∞ e2
nonlinear Poisson–Boltzmann equation,
Here, σ1 and σ2 are the surface charge which is also valid for potentials above
densities of probe and sample, ε0 is 50 mV. A salt concentration of 1 mM
the permittivity of free space, ε is the was chosen resulting in a Debye length
dielectric constant, e is the unit charge. λD = 9.6 nm. At close distances, the force
The Debye length λD is determined by calculated with constant charge bound-
the salt concentration; n∞ is the bulk ary conditions is much higher than the
concentration of a monovalent salt in force calculated assuming constant poten-
molecules per volume. For more detailed tial. For D λD , the force calculated with
10 000
σ 1 = 0.036 e nm−2, σ 2 = 0.023 e nm−2
1000 1 mM salt
[µN m−1]
λ D = 9.6 nm
F/R
Ψ1 = 80 mV
100
Ψ2 = 50 mV
10
0 10 20 30 40 50 60
Distance
[nm]
Fig. 6 Electrostatic double-layer force acting on a sphere of 3-µm
radius in water containing 1 mM monovalent salt. The force was
calculated with the nonlinear Poisson–Boltzmann equation and
the Derjaguin approximation for constant potentials (1 = 80 mV,
2 = 50 mV) and for constant surface charge (σ1 = 0.0058 C m−2
= 0.036 e nm−2 , σ2 = 0.0036 C m−2 = 0.023 e nm−2 ). The
surface charge was adjusted by σ1/2 = εε0 1/2 /λD , so that at large
distances both lead to the same potential. In addition, the force
expected with approximation (8) is plotted as a dotted line.
constant charge and constant potentials In this model the surface charge is caused
are equal and surface charges and po- by the dissociation of ions from surface
tentials can be converted according to groups. The dissociation constant and as a
σ1/2 = εε0 1/2 /λD . Also the force calcu- consequence, the surface charge depends
lated with approximation (8) is plotted. It on the potential. This also explains the
lies between the two extreme cases. often observed dependency of the surface
Which boundary condition is more real- charge on pH and the salt concentration.
istic depends on the materials used. In The charge regulation model is success-
addition, the electrolyte and the speed fully applied to many materials, especially
of the approach might have an influ- those in which the surface charge is mainly
ence. Prica and coworkers measured force determined by pH, such as oxides, sil-
curves between zirconia, which showed icon nitride [51, 52, 123, 124], mica [34,
a constant charge behavior [84]. Also the 157–159], and biological materials.
force between two surfaces coated with
densely packed carboxylic groups followed 2.5.4.3 Other Surface Forces in Aqueous
constant charge conditions [94]. In other Electrolyte
cases, constant potential conditions are With any surface force–measuring de-
more appropriate. Most cases, however, vice, the total force between two surfaces
lay between the two extremes. Then often, is determined. The origin of all surface
a charge regulation model is applied [156]. forces is the interaction between electric
charges. Under normal circumstances it progress has been made during the last
is practical, however, to separate the to- few years in calculating Hamaker con-
tal force into several components and take stants from spectroscopic data. A list of
the components as being additive (for a Hamaker constants for different combina-
critique see Ref. [160]). These components tions of the three media involved is given
are the van der forces, the electrostatic in Refs. [163, 164]. There are some cases
double-layer force, the hydration repulsion in which the van der Waals force is re-
(between hydrophilic surfaces), and the hy- pulsive [165]. A repulsive van der Waals
drophobic attraction (between hydropho- force was, for example, detected between
bic surfaces). To analyze the electrostatic a silicon nitride tip and a silicon oxide
double-layer force, one needs to separate it surface in diiodomethane [166]. Also the
from all other components. This is usually van der Waals interaction between many
done by assuming a certain distance de- solids and vapors in water is often repul-
pendency for the van der Waals force and sive and causes the adsorption of water to
the electrostatic force. In addition, it can solid surfaces.
be identified by the influence of salt. Hydration forces [35–38] are relatively
The van der Waals potential per unit area short-ranged, so that at salt concentra-
is given by tions below 0.1 M, they can easily be
distinguished from the longer range elec-
H trostatic and van der Waals forces. They
VA = − (9)
12πD 2 are repulsive and, except for the case
Using Eq. (7), the force between a sphere of molecularly smooth surfaces and low
interacting with a plane can be calculated: salt concentrations where force oscillations
were observed [167], decay exponentially
HR with distance:
Fv dW = − (10)
6D 2
V A = Ae−D/λ0 (11)
For D 10 nm, retardation effects have
to be considered, and the force decays Characteristic decay lengths λ0 deter-
with D −3 instead of D −2 . H is the so- mined with the SFA, the osmotic stress
called Hamaker constant that depends on method, or the AFM range from 0.2 to
the dielectric properties of the interacting 1.4 nm. Typical amplitudes are A = 10−3
materials and the medium between. De- to 10 J m−2 . In contrast to the electrostatic
tailed calculations showed that Derjaguins double-layer force, hydration forces tend
approximation leads to errors that are usu- to become stronger and longer ranged
ally less than 10% [161]. For large distances with increasing salt concentration, espe-
the precise shape of the probe has to be cially for divalent cations.
considered [19, 162]. Hydrophobic forces occur between sur-
The van der Waals force is usually at- faces that are hydrophobic and show a
tractive. It is always attractive between two contact angle around 90◦ or higher. They
identical materials immersed in any third are attractive and often lead to a jump-in
material. For hydrocarbons immersed in at distances between 5 and 100 nm. They
water, the Hamaker constant is about can easily be avoided by using hydrophilic
0.2 − 1 × 10−20 J, for many oxides in wa- probes. For contact angles below 70 to
ter, it is 0.8 − 5 × 10−20 J, and for metals 80◦ , they are usually much smaller then
in water, it is 30 − 40 × 10−20 J. Great DLVO force and can be neglected [168].
Microfabricated tips of silicon nitride or structures [54, 61, 62], adsorbed protein
silicon oxide are naturally hydrophilic. One layers [99], or even whole cells [63, 101]. In
should, however, keep in mind that they all cases the force could be well described
might contaminate and as a result become by DLVO theory; independently measured
hydrophobic. surface potentials (or charge densities) and
Hamaker constants agreed with AFM re-
2.5.5 sults. Only at short separation below 1
AFM Force Measurements of the Electric to 5 nm, differences occurred [10, 65, 79,
Double Layer 176], particularly at high ionic strength.
These are usually attributed to hydration
2.5.5.1 Between Solid Surfaces forces.
With the AFM, DLVO forces were mea- As one example, the force-versus-
sured between several materials that are of distance between a silica particle and a
interest especially in colloidal science, for titania flat is shown in Fig. 7. Like in
example, glass, silica, and silicon nitride [4, many publications, the force is scaled
9, 10, 51, 53, 64–69], gold [64, 89–91], by division through the radius of the
zinc, and lead sulfide [86–88], titanium silica microsphere. According to Eq. (7),
oxide [73, 74, 82], zirconia [83–85], alu- the scaled force, F /R, is equal to 2πVA .
mina [10, 71, 72], different polymers [56, Force curves were recorded at different
76, 77, 97, 98, 169–171], LB and pH values ranging from pH 8.8 for the
other thin organic layers [57, 58, 94, 95, top curve to pH 3.0 for the bottom curve.
172], surfactant adsorbed to solid sur- The surface charges of both materials are
faces [92, 173–175], biological membrane mainly determined by the pH. Silica has
1.5
1.0 pH
[mN m−1]
F/R
0.6
0.1
−0.3
0 10 20 30 40 50 60
Distance
[nm]
Fig. 7 Force between a silica (SiO2 ) microsphere of 2.5-µm radius
and a titania (TiO2 ) crystal versus distance. The force is scaled by the
radius of the sphere. The curves were recorded at pH values of 8.8,
7.2, 6.3, 5.3, and 3.0 from top to bottom with 1 mM KNO3
background electrolyte. (The figure is reproduced with kind
permission from C. J. Drummond [73].)
an isoelectric point around pH 3.0, while Campbell and Hillier determined the force
the isoelectric point of titania is pH 5.6. between a silica sphere as probe and
As a consequence, at high pH, where carbon [75]. Figure 8 shows the result of
both materials are negatively charged, an Hillier and coworkers [179] who measured
electrostatic repulsion is observed. The re- the force between a gold sample and a
pulsion decreases as the pH decreases, silica sphere attached to the end of a
and at pH 3.0, that is, below the isoelectric cantilever. Silica is negatively charged at
point of titania, there is an electrostatic the pH ≈ 5.5 used in these experiments.
attraction as well as a van der Waals force As expected, the silica microsphere is re-
resulting in an overall attraction between pelled at negative potentials of the gold.
the two surfaces. When changing the potential to more pos-
itive values, the repulsion decreased and
2.5.5.2 High Surface Potentials finally attractive forces were observed. For
With classical surface force–measuring separations larger than 10 nm, the force
techniques, it is impossible to analyze the curves could be described by DLVO theory.
electrostatic double layer at surface poten- For smaller separations, measured force
tials significantly higher than 0.1 V. The curves deviated significantly from calcu-
reason is that high surface potentials can lated force-versus-distance curves. Hillier
only be obtained by applying an external and coworkers attribute this discrepancy
potential to a metallic or conducting sam- to several factors: an overestimation of
ple. All materials that are accessible to the Hamaker constant, the roughness of
the SFA or MASIF are not conducting. the two interacting surfaces, and hydration
Therefore it was a particular challenge to effects.
measure electrostatic double-layer forces Although the results of Hillier and
at high surface potentials with the AFM. coworkers qualitatively agree with results
In first experiments, Ishino and cowork- of Raiteri and coworkers [178] obtained
ers [177] coated a microfabricated AFM tip with silicon nitride tips, there are also
with gold and used it as working elec- some discrepancies. First, the force ob-
trode. They indeed observed a variation served between silica and gold changed
of the force with the applied potential almost linearly with the applied poten-
when measuring the interaction between tial in a range from −0.7 to +0.4 VSCE .
the gold-coated tip and a stearic acid Raiteri and coworkers observed a change
monolayer. An electrostatic repulsion was of the force only in a narrow potential
observed at negative applied potentials range of approximately 0.4 V. Above and
and attractive forces were measured at below this voltage range, the force sat-
positive potentials. Since the stearic acid urated. This saturation behavior is also
monolayer was negatively charged, such a predicted by DLVO theory. Second, Hillier
behavior is expected. and coworkers [179] always measured de-
A more convenient setup consists of cay lengths that agreed with the calculated
a conducting sample, which serves as Debye length. Raiteri and coworkers ob-
a working electrode, and an insulating served the same for potentials that lead to
probe. Raiteri and coworkers and Döppen- a repulsive force. In the attractive regime,
schmidt and coworkers [55, 178] measured however, the decay length was significantly
the force between a gold, platinum, or larger than the Debye length. These two
graphite sample and a silicon nitride tip. questions are still open.
Cantilever with
silica microsphere
RE
Potentiostat CE
Gold
sample
WE
Sample stage
1
−700 1
−500 0.8 −700
0.8 V −400
−400
−300 0.6
[mV] −200 V −200
[mN m−1]
−100 0.4 [mV] −100

F/R
0.6 0 0
+100 0.2 +100
0
[mN m−1]
0.4 −0.2
F/R
−0.4
0.2 0 2 4 6 8 10 12 14 16 18 20
Distance
[nm]
0
−0.2
−0.4
0 10 20 30 40 50 60
Distance
[nm]
Fig. 8 Force between a silica sphere and gold electrode in an aqueous
electrolyte solution of 1 mM KCl and pH ≈ 5.5 as a function of the applied
potential at the gold electrode. The curves correspond to, from top to bottom,
electrode potentials of −0.7, −0.5, −0.4, −0.3, −0.2, −0.1, 0, and +0.1 V
(versus SCE). Inset: Force curves recorded in a 10 mM KCl solution. (The figure
is reproduced with kind permission from A. C. Hillier [179].)
Also, preliminary experiments are done 2.5.5.3 Between Deformable Surfaces

for the symmetric case of gold interacting While the electric double layer on a solid
with gold [93]. Both gold surfaces were surface is relatively well understood and
shortcut and served as working electrodes. theories are able to account for colloidal
In this case a repulsion was observed at stability and coagulation kinetics quite
high positive potentials and high negative well, there has been much less success in
potentials. In between, around the point understanding the double-layer structure
of zero charge, the repulsion was reduced. at liquid–liquid or liquid–gas interfaces.
This behavior agrees with predictions of This is despite the fact that the stability of
continuum theory. emulsions or dispersion of particles and
gas bubbles play a central role in many experiments were not very precise and
industrial processes such as flotation or the the distance resolution was below 10 nm.
deinking of paper. With the AFM or with The first precise experiments were done by
AFM-related setups, the electric double Ducker and coworkers [182] who observed
layers at such deformable interfaces can an electrostatic double-layer force that de-
be analyzed. It is possible to measure cayed with the expected Debye length. This
the force between a solid microsphere observation that the electric double layer
and an oil drop (or another immiscible decays exponentially with the Debye length
liquid) or a bubble in aqueous medium. even at air–water interfaces was later con-
These experiments are, however, more firmed with devices that were equipped
difficult to perform and to interpret. First, with capacitatively calibrated piezo trans-
since the interface is deformable, it is lators (Fig. 9). Ducker and coworkers de-
difficult to determine zero distance. In tected, however, an unexpected attractive
principle, the shape of the interface can be force between a presumably hydrophilic
calculated using the Laplace equation [180, silica particle and a bubble. Fielden and
181]. This is, however, not trivial and in coworkers confirmed that between hy-
many cases practically impossible. Second, drophilic particles and air bubbles only
the tip or particle can even penetrate into repulsive forces are acting [183] and their
the bubble or oil drop. In this case, a experiments can be considered as the first
three-phase contact line is formed and accurate measurements. As expected from
the capillary force completely dominates adsorption experiments, a hydrophilic par-
the interaction. Electrostatic double-layer ticle is repelled by an air bubble and a
forces can only be detected before a three- stable water film remains on the particle
phase contact is formed. surface. The previously observed attrac-
First experiments were done between tion was probably due to contamination.
differently treated silica particles and an In contrast, a hydrophobic particle imme-
air bubble [132]. As expected, they showed diately forms a three-phase contact and
that a hydrophilic particle is repelled by snaps into the bubble (or vice versa: the
an air bubble, while a hydrophobic par- bubble engulfs the particle). Before con-
ticle jumps into the bubble. These early tact, the electrostatic double-layer force
100
Decay length
Fig. 9 Decay lengths of the

[nm]
exponentially decaying component

versus one over the square root of the
10
salt concentration. The dashed line
corresponds to the calculated Debye
length. The figure summarizes results of
measurements with hydrophilic and
hydrophobic silica microspheres in
solutions with different concentrations 10 100
of SDS (sodium dodecylsulfate) or 0.3/sqrt (Conc. mM−1)
DTAB (from Ref. [143]). [nm]
decaying with the Debye length could be reverses its surface charge from nega-
observed in both cases. tive to positive and it makes the surface
Ducker and coworkers [182] and Fielden hydrophobic.
and coworkers [183] performed their ex- When increasing the DTAB concen-
periments with commercial AFMs. For tration to 5.4 mM, a small electrostatic
routine measurements this is not conve- repulsion was observed before the jump-
nient because the liquid cells of commer- in. The repulsion was at least partly due
cial instruments are small and difficult to to electrostatic repulsion since it decayed
access manually. In addition, a thorough exponentially with the Debye length. The
cleaning procedure is hampered by the reason is probably that at DTAB concen-
fact that the cell consists of many different trations above 5 mM also, the air–water
materials. Therefore, Preuss and cowork- interface becomes positively charged be-
ers built their own device to measure the cause of adsorbed DTAB. This leads to an
force between particles and bubbles [143, electrostatic repulsion between the posi-
184–186]. With this device, the influence tively charged air–water interface and the
of different surfactants on the interaction positively charged silica surface. Above
of hydrophobic and hydrophilic particles approximately 6 mM DTAB, the jump-in
with bubbles was studied. disappeared and only repulsive force were
As one example, the force between observed.
a hydrophilic silica particle and an air The interaction of a particle with an oil
bubble at different concentrations of dode- drop is relatively similar to that with an
cyltrimethylammonium bromide (DTAB) air bubble [192–194]. Both interfaces, the
is shown in Fig. 10. Without surfactant, oil–water and the water–gas, are nega-
the particle is repelled by the air bubble. tively charged around neutral pH. There-
At distances above 5 nm, the electro- fore an electrostatic repulsion is observed
static repulsion dominates. The reason when interacting with a silica particle. As
being the negative surface charges on the expected from DLVO theory, this repulsion
silica surface and at the water–air inter- decays roughly exponentially with a decay
face [187–190]. Even at close distance, a length equal to the Debye length [195].
stable water remains on the particle sur-
face and no three-phase contact is formed. 2.5.5.4 Imaging Charge Distributions
Adding even small amounts of the cationic Besides measuring the distance depen-
surfactant DTAB changes the interac- dency of a force, it is sometimes interesting
tion drastically. At concentrations between to know how the force changes from
0.1 mM and typically 5 mM DTAB (criti- one place on the sample to another. To
cal micellar concentration is ≈16 mM), no obtain this lateral information, Senden,
repulsion was observed. When the parti- Drummond, and Kékicheff [196] imaged a
cle comes into contact with the air–water silicon nitride surface at low force. When
interface, it jumps into the bubble and a imaging at very low forces (about 0.1 nN),
three-phase contact is formed. Such a be- the tip did not touch the surface because
havior can be explained with the strong electrostatic repulsion kept the tip sev-
adsorption of long-chain alkyltrimethy- eral nanometers above the surface. The tip
lammonium ions to silica [191]. At a ‘‘rode’’ on top of the double layer. In this
concentration of 0.1 mM, DTA+ forms way a constant electrostatic force image is
a monolayer on the silica surface. This obtained. When increasing the force, the
1.5
0 mM DTAB
1.0
0.5
0.0
0.0 0.1 0.2 0.3 0.4 0.5
5
0.1 mM DTAB
0
−5
−10
−15
[mN m−1]
F/R
0 2 4 6 8
0.4
15
0.2 5.4 mM
10
0.0
5
0,00 0,05 0,10
0
−5
−10
−15
−20
0 2 4 6 8
12
10 13.2 mM DTAB
8
6
4
2
0
0.0 0.1 0.2 0.3 0.4 0.5

Distance
[nm]
Fig. 10 Normalized force-versus-distance curve measured with a hydrophilic silica
particle in aqueous electrolyte with no added DTAB, 0.1 mM DTAB, 5.4 mM DTAB, and
13.2 mM DTAB (from Ref. [143]). The electrolyte contained 0.3 mM KCl, the pH was
around 5.5. The insert at 5.4 mM shows the electrostatic repulsion before the jump-in.
The particle radius R was 2.5 µm.
tip came into contact with the sample and covered with micells or hemimicells of two
a topographic image was obtained. gemini surfactants. AFM images taken in
Using the electrostatic repulsion mode, the electrostatic repulsion mode allowed
Manne and coworkers studied the struc- to analyze the supramolecular structure of
ture of surfactants adsorbed to solid surfactants on solid surfaces and greatly
surfaces. They imaged the cationic sur- advanced our knowledge about surfactant
factant cetyltrimethylammonium bromide adsorption. The advantage over contact
(CTAB) adsorbed to a hydrophobic grap- mode is the significantly less destruc-
hite surface [197]. When imaging with tive interaction between tip and sample.
a very low force, they observed paral- Burgess and coworkers [201] could even
lel stripes spaced about 4.2 nm apart, visualize changes in the structure of SDS
that is, about twice the length of the adsorbed to gold(111) upon variation of
CTA+ ion. The adsorbed stripes were the potential. At negative and neutral sur-
generally observed in three orientations. face charge, SDS forms a hemimicellar
On the basis of observations like these, structure. At positive surface charge, a
the authors proposed that CTAB forms condensed monolayer is formed.
hemicylindrical hemimicells on graphite. In the described experiments the images
Manne and coworkers and others extended are not solely images of the charge
this work to other solids such as sil- distribution but they are a mixture of
ica [198], mica [175], or gold [199] and to topography and charge density. Heinz
other surfactants such as sodium dodecyl and Hoh [203] developed a protocol, called
sulfate (SDS) or dodecyldimethylammo- D minus D mapping, on how the two
niopropanesulfonate (DDAPS) [200]. As contributions can be separated. Therefore
an example, Fig. 11 shows mica surfaces they take isoforce images at different salt
(a) (b)
Fig. 11 To study the structure surfactants on surface, these AFM images (150 × 150 nm)
were recorded in electrostatic repulsion mode. They show two different gemini surfactants,
(Cm H2m+1 )[N+ (CH3 )2 ](CH2 )n [N+ (CH3 )2 ](CH2 )j with m-n-j of 18-3-1 (a) and 12-4-12 (b), in
aqueous medium on mica. The concentrations were 3.0 and 2.2 mM, respectively. Both are
above the critical micellar concentration. (The figure is reproduced with kind permission
from S. Manne [202].)
concentrations to remove topography and to calibrate the tip and then the charge
isolate electrostatic contributions to the density of the deposited material can eas-
tip–sample interaction. ily be calculated. In this way the author
To get more quantitative information on determined the charge density of pur-
the surface charge density, whole force ple membranes deposited on alumina [54].
curves must be taken at each individ- Rotsch and coworkers verified the positive
ual point of the sample. In this way, surface charge of dimethyldioctadecylam-
differences between the charge densities monium bromide (DODAB) patches on
of different regions on a sample can be negatively charged mica (Fig. 12). At neu-
obtained. This not only allows to distin- tral pH they observed a repulsive force
guish regions on the basis of their electric between the negatively charged mica and
properties but also opens the possibility to the negatively charged silicon nitride tip.
measure quantitative charge densities with On a DODAB patch an electrostatic at-
a standard AFM tip. Usually, it is practi- traction was observed, indicating that the
cally impossible to measure local charge patch is positively charged.
densities with standard tips directly. The Taking many force curves can nowadays
surface chemistry of silicon nitride and the be done automatically with commercial
size of the tip are not known accurately instruments. Usually it requires 0.1 to 1 s
enough to allow quantitative measure- to take one force curve. To obtain say
ments. To circumvent this problem, the 64 × 64 force curves lasts 7 to 70 min.
sample is adsorbed on a substrate with To increase the speed, Miyatani and
known charge density. Then, force curves coworkers developed a procedure in which
are taken on the substrate and on the de- the cantilever is moved in a sinusoidal
posited material. This substrate is used form rather than a triangular pulse [205].
30 1 µm
20
Cantilever deflection
10
[nm]
−10
2 µm
−20
−20 0 20 40 60
Translator position
(a) (b) [nm]
Fig. 12 AFM contact mode image showing a typical assembly of DODAB bilayer patches (a).
The patches were self-assembled from a vesicle suspension and imaged in pure water. To the
right, a series of approaching force curves are shown that were taken along a line across one of
the patches in 5 mM LiNO3 (b). The different electrostatic interaction – repulsive on the mica
substrate at the beginning and end of the line and attractive on the DODAB patch – is visible.
(The figure is reproduced with kind permission from M. Radmacher [204].)
This method is, however, limited by issues, the AFM has become a routine tool
the increasing viscous drag when the to measure interaction force for specific
cantilever approaches a sample [134]. particles or surfaces relevant to practical
applications.
2.5.6
Conclusion and Outlook References
Since the first measurements of the elec- 1. G. Binnig, C. F. Quate, C. Gerber, Phys.
trostatic double-layer force with the AFM Rev. Lett. 1986, 56, 930–933.
2. A. L. Weisenhorn, P. K. Hansma, T. R.
not even 10 years ago, the instrument has Albrecht et al., Appl. Phys. Lett. 1989, 54,
become a versatile tool to measure sur- 2651–2653.
face forces in aqueous electrolyte. Force 3. E. Meyer, H. Heinzelmann, P. Grütter
measurements with the AFM confirmed et al., Thin Solid Films 1989, 181, 527–544.
that with continuum theory based on the 4. A. L. Weisenhorn, P. Maivald, H.-J. Butt
et al., Phys. Rev. B 1992, 45, 11 226–11 232.
Poisson–Boltzmann equation and applied 5. J. H. Hoh, J. P. Revel, P. K. Hansma, Nan-
by Debye, Hückel, Gouy, and Chapman, otechnology 1991, 2, 119–122.
the electrostatic double layer can be ade- 6. S. Manne, H. E. Gaub, Curr. Opin. Colloid
quately described for distances larger than Interface Sci. 1997, 2, 145–152.
7. B. Cappella, G. Dietler, Surf. Sci. Rep. 1999,
1 to 5 nm. It is valid for all materials in-
34, 1–104.
vestigated so far without exception. It also 8. H.-J. Butt, M. Jaschke, W. Ducker, Bioelec-
holds for deformable interfaces such as trochem. Bioenerg. 1995, 38, 191–201.
the air–water interface and the interface 9. W. A. Ducker, T. J. Senden, R. M. Pashley,
between two immiscible liquids. Even the Nature 1991, 353, 239–241.
10. H.-J. Butt, Biophys. J. 1991, 60, 1438–1444.
behavior at high surface potentials seems
11. H. Freundlich, Z. Phys. Chem. 1910, 73,
to be described by continuum theory, al- 385–423.
though some questions still have to be 12. B. Derjaguin, L. Landau, Acta Physicochim.
clarified. For close distances, often the URSS 1941, 14, 633–662.
hydration force between hydrophilic sur- 13. E. J. W. Verwey, J. T. G. Overbeek, Theory of
the Stability of Lyophobic Colloids, Elsevier
faces influences the interaction. Between Publishing, New York, 1948.
hydrophobic surfaces with contact angles 14. G. Gouy, J. Phys. (Paris) 1910, 9, 457–468.
above 80◦ , often the hydrophobic attraction 15. D. L. Chapman, Philos. Mag. (London) 1913,
dominates the total force. 25, 475–481.
16. P. Debye, E. Hückel, Phys. Z. 1923, 24,
In the future, more fundamental investi-
185–206.
gations are probably going to focus on the 17. J. Mahanty, B. W. Ninham, Dispersion
interaction at close distances where only Forces, Academic Press, New York, 1976.
few molecular layers are present between 18. J. N. Israelachvili, Intermolecular and Sur-
the interacting surfaces. Then, however, face Forces, 2nd ed., Academic Press, Lon-
don, 1992.
the separation into electrostatic, van der
19. R. J. Hunter, Foundations of Colloid Science
Waals, and hydration forces might not be I+II, Clarendon Press, Oxford, 1995.
useful anymore. Also the investigation of 20. I. E. Dzyaloshinskii, E. M. Lifshitz, L. P.
shear forces (friction) or the change of Pitaevskii, Adv. Phys. 1961, 10, 165–209.
interaction under shear is a promising 21. P. M. Claesson, T. Ederth, V. Bergeron
et al., Adv. Colloid Interface Sci. 1996, 67,
area of research and many important fun- 119–183.
damental and applied questions can be 22. N. V. Churaev, Adv. Colloid Interface Sci.
addressed. Apart from these fundamental 1999, 83, 19–32.
23. B. V. Derjaguin, A. S. Titijevskaia, I. I. Abri- 48. K. L. Johnson, K. Kendall, A. D. Roberts,

cossova et al., Discuss. Faraday Soc. 1954, 18, Proc. R. Soc. London, A 1971, 324, 301–313.
24–41. 49. K. L. Johnson, Contact Mechanics, Cam-
24. J. T. G. Overbeek, M. J. Sparnaay, Discuss. bridge University Press, Cambridge, 1985.
Faraday Soc. 1954, 18, 12–24. 50. L. E. McNeil, M. Grimsditch, J. Phys.: Con-
25. W. Black, J. G. V. de Jongh, J. T. G. Over- dens. Matter 1993, 5, 1681–1690.
beek et al., Trans. Faraday Soc. 1960, 56, 51. T. J. Senden, C. J. Drummond, Colloids
1597–1608. Surf., A 1994, 94, 29–52.
26. G. C. J. Rouweler, J. T. G. Oberbeek, Trans. 52. B. V. Zhmud, A. Meurk, L. Bergström, J.
Faraday Soc. 1971, 67, 2117–2121. Colloid Interface Sci. 1998, 207, 332–343.
27. B. V. Derjaguin, Y. I. Rabinovich, N. V. 53. X. Y. Lin, F. Creuzet, H. Arribart, J. Phys.
Churaev, Nature 1978, 272, 313–318. Chem. 1993, 97, 7272–7276.
28. R. G. Horn, D. T. Smith, J. Non-Cryst. Solids 54. H.-J. Butt, Biophys. J. 1992, 63, 578–582.
1990, 120, 72–81. 55. A. Döppenschmidt, H.-J. Butt, Colloids Surf.,
29. J. Keldenich, G. Peschel, Colloid Polym. Sci. A 1999, 149, 145–150.
1991, 269, 1064–1070. 56. P. Weidenhammer, H. J. Jacobasch, J. Col-
30. D. Tabor, R. H. S. Winterton, Nature 1968, loid Interface Sci. 1996, 180, 232–236.
219, 1120–1121. 57. T. Ishino, H. Hieda, K. Tanaka et al., Jpn. J.
31. J. N. Israelachvili, D. Tabor, Proc. R. Soc. Appl. Phys. 1994, 33, 4718–4722.
London, A 1972, 331, 19–38. 58. T. Arai, D. Aoki, Y. Okabe et al., Thin Solid
32. J. N. Israelachvili, G. E. Adams, J. Chem. Films 1996, 273, 322–326.
Soc., Faraday Trans. 1 1978, 74, 975–1001. 59. Y. F. Dufrene, T. Boland, J. W. Schneider
33. S. A. Joyce, J. E. Houston, Rev. Sci. Instrum. et al., Faraday Discuss. 1998, 111, 79–94.
60. H. Mueller, H.-J. Butt, E. Bamberg, J. Phys.
1991, 62, 710–715.
Chem. 2000, in press.
34. R. M. Pashley, J. Colloid Interface Sci. 1981,
61. D. J. Müller, A. Engel, Biophys. J. 1997, 73,
83, 531–546.
1633–1644.
35. G. Cevc, J. Chem. Soc., Faraday Trans. 1991,
62. D. J. Müller, D. Fotiadis, S. Scheuring et al.,
87, 2733–2739.
Biophys. J. 1999, 76, 1101–1111.
36. S. Leikin, V. A. Parsegian, D. C. Rau et al.,
63. R. F. Considine, D. R. Dixon, C. J. Drum-
Annu. Rev. Phys. Chem. 1993, 44, 369–395.
mond, Langmuir 2000, 16, 1323–1330.
37. L. Fischer, J. Chem. Soc., Faraday Trans. 64. W. A. Ducker, T. J. Senden, R. M. Pashley,
1993, 89, 2567–2582. Langmuir 1992, 8, 1831–1836.
38. J. Israelachvili, H. Wennerström, Nature 65. L. Meagher, J. Colloid Interface Sci. 1992,
1996, 379, 219–225. 152, 293–295.
39. D. M. LeNeveu, R. P. Rand, V. A. Parsegian, 66. M. W. Rutland, T. J. Senden, Langmuir 1993,
Nature 1976, 259, 601–603. 9, 412–418.
40. V. A. Parsegian, R. P. Rand, N. L. Fuller 67. G. Hüttl, D. Beyer, E. Müller, Surf. Interface
et al., Methods Enzymol. 1986, 127, 400–416. Anal. 1997, 25, 543–547.
41. V. A. Parsegian, N. Fuller, R. P. Rand, Proc. 68. G. Toikka, R. A. Hayes, J. Colloid Interface
Natl. Acad. Sci. U.S.A. 1979, 76, 2750–2754. Sci. 1997, 191, 102–109.
42. J. P. Chapel, Langmuir 1994, 10, 4237–4243. 69. B. V. Zhmud, A. Meurk, L. Bergström, Col-
43. A. Grabbe, R. G. Horn, J. Colloid Interface loids Surf., A 2000, 164, 3–7.
Sci. 1993, 157, 375–383. 70. Y. Kanda, T. Nakamura, K. Higashitani,
44. J. L. Parker, Langmuir 1992, 8, 551–556. Colloid Surf., A 1998, 139, 55–62.
45. J. L. Parker, Prog. Surf. Sci. 1994, 47, 71. S. Veeramasuneni, M. R. Yalamanchili,
205–271. J. D. Miller, J. Colloid Interface Sci. 1996,
46. H. K. Christenson, The long-range attrac- 184, 594–600.
tion between macroscopic hydrophobic sur- 72. I. Larson, C. J. Drummond, D. Y. C. Chan
faces, in Modern Approaches to Wettability et al., Langmuir 1997, 13, 2109–2112.
(Eds.: M. E. Schrader, G. Loeb), Plenum 73. I. Larson, C. J. Drummond, D. Y. C. Chan
Press, New York, 1992, pp. 29–51. et al., J. Phys. Chem. 1995, 99, 2114–2118.
47. J. Israelachvili, R. Pashley, Nature 1982, 74. K. Hu, F. R. F. Fan, A. J. Bard et al., J. Phys.
300, 341–343. Chem. B 1997, 101, 8298–8303.
75. S. D. Campbell, A. C. Hillier, Langmuir 102. C. A. J. Putman, B. G. De Grooth, N. F. Van

1999, 15, 891–899. Hulst et al., J. Appl. Phys. 1992, 72, 6–12.
76. L. Meagher, R. M. Pashley, Langmuir 1995, 103. H.-J. Butt, M. Jaschke, Nanotechnology 1995,
11, 4019–4024. 6, 1–7.
77. L. Meagher, R. M. Pashley, J. Colloid Inter- 104. T. R. Albrecht, S. Akamine, T. E. Carver
face Sci. 1997, 185, 291–296. et al., J. Vac. Sci. Technol. 1990, A8,
78. W. R. Bowen, N. Hilal, M. Jain et al., Chem. 3386–3390.
Eng. Sci. 1999, 54, 369–375. 105. H.-J. Butt, P. Siedle, K. Seifert et al., J.
79. R. H. Yoon, S. Vivek, J. Colloid Interface Sci. Microsc. 1993, 169, 75–84.
1998, 204, 179–186. 106. G. Y. Chen, R. J. Warmack, T. Thundat et al.,
80. M. Mizukami, M. Kurihara, Prog. Colloid Rev. Sci. Instrum. 1994, 65, 2532–2537.
Polym. Sci. 1997, 106, 266–269. 107. F. J. Elmer, M. Dreier, J. Appl. Phys. 1997,
81. R. Ishiguro, D. Y. Sasaki, C. Pacheco et al., 81, 7709–7714.
Colloids Surf., A 1999, 146, 329–335. 108. T. Thundat, R. J. Warmack, G. Y. Chen
82. I. Larson, C. J. Drummond, D. Y. C. Chan et al., Appl. Phys. Lett. 1994, 64, 2894–2896.
et al., J. Am. Chem. Soc. 1993, 115, 109. H.-J. Butt, J. Colloid Interface Sci. 1996, 180,
11 885–11 890. 251–260.
83. H. G. Pedersen, L. Bergström, J. Am. Ce- 110. J. E. Sader, L. White, J. Appl. Phys. 1993, 74,
ram. Soc. 1999, 82, 1137–1145. 1–9.
84. M. Prica, S. Biggs, F. Grieser et al., Colloids 111. J. E. Sader, Rev. Sci. Instrum. 1995, 66,
Surf., A 1996, 119, 205–213. 4583–4587.
85. S. Biggs, Langmuir 1995, 11, 156–162. 112. J. N. Neumeister, W. A. Ducker, Rev. Sci.
86. D. T. Atkins, R. M. Pashley, Langmuir 1993, Instrum. 1994, 65, 2527.
9, 2232–2236.
113. M. Sasaki, K. Hane, S. Okuma et al., Rev.
87. G. Toikka, R. A. Hayes, J. Ralston, J. Chem.
Sci. Instrum. 1994, 65, 1930–1934.
Soc., Faraday. Trans. 1997, 93, 3523–3528.
114. Y. I. Rabinovich, R. H. Yoon, Colloids Surf.,
88. G. Toikka, R. A. Hayes, J. Ralston, Lang-
A 1994, 93, 263–273.
muir 1996, 12, 3783–3788.
115. T. A. Senden, W. A. Ducker, Langmuir 1994,
89. S. Biggs, M. K. Chow, C. F. Zukoski et al.,
10, 1003–1004.
J. Interface Colloid Sci. 1993, 160, 511–513.
116. J. P. Cleveland, S. Manne, D. Bocek et al.,
90. S. Biggs, P. Mulvaney, C. F. Zukoski et al.,
J. Am. Chem. Soc. 1994, 116, 9150–9157. Rev. Sci. Instrum. 1993, 64, 403–405.
91. I. Larson, D. Y. C. Chan, C. J. Drummond 117. A. Torii, M. Sasaki, K. Hane et al., Meas. Sci.
et al., Langmuir 1997, 13, 2429–2431. Technol. 1996, 7, 179–184.
92. S. Biggs, P. Mulvaney, J. Chem. Phys. 1994, 118. J. E. Sader, J. W. M. Chou, P. Mulvaney,
100, 8501–8505. Rev. Sci. Instrum. 1999, 70, 3967–3969.
93. R. Raiteri, M. Preuss, M. Grattarola et al., 119. J. L. Hutter, J. Bechhoefer, Rev. Sci. In-
Colloids Surf., A 1998, 136, 191–197. strum. 1993, 64, 1868–1873.
94. K. Hu, A. J. Bard, Langmuir 1997, 13, 120. P. Siedle, H.-J. Butt, E. Bamberget et al.,
5114–5119. Inst. Phys. Conf. Ser. 1992, 130, 361–364.
95. D. V. Vezenov, A. Noy, L. F. Rozsnyai et al., 121. S. Xu, M. F. Arnsdorf, J. Microsc. 1994, 173,
J. Am. Chem. Soc. 1997, 119, 2006–2015. 199–210.
96. A. Noy, D. V. Vezenov, C. M. Lieber, Annu. 122. J. A. DeRose, J. P. Revel, Microsc. Microanal.
Rev. Mater. Sci. 1997, 27, 381–421. 1997, 3, 203–213.
97. M. E. Karaman, L. Meagher, R. M. Pashley, 123. M. Grattarola, G. Massobrio, S. Martinoia,
Langmuir 1993, 9, 1220–1227. IEEE Trans. Electron Devices 1992, 39,
98. Y. Q. Li, N. J. Tao, J. Pan et al., Langmuir 813–819.
1993, 9, 637–641. 124. D. L. Harame, L. J. Bousse, J. D. Shott et al.,
99. W. R. Bowen, N. Hilal, R. W. Lovit et al., J. IEEE Trans. Electron Devices 1987, 34,
Colloid Interface Sci. 1998, 197, 348–352. 1700–1707.
100. H. Mueller, H.-J. Butt, E. Bamberg, Bio- 125. H.-J. Butt, T. Müller, H. Gross, J. Struct.
phys. J. 1999, 76, 1072–1079. Biol. 1993, 110, 127–132.
101. Y. L. Ong, A. Razatos, G. Georgiou et al., 126. M. Hegner, P. Wagner, G. Semenza, Surf.
Langmuir 1999, 15, 2719–2725. Sci. 1993, 291, 39–46.
127. N. A. Burnham, R. J. Colton, H. M. Pollock, 155. E. Barouch, E. Matijevic, J. Chem. Soc.,

J. Vac. Sci. Technol., A 1991, 9, 2548–2556. Faraday Trans. 1 1985, 81, 1797–1817.
128. G. L. Miller, J. E. Griffith, E. R. Wagner, 156. D. E. Yates, S. Levine, T. W. Healy, J. Chem.
Rev. Sci. Instrum. 1991, 705. Soc., Faraday Trans. 1 1974, 70, 1807–1818.
129. P. J. Bryant, H. S. Kim, R. H. Decken et al., 157. P. M. Claesson, P. Herder, P. Stenius et al.,
J. Vac. Sci. Technol., A 1990, 8, 2502. J. Colloid Interface Sci. 1986, 109, 31–39.
130. D. A. Grigg, P. E. Russel, J. E. Griffith, Ul- 158. P. J. Scales, F. Grieser, T. W. Healy, Lang-
tramicroscopy 1992, 42, 1504. muir 1990, 6, 582–589.
131. J. Mertz, O. Marti, J. Mlynek, Appl. Phys. 159. S. Nishimura, H. Tateyama, K. Tsunematsu
Lett. 1993, 62, 2344–2346. et al., J. Colloid Interface Sci. 1992, 152,
132. H.-J. Butt, J. Colloid Interface Sci. 1994, 166, 359–367.
109–117. 160. B. W. Ninham, V. Yaminsky, Langmuir
133. D. Y. C. Chan, R. G. Horn, J. Chem. Phys. 1997, 13, 2097–2108.
1985, 83, 5311–5324. 161. P. Johansson, P. Apell, Phys. Rev. B 1997,
134. O. I. Vinogradova, H.-J. Butt, G. Yakubov 56, 4159–4165.
et al., Rev. Sci. Instrum. 2001, 72, 2330–2339. 162. C. Argento, R. H. French, J. Appl. Phys.
135. J. B. Pethica, A. P. Sutton, J. Vac. Sci. 1996, 80, 6081–6090.
Technol., A 1988, 6, 2490–2494. 163. L. Bergström, Adv. Colloid Interface Sci.
136. H.-J. Butt, A. Döppenschmidt, G. Hüttl 1997, 70, 125–169.
et al., J. Chem. Phys. 2000, 113, 1194–1203. 164. H. D. Ackler, R. H. French, Y. M. Chiang,
137. G. J. C. Braithwaite, A. Howe, P. F. Luck- J. Colloid Interface Sci. 1996, 179, 460–469.
165. A. Milling, P. Mulvaney, I. Larson, J. Colloid
ham, Langmuir 1996, 12, 4224–4237.
Interface Sci. 1996, 180, 460–465.
138. V. S. J. Craig, A. M. Hyde, R. M. Pashley,
166. A. Meurk, P. W. Luckham, L. Bergström,
Langmuir 1996, 12, 3557–3562.
Langmuir 1997, 13, 3896–3899.
139. R. M. Pashley, M. E. Karaman, V. S. J. Craig
167. J. N. Israelachvili, R. M. Pashley, Nature
et al., Colloids Surf., A 1998, 144, 1–8.
1983, 306, 249–250.
140. N. P. Costa, J. H. Hoh, Rev. Sci. Instrum.
168. R. H. Yoon, D. H. Flinn, Y. I. Rabinovich, J.
1995, 66, 5096–5097.
Colloid Interface Sci. 1997, 185, 363–370.
141. M. Pierce, J. Stuart, A. Pungor et al., Lang-
169. L. Meagher, V. S. J. Craig, Langmuir 1994,
muir 1994, 10, 3217–3221. 10, 2736–2742.
142. S. M. Hues, C. F. Draper, K. P. Lee et al., 170. J. Nalaskowski, S. Veeramasuneni, J. D.
Rev. Sci. Instrum. 1994, 65, 1561–1565. Miller, Interaction forces in the flotation of
143. M. Preuss, H.-J. Butt, Langmuir 1998, 14, colemanite as measured by atomic force mi-
3164–3174. croscopy, in Innovations in Mineral and Coal
144. B. Derjaguin, Kolloid Z. 1934, 69, 155–164. Processing (Eds.: S. Atak, G. Önal, S. Celik),
145. D. Y. C. Chan, D. J. Mitchell, J. Colloid In- Brookfield, Rotterdam, 1998, pp. 159–165.
terface Sci. 1983, 95, 193–197. 171. S. Nishimura, M. Kodama, H. Noma et al.,
146. V. A. Parsegian, D. Gingell, Biophys. J. Colloids Surf., A 1998, 143, 1–16.
1972, 12, 1192–1204. 172. V. Kane, P. Mulvaney, Langmuir 1998, 14,
147. R. Hogg, T. W. Healy, D. W. Fuerstenau, 3303–3311.
Trans. Faraday Soc. 1966, 62, 1638–1651. 173. C. J. Drummond, T. J. Senden, Colloids
148. H.-J. Butt, Biophys. J. 1991, 60, 777–785. Surf., A 1994, 87, 217–234.
149. H.-J. Butt, Nanotechnology 1992, 3, 60–68. 174. M. E. Karaman, R. M. Pashley, N. K. Bolon-
150. S. L. Carnie, D. Y. C. Chan, J. S. Gunning, kin, Langmuir 1995, 11, 2872–2880.
Langmuir 1994, 10, 2993–3009. 175. W. A. Ducker, E. J. Wanless, Langmuir 1999,
151. H. Ohshima, J. Colloid Interface Sci. 1994, 15, 160–168.
162, 487–495. 176. S. Veeramasuneni, M. R. Yalamanchili,
152. J. Stankovich, S. L. Carnie, Langmuir 1996, J. D. Miller, Colloids Surf., A 1998, 131,
1996, 1453–1461. 77–87.
153. A. V. Nguyen, G. M. Evans, G. J. Jameson, 177. T. Ishino, H. Hieda, K. Tanaka et al., Jpn. J.
J. Colloid Interface Sci. 2000, 230, 208–212. Appl. Phys. 1994, 33, L1552–L1554.
154. V. G. Levadny, M. L. Belaya, D. A. Pink 178. R. Raiteri, M. Grattarola, H.-J. Butt, J. Phys.
et al., Biophys. J. 1996, 70, 1745–1752. Chem. 1996, 100, 16 700–16 705.
179. A. C. Hillier, S. Kim, A. J. Bard, J. Phys. 193. B. A. Snyder, D. E. Aston, J. C. Berg, Lang-
Chem. 1996, 100, 18 808–18 817. muir 1997, 13, 590–593.
180. S. J. Miklavcic, R. G. Horn, D. J. Bachmann, 194. G. E. Yakubov, O. I. Vinogradova, H.-J.
J. Phys. Chem. 1995, 99, 16 357–16 364. Butt, J. Adhes. 2000, 14, 1783–1799.
181. D. J. Bachmann, S. J. Miklavcic, Langmuir 195. P. G. Hartley, F. Grieser, P. Mulvaney et al.,
1996, 12, 4197–4204. Langmuir 1999, 15, 7282–7289.
182. W. A. Ducker, Z. Xu, J. N. Israelachvili, 196. T. J. Senden, C. J. Drummond, P. Kékicheff,
Langmuir 1994, 10, 3279–3289. Langmuir 1994, 358–362.
183. M. L. Fielden, R. A. Hayes, J. Ralston, Lang- 197. S. Manne, J. P. Cleveland, H. E. Gaub et al.,
muir 1996, 12, 3721–3727. Langmuir 1994, 10, 4409–4413.
184. M. Preuss, H.-J. Butt, Int. J. Miner. Process 198. S. Manne, H. E. Gaub, Science 1995, 270,
1999, 56, 99–115. 1480–1482.
185. M. Preuss, H.-J. Butt, J. Colloid Interface Sci. 199. M. Jaschke, H.-J. Butt, H. E. Gaub et al.,
1998, 208, 468–477. Langmuir 1997, 13, 1381–1384.
186. S. Ecke, M. Preuss, H.-J. Butt, J. Adhes. Sci. 200. W. A. Ducker, L. M. Grant, J. Phys. Chem.
Technol. 1999, 13, 1181–1191. 1996, 100, 11 507–11 511.
187. S. Usui, H. Sasaki, H. Matsukawa, J. Colloid 201. I. Burgess, C. A. Jeffrey, X. Cai et al., Lang-
Interface Sci. 1981, 81, 80–84. muir 1999, 15, 2607–2616.
188. R. H. Yoon, J. L. Yordan, J. Colloid Interface 202. S. Manne, T. E. Schäffer, Q. Huo et al.,
Sci. 1986, 113, 430–438. Langmuir 1997, 13, 6382–6387.
189. P. Saulnier, J. Lachaise, G. Morel et al., J. 203. W. F. Heinz, J. H. Hoh, Biophys. J. 1999, 76,
Colloid Interface Sci. 1996, 182, 395–399. 528–538.
190. H. J. Schulze, C. Cichos, Z. Phys. Chem. 204. C. Rotsch, M. Radmacher, Langmuir 1997,
(Leipzig) 1972, 251, 252–268. 13, 2825–2832.
191. L. G. T. Eriksson, P. M. Claesson, J. C. Erik- 205. T. Miyatani, M. Horii, A. Rosa et al., Appl.
sson et al., J. Colloid Interface Sci. 1996, 181, Phys. Lett. 1997, 71, 2632–2634.
476–489.
192. P. Mulvaney, J. M. Perera, S. Biggs et al., J.
Colloid Interface Sci. 1996, 183, 614–616.
2.6 Comparison Between Liquid State and Solid State Electrochemistry 253
2.6 example is AgI (at T > 149 ◦ C), where only

Comparison Between Liquid State and Solid Ag+ ions are mobile. (3) In liquid-state
State Electrochemistry cells one has the option to stir and replen-
ish the LE. This is not possible, of course,
Ilan Riess when using SEs. Contrary to SSE in which
Technion-Israel Institute of Technology, Haifa,
the electrode can be complex, in LSE it is
Israel
usually the LE that has a complicated com-
position, containing many different kinds
Abstract
of ions coming from dissolution of dif-
ferent compounds. These differences and
We discuss the similarities and differ-
others will be discussed in more detail in
ences between liquid-state electrochem-
the coming sections. In Sect. 2 we discuss
istry (LSE) and solid-state electrochemistry
the geometry of liquid and solid ECs; in
(SSE). Although based on the same ther- Sect. 3 material properties, in particular,
modynamic principles, the properties of charge carriers, their nature, concentra-
these cells are quite distinct. Differences tion and conduction mechanism; in Sect. 4
exist in the bulk conduction mechanism, the theoretical current-voltage relations in
partially in electrode reaction and in cell the bulk and the relevant boundary con-
construction and morphology. This also ditions. In Sect. 5 experimental methods
leads to differences in applications. for characterizing materials properties and
overpotentials are discussed, and in Sect. 6
2.6.1 electrode processes. Two short sections
Introduction follow, Sect. 7 in which the use of the Gal-
vani potential is considered, and Sect. 8
LSE, the classical electrochemistry, is con- in which catalysis is briefly considered. A
cerned with electrochemical cells (ECs) summary is given in Sect. 9.
based on liquid ionic-conductors (liquid LSE is a well-established field, for which
electrolytes (LEs)). Solid-state electrochem- there are many good, comprehensive text
istry is concerned with ECs in which the books and we mention only four: Bock-
ionic conductor (electrolyte) is a solid. Both ris and Ready [1], Bard and Faulkner [2],
fields are based on common thermody- Oldham and Myland [3] and Gileadi [4]. In
namic principles. Yet, the finer character- SSE, being a younger field, the informa-
istics of ECs in the two fields are different tion is scattered in many review papers on
because of differences in the materials different topics, books containing review
properties, conduction mechanisms, mor- chapters [5–7], books dedicated to par-
phology and cell geometry. Differences ticular topics: on solid-state batteries [8],
that come immediately to mind are: (1) solid-oxide fuel cells [9], and solid-state
The lack of electronic (electron/hole) con- sensors [10]. A text book that summarizes
duction in most LEs, while electronic the state of the art up to year 1981 is the
conduction exists to some extent in all solid book by Rickert [11]. A recent book (in Ger-
electrolytes (SEs). (2) In LEs both cations man) was published by Maier. It presents
and anions are mobile, while in SEs only a broad overview [12]. Chemical reactions
one kind of ions is usually mobile while the in ionic solids are closely related to SSE as
other forms a rigid sublattice serving as a they are based on interdiffusion of ions and
frame for the motion of the mobile ion. An sometimes also on charge transfer. The
driving forces are similar. The reactions are then chemically inert, serving only
can be driven by both chemical potential for the supply and removal of electrons.
differences and electrochemical potential Furthermore, the LE tolerates the addition
differences. A recent book on solid-state of other materials that contribute mobile
reaction is by Schmalzried [13]. Properties ions which, however, do not participate in
of many relevant ionic materials as well the electrical current under steady state
as an excellent summary of the theory of conditions. The main use of this feature
point defect in solids can be found in the is in adding a supporting electrolyte, the
book by Kröger [14]. An outstanding con- importance of which will be discussed in
tribution to the field up to the year 1977 the theoretical Sect. 4.
was made by Wagner, whose results are In other arrangements, the electrodes
quoted in the reviews and books written provide material or serve as a sink.
later and mentioned in the preceding text. An electrode that serves as a source of
material is, for example, a metal anode that
2.6.2 dissolves and liberates cations into the LE.
Electrochemical Cell Geometry A metal electrode serves as a sink when it
is oxidized to form, for instance, an oxide
2.6.2.1 Cells Based on LEs layer or when electroplating occurs. This
An EC based on a LE is shown schemati- is also the case when insertion takes place
cally in Fig.1. It is composed of a liquid that as hydrogen into a palladium electrode.
conducts ions and two current-carrying Material can be exchanged also with a gas
electrodes, an anode and a cathode (E1 , phase. This is the case when bubbling
E2 ). An electrode can be composed of more hydrogen on a Pt electrode.
than one piece of one kind of metal as A key limitation of the existing LE ECs
when using a sacrificial anode. Additional is that the operation temperature must
electrodes (Eref ) that do not carry currents be kept low, compatible with the vapor
are added in most cases to serve as refer- pressure of the liquid. (An exception is the
ence electrodes in voltage measurements use of a few high-temperature, molten salt
and analysis of cell operation. LEs, for example, molten carbonates) [1,
The LE serves not only as a medium for 15, 16]. While low-operating temperature
selective transport of ionic charge but also, has obvious advantages, it may also
in many cases, as the source and sink for introduce limitations because of two very
the reactants and products involved in the different reasons. (1) high temperatures
electrochemical reaction. The electrodes allow to overcome electrode overpotentials,
E1 E ref E2
Fig. 1 Schematics of a
liquid-state electrochemical cell.
LE – liquid electrolyte; E1 ,
E2 – current-carrying electrodes;
Eref – reference electrode.
and (2) the heat generated as a by-product and the electrodes, E1 , E2 , being solids.
of the electrochemical reaction in hot These electrodes serve both as source and
cells can be used for heating much sink for electrons and for material. An ex-
more efficiently. This is the case in high- ample is the battery Lix C|SE(Li+ )|Liy TiS2 ,
temperature solid-oxide fuel cells in which where the anode is graphite intercalated
the hot exhaust gases can be used for with Li, and the cathode is TiS2 , a lay-
cogeneration of electricity by gas turbine ered compound that allows intercalation
driven electrical generators. A second key of Li into it, serving both as a conduc-
limitation of LE-based ECs is that they tor and a source/sink of Li+ ions and
cannot be used in thin-film devices and electrons. The SE conducts Li+ ions, and
miniature cells in microelectronics, while can be, for example, a glass made of
solid-state cells are being developed for Li2 O−Li2 SO4 −B2 O3 [8]. In this case, the
these applications. electrodes materials serve for the supply
and removal of both charge and matter.
2.6.2.2 Cells Based on SEs They must allow the diffusion (intercala-
Solid electrochemical cells are more re- tion) of the material into them and they
stricted in the structure because of the solid must also be electronic (electron/hole)
nature of the ionic conductor. Schematic conductors. The SE serves solely as a
configurations of solid-state cells are pre- membrane that facilitates the selective con-
sented in Fig. 2(a–c). Figure 2(a) shows duction of certain ions (e.g. Li+ ), but not
an all-solid cell with the SE, SE(M+ ), that of electron/holes and of other ions
E1 SE(M +) E2
(a)
Inert metal Inert metal

anode cathode
SE Inert metal
P(X2)low P(X2)high
(−) SE(X−) (+)
TPB
(b)
(c)
MIEC anode MIEC cathode
P(X2)low P(X2)high
(−) (+)
SE(X−)
(d)
Fig. 2 Schematics of solid-state electrochemical cells. (a) Cell with two, solid, current-carrying
electrodes E1 and E2 , SE(M+ ) – solid electrolyte conducting the cation M+ . (b) Cell with
electrodes composed of a porous metal layer for electronic charge exchange and the gas phase,
X2 , for material exchange. PX2 –partial pressure of the gas X2 . SE(X− ) – solid electrolyte
conducting the anion X− . (c) The triple phase boundary (TPB) enlarged. (d) Cell with
mixed – ionic – electronic conducting electrodes for supply of electronic charge permitting
diffusion of ions through them, enabling material exchange with the gas phase.
that may come from the electrodes. SEs the rates of the different and competing
usually do not tolerate the addition of a elementary steps in the electrode reaction.
second type of mobile ions to serve as The geometry of this cell is again different
supporting electrolytes. All this is quite from that of a LE-based cell.
different from the case of LSE. There are A modification of the cell shown in
additional differences. Ions have to cross Fig. 2(b) is presented in Fig. 2(d) in
the SE/electrode interface. The mobility of which nonporous MIECs are used as
ions in SEs is, for most cases, lower than electrodes. In a fuel cell, oxygen has then
in LE, and is realized usually only at el- to diffuse from the gas phase through
evated temperatures. Because of the low the MIEC cathode to reach the SE and,
mobility, SEs contain high concentrations correspondingly, the fuel and the exhaust
(>1 atomic%) of mobile ions. The highest gas have to diffuse through the anode. (In
conductivity obtained at elevated T reaches reality, the MIEC electrodes are porous,
∼1S/cm for some materials (e.g. for Ag+ which increases the electrode active area).
in AgI at T > 149 ◦ C and for O2− in Bi2 O3 The MIEC can be either a semiconductor
at ∼800 ◦ C [11]). However, for most SEs or a metal. An example of a continuous
the conductivity is at least one order of cathode is: YSZ|Ag|air, with the Ag layer
magnitude lower. as the MIEC electrode. This is possible
Figure 2(b) exhibits a solid electrochem- because silver is permeable to oxygen at
ical cell with porous, inert electrodes in elevated temperature [17, 18].
contact with a gas phase at each elec-
trode that interacts with the electrode (by 2.6.2.3 Comparing the Two Groups of
electron charge transfer) and with the Cells
SE. An example is the fuel cell arrange- In LSE salt bridges are used to connect elec-
ment, H2 ,H2 O,Pt|YSZ|Pt,O2 . The SE yt- trically (by selective ion motion) but to sep-
tria stabilized zirconia (YSZ) (the oxide: arate physically and to a good approxima-
(Y2 O3 )x (ZrO2 )1−x , x ∼ 0.1) is an oxygen- tion also chemically, two compartments.
ion conductor. One electrode is exposed Owing to ion leaks, these are useful only
to a high concentration of oxygen (pure for low-current measurements, in partic-
oxygen or air) and the other to a fuel ular, open circuit voltage measurements.
such as H2 . (In modern fuel cells, Pt for In SSE, salt bridges are not needed. SEs
the electrodes is replaced by other metals, usually conduct only one kind of ions. Let
mixed ionic electronic conductor (MIECs) us consider the Daniel cell Cu|LE(Cu2+ ,
or cermets [9].) In the cell mentioned here, SO4 2− )|Bridge|LE(Zn2+ ,SO4 2− )|Zn. The
each electrode is a mixture of phases, one bridge is required to stop mixing of the
(Pt) supplying the electronic charge and two solutions. It allows a slow passage of
the second one (the gas phase) serving as ions, thus facilitating a continuous electri-
the source or sink of material. The active cal circuit. If the bride would be absent,
electrode area is around the triple phase this would destroy the cell as mixing
boundary (TPB) in which the three phases: would occur followed by electroplating of
gas, metal, and SE involved in the electrode Cu onto the Zn electrode. On the other
reaction, meet (see Fig. 2c). The length of hand, in SSE, if one would build a Cu,
the TPB is the length of the pore edges in Zn cell, one would use a SE that con-
the porous metal. The width of this area ducts only one kind of ions, say Cl− .
varies from case to case and depends on This could be SrCl2 [19]. The cell would
then be: CuCl2 ,Cu|SrCl2 |Zn,ZnCl2 . This be achieved by supplying a liquid to an

cell would function without the need of electrode.
a bridge. The SE (SrCl2 ) serves as the One expression of the flexibility of the
separating membrane. We notice that the geometry in LSE, in contrast to SSE, is
phases CuCl2 and ZnCl2 are added to the the possibility to use a dropping mercury
electrodes in order to fix the chemical po- electrode.
tential of Cl2 there.
If a concentration cell is considered, 2.6.3
then in LSE the cell would be, for exam- Charge Transport in Electrolytes
ple: Zn|LE(Zn2+ ,Cl− )|Bridge(K+ ,Cl− )|LE
(Zn2+ ,Cl− )|Zn, with the Cl− concen- 2.6.3.1 Charge Transport in LEs
tration different in the two LEs. In In LSE charge is transported by ions in
SSE one would use a cell such as: the LE and by electrons in the metallic
I ),Zn|SE(Cl− )|Zn,Cl (P II )
Cl2 (PCl with electrodes. The LE is usually a solvent
2 Cl2
2 that contains an ionic solute. In many
different partial pressures, PCl2 , of Cl2 .
cases, the solvent is water and the solute
No bridge is again needed since the
a salt, acid or base. A different example
SE serves as the separating membrane.
I and is molten carbonates in which the salt
(For this cell to work properly PCl
II should be kept below the equilib-
2 is molten and the liquid serves also as
PCl the source for ions [1, 15, 16]. In some
2
rium pressure of Zn/ZnCl2 , otherwise cases, the water in an ionic solution
the chloride ZnCl2 would be formed at also takes part in the reaction. For
the electrodes, and both PCl2 -s would ob- example, the reaction: Cr(s) + 4H2 O(l) →
tain the (same) equilibrium value and the CrO4 2− (aq) + 8H+ (aq) + 6e− (s) that can
voltage would vanish). In SSE the metal take place at a Cr electrode. In SSE on the
electrodes in concentration cells are, in other hand, the SE is always chosen to be
many cases, inert. A typical example is the chemically stable (in the steady state).
cell: O2 (POI 2 ),Pt|SE(O2− )|Pt,O2 (POII2 ). In LEs there are at least two kinds of
When the source of material is a solid mobile ions, the cations and anions. They
electrode, it is consumed under current. have usually comparable mobilities and
Therefore the lifetime of the electrode is both may contribute to the electrical cur-
limited. In SSE, care must be taken to press rent. There is rarely an electron current
the electrode onto the SE continuously. At in a LE, though a few exceptions exist
the other solid electrode, space must be (e.g. sodium metal dissolved in liquid am-
left for the electrode to grow. In contrast, monia contributes solvated electrons [20]).
in LSE the LE stays always in good contact Electrons can also propagate in a liquid
with the electrode. Furthermore, when the by hopping between a redox couple say:
source of material is the LE, then material M+ + M2+ → M2+ + M+ , which is equiv-
can be supplied indefinitely since the LE alent to the propagation of an electron from
can be replenished. The geometry of LSE left to right [21]. This is analogous to hop-
cells closest to all solid cells is when the LE ping conduction in an impurity band in
is soaked in a porous solid. solids. The rate of electron hopping in a
A continuous supply of material is liquid is low due to the relatively large aver-
possible in SSE and in LSE, when it comes age distance between the ions for common
from the gas phase. In SSE this can also concentrations and in view of the large
distance of closest approach due to the the disordered system is very large and
cloud of solvating molecules. We can there- the electron states highly localized with
fore safely consider LEs as ionic conduc- negligible mobility.
tors with negligible electronic conduction. The ions and electrons in SEs and in
The motion of ions in a LE is impeded. MIECs polarize the ionic neighborhood.
The ion or the ion and its solvating sphere Therefore, the ions, electrons, and the
move in a viscous liquid. The force of polarized surrounding form the corre-
retardation can be calculated according sponding small polarons [22]. This would
to Stokes equation and the corresponding be analogous to polarization of an ionic
diffusion coefficient or mobility is related liquid. However, the details are different
to the viscosity. and in the sphere of solvating molecules
Protons can propagate by two mecha- that exists in LSE, neutral molecules with
nisms. One is by the viscous flow of a an electric dipole are attracted to the
complex H+ (H2 O)n where n = 1, 2, 3, 4, ion and also change orientation. For SEs
the second is by the Grotthuss mechanism, with a high concentration of mobile ions
hopping of a proton from one hydronium (>10 ion%), for example, Ag+ ion in AgI
H3 O+ to a nearby water molecule. The at T > 149 ◦ C, the interaction between the
latter mechanism has some similarity to mobile ions forces local relaxation within
hopping conduction of ions in solids as the sea of these ions, which affects the con-
discussed in the following text. ductivity and its dependence on frequency
for ac signals [23].
2.6.3.2 Charge Transport in SEs The high concentration of ions both in
SEs are those solid ionic conductors LEs and in SEs affect the width of the
for which ionic conductivity, σion , is space charge that exists at the boundary
significant and the electronic conduction, of the electrolyte near the contact interface
σel , is small. There is no exact definition of with the electrode. Any abrupt change in
what is a significant ionic conductivity, but the concentration of charged species must
usually σion > 10−4 S/cm is considered result in a space charge [24, 25]. The width
significant with the lower limit applying of that space charge is roughly the Debye
to cells with thin-layer SEs. Electronic length and for typical LEs, it is only 0.3
conduction in solids is, usually, present to 10 nm; while for typical SEs, it is 0.3
to a higher degree than in LEs. SEs to 1 nm. The detailed characteristics of the
are those MIECs for which σel σion . space charge are not identical for LEs and
The reasons for low electronic conduction SEs. One difference arises from the fact
in ionic solids are: (1) The energy gap that in SEs usually only one type of ions
is large and intrinsic electron and hole is mobile. There is therefore a difference
concentrations are very low. (2) The between enriching and depleting that ion,
occupied electron states in the conduction and the profile of the space charge when
band and hole states in the valence it is enriched or depleted is different [12].
band are localized rather than delocalized. The other differences are due to the or-
Therefore, the electron and hole mobilities dered structure of the SE and the absence
are thermally activated and low. The of a solvating cloud in SEs, as discussed in
lack of electron conduction in LE can Sect. 6.
be discussed in the same terms. In In SEs, ionic and electronic charge
these terms, the quasi–energy gap of transport is associated with deviations
from the normal state. This can be three ways, similar to those for electronic
best understood by considering crystalline defects:
solids. Then conduction is associated with
1. By thermal excitation. The result, in
point defects [14]. The perfect crystal is
an elementary solid, is a vacancy
defined as that with perfect lattice pe-
and an interstitial. In a binary-ionic
riodicity, at zero temperature. In semi-
compounds, say MX (a compound
conductors, electrons in the conduction
composed of cations M and anions X),
band are thus defects formed by ex-
thermal excitation may result in one
citation into states that are empty at
of five combinations of a vacancy, an
T = 0. Conduction in LEs can also be
interstitial and/or a misplaced ion [14].
formulated as a deviation from a stan-
The most widely discussed defects are
dard liquid state [26]. However, this is
Frenkel pairs, an interstitial of, for
uncommon and is not discussed fur-
instance, a cation and the vacancy
ther here.
left behind, Mi + VM , and Schottky
The electronic charge carriers can
pairs of vacancies of the cation M
be generated, under equilibrium, in
and anion X: VM + VX . We use here,
three ways:
the Kröger–Vink notation of point
1. By thermal excitation across the band defects. Charge is measured relative
gap. to the perfect crystal. A dot denotes
2. By extrinsic doping with donors or a positive relative charge, a prime
acceptors. denotes a negative relative charge and
3. By change of stoichiometry. The ionic x denotes a zero relative charge [14].
defects formed in this way act as native Thus, the Frenkel pairs in AgBr,
donors or acceptors. interstitial silver ions and silver ion
vacancies, are, in the Kröger–Vink
Of course, any combination of these three notation: Agi • + VAg \ .
causes may be realized simultaneously in 2. By doping. Charged ionic defects can be
any given material. In addition to elec- generated also by doping with charged
trons/holes generated under thermal equi- impurities. For substitutional doping,
librium, they may also be injected from they must be aliovalent. For example,
electrodes and excited by optical pump- doping substitutionally ZrO2 with CaO
ing under nonequilibrium conditions. As forming the defect CaZr \\ generates
mentioned before, the mobility of the elec- also ionic defects in the form of charged
trons and holes is thermally activated as oxygen vacancies, VO •• . The later are
they form small polarons. mobile at elevated temperatures. We
For ionic conduction to exist in a crystal, have mentioned before that doping may
an ion has to move out of its normal po- result also in generating electrons or
sition either into an interstitial site or find holes. Whether doping will result in
an existing defect in the form of a vacancy. ionic defects (called self compensation)
The conductivity increases therefore lin- or electronic ones depends on the
early with the defects concentration, in the materials involved, and can be predicted
dilute limit (as long as interaction between if the nature of the point defects and
the defect can be neglected). The defects their concentrations in the host are
can be generated, under equilibrium, in known in detail [27].
3. By change of stoichiometry. For binary hopping conductivity is proportional to the

compounds, say oxides, the oxygen product of concentrations: [Di • ][Di •• ].
vacancies can be generated by reducing The conduction-electron conductivity
the oxygen partial pressure, PO2 , of is proportional to the concentration of
the gas in equilibrium with the oxide, electrons in the conduction band, which,
at elevated temperatures. This yields under equilibrium, is proportional to the
for instance, CeO2−y , y > 0. Cation ratio of concentrations [Di • ]/[Di •• ], as can
vacancies or oxygen interstitials can be be deduced from the mass action law for
formed under high PO2 as is the case in the reaction Di •• + e\ ↔ Di • . The con-
Cu2 O, which contains then VCu x , V \ centration of the electrons in the conduc-
Cu
\\
and Oi [28, 29]. tion band may be quite low, though both
the ionic defects concentrations may be
The propagation of ionic defects makes use high. Then, the hopping conductivity may
of the existing point defects in various com- be higher than the conduction-electron
binations [30], provided they form a contin- conductivity. Not only impurities but also
uous, percolating path. Let us consider two native defects with variable charged states
simple but important paths: (1) Vacancy allow hopping conduction in a defect
propagation by the hopping of a nearby band [32].
ion (of the correct type) into the vacant In SSE MIECs play an important
site. This leaves a new vacancy behind. For role [33]. They are characterized by the rel-
example MM x + VM \\ → VM \\ + MM x de- ative concentrations of: (1) mobile ions
scribes the motion of a cation with valence (fixed or variable charge), (2) electron and
2+ from left to right or, equivalently, the holes, (3) immobile charged impurities
(fixed or variable charge). The most widely
motion of a vacancy VM \\ from right to left.
considered MIECs (including SEs), both
(2) The other example is that of interstitial
theoretically and experimentally, are those
cation moving via interstitial sites. In all
with a high concentration of one kind of
cases, the mobility is thermally activated
mobile ionic defects (Nion ) and a low con-
with the thermal energy required to over-
centration of electrons (n) and holes (p): n,
come the potential energy (or rather the
p Nion . A typical example is YSZ with
enthalpy) barriers between allowed sites in
∼5 anion% mobile oxygen vacancies, a
the conduction path. low but nonnegligible hole concentration
Another conduction mechanism in at T ∼ 1000 ◦ C for PO2 ∼ 1 atm and a low
solids that has to be considered, be- but significant electron concentration at
sides electron and hole conduction in the T ∼ 1000 ◦ C for PO2 ∼ 1025 atm. The elec-
corresponding energy bands and ion con- tron and hole conductivities in YSZ are low
duction, is electron hopping in an impurity at T ∼ 1000 ◦ C for intermediate PO2 . In
band [31]. For example, let D be the impu- Ag2 S, on the other hand, the concentration
rity then, Di • + Di •• → Di •• + Di • (Di • , ratio is again n, p Nion but due to a high
Di •• are the interstitial cations in the electron mobility, the electron conductivity
Kröger–Vink notation) describes the ef- is higher than the ionic one [11]. Another
fective motion of an electron from left to defect model is p n = zion Nion , where
right. This occurs in addition to any con- zion q is the charge of the mobile ionic de-
duction due to electrons in the conduction fect (q is the elementary charge). This is
band or holes in the valence band [27]. The the case for example in CeO2−x where p
n = 2[VO •• ]. In this case, the electron con- σion ), whenever ambipolar motion of
ductivity is higher than the ionic one due different charge carriers occurs, as is the
to a significantly higher electron mobility. case in chemical diffusion and under
In amorphous condensed phases, the conditions in which the driving force
interpretation of the conduction is rather for ionic motion is small. The existence
similar to that in crystalline materials if one of electronic conduction in addition to
views the amorphous phase as a crystalline an ionic one yields a very important
solid that is highly disordered. In this view, class of materials: MIECs.
the solid contains a high concentration of 3. The concentration of ionic defects in
defects with at least one kind being mobile. SEs must be high compared to that in
This explains why glasses exhibit, in many LEs in order to obtain comparable ionic
cases, a higher ionic conductivity than the conductivity because the mobility of the
corresponding crystalline solid with the ions in SEs is usually lower than in
same composition. LEs. Even so, in most cases, the SEs
Polymer ion and electron conductors are will exhibit a high conductivity only at
of much current interest in SSE because elevated temperatures.
of their ease of production, low cost and 4. The path for ion motion and the temper-
relative high conductivity at room tem- ature dependence of ionic conductivity
perature. The conduction mechanism for in SEs and LEs are very different.
ions is different from that in inorganic
solids and is based either on segmental 2.6.4
motion in the polymer and hopping of the Theoretical Current Voltage Relations for
ion (in a remote similarity to the Grot- Ideal ECs (No Electrode Polarization)
thuss mechanism for proton conduction
in water) or on a LE embedded in the poly- 2.6.4.1 General Current Density Equations
mer. Electron conduction in polymers can The current density equation for the ions
be discussed in terms used for conduction in the bulk of the LE or SE are the same:
in inorganic semiconductors. Single-phase σion
ji = − ∇ µ̃ion (1)
MIECs polymer with significant conduc- zion q
tivities do not exist yet. Quasi-MIECs are
formed by blending two polymers, one where µ̃ion is the electrochemical poten-
conducting ions and the other conducting tial of the ions, zion the charge of the ion
electrons/holes [34]. and q is the elementary charge of a pro-
ton. For SEs the charge considered is the
2.6.3.3 Summarizing the Differences
effective one in the Kröger–Vink notation
Between Transport in LEs and SEs (though absolute charges could also be dis-
The key differences between conduction cussed [35]). The convention used here is
in LEs and in SEs are the one used in SSE referring the various
quantities to a single particle rather than
1. There is usually only one kind of mobile to a mole as done in LSE. Thus, the chem-
ions in SEs, as compared to at least two ical potential is per particle, the charge is
in LEs. counted in elementary charges, q, rather
2. There is an electronic conductivity, σel , than in Faraday units and one uses the
present in SEs that has to be taken Boltzmann constant rather than the gas
into consideration, even if small (σel constant.
For LEs one has to write at least reaction, and we shall use this approach
two current-density equations, one for here also. Reference electrodes are also
each mobile ionic specie. If a supporting used in SSE, but they are not universal
electrolyte is added to the LE, two more standard ones, but rather chosen ad hoc
current-density equations have to be added for each particular system.
for the additional ions. The gradient in the
latter two equations vanishes under steady 2.6.4.3 Supporting Electrolyte
state conditions because the electrodes are Let us first examine the effect of a
blocking for the ions of the supporting supporting electrolyte in a LEs and its
electrolyte. On the other hand, for SEs, absence in SEs. The supporting electrolyte
one has to consider the current density of does not participate in the electrochemical
electrons and/or holes. For electrons: reaction, by definition. Thus, jion = 0
σe (where j is the current density) for the
je = ∇ µ̃e (2)
q ions of the supporting electrolyte. Hence,
by Eq. (1)
where µ̃e is the electrochemical potential
of the electrons. µ̃e is equal to the redox ∇ µ̃ion = 0, jion = 0 (3)
potential of a redox couple, and it is equal
to the Fermi level in solids. An electrochemical potential, µ̃, can be
Quasi-coupling between the ionic and expressed in terms of a chemical potential,
electronic currents is sometimes intro- µ, and the inner (Galvani) electrical
duced by adding a term Lie ∇ µ̃e to Eq. (1) potential ϕ,
and a term Lei ∇ µ̃i to Eq. (2). However, it
µ̃ = µ + zqϕ (4)
was shown that these corrections are not
really necessary (except perhaps in met- where zq is the charge of the relevant
als [33]) if the current density of all mobile specie. In the limit of relative high con-
species, as well as interactions between de- centrations of the supporting electrolyte,
fects are all taken into consideration [36]. gradients in the chemical potential of its
ions can be neglected even when a voltage
2.6.4.2 Half Cells is applied to the electrodes, and/or there
In LSE it is common to discuss an EC as exist a difference in the composition of the
if separated into halves, and to consider electrodes,
the potential at each electrode with respect ∇µion = 0 (5)
to a third standard reference electrode.
Combining Eqs. (3–5) yields:
Furthermore, many of the ECs discussed
in LSE contain supporting electrolytes. ∇ϕ = 0 (6)
This is quite different from SSE, in which
supporting electrolytes are absent. Thus, that is, the Galvani potential is uniform,
in LEs the voltage as well as the Nernst the inner electric field vanishes in the LE
theoretical voltage can be expressed in due to the supporting electrolyte. In this
terms of ion concentration in the LE. case, the other (minority) ions flow under
In SSE the analysis is done in terms of a chemical potential gradient only.
the chemical potentials of the materials In SSE, on the other hand, the mobile
and electrochemical potentials of the ions ions are those participating in the elec-
and electrons/holes involved in the cell trochemical reaction. The concentrations
of the mobile ions can be high as is the Xi \\ ←−→ X + 2e\ (10)

case with Agi • (interstitial silver ion, in the
Kröger–Vink notation) in Ag2 S and with If equilibrium prevails locally at each
VO •• in YSZ. Eq. (5) holds for these ions. electrode, then,
However, their current does not vanish, in
general, and Eqs. (3) and (6) do not hold. µX = µ̃ion + zion µ̃e (11)
Instead, Eqs. (1) and (5) yield,
with zion = −2 for Xi \\ . When the ionic
jion = −σion ∇ϕ (7) current of Xi \\ vanishes, then by Eq. (1),
∇ µ̃ion = 0 and the voltage measured on
that is, the ionic current through the SE is the cell is equal to the Nernst voltage (cf.
driven by an electric field: −∇ϕ. Eqs. (8), (9), and (11)), despite the existence
In other types of MIECs, such as CeO2−x of an electronic conductivity,
(p n = zion Nion ), Eq. (5) does not hold.
Thus, Eq. (7) is not true anymore and one V (jion = 0) = Vth (12)
has to revert to Eq. (1) [37].
As we shall see shortly, V (ji = 0) need
not be equal to the open-circuit voltage of
2.6.4.4 Boundary Conditions the cell.
The voltage measured on an EC is given When µX refers to a dilute concentration
by the difference in the electrochemical or an ideal gas, for example, oxygen at
potential of the electrons at the two PO2 ≤ 1 atm, then Boltzmann statistics
electrodes connected to the voltmeter, holds, and

−qV = µ̃e (8) 0 1 [X2 ]1
µX = µX (T ) + kB T ln
2 [X2 ]2
The Nernst voltage, Vth , is, in SSE, a the- (13)
oretical quantity, having the dimension of where kB is the Boltzmann constant, [X2 ]1
voltage, that reflects a chemical potential and [X2 ]2 the concentrations of X2 near
difference. For a given chemical compo- the two electrodes. The Nernst voltage
nent, X, with mobile ions of charge zion , depends then on the concentrations [X2 ]1
Vth is defined as, and [X2 ]2 . From Eqs. (9) and (13),

µX RT [X2 ]1
−qVth = (9) −qVth = ln (14)
zion 2zion q [X2 ]2
where µX is the difference in the We notice that Vth does not depend on the
chemical potential of the component X at concentration of ions in the SE (which for
the two electrodes. Thus, for example, for YSZ, for example, is fixed and uniform to
YSZ one can consider µ(O) = 12 µ(O2 ) a good approximation). µX can be fixed by
and zion = −2, while for Ag2 S one would a single phase, for example, O2 gas with a
consider µ(Ag) and zion = 1. given PO2 or Ag metal or it can be fixed by
a reaction in the electrode. For example,
2.6.4.5 Relations Between Open-Circuit PO2 may be fixed by the couple of reacting
Voltage and the Nernst Voltage solids Fe/FeO when in equilibrium. µ(Na)
µX is related to µ̃ion and µ̃e . Let the can be fixed by SnO2 /Na2 SnO3 /O2 under
mobile ion be, for example, Xi \\ . Then, a PO2 value for which SnO2 is stable or it
can be fixed by CO2 /Na2 CO3 /O2 under a with fixed ionic charges and with negligible
PO2 value for which CO2 is stable [38]. µX electronic conductivity.
may also be fixed by an auxiliary EC [33]. The way the Nernst voltage is calculated
The open-circuit voltage is the voltage in LSE is by considering the electrode
when the total current through the EC reaction under open-circuit condition,
vanishes. In SSE the total current, It , rather as in SSE from µX (Eq. 9).
may contain a nonnegligible electronic Let us consider, for example, the cell,
contribution, Iel and Ag|AgCl|LE(H2 O,H+ ,Cl− )|Pt,H2 , and cal-
culate the Nernst voltage using first the
It = Iion + Iel (15) LSE approach and then the SSE approach.
The voltage of the cell is, according to
Under open circuit conditions, It = 0 and
Eq. (8),
the partial currents Iion = −Iel need not
vanish. Therefore, the voltage for It = 0 −qV = µ̃e (Pt) − µ̃e (Ag) (17)
is different from the voltage for Iion = 0.
It can be shown that when a single ionic µ̃e (Ag) is fixed by the electrode reac-
species is mobile then [37], tion: Ag + Cl− ↔ AgCl + e− , which un-
der equilibrium yields,
Voc < V (jion = 0) = Vth ,
µ̃e (Ag) = µ0 (Ag) + µ̃(Cl− )Ag,AgCl
Voc −−−→ V (jion = 0) only when
− µ0 (AgCl) (18)
Iel −−−→ 0. (16a)
where µ̃(Cl− )Ag,AgCl is the value of the
Eq. (16a) holds whether the electron/hole
electrochemical potential of Cl− near the
transport is via the conduction/valence
interface Ag/AgCl and µ0 denotes the
band or via defect band hopping.
standard chemical potential at the given
When variable charge, mobile ions exist,
temperature, T , which for solids is equal to
for example, Cu+ and Cu++ , then there
the chemical potential. The measurement
are different Nernst voltages, one for each
of the open-circuit voltage should be done
zion value (see Eq. 9). We neglect now any
under steady state. Then the Cl− current
electronic current. Then, the open-circuit
must vanish in the two, purely ionic
voltage is not equal to either of the Nernst
conductors AgCl and LE(H2 O,H+ ,Cl− )
voltages. For the example with zion = 1
placed in series. Thus, µ̃(Cl− ) is uniform
and 2 [33, 39],
throughout the cell and we may remove
Vth (zion = 2) < Voc < Vth (zion = 1) the subscript Ag/AgCl and we may add
the subscript LE.
(16b)
µ̃e (Pt) is fixed by the electrode reaction
In LSE the Nernst voltage is, in princi-
H2 ↔ 2H+ + 2e− , which, under equilib-
ple, defined similarly (with respect to a
rium, yields,
standard-reference electrode) and called
the equilibrium potential. However, in µ̃e (Pt) = 12 µ(H2 ) − µ̃(H+ )Pt,LE (19)
practice one usually refers to the open-
circuit voltage of a half cell with respect where µ̃(H+ )Pt,LE is the electrochemical
to a standard electrode. The open-circuit potential of the protons near the H2 ,Pt/LE
voltage is equal to the Nernst voltage when interface. µ̃(H+ ) turns out to be uniform
it is measured on purely ionic conductors, in the LE, since both µ̃(Cl− ) and µ(HCl)
are uniform in the LE, thus the subscript µ(Cl2 )Pt,LE , the chemical potential of Cl2
Pt,LE can be replaced by LE. Combining at the LE/Pt,H2 interface, is fixed by the
Eqs. (17), (18) and (19) yields, reaction of H2 and HCl in the LE: H2 +
Cl2 ↔ 2HCl ↔ 2H+ + 2Cl− . Under equi-
−qVoc = 12 µ(H2 ) − µ0 (Ag) + µ0 (AgCl)
librium,
− µ̃(H+ )LE − µ̃(Cl− )LE (20)
1
2 µ(Cl2 )Pt,LE = µ(H+ )Pt,LE + µ(Cl− )Pt,LE
Under equilibrium one can make the
following substitution: − 12 µ(H2 ) (26)
µ̃(H+ )LE + µ̃(Cl− )LE = µ(HCl)LE Combining Eqs. (25) and (26) and noticing
= µ(H+ )LE + µ(Cl− )LE (21) that µ(H+ ) and µ(Cl− ) are uniform in the
LE, as mentioned before, yields,
in Eq. (20), then,
−qVoc = 12 µ(H2 ) − µ0 (Ag) + µ0 (AgCl) −qVth = 12 µ(H2 ) − µ0 (Ag) + µ0 (AgCl)
− µ(H+ )LE − µ(Cl− )LE (22) − µ(H+ )LE − µ(Cl− )LE (27)
or when expressed in terms of activities, Vth equals Voc of Eq. (22) since the elec-
tronic component in the combined elec-
−qVoc = 12 µ0 (H2 ) − µ0 (Ag) + µ0 (AgCl)
trolyte AgCl/LE vanishes (see Eq. (16a)).
− µ0 (H+ )LE − µ0 (Cl− )LE This demonstrates that the two definitions
a(H2 )1/2 of the Nernst voltage are indeed equal
+ kB T ln (23) when the membrane is a pure ionic con-
a(H+ )LE a(Cl− ) LE
ductor. This would not be so if electronic
For low ion concentrations a(H+ )LE and conduction would be present (as can be the
a(Cl− )LE may be replaced by the corre- case in SSE), as then Vth = Voc (Eq. 16a).
sponding (uniform) concentration [H+ ] Different values for the Nernst voltage
and [Cl− ] in the LE. defined as Voc rather than Vth would be
The argumentation in the SSE approach obtained also when the LE contains vari-
would be as follows: There is a selective able charge-mobile ions, for example, Cu+
membrane for Cl− ions, that is, an and Cu++ , as then Vth = Voc as can be
electrolyte, made of two parts connected in seen in Eq. (16b).
series: AgCl and the LE. The Nernst voltage
is determined by µ(Cl2 ) according to 2.6.4.6 Chemical Diffusion Coefficient
Eq. (9), Chemical diffusion of a single component
2qVth = µ(Cl2 )Pt,LE − µ(Cl2 )Ag,AgCl occurs, in ionic solids, by the motion of
(24) ions. This must be accompanied by an
µ(Cl2 )Ag,AgCl , the chemical potential of Cl2 equal current of electronic charge to avoid
at the Ag/AgCl interface is governed by charge accumulation. Thus diffusion in
the reaction: 2Ag + Cl2 ↔ 2AgCl. Under ionic solids with a single type of mobile
equilibrium, ions is an ambipolar motion of ions
and electrons (or holes). This motion is
1
2 µ(Cl2 )Ag,AgCl = µ0 (AgCl) − µ0 (Ag) governed by Eqs. (1) and (2) and the
(25) condition It = 0. The chemical diffusion
coefficient is then [12, 40], Vth − V

Ie = −Sσe1
kB T σion σel 1 ∂ ln(aX ) L
Dchem = e−q(Vth −V )/kB T − e−qVth /kB T
zion q σion + σel cX ∂ ln(cX ) × (30)
(28) 1 − e−q(Vth −V )/kB T
where cX is the total, local, concentration where σe1 is the conductivity in the MIEC
of material X (or defects), which, in ionic at the end, where V ≡ 0, L is the length
solids, may be equal to the concentration of and S the cross-sectional area of the MIEC.
the mobile ions (or defects), aX is the local In SSE 1/kB T is normally denoted as β.
activity of the component X. Dchem need We shall not use this notation here to
not be uniform. It is dominated by the avoid confusion with the notation β of the
smaller of the conductivities with Dchem ∝ symmetry factor in LSE.
σel for σel σion and Dchem ∝ σion for A different defect model for a wide group
σion σel . of MIECs is one with comparable concen-
tration of electrons (or holes) and mobile
2.6.4.7 I–V Relations ions, (p n = zion Nion ). The inner elec-
The ionic and electronic currents in the tric field −∇ϕ in this case is not uniform
bulk of a MIEC are governed by the and the I –V relations are different from
current-density equations of the kind of the previous ones [33, 37],
Eqs. (1) and (2), the boundary values of µX ,
kB T Vth − V
of the mobile species, at the two electrodes Iion = 2Sνion (n1 − n2 ) (31)
Vth L
and the voltage on the EC, Eq. (8). The I –V
relations are actually two relations: Iion vs and
µX1 , µX2 , and V , and Iel vs µX1 , µX2 , and kB T V
Ie = −2Sνe (n1 − n2 ) (32)
V . These relations are not universal but Vth L
depend on the defect model of the MIEC.
For an EC based on an MIEC with high where ν is the mobility, n1 and n2 the
and uniform concentrations of mobile ions concentration of electrons (and of the
(n,p Nion ) and thus a uniform ionic mobile-ionic defects) at the two ends of the
conductivity, MIEC. When the electrodes are reversible,
n1 and n2 are then fixed by µX,1 and µX,2
Vth − V also under current, Iion is linear in V
Iion = (29)
Rion and Ie is proportional to V . On the other
hand, when the electrodes block the ionic
where Rion is the resistance of the MIEC to current, (e.g. when full blocking occurs
ionic current. The inner electric field, −∇ϕ Iion = 0, hence V = Vth ) n1 and/or n2
is uniform, in the steady state, as can be depend then on V and the Ie − V relations
deduced from Eq. (7) and the uniformity are not linear.
of jion and σion .
The electronic current, Iel , is not linear 2.6.5
in V . The electron and hole concentrations Methods for Characterizing ECs
are not uniform and are dependent on
the boundary values, µX1 , µX2 , and V . 2.6.5.1 Methods for Characterizing LEs
In particular, for MIECs with one type of and SEs/MIECs
minority electronic carriers, for instance, The key parameters of interest in LEs
electrons, (p n Nion ) [33, 37], and SEs/MIECs in ECs for the purpose
of discussing the I –V relations are useful to combine coulometric-titration

the nature of the charge carriers, their measurements with in situ conductivity
concentrations and mobilities. measurements [28, 29]. Hall coefficient
In LSE the concentrations of the mobile measurements to determine the charge-
ions are usually known and fixed by the carrier concentration are difficult because
composition of the LE. This is certainly of the low mobilities found in ionic solids
true for strong electrolytes where the both for ionic and electronic charge car-
concentration of the mobile ions is equal riers. Only few Hall measurements were
to the concentration of the corresponding done on ionic solids to determine the elec-
material added to the solution, times the tron/hole concentrations and we are aware
stoichiometry coefficient. of only work done by Funke that measured
In SSE on the other hand, the nature the ion Hall coefficient in an SE [42, 43].
and concentration of mobile defects can The mobility can be determined from the
be guessed a priori with reasonable accu- electrical conductivity once the defect con-
racy only when they are introduced by centration has been determined, or directly
doping, provided the dopants are fully from a drift experiment.
ionized. When the mobile defects are gen- In SSE one has to determine both the
erated by thermal excitation, for example, ionic conductivity, σion , and the electronic
Frenkel pairs, or by deviation form stoi-
conductivity, σel . For that purpose, one
chiometry, they have then to be identified
can use blocking electrodes that are per-
and their concentration measured. The ex-
meable to either ions or electrons, and
perimental methods used can be direct
measure the I –V relation when a steady
ones, that is, spectroscopic ones, such
state is reached. This is the basis for
as characteristic ultra-violet (UV), visible
the Hebb–Wagner polarization method
(VIS), and infrared (IR) absorption; elec-
for determining σel [33, 44–46]. In the
tron spin resonance (ESR); and nuclear
Hebb–Wagner method, the ion-blocking
magnetic resonance (NMR). The methods
can be indirect ones such as conductivity electrode is an inert conductor, for exam-
measurement, which is linear in the de- ple, graphite or platinum. The blocking
fect concentration (for low concentration), occurs either because a continuous-inert
thermogravimetry (TGA) that follows the electrode can block the access of mate-
changes in the stoichiometry by weight rial to the SE or because the gas phase
changes, and by coulometric titration that contains a limited amount of the rele-
follows the stoichiometry changes by elec- vant material. The counter electrode in the
trical measurements and can be done Hebb–Wagner experiment is a reversible
more accurately than TGA. Other meth- one. It fixes the composition in the MIEC
ods used, follow changes in stoichiometry being tested, near the electrode/MIEC con-
such as pressure measurements, IR mea- tact. This allows to relate the measured
surements and chemical-titration mea- local σel to a well-defined composition
surements of the gases evolving from the of the MIEC. An example of such a cell
sample [41]. The indirect methods require is: (−)Ag|AgBr|C(+), which allows the
a model to relate the measured quantities determination of the distribution of σel
to the defect nature and concentrations. throughout the AgBr. The voltage applied
In particular, in order to decide whether has to be kept below the decomposition
the defects are charged or neutral, it is voltage. When this voltage is exceeded the
MIEC decomposes, in analogy to electrol- measured in addition (using coulometric

ysis of an LE in LSE. titration in combination with EMF mea-
The partial ionic or electronic currents surements [33]).
and conductivities can also be measured
by setting the driving force for the other 2.6.5.2 Methods for Characterizing
partial current to zero. If one examines Elements in the EC
the Hebb–Wagner method, it becomes Working ECs are commonly characterized
apparent that this nullification also occurs using ac-impedance measurements. This
there. Because of the extreme resistance of allows to identify the different contribution
the ion-blocking electrode to ionic current, to the impedance by the various elements
the driving force for ionic motion in the of the cell [49–51]. This method, first
MIEC vanishes (∇ µ̃ion = 0). developed for LSE, is also extensively
A simple way to make the driving force used in SSE. As a result of the different
for electrons vanish (∇ µ̃e = 0) is to short geometry in LSE and SSE the processes
circuit the electrodes [47]. This sets the and their sequence are different. Yet
difference µ̃e = 0 and for most common the analysis can use the same electrical
MIECs, it also sets ∇ µ̃e = 0. elements, namely, resistors, capacitors,
Another method for separating σion from and Warburg elements (for simulating
σel is by their different dependence on diffusion). In both cases non Faradaic
the chemical potential, µX , applied to currents may exist, which originate from
the electrodes. This method works well charging double layers without a chemical
when σion σel or vice versa, and allows reaction taking place.
to identify which conductivity is being Voltammetry is commonly used in LSE
measured [22]. not so in SSE probably because of the dif-
The chemical diffusion coefficient, ficulty in the interpretation of the results.
Dchem , can be determined using time-
dependent electrochemical methods. The 2.6.6
best known is the galvanostatic-intermit- Electrode Processes and Overpotentials
tent-titration-technique (GITT) [40]. In this
measurement, ions are pumped through 2.6.6.1 General
an SE into (or out of) an MIEC for The details of electrode processes in LSE
a short time. The pumping is stopped and SSE are different, as mentioned
and the relaxation of the cell open-circuit before. To examine the electrode processes
voltage is followed. The time dependence in SSE let us examine a few examples. The
of the cell voltage during pumping and EC, (+)Ag|AgI|Ag2 S(−) represents the EC
during relaxation, both yield Dchem . To shown in Fig. 2(a). AgI is an SE that
overcome contact overpotential a modified conducts Ag+ ions (at T > 149 ◦ C [11]).
arrangement was suggested in which the The electrodes Ag and Ag2 S are a source
chemical potential of the diffusing specie or sink for both the material (silver) and for
is monitored as it arrives at the remote the electrons. In the oxidation reaction of
end [48]. Ag at the Ag/AgI interface, silver atoms or
Through the dependence of Dchem ions have to leave the silver metal and enter
on σion and σel (Eq. 28) Dchem can be the AgI lattice as ions. There is charge
used to determine the minority con- transfer of ions across the interface and
ductivity provided cX −1 ∂ ln(aX )∂ ln(cX ) is there may also be electron transfer to form
the ions there. At the AgI/Ag2 S interface, The distribution ϕ( r ) and the concentra-
Ag+ ions are transferred from the AgI tion of the charges that fix ϕ( r ) therefore
lattice into the Ag2 S lattice. deserve an over view. In LSE the LE is
For porous gas electrodes, shown terminated abruptly at the contact with
schematically in Fig. 2(b), there can be the electrode (and at the upper side in
many different steps involved. For exam- contact with the atmosphere). Charge is
ple, the overall cathodic reaction on a Pt transferred, also under equilibrium, from
electrode on YSZ is: O2 (gas) + 4e− (Pt) ↔ the metal electrode into the LE (or vice
2O2− (YSZ). One can suggest that the over- versa) leaving behind the opposite charge
all reaction consists of the following series and thus forming a double layer. In the
of elementary steps: diffusion of O2 in the metal the excess or missing charge (in the
gas phase toward the TPB, O2 adsorption ‘‘sea’’ of electrons) concentrates near the
on the Pt, O2 dissociation, electron transfer M/LE interface, in a very thin layer, less
to Oad to form Oad 2− , diffusion of Oad 2− than a mono atomic–layer wide. In the LE
along the Pt electrode surface toward the the charge is spread out a little more over
SE (YSZ), and then entering of the ion into a length determined by the Debye length,
the YSZ. There can be other series of steps λD (which is the inverse of the Debye re-
that yield the same overall reaction [52]. ciprocal length, κ commonly used in LSE).
A different electrode is one that is not λD is about 0.3 to 10 nm wide for most
porous but a continuous MIEC and the LEs due to the high concentration of the
source or sink for material is not the MIEC mobile ions.
but the gas phase, as shown schematically In the LE the local dielectric constant at
in Fig. 2(d), Then diffusion of material in the interface may be significantly modified
the form of ions through the MIEC, as well as a result of orientation of the water-
as charge transfer of ions at the MIEC/SE molecule dipole in the electric field of the
double layer. Furthermore, ions can flow
interface has also to be considered.
to the interface and be adsorbed there, in
It is evident from those examples
the double-layer region.
that electrode processes in SSE involve
The explanation for the charge transfer
elementary steps that may be limited by
on contact of the metal and the LE, un-
diffusion, others by slow-charge transfer
der equilibrium, depends on the approach
of electrons in an electronation or de-
taken. Of course, the different explana-
electronation step, and yet others by a slow
tions are equivalent. From the SSE point
transfer of ions across an interface. The
of view, the charge in the double layer
latter step is not common in LSE. On the
(for zero applied voltage) originates from
other hand convection exits in LSE but not
a difference in two properties between the
in SSE.
electrode material and the LE, the work
function and the chemical potential of the
2.6.6.2 Inner (Galvani) Electrical Potential mobile component. For inert metals it is
only the difference in the electron work
2.6.6.2.1 Inner Electrical Potential in LSE functions that matters. Electrons are trans-
The analysis of elementary steps that inferred until the electrochemical potential
volve charge transfer, whether of electrons of the electrons in the metal and the nearby
or of ions, require a detailed discussion of LE is the same. Then the potential of zero
the inner (Galvani) electrical potential, ϕ. charge, used in LSE, equals the difference
in the work functions (divided by the el- translational symmetry at the surface. Elec-
ementary charge, q). When the electrode trons or holes trapped at these states cause
can be dissolved (e.g. a Zn electrode) then a double layer between the trapped charges
material and electrons are transferred, un- and the opposite charge left behind in the
til the chemical potential of the material bulk of the semiconductor and distributed
as well as the electrochemical potentials of nearby [53]. Due to quantum effects, elec-
the corresponding ions and of the electrons trons can be found outside the free surface.
all are uniform across the interface. This also contributes to a double layer.
At the free surface of the LE facing In SEs and MIECs of interest the
the gas phase a double layer can also be contributions to surface charges come
formed, in analogy to the free surface of a mainly from mobile ions that exist in
solid, discussed in the following text. Our large concentrations rather than from
main concern is the interface between the electrons or holes, when they are the
LE and the metal electrons. minority charge carriers. These ions can
accumulate or be depleted near the surface
2.6.6.2.2 Inner Electrical Potential in SSE forming a charged layer there (the ‘‘core’’
In SSE the situation is somewhat more in Fig. 3a) [54, 55]. The position of the
complex because of the rigid structure of ‘‘core’’ may vary to be in between the first
the solid. The Galvani potential distribu- and second atomic layer, within the first
tion is shown in Fig. 3. Let us start with atomic layer or on the first atomic layer.
the free surface of a solid. A semicon- The opposite ionic charge left behind in
ductor may have surface states for the the bulk is distributed nearby. For a core
electrons because of the breaking of the centered at xc (in Fig. 3), the opposite
ϕ ϕ
ϕ
X X X
+ Metal
+ Metal +
+ + +
+
+ − + − −
+
+ − − + + − − − −
+
+ − − − + + − − − + − − −
0 Xc X1 X 0 Xc X1 X 0 X1 X
(a) (b) (c)
Fig. 3 Distribution of charge and Galvani potential, ϕ, at a solid interface. xc
- the position of the core. In (b) and (c) the metal is left of x = 0. (a) Free
surface; (b) metal/MIEC interface that contain a core and (c) metal/MIEC
interface without a core.
charge is distributed beyond x1 to the are adsorbed in the double-layer region.

right due to the finite size of the ions. In When ions do exist in this region in LSE
the double layer formed one layer is thin, then Fig. 3(b) is the appropriate analogous
of a mono-atomic width and the opposite presentation and the core is the analogue
charge is spread out over a few atomic of the adsorbed-ions layer.
distances. When applying a voltage on the EC
When an inert metal electrode is ap- a part η of this voltage appears across
plied on the SE or MIEC the difference in the interface and modifies the Galvani
the work functions between the metal and potential drop there. The modification is
the SE/MIEC, contributes an additional generated by a change of the double-layer
charge transfer that charges the metal charge.
surface and changes the charges in the
SE/MIEC (i.e. in the core and the space 2.6.6.3 I–V Relations for Charge-Transfer
charge). This modifies the electric poten- Processes in LSE
tial distribution as shown schematically in
Fig. 3(b). The charge in the metal, which is 2.6.6.3.1 The Butler–Volmer Equation for
compensated in the SE/MIEC by a change an Elementary Step Charge-transfer I–V
in distribution of, mainly, majority ionic relations are given in LSE by the But-
species, is electronic. The potential dis- ler–Volmer type equation. For an elemen-
tribution is different and determined by tary step in which a species with charge zq
redistribution of the charged species. In is transferred,
particular an electric field appears in the
I = I0 (e(1−β)zqδ(ϕ)/kB T − e−βzqδ(ϕ)/kB T )
region between the surface of the metal
and the core (between x = 0 and xc in (33a)
Fig. 3b). This region has a width of the or- where ϕ is the difference in the Galvani
der of 0.1 nm. The field for x > xc is also potential across the interface, δ(ϕ) is
modified. When the electrode applied is the deviation of ϕ from the equilibrium
not inert (e.g. an Ag electrode on AgI), ma- value ϕeq
terial transfer as well as electron transfer
δ(ϕ) = ϕ − ϕeq (33b)
occurs. This modifies the concentration of
ionic defects in the core and in the space z is unity for an electron or an ion of
charge in the SE/MIEC. The distribution unit charge, z > 1 for an ion with a larger
of ϕ looks, however, qualitatively as shown charge. In Eq. (33a) the concentration of
in Fig. 3(b). the species involved are assumed not to be
A simplified situation exists when one altered under current. I0 depends on the
can neglect the core and the corresponding concentration of these species [56].
space charge that exists in the free surface In these I –V relations β is not neces-
of a solid (shown in Fig. 3a). Then a sarily an integer. This is due to the form
space charge is formed only due to of the potential barrier the species have
the contact between the SE/MIEC and to overcome. A potential barrier may ex-
the metal electrode. This situation is hibit a maximum as shown in Fig. 4(a),
shown in Fig. 3(c) for an inert metallic or be in the form of just one slope as
electrode. This last case is similar to the shown in Fig. 4(b), which might represent
situation in the metal-LE double layer a step potential or be one side of a mini-
in LSE mentioned before, when no ions mum in the potential shown in Fig. 4(c).
Potential energy
X X X
(a) (b) (c)
Fig. 4 Potential energy barriers of different forms.
Potential barriers of type (1) may have a thermal activation but not necessarily
contribution from the electrical potential the same energy. Under equilibrium
energy as shown in Figs. 5(b) and (c). This the two transfers (which depend on
then leads to expressions of the form of the barrier heights, local concentrations
Eq. (33a) with 0 < β < 1. The parameter of reactants and products and reaction
β is associated with the symmetry of the constants) exhibit the same rate. The
potential maximum, therefore β is denoted electrical potential contribution is, under
as the symmetry factor. equilibrium, ϕeq (Fig. 5b). Under current
Let us first see why a potential barrier a deviation δ(ϕ) from the equilibrium
of the form of Fig. 4(a) arises in LSE. value ϕeq is generated. Part −βδ(ϕ)
For ion transfer between the solution modifies one side barrier height and a
and the metal electrode, say a solvated part (1-β)δ(ϕ) modifies the other side
proton, the position, sufficiently removed barrier height (Fig. 5c).
from the metal electrode, where it is fully
When the charge transferred is an
surrounded with water molecules, is of
electron, Eq. (33a) still holds. An example
relative low energy. The position in which
for such a reaction is: HLE + + eM − ↔
the proton is attached to the metallic elec-
Had,M , where: HLE + is a proton in the
trode with most of the water molecule
LE, eM − an electron in the metal electrode,
surrounding it removed, is also of relative
and Had,M an hydrogen atom adsorbed
low energy due to the interaction of the
proton with the metal. The positions in on the metal electrode. It is not obvious
between these two ones are of higher en- that this equation holds for an electron
ergy forming a maximum in the potential transfer as an electron can easily tunnel
energy. The barrier may have a contribu- through a potential barrier that is a few Å
tion of electrical potential energy, which wide and only ∼1 eV high, as is typically
is linear in position x. The subdivision the case at the Metal/LE interface [57, 58].
into different energy contributions looks However, the barrier that the electron
qualitatively then, as shown in Figs. 5(b) has to cross is not the only factor to be
and (c). considered. For tunneling to take place
The transfer of a proton from left to the initial and final states of the electron
right over the barrier in Figs. 4(b), 5(b) have to be degenerate, that is, have the
and (c) requires thermal activation and the same energy. Furthermore, the state the
transfer from right to left also requires electron is transferred to must be empty
Potential
~
µ
µ0
µ0
µ0 + zqϕ eq
zqϕeq
X X
(a) (b)
~
µ
µ0 + zqϕ
µ0
zqϕ µ0 + zqϕ eq
X
(c)
Fig. 5 Potential energies (µ0 , µ̃, zqϕ) in SSE. (a) Inside the SE (or
MIEC); (b) at the SE/electrode interface under equilibrium and (c)
at the SE/electrode interface under current.
to be able to receive the electron. It turns elementary steps some transferring charge
out that the position of the proton and and some not. The products of one step are
hydrogen atom with the right energy for the reactants for the following step. When
electron tunneling lies at a relative high the charge transfer occurs across exactly
energy and the motion toward that position the same potential difference, ϕ then the
forms a potential barrier similar to the one I –V relations can be related through I0
shown in Figs. 5(b) and (c). This results in to the concentrations of the species that
Eq. (33a) also for electron transfer [56]. exist in the first and the last step in the
Charge transfer to a neutral species series. Taking one elementary step to be
that approaches the metal electrode from rate determining (rds) [56],
beyond the outer Helmholtz plane, for
I = I0 (eαa qδ(ϕ)/kB T − e−αc qδ(ϕ)/kB T )
example, O2 dissolved in the LE, may
also follow the Butler–Volmer Eq. (33a). (34a)
This is expected to be different from the where,
I –V relation for an electron transfer to O2 γc
αc = + rβ,
already adsorbed onto the metal (see in the ν
following text). nt − γc
αa = + rβ (34b)
ν
2.6.6.3.2 The Butler–Volmer Equation for where, nt is the total number of elementary
Multistep Reactions An over all process at charges transferred, γc is the number
an interface consists usually of a series of of elementary charges transferred before
the rds, r is the number of elementary by the average over a volume of atomic
charges transferred in each rds, ν is size. However, for charge transfer at an
the number of times the rds has to be interface the detailed distribution of the
repeated to complete one reaction and β energies on an atomic scale must be con-
is the symmetry factor of the rds. Thus, sidered, as these exhibit potential barriers
for example, r = 1 for an electron being that the ion has to overcome [66].
transferred; r = 2 for an O2− ion being To be specific let us consider the case of
transferred; and r = 0 if the rds is a dilute defect concentration, [i], so that the
chemical reaction without charge transfer. chemical potential is:
2.6.6.4 I–V Relations for Charge-Transfer µ = µ0 + kT ln[i] (36)

Processes in SSE
where kT ln[i] is contributed by the (config-
2.6.6.4.1 General Equation (33a) for an urational) entropy and µ0 = u + P v − Ts0
elementary charge-transfer step has been (where s0 is the standard entropy per par-
applied in SSE [51, 59–61] as well as ticle) is equal the enthalpy per particle up
Eqs. (34a) and (34b) for a multistep reac- to the constant − Ts0 , u is the energy per
tion that contains charge transfer [62–65]. particle, P the pressure and v the vol-
This application to SSE is justified for ion ume per particle. On the atomic scale, µ0
transfer across an interface. However, it varies mainly due to variations in u(x).
may be problematic in two cases: (1) The Variation of µ0 vs. x are shown schemat-
transfer of an electron may not be asso- ically in Fig. 5(a). The changes within a
ciated with a symmetry factor 0 < β < 1, crystalline solid are periodic, on an atomic
but rather with β = 0 or 1. (2) The more scale. The periodicity ends at the inter-
complicated morphology may not allow face. A potential energy maximum occurs
the application of Eq. (34a) to a multistep at the boundary as follows: the energy of
reaction because the charge transfer ele- the ion exhibits a local minimum at the
mentary steps do not occur across exactly normal site inside the solid and also at an
the same interface. interstitial site, the ion is at a local energy
minimum outside as an adsorbed ion, and
the energy exhibits a maximum as the ion
2.6.6.4.2 Ion Transfer Let us first con-
moves between such two minima. The en-
sider an ion being transferred from one
ergy exhibits a maximum as the ion enters
solid into the other. The transfer is driven
the solid, since it has to squeeze in be-
by a difference in the electrochemical po-
tween the ions of the solid in the outermost
tential of the ions across the interface.
atomic layer before it reaches a normal or
Except for a possible difference in the elec-
interstitial site inside. The concentration [i]
tric field, ϕ, a difference in the chemical
is not uniform either. Under equilibrium
potential of the ions may also exist. The
conditions, µ̃ = µ + zqϕ is uniform with
chemical potential, µ, is composed of the
variations in kT ln[i] compensating for a
standard one, µ0 , and the activity term:
nonuniformity in µ0 and zqϕ (see Fig. 5b).
µ = µ0 + kT ln(a) (35) Figure 5(c) shows µ0 (x) + zqϕ(x), zqϕ(x)
and µ̃(x) under current with ∇ µ̃ = 0. The
For macroscopic considerations the stan- moving ion has to overcome a potential
dard chemical potential can be represented barrier of the type shown in Fig. 5(c), and
therefore, the I –V relations have to follow εa , in an adsorbed atom/ion occupied with

Eq. (33a). an electron that is to be transferred to the
metal and the energy potential barrier, Ep ,
2.6.6.4.3 Electron Transfer For electrons the electron has to tunnel through. In the
the distributions of µ0 (x), µ0 (x) − case shown in Fig. 6(a), the energy of the
zqϕ(x), and µ̃(x) are qualitatively similar electron has to be raised at least by the
to those for ions as presented schemati- amount εF − εa to reach degeneracy with
cally in Fig. 5. However, the electron can an empty state in the conduction band
tunnel through the potential barrier if of the metal to allow tunneling. On the
the initial and final states are degener- other hand, in the case shown in Fig. 6(b),
ate. In LSE a potential of the shape of Figs. the electron can tunnel into a degenerate,
5(b) and (c) arose due to the motion of empty state in the conduction band of
the ions and atoms involved in the elec- the metal without thermal excitation. It is
tronation and de-electronation process. In expected that the potential barrier, Ep , and
SSE, however, the motion of the ions is the difference in energy of the electron
much more restricted. One expects, for in the metal and in the atom/ion, depend
instance, that an ion/atom adsorbed on a on a gradient in the electrical potential
metal electrode or an ion/atom at the inter- (not shown). The I –V relations for this
face between the two solids (e.g. SE/metal elementary step are,
electrode) will only oscillate about a mini-
mum. Thus the electron transfer is over a I = I0 f (δ(ϕ))

short distance of ∼1 Å, while the adsorbed 1 − pr pr −qδ(ϕ)/kB T
ion/atom/molecule hardly moves. × − e or
1 − pr,eq pr,eq
Tunneling to a degenerate state, in
SSE, requires thermal excitation only in I = I0 f (δ(ϕ))
one tunneling direction. To see this let
1 − pr qδ(ϕ)/kB T pr
us consider Fig. 6, which exhibits the × e −
conduction band of a metal, the Fermi 1 − pr,eq pr,eq
level, εF , the upper discrete energy level, (37a)
ε ε
Ep Ep
εa
εF εF
εa
Metal Metal
(a) (b)
Fig. 6 Tunneling of an electron through a potential energy
barrier into a metal. (a) Thermally assisted tunneling and (b)
tunneling not requiring thermal excitation.
where f (δ(ϕ)) yields the dependence of is true whether it is the free metal/gas
the tunneling current on the distortion interface or the metal/SE interface. Thus
in the potential barrier by δ(ϕ), pr is the tunneling distance for an electron is
the probability that the acceptor level is significantly smaller in SSE (except when
occupied by an electron and pr,eq is pr un- in LSE the electron transfer is also to a
der equilibrium. The relevant expression species directly adsorbed onto the metal
depends on the transfer direction being electrode. However, the concentration of
thermally activated relative to the polarity uncharged adsorbed acceptors is expected
of the applied voltage. Eq. (37a) also holds to be relatively low). The energy barrier
when the electron is transferred from the heights are expected to be similar in SSE
metal to the atom/ion. The temperature de- and LSE; however, ε, the energy required
pendence is given by the exponential term to achieve degeneracy is lower than the
exp(±qδ(ϕ)/kB T ). Neglecting the effect energy barrier height. We therefore expect
of δ(ϕ) on f and pr , Eq. (37a) reduces to, that electron tunneling at the electrode in
SSE is faster than in LSE.
I = I0 (1 − e−qδ(ϕ)/kB T ) or For comparing the transfer rate of ions
I = I0 (eqδ(ϕ)/kB T − 1) (37b) and electrons in SSE across a potential
barrier, let us consider an example of
Eq. (37a) is different from Eq. (33a) as one oxygen transfer from the gas phase into
term in the brackets does not depend on YSZ, first adsorbing onto a metal electrode
δ(ϕ). with electron transfer and diffusion to the
TPB occurring there:
2.6.6.4.4 Further Discussion of Ion and O2 −−−→ 2Oad −−−→ 2O− ad
Electron Transfer It is questionable if the
rate of an electron transfer in SSE, from −−−→ 2O2− ad −−−→ 2O2− ad,TPB
a metal to an atom/ion (or vice versa), −−−→ 2O2− (YSZ) (38)
can be slower than an ion transfer at
the electrode. The electron transfer is by We notice (see Fig. 6) that electron transfer
tunneling. For a given atom/ion as the in one direction does not require thermal
acceptor (or donor), four factors govern excitation at all and in the opposite
the tunneling rate. First, availability of direction it needs only to be excited by
the electrons (or of an empty site for ε, a fraction of the energy potential
the reverse transfer). Electrons are ample barrier Ep , to reach a degenerate state.
in a metal. Then comes the distance ε is expected to be also lower than the
for tunneling, the energy barrier height, potential barrier that ions have to cross,
and the energy ε required to achieve since the later should be similar to the
degeneracy. Let us first compare the total barrier height in Ep of the electrons.
electron tunneling process in SSE with The high concentration of electrons in
the one in LSE, in which the electron the metal, the short tunneling distance of
tunneling process is different and can be ∼0.1 nm, the barrier height of ∼1 eV, and
rate determining. In LSE the ion is solvated the relative low value of ε should result
and the cloud of solvating molecules keeps in a high value of the exchange current
the ion quite far from the metal electrode. I0 for electron transfer and I0 should be
In SSE, both the ion and the atom are higher than that for ion transfer. Thus if
adsorbed onto the metal electrode. This charge transfer is found experimentally to
be the rate-determining step in SSE then it and that a charge-transfer rds is thus due
is, most likely, an ion transfer that is rate- to ion transfer. Then there is a negligible
determining not an electron transfer [60]. difference µ̃e at the interface. An overpo-
The electron as well as ion-transfer tential is still measured because of the way
currents depend not only on the rate overpotentials, η, at the electrodes are mea-
per adsorbed atom/ion but also on the sured [59, 60, 62, 68]. η is not measured by
atom/ion coverage. If this is low it lowers inserting an inert metallic electrode into
both exchange currents. There can be two the SE and measuring η with respect to
reasons for low coverage: either another the working electrode, that is, at the two
elementary step, which governs the supply sides of the interface. Such a measure-
of the atom/ion, is the rate determining ment would yield η = −µ̃e /q̃. Rather,
step, or the equilibrium coverage is low. the measurement is done between a work-
The latter has been suggested for gold ing electrode and a reference electrode put
cathodes on YSZ for oxygen transfer [67]. on top of the SE exposed to the same
In this case a following rate- determining material (for instance, gas) as the work-
charge-transfer step is, most likely, an ion- ing electrode, thus adopting a well-defined
transfer step. chemical potential µX . The overpotential
The expression for multistep reactions then yields changes in the electrochemical
assumes that all charge transfers occur potential of the ions between the working
at a single interface across the same ϕ. and reference electrodes.
In view of the complexity of the interface The way the measurement of the overpo-
in SSE the multistep reaction can take tential is arranged δ(ϕ) at the rds is equal
place spread over different sites with dif- to the measured potential between the
ferent Galvani potential differences. If so, working electrode and the reversible refer-
Eq. (34a) cannot be applied. For instance, ence electrode, δ(ϕ) = η. Then Eq. (33a)
let us consider the overall reaction in which can be written as,
a neutral adsorbed oxygen atom on a metal
electrode obtains two electrons and finally I = I0 (e(1−β)zqη/kB T − e−βzqη/kB T )
ends inside the SE (for instance, YSZ) at (39a)
an interstitial site: Oad + 2e− → Oi 2− (SE). Eq. (34a) as,
Let us assume that the series of elemen-
tary steps involved are: (1) electronation of I = I0 (eαa qη/kB T − e−αc qη/kB T ) (39b)
the adsorbed oxygen atom: Oad + e− M →
and Eq. (37b) as
Oad − , followed by (2) diffusion along the
metal, (3) transfer of the ion into the SE I = I0 (1 − e−qη/kB T ) or
to an interstitial site, Oad − → Oi − (SE),
and (4) the adding of an electron there: I = I0 (eqη/kB T − 1) (39c)
Oi − (SE) + e− (M) → Oi 2− (SE). It is obvi-
ous that charge transfer occurs at four 2.6.6.4.5 Relating the Overpotential to a
different locations (though all are associ- Difference in a Chemical or Electrochemical
ated with the same electrode reaction) with Potential If one wants to allocate to δ(ϕ)
different ϕ values and Eq. (34a) cannot a relation to a total driving force, this is not
be applied. trivial. A total driving force is associated
It was argued before that in SSE the with a gradient in the electrochemical
transfer of electrons is usually rather rapid potential of electrons, ions, or of the
chemical potential of neutral species. associated with each step and use for the
For example, for electrons, it is the charge transfer Eq. (33a) or Eq. (37b) and
difference in the electrochemical potential Eq. (39a) or Eq. (39c). δ(ϕ) can, in certain
of the electrons, −δ µ̃e /q = −δ(µe )/q + cases, be related to δ µ̃e , δ µ̃ion or δµX across
δ(ϕ), which is not necessarily δ(ϕ). the interface.
However, in some cases δ(µe ) = 0 and In LSE the specie adsorbed on the
−δ µ̃e /q = δ(ϕ), as is the case in LSE metal electrode, M, are in many cases
when the electron is exchanged between neutral ones. Ions are solvated in the
a metal and a redox couple in the LE, LE. For example, in the process Cl− (LE)
the redox potential of which is fixed by →Clad,M + eM − the adsorbed specie is
a high concentration of the redox couple, neutral Cl. In SSE charged specie are ex-
as then δ(µe ) = 0 also under current. In pected to be adsorbed in key reactions, for
SSE, on the other hand, δ(µe ) = 0, is not example, Oad − on Pt is obtained in the re-
necessarily the case for electrons injected action: Oad + eM − → Oad − . Furthermore,
into an SE and electron exchange between adsorbed species with different charge may
a metal and adsorbed species. coexist.
ϕ may represent δ µ̃ion . For a fixed We have mentioned that reference elec-
concentration of ionic defects at the trodes used for electrode characterization
interface δ µ̃ion = zqδ(ϕ). Furthermore, in SSE (and LSE) are chemically active
if electron transfer is fast and δ µ̃e = 0, ones. For determining the distribution of
one can express zqδ(ϕ) in terms of µ̃e in the SE or MIEC, inert metal elec-
δµX = δ µ̃ion . This yields [51, 60, 68], trodes are used [46, 69, 70].
I = I0 (e(1−β)δµX /kB T − e−βδµX /kB T ) (40) 2.6.6.6 Diffusion

Diffusion in both LSE and SSE is governed
2.6.6.5 Conclusions for Charge-Transfer by the gradient of the concentrations of
Processes the relevant species. Owing to the more
To summarize these results for SSE: If a complex morphology of electrodes in SSE,
Butler–Volmer I –V relation is observed the paths for diffusion in SSE are more
experimentally with a noninteger αa and diverse as mentioned before.
αc , then the rds is most likely an ionic
charge transfer. It is important to measure 2.6.7
the I –V relations for both direction of the Galvani Potential Distribution versus
current in order to determine accurately Electrochemical Potential Distribution
αc and αa . If αa and αc are integers then
the rds may be one of the following: (1) an In LSE it is customary to follow the
electron transfer and β = 0 or 1; or more distribution of the Galvani potential, ϕ,
likely (2) an ion transfer with z = 2 and along the EC. ϕ changes within the bulk of
β = 0.5; or (3) the rds is not a charge a single phase (except when a supporting
transfer one. We have noticed that in SSE electrolyte is used) as well as at interfaces.
a multistep reaction can be spread over The drop at interfaces exists also under
close but different interfaces in which case equilibrium. A detailed discussion of the
Eqs. (34a) and (39b) cannot be applied. It distribution of ϕ requires the knowledge
is safe then to calculate each elementary of the specific changes at interfaces for the
step separately, determine the parameters materials involved.
In SSE it is customary to follow the distri- supply of ions through the SE is usually
bution of the electrochemical potential of negligible and does not contribute signifi-
the electrons, µ̃e , and of the ions, µ̃ion . This cantly the reaction. Instead the ions cover
has the advantage that under equilibrium, the electrode material and thus change the
µ̃e and µ̃ion are uniform. Furthermore, un- catalytic activity of the electrode. The ap-
der current for a reversible electrode with plied voltage governs that coverage. This
respect to electron transfer: µ̃e = 0 and is the so-called NEMCA effect developed
for a reversible electrode with respect to ion by Vayenas and coworkers [72–74]. For ex-
transfer: µ̃ion = 0 at the interface. µ̃e ample, a cell used for the oxidation of CH4
can be directly related to the measurable will be: (−)air,Pt|YSZ(O2− )|Pt,CH4 ,O2 (+)
voltage, V , and µ̃ion + zµ̃e to the chem- where CH4 reacts with O2 supplied to-
ical potential difference at the electrodes, gether with it. The rate of reaction is
µX , that can also be measured. governed by the voltage on the EC, which
controls the coverage of the working Pt
2.6.8 electrode by oxygen coming through the
Heterogeneous Catalysis in SSE YSZ SE.
Catalysis plays an important role in SSE 2.6.9

as it does in LSE. At low temperatures, Summary
relevant to LSE, the catalysts are relative
few, usually, precious metals, such as Although based on the same thermo-
Pt. At elevated temperatures, normally dynamic principles, LSE and SSE are
encountered in SSE, the reactions are different in many aspects. This arises be-
accelerated and a wider range of materials cause of difference in material properties,
can serve as catalysts many of them quite morphology, and the geometry of the ECs.
inexpensive. The differences are in bulk properties of
ECs can be used to change the rate the ionic conductors and in the elementary
of chemical reactions. Two modes of steps of electrode processes. Differences
operation are known. In the first one, arise also in applications. Solid state cells
a reactant is supplied via the SE to have an advantage where dry cells are
the working electrode (from a counter required, in microelectronics and at el-
electrode that serves as a reservoir of the evated temperatures. Liquid-state devices
material). That reactant reacts with other are richer in the variety of ions, which can
reactants supplied directly to the working be transported in a large range of LEs, and
electrode via the gas phase. The state of the have significant advantage in large pro-
material that is transferred through the SE duction electrochemical plants where high
(i.e. its chemical potential) is controlled by volumes of materials have to be processed.
the applied voltage [71]. This controls also
the reaction rate at the working electrode. References
In the second mode of operation, all the
reactants are supplied via the gas phase in 1. J. O’M. Bockris, A. K. N. Ready, Modern Elec-
the working electrode compartment. The trochemistry, Plenum/Rosetta, N.Y., 1970,
Vol. 1, 2.
working electrode acts as a catalyst. The 2. A. J. Bard, L. R. Faulkner, Electrochemical
catalytic activity of the electrode is con- Methods, Fundamentals and Application, John
trolled by the voltage applied to the EC. The Wiley & Sons, New York, 1980.
3. K. B. Oldham, J. C. Myland, Fundamentals of 21. J. O’M. Bockris, A. K. N. Ready, Modern Elec-

Electrochemical Science, Academic Press, New trochemistry, Plenum/Rosetta, N.Y., 1970,
York, 1994. Vol. 2, p. 983.
4. E. Gileadi, Electrode Kinetics for Chemists, 22. H. L. Tuller, Mixed conduction, in Nonstoi-
Chemical Engineers and Materials Scientists, chiometric Oxides (Ed.: Toft-Sørensen), Aca-
VCH, New York, 1993. demic Press, New York, 1981, pp. 271–335.
5. P. G. Bruce, (Ed.), Solid State Electrochem- 23. K. Funke, Prog. Solid State Chem. 1993, 22,
istry, Cambridge University Press, New York, 111.
1995. 24. W. Shockley, Bell Syst.Tech. J. 1949, 28, 435.
6. P. J. Gelling, H. J. M. Bouwmeester, (Eds.), 25. I. Riess Solid State Ionics 1994, 69, 43.
The CRC Handbook of Electrochemistry, CRC 26. J. Maier, Angew. Chem., Int. Ed. Engl. 1993,
32, 313.
Press, Boca Raton, 1997.
27. Y. Tsur, I. Riess, Phys. Rev. B 1999, 60, 8138.
7. T. Takahashi, (Ed.), High Conductivity Solid
28. O. Porat, I. Riess, Solid State Ionics 1994, 74,
Ionic Conductors. Recent trends and Applica-
229.
tions, World Scientific, Rive Edge, 1989.
29. O. Porat, I. Riess, Solid State Ionics 1995, 81,
8. C. Julien, G.-A. Nazri, Solid state batteries: 29.
materials, design and optimization, in Elec- 30. J. R. Manning, Diffusion Kinetics for Atoms
tronic Materials: Science and Technology Series in Crystals, van Nostrand, New York, 1968,
(Ed.: H. L. Tuller), Kluwer Academic Pub- pp. 2–9.
lisher, Norwell, 1994. 31. N. F. Mott, E. A. Davis, Electronic Processes in
9. N. Q. Minh, T. Takahashi, Science and Tech- Non-Crystalline Materials, Clarendon Press,
nology of Ceramic Fuel Cells, Elsevier, Amster- Oxford, 1971, Chap. 6.
dam, New York, 1995. 32. F. Lange, M. Martin, Ber. Bunsen-Ges. Phys.
10. K. S. Goto, Solid State Electrochemistry and its Chem. 1997, 101, 176.
Applications to Sensors and Electronic Devices, 33. I. Riess Electrochemistry of mixed ionic elec-
Elsevier, Amsterdam, New York, 1988. tronic conductors, in CRC Handbook of Solid
11. H. Rickert, Solid State Electrochemistry, State Electrochemistry (Eds.: P. J. Gellings,
Springer-Verlag, New York, 1982. H. J. M. Boumeester), CRC Press, Boca Ra-
12. J. Maier, Festkörper-Fehler und Funktion, ton, 1997, pp. 223–268.
Teubner Studienbücher, Stuttgart, 2000. 34. I. Riess, Solid State Ionics, 2000, 136–137,
13. H. Schmalzried, Chemical Kinetics of Solids, 1119.
VCH, New York, 1995. 35. J. Maier, J. Am. Ceram. Soc. 1993, 76, 1212.
14. F. A. Kröger, The Chemistry of Imperfect 36. H.-D. Weimhöfer, Solid State Ionics 1990,
Crystals, North Holland, New York, 1974, 40/41, 530.
Vol. II. 37. I. Riess, J. Phys. Chem. Solids 1986, 47, 129.
15. G. Mamantov in Materials for Advanced 38. J. Maier, M. Holzinger, W. Sitte, Solid state
Batteries (Eds.: D. W. Murphy, J. Broadhead, Ionics 1994, 74, 5.
39. J. Maier, J. Am. Ceram. Soc. 1993, 76, 1218.
B. C. H. Steele), Plenum Press, New York,
40. W. Weppner, R. A. Huggins, Annu. Rev.
1979, pp. 111–122.
Mater. Sci. 1978, 8, 269.
16. A. Bélanger, F. Morin, M. Gauthier et al., in
41. I. Riess, D. S. Tannhauser, Control and mea-
Materials for Advanced Batteries (Eds.: D. W.
surement of stoichiometry in oxides, in
Murphy, J. Broadhead, B. C. H. Steele), Transport in NonStoichiometric Compounds,
Plenum Press, New York, 1979, pp. 211–222. Materials Science Monographs 15 (Ed.:
17. T. A. Ramanarayanan, R. A. Rapp, Metall. J. Nowotny), Elsevier Scientific Publishing
Trans. 1972, 3, 3239. Company, New York, 1982, pp. 503–517.
18. J. Van herle, A. J. McEvoy, Phys. Chem. Solids 42. K. Funke, R. Hackenberg, Ber. Bunsen-Ges.
1994, 55, 339. Phys. Chem. 1972, 76, 883.
19. S. Fujitu, K. Koumoto, H. Yanagida, Solid 43. J. Sohége, K. Funke, Ber. Bunsen-Ges. Phys.
State Ionics 1986, 18, 19, 1146. Chem. 1984, 88, 657.
20. K. B. Oldham, J. C. Myland, Fundamentals of 44. M. H. Hebb, J. Chem. Phys. 1952, 20, 185.
Electrochemical Science, Academic Press, New 45. C. Wagner, Z. Elektrochem. 1956, 60, 4.
York, 1994, p. 9. 46. I. Riess, Solid State Ionics 1996, 91, 221.
47. I. Riess, Solid State Ionics 1991, 44, 207. 62. Da. Yu. Wang, A. S. Nowick, J. Electrochem.
48. K. D. Becker, H. Schmalzried, V. von Soc. 1979, 126, 1155.
Wurmb, Solid State Ionics 1983, 11, 213. 63. F. H. van Heuveln, H. J. M. Boumeester, J.
49. J. R. Macdonald, D. R. Franceschetti, J. Electrochem. Soc. 1997, 144, 134.
Chem. Phys. 1978, 68, 1614. 64. B. A. Boukamp, B. A. van Hassel, I. C. Vinke
50. J. R. Macdonald, (Ed.), Impedance Spec- et al., Electrochim. Acta 1993, 38, 1817.
troscopy, John Wiley & Sons, New York, 1987. 65. Y. Jaing, S. Wang, Y. Zhang et al., Solid State
51. S. B. Adler, J. A. Lane, B. C. H. Steele, J. Ionics 1998, 110, 111.
Electrochem. Soc. 1996, 143, 3554. 66. J. Maier, Angew. Chem. 1993, 105, 558.
52. I. Riess, J. Schoonman in CRC Hand- 67. M. Gödickemeier, K. Sasaki, L. J. Gauckler
book of Solid State Electrochemistry (Eds.: et al., J. Electrochem. Soc. 1997, 144, 1635.
P. J. Gellings, H. J. M. Boumeester), CRC 68. I. Riess, Electrode processes in solid oxide
Press, Boca Raton, 1997, pp. 269–294. fuel cells, in Oxygen Ion and Mixed Con-
53. N. W. Ashcroft, N. D. Mermin, Solid State ductors and their Technological Applications,
Physics, Holt-Saunders Intl., Philadelphia, NATO Advanced Study Institute Series E
1976, pp. 367–371. (Eds.: H. L. Tuller, J. Schoonman, I. Riess),
54. R. N. Ediriweera, P. R. Mason, Solid State Kluwer Academic Publishers, Norwell, 2000,
Ionics 1985, 17, 213. pp. 21–56, Vol. 338.
55. J. Maier in Recent Trends in Superionic 69. J. Mizusaki, K. Fueki, T. Mukaibo, Bull.
Solids and Solid Electrolytes (Eds.: S. Chandra, Chem. Soc. Jpn. 1975, 48, 428.
A. Laskar), Academic Press, New York, 1989, 70. C. Rosenkranz, J. Janek, Solid State Ionics
p. 137. 1995, 82, 95.
56. J.O’M. Bockris, A. K. N. Ready, Modern Elec- 71. T. M. Gür, R. A. Huggins in Fast Ion
trochemistry, Plenum/Rosetta, N.Y., 1970, Transport in Solids (Eds.: P. Vashishta,
Vol. 1, 2, Chaps. 8 & 9. J. N. Mundy, G. K. Shenoy), Elsevier, Ams-
57. R. H. Fowler, L. Nordheim, Proc. R. Soc. terdam, New York, 1979.
London, Ser. A 1928, 119, 173. 72. C. G. Vayenas, S. Bebelis, I. V. Yentekakis
58. S. M. Sze, The Physics of Semiconductor et al., Catal. Today 1992, 11(3) 303–442.
Devices, Wiley, N.Y., 1969. 73. D. Tsiplakides, S. G. Neophytides, O. Enea
59. J. Mizusaki, K. Amano, S. Yamauchi et al., et al., J. Electrochem. Soc. 1997, 144, 2072.
Solid State Ionics 1987, 22, 313. 74. S. G. Neophystides, D. Tsiplakides, P. Stone-
60. I. Riess, M. Gödickemeier, L. J. Gauckler, hart et al., Nature 1994, 370, 45.
Solid State Ionics 1996, 90, 91.
61. A.-M. Svensson, S. Sunde, K. Nisancioglu, J.
Electrochem. Soc. 1997, 144, 2719.
2.7 then proceed and discuss the adsorption

Polyelectrolytes in Solution and at Surfaces of PEs from semidilute solutions. Mean-
field theory profiles of PE adsorption are
Roland R. Netz calculated as a function of surface charge
Max-Planck Institute for Colloids and Inter- density (or surface potential), the amount
faces, Potsdam, Germany
of salt in the system, and the charge frac-
David Andelman tion on the chains. The phenomenon of
School of Physics and Astronomy Raymond charge inversion is reviewed and its rel-
and Beverly Sackler Faculty of Exact Sciences evance to the formation of multilayers is
Tel Aviv University, Ramat Aviv, Israel explained. The review ends with a short
overview of the behavior of grafted PEs.
2.7.1
Abstract
Introduction
This chapter deals with charged poly-
PEs are charged macromolecules that are
mers (polyelectrolytes) in solution and at
extensively studied not only because of
surfaces. The behavior of polyelectrolytes
their numerous industrial applications but
(PEs) is markedly different from that of
also from a purely scientific interest [1–4].
neutral polymers. In bulk solutions, that
The most important property of PEs is
is, disregarding the surface effect, there are
their water solubility giving rise to a
two unique features to charged polymers:
wide range of nontoxic, environmentally
first, owing to the presence of long-ranged
friendly, and cheap formulations. On the
electrostatic repulsion between charged theoretical side, PEs combine the field of
monomers, the polymer conformations statistical mechanics of charged systems
are much more extended, giving rise to with the field of polymer science and
a very small overlap concentration and offer quite a number of surprises and
high solution viscosity. Second, the pres- challenges.
ence of a large number of counterions The polymers considered in this review
increases the osmotic pressure of PE solu- are taken as linear and long polymer
tions, making such polymers water soluble chains, containing a certain fraction of
as this is of great importance to many ap- electrically charged monomers. Chemi-
plications. At surfaces, the same interplay cally, this can be achieved, for example,
between monomer–monomer repulsion by substituting neutral monomers with
and counterion degrees of freedom leads acidic ones. Upon contact with water,
to a number of special properties. In par- the acidic groups dissociate into positively
ticular, the adsorption behavior depends charged protons, which bind immediately
on both the concentration of polymers and to water molecules, and negatively charged
the added salt in the bulk. We first de- monomers. Although this process effec-
scribe the adsorption behavior of single tively charges the polymer molecules, the
PE molecules and discuss the necessary counterions make the PE solution electro-
conditions to obtain an adsorbed layer and neutral on macroscopic length scales.
characterize its width. Depending on the The counterions are attracted to
stiffness of the PE, the layer can be flat and the charged polymers via long-ranged
compressed or coiled and extended. We Coulomb interactions, but this physical
2.7 Polyelectrolytes in Solution and at Surfaces 283
association typically only leads to thermodynamics of flexible chains is well

rather loosely bound counterion clouds developed and the theoretical concepts can
around the PE chains. Because PEs be applied with a considerable degree of
are present in a background of a confidence [5–9]. Long and flexible chains
polarizable and diffusive counterion have a large number of conformations,
cloud, there is a strong influence of a fact that plays a crucial role in deter-
the counterion distribution on the PE mining their behavior in solution. When
structure, as will be discussed at length flexible chains adsorb on surfaces, they
in this review. Counterions contribute form a diffusive adsorption layer extending
significantly toward bulk properties, such away from the surface into the solution.
as the osmotic pressure, and their This is in contrast to semiflexible or rigid
translational entropy is responsible for the chains, which can form dense and compact
generally good water solubility of charged adsorption layers.
polymers. In addition, the statistics of The main parameters used to describe a
PE chain conformations are governed by flexible polymer chain are the polymeriza-
intrachain Coulombic repulsion between tion index N , which counts the number of
charged monomers, resulting in more repeat units or effective monomers along
extended and swollen conformations of the chain, and the Kuhn length a, being
PEs as compared to neutral polymers. the size of one effective monomer or the
All these factors combined are of great distance between two neighboring effec-
importance when considering PE adsorptive monomers. The effective monomer
tion to charged surfaces. We distinguish size ranges from a few Å for synthetic
between physical adsorption, where chain polymers to a few nanometers for biopoly-
monomers are attracted to surfaces via mers [7]. The effective monomer size a is
electrostatic or nonelectrostatic interac- not to be confused with the actual size b
tions, and chemical adsorption, where a of one chemical monomer; in general, a
part of the PE (usually the chain end) is greater than b owing to chain stiffen-
is chemically bound (grafted) to the suring effects, as will be explained in detail
face. In all cases, the long-ranged repulsion later on. In contrast to other molecules
of the dense layer of adsorbed PEs and or particles, a polymer chain contains not
the entropy associated with the counterion only translational and rotational degrees of
distribution are important factors in the freedom but also a vast number of confor-
theoretical description. mational degrees of freedom. For typical
polymers, different conformations are pro-
2.7.2 duced by torsional rotations of the polymer
Neutral Polymers in Solution backbone bonds.
A simple description of flexible chain
Before reviewing the behavior of charged conformations is achieved with the freely
polymers, let us describe some of the jointed chain model in which a polymer
important ideas underlying the behavior consisting of N + 1 monomers is repre-
of neutral polymer chains in solution. sented by N bonds defined by bond vectors
rj with j = 1, . . . N . Each bond vector has
2.7.2.1 Flexible Chain Statistics a fixed length |rj | = a corresponding to
The chains considered in this review are ei- the Kuhn length, but otherwise is allowed
ther flexible or semiflexible. The statistical to rotate freely. For the freely jointed chain
model, the monomer size b equals the R aN ν (3)

effective monomer size a, b = a. Fixing
one of the chain ends at the origin, the po- For an ideal polymer chain (no interactions
sition of the (k + 1)-th monomer is given between monomers), the above result
by the vectorial sum implies ν = 1/2. This result holds only
for polymers in which the attraction

k
between monomers (as compared with
Rk = rj (1) the monomer–solvent interaction) cancels
j =1
the steric repulsion (which is due to the
Because two arbitrary bond vectors are fact that the monomers cannot penetrate
uncorrelated in this simple model, the each other). This situation can be achieved
thermal average over the scalar product in special solvent conditions called theta
of two different bond vectors vanishes, solvents.
rj · rk = 0 for j = k, while the mean- More generally, polymers in solution can
squared bond vector length is simply given experience three types of solvent condi-
by rj2 = a 2 . It follows then that the mean- tions, with theta solvent condition being
squared end-to-end radius is proportional intermediate between ‘‘good’’ and ‘‘bad’’
to the number of monomers solvent conditions. The solvent quality
depends mainly on the specific chem-
2
RN = N a2 (2) istry determining the interaction between
the solvent molecules and monomers.
The same result is obtained for the It also can be changed by varying the
mean quadratic displacement of a freely temperature.
diffusing particle and eludes to the same The solvent is called good when the
underlying physical principle, namely, the monomer–solvent interaction is more
statistics of Markov processes. favorable than the monomer–monomer
In Fig. 1(a) we show a snapshot of one. Single polymer chains in good sol-
Monte-Carlo simulations of a freely jointed vents have ‘‘swollen’’ spatial configura-
chain consisting of 100 monomers, each tions, reflecting the effective repulsion
being represented by a sphere of diameter between monomers. For good solvents, the
b (being equal here to a, the effective steric repulsion dominates and the poly-
monomer size). The bar represents a mer coil takes a more swollen structure,
length of 10b, which according to Eq. (2) characterized by an exponent ν 3/5 [7].
is the average distance between the chain This spatial size of a polymer coil is
ends. Indeed, the end-to-end radius gives much smaller than the extended con-
a good idea of the typical chain radius. tour length L = aN but larger than the
The freely jointed chain model describes size of an ideal chain aN 1/2 . The rea-
ideal Gaussian chains and does not ac- son for this peculiar behavior is entropy
count for interactions between monomers combined with the favorable interaction
that are not necessarily close neighbors between monomers and solvent molecules
along the backbone. Including these in- in good solvents. As we will see later, it
teractions will give a different scaling is the conformational freedom of poly-
behavior for long polymer
chains. The end- mer coils that leads to salient differ-
to-end radius, R = RN2 , can be written
ences between polymer and simple liquid
more generally for N 1 as adsorption.
(a) (b)
(c)
(d)
Fig. 1 Snapshots of Monte-Carlo simulations of a neutral and semiflexible chain

consisting of N = 100 monomers of diameter b, which defines the unit of length. The
theoretical end-to-end radius R is indicated by a straight bar. The persistence lengths used
in the simulations are (a) 0 = 0, corresponding to a freely jointed (flexible) chain, leading
to an end-to-end radius R/b = 10; (b) 0 /b = 2, leading to R/b = 19.8; (c) 0 /b = 10,
leading to R/b = 42.4; and (d) 0 /b = 100, leading to R/b = 85.8.
In the opposite case of ‘‘bad’’ (sometimes 2.7.2.2 Semiflexible Chain Statistics

called poor) solvent conditions, the effective Beside neglecting monomer–monomer
interaction between monomers is attrac- interaction, the freely jointed chain model
tive, leading to collapse of the chains and does not take into account the chain elas-
to their precipitation from the solution ticity, which plays an important role for
(phase separation between the polymer some polymers, and leads to more rigid
and the solvent). It is clear that in this case, structures. This stiffness can be conve-
the polymer size, like any space-filling niently characterized by the persistence
object embedded in three-dimensional length 0 , defined as the length over which
space, scales as N ∼ R 3 , yielding ν = the tangent vectors at different locations
1/3. on the chain are correlated. In other words,
the persistence length gives an estimate for monomer size. The mean-squared end-to-
the typical radius of curvature, while tak- end radius of a semiflexible chain is known
ing into account thermal fluctuations. For and reads [6]
synthetic polymers with trans-cis confor-
mational freedom of the chain backbone, R 2 = 20 L + 220 (e−L/0 − 1) (5)
the stiffness is due to fixed bond angles
where the persistence length is 0 and the
and hindered rotations around individual
total contour length of a chain is L. Two
backbone bonds. This effect is even more
limiting behaviors are obtained for R from
pronounced for polymers with bulky side
Eq. (5): for long chains, L 0 , the chains
chains, such as poly-DADMAC, because
behave as flexible ones, R 2 20 L; while
of steric constraints, and the persistence for rather short chains, L 0 , the chains
length is of the order of a few nanome- behave as rigid rods, R L. Comparison
ters. with the scaling of the freely jointed chain
Biopolymers with a more complex struc- model (Eq. 2) shows that a semiflexible
ture on the molecular level tend to be chain can, for L 0 , be described by a
stiffer than simple synthetic polymers. freely jointed chain model with an effective
Some typical persistence lengths encoun- Kuhn length of
tered in these systems are 0 ≈ 5 mm
for tubulin, 0 ≈ 20 µm for actin, and a = 20 (6)
0 ≈ 50 nm for double-stranded DNA. Be-
cause some of these biopolymers are and an effective number of segments or
charged, we will discuss at length the monomers
L
dependence of the persistence length on N= (7)
20
the electrostatic conditions. In some cases
the main contribution to the persistence In this case the Kuhn length takes
length comes from the repulsion between into account the chain stiffness and is
charged monomers. independent from the monomer length.
To describe the bending rigidity of This monomer size is denoted by b
neutral polymers, it is easier to use a whenever there is need to distinguish
continuum model [6], in which one ne- between the monomer size b and the
glects the discrete nature of monomers. persistence length 0 (or Kuhn length a).
The bending energy (rescaled by the ther- In Fig. 1 we show snapshots taken from
mal energy, kB T ) of a stiff or semiflexible Monte-Carlo simulations of a semiflexible
polymer of contour length L is given by chain consisting of 100 polymer beads
of diameter b. The persistence length
2
0 L d2 r(s) is varied from 0 = 2b (Fig. 1b), over
ds (4)
2 0 ds 2 0 = 10b (Fig. 1c), to 0 = 100b (Fig. 1d).
Comparison with the freely jointed chain
where d2 r(s)/ds 2 is the local curvature model (having no persistence length)
of the polymer. We assume here that is given in Fig. 1(a) (a = b, 0 = 0). It is
the polymer segments are nonextendable, seen that as the persistence length is
that is, the tangent vectors are always increased, the chain structure becomes
normalized, |dr(s)/ds| = 1. Clearly, this more expanded. The theoretical prediction
continuum description will only be good for the average end-to-end radius R (Eq. 5)
if the persistence length is larger than the is shown as the black bar on the figure and
gives a good estimate on typical sizes of water is polarizable and reacts to the
semiflexible polymers. presence of a charge with polarization
charges. In addition, and this is by
2.7.3 far a more important effect, water
Properties of Polyelectrolytes in Solution molecules carry a permanent dipole
moment that partially orients in the
For PEs, electrostatic interactions provide vicinity of charged objects. Note that
the driving force for their salient features for water, ε ≈ 80, so that electrostatic
and have to be included in any theoreti- interactions and self-energies are much
cal description. The reduced electrostatic weaker in water than in air (where ε ≈
interaction between two pointlike charges 1) or some other low-dielectric solvents.
can be written as z1 z2 v(r) where Still, the electrostatic interactions are
especially important in polar solvents
e2 because in these solvents, charges
v(r) = (8)
kB T εr dissociate more easily than in unpolar
is the Coulomb interaction between two solvents.
elementary charges, z1 and z2 are the va- 3. In biological systems and most indus-
lences (or the reduced charges in units trial applications, the aqueous solution
of the elementary charge e), and ε is contains mobile salt ions. Salt ions
the medium dielectric constant. Through- of opposite charge are drawn to the
out this review, all energies are given in charged object and form a loosely
units of the thermal energy kB T . The in- bound counterion cloud around it. They
teraction depends only on the distance r effectively reduce or screen the charge
between the charges. The total electrostatic of the object. The effective (screened)
energy of a given distribution of charges electrostatic interaction between two
is obtained from adding up all pairwise charges z1 e and z2 e in the presence
interactions between charges according of salt ions and a polarizable sol-
to Eq. (8). In principle, the equilibrium vent can be written as z1 z2 vDH (r),
behavior of an ensemble of charged parti- with the Debye–Hückel (DH) poten-
cles (e.g. a salt solution) follows from the tial vDH (r) given on the linear-response
partition function, that is, the weighted level by
sum over all different microscopic con- B −κr
figurations, which – via the Boltzmann vDH (r) = e (9)
r
factor – depends on the electrostatic en-
ergy of each configuration. In practice, The Bjerrum length B is defined as
however, this route is very complicated for
several reasons: e2
B = (10)
εkB T
1. The Coulomb interaction (Eq. 8) is
long-ranged and couples many charged and denotes the distance at which the
particles. Electrostatic problems are Coulombic interaction between two unit
typically many-body problems, even for charges in a dielectric medium is equal
low densities. to thermal energy (kB T ). It is a measure
2. Charged objects in most cases are of the distance below which the Coulomb
dissolved in water. Like any material, energy is strong enough to compete with
the thermal fluctuations; in water at room over the two-body potential (Eq.9). How-
temperatures, one finds B ≈ 0.7 nm. The ever, we will at various points in this review
exponential decay is characterized by the also discuss how to go beyond the DH ap-
so-called screening length κ −1 , which is proximation, for example, in the form of
related to the salt concentration csalt by the nonlinear Poisson–Boltzmann theory
(see Sect. 2.7.5) or a box model for the
κ 2 = 8πz2 B csalt (11) counterion distribution (see Sect. 2.7.6).
where z denotes the valency of the 2.7.3.1 Isolated Polyelectrolyte Chains

screening ions (z : z salt). At physiological We discuss now the scaling behavior of
conditions the salt concentration is csalt ≈ a single semiflexible PE in the bulk, in-
0.1 M and for monovalent ions (z = 1) this cluding chain stiffness and electrostatic re-
leads to κ −1 ≈ 1 nm. pulsion between monomers. For charged
The so-called DH interaction (Eq. 9) polymers, the effective persistence length
embodies correlation effects due to the is increased owing to electrostatic re-
long-ranged Coulomb interactions in a pulsion between monomers. This effect
salt solution using linear-response theory. modifies considerably not only the PE be-
In the following we estimate the range havior in solution but also their adsorption
of validity of this approximation using characteristics.
simple scaling arguments. The number The scaling analysis is a simple exten-
of ions that are correlated in a salt solution sion of previous calculations for flexible
with concentration csalt is of the order (Gaussian) PEs [10–12]. The semiflexible
of n ∼ κ −3 csalt , where one employs the polymer chain is characterized by a bare
screening length κ −1 as the scale over persistence length 0 and a linear charge
which ions are correlated. Using the density τ . Using the monomer length b
definition κ 2 = 8πz2 B csalt , one obtains and the fraction of charged monomers f
n ∼ (z2 B csalt )−3/2 . The average distance
1/3
as parameters, the linear charge density
−1/3
between ions is roughly r ∼ csalt . The can be expressed as τ = f/b. Note that
typical electrostatic interaction between in the limit where the persistence length
two ions in the solution thus is U ∼ is small and comparable to a monomer
1/3
z2 B /r ∼ z2 B csalt , and we obtain U ∼ size, only a single length scale remains,
n −2/3 . Using these scaling arguments, 0 a b.
one obtains that either (1) many ions are Many interesting effects, however, are
weakly coupled together (i.e. n 1 and obtained in the general case treating the
U 1) or (2) a few ions interact strongly persistence length 0 and the monomer
with each other (n U 1). In the first size b as two independent parameters.
case, and in the absence of external fields, In the regime where the electrostatic
the approximations leading to the DH energy is weak, and for long enough con-
approximation (Eq. 9) are valid. tour length L, where L 0 , a polymer
The DH approximation forms a conve- coil will be formed with a radius R un-
nient starting point for treating screening perturbed by the electrostatic repulsion
effects, since (owing to its linear character) between monomers. According to Eq. (5),
the superposition principle is valid and the we get R 2 20 L. To estimate when the
electrostatic free energy is given by a sum electrostatic interaction will be sufficiently
strong to swell the polymer coil, we re- expected to have a conformation charac-
call that the electrostatic energy (rescaled terized by an effective persistence length
by the thermal energy kB T ) of a homoge- eff , larger than the bare persistence length
neously charged sphere of total charge Q 0 , that is, one expects eff > 0 .
and radius R is This effect is clearly seen in Fig. 2,
where we show snapshots of Monte-Carlo
3B Q2
Wel = (12) simulations of a charged chain of 100
5R monomers of size b each and bare persis-
The exact charge distribution inside the tence length 0 /b = 1 and several values
sphere only changes the prefactor of order of κ −1 and τ . The number of monomers in
unity and is not important for the scaling an electrostatic blob can be written accord-
arguments. For a polymer of length L ing to Eq. (14) as Lel /0 = (τ 2 B 0 )−2/3
and line charge density τ , the total charge and yields for Fig. 2(a) Lel /0 = 0.25,
is Q = τ L. The electrostatic energy of a for Fig. 2(b) Lel /0 = 0.63, for Fig. 2(c)
(roughly spherical) polymer coil is then Lel /0 = 1.6, and for Fig. 2(d) Lel /0 = 4.
−1/2
Accordingly, in Fig. 2(d) the electrostatic
Wel B τ 2 L3/2 0 (13) blobs consist of four monomers, and the
weakly charged chain crumples at small
The polymer length at which the electro-
length scales. A typical linear charge den-
static self-energy is of order kB T , that is,
sity reached with synthetic PEs such as
Wel 1, follows as
Polystyrenesulfonate (PSS) is one charge
Lel 0 (B 0 τ 2 )−2/3 (14) per two carbon bonds (or, equivalently, one
charge per monomer), and it corresponds
and defines the electrostatic blob size or to τ ≈ 4 nm−1 . Since for these highly flex-
electrostatic polymer length. We expect a ible synthetic PEs the bare persistence
locally crumpled polymer configuration if length is of the order of the monomer
Lel > 0 , that is, if size, 0 b, the typical charge parameter
for a fully charged PE therefore is roughly
τ B 0 < 1 (15) τ 2 B 0 ≈ 3 and is intermediate between
Fig. 2(a and b). Smaller linear charge den-
because the electrostatic repulsion be-
sities can always be obtained by replacing
tween two segments of length 0 is smaller
some of the charged monomers on the
than the thermal energy and is not suf-
polymer backbone with neutral ones, in
ficient to align the two segments. This
which case the crumpling observed in
is in accord with more detailed calcula-
Fig. 2(d) becomes relevant. Larger bare
tions by Joanny and Barrat [11]. A recent
persistence lengths can be achieved with
general Gaussian variational calculation
biopolymers or synthetic PEs with a con-
confirms this scaling result and in ad-
jugated carbon backbone.
dition yields logarithmic corrections [12].
The question now arises as to what
Conversely, for
are the typical chain conformations at

τ B 0 > 1 (16) much larger length scales. Clearly, they
will be influenced by the repulsions.
electrostatic chain–chain repulsion is al- Indeed, √ in the persistent regime, obtained
ready relevant on length scales comparable for τ (B 0 ) > 1, the polymer remains
to the persistence length. The chain is locally stiff even for contour lengths larger
(a)
(b)
(c)
(d)
Fig. 2 Snapshots of Monte Carlo simulations of a polyelectrolyte chain of N = 100

monomers of size b, taken as the unit length. In all simulations the bare persistence
length is fixed at 0 /b = 1, and the screening length and the charge interactions are
tuned such that the electrostatic persistence length (OSF ) is constant
√ and
OSF /b = 100,√ see Eq. (18). The parameters used√ are (a) κ −1 /b = 50 and τ 2 B 0 = 8;
(b) κ −1 /b = √200 and τ 2 B 0 = 2; (c) κ −1 /b = 800 and τ 2 B 0 = 1/2; and
(d) κ −1 /b = 3200 and τ 2 B 0 = 1/8. Noticeably, the weakly charged chains crumple
at small length scales and show a tendency to form electrostatic blobs.
than the bare persistence length 0 , and Skolnick and Fixman, reads [13–15]
the effective persistence length is given by
B τ 2
OSF = (18)
eff 0 + OSF (17) 4κ 2
and is calculated from the electrostatic
The electrostatic persistence length, first energy of a slightly bent polymer using
derived by Odijk and independently by the linearized DH approximation (Eq. 9).
It is valid only for polymer conformations Gaussian √ regime is obtained in the opposite
that do not deviate too much from the limit of τ (B 0 ) < 0 κ.
rodlike reference state. The electrostatic The electrostatic persistence length is
persistence length gives a sizable contri- visualized in Fig. 3, in which we present
bution to the effective persistence length snapshots of Monte-Carlo simulations of a
only for OSF > 0 . This is equivalent to charged chain consisting of 100 monomers
the condition of size b. The bare persistence length was
fixed at 0 = b, and the charge-interaction
τ B 0 > 0 κ (19) parameter was chosen to be τ 2 B 0 = 2,
close to the typical charge density in
The persistent regime is obtained for experiments on fully charged synthetic
parameters satisfying both conditions (16) PEs. The snapshots correspond to vary- √
and (19). Another regime called the ing screening lengths of (1) κ −1 /b = 2,
(a)
(b)
(c)
Fig. 3 Snapshots of Monte-Carlo simulations of a PE chain consisting of N = 100
monomers of size b. In all simulations, the bare persistence length is fixed at
0 /b = 1 and the charge-interaction parameter is chosen to be τ 2 B√ 0 = 2. The
snapshots correspond to varying screening lengths of (a) κ −1 /b = 2, leading √ to an
electrostatic contribution to the persistence
√ length of OSF /b = 1; (b) κ −1 /b = 18,
leading to OSF /b = 9; and (c) κ −1 /b = 200, leading to OSF /b = 100. According
to the simple scaling principle (Eq. 17), the effective persistence length in the
snapshots (a–c) should be similar to the bare persistence length in Fig. 1(b–d).
leading to an electrostatic contribution to a chain stiffening, so that the PE forms a

to the persistence length√ of OSF = b linear array of electrostatic blobs. To quan-
(Fig. 3a), (2) κ −1 /b = 18, √or OSF = 9b tify this effect, one defines an effective line
(Fig. 3b), and (3) κ −1 /b = 200, equiva- charge density of a linear array of electro-
lent to OSF = 100b (Fig. 3c). According static blobs with blob size Rel (0 Lel ),
to the simple scaling principle (Eq. 17), 1/2
the effective persistence length in the τ Lel Lel
τ̃ τ (20)
snapshots (Fig. 3a–c) should be simi- Rel 0
lar to the bare persistence length in
Fig. 1(b–d), and indeed, the chain struc- Combining Eqs. (18) and (20) gives the
tures in Figs. 3(c) and 1(d) are very similar. effective electrostatic persistence length
Figs. 3(a) and 1(b) are clearly different, al- for a string of electrostatic blobs,
though the effective persistence length 1/3 −2/3
should be quite similar, mostly because τ 2/3
KK B 02 (21)
of self-avoidance effects that are present in κ
charged chains and are discussed in detail
in Sect. 2.7.3.1.2.
For the case in which the polymer crum-
ples on length scales larger than the bare Rel ∼ (ᐉ0 Lel)1/2
bending rigidity, that is, for Lel > 0 or Gaussian
√
τ (B 0 ) < 1, the electrostatic repulsion
between polymer segments is not strong
enough to prevent crumpling on length
scales comparable to 0 , but can give rise
to a chain stiffening on larger length
scales, as explained by Khokhlov and Persistent
Khachaturian [10] and confirmed by Gaus- ᐉKK
sian variational methods [12]. Figure 4
schematically shows the PE structure in
this regime, where the chain on small
scales consists of Gaussian blobs of chain
length Lel , within which electrostatic in-
teractions are not important. On larger Gaussian
length scales, electrostatic repulsion leads
Fig. 4 Schematic view of the four RSW ∼ L1/2

SW
scaling ranges in the Gaussian-
persistent regime. On spatial scales
smaller than Rel , the chain behavior is
Gaussian; on length scales larger than
Rel but smaller than KK , the Gaussian
blobs are aligned linearly. On length
scales up to Lsw , the chain is Swollen
isotropically swollen with an exponent
ν = 1/2, and on even larger length
scales, self-avoidance effects become
R ∼ Ln
important and ν changes to ν = 3/5.
This electrostatic stiffening is only relevant chain crumpling at small length scales
for the so-called Gaussian-persistent regime and chain stiffening at large length scales
valid for KK > Rel , or equivalently occur simultaneously [16–20]. However,
extremely long chains are needed in order
τ B 0 > (0 κ)3/2 (22) to obtain reliable results for the persistence
length, since the stiffening occurs only at
When this inequality is inverted, the
intermediate length scales and therefore
Gaussian-persistence regime crosses over
fitting of the tangent–tangent correlation
to the Gaussian one.
function is nontrivial. Nevertheless, sim-
The crossover boundaries (Eqs. 16, 19,
ulations point to a different scaling than
22) between the various scaling regimes
that in Eq. (21), with a dependence on the
are summarized in Fig. 5. We obtain three
screening length closer to a linear one, in
distinct
√ regimes. In the persistent√ regime,
qualitative agreement with experimental
for τ (B 0 ) > 0 κ and τ (B 0 ) > 1,
results [3]. The situation is complicated by
the polymer takes on a rodlike struc-
the fact that more recent theories for the
ture with an effective persistence length
single PE chain give different results, some
larger than the bare persistence length
confirming the simple scaling results de-
and given by the OSF expression (Eq. 18).
scribed in Eqs. (18) and (21) [12, 21, 22],
In√ the Gaussian-persistent
√ regime, for
some confirming Eq. (18) while criticiz-
τ (B 0 ) < 1 and τ (B 0 ) > (0 κ)3/2 ,
ing Eq. (21) [11, 23, 24]. This issue is not
the polymer consists of a linear array of
resolved and is under intense current in-
Gaussian electrostatic blobs, as shown in
vestigation. For multivalent counterions,
Fig. 4, with an effective persistence length
fluctuation effects can even give rise to a
KK larger than the electrostatic blob size
and given by Eq. (21). PE collapse purely due to electrostatic in-
√ Finally, in the3/2 Gaus-
teractions [25–27], which is accompanied
sian
√ regime, for τ ( B 0 ) < ( 0 κ) and
τ (B 0 ) < 0 κ, the electrostatic repul- by a negative contribution to the effective
sion does not lead to stiffening effects at persistence length [28].
any length scale.
The persistence length KK was also 2.7.3.1.1 Manning Condensation A pe-
obtained from Monte-Carlo simulations culiar phenomenon occurs for highly
with parameters shown in Fig. 2(d), where charged PEs and is known as the Manning
Fig. 5 Schematic phase diagram of a single

semiflexible PE in bulk solution with bare persistence
−1 −1 Gaussian-
length 0 and line charge density τ , exhibiting various κ ᐉ0 persistent
scaling regimes. High salt concentration and small τ 2/3
Persistent
correspond to the Gaussian regime, where the
electrostatic interactions are irrelevant. In the
persistent regime, the polymer persistence length is 1
increased, and in the Gaussian-persistent regime, the Gaussian
polymer forms a persistent chain of Gaussian blobs as
indicated in Fig. 4. The broken line indicates the 1
Manning condensation, at which counterions
condense on the polymer and reduce the effective τ ᐉ1/2
B ᐉ0
1/2
polymer line charge density. We use a log–log plot, 1 1/2
ᐉ0
and the various power-law exponents for the crossover 1/2
zᐉB
boundaries are denoted by numbers.
condensation of counterions [29, 30]. For a into play, concentrating

√ on the persistent
rigid PE represented by an infinitely long regime defined by τ (B 0 ) > 1. The end-
and straight cylinder with a linear charge to-end radius R of a strongly charged PE
density larger than chain shows three distinct scaling ranges.
For a chain length L smaller than the
B τ z = 1 (23) effective persistence length eff , which
where z is the counterion valency, it according to Eq. (17) is the sum of the
was shown that counterions condense bare and electrostatic persistence lengths,
on the oppositely charged cylinder even R grows linearly with the length, R ∼ L.
in the limit of infinite solvent dilution. Self-avoidance plays no role in this case
Real polymers have a finite length, and because the chain is too short to fold back
are neither completely straight nor in on itself.
the infinite dilution limit [31, 32]. Still, For much longer chains, L eff , we
Manning condensation has an experi- envision a single polymer coil as a solution
mental significance for polymer solutions of separate polymer pieces of length eff ,
because thermodynamic quantities, such and treat their interactions using a virial
as counterion activities [33] and osmotic expansion. The second-virial coefficient v2
coefficients [34], show a pronounced sig- of a rod of length eff and diameter d scales
nature of Manning condensation. Locally, as v2 ∼ 2eff d [10]. The chain connectivity is
polymer segments can be considered as taken into account by adding the entropic
straight over length scales comparable chain elasticity as a separate term. The
to the persistence length. The Manning standard Flory theory [7] for a semiflexible
condition (Eq. 23) usually denotes a re- chain is based on writing the free energy
gion where the binding of counterions to F as a sum of two terms
charged chain sections begins to deplete
R2 L/eff 2
the solution from free counterions. F + v2 R 3 (24)
Within the scaling diagram of Fig. 5, the eff L R3
Manning threshold (denoted by a vertical where the first term is the entropic elastic
broken line) is reached typically for charge energy associated with swelling a polymer
densities larger than the one needed to chain to a radius R and the second
straighten out the chain. This holds for term is the second-virial repulsive energy
monovalent ions provided 0 > B , as is proportional to the coefficient v2 and the
almost always the case. The Manning con- segment density-squared. It is integrated
densation of counterions will not have over the volume R 3 . The optimal radius
a profound influence on the local chain R is calculated by minimizing this free
structure since the chain is rather straight energy and gives the swollen radius
already due to monomer–monomer repul-
sion. A more complete description of var- 1/5
v2
ious scaling regimes related to Manning R∼ Lν (25)
eff
condensation, chain collapse, and chain
swelling has recently been given [35, 36]. with ν = 3/5. This swollen radius is
only realized above a minimal chain
2.7.3.1.2 Self-avoidance and Polyelectrolyte length L > Lsw ∼ 7eff /v22 ∼ 3eff /d 2 . For
Chain Conformations Let us now consider elongated segments with eff d or,
how the self-avoidance of PE chains comes equivalently, for a highly charged PE, we
obtain an intermediate range of chain 2.7.3.2 Dilute Polyelectrolyte Solutions

lengths eff < L < Lsw for which the It is natural to generalize the discus-
chain is predicted to be Gaussian and the sion of single-chain behavior to that of
chain radius scales as many PE chains at dilute concentrations.
The dilute regime is defined by cm < cm ∗ ,
1/2
R ∼ eff L1/2 (26) where cm denotes the monomer con-
centration (per unit volume) and cm ∗ is
For charged chains, the effective rod diam-
eter d is given in low salt concentrations the concentration where individual chains
by the screening length, that is, d ∼ κ −1 start to overlap. Clearly, the overlap con-
plus logarithmic corrections. The condi- centration is reached when the average
tion to have a Gaussian scaling range bulk monomer concentration exceeds the
(Eq. 26) thus becomes eff κ −1 . Within monomer concentration inside a polymer
the persistent and the Gaussian-persistent coil. To estimate the overlap concentra-
tion cm∗ , we simply note that the average
scaling regimes depicted in Fig. 5, the ef-
fective persistence length is dominated by monomer concentration inside a coil with
the electrostatic contribution and given by radius R bN ν is given by
Eqs. (18) and (21), respectively, which in N
∗
turn are always larger than the screen- cm N 1−3ν b−3 (27)
R3
ing length κ −1 . It follows that a Gaussian
scaling range (Eq. 26) always exists be- For ideal chains with ν = 1/2, the over-
lap concentration scales as cm ∗ ∼ N −1/2
low the asymptotic swollen scaling range
(Eq. 25). This situation is depicted in and thus decreases slowly as the polymer-
Fig. 4 for the Gaussian-persistent scaling ization index N increases. For rigid poly-
regime, where the chain shows two dis- mers with ν = 1, the overlap concentration
∗ ∼ N −2 and decreases strongly
scales as cm
tinct Gaussian scaling ranges at the small
and large length scales. This multihier- as N increases. This means that the dilute
archical scaling structure is only one of regime for stiff PE chains corresponds
the many problems one faces when trying to extremely low monomer concentra-
to understand the behavior of PE chains, tions. For example, taking a monomer size
be it experimentally, theoretically, or by b = 0.254 nm and a polymerization index
simulations. of N = 104 , the overlap concentration be-
comes cm ∗ ≈ 6 × 10−7 nm−3 ≈ 10−3 mM,
A different situation occurs when the
polymer backbone is under bad-solvent which is a very small concentration.
conditions, in which case an intricate inter- The osmotic pressure in the dilute
play between electrostatic chain swelling regime in the limit cm → 0 is given by
and short-range collapse occurs [37]. Quite f cm cm
recently, this interplay was theoretically = + (28)
kB T z N
shown to lead to a Rayleigh instability in
the form of a necklace structure consist- and consists of the ideal pressure of non-
ing of compact beads connected by thin interacting counterions (first term) and
strings [38–41]. Small-angle X-ray scatter- polymer coils (second term). Note that
ing on solvophobic PEs in a series of polar since the second term scales as N −1 , it
organic solvents of various solvent quality is quite small for large N and can be
could qualitatively confirm these theoreti- neglected. Hence, the main contribution
cal predictions [42]. to the osmotic pressure comes from the
counterion entropy. This entropic term ex- will be discussed below. An important
plains also why charged polymers can be length scale, schematically indicated in
dissolved in water even when their back- Fig. 6, is the mesh size or correlation
bone is quite hydrophobic. Precipitation length ξ , which measures the length
of the PE chains will also mean that the below which entanglement effects between
counterions are confined within the pre- different chains are unimportant. The
cipitate. The entropy loss associated with mesh size can be viewed as the polymer
this confinement is too large and keeps the (blob) scale below which single-chain
polymers dispersed in solution. In con- statistics are valid. A semidilute solution
trast, for neutral polymers there are no can be thought of being composed of
counterions in solution. Only the second roughly a close-packed array of polymer
term in the osmotic pressure exists and blobs of size ξ .
contributes to the low osmotic pressure of The starting point for the present
these polymer solutions. In addition, this discussion is the screened interaction
can explain the trend toward precipitation between two charges immersed in a
even for very small attractive interactions semidilute PE solution containing charged
between neutral polymers. polymers, their counterions, and, possibly,
additional salt ions. Screening in this case
2.7.3.3 Semidilute Polyelectrolyte Solution is produced not only by the ions but also
In the semidilute concentration regime, by the charged chain segments that can
cm > cm ∗ , different polymer coils are be easily polarized and shield any free
strongly overlapping, but the polymer so- charges.
lution is still far from being concentrated. Using the random-phase approximation
This means that the volume fraction of the (RPA), the effective DH interaction can be
monomers in solution is much smaller written in Fourier space as [49, 50]
than unity, b3 cm 1. In this concentra-
tion range, the statistics of counterions and
+
polymer fluctuations are intimately con-

+
nected. One example in which this feature

+
−
+
is particularly prominent is furnished by −

−
+
+
neutron and X-ray scattering from semidi- −

− −
+
lute PE solutions [43–48]. −

+
−
+
− −
The structure factor S(q) shows a
+
+
+
pronounced peak, which results from − + + +

+ + +
+ + +
+ + − −
+
a competition between the connectivity

+
−
of polymer chains and the electrostatic − −
+
− −
+
repulsion between charged monomers, as −

− − −
+
+
−
− −
+
−
+
−
−
+
Fig. 6 Schematic view of the chain − − −

structure in the semidilute +
+
+ +
+ + +
+ + + + + +
concentration range. The mesh size ξ is +
+
about equal to the effective polymer

persistence length eff and to the
screening length κ −1 (if no salt is added
ξ ≈ ᐉeff ≈κ −1
to the system).
vRPA (q) = length b, by its persistence length eff . In

1 + v2 cm S0 (q) general, this form factor is a complicated
−1 −1 function that cannot be written down in
cm f 2 S0 (q) + vDH (q) + v2 cm vDH (q)S0 (q)
closed form [51, 52]. However, one can
(29)
separate between three different ranges of
recalling that cm is the average density
wave numbers q, and within each range
of monomers in solution and f is the
the form factor shows a rather simple scal-
fraction of charged monomers on each of
ing behavior, namely,
the PE chains. The second-virial coefficient
of monomer–monomer interactions is v2 S0−1 (q)
and the single-chain form factor (discussed 
 6
below) is denoted by S0 (q). In the case 
 N −1 for q2 <

 N beff
in which no chains are present, cm = 0, 

 q 2 beff 6 36
the RPA expression reduces to vRPA (q) = for < q2 <
vDH (q), the Fourier-transform of the DH 
 6 N beff π 2 2eff


potential of Eq. (9), given by 
 qb 36

 π for < q2
π 2eff
2
4πB
vDH (q) = (30) (32)
q2 + κ2
For small wave numbers the polymer
As before, κ −1 is the DH screening length, acts like a point scatterer, while in the
which is due to all mobile ions. We intermediate wave number regime the
can write κ 2 = κsalt
2 + 4πz f c , where
B m polymer behaves like a flexible, Gaussian
2 = 8πz2 c polymer, and for the largest wave numbers
κsalt B salt describes the screen-
ing due to added salt of valency z : z the polymer can be viewed as a stiff rod.
and concentration csalt , and the second One of the most interesting features of
term describes the screening due to semidilute PE solutions is the fact that the
the counterions of the PE monomers. structure factor S(q) shows a pronounced
Within the same RPA approximation, the peak. For weakly charged PEs, the peak po-
1/4
monomer–monomer structure factor S(q) sition scales as q ∼ cm with the monomer
of a polymer solution with monomer den- density [45], in agreement with the afore-
sity cm is given by [49, 50] mentioned RPA results for charged poly-
mers [49, 50]. We now discuss the scaling
S0−1 (q) of the characteristic scattering peak within
S −1 (q) = f 2 vDH (q) + + v2
cm the present formalism. The position of the
(31) peak follows from the inverse structure
The structure factor (or scattering func- factor (Eq. 31), via ∂S −1 (q)/∂q = 0, which
tion) only depends on the form factor of leads to the equation
an isolated, noninteracting polymer chain,
S0 (q), the second-virial coefficient, v2 , the q 2 + κsalt
2
+ 4πzB f cm
fraction f of charged monomers, and the 1/2
interaction between monomers, which in 8πqB f 2 cm
= (33)
the present case is taken to be the DH ∂S0−1 (q)/∂q
potential vDH (q). The structure factor of
a noninteracting semiflexible polymer is In principle, there are two distinct scaling
characterized, in addition to the monomer behaviors possible for the peak, depending
on whether the chain form factor of fraction f = 0.5 and with no added salt.
Eq. (32) exhibits flexible-like or rigid-like As the monomer density decreases (bot-
scaling. Concentrating now on the flexible tom to top in the figure), the peak moves
case, that is, the intermediate q-range in to smaller wave numbers and sharpens, in
Eq. (32), the peak of S(q) scales as agreement with previous implementations
1/4 of the RPA. In Fig. 7(b) we show the same
24πB f 2 cm
q∗ (34) data in a different representation. Here we
beff clearly demonstrate that the large-q region
in agreement with experimental results. A already is dominated by the 1/q behavior
peak is only obtained if the left-hand side of of the single-chain structure factor, S0 (q).
Eq. (33) is dominated by the q-dependent Since neutron-scattering data easily extend
part, that is, if (q ∗ )2 > κsalt
2 + 4πz f c .
B m
to wave numbers as high as q ∼ 5 nm−1 ,
In Fig. 7(a) we show density-normalized the stiff rodlike behavior in the high q-
scattering curves for a PE solution char- limit, exhibited on such a plot, will be
acterized by the persistence length eff = important in interpreting and fitting ex-
1 nm (taken to be constant and thus inde- perimental data even at lower q-values.
pendent of the monomer concentration), In a semidilute solution there are three
with monomer length b = 0.38 nm (as ap- different, and in principle, independent
propriate for poly-DADMAC) and charge length scales: the mesh size ξ , the
30
20
S(q) /cm
10
0
0 1 2 3 4 5
q
(a) [nm−1]
20
Fig. 7(a) RPA prediction for the
rescaled structure factor S(q)/cm of a
15 semidilute PE solution with persistence
length eff = 1 nm, monomer length
q S(q) /cm
b = 0.38 nm, and charge fraction

10 f = 0.5 in the salt-free case. The
monomer densities are (from bottom to
5
top) cm = 1 M, 0.3 M, 10 mM, 3 mM,
1 mM, and 0.3 mM. (b) For the same
series of cm values as in (a), the
0 structure factor is multiplied by the wave
0 1 2 3 4 5 number q. The semiflexibility becomes
q more apparent because for large q the
(b) [nm−1] curves tend toward a constant.
screening length κ −1 , and the persistence semidilute PE solutions, all three length-
length eff . In the absence of added salt, scales scale in the same manner with cm ,
−1/2
the screening length scales as namely, as ∼ cm , as is known also from
experiments [43, 44, 54] and previous theo-
κ −1 ∼ (zB f cm )−1/2 (35) retical calculations [55, 56]. In simulations
of many PE chains, the reduction of the
Assuming that the persistence length is
chain size due to screening by PE chains
larger or of the same order of magnitude as
was clearly seen [57–60].
the mesh size, as is depicted in Fig. 6, the
polymer chains can be thought of straight
2.7.4
segments between different crossing links.
Adsorption of a Single Polyelectrolyte Chain
Denoting the number of monomers inside
a correlation blob as g, this means that ξ ∼
After reviewing bulk properties of PE
bg. The average monomer concentration
solutions, we elaborate on the adsorption
scales as cm ∼ g/ξ 3 , from which we
diagram of a single semiflexible PE on an
conclude that
oppositely charged substrate. In contrast
ξ ∼ (bcm )−1/2 (36) to the adsorption of neutral polymers, the
resulting phase diagram shows a large
Finally, the persistence length within a region where the adsorbed polymer is
semidilute PE solution can be calculated flattened out on the substrate and creates
by considering the electrostatic energy cost a dense adsorption layer.
for slightly bending a charged rod. In PE Experimentally, the adsorption of char-
solutions, it is important to include in ged polymers on charged or neutral sub-
addition to the screening by salt ions the strates has been characterized as a function
screening due to charged chain segments. of the polymer charge, chemical com-
This can be calculated by using the RPA position of the substrate, pH, and ionic
interaction (Eq. 29). Since the screening strength of the solution [61, 62], as well as
due to polymer chains is scale-dependent the substrate charge density [63–68]. Re-
and increases for large separations, a q- peated adsorption of anionic and cationic
dependent instability is encountered and PEs can lead to well-characterized mul-
leads to a persistence length [53] tilayers on planar [69–74] and spherical
substrates [75–77]. Theoretically, the ad-
−1/2
sd
OSF ∼ (bcm ) (37) sorption of PEs on charged surfaces
poses a much more complicated prob-
where the ‘‘sd’’ superscript stands for
lem than the corresponding adsorption
‘‘semidilute.’’ This result is a generaliza-
of neutral polymers. The adsorption pro-
tion of the OSF result for a single chain
cess results from a subtle balance between
and applies to semidilute solutions. Com-
electrostatic repulsion between charged
paring the three lengths, we see that
monomers, leading to chain stiffening,
and electrostatic attraction between the
sd zB f −1
ξ ∼ OSF ∼ κ (38) substrate and the polymer chain. The
b
√ adsorption problem has been treated the-
Since the prefactor (B f/b) for syn- oretically employing the uniform expan-
thetic fully charged polymers is roughly sion method [78] and various continuous
of order unity, one finds that for salt-free mean-field theories [79–83]. In all these
works, the polymer density is taken to by the Gouy-Chapman (GC) theory for
be constant in directions parallel to the counterion distribution close to a charged
surface. surface. Although these effects are clearly
The adsorption of a single semiflexible important, it is difficult to include them
and charged chain on an oppositely systematically, and we remain at the lin-
charged plane [84] can be treated as a earized DH level.
generalization of the adsorption of flexible Because of the confinement in the
polymers [85]. A PE characterized by the adsorbed layer, the polymer feels an
linear charge density τ is subject to entropic repulsion. If the layer thickness
an electrostatic potential created by σ , δ is much smaller than the effective
the homogeneous surface charge density persistence length of the polymer, eff ,
(per unit area). Because this potential as depicted in Fig. 8(a), a new length scale,
is attractive for an oppositely charged the so-called deflection length λ, enters the
substrate, we consider it as the driving description of the polymer statistics. The
force for the adsorption. More complex deflection length λ measures the average
interactions are neglected. They are due to distance between two contact points of
the dielectric discontinuity at the substrate the polymer chain with the substrate. As
surface and are due to the impenetrability shown by Odijk, the deflection length
of the substrate for salt ions. 1/3
scales as λ ∼ δ 2/3 eff and is larger than
An ion in solution has a repulsive the layer thickness δ but smaller than
interaction from the surface when the the persistence length eff [86, 87]. The
solution dielectric constant is higher than entropic repulsion follows in a simple
that of the substrate. This effect can lead to manner from the deflection length by
desorption for highly charged PE chains. assuming that the polymer loses roughly
On the contrary, when the substrate is a free energy of one kB T per deflection
a metal, there is a possibility to induce length.
PE adsorption on noncharged substrates On the other hand, if δ > eff , as shown
or on substrates bearing charges of the in Fig. 8(b), the polymer forms a random
same sign as the PE. See Ref. [84] for more coil with many loops within the adsorbed
details. layer. For a contour length smaller than
Within the linearized DH theory, the L ∼ δ 2 /eff , the polymer obeys Gaussian
electrostatic potential of a homogeneously statistics and decorrelates into blobs with
charged plane is an entropic cost of one kB T per blob. The
entropic repulsion force per monomer unit
Vplane (x) = 4πB σ κ −1 e−κx (39)
length is thus [86, 87]
Assuming that the polymer is adsorbed −5/3 −1/3
δ eff for δ eff
over a layer of width δ smaller than frep ∼
the screening length κ −1 , the electrostatic eff δ −3 for δ eff
(41)
attraction force per monomer unit length
where we neglected a logarithmic correc-
can be written as
tion factor that is not important for the
fatt = −4πB σ τ (40) scaling arguments. As shown in the pre-
ceding section, the effective persistence
We neglect nonlinear effects due to coun- length eff depends on the screening
terion condensation. They are described length and the line charge density; in
Fig. 8 (a) Schematic picture of the λ

adsorbed polymer layer when the
effective persistence length is larger δ
than the layer thickness, eff > δ. The
distance between two contacts of the (a)
polymer with the substrate, the so-called
deflection length, scales as
1/3
λ ∼ δ 2/3 eff . (b) Adsorbed layer for the
δ
case when the persistence length is
smaller than the layer thickness,
eff < δ. In this case the polymer forms
a random coil with many loops and a (b)
description in terms of a flexible
polymer model becomes appropriate.
essence, one has to keep in mind that (for κeff > 1)

eff is larger than 0 for a wide range of pa-
rameters because of electrostatic stiffening κ 5/3
σ∗ ∼ 1/3
(43)
effects. τ B eff
The situation is complicated by the
fact that the electrostatic contribution to For σ > σ ∗ , the polymer is adsorbed and
the persistence length is scale-dependent localized over a layer with a width smaller
and decreases as the chain is bent at than the screening length (and with the
length scales smaller than the screening condition eff > κ −1 also satisfying δ <
length. This leads to modifications of eff ). As σ is decreased, the polymer
the entropic confinement force (Eq. 41) abruptly desorbs at the threshold σ = σ ∗ .
if the deflection length is smaller than In the Gaussian regime, the effective
the screening length. As can be checked persistence length eff is given by the bare
explicitly, all results reported here are not persistence length 0 and the desorption
changed by these modifications. threshold is obtained by replacing eff by
The equilibrium layer thickness follows 0 in Eq. (43), that is,
from equating the attractive and repulsive
forces (Eqs. 40 and 41). For rather stiff κ 5/3
σ∗ ∼ 1/3
(44)
polymers and small layer thickness, δ < τ B 0
κ −1 < eff , we obtain
In the persistent regime, we have eff ∼
δ ∼ (B σ τ eff )−3/5
1/3
(42) OSF with OSF given by Eq. (18). The ad-
sorption threshold follows from Eq. (43) as
For a layer thickness corresponding to
the screening length, δ ≈ κ −1 , scaling κ 7/3
σ∗ ∼ 4/3
(45)
arguments predict a rather abrupt desorp- τ 5/3 B
tion transition [84]. This is in accord with
previous transfer-matrix calculations for Finally, in the Gaussian-persistent regime,
a semiflexible polymer bound by short- we have an effective line charge density
ranged (square-well) potentials [88–91]. from Eq. (20) and a modified persis-
Setting δ ∼ κ −1 in Eq. (42), we obtain an tence length (Eq. 21). For the adsorption
expression for the adsorption threshold threshold, we obtain from Eq. (43)
5/9 a flexible polymer (Eqs. 47–48) are in

∗ κ 7/3 0
σ ∼ 7/9
(46) agreement with previous results [78].
τ 5/9 B In Fig. 9 we show the desorption transi-
Let us now consider the opposite limit, tions and the line at which the adsorbed
eff < κ −1 . From Eq. (42), we see that the layer crosses over from being flat, δ < eff ,
layer thickness δ is of the same order to being crumpled or coiled, δ > eff . The
as eff for B σ τ 2eff ∼ 1, at which point underlying PE behavior in the bulk, as
the condition δ eff used in deriving shown in Fig. 5, is denoted by broken lines.
Eq. (42) breaks down. If the layer thickness We obtain two different phase diagrams,
is larger than the persistence length but depending on the value of the parameter
smaller than the screening length, eff < 3/2 1/2
δ < κ −1 , the prediction for δ obtained from = σ 0 B (49)
balancing Eqs. (40) and (41) becomes
For strongly charged surfaces, > 1, we
1/3
eff obtain the phase diagram as in Fig. 9(a),
δ∼ (47) and for weakly charged surfaces, < 1, as
B σ τ
in Fig. 9(b). We√ see that strongly charged
From this expression we see that δ has the PEs, obeying τ (0 B ) > 1, always adsorb
same size as the screening length κ −1 for in flat layers. The scaling of the desorption
transitions is in general agreement with
eff κ 3
σ∗ ∼ (48) recent computer simulations of charged
τ B PEs [92].
This in fact denotes the location of a
continuous adsorption transition at which 2.7.4.1 Adsorption on Curved Substrates
the layer grows to infinity. The scaling Adsorption of PEs on curved sub-
results for the adsorption behavior of strates is of importance because PEs are
κ−1ᐉ0−1
κ−1ᐉ 0−1 Adsorbed-
Adsorbed coiled Adsorbed-flat
coiled 5/21
Adsorbed Σ −2/3 1/3
1/3 flat
1 Desorbed 5/7
1
3/5
Σ −2/3 Desorbed 5/7
τ ᐉ1/2
Β ᐉ0
1/2 τ ᐉ1/2
Β ᐉ0
1/2
(a) Σ −1 1 Σ 3/2 (b) Σ 1
Fig. 9 Adsorption scaling diagram shown on a log–log plot for (a) strongly charged
3/2 1/2
surfaces, = σ 0 B > 1, and for (b) weakly charged surfaces, < 1. We find a
desorbed regime, an adsorbed phase in which the polymer is flat and dense, and an
adsorbed phase in which the polymer shows loops. It is seen that a fully charged PE is
expected to adsorb in a flat layer, whereas charge-diluted PEs can form coiled layers with
loops and dangling ends. The broken lines denote the scaling boundaries of PE chains in
the bulk as shown in Fig. 5. The numbers on the lines indicate the power-law exponents of
the crossover boundaries between the regimes.
Fig. 10 Numerically determined 150

adsorption diagram for a charged
semiflexible polymer of length
Wrapped
L = 50 nm, linear charge density
τ = 6 nm−1 , persistence length 100
0 = 30 nm, interacting with an
oppositely charged sphere of radius
Z
Rp = 5 nm. Shown is the main
transition from the unwrapped 50
configuration (at the bottom) to the
wrapped configuration (at the top) as
a function of sphere charge Z and Unwrapped
inverse screening length κ. Wrapping 0
0 2 4 6 8 10
is favored at intermediate salt concen-
trations. The parameters are chosen κ
for the problem of DNA-histone com-
[nm−1]
plexation. (Adapted from Ref. [99].)
widely used to stabilize colloidal suspen- attraction between the sphere and the
sions [93] and to fabricate hollow poly- DNA, and electrostatic repulsion between
meric shells [75–77]. When the curvature the DNA segments with respect to the
of the small colloidal particles is large chain configuration [99]. Configurational
enough, it can lead to a much more fluctuations away from this ground state
pronounced effect for PE adsorption as are unimportant for such stiff polymers.
compared with neutral polymer. This is We show in Fig. 10 the main transition
mainly due to the fact that the electrostatic between an unwrapped state, at low sphere
energy of the adsorbed PE layer depends charge Z, and the wrapped state, at large
sensitively on curvature [94–98]. Bending sphere charge Z. It is seen that at values
a charged polymer around a small sphere of the sphere charge between Z = 10 and
costs a large amount of electrostatic en- Z = 130, the wrapping only occurs at in-
ergy, which will disfavor adsorption of termediate values of the inverse screening
1/2
long, strongly charged PE at very low salt length κ ∼ csalt . At low salt concentrations
concentrations. (lower left corner in the phase diagram),
In Fig. 10 we show the adsorption phase the self-repulsion between DNA segments
diagram of a single stiff PE of finite length prevents wrapping, while at large salt
that interacts with an oppositely charged concentrations (lower right corner in the
sphere of charge Z (in units of e). The diagram), the electrostatic attraction is not
specific parameters were chosen as ap- strong enough to overcome the mechani-
propriate for the complexation of DNA (a cal bending energy of the DNA molecule.
negatively charged, relatively stiff biopoly- These results are in good agreement
mer) with positively charged histone pro- with experiments on DNA/histone com-
teins, corresponding to a DNA length of plexes [100]. Interestingly, the optimal salt
L = 50 nm, a chain persistence length of concentration, where a minimal sphere
0 = 30 nm, and a sphere radius of Rp = charge is needed to wrap the DNA, oc-
5 nm. The phase diagram was obtained by curs at physiological salt concentration,
minimization of the total energy including for κ −1 ≈ 1 nm. For colloidal particles
bending energy of the DNA, electrostatic of larger size and for flexible synthetic
polymers, configurational fluctuations be- In many experiments, the total amount

come important. They have been treated of polymer surface excess is measured
using a mean-field description in terms as a function of the bulk polymer concen-
of the average monomer density profile tration, pH, and/or ionic strength of the
around the sphere [101, 102]. bulk solution [103–110]. For reviews see,
for example, Refs. [61, 62, 111, 112]. More
2.7.5 recently, spectroscopy [105] and ellipsom-
Adsorption from Semidilute Solutions etry [109] have been used to measure the
width of the adsorbed PE layer. Other tech-
So far we have been reviewing the niques such as neutron-scattering can be
behavior of single PE chains close to a employed to measure the entire profile
charged wall (or surface). This will now cm (x) of the adsorbed layer [113, 114].
be extended to include adsorption of PEs In spite of the difficulties in treating
from bulk (semidilute) solutions having theoretically PEs in solution because of
a bulk concentration cm b . As before, the the delicate interplay between the chain
chains are assumed to have a fraction connectivity and the long range nature
f of charged monomers, each carrying of electrostatic interactions [1, 9, 115,
a charge e, resulting in a linear charge 116], several simple approaches treating
density τ = f/b on the chain. The solution adsorption exist.
can also contain salt (small ions) of One approach is a discrete multi-Stern
concentration csalt , which is directly related layer model [117–121], where the system
to the DH screening length, κ −1 . For is placed on a lattice whose sites can
clarity purposes, the salt is assumed to be be occupied by a monomer, a solvent
monovalent (z = 1) throughout Sect. 2.7.5. molecule, or a small ion. The electrostatic
We will consider adsorption only onto potential is determined self-consistently
a single flat and charged surface. Clearly, (mean-field approximation) together with
the most important quantity is the profile concentration profiles of the polymer and
of the polymer concentration cm (x) as small ions.
function of x, the distance from the wall. In another approach, the electrostatic
Another useful quantity is the polymer potential and the PE concentration
surface excess per unit area, defined as are treated as continuous functions [78,
∞ 80–82, 122–125]. These quantities are
= [cm (x) − cm b
] dx (50) obtained from two coupled differential
0 equations derived from the total free
Related to the surface excess is the energy of the system. We will review
amount of charges (in units of e) carried the main results of the latter approach,
by the adsorbing PE chains, f . In presenting numerical solutions and
some cases the polymer carries a higher scaling arguments of the mean-field
charge (per unit area) than the charged profiles.
surface itself, f > σ , and the surface
charge is overcompensated by the PE, 2.7.5.1 Mean-field Theory and its Profile
as we will see later. This does not Equations
violate charge neutrality in the system The charge density on the polymer
because of the presence of counterions chains is assumed to be continuous and
in solution. uniformly distributed along the chains.
Further treatments of the polymer charge where the second-virial coefficient v2 is of

distribution (annealed and quenched mod- order a 3 . The last term couples the system
els) can be found in Refs. [81, 82, 123]. to a polymer reservoirvia a chemical
Within mean-field approximation, the potential µp , and φb = cm b is related to
free energy of the system can be ex- the bulk monomer concentration, cm b .
pressed in terms of the local electrostatic The entropic contribution of the small
potential ψ(r), the local monomer concen- (monovalent) ions is
tration cm (r), and the local concentration

of positive and negative ions c± (r). The Fions (r) = [ci ln ci − ci − csalt ln csalt
mean-field approximation means that the i=±
influence of the charged surface and the
interchain interactions can be expressed + csalt ] − µi (ci − csalt ) (53)
in terms of an external potential that will
determine the local concentration of the where ci (r) and µi are, respectively,
monomers, cm (r). This external potential the local concentration and the chemical
depends both on the electrostatic potential potential of the i = ± ions, while csalt is
and on the excluded volume interactions the bulk concentration of salt.
between the monomers and the solvent Finally, the electrostatic contributions
molecules. The excess free energy with re- (per kB T ) are
spect to the bulk can then be calculated
using another important approximation, Fel (r)
ε
the ground state dominance. This ap-
f eφ 2 ψ + ec+ ψ − ec− ψ − |∇ψ|2
proximation is used often for neutral = 8π
polymers [9] and is valid for very long poly- kB T
mer chains, N 1. It is then convenient (54)
to introduce the polymer order parameter The first three terms are the electrostatic
φ(r), where cm (r) = |φ(r)|2 , and to express energies of the monomers, the positive
the adsorption free energy F in terms of ions, and the negative ions, respectively;
φ and ψ (and in units of kB T ) [80–82, f is the fractional charge carried by
122–124] one monomer. The last term is the self-
energy of the electric field where ε is
F= dr{Fpol (r) + Fions (r) + Fel (r)} the dielectric constant of the solution.
(51) Note that the electrostatic contribution
The polymer contribution is (Eq. 54) is equivalent to the well-known
result: (ε/8πkB T ) dr|∇ψ|2 plus surface
a2 1 terms. This can be seen by substituting the
Fpol (r) = |∇φ|2 + v2 (φ 4 − φb4 )
6 2 Poisson–Boltzmann equation (as obtained
− µp (φ 2 − φb2 ) (52) below) into Eq. (54) and then integrating
by parts.
where the first term is the polymer elastic Minimization of the free energy
energy. Throughout this section we restrict (Eqs. 51–54) with respect to c± , φ, and
ourselves to flexible chains and treat the ψ yields a Boltzmann distribution for
Kuhn length a and the effective monomer the density of the small ions, c± (r) =
size b as the same parameter. The second csalt exp(∓eψ/kB T ), and two coupled
term is the excluded volume contribution differential equations for φ and ψ:

2 8πe eψ The boundary conditions for φ(x) depend
∇ ψ(r) = csalt sinh
ε kB T on the nature of the short-range non-
4πe electrostatic interactions of the monomers
− (f φ 2 − f φb2 eeψ/kB T ) and the surface. For simplicity, we take
ε
(55) a nonadsorbing surface and require that
the monomer concentration will vanish
a2 2 f φeψ there:
∇ φ(r) = v2 (φ 3 − φb2 φ) +
6 kB T
(56) φ|x=0 = 0 (58)
Equation (55) is a generalized Pois-
son–Boltzmann equation including the We note that the boundary conditions
free ions and the charged polymers. The chosen in Eqs. (57) to (58) model the
first term represents the salt contribu- particular situation of electrostatic attrac-
tion and the second term is due to the tion in competition with a short-range
charged monomers and their counteri- (steric) repulsion of nonelectrostatic ori-
ons. Equation (56) is a generalization of gin. Possible variations of these boundary
the self-consistent field equation of neu- conditions include surfaces with a constant
tral polymers [9]. In the bulk, the above surface charge (discussed later) and sur-
equations are satisfied by setting ψ → 0 faces with a nonelectrostatic short-range
and φ → φb . attractive (or repulsive) interaction with
the polymer [83, 127]. Far from the sur-
face (x → ∞), both ψ and φ reach their
2.7.5.2 Numerical Profiles: Constant ψs bulk values and their derivatives vanish:
When the surface is taken as ideal, that ψ |x→∞ = 0 and φ |x→∞ = 0.
is, flat and homogeneous, the physical The numerical solutions of the mean-
quantities depend only on the distance field Eqs. (55) and (56) together with the
x from the surface. The surface imposes boundary conditions discussed above are
boundary conditions on the polymer order presented in Fig. 11, for several different
parameter φ(x) and electrostatic potential physical parameters.
ψ(x). In thermodynamic equilibrium, all The polymer is positively charged and
charge carriers in solution should exactly is attracted to the nonadsorbing surface
balance the surface charges (charge neu- held at a constant negative potential. The
trality). The Poisson–Boltzmann Equa- aqueous solution contains a small amount
tion (55), the self-consistent field Equa- of monovalent salt (csalt = 0.1 mM). The
tion (56), and the boundary conditions reduced concentration profile cm (x)/φb2
uniquely determine the polymer concen- is plotted as a function of the distance
tration profile and the electrostatic poten- from the surface x. Different curves
tial. In most cases, these two coupled correspond to different values of the
nonlinear equations can only be solved reduced surface potential ys ≡ eψs /kB T ,
numerically. the charge fraction f , and the effective
We present now numerical profiles monomer size a.
obtained for surfaces with a constant Although the spatial variation of the
potential ψs : profiles differs in detail, they all have a
single peak that can be characterized by
ψ|x=0 = ψs (57) its height and width. This observation
200
cm(x)/cbm
100
0
0 10 20 30 40 50
x
[Å]
Fig. 11 Adsorption profiles obtained by numerical solutions
of Eqs. (55) and (56) for several sets of physical parameters in
the low-salt limit. The polymer concentration scaled by its bulk
value cmb = φ 2 is plotted as a function of the distance from the
b
surface. The different curves correspond to: f = 1, a = 5 Å and
ys = eψs /kB T = −0.5 (solid curve); f = 0.1, a = 5 Å and
ys = −0.5 (dots); f = 1, a = 5 Å and ys = −1.0 (short dashes);
f = 1, a = 10 Å and ys = −0.5 (long dashes); and f = 0.1,
a = 5 Å and ys = 1.0 (dot–dash line). For all cases,
φb2 = 10−6 Å−3 , v2 = 50 Å3 , ε = 80, T = 300 K, and
csalt = 0.1 mM. (Adapted from Ref. [124].)
serves as a motivation to using scaling where h(x) is a dimensionless function

arguments. normalized to unity at its maximum and
cM sets the scale of polymer adsorption,
√
2.7.5.3 Scaling Results such that φ(D) = cM . The free energy
The numerical profiles of the previous can now be expressed in terms of D and
section indicate that it may be possible cM , while the exact form of h(x) affects
to obtain simple analytical results for only the numerical prefactors.
the PE adsorption by assuming that In principle, the adsorption length D
the adsorption is characterized by one depends also on the ionic strength through
dominant length scale D. Hence, we write κ −1 . As discussed below, the scaling
the polymer order parameter profile in the assumption (Eq. 59) is only valid as
form long as κ −1 and D are not of the
same order of magnitude. Otherwise,
√
φ(x) = cM h(x/D) (59) h should be a function of both κx
and x/D. We concentrate now on two are much stronger than the excluded
limiting regimes where Eq. (59) can be volume ones.
justified: (1) the low-salt regime D κ −1 Neglecting the latter interactions and
and (2) the high-salt regime D κ −1 . minimizing the free energy with respect
We first discuss the case of constant to D and cM gives
surface potential, which can be directly
compared to the numerical profiles. Then a2 1
D2 ∼ (61)
we note the differences with the constant f |ys | f |ψs |
surface charge boundary condition in and
which the interesting phenomenon of |ys |2
charge overcompensation is discussed in cM ∼ |ψs |2 (62)
4πlB a 2
detail.
recalling that ys = eψs /kB T . As discussed
above, these expressions are valid as
2.7.5.3.1 Low-salt Regime D κ −1 and
long as (1) D κ −1 and (2) the ex-
ψs = constant In the low-salt regime the
cluded volume term in Eq. (60) is negli-
effect of the small ions can be neglected
gible. Condition (1) translates into csalt
and the free energy (per unit surface area)
f |ys |/(8πlB a 2 ). For |ys | 1, a = 5 Å and
(Eqs. 51–54) is approximated by (see also
lB = 7 Å this limits the salt concentra-
Refs. [80, 124])
tion to csalt /f 0.4 M. Condition (2)
on the magnitude of the excluded vol-
a2 ume term can be shown to be equivalent
F cM − f |ys |cM D
6D to f v2 |ys |/ lB a 2 . These requirements
1 are consistent with the data presented in
+ 4πlB f 2 cM
2
D 3 + v2 cM
2
D (60)
2 Fig. 11.
We recall that the profiles presented
In the above equation and in what follows, in Fig. 11 were obtained from the nu-
we neglect additional prefactors of order merical solution of Eqs. (55) and (56),
unity in front of the various terms that arise including the effect of small ions and
from inserting the scaling profile (Eq.59) excluded volume. The scaling relations
into the free energy. The first term of are verified by plotting in Fig. 12 the
Eq. (60) is the elastic energy characterizing same sets of data as in Fig. 11, using
the response of the polymer to concentra- rescaled variables as defined in Eqs. (61)
tion inhomogeneities. The second term and (62). That is, the rescaled electro-
accounts for the electrostatic attraction of static potential ψ(x)/ψs and polymer con-
the polymers to the charged surface. The centration cm (x)/cM ∼ cm (x)a 2 /|ys |2 are
third term represents the Coulomb repul- plotted as functions of the rescaled dis-
sion between adsorbed monomers. The tance x/D ∼ xf 1/2 |ys |1/2 /a. The different
last term represents the excluded volume curves roughly collapse on the same curve.
repulsion between adsorbed monomers, In many experiments the total amount
where we assume that the monomer con- of adsorbed polymer per unit area
centration near the surface is much larger is measured as function of the physical
than the bulk concentration cM φb2 . characteristics of the system such as the
In the low-salt regime and for highly charge fraction f , the pH of the solution,
charged PEs, the electrostatic interactions or the salt concentration csalt [103–110].
0.8
0
0.6
cm(x)/c M
ψ (x) / | ψs |
−0.5 0.4
0.2
−1
0
0 1 2 3 0 1 2 3
x/D x/D
(a) (b)
Fig. 12 Scaling behavior of PE adsorption in the low-salt regime (Eqs. 61 and 62). (a) The
rescaled electrostatic potential ψ(x)/|ψs | as a function of the rescaled distance x/D. (b) The
rescaled polymer concentration cm (x)/cM as a function of the same rescaled distance. The
profiles are taken from Fig. 11 (with the same notation). (Adapted from Ref. [124].)
This quantity can be easily obtained from 2.7.5.3.2 High-salt Regime D κ −1 and
our scaling expressions yielding ψs = constant Let us now consider the
opposite case of a high ionic strength
∞
solution. Here, D is much larger than
= [cm (x) − φb2 ] dx
0 κ −1 , and the electrostatic interactions are
|ys |3/2 |ψs |3/2 short ranged with a cutoff κ −1 . The free
DcM 1/2
∼ (63) energy of the adsorbing PE layer (per unit
lB af f 1/2
surface area) then reads
The adsorbed amount (f ) in the low-
a2
salt regime is plotted in the inset of F cM − f |ys |cM κ −1
Fig. 13(a). As a consequence of Eq. (63), 6D
decreases with increasing charge frac- 1
+ 4πlB f 2 κ −2 cM
2 2
D + v2 cM D (64)
tion f . Similar behavior was also re- 2
ported in experiments [106]. This effect
is at first glance quite puzzling be- The electrostatic cutoff enters in two
cause as the polymer charge increases, places. In the second term, only the
the chains are subject to a stronger at- first layer of width κ −1 interacts elec-
traction to the surface. On the other trostatically with the surface. In the
hand, the monomer–monomer repulsion third term, each charged layer situated
is stronger and, indeed, in this regime, at point x interacts only with layers at
the monomer–monomer Coulomb repul- x for which |x − x | < κ −1 . This term
sion scales as (f cM )2 and dominates can be also viewed as an additional
over the adsorption energy that scales electrostatic excluded volume with vel ∼
as f cM . lB (f/κ)2 .
0.002 0.003
0.01
csalt −> 0 csalt −> 0
0.002 0.001 0.005
0 0.002 0
0 0.5 1 −4 −2 0 2 4
[Å−2]
[Å−2]
csalt = 1 M
Γ
Γ
0.001
0.001 csalt = 1 M
csalt = 0.1 M
csalt = 0.1 M
0 0
0 0.2 0.4 0.6 0.8 1 −4 −2 0 2 4
(a) f (b) pH − pK0
Fig. 13 Typical adsorbed amount as a used for ε, T, and v2 are the same as in Fig. 11,
function of (a) the charge fraction f and (b) the while ys = eψs /kB T = −0.5 and a = 5Å. The bulk
pH − pK0 of the solution for different salt concentration cmb = φ 2 is assumed to be much
b
concentrations (Eq. 67). The insets correspond smaller than cM . (Adapted from Ref. [124].)
to the low-salt regime (Eq. 63). The parameters
Minimization of the free energy gives given by the pH of the solution through
the expression
1/2
κa 2 c
D ∼ salt (65) 10pH−pK0
f |ys | f |ψs | f = (68)
1 + 10pH−pK0
and
f 2 |ys |2 /(κa)2 where pK0 = − log10 K0 and K0 is the ap-
cM ∼ (66) parent dissociation constant. We note that
f 2 /csalt + αv2
this relation is only strictly valid for in-
yielding finitely dilute monomers and that distinct
deviations from it are observed for PEs
−1/2
f |ys |csalt f |ψs | −1/2 because of the electrostatic repulsion be-
∼ ∼ c tween neighboring dissociating sites [126].
salt + αv2
f 2 /c vel + αv2 salt
(67) Still, the results in Fig. 13(b) capture the
where α is a numerical constant of order main qualitative trends of pH-dependent
unity that depends on the profile details. PE adsorption.
The adsorption behavior is depicted in Another interesting observation that can
Figs. 13 and 14. Our scaling results are be deduced from Eq. (67) is that is only a
−1/2
in agreement with numerical solutions function of f csalt . Indeed, as can be seen
of discrete lattice models (the multi- in Fig. 13, csalt only affects the position of
Stern layer theory) [61, 62, 111, 117–120]. the peak and not its height.
In Fig. 13, is plotted as function of The effect of salt concentration is shown
f (Fig. 13a) and the pH (Fig. 13b) for in Fig. 14, where is plotted as function of
different salt concentrations. The behavior the salt concentration csalt for two charge
as seen in Fig. 13(b) represents annealed fractions f = 0.01 and 0.25. The curves
PEs where the nominal charge fraction is on the right-hand side of the graph are
0.03 Low salt High salt
f = 0.01
0.02
[Å−2]
Γ
0.01
f = 0.25
0
10−6 10−5 10−4 10−3 10−2 10−1 1 10
csalt
[M]
Fig. 14 The adsorbed amount as a function of the salt
concentration csalt (Eq. 67) for f = 0.01 and 0.25. The solid curves
on the right-hand side correspond to the scaling relations in the
high-salt regime (Eq. 67). The horizontal lines on the left-hand side
mark the low salt values (Eq. 67). The dashed lines serve as guides
to the eye. The parameters used are ε = 80, T = 300 K, v2 = 50Å3 ,
a = 5Å, ys = −2.0. (Adapted from Ref. [124].)
calculated from the high-salt expression For strongly charged PEs, on the other
for (Eq. 67). The horizontal lines on the hand, the balance between these two elec-
left-hand side of the graph indicate the trostatic terms depends on the amount of
low-salt values of (Eq. 63). The dashed salt. At low salt concentrations, the domi-
lines in the intermediate salt regime serve nant interaction is the monomer–surface
only as guides to the eye since our scaling Coulomb repulsion. Consequently, addi-
approach is not valid when D and κ −1 are tion of salt screens this interaction and
of the same order. increases the adsorbed amount. When the
Emphasis should be drawn to the dis- salt concentration is high enough, this
tinction between weakly and strongly Coulomb repulsion is screened out and the
charged PEs. For weak PEs, the adsorbed effect of salt is to weaken the surface at-
amount is a monotonously decreas- traction. At this point the adsorbed amount
ing function of the salt concentration starts to decrease. As a result, the behav-
csalt in the whole range of salt con- ior over the whole concentration range is
centrations. The reason being that the nonmonotonic with a maximum at some
monomer–monomer Coulomb repulsion, ∗ , as seen in Fig. 14.
optimal value csalt
proportional to f 2 , is weaker than the From this analysis and from Figs. 13,
monomer–surface interaction, which is 14 and Eq. (67), it is now natural to divide
linear in f . the high-salt regime into two subregimes
according to the PE charge. At low diagram. For weak PEs, this cut goes
charge fractions (subregime HS I), f through the left-hand side of the diagram
f ∗ = (csalt v2 )1/2 , the excluded volume starting from the low-salt regime and,
term dominates the denominator of upon addition of salt, into the HS I regime.
Eq. (67) and Such a path describes the monotonous be-
−1/2
havior inferred from Fig. 14 for the weak
∼ f |ψs |csalt (69) PE (f = 0.01). For strong PEs, the cut
goes through the right-hand side of the di-
whereas at high f (subregime HS II), f
agram, starting from the low-salt regime,
f ∗ , the monomer–monomer electrostatic
passing through the HS II regime, and
repulsion dominates and decreases with
ending in the HS I. The passage through
f and increases with csalt :
the HS II regime is responsible for the
∼ csalt |ψs |f −1
1/2 nonmonotonous behavior inferred from
(70)
Fig. 14 for the strong PE (f = 0.25).
The various regimes with their crossover Similarly, Fig. 13(a,b) correspond to hor-
lines are shown schematically in Fig. 15. izontal scans through the top half of the
Keeping the charge fraction f constant diagram. As long as the system is in the
while changing the amount of salt cor- HS I regime, the adsorbed amount in-
responds to a vertical scan through the creases when the polymer charge fraction
1
High salt HS II
(HS I)
0.1
[M]
csalt
0.01
0.001
0.0001 Low salt
10−3 10−2 10−1 1

f
Fig. 15 Schematic diagram of the different adsorption regimes as
function of the charge fraction f and the salt concentration csalt .
Three regimes can be distinguished: (a) the low-salt regime
D κ −1 ; (b) the high-salt regime (HS I) D κ −1 for weak PEs
f f ∗ = (csalt v2 )1/2 ; and (c) the high-salt regime (HS II) D κ −1
for strong PEs f f ∗ .
increases. As the polymer charge fur- 2.7.5.4.1 Low Salt Limit: D κ −1 As-
ther increases, the system enters the suming that there is only one length scale
HS II regime and the adsorbed amount characterizing the potential behavior in the
decreases. Thus, the nonmonotonous be- vicinity of the surface, as demonstrated in
havior of Fig. 13. We finally note that Fig. 12(a), we find that the surface poten-
the single-chain desorption transition for tial ψs and the surface charge σ are related
a flexible chain, which was discussed in by ψs ∼ σ eD. In the low salt limit we find
Sect. 2.7.4, is also valid for adsorption from from Eq. (61)
solutions and will lead to a desorption tran-
sition at very high salt concentrations in D ∼ (f σ lB )−1/3 (71)
Fig. 15 (which is not shown for clarity).
in agreement with Eq. (47).
2.7.5.4 Overcompensation of Surface Let us define two related concepts via the
Charges: Constant σ effective surface charge density defined
We turn now to a different electrostatic as σ = f − σ , which is the sum of
boundary condition of constant surface the adsorbed polymer charge density and
charge density and look at the interesting the charge density of the bare substrate.
phenomenon of charge compensation by For σ = 0, the adsorbed polymer charge
the PE chains in relation to experiments for exactly compensates the substrate charge.
PE adsorption on flat surfaces, as well as on If σ is positive, the PE overcompensates
charged colloidal particles [72, 73, 75–77]. the substrate charge and more polymer
What was observed in experiments is that adsorbs than is needed to exactly cancel
PEs adsorbing on an oppositely charged the substrate charge. If σ is positive and
surface can overcompensate the original reaches the value σ = σ , it means that
surface charge. Because PEs create a thin the PE charge is f = 2σ and leads to
layer close to the surface, they can act as a charge inversion of the substrate charge.
an effective absorbing surface to a second The effective surface charge consisting of
layer of PEs having an opposite charge the substrate charge and the PE layer has
compared to the first layer. Repeating the a charge density that is exactly opposite to
adsorption of alternating positively and the original substrate charge density σ .
negatively charged PEs, it is possible to Do we obtain overcompensation or
create a multilayer structure of PEs at the charge inversion in the low salt limit
surface. Although many experiments and within mean-field theory? Using scaling
potential applications for PE multilayers arguments, this is not clear since we find
exist, the theory of PE overcompensation that σ ∼ f ∼ σ . That is, each of the two
is only starting to be developed [83, 84, terms in σ scales linearly with σ , and
123–125, 127, 128]. the occurrence of overcompensation or
The scaling laws presented for constant charge inversion will depend on numerical
ψs can be used also for the case of constant prefactors determining the relative sign
surface charge. A surface held at a constant of the two opposing terms. However, if
potential ψs will induce a surface charge we look at the numerical solution for the
density σ . The two quantities are related by mean-field electrostatic potential (Fig. 12),
dψ/ dx = 4πσ e/ε at x = 0. We will now we see indeed that all plotted profiles have
consider separately the two limits: low salt a maximum of ψ(x) as function of x. An
D κ −1 and high salt D κ −1 . extremum in ψ means a zero local electric
field. Or equivalently, using Gauss law, this relative strength of the surface charge den-
means that the integrated charge density sity σ and the nonelectrostatic interaction,
from the wall to this special extremum it is also possible to cause a charge over-
point (including surface charges) is exactly compensation and even an exact charge
zero. At this point the charges in solution inversion in a special case.
exactly compensate the surface charges. Finally, we note that the dependence of
the charge parameter σ on the amount of
salt, csalt , is different for constant surface
2.7.5.4.2 High Salt Limit: D κ −1
charge and constant surface potential
When we include salt in the solution
cases. While for the former, σ is
and look at the high salt limit, the
nonmonotonous and has a maximum (as
situation is more complex. The only length
mentioned above) as function of the salt
characterizing the exponential decay of ψ
concentration, in the latter case, σ/σ is
close to the surface is the DH screening
a monotonic decreasing function of csalt
length. Hence, using dψ/ dx|s ∼ σ e yields
(Eq. 73). This can be explained by the extra
ψs ∼ σ eκ −1 , and therefore from Eq. (65),
powers of csalt in the latter case coming
−1/2
from the relation ψs ∼ σ csalt ∼ σ κ −1 .
κ 2a2
D∼ ∼ κ 2 f −1 σ −1 (72) Let us remark that in other theories the
f σ lB overcharging is due to lateral correlations
between adsorbed PEs, which in conjunc-
The estimation of the PE layer charge can
tion with screening by salt ions leads to
be obtained by using the expression for D
strongly overcharged surfaces [84, 128].
and cM in this high salt limit (Eqs. 63–65),
yielding
2.7.5.5 Final Remarks on Adsorption from
κ −2 )
βσ (8πlB csalt βσ Semidilute Solutions
f = The results presented earlier (Sect. 2.7.5)
1 + αv2 csalt f −2 1 + v2 /vel
for adsorption from solutions have been
(73) derived using mean-field theory. Hence,
where vel = f 2 /αcsalt is the electrostatic lateral fluctuations in the polymer and
contribution to the second-virial coefficient ionic concentrations are neglected. In ad-
v2 and α > 0 and β > 1 are positive dition, we neglect the delicate influence of
numerical factors. the charges on the PE persistence length
We see that σ = f − σ is a decreas- and any deviations from ground state dom-
ing function of v2 . Charge overcompensa- inance [9]. The region of validity of the
tion can occur when v2 is smaller than vel theory is for long and weakly charged
(up to a prefactor of order unity). When polymer chains in contact with a mod-
v2 can be neglected in the vicinity of the erately charged surface. The PE solution is
surface, or when v2 = 0 (theta solvents), placed in contact with a single and ideal
there is always charge overcompensation, surface (infinite, flat, and homogeneous).
σ = (β − 1)σ > 0. This is the case of The problem reduces then to an effec-
strongly charged PEs. Similar conclusions tive one-dimensional problem depending
have been mentioned in Refs. [83, 127] only on the distance from the charged
where v2 was taken as zero but the sur- surface. We take very simple boundary
face has a nonelectrostatic short range conditions for the surface assuming that
interaction with the PE. By tuning the the polymer concentration is zero on the
surface and keeping the surface in con- The scaling arguments are then repeated
stant potential or constant surface charge for constant σ boundary conditions. It
conditions. is found that the PE can possibly cause
We find numerical solutions for the charge overcompensation and even in-
PE profile equations in various cases. version of the nominal substrate charge,
These numerical solutions agree well with leading the way to multilayer formation of
simple scaling assumptions describing the positively and negatively charged PEs. The
adsorption of PEs. Scaling expressions for scaling approach can serve as a starting
the amount of adsorbed polymer and point for further investigations. The ana-
the width D of the adsorbed layer, as a lytical and approximated expressions are
function of the fractional charge f and the valid only in specific limits. Special atten-
salt concentration csalt , are obtained for tion should be directed to the crossover
two cases: constant ψs and constant σ . regime where D and κ −1 are of compara-
For constant ψs and in the low-salt ble size.
regime, a f −1/2 dependence of is The problem of charge inversion is not
found. It is supported by our numerical well understood at present. Alternative
solutions of the profile Eqs. (55) and (56) approaches rely on lateral correlations be-
and is in agreement with experiment [106]. tween semiflexible adsorbed PE chains,
This behavior is due to strong Coulomb which also can lead to strong overcompen-
repulsion between adsorbed monomers sation of surface charges [84, 128].
in the absence of salt. As f decreases,
the adsorbed amount increases until the 2.7.6
electrostatic attraction becomes weaker Polyelectrolyte Brushes
than the excluded volume repulsion, at
which point, starts to decrease rapidly. Charged polymers that are densely end-
At high salt concentrations it is not grafted to a surface are called polyelectrolyte
possible to neglect the excluded vol- brushes or charged brushes. They have
ume interaction of the monomers since been the focus of numerous theoreti-
the electrostatic interactions are screened cal [129–138] and experimental [139–142]
by the salt. We obtain two limit- studies. In addition to the basic inter-
ing behaviors: (1) For weakly charged est, charged brushes are considered for
PEs, f f ∗ = (csalt v2 )1/2 , the adsorbed their applications as efficient means for
amount increases with the fractional preventing colloids in polar media (such
charge and decreases with the salt con- as aqueous solutions) from flocculating
centration, ∼ f/ (csalt ), owing to the and precipitating out of solution [93]. This
monomer–surface electrostatic attraction. stabilization arises from steric (entropic)
(2) For strong PEs, f f ∗ , the ad- and electrostatic (energetic) repulsion. A
sorbed amount decreases with the frac- strongly charged brush is able to trap
tional charge and increases
with the salt its own counterions and generates a
concentration, ∼ (csalt )/f , owing to layer of locally enhanced salt concen-
the dominance of monomer–monomer tration [131]. It is thus less sensitive to
electrostatic repulsion. Between these two the salinity of the surrounding aqueous
regimes, we find that the adsorbed amount medium than a stabilization mechanism
reaches a maximum in agreement with based on pure electrostatics (i.e. without
experiments [107, 110]. polymers).
Neutral brushes have been extensively or be confined inside the brush, H ≈ L,

studied theoretically in the past using scal- as shown in Fig. 16(b). As we show now,
ing theories [143, 144], strong-stretching case (b) is indicative of strongly charged
theories [145–149], self-consistent field brushes, while case (a) is typical for weakly
theories [150, 151], and computer simu- charged brushes.
lations [152–155]. Little is known from The free energy per unit area (and in
experiments on the scaling behavior of units of kB T ) contains several contribu-
PE brushes as compared to uncharged tions. We denote the grafting density of
polymer brushes. The thickness of the PEs by ρ, the counterion valency by z,
brush layer has been calculated from recalling that N is the polymerization in-
neutron-scattering experiments on end- dex of grafted chains and f their charge
grafted polymers [139] and charged diblock fraction. The osmotic free energy, Fos ,
copolymers at the air–water interface [141, associated with the ideal entropy cost of
142]. confining the counterions to a layer of
Theoretical work on PE brushes was thickness H is given by
initiated by the works of Miklavic and
Marcelja [129] and Misra and cowork-
Nfρ Nfρ
ers. [130]. In 1991, Pincus [131] and Fos ln (74)
z zH
Borisov, Birshtein and Zhulina [132] pre-
sented scaling theories for charged brushes
Fv2 is the second-virial contribution
in the so-called osmotic regime, where the
to the free energy, arising from steric
brush height results from the balance be-
repulsion between the monomers (contri-
tween the chain elasticity (which tends to
butions due to counterions are neglected).
decrease the brush height) and the repul-
Throughout this review, the polymers are
sive osmotic counterion pressure (which
assumed to be in a good solvent (posi-
tends to increase the brush height). In
tive second-virial coefficient v2 > 0). The
later studies, these works have been gener-
contribution thus reads
alized to the poor solvents [133, 134] and to
the regime where excluded volume effects 2
become important, that is, the so-called 1 Nρ
Fv2 Lv2 (75)
quasi-neutral or Alexander regime [137]. 2 L
In what follows we assume that the
charged brush is characterized by two Finally, a direct electrostatic contribution
length scales: the average vertical exten- Fel occurs if the PE brush is not locally
sion of polymer chains from the wall L electro-neutral throughout the system, as
and the typical extent of the counterion for example is depicted in Fig. 16(a). This
cloud, denoted by H . We neglect the pres- energy is given by
ence of additional salt, which has been
discussed extensively in the original liter- 2πB (Nfρ)2 (L − H )2
Fel = (76)
ature, and only consider screening effects 3 H
due to the counterions of the charged
brush. Two different scenarios emerge, This situation arises in the limit of
as is schematically presented in Fig. 16. low charge, when the counterion density
The counterions can either extend outside profile extends beyond the brush layer, that
the brush, H L, as shown in Fig. 16(a), is, H > L.
Fig. 16 Schematic PE brush structure. H >> L

In (a), we show the weak-charge limit
where the counterion cloud has a + + + −
+ + −
− − −
thickness H larger than the thickness of −
+ + + + −
the brush layer, L. In (b), we show the + −
− − −
opposite case of the strong-charge limit, −
+ + +
where all counterions are contained + + − −
+ −
inside the brush and a single length −
− − − −
scale L ≈ H exists. + + + +
+ −
− − + −
− −
−
+ + +
+ +
+ −
− − −
+ − −
+ + + −
+ +
(a) L
+ + +
+ +
−
− − − − − −
+ + + +
+
− − − −
−
+ − + +
+ +
+
− −
− − −
−
+ + − + +
−
+
− − − +
− −
− −
+ + +
+ +
+
− − −
+ − −
+ + +
+ +
L≈H
(b)
The last contribution is the stretching In this case, minimization of Fos + Fel
energy of the chains, which is with respect to the counterion height H
leads to
3L2 1
Fst = ρ (77) H ∼ (78)
2N a 2 zB Nfρ
which is the Gouy-Chapman length for z-
Here, a is the monomer size or Kuhn
valent counterions at a surface of surface
length of the polymer, implying that
charge density σ = Nfρ. Balancing now
we neglect any chain stiffness for the
the polymer stretching energy Fst and the
brush problem. The different free en-
electrostatic energy Fel , one obtains the
ergy contributions lead, upon minimiza-
so-called Pincus brush
tion with respect to the two length
scales H and L, to different behaviors. L N 3 ρa 2 B f 2 (79)
Let us first consider the weak charg-
ing limit, that is, the situation in which In the limit of H ≈ L, the PE brush
the counterions leave the brush, H > L. can be considered as neutral and the
electrostatic energy vanishes. There are 1

ρ∗ ∼ 1/2 1/2
(83)
two ways of balancing the remaining N B v2
free energy contributions. The first is
obtained by comparing the osmotic energy
For large values of the charge fraction
of counterion confinement, Fos , with the
f and the grafting density ρ, it has
polymer stretching term, Fst , leading to
been found numerically that the brush
the height
height does not follow any of the scaling
N af 1/2 laws discussed here [156]. This has been
L∼ (80) recently rationalized in terms of another
z1/2
scaling regime, the collapsed regime. In
constituting the so-called osmotic-brush
this regime, one finds that correlation and
regime. Finally, comparing the second-
fluctuation effects, which are neglected in
virial free energy, Fv2 , with the polymer
the discussion in this section, lead to a
stretching energy, Fst , one obtains
net attraction between charged monomers
L ∼ N a(v2 ρ/a)1/3 (81) and counterions [157].
and the PE brush is found to have the 2.7.7

same scaling behavior as the neutral Conclusion
brush [143, 144]. Comparing the heights
of all three regimes, we arrive at the In this chapter we have reviewed the
phase diagram shown in Fig. 17. The three behavior of charged polymers (PEs) in so-
scaling regimes meet at the characteristic lution and at interfaces, concentrating on
charge fraction aspects that are different from the corre-
1/3 sponding behavior of neutral polymers.
zv2
f∗ ∼ (82) Because charged biopolymers and iso-
N 2 a 2 B
lated PE chains tend to be quite stiff
and the characteristic grafting density due to electrostatic monomer–monomer
ρ Fig. 17 Scaling diagram for PE brushes

3/2 on a log–log plot as a function of the
grafting density ρ and the fraction of
charged monomers f . Featured are the
Pincus-brush regime, where the
ρ∗ Osmotic brush counterion layer thickness is much
larger than the brush thickness, the
osmotic-brush regime, where all
−3/2 counterions are inside the brush and the
Neutral
brush brush height is determined by an
Pincus brush
equilibrium between the counterion
osmotic pressure and the PE stretching
energy, and the neutral-brush regime,
−3 where charge effects are not important
f and the brush height results from a
f∗ balance of PE stretching energy and
second-virial repulsion. The power-law
exponents of the various lines are
denoted by numbers.
repulsions, their chain statistics is re- will be depleted from the surface. We
lated to that of semiflexible polymers. also briefly review the phenomenon of
Neutral and charged semiflexible polymers charge overcompensation and inversion,
are controlled by their bending rigidity, when the adsorbed PE layer effectively
which is usually expressed in terms of inverses the sign of the surface charge
a persistence length (see Sect. 2.7.2.2). leading the way to formation of PE multi-
For PEs, the electrostatic interaction layers.
considerably influences this persistence Important topics that we have left out
length. are the dynamics of PE solutions, which
In solution, we have considered the scal- is reviewed in Ref. [4], and the behavior of
ing behavior of a single PE (Sect. 2.7.3.1). PEs under bad-solvent conditions [39–41].
The importance of the electrostatic persis- In the future we expect that studies of
tence length was stressed. The Manning PEs in solutions and at surfaces will be
condensation of counterions leads to a directed more toward biological systems.
reduction of the effective linear charge We mentioned in this review the complex-
density (Sect. 2.7.3.1.1). Excluded volume ation of DNA and histones (Sect. 2.7.4.1).
effects are typically less important than This is only one of many examples of
for neutral polymers (Sect. 2.7.3.1.2). Di- interest in which charged biopolymers,
lute PE solutions are typically dominated receptors, proteins, and DNA molecules
by the behavior of the counterions. So interact with each other or with other
is the large osmotic pressure of dilute cellular components. The challenge for
PE solutions due to the entropic contri- future fundamental research will be to
bution of the counterions (Sect. 2.7.3.2). try to understand the role of electrostatic
Semidilute PE solutions can be described interactions combined with specific bi-
by the RPA, which in particular yields ological (lock–key) mechanisms and to
the characteristic peak of the structure
infer on biological functionality of such
factor.
interactions.
At surfaces, we discussed in detail
the adsorption of single PEs (Sect. 2.7.4),
the adsorption from semidilute solutions Acknowledgment
(Sect. 2.7.5), and the behavior of end-
grafted PE chains (Sect. 2.7.6). We tried It is a pleasure to thank our collabora-
to express the PE behavior in terms of tors I. Borukhov, J.F. Joanny, K. Kunze,
a few physical parameters such as the L. Leibler, R. Lipowsky, H. Orland, M.
chain characteristics (persistence length), Schick, and C. Seidel with whom we have
ionic strength of the solutions, and surface been working on PE problems. We ben-
characteristics. The shape and size of the efitted from discussions with G. Ariel, Y.
adsorbing layer is, in many instances, gov- Burak, and M. Ullner. One of us (DA)
erned by a delicate balance of competing would like to acknowledge partial sup-
mechanisms of electrostatic and nonelec- port from the Israel Science Foundation
trostatic origin. In some cases, it is found funded by the Israel Academy of Sciences
that the adsorbing PE layer is flat and and Humanities – Centers of Excellence
compressed, while in other cases, it is Program and the Israel–US Binational
coiled and extended. Yet, in other situa- Science Foundation (BSF) under grant
tions, the PEs will not adsorb at all and no. 98-00429.
References 26. N. V. Brilliantov, D. V. Kuznetsov, R. Klein,

Phys. Rev. Lett. 1998, 81, 1433.
1. F. Oosawa, Polyelectrolytes, Marcel Dekker, 27. M. O. Khan, B. Jönsson, Biopolymers 1999,
New York, 1971. 49, 121.
2. H. Dautzenberg, W. Jaeger, B. P. J. Kötz 28. T. T. Nguyen, I. Rouzina, B. I. Shklovskii,
et al., Polyelectrolytes: Formation, character- Phys. Rev. E 1999, 60, 7032.
ization and application, Hanser Publishers, 29. G. S. Manning, J. Chem. Phys. 1969, 51, 924.
Munich, 1994. 30. G. S. Manning, J. Chem. Phys. 1969, 51, 934.
3. S. Förster, M. Schmidt, Adv. Polym. Sci. 31. G. S. Manning, U. Mohanty, Physica A
1995, 120, 50. 1997, 247, 196.
4. J.-L. Barrat, J.-F. Joanny, Adv. Chem. Phys. 32. M. Deserno, C. Holm, S. May, Macro-
1996, 94, 1. molecules 2000, 33, 199.
5. E. Eisenriegler, Polymers Near Surfaces, 33. C. Wandrey, D. Hunkeler, U. Wendler
World Scientific, Singapore, 1993. et al., Macromolecules 2000, 33, 7136.
6. A. Yu. Grosberg, A. R. Khokhlov, Statistical 34. J. Blaul, M. Wittemann, M. Ballauff et al., J.
Physics of Macromolecules, AIP Press, New Phys. Chem. B 2000, 104, 7077.
York, 1994. 35. H. Schiessel, P. Pincus, Macromolecules
7. P. J. Flory, Principles of Polymer Chemistry, 1998, 31, 7953.
Cornell University, Ithaca, 1953. 36. H. Schiessel, Macromolecules 1999, 32, 5673.
8. H. Yamakawa, Modern Theory of Polymer 37. A. R. Khokhlov, J. Phys. A 1980, 13, 979.
Solutions, Harper and Row, New York, 1971. 38. Y. Kantor, M. Kardar, Phys. Rev. E 1995, 51,
9. P. G. de Gennes, Scaling Concepts in Polymer 1299.
Physics, Cornell University, Ithaca, 1979. 39. A. V. Dobrynin, M. Rubinstein, S. P.
10. A. R. Khokhlov, K. A. Khachaturian, Poly- Obukhov, Macromolecules 1996, 29, 2974.
mer 1982, 23, 1742. 40. A. L. Lyulin, B. Dünweg, O. V. Borisov et al.,
11. J.-L. Barrat, J.-F. Joanny, Europhys. Lett. Macromolecules 1999, 32, 3264.
1993, 24, 333. 41. U. Micka, C. Holm, K. Kremer, Langmuir
12. R. R. Netz, H. Orland, Eur. Phys. J. B 1999, 1999, 15, 4033.
8, 81. 42. T. A. Waigh, R. Ober, C. E. Williams et al.,
13. T. Odijk, J. Polym. Sci., Polym. Phys. Ed.
Macromolecules 2001, 34, 1973.
1977, 15, 477.
43. M. Nierlich, F. Boue, A. Lapp et al., Coll.
14. T. Odijk, Polymer 1978, 19, 989.
Polym. Sci. 1985, 263, 955.
15. J. Skolnick, M. Fixman, Macromolecules
44. M. Nierlich, F. Boue, A. Lapp et al., J. Phys.
1977, 10, 944.
(France) 1985, 46, 649.
16. G. A. Christos, S. L. Carnie, J. Chem. Phys.
45. A. Moussaid, F. Schosseler, J. P. Munch
1990, 92, 7661.
17. C. Seidel, H. Schlacken, I. Müller, Macro- et al., J. Phys. II (France) 1993, 3, 573.
mol. Theory Simul. 1994, 3, 333. 46. W. Essafi, F. Lafuma, C. E. Williams, J.
18. M. Ullner, B. Jönsson, C. Peterson et al., J. Phys. II (France) 1995, 5, 1269.
Chem. Phys. 1997, 107, 1279. 47. K. Nishida, K. Kaji, T. Kanaya, Macro-
19. U. Micka, K. Kremer, Phys. Rev. E 1996, 54, molecules 1995, 28, 2472.
2653. 48. W. Essafi, F. Lafuma, C. E. Williams, Eur.
20. U. Micka, K. Kremer, Europhys. Lett. 1997, Phys. J. B 1999, 9, 261.
38, 279. 49. V. Y. Borue, I. Y. Erukhimovich, Macro-
21. H. Li, T. Witten, Macromolecules 1995, 28, molecules 1988, 21, 3240.
5921. 50. J. F. Joanny, L. Leibler, J. Phys. (France)
22. B.-Y. Ha, D. Thirumalai, J. Chem. Phys. 1990, 51, 545.
1999, 110, 7533. 51. J. des Cloizeaux, Macromolecules 1973, 6,
23. B.-Y. Ha, D. Thirumalai, Macromolecules 403.
1995, 28, 577. 52. T. Yoshizaki, H. Yamakawa, Macromolecules
24. T. B. Liverpool, M. Stapper, Europhys. Lett. 1980, 13, 1518.
1997, 40, 485. 53. R. R. Netz, to be published.
25. R. G. Winkler, M. Gold, P. Reineker, Phys. 54. M. N. Spiteri, F. Boue, A. Lapp et al., Phys.
Rev. Lett. 1998, 80, 3731. Rev. Lett. 1996, 77, 5218.
55. T. A. Witten, P. Pincus, Europhys. Lett. 1987, 82. I. Borukhov, D. Andelman, H. Orland in
3, 315. Short and Long Chains at Interfaces (Eds.:
56. J.-L. Barrat, J.-F. Joanny, J. Phys. II (France) J. Daillant, P. Guenoun, C. Marques et al.),
1994, 4, 1089. Edition Frontieres, Gif-sur-Yvette, 1995,
57. M. J. Stevens, K. Kremer, J. Chem. Phys. pp. 13–20.
1995, 103, 1669. 83. J. F. Joanny, Eur. Phys. J. B 1999, 9, 117.
58. M. J. Stevens, K. Kremer, J. Phys. II 84. R. R. Netz, J. F. Joanny, Macromolecules
(France) 1996, 6, 1607. 1999, 32, 9013.
59. H. Schäfer, C. Seidel, Macromolecules 1997, 85. O. V. Borisov, E. B. Zhulina, T. M. Bir-
30, 6658. shtein, J. Phys. II (France) 1994, 4, 913.
60. M. J. Stevens, S. J. Plimpton, Eur. Phys. J. B 86. T. Odijk, Macromolecules 1983, 16, 1340.
1998, 2, 341.
87. T. Odijk, Macromolecules 1984, 17, 502.
61. M. A. Cohen Stuart, J. Phys. (France) 1988,
88. A. C. Maggs, D. A. Huse, S. Leibler, Euro-
49, 1001.
phys. Lett. 1989, 8, 615.
62. M. A. Cohen Stuart, G. J. Fleer, J. Lyklema,
89. G. Gompper, T. W. Burkhardt, Phys. Rev. A
et al., Adv. Colloid Interface Sci. 1991, 34,
477. 1989, 40, 6124.
63. Y. Fang, J. Yang, J. Phys. Chem. B 1997, 101, 90. G. Gompper, U. Seifert, J. Phys. A 1990, 23,
441. L1161.
64. J. O. Rädler, I. Koltover, T. Salditt et al., 91. R. Bundschuh, M. Lässig, R. Lipowsky, Eur.
Science 1997, 275, 810. Phys. J. E 2000, 3, 295.
65. T. Salditt, I. Koltover, J. O. Rädler et al., 92. V. Yamakov, A. Milchev, O. Borisov et al., J.
Phys. Rev. Lett. 1997, 79, 2582. Phys.: Condens. Matter 1999, 11, 9907.
66. B. Maier, J. O. Rädler, Phys. Rev. Lett. 1999, 93. D. H. Napper, Polymeric Stabilization of
82, 1911. Colloidal Dispersions, Academic Press, New
67. H. von Berlepsch, C. Burger, H. Dautzen- York, 1983.
berg, Phys. Rev. E 1998, 58, 7549. 94. T. Wallin, P. Linse, Langmuir 1996, 12, 305.
68. K. de Meijere, G. Brezesinski, H. Möhwald, 95. T. Wallin, P. Linse, J. Phys. Chem. 1996, 100,
Macromolecules 1997, 30, 2337. 17 873.
69. G. Decher, J. D. Hong, J. Schmitt, Thin 96. T. Wallin, P. Linse, J. Phys. Chem. B 1997,
Solid Films 1992, 210/211, 831. 101, 5506.
70. R. von Klitzing, H. Möhwald, Langmuir 97. E. M. Mateescu, C. Jeppesen, P. Pincus, Eu-
1995, 11, 3554. rophys. Lett. 1999, 46, 493.
71. R. von Klitzing, H. Möhwald, Macro- 98. R. R. Netz, J. F. Joanny, Macromolecules
molecules 1996, 29, 6901. 1999, 32, 9026.
72. G. Decher, Science 1997, 277, 1232. 99. K. K. Kunze, R. R. Netz, Phys. Rev. Lett.
73. M. Lösche, J. Schmitt, G. Decher et al., 2000, 85, 4389.
Macromolecules 1998, 31, 8893. 100. T. D. Yager, C. T. McMurray, K. E. van
74. F. Caruso, K. Niikura, D. N. Furlong et al.,
Holde, Biochemistry 1989, 28, 2271.
Langmuir 1997, 13, 3422.
101. F. von Goeler, M. Muthukumar, J. Chem.
75. E. Donath, G. B. Sukhorukov, F. Caruso
Phys. 1994, 100, 7796.
et al., Angew. Chem., Int. Ed. 1998, 110, 2323.
102. E. Gurovitch, P. Sens, Phys. Rev. Lett. 1999,
76. G. B. Sukhorukov, E. Donath, S. A. Davis
et al., Polym. Adv. Technol. 1998, 9, 759. 82, 339.
77. F. Caruso, R. A. Caruso, H. Möhwald, Sci- 103. P. Peyser, R. Ullman, J. Polym. Sci. A 1965,
ence 1998, 282, 1111. 3, 3165.
78. M. Muthukumar, J. Chem. Phys. 1987, 86, 104. M. Kawaguchi, H. Kawaguchi, A. Taka-
7230. hashi, J. Colloid Interface Sci. 1988, 124, 57.
79. X. Chatellier, J.-F. Joanny, J. Phys. II 105. J. Meadows, P. A. Williams, M. J. Garvey
(France) 1996, 6, 1669. et al., J. Colloid Interface Sci. 1989, 132, 319.
80. R. Varoqui, J. Phys. (France) II 1993, 3, 106. R. Denoyel, G. Durand, F. Lafuma et al., J.
1097. Colloid Interface Sci. 1990, 139, 281.
81. I. Borukhov, D. Andelman, H. Orland, Eu- 107. J. Blaakmeer, M. R. Böhmer, M. A. Cohen
rophys. Lett. 1995, 32, 499. Stuart et al., Macromolecules 1990, 23, 2301.
108. H. G. A. van de Steeg, A. de Keizer, 132. O. V. Borisov, T. M. Birstein, E. B. Zhulina,

M. A. Cohen Stuart et al., Colloid Surf., A J. Phys. II (France) 1991, 1, 521.
1993, 70, 91. 133. R. S. Ross, P. Pincus, Macromolecules 1992,
109. V. Shubin, P. Linse, J. Phys. Chem. 1995, 99, 25, 2177.
1285. 134. E. B. Zhulina, T. M. Birstein, O. V. Borisov,
110. N. G. Hoogeveen, Ph.D. Thesis, Wagenin- J. Phys. II (France) 1992, 2, 63.
gen Agricultural University, The Nether- 135. J. Wittmer, J.-F. Joanny, Macromolecules
lands, unpublished, 1996. 1993, 26, 2691.
111. G. J. Fleer, M. A. Cohen Stuart, J. M. H. M. 136. R. Israels, F. A. M. Leermakers, G. J. Fleer
Scheutjens et al., Polymers at Interfaces, et al., Macromolecules 1994, 27, 3249.
Chapman & Hall, London, 1993, Chap. 11. 137. O. V. Borisov, E. B. Zhulina, T. M. Birstein,
112. C. A. Haynes, W. Norde, Colloid Surf., B Macromolecules 1994, 27, 4795.
1994, 2, 517. 138. E. B. Zhulina, O. V. Borisov, J. Chem. Phys.
113. P. Auroy, L. Auvray, L. Léger, Macro- 1997, 107, 5952.
molecules 1991, 24, 2523. 139. Y. Mir, P. Auvroy, L. Auvray, Phys. Rev. Lett.
114. O. Guiselin, L. T. Lee, B. Farnoux et al., J. 1995, 75, 2863.
Chem. Phys. 1991, 95, 4632. 140. P. Guenoun, A. Schlachli, D. Sentenac, J. M.
115. T. Odijk, Macromolecules 1979, 12, 688. Mays et al., Phys. Rev. Lett. 1995, 74, 3628.
116. A. V. Dobrynin, R. H. Colby, M. Rubinstein, 141. H. Ahrens, S. Förster, C. A. Helm, Macro-
Macromolecules 1995, 28, 1859. molecules 1997, 30, 8447.
117. H. A. van der Schee, J. Lyklema, J. Phys. 142. H. Ahrens, S. Förster, C. A. Helm, Phys.
Chem. 1984, 88, 6661. Rev. Lett. 1998, 81, 4172.
118. J. Papenhuijzen, H. A. van der Schee, 143. S. Alexander, J. Phys. (France) 1977, 38, 983.
G. J. Fleer, J. Colloid Interface Sci. 1985, 104, 144. P. G. de Gennes, Macromolecules 1980, 13,
540. 1069.
119. O. A. Evers, G. J. Fleer, J. M. H. M. Scheut- 145. A. N. Semenov, Sov. Phys. JETP 1985, 61,
jens et al., J. Colloid Interface Sci. 1985, 111, 733.
446. 146. S. T. Milner, T. A. Witten, M. E. Cates, Eu-
120. H. G. M. van de Steeg, M. A. Cohen Stuart, rophys. Lett. 1988, 5, 413.
A. de Keizer et al., Langmuir 1992, 8, 8. 147. S. T. Milner, T. A. Witten, M. E. Cates,
121. P. Linse, Macromolecules 1996, 29, 326. Macromolecules 1988, 21, 2610.
122. R. Varoqui, A. Johner, A. Elaissari, J. Chem. 148. S. T. Milner, Science 1991, 251, 905.
Phys. 1991, 94, 6873. 149. A. M. Skvortsov, I. V. Pavlushkov, A. A.
123. I. Borukhov, D. Andelman, H. Orland, Eur. Gorbunov et al., Polym. Sci. 1988, 30, 1706.
Phys. J. B 1998, 5, 869. 150. R. R. Netz, M. Schick, Europhys. Lett. 1997,
124. I. Borukhov, D. Andelman, H. Orland, 38, 37.
Macromolecules 1998, 31, 1665. 151. R. R. Netz, M. Schick, Macromolecules 1998,
125. I. Borukhov, D. Andelman, H. Orland, J. 31, 5105.
Phys. Chem. B 1999, 24, 5057. 152. M. Murat, G. S. Grest, Macromolecules 1989,
126. C. Tanford, J. C. Kirkwood, Am. J. Chem. 22, 4054.
Soc. 1957, 79, 5333. 153. A. Chakrabarti, R. Toral, Macromolecules
127. D. Andelman, J. F. Joanny, C. R. Acad. Sci. 1990, 23, 2016.
(Paris) 2000, 1, 1153. 154. P. Y. Lai, K. Binder, J. Chem. Phys. 1991, 95,
128. T. T. Nguyen, A. Y. Grosberg, B. I. Shklov- 9288.
skii J. Chem. Phys. 2000, 113, 1110. 155. C. Seidel, R. R. Netz, Macromolecules 2000,
129. S. J. Miklavic, S. Marcelja, J. Phys. Chem. 33, 634.
1988, 92, 6718. 156. F. S. Csajka, C. Seidel, Macromolecules 2000,
130. S. Misra, S. Varanasi, P. P. Varanasi, 33, 2728.
Macromolecules 1989, 22, 5173. 157. F. S. Csajka, R. R. Netz, C. Seidel et al., Eur.
131. P. Pincus, Macromolecules 1991, 24, 2912. Phys. J. E 2001, 4, 505.
327
3.1 consisting of adsorbate molecules A and

Introduction solvent molecules S, for which Eq. (1) takes
the form
Boris B. Damaskin and Oleg A. Petrii
Moscow State University, Moscow, Russia dγ = −A dµA − S dµS (2)
The adsorption of the ith solution com- However, the values µA and µS are
ponent is defined as a phenomenon that not independent, they are interrelated by
manifests itself in the difference of this Gibbs-Duhem equation [2, 3]:
component concentration in the vicinity
of the interface (ci ) and in the solu- NA dµA + NS dµS = 0 (3)
(b)
tion bulk (ci ). If the concentration of where NA and NS = 1 − NA are the mole
the ith component becomes higher than fractions of A and S. It follows from (2)
(b)
ci when the distance x decreases (as we and (3) that
approach the interface), its adsorption is
positive; in the opposite case, it is negative S N A
dγ = − A − dµA (4)
(curves 1 and 2, respectively, in Fig. 1). NS
The quantity characterizing the re-
versible adsorption of the ith component is The combination in parentheses defines
its surface excess i defined by the Gibbs the so-called relative surface excess:
adsorption isotherm [1] at constant tem- (S) S N A
A = A − (5)
perature (T = const) and pressure (p = NS
const):
which can be determined from experimen-
dγ = − i dµi (1) tal data. Particularly, for the system under
i
consideration

The quantity γ is the work consumed in (S) ∂γ
increasing the interface by unit surface A = − (6)
∂µA T ,p
area at T = const and p = const under re-
versible conditions, and µi is the chemical In order to determine separately the val-
potential of the ith component. ues of A and S , the second equation
In order to clarify the physical meaning interrelating these values should be de-
of i , let us consider a binary system rived. For this purpose, i is defined as the
328 3 Specific Adsorption
Fig. 1 Schematic dependence

ci
of concentration of the 1 and 2
components on the distance
from the electrode surface.
1
Concentration
[mol]
S1
(b)
ci
S2
2
0
Coordinate x
difference between the amounts of compo- where MA and MS are the molar masses
nent i in two equal solution cylinders with of components. A simultaneous solution
unit cross section, the first cylinder being of Eqs. (5) and (9) results in the following
normal and adjacent to the interface, and expressions:
the second one being completely in the MS NS
(M) (S)
solution bulk. This definition remains am- A = A
MA NA + MS NS
biguous if we do not clarify the meaning
of ‘‘equal.’’ If both cylinders contain equal (M) (S) MA NS
S = −A (10)
numbers of solvent moles, then S = 0. MA NA + MS NS
As follows from Eq. (5), the surface excess Finally, if both cylinders are of equal
of A defined by imposing this condition volumes, and the molar volumes of the
(S)
represents A (relative surface excess). If components VA and VS do not change in
both cylinders contain equal numbers of the course of adsorption, Eq. (5) should be
moles of A and S, the equation supplemented by the equation
(V ) (V )
A
(N )
+ S
(N )
=0 (7) V A A + VS S =0 (11)
The corresponding simultaneous solution

combined with Eq. (5) allows us to calcu-
of Eqs. (5) and (11) gives the surface
late another type of surface excess values:
excesses
A
(N ) (S)
= A (1 − NA ) (V ) (S) VS N S
A = A
VA N A + VS N S
(N ) (S)
S = −A (1 − NA ) (8) VA N S
(V ) (S)
S = −A (12)
VA N A + VS N S
Next, for two cylinders of equal masses,
the Eq. (5) should be supplemented by As follows from Eqs. (8), (10), and
equation (12), for adsorption from diluted so-
(S) (N ) (M)
lutions (NA 1), A ≈ A ≈ A ≈
(M) (M) (V )
M A A + M S S =0 (9) A . Under these conditions, the physical
3.1 Introduction 329
meaning of the surface excess i can be specific properties, the difference appears
represented by the areas S1 and S2 en- to be not so dramatic as that which we
closed by the curve corresponding to ci found when going to typical d-metals
versus x dependence in the vicinity of (platinum group metals are the most
(b)
interface and the line ci = const (see thoroughly studied representatives).
shaded areas in Fig. 1). For positive ad- In aqueous solutions, platinum-group
sorption, i > 0 whereas, for the negative metals adsorb hydrogen and oxygen and
adsorption, i < 0. In agreement with this can be classified as the so-called ‘‘per-
interpretation of surface excess, its dimen- fectly polarizable electrodes’’ in terms of
sionality is mol m−2 or mol cm−2 , which Planck [4]. For perfectly polarizable elec-
coincides with the dimensionality of the trode, one portion of electric charge de-
surface concentration. The latter can be livered to its surface is expended in the
defined as the number of moles of the change of electrode free charge, when an-
ith component that find themselves in a other portion is expended in the oxidation
monomolecular layer of substance closest of hydrogen adatoms or in the generation
to the surface. In contrast to the surface of oxygen adatoms. The complete quantity
excess, the surface concentration is ei- of electric charge delivered to the surface
ther positive or zero. For calculating the is said to be expended in the change of
surface concentration from experimental total electrode charge. It means that the
i values, it is essential to know how ci state of electrode is uniquely determined
changes with x. The problem may be sub- by its initial state and the change of its total
stantially simplified, if the surface excess charge.
of the ith component is located inside In the text that follows, we start with
an adsorption monolayer, and its bulk the adsorption phenomena on ideally
concentration is sufficiently low (Ni 1). polarizable electrodes, and then pass to
Under these conditions, typical for ad- perfectly polarizable electrodes.
sorption of many organic substances on Adsorption of various ions and mole-
electrodes, the surface concentration prac- cules on electrodes is favored by their
tically coincides with the surface excess. interaction with the electrode surface, and
Originally, the phenomenon of specific also by the squeezing of adsorbate species
adsorption was considered in connection out of the solution bulk. The latter phe-
with studies on mercury electrodes, the nomenon is induced by the difference in
latter being ideally polarizable in a wide free solvation energies G between the
potential range. It should be mentioned surface layer and in the bulk of the so-
that for ideally polarizable electrode, the lution. For positive adsorption, G < 0,
total quantity of electric charge delivered which, in its turn, can be caused by a de-
to its surface is expended in the change of crease in enthalpy (H < 0) and by an
the free electrode charge, that is, the charge increase of entropy (S > 0) as the adsor-
defined by excess or lack of electrons. Quite bate species pass from the solution bulk
a number of s,p-metals (Bi, Ga, Cd, Pb, to the surface. In principle, electrostatic
Sn, Sb, In, Tl) can also exhibit similar forces can solely induce the interaction
behavior in a certain potential interval, and of ions with the electrode surface: the
are sometimes integrated into the group of coulombic attraction and repulsion should
mercury-like metals. Although any metal result in positive and negative adsorption,
of this group undoubtedly exhibits certain respectively. In addition, at a direct contact
of an ion with the metal surface (in the Gouy was the first to mention the
absence of separating solvent molecules), specific behavior of anions on mercury [5];
the ion–metal interaction can include however, the term ‘‘specific adsorption’’
other contributions: image forces (inter- was introduced by Stern [6] who made an
action with the charge of the opposite attempt to treat specific adsorption quan-
sign induced in the metal phase) and a titatively. Generally, any ion can be specif-
covalent bonding, which is sometimes con- ically adsorbed under definite conditions,
sidered in the terms of donor–acceptor and only under certain circumstances can
interaction (which can result in a partial specific adsorption be ignored in the first
charge transfer between adsorbed ions and approximation.
metal). It is conventionally supposed that the
If the ion–metal interaction is com- cations like Li+ , Na+ , and K+ , and anions
pletely electrostatic and the squeezing out like F− , having small sizes and surrounded
is absent, the electrolyte is surface-inactive. by dense solvation shells (also complex
In the opposite case, specific adsorp- anions like HF2 − , PF6 − , BF4 − ), are not
tion takes place, and the corresponding adsorbed specifically. Also, it is assumed
ion is surface-active. Figure 2 shows the that specific adsorption of anions is more
electrocapillary curves (dependences of γ pronounced than that of cations. These
on electrode potential E) for a mercury suppositions stemmed from the studies on
electrode in 0.01, 0.1, and 1 M aqueous a mercury electrode in aqueous solutions
solutions of a surface-inactive electrolyte and cannot be generalized a priori to other
(curves 1, 2, and 3) and the corresponding systems.
γ versus E-curves in solutions containing Historically, the term ‘‘specific adsorp-
surface-active anions (1 , 2 and 3 ). tion’’ goes back to the studies of adsorption
γ 0max
400
1
Surface tension
2
375 3
[mN m−1]
350
1
2 1′
325 3 3′
2′
−1.0 −0.5 0.0 0.5

Potential
[V]
Fig. 2 Electrocapillary curves of mercury in aqueous KF (1–3)
and KI (1 –3 ) solutions, salt concentrations are 0.01 M (1,1 );
0.1 M (2,2 ); 1 M (3,3 ).
of inorganic and organic ions, but it is interface, which is equal to the pzc
worthwhile to apply it also to adsorp- shift and increases with − . The spe-
tion of neutral organic molecules. The cific adsorption of cations shifts the pzc
reason lies not only in the squeezing towards positive values. For simultane-
out from solution, which occurs in both ous specific adsorption of anions and
cases, but also in the existence of spe- cations, the magnitude and the sign of
cific interactions with the metal. That is, the pzc shift can change with relative ionic
the molecules of aromatic compounds are concentrations.
bound to the electrode via the so-called The examples of pzc shifts induced by
π-electron interaction [7–10]; functionally specific adsorption of anions on mercury
substituted molecules (thiolates, amines, in aqueous solutions are presented in
and so forth) assume certain interfacial Table 1. As demonstrated by Frumkin [13],
orientations owing to specific interaction the specific features of metal–ion interac-
of substituting groups with the metal, par- tion can be elucidated by comparing the
ticularly mercury [11]. pzc shifts and the changes of surface po-
According to the Lippman equation [12] tential at the solution/air interface. The
adsorption of ion at the latter interface re-
∂γ
σM = − (13) sults only from squeezing-out effect. For
∂E T ,p,µ example, the change of surface potential
when going from 0.01 M KCl to 2 M KI
the charge density on the metal surface
solution reaches only 52 mV [14], while
σM is determined by the slope of the
the pzc shift for Hg/1 M KI (in compari-
electrocapillary curve. The maximum of
son with pzc in surface-inactive electrolyte
γ , E-curve corresponds to σM = 0; hence,
solution) is about 350 mV.
the position of this maximum determines
The pzc shift resulting from the ad-
the potential of zero charge (pzc). In
sorption of polar organic molecules is
surface-inactive electrolyte solutions, the
determined by the orientation of their
value of the pzc does not depend on
dipoles on the electrode surface.
concentration (Fig. 2). On the other hand,
with an increase in the concentration
of surface-active anion, the pzc shifts Tab. 1 The pzc and the pzc shift values for a
towards negative potentials, because of the mercury electrode in aqueous unimolar
solutions as given by Frumkin [13]
increasing specific adsorption of anions.
Indeed, the condition of electrical neu-
Electrolyte pzc (MCE) [V] pzc shifts [V]
trality in the 1,1-electrolyte solution is
σM = F − − F + (14) KOH −0.47 0.00

K2 CO3 −0.48 −0.01
K2 SO4 −0.48 −0.01
where F stands for Faraday constant, and
K2 HPO4 −0.52 −0.05
the subscripts correspond to the charges NaNO2 −0.54 −0.07
of the ions. Hence, at the pzc, − = + . KCl −0.56 −0.09
However, the specifically adsorbed anions KNO3 −0.56 −0.09
are located closer to the electrode sur- KBr −0.65 −0.18
NaCNS −0.72 −0.25
face as compared with surface-inactive
KI −0.82 −0.35
cations. Thus, a potential drop arises K2 S −0.98 −0.51
near the uncharged electrode/solution
The specific adsorption reduces the obtain the following expression from (17):
work of surface formation because of the
∂E0 ∂(+ + − )
difference in the energetic states of solvent = −RT
molecules on the surface and in the bulk ∂ ln a± σM ∂σM a±
of the solution, and also because of the (18)
lateral interactions in the adsorbed layer. At the pzc, according to Eq. (14), − = + ,
For neutral organic molecules, as a rule, and from Eq. (18) one obtains
the decrease in γ is observed only in the
vicinity of the pzc, because at high |σM | the ∂E0 ∂−
= −2RT
organic molecules are squeezed out by the ∂ ln a± σM =0 ∂σM a±
solvent of a higher permittivity. (19)
The electrocapillary theory developed by Equation (19) relates the dependence
Frumkin [15, 16] makes it possible to de- of the pzc on the concentration of
termine not only the electrode charge and the electrolyte solution in the presence
pzc but also the relative surface excesses of specific adsorption (− = 0 when
of various ions by treating the γ , E-curves. σM = 0) and its variation with the electrode
Namely, in a binary 1,1-electrolyte solu- charge. The dependence of the pzc of
tion, when we use a reference electrode a mercury electrode on the logarithm
reversible with respect to anion (see above of KI concentration was used for the
in Electrode Potentials), first time for studying the iodide specific
adsorption in [17] and later was named
dγ = −σM dE− − 2RT + d ln a± (15) the Esin-Markov effect. As follows from
the model theories of the electric double
whereas, in the same solution with the
layer (see Sect. 3.2), the limiting slope of
reference electrode reversible in respect to
the aforementioned dependence should
cation,
tend to the value −RT /λF , where the
dγ = −σM dE+ − 2RT − d ln a± (16) coefficient λ(0 < λ ≤ 1) characterizes the
discrete nature of the charge of specifically
In the formulae given above, R is the adsorbed anions.
universal gas constant, and a± is the The combination of Eqs. (13), (14), and
mean activity of the electrolyte solution. (17) make it possible to calculate the de-
By combining Eqs. (15) and (16) and pendencies of − and + on E0 from the
introducing the notation dE0 = d(E+ + experimental γ , E-curves. Figure 3 shows
E− )/2, one obtains the corresponding dependencies of F −
and F + on E0 for 0.1 M solutions of
dγ = −σM dE0 − RT (+ + − ) d ln a± surface-inactive and surface-active elec-
(17) trolytes. The positive adsorption of cations
On the combined potential, scale intro- at E0 < 0 and of anions at E0 > 0 in
duced above the pzc value does not depend a former solution are induced only by
on the concentration of the surface- the coulombic attraction of ions to the
inactive electrolyte. Hence, it is convenient oppositely charged electrode surface. In
to refer the potential to this zero point (see contrast, the negative adsorption of cations
Fig. 2). at E0 > 0 and of anions at E0 < 0 is caused
The properties of the total differential by the repulsion from the surface of the
of a function of two variables, allow us to same sign. In 0.1 M KF solution at E0 = 0,
− = + ≈ 0. However, in concentrated only at sufficiently high negative E0 . In ac-

(>1 M) solutions of surface-inactive elec- cordance with Eq. (14), the positive values
trolytes at E0 = 0, the value of γ tends to of + at σM > 0 correspond to F − > σM ,
increase with concentration. In accordance which is known as a phenomenon of the
with Eq. (17), this corresponds to the nega- recharging of the surface by specifically
tive adsorption: − = + < 0. The latter is adsorbed ions. The positive values of +
caused by hydration of ions, when a layer of often pass through a minimum (as shown
solvent molecules separates ions from the in Fig. 3), and the growth of + with the
electrode surface. By designating the thick- positive potential at σM > 0 points to the
ness of this layer as δ and assuming that the enhancement of the surface recharging.
concentration of ions near the uncharged Under these conditions, the discreteness
surface sharply increases from zero to c(b) coefficient of the charge of specifically
at x = δ, we obtain − = + = −δc(b) . Ex- adsorbed anions is λ < 1. Specific adsorp-
perimental results confirm that, at E0 = 0, tion of ions induces more complicated
the negative adsorption of a salt varies potential-distance dependence than com-
proportionally as the salt concentration in pared with the corresponding decay in the
the bulk. For Hg/water interface, the slope absence of specific adsorption. For exam-
of this line corresponds approximately to ple, when the surface is recharged by the
0.3–0.4 nm, namely, a monolayer of water specifically adsorbed anions, the potential
molecules. versus distance dependence should pass
As seen from Fig. 3, the specific ad- through a minimum. When both anions
sorption of anions substantially affects not and cations are specifically adsorbed, the
only − versus E0 but also + versus recharging can take place at any electrode
E0 dependencies. The coulombic repul- charge. However, at σM > 0, the anions are
sion of anions from the negatively charged located closer to the surface than cations;
electrode surface can fully eliminate the and, when at σM < 0, the opposite situa-
manifestations of their specific adsorption tion takes place.
30
1′
20
2′
[µC cm–2]
F Γi
10
1
0
Fig. 3 Components of electrode charge
2
associated with the surface excesses of
anions (1,1 ) and cations (2,2 ), as −0.6 –0.3 0.0 0.3
function of the potential of a mercury
electrode in 0.1 M aqueous solutions of Potential
KF (1,2); KI (1 2 ). [V]
Certain model assumptions are neces- the diffuse layer, then the value of +
sary in order to reveal the surface con- determined from the experimental data
(2)
centration of specifically adsorbed ions in should be equal to + . Hence, we can
the total surface excess i . Usually, the calculate the value of ϕ2 from Eq. (21)
ionic component of the electrical double and, then, find the contribution of anions
(2)
layer (EDL) is assumed to consist of the residing in the diffuse part, − from
dense part and the diffuse layer separated Eq. (22). Finally, we determine the specific
by the so-called outer Helmholtz plane. (1)
adsorption of anions, − from the
Only specifically adsorbing ions can pen- (1)
experimental value of − = − + − .
(2)
etrate into the dense layer close to the This technique of calculating the surface
surface (e.g. iodide ions), with their elec- concentration of anions was pioneered
tric centers located on the inner Helmholtz by Grahame and Soderberg [21] using
plane. The charge density of these specif- the experimental dependencies of the
ically adsorbed ions σ1 is determined by differential capacitance C on the potential,
(1)
their surface concentration i . Namely, where C is given by
for single-charged anions: 2
∂σM ∂ γ
(1) C= =− (24)
σ1 = −F − (20) ∂E a± ∂E 2 a±
On the other hand, the surface excesses of In contrast to γ , the value of C can be mea-
(2)
all ions in the diffuse layer i are defined sured not only for liquid electrodes but also
by their coulombic interaction with the for solid ones. Usually, the specific adsorp-
net charge (σM + σ1 ). According to the tion of ions increases the capacitance of the
classical theory of the diffuse layer [18–20], EDL, whereas the adsorption of organic
the behavior of 1,1-electrolyte solution with molecules decreases it. In the framework
concentration c is described by: of a simplified model of the double layer as
a capacitor, this corresponds to varying the
(2) √ F ϕ2
F + = A c exp − − 1 (21) distance between the capacitor plates and
2RT decreasing the permittivity. In dilute solu-

(2) √ F ϕ2 tions of a surface-inactive 1,1-electrolyte,
F − = A c exp −1 (22)
2RT a capacitance minimum appears in the
vicinity of pzc (see curve 1 in Fig. 4). The
where A = (2RT ε0 ε)1/2 , ε is the permittiv- position of this minimum is determined
ity of solvent, ε0 = 8.854 × 10−12 F m−1 , by the potential E0 = 0. In solutions with
and ϕ2 is the potential of the outer relatively high (≥0.1 M) concentrations
Helmholtz plane referred to the solution of surface-inactive electrolytes, this min-
bulk and related to the value of (σM + σ1 ) imum of C, E0 -curve disappears, but the
by the formula pzc position still corresponds to E0 = 0
(curve 2 in Fig. 4). Numerical integration
2RT σM + σ1
ϕ2 = sinh−1 √ (23) of these C, E0 -curves
F 2A c E0
If we suppose that cations in solutions σM = C dE0 (25)
0
of lithium, sodium, and potassium salts
are surface-inactive, that is, at any given results in the corresponding σM , E0 -curves
E0 , their surface excesses are located in (see curves 1 and 2 in Fig. 5).
Fig. 4 Dependencies of the 60

differential capacitance of a
mercury electrode on its 3
potential in aqueous solutions
of: 0.003 M KF (1); 0.1 M KF (2);
Differential capacitance
0.1 M KI (3). 40
[µF cm–2]
2
20
1
0
−1.0 –0.5 0.0 0.5
Potential
[V]
Fig. 5 Dependencies of the 20

charge of a mercury electrode
on its potential in aqueous 3
2
solutions of: 0.003 M KF (1);
0.1 M KF (2); 0.1 M KI (3). 10
1
[µC cm–2]
0
Charge
−10
−20
−1.0 −0.5 0.0 0.5
Potential
[V]
For nonsymmetrical surface-inactive even at extreme dilution (limited by the po-

electrolytes, the minimum of C shifts tentialities of capacitance measurements).
from the pzc. If, in the absolute mag- In this case, the pzc should be calcu-
nitude the anion charge is higher than lated by means of back integration of
the cation charge, the minimum shifts C, E0 -curves. The idea of this method is as
from pzc to more negative values, and vice follows.
versa. Specific adsorption of ions also in- At a certain sufficiently negative poten-
duces a shift of the capacitance minimum tial E ∗ , the specific adsorption of anions
from the pzc. Moreover, when the spe- disappears. Under these circumstances,
cific adsorption is sufficiently pronounced, the C,E0 -, σM ,E0 -, and γ ,E0 -curves mea-
no minimum appears in the C, E0 -curves, sured, on the one hand, in solutions of
surface-inactive electrolytes and, on the ion of the same sign − by 2, and

other hand, in those of surface-active elec- their common counterion of a constant
trolytes of equal concentrations with the concentration − by 3. Bearing in mind
same cation merged together (see curves 2 that, to a first approximation, the activity
and 3 in Figs. 4 and 5, and also see curves coefficient in solutions of a constant ionic
1 and 1 , 2 and 2 , 3 and 3 in Fig. 2). strength does not change with the ratio
This coincidence provides a way of calcu- c1 /c2 , we can write
lating the σM ,E0 - and γ ,E0 -curves from
the curves of differential capacitance for d ln a3 ≈ d ln c3 = 0;
solutions containing specifically adsorbed dx
anions d ln a1 ≈ d ln(xc3 ) = d ln x = (29)
x
E0
∗ d ln a2 ≈ d ln[(1 − x)c3 ]
σM = σM + C dE0 ;
E∗ dx
= d ln(1 − x) = − (30)
E0
1−x
γ = γ0∗ − σM dE0 (26)
E∗
Hence, the electrocapillarity equation for
In this equation, from the data for a the systems like (28) takes the form
solution of the surface-inactive electrolyte,
E0 (1) (2) dx
dγ = −σM dE − RT 1 d ln x + 1
∗ max x
γ0 = γ0 − σM dE0 (27)
0
(2) dx
−2 (31)
where γ0max is the arbitrary constant. The 1−x
choice of its value does not affect the values
of + and − , because, according to the The surface excesses of ions 1 and 2
Eq. (17), these values are determined not in the diffuse layer are caused only by
by the absolute γ values but by the changes the coulombic interaction of these ions
of the surface tension with a± . with the net charge (σM + σ1 ), hence,
(2) (2)
Another way of estimating the surface 1 / 2 = x/(1 − x). Under these con-
concentration of the specifically adsorbed dition, the sum of two last terms inside the
ions was proposed in Refs. [22, 23], and brackets of Eq. (31) is equal to zero, so that
is based upon using binary mixtures of
(1)
electrolytes of a constant total concentra- dγ = −σM dE − RT 1 d ln x (32)
tion, like
Hence, for the systems of the type (28),
xc CA∗ + (1 − x)c CA or

xc C∗ A + (1 − x)c CA (28) (1) 1 ∂γ
1 = − (33)
RT ∂ ln x E
in which x is the mole fraction of the
surface-active ions (A− +
∗ or C∗ ) in the The γ versus ln x dependence at E =
mixture. It is assumed that ions C+ and const, which is required for using for-
A− are not specifically adsorbed, and their mula (33), can be obtained either from
surface excesses are located in the diffuse experimental electrocapillary curves, or on
layer. Let us designate the surface-active the basis of a series of C,E- or σM ,E-curves
ion by a subscript 1, the surface-inactive for systems like (28) with different x.
In principle, σM ,E-curves can be ob- the free adsorption energy G0A :

tained by various electrochemical tech-
niques. For example, according to Ref. [24], −G0A = const + RT ln β (35)
we can use the chronocoulometry tech-
nique, in which current transients are where the value of const depends on
measured following a potential step E, the choice of the standard state. The
and the variation of charge σM are ob- attraction constant reflects the set of lateral
interactions between the molecules of
tained by integrating these transients. This
adsorbate A and solvent S in the surface
technique has made good use of pyridine
layer:
adsorption studies on the single crystalline
silver and gold [25–30]. −GAA + GAS − GSS
It should be noted that deviations from a= (36)
RT
Eqs. (32) and (33) can be induced by the
noncoincidence of the outer Helmholtz The G0A , GAA , GAS and GSS values,
planes for ions 1 and 2 of a system like and, correspondingly, ln β and a values
28 [31]. depend on the electric state of the surface,
When the species i are adsorbed re- that is, on the electrode potential or
versibly, their electrochemical potentials charge. However, these dependences can
(ads) be obtained in an explicit form only on the
in the adsorption layer µ̄i and in the so-
(b) (ads) basis of some model of the surface layer
lution bulk µ̄i should be equal: µ̄i = consisting of the mixture of organic and
(b)
µ̄i . This simple relationship does not al- water molecules. Unfortunately, so far, the
low one to link unambiguously the surface models of various authors [33–41] have not
and bulk concentrations of ith species, be- resulted in any unified approach.
cause the activity coefficients in the bulk Equation (34) was deduced by Frum-
of the solution and especially in the sur- kin [32] as a general case of Langmuir
face layer are concentration-dependent, isotherm, which corresponds to a = 0. A
and, in general, these dependencies are statistical derivation of Eq. (34) was carried
ambiguously determined. Therefore, the out by Fowler and Guggenheim [42]; how-
dependence of the surface concentration ever, the applicability of Frumkin isotherm
of adsorbate or the surface coverage θ on to the experimental data appears to be sub-
(b) stantially wider than it follows immediately
the bulk concentration ci is usually de-
scribed by an equation providing the best from the derivation conditions.
approximation of experimental data with Figure 6 shows the dependencies of
(b)
a number of fitting parameters being as θ versus ln(βci ) calculated by Eq. (34)
low as possible. The adsorption of neu- at various values of a. When a = 2,
(b)
tral organic molecules is most frequently at θ = 0.5, ∂θ/∂ ln(βci ) = ∞. When
described by the Frumkin isotherm [32]: a > 2, a steplike transition from low θ
values to θ ≈ 1 appears on the isotherm,
(b) θ which corresponds to 2D-condensation of
βci = exp(−2aθ) (34)
1−θ adsorbate molecules (see Sect. 3.3.2.1).
When the ions of the ith species are
where β is the constant of adsorption equi- adsorbed in place of neutral molecules,
librium, and a is the so-called attraction their concentration at the outer Helmholtz
(2)
constant. The value of ln β characterizes plane ci should be substituted for the
Fig. 6 Dependencies of the

1.0 surface coverage on In(βci(b) )
5 4 3 2 1 calculated from Eq. (34) for
various values of a: −2(1); 0(2);
1(3); 2(4); 4(5).
Surface coverage
0.5
0
–4 0 4 8
ln(βci(b))
(b)
bulk concentration ci [43]. According to charge σM , and the potential drop in the in-
Boltzmann formula, ner part of the EDL. The model applied for
describing the EDL determines the choice
(2) (b) −zi F ϕ2 of the independent variable. Namely, in
ci = ci exp (37)
RT the framework of Frumkin model [33] for
adsorption of neutral organic molecules,
According to Eq. (23), the value of ϕ2 , is ln β is a square function of E0 ; in terms
a function of the charge of specifically of Vorotyntsev’s model [44] for ionic ad-
adsorbed ions σ1 = zi F max θ, where max sorption, ln β is also a function of E0 ,
is the limiting value of their surface but substantially more complicated [45]; in
concentration corresponding to θ = 1. Grahame-Parsons model [46, 47] for ionic
Thus, the adsorption isotherm equation adsorption, the value of ln β is considered
for specifically adsorbed ions appears to be as a function of σM ; in contrast, in Kolo-
very complex. tyrkin–Alekseev–Popov model [48], put
At low surface coverage, when ln(1 − forward later in Refs. [49, 50], and also
θ) ≈ −θ, the Frumkin isotherm (34) is in a similar Nikitas’ model [40, 41], ln β is
rearranged into a virial isotherm: a linear or square function of the potential
(b) (1) (1) drop in the inner layer: ϕ = E0 − ϕ2 .
ln βν + ln ci = ln i + 2Bν i (38) As demonstrated in Refs. [51, 52], in
the case of reversible specific adsorption
where
of Cl− , Br− , and I− anions on various
ln βν = ln β + ln max ;
(1)
i = θmax mercury-like electrodes from different
solvents, the values of ln β at σM = 0 and
1 − 2a (b)
EσM =0 at ci = const can be correlated
Bν = (39)
2max with a linear combination of donor (DN)
As mentioned in the preceding text, free and acceptor (AN) numbers of solvents:
energy of adsorption of ions or molecules, y = k0 − k1 DN − k2 AN (40)
and, correspondingly, the value of ln β
(see Eq. 35) depends on one of electrical The quantity of y in Eq. (40), which is
variables: electrode potential E, electrode equal to ln β at σM = 0 [51] or to EσM =0
(b)
at ci = const [52], is the measure of value (namely, for gallium in aprotic sol-
specific adsorption energy of the ith vents), or, finally, remain independent of
ion, and k0 , k1 , and k2 are constants the anion nature (e.g. on In-Ga/acetonitrile
independent of the nature of solvent interface).
for the given metal and ion. In its Let us consider the specific adsorption
physical meaning, the value of k0 should phenomena on the d-metals.
represent the metal-anion bond energy in As was mentioned in the preceding text,
the absence of solvent. the main difference between ‘‘perfectly
In the physical sense, Eq. (40) is based on polarizable’’ and ‘‘ideally polarizable’’ elec-
the supposition that the greater the specific trodes consists in the fact that the potential
adsorption energy of anions should be, the of the former is determined by the pres-
smaller their solvation energy (presented ence of a certain redox system,
by the value of AN) and the weaker
the bond of solvent molecules with the Ox + nē ⇐⇒ Red (41)
uncharged electrode surface (presented
by the value of DN). This is induced by and the state of the system is unam-
the donor–acceptor nature of ion–dipole biguously determined by its initial state
interactions, and also by the interaction of and a quantity of charge passed across
metal with anions and solvent molecules in the interface. For platinum metals in
the course of their competitive adsorption aqueous solutions of surface-inactive elec-
on the uncharged electrode surface. trolytes, Ox1 = H+ , Red1 = Hads (n = 1);
In Ref. [52], the Eq. (40) was com- Ox2 = Oads + 2H+ , Red2 = H2 O (n = 2).
pared with the experimental values of A consistent thermodynamic theory of per-
(b)
y = EσM =0 at ci = 0.1 M, which re- fectly polarizable electrodes was developed
sulted from the specific adsorption of Cl− , in [53–60] on the basis of the initial theory
Br− , and I− anions on the electrodes made of amalgam [13, 61, 62] and platinum hy-
of mercury, gallium, and eutectic alloys In- drogen [63] electrodes. In this case, Gibbs
Ga (16.4 at % In) and Tl-Ga (0.02 at % Tl) equation at T , p = const takes the follow-
in water, acetonitrile, dimethylformamide, ing form:
N -methylformamide, and dimethyl sulfox-
ide. The correlation analysis demonstrated dγ = −Ox dµOx − Red dµRed
that the bond energy in the absence of sol-
vent (k0 ) increases in the sequence I− < − j dµj (42)
Br− < Cl− , in agreement with the row j
for dissociation energy of gaseous mono-
halogenides of mercury and gallium. At in which µOx , µRed , Ox , Red are the
the same time, for various donor–acceptor chemical potentials and surface excesses
properties of solvent, the energy of specific of Ox and Red components, µj , j are the
adsorption on the metal/solution interface chemical potentials and surface excesses
at σM = 0 can either increase in a sequence of other solution components.
Cl− < Br− < I− (this ‘‘usual’’ sequence It follows from Eq. (42) that, when
was observed in aqueous solutions on mer- µj = const, the function γ is a 3D surface,
cury and mercury-like metals), or decrease its intersects with the planes µOx = const
in the same sequence in parallel with k0 and µRed = const being the characteristic
electrocapillary curves described by Lipp- the Nernst Equation for the corresponding
man equations: redox system.
In order to divide the values of F Ox
∂γ and F Red into bound charges (AOx
= +F Red (43)
∂E T ,p,µj ,µOx and ARed ) and free charge σM , certain
assumptions going beyond the frames of
∂γ
= −F Ox (44) thermodynamics are necessary.
∂E T ,p,µj ,µRed In a certain case of a platinum hydro-
gen electrode, on which the equilibrium
The quantities of F Ox and F Red are Hads ⇔ H+ + ē is established, at µj =
called the total charges of the surface at the const,
corresponding choice of values remaining
constant, because these values characterize dγ = −H dµH − H+ dµH+ (45)
the quantities of electricity required for
increasing the surface by a unit area where
without introducing any changes in the
composition of both contacting phases. F H = −Q1 = AH − σM (46)
It is evident that for redox-systems F H+ = Q2 = σM + AH+ (47)
more complex than (41), Gibbs adsorp-
tion equation should also become more that is, total charges F H and F H+ are
complicated. The number of characteris- the algebraic sum of bound (AH , AH+ )
tic Lippman Equations should be equal to and free (σM ) charges. At an excess of
the number of independent variables in surface-inactive electrolyte, AH+ can be
1
0
2
1
Surface tension
3
−200 2
[mN m–1]
4
−400
3 5
−0.8 0 0.8 −0.8 0 0.8

Potential Potential
(a) [V] (b) [V]
Fig. 7 Electrocapillary curves of the first (1a, 2a, 1b, 2b, 3b)
and second (3a, 4b, 5b) (at E = 0 (RHE)) kind for a platinized
platinum electrode in the following solutions:
0.005 M H2 SO4 + 0.5 M Na2 SO4 (1a, 1b);
0.01 M NaOH + 0.5 M Na2 SO4 (2a); 0.01 M HC1 + 1 M
KC1 (2b); 0.01 M HBr + 1 M NaBr (3b); 0.5 M Na2 SO4
(3a, 4b); 0.1 M KI (5b) [64].
equated to zero, and F H+ = σM , that is, allows us to describe the effect of specif-
under certain conditions, F H+ can be ically adsorbed ions on the hydrogen
considered as a free surface charge. adsorption. Inasmuch as ionic adsorp-
The electrocapillary curves correspond- tion affects the hydrogen and oxygen
ing to Eq. (43) are called the electrocap- adsorption, an attempt should be made
illary curves of the first kind, and those in using these effects for determination
corresponding to the Eq. (44) are called of ionic adsorption on platinum group
the electrocapillary curves of the second metals [67].
kind [60, 64]. Figure 7 shows the examples In contrast to mercury-like metals,
of curves of these kinds for Pt electrode. the adsorption of ions on platinum
This figure demonstrates also the effects metals appears to be relatively slow [68],
the solution pH and the anion adsorp- the adsorption rate νads exponentially
tion on electrocapillary curves. It should decreasing with the surface coverage θ
be stressed that pH-dependent adsorption in accordance with Roginskii-Zel’dovich
behavior of Pt hydrogen electrode is a typ- equation
ical feature of hydrogen-adsorbing metals,
(b)
it is governed by Eq. (45), for the latter νads = Kads ci exp(−αf θ) (49)
no assumptions are made on the specific
adsorption of OH− ions. where Kads and α are constants (0 < α <
Therefore, for Pt hydrogen electrode, two 1) and f is the inhomogeneity factor [67].
pzc values should be introduced, which For the middle coverage of platinum
correspond to H = 0 and H+ = 0 (and metals by ions, a logarithmic dependence
to the maxima of the electrocapillary curves of θ on adsorption time tads is observed:
of the 1st and 2nd kinds), respectively. 1
When the concentration of a specifically θ = const1 + ln tads (50)
αf
adsorbed ion i is sufficiently low as
compared with the supporting-electrolyte These data confirm that the adsorption
concentration, so that it can be changed of ions on platinum metals follows the
without changing the chemical potential general tendencies of adsorption on the en-
of supporting ions, the Gibbs Equation ergetically inhomogeneous surfaces. This
takes the following form: conclusion agrees also with the results of
analyzing the isotherms found for ionic
dγ = −H dµH − i dµi (48) adsorption on platinum [68], which follow
where µi is a chemical potential of ad- the logarithmic law (Temkin isotherm):
sorbed ion, and i –its surface excess. 1
From this equation, the quantitative re- θ = const2 + ln c(b) (51)
f
lationships for determination of i versus
electrode potential dependencies by using and also with the data on the isotherms of
the electrode charging curve and the shifts displacement of one ion by another. The
of electrode potential with the concentra- latter proved to be helpful when comparing
tion of adsorbed ions (so-called adsorption the surface activities of ions. Particularly,
shift of potential) [65]. The measurements it was found (Fig. 8) that Cs+ ions
of this sort gave the results that agree with displace Na+ ions adsorbed on platinum
direct data on ionic adsorption obtained by at substantially lower bulk concentration
radiotracer techniques [66]. Equation (48) than that corresponding to the opposite
Fig. 8 Isotherms of displacement of

10 Cs+ cations by Na+ (1), and of Na+
cations by Cs+ (2) on platinized
platinum electrodes at E = 0 (RHE) in
the following solutions:
0.0005 M H2 SO4 + 0.001 M
F ΓCs+, F ΓNa+
1 Cs2 SO4 + x M Na2 SO4 (1); 0.0005 M

[µC cm−2]
5 2 H2 SO4 + 0.001 M Na2 SO4 + x M

Cs2 SO4 (2) [69].
0
−5 −4 −3 −2 −1 0
log c Na2SO4, log c Cs2SO4
[mol l−1]
situation (displacement of adsorbed Cs+ A complete charge transfer makes it

by Na+ ), while the adsorption values of possible to introduce the term ‘‘adatom,’’
Cs+ and Na+ differ only slightly [69]. whereas the phenomenon of adatom for-
It should be stressed that the adsorp- mation, which occurs before the thermo-
tion of isotherms accompanied by charge dynamic potential of the corresponding
transfer differ from those measured in system is reached, was called ‘‘underpo-
the absence of the latter even on uniform tential deposition’’ (upd) (see Chap. 6).
surfaces. In Refs. [70–74], the isotherm of A typical example of adatom forma-
adsorption with charge transfer are consid- tion known for a long time is the ad-
ered, and the field effect is introduced by sorption of hydrogen ions on platinum,
two different ways: via the Nernst Equation which is accompanied by charge trans-
and by considering the displacement of wafer and consequent transition to the
ter molecules whose adsorption changes atomic state. Oxygen adsorption can also
with potential. This is why the author be considered to be reversible in a cer-
refers to this as combined isotherm. tain potential region preceding the oxygen
In as much as the specific adsorp- evolution.
tion is frequently associated with the On the basis of the generalized concept
donor–acceptor ion–metal interaction, of the electrode charge, a consistent
the state of specifically adsorbed ions re- phenomenological treatment of charge
quires special attention. It is supposed that transfer during the chemisorption was
specific adsorption can result either in par- developed [57, 77]. Let us assume that the
tial or in complete charge transfer between process of type (41) but involving only one
adsorbed species and metal. Although the electron occurs through an intermediate
idea of partial charge transfer was put for- adsorption state A:
ward by Lorenz [75, 76] in connection with
ionic adsorption on mercury, the most im-
Ox + n1 ē ⇐⇒ A (52)
pressive confirmations were obtained for
d-metals. A + n2 ē ⇐⇒ Red (53)
It is evident that n1 + n2 = 1. The constant potential. This is why, it is hardly

quantities n1 and n2 are called the formal appropriate to consider nj as the ‘‘elec-
(thermodynamic, macroscopic) charge- trosorption valence’’ [78–81]. If n electrons
transfer coefficients. In the case of ad- take part in redox process, we should use
sorption of Ox, the process may stop at (n1 /n) and (n2 /n) instead of n1 and n2 .
stage (52), and, in the case of adsorption The formal charge-transfer coefficient
of Red, it may stop at stage (53). It can be can be experimentally determined;
shown [57] that the formal charge-transfer whereas, the determination of the
coefficients are determined by the relation- true transfer coefficient based on
ships: electrochemical data requires making
assumptions outside the scope of
1 ∂Q1 ∂Red thermodynamics [82, 83]. In principle,
n1 = − = (54)
F ∂ E ∂ E the latter value can be determined

1 ∂Q2 ∂Ox from nonelectrochemical, particularly
n2 = = (55) spectroscopic techniques sensitive directly
F ∂ E ∂ E
to the electronic states of species and/or
where = Ox + Red . the metal.
The values n1 and n2 designate the Charge transfer can be evidenced by the
formal coefficients of charge transfer, enhanced adsorption of ions that proceeds
which characterize the overall effect of simultaneously with the less-pronounced
introducing the species into the surface adsorption of oppositely charged ions, the
layer – both the true (microscopic) charge latter compensating the superequivalent
transfer from species to the electrode or adsorption [66]. Namely, the adsorption of
vise versa, and any charge redistribution Na+ on platinized platinum in the re-
in the EDL. Thus, ni differs from zero gion of positive charges increases in the
when the charge of an adsorbed species re- sequence Na2 SO4 < NaI < NaCl < NaBr
mains unchanged, however, its adsorption (Fig. 9), whereas the adsorption of an-
changes the double-layer capacitance at a ions grows in a row SO4 2− < Cl− <
6
[µC cm−2]
4 3
F ΓNa+
2
Fig. 9 Dependencies of Na+ 2
adsorption on the potential of a 4
platinized platinum electrode in the
1
following solutions: 0.0005 M 0
H2 SO4 + 0.0015 M Na2 SO4 (1); 0 0.2 0.4 0.6 0.8 1.0
0.001 M HC1 + 0.003 M NaC1 (2);
0.001 M HBr + 0.003 M NaBr (3); Potential
0.0005 M H2 SO4 + 0.003 M NaI [66]. [V, RHE]
Br− < I− . Thus, the sequences for specific solution of this problem, the conclusion
adsorbability and superequivalent adsorp- on the pzfc (potential of zero free charge)
tion disagree. This phenomenon can be shift induced by specific adsorption can
explained under a supposition that the change. As demonstrated in Ref. [84],
charge-transfer degree increases with the the pzfc of a platinum electrode can
adsorbability. The approximately constant shift substantially towards more positive
or even slightly decreasing adsorption of and more negative potential values when
Na+ observed with the growth of potential I− and Tl+ , respectively, are adsorbed.
indicates that the charge-transfer degree Thus, the direction of the pzfc shift is
increases as the potential shifts towards opposite as compared with that induced
more positive values. by the adsorption of the same ions on
The charge transfer, which accompanies mercury-like metals. Hence, under the
the adsorption processes, can affect the conditions of charge transfer, we can
latter because the ionic charges are discuss only conventional zero free charge
spread over the surface in a peculiar potentials, because, actually, the nature of
manner. Consequently, the Esin-Marov the electrode surface changes.
effect can be substantially weakened. Complete charge transfer (upd) is possi-
A direct determination of Esin-Markov ble when the support-adsorbate interaction
coefficient on platinum group metals [54] is stronger than the lateral interaction in
demonstrates no effect altogether, which the adlayer. Usually, upd induces mono-
can be explained by a pronounced charge layer coverage; however, for certain sys-
transfer from adsorbed anions. According tems the formation of second and even
to the thermodynamic theory of the third layer turns out to be possible, and
platinum hydrogen electrode, there are two these layers do not resemble the corre-
Esin-Markov coefficients corresponding to sponding bulk deposits.
the constant values of free and total charges Consideration of perfectly polarizable
respectively. When comparing the data for electrodes presented in the preceding text
Pt and Hg electrodes, one should use the is related to equilibrium conditions (E =
former coefficient. const). When dealing with nonpotentio-
Charge transfer manifests itself also static modes (linear, stepwise, sinusoidal,
by a slow establishment of adsorption etc.), one always observes not only double-
equilibria mentioned in the preceding layer capacity but also pseudocapacity
text, a slow exchange between adsorbed induced by the changes of surface cov-
ions and solution, temperature effects, erage with time in the course of reactions
and the adsorption isotherms of strongly (52, 53). Conway, Gileadi and coworkers
surface-active ions. The latter isotherms intensively studied this problem Refs. [85,
demonstrate a slope close to that known 86].
for the adsorption of neutral and low- An application of the thermodynamic
polar species, namely, Hads , for which the theory of Pt hydrogen electrode to the re-
Temkin isotherm is applicable [67]. versible adsorption of organic substances
For charge transfer from adsorbed ions leads to a conclusion [87] that the max-
to the metal surface, a problem arises as to imum adsorption of neutral molecules
whether this charge should be attributed should be observed in the potential region,
to the ‘‘electrode’’ part of the double-layer where the adsorption of both hydrogen
or to its ionic part. Depending on the and oxygen is low (so-called double-layer
region). However, this conclusion was ions, makes it possible to clarify the roles
never checked, because for the majority the ion solvation and the metallic surface
of adsorbates, the destruction of organic solvation play in the interfacial structure,
molecules in the course of adsorption and, particularly, understand the mecha-
was found [88], which was accompanied nism of solvophilicity of metals.
by their hydridization, dehydridization, as- In the potential region where forma-
sociation of adsorption products, and so tion of chemisorbed oxygen layers and
forth. Hence, their adsorption appears to phase oxides takes place, the adsorption
be irreversible, and the composition of phenomena become irreversible and ex-
adsorbed species differs from that in the tremely complicated, because almost all
(ads) (b)
solution bulk. In this case, µ̄i = µ̄i ; solution components can take part in
no exchange between adsorbed layer and electrode reactions and undergo electro-
solution occurs, the species cannot be chemical transformations. The adsorption
desorbed in pure solvent, and no unam- peculiarities under these conditions differ
biguous relationship exists between the strongly from those known for d-metals
bulk and surface concentrations of adsor- in the absence of oxygen adsorption. For
bates. Certain inorganic ions (like I− in example, the adsorbability sequence for
acid media) are also adsorbed irreversibly halide ions on Pt at E > 1.5 V (RHE) trans-
on platinum. forms into I− < Br− < Cl− < F− [93]. It
In parallel with complications induced is assumed that anions penetrate into
by irreversible adsorption, new approaches the oxide layer with parallel dehydration,
to the experimental studies of surface and their susceptibility to injection corre-
excesses are developed. Namely, the adsor- lates with the reciprocal crystallographic
bate’s state and coverage usually remain radia. The adsorption versus potential de-
unchanged when the potential is switched pendence appears to be nonmonotonous
off and the electrode withdrawn from so- and demonstrates a number of maxima;
lution. Moreover, some adsorbed layers moreover, in a number of cases, the max-
remain stable even under UHV conditions, ima of anion adsorption correspond to the
and a number of highly informative ex situ maxima of cation adsorption [94–98]. It
spectroscopic techniques were success- was also found that organic substances
fully applied to the studies of adsorption could adsorb at high anodic potentials on
on d-metals [89–91]. platinum group metals [99]. The mecha-
Modeling of electrochemical interfaces nisms of these phenomena still remain
under UHV conditions supplements the unclear.
aforementioned approaches. For example, In principle, the theory of reversible elec-
it exploits the controlled modification of trodes can serve as the basis for developing
the metal/vacuum interface by the various the thermodynamic approach to the sur-
components typical of electrochemical in- face electrochemistry of oxide materials
terfaces (ions, solvent molecules) and reg- with metallic conductivity (iridium, ruthe-
istration of infrared reflection-absorption nium, tin dioxides, and so on) [100]. In
spectra along with work function mea- addition, the properties of these interfaces
surements [92]. A gradual increase in the can be considered within the framework of
surface coverage by solvent molecules, three models: the classical model of bound
which changes from a submonolayer up sites [101], which can be modified by con-
to multilayers at various doses of adsorbed sidering several types of surface groups
(MUSIC model) [102, 103]; various modi- and nanoheterogeneity should be stressed
fications of Grahame-Parsons model [104]; because rather unexpected features of
and the model of nonlocal electrostat- adlayers result from comparable sizes of
ics [105]. On oxide surfaces, functional adsorbate molecule and surface fragment.
groups are present, which can be ionized One of the aspects of this problem
reversibly upon their contact with the elec- is the effect of long-living metastable
trolyte solution: surface defectiveness that is responsible
for ‘‘adsorption traps,’’ the surface centers
MOH2 + ⇐⇒ MOH + H+ (56) with anomalously high adsorption ability.
− + Specific adsorption is rated among the
MOH ⇐⇒ MO + H (57)
most important phenomena in surface
MOH2 + . . . A− ⇐⇒ MOH + H+ + A− science. Its unique importance for electro-
(58) chemistry is dictated by the role of specific
adsorption of reactants and products, and
MOH + C+ ⇐⇒ MO− . . . C+ + H+ also of inert solution components in the
(59) kinetics of electrode reactions. Adsorption
where C+ and A− are cations and anions can change not only reaction rates but also
of the electrolyte. In this case, the specific the mechanisms.
adsorption of certain components can be
considered as an increase in the binding References
degree of this component with the surface
functional groups [106]. 1. J. W. Gibbs, The Collected Works of J. W.
For oxide/solution interfaces, a pro- Gibbs, Longmans, New York, 1931, p. 219,
nounced correlation was revealed be- Vol. 1.
2. E. A. Guggenheim, N. K. Adam, Proc. R.
tween the type of organic compound and Soc. London 1933, A139, 218–236.
their adsorption [107]. Such selectivity was 3. A. W. Adamson, Physical Chemistry of Sur-
noted for adsorption on metals and can faces, John Wiley & Sons, New York, 1976,
point to the chemisorption nature of the p. 68.
4. M. Planck, Ann. Phys. 1891, 44, 385–428.
metal–adsorbate interaction [108].
5. G. Gouy, Compt. Rend. 1908, 146,
A modern stage in the development of 1374–1376.
notions of specific adsorption is masked 6. O. Stern, Z. Elektrochem. 1924, 30, 508–516.
by a changeover to describing this phe- 7. M. A. Gerovich, Doklady AN SSSR 1954, 96,
nomenon on a microscopical level, using 543–546.
8. E. Blomgen, J. O’M. Bockris, J. Phys. Chem.
computational approaches to simulation 1959, 63, 1475–1484.
of adsorbed layers. We can expect a rapid 9. M. A. Gerovich, Doklady Akad. Nauk SSSR
progress in these directions, because the 1955, 105, 1278–1281.
information on the microscopic structure 10. E. Blomgen, J. O’M. Bockris, C. Jesch,
J. Phys. Chem. 1961, 65, 2000–2010.
of adlayers becomes progressively more
11. H. O. Finklea in Electroanalytical Chemistry
available because of studies on single- (Eds.: A. J. Bard, I. Rubinstein), Marcel
crystal electrodes and the results of physi- Dekker, New York, 1996, pp. 109–335.
cal surface science techniques. 12. G. Lippmann, Ann. Chim. Phys. 1875, 5,
There is a number of novel problems in 494–549.
13. A. N. Frumkin, Elektrokapillyarnye yavleniya
the field of specific adsorption induced i elektrodnye potentsialy (Electrocapillary
by fast progress of material science. Phenomena and Electrode Potentials), Ti-
Particularly, the role of nanostructure pographiya Sapozhnikova, Odessa, 1919.
14. A. Frumkin, Z. Phys. Chem. 1924, 109, 41. P. Nikitas, Electrochim. Acta 1996, 41,
34–48. 2159–2170.
15. A. Frumkin, Z. Phys. Chem. 1923, 103, 42. R. Fowler, E. A. Guggenheim, Statistical
43–70. Mechanics, Cambridge University Press,
16. A. Frumkin, Zh. Fiz. Khim. 1956, 30, Cambridge, 1956.
2066–2069. 43. R. de Levie, J. Electrochem. Soc. 1971, 118,
17. O. A. Esin, B. F. Markov, Acta Physicochim. 185C–192C.
URSS 1939, 10, 353–364. 44. M. A. Vorotyntsev, J. Res. Inst. Catalysis,
18. G. Gouy, J. Phys. Radium 1910, 9, 457–466. Hokkaido Univ. 1982, 30, 167–177.
19. D. L. Chapman, Philos. Mag. 1913, 25, 45. B. Damaskin, U. Palm, M. Salve, J. Elec-
475–483. troanal. Chem. 1987, 218, 65–76.
20. D. C. Grahame, Chem. Rev. 1947, 41, 46. D. C. Grahame, R. Parsons, J. Am. Chem.
441–501. Soc. 1961, 83, 1291–1296.
21. D. C. Grahame, B. A. Soderberg, J. Chem. 47. J. M. Parry, R. Parsons, Trans. Faraday Soc.
Phys. 1954, 22, 449–460. 1963, 59, 241–256.
22. H. D. Hurwitz, J. Electroanal. Chem. 1965, 48. Ya. M. Kolotyrkin, Yu. V. Alekseev, Yu. A.
10, 35–41. Popov, J. Electroanal. Chem. 1975, 62,
23. E. Dutkiewicz, R. Parsons, J. Electroanal. 135–149.
Chem. 1966, 11, 100–110. 49. B. Damaskin, S. Karpov, S. Dyatkina et al.,
24. J. Lipkowski, C. Nguen Van Huong, C. Hin- J. Electroanal. Chem. 1985, 189, 183–194.
nen et al., J. Electroanal. Chem. 1983, 143, 50. B. B. Damaskin, O. A. Baturina, Russ. J.
375–396. Electrochem. 1999, 35, 1186–1193.
25. L. Stolberg, J. Lipkowski, D. E. Irish, J. Elec- 51. W. R. Fawcett, A. Motheo, J. Electroanal.
troanal. Chem. 1987, 238, 333–353.
Chem. 1994, 375, 319–338.
26. A. Hamelin, S. Morin, J. Richter et al.,
52. B. B. Damaskin, Russ. J. Electrochem. 1994,
J. Electroanal. Chem. 1989, 272, 241–252.
30, 91–94.
27. L. Stolberg, J. Lipkowski, D. E. Irish, J. Elec-
53. A. Frumkin, N. Balashova, V. Kazarinov, J.
troanal. Chem. 1990, 296, 171–189.
Electrochem. Soc. 1966, 113, 1011–1018.
28. L. Stolberg, J. Lipkowski, D. E. Irish, J. Elec-
54. A. Frumkin, O. Petry, A. Kossaya et al., J.
troanal. Chem. 1992, 322, 357–372.
29. A. Hamelin, S. Morin, J. Richter et al.,
J. Electroanal. Chem. 1990, 285, 249–262. 55. A. Frumkin, O. Petry, Electrochim. Acta
30. A. Hamelin, S. Morin, J. Richter et al., 1970, 15, 391–403.
J. Electroanal. Chem. 1991, 304, 195–209. 56. A. N. Frumkin, O. A. Petrii, B. B. Damas-
31. W. R. Fawcett, J. Electroanal. Chem. 1972, kin, J. Electroanal. Chem. 1970, 27, 81–100.
39, 474–477. 57. B. Grafov, E. Pekar, O. Petrii, J. Electroanal.
32. A. Frumkin, Z. Phys. Chem. 1925, 116, Chem. 1972, 40, 179–186.
466–484. 58. A. N. Frumkin, B. B. Damaskin, O. A. Petrii,
33. A. Frumkin, Z. Physik 1926, 35, 792–802. J. Electroanal. Chem. 1974, 53, 57–65.
34. R. Parsons, J. Electroanal. Chem. 1964, 7, 59. A. N. Frumkin, Potentsialy nulevogo zaryada
136–152. (Potentials of Zero Charge), Nauka, Moscow,
35. J. O’M. Bockris, M. A. V. Devanathan, 1982.
K. Müller, Proc. R. Soc. London 1963, A274, 60. A. N. Frumkin, O. A. Petrii, Electrochim.
55–79. Acta 1975, 20, 347–359.
36. J. O’M. Bockris, E. Gileadi, K. Müller, Elec- 61. A. Frumkin, Philos. Mag. 1920, 40, 363–375.
trochim. Acta 1967, 12, 1301–1321. 62. A. Frumkin, Philos. Mag. 1920, 40, 375–385.
37. B. Damaskin, A. Frumkin, A. Chizhov, 63. A. Frumkin, A. Slygin, Acta Physicochim.
J. Electroanal. Chem. 1970, 28, 93–104. URSS 1936, 5, 819–840.
38. M. V. Sangaranarayanan, S. K. Rangarajan, 64. A. N. Frumkin, O. A. Petrii, Doklady Akad.
Can. J. Chem. 1981, 59, 2072–2079. Nauk SSSR 1971, 196, 1387–1390.
39. R. Guidelli, G. Aloisi, J. Electroanal. Chem. 65. O. A. Petrii, Yu. G. Kotlov, Elektrokhimiya
1992, 329, 39–58. 1968, 4, 1256–1260.
40. P. Nikitas, J. Electroanal. Chem. 1994, 375, 66. V. E. Kazarinov, O. A. Petrii, V. V. Topolev
319–338. et al., Elektrokhimiya 1971, 7, 1365–1368.
67. V. S. Bagotzky, Yu. B. Vassiliev, J. Weber 89. A. Hubbard, J. Stickney, M. Soriaga et al.,
et al., J. Electroanal. Chem. 1970, 27, 31–46. J. Electroanal. Chem. 1984, 168, 143–166.
68. N. A. Balashova, V. E. Kazarinov in Electro- 90. P. Andricacos, P. Ross, J. Electroanal. Chem.
analytical Chemistry (Ed.: A. J. Bard), Marcel 1984, 167, 301–308.
Dekker, New York, 1969, pp. 135–197, 91. A. Hubbard, Heterog. Chem. Rev. 1994, 1,
Vol. 3. 3–39.
69. B. B. Damaskin, O. A. Petrii, V. E. Kazari- 92. M. J. Weaver, I. Villegas, Langmuir, 1997,
nov, Elektrokhimiya 1972, 8, 1373–1377. 13, 6836–6844.
70. E. Gileadi, Collect. Czech. Chem. Commun. 93. V. E. Kazarinov, Elektrokhimiya 1966, 2,
1971, 36, 464–475. 1389–1394.
71. E. Gileadi, Isr. J. Chem. 1971, 9, 405–412. 94. N. V. Pospelova, A. A. Rakov, V. I. Veselov-
72. E. Gileadi, G. E. Stoner, J. Electrochem. Soc. skii, Elektrokhimiya 1969, 5, 797–803.
1971, 118, 1316–1319. 95. A. A. Yakovleva, R. K. Bairamov, V. I. Vese-
73. E. Gileadi, Discuss. Faraday Soc. 1973, 56, lovskii, Elektrokhimiya 1975, 11, 515–518.
228–234. 96. V. I. Naumov, Yu. M. Tuyrin, Elektrokhimiya
74. E. Gileadi, Electrochim. Acta 1987, 32, 1973, 9, 1011–1016.
221–229. 97. V. I. Naumov, V. E. Kazarinov, Yu. M.
75. W. Lorenz, Z. Phys. Chem. 1961, 218, Tuyrin, Elektrokhimiya 1973, 9, 1412–1413.
272–276. 98. N. V. Pospelova, A. A. Rakov, V. I. Vese-
76. W. Lorenz, Z. Phys. Chem. 1973, 254, lovskii, Elektrokhimiya 1969, 5, 1318–1320.
123–133. 99. L. M. Mirkind, M. I. Fioshin, Doklady Akad.
77. A. Frumkin, B. Damaskin, O. Petrii, Z. Nauk SSSR 1964, 154, 1163–1166.
Phys. Chem. 1975, 256, 728–736. 100. O. A. Petrii, Electrochim. Acta 1996, 41,
78. K. Vetter, J. Schultze, Ber. Bunsen-Ges. Phys. 2307–2312.
Chem. 1972, 76, 920–927. 101. D. E. Yates, S. Levine, T. W. Healy, J. Chem.
79. J. Schultze, K. Vetter, J. Electroanal. Chem. Soc., Faraday Trans. 1 1974, 70, 1807–1818.
1973, 44, 63–81. 102. T. Hiemstra, W. H. Van Reimsdijk, G. H.
80. K. Vetter, J. Schultze, Ber. Bunsen-Ges. Phys. Bolt, J. Colloid Interface Sci. 1989, 133,
Chem. 1972, 76, 927–933. 91–104.
81. K. Vetter, J. Schultze, J. Electroanal. Chem. 103. T. Hiemstra, W. H. Van Reimsdijk, G. H.
1974, 53, 67–76. Bolt, J. Colloid Interface Sci. 1989, 133,
82. J. Schultze, F. Koppitz, Electrochim. Acta 105–117.
1976, 21, 327–336. 104. B. B. Damaskin, Elektrokhimiya 1989, 25,
83. F. Koppitz, J. Schultze, Electrochim. Acta 1641–1648.
1976, 21, 337–343. 105. M. A. Vorotyntsev, A. A. Kornyshev, Elek-
84. A. N. Frumkin, Zh. N. Malysheva, O. A. trostatika sred s prostranstvennoi dispersiei (in
Petrii et al., Elektrokhimiya 1972, 8, 599–603. Russian) (Electrostatics of the Spatially Dis-
85. B. E. Conway, E. Gileady, Trans. Faraday persed Media), Nauka, Moscow, 1993.
Soc. 1962, 58, 2493–2509. 106. H. Tamura, T. Oda, M. Nagayama et al.,
86. B. E. Conway, E. Gileady in Modern Aspects J. Electrochem. Soc. 1989, 136, 2782–2786.
of Electrochemistry (Eds.: J. O’M. Bockris, 107. E. A. Nechaev, Zh. Fiz. Khim. 1978, 52,
B. E. Conway), Butterworths, London, 1964, 1494–1496.
pp. 347–442, Vol. 3. 108. M. I. Urbakh, E. A. Nechaev, Elektrokhimiya
87. A. N. Frumkin, Doklady Akad. Nauk SSSR 1980, 16, 1264–1268.
1964, 154, 1432–1436.
88. B. B. Damaskin, O. A. Petrii, V. V. Batrakov,
Adsorption of Organic Compounds on Elec-
trodes, Plenum Press, New York, 1971.
3.2 State of Art: Present Knowledge and Understanding 349
3.2 basis for the elucidation of the problems

State of Art: Present Knowledge and connected with electrified interphases.
Understanding In constructing models for interfacial
structure, only long-range electrostatic
G. Horányi
effects were considered and aside from
Institute of Chemistry, Budapest, Hungary
the magnitude of the charges on the ions
3.2.1 (and their radii) the differences in the
Historical Background chemical nature of the different ions were
ignored. The well-known Gouy–Chapman
The thermodynamic treatment of ideally (GC) model is based on this approach
polarizable interphase (the term used (see Chapter 2.2, Sect. 2.2.2). However, the
when no charged component is common experimental observations furnished more
in two adjoining phases) leading to and more evidences that the behavior
the Gibbs adsorption equation owing of several systems deviates significantly
to the nature of the thermodynamic from that expected on the basis of GC
approach conveys no information about theory. One of the well-known examples
the structure of the interphase and has was the significant deviation of the surface
nothing to do with the forces playing role excess of anions measured at Hg electrode
in the formation of this structure. at positive potentials. It was reasonable
The strict thermodynamic analysis of the to assume that these deviations should
interphase is based on data available from be ascribed to short-range interactions
the bulk phases (concentration variables) between the ions and the surface of the
and the total amount of material involved metal. As these types of interactions should
in the whole system figuring in the be specific for both metal and ions, the
relations expressing the relative surface adsorption resulted by these forces could
excess of suitably chosen (charged or be named as specific adsorption.
not charged) components of the system. Although different definitions and
In addition, the Gibbs equation for a names for this phenomenon have been
polarizable interphase contains a member proposed, the most general definition
related to the potential difference between (accepted by IUPAC as well) is an
one of the phases (metal) and a suitably operative one defining specific adsorption
chosen reference electrode immersed in as the deviation from the extent of
the other phase (solution) (and attached to the adsorption predicted by GC theory.
a piece of the same metal that forms one In contrast to this, the adsorption of
of the phases). ions ‘‘obeying’’ GC theory is termed as
The only model used in the case of nonspecific adsorption. It is important to
such a treatment is the assumption that emphasize that the distinction between
the existence of charged particles (ions specific and nonspecific adsorption is
and electrons) in the bulk phases can be based on an arbitrarily chosen model
considered as a physical reality. assumption. It should, however, be borne
In the historical development of the con- in mind that it is impossible, by means of
cept of electrified interphase, the modeling thermodynamic arguments, to determine
preceded the strict thermodynamic de- the extent of the contribution of the specific
scription and thus for a long period, only adsorption to the overall surface excess
the double-layer model served as a unique concentration and to the charge density.
This means that the concept of specific potentials of the metal and electrolyte, re-
adsorption involves the use of nonther- spectively. It is often convenient to take φ S
modynamic methods and model assump- as zero.
tions. The surface free charge density in the
For the orientation of the reader, we diffuse layer is denoted by σ d and it
summarize the fundamental elements of may be regarded as made up of the
the characteristic features of the double- charges contributed by the surface excess
layer model relating to the specific ad- or deficiency of each ionic species (i )
sorption on the basis of the correspond-
ing IUPAC recommendations [1]. For a σd = σid = zi ndi = zi i (1)
detailed survey including historical back-
the sum being overall ionic species in the
ground, see elsewhere [2–7]. solution.
On the basis of the formal relationship
3.2.2
The Structure of the Double Layer d(φ M − φ S ) d(φ M − φ2 )
=
dσ M dσ M
Recapitulation of fundamental notions In
d(φ2 − φ S )
the case of nonspecific adsorption it is + (2)
dσ M
assumed that ions retain their solvation
shell, and in the position of closest the differential capacitance of the whole
approach to the interface they are separated interphase can be conceived as composed
from it by one or more solvent layers. of two components, so Eq. (2) may be
The locus of the electrical centers of written as
nonspecifically adsorbed ions in their
C −1 = (C i )−1 + (C d )−1 (3)
position of closest approach is the outer
Helmholtz plane (OHP). that is, the interphase behaves as a
The position of this plane for the anion series combination of C i , the inner-
and the cation of the electrolyte may differ. layer capacitance, and C d , the diffuse-layer
It may also differ for the different cations capacitance.
and the different anions of solution. The
region in which nonspecifically adsorbed The double layer in the case of specific
ions are accumulated and distributed by adsorption In the case of specifically ad-
the contrasting action of the electric field sorbed ions it is assumed that they pen-
and thermal motion is called the diffuse etrate into the inner layer and may (but
layer. The region between the OHP and not necessarily) come in contact with the
the interface is called the inner (compact) metal surface. They are usually assumed
layer. to form a partial or complete monolayer.
In accordance with the IUPAC defi- The locus of the electrical centers of this
nitions [1] denoting by x2 , the distance layer of specifically adsorbed ions is the
of OHP from the interface, and by φ2 , inner Helmholtz plane (IHP) assumed to be
the mean electric potential at OHP, the at a distance x1 from the metal surface. In
potential drop across the inner layer is certain cases the amount of charge of the
(φ M − φ2 ), while in the diffuse layer it is ions specifically adsorbed in the IHP (σ i ) is
(φ2 − φ S ) where φ M and φ S are the inner higher than the charge on the metal phase
(with the reverse sign); thus the charge of adsorption. The potential drop across the
the diffuse layer has the same sign as that inner layer may be assumed to depend on
on the metal (superequivalent adsorption). the charge σ M and specifically adsorbed
If the contribution to the solution charges σji . From this assumption it fol-
charge from ions whose centers lie in lows generally that
the region x1 < x < x2 is denoted by
σ i , the electroneutrality condition for the M ∂(φ M − φ2 )
d(φ − φ2 ) = dσ M
interphase will be as follows: dσ M σji

σ M = −σ S = −(σ i + σ d ) (4) ∂(φ M − φ2 )
+ dσji (9)
dσji M i
The separation of σ S into σ i , and σ d σ ,σk
cannot be made without the introduction or
of a model. Use of the GC theory and

the assumption that one of the ionic d φ M − φ2 1 ∂ φ M − φ2
≡ i =
components is not specifically adsorbed dσ M C ∂σ M
σji
enables the other σjd to be calculated and

so the specifically adsorbed charge due to ∂ φ M − φ2 dσji
the other ion to be obtained: +
∂σji dσ M
σ M ,σki
σji = σjS − σjd (5) (10)
Thus the inner-layer capacitance may be
In view of the short-range of the forces related to a series of partial capacitances
causing specific adsorption, the quantities that depend on σ M :
σji are often equivalent to less than one
monolayer of ions. Consequently the real ∂σ M
i
C σ M
=
(11)
distribution is often represented in models ∂ φ M − φ2 i
σj
as a monolayer of ions with their centers
on the IHP.
and each σji for the specifically ad-
The charge in the diffuse layer is now
sorbed ions:

σ d = −σ M − σji (6) ∂σji
i
C σj =i
(12)
and the corresponding differential capaci- ∂ φ M − φ2 M i σ ,σk
tance (Csd ) of the diffuse layer is
Here the subscript σki means that all σji are
dσ d kept constant except one.
Csd = − (7)
d(φ2 − φ S )
3.2.3
Thus Eq. (3) can be replaced by Thermodynamics and Models

i −1
1+d σj 3.2.3.1 The Extent of Adsorption
C −1 = (C i )−1 + (C d )−1
dσ M In the previous section it was emphasized
(8) that the charge distribution in the double
The inner-layer capacity can be further ana- layer cannot be described without model
lyzed to express its dependence on specific assumptions concerning both the specific
adsorption and the electrostatic adsorp- In the case of models, an adsorption

tion. It should, however, be taken into layer of some definite thickness is as-
consideration that a strict thermodynamic sumed; therefore, we can write
treatment of the interfacial phenomena
lead to the determination of the extent ni = nai + nSi
of adsorption in terms of relative surface nw = naw + nSw (15)
excesses, quantities that are not depend-
ing on the choice of the position of the where the indices a and S refer to the
plane of the interface; however, through adsorption and liquid phase, respectively.
the Gibbs–Duhem equation they depend Since
on which component is taken as a basis xi nS
= S (16)
for the expression of the relative surface xw nw
excesses of the other components [8]. In
electrochemistry, mostly the solvent is con- we obtain the relation
sidered as such basis.
1 a a xi
Thus, the relative surface excess con- i = ni − nw (17)
A xw
centration (i ) of some ith component
referred to the solvent (water) is given by If a monomolecular adsorption layer is
the equation: assumed, for the total surface area A
we have

1 xi xi Ai nai = A (18)
i = ni + nw = i − w
A xw xw
(13) where Ai is the surface requirement of one
in the case of any arbitrarily chosen mole of the ith component. From this
reference system, where ni and nw are the
numbers of moles of the ith component A− Ai nai
and of the water, respectively, in the entire i =w
naw = (19)
system, xi and xw are the corresponding Aw
equilibrium concentrations, i the surface
excess of component i with respect to and hence
the arbitrarily chosen reference system  
1 A− Ai nai xi
defined by i = nai −  (20)
A Aw xw
1
i = (ni − nr xi ) (14)
A In dilute solutions xi xw and xw ≈ 1,
and therefore
where nr is the total amount (in moles)
1 a
of the chosen reference system. It may i ≈ n (21)
A i
be seen from these simple relationships
that while the expression for relative sur- That is, the thermodynamically deter-
face excess contains only easily measurable minable relative surface excess amount is
variables characterizing the system stud- practically the same as the total amount
ied, the i surface excess strongly depends of substance in the monomolecular layer,
on an arbitrary chosen parameter, nr . and on the basis of the aforementioned
text, the coverage i relating to the ith by the equation:

component is
1
Ai nai i = (ni − nri ) (27)
A
Ai i = = i (22)
A
where ni is the amount of component i
In the sense of the outlined model, the in the whole system and nri is the amount
surface is always completely covered, and of component i in the reference system.
thus the adsorption of some dissolved The difference between i and i can be
component S can in essence be regarded expressed as
as an exchange process, characterized in
1 xi a
the simplest case by the equation i = i − i = n
A xw w
SS + H2 Oa −
−
−−−
− Sa + H 2 O S (23) 1 xi
xi
=− nw − nrw = − w (28)
In a system consisting of solvent (water) A xw xw
and a single solute, we have for this According to Eq. (28), at a constant solu-
equilibrium tion composition, the smaller the absolute
value of w , the smaller is the absolute
aSa awS fSa fwS xSa xwS
K= = · (24) value of the difference between surface ex-
aSS awa fSS fwa xSS xwa cess concentrations and (relative) surface
excesses. Since the extent of the refer-
The a values are the activities referring to
ence phase can be selected arbitrarily, and
the adsorption and solution phases, while
the value of i depends directly on this
the f values are the activity coefficients.
choice, the latter is ill defined from a
From the above relations it follows that
thermodynamic point of view. However,
xSS f a f S na f af S Aw naS by introducing some nonthermodynamic
K = S wS · aS = SS w ·
xwS fwa fS nw fS fwa A − AS naS assumptions, the value of i can be es-
timated for several special cases. If the
fSa fwS Aw S above-mentioned monolayer model can be
= · · (25)
fw fw AS 1 − S
S a considered as a reliable description of the
interphase, the value of w is small, and
In dilute solutions, fSS /fwS can be regarded the concentration of the species i in the
as constant, and in ideal mixtures as unity. solution (xi ) is also low, therefore i is
Further, xw ≈ 1, and thus negligible, and Eq. (22) is always true.
fSa Aw S The thickness of the adsorption layer, the
KxSS = · (26) diffuse part of the double layer, strongly de-
fwa AS 1 − S
pends on the total ion concentration, on the
This is in effect the equation describing ionic strength. Consequently, if the ions in-
the adsorption isotherm (see later). If the volved in specific adsorption are present at
adsorption phase is also regarded as an a low concentration and the total ionic con-
ideal mixture, that is, (fSa /fwa ) ≈ 1, we centration is also rather low, the monolayer
obtain the Langmuir isotherm. model cannot be used without restrictions.
If we consider the problem in a more There is a very important point to be
generalized form, the surface excess con- stressed. Although i values are arbitrary
centrations i per definition can be given depending on the choice of the reference
system, the charge on both sides of the amount adsorbed (or surface excess) as
double layer expressed through zi i a function of bulk activity of the same
according to species at constant temperature. In the
case of electrified interphases the electrical
σ M = −σ S = − zi i = − zi i state of the interphase must also be
(29) kept constant for the determination of
does not depend on this choice as the relation between surface and bulk
1 concentrations.
zi i = ni zi − nr xi zi Two ways for this are followed in the
A
practical determination of the isotherms.
1
= ni zi (30) These are the determination at constant
A σ M or at constant (φ M − φ S ) (or at

1 nw some constant cell potential). From a
zi i = ni zi − xi zi
A xw thermodynamic point of view, isotherms
1 with respect to relative surface excesses
= ni zi (31) may be determined at constant charge
A
or at any well-defined constant potential.
considering the very fact that owing to However, the interpretation and physical
the electroneutrality
of the bulk solution meaning of the results may be significantly
phase, xi zi = 0. more difficult in the case when constant
cell potential is used.
3.2.3.2 Adsorption Isotherm In the case of adsorbed layers, by analogy
Adsorption in systems of neutral mole- with the two-dimensional pressure of
cules is frequently expressed in terms spread films, a surface pressure defined
of an adsorption isotherm that gives the as the change of interfacial tension (γ )
Tab. 1 Adsorption isothermsa
Name Isotherms
Henry βai = RTθ

θ
Langmuir βai =
1−θ
RTi bi
Volmer βai = exp
1 − bi 1 − bi
RTi bi
van der Waals βai = exp exp(−2rθ)
1 − bi 1 − bi
Virial βai = i exp(−2rθ)
θ
Frumkin βai = exp(−2rθ)
1−θ
θ b 1 + θ(1 − θ)
Modified H.F.L. (Helfand, Frisch, and βai = exp exp(−2rθ)
Lebowitz, modified by Parsons) 1 − θ 0.907 (1 − θ )2
a a = activity of the ions; r = interaction parameter determinable from experiment in isotherms

i
1–7; b = van der Waals constant; and θ = fraction of the surface covered by ions.
caused by the addition of a given species to For this case the Gibbs adsorption
a base solution is introduced. At constant equation takes the form
cell potential this is

−dγ = σ M dE+ + − dµsalt (35)
π = γbase − γ (32)
−dγ = σ M dE− + + dµsalt (36)
where γbase is the interfacial tension of the
base solution. where E+ and E− are the potentials with
The analogue quantity at constant respect to a reference electrode reversible
for the cation and anion, respectively, −
charge is
= ξbase − ξ (33) and + are the corresponding relative
surface excesses (see Eq. 13) of the ions.
where The relative surface excesses can be
ξ = γ + σE (34) obtained from the following relationship:

In both cases, all other variables should dγ
− = ± (37)
be kept constant. However, in reality the dµsalt E±
interfacial concentration of other species
may be changed by the introduction of one A very important indicator of the occur-
of the components. rence of specific adsorption is the so-called
Various isotherms were proposed on the Esin–Markov coefficient, which in a gen-
basis of different physical models. Three eralized form can be given by the following
approaches should be distinguished [1]. equation:

1. The solution phase is considered as ∂E± 1 ∂E±
= (38)
a continuum and the behavior of the ∂µsalt σ M RT ∂ ln asalt σ M
adsorbed species is similar to that of
Originally the notion of Esin–Markov co-
the adsorbate at the vapor–condensed
efficient was introduced for the condition
phase interface.
of σ M = 0, that is, for the effect observed
2. The adsorption process is a replacement
at the potential of zero charge (Epzc ).
reaction in which the adsorbing species
In the absence of specific adsorption the
replaces another species, normally the
Esin–Markov coefficient is zero, that is,
solvent (see previous section).
no shift of the PZC should occur when the
3. The interphase is regarded as a two-
concentration (activity) of the electrolyte is
dimensional solution.
changed. The origin of the shift at σM = 0
Some of the isotherms described in the in the case of specific adsorption can be
literature are shown in Table 1. understood easily. Holding the potential
at Epzc , if the addition of the salt is fol-
3.2.3.3 Indication and Evaluation of lowed by the specific adsorption of the
Specific Adsorption anion or the cation, this process should re-
In order to show how the occurrence of sult in a change in the electric state of the
specific adsorption can be demonstrated interphase. The charge of the specifically
through thermodynamic relationships, we adsorbed species should be compensated
consider a simple case: solution of a salt by the charge of the counterions on the so-
MeX containing Me+ and X− ions in lution side of the double layer. This means
contact with Hg. that at the preselected potential value, σ M
does not remain zero following the adsorp- In the case of mercury it is accepted
tion. In order to attain again the σ M = 0, that specific adsorption of F− plays no role
the potential of the electrode should be under certain conditions.
shifted to a new value. In general, specific However, this conclusion for mercury
adsorption of anions is indicated by nega- electrodes cannot automatically be carried
tive shifts in potential at constant charge over to other metals shown, for instance,
density, while specific cationic adsorption in studies of Ag electrodes [9, 10].
results in the opposite shift. The ‘‘method of mixed electrolytes’’ first
The determination of the extent of presented in two fundamental papers by
specific adsorption, remaining at the Hurwitz [11] and Dutkiewicz and Par-
above-mentioned simplest case (MeX), sons [12] was an important contribution
follows the following pattern. to the methods of evaluation of specific ad-
In these simplest cases the evaluation sorption. In Ref. [11], Gibbs’s adsorption
of the specific adsorption is based on equation has been derived in a general
the assumption that one type of ion form for the case of mixed solutions of
in the electrolyte is not involved in the strong electrolytes at constant molal ionic
specific adsorption. Mostly anionic specific strength.
adsorption is calculated assuming that It was predicted that in the case of
the cation (often an alkali ion) is not the binary systems NaF-NaX or KF-
specifically adsorbed. This means that KX (X = Cl− , Br− , I− ), investigated at
the cation is only in the diffuse layer. constant molal ionic strength at a mercury
Determining − and from Eq. (37) electrode, the treatment would lead to
+
(by measuring the interfacial tension), it is the specifically adsorbed surface excess of
possible to calculate, from + the extent Cl− , Br− , I− , provided that the specific
d
of − , excess concentration of the anion adsorption of F− is considered negligible,
in the diffuse layer, by the relationships of which seems a reasonable assumption.
GC theory. Likewise, it was demonstrated that the
The amount of specifically adsorbed determination of Cs+ for Cs+ from
anion is then obtained by the difference mixtures of the type LiCl-CsCl is possible
if the specific adsorption of Li+ is assumed
sp d
− = − − − (39) to be weak.
In Ref. [12] the specific case of KI + KF
(In the case of dilute solutions − ≈ ).
− system was discussed in detail.
The reliability of such a determination Their derivation, which could be an illus-
depends on the validity of GC theory. In the trative example on how to treat a problem
case of strong specific anion adsorption, simultaneously in terms of thermodynam-
this method furnishes reliable results. The ics and model approach, is as follows.
errors in the determination of − d are
The electrocapillary equation for a pure
greatest when specific adsorption is weak mercury electrode in contact with an
and the solution is dilute. aqueous solution containing KI and KF
For the interpretation of inner-layer at constant temperature and pressure may
models in the absence of specific adsorp- be written
tion, it is very important to find cases and
systems where there is no specific anion −dγ = K + dµ̄K+ + I − dµ̄I− + F − dµ̄F−
adsorption at all. (40)
where γ is the interfacial tension, µ̄i the diffuse layer, then it follows that
electrochemical potential of species i, and
i is the surface excess of species i relative Id− x d−
= = I (45)
to the water. If we introduce the chemical F− 1−x F−
potentials of the salts and eliminate the
electrochemical potentials of the anions, because the concentration of ions of the
we may write Eq. (40) in the form same charge in the diffuse layer is directly
proportional to their bulk concentration.
−dγ = σ dE+ + I − dµKI + F − dµKF Thus it may be seen from Eq. (45) that
(41) the second term in the coefficient of
where σ is the charge per unit area on d ln x in Eq. (44) is simply the surface
the mercury surface, E+ is the potential concentration of iodide in the diffuse layer,
of the mercury electrode with respect to and Eq. (44) reduces to
a reference electrode reversible to the
−dγ = σ dE+ + Ii − RT d ln x (46)
potassium ion in the working solution,
and µj is the chemical potential of the Finally, since the concentration of K+ is
salt j . constant and the activity coefficients are
Making the very plausible assumption constant to a good approximation,
that in a mixed-salt solution of constant
total concentration, we may write dE+ = dEN.C.E. (47)
dµKI = RT d ln mKI = RT d ln x (42) where EN.C.E. is the potential of the mer-

cury electrode with respect to the normal
where mKI is the molal concentration of KI, calomel electrode. Thus the coefficients
x = mKI /(mKI + mKF ), and mKI + mKF is
∂γ
constant. Similarly, = −Ii − (48)
RT ∂ ln x EN.C.E.
dµKF = RT d ln mKF = RT d ln(1 − x)
∂ξN.C.E.
= −Ii − (49)
x RT ∂ ln x σ
= −RT d ln x (43)
1−x
give directly the surface excess of iodide
Substituting Eqs. (42) and (43) in Eq. (41), ion present in the inner layer, if
we obtain
ξN.C.E. = γ + σ EN.C.E. (50)
−dγ = σ dE+
3.2.4
x
+ I − − Specific Adsorption at Solid Electrodes
1−x

3.2.4.1 The End of the ‘‘Mercury Era’’
× F − RT d ln x (44) During the first half of the last century,
both the theoretical and experimental work
Now the total surface excess of iodide ion in connection with electrified interphases
can be divided into that present in the were mainly centered around the mer-
inner layer, Ii − , and that present in cury–water system.
the diffuse layer, Id− . If it is assumed that Considering the potentialities of the
the fluoride ion is present only in the available experimental methods, this was
the only feasible way to understand double- The direct in situ study of the specific
layer phenomena and to create a reliable adsorption of ions and organic species
theoretical basis for the interpretation of was dictated by the desire to obtain more
the phenomena observed. and more direct information on the pro-
In principle, in the case of nonspe- cesses occurring on the surface of the
cific adsorption the nature of the electrode electrode. This was one of the most impor-
metal should play a secondary role; there- tant factors leading to the elaboration of
fore, the fundamental relationships found nonelectrochemical methods for the study
for this type of adsorption in the case of of specific adsorption. For instance, radio-
mercury should be of general validity, in- tracer method, ellipsometry, and various
dependent of the nature of the metal. In spectroscopic methods (such as Raman
reality, the experimental results obtained spectroscopy) should be mentioned here.
with mercury, the methods for the evalu- The introduction of new methods re-
ation of these experimental data, and the sulted in a change in the philosophy on
theoretical efforts to interpret them served how to separate specific and nonspecific
as very important basis for the study of adsorption. According to the approach
any kind of electrode. Some of these, for presented in the previous section, the
instance, methods of Hurwitz and of Par- determination of specific adsorption re-
sons and coworkers and the Esin–Markov
quired the deduction of the adsorption in
effect, were mentioned in the previous
the diffuse double layer, calculated on the
section.
basis of model assumption, from the to-
It is evident that such methods as the
tal adsorption determined by well-defined
application of capillary electrometer or the
and reliable thermodynamic methods. In
dropping electrode cannot be used in the
the case of some new methods, the specific
case of solid electrodes. On the other hand,
adsorption of a given species is studied in
the measurement of surface tension and
the presence of a great excess of a sup-
surface stress of solid electrodes is not easy
and in addition, the interpretation of the porting electrolyte. If this excess is several
data is very problematic [13–24]. orders of magnitude higher than the con-
Beginning from the fifties of the last centration of the species studied, there
century, the investigation of the behav- could be no doubt that in the case of a mea-
ior of solid electrodes came into the surable adsorption of these species (more
foreground. The available methods, for in- than 10−11 to 10−12 mol cm−2 ) the value
stance, wetting (contact) angle method, the obtained should be ascribed to specific ad-
measurement of differential capacity, were sorption. This consequence can be clearly
far from furnishing reliable data in the case demonstrated in the case of radiotracer
of solid electrodes. adsorption studies.
The practical aspects of electrocataly- For instance, if the adsorption of 35 S-
sis (fuel cell–oriented research) required labeled H2 SO4 is studied in the presence of
data on the specific adsorption at elec- 1 mol dm−3 HClO4 and the concentration
trodes with high real-surface area, while of labeled sulfate is 10−4 mol dm−3 , the ra-
for theoretical considerations, results ob- tio of the adsorption of HSO4 − and ClO4 −
tained with well-defined electrode surfaces ions in the diffuse part of the double layer
(single-crystal surfaces) were of great im- should be 10−4 as well. This means that
portance. the extent of the nonspecific adsorption
of HSO4 − species under such conditions the choice of the components, the ther-
should be less than 10−13 mol cm−2 . modynamic charge Q (the total charge) of
Despite the development and application the electrode at a given potential should
of new mostly nonelectrochemical meth- be different. This total charge should
ods, ‘‘classical’’ electrochemical methods be distinguished from the so-called ‘‘free
remained in use for solid electrodes. charge’’ corresponding to the double-layer
Whatever the method used should be, model believed to exist on the solution side
the study of the specific adsorption in and (with opposite sign) on the metal side
the presence of a supporting electrolyte of the double layer formed at the interface
has the advantage that the results can be (see later).
obtained directly without any calculations The study of the UPD of metal ions
based on model assumptions concerning belongs to the group of adsorption studies
the structure of the double layer. in the presence of a great excess of
supporting electrolyte as in most cases
3.2.4.2 Underpotential Deposition: A studied the concentration of metal ions
Limiting Case of Specific Adsorption is chosen to be very low in comparison
In the literature the specific adsorption and with that of the ions of the supporting
the underpotential deposition (UPD) pro- electrolyte.
cesses are discussed separately. For a long The most important exception to this
period the adsorption of hydrogen on Pt, rule is the specific adsorption, the UPD, of
Rh, Pd electrodes at more positive potential H+ ions or hydrogen adsorption on some
values than the hydrogen equilibrium po- noble metal electrodes. These systems
tential was considered as some specific and were studied from several molar acid
interesting phenomenon characteristic for concentrations to very high pH values. For
some noble metals and hydrogen. detailed discussion of UPD see Sects. 3.2
However, beginning from the sixties it and 3.3.
became quite evident that the adsorption
of metal ions on foreign metal substrates
3.2.4.2.1 A Widely Used and Misinter-
is often accompanied by significant charge
preted Misconception: The Electrosorp-
transfer commensurable with the charge
tion Valency In the classical double-layer
involved in hydrogen adsorption at noble
models it is assumed that specifically ad-
metal electrodes [25–28]. These phenom-
sorbed ions retain their original charge
ena were termed as UPD characterized by
during the adsorption process.
the general equation
In the sixties, Lorenz and coworkers have
MSn+ + ne−
−−
−−−
−Mad (51) suggested that a purely electrostatic bond
between the ion and the metal cannot be
From thermodynamic point of view, it is considered as a physical reality and there
not possible to decide whether we have should be a degree of charge transfer
M n+ and ne− or Mad at the interface. If similar to the formation of a covalent
the equilibrium characterized by Eq. (51) bond [29–33]. The specific adsorption on
is considered as a reality, only two from the basis of this concept was formulated as
the three reaction partners can be consid- follows
ered as independent components of the
z(1−λ)
(thermodynamic) system. Depending on R z = Rads + λze (52)
where R z is the adsorbing species of charge equation:

ze and λ is a dimensionless coefficient,
denoted as the partial-charge coefficient. νM − OH2 + Szaq −
←−−
−−
→− M−S
z+λ
Thus in the generally accepted model, + λe− + νH2 O(aq) (54)

λ = 0. If λ = 1, then this would imply
an entirely covalent bond. where ν is the stoichiometric number,
The Lorenz concept of partial charge Sz is the adsorbing species, λ is the
transfer was subject to general criticism by partial charge coefficient, M stands for the
Parsons, Damaskin, and other authors and metallic atom, and Sz+λ is the specifically
these critical views are well summarized adsorbed substance.
in the IUPAC recommendations [1] in For characterizing this process they
connection with these matters: introduced the concept of ‘‘electrosorption
‘‘The fraction of charge shared by the ad- valency,’’ which is defined by the relation
sorbed particle may be characterized by a
partial charge number. This term, although 1 ∂σ M 1 ∂µs
γ =− =
conceptually correct when applied to the F ∂ φ F ∂φ
model, is not adequate to indicate the (55)
quantity accessible to experimental mea- where µs is the chemical potential of
surements on a thermodynamic basis. the adsorbing species in solution and
The accessible quantity is the formal φ is the potential difference across the
partial charge number defined by interface.
The term ‘‘electrosorption valency’’ was

− ∂σ M /∂B E,µ =µ chosen because of the analogy between the

lB = i B
value, γ as defined above, and the charge
F
on the ion, z. If we compare Eqs. (53)
(∂µB /∂E)B ,µi =µB
= (53) and (55), the only difference is that in-
F stead of E in the former φ can be
Physically, lB measures the average num- found in the latter. Although E is a po-
ber of unit charge supplied to the elec- tential referred to a reference electrode
trode from the external circuit when one while φ is considered as a potential dif-
molecule of species B is adsorbed at conference across the interface, the physical
stant potential.’’ content of the two equations is almost
Considering that in the overall process the same. Considering the conclusions
the adsorption of B could be accompa- drawn in connection with Eq. (53), the
nied by the simultaneous adsorption and electrosorption valency values determined
desorption of other species changed, for according to the preceding definition equa-
instance, oppositely as B, the interpretation could be very misleading if the
tion of the lB values could be very difficult possible simultaneous specific adsorption
(see Sect. 3.2.4.2.2) of counterions is not taken into considera-
A similar criticism refers to the concept tion. This question will be discussed in the
of electrosorption valency introduced by next section.
Vetter and Schultze starting from the
partial charge concept [34, 35]. 3.2.4.2.2 Interrelation of Anion Specific
They described the adsorption equilib- Adsorption and UPD of Metal Ions The
rium of a charged species by the following simultaneous occurrence of the specific
adsorption of anions and the formation a great excess of supporting electrolyte

of adatoms on a surface could result in containing the metal ions forming the
significant difficulties in the calculation of adatoms present also in a low concentra-
the charge and mass balance required for tion [36–42].
the reliable interpretation of the overall In the last three to four years the
adsorption phenomena. A number of study of the coadsorption of anions and
situations should be distinguished in such other solution components with adatoms
cases; however, we consider two limiting plays a central role in order to clarify
cases. the structure of adlayers formed on
single-crystal surfaces. In most of these
1. The UPD takes place at a surface (in a
cases the layer of Cu adatom formed on
potential range) where the role of the
Pt(111) [43–51] and Au(111) [50, 52–61]
specific adsorption of anions can be
surfaces was studied. Several studies
neglected. In such cases the mass and
were devoted to UPD of Ag on Au(111)
charge balance can be relatively easily
surfaces [62–66].
treated.
2. The UPD takes place on a surface (in
a potential range) where the specific 3.2.4.3 The Generalization of the
Thermodynamics of Monolayer Adsorption
adsorption of the anions present in
in Electrochemical Systems
the system is significant. This group
The thermodynamics of adsorbed mono-
of phenomena can be divided into two
layers was first discussed in detail by
subgroups.
Frumkin and coworkers for the particular
– Adatoms displace adsorbed anions, case of hydrogen adsorbed on a platinum
that is, as the coverage with respect metal [67]. Starting from these studies, a
to adatom increases the adsorbed generalized derivation was attempted by
anions decrease. Parsons [68] through an illustrative exam-
– Specific adsorption of the anions ple considering a particular system: a gold
takes place as a result of interaction electrode in contact with an aqueous so-
with adatoms. The extent of anion lution containing a low concentration of
adsorption is proportional to the cov- thallium perchlorate (1 mM) and a much
erage with respect to the adatoms. higher concentration of potassium per-
Even if the adatom displaces the an- chlorate (0.1 M). The main stages of this
ions adsorbed on the support metal, derivation will be presented here.
this effect could be compensated or The interphase was described at constant
overcompensated by the extent of ad- temperature and pressure by the Gibbs
sorption induced by the adatoms. In equation in the form
the latter case an increase in the
−dγ = TA dµTA + KA dµKA + T dµT
overall anion adsorption can be ob-
served and this increase can be called (56)
induced anion adsorption. where i is the surface excess and µi
is the chemical potential of species i. T
Radiotracer adsorption studies furnish represents thallium, A the anion, perchlo-
rapid and reliable information on these rate, and K the cation, potassium. The
phenomena, using labeled anions in three components are taken as neutral
low concentration in the presence of species so that the electroneutrality of
the interphase is maintained. The chem- 2. Reference electrode reversible with re-
ical potential of the gold is, of course, spect to K ions (EK )
invariant at constant temperature and
pressure. ∂γ
QK = z+ ν+ F KA =
The chemical potential of the monolayer ∂EK µTA ,µT
T may be determined from the potential (62)
(ET ) of the working electrode with respect 3. Reference electrode reversible with re-
spect to T ions (ET )
to an electrode of pure T, both in
equilibrium with T ions in the solution,
∂γ
that is, the electromotive force (emf) of the QT = zT F T = (63)
∂ET µKA ,µTA
cell T|T+ |T|Au is considered
z− , z+ and zT are the charge of ions, ν− is
(µT (θ) − µoT ) the number of anions in salt TA, and ν+ is
ET = − (57)
zT F the number of cations in salt KA. QA , QK ,
and QT are the so-called thermodynamic
where µT (θ) is the chemical potential of T or total charges of the electrode. This
at coverage θ on the gold surface and µoT is thermodynamic charge in general differs
the standard chemical potential of pure T. from the ‘‘free charge’’ believed to exist
It follows directly from Eq. (56) that on the two sides of the double layer.
three surface excesses are accessible to They can be considered equal to each
experiment: other in the case of ideally polarizable
electrodes without any charge transfer. In
such situation the charge of the electrode
∂γ
TA = − (58) has a single value.
∂µTA µKA ,µT
However, in the case of adsorption

∂γ processes connected with charge transfer
KA = − (59) (for instance, UPD), the thermodynamic
∂µKA µTA ,µT
charge of the electrode has more than one
∂γ value as owing to the equilibrium pro-
T = − (60)
∂µT µTA ,µKA cesses (with charge transfer), the number
of the degrees of freedom for the choice
From these equations, three Lippmann of thermodynamically independent com-
equations can be derived depending on ponents decreases, that is, it will be less
the choice of reference electrode reversible than the number of chemical components
with respect to one of the ions present in distinguishable in the system.
the system. In the case discussed above we have the
following choices.
1. The potential of the working electrode is The system is built up from
measured against a reference electrode
reversible with respect to A ions (EA ) 1. KA and T
2. TA and T
3. KA and TA
∂γ
QA = z− ν− F TA =
∂EA µKA ,µT The QA , QK , and QT charge values cor-
(61) respond to the charge to be introduced
electrically together with the correspond- at the conclusion that the results obtained
ing material to create 1 cm2 of interphase confirm that the charge associated with
of the electrode keeping constant all the the unusual states is at least formally
parameters characterizing the system. compatible with the exchange of one
electron per three surface platinum atoms.
3.2.4.4 Anion Adsorption on Well-defined In accordance with this view, the oc-
Crystal Surfaces currence of the unusual peaks on the
Fundamental questions connected with cyclic voltammograms observed in acetic
the characterization of the double layer acid–acetate system was ascribed to a
at well-defined single-crystal faces of solid one-electron-transfer process of acetic
metals have been reviewed and surveyed acid/acetate ions according to the equa-
several times by Hamelin [6, 69–72]. tions [83]
Since the early 1980s, cyclic voltammetry
−− + −
has been used to characterize single-crystal CH3 COOH−−−
−CH3 COOads + H + e
electrodes in terms of surface order, pres-
at pH < 2
ence or absence of defects contaminations,
and so on. In some cases the voltammo- CH3 COO −−−CH3 COOads + e−
− −−−
grams have also been used for identifica-
tion of adsorbates, mostly on the basis of at pH > 5
electric charge calculation. However, since
the double-layer charging and surface re- There are, however, many evidences
actions such as UPD and anion adsorption demonstrating that the existence of the
may occur in parallel, it is difficult to break unusual states cannot be easily explained
down the total voltammetric charge into all with the charge transfer to or from the
of its individual components. Therefore, anions (or other species). For instance, re-
interpretation from voltammetry alone, cent spectroscopic studies do not support
used as a tool for the adsorbed species the occurrence of electron transfer in the
identification, may be ambiguous. This case of anion adsorption [77, 79, 84–88].
has been, for instance, the case with the A very important and confusing ob-
interpretation of the ‘‘unusual adsorption servation is that the charge involved in
states’’ on the Pt(111) electrode [73–82]. the unusual states in various solutions
Although at a very early stage of the is nearly independent of the nature of
studies the appearance of the unusual the anion, including organic anions. This
peak on the voltammograms of Pt(111) was aspect of the unusual states cannot be
explained by anion adsorption for a long easily reconciled with the charge trans-
period, the view that it should be ascribed fer to or from the anion as a mechanistic
to hydrogen adsorption (or incorporation origin of the pseudocapacitive current gen-
of hydrogen) was widely accepted by erated by the reversibly adsorbed anions on
many researchers. Recently the hydrogen Pt(111).
adsorption model is rejected on the Results obtained from careful exam-
basis of reexamination of the problem ination of the pH effects on the po-
by the most eminent proponents of the tentiodynamic behavior of the Pt(111)
model studying the unusual desorption electrode in acidified perchlorate solu-
states on Pt(111) electrodes by CO and I tions [82] advocate in the favor of the
displacement method. Firstly, they arrived assumption that water and not ClO4 −
anions are involved in the unusual ad- (ClO4 − ) is kept constant when the concen-
sorption states in the double-layer region tration of the adsorbing anion is changed.
of Pt(111). Presumably H2 O dissociation It is considered that the effect of the ad-
and OH deprotonation are responsible for sorption of the anion of the supporting
the two main electrode processes in the electrolyte on the measured surface excess
double-layer region of Pt(111) in NaClO4 of the investigated anion can be mini-
solutions. mized under this condition. In addition,
For instance, in situ Fourier transform it is assumed that in the presence of an
infrared (FTIR) spectroscopy has been excess of supporting electrolyte the mea-
used by Faguy and coworkers [79] to study sured surface excess is essentially equal to
the potential dependent changes in anion the surface concentration.
structure and composition at the surface In the case of sulfate adsorption the sur-
of Pt(111) electrodes in HSO4 − -containing face excesses determined from the charge
solutions. From the infrared (IR) differen- density data were compared with the Gibbs
tial normalized relative reflectance data, excesses determined from radiochemical
the maximum rate of intensity changes measurements using 35 S-labeled sulfate
for three IR bands can be obtained. Two solutions. Very good agreement of the data
modes associated with the adsorbed an- was reported [89].
ion and one mode assigned to species is In order to obtain insight into the nature
not adequately described as either sulfate of the adsorbed species, Esin–Markov co-
or bisulfate; the data are more consis- efficients for SO4 2− adsorption from two
tent with an adsorbed H3 O+ −SO4 2− ion series of solutions were determined: (1) at
pair, possibly with the three unproto- constant pH and variable K2 SO4 concen-
nated sulfate oxygens interacting with Pt tration; (2) at constant K2 SO4 concentra-
sites. tion and variable pH. The Esin–Markov
In a series of papers by Lipkowski and coefficients were used to identify the na-
coworkers, specific adsorption SO4 2− , Cl− , ture of the adsorbed species (SO4 2− or
and Br− on Au(111) electrode from 0.1 M HSO4 − ). The authors arrived at the con-
HClO4 + x M K2 SO4 and 0.1 M HClO4 + clusion that SO4 2− ion is the adsorbed
x M KCl or KBr or KI solution was studied species even if HSO4 − predominates in
using chronocoulometry and evaluating the bulk of the solution. Similar results
the data on the basis of thermodynamic were reported in Ref. [96].
of ideally polarizable electrode (x varies This conclusion, however, seems to be
between 5 × 10−6 and 5 × 10−3 M) [53, 56, questionable, in the light of thermody-
89–94]. A further study was devoted to namic considerations and other experi-
OH− adsorption [95]. mental results (see later).
We should make a distinction between Results obtained for Br− , Cl− , and
these cases and the case of adsorption −
I adsorption were compared. The main
from a mixed electrolyte at constant ionic conclusions are as follows: All the three
strength discussed in Sect. 3.2.3.3. In this halides form a chemisorption bond with
latter case, the investigated anion replaces the gold surface. The bonds are quite
progressively the anion of the supporting polar at the negatively charged surface;
electrolyte, while in the method by Lip- however, its polarity drops significantly
kowski and coworkers the concentration at the positively charged surface. At low
of the anion of the supporting electrolyte charge densities and coverages, the bond
polarity is determined by the ability of free and neutral media. The adsorption of sul-
electrons to screen the dipole formed by fate in alkaline solutions is inhibited by
the adsorbed anion and its image charge the more strongly specifically adsorbed
in the metal. At high charge densities and OH− ions. The potential dependence of
coverages, the chemisorption bond has surface coverage, estimated on the basis
a predominantly covalent character. The of spectroelectrochemical data, indicated
strength of the halide adsorption and the that a considerable repulsion exists among
covalent character of the chemisorption the adsorbed sulfate species. The spectral
bond increase progressively by moving data were compared to those obtained for
from chloride to iodide. sulfate adsorption on Pt(111) and on un-
An interesting comparative study of the derpotentially deposited Ag on the Pt(111)
adsorption of bisulfate anion on different surface.
metals was carried out by Shingaya and Voltammetric study of the adsorption
Ito [88]. Adsorption of sulfuric acid species of 14 anions on Ag(111) was reported in
on various metal(111) electrodes in a 0.5- Ref. [98].
M H2 SO4 solution was investigated by IR The specific anion adsorption of chloride
reflection absorption spectroscopy (IRAS). and sulfate on Cu(111) from acidic aque-
Pt, Au, Ag, Cu, and Rh(111) were used as ous electrolytes has been studied by Broek-
electrodes. According to this study, HSO4 − mann and coworkers [99] using scanning
is adsorbed on all of the surfaces including tunneling microscopy (STM) and cyclic
Pt, Rh, Au, Ag, and Cu in a 0.5-M H2 SO4 voltammetry. It was found that at posi-
solution. It was found that considerable tive potentials √
adsorbed
√ chloride forms a
amounts of sulfate in addition to bisulfate well-ordered ( 3 × 3)R30◦ superstruc-
coexist on Ag(111) in a 0.5-M H2 SO4 ture. At negative potentials the bare copper
solution. An interesting finding that both surface can be observed. In the case of
DSO4 − and SO4 2− are coadsorbed on sulfate adsorption at positive potentials,
Au(111) in 0.5-M D2 SO4 + D2 O solution. high-resolution STM images revealed an
The interconversion of HSO4 − /H3 O+ and additional species that was assigned to
H2 SO4 that was seen on Pt(111) was not coadsorbed water molecules. Close-packed
observed on the other electrodes. The sulfate rows are separated by zigzag chains
frequencies of the SO3 symmetric and of these water molecules.
S−OH stretching bands of a bisulfate Adsorption of sulfate ions on the three
adsorbed on the electrode surfaces are basal planes of monocrystalline copper and
related to the electronic state of each silver electrodes in 0.1 M HClO4 was stud-
M(111) surface. ied by radiometric and electrochemical
A somewhat different conclusion was re- methods [100, 101].
ported for Ag(111) by Adzic and coworkers The adsorption of OH− ions on
in a study [97] in which cyclic voltamme- the Pt(111) plane has been studied
try and in situ IR spectroelectrochemistry by Drazic and coworkers [102] using
were employed to investigate the ionic ad- fast cyclic voltammetry in sodium hy-
sorption on the Ag(111) electrode surface droxide solutions (0.03 to 1 M), under
in sulfate-containing solutions of differ- quasi-equilibrium and Tafel approxima-
ent pH. It has been concluded that the tion conditions. It was shown that the
sulfate anion is the predominant species OH− ion adsorption is an electrosorp-
adsorbed at the electrode surface in acidic tion process with one electron exchanged
between an OH− ion and the platinum electrode potential and constant charge
surface. The electrosorption process fol- coincide.
lows the Frumkin adsorption isotherm
with low-intensity repulsive interactions of 3.2.5
the adsorbed species (f = 2–3). The esti- Methods of Investigation and Illustrative
mated values of the standard electrochemi- Examples
cal rate constant (k o = 5.6 × 10−4 cm s−1 )
and the standard exchange current den- 3.2.5.1 General Remarks
sity (j00 = 5.45 × 10−2 A cm−2 ) indicate a During the last two to three decades, so-
rather fast electrochemical process. phisticated methods of solid-state surface
The thermodynamic and inner-layer science were introduced to study the ad-
characteristics of halide ions adsorbed on sorption phenomena at electrochemical
bismuth single-crystal planes from so- interfaces. These new techniques of inves-
lutions of methanol and propanol was tigation enable us to interpret adsorption
studied by an Estonian group in a se- phenomena at molecular level.
ries of publications [103–108]. The ad- The main characteristic of most of the
sorption of Cl− , Br− , and I− ions from new experimental methods is the simul-
methanolic solutions on different faces taneous use (coupling) of electrochemical
of a bismuth single-crystal electrode was techniques with other nonelectrochemical
studied by impedance and chronocoulo- methods. Only some of these methods
metric techniques. The charge of specif- can be very briefly mentioned in the
ically adsorbed anions σ1 was calculated following sections. More detailed descrip-
using the Hurwitz–Parsons–Dutkiewicz tion of the methods and their application
method. Fitting the results to the simple can be found in monographs and text-
virial isotherm, the values of the stan- books [109–111].
dard Gibbs energy of adsorption −GoA
and the mutual repulsion coefficient B 3.2.5.2 In Situ Techniques
of the adsorbed anions were estimated.
The dependence of the integral capaci- 3.2.5.2.1 Optical Techniques
tance of the inner layer on the electrode
charge σ was calculated according to UV-visible Techniques
the traditional inner-layer model using
the electrode charge as the independent Ellipsometry [112, 113] Ellipsometry is one
electrical variable. It was found that the of the earliest optical technique to be
adsorption of halide ions decreases in the applied to the study of adsorption pro-
sequence of the Bi single-crystal planes cesses [112, 113]. It involves the analysis
(011) > (001) > (111). of the phase change and the change in
As a continuation of this work, in accor- amplitude ratio of polarized light reflected
dance with the method suggested by Lip- from a surface.
kowski and coworkers, instead of electrode A significant limitation of the method
charge the electrode potential was consid- is that a model for the adsorbed layer
ered as independent electrical variable for should be made for the evaluation of
the evaluation of the experimental data. the experimental data. As the theory of
It was found that under comparable ellipsometry is based on the assumption
conditions the results obtained at constant that the reflection occurs from an ideally
smooth surface, the interpretation of data One of the most commonly applied IR
obtained with surfaces with nonideal techniques developed to overcome these
topography encounters difficulties. problems is the external reflectance tech-
nique. In this method, the strong sol-
Specular reflectance spectroscopy [114] The vent absorption is minimized by simply
principle of the method is the recording of pressing a reflective working electrode
a reflection of the incident monochromatic against the IR transparent window of
beam by the surface of the electrode as a the electrochemical cell. The sensitivity
function of wavelength, potential, or time. problem, that is, the enhancement of
Similar to ellipsometry, a model of the the signal/noise ratio in the case of ex-
interface is required to evaluate the external reflectance techniques is solved
perimental data. Several techniques were by various approaches. These are, for
elaborated to enhance the sensitivity of instance, electrochemically modulated in-
reflectance measurements (for instance, frared spectroscopy (EMIRS), in situ FTIR
multiple reflectance). (which use potential modulation), and
In order to avoid the problems connected polarization modulation infrared reflec-
with measurement of absolute reflectivi- tion absorption spectroscopy (PM-IRAS,
ties, changes in reflectivity are detected as FTIR) [86, 117–123].
the potential of the electrode is modulated A new development is the elabora-
in the case of the so-called ‘‘modulated tion of synchrotron far infrared spec-
specular reflectance spectroscopy.’’ troscopy (SFIRS) for adsorption stud-
ies [124].
Vibrational Spectroscopy (Infrared and Ra-
man Scattering Spectroscopies) Although Raman Spectroscopy The major problems
UV-visible techniques were proven useful, in the case of the application of IR
they are not suitable for identifying ad- spectroscopy, the strong absorption by
sorbates or orientation of adsorbates. This the solvent and window material, can
type of information could be provided by be avoided using Raman scattering tech-
the various versions of vibrational spec- niques, since only visible radiation is
troscopy. involved in this case.
During the last decade, vibrational spec- Raman scattering cross sections of typ-
troscopy has contributed to a great extent ical adsorbates are low, therefore two
to the advance in electrochemical surface fundamental techniques were elaborated
science to enhance them: resonance Raman scat-
tering (RRS) and surface-enhanced Raman
Infrared (IR) spectroscopy [115, 116] For in scattering (SERS) [125–127].
situ application of IR spectroscopy for
the study of adsorption, there were two Second harmonic generation anisotropy from
fundamental problems to overcome: single-crystalline electrode surfaces [128] In-
terfacial second harmonic generation
1. Most solvents (especially water) absorb (SHG) is an optical spectroscopy that
IR very strongly. is inherently surface-sensitive. All effects
2. How to separate the signals coming that modify the electron density profile
from the adsorbed layer from the of the interface affect the SHG pro-
noises. cess. Therefore, SHG depends on the
exciting and SH frequency, on surface X-ray Methods

structure, and on adsorbates and adlayers.
For metal electrodes there is a remarkable Surface X-ray scattering (SXS) The first
dependence of SHG on potential, adsorp- SXS study of an underpotentially deposited
tion, and charge density at the metal, metal monolayer was reported more than
and on surface structure. Using the ad- ten years ago. In a recent review [132]
vanced form of the SHG anisotropy, the it is demonstrated that this method
interference SHG anisotropy, one can de- is well suited for the study of the
termine the anisotropy of the SH field structure of metals, halides, and metal-
itself and can pursue how the various halide adlayers on single-crystal electrodes.
As another example, the study of the
isotropic and anisotropic sources con-
distribution of water at Ag(111) surface
tributing to this field evolve with potential
can be mentioned [133].
and adsorption. This yields information on
In situ X-ray absorption fine structure
the geometric and electronic structure of
method (XAFS) [134] gives insight into
the interface and adlayer to unparalleled
the interaction between adsorbate and
details.
substrate.
There are mutual links between SHG at
In situ X-ray surface diffraction (SRSD)
electrodes and electrochemistry, and the
method was applied for the study of UPD
results of classical electrochemical mea-
of copper on Au [135].
surements, simply because these methods
probe different physical properties of the
interfacial regime. Adsorption and poten- 3.2.5.2.2 Radiotracer Methods Of all the
tial, charge density, surface structure, and in situ techniques used to obtain informa-
the formation of (ordered) adlayers have tion about adsorption phenomena at elec-
a significant impact on both the behavior trode surfaces, the radiotracer technique
of the electric double layer and the SH re- was the first direct method furnishing re-
sponse. SHG was applied to study sulfate liable data on the extent of adsorption of
adsorption at Au(111) [129]. a specific preselected species. Generally,
in situ radiotracer methods provide direct
information on adsorption|accumulation
Sum and difference frequency generation at processes on a great a variety of electrodes
electrode surfaces [130] The investigation under various experimental conditions.
of vibrational and electronic properties of The in situ experimental techniques
the electrochemical interface by using non- used by various authors are described in
linear optical techniques of visible-infrared the relevant literature (papers and reviews)
sum (SFG) and difference (DFG) fre- in detail [100, 136–155].
quency generation constitutes an interest- The main problem of the in situ radio-
ing aspect of the study of adsorbed species. tracer study of electrosorption phenomena
The vibrational behavior of H-Pt(hkl) sys- originates from the very nature of the
tem, both in the underpotential and over- method, because the radiation measured
potential regime, adsorption of CN− on consists of two main parts. The first one
Au(hkl) electrode (using DFG), electrode is that coming from the solution phase or
surface electronic properties were stud- from the solution layer contacting the elec-
ied and measurements of the PZC were trode. The second radiation component is
carried out [131]. the radiation coming from the adsorption
layer of interest for the measurement of 36 Cl),the self-absorption of the radiation

adsorption. in the solution phase is so high that the
The elaboration of a technique for the thickness of the solution layer effective
radiotracer adsorption measurements is in the measured solution background
equal to the search of the optimal con- radiation is very low. (Similar phenomena
ditions for the minimalization of the role could be observed for isotopes emitting
of the radiation coming from the solution low-energy X ray.)
background. The usual classification of the In cases in which the radiation coming
various methods follows this pattern, that from the solution background would be too
is, how this requirement is fulfilled. Ac- high in comparison with that originating
cording to this principle, the methods can from the adsorbed layer (smooth surfaces,
be divided in two main groups: the radi- γ -radiation), some kind of mechanical
ation of solution background is governed
means should be applied to reduce the
and minimized by self-absorption of the
role of the background radiation.
radiation (thin-foil method) and the back-
The mechanical control of the back-
ground radiation intensity is minimized by
ground radiation can be achieved by two
mechanical means (thin-gap method). The
different ways: fixed rigid reduction of so-
most important and widely used represen-
lution layer thickness (classical thin-layer
tative of the first approach is the so-called
or gap method) and flexible mechanical
foil method.
or temporary reduction of the thickness
In the case of the thin-foil method, the
detector ‘‘sees’’ simultaneously both com- of the solution layer (electrode-lowering
ponents of the radiation coming through a technique). In the case of the first version,
thin foil (metal or metal(gold)-plated plas- specially designed cells are used where
tic film) forming the bottom of the cell. The only a very thin-solution layer flows be-
adsorbent is either the foil itself or a thin tween the electrode and the detector as
deposited layer on the bottom of the cell shown by Scheme 2.
serving as a mechanical support and elec- This method is not used very often
tric conductor (if the foil is metal-plated). because of technical problems. These prob-
The situation can be visualized by lems are avoided by the application of a
Scheme 1. flexible version, the so-called electrode-
The radiation measured is IT = Is + Ia lowering technique. In the case of this
(where Is and Ia are the intensities of the technique, the solution gap between the
radiation coming from the solution phase electrode and detector is minimized tem-
and from the adsorbed layer, respectively). porarily only for the time of the mea-
Using isotopes emitting soft β-radiation surement of the adsorption. The electrode
(radiations characterized by high mass- is placed in two positions as shown by
absorption coefficient (µ) (14 C, 35 S, and Scheme 3.
Solution
Adsorbed layer
Is Ia
D
Scheme 1
Detector
Solution flow
Electrode
Scheme 2
Electrode Electrode
Bottom of
the cell
Solution Gap
D D
Electrode in lifted up Electrode pressed down to

position the bottom of the cell
Scheme 3
In the ‘‘lifted-up’’ position, the attain- It is important to emphasize that al-

ment of the adsorption and the elec- though from the data obtained the ra-
trochemical equilibrium (or steady state) tio between the electric charge and the
proceeds without any disturbance, the corresponding mass change can be de-
detector measures only the solution back- fined for every experimental point as a
ground radiation. In the ‘‘pressed-down’’ function of the potential applied, one
position, the intensity measured comes should be very cautious, at least in the
from the species adsorbed at the electrode case of adsorption, to draw conclusions,
surface and from the solution layer present particularly the application of the no-
in the gap between the bottom of the cell tion of electrosorption valency should be
and the electrode. The thickness of this avoided.
gap depends on the mechanical state of The mass change could be the result of
the electrode surface, the stability of the several simultaneous adsorption and des-
bottom of the cell, and so on. orption processes involving the molecules
of the solvent as well. Sometimes the
3.2.5.2.3 Electrochemical Quartz Crystal frequency shifts observed could not be
Microbalance (EQCM) and Nanobalance ascribed to specific adsorption.
(EQCN) Techniques The first application Tsionsky and coworkers [158] studied
of quartz crystal microbalance (QCM) in the double-layer region of gold electrode
electrochemistry dates back to 1981 [156]. in 0.1 M HClO4 , KNO3 , and KOH using
A review on the application of these EQCN method. The authors observed
techniques can be found in Ref. [157]. frequency shifts even though the ions
One of the main advantages of the of these electrolytes are not specifically
EQCM and EQCN techniques in the study adsorbed and there is no faradaic reaction
of adsorption phenomena is the possi- that could lead to the formation of
bility of obtaining simultaneous voltam- adsorbed species through charge transfer.
metric and gravimetric response through Tsionsky and coworkers [158] developed a
frequency changes. model in which they assumed that the
liquid in a thin layer near the surface Au(111) was investigated and interpreted
has a higher viscosity because of the in Ref. [164]. Other interesting results
high concentration of ions and the high obtained in various systems have been
electrical field in this region. reported in Refs. [48–50, 165–175].
Some of the illustrative examples for the Atomic force microscopy (AFM) similar
application of EQCN technique are col- to STM is a scanning atomic-scale probe
lected in [157]. The electrical double-layer microscopy; however, AFM does not rely
region of optically polished gold in 0.1-M on tunneling. It detects the interatomic
NaF, KF, RbF, and CsF solutions in the forces between the tip and surface atoms.
absence and with addition of specifically This method provides a very powerful
adsorbing anions SO4 2− , OH− , Cl− , and complement to STM. It can be used in
Br− was studied, arriving at the conclu- cases when faradaic current is flowing,
sion that EQCN can be used to investigate that is, it offers the possibility of studying
the electrical double-layer phenomena on surfaces in the course of electrochemical
metal electrodes. processes. Nonconducting samples can be
A combined interferometric and EQCN imaged as well.
technique was elaborated for the study
of the energetics of adsorption phenom- 3.2.5.3 Ex Situ Methods
ena [159]. In the application of in situ methods, the
EQCN technique is very suitable for presence of the bulk electrolyte consti-
the study of the deposition and disso- tutes a significant limitation. As it was
lution processes of adatoms on various shown in the previous sections, by the
electrodes [55, 159–163]. use of thin-layer cells or carrying out
the measurements in the presence of a
3.2.5.2.4 Scanning Tunneling Microscopy thin-solution layer, the role of the bulk
(STM) and Atomic Force Microscopy (AFM) electrolyte phase can be reduced.
STM was introduced into electrochemistry An alternative route suggested in the lit-
during the late 1980s. The method is erature [114, 176–178] to overcome these
based on the phenomena of quantum- limitations is the emersion of the elec-
mechanical tunneling. It allows atomic trode from the electrochemical cell and
imaging of the structural properties of both transferring it into an ultrahigh vacuum
bare and adsorbate-covered surfaces. The (UHV) chamber, where studies using
principle of the method is as follows: surface-sensitive methods can be car-
A sharp tip of a chemically inert metal ried out.
is brought so close to the electrode However, this procedure involves the
surface that tunneling can occur and a loss of the electrochemical environment
tunneling current flows maintaining the and potential control. Auger electron
potential difference between the tip and spectroscopic (AES) studies [179] show
the electrode. Scanning the tip across that despite the evacuation required by
the surface of the sample at constant AES, chemisorbed species remain at
current or height from the surface, a the surface and their identification and
topographical image of the surface with quantitization can be carried out in
atomic resolution can be obtained. vacuum.
The coadsorption structure of Cu According to Hubbard [179], cationic
and halides on Pt(111), Pt(100), and species present in the emersed layer
of liquid or in the electrical double solution. This common potential scale rep-
layer (EDL) can also be determined us- resents an important link between UHV
ing AES. However, anions X− , present studies and electrochemistry. The poten-
in the emersed EDL and solution, typ- tial range accessible by the investigation
ically undergo hydrolysis to volatile HX of substrates with different work functions
species, which are subsequently lost to the can be expanded by the coadsorption of
vacuum. electronegative and electropositive species,
In contrast to the latter statement, Kolb such as halogens or alkali metals, which
and coworkers reported somewhat differ- induce large work function changes. The
ent results claiming that under certain adsorption of anions or cations from
conditions the electrode can be removed electrolytes, achieved in UHV at low tem-
from the electrolyte into an inert atmo- perature, is of particular interest. Solvation
sphere or even into vacuum with the processes and surface reactions, which
double layer apparently intact and the bulk result from the electrostatic interaction
electrolyte completely absent [114]. They between the partially ionically adsorbed
measured the work function of a gold species and coadsorbed H2 O, the most
electrode after emersion as a function of frequently used solvent, can be studied
potential in order to prove that the poten- by applying surface-sensitive techniques.
tial drop across the double layer remains A variety of methods have been developed
unchanged. allowing the investigation of the surface
The chemical composition of the structure and chemical composition.
emersed double layer was tested by ESCA. The structure of sulfuric acid species
It was found for nonspecific adsorbing adsorbed on a Pt(111) electrode has
ions that the ex situ obtained data are been successfully determined by using
in very good agreement with predictions the UHV modeling [186]. The micro-
from classical double-layer theory. scopic information about the structure
In the case of immersion technique of an adsorbed electrolyte anion and wa-
the opposite way (from UHV chamber ter as a solvent is indispensable to a
to the solution phase) is followed for true molecular level understanding of
the determination of the PZC of Au(111), the electrochemical double layers and
Pt(111) electrodes [180]. chemical reactions at the electrode sur-
faces.
3.2.5.3.1 Electrochemical Double-layer
Modeling Under Ultrahigh Vacuum 3.2.5.4 Illustrative Examples
Conditions Simulations of the elec- 1. The influence of the surface structure
trochemical double layer under UHV on the anion adsorption. Potential
conditions allow the characterization dependence of the adsorption of sulfate
of the chemical interaction between ions at various crystal faces of copper as
electrode surface and electrolyte con- studied by radiotracer technique [101]
stituents [181–186]. The electron work is shown in Fig. 1.
function measured for an adsorbate- 2. Effect of the nature of the an-
covered metal in UHV is considered to be ion on the potential dependence of
equivalent to the electrochemical potential the adsorption. Data obtained by the
for the same adsorbate-–electrode pair in method of Lipkowski and coworkers at
Au(111) electrode are summarized in 5. Anion specific adsorption induced by

Fig. 2 [93]. adatoms. Figure 6 demonstrates the
3. The unusual peak on Pt(111). Figure 3 increase in the sulfate adsorption (deter-
shows the cyclic voltammograms of mined by radiotracer technique) follow-
Pt(111) at various pH values [82]; A-1 ing the formation of Cd, Cu, Hg adlayer
and C-1 are the unusual peaks. Figure 4 on a polycrystalline Au support [147].
shows the shift of the unusual peak 6. The application of EQCM technique
following the addition of acetic acid to coupled with voltammetry. Figure 7
a HClO4 solution [83]. shows the simultaneous recording of
4. The formation of a layer of adatoms. the cyclic voltammogram and the fre-
Figure 5 shows the voltammetric curve quency change of the Au(111) EQCM
for the UPD of Ag+ ions on Au(100) electrode in the course of UPD of
and the (E) relationship [187]. Cu [55].
Cu(111) Cu(110)
4
1
10−14 Γ/ ions cm−2
Cu(100) Cu(poly)
4
0
−0.7 −0.6 −0.5 −0.4 −0.3 −0.2 −0.1 0 −0.7 −0.6 −0.5 −0.4 −0.3 −0.2 −0.1 0
E
[V]
Fig. 1 Potential dependence of sulfate adsorption on Cu electrodes from 5 × 10−4 M H2 SO4
in 0.1 M HClO4 ; (•), positive scan; (), negative scan. Potentials were measured versus
Ag|AgCl|1 M Cl− . (Reprinted from Ref. [101] with permission from Elsevier Science.)
80
I−
10−13 Γ/ions cm−2
60
Br−
Cl−
40
SO42−
20
0
−800 −400 0 400 800
E vs SCE
[mV]
Fig. 2 Comparison of SO4 2− , Cl− , Br− , and I− adsorption at the
Au(111) electrode surface from 0.1 M HClO4 + 10−3 M K2 SO4 and
0.1 M HClO4 + 10−3 M KCl, KBr, and KI solutions. (Reprinted from
Ref. [93] with permission from Elsevier Science.)
Pt(111)/0.1M NaClO4 pH = 1.14

pH = 1.47
pH = 2.27
80 pH = 3.12
A-2
pH = 3.88
pH = 4.07
60
A-1
40
20
[µA cm−2]
i
−20
C-2
−40
−60 C-1
−0.6 −0.4 −0.2 0.0 0.2 0.4 0.6 0.8 1.0

E vs SCE
[V]
Fig. 3 Cyclic voltammograms (ν = 40 mV s−1 ) of Pt(111) electrode in 0.1 M
NaClO4 solutions of lower pH values. (Reprinted from [82] with permission
from Elsevier Science.)
3.2.6 Information on the structure of wa-

Trends and Perspectives ter next to metal surfaces is obtained
from molecular dynamics computer sim-
The classical models of the double layer ulations [191]. Such simulations, although
were based on the primitive model in able to mimic the reality in more detail
which the solvent is regarded as a dielectric compared with simple models used in an-
continuum and the solid phase was alytical theories, are still limited to the
regarded as a charged hard wall. description of models. From the simu-
Recently, progress has been made to- lations it follows that water next to an
ward replacing this primitive model by uncharged metallic surface is perturbed to
a microscopic or molecular model of the a distance of ∼1 nm. Next to the charged
solvent and an electronic model of the surface, water is reorienting and, when
metal [188–192]. The assumption that the the external field is strong, undergoes a
solvent molecules are dipolar hard spheres layering transition.
and that the metal can be modeled as jel- In Ref. [192] a simple model for inter-
lium leads to results that are nearly free of facial adsorption is presented in which
empirical parameters and that are in close the three-dimensional system reduces to
agreement with experiment. an equivalent two-dimensional lattice gas.
Computer simulations of the electro- This model is applied to study the UPD of
chemical double layer could contribute to copper on the (111) surface of a crystalline
a better understanding of the double-layer gold electrode in the presence of bisul-
structure [193–197]. fate. A cluster variation approximation
20
III II
I
[µA cm−2]
0
j
−20
−200 0 200 400 600

Potential (SCE)
[mV]
Fig. 4 Cyclic voltammograms at 20 mV s−1 on Pt(111) in
0.2 M HClO4 at pH 0.8 with 0 mM (− · − · −), (I) 1 mM
(- - - - -), (II) 3 mM ( . . . . .) and (III) 11 mM ( )
CH3 COOH. (Reprinted from Ref. [83] with permission from
Elsevier Science.)
−E H
[mV]
600 700 800 900 1000 1100 1200 1300 1400
10
8 D1
6 D3
D2
4
[µA cm−2]
2
0
i
−2
−4 A1
A3 A2
−6
−8
−10
−100 0 100 200 300 400 500 600 700 800
∆E
(a) [mV]
180
160 Adsorption
Desorption
140
120
[µC cm−2]
100
zF Γ
80
60
40
20
0
−100 0 100 200 300 400 500 600 700 800
∆E
(b) [mV]
Fig. 5 (a) Cyclic voltammogram in the system
Au(100)/5 × 10−3 M Ag2 SO4 + 0.5 MH2 SO4 at T = 298 K.
|dE/dt| = 7 mV s−1 . An and Dn with n = 1, 2, 3 denote cathodic
adsorption and anodic desorption peaks, respectively. (b) (E)
isotherm of the system Au(100)/4.2 × 10−4 M Ag2 SO4
+ 0.5M H2 SO4 at T = 298 K. (Reprinted from Ref. [187] with
permission from Elsevier Science.)
Fig. 6 Effect of the UPD of Cd2+ (1),

Cu2+ (2), and Ag+ (3) on the potential
10−10 Γ/mol cm−2

0.75 3
dependence of the adsorption of HSO4 −
ions on polycrystalline gold. Curve 4
0.5 2
(- - - - -) corresponds to the potential 1
dependence without any addition: 4
cH2 SO4 = 4 × 10−4 mol dm−3 , 0.25
cMe = 8 × 10−4 mol dm−3 in
1 mol dm−3 HClO4 . (Reprinted from
Ref. [147] with permission from Elsevier −400 0 400 800 1200
Science.) E
[mV]
Fig. 7 Cyclic voltammogram and 0.10

simultaneously recorded frequency P1′
change of the Au(111) EQCM electrode
P2′
Current density
in 0.05 M H2 SO4 (- - - - -) and in 0.05

[mA cm−2]
0.05 M H2 SO4 + 5 mM CuSO4 . Sweep

rate = 10 mV−1 . The potential was
scanned from 0.75 V and reversed at 0
0.33 V. (Reprinted from Ref. [55] with
0.05
P2 P1
4
2
0
−2
[Hz]
∆f
−4
−6
100 ng cm−2
−8
−10
0.3 0.4 0.5 0.6 0.7 0.8

Potential vs RHE
[V]
to the model system yields realistic ad- References

sorption isotherms, and the resulting
model voltammogram agrees well with 1. S. Trasatti, R. Parsons, Pure Appl. Chem.
experiment. 1986, 58, 437–454.
2. M. A. Habib, J. O’M. Bockris, in Compre-
hensive Treatise of Electrochemistry, Double
Acknowledgment
Layer (Eds.: J. O’M. Bockris, B. E. Conway,
E. Yeager), Plenum Press, New York, 1980,
Financial support from Hungarian Sci- pp. 135–219, Vol. 1.
ence Foundation (Grants T023056 and 3. A. J. Bard, L. R. Faulkner, Electrochemical
031703) is acknowledged. Methods, Fundamentals and Applications,
John Wiley & Sons, New York, 1980, 25. D. M. Kolb, in Advances in Electrochem-
pp. 488–552. istry and Electrochemical Engineering (Eds.:
4. S. Trasatti, in Trends in Interfacial Elec- H. Gerischer, C. W. Tobias), John Wiley
trochemistry (Ed.: A. F. Silva), D. Reidel & Sons, New York, 1978, pp. 125–271,
Publishing Company, Dordrecht, 1986, Vol. 11.
pp. 25–48. 26. R. R. Adžić, in Advances in Electrochem-
5. R. Parsons, in Trends in Interfacial Elec- istry and Electrochemical Engineering (Eds.:
trochemistry (Ed.: A. F. Silva), D. Reidel H. Gerischer, C. W. Tobias), John Wiley
Publishing Company, Dordrecht, 1986, & Sons, New York, 1985, pp. 159–260,
pp. 71–81. Vol. 13.
6. A. Hamelin, in Trends in Interfacial Elec- 27. A. Aramata, in Modern Aspects of
trochemistry (Ed.: A. F. Silva), D. Reidel Electrochemistry (Eds.: J. O’M. Bockris
Publishing Company, Dordrecht, 1986, et al.), Plenum Press, New York, 1977,
pp. 83–102. pp. 181–250, Vol. 31.
7. R. Parsons, in Trends in Interfacial Elec- 28. S. Szabó, Int. Rev. Phys. Chem. 1991, 10,
trochemistry (Ed.: A. F. Silva), D. Reidel 207–248.
Publishing Company, Dordrecht, 1986, 29. W. Lorenz, Z. Phys. Chem. 1962, 219,
pp. 373–385. 421–423.
8. D. H. Everett, Pure Appl. Chem. 1970, 21, 30. W. Lorenz, Z. Phys. Chem. N. F. 1967, 54,
583–596. 191–195.
9. G. Valette, R. Parsons, J. Electroanal. Chem. 31. W. Lorenz, Z. Phys. Chem. 1970, 244, 65–84.
1986, 204, 291–297. 32. W. Lorenz, Z. Phys. Chem. 1971, 248,
10. G. Valette, R. Parsons, J. Electroanal. Chem.
161–170.
1985, 191, 377–386.
33. C. Engler, W. Lorenz, Z. Phys. Chem. 1978,
11. H. D. Hurwitz, J. Electroanal. Chem. 1965,
259, 1188–1190.
10, 35–41.
34. K. J. Vetter, J. W. Schultze, Ber. Bunsen-Ges
12. E. Dutkiewicz, R. Parsons, J. Electroanal.
Phys. Chem. 1972, 76, 927–933.
Chem. 1966, 11, 100–110.
35. K. J. Vetter, J. W. Schultze, J. Electroanal.
13. G. Láng, K. E. Heusler, J. Electroanal. Chem.
Chem. 1974, 53, 67–76.
1995, 391, 169–179.
36. A. Aramata, S. Taguchi, T. Fukuda et al.,
14. K. E. Heusler, G. Láng, Electrochim. Acta
1997, 42, 747–756. Electrochim. Acta 1998, 44, 999–1007.
15. G. Valincius, Langmuir 1998, 14, 37. G. Horányi, A. Aramata, J. Electroanal.
6307–6319. Chem. 1997, 437, 259–262.
16. J. Lipkowski, W. Schmickler, D. M. Kolb 38. G. Horányi, A. Aramata, J. Electroanal.
et al., J. Electroanal. Chem. 1998, 452, Chem. 1997, 434, 201–207.
193–197. 39. G. Horányi, ACH Models Chem. 1997, 134,
17. W. Haiss, R. J. Nichols, J. K. Sass et al., J. 33–47.
Electroanal. Chem. 1998, 452, 199–202. 40. G. Horányi, M. Wasberg, J. Electroanal.
18. W. Schmicler, E. Leiva, J. Electroanal. Chem. Chem. 1996, 413, 161–164.
1998, 453, 61–68. 41. P. Zelenay, M. Gamboa-Aldeco, G. Horányi
19. R. Guidelli, J. Electroanal. Chem. 1998, 453, et al., J. Electroanal. Chem. 1993, 357,
69–78. 307–326.
20. G. Valincius, J. Electroanal. Chem. 1999, 42. K. Varga, P. Zelenay, G. Horányi et al., J.
478, 40–49. Electroanal. Chem. 1992, 327, 291–306.
21. B. M. Grafov, Russ. J. Electrochem. 1999, 35, 43. N. M. Marković, H. A. Gasteiger, P. N. Ross
1029–1032. Jr., Langmuir 1995, 11, 4098–4108.
22. G. Láng, K. E. Heusler, J. Electroanal. Chem. 44. Y. Shingaya, H. Matsumoto, H. Ogasawara
1999, 472, 168–173. et al., Surf. Sci. 1995, 335, 23–31.
23. R. Guidelli, J. Electroanal. Chem. 1999, 472, 45. R. Gómez, H. S. Yee, G. M. Bommarito
174–177. et al., Surf. Sci. 1995, 335, 101–109.
24. B. M. Grafov, J. Electroanal. Chem. 1999, 46. N. M. Marković, H. A. Gasteiger, C. A. Lucas
471, 105–108. et al., Surf. Sci. 1995, 335, 91–100.
47. L. J. Buller, E. Herrero, R. Gómez et al., 71. A. Hamelin, A. M. Martins, J. Electroanal.

J. Chem. Soc., Faraday Trans. 1996, 92, Chem. 1996, 407, 13–21.
3757–3762. 72. A. Hamelin, J. Electroanal. Chem. 1995, 386,
48. I. Oda, Y. Shingaya, H. Matsumoto et al., J. 1–10.
Electroanal. Chem. 1996, 409, 95–101. 73. J. Clavilier, R. Faure, G. Guinet et al., J.
49. J. Inukai, Y. Osawa, M. Wakisaka et al., J. Electroanal. Chem. 1980, 107, 205–209.
Phys. Chem. B 1998, 102, 3498–3505. 74. F. T. Wagner, R. N. Ross Jr., J. Electroanal.
50. E. Herrero, S. Glazier, L. J. Buller et al., J. Chem. 1983, 150, 141–164.
Electroanal. Chem. 1999, 461, 121–130. 75. K. al Faaf-Golze, D. M. Kolb, D. Scherson,
51. M. S. Zei, K. Wu, M. Eiswirth et al., Elec- J. Electroanal. Chem. 1986, 200, 353–362.
trochim. Acta 1999, 45, 809–817. 76. N. Marković, M. Hanson, G. McDougall
52. J. Hotlos, O. M. Magnussen, R. J. Behm, et al., J. Electroanal. Chem. 1986, 214,
Surf. Sci. 1995, 335, 129–144. 555–566.
53. Z. Shi, S. Wu, J. Lipkowski, Electrochim. 77. P. W. Faguy, N. Marković, R. R. Adžić et al.,
Acta 1995, 40, 9–15. J. Electroanal. Chem. 1990, 289, 245–262.
54. J. G. Gordon, O. R. Melroy, M. F. Toney, 78. E. Herrero, J. M. Feliu, A. Wieckowski et al.,
Electrochim. Acta 1995, 40, 3–8. Surf. Sci. 1995, 325, 131–138.
55. M. Watanabe, H. Uchida, M. Miura et al., J. 79. P. W. Faguy, N. S. Marinković, R. R. Adžić,
Electroanal. Chem. 1995, 384, 191–195. J. Electroanal. Chem. 1996, 407, 209–218.
56. Z. Shi, S. Wu, J. Lipkowski, J. Electroanal. 80. A. Zolfaghari, G. Jerkiewicz, J. Electroanal.
Chem. 1995, 384, 171–177. Chem. 1997, 422, 1–6
57. E. D. Chabala, J. Cairns, T. Rayment, J. Elec- 81. T. Fukuda, A. Aramata, J. Electroanal. Chem.
troanal. Chem. 1996, 412, 77–84. 1997, 440, 153–161.
82. V. Lazarescu, J. Clavilier, Electrochim. Acta
58. W. Haiss, J. K. Sass, J. Electroanal. Chem.
1998, 44, 931–941.
1996, 410, 119–124.
83. T. Fakuda, A. Aramata, J. Electroanal. Chem.
59. M. Legault, L. Blum, D. A. Huckaby, J. Elec-
1999, 467, 112–120
troanal. Chem. 1996, 409, 79–86.
84. F. C. Nart, T. Iwasita, M. Weber, Electrochim.
60. S. Wu, Z. Shi, J. Lipkowski et al., J. Phys.
Acta 1994, 39, 961–968.
Chem. B 1997, 101, 10 310–10 322.
85. T. Iwasita, A. Rodes, E. Pastor, J. Elec-
61. H. Uchida, M. Hiei, M. Watanabe, J. Elec-
troanal. Chem. 1995, 383, 181–189.
troanal. Chem. 1998, 442, 97–106. 86. E. Pastor, A. Rodes, T. Iwasita, J. Elec-
62. P. Mrozek, Y.-E. Sung, A. Wieckowski, Surf. troanal. Chem. 1996, 404, 61–68.
Sci. 1995, 335, 44–51. 87. Y. Shingaya, K. Hirota, H. Ogasawara et al.,
63. A. J. Motheo, J. R. Santos Jr., A. Sadkowski J. Electroanal. Chem. 1996, 409, 103–108.
et al., J. Electroanal. Chem. 1995, 397, 88. Y. Shingaya, M. Ito, J. Electroanal. Chem.
331–334. 1999, 467, 299–306.
64. P. Mrozek, Y. E. Sung, M. Han et al., Elec- 89. Z. Shi, J. Lipkowski, M. Gamboa et al., J.
trochim. Acta 1995, 40, 17–28. Electroanal. Chem. 1994, 366, 317–326.
65. E. D. Chabala, A. R. Ramadan, T. Brunt 90. Z. Shi, J. Lipkowski, S. Mirwald et al., J.
et al., J. Electroanal. Chem. 1996, 412, 67–75. Electroanal. Chem. 1995, 396, 115–125.
66. E. D. Chabala, T. Rayment, J. Electroanal. 91. Z. Shi, J. Lipkowski, J. Electroanal. Chem.
Chem. 1996, 401, 257–261. 1996, 403, 225–239.
67. A. N. Frumkin, O. A. Petrii, B. B. 92. Z. Shi, J. Lipkowski, S. Mirwald et al., J.
Damaskin, J. Electroanal. Chem. 1970, 27, Chem. Soc., Faraday Trans. 1996, 92,
81–90. 3737–3746.
68. R. Parsons, J. Electroanal. Chem. 1994, 376, 93. J. Lipkowski, Z. Shi, A. Chen et al., Elec-
15–20. trochim. Acta 1998, 43, 2875–2888.
69. A. Hamelin, in Modern Aspects of Electro- 94. A. Chen, Z. Shi, D. Bizzotto et al., J. Elec-
chemistry (Eds.: B. E. Conway, R. E. White, troanal. Chem. 1999, 467, 342–353.
J. O’M. Bockris), Plenum Press, New York, 95. A. Chen, J. Lipkowski, J. Phys. Chem. B
1985, pp. 1–101, Vol. 16. 1999, 103, 682–691.
70. A. Hamelin, J. Electroanal. Chem. 1996, 407, 96. I. R. de Moraes, F. C. Nart, J. Electroanal.
1–11. Chem. 1999, 461, 110–120.
97. N. S. Marinković, J. S. Marinković, R. R. 118. M. Weber, F. C. Nart, Electrochim. Acta

Adžić, J. Electroanal. Chem. 1999, 467, 1996, 41, 653–659.
291–298. 119. F. Kitamura, T. Ohsaka, K. Tokuda, J. Elec-
98. K. J. Stevenson, X. Gao, D. W. Hatchett troanal. Chem. 1996, 412, 183–188.
et al., J. Electroanal. Chem. 1998, 447, 43–51. 120. T. Iwasita, X. Xia, J. Electroanal. Chem. 1996,
99. P. Broekmann, M. Wilms, M. Kruft et al., J. 411, 95–102.
Electroanal. Chem. 1999, 467, 307–324. 121. J. J. Calvente, N. S. Marinković, Z. Kováčová
100. S. Smoliñski, P. Zelenay, J. Sobkowski, J. et al., J. Electroanal. Chem. 1997, 421, 49–57.
Electroanal. Chem. 1998, 442, 41–47. 122. W. R. Fawcett, A. A. Kloss, J. J. Calvente
101. S. Smoliñski, J. Sobkowski, J. Electroanal. et al., Electrochim. Acta 1998, 44, 881–887.
Chem. 1999, 463, 1–8. 123. F. Kitamura, N. Nanbu, T. Ohsaka et al., J.
102. D. M. Dražić, A. V. Tripković, K. D. Electroanal. Chem. 1998, 452, 241–249.
Popović et al., J. Electroanal. Chem. 1999, 124. C. A. Melendres, F. Hahn, J. Electroanal.
466, 155–164. Chem. 1999, 463, 258–261.
103. M. Väärtnou, P. Pärsimägi, E. Lust, J. Elec- 125. G. M. Brown, G. A. Hope, J. Electroanal.
troanal. Chem. 1995, 385, 115–119. Chem. 1996, 405, 211–216.
104. M. Väärtnou, P. Pärsimägi, E. Lust, J. Elec- 126. S. Z. Zou, Y. X. Chem, B. W. Mao et al., J.
troanal. Chem. 1996, 407, 227–232. Electroanal. Chem. 1997, 424, 19–24.
105. M. Väärtnou, E. Lust, J. Electroanal. Chem. 127. M. Bron, R. Holze, Electrochim. Acta 1999,
1999, 469, 182–188. 45, 1121–1126.
106. M. Väärtnou, E. Lust, Electrochim. Acta 128. B. Pettinger, C. Bilger, J. Lipkowski et al.,
1999, 44, 2437–2444. in Interfacial Electrochemistry, Theory, Experi-
107. J. Ehrlich, T. Ehrlich, A. Jänes et al., Elec- ment, and Applications (Ed.: A. Wieckowski),
trochim. Acta 1999, 45, 935–943. Marcel Dekker, New York, 1999,
108. M. Väärtnou, E. Lust, Electrochim. Acta pp. 373–404.
2000, 45, 1623–1629. 129. S. Mirwald, B. Pettinger, J. Lipkowski, Surf.
109. R. Greef, R. Peat, L. M. Peter et al., Instru- Sci. 1995, 335, 264–272.
mental Methods in Electrochemistry, John 130. A. Tadjeddine, A. le Rille, in Interfacial
Wiley & Sons, New York, 1985. Electrochemistry, Theory, Experiment, and
110. P. A. Christensen, A. Hamnett, Techniques Applications (Ed.: A. Wieckowski), Marcel
and Mechanisms in Electrochemistry, Blackie Dekker, New York, 1999, pp. 317–343.
Academic & Professional, London, 1994. 131. A. Tadjeddine, A. le Rille, Electrochim. Acta
111. A. Wieckowski, (Ed.), Interfacial Electro- 1999, 45, 601–609.
chemistry. Theory, Experiment, and Applica- 132. J. X. Wang, R. R. Adžić, B. M. Ocko, in In-
tions, Marcel Dekker, New York, 1999. terfacial Electrochemistry, Theory, Experiment,
112. Y. Chiu, M. A. Genshaw, J. Phys. Chem. and Applications (Ed.: A. Wieckowski), Mar-
1969, 73, 3571–3577. cel Dekker, New York, 1999, pp. 175–186.
113. W.-K. Puik, M. A. Genshow, J. O’M. Bockris, 133. M. F. Toney, J. N. Howard, J. Richer et al.,
J. Phys. Chem. 1970, 74, 4266–4275. Surf. Sci. 1995, 335, 326–332.
114. D. M. Kolb, in Trends in Interfacial Elec- 134. O. Endo, M. Kiguchi, T. Yokoyaama et al.,
trochemistry (Ed.: A. F. Silva), D. Reidel J. Electroanal. Chem. 1999, 473, 19–24.
Publishing Company, Dordrecht, 1986, 135. M. Cappadonia, K. M. Robinson, J. Schmid-
pp. 301–330. berger et al., J. Electroanal. Chem. 1997, 436,
115. A. Bewick, in Trends in Interfacial Elec- 73–78.
trochemistry (Ed.: A. F. Silva), D. Reidel 136. N. A. Balashova, V. N. Kazarinov, in Electro-
Publishing Company, Dordrecht, 1986, analytical Chemistry (Ed.: A. J. Bard), Marcel
pp. 331–358. Dekker, New York, 1969, pp. 135–197,
116. T. Iwasita, E. Pastor, in Interfacial Electro- Vol. 3.
chemistry, Theory, Experiment, and Applica- 137. G. Horányi, Electrochim. Acta 1980, 25,
tions (Ed.: A. Wieckowski), Marcel Dekker, 43–57.
New York, 1999, pp. 353–372. 138. V. E. Kazarinov, V. N. Andreev, in Compre-
117. M. Weber, F. C. Nart, Langmuir 1996, 12, hensive Treatise on Electrochemistry, (Eds.:
1895–1900. E. Yeager, J. O’M. Bockris, B. E. Conway),
Plenum Press, New York, 1990, 157. M. Hepel, in Interfacial Electrochemistry,

pp. 393–443, Vol. 9. Theory, Experiment, and Applications (Ed.:
139. A. Wieckowski, in Modern Aspects of Elec- A. Wieckowski), Marcel Dekker, New York,
trochemistry (Eds.: J. O’M. Bockris, B. E. 1999, pp. 599–630.
Conway, R. E. White), Plenum Press, New 158. V. Tsionsky, L. Daikhin, E. Gileadi, J. Elec-
York, 1990, pp. 65–119, Vol. 21. trochem. Soc. 1996, 143, 2240–2245.
140. P. Zelenay, A. Wieckowski, in Electrochem- 159. L. Jaeckel, G. Láng, K. E. Heusler, Elec-
ical Interfaces: Modern Techniques for In trochim. Acta. 1994, 39, 1031–1038.
Sity Surface Characterization (Ed.: H. D. 160. H. W. Lei, H. Uchida, M. Watanabe, J. Elec-
Abruna), VCH Publishers, New York, 1991, troanal. Chem. 1996, 413, 131–136.
pp. 479–527. 161. H. Uchida, N. Ikeda, M. Watanabe, J. Elec-
141. E. K. Krauskopf, A. Wieckowski, in Fron- troanal. Chem. 1997, 424, 5–12.
tiers of Electrochemistry (Eds.: P. N. Ross, 162. B. K. Niece, A. A. Gewirth, J. Phys. Chem. B
J. Lipkowski), VCH Publishers, New York, 1998, 102, 818–823.
1992, pp. 119–169. 163. G. Gloaguen, J.-M. Léger, C. Lamy, J. Elec-
142. M. E. Gamboa-Aldeco, K. Franaszczuk, troanal. Chem. 1999, 467, 186–192.
A. Wieckowski, in The Handbook of Surface 164. H. Matsumoto, J. Inukai, M. Ito, J. Elec-
Imaging and Visualization (Ed.: A. T. Hub- troanal. Chem. 1994, 379, 223–231.
bard), CRC Press, New York, 1995, p. 635. 165. G. Nagy, Electrochim. Acta 1995, 40,
143. G. Horányi, in Interfacial Electrochemistry, 1417–1420.
Theory Experiment, and Applications (Ed.: 166. I. Villegas, X. Gao, M. J. Weaver, Electro-
A. Wieckowski), Marcel Dekker, New York, chim. Acta 1995, 40, 1267–1275.
1999, pp. 477–491. 167. F. Möller, O. M. Magnussen, R. J. Behm,
144. Gy. Horányi, in A Specialist Periodical Re- Electrochim. Acta 1995, 40, 1259–1265.
port, Catalysis (Ed.: J. J. Spivey), The Royal 168. K. Ogaki, K. Itaya, Electrochim. Acta 1995,
Society of Chemistry, Cambridge, 1996, 40, 1249–1257.
p. 254–301, Vol. 12. 169. D. Carnal, P. I. Oden, U. Müller et al., Elec-
145. G. Horányi, Rev. Anal. Chem. 1995, 14, trochim. Acta 1995, 40, 1223–1235.
1–58. 170. L. J. Wan, S. L. Yau, G. M. Swan et al., J.
146. A. Wieckowski, P. Zelenay, K. Varga, J. Electroanal. Chem. 1995, 381, 105–111.
Chim. Phys. 1991, 88, 1247–1270. 171. J. C. Bondos, A. A. Gewirth, R. G. Nuzzo, J.
147. G. Horányi, E. M. Rizmayer, P. Joó, J. Elec- Phys. Chem. 1996, 100, 8617–8620.
troanal. Chem. 1983, 152, 211–222. 172. J. Inukai, S. Sugita, K. Itaya, J. Electroanal.
148. Y.-E. Sung, A. Thomas, M. Gamboa-Aldeco Chem. 1996, 403, 159–168.
et al., J. Electroanal. Chem. 1994, 378, 173. A. M. Funtikov, U. Stimming, R. Vogel, J.
131–142. Electroanal. Chem. 1997, 428, 147–153.
149. D. Poškus, G. Agafonovas, J. Electroanal. 174. W. H. Li, R. J. Nichols, J. Electroanal. Chem.
Chem. 1995, 393, 105–112. 1998, 456, 135–160.
150. A. Kolics, A. E. Thomas, A. Wieckowski, J. 175. P. A. Christensen, A. Hamnett, in Tech-
Chem. Soc., Faraday Trans. 1996, 92, niques and Mechanisms in Electrochemistry,
3727–3736. Blackie Academic, Chapman & Hall, Lon-
151. D. Poškus, A. Agafonovas, I. Jurgaitiene, J. don, 1994, pp. 67–89.
Electroanal. Chem. 1997, 425, 107–115. 176. D. Hecht, J.-H. Strehblow, J. Electroanal.
152. D. Poškus, J. Electroanal. Chem. 1998, 442, Chem. 1997, 440, 211–217.
5–7. 177. D. Hecht, H.-H. Strehblow, J. Electroanal.
153. A. Kolics, J. C. Polkinghorne, A. Wiec- Chem. 1997, 436, 109–118.
kowski, Electrochim. Acta 1998, 43, 178. D. Lützenkirchen-Hecht, H.-H. Strehblow,
2605–2618. Electrochim. Acta 1998, 43, 2957–2968.
154. P. Waszczuk, A. Wnuk, J. Sobkowski, Elec- 179. A. T. Hubbard, in Interfacial Electrochem-
trochim. Acta 1999, 44, 1789–1795. istry, Theory, Experiment, and Applications
155. A. Wieckowski, A. Kolics, J. Electroanal. (Ed.: A. Wieckowski), Marcel Dekker, New
Chem. 1999, 464, 118–122. York, 1999, pp. 211–229.
156. T. Nomura, M. Iijima Anal. Chim. Acta 180. U. W. Hamm, D. Kramer, R. S. Zhai et al.,
1981, 131, 97–102. J. Electroanal. Chem. 1996, 414, 85–89.
181. I. Villegas, M. J. Weaver, J. Phys. Chem. and Applications (Ed.: A. Wieckowski), Mar-
1996, 100, 19 502–19 511. cel Dekker, New York, 1999, pp. 1–17.
182. Y. Shingaya, M. Ito, Electrochim. Acta 1998, 190. L. Blum, M. D. Legault, D. A. Huckaby in
44, 889–895. Interfacial Electrochemistry, Theory, Experi-
183. G. Pirug, H. P. Bonzei, Surf. Sci. 1998, 405, ment, and Applications (Ed.: A. Wieckowski),
87–103. Marcel Dekker, New York, 1999, pp. 19–31.
184. E. M. Stuve, K. Bange, J. K. Sass, in Trends 191. M. L. Berkowitz, I. C. Yeh, E. Spohr, in In-
in Interfacial Electrochemistry (Ed.: A. F. terfacial Electrochemistry, Theory, Experiment,
Silva), D. Reidel Publishing Company, and Applications (Ed.: A. Wieckowski), Mar-
Dordrecht, 1986, pp. 255–280. cel Dekker, New York, 1999, pp. 33–45.
185. G. Pirug, H. P. Bonzel, in Interfacial Electro- 192. G. Brown, P. A. Rikvold, S. J. Mitchell et al.,
chemistry, Theory, Experiment, and Applica- in Interfacial Electrochemistry, Theory, Experi-
tions (Ed.: A. Wieckowski), Marcel Dekker, ment, and Applications (Ed.: A. Wieckowski),
New York, 1999, pp. 269–286. Marcel Dekker, New York, 1999, pp. 47–61.
186. Y. Shingaya, M. Ito, in Interfacial Electro- 193. A. Ignaczak, J. A. N. F. Gomes, J. Elec-
chemistry, Theory, Experiment, and Applica- troanal. Chem. 1997, 420, 209–218.
tions (Ed.: A. Wieckowski), Marcel Dekker, 194. M. T. M. Koper, Electrochim. Acta 1998, 44,
New York, 1999, pp. 287–300. 1207–1212.
187. S. Garcia, D. Salinas, C. Mayer et al., Elec- 195. W. Schmickler, (Ed.), J. Electroanal. Chem.
trochim. Acta 1998, 43, 3007–3019. 1998, 450(2), 157–351.
188. S. Trasatti, (Ed.), Electrochim. Acta 1996, 196. P. A. Bopp, A. Kohlmeyer, E. Spohr, Elec-
41(14), 2071–2338. trochim. Acta 1998, 43, 2911–2918.
189. J. W. Halley, S. Walbran, D. Lee Price, in In- 197. E. Spohr, Electrochim. Acta 1999, 44,
terfacial Electrochemistry, Theory, Experiment, 1697–1705.
3.3 Phase Transitions in Two-dimensional Adlayers at Electrode Surfaces 383
3.3 anodic films [17]. Other liquid metal elec-

Phase Transitions in Two-dimensional trodes employed in these studies have been
Adlayers at Electrode Surfaces: gallium [18] and amalgams [19]. The use of
Thermodynamics, Kinetics, and Structural well-defined single-crystal electrodes, such
Aspects as Au(hkl), Ag(hkl), and Pt(hkl), demon-
strated the important influence of the
Thomas Wandlowski
substrate material and its surface struc-
Institute of Surfaces and Interfaces, Juelich,
ture (crystallographic orientation, defect,
Germany
and reconstruction pattern) on electrode
3.3.1 kinetics and 2D phase transitions in elec-
Introduction trochemically formed adlayers of organic
molecules [20], ions [21], or metals [22].
Two-dimensional (2D) phase transitions Additional contributions arise from sol-
on surfaces or in adlayers have received vent molecules and the codeposition of
increased attention in recent years [1–4] ions of the supporting electrolyte. The
as they are related to important as- combination of classical electrochemical
pects in surface, interfacial and mate- experiments, based on measurements of
rials science, and nanotechnology, such current, charge density, or impedance as
as ordered adsorption, island nucleation a function of the applied electrode poten-
and growth [2, 5–7], surface reconstruc- tial or time, with structure-sensitive in situ
tion [8], and molecular electronics [9]. Ki- techniques, such as scanning probe mi-
netic phenomena such as catalytic activity croscopy [23], surface X-ray scattering [24],
electroreflectance [25], second harmonic
and chirality of surfaces [10–12], selective
generation [26] and/or vibrational spec-
recognition of molecular functions [13], or
trocopies (FTIR [27], surface-enhanced in-
oscillating chemical reactions [14] are di-
frared reflection absorption spectroscopy
rectly related to phase-formation processes
(SEIRAS) [28], Raman [29], and sum fre-
at interfaces.
quency generation (SFG) [30]) provide a
Concepts of ordering and reactivity in
powerful strategy (1) to access macro-
two dimensions can be also addressed
scopic data and (2) to develop a molecu-
in studies with organic, ionic, or metal- lar/atomic level structural and mechanistic
lic (sub)-monolayers at potentiostatically understanding of 2D phase formation in
controlled electrode–electrolyte interfaces. these systems. The structure and stability
This approach offers the advantage, in of the adsorbed monolayers is affected by
comparison to a nonelectrochemical envi- the symmetry of the substrate, the ad-
ronment, that the structural and dynamic sorbate–adsorbate interactions, and the
properties of the adsorbate and the sub- corrugations in the adsorbate–substrate
strate can be directly tuned through the interaction potential.
applied electrode potential. The first elec- The present chapter starts with some
trochemical studies of 2D phase transi- general remarks on the thermodynamics
tions were mostly confined to processes and kinetics of 2D phase transitions in po-
occurring at ideally smooth mercury elec- tentiostatically generated adlayers on well-
trodes. Typical examples are the forma- defined metal–electrolyte interfaces. Sub-
tion of compact monolayers of organic sequently, three main groups of systems
molecules or salts [15, 16] and so-called will be considered: organic and (an-)ionic
monolayers and underpotential-deposited the intensity or position of a characteristic

(UPD) metal films. Each subtopic is di- diffraction spot [39].
vided into an introductory overview and The behavior of the order parameter
several case studies. The latter have been near the transition potential or temper-
selected in order to illustrate typical ap- ature distinguishes two rather different
proaches and the present understanding transformation scenarios. In a first-order
of thermodynamic, kinetic, and structural phase transition, the particle configura-
aspects of phase formation processes in tion, and consequently the order pa-
2D electrochemical systems. The review rameter, changes discontinuously. This
aims at covering main developments in implies that quantities such as the in-
this field, but does not attempt to be all- ternal energy U or entropy S change
inclusive. We apologize for unavoidable discontinuously (but the free energy F
omissions. remains continuous!), and a latent heat
is involved in the transition [40, 41]. In
3.3.2 this case, two distinct free energy curves
Thermodynamic Aspects cross one another. The system abruptly
3.3.2.1 Phase Transitions and Order
changes from one equilibrium phase to a
Parameter second equilibrium phase [42, 43]. Anal-
Two-dimensional phase transitions occur ogous to Langmuir–Blodgett (LB) films
at a clean or adsorbate-covered surface or insoluble monolayers at the air–water
because a system in thermodynamic interface [44], one may distinguish be-
equilibrium seeks to minimize its free tween gaseous, liquidlike (liquid-expanded
energy, F [1, 31–33] and/or liquid-condensed), and solidlike
films [2, 3, 5, 42, 43]. First-order phase
F = U −TS (1) transitions are characterized by (1) the
phase coexistence and (2) the formation
One phase will supplant another at a of nuclei that subsequently grow. Typical
given electrode potential E and/or temper- electrochemical responses comprise sharp
ature T because different states partition current peaks (‘‘spikes’’), discontinuities
their free energy between the internal en- in the charge density or capacitance versus
ergy U (T , E, c) and the entropy S(E, T , c) potential curves (‘‘capacitance pits’’ [45]),
in different ways [34–36]. The competing or the steplike blocking of charge trans-
phases are characterized by the so-called fer reactions [46]. Examples of first-order
order parameter [1]. By construction, the phase transitions in electrochemical sys-
order parameter has a nonzero value in tems are the potential-induced formation
one phase (usually low temperature/low of a hydrogen-bonded uracil film on
symmetry state) and vanishes in the Hg [47] or Au(hkl) electrodes [20] from
other (high-temperature/high-symmetry) its disordered adsorbed monomers, the
phase. For liquid ↔ gas or liquid ↔ solid deposition of the rotated hexagonal Pb
phase transitions, the order parameter is UPD monolayer on Ag(111) from citrate-
the difference in density or the respective containing electrolyte [48–50], or the
√ √
high-density coverage [37, 38]. Structural gaseous-like ↔ c( 2 × 2 2)R45◦ transi-
phase transitions, such as orientational or tion of bromide adsorbed at the aqueous
positional changes, may be represented by electrolyte–Au(100) interface [51].
By contrast, the two competing phases for bromide onto a single-crystalline

become indistinguishable at the critical Ag(100)-electrode a potential-induced con-
values of electrode potential (Ec ) or tem- tinuous (second-order) phase transition
perature (Tc ) for a continuous transition between a disordered, low-coverage halide
* (In modern theories, one typically dis- adlayer at more negative potentials, and a
tinguishes between discontinuous and doubly degenerate c(2 × 2) ordered phase
continuous phase transitions. This clas- with coverage 0.5 at more positive poten-
sification is more appropriate than the tials (paragraph 3.3.4.1.2). A temperature-
classical approach of Ehrenfest, which dis- dependent low energy electron diffraction
tinguishes between first-, second-, and (LEED) study of Ag(100)/Br revealed the
higher-order phase transitions [52]). The same type of phase transition under UHV-
particle configuration and the order pa- conditions [58]. The transition was shown
rameter change continuously, while the to belong to the 2D-square Ising univer-
symmetry of the system changes discon- sality class, for which the exact solution
tinuously. Close to the continuous phase has been known since the work of On-
transition, strong fluctuations in the or- sager [59, 60]. The other example of a
der parameter and in its values around continuous phase transition is the tem-
its mean occur. The correlation length ξ , perature dependence of a 2D condensed
which is a measure of the domain size, di- film of 5-iodocytosine adsorbed at a mer-
verges. The Landau–Lifshitz theory states cury electrode in the neighborhood of the
that a phase transition can be continuous critical temperature, at constant adsorbate
only if the low-temperature/low-symmetry concentration and electrode potential [35]
surface space group G is a subgroup (paragraph 3.3.4.3.2).
of the high-temperature/high-symmetry Exact solutions, as obtained for uni-
phase Go , and if the respective transform versality classes of 2D systems such as
function is an irreducible representation the 2D-Ising (isomorphic to the lattice
of Go [41]. gas) or 3-state Potts model, are impor-
One typically finds that the order param- tant but do not contain any qualitative
eter of a continuous phase transition varies insight into the nature of the respective
in the critical region as (E − Ec )β [53] or continuous phase transitions. Such an in-
(T − Tc )β [1, 33]. The numerical value of sight is given by the (1) renormalization
the critical exponent β depends only on a group [61], which explains how the quali-
few physical properties, such as the dimen- tative features of the cooperative behavior
sion of the local variable (order parameter) arise, (2) series expansion techniques [60],
in the Hamiltonian, the symmetry of the (3) MC simulations [33, 38, 62] and in
coupling between the local variables, and situ structure-sensitive experimental stud-
the dimensionality of the system (here ies with the spatial resolution of individual
2D). This property is called universal- atoms and/or molecules.
ity [32, 33, 54]. Systems with identical crit-
ical behavior form one universality class. 3.3.2.2 Adsorption Isotherms and Lateral
Only two examples have been reported Interactions
for interfacial electrochemical systems: In Adsorption phenomena including 2D
situ surface X-ray scattering (SXS) [53], phase transitions at electrochemical inter-
chronocoulometry [55], and Monte Carlo faces may be classified, according to the
(MC) simulations [56, 57] demonstrated binding energy of the adsorbate onto the
surface, in two categories. Physisorption maximum surface excess m or the maxi-

corresponds to small binding energies, mum number of substrate sites), adsorbate
the substrate–adsorbate interactions are concentration (activity), electrode poten-
mainly due to van der Waals forces and tial, and temperature allow the construc-
involve almost no mixing between the or- tion of various types of phase diagrams.
bitals of the adsorbate and the substrate [1]. Graphs of θ = f (c)T ,E and θ = f (E)T ,c
Lateral interactions between the adsorbed versus temperature as the parameter are
species are dominated by dipole–dipole, identified as isotherms [1, 63].
dipole–induced dipole and hydrophobic Phase-formation phenomena at elec-
interactions, π-stacking and/or hydrogen trode–electrolyte interfaces can be con-
bonding [63, 64]. veniently treated with lattice gas con-
Chemisorption represents the formation cepts [38, 60, 67]. Such models consider
of a surface chemical bond, which is ei- that the entities, atoms, ions, or molecules,
ther covalent (sharing of electrons) or are fixed to particular ‘cells’ i, j . (M. Fisher
ionic (electron transfer). The understand- explicitly pointed out ‘‘. . . that instead of
ing of chemisorption phenomena is rather imagining the particles confined to lattice
complex and requires knowledge on the sites, one may suppose that they move
geometrical structure of the system, adsor- continuously in space divided into ‘cells,’
bate binding and charge transfer, the elec- but that their interactions are determined
tronic structure of adsorbate and substrate, solely by which particular cells are occu-
pied’’ [52].) The configurational energy of
as well as vibrational frequencies [63].
the adlayer on a (L × L) square lattice is
Three dominant types of interactions
given, as an example, by the following
may occur between chemisorbed species:
Grand Canonical Hamiltonian [38, 56, 57,
dipole–dipole (direct and screened by the
63]
electrolyte), electron–electron (indirect via
substrate electrons or direct at short dis- 2

L
tances), and elastic (via substrate ions) [31, H=− εij si sj − µ̄ si (2)
32, 65, 66]. i,j ij
Adsorption and phase formation is
treated in three complementary regimes: where i and j denote sites on the lattice,
the thermodynamic approach is applied to si is the occupation state of site i, which

derive macroscopic properties of a system is either 0 (empty) or 1 (occupied). ij
at equilibrium; principles of quantum me- is the sum over all pairs of sites on
chanics are used to develop a microscopic the lattice, and εij is the respective pair
understanding; and statistical mechanics interaction energy (three-body and higher
establishes the connection between macro- interactions require additional terms). The
scopic and microscopic quantities, and geometry of the lattice may vary with
relates the two previous approaches [63]. the substrate surface crystallography. The
Intriguing aspects of both physisorption surface coverage is given by [38, 63]
and chemisorption at electrode–electrolyte
θ = N −1 si (3)
interfaces are the various phases, which
i
may exist at the surface, and the transi-
tions between them. The relations between with N as the total number of surface
coverage θ (with θ referred either to the sites on the lattice. The electrochemical
potential µ̄ is the intensive variable, which is likely to undergo a series of struc-

controls the coverage. In the weak solution tural modifications before the maximum
approximation, it is related to the electrode packing density of the first monolayer is
potential E by attained. Phase diagrams of systems with
lateral repulsion are more diverse (cf. Sect.
c
µ̄ = µ̄o + RT ln − FγE (4) 3.3.4.1). For smaller degrees of coverage,
co one may expect the formation of 2D lattices
with µ̄o and co as suitably chosen reference with large interatomic spacing, and more
states, and γ as the electrosorption valency. compact lattices may appear in a fairly
In the limit of εij → 0 and θ → 0, complex sequence. An example is bromide
Eq. (2–4) yields the Henry isotherm, and on Au(100)-(1 × 1) [51]. A gaseous-like
εij → 0 combined with finite values of θ bromide adlayer undergoes a first-order
results in the Langmuir isotherm [5, 60, phase transition to form a commensurate
√ √
63]. c( 2 × 2 2)R45◦ structure. Upon further
The accurate theoretical description of increasing the coverage, an aligned, uni-
√
the lateral interactions in the adsorbed axially incommensurate c( 2 × p)R45◦
layer requires that the complete substrate- phase is formed (see section on Ad-
adlayer-solution system be treated. In the layer phases and phase transitions on
simplest case, one considers only pair fcc (100) surfaces). The commensu-
interactions between nearest-neighbor ad- rate/uniaxially incommensurate transition
sorbed species. In general, attractive (C/UIC) has been described theoretically
(short-range) interactions will induce first- by Frenkel–Kontorova [68] and Frank–van
order phase transitions with typical fea- der Merwe [69] on the basis of a model
tures such as the existence of a critical of harmonically bound atoms in a sinu-
temperature Tcrit , phase coexistence at soidal corrugation potential. The compres-
T < Tcrit , and island formation controlled sion may be either uniform or associated
by nucleation and growth. If the inter- with the formation of localized regions of
actions are repulsive (long range), then high density (domain walls or solitons).
continuous phase transitions may oc- Pokrovsky and Talapov [70] found that
cur [63, 67]. In case of lateral attractive the incommensurability is proportional
interactions, the adsorbed species occupy to both the reduced temperature and the
preferentially nearest-neighbor sites even (electro-)chemical potential with the same
at small concentrations Typical attractive power law exponent β = 0.5. This theo-
interactions are π – stacking, hydrogen retical prediction is also supported by the
bonding, hydrophobic forces, and so on C/UIC–transition of bromide and chloride
(cf. Sect. 3.3.4.3). If the minimum of the adlayers on Au(100) in an electrochemical
lateral interactions is located at a dis- environment [51].
tance larger than the spacing between Further compression of the uniaxial
the nearest adsorption sites, the initially incommensurate (UI) phase may result
growing 2D phase will not match the in (1) an incommensurate phase in both
densest packing of the monolayer. After directions or (2) an incommensurate ro-
the former phase covers the whole sur- tated phase, because shear waves typi-
face, further deposition of the adsorbate cally cost less energy than compression
(for instance at higher electrode poten- waves if the misfit is large enough [71,
tials) results in a denser film, which 72]. The deregistry transition results from
the competition between the depth and of nearest neighbors, while treating the
width of substrate potential wells, and other interactions, as well as interactions
the dimension and magnitude of the lat- of nearest neighbors with their neighbors
eral interaction energy of the adsorbed in the MFA. Compared to the MFA
system. (or BWA) and the QCA, a further
The consideration of interactions be- advancement, especially at T < Tcrit , is
tween the adsorbed particles is a sig- represented by the low-temperature series
nificant improvement of the Langmuir expansion (LTSE) [60, 73, 77–80]. This
approach. In the simplest case, one takes approach not only considers equilibrium
into account only pair interactions between adsorbate clusters up to a certain size,
nearest neighbors, which may be attrac- but also various configurations of a
tive or repulsive, respectively. With the cluster with constant size, for example,
assumption of randomly distributed ad- the lattice geometry is taken explicitly
species, which are not correlated, one ob- into account. Exact isotherms within the
tains the Bragg–Williams Approximation lattice gas treatment may also be obtained
(BWA) [60, 63, 73–76]. The interactions by MC simulation [62], usually based on
of the test particle are considered in the an MC algorithm in a Grand Canonical
mean field of all other particles being pro- ensemble [38, 56, 57].
portional to the coverage θ multiplied by The detailed discussion and application
the total number of particles in the lat- of the aforementioned isotherms will be
tice (mean-field approximation or MFA). given in the paragraph on phase transi-
Combination with Eq. (4) leads directly to tions in organic monolayers (Sect. 3.3.4.3).
Frumkin-type equations. It involves the
statistical distribution of adspecies among 3.3.3
energetically uniform sites and neglects Kinetics of 2D Phase Formation and
local fluctuations and correlations. The Dissolution
loop in the isotherm below the critical
temperature Tcrit is related to an artificial 3.3.3.1 Diffusion, Adsorption,
system forced to be homogeneous under Autocatalysis
all conditions. Especially, the coexistence The formation of 2D ordered, steady state
of the two phases with different densi- adlayers at electrified interfaces is the
ties at the equilibrium phase transition result of several complex, nonequilibrium
under conditions of short-range attractive processes. In the simplest case, this
interactions is excluded. The models based sequence involves mass transfer of the
on the BWA are more reliable if the lat- molecules/ions from the bulk electrolyte
eral interactions can be neglected, or if towards the surface, adsorption and/or
the range of interactions is large [52]. The charge transfer, and association at the
MFA breaks down severely under con- electrode surface [15, 16] (Fig. 1).
ditions in which phase transitions occur
and ordered adlayers form [34–36, 54, 56, Mass transport controlled by diffusion The
60, 73]. limiting case of nonfaradaic adlayer for-
Isotherms based on the so-called mation controlled exclusively by diffusion
quasi-chemical approximation (QCA) are as r.d.s. was discussed by Koryta [81] and
a significant improvement over the Delahay [82]. The time dependence of the
MFA [77, 78]. This gives an exact treatment surface excess , in case of semiinfinite
Fig. 1 Schematic
representation of the interfacial
equilibria involved in the A
formation of ordered A
monolayers.
A ad A A
A
A ad A
Γ cx = 0 δc /δx c
x
III II I
linear diffusion, is obtained from Fick’s The former yields, in the limit of a
first law according to negligible rate of desorption,
t
∂c θ = 1 − exp(−kad cx=0 t) (7)
(t) = D dt (5)
0 ∂x x=0
where kad is the adsorption rate.
with x as the distance from the plane This theory was extended by Schuh-
of adsorption at the electrode. If the mann to an adsorption process obeying
rate of adsorption is sufficiently high, so a Frumkin isotherm [91]. The same treat-
that the subsurface concentration cx=0 ment is also valid for interfacial faradaic
is zero during the phase formation, one reactions [90, 92].
obtains [82, 83] Several groups derived approximate
expressions for the coupling of diffu-

2c Dt sion/adsorption [85, 87, 88, 93–96] or
θ= = (6) diffusion/charge transfer-controlled pro-
m m π
cesses [2, 92, 97] either in the time or in
m is the maximum surface excess the frequency domain.
of the condensed phase. Equation (6)
is formally equivalent to the well- Autocatalytically controlled surface reactions
known Cottrell equation for faradaic pro- Michailik and coworkers [98] proposed an
cesses [84]. Diffusion-controlled adsorp- autocatalytic surface process representing
tion in the frequency domain was treated the time dependence of formation of a 2D
theoretically and experimentally by Me- condensed film of triphenylethyl phospho-
lik–Gaikazyan [85, 86], Lorenz [87], and nium sulfate ((TPEP)2 SO4 ) deposited at a
Armstrong [88]. mercury–electrolyte interface [99]:
∂θ
Adsorption control The kinetics of the ad- = ka (1 − θ)l θ p (8)
sorption step in the absence of mass trans- ∂t
port control was treated by Lorenz [87] l and p are integers (l, p ≥ 0), with l as the
and Delahay [89, 90] assuming either a reaction order with respect to the initial
Langmuir- or a Temkin-type formalism. state and p the order with respect to the
products. The analytical rate equation θ(t), Frenkel [112]. The total Gibbs energy re-
derived with l = 2 and p = 1, was shown quired to create a 2D cluster by deposition
to describe quantitatively the deposition of of N particles (ions, molecules) from solu-
(TPEP)2 SO4 onto a mercury electrode [98]. tion is:
Lorenz [100] combined diffusion, intrin-
sic adsorption, and surface association GT (N ) = GP (N ) + GPB (N ) (9)
as separate contributions and developed
The first (favorable) term represents the
the first model, which attributes a slow
free energy gain due to the formation of the
(dynamic) adsorption step to the 2D asso-
new phase. The second (unfavorable) term
ciation of adsorbed species at an electro-
accounts for the creation of a new phase
chemical interface.
boundary. GT (N ) reaches its maximum
The above-mentioned approaches suffer
for a critical particle number Nc . Clusters
from a conceptual difficulty assuming that
with a smaller number of particles than Nc
the adsorbed monomers are distributed
will tend to dissolve, while larger clusters
homogeneously at the electrode. This is
will grow further. The relation (9) is valid
not unambiguously valid in case of 2D or-
for any arbitrary geometrical form of the
dered ML (cf. experimental section 3.3.4).
cluster [113]. The special case of a circular
On the basis of i–t transients for 1 M
patch of radius r and height h shall be
pyridine in 1 M NaOH, Armstrong [101] considered in more detail.
showed that the formation of the adsorbate Phase formation controlled by a charge
film proceeds according to a 2D nucle- transfer reaction Men+ + ne− ↔ M across
ation and growth mechanism, analogous the interface, such as 2D electrocrystal-
to faradaic processes such as electrocrys- lization [114, 115], UPD [2, 22, 116], and
tallization and the deposition of anodic 2D anodic passivation [117, 118] gives the
films [17]. This approach accounts for the following expression for GT
inhomogeneous distribution of adspecies
at the electrode–electrolyte interface, and πr 2 h
shall be discussed in the following sections GT (r) = neo η + 2πrε (10)
v̄
in more detail.
v̄ is the specific particle volume related
3.3.3.2 Nucleation and Growth to the number N of deposited particles
Mechanisms by N = (πr 2 h)/v̄. η = (E − Erev ), with
3.3.3.2.1 Nucleation Erev being the equilibrium potential, is
Homogeneous nucleation – concept of the the overvoltage and represents the driving
critical cluster A metal surface that is force of the faradaic process. ε is the line
homogeneously flat, like mercury or quasi- tension, which has to be overcome by
perfect silver single crystals [2], offers no the expansion of the nucleus [118]. Metal
specific adsorption sites. In analogy to 3D deposition can occur only if η is negative,
phenomena, the nucleation of a new 2D so that the Gibbs energy GT (r) of a
phase may be treated on the basis of cluster as a function of its size r first rises,
the classical nucleation theory (CNT) as reaches a maximum and then decreases
formulated by Gibbs [102] and further de- (Fig. 2). The critical radius rc and the
veloped by Volmer [103, 104], Farkas [105], critical free energy GTc (r) are given by
Stranski and Kaishev [106–109], Becker εv
rc = − (11a)
and Döring [110], Zeldovich [111], and neo η
r Fig. 2 Dependence of the free energy

h change of formation of a disc-shaped
cluster GT (r) on radius r for a charge
transfer-controlled process.
∆G T(r ) 2πr ε and concentration according to

E E
γ # = qf dE − qm dE (13)
Erev Erev
∆G Tc(r )
with Erev being the equilibrium poten-
tial at which γ # = 0 [3, 119]. γ # is
a nonlinear function of the applied elec-
rc r trode potential. The free energy GT c (r)
and radius rc of the critical cluster are
derived as
πε2
GT c = − (14a)
γ #
π r 2hneoη/v and
ε
rc = − (14b)
γ #
and The formation of the critical nuclei occurs

πhε2 v by thermal fluctuations and may be consid-
GTc = − (11b) ered as an activation process [103]. Becker
neo η
and Döring treated the homogeneous nu-
cleation process as a series of consecutive
GTc is proportional to 1/η.
bimolecular reactions [110]
The formation and transformation of
electroinactive 2D condensed organic mono- α1
layers adsorbed at the electrode–electrolyte M1 + M1 ←−→ M2 . . . Mi + M1
β2
interface are predominantly controlled by
αi
electrocapillary forces and yield [16], instead ←−→ Mi+1 + M1 ←−→ . . . (15)
βi + 1
of Eq. (11):
in which the growth or decay of clusters
GT (r) = πr 2 h γ # + 2πrε (12) proceed by attachment or detachment of
monomers M1 to an i-mer to form an
γ # represents the supersaturation given (i + 1) mer, respectively. The interfacial
by the difference in surface energy per concentration of monomers is kept con-
unit area between the final (f ) and the stant by rapid exchange with the adjacent
metastable (m) states [3, 119, 120]. The bulk phase. The rate-determining step is
numerical values of the driving force assumed to be the incorporation of one
are obtained by integration of the charge additional monomer to the critical cluster
density curves of the two distinct phases f Mnc . The solution of the set of steady state
and m at constant pressure, temperature, equations based on Eq. (15) then leads to
the steady state nucleation rate J as: according to
J = Zα ∗ cnc α1
S + M1 ←−→ S
βs1
− GT c
= Zα ∗ c1 exp (16) − M1 + M1 . . . S − Mi−1
kT
αi
+ M1 ←−→ S − Mi + · · · (17)
where α ∗ is the rate at which the βsi
monomers are added to the critical clus-
ter of nc atoms, ions, or molecules; Instead of the steady state of (quasi)
cnc is the equilibrium concentration of homogeneous nucleation, during which
critical clusters, and Z is the nonequi- nuclei appear at constant rate, both the
librium Zeldovich factor [111, 121]. The total number of critical nuclei and the
equilibrium concentration of critical clus- duration of the nucleation process are
ters is related to the concentration of confined by the maximum number of
monomers, c1 , by the Boltzmann ex- preexisting sites No . If there is a uniform
pression with the energy barrier given probability with time of converting these
by the free energy of formation of the sites into critical nuclei S − Mnc , one
critical nucleus. The period required to obtains the following first-order nucleation
reach a steady state of critical nuclei at law as an approximate solution of the
constant supersaturation, for example, a reaction sequence (17)
stable distribution of clusters of the new
phase controlled by ‘‘monomer’’ transport N (t) = No [1 − exp(−J t)] (18)
in the ambient phase and incorporation
of building units of clusters, is identi- The quantity J is the steady state
fied as initial delay or induction period nucleation rate per active site and may be
t [122–127]. Toshev and coworkers re- expressed analogous to J in Eq. (16) [17,
ported that t in homogeneous nucleation 122, 123]. Two limiting cases of the
is usually longer than for heterogeneous exponential law of nucleation, as given
nucleation [128]. by Eq. (18), are of particular importance:
(1) J 1 : instantaneous nucleation

Heterogeneous nucleation on active sites
Real surfaces, for instance obtained with N (t) = No , (19)
melt-grown gold, silver, or platinum
single-crystal electrodes, exhibit various Within the timescale considered, all avail-
sites and structural imperfections of dif- able sites are converted to critical nuclei
ferent dimensionality, such as kinks, va- immediately.
cancies, monatomic steps, reconstructed
(2) J
1 : progressive nucleation
surface domains, 2D islands, and holes [2,
129, 130]. Defining the number of ac- N (t) = J No t (20)
tive sites S under particular experimental
conditions as No , one may consider the where the number of critical nuclei
heterogeneous nucleation process as the increases linearly with time and N
No .
successive incorporation of monomers The steady state rate of 2D nucleation is
M1 into site-confined clusters S − Mi strongly dependent on the dimensionality
of the active sites. Staikov and cowork- on lose their physical meaning and the
ers reported that, at constant supersatu- use of atomic forces of interactions be-
ration, J (terrace) < J (step) < J (kink) [2, comes more reasonable. The atomistic
130]. Active sites represent not only surface approach for the calculation of the nu-
imperfections, but may also be gener- cleation rate on supersaturation was first
ated by surface oxidation/reduction, reintroduced by Walton [140] and then de-
construction processes or ad/desorption veloped to a general nucleation theory by
of anions or solvent molecules [131, 132]. Stoyanov and Milchev [141, 142]. These
Fleischmann and coworkers [17, 133] authors considered the excess free energy
and Retter [134] pointed out that the of creation of a new ‘‘surface’’ (second
combination of active intermediates to term in Eq. (9)) by the difference be-
form a critical nucleus at a specific site may tween the individual binding energies of
lead to the power law of n–step nucleation N monomers (atoms, molecules) in the
bulk, N 1/2 , and the dissociation ener-
N (t) = J No t n (21) gies of an N –mer cluster by including the
binding energies of the monomers in the
J represents the modified nucle-
cluster as well
as the interaction with the
ation rate.
substrate, i . Substrate-induced strain
In case of homogeneous nucleation,
is accounted by the introduction of a
No may be attributed to the number of
strain energy per monomer, εM , which
available lattice sites [135]. The nucleation
yields
mechanisms described earlier assume that
the surface concentration of monomers re-

mains constant throughout the transition. GPB (N ) = N 1/2 − i + εM N
The formation of 2D condensed films of 5- N
bromocytosine [136, 137] on mercury and (22)
anodic adenine-mercury complexes [138] Applications of this atomistic approach to
are examples of the so-called ‘‘truncated’’ metal deposition on foreign substrates are
nucleation. This mechanism is based on summarized in [2]. Buess–Herman [143]
the condition that the formation of critical pointed out that, in contrast to charge
nuclei and their competition with growth transfer–induced 2D adlayers, critical nu-
and/or ad/desorption processes tend to clei of surfactant molecules are quite large,
decrease the available monomer concen- often being composed of more than 100
tration with time. At a certain moment, molecules. In these cases, the CNT is still
the nucleation process ceases. valid.
Atomistic approach – small cluster model of 3.3.3.2.2 Growth The expansion of a su-
nucleation With increasing supersatura- percritical nucleus through continued in-
tion the number of monomers constituting corporation of monomers is called growth.
the critical cluster reduces up to a few If the resulting new phase is isotropic,
atoms or molecules, and in some partic- as for instance in the case of a cir-
ular cases of nucleation on active sites, cular island, a single, time-independent
this number was found to be close to growth rate, kG , suffices to characterize
zero [2, 139]. Macroscopic quantities, such the process. For anisotropic growth, there
as surface, surface free energies, and so will be several growth rates [2]. For 2D
growth, expansion of the clusters is pos- coefficient [145, 146]. The corresponding
sible only at its periphery. Referring to coverage function is given by
the elementary process involved on an ide-
ally smooth surface, one may distinguish
∂θ 2πhA2 Dρ
two main mechanisms: (1) growth con- = = K1 (27)
trolled by the rate of monomer incorporation ∂t MSE m
and (2) mass transport (surface diffusion)-
controlled growth [17, 118]. When more than one critical nucleus is
Assuming that the incorporation is the formed, competition between the grow-
rate–determining step, the number N (t) ing centers for the limited areas SE takes
of monomers belonging to an isolated place. Growth ceases at the point where
circular cluster of height h obeys the these patches merge, domain bound-
equation aries are created, coalescence may often
be ruled out [147]. Computer simulations
∂N (t)
= 2πkG r(t)h (23) predict for the 2D growth controlled
∂t
by direct incorporation of monomers
with (Eden model), rather compact patches
kG
r(t) = t (24) with fractal dimensions close to 2 [148].
ρ
In situ scanning tunneling microscopy
ρ is the number of particles per unit (STM) experiments with hydrogen-bonded
area [17]. Normalization to the real sur- 2D condensed films of purine bases
face area SE and to the maximum on defect-free HOPG-surfaces confirmed
surface excess of the condensed phase these findings [149]. In the diffusion-
m yields the fractional coverage of this limited aggregation model, proposed by
phase as Witten and Sander, the resulting clus-
2M ter has an open structure with fractal
∂θ 2πhkG
= t = K1 t (25) dimensions of 1.66 [150]. More complex
∂t ρSE m mechanisms involve anisotropic and/or
The first in situ atomic-scale visualization multistep growth, induction periods of
of this growth mechanism was recently growth [117, 133, 151–154], as well as
demonstrated for the incorporation of defect-mediated processes [2, 17]. The case
chloride ions at kink positions of the of 1D growth was treated quantitatively
c(2 × 2) chloride adlayer on Cu(100) in in the formation of surface hemimicelles
0.01 M HCl [144]. of amphiphils [155]. Lorenz and cowork-
If the rate of advance of the growing ers pointed out that type and density of
circular center is controlled by symmetrical surface defects, such as monatomic steps
hemicylindrical (surface) diffusion about on single-crystal surfaces, influence dra-
an axis perpendicular to the 2D nucleus matically the steady state structures as
one obtains well as the kinetics of 2D phase forma-
√ tion processes. They demonstrated, for
r(t) = A Dt (26)
instance, that barrierless nucleation and
where A is a constant depending on 2D growth on an equidistantly spaced
potential, molecular weight, and density stepped surface predicts rectangular i–t
of the monomer, and D is the diffusion transients [2].
3.3.3.2.3 Coupling of Nucleation and Mononucleation Mononucleation events

Growth The rate of nucleation J and at electrochemical interfaces have been
the growth rate kG are both functions observed in potential step experiments
of the applied supersaturation (e.g. by monitoring current, charge, or ca-
electrode potential [2] or difference in pacitance as a function of time under
surface energy [3], respectively). Three conditions of (1) low supersaturation and
basic regimes may be distinguished: (2) small, defect-free electrode surfaces.
If the formation period of the critical These experiments offer the unique op-
cluster is much longer than the time portunity to determine simultaneously
required to cover the available electrode absolute rates of the nucleation and growth
surface by its subsequent growth, the processes. The formation of the critical
transformation involves just one nucleus, nucleus appears stochastically in time and
and is called mononucleation. If the space (Fig. 3). The time distribution (con-
rate of nucleation is much faster than tinuous) of the birth events of critical
the subsequent growth process, many nuclei (discrete), as indicated by the on-
nuclei contribute to the creation of set of the subsequent growth transient
the new phase. This regime is known (current/charge increase, decrease of in-
as polynucleation. The intermediate case, terfacial capacitance), within a sufficient
which involves just a few nuclei, is named long train of repetitive transient experi-
oligonucleation. ments under identical conditions, leads
40
(a) (b) (c)
30
[nA]
20
−i
10
0
0 10 20 30 40 50 60
t
(a) [s]
Fig. 3 (a) Current–time curves
associated with the formation of single
monolayers on a rectangular
quasi-perfect Ag(100) electrode in 6 M 4
AgNO3 at 45 ◦ C and an overvoltage of
2 mV, as reported by Bostanov and 3
−ln Po
coworkers in Ref. [154]. The inserts

show possible nucleation sites. (b) Time 2
dependence of − ln(Po ) during the
growth of a screw dislocation-free (100) 1
face of a silver single crystal at 6.0 mV.
The solid line is drawn with 48
12 24 36
JSE = 0.0864 nucleation per second
(reprinted with permission from t
Ref. [159], copyright 1982 by Elsevier). (b) [s]
directly to the steady state nucleation rate 6 mV is shown in Fig. 3(b). The nucleation
J . The statistical analysis is based on the rate is strongly potential-dependent with
Poisson distribution: ln J ∼ 1/η (cf. Eqs. (11 and 16)) [156–159].
The shape of the mononucleation tran-
µN exp(−µ) sient is directly related to the absolute rate
PN = (28)
N! of growth, since no overlapping of nu-
where clei takes place. Figure 3(a) illustrates the
isotropic growth of an expanding circular
µ = J SE t; J in (m−2 s−1 ) (29) silver nucleus. Growth stops at the edges
of the rectangular substrate surface. The
PN represents the probability of having growth rate constant was found to be of
N nuclei formed at a time t, where t the order of 1 cm s−1 V−1 [156–159]. The
is a continuous variable, whereas N is carefully chosen geometry of the silver
discrete: N = 0, 1, 2, and so on [123]. template provides, in addition, direct ac-
Usually one evaluates the probability of cess to the location of the initially formed
observing no nucleation event, Po , dur- critical nucleus.
ing a given time interval t. These data The growth of a 2D condensed organic
are experimentally accessible and rather film created by mononucleation on a
unambiguous [123, 156–164]. According hemispherical mercury electrode is often
to Eq. (25), ln Po should be directly pro- expressed in terms of a growing spher-
portional to t, with the proportionality ical disk by expansion of its periphery
constant J (Fig. 3b). (Fig. 4a) [160, 164]:
Typical examples of mononucleation are
∂q kG (t − to )
the deposition of silver on electrolytically i= = 2πrqm kG sin
∂t r
grown Ag(100) crystals (Fig. 3) [156–159],
and the formation of 2D condensed or- (30)
ganic films, such as isoquinoline [160], where r denotes the radius of the electrode,
thymine [163], or coumarin [162–164] on qm the monolayer charge, kG the growth
mercury electrodes. These experiments rate, and to the onset of growth. Figure 4(b)
demonstrate that the rate of nucleation shows a set of experimental and simulated
is more strongly dependent on the ap- transients for coumarin on mercury [164].
plied supersaturation than the growth The stochastic nature of the experiment
rate. Extreme care has to be taken in is represented by the scatter of the
mononucleation experiments to avoid arti- mononucleation transients along the time
facts from edges or defect contributions of axis. The shape of the i–t traces is identical
the substrate or catalytic effects of impuri- because of the use of a small, spherical
ties [161, 162]. (mercury) electrode.
Figure 3 illustrates, as just a small
section of a train, three individual i–t tran- Polynucleation At high supersaturation,
sients for the deposition of a single silver many nuclei are created independently.
monolayer on a dislocation-free Ag(100) The completion of the transition proceeds
surface having a rectangular cross section. by subsequent growth of a large number
The formation of the critical nuclei is in- of nuclei. The resulting transients, for
dicated by arrows. The corresponding plot instance triggered by a potential step, are
of −lnPo (t) versus t for an overvoltage of no longer stochastic, but deterministic,
1.5
1.0
[µA]
0.5
i
0
−0.5
0 0.5 1.0 1.5 2.0 2.5
t
(a) (b) [s]
Fig. 4 (a) Scheme illustrating the growth of a single nucleus
according to constant radial growth. (b) Stochastic current–time
transients obtained for 0.5 M MaF + 5 mM coumarin at a hanging
mercury drop electrode (HMDE) following a single potential step
from E1 = −0.400 V to E2 = −0.520 V. The lines were drawn using
Eq. (30) with kG /r = 17.5 s−1 , qm = 108 nC, and appropriate values
for to (reprinted from Ref. [164], copyright 1995 by Bulgarian
Chemical Society).
and do not depend on the shape and This statistical law makes it possible to
area of the electrode surface. In the initial correlate single cluster behavior to that
stages of formation of the new phase, of a collection of many clusters and,
the individual centers can be assumed to alternatively, to derive nucleation and
grow rather independently of each other. growth laws from observations involving
The extended fractional coverage θx , which many overlapping clusters. The validity of
represents the hypothetical coverage of the Eq. (32) requires a random distribution of
unhindered growth of individual nuclei, a large number of clusters, in which, each
is obtained by solving the convolution must be small with respect to the total
integral [17, 134, 165] available electrode area.
The combination of Eqs. (31 and 32)
∂θx ∂N with the exponential law of nucleation
= f1 (t − x) dx (31)
∂t ∂t t=x (Eq. (18)), and growth of circular 2D clus-
ters by direct incorporation of monomers
with ∂N /∂t given by the respective
(Eq. (25)) yields, as an example [172, 173]
law of nucleation (cf. Eq. (18–21)) and
f1 (t − x) representing the growth (cf.
2t
Eq. (25, 27)). The extended coverage θx θ = 1 − exp −2K1 No t 2 −
J
and the ‘‘true’’ coverage θ, as obtained

2
in reality by internuclear collision and + 2 [1 − exp(−J t)] (33)
overlap of the growing centers at an ad- J
vanced stage of the transition, are related
and the limiting cases
within the framework of the Avrami theo-
rem [166–171] θ = 1 − exp(−2K1 No J t 2 )
θ = 1 − exp(−θx ) (32) for J −−−→ ∞ (33a)

2K1 No J 3 The same strategy was applied in the
θ = 1 − exp − t derivation of rate equations for n-step nu-
3
cleation according to a power law (cf.
for J −−−→ 0 (33b) Eq. (21)) [133, 134], the combination of
nucleation laws with anisotropic growth
for instantaneous or progressive one-step regimes [153], as well as truncated nucle-
nucleation [3, 17, 118]. ation due to time-dependent concentration
Examples of the kinetic mechanisms gradients of monomers [136]. MC simu-
represented by Eq. (33) and its lim- lations verified that the Avrami theorem
iting cases are the 2D formation of is valid for instantaneous [184], progres-
organic and ionic monolayers of cy- sive [185], and n-step nucleation according
tosine [174], 5-bromocytosine [173], iso- to a power law [184–187].
quinoline [175], coumarin [164], guani- The above-mentioned expressions of
dinium nitrate [176], for chloride [177] polynucleation and growth processes, as
and hydroxide [178] and several other so- triggered by single potential step experi-
called anodic films on mercury [179, 180], ments E1 → E2 , may be represented by
as well as the transition between the the general Avrami equation [15, 16, 188]
√ √
( 3 × 3) and the (1 × 1) Cu–UPD layer
on Au(111) [181]. θ = 1 − exp(−bf t m ) (35)
In case of surface diffusion-controlled bf is a constant incorporating both rates
growth (Eq. (27)), one obtains accordingly of nucleation, J , and growth, kw ; m is re-
lated to the sum of the dimension and the
θ = 1 − exp − 2K1 No time exponent in the nucleation law. The
values of m provide a set of diagnostic cri-

1 teria to determine possible mechanisms
× t − [1 − exp(−J t)] (34) of the nucleation and growth processes
J
involved (Table 1). Unfortunately, there ex-
and ists no unique relation between m, bf ,
and the actual kinetic mechanism on the
θ = 1 − exp(−2K1 No t) basis of single potential step experiments.
However, in carefully designed secondary
for J −−−→ ∞ (34a) growth double potential step experiments
θ = 1 − exp(−K1 No J t 2 ) (E1 → E2 → E3 ), one may separate nucle-
ation and growth exploiting their different
for J −−−→ 0 (34b) dependence on supersaturation [156, 157,
189, 190]. The rate of nucleation is a much
Phase formation kinetics based on sur- steeper function of potential than the rate
face diffusion-controlled growth processes of growth. The program sequence starts
have been suggested for 2D monolay- with a potential pulse E1 → E2 of large
ers of camphor-10-sulfonate on mer- amplitude and short duration τ to create
cury [182], physisorbed uracil films on a fixed number of critical nuclei, while
Au(hkl) [183], as well as the commensu- ensuring their insignificant growth. The
√
rate/incommensurate transition c( 2 × second step, E2 → E3 , is chosen such
√ √
2 2)R45◦ → c(p × 2 2)R45◦ of bromide that (1) further nucleation is unlikely and
on Au(100) [51]. (2) only preformed nuclei start to grow
Tab. 1 Diagnostic criteria to determine possible mechanisms of

nucleation and growth based on the general Avrami equation (35)
Exponent Nucleation Growth Example
1 I SD 154
1<m<2 EL SD 51, 154, 182, 183
2 I DI 173, 181
P SD
2<m<3 EL DI 173, 174
3 P DI 16, 164, 175, 176
m>3 PL DI or SD 134
Note: Abbreviations: I (P) instantaneous (progressive) nucleation;

EL: exponential law of nucleation; PL: power law of nucleation; SD:
surface diffusion; DI: direct incorporation.
Fig. 5 (a) Schematic charge,

qM qM (b) capacitance, and (c) current
q∞
transients resulting from a single
qm qm potential step from E1 to E2 . Panel (c)
qo also shows the regime of a secondary
growth double potential step
q dl q dl experiment E1 → E2 → E3 .
0 t
(a) E1 Ec E2 The formation of 2D ordered mono-

layers at electrochemical interfaces may
C C be triggered by an appropriate change of
Co
monomer concentration, temperature, or
potential [16]. Potential step experiments
are preferred for experimental simplic-
C∞ ity. The kinetic analysis of nonfaradaic
reactions is often based on capacitance
(b) E1 Ec E2 0 t or charge density transients. The surface
coverage is obtained according to
i i Co − C(t) q(t) − qo
θ(t) = = (36)
E1 E2 Co − C∞ q∞ − qo
t where Co , C∞ and qo , q∞ are the values
Hysteresis
E3 of the differential capacitance and charge
density before and after the transition
(c) 0 t (Fig. 5). In case of monolayer formation
involving faradaic processes (anodic films,
UPD, electrocrystallization), the current
(Fig. 5). Thus, nucleation and growth are density is preferred. Numerical integration
separately controlled by the parameters of yields
the prepulse (E2 , τ ) and of the secondary 1
θ(t) = j dt (37)
pulse to E3 , respectively. qm
qm is the charge of a complete monolayer. with time-resolved structure-sensitive ex-

Differentiation of Eq. (37) represents the periments is desirable. New perspectives
strategy to transform the model equations along these strategies developed also with
for the time dependence of the coverage the advent of time-resolved SEIRAS [199],
θ (cf. Eqs. (33, 34)) into the corresponding surface X-ray scattering (SXS), [200] and
current density functions. Typical expres- in situ scanning probe techniques [201].
sions are summarized in [17, 118, 146, 165, Some examples will be described in the
172, 191]. application sections.
Transients controlled by polynucleation
and 2D growth exhibit typically sigmoidal 3.3.3.3 Dissolution of 2D Condensed
shape of the capacitance and charge den- Monolayers
sity as a function of time (nonfaradaic pro- The dissolution of 2D condensed mono-
cess) or a bell-shaped current density re- layers represents a disorder/order tran-
sponse (faradaic process), provided that the sition of a high-coverage phase into a
interference with bulk diffusion and strong low-coverage phase. The process may be
coupling with ad/desorption of monomers triggered by applying a potential and/or
is negligible (Fig. 5, Sect. 3.3.4). The role temperature perturbation [15, 16, 202].
of bulk diffusion in 2D phase forma- Potential step experiments are often pre-
tion kinetics was analyzed by Lorenz [2], ferred at electrochemical interfaces be-
Bosco [192], and Guidelli and cowork- cause of experimental simplicity. The
ers [193]. Rangarajan [165] and Pohlman resulting transient is usually faster than
[194] explored the consequences of the the respective adlayer formation [3]. The
strong coupling between adsorption and overall dissolution process includes the
nucleation/growth processes. following elementary steps (cf. Fig. 1):
With the exception of the elegant (1) disintegration or 2D melting of the or-
mononucleation transients on silver depo- dered patches on the electrode surface,
sition at defect–free silver electrodes [115, (2) surface diffusion and desorption of
156–159], the analysis of the macroscopic monomers, (iii) bulk diffusion into the
current, charge density, and/or capaci- electrolyte.
tance transients does not provide direct ac- Assuming interfacial equilibrium be-
cess to structural information and molec- tween the condensed phase and the ad-
ular/atomistic mechanisms of 2D phase sorbed noncondensed monomer species
formation. Employing dynamic MC sim- on the surface and within the Helmholtz
ulations of microscopic models, Rikvold layer, de Levie [83] described the dissolu-
and coworkers pointed out that mean-field tion transients of an ordered camphor film
rate equations, such as the ones based on at a stationary mercury electrode in very
the Avrami ansatz, are limited especially dilute solution (10 µM) by semiinfinite pla-
in the later stage of the overall transi- nar diffusion according to
tion [195]. For systems in which ordered

phases are involved, the microscopic c − c(x = 0) 2Dt
adlayer structure and the dynamic details θdiss = 1 − θ =
m π
of the adsorption, phase formation, and (38)
lateral diffusion processes should become θ as the fraction of the electrode covered by
important [196–198]. The combination of the condensed layer with m as (constant)
time-resolved dynamical MC simulations surface excess of this phase and c(x = 0) as
the interfacial concentration of monomers the disaggregation process on (1) the

immediately adjacent to the electrode. dissolution overvoltage and (2) on the
Under conditions of fast surface melting history of the preceding 2D condensation.
of the 2D condensed layer and negligible Dissolution transients of ordered films
desorption of monomers into the outer formed according to polynucleation and
Helmholtz region, Lorenz [87] and Ran- incorporation-controlled growth indicate
garajan [165] suggested an ansatz based slower kinetics with increasing ageing
on the Langmuir isotherm time at the same potential (Fig. 6). This
observation is attributed to annealing
θdiss = 1 − θ = 1 − exp(−kd t) (39) processes of adlayer defects [164, 206,
207]. Dissolution transients of ordered
This model predicts exponential capaci- monolayers created from a single nucleus on
tance, charge density, or current versus defect-free substrates could be modeled by
time curves, which have been, for instance, an inverse growth mechanism. The shape
reported for the dissolution of ordered of these curves deviates significantly from
thymine [203] and uracil [204] films on liq- that of a single exponential [163].
uid electrodes. Another example, based on Faradaic admittance experiments on
rather similar assumptions, is the elec- a mercury surface blocked by tribenzy-
trodesorption of the first ML of anodically lammine [208] and in situ STM experi-
formed HgS [205]. ments with physisorbed hydrogen-bonded
Both models do not explain the monolayers of guanine and adenine on
experimentally observed dependence of HOPG [209, 210] revealed an average
0.15
1 2 3 4 5
0.10
[Fm−2]
C
0.05
0
0 0.5 1.0 1.5 2.0 2.5
t
[s]
Fig. 6 Capacitance transients for 0.5 M aqueous NaF + 0.5 mM
coumarin following a single potential step from E1 = −1.000 V
(inside the capacitance pit III) to E2 = −0.4265 V at different
waiting (aging) times to at E1 : (1) 2.5s, (2) 5.0s, (3) 10s, (4) 20s,
(5) 50s. The experimental transients could be well represented by
Eq. (40) in combination with Eq. (36) (reprinted from Ref. [164],
copyright 1995 by Bulgarian Chemical Society).
diameter of inhomogeneously distributed ad/desorption of foreign species such as

2D-ordered patches on these defect-free solvent molecules and ions of the support-
substrates of 1 µm to 3 µm. The domain ing electrolyte may trigger the creation
size of organic, ionic or metal ML on real and expansion of circular holes, which fi-
single crystals is significantly smaller, and, nally cause the complete dissolution of the
depending on the defect density of the 2D adlayer [199]. The mathematical treat-
substrate and the substrate–adsorbate in- ment of this mechanism is isomorphic to
teractions, may range between 20 nm up the equations derived for polynucleation
400 nm [20, 23]. and growth (cf. Sect. 3.3.3.2), and predicts
Referring to the above-mentioned exper- for the dissolution of 2D solidlike films
imental observations, Buess–Herman and sigmoidal q M –t and C–t transients, or
Badialli suggested a defect-mediated disor- i–t curves exhibiting a characteristic max-
dering process [3, 143, 207]: A 2D-ordered imum. Models based on the nucleation
ML, formed from a large number of nu- and spreading of holes have been applied
clei, is composed of polygonal patches to the reduction and dissolution of an-
separated by domain boundaries, which odic films on mercury electrodes [205, 212,
represent a lattice arrangement of line 213], the dielectric breakdown of bilayer
and point defects [207, 211]. These bound- membranes [214, 215], and to the transfor-
√ √
aries comprise sites, where the activation mation of the ( 3 × 3) Cu–UPD phase
free energy of the phase transformation on Au(111) [181]. Recently, hole nucleation
is probably lower than that required to and growth was also considered to describe
form a hole inside the patch. In the two the cathodic desorption of alkylthioles
limiting cases, dissolution either starts on Au(111) [216–218] and the disordering
from point defects or lines (Fig. 7). Point of uracil [219], cytosine [201], and uri-
defects resulting from intersection lines, dine [220, 221] monolayers chemisorbed
substrate defects (vacancies, atomic disor- on Au(hkl) (paragraph 3.3.4.3.5). Exper-
der, chemical impurities, etc.) [2], or the imental i–t transients have been fitted
(a)
Fig. 7 Simplified scheme of

disordering of a
two-dimensional condensed
film starting from point (a) or
line defects (b). The ordered
phase is represented by the
(b) hatched area.
to expressions combining instantaneous of the disordered phase θdiss increases,

or hole nucleation according to an ex- and the remaining adlayer fraction to be
ponential law combined with linear or transformed reduces, which adds a positive
surface diffusion-controlled growth [181, contribution to γ #. Assuming a squared
212, 213, 216–221]. Employing stepped array (average length l) of domain bound-
gold electrodes with (111)-oriented ter- aries of the initial 2D-ordered phase leads
races of different width (26, 10, 6, or 4 to the following time dependence of θdiss
atoms, (110)-oriented steps) van Krieken
and coworkers concluded that an increas- θdiss = 1 − θ = 1
ing step density enhances the nucleation of  2
holes within an ordered uridine layer, and  α 
simultaneously slows down their subse- −
 

2αvd (0)
quent growth rate [220]. Derivations from 1 − (1 − α) exp t
(α − 1)l
the classical Avrami-type equations were
attributed to the strong coupling of the (40)
phase dissolution with (bulk) diffusion with
ε
or partial desorption [165]. Recent in situ α= (41)
l γ #
STM experiments with physisorbed gua-
nine ML on HOPG [210] and chemisorbed and vd (0) being the rate of disorder
2,2 -bipyridine on Au(111) [222] demon- propagation at t = 0. This model can
strated clear evidence for hole nucleation be adapted to any geometrical shape of
and growth controlled disordering of a 2D domain boundaries [207]. If ε and γ # are
condensed phase triggered by the applied known, the value of the parameter α allows
electrode potential and/or temperature (cf. the determination of L, which provides an
Sect. 3.3.4.3). estimation of the initial average size of the
Disordering starting from line defects or condensed phase.
grain boundaries within a 2D adlayer was Equation (40) was successfully applied to
first described quantitatively by Mulder, represent the disordering of physisorbed
who considered the progressive shrinkage films on mercury and Au(hkl) elec-
of circular patches with constant rate [211]. trodes, such as coumarin, thymine, and
The model predicts a first-order melting uracil [207, 225]. The experimental tran-
transition [223, 224]. Taking the differ- sients also support the theoretically pre-
ence of interfacial energy γ # between the dicted trends with ageing or healing out
growing (dilute) regions, the dissolving of domain boundaries during the polariza-
2D condensed phase, and the line ten- tion of the high-coverage phase (Fig. 6). In
sion ε between the two competing phases situ STM experiments of the potential-
into account, Badiali and coworkers ex- induced disordering of a cytosine ML
tended this approach by introducing a on Au(111) indicate that dissolution ini-
time-dependent propagation rate of the tiated by point and line defects may
disordered patches [207]. The physical rea- occur even simultaneously [201]. Rather
son of this approach is based on the little direct experimental evidence exists
argument that dissolution leads to a pro- at present at electrode–electrolyte in-
gressive disappearance of line defects and terfaces for the dissolution of solidlike
consequently to a decrease of the line en- films by subsequent loss of long-range
ergy ε · l. In parallel, the surface coverage positional and orientation order, caused
by the unbinding of dislocations and example, the following sequence of specif-

discommensurations as predicted by the ically adsorbed anions was found on Hg
KTHNY-theory [226–228]. Similarly, the and Au(111) on the basis of the Gibbs ener-
interactions between adlayer and substrate gies of adsorption and solvation [229–232,
surface as well as the role of substrate de- 234, 241]:
fects, such as vacancies, steps, and holes
on the disordering of adsorbed ML at F− < ClO4 − < SO4 2− < Cl−
electrified interfaces have not yet been con- < Br− < I− (42)
sidered in detail, neither experimentally
nor theoretically. The combination of classical electrochem-
ical measurements with ex situ transfer
3.3.4 experiments into UHV [242], and in situ
Examples structure-sensitive studies such as elec-
troreflectance [25], Raman and infrared
3.3.4.1 Phase Transitions in Anionic (IR)-spectroscopies [29, 243], and more
Adlayers recently STM and SXS [39] provided de-
tailed knowledge on energetic, electronic
3.3.4.1.1 Introduction Anions have a and structural aspects of (ordered) anion
strong tendency to adsorb specifically at adsorption and phase formation. These
metal surfaces, for example, to establish experimental studies have been comple-
a direct bond with the electrode by par- mented by various theoretical approaches:
tial loss of their hydration shell. As a (1) quantum-chemical model calculations
consequence of the contact with the elec- to explore substrate–adsorbate interac-
trode, the ionic character of the anions is tions [244–246]; (2) computer simulation
markedly reduced, resulting in a higher techniques to analyze the ion and sol-
surface concentration than in case of non- vent distribution near the interface [247];
specific adsorption. This effect was first (3) statistical models [67]; and (4) MC sim-
observed in double-layer studies on mer- ulations [38] to describe phase transitions
cury [229, 230] and later confirmed and in anionic adlayers.
studied in detail on single-crystal solid elec- Specifically adsorbed anions can also
trodes [231–234]. Specifically adsorbed an- significantly affect the stability and elec-
ions can form various types of ordered trochemical reactivity of metal electrodes.
structures, either more open (cf. sulfate Examples are the alteration of the potential
on Au(hkl) [235, 236]) or close-packed as (charge) distribution in the double layer
reported for halides on different solid elec- known as Frumkin effect [248], the lifting
trodes [21]. Cyclic current-potential curves of the surface reconstruction of Au(hkl)
often reveal sharp current peaks, indica- electrodes [249], the electrochemical depo-
tive of phase transitions within the anionic sition and dissolution of metals in the
adlayers and hence of the existence of or- presence of anionic ligands [250], and the
dered phases [21, 237]. Thermodynamic role of halides and sulfate on the oxygen
data of specific anion adsorption was reduction on Pt and Au [251].
obtained in surface tension studies (on The following chapter is focused on the
mercury only! [229, 238–240]), capacitance structure and phase behavior of specif-
measurements [231–233], cyclic voltam- ically adsorbed anion adlayers on solid
metry, and chronocoulometry [234]. As an single-crystal metal electrodes. On the
Tab. 2 Adlayer structures of halides on Au(hkl) and Ag(hkl), adapted from Ref. [21]
Electrode Electrolyte Structure Coverage Technique Reference
Au(111) Cl Aligned-hex 0.508–0.5277 SXS 253, 254

Br Rotated-hex
√ √ 0.462–0.515 SXS/STM 254
I ( 3 ×√ 3)R30◦ 1/3 STM/LEED 255, 258
c(p × 3) 0.33–0.41
rotated-hex 0.41–0.45 SXS
√ √
Au(100) Cl, Br c(√2 × 2 2)R45◦ 0.50 SXS, STM 51, 259, 261
c( 2 ×√p)R45◦ 0.50–0.62
I c(p × 2 2)R45◦ 0.46–0.49 STM 257, 263
rotated-hex
√ √ SXS
c( 2 × 2 2)R45◦ 0.50
Au(110) Br (1 × 3) 2/3 STM 262
(1 × 4) 3/4
rotated 3/4
pseudohex
c(2 × p)
I c(2 × p) 0.39–0.416 SXS 260
Ag(111) Cl Aligned-hex 0.53 STM 266
Br √× 7)√
(7 0.51 SXS 265
I ( 3 ×√ 3)R30◦ 1/3 SXS 265
c(p × 3) 0.33–0.38 STM/LEED 267
rotated-hex 0.41
Ag(100) Cl, Br, I c(2 × 2) 0.50 SXS 53, 55, 268
Ag(110) Cl, Br c(p × 2) 0.72–0.76 SXS 263
basis of selected examples with halide given in Ref. [21]. Selected results of these
or oxyanion–containing adlayers, it will studies, as obtained by in situ STM and/or
be demonstrated that the observed struc- SXS for ordered halide adlayers on Au(hkl)
tures and 2D phase transitions can be and Ag(hkl), are summarized in Table 2.
understood as a result of competing adsor- Before the detailed discussion of specific
bate–adsorbate and adsorbate–substrate examples, several general trends shall be
interactions. pointed out.
Ordered halide adlayers are only ob-
3.3.4.1.2 Phase Formation in Halide served above a critical potential Ec , which
Adlayers often represents rather high anion cover-
ages. The potentials of adlayer formation
In Situ Results on Adlayer Structures For and phase transitions between different
Cl− , Br− , and I− ordered structures have ordered adlayers shift ∼60 mV per decade
been observed in in situ studies on single- of halide concentration. The potential
crystal electrodes of the face centered cubic ranges of the ordered phases depend
(fcc) metals Au [51, 252–264], Ag [53, 55, strongly on Epzc and the strength of
251, 265–269], Cu [270–272], Pt [23, 251, metal–halide interactions. The ordered
273, 274], [275–277], Pd [278], Rh [279], phases are stabilized by repulsive ad-
and Ni [280]. A comprehensive review is sorbate–adsorbate interactions. Halides
form predominantly close-packed hexag- The chemisorbed halide (or halogen)

onal adlayers on the (111) surfaces of adlayers, in particular those on coinage
fcc metals reflecting isotropic adsor- metal surfaces, exhibit a structural phase
bate–adsorbate interactions [259, 268]. At behavior, which closely resembles that
saturation coverages the adlayer spacings of classical 2D systems, such as ph-
are typically slightly above the halide ysisorbed noble gases on graphite and
van der Waals diameter [254]. For halide metal surfaces [33, 285]. An advantage of
adlayers on coinage metal surfaces such the electrochemical environment, as com-
as Au(111), Ag(111), and Cu(111) often pared to the solid–vacuum interface, is
incommensurate structures with lattice that the adlayers are close to equilibrium.
spacings, which decrease with increas- Since coverage and adlayer structure can
ing potential were found (electrocompres- be directly controlled via the electrode
sion) [51, 254, 255, 259]. On the (111) potential, metal electrodes in halide so-
surfaces of transition metals well-ordered lutions are interesting model systems for
fundamental studies of phase transitions
adlayers exhibit preferentially commen-
in two dimensions. Most phase transi-
surate structures indicating a stronger
tions in ordered adlayers of halides on
preference for energetically favorable ad-
smooth (111) and (100) surfaces are first
sorption sites [273–277]. Halide adlayers
order, often manifested as sharp peaks
on fcc (100) surfaces exhibit a strong
in the cyclic voltammogram or double
trend towards commensurate structures, layer capacitance, and discontinuities in
with a predominant occurrence of sim- the adsorption isotherms. They involve
ple c(2 × 2) adlattices. This reflects the phase coexistence and proceed via nucle-
higher substrate corrugation potential of ation and growth mechanisms [51]. Phase
the more open, square (100) lattice [259]. transitions, which can be described by
Incommensurate phases are only ob- continuous changes of the order parame-
served on Au(100), which represent similar ter, were reported for two electrochemical
trends as observed on (111) surfaces [51, systems on (100) surfaces: the disorder-
259, 268]. Only few studies of halides order transition for Ag(100)/Br− [53, 55,
exist on fcc (110) surfaces, which often 265] and the commensurate – UI transi-
report pseudohexagonal adlayers, where tions for Au(100)/Br− [51, 259].
the anions reside in the atomic rails
along the (11̄0)-direction. An exception is Adlayer Phases and Phase Transitions on fcc
Au(110)/I− , where a rotated incommen- (111) Surfaces The structure of adsorbed
surate phase was observed [257, 262, 264]. monolayers, in particular the registry be-
The structure and phase behavior of halide tween the adlayer and the substrate, is af-
adlayers at the electrochemical interface is fected by the symmetry of the substrate, the
very similar to the adlayers formed by dis- adsorbate–adsorbate interactions, and the
sociative desorption in the gas phase in the corrugations in the adsorbate–substrate
high-coverage regime [281, 282]. Differ- interaction potential [259]. The later is
ences are found only at low coverages [283, rather weak on the (111) faces of fcc met-
284]. These differences can be related to als, and of comparable magnitude to the
solvation effects at the electrochemical in- lateral interactions between the halide ad-
terfaces, which are most pronounced in sorbates in high-density adlayers. If the
the low-coverage regime. average lattice constants of adlayer and
substrate are different, the resulting phase The iodide species are arranged in a
is a compromise between a simple com- centered rectangular unit cell with sides
√
mensurate structure, which minimizes the 3 and p, where p decreases with in-
interface energy (determined by the cor- creasing potential. The adsorbate coverage
rugation in the adsorption energy), and changes from about 0.355 to 0.40 in
a uniformly compressed incommensurate a 0.56 V wide potential range. At the
√
phase, which is favored by the elastic lat- most positive extreme of the c(p × 3)
eral interactions in the adlayer. Frenkel phase, the structure corresponds to the
√
and Kontorova [68] and Frank and van c(5 × 3) high-order commensurate unit
der Merwe [69] showed the lowest energy cell with a nearest-neighbor iodide spac-
state corresponds to a system of coming of 0.439 nm. Qualitatively, the SXS
mensurate regions separated by domain data show two discontinuous changes in
walls (also called solitons or misfit dis- coverage, indicating first-order phase tran-
√ √ √
locations). When the domain walls are sitions for ( 3 × 3)R30◦ ↔ c(p × 3)
√
diffuse, such that the adsorbed species are and c(p × 3) ↔ rotated-(hex) (Fig. 8).
equally spaced along the incommensurate The phase behavior of Ag(111)/I− is com-
direction, the structure is referred to as pared with that of iodide on an Au(111)
a uniformly compressed incommensurate surface, which exhibits an almost identical
phase. For these phases and dislocation substrate geometry. Both incommensu-
√
structures, the displacement of the adsor- rate phases, the c(p × 3) and the ro-
bates from the commensurate positions tated – (hex) phase, were observed by in
can result in a long-range vertical modula- situ STM, as well as SXS [252, 255, 258].
√ √
tion (Moire-pattern) of the adlayer, which The commensurate ( 3 × 3)R30◦ phase
is often observed directly by STM [252, 253, was only found in a limited potential range
258, 261, 267]. by STM and ex situ LEED studies [258],
A whole sequence of adlayer structures which might suggest that this structure
is illustrated by the system Ag(111)/I− , exists only as a short-range ordered phase.
where ordered phases are observed in The potential-dependent coverage of the
a wide range of potentials and surface incommensurate phases on Au(111) is also
coverages due to the strong adsorption plotted in Fig. 8 (filled circles). For com-
and high polarizability of iodide [259, 267]. parison, the potential scale for the data on
Starting at negative potentials in which Au(111) was shifted by −0.50 V to account
the halide layer is disordered, a commen- for the difference in Epzc of Ag(111) and
√ √ √
surate ( 3 × 3)R30◦ , an UI c(p × 3), Au(111) [259].
and an incommensurate rotated hexago- The phase behavior of iodide on Ag(111)
nal structure were found with increasing and Au(111) is an elegant example of
potential, independently in SXS- and STM the stepwise loss of commensuration with
studies [259, 267] (Fig. 8). The coverage increasing potential accompanied by two
θ increases from 1/3 in the commensu- first-order phase transitions. This phase
rate phase with iodide species residing sequence has been predicted by Bak and
in threefold hollow sites to 0.442 at the coworkers [286] using Landau theory and
√
most positive potentials. The c(p × 3) Kadar [287] using a generalized three-state
structure is obtained by uniaxial com- Potts model.
√ √
pression of the ( 3 × 3)R30◦ adlattice The orientation of the adlayers in
along the (11̄0) direction of the substrate. the incommensurate hexagonal phases
(a1) (b1) (c1) Fig. 8 (a) The structures of iodide on

Ag(111) and Au(111) deduced from
( 3 × 3)R30° c(p × 3) Rotated-hex X-ray scattering results. They
correspond
√ √ to (1) the commensurate √
( 3 × 3)R30◦ , (2) the UI c(p × 3),
and (3) the rotated hexagonal phase.
(b) Corresponding in situ STM images
of iodide on Au(111) (from Ref. [258]).
(a)
(c) The potential-dependent coverages,
θ, of iodide, determined from the
in-plane diffraction, are shown versus
the applied potential for Ag(111) in
0.1 M NaI (open symbols) and for
Au(111) in 0.1 M KI (filled circles). The
(b) potential scale of the Au(111) data has
been shifted negatively by 0.50 V to
facilitate the comparison with the
0.44 Ag(111) data (reprinted with permission
from Ref. [265], copyright 1996 by
Ag(111) Elsevier Science Ltd.).
Au(111)
0.42 C
0.40
θ
0.38 B
0.36
A
0.34
−1.0 −0.8 −0.6 −0.4 −0.2

EAg/AgCl
(c) [V]
exhibits a complex behavior. The adlayers a rotation from a high symmetry,

√ √
are rotated relative to the ( 3 × 3)R30◦ commensurate orientation is energetically
o
direction by an angle φ < 5 , which varies preferred if the transverse strain in the
continuously with adlayer density, 1◦ to adlayer is sufficiently lower than the
2.5◦ for Ag(111)/I− [259] and 0.5◦ to longitudinal strain. The total energy is
4.0◦ for Au(111)/I− [255]. The epitaxial minimized at a nonzero angle because
rotation is a well-known phenomenon of the presence of periodic lateral
in heteroepitaxial growth [285], and was distortions of the adsorbate lattice (static
first studied theoretically by Novaco and distortion waves). The Novaco–McTague
McTague [71, 72]. They showed that theory predicts an analytical relationship
for an infinite incommensurate adlayer, between the rotation angle φ and
√
the incommensurability ε = 3θ in the can occupy the preferred hollow sites,
rotated (hex) phase, and provides a minimizes the interfacial energy between
reasonable description of the experimental halide adlayer and metal, the resulting
data for iodide on Ag(111) and square adlattice is energetically unfavor-
Au(111) [255, 259]. able in terms of adsorbate–adsorbate
The continuous variation of the ad- interactions. An instructive example for
layer coverage in the UI or incommen- illustrating the competing contributions
surate phase with potential is called of lateral interactions and corrugation
electrocompression [255]. Assuming an potential are bromide adlayers on the struc-
electrosorption valency γ = −1, the lat- turally isomorphic Ag(100) and Au(100)
eral compression κ2D of the adlayer can surfaces.
be calculated via κ2D = (γ eo )−1 (δA/δE),
where A = −1 is the area per adsorbed Ag(100)/Br Recently, Wandlowski and
iodide species [255, 288]. The diffraction coworkers demonstrated in a combined
data yield halide adlayer compressibil- electrochemical and in situ SXS study
ity’s in the range 1.8 to 6.9 Å2 eV−1 and the existence of a low-coverage lattice
4.47 to 9.40 9 Å2 eV−1 for the UIC and gas phase for bromide on Ag(100), and
the IC phases, respectively [21]. This is an Ising-type transition into the c(2 × 2)
higher than the compressibility of elec- phase [53, 55]. Figure 9 shows the capac-
trochemically deposited metallic monolay- itance curve and the potential-dependent
ers (κ2D = 1–2 Å2 eV−1 ) [288], but signifi- SXS intensity at (1/2, 1/2, 0.12), which
cantly lower than that of physisorbed noble is a first order diffraction peak of the
gas adlayers (κ2D = 1 − 2 Å2 eV−1 ) [285]. c(2 × 2) phase, and the intensity at (1,
The compressibility of metal and noble 0, 0.12) representing the occupation of
gas adlayers can be explained by the the fourfold hollow sites of the substrate.
free electron model. The latter is not The position of the sharp peak P2 in the
applicable to quantitatively describe the capacitance curve at −0.80 V (vs. SCE) cor-
electrocompression of halides on Au(111). relates well with the onset of the c(2 × 2)
Taking into account substrate-mediated, long-range order. For E > −0.80 V, the
indirect interactions, electrostatic inter- intensity at (1/2, 1/2, 0.1), which is pro-
actions, Lennard–Jones interactions, in- portional to the order parameter squared,
teractions between induced dipole mo- increases according to (E − Ec )2β with the
critical exponent β = 0.125 in agreement
ments, and three-body interactions, Wang
with the prediction of the 2D Ising model.
and coworkers calculated θ(E) in rea-
The same power law was also obtained
sonable agreement with the experimental
under UHV–conditions (LEED), in which
data [289].
the chemical potential was obtained from
the coverage [58]. The intensity at (1, 0,
Adlayer Phases and Phase Transitions on 0.12) slowly decreases between −1.25 V
fcc (100) Surfaces With the exception of and −0.60 V, for example, in the potential
Au(100) the only stable low-order com- range of the broad capacitance peak P1 . No
mensurate phase studied on (100) metal discontinuity is observed around the dis-
surfaces is the c(2 × 2) structure [53, 55, order/order phase transition (P2 ), which
268–270, 278]. Although the commensu- indicates a continuous change in cover-
rate c(2 × 2) phase, where all adsorbates age, in support of a second-order phase
250 Fig. 9 (a) Capacitance versus potential

curves for Ag(100) in (0.05–x) M
200 P2 KClO4 + x M KBr; x represents the KBr
P1 concentrations: 1–0, 2–10−2 M. Scan
[µF cm−2]
150 rate 10 mV s−1 , 18 Hz and 10 mV

2 peak-to-peak ac-amplitude. (b) Potential
C
100 dependence of the normalized

scattering intensity of Ag(100) in 0.04 M
50 1 KClO4 + 0.01 M KBr at (1/2, 1/2, 0.12)
(◦, reflection of the c(2 × 2) adlayer)
0 and at (1, 0, 0.12) (dashed line,
(a) −1.4 −1.2 −1.0 −0.8 −0.6 −0.4
Ag(100)-(1 × 1) surface rod). (c) The
1.2 real space structures of the lattice gas
configuration (left) and the
1.0
commensurate c(2 × 2) bromide
(1, 0, 0.12)
0.8 adlayer (right) (reprinted with
Intensity
permission from Ref. [55], copyright

0.6 Ag(100) 2001 by Elsevier Science Ltd.).
(1/2, 1/2, 0.12)
0.4
0.2
0.0
−0.2
−1.4 −1.2 −1.0 −0.8 −0.6 −0.4
ESCE
(b) [V]
(c)
transition. Comparison of the bromide simulations assuming lattice gas adsorp-

coverage in hollow sites, θsxs , obtained tion with a pure hard square model [290].
from the SXS experiments with the to- On the basis of an MFA, Ocko and
tal coverage, θtot , measured by chrono- coworkers estimated a repulsive inter-
coulometry, revealed good agreement at action energy of 0.110 eV between bro-
the highest potentials, but lower val- mide adsorbates on next-nearest neighbor
ues of θsxs at intermediate coverages, sites [268]. Koper [56] and Mitchell and
which is attributed to a significant dis- coworkers [57, 291] performed MC sim-
placement of the adsorbates from hollow ulations for a lattice gas model (LGM).
sites at low coverages [55] (Fig. 10a). The These authors found from fits to the
disorder-order transition occurs at θsxs = Ag(100)/Br− adsorption isotherms that the
0.25 and θtot = 0.35. The latter is close to lateral interactions are quite adequately de-
the critical coverage of 0.368 obtained from scribed with a nearest neighbor excluded
Fig. 10 (a) Potential dependence of the

normalized bromide coverage obtained 0.5
II
chronocoulometrically (θtot , ) or by
analyzing the scattering intensity in (1, 0.4
0, 0.1) position (θSXS , ◦) for Ag(100) in
Coverage
0.04 M KClO4 + 0.01 M KBr. (b) θtot 0.3
(referenced to the Ag(100)-(1 × 1) Chrono (θ tot)
0.2 I
lattice) for three KBr concentrations θSXS
in (0.05 − x) M KClO4 on Ag(100). The
solid lines represent the 0.1
Monte-Carlo-simulation isotherms as
calculated by Rikvold and coworkers in 0
Ref. [291] (reprinted with permission −1.4 −1.2 −1.0 −0.8 −0.6 −0.4
from Ref. [55], copyright 2001 by
Elsevier Science Ltd.). ESCE
(a) [V]
0.5
0.4 10 mM
1 mM
Coverage
0.3 0.1 mM
0.2
0.1
0
−1.4 −1.2 −1.0 −0.8 −0.6 −0.4
ESCE
(b) [V]
volume interaction and a (dominating) the negative one. Comparison with in

long-range dipole–dipole repulsion. The situ SXS experiments revealed that the
simulations quantitatively reproduced the peak P1 represents the transition be-
electrochemical data and provided an ex- tween the hexagonal (reconstructed) and
cellent description of the critical properties the quadratic (unreconstructed) arrange-
of this continuous phase transition [57, ment of the gold atoms in the top layer
291] (Fig. 10b). Koper also pointed out that of the substrate [51]. No other diffrac-
the QCA is a more appropriate analyti- tion peaks were found at E < P2 . At the
cal description of the experimental θ − E voltammetric peak P2 , a first-order phase
isotherm than the mean-field treatment transition between gaseous-like, randomly
(MFA or Frumkin isotherm) [56]. adsorbed bromide and a commensurate
√ √
c( 2 × 2 2)R45◦ halide adlayer (coverage
Au(100)/Br Figure 11 illustrates a typical θ = 0.5), which is stable in the poten-
current versus potential curve of bromide tial region between P2 and P3 , takes
on Au(100) [51, 259, 261, 292]. Three char- place [51]. The rectangular unit cell con-
acteristic current peaks are observed in tains two bromide species with nearest
the positive scan direction, and two in and next-nearest neighbor distances of
Au(100)-(hex) Au(100)-(1 × 1) Fig. 11 (a) First current versus

potential scan of a freshly
c ( 2 × 2 2) c ( 2 × p ) flame-annealed Au(100) electrode,
10 immersed at −0.80 V in 0.05 M NaBr,
scan rate 10 mV s−1 . The stability
I II III IV regions of the various substrate surface
(Au) and adlayer structures (Br) are
5 indicated. The transitions are labeled
[mA cm−2]
P1 P2 P3 P1 , P2 /P2 , and P3 /P3 , and the stability

regions are named I, II, III, and IV.
j
(b) Potential dependence of the

0 bromide adlayer coverage as obtained
by√SXS for√the C/UIC transition√ between
c( 2 × 2 2)R45◦ and c( 2 × p)R45◦ .
P2′ P3′
The open (filled) circles represent the
−5 negative (positive) going potential scan.
−1.0 −0.6 −0.2 0.2 0.6 1.0 The solid line was calculated with a
ESCE power law 0.122(E − Ec )0.40 + 0.5,
(a) [V] where Ec is the concentration-dependent
thermodynamically defined transition
0.57 potential, as indicated in the figure
(reprinted with permission from
Ref. [51], copyright 1996 by American
0.55 Chemical Society).
c ( 2 × p)
θSXS
0.53
Negative
c ( 2 × 2 2)
0.51 Positive
Ec
0.49
0.30 0.40 0.50 0.60
ESCE
(b) [V]
4.08 Å and 4.56 Å, respectively. The crys- Ag(100)/Br− [53]), suggests that the elastic
tallographic analysis of the X-ray peak interactions between the Br− adsorbates,
intensities suggests a registry of the which favor hexagonal packing, are more
√ √
c( 2 × 2 2)R45◦ adlayer with all adsor- significant than the adsorbate–substrate
bates residing in bridge sites, rather than interaction energy difference between the
a mixed occupation of top and hollow two phases. The corrugation potential for
sites [259]. This result is also supported bromide on Au(100) appears to be weaker
by an in situ STM study of Cuesta and than on Ag(100). Similar trends were
coworkers [261], in which no differences found in halide adlayers on Au(111) and
in the apparent height of the halide Ag(111) [21], as well as in the corrugation
√ √
adsorbate in the c( 2 × 2 2)R45◦ unit potentials extracted from the calculations
cell were observed. The preference for the of Ignaczak [245].
√ √ √ √
c( 2 × 2 2)R45◦ phase, with respect to The commensurate c( 2 × 2 2)R45◦
the higher coordinated, and commonly on phase transforms at E > 0.40 V into
√
fcc(100) surfaces found c(2 × 2) phase (cf. a UI c( 2 × p)R45◦ adlayer, where
√
the incommensurability ε = 2 2/(p − 1) This behavior is consistent with a cu-
√
continuously varies from 0 (commensurate mulative disorder in the c( 2 × p)R45◦
phase) to 0.12. The nearest-neighbor spac- phase [259]. The results for bromide ad-
√
ing in the c( 2 × p)R45◦ structure is al- sorption on Au(100) under electrochemical
√
ways the commensurate spacing aAu 2 = control indicate that continuous C → UIC
4.078 Å, the next-nearest-neighbor spacing transitions are not restricted to the vacuum
decreases from 4.56 Å in the commensu- environment.
rate phase to 4.13 Å resulting in an almost The kinetics of the phase transitions
undistorted hexagonal lattice at the most of bromide adlayers on Au(100)-(1 × 1)
positive potentials [51, 259]. As the lat- were studied by chronocoulometric poten-
tice compresses the bromide species slide tial step experiments [51]. Current tran-
√
along the ‘‘rails’’ determined by the under- sients indicate for the disorder ↔ c( 2 ×
√
lying gold atoms. The adlayer compression 2 2)R45◦ transitions in both directions
is similar to that in the rotated hexagonal an instantaneous (hole) nucleation pro-
phase on Au(111) [254]. cess coupled with a Langmuir-type ad-
The continuous change of the bromide sorption/desorption mechanism. Rising
coverage with potential upon the transi- transients, which could be modeled with
√ √ √
tion c( 2 × 2 2)R45◦ → c( 2 × p)R45◦ the exponential law of nucleation and 1D or
can be approximated by the power law θ = surface diffusion-controlled growth, were
√ √
0.122 (E − Ec )β + 0.50 with β = 0.40 and obtained for the c( 2 × 2 2)R45◦ →
√ ◦
Ec = 0.375 V. The measured exponent is c( 2 × p)R45 transition when stepping
smaller than the theoretical value β = 0.50 into the potential regime of high incom-
predicted by Pokrovsky and Talapov for mensurability, (the uniformly compressed
the C → UIC transition within a model adlayer phase). The signal-to-noise ratio
of noninteracting domain walls [70]. The in the range of the weakly incommen-
theoretical prediction is only supposed surate phase was too high, and hence
to apply close to the transition, where no direct conclusion on the ‘‘continu-
the domain walls are narrow relative to ous’’ character of this phase transition
their separation. The experimental pre- around Ec (as predicted from the above-
cision does not permit to quantitatively mentioned SXS results) from transient
extract the exponent close to the transi- experiments was accessible. The dissolu-
√
tion, in which the theoretical prediction tion of the UIC ( 2 × p)R45◦ phase is
could be unambiguously tested. At higher characterized by monotonous i–t curves,
which may be related to the consider-
incommensurabilities the bromide cover- √ √
age is not only determined by the proximity able disorder of the c( 2 × 2 2)R45◦
to the C/UIC phase transition, but rather when formed from the incommensurate
by the compressibility of the bromide ad- phase [51].
layer, which may explain the deviations
in the exponent. The width of the X-ray 3.3.4.1.3 Phase Transitions in Adlay-
diffraction (XRD) peaks in the incommen- ers of Oxoanions The specific ad-
surate direction scale quadratically with sorption of oxoanions, such as sul-
the wave vector component in this di- fate/bisulfate, hydroxide, or phosphate
rection, and increase continuously by an considerably affects the shape of voltam-
order of magnitude as the C/UIC tran- metric and capacitance curves [21, 23,
sition is approached from the UIC side. 231–233, 293]. The presence of sharp
current spikes, reported, for instance, have focused on sulfate or bisulfate

for Au(111) or Pt(111)/H2 SO4 [237, 294, adsorbed on noble metal single-crystal
295] and HPO4 2− [296, 297] was at- electrodes, and in particular on the
tributed to first-order phase transition Au(111) surface [235, 298–309]. The ad-
involving specifically adsorbed anions. sorption behavior of sulfate species
Most structural studies with oxoanions on Au(111) was investigated by cyclic
Au(111)-( 3 × p ) Au(111)-(1 × 1)
Disordered ( 3 × 7)
3
I II III
2
P1 P2
[µA cm−2]
1
j
−1
P2′
−2
−0.4 −0.2 0 0.2 0.4 0.6 0.8 1.0
ESCE
(a) [V]
(1 × 1) ( 3 × 7)
(b) 110 nm (c) 10 nm

Fig. 12 (a) Cyclic voltammogram range for Au(111) in 0.05 m
H2 SO4 , scan rate 10 mV s−1 . The stability of the various substrate
(Au) and adlayer structures (HSO4 − or SO4 2− ) is indicated. (b)
√ In
situ STM image of the thermally reconstructed Au(111)-(p × 3)
surface at −0.20 V. (c) First-order phase transition between
randomly
√ √ adsorbed sulfate species and the ordered
( 3 × 7)19.1◦ phase on Au(111)-(1 × 1) during a potential scan
from E1 = 0.750 V to E2 = 0.820 V. The scanning direction is from
top to bottom. The arrow indicates the phase transition upon
passing P2 (reprinted from Ref. [299], copyright 1997 by VCH
Verlagsgesellschaft mbH Weinheim).
voltammetry [237, 310], chronocoulometry The macroscopic kinetics of the disor-

and radiochemistry [311], microgravime- der/order phase transition, giving rise to
√ √
try [312], IR-spectroscopy [298, 313–315], the ( 3 × 7)R19.1◦ adlayer on Au(111),
SHG [316], and in situ STM [235, 298, 299, was studied by chronoamperometric po-
317] (Fig. 12). Three potential regions can tential step experiments [299]. The ob-
be distinguished in the steady state voltam- tained experimental current transients
mogram for Au(111) in 0.05 M H2 SO4 , could be represented by the exponen-
separated by well-defined current peaks. tial law of nucleation in combination
P1 represents the lifting of the recon- with surface diffusion-controlled growth
struction of the gold surface [249], and (Fig. 13). The first-order nature of this
implies that region I can be assigned to transition is supported by a recent in
√ situ STM study [235]. Magnussen and
the reconstructed Au(111)-(p × 3) sur-
coworkers reported on the coexistence
face with minor adsorption of sulfate. √ √
Regions II and III represent the un- of fluctuating ( 3 × 7) adlayer islands
reconstructed Au(111)-(1 × 1) phase. At and patches displaying the bare Au(111)
potentials more positive than the sharp substrate on terraces at potentials just
positive of P2 . Domain boundaries were
current spikes P2 /P2 , which exhibit a
found to be mobile, and the shape of
small but distinct hysteresis, (bi)sulfate
the islands was changing rapidly. In situ
forms a highly ordered commensurate
√ √ SEIRAS studies employing the step-scan
( 3 × 7)R19.1◦ superstructure. Chrono-
technique might reveal further structural
coulometric and radiochemical experi-
details of the sulfate and water species
ments [311] yielded in this potential region
of this phase transition with a time res-
a maximum surface coverage of 0.2, which
olution of up to 5 µs [199]. Increasing
is significantly below the value expected
the step density of the substrate surface
for a close-packed sulfate adlayer. Infrared causes (1) the cumulative disorder of the
√ √
data revealed a C3ν symmetry, in which ( 3 × 7)R19.1◦ adlayer and (2) an in-
three oxygen atoms of the sulfate species crease of the nucleation rate [299] (Fig. 14).
interact with the substrate reflecting the No current spikes P2 /P2 were observed
symmetry match of the tetragonal an- at a critical miscut angle of 4.7◦ , corre-
ion with the trigonal gold surface [313, sponding to a terrace width of 2.9 nm.
314]. This assignment is in agreement This trend is supported by similar ob-
with recent calculations [318]. The same servations of ordered sulfate and phos-
ordered sulfate structure was found dur- phate structures on defect-rich Au(100)
ing in situ STM studies on (111) surfaces electrodes [309].
of Pt [300, 301], Rh [302], Cu [303–306],
Ir [307], and Pd [308]. Two kinds of max- 3.3.4.2Phase Transitions in
ima can be seen in the STM images of UPD – Adlayers
√ √
the ( 3 × 7)R19.1◦ structure, but only
the brighter one corresponds to the ad- 3.3.4.2.1 Introduction UPD is referred to
sorbed sulfate, the secondary maxima have as the deposition of a metal monolayer
been assigned either to coadsorbed wa- onto a foreign metal substrate positive
ter [302, 314] or hydronium ions [308, 309] of the respective bulk (Nernst) potential
to stabilize the oxoanion adlattice by direct Erev . The metal adatoms are bound more
hydrogen bridges. strongly to the foreign metal substrate than
1.6
log tmax
log imax
[ms]
[µA]
0.750 0.816
1.2
−0.4 1.6
0.812 1.4
−0.8 1.2
0.800 0.808 0.816
[µA]
0.8
i
ESCE
0.808 [V]
0.804
0.4
0.800
0.796
0
0 100 200 300 400 500
t
[ms]
Fig. 13 Current transients i(t) for Au(111), miscut < 0.5◦ , in 0.05 M H2 SO4
obtained after a single potential step from E1 = 0.75 V (region II) to various
final potentials in region III. The experimental traces are given as individual data
points, the solid lines represent theoretical curves calculated with the
parameters of the numerical fit to a model combining (a) an adsorption process
(Eq. 7) and (b) one-step nucleation according to an exponential law with
surface diffusion-controlled growth (Eq. 34), (reprinted from Ref. [299].
Copyright 1997 by VCH Verlagsgesellschaft mbH Weinheim).
to a substrate of its own kind, for example, deposit [116, 319]. This correlation is not
the chemical potentials of monolayer valid at single-crystal electrodes [320–322].
and bulk are different. In comparison New theories of UPD involving semiem-
to molecular adsorption, one observes a pirical models, DFT- and band structure
significant charge transfer (faradaic reaction calculations taking into account contri-
!) between the deposited metal ion and the butions of the electronic subsystem (re-
polarized electrode [116]. lated to work function differences) and
The UPD is often demonstrated in of the ionic cores of the adsorbate, have
cyclic voltammetry. The formation (dis- been developed to model the trends ob-
solution) of the first monolayer is seen served on single-crystal electrodes (cf.
by pronounced current peaks at E > Ref. [323]). These quantum statistical ap-
Erev , the bulk deposition occurs at E < proaches are complemented by statistical
Erev . For many polycrystalline surfaces, mechanical models [67] and (dynamic)
it has been shown that the differences Monte Carlo simulations [197] on UPD
in peak potentials for the oxidative dis- systems.
solution of metal monolayers and bulk The use of single-crystal electrodes
correlate linearly with the difference in in UPD–studies revealed pronounced
work function, , of substrate and structure specificity in metal monolayer
0.8
< 0.5°
0.6
0.750 V 0.804 V
[µA] 0.4
0.7°
i
3.0°
1.6°
0.2
4.7°
0
0 100 200 300 400 500
t
[ms]
Fig. 14 Typical set of i–t transients with E1 = 0.75 V/E2 = 0.804 V for
stepped Au(111) s crystals with various miscut angles (as indicated) in
0.05 M H2 SO4 . The solid lines represent calculated curves as obtained
with the parameters of the numerical fit to an adsorption-nucleation
model (cf. Fig. 13), (reprinted from Ref. [299], copyright 1997 by VCH
Verlagsgesellschaft mbH Weinheim).
formation as was shown in the early coadsorption onto the exact determination
work of Schultze [320–322], Lorenz [324], of coverage-potential isotherms was ex-
Bewick and coworkers [48–50], or plored by the twin-electrode thin-layer
Adzic [325]. Meanwhile a considerable technique [329] and studies with the quartz
amount of data is available, which crystal microbalance (QCM) [330]. Multi-
illustrate the role of electrode material step adsorption isotherm was taken as a
and surface crystallography, the influence first evidence for ordered UPD films [2,
of coadsorbed anions and/or solvent 22, 320–322]. The kinetics of adlayer
molecules, as well as intermetallic deposition/dissolution, as monitored in
electronic and ionic interactions on (1) chronoamperometric potential step ex-
the formation/dissolution and (2) on periments, was phenomenologically in-
the structural properties of these metal terpreted with homogeneous diffusion-
monolayers (cf. reviews in [2, 22, 116, adsorption models (cf. summary in [2])
326–328]). or as first-order phase transitions con-
UPD of metals has been extensively trolled by nucleation and growth [48–50].
studied with a variety of methods. Cyclic Optical properties of the metal mono-
voltammetry is frequently employed to layers were studied in the visible and
phenomenologically describe the phase be- near UV-range by electroreflectance spec-
havior in UPD systems and to determine troscopy [25] and second harmonic gen-
adsorption isotherms assuming full dis- eration [331]. The use of ex situ UHV-
charge of the metal ions [2, 116, 326–328]. techniques (LEED, reflection high-energy
The role of partial discharge and anion electron diffraction (RHEED), XPS, UPS,
Auger electron spectroscopy (AES)) af- step-scan SEIRAS in exploring kinetic

ter controlled emersion of the electrode mechanisms of metal deposition pro-
provided important direct information on cesses.
the structure, energetics, and composition The following sections are not meant to
of the various metal adlayers [242, 332, review the entire field of UPD. Emphasis
333]. Major breakthroughs in the determi- will be rather given to three model systems,
nation of adsorption sites, UPD adlayer Cu UPD on Au(hkl), Pt(hkl) and Pb UPD
structures, and their dynamics and re- on Ag(hkl) to describe, based on these
activity properties were achieved by the examples, typical structural aspects and
use of in situ SXS techniques [24, 39, kinetic processes of 2D phase formation
334], as well as in situ scanning probe with low-dimensional metallic adlayers on
microscopy such as STM and atomic foreign substrates.
force microscopy (AFM) [23, 335, 336].
Very recently, several structure-sensitive 3.3.4.2.2 Underpotential Deposition of
in situ techniques have been applied to Copper Ions on Au(hkl)
monitor the kinetics of UPD or dis-
solution. Finnefrock and coworkers [200] Cu–UPD on Au(111) in sulfuric acid – struc-
performed first SXS-measurements for Cu tural aspects The UPD of Cu on
UPD on Pt(111), and Ataka and cowork- Au(111) in sulfuric acid electrolyte
ers [199] demonstrated the usefulness of has been studied using (1) classical
20
A2
[µA cm−2]
10 B2
j
B1 A1
Fig. 15 Cyclic voltammogram for Cu
−10
UPD on a well-ordered Au(111)
1.0 electrode in 0.1 M H2 SO4 + 1 mM
CuSO4 , scan rate 1 mV s−1 (reproduced
from L. Kibler, Preparation and
Coverage
( 3 × 3)R30°
Characterization of Noble Metal
0.6 Single-Crystal Electrodes. Copyright
Random 2000 by International Society of
adsorption Electrochemistry), and electrochemically
0.2 (1 × 1) derived Cu coverage (normalized charge
referring to one complete Cu UPD ML)
0.0 0.1 0.3 0.4
as a function of potential, determined by
0.2
potential steps in the positive direction.
ESCE The Cu adlayer structures are also
[V] shown (adapted from Ref. [360]).
electrochemical methods, such as voltam- The coverage was estimated to 0.2 ML em-
metry [320–322, 337, 338], chronocoulom- ploying QCM [351, 352] and chronocoulo-
etry [339, 340], (2) in situ techniques such metric [339, 340] measurements. These
as STM [341–343], and AFM [344–346], experimental results are corroborated by
FTIR [199, 347], X-ray absorption spec- a recent theoretical work of Blum and
troscopy [348, 349], SXS [350, 351], and coworkers [67, 356–358] and Rikvold and
QCM [352, 353] and (3) ex situ UHV tech- coworkers [197, 359].
niques (LEED, RHEED, AES) [242, 354,
355]. Current transients and adsorption kinetics
The cyclic voltammogram of Cu UPD The shape of the current peaks, the
on Au(111) shows two well-defined hysteresis in the peak positions between
pairs of current peaks A1 /A2 and the cathodic and anodic potential sweeps
B1 /B2 corresponding to energetically dif- (particularly for B1 and B2 ) and
ferent adsorption/desorption processes lattice gas simulations [197, 359] suggest
(Fig. 15) [320–322, 337]. In the first step that monolayer formation occurs via
(peak A1 ), the transition between ran- several first-order 2D phase transitions.
domly adsorbed copper and (hydrogen) Single potential step experiments revealed
sulfate ions into an ordered layer of monotonously falling transients for peak
copper atoms (electrosorption valency √
A (disordered Cu adlayer → ( 3 ×
γ ∼ 1.8 [339, 340]) and coadsorbed sul- √1 ◦
3)R30 ) and rising transients for peaks
√ √
fate ions takes place. The resulting
√ √ A2 (( 3 × 3)R30◦ → disordered Cu
( 3 × 3)R30◦ structure was first ob- √ √
adlayer), B1 (( 3 × 3)R30◦ → (1 × 1)),
√ √
served by ex situ LEED and RHEED and B2 ((1 × 1) → ( 3 × 3)R30◦ )
experiments [354], and later confirmed (Fig. 16) [181]. If the potential step width
by in situ SXS [350], STM [341–343], is increased (‘‘higher overpotentials’’),
and AFM [344]. QCM [352, 353], chrono- the surface reaction becomes faster,
coulometric [339, 340], and FTIR- and the current maxima are shifted
measurements [199, 347] proved the to shorter times and higher current
coadsorption of sulfate. The SXS study densities. At low overpotentials, the entire
of Toney and coworkers [350] revealed phase transition process can be slowed
that copper atoms form a commensu- down into the timescale of seconds
rate honeycomb lattice (occupation of or minutes. The transients for peaks
threefold hollow sites, 2/3 ML coverage), A2 , B1 , and B2 could be modeled
while sulfate species are adsorbed in the by instantaneous (hole) nucleation and
centers (1/3 ML coverage) above the plane 2D growth (Bewick–Fleischman–Thirsk
of the copper atoms. Three oxygen atoms (BFT) – theory) in parallel to a Langmuir-
of each sulfate species are chemically type adsorption/desorption process of
bound to Cu atoms, and one points to- species, with both proceeding at different
wards the electrolyte [350]. In the second electrode sites [181]. Models based on site-
step (peak B), a pseudomorphic mono- unspecific coupling of adsorption and
layer of Cu (1 × 1) on Au(111) (occupation nucleation [361] or solely on homogeneous
of threefold hollow sites) is formed [343]. adsorption processes with lateral attraction
EXAFS experiments indicate the disor- failed [362].
dered coadsorption of sulfate species on Additional support for the nucleation
top of the copper monolayer [348, 349]. and growth controlled phase formation
−1200 300
190 mV 237 mV
A1 A2
[µA cm−2]
−800 200 236 mV

j
233 mV
−400 350 mV 190 mV 100
229 mV
0 0
0.0 0.2 0.4 0.6 0.8 0.0 0.5 1.0 1.5
t t
[s] [s]
40 mV 115 mV
−40 90 mV 30 mV 150 150
B1 B2
[µA cm−2]
100 104 mV 100

34 mV
j
−20 41 mV
98 mV
50 50
43 mV 90 mV
0 0
0.0 2 4 6 8 0.0 0.5 1.0 1.5 2.0
t t
[s] [s]
Fig. 16 Measured current transients ( ) for (a) Langmuir-like adsorption according to
the two-dimensional phase transitions during Cu Eq. (7) and instantaneous (hole) nucleation/2D
UPD on Au(111) in 0.05 M H2 SO4 + 1 mM growth represented by Eq. (33a) as two parallel
CuSO4 (peaks A1 , A2 , B1 , and B2 in Fig. 15). The processes are also shown (•). (Reprinted from
potential steps are indicated in the figure. The Ref. [181], copyright 1994 by Elsevier Science
results of a least-square fit to a model combining Ltd.)
process of Cu UPD on Au(111) is given the disordered, low-coverage phase (A2 )

by a recent dynamic MC simulation of is described by a sequence of two first-
Rikvold and coworkers [359]. The authors order phase transitions, the first involving
employed a two-component lattice gas the desorption of 1/3 ML of Cu followed by
√ √
(Cu2+ , SO4 2− ). The microscopic dynamics the decay of a ‘‘long-lived’’ ( 3 × 3)R30◦
includes adsorption/desorption and one- phase of 1/3 ML Cu and 1/3 ML sul-
step lateral diffusion of both adsorbate fate species [359]. Time-resolved SEIRAS
species with energy barriers according to experiments demonstrated that the first
the Volmer–Erdey–Gruz theory. The sim- step may already involve the partial des-
ulated current profiles for the transitions orption of sulfate [199].
√ √
( 3 × 3)R30◦ → (1 × 1) (B1 ) and (1 × The reversed, qualitatively different
√ √
1) → ( 3 × 3)R30◦ (B2 ) reflect one-step monotonically decaying profiles of the
nucleation and growth mechanisms with transition between randomly adsorbed
√ √
the same overpotential-dependence as ob- species and ( 3 × 3)R30◦ (A1 ) are mod-
served in the original experimental data. eled by an initial fast adsorption of
√ √
The transition of the ( 3 × 3)R30◦ into Cu, which cause the collapse of the
t = 0.0 s
t = 0.25 s
t = 1.00 s
Fig. 17 A series of snapshots of a dynamic Monte-Carlo

simulation after a negative-going potential step to 20 mV
below
√ the √ transition between the low coverage and mixed
( 3 × 3) layer (peak A1 ). Cu is represented by the filled
circles (•), sulfate by a triangle ( ), and the unoccupied sites
are open circles (.). After the step, a fraction of sulfate
desorbs, but the remaining sulfate combines with the newly
adsorbed copper to form a loose domain. With time, this
domain fills in and grows. The MC-current response
reproduces qualitatively the monotonically decreasing
experimental transient in panel A1 , Fig. 16 (with permission
by P. A. Rikvold, Ref. [359], copyright 1999, The
Electrochemical Society).
already present sulfate into domains Time-resolved in situ STM experi-

√ √
of a metastable ( 3 × 3) phase with ments [360] performed in the so-called
1/3 ML Cu and sulfate species, respectively x –t mode (one scanline is recorded
(Fig. 17). Later this fills with Cu form- as a function of time) revealed first
ing the equilibrium phase with 2/3 ML structural details of these phase forma-
Cu [359]. tion processes. Figure 18 indicates that
t = 12.7 s
B1
25 mV
180 mV
t=0s 100 nm
t = 12.7 s
B2
76 mV
37 mV
t=0s 100 nm
Fig. 18 (a) In situ x –t STM images of the phase transition

√ √
( 3 × 3)R30◦ → (1 × 1) for Cu UPD at B1 and the reversed
process at B2 in the neighborhood of a monatomic step. The
potential step regime and the corresponding time sequence are
..
given in the figure (with permission by M. Holzle, Ref. [360]).
√
the potential-induced transition ( 3 × adlayer structure and on the deposition
√
3)R30◦ → (1 × 1) (peak B1 ) starts at kinetics (cf. review in [24]). Coadsorption
monatomic steps and on flat terraces. with Cu atoms was proven by in situ
Nucleation on steps appears to be ener- radiotracer [363], ex situ AES [364], in situ
getically preferred and somewhat faster. EXAFS, and SXS [348, 349, 365, 366] and
Growth proceeds with a constant rate. chronocoulometry [339, 340, 367, 368]. In
This observation is directly reflected diffraction studies (LEED, RHEED [364,
in the corresponding current–potential 369], SXS [365]) it was demonstrated that
and current–time curves. Measurements the anions are structure-determining for
on well-oriented Au(111) electrodes with Cu UPD on Au(111). The adlayers at
miscut angles
0.1◦ revealed a split intermediate coverages are distinctly dif-
peak B as a results of energetically ferent for perchlorate, sulfate, chloride,
well-separated nucleation on steps (prefer- and bromide solutions (Figs. 15, 19), and
entially more positive) and terraces [337, altogether do not resemble the pseudomor-
342] (cf. Fig. 15). This separation van- phic adlayer structure for Cu evaporated
ishes and peak B broadens with miscut onto Au(111) in UHV [354]: (2.2 × 2.2)
√ √
angles >1◦ (increasing step density). Si- with ClO4 − , ( 3 × 3)R30◦ with SO4 2− ,
multaneously, the maximum of the cur- (5 × 5) with Cl− [364], and (4 × 4) with
rent transients changes into monotonously Br [365]. The strength of copper adsorp-
falling traces indicating dominant con- tion apparently decreases in the order
tributions from nucleation at step edges Cl− > Br− > SO4 2− for equal concentra-
coupled with 1D growth from these sites tions of the anion in the bulk. Since Br−
and/or increasing disorder of the ad- is more strongly adsorbed an Au(hkl) than
layer. Cl− indicates that the synergetic effect of
STM-measurements of the reversed the anion on the Cu adsorption does not
√ √
process (1 × 1) → ( 3 × 3)R30◦ (peak correlate directly with the strength of anion
B2 ) clearly show that the terraces are adsorption [370].
preferential (hole) nucleation sites, and The deposition kinetics is also strongly
that the new phase grows linearly with influenced by coadsorbed anions, as re-
time. Increasing the substrate step den- flected in the current densities of the
sity modifies this mechanism; the maxima positive UPD peak (A1 ). Cu deposition
of the rising part of the correspond- in ClO4 − solution is slow, requiring up
ing current transients appear distorted. to 20 min to form a full monolayer. Cl−
Similar trends were observed for the shifts the onset of Cu deposition towards
‘‘desorption’’ process around A2 . Unfor- more positive values. At potentials negative
tunately, the STM and chronoamperomet- of the peak at 0.265 V, a (5 × 5) structure
ric experiments did not yet allow the was observed [371], which is attributed to
derivation of a quantitative correlation Cl− forming a bilayer with coadsorbed
between step density and the elemen- Cu (cf. discussion in [367]). Cu deposi-
tary steps of 2D phase formation of tion in Br− solution causes, around the
Cu–UPD. sharp current peak at 0.260 V, the forma-
tion of a Cu-induced (4 × 4) Br− adlayer,
Influence of anions Cyclic voltammetry re- which fills up with increasing Cu coverage
veals that the anions of the supporting into a stochiometric CuBr pattern. At the
electrolyte have a strong impact on the second sharp current spike, around 0.05 V,
40 60
ClO4− Cl − 80 Br −
30 40
40
[µA cm−2 ]
20
20
j
0 0
10
−20
0 −40 (4 × 4)
(5 × 5)
(2.2 × 2.2) −40
−10 −80
−0.2 0.0 0.2 0.4 0.6 0.0 0.1 0.2 0.3 0.4 −0.1 0.1 0.3 0.4
ESCE ESCE ESCE
[V] [V] [V]
Fig. 19 Cyclic voltammograms for UPD of Cu on Au(111) in 0.1 M HClO4 , 0.05 M H2 SO4 , 0.1 M
HClO4 + 1 mM NaCl, and 0.1 M HClO4 + 1 mM NaBr. The copper concentration was always 1 mM,
scan rate 5 mV s−1 . The ordered structures in the medium coverage region are also indicated (with
..
permission by M. Holzle, Ref. [360]).
4 4
Au(100) Au(110)
2 2
[µA cm−2]
0 0
j
−2 −2
−4 −4
0.0 0.1 0.2 0.3 0.4 0.5 0.0 0.1 0.2 0.3 0.4 0.5
ESCE ESCE
[V] [V]
Fig. 20 Cyclic voltammograms for UPD of Cu on Au(100) and Au(110) in 0.1 M
H2 SO4 + 1 mM Cu2+ , scan rate 1 mV s−1 .
a phase transition into a (1 × 1) Cu layer orientation on Cu–UPD on Au(hkl) in

with a (4 × 4) Br− structure on top of it sulfuric acid solution. The Cu ML on
takes place. Both transitions are quite re- Au(111) is formed in two energetically
versible [365]. distinctly different steps in 0.35 V > E >
0.05 V, with the second peak being split for
quasi-perfect Au(111) surfaces into con-
Role of substrate crystallography on Cu–UPD tributions from step edges and terraces
Figures 15 and 20 illustrate the influ- (Fig. 15). The formation of the pseudo-
ence of the substrate crystallographic morphic Cu ML on Au(100) proceeds in
two steps in 0.40 V > E > 0.17 V. The ma- 3.3.4.2.3 Underpotential Deposition of
jor charge contribution is consumed in the Copper Ions on Pt(hkl) The UPD of Cu
broad, more positive feature. No ordered on Pt(hkl) has been studied by electro-
adlayer has been reported in this poten- chemical techniques [375–386], electrore-
tial region. At potentials negative of the flectance [375], radioactive labeling [387,
sharp current spike, an ordered Cu–UPD 388], IR-spectroscopy [389, 390], electro-
(1 × 1) ML was found [372]. The overall chemical quartz crystal microbalance
deposition process appears to be strongly (EQCM) [391], in situ STM [389, 392–396],
influenced by the presence of monatom- and a variety of X-ray techniques [200, 385,
ically high gold islands, created during 397–401], as well as ex situ UHV-methods
the lifting of the Au(100)-(hex) recon- such as LEED, AES, and XPS [389, 399,
struction [373]. On Au(110) the Cu ML is 402–406].
formed in a broad, single peak adsorption Figure 21 shows a slow scan voltam-
process [374]. mogram of Cu–UPD on a high-quality
40
A2
30 0
[µC cm−2]
−200
qM
20
−400
−600
[µA cm−2]
10 0.3 0.5 0.7

ERHE
j
[V]
B2
−10
B1 A1
(1 × 1) ( 3 × 3) ( 3 × 7)
−20
( 3 × 3)
0.3 0.4 0.5 0.6 0.7 0.8

E RHE
[V]
Fig. 21 UPD of Cu on a Pt(111) bead electrode in 1 mM
Cu2+ + 0.1 M H2 SO4 , sweep rate 1 mV s−1 . The inset shows the
integrated charge density for the positive and negative going potential
scans, respectively [407].
Pt(111) electrode in 0.1 M H2 SO4 [407]. 390, 402]. During the second UPD peak,
Two pairs of narrow deposition (A1 , B1 ) B2 , the sulfate ions are displaced from
and dissolution (A2 , B2 ) peaks, which the honeycomb center, and a (1 × 1) pseu-
are separated by a small, but distinct domorphic ML is formed with the Cu
hysteresis, are developed clearly. Inte- residing in the threefold hollow sites of
grating the current between 0.70 V and the Pt surface, on top of which the sul-
√ √
0.55 V (RHE) gives a monolayer charge of fate ions arrange in a ( 3 × 7) pattern,
approximately 500 µC cm−2 . The mecha- θ = 0.20 [389, 396, 397.] The reverse pro-
nism of Cu deposition on Pt(111) in the cesses take place during the dissolution
presence of sulfuric acid is described as of the UPD phase upon positive potential
follows: Bisulfate ions adsorbed on the excursion.
√ √
Cu-free Pt(111) surface form a ( 3 × 7) Systematic studies with stepped
structure [301] which gradually transforms Pt(111) s electrodes having (111) terraces
√ √
into ( 3 × 3)R30◦ [389, 390, 402]. Upon and monatomic (110) [386, 407] or
passing the first UPD peak A1 , sulfate ions (100) [379, 380] steps demonstrate that
and Cuδ+ coadsorb in a honeycomb lattice the fingerprint region of Cu–UPD
(2/3 ML of Cu, threefold hollow sites). The depends strongly on terrace size and step
√ √
resulting ( 3 × 3)R30◦ structure was density. At low coverages, Cu adsorbs
observed by in situ STM [390, 395, 396] preferentially on steps [382, 383, 407].
and SXS [408], and is in agreement with Figure 22 illustrates, as an example, typical
ex situ XPS and LEED investigations [389, current versus potential profiles for the
30
B2 A2
[µA cm−2]
20
j
10
D2 C2
0
Pt(111)
Pt(10.10.9)
Pt(776)
Pt(554)
Pt(775)
Pt(332)
Pt(221)
Pt(331)
0.4 0.5 0.6 0.7
E RHE
[V]
Fig. 22 Current versus potential profiles of the stripping of a Cu–UPD ML deposited on
stepped Pt(n n n-2) electrodes as recorded during a positive potential scan from 0.35 V to
0.80 V, 1 mV s−1 , with 1 mM Cu2+ in 0.1 M H2 SO4 [407].
0.8
0.8 4
0.6
[mA cm−2]
0.4
j
0.6 3
0.2
[mA cm−2]
[mA cm−2]
0.0
0.4 2
j
j
0 2 4
t
[s]
0.2 1
E = 0.65 V E = 0.68 V
E = 0.64 V E = 0.67 V
E = 0.63 V E = 0.66 V
0.0 0
0 5 10 15 0.0 0.5 1.0 1.5

t t
[s] [s]
Fig. 23 Current transients of the dissolution of a Cu UPD ML on Pt(111) in 1 mM Cu2+ + 0.1 M
H2 SO4 , obtained when stepping the potential from E1 = 0.50 V to various final potentials as
indicated in the figure. The transients with final potentials lower than 0.67 V could be modeled
by assuming two successive hole nucleation processes according to an exponential law coupled
with surface diffusion-controlled growth (cf. Eq. (30) and Eq. (33)). The inset shows, as an
example, the fit ( ) for the experimental transient (- - - - ) E1 = 0.50 V → E2 = 0.65 V [407].
dissolution of the Cu–UPD phase during (Fig. 23). Dissolution according to a 1D

a slow positive potential scan (1 mV s−1 ). line-by-line controlled mechanism was
The narrow peaks A2 and B2 representing recently proposed for stepped Pt-electrodes
the dissolution from terrace sites decrease with a terrace width <6 atoms. No detailed
and merge in a broader feature with analysis of the kinetic mechanism of the
decreasing terrace width (increasing step slow deposition process has been reported.
density). Simultaneously, two additional Cu–UPD on Pt(hkl) depends strongly on
current peaks, D2 and C2 , develop around substrate crystallography and anions of the
0.45 V and 0.73 V, which correspond to supporting electrolyte [378, 386, 402, 409].
sulfate desorption on copper-decorated On defect-free, hydrogen-cooled Pt(100)-
monatomic steps and, at more positive (1 × 1) electrodes a pseudomorphic
potentials than B2 and A2 , to the Cu adlayer of fully discharged Cu-ions is
dissolution from previously decorated formed (dissolved) in sulfuric acid in one
step sites [382]. The i–t transients of single step consuming 446 µC cm−2 [404,
Cu–UPD dissolution in sulfuric acid could 410, 411]. The stripping process is
be modeled within the BFT-theory by controlled by (hole) nucleation and 2D
two successive (Pt(111), low overvoltages) growth [407]. Cu deposition onto Pt(110)-
or one single (higher index, stepped (1 × 1) surfaces takes place in two distinct
electrodes) hole nucleation processes steps giving rise to a (2 × 1) and (1 × 1)
according to an exponential law coupled overlayer periodicity [394, 403]. Cu–UPD
with surface diffusion-controlled growth adlayers on Pt(hkl) are stabilized by
coadsorbed anions, which shift the onset these UPD phases were reported by
of deposition towards more negative Finnefrock and coworkers [200]: At the
potentials according to the following initial potential E1 = 0.20 V the epitaxial
sequence: F− < ClO4 − < (H)SO4 (2)− < Cu(1 × 1) layer is present at the surface.
Cl− < Br− < I− [390]. Rather complex At the final potential E2 = 0.35 V (past
bilayer structures have been reported B2) the incommensurate CuCl structure
(Fig. 24, cf. summary in [22]). The is formed (Fig. 25a). The corresponding
voltammetric profiles represent two-stage transient current, which is related to the
processes. At potentials above the first dissolution of Cu and the reorganization
UPD peak, chloride ions are adsorbed in of the Cl- adlayer, decays exponentially to
a disordered phase. During the first UPD- zero at times well below 1 s. The X-ray
step (disorder–order phase transition) a response at the scattering position (0.765,
Cu ML is formed, which is covered 0, 0.5), which is sensitive to the 2D order
by an incommensurate, close to (4 × of the Cu species within the UPD phase, is
4) Cl− -adlayer. The latter transforms slower and completely different in shape
during the second UPD peak at negative (Fig. 25c). The sigmoidal SXS-transients
potentials, and on top of an epitaxial (1 × were attributed to a 2D nucleation-
√
1) Cu adlayer, into a (2 × 3) phase and-growth process. These experimental
accompanied by partial desorption of results indicate that the processes of
Cl− [402, 405, 412, 413]. The extend of Cu and anion adsorption/desorption and
Cu and Cl− codeposition on Pt(111) s- the formation of the ordered adlayer
electrodes depends critically on terrace occur at rather different timescales.
width and step density [386]. Pioneering This conclusion is not accessible from
time-resolved SXS experiments on the the analysis of just the electrochemical
kinetics of phase transitions between current response! The reversed process,
120
1ML-(1 × 1) ∼ 0.5 ML Cu 1ML-(1 × 1) ∼ 0.5 ML Cu Cu
A2 A2 A2
80
B2
[µA cm−2]
40 B2
j
0
Cu(1 × 1)
B1
−40 B1 B1
A1 A1
Incomm-(hex)
A1 Cl− Incomm-(hex) Br − ( 3 × 3)R30° I−
−80
0.0 0.2 0.4 0.6 0.0 0.2 0.4 0.6 0.0 0.2 0.4 0.6
E SCE
[V]
Fig. 24 Cyclic voltammograms for UPD of Cu on Pt(111) in 0.05 M H2 SO4 + 1 mM NaCl, 0.05 M
..
H2 SO4 + 0.1 mM NaBr, and 0.05 M H2 SO4 + 0.1 mM NaI (with permission by M. Holzle,
Ref. [360]). The halide adlayer structures at intermediate Cu coverages between the deposition peaks
A1 and B1 , and the Cu coverage within the respective bilayers are also indicated.
350
Disordered Cl Incomm-(hex) Disordered Cl
EAg/AgCl on Cu(1 × 1) CuCl bilayer on Cu(1 × 1)
[V] 300
250
(a)
200
4
4
2 2
0
[mA]
0 10.0 10.5
i
−2
(b)
−4
(c)
1300
1200
[a.u.]
I
1100
1000
0 10 20 30 40
t
[s]
Fig. 25 Potential step experiment of the transition between the commensurate
Cu(1 × 1) and the incommensurate CuCl UPD adlayer on Pt(111) in 0.1 M
H2 SO4 + 1.0 mM Cu2+ : (a) potential perturbation, (b) current transients (the
inset shows a magnified region of the entire transient) and (c) time dependence
of the scattered X-ray intensity at (0.765, 0, 0.5), which represents a
characteristic diffraction rod of the incommensurate CuCl bilayer (reprinted from
Ref. [200], copyright 1998 by American Physical Society).
the formation of the epitaxial (1 × 1) Cu with Br− arranged in an incommensurate

adlayer from the incommensurate CuCl hexagonal structure aligned along the (1,0)
bilayer is reflected in the electrochemical surface direction. At potentials negative
and the SXS response as a single decaying of the second UPD peak, a pseudomor-
function with the same time constant. phic Cu (1 × 1) structure covered by a
Cu–UPD on Pt(111) electrodes in the disordered bromide layer is formed [365,
presence of bromide anions proceeds also 402, 413]. The structural transitions of
in two steps: An incommensurate bromide Cu–UPD in I− -containing electrolyte are
adlayer transforms first into a CuBr-bilayer not yet fully understood [22].
3.3.4.2.4 Underpotential Deposition of systems is represented by just one char-

Lead on Ag(hkl) In comparison to Cu acteristic pair of sharp current peaks, and
UPD (d = 2.560 Å) on Au(hkl) (d = the onset of metal deposition shifts to-
2.885 Å) or Pt(hkl) (d = 2.450 Å), lead wards more negative potentials according
(d = 3.490 Å) assumes a larger lattice to the following sequence: I− < Br− <
constant than the substrate material Cl− < citrate < acetate < ClO4 − .
Ag(hkl) (d = 2.885 Å). Potential step experiments of the Pb ML
The Pb–UPD on low-index silver formation in ClO4 − and acetate electrolyte
single crystals has been studied exhibit an initial fast decay due to double
extensively, employing electrochemical layer charging followed by a plateau re-
techniques [48–50, 130, 414–424] electro- gion and a subsequent accelerating decay
reflectance [420], SHG [425], STM [2, 130, to zero [130, 360, 421] (Fig. 26a). Initially,
426–428], EXAFS and SXS [429–431], as these transients were attributed to a homo-
well as ex situ UHV-methods such as LEED geneous Frumkin-type phase formation
and AES [420, 432]. process, involving the discharge of lead
Figure 26 shows typical cyclic voltam- and bulk diffusion as rate-determining
mograms of Pb–UPD on Ag(111) in three steps [416]. Recent electrochemical and
electrolytes containing perchlorate, ac- in situ STM experiments with high-
etate, or citrate ions [360, 422]. Three pairs quality stepped Ag(111)-surfaces [2, 130] or
thermally, in a hydrogen flame-annealed
of current peaks with a small, but distinct
electrodes [360], indicate that the plateau
hysteresis between the negative and the
region is most pronounced and does not
positive potential scans developed in the
represent a classical adsorption process.
presence of weakly specifically adsorbed
Instead, Lorenz and coworkers revised
perchlorate [48–50, 415–419, 421, 422].
their original interpretation and described
The electrosorption valency of lead, γ , is 2,
the formation transients of a Pb–UPD
which indicates that the deposited metal is
ML in ClO4 − and acetate electrolyte with
completely discharged and no cosorption
a model involving the lateral growth of
with the anion takes place [417]. Com- a 2D condensed ML, which starts ex-
parison of Ag(111)-substrates prepared clusively at monatomic steps [130, 336].
by chemical etching [415] with electrolyt- The current decay is critically dependent
ically grown quasi-perfect electrodes [419] on step density and distribution. Forma-
or with thermally annealed films [130, 424] tion transients of Pb UPD adlayers in
indicates that the first and third pair of solution with more strongly adsorbing an-
peaks represent processes on defect sites, ions, which block step sites (cf. citrate
while the middle peak is attributed to ter- in [48–50, 419, 421]) exhibit characteris-
race sites. Several authors proposed origi- tic bell-shaped morphology, which was
nally, on the basis of electrochemical stud- quantitatively modeled by instantaneous
ies, the 1D decoration of monatomic steps hole nucleation in combination with a
followed subsequently by the 2D phase for- 2D growth process with constant [418,
mation of lead on terraces [2, 48–50, 415, 421] or decreasing rate [360]. The criti-
418]. The presence of more strongly ad- cal nuclei were thought to be created
sorbing anions such as citrate, acetate [418, upon replacement of the prior, specif-
421–423], or halides [433] blocks the sur- ically adsorbed (citrate) anions by lead
face defect sites. The Pb–UPD of these species [418].
ClO4− D2 60 D2 NaOAc D2 Na2HCit
20 40
D3 D1
j
0 0 0
A3 A1
[µA cm−2]
−20 −60 −40
A2 A2 A2
−0.6 −0.4 −0.2 −0.6 −0.4 −0.2 −0.6 −0.4 −0.2

E SCE
(a) (b) [V] (c)
−12 0.4
15 −0.350 −0.449
−0.350 −0.450 0.3
−8
−0.200 −0.480
10 −0.446
j
−0.430 0.2
−0.450 −4
[mA cm−2]
5 −0.420 −0.443
0.1
−0.390
0 0
10 30 50 70 0 30 60 90 0 0.5 1.0 1.5 2.0 2.5
t t t
(a) [ms] (b) [ms] (c) [s]
Fig. 26 Cyclic voltammograms for UPD of Pb on well-prepared Ag(111) electrodes in different electrolytes. (a) 0.5 M NaClO4 + 1 mM HClO4 + 3 mM
PbClO4 , (b) 0.5 M NaAc + 1 m MHClO4 + 3 mM PbOAc, and (c) 0.5 M NaClO4 + 0.1 M Na2 HCit + 3 mM Pb(ClO4 )2 , scan rate 1 mV s−1 . The lower
panels a, b, and c show for each electrolyte typical current transients of formation of a complete UPD Pb ML. The transients in (a) have been reproduced
from Ref. [130], copyright 1997, NRC Canada. The transients of Pb UPD in citrate-containing solution (panel (c) solid lines) were fitted to a model based
..
on instantaneous hole nucleation combined with a time-dependent rate of growth (with permission by M. Holzle, Ref. [360]).
3.3 Phase Transitions in Two-dimensional Adlayers at Electrode Surfaces
431
2D Me
Time
1D Me
Terrace
Terrace
E 3D Me E 2D Me E1D Me 0 nm 20 nm
(a)
Time
Terrace Pit
E3D Me E 2D Me E1D Me 0 nm 80 nm
(b)
Fig. 27 (a) In situ STM line scan plot showing the step decoration by a 1D Pb
phase and the initial stage of 2D phase formation on a stepped Ag(111) substrate
and (b) line scans of formation of a 2D Pb UPD phase near a terrace and a
monatomic deep pit for Pb UPD on Ag(111) in 4 mM Pb(ClO4 )2 + 0.01 M HClO4
(by courtesy of E. Amman from Ref. [434]).
Combined electrochemical and in situ terrace into a 2D-condensed lead phase at

STM experiments gave clear evidence more negative potentials around the main
for the stepwise formation of low- current peak. The formation of the 2D
dimensional adlayer phases in the sys- Pb-ML represents a first-order phase tran-
tem Ag(111)/Pb2+ /ClO4 − , H+ [427, 434]. sition process, as reflected by the hysteresis
Figure 27(a) shows a series of potential- between the peaks A2 /D2 in the voltammo-
dependent line scans representing the gram. The growth process is illustrated in
initial stage of Pb–UPD on Ag(111). At Fig. 27(b). One may notice that a restricted
potentials just past the first current peak region around the upper part of the step
in the voltammogram of Fig. 26, the dec- edge having a width of around 2 nm re-
oration of monatomic steps occurs. The mains uncovered. Nucleation on the flat
width of the decoration extends to about terrace (e.g. on nondefect sites) is delayed
2 nm, which is in dynamic equilibrium and starts only at more negative potentials
with an expanded 2D Pb phase at the ter- corresponding to higher supersaturation.
race. The 1D phase either dissolves upon At potentials around the most negative
returning the potential towards more posi- voltammetric peak, the completion of the
tive values, or grows further onto the lower Pb–UPD phase on the upper terrace takes
place. The dissolution of the metal adlayer the UPD-range may cause surface alloy-
proceeds in the reverse regime. ing [427].
The sequence of formation/dissolution The completed Pb UPD is metallic, and
of the experimentally observed 1D and 2D represents an incommensurate, hexagonal
Pb-phases is in agreement with theoret- ML that is compressed compared with the
ical predictions based on a Nernst-type bulk metal by 0.1–3.2%, and rotated from
equation [2]: the substrate (011̄)-direction by ±4.5◦ [426,
427, 429–431]. The rotation of the ad-

o RT aPb2+ layer with respect to the substrate lattice
EPb,nD = EPb,3D + ln gives rise to a characteristic Moire pat-
zF aPb,nD
(43) tern as observed in several in situ STM
with the dimensionality n, EPb,3D o the studies [360, 426, 427] (Fig. 28). The in-
formal potential of Pb in the bulk phase, teraction between solvent molecules and
and aPb2+ the Pb activity in the elec- the Pb adatoms does not influence the
trolyte. The activity of the condensed metal structure of the complete ML deposited
phases shift towards lower values with de- in ClO4 − or acetate-containing electrolyte,
creasing dimensionality due to stronger since the UPD phase is essentially iden-
substrate–adsorbate interactions, and in tical to that of vapor-deposited Pb on
consequence, the corresponding equilib- Ag(111) at full coverage [420, 435]. The
rium potentials are shifted in the positive monolayer compression in the vacuum
direction [2]. Extended polarization within experiment (1–2%) is slightly less than for
3.46
Acetate, cathodic
Acetate, anodic
Perchlorate, cathodic
3.44 Perchlorate, anodic
a nn
[A]
3.42
3.40
10 nm × 10 nm 0 50 100 150
EPb/Pb2+
(a) (b) [mV]
Fig. 28 (a) Atomic resolution STM image of a plot. The solid and dashed lines are linear
Pb UPD ML in citrate-containing electrolyte at least-squares fits to the data for perchlorate and
potentials past the main deposition peak A2 acetate, respectively. These lines have slopes of
..
(with permission by M. Holzle, Ref. [360]). 0.424 ± 0.01 Å/V (perchlorate) and
(b) Dependence of the Pb ML near-neighbor 0.420 ± 0.01 Å/V (acetate); their offsets are
distance ann on the electrode potential E for Pb 3.390 Å (perchlorate) and 3.395 Å (acetate).
UPD on Ag(111). The near-neighbor spacing for (Reprinted from Ref. [431], copyright 1995 by The
bulk Pb is 3.501 Å, well above the range of the American Chemical Society.)
Pb UPD, probably caused by the stronger The order of these 2D systems is produced
Pb–Ag bonds compared to Pb–Pb [319], by a chemical reaction at the interface
although the rotational angle φ is about as the system approaches equilibrium.
the same. Unfortunately, they appear to contain a
The equilibrium change of the ML significant number of pinholes and other
Pb–Pb distance, aPb – Pb with the applied defects [443–445]. Constrains exist also for
electrode potential provides an estimate of monolayers based on the LB technique.
the 2D compressibility of the lead ML These studies are limited to the choice of
√ molecules with a very low solubility in the
aPb−Pb 3 ∂aPb−Pb subphase [438, 439].
κ2D = (44)
zeo ∂E T Alternatively, molecular and ionic mono-
layers can also be obtained on conducting
The experimental value κ2D = (1.25 ± surfaces in an electrochemical environ-
0.05) Å2 /eV is in qualitative agreement ment [3, 15, 16, 20, 143]. This approach
with the results of a simplified model based offers the advantage that formation and
on a 2D free electron gas [334, 431]. properties of a wide variety of adlay-
The rotational epitaxy angle estimated ers can be controlled as required by
by Toney and coworkers from in situ the applied electrode (substrate) poten-
SXS investigations (4.5◦ ) is only slightly tial and subsequently characterized by
smaller than the value predicted by the structure-sensitive in situ techniques in
model of Novaco and McTague [71, 72]. real space and real time. For these
The adlayer orientation with the lowest reasons, potentiostatically or galvanos-
energy is not the high symmetry direction. tatically generated monolayers on well-
Instead, the adlayer prefers to rotate to defined metal electrodes have become
nonsymmetry directions, taking advantage attractive model systems and provide an
of the lowest energy shear wave. important testing ground for fundamen-
A comparable discussion of Pb-UPD on tal issues in 2D physics and chemistry,
the other low-index phases of silver is such as phase transitions in adlayers
summarized in [2, 22]. and substrates surfaces [3, 7, 15, 16, 20,
143, 446]. Since the pioneering work
3.3.4.3Phase Transitions in Organic of Frumkin [447, 448], many ‘‘equilib-
Monolayers rium’’ and ‘‘dynamic’’ adsorption stud-
ies at metal–electrolyte interfaces have
3.3.4.3.1 General Aspects Organic mo- been performed with mercury [3, 15, 16,
nolayers on well-defined metal substrates 20, 449], gallium [18], and several amal-
may be obtained in various ways. Typ- gams [19], or low-melting-point electrodes
ical strategies are molecular beam epi- of sp-metals, such as Bi, Pb, Sn, or Zn,
taxy (MBE) [436], so-called ‘‘self-assembled onto which organic molecules are usually
monolayers’’ (SAM) [437] or LB films weakly adsorbed [449, 450]. The potentio-
(LB) [438, 439]. SAM’s are molecular as- static formation of 2D condensed mono-
semblies formed by the spontaneous in- layers was first reported by Lorenz [451]
teraction of a surfactant with a solid or and Vetterl [45] who found an unusual ca-
liquid substrate. Examples are thiols on pacitance versus potential hysteresis for
gold or silver [437, 440, 441] as well as alkyl- saturated aqueous solutions of nanoic
trichlorosilanes on oxide surfaces [442]. acid and the so-called ‘‘capacitance pits’’
for various purine and pyrimidine bases. pyridine [460] and several derivatives of
Since then, many other systems have uracil [469]. The phenomenological kinet-
been reported and quantitatively analyzed. ics of film formation and dissolution, as
Examples include carboxylic acids, cam- monitored by i–t, q M –t, or C –t tran-
phor and related compounds, purine and sients after single or multiple potential
pyrimidine and their derivatives, pyridine step protocols, have been described by
and bipyridines, quinolines, coumarin, models that involve (hole) nucleation and
thiourea, tetraalkyl-ammonium salts, and growth mechanism [183, 219, 221, 466].
so on. The corresponding literature is re- The results of these ‘‘macroscopic’’ elec-
viewed in [3, 15, 16, 20, 143]. trochemical experiments are still scarcely
Despite the rather detailed phenomeno- complemented by structural studies, such
logical knowledge on the 2D phase as IR- or Raman Spectroscopies or the
formation in adlayers, as obtained at use of UHV techniques after emersion
the mercury–electrolyte interface, fur- of the electrode [454, 463, 472–476]. Re-
ther progress has been hampered by markable progress in developing a ‘‘true’’
the lack of information on the struc- atomistic/molecular picture of potential-
ture and molecular mechanisms of film induced 2D organic phase formation
formation. New experimental and theoret- on adsorbate-modified single-crystal elec-
ical perspectives became available when trodes was achieved by the application of
single-crystal electrodes were employed in situ STM and AFM. Direct evidence
as substrate materials. This step offers for the formation of 2D long-range or-
the advantage to combine classical elec- dered structures on Au(111) has been
trochemical experiments with the power reported for several purine and pyrimi-
of structure sensitive in situ techniques, dine bases [201, 219, 454, 463, 477–480],
such as scanning probe microscopies phenol [481], pyridine [482, 483], 2,2 -
(STM, AFM) [23, 452, 453], vibrational and 4,4 -bipyridine [484–490], phenan-
spectroscopies (IRAS [454], SEIRAS [28], throline [491, 492], octylthiole [493], tetra-
SFG [30],. . .), electroreflectance [455], and methyl-thiourea [494, 495], cysteine [496],
surface plasmon microscopy [456], neu- mercaptopyridine [497], and benzenethi-
tron and surface X-ray scattering [457, ole [498]. Tao and coworkers [484, 491] and
458]. The formation of condensed organic Wandlowski and coworkers [201, 478, 479,
films on solid electrodes was suggested 486, 487] have demonstrated that the dy-
by Batrakov in 1974 for the adsorption namics of these phase formation processes
of camphor on Zn(0001) [459]. Few other can be studied successfully as a function
examples have been reported in the lit- of potential and temperature, even at a
erature during the last decade, mostly nanoscale level. Itaya and coworkers have
based on voltammetric and capacitance employed iodine-modified Au(111)- and
measurements: pyridine on Ag(210) [460], Ag(111)-electrodes to image self-organized
thymine on Cd(0001) [461], Au(100) [462], arrays of crystal violet, methyl-
Au(111) [463], and Ag(hkl) [464], coumarin pyridinium-phenylendivinylene and por-
on Au(111) and Au(100)-(hex) [465], uri- phyrine derivatives [499–501]. The same
dine on Au(hkl) [220, 466], uracil and group also reported 2D ordered struc-
camphor on Au(hkl) and Ag(hkl) [183, tures of benzene, naphthalene, and an-
193, 219, 221, 467–471]. Quantitative ther- thracene on Pt(111), Rh(111), and Cu(111)
modynamic studies were carried out for electrodes [502, 503]. These results are
complemented by a few structural stud- exclude interfacial water, recent in situ

ies with physisorbed films of purine SEIRAS studies on structural transitions
bases and porphyrine derivatives at the of 4,4 -bipyridine on Au(111) demonstrate
HOPG(0001)–aqueous electrolyte inter- that the ordered organic adlayer is sta-
face [210, 480, 504–508]. Commensurate bilized by hydrogen-bonded coadsorbed
2D ordered organic adlayers were also water species [476].
found on surfaces with quadratic sym-
metry, such as uracil and 2,2 -bipyridine 3.3.4.3.2 Thermodynamic Aspects The
on Au(100)-(1 × 1) [479, 487], and benzene adsorption of organic molecules on
and pyridine on Cu(100)[509]. ideally polarizable electrode–electrolyte
Basic ordering principles of the above- interfaces is based on the Gibbs equation,
mentioned monolayers appear to be (1) which relates, at constant temperature and
the ability to create strong and intermolec- pressure, the specific surface work γ to the
ular hydrogen bonds between adjacent electrode potential E and the bulk activity
molecules [463, 479, 507], (2) packing con- (concentration) ai of adsorbate i.
strains, molecular geometry, and dipole

forces [510], (3) ion pairing [511, 512], (4) ∂γ = −q M ∂E − RT i ∂ ln ai (45)
the formation of interfacial stacks due to π- i
electron attraction and London dispersion
forces [453, 479, 484, 491], (5) hydrophobic i represents the surface excess of the
interactions [513], as well as (6) substrate- component i, relative to a reference
adsorbate coordination chemistry [20, 463, species, usually the solvent. Equation (45)
493]. Complementary in situ STM studies is exact at liquid electrodes [515], and
with uracil and 2,2 -bipyridine, potentio- represents a first-order approximation
statically deposited on hexagonal Au(111) for solid electrodes assuming that the
or quadratic Au(100)-(1 × 1) electrodes, elastic surface strain terms (∂ε/∂E)T ,p,ai
demonstrate that the structure of the and (∂ε/∂ ln a)T ,p,E are negligible (cf.
adsorbed monolayers, and in particular discussion in Ref. [516, 517]). The specific
the registry between the adlayer and the surface work (interfacial tension in case
substrate, is affected by the symmetry of liquid electrodes), charge density, and
of the substrate (nature of the metal, differential capacitance are related by
crystallographic orientation), the adsor-
M ∂γ
bate–adsorbate interactions and the cor- q =− and (46)
∂E ai
rugations in the adsorbate–substrate in-
2 M
teraction potential [478, 479, 486, 487]. ∂ γ ∂q
The influence of the solvent on the C=− 2
= (47)
∂E ai ∂E ai
formation and stability of 2D condensed
organic phases is rather unexplored and The adsorption of neutral molecules at
controversial [3, 16]. Capacitance measure- electrified interfaces is characterized by
ments of ordered 2-thiouracil films at a smooth variation of γ , q M , C and i
mercury–acetonitrile interfaces showed as a function of electrode potential or
that the addition of small amounts of temperature (dotted and dashed curve in
water shifts the stability range and satu- Fig. 29) [449, 518]. The relation between
ration capacitance of the solidlike organic surface excess i (or coverage θi = i /im
phase [514]. While camphor and analogs with im as the maximum surface excess
Fig. 29 Schematic representation of γ C

the adsorption in the absence
(dotted-and-dashed line) and in the
presence ( ) of two-dimensional
phase transitions: interfacial tension γ ,
differential capacitance C, charge
density qM , and surface excess as a
function of the electrode potential E. E E
The dashed lines represent the
‘‘adsorbate-free’’ base electrolyte. qM Γ
E E
of i) and bulk activity ai at constant case of one-state adsorption, the classical

temperature is expressed by adsorption Frumkin isotherm
isotherms, which enable the estimation of
θ
energetic parameters such as Gibbs energy Bc = z = exp(−2aF θ) (48)
of adsorption, GA , or nearest-neighbor 1−θ
interaction energies, εAA . 2D first-order
degenerates into a vertical discontinuity if
phase transitions in the organic adlayer,
the Frumkin lateral interaction coefficient
such as disorder/order or reorientation
aF ≥ 2 [449]. The latter is related to the av-
processes, are represented by discontinu-
erage nearest-neighbor interaction energy
ous changes of the interfacial properties F as follows
εAA
(solid curves in Fig. 29) [3, 15, 16, 20,
143]. Often one observes a distinct hys- RT aF
F
teresis between the positive and negative εAA =− (49)
2
going potential scans [519]. Exact values of
the equilibrium transition potentials Ec , The derivation of aF is based on the
which corresponds to the coexistence of Bragg–Williams (BWA) or MFA assuming
the two phases (for instance liquidlike and (1) the homogeneous (statistical) distri-
solidlike patches), can be obtained by dou- bution of molecules among energetically
ble potential step experiments [143]. uniform sites, and (2) neglecting local fluc-
Historically, Frumkin-type models, tuations and correlations [74–76]. Retter
which represent the Helmholtz region pointed out that a more realistic isotherm
by a network of two or three treatment of 2D condensation in or-
condensers [449, 520–523] and classical ganic adlayers requires the consideration
thermodynamics based on a mean-field of localized adsorption and short-range
treatment [524–526], were applied first to (nearest-neighbor interactions) [527]. This
describe 2D phase transitions in organic improvement was achieved by introduc-
adlayers at metal–electrolyte interfaces as ing LGMs with a square symmetry, such
a function of concentration, potential, and as the two-state Ising model [528–531].
temperature (Sect. 3.3.2.2). In the simplest Each state of the lattice is assumed to be
vacant or occupied. Below a critical tem- phase for the square lattice is given by
perature Tc , the 2D lattice gas can split
5
into two phases, a gaseous phase and θ = 1 − x z − 2 2x − 16 z−2
8 −1 14
a condensed phase. The corresponding 2x

first-order equilibrium phase transition in- 31
volves a latent heat [528]. Two advanced − 3 6x 20 − 16x 22 + 24 z−3 . . .
3x
approximations of lattice gas isotherms
(52)
have been applied to describe 2D con-
with z = Bc and x = exp(−aLTE /4). Equa-
densation in organic adlayers: The QCA
tion (52) represents a truncated series
considers nearest-neighbor site pairs, but
expansion, which is obtained with the
treats them (still unrealistically) as inde-
assumption that contributions of clusters
pendent of each other [60, 73]:
larger than trimers can be neglected.
Figure 30 compares the adsorption
ln(Bc) = ln z
isotherms (ai + ln(Bc)) versus θ, i = F,

(β − 1 + 2θ)(1 − θ) 2 θ QCA, LTSE, and the temperature depen-
=
(β + 1 − 2θ)θ 1−θ dence of θ in the vicinity of the phase
transition for the above three approxima-
(50) tions and the exact LGM. According to

β = 1 − 4θ(1 − θ) the LGM, a first-order phase transition
takes place at (aG + ln(Bc)) = 0 and at
a 1/2
Qc temperatures lower than Tc [60, 528, 532].
× 1 − exp (51)
2 The degree of coverage of the condensed
phase at the equilibrium phase transition
The LTSE [529] considers equilibrium depends on aG , the lattice gas pair interac-
adsorbate clusters (up to 15 monomers) as tion coefficient, as follows:
well as various configurations of a cluster
a −4 1/8
with constant size indicating that the G
lattice geometry is taken into account. The θ = 0.5 + 0.5 1 − sin h
4
corresponding isotherm of the condensed (53)
θ θ
1.0
1.0
1
1 2
2 3
0.75 4
1 0.5 3
2
3
0.5 Tc /Tcrit
−2 −1 0 1 ai + ln(Bc) 0 0.5 1.0
(a) (b)
Fig. 30 (a) Coverage of the condensed phase as a function of the activity term (ai + ln(Bc)) and
(b) temperature dependence (Tc /Tcrit ) of the coverage of the condensed phase, θ, for the following
models: (1) BWA or MFA (Frumkin isotherm), (2) QCA, (3) LTSE, and (4) LGM. (Reprinted from
Ref. [528], Copyright 1987 by Elsevier Science Ltd.)
with aG = −2εAA /kT , εAA as the lateral of 2D condensation [527–530]. Figure 31

interaction energy. The square lattice gas illustrates experimental and theoretical
predicts the following critical values of C –E and θ –T phase diagrams for 3 mM
θ and aG at T = Tc : θc = 0.5 and aGc = iodocytosine/0.5 M KCl at various tem-
3.52549 [528]. Kharkats and coworkers ex- peratures [527]. The isotherm, based on
amined critical parameters and phase dia- the low-temperature series expansion rep-
grams of other 2D lattice geometries [533, resents the best fit of the experimen-
534]. tal data. The analysis of this system
The degree of coverage of the condensed also demonstrates that the Frumkin ad-
phase can be determined experimentally sorption model (MFA) is not appropri-
at constant potential from the saturation ate to model the phase diagram. These
capacity Cs , the capacity of the noncon- conclusions are supported by related
densed phase Co , and the capacity of the studies with camphor-10-sulfonate [536],
film C according to [529, 535] adenine [537], uracil [538], borneol, and
Co − C 3-hydroxyadamantane [539] adsorbed on
θ= (54) mercury electrodes as well as uracil deriva-
Co − Cs
tives on Ag(111) and Au(111) [469]. All
The potential dependence of θ at constant studies reported nearest-neighbor inter-
concentration results from combining the action energies at the respective critical
respective adsorption isotherms (48), (50) temperatures ranging between −4.0 and
or (52) with the potential dependence of −6.0 kJ mol−1 .
the adsorption coefficient B of neutral
organic molecules at metal–aqueous
3.3.4.3.3 Kinetic Aspects The kinetics of
electrolyte interfaces, which is defined
2D phase formation and dissolution of or-
as [449, 529]
ganic adlayers were mostly studied by i–t,
B = Bm exp(−α(E − Em )2 ) (55) q M –t or C –t single or multiple potential
step experiments, and analyzed on the ba-
1 GA
Bm = exp (56) sis of macroscopic models according to
55.5 RT strategies described in Chapter 3.3.3. Only
Co − Cs rather recently, modern in situ techniques
α= (57)
2RT m such as STM [20, 201, 453, 478, 479, 484,
487, 488] and time-resolved infrared spec-
with Em , Bm , GA representing the elec-
troscopy (SEIRAS) [475, 476] were applied
trode potential at maximum adsorption,
to study structural aspects of these phase
the adsorption coefficient at E = Em , and
transitions at a molecular or atomistic
the Gibbs energy of adsorption.
level.
Equations (54)–(57) and the correspond-
ing isotherms (48), (50) or (52) predict
two linear dependencies: (1) (Ec − Em ) 3.3.4.3.4 Case Study I – Coumarin at
versus Tc at constant adsorbate concen- the Mercury–Aqueous Electrolyte Interface
tration, and (2) 1/Tc versus ln i at E = Em The adsorption behavior of coumarin (2H-
with Tcrit referring to the maximum con- 1-benzopyran-2-one) at the mercury–elec-
densation temperature Tc . Furthermore, trolyte interfaces was investigated by
they permit to model the capacity ver- electrocapillary, capacitance, and tran-
sus potential dependence in the region sient experiments [162–164, 540–543].
25
20
[µF cm−2]
C
15
3 2 1
10
−0.7 −0.8
E SCE
(a) [V]
1.0
0.8
θ
Tcrit(MFA)
0.6
Tcrit(LGM)
280 290 300 310 320 330

T
(b) [K]
Fig. 31 (a) Capacitance-potential dependence for the system 3 mM
5-iodocytosine in 0.5 M KCl (Mc Ilvaine buffer, pH 7) at a mercury electrode at
different temperatures: (1) 298 K, (2) 308 K, (3) 315 K, and (4) 319 K. The bold
solid lines show the theoretical curves according to Eq. (52) calculated by
nonlinear regression for the square Ising lattice (reprinted from Ref. [536],
copyright 1993 by Elsevier Science Ltd.). (b) Temperature dependence of the
condensation degree of coverage θ for 3 mM (o) and 15 mM (+) 5-iodocytosine in
0.5 M KCl (McIllvaine buffer, pH 7) at a mercury electrode. (- - - - ) theoretical
curve according to the Frumkin or MFA adsorption model; ( ) theoretical
curve according to the LGM (Eq. 53). (Reprinted from Ref. [527] and [529],
Copyright 1984 and 1993 by Elsevier Science Ltd.)
Figure 32 shows a typical capacitance ver- with the coumarin molecule slightly
sus potential curve of 5 mM coumarin inclined with respect to the interface
in 0.5 M NaF at a HMDE. Three dif- and mI = 3.1 10−10 mol cm−2 [163]. The
ferent potential regions exist. Region capacitance decreases at more negative
I is interpreted as planar orientation potentials, and corresponds in region
0.4
O
0.3 O
[Fm−2]
C
0.2 I
II
0.1
III
−0.3 −0.5 −0.7 −0.9 −1.1 −1.3

E SCE
[V]
Fig. 32 Interfacial capacitance of mercury in contact with
aqueous 0.5 M NaF in the absence and in the presence of
5 mM coumarin, temperature 5 ◦ C, scan rate 5 mV s−1 . The
scan directions are indicated by arrows in the figure. The open
circles show the initial capacitance Ci observed immediately
after stepping the potential from region I into region III.
(Reprinted from Ref. [164], copyright 1994 by Bulgarian
Chemical Society.)
II to mII = 6.1 10−10 mol cm−2 with be quantitatively described by progressive

coumarin molecules aligned with the nucleation with 2D growth as rate limiting
C(5) and C(6) carbon atoms towards process, for example, m = 3 in the general
the electrode surface [162]. At sufficiently Avrami-formalism as expressed by Eq. 35
high adsorbate concentrations a 2D con- (Fig. 33a). At smaller overvoltages with
densed film is formed in region III, final potentials ranging between −0.530 V
giving rise to a characteristic capacitance and −0.555 V, stochastic behavior is
pit with a pronounced hysteresis at both encountered. All capacitance transients
edges [162–164]. Thomas and cowork- exhibit the same values for Ci and Cf ,
ers proposed a perpendicular orientation but there is stochasticity in the onset
with coumarin carbons C(6) and C(7) of each transient as well as in its
pointed towards the electrode and mIII = shape (Fig. 33b). The observed differences
8.4 10−10 mol cm−2 . The pit width de- between the individual transients are
creases linearly with increasing tempera- indicative of multiple nucleation, and
ture according to (E+ − E− )2 versus T , represent the domain of oligonucleation.
and a critical temperature Tcrit = 28 ◦ C By selecting Ef to be larger than −0.530 V,
could be estimated at coumarin saturation mononucleation events occur. Nucleation
concentration [162]. is so slow that there is negligible chance
Single potential steps I → III towards for a second nucleation during the
the center of the capacitance pit time necessary for the first nucleus to
(Ef < −0.560 V) yielded deterministic form a complete film on the hanging
sigmoid-shape transients, which could mercury drop (Fig. 33c). For an isotropic
442
0.4 0.3
0.3
0.2
3 Specific Adsorption
C
0.2
17 5 44 23 29 20
[Fm−2]
[Fm−2]
3 2 1
0.1
0.1
0 0
0 0.01 0.02 0.03 0.04 0.05 0.06 0 0.2 0.4 0.6 0.8 1.0
t t
(a) [s] (b) [s]
1.0 100
0.2 Small
0.8 Medium 80
Large
0.6 60
J
C
vg
1 2 3 4 5 6
0.1
[Fm−2]
0.4 40
[cms−1]
[cms−1]
0.2 20
0
0 0.5 1.0 1.5 2.0 2.5 −0.40 −0.45 −0.50 −0.55 −0.60
t ESCE
(c) [s] (d) [V]
Fig. 33 Capacitance transients obtained from single potential step experiments for 5 mM coumarin in 0.5 M NaF at a stationary mercury electrode,
5 ◦ C.(a) deterministic transients with E1 = −0.400 V to E2 = −0.560 V (1), −0.565 V (2), and −0.570 V (3) The solid lines were calculated by nonlinear
last squares fitting to Eq. (35) with m = 3. (b) Transients as in (a), except that E2 = −0.540 V. The transients are now stochastic. The numbers represent
their sequence number in a set of 50 repeat experiments. (c) Transients as in (a), except that E2 = −0.520 V. All transients reflect mononucleation and
can be represented by Eq. (30). The coverage was obtained with Eq. (35). (Reprinted from Ref. [164], copyright 1994 by Bulgarian Chemical Society.)
sphere, the growth of the compact film, capacitance pits, thermodynamic and ki-
once it has been nucleated, is given netic aspects of the phase formation
by [544] of uracil and its derivatives have been
studied comprehensively at the mer-
2θ = 1 − cos(vg (t − to )/r) (58) cury–electrolyte interface (cf. reviews in [3,
15]). In 1995, Hölzle and coworkers [219]
where vg is the linear growth velocity, presented convincing evidence that 2D
to the moment of nucleation, and r the condensation of uracil is also observed
radius of the mercury droplet, θ is obtained on low-index gold single-crystal electrodes.
from Eq. (36). The potential dependence of On the basis of voltammetric, capacitance
the growth rate is plotted in Fig. 33(d) for and transient experiments [183, 219–221,
three different drop sizes. The nucleation 461, 464, 462, 467, 547] and their combina-
rate was obtained for each potential by tion with structure-sensitive methods such
analyzing a series of 550 mononucleation as in situ STM [454, 478–480], vibrational
transients. Extrapolating the growth curves spectroscopy [454, 473, 475,], SXS [458],
to to , and sorting in bins, provides the and UHV-based characterization strate-
probability Po that nucleation has not gies [463], a rather comprehensive un-
yet occurred in a given time interval t derstanding of the interfacial properties
(cf. Sect. 3.3.3.2.3). Representing − ln Po of uracil and several of its derivatives
as a function of time then produces on Au(hkl), Ag(hkl), and Cd(0001) has
a plot from which both, the induction emerged.
time and the steady state nucleation rate Figure 34 shows a typical cyclic voltam-
is obtained [163, 164, 545]. The latter is mogram for 12 mM Uracil in 0.05 M
plotted in Fig. 33(d). The experiments H2 SO4 . The first voltammetric scan, as
were carefully designed to ensure that obtained with a flame-annealed, thermally
only a small fraction of the observed √
reconstructed Au(111)-(p × 3) electrode
nucleation appears to originate at the after immersion at −0.100 V, is indicated
lumen of the capillary (e.g. at defect as dashed line. The solid line repre-
sites!) [164]. sents the steady state voltammogram,
The dissolution transients of the con- recorded after five complete potential cy-
densed coumarin film III depend strongly cles [478]. The phenomenological analysis
on the applied potential step regime, of these electrochemical data yields the
and show strong aging effects, which existence of four distinct interfacial re-
were attributed to the healing of domain gions labeled I to IV in Fig. 34 [20]:
boundaries. The corresponding ‘‘grain- In region I the molecules are ran-
boundary’’ dissolution model as developed domly (gaseous-like) adsorbed on the
recently by Poelman and coworkers [546] electrode until complete desorption oc-
is applicable (cf. Fig. 6) curs at sufficiently negative potentials.
The formation of condensed coumarin The formation of a 2D condensed phy-
monolayers was also reported on Au(111)- sisorbed uracil film at more positive
(1 × 1) and Au(100)-(hex) [465]. potentials is indicated by typical current
spikes P1 /P1 and P2 /P2 , which limit
3.3.4.3.5 Case Study II – Uracil on Au(hkl) the pit region II. A marked hystere-
Since the pioneering work of Vetterl [45], sis is observed at both pit edges when
who reported for the first time so-called changing the direction of the potential
(b) 30 nm
P2
(a) 100 nm (c) 100 nm
P1'
I II IV
2 µA cm−2
0.0 0.2 0.6 ESCE

[V]
Start
P1
III
P3
P2'
(e) 100 nm (d) 20 nm

Fig. 34 Cyclic voltammograms for Au(111)/0.05 M H2 SO4 in the presence of 12 mM
uracil, scan rate 10 mV s−1 . The first scan as obtained with a flame-annealed,
reconstructed electrode after immersion at −0.10 V is plotted as dashed line. The
steady state voltammogram is shown as solid curve. The stability regions of the various
adlayer phases are labelled I to IV, and they are illustrated
√ with typical in situ STM
images: (a) thermally reconstructed Au(111)-(p × 3) surface, (b) physisorbed uracil
film, (c) unreconstructed Au(111)-(1 × 1) surface in the presence of the chemisorbed
uracil film, (d) as obtained after scanning the electrode potential from √ II → IV,
(e) island-free area of a potential-induced reconstructed Au(111)-(p × 3) electrode.
(Reprinted from [478], copyright 1997, American Chemical Society.)
scan. The stability range of II increases Hg < Ag(111) < Ag(100) < Au(111)
with decreasing temperature. The respec- √
−(p × 3) < Au(100)−(1 × 1)
tive critical temperatures assumes the
following substrate sequence [469, 470] ∼ Au(100)−(hex) (59)
The chronocoulometrically determined (111) terraces and monatomic (110)-

surface excess mII = 4.2 10−10 mol cm−2 oriented steps revealed that no 2D
(A = 0.40 nm2 ) and the small neg- condensation occurs at miscut an-
ative shift of Epzc are consistent gles >4◦ because of limited terrace
with a planar surface orientation size [475].
of the molecules [469]. Comparative Single and multiple potential step ex-
electrochemical [469], spectroscopic [475, periments demonstrated that the macroki-
454], and in situ STM experiments [478, netics of the formation of the phy-
479] on quasi-ideal Au(111) and Au(100) sisorbed uracil film represents a first-order
electrodes indeed revealed the existence phase transition and follows the expo-
of a highly ordered monolayer of uracil nential law of nucleation (cf. Eq. (34))
molecules interconnected via a network of in combination with surface diffusion-
directional hydrogen bonds, similar to that controlled growth [183]. In situ STM [20,
in the 3D solid. A typical high-resolution 478, 479] and time-resolved SEIRAS stud-
STM image and the proposed packing ies [475] suggest that these processes are
model with the high-order coincidence strongly related to the formation/breaking
mesh are plotted in Fig. 35. One may of uracil–water and water–water hy-
notice that the same long-range order drogen bonds within the Helmholtz
is found on hexagonal, densely packed region.
(111) as well as on the more open, The physisorbed uracil films II changes
quadratic (100) substrate surfaces, which upon passing the sharp current peak P2
point to the structure-determining role towards positive potentials due to the
of lateral molecular interactions, and break-off of the hydrogen-bonded network
to less important specific contributions accompanied by a partial charge transfer
of the substrate surface [478, 479]. Sys- process, which is strongly dependent on
tematic studies with stepped electrodes the solution pH and consumes up to
Au(n, n, n-2) = Au((n(111)-(110)) having 60 µC cm−2 [219, 466, 469, 475, 478].
1 nm
(a) 10 nm (b)
Fig. 35 Unfiltered in situ STM images of the physisorbed uracil film on
√
Au(111)-(p × 3)/0.05 M H2 SO4 + 3 mM uracil: (a) high-resolution image
at E = −0.05 V, iT = 2 nA, vT = +0.01 V. The primitive unit cell is indicated;
(b) proposed packing model. (Reprinted from [478], copyright 1997,
American Chemical Society.)
At sufficiently positive potentials, in densely packed arrangement of molecules

√
region IV, uracil forms on low- and higher- giving rise to a commensurate ( 3 × 3)
index Au(hkl) a chemisorbed, highly or- unit cell, which contains two mole-
dered adlayer, which (1) occurs rather cules (mIV (111) = 7.9 10−10 mol cm−2 ,
independent of the adsorbate concentra- AIV (111) = 0.21 nm2 ). On Au(100)-(1 ×
tion, (2) is stable at temperatures as 1) electrodes, a regular sticklike pat-
high as 100 ◦ C, and (3) inhibits the on- tern was obtained in region IV after
set of gold oxidation. This process is equilibration. Each stick consists of four
accompanied by the lifting of the sub- uracil molecules (mIV (100) = 7.2 10−10
strate surface reconstruction, and gives mol cm−2 , AIV (111) = 0.23 nm2 ) (Fig. 36).
rise to monatomic high gold islands as The steady state structures of the chemis-
shown in Fig. 34(c). In situ STM exper- orbed phase IV are determined by the
iments revealed on Au(111) a hexagonal competitive interplay of (1) short-range,
b
α
a
a
(a) 7 nm (b) 7 nm
a b
b α
a
[112]
[110]
1 nm 1 nm
(c) (d)
Fig. 36 High-resolution STM images and packing models of chemisorbed uracil
on Au(111)-(1 × 1) (a); E = 0.60 V, iT = 2 nA, vT = −0.65 V and
Au(100)-(1 × 1) (b); E = 0.60 V, iT = 1.8 nA, vT = −0.52 V in 0.05 M H2 SO4 .
Characteristic dimensions of the adsorbate mesh and the respective elementary
cells are indicated in the experimental images and in the proposed packing
models depicted in (c) and (d), respectively.
attractive π-stacking between adjacent (υC4=O ) with a time resolution of

uracil molecules, and (2) substrate-adsor- 200 µs reveals that the spectroscopic
bate coordination [478, 479]. The analysis transient response is systematically
of the STM-contrast pattern, in com- longer than the corresponding current
bination with electrochemical and IR response of the same potential step.
spectroscopic experiments of uracil and Both approaches indicate a dissolution
several of its N(1) and/or N(3)-substituted mechanism controlled by hole nucleation
methylderivatives demonstrates that the and diffusion-controlled growth, but
chemisorbed phase IV corresponds to per- reflect complementary properties of
pendicularly oriented molecules coordi- the phase transition [475]. Time-resolved
nated with the positively charged electrode SEIRA spectroscopy at electrochemical
surface via the N(3) nitrogen and both interfaces currently starts to develop as
C(2)=O, C(4)=O oxygen’s [469, 478, 479, a powerful approach to unravel the
454]. atomistic/molecular level details of these
The dissolution kinetics of the complex interfacial processes, such as
chemisorbed uracil phase IV on Au(111) 2D phase transitions, at solid–liquid
(towards III and II) was studied interfaces.
by potential step measurements. The The brief description of uracil adlay-
i–t transients could be modeled ers on Au(hkl) also demonstrates that the
by the exponential law [475] or understanding of 2D phase transition pro-
instantaneous [221] hole nucleation in cesses on defined solid electrodes, such as
combination with surface diffusion- single-crystalline substrates, requires the
controlled growth. The corresponding two simultaneous consideration of interfacial
rate constants vary systematically with the processes associated with the adlayer as
defect density (1) of the chemisorbed layer well as with the substrate surface.
(domain size, local disorder) and (2) of
the substrate surface (step orientation and
distribution, terrace size) [3, 221, 475]. 3.3.4.3.6 Case Study III – 2,2 -Bipyridine
Increasing defect density decreases the (2,2 -BP) on Au(hkl) 2,2 -bipyridine is a
half transition time of the ‘‘dissolution’’ bidentate ligand used in coordination
process, and the corresponding transients chemistry and a typical representative
finally approach the instantaneous limit of aromatic, nitrogen-containing hetero-
of hole nucleation. The STM experiments cycles, which act as basic building blocks
suggest that the critical hole nuclei are in highly specific and functionalized host
most probably represented by domain or lattices at defined surfaces [548, 549]. Two
point defect sites [478, 479]. Simultaneous planar pyridine rings are connected via a
electrochemical i–t and time-resolved C−C bond with 10% double-bond charac-
ATR-SEIRAS experiments in the rapid ter [550]. Electrochemical studies revealed
scan or the step-scan regime (Fig. 37) [475] that n, n -bipyridine [551–553] and several
with quasi Au(111) demonstrate that the of its Co2+ and Ni2+ complexes [554]
dissolution of the chemisorbed uracil form 2D condensed films at the atom-
film is associated with an orientational ically smooth mercury–electrolyte inter-
change of uracil from perpendicular face. The adsorption of 2,2 -bipyridine
to planar. Monitoring the integrated on solid electrodes, such as Ag(poly),
intensity of a characteristic uracil vibration Au(hkl), and Cu(111) was studied by
nC2 = O Fig. 37 Time-resolved SEIRAS–ATR
nC4 = O
experiment for the dissolution of the
chemisorbed uracil monolayer on
−0.005/a.u.
quasi-Au(111) thin-film electrodes in

0.1 M H2 SO4 + 12 mM uracil. The
potential was stepped from E1 = 0.60 V
120 (t1 = 30 s) to E2 = 0.20 V. (a) original
set of time-resolved SEIRAS spectra
80 (step-scan technique) with
characteristic uracil vibrations in the
]
40 tC=O stretching region (For clarity, the
[ms ‘‘loss’’ spectra are plotted with an
1750 1650 1550 1450
inverted sign !) (b) Time dependence of
the integrated intensity of the uracil
n
(a) [cm ]−1 vibration υC4=O (1590 cm−1 ). The solid
lines represent fits to an Avrami-type
0.12 equation (cf. Eq. (35)) with m = 1.12.
(c) Corresponding current transient.
Intensity (n C2 = O)
IR The Avrami-exponent m = 1.48 points

0.08 to nucleation according to an
[a.u.]
exponential law coupled with surface

0.04 I = Io(1-exp(-b ft m) diffusion-controlled growth (Eq. (34)
m = 1.12 and Eq. (35), solid line). The
(b) experimental results in panels (b) and
0 (c) indicate that the spectroscopic and
0 the electrochemical transients probe
different interfacial properties of the
dissolution process and illustrate the
−4 complementary information of both
i = const m t m−1 exp(-b ft m)
[mA]
approaches (reproduced from

i
m = 1.48
−8 Ref. [475]).
EC (c)
−12
0 20 40 60 80 100
t
[ms]
Raman scattering [555], chronocoulome- electrode. Molecular stacks are formed via
try, and SHG [556], SXS [458], FTIR [557], the coordination of the two ring nitrogen
SEIRAS [489], and in situ STM [484, atoms with the positively charged gold
486–489]. Lipkowski and coworkers [556] electrode in region I (Fig. 38a, b). The
found, on the basis of electrochemical data, distance between adjacent rows amounts
that 2,2 -bipyridine exhibits multistate to (0.96 ± 0.5) nm, the intermolecular
adsorption on Au(111). Structural details distance is (0.38 ± 0.2) nm, and the
of this complex phase behavior were molecules are tilted from the normal
resolved by in situ STM (Fig. 38) and to the axis of the chain by an
SEIRAS. angle α = (28 ± 2)o [484, 486, 488]. The
2,2 -BP was dissolved in aqueous simultaneous imaging of the hexagonal
electrolyte and potentiostatically deposited substrate surface and of the organic adlayer
onto the unreconstructed Au(111)-(1 × 1) pattern pointed to a commensurate (4 ×
N N
+ + + + + +
Stacking phase I
(e) 7.0 nm
Stacking phase II
(d) 70 nm (a) 8.0 nm
II I
2 µA cm−2
ESCE
-0.2 0 0.2 0.4 0.6 [V]
(c) 100 nm (b) 70 nm

Fig. 38 Cyclic current versus potential curves for Au(111)/0.05 M H2 SO4 in the absence
(dotted curve) and in the presence (full line) of 3 mM 2,2 -bipyridine, scan rate 10 mV s−1 .
(a) Unfiltered high-resolution image of the 2,2 -BP stacking structure at E = 0.50 V,
iT = 2 nA, vT = 10 mV. The suggested commensurate unit cell is indicated. (b) Large
domain of the anodic stacking phase I, (c) monatomic high gold ‘‘structure’’ decorating a
step of the Au(111)-surface after a single potential scan (10 mV s−1 ) from 0.50 V to
−0.05 V (negatively charged electrode). (d, e) large scale
√ and high-resolution image of the
cathodic 2,2 -BPH+ -stacking layer II on Au(111)-(p × 3) at −0.20 V. (Reprinted from
Ref. [486], copyright 1998 by Elsevier Science Ltd.)
√
2 3) unit cell, which contains three positions [486]. The structure-determining
molecules, for example, AI ≈ 0.38 nm2 . role of the substrate crystallography onto
The nitrogen atoms coordinate with the 2,2 -BP stacking pattern is supported
the substrate surface in twofold bridge by experiments with the quadratic
Au(100)-(1 × 1) surface. Dretschkow and is immediately reconstructed (Fig. 38d),

coworkers reported kinked stacking and soon after, a new STM-contrast
rows due to the misfit between high- pattern develops at negative charge den-
order substrate coordination sites, and sities in region II (Fig. 38e), which is
the potential molecular coordination composed of parallel molecular stacking
sites [487]. rows with alternating molecular tilt an-
After application of a single potential gels of +23◦ or −16◦ between adjacent
step/scan towards negative potentials, the rows. SEIRAS-experiments indicate a tor-
ordered 2,2 -BP adlayer dissolves, and the sional or inclined orientation of the two
entire substrate surface appears very mo- ring nitrogen atoms with respect to the
bile. Fractal, monatomic high gold features electrode surface [486, 489]. Both highly
grow at step edges (Fig. 38c), their surface ordered 2,2 -BP adlayers, in regions I and
35 nm × 35 nm
(a) (b)
(c) (d)
Fig. 39 Sequence of STM images for 3 mM 2,2 -BP on Au(111)/0.05 M H2 SO4
after a single potential scan (10 mV s−1 ) from 0.50 V to 0.24 V (potential-induced
dissolution of I): (a) E = 0.50 V, (b) E = 0.24 V, t = 3 min, (c) E = 0.17 V,
t = 4.5 min, (d) E = 0.17 V, t = 8.0 min; iT = 2 nA. (Reprinted from Ref. [486],
copyright 1998 by Elsevier Science Ltd.)
II, have been observed in sulfate- and stacking rows, and rather slow perpen-
in perchlorate-containing neutral or acidic dicular to it. Positional and directional
electrolytes [484, 486–489]. fluctuations within the ordered and the dis-
Details of the dynamics of the ordered phases appear. Under the shown
rather slow potential and/or temperature- experimental conditions the ordered 2,2 -
induced structural transitions could be BP stacking layer dissolved completely at
monitored by in situ STM in real space and 33.60 ◦ C (Fig. 40d). Closer inspection of
time [486]: The first example starts with a the disordered regions reveals a granular
highly ordered 2,2 -BP stacking layer I, structure on a length scale exceeding the
equilibrated at Ei = 0.500 V on Au(111)- pixel resolution of the image in the x –y
(1 × 1) (Fig. 39a), and then the potential direction. This means that a noticeable
was scanned slowly towards negative part of the modulation in the disordered
values. The range was restricted ensuring area is not a time effect caused by the fast
that the gold electrode still bears a positive motion of adsorbed species, but rather re-
charge. At E = 0.240 V the first point flects, at least in part, the static positions of
dislocations occur (Fig. 39b). The active particles in the disordered phase. This sug-
centers are formed not only on defects of gests the existence of adsorbate–substrate
the adsorbate or substrate lattices, such coordination complexes. Similar observa-
as point defects or domain boundaries, tions were reported for the 2D fluid–solid
steps or kink positions. They are equilibrium of cesium and oxygen coad-
distributed evenly across entire terraces. sorbed on Ru(0001) under UHV condi-
The nucleated holes grow anisotropically tions [558].
with time and more negative electrode In the present case, the adsorbed species
potentials until the ordered adlayer is ‘‘feel’’ the effect of the substrate corru-
dissolved completely. Growth proceeds gation potential. This situation is differ-
preferentially along the directions of the ent from purely 2D system, like gua-
commensurate molecular stacking rows. nine or adenine on HOPG [210] or the
During this transition one can identify two physisorbed films of uracil or thymine on
phases clearly, solidlike patches exhibiting Au(hkl) [219, 478, 479], where the adsor-
characteristic long-range order of 2,2 - bate–substrate interactions are very weak.
BP in region I, and a fluid medium There, the respective phase transitions are
disordered on the atomic scale (Fig. 39b). governed entirely by the applied electri-
The observed order/disorder transition is cal field and directional, lateral attrac-
first order [486]. tive interactions via hydrogen bonds [20,
The above interpretation is supported by 453].
experiments on the temperature-induced The potential and/or temperature con-
dissolution of the ordered 2,2 -BP phase trolled dissolution of the above order/
I [486]. The system was stabilized at Ei = disorder transition is completely reversible
0.200 V and T = 29.7 ◦ C (Fig. 40a), and in sulfuric acid solution. A representa-
then the temperature was slowly ramped tive time sequence, after application of
with 0.02 ◦ C min−1 . Firstly, point and line a potential scan from Ei = 0.100 V to
dislocations (‘‘one-dimensional holes’’) ap- Ef = 0.500 V, is plotted in Fig. 41. Within
pear (Fig. 40b). They grow anisotropi- a few seconds at Ef the 2,2 -BP molecules
cally by successive stripping of molecular start to order. No preferential nucleation
stacks, fast along the directions of the sites, which might correlate with defect
40 nm × 40 nm
(a) (b)
(c) (d)
Fig. 40 Temperature-induced dissolution of the 2,2 -BP stacking phase I for
3 mM 2,2 -BP in 0.05 M H2 SO4 /Au(111). (a) The system was equilibrated at
E = 0.20 V and 29.7 ◦ C for 20 min. Then the temperature was ramped with
0.02 ◦ C min−1 , (b) 33.14 ◦ C, (c) 33.51 ◦ C, (d) 33.60 ◦ C. All images were
obtained in constant current mode (iT = 0.25 nA) at the same surface area.
(Reprinted from Ref. [486], copyright 1998 by Elsevier Science Ltd.)
structures of the substrate surface, were positions at the end of the growing front.
found. First stacks are formed; their direc- (2) The second growth mechanism pro-
tions correlate but exhibit a ‘‘missing row’’ ceeds via incorporation of entire molecular
arrangement (Fig. 41b). With increasing rows perpendicular to the main stack di-
observation time this metastable phase is rection. The growth and the existence of
displaced by a densely packed stacking do- other metastable structures are strongly
main (Fig. 41c) with dimensions typical influenced by the nature of the support-
for phase I. The latter grows anisotropi- ing electrolyte anions [20, 484, 486, 488,
cally in two directions: (1) parallel along 489]. Defects within the adlayer change
the main stacking direction. Individual their shape or heal with increasing ob-
molecules and small stacks are still in servation time. The detailed mechanism
search of energetically favorable adlattice most probably involves place exchange
(a) 35 nm x 35 nm (b)
(c) (d)
(e) (f)
Fig. 41 Potential-induced formation of the 2,2 -BP stacking structure I on
Au(111) after a potential scan from region II (10 mV s−1 ): (a) Ei = 0.10 V to
Ef = 0.50 V, (b) E = 0.50 V, t = 1 min, (c) E = 0.50 V, t = 5 min,
(d) E = 0.50 V, t = 8 min, (e) E = 0.50 V, t = 11 min, (f) E = 0.50 V,
t = 42 min; iT = 2 nA. Solution composition 3 mM 2,2 -BP/0.05 M H2 SO4 .
(Reprinted from Ref. [486], copyright 1998 by Elsevier Science Ltd.)
processes and surface diffusion of 2,2 -BP- A main field of activities is focused on
gold complexes. A similar mechanism is structure and reactivity in two-dimensional
proposed for the ‘‘aging’’ of SAMs formed adlayers at electrode surfaces. Significant
by alkanthioles on gold surfaces [559]. new insights were obtained into the spe-
At the end of this chapter it is remarked cific adsorption and phase formation of
that the described structural details of the anions and organic monolayers as well as
above-mentioned phase transitions are not into the underpotential deposition of metal
accessible by macroscopic electrochemical ions on foreign substrates. The in situ
approaches, such as measurements of the application of structure-sensitive methods
interfacial capacitance or charging current, with an atomic-scale spatial resolution, and
but can be attributed to the often observed a time resolution up to a few microseconds
short or long-time transient responses. revealed rich, potential-dependent phase
behavior. Randomly disordered phases,
3.3.5 lattice gas adsorption, commensurate and
Conclusion and Outlook incommensurate (compressible and/or ro-
tated) structures were observed. Attempts
The macroscopic approach towards the have been developed, often on the basis of
‘‘static’’ electrochemical interface as de- concepts of 2D surface physics, to ratio-
veloped through the work of Lippmann, nalize the observed phase changes and
Gouy, Grahame, Frumkin/Damaskin, and transitions by competing lateral adsor-
Parsons is starting to change deeply. Over bate–adsorbate and adsorbate–substrate
the last decade a detailed microscopic pic- interactions.
ture of the bare and adsorbate-covered The thermodynamic and statistical me-
electrochemical interface has emerged. chanical analysis of first-order and/or con-
These advances could be attributed to four tinuous phase transitions demonstrated
major developments: (1) The use of well- clearly that models based on the MFA
defined single-crystal electrodes instead are often inadequate. Lattice gas mod-
of polycrystalline material. (2) The com- els (LGM), and approximations based
bination of complementary classical elec- on the quasi-chemical treatment (QCA)
trochemical techniques, based on macro- or series expansion techniques (e.g. the
scopic measurements of current, charge low-temperature series expansion (SE))
density, or interfacial capacitance, with are more appropriate. MC simulations
structure-sensitive in situ techniques, have also been developed as a powerful
such as vibrational spectroscopies (FTIR, approach to unravel the role of lateral
SEIRAS, surface-enhanced Raman scatter- interactions in various phase formation
ing (SERS), SFG), X-ray methods (SXS, processes.
EXAFS, . . .) and scanning probe micro- Impressive atomic- and molecular-scale
scopies (STM, AFM and related tech- details of steady state properties of 2D
niques). (3) The use of UHV techniques phase changes in ionic and molecular
for structure-sensitive investigations af- adlayers, and substrate surfaces have been
ter controlled emersion of the electrode obtained, mostly based on in situ STM,
and the transfer into a vacuum chamber. IR- and SXS-studies. Clear evidence was
(4) The development of modern theoreti- presented of the effect of crystallographic
cal concepts of the electrochemical double orientation, the coadsorption of ions of the
layer, and of processes taking place there. supporting electrolyte and of the solvent
molecules on the adlayer structures, and electrodes. Fascinating perspectives to-

in a few systems on the electrode reactivity. wards these goals are anticipated by new
Especially, scanning probe experiments methodological developments aimed to
provided important new insight into the improve the time- and spatial resolution
role of defects, and their control and of structure-sensitive in situ techniques.
manipulation in 2D phase formation The ‘‘ultimate goal’’ of true insights into
processes. The extension of these studies the physico-chemical nature of phase tran-
to more reactive metals, such as Fe, Ni, sitions, structural and dynamic changes in
or Al is almost nonexistent, despite their adlayers at electrodes calls for a multidis-
important technological relevance. ciplinary approach, combining high-level
Pioneering approaches were developed, experimental techniques with advanced
on the basis of time-resolved in situ theoretical treatments of clean and adsor-
STM, SEIRAS, and SXS-studies, to explore bate–modified electrified interfaces. The
structural details of dynamic processes basic knowledge obtained in these studies
involved in phase transitions in adlayers comprises a challenging testing ground
and substrate surfaces, and to relate for the electrochemical nanostructuring
these results to macrokinetic models of of surfaces and the tailoring of surface
nucleation and growth. reactivity.
In spite of this progress, major chal-
lenges exist in understanding the relation-
ship between structure and dynamics of Acknowledgment
2D phase changes and transitions in ad-
layers at electrochemical interfaces at a The present work was supported by the
microscopic level. These include topics Volkswagen Foundation, the Deutsche
such as (1) phase changes in functional- Forschungsgemeinschaft, and the Re-
ized organic adlayers induced by controlled search Center Jülich. I am deeply indebted
electrical, magnetic, or temperature per- to my collaborators in the groups in Ulm,
turbations; (2) the structure-determining and in Jülich, namely, Th. Dretschkow,
role of interfacial water as well as dy- M. H. Hölzle, D. Mayer, K. Ataka, G. Nagy,
namic relaxations; (3) the molecular-level and S. Pron’kin. Results of their exper-
control and modification of electrode re- imental work are partially incorporated
activity upon phase changes in adlayer. in this chapter. In addition, I would
At present, the microscopic mechanisms like to thank J. X. Wang, B. M. Ocko, and
of reactions at adlayer-covered electrodes R. R. Adzic very much for their pleas-
are only studied on a qualitative level. The ant and fruitful cooperation during SXS-
‘‘controlled’’ tailoring of structural prop- measurements at Brookhaven National
erties of clean and adsorbate-covered elec- Laboratory. Furthermore, I gratefully ac-
trodes offers fascinating opportunities to knowledge the critical comments and
design molecular- or atomic-scale sites to interesting discussions with N. J. Tao,
trigger specific reaction pathways, as well G. Nagy and U. Retter, and O. Magnussen
as sensitivity and selectivity of interfacial for sending Ref. [21] prior to publication.
processes. One may imagine electrochem- Finally, I would like to thank R. de Levie,
ical switches or nanoscale storage units D. M. Kolb, and H. Ibach for the encour-
based on phase transitions in functional- agement and the continuous support in
ized adlayers deposited on well-structured these studies.
References 21. O. M. Magnussen, Chem. Rev. 2002, 102,

679–725.
1. A. Zangwill, Physics at Surfaces, Cambridge 22. E. Herreo, L. J. Buller, H. Abruna, Chem.
University Press, Cambridge, 1988. Rev. 2001, 101, 1897–1930.
2. E. Budevski, G. Staikov, W. J. Lorenz, Elec- 23. K. Itaya, Prog. Surf. Sci. 1998, 58, 121–247.
trochemical Phase Formation and Growth, 24. E. Herrero, in Encyclopedia of Analytical
VCH, Weinheim, 1996. Chemistry (Ed.: R. A. Meyers), John Wiley
3. C. Buess Herman, S. Bare, M. Poelman & Sons, New York, 2002, in press.
et al., in Interfacial Electrochemistry (Ed.: 25. D. M. Kolb, in Spectroelectrochemistry (Ed.:
A. Wieckowski), Marcel Dekker, New York, R. J. Gale), Plenum Press, New York, 1988,
1999, pp. 427–452. pp. 87–188.
4. A. Wieckowski, Interfacial Electrochemistry, 26. R. Corn, in Adsorption of Organic Molecules
(Eds.: J. Lipkowski, P. N. Ross), VCH, New
Marcel Dekker, New York, 1999.
York, 1992, pp. 391–408.
5. J. A. Venables, Introduction to Surface and
27. M. J. Weaver, S. Zou, in Advances in Spec-
Thin Film Processes, Cambridge University
troscopy (Eds.: R. J. H. Clark, R. E. Hesters),
Press, Cambridge, 2000.
John Wiley & Sons, Chichester, 1998,
6. W. J. Lorenz, W. Plieth, Electrochemical
Vol. 26, Chap. 5.
Nanotechnology, Wiley/VCH, Weinheim,
28. M. Osawa, Bull. Chem. Soc. Jpn. 1997, 70,
1998.
2861–2880.
7. J. Lipkowski, P. N. Ross, Adsorption of Or-
29. Z. Q. Tian, B. Ren, in Encyclopedia of Analyt-
ganic Molecules at Metal Electrodes, VCH,
ical Chemistry (Ed.: R. A. Meyer), John Wiley
New York, 1992.
& Sons, New York, 2001, pp. 9162–9201.
8. D. M. Kolb, Prog. Surf. Sci. 1996, 51, 30. A. Tadjeddine, A. LeRille, in Interfacial Elec-
109–173. trochemistry (Ed.: A. Wieckowski), Marcel
9. H. S. Nalwa, Handbook of Nanostructured Dekker, New York, 1999, pp. 317–343.
Materials, Academic Press, San Diego, 2000, 31. I. Lyuksyukov, A. G. Naumovets, V. Pokro-
Vol. 5. vsky, Two-dimensional Crystals, Academic
10. G.A. Somorjai, Introduction to Surface Press, Boston, 1992.
Chemistry and Catalysis, John Wiley & Sons, 32. T. L. Einstein, in Chemistry and
New York, 1994. Physics of Solid Surfaces VII (Eds.:
11. A. Baiker, J. D. Grunwaldt, Ch. A. Müller R. Vanselow, R. Howe), Springer, Berlin
et al., Chimia 1998, 52, 517–524. 1988, pp. 307–336.
12. A. Ahmadi, G. Attard, J. Feliu et al., Lang- 33. B. N. J. Person, Surf. Sci. Rep. 1992, 15,
muir 1999, 15, 2420–2424. 1–135.
13. N. J. Tao, Phys. Rev. Lett. 1996, 76, 34. U. Retter, J. Electroanal. Chem. 1987, 236,
4066–4069. 21–30.
14. J. Lipkowski, P. N. Ross, Imaging of Surfaces 35. U. Retter, J. Electroanal. Chem. 1984, 165,
and Interfaces, Wiley-VCH, New York, 1999. 221–230.
15. R. de Levie, Chem. Rev. 1988, 88, 599–609. 36. U. Retter, Electrochim. Acta 1996, 41,
16. C. Buess-Herman, Prog. Surf. Sci. 1994, 46, 2171–2174.
335–375. 37. W. Gebhardt, U. Krey, Phasenübergänge und
17. M. Fleischmann, H. Thirsk in Advances in kritische Phänomene, F.Vieweg & Sons,
Electrochemistry and Electrochemical Engi- Braunschweig, 1980.
neering (Ed.: P. Delahay), John Wiley & 38. G. Brown, P. A. Rikvold, S. M. Mitchell
Sons, New York, 1993, pp. 123–210. et al., in Interfacial Electrochemistry (Ed.:
18. M. R. Moncelli, M. Innocenti, R. Guidelli, A. Wieckowski), Marcel Dekker, New York,
J. Electroanal. Chem. 1990, 295, 275–290. 1999, pp. 47–61.
19. M. Freymann, M. Poelmann, C. Buess- 39. J. X. Wang, R. R. Adzic, B. M. Ocko, in Inter-
Herman, Isr. J. Chem. 1997, 37, 241–246. facial Electrochemistry (Ed.: A. Wieckowksi),
20. Th. Dretschkow, Th. Wandlowski, The Marcel Dekker, New York, 1999,
Solid-Liquid Interfaces – A Surface Science pp. 175–186.
Approach (Ed.: K. Wandelt), Springer, 40. P. Ehrenfest, Leiden Comm. Suppl. 1933,
Berlin, 2002, in press. 75b, 1–10.
41. L. D. Landau, E. M. Lifshitz, Statistical 65. N. D. Lang, A. R. Williams, Phys. Rev. B

Physics, Pergamon Press, New York, 1980. 1978, 18, 616–636.
42. Cl. Buess-Herman, J. Electroanal. Chem. 66. T. L. Einstein, CRC Crit. Rev. Solid State
1985, 186, 27–39. Mater. Sci., 1978, 7, 261–299.
43. P. Nikitas, J. Electroanal. Chem. 1991, 300, 67. L. Blum, D. A. Huckaby, M. Legault, Elec-
607–618. trochim. Acta 1996, 41, 2207–2227.
44. G. L. Gaines, Insoluble Monolayers at Liq- 68. Y. I. Frenkel, T. Kontorova, Zh. Eksp. Teor.
uid/gas Interfaces, Wiley-Interscience, New Fiz. 1938, 8, 1340–1345.
York, 1966. 69. F. C. Frank, J. H. van der Merwe, Proc. R.
45. V. Vetterl, Collect. Czech. Chem. Commun. Soc. London 1949, 198, 205–216.
1966, 31, 2105–2111. 70. V. L. Pokrofsky, A. L. Talapov, Phys. Rev.
46. J. Lipkowski, in Modern Aspects Lett. 1979, 42, 66–67.
of Electrochemistry (Eds.: R. E. White, 71. A. D. Novaco, J. P. McTague, Phys. Rev. Lett.
J. O’. M. Bockris, B. E. Conwa), Plenum 1977, 38, 1286–1289.
Press, New York, 1992, pp. 1–99, Vol. 23. 72. J. P. McTague, A. D. Novaco, Phys. Rev. B
47. Th. Wandlowski, M. Heyrovski, L. Novot- 1979, 19, 5299–5306.
ney, Electrochim. Acta 1992, 37, 2663–2672. 73. T. Hill, Statistical Thermodynamics, Addison
48. A. Bewick, B. Thomas, J. Electroanal. Chem. Wesley, Reading, 1962.
1975, 65, 239–244. 74. W. Bragg, E. Williams, Proc. R. Soc. London,
49. A. Bewick, B. Thomas, J. Electroanal. Chem. Ser. A 1934, 145, 699–730.
1977, 84, 127–140. 75. W. Bragg, E. Williams, Proc. R. Soc. London,
50. A. Bewick, B. Thomas, J. Electroanal. Chem. Ser. A 1935, 151, 540–566.
1977, 85, 329–337. 76. R. Fowler, E. Guggenheim, Statistical Ther-
modynamics; Cambridge University Press,
51. Th. Wandlowski, J. X. Wang, O. M. Mag-
London, 1939.
nussen et al., J. Phys. Chem. 1996, 100,
77. E. Guggenheim, Proc. R. Soc. London, Ser.
10 277–10 287.
A 1935, 148, 304–312.
52. M. Fisher, Rep. Phys. London 1967, 30, 615.
78. H. Bethe, Proc. R. Soc. London, Ser. A 1935,
53. B. M. Ocko, J. X. Wang, T. Wandlowski,
150, 552–575.
Phys. Rev. Lett. 1997, 79, 1511–1574.
79. C. Domb, in Phase Transitions and Critical
54. M. Schick, Prog. Surf. Sci. 1981, 11,
Phenomena (Eds.: C. Domb, M. Green),
245–292. Academic Press, New York, 1974, pp. 357,
55. Th. Wandlowski, J. X. Wang, B. M. Ocko, Vol. 3.
J. Electroanal. Chem. 2001, 500, 418–434. 80. M. Sykes, J. Essam, D. Gaumit, J. Math.
56. M. T. M. Koper, J. Electroanal. Chem. 1998, Phys. 1983, 6, 283–287.
450, 189–201. 81. J. Koryta, Collect. Czech. Chem. Commun.
57. S. J. Mitchell, G. Brown, P. A. Rikvold, Surf. 1953, 18, 206–213.
Sci. 2001, 471, 125–142. 82. P. Delahay, I. Trachtenberg, J. Am. Chem.
58. D. E. Taylor, E. D. Williams, R. L. Park Soc. 1957, 79, 1362–2355.
et al., Phys. Rev. B 1985, 32, 4653–4659. 83. R. Srinivasan, R. de Levie, J. Electroanal.
59. L. Onsager, Phys. Rev. 1944, 665, 117–125. Chem. 1986, 205, 303–307.
60. T. Hill, Statistical Mechanics, McGraw Hill, 84. A. J. Bard, L. Faulkner, Electrochemical
New York, 1956, pp. 286–353, Chap. 7. Methods and Applications, John Wiley &
61. K. G. Wilson, J. Kogut, Phys. Rep. 1974, 12, Sons, New York, 1980.
75–199. 85. A. N. Frumkin, V. I. Melik-Gaikazyan, Dokl.
62. K. Binder, D. W. Hermann, Monte Carlo Akad. Nauk USSR 1951, 77, 855–858.
Simulation in Statistical Physics, Springer Se- 86. V. I. Melik-Gaikazyan, Zh. Fiz. Khim. 1952,
ries Solid State Sci., Springer, Heidelberg, 26, 560–580, 1184–1189.
1992, Vol. 80. 87. W. Lorenz, F. Möckel, Z. Elektrochem. 1956
63. M. C. Desjonqueres, D. Spanjaard, Con- 60, 507–515, 939–944.
cepts in Surface Physics, Springer, Berlin, 88. R. D. Armstrong, W. P. Race, H. R. Thirsk,
1996. J. Electroanal. Chem. 1968, 16, 517–529.
64. J. Israelashvilli, Intermolecular and Surface 89. P. Delahay, D. Mohilner, J. Am. Chem. Soc.
Forces, Academic Press, London, 1992. 1962, 84, 4247–4252.
90. P. Delahay, J. Phys. Chem. 1963, 67, 115. E. Budevski, V. Bostanov, T. Vitanov et al.,
91. D. Schuhmann, in Proprietes Electriques des 116. D. M. Kolb, in Advances in Electrochem-
Interfacial Charges (Ed.: D. Schuhmann), istry and Electrochemical Engineering (Eds.:
Masson, Paris, 1978, pp. 166–192. H. Gerischer, Ch. Tobias), John Wiley &
92. M. Noel, K. I. Vasur, Cyclic Voltammetry Sons, New York, pp. 125–272, Vol. 11.
and Frontiers in Electrochemistry, Aspect 117. R. Philipp, U. Retter, Thin Solid Films 1992,
Publications, London, 1990. 207, 42–50.
93. P. Delahay, C. T. Fike, J. Am. Chem. Soc. 118. Southhampton Electrochemistry Group, In-
1958, 80, 2628–2630. strumental Methods in Electrochemistry, Ellis
94. T. Berzins, P. Delahay, J. Am. Chem. Soc. Horwood, New York, 1990.
1955, 49, 906–909. 119. Cl. Buess-Herman, L. Gierst, Colloids Surf.
95. P. Delahay, I. Trachtenberg, J. Am. Chem. 1984, 12, 137–150.
Soc. 1957, 79, 2355–2362. 120. U. Retter, J. Electroanal. Chem. 1990, 296,
96. L. Rampazzo, Electrochim. Acta 1965, 14, 445–451.
733–741. 121. D. Kashchiev, Surf. Sci. 1969, 14, 209–220.
97. K. Jüttner, G. Staikov, W. J. Lorenz et al., 122. D. Kashchiev, Nucleation – Basic Theory
J. Electroanal. Chem. 1977, 80, 67–80. and Applications, Butterworth-Heinemann,
98. J. Michailik, B. B. Damaskin, Sov. Elec- Oxford, 2000.
trochem. 1979, 15, 478–481. 123. R de Levie, in Advances in Electrochem-
99. N. Emanuel, D. Knorre, Kurs Chimitschesoi istry and Electrochemical Engineering (Eds.:
Kinetiki; Wyschaja schkola, Moskwa, 1962. H. Gerischer, Ch. Tobias), John Wiley &
100. W. Lorenz, Z. Elektrochem. 1958, 62, Sons, New York, 1985, pp. 1–67, Vol. 13.
192–200. 124. S. Toshev, I. Markov, J. Cryst. Growth 1968,
101. R. D. Armstrong, J. Electroanal. Chem. 1969, 34, 436–445.
20, 168–170. 125. S. Toshev, I. Markov, Ber. Bunsen-Ges. Phys.
102. J. W. Gibbs, Collected Works, Longman’s Chem. 1969, 73, 184–192.
Green, London (1878), 1978. 126. D. Kashchiev, Surf. Sci. 1969, 18, 389–397.
103. M. Volmer, A. Weber, Z. Phys. Chem. 1926, 127. J. H. Sluyters, J. H. O. J. Wijenberg, W. H.
119, 277–301. Mulkder et al., J. Electroanal. Chem. 1989,
104. M. Volmer, Kinetik der Phasenbildung, 261, 263–272.
Steinkopf, Dresden, 1939. 128. S. Toschev, J. Gutzow, Phys. Status Solidi
105. L. Farakas, Z. Phys. Chem. 1927, 125, 239. 1967, 21, 683–689.
106. I. N. Stranski, R. Kaishev, Z. Phys. Chem. 129. I. Stranski, Z. Phys. Chem. 1928, 136,
1934, B26, 100–113. 209–220.
107. I. N. Stranski, R. Kaishev, Z. Phys. Chem. 130. G. Staikov, W. J. Lorenz, Can. J. Chem.
1934, B26, 114–131. 1997, 75, 1624–1634.
108. I. N. Stranski, R. Kaishev, Z. Phys. Chem. 131. A. Milchev, Electrochim. Acta 1986, 31,
1934, B26, 312–316. 977–980.
109. I. N. Stranski, R. Kaishev, Z. Phys. Chem. 132. A. Milchev, J. Electroanal. Chem. 1998, 457,
1934, A170, 295–299. 35–46.
110. R. Becker, W. Döring, Ann. Phys. 1935, 24, 133. P. M. Jacobs, F. C. Tompkins, in Chemistry
719–753. of the Solid State (Ed.: W. E. Garner), Butter-
111. J. B. Zeldovich, Acta Physicochim. (URSS) worths, London, 1955, pp. 181–212.
1943, 18, 1–12. 134. U. Retter, J. Electroanal. Chem. 1978, 87,
112. I. Ya. Frenkel, J. Chem. Phys. 1939, 7, 181–188.
200–210; 538–546. 135. W. Obretenov, I. Petrov, I. Nachev et al., J.
113. E. Budevski, G. Staikov, W. J. Lorenz, Elec- Electroanal. Chem. 1980, 109, 195–198.
trochim. Acta 2000, 45, 2559–2574. 136. C. Donner, H. Baumgärtel, L. Pohlmann
114. E. Budevski, in Comprehensive Treatise et al., Ber. Bunsen-Ges. Phys. Chem. 1996,
of Electrochemistry (Eds.: B. E. Conway, 100, 403–412.
J. O. M. Bockris, E. Yeager et al.), Plenum, 137. C. Donner, L. Pohlmann, H. Baumgärtel,
New York, 1983, pp. 399–450, Vol. 7. Surf. Sci. 1996, 345, 363–372.
138. R. Philipp, U. Retter, Z. Phys. Chem. 1996, 163. R. Srinivasan, R. de Levie, J. Phys. Chem.
196, 125–134. 1987, 91, 2904–2908.
139. I. Markov, Crystal Growth for Beginners, 164. Th. Wandlowski, R. de Levie, Bulg. Chem.
World Scientific, Singapore, 1995, Commun. 1995, 27, 231–246.
pp. 63–146. 165. E. Bosco, S. K. Rangarajan, J. Chem. Soc.,
140. D. Walton, J. Chem. Phys. 1962, 37, Faraday Trans. 1981, 177, 483–495.
2182–2190. 166. F. Canac, C. R. Acad. Sci. 1933, 196, 51–62.
141. S. Stoyanov, Thin Solid Films 1973, 18, 167. A. N. Kolmogoroff, Bull. Acad. Sci. USSR,
91–98. Ser. Math. Nat. 1937, 3, 355–361.
142. A. Milchev, S. Stoyanov, R. Kaischev, Thin 168. M. Avrami, J. Chem. Phys. 1939, 7,
Sold Films 1974, 22, 255–265, 267–274. 1103–1111.
143. C. Buess Herman, in Adsorption of 169. M. Avrami, J. Chem. Phys. 1940, 8, 212–224.
Organic Molecules at Metal Electrodes (Eds.: 170. M. Avrami, J. Chem. Phys. 1941, 9, 177–184.
J. Lipkowski, P. N. Ross), VCH, New York, 171. U. R. Evans, Trans. Faraday Soc. 1945, 41,
1992, pp. 77–118. 365–374.
144. O. M. Magnussen, L. Zitzler, B. Gleich 172. M. Y. Abyaneh, M. Fleischmann, Elec-
et al., Electrochim. Acta 2001, 46, trochim. Acta 1982, 27, 1513–1518.
3725–3733. 173. U. Retter, J. Electroanal. Chem. 1984, 179,
145. F. C. Frank, Proc. R. Soc. London, Ser. A 25–29.
146. S. Rangarajan, Faraday Symp. Chem. Soc. 371–375.
1978, 12, 101–114. 175. G. Quarin, Cl. Buess-Herman, L. Gierst,
147. S. Fletcher, A. Smith, Electrochim. Acta J. Electroanal. Chem. 1981, 123, 35–58.
176. Th. Wandlowski, G. R. Jameson,
1980, 25, 1019–1024.
R. de Levie, J. Phys. Chem. 1993, 97,
148. S. H. Liu, Solid State Phys. 1986, 39,
10 119–10 126.
207–211.
177. A. Bewick, M. Fleischmann, H. R. Thirsk,
149. N. J. Tao, C. Z. Li, F. Cunha et al., Scann.
Trans. Faraday Soc. 1962, 58, 2200–2209.
Microsc. 2002, in press.
178. R. D. Armstrong, M. Fleischmann, H. R.
150. T. A. Witten, L. M. Sander, Phys. Rev. Lett.
Thirsk, J. Electroanal. Chem. 1966, 11,
1981, 47, 1400–1403.
208–215.
151. L. Pospisil, J. Electroanal. Chem. 1986, 206, 179. J. A. Harrison, H. R. Thirsk, in Electroan-
269–283. alytical Chemistry (Ed.: A. J. Bard), Mar-
152. L. Pospisil, J. Phys. Chem. 1988, 92, cel Dekker, New York, 1971, pp. 67–148,
2501–2506. Vol. 5.
153. R. Philipp, J. Electroanal. Chem. 1990, 290, 180. C. Müller, J. Claret, M. Sartet, J. Electroanal.
67–78. Chem. 1987, 227, 147–158.
154. Th. Wandlowski, J. Electroanal. Chem. 1992, 181. M. H. Hölzle, U. Retter, D. M. Kolb, J. Elec-
333, 77–91. troanal. Chem. 1994, 371, 101–109.
155. U. Retter, Langmuir 2000, 16, 7752–7756. 182. R. Philipp, J. Dittrich, U. Retter et al., J.
156. E. Budevski, W. Bostanov, T. Vitanov et al., Electroanal. Chem. 1988, 250, 159–164.
Electrochim. Acta 1966, 11, 1697–1707. 183. Th. Wandlowski, Th. Dretschkow, J. Elec-
157. E. Budevski, W. Bostanov, T. Vitanov et al., troanal. Chem. 1997, 427, 105–112.
Phys. Status Solidi 1966, 13, 577–588. 184. R. G. Barradas, T. J. Vandernoot, J. Elec-
158. V. Bostanov, G. Staikov, D. K. Roe, J. Elec- troanal. Chem. 1982, 142, 107–119.
trochem. Soc. 1975, 122, 1301–1305. 185. R. G. Barradas, T. J. Vandernoot, J. Elec-
159. W. Obretenov, V. Bostanov, V. Popov, troanal. Chem. 1984, 176, 151–167.
J. Electroanal. Chem. 1982, 132, 273–276. 186. R. G. Barradas, M. Rennie, T. Vandernoot,
160. L. Gierst, C. Frank, G. Quarin et al., J. Electroanal. Chem. 1983, 144, 455–458.
J. Electroanal. Chem. 1981, 129, 353–363. 187. W. H. Mulder, J. H. O. J. Wijenberg,
161. R. Sridharan, R. de Levie, J. Electroanal. M. Sluyters-Rehbach et al., J. Electroanal.
Chem. 1987, 218, 287–296. Chem. 1989, 270, 7–19.
162. F. Thomas, Cl. Buess-Herman, L. Gierst, 188. Th. Wandlowski, J. Electroanal. Chem. 1992,
J. Electroanal. Chem. 1986, 214, 597–613. 333, 77–91.
189. R. Kaishev, B. Mutafschiew, Electrochim. 214. I. G. Abidov, V. Arakelyan, L. V. Cher-

Acta 1965, 19, 643–652. nomordik et al., Bioelectrochem. Bioenerg.
190. U. Retter, J. Electroanal. Chem. 1982, 136, 1979, 6, 37–63.
167–174. 215. D. Exerova, D. Kashchiev, Contemp. Phys.
191. W. Obretenov, I. Petrov, I. Nachev et al., J. 1986, 27, 429–436.
Electroanal. Chem. 1980, 109, 195–198. 216. J. J. Calvente, Z. Kovacova, M. D. Sanchez
192. R. G. Barradas, E. Bosco, Electrochim. Acta et al., Langmuir 1996, 12, 5696–5903.
1986, 31, 949–963. 217. D. F. Yang, M. Morin, J. Electroanal. Chem.
193. R. Guidelli, M. L. Foresti, M. Innocenti, J. 1997, 429, 1–5.
Phys. Chem. 1996, 100, 47–53. 218. D. F. Yang, M. Morin, J. Electroanal. Chem.
194. C. Donner, L. Pohlmann, Langmuir 1999, 1998, 441, 173–181.
15, 4898–4915. 219. M. H. Hölzle, Th. Wandlowski, D. M. Kolb,
195. R. A. Ramos, P. A. Rikvold, M. A. Novotney, Surf. Sci. 1995, 335, 281–290.
Phys. Rev. B 1999, 59, 9053–9069. 220. M. van Krieken, C. Buess-Herman, Elec-
196. P. A. Rikvold, A. Wieckowski, R. A. Ramos, trochim. Acta 1998, 43, 2831–2841.
Mater. Res. Soc. Symp. Proc. 1997, 451, 221. S. Bare, C. Buess-Herman, Colloids Surf., A
69–81. 1998, 134, 181–191.
197. P. A. Rikvold, G. G. Brown, M. A. Novotny 222. Th. Dretschkow, D. Lampner, Th. Wand-
et al., Colloids Surf. A 1998, 134, 3–14. lowski, J. Electroanal. Chem. 1998, 458,
198. G. Brown, P. A. Rikvold, M. A. Novotney 121–138.
et al., J. Electrochem. Soc. 1999, 146, 223. S. T. Chui, Phys. Rev. Lett. 1982, 48,
1035–1040. 933–935.
199. K. Ataka, G. Nishina, W. B. Cai et al., Elec- 224. K. I. Strandburg, Rev. Mod. Phys. 1988, 60,
161–207.
trochem. Commun. 2000, 2, 417–421.
225. Th. Wandlowski, 2001, unpublished
200. A. C. Finnefrock, K. L. Ringland, J. D. Brock
226. J. M. Kosterlitz, D. J. Thouless, J. Phys. C
et al., Phys. Rev. Lett. 1998, 81, 3459–3462.
1973, 6, 1181–1203.
201. Th. Wandlowski, D. M. Lampner, S. M.
227. B. I. Halperlin, D. R. Nelson, Phys. Rev. Lett.
Lindsay, J. Electroanal. Chem. 1996, 404,
1978, 41, 121–124, 519–521.
215–226.
228. A. P. Young, Phys. Rev. B 1979, 19,
202. J. G. Dash, Rev. Mod. Phys. 1999, 71,
1855–1866.
1737–1743. 229. D. C. Grahame, Chem. Rev. 1947, 41,
203. R. Sridharan, R. de Levie, J. Electroanal. 441–501.
Chem. 1997, 230, 241–256. 230. J. O’. M. Bockris, B. E. Conwasy, E. Yeager,
204. Th. Wandlowski, Habilitatition, University (Eds.), Comprehensive Treatise of Electrochem-
Press Halle, Halle, 1990. istry, Plenum Press, New York, 1980, Vol. 1.
205. U. Retter, W. Kant; Thin Solid Films 1995, 231. A. Hamelin, T. Vitanov, E. Sevastyanonov
256, 89–93, 265, 101–106. et al., J. Electroanal. Chem. 1983, 145,
206. C. Franck, PhD Thesis, ULB, Brussels, 225–264.
1986. 232. A. Hamelin, in Modern Aspects of Electro-
207. M. Poelman, C. Buess-Herman, J. P. Badi- chemistry (Ed.: J. O’. M. Bockris), Plenum
ali, Langmuir 1999, 15, 2194–2201. Press, New York, 1987, pp. 1–101.
208. H. Jehring, U. Retter, E. Horn, J. Elec- 233. A. Hamelin, Valette, R. Parsons, Z. Phys.
troanal. Chem. 1983, 149, 153–166. Chem. N.F. 1978, 113, 71–89.
209. R. Srinivasan, P. Gopalan, J. Phys. Chem. 234. J. Lipkowski, Z. Shi, A. Chen et al., Elec-
1993, 97, 8770–8775. trochim. Acta 1998, 43, 2875–2888.
210. N. J. Tao, Z. Shi, J. Phys. Chem. 1994, 98, 235. O. M. Magnussen, J. Hageböck, J. Hotlos
1464–1471. et al., Faraday Discuss. 1992, 94, 329–338.
211. W. H. Mulder, J. Electroanal. Chem. 1994, 236. M. Kleinert, A. Cuesta, L. A. Kibler et al.,
366, 287–293. Surf. Sci. Lett. 1999, 430, L521–L526.
212. M. Fleischmann, H. R. Thirsk, Electrochim. 237. D. A. Scherson, D. M. Kolb, J. Electroanal.
Acta 1964, 9, 757–771. Chem. 1984, 176, 353–357.
213. M. Fleischmann, H. R. Thirsk, J. Elec- 238. G. Gouy, Ann. Chim. Phys. 1903, 29,
trochem. Soc. 1963, 110, 688–698. 145–242.
239. G. Gouy, Ann. Chim. Phys. 1906, 8, 263. J. X. Wang, R. R. Adzic, T. Wandlowski
291–364. et al., unpublished.
240. G. Gouy, Ann. Chim. Phys. 1906, 9, 75–139. 264. J. X. Wang, G. M. Watson, B. M. Ocko,
241. W. Schmickler; Chem. Rev. 1996, 96, unpublished.
3177–3200. 265. B. M. Ocko, O. M. Magnussen, J. X. Wang
242. D. M. Kolb, Z. Phys. Chem. N.F. 1987, 154, et al., Physica B 1996, 221, 238–244.
179–199. 266. G. Aloisi, A. M. Funtikov, T. Will, J. Elec-
243. T. Iwasita, F. C. Nart, in Advances in Elec- troanal. Chem. 1994, 370, 297–300.
trochemical Science and Engineering (Eds.: 267. T. Yamada, K. Ogaki, S. Okubo et al., Surf.
H. Gerischer, C. W. Tobias), VCH, New Sci. 1996, 369, 321–335.
York, 1995, pp. 126–216. 268. B. M. Ocko, Th. Wandlowski, Mater. Res.
244. G. Pacchioni, P. S. Bagus, M. R. Philpott, Soc. Symp. Proc. 1997, 451, 55–64.
Z. Phys. D 1989, 12, 543–546. 269. T. Teshima, K. Ogaki, K. Itaya, J. Phys.
245. A. Ignaczak, J. A. N. F. Gomes, J. Elec- Chem. B 1997, 101, 2046–2053.
troanal. Chem. 1997, 420, 71–78. 270. P. Broekmann, M. Wilms, M. Kruft et al., J.
246. M. T. M. Koper, R. A. van Santen, Surf. Sci. Electroanal. Chem. 1999, 467, 307–324.
1999, 422, 118–131. 271. J. Inukai, Y. Osawa, K. Itaya, J. Phys. Chem.
247. E. Spohr, in Advances in Electrochemical B 1998, 102, 1034–1040.
Sciences and Engineering (Eds.: R. C. Alkire, 272. P. Broekmann, M. Anastasescu, A. Spaenig
D. M. Kolb), Wiley-VCH, Weinheim, 1999, et al., J. Electroanal. Chem. 2001, 500,
pp. 1–76, Vol. 6. 241–254.
248. A. N. Frumkin, Z. Phys. Chem. 1933, 164A, 273. S. L. Yau, C. M. Vitus, B. C. Schardt, J. Am.
121–133. Chem. Soc. 1990, 112, 3677–3679.
274. C. A. Lucas, N. M. Markovic, P. N. Ross,
249. D. M. Kolb, Prog. Surf. Sci. 1996, 51,
Phys. Rev. B 1997, 55, 7964–7971.
109–173.
275. R. Vogel, I. Kamphausen, H. Baltruschat,
250. A. R. Despic, in Kinetics and Mechanisms
Ber. Bunsen-Ges. Phys. Chem. 1992, 96,
of Electrode Processes (Eds.: B. E. Conway,
525–530.
J. O’. M. Bockris, E. Yeager et al.), Plenum
276. J. Inukai, Y. Osawa, M. Wakisaka et al., J.
Press, New York, 1993, pp. 451–528.
Phys. Chem. B 1998, 102, 3498–3505.
251. R. R. Adzic, J. X. Wang, Electrochim. Acta
277. J. M. Orts, R. Gómez, J. M. Feliu, J. Elec-
2000, 45, 4203–4210. troanal. Chem. 1999, 467, 11–19.
252. X. Gao, M. J. Weaver, J. Am. Chem. Soc. 278. K. Sashikata, Y. Matsui, K. Itaya et al., J.
1992, 114, 8544–8551. Phys. Chem. 1996, 100, 20 027–20 034.
253. O. M. Magnussen, J. X. Wang, R. R. Adzic 279. L. J. Wan, S. L. Yau, G. M. Swain et al., J.
et al., J. Phys. Chem. 1996, 100, 5500–5508. Electroanal. Chem. 1995, 381, 105–111.
254. O. M. Magnussen, B. M. Ocko, R. R. Adzic 280. P. Müller, S. Ando, T. Yamada et al., J.
et al., Phys. Rev. B 1995, 51, 5510–5513. Electroanal. Chem. 1999, 467, 282–290.
255. B. M. Ocko, G. M. Watson, J. Wang, J. Phys. 281. H. H. Farell, in The Chemical Physics of Solid
Chem. 1994, 98, 897–906. Surfaces and Heterogeneous Catalysis (Eds.:
256. X. Gao, G. J. Edens, F. C. Liu et al., J. Phys. D. A. King, D. P. Woodruff), Elsevier, New
Chem. 1994, 98, 8086–8095. York, 1987, pp. 225–272, Vol. 36.
257. X. Gao, G. J. Edens, M. J. Weaver, J. Phys. 282. P. A. Dowden, Crit. Rev. Solid State Mater.
Chem. 1994, 98, 8074–8085. Sci. 1987, 13, 191–210.
258. T. Yamada, N. Batina, K. Itaya, J. Phys. 283. G. N. Kastanas, B. E. Koel, Appl. Surf. Sci.
Chem. 1995, 99, 8817–8823. 1993, 64, 235–249.
259. B. M. Ocko, O. M. Magnussen, J. X. Wang 284. P. J. Goddard, R. M. Lambert, Surf. Sci.
et al., Phys. Rev. B 1996, 53, R7654–R7657. 1977, 67, 180–194.
260. J. X. Wang, G. M. Watson, B. M. Ocko, J. 285. K. Kern, G. Comsa, in Chemistry
Phys. Chem. 1996, 100, 6672–6677. and Physics of Solid Surfaces (Eds.:
261. A. Cuesta, D. M. Kolb, Surf. Sci. 2000, 465, R. Vanselow, R. Howe), Springer, Berlin,
310–316. 1988, pp. 65–108.
262. S. Zou, X. Gao, M. J. Weaver, Surf. Sci. 286. P. Bak, D. Mukamel, I. Villain et al., Phys.
2000, 452, 44–57. Rev. B 1979, 19, 1610–1613.
287. M. Kardar, A. N. Berker, Phys. Rev. Lett. 312. H. Uchida, N. Ikeda, M. Watanabe, J. Elec-
1982, 48, 1552–1555. troanal. Chem. 1997, 424, 5–12.
288. M. F. Toney, J. G. Gordon, M. G. Samant 313. Y. Shingaya, M. Ito, Electrochim. Acta 1998,
et al., Phys. Rev. B 1992, 45, 9362–9374. 44, 745–751.
289. X. Wang, R. Chen, Y. Wang et al., J. Phys. 314. K. Ataka, M. Osawa, Langmuir 1998, 14,
Chem. B 1998, 102, 7568–7576. 951–959.
290. Z. Racz, Phys. Rev. B 1980, 21, 4012–4016. 315. I. R. de Moraes, F. C. Nart, J. Electroanal.
291. S. J. Mitchell, G. Brown, P. A. Rikvold, J. Chem. 1999, 461, 110–120.
Electroanal. Chem. 2000, 493, 68–74. 316. Z. Shi, J. Lipkowski, S. Mirwald et al.,
292. W. Haiss, PhD Thesis, Free University J. Electroanal. Chem. 1995, 396, 115–124.
Berlin, Berlin, 1994. 317. T. Nishizawa, T. Namata, Y. Kinoshita
293. F. Silva, A. Martins, in Interfacial Elec- et al., Surf. Sci. 1996, 367, L73–L78.
trochemistry (Ed.: A. Wieckowski), Marcel 318. E. M. Patrito, P. Paredes Olivera, H. Sellers,
Dekker, New York, 1999, pp. 427–461. Surf. Sci. 1997, 380, 264–282.
294. H. Angerstein-Kozlowska, B. E. Conway, 319. D. M. Kolb, M. Przasnyski, H. Gerischer,
A. Hamelin et al., J. Electroanal. Chem. 1987, J. Electroanal. Chem. 1974, 54, 25–38.
228, 429–453. 320. J. W. Schultze, D. Dickertmann, Faraday
295. J. Clavilier, J. Electroanal. Chem. 1980, 107, Symp. 1977, 12, 36.
211–216. 321. J. W. Schultze, D. Dickertmann, Faraday
296. J. M. Feliu, M. T. Valls, A. Aldaz et al., Symp. 1977, 12, 172.
J. Electroanal. Chem. 1993, 345, 475–481. 322. J. W. Schultze, D. Dickertmann, Surf. Sci.
297. F. Silva, M. J. Sottomayor, A. Martins, 1976, 54, 489–505.
J. Electroanal. Chem. 1994, 375, 395–399. 323. E. Leiva, Electrochim. Acta 1996, 41,
298. G. J. Edens, X. Gao, M. J. Weaver, J. Elec- 2185–2206.
troanal.Chem. 1994, 375, 357–366. 324. W. J. Lorenz, H. D. Hermann, N. Wütrich
299. Th. Dretschkow, Th. Wandlowski, Ber. et al., J. Electrochem. Soc. 1974, 121,
Bunsen-Ges. Phys. Chem. 1997, 101, 1167–1177.
749–757. 325. R. Adzic, E. Yeager, B. D. Cahan, J. Elec-
300. A. M. Funtikov, U. Linke, U. Stimming trochem. Soc. 1974, 121, 474–484.
et al., Surf. Sci. 1995, 324, L343–L378. 326. K. Jüttner, W. J. Lorenz, Z. Phys. Chem. N.F.
301. A. M. Funtikov, U. Stimming, R. Vogel, J. 1980, 122, 163–185.
Electroanal. Chem. 1997, 428, 147–153. 327. G. Kokkinidis, J. Electroanal. Chem. 1986,
302. L. J. Wan, S. L. Yau, K. Itaya, J. Phys. Chem. 201, 217–236.
1995, 99, 9507–9513. 328. R. Adzic, in Advances in Electrochemistry
303. M. Wilms, P. Broekmann, M. Kruft et al., and Electrochemical Engineering (Eds.:
Surf. Sci. 1998, 404, 83–86. H. Gerischer, C. Tobias), John Wiley &
304. M. Wilms, P. Broekmann, C. Stuhlmann Sons, New York, 1984, pp. 159, Vol. 13.
et al., Surf. Sci. 1998, 416, 121–140. 329. K. Engelsmann, W. J. Lorenz, E. Schmidt,
305. P. Broekmann, M. Wilms, M. Kruft et al., J. Electroanal. Chem. 1980, 114, 1–10.
J. Electroanal. Chem. 1999, 467, 307–324. 330. O. Melroy, K. Kanazawa, J. J. G. Gordon
306. W. H. Li, R. J. Nichols, J. Electroanal. Chem. et al., Langmuir 1986, 2, 697–700.
1998, 456, 153–160. 331. G. L. Richmond, M. Robinson, V. L. Shan-
307. L. J. Wan, M. Hara, J. Inukai et al., J. Phys. non, Prog. Surf. Sci. 1988, 28, 1–70.
Chem. B 1999, 103, 6978–6983. 332. G. Lehmpfuhl, Y. Uchida, M. S. Zei et al.,
308. L. J. Wan, T. Suzuki, K. Sashikata et al., in Imaging of Surfaces and Interfaces (Eds.:
J. Electroanal. Chem. 2000, 484, 189–193. J. Lipkowski, P. N. Ross), John Wiley &
309. A. Cuesta, M. Kleinert, D. M. Kolb, Phys. Sons, New York, 1999, pp. 57–98.
Chem. Chem. Phys. 2000, 2, 5684–5690. 333. R. Kötz in Advances in Electrochemical Sci-
310. H. Angerstein-Kozlowska, B. E. Conway, ence and Engineering (Eds.: H. Gerischer,
A. Hamelin et al., Electrochim. Acta 1986, Ch. Tobias), VCH, Weinheim, 1990, pp.
31, 1051–1061. 75–109, Vol. 1.
311. Z. Shi, J. Lipkowski, M. Gamboa et al., 334. M. Toney, O. R. Melroy, in Electrochemical
J. Electroanal. Chem. 1994, 366, 317–326. Interfaces – Modern Techniques for in situ
Characterization (Ed.: H. D. Abruna), VCH, 358. L. Blum, M. D. Legault, D. A. Huck-

Berlin, 1991, pp. 57–132. aby, in Interfacial Electrochemistry (Ed.:
335. A. A. Gewirth, B. K. Niece, Chem. Rev. 1997, A. Wieckowski), Marcel Dekker, New York,
97, 1129–1162. 1999, pp. 19–31.
336. G. Staikov, W. J. Lorenz, E. Budevski, in 359. G. Brown, P. A. Rickvold, M. A. Novotny
Imaging of Surfaces and Interfaces (Eds.: et al., J. Electrochem. Soc. 1999, 146,
J. Lipkowski, P. N. Ross), John Wiley & 1035–1040.
Sons, New York, 1999, pp. 1–56. 360. M. H. Hölzle, PhD Thesis, University of
337. M. H. Hölzle, V. Zwing, D. M. Kolb, Elec- Ulm, Germany, 1995.
trochim. Acta 1995, 40, 1237–1247. 361. E. Bosco, S. K. Rangarajan, J. Chem. Soc.,
338. I. H. Omar, H. J. Pauling, K. Jüttner, J. Elec- Faraday Trans. 1981, 77, 1673–1696.
trochem. Soc. 1993, 140, 2187–2192. 362. K. Jüttner, G. Staikov, W. J. Lorenz et al.,
339. Z. Shi, J. Lipkowski, J. Electroanal. Chem. J. Electroanal. Chem. 1977, 80, 67–80.
1994, 364, 289–294. 363. G. Horanyi, E. M. Rizmayer, P. Jovi, J. Elec-
340. Z. Shi, J. Lipkowski, J. Electroanal. Chem. troanal. Chem. 1983, 152, 211–222.
1994, 365, 303–309. 364. M. S. Zei, G. Qiao, G. Lehmpfuhl et al., Ber.
341. O. M. Magnussen, J. Hotlos, R. Nichols Bunsen-Ges. Phys. Chem. 1987, 91, 349–353.
et al., Phys. Rev. Lett. 1990, 64, 2929–2932. 365. E. Herrero, S. Glazier, H. D. Abruna,
342. T. Hachiya, H. Honbo, K. Itaya, J. Elec- J. Phys. Chem. B 1998, 102, 9825–9833.
troanal. Chem. 1991, 315, 275–291. 366. S. Wu, J. Lipkowski, T. Tyliszcza et al., Prog.
343. T. Will, M. Dietterle, D. M. Kolb, in Surf. Sci. 1995, 50, 227–236.
Nanoscale Probes of Solid/Liquid Interfaces, 367. Z. Shi, S. Wu, J. Lipkowski, J. Electroanal.
NATO ASI Series E (Eds.: A. A. Gewirth, Chem. 1995, 284, 171–177.
H. Siegenthaler), Kluwer, Dordrecht, 1995, 368. Z. Shi, J. Lipkowski, J. Electroanal. Chem.
pp. 137–162, Vol. 288. 1994, 369, 283–287.
344. S. Manne, P. K. Hansma, J. Massie et al., 369. R. Michaelis, PhD Thesis, Free University
Science 1991, 251, 183–186. Berlin, Berlin, 1990.
345. N. Ikemiya, S. Miyaoka, S. Hara, Surf. Sci. 370. Z. Shi, S. Wu, J. Lipkowski, Electrochim.
1994, 311, L641–L648. Acta 1995, 40, 9–15.
346. N. Ikemiya, S. Miyaoka, S. Hara, Surf. Sci. 371. N. Batina, T. Will, D. M. Kolb, Faraday Dis-
1995, 327, 261–273. cuss. 1992, 94, 93–106.
347. D. B. Parry, M. G. Samant, H. Seki et al., 372. F. A. Möller, O. M. Magnussen, R. J. Behm,
Langmuir 1993, 9, 1878–1887. Phys. Rev. B 1995, 51, 2484–2490.
348. L. Blum, H. D. Abruna, J. White et al., 373. R. Randler, M. Dietterle, D. M. Kolb,
J. Chem. Phys. 1986, 85, 6732–6738. Z. Phys. Chem. 1999, 208, 43–56.
349. O. R. Melroy, M. J. Samant, G. L. Borges 374. F. A. Möller, O. M. Magnussen, R. J. Behm,
et al., Langmuir 1988, 4, 728–732. Electrochim. Acta 1995, 40, 1259–1265.
350. M. F. Toney, J. N. Howard, J. Richer et al., 375. D. M. Kolb, R. Kötz, K. Yamamoto, Surf.
Phys. Rev. Lett. 1995, 75, 4472–4475. Sci. 1979, 87, 20–30.
351. J. G. Gordon, O. R. Melroy, M. F. Toney, 376. C. L. Scortichini, C. N. Reilley, J. Elec-
Electrochim. Acta 1995, 40, 3–8. troanal. Chem. 1982, 139, 233–245.
352. G. L. Borges, K. K. Kanazawa, J. G. Gordon 377. C. L. Scortichini, C. N. Reilley, J. Elec-
et al., J. Electroanal. Chem. 1994, 364, troanal. Chem. 1982, 139, 255–260.
281–284. 378. D. M. Kolb, K. A. Jaaf-Golze, M. S. Zei,
353. H. Uchida, M. Hiei, M. Watanabe, J. Elec- Dechema – Monographien 1986, 12, 53–64.
troanal. Chem. 1998, 452, 97–106. 379. Ch. Nishihara, A. Nozoye, J. Electroanal.
354. Y. Nakai, M. S. Zei, D. M. Kolb et al., Ber. Chem. 1995, 386, 75–82.
Bunsen-Ges. Phys. Chem. 1984, 88, 340–345. 380. Ch. Nishihara, A. Nozoye, J. Electroanal.
355. J. Zhang, Y. E. Sung, P. A. Rikvold et al., Chem. 1995, 396, 139–142.
J. Chem. Phys. 1996, 104, 5699–5712. 381. R. Gomez, J. M. Feliu, H. D. Abruna, Lang-
356. M. Legault, L. Blum, D. A. Huckaby, J. Elec- muir 1994, 10, 4315–4323.
troanal. Chem. 1996, 409, 79–86. 382. L. Buller, E. Herrero, R. Gomez et al.,
357. D. A. Huckaby, M. D. Legault, L. Blum, J. Chem. Soc., Faraday Trans. 1996, 92,
J. Chem. Phys. 1998, 109, 3600–3606. 3757–3762.
383. P. Berenz, S. Tillmann, H. Massong et al., 410. A. Bittner, J. Winterlin, G. Ertel, Surf. Sci.
Electrochim. Acta 1998, 43, 3035–3043. 1997, 376, 267–278.
384. E. A. Meguid, P. Berenz, H. Baltruschat, 411. A. A. Aki, G. A. Attard, R. Price et al.,
J. Electroanal. Chem. 1999, 467, 50–59. J. Chem. Soc., Faraday Trans. 1995, 91,
385. E. Herrero, S. Glazier, L. J. Buller et al., 3585–3595.
J. Electroanal. Chem. 1999, 461, 121–130. 412. N. Markovic, H. A. Gasteiger, P. N. Ross,
386. L. Buller, E. Herrero, R. Gomez et al., Langmuir 1995, 11, 4098–4108.
J. Phys. Chem. B 2000, 104, 5932–5939. 413. I. A. Tidswell, C. A. Lucas, N. M. Markovic
387. K. Varga, P. Zelenay, A. Wieckowski, J. Elec- et al., Phys. Rev. B 1995, 51, 10 205–10 208.
troanal. Chem. 1992, 330, 453–467. 414. D. Dickertmann, F. D. Koppitz, J. W.
388. G. Horanyi, J. Electroanal. Chem. 1974, 55, Schultze, Electrochim. Acta 1976, 21,
45–51. 967–971.
389. Y. Shingaya, H. Matsumoto, H. Ogasawara 415. W. J. Lorenz, E. Schmidt, G. Staikov et al.,
et al., Surf. Sci. 1995, 335, 23–31. Faraday Symp. Chem. Soc. 1977, 12, 14–23.
390. I. Oda, Y. Shingaya, H. Matsumoto et al., 416. K. Jüttner, G. G. Staikov, W. J. Lorenz et al.,
J. Electroanal. Chem. 1996, 409, 95–101. J. Electroanal. Chem. 1977, 80, 67–80.
391. M. Watanabe, H. Uchida, N. Ikeda, J. Elec- 417. H. Bort, K. Jüttner, W. J. Lorenz et al.,
troanal. Chem. 1995, 380, 255–260. J. Electroanal. Chem. 1978, 90, 413–424.
392. H. Matsumoto, I. Oda, J. Inukai et al., 418. G. Staikov, K. Jüttner, W. J. Lorenz et al.,
J. Electroanal. Chem. 1993, 356, 275–280. Electochim. Acta 1978, 23, 305–313.
393. H. Matsumoto, I. Oda, J. Inukai et al., 419. G. Staikov, K. Jüttner, W. J. Lorenz et al.,
J. Electroanal. Chem. 1994, 379, 223–231. Electochim. Acta 1978, 23, 319–324.
394. G. Beitel, O. M. Magnussen, R. J. Behm, 420. K. Takayanagi, D. M. Kolb, K. Kambe et al.,
Surf. Sci. 1995, 336, 19–26. Surf. Sci. 1980, 100, 407–422.
395. K. Sashikata, N. Furuya, K. Itaya, J. Elec- 421. J. N. Jovicevic, V. D. Jovic, A. R. Despic,
troanal. Chem. 1991, 316, 361–368. Electrochim. Acta 1984, 29, 1625–1632.
396. Z. L. Wu, Z. H. Zang, S. L. Yau, Langmuir 422. V. D. Jovic, J. N. Jovicevic, A. R. Despic,
2000, 16, 3522–3528. Electrochim. Acta 1985, 30, 1455–1464.
397. H. S. Yee, H. D. Abruna, J. Phys. Chem. 423. V. D. Jovic, B. M. Jovic, A. R. Despic, J. Elec-
1993, 97, 6278–6288. troanal. Chem. 1990, 288, 229–243.
398. R. Durand, R. Faure, D. Aberdam et al., 424. K. J. Stevenson, D. W. Hatchett, H. S.
Electrochim. Acta 1992, 37, 1977–1982. White, Langmuir 1996, 12, 494–499.
399. N. Markovic, H. A. Gasteiger, C. A. Lucas 425. D. A. Koos, V. L. Shannon, G. L. Richmond,
et al., Surf. Sci. 1995, 335, 91–100. J. Phys. Chem. 1990, 94, 2091–2098.
400. C. A. Lucas, N. M. Markovic, I. M. Tidswell 426. U. Müller, D. Carnal, H. Siegenthaler et al.,
et al., Physica B 1996, 221, 245–250. Phys. Rev. B 1992, 46, 12 899–12 901.
401. E. Herrero, L. J. Buller, J. Li et al., Elec- 427. D. Carnal, P. I. Oden, U. Müller et al., Elec-
trochim. Acta 1998, 44, 983–992. trochim. Acta 1995, 40, 1223–1235.
402. R. Michaelis, M. S. Zei, R. S. Zhai et al., 428. U. Schmidt, S. Vinzelberg, G. Staikov, Surf.
J. Electroanal. Chem. 1992, 339, 299–310. Sci. 1996, 348, 261–279.
403. R. Michaelis, D. M. Kolb J. Electroanal. 429. M. G. Samant, G. L. Borges, J. G. Gordon
Chem. 1992, 328, 341–348. et al., J. Am. Chem. Soc. 1987, 109,
404. D. Aberdam, Y. Gauthier, R. Durand et al., 5970–5974.
Surf. Sci. 1994, 306, 114–124. 430. M. G. Samant, M. F. Toney, G. L. Borges
405. M. S. Zei, K. Wu, M. Eiswirthz et al., Elec- et al., Surf. Sci. 1988, 193, L29–L36.
trochim. Acta 1999, 45, 809–817. 431. M. F. Toney, J. G. Gordon, M. G. Samant
406. M. S. Zei, Z. Phys. Chem. 1999, 208, 77–91. et al., J. Phys. Chem. 1995, 99, 4733–4744.
407. Th. Wandlowski, E. Herrero, J. Feliu, 432. M. E. Hansen, E. Yeager, ACS Symp. 1988,
J. Electroanal. Chem., in preparation 378, 398–407.
408. C. A. Lucas, N. M. Markovic, P. N. Ross, 433. L. Laguren–Davidson, F. Lu, G. Salaita
Phys. Rev. B 1997, 56, 3651–3654. et al., Langmuir 1988, 4, 224–232.
409. A. A. Aki, G. A. Attard, J. Phys. Chem. B 434. E. Amman, MS Thesis, University of Bern,
1997, 101, 4597–4606. Bern, Switzerland, 1995.
435. K. Takayanagi, Surf. Sci. 1981, 104, 460. A. Hamelin, S. Morin, J. Richer et al.,
527–548. J. Electroanal. Chem. 1991, 304, 195–209.
436. E. Altman, R. Colton, Surf. Sci. 1993, 295, 461. A. Popov, R. Naneva, K. Dimitrov et al.,
437. A. Ullman, Chem. Rev. 1996, 96, 1533–1554. 462. B. Roelfs, H. Baumgärtel, Ber. Bunsen-Ges.
438. R. E. Pagano, R. Miller, J. Colloid. Interface Phys. Chem. 1994, 99, 677–681.
Sci. 1973, 45, 126–137. 463. B. Roelfs, E. Bunge, C. Schröter et al.,
439. A. Nelson, N. Auffret, J. Electroanal. Chem. J. Phys. Chem. B 1997, 101, 754–765.
1988, 244, 99–113. 464. M. H. Hölzle, D. Krznaric, D. M. Kolb,
440. M. Porter, T. B. Bright, D. Allara et al., J. Electroanal. Chem. 1995, 386, 235–239.
J. Am. Chem. Soc. 1987, 109, 3559–3568. 465. M. H. Hölzle, D. M. Kolb, Ber. Bunsen-Ges.
441. L. Strong, G. M. Whitsides, Langmuir 1988, Phys. Chem. 1994, 98, 330–335.
4, 546–558. 466. M. Scharfe, A. Hamelin, C. Buess-Herman,
442. J. Sagiv, J. Am. Chem. Soc. 1980, 102, 92–98. Electrochim. Acta 1995, 40, 61–67.
443. H. Finklea, D. A. Snider, J. Fedyk, Lang- 467. M. H. Hölzle, Th. Wandlowski, D. M. Kolb,
muir 1993, 9, 3360–3367. J. Electroanal. Chem. 1995, 394, 271–275.
444. P. Fenter, P. Eisenberger, K. S. Liang, Phys. 468. Th. Wandlowski, J. Electroanal. Chem. 1995,
Rev. Lett. 1993, 70, 2447–2450. 395, 83–89.
445. O. M. Magnussen, B. M. Ocko, M. Deutsch 469. Th. Wandlowski, M. H. Hölzle, Langmuir
et al., Nature 1996, 384, 250–252. 1996, 12, 6597–6615.
446. A. A. Kornyshev, I. Vilfan, Electrochim. Acta 470. D. Krznaric, B. Cosovic, M. H. Hölzle et al.,
1995, 40, 109–127. Ber. Bunsen-Ges. Phys. Chem. 1996, 100,
447. A. N. Frumkin, Z. Phys. Chem. 1925, 116, 1779–1790.
466–485. 471. H. Striegler, M. H. Hölzle, Th. Wandlowski
448. A. N. Frumkin, Z. Phys. 1926, 35, 792–802. et al., Poster 1a-41, presented at the 47th
449. B. Damaskin, O. A. Petrii, V. V. Batrakov, Annual Meeting of the ISE, November 1–6,
Adsorption of Organic Compounds on Elec- Budapest, 1996.
trodes, Plenum Press, New York 1972. 472. Th. Boland, B. Ratner, Langmuir 1994, 10,
450. R. Parsons, Chem. Rev. 1990, 90, 813–826. 3845–3852.
451. W. Lorenz, Z. Electrochem. 1958, 62, 473. W. Haiss, B. Roelfs, S. N. Port et al., J.
452. D. A. Bonell, Scanning Tunneling Microscopy 474. M. Futamata, J. Phys. Chem. B 2001, 105,
and Spectroscopy, VCH, New York, 1993. 6933–6942.
453. N. J. Tao, in Imaging of Surfaces and In- 475. S. Promkin, Th. Wandlowski, unpublished.
terfaces (Eds.: J. Lipkowski, P. N. Ross), 476. Th. Wandlowski, K. Ataka, D. Mayer, Lang-
John Wiley & Sons, New York, 1999, muir 2002, la u255854.
pp. 211–248. 477. N. J. Tao, J. A. DeRose, S. M. Lindsay,
454. W. H. Liu, W. Haiss, S. Floate et al., Lang- J. Phys. Chem. 1993, 97, 910–919.
muir 1999, 15, 4875–4883. 478. Th. Dretschkow, A. S. Dakkouri, Th. Wand-
455. K. Takamura, F. Kusu, in Methods od Bio- lowski, Langmuir 1997, 13, 2843–2856.
chemical Analysis (Ed.: D. Glick), John Wiley 479. Th. Dretschkow, Th. Wandlowski, Elec-
& Sons, New York, 1987, pp. 155, Vol. 32. trochim. Acta 1998, 43, 2991–3006.
456. H. H. Rotermund, K. Krischer, B. Pettinger, 480. S. Soverby, personal communication. 1998.
in Imaging of Surfaces and Interfaces (Eds.: 481. K. M. Richard, A. A. Gewirth, J. Phys. Chem.
J. Lipkowski, P. N. Ross), John Wiley & 1995, 99, 12 288–12 293.
Sons, New York, 1999, pp. 139–210. 482. G. Andreasen, M. E. Vela, R. C. Salvarezza
457. I. Burges, V. Zamlymny, G. Szymanski et al., Langmuir 1997, 13, 6814–6819.
et al., Langmuir 2001, 17, 3355–3367. 483. W. B. Cai, L. J. Wan, H. Noda et al., Lang-
458. Th. Wandlowski, B. M. Ocko, O. M. Mag- muir 1998, 14, 6992–6998.
nussen et al., J. Electroanal. Chem. 1996, 484. F. Cunha, N. J. Tao, Phys. Rev. Lett. 1995,
409, 155–164. 75, 2376–2379.
459. V. V. Batrakov, B. B. Damaskin, Y. B. Ipa- 485. F. Cunha, N. J. Tao, X. W. Wang et al.,
tov, Elektrokhimiya 1974, 10, 216–220. Langmuir 1996, 12, 6410–6418.
486. Th. Dretschkow, D. Lampner, Th. Wand- 510. Cl. Bues-Herman, L. Gierst, N. Vanlae-
lowski, J. Electroanal. Chem. 1998, 458, them-Meuree, J. Electroanal. Chem. 1981,
121–138. 123, 1–19.
487. Th. Dretschkow, Th. Wandlowski, J. Elec- 511. Th. Wandlowski, G. B. Jameson, R. de
troanal. Chem. 1999, 467, 207–216. Levie, J. Phys. Chem. 1993, 97,
488. Th. Dretschkow, Th. Wandlowski, Elec- 10 119–10 126.
trochim. Acta 1999, 45, 731–740. 512. Th. Wandlowski, R. de Levie, J. Electroanal.
489. H. Noda, T. Minoha, L. J. Wan et al., Chem. 1992, 329, 103–127.
J. Electroanal. Chem. 2000, 481, 62–68. 513. R. Guidelli, M. L. Foresti, J. Electroanal.
490. D. Mayer, K. Ataka, Th. Wandlowski, Chem. 1986, 197, 123–141.
J. Electroanal. Chem. 2002, in press. 514. Th. Wandlowski, P. Chaiyasith, H. Baum-
gärtel, J. Electroanal. Chem. 1992, 346,
491. F. Cunha, Q. Jing, N. J. Tao, Surf. Sci. 1997,
271–279.
389, 19–28.
515. D.M. Mohilner, in Electroanal. Chem. (Ed.:
492. O. Dominguez, L. Echegoyen, F. Cunha
A. J. Bard), Marcel Dekker, New York, 1966,
et al., Langmuir 1998, 14, 821–824.
pp. 241–405, Vol. 1.
493. J. Pan, S. M. Lindsay, N. J. Tao, Langmuir 516. J. Lipkowski, W. Schmickler, D. M. Kolb
1993, 9, 1556–1560. et al., J. Electroanal. Chem. 1998, 452,
494. E. Bunge, R. J. Nichols, H. Baumgärtel 193–197.
et al., Ber. Bunsen-Ges. Phys. Chem. 1995, 517. S. Wu, J. Lipkowski, O. M. Magnussen
99, 1243–1246. et al., J. Electroanal. Chem. 1998, 446, 67–77.
495. E. Bunge, R. J. Nichols, B. Roelfs et al., 518. L. Stolberg, J. Lipkowski, in Adsorption of
Langmuir 1996, 12, 3060–3066. Organic Molecules at Metal Electrodes (Eds.:
496. A. S. Dakkouri, D. M. Kolb, R. Edelstein, J. Lipkowski, P. N. Ross), VCH, New York,
Shima et al., Langmuir 1996, 12, 1992, pp. 171–238.
2849–2852. 519. U. Retter, H. Jehring, V. Vetterl, J. Elec-
497. L. J. Wan, H. Noda, Y. Hara et al., J. Elec- troanal. Chem. 1974, 57, 391–397.
troanal. Chem. 2000, 489, 68–75. 520. B. B. Damaskin, N. K. Akhmetov, Sov. Elec-
498. L. J. Wan, M. Terashima, H. Noda et al., J. trochem. 1979, 15, 478–481.
Phys. Chem. B 2000, 104, 3563–3569. 521. Yu. Kharkats, J. Electroanal. Chem. 1980,
499. N. Batina, M. Kunikate, K. Itaya, J. Elec- 115, 75–88.
troanal. Chem. 1995, 405, 245–250. 522. Yu. Ya. Gurevich, Yu. I. Kharkats, J. Elec-
500. M. Kunikate, N. Batina, K. Itaya, Langmuir troanal. Chem. 1978, 86, 245–258.
1995, 11, 2337–2340. 523. U. Retter, H. Lohses, J. Electroanal. Chem.
501. K. Ogaki, N. Batina, M. Kunikate et al., J. 1982, 134, 243–250.
Phys. Chem. 1996, 100, 7185–7190. 524. P. Nikitas, J. Electroanal. Chem. 1991, 300,
502. S. L. Yau, Y. G. Kim, K. Itaya, J. Am. Chem. 607–628.
Soc. 1996, 118, 7795–7803. 525. P. Nikitas, Electrochim. Acta 1991, 36,
447–457.
503. L. J. Wan, K. Itaya, Langmuir 1997, 13,
526. P. Nikitas, S. Andoniou, J. Electroanal.
7173–7179.
Chem. 1994, 375, 339–356.
504. R. Srinivasan, J. C. Murphy, R. Fainchtein
527. U. Retter, J. Electroanal. Chem. 1984, 165,
et al., J. Electroanal. Chem. 1994, 312,
221–230.
293–300. 528. U. Retter, J. Electroanal. Chem. 1987, 236,
505. R. Srinivasan, R. Gopalan, J. Phys. Chem. 21–30.
506. N. J. Tao, Z. Shi, Surf. Sci. 1994, 321, 41–48.
L149–L156. 530. U. Retter, Electrochim. Acta 1996, 41,
507. N. J. Tao, Z. Shi, J. Phys. Chem. 1994, 98, 2171–2174.
7422–7426. 531. R. Srinivasan, R. de Levie, S. K. Rangarajan,
508. N. J. Tao, Phys. Rev Lett. 1996, 76, Chem. Phys. Lett. 1987, 142, 43–47.
4066–4069. 532. C. Young, Phys. Rev. 1952, 85, 808–816.
509. L. J. Wan, C. Wan, C. Bai et al., J. Phys. 533. Y. I. Kharkats, U. Retter, J. Electroanal.
Chem. B 2001, 105, 8399–8402. Chem. 1990, 287, 363–367.
534. U. Retter, J. Electroanal. Chem. 1992, 329, 548. P. Steel, Coord. Chem. Rev. 1990, 106,
81–89. 227–265.
535. R. Sridharan, R. de Levie, J. Phys. Chem. 549. E. B. Constable, Adv. Inorg. Chem. 1989, 34,
1982, 86, 4489, 4490. 67–138.
536. R. Philipp, U. Retter, J. Dittrich et al., Elec- 550. L. L. Merritt, E. Schroeder, Acta Crystallogr.
trochim. Acta 1987, 32, 1671–1677. 1956, 9, 801–804.
537. U. Retter, V. Vetterl, J. Jursa, J. Electroanal. 551. E. A. Mambetkaziev, A. M. Shaldybaeva,
Chem. 1989, 274, 1–9. V. N. Statsyuk et al., Sov. Electrochem. 1975,
538. Th. Wandlowski, J. Electroanal. Chem. 1991, 11, 1643–1645.
302, 233–253. 552. N. K. Akhmetov, R. I. Kaganovich, E. A.
539. S. Laushera, E. Stenina, Sov. Electrochem. Mambetkaziev et al., Sov. Electrochem. 1977,
1992, 28, 78–87. 13, 248–250.
540. V. S. Griffiths, J. B. Westermore, J. Chem. 553. N. K. Akhmetov, R. I. Kaganovich, E. A.
Soc. 1963, 4941–4945. Mambetkaziev et al., Sov. Electrochem. 1978,
541. L. K. Partridge, A. C. Tansley, A. C. Porter, 14, 1534–1537.
Electrochim. Acta 1967, 12, 1573–1580. 554. L. Pospisil, J. Kuta, J. Electroanal. Chem.
542. B. B. Damaskin, S. L. Dyatkina, S. I. 1979, 101, 391–398.
Petrochenko, Sov. Electrochem. 1969, 5, 555. M. Kim, K. Itoh, J. Phys. Chem. 1987, 91,
935–941. 126–132.
543. A. A. Moussa, H. A. Ghaly, M. M. Abou- 556. D. Yang, D. Bizotto, J. Lipkowski et al.,
Romia, Electrochim. Acta 1975, 20, 485–497. J. Phys. Chem. 1994, 98, 7083–7089.
544. L. Gierst, C. Franck, G. Quarin et al., 557. M. Hoon-Khosla, W. R. Fawcett, J. D. God-
J. Electroanal. Chem. 1981, 129, 353–363. dard et al., Langmuir 2000, 16, 2356–2369.
545. W. Obretenov, V. Bostanov, V. Popov, 558. J. Trost, J. Wintterlin, G. Ertl, Surf. Sci.
J. Electroanal. Chem. 1982, 132, 273–276. 1995, 329, L583–L587.
546. M. Poelman, C. Buess-Herman, J. P. Badi- 559. G. E. Piorier, Chem. Rev. 1997, 97,
ali, Langmuir 1999, 15, 2194–2201. 1117–1129.
547. A. Popov, N. Dimitrov, T. Vitanov, Elec-
trochim. Acta 1992, 37, 2373–2376.
471
4.1 Thus, it is essential to clarify the effects

Atomically Controlled Electrochemical of these factors on the growth mode to
Deposition and Dissolution of Noble Metals obtain highly ordered metal layers.
Quantitative understanding of the elec-
Shen Ye and Kohei Uosaki trochemical deposition and dissolution of
Hokkaido University, Sapporo, Japan
metals became possible since Clavilier
4.1.1 established the technique to prepare a well-
Introduction defined surface of a platinum single-crystal
electrode in 1980 [9]. Furthermore, the dis-
Deposition and dissolution processes of covery of scanning tunneling microscopy
noble metals are important not only (STM) [10] and atomic force microscopy
for fundamental science but also for (AFM) [11] soon after, made the study on
applications. As shown in Fig. 1, the noble the surface structure of the electrode in situ
metals are widely used in a variety of possible at atomic resolution. The surface
fields such as catalysis, for automobile structures of many noble metal single-
and chemical reactions, fabrication of crystal electrodes, such as Au(hkl) [12, 13],
electronics devices and jewelry, batteries Pt(hkl) [14, 15], Pd(hkl) [16], Rh(hkl) [17],
and corrosion protection [1, 2]. Deposition and Ir(hkl) [18], have been investigated
of a highly ordered noble metal is usually by STM under electrochemical control
carried out by physical means such as at atomic resolution. Reality involves un-
vacuum deposition [3–6] and chemical derstanding the relationship between the
vapor deposition (CVD) [7, 8]. electrochemical behavior and the surface
Electrochemical deposition is more eco- structure of these electrodes.
nomical and convenient than the growth in Recently, in addition to the in situ
vacuum and has been used for a long time, STM/AFM, many other surface-analysis
but the quality of an electrodeposited metal techniques such as surface X-ray scat-
layer is generally poorer than that of a phys- tering (SXS) [19, 20] and electrochemical
ically deposited layer. The morphology of quartz crystal microbalance (EQCM) [21,
the electrodeposited metal layer is con- 22] have also been employed to investi-
trolled by many factors such as deposition gate the electrochemical deposition and
potential, current density, temperature, dissolution processes at atomic resolution.
concentration of metal ion/complex, pH, Atomically controlled electrochemical epi-
nature of the additives and substrates. taxial growth and layer-by-layer dissolution
472 4 Underpotential Deposition
10000
Other
Dental
×10 Glass
8000 Electronics
Chemical
Jewellery
Autocatalyst
[kilo-ounce]
6000
Demand
×10
4000
2000
×10
0
um um um um um
tin di di ni di
a lla o e Iri
Pl Pa Rh th
Ru
Fig. 1 Principal uses of noble metals (platinum, palladium, rhodium, ruthenium and iridium) in the
world in 2000 [2]. The scales for rhodium, ruthenium and iridium are enlarged by 10 times.
of many metals have become possible than that of the metal itself, a three-
now [23, 24]. dimensional (3D) island growth mecha-
In this chapter, we describe the recent in nism, namely, the VW mode is favored in-
situ atomic resolution studies on the elec- dependent of crystallographic mismatch.
trochemical deposition and dissolution of When the binding energy between the
noble metals, which show extremely high metal and the substrate is higher than that
catalytic activity for many chemical reac- of the metal itself, a two-dimensional (2D)
tions [1, 2]. metallic monolayer can be formed in
the underpotential deposition (UPD) re-
4.1.2 gion, and there are two possible growth
Atomically Controlled Deposition of Noble mechanisms for the subsequent growth
Metals of the metallic layers depending on the
crystallographic mismatch. When the crys-
Depending on the binding energy and tallographic mismatch of the metal and
crystallographic mismatch between the the substrate is negligibly small, metallic
deposited metal and the substrate, the layers are formed on the UPD layer on
growth of a metallic layer on the sub- the substrate by a layer-by-layer growth,
strate near equilibrium condition can namely, the FM mode. When the crys-
be classified into three mechanisms, tallographic mismatch is relatively large,
namely, the Vomer-Weber (VW), the the UPD layer contains considerable in-
Stranski-Krastanov (SK), and the Frank- ternal strain, and therefore, the growth of
van der Merwe (FM) modes [25–27]. Gen- the unstrained 3D metal islands on top
erally, when the binding energy between of the strained UPD layers is energetically
the metal and the substrate is lower favored and this mechanism is called the
Tab. 1 Physical properties and electrochemical growth modes of noble metal elements on gold substrate
Au Pt Pd Rh Ru Os Ir Ag
Atomic number 79 78 46 45 44 76 77 47
Atomic weight 196.97 195.08 106.42 102.90 101.07 190.2 192.2 107.87
Lattice structure fcc fcc fcc fcc hcp hcp fcc fcc
Electronic structure 5d10 6s1 5d9 6s1 4d10 5s0 4d8 5s1 4d7 5s1 5d6 6s2 5d7 6s2 4d10 5s1
Atomic diameter rA [nm] 0.288 0.277 0.275 0.269 0.270 0.273 0.271 0.289
Special gravity [g cm−3 ] 19.32 21.45 12.01 12.41 12.37 22.59 22.56 10.50
Melting point [◦ C] 1064 1768 1554 1963 2250 3045 2457 962
Thermal conductivity [Wcm−1 K−1 ] 2.93 0.72 0.76 1.50 1.05 0.87 1.47 4.19
Resistivity [µ cm, 20 ◦ C] 2.2 10.6 10.8 4.6 7.3 9.5 5.1 1.62
Tensile strength [Mpa]a 108 123 170 695 – – 1088 147
Hardness [Vickers]a 22 41 41 101 – 350 220 24
Surface energy νA [Jm−2 ] [28] 1.488 2.691 2.043 2.828 3.409 3.947 3.231 1.302
Geometry parameter rAB b 1 0.962 0.955 0.934 0.937 0.948 0.941 1.00
Lattice mismatch AB c 0 −3.8% −4.6% −6.7% −6.2% −5.2% −5.8% 0.17%
(continued overleaf )
4.1 Atomically Controlled Electrochemical Deposition and Dissolution of Noble Metals
473
474
Tab. 1 (continued)
4 Underpotential Deposition
Au Pt Pd Rh Ru Os Ir Ag
Surface energy mismatch AB d 0 0.575 0.314 0.620 0.802 0.905 0.739 0.133
Electrochemical growth mode on FM SK FM SK VW FM
gold substratee [29–31] [32] [33–36] [37] [38–40] [41–45]
a Annealed;
b Geometric parameter rAB = rB /rA [26] where A and B denote substrate and metal, respectively. Values in the table are calculated with
respect to the gold substrate;
c Lattice mismatches
AB [26] are shown with respect to that of gold;
d The surface-energy mismatches are calculated with respect to a gold substrate following the definition given by Bauer and
coworkers [26] as: AB = 2|(νA − νB )/(νA + νB )|;

e VW mode: Volmer-Weber mode, namely, 3D island growth; FM mode: Frank-van der Merwe mode, namely, layer-by-layer 2D mode; SK
mode: Stranski-Krastanov mode, unstrained 3D islands are formed on top of strained 2D layers. See text for details.
Note: fcc: face centered cubic; hcp: hexagonal close packed.
4.1 Atomically Controlled Electrochemical Deposition and Dissolution of Noble Metals 475
SK mode. Thus, the metal growth mode on If one examines the parameters shown in
the substrate can be estimated qualitatively Table 1, one can expect that electrochem-
from the properties of both deposited and ical deposition of gold and palladium on
substrate metals. a gold electrode should be the best candi-
Physical properties such as lattice and date for the layer-by-layer growth process
electronic structures, atomic diameters, among these noble metal elements. The
melting point, thermal conductivity, re- observed growth modes shown in the table
sistivity, tensile strength, hardness as well are in qualitative agreement with expecta-
as surface energy (νA ) [28] of the noble tions based on the physical parameters.
metal elements gold, platinum, palladium,
rhodium, ruthenium, osmium, and irid- 4.1.2.1 Platinum
ium are summarized in Table 1. The same Platinum is the most important mate-
properties for silver are also shown in rial in catalysis including electrocatal-
the same table for comparison. Following ysis [46–48]. For example, platinum is
the definition given by Bauer and cowork- widely used as a catalyst for automo-
ers [26], geometry parameter (rAB ), lattice biles [2]. Platinum and its alloys are also
mismatch (AB ) and surface-energy mis- employed as the main catalytic materials
match (AB ) in the table are given with in fuel cells, which are attracting more
respect to gold, which is widely used as a attention due to their high-energy conver-
substrate in studies of electrochemical epi- sion efficiency, flexibility, and low/clean
taxial growth of noble metals. The experi- emission than those of traditional internal-
mentally observed electrochemical growth combustion engines [49, 50]. The pla-
modes of these metals on gold are also tinized platinum electrode [51] has been
listed in the table. These parameters play used for standard reversible hydrogen elec-
significant roles in the growth mechanism trodes (RHE) because of its extremely high
for the electrochemical deposition of these reversibility for hydrogen evolution and
metals on various substrates. As shown in oxidation reactions.
the table, the lattice mismatches (AB ) are Ultrathin platinum layers on a chemi-
observed from −3.8% and −6.7% among cally inactive substrate are very important
these noble metal elements, and platinum systems in the field of catalysis and electro-
shows the smallest mismatch (−3.8%) rel- catalysis. Many studies on epitaxial growth
ative to the gold substrate. The value of of platinum thin layers on various sub-
AB , which is calculated from the sur- strates have been reported in the UHV
face energy of each metal with respect environment [4, 52–55]. The electrochem-
to a gold substrate [28], decreases in the ical growth of platinum layers is achieved
sequence of gold (0), palladium(0.314) < by using a number of commercially avail-
platinum(0.575) < rhodium(0.620) < iri- able electroplating baths [56–58]. These
dium(0.739) < ruthenium(0.802) < osm- studies are mainly limited to the purpose
ium(0.905). As proposed by Bauer and of electroplating where film thickness, pu-
coworkers, the value of AB can be em- rity, film strength, adhesive force to the
ployed to judge the growth mode of metal substrate and growth rate are more im-
deposition on various substrates [26]. AB portant than surface atomic structure [46,
values <0.5 with small lattice mismatch 47, 51, 59–62]. The atomically controlled
(rAB ≤ 1.00) are required for the 2D crys- electrochemical deposition of platinum on
talline growth of the metal on the substrate. a gold single-crystal substrate was realized
50 400
End
0 300
Start
Mass change
[µA cm−2]
−50
Current
End 200
[ng]
−100
100
−150
Start 0
−200
500 600 700 800 900 1000 1100
Potential vs RHE
[mV]
Fig. 2 Potential dependence of current (solid line) and the mass
change (dotted line) at a (111)-such as gold/EQCM electrode in
50 mM HClO4 + 0.6 mM H2 PtCl6 . Sweep rate: 20 mV s−1 [32].
recently by Uosaki and colleagues [32, 63, PtCl6 2− :

64].
Figure 2 shows the potential depen- PtCl6 2− + 4e− −−−→ Pt + 6Cl−
dence of the current and the surface mass
That is, 48.8 g F−1 -electron, which is equal
change at an Au(111) electrode in the first
to the atomic weight of platinum (195.08),
potential cycle between +1.0 V and +0.6 V
divided by 4 [32, 63, 64].
in 50 mM HClO4 solution after 0.6 mM
An STM image obtained at +0.95 V
H2 PtCl6 solution was added at +1.0 V (vs.
where neither cathodic nor anodic cur-
RHE). Cathodic current (solid line) started
rent flowed showed an ordered adlayer
to flow around +0.80 V and increased sig-
structure (40 × 40 nm2 , Fig. 3a), which
nificantly as the potential became more is totally different from that of the
negative than +0.70 V. The surface mass Au(111) substrate. The spots with the
(dotted line) significantly increased when same brightness showed a hexagonal sym-
the potential became more negative than metry with a nearest-neighbor distance
+0.65 V. Even after the sweep direction of about 0.76 nm, while the symmetry
was reversed at +0.60 V, a cathodic current of the adlattice was rotated by about 20
flowed and the surface mass continued to degrees from that of the Au(111) sub-
increase until the potential became more strate. This structure can be assigned
positive than +0.80 V. The relationship be- to an adlayer of √the PtCl 2− complex
√ 6
tween the electric charge passed and the that forms the 7 × 7R19.1◦ struc-
surface mass change showed that the sur- ture on the Au(111) surface [32, 63, 64].
face mass increased by 48.5 g as a cathodic The adsorption of the PtCl6 2− com-
charge of 1 F was passed. This value is in plex on the Au(111) surface has also
good agreement with the calculated value been confirmed by the mass increase
for the four-electron reduction process of after injection of the PtCl6 2− complex
(a) 0 20 40 nm (b) 0 20 40 nm
(c) 0 20 40 nm
Fig. 3 STM images (40 × 40 nm2 ) of platinum deposition process on an Au(111) substrate
(a) at +0.95 V, (b) 10 min and (c) 30 min, respectively, after the potential was stepped from
+0.95 V to +0.70 V in 50 mM H2 SO4 + 0.05 mM H2 PtCl6 [32].
into the blank solution√ at the √ same from the top-right portion of the image
potential [64]. The same 7 × 7R19.1◦ (Fig. 3b) and covered the entire area of the
adlattice structure of the PtCl6 2− com- image after prolonged deposition (Fig. 3c).
plex on an Au(111) electrode has been The bottom-left portion of Fig. 3(b) is the
confirmed by Kolb and colleagues re- Au(111)
√ covered with a PtCl6 2− adlayer
√
cently [65]. As will be described later, of 7 × 7R19.1◦ structure (c.f. Fig. 3a).
similar adlattice structures of metal com- The terrace size of the deposited platinum
plexes, such as PdCl4 2− [33], RhCl6 2− [37], layer seemed to be much smaller than that
AuCl4 − [66] and PtCl4 2− [67], have also of the Au(111) surface before the platinum
been observed on the Au(III) electrode deposition. A number of small clusters
surface. were also observed on the platinum layer.
As shown in Figs. 3(b) and (c), the depo- These clusters are higher than the first plat-
sition of a platinum layer of monoatomic inum layer by a monoatomic height of the
height was initiated as soon as the potential platinum layer. The third platinum layer
was stepped to +0.70 V. The growth of the started to grow before the growth of the
first layer seemed to be essentially in a 2D second platinum layer was completed. The
mode. The deposited platinum layer grew morphology of the subsequent platinum
m
3n
0.3
m
6n
0.7
m
6n
0.7
(a) 0 2.5 5.0 nm (b) 0 2.5 5.0 nm

Fig. 4 High resolution STM images (5 × 5 nm2 ) of (a) an Au(111) electrode obtained at
+0.70 V in 50 mM HClO4 + 0.6 mM H2 PtCl6 and (b) the same electrode in 50 mM HClO4
after the solution in the STM cell was replaced several times by 50 mM HClO4 , while the
potential was held at +0.70 V after the STM image (a) was obtained. Note that the imaged
portion was not the same [32].
√ √
layers became rougher than that of the the adlattice structure of 7 × 7R19.1◦
first layer. This means that the growth of disappeared but a highly ordered surface
the platinum layer on an Au(111) electrode structure of a hexagonal array with a neigh-
is not a perfect 2D growth, namely, the FM bor atomic distance of 0.28 nm (Fig. 4b).
mode. However, Friedrich and colleagues Although it is difficult to distinguish the
observed that platinum deposition on an distance of the unit lattice of Pt(111)
Au(111) surface was mainly observed at (0.277 nm) and Au(111) (0.288 nm) sur-
the step edges of the substrate as isolated faces by the present STM measurement,
platinum particles at the lower coverage the image shown in Fig. 4(b) should cor-
by ex situ STM measurements [68]. Recent respond to the Pt(111) – (1 × 1) structure
investigation on the initial stage of plat- formed on the Au(111) substrate as both
inum deposition on Au(hkl) surfaces by EQCM and XPS measurements confirmed
Kolb and colleagues also showed that the the existence of platinum on the Au(111)
nucleation of platinum occurred preferen- surface [32]. Adsorption of the PtCl6 2−
tially at the surface defects and that the complex is considered to inhibit the 3D
mobility of platinum was apparently in- growth but to promote the initial 2D
creased by the addition of chloride into the growth of platinum [32].
solution [65]. √ On the other hand, X-ray diffraction
√The adlayer with a structure of 7 × (XRD) measurements of the deposited
7R19.1◦ was also observed on the surface platinum films (ca. 250 ML) on the
of deposited platinum, demonstrating that Au(111) substrate demonstrated that a
the PtCl6 2− complex adsorbed not only on (111)-orientated platinum-bulk phase was
the gold substrate but also on the plat- definitely formed on the Au(111) elec-
inum surface during the electrochemical trode surface by electrochemical epitaxial
deposition process (Fig. 4a). In situ STM growth [32].
measurement of the platinum-deposited The aforementioned experimental re-
Au(111) surface in 50 mM HClO4 solution sults suggest that the platinum layer
without the PtCl6 2− complex showed that grows on the Au(111) electrode by the
SK mechanism. This may be related to the market price of palladium since 1999 may
relatively large AB of platinum with re- affect its applications in the industry [2].
spect to the gold substrate (0.575) and the Since Clavilier proposed a simple way
moderately slower surface diffusion rate of to prepare a noble metal single crystal [9],
platinum atoms [69, 70]. many electrochemical studies have been
Electrochemical characterizations of the carried out on platinum, gold, and irid-
UPD of hydrogen and copper show that ium as well as palladium single-crystal
the surface structure of the electrochem- electrodes. The preparation of a palladium
ically deposited platinum layer on the single-crystal is, however, difficult because
Au(111) substrate was quite different from palladium cannot be subjected to the usual
that of a polycrystalline platinum elec- convenient flame-annealing treatment to
trode, but more similar to that of the prepare the electrode [73]. It is of great in-
Pt(111) single-crystal electrode [32]. The terest if one can grow a palladium epitaxial
difference in the electrochemical behavior layer on other noble metal single-crystal
between the platinum-deposited Au(111) surfaces, which are easily prepared by Clav-
and an ideal Pt(111) electrode is attributed ilier’s method. This provides not only a
to the relatively small (111) domain and convenient method of preparing an or-
the higher step density on the platinum- dered palladium surface but also a way
deposited surface. Recently, Watanabe and of investigating the chemical and physical
colleagues reported that the (111)-like sur- properties of a thin metallic layer, which
face structure of a platinum thin-film may be different from those of the bulk
electrode prepared by a sputtering method material.
can be improved by electrochemical an- The investigation on the electrochem-
nealing at the potential region of hydrogen ical epitaxial growth of palladium on
adsorption in HClO4 solution [71]. gold and platinum single-crystal sub-
strates was started recently. Attard and
coworkers reported the irreversible depo-
4.1.2.2 Palladium sition of submonolayer and monolayer
As shown in Fig. 1, palladium is mainly coverage of palladium on Pt(111) elec-
consumed as an autocatalyst now [1, 2]. trode by the immersion technique for
Palladium is also an important noble metal the first time [74]. Llorca and coworkers
element widely applied in the development investigated the irreversibly adsorbed pal-
of electrocatalytic materials [46, 48, 72]. It ladium on Pt(hkl) in acidic solution [75]
is widely used as a promoter for electroless and reported that electrocatalytic activity
deposition of metals on various substrates. for the oxidation of formic acid on the
Palladium is also known for its extraor- Pt(100) electrodes was improved drasti-
dinary ability to absorb a large amount cally by the palladium-adlayer modifica-
of hydrogen. The growth of palladium tion, while that on the Pt(111) electrode
thin layers on various substrates in UHV was not greatly affected by the Pd-
has been investigated in detail [3–8]. Elec- modification [76]. However, it is difficult
trodeposition of palladium is widely used, to prepare an ultrathin film of palladium
and a number of commercially available with various thickness by the immersion
electroplating baths of palladium for differ- technique, and STM observation of the
ent purposes have been developed [56–58, atomic structure of these surfaces is not
61, 72]. However, the rapid increase in the available yet.
The electrochemical 2D growth of thin Figure 5 shows the characteristic po-

palladium layers on gold single-crystal tential dependence of the current and
electrodes was reported by Kolb’s group the surface mass change at an Au(111)
for the first time [77]. They also found electrode in 50 mM H2 SO4 solution con-
that the hydrogen absorption reaction at taining 0.1 mM PdCl4 2− complex [33]. A
the palladium thin layers takes place at cathodic current started to flow as soon
more cathodic potentials than that at a as the potential became more negative
bulk-palladium electrode and does not oc- than +0.95 V, reaching a maximum at
cur at palladium films thinner than 2 ML. +0.88 V and decreasing to a limiting value.
The observation of the electrochemical When the potential scan was reversed in
deposition of palladium at atomic res- the positive direction, an anodic current
olution was not available until detailed started to flow at +0.8 V with two anodic
STM and EQCM studies of the electro- peaks at +0.89 and +0.96 V. The mass re-
chemical epitaxial growth of palladium spectively increased and decreased when
on Au(hkl) electrodes were carried out the cathodic and anodic currents flowed.
by Uosaki and colleagues [33, 34, 78–80] Quantitative analysis between the charge
and later by Kolb and colleagues [35, and the mass change show that the ca-
36]. thodic and anodic currents are respectively
30
[µA cm−2]
Current
(a) −30
500
400
300
[ng cm−2]
∆m
200
100
0.0 0.5 1.0 1.5

E vs RHE
(b) [V]
Fig. 5 Potential dependence of (a) current and (b) mass change
(m) at an Au(111) electrode in 50 mM H2 SO4 + 0.1 mM
PdCl4 2− . Sweep rate: 5 mV s−1 [33].
105 nm × 105 nm 23 nm × 23 nm
0 10 nm
(a) (b)
Fig. 6 (a) STM image (10 × 10 of an Au(111) surface obtained at +0.95 V in

nm2 )
50 mM H2 SO4 + 0.1 mM PdCl4 2− [33] and (b) STM image (105 × 105 nm2 ) of an
Au(111) surface at 0.63 V in +0.95 V in 0.1 M H2 SO4 + 0.1 mM PdCl4 2− . The inset
(23 × 23 nm2 ) shows the same structure at a higher resolution [35].
related to deposition and dissolution of in very good agreement with the results
palladium on the Au(111) surface [33]. of coverage estimated from the EQCM
Kolb and colleagues reported recently that measurements [33]. A similar but not iden-
the characteristic features of the current- tical adlattice structure of the PdCl4 2−
potential curves for palladium deposition complex on an Au(111) substrate was ob-
on gold single-crystal electrodes depend served by Kolb and colleagues as shown in
significantly on the surface structure and Fig. 6(b) (105 × 105 nm2 ) [35]. This image
concentration of the PdCl4 2− complex as reveals the existence of an ordered adlayer
well as the chloride ion contained in the with parallel stripes separated by about
solution [35]. 1.8 nm and a distorted hexagonal struc-
Figure 6 shows STM images of the ture (inset, 23 × 23 nm2 ) incommensurate
Au(111) electrodes in dilute H2 SO4 so- with the Au(111) substrate. The approxi-
lution containing 0.1 mM PdCl4 2− at the mate next-neighbor distances were 0.81,
potential where no palladium deposition 0.68 and 0.73 nm. Domain boundaries and
takes place as reported by (a) Uosaki and small domains with three directions are
colleagues [33] and (b) later by Kolb and clearly observed.
colleagues [35]. Figure 6(a) (10 × 10 nm2 ) Figure 7 shows the surface structure on
shows an adlattice in hexagonal symme- the Au(100) electrode in dilute H2 SO4 so-
try with a nearest-neighbor distance of lution containing the PdCl4 2− complex at
0.78 nm [33]. This structure is very similar the potential where no palladium deposi-
to that of the PtCl6 2− adlayer on an Au(111) tion takes place as reported by (a) Uosaki
electrode (Fig. 3a) and is attributed to the and colleagues [34] and (b) later by Kolb
structure of the adsorbed PdCl4 2− complex and colleagues [36]. The two images show
on the Au(111)
√ √ with an adlattice struc- essentially the same structure of an ad-
ture of 7 × 7R19.1◦ . This result is lattice with long and short axis lengths of
4.5 nm × 4.5 nm
(001)
(010)
(a) 0 5 nm (b)
(c)
Fig. 7 (a) Lowpass filtered-STM white box [34]; (b) STM image (4.5 × 4.5 nm2 )
image (5 × 5 nm2 ) of an Au(100) surface at of an Au(100) surface at 0.61 V in 0.1 M
+1.05 V in 50 mM H2 SO4 + 0.5 mM PdCl4 2− . H2 SO4 + 0.5 mM PdCl4 2− + 0.6 mM HCl [36];
The inset is the Fourier transformed 2D (c) A ball model of the PdCl4 2− complex
spectrum and a unit lattice determined from the adsorbed on Au(100) surface [34].
spectrum is superimposed on the image with a
1.47 nm and 1.03 nm and can be attributed conditions, suggesting that the adsorption
to the adlattice structure of the adsorbed of PdCl4 2− on Au(100) is weaker than that
PdCl4 2− complex on the Au(100) surface. on Au(111) [34].
An adlattice model is given in Fig. 7(c). The Figure 8 shows STM images (300 ×
STM image of the ordered-adlayer struc- 300 nm2 ) of the Au(111) electrode sur-
ture on the Au(100) surface was hard to face during the palladium deposition
obtain when the concentration of PdCl4 2− in 50 mM H2 SO4 solution containing
species was lower than 0.05 mM, while 0.5 mM PdCl4 2− [33]. Atomically flat and
the ordered PdCl4 2− adlayer was observed large terraces (Terrace-1 and Terrace-2)
on the Au(111) surface under the same with monoatomic steps on the Au(111)
0.80 V
0.95 V
(a) (b)
1.0 Pd-1
1.0 Pd-1
nm
nm
0 0
Au Au
Substrate Substrate
−1.0 −1.0
(a') 0 300 nm (b') 0 300 nm
1 min 2 min
(c) (d)
1.0 Pd-2 1.0
nm
nm
0 Pd-1 0 Pd-2 Pd-2

Pd-1 Pd-1 Pd-2 Pd-1
Au Au
−1.0 Substrate −1.0
Substrate
(c') 0 300 nm (d') 0 300 nm
Fig. 8 (a) STM image (300 × 300 nm2 )
of an surface (b) 0 min, (c) 1 min, (d) 2 min,
Au(111) surface at +0.95 V in 50 mM (e) 3 min, (f) 28 min, (g) 48 min and (h) 90 min
H2 SO4 + 0.1 mM PdCl4 2− . Sequentially after the potential was stepped from +0.95 V to
obtained STM images (300 × 300 nm2 ) of the +0.80 V in the same solution. A cross section
palladium deposition process on the Au(111) along the white dotted line in each figure is
shown below each image [33].
3 min 28 min
(e) (f)
1.0 1.0 Pd-3
nm
nm
0 Pd-2 0 Pd-2
Pd-1 Pd-2 Pd-1 Pd-1 Pd-2
Pd-1
Au Au
−1.0 −1.0
Substrate Substrate
(e') 0 300 nm (f') 0 300 nm
48 min 90 min
(g) (h)
1.0 1.0 Pd-4
Pd-3 Pd-4
Pd-4
nm
nm
0 Pd-2 Pd-3 0 Pd-3

Pd-1 Pd-2 Pd-2 Pd-3
Au Pd-1 Pd-2
−1.0 −1.0 Au Pd-1
Substrate Substrate
(g') 0 300 nm (h') 0 300 nm
Fig. 8 (Continued)
substrate were observed at +0.95 V (d) 2 min, (e) 3 min, (f) 28 min, (g) 48 min
where no palladium deposition takes and (h) 90 min after the potential of the
place (Fig. 8a). Figures 8(b–h) show se- Au(111) was stepped from +0.95 V to
quentially obtained STM images of +0.80 V as indicated by the thick arrow
the same area at (b) 0 min, (c) 1 min, in Fig. 8(b) where palladium deposition is
expected to take place. The palladium nu- reactant, namely, PdCl4 2− , between the
clei were being generated on the Au(111) STM tip and the electrode surface, leading
substrate immediately after the potential to a negative shift in the equilibrium
step. The cross section (Fig. 8b ) along the potential for palladium deposition [78].
dotted line shows the formation of palla- The palladium layer, which was originally
dium islands with a monoatomic height first deposited on Terrace 1, was also
on the large terraces of the gold substrate. grown laterally and finally merged with
In the upper part of Fig. 8(b), 2D growth Pd-3 on Terrace 2 (Figs. 8f,g,h). The STM
of the first palladium monolayer (Pd-1) images in Fig. 8 demonstrate that the
progressed not only on the large terrace electrochemical deposition of palladium
but also on a narrow terrace between the on the Au(111) surface is an epitaxial layer-
step lines. This result is in contrast to that by-layer process.
observed by Kolb and colleagues, in which Figure 9 shows STM images (300 ×
the palladium layer started to preferentially 300 nm2 ) of an Au(100) electrode ob-
grow from the step-and-edge sites [35, 77]. tained in 50 mM H2 SO4 solution con-
This discrepancy should be due to the
taining 0.5 mM PdCl4 2− [34]. A number
differences in the deposition potential and
of monoatomic-high gold islands were
solution conditions.
observed in the STM image at +1.0 V
As shown in Figs. 8(c) and (c ), after the
(Fig. 9a) when no palladium deposition
first palladium layer (Pd-1) was completely
took place. These gold islands were
formed on the gold substrate, islands
formed as a result of the reconstruc-
of the second palladium layer (Pd-2)
tion lifting of the Au(100) surface due
were generated on both Terrace-1 and
to the large difference (25%) in sur-
Terrace-2, and the first palladium layer
face densities between the reconstructed
(Pd-1) on the narrow terrace started
to grow laterally further onto the Pd-1 and unreconstructed Au(100) surface [81].
layer on Terrace-2 from the step line. Figures 9(b)–(g) show STM images se-
Both the Pd-2 layer on Terrace-2 and quentially obtained during the palladium
the palladium layer extended from the deposition process. As soon as the poten-
Pd-1 layer on the narrow terrace grew tial of the Au(100) substrate was stepped
two-dimensionally and merged together, from +1.0 to +0.80 V (thick arrow in
resulting in a large flat terrace of a second Fig. 9b), palladium nuclei were generated
palladium layer on Terrace-2 (Figs. 8d, e). both on the wide terrace and on the gold
The cross section (Figs. 8c ,d ,e ) clearly islands (cross section I). The height of the
shows the 2D growth of the palladium palladium nuclei corresponded to that of
monoatomic layer. After the complete a palladium monoatomic step (0.19 nm).
second palladium layer was formed on Kolb and colleagues reported that the pal-
both Terrace-1 and Terrace-2, the third ladium nucleated at the steps and island
palladium layer (Pd-3) started to grow two- rims on the island-covered Au(100) elec-
dimensionally on the terraces and from trode surface and that the first layer of
the step line (Figs. 8f,g). The growth rate palladium two-dimensionally grew from
of the third palladium layer was, however, these nuclei [36].
slower than that of the first and the As shown in the cross section II in
second layers. This should be due to the Fig. 9(b), the growth of the first pal-
decrease in the local concentration of the ladium layer on the Au(100) surface
10 s
(II)
0.8 V
(I)
1.0 V
1.0
nm
0.0 Au Au
Au substrate
−1.0
(a) 0 300 nm 1.0
Pd
11 s
nm
0.0 1st Pd
Au Au 1st Pd
Au substrate
Cross section (II)
−1.0
1.0 Pd
Pd
nm
0.0 Au Au
Au substrate
Cross section (I)
56 s −1.0
(b) 0 300 nm
0.5
1st Pd
nm
0.0 2nd Pd
2nd Pd 1st Pd
Au 1st Pd Au
Au substrate
−0.5
(c) 0 300 nm
Fig. 9 (a) STM image (300 × 300 nm2 )

of an surface in the same solution. Time after the
Au(100) surface at +1.0 V in 50 mM potential pulse corresponding to the top and
H2 SO4 + 0.5 mM PdCl4 2− . (b)∼(g) STM image bottom of the STM image is indicated on the
(300 × 300 nm2 ) obtained sequentially during right side of each image. The cross section along
the palladium deposition process on an Au(100) broken line in each image is shown below [34].
was completed within 10 s after the po- the second layer of palladium seemed
tential step, and the second palladium to grow laterally from the deposited pal-
layer then started to grow (Fig. 9c). As ladium layer just on top of the gold
shown in the cross section in Fig. 9(c), islands. Growth of the second layer of
103 s 291 s
148 s 335 s
0.5 0.5
2nd Pd
2nd Pd 3rd Pd 3rd Pd
3rd Pd
nm
nm
0.0 3rd Pd 0.0

1st Pd 1st Pd
2nd Pd 2nd Pd 2nd Pd 2nd Pd
Au
1st Pd Au 1st Pd 1st Pd 1st Pd
−0.5 −0.5
(d) 0 300 nm (e) 0 300 nm
635 s 1110 s
680 s 1155 s
0.5 0.5
2nd Pd 4th Pd layer

nm
nm
0.0 0.0 3rd Pd 2nd Pd

3rd Pd 3rd Pd
3rd Pd
1st Pd 2nd Pd 1st Pd 3rd Pd
2nd Pd 2nd Pd 2nd Pd
−0.5 −0.5 1st Pd Au
(f) 0 300 nm (g) 0 300 nm
Fig. 9 (Continued)
palladium on the gold terrace was com- reflecting the difference in atomic di-
pleted within about 30 s. It was noted ameters of gold (0.288 nm) and palla-
that, although the gold terraces were cov- dium (0.275 nm).
ered with two layers of palladium, the The cross section of Fig. 9(d) (103 ∼
gold islands were covered with only one 148 s) clearly shows that the third layer
monolayer of palladium. At this stage, of palladium was also initiated from
the height difference between the two re- the deposited palladium layer on top of
gions is very small but still distinguishable, the gold islands and grew laterally. As
shown in the subsequent STM images also observed in the palladium-deposition

in Figs. 9(e) (291 ∼ 395 s) and 9(f) (635 ∼ process on an Au(111) electrode and
680 s), each of the palladium islands grew could be attributed to the concentration
laterally and eventually merged together. decrease in PdCl4 2− between the STM
Completion of the third palladium layer tip and the electrode surface [33, 78].
took more than 1000 s, which is much The 2D growth process of the fourth
longer than the time required for the palladium layer was clearly observed on
growth of the first and second layers of the right portion of the STM images shown
palladium. The same phenomena were in Figs. 9(f) and 9(g) (1110 ∼ 1155 sec).
500 nm × 500 nm
0.65 V
0.52 V
0.65 V
(a) 0 min (b) 5 min
0.52 V
0.47 V
0.47 V
0.43 V
(c) 10 min (d) 15 min
0.43 V 0.39 V
(e) 20 min (f) 27 min
Fig. 10 Sequence of STM images for palladium deposition on island-free Au(100) in 0.1 M
H2 SO4 + 0.1 mM PdCl2 + 0.6 mM HCl solution [36]. Deposition potentials as indicated in
the figure. IT = 2nA.
Even after four layers of palladium were islands were much smaller than the ini-
deposited on the Au(100) terrace and three tial gold islands. Kolb and colleagues
layers of palladium on the gold islands, attributed this phenomenon to the for-
the shape of the gold islands was still mation of a palladium-gold surface al-
distinguishable. loy on the Au(100) electrode, although
Recently, Kolb and colleagues com- such an effect was not observed on the
pared the electrochemical deposition be- Au(111) electrode [36]. The surface alloy
havior of palladium on the island-covered was considered to be the main reason
and island-free unreconstructed Au(100) for the deviations in the electrocatalytic
electrode surfaces [36]. By electrochemical behavior of massive Pd(100) and thin
annealing in 0.1 M H2 SO4 + 0.1 mM palladium layers (<3 ML) on Au(100).
PdCl2 + 0.6 mM HCl at +0.80 V (vs. SCE; However, the formation of a surface al-
same for Kolb’s work) where no palladium loy during the dissolution of palladium
deposition takes place, the monoatomic- deposited on an Au(100) electrode has
high gold islands generated by the lift- not been observed by Uosaki and col-
ing of the reconstruction disappeared leagues in their in situ STM [34] and SXS
instantaneously. Figure 10 shows in situ studies [82, 83]. This may be related to
STM images for palladium deposition the differences between the experiments
on an island-free Au(100) surface in the carried out in the two groups, such as
same solution. Palladium deposition com- electrode potential and chloride concen-
menced at +0.52 V, and palladium nu- tration as well as the in situ STM cell
cleated selectively at the monoatomic-high configuration.
steps and grew onto the terraces (Fig. 10b). Because STM is sensitive only to the out-
The second layer of palladium started ermost surface and the spatial resolution
to grow at +0.47 V by creation of small of STM is limited to about ±0.1 nm, the
palladium islands at the steps (Fig. 10c). crystallographic relationship between the
The fourth monolayer started to grow be- palladium layers and the gold substrate
fore the third monolayer was completed, has remained unclear. Recently, Uosaki
where palladium also nucleated on the and colleagues carried out the in situ SXS
terraces (Figs. 10e–f). Square-type struc- measurements to elucidate the detailed
tures of palladium deposits were generated structure of palladium electrochemically
with the third monolayer, revealing the deposited on Au(111) and Au(100) elec-
symmetry of the Au(100) substrate and trode surfaces [82, 83]. The pseudomor-
showing a layer-by-layer epitaxial growth phic layer structure was found only on
of palladium (Figs. 10e–f). The behav- the first monolayer of palladium on the
ior is markedly different from those Au(111) substrate, while the pseudomor-
observed on the island-covered Au(100) phic layer-by-layer structure was observed
electrode surface (Fig. 9) [34, 36]. Fur- on the Au(100) substrate up to 14 ML. The
thermore, when the palladium deposited relaxation of the palladium layer on the
on both island-covered and island-free Au(111) electrode was found between the
Au(100) surfaces was stripped anodically, second and fourth layer, and the palla-
many small monoatomic-high islands and dium then started to grow from the fifth
holes were created at those places where layer with (111) orientation. However, the
palladium had been deposited. These relaxation of palladium on the Au(100)
surface takes place after 14 ML and the sur- 4.1.2.3 Rhodium

face became rough when its thickness was The autocatalyst is the main application for
higher than 40 ML. These differences have rhodium now (Fig. 1) [1, 2]. Electrodeposi-
been explained by the different adsorption tion of rhodium is also widely used both in
sites on Au(111) (fcc and hcp sites) and corrosion prevention and in the decorative
Au(100) (only 4-hollow sites) surfaces [82, field [46, 61, 85]. Details about the progress
83]. in electroplating baths for rhodium have
In conclusion, epitaxial layer-by-layer been summarized elsewhere [56–58, 61,
growth of thin palladium layers follow- 85].
ing the FM mode has been observed on The electrochemical deposition pro-
both Au(111) and Au(100) electrodes. The cesses of rhodium at atomic resolution
ordered adlayer of the PdCl4 2− complex have been investigated first on platinum
was observed on the surface of the pal- single-crystal electrodes [86–89]. From the
ladium deposited on the Au(111) surface electrochemical characterization and in
by Uosaki and colleague [33] and by Kolb situ infrared reflection-absorption spec-
and colleague [35]. The adlayer structure troscopy (IRRAS) spectra of CO adsorbed
of the PdCl4 2− complex has not been on the rhodium deposited electrochem-
observed clearly on the surface of the pal- ically on a Pt(111) electrode, Inukai
ladium deposited on the Au(100) electrode and coworkers suggested that the epi-
by STM so far [36, 78]. Kolb and colleague
taxial growth of rhodium monolayers
concluded that the PdCl4 2− complex is
occurs in a pseudomorphic (111) struc-
displaced by more strongly adsorbing chlo-
ture [86]. Figure 11 compares the voltam-
ride on the palladium overlayers on the
mograms of a bulk Rh(111), an Rh mono-
Au(100) electrode. Uosaki and colleagues
layer/Pt(111), and Rh overlayers/Pt(111) in
suggested that the same adlayer of the
0.5 M H2 SO4 solution [89]. Although there
PdCl4 2− complex should also exist on the
are some small differences, the voltam-
palladium overlays on Au(100) as that on
mogram of monolayer Rh deposited on
the bare Au(100) surface and proposed
that the adlayer plays a role in the 2D Pt(111) is very similar to that of a mas-
epitaxial growth during electrochemical sive Rh(111) electrode, some additional
deposition of palladium on the Au(100) peaks in the hydrogen-adsorption region
surface similar to that observed on the were observed on the thicker rhodium
Au(111) surface [34]. layers. From these types of electrochem-
The electrocatalytic behavior of the thin ical behavior of rhodium deposited on
palladium layers deposited on Au(hkl) a Pt(111) surface, Gomez and cowork-
surfaces for hydrogen adsorption/absorp- ers concluded that the first rhodium layer
tion [77, 80], oxygen reduction [79], grows in a pseudomorphic way, whereas
oxide formation/reduction [80], copper the second and subsequent layers grow in
UPD [33, 77], electrochemical oxidation a pseudo-layer-by-layer growth: the third
of formic acid [84] as well as formal- layer begins to form before the second is
dehyde [80] has been investigated in detail. completed (SK growth) [89]. In the case
These electrocatalytic activities depended of ultrathin rhodium layers prepared by
significantly on the surface structure metal evaporation in UHV, the second
and thickness of the ultrathin palladium rhodium layer begins to form prior to the
layers [80, 84]. completion of the first layer, although both
Rh (111)
200
Rh/Pt (111)
200
[µA cm−2]
[µA cm−2]
j
j
0.2 0.6 1.0 0.2 0.7 1.0
E RHE E RHE
[V] [V]
−400 −400
(a) (b)
200
0.2 0.9
[µA cm−2]
E RHE
[V]
j
−400
(c)
Fig. 11 CVs in 0.5 M H2 SO4 solution for (a) a prepared by electrochemical deposition onto a
massive Rh(111) electrode, (b) Pt(111) electrode monolayer Rh-modified Pt(111) electrode in
modified by one monolayer of rhodium using 0.06 mM Rh3+ at +0.6 V for 620 s [89].
electroless deposition and (c) rhodium overlayer
layers give rise to pseudomorphic (1 × 1) a platinum substrate clean during the in

islands [90]. situ STM measurements.
These studies are, however, character- Recently, Kolb and colleague investi-
ized mainly by electrochemical measure- gated the initial stages of electrochemical
ments. The surface morphology deter- deposition of rhodium on Au(111) elec-
mined by STM is still very limited, which trodes by in situ STM measurements [37].
may be related to the difficulties of keeping Figure 12 shows current-potential curves
10 Fig. 12 CVs for a well-ordered Au(111)

surface (solid line) and a 5◦ stepped
Au (111) Au(111) surface (dashed line) in 0.1 M
H2 SO4 + 0.1 mM RhCl3 + 1 mM HCl.
0
Sweep rate: 10 mV s−1 . The scans
[µA cm−2]
started at +0.5 V and finished at

0 V [37].
j
−10
−20
0.0 0.2 0.4 0.6

E SCE
[V]
30 nm × 30 nm 10 nm × 10 nm
550 mV
480 mV
(a) (b)
Fig. 13 STM images for Au(111) √ in 0.1√M H2 SO4 + 0.1 mM RhCl3 + 1 mM HCl.
(a) E = +0.65 V, showing the ( 13 × 13) R13.9◦ -superlattice pf RhCl6 3− and (b) the same
structure at +0.55 V with the atomically resolved Au(111) surface after a potential step to
+0.48 V at the moment shown by the arrow [37].
of (a) well-defined (solid line) and (b) 5◦ observed at the stepped-Au(111) surface,
stepped Au(111) electrodes (dashed line) suggesting that the rhodium deposition
in 0.1 M H2 SO4 + 0.1 mM RhCl3 + 1 mM proceeds easier at surface defects. The
HCl. A pair of spikes was only observed rhodium deposited on a gold-electrode sur-
on the well-defined Au(111) surface at face cannot be dissolved again because
+0.52 V (vs. SCE). The deposition of a passivating oxide is formed on the
rhodium at the well-defined Au(111) sur- rhodium layer when the potential is swept
face was started around +0.20 V with a to the positive region [37].
large (+0.10 V) and a small cathodic peak Figure 13 shows in situ STM images at
(+0.18 V). The peak position of the small the potential region around the current
peak coincided with the reduction peak spikes. An STM image with a hexagonal
structure can be observed in the potential to the adsorption of the RhCl6 3− com-
region more positive than that of the spike plex. When the potential was stepped to
(Fig. 13a).
√ This√ structure is determined +0.48 V, that is, negative relative to the
to be 13 × 13R13.9◦ and is attributed spikes, this adlattice structure disappeared
Rh
bilayer
1 ML Rh
(a) (b)
(c) (d)
(e)
Fig. 14 Sequence of STM images for the deposition of rhodium on Au(111) surface at
+0.19 V (400 × 400 nm2 ) in. 1 M H2 SO4 + 0.1 mM RhCl3 + 1 mM HCl. (a) Bare Au(111)
surface; (b–d) growth of a rhodium bilayer on Au(111) and (e) rhodium cluster formation
on the bilayer [37].
and the Au(111) substrate could be imaged a direct methanol fuel cell (DMFC) [49,
atomically (Fig. 13b). 50], which is a key candidate for use
Figure 14 shows a sequence of STM as a pollution-free power supply in the
images (400 × 400 nm2 ) during electro- near future. In fact, the most efficient
chemical deposition of rhodium on an catalyst for the practical DMFC is the
Au(111) electrode in the same solution Pt/Ru based electrode, and ruthenium is
as Fig. 13 [37]. When the potential was known to promote the methanol oxidation
stepped to +0.19 V, a number of rhodium on a platinum surface by providing OHad
islands were formed at the terrace of the species on its surface.
gold substrate (Fig. 14b). Most of the small The electrochemical deposition of a
rhodium islands are immediately covered metallic-ruthenium film is very difficult
by a rhodium bilayer with a total height compared with that of other platinum-
of 0.3 nm. The adsorption of the RhCl6 3− group elements [46, 61, 91–93]. One of the
complex was also observed on the surface reasons may be related to the complicated
of the rhodium bilayer. A rhodium film electrochemistry of ruthenium deposition
with a two-monolayer height covered the and the stability of the Ru-chloro com-
gold surface almost completely (Fig. 14d). plex [92]. For example, it has been reported
The formation of the rhodium islands that the RuCl3 species in HClO4 solution is
shows a fractal growth, indicating a low decomposed partly into RuO2+ , in which
mobility of rhodium atoms on the Au(111) Ru(IV) is present [94].
surface. A number of small rhodium The controlled deposition of ruthe-
clusters were observed on the flat rhodium nium on well-defined surfaces, such
bilayer (Fig. 14e). These results suggest as Pt(hkl) [95–103] and Au(hkl) [38–40],
that the rhodium deposition proceeds has been characterized by electrochemi-
through the SK mode [37]. This behav- cal measurements, Fourier transform in-
ior contrasts significantly with palladium frared reflection-absorption spectroscopy
on Au(111) and Au(100) electrodes where (FT-IRRAS), XPS and STM measure-
a layer-by-layer growth (Frank-van der ments. The interest in these studies is
Merwe growth mode) of palladium was mainly concentrated on the ruthenium
observed clearly. modification of a platinum surface be-
Electrochemical characterization also cause of its extreme importance in
showed that the rhodium layers on Au(111) electrocatalysis. It has been demon-
with a thickness of two monolayers or strated that a ruthenium-deposited Pt(111)
less behave similarly to a well-ordered substrate showed an extremely high
Rh(111) single-crystal electrode, while activity in methanol oxidation com-
thicker rhodium layers are similar to elec- pared to ruthenium-deposited Pt(hkl)
trochemically disordered Rh(111), where electrodes with other crystallographic
the surface has a higher defect density [37]. orientations [98, 99].
The studies of ruthenium deposition
4.1.2.4 Ruthenium are mainly carried out by electrochemi-
The principle application of ruthenium cal and spontaneous (electroless) deposi-
is in the electronics industry for the tion procedures. Friedrich and colleagues
manufacture of resistors [2]. Ruthenium found that the well-reproduced current-
is attracting more attention because of its potential curves of ruthenium-modified
crucial role in electrooxidation catalysis in Pt(111) electrodes were obtained when
ruthenium was deposited electrochem- worse when 0.1 M HClO4 was used as
ically in the potential region between supporting electrolyte during the electro-
+0.8 V and +0.3 V (vs. RHE) from a chemical deposition [95–97]. On the other
freshly prepared 0.1 M H2 SO4 containing hand, Wieckowski and colleague deposited
5 mM RuCl3 . The reproducibility became ruthenium from 0.1 M HClO4 containing
30
20
10
Current density
[µA cm−2]
−10
−20
Pt(111)
Ru/Pt(111) Edep = 0.6 V
−30
Ru/Pt(111) Edep = 0.3 V
Fig. 15 CVs of Pt(111) modified by
ruthenium in 0.1 M H2 SO4 + 5 mM −40
RuCl3 at +0.6 and +0.3 V in 0.1 M 0 200 400 600 800 1000
HClO4 in comparison to the bare
Pt(111) voltammogram in 0.1 M HClO4 . E (RHE)
Sweep rate: 20 mV s−1 [97]. [mV]
25 nm 25 nm
(a) (b)
Fig. 16 STM images of ruthenium-modified Pt(111) electrodes recorded in 0.1 M HClO4 at
+0.5 V. Deposition was performed prior to imaging in a standard glass cell from 0.1 M
H2 SO4 + 5 mM RuCl3 at +0.6 V for (a) 5 min and (b) 30 min, respectively [97].
dilute RuCl3 by a spontaneous-deposition the various oxidation states of deposited

method and suggested that the RuOad ruthenium, the coverage of ruthenium
species was formed on the Pt(111) sur- cannot be estimated reliably from the elec-
face [98]. Figure 15 compared the CVs trical charge passed. The coverage of ruthe-
of Pt(111) before (solid line) and after nium in these experiments was mainly
(dashed and dotted lines) the electro- determined from STM, XPS, or Auger
chemical deposition of ruthenium [97]. experiments instead of electrochemical
The characteristic ‘‘butterfly’’ peak ob- measurements.
served at a clean Pt(111) changed, and Figure 16 shows in situ STM images
a new peak was observed in the hydrogen- of Pt(111) electrodes in 0.1 M HClO4 at
adsorption region after ruthenium mod- +0.5 V after electrochemical deposition
ification. Similar changes in CVs were of ruthenium in 0.1 M H2 SO4 + 5 mM
also observed by Wieckowski and col- RuCl3 at +0.6 V for (a) 5 min and
leagues, where a spontaneous-deposition (b) 30 min, respectively [96, 97]. The for-
method was employed [98–100, 104]. The mation of ruthenium islands of 3 ∼ 5 nm
formation of ruthenium on the platinum in diameter were conformed on the ter-
surface has been confirmed by XPS, race. It seemed to be difficult to control
Auger, and FT-IRRAS as well as STM the size of the ruthenium islands. The cov-
measurements [96–101, 104]. Because of erage of ruthenium in Fig. 16(a) and (b)
50.0 0.5 nm 50.0 0.5 nm
0.2 nm 0.2 nm
25.0 0.0 nm 25.0 0.0 nm
0 0
0 25.0 50.0 0 25.0 50.0
(a) nm (b) nm
50.0 1.0 nm 100 1.0 nm
0.5 nm 75 0.5 nm
25.0 0.0 nm 50 0.0 nm
25
0 0
0 25.0 50.0 0 25 50 75 100
(c) nm (d) nm
Fig. 17 STM images of the Pt(111) after spontaneous deposition of ruthenium for (a) 10 s, (b) 20 s,
(c) 40 s (50 × 50 nm2 ) and (d) 90 s (100 × 100 nm2 ). See Table 2 for details [100].
was estimated to be 0.25 and 0.6 from the the inner monoatomic layer has not yet
STM images and was in agreement with been completed. The bilayer islands, cor-
the calibration by XPS measurement. The responding to 10% of the overall islands,
density of the ruthenium islands depends were observed clearly at coverage of 0.14 to
on the deposition potential and deposition 0.19 ML (Fig. 17c,d). Similar bilayer island
time [97]. structures of vapor-deposited ruthenium
Figure 17 shows STM images of a on a Pt(111) surface have been observed
Pt(111) electrode obtained by Wieckowski by Iwasita and coworkers [103]. Extend-
and colleagues in air after immersion in ing the immersion time more than 90 s
0.1 M HClO4 solution containing 0.5 mM does not result in higher ruthenium cov-
RuCl3 for (a) 10 s, (b) 20 s, (c) 40 s and erage. The limit of ruthenium coverage on
(d) 90 s [100]. In order to keep the Pt(111) the Pt(111) surface was approximately 0.2,
substrate clean for longer STM-imaging much less than that from electrochemical
times, after ruthenium deposition, the deposition. When an Ru-modified elec-
platinum surface was covered with an io- trode was polarized in hydrogen-evolution
dide monolayer to keep the surface free region, the amount of the ruthenium de-
from contamination since the iodide layer posited also decreased, indicating that a
has no effect in the STM topography of ruthenium oxide-reduction process in that
the Ru-modified surface. Ruthenium is- potential region may be accompanied by
lands of nanometer size were observed partial ruthenium dissolution from the
on the Pt(111) surface and their number surface [100].
increased with an increase in immer- Recently, Behm and colleagues investi-
sion time. Table 2 summarizes the results gated the electrodeposition of ruthenium
obtained from these STM images. The ison reconstructed [38] and unreconstructed
lands are not uniformly distributed over Au(111) electrodes in detail [39]. It was
the surface when the coverage is low found that the nucleation of ruthenium
(Fig. 17a and b). There is no selective islands during the electrodeposition on
growth of islands at the steps, indicat- a reconstructed Au(111) electrode occurs
ing that the electroless deposition is not exclusively at the elbows of the Au re-
nucleated by the crystallographic defects construction and that the fcc area of the
at the surface. When the coverage is above reconstructed Au(111) surface is decorated
0.14 ML, a second layer deposit on top at higher coverage [38]. However, a ran-
of the inner layer was observed, although dom nucleation process was observed on
Tab. 2 Ruthenium coverage and the ruthenium island heights as

determined by STM as a function of deposition time (Fig. 18) [100]
Deposition time [s] Ru coverage [ML] Relative height in the

island center [nm]
10 0.01 0.22
20 0.08 0.22
40 0.14 0.22 and 0.45
90 0.19 022 and 0.45
120 0.19 0.22 and 0.45
Fig. 18 In situ STM image

(4000 × 4000 nm2 ) of Au(111) surface
at +0.3 V in 0.05 M H2 SO4 + 1.6 mM
RuNO(NO3 )x (OH)y (x + y = 3). STM
imagining was carried out in the
2000 nm region (shown by the white
box) in the beginning when the potential
was swept between +0.3 V
and – 0.3 V (2 mV s−1 ) for
approximately 2 hr. When the imaging
region was enlarged to 4000 nm,
ruthenium deposition was observed
outside of the tip-scan region [40].
0 2.00 4.00 µm
an unreconstructed Au(111) surface [39] STM-tip effect during the ruthenium de-
as was observed on a Pt(111) electrode position.
(cf. Figs. 16 and 17). The ruthenium is As described in the preceding text, al-
deposited in the entire Au double-layer though it became possible to grow electro-
potential region and a multilayer of ruthe- chemically the ultrathin ruthenium films,
nium can be deposited when the potential which can be applied in electrocatalysis,
becomes more negative than – 0.1 V (vs. the epitaxial growth of a metallic ruthe-
Ag/AgCl). nium film seems to be difficult at the
The similar random nucleation pro- present stage.
cess of ruthenium deposition on an It should be mentioned here that the
unreconstructed Au(111) electrode was composition of the electrolyte solution
observed by Uosaki and colleague in- used in these in situ STM studies of the
dependently [40]. Figure 18 shows STM electrochemical deposition of noble metals
images (4000 × 4000 nm2 ) at +300 mV (platinum, palladium, rhodium and ruthe-
nium) are quite different from those used
(vs. Ag/AgCl) in 0.05 M H2 SO4 solution
in the real electroplating industry [56–58].
containing 1.6 mM RuNO (NO3 )x (OH)y
Although some experimental conditions
(x + y = 3). It has been reported that the
(temperature, concentration) may be dif-
RuNO complex is more stable than that
ficult for the in situ STM measurements,
of the RuCl3 species in solution [104].
electrodeposition in a practical electrolyte
Before this image was obtained, STM
bath should provide more information
imaging was carried out at the nar-
both in application and in fundamental
row center portion of the image (shown fields.
by the dotted white box) for approxi-
mately 30 min, and no clear ruthenium- 4.1.3
nucleation island was observed. When Atomically Controlled Dissolution of Noble
the tip scan region was enlarged, how- Metals
ever, many ruthenium-nucleation islands
were randomly formed outside of the re- Knowledge about the metal-dissolut-
gion originally scanned, indicating a large ion mechanism is very important for
500
400
I-Pd(100)
300
[µA cm−2] 200

I
100
Pd(100)
−100
0 0.5 1.0 1.5

E vs RHE
[V]
Fig. 19 CV of I-free and I-modified Pd(100) electrode in
0.05 M H2 SO4 solution [117].
controlling the metal surface structure. electrodes in sulfuric solution was cat-
Studies of anodic dissolution of metal alyzed by a chemisorbed iodine mono-
at atomic resolution have been reviewed layer for the first time [113–116]. The
elsewhere [23, 24, 105]. In this section, electrochemical dissolution processes of
the discussion will be concentrated on palladium modified by iodine have been
the anisotropic layer-by-layer dissolution investigated in detail by a combination of
behavior of noble metal electrodes, ultrahigh vacuum electrochemistry (UHV-
palladium and gold, catalyzed by an EC) and Auger as well as low-energy
adlayer of adsorbate on the substrate. electron diffraction (LEED) [117, 118].
Similar anisotropic anodic dissolution Figure 19 compares the CVs of a bare
processes catalyzed by the ordered Pd(100) and an I-modified Pd(100) elec-
adsorbed layer have also been observed trode in 0.05 M H2 SO4 solution [117]. A
on S-Ni(100), [106] I-Ni(111), [107] I- small anodic peak due to the oxidation
Ag(100) [108] and Cl-Cu(hkl) [109–112] as of the Pd(100) surface was observed at
well as Cl-Au(hkl) [29–31], which will be +0.90 V (vs. RHE) on the bare Pd(100)
discussed in the next section in detail. electrode (Reactions 1 and 2):
Pd + H2 O −−−→ Pd − OHs + H+ + e−
4.1.3.1 Palladium
(1)
Soriaga and colleagues reported that the
dissolution of palladium single-crystal Pd − OH(s) −−−→ PdO(s) + H+ + e− (2)
600
I-Pd (111)
500
I-Pd (100)
I-Pd (110)
400
[µA cm−2]
300
I
200
100
−100
0.5 1.0 1.5
E vs RHE
[V]
Fig. 20 CVs of I-modified Pd(111), Pd(100) and Pd(110) electrodes
in 0.05 M H2 SO4 solution [117].
This anodic peak was suppressed com- coverage remains constant throughout the
pletely on the I-modified Pd(100) electrode, dissolution process [117].
and a new anodic peak with an order-of- Figure 20 compares the structure depen-
magnitude larger peak current than that of dence of the I-catalyzed electrochemical
the I-free Pd(100) electrode was observed dissolution of Pd(hkl) electrodes in 0.05 M
at +1.25 V. This new anodic peak has been H2 SO4 solution. The dissolution rate
attributed to the two-electron dissolution slightly depends on the crystallographic
process of palladium as: orientation and decreases in the order of
I-Pd(110) > I-Pd(111) > I-Pd(100) [117].
Pd(s) −−−→ Pd(aq) 2+ + 2e− (3) Figure 21 shows a sequence of in situ
STM images of an I-modified Pd(100)
Therefore, the anodic dissolution of pal- electrode after the potential was stepped
ladium was enhanced significantly when to +1.05 V for (a) 5, (b) 6, and (c) 7 min,
the iodine adlayer was adsorbed on the where the palladium started to dissolved
palladium surface. The rate of the dissolu- anodically (Fig. 19) [117]. An atomically
tion is directly proportional to the coverage flat Pd(100) surface with a monoatomic
of iodine on the surface, and the iodine step was observed, and the I-catalyzed
T2
T3 T2 T3
[001]
[010]
T1
T1
(a) (b)
T2
0 40 80 nm
(c)
Fig. 21 STM images (80 × 80 nm2 ) of Pd(100)-c(2 × 2)-I obtained after (a) 5, (b) 6 and
(c) 7 min of dissolution at +1.05 V in 0.05 M H2 SO4 [117].
dissolution take places only at the step in the corner of the top terrace can be re-
edges in a layer-by-layer mode, while garded as the most reactive ones for the
etching from the pit formation was not dissolution. Removal of atom A (dissolu-
observed at the surface. As shown in tion along [001] direction) easily permits
Fig. 21, terraces T1 , T2 and T3 became the iodine atom to slide down from a 4-
narrow along [001]-directed steps but only fold site at the upper terrace onto another
a minimal change was observed in the hollow site (H) at the lower terrace. On
direction perpendicular to it, that is, an the other hand, iodine drop onto site H
anisotropic dissolution of palladium took cannot be accomplished upon removal of
place. Itaya and colleagues interpret the atom B (dissolution along [010] direction)
dissolution behavior using a schematic because of steric hindrance by atom A.
model (Fig. 22) where iodine formed a Therefore, iodine is easier to diffuse from
c(2 × 2) adlattice on the Pd(100) sur- a high-symmetry site from the upper to
face [117]. The palladium atoms A and B a lower terrace and is considered to be
Fig. 22 Schematic model to show the

anisotropic dissolution of
I-Pd(100) [117].
iodine adlayers play a significant role in

H
A the iodine-catalyzed dissolution process.
B Therefore, the dissolution process of the
C I-modified Pd(111) and Pd(100) electrodes
take place in the layer-by-layer sequence
without disruption of the iodine-adlattice
structure.
4.1.3.2 Gold
Gold is known to have excellent chemical
stability and is widely used in industry
and as an accessory material [119, 120].
I on the upper terrace
It becomes unstable and dissolves in
[010]
[011] the positive-potential region, especially in
I on the lower terrace
solutions containing Cl− .
[001] Figure 23 shows CVs of (a) Au (111) and
Pd at upper terrace
(b) Au(100) electrodes in 0.1 M HClO4
Pd at lower terrace solution containing 1 mM Cl− in the
potential region between +0.2 and +1.7 V.
A significant increase in the anodic current
the major driving force in the anisotropic started at +1.25 V, and two anodic peaks
dissolution [117]. were found at +1.40 V and +1.51 V at
The step-selective layer-by-layer dissolu- Au(111) electrode. When the potential
tion behavior was also observed on an I- became more positive than +1.55 V, the
Pd(111) surface. However, the anisotropic- anodic current quickly decreased. The
dissolution features are not as obvious pronounced anodic current in the positive-
as that observed on an I-Pd(100) sur- going potential sweep was attributed to
face. The pit formation on the I-modified the 3e− oxidative dissolution of gold
Pd(110) precludes layer-by-layer dissolu- (Reaction 4, forward) [29].
tion and leads to progressive disorder [117,
118]. Au + 4Cl− ←−→ AuCl4 − + 3e− (4)
The ex situ LEED [113–116] and in + −
Au + H2 O ←−→ AuO + 2H + 2e (5)
situ STM measurements with high res-
olution [105, 117] demonstrated that the In the more positive potential region, the
ordered iodine adlayers were formed √ dissolution reaction competed with oxide
on the palladium surfaces as ( 3 ×
√ formation (Reaction 5, forward) on the
3R30◦ )-I-Pd(111), c(2 × 2)-I-Pd(100) and gold surface and passivation was com-
pseudohexagonal-I-Pd-(110). It is notewor- pleted around +1.7 V. The cathodic peaks
thy that the same adlattice structure of observed at +1.21 and +1.10 V in the
iodine was observed on the palladium ter- negative-going potential sweep were re-
race after anodic dissolution. The ordered lated to the simultaneous oxide reduction
10 µA cm−2 10 µA cm−2
200 200
Peak I Peak II Peak I Peak II
0 0
0.5 1.0
0.5 1.0
Current density
100
100
[µA cm−2]
0
0
−100
0 0.5 1.0 1.5 0 0.5 1.0 1.5

Potential vs RHE Potential vs RHE
(a) [V] (b) [V]
Fig. 23 CVs of (a) Au(111) and (b) Au(100) electrodes in 0.1 M HClO4
solution containing 1 mM Cl− in the potential region between +0.2 V and
+1.7 V. Insets show CVs observed in the potential region between +0.2 and
+1.2 V. Sweep rate: 20 mV s−1 [31].
(Reaction 5, backward)/gold dissolution were parallel to the [110] direction, crossed

(Reaction 4, forward), and the redeposi- each other at 60 or 120 degrees. When the
tion of the dissolved gold species (Reac- potential was slowly swept (2 mV s−1 ) in
tion 4, backward), respectively, as already the positive direction, drastic changes in
confirmed by EQCM measurements [29]. the shape of the step lines were observed.
Similar 3e− anodic dissolution of Au(100) Figure 24 shows an STM image obtained
electrode in acidic solution containing Cl− during the potential sweep from +0.99 V
(Fig. 23b) is expected [31]. (top) to +1.09 V (bottom), close to that
Enlarged CVs obtained in the potential of Peak II observed at Au(111) (Fig. 23a).
region between +0.2 and +1.2 V in Although no change was found on the
Fig. 23 show an anodic peak around terraces, several kinks were formed in the
0.5 V (Peak I) and a pair of redox spikes originally straight step lines. Thus, the gold
around +1.0 V (Peak II). The Peak I can surface really started to dissolve from the
be attributed to Cl− -induced lifting of the step sites at this potential. The outline of
surface reconstruction on both Au(111) the step lines changed drastically on the
and Au(100) electrodes [81]. Peak II is not bottom part compared to that on the top
observed at all in the Cl− -free solution and part in the STM image, demonstrating
has been attributed to the formation of an that the gold-dissolution rate increased
ordered chloride adlayer on the Au(111) as the electrode potential became more
and Au(100) surfaces. positive. The SXS measurement showed
Figure 24 shows sequentially obtained that the chloride coverage on the Au(111)
in situ STM images (1 × 1 µm2 ) of an electrode increased with a slope of about
Au(111) surface in a 0.1 M HClO4 solution 1.5%/V after the ordered chloride layer was
containing 1 mM Cl− [30, 31]. Stable large constructed at this potential region [121].
terraces and monoatomic step lines were Thus, potential induced ‘‘compression’’ of
observed at +0.8 V as was the case in the chloride adlayer and initial dissolution
Cl− free solution. The step lines, which on the Au(111) electrode seemed to
(a) (b)
(c) (d) 0 0.5 1 µm
Fig. 24 A sequence of in situ STM images (1 × 1 µm2 ) of the Au(111) surface in 0.1 M HClO4
containing 1 mM Cl− . (a) At +0.8 V, (b) +0.99 V ∼ +1.09 V during a potential
sweep (2 mV s−1 ) with rastering the tip downwards, (c) 0 min, (d) 6 min and (e) 40 min after
the potential was swept to +1.27 V, (f) potential was stepped to +1.36 V at the moment shown
by the arrow with rastering the tip downwards, (g) next scan after (f) with rastering the tip
upwards, (h) 3 min later from (f) with rastering tip downwards. Etip = +1.2 V. iT = 5 nA. A
compass was drawn to indicate the orientation of the surface. See the text for details [30].
take place in this potential region. The was etched mainly from its step edge
dissolution proceeded mainly at the step in prolonged etching at +1.27 V (6 and
sites, and the originally observed straight 40 min., Figs. 24d–e). At the beginning of
step lines were etched like the teeth the dissolution, only steps along the [110]
of a saw at more positive potential direction were observed (Figs. 24b,c). As
(+1.27 V, Fig. 24c). The dissolution also shown by the arrows in Fig. 24d, some
proceeded on the terrace (Fig. 24c) but of the step lines were rotated about
only from spots where small pits or 30 degrees, and new step lines running
defects were found at more negative along the [211] direction of the Au(111)
potentials (Figs. 24a,b) as indicated by substrate were observed. The step edges
the black circles. The Au(111) surface along the [211] direction, however, became
(e) (f)
(g) (h) 0 0.5 1 µm
[110]
[211]
Fig. 24 (Continued)
dominant after the electrode was kept at part of the image, rectangular-shaped pits
+1.27 V for a long time (Fig. 24e). were clearly observed. In the next STM
As the potential was stepped to a more scan (Figs. 24f,g), rectangular pits were
positive values (+1.36 V, Figs. 24f–h), clearly observed on all terraces. It is
where gold still dissolves without oxide interesting to note that the rectangular
formation, at the moment shown by an pits were parallel to each other within the
arrow in Fig. 24(f), acceleration of the same domain. Furthermore, the direction
dissolution rate, especially on the terraces, of the long axis of the rectangular pits
was observed. Many small pits were was parallel to the [211] direction of the
formed on the terraces. In the bottom substrate, that is, the new step direction
after dissolution (Fig. 24e). The pits grew for longer periods (Figs. 25e,f). These
and then merged. Figure 24(h) shows an results suggest that anodic dissolution of
STM image captured about 2 min after the Au(100) electrode surface is also an
Fig. 24(g). The top layer of the Au(111) anisotropic layer-by-layer etching process
was almost etched away, and the next in which the step edges preferentially
layer of the gold surface started to dissolve. retreated along the [100] direction of the
Nearly all of the step lines were orientated Au(100) substrate [31].
along the [211] direction, and the step As described in the preceding text,
line along the [110] direction was absent. the anisotropic-dissolution behavior was
These results demonstrated that gold was observed both on Au(111) and Au(100)
dissolved layer by layer in an anisotropic electrode surfaces only in acidic solu-
way in which the step edges preferentially tion containing Cl− . The structure of
retreated along the [211] direction of the the chloride adlayer on the gold surface
Au(111) substrate. is considered to be one of the impor-
STM observations were also carried out tant origins of the anisotropic-dissolution
in solutions of various Cl− concentrations. process. Suggs and Bard observed that cop-
It was found that the critical potential of the per dissolved in aqueous-chloride solution
dissolution depended on the Cl− concen- from the step edges respectively retreat-
tration. The higher the Cl− concentration, ing along the [211] and [100] directions on
the more negative the critical potential. Cu(111) [109] and Cu(100) electrode sur-
This potential dependence is consistent faces [110]. Vogt and coworkers observed
with that of Peak II. In solutions contain- a similar anisotropic dissolution on a
ing Cl− at less than 0.1 mM, the dissolu- Cu(100) electrode surface in HCl solution
tion rate was very low, and both the [211] and proposed that the preferential forma-
and [110] step lines were observed after the tion of the [100]-step edge resulted from
electrode was kept at +1.36 V for 20 min. its higher stability because the [100]-step
In a solution of higher Cl− concentration, edge of the Cu(100) was more stable than
that is, more than 10 mM, the anisotropic that of the [110]-step edges by 3 to 4 orders
characteristics were hard to observe be- of magnitude [111, 112].
cause of the very high dissolution rate. Figure 26 illustrates a real-space model
The anisotropic dissolution behavior was of the ordered chloride adlayer and the un-
also observed on the Au(100) electrode derlying Au(111) at the potential of Peak II,
in the similar solutions (Fig. 25). The based on the SXS measurements [121]. A
dissolution takes place only on the step chloride adlayer with a close-packed hexag-
edges in the potential region around onal structure is not commensurate with
Peak II (Fig. 25a). At the beginning of the gold substrate, and the interatomic
the dissolution, steps only along the [110] distance was close to the van der Waals
direction were observed (Figs. 25b,c). As diameter of chloride. The highest density
shown by the arrows in Fig. 25(d), many of the chloride adatom was found in the
step lines rotated 45 degrees from the step lines along the [211] direction, and
original [110] direction, that is, the [100] the adlayer was compressed more with
direction of the Au(100) substrate, after increasing potential. As the STM results
10 min etching at +1.30 V. Almost all of demonstrated in Fig. 24, the original step
the step lines were running along the [100] lines along the [110] direction disappeared
direction after being polarized at +1.30 V and new [211] step lines were observed
(a) (b) (c)
(d) (e) (f)
[110]
[100]
Fig. 25 A sequence of in situ STM images of the Au(100) surface in 0.1 M HClO4 containing 1 mM Cl− . (a) At +0.8 V (1 × 1 µm2 ), (b) At +1.1 V,
4.1 Atomically Controlled Electrochemical Deposition and Dissolution of Noble Metals
(c) 0 min, (d) 10 min, (e) 20 min, and (f) 40 min after the potential was swept to +1.30 V (0.5 × 0.5 µm2 ). Etip = +1.40 V. iT = 3 nA. A compass is
drawn to indicate the orientation [31].
507
[211]
[110]
Gold atom (1st layer) Chloride (1st layer)
Gold atom (2nd layer) Chloride (2nd layer)
Fig. 26 A schematic model illustrating the chloride adlayer

on an Au(111) surface based on the surface X-ray diffraction
measurement. See the text for details [30].
on the surface after the anodic dissolu- chloride adlayer structure has also been
tion. The lateral interactions between the observed using in situ STM by Kolb and
adsorbed chloride and gold were consid- colleagues [123]. The chloride adlayer with
ered to stabilize the atom rows along a higher Cl− adatom density on the step
the [211] direction more than those in edges of the [100] direction will make these
the other directions because of its highest step edges more stable than those of [110]
chloride-adatom density. In other words, a and other directions on the Au(100) elec-
much lower etching rate on the [211] step trode surface in the positive potential
lines was expected, and as a result, the region. Because of the stability, the [110]-
anisotropic dissolution of Au(111) in the step lines, which were originally observed
Cl− containing solution was observed. before dissolution, disappeared and only
In the case of an Au(100) electrode, the [100]-step lines can be observed after
Wang and coworkers observed using SXS the anodic dissolution.
measurements that chloride√formed uni-
axially incommensurate c( 2 × p)R45◦ 4.1.4
adlayer structure on the Au(100) elec- Summary
trode surface and was compressed along
the [100] direction when the potential Electrochemical deposition of noble metal
became more positive [122]. A similar elements (platinum, palladium, rhodium,
and ruthenium) on substrates (mainly gold and Technology Foundation, the Japan Se-
and platinum single-crystal electrode sub- curities Scholarship Foundation and the
strates) has been investigated by many Tokuyama Science and Technology Foun-
modern instrumental analysis methods, dation.
such as STM, EQCM, and SXS measure-
ments at an atomic level. As summarized References
in Table 1, electrochemical deposition of
palladium on a gold substrate takes places 1. F. R. Hartley, (Ed.), Chemistry of the Plat-
inum Group Metals, Elsevier Scientific Pub-
in the FM mode, similar to that of silver.
lishers B. V., Amsterdam, 1991.
Platinum and rhodium grow electrochem- 2. A. Cowley, M. Steel, Platinum 2001, John-
ically on a gold substrate in the SK mode. son Matthey, London, 2001.
The electrochemical growth of ruthenium 3. G. A. Somorjai, Surface Chemistry and Catal-
on a gold substrate prefers the SK mode, ysis, John Wiley & Sons, New York, 1994.
4. C. Argile, G. E. Rhead, Surf. Sci. Rep. 1989,
and a metallic-ruthenium film is still diffi- 10, 277–356.
cult to grow at the moment. These types of 5. J. A. Rodriguez, Surf. Sci. Rep. 1996, 24,
behavior can be understood qualitatively 223–287.
using the mismatch parameter of lattice 6. H. Brune, Surf. Sci. Rep. 1998, 31, 121–229.
7. T. T. Kodas, M. J. Hampden-Smith, The
and surface energy between the deposited Chemistry of Metal CVD, VCH, Weinheim,
metal and the gold substrate. 1994.
The anistropic layer-by-layer dissolution 8. M. K. Hampden-Smith, T. T. Kodas, Chem.
of palladium and gold are enhanced by Vapor Deposition 1995, 1, 8–23.
9. J. Clavilier, R. Faure, G. Guinet et al.,
an adlayer of iodine and chloride on the
substrate surfaces. 10. G. Binnig, H. Rohrer, C. Gerber et al., Appl.
The present information is very useful Phys. Lett. 1982, 40, 178.
for creating electrocatalytic materials with 11. G. Binning, C. F. Quate, C. Gerber, Phys.
high activity and selectivity. This is also Rev. Lett. 1986, 12, 930.
12. O. M. Magnussen, J. Hagebock, J. Hotlos
very important for understanding the et al., Faraday Discuss. 1992, 94, 329–338.
reaction mechanism of the metal epitaxy in 13. A. Cuesta, M. Kleinert, D. M. Kolb, Phys.
the field of surface electrochemistry. Very Chem. Chem. Phys. (PCCP) 2000, 2,
rapid progress is expected in this field. 5684–5690.
14. A. M. Funtikov, U. Stimming, R. Vogel,
Surf. Sci. 1995, 324, L343–L348.
Acknowledgment 15. A. M. Funtikov, U. Stimming, R. Vogel,
16. L. J. Wan, T. Suzuki, K. Sashikata et al.,
This work was partially supported by J. Electroanal. Chem. 2000, 484, 189–193.
Grants-in-Aid for Scientific Research on 17. L. J. Wan, S. L. Yau, K. Itaya, J. Phys. Chem.
Priority Area of ‘‘Electrochemistry of Or- B 1995, 99, 9507–9513.
dered Interfaces’’ (No. 09237101) and for 18. L. J. Wan, M. Hara, J. Inukai et al., J. Phys.
Chem. B 1999, 103, 6978–6983.
Encouragement of Young Scientists (No. 19. M. F. Toney, J. McBreen, Interface 1993, 1,
12750723, 10740314) from the Ministry 22–31.
of Education, Science, Sports and Cul- 20. J. X. Wang, R. R. Adzic, B. M. Ocko in Inter-
ture, Japan. SY acknowledges the support facial Electrochemistry (Ed.: A. Wieckowski),
Marcel Dekker, New York, 1999,
from PRESTO, Japan Science and Tech- pp. 175–186.
nology Corporation (JST). SY acknowl- 21. D. A. Buttry, M. D. Ward, Chem. Rev. 1992,
edges grants from the Iketani Science 92, 1355–1379.
22. M. Hepel in Interfacial Electrochemistry (Ed.: 45. M. J. Esplandiu, M. A. Schneeweiss, D. M.

A. Wieckowski), Marcel Dekker, New York, Kolb, Phys. Chem. Chem. Phys. (PCCP)
1999, pp. 599–630. 1999, 1, 4847–4854.
23. A. A. Gewirth, B. K. Niece, Chem. Rev. 1997, 46. F. H. Reid, Trans. Inst. Metal Finishing 1970,
97, 1129–1162. 48, 115–123.
24. K. Itaya, Prog. Surf. Sci. 1998, 60, 121–248. 47. J. F. Llopis, F. Colom in Encyclopedia of
25. E. Bauer, Z. Kristallogr. 1958, 110, 372–394. Electrochemistry of the Elements (Ed.:
26. E. Bauer, J. H. van der Merwe, Phys. Rev. B A. J. Bard), Marcel Dekker, New York, 1976,
1986, 33, 3657–3671. pp. 169–219, Vol. VI.
27. E. Budevski, G. Staikov, W. J. Lorenz, 48. S. Tanaka in Science of Noble Metal, Applica-
Electrochemical Phase Formation and tion (Ed.: S. Tanaka), Tanaka Nobel Metal,
Inc., Tokyo, 1985, pp. 261–321.
Growth – An Introduction to the Initial Stages
49. L. Carrette, K. A. Friedrich, U. Stimming,
of Metal Deposition, VCH, Weinheim, 1996.
Chem. Phys. Chem. 2000, 1, 162–193.
28. L. Z. Mezey, J. Giber, Jpn. J. Appl. Phys.
50. S. Wasmus, A. Kuever, J. Electroanal. Chem.
1982, 21, 1569–1571.
1999, 461, 14–31.
29. S. Ye, C. Ishibashi, K. Uosaki, J. Elec- 51. A. M. Feltham, M. Spiro, Chem. Rev. 1971,
trochem. Soc. 1998, 145, 1614–1623. 71, 177–193.
30. S. Ye, C. Ishibashi, K. Uosaki, Langmuir 52. J. W. A. Sachtler, M. A. Van Hove, J. P.
1999, 15, 807–812. Biberian et al., Surf. Sci. 1981, 110, 19–42.
31. S. Ye, C. Ishibashi, K. Uosaki in Scanning 53. R. F. C. Farrow, G. R. Harp, R. F. Marks
Probe Microscopy for Electrode Characteri- et al., J. Cryst. Growth 1993, 133, 47–58.
zation and Nanometer Scale Modification, 54. T. R. Linderoth, J. J. Mortensen, K. W.
Electrochemical Society Proceedings (Eds.: Jacobsen et al., Phys. Rev. Lett. 1996, 77,
D. C. Hansen, H. S. Isaacs, K. Sieradzki), 87–90.
Electrochemical Society, New York, 2001, 55. M. O. Pedersen, S. Helveg, A. Ruban et al.,
pp. 133–147, Vol. PV 2000-35. Surf. Sci. 1999, 426, 395–409.
32. K. Uosaki, S. Ye, H. Naohara et al., J. Phys. 56. Technical Catalog for Metal Plating, Tanaka
Chem. B 1997, 101, 7566–7572. Noble Metal, Inc., 2000.
33. H. Naohara, S. Ye, K. Uosaki, J. Phys. Chem. 57. Precious Metal Plating Process Catalog, N. E.
B 1998, 102, 4366–4373. CHEMCAT, 2001.
34. H. Naohara, S. Ye, K. Uosaki, J. Electroanal. 58. Technical Catalog for Metal Plating, Japan
Chem. 1999, 473, 2–9. Pure Chemical (JPC), 2001.
35. L. A. Kibler, M. Kleinert, R. Randler et al., 59. P. E. Skinner, Platinum Met. Rev. 1989, 33,
Surf. Sci. 1999, 443, 19–30. 102–105.
36. L. A. Kibler, M. Kleinert, D. M. Kolb, Surf. 60. A. J. Gregory, W. Levason, R. E. Noftle
Sci. 2000, 461, 155–167. et al., J. Electroanal. Chem. 1995, 399,
37. L. A. Kibler, M. Kleinert, D. M. Kolb, J. Elec- 105–113.
61. S. Tanaka in Science of Noble Metal, Applica-
troanal. Chem. 1999, 467, 249–257.
tion (Ed.: S. Tanaka), Tanaka Noble Metal,
38. S. Strbac, O. M. Magnussen, R. J. Behm,
Tokyo, 1985, pp. 225–247.
Phys. Rev. Lett. 1999, 83, 3246–3249.
62. B. Gollas, J. M. Elliott, P. N. Bartlett, Elec-
39. S. Strbac, F. Maroun, O. M. Magnussen
trochim. Acta 2000, 45, 3711–3724.
et al., J. Electroanal. Chem. 2001, 500, 63. K. Uosaki, S. Ye, Y. Oda, T. Haba et al.
479–490. in Proceedings of the Sixth International
40. K. Uosaki, J. Ueda, S. Ye, in preparation. Symposium on Electrode Processes (Eds.:
41. N. Ikemiya, K. Yamada, S. Hara, J. Vac. Sci. A. Wieckowski, K. Itaya), Electrochemical
Technol., B 1995, 14, 1369–1375. Society, Pennington, New Jersey, 1996,
42. N. Ikemiya, K. Yamada, S. Hara, Surf. Sci. pp. 168–179, Vol. PV 96-8.
1996, 348, 253–260. 64. K. Uosaki, S. Ye, Y. Oda et al., Langmuir
43. W. J. Lorenz, G. Staikov, Surf. Sci. 1995, 1997, 13, 594–596.
335, 32–43. 65. H.-F. Waibel, M. Kleinert, D. M. Kolb in
44. K. Ogaki, K. Itaya, Electrochim. Acta 1995, 199th Meeting of the Electrochemical Society,
40, 1249–1257. Washington, 2001, pp. 354.
66. W. Haiss, J. K. Sass, J. Electroanal. Chem. 90. G. A. Attard, R. Price, A. Al-Akl, Surf. Sci.
1997, 431, 15–17. 1995, 335, 52–62.
67. T. Katsuzaki, K. Uosaki, unpublished. 91. F. H. Reid, J. C. Blake, Trans. Inst. Metal
68. K. A. Friedrich, A. Marmann, U. Stimming Finishing 1961, 38, 45–51.
et al., Fresenius’ J. Anal. Chem. 1997, 358, 92. J. F. Llopis, I. M. Tordesillas in Encyclopedia
163–165. of Electrochemistry of the Elements (Ed.:
69. E. G. Seebauera, C. E. Allen, Prog. Surf. Sci. A. J. Bard), Marcel Dekker, New York, 1976,
1995, 49, 265–330. pp. 277–298, Vol. VI.
70. N. Kimizuka, T. Abe, K. Itaya, Denki Kagaku 93. K. R. Seddon, Platinum Met. Rev. 1996, 40,
1993, 61, 796–799. 128–134.
71. S.-L. Yau, T. Moriyama, H. Uchida, M. Wat- 94. F. P. Gortsema, J. W. Cobble, J. Am. Chem.
anabe, Chem. Commum. 2000, 2279–2280. Soc. 1961, 83, 4317–4321.
72. J. F. Llopis, F. Colom in Encyclopedia 95. K. A. Friedrich, K.-P. Geyzers, U. Linkeb
of Electrochemistry of Elements (Ed.: et al., J. Electroanal. Chem. 1996, 402,
A. J. Bard), Marcel Dekker, New York, 1976, 123–128.
pp. 253–275, Vol. VI. 96. S. Cramm, K. A. Friedrich, K.-P. Geyzers
73. A. Cuesta, L. A. Kibler, D. M. Kolb, J. Elec- et al., Fresenius’ J. Anal. Chem. 1997, 358,
troanal. Chem. 1999, 466, 165–168. 189–192.
74. G. A. Attard, A. Bannister, J. Electroanal. 97. K. A. Friedrich, K. P. Geyzers, A. Marmann
Chem. 1991, 300, 467–485. et al., Zeitschrift fur Physikalische Chemie
75. M. J. Llorca, J. M. Feliu, A. Aldaz et al., 1999, 208, 137–150.
J. Electroanal. Chem. 1993, 351, 299–319. 98. W. Chrzanowski, A. Wieckowski, Langmuir
76. M. J. Llorca, J. M. Feliu, A. Aldaz et al.,
1997, 13, 5974–5978.
99. W. Chrzanowski, A. Wieckowski in Interfa-
77. M. Baldauf, D. M. Kolb, Electrochim. Acta
cial Electrochemistry (Ed.: A. Wieckowski),
1993, 38, 2145–2153.
Marcel Dekker, New York, 1999,
78. H. Naohara, S. Ye, K. Uosaki, Colloids Surf.,
pp. 937–954.
A 1999, 154, 201–208.
100. E. Herrero, J. M. Feliu, A. Wieckowski,
79. H. Naohara, S. Ye, K. Uosaki, Electrochim.
Langmuir 1999, 15, 4944–4948.
Acta 2000, 45, 3305–3309.
101. H. Kim, I. R. de Moraes, G. Tremiliosi
80. H. Naohara, S. Ye, K. Uosaki, J. Electroanal.
Chem. 2001, 500, 435–445. et al., Surf. Sci. 2001, 474, L203–L212.
81. A. S. Dakkouri, D. M. Kolb in Interfacial 102. W. F. Lin, M. S. Zei, M. Eiswirth et al., J.
Electrochemistry (Ed.: A. Wieckowski), Mar- Phys. Chem. B 1999, 103, 6968–6977.
cel Dekker, New York, 1999, pp. 151–173. 103. T. Iwasita, H. Hoster, A. John-Anacker
82. M. Takahashi, Y. Hayashi, J. Mizuki et al., et al., Langmuir 2000, 16, 522–529.
Surf. Sci. 2000, 461, 213–218. 104. G. Tremiliosi-Filho, H. Kim, W. Chrzan-
83. T. Kondo, K. Tamura, M. Takahashi, owski et al., J. Electroanal. Chem. 1999, 467,
J. Mizuki, K. Vosaki, Electrochim. Acta, in 143–156.
press (2002). 105. K. Itaya in Interfacial Electrochemistry (Ed.:
84. M. Baldauf, D. M. Kolb, J. Phys. Chem. 1996, A. Wieckowski), Marcel Dekker, New York,
100, 11 375–11 381. 1999, pp. 187–210.
85. J. F. Llopis, I. M. Tordesillas in Encyclopedia 106. T. Suzuki, T. Yamada, K. Itaya, J. Phys.
of Electrochemistry of the Elements (Ed.: Chem. 1996, 100, 8954–8961.
A. J. Bard), Marcel Dekker, New York, 1976, 107. P. Muller, S. Ando, T. Yamada et al.,
pp. 299–322, Vol. VI. J. Electroanal. Chem. 1999, 467, 282–290.
86. J. Inukai, M. Ito, J. Electroanal. Chem. 1993, 108. T. Teshima, K. Ogaki, K. Itaya, J. Phys.
358, 307–315. Chem. 1997, 101, 2046–2053.
87. R. Gomez, A. Rodes, J. M. Perez et al., Surf. 109. D. W. Suggs, A. J. Bard, J. Am. Chem. Soc.
Sci. 1995, 327, 202–215. 1994, 116, 10 725.
88. R. Gomez, A. Rodes, J. M. Perez et al., Surf. 110. D. W. Suggs, A. J. Bard, J. Phys. Chem.
Sci. 1995, 344, 85–97. 1995, 99, 8349.
89. R. Gomez, J. M. Feliu, Electrochim. Acta 111. O. M. Magnussen, M. R. Vogt, F. A. Moller
1998, 44, 1191–1205. et al., Surf. Sci. 1996, 367, L33–L41.
112. M. R. Vogt, A. Lachenwitzer, O. M. Mag- Marcel Dekker, New York, 1999,

nussen et al., Surf. Sci. 1998, 399, 49–69. pp. 249–267.
113. J. A. Schimpf, J. R. MacBride, M. P. Soriaga, 119. G. M. Schmid, M. E. Curley-Fiorino, Gold,
J. Phys. Chem. 1993, 97, 10 518–10 520. Marcel Dekker, New York, 1975.
114. J. A. Schimpf, J. B. Abreu, M. P. Soriaga, 120. R. J. Puddephatt, The Chemistry of Gold,
Langmuir 1993, 9, 3331–3333. Elsevier Scientific Publishing Company,
115. A. Carrasquillo Jr., J. J. Jeng, R. J. Barriga Amsterdam, 1978.
et al., Inorg. Chim. Acta 1997, 255, 249–254. 121. M. Magnussen, B. M. Ocko, R. R. Adzic
116. M. P. Soriaga, J. A. Schimpf, A. J. Car- et al., Phys. Rev. B 1995, 51, 5510–5513.
rasquillo et al., Surf. Sci. 1995, 335, 122. J. X. Wang, A. R. Adzic, M. Ocko in Interfa-
273–280. cial Electrochemistry (Ed.: A. Wieckowski),
117. K. Sashikata, Y. Matsui, K. Itaya et al., J. Marcel Dekker, New York, 1999,
Phys. Chem. B 1996, 100, 20 027–20 034. pp. 175–186.
118. M. P. Soriaga, Y. G. Kim, J. E. Soto in Inter- 123. A. Cuesta, D. M. Kolb, Surf. Sci. 2000, 465,
facial Electrochemistry (Ed.: A. Wieckowski), 310–316.
4.2 Electrodeposition of Compound Semiconductors by Electrochemical Atomic Layer Epitaxy (EC-ALE) 513
4.2 various compounds. It describes the sta-

Electrodeposition of Compound tus of device formation using EC-ALE,
Semiconductors by Electrochemical Atomic and goes over some of the problems
Layer Epitaxy (EC-ALE) and issues involved in developing cycles
and growing films. Finally, given that
John L. Stickney, Travis L. Wade, Billy H.
EC-ALE is based on surface-limited elec-
Flowers, Raman Vaidyanathan, and Uwe
trochemical reactions, studies of relevant
Happek
electrodeposit surface chemistry are dis-
University of Georgia, Athens, Georgia, USA
cussed.
Abstract
4.2.1
Introduction
This chapter concerns the state of devel-
opment of electrochemical atomic layer Nanometer control in growth is a ma-
epitaxy (EC-ALE), the electrochemical ana- jor frontier of Material Science. By con-
log of atomic layer epitaxy (ALE). EC-ALE structing superlattices, nanowires, nan-
is being developed as a methodology for oclusters, and nanocrystalline materials,
the electrodeposition of compound semi- the electronic structure (band gap) of a
conductors with nanoscale control. ALE is semiconductor can be engineered, the
based on the formation of compounds, one wavelength of light emitted or absorbed
monolayer (ML) at a time, using surface- by a compound can be adjusted over
limited reactions. An atomic layer of one a broad range. By direct analogy with
element can be electrodeposited at a po- the quantum-mechanical model of a par-
tential under that needed to deposit the ticle in a box, it is known that the
element on itself, and this process is smaller the box containing an electron
referred to as underpotential deposition the further apart are energy levels. This
(UPD). EC-ALE is the use of UPD for the translates directly for some semiconduc-
surface-limited reactions in an ALE cycle. tor structures – the smaller the thick-
Electrodeposition is generally performed ness of the layers, or the dimensions
near room temperature, avoiding prob- of a particle, the larger the resulting
lems with interdiffusion and mismatched band gap.
thermal expansion coefficients. This The primary methodologies for form-
makes EC-ALE a good candidate to form ing thin-film materials with atomic
superlattices, where the compound de- level control are molecular beam epi-
posited is modulated on the nanometer taxy (MBE) [1–6], vapor phase epitaxy
scale. (VPE) [7–9], and a number of deriva-
This chapter describes the basics of tive vacuum-based techniques [10]. These
the EC-ALE cycle, the elements that methods depend on controlling the flux
have been used to form deposits, as of reactants and the temperature of the
well as the solutions, rinsing, and po- substrate and reactants. If the growth tem-
tential changes. It describes the hard- perature in MBE and VPE is too high,
ware presently being used by this group especially with the formation of nano-
and other research laboratories, the com- structured materials, even 200 to 500 ◦ C,
pounds that have been formed, and interdiffusion of the component elements
the development of deposition cycles for can result. In addition, different materials
have different thermal expansion coeffi- can be formed electrochemically on Si

cients. If a heterojunction is formed at wafers, using the Dual Damascene copper-
an elevated temperature and then cooled electroplating method [11, 12], and the
to room temperature, stresses and de- large semiconductor companies, such as
fects can develop if the coefficients do not IBM, Intel, AMD, Motorola, and so on, are
match. installing wafer-electroplating machines
Electrodeposition is not ‘‘heat and beat,’’ in their fabrication lines.
it is not a heat-driven reaction; it is gen- Epitaxy in electrodeposition has been
erally performed near room temperature. studied for 40 or more years [13–17], and
Ideally, electrodeposition involves control is similar to MBE and VPE in that the
of the deposition by activity of the electrons rate of deposition should be limited and
at the interface, by the potentials applied have as much surface diffusion as possi-
to the deposition cell. ble. Surface diffusion is limited near room
The work described here concerns devel- temperature, but may be enhanced due
opment of electrochemical methodologies to solvation effects. Epitaxial electrodepo-
to grow compound semiconductors with sition is normally best carried out near
nanoscale or atomic layer control. That equilibrium, where the deposition rate is
thin films of some compounds can be lower and the exchange current high. The
formed electrochemically is clear. The exchange current acts, in addition to sur-
questions are: How much control is there face diffusion, to move depositing atoms
over composition, structure, and morphol- around. Near equilibrium, electrodeposi-
ogy? What compounds can be formed, and tion is dynamic, with slightly more atoms
with what quality? depositing than dissolving at a given time.
The issue of epitaxy is central to the In principle, atoms deposited in less than
formation of high-quality electronic and optimal sites redissolve, leaving only those
optoelectronic structures. There are a atoms in more optimal sites. As the poten-
number of definitions, mostly coming tial is driven from equilibrium (the formal
down to the growth of single crystals potential), deposition rates increase, and
on single-crystal surfaces. In the vast the exchange current drops, promoting
majority of cases, the lattice constants of three-dimensional (3D) growth, and gen-
the deposit do not exactly match those erally degrading epitaxy.
of the substrate, so strain will build, and The quality of an elemental deposit is a
defects are generated. The question then function of the rate at which it deposits,
arises: single-crystal grains of what size the surface diffusion, its exchange current,
need to be formed on a substrate to call it and the substrate quality and structure.
epitaxy? Electrodeposition of a compound thin
film requires all these things, as well as
4.2.1.1 Electrodeposition stoichiometry.
Electrodeposition has been used since the Several methods and variations have
beginning of the century to form high- been developed to electrodeposit com-
quality, mostly metallic, thin films: car pounds. Most of the work described in this
bumpers, decorative plating, metal con- article concerns the formation of nonox-
tacts, and so on. It has recently been shown ide compounds such as II to VI and III
that high-quality copper interconnects for to Vs. Oxides are probably the largest
ultra–large scale integration (ULSI) chips group of electrodeposited compounds (e.g.
aluminum anodization), but will not be referred to as UPD [34–39]. In the deposi-
discussed here. The electrodeposition of tion of one element on a second, frequently
II to VI compounds has been extensively the first element will form an atomic layer
studied and is well reviewed in a number of at a potential under, or prior to, that needed
articles [18–23]. The most prominent com- to deposit the element on itself. One way
pound electrodeposition methods include of looking at UPD is that a surface com-
codeposition, precipitation, and various pound, or alloy, is formed, and the shift
two-stage techniques. in potential results from the favorable free
The focus of the work described here energy of formation of the surface com-
is on understanding the mechanisms of pound.
compound electrodeposition and how to EC-ALE is the combination of UPD
control structure, morphology, and com- and ALE. Atomic layers of a compound’s
position. The primary tool for understand- component elements are deposited at
ing compound electrodeposition and for underpotentials in a cycle, to directly
improving control over the process has form a compound. It is generally a
been the methodology of EC-ALE [23–26]. more complex procedure than most of
Historically, EC-ALE has been developed the compound electrodeposition methods
by analogy with ALE [27–33]. ALE is a described earlier, requiring a cycle to form
methodology used initially to improve each ML of the compound. However,
epitaxy in the growth of thin films by it is layer-by-layer growth, avoiding 3D
MBE and VPE. The principle of ALE is nucleation, and offering increased degrees
to use surface-limited reactions to form of freedom, atomic level control, and
each atomic layer of a deposit. If no more promoting of epitaxy.
than an atomic layer is ever deposited, In addition, EC-ALE offers a way of better
the growth will be 2D, layer-by-layer, understanding compound electrodeposi-
and epitaxial. Surface-limited reactions tion, essentially a way of breaking it down
are developed for deposition of each into its component pieces. It allows com-
component element, and a cycle is formed pound electrodeposition to be deconvolved
with them. With each cycle, a compound into a series of individually controllable
ML is formed, and the deposit thickness is steps, resulting in an opportunity to learn
controlled by the number of cycles. more about the mechanisms, and gain
As noted, in techniques such as MBE a series of new control points for elec-
and VPE, surface-limited reactions are trodeposition. The main problem with
generally controlled by the temperatures compound electrodeposition by codeposi-
of the reactants and substrate. The tem- tion is that the only control points are the
perature is kept high enough so that solution composition and the deposition
deposition over a ML sublimes, leaving potential, or current density, in most cases.
only the atomic layer, forming the com- In an EC-ALE process, each reactant has its
pound. Problems are encountered when own solution and deposition potential, and
the temperatures needed to form atomic there are generally rinse solutions as well.
layers of different elements are not the Each solution can be separately optimized,
same, as changing the temperature be- so that the pH, electrolyte, and additives
tween layers is difficult. or complexing agents are tailored to fit
Surface-limited reactions are well known the precursor. On the other hand, code-
in electrochemistry, and are generally position uses only one solution, which is
a compromise, required to be compatible phase with potential control instead of

with all reactants. temperature. This increases the variable
Finally, electrodeposition is orthogonal space for producing materials, the diversity
to MBE and chemical vapor deposition of conditions under which a compound
(CVD), as it involves growth in a condensed can be formed.
Waste
Working, counter, and

reference electrode
connections to potentiostat
Deposition
cell
Valve Block
Computer
control
Deposition
solution
Pump
Fig. 1 Schematic diagram of automated flow electrodeposition

system, for forming thin films using EC-ALE.
4.2.2 the working electrode, allowing observa-

Hardware tion of the deposit during growth, as well
as providing an optimal current distribu-
Figure 1 is a schematic diagram of a bation.
sic electrochemical flow-deposition system Villegas and coworkers used a wall-jet
used for electrodepositing thin films using configuration, with very good results, in
EC-ALE, and Fig. 2 is a picture showing the the formation of CdTe deposits on Au [43].
solution reservoirs, pumps, valves, electro- The quality of their deposits appear equiv-
chemical cell, potentiostat, and computer. alent to those recently produced by this
A number of electrochemical cell de- group with the thin-layer flow cell (Fig. 3c).
signs have been tried. A larger thin-layer Figure 4 is a transmission electron mi-
electrochemical flow cell is now used crograph(TEM) of a CdTe deposit formed
(Fig. 3c) [40], with a deposition area of using their cell and 200 EC-ALE cycles [43].
about 2.5 cm2 and a cell volume of 0.1 mL, A wall-jet configuration has also been used
resulting in a two order of magnitude drop by Foresti and coworkers, in their studies
in solution volume, compared with the of thin-film growth using EC-ALE on Ag
H-cell (Fig. 3b). The cell includes an in- electrodes [44].
dium tin oxide (ITO) auxiliary electrode, The pumps used in the flow system in
as the opposite wall of the cell from Fig. 2 are standard peristaltic pumps (Cole
Fig. 2 Picture of one of the

flow-cell systems presently used
by this group.
Parmer). The main requirement for the coworkers have used pressurized bottles,
pumps is that they are clean. If smoother without pumps, to deliver solution [44, 45].
pumping is required, pulse dampening or In Fig. 2, there is one pump for each line,
syringe pumps could be used. Foresti and to push solution through the cell. Villegas
(a)
Fig. 3 Diagrams of electrochemical cells used in flow systems for
thin-film deposition by EC-ALE. (a) First small thin-layer flow cell
(modeled after electrochemical liquid chromatography detectors). A
gasket defined the area where the deposition was performed, and
solutions were pumped in and out through the top plate. (Reproduced
by permission from Ref. [41].) (b) H-cell design where the samples were
suspended in the solutions, and solutions were filled and drained from
below. (Reproduced by permission from Ref. [42].) (c) Larger thin-layer
flow cell. This is very similar to that shown in (a), except that the
deposition area is larger and laminar flow is easier to develop because
of the solution-inlet designs. In addition, the opposite wall of the cell is
a piece of ITO, used as the auxiliary electrode. It is transparent, so the
deposit can be monitored visually, and it provides excellent current
distribution. The reference electrode is incorporated into the cell as well.
(Adapted from Ref. [40].)
Sample holder Reference/

auxillary
Electrochemical compartment
cell
Sample
(b)
Plexiglass BAS reference

back
Substrate
Teflon nut
Gasket
Inlet Outlet
(c) ITO counter
Fig. 3 (Continued)
and coworkers used a single pump on the sufficient to prevent problems with oxy-
outlet to suck each solution into the cell, gen, as most tubing has some oxygen
in an elegant simplification [43]. permeability. To better avoid this problem,
There are a number of vendors that sell the solution delivery tubes were threaded
solenoid-actuated Teflon valves, which are through larger ID tubes and into the Plexi-
easily interfaced to a computer. Care must glas box (Fig. 2) that houses the pumps and
be taken to choose a design in which the valves. The sparging N2 was made to flow
internal volume at the valve outlet can be out of the solution reservoirs, through the
flushed easily between steps, however [41]. large ID tubes (around the outside of the
Rotary selection valves have been used as solution-delivery tubes) and into the box,
well, but given the number of rotations greatly decreasing oxygen exposure. The
needed for a 200-cycle deposit, various measured oxygen content of the N2 leav-
failure modes revealed themselves. ing the Plexiglas box was 10 to 30 ppm, as
As deposition of most of the relevant measured with a glove box oxygen analyzer
atomic layers involves reduction at rela- (Illinois Instruments, model 2550).
tively low potentials, oxygen has proven to The majority of deposits formed in this
be a major problem. It has been shown group have been on Au electrodes, as they
repeatedly that oxygen not rigorously ex- are robust, easy to clean, have a well-
cluded results in thinner deposits, if they characterized electrochemical behavior,
are formed at all. For this reason, exten- and reasonable quality films can be formed
sive sparging of the solution reservoirs is by a number of methodologies. However,
critical. Sparging alone is generally not Au it is not well lattice-matched to
5 nm
Fig. 4 Transmission electron micrograph (TEM) of a CdTe deposit

formed using 200 cycle of CdTe via EC-ALE. The regular layered
structure, parallel with the substrate Au lattice planes, suggest the
epitaxial nature of the deposit. (Reproduced by permission from
Ref. [43].)
most of the compounds being formed Au vapor deposited on mica at 300

by EC-ALE. to 400◦ is known to form large (111)
Some deposits have been formed on Au terraces [46–51]. Several attempts to use
single crystals. However, single crystals these films in the flow cell (Fig. 3c) resulted
are too expensive to use in forming in delamination, due to the constant
larger deposits, and have to be recycled. rinsing.
A number of disposable substrates have Presently, Au on glass is being used to
been investigated, so that the substrates do form most of the deposits in the flow cells.
not have to be recycled, and the deposits These substrates are microscope slides,
can be kept around. The first deposits were etched in HF, coated with 8 nm of Ti
made on cold-rolled gold foil, which proved and then 200 nm of Au, at about 300◦ .
too expensive, and polycrystalline after The deposits do not, in general, show
etching. Au vapor deposited on Si(100) as many large terraces as Au on mica;
at room temperature was used extensively, however, they are orders of magnitude
as the films were more reproducible and better than Au on Si (Fig. 5b). To improve
resembled Au mirrors. On the nanometer terrace sizes, the substrates are annealed
scale, however, the films consisted of in a tube furnace at 550◦ for 12 h with
40 nm hemispheres (Fig. 5a). flowing N2 . In addition, substrates are
200 300
200
100
100
0 0
0 100 200 0 100 200 300
nM nM
Z range: 15 nm Z range: 2 nm
Fig. 5 (a) AFM image of Au vapor deposited on Si(100) at room temperature and (b) AFM image of
Au on glass, annealed at 550 C for 12 h, in a tube furnace, in a flow of N2 .
given a brief flame anneal, in the dark used as a substrate. The problems in-
with a H2 flame to a dull orange glow, volve adequately preparing the substrate
prior to use. surfaces and understanding the electro-
Some Cu substrates have been used, chemistry of a compound semiconductor
including Cu foils, etched foils, and vapor- substrate. Work is progressing in this di-
deposited Cu on glass. There does not rection. Good quality deposits of CdSe
appear to be a significant difference in the have been formed on InP and GaAs
quality of deposits formed on Cu versus substrates using codeposition, by Mau-
Au, beyond that expected to result from rin and coworkers [55, 56]. They used
considerations of lattice matching. reflection high-energy electron diffraction
Single-crystal silver substrates have been (RHEED) and TEM to follow the habit of
used exclusively by Foresti and cowork- CdSe growth. Their work clearly shows
ers [44, 45, 52, 53]. They use macroscopic the applicability of high-quality com-
Ag single crystals, formed in-house. They mercial compoundsemiconductor wafers
have formed a number of II to IV com- as substrates for compound electro-
pounds using EC-ALE, including ZnSe, deposition.
CdS, and ZnS.
Semiconductors such as polycrystalline 4.2.3
ITO on glass have been used to form Deposition Programs
deposits of ZnS [54], CdS, and CdTe,
by this group, with no obvious prob- 4.2.3.1 Cycle Steps
lems. Ideally, lattice-matched semicon- As described in the introduction, UPD is
ductor substrates could also be used the formation of an atomic layer of one
to form deposit. For instance, InSb is element on a second element at a potential
lattice-matched with CdTe, and could be under that required to deposit the element
on itself: as Cl− , Br− , and I− oxidatively adsorb

on metal surfaces as halide atoms. With
M2+ + 2e− ⇐⇒ MUPD (1) respect to compound formation, oxidative
The formal potential is the potential at UPD from sulfide is a good example:
which the element begins to be stable
HS− ←−→ S(UPD) + H+ + 2e− (3)
depositing on itself:
◦ The other chalcogenides (H2 Te and H2 Se)
M(s) + M2+ + 2e− ⇐⇒ 2M(s) EM2+ /M
work as well; however, their aqueous
(2)
solutions tend to be unstable, relative to
The point of view in this group is that the
sulfide. Solutions of reduced forms of the
element is more stable because a surface
pnictides, such as H3 As, H3 Sb and H3 P,
compound is formed, so the free energy of
compound formation promotes deposition are of interest, and expected to work for
at an underpotential, accounting directly oxidative UPD of atomic layers. As these
for the potential shift. This can be seen species are generally gases, it is better to
in oxidative stripping of Zn chalcogenide use a basic solution to prevent out-gassing
MLs on Au electrodes. The Zn stripping from solution.
features shift to more positive potentials, The same equilibrium used to described
as the chalcogenide layer is changed from oxidative UPD (Eq. 3) can frequently
Te to Se to S [57]. The order and magni- be used to form atomic layers starting
tudes of the shifts are consistent with the with more stable, oxidized forms of the
differences in heats of formation for the elements, if a stripping step is added to
corresponding compounds, suggesting the the cycle. This approach has been used
relationship between their electrochemical extensively to form atomic layers of Te [24,
stability on the surface and the compounds 41, 42, 58].
formed. It would be nice to form CdTe using
Reductive UPD is the major atomic layer reductive UPD of both Te and Cd. If,
deposition process used in EC-ALE to form however, acidic Te solutions are used,
atomic layers (Eq. 1). Many metals can reductive Te UPD requires a potential
be obtained in a soluble oxidized form, near 0.0 V (Fig. 6b) in order to avoid bulk
from which atomic layers can be deposited deposition. Cd UPD is optimal between
at underpotentials. Control points are the −0.4 and −0.6 V. The problem is then
concentration of the reactant, the potential that the Cd atomic layers strip during the
used for the deposition, the pH, and the Te deposition step. On the other hand,
presence of other additives or complexing Te can be deposited at −0.5 V, where the
agents, which can change the activity of Cd remains stable, but some bulk Te is
the reacting species. formed along with the Te(UPD) :
Oxidative UPD involves the oxidation
of species to form an atomic layer where 2HTeO2 + + 6H+ + 8e− ←−→
the precursor contains the element in a Teo + Te(UPD) + 4H2 O (4)
negative oxidation state. A classic example
is the formation of oxide layers on Pt and The trick is that bulk Te (Teo ) can be
Au, where water is oxidized to form atomic stripped at a more negative potential in
layers of oxygen. Halide adsorption can be supporting electrolyte, a blank solution. In
thought of similarly, where species such that way, only Te(UPD) is left, owing to
20.0 5 mV s−1 Alloy stripping

Bulk Cd
stripping UPD stripping
Current
0.0
[µA]
Cd UPD
−20.0
Alloy dep
−0.80 −0.40 0.00 0.40

Potential vs Ag/AgCl
(a) [V]
(c) (b)
3′ Te
Te pH 2.2
pH 9.2
1′ 1′
80 3′
2′
Current
2′
[µA]
0
1 0
2 2
1
3
−80
3
−0.80 −0.40 0.00 0.40

[V]
Fig. 6 Voltammetry showing examples of irreversible behavior in the formation of atomic
layers on Au. (a) Cd, (b) Te, pH 2, (c) Te, pH 10, (d) Se, (Adapted from Ref. [59],) (e) As.
Bulk Se removal
Se UPD removal
10 µA
Further bulk Se Se UPD
Limited bulk Se
0 200 400 600 800

Potential vs Ag/AgCl (1-M NaCl)
(d) [mV]
20 µA
−0.4 −0.2 0.0 0.2

Potential
(e) [V]
Fig. 6 (Continued)
stabilization by bonding with previously Equations (4) and (5) together have essen-
deposited Cd, forming CdTe: tially the same result as Eq. (3), oxidative
UPD or formation of a Te atomic layer,
Teo + Te(UPD) + 2H+ + 2e− ←−→
along with conversion of some tellurite
Te(UPD) + H2 Te (5) to telluride. The advantage is that F and G
can be accomplished using the more stable Reactants can also be generated electro-
tellurite solutions, instead of the unstable chemically, prior to pumping the solution
telluride. into the cell, or generator electrodes can be
incorporated into the cell, so a precursor
could be generated on the opposite wall of
4.2.3.2 Cycles
the cell, and allowed to diffuse across the
A cycle consists of the steps needed to
thin solution layer to deposit. This method
form a ML of the compound. Figure 7 is a
would have advantages in that species with
cycle diagram for CdTe and describes the
limited stability could be used, as they need
deposition, stripping, and rinsing steps, as
only last long enough to diffuse across, and
well as times and potentials.
only enough would be generated to form
In principle, each solution can be inde-
the atomic layer. This has been suggested
pendent, composed of different solvents, as a way of obtaining H3 As, H2 Te, and so
reactants, electrolytes, buffers, and addi- on as precursors for oxidative UPD.
tives. In general, aqueous solutions have For some elements, there are a variety
been used, as their electrochemical be- of possible precursors, sulfur for instance
havior is better understood, and ultrapure (Fig. 8). Metal-organic precursors, such as
water can be obtained by a number of used in metallo organic molecular beam
methods. epitaxy (MOMBE) or metallo organic va-
The choice of reactant has to do with por phase epitaxy (MOVPE), are possible
whether you want to perform oxidative or precursors that can be used, if they can be
reductive UPD, oxidation state, solubility, made soluble in water or if a nonaqueous
availability, purity, and price. In general, solvent were used. Mixed aqueous-organic
salts are used, simplifying solution prepa- solvents could improve solubility. Overall,
ration. Gases can be used by saturating the there would be an increased probability of
solution, prior to pumping it into the cell. carbon contamination. One of the benefits
.
cs .
se cs
2 se
V, 2
.7
0 V,
−0.8 0
−0 0.
7
@ − .
fill @ cs cs
.
n e se se
tio ns 2
Potential (vs Ag/AgCl)
lu ri V, 2
so nk 60 V,
−0.7 a 0. 0
Te Bl − .6
@ −0
fill @
n se
Te deposition utio rin
20 secs. @ −0.70 V s ol nk
−0.6 a
quiescent Cd Bl
Cd deposition
20 secs. @ −0.60 V
quiescent
−0.5
0 10 20 30 40 50 60
Time
[s]
Fig. 7 Diagram of an EC-ALE cycle for the formation of CdTe. (Adapted from Ref. [60].)
2.5
2
Coverage
[ML]
Sulfide
1.5
Thiosulfate
1
0.5
0
−1.5 −1 −0.5 0 0.5
[V]
Fig. 8 Graphs of the Auger signals resulting from exposure to various
sulfur-containing precursors as a function of potentials. (Adapted from Ref. [61].)
of using inorganic salts in aqueous solu- concentrations, although electrolyte con-

tions is that the number of constituents centrations of 50 mM have been tried and
is limited, and thus the possible sources look promising. Currents can be kept very
of contamination. The purities of cor- low, given that no more than an atomic
responding inorganic salts are generally layer is deposited at a given time. Draw-
better than metal-organic precursors and backs to high electrolyte concentrations are
cheaper. cost, waste, and contamination. It is antici-
The concentration of reactants can pated that when systematic studies of trace
be kept very low, as only a nanomole contaminants are performed on deposits,
cm−2 of material deposits in a given many will be found to result from the elec-
step. Use of low concentrations conserves trolytes, pointing out the need for such
reactants, simplifies waste handling, and studies.
minimizes the concentrations of reactant Acid-based electrolytes are desirable, as
salt-born contaminants. In studies of InSb the mobility of the proton allows use of
deposition, 0.02-mM solutions of the Sb more dilute solutions. However, they do
precursor have been used [62]. promote hydrogen evolution, which can
The electrolyte is usually the major com- form bubbles in the cell. Good results
ponent, besides solvent, suggesting the with both sulfate and perchlorate salts have
use of the highest purity feasible. Fil- been seen. Recent results with chloride in
tering of solutions is advised to remove the deposition of CdTe are encouraging as
any particulates. The concentration deter- well [63].
mines the conductivity of the solutions, Solution pH is an important variable, as
and the conductivity needed is a function it controls solubility, deposition potential,
of the cell design, currents expected, and and precursor speciation [64]. As noted,
range of deposition potentials to be used. one of the advantages of using an ALE
At present, most studies in this group process is that very different solutions can
have involved 0.1- to 0.5-M electrolyte be used for each step in the cycle. For
instance, pH 4 Cd solutions have been The amount of rinsing needed depends

used with pH 10 Te solutions to form on factors such as cell design and the
CdTe. However, some care must be taken amount of a previous reactant that can be
while rinsing between solutions. On the tolerated in a subsequent cycle step. Ide-
other hand, Foresti and coworkers have ally, the cell will have good laminar flow
used the same pH for both the metal and and will rinse out easily. As noted previ-
chalcogenide in the formation of II to VI ously, there are multiple reasons to limit
compounds, such as CdS and CdSe on Ag the concentrations of reactants. Similar
single crystals [44, 45, 52, 65], in order to points can be made concerning buffers as
simplify solution exchange. To keep the they control pH. It is very important to re-
Cd from precipitating in a basic solution, move buffers between steps in a cycle. In
it was complexed with pyrophosphate. the deposition of CdTe, for instance, pH 4
The drawback to change the pH in each Cd solutions were used along with pH 10
step is evident in the formation of CdTe. Te solutions. As can be seen in Figs. 6b
After Cd atomic layer formation, the Cd2+ and 6c, the potential at which bulk Te de-
solution is removed, and replaced with a position occurs shifts from 0.0 V, at pH 2,
basic TeO3 2− solution. Cd deposition is to −0.7 V, at pH 10. Problems were en-
a relatively reversible process (Fig. 6a), so countered when insufficient rinsing was
used between the pH 4 Cd solution and
there are two problems: first, the Cd2+
the pH 10 Te solution. Some Te was de-
activity is lost and second, the pH is
positing from solution with a pH close to
shifted from 4 to 10. First, the simple
6. Under those conditions, some bulk Te
loss of Cd2+ activity shifts the equilibrium
was deposited each cycle, instead of just
and can cause stripping of some of the
an atomic layer, promoting 3D growth and
atomic layer. In addition, shifting the pH
rough deposits.
to 10 has a similar result, as the Cd2+ /Cd
To get an idea of the amount of
equilibrium shifts negatively with pH.
rinsing required, an electroactive species
Thus maintaining the same potential and
can be used, where the amount left
increasing the pH results in Cd formed at in the cell after a given amount of
pH 4 stripping from the surface. rinsing can be determined coulometrically.
The history of electrodeposition involves Such measurements can be misleading if
control of deposit structure and morphol- reactant is trapped under the gasket at
ogy using additives. These can be things as the sides of the deposit (Fig. 3c). This
simple as chloride, sulfides, and glycerol, can be avoided by using a resist to coat
or complex organic compounds, or traces the substrate under the gasket, preventing
of other metals. In EC-ALE, it is desirable electrochemical reaction of species trapped
to control the structure and morphology of under the gasket.
deposits using surface-limited reactions, Insufficient rinsing can also result in
not additives. Some of these species may some codeposition if the previous reactant
be beneficial to an EC-ALE cycle, but very is not fully removed. The main drawback
little work has been done along this line. is the possibility of 3D growth, which can
Most additive work has involved complex- be hard to identify with very thin deposits.
ing agents, to shift the potential used to Alternatively, the rinse solution may not
form an atomic layer, pH or Cl− being be important. Some high-quality CdTe
good examples. films were formed in this group without
using a rinse solution at all. That is, the be used. In general, the same elec-
reactant solutions were exchanged by each trolyte concentrations have been used
other, underpotential control, suggesting for rinsing and deposition solutions
that some small amount of codeposition so far.
probably did occur. The starting potentials for most atomic
The amount of electrolyte needed in a layers, in this group, were obtained by
rinse solution depends on current flow studies of the voltammetry for an element
during the rinse. Rinsing can be per- on an Au electrode [57, 66–68], usually
formed at open circuit in some cases, using a thin-layer electrochemical cell
so that no electrolyte is needed. If the (TLEC) (Fig. 9) [69, 70]. UPD potentials
amount of current during the rinse on Au are not expected to be optimal for
under controlled potential is very low, growth of a compound; however, they are
low-electrolyte concentrations could again generally a good start.
Solution
gap
0.001′
Electrode
+
+ +
+
+
+
+
+ + + +
+ + + + Pin hole
+ + + solution
+ + + Thin layer
+ + in and out
+ electrode
10 mm ID
teflon
thermometer
adaptor
1 mm capillary
tubing
24 / 40
joint
Male
ER Varibor Reference
teflon electrode
stopcock compartment
2 mm
10 mm
fine
frit
Fig. 9 Schematic diagram of a thin-layer electrochemical cell (TLEC).

The simplest model for EC-ALE is that a Reasonable quality deposits have been
set of conditions is chosen for a cycle, and obtained for 200-cycle runs, using these
each cycle produces one compound ML. more negative ‘‘steady state’’ potentials
Ideally, the same potentials and solutions from the beginning. However, that more
are used for each cycle of a deposition. than an ML/cycle is formed during the
Recently, it has become clear that this is first few cycles is a significant problem, as
not the optimum case for EC-ALE growth it suggests bulk deposition, non–layer-by-
of many compounds on Au, as the initial layer, or 3D growth. That high growth rates
conditions do not appear optimal from were observed initially is understandable,
start to finish. in that the potentials were close to bulk
Graphs of the charges for Cd and Te growth conditions for Cd and Te.
versus the cycle # (Fig. 10) show that if a Using these more negative potentials
single set of potentials is used for each did result in reasonable deposits. Overall
cycle, the first cycles are much larger they were stoichiometric, displayed the
than the succeeding cycles. If potentials expected X-ray diffraction (XRD) patterns
determined from using a TLEC are used, without annealing, and the graphs of
the first cycle looks good, the charges growth versus the number of cycles
correspond to the formation of an ML were linear (after the first few). Deposit
of the compound, but the charges quickly morphologies appeared reasonable but
die away to nothing. On the other hand, probably not optimal, given some 3D
if potentials are used that will result in growth during the first few cycles.
something close to an ML of the compound It has become clear that the potentials
for each cycle once steady state has been needed to form atomic layers shift neg-
reached, the first few cycles will result in a atively as the semiconductor films grow,
great excess of deposition (Fig. 10). especially over the first 25 cycles. The
4.00 To
14.5 ML
3.50 Monolayers Te deposited
Monolayers deposited
Monolayers Cd deposited
3.00 To
9.5 ML
2.50
2.00
1.50
1.00
0.50
0.00
0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 6.5 7
Cycle number
Fig. 10 Bar graphs showing the charges, in terms of monolayers, where 1 ML would
be optimum, as a function of the cycle, for the first 7 cycles of CdTe deposition. The
first half cycle was Te deposition, and the second Cd. A single set of potentials was
used for all the cycles, and during the first cycle, the amounts deposited were excessive.
(Adapted from Ref. [60].)
most logical reason for this shift is that of deposition agreed closely with the de-
a junction potential is developing between posited amounts, suggesting that proton
the Au substrate and the depositing com- and oxygen reduction were not significant.
pound semiconductor. When a program Finally, no elemental Cd and Te were ob-
is used where the potentials are initially served in XRD patterns; only diffraction
shifted for each cycle, steady state poten- peaks for CdTe and the substrate, for 200-
tials are generally achieved after about 25 cycle deposits, were observed.
cycles, which can be maintained through One explanation for the observed charge
the rest of the deposit without the de- disparity is that excess Cd is being
posited amounts changing for each cycle. deposited, and when the TeO3 2− is
Similar procedures were used by Ville- introduced, some Te is exchanged for Cd.
gas and coworkers [43] in the formation The TeO3 2− may react chemically with the
of CdTe. A reasonable potential ramp can cadmium, or the current observed may be
be determined by inspection, and the po- the sum of that for oxidation of part of the
tential changes programmed through the previously deposited excess Cd and that for
first 25 cycles or so. Some form of feed- reductive deposition of the Te atomic layer.
back within the cycle would be better, to In general, the currents observed during a
regulate deposition on the fly. The most cycle in the formation of a compound have
not been easy to understand, and they
obvious feedback would be the deposition
are presently a major topic of study. For
currents (or charges).
instance, in the formation of CdSe, a cycle
Figure 10 displays the coverages for Cd
similar to that for CdTe is used; however,
and Te deposited in each of the first few
it is the charge during Se deposition that is
CdTe cycles, using a program in which
in excess and that for Cd, which is almost
the potentials remained constant. The cov-
nonexistent. Again, the deposit is nearly
erages are the result of coulometry, two
stoichiometric, within a percent of 1 : 1,
electrons for Cd and four electrons for Te.
using EPMA. In summary, the relative
Note the excessive deposition for the first charges for Cd and Te deposition are
few cycles, significantly more than a ML, not easily interpreted in terms of deposit
and the disparity between Cd and Te cover- stoichiometry. They are, however, a key to
ages. The relative coverages for Cd and Te incorporation of feedback in the EC-ALE
should be 1 : 1 in Fig. 10, but seldom are cycle.
in such graphs. It is not yet clear why the Ideally, each component element can be
stoichiometry is not 1 : 1 from coulometry, deposited at a potential optimized sepa-
while the stoichiometries from EPMA for rately, independent of the conditions used
the resulting deposits are 1 : 1. Some pos- to deposit subsequent atomic layers in the
sible reasons for this discrepancy (Fig. 10) cycle. In some cases this independence
that have been considered include solvent is nearly realized, as in the deposition of
or proton reduction, oxygen reduction, and Te atomic layers (Fig. 11) discussed pre-
charging currents. However, blank solu- viously. A broad potential region (0.7 V)
tions provide a gauge of proton, solvent, is evident in Fig. 11, where the thick-
and oxygen reduction, as well as charg- nesses of 200-cycle CdTe deposits do not
ing currents, and they do not account change appreciably as the Te deposition
directly for the asymmetry. In addition, potential is adjusted. However, Te depo-
coulometric stripping after 5 to 10 cycles sition is a relatively irreversible process
240
220
210
200
180
160
Thickness
140
[nm]
120
100
80 78
68 70 68
60 62 58 57
40
20
0 2 3
−0.9 −0.8 −0.7 −0.6 −0.5 −0.4 −0.3 −0.2 −0.1 0 0.1 0.2
Te deposition potential (vs Ag/AgCl)
Fig. 11 Graph of CdTe deposit thickness as a function of the potential used to deposit Te. Each
point represents a deposit formed with 200 cycles. (Adapted from Ref. [60].)
at pH 9 (Fig. 6c), suggesting that once it formed by stripping should still be close
is deposited, the potential can be shifted to the deposit surface, in the barrier layer,
significantly more positively (Fig. 6c) with- where they may redeposit as the second
out its stripping from the surface. On the element deposits. On the other hand, given
other hand, a more reversible species, such that the second element is reversible, the
as Cd, may strip from the surface if the potential can be stepped positively after its
potential is shifted positively in a subse- introduction to the cell. Even if some bulk
quent cycle step. This is exaggerated after deposit is formed at the more negative
a rinse, where the activity for Cd2+ has potential, it should strip reversibly when
been removed. the potential is stepped positively.
In a system where both elements The inverse process, in which the
deposit reversibly, at significantly different potential is shifted negatively, is not
potentials, the rinsing procedures must generally a problem. The potential cannot
be carefully thought through. If, for be shifted prior to rinsing, or else bulk
instance, a reversible species is reductively deposition of the first element would occur
deposited, and a more positive potential before the solution is rinsed from the cell.
is to be used to deposit the next element, It is best to shift the potential after the
the rinse step can be performed at the blank rinse, just prior to introduction of
more negative potential, and then shifted the second element. In fact, the potential
positively upon introduction of the second can be shifted after introduction of the
reactant, hopefully avoiding material loss second element, as it should not deposit at
during rinsing and minimizing losses the more positive potential, but must wait
during introduction. If the second element until the potential is shifted negatively.
is pumped in rapidly, and with a minimum Keeping the potentials close can be bene-
of solution, ions of the first element ficial, using complexing agents or the pH.
The ideal process is represented by the workable. As noted earlier, deposition

formation of CdS, where the Cd is first times are controlled by the concentration
deposited from a Cd2+ solution by reduc- of reactants and the kinetics of the
tive UPD, and then the potential is shifted deposition process. High concentrations
negatively to where S(UPD) is formed from result in increased atomic layer formation
a HS− solution by oxidative UPD. In this rates; however, they are wasteful, and
way, both elements are stable on the sur- may be difficult to achieve in some
face during the deposition of the other. cases. TeO2 solutions, for instance, are
What is the optimal time for a given difficult to make, as are mM solutions
step in a cycle? What is the minimum at most pHs because of poor solubility.
time needed to form a deposit? How High concentration solutions also take
much time is needed for deposition of longer exchange in the cell. In general,
an atomic layer and how long does it 20-s deposition times are presently used,
take to change solutions? Each step should probably much more than enough, with a
be independently optimized. Pumps can couple of seconds for pumping (Fig. 7).
easily exchange solution within a fraction Of course, deposition times can be
of a second. The time required for decreased by using a larger driving force,
deposition of an atomic layer will depend but that runs the risk of bulk deposition.
on the concentration of the solution and It is easy to envision a cycle in which
the kinetics for deposition. The deposition overpotentials are used, and the deposition
current should probably be an exponential is simply stopped after a ML of charge has
decay, raising the question of at what passed. Such a cycle would not involve
point do you cut it off. This is related surface-limited reactions and 3D growth
to fundamental questions such as: If you would be expected.
do not complete an ML, will defects result? Faster pumps result in a shorter cycle.
There are indications that it is better to Exceeding the Reynolds number, however,
deposit less than an ML than too much, results in turbulent flow. Flow patterns
as too much will produce 3D growth. It are frequently seen in deposits where the
is anticipated that an incomplete deposit gasket has not been set correctly. Anything
will work fine, that cutting off the deposit that perturbs laminar flow in the cell affects
sooner than later will probably not be a the deposit homogeneity: bubbles are a
problem. The Foresti group has repeatedly good example. If bubbles are present, even
developed cycles that produced only a temporarily, they frequently leave their
small fraction of an ML/cycle [44, 45, 52, shadow in the deposit, as areas where less
53, 65]. It is probably a good idea to allow deposit was formed. If the gasket is not
a given atomic layer deposition process to smooth, or something gets in the way
proceed as close to completion as possible, of part of the inlet or outlet, the deposit
however. quickly shows a flow pattern. Detailed
With the 0.2 mM reactant solutions studies have not yet been performed
used in this group, deposition times as exploring the relationship between flow
short as a couple of seconds have been rates, cell design, and flow patterns, but
used with little problem, resulting in it is clear from experience that the more
cycle times close to 10 s. With slightly laminar the flow the more homogeneous
higher concentrations, and fast pumps, the deposit. CdTe, for example, is sensitive
cycle times of a few seconds appear to flow rate, and a study of the dependence
on flow rate is underway. Most of the include the chalcogenides: S, Se, and Te;
compounds, described in the next section, the pnictides: As and Sb; the group III
form homogeneously in the deposit area. metals: Ga and In; the group II metals:
CdTe, on the other hand, does not deposit Zn, Cd, and Hg; as well as Cu and Co. The
well at the inlet and outlet, suggesting range of compounds accessible by EC-ALE
turbulence may affect the deposit. Slower is not clear, although the majority of work
pumps appear to help, but a detailed study has been performed on II to VI compounds
is needed to determine what step in the (Table 1). The III to V compounds InAs
cycle is causing the problem, and why. and InSb have recently been formed [40,
Besides the flow rate, rinse times are 62]. In addition, Shannon and coworkers
very important. As discussed previously, have begun studies of CoSb3 [73] with the
if rinsing is insufficient to achieve the intent of forming thermoelectric materials.
desired pH, deleterious effects can result. Initially, EC-ALE was developed on the
The example was given that if Te solutions principle that reductive UPD of a metal and
are too acidic, bulk Te will form. oxidative UPD of a main group element
Generally, this group has used stop flow were required to form a working cycle. This
to provide adequate deposition times and would then limit compounds that could be
to limit the volumes of solution needed. formed to those containing a chalcogenide
Villegas and coworkers have used con- or a pnictide, as reduced forms of some
tinuous pumping, with good results [43], of these elements were reasonably stable
which may be an important alternative. in aqueous solutions. Recently, it has
Which element should be deposited first been shown that reductive UPD of both
is an ongoing question. In the forma- elements is possible [40], suggesting a
tion of II to VI compounds by this group, much larger range of compounds or
the chalcogenides have generally been de- possibly alloys might be formed.
posited first. The reasons for this have Table 1 is a listing of compounds formed
more to do with the rich surface chemistry using EC-ALE. The first EC-ALE studies
of those atomic layers than a feeling that focused on CdTe [24, 25] for the historical
better deposits will result. Results from reason that its electrodeposition had been
two groups using Raman spectroscopy, studied the most [98–109]. The genesis
Shannon and coworkers [71] and Weaver for most of that work has been the desire
and coworkers [72], have suggested that to electrodeposit CdTe photovoltaics. The
higher-quality deposits of CdS are formed majority of the CdTe work is codeposition,
when the Cd is deposited first. Recent re- after the classic paper by Kroger and
sults by this group [63] indicate that no coworkers [99].
mater which is deposited first, Cd ends The first studies of EC-ALE were stimu-
up being the atomic layer next to the Au lated by Mike Norton [66], who suggested
substrate, suggesting it is probably better that an electrochemical form of ALE might
to deposit the Cd first. be possible, which prompted studies of
the UPD of Cd and Te on Cu, Au, and
4.2.4 Pt substrates, by this group. Those studies
Compound Formation involved the use of a TLEC [69, 70] (Fig. 9),
which consisted of a polycrystalline rod
At present, the elements used in the of the substrate metal, inserted into a
formation of compounds by EC-ALE vacuum-shrunk glass compartment. The
Tab. 1 EC-ALE studies
Compound Study Year References
ZnTe TLEC 1996 57

ZnSe TLEC 1996 57
ZnSe Flow cell deposition 1999 44
ZnS TLEC 1996 57
ZnS Flow cell deposition 1997 54
ZnS TLEC 1996 57
ZnS STM studies of monolayers 1999 74
ZnS Growth of superlattice with EC-ALE 1999 75
ZnS Size-quantized film, photoelectrochem. 2000 76
CdTe UHV-EC of first few monolayers 1992 77

CdTe UHV-EC of first few monolayers 1993 78
CdTe STM 1993 79
CdTe STM and UHV-EC 1995 80
CdTe Flow cell 1995 41
CdTe Review 1995 81
CdTe Flow cells, H-cell 1998 42, 58
CdTe UHV-EC, in situ STM 1998 82
CdTe Wall-jet flow cell growth of thin films 1999 43
CdSe STM 1996 83

CdSe STM, UHV-EC 1997 84
CdSe Flow cells, H-cell 1998 42
CdSe SERS 1999 85
CdSe Flow cells, large thin-layer cell 2000 86
CdS STM 1994 87

CdS TLEC 1994 67
CdS Voltammetry 1994 88
CdS STM 1996 89
CdS STM, RRDE 1996 90
CdS STM 1996 89
CdS STM, surface study 1997 91
CdS Flow cells, H-cell 1998 42
CdS STM, voltammetry 1998 52
CdS Raman study 1998 92
CdS Photoelectrochemical studies of films formed by EC-ALE 1998 93
CdS Resonance Raman study 1999 71
CdS Photoluminescence study of a heterojunction 1999 94
CdS Flow cells 1999 45
CdS Growth of superlattice with EC-ALE 1999 75
HgS Photoluminescence study of a heterojunction 1999 94

Tab. 1 (continued)
Compound Study Year References
GaAs UHV-EC 1992 68, 95

InAs Thin films with TLEC flow system 1999 40
InSb Thin films with TLEC flow system 2001 62
CuInSe2 TLEC 1996 96
In2 Se3 Flow cells 2001 97
CoSb Overview of EC-ALE of CoSb 1999 73
CdS/HgS Photoluminescence study of a heterojunction 1999 94
CdS/ZnS Growth of superlattice with EC-ALE 1999 75
CdS/CdSe Growth of superlattice with EC-ALE, studied by SERS 1999 85
InAs/InSb Superlattice 2001 62
Note: STM: scanning tunneling microscopy; EC-ALE: electrochemical atomic layer epitaxy; UHV-EC:
ultrahigh vacuum electrochemistry; SERS: surface-enhanced Raman scattering; RRDE: rotating
ring-disk electrode.
compartment was designed to hold the the deposit to air. The TLEC is used to
electrode in the center of a roughly cylin- determine the potentials needed to form
drical glass enclosure, about 25 µm away atomic layers of the elements on the sub-
from the glass. Two pinholes were ground strate.
into the bottom of the cell, to allow so- Following the codeposition work of
lutions to flow in and out, and for ionic Kroger and coworkers [99] and many
conductivity. The total volume of the cell others, acidic solutions of HTeO2 + were
was 3.0 µL, with a 1-cm2 electrode area. initially used [24, 25]. At that time, it was
The TLEC provided a defined envi- assumed that if Cd was deposited by
ronment for studies of surface-limited reductive UPD, Te should be deposited
reactions. The large surface area to volume using oxidative UPD, from a solution
ratio greatly limits background reactions of H2 Te or a related telluride species.
from traces of oxygen and other contami- It was not felt that reductive UPD
nants. Limited and predictable background of both elements could be performed.
currents facilitate coulometry and allow Solutions of telluride proved unstable,
for more detailed coverage measurements. however, oxidizing and turning purple
The reactant amounts are also limited, with flecks of elemental Te, with even
allowing studies with only slightly more traces of oxygen in solution. This led
than that needed to form atomic layers, to the use of the process described
limiting bulk-deposition and preventing by Eqs. (4 and 5) [24, 25], where more
bulk-stripping currents from swamping than an atomic layer of Te was first
out UPD stripping. In addition, the TLEC deposited, from a HTeO2 + solution, and
facilitated solution exchange, which is very the excess Te was stripped reductively
important for studies of EC-ALE. One solu- at a more negative potential using a
tion can be flushed out and another rinsed blank solution in a separate step. The
in within a few seconds, without exposing only analysis performed on those initial
deposits was coulometry, as the substrates the bubbles floated to the top. Sub-
were not easily removed from the Pyrex strates used in that study were Au on
cells (Fig. 9), and the substrate’s cylindrical Si(100), which looked like a mirror but
geometry did not lend itself to analysis by were composed of 40 nm hemispheres
many techniques. (Fig. 5a). The importance of oxygen ex-
A flow-deposition system was subse- clusion was discovered while draining
quently developed to grow thicker films solutions, as some oxygen was sucked
on flat disposable substrates [41]. The into the cell, unless an N2 gas blanket
first edition was a small thin-layer flow was kept above the solution. This helped
cell (Fig. 3a). The majority of the sys- account for one of the major problems
tem was very similar to that used to- early on, reproducibility. Deposits were
day: pumps, electrochemical cell, valves, formed one day, but not the next, using
solution reservoirs, and potentiostat, all the same conditions. Progress was exceed-
controlled by computer. Those first de- ingly slow until the oxygen problem was
posits were excessively heterogeneous, better understood.
however [41]. The deposits suffered from a The deposit quality was greatly im-
lack of adequate oxygen exclusion, gasket proved with the H-cell [42, 58], and Fig. 12
effects, and bubble problems. In addition, shows XRD patterns for deposits of CdTe,
CdSe, and CdS made using the system.
those deposits were grown with a sin-
In each case, the deposits have a [110] pre-
gle set of unoptimized deposition poten-
ferred orientation, and crystallize in the
tials. Scanning electron microscopy (SEM)
zinc blende (cubic) structure. Graphs of
images revealed extensive 3D growth.
deposit thickness versus the number of
Interestingly, many deposits were com-
cycles were linear, a good indication of a
posed of smooth bumps, suggesting they
surface-limited growth. The best deposits
might result from layer-by-layer growth
of CdTe were a little thin, however, cor-
on a series of nucleation sites. The de-
responding to the growth of only about
posits were stoichiometric by EPMA, but 0.4 ML/cycle. In hindsight, oxygen exclu-
graphs of coverage versus cycle num- sion was probably still a problem, the
ber were not linear, as expected, but samples were being overrinsed each cycle,
increased exponentially, clearly demon- and a constant set of deposition potentials
strating increasing surface roughness or were used.
3D growth. As noted, the most important advance
There were a number of excuses made, resulting from changing to the H-cells
and a number of solutions suggested was that CdTe deposits were more re-
for those first deposits, such as prob- producible. Studies of the dependence of
lems with valves, with gaskets, and with deposit thickness on various cycle variables
substrates [41]. The next set of CdTe thin- were possible using the H-cell design [58],
film growth studies were made using a including the dependency on the poten-
H-cell (Fig. 3b) [42, 58]. The substrates tials used for Te deposition, Te stripping,
were simply suspended in the H-cells, and Cd deposition. Plateaus in coverage
and the solutions were pumped in from as a function of those potentials were
the bottom and then drained out. This observed, confirming the surface-limited
eliminated problems with gaskets and nature of the growth process [58]. That
bubbles, as there were no gaskets, and is, the deposit thickness was zero if the
Omega optimized
CdTe
(a)
Relative intensity
CdSe
(b)
CdS
(c)
5 25 45 65
2θ
Fig. 12 XRD patterns for (a) CdTe, (b) CdSe, (c) CdS. (Reproduced with
permission from Ref. [42].)
Cd deposition potential was too far pos- inside an N2 -purged Plexiglas box, greatly
itive, and if the potential was too far improving oxygen exclusion, and an
negative, some bulk Cd was formed for oxygen analyzer (Illinois Instruments,
each cycle, resulting in a thick rough de- model 2550) was used to measure the oxy-
posit. In between was a range of potentials gen levels in the box, 10 to 30 ppm. Use
where the deposit thickness was constant, of the thin-layer flow cell dropped the so-
close to 0.4 ML/cycle, suggesting the de- lution volumes used from 100 mL/cycle
position was controlled by surface-limited to about 1 mL/cycle, and allowed potential
reactions. control to be maintained throughout the
There were some major problems with deposition.
the H-cells, however, such as the need At that time, the cycle program was also
for 50-gallon drums to hold the re- changed, so that cathodic UPD was used
sulting chemical waste, as 100 mL of for deposition of both elements. Use of a
solution was used for each cycle. In pH 10 tellurite solution, TeO3 2− , instead
addition, potential control was lost for of pH 2, shifted the Te UPD potential to
the deposits each time the solution was better coincide with that for Cd UPD. A
drained, which can result in some de- program similar to that presently used
posit loss. for depositing CdTe is shown in Fig. 7.
Subsequently, the H-cell was replaced A comparison of the programs in Fig. 13
with a larger thin-layer flow cell (Fig. 3c), clearly shows the relative simplicity of the
with a 1 × 3 cm deposition area. The new program and hardware, compared to
pumps, valves, and cell were placed the older.
Cd-16 = 16 ml of Cd solution
Cd, B-16 = 16 ml of Cd solution
Cd, B-24 = 24 ml of Cd solution
−1.20 Te-16 = 16 ml of Te solution
Te Te, B-24 Te, B-24 = 24 ml of Te blank solution
strip O. C. O. C. = Open circuit
Potential
−1.00
Te, B-24 Te-16 Te Te, B-24
[V]
O. C. O. C. dep O. C.
−0.80
Cd
Cd, B-16 Cd-16 Cd, B-24
dep
−0.60
1 2 3 4 5
(a) min
B-1 = 1 ml of Cd blank
Cd-1 = 1 ml of Cd solution
Te-1 = 1 ml of Te solution
−0.75
Potential
Te deposition B-1
[V]
−0.70
Cd-1
Te-1
Te-1
B-1
B-1
Cd deposition
−0.65
30 60
(b) s
Fig. 13 Diagrams of the cycle program for use with (a) the H-cell (Fig. 3b) and (b) the large
thin-layer flow cell (Fig. 3c).
As noted above, the deposits made with Figure 11 is a study of the potential
the H-cell design were thin, only about dependence of deposit thickness, using
0.4 ML/cycle. For a 200-cycle deposit, 200-cycle CdTe deposits, formed with the
they appeared deep blue, the result of large thin-layer flow cell, as a function of
interference effects in the 30-nm-thick the Te deposition potential. A 0.7 V plateau
films. With the new cycle program and in deposit thickness suggests broad flexi-
large thin-layer cell, the best deposits bility in the choice of the Te deposition
appeared gold in color, and ellipsometric potential. Those results were very encour-
measurement indicated they were very aging. At positive potentials, the coverage
close to 1 ML/cycle. A study of the dropped dramatically, as no Te was de-
thickness as a function of the number posited, so there was nothing for Cd to
of cycles is shown in Fig. 14, where the deposit on. At potentials below −0.7 V,
line is straight, indicating surface-limited some bulk Te appears to have deposited
control of growth. Figure 15 is a graph of for each cycle, as the thickness greatly ex-
the square of absorptivity X photon energy ceeded that expected for ML/cycle growth,
versus energy, from which a band gap of and the morphology became ‘‘sandy’’
1.55 eV has been extrapolated, equivalent (Fig. 16). Sandy is a term used in this group
to literature values for CdTe. to describe deposits where 3D growth is
(Cd −0.700 V, Te −0.700 V, open circuit fill, no blank rinse)
200 188
180
160
140
Film thickness
120
[nm]
100
80
72.6
60
40
35.5
20 14.4
0
0 50 100 150 200 250 300 350 400 450 500
Deposition cycles
Fig. 14 Thickness of CdTe deposits formed with the thin-layer flow cell, as a function of the number
of cycles. (Adapted from Ref. [60].)
3.0
CdTe 190-nm thickness
Gold on silicon
2.5 Experimental band gap EG = 1.55 eV
2.0
(α h ν)2
1.5
1.0
A
0.5
0.0
1.2 1.4 1.6 1.8
hν
[eV]
Fig. 15 Graph showing extrapolation to the band gap for CdTe films. (Adapted from
Ref. [60].)
evident in an optical microscope, appear- This is a strong indicator that the growth
ing as dark or multicolored dots, sand, mode is no longer layer by layer, and
when a 200-cycle deposit is examined at the deposit is roughening. Ellipsometric
1000 X with a metalographic microscope. thickness measurements are only accurate
Good Sandy
Fig. 16 Optical micrograph, 1000 X, taken with a
metalographic microscope of CdTe deposits. The good deposit
is on the left, and was produced using reasonable potentials.
The bad deposit is on the right, and was produced with a
program in which the potential for Te was excessively negative,
leading to roughening of the deposit, and what is referred to
here as sand, owing to its appearance in the microscope.
for smooth films [111], and became mis- described above, and Fig. 4 is a TEM of
leading for the films formed at such one of their deposits. Their cell was a wall
negative Te potentials. In between these jet design, and they used a single pump
potential extremes, the 0.7-V wide plateau to suck the solution from a distribution
was evident (Fig. 11), where the deposit valve, through their cell. The deposit was
thickness was consistent with ML/cycle formed in a continuous-flow mode, not
growth. The presence of this plateau is the stop flow used by this group. Their
consistent with surface-limited control of design greatly simplifies the hardware and
the deposition. Activation-limited growth appears to work just as well as the thin-
would have resulted in increased deposi- layer flow system. In addition, Villegas
tion as the potential was decreased. Mass was the first to notice the need to adjust
transfer–limited growth would have re- potentials as the deposits grow. They used
sulted in a plateau as well, but at a a cycle incorporating the reactions shown
much higher growth rate, and would not in Eqs. (4 and 5).
have been expected to increase again at Films of CdSe have been grown with
still more negative potentials. In addi- the automated flow-cell systems, both us-
tion, the deposit thickness for the plateau ing the H-cell [42] (Fig. 3b), and recently
was not a function of the precursor con- with the large thin-layer cell (Fig. 3c) [86].
centration, consistent with surface-limited Comparisons of XRD patterns have sug-
control. gested that in both cases the quality of
Thin films of CdTe have also been the CdSe deposits were better than equiv-
produced using an automated flow-cell alent CdTe films formed under similar
system by Villegas and coworkers [43]. conditions (Fig. 12b). Figure 17 shows an
Their films were very similar to those XRD pattern for a 200-cycle CdSe deposit,
10 000 Au
(111)
CdSe
8000 (111)
Intensity
6000
[counts]
Au
(200)
4000
CdSe Au
(220) CdSe (220)
2000 (311)
0
20 30 40 50 60 70
2θ
[degrees]
Fig. 17 XRD of CdSe formed using the large thin-layer flow cell, and 200 cycles. (Adapted from
Ref. [86].)
formed using cathodic UPD for both el- in growth and measurement at that time,
ements. The probable reason for CdSe rather than any significant variation in
forming higher quality deposits than CdTe coverage as a function of potential. The
is the lattice-match with the Au substrate. plateau and trends in the graph are as
That is, from atomic level studies, it is expected. At potentials above −0.55 V,
known that an ML and a half of CdTe [63, the deposits die off, while at potentials
78, 79], or CdSe [83, 84], will form a (3 × 3) below −0.68 V, some bulk deposition of
surface unit cell on Au(111). The struc- Se results, the growth rate exceeds an
tures proposed to account for these (3 × 3) ML/cycle, and the surface roughens (sand
unit cells involve superposition of an ML is evident).
and a half of these compounds, with the Typical 200-cycle CdSe deposits appear
zinc blende structure, on the Au surface, gold in color, as did the CdTe deposits.
where three Au lattice constants match up Given that the band gaps of CdSe and
with two times the Cd–Cd distance in the CdTe are both direct and about 1.6 eV,
(111) plane of CdTe or CdSe (Fig. 18). In similarities in appearance are expected.
the case of CdTe, this results in a 6% lat- Visual inspection of the 1 cm × 3 cm
tice mismatch, while it is less than 1% for CdSe deposits shows them to be more
CdSe on the same Au(111) surface. This homogeneous than corresponding CdTe
suggests there should be fewer defects, or deposits, in general.
a larger critical thickness for CdSe deposits Weaver and coworkers formed superlat-
compared with CdTe [112]. tices with CdSe and CdS, using EC-ALE,
Studies of the potential dependence of without an automated system [85]. They
CdSe deposit thicknesses show a much studied their relatively thin deposits by
shorter plateau region with CdSe than surface-enhanced Raman (SERS), examin-
CdTe (Figs. 11 and 19). The variability in ing stress build-up in the deposits.
thickness evident in the plateau (Fig. 19) CdS growth, by EC-ALE, has been
probably represents the standard deviation studied by more groups than any other
(a) (b)
√ √
Fig. 18 Structures proposed to account for the (a) ( 7 × 7)R19.1◦ and (b) (3 × 3) structures
observed in the formation of monolayer of CdTe. (Adapted from Ref. [61].)
200.0 190.6
155.6
150.0
Thickness
[nm]
100.0 93.5 91.5

83.4 84.9
81.9 71.2
68.1 65.4 61.4
62.9
50.0 41.9
15.0
0.0
−0.74 −0.69 −0.64 −0.59 −0.54
Se deposition potential (vs Ag/AgCl reference)
Fig. 19 Graph of the coverage of CdSe versus the potential used for Se deposition in the formation
of 200-cycle deposits. (Adapted from Ref. [86].)
compound (Table 1) [42, 52, 65, 67, 75, 85, as a function of the numbers of cycles
87, 89–91, 94, 113]. Initial EC-ALE studies performed. The dependence of thickness
in this group of CdS were performed with on the Cd deposition potential, for CdS
a TLEC (Fig. 9), to determine potentials deposits, revealed a plateau between −0.3
for a cycle [67]. Cd and S coverages were and −0.55 V, with the best deposits formed
determined coulometrically for deposits at −0.5 V, using a pH 5.9, 10 mM, CdSO4
solution and a pH 11, 11 mM, Na2 S system. At present there is no reason

solution. Reductive UPD was used for the to believe that the cycle developed for
Cd atomic layers and oxidative UPD for S 20 cycles will not produce good quality
atomic layers. deposits of any given thickness using the
The H-cell (Fig. 3b) was used to form automated flow-cell system.
CdS thin films with 200 cycles of deposi- There has been a report of ZnSe
tion [42]. The films were a transparent yel- thin films formed using EC-ALE in an
low from visual inspection. XRD indicated automated deposition system, by the
the deposits were again (111) oriented, zinc Foresti group [44]. Films with up to 31
blende CdS, but of a significantly lower cycles were formed, and produced the
quality than corresponding CdSe and CdTe expected linear graph of stripping charge
(Fig. 12c). The films were also fairly thin, versus number of cycles. The slope of the
only about 0.2 ML/cycle, compared with graph suggested that the deposits grew
the 0.4 ML/cycle obtained with CdTe us- at a rate of only about 0.14 ML/cycle. The
ing the H-cell, and 1 ML/cycle obtained reasons for this are unclear; however, most
for CdSe. EPMA suggested that the de- of the studies produced by the Foresti
posits were about 20% rich in S, as well. group have resulted in low coverages per
The cycle for CdS deposition had not been cycle. It appears that they are conservative
extensively optimized at that time. Some
with their deposition potentials, and do
reasonable CdS films were formed using
not shift them as the deposits grow. As
the large thin-layer flow cell, as well, but
discussed earlier, as long as depositing
have not yet been well characterized. More
less than an ML/cycle does not result in
work must be done on the cycle for CdS,
defects, higher-quality deposits may be a
and it is probable that a program in which
by-product of such cycles. Their report
the potentials are systematically shifted for
describes very careful coulometric studies
the first 30 cycles will greatly improve the
of both Zn and Se coverages.
deposit quality.
An automated flow system has also been TLEC studies of the first 14 cycles of
used by Foresti and coworkers to form ZnSe growth have been performed by this
CdS layers, with up to 150 cycles, using group [57]. That study has resulted in an
pH 9.2 solutions for both elements on initial set of conditions for an EC-ALE
Ag(111) electrodes [45]. In their case, the cycle, but no thicker films have as yet
deposits appeared stoichiometric, without been formed using an automated flow-cell
the excess S previously observed by this deposition system, by this group.
group [42]. Their cycle produced relatively Several groups have worked on the for-
thin deposits, similar to this author’s, or mation of ZnS using EC-ALE (Table 1) [23,
about 1/3 ML/cycle of CdS. 54, 57, 65, 75, 76, 81, 113]. The first study
EC-ALE studies of ZnTe using a TLEC was again by TLEC, where it proved diffi-
were performed with up to 20 cycles of cult to quantify stripping coulometry for
deposition [57]. Coulometry was the only both Zn and S separately for deposits
analysis performed on the deposits. A plot formed with more than 5 cycles. Foresti
of coverage as a function of number of and coworkers used a procedure in which
cycles was linear, as expected for a surface- S was reductively stripped and Zn oxida-
limited process. No thicker films have as tively stripped to accurately determine both
yet been formed using the flow-deposition Zn and S coverages from coulometry for
deposits formed with a greater number of time, however. A paper looking at the pho-
cycles. toelectrochemistry of ZnS films as a func-
A ZnS cycle was automated using the tion of the layer thickness was published
H-cell design, with the idea of form- by Yoneyama and coworkers in 2000 [76].
ing phosphor screens for field-emission Their study involved the growth of films
displays (FED) [54]. Doping studies were of increasing thickness with an increasing
performed with Ag, Cu, and Mn, and will number of EC-ALE cycles. The deposits
be discussed in a subsequent section. The were studied with photoelectrochemistry
films showed evidence of 3D growth and (photocurrent curves) and clearly showed
roughening, and were rich in Zn (a Zn/S the dependence of the band gap on the
ratio of 1.35 from EPMA). XRD showed thickness of the ZnS films. The band gaps
the deposits to be zinc blende ZnS, of of deposits formed with more than 4 cy-
variable quality, depending on the doping cles, about 1.5 nm, showed the expected
scheme. Deposits formed with as many dependence on thickness. The thinner
as 200 cycles were produced. A plot of deposits were blueshifted from the bulk
coverage versus cycle number was expo- value of 3.6 eV, to about 4 eV. However,
nential; the coverage per cycle increased deposits formed with less than 4 cycles
as more cycles were performed, consistent did not display the theoretical exponen-
tial dependence expected. This work was
with some 3D growth and roughening.
extended by the formation of CdS/ZnS su-
SEM of the deposits grown on ITO re-
perlattices [75], discussed in a subsequent
vealed small smooth bumps, suggesting
section.
that the deposits could be layer-by-layer
InAs was the first III to V compound
growth on 3D nucleation sites.
grown using EC-ALE, although there was
From the exponential graph of coverage
some early work concerned with the for-
versus cycle number, and the presence
mation of GaAs [68, 95]. In those early
of excess Zn, it appears that the Zn
GaAs studies, no more than a single
potential was pushed too far negative, ML was ever formed. The major prob-
causing some bulk Zn deposition and 3D lems at that time were the reactivity of
growth. These ZnS deposits, formed five Ga, and the hardware problems previ-
years ago, were deposited using the same ously discussed: lack of oxygen exclu-
cycle potentials throughout the deposition. sion and loss of potential control during
Graphs of Zn/S ratio, as a function of rinsing.
the number of cycles, indicated that after Indium, on the other hand, is less re-
50 cycles the deposits were very close active than Ga, making InAs easier to
to stoichiometric, possibly a little S rich. form than GaAs. Deposits of InAs of good
Excess Zn began to show up at the 100- quality have been formed using the au-
cycle point. It is probable that the ZnS tomated thin-layer flow-deposition system
deposits would be greatly improved, using (Fig. 3c) [40, 62]. Homogeneous deposits
the present hardware and shifting the have been formed with close to 1 : 1 stoi-
potentials through the first 30 cycles. chiometry. EPMA suggested the deposits
In 1999, Foresti and coworkers pub- were somewhat rich in As, as much
lished a paper describing the growth ZnS as 20%, but attempts to better optimize
and CdS, using their automated deposition the deposition potentials did not greatly
system [45]. Few details were given at that improve the EPMA results. Elemental
analysis, using inductively coupled plasma the (αhν)2 versus energy, the band gap was
mass spectroscopy (ICP-MS), suggested estimated to be 0.36 eV, in agreement with
these deposits were closer to 5% rich in literature values. Band gaps for the InAs
As and that there may have been some In deposits appear to be sensitive functions of
loss during EPMA analysis. Figure 20 is an a number of cycle variables. Several sam-
XRD pattern for zinc blende InAs, formed ples resulted in band gaps closer to 0.44 eV.
with 500 cycles. There are no indications These blueshifts appear to result from
of As in the XRD. From Fig. 21, a plot of smaller crystallites, nanoclusters, formed
Au
6000 (111)
5000
4000
Intensity
[counts]
InAs Au
3000 (111) (200)
2000 InAs Au
(220) InAs (220)
1000 (311)
0
20 30 40 50 60 70
2θ
[degrees]
Fig. 20 Glancing-angle XRD pattern of a 270-cycle deposit of InAs. (Adapted from Ref. [62].)
InAs, 0.36 eV
[arb. units]
(αhν )2
0.1 0.2 0.3 0.4 0.5 0.6

hν
[eV]
Fig. 21 Band gap measurement on InAs films formed with 200 cycles,
obtained from absorption data. (Adapted from Ref. [40].)
when the deposition conditions were less potential is shown in Fig. 24. At posi-
than optimal. tive potentials, above −0.6 V, little de-
AFM images of the InAs films show posit is formed, as would be expected.
them to be relatively conformal with the Au Below −0.6 V, a relative plateau is ob-
substrate, with some texture, but little 3D served, but which gradually increases be-
growth for a 250-cycle deposit. Figure 22(a) tween −0.625 and −0.775 V. Below about
is an image of an Au on glass substrate, −0.7 V, the deposits correspond to more
annealed at 550 ◦ C for 12 h. Figure 22(b) is than 1 ML/cycle, and some roughening
of an equivalent substrate onto which 250 is evident with optical microscopy. Be-
cycles of InAs have been grown. low −0.775 V, the coverages measured
The InAs program involved slowly with ellipsometry drop to the 1-ML/cycle
shifting the potentials for In and As level, but microscopy shows the deposits
deposition negatively for the first 25 cycles, to be roughened and sandy. Coverage
and then holding them constant. Figure 23 measurements with EPMA also indicate
is a graph of cycle potentials as a function that these deposits are rich in As, and
of the number of cycles. The progression there is a significant increase in cov-
is essentially exponential, asymptotically erage, as expected, demonstrating that
approaching the steady state values, as the the ellipsometry readings were faulty at
junction potential is built up. these negative potentials. The As poten-
The thickness dependence of the InAs tial was driven too far negative, and some
deposits as a function of the As deposition bulk As was formed for each cycle. The
0 1.00 0 1.00
µM µM
Data type Height Data type Height
Z range 20.0 nM Z range 20.0 nM
(a) (b)
Fig. 22 AFM images of (a) the Au on glass substrate, annealed at 550 ◦ C for 12 h and
(b) equivalent substrate onto which an InAs deposit has been formed with 200 cycles of
deposition.
0 20 40
− 0.400
Potential
[V]
− 0.600
− 0.800
Cycle no.
Fig. 23 A graph of the deposition potentials for In and As as a function of
the cycle number in the growth of InAs. (Adapted from Ref. [62].)
100
80
60
40
20
0
− 0.900 − 0.850 − 0.800 − 0.750 − 0.700 − 0.650 − 0.600 − 0.550
As deposition potential vs Ag/AgCl
[V]
Fig. 24 A graph of deposit thickness as a function of the potential used to
form As atomic layers. Each deposit was made with 200 cycles of
deposition. (Adapted from Ref. [62].)
best deposits were those where the As Again, the thickness measurements at
potential was between −0.6 and −0.7 V: these negative potentials, determined with
where the deposits did not appear rough, ellipsometry, appear faulty, consistent
had close to 1 : 1 stoichiometry, and IR with the roughening observed. The best
absorption indicated a band gap close deposits were grown in the potential range
to 0.36 eV. between −0.725 and −0.8 V.
The dependence of the InAs film The importance of the substrate struc-
thickness on the In potential used is ture on the quality of the resulting de-
shown in Fig. 25. Again, little deposition posit cannot be overemphasized. Figure 26
occurs for potentials above −0.7 V. There shows two reflection transmission IR spec-
is a plateau between −0.725 and −0.8 V, tra for InAs. The InAs was grown on an Au
where deposits grow at close to 1 ML/cycle, on glass substrate that had been carefully
and the deposit thickness increased below annealed. Then, half the substrate was cov-
−0.8 V, as some bulk In began to form ered, and more Au was vapor-deposited
for each cycle, creating a rough deposit. on one half at room temperature. The
2
Monolayers/cycle
1.5
0.5
0
− 0.9 − 0.85 − 0.8 −0.75 −0.7
In deposition potential vs Ag/AgCl
[V]
Fig. 25 Graph of the deposit thickness as a function of the In potential
used. (Adapted from Ref. [62].)
1.0 Unannealed
Annealed
Transmission
Substrate
[arb. units]
Annealed Unannealed
0.8
Deposit
0.6
2000 4000 6000 8000
Frequency
[cm −1]
Fig. 26 IR reflection transmission spectra for InAs films formed on smooth
annealed Au surface and roughened Au surface. Substrate was formed by taking
a well-annealed, smooth Au on glass substrate and vapor-depositing more Au
on one half, at room temperature, so that a series of 40 bumps were formed
over one half of the surface. The two deposits were subsequently formed in the
same electrodeposition run. (Adapted from Ref. [114].)
room-temperature Au greatly roughened that on the roughened side is signifi-

half of the substrate (Fig. 27). The results cantly blueshifted. The blueshift is the
are evident in Fig. 26, where the absorp- result of the substrate structure causing
tion of the InAs grown on the smooth the formation of a nanocrystalline InAs
side is characteristic of bulk InAs, while deposit.
0 1.00 0 1.00
µM µM
10061844.001 10061753.001
Au/glass, flame annealed Au/glass, unannealed
(a) (b)
Fig. 27 AFM images of an Au on glass substrate that had been well annealed, and then half
of it was covered. More Au was vapor-deposited on the other half at room temperature,
forming a surface composed of 40-nm Au hemispheres. (Adapted from Ref. [114].)
Given positive results with InAs, it was 115–119]; however, reduction of Sb, under
felt that InSb could be grown using EC- the conditions used by this group, was very
ALE, since the voltammetry of Sb is difficult. The stability of elemental Sb over
more straightforward than that for As. a large potential range suggested that InSb
That is, Sb+3 solutions have well-defined, formation would be tractable.
relatively reversible UPD features (Fig. 28), The potentials used to form InSb
while As+3 solutions show significant were applied in a roughly exponential
irreversibility (Fig. 6e). In addition, it has progression, as with InAs, through the
been shown that As is easily reduced first 30 cycles. Figure 29 shows the (111)
to arsine species at modestly negative XRD peak for zinc blende InSb, from
potentials [68, 95], making it difficult to a film formed with 200 cycles on a Cu
keep in the deposit if other steps in the substrate. The poor quality of the XRD
cycle are performed at overly negative appears to result from the small grain size
potentials, one of the basic problems in the deposit, as seen in the AFM image
with GaAs formation by EC-ALE. The in Fig. 30(b). The surface is composed of
potentials for Ga deposition were too smaller crystallites than the corresponding
similar to those where As is reduced to InAs deposits (Fig. 30a). The origin of
arsine. There are reports in the literature this difference appears to be the lattice
describing the reductive conversion of Sb constants, again. If we assume that the
to stibine, H3 Sb, or related species [110, interface is similar to that for CdTe (zinc
20.0
Current
0.0
[ µA]
−20.0
−0.80 −0.40 0.00 0.40

[V]
Fig. 28 Voltammetry for Sb deposition on Au, from a solution of Sb2 O3 , pH 5,
0.10 M NaClO4 , and approximately 50 µM Sb. (Adapted from Ref. [62].)
InSb
2500 (111)
Intensity
[counts]
1500
0
24 26 28 30 32
2θ
[degrees]
Fig. 29 XRD pattern of the (111) peak for zinc blende InSb, formed with 200
cycles on a Cu substrate. (This figure was adapted from Ref. [62].)
blende) on Au(111), then there should formation of CuInSe2 , CIS, over the last
again be the 2 : 3 match between the 17 years [120–132]. A number of electrode-
compound and the Au surface. The lattice position methodologies have been used,
mismatch for InSb would be 6%, as it including codeposition and a number of
was for CdTe and Au, as CdTe and InSb two-stage methodologies. The impetus of
have the same lattice constant (0.648 nm), that work has been the excellent photo-
while InAs and CdSe have the same lattice voltaic properties of CIS.
constant (0.606 nm), resulting in a much Initial TLEC studies developing an EC-
better lattice mismatch, only 1% with Au. ALE cycle to form CIS were published
There have been a number of stud- in 1996 [96]; however, a complete cycle
ies directed toward the electrochemical was not accomplished at that time. Steps
0 500 0 500
nM nM
inas 05051659.001
InAs 7-23-99, 200 cycles InSb 12-15-99, 200 cycles on annealed Au
(a) (b)
Fig. 30 AFM images of (a) 200 cycles of InAs and (b) 200 cycle of InSb, both on Au on glass
substrates.
involving Se UPD, followed by Cu, Se, In, are more problematic. Deposits have been
and then Se again were successful. How- made, which visually look very good, but no
ever, keeping In on the electrode while the peaks have as yet been observed with XRD,
next atomic layer of the significantly more besides those due to the Au substrate. Pos-
noble Cu was deposited, has proven prob- sible reasons for the absence of XRD would
lematic. Presently, these studies are being be the fact that the films are probably less
revisited using the automated flow-cell de- than 100-nm thick, have small crystallites,
position system, and Cu complexation is and the pattern for In2 Se3 has many peaks.
being used to shift its deposition potential Figure 31 is an AFM image of one of the
to more closely match that of In. However, deposits. EPMA indicates that the deposits
problems persist in keeping the In in the are close to stoichiometric. Reflection ab-
deposit. To better understand the mech- sorption measurements give a band gap of
anism for CIS formation, studies of the
1.6 eV, while there are various reports of
EC-ALE formation of In2 Se3 and Cu2 Se
band gaps between 1.5 and 1.8 eV in the
have begun.
literature [97].
It appears that some Cu2 Se has been
formed, but not a pure phase. Work is con-
tinuing on the deposition conditions to 4.2.4.1Toward Growing Device Structures
see if different phases can be formed, and There are a number of ways to intro-
some very interesting deposit morpholo- duce dopants into an EC-ALE deposit.
gies have been produced. The In2 Se3 films For instance, they can be introduced
500
250
0
0 250 500
nM
Fig. 31 An AFM image of a 200-cycle deposit of In2 Se3 . (Adapted
from Ref. [97].)
homogeneously throughout the deposit commercial phosphors are based on ZnS,

or delta doped into the structure. For a as a host material, with metals such as
relatively homogeneous distribution, low Ag, Cu, and Mn as activators. In that ini-
concentrations of oxidized precursors can tial EC-ALE doping study, ZnS films were
be incorporated into the reactant solu- formed on Au and ITO substrates. Two
tions. By using very low concentrations, methods were used, codeposition of the
the amounts incorporated in each atomic dopant with the Zn atomic layers and use
layer will be limited. The dopant can also be of a separate delta-doping step every tenth
incorporated in its own cycle step. Again, cycle. No detectable dopant was found
a low concentration would be used so that using EPMA, except for the case where
some fraction of an atomic layer is in- Ag was codeposited with the Zn atomic
troduced for each cycle. Alternatively, a layers. In that case, about 1/5 ML/cycle
delta-doping scheme can be constructed of Ag was deposited, probably replacing
where a fraction of an atomic layer of Zn. From XRD, the quality of the de-
dopant is deposited for every set number posit was significantly degraded by the
of cycles. Any of these scenarios is a simple excess Ag incorporation. The amount of
modification of the EC-ALE program. activator (Ag) was greatly in excess, judg-
Initial doping studies of ZnS were run ing by commercial ZnS phosphors. As
with the idea of forming phosphor screens noted, the Mn codeposited with Zn was not
for flat-panel display applications [54], as detectible with EPMA; however, cathodo-
noted earlier. Some of the most important luminescence of a deposit grown on ITO
Intensity
400 500 600 700 800

Wavelength
[nm]
Fig. 32Cathodoluminescence of a Mn-doped ZnS thin film,
grown on ITO. (Adapted from Ref. [54].)
resulted in the emission spectrum shown growth of CdS on Au, using EC-ALE,
in Fig. 32, which is characteristic of Mn. which was then capped with a single ML
of HgS [94]. The HgS capping layer was
4.2.4.2 Diodes deposited in two ways, one using EC-ALE
There are a number of papers in the litera- and the other by chemical exchange. The
ture concerning formation of compound chemical step involved exchange of the last
semiconductor diodes by electrodeposi- layer of Cd in a solution containing Hg
tion, the most popular structure being a ions. The deposits were studied with STM,
CdS-CdTe-based photovoltaic. In general, electrochemistry, and photoluminescence.
CdS was deposited first on an ITO on glass Photoluminescence showed excellent cou-
substrate, followed by a layer of CdTe, pling between the electrodeposited HgS
usually by codeposition [133–143]. layer and the underlying CdS, while photo-
So far, there has been little work done luminescence from the chemically formed
on the formation of heterojunctions using layer was not nearly as good.
EC-ALE. One study, however, performed Sailor and Martin and coworkers grew
by Shannon and coworkers, involved the an array of CdSe-CdTe nanodiodes [144]
in alumite (anodized aluminum, which lattices [75, 113]. These EC-ALE-formed

forms hexagonal arrays of 200-nm pours superlattices were deposited by hand,
[145–149]) using a compound electrodepo- where the cycles involved manually dip-
sition methodology called sequential mono- ping or rinsing the substrate in a sequence
layer electrodeposition (SMED) [144]. SMED of solutions.
involves a potential sequence in a single Thicker EC-ALE-grown superlattices
electrochemical bath to promote layer-by- have recently been formed using the au-
layer growth and to minimize accumula- tomated flow-deposition system [62]. A
tion of elemental Se [150]. Current voltage number of InAs/InSb superlattices have
measurements on the array revealed the been formed, using a combination of cy-
rectifying behavior expected of diodes. cles similar to those described in the
The first electrodeposition of a com- preceding text for InAs and InSb. As su-
pound superlattice was by Rajeshwar and perlattices are composed of alternating
coworkers [151], where layers of CdSe layers of two or more compounds, they
and ZnSe were alternately formed us- are characterized by their period, or the
ing codeposition in a flow system. That repeat thickness (Fig. 33). The superlat-
study was proof of concept, but resulted tices of InAs/InSb formed in this group
in a superlattice with a period signif- were built with 10 cycles of InAs, fol-
icantly greater than would be expected lowed by 10 cycles of InSb, making one
to display quantum confinement effects. period. A 30-cycle layer of InAs was first
There have since been several reports of formed on an Au on glass substrate, before
very thin superlattices formed using EC- the lattices were grown, in order to help
ALE [73, 75, 85, 113]. Surface-enhanced develop the junction potential. However,
Raman (SERS) was used to characterize every time the compound was changed,
a lattice formed from alternated layers twice each period, a new junction was
of CdS and CdSe [85]. Photoelectrochem- formed. Experimentally, it appeared that
istry was used to characterize CdS/ZnS the potentials needed to deposit the next
InSb
Repeat
unit
InAs
Fig. 33 Schematic diagram of a

superlattice.
compound layer had to be adjusted with Superlattices should display satellite peaks
each cycle. at angles corresponding to the period
Figure 34 is a graph of the potential of the lattice [152]. Given that the in-
program used to grow the first 100 cycles terplaner spacings, in the [110] direction,
of an InAs/InSb superlattice. Experience are 0.35 nm for InAs and 0.374 nm for
showed that the potentials had to be InSb, ideally, a period composed of 10 ML
changed for both elements for each of each should be 7.24-nm thick. Using
cycle. Those potentials and changes were Bragg’s law, with 0.1789-nm X rays, the
determined experimentally, and were not angle between the primary peak and its
extensively optimized. Figure 34 shows satellites should be 0.71◦ . Figure 35 sug-
that the potentials for As and Sb were each gests the presence of two shoulders, each
shifted to lower potentials with successive about 0.93◦ from the central peak. The
cycles in each half period. On the other quality of this preliminary XRD pattern
hand, the In potential was increased is bad, but the symmetric shoulders are
during the InSb half period, and decreased evident. On the basis of 0.93◦ , the pe-
during the InAs half period. At present, riod for the superlattice appears to be
these potential changes were arrived at closer to 5.5 nm, suggesting that about
by observation. It is hoped that feedback 3/4 ML/cycle was actually deposited.
can be incorporated to account for and Elemental analysis with ICP-MS was
optimize the progressions or that the performed on a dissolved piece of the
potential shifts can at least be modeled. superlattice, and indicated a deposit com-
Given the lattice mismatch of InAs position of In0.49 As0.37 Sb0.14 . It was grat-
(a = 0.606 nm) with InSb (a = 0.648 nm), ifying to find the deposit close to 50%
about 6.5%, defects are expected. At best In, as expected. However, there appears
a strained-layer superlattice would result. to be significantly more As than Sb. The
Figure 35 is an XRD pattern for a 41- deposit was started with 30 cycles of an
period InAs/InSb deposit, where each InAs buffer layer, so the deposit should
period was 10 cycles of InAs followed by be about 10% higher in As than Sb, but
10 cycles of InSb. The central [110] re- there was significantly more than 10% ex-
flection is near 28◦ and is quite broad. cess As over Sb. The absolute coverages
0 50 100
−0.35
Deposition potentials
−0.45
In potential
−0.55
As potential
−0.65 Sb potential
−0.75
−0.85
Cycle no.
Fig. 34 Deposition potential versus cycle number for an InAs10/InSb10-41X
superlattice: In (solid line), As (dot dash), Sb (dash). (Adapted from Ref. [62].)
105 27.887
90 26.876 28.779
[counts/second]
75
Intensity
60
45
30
25 27 29 31 33
2θ
[degrees]
Fig. 35 Grazing incidence angle X-ray diffraction (XRD) pattern of an
InAs10/InSb10-41X superlattice. The incident angle was 0.50◦ . (Adapted
from Ref. [62].)
1.0
0.8
Signal
0.6
0.4
0.2
0.0
1000 1500 2000 2500 3000 3500
Frequency
[cm−1]
Fig. 36 Reflection IR of an InAs10/InSb10-41X superlattice. (Adapted from
Ref. [62].)
from ICP-MS were about 10% less than redshifted from pure InSb, it is probably a
would be expected if an ML was deposited Type II superlattice [154].
for each cycle, suggesting somewhat more
than the 3/4 ML/cycle, as indicated by the Acknowledgment
XRD satellite peaks, was deposited. The
extra material is easily understood if the Acknowledgment is made to the National
substrate is considered slightly rough. A Science Foundation, Divisions of Chem-
roughness factor (RF) of 1.15 would ratio- istry and Materials for support of this work.
nalize the ICP-MS and XRD results, where
an RF of 1.0 corresponds to an atomically References
flat surface. One explanation of the dispar-
ity between As and Sb is that close to a full 1. A. Y. Cho, J. Vac. Sci. Technol. 1971, 8, s31.
ML of InAs was deposited for each cycle, 2. A. Y. Cho, J. R. Arthur, Prog. Solid State
while the coverage of InSb was closer to Chem. 1975, 10, 157.
3. J. R. Arthur, J. Appl. Phys. 1968, 39, 4032.
a 1/2 ML/cycle. This conclusion is consis- 4. J. Y. Tsao, Materials Fundamentals of Molec-
tent with the fact that the InSb cycle had ular Beam Epitaxy, Academic Press, Boston,
not been well optimized. 1993.
Figure 36 is an IR reflection- 5. M. B. Panish, H. Temkin, Annu. Rev. Mater.
Sci. 1989, 19, 209.
transmission spectrum taken from
6. M. A. Herman, H. Sitter, Molecular Beam
the superlattice deposit. Very strong Epitaxy: Fundamentals and Current Status,
adsorption is evident at higher wave Springer-Verlag, Berlin, 1989.
number, nearly 0% transmission, for the 7. H. O. Pierson, Handbook of Chemical Vapor
250-nm-thick superlattice deposit. The Deposition, Noyes Publications, Park Ridge,
1992.
deposit starts to absorb at very low 8. F. S. Galasso, Chemical Vapor Deposited
energies, near 1000 cm−1 . Problems with Materials, CRC Press, Boco Raton, 1991.
the spectrometer’s beam splitter make it 9. W. Kern in Microelectronic Materials and Pro-
hard to get a clear measure of the leading cesses (Ed.: R. A. Levy), Kluwer Academic,
Dordrecht, 1989.
edge of the absorption, and thus a band gap
10. K. K. Schuegraf, Handbook of Thin-Film
measurement. However, a comparison Deposition Processes and Techniques, Noyes,
was made between the absorptivity of the Park Ridge, 1988.
superlattice and that of a single crystal of 11. P. C. Andricacos, C. Uzoh, J. O. Dukovic
InSb, the superlattice component with the et al., IBM J. Res. Dev. 1998, 42, 567.
12. P. C. Andricacos, Interface 1999, 8, 32.
lower band gap, around 1300 cm−1 . The 13. M. Fleischmann, H. R. Thirsk, (Eds.), Metal
absorptivity of the superlattice was about Deposition and Electrocrystallization, Ad-
20 times greater than that for single crystal vances in Electrochemistry and Electro-
InSb, indicating that the superlattice is chemical Engineering, Interscience Pub-
lishers, John Wiley & Sons, New York, 1963,
redshifted from InSb. If the deposit was
Vol. 3.
an alloy, InAsx Sb1−x , the band gap should 14. P. E. Light, D. Shanefield, J. Appl. Phys.
be blueshifted from InSb, assuming it is 1963, 34, 2233.
a linear function of the mole fractions of 15. K. R. Lawless, J. Vac. Sci. Technol. 1965, 2,
As and Sb, and the band gaps for the two 24.
16. H. J. Choi, R. Weil in Electrochemical So-
compounds, or about 0.3 eV [153]. These ciety National Meeting (Eds.: R. Weil,
results strongly indicate that the deposit R. G. Barradas), The Electrochemical Soci-
is not an alloy, and since the band gap is ety, 1981, p. 169, Vol. 81-6.
17. J. P. G. Farr, A. J. S. McNeil, C. A. Loong, 39. E. Herrero, L. J. Buller, H. D. Abruna, Chem.

Surf. Technol. 1981, 12, 13. Rev. 2001, 101, 1897.
18. G. F. Fulop, R. M. Taylor, Annu. Rev. Mater. 40. T. L. Wade, L. C. Ward, C. B. Maddox et al.,
Sci. 1985, 15, 197. Electrochem. Solid State Lett. 1999, 2, 616.
19. K. Rajeshwar, Adv. Mater. 1992, 4, 23. 41. B. M. Huang, L. P. Colletti, B. W. Gregory
20. G. Hodes, Sol. Energy Mater. 1994, 32, 323. et al., J. Electrochem. Soc. 1995, 142, 3007.
21. R. K. Pandey, S. N. Sahu, S. Chandra, 42. L. P. Colletti, B. H. Flowers, J. L. Stickney,
Handbook of Semiconductor Electrodeposition, J. Electrochem. Soc. 1998, 145, 1442.
Marcel Dekker, New York, 1996. 43. I. Villegas, P. Napolitano, J. Electrochem.
22. G. Hodes in Physical Electrochemistry (Ed.: Soc. 1999, 146, 117.
I. Rubinstein), Marcel Dekker, New York, 44. G. Pezzatini, S. Caporali, M. Innocenti et al.,
1995, p. 515. J. Electroanal. Chem. 1999, 475, 164.
23. J. L. Stickney in Electroanalytical Chemistry 45. M. Innocenti, G. Pezzatini, F. Forni et al.
(Eds.: A. J. Bard, I. Rubenstein), Marcel in 195th meeting of the Electrochemical So-
Dekker, New York, 1999, p. 75, Vol. 21. ciety (Eds.: P. C. Andricacos, P. C. Searson,
24. B. W. Gregory, D. W. Suggs, J. L. Stickney, C. R. Simpson et al.), The Electrochemical
J. Electrochem. Soc. 1991, 138, 1279. Society, Seattle, 1999, p. 294, Vol. 99-9.
25. B. W. Gregory, S. J. L., J. Electroanal. Chem. 46. C. E. D. Chidsey, D. N. Loiacono, T. Sleator
1991, 300, 543. et al., Surf. Sci. 1988, 200, 45.
26. J. L. Stickney, B. W. Gregory, I. Villegas, 47. R. Emch, J. Nogami, M. M. Dovek et al., J.
U.S. Patent, University of Georgia, 1994, Appl. Phys. 1988, 65, 79.
p. 5320736, Vol. 6. 48. A. L. Putnam, B. L. Blackford, M. H. Jeri-
27. T. Suntola, J. Antson, U.S. Patent, USA, cho et al., Jpn. J. Appl. Phys. 1989, 217,
276.
1977, p. 4058430.
49. S. Buchholz, H. Fuchs, J. P. Rabe, J. Vac.
28. C. H. L. Goodman, M. V. Pessa, J. Appl.
Sci. Technol., B 1991, 9, 857.
Phys. 1986, 60, R65.
50. J. A. DeRose, T. Thundat, L. A. Nagahara
29. S. P. DenBaars, P. D. Dapkus, J. Cryst.
et al., Surf. Sci. 1991, 256, 102.
Growth 1989, 98, 195.
51. E. Holland-Moritz, J. Gordon-II, G. Borges
30. T. F. Kuech, P. D. Dapkus, Y. Aoyagi, Ato-
et al., Langmuir 1991, 7, 301.
mic Layer Growth and Processing, Materials
52. M. L. Foresti, G. Pezzatini, M. Cavallini
Research Society, Pittsburgh, 1991. et al., J. Phys. Chem. B 1998, 102, 7413.
31. A. Usui, H. Watanabe, Annu. Rev. Mater. 53. M. L. Foresti, M. Innocenti, F. Forni et al.,
Sci. 1991, 21, 185. Langmuir 1998, 14, 7008.
32. S. Bedair, Atomic Layer Epitaxy, Elsevier, 54. L. P. Colletti, R. Slaughter, J. L. Stickney,
Amsterdam, 1993. J. Soc. Info. Display 1997, 5, 87.
33. L. Niinisto, L. M., Thin Solid Films 1993, 55. H. Cachet, R. Cortes, M. Froment et al.,
225, 130. Philos. Mag. Lett. 1999, 79, 837.
34. D. M. Kolb in Advances in Electrochem- 56. L. Beaunier, H. Cachet, M. Froment et al.,
istry and Electrochemical Engineering (Eds.: J. Electrochem. Soc. 2000, 147, 1835.
H. Gerischer, C. W. Tobias), John Wiley & 57. L. P. Colletti, S. Thomas, E. M. Wilmer
Sons, New York, 1978, p. 125, Vol. 11. et al., MRS Symp. Boston 1996, 451, 235.
35. K. Juttner, W. J. Lorenz, Z. Phys. Chem. N. 58. L. P. Colletti, J. L. Stickney, J. Electrochem.
F. 1980, 122, 163. Soc. 1998, 145, 3594.
36. R. R. Adzic in Advances in Electrochem- 59. B. M. Huang, T. E. Lister, J. L. Stickney,
istry and Electrochemical Engineering (Eds.: Surf. Sci. 1997, 392, 27.
H. Gerishcher, C. W. Tobias), Wiley-Inter- 60. B. H. J. Flowers, T. L. Wade, M. Lay et al., J.
science, New York, 1984, p. 159, Vol. 13. Electroanal. Chem. 2002, in press.
37. A. T. Hubbard, V. K. F. Chia, D. G. Frank 61. K. Varazo, M. D. Lay, J. L. Stickney, Lang-
et al. in New Dimensions in Chemical Analysis muir 2002, in prep.
(Ed.: B. L. Shapiro), Texas A & M University 62. T. L. Wade, R. Vaidyanathan, U. Happek
Press, College Station, 1985, p. 135. et al., J. Electroanal. Chem. 2001, 500, 322.
38. A. A. Gewirth, B. K. Niece, Chem. Rev. 1997, 63. K. Varazo, M. Lay, J. L. Stickney, J. Electro-
97, 1129. anal. Chem., 2002, in press.
64. M. J. N. Pourbaix, Atlas of Electrochemical 88. E. S. Streltsov, I. I. Labarevich, Dokl. Akad.

Equilibria in Aqueous Solutions, Pergamon Nauk Bel. 1994, 38, 64.
Press, Oxford, 1949. 89. U. Demir, C. Shannon, Langmuir 1996, 12,
65. M. Innocenti, G. Pezzatini, F. Forni et al., 594.
J. Electrochem. Soc. 2001, 148, c357. 90. U. Demir, C. Shannon, Langmuir 1996, 12,
66. B. W. Gregory, M. L. Norton, J. L. Stickney, 6091.
J. Electroanal. Chem. 1990, 293, 85. 91. G. D. Aloisi, M. Cavallini, M. Innocenti
67. L. P. Colletti, D. Teklay, J. L. Stickney, et al., J. Phys. Chem. B 1997, 101, 4774.
J. Electroanal. Chem. 1994, 369, 145. 92. A. Gichuhi, B. E. Boone, U. Demir et al.,
68. I. Villegas, J. L. Stickney, J. Electrochem. Soc. J. Phys. Chem. B 1998, 102, 6499.
1992, 139, 686. 93. T. Torimoto, S. Nagakubo et al., Langmuir
69. A. T. Hubbard, F. C. Anson in Electroana- 1998, 14, 7077.
lytical Chemistry (Ed.: A. J. Bard), Marcel 94. A. Gichuhi, B. E. Boone, C. Shannon, Lang-
Dekker, New York, 1970, p. 129, Vol. 4. muir 1999, 15, 763.
70. A. T. Hubbard, Crit. Rev. Anal. Chem. 1973, 95. I. Villegas, J. L. Stickney, J. Vac. Sci. Tech-
3, 201. nol., A 1992, 10, 3032.
71. B. E. Boone, A. Gichuhi, C. Shannon, Anal. 96. R. D. I. Herrick, J. L. Stickney in New Di-
Chim. Acta 397, 1999, 43. rections in Electroanalytical Chemistry (Eds.:
72. S. Z. Zou, M. J. Weaver, J. Phys. Chem. B J. Leddy, M. Wightman), The Electrochem-
1999, 103, 2323. ical Society, Pennington, 1996, p. 186,
73. C. Shannon, A. Gichuhi, P. A. Barnes et al. Vol. 96-9.
in National Meeting of the Electrochemical So- 97. R. Vaidyanathan, J. L. Stickney, U. Happek,
ciety (Eds.: P. C. Andricacos, P. C. Searson, J. Electroanal. Chem. 2002, in press.
C. Reidsema-Simpson et al.), ECS, Seattle, 98. W. J. Danaher, L. E. Lyons, Nature 1978,
1999, p. 282, Vol. 99-9. 271, 139.
74. A. Gichuhi, C. Shannon, S. S. Perry, Lang- 99. M. P. R. Panicker, M. Knaster, F. A. Kroger,
muir 1999, 15, 5654. J. Electrochem. Soc. 1978, 125, 566.
75. H. Yoneyama, A. Obayashi, S. Nagakubo 100. G. Fulop, M. Doty, P. Meyers et al., Appl.
et al., Abstracts Electrochem. Soc. Meeting Phys. Lett. 1982, 40, 327.
1999, 99-2, 2138. 101. B. M. Basol, J. Appl. Phys., 1984, 55, 601.
76. T. Torimoto, A. Obayashi, S. Kuwabata 102. R. N. Bhattacharya, K. Rajeshwar, J. Electro-
et al., Langmuir 2000, 16, 5820. chem. Soc. 1984, 131, 2032.
77. D. W. Suggs, I. Villegas, B. W. Gregory 103. S. S. Ou, O. M. Stafsudd, B. M. Basol,
et al., J. Vac. Sci. Technol., A 1992, 10, 886. J. Appl. Phys. 1984, 55, 3769.
78. D. W. Suggs, J. L. Stickney, Surf. Sci. 1993, 104. M. Takahashi, K. Uosaki, H. Kita, J. Appl.
290, 362. Phys. 1984, 55, 3879.
79. D. W. Suggs, J. L. Stickney, Surf. Sci. 1993, 105. K. Uosaki, M. Takahashi, H. Kita, Elec-
290, 375. trochim. Acta 1984, 29, 279.
80. L. B. Goetting, B. M. Huang, T. E. Lister 106. R. D. Engelken, T. P. V. Doren, J. Electro-
et al., Electrochim. Acta 1995, 40, 143. chem. Soc. 1985, 132, 2904.
81. C. K. Rhee, B. M. Huang, E. M. Wilmer 107. G. Maurin, O. Solorza, H. Takenouti, J. Elec-
et al., Mater. Manufact. Proc. 1995, 10, 283. troanal. Chem. 1986, 202, 323.
82. B. E. Hayden, I. S. Nandhakumar, J. Phys. 108. H. Minoura, M. Kitakata, T. Sugiura et al.,
Chem. B 1998, 102, 4897. Bull. Chem. Soc. Jpn. 1987, 60, 2373.
83. T. E. Lister, J. L. Stickney, Appl. Surf. Sci. 109. J. v. Windheim, A. Darkowski, M. Cocivera,
1996, 107, 153. Can. J. Phys. 1987, 65, 1053.
84. T. E. Lister, L. P. Colletti, J. L. Stickney, Isr. 110. H. W. Salzberg, A. J. Andreatch, J. Electro-
J. Chem. 1997, 37, 287. chem. Soc. 1954, 101, 528.
85. S. Zou, M. J. Weaver, Chem. Phys. Lett. 111. H. G. Tompkins, A User’s Guide to Ellipso-
1999, 312, 101. metry, Academic Press, Boston, 1993.
86. B. H. Flowers Jr., T. L. Wade, K. Mathe 112. J. H. van-der-Merwe, Crit. Rev. Solid State
et al., MRS Symposium proceedings. 2002. Mater. Sci. 1978, 7, 209.
87. U. Demir, C. Shannon, Langmuir 1994, 10, 113. T. Torimoto, A. Obayashi, S. Kuwabata
2794. et al., J. Electrochem. Soc. 2000, in press.
114. T. L. Wade, B. H. Flowers Jr., K. Varazo 134. R. N. Bhattacharya, K. Rajeshwar, J. Appl.

et al., Conference Proceedings of the Electro- Phys. 1985, 58, 3590.
chemical Society, Washington, 2001, 8. 135. G. C. Morris, A. Tottszer, S. K. Das, Mater.
115. A. Reisman, M. Berkenblit, E. C. Haas et al., Forum 1991, 15, 164.
J. Electrochem. Soc. 1954, 101, 387. 136. S. K. Das, G. C. Morris, J. Appl. Phys. 1992,
116. A. L. Pitman, M. Pourbaix, N. D. Zoubov, 72, 4940.
J. Electrochem. Soc. 1957, 104, 594. 137. G. C. Morris, S. K. Das, P. G. Tanner,
117. L. Tomlinson, Anal. Chim. Acta 1964, 31, J. Cryst. Growth 1992, 117, 929.
545. 138. T. Yoshida, J. Electrochem. Soc. 1992, 139,
118. L. Tomlinson, J. Electrochem. Soc. 1964, 111, 2353.
592. 139. S. K. Das, Thin Solid Films 1993, 226, 259.
119. M. C. Hobson Jr., H. Leidheiser Jr., Trans. 140. S. K. Das, G. C. Morris, J. Appl. Phys., 1993,
Metallurg. Soc. AIME 1965, 233, 482. 73, 782.
120. R. N. Bhattacharya, J. Electrochem. Soc. 1983, 141. S. K. Das, G. C. Morris, Sol. Energy Mater.
130, 2040. 1993, 28, 305.
121. T. L. Chu, S. S. Chu, S. C. Lin et al., J. Electro- 142. S. Dennison, J. Mater. Chem. 1994, 4, 41.
chem. Soc. 1984, 131, 2182. 143. D. Lincot, A. Kampmann, B. Mokili et al.,
122. G. Hodes, T. Engelhard, C. R. Herrington Appl. Phys. Lett. 1995, 67, 2355.
et al., Prog. Cryst. Growth Charact. 1985, 10, 144. J. D. Klein, R. D. Herrick, D. Palmer et al.,
345. Chem. Mater. 1993, 5, 902.
123. R. P. Singh, S. L. Singh, S. Chandra, J. Phys. 145. J. W. Diggle, T. C. Downie, C. W. Goulding,
D: Appl. Phys. 1986, 19, 1299. Chem. Rev. 1969, 69, 365.
124. C. D. Lokhande, J. Electrochem. Soc. 1987, 146. G. E. Thompson, Y. Xu, P. Skeldon et al.,
134, 1728. Philos. Mag. B 1987, 55, 651.
125. C. X. Qiu, I. Shih, J. Appl. Phys. 1988, 64, 147. A. Despic, V. P. Parkhutic, (Eds.), Electro-
758. chemistry of Aluminum in Aqueous Solutions
126. D. Pottier, G. Maurin, J. Appl. Electrochem. and Physics of its Anodic Oxide, Modern As-
1989, 19, 361. pects of Electrochemistry, Plenum Press,
127. B. M. Basol, J. Vac. Sci. Technol., A 1992, 10, New York, 1989, Vol. 20.
2006. 148. R. M. Metzger, V. V. Konovalov, M. Sun
128. S. Menezes, Appl. Phys. Lett. 1992, 61, 1564. et al., IEEE Trans. Magn. 2000, 36, 30.
129. J. F. Guillemoles, P. Cowache, A. Lusson 149. G. Zangari, D. N. Lambeth, IEEE Trans.
et al., J. Appl. Phys. 1996, 79, 7293. Magn. 1997, 33, 3010.
130. R. P. Raffaelle, J. G. Mantovani, R. Fried- 150. A. M. Kressin, V. V. Doan, J. D. Klein et al.,
feld, IEEE Photovoltaic Specialists Conference, Chem. Mater. 1991, 3, 1015.
IEEE, Anaheim, 29 September, 1997. 151. C. Wei, K. Rajeshwar, J. Electrochem. Soc.
131. A. A. I. Al-Bassam, Physica B 1999, 266, 1992, 139, L40.
192. 152. Y. Sakuma, M. Ozeki, K. Kodama, J. Cryst.
132. M. Kemell, M. Ritala, H. Saloniemi et al., Growth 1991, 115, 324.
J. Electrochem. Soc. 2000, 147, 1080. 153. R. Pierret, Semiconductor Device Fundamen-
133. B. M. Basol, E. S. Tseng, R. L. Rod, Proc. tals, Addision-Wesley, Reading, 1996.
16th IEEE Photovolt. Special. Conf. San Diego 154. P. J. P. Tang, M. J. Pullin, S. J. Chung et al.,
1982, 805. Semicond. Sci. Technol. 1995, 10, 1177.
3
1.1 section of this volume covers kinetics and

The Current–Potential Relationship mechanisms of selected electrochemical
processes.
Ernesto J. Calvo
Universidad de Buenos Aires, Buenos Aires, 1.1.2
Argentina Introduction
The current–potential relationship at elec-

1.1.1 trodes is set by a number of com-
Scope plex physical and chemical phenom-
ena, depending on the experimental
This first chapter to Volume 2 ‘‘Interfacial conditions.
Kinetics and Mass Transport’’ introduces Electrode reactions are heterogeneous
the following sections, with particular fo- chemical processes that may involve one
cus on the distinctive feature of electrode or more electron-transfer steps across the
reactions, namely, the exponential cur- electrochemical double layer [1, 2]. Elec-
rent–potential relationship, which reflects trode reactions provide a switch for charge
the strong effect of the interfacial electric to flow between phases of different type of
field on the kinetics of chemical reactions electrical conductivity: electrodes and elec-
at electrode surfaces. We then analyze the trolyte [3]. Therefore, their response can
consequence of this accelerating effect on be analyzed either on the basis of electrical
the reaction kinetics upon the surface con- or chemical models. The distinctive fea-
centration of reactants and products and ture of reactions at electrodes is the strong
the role played by mass transport on the dependence of both the surface concen-
current–potential curves. The theory of trations and the kinetics on the electrode
electron-transfer reactions, migration, and potential [4–10].
diffusion processes and digital simula- The following reactions at electrified
tion of convective-diffusion are analyzed interfaces are examples of electrode re-
in the first four chapters. New experi- actions:
mental evidence of mechanistic aspects in
electrode kinetics from different ‘‘in-situ’’
Feaq 2+ −−−→ Feaq 3+ + e(metal) (1)
spectroscopies and structural studies are
2+
discussed in the second section. The last Cu −−−→ Cuaq + 2e(metal) (2)
4 Fundamentals
Xaq ± −−−→ Xorg ± (3) which is proportional to the rate of the

electrochemical reaction (in moles per
The first reaction is an electron transfer second):
across the double layer at the elec- dN
trode–electrolyte interface between redox I = nF (5)
dt
species in the electrolyte that exchange
electrons with a metal electrode, the sec- Since we are dealing with heterogeneous
ond one is an ion-transfer reaction across reactions, we need to normalize the
the double layer since the electron lost total current dividing by the electrode
by the Cu atom remains at the metal. area, A and thus the current density
The third one is an ion-transfer process is i = I /A. It should be noticed that
across the water–organic solvent interface the current measures an average value
or ion transfer at immiscible electrolyte of the reaction rate over the electrode
solutions (ITIES) without the transfer of area, which is important to keep in
electrons. In all cases the electrochem- mind given our present knowledge of
ical reaction takes place at an electri- heterogeneities at electrode surfaces in
fied interface and therefore the rate of micro- and nanoscale. (At present it is
these reactions follow similar exponen- possible to study electrode surfaces ‘‘in
tial dependence on the interfacial electrical situ’’ in contact with electrolytes and under
potential. potential control using scanning probe
While electrodes can be metals, car-
microscopes with spatial resolution in the
bon or semiconductors, and so on and
micro and nanometer.)
the charge carriers are electrons or holes,
In 1905, Tafel [11] described the exper-
the charge carriers in liquid and solid
imental relationship between current, I ,
electrolytes are solvated or nonsolvated
and the overpotential η – the difference be-
ions in molten salts and solid electrolytes.
tween the working electrode potential, E,
Therefore the flow of electricity is accom-
and the equilibrium potential Ee – during
plished by the electrode reactions at the
the electrochemical reduction of protons to
interface between these electrode and elec-
molecular hydrogen on different electrode
trolyte phases.
materials such as Hg, Sn, Bi, Au, Cu, Ni,
The relation between the total quan-
and so on:
tity of electricity, q, and the number of
moles of each chemical being transformed
during the electrolysis has been estab- η = a + b log I (6)
lished by Michael Faraday in the eighteen
century Figure 1 depicts a reproduction of the
q = nF N (4) original current–potential curves reported
by Julius Tafel for different metals, which
where n is the number of charges demonstrates the electrocatalytic nature of
transferred per molecule or ion, N the this reaction [11].
number of moles and q the total quantity Electrode reactions are heterogeneous
of electricity passed through the interface chemical reactions in solution and may
and thus through the external circuit. include elementary electron-transfer steps,
Differentiating the amount of charge ion transfer, potential-independent, or
with respect to time yields the current, chemical steps, and so on.
1.1 The Current–Potential Relationship 5
Fig. 1 Steady state

overpotential versus log I curves Hg
2.0
for the hydrogen evolution on
different metals. (Taken from
Ref. [11].) Sn
1.8
Bi
Au
1.6
Cu
Ni
1.4
η
1.2
1.0
0.0
0.00 0.05 0.10 0.15

log I
1.1.3 electrons transferred from the electrode,

Thermodynamics of Electrochemical Cells the standard free-energy change for the
overall reaction with hydrogen:
The electrode potential can be expressed by
n
O+ H2 = R + nH+ (9)
E = (φM − φS ) + Const (7) 2
where φM is the inner or Galvani potential is given by
of the metal phase and φS the inner or ◦ ◦
Galvani potential of the electrolyte solu- −G = nF E = RT ln Keq (10)
tion phase. In electrokinetic experiments,
where E ◦ is the standard potential of
the electrode potential is measured with
respect to a reference electrode and there- the O/R redox couple with respect to
fore an undetermined constant adds to the the standard hydrogen reference electrode
electrode potential [10]. (SHE) and Keq is the equilibrium constant
If we consider the generic electroreduc- for Reaction (9).
tion half-reaction: The dependence of the electrode po-
tential on the activities of O and R at
O + ne −
←−−
−−
→−R (8) equilibrium can be described by the Nernst
equation:
where O is the soluble oxidized and

R the reduced species present in the ◦ RT ao
Ee = E + ln (11)
electrolyte and n is the number of nF aR
6 Fundamentals
where aO and aR denote the activities of O and,

and R, respectively. The use of formal or ia
vb = kb CR∗ = (15)
conditional equilibrium potentials, E ◦ , is

F
more convenient in electrode kinetics [2]
where kf and kb are the corresponding
since concentrations instead of activities
rate coefficients and ic and ia the cor-
can be used, but it must be kept in mind
responding formal cathodic and anodic
that activity coefficients are included in the
partial current densities that flow at the
rate coefficients.
electrode–electrolyte interface. These par-
The steady state current–potential rela-
tial currents cannot be measured since
tionship under electrode kinetics condi-
we cannot distinguish in the external cir-
tions has been described by Butler [12, 13]
cuit those electrons flowing towards the
and Volmer [14].
electrolyte from those flowing towards the
electrode, and therefore we can only detect
1.1.4 the net current, I = ic − ia . This difference
The Butler–Volmer Equation
implies that we must adopt a convention
of signs for the reduction and oxidation
In this section, we shall derive an expres-
currents [2].
sion for the current–potential relationship
The net reaction rate is
in electrode reactions. Since multielectron-
transfer reactions at interfaces are most i
v = vf − vb = (16)
likely to occur in a series of one-electron F
steps, we shall consider first the general
According to the absolute reaction rate
one-electron electrode reaction:
theory of Eyring, first applied to electrode
kf reactions, the rate coefficient of a chemical
O+e −
←−−
−−
→−R (12) reaction, k, can be expressed as
kb

=

the rate of which is considered to be kB T Gi
ki = κ exp − (17)
totally determined by the charge-transfer h RT
process, and mass-transport limitations

=
or chemical step limitations will not be where Gi is the free energy of activa-
considered here so that the concentrations tion for the reaction, kB the Boltzmann
of reactant and product at the interface will constant, h the Plank constant and κ the
be the same as the analytical concentration transmission coefficient which for simplic-
in the bulk of the electrolyte. ity we shall consider here equal to one.
The net rate of Reaction (12) will be given In the case of electrode reactions that oc-
by the difference between the forward and cur at an electrified interface and therefore
backward rates: under the influence of a strong elec-
trical field (c.a. 107 V cm−1 ), the driving
v = vf − vb (13) force for the reaction of charged species

=
Gi has a term of electrostatic work
which, for reaction order one for reactant
wel = −F E as shown in the scheme of
and product, can be expressed as
Fig. 2. The negative sign corresponds to
ic the negatively charged electron crossing
∗
vf = kf CO = (14) the electrical double layer.
F
A′ aFE
C FE
(1 − a)FE
∆G ≠f (E) A
∆G ≠b (E)
∆Gf (Ee) FE
O + e− R
Fig. 2 Schematic representation of the effect of electrode potential
on the free energy versus reaction coordinate curves for an electron
reactant at two electrode potentials: E = Ee (solid line) and E < Ee
(broken line).
For convenience, we divide the free It is convenient to define

energy of activation into a potential-
=
independent and a potential-dependent kB T Gf (Ee )
kf ,o = exp − (22)
contribution: h RT

=
=
Gf (E) = Gf (Ee ) + αF E (18) and

=
and, kB T Gb (Ee )
kb,o = exp − (23)

=
=
h RT
Gb (E) = Gb (Ee ) − (1 − α)F E
(19) And therefore,
where 0 ≤ α ≤ 0 is the fraction of the
−αF E
applied potential that drives the reaction kf (E) = kf ,o exp and
and a measure of the symmetry of the RT

energy barrier. (1 − α)F E
kb (E) = kb,o exp (24)
Replacing in Eq. (17): RT

=
kB T Gf (Ee ) Thus from Eq. (16), the net current density
kf = exp − at the electrode interface is
h RT
∗ −αF E
−αF E i = F kf ,o CO exp − kb,o CR∗
× exp (20) RT
RT
(1 − α)F E
kB T

=
Gb (Ee ) × exp (25)
kb = exp − RT
h RT
which has an exponential dependence on

(1 − α)F E the electrode potential, E, as observed in
× exp (21)
RT experiments.
8 Fundamentals
We note that at equilibrium i = 0 Eq. (27), we obtain the Butler–Volmer

and E = Ee and therefore from Eq. (25) equation (29):
we obtain
−αF η (1 − α)F η
RT kf ,o CO i = io exp − exp
Ee = ln + ln (26) RT RT
F kb,o CR
(29)
which is identical to the Nernst equa- named after Butler and Volmer, the
tion for the reaction between O and R pioneers in this field. The validity of
at the electrode surface. Note that the first the Butler–Volmer equation has been
term in Eq. (26) is the conditional or for- verified by isotopic labeling [8]. It relates
mal standard electrode potential for the the current density that flows through
reaction, E ◦ , which contains the nonelec-

the electrode–electrolyte interface due to
trical terms of the free energy of activation the electrode reaction to the overpotential
including the activity coefficients for the in terms of two kinetic parameters: the
species O and R. exchange current density, io , and the
We can group all rate coefficients by transfer coefficient, α.
substituting E = Ee in Eqs. (24 and 26) The exchange current density intro-
and multiplying by F and obtain the duced by Butler [12, 13] is a measure of
exchange current density, io : the exchange rate between O and R at
the electrode–electrolyte interface at equi-
E◦

1−α α
librium and measures the height of the
io = F kf ,o CO CR exp −αF activation barrier when both the reactant
RT
and the product of an electrode reaction
E◦

1−α α are at the same free-energy level (as shown
= F kb,o CO CR exp (1 − α)F in Fig. 2).
RT
The transfer coefficient, α, introduced
(27)
in electrode kinetics by Erdey–Gruz and
we can further define a standard state both
Volmer [14] for the hydrogen evolution
for the electrode potential E = Ee and
reaction (HER), measures the symmetry of
the reactant and product concentrations
the free-energy curves at their intersection
CO = CR = 1 M:
in the transition state. Figure 3 depicts a

E◦

plot of ln i versus η according to Eq. (29).
◦
ko = kf ,o exp −αF Rewriting the Butler–Volmer equation,
RT
one obtains an expression for the over-

E◦

potential that explains the experimental
= kb,o exp (1 − α)F (28) findings of Tafel:
RT
2.303aRT
and thus, io = F ko◦ CO CR . ko◦ is the
1−α α η=
αF
standard rate constant for unit surface    

 

concentration of reactants at the standard   i  
equilibrium potential. × − log 
  + log io


 Fη 

Substituting in Eq. (25) and introducing  1 − exp 
the overpotential for the reaction as RT
η = (E − Ee ) and substituting io into (30)
log ( j/jo)
1
−1
−2
−150 −100 −50 0 50 100 150
E − Eo
Fig. 3 Plots of log |i/io | versus η calculated with Eq. (29).
The Tafel relationship is a linear free- which resembles the ohmic current–
energy relation of the rate coefficient for a potential behavior observed in conductor
net electrode reaction (neglecting the back phases (metals and electrolytes). Notice
reaction) and the transfer coefficient: that the slope of the current potential plot
near equilibrium is independent of α.

=
1 ∂Gf By analogy with Ohm’s law, the charge-
α=− (31)
F ∂E transfer resistance, Rt , can be defined
Ee
as a particular case of the polarization
◦
and since E = −G /F , the Tafel depen- resistance: Rp = (dη/di)i→0 :
dence is a linear free-energy relationship
RT 1
between the kinetics and the thermody- Rt = · (33)
namic driving force like the Brönsted nF io
relationship between the rate of proton which is a measure of io and is also
transfer and the acid pKa in acid–base independent of α.
catalysis [10]. It should be noted that in deriving
At low overpotentials (η RT /αnF ), Eqs. (18 and 19), only the first term of the
which fulfills for 5–10 mV excursion series expansion of the free-energy curves
around the equilibrium potential, the has been taken into account (i.e. the re-
exponential expressions in Eq. (29) can be actants and products parabolas have been
expanded in series, and the first series considered linear at the crossing point) and
term in the expansion gives therefore α results independent of the elec-
nF η trode potential; this is not the case when
i∼
= io (32) microscopic theories like the Marcus the-
RT
ory based on a parabolic description are
a linear i –η relationship for the flow of introduced. Thus, a limitation of the But-
charge at the interface near equilibrium, ler–Volmer current–potential relation is
10 Fundamentals
that it cannot account for the curvature the vibrational, λi contributions:

of α with the electrode potential far from
the equilibrium potential and does not λ = λo + λi (35)
incorporate the temperature dependence
of k o explicitly. Assuming the solvent as a dielectric
continuum, the outer sphere component,
1.1.5 λo , can be expressed in terms of the ionic
Theory of Electron Transfer radii a1 and a2 of the inner coordination
shell (a2 → ∞ for reactions at electrodes),
Rudolf Marcus was Chemistry Nobel lau- the center-to-center separation distance of
reate in 1992 for his work on the theory the reactants R, the optical (εop ) and static
of homogeneous and heterogeneous elec- (εs ) dielectric constants of the solvent and
tron transfer (ET) in such diverse processes the charge transferred (e) between the
as the bioenergetic of photosynthesis and reactants:
respiration, homogeneous redox and het-
1 1 1
erogeneous electrode reactions, corrosion, λo = (e)2 + +
2a1 2a2 R
color photography, and so on. Marcus the-
ory provides an insight on the physical 1 1
× − (36)
and chemical phenomena of the heavy εop εs
particle subsystem during electron trans-
fer in redox reactions. Furthermore, it can The inner sphere component, λi is
accommodate disparate rates over many given by
orders of magnitude for soluble redox
1 p
species, intramolecular electron transfer λi = kj (qjr − qj )2 (37)
2
in biological reactions of utmost impor- j
tance, and a common framework for a
p
comparison of redox processes in solution where qjr and qj are the equilibrium
and at electrode surfaces. values for the j th normal mode coordinate
In its simplest form the Marcus expres- q and kj is a reduced force constant,
p p
sion for the electron-transfer rate coeffi- kj = 2kjr kj /(kjr + kj ).
cient is given by [15–20]: For interfacial electrode reactions, the
  free energy of activation is thus given by
G◦ 2
 λ 1+ λ 
 
= λ F (E − E ◦ ) 2
kET = Z exp −  (34) Gi = 1+ (38)
 4RT  4 λ
since G◦ = −FE ◦ .

with λ the reorganization free energy The Marcus free-energy relation predicts
of the reactants to yield products and a nonlinear dependence of the transfer
G◦ the standard free energy of the coefficient, α on electrode potential:
reaction (thermodynamic driving force)
=
and Z a frequency factor taken by Marcus 1 ∂Gi 1 F (E − E ◦ )
α= = +
as 104 cm−1 for heterogeneous electrode F ∂E 2 2λ

reactions [19]. The reorganization free w o − wR
− (39)
energy, λ comprises the solvation, λo , and 2λ
where wo and wR are the electrostatic distance like in intramolecular electron

work terms due to the electrical work transfer. The first evidence of the inverted
involved in bringing charged reactants at Marcus region at a polarized interface was
the reaction site or a charged reactant close provided by the study of ET at liquid–liquid
to the charged electrode surface (double interfaces with an adsorbed C-10 lipid layer
layer or Frumkin effects, see below). by scanning electrochemical microscopy
Notice that while the Butler–Volmer (SECM) [22–25] and is described in
equation predicts α = 0.5 independent Chapter 4.2.
of the electrode potential, Marcus theory Since the appearance of the Marcus
predicts curvature dependent on electrode theory, for many years a clear potential
potential and reorganization energy. Only dependence of α with potential has been
when |G◦ /2λ| 1, the Marcus theory difficult to prove experimentally with
predicts α = 0.5, which corresponds to low soluble redox species reacting at electrodes
polarization and slow reactions. because the diffusion limit is reached at
The Marcus theory also allows to com- high ET transfer rates where one would
pare homogeneous electron-transfer re- expect the quadratic term to become
actions with the same reactions at elec- important.
trode surfaces and cross-reaction rates For heterogeneous electron transfer,
in terms of the individual self-exchange the use of ordered organic monolayers
reactions, that is, λ1,2 ∼
= 1/2(λ1,1 + λ2,2 ). (self-assembled monolayers or SAMs)
Experimental verification over 20 orders of at electrode surfaces as blocking films
magnitude has been the greatest success either with electroactive species in the
of this theory. electrolyte [26] or with electroactive groups
Another, very impressive prediction of tethered at the opposite end of the blocking
the Marcus theory is the inverted region, molecule from the covalent attachment
coined by Marcus in 1960 [17]: For a se- end [27] has provided a method to study
ries of related reactions of similar λ but the effect of electrode-redox center distance
different driving force, G◦ , as G◦ be- and the effect of the electrode potential on
comes negative the activation free energy, the electron-transfer rate.
G
= , first decreases as one would ex- The effect of tunneling distance in het-
pect in a Brönsted free-energy plot for erogeneous electron transfer at electrode
acid-base catalyzed reactions, and reaches surfaces has been shown by using block-
G
= = 0 when G◦ = −λ. Then the ac- ing organized thin films (thickness less
tivation energy increases again when G◦ than 1.5 nm) where the electrons can tun-
becomes even more negative, leading to nel through the films in the absence of
an inversion in the plot of logarithm of defects or pinholes that would otherwise
kET versus G◦ . An experimental test of allow the access of electroactive species in
this inverted region was provided in 1984 direct contact with the electrode surface.
by Miller, Calcaterra, and Closs for elec- For ω-hydroxy-alkylthiol monolayers of dif-
tron transfer between a biphenyl group to ferent number of methylene units Miller
an acceptor center [21]. The difficulty with and coworkers [26] showed that the ET
detecting the inverted region has been for heterogeneous rate constant, kET , decays
many years the mass transport limit in very exponentially with the number of methy-
fast redox reactions unless the electron lene groups in the alkylthiol, n, which is
donor and acceptor are positioned at fixed equivalent to the distance from the redox
12 Fundamentals
species to the electrode, x: rates of electron transport across organic

−βx thin films with a range of structures by us-
kET (x) = ko (x = 0)e (40)
ing a Metal-SAM(1)SAM(2)-Metal junction
where ko is the preexponential rate co- using a mercury drop covered with C-16
efficient and β the tunneling coefficient, alklylthiol. The current measured across
assumed potential independent, and close these junctions obeys an exponential re-
to 1 Å−1 [28, 29] for saturated alkylthiols lation with a decay constant dependent
in analogy to intramolecular through-bond on the molecular structure. Similar stud-
electron transfer, and 0.4–0.6 Å−1 for π- ies have been reported by Schiffrin [31],
conjugated molecular spacers. Electron Lindsay [32], and Bard [33] with scanning
tunnel factors in ET through molecules tunneling microscopy (STM) examination
and molecular interfaces is considered of SAMs at electrode surfaces.
in Chapter 1.3 and a similar exponential For ferrocene sites at the end of long
decay has been observed in tunneling alkanethiols self-organized at gold elec-
spectroscopy for tip-to-molecule distance trodes and diluted with unsubstituted
as described in Chapter 3.3. Ratner and thiols with the redox moiety in contact
Whitesides [30] have recently introduced a with the electrolyte (Fig. 4a), Chidsey has
simple experimental procedure to measure reported [34] curved Tafel plots (Fig. 4b),
which could be fitted by equations de-
rived from Marcus theory with values of
Fe λ = 0.85 eV and Z = 6.73 × 104 s−1 eV−1
for a reaction rate of k = 2.5 s−1 at E ◦ in

Fig. 4(b). Similar curvature in Tafel plots

has been reported by Faulkner and cowork-
ers [35] for adsorbed osmium complexes at
ultramicro-electrodes (UME). The temper-
ature dependence of the rate coefficient
could also be fitted from Marcus equation
and electron states in the metal and cou-
S S S
pling factors given by quantum mechanics.
S S
(a) AuAu Au Au Au Au Au AuAu At high driving forces in Fig. 4(b), a
plateau is reached instead of the in-
105 verted region predicted by Marcus theory
due to the continuum of states in the
104
Fig. 4 (a) Schematic representation of
103 thiolated gold with ferrocene covalently
kf + k b
[s−1]
attached to the alkylthiol end opposite

102 to the S-Au bond. (b) Log (kf + kb )
versus (E − Eo ) for the
oxidation–reduction of terminal
101 ferrocene sites at SAM modified gold
electrode (taken from Ref. [34]). The
100 dotted line corresponds to calculation
−1.0 −0.5 0.0 0.5 1.0
with the Butler–Volmer equation (29),
E − Eo the dashed line calculated curves from
(b) [V] Marcus–Levich–Dogonazde theory.
Fig. 5 Log of current density versus

potential curve of 0.1 M ZnSO4 at 0.6 Zn Hgx/Zn2+
atom % Zn amalgam electrode with
50 µM [(C4 H9 )4 N]2 SO4 . Current den-
sities were obtained from radiolabe-
led 65 Zn amalgam. (Taken from Ref. [8].)
−4
metal electrode unlike the inversion found

for electron-transfer rates between two
molecular states. The electronic states
at the metal surface are considered in
the Levich–Dogonadze theory [58], which
[a cm−2]
i−(calculated)
log i 
takes into account the electronic states −5
overlapping between the electrode and re-
dox electrolyte and also describes the tem- i (experimental)
perature dependence (see Chapter 1.2).
Gerischer developed a quantum mechan-
ics treatment for redox systems at metal
and semiconductor electrodes, where the
difference between metal and semicon- −6
ductor electrodes is the integration over i+ (from Zn55)
the electronic states in the electrode as
in the Levich–Dogonadze theory [59]. In −0.80 −0.75 −0.70 V −0.65
Gerischer model the electronic levels in eh
the redox electrolyte are oxidized species
(unoccupied states) and reduced species
(occupied states) with a Gaussian distri- relationship are interfacial ion processes
that can take place at
bution. The electronic overlap of these
redox states with the states in the semi- 1. solid–liquid interface like in metal
conductor or metal electrode describe the dissolution in liquid electrolyte
different contributions to the exchange 2. liquid–liquid interfaces like the inter-
current density and the current–potential faces between two immiscible elec-
dependence [3]. The current–potential re- trolyte solutions (ITIES).
lationship for electrode reactions at semi-
conductor electrodes present asymmetric During the dissolution of a metal for in-
curves due to the overlapping of the stance, metal cations belonging to the
electronic structure of the semiconductor metal crystallographic lattice can reach
(electron or hole density of states) and the the surface and become ad-atoms, a term
electronic states in the redox electrolyte. coined by Lorenz [38] to describe metal
atoms adsorbed on surfaces. These ad-
1.1.6 sorbed cations can loose electrons into
Ion-Transfer Reactions the metal surface electronic plasmon and
cross the electrical double layer to become
Electrode reactions also character- solvated cations in the electrolyte. A very
ized by exponential current–potential important role is the solvation process
14 Fundamentals

that stabilizes the cation in solution. Note, (1 − α)zF
= k− co exp E (43)
however, that there is no electron trans- RT
fer across the electrical double layer in
the dissolution of a metal. Likewise, at therefore the total current that flows at the
ITIES not only electrons can be transferred interface is given by
across the liquid–liquid electrolyte bound-
αzF
ary from donor redox states in one liquid i = k+ cM cT exp E
RT
phase to an acceptor redox states in the
other, and vice versa; but also ion trans- (1 − α)zF
− k− co exp E (44)
fer at ITIES is possible with resolvation of RT
the ion being transferred as described in
Chapter 4.2. The current–potential curve or defining an exchange current density,
is also exponential in that case and can be io and overpotential, η:
described in a similar way to ion transfer
αzF
at solid–liquid interfaces with surmount- i = io exp η
RT
ing an activation barrier by a Boltzman
population of species with enough energy. (1 − α)zF
− exp − η (45)
Gerischer [39] considered the process RT
described by
a typical ion-transfer electrode reaction
is the dissolution of zinc amalgam into
M(metal) −−−→ M z+
(solvated)
solvated Zn2+ cations in the aqueous
+ ze(metal) (41) electrolyte:
where a metal ion at the surface metal Zn(Hg) −−−→ Zn2+ (aq) + 2e(Hg) (46)
lattice loses an electron (which remains
in the metal) and crosses the double For this reaction a Tafel plot is shown in
layer to become a solvated cation in the Fig. 5 as predicted by Eq. (45). A kinetic
electrolyte. For this process an exponential analysis of this data yields for the trans-

current–potential curve is observed, which fer coefficient: α = RT /zF ∂ ln io /∂εo =
can be derived by considering the anodic 0.72 for the charge-transfer valence z = 2.
(i+ ) and cathodic (i− ) partial current The free-energy barrier for the flow of
densities proportional to the concentration ionic charge across the oxide–electrolyte
of the reacting species and to the Boltzman interface has an electrical contribution and
factor e−Ei/kT : consequently the current–potential curve
  can be described by a Butler–Volmer type
−
→ equation [8].
−Ei
i+ = k+ cM cT exp   The ion-transfer reactions at the ox-
RT
ide–electrolyte interface:

αzF
= k+ cM cT exp E (42) Mz+ (oxide) −−−→ Mz+ (electrolyte) (47)
RT
and and,
 
←−
− E O2− (oxide) + 2H+ (electrolyte) −−−→
i− = k− co exp  
i
RT H2 O(electrolyte) (48)
are also examples of interfacial ion-transfer reorganization in the metal-ligand vibra-

reactions. Under steady state conditions, tion scale, 10−13 s or slower, the probability
the net current at the interface is given of simultaneous electron transfer in a sin-
by the sum of cation and anion partial gle step is extremely low. Therefore, two
ionic currents. Deposition and dissolution one-electron steps with the stabilization
of the oxide requires that both cations and of an intermediate state are more likely
anions flow in the same direction while the to occur than a single two-electron step.
corresponding electrical currents are oppo- Figure 6 depicts the free-energy profile for
site [39]. The steady state potential of the a two-step electron-transfer electrode re-
oxide electrode at a particular composition action, where the relative heights of the
MOn lies between the formal equilibrium
activation barriers for the elementary steps
potential with respect to oxygen and metal
is differently influenced by the electrode
electrodes as required by the electroneu-
potential. At E1 , the rate-determining step
trality condition [40].
corresponds to the first electron transfer
The transfer of ions across immisci-
while at E2 , the rds is the second electron
ble electrolyte interfaces (ITIES) can be
described by a similar Butler–Volmer for- transfer and at E ∗ both transition states
malism and involves resolvation of the have the same activation barrier (k2 = k−1 )
transferred ions. The flux of ions can be and therefore there is a shift in the position
measured by the electronic current in the of the transition state from the first to the
external circuit generated at symmetrical second electron-transfer process.
reference electrodes reversible to one of Note that both rate constants k1 and k2
the ions. may have different kinetics with io,1 and α1
and io,2 and α2 respectively and a general
1.1.7 expression for a sequential two-electron
Multiple Electron Transfer reaction has been derived by Vetter [8]:
In general, reactions at electrodes involv-

α1 F η
ing more than one electron-transfer events i = 2io,1 exp
RT
occur in successive steps with participation
of intermediaries, which in most cases are 2F η
1 − exp −
adsorbed at the electrode surface. For in- RT
×
stance, consider io,1 (1 + α2 − α1 )F η
1+ exp −
k1 io,2 RT
H+ (aq) + e −−−→ Hads (metal) (51)
where η = E − E ◦ and E ◦ is the standard

k2
+ H+ (aq) + e −−−→ H2 (49) electrode potential for the overall two-
electron process. In Fig. 7 we represent
or, the Tafel plot for the current potential
k1 k2 relationship described by Eq. (51).
Cuaq 2+ + e −−−→ Cuads + + e −−−→ Cu We can consider two limiting cases:
(50)
Since the time scale for the electron- 1. The first electron-transfer step is rds,
transfer event, ca. 10−16 s is much smaller then k2 k−1 in Eq. (51), and the
than the time scale of the fastest chemical forward rate coefficient kf ≈ k1 and
16 Fundamentals
=/ 1 Fig. 6 Standard free-energy diagram

for a two-step electron-transfer
E1 electrode reaction with two potential
dependent transition states.
The relative height of each individual

activation barrier changes with electrode
potential, as illustrated in Fig. 6.
As depicted in Fig. 7, at potentials more
E* positive than E ∗ , k2 > k−1 and the first
electron transfer is rds while for E < E ∗ ,
k−1 > k2 and the second electron transfer
becomes rate-determining. The Tafel slope
=/ 2 in the first case is 2RT/F and 3RT /2F for
the preequilibrium in the second case.
In multiple-electron electrode reactions
E2 extrapolation of Tafel lines from high
overpotential regions to reversible poten-
tial for the overall reaction Eo , does not
yield the same apparent standard rate co-
efficient (exchange current density) and
the backward rate coefficient kb ≈ may be misleading: io,2 is larger than
(k2 /k−2 )/k−1 . io,1 , the exchange current density for the
2. Second electron-transfer step in kinetics that change more rapidly with
Eq. (51) being rds. k2 k−1 then kf ≈ potential. This feature distinguishes an
(k1 /k−1 )k2 and, kb ≈ k−2 . n-electron sequential multistep reaction
from a single n-electron charge-transfer
1 2 step [8].
k−1 k2 The rate-determining step, rds, for
a given mechanism may change with
k2 = k1 potential, but this does not violate the
k−2 principle of microscopic reversibility since
inspection of Fig. 7 shows that, at any
i0.2
k1 potential, the transition state is always
i0.1 the same for the forward and backward
Log i
reactions in Eq. (51). Transition states at

different electrode potentials, on the other
hand, need not be the same (see Fig. 6).
i0 Furthermore, there is no reason, why the
potential E ∗ , determined by kinetic factors,
Fig. 7 Schematic representation of the

variation of the individual rate
coefficients ki (broken lines) and the
overall rate coefficient (solid line) for a
Eo2′ E*E o′ Eo1′
two-electron electrode reaction as a
E function of the electrode potential.
at which the transition of rds occurs, The reactants and products in the rate-
should be the same as the thermodynamic determining step (rds) will be identified by
equilibrium potential Ee . R and P , respectively.
In Fig. 7, each transition state can be We shall assume that the completion of
identified and is characterized by io,i , Eo,j the overall process represented by Eq. (52)
and αj ; io,i measures the barrier height requires the formation and decomposition
of each transition state in the respective of ν identical activated complexes.
standard conditions, whereas at the formal According to Parsons the rate-deter-
standard potential, E ◦ , for the overall

mining step can be represented [46] by
reaction, kf = kb and this potential value is
a b n c d
half way between E1◦ and E2◦ . The change

A + B + .... e = C + D + ....
in transition states with electrode potential ν ν ν ν ν
(53)
is observed, for instance, in the complex The stoichiometric number indicates the
oxygen electroreduction (4-electron and 4- number of times the rate-determining
proton) reaction, the mechanism of which step occurs in the overall stoichiometric
may also change with pH. reaction. For example, the Tafel mecha-
nism for the reduction of H+ on metals
1.1.7.1 Stoichiometric Number occurs in two steps with Reaction (54) oc-
The concept of stoichiometric number curring twice for each time Reaction (56)
was introduced in electrochemistry by takes place.
Horiuti and Ikusima [42, 43] for the
2 H+ + e −−−→ Had (54)
hydrogen electroreduction reaction. We
need to introduce the stoichiometric num- 2Hads −−−→ H2 (55)
ber ν in complex multielectron elec-
trode kinetics in order to distinguish where Had indicates an adsorbed hydrogen
different possible mechanisms. The In- atom (ad-atom) on the metal electrode.
ternational Union of Pure and Ap- If the overall reaction is written in
plied Chemistry (IUPAC) defines the the form:
stoichiometric number in electrochem-
2H+ + 2e = H2 (56)
istry as a positive integer that indi-
cates the number of identical activated ν = 2 and n = 2; but if the reaction is
complexes formed and destroyed in the represented as
completion of the overall reaction as
formulated with the charge number, 1
H+ + e = H2 (57)
n [44, 45] 2
A general expression can be derived for
then, ν = 1 and n = 1.
the electrode reaction:
It is worthwhile noticing that, while ν
and n are arbitrary quantities that depend
aA + bB + . . . .ne = cC + dD + . . . . on the way we write the stoichiometric
(52) equation, the ‘‘electron number’’ n/ν is
which takes place in several consecutive characteristic of the electrode reaction
intermediate steps, the rate of one of kinetics. In the Tafel hydrogen reduction
them controlling the overall kinetics. mechanism, n/ν = 1 and indicates the
18 Fundamentals

=
number of electrons transferred for the kB T Gb (1 − α) nF E
completion of the rate-determining step. kf = exp − +
h RT ν RT
The forward and backward reaction rate
coefficients can be expressed by Eq. (17) (64)

=
with G
= , the standard free energy of where Gi represents the nonelectrical
formation of one mole of activated complex contribution to the standard free energy of
from reactants in Eq. (52), and should be activation and E = (φM − φS ) + const. It
replaced by the electrochemical free energy should be noted that the right-hand side of

= Eqs. (63 and 64) should be divided by the
of activation, Gi , for charged particles.
activity coefficient of the activated complex,

=
Gf = µ◦± − µ◦R (58) which could be potential dependent [46].
Introducing the overpotential, η = E −

=
Gb = µ◦± − µ◦P (59) Ee , and replacing

The standard electrochemical potentials of α nF η
i = io exp −
reactants and products may be written ν RT

in terms of the chemical and electrical (1 − α)nη
contributions: − exp (65)
ν
a ◦ b ◦
µ◦R = µA + µB and the exchange current density, io
ν ν
n ◦ n n 1−α 1−α α α
+ µe + F φM (60) io = F ko [A] ν [B] ν [C] ν [D] ν (66)
ν ν ν
c ◦ d ◦ n for a first-order reaction.
µ◦P = µC + µD + F φS (61)
ν ν ν For |η| νRT /αnF , the Tafel approxi-
mation is valid and
In order to evaluate the potential depen-
dence of the free energy of activation, ∂η νRT
=−
without knowledge of the structure of the ∂ ln i T,P,ci αnF
activated complex, it is assumed that the
and
electrical contribution to the standard free
energy of the transition state lies between ∂η νRT
= (67)
that of the standard free energy of P and ∂ ln i T,P,ci (1 − α)nF
that to the standard free energy of R in
and therefore the stoichiometric number,
the rds. The symmetry factor or transfer
ν, can be obtained from anodic and
coefficient, α, for the rds in analogy to the
cathodic Tafel plots,
case for one-electron reactions is given by
−1
◦ α nF ∂ ln i ∂ ln i
µ◦± = µ± + nF E (62) ν=
RT ∂E a
−
∂E c
ν
(68)
which after substitution in Eq. (17) with if the rate-determining step is the same
the free energy of activation given by over the whole range of potentials from
Eqs. (58 and 59) is which Tafel slopes are obtained. Note

= that from a Tafel plot we can only
kB T Gf α nF E determine the product αn/ν, not the
kb = exp − + (63)
h RT ν RT quantities separately.
For small overpotentials, |η| νRT / acquire a component of their velocity along
αnF in the linear polarization region, an the direction of the field. These effects
explicit expression for the stoichiometric will be treated in detail in Chapter 2.1:
number can be found: Diffusion and migration. However, in
electrokinetic experiments, the migration
nF io ∂η nF io ∂η contribution can be reduced by using an in-
ν=− =
RT ∂ic η=0 RT ∂ia η=0 ert supporting electrolyte at a much higher
(69) concentration than that of the electroactive
which is valid if the reversible potential is species so that the transport number of the
far from the pzc, otherwise double-layer electroactive species is negligible. Forced
effects are important (see Sect. 1.1.10). convection combined with diffusion can be
controlled with hydrodynamic electrodes
1.1.8 with a good mathematical description of
Mass Transport at Electrodes the flow pattern and thus of the steady
state current–potential relation. [9].
In electrode reactions, the reactant has to
∂c
find the electrode surface where electrons = D∇ 2 c − Vx ∇c (71)
are taken or released, and therefore ∂t
the mass transport of reactants and The solution of this partial differential
products becomes very important in the second-order equation depends on the
description of electrode reactions where initial and boundary conditions of the
we will be interested in current–time and particular experiment, giving rise to a
current–potential relations. The master multitude of techniques. In Chapter 2.2,
equation for mass transport to an electrode the digital simulation of voltammetry un-
surface is the Nernst–Planck equation: der stagnant and hydrodynamic conditions
is described. By changing the electrode
−→ −
→ zi F −
→ −
→ potential one can modify the boundary
Ji = −Di ∇ ci + ci ∇ φ + ci ν
RT conditions and transient effects arise until
(70) a new steady state is reached.
which accounts for the flux of species i The steady state diffusion is reached
(mol cm−2 s−1 ) at a given distance from when
the electrode surface, Di is the diffusion ∂c
=0 (72)
coefficient of the diffusing electroactive ∂t
species (cm−2 s−1 ). The first term on the This condition can be accomplished when
right hand side of Eq. (70) represents the the concentration gradient close to the
transport mechanism by diffusion in a electrode surface is constant (linear) or
−→
concentration gradient, ∇ ci , by random its curvature is compensated by a term X:
thermal motion of the particles, which
∂c
tends towards uniform concentration of = D∇ 2 c ± X (73)
mobile species. The second and third term ∂t
on the right hand side describe, respec- where X can arise from the diffusion field
tively, migration under the influence of the geometry, that is, nonlinear terms like in
−
→ − → cylindrical or spherical electrodes:
electric field, E = ∇ φ, and convection
under the influence of a flow field, with the 2 ∂c
fluid velocity −→
ν (cm s−1 ) and the particles Xspherical = D
r ∂r
(74)
20 Fundamentals
or, is to be determined from the limiting

1 ∂c
Xcylindrical = D (75) convective-diffusion current.
r ∂r According to the classical treatment by
or approximations to these terms in ultra Randles [47] for a simple electroreduction
microelectrodes where the electrode size is of O to R in solution and assuming,
smaller or comparable to the characteristic for simplicity, that the mass transport
diffusion length at a given time, namely, rate coefficients, kd,i for the oxidized
√ and reduced species are the same, the
r ≤ (2Dt).
Two other sources of a term X that net current density under steady state
can compensate for the curvature of the conditions is
concentration gradient at the electrode
i = nF kd co∗ − cos = nF kd cRs − cR∗
surface are a convective term X = ci − →
ν,
−
→ (79)
where ν is the fluid velocity or a chemical
with the diffusion-limiting current densi-
reaction (not involving a charge-transfer
ties for the anodic and cathodic processes
step) term X = k1 ci coupled to the charge-
given by.
transfer reaction, that is,
iL = nF kd ci∗ (80)
k1
− +
AH −
←−−
−−
→−A +H (76) A diffusion layer of thickness δ, which
k−1 has a purely formal significance, can be
defined as
followed by the electron-transfer reaction:
D (nF DAci )
1 δ= = (81)
+ Kd,i i
H (aq) + e −−−→ H2 (g) (77)
2
In Eq. (25) we have considered the surface
The diffusion equation is concentration of the reacting species equal
to the analytical concentration in the bulk
∂cAH ∂ 2 CH electrolyte. If we replace in Eq. (25) the
= DH − k1 cHA + k−1 cA cH+
∂t ∂x 2 surface concentrations from Eqs. (79–81):
(78)
which has a steady state at the electrode
ci (0, t) i
characterized by a reaction layer, xR = = 1− (82)
√ ci∗ iL
(DH /k−1 cA ), a distance over which a
proton once formed from dissociation of we obtain
HA has a chance to diffuse before reacting
i i αnF η
with A− [4]. = 1− exp −
io iL , c RT
In most practical applications, where the
maximum yield of a product or electricity i (1 − α)nF η
− 1− exp
in electrochemical energy conversion sys- iL , a RT
tems is desired, the rate of mass transport (83)
should be fast enough in order not to limit Figure 8 depicts a plot of the cur-
the overall rate of the process. Conversely, rent–potential curve calculated with
in electroanalytical techniques, such as po- Eq. (83), i/iL versus η and the correspond-
larography or gas sensing, the reaction ing normalized concentration profiles at
is limited by the transport of the reac- the limiting current and at equilibrium.
tant since the analyte bulk concentration The mass transport free Butler–Volmer
C0/C*0
i/iL 1
Butler–Volmer
1 X
id,a
0 h
C0/C*0
1
id, c
−1
X
C0/C*0
X
Fig. 8 Plot of i/iL versus η calculated with Eq. (83) (solid line) and
Butler–Volmer current–potential curve calculated with Eq. (29) for
α = 0.5 and n = 1. Inset: concentration profiles of the oxidized species
normalized to the analytical concentration at limiting current and at
equilibrium.
predicted current–potential curve calcu- potential, kobs :

lated with Eq. (29) (dashed line) is shown
i
for comparison. kobs = (86)
nF ci∗
We can reexamine Eq. (83), both in the
Tafel region, for |η| RT /αnF : can be separated into the electrode reaction
apparent rate coefficient for extrapolated
i αnF η infinitely fast mass transport conditions, k,
log i = log io + 1 − exp −
iL,c RT and the mass transport rate coefficient, kd :
(84)
1 1 1
and for the linear polarization region, for = + (87)
|η| RT/αnF: kobs k kd
  Heterogeneous electrode reactions can be

compared with homogeneous kinetics in
nF η 
 1

i=  solution, with regard to mass transport.
RT  1 1 1  The second-order rate coefficient for a
+ +
io iL,c iL,a fast homogeneous reactions in solution,
(85) k(hom), which would be observed if dif-
which in the absence of concentration fusion were infinitely fast, can be related
polarization is coincident with Eq. (32). to the measured rate coefficient, kobs (hom)
Expressing Eq. (84) in terms of rate by application of Fick’s first law in a spher-
coefficient rather than current densities ical continuum diffusion field around
and rearranging terms, the rate coeffi- the reacting molecule. At a collision dis-
cient obtained experimentally at constant tance rAB , this corresponds to the average
22 Fundamentals
concentration of molecules undergoing defined by

encounters equal to the average analytical
concentration in the bulk solution [48]. ∂ log ij
pk = (90)
∂ log ck ci
=k ,E
4πrAB DAB
kobs (hom) = (88) where ij is a partial anodic or cathodic
4rAB DAB
1+ current density and ck the concentration
k(hom)
of species k.
where rAB is the encounter distance of Because of the strong dependence of the
molecules A and B in solution and DAB is kinetics on potential, the determination
the relative diffusion coefficient. of the electrochemical reaction order
Rewriting Eq. (88) with kd (hom) = requires that the partial anodic or cathodic
4πrAB DAB NA where NA is Avogadro’s current densities be measured at constant
number: potential in addition to the concentration
of the other species being kept constant.
1 1 1 Vetter used extensively the concept of
= + (89) reaction order, for instance, as diagnostic
kobs (hom) k(hom) kd (hom)
of the mechanism of the electrode reaction:
The analogy in the mass transport effects
in electrode reaction and homogeneous Mn(IV) + e −−−→ Mn(III) (91)
second-order fast reactions in solution
becomes clear. In electrode kinetics, how- on platinum electrode [8], which looks
ever, the charge-transfer rate coefficient simple at first sight. However the reaction
orders found experimentally are PMn(IV) =
can be externally varied over many orders
0 and PMn(III) = 1, which suggests the
of magnitude through the electrode poten-
following mechanism:
tial and kd can be controlled by means of
hydrodynamic electrodes. For instance the
Mn(III) + e −−−→ Mn(II)
mass transport rate coefficient, kd , for a ro-
tating disc electrode at the maximum prac- Mn(II) + Mn(IV) −−−→ 2Mn(III)
tical rotation speed of 10 000 per min−1 is (92)
approximately 2 × 10−2 cm s−1 . with the first electron transfer to Mn(III)
rate-determining step.
1.1.9 For the HER on mercury the reaction
Electrochemical Reaction Order order PH + = 1 at pH less than 8 and
PH + = 0 at pH larger than 9, allows
The order of an electrochemical reaction is the following elementary steps to be
an empirical factor widely used in chemical distinguished:
kinetics, which may give an insight into the
events at the molecular level for kinetics of Haq + + e −−−→ Hads at pH < 8 (93)
complex reactions. It relates the reaction
rate to the concentration of a particular and
species in the kinetic equation.
In electrode reactions, the reaction order H2 O + e −−−→ Hads + HO− at pH > 9
with respect to the species k can be (94)
1.1.10 are maximum. For the electroreduction

Double-Layer Effects on the reaction:
Current–potential Relationship
OzOo + ne −−−→ RzRR (95)
Since the electrode reactions occur
at charged interfaces the structure of two effects due to the double-layer struc-
this electrified interface influences the ture can be described:
electrode kinetics of charged species. 1. the concentration of the reacting ion,
Frumkin [49] pointed out the importance p
CO in the outer Helmholtz plane, is dif-
of the double-layer structure on electrode ferent from the analytical concentration
kinetics for the hydrogen reduction S,
in the solution, CO
reaction on mercury. The electrical

potential at the plane where the p S zi Fφ2
CO = CO exp − (96)
reactant undergoes electron transfer RT
to become a product influence the
concentration of charged species different The electrical work to place a mole of
from the analytical concentration the O from the bulk solution in the OHP is
bulk electrolyte due to the electrostatic zi Fφ2 where φ2 is the potential of the
work required for the charged reactant OHP with respect to the potential of the
(product) to reach the electrode surface. solution.
Furthermore, the inner (Galvani) potential 2. The driving force for the electrochem-
difference with respect to the solution ical reaction is the potential jump
(φp –φS ) is less than the applied electrode- between the metal and the reaction site
electrolyte potential difference (φM –φS ). (in the absence of specific adsorption,
The electrostatic work necessary to identified with the OHP). The appar-
bring the electroactive ion to a charged ent rate coefficient for the reduction
interface, where the electrode reaction reaction involving n electrons is
occurs, is equivalent to the electrostatic φ
kapp = kφM =φ2 CO2
contribution to the free energy of activation
in homogeneous reactions between ions αnF (φM − φ2 )
× exp − (97)
in solution. When two ions of the same RT
charge react, their kinetics are slower as
compared to the uncharged species and By combining Eqs. (96 and 97) and replac-
conversely when two ions of different ing the concentration of O at the plane of
charge react, the apparent kinetics are reaction, the apparent rate constant can be
faster than for the uncharged molecules. expressed in terms of the applied electrode
When no specific adsorption occurs, the potential.
closest the reactant can get to the electrode
S αnF (φM − φS )
surface is assumed to be the outer kapp = kφM =φ2 CO exp −
RT
Helmholtz plane (OHP) and the reaction
plane is identified with this OHP, φp ≈ φ2 . (αn − z)F (φ2 − φS )
× exp
The simplest case is when the reactant is RT
the only ionic component of the solution, (98)
then the effects on electrode kinetics The first exponential term shows the
due to the properties of the interface potential dependence of the apparent
24 Fundamentals
rate constant on the applied poten- Nonlinear Tafel plots are predicted by
tial difference (metal–electrolyte solution), Eq. (101) when φ2 changes appreciably
(φM – φS ). The second exponential term is with electrode potential: at low ionic
the double-layer correction to the apparent concentrations and close to the pzc.
rate constant due to concentration at the The apparent electrochemical reaction
reaction plane and potential φ2 . It is also order is influenced by double-layer effects:
shown that the same correction applies for
the reverse reaction for zr = zo ± n since app ∂ ln iapp
PO = = PO
the transition state and hence the reaction ∂ ln cO E
plane must be the same for the forward
F ∂φ2
and backward processes. + (αn − zo )
It should be noticed that the Galvani RT ∂ ln CO E
potential difference, φ2 , cannot be mea- (102)
sured directly but its value can be derived the double-layer correction in the second
from the Gouy–Chapman theory of the term of Eq. (102) due to interfacial poten-
diffuse double layer. If the excess charge tial distribution is zero at the pzc.
in the metal, q m , is determined experi-
mentally, that is, from the integration of 1.1.11
differential capacity curves, then for a 1 : 1 Adsorption and Electrocatalysis
electrolyte of concentration cs , then φ2
can be calculated: Adsorption of reactants, intermediates and
m products at electrode surfaces significantly
2RT q (2π)1/2
φ2 = sinh−1 influence the current–potential relation-
zi F (2RT cs ε)1/2 ship for electrode reactions. In particular
(99) species confined to electrode surfaces can
The double-layer correction of the observed be oxidized and reduced without the need
electrode kinetics in Eq. (98) is more for the reactants to diffuse towards the
important at low ionic concentration, high
electrode surface. The adsorption free en-
ionic charge and close to the potential of
ergy plays a key role in electrode kinetics
zero charge (pzc). This correction can be
and for charged species on the potential
minimized by using a supporting or inert
dependence.
electrolyte of high concentration so that
An electrocatalytic reaction is an elec-
most of the potential drops operates in the
trode reaction sensitive to the properties
inner Helmholtz plane and φ2 ≈ 0.
of the electrode surface. An electrocatalyst
The consequence of double-layer cor-
participates in promoting or suppressing
rection on the exchange current density
an electrode reaction or reaction path with-
is
out itself being transformed. For example,
(αn − z)Fφ2 oxygen reduction electrode kinetics are
io = io (app) exp −
RT enhanced by some five orders of mag-
(100) nitude from iron to platinum in alkaline
and on the apparent charge-transfer coef- solutions or from bare carbon to carbon
ficient, electrodes modified with adsorbed iron or
∂ ln iapp F ∂φ2 cobalt phthalocyanines or porphyrins and
αapp = = (αn − zo ) when certain metals are under potential
∂E RT ∂E
(101) deposited (upd).
As early as 1905, Tafel [11] showed reduction and CO and methanol oxida-
that the kinetics of the proton reduction tion. In particular, Chapter 5.1 analyzes
strongly depends on the nature of the electrocatalysis in depth.
electrode material. Thirty years later,
Horiuti and lkusima used the word 1.1.12
‘‘catalyst’’ instead of electrode [42], and Mixed Potential
Butler pointed out the very important role
of adsorbed intermediates in the evolution At the equilibrium potential, both anodic
of hydrogen [12, 13]. and cathodic processes of a single electron-
The main catalytic influence of the na- transfer reaction take place at the same
ture of the electrode material is through exchange rate (a measure of which is
the adsorption of intermediates of com- io ) and the current through the external
plex electrode reactions. Hortiuti and circuit is zero. The open-circuit potential is
Polanyi [50] suggested that the activation the thermodynamic equilibrium potential.
energy of an electrode reaction should When more than one reaction can take
be related to the heat of adsorption of place at the electrode surface, on the
adsorbed intermediates by a relationship other hand, the open-circuit potential
of the form of the Brönsted rule in ho- is not a thermodynamic quantity but a
mogeneous solutions. This corresponds ‘‘mixed potential’’ set up by the kinetics
to a vertical shift of the potential energy of both simultaneous electrode reactions
curves by an amount βH◦ads with β a that exchange electrons with the metal.
symmetry factor. Correlation of the acti- At open circuit all the charge transferred
vation energy or overpotential at a given from the electrode to the solution must be
current for a particular rds on different counterbalanced by the flow of charge in
substrates with relative measures of the the opposite direction (Ia = Ic ) with no net
heat of chemisorption of intermediates
flow of charge through the interface. Thus,
have been successfully made for hydro-
anodic and cathodic processes are coupled
gen evolution [51], oxygen evolution [52],
through the charge exchanged and a net
and oxygen reduction [53], and so on. For
chemical reaction proceeds.
the hydrogen electrode reaction (HER) on
In order to explain the corrosion process
different metals, it is predicted that the
of metals, Wagner and Traud [54] devel-
curve log io vs. G◦ads goes through a
oped the ‘‘mixed potential’’ theory, which
maximum with decreasing log io both at
assumes that the current–potential rela-
positive and negative values of G◦ads due
tionship is given by
to the opposite effects of the free energy of
adsorption on the geometric and exponen-
tial factors in the electrokinetic equation I (E) = |Ia,i (E)| − |Ic,i (E)| = 0
(volcano plots). (103)
Several chapters in this volume describe The additive individual component cur-
new experimental evidence of adsorbate rents are based on the simplifying assump-
bonding with different ‘‘in situ’’ spectro- tion that the anodic and cathodic processes
scopic techniques, the effect of the interfa- are statistically independent and that the
cial structure on kinetics and mechanisms electrode surface sites are indistinguish-
of selected electrode reactions with bond able for both reactions. Since the total
breaking and surface bonding like oxygen anodic and cathodic currents equal at the
26 Fundamentals
O → R + n e− O′ → R′ + n′ e−
ja
Ee′1 Ee′2
EM
jc
R + n e− → O R′ + n′ e− → O′
Fig. 9 Schematic representation of the establishment of mixed

potential, EM , at an electrode. Partial current densities (solid lines) and
total current density curve (broken line).
open-circuit or mixed potential, EM properties; and consequently the mixed

potential EM may be susceptible to those
|ia,i |Aa,i = |ic,i |Ac,i (104) kinetic factors, unlike the equilibrium po-
tential of each partial electrode reaction,
where ik,i represents the partial anodic which is fixed by thermodynamics and
and cathodic current densities and Ak,i the activities in the bulk solution. Unlike
the respective reaction areas. consecutive electron-transfer reactions the
In Fig. 9 we consider two simultane- kinetics of which are determined by the
ous electrode reactions (O/R and O /R ) slowest process, mixed potentials are de-
with transfer of n and n electrons, respec- termined by the fastest of several possible
tively and formal equilibrium potentials electrode reactions.

E1e and E2e . At open circuit, the mixed If one of the partial electrode reactions
potential, EM , adopts intermediate values is the dissolution of the electrode (i.e.
E1e < EM < E2e , closer to the equilibrium metal or semiconductor) the open-circuit
potential of the faster partial reaction. The potential is a corrosion potential and the
overall current–potential curve is indi- system undergoes corrosion at a level
cated by the broken line. Notice that EM given by the corrosion current (icorr ),
is not determined by the thermodynamic which measures the dissolution rate. The
values of E1e and E2e but by the kinet- magnitude of icorr for a corroding system
ics of the respective reactions, that is, by may vary significantly over several orders
the respective anodic and cathodic compo- of magnitude, like the exchange current
nent curves in Fig. 9 with the condition of densities of different individual electrode
Eq. (104). These curves may be altered by reactions span a range of orders of
mass transport, surface area and specific magnitude.
In Fig. 9, the occurrence of two simul- due to the limited solubility of oxygen
taneous redox reactions at an electrode in water. At even more negative poten-
surface has been considered; however, in tials, hydrogen evolution takes place on
most corrosion problems, more than two the electrode surface. The cathodic reduc-
reactions may take place and both for- tion of some dissolved metal ions in the
ward and backward individual electrode solution adjacent to the electrode as a con-
reactions may not take place at signifi- sequence of corrosion is also considered
cant rates in the potential range where the in Fig. 10 but the rate of oxygen reduction
mixed potential is observed, owing to the is the controlling factor in the overall rate.
slow kinetics of the participating reactions The measurable current–potential curve
under those conditions. results from the balance of the anodic and
Figure 10 illustrates the corrosion of two cathodic processes as indicated by the bro-
metals, M and M , in aqueous aerated (oxy- ken line and the corrosion potential, EM ,
genated) solutions. Three anodic partial is also shown. Metal M is more noble than
reactions are considered: the active dis- M and is not corroded because the oxygen
solution of two metals M and M with reduction reaction is far too slow to bring
different kinetics in the absence of their M to a potential where it can be corroded.
ions in the bulk electrolyte and decomposi- Hydrogen evolution is a faster reaction but
tion of water with the evolution of oxygen. cannot corrode M because its equilibrium
The kinetics of the latter process is so potential is more negative than EM e and
slow on most corroding metals that only therefore the reaction cannot proceed on
at very negative potentials can dissolved M under those conditions.
oxygen be electroreduced and this eventu- It is well known that the reduction of
ally becomes limited by mass transport organic molecules such as nitrobenzene
M′ → M′Z+ + z′ e−
M → MZ+ + z e−
jcorr
Eo′1 Eo′2 Eo′4 Eo′3

Ecorr
H2O → O2 + 4H+ + 4e−
MZ+ +z e− →M jcorr
2H+ + 2e → H2
O2 + 4e + 4H+ → 2H2O
Fig. 10 Representation of the formation of the corrosion

potential, Ecorr , by simultaneous occurrence of metal
dissolution, hydrogen evolution, and oxygen reduction. Partial
current densities (solid lines) and total current density curve
(broken line).
28 Fundamentals
with metals in acid leads to different final nuclei and oxidation of organic and inor-
products depending on the nature of the ganic reductants. A quantitative treatment
metal employed and the composition of has been derived by Spiro [56] and practi-
the electrolyte. The kinetics of hydrogen cal aspects have been recently reviewed by
evolution, either from proton or from O’Sullivan [57].
water, as well as the kinetics of dissolution
of some metals, particularly those like Zn 1.1.13
that can be complexed by HO− , strongly Electrocrystallization
depend on the pH and metal surface; the
mixed potential is then fixed by the kinetics The formation or dissolution of a new
of hydrogen reduction and dissolution phase during an electrode reaction such as
of the metal, which are relatively fast metal deposition, anodic oxide formation,
reactions. Organic electroreductions are precipitation of an insoluble salt, and so
slower processes and not likely to be on involves surface processes other than
potential-determining reactions, but they charge transfer. For example, the incorpo-
proceed through different reaction paths ration of a deposited metal atom into a
at different potentials depending on the stable surface lattice site introduces an
mixed potential of the metal surface. extra hindrance to the flow of electric
The theory of mixed potentials is based charge at the electrode–solution interface
on the independence of partial anodic and therefore the kinetics of these elec-
and cathodic reactions. However, in real trocrystallization processes are reflected in
corroding systems, coupling of these the overall electrode reaction.
reactions, not only through exchanged During the initial stages of phase for-
charge at the interface, may sometimes mation at electrodes (nucleation) incor-
occur as a result of (1) geometric surface poration of an adsorbed species into the
limitations due to the blocking effect surface lattice occurs only at ‘‘active sites’’
of the intermediates of one reaction (crystal plane edges or kinks) where in-
on the kinetics of others taking place teractions with other species already in
simultaneously at the same electrode the lattice are possible. The process of
surface and (2) changes in local pH formation of such nuclei may become ki-
due to the fact that some corrosion netically limiting and supersaturation or
reactions such as hydrogen evolution may overpotential is required for nuclei to be
promote changes in the kinetics of other formed and survive. Depending on the rel-
participating reactions. ative time scale of nuclei formation and
Electroless deposition of metals like further growth or thickening of the de-
nickel-phosphorus, introduced in 1946 posit local potential distribution and mass
by Brenner and Riddell [55], is of great transport conditions determine the mor-
industrial importance and allows to de- phology.
posit metals on nonconducting substrates If the electrocrystallization is con-
such as printed circuit boards (copper), trolled by the formation of two or three-
contacts (silver), and so on. Its mecha- dimensional isolated nuclei, the cur-
nism is based on a particular case of rent–potential relationship has a stronger
mixed-potential where the partial electrode overpotential dependence than predicted
reactions are metal deposition at metal by the Butler–Volmer equation [58],
that is, 5. J. O’M. Bockris, K. N. Reddy, Modern Electro-

B chemistry, Plenum Press, New York, 1970.
i = A exp (105) 6. B. E. Conway, Theory and Principles of Elec-
ηn
trode Processes, Ronald Press, New York, 1964.
with n = 2 for three-dimensional and n = 7. P. Delahay, Double Layer and Electrode Kinet-
1 for two-dimensional nucleation. Growth ics, 2nd edn., Interscience Publishers, New
York, 1966.
of isolated nuclei at an electrode surface
8. K. J. Vetter, Electrochemical Kinetics (English
is eventually limited when they start to Translation), Academic Press, New York,
coalesce due to their number and size and 1967.
the size of the electrode area. 9. C. M. A. Brett, A. M. Oliveira Brett, Electro-
In metal deposition, depending on the chemistry: Principles, Methods and Applica-
binding energy of the metal to be de- tions, Oxford University Press, Oxford, 1993.
10. E. J. Calvo in Comprehensive Chemical Kinet-
posited with respect to the foreign metal
ics (Eds.: C. H. Bamford, R. G. Compton),
substrate, two limiting cases have been Elsevier, Amsterdam, 1986, Vol. 26.
observed: (1) formation of structured un- 11. J. Tafel, Z. Phys. Chem. 1905, 50A, 641.
derpotential deposited layers (upd) and 12. J. A. V. Butler, Trans. Faraday Soc. 1932, 28,
further growth of the bulk metal phase 379.
on top of the two-dimension upd layer 13. J. A. V. Butler, Proc. R. Soc. London, Ser. A
1936, 157, 423.
or (2) deposition in the overpotential re- 14. T. Erdey-Gruz, M. Volmer, Z. Phys. Chem.
gion involving relatively few nuclei by Abt. A 1931, 157, 165.
three-dimensional nucleation and further 15. R. A. Marcus (Nobel Lecuture), Angew.
growth. The influence of upd ad-adatoms Chem., Int. Ed. Engl. 1993, 32, 1111.
on electrode kinetics is treated in Chap- 16. R. A. Marcus, J. Chem. Phys. 1956, 24, 966.
ter 4.3 and electrochemical nucleation and 17. R. A. Marcus, Discuss. Faraday Soc. 1960, 29,
21.
growth is described in Chapter 5.3. 18. R. A. Marcus, J. Phys. Chem. 1963, 67, 853.
19. R. Marcus, J. Chem. Phys., 1965, 43679.
Acknowledgments 20. R. Marcus, J. Phys. Chern. 1968, 72, 891.
21. J. R. Miller, L. T. Calcaterra, G. L. Closs, J.
Am. Chem. Soc. 1994, 106, 3047.
The author wishes to acknowledge finan- 22. A. L. Barker, J. V. Macpherson, C. J. Slevin
cial support during the preparation of et al., J. Phys. Chem. B 1998, 102, 1586.
this chapter from Conicet, ANPCyT (Ar- 23. A. L. Barker, P. R. Unwin, S. Amemiya et al.,
gentina), Fudetec, Motorola SPS, and the J. Phys. Chem. B 1999, 103, 7260.
Guggenheim Foundation for a fellowship 24. M. Tsionsky, A. J. Bard, M. V. Mirkin, J.
Phys. Chem. 1996, 100, 17 881.
2000/2001. 25. M. Tsionsky, A. J. Bard, M. V. Mirkin, J. Am.
Chem. Soc. 1997, 119, 10 785.
References 26. C. J. Miller in Physical Electrochemistry, Prin-
ciples, Methods and Applications (Ed.: I. Rubin-
1. E. Gileadi, M. Urbakh, (Eds.), Thermodynam- stein), Marcel Dekker, New York, 1995,
ics and Electrified Interfaces, Vol. 1. EoE, Chap. 2.
Wiley-VCH, Weinheim, Germany, 2002. 27. H. O. Finklea, Electroanal. Chem. 1996, 19,
2. A. J. Bard, L. R. Faulkner, Electrochemical 109.
Methods. Fundamentals and Applications, 2nd 28. C. E. D. Chidsey, J. Am. Chem. Soc. 1997,
ed., Wiley, New York, 2001. 119, 10 563.
3. H. Gerischer, D. M. Kolb, J. K. Saas, Adv. 29. S. Creager, S. J. Yu, D. Bamdad et al., J. Am.
Phys. 1978, 27, 437. Chem. Soc. 1999, 121, 1059.
4. W. J. Albery, Electrode Kinetics, Oxford Uni- 30. R. Holmlin, R. F. Ismagilov, R. Haag et al.,
versity Press, Clarendon Press, Oxford, 1975. Angew. Chem., Int. Ed. 2001, 40, 2316.
30 Fundamentals
31. D. I. Gittins, D. Bethell, D. J. Schiffrin et al., 46. R. Parsons, Trans. Faraday Soc. 1951, 47,
Nature 2000, 408, 67. 1332.
32. X. D. Cui, A. Primak, X. Zarate, J. Tomfohr, 47. J. E. B. Randles, Can. J. Chem. 1959, 37, 238.
O. F. Sankey, A. L. Moore, T. A. Moore, 48. J. D. Clark, R. P. Wayne in Comprehensive
D. Gust, G. Harris, S. M. Lindsay, Science Chemical Kinetics (Eds.: C. H. Bamford,
2001, 294, 571. C. F. H. Tipper), Elsevier, Amsterdam, 1969,
33. F. F. Fan, J. Y. S. M. Dirk, D. W. Price et al., Vol. 2, Chap. 4, 377–462.
J. Am. Chem. Soc. 2001, 123, 2454. 49. A. N. Frumkin, Z. Phys. Chern. 1933, 164,
34. C. E. D. Chidsey, Science 1991, 251, 919. 121.
35. R. J. Forster, L. R. Faulkner, J. Am. Chem. 50. J. Horiuti, M. Polanyi, Acta Physicochim.
Soc. 1994, 116, 5444. URSS 1935, 2, 505.
36. V. G. Levich in Physical Chemistry. An Ad- 51. J. O’M. Bockris, B. E. Conway, J. Chern. Phys.
vanced Treatise (Eds.: H. Eyring, D. Hender- 1956, 26, 532.
son, W. Jost), Academic Press, New York, 52. P. Ruetschi, P. Delahay, J. Chern. Phys. 1955,
1970, pp. 985–1074, Vol. IXB. 23, 1167.
37. H. Gerischer, Adv. Electrochem. Electrochem. 53. A. J. Appleby, Mod. Aspects Electrochem. 1974,
Eng. 1961, 1, 139. 9, 369.
38. W. Lorenz, Z. Elektrochem. 1953, 57, 382. 54. C. Wagner, M. Traud, Z. Elektrochem. 1938,
39. H. Gerischer, Z. Elektrochem. 1953, 57, 604. 44, 391.
40. K. Heusler, Electrochim. Acta 1983, 28, 439. 55. A. Brenner, G. Riddell, J. Res. Natl. Bur.
41. E. J. M. O’ Sullivan, E. J. Calvo in Reactions at Stand. 1946, 37, 31.
Metal Oxide Electrodes Comprehensive Chem- 56. M. Spiro, J. Chem. Soc., Faraday Trans. I
ical Kinetics (Ed.: R. G. Compton), Elsevier, 1979, 1507.
New York, 1987. 57. E. J. O’ Sullivan in Advances in Electrochemi-
42. J. Horiuti, M. Ikusima, Proc. Imp. Acad. cal Science and Engineering (Eds.: R. C. Alkire,
(Tokyo) 1939, 15, 39. D. M. Kolb), Wiley-VCH, Weinheim, Ger-
43. J. Horiuti, J. Res. Inst. Catal. Hokkaido Univ. many, 2002.
1948, 1, 8. 58. T. Erdey-Gruz, M. Volmer, Z. Phys. Chem.
44. R. Parsons, Pure Appl. Chem. 1974, 37, 501. Abt. A 1931, 157, 177.
45. R. Parsons, Pure Appl. Chem. 1979, 52, 233;
reprinted in Electrochim. Acta 1981, 26, 1869.
1.2 Quantum Theory of Electrochemical Electron-Transfer Reactions 31
1.2 These surfaces will also serve as the

Quantum Theory of Electrochemical basis for a discussion of solvent dynam-
Electron-Transfer Reactions ics. The model Hamiltonian will then be
extended to describe bond-breaking re-
Wolfgang Schmickler and Stefan Frank actions, which will conclude the main
Abteilung Elektrochemie, University of Ulm, part of this review. The last section is
Ulm, Germany
devoted to computer simulations, which
1.2.1 can fill in some of the gaps that are
Introduction missing in proper theory, and give quanti-
tative results.
Electron-transfer reactions are amongst In this chapter, we attempt to address
the most important processes in elec- a broad audience that comprises both
trochemistry. Indeed, all electrochemical experimentalists and theoreticians. We
reactions involve an exchange of electrons therefore have to strike a compromise
between a particle and the electrode at between readability and rigor and will
some stage. Therefore, electron transfer not be able to cover some of the more
has been the subject of intensive research formal aspects and second-order effects,
ever since the rise of electrochemical kinet- for which the book by Kuznetsov [3] is
ics toward the middle of the last century. a good reference. As an introduction,
Much of our present understanding we start with a simple presentation of
of electron-transfer reactions rests on the Marcus theory, which should provide
concepts that were established in the the necessary background for the more
pioneering papers by Marcus [1] and by formal sections.
Hush [2] in the 1950s. In particular, these
1.2.2
works pointed out the important role
A Simple Introduction to Marcus Theory
that the reorganization of the solvent
plays. Since that time the theory of
We consider electron transfer from a metal
these reactions has developed in various
electrode to a solvated species nearby
directions. A good review of the present
according to the scheme
state of the theory and its applications
to physics, chemistry and biology can be −
red −
−
−−−
− ox + e (Me) (1)
found in the book by Kuznetsov [3]. Here
we shall focus on a few topics relevant to At first, we limit ourselves to the so-called
electrochemical electron transfer at metal outer sphere reactions in which chemical
electrodes – semiconductors are treated in bonds are neither formed nor broken.
a recent monograph by Memming [4]. During such a reaction, the charge on
The formal development of the topic will the reactant, and hence the configuration
rest on a model Hamiltonian proposed by of its solvation sheath, changes. The
one of us in the 1980s. From this we will order in which electron transfer and
first derive the Levich and Dogonadze the- solvent reorganization occur is governed
ory [5], which was the first quantum theory by the Frank–Condon principle: electron
for electron transfer in condensed media, transfer is fast and occurs practically at
and then obtain the classical potential- a fixed position of the solvent. Obviously,
energy surfaces that are generalizations it would be energetically unfavorable if
of those familiar with Marcus and Hush. electron transfer were to take place before
32 1 Fundamentals
Fig. 1 Potential energy vs.

generalized solvent coordinate q
Red ox + e − in Marcus theory.
Energy
Eact
∆G
qi qf q
Reaction coordinate
the solvent had started to reorganize, since justified within an explicit model for the
the reactant would be surrounded by a solvation [6].
solvation sheath pertaining to a different The minima for the two parabolas cor-
charge. Likewise, a solvent reorganization respond to the equilibrium configurations
prior to electron transfer would require for the initial and the final states, and
high energy, and is equally unlikely. the difference between the values at the
Therefore, the solvation sheath must first minima gives the free energy G, of the
take up a configuration between those reaction. Changing the electrode poten-
for the initial and the final charge state, tial affects the energy of the transferred
then the electron is transferred, and the electron, and hence G. The free energy
system relaxes to its new equilibrium of activation is the difference between the
configuration. energies of the crossing point and the min-
These ideas can be put into quantitative imum of the initial state; it can easily be
terms within a simple, one-dimensional calculated within the harmonic approxima-
model in which the (free) energy of tion: Let the two parabolas for the initial (i)
the system is plotted as a function of and the final (f ) states be
a generalized solvent coordinate q char-
acterizing the solvation of the reactant. Ui (q) = 12 k(q − qi )2 ,
There are two such curves (see Fig. 1), Uf (q) = 12 k(q − qf )2 (2)
one for each side of Eq. (1). Each curve
attains its minimum at the equilib- where qi and qf denote the two minima.
rium configuration for the corresponding A simple calculation gives, for the energy
state. If we develop each curve into a of activation
Taylor series about its minimum and
(λ + G)2
keep terms up to second order, we ob- Eact = (3)
tain the harmonic approximation familiar 4λ
through many branches of physics and where
chemistry in which each curve is rep- λ = 12 k(qi − qf )2 (4)
resented by a parabola. We make the
further assumption that the curvatures is the so-called energy of reorganization for
of both parabolas are equal – this can be the solvent.
The preexponential factor Z for the factor C = 1/κ ≈ 10−8 cm to the pre-
rate constant can be calculated from the exponential factor.
activated complex or from Kramer’s theory 2. For electrochemical reactions at metal
(vide infra); it is independent of G electrodes, the reaction-free energy G,
or, in the electrochemical context, of the refers to the exchange with the Fermi
overpotential η. Thus, the rate constant can level, and can then be related to the
be written as overpotential η by G = ±e0 η, where
the sign depends on the direction of the
(λ + G)2 current, anodic or cathodic. However,
k = Z exp − (5)
4λkB T in a reduction reaction the electron can
come from any occupied state, and in
where kB is the Boltzmann’s constant. an oxidation it can be transferred to
This expression gives the rate for a any empty level. Let us denote by ε
homogeneous electron-transfer reaction; the energy of a metal electron with
it is usually associated with the name respect to the Fermi level, and by ρ(ε),
of Marcus, but the independent work the density of electronic states at the
of Hush is also equivalent. The sign surface. To be specific, we consider
convention in Eq. (3) is the usual one: electron transfer from the metal to a
a negative sign denotes a lowering of redox partner in the solution. The free
the free energy during the reaction; thus, energy for the transfer of an electron
−G measures the driving force for with energy ε is by an amount −ε
the reaction. An unusual feature is the lower than that for electron transfer
quadratic dependence of the activation from the Fermi level. The probability
energy on G. As long as λ + G > 0, of finding an electron with that energy
the reaction gets faster as the driving force is given by ρ(ε)f (ε), where f (ε) is the
increases, but for very large driving forces, Fermi–Dirac distribution. Hence, the
where λ + G < 0, the reaction becomes rate of electron transfer from the metal
slower. This is the so-called inverted region, to the reactant is given by
whose existence has been established for

homogeneous reactions. For reasons that
will become apparent below, it cannot kred = ZC ρ(ε)f (ε)
be accessed in heterogeneous reactions at
(λ − ε + e0 η)2
metal electrodes. × exp − dε (6)
In order to apply Eq. 5 to electrochemical 4λkB T
reactions, we must make two modifica-
where we have used the fact that a neg-
tions:
ative overpotential favors a reduction.
1. Equation 5 gives the rate for a reac-
tant sitting at a given distance from Equation (6) has also been derived by
the electrode. In order to get the to- Gerischer [7, 8], but with a different
tal rate, we must integrate over all interpretation. He defines
distances. Typically, the rate decreases
exponentially with the distance, with Wox (ε, η) = (4πλkB T )(−1/2)
a decay constant of γ ≈ 1 A−1 . Inte- (λ − ε + e0 η)2
gration, therefore, adds an additional × exp − (7)
4λkB T
34 1 Fundamentals
as the normalized density of oxidized to the integral comes from the vicinity
or empty states in the solution. Since of the Fermi level. If the density of
ρ(ε)f (ε) is the density of occupied states states ρ(ε) is taken to be a constant,
on the metal, Eq. (6) can be interpreted in and the Fermi–Dirac distribution replaced
the following way: The rate of exchange by a step function, the integral can be
of electrons of energy ε is proportional performed explicitly:
to the density of occupied states on the
metal times the density of empty states in e0 η + λ
kred = ZCρ πλkB T erfc
the solution. Changing the overpotential (4λkB T )1/2
simply shifts the density of oxidized states (9)
with respect to the Fermi level. Here erfc denotes the complement of
The form of Eq. (6), where the total rate the error function. Equations (6) and (9)
is written as an integral over the energy predict rate constants with a transfer coef-
ε of the transferring electron is found in ficient of 1/2 for small overpotentials, and
several versions of the electron-transfer a constant rate klim for high overpotentials
theory. For future reference, we write it as (see Fig. 2).
From the form of the current-potential
kred = ZC ρ(ε)f (ε)kr (ε) dε (8) curve, it is clear that the Marcus-inverted
region cannot be accessed in a simple
where kr (ε) is the energy-resolved rate. redox reaction on a metal electrode. For
The integration in Eq. (6) is to be large overpotentials, the electrons are
performed over the conduction band of the simply transferred to energy levels above
electrode; in practice, it may be extended the Fermi level so that the inverted part of
to ±∞, since the major contribution the energy-resolved rate plays no role.
0
λ = 0.5 eV
−4 λ = 0.75 eV
ln (kred/klim)
λ = 1 eV
−8
−12
0.0 0.4 0.8 1.2 1.6

η
[V]
Fig. 2 Logarithm of the normalized rate constant kred /klim vs.
overpotential for the reduction of a redox couple at a metal electrode
according to Marcus theory (Eq. 9).
1.2.3 of nonlocal electrostatics or involving the

Energy of Reorganization whole frequency spectrum of the dielectric
constant, have been suggested, but the best
As outlined above, the energy of reorgani- estimates nowadays come from computer
zation plays a major role in the electron- simulations.
transfer theory. Generally it contains two The contribution from inner-sphere
contributions: one from the surrounding modes are easily obtained within the
solvent and the other arising from changes harmonic approximation. Let q be the
in the bond lengths and vibration frequen- change in the bond length of a mode
cies of the reacting complex itself. They are with frequency ω and effective mass m;
also referred to as outer-sphere and inner- then its contribution to the energy of
sphere reorganization, respectively. reorganization is
The reorganization of the solvent can
be expressed through its polarization, λi = 12 mω2 q 2 (11)
which also contains two contributions: one
from the electronic polarizability of the When electron transfer is accompanied
solvent molecules, and the other from the by a significant change in frequency
librational and vibrational motion. Only from ω1 to ω2 , an effective frequency
the latter are slower than the electron 2ω1 ω2 /(ω1 + ω2 ) can be used to a good
exchange as such, and contribute to the approximation.
solvent reorganization energy λout . This
takes a form that is reminiscent of the 1.2.4
Born formula for the energy of solvation: Quantum-mechanical Theory

1 1 1 1.2.4.1 A Model for Electron Exchange
λout = − Between a Metal and a Solvated Reactant
2ε0 ε∞ εs
Before we set up the model Hamiltonian
× (Dox − Dred )2 dV (10) for electrochemical electron transfer, we
have to specify the models for the vari-
Here, ε∞ and εs are the optical and the ous parts of the system. For the electrons
static dielectric constants of the solution; in the metal, we use the quasi-free elec-
the former appears because the contribu- tron model in which the electronic states
tion from the electronic polarizability has are labeled by their quasi-momentum k.
been subtracted. Dox and Dred are the di- For outer-sphere electron transfer on met-
electric displacements when the reactant is als, it is usually permissible to ignore the
in the reduced and in the oxidized form, re- spin index – keeping it would introduce an
spectively. The integral is to be performed additional factor of two, which can be in-
over the space filled by the solution. When corporated into the interaction constants.
the reactant is close to the electrode sur- On the reactant, we consider a single or-
face, image terms arise, which contribute bital, labeled a, with which the electrons
to the displacement. are exchanged.
The energy of reorganization is a molec- An important aspect of electron transfer
ular concept, and an equation such as (10), is the accompanying reorganization of the
which is based on macroscopic electrostat- solvent. The latter can be modeled as a
ics, can only be a rough approximation. phonon bath, or a collection of harmonic
Several other expressions, on the basis oscillators. Other models in terms of the
36 1 Fundamentals
solvent polarization [9] or the orientation and the reactant is incorporated in the
of the solvent dipoles [10] are equivalent. third term; the amplitudes Vk characterize
The formulation in terms of harmonic os- the strength of the electronic coupling
cillators has the additional advantage that between the two partners. The fourth term
it is easy to include quantum modes: Most describes the bath of harmonic oscillators;
reactions are coupled to a few quantum qν and pν are dimensionless coordinates
modes, such as inner-sphere modes of the and momenta, and ων the frequencies. The
reactant, or even solvent modes in the summation runs over both the classical
first solvation shell [11]. These modes can solvent modes and over the quantum
be included in the collection of harmonic (inner-sphere) modes. The coupling of the
oscillators; of course, in contrast to the oscillators to the redox couple is assumed
classical solvent modes they have to be to be linear and proportional to the charge,
treated by quantum mechanics. as given in the last term.
The solvent modes can be separated This Hamiltonian, which was intro-
into the slow librational and vibrational duced by Schmickler [12], is equivalent to
modes, and the fast electronic modes. Only earlier formulations by Levich and Dogo-
the slow modes are represented in the nadze in terms of wave mechanics [5]; it is
phonon bath. The fast modes are supposed also related to the spin–boson model for
to be in electronic equilibrium with the homogeneous electron exchange [13] and
transferring electron so that they simply to the Anderson–Newns model for specific
renormalize the energy of the reactant’s adsorption [14].
electronic states. Further treatment of this model depends
on the magnitude of the electron-transfer
1.2.4.2 The Model Hamiltonian term: if the amplitudes Vk are small,
The quantitative formulation of the model the first-order perturbation theory can be
outlined above is conveniently written in applied; if they are large, it is better
terms of second quantization. For this to start from the adiabatic potential-
purpose, we introduce number operators energy surfaces.
n and creation and annihilation operators
c+ and c for the various electronic 1.2.5
states. Then the Hamiltonian takes the First-order Perturbation Theory
form
1.2.5.1 Semiclassical Systems
H = εa na + εk nk + [Vk ck+ ca We next consider the case in which the
k k electron-transfer terms are small so that
1 the first-order perturbation theory can be
+ Vk∗ ca+ ck ] + h̄ων (pν2 + qν2 ) applied and assuming that all the phonon
2 ν
modes are classical – quantum modes will
be considered below. In this case, the
− na h̄ων gν qν (12) coordinates qν of the oscillators can be
ν
treated as external parameters. It is then
The first term describes the electronic state convenient to include the electron-phonon
of the reactant with energy εa , the next coupling – the last term in Eq. (12) – into
term represents the electrons in the metal. the electronic energy of the reactant,
Electron exchange between the metal and rewrite the first and last term as
ε̃a na , where • Finally, we relate εa to the overpotential

η by noting that for η = 0 the forward
ε̃a = εa − h̄ων gν qν (13) and backward rate must be equal;
ν
this gives: εa = λ + e0 η. Further, we
To first order in the coupling constants introduce the factor C converting bulk to
Vk , the rate of electron transfer from a surface concentrations (see Sect. 1.2.2),
state k on the metal to the reactant at a and take the coupling elements Vk
given configuration qν is given by Fermi’s as constant.
golden rule:
This procedure results in the following
|Vk |2 expression for the electrochemical rate
Wka (qν ) = δ(ε̃a − εk ) (14) constant for the reduction:
h̄
The full derivation of the rate constant 1/2
|Vk |2 π
for this case can be found in Refs. [6, 15]. kred = C
h λkB T
Here we list only the main steps involved:
(λ − ε + e0 η)2
× ρ(ε)f (ε) exp −
• The delta function in Eq. (14) is replaced 4λkB T
by its Fourier transform; this greatly (18)
simplifies the mathematics. which differs from the Marcus–Hush
• Next the thermal average over all expression in the preexponential factor,
configurations qν is performed. This
which is here determined by the square
results in an energy-resolved rate for
of the transition amplitude Vk . This is
the reduction:
characteristic for the nonadiabatic, weak
1/2
|Vk |2 λ coupling case. Solvent dynamics, which
kr (εk ) = is so important for the adiabatic, strong
h̄ 4kB T
coupling case plays no role.
(εa − εk )2
× exp − (15)
4λkB T
1.2.5.2 Quantum Modes
for the rate at a given energy εk , which When quantum modes are reorganized
is to be compared with Eq. (5) of the during the electron exchange, they have
simple Marcus theory. Here to be treated separately from the classical
1 modes. We consider explicitly the case
λ= h̄ων gν2 (16) in which one such mode is present;
2 ν
the generalization to many modes is
is the energy of reorganization of the straightforward, though it results in rather
phonon bath. Thus, in the semiclassical cumbersome formulae.
case the coupling to the bath is charac- The situation is simplest if the frequency
terized by a single constant. ωi of the quantum mode is so high that its
• To get the total rate, we have to sum quantum of energy is much higher than
over all occupied levels on the metal. the thermal energy, h̄ωi kT , so that it
This sum is converted into an integral: is not excited in the initial state. In this
case, we only have to consider transitions
→ dε f (ε)ρ(ε) (17) from the ground state 0 to the nth excited
k,occ. state. Let
38 1 Fundamentals
n
f 1 λi where . . . denotes a thermal average.
M0n = |ψ0i |ψn |2 = A good general treatment of reactions
n! h̄ωi
involving quantum modes has been given
λi
× exp − (19) by Tanaka and Hsu [16].
h̄ωi
be the Frank–Condon factors for these 1.2.6 Adiabatic Reactions

f
transitions. Here, ψni , ψn are the vibronic When the electronic interaction is large,
wave functions for the initial and final perturbation theory does not suffice. We
states; λi is the energy of reorganization consider the adiabatic limit in which
pertaining to the quantum mode. The the system is always in electronic equi-
rate of electron transfer involving the librium – the general case of an inter-
transition 0 → n is then obtained by action with an intermediate strength
multiplying the classical rate of Eq. (17) by has not yet been solved – and con-
the corresponding Frank–Condon factor, struct the potential-energy surface for
and by noting that the energy n h̄ωi , the reaction.
which goes into the excitation of the The electronic interaction with the metal
quantum mode, is no longer available for induces a broadening of the electronic
the classical reorganization. The energy- level of the redox partner, which is
reduced rate for the reduction is then determined by the matrix elements Vk :
1/2
|Vk |2 π
kr (0 → n) = M0n
h̄ λkB T =π |Vk |2 δ(ω − εk ) (23)
k
(λ + n h̄ωε − e0 η)2
× exp − (20)
4λkB T In principle, is a function of the
electronic energy ε. When the reactant
The total rate is obtained by summing over
all final states n: interacts with a broad electronic band, the
width usually does not vary much in

kr = kr (0 → n) (21) the relevant region near the Fermi level
n and can be taken as constant; this is the
so-called wide-band approximation, which
As long as e0 η < h̄ω, transitions from will be used here.
ground state to ground state dominate. So far the adiabatic case has only been
When the quantum of energy is of the fully worked out for the semiclassical
same order of magnitude as the thermal case in which quantum modes play no
energy, h̄ω ≈ kT , we have to consider role. In this case, it is again convenient
transitions from excited initial states m to use the q-dependent electronic en-
to final states n, and perform a thermal av- ergy ε̃a introduced in Eq. (13). In the
erage over the initial states of the quantum wide-band approximation, the broaden-
modes: ing of this energy level takes the form
of a Lorentzian so that its density of
|Vk |2
kr = Mmn (π/λkB T )1/2 states becomes
h̄

(λ+(n − m) h̄ωε−e0 η)2 2
× exp − (22) ρ(ω) = (24)
4λkB T [ω − ε̃(qν )]2 + 2
The reactant orbital is filled only partially charge states of the reactant. In an
up to the Fermi level of the metal, and its electron-transfer reaction, the system
occupancy is given by moves from one minimum via the
barrier to the other minimum.
1 ε̃a
n(qν ) ≡ na (qν ) = arccot (25) In mathematical terms, three stationary
π
points exist if both the conditions
where the Fermi level has been taken as
2λ
the energy zero. The potential energy of the >1 (28)
system, as a function of the coordinates qν , π
is then 2λ
|ε − λ| ≤ |arccos(γ ) − γ (1 − γ 2 )1/2 |
1 π
E[qν ] = ε̃a n(qν ) + h̄ων qν2 (29)
2 ν are fulfilled, where γ 2 = π/2λ.
Within this formalism, it is in prin-
+ ln(ε̃a2 + 2 ) (26) ciple possible to calculate the potential-
2π
energy surfaces for electron-transfer re-
where a constant term has been neglected. actions [17]. The important system pa-
With the aid of Eqs. (25 and 26) the rameters are λ, and εa ; they can be
potential energy of the system can be estimated from quantum chemical cal-
calculated as a function of the coordi- culations and computer simulations. At
nates qν . Of particular interest are the present, such surfaces involve fairly rough
stationary points of this curve, at which approximations; however, they are useful
the occupancy of the reactant obeys the visualizations for understanding the dy-
self-consistency equation: namics of electron-transfer reactions. It
is convenient to restrict the system to a
1 εa na − 2λn0
n0 = arccot (27) minimum set of coordinates using a gen-
π eralized solvent coordinate that follows a
which must be solved numerically. De- reaction path from one minimum of the
pending on the values of the system surface via a saddle point to the other
parameters, this equation has either one or minimum. This coordinate can be nor-
three solutions [17]. In the case in which malized in such a way that its value at the
the system is not too far from equilibrium, two minima corresponds to the charge on
they can be stated in the following form: the reactant in these configurations [18].
It, then, is identical to the reaction coordi-
• When the electronic interaction is nate in the theory of Hush [2]. If quantum
large and the energy of reorganization modes are reorganized during the electron
λ is small, there is only one stationary transfer – typically, these would be inner-
state. This corresponds to an adsorbed sphere modes – they have to be singled out
state, which sits in a single potential well and treated separately.
on the electrode surface. Figure 3 shows an adiabatic potential-
• When the electronic interaction is weak energy surface for one single solvent
and the energy of reorganization is large, coordinate q. Two minima are separated by
there are three stationary solutions: two a barrier. This barrier is the lower and the
minima separated by a barrier. The flatter the larger the electronic coupling
two minima correspond to two different between the reactant and the metal. For
40 1 Fundamentals
15.0
12.5
10.0
Energy
[kB T ]
7.5
5.0
2.5
0.0
−1 0 1 2 3
Solvent coordinate q
Fig. 3Adiabatic potential-energy curves for various values of the energy
broadening ; full line: = 0.01 eV; long dashes: = 0.05 eV; short dashes:
= 0.1 eV. The energy of reorganization was taken as 6 kB T.
the lowest coupling shown, the surface case involving an inner-sphere mode has
resembles two intersecting parabola. This been treated, amongst others, by Sumi
is the weakly adiabatic regime in which and Marcus [19]. The simplest estimate
the electronic coupling is strong enough for the rate of an adiabatic reaction is
to establish electronic equilibrium, but too based on the transition-state theory. If the
weak to lower the barrier. potential-energy well of the initial state is
approximated by a harmonic oscillator of
1.2.7 frequency ωi , the rate in that theory is
Solvent Dynamics given by
ωi Eact
In the nonadiabatic case, electron transfer k= exp − (30)
is a rare event. Therefore, the equilibrium 2π kB T
distribution in the initial well is undis- where Eact is the energy of activation. How-
turbed, and the preexponential factor is ever, as was pointed out by Kramers [20],
determined by the electronic coupling be- the transition-state theory overestimates
tween the reactants. In adiabatic reactions, the rate since the population at the barrier
the electronic system is in equilibrium at may be depleted, and also because a sys-
all points of the trajectory; therefore, the tem that has passed the saddle point can
rate is not limited by electron transfer but return, a phenomenon denoted as barrier
by solvent dynamics. recrossing. These effects can be described
The main points can be explained within in terms of the solvent friction, which en-
the one-dimensional model outlined in the compasses the influence of all the solvent
preceding section in which the free energy modes that are not explicitly considered in
is taken to be a function of a generalized the one-dimensional model. As a result,
solvent coordinate – the two-dimensional the preexponential factor is modified by an
additional transition factor: represent the results; however, instead of

the coordinate q defined in Sect. 1.2.6
(γ 2 /4 + ωb2 )1/2 − γ /2
κ= (31) most researchers use an electronic shift
ωb δE = −2λq. As electron transfer requires
where γ is the friction coefficient, and relatively large fluctuations of the ion solva-
ωb is the barrier frequency obtained tion, umbrella sampling techniques must
by approximating the barrier region by be applied and in most cases it is not pos-
an inverted parabola. In effect, Eq. (31) sible to obtain the electron-transfer rates
describes a competition between the force directly from the MD trajectories. Instead,
exerted by the barrier, which is determined one applies transition-state theory or per-
by ωb and the stochastic interaction with forms simulations near the saddle point
the surrounding bath characterized by γ . in order to explore solvent dynamics near
When the friction is weak, γ ωb , the the barrier.
transition factor reduces to unity, and Rose and Benjamin [22] have computed
the transition-state theory holds. In the the diabatic free energy surface of the re-
opposite limit, the case of strong friction, dox pair Fe2+ /Fe3+ at a fixed distance
the transition factor simplifies to from a Pt(100) surface (Fig. 4). Differ-
γ ences to a parabolic representation of the
κ= (32) free energy are only significant far from
2ωb
equilibrium, which affects the reorgani-
For electron transfer in solutions, it is zation energy λ, but hardly affects the
often assumed that the strong friction free activation energy. In contradiction to
limit holds. The friction coefficient has a common assumption, the curvature at
been identified with the inverse of the the potential minimum for the oxidized
longitudinal relaxation time: and the reduced states are slightly differ-
ent. Both the electron-transfer rate k and
1 ε∞
γ = , where τl = τD (33) the transfer coefficient α depend on the
τl εs applied overpotential η. In the regime of
and τD is the usual relaxation time. strong electronic coupling, they elucidate
a classical transmission coefficient κ in
1.2.8 agreement with Grote–Hynes theory [28],
Computer Simulations which is a generalization of Kramer’s the-
ory; in particular, κ increases as the barrier
We have already emphasized the impor- becomes sharper.
tant role of the solvent dynamics and struc- Smith and Halley [23] performed sim-
ture in the kinetics of electron-transfer ulations on a similar system for various
reactions. During the last few years, a fixed distances between the ion and the
number of classical molecular dynamic metal surface. They found a weak depen-
(MD) simulations have been performed dence of the reorganization energy on the
to obtain free energy surfaces of the re- separation, but strong variations of the
action [21–27]. These simulations require thermodynamic driving force; even the di-
explicit interaction potentials between the rection of the reaction changed when the
constituents of the system: the reactants, ion was moved away from the surface.
the solvent, and the electrode. Again, a Since their model includes an external
generalized solvent coordinate is used to electric field, but no realistic screening,
42 1 Fundamentals
100
80
60
[kcal mol−1]
G
40
20
Fe3+ Fe2+
0
−4 −3 −2 −1 0 1 2 3 4
(∆E − ∆E ‡)
[eV]
Fig. 4 Diabatic solvent reorganization free energy curves for Fe2+
and Fe3+[22]. The bold solid lines represent the molecular dynamics
result, the thin solid lines the best parabolic fit of the region, near the
bottom of each well.
it is difficult to assess the significance of of the high-frequency degrees of freedom

their results. We note that their results of water leads to a major shift in the ther-
contradict an earlier work of Straus and modynamic driving force, and hence, in
Voth [21]. the solvation of the redox couple. Remark-
In a series of publications, Voth and ably, the reorganization energy λ is almost
coworkers [21, 24, 25, 26, 27] explored var- unaffected; this means that the activation
ious aspects of electron-transfer reactions. energy at equilibrium does not change,
Their calculations are based on a version but only the equilibrium potential (Fig. 5).
of the Anderson–Newns Hamiltonian pre- When the constraint of a fixed ion po-
sented in Sect. 1.2.4.2, which allows the sition is released and the ion is allowed
direct computation of adiabatic free en- to move [25], the adiabatic free energy
ergy surfaces. For a Fe2+ /Fe3+ couple surface is still very similar to the earlier
at a fixed distance to a Pt(111) surface, case, but the barrier close to equilibrium
they compared a classical and a quan- is slightly higher. The ion is on average
tized water model [24]. The quantization further away from the surface, weakening
40
30
20
Classical
G(∆E)
[kcal]
10
Quantum
−10
250 300 350 400 450 500

∆E
[kcal]
Fig. 5 Adiabatic free energy curves for the Fe2+ /Fe3+ couple from
molecular dynamics [25]. The solid line represents results for the
quantized high-frequency degrees of water, the dashed line for the
classical degrees of freedom only.
the electronic coupling. When the ion is for the uncharged reactants, two for the
moved away from the surface, the thermo- transfer of an electron to one of the re-
dynamic driving force can change for the actants, the other remaining neutral, and
same reason. The presence of a counte- one for the transfer of one electron to each
rion causes nonmonotonical shifts in the of the reactants (Fig. 6). Agreement with
barrier height and driving force depend- an analytic model for such a process [29]
ing on the separation of the two ions. is satisfying, showing similar trends in the
For short distances, the reorganization en- driving force of the elementary steps.
ergy is markedly increased. The influence The solvent dynamics of a one-dimen-
of the electronic polarizability has been sional system was explored by Kuznetsov
investigated for a Na0 /Na+1 couple by in- and Schmickler [30]. They singled out
troducing a fluctuating charge on the water one important solvent coordinate and
molecules [26]. The electronic polarizabil- represented all the others by a heat bath.
ity causes a stronger solvation of the ion This is a similar concept to analytical
and a lower reorganization energy λ. The theories such as Kramer’s [20]. Adiabatic
polarizable water can accommodate the potential-energy surfaces were calculated
change of the ion charge to some extent from the Anderson–Newns Hamiltonian
and less solvent reorganization is needed. (see Fig. 3). The result is a symmetric
Also, changes in the thermodynamic driv- double well, the barrier being the lower
ing force are observed. For two different and the less sharp, the higher the electronic
neutral redox species [27], the free en- coupling . MD-simulations of a motion
ergy surface shows four distinct wells, one on this surface coupled to the heat bath
44 1 Fundamentals
I(2)
30
25
F(∆E1, ∆E2)
20
15
[kcal]
10
5
P R
50
−200 0
−150 I(1) −50
−100 −100
∆ E1 −50 −150 ∆E2
[kcal] 0 −200 [kcal]
50
Fig. 6 Adiabatic free energy surface for the electron transfer between two different redox
couples and a metal electrode from molecular dynamic simulations [27].
were performed. The time that was needed electrochemical reactions, bond breaking
to cross the transition state starting from was introduced into the Marcus–Hush
the reactant well with thermal velocity framework by Savéant [33, 34]. Koper
yields the reaction rate. The higher the and Voth [35, 36] reformulated these
friction exerted by the heat bath, the more ideas in terms of an extension of the
the motion resembles a random walk, model Hamiltonian of Eq. (12). The model
and the more likely is barrier recrossing applies to a reaction of the type:
(Fig. 7). The authors compare the reaction
R − X + e− −−−→ R• + X− (34)
rates to the predictions of transition-
state theory, and obtain a transmission Typically, R is an organic fragment, and
coefficient κ. The results are in qualitative, X− a halide anion. The electron-transfer
but not in quantitative agreement with step and bond breaking are assumed
Kramer’s theory [20]. to be concerted; an example for such
a reaction is the reductive cleavage of
1.2.9 alkyl halides. The authors present both
Bond Breaking Reactions an analytical theory, [35] and molecular
dynamic computer simulations [36], and
The models presented in the preceding calculate adiabatic energy surfaces. In both
section consider only elementary electron- cases, the bond breaking is introduced
transfer steps. In the same sense, many in a similar way following Savéant’s
theories such as those of Marcus and Hush ideas. The R − X bond is described by a
are restricted to outer-sphere reactions. Morse potential, the interaction between
For homogeneous electron transfer, bond the fragments R and X− is described
breaking was considered in early papers by the repulsive part of the same Morse
by German and Dogonadze [31], and potential. The following term is added to
later by German and Kuznetsov [32]. For the Anderson–Newns Hamiltonian:
0.80
g = 1.98
0.75
0.70
0.65
(a) 0.60
g = 1.98
0.72
Distance
0.70
0.68
0.66
(b) 0.64
g = 18.1
0.750
0.625
0.500
0.375
0.00 0.25 0.50 0.75 1.00
(c) Time
Fig. 7 Molecular dynamic trajectories of a particle coupled to a heat
bath on the potential-energy surface of Fig. 1. The saddle point is
indicated by the dashed line. (a) and (b) are for relatively high
friction and (c) low friction. = 0.05 eV. The unit of time has been
chosen such that the period of the harmonic oscillator is 2π , and the
friction coefficient γ is given in terms of this unit.
46 1 Fundamentals
HRX = (1 − na )D{1 − exp[−κ(r − r0 )]}2 ground state of the bonding mode, but
the energy of the final state is much
+ na D exp[−2κ(r − r0 )] (35) larger than the thermal energy kB T .
where D is the dissociation energy of the This is energetically unfavorable, but the
bond, and r0 its equilibrium distance. The overlap with higher energetic states is
parameter κ determines the vibration fre- better. As a consequence, the transfer
quencies and the anharmonicity. Since coefficient α at equilibrium is larger
the bond is broken when the orbital a than 1/2. At higher overpotentials, α
is occupied, that orbital is antibonding. All approaches 0 as familiar from outer-sphere
vibrations are assumed to be classical. The electron-transfer reactions. In comparison
potential-energy surface calculated from to Savéant’s theory, the activation energy
analytical theory [35] for small electronic is smaller owing to the inclusion of
coupling shows a distinct well for the tunneling effects.
reactants and an exit channel for the prod-
ucts; both are separated by a saddle. These 1.2.10
qualitative features are found again on the Conclusion
free energy surface calculated from MD
simulations [36]; however, there are quali- After fifty years of research, the theory
tative differences (Fig. 8). The product exit of electrochemical electron-transfer reac-
channel shows a shallow well for short dis- tions is well developed, and has provided
tances between the fragments. This is a us with a basic understanding of the es-
consequence of the microscopic solvation sentials. However, a number of important
effects, which are not included in the an- questions are still open, and remain the
alytical theory. However, the barrier that subject of ongoing research. We list what
has to be overcome to separate the prod- we believe to be the most important points:
ucts is much lower than the barrier for
the electron-transfer reaction itself, so frag- • At present, we have a theory for weak
ment dissociation is never the rate-limiting coupling on the basis of first-order per-
step. Comparison to a linear response turbation, and for strong coupling, on
model on the basis of the results of the ana- the basis of the adiabatic limit. However,
lytical theory, show quantitative deviations the intermediate case is unexplored; in
for the minima and curvature of the sur- particular, we do not fully know how
face, as well as for the overpotential. These the rate constant will behave as a func-
differences arise from nonlinear response tion of the electronic coupling. A recent
of the solvent to the change in the charge paper by Mohr and Schmickler [38] in-
on the redox center and the volume exclu- vestigates this behavior, but the effects
sion of the R fragment. Thus, the analytical of solvent dynamics are not included in
theory gives rather a rough qualitative pic- this work so that it will apply to very
ture, but serves as a helpful visualization short timescales only.
of the basic characteristics of the reaction. • Although computer simulations have
The case of weak coupling, in which provided us with some insights into
the first-order perturbation theory applies, deviations from the harmonic approx-
was explored by Schmickler [37]. For the imation, much more work needs to
parameters chosen, the major contribution be done in order to provide us with
to the reaction rate comes from the accurate potential-energy surfaces, and
F (∆E, r )
[kcal mole−1]
100
50
0
50
0
−50
−100
[kcal mole−1] −150
−200
−250 3 2
5 4
−300 7 6
9 8 r
(a) [Angstroms]
F (∆E, r )
[kcal mole−1]
100
50
0
50
0
−50
−100
∆E −150
[kcal mole−1]
−200
−250 3 2
5 4
−300 7 6
9 8 r
(b) [Angstroms]
Fig. 8 Adiabatic free energy surface for simultaneous electron transfer and
bond breaking from MD simulations (upper) and a linear response solvent
model (lower panel) [36].
also to explore subtle effects such as the its effect on quantum modes re-
variation of the electronic overlap with mains unexplored.
electrode potential. • During the last decade, long-range elec-
• Solvent dynamics have been explored tron transfer along molecular chains
mainly within a wholly classical context; has become very interesting, not least
48 1 Fundamentals
because of its application to biological 14. D. M. Newns, Phys. Rev. 1969, 178, 1123.
systems. While the superexchange the- 15. P. P. Schmidt, Specialist Periodical Reports,
Electrochemistry, The Chemical Society, Lon-
ory [39, 40] is a good first approximation,
don, 1975, Vol. 5, pp. 21–131.
it leaves many questions, in particular, 16. S. Tanaka, C.-P. Hsu, J. Chem. Phys. 1999,
the coupling to solvent and vibrational 111, 11 117.
modes, open. 17. W. Schmickler, Chem. Phys. Lett. 1995, 237,
152.
With these and other problems still 18. W. Schmickler, Electrochim. Acta 1996, 41,
open, the electron-transfer theory remains 2329.
19. H. Sumi, R. Marcus, J. Chem. Phys. 1986, 84,
a vibrant field in physics, chemistry and
4894.
biology alike. 20. H. A. Kramers, Physica 1940, 7, 284.
21. J. B. Straus, G. A. Voth, J. Phys. Chem. 1993,
Acknowledgment 97, 7388.
22. D. A. Rose, I. Benjamin, J. Chem. Phys. 1994,
100, 3545.
Financial support by the Volkswagens- 23. B. B. Smith, J. W. Halley, J. Chem. Phys.
tiftung is gratefully acknowledged. 1994, 101, 10 915.
24. J. B. Straus, A. Calhoun, G. A. Voth, J. Chem.
Phys. 1995, 102, 529.
References
25. A. Calhoun, G. A. Voth, J. Phys. Chem. 1996,
100, 10 746.
1. R. A. Marcus, J. Chem. Phys. 1956, 24, 966. 26. A. Calhoun, G. A. Voth, J. Electroanal. Chem.
2. N. S. Hush, J. Chem. Phys. 1958, 28, 962. 1998, 450, 253.
3. A. N. Kuznetsov, Charge Transfer in Physics, 27. A. Calhoun, G. A. Voth, J. Chem. Phys. 1998,
Chemistry and Biology, Gordon & Breach, 109, 4569.
Longhorne, Pa., 1995.
28. R. F. Grote, J. T. Hynes, J. Chem. Phys. 1980,
4. R. Memming, Semiconductor Electrochem-
95, 2715.
istry, Wiley-VCH, Weinheim, Germany,
29. Y. G. Boroda, A. Calhoun, G. A. Voth, J.
2000.
Chem. Phys. 1997, 107, 8940.
5. V. G. Levich, Kinetics of reactions with
30. A. M. Kuznetsov, W. Schmickler, Chem.
charge transport in Physical Chemistry,
Phys. Lett. 2000, 327, 314.
An Advanced Treatise (Eds.: H. Eyring,
D. Henderson, W. Jost), Academic Press, 31. E. D. German, R. R. Dogonadze, Int. J.
New York, 1970, Vol. IXb, pp. 986–1014. Chem. Kinet. 1974, 6, 467.
6. W. Schmickler, Interfacial Electrochemistry, 32. E. D. German, A. M. Kuznetsov, J. Phys.
Oxford University Press, New York, 1996. Chem. 1994, 98, 6120.
7. H. Gerischer, Z. Phys. Chem. N.F. 1960, 26, 33. J. M. Savéant, J. Am. Chem. Soc. 1987, 109,
223. 6788.
8. H. Gerischer, Z. Phys. Chem. N.F. 1961, 27, 34. J. M. Savéant, Acc. Chem. Res. 1993, 26, 455.
40, 48. 35. M. T. M. Koper, G. A. Voth, Chem. Phys. Lett
9. R. R. Dogonadze, Theory of molecular elec- 1998, 282, 100.
trode kinetics in Reactions of Molecules at Elec- 36. A. Calhoun, M. T. M. Koper, G. A. Voth, J.
trodes (Ed.: N. S. Hush), Wiley-Interscience, Phys. Chem. B 1999, 103, 3442.
London, 1971, pp. 135–227. 37. W. Schmickler, Chem. Phys. Lett. 2000, 317,
10. W. Schmickler, Ber. Bunsen-Ges. Phys. Chem. 458.
1976, 80, 834. 38. J. H. Mohr, W. Schmickler, Phys. Rev. Lett.
11. J. S. Bader, R. A. Kuharski, D. Chandler, J. 2000, 84, 1051.
Chem. Phys. 1990, 93, 230. 39. H. Taube, Electron Transfer Reactions of
12. W. Schmickler, J. Electroanal. Chem. 1986, Complex ions in Solutions, Academic Press,
204, 31. New York, 1970.
13. A. J. Legget, S. Chakravarty, A. T. Dorsey 40. D. N. Beratan, J. J. Hopfield, J. Am. Chem.
et al., Rev. Mod. Phys. 1987, 59, 1. Soc. 1984, 106, 1584.
1.3 Electronic Tunnel Factors in Molecular Electron Transfer and Molecular Conduction 49
1.3 1.3.1
Electronic Tunnel Factors in Molecular Introduction
Electron Transfer and Molecular
Conduction Electron transfer, a fundamental chemi-
cal process underlying all redox reactions,
Abraham Nitzan has been under experimental and theoreti-
Tel Aviv University, Tel-Aviv, Israel cal study for many years [1–6]. Theoretical
studies of such processes seek to under-
Abstract stand the ways in which their rate depends
on donor and acceptor properties, on the
Electron transmission through molecules solvent, and on the electronic coupling
and molecular interfaces has been a sub- between the states involved. The differ-
ject of intensive research owing to recent ent roles played by these factors and the
interest in electron transfer phenomena way they affect qualitative and quantita-
underlying the operation of the scan- tive aspects of the electron transfer process
ning tunneling microscope (STM) on one have been thoroughly discussed in the past
hand, and in the transmission properties half-century. This kind of processes, which
of molecular bridges between conducting dominate electron transitions in molecular
leads, on the other. In these processes, systems, is to be contrasted with electron
the traditional molecular view of elec- transport in the solid state, that is, in met-
tron transfer between donor and acceptor als and semiconductors. Electrochemical
species give rise to a novel view of the reactions that involve both molecular and
molecule as a current carrying conductor, solid-state donor/acceptor systems, bridge
and observables such as electron trans- the gap between these phenomena [6].
fer rates and yields are replaced by the Here, electron transfer takes place between
conductivities, or more generally by cur- quasi-free electronic states on one side,
rent–voltage relationships, in molecular and bound molecular electronic states on
junctions. Such investigations of electri- the other.
cal junctions in which single molecules The focus of the present discussion is
or small molecular assemblies operate as another class of electron transfer phe-
conductors, constitute a major part of what nomena: electron transmission between
has become the active field of molecular two regions of free or quasi-free electrons
electronics. through molecules and molecular layers.
In this chapter, the author reviews Examples for such processes are photoe-
the current knowledge and understand- mission (PE) through molecular overlay-
ing of this field with particular empha- ers, the inverse process of low energy
sis on theoretical issues, and on the electron transmission (LEET) into met-
relationship between this new look at als through adsorbed molecular layers and
electron transfer phenomena and the tra- electron transfer between metal, and/or
ditional study of molecular electron trans- semiconductor contacts, through molecu-
fer in solution. Different approaches to lar spacers. Figure 1 depicts a schematic
computing the conduction properties of view of such systems. The ‘‘standard’’
molecules and molecular assemblies are electron transfer model in Fig. 1(a) shows
reviewed, and the relationships between donor and acceptor sites connected by a
them is discussed. molecular bridge. In Fig. 1(b), the donor
50 1 Fundamentals
B Fig. 1 Schematic views of typical

D A electron transmission systems. (a) A
‘‘standard’’ electron transfer system
containing a donor, an acceptor, and a
(a)
molecular bridge connecting them (not
shown are nuclear motion baths that
must be coupled to the donor and
acceptor species). (b) A molecular
bridge connecting two electronic
L R continua, L and R, representing, e.g.,
two metal electrodes (c) same as
(b) with the bridge replaced by a
molecular layer.
(b)
L R
(c)
and the acceptor are replaced by a continua free or quasi-free electron states in the
of electronic states representing free space substrates (or, depending on the process,
or metal electrodes. (This replacement can in vacuum). In addition, coupling to the
occur on one side only, representing elec- thermal environment (represented by the
tron transfer between a molecular site and box ) may affect transmission through
an electrode). In Fig. 1(c), the molecular the bridge. The double arrows in the Figure
bridge is replaced by a molecular layer. represent the couplings between these dif-
A schematic view of the electronic states ferent subsystems.
involved is shown in Fig. 2. The middle The first two of the examples given
box represents the bridging molecule or above, PE and LEET, involve electrons
molecular layer and a set of levels {n} of positive energy (relative to zero kinetic
represents the relevant molecular orbitals. energy in vacuum), and as such are re-
In a ‘‘standard’’ electron transfer system, lated to normal scattering processes. The
(Fig. 2a) this bridge connects the donor third example, transmission between two
and acceptor species, now represented conductors through a molecular layer, in-
by potential surfaces associated with the volves negative energy electrons, and as
vibronic structure of the corresponding such is closely related to regular electron
intramolecular and solvent nuclear mo- transfer phenomena. The latter type of
tions. When the bridge connects two metal processes has drawn particular attention
electrodes (or separates a metal substrate in recent years owing to the growing inter-
from vacuum), these nuclear baths are re- est in conduction properties of individual
placed by manifolds of electronic states molecules and of molecular assemblies.
{l} and {r} that represent the continua of Such processes have become the subjects
{n}
D Θ A
(a)
{l} {n} {r}
Θ
(b)
Fig. 2 A schematic view of the electronic and nuclear states involved in
typical electron transmission systems. See text for details.
of intensive research owing to recent properties. Such investigations of electri-

interest in electron transfer phenomena cal junctions in which single molecules or
underlying the operation of the STM small molecular assemblies operate as con-
on one hand, and in the transmission ductors connecting ‘‘traditional’’ electrical
properties of molecular bridges between components, such as metal or semicon-
conducting leads on the other. In the lat- ductor contacts, constitute a major part
ter case, the traditional molecular view of of what has become the active field of
electron transfer between donor and ac- molecular electronics [7–17]. Their diver-
ceptor species, gives rise to a novel view sity, versatility and amenability to con-
of the molecule as a current carrying control and manipulation, make molecules
ductor, and observables such as electron and molecular assemblies potentially im-
transfer rates and yields are replaced by portant components in nano-electronic
the conductivities, or more generally, by devices. Indeed, basic properties pertain-
current–voltage relationships, in molecu- ing to single electron transistor behavior
lar junctions. Of primary importance is and to current rectification have already
the need to understand the interrelation- been demonstrated. At the same time,
ship between the molecular structure of major difficulties lie on the way to real
such junctions and their function, that technological applications [18]. These dif-
is, their transmission and conduction ficulties stem from problems associated
52 1 Fundamentals
with the need to construct, characterize, given (under the Condon approximation)
control and manipulate small molecular by the golden rule based expression
structures at confined interfaces with a
high degree of reliability and reproducibil- 2π
ket = |VDA |2 F (1)
ity, and from issues of stability of such h̄
small junctions.
It should be obvious that while the where VDA is the coupling between the
different processes outlined above corre- donor (D) and acceptor (A) electronic states
spond to different experimental setups, and where
fundamentally they are controlled by sim-
ilar physical factors. Broadly speaking, we F = F(EAD ) = Pth (εD (νD ))
νD νA
may distinguish between processes for
which lifetimes or rates (more generally, the × |νD |νA |2 δ(εA (νA ) − εD (νD ) + EAD )
time evolution) are the main observables, (2)
and those that monitor fluxes or currents. is the thermally averaged and Franck
In this review, we focus on the second Condon (FC) weighted density of nuclear
class, which may be further divided into states. In Eq. (2), νD and νA denote donor
processes that measure current–voltage and acceptor nuclear states, Pth is the
relationships, mostly near equilibrium, Boltzmann distribution over donor states,
and those that monitor the nonequilibrium εD (νD ) and εA (νA ) are nuclear energies
electron flux, for example, in photoemis- above the corresponding electronic origin
sion experiments. and EAD = EA − ED is the electronic
The focus of the following discussion is energy gap between the donor and acceptor
electron transfer between two conducting states. In the classical limit, is given by
electrodes through a molecular medium.
Such processes bear strong similarity to e−(λ+EAD ) /4λkB
2
the more conventional systems that involve F(EAD ) = √ (3)

4πλkB
at least one molecular species in the
donor/acceptor pair. However, important
where kB is the Boltzmann constant and
conceptual issues arise from the fact that
is the temperature, and where λ is the
such systems can be studied as part
reorganization energy, a measure of the
of complete electrical circuits, providing
electronic energy that would be dissipated
current–voltage characteristics that can be
after a sudden jump from the electronic
analyzed in terms of molecular resistance,
state describing an electron on the donor
conductance, and capacitance.
to that associated with electron on the
acceptor. If the donor (say) is replaced
1.3.2
by an electrode [6, 19, 20], we have to sum
Standard Electron Transfer Theory
over all occupied electrode states
To set the stage for our later discussion,
|VDA |2 F ⇒ f (εk )F(εk − e)|VkA |2
we first briefly review the rate expressions
k
for ‘‘standard’’ electron transfer processes
(Figs. 1a, 2a, 3a). We focus on the = dεf (ε)F(ε − e) δ(ε − εk )|VkA |2
particular limit of nonadiabatic electron k
transfer, where the electron transfer rate is (4)
1 2 N
∆E
D A
(a)
1 2 N
∆E
N+1
0=D =A
(b)
Fig. 3 Simple level structure models (a) for molecular electron transfer and (b) for electron
transmission. The molecular bridge is represented by a simple set of levels that represent local
orbitals of appropriately chosen bridge sites. This set of levels is coupled to the donor and
acceptor species (with their corresponding nuclear environments) in (a), and to electronic
continua representing metal leads (say) in (b). In the latter case, the physical meaning of states 0
and N + 1 depends on the particular physical problem: They can denote donor and acceptor
states coupled to the continua of environmental states (hence the notation 0 = D, N + 1 = A),
surface localized states in an metal-molecule-metal junction, or they can belong to the right and
left scattering continua.

e−(λ−e+ε) /4λkB
2
where 2π
1 ket = dε √
f (ε) = (5) h̄ 4πλkB
1 + eε/kB
is the Fermi–Dirac distribution function × |V (ε)|2 f (ε) (7)
with ε measured relative to the electron
chemical potential µ in the electrode, Note that the reorganization energy that
and , which determines the position appears in Eq. (7) is associated with the
of the acceptor level relative to µ is the change in the redox state of the molecu-
overpotential. Defining lar species only. The nominal change in
the ‘‘oxidation state’’ of the macroscopic
δ(ε − εk )|VkA |2 ≡ |V (ε)|2 (6) electrode does not affect the polarization
k state of the surrounding solvent because
the electron transfer rate takes the the transferred electron or hole does not
form stay localized.
54 1 Fundamentals
Much of the early work on electron trans- in which the spacer is a well-characterized
fer have used expressions such as (1 and Langmuir–Blodgett film [59–61]. The
7) with the electronic coupling term VDA STM provides a natural apparatus for such
being used as a fitting parameter. More studies [58, 62–76]. Other approaches in-
recent work has focused on ways to char- clude break junctions [77–79] and mer-
acterize the dependence of this term on the cury drop contacts [80–84].
electronic structure of the donor/acceptor Simple theoretical modeling of VDA
pair and on the environment. In particular, usually relies on a single electron (or hole)
studies of bridge-mediated electron trans- picture in which the donor-bridge-acceptor
fer, where the donor and acceptor species (DBA) system is represented by a set of
are rigidly separated by molecular bridges levels: |D, |A, {|1, . . . |N } as depicted
of well-defined structure and geometry, in Fig. 3. In the absence of coupling
have been very valuable for characteriz- of these bridge states to the thermal
ing the interrelationship between structure environment, and when the energies En
and functionality of the separating envi- (n = 1, . . . , N ) are high relative to the
ronment in electron transfer processes. As energy of the transmitted electron (the
expected for a tunneling process, the rate donor/acceptor orbital energies in Figs. 1a,
is found to decrease exponentially with the 2a and 3a, or the incident electron energy
donor-acceptor distance in Figs. 1b to c, 2b and 3b), this is
the superexchange model for electron

ket = k0 e−β RDA (8) transfer [85]. Of particular interest are
situations where {n} are localized in space,
where β is the range parameter that char- so that the state index n corresponds to the
acterizes the distance dependence of the position in space between the donor and
electron transfer rate. The smallest val- acceptor sites (Fig. 3a) or between the two
ues for β are found in highly conjugated electron reservoirs (Fig. 3b). These figures
organic bridges for which β is in the depict generic tight binding models of this
range of 0.2 to 0.6 Å−1 [21–32]. In con- type, where the states n = 1, . . . , N are the
trast, for free space, taking a characteristic bridge states, here taken as degenerate
ionization
barrier UB = 5 eV, we find β = in zero order. Their localized nature
8mUB / h̄ ≈ 2.4 Å−1 (m is the electron
2
makes it possible to assume only nearest-
mass). Lying between these two regimes neighbor coupling between them, that is,
are many motifs, both synthetic and nat- Vn,n = Vn,n±1 δn ,n±1 . We recall that the
ural, including cytochromes and docked appearance of VDA in Eq. (1) is a low-
proteins [33–41], DNA [42–50], and sat- order perturbation theory result. A more
urated organic molecules [51–57]. Each general expression is obtained by replacing
displays its own characteristic range of β VDA by TDA where the T operator is
values, and hence its own timescales and defined by T (E) = V + V G(E)V , with
distance dependencies of electron trans- G(E) = (E − H + (1/2)i#)−1 and where
fer. A direct measurement of β along # stands for the inverse lifetime matrix
a single molecular chain was recently of bridge levels. Assuming for simplicity
demonstrated [58]. that the donor level |D is coupled only
In addition, to bridge assisted transfer to bridge state |1 and that the acceptor
between donor and acceptor species, elec- level |A is coupled only to bridge level N ,
tron transfer has been studied in systems the effective coupling between donor and
acceptor is given by These results imply a simple form for the

distance parameter β of Eq. (8)
TDA (E) = VDA + VD1 G1N (E)VN A (9)
2 %EB
β = ln (13)
This naturally represents the transition a VB
amplitude as the sum of a direct contri-
where a measures the segment size, so that
bution, VDA , which is usually disregarded
the bridge length is Na. The exponential
for long bridges and a bridge mediated con-
dependence on the bridge length is a man-
tribution. In using TDA instead of VDA in
ifestation of the tunneling character of this
Eq. (1), the energy parameter E in Eq. (9)
process. For typical values, for example,
should be taken to be equal to ED = EA
%EB /VB = 10 and a = 5 Å, Eq. (13) gives
at the point where the corresponding
β = 0.92 Å−1 . More rigorous estimates of
potential surfaces cross (or go through
the electronic coupling term in electron
an avoided crossing). In the level struc-
transfer processes involve electronic struc-
ture of Fig. 3(a) that corresponds to the ture calculation for the full DBA system.
DBA system in Fig. 1(a), making the tight Such calculations, in the context of molec-
binding approximation and in the weak ular conduction, will be discussed later.
coupling limit, max |V | min(EB − E),
the Green’s function element in (9) is 1.3.3
given by Transmission Between Conducting Leads

N −1
1 Vn,n+1 Equations (1, 7 and 11) are expressions
G1N (E) = (10)
E − EN E − En for the rate of electron transfer between
n=1
donor and acceptor molecules or between a
For a generalization of Eq. (10) that molecule and a metal electrode. As already
does not assume weak coupling see mentioned, for electron transfer in metal-
Refs. [86–88]. For a model with identical molecule-metal (MMM) junctions, the
bridge segments, En and Vn,n+1 are primary observable is the current–voltage
independent of n and will be denoted characteristics of the system. Putting it in
En = EB and Vn,n+1 = VB . Using this in another way, while the primary observable
Eq. (1) leads to in ‘‘standard’’ charge transfer processes in-
volving molecular donors and/or acceptors

2π V1D VN A 2 VB 2N is a transient quantity, in MMM junctions
ket = F
h̄ V %E we focus on the steady state current through
B B
(11) the junction for a given voltage difference
where %EB = EB − E. Similarly, for between the two metal ends. Note that for
a bridge-assisted transfer between a electron transfer processes, in addition to
molecule and an electrode, Eq. (7) applies rates, other observables are the yields of
with |V (ε)|2 given by different products of the electron transfer
2N reaction. Furthermore, for light induced
2 VB electron transfer processes, the steady state
|V (ε)| = δ(ε − εk )
%EB current under a constant illumination can
k
be monitored.
V1k VN A 2 Consider first a simple model for a

× (12)
VB metal-insulator-metal (MIM) system in
56 1 Fundamentals
which the insulator is represented by a and when → 0, f (E) − f (E + e) =

continuum characterized by a dielectric eδ(E − EF ). Equations (15 and 16) then
constant ε [89]. For specificity, assume that lead to an expression for the conduction
the electrode surfaces are infinite parallel per unit area, that is, the conductivity per
planes perpendicular to the x direction. unit length
In this case, the transmission problem is
4πme2 EF
essentially one-dimensional and depends σx = dEx T (Ex ) (17)
only on the incident particle velocity in (2π h̄)3 0
√
the x direction, υx = 2Ex /m. In the For finite , these expressions pro-
WKB approximation, the transmission vide a framework for predicting the
probability is given by current–voltage characteristics of the junc-
tion; explicit approximate expressions
4π s2 were given by Simmons [89]. Here we only
T (Ex ) = exp − [2m(UB (x)
h̄ s1 emphasize [89], that the dependence on
arises partly from the structure of Eqs. (15)
− Ex )]1/2 dx (14) and (16), for example, at zero temperature
EF −e
4πm2 e2
where UB (x) is the barrier potential that J = e dEx T (Ex )
determine the turning points s1 and s2 (2π h̄3 ) 0
EF
and m is the mass of the tunneling
particle. The√tunneling flux is given by + dEx (EF − Ex )T (Ex )
EF −e
T (Ex )n(Ex ) 2Ex /m, where n(Ex ) is the (18)
density per unit volume of electrons of but mainly from the voltage dependence of
energy Ex in the x direction. n(Ex ) is T. The simplest model for a metal-vacuum-
obtained by integrating the Fermi–Dirac metal barrier between identical electrodes
function with respect to Ey and Ez . When without an external field is a rectangular
a potential is applied so that the right barrier of height above the Fermi energy
electrode (say) is positively biased, the net given by the metal work function. When a
current density is obtained in the form [89] uniform electric field is imposed between
∞ the two metals, a linear potential drop
J = dEx T (Ex )ξ(Ex ) (15) from EF on one electrode to EF − e on
0 the other is often assumed (see Fig. 4). In
where addition, the image potential experienced
∞ ∞ by the electron between the two metals will
2m2 e
ξ(Ex ) = dυy dυz [f (E) considerably modify the potential barrier.
(2π h̄)3 −∞ −∞ For a point charge e, located at position
∞
4πme x between two conducting parallel plates
− f (E + e)] = dEr that are a distance d apart, the image po-
(2π h̄)3 0
tential is
× [f (E) − f (E + e)] (16)
e2
VI = −
and where Er = E − Ex = (1/2)m(υy2 + 4πε
υz2 ) is the energy in the direction per- ∞

1 nd 1
pendicular to x. In obtaining this result, × + −
it is assumed that the electrodes are 2x [(nd)2 − x 2 ] nd
n=1
chemically identical. At zero temperature, (19)
EF
EF EF
EF
Fig. 4 Tunneling gap between two metal electrodes in an unbiased

(left) and a biased (right) situation. The bare gap, given by the work
function W, is modified by the image interaction – the resulting
barriers are represented by the curved lines.
where ε is the dielectric constant of the The planar geometry implied by the
spacer. For x = d/2, this becomes assumption that transmission depends
only on the energy of the motion parallel
e2 ln 2 to the tunneling direction, as well as the
VI = − (20)
2πεd explicit form of Eq. (14) are not valid for a
This negative contribution to the elec- typical STM configuration that involves a
tron’s energy reduces the potential barrier tip on one side and a structured surface on
(Fig. 4), and has been invoked to explain the other. To account for these structures,
the lower than expected barrier observed Tersoff and Hamman [103] have applied
in STM experiments [90, 91]. Some points Bardeen’s formalism [104], which is a
should however, be kept in mind. First, perturbative approach to tunneling in
the classical result (19) fails close to arbitrary geometries. Bardeen’s formula
for the tunneling current is
the metal surface where quantum me-
chanical and atomic size effects change 4πe
I= [f (El )(1 − f (Er + e))
both the position of the reference im- h̄
l,r
age plane and the functional form of the
image potential [92–96]. Second, consid- − (1 − f (El ))f (Er + e)]|Mlr |2
eration of the dynamic nature of the image 2πe
× δ(El − Er ) = [f (El )
response should be part of a complete the- h̄
l,r
ory [97–100]. The timescale of electronic 2
response of metals can be roughly esti- − f (Er + e)]|Mlr | δ(El − Er ) (21)
mated from the plasma frequency to be where
∼10−16 s. This should be compared to the
time during which a tunneling particle can h̄2 l∗ ∇ψr − ψl ∇ψr∗ ) (22)
Mlr = dS(ψ
respond to interactions localized in the bar- 2m
rier. For transmitted particles, this is the is the transition matrix element for the
traversal time for tunneling [101, 102] that, tunneling process. This is just the Golden
for an electron traversing a 10 Å wide 1 eV rule rate expression multiplied by the
barrier is of the order of ∼1 fs. This com- electron charge e, with M playing the
parison would justify the use of the static role of coupling. In Ref. [103], only the
image approximation in this context, but first term in the square brackets of the
this approximation becomes questionable first line appears. This gives the partial
for deeper tunneling or narrower barriers. current from the negative to the positive
58 1 Fundamentals
electrode. The net current is obtained by 1.3.4

subtracting the reverse current as shown The Landauer Formula
in Eq. (21). Also, compared with Ref. [103],
Eq. (21) contains an additional factor of 2 The results (14)–(17) and (21)–(23) are
that accounts for the spin multiplicity of special cases of a more systematic rep-
the electronic states. resentation of the conduction and the
In Eqs. (21) and (22), ψ4 and ψr are current–voltage characteristic of a given
electronic eigenstates of the negatively junction due to Landauer [105, 106]. Lan-
biased (left) and positively biased (right) dauer’s original result was obtained for a
electrodes, respectively, is the bias system of two one-dimensional leads con-
potential, and the integral is over any necting two macroscopic electrodes (‘‘elec-
surface separating the two electrodes tron reservoirs’’) via a scattering object or
and lying entirely in the barrier region. a barrier characterized by a transmission
The wave functions appearing in Eq. (22) function T (E). The zero-temperature con-
are eigenfunctions of Hamiltonians that ductance, measured as the limit → 0 of
describe each electrode in the absence of the ratio I / between the current and the
the other, that is, interfaced with an infinite voltage drop between the reservoirs, was
spacer medium. These functions therefore found to be
decay exponentially in the space between
the two electrodes in a way that reflects e2
g= T (EF ) (25)
the geometry and chemical nature of the π h̄
electrodes and the spacer. For → 0,
Eq. (21) yields the conduction (The corresponding resistance, g −1 , can
be represented as a sum of the in-
I 4πe2 trinsic resistance of the scatterer itself,
g≡ = |Mlr |2 δ(El − EF )
h̄ [(e2 /π h̄)(T/(1 − T ))]−1 and a contribu-
l,r
tion (e2 /π h̄)−1 from two contact resis-
× δ(Er − EF ) (23) tances between the leads and the reser-
Tersoff and Hamman [103] have used voirs. See Chapter 5 of Ref. [107] for a
substrate wave functions that correspond discussion of this point.) This result is
to a corrugated surface of a generic metal obtained by computing the total unidi-
while the tip is represented by a spherical rectional current carried in an ideal lead
s orbital centered about the center r0 of the by electrons in the energy range (0, E) =
tip curvature. In this case, they find (0, h̄2 kE
2 /(2m)). In a one-dimensional sys-
tem of length L, the density of elec-

I ∝ |ψν (r0 )|2 δ(Eν − EF ) (24) trons, including spin, with wave vec-
ν tors in the range between k and k +
dk is n(k) dk = 2(1/L)(L/2π)f (Ek ) dk =
The rhs of Eq. (24) is the local density
f (Ek ) dk/π. The corresponding velocity is
of states of the metal. While this result is
υ = h̄k/m. Thus
useful for analysis of spatial variation of the
tunneling current on a given metal surface, kE kE
h̄k
the contributions from the coupling matrix I (E) = e dkυ(k)n(k) = e dk
0 0 m
elements in Eq. (23) cannot be disregarded
E
when comparing different metals and or f (Ek ) e
× = dE f (E ) (26)
different adsorbates [20]. π π h̄ 0
At zero temperature, the net current e2

carried under bias is g(E) = Tij (E) (31)
π h̄
∞ i,j
e
I = dE(f (E) − f (E + e)) As an example, consider the case of
π h̄ 0
a simple planar tunnel junction (see
→0 e2 Eqs. (14)–(17)), where the scattering pro-
−−−−→ (27)
π h̄ cess does not couple different transversal
modes. In this case, the transmission func-
Thus, the conductance of an ideal
tion depends only on the energy in the
one-dimensional lead is I / = e2 /π h̄ =
tunneling direction
(12.9K8)−1 . In the presence of the scat-
terer, this is replaced by Ly Lz
Tij (E) = Tii (E) = dky
∞ (2π)2
e i,j i
I = dE T (E)(f (E) − f (E + e))
π h̄ 0 h̄2 2 2
× dkz T E − (ky + kz )
→ 0, → 0 e2 2m
−−−−−−−−−→ T (EF ) (28)
π h̄ E
Ly Lz 2πm
= dEr T (E − Er ) (32)
which leads to (25). This result is valid for (2π)2 h̄2 0
one-dimensional leads. When the leads
Er is defined below Eq. (16). Using this
have finite size in the direction normal
in Eq. (29) yields the conductivity per
to the propagation so that they support
unit length
traversal modes, a generalization of (25) to
this case yields [108]
g 4πme2 EF
≡σ = dEx T (Ex )
Ly Lz h3
e2 e2 0
g= Tij (EF ) = Tr(SS † )EF (33)
π h̄ π h̄ in agreement with Eq. (17).
i,j
(29) Similarly, Eqs. (21 and 23) are easily seen
where Tij = |Sij |2 is the probability that to be equivalent to Eqs. (25 or 31) if we
a carrier coming from the left (say) of identify Mlr with Tlr in Eq. (37) below. An
the scatterer in transversal mode i will be important difference between the results of
transmitted to the right into transversal (29 and 31) and results based on Bardeen’s
mode j (Sij , an element of the S matrix, is formalism, that is Eqs. (21 to 23), is that the
the corresponding amplitude). The sum former are valid for any set of transmission
in (29) is over all transversal modes probabilities, even close to 1, while the
whose energy is smaller than EF . The latter yields a weak coupling result.
analog of Eq. (29) for the microcanonical Another important conceptual difference
chemical reaction rate was first written by is the fact that the sums over l and r in
Miller [109]. More generally, the current Eqs. (21 to 23) is over zero order states
for a voltage difference between the defined in the initial and final subspaces,
electrodes is given by while the sums in Eqs. (29 to 31) is over
∞ scattering states, that is, eigenstates of
g(E) the exact system’s Hamiltonian. It is the
I= dE[f (E) − f (E + e)]
0 e essence of Bardeen’s contribution [104]
(30) that in the weak coupling limit (i.e.
60 1 Fundamentals
high/wide barrier), it is possible to write (An additional factor of 2 on the rhs

the transmission coefficient Tij in terms accounts for the spin degeneracy). It is
of a golden rule expression for the convenient to recast this result in the form
transition probability between the zero
4πe ∞
order standing wave states |l and |r I= dE[f (E) − f (E + e)]
h̄ 0
localized on the left and right electrodes,
thus establishing the link between the × |Tlr |2 δ(E − El )δ(E − Er )
two representations. (For an alternative l,r
formulation of this link, see Ref. [110]) ∞ g(E)
To explore this connection on a more = dE[f (E) − f (E + e)]
0 e
formal basis, we can replace the expres- (36)
sion based on transmission coefficients where
T by an equivalent expression based on
scattering amplitudes, or T matrix ele- 4πe2
g(E) ≡ |Tlr |2
ments, between zero order states localized h̄
l,r
on the electrodes. This can be derived
× δ(E − El )δ(E − Er ) (37)
directly from Eqs. (29 or 31) by using
the identity Note that Eqs. (34 and 37) imply again
Eq. (31). For → 0, Eqs. (36 and 37) lead
Tij (E) = 4π 2 |Tlr |2 to I = g with
i,j l,r
g = g(EF ) (38)
× δ(E − El )δ(E − Er ) (34)
The analogy of this derivation of the result
On the left side of Eq. (34), a pair of (23) is evident.
indices (i, j ) denote an exact scattering
state of energy E, characterized by an 1.3.5
incoming state i on the left (say) electrode Molecular Conduction
and an outgoing state j on the right
electrode. On the right, l and r denote Equations (36 to 38) provide a convenient
zero order states confined to the left and starting point for most treatments of cur-
right electrodes, respectively. T is the rents through molecular junctions where
corresponding transition operator whose the coupling between the two metal elec-
particular form depends on the details trodes is weak. In this case, it is convenient
of this confinement. Alternatively, we can to write the system’s Hamiltonian as the
start from the golden-rule-like expression sum, H = H0 + V , of a part H0 that repre-
sents the uncoupled electrodes and spacer
4π and the coupling V between them. In the
I =e [f (El )(1 − f (Er + e))
h̄ weak coupling, limit the T operator
l,r
− f (Er + e)(1 − f (El ))]|Tlr |2 T (E) = V + V G(E)V ;

4πe G(E) = (E − H + iε)−1 (39)
× δ(El − Er ) = [f (El )
h̄
l,r
is usually replaced by its second term
2
− f (Er + e)]|Tlr | δ(El − Er ) (35) only. The first ‘‘direct’’ term V can be
disregarded if we assume that V couples

N
N
the states l and r only via states of the VB = Vn,n |nn | (44)
molecular spacer. Consider now a simple n=1 n =1
model where this spacer is an N -site and where, in the basis of eigenstates of
bridge connecting the two electrodes so HB0
that Site 1 of the bridge is attached to the
left electrodes and Site N – to the right ;nn (E) = δn,n (δn,1 + δn,N )[<n (E)
electrode, a variant of Fig. 3(b). In this
− 12 i#n (E)] (45)
case, we have Tlr = Vl1 G1N VN r , so that at
zero temperature [111] #n (E) = 2π |Vnj |2 δ(E − Ej ) (46)
j
Tij (E) = |G1N (EF )|2 ∞
PP #n (E )
i,j <n (E) = dE (47)
2π −∞ (E − E )
(L) (R)
× #1 (EF )#N (EF ) (40)
In Eq. (46), the sum is over both the right
and (using Eqs. 36 and 37) and the left manifolds (that is, j goes over
all states {l} and {r} in these manifolds),
(L) (R)
e EF
(L) so that #n = #n + #n ; n = 1, N . The
I () = dE|G1N (E, )|2 #1 (E) transmission problem is thus reduced
π h̄ EF −e
to evaluating a Green’s function matrix
(R)
× #N (E + e) (41) element and two width parameters. The
first calculation is a simple inversion of
with a finite (order N ) matrix. The width #
(L)
and the associated shift < represent the
#1 (E) = 2π |Vl1 |2 δ(El − E); finite lifetime of an electron on a molecule
l adsorbed on the metal surface, and

(R)
#N (E) = 2π |VN r |2 δ(EN − E) (42) can be estimated, for example Ref. [111],
r using the Newns–Anderson model of
chemisorption [112]. In the simple tight-
The Green’s function in Eq. (40) is it- binding model of the bridge and in the
self reduced to the bridge’s subspace weak coupling limit, G1N is given by
by projecting out the metals’ degrees of Eq. (10) modified by the inclusions of the
freedom. This results in a renormaliza- self-energy terms
tion of the bridge Hamiltonian: in the
bridge subspace G1N (E)
V1,2
=
(E − H + iη)−1 (E − E1 − ;1 (E))(E − EN − ;N (E))
−−−→ (E − HB − ;B (E))−1 (43)
N−1
Vj,j +1
× (48)
E − Ej
where HB = HB0 + VB is the Hamiltonian j =2
of the isolated bridge entity with
Equations (40 to 48) thus provide a
complete simple model for molecular

N
HB0 = En |nn|; conduction, equivalent to similar approx-
n=1 imations used in theories of molecular
62 1 Fundamentals
electron transfer (e.g. [113, 114] and refer- the process is dominated by the change
ences therein). For applications of variants in the nuclear configuration between the
of this formalism to electron transport in two localization states of the electron. Sup-
specific systems, see Refs. [86, 87, 115, pose now that the same DBA complex
116]. In the following text, we discuss more is used to connect between two metal
general forms of this formulation. contacts such that the donor and ac-
ceptor species are chemisorbed on the
1.3.6 two metals (denoted ‘‘left’’ and ‘‘right’’,
Relation to Electron Transfer Rates respectively). We wish to calculate the con-
duction of this junction and its relation
It is interesting to examine the rela- to kD→A . First note that the conduction
tionship between the conduction of a process does not involve localized states
molecular species and the electron trans- of the electron on the donor or the ac-
fer properties of the same species [117]. ceptor, so the factor F will be absent.
We should keep in mind that because (We will disregard energy loss arising
of tunneling there is always an Ohmic from transient distortions of the nuclear
regime near zero bias, with conduction configuration associated with transient
given by the Landauer formula. Obvi- populations of electronic states of the
ously, this conduction may be extremely DBA complex). Assuming as before that
low, indicating in practice an insulating states of the molecular complex are cou-
behavior. Of particular interest, is to esti- pled to the metal only via the D and A
mate the electron transfer rate in a given orbitals, and that the latter are coupled
DBA system that will translate into a only to their adjacent metal contacts, the
measurable conduction of the same sys- conduction is given by an equation sim-
tem when used as a molecular conductor ilar to Eq. (40), except that the bridge
between two metal leads. To this end, (1, . . . , N ) is replaced by the complex
consider a DBA system with a bridge DBA = (D, 1, . . . , N , A)
that consists of N identical segments (de- e2 (L) (R)
noted 1,2, . . . , N ) with nearest neighbor g(E) = |GDA (E)|2 #D (E)#A (E)
π h̄
coupling VB . The electron transfer rate (51)
is given by Eq. (11) that we rewrite in where, in analogy to Eq. (48)
the form
GDA (E)
2π
kD→A = |VD1 VN A |2 |G1N (ED )|2 F VD1 VN A
h̄ =
(49) (E − ED − ;D (E))(E − EA − ;A (E))
where, in the weak coupling limit, |VB | × G1N (E) (52)
|EB − E| (cf. Eq. 10)
Since the donor and acceptor species are
|VB |N −1 chemisorbed on their corresponding metal
G1N (E) = (50) contacts, their energies shift closer to the
(EB − E)N
Fermi energies. We assume that this shift
and where F is the Franck–Condon- occurs uniformly in the DBA complex
weighted density of nuclear states, given without distorting its internal electronic
in the classical limit by Eq. (3). The ap- structure (strictly speaking this can happen
pearance of F in Eq. (49) indicates that only in the symmetric case of identical
donor and acceptors and identical metal 1.3.7

electrodes, but the result of Eq. (53) below Quantum Chemical Calculations
is probably a good approximation for more
general cases because G1N (E) is often The simple models discussed above
not strongly dependent on E). Assuming are useful for qualitative understanding
therefore that the denominator in Eq. (52) of molecular conductivity; however, the
is dominated by the imaginary parts of the Landauer formula or equivalent for-
self-energies ;, we get mulations can be used as a basis for
more rigorous molecular calculations
16e2 |VD1 VN A |2 using extended Huckel (EH) based
g = g(EF ) = computations [70–72, 79, 119–133] or
π h̄ # (L) (EF )# (R) (EF )
D A Hartree–Fock (HF) [134–137]. These ap-
× |G1N (EF )| ;2
EF = ED = EA (53) proaches follow similar semiempirical and
ab initio calculations of electron transfer
Comparing to Eq. (49) we get rates in molecular systems [138]; how-
ever, instead of focusing on the com-
putation of the electronic coupling VDA
e2 kD→A 8 h̄
g= (54) needed in Eq. (1), the sum in Eq. (34)
π h̄ F π# (L) # (R) is calculated directly. Structural stability
D A
considerations suggest that useful MMM
It has been argued that, Eq. (52) and bridges should involve strong chemisorp-
therefore Eq. (54) holds the energy spac- tion bonding between the molecule and
ing EB − EF between the bridge levels the metal substrate, implying large elec-
and the Fermi energy is large relative tronic coupling between them [139]. It
to kB T , Eq. (54) holds also when the is, therefore, preferable to use a ‘‘su-
electron transfer process involves ther- permolecule’’ approach, in which the
mal activation into the bridge states (and quantum chemical calculations are car-
not only for the bridge assisted tun- ried out for a species that comprises
neling implied by Eq. 53) [118]. Using the molecule and two clusters of metal
the classical expression for F, Eq. (3), atoms, so that the reduction that intro-
we
√ have in the present case F = duces the self energy ; Eq. (43) is done
−1
( 4πλkB T ) exp(−λ/4kB T ). For a typ- at some deeper metal–metal contact. Such
ical value of the reorganization energy atomic level calculations usually start from
∼0.5 eV, and at room temperature, this a (nonorthogonal) basis set of atomic or-
is ∼0.02 (eV)−1 . Taking also #D = #A bitals, so the formalism described above
(L) (R)
−13
∼0.5 eV leads to g ∼ (e /π h̄)(10 kD→A
2 has to be generalized for this situation.
(s −1 )) ∼
= [10−17 kD→A (s −1 )] 8−1 . This sets (Alternatively, it has been shown by Em-
a criterion for observing Ohmic behav- berly and Kirczenow [140, 141] that one
ior for small voltage bias in molecular can map the problem into a new Hilbert
junctions: With a current detector sensi- space in which the basis states are orthog-
onal.) We also relax the assumption that
tive to pico-amperes, kD→A has to exceed
the molecule–metal contact is represented
∼106 s−1 (for the estimates of F and #
by coupling to a single molecular orbital.
given above) before measurable current
Defining the operator
can be observed at 0.1 V voltage across
such a junction. H(E) = EZ − H with Zij = i|j (55)
64 1 Fundamentals
The Green’s function is G(E) = H(E)−1 . electron representations such as the EH

In Eq. (55), i and j denote atomic orbitals approximation, and can be similarly done
that may be assigned to the supermolecule at the HF level using, after convergence,
(M), the left metal (L) and the right metal the Fock rather then the Hamiltonian
(R) subspaces. Denoting formally the matrix in Expressions (55 to 60). (Note that
coupling between the subspace M and the the Fock operator depends on the ground
subspaces K = L, R by the corresponding state electronic configuration. The latter
submatrices HMK , the Green’s function for is taken in Refs. [134, 135] to be that of
the supermolecule subspace is the isolated supermolecule, assuming that
the contact with the bulk electrodes does
G(M) (E) = (H − ; (L) − ; (R) )−1 (56) not affect it appreciably. In particular, the
supermolecule is usually assumed neutral
with ; (K) is a matrix in the molecular
in these calculations.)
subspace and
An important attribute of the approach
; (K) = HMK (H−1 )KK HKM (57) described above is that, within the approx-
imation used, it provides the total current
Eq. (57) is a compact notation for carried by the system, both through the
(; (K) )n,n = k,k Hnk (H−1 )kk Hk n where unoccupied molecular levels (electron con-
k and k are states in the metal K subspace. duction) and the occupied ones (hole
Using also conduction). This results from the fact that
the trace in Eq. (59) is over all the atomic or-
Tlr = Hln Gnn Hn r (58) bitals that comprise the (super)molecular
n,n basis set, that upon diagonalization in the
(l and r in the metal L and R subspaces, (super)molecular Hamiltonian will yield
respectively; n, n in the supermolecule both occupied and unoccupied molecu-
subspace) in Eq. (57) leads to lar orbitals. In a one-electron theory such
as the EH approximation, both types of
e2 orbitals contribute in the same way. For
g(E) = T r G(M) (E)# (R) (E)
π h̄ example, the terms in Eq. (59) that de-
scribe an electron moving from the highest
× G(M)† (E)# (L) (E) (59)
occupied molecular orbital (HOMO) into
where, for example, for the ‘‘left’’ metal empty states of the anode, followed by an
electron moving from the cathode into the
(L)
#n,n = 2π Hnl Hln δ(E − El ) HOMO (hole transport), and an electron
l moving from the cathode to the lowest
(n and n in the molecular subspace) unoccupied molecular orbital (LUMO),
(60) then moving on into the cathode (elec-
In practice, ; and # = −2Im(;) can tron transport), are similar (their values
be computed by using closure relations depend on the energies of the molecular
based on the symmetry of the metal orbitals involved with respect to the Fermi
lattice [130]. The trace in Eq. (59) is over energies), irrespective of whether the cor-
all basis states in the (super)molecular responding orbitals are occupied or not.
subspace. The evaluation of the Green’s The same is true in the HF calculation
function matrix elements and of this trace if the Koopman’s theorem [142], stating
is straightforward in semiempirical single that the HF orbital energies represent
the actual energies involved in removing calculations should be placed not in the
an electron from an occupied orbital or absolute numbers obtained, but rather in
putting an electron into an unoccupied highlighting the importance of the model
one, holds. The Koopman’s theorem is parameters in determining the junction
accurate only for large systems, and the conduction behavior. Examples for qual-
approximation involved in applying it to itative issues that were investigated with
small systems is one reason why HF is not these types of calculations include the ef-
necessarily superior to EH for calculating fect of the nature (length and conjugation)
the conduction properties of small molec- of the molecular bridge [126, 127], the ef-
ular junctions. This is true particularly fect of the molecule–electrode binding and
for LUMO dominated conduction, because of the molecular binding site [130], the re-
the HF is notoriously inadequate for elec- lation of conductance spectra to molecular
tron affinities [143–146]. See Ref. [137] for electronic structure [79] and the effect of
further discussion of this point. Another bonding molecular wires in parallel [128].
potential (but in principle avoidable) prob- (See also Ref. [147])
lem in these calculations is associated
with the finite, relatively small basis of 1.3.8
atomic orbitals used. Close to resonance, Spatial-grid Based Pseudopotential
when the electrode electrochemical poten- Approaches
tial µ = EF + e approaches the HOMO
or LUMO energies, the corresponding Another way to evaluate the expressions
HOMO or LUMO orbitals dominate the appearing in Eqs. (34 and 37) as well as
electron transfer and a small basis that related partial sums is closely related to the
correctly describes these orbitals is suffi- discrete variable representation of reaction
cient. When EF is a distance %E away probabilities as formulated by Seideman
from EHOMO or ELUMO , all molecular and Miller [148–150]. We have already
orbitals in the range %E below EHOMO seen that the sum
and in a similar range above ELUMO can
contribute to the transmission probabil- s(E) ≡ |Tlr |2 δ(E − El )δ(E − Er )
l,r
ity and cannot be ignored, implying the
(61)
need for a larger molecular basis [110]. which is related to the conduction by
We note in passing that the recently dis- g(E) = (4πe2 / h̄)s(E) (c.f Eq. 37) can be
cussed transmission antiresonances [140, represented by (c.f. Eq. 59)
141] associated with the nonorthogonality
of the atomic orbital basis sets, have been 1
s(E) = T r G(M) (E)# (R) (E)
shown [110] to be sometimes artifacts of a 4π 2
small basis calculation.
× G(M)∗ (E)# (L) (E) (62)
In spite of these limitations, EH- and
HF-based calculations have provided im- If, instead of considering transitions from
portant insight into the conduction proper- ‘‘left’’ to ‘‘right’’ electrode, we think
ties of molecular junctions. In view of other of Eq. (61) as expressing a sum over
unknowns, associated both with the un- transition probabilities from all initial
certainty about the junction structure and (i) states of energy E in the reac-
with the simplified computation, the main tant space to all final (f ) states of the
value of the results obtained from such same energy in the product space, s(E)
66 1 Fundamentals
is also associated with the so called a good numerical accuracy. Equation (62)
cumulative reaction probability [148–150], then becomes
which in terms of the reaction
S matrix
is defined by N (E) = i,f |Sif (E)|2 = s(E) = 4T r GABC (E)εR GABC∗ (E)εL
(63)
4π 2 s(E), that is, N (E) = i,f Tif (E).
where GABC (E) = (E − H + iε)−1 ; ε =
Equation (62) now expresses the important
εR + εL and where εR and εL are different
observation that the cumulative reaction
from zero only on grid points near the right
probability for a reactive scattering process
side (more generally the product side),
can be expressed as a trace over states,
and the left (reactant) side of the inner
defined in a finite subspace that contains the
subspace, respectively.
interaction region, of an expression that de-
A similar development can be done for
pends on the reduced Green’s function
the partial sum
and the associated self energy defined in

that subspace. Following Seideman and sl (E) ≡ |Tlr |2 δ(E − Er ) (64)
Miller we can use a spatial grid representa- r
tion for the states in this subspace, so that
the trace in Eq. (62) becomes a sum over which, provided that l is taken as an
grid points. Also, in this representation the eigenstate of the Hamiltonian describ-
overlap matrix Z is zero. In general, any ing the left electrode (or the reactant
subspace of position space that separate subspace), is related the ‘‘one to all’’
reactants from products (i.e. that encom- rate, kl (E) to go from an initial state
passes the entire interaction region; the of energy E on the left electrode (or in
molecular bridge in our application) can reactant space) to all possible states on
be used in Eq. (62), provided that the con- the right one (product space) according to
sequences of truncating the ‘‘rest of the kl = (2π/ h̄)sl . (The ‘‘microcanonical
rate’’
universe’’, expressed by the corresponding is defined by k(E) = ρL −1 (E) l kl δ(E −
; and # can be computed. The absorb- El ) = (2π h̄ρL (E))−1 4π 2 s(E) = (π h̄ρL
ing boundary condition Green’s function (E))−1 N (E).) We use the same definition
(ABCGF) method of Seideman and Miller of the coupling V between our subspace
is based on the recognition that if this (bridge) and the reactant and product (elec-
subspace is taken to be large enough so trode) states. Putting T = V GV in Eq. (64)
that its boundaries are far from the inter- we get
action region, the detailed forms of ; and 1
# are not important; the only requirement sl (E) = l|V G(M) # (R) G(M)∗ V |l (65)
2π
is that scattered waves that approach these
boundaries will be absorbed and not re- Using again a position grid representation
flected back into the interaction zone. In of the intermediate states used to evaluate
the ABCGF method, this is accomplished this expression, and applying the same
by taking ; = −(1/2)i# = −iε(r), a local methodology as above, Eq. (65) can be
function in position space, taken to be zero recast in the form
in the interaction region and gradually in- 1
sl (E) = l|V GABC (E)εR GABC∗ (E)V |l
creasing from zero when approaching the π
subspace boundaries. Its particular form 1
= l|εL GABC (E)εR GABC∗ (E)εL |l
is chosen to effect complete absorption π
of waves approaching the boundary to (66)
The second part of Eq. (66) is ob- Green’s function of the bare electrode sys-
tained by using the identity εR |l = tem and δV is the difference between
0 to write εR G∗ V |l = εR (1 + G∗ V )|l = the potential of the full system contain-
εR G∗ (G∗−1 + V )|l, which, together with ing an atomic or a molecular spacer and
G∗−1 = E − H0 − V + iε, (E − H0 )|l = 0 that of the bare electrodes. In atomic units
and ε|l = εl |l, yields the desired result. (|e|, η, m = 1), it is
The results (63 and 66) are very useful

for computations of transmission proba- δV (r, r ) = Vps (r, r ) + Vxc (n(r)) − Vxc
bilities in models in which the interaction

between the transmitted particle and the δn(r )
× n0 (r) + dr (68)
molecular spacer is given as a position de- |r − r |
pendent pseudopotential. Applications to
where Vps is the sum of nonlocal pseu-
electron transmission through water and
dopotentials representing the cores of
other molecular layers are discussed in
the spacer atoms and Vxc is the lo-
Refs. [151–158].
cal density approximation (LDA) of the
1.3.9
exchange-correlation potential. n is the
Density Functional Calculations electron number density for the full system
(electrodes and atoms) and δn = n − n0 .
Density functional methods provide a con- Equation (67) yields scattering states that
venient framework for treating metallic in- can be labeled by their energy E, mo-
terfaces [100]. Applications of this method- mentum k in the direction (yz) parallel
ology to the problem of electron transport to the electrodes and spin. In addition,
through atomic and molecular bridges Lang distinguishes between wave func-
have been advanced by several workers. In tions that in the electrode regions carry
particular, Lang’s approach [90, 159–165] positive (+) or negative (−) momentum
in the tunneling direction. Denoting by
is based on the density functional formal-
µL and µR the electron electrochemical
ism [166, 167] in which the single electron
potential in the left and right electrode,
wave functions ψ0 (r) and the electron den-
respectively, the zero-temperature electri-
sity n0 (r) for two bare metal (jellium)
cal current density from left to right (for
electrodes is computed, then used in the
µL > µR ) is then
Lippman–Schwinger equation
µR
∗
J (r) = −2 dE d2 K I m{ψ+ ∇ψ+ }
ψ(r) = ψ0 (r) + dr dr G0 (r, r )δV µL
(69)
× (r , r )ψ(r ) (67) The factor 2 accounts for the double oc-
cupancy of each orbital. This approach
to get the full single electron scattering was used recently [169] to calculate cur-
wave functions ψ(r) in the presence of rent through a molecular species, Benzene
the additional bridge. (Lang’s earlier cal- 1,4-dithiolate molecule (as used in the ex-
culations [90, 168] use a related density periment described in Ref. [67]), between
functional approach to calculate the tun- two jellium surfaces. The result demon-
neling current between an atomic tip and strates the large sensitivity of the computed
a jellium electrode without an atomic or current to the microscopic structure of the
molecular bridge). In Eq. (67), G0 is the molecule–metal contacts.
68 1 Fundamentals
A similar density functional approach, To end this brief overview of density-

using an atomic level description of the functional-based computations of molecu-
electrodes, was described by Di ven- lar conduction, we should note that this
tra and Pantelides [170]. These authors approach suffers in principle from prob-
used density-functional based ground state lems similar to those encountered in using
molecular dynamics [171] in order to get the HF approximation, that is, the inher-
the relaxed structure of the metal-atomic ent inaccuracy of the computed LUMO
system-metal junction; then evaluated the energy and wave functions. The errors
current through the relaxed structure. are different, for example HF overesti-
The density-functional based calcula- mates the HOMO–LUMO gap (since the
tions described above were done for small HF LUMO energy is too high [143–146,
applied potential bias between the elec- 174–176]) while density functional theory
trodes. In contrast, the density functional (DFT) underestimates it [167, 177]. Com-
approach of Hirose and Tsukada [172] mon to both approaches is the observation
calculates the electronic structure of a that processes dominated by the HOMO
metal-insulator-metal system under strong level will be described considerably better
applied bias. The main difference from the by these approaches than by processes con-
density functional approaches described trolled by coupling to the LUMO [137, 178].
above comes in the way the effective
one-electron potential is calculated. The 1.3.10
potential used in this work contains the Potential Profiles
usual contributions from the Coulomb
and the exchange-correlation interactions The theoretical and computational ap-
as well as from the ionic cores. How- proaches described above are used to
ever, the Coulomb (Hartree) contribution compute both the Ohm-law conduction,
is obtained from the solution of a Pois- g(EF ) of a molecular bridge connect-
son equation ing two metals, (Eqs. 37 or 59), and
the current–voltage characteristics of the
∇ 2 VH (r) = −4π[ρ(r) − ρ+ (r)] (70) junction, also beyond the Ohmic regime
(Eq. 36). We should keep in mind that
in the presence of the applied potential these calculations usually disregard a po-
boundary conditions. ρ+ (r) is the fixed tentially important factor – the possible
positive charge density, and the electron effect of the imposed electrostatic field on
density ρ(r) is constructed by summing the nuclear structure as well as on the elec-
the squares of the wave functions over tronic structure of the bridge. A change
the occupied states. At the same time, in nuclear configuration under the im-
the exchange-correlation potential is cal- posed electrostatic field is, in fact, not very
culated in the standard LDA, neglecting likely for stable chemisorbed molecular
the effect of the finite current that exists in bridges. On the other hand, the electronic
the steady state system. The resulting for- wave functions can be distorted by the
malism thus accounts approximately for imposed field, and this in turn may af-
nonequilibrium effects within the density fect the electrostatic potential distribution
functional calculation. (A simplified ver- along the bridge, the electronic coupling
sion of the same methodology has recently between bridge segments and the position
been presented by Mujica et al. [173].) of the molecular energy levels vis-a-vis
the metal’s Fermi energies. (In a single under discussion ever since Aviram and
electron description, the local electrostatic Ratner [179] suggested that an asymmetric
potential will be an input, associated with donor-bridge-acceptor system connecting
the underlying many-electron response of two metal leads could rectify current. The
the molecular bridge, to the position de- proposed mechanism of operation of such
pendent energies of the bridge electronic a device is shown in Fig. 5. When the left
states in the site representation.) These electrode is negatively biased, that is, the
effects were, in fact, taken approximately corresponding electrochemical potentials
into account by Hirose and Tsukada [172] satisfy µL > µR as shown, electrons can
and by Mujica, Roitberg and Ratner [173] move from this electrode to the LUMO of
by solving simultaneously, the coupled molecular segment A, as well as from the
Schrödinger and Poisson equations. The HOMO of molecular segment D, to the
latter yields the electrostatic potential for right electrode. Completion of the transfer
the given electron density and under the by moving an electron from A to D is as-
imposed potential boundary conditions. sisted by the intermediate bridge segment
The importance of the electrostatic po- B. When the polarity of the bias is reversed,
tential profile on the molecular bridge the same channel is blocked. This simple
in determining the conduction proper- analysis is valid only if the molecular en-
ties of a metal-molecule-metal junction ergy levels do not move together with the
was recently discussed by Tian et al [72] metal electrochemical potentials, and if the
in conjunction with the current–voltage coupling through the intermediate bridge
characteristics of a junction comprised of is weak enough so that the orbitals on the D
an STM tip, a gold substrate and a molecule and A species maintain their local nature.
with two bonding sites (e.g. α, α -xylyl Other models for rectification in molec-
dithiol) connecting the two. ular junctions have been proposed [180].
As discussed above, the expected recti-
1.3.11 fying behavior can be very sensitive to
Rectification the actual potential profile in the ABD
complex, which in turn depends on the
The possibility of constructing molecular molecular response to the applied bias [72,
junctions with rectifying behavior has been 181]. This explains why rectification is
Fig. 5 A model for current rectification

in a molecular junction: Shown are the
chemical potentials µL and µR in the mL
two electrodes, and the HOMO and
LUMO levels of the donor, acceptor,
and bridge. When the right electrode is
positively biased, (as shown) electrons D
can hop from left to right as indicated by mR
A
the dotted arrows. If the opposite bias B
can be set without affecting the
electronic structure of the DBA system
too much, the reverse current will
be blocked.
70 1 Fundamentals
often not observed even in asymmetric two metals, intercarrier interactions ap-
molecular junctions [181]. However, recti- pear as suppression of current owing to
fication has been observed in a number film charging [186, 187]. In nano-junctions
of metal-molecule-metal junctions, as well involving double barrier structures, in-
as in several STM experiments involving creased electron population in the interme-
adsorbed molecules [61, 65, 182–185]. diate well under resonance transmission
should affect the transport process for sim-
1.3.12 ilar reasons. For example, consider a small
Carrier–carrier Interactions metal sphere of radius R in the space be-
tween two metal electrodes (Fig. 6), and
The models and calculations discussed assume that both sphere and electrodes
so far focus on processes for which the are made of the same metal of work
probability that a charge carrier populates function W. Neglecting possible proxim-
the bridge is low so that carrier–carrier ity effect of these electrodes, the classical
energy for removing an electron from
interactions can be disregarded. Elec-
the sphere to infinity is W + e2 /2R and
tron–electron interactions were taken into
the classical energy for the opposite pro-
account only in so far that they affected
cess is W − e2 /2R. (From experimental
the single electron states, either in con-
and theoretical work on ionization poten-
structing the molecular spectrum (in the
tials of small metal clusters [188] we know
ab initio HF or DFT calculations) or in that the actual energies are approximately
affecting the junction electrostatic poten- W + 0.4e2 /R and W − 0.6e2 /2R, respec-
tial through the electronic polarization tively; with the differences arising from
response of the molecule or the metal con- quantum size effects.) Here, the sphere
tacts. When the density of carriers in the plays the role of a molecular bridge in
space between the metal contacts becomes assisting electron tunneling between the
large, Coulomb interactions between them two electrodes, and these energies now
have to be taken into account explicitly. play the same role as the corresponding
Here we briefly discuss the effect of such HOMO and LUMO energies of the bridge.
interactions. This implies that a finite voltage differ-
In classical (hopping) transport of car- ence is needed before current can flow in
riers through insulating films separating this sphere-assisted mode between the two
EF + e2/(2R)
EF EF
EF − e2/(2R)
Fig. 6A nano-dot between two

conducting leads: A model for
Coulomb blockade phenomena.
metals, a phenomenon known as Coulomb Hubbard model with [199–201]or with-

blockade. For a larger potential bias, other out [202] the mean field approximation.
conduction channels, corresponding to In particular, Malysheva and Onipko have
more highly charged states of the sphere derived a tight binding analog of the
give rise to the phenomenon of Coulomb model for negative differential resistance
steps [189]. For experimental manifesta- originally proposed by Davydov and Er-
tions of such and related phenomena, makov [203] (see also Refs. [204–206]).
see, for example, Refs. [190–196]. The pos- Numerical simulations [207] can assist in
sibility to observe such phenomena in gauging the performance of the mean field
electrochemical systems was discussed by approximations used in these calculations.
Kuznetsov and Ulstrup [197] and possibly Such models may be relevant to the un-
demonstrated by Fan and Bard [198]. derstanding of recent experimental obser-
When the junction consists of a vation of negative differential resistance in
molecule or a few molecules connecting a metal-self-assembled monolayer–metal
two metal leads, such Coulomb blockade junction with the SAM containing a ni-
phenomena are not expected to appear troamine redox center [208].
so clearly. The first Coulomb threshold is We conclude this discussion by empha-
replaced, as just described, by the gap as- sizing again that understanding correlated
sociated with the position of the metal’s carrier transport in molecular junctions
Fermi energies relative to the molecular continues to be an important experimen-
HOMO and LUMO levels (modified by tal and theoretical challenge. Recent work
appropriate electron correlations). How- by Gurvitz et al [209–211], using exactly
ever, the discreteness (in the sense that solvable models of electron transport in
%E kB T ) of the molecular spectrum two and three barrier structures, has indi-
implies that for any given charging state of cated that new phenomenology may arise
the molecule, for example, a molecule with from the interplay of inelastic transitions
one excess electron or one excess hole, and intercarrier interactions in the barrier.
there will be several distinct conduction In fact, dephasing transitions in the barrier
channels that will appear as steps in the may prove instrumental in explaining the
current versus voltage plot. It will be hard charge quantization that gives rise to the
to distinguish between this structure and single electron transport behavior of such
between a ‘‘genuine’’ Coulomb blockade junctions (Sect. 6.3 in Ref. [212]).
structure. It should be emphasized that
for potential applications, such as using 1.3.13
the molecular junction in single electron Some Open Issues
transistor devices, the distinction between
the origins of these conduction structures This section discusses some subtle dif-
is in principle, not important. ficulties that are glossed over in most of
However, understanding the role played the treatments of electron transmission us-
by electron–electron interactions (in paring the formalisms described above. These
ticular, correlation effects beyond the should be regarded as open theoretical is-
HF approximation) remains an important sues that should be addressed in future de-
challenge in the study of molecular nano- velopments. The source of these problems
junctions. Several recent theoretical works is our simplified treatment of what is ac-
have addressed this problem within the tually a complex many-body open system.
72 1 Fundamentals
In particular, common ways of incorpo- have a choice: either imposing the reac-
rating many-body-effect using single-body tion field on the electronic Hamiltonian
effective potentials become questionable in the position representation, thus modi-
in particular limits of timescales and inter- fying all Coulomb interaction terms, then
action strengths. calculate the electronic wave functions un-
One such issue, already mentioned, is der the new potential, or compute the
the use of static image to account for electronic wave functions with the original
the effect of metal polarizability (that Hamiltonian under the imposed dielectric
is, the response of the metal electrons) boundary conditions. The fact that the two
on charge transfer processes at metal representations are not equivalent is as-
surfaces. Timescale estimates for electron sociated with the approximate nature of
tunneling in molecular systems are of the the approach, which replaces a detailed
same order as metal plasma frequencies treatment of the electronic structure of the
that measure the electronic response solvent by its electronic dielectric response
time of metals. However, static image and with the fact that the Schrödinger
theory has been used in the analysis equations derived from them are nonlin-
of Sect. 1.3.3, and in other treatments ear in the electronic wave functions).
of electron injection from metals into Examination of the energies and time-
scales involved suggests that assuming
insulating phases [213]. To what extent
instantaneous metal response to the elec-
dynamic image effects are important is not
tron position is more suitable in most
known, though theories that incorporate
situations than taking instantaneous re-
such effects have been developed [97–100].
sponse to the charge distribution defined
Assuming that image interactions at
by a molecular orbital, but the correspond-
metal surfaces should be accounted for
ing timescales are not different enough to
in the static limit, namely that the metal
make this a definite statement.
responds instantaneously to the tunnel-
A similar issue appears in attempts to
ing charge, opens other questions. Many account for the electronic polarizability
calculations of electronic processes near of a solvent in treating fast electronic
metal surfaces [e.g. [89] (See Sect. 1.3.3 processes involving solute molecules or
above)] assume that the metal electrons re- excess electrons in this solvent. For
spond instantaneously to the position of the example, in treating electron transmission
tunneling electron. Other calculations use in MIM junctions, the potential barrier that
atomic or molecular orbitals, or more gen- enters into expressions such as Eq. (14)
eral electronic charge distributions, and depends on the electronic structure of the
computing these under the given potential insulating spacer. For vacuum tunneling,
boundary conditions (see, e.g. Ref. [178]) a rectangular barrier, whose height above
implies that the corresponding orbitals or the metal Fermi energy is the metal work
charge distributions are well defined on function, modified by image interactions
timescales shorter than the metal response as discussed above and in Sect. 1.3.3,
times. (Computing molecular orbitals self- seems appropriate. For a dielectric spacer,
consistently with image interactions is the the barrier should be further modified
common practice in quantum chemistry by the fast (electronic) dielectric response
calculations for solvated molecules using of this spacer in the same way that it
reaction field (cavity) models. Again we is modified by the electronic response of
the metal, raising issues similar to those hand, it appears that for weakly coupled
discussed above. contacts the perturbative approach that
An interesting point of concern is related leads to Eq. (21) is valid. This approach
to the way the Fermi distribution functions describes the transmission in terms of
enter into the current equations. For exam- electron states localized on the two elec-
ple, Bardeen’s transmission formula (21) trodes where unidirectional rates appear
is based on weak coupling between states with f (1 − f ) factors, and can in principle
localized on the two electrodes, the par- be carried over to the inelastic regime. The
tial or unidirectional currents contain a exact correspondence between these differ-
product, f (1 − f ),that is, the probability ent representations needs further study.
that the initial state is occupied multi- Finally, the theoretical and computa-
plied by the probability that the final state tional issues discussed in this chapter are
is not. In this viewpoint, the transitions based on the assumption that the electron
occur between two weakly coupled sys- tunnel factors do not depend on the inter-
tems, each of them in internal thermal action between the molecular system and
equilibrium, which are out of equilibrium its thermal environment. This assumption
with each other because of the potential covers one important scenario of electron
bias. Alternatively, we could work in the transmission. On the other extreme are
basis of exact eigenstates of the whole sys- systems in which electron transfer is ther-
tem comprising the two electrodes and mally activated and long range electron
the spacer between them. This system transfer is dominated by hopping. An im-
is in an internal nonequilibrium state in portant example is provided by the recent
which transmission can be described as results on long range electron transport
a scattering problem. The relevant eigen- in DNA [215–219]. The transition from
states correspond to incident (incoming)
tunneling to hopping and other issues
waves in one electrode and transmitted
involving thermal effects in long-range
waves in the other. The flux associated
electron transfer have recently been dis-
with those scattering states arising from
cussed by several workers [220].
an incident state in the negatively biased
electrode is proportional to f (E), while
that associated with incoming waves in the Acknowledgments
positively biased electrode is proportional
to f (E + e). The net flux is therefore, This work was supported by the the
found again to be proportional to the differ- Israel Science Foundation, the U.S. Israel
ence f (E) − f (E + e). This argument Binational Science foundation and by the
cannot be made unless the process can Israel Ministry of Science.
be described in terms of coherent scatter-
ing states defined over the whole system. References
When inelastic scattering and dephasing
processes take place, the description in 1. J. Jortner, M. Bixon, (Eds.), Advances in
terms of exact scattering states of the Chemical Physics: Electron Transfer – from
Isolated Molecules to Biomolecules, John
whole system becomes complicated [212,
Wiley & Sons, New York, 1999, Vol. 106.
214], although kinetic equations for elec- 2. A. M. Kuznetsov, Charge Transfer in Physics,
tron transport can be derived for relatively Chemistry and Biology, Gordon & Breach,
simple situations [209–211]. On the other New York, 1995.
74 1 Fundamentals
3. A. M. Kuznetsov, J. Ulstrup, A. M. Kuzne, 27. S. Woitellier, J. P. Launay, C. W. Spangler,

T. Sov, Electron Transfer in Chemistry and Inorg. Chem. 1989, 28, 758.
Biology: An Introduction to the Theory, John 28. A. Helms, D. Heiler, G. Mclendon, J. Am.
Wiley & Sons, New York, 1998. Chem. Soc. 1992, 114, 6227.
4. D. R. Lamb, Electrical Conduction Mecha- 29. L. M. Tolbert, Acc. Chem. Res. 1992, 25, 561.
nisms in Thin Insulating Films, Methuen, 30. T. S. Arrhenius, M. Balnchard-Desce,
London, 1967. M. Dvolaitzky et al., Proc. Natl. Acad. Sci.
5. R. J. Miller, G. McLendon, A. Nozik et al., 1986, 83, 5355.
Surface Electron Transfer Processes, VCH 31. P. Finkh, H. Heitele, M. Volk et al., J. Phys.
Publishers, New York, 1995. Chem. 1988, 92, 6584–6590.
6. F. Schmickler, Interfacial Electrochemistry, 32. R. F. Ziessel, Chem. Educ. 1997, 74,
Oxford University Press, Oxford, 1996. 673–679.
7. R. M. Metzger in Molecular Electronics – Sci- 33. M. J. Bjerrrum, D. R. Casimiro, I. J. Chang
ence and Technology (Ed.: A. Aviram), Amer- et al., J. Bioenerg. Biomembr. 1995, 27, 295.
ican Institute of Physics Conference Pro- 34. R. Langen, I. J. Chang, J. P. Germanas et al.,
ceedings, New York, 1992, Vol. 262, p. 85. Science 1995, 268, 1733–1735.
8. C. A. Mirkin, M. A. Ratner, Annu. Rev. 35. J. R. Winkler, H. B. Gray, Chem. Rev. 1992,
Phys. Chem. 1992, 43, 719–754. 92, 369.
9. K. Sienicki, (Ed.), Molecular Electronics and 36. M. Y. Ogawa, J. F. Wishart, Z. Young et al.,
Molecular Electronic Devices, CRC Press, J. Phys. Chem. 1993, 97, 11 456.
New York, 1994. 37. S. S. Isied, M. Y. Ogawa, J. F. Wishart,
10. M. C. Petty, M. R. Bryce, D. Bloor, (Eds.), Chem. Rev. 1992, 92, 381.
An Introduction to Molecular Electronics, 38. K. S. Schanze, L. A. Cabana, J. Phys. Chem.
Oxford University Press, Oxford, 1995. 1990, 94, 2740.
11. C. Joachim, S. Roth, (Eds.), Atomic and 39. J. M. Vanderkooi, S. W. Englander, S. Papp
Molecular Wires, Kluwer, Dordrecht, The et al., Proc. Natl. Acad. Sci. 1990, 87, 5099.
Netherlands, 1997, Vol. 341. 40. C. E. Moser, C. C. Page, R. Farid et al., J.
12. J. Jortner, M. Ratner, (Eds.), Molecular Elec- Bioenerg. Biomembr. 1995, 27, 263–274.
tronics, Blackwell Science, Oxford, 1997. 41. M. W. Mutz, M. A. Case, J. F. Wishart et al.,
13. L. Kouwenhoven, Science 1997, 275, 1896, J. Am. Chem. Soc. 1999, 121, 858, 859.
1897. 42. F. D. Lewis, T. F. Wu, Y. F. Zhang et al.,
14. A. Aviram, M. Ratner, (Eds.), Molecular Science 1997, 277, 673–676.
Electronics: Science and Technology, New 43. R. E. Holmlin, P. J. Dandliker, J. K. Barton,
York Academy of Sciences, New York, 1998. Angew. Chem., Int. Ed. Engl. 1997, 36, 2714.
15. C. Dekker, Phys. Today 1999, 52, 22–28. 44. P. Lincoln, E. Tuite, B. Norden, J. Am.
16. M. A. Reed, Proc. IEEE 1999, 87, 652–658. Chem. Soc. 1997, 119, 1454, 1455.
17. A. G. Davies, Philos. Trans. R. Soc. London 45. A. M. Brun, A. Harriman, J. Am. Chem. Soc.
Ser. A – Phys. Sci. Eng. 2000, 358, 151–172. 1994, 116, 10 383.
18. R. Landauer, IEEE Trans. Electron Devices 46. T. J. Meade, J. F. Kayyem, Angew. Chem.,
1996, 43, 1637–1639. Int. Ed. Engl. 1995, 34, 352–354.
19. R. A. Marcus, J. Chem. Phys. 1965, 43, 679. 47. P. F. Barbara, E. J. C. Olson, Experimental
20. S. Gosavi, R. A. Marcus, J. Phys. Chem. electron transfer kinetics in a DNA envi-
2000, 104, 2067–2072. ronment in Advances in Chemical Physics:
21. M. R. Wasielewski, M. P. Niemczyk, D. G. Electron Transfer – from Isolated Molecules
Johnson et al., Tetrahedron 1989, 45, 4785. to Biomolecules, Pt 2 (Eds.: M. Bixon,
22. S. B. Sachs, S. P. Dudek, L. R. Sita et al., J. J. Jortner), Horizon Pubs & Distributors
Am. Chem. Soc. 1997, 119, 10 563. Inc., 1999, pp. 647–676, Vol. 107.
23. V. Grosshenny, A. Harriman, R. Ziessel, 48. K. Fukui, K. Tanaka, Angew. Chem., Int. Ed.
Angew. Chem., Int. Ed. Engl. 1996, 34, 2705. Engl. 1998, 37, 158–161.
24. J. M. Tour, Chem. Rev. 1996, 96, 537. 49. E. Meggers, M. E. Michel-Beyerle, B. Giese,
25. A. Osuka, N. Tanade, S. Kawabata et al., J. J. Am. Chem. Soc. 1998, 120, 12 950–12 955.
Org. Chem. 1995, 60, 7177. 50. E. Meggers, D. Kusch, M. Spichty et al.,
26. A. C. Ribou, J. P. Launay, K. Takahashi Angew. Chem., Int. Ed. Engl. 1998, 37,
et al., Inorg. Chem 1994, 33, 1325. 460–462.
51. K. P. Ghiggino, A. H. A. Clayton, J. M. 75. Y. Manassen, D. Shachal, Scanning tun-

Lawson et al., New J. Chem. 1996, 20, 853. neling microscopy (STM) on physisorbed
52. H. Oevering, M. N. Paddon-Row, M. Hepp- chemical groups of individual immobilized
ener et al., J. Am. Chem. Soc. 1987, 109, molecules, Molecular Electronics: Science and
3258. Technology, 1998, Vol. 852, pp. 277–289.
53. M. T. Carter, G. K. Rowe, J. N. Richardson 76. G. Nagy, J. Electroanal. Chem. 1996, 409,
et al., J. Am. Chem. Soc. 1995, 117, 2896. 19–23.
54. H. L. Guo, J. S. Facci, G. Mclendon, J. Phys. 77. J. M. Tour, L. Jones, D. L. Pearson et al., J.
Chem. 1995, 99, 8458. Am. Chem. Soc. 1995, 117, 9529–9534.
55. B. Paulson, K. Pramod, P. Eaton et al., Phys. 78. C. Zhou, C. J. Muller, M. A. Reed et al.,
Chem. 1993, 97, 13 042. Mesoscopic phenomena studied with me-
56. M. D. Johnson, J. R. Miller, N. S. Green chanically controllable break junctions at
et al., J. Phys. Chem. 1989, 93, 1173. room temperature in Molecular Electronics
57. C. A. Stein, N. A. Lewis, G. Seitz, J. Am. (Eds.: J. Jortner, M. Ratner), Blackwell Sci-
Chem. Soc. 1982, 104, 2596. ence, Oxford, 1997, pp. 191–213.
58. V. J. Langlais, R. R. Schlittler, H. Tang et al., 79. C. Kergueris, J. P. Bourgoin, S. Palacin
Phys. Rev. Lett. 1999, 83, 2809–2812. et al., Phys. Rev. B: Condens. Matter 1999,
59. E. E. Polymerpopoulos, D. Mobius, 59, 12 505–12 513.
H. Kuhn, Thin Solid Films 1980, 68, 173. 80. J. D. Porter, A. S. Zinn, J. Phys. Chem. 1993,
60. M. Fujihira, K. Nishiyama, H. Yamada, Thin 97, 1190–1203.
Solid Films 1985, 132, 77. 81. K. Slowinski, R. V. Chamberlain, C. J.
61. S. Roth, M. Burghard, C. M. Fischer, Res- Miller et al., J. Am. Chem. Soc. 1997, 119,
onant tunneling and molecular recti- 11 910–11 919.
fication in Langmuir-Blodgett films in 82. K. Slowinski, K. U. Slowinska, M. Majda, J.
Molecular Electronics (Eds.: J. Jortner, Phys. Chem. 1999, 103, 8544–8551.
M. Ratner), Blackwell Science, Oxford, 83. K. Slowinsky, H. K. Y. Fong, M. Majda, J.
1997, pp. 255–280. Am. Chem. Soc. 1999, 121, 7257–7261.
62. E. Delamarche, B. Michel, H. A. Biebuyck 84. R. E. Holmin, R. F. Ismagilov, R. Haag,
et al., Adv. Mater. 1996, 8, 718. et al., Angewandte Chemie 2001, 40, 2316.
63. L. A. Bumm, J. J. Arnold, M. T. Cygan et al., 85. H. M. McConnel, J. Chem. Phys. 1961, 35,
Science 1996, 271, 1705–1707. 508–515.
64. U. Durig, O. Zuger, B. Michel et al., Phys. 86. A. Onipko, Y. Klymenko, L. Malysheva et al.,
Rev. B 1993, 48, 1711. Solid State Commun. 1998, 108, 555–559.
65. A. Dhirani, P. H. Lin, P. Guyot-Sionnest, J. 87. A. Onipko, Y. Klymenko, J. Phys. Chem. A
Chem. Phys. 1997, 106, 5249–5253. 1998, 102, 4246–4255.
66. L. Ottaviano, S. Santucci, S. D. Nardo et al., 88. B. L. Burrows, A. T. Amos, S. G. Davison,
J. Vac. Sci. Technol., A 1997, 15, 1014. Int. J. Quantum Chem. 1999, 72, 207–220.
67. M. A. Reed, C. Zhou, C. J. Muller et al., 89. J. G. Simmons, J. Appl. Phys. 1963, 34,
Science 1997, 278, 252–254. 1793–1803.
68. D. Porath, O. Millo, J. Appl. Phys. 1997, 81, 90. N. D. Lang, Phys. Rev. B 1987, 36, 8173.
2241–2244. 91. N. D. Lang, Phys. Rev. B 1988, 37, 10 395.
69. D. Porath, Y. Levi, M. Tarabiah et al., Phys. 92. N. Lang, W. Kohn, Phys. Rev. B 1973, 7,
Rev. B 1997, 56, 9829–9833. 3541.
70. C. Joachim, J. K. Gimzewski, R. R. Schlittler 93. R. Monnier, J. P. Perdew, Phys. Rev. B 1978,
et al., Phys. Rev. Lett. 1995, 74, 2102. 17, 2595.
71. S. Datta, W. D. Tian, S. H. Hong et al., 94. R. Monnier, J. P. Perdew, Phys. Rev. B 1980,
Phys. Rev. Lett. 1997, 79, 2530–2533. 22, 1124(E).
72. W. D. Tian, S. Datta, S. H. Hong et al., J. 95. A. Kiejna, Phys. Rev. B 1991, 43,
Chem. Phys. 1998, 109, 2874–2882. 14 695–14 698.
73. J. R. Hahn, Y. A. Hong, H. Kang, Appl. 96. K. L. Jensen, J. Vac. Sci. Technol., B 1998,
Phys. A -Mater. Sci. Process. 1998, 66, 17, 515–519.
S467–S472. 97. P. A. Serena, J. M. Soler, N. Garcia, Phys.
74. W. Han, E. N. Durantini, T. A. Moore et al., Rev. B 1986, 34, 6767.
J. Phys. Chem. 1997, 101, 10 719–10 725. 98. A. Liebsch, Phys. Scr. 1986, 35, 354.
76 1 Fundamentals
99. B. G. Rudberg, M. Johnson, Phys. Rev. B 127. M. Magoga, C. Joachim, Phys. Rev. B 1998,
1991, 34, 9358. 57, 1820–1823.
100. A. Liebsch, Electronic Excitations at Metal 128. M. Magoga, C. Joachim, Phys. Rev. B 1999,
Surfaces, Plenum Press, New York, 1997. 59, 16 011–16 021.
101. M. Buttiker, R. Landauer, Phys. Rev. Lett. 129. M. P. Samanta, W. Tian, S. Datta et al.,
1982, 49, 1739–1742. Phys. Rev. B 1996, 53, R7626–R7629.
102. M. Buttiker, Phys. Rev. B 1983, 27, 130. S. N. Yaliraki, M. Kemp, M. A. Ratner, J.
6178–6188. Am. Chem. Soc. 1999, 121, 3428–3434.
103. J. Tersoff, D. R. Hamman, Phys. Rev. B 131. F. Biscarini, C. Bustamante, V. M. Kenkre,
1985, 5031, 805. Phys. Rev. B 1995, 51, 11 089–11 101.
104. J. Bardeen, Phys. Rev. Lett. 1961, 6, 57. 132. E. G. Emberly, G. Kirczenow, Phys. Rev. B
105. R. Landauer, IBM J. Res. Dev. 1957, 1, 223.
1999, 60, 6028–6033.
106. R. Landauer, Philos. Mag. 1970, 21,
133. L. E. Hall, J. R. Reimers, N. S. Hush et al.,
863–867.
J. Chem. Phys. 2000, 112, 1510–1521.
107. Y. Imry, Physics of mesoscopic systems
134. F. Faglioni, C. L. Claypool, N. S. Lewis et al.,
in Directions in Condensed Matter Physics
(Eds.: G. Grinstein, G. Mazenko), World J. Phys. Chem. 1997, 101, 5996–6020.
Scientific, Singapore, 1986, p. 101. 135. C. L. Claypool, F. Faglioni, W. A. Goddard
108. Y. Imry, Introduction to Mesoscopic Physics, et al., J. Phys. Chem. 1997, 101, 5978–5995.
Oxford University Press, Oxford, 1997. 136. S. Larsson, A. Klimkans, Theochem. J. Mol.
109. W. H. Miller, S. D. Schwartz, J. W. Tromp, Struct. 1999, 464, 59–65.
J. Chem. Phys. 1983, 79, 4889–4898. 137. S. N. Yaliraki, A. E. Roitberg, C. Gonzalez
110. M. Galperin, D. Segal, A. Nitzan, J. Chem. et al., J. Chem. Phys. 1999, 111, 6997–7002.
Phys. 1999, 111, 1569–1579. 138. M. D. Newton, Chem. Rev. 1991, 91, 767.
111. V. Mujica, M. Kemp, M. A. Ratner, J. Chem. 139. T. Vondrak, C. J. Cramer, X. Y. Zhu, J.
Phys. 1994, 101, 6849–6864. Phys. Chem. B 1999, 103, 8915–8919.
112. D. M. Newns, Phys. Rev. 1969, 178, 1123. 140. E. Emberly, G. Kirczenow, Phys. Rev. Lett.
113. J. N. Onuchic, D. N. Beratan, J. Chem. Phys. 1998, 81, 5205–5208.
1990, 92, 722. 141. E. G. Emberly, G. Kirczenow, J. Phys. C:
114. P. C. P. D. Andrade, J. N. Onuchik, J. Chem. Condens. Matter 1999, 11, 6911–6926.
Phys. 1998, 108, 4292–4298. 142. T. Koopmans, Physica 1933, 1, 104.
115. A. Onipko, Phys. Rev. B 1999, 59, 143. M. F. Falcetta, K. D. Jordan, J. Phys. Chem.
9995–10 006. 1990, 94, 5666.
116. A. Barraud, P. Millie, I. Yakimenko, J. Chem. 144. M. F. Falcetta, K. D. Jordan, J. Am. Chem.
Phys. 1996, 105, 6972–6978. Soc. 1991, 113, 2903.
117. A. Nitzan, J. Phys. Chem. A 2001, 105, 2677. 145. C.-S. Chen, T. H. Feng, J. S.-Y. Chao, J.
118. D. Segal, A. Nitzan, M. A. Ratner et al., J. Phys. Chem. 1995, 99, 8629.
Phys. Chem. 2000, 104, 2790. 146. P. D. Burrow, A. E. Howard, A. R. Johnston
119. P. Sautet, C. Joachim, Chem. Phys. 1989,
et al., J. Phys. Chem. 1992, 96, 7570.
135, 99.
147. Z. I. Miskovic, R. A. English, S. G. Davison
120. P. Sautet, C. Joachim, Chem. Phys. Lett.
et al., J. of Phys. Condensed Matter 1997, 9,
1991, 185, 23.
10 749–10 760.
121. P. Sautet, J. C. Dunphy, D. F. Ogletree et al.,
Surf. Sci. 1994, 315, 127. 148. T. Seideman, W. H. Miller, J. Chem. Phys
122. C. Chavy, C. Joachim, A. Altibeli, Chem. 1992, 97, 2499.
Phys. Lett. 1993, 214, 569. 149. T. Seideman, W. H. Miller, J. Chem. Phys
123. P. Doumergue, L. Pizzagalli, C. Joachim 1992, 96, 4412.
et al., Phys. Rev. B 1999, 59, 15 910–15 916. 150. W. H. Miller, T. Seidman, Cumulative and
124. C. Joachim, J. K. Gimzewski, Europhys. Lett. state to state reaction probabilities via a
1995, 30, 409. discrete variable representation-absorbing
125. C. Joachim, J. F. Vinuesa, Europhys. Lett. boundary condition green’s function in
1996, 33, 635–640. Time-Dependent Quantum Molecular Dy-
126. M. Magoga, C. Joachim, Phys. Rev. B 1997, namics: Experiment and Theory (Ed.: J. Broe-
56, 4722–4729. ckhove), Plenum Press, New York, 1992.
151. D. Evans, I. Benjamin, T. Seidman et al., 177. H. Burke, E. K. U. Gross in Density Func-
Abstr. Papers Am. Chem. Soc. 1996, 212, tionals: Theory and Applications (Ed.: D. Jou-
194. bert), Springer, Berlin, 1998.
152. A. Haran, A. Kadyshevitch, H. Cohen et al., 178. D. Lamoen, P. Ballone, M. Parrinello,
Chem. Phys. Lett. 1997, 268, 475–480. Phys. Rev. B: Condens. Matter 1996, 54,
153. A. Mosyak, A. Nitzan, R. Kosloff, J. Chem. 5097–5105.
Phys. 1996, 104, 1549–1559. 179. A. Aviram, M. A. Ratner, Chem. Phys. Lett.
154. A. Mosyak, P. Graf, I. Benjamin et al., J. 1974, 29, 277.
Phys. Chem. A 1997, 101, 429–433. 180. D. H. Waldeck, D. N. Beratan, Science 1993,
155. A. Nitzan, I. Benjamin, Acc. Chem. Res. 261, 576, 577.
1999, 32, 854–861. 181. R. A. Marcus, J. Chem. Soc., Faraday Trans.
156. I. Benjamin, D. Evans, A. Nitzan, J. Chem. 1996, 92, 3905–3908.
Phys. 1997, 106, 1291–1293, 6647–6654. 182. M. Pomerantz, A. Aviram, R. A. McCorkle
157. R. Naaman, A. Haran, A. Nitzan et al., J. et al., Science 1992, 255, 1115.
Phys. Chem. B 1998, 102, 3658–3668. 183. A. S. Martin, J. R. Sambles, Adv. Mater.
158. U. Peskin, A. Edlund, I. Bar-On et al., J. 1993, 5, 580–582.
Chem. Phys. 1999, 111, 7558–7566. 184. A. S. Martin, J. R. Sambles, G. J. Ashwell,
159. N. D. Lang, Phys. Rev. B 1988, 38, 10 395. Phys. Rev. Lett. 1993, 70, 218–221.
160. N. D. Lang, A. Yacoby, Y. Imry, Phys. Rev. 185. C. M. Fischer, M. Burghard, S. Roth et al.,
Lett. 1989, 63, 1499. Europhys. Lett. 1994, 28, 129–134, 375.
161. N. D. Lang, Phys. Rev. B 1992, 45, 13 599. 186. K. Nagesha, J. Gamache, A. D. Bass et al.,
162. N. D. Lang, Phys. Rev. B 1995, 51, 2029(E). Rev. Sci. Instrum. 1997, 68, 3883–3889.
163. N. D. Lang, Phys. Rev. B 1995, 52, 187. R. M. Morsolais, M. Deschenes, L. Sanche,
5335–5342. Rev. Sci. Instum. 1989, 60, 2724–2732.
164. N. D. Lang, P. Avouris, Phys. Rev. Lett. 1998, 188. G. Makov, A. Nitzan, L. E. Brus, J. Chem.
81, 3515–3518. Phys. 1988, 88, 5076–5085.
165. N. D. Lang, P. Avouris, Phys. Rev. Lett. 2000, 189. D. K. Ferry, S. M. Goodnick, Transport
84, 358–361. in Nanostructures, Cambridge University
166. R. G. Parr, W. Yang, Density Functional The- Press, Cambridge, 1997, Vol. 6.
ory of Atoms and Molecules, Oxford Univer- 190. R. Wilkins, E. Ben-Jacob, R. C. Jaklevic,
sity Press, Oxford, 1989. Phys. Rev. Lett. 1989, 63, 801–804.
167. O. Gunnarsson, R. O. Jones, Rev. Mod. 191. M. A. Kastner, Rev. Mod. Phys. 1992, 64,
Phys. 1989, 61, 689. 849.
168. N. Lang, Phys. Rev. Lett. 1986, 56, 1164. 192. S. H. M. Persson, L. Olofsson, L. Gunnar-
169. M. Di Ventra, S. T. Pantelides, N. D. Lang, sson, Appl. Phys. Lett. 1999, 74, 2546–2548.
Phys. Rev. Lett. 2000, 84, 979–982. 193. M. Dorogi, J. Gomez, R. Osifchin et al.,
170. M. Di Ventra, S. T. Pantelides, Phys. Rev. B Phys. Rev. B: Condens. Matter 1995, 52,
1999, 59, R5320–R5323. 9071–9077.
171. G. Galli, M. Parrinello in Computer Simula- 194. R. P. Andres, S. Datta, M. Dorogi et al., J.
tions in Material Science (Eds.: V. Pontikis, Vac. Sci. Technol., A 1996, 14, 1178–1183.
M. Meyer), Kluwer, Dordrecht, The Nether- 195. R. P. Andres, T. Bein, M. Dorogi et al., Sci-
lands, 1991. ence 1996, 272, 1323–1325.
172. K. Hirose, M. Tsukada, Phys. Rev. B 1995, 196. A. N. Korotkov, Coulomb blockade and digi-
51, 5278–5290. tal single electron devices in Molecular Elec-
173. V. Mujica, A. E. Roitberg, M. A. Ratner, J. tronics (Eds.: J. Jortner, M. Ratner), Black-
Chem. Phys. 2000, 112, 6834–6839. well Science, Oxford, 1997, pp. 157–189.
174. A. Szabo, N. S. Ostlund, Modern Quantum 197. A. M. Kuznetsov, J. Ulstrup, J. Electroanal.
Chemistry: Introduction to Advanced Elec- Chem. 1993, 362, 147–152.
tronic Structure Theory, McGraw-Hill, New 198. F.-R. F. Fan, A. J. Bard, Science 1997, 277,
York, 1989. 1791–1793.
175. H. J. Silverstone, M. L. Yin, J. Chem. Phys. 199. L. I. Malysheva, A. I. Onipko, Phys. Rev. B
1968, 49, 2020. 1992, 46, 3906–3915.
176. S. Huzinaga, C. Arnan, Phys. Rev. A 1970, 200. V. Mujica, M. Kemp, A. Roitberg et al., J.
1, 1285. Chem. Phys. 1996, 104, 7296–7305.
78 1 Fundamentals
201. Y. Kawahito, H. Kasai, H. Nakanishi et al., 212. S. Datta, Electric Transport in Mesoscopic
J. Appl. Phys. 1999, 85, 947–952. Systems, Cambridge University Press, Cam-
202. Y.-Q. Li, C. Gruber, Phys. Rev. Lett. 1998, 80, bridge, 1995.
1034–1037. 213. E. L. Wolf, Principles of Electron Tunneling
203. A. S. Davidov, V. M. Ermakov, Physica 1987, Spectroscopy, Oxford University Press, New
28D, 168–180. York, 1985.
204. H. L. Berkowitz, R. A. Lux, J. Vac. Sci. 214. E. Emberly, G. Kirczenow, Phys. Rev. B
Technol. 1987, B5, 967–970. 2000, 61, 5740–5750.
205. F. W. Sheard, G. A. Toombs, Appl. Phys. 215. J. Jortner, M. Bixon, T. Langenbacher et al.,
Lett. 1988, 52, 1228–1230. Proc. Natl. Acad. Sci. USA 1998, 95,
206. N. C. Kluksdahl, A. M. Kriman, D. K. Ferry 12 759–12 765.
et al., Phys. Rev. B 1989, 39, 7720–7735. 216. M. Bixon, B. Giese, S. Wessely et al.,
207. A. Nakano, R. K. Kalia, P. Vashishta, Phys. Proc. Natl. Acad. Sci. USA 1999, 96,
Rev. B: Condens. Matter 1991, 44, 11 713–11 716.
8121–8128. 217. B. Giese, S. Wesswly, M. Spormann et al.,
208. J. Chen, M. A. Reed, A. M. Rawlett et al., Angew. Chem. Int., Ed. Engl. 1999, 38,
Science 1999, 286, 1550–1552. 996–998.
209. S. A. Gurvitz, H. J. Lipkin, Y. S. Prager, 218. Y. A. Berlin, A. L. Burin, M. A. Ratner, J.
Mod. Phys. Lett. B 1994, 8, 1377. Phys. Chem. 2000, 104, 443–445.
210. S. A. Gurvitz, H. J. Lipkin, Y. S. Prager, 219. B. Giese, Acc. Chem. Res. 2000, 33, 631–636.
Phys. Lett. A 1996, 212, 91–96. 220. A. Nitzan, Annu. Rev. Phys. Chem. 2001, 52,
211. S. A. Gurvitz, Y. S. Prager, Phys. Rev. B 681–750.
1996, 53, 15 932–15 943.
81
2.1 a couple of sections, it is assumed that

Diffusion and Migration there are no homogeneous reactions. In-
soluble species are also absent. Fourth,
José A. Manzanares
since the chapter is mainly of interest
University of Valencia, Burjasot, Spain
.. to cases in which mass transport is rate
Kyosti Kontturi controlling, reversible electrode reactions
Helsinki University of Technology, Espoo, are assumed and no kinetic effects are
Finland
considered. Fifth, most sections involve
Introduction
steady state transport because the theoret-
ical description of transient experiments
This chapter describes mass transport to in the presence of migration is difficult
electrodes by diffusion and migration. It and quite often requires numerical solu-
is assumed throughout that there is no tions that bring much more mathematical
convective motion of the solution, and complexity than new physical insights.
transport is described with respect to a Nevertheless, the effect of migration is ana-
reference frame fixed to the laboratory lyzed in a simple transient case. Finally, the
[1, 2]. Thus, many of the equations derived local electroneutrality (LEN) assumption is
in this chapter cannot be directly applied used. This is indeed a good approximation
to systems with bulk motion. for most experimental situations, although
The chapter emphasizes migrational ef- deviations from LEN occur under some
fects and provides analytical expressions conditions. These deviations are discussed
for the electric potential drop in the in the last section.
solution under different experimental con-
ditions. The fundamental concepts are 2.1.1
discussed in detail, and a number of Fundamental Concepts
important restrictions are introduced for
the sake of clarity. First, diluted solu- This section presents the basic equations
tions are considered throughout and the describing mass transport to electrodes:
flux equations incorporate neither cross the flux equations, the continuity equa-
terms nor activity coefficients [2–5]. Sec- tion, and the LEN assumption. The flux
ond, one-dimensional systems are consid- equations are derived from the concept of
ered, except when presenting the transport electrochemical potential, which is used as
equations in Sect. 2.1.1. Third, except in a general term valid for both charged and
82 2 Transport Phenomena

neutral species. The connection between dU = T dS + µ̃i dni (2)
electric conduction and the basic transport i
mechanisms, diffusion and migration, is
discussed in detail, and the difference be- where T is the thermodynamic tempera-
tween ohmic and total electric potential ture, S the entropy, and ni the number of
gradient is emphasized. The spatial re- moles of species i. Moreover, considering
gion in which mass transport takes place that the exchange of matter between V and
and the coupling between mass transport V takes place without energy exchange
and the electrode reaction are discussed with their surroundings, it is satisfied that
next. At this point, the one-dimensional dU + dU = 0 and dni + dni = 0. The
geometry to be used henceforth is intro- Second Law of Thermodynamics requires
that the entropy of the isolated system
duced. The electric potential drops in the
formed by the two volume elements must
system and the sign conventions are also
increase
defined. The boundary conditions for the

transport equations are written in terms of T (dS + dS ) = (µ̃i − µ̃i )dni ≥ 0
the integral transport numbers. For the i
sake of simplicity, parallel reactions [6] (3)
and adsorption phenomena [7] are not When the transport of species i is not
considered. coupled to that of other species, each
individual term of this sum is positive.
2.1.1.1 Diffusion–migration Flux In this case, dni is determined only by
Equations (µ̃i − µ̃i ) and species i moves towards
The flux equations describing the behavior the region in which its electrochemical
of electrochemical systems are intimately potential is lower, that is, dni < 0 when
related to the Second Law of Thermody- µ̃i < µ̃i and vice versa. This happens, for
namics. Consider two neighboring volume instance, in the case of ionic species in
elements V and V of a solution hav- diluted solutions. When the transport of
ing the same temperature and pressure, different species is coupled, one or more
but different electrochemical potentials of terms in the sum can be negative, but the
their constituents. The difference between sum is always positive.
the electrochemical potential of species The transport of species i can be
i, µ̃i and µ̃i , in these volume elements described in terms of either its velocity vi or
implies that there is no distribution equi- its flux density Ji , which are related to each
librium and this species tends to move other by the molar concentration, Ji =
from one volume element to the other. ci vi . If the area of the surface separating
This motion of species i from one volume the two volume elements is dA and its
element to its neighbor is generically called orientation is given by the unit vector
transport of species i. Since the transport n̂(from V to V ), the number of moles
of the different species in solution takes of species i crossing the surface in a time
place under thermal and mechanical equi- dt is dni = −Ji n̂ dA dt.
librium, the change in internal energy U When the difference between µ̃i and µ̃i
of these two volume elements is [1, 8, 9] is not too large, it can be assumed [10]
that the rate of change of the amount of

dU = T dS + µ̃i dni (1) species i, dni /dt, is proportional to the
i difference µ̃i − µ̃i or, more exactly, to
2.1 Diffusion and Migration 83
the gradient of this potential normal to 2.1.1.2 Poisson Equation and the LEN
µ̃i n̂. In this linear
the surface, ∂ µ̃i /∂n ≡ ∇ Assumption
approximation, the velocity of species i is Equation (5) states that the flux density of
expressed as species i in a diluted solution depends on
∇ µ̃i but not on other gradients ∇ µ̃j =i . In
µ̃i
vi = −ui ∇ (4) this sense, there is no correlation between
the motions of two different species, i
where ui is its mobility. Similarly, the flux and j . The migrational term in Eq. (7),
density takes the form however, imposes a strong constraint on
the motion of charged species by coupling
µ̃i = − Di ci ∇
Ji = ci vi = −ui ci ∇ µ̃i (5) their flux densities Ji .
RT The electrical state of the system is so
where R is the universal gas constant sensitive to small changes in composi-
tion [9] that the gradient ∇φ in Eq. (7)
and the Einstein relation between mobility
and diffusion coefficient, Di = ui RT , has cannot be assumed, in general, to be a
been used. simple constant [12], and an additional
At constant temperature and pressure, equation is required to determine it. Given
µ̃i originates from changes
the gradient ∇ the slowness of particle motion in solution,
in composition and electric potential φ it is justified to use the Poisson equation
so that from electrostatics to relate the changes in
electric potential to the local electric charge
∇ ln ci + zi F ∇φ
µ̃i = RT ∇ (6) density ρe
ρe F
where F is the Faraday constant and zi ∇2φ = − =− zi ci (8)
is the charge number of species i. Taking ε ε
i
Eq. (6) to Eq. (5), the Nernst–Planck flux
equation [11] where ε is the dielectric permittivity of
the solution.
i + zi ci f ∇φ)
Ji = −Di (∇c (7) Consider, for example, a 10 mM 1 : −1
electrolyte solution that contains a positive
is obtained, where f denotes the ratio charge density of the order of 106 C m−3
F /RT . The two terms in the right- counterbalanced by a negative charge of
hand side of this equation represent the −106 C m−3 . Since the dielectric permit-
transport mechanisms of diffusion and tivity of aqueous electrolyte solutions is of
migration, respectively. Diffusion is a con- the order of 10−9 C V−1 m−1 , the right-
sequence of the random thermal motion hand side of Eq. (8) is roughly the dif-
of the particles, which tends to make the ference between two terms whose order
concentration of all species uniform. Mi- of magnitude is 1015 V m−2 . The typical
gration accounts for the influence of the order of magnitude of the electric fields,
= −∇φ,
electric field, E on the random however, can be estimated as 10 V m−1
motion of the charged particles, and Eq. (7) and they are practically independent on
shows that the particles acquire a compo- position (except at the interfacial regions),
nent of their velocity along the direction so that ∇ 2 φ is many orders of magni-
of the electric field as a result of this tude smaller than 1015 V m−2 . This means
influence. that the imbalance between cation and

anion concentrations in the system that d ∂ci
ci (r , t)dV = dV
is required to set up the typical electric dt ∂t
potential distributions is many orders of V V

magnitude smaller than the solution con-
= ri dV − Ji n̂dA
centration. Therefore, the space charge
density can be neglected and use can be V S

made of the LEN assumption [13] Ji )dV
= (ri − ∇ (10)

zi ci = 0 (9) V
i
where Gauss’ divergence theorem has
This equation replaces Eq. (8) in the been used. Moreover, since no restric-
description of transport processes. Though tions apply to the volume V , Eq. (10)
it is customary to use the equal symbol implies that
in Eq. (9), the electric charge density is
not strictly zero, and the Poisson equation ∂ci Ji = ri
+∇ (11)
cannot be reduced to the Laplace equation ∂t
∇ 2 φ = 0.
which is the continuity equation for
species i.
2.1.1.3 Continuity Equation
The spatial variation of the flux density of 2.1.1.4 Ohm’s Law and Migrational
species i is determined by the principle of Transport Numbers
conservation of the amount of substance In a diluted electrolyte solution of uniform
for this species, which is represented by the composition, the flux density Ji = ci vi is
continuity equation. Consider an arbitrary proportional to ∇φ, and the conduction
volume V in an electrolyte solution, in electric current density
which the molar concentration of species

i is ci (r , t). The amount of substance of I ≡ F zi Ji (12)
species i within V can change as a result i
of either a net flow of particles through
the surface S enclosing the volume V , or satisfies Ohm’s law
due to homogeneous chemical reactions
occurring within V . If the surface S is I = −κ ∇φ,
i = 0, ∀i)
(∇c (13)
divided into surface elements of area dA
perpendicular to the unit vector n̂ (oriented where
outwards from V ), the number of moles of F2 2
κ≡ zi Di ci (14)
species i entering V through this surface RT
i
element in a time dt is dni = −Ji n̂dAdt.
The homogeneous reactions generate or is the electrical conductivity of the solution.
consume this species at a net rate ri (r , t). The minus sign in Eq. (13) means that the
Since the (net) accumulation rate of electric current flows in the direction of
species i within V is given by the sum decreasing electric potential. Equation (14)
of the (net) generation rate due to the evidences that the conductivity of the so-
chemical reaction and the (net) incoming lution is determined by all ionic species in
flux through S, it is concluded that solution, no matter whether they are active
or not. The migrational transport number This is due to the diffusion potential gra-
dient that originates from the diffusion
z 2 D i ci process, and therefore an expression sim-
ti ≡ i (15)
zj2 Dj cj ilar to Ohm’s law can only be recovered
j by subtracting the diffusion potential gra-
dient from the total potential gradient, as
represents the relative contribution of shown by Eq. (17). A theoretical simpli-
species i to the conductivity of the solution. fication sometimes used in the literature
Unfortunately, concentration gradients consists in neglecting the diffusion poten-
develop under most experimental condi- tial by assuming all ionic diffusivities to be
tions, and the motion of charge carriers equal [17, 18].
takes place by both diffusion and migration
as described by the Nernst–Planck equa-
2.1.1.5 Diffusion-conduction Flux
tion, Eq. (7). In this case, Eqs. (14 and 15)
Equation
provide the local (i.e. position-dependent)
The introduction of the modified Ohm’s
values of the conductivity and the transport
law, Eq. (17), in the Nernst–Planck equa-
number of species i, respectively. By writ-
tion, Eq. (7), leads to the following form of
ing the Nernst–Planck equation in terms
the flux equation
of ti as [1, 14]

ti κ
Ji = − 2 ∇ µ̃i (16) Ji = −Di (∇c dif ) + ti I
i + zi ci f ∇φ
zi F 2 zi F
(20)
Eq. (12) leads to the modified Ohm’s For the sake of simplicity, consider the case
law [5, 15] of a binary solution obtained by complete
dissociation of a strong electrolyte Aν1 Bν2
I = −κ(∇φ
− ∇φ ohm (17)
dif ) ≡ −κ ∇φ
into ν1 ions Az1 of charge number z1
and ν2 ions B z2 of charge number z2 ;
in which
obviously, the relation z1 ν1 + z2 ν2 = 0 is
ti
ohm = 1
∇φ µ̃i
∇ (18) satisfied. The ionic concentrations are then
F zi related to the stoichiometric electrolyte
i
concentration c12 by the relations c1 =
is the ohmic potential gradient, and ν1 c12 and c2 = ν2 c12 which allows Eq. (20)
to be rewritten as
dif = − F
∇φ i
zi Di ∇c
κ ti I
12 + ti I
Ji = νi J12 +
i
= −νi D12 ∇c
RT ti zi F zi F
=− ∇ ln ci (19) (21)
F zi
i where J12 is the electrolyte flux den-
sity, and
is the diffusion potential gradient.
That is, the concentration gradients (ν1 + ν2 )D1 D2
break the proportionality between the D12 = (22)
ν2 D1 + ν1 D2
current density and the electric field; fur-
thermore, in some cases, I and ∇φ are not is the Nernst–Hartley expression for the
even parallel vectors, a fact that is relevant diffusion coefficient of the dissociated
to the tertiary current distribution [16]. electrolyte [3].
Equations (20 and 21) describe ionic the concentration variations toward the
transport as a combination of electrolyte interior of the solution, which affects
diffusion and ionic conduction, and hence a region that grows in thickness. It
the name diffusion–conduction flux equa- is possible, however, to confine the
tion is suggested. Its comparison with concentration variations to a ‘‘thin’’ region
the diffusion–migration flux equation, near the electrode by vigorous stirring,
Eq. (7), evidences the difference between ‘‘thin’’ here means of the order of tenths
migration and conduction. Migration is of a millimeter. The solution can then
due to electric fields, either external or be divided in two regions: the diffusion
internal, and does not require a nonzero boundary layer (DBL) adjacent to the
current density. Conduction is the ionic electrode surface, in which concentration
motion associated to the part of the elec- varies with position and diffusion affects
tric field that is controlled externally. Note mass transport; and the bulk solution, in
that the electric current density is a mea- which concentration is made uniform by
surable magnitude but not the local electric stirring and transport is due to migration
field. Conduction requires a nonzero elec- and convection [5, 19, 20]. Although the
tric current density. transition between these two regions
Note, finally, that the dissociated elec- is actually smooth, it is practical to
trolyte satisfies Fick’s law, J12 = −D12 assume that the transition is abrupt and
12 , because both ions move at the same
∇c to assign a thickness δ to the DBL,
velocity in the absence of current (I = 0), which can be empirically determined from
and the dissociated electrolyte is the only the stirring rate or other hydrodynamic
independently moving substance [1]. This conditions [21, 22]. The thickness δ is
is not in contradiction with the fact that the considered to be finite when describing
ionic diffusion coefficients are different to steady state transport (indeed, this is a
each other, D1 = D2 , because the diffu- necessary condition for reaching the steady
sion potential gradient affects the ionic state); infinite DBL thickness, however, is
motion even when I = 0. In conclusion, assumed in Sect. 2.1.5.
although the Nernst–Planck equation es- In the following sections, the one-
tablishes that the flux of species i in a dimensional geometry described by Fig. 1
diluted solution is only proportional to the is considered. The electrode is located at
gradient ∇ µ̃i , the contribution of the elec- x = 0 and the DBL in the region 0 < x < δ.
tric potential gradient to ∇ µ̃i couples the The electric current density I is positive for
transport of different species. anodic oxidations and negative for cathodic
reductions, in agreement with the IUPAC
2.1.1.6 Diffusion Boundary Layer convention. The electric potential drop in
In general, the electrode reaction makes the DBL is defined as the potential in the
the concentration of species i to vary with bulk solution (x = δ) relative to that at the
the distance to the electrode surface. These electrode surface (x = 0), %φ ≡ φ b − φ s .
concentration changes occur not only for
the species participating in the electrode 2.1.1.7 Faraday’s Law and Integral
reaction, but also to other charged species Transport Numbers
because of their electrostatic coupling. The charge transfer between an electrolyte
The progress of the electrode reaction solution and an electrode (quite frequently,
is accompanied by a propagation of a metal phase) can be represented by the
Electrode DBL Bulk solution
I
fb
f(x)
Potential drop in DBL
∆f ≡ fb − fs
fs
Electrode potential
E ≡ fe − fs
fe
0 d x
Fig. 1 Schematic drawing of the system considered. The electric
potential distribution has been shown for a cathodic process (I < 0).
Superscripts b, s, and e denote, respectively, bulk solution (x = δ),
solution adjacent to the electrode surface (x = 0), and electrode phase.
general electrode reaction equation conduction electric current density and,

therefore, that the flux density of species i
νir Bizi + ne− = 0 (23) at the electrode surface is given by
i
1 dni ν r dξ I
where Bi and νir are the molecular formula Jis = = i = νir (24)
A dt A dt nF
and the stoichiometric number of species
i, respectively. The value of νir is positive where A is the electrode surface area.
for products and negative for reactants. Equation (24) shows clearly that, at the
The stoichiometric number electrode surface, the fraction of electric
of the electron current density transported by species i is
in the reaction, n = i zi νir , is positive
for (anodic) oxidations and negative for given by the magnitude
(cathodic) reductions. The electrons in zi F Jis zi νir
Eq. (23) are considered to be in the metal Tis ≡ = (25)
I n
phase, and species Bi can be either in
which is called the integral transport
solution or in the electrode phase; it
number of species i [23]; this magnitude
is obviously necessary that at least one
is zero for the species that are not
active species Bi is in solution because
involved in the electrode reaction (i.e. the
Eq. (23) describes charge transfer between
electroinactive species).
It is important
two phases.
to observe that i Tis = 1 but Tis is not
The evolution of the electrode reaction
bounded between 0 and 1. For example, in
needs the transport of reactants from the
the electrode reaction
solution to its surface and products vice
versa. Faraday’s law establishes that the Fe(CN)6 3− (aq) − Fe(CN)6 4− (aq) + e− = 0
reaction rate dξ/dt is proportional to the (26)
the transport numbers at the elec- The integral transport number of species
trode surface are T s = −3 and i can also be defined at locations differ-
Fe(CN)6 3−
T s = 4. ent from the electrode surface as Ti ≡
Fe(CN)6 4−
Some electrode reactions, such as the zi F Ji /I [23]. Under steady state condi-
anodic dissolution of copper tions and in the absence of homogeneous
reactions, both the current density I and
Cu+ (aq) − Cu(s) + e− = 0 (27) the flux density Ji do not vary with position,
and Ti takes the value specified by Eq. (25)
the hydrogen evolution at a platinum throughout the whole system even when
electrode concentration gradients are present. Oth-
erwise, Ti is position-dependent. It is
H2 (g) − 2H+ (aq) + 2e− = 0 (28)
important to observe that the integral
or the oxidation of silver at a Ag|AgCl transport numbers Ti are equal to the
electrode migrational transport numbers ti in the
absence of concentration gradients only;
AgCl(s) − Ag(s) − Cl− (aq) + e− = 0 indeed, concentration gradients originate
(29) due to the fact that Ti = ti .
involve only one electroactive species
(Cu+ , H+ , and Cl− , respectively), which is 2.1.1.8 Nernst Equation and
denoted by subscript 1 henceforth. Since Concentration Overpotential
the current density in these processes is In order to drive an electric current den-
due to the single electroactive species, sity across the system depicted in Fig. 1,
n = z1 ν1r and T1s = 1, while Tis = 0 for an electric potential difference must be ap-
i = 1. In these simple cases, the DBL plied between the electrode and the bulk
becomes depleted of this species when solution. This potential difference can be
J1 < 0 (i.e. the species is consumed at the expressed in terms of three contributions:
electrode surface), and the potential drop the potential drop in the DBL, the equilib-
satisfies z1 %φ > 0. On the contrary, when rium interfacial or Nernst potential drop,
J1 > 0, the DBL becomes concentrated in and the activation overpotential. For large
species 1 and z1 %φ < 0. enough kinetic currents, the latter contri-
Note also that Tis is zero for neutral bution can be neglected [7, 25] and the
species but this does not imply that potential difference required for the elec-
they are not transported towards or trode process is Eeq − %φ, where Eeq is
from the electrode. For example, in the the equilibrium electrode potential (i.e. the
case of nitrous acid in the electrode electric potential of the electrode relative
reaction to that at the adjacent electrolyte solution).
The value of Eeq is given by the Nernst
3H+ (aq) + NO3 − (aq) − HNO2 (aq) equation. For the electrode Reaction (23),
− H2 O(aq) + 2e− = 0 (30) the Nernst equation takes the form
s
THNO = 0 but JHNOs = −I /2F , as can ◦ RT r cis
2 2 Eeq = E + νi ln ◦ (31)
be easily deduced from Eq. (24). (For the nF c
i
neutral species, the transference number
τi ≡ Ti /zi constitutes a good alternative to where E ◦ is the standard (or, more exactly,
Ti [24].) the formal) electrode potential and c◦ is
the standard concentration. The concen- 2.1.2

trations cis of the active species in Eq. (31) Steady State Current–voltage Curves of
are those at the electrode surface, which Systems with One Active Species
are evaluated by extrapolation to the elec-
trode surface of the concentration profile The finite transport rate of the electroac-
within the DBL [19]. In other words, the tive species toward the electrode surface
electrode surface at position x = 0 should implies the existence of a limiting current
be understood as the neutral solution just density, that is the most important con-
outside the electrical double layer (EDL). sequence of mass transport. This limiting
Similarly, %φ refers to the potential dif- current density is defined by the condition
ference between the bulk solution and of maximum concentration gradient of the
that point in the solution (that is, between electroactive species, which is achieved
the bulk solution|DBL and the EDL|DBL when the concentration of this species
boundaries). Hence, no details of the EDL tends to zero at the electrode | depleted
structure are considered when evaluating DBL interface (or, more accurately, at the
the equilibrium interfacial potential drop EDL | DBL boundary).
and the surface concentrations of the ac- In this section, the mass transport effects
tive species. on the steady state current–voltage curves
Since the interfacial concentrations of are described for the (relatively common)
the active species vary with the electric case of systems with only one electroactive
current density, Eeq also does and its species in solution. This species is denoted
variation is often described in terms of by subscript 1. The integration procedure
the concentration overpotential [19, 26] of the transport equations in multi-ionic
systems is presented first and the general
b RT r cis current–voltage curve is derived. Except
ηc ≡ Eeq − Eeq ≡ νi ln b (32)
nF ci in Sect. 2.1.2.5, homogeneous reactions
i
are assumed to be absent. The interest-
which represents the potential of the elec- ing case of excess supporting electrolyte,
trode in equilibrium with the interfacial in which the active species is present in
concentrations that are established when very small concentration and transport
a current density I is driven through the takes place mostly by diffusion, is dis-
system, Eeq , relative to the (constant) po- cussed. Finally, the simple cases of binary
tential of a similar electrode equilibrated and ternary solutions are considered to il-
with the bulk solution, Eeq b . The potential lustrate the conclusions drawn from the
difference between the electrode and the general solution.
bulk solution is then Eeqb + η − %φ.
c
Note, finally, that when the supporting 2.1.2.1 Integration of the Transport
electrolyte is used, the ionic strength Equations
of the DBL is not significantly affected The fact that the flux density of the
by the concentration polarization of the electroinactive species, i = 1, is zero
active species, and therefore, their activity simplifies significantly the solution of the
coefficients are practically independent of transport equations. These species are
the current density [26]. This justifies the in electrochemical equilibrium within the
use of concentrations instead of activities DBL, and hence their concentrations ci
in the above equations. are related to the electric potential φ by the
Boltzmann equation where %ci ≡ cib − cis . Introducing the lim-

iting diffusion current density
dci dφ
= −zi ci f (i = 1, Ji = 0) (33)
dx dx z1 F D1 c1b
ILd,1 ≡ − (40)
Thus, the ionic concentrations at the δ
electrode surface, x = 0, are related to
and using Eqs. (34 and 36), the cur-
those in the bulk solution, x = δ, by
rent–voltage relation takes the form
cis = cib ezi f %φ (i = 1, Ji = 0) (34) cb
zi
I = ILd,1 i
1 − (1 − ezi f %φ )
where %φ ≡ φ b − φ s is the electric po- c
i =1 1
b z1
tential drop across the DBL (see Fig. 1). (41)
Furthermore, the LEN assumption Note that the definition of ILd,1 has not
been deduced from the condition c1s = 0,
zi ci = z1 c1 + zi ci = 0 (35) but it rather corresponds to a situation
i i =1 in which the transport of species 1 takes
implies that not only ci (i = 1) but also c1 place mostly by diffusion, and hence its
is determined by φ. Thus, for instance, the name. Moreover, Eq. (40) corresponds to
surface concentration of the electroactive the particular case in which there is only
species is one active species; see Eq. (78) below for
the general definition.
1 s 1 b zi f %φ It is interesting to observe that species
c1s = − zi ci = − zi ci e
z1 z1 1 is the only one that moves, while the
i =1 i =1
(36) inactive species are standing still over
The current–voltage relation can be ob- the system. In a sense, this is similar to
tained from the flux equation of the the conduction of electrons in a metal,
electroactive species in which electrons move and the ions
are standing on the crystal lattice and,
I dc1 dφ therefore, Ohm’s law is expected to be
J1 = = −D1 + z1 c1 f
z1 F dx dx satisfied. Indeed, Eq. (37) shows that the
(37) electric current density is proportional
by transforming the migration term with to the electric field, I ≡ −κeff (dφ/dx),
the help of Eq. (35) to in which
dφ dφ dci F2
z1 c1 f =− zi ci f = κeff (x) = D1 zi2 ci (x) (42)
dx dx dx RT
i =1 i =1 i
(38)
Integration of Eq. (37) over the DBL is the local effective conductivity and
then gives Eqs. (33 and 35) have been used. There is,
however, a difference between these two
Iδ
− = %c1 + %ci systems. The ‘‘static’’ distribution of the in-
z1 F D1 active species is determined by the electric
i =1
zi
potential distribution and hence, although
= 1− %ci (39) I is only due to species 1, all species are
z1 relevant to the conduction process.
i =1
The local ohmic behavior is not in conditions c1s = 0, satisfies

contradiction with the fact that Eq. (41) 1/2
shows a nonlinear variation of I with −z1 f %φL c1b
e = 1− b (45)
the potential drop %φ. At low potentials, C
f %φ 1, the DBL is practically nonpo-
larized (i.e. the ionic concentrations take Thus, when the ionic strength of the
their bulk values throughout the DBL) and solution is much larger than the concentra-
the slope of the current–voltage curve is tion of the electroactive species (C b c1b ),
b . When increasing %φ, the slope of the
κeff that is, in the presence of excess support-
current–voltage curve decreases owing to ing electrolyte case, z1 %φL tends to zero
the development of concentration polariza- and the electric potential drop is then neg-
tion. The effective conductivity varies then ligible [5, 26, 34]. On the contrary, in the
with position and this makes the overall absence of a supporting electrolyte, the
system behavior to be nonohmic. electroactive species is the only species of
its class, C b = c1b , and z1 %φL diverges.
The conclusion that migration is absent
2.1.2.2 Solutions of Homovalent Ions,
when the ionic strength of the solution
|zi | = z
is much larger than the concentration of
The theoretical analysis of ion transport
the electroactive species is not restricted to
requires the grouping of ions according
the homovalent case; note that the equa-
to their valencies [27–31]. All ions of
tion I ≡ −κeff (dφ/dx) implies that dφ/dx
the same valency (a term used here as
decreases as κeff increases.
equivalent to charge number) are grouped
In homovalent systems, the current–vol-
together to constitute a ‘‘class’’ and the
tage curve, Eq. (41), takes the form
sum of their concentrations is termed the
‘‘class concentration’’ [32], denoted by C. 1 − e−z1 f %φ
I = IL (46)
The simplest transport problem involves 1 − e−z1 f %φL
only two classes corresponding to valencies where
z and −z. This is sometimes referred to as  1/2 
the homovalent case [33]. Some interesting 2C b c1b
IL = ILd,1 1 − 1 − 
conclusions are drawn from Eqs. (36 and c1b Cb
41) in this situation.
Denote the class of the electroactive (47)
species by subscript 1 and the other class by is the limiting current density (i.e. that
subscript 2. The LEN assumption requires corresponding to c1s = 0 and %φ = %φL ).
the two class concentrations to be equal to The value of IL varies between 2ILd,1
each other in the absence of supporting electrolyte
(C b = c1b ) and ILd,1 in the presence
C1 ≡ ci = ci ≡ C2 ≡ C (43) of excess supporting electrolyte (C b
zi =z1 zi =−z1 c1b ). Migrational effects in homovalent
solutions can therefore account for up to a
Equation (36) then simplifies to
factor 2 in IL .
c1s = C b e−z1 f %φ + (c1b − C b )ez1 f %φ
(44) 2.1.2.3 Binary Electrolyte Solutions
and the maximum potential drop in the When the solution contains only two
DBL, which is attained under limiting ionic species, the concentration profiles
are linear and the diffusion and migration includes the case of both cathodic and
terms in the flux equations are indepen- anodic processes. Consider first that the
dent on position within the DBL. The elec- electroactive species is an anion and,
troinactive species can be thought to be in consequently, that IL > 0; note that the
equilibrium as a result of the balance of dif- charge numbers have been defined with
fusional and migrational terms in its flux sign. In the depleted DBL adjacent to
equation. The migrational contribution to the anode, the current density is positive
the flux of the electroactive species is equal and tends to the limiting value when
(due to the LEN assumption) to that of the the (negative) potential drop increases
inactive species, except for a sign rever- in absolute value. In the concentrated
sal, and hence the flux of the electroactive DBL adjacent to the anode (for the
species can be described as the sum of two same geometry described in Fig. 1, that
diffusional contributions, Eq. (38). This is, electrode located at x = 0 and DBL
means that the slopes of the concentration extending from x = 0 to δ), the current
profiles are proportional to the electric cur- is negative, and the potential drop is
rent, Eq. (39), and the occurrence of a limit- positive. Increasing the potential drop
ing current is clearly associated to the max- now leads to an increase of the electrical
imum concentration gradient that can be conductivity of the DBL, and hence of the
established over the DBL, that is, to vanish- slope of the current–voltage curve. When
ing concentrations at the electrode surface. the electroactive species is a cation, these
The current–voltage curve of this system comments must be changed accordingly,
is a particular case of Eq. (41), which takes but Fig. 2 remains valid.
the simple form [34] Figure 3 shows the concentration, ci (x),
and electric potential, φ(x), profiles
I = IL (1 − ez2 f %φ ) (48)
given by
where c1 c2 z2 f (φ b −φ) I x
= = e = 1 − 1 −
z1 (z1 − z2 )F D1 c2b c1b c2b IL δ
IL = 1 − ILd,1 = − (50)
z2 δ
(49) which can be obtained by integration of
is the limiting current density. This curve Eq. (37) from x to δ, after use of Eq. (38).
has been plotted in Fig. 2. The figure In the depleted DBL, the interfacial
1.0
Concentrated
Depleted DBL
DBL
0.5
I/IL
0.0 Fig. 2 Current–voltage curves of a DBL

containing a strong binary electrolyte.
Magnitudes z2 , %φ, I, and IL are defined
−0.5
with sign; subscript 2 denotes the
electroinactive species. Negative values
−1.0 of I/IL correspond to electrode
−1 0 1 2 3 4 5 dissolution (anodic or cathodic), and
positive values to single ion discharge at
−z2 f∆f
the electrode.
Depleted DBL Concentrated DBL

2.0
1.0
ci(x)/cbi
ci(x)/cbi
1.5
0.5
1.0
0.0
−z2 f [fb − f(x)]
−z2 f [fb − f(x)]

4 0.0
2 −0.4
0 −0.8
0.0 0.5 1.0 0.0 0.5 1.0 1.5
x/d x/d
Fig. 3 Concentration and electric potential profiles for depleted
(I/IL > 0) and concentrated DBL (I/IL < 0). The continuous lines
correspond to |I/IL | = 0.5 and the dashed lines to |I/IL | = 1.0.
concentration cis decreases with increasing 2.1.2.4 Ternary Electrolyte Solutions. The
%φ, and the limiting current IL is Supporting Electrolyte
reached when z2 %φ → −∞ and cis → 0. Consider a ternary system in which the
In the concentrated DBL, the interfacial common ion, i = 3, is an inactive species.
concentration cis increases with increasing (If the electroactive species were the
%φ (in absolute value). common ion, the system behavior would
It is important to stress that %φ is be similar to that of a binary solution.)
the potential drop in the DBL only. A key feature of systems with only one
There is also a potential drop at the electroactive species is that there is a
electrode|DBL interface that changes with background concentration of indifferent
concentration polarization. The variation electrolytes that maintain a fairly uniform
of the latter is given by the concentration conductivity of the DBL. This implies
overpotential that, in this case, takes that the electric potential drop is small
the form and is quite uniformly distributed over
the DBL. The concentration profile of the
1 cs 1 I electroactive species is (slightly) nonlinear,
ηc = ln 1b = ln 1 −
z1 f c1 z1 f IL and the same is true in general for the
(51) inactive ions, but the system behavior
Combining Eqs. (48 and 51), the cur- is still quite simple. In the particular
rent–voltage curve can be expressed as [34] case z1 = z2 , the concentration profile
of the common ion is linear and is
z2 − z1 given by
I = IL 1 − exp f (ηc − %φ)
z1 z2
(52) z1 c1b I x
where ηc − %φ is the change in the c3 (x) = c3b − 1−
z1 − z3 ILd,1 δ
potential difference between the electrode (53)
and the bulk solution due to the passage which has been obtained by integration of
of the electric current. Eq. (37) and making use of Eq. (38). The
distributions of the electric potential and when I /ILd,1 = 1.0, which is due to the
concentration of the other inactive ion can fact that there is a small migrational con-
be obtained from Eq. (33) as tribution to the transport of this species. In
other words, the limiting current is slightly
1 c3 (x) 1 c2 (x) larger than ILd,1 under these conditions.
f [φ b − φ(x)] = ln b = ln b
z3 c3 z2 c2 For the discussion of the current–voltage
(54) relation, the homovalent case of Eq. (46) is
Finally, the concentration profile of the considered. In a ternary case, this equation
electroactive species is determined from becomes
the LEN assumption, Eq. (35). Figure 4
2z1 F D1 c3b
shows these profiles for charge num- I =− (1 − e−z1 f %φ ) (55)
bers z1 = z2 = 1, z3 = −2, a concentration δ
ratio c1b /c2b = 0.2, and current densities while Eqs. (45 and 47) simplify to
I /ILd,1 = 0.5 and 1.0. The most signif-
icant characteristic shown by this figure RT
%φL = ln(1 + c1b /c2b ) (56)
is that the electric potential drop is very 2z1 F
small, even for the case I /ILd,1 = 1.0. As
and
a consequence, ion transport takes place 
mostly by diffusion and the concentra- 1/2 
2z1 F D1 c3b c1b
tion profiles are practically linear. As could IL = − 1 − 1 − 
be expected from the ratio c1b /c2b = 0.2,
δ c3b
the concentration changes associated to (57)
current transport are more important in Equations (56 and 57) have been repre-
relative terms for the electroactive species sented in Fig. 5 as a function of the elec-
than for the inactive ones. A detail that can troactive/supporting electrolyte concentra-
(hardly) be observed in Fig. 4 is that c1s = 0 tion ratio. When the electroactive species
6 0.06
2
4 0.04
f [fb − f(x)]
ci(x)/cb1
2 0.02
1
f
0 0.00
0.0 0.5 1.0 1.5
x/d
Fig. 4 Electric potential and ionic concentration profiles
(marked with the respective subscripts 1, 2, 3) in a ternary
system with z1 = z2 = 1, z3 = −2, for a concentration ratio
electroactive to inactive cation c1b /c2b = 0.2, and two values of the
electric current density: I/ILd,1 = 0.5 (continuous line) and 1.0
(dashed line).
IL/ILd, 1
z1 f∆fL
2.0 2
1.5 1
1.0 0
10−2 100 102 104
cb1/cb2
Fig. 5 Dependence of the limiting current density

(continuous line) and the electric potential drop (dashed
line) in the DBL under limiting conditions on the
electroactive/supporting electrolyte concentration ratio.
is a tracer ion, c1b /c2b 1, %φL ≈ 0 and ratio c1b /c2b , when the value of ILd,1 is fixed
(i.e. c1b is fixed). For very low values of this
z1 F D1 c1b c1b ratio, %φ is very small. It is then more
IL ≈ − 1 + b ≈ ILd,1
δ 4c2 relevant to consider also the variation of
(58) the electrode potential. Figure 7 shows the
On the contrary, in the absence of same curves as in Fig. 6 though presented
supporting electrolyte, c1b /c2b 1, IL ≈ in terms of %φ − ηc , where
2ILd,1 and %φL diverges; the ratio IL /ILd,1
tends to (1 − z1 /z3 ) when c1b /c2b 1 in 1 cs
ηc ≡ ln 1b (59)
systems with other charge numbers, as z1 f c1
can be deduced from Eq. (49).
The current–voltage curves given by is the concentration overpotential. The po-
Eq. (55) have been represented in Fig. 6. tential drop %φ − ηc is the change in elec-
The initial slope of these curves is κeff b , trode potential (relative to bulk solution)
which obviously increases with decreasing due to the concentration polarization.
2.0 1000
50
1.5 5
1
I/ILd, 1
0.1
1.0
Fig. 6 Current–voltage curves of a 0.5

depleted DBL for different values of the
electroactive/supporting electrolyte
concentration ratio, c1b /c2b , shown on the 0.0
0 1 2 3 4
curves. %φ is the electric potential drop
z1 f∆f
in the DBL.
Fig. 7 Current–overall voltage

2.0 1000 drop for different values of the
50 electroactive/supporting
1.5 5 electrolyte concentration
ratio, c1b /c2b .
I/ILd, 1
1
1.0
0.1
0.5
0.0
0 2 4 6
z1 f(∆f − hc)
A validity test for the excess supporting is not affected by the passage of the elec-
electrolyte assumption (in the homovalent tric current. It must be stressed, however,
case) can be obtained by calculating the that this statement refers to the concentra-
migrational flux from Eqs. (37 and 38) as tion gradients in relative terms to the bulk
concentration of the respective species. In
dφ z 2 c1 absolute terms, the concentration gradi-
J1,mig ≡ −z1 D1 c1 f = 1 J1
dx zi2 ci ents of the inactive species are similar to
i that of the active species, and by no means
c1 the concentration gradient of the active
= J1 (60) species can be neglected. In fact, in the
2c3
homovalent case, it can be easily shown
and comparing it to the diffusional flux that dc1 /dx ≈ 2dc3 /dx ≈ −2dc2 /dx when
migration is negligible. A straightforward
c1 implication of these comments is that
J1,dif = J1 − J1,mig = J1 1 −
2c3 the LEN assumption cannot be used in
(61) conjunction with the excess supporting
Equations (60 and 61) show clearly that electrolyte assumption.
the migrational contribution to J1 is
negligible compared to the diffusional one
in as much as c1 is negligible compared 2.1.2.5 Weak Binary Electrolyte
to c3 . Alternatively, the importance of Consider an electrolyte Aν1 Bν2 dissociat-
migration can be estimated from Eq. (58), ing into ν1 ions Az1 of charge number z1
since the term c1b /4c2b accounts for the and ν2 ions B z2 of charge number z2
migrational contribution to IL .
When the supporting electrolyte is elec- −−−
Aν1 Bν2 ← −−
→− ν1 A + ν2 B
z1 z2
(62)
troinactive (and chemically inert), its pres-
ence can be totally ignored in the descrip- where z1 ν1 + z2 ν2 = 0. Denote by c12 =
tion of transport processes. The absence αc12,T and c12,u = (1 − α)c12,T the con-
of migration and the zero flux condition centrations of dissociated and undisso-
for the inactive species implies that these ciated electrolyte fractions, respectively,
species have an approximately uniform being α the degree of dissociation and
concentration throughout the DBL, which c12,T = c12 + c12,u the total electrolyte
concentration. When the dissociation Re- If only species 1 is electroactive, the

action (62) is fast compared with the trans- boundary conditions at the electrode
s
surface are J12,u = 0, J2s = 0, and J1s =
port process, it can be assumed that the
dissociated and undissociated electrolyte I /z1 F , and therefore the flux density
fractions are in equilibrium and the mass J12,T can be evaluated, from Eq. (21)
action law and J2s = 0, as
c1ν1 c2ν2 ν1 ν2 ν12
ν12 −1 ν1 ν2 α t2 I
K= = c12,T (63) J12,T = − (67)
c12,u 1−α z2 ν2 F
allows then to determine α as a function The concentration profile of the electrolyte

of c12,T . In deriving Eq. (63), the following c12,T could then be determined from
relations have been used, c1 = ν1 c12 , c2 = Eqs. (63, 65, and 67), and the boundary
b
value c12,T . Note that the transport num-
ν2 c12 , and ν12 ≡ ν1 + ν2 .
Because of the electrolyte dissociation, bers are constant in a binary system.
the flux densities Ji of the ions and that of For the determination of the cur-
the undissociated electrolyte rent–voltage characteristics of this system,
the integration of Eq. (65) over the DBL
dc12,u
J12,u = −D12,u (64) t2 I δ
dx −J12,T δ =
z2 ν2 F
vary with position. The same applies to the = D12,u %c12,u + D12 %c12 (68)
flux density of the dissociated electrolyte,
J12 [35]. The total electrolyte flux can be conveniently rewritten in the form
s s

dc12,u D12,u c12,u + D12 c12
J12,T = J12,u + J12 = −D12,u I = IL,T 1 −
dx b
D12,u c12,u b
+ D12 c12
dc12 dc12,T (69)
− D12 = −D12,T (65) where
dx dx
however, is constant throughout the sys- z2 ν2 F b b
IL,T = (D12,u c12,u + D12 c12 )
tem as required by the continuity equation, t2 δ
(70)
Eq. (11). In spite of this fact, the total elec-
is the limiting current density defined
trolyte concentration c12,T does not vary
from the condition that the electrolyte
linearly with position because the effective
concentration (both dissociated and undis-
diffusion coefficient of the electrolyte
sociated) becomes zero at the electrode
ν12 (1 − α)D12,u + αD12 surface when I → IL,T . This limiting cur-
D12,T ≡ (66) rent density reduces to
ν12 (1 − α) + α
b
z2 ν2 F D12,u c12,T
is a function of α, and α is a function
IL,u ≡ (71)
of c12,T . Exception is made for the t2 δ
special cases of very weak electrolyte (α →
and
0, K → 0, D12,T → D12,u ), very strong
electrolyte (α → 1, K → ∞, D12,T → b
z2 ν2 F D12 c12,T b
z1 ν12 F D1 c12,T
D12 ), and D12,u = D12 , when D12,T is a IL ≡ =−
t2 δ δ
constant and c12,T varies linearly. (72)
in the limits of very weak (α → 0) and very constant K when D12,u = D12 , for in this
strong electrolyte (α → 1), respectively. In- case, IL,T = IL,u = IL .
deed, Eq. (70) can be easily transformed to The paradoxical results mentioned above
can be explained in terms of the electric po-
D12,u tential drop across the DBL (see Appendix
IL,T = IL (1 − α b ) + α b (73)
D12 A for its derivation)
where α b is the degree of dissociation at RT cb RT ν12 D12,u
x = δ. %φ = − ln 12
s − F z (ν − 1)D
z2 F c12 2 12 2
Figure 8 shows the dependence of the b
c12,u s
c12,u
limiting current on the dissociation de-
× − s (74)
gree for the cases of 1 : −1 and 2 : −1 b
c12 c12
electrolytes and different values of the ratio
D12,u /D12 . At very low electrolyte concen- This potential drop increases when K
tration or very high dissociation constant, decreases and there are less ions to
the electrolyte is completely dissociated conduct the electric current. In particular,
and IL,T = IL . At very high electrolyte %φdif vanishes and %φohm diverges when
concentration or very low dissociation K → 0, so that the sentence in the
rate constant, the electrolyte is practically previous paragraph should have been
undissociated and IL,T = IL,u . The ra- written as follows: the system is able to
tio IL,T /IL is then equal to D12,u /D12 . conduct an electric current in the limiting
Surprisingly, in this limit of negligible case α → 0, provided that an infinite
dissociation when there are no ions in potential difference is applied across the
solution, the system conducts electric cur- depleted DBL.
rent and, moreover, the limiting current The current–voltage curves of this
can be greater than for a fully dissociated system can be obtained by combining
electrolyte if D12,u > D12 . Note also from Eqs. (69 and 74). Figure 9 shows these
Fig. 8 that the limiting current density curves for a 1 : −1 electrolyte, with an
becomes independent of the dissociation electroactive cation, and different values of
(cb12)2/K
10−2 100 102 104
2
2
Fig. 8 Limiting current density in a

IL, T /IL
depleted DBL containing a weak

1 electrolyte solution relative to the value
1 corresponding to complete dissociation.
0.5 The lower abcisa scale and the
continuous lines correspond to a 1 : −1
electrolyte. The upper abcisa and the
0 dashed lines correspond to a 2 : −1
10−2 100 102 104 electrolyte. The ratio D12,u /D12 has been
cb12/K given the values 0.5, 1, and 2, as shown
on the curves.
Fig. 9 Current–voltage curves of a

0.01
depleted DBL containing a weak 1 : −1 1.0
electrolyte for the values of the ratio 0.1
b
c12,T /K, which are shown on the curves. 1
Both the ratio D12,u /D12 and the 10
IL, T /IL
transport number t1 have been set equal
to 0.5. 0.5 100
1000
0.0
0 2 4 6 8 10
f∆f
the bulk electrolyte concentration relative s

that the value of α s is independent of c12,T ,
to the dissociation constant, c12,T b /K. which is inconsistent with Eq. (63). It is
When the dissociation degree is close then concluded that Eq. (63) is not valid in
to one, that is at low values of the the vicinity of the electrode. The study of
b
ratio c12,T /K, the curves are practically the limiting current without the assump-
identical to that in Fig. 2. However, tion that the dissociation reaction is fast
when the ratio c12,T b /K is high and compared with the electrodiffusion pro-
the electrolyte is mostly undissociated, cess has been carried out by Kharkats and
the electric conduction becomes difficult. Sokirko [36, 37]. The validity of the dis-
b
Thus, for instance, in the case c12,T /K = sociation equilibrium assumption and the
100 where α = 0.031 and IL,T /IL =
b drastic changes in concentrations that oc-
0.516, the electric potential drop in the cur when I → IL have been considered by
DBL, %φ, must be of the order of 15 times Vorotyntsev [38].
RT/F for the current density to be close to
this limiting value. No need to mention 2.1.3
that large values of the ratio c12,T b /K Steady State Current–overpotential Curves
also imply lower concentration of the in the Presence of Supporting Electrolyte
electroactive species close to the electrode
and, therefore, larger electrode|solution The use of an excess amount of inactive
potential drops (i.e. Nernstian potential electrolyte makes negligible the migra-
drops in the case of reversible electrode tional contribution to mass transport and
reactions). However, the current–voltage therefore simplifies the theoretical mod-
curves do not change qualitatively when eling of experimental observations. In
the electrode potential is taken into particular, the absence of potential drop
account in the voltage drop. in the DBL eliminates the coupling of
Note, finally, that the boundary condi- ionic fluxes. The current–voltage curves
s
tion J12,u = 0 requires (except in the trivial are then presented in terms of the concen-
case I = 0, which lacks interest) that the tration overpotential. Mass transport in the
dissociation degree α is unity in the vicinity presence of homogeneous reactions can be
of the electrode, independent of the values described with relative simplicity in these
of I and K. This is equivalent to stating systems, as shown in the last section.
2.1.3.1 Systems with One Electroactive The concentration overpotential in Eq. (32)
Species then takes the form
In the absence of migration, the flux
1 r I
density of the electroactive species is ηc = νi ln 1 − (80)
nf ILd,i
i
I dc1 cb − c1s
J1 = ≈ −D1 = −D1 1 Consider, for example, the anodic oxida-
z1 F dx δ
(75) tion of iron(II)
and the interfacial concentration can be
Fe3+ (aq) − Fe2+ (aq) + e− = 0 (81)
expressed as
in a situation in which the bulk (aqueous)
s b I
c1 = c1 1 − (76) solution is equimolar in ferrous and ferric
ILd,1 ions. Since the ferrous ion is consumed
where ILd,1 is the limiting diffusion at the electrode, the limiting current
current density defined by Eq (40). Equa- exhibited by the system is the diffusion
tion (76) can be rewritten in terms of the limiting current of this ion and the
concentration overpotential, Eq. (32), as current–overpotential curve is
1 1 − I /ILd,Fe3+
I = ILd,1 (1 − ez1 f ηc ) (77) ηc = ln
f 1 − I /ILd,Fe2+
This equation is valid both for a depleted 1 + (DFe2+ /DFe3+ )I /ILd,Fe2+
1
DBL, in which I /ILd,1 > 0 and z1 ηc < 0, = ln
f 1 − I /ILd,Fe2+
and for a concentrated DBL, in which
I /ILd,1 < 0 and z1 ηc > 0. (82)
which has been represented in Fig. 10.
2.1.3.2 Systems with Several Electroactive
Species 2.1.3.3 Diffusion-reaction Processes
When several species are involved in the The general strategy for solving transport
electrode reaction described by Eq. (23), problems in which the electrode reaction,
the reaction rate is determined by the Eq. (23), is coupled to a homogeneous
reactant (νir < 0) with slower diffusion reaction in the DBL
rate. Thus, the limiting current exhibited
by the system is the smallest of the limiting νi Cizi = 0 (83)
diffusion current densities [26] i
nF Di cib is outlined in this section. The stoichio-

ILd,i ≡ − (78) metric numbers are defined to be positive
νir δ for products and negative for reactants.
of the reactants. Note that I can be either Chemical equilibrium is assumed for both
positive or negative, but it is always of the reactions, which is justified from the fact
same sign as ILd,i for the reactants. that chemical kinetics is usually much
From Eqs. (25 and 78), the interfacial faster than diffusion processes.
concentration of species i is given by The homogeneous reaction is con-
sidered to be reversible. The chemical
cis = cib +
Ji δ
= cib 1 −
I
(79)
equilibrium
assumption, i νi µ̃i ≈ 0 or
Di ILd,i K ≈ i ciνi , applied to transport problems
Fig. 10 Current density versus 1.0

concentration overpotential for the
anodic oxidation of iron(II) from an
equimolar solution of ferrous and ferric
ions. The diffusion coefficients
DFe3+ = 0.604 × 10−5 cm2 s−1 and
I/ILd, Fe2+
DFe2+ = 0.719 × 10−5 cm2 s−1 [39] 0.5
have been used.
0.0
0.0 2.0 4.0 6.0
fhc
means that the forward and backward re- given chemical element is not modified.
action rates are much larger than the It is then correct to state that the sum
mass transport by diffusion, and hence of the flux densities, multiplied by a
their difference can be neglected in a convenient stoichiometric coefficient, of
first approximation. Actually, there is a all species that contain a given chemi-
small difference between the forward and cal element involved in Reaction (83) is
backward reaction rates, the net reac- independent of position. By considera-
tion rate tion of all the chemical elements, a large
number of equations are obtained. The
ri 1 dJi
r≡ = (84) elimination of those that are simply lin-
νi νi dx
ear combinations of the others leads to
which is responsible for the spatial a much smaller number of independent
changes of the flux densities of the equations. Alternatively, those groups of
reactive species. In the description of atoms that are not modified by the re-
transport processes, this equation can be action can be considered as a unique
used to set up relations among the flux entity. This is the so-called formalism
density gradients of the reactive species of constituents, which leads faster to the
and to evaluate the reaction rate. If same number of independent equations.
the reaction rate constants are known, Consider, for instance, that the chemical
the validity of the fast chemical equilib- reaction involves several species with a
rium assumption should also be checked cyanide group. The reaction cannot make
a posteriori. the cyanide group disappear but it can
The continuity equation under steady simply bind it to one or another chemical
state conditions, Eq. (84), states that the species. Thus, the flux of cyanide con-
flux densities of the reactive species change stituent (defined as the group of chemical
with position because they are either con- species that contain cyanide) is not affected
sumed or generated in the homogeneous by the chemical reaction and is constant
reaction. The reaction involves the trans- throughout the DBL (under steady state
formation of chemical species by the conditions).
exchange of atoms or groups of atoms These conservation equations (one for
among them, but the total amount of a each constituent) are generically written in
the form where ν2r = ν3r = νI,1 = νCd,2 = 0, ν1r =

νCd,1 = νCd,3 = νI,2 = 1, n = 2, and νI,3 =
dJj dJi
= νj,i =0 (85) 4. These two equations are to be solved
dx dx together with the chemical equilibrium
i
condition
where the sum extends over all species i c3
K≈ (91)
that contains the constituent (or chemical c1 c24
element) j , and νj,i is the stoichiometric
using the bulk concentration val-
number of constituent j in species i. Tak-
ues as boundary conditions. The dif-
ing into account that the flux densities of
fusion coefficients and complexation
the different species at the electrode sur-
equilibrium constants are D1 = 0.7 ×
face are given by Eq. (24), and neglecting
10−5 cm2 s−1 , D2 = 2.0 × 10−5 cm2 s−1 ,
migration owing to the presence of ex-
D3 = 0.5 × 10−5 cm2 s−1 , and K = 2 ×
cess supporting electrolyte, Eq. (85) can be
106 M−4 [39, 40].
written as
A simple way to solve this equation
dci system is to consider c3 as the indepen-
νj,i Ji = − νj,i Di dent variable and evaluate c2 and c1 from
dx
i i
Eqs. (90 and 91), respectively. Finally, in-
I tegration of Eq. (89) yields the position
= νj,i νir (86)
nF variable. Figure 11 shows the concentra-
i
tion profiles thus obtained for the cases
Note that the equation for the conduction of anodic dissolution (I > 0) and cathodic
current density, Eq. (12), is implicit in the discharge of cadmium (I < 0). In the first
system of Eqs. (86). case, the cadmium and complex ion con-
As an example, consider the dissolution centration gradients are negative (i.e. sur-
of a cadmium metal electrode face concentration larger than bulk value)
and the iodide concentration gradient is
Cd2+ (aq) − Cd(metal) + 2e− = 0 (87) positive. The DBL is then concentrated in
the cadmium constituent. In the second
in a solution containing iodide ion, case, the cadmium ion is consumed by
which can complex to cadmium to form the electrode reaction and the DBL is de-
tetraiodocadmiate pleted in the cadmium constituent. The
reaction rate evaluated from Eq. (84) has
[CdI4 ]2− (aq) − Cd2+ (aq) − 4I− (aq) = 0 also been represented. It is thus evidenced
(88) that the rate of the complexation reaction
Denoting the species Cd2+ , I− , and is large in the case of the anodic disso-
[CdI4 ]2− by subscripts 1 to 3, respectively, lution and that the reaction front shifts
and considering cadmium Cd and iodine towards the bulk solution with increasing
I as two constituents, Eq. (86) takes current. On the contrary, the reaction rate
the form is negligible in the case of the cathodic dis-
charge, in which the concentration profiles
dc1 dc3 I are practically linear. The electric current
D1 + D3 =− (89)
dx dx 2F values have been presented relative to the
dc2 dc3 limiting diffusion current density of cad-
D2 + 4D3 =0 (90) mium ion, ILd,1 = −2F D1 c1b /δ. It can be
dx dx
100 I− I−
1.0
ci(x)/cbi
[mM]
ci(x)
50 −104 [CdI4]2− 102 0.5
[CdI4]2−
Cd2+
Cd2+
0 0.0
100 I− −2 × 104 2
[mol L−1 s−1]

[CdI4]2−
[mM]
ci(x)
−2 × 104
r
50 1
−104
Cd2+
0 102 0
0.0 0.5 1.0 0.0 0.5 1.0 1.5
x/d x/d
Fig. 11 Concentration profiles in a cadmium iodide solution adjacent to a
cadmium metal electrode and the rate of formation of tetraiodocadmiate
ion. The bulk concentrations are c1b = 5 µM, c2b = 100 mM, c3b = 1 mM, and
the three values of the electric current density have been considered:
I/ILd,1 = 102 (cathodic cadmium discharge), −104 and −2 × 104 (anodic
cadmium dissolution).
deduced from Eq. (89) that the limiting first the electric potential distribution and
current density (corresponding to c1s = then the current–voltage curves or the
c3s = 0) is IL = −2F [D1 c1b + D3 c3b ]/δ. For ionic concentration distributions in the
the bulk concentrations considered in DBL. The generality of the method, and
Fig. 11, c1b = 5 µM, c2b = 100 mM, c3b = the inherent complexity of the nonlinear
1 mM, and a DBL thickness δ = 10−3 cm, migration term in the flux equations
these currents are ILd,1 = −0.67 and IL = could make it not very attractive for the
−97.2 mA cm−2 . nonexpert, and therefore simpler solutions
based on the Goldman constant field
2.1.4 (GCF) assumption are also presented. For
Steady State Current–voltage Curves the sake of simplicity, no homogeneous
of Systems with Several Active Species reactions are considered.
There are electrochemical processes in 2.1.4.1 Kramer’s Integration Method

which little or no supporting electrolyte The mathematical difficulty of the elec-
is used. This is the case, for instance, trodiffusion equations, Eq. (7), arises from
of the study of overlimiting currents [34, the nonlinear migration term, which
41, 42], and of microelectrode [43–45], and couples the Nernst–Planck equations of
different voltammetric techniques [17, 18, the different ionic species. By assuming
46–48], in which the absence of supporting that the electric field is constant (GCF
electrolyte increases the sensitivity of the assumption), this difficulty is removed
detection of the redox species. In this and analytical solutions are easy to find.
section, a general procedure for solving the Unfortunately, the validity conditions for
electrodiffusion equations is presented. this approximation are not always satisfied
The procedure is based on determining in electrochemical systems [12, 49], and it
is then needed either to derive approxi- method, that uses the concept of va-
mate expressions for the electric potential lency classes [32], these authors proposed
distribution in some other way [50] or to rewrite the Nernst–Planck equation
to calculate the exact distribution as de- system for N species in terms of the
scribed below. In any case, if the electric magnitudes
field distribution E(x) or, equivalently, the
electric potential distribution
N
Sk ≡ zik ci , k = 0, 2, 3, . . . , N (95)
δ i=1
φ(x) = φ b + Edx (92)
x
N
Gk ≡ zik Ji /Di ,
is considered to be known, the transport i=1
equations can be formally integrated by
k = 0, 1, 2, . . . , N − 1 (96)
Kramer’s transformation [51]. This con-
sists of multiplying both terms of Eq. (7)
instead of the N concentrations ci and flux
by ezi f φ and integrating from x to δ to
densities Ji . The LEN assumption
in this
obtain [8, 52, 53]
notation takes the form S1 ≡ i zi ci = 0,
δ and this justifies the absence of index
Ji
ezi f [φ(x )−φ ] dx
b
ci (x) = cib + k = 1 in Eq. (95). The aim of the method
Di x
is obtaining an equation for the electric
b
−φ(x)]
× ezi f [φ (93) field in which the unknown concentrations
are eliminated by using their relation to
The surface concentration cis is obtained the flux densities, which are known from
by setting x = 0 in this equation. Eq. (24).
When the GCF approximation φ(x) = Multiplying the Nernst–Planck equa-
φ b + (δ − x)E is valid, the integral in tion, Eq. (7), by −1/Di and summing
Eq. (93) can be evaluated and this equa- over species i, the electric field is
tion becomes eliminated and the concentration sum

Ji S0 ≡ i ci is shown to follow the lin-
ci (x) = cib − ezi f E(x−δ) ear profile
zi Di f E
Ji S0 (x) = S0b + (δ − x)G0 (97)
+ (94)
zi Di f E
Similarly, multiplying the Nernst–Planck
2.1.4.2 The Equation for the Electric Field equation by zi /Di and summing over
under the LEN Assumption species i, it is obtained that
Nikonenko and Urtenov [54, 55] have
presented an algorithm for the solu- G1 = f ES2 (98)
tion of the steady state transport equa-
tions in multicomponent systems, which which could be used to evaluate the electric
is based on calculating first the elec- field if the relation between S2 and S0 were
tric field E and then using Eqs. (92 known. Note that S2 is equal to double
and 93). Their method is outlined here the ionic strength of the solution, and
for the particular case when the LEN the advantages of using it as a variable
assumption is used. Unlike Schlögl’s in the formulation of transport problems

have also been emphasized by other IL
authors [17, 18]. Some simple cases are cis = cib 1+
(1 + zi 3)ILd,i
considered next and the general approach

is presented later. I 1+zi 3 I −zi 3
In the case of a binary solution, S2 = × 1− −1 1−
IL IL
−z1 z2 S0 the electric potential can be
obtained from Eq. (98) as (104)
which reduces to cis = cib (1 − I /IL ) in the
3 S0 (x) binary case. In deriving Eq. (104), note that
φ(x) = φ b + ln (99)
f S0b −Ji δ/Di cib = I /ILd,i .
In the case of nonhomovalent multi-
where ionic systems, the solution procedure
1 G1
3≡− (100) is necessarily more complicated because
z1 z2 G0 there is no simple relation such as S2 =
Moreover, introducing the limiting cur- z2 S0 , a problem that was first encountered
rent density and solved in different ways by Schlögl [59]
and Brady and Turner [60]. If the system
S0b I contains N different ionic species, the N
IL ≡ − (101) Nernst–Planck equations for the N flux
δ G0
densities Ji are replaced by
the steady state current–voltage curve can dSk
be deduced from Eq. (99) as Gk = − + f ESk+1 ,
dx
3 cs k = 0, 1, 2, . . . , N − 1 (105)
%φ = − ln bi
f ci In order to solve the system of Eqs. (105),

3 I a closure relation between SN and other
= − ln 1 − (102) sums is needed. Such a relation is
f IL
SN = q1 SN −1 − q2 SN −2 + q3 SN −3
In a multi-ionic, homovalent solution,
|zi | = z, S2 = z2 S0 and Eq. (99) is again − · · · + (−1)N −1 qN S0 (106)
obtained [56–58], though 3 is now defined
as 3 ≡ G1 /z2 G0 . Since the integral in where
Eq. (93) can be evaluated by using Eqs. (97 q1 = z1 + z2 + · · · + zN (107)
and 99) as
q2 = z1 z2 + z1 z3 + · · · + z1 zN + z2 z3 + · · ·
δ
δ S0zi 3 dS0 + z2 zN + z3 z4 + · · · + zN −1 zN (108)
zi f (φ−φ b ) 0
e dx = − q3 = z1 z2 z3 + · · · + z1 z2 zN + z1 z3 z4 + · · ·
0 G0 (S0b )zi 3
 1+zi 3  + z1 z3 zN + · · · + zN −2 zN −1 zN (109)
S0b δG0
=  1+ − 1 ..
G0 (1 + zi 3) .
S0b
N +2−k
+1−k N
N
(103) qk = ··· zi1 zi2 . . . zik
the interfacial concentration is finally i1 =1 i2 =i1 +1 ik =ik−1 +1
given by (110)
Some particular cases are considered in dimensionless electric field e ≡ f Eδ as

the next sections.
de x
2
= e 0.2729 + e 0.7271 1 −
d(x/δ) δ
2.1.4.3 Ternary Electrolyte Solutions

In a ternary system (N = 3), Eqs. (97 and 4cb ILd,1
98) must be solved together with + 11.456 1 + 2b (118)
3c1 I
dS2
G2 = − + f E S3 (111)
dx where ILd,1 = F D1 c1b /δ is the limiting
S 3 = q1 S 2 + q3 S 0 (112) diffusion current density of ferrous ion.
Equation (118) can be integrated by
After some algebra, Eq. (111) can be standard numerical methods, such as
transformed into the following equation fourth-order Runge–Kutta method [61],
for the electric field starting from position x = δ, in which
G1 dE the boundary condition eb = G1 δ/S2b is
= (G2 − q1 G1 )f E − q3 (f E)2 S0 (x) applied as
E dx
(113) x I 0.2619
which has to be integrated numerically e =1 = (119)
from x = δ to x using the boundary δ ILd,1 1 + 2c2b /c1b
condition Eb = G1 /f S2b . The potential drop %φ (and its distribu-
As an example, consider the case of a tion, if needed) can then be calculated by
redox system such as FeCl2 + FeCl3 in numerical integration of the electric field
the absence of supporting electrolyte in a using, for example, a fourth-order New-
potential range in which the only possible ton–Cotes method [61]. This procedure
electrode process is the anodic oxidation yields the current–voltage curves shown
of ferrous ion in Fig. 12 in the case of a DBL depleted
Fe3+ (aq) − Fe2+ (aq) + e− = 0 (114) of ferrous ions, that is, for I /ILd,1 > 0
and %φ < 0. These curves end when they
Denoting the ions Fe2+ , Fe3+ , and Cl− meet the dashed line that corresponds to
by subscripts 1, 2, and 3, respectively, the the limiting current density, that is, to
ion fluxes are J1 = −J2 = −I /F , and the vanishing surface concentration of ferrous
parameters Gk take the values ion. The limiting current density is a bit
smaller than the limiting diffusion current
1 1 I I
G0 = − = 0.1904 density ILd,1 because the ferrous cation is
D2 D1 F F D1
(115) driven towards the anode by its concentra-
tion gradient against an electric field that
z2 z1 I I
G1 = − = 1.5712 pushes it away from the electrode. The
D2 D1 F F D1
strength of this electric field is reduced
(116)
when the solution conductivity increases,
z22 z12 I I that is, when the bulk concentration ratio
G2 = − = 6.7136
D2 D1 F F D1 c2b /c1b increases for a fixed value of c1b , and
(117) the limiting current density tends to ILd,1
where D1 = 0.719 × 10−5 cm2 s−1 and when this occurs.
D2 = 0.604 × 10−5 cm2 s−1 [39]. Equa- The ionic concentration distributions
tion (113) can be rewritten in terms of the in the DBL are represented in Fig. 13.
Fig. 12 Current–voltage curves of a 1.0

ternary system FeCl2 + FeCl3 in the
absence of supporting electrolyte for 10
2
different values of the bulk 1
0.5
concentration ratio c2b /c1b shown near 0.1 0
the curves. The dashed line represents
I/ILd, 1
IL /ILd,1 obtained from the (exact) 0.5
numerical solution, and the dotted line
represents IL /ILd,1 obtained from the
GCF assumption.
0.0
0.00 0.05 0.10 0.15
−f∆f
The ferrous ion concentration is ob- estimated as

tained from Eq. (93) and using the rela- IL 2f %φ
tion −J1 δ/D1 c1b = I /ILd,1 . The chloride = (120)
ILd,1 1 − e−2f %φ
ion follows Boltzmann’s equation c3 =
b
c3b ef (φ−φ ) , and the ferric ion concentra- which has been represented in Fig. 12
tion is obtained from the LEN assumption and agrees very well with the exact
c2 = (c3 − 2c1 )/3. numerical solution.
Figures 12 and 13 evidence that the
potential drop in the DBL is small and 2.1.4.4 Quaternary Electrolyte Solutions
is distributed quite uniformly. It is then Equations (97, 98 and 111), and
expected that the GCF assumption, E =
−%φ/δ, provides reasonable estimates. dS3
G3 = − + f E S4 (121)
Indeed, by setting c1s = 0 in Eq. (94), dx
the limiting current density can be S 4 = q1 S 3 − q2 S 2 − q4 S 0 (122)
6 0.09
Cl−
4 0.06
f[f(x) − fb]
f
ci(x)/c1b
2 Fe3+ 0.03
Fe2+
0 0.00
0.0 0.5 1.0 1.5
x/d
Fig. 13 Electric potential and concentration profiles of Fe2+ ion
(i = 1), Fe3+ ion (i = 2), and Cl− ion (i = 3) for a bulk
concentration ratio c2b /c1b = 1.0 and the limiting current density
I = IL = 0.9261ILd,1 .
lead in this case to the following equation 2.1.2.4 was restricted to steady state condi-
for the electric field tions, but it is expected that a large excess
of supporting electrolyte makes migration
G1 d2 E 3G1 dE 2 negligible under transient conditions too.)
− + (G2 − q1 G1 )
E dx 2 E2 dx This section describes diffusion and mi-
dE gration under transient conditions in some
×f − (G3 − q1 G2 + q2 G1 simple illustrative cases. For the sake
dx
of simplicity, an infinite DBL thickness
+ q4 S0 f E)(f E)2 = 0 (123) δ → ∞ is considered, which is justi-
fied in the absence of stirring. Current
which has to be integrated numerically and voltage steps in semi-infinite pla-
from x = δ to x using the boundary nar geometry (and in the absence of
conditions Eb = G1 /f S2b and homogeneous reactions) are described.
Both Laplace transforms and Boltzmann’s
dE G1 S3b change of variables are used. The influ-
= G2 − G1 b (124)
dx x=δ f (S2b )2 S2 ence of migration is discussed analytically
(whenever possible) by using the latter
Although the mathematical difficulty of technique, which is introduced in Ap-
this equation system is not very high, pendix B. Other solution techniques of
the GCF assumption, with the electric the diffusion equation [62–64], and the
field estimated, for example, as E ≈ Eb = electrodiffusion equations are available in
G1 /f S2b , is a much simpler alternative to the literature [13, 65–72]. Further infor-
be considered when the ionic strength is mation on electrochemical techniques can
roughly uniform over the DBL. be found in Ref. [73].
2.1.5 2.1.5.1 Fick’s Second Law for a Strong

Ion Transport under Transient Conditions Binary Electrolyte
The diffusion–conduction equation,
Needless to say, transient transport con- Eq. (21), expresses the flux density of
ditions involve a higher level of mathe- an ionic species from a strong binary
matical complexity to the description of electrolyte in terms of the electrolyte
ion transport in electrochemical systems. flux J12 and the current density I .
The only case that can be worked out with Under transient conditions, both Ji
ease is that of binary solutions. Ternary and J12 are position-dependent, but
and multi-ionic solutions necessarily in- I is not. Thus, taking into account
volve systems of partial differential equa- Eq. (21) and the relation ci = νi c12 , the
tions coupled through the migration term continuity equation, Eq. (11), leads to
that require advanced numerical methods. Fick’s second law
Given this situation, it is by no means
∂c12 ∂ 2 c12
surprising that most theoretical studies of = D12 (125)
transient problems neglect migration and ∂t ∂x 2
use the diffusion equation (or Fick’s sec- The transient ionic electrodiffusion can
ond law) to describe the transport of ionic then be described as electrolyte dif-
species. (The analysis of the role of sup- fusion and solved by standard tech-
porting electrolytes in Sects. 2.1.2.2 and niques [62–64].
2.1.5.2 Systems with One Active Species 2.1.5.3 Voltage Step in the Presence of
and Supporting Electrolyte Supporting Electrolyte
The transport of the active species in the The boundary condition, Eq. (128), for a
presence of excess supporting electrolyte voltage step perturbation reads
under transient conditions is described by
the equation c1s (t) = c1b exp[z1 f %EH (t)] (129)
where H (t) is the step function [H (t) = 0

∂c1 ∂ 2 c1
= D1 2 (126) if t < 0, H (t) = 1 if t > 0], and %E is the
∂t ∂x width of the voltage step.
Using Laplace transformation with re-
which is obtained from the continuity
spect to time, the general solution of
equation, Eq. (11), by neglecting the migra-
Eq. (126) in the Laplace domain is [62]
tional contribution to the flux density J1 .
Although Eq. (126) resembles Fick’s sec- c1b 1/2
ond law, it should be observed that Fick’s c̃1 (x, s) = + Aex(s/D1 )
s
laws are valid for (neutral) electrolytes. Ion
+ Be−x(s/D1 )
1/2
transport takes place by electrodiffusion (130)
and satisfies pseudo-Fick’s laws only when
where c̃1 (x, s) is the Laplace transformed
migration is negligible.
concentration and s is the Laplace variable.
For a simple electrode process, like
Since the concentration must be bounded,
cathodic deposition or anodic dissolution
the integration constant A vanishes. The
(without contribution of nucleation), the
boundary condition, Eq. (129), gives the
initial value is c1 (x, t = 0) = c1b and the
value for B, resulting in
boundary value at the bulk solution is
c1b (t) = c1b . The boundary condition at the c1b
electrode surface can be either Eq. (24), in c̃1 (x, s) = 1 + (ez1 f %E − 1)
s
the form
× e−x(s/D1 )
1/2
(131)
∂c1 I (t)
=− (127)
∂x x=0 z1 F D1 By inverse Laplace transformation, the
concentration profile is obtained as
or the Nernst equation, Eq. (31), for the
c1 (x, t) = c1b [1 + (ez1 f %E − 1)erfc(ζ )]
case of a single active species
(t > 0) (132)
◦ RT cs (t)
Eeq (t) = E + ln 1 ◦ (128) 1/2
where ζ ≡ x/(2D1 t 1/2 ) is the Boltzmann
z1 F c
variable (see Appendix B) and erfc stands
When the system is perturbed by apply- for the complementary error function.
ing a known function, I (t), Eq. (127) is The electric current response is obtained
used to evaluate the chronopotentiomet- from Eqs. (127 and 132) as
ric response of the system. Analogously,
∂c1
when the system is perturbed by applying a I (t) = −z1 F D1
∂x x=0
known function Eeq (t), Eq. (128) describes
1/2
the chronoamperometric response of the D1
= z1 F c1b (ez1 f %E − 1) (133)
system. πt
If the potential step width is large enough, 2.1.5.5 Voltage Step in the Absence of
−z1 f %E 1, the surface concentration Supporting Electrolyte
of the electroactive species is negligible Although it is important for practical
and Eq. (133) reduces to the Cottrell purposes to attempt the treatment of
equation [73] systems which are not stripped of so
1/2 many features that they no longer bear
D1 resemblance to reality, the simultaneous
I (t) = −z1 F c1b (134)
πt treatment of a reasonable fraction of
these features makes the mathematical
2.1.5.4 Current Step in the Presence of problem so complex that there is a danger
Supporting Electrolyte of obscuring the real problem [74]. It is
Consider that the function I (t) in Eq. (127) then suggested to simplify the present
is a current step I (t) = I0 H (t), where I0 analysis of a large voltage step in a
is the constant current applied from t = 0. ternary solution (where the common ion
Equation (130) reduces now to is inactive) by making equal all diffusion
coefficients, Di = D, for i = 1,2,3, and
c1b I0 hence eliminating the diffusion potential.
s −3/2 e−x(s/D1 )
1/2
c̃1 (x, s) = +
s z1 F D1
1/2 This allows for a better understanding of
(135) the effect of migration on the transient
whose inverse transform provides the response of the system.
concentration distribution The treatment in Sect 2.1.4.3 is followed
and magnitudes Sk and Gk are used in-
2I0 t 1/2 stead of ionic concentrations and flux
c1 (x, t) = c1b + 1/2
z1 F D1 densities, respectively. However, the re-
lations between c1 , c2 , c3 , and S0 , S2 , are
× [π −1/2 e−ζ − ζ erfc(ζ )]
2
(136) still needed because the Nernst equation,
Eq. (31), involves ionic species.
The potential response is given by Eq. (128)
The basic equations describing this
with the surface concentration determined
problem are the continuity equations,
from Eq. (136) at ζ = 0, that is,
Eq. (11), which owing to the simplifying
◦ RT assumption Di = D reads as
Eeq (t > 0) = E +
z1 F
∂Sk ∂Gk
1/2 = −D (139)
cb 2I0 t ∂t ∂x
× ln 1◦ + (137)
c z1 F c◦ πD1
and Eq. (105) for k = 0, 1, and 2,
In cathodic processes (I0 < 0), the surface ∂S0 (x, t)
concentration becomes zero at the transi- G0 = − (140)
∂x
tion time
∂φ(x, t)
2 G1 = −f S2 (x, t) (141)
z1 F c1b ∂x
τ = πD1 (138) ∂S2 (x, t)
2I0 G2 = − + q1 G1
∂x
This relationship is called the Sand equa- S0 (x, t)
tion [73]. + q3 G1 (142)
S2 (x, t)
where the closure relation, Eq. (112), The difficulty introduced by the second
and Eq. (141) have been used to obtain term in the right-hand side of Eq. (147)
Eq. (142). Equation (141) is Ohm’s law, makes it very convenient to introduce a
Eq. (13), for this system. further simplifying assumption.
At the electrode surface, Eq. (24) re- When the number of electroactive
quires that species is conserved by the electrode reac-

tion, i νir = 0, it is satisfied that Gs0 = 0,
I (t) r
Gs0 (t) = νi (143) and the only solution to Eq. (145) is the
nF D
i trivial result S0 = S0b . This is the prin-
I (t) 2 r I (t) ciple of unchanging total concentration
Gs2 (t) = zi νi = (z1 + z2 ) established by Oldham and Feldberg [17].
nF D FD
i Equation (147) then reduces to
(144)
which are used as boundary conditions ∂S2 ∂ 2 S2 q3 I S0b 1 ∂S2
in chronopotentiometric techniques or to =D 2 + (148)
∂t ∂x F S22 ∂x
evaluate the system response in chronoam-
perometric techniques. In the second The initial and boundary conditions for
equality of Eq. (144), the common ion this equation are Eq. (31), S2 (x, t = 0) =
is considered to be inactive, ν3r = 0. The
S2b and S2 (x → ∞) = S2b .
parameter G1 = I (t)/FD is only a func-
In the case of a large voltage step,
tion of t.
f |%E| 1, Eq. (31) is approximated by
Combining Eq. (139) for k = 0 and
the condition of zero surface concentra-
Eq. (140), the partial differential equation
tion of the reactant, c1s (t) = 0 (or S2s =
for G0 is
−z2 z3 S0b ), and both boundary conditions
∂S0 ∂G0 ∂ 2 S0 for S2 turn out to be independent of time.
= −D =D 2 (145) This implies that S2 is a function of x and t
∂t ∂x ∂x
1/2
through variable ζ ≡ x/(2D1 t 1/2 ). More-
which is subject to the initial and boundary
over, since the electric current density is
conditions S0 (x, t = 0) = S0b and S0 (x →
obtained from Eqs. (142) and (144) as
∞, t) = S0b , in addition to Eq. (143). The

DBL thickness is considered to be infinite ∂S2 I (t)
due to the absence of stirring. The solution = (z3 − z1 ) (149)
∂x x=0 FD
of Eq. (145) can be obtained by following
the method explained in Sect. 2.1.5.3 as it is concluded that I (t) is proportional to
t −1/2 .
S0 (x, t) = S0 (ζ ) = S0b + (S0s − S0b )erfc(ζ )
Equation (148) can then be transformed
(146)
Combining Eq. (139) for k = 2 and into the ordinary differential equation for
Eq. (142), the partial differential equation S2 (ζ ) (see Appendix B)
for S2 is
d2 S2 dS2 (S b )2 dS2
∂S2 ∂ 2 S2 (x, t) 0= 2
+ 2ζ + γ 22 (150)
=D dζ dζ S2 dζ
∂t ∂x 2

∂ S0 (x, t) Parameter γ in Eq. (150) is related to the
− q3 DG1 (147)
∂x S2 (x, t) current density by
Fig. 14 Current density in a voltage

4, 1, −1
2.0 step of large width relative to that in
absence of migration, given by Cottrell
4, 1, −2 Reduction
Eq. (152), as a function of the bulk
1.5 concentration ratio product/reactant.
I/ICottrell
2, 1, −1 The charge numbers z1 , z2 , z3 are

shown close to the curves.
3, 2, −1
1.0
2, 3, −1
1, 2, −1 Oxidation
0.5 1, 4, −1
−3 −2 −1 0 1 2
Log10(cb2/cb1)
1/2
F (S2b )2 D smaller in the case of oxidations. It is
I =γ observed that, for a given stoichiometric
2q3 S0b t
number of the electron n = z2 − z1 in the
(z3 − z1 )(S2b )2 π 1/2 electrode reaction, increasing the charge
=γ ICottrell (151) numbers of either the inactive anion
2q3 S0b (S2b + z2 z3 S0b )
or the active cations reduces the effect
where of migration. Moreover, the effect of
1/2 migration is larger when |n| increases.
D The results of these numerical simula-
ICottrell = (z2 − z1 )F c1b (152)
πt tions can be understood by the following
intuitive arguments. Consider, for exam-
is the classical result obtained by Cottrell
ple, the oxidation of a cationic species
in the absence of migration. The value of γ
[I > 0, n > 0, z1 > 0, z2 = (n + z1 ) > 0,
must be obtained from the numerical so-
z3 < 0]. According to Ohm’s law, the elec-
lution of Eq. (150) by the shooting method
tric potential gradient in solution pushes
starting from S2 (0) = S2s = −z2 z3 S0b and
the cations away from the anode, so that
their gradients are smaller (in absolute
dS2 (z1 − z3 )(S2b )2
= γ (153) value) than in absence of migration. Since
dζ ζ =0 q3 S0b the voltage step makes the surface concen-
tration of the reduced species zero, c1s = 0,
and modifying it by iteration until
the flux density of this species at the elec-
S2 (∞) = S2b is satisfied.
trode surface is only due to diffusion,
Myland and Oldham [18] have compiled

the results for the case in which the dc1 I
=− (154)
product of the electrode reaction is absent dx x=0 (z1 − z2 )F D
in bulk solution, c2b = 0. Figure 14 shows
the current density as a function of the and a smaller concentration gradient im-
bulk product/reactant concentration ratio plies then smaller current density. Con-
for several values of the charge numbers trarily, in the case of reduction of cations,
in the case of cationic redox couples. The migration pushes the cations towards the
current is then larger than in the absence cathode, so that the concentration gra-
of migration in the case of reductions and dients are steeper. Since Eq. (154) still
applies, this implies larger current den- ∂ρe /∂t can be evaluated from the conti-
sities than in absence of migration. nuity equation, Eq. (11) with ri = 0, as
Consider, finally, the oxidation of an an- ∂ci
∂ρe
ion [I > 0, n > 0, z1 < 0, z2 = (n + z1 ) < =F zi
0, z3 > 0]. The electric potential gradient ∂t ∂t
i
in solution pushes the anions towards the ∂Ji ∂I
anode, so that the concentration gradients = −F zi =− (156)
∂x ∂x
are steeper and the current density is larger i
than in the absence of migration. The op- where I is the conduction electric cur-
posite effect is observed in the reductions rent density. Equation (156) represents the
of anions. conservation of electric charge and its com-
bination with Eq. (155) leads to
2.1.6
Deviations from Local Electroneutrality ∂IT ∂I ∂Id
= + =0 (157)
∂x ∂x ∂x
This last section is devoted to the space
which states that the total electric current
charge density and its variation with time
density IT that flows through an electro-
and position. The dielectric permittivity
chemical cell is independent of position,
of the solution is assumed to be con-
and is made up of two contributions, the
stant, which is a good approximation
conduction and the displacement current
for moderate concentrations and electric
densities [66–69, 79]
fields [75–77]. It is further assumed that
the relaxation of the dielectric (i.e. the re- ∂ 2φ
orientation of water molecules) is much IT = I + Id = F zi Ji − ε
∂t∂x
faster than the redistribution of the space i
(158)
charge density [78].
The current density IT is equal to that
due to the electrons flowing through the
2.1.6.1 Space Charge Density external circuit connected to the electrodes.
The Poisson equation, Eq. (8), identifies In the absence of concentration gradi-
the space charge density in an electrolyte ents, no space charge density can exist in
solution, ρe = F i zi ci , as the source of the solution under steady state conditions.
the electric field. Thus, whenever the elec- Moreover, if by any means some electrical
tric field changes with position, there is a charge density is generated at any position
space charge distribution over the system. within the system, the time required for
In general, ρe is a function of both this charge to disappear (in fact, to migrate
position and time. Its time derivative is to the system boundaries) is of the order
of nanoseconds. This time is the so-called
∂ρe ∂ 3φ
= −ε electrical relaxation time of the system, τe ,
∂t ∂t∂x 2 and can be obtained from Eqs. (8, 13, and

∂ ∂ ∂φ ∂Id 156) as
= −ε = (155)
∂x ∂t ∂x ∂x
∂ρe ∂I ∂ 2φ
=− =κ 2
where Id is the displacement current, ∂t ∂x ∂x
that is, the time derivative of the elec- ρe ρe
tric displacement D = εE. Alternatively, = −κ ≡− (159)
ε τe
so that τe ≡ ε/κ ∼ 10−9 s for a 100 mM equation, Eq. (8) tells that the field associ-
aqueous solution. ated to this charge density over a region
Concentration gradients, however, are of thickness LD is of the order of ρe LD /ε.
present in most situations of electrochem- The migrational flux density associated to
ical interest, and the time evolution of ρe this field is of the order of F Dcρe LD /εRT .
is then determined by the time variation The condition for the existence of an equi-
of the conditions imposed on the system, librium space charge density is Dc/LD ≈
which always involve times much larger F Dcρe LD /εRT , which yields the expres-
than τe . sion of the Debye length
Usually, space charge regions exist 1/2 1/2
εRT εRT
close to the system boundaries, such LD ≈ ≈ (161)
as the EDL at a planar electrode. In Fρe F 2c
the presence of concentration gradients, Space charge distributions are then ex-
Eq. (159) becomes pected to exist over regions of thickness
LD , which is of the order of 10−7 cm
∂ρe ∂ ∂φohm
= κ for a 100 mM aqueous solution. A prac-
∂t ∂x ∂x tical consequence of this comment is that

∂ ∂φ ∂ ∂φdif the behavior of electrochemical systems
= κ − κ (160) comprising microgeometries is affected by
∂x ∂x ∂x ∂x
space charge layers [80–82].
which is much smaller, in magnitude, Note, finally, that ionic motions associ-
than ρe /τe . Under equilibrium conditions, ated to space charge redistribution involve
I = 0, the two terms on the right-hand side distances of the order of LD , which sug-
cancel out. In the presence of conduction gests that τe can be interpreted as the time
current, these terms are very similar to τe ≈ L2D /D required for the ions to diffuse
each other and much larger in magnitude over LD [79, 83–85].
than ∂ρe /∂t. An order of magnitude analy-
sis of these two terms allows us to estimate 2.1.6.2 Deviations from Local
the characteristic length for the space Electroneutrality
charge distribution, LD . Because of the Planck suggested that the electric poten-
large electric fields existing in space charge tial calculated from the LEN assumption,
regions, the ions have a strong tendency φ (0) , could be used to check the validity
to migrate. If the ions were indeed migrat- of this assumption [27, 31]. When this ap-
ing, the space charge distribution would proximate electric potential is introduced
disappear, so that the necessary condition in the Poisson equation, the space charge
for an equilibrium space charge distribu- density obtained
tion is that this tendency to migrate has
∂ E(0) ∂ 2 φ (0)
to be compensated by a strong tendency ρr ≡ ε = −ε (162)
to diffuse in the opposite direction [68]. ∂x ∂x 2
The diffusional flux associated to concen- provides an idea of the deviations from
tration changes is of the order of Dc/LD , LEN, which would be required to create
where c is a typical ion concentration. The the distribution φ (0) (x, t). To make clear
magnitude of the charge density can be the difference between the true ρe in the
estimated as ρe ≈ F c, and the Poisson Poisson equation, Eq. (8), and ρr defined
by Eq. (162), the latter is often named and introducing the limiting current den-
residual space charge density [13, 86]. sity IL ≡ −S0b I /(δG0 ) defined in Eq. (101),
Consider, for instance, the steady state the equation for E becomes
electric field in a binary solution under
current flow conditions with species 2 L2D d2 E I 1 IL x
= z1 − z2 + −1
being inactive. From Sect. 2.1.4.2, E(0) = E dx 2 IL f Eδ I δ
S2 /f G1 = −z2 S0 /f G0 and the residual
IL
space charge density required to establish + (f ELD )2 (168)
2I
this approximate electric field is
b 2 where LD ≡ (εRT /F 2 S0b )1/2 is the Debye
εRT I 2 c2 length.
ρr = −
z2 F IL δ c2 (x) Analytical studies of Eq. (168) have
−2 used different changes of variables to
εRT IL x write the electric field in terms of either
=− + −1 (163)
z2 F δ 2 I δ Painlevé transcendents [68] or Jacobian el-
liptic functions [88]. Alternatively, asymp-
where Eqs. (97 and 101) have been used;
totic expansions have also been used [68,
note also that G1 = z1 G0 = I /F D1 . This
87, 89, 90]. The approximate solution
shows that significant deviations from the
methods have neglected different terms
LEN only occur at the limiting current
of Eq. (168). Thus, while Urtenov and
and close to the electrode surface x = 0,
Nikonenko [55, 91, 92] have considered
in which the ionic concentrations tend to
that the space charge density is quasi-
zero and ρr diverges [34, 42, 87].
uniform, dS1 /dx ≈ 0, Bass [93] assumed
The actual space charge density in this
that εE2 RT S0b and I /IL δ/LD to ex-
binary solution can be obtained from the
press the electric field as the sum of the
(exact) electric field E by using the same
electroneutral electric field, which can be
formalism and the Poisson equation
derived from Eq. (99), and a modified Gouy
dE ρe F S1 distribution corrected for the presence of
= = (164)
dx ε ε current. In any case, both analytical and
numerical [94, 95] solutions have shown
First, integrating Eq. (105) for k = 0
that in the range of underlimiting currents,
dS0 the deviations from LEN are confined to
−G0 = − f E S1 (165) the EDL and that the latter is not signifi-
dx
cantly perturbed by the electric current.
from x to δ, and using Eq. (164), Eq. (97) In conclusion, ρr in Eq. (163) is a
is modified to physically meaningful magnitude that is
ε E2 very similar to the actual ρe when LEN is
S0 (x) = S0b + (δ − x)G0 + (166) a good approximation. Contrarily, when
2RT
deviations from LEN occur, the actual
where the electric field in bulk solution electric potential surely differs from that
has been neglected. Then, substituting obtained by using the LEN, and then the
Eqs. (164 and 166) into Eq. (105) for k = 1 actual ρe is likely to show no resemblance
dS1 dS1 to ρr .
−G1 = − f E S2 = + z1 z2 f ES0 Although the equations presented in
dx dx
(167) this chapter cannot account for this fact,
overlimiting currents can be observed in Since φ (0) is the electric potential calcu-
practice [41]. The study of these currents lated from the transport equations at zero
requires not only to abandon the LEN as- order in λ (that is, making use of the LEN
sumption but also to take into account assumption), the residual space charge
the electroconvective motions that origi- density defined in Eq. (162) is identified as
nate due to the electrical force ρe E acting (1)
ρe . The use of higher-order terms in the
on the electrolyte solution in space charge Expansions (170 and 171) is only needed
regions [68, 90, 91, 95–97]. (1)
when ρe is of the order of F c [13, 27, 31,
67]. A different validity test for the LEN as-
(2) (1)
2.1.6.3 Validity of the LEN Assumption sumption, namely λρe ρe , has been
MacGillivray [98] was the first to provide a recently proposed by Feldberg (with a dif-
mathematical justification for the LEN as- ferent notation) [99].
sumption on the basis of the perturbation Note that it is possible to use either the
theory. The idea behind such a justification full Poisson equation or the terms of its
is that an order of magnitude analysis of asymptotic expansion, such as the LEN as-
(0)
the Poisson equation, Eq. (8), shows that sumption or Eq. (172), but ρe can never
be interpreted as the full ρe in the Poisson
ρe ε d2 φ ε RT /F equation. This would lead to the wrong
=− ≈
Fc F c dx 2 F c L2 conclusion that the LEN assumption im-
plies a constant electric field.
LD 2
= ≡λ1 (169)
L
Appendix A
where L is a typical length for variation of
the electric potential, which is considered The total potential drop across the DBL in
to be of the order of the size of the elec- Eq. (74) can be obtained as the sum of the
trochemical cell and hence much larger diffusion and ohmic potential drops. The
than the Debye length LD . This suggests former is given by Eq. (19) as
that the solution of the Nernst–Planck and
RT t1 t2
Poisson equations can be found by writing %φdif = − + % ln c12
F z1 z2
the concentrations and the electric poten- (173)
tial as an asymptotic expansion in terms of To integrate the ohmic potential gradient,
the ‘‘small parameter’’ λ ≡ (LD /L)2 , it is convenient to introduce first the molar
ci = ci
(0) (1)
+ λci
(2)
+ λ 2 ci + ··· (170) conductivity of the electrolyte as
κ F2
φ = φ (0) + λφ (1) + λ2 φ (2) + · · · (171) ;12 ≡ = −z1 z2 (ν2 D1 + ν1 D2 )
c12 RT
Thus, the zeroth-order approximation (in (174)
(0)
λ) of the Poisson equation is ρe = 0. where the last equality is known as the
In other words, the LEN assumption is a Nernst–Eintein relation [3]. Equation (17)
consequence of the Poisson equation when then implies
λ 1. Moreover, the first-order term of I I dx
the Poisson equation is dφohm = − dx = − (175)
κ ;12 c12
d2 φ (0) Eliminating the position variable from
λρe(1) = −ε (172) Eqs. (65) and (67) as
dx 2

z1 ν1 F ∂c1 ∂c1 ∂ζ ∂c1
−dx = (D12,u dc12,u + D12 dc12 ) = +
t2 I ∂t x ∂ζ t ∂t x ∂t ζ
(176)
and using Eq. (63) to deduce that d lnc12,u= ζ ∂c1 ∂c1
=− + (180)
ν12 d ln c12 , Eq. (175) is transformed to 2t ∂ζ t ∂t ζ
z1 ν1 F 1 and the diffusion equation becomes
dφohm = (D12,u dc12,u
t2 ;12 c12
∂ 2 c1 ∂c1 ∂c1
RT ν12 t1 dc12 + 2ζ = 4t (181)
+ D12 dc12 ) = ∂ζ 2 ∂ζ ∂t
F z1 ν1 c12
When the boundary conditions imposed by
D12,u dc12,u RT ν12 t1 the experiment allow for the separation of
− = d ln c12
z2 D2 c12 F z1 ν1 variables ζ and t (linear sweep voltamme-
try, for instance, does not), a solution of the
ν12 D12,u c12,u
− d (177) form c1 (ζ, t) = c1b + Z(ζ )T (t) transforms
z2 (ν12 − 1)D2 c12
Eq. (181) into
Straightforward integration over the DBL
and addition of %φdif yields Eq. (74). 1 d2 Z dZ 4t dT
+ 2ζ = = λ (182)
Z dζ 2 dζ T dt
Appendix B where λ is a constant.
In the chronoamperometric case, the
The perturbation of the concentration field boundary condition, Eq. (129), does not
introduced by the electrode propagates at introduce any dependence on t, and λ = 0
a finite rate determined by the diffusion in Eq. (182). Then, the solution of this
coefficients and only affects a region close equation is
to the electrode whose thickness is of the
1/2 Z(ζ ) = c1 (ζ, t) − c1b = Aerfc(ζ ) + B
order of the diffusion length 2D1 t 1/2 .
(183)
Boltzmann showed that the diffusion
where erfc(ζ ) is the complementary error
equation, Eq. (126), can be reduced to
function. The boundary condition c1 (x →
an ordinary differential equation under
∞, t) = c1b imposes B = 0, and Eq. (132)
certain conditions if expressed in terms of
1/2 is obtained.
the variable ζ ≡ x/(2D1 t 1/2 ) [62, 100].
In the chronopotentiometric case, the
To obtain the diffusion equation in terms
boundary condition, Eq. (127), becomes
of the variables (ζ , t), the elementary con-
centration variation is first expressed as ∂c1 dZ 2I0
= T (t) =− t 1/2
∂ζ ζ =0 dζ ζ =0 z1 F D1
1/2
∂c1 ∂c1
dc1 = dζ + dt (178) (184)
∂ζ t ∂t ζ which makes λ = 2 in Eq. (182). The so-
The partial derivatives of c1 with respect to lution of the ordinary differential equation
x and t (at constant x) are then given by for the function Z(ζ ) are the first integrals
2 of the complementary error function [61]
∂ c1 ∂ζ ∂ ∂ζ ∂c1
=
Z(ζ ) = A[π −1/2 e−ζ − ζ erfc(ζ )] (185)
2
2
∂x t ∂x t ∂ζ ∂x t ∂ζ t t
2
1 ∂ c1 where the boundary condition c1 (x →
= (179)
4D1 t ∂ζ 2 t ∞, t) = c1b has been taken into account.
Finally, the use of Eq. (184) to determine x position (m)

A yields Eq. (136). zi charge number, defined with sign
List of important symbols Greek symbols

α degree of dissociation, Eq. (63)
Latin symbols δ DBL thickness (m)
C class concentration (mol m−3 ), % difference defined as bulk minus
Eq. (43) surface value
ci molar concentration (mol m−3 ) ε dielectric permittivity (C V−1 m−1 ),
D diffusion coefficient (m2 s−1 ) Eq. (8)
Eeq equilibrium φ electric potential (V)
electrode potential (V), Eq. (31) 3 ≡ −G1 /z1 z2 G0
,E
E electric field (V m−1 ) auxiliary variable, Eq. (100)
e ≡ f Eδ dimensionless ηc concentration
electric field, Eq. (118) overpotential (V), Eq. (32)
F Faraday constant (C mol−1 ) κ conductivity (=−1 m−1 ), Eq. (14)
f ≡ F /RT (V−1 ) κeff ≡ −I /(dφ/dx) effective

Gk ≡ i zik Ji /Di (mol m−4 ), Eq. (96) conductivity (=−1 m−1 ), Eq. (42)
I electric current ;12 molar conductivity
density (A m−2 ), Eqs. (12, 158) (m2 =−1 mol−1 ), Eq. (174)
Ji ionic flux density (mol m−2 s−1 ), µ̃i electrochemical potential (J mol−1 )
Eq. (5) ν12 ≡ ν1 + ν2 , Eq. (63)
K equilibrium constant νi stoichiometric
−3
[(mol m ) νi ], Eq. (63) number, Eqs. (62, 83)
LD Debye length (m), Eq. (161) νir stoichiometric number
N number of ionic species, Eq. (95) in electrode reaction, Eq. (23)
ρ charge density (C m−3 ), Eq. (8)
n = i zi νir stoichiometric
number of the electron, Eq. (23) τe electrical relaxation
ni number of moles of species time (s), Eq. (159)
i (mol) ξ reaction coordinate (mol), Eq. (24)
qk auxiliary variable, Eq. (110) ζ ≡ x/(2D 1/2 t 1/2 )
R gas constant (J mol−1 K−1 ) Boltzmann variable, Eq. (132)
r chemical reaction
rate (mol m−3 s−1 ), Eqs. (11, 84) Subscripts and superscripts

Sk ≡ i zik ci (mol m−3 ), Eq. (95) d displacement, Eq. (155)
s Laplace variable (s−1 ), Eq. (130) dif diffusion, Eqs. (19, 61)
T temperature (K) e electrical, Eq. (8)
Ti integral transport number, Eq. (25) i ionic species
t time (s) ij dissociated electrolyte, Eq. (63)
ti migrational transport number, ij, T total electrolyte, Eq. (63)
Eq. (15) ij, u undissociated electrolyte, Eq. (63)
ui ionic mobility (m2 V−1 s−1 ), j constituent, Eq. (85)
Eq. (4) k power of charge number, Eq. (95)
vi ionic velocity (m s−1 ), Eq. (4) L limiting, Eq. (47)
Ld, i limiting diffusion B. E. Conway et al.), Plenum Press, New

(current density), Eq. (78) York, 1983, Vol. 6, Chap. 4, pp. 239–315.
17. K. B. Oldham, S. W. Feldberg, J. Phys.
mig migration, Eq. (60)
Chem. B 1999, 103, 1699–1704.
ohm ohmic, Eq. (18) 18. J. C. Myland, K. B. Oldham, Electrochem.
T total, Eq. (158) Commun. 1999, 1, 467–471.
(i) perturbation order, Eq. (170) 19. N. Ibl, Pure Appl. Chem. 1981, 53,
b bulk solution (x = δ) 1827–1840.
20. E. L. Cussler, Diffusion, Mass Transfer in
◦ standard state, Eq. (31) Fluid Systems, 2nd ed., Cambridge Uni-
r electrode reaction, Eq. (23) versity Press, Cambridge, 1997, Chap. 13,
s electrode surface (x = 0) pp. 331–355.
21. J. N. Agar, Discuss. Faraday Soc. 1947, 1,
26–37.
References
22. L. L. Bircumshaw, A. C. Riddiford, Q. Rev.
Chem. Soc. London 1952, 6, 157–185.
1. R. Haase, Thermodynamics of Irreversible 23. J. A. Manzanares, K. Kontturi in Surface
Processes, Addison-Wesley, New York, 1969. Chemistry and Electrochemistry of Membranes
2. K. Kontturi, Acta Polytech. Scand. 1983, 152, (Ed.: T. S. Sørensen), Marcel Dekker, New
1–40. York, 1999, Chap. 11, pp. 399–435.
3. R. A. Robinson, R. H. Stokes, Electrolyte So- 24. K. Hattenbach, Terminology for Electrodialy-
lutions, Butterworths, London, 1955. sis, European Society of Membrane Science
4. N. Ibl in Comprehensive Treatise of Electro- and Technology, Geesthacht, Germany,
chemistry (Ed.: E. Yeager, J. O’M. Bockris, 1988.
B. E. Conway et al.), Plenum Press, New 25. P. Delahay, Double Layer and Electrode Ki-
York, 1983, Vol. 6, Chap. 1, pp. 1–63. netics, Interscience Publishers, New York,
5. J. S. Newman, Electrochemical Systems, 1965.
Prentice-Hall, Englewood Cliffs, N.J., 1991. 26. K. J. Vetter, Electrochemical Kinetics, Aca-
6. A. V. Sokirko, J. Electroanal. Chem. 1994, demic Press, New York, 1967, Chap. 2,
364, 51–62. pp. 157–184.
7. E. J. Calvo in Comprehensive Chemical Kinet- 27. M. Planck, Ann. Phys. 1890, 39, 161–186.
ics (Eds.: C. H. Bamford, R. G. Compton), 28. K. R. Johnson, Ann. Phys. (Leipzig) 1904,
Elsevier, Amsterdam, 1986, Vol. 26, 14, 995–1003.
Chap. 1, pp. 1–78. 29. H. Pleijel, Z. Phys. Chem. 1910, 72, 1–37.
8. R. Schlögl, Stofftransport durch Membra- 30. D. A. MacInnes, The Principles of Elec-
nen, Steinkopff-Verlag, Darmstadt, 1964, trochemistry, Reinhold, New York, 1939,
Chap. 1, pp. 6–15, and Sect. 6.5, pp. 79–93. Chap. 13, pp. 220–245.
9. E. A. Guggenheim, Thermodynamics, North 31. M. Planck, Ann. Phys. 1890, 40, 561–576.
Holland, Amsterdam, 1967, Chap. 8, pp. 32. R. Schlögl, Ber. Bunsen-Ges. Phys. Chem.
298–302. 1978, 82, 225–232.
10. D. Kondepundi, I. Prigogine, Modern Ther- 33. K. B. Oldham, Z. G. Zoski in Comprehensive
modynamics, John Wiley & Sons, Chich- Chemical Kinetics (Eds.: C. H. Bamford,
ester, 1998. R. G. Compton), Elsevier, Amsterdam,
11. R. P. Buck, J. Membr. Sci. 1984, 17, 1–62. 1986, Vol. 26, Chap. 2, pp. 79–143.
12. J. Pellicer, S. Mafé, V. M. Aguilella, Ber. 34. V. G. Levich, Physicochemical Hydrodynam-
Bunsen-Ges. Phys. Chem. 1986, 90, 867–872. ics, Prentice-Hall, Englewood Cliffs, N.J.,
13. S. Mafé, J. Pellicer, V. M. Aguilella, J. Phys. 1962, Chap. 6, pp. 231–371.
Chem. 1986, 90, 6045–6059. 35. Yu. I. Kharkats, Sov. Electrochem. 1988, 25,
14. D. G. Miller, Chem. Rev. 1960, 60, 15–37. 503–506.
15. D. G. Miller, J. Phys. Chem. 1967, 71, 36. A. V. Sokirko, Yu. I. Kharkats, Sov. Elec-
3588–3592. trochem. 1989, 25, 287–292.
16. N. Ibl in Comprehensive Treatise of Electro- 37. Yu. I. Kharkats, A. V. Sokirko, J. Electroanal.
chemistry (Eds.: E. Yeager, J. O’M. Bockris, Chem. 1991, 303, 17–25.
38. M. A. Vorotyntsev, Sov. Electrochem. 1988, 63. S. L. Marchiano, A. J. Arvia in Compre-

24, 1150–1154. hensive Treatise of Electrochemistry (Eds.:
39. Y. Marcus, Ion Properties, Marcel Dekker, E. Yeager, J. O’M. Bockris, B. E. Conway
New York, 1997. et al.), Plenum Press, New York, 1983,
40. J. G. Stark, H. G. Wallace, Chemistry Data Vol. 6, Chap. 2, pp. 65–132.
Book, John Murray, London, 1982. 64. D. Britz, Digital Simulation in Electrochem-
41. V. K. Indusekhar, P. Meares in Physico- istry, Springer-Verlag, Berlin, Germany,
chemical Hydrodynamics (Eds.: D. B. Spald- 1988.
ing), Building & Sons, Guilford, 1977, 65. H. Cohen, J. W. Cooley, Biophys. J. 1965, 5,
pp. 1301–1313, Vol. II. 145–163.
42. K. Aoki, J. Electroanal. Chem. 2000, 488, 66. J. R. Sandifer, R. P. Buck, J. Electroanal.
25–31. Chem. 1975, 79, 384–391.
43. K. B. Oldham, T. J. Cardwell, J. H. Santos 67. J. R. Sandifer in Ion-Transfer Kinetics (Ed.:
et al., J. Electroanal. Chem. 1997, 430, 25–37, J. R. Sandifer), VCH Publishers, New York,
39–46. 1995, Chap. 6, pp. 115–138.
44. M. F. Bento, L. Thouin, C. Amatore et al., J. 68. I. Rubinstein, Electro-diffusion of Ions, SIAM,
Electroanal. Chem. 1998, 443, 137–148. Philadelphia, Pa., 1990, Chap. 1, pp. 1–21.
45. M. F. Bento, L. Thouin, C. Amatore, J. Elec- 69. W. D. Murphy, J. A. Manzanares, S. Mafé
troanal. Chem. 1998, 446, 91–105. et al., J. Phys. Chem. 1992, 96, 9983–9991.
46. A. M. Bond, S. W. Feldberg, J. Phys. Chem. 70. G. R. Engelhardt, A. D. Davydov, Russ. J.
B 1998, 102, 9966–9974. Electrochem. 1994, 30, 953, 954.
47. C. Amatore, L. Thouin, M. F. Bento, J. Elec- 71. A. A. Moya, J. Castilla, J. Horno,
troanal. Chem. 1999, 463, 45–52. Ber. Bunsen-Ges. Phys. Chem. 1995, 99,
48. K. B. Oldham, J. Phys. Chem. B 2000, 88, 1037–1042.
4703–4706. 72. A. A. Moya, J. Horno, Electrochim. Acta
49. A. D. MacGillivray, D. Hare, J. Theor. Biol. 1996, 99, 1037–1042.
1969, 25, 113–126. 73. A. J. Bard, L. R. Faulkner, Electrochemical
50. A. V. Sokirko, J. A. Manzanares, J. Pellicer, Methods, John Wiley & Sons, New York,
J. Colloid Interface Sci. 1994, 168, 32–39. 1980.
51. H. A. Kramers, Physica 1940, 7, 284–330. 74. I. C. Bassignana, H. Reiss, J. Phys. Chem.
52. W. E. Morf, Anal. Chem. 1977, 49, 810–813. 1983, 87, 136–149.
53. U. Behn, Ann. Phys. Chem. N.F. 1897, 62, 75. D. C. Grahame, J. Chem. Phys. 1950, 18,
54–67. 903–909.
54. M. Kh. Urtenov, Methods of Solution of the 76. F. Booth, J. Chem. Phys. 1951, 19, 391–394,
Nernst-Planck-Poisson Equation System [in 1327–1328, 1615.
russian], Kuban State University, Krasnodar, 77. I. C. Yeh, M. L. Berkowitz, J. Chem. Phys.
Russia, 1998. 1999, 110, 7935–7942.
55. V. V. Nikonenko, M. K. Urtenov, Russ. J. 78. J. M. G. Barthel, H. Krienke, W. Kunz, Phys-
Electrochem. 1996, 32, 187–194. ical Chemistry of Electrolyte Solutions. Modern
56. A. Guirao, S. Mafé, J. A. Manzanares et al., Aspects, (Ed.: Deutsche Bunsen-Gesellschaft
J. Phys. Chem. 1995, 99, 3387–3393. für Physikalische Chemie e.V.), Steinkopff,
57. P. Ramı́rez, S. Mafé, A. Tanioka et al., Poly- Darmstadt, 1998, Sect. 2.6.
mer 1997, 38, 4931–4934. 79. S. Mafé, J. A. Manzanares, J. Pellicer, J.
58. J. A. Manzanares, G. Vergara, S. Mafé et al., Electroanal. Chem. 1988, 241, 57–77.
J. Phys. Chem. B 1998, 102, 1301–1307. 80. J. D. Norton, H. S. White, S. W. Feldberg, J.
59. R. Schlögl, Z. Phys. Chem. N. F. 1954, 1, Phys. Chem. 1990, 94, 6772–6780.
305–339. 81. C. P. Smith, H. S. White, Anal. Chem. 1993,
60. J. F. Brady, J. C. R. Turner, J. Chem. Soc., 65, 3343–3353.
Faraday Trans. 1 1978, 74, 2839–2849. 82. D. R. Baker, M. W. Verbrugge, Proc. R. Soc.
61. M. Abramowitz, I. A. Stegun, Handbook of London, A 1998, 454, 1805–1829.
Mathematical Functions, Dover Publica- 83. R. P. Buck, J. Electroanal. Chem. 1969, 23,
tions, New York, 1965. 219–240.
62. J. Crank, The Mathematics of Diffusion, 2nd 84. A. D. MacGillivray, J. Chem. Phys. 1970, 52,
ed., Oxford University Press, Oxford, 1977. 3126–3132.
85. J. L. Jackson, J. Phys. Chem. 1974, 78, 93. L. Bass, Trans. Faraday Soc. 1964, 60,
2060–2064. 1656–1663.
86. V. M. Aguilella, S. Mafé, J. Pellicer, Elec- 94. I. Rubinstein, L. Shtilman, J. Chem. Soc.,
trochim. Acta 1987, 32, 483–488. Faraday Trans. 2 1979, 75, 231–246.
87. W. H. Smyrl, J. Newman, Trans. Faraday 95. J. A. Manzanares, W. D. Murphy, S. Mafé
Soc. 1967, 63, 207–216. et al., J. Phys. Chem. 1993, 97, 8524–8530.
88. R. H. Tredgold, Biochim. Biophys. Acta 96. I. Rubinstein, React. Polym. 1984, 2,
1972, 274, 563–574. 117–131.
89. J. Newman, Trans. Faraday Soc. 1965, 61, 97. V. V. Nikonenko, V. I. Zabolotskii, N. P.
2229–2237. Gnusin, Sov. Electrochem. 1989, 25,
90. A. V. Listovnichii, Russ. J. Electrochem. 1989, 262–266.
25, 1479–1482. 98. A. D. MacGillivray, J. Chem. Phys. 1968, 48,
91. M. Kh. Urtenov, V. V. Nikonenko, Russ. J. 2903–2907.
Electrochem. 1993, 29, 314–324. 99. S. W. Feldberg, Electrochem. Comm. 2000, 2,
92. V. V. Nikonenko, M. Kh. Urtenov, Russ. J. 453–456.
Electrochem. 1996, 32, 195–198. 100. L. Boltzmann, Ann. Phys. 1894, 53, 959–964.
2.2 and the boundary element method (BEM))

The Digital Simulation of Voltammetry are discussed. Each has been used in vary-
under Stagnant and Hydrodynamic ing degrees for the solution of the partial
Conditions differential equations relating to the mass
transport of material during electrolysis. At
Kerry A. Gooch, Fulian L. Qiu, and Adrian
least one formulation strategy is outlined
C. Fisher
University of Bath, Bath, United Kingdom in detail for each method, to give work-
ers, who are new to the field, a jump-off
2.2.1 point to the extensive literature available.
Introduction Although not included within this re-
view, sample codes for the FDM can be
2.2.1.1 Digital Simulations found in a variety of publications, includ-
The advent of the personal computer has ing, Feldberg [1], Bard and Faulkner [2],
led to the rapid development of digital Britz [3], and Gosser [4]. Sample codes for
simulations in the field of electrochem- the FEM and BEM are available on our
istry. The motivation for these advances website [5]. In each case, the general for-
has been driven by many factors includ- mulation stages are followed by a review
ing: the desire to gain quantitative insights of the digital simulation of mass trans-
from electroanalytical measurements, ex- port using the specific technique. Special
tract kinetic and mechanistic information emphasis is placed on the literature asso-
regarding the pathways of electrolysis re- ciated with hydrodynamic techniques (e.g.
actions, and assist in the design of new the rotating disc, channel flow cell, or wall
quantitative electrochemical methods. jet electrode), since there are already excel-
Quantification of the related phenom- lent overviews of diffusional applications
ena proceeds by the definition of a model
using FDM (e.g. Ref. [3]).
that describes the physical processes (e.g.
mass transport, chemical reactivity, etc.)
2.2.1.2 Simulation Methods. The
in terms of a set of mathematical ex-
Fundamentals
pressions. Traditionally, workers have at-
Irrespective of the simulation technique
tempted to solve these relationships di-
rectly via standard mathematical methods; chosen, the key stages in the development
however, in many cases the complex- of a numerical model and its validation are
ity of the problem does not permit this shared and these are indicated in Fig. 1.
analytical-solution approach. Digital simu- The procedure begins with a proposed,
lation breaks down (discretizes) the prob- realistic, or preliminary working physi-
lem into a series of steps that can be cal model. A mathematical description
solved sequentially by the composition of is formulated, which in the case of
a suitable computer program. This dis- transport-related problems proceeds via
cretization process gives rise to a variety conservation of momentum, mass, charge,
of different digital strategies with which and energy.
electrochemical or related problems can For an electrolysis reaction, in which an
be solved. electrode is sited in an electrolyte solution
In this overview, three digital simulation containing an electroactive species, the
approaches (the finite difference method species distribution (in one spatial dimen-
(FDM), the finite element method (FEM), sion) may be described by an expression of
2.2 The Digital Simulation of Voltammetry under Stagnant and Hydrodynamic Conditions 123
O
R
Reactant
y
Physical model transport
Product
O+ e R transport
x Electrode
∂c0 ∂2c ∂2c

Mathematical description = D 20 + D 20
∂t ∂x ∂y
Numerical discretisation Grid points

x Electrode
c0
Computational predictions/ Computational prediction

model validation Analytical solution
Distance
Fig. 1 Simulation procedure.
the form To give a more general solution to

the expressions, it is common to cast
∂c the equation variables in a dimensionless
= D∇ 2 c − Vx ∇c ± k (1)
∂t form. For example, the concentration, c, is
cast dimensionless using
In Eq. (1), c is the reagent concentration, t
the time, Vx the velocity, and k any kinetic c
c∗ = (2)
terms (e.g. first-, second-order reactions, cBULK
etc.). For more complex geometries, such
as disc or band electrodes, the form of the where cBULK is the concentration of a
transport equation becomes more complex specific starting material. Similarly, the
but still retains terms of the same form as distance parameter x can be cast into a
Eq. (1). dimensionless (X) form
Next, digital solution of the model
is required and this is achieved by x
X= (3)
discretization of Eq. (1). This process δ
results in a set of algebraic equations that
can then be employed to solve for the with δ representing a fixed distance. The
unknown variables. time variable, t, can be normalized by the
introduction of a fixed observation time, T , appears as the second derivative. The so-
t lution procedure requires discretization
τ= (4) of the domain into a two-dimensional
T
grid of points, at which the concentration
Additionally, dimensionless current, pois computed.
tential, reaction kinetic, and scan rate Implementation of the FEM requires
parameters may also be employed. The integration of the differential equations
choice of δ, T , and so forth depends upon once (by parts), to produce what is
the specific problem tackled and a good often referred to as the weak formulation
overview may be found in the text by (Sect. 2.2.3.2). In contrast to the FDM, the
Britz [3]. integral expression contains the variable
Solution of Eq. (1) in the dimensionless and its first derivative, thus reducing the
form leads to the prediction of the temporal order of the differential by one with respect
and spatial variation of c. These may be to the variable. Domain discretization uses
easily returned to dimensional variables a mesh of interconnecting elements and
by reverse transformation. the variable approximated using a chosen
Once the predictions have been ob- interpolation function.
tained, they need to be validated and this The BEM involves integration of the dif-
is often the most challenging aspect of ferential equation twice; this yields an ex-
the simulation procedure. Workers often pression that includes the variable and its
configure codes so that they initially tackle derivative. Discretization employs a mesh
documented problems with analytical so- of elements with the variable approximated
lutions, providing a quantitative check on using an interpolation function. In the
the simulations before extension to more BEM, the formulation permits (in some
complex problems. cases) the variable distribution within the
domain to be related to the values only at
2.2.1.3 The Methods. An Overview the boundary. The integration procedure
Before proceeding to the detailed develop- therefore effectively reduces the order of
ment and numerical procedures for each the digital simulation from a two- to a one-
method, it is of interest to note the re- dimensional problem. This reduces the
lationship between the FDM, FEM, and complexity of the grid and number of ele-
BEM formulations. Although fundamen- ments required in comparison to the FDM
tally different in their implementation, and the FEM.
they are linked mathematically (Fig. 2). The choice of the optimal method de-
By way of example, consider the numer- pends on the problem of interest. The
ical procedure for the solution of Fick’s FDM has found widespread application in
second law of diffusion in two dimensions the electrochemical field due to the relative
∂c ∂ 2c ∂ 2c ease of application, the routine modifi-
=D 2 +D 2 (5) cation for a variety of transport-related
∂t ∂x ∂y
problems, and the simple modification
The FDM uses a truncated Taylor series for chemical reactivity issues. Apart from
(Sect. 2.2.2.2) to approximate the differ- its popularity in the engineering fields,
ential equation. This approach uses the the FEM (Workers have used a variety of
original formulation of Fick’s law and approaches based around the method of
the variable of interest (concentration) weighted residuals of which further detail
Finite
difference
Original
formulation
Electrode
Series of grid
points
Integration
Finite element
Weak formulation
Electrode
Mesh of interconnecting
triangular elements
Integration
Boundary
element
Inverse
formulation
Electrode
Mesh of elements around
boundary of domain
Fig. 2 The relationship between the FDM, FEM, and BEM formulations.
can be found in Sect. 2.2.3.2.) has, more BEM. This in part has been due to the
recently, also gained popularity in electro- perceived difficulty [6] of implementation
chemical applications. Merits of the FEM with some relatively involved mathemat-
include a considerably reduced number ical steps required before numerical so-
of discretization steps within the domain lution. In addition, the application of the
and the ability to address complex or irreg- method to problems of complex chemi-
ular geometries. cal reactivity is challenging. However, the
Perhaps the most neglected of the three method does possess some benefits over
techniques within electrochemistry is the both the FEM and FDM. In particular,
the reduction in dimensionality brought FDM along with the application to a simple
about by the formulation procedure electrochemical simulation.
can give rise to significant time-savings
and reduced grid-generation requirements 2.2.2.2The Explicit Finite Difference (EFD)
when complex geometries and/or three- Method
dimensional analysis of transport prob- Consider Fick’s second law of diffusion (in
lems are required. one dimension):
2.2.2 ∂c ∂ 2c
=D 2 (6)
Finite Difference Methods ∂t ∂x
2.2.2.1 Introduction This expression may be cast into a finite

The basis of the FDM was described as difference form via the following. The
early as 1911 [7] and underwent significant simulation domain (x = 0 to x = δ) is
development throughout the first half divided into a series of equally spaced
of the 1900s. Courant and coworkers [8] (x) points as indicated in Fig. 3. The
t
concentration close (e.g. ci−1 t ) to an
or ci+1
reported a detailed implementation and
t
individual point (ci ) is approximated using
later Emmons [9] applied the method to
a range of problems. In 1948, Randles the Taylor expansion
simulated an electrochemical linear-sweep
(x)2
experiment [10] (with all of the calculations f (x ± x) = f (x) ± xf (x) +
performed by hand). However, it was 2!
not until the 1960s with the work of (x)3
× f (x) ± f (x)
Feldberg [1] that the digital simulation of 3!
electrode processes became a focused area (7)
of electrochemical research (Feldberg used where f , f , and f are the first, second,
a flux balance of material into and out and third derivatives of f respectively.
of a box to derive the discretized form If ∂x is small, then the terms from
of the difference equations required for (x)3 /3! may reasonably be neglected and
expressions for the concentration at (ci−1 t
solution.). Many formulations using the
t
and ci+1 ) predicted
FDM have since been reported and a
number of the key methods have been
highlighted below. The review begins ∂c (x)2 ∂ 2c
t
ci+1 = cit + x +
with a detailed overview of the key steps ∂x 2! ∂x 2
required for the formulation of the explicit (8)
cti−1 cti cti+1

Bulk
Electrode
solution
∆x
X=0 X=d
Fig. 3 The discretization of a domain into a series of sample points,
where the cit represents the concentration at the point i at a specific time t.

∂c (x)2 ∂ 2c an individual point (i) via the values at the
t
ci−1 = cit − x + previous time-step; hence this approach
∂x 2! ∂x 2
(9) is termed an explicit formulation. Note
summing Eqs. (8 and 9) yields that Eqs. (10 to 12) have been written in
terms of the concentrations at the time
∂ 2c 1 interval t, this is referred to as a backwards
= (ct − 2cit + ci−1
t
) (10) difference approximation. Application of
∂x 2 (x)2 i+1
Eq. (12) permits the simulation of a variety
Similarly the variation of concentration at of voltammetric measurements under
a point i during a small increment in time diffusional transport control.
(t) can be approximated as Consider a potential step experiment
performed under conditions in which
∂c ct+t − cit the analytical Cottrell equation is valid.
= i (11)
∂t t Assuming that initially the solution con-
tains a single electroactive reagent of
cit+t and cit are the concentrations at the
concentration cBULK . At t = 0 the applied
point i at the new and old time intervals,
potential is insufficient to drive the elec-
respectively. Substituting Eqs. (10 and 11)
trolysis reaction, following the potential
into Eq. (6) followed by rearrangement
step at t > 0, the reagent is destroyed
gives the finite difference form of Fick’s
at the electrode surface at a transport
second law
limited rate. Digital simulation proceeds
cit+t = (λci−1
t
+ (1 − 2λ)cit + λci+1
t
) by dividing the domain δ into N I finite
(12) difference points (Fig. 4a) and applying
where boundary conditions
Dt
λ= (13) t=0 cit = cBULK i = 0, N I (14)
(x)2
t>0 cit =0 i=0 (15)
The above expression provides a direct
method of obtaining the concentrations at cit = cBULK i = NI (16)
i=1 i = NI
Electrode Bulk
solution
(a) X=0 X=d
cb
c
0
(b) x
Fig. 4(a) Grid and (b) concentration distribution calculated using
an EFD simulation.
The boundary condition (14) permits the and is known as the von Neumann stabil-
calculation of new concentrations at all ity criterion [17], and it has been converted
points (i) for the time increment t + t, into a complete proof by John [18]. The
subject to the conditions specified in stability criterion can prove problematic
Eqs. (15 and 16). The current may be when a mesh contains small increments
predicted by taking the concentration of x. For a fixed diffusion coefficient,
gradient at x = 0 this requires a small value of t, thus
considerably increasing the real simula-
∂c
i = nFAD (17) tion time.
∂x x=0 Extension of the EFD method to tackle
The choice of δ should ensure that convection and chemical reactions pro-
concentration changes from their bulk vides no conceptual problems. The con-
values occur within the simulation region. vective term from Eq. (1) can be cast into
The simplest value for x would be the EFD form
given by δ/N I , however, since the greatest
∂c cit − ci−1
t
change in concentration is usually found = Vx (18)
at or near the electrode surface, a grid ∂t x
that concentrates mesh points close to the
and summed with Eq. (10) to yield the EFD
surface would be desirable (Fig. 4a).
expression for the convective–diffusion
For one-dimensional simulations, Joslin
equation. Similarly, kinetic complications
and Pletcher [11] and Feldberg [12] have
proposed alternative strategies based on arising from homogeneous chemical re-
a smooth spacing function and expo- actions can be introduced (e.g. for a
nentially expanding grid, respectively. Re- first-order chemical reaction)
cently Bieniasz [13] has reported a more ∂c
sophisticated adaptation. The approach ex- = −kcit (19)
∂t
ploits a set of steps that recalculate the
mesh point positions following a concen- The ease of introduction of chemical re-
tration distribution calculation. Irrespec- activity into digital simulation methods
tive of the choice of function, the overall provides a key advantage over analyti-
goal is the accurate calculation of the con- cal procedures that are difficult to solve
centration distribution – too few sample for complex chemical problems. How-
points yield an inaccurate (nonconverged) ever, introduction of chemical reactivity
approximation. needs to be tackled with care (especially
Selection of t in electrochemical sim- when the rate constants are large). The
ulations has generally remained a fixed rapid chemical depletion of one or more
value, although, a few groups have re- reagents within the simulation can result
ported a range of time-dependent varia- in finer mesh requirements or shorter
tions [14–16]. However, with all of these time increments to retain stability and
approaches, one must be aware of an accuracy within the simulation. Readers
operating restriction of the EFD, which are directed to a review by Speiser [19]
arises from the parameter λ Eq. (13). If in which these issues are highlighted in
its value exceeds 1/2, then the simulation some detail.
will not give a stable solution. This condi- The EFD method can be routinely app-
tion arises from the formulation procedure lied to multidimensional spatial problems,
however the restriction imposed by λ has can be written in the tridiagonal form
led workers to look for alternative formu-
2(λ + 1) t+t
lation strategies to overcome these diffi- c0t+t − c1 + c2t+t
culties. In the next sections, some popular λ
alternative formulations are examined. 2(λ − 1) t
= −c0t + c1 − c1t (23)
λ
2.2.2.3 The Crank & Nicolson (CN) t+t 2(λ + 1) t+t t+t
Method ci−1 − ci + ci+1
λ
The CN method [20, 21] was used by 2(λ − 1) t
Randles [10] in his work carried out before = −ci−1
t
+ ci − ci+1
t
(24)
λ
the advent of modern computers. A central
t+t 2(λ + 1) t+t t+t
I −1 − cN I + cN
difference scheme is applied in which the cN I +1
diffusional term (10) is centered at the λ
midpoint of the time-step 2(λ − 1) t
= −cN t
I −1 + cN I − cN
t
I +1
λ
∂ 2 c(t+t/2) 1 t (25)
= c − 2cit + ci+1
t
Application of appropriate boundary con-
δx 2 2(x)2 i−1
t+t
ditions yields a set of simultaneous equa-
t+t
+ ci−1 − 2cit+t + ci+1 tions that can be solved using a variety of
(20) methods including Gaussian elimination,
The concentration at (t + t/2) is evalu- of which the Thomas Algorithm [22] is a
ated via linear interpolation using simplified version. Although formulation
and implementation of the CN scheme is
(t+t/2) (cit + cit+t ) considerably more involved than the EFD,
ci = (21) it does have advantages. In principle, the
2
stability of the method is not restricted
and leads to an equivalent expression to by the value of λ, therefore permitting
that of Eq. (12) for the explicit method the use of larger time increments within
simulations. It has been noted that the CN
t+t 2(λ + 1) t+t method does possess some limitations that
ci−1 − ci t+t
+ ci+1
λ can be overcome by modification and the
2(λ − 1) t reader is directed to the text by Britz [3] for
= −ci−1
t
+ ci − ci+1
t
(22) further details. The CN technique has been
λ
applied to simulate one-dimensional diffu-
where λ is given by Eq. (13). Unlike sion to planar and spherical electrodes and
Eq. (11), the new concentrations cannot diffusion in redox polymers [23].
be directly obtained from Eq. (20) and
hence the method is generally referred 2.2.2.4 The Alternating Direction Implicit
to as implicit (Some workers refer to the (ADI) Method
CN scheme as a semi-implicit method The ADI method was reported by Peace-
since the resulting expression contains man and Rachford [24] in 1955 and intro-
both concentration terms centered at t duced into electrochemical applications by
and t + t.). Solution of Eq. (20) can be Heinze [25–27]. A fully implicit procedure
achieved, since for a general problem of is used to solve multidimensional dif-
N I points, a set of simultaneous equations fusion or convective-diffusion problems.
One reported strategy [28] breaks each in- in an identical manner to those for the
dividual time increment into two half steps CN scheme.
(in the case of a two-dimensional prob- The ADI method was used by Hei-
lem). In the first half-time increment, nze [29] to assess the voltammetric proper-
one of the dimensions is solved explic- ties at microdisc electrodes. Gavaghan [30]
itly whilst the other is retained in implicit reported a thorough examination of opti-
form, for example, mum mesh generation and errors present
in the FD analysis of the disc electrodes
t+t/2
ci,j − ci,j
t
using the ADI method. More recently,
the technique has been widely adopted
t+t/2 t+t/2 t+t/2
= λx ci,j +1 − 2ci,j + ci,j −1 in the field of scanning electrochemical
microscopy [31–37] for the simulation of
+ λy ci+1,j
t
− 2ci,j
t
+ ci−1,j
t
(26) approach curve response and interfacial
chemical reactivity.
where The ADI method has also been applied
Dt Dt to convective-diffusion problems to assess
λx = λy = (27)
(x)2 (y)2 the influence of convective transport on the
voltammetric response of microdisc elec-
and i, j , x, and y are defined in Fig. 5. trodes [38] in channel cells. The convective
In the following half-time increment, the component may be routinely introduced
implicit/explicit directions are swapped to Eq. (28) in an analogous manner to
that noted for the EFD Eq. (18). The
t+t t+t/2
ci,j − ci,j
three-dimensional transport problem was

t+t/2 t+t/2 t+t/2 addressed in an analogous manner to the
= λx ci,j +1 − 2ci,j + ci,j −1
two-dimensional problem by splitting each

t+t t+t t+t full time increment into three time-steps of
+ λy ci+1,j − 2ci,j + ci−1,j (28)
one third, to allow the x, y, and z directions
Values at t + t/2 are known from the to be solved implicitly.
previous time-step and the implicit terms
are calculated at the interval t + t. The set 2.2.2.5 The Hopscotch (HS) Method
of tridiagonal matrix equations obtained The HS method [39] was reported as an
in each half-time increment may be solved alternative strategy to ADI by Shoup and
ci, j+1
ci−1, j ci, j ci+1, j
∆y
y
ci, j−1
∆x
Fig. 5 Two-dimensional mesh
x for diffusional calculations.
Szabo [40–42]. An implicit and explicit adopted by workers in the electrochemical

scheme is employed at alternating points field. Modification of Eqs. (29 and 30)
within the two-dimensional mesh (Fig. 5), to include convective transport has been
which allows the new concentrations to be reported for hydrodynamic applications at
calculated explicitly, but with the advantage a band ultramicroelectrode in a channel
of an implicit stability criteria [22]. A flow cell [43], and at a tubular band
counter (a sum of the number of time microelectrode in which the conditions for
increments plus the i and j pointers) is the current to be independent of flow rate
used to determine whether the calculation are established [44].
is performed explicitly (if the counter
is odd) 2.2.2.6 The Backwards Implicit (BI)
Method
t+t
ci,j − ci,j
t
The fully implicit BI method was used by
t Anderson [45, 46] to simulate the steady
= λx ci,j +1 − 2ci,j + ci,j −1
t t
state current response to flow rate within
t a channel flow cell. Subsequently, this was
+ λy ci+1,j − 2ci,j
t
+ ci−1,k
t
(29)
implemented by Fisher and Compton [47]
or implicitly (if the counter is even) at a to evaluate the time-dependent convective-
specific grid point. diffusion problem

t+t ∂c ∂ 2c ∂c
ci,j − ci,j
t
= D 2 − Vx (31)
∂t ∂y ∂x
t+t t+t t+t
= λx ci,j +1 − 2ci,j + ci,j −1 Cast into fully implicit form, the FD form
of Eq. (31) is given by
t+t t+t t+t
+ λy ci+1,j − 2ci,j + ci−1,j (30)
t+t
ci,j − ci,j
t
The values of λx and λy are given by

t+t t+t t+t
Eq. (27) and i, j , x, y are defined = λy ci+1,j − 2ci,j + ci−1,j
in Fig. 5. When an implicit calculation
t+t t+t
is performed at the point i, j , the −λc ci,j +1 − ci,j (32)
surrounding mesh points will have been
estimated explicitly at the new time where i and j are pointers as shown in
increment using Eq. (29). The counter Fig. 5, λy is given by Eq. (27) and
system ensures that at the next time-
Vx t
increment, the points that have previously λc = (33)
(x)
been calculated explicitly will now be
evaluated implicitly and vice versa. The Following application of the boundary con-
approach allows unrestricted values of ditions, the set of tridiagonal simultaneous
λ to be used and was demonstrated equations may be solved routinely as noted
by Shoup and Szabo to give a good above. Compton and coworkers have used
level of accuracy for the solution of the technique extensively in mechanistic
transport limited current measurements analysis including ECE reactions [48–53]
at the microdisc electrode. Despite its and coupled kinetics reactions [54–59].
relative ease of application and proven The technique has also been used to as-
stability, the method has not been widely sess the influence of hydrodynamics on the
current density at microband, microstrip, transport-related problems. The SIP was

and microdisc electrodes operating in originally reported by Stone [74] in 1968
channel-based cells [60–63]. In addition, as an efficient approach to the solu-
the technique has been used to simulate tion of two- and three-dimensional trans-
voltammetric responses at wall jet and wall port problems. The SIP has been devel-
jet ring disc electrodes [64–67]. oped to tackle a range of steady state
and time-dependent problems related to
2.2.2.7 The Fast Implicit Finite Difference voltammetric measurements. In this ap-
(FIFD) Method proach, the fully implicit form of the
The Fast Implicit Finite Difference method, two-dimensional transport equation is
implemented by Rudolph [68, 69] has solved directly:
marginally higher computational require-
t+t t+t
ments but has higher potential accuracy − λy ci+1,j − λx ci,j +1 + (2λx + 2λy + 1)
when tackling demanding problems. The t+t t+t t+t
FIFD formulation expresses the diffu- × ci,j − λx ci,j −1 − λy ci−1,j = ci,j
t
sional transport equation on an exponen- (38)

tially expanding spatial grid via where λy and λx are given by Eq. (27).
Using this approach, the system equations
∗
cit+t − cit = D2i t+t
ci+1 − cit+t were cast into matrix form to yield

∗
− D1i cit+t − ci−1
t+t
[A]{p} = {q} (39)
(34)
where where p is the vector of unknown con-

centrations, q the vector of known values,
∗ 4
D2i = D ∗ exp 2β −i i > 1 (35) and [A] contains the coefficients from the
3 right-hand side of Eq. (38). To overcome

∗ 5 the traditional difficulties associated with

D1i = D ∗ exp 2β −i i ≥ 2 (36) solution of the pentadiagonal matrices, the
4
matrix [A] can be modified by the addi-
with β, the grid expansion factor and tion of a further matrix [B] so that the
Dt summed matrices [A] + [B] can be fac-
D∗ = (37) torized into upper and lower triangular
(x)2
matrices ([L] + [U ]).
The approach has undergone extensive
development and interested readers are di- [A][B] = [L][U ] (40)
rected to the recent review by Speiser [19].
Feldberg and Rudolph [70] have adopted This approach permits the efficient itera-
one of the developments to create a com- tive solution of the matrix equations using
mercial electrochemical software package a standard NAG routine (NAG FORTRAN
DigiSim [71, 72]. Library (D03EBF)). The approach has been
compared to the ADI and HS methods with
2.2.2.8The Strongly Implicit Procedure the authors concluding that the SIP pro-
(SIP) vides a highly efficient competitor to these
Compton and coworkers recently pro- strategies in both diffusion and convective-
posed the SIP [73] for a variety of diffusion problems [75].
2.2.3 2.2.3.2 Formulation

Finite Element Methods The key steps involved in the formulation
of a one-dimensional diffusional problem
2.2.3.1 Introduction under steady state conditions now follow:
The FEM is part of a larger group of
techniques that exploit the method of ∂ 2c
D =0 (41)
weighted residuals (MWR) [76]. These use ∂x 2
a set of weighting functions to allow
The FEM uses a mesh of elements to cover
the approximation of a variable over a the domain, as shown in Fig. 6
domain. The choice of weighting function The variation of concentration c(x)
leads to a number of different alternative along the element length l is defined
formulations, including the collocation by an interpolation function with the
method, subdomain method, method of concentrations related to the values at the
moments, and the Galerkin method. nodal points (Fig. 6) ca and cb
The FEM traces its lineage back to classi-
cal structural analysis [77] and Argyris [78], c(x) = Na ca + Nb cb (42)
who introduced the concept underlying the
FEM in 1954. In 1956, Turner and cowork- where the linear interpolation functions
ers [79] made a key contribution to the are given by
area when they reported the use of pin- 1
jointed bars and triangular plates for the Na = (xb − x) (43)
l
analysis of aircraft structures. The term,
1
finite element, however, was not widely Nb = (x − xa ) (44)
adopted until Clough [80] used it in the l
title of a paper in 1960. The method and xa and xb are the bounds of each
has subsequently been applied to a vast element on Fig. 6. For an individual ele-
range of problems, including, soil me- ment, these relationships may be written
chanics [81], aerodynamics [82], heat trans- in matrix form
fer [83], electric circuit analysis [84], and
has also been described in a large num- c = N e ce (45)
ber of texts [85–88]. Outlined below is
where
one formulation strategy for the FEM

analysis of one-dimensional diffusional ca

ce = N e = [ Na Nb ] (46)
transport. cb
x
Node
numbers
1 2 3 4 5 6
Electrode 1 2 3 4 5
a l b Element
numbers
ca cb
Fig. 6 One-dimensional element mesh.

Summing the contribution from ne ele- and this is now integrated (by parts)
ments to yield

ne
c= N e ce b
(47) b ∂ N e ∂C

e ∂C
D dx − D N =0
∂x a
e=1
a ∂x ∂x
A weighted residual formulation [89, 90] (52)
is used in order to solve Eq. (41). This ap- From Eq. (47)
proach seeks to approximate the variation
∂C ∂ N e
of concentration over a region in space = C (53)
using a residual R(x) such that ∂x ∂x
and Eq. (52) can be rewritten
∂ 2c
R(x) − 2 = 0 (48)
∂x ∂ N e ∂ N e
b ∂c e b
D dx C − D (N ) = 0
The aim of the numerical procedure is a ∂x ∂x ∂x a
(54)
to ensure that the residual is minimized
Noting the second term of Eq. (54) relates
within the domain (ideally it will be zero).
to the boundary of the element and may
Mathematically, this is achieved by forcing
be applied later; evaluation of Eq. (54)
R(x) to be zero (on average) over a given
now requires the remaining integral. From
finite region in space by multiplying by a
Eq. (45), we can calculate
weighting function w(x)
∂ N 1 1
b
= − (55)
w(x)R(x) dx = 0 (49) ∂x l l
a
it follows from Eq. (54)
The weighting function effectively dis-

tributes any residual error over the space D 1 −1

C=0 (56)
interval defined by the bounds a and b. l −1 1
As noted, a number of possible routes to
apply Eq. (49) exist, including: orthogonal written in matrix form for an individ-
collocation, the subdomain, and Galerkin ual element
methods [89 pg. 78, 91]. In the FEM, the
Galerkin method exploits weighting func- [K]e C = P (57)
tions that take the same form as the
interpolation functions Eqs. (43, 44). The
where
1 −1
residual takes the same form as the origi- [K] =
e
(58)
−1 1
nal differential
([K]e is usually referred to as the charac-
∂ 2C teristic matrix of the element).
R(x) − =0 (50)
∂x 2 For a mesh of ne elements (Fig. 6),
where C is now an approximate solution a global matrix (Fig. 7) is now con-
for the concentration. Recasting Eq. (41) structed with contributions from each
in weighted residual form gives node summed into the relevant ma-
trix position.
b ∂ 2C At this stage, any boundary conditions
D Ne dx = 0 (51)
a ∂x 2 for the problem are introduced (e.g. flux
Node
numbers
1 2 3 4 5 6
1 1−1 c1 k1
2 −1 1 + 1 −1 c2 k2
3 −1 1+1 −1 c3 k3
4 −1 1+1 −1 c4 k4
5 −1 1+1 −1 c5 k5
6 −1 1 c6 k6
Contributions
from each
element Unknown Known
concentration concentration
vector vector
Fig. 7 Global matrix for a problem in which ne = 5.
or fixed values of concentration). For the 2.2.3.3 Applications

internal nodes, it is assumed that the Early applications of the MWR/FEM in
flux across the boundaries between each electrochemistry were reported by Whit-
element are in balance leaving only the ing and coworkers [98] and Speiser and
two values of flux at the external points of coworkers [99–105], who exploited the
the mesh to be defined (if required). Left method for electroanalytical purposes. In
undefined, the values correspond to the 1984, Penczek and coworkers [91, 106]
Neumann boundary condition of no-flux. demonstrated the application of the FEM
However, a fixed concentration, Dirichlet to finite and infinite one-dimensional
boundary condition, may be routinely diffusion problems relating to the voltam-
introduced and further details can be metry at mercury amalgam films.
found in [86 pg. 184]. It is apparent that the Early two-dimensional simulations fo-
matrix formed is symmetric and banded cused on the evaluation of the current
around the center diagonal and this distribution at microdisc electrodes [107,
permits standard Gaussian elimination 108] and simulations of a variety of
routines to be employed for solution of electrode geometries [109–111] including
the matrices [92–94]. the influence of recessed microelectrode
Formulation to introduce convective configurations [112]. Work has been also
transport into electrochemical calculations extended to cases involving coupled ho-
provides no conceptual difficulties and mogeneous kinetics, adsorption [113], and
readers are referred to the article by time-dependent redox polymer electro-
Stevens [95]. Similar procedures may be chemistry [114].
used to formulate two-dimensional analo- More recently, workers have reported
gies and these have been described previ- the use of automatic mesh generation
ously [96, 97]. routines [115] and strategies that permit
the solutions to be obtained to a specified that approximate the (nonlinear) variation

accuracy [116]. The FEM has been utilized of the variable at the boundary. Below, one
to simulate hydrodynamic voltammetry at formulation strategy for a one-dimensional
the rotating disc, channel, wall jet, and diffusion-controlled problem is outlined
related electrode geometries [95, 117–119] Eq. (41).
and as a computer-aided design tool
by Fisher and coworkers [120–124] for 2.2.4.2 Formulation
the development of new electroanalytical
The formulation procedure begins in an
devices and to quantify voltammetry under
analogous manner to that of the FEM.
microfluidic control.
For the case of the diffusional operator
Workers have also used a number
Eq. (41), the weighted residual form is
of commercially available FEM pack-
given by Eq. (51) as before, integration (by
ages to allow the simulation of immis-
parts) yields
cible liquid–liquid interfacial measure-
ments [125–127] and approach curves for
∂w ∂Cb ∂C b
scanning electrochemical microscope ap- D dx − D w(x) =0
plications [128]. Compton and coworkers a ∂x ∂x ∂x a
(59)
have also used a commercial FEM package
A second integration gives the inverse
to model coupled heat and mass transport
formulation
at a wire [129], and dissolution kinetics in
flow-through devices [130].
b ∂ 2w ∂C b
D C dx + D w(x)
a ∂x 2 ∂x a
2.2.4

Boundary Element Method ∂w b

−D C = 0 (60)
∂x a
2.2.4.1 Introduction
The BEM forms part of a group of The procedure yields an integral in which
integral methods [131–142] that have been the differential operator acts on the
used sporadically for the simulation of weighting function and not C, plus two
electrochemically related problems. The terms that are evaluated at the boundary.
BEM was described by Banerjee and Careful selection of w such that
coworkers and Brebbia and coworkers and
has been outlined by them and others in ∂ 2w
=0 (61)
detail in a number of engineering texts [6, ∂x 2
143–146]. A wide range of engineering-
related problems have been tackled using reduces Eq. (60) to an integral expression
the BEM including: heat transfer, fluid relating to the boundary values of concen-
mechanics, and structural analysis. tration, the weighting function, and the
The BEM uses an integral approach gradient of concentration. This procedure
to solve the differential equations related is referred to as the homogeneous solution
to the transport of material within a method and has been applied to a range of
domain. The differential equations are one-dimensional problems. An alternative
transformed into an equivalent integral strategy usually adopted for solving two-
equation at the boundary of the domain and three-dimensional transport problems
and discretized using a series of elements is the use of the Dirac delta function
−δ(x − ξ ) such that 2.2.4.3 Applications

Despite the potential of the BEM to reduce
∂ 2w the dimensionality of the numerical solu-
= −δ(x − ξ ) (62)
∂x 2 tion and provide a direct measure of the
when applied at the point x = ξ , further interfacial flux, it has been poorly exploited
details may be found in [6 pg. 30]. The by workers in the electrochemical field in
choice of w is discussed elsewhere [144] comparison with the FDM and the FEM.
but is dependent upon the specific differ- By comparison, heat and mass transfer
ential to be solved. A general solution that have been widely treated using the BEM
satisfies Eq. (61) is in the engineering literature [148–151]. In
1984, the BEM was employed to calculate
w = Ax + B (63) the primary current distribution during an
electropolymerization reaction [152], the
where A and B are arbitrary con- potential of the BEM for applications to ir-
stants. Selecting two solutions of Eq. (63) regular geometries was also noted. Hume
for example, and coworkers [153] used the approach to
w1 = x (64) analyze mass transport effects of electrode-
position through polymeric masks.
w2 = B (65) Of related interest, Ramachandran
and coworkers have reported a range
and then substituting these into Eq. (60)
of papers on the influence of mass
yields two simultaneous equations
transport [154–156], including diffusion-

∂Ca ∂Cb reaction problems [157]. Further work was
−a + Ca + b − Cb = 0 reported on a variety of current distribution
∂x ∂x
(66) problems [158–161] in the early 1990s,
with a comparison of the FEM and the
∂Ca ∂Cb
− + =0 (67) BEM efficiency reported by Matlosz and
∂x ∂x
coworkers [162]. A two-dimensional study
where Ca , Cb (∂Ca /∂x) and (∂Cb /∂x) cor- of coplanar auxiliary electrodes was re-
respond to the concentrations and fluxes ported by Mehdizadeh and coworkers [163]
at the boundary points a and b respec- and was used to assess the influence of the
tively. Application of boundary conditions electrode configuration on uniform growth
(defining the flux or concentration at each over the cathode electrode. Electroplating
boundary) permits the solution for the un- and corrosion protection in industrial cell
known values. Of particular interest to the configurations have also been addressed
electrochemist is the ability to gain a direct by Druesne and coworkers [164, 165].
measure of the flux from the simulation In 1994, Barbero and coworkers [166]
unlike the FEM and the FDM that rely on extended BEM applications in the elec-
a fine mesh normal to the surface. trochemical field to investigate convec-
Extension of the approach to convective tive electrodiffusion problems in charged
and indeed migratory transport has been membranes. Qiu, Wrobel, and Power have
achieved in electrochemical applications also outlined the application of the BEM to
and readers are referred to the article transport-related problems including de-
by Qiu, Wrobel, and Power [147] for tailed procedures for assessing current
further details. distribution effects controlled by combined
diffusion, convection, migration [147], ex- References

amining binary and three electrolyte ion
systems [167], and in the prediction of elec- 1. S. W. Feldberg in Electroanalytical Chemistry
trode topography [168]. (Ed.: A. J. Bard), Marcel Dekker, New York,
1969, pp. 199–296.
The BEM has been applied in scanning 2. A. J. Bard, L. R. Faulkner, Electrochemical
electrochemical microscopy applications Methods, John Wiley & Sons, New York,
by Fisher and Denuault [169] to examine 1980.
the influence of probe and substrate 3. D. Britz, Digital Simulation in Electrochem-
surface topography. In addition, time- istry, 2nd ed., Springer-Verlag, Berlin, 1988.
4. D. K. Gosser, Cyclic Voltammetry. Simula-
dependent phenomena have been assessed tion and Analysis of Reaction Mechanisms,
in oil droplets [170, 171] and a range of Wiley-VCH, Weinheim, Germany, 1993.
microelectrode geometries using the dual 5. http://www.bath.ac.uk/∼chsacf.
reciprocity method (DRM) [172] closely 6. P. A. Ramachandran, Boundary Element
related to the BEM [173]. Methods in Transport Phenomena, 1st ed.,
Elsevier, Southampton, UK, 1994.
7. L. F. Richardson, Philos. Trans. 1911, 210,
2.2.5 307–357.
Alternative Simulation Strategies 8. R. Courant, K. Friedrichs, H. Lewy, Math.
Ann. 1928, 100, 32.
9. H. W. Emmons, Q. Appl. Math. 1944, 2,
In addition to the topics reviewed above,
173–195.
which form the vast majority of the articles 10. J. E. B. Randles, Trans. Faraday Soc. 1948,
published to date in the field of electro- 44, 327–338.
chemical simulation, there are a number 11. T. Joslin, D. Pletcher, J. Electroanal. Chem.
of other alternative methods that have 1974, 49, 171.
been exploited by workers. These include, 12. S. W. Feldberg, J. Electroanal. Chem. 1981,
127, 1.
statistical techniques such as the Monte 13. L. K. Bieniasz, J. Electroanal. Chem. 1993,
Carlo method [174–179], which has been 360, 119–138.
exploited to examine the fractal nature 14. J. B. Flanagan, K. Takahashi, F. Anson, J.
of electrode surfaces and electrodeposited Electroanal. Chem. 1977, 81, 261.
polymer film growth. The finite volume 15. J. W. Dillard, J. A. Turner, R. A. Ostery-
oung, Anal. Chem. 1977, 49, 1246.
method, which has found significant ap- 16. R. Seeber, S. Stefani, Anal. Chem. 1981, 53,
plication in the engineering literature [180, 1011–1016.
181], remains poorly exploited in the 17. J. von Neumann, R. D. Richtmyer, J. Appl.
electrochemical field [182, 183] as does Phys. 1950, 53, 1021–1099.
the multidimensional upwinding method, 18. F. John, Common Pure Appl. Math. 1952, 5,
155–211.
which has been applied by Van Den Boss- 19. B. Speiser in Electroanalytical Chemistry
che and coworkers [184, 185] to multi-ion (Eds.: A. J. Bard, I. Rubinstein), Marcel
systems at the rotating disc electrode. For Dekker, New York, 1996, pp. 1–108.
recent advances, readers are referred to the 20. J. Crank, P. Nicolson, Proc. Cambridge Phi-
review of Speiser [19]. los. Soc. 1947, 43, 50–67.
21. J. Crank, The Mathematics of Diffusion,
Clarendon Press, Oxford, 1956.
Acknowledgments 22. L. Lapidus, G. F. Pinder, Numerical Solution
of Partial Differential Equations in Science and
Engineering, John Wiley & Sons, New York,
FLQ (GR/N00548) and KAG (studentship 1982.
No. 9930027X) would like to thank the 23. M. Störzbach, J. Heinze, J. Electroanal.
EPSRC for their support. Chem. 1993, 346, 1–27.
24. D. W. Peaceman, H. H. Rachford, J. Soc. 50. J. A. Alden, R. G. Compton, J. Phys. Chem.

Ind. Appl. Math. 1955, 3, 28. B 1997, 101, 9606–9616.
25. J. Heinze, M. Störzbach, Ber. Bunsen-Ges. 51. J. A. Alden, M. A. Feldman, E. Hill et al.,
Phys. Chem. 1986, 90, 1043–1048. Anal. Chem. 1998, 70, 1707–1720.
26. J. Heinze, M. Störzbach, J. Mortensen, J. 52. A. B. Miles, R. G. Compton, J. Phys. Chem.
Electroanal. Chem. 1984, 165, 61–70. B 2000, 104, 5331–5342.
27. J. Heinze, M. Störzbach, J. Mortensen, J. 53. A. B. Miles, R. G. Compton, J. Electroanal.
Electroanal. Chem. 1988, 240, 27–43. Chem. 2001, 499, 1–16.
28. M. K. Jain, Numerical Solution of Differential 54. R. G. Compton, M. B. G. Pilkington, G. M.
Equations, Wiley Eastern Limited, New Stearn, J. Chem. Soc. 1988, 84, 2155–2171.
Delhi, 1979. 55. J. A. Alden, R. G. Compton, R. A. W. Dryfe,
29. J. Heinze, J. Electroanal. Chem. 1981, 124, J. Electroanal. Chem. 1995, 397, 11–17.
73. 56. B. A. Coles, R. A. W. Dryfe, N. V. Rees et al.,
30. D. J. Gavaghan, J. Electroanal. Chem. 1998, J. Electroanal. Chem. 1996, 411, 121–127.
456, 1–35. 57. J. A. Alden, R. G. Compton, J. Phys. Chem.
31. A. J. Bard, F. R. F. Fan, F. F. Fu-Ren et al., B 1997, 101, 8941–8954.
Anal. Chem. 1989, 61, 132. 58. F. Prieto, W. J. Aixill, J. A. Alden et al., J.
32. C. Demaille, P. R. Unwin, A. J. Bard, J. Phys. Chem. B 1997, 101, 5540–5544.
Phys. Chem. 1996, 100, 14 137–14 143. 59. W. J. Aixill, J. A. Alden, F. Prieto et al., J.
33. J. L. Amphlett, G. Denuault, J. Phys. Chem. Phys. Chem. B 1998, 102, 1515–1521.
B 1998, 102, 9946–9951. 60. R. G. Compton, A. C. Fisher, R. G. Welling-
34. A. J. Bard, M. V. Mirkin, P. R. Unwin et al., ton et al., J. Phys. Chem. 1993, 97,
J. Phys. Chem. 1992, 96, 1861–1868. 10 410–10 415.
35. P. R. Unwin, A. J. Bard, J. Phys. Chem. 1991, 61. C. W. Davies, M. K. Walters, A. C. Fisher
95, 7814–7824. et al., J. Phys. Chem. 1995, 99, 10 942.
36. J. V. Macpherson, P. R. Unwin, J. Phys. 62. W. J. Aixill, A. C. Fisher, Q. Fulian, J. Phys.
Chem. 1996, 100, 19 475–19 483. Chem. 1996, 100, 14 067–14 073.
37. J. V. Macpherson, P. R. Unwin, J. Phys. 63. C. W. Davies, Q. Fulian, M. K. Walters et al.,
Chem. 1994, 98, 1704–1713. Electroanalysis 1996, 11, 849.
38. J. Booth, R. G. Compton, J. A. Cooper et al., 64. R. G. Compton, A. C. Fisher, G. P. Tyley, J.
J. Phys. Chem. 1995, 99, 10 942–10 947. Appl. Electrochem. 1991, 21, 2–5.
39. A. R. Gourlay, J. Inst. Math. Appl. 1970, 6, 65. A. C. Fisher, R. G. Compton, J. Appl. Elec-
375. trochem. 1991, 21, 208–212.
40. D. Shoup, A. Szabo, J. Electroanal. Chem. 66. R. G. Compton, A. C. Fisher, M. H. Latham
1982, 140, 237–245. et al., J. Phys. Chem. 1992, 96, 8363–8367.
41. D. Shoup, A. Szabo, J. Electroanal. Chem. 67. J. C. Ball, R. G. Compton, C. M. A. Brett, J.
1984, 160, 1–17. Phys. Chem. B 1998, 102, 162–166.
42. D. Shoup, A. Szabo, J. Electroanal. Chem. 68. M. Rudolph, J. Electroanal. Chem. 1991, 314,
1986, 199, 437–441. 13–22.
43. P. Pastore, F. Magno, J. Lavagnini et al., J. 69. M. Rudolph, J. Electroanal. Chem. 1992, 338,
Electroanal. Chem. 1991, 301, 1–13. 85–98.
44. H. R. Corti, D. L. Goldfarb, A. S. Ortiz et al., 70. S. W. Feldberg, C. I. Goldstein, M. Rudolph,
Electroanalysis 1995, 7, 569–573. J. Electroanal. Chem. 1996, 413, 25–36.
45. S. Moldoveanu, J. L. Anderson, J. Elec- 71. J. Mocak, S. W. Feldberg, J. Electroanal.
troanal. Chem. 1984, 175, 67. Chem. 1994, 378, 17–29.
46. J. L. Anderson, S. Moldoveanu, J. Elec- 72. M. Rudolph in Physical Electrochemistry (Ed.:
troanal. Chem. 1984, 179, 107–119. I. Rubinstein), Marcel Dekker, New York,
47. A. C. Fisher, R. G. Compton, J. Phys. Chem. 1995, pp. 81–129.
1991, 95, 7538–7542. 73. R. G. Compton, R. A. W. Dryfe, R. G. Well-
48. A. C. Fisher, R. G. Compton, J. Appl. Elec- ington et al., J. Electroanal. Chem. 1995, 383,
trochem. 1992, 22, 38–42. 13–19.
49. R. G. Compton, B. A. Coles, A. C. Fisher, J. 74. H. L. Stone, Siam J. Numer. Anal. 1968, 5,
Phys. Chem. 1994, 98, 2441–2445. 530.
75. J. A. Alden, R. G. Compton, J. Electroanal. Cambridge University Press, Cambridge,

Chem. 1996, 402, 1–10. 1996.
76. B. A. Finlayson, Br. Chem. Eng. 1969, 14, 95. N. P. C. Stevens, A. C. Fisher, J. Phys.
53–57. Chem. B 1997, 101, 8259–8263.
77. R. Courant, Bull. Am. Math. Soc. 1943, 49, 96. A. J. Chapman, Heat Transfer, 4th ed.,
1–23. Macmillan Publishing, New York, 1984.
78. J. H. Argyris, Aircraft Eng. 1954, 26, 347. 97. J. P. Holman, Heat Transfer, 8th ed.,
79. M. J. Turner, R. W. Clough, H. C. Martin McGraw-Hill, New York, 1997.
et al., J. Aerosol Sci. 1956, 23, 805. 98. L. F. Whiting, P. W. Carr, J. Electroanal.
80. R. W. Clough, Proceedings of 2nd ASCE Chem. 1977, 81, 1–20.
Conference on Electronic Computation, 99. B. Speiser, A. Rieker, J. Electroanal. Chem.
Pittsburg, Pa., 1960. 1979, 102, 1–20.
81. C. V. Girijavallabhan, L. C. Reese, J. Soil 100. B. Speiser, J. Electroanal. Chem. 1980, 110,
Mech. Proc. ASCE 1968, 94, 473–496. 69–77.
82. S. S. Rao, Proc. Int. Symp. Disc. Methods Eng. 101. S. Pons, B. Speiser, J. F. McAleer, Elec-
Milan, 1974, pp. 512–525. trochim. Acta 1982, 27, 1177–1179.
83. S. G. Ravikumaur, K. N. Seetharamu, P. A. 102. B. Speiser, S. Pons, A. Rieker, Electrochim.
Aswathanarayana, Proc. Int. Conf. On Finite Acta 1982, 27, 1171–1176.
Elements in Computational Mechanics 1985, 103. B. Speiser, J. Electroanal. Chem. 1984, 171,
861, Vol. 2. 95–109.
84. P. P. Silvester, R. L. Ferrari, Finite Elements 104. P. Hertl, B. Speiser, J. Electroanal. Chem.
for Electrical Engineers, 3rd ed., Cambridge 1987, 217, 225–238.
University Press, Cambridge, 1996.
105. P. Hertl, B. Speiser, J. Electroanal. Chem.
85. O. C. Zienkiewicz, R. L. Taylor, The Finite
1988, 250, 237–256.
Element Method, 5th ed., Butterworth-
106. M. Penczek, Z. Stojek, J. Electroanal. Chem.
Heinemann, Oxford, 2000.
1984, 181, 83–91.
86. S. S. Rao, The Finite Element Method in
107. J. Galceran, D. J. Gavaghan, J. S. Rollet, J.
Engineering, Pergamon Press, Oxford, 1982.
Electroanal. Chem. 1995, 394, 17.
87. G. Dhatt, G. Touzot, The Finite Element
108. M. Penczek, Z. Stojek, J. Electroanal. Chem.
Method Displayed, John Wiley & Sons, New
1987, 227, 271–274.
York, 1984.
88. C. Cuvelier, A. Segal, A. A. van Steenhoven, 109. M. M. Stephens, E. D. Moorhead, J. Elec-
Finite Element Methods and Navier-Stokes troanal. Chem. 1987, 220, 1–30.
Equations, D. Reidel, Dordrecht, The 110. N. P. C. Stevens, S. J. Hickey, A. C. Fisher,
Netherlands, 1986. An. Quim. 1997, 93, 225–232.
89. J. Villadsen, M. L. Michelsen, Solution of 111. N. P. C. Stevens, F. L. Qiu, K. A. Gooch
Differential Equations by Polynomial Approxi- et al., J. Phys. Chem. B 2000, 104,
mation, Prentice Hall, Englewood Cliffs, NJ, 7110–7114.
1978. 112. P. N. Bartlett, S. L. Taylor, J. Electroanal.
90. G. Fairweather, Finite Element Galerkin Chem. 1998, 453, 49–60.
Methods for Differential Equations, John 113. E. D. Moorhead, M. M. Stephens, J. Elec-
Wiley & Sons, New York, 1978. troanal. Chem. 1990, 282, 1–26.
91. M. Penczek, Z. Stojek, J. Osteryoung, J. 114. E. Deiss, O. Haas, C. Daul, J. Electroanal.
Electroanal. Chem. 1984, 170, 99–108. Chem. 1992, 337, 299–324.
92. E. Isaacson, H. B. Keller, Analysis of Numer- 115. T. Nann, J. Heinze, Electrochem. Commun.
ical Methods, John Wiley & Sons, New York, 1999, 1, 289–294.
1966. 116. K. Harriman, D. Gavaghan, P. Houston
93. L. W. Johnson, R. D. Riess, Numerical Anal- et al., Electrochem. Commun. 2000, 2,
ysis, 2nd ed., Addison-Wesley, Reading, 150–170, 567–585.
Mass., 1982. 117. R. Ferrigno, P. F. Brevet, H. H. Girault, J.
94. W. H. Press, S. A. Teukolsky, W. T. Vetterl- Electroanal. Chem. 1997, 430, 235–242.
ing et al., Numerical Recipes in Fortran 77: 118. N. P. C. Stevens, A. C. Fisher, Electroanaly-
The Art of Scientific Computing, 2nd ed., sis 1998, 10, 16–20.
119. B. A. Coles, R. G. Compton, C. M. A. Brett 144. C. A. Brebbia, J. C. F. Telles, L. C. Wrobel,

et al., J. Electroanal. Chem. 1995, 381, Boundary Element Techniques. Theory and
99–104. Application in Engineering, Springer-Verlag,
120. Q. Fulian, N. P. C. Stevens, A. C. Fisher, J. Berlin, Germany, 1984.
Phys. Chem. B 1998, 102, 3779–3783. 145. C. A. Brebbia, Boundary Element Method for
121. K. A. Gooch, N. A. Williams, A. C. Fisher, Engineers, Pentech, London, 1984.
Electrochem. Commun. 2000, 2, 51–55. 146. B. E. Beskos, Boundary Element Methods
122. Q. Fulian, A. C. Fisher, D. J. Riley, Electro- in Mechanics, Computational Methods in
analysis 2000, 12, 503–508. Mechanics, North Holland, Amsterdam,
123. F. L. Qiu, K. A. Gooch, A. C. Fisher et al., 1987, 1–25, Vol. 3.
Anal. Chem. 2000, 72, 3480–3485. 147. Z. H. Qiu, L. C. Wrobel, H. Power, Eng.
124. N. P. C. Stevens, K. A. Gooch, A. C. Fisher, Anal. Boundary Elements 1995, 15, 299.
J. Phys. Chem. B 2000, 104, 1241–1248. 148. P. H. L. Groenenboom, Appl. Math. Modell.
125. J. Josserand, J. Morandini, H. J. Lee et al., J. 1982, 6, 35.
Electroanal. Chem. 1999, 468, 42–52. 149. C. A. Brebbia, P. Skerget, J. Appl. Phys.
126. R. Ferrigno, H. H. Girault, J. Electroanal. 1982, 53, 8366.
Chem. 2000, 492, 1–6. 150. W. L. Wendland, J. Zhu, Math. Comput.
127. H. J. Lee, C. Beriet, R. Ferrigno et al., J. Modell. 1991, 15, 19.
Electroanal. Chem. 2001, 502, 138–145. 151. S. Y. Long, X. C. Kuai, J. Chen et al., Eng.
128. Y. Lee, S. Amemiya, A. J. Bard, Anal. Chem. Anal. Boundary Elements 1993, 12, 293.
2001, 73, 2261–2267. 152. R. Bialecki, R. Nahlik, M. Lapkowski, Elec-
129. A. Beckmann, B. A. Coles, R. G. Compton trochim. Acta 1984, 29, 905.
et al., J. Phys. Chem. B 2000, 104, 153. E. C. Hume, W. M. Deen, R. A. Brown, J.
Electrochem. Soc. 1984, 131, 1251.
764–769.
154. P. L. Mills, S. Lai, M. P. Dudukovic et al.,
130. B. A. Coles, R. G. Compton, M. Suarez
Siam J. Sci. Stat. Comput. 1988, 9, 271.
et al., Langmuir 1998, 14, 218–225.
155. P. A. Ramachandran, Int. J. Numer. Methods
131. S. Coen, D. K. Cope, D. E. Tallman, J. Elec-
Eng. 1990, 29, 1021.
troanal. Chem. 1986, 215, 29–48.
156. P. A. Ramachandran, Chem. Eng. J. 1990,
132. M. Fleischmann, J. Daschbach, S. Pons, J.
45, 49.
157. P. A. Ramachandran, J. Comput. Phys. 1992,
133. J. Daschbach, S. Pons, M. Fleischmann, J. 102, 63.
Electroanal. Chem. 1989, 263, 205–224. 158. J. F. Yan, S. N. R. Pakalapati, T. V. Nguyen
134. M. Rudolph, J. Electroanal. Chem. 1990, 292, et al., J. Electrochem. Soc. 1992, 139, 1932.
1–7. 159. P. Cicognani, F. Gasparoni, B. Mazza et al.,
135. U. Kalapathy, D. E. Tallman, D. K. Cope, J. J. Electrochem. Soc. 1990, 137, 1689.
Electroanal. Chem. 1990, 285, 71–77. 160. J. Horkins, L. T. Romankiw, J. Electrochem.
136. D. K. Cope, C. H. Scott, D. E. Tallman, J. Soc. 1989, 136, 756.
Electroanal. Chem. 1990, 285, 49–69. 161. E. K. Yung, L. T. Romankiw, J. Electrochem.
137. D. K. Cope, D. E. Tallman, J. Electroanal. Soc. 1989, 136, 764.
Chem. 1990, 285, 79–92. 162. M. Matlosz, C. Creton, C. Clerc et al., J.
138. D. R. Baker, M. W. Verbrugge, J. Newman, Electrochem. Soc. 1987, 134, 3015.
J. Electroanal. Chem. 1991, 314, 23–44. 163. S. Mehdizadeh, J. Dukovic, P. C. Andricacos
139. M. V. Mirkin, A. J. Bard, J. Electroanal. et al., J. Electrochem. Soc. 1990, 137, 110.
Chem. 1992, 323, 1–27, 29–51. 164. F. Druesne, P. Paumelle, Corros. Prevention
140. L. K. Bieniasz, Comput. Chem. 1992, 16, Control 1998, 45, 118.
311–317. 165. F. Druesne, P. Paumelle, P. Villon, Eng.
141. L. K. Bieniasz, J. Electroanal. Chem. 1993, Anal. Boundary Elements 2000, 24, 615.
347, 15–30. 166. A. J. Barbero, S. Mafe, P. Ramirez, Elec-
142. B. Pillay, J. Newman, J. Electrochem. Soc. trochim. Acta 1994, 39, 2031.
1993, 140, 414–420. 167. Z. H. Qiu, L. C. Wrobel, H. Power, J. Appl.
143. P. K. Banerjee, R. Butterfield, Boundary El- Electrochem. 1997, 27, 1333.
ement in Engineering Science, McGraw-Hill, 168. Z. H. Qiu, H. Power, J. Appl. Electrochem.
New York, 1981. 2000, 30, 575.
169. F. Qiu, A. C. Fisher, G. Denuault, J. Phys. 179. G. Nagy, Y. Sugimoto, G. Denuault, J. Elec-
Chem. B 1999, 103, 4387, 4393. troanal. Chem. 1997, 433, 167–173.
170. F. L. Qiu, J. C. Ball, F. Marken et al., Electro- 180. H. K. Versteeg, W. Malalasekera, An Intro-
analysis 2000, 12, 1012–1016. duction to Computational Fluid Dynamics:
171. J. C. Ball, F. Marken, F. L. Qiu et al., Electro- The Finite Volume Method, Longman Group,
analysis 2000, 12, 1017–1025. Harlow, UK, 1995.
172. F. L. Qiu, A. C. Fisher, Electrochem. Com- 181. L. Ronghua, W. Wei, C. Zhongying, Gen-
mun. 2000, 2, 738–742. eralised Difference Methods for Differential
173. D. Nardini, C. A. Brebbia in Boundary Equations: Numerical Analysis of Finite Vol-
Element Methods in Engineering (Ed.: ume Methods, Marcel Dekker, New York,
C. A. Brebbia), Springer-Verlag, Berlin, 2001.
1982, pp. 312–326. 182. S. Bacha, A. Bergel, M. Comtat, J. Elec-
174. B. Marner, W. Schmickler, J. Electroanal. troanal. Chem. 1993, 359, 21–38.
Chem. 1986, 214, 589–596. 183. Yu. V. Benderskii, V. G. Mairanosvkii, Elek-
175. B. Aurian-Blajeni, M. Kramer, M. Tomkie- trokhimiya 1992, 28, 835–841.
wicz, J. Phys. Chem. 1987, 91, 600–605. 184. B. Van Den Bossche, L. Bortels, J. Decon-
176. T. Pajkossy, L. Nyikos, Electrochim. Acta inck et al., J. Electroanal. Chem. 1995, 397,
1989, 34, 171–179. 35–44.
177. F. Sagués, J. M. Costa, J. Chem. Educ. 1989, 185. B. Van Den Bossche, L. Bortels, J. Decon-
66, 502–506. inck et al., J. Electroanal. Chem. 1996, 411,
178. S. Záliš, N. Fanelli, L. Pospı́šil, J. Elec- 129–143.
troanal. Chem. 1991, 314, 1–11.
147
3.1 for many years for scrutinizing complex

NMR Spectroscopy in Electrochemistry surface processes [2]. Unfortunately, few
of these techniques are readily applicable
YuYe Tong
to investigating problems at the electro-
Georgetown University, Washington, Washing-
chemical interface, because of the ubiq-
ton DC
uitous presence of the condensed-phase
Eric Oldfield and Andrzej Wieckowski electrolyte. In electrocatalysis/fuel cell ap-
University of Illinois at Urbana-Champaign, plications, for instance, specific chemical
Urbana, Illinois and physical properties of the electrode be-
come of paramount importance in terms
3.1.1 of understanding the fundamental as-
Introduction pects of such interfacial electrochemical
phenomena [3]. It is in this field of in-
The fundamental physical origins of the terfacial electrochemistry that the newly
different chemical reactivities of catalytic developed electrochemical nuclear mag-
or functional metal surfaces – be they used netic resonance (EC-NMR) spectroscopy
as heterogeneous catalysts, electrocata- technique [4] is mainly applicable. This
lysts, substrates for materials engineering, development is now beginning to bring
or simply as templates for corrosion stud- results that illuminate both electronic and
ies – are electronic [1]. Most surface pro- dynamic aspects of processes in real-world
cesses can be considered to involve the catalytic systems [4–10].
optimization of the electronic interaction Technically, however, NMR probably
energy of the system through suitable ge- possesses the lowest mass-detection sen-
ometric rearrangements. These surfaces sitivity of any spectroscopic technique [11]
processes necessarily involve the activa- and is still quite a challenging prospect.
tion of one or more chemical ingredients To appreciate the difficulties, note that a
together with the concomitant excitation typical high-field NMR instrument needs
of the many-electron metal surface. They ∼1018 –1019 NMR-active atoms, for exam-
are, therefore, in general rather complex ple 13 C spins, to make a signal detectable
processes. within a reasonable time period. However,
In the gas (dry) phase, ultrahigh vacuum 1 cm2 of a single-crystal metal surface con-
(UHV)-based surface science characteri- tains only about 1015 atoms. Therefore,
zation techniques have been developed at least 1 m2 surface area of NMR-active
148 3 New Experimental Evidences
atoms is needed to meet the sensitivity more challenging than the more standard
requirement, which is the primary reason ‘‘gas-phase’’ techniques because of the
that most surface NMR investigations re- basic incompatibility between NMR de-
ported so far have used high-surface-area tection and the conducting requirements
materials. Nevertheless, there are several for proper electrode function. That is, the
advantages to using NMR to investigate presence of conducting material in the
(electrochemical) surface processes. For NMR-detection coil greatly degrades the
example, NMR is extremely sensitive to so-called quality factor of the NMR probe,
local static (geometric), electronic, and which substantially decreases sensitivity.
dynamic structure [12]. Also, NMR can Nevertheless, the potential advantages of
characterize samples in most types of cat- interfacing in situ solid-state NMR with
alysts – metal powders, oxide-supported electrochemistry for investigating the elec-
metal catalysts, carbon-supported cata- trochemical interface are many. For ex-
lysts, and both metal and adsorbate struc- ample: NMR gives a direct probe of the
tures can be probed. In addition, it is Fermi level local densities of states – for
a nondestructive technique and can be both adsorbate and substrate; structure
applied under technically demanding en- and bonding in an electrochemical en-
vironmental conditions close to real-world vironment can be investigated using a
operating conditions. For example, it is variety of NMR interactions; surface dy-
possible to investigate graphite-supported namics can be readily probed; and, of
metal catalysts in an electrochemical en- course, the effects of potential control can
vironment under potential control. Given be investigated in working electrochemi-
adequate mass sensitivity then, NMR has cal systems – even where the adsorbates
a unique ability to provide a wealth of (such at CN on Pt) would not be ac-
electronic and structural information on cessible in conventional ‘‘gas-phase’’ or
the atomic level, and in addition per- dry systems. Furthermore, by using high
mits one to access motional information magnetic fields and high-surface-area elec-
on adsorbates over a time-span unattain- trode materials, the current state-of-the-art
able by any other spectroscopic technique. signal-detection sensitivity is in fact quite
As such, NMR of oxide-supported metal reasonable, and permits the routine acqui-
catalysts has therefore played an impor- sition of NMR spectra of surface species.
tant historical role in the development of As an example, we show in Fig. 1 13 C
many fundamental aspects of gas-phase NMR spectra of 13 CO chemisorbed onto
heterogeneous catalysis at transition metal a 10-nm Pt sample having a CO coverage
surfaces [13–16], and the techniques de- of about 0.8, recorded at room tempera-
veloped, primarily by Slichter and cowork- ture and in, respectively, Fig. 1(a), 14.1 and
ers [17], can be readily adapted to electro- Fig. 1(b), 8.5 T magnetic fields. The num-
chemical systems. bers of scans were 1600 and 2000, and the
EC-NMR was pioneered by Wieck- signal-to-noise (S/N) ratios (after using a
owski’s group [18] at the University of 1-kHz line broadening) were 50 and 10,
Illinois at the beginning of 1990s and, respectively, showing a significant S/N en-
as noted above, has its roots in metal hancement in the higher magnetic field.
NMR of oxide-supported metal catalysts. Sensitivity is also greatly enhanced at lower
However, interfacing solid-state NMR with temperatures at which 13 C NMR spectra
interfacial electrochemistry is somewhat can now be recorded in as little as 10 min.
3.1 NMR Spectroscopy in Electrochemistry 149
Fig. 1 13 C NMR spectra of CO

adsorbed on a 10-nm Pt electrocatalyst,
recorded at room temperature and at
(a) 14.1 T with 1600 scans and (b) 8.5 T
with 2000 scans. The S/N ratios are
ca. 50 and 10, respectively.
3.1.2
NMR amplitude
Principles of Solid-state Surface EC-NMR
(a)
Spectroscopy
3.1.2.1 Metal NMR Basics

In electrochemistry, electrodes are indis-
pensable parts of an electrical circuit, so the
EC-NMR of electrode surfaces in general
(b)
has to deal with metallic systems. In NMR,
in general, a specific resonance position (a
frequency shift) with respect to a well-
defined reference corresponds to a specific
chemical environment (a structure) [12].
This shift is the chemical shift (δCS ) in a 800 600 400 200 0
molecule or the Knight shift (K) [19] in a d
metal, and can be considered a fingerprint [ppm]
of the corresponding site or structure [20].
While the familiar chemical or orbital shift
Theoretically, the Hamiltonian of the nu-
is ubiquitous, arising from the shielding
clear–electron hyperfine interaction can
of the external magnetic field at the nu-
be written for a nuclear spin I = 1/2 as
cleus by the otherwise quenched orbital

angular momentum of the surrounding 8π
electrons, the Knight shift is metal-specific, H = γn h̄I µB S(r)δ(r)
3
being produced by nonzero magnetic in-
S 3r(Sr)
teractions between nuclear and electronic − 3− − 3 (1)
spins since only electron spins at the r r5 r
Fermi energy can be polarized and thus where µB is the Bohr magneton and γn is
create a nonzero spin density in an ex- the nuclear gyromagnetic ratio. I , S, and l
ternal magnetic field (the so-called Pauli are the nuclear spin, electron spin, and
paramagnetism). In addition, the Korringa electron orbital moments, respectively.
relationship [21] (vide infra), which gov- The symbol r is the radius vector of an
erns the nuclear spin–lattice relaxation in electron with respect to the nucleus at the
metals, is also metal-specific. Together, origin. The first term in the brace in Eq. (1)
the Knight shift and the Korringa relation is the Fermi contact term, the second
represent the two primary NMR probes (square bracket) is the dipole term, and the
of electronic structure in metals, and of third (l/r 3 ) is due to the orbital movement
surface molecular bonding in general. of electrons. In a diamagnetic molecule,
since all electron spins are paired up, the where Hhf is the hyperfine field produced
expectation values of the first and second by an electron at the site of the nucleus,
terms become zero, leaving only the third and D(EF ) is the (local) density of
term, which produces the usual so-called states at the Fermi level. Despite their
chemical or orbital shift, δCS . In the case simplicity, Eqs. (2 and 3) already give the
of a metal, the total line shift is the sum simple physical picture that the NMR of
of the first term (K) and the third term metallic species can measure, in principle,
(δCS ) in Eq. (1). In many cases, K, which D(EF ). For systems having more than
arises from the hyperfine interaction of one band, which cuts the Fermi level,
the observed nucleus with electron spins, the deduction of multiple D(Ef )s becomes
dominates the purely chemical shift term, more involved, though the principles are
δCS . The dipolar term normally contributes the same. More specifically, for 195 Pt [22]
to relaxation, but not to the shift, since the or chemisorbed 13 CO [10], one can express
trace of the operator is zero. the two primary metal NMR observables,
The Korringa relationship [21] indicates the Knight shift K and the Korringa
that 1/T1 ∝ T , where T1 is the spin–lattice constant S(T1 T )−1 , in terms of D(EF )s.
relaxation time and T is the absolute This is the so-called two-band model, the
temperature of the sample. This unique bands corresponding to metal s and d
temperature dependence of 1/T1 is the bands or to ligand (13 CO) 5σ and 2π ∗
NMR fingerprint of a metallic state. It bands. By experimentally measuring K
results from the fact that only conduc- and S(T1 T )−1 (for the metal or ligand)
tion electrons around the Fermi level and by solving two equations for two
can satisfy energy conservation for the unknowns, s (5σ )- and d (2π ∗ )-D(EF ) can
electron-nuclear spin ‘‘flip-flop’’ relaxation be deduced. For more details, the reader is
process, and the fraction of these electrons referred to Refs. [10, 20, 22].
is proportional to kB T . When all relaxation It is important to note that many metal-
mechanisms other than the first term in lic properties, such as the Knight shift
Eq. (1) can be neglected, and there are only and the Korringa relationship, are deter-
s-like electrons at the Fermi level (such as mined by the finite and quasi-continuous
in the alkali metals), the Korringa relation- nature of the Fermi level local density
ship takes its simplest form: of states (EF -LDOS). In the approxima-
tion most familiar to chemists, what this
2 means is that the highest occupied molec-
2 γe h
K T1 T = =S (2) ular orbital–lowest unoccupied molecular
γn 4πkB
orbital (HOMO–LUMO) gap in metals is
Here, γe and γn are the electronic and much smaller than the thermal energy
nuclear gyromagnetic ratios, respectively, kB T , and the value of the EF -LDOS re-
and kB is the Boltzmann constant. The flects the frontier orbital contributions in
Knight shift, K, also has a simple a metallic system [23]. The EF -LDOS also
relationship with the electronic density of represents a crucial metal surface attribute
states at the Fermi level through the Pauli that can serve as an important concep-
susceptibility, χPauli : tual bridge between the delocalized band
structure (physics) picture and the local-
Hhf ized chemical bonding (chemical) picture
K = χPauli = µB Hhf D(EF ) (3) of metal–adsorbate interactions.
µB
In addition to the determination of oxygen-free N2 or Ar, are transferred into

the chemical and/or Knight shift and a precleaned glass ampoule and flame-
the spin–lattice relaxation time in the sealed. The potential drift of a sample in
laboratory frame (T1 ), there is another im- a conventional three-electrode cell, in an
portant NMR observable – the spin–spin oxygen-free environment, is only a few
relaxation time (T2 ). While the chemical millivolts over 12 h. The same observation
and/or Knight shift contains essentially holds after transferring the electrode ma-
static structural information, the tempera- terial from the NMR ampoule back into
ture and/or magnetic field dependence of the cell. These observations suggest that
the relaxation times, both T1 and T2 , are the surface potential does not change sig-
related to the dynamics of the observed nificantly once a sample is sealed (and
nucleus. T1 measures the rate at which does not change at all when the sample
the spin system returns to thermal equi- is frozen).
librium with its environment (the lattice) For room temperature electrode po-
after a perturbation, while T2 measures tential–dependence studies, we have in-
the rate of achieving a common spin tem- corporated an electrochemical cell inside
perature within the spin system. Both T1 the NMR probe [6]. This permits running
and T2 provide exceptionally important in- NMR measurements while an (variable)
formation on motions, and can cover the external electrode potential is applied.
timescale from ∼ 10−9 to 102 s. Moreover, Figure 2 shows a picture of a real EC-NMR
the temperature dependence of these mo- probe as well as a schematic diagram of
tions provides important thermodynamic the setup. The working electrode mate-
information in the form of activation en- rial was loosely packed in the cell that
ergies for ligand motion on the catalyst was placed inside an NMR coil and elec-
surface. trical connection to an external circuit
(potentiostat) made by inserting a Pt wire
3.1.2.2 EC-NMR Instrumentation into the sample. (A very similar setup
Two types of EC-NMR experiments has also been developed by the Lausanne
have been carried out: electrode poten- group [8, 24].) Our cell design permits both
tial–dependent studies of an adsorbate electrochemical sample preparation and
(13 CO, 13 CN) at room temperature [6, 8, characterization (voltammetry) and NMR
24], and temperature-dependent studies data acquisition under active potential con-
of an adsorbate [5, 10] and of platinum trol, avoiding sample transfer from the
electrocatalysts [7, 25], down 10 K. For preparative electrochemical cell. A very im-
all of the EC-NMR measurements, the portant technical issue here is that while
electrode materials, either polycrystalline the potentiostat is on, there is consider-
platinum black or carbon-supported com- able ‘‘noise’’ injected into the NMR probe,
mercial fuel cell grade platinum electrodes, since the potentiostat leads act as excel-
were immersed in a supporting electrolyte, lent ‘‘antennas’’ right inside the NMR coil!
typically 0.5 M H2 SO4 . For temperature- This noise has to be removed by using
dependence studies, EC-NMR samples are extensive electronic filtering of the electro-
prepared in a conventional three-electrode chemical leads entering the probe, in order
flow cell with oxygen-free N2 or Ar as a to eliminate extensive radio frequency (rf)
protecting gas, then together with support- pickup. The low-pass filter circuit elements
ing electrolyte and under the protection of Lcounter Ccounter , Lref . Cref . , and Lwk. Cwk.
NMR spectrometer
Cmatching
Ctuning
Counter NMR coil Working

electrode electrode
Reference
Lcounterelectrode Lwk.
Ccounter Lref. Cwk.
Cref.
Potentiostat
Fig. 2Schematic diagram (left) and picture (right) of an EC-NMR probe and its circuitry,
showing the interface between NMR and electrochemistry.
in Fig. 2 were designed for this purpose, were then connected to each other and
plus they serve a second purpose of acting to the external potentiostat. Cyclic voltam-
as a high impedance to the rf pulses from mograms of high quality are obtained by
the NMR spectrometer. These are typically both types of EC-NMR cell.
∼1 kW and may be incompatible with long
potentiostat life. 3.1.3
Figure 3 shows another type of setup for Applications of Solid-state Surface EC-NMR
an EC-NMR cell, developed by the Berkeley Spectroscopy
group to investigate 13 CO adsorption from
3.1.3.1 195 Pt
NMR of Carbon-supported
the gas phase onto a carbon-supported
platinum electrocatalyst [26, 27]. In this Electrocatalysts and a Potential-scan-
system, a platinum electrocatalyst was generated Sintering Effect
pressed into a graphitized carbon cloth At the beginning of the 1980s, Slichter
that was then cut into pieces and stacked and coworkers discovered several unique
into a cylindrical form that fits snugly features of the 195 Pt NMR of oxide-
into the cell’s compartment. In order to supported small platinum particles [28].
reduce degradation in the quality factor They found that the overall 195 Pt NMR
of the NMR coil due to the presence lineshape was extremely broad, extending
of bulk electronic or ionic conductivity, downfield some 4 kG from the position
the electrode layers were also alternately of bulk platinum (1.138 G kHz−1 ), and
sandwiched with nonconductive fiberglass contained a feature on the low-field side
cloth. All the electrode layers were in- (1.089 G kHz−1 ), which arose from the
dividually stitched with gold wire and oxidized Pt surface atoms. Later on, van
F
E
G
H
A
D
Fig. 3 Schematic representation of the EC-NMR probe cell

developed by the Berkeley group [27]: (A) reference electrode
compartment; (B) NMR coil; (C) working electrode compartment;
(D) counterelectrode compartment; (E) electrolyte inlet;
(F) electrolyte outlet; (G) carbon-supported Pt electrocatalysts; and
(H) fiberglass separator. (Reproduced with permission from
Ref. [27], Copyright by  2001 Electrochemical Society.)
der Klink and coworkers confirmed these is this distinguishable surface signal that
observations, finding that the signal from makes 195 Pt NMR unique in investigat-
clean-surface Pt atoms was centered at ing the surface physics and chemistry of
1.100 G kHz−1 [29], a position clearly very nanoscale platinum particles.
different to the 1.138 G kHz−1 bulk posi- We have observed that carbon-supported
tion (the frequency difference is magnetic nanoscale platinum electrodes retain quite
field dependent and is about 2.5 MHz in a closely the 195 Pt NMR spectral character-
field of 8.5 T). Ab initio theoretical calcu- istics of isolated small platinum particles
lations on a five-layer Pt (001) cluster [40] supported on oxides [4, 7, 25, 30], with the
then demonstrated that the surface shift clean-surface platinum atoms also resonat-
must be due to a gradual drop in the d- ing at 1.100 G kHz−1 with respect to the
like Fermi level LDOS on moving from value for bulk atoms, 1.138 G kHz−1 . We
the inside of the particle to the surface. It show in Fig. 4(a–c), 195 Pt NMR spectra
a b
NMR Amplitude
c d
1.08 1.10 1.12 1.14 1.08 1.10 1.12 1.14

Field/frequency
[G kHz−1]
Fig. 4 Typical point-by-point 195 Pt NMR spectra showing electrochemical
cleaning, sintering by potential cycling, and a layer-model analysis of the
2.5-nm sample: (a) as-received catalyst; (b) electrochemically cleaned in
0.5 M H2 SO4 by holding electrode potential at 0.45 V versus reversible
hydrogen electrode (RHE); (c) cleaned by extensive potential cycling; and
(d) layer-model deconvolution of spectrum (b). The solid line in (b) is the
result of the simulation.
of a 2.5 nm carbon-supported commercial for the clean-surface spectrum (Fig. 4b) are
Pt electrocatalyst after different surface shown in Fig. 4(d). The NMR layer model
treatments. Figure 4(a) shows the spec- assumes that the nanoscale platinum
trum of an as-received sample, while particles can be represented by ideal
Fig. 4(b) shows a sample that has been cubo-octahedral particles built up layer-
electrochemically cleaned by holding the by-layer from a central atom, that NMR
electrode potential within the electrochem- signals from atoms within a given layer
ical double-layer region with no potential can be approximated by a Gaussian, and
cycling until the reduction current could that the average Knight shift of the nth
no longer be measured. Figure 4(c) shows layer, Kn , which is the center of the
the spectrum obtained after extensive po- corresponding Gaussian, ‘‘heals’’ back
tential cycling (which is still widely used exponentially towards the bulk platinum
to clean electrode surfaces), and in which position when moving inwards. That is,
there are clearly discernable spectral dif-
−n
ferences, as discussed below. Kn = K∞ + (K0 − K∞ ) exp (4)
m
In order to further analyze these results,
we use a layer-model [31] (vide infra) where n is the layer number, counting
deconvolution of the lineshape, and results inwards from the surface layer (where
n = 0), m is the characteristic number lowering provides clear evidence for sin-
(of layers) defining the ‘‘healing length’’ tering produced by potential cycling, an
for the Knight shift, and K0 and K∞ effect that clearly needs to be borne in
(= 3.34%) are the Knight shifts of the mind when analyzing the results of both
surface layer and the bulk, respectively. ligand and metal EC-NMR experiments.
The relative contribution of each Gaussian
is dictated by the fraction of the atoms 3.1.3.2 Correlation between the Knight
within the corresponding layer, which can Shift of a Platinum Surface and the
be determined from the size distribution Electronegativity of the Adsorbate
of the sample. The healing length, de- Understanding how adsorbates modify the
fined as m times the distance between two chemical and physical properties of metal
consecutive layers (0.229 nm for Pt), is surfaces has long been one of the cen-
0.46 nm (∼ =2 Pt layers) from the deconvo- tral themes of surface science, and is of
lution shown in Fig. 4(d). In addition, the importance in relation to the ability to
shape of a 195 Pt NMR spectrum conveys scientifically engineer metal surface prop-
information on the size of the electrocat- erties for targeted applications, such as
alyst. That is, the stronger is the surface enhancing the CO tolerance of electrocat-
peak, the higher is the dispersion, or the alysts used in fuel cells, or more generally
smaller is the particle size. in promoting metal surface catalysis. As
The surfaces of the as-received ma- demonstrated in Sect. 3.1.3.1, interfacial
terials are covered by O and/or OH electrochemistry offers an elegant way to
species, as indicated by the peak at modify electrode surfaces while 195 Pt NMR
1.098 G kHz−1 . Remarkably, these sur- provides a powerful probe to follow the
faces were effectively cleaned simply by variations in the physical properties of
holding the electrode potential within the metal surfaces caused by such modifica-
electrochemical double-layer region. This tions. By using this strategy, the influence
mild surface-reduction procedure should of a series of different ligands (H, O, S,
be contrasted to the rigorous and often CN− , CO, and Ru), electrochemically ad-
technically demanding methods employed sorbed onto carbon-supported nanoscale
at the solid–gas interface, which usually Pt particles from the same starting batch,
involve several cycles of high-temperature has been investigated by 195 Pt NMR, in an
calcination and reduction, limiting the electrochemical environment [10].
choice of catalyst support to, typically, non- The starting electrode material used for
conductive oxides such as alumina, titania, this series of experiments was the same
and silica. as that used in the previous section. As
However, extensive potential cycling, shown there, the frequency difference be-
which is a widely used in electrode-surface tween the resonance positions of surface
cleaning procedure, sinters the nanoscale and bulk platinum atoms at 8.5 T is about
electrocatalyst. Although the sample in 2.5 MHz, large enough to provide a very
Fig. 4(c) started with the same as-received convenient spectral visualization of how
electrocatalyst as that shown in Fig. 4(b), deep the influence of an adsorbate can
the dispersion estimated from the corre- go, as illustrated by the point-by-point
sponding 195 Pt NMR spectrum is only 195 Pt NMR spectra (recorded at 80 K)
about 27% [30], far lower than the 50% presented in Fig. 5. These spectra are nor-
obtained from Fig. 4(b). This dispersion malized by equalizing their amplitudes
Pt/clean
Pt/Ru
Pt/CO
Pt/O
NMR amplitude
Pt/H
Pt/CN−
Pt/S
1.08 1.10 1.12 1.14

Field/frequency
[G kHz−1]
Fig. 5 Superimposition of point-by-point, 8.47 T 195 Pt NMR spectra of
a 2.5 nm, carbon-supported Pt electrocatalyst without and with
different chemisorbed ligand: Ru, CO, O, H, CN− , and S. The spectra
were normalized by equalizing the amplitude at 1.131 G kHz−1
(indicated by the arrow). The invariance of signals beyond
1.131 G kHz−1 provides experimental confirmation of the
Friedel–Heine invariance theorem. The surface peaks range over
∼11 000 ppm. (Reproduced with permission from Ref. [10], Copyright
by  2000 American Chemical Society.)
at 1.131 G kHz−1 , which is indicated by good approximation, do not go beyond

the arrow. The interesting observation their next-nearest neighbors. That is, the
here is that the position of the high-field high-field signals are invariant towards
signal intensity (above 1.131 G kHz−1 ), ‘‘boundary’’ changes.
which is due to Pt atoms within the In complete contrast to the Friedel–
three innermost layers, is independent of Heine invariance of the electronic proper-
adsorbate type. This independence pro- ties observed for the innermost platinum
vides direct experimental evidence for particle layers, the surface and subsur-
the validity of the Friedel–Heine invari- face NMR signals undergo major fre-
ance theorem [32, 33], which states that quency shifts as different chemical species
the integral electronic properties at an are adsorbed, as can be seen in Fig. 5.
atom, such as its LDOS, are determined The NMR layer-model spectral deconvo-
primarily by the surrounding medium, lution [31] technique was applied to these
within a few electronic wavelengths. This spectra in order to obtain the variations
is exactly what is observed – although the in the surface and subsurface Knight
chemical identity of the adsorbates is shifts. Because of the Friedel–Heine
quite varied, their influences, to a very invariance, the NMR parameters (peak
position, width, relative intensity) for the 1/ ln[(K0 − K∞ )/(K1 − K∞ )]. By replac-
three central layers were fixed to the ing K1 and K0 by the correlation lines
clean-surface values, while the surface and shown in Fig. 7,
subsurface peak positions were allowed to
vary. For Pt/O, the position of the third K0 (ppm) = (−10.2 + 5.3χ) × 103 (5)
layer was also varied, as a result of the K1 (ppm) = (−21.5 + 5.8χ) × 103 (6)
long healing length in this system, plus
the surface peak areas were also varied in m (the healing length) can then be
some simulations, as a result of saturation directly expressed as a function of the
effects caused by longer T1 values. The electronegativity χ:
results of these simulations are shown in
1 (23.2 + 5.3χ)
Fig. 6 and the surface and subsurface peak m= (7)
ln (11.9 + 5.8χ)
shift values obtained are plotted in Fig. 7
as a function of the Allred–Rochow elec- In Fig. 8, the healing length (= 0.229 m,
tronegativity (defined as the electrostatic where 0.229 nm is the layer thickness for
force exerted by the nucleus of the atom on Pt particles) is plotted as a function of
its valence electrons) of the bonded adsor- the electronegativity, Eq. (7). This result
bate atom. Remarkably, both the surface shows that the larger the electronegativity
and subsurface peak positions map almost of the adsorbate, the longer the healing
linearly the electronegativity of the adsor- length, that is, the deeper the influence of
bate. The reasons for this are as follows: the adsorbate goes. It is also interesting to
According to the layer model, Eq. (4), one note that the value of the healing length
has K1 = K∞ + (K0 − K∞ ) exp(−1/m) m for hydrogen (χ = 2.2) adsorption
for the subsurface layer, n = 1. A sim- obtained via Eq. (7) is about 2.9, which
ple mathematical conversion gives m = is very close to the m = 2.6 value found
a Pt/clean b Simulation c Pt/Ru d Pt/H

NMR amplitude
e Pt/CO f Pt/CN− g Pt/S h Pt/O
1.08 1.10 1.12 1.14 1.08 1.10 1.12 1.14 1.08 1.10 1.12 1.14 1.08 1.10 1.12 1.14
Field/frequency
[G kHz−1]
Fig. 6 The NMR layer-model simulations of the adsorbed H; (e) Pt with adsorbed CO; (f) Pt with
195 Ptspectra shown in Fig. 5 and the small dots adsorbed cyanide; (g) Pt with adsorbed S (from
are the variances between the experimental and Na2 S solution); (h) Pt with adsorbed O.
fitted values. (a) clean-surface Pt; (b) Simulation (Reproduced with permission from Ref. [10],
of (a); (c) Pt with adsorbed Ru; (d) Pt with Copyright by American Chemical Society 2000.)
Surface
0
d (× 103)
[ppm]
−10
Subsurface
−20
Pt bulk shift
−30
1.5 2.0 2.5 3.0 3.5

Allred–Rochow electronegativity
Fig. 7 Correlation between surface/subsurface frequency shifts
(with respect to the Pt NMR reference H2 PtCl6 ) and the
Allred–Rochow electronegativity. The dashed horizontal line
indicates the Knight shift of bulk platinum atoms. The solid
straight lines are linear fits to the surface and subsurface shifts
as a function of the electronegativity. Both have R2 values of
ca. 0.92. (Reproduced with permission from Ref. [10], Copyright
by American Chemical Society 2000.)
0.9
0.8
0.7
0.6
Li H
0.5
K
0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0

Electronegativity
Fig. 8 Solid curve showing the relationship between the
healing length and electronegativity as determined from
Eq. (7). For comparison, we also show the three experimental
points based on previous observations. See the text for
details. (Reproduced with permission from Ref. [10], 
Copyright by American Chemical Society 2000.)
by a detailed layer-model analysis [31]. has been applied to investigate the

In addition, for the adsorption of the metal–CO interactions in a number of
alkali metal elements (χ ∼ 1), Eq. (7) recent studies [25].
predicts m ∼ 2, virtually the same m as The discussions in the two previous
obtained for clean-surface Pt particles in an sections demonstrate that there is an un-
electrochemical environment. One would, ambiguous NMR discrimination between
therefore, expect that alkali adsorption the surface peak (at 1.100 G kHz−1 for
would not significantly change the 195 Pt nanoscale platinum particles) and the rest
NMR spectrum of small Pt particles after of the platinum sites. By carrying out
adsorption. This is exactly what has been detailed temperature-dependent T1 relax-
observed experimentally (in a gas-phase ation measurements on a given Pt sample,
system) [34]. For comparison, these three a value for S(T1 T )−1 can be obtained,
experimentally estimated points are also and by using the two-band model (see
plotted in Fig. 8. Sect. 3.1.2.1), a pair of surface Pt s- and
The invariance of the frequency shifts d-D(EF )s can therefore be deduced. Sim-
of the more buried (deeper than the ilarly, S(T1 T )−1 and K measurements on
13 CO adsorbed onto the metal surfaces
third layer) Pt atoms to surface adsor-
bate electronegativity provides strong ev- can give the corresponding 13 C 5σ and
idence demonstrating the applicability of 2π ∗ (EF )s (see Sect. 3.1.3.4).
Friedel–Heine invariance of the EF -LDOS It is of fundamental interest to see if
in these nanoscale electrocatalyst systems. there exists any type of correlation be-
This is a rather interesting extension tween the clean-surface EF -LDOS of the
of the original Friedel–Heine invariance metal and the 13 C Knight shift of the
theorem to nanoscale systems and may chemisorbed CO. The former is an impor-
have utility in describing the electronic tant surface attribute defining the ability
surface structure of metal catalysts in of the metal surface to donate (to act as
different chemical environments. These a HOMO) as well as to accept (to act
results may also be expected to lead to as a LUMO) electrons, while the latter
useful general correlations between elec- measures, at least to a certain extent, the
tronic properties and more conventional degree of metallization of chemisorbed CO
chemical descriptors (such as ligand elec- (see Sect. 3.1.3.4). We show in Table 1, a
tronegativity), which could be helpful in set of data for atop CO on Pt and Pd,
understanding the electronic structure of shown graphically in Fig. 9. The purely
metal–adsorbate interfaces by providing orbital contributions (chemical shifts) for
guidelines for engineering new electrode atop CO on Pt and on Pd were calculated
surfaces. by using density functional theory (DFT)
calculations on model COPt7 and COPd7
3.1.3.3 Correlation between the clusters in which CO sits atop the central
Clean-surface EF -LDOS of Metals and the metal atom, which is coordinated by the
Knight Shift of Adsorbates remaining six metal atoms. For COPt7 ,
Because NMR has elemental specificity a value of 160 ± 20 ppm from tetram-
and is also nondestructive, it offers unique ethylsilane (TMS) was obtained, while
opportunities for investigating electro- for COPd7 , the value was 203 ± 20 ppm.
chemical problems from both sides of These calculated values are in good agree-
the electrochemical interface. This strategy ment with available experimental data
Tab. 1 Correlation between clean-surface EF -LDOS and 13 C shift of chemisorbed

CO [25]. The Knight shift = total shift − chemical shift
Sample Clean-surface EF -LDOS Total 13 C shift of COchemisorbed

(Ry−1 atom−1 ) (ppm w.r.t TMS)
PtCO7 0 160
CO–Pt/oxides (dry) 14.8 330
CO–Pt/carbon (wet) 17.7a 351a
CO–Pt/carbon (wet) 18.6 383
PdCO7 0 203
CO–Pd/oxides (dry) 24 500
a For an 8.8-nm sample.
500
400
[ppm]
Shift
300
200
0 5 10 15 20 25 30
Clean surface LDOS (Ry−1 atom−1)
Fig. 9 Correlation between the Fermi level total electronic
densities of states at transition metal (Pt and Pd) surfaces (in
Ry−1 atom−1 ) and the corresponding 13 C Knight shift of
chemisorbed CO. (Partially reproduced with permission from
Ref. [25],  Copyright by American Chemical Society.)
while the standard deviations account for clean-surface EF -LDOS before adsorption.
uncertainties caused by the use of dif- The straight line is a linear fit to the data,
ferent functionals and basis sets. The giving a slope of 12 ppm/Ry−1 atom−1 .
clean-surface EF -LDOS for Pd was esti- This linear relationship can be readily
mated on the basis of the changes in related to the frontier orbital interpretation
magnetic susceptibility of small Pd par- of the Blyholder model [35] – a higher
ticles with respect to their bulk value, clean-surface EF -LDOS means more metal
by application of the exponential heal- electrons and holes are available to engage
ing model. in 5σ -forward and 2π ∗ -back donation.
As shown in Fig. 9, the 13 C NMR As the EF -LDOS increases, CO becomes
shift of CO responds linearly to the more metallic, which results in a larger
Knight shift. A correlation very similar in Fig. 10 is a typical 13 C NMR spec-

in nature to this has also been found to trum and temperature-dependent T1 data
exist in the ‘‘gas phase’’, in which the for the 8.8-nm sample [10]. The straight
C–O vibrational stretch frequency (after line through the origin is characteristic of
chemisorption) was shown to correlate the Korringa relationship, indicating the
linearly with the Pt clean-surface EF -LDOS metallic state of adsorbed CO in this elec-
before chemisorption [36]. The higher the trocatalyst system.
surface EF -LDOS, the lower the C–O Table 2 collects all of the 13 C NMR
stretching frequency. results together with values of the 5σ -
The linear relationship shown in Fig. 9 and 2π ∗ -D(EF ) deduced from these re-
is important because it demonstrates the sults, in addition to the corresponding
validity of the frontier orbital interac- infrared data [37] (vide infra). The results
tion picture of metal surface chemistry, shown in Table 2 indicate that the ma-
in which the importance of the clean- jor part of the total EF -LDOS at 13 C is
surface EF -LDOS is highlighted. In ad- from the 2π ∗ -like electrons, with 2π ∗ -
dition, it puts 13 C NMR spectroscopy D(EF ) being about 10 times larger than
of chemisorbed CO on a firmer foot- the 5σ − D(EF ). These values are in good
ing, by probing the electronic properties agreement with theoretical band struc-
of transition metal surfaces before CO ture calculations. The contribution to the
chemisorption. spin–lattice relaxation rate (1/T1 ) is dom-
inated by orbital and dipolar interactions,
consistent with the dominance of π-like
3.1.3.4 Coupling EC-NMR with In Situ
electrons at the Fermi level. Notice that
Infrared Spectroelectrochemistry
while 5σ − D(EF ) is almost a constant,
It is now clear that CO acquires metallic
the 2π ∗ -D(EF ) varies noticeably from sam-
properties upon adsorption onto transition
ple to sample. However, since NMR only
metal surfaces, as indicated by the pres-
measures the electron density at a given
ence of a Knight shift and a Korringa energy level, that is, the Fermi level, it does
relationship, measured via 13 C NMR. not provide complete information about
Furthermore, by using the experimen- the total electron densities for forward-
tally measured 13 C Knight shift and the
Korringa constant, a phenomenological Tab. 2 The 5σ - and 2π ∗ -D(EF )s deduced by the
two-band model [10] permits a quantita- two-band model from 13 C NMR data of CO
tive partitioning between the 5σ - and chemisorbed on Pt electrocatalysts [10] and the
2π ∗ -D(EF ) at 13 C to be obtained, pro- corresponding SNIFTIRS C–O vibrational
stretching frequencies [37]
viding insight into metal–CO bonding at
the electronic level. These types of 13 C
Size 5σ -D(EF ) 2π ∗ -D(EF ) IR
NMR study have been carried out on 13 CO (nm) (Ry (Ry frequency
chemisorbed (ex MeOH) onto a series of molecule)−1 molecule)−1 (cm)−1
carbon-supported, fuel cell grade, com-
mercial platinum electrocatalysts having 8.8 0.6 6.4 2044
different average particle size: 2.0, 2.5, 3.9 0.6 6.6 2043
3.2 0.6 6.5 2044
3.2, 3.9, and 8.8 nm, in which both metal
2.5 0.7 6.8 2038
and adsorbation EF -LDOS might be ex- 2.0 0.6 7.3 2028
pected to vary with particle size. Shown
4
Pt/13CO
(s−1)
1/T1
Carbon support
0
0 200
Temperature
[K]
1400 1200 1000 800 600 400 200 0 −200

d (ppm from TMS)
Fig. 10 A typical 13 C NMR spectrum and temperature-dependent T1 data
for an 8.8-nm Pt/C sample. The straight line through the origin is
characteristic of the Korringa relationship, indicating the metallic state of
adsorbed CO in this electrocatalyst system. (Adapted from Ref. [10].)
and backdonation, which are the inte- yield a linear correlation between νCO and
grals of the LDOS from the bottom of the 2π ∗ -D(EF ), Table 2, with a slope of
the conduction band to the Fermi level. −19 ± 2 cm−1 /(Ry molecule)−1 and an
Nevertheless, as can be seen in Table 2, intercept of 2167 ± 20 cm−1 , as shown
the variation in the total EF -LDOS bein Fig. 11. The latter value is quite close
tween different samples are mainly due to to the value expected for free CO, ν =
changes in the 2π ∗ -D(EF ). This suggests 2143 cm−1 . On the basis of these results,
that the changes in chemical properties the following overall physical picture of
of adsorbed CO are mainly determined carbon monoxide adsorbed onto a plat-
by variations in backbonding. In particu- inum electrode can be proposed: Since the
lar, if the Fermi level cuts the tail of the 5σ orbital of CO is essentially nonbonding,
2π* band as it rises, then the increase in with the lone pair concentrated on carbon
the 2π ∗ -D(EF ) would indicate an enhance- and the 2π ∗ orbital is a much less polar-
ment in backdonation, and a decrease in ized antibonding orbital, the change in CO
the corresponding vibrational C–O stretch stretch frequency is mainly governed by
frequency. changes in 2π ∗ backdonation: the higher
In situ subtractively normalized interfa- the backdonation, the lower the CO stretch
cial Fourier transform infrared reflectance frequency.
spectroelectrochemistry (SNIFTIRS) stud- The above example demonstrates the
ies confirm this prediction [37]. They also complementary nature of in situ infrared
Fig. 11 Correlations between νCO and 2050

2π ∗ -D(EF ). The straight line is the linear
fit that gives a slope of
−19 cm−1 /Ry−1 molecule−1 and an
intercept of 2167 cm−1 , with a 2040
R2 = 0.98. (Part of the results shown in
(cm−1)
this figure has been published in
Refs. [10, 37].)
2030
2020
6.0 6.5 7.0 7.5 8.0
D2p*(Ef)
[Ry−1 molecule−1]
spectroscopy and solid-state NMR, in pro- 3.1.3.5 Surface Diffusion

viding insights into the details of surface The rationale behind using NMR to study
chemistry of electro-chemisorbed CO on surface diffusion is that the correlation
platinum. While SNIFTIRS – a major re- time for diffusion, τ , can be related to the
search technique of the interfacial elec- temperature dependence of the NMR re-
trochemical community – provides impor- laxation rates observed for diffusing atoms
tant qualitative information regarding the or molecules [12]. Under certain circum-
bond strengths of adsorbed CO, it offers stances, such a relation can be expressed
somewhat less insight into the electronic analytically, permitting a quantitative anal-
relocations that occur within the bond, the ysis of the data in dynamic and thermody-
electron distribution as a function of sam- namic terms.
ple composition and morphological detail. When using the temperature depen-
For instance, how the bond changes as a dence of the nuclear spin–spin or nuclear
function of particle size is difficult to learn spin–lattice relaxation rate to study molec-
using infrared alone. However, a 13 C NMR ular motion, as is the case with the
analysis of the 5σ - and 2π ∗ -EF -LDOS at surface diffusion we are dealing with here,
the carbon atom offers a quantitative de- there exist so-called ‘‘strong’’ and ‘‘weak’’
scription of metal–CO bonding in terms collision limits. Different mathematical re-
of changes in the EF -LDOS. The joint lationships are needed to describe these
SNIFTIRS/NMR approach provides new limits. Consider the nuclear spin–spin
information on the electronic structure relaxation rate (1/T2 ) as measured by a con-
of the metal–solution interface, which ventional Hahn-echo pulse sequence, and
is relevant to the characterization of in- suppose that ω is the amplitude of the
dustrial fuel cell catalysts, as well as local field fluctuation responsible for relax-
other interfacial electrochemical systems ation. Also assume that τ is the correlation
of practical importance, by combining time for the motion, say a jump, which
bond strength (IR) and quantitative EF - causes the local field to fluctuate. The
LDOS results (NMR). strong collision limit is defined such that
τ
1/ω. At t = 0, that is, just after the platinum particles, 1/T2 can be ex-
first π/2 pulse, which flips all nuclear spins pressed as
into the xy plane, all spins are in phase, and
1 1 1 1 1
they retain their phase memory until t = τ , = + + +
T2 T2 RL T2 Pt T2 dip T2 diff
when the collision changes the direction of
the local field. Since τ is sufficiently long 1 (ω)2 τ 1
= RL
+aT + +
to dephase the spins, they lose all phase T2 (1+(ω)2 τ 2 ) (N τ )
memory before the next collision. There- (11)
fore, 1/T2 = 1/τ . In contrast, in the weak where 1/T2 RL is the temperature-indepen-
collision limit, τ 1/ω. Since τ is rela- dent ‘‘rigid lattice’’ contribution, and
tively short, insufficient dephasing builds 1/T2 Pt (= aT , where a is a constant
up between two consecutive jumps. When and T the absolute temperature) ac-
the spin jumps randomly, many jumps are counts for the effects of platinum spins.
needed in order to accumulate sufficient 1/T2 dip = (ω)2 τ/(1 + (ω)2 τ 2 ), where
dephasing, and one has 1/T2 = (ω)2 τ . ω is the dipolar field in Hertz and
In summary then, we have for these two τ = τ0 exp(E/kB T ), with τ0 a preexponen-
limiting cases: tial factor, chosen to be 10−13 s, and E
is the activation energy for surface dif-
1 1 1 fusion), arises from dipolar interactions
= for τ
(8) between neighboring carbon- 13 spins.
T2 τ ω
1/T2 diff (= 1/N τ , where N is the number
and of jumps needed to dephase the carbon
spins) comes from the field difference
1 1 (positive or negative) between adjacent ad-
= (ω)2 τ for τ (9) sorption sites, due to the presence of a
T2 ω
significant magnetic field inhomogeneity
Fortunately, the function on the particle surface.
The solid lines in Fig. 12 are the fit to
Eq. (11) with 1/T2 RL , a, ω, E, and N as
1 (ω)2 τ
= (10) fitting parameters. The results of the fits
T2 (1 + (ω)2 τ 2 ) (in which 1/T2 RL are 293 and 930 s−1 for
CO and CN, respectively) are also shown
satisfies both limiting cases. In fact, in Fig. 12. Clearly, this simple model de-
Eq. (10) is similar to the typical spectral scribes the experimental data very well, for
density function used for a randomly both 13 CO and 13 CN. It is interesting to
fluctuating local field at a frequency ω, note that the activation energy found for
and as we show below, the use of these CO (7.8 kcal mol−1 ) is close to that found
simple relations leads to information on for CO on ‘‘dry’’ alumina-supported Pt
both the rates of surface diffusion and the clusters, as well as that found in single-
associated energetics. crystal studies. The results of the fits also
The temperature dependence of 1/T2 indicate that the activation energy for sur-
for 13 CN (a) [6] and for 13 CO (b) [5] on face diffusion is higher for CN than for
polycrystalline platinum black at satu- CO, while the difference in the constant a
ration coverage are shown in Fig. 12. suggests, paradoxically, that platinum has
For CO or CN adsorbed onto small more of an electronic influence on 13 C
Fig. 12 Temperature dependence of

spin–spin relaxation rates of 1600
a
chemisorbed and (panel a) 13 CN and Pt/13CN
(panel b) 13 CO on 10-nm Pt E = 10.3 kcal/mol
electrocatalyst surfaces. The solid lines 1400 a = 0.8 s−1 •K−1
are the fits to Eq. (11) and the numbers ∆w = 966 Hz
are the corresponding results of the fits. N = 1222
(Adapted from Refs. [5, 6].) 1200
1000
(s−1)
1/T2
b
2500 Pt/13CO
E = 7.8 kcal/mol
a = 5.4 s−1 •K−1
2000 ∆w = 1053 Hz
N = 977
1500
1000
500
100 200 300
Temperature
[K]
spin–spin relaxation of CO than on that with a similar adsorbate coverage for

of CN. The latter may, however, be related both samples.
to the fact that the J -coupling is medi-
ated by orbital electrons between 13 C and 3.1.3.6 Isotopic Labeling for Site
195 Pt and is expected to be primarily influ-
Selectivity
enced via σ -bonding rather than through Isotopic labeling is a very useful and
π-bonding, for |ψ|2 = 0 at C for the 2π ∗ geometrically specific NMR technique
orbital. If the 195 Pt– 13 C dipolar interaction when the natural abundance of the nucleus
were the only mechanism for 1/T2 Pt , the of interest is too low to be detected
ratio in a values would yield rPt−CN /rPt−CO under normal conditions, such as with
equal to 1.9, which is unrealistic. Thus, 13 C, 15 N, and 17 O. As an example,
while the activation values may indicate an we consider here the bonding of a
overall stronger metal surface bonding for larger species, acetonitrile, bonded to Pt
Pt–CN than for Pt–CO, the a values sug- in an electrochemical environment [40].
gest that σ -bonding is stronger in the latter 13 C 1/T versus T data for CH 13 CN
1 3
case. Finally, ω for the 13 C– 13 C dipolar and 13 CH3 CN adsorbed on platinum are
interaction is essentially identical in each shown in Fig. 13. A linear 1/T1 versus
of the systems investigated, consistent T plot, reflecting Korringa behavior with
Pt-CN13CH3
2.0
Pt-13CNCH3
1.5
(s−1)
1/T1
1.0
0.5
0.0
0 50 100 150 200 250 300
Temperature
[K]
Fig. 13 Temperature dependence of spin–lattice relaxation rates
for chemisorbed CH3 13 CN and 13 CH3 CN. The clear absence of a
Korringa relationship for 13 CH3 CN (solid circles) indicates that the
carbon atom in the methyl group does not acquire any metallic
character. (Adapted from Ref. [38].)
T1 T = 170 s K, is only obtained with the any metallic character. This implies that
CH3 13 CN adsorbate. This proves that at neither C–C bonding nor antibonding
least the carbon of the CN group is orbitals are involved with bonding to the
directly attached to platinum, and that platinum surface, while both 13 C and
the 13 C has acquired metallic character. 15 N atoms in the CN group are strongly
Note that the value of T1 T = 170 s K for involved, in contrast to the situation with
CH3 13 CN is quite close to that of 13 CN the CN itself, in which cyanide has a C-
(135 s K). The T1 T product for 15 N in down on-top bonding.
CH3 C15 N at room temperature is 224 s K,
showing that the spin–lattice relaxation
3.1.3.7 Potential-dependent EC-NMR
of 15 N in chemisorbed CH3 C15 N is
much more efficient than in chemisorbed Understanding the nature of the electric
C15 N, where T1 T varied in the range field (electrode potential) effects on the
∼3000–4000 s K. Indeed, after taking into electronic structure at the solid–liquid
account the γ13C /γ15N ∼ 2.5 factor in interface is an outstanding issue in
Eq. (3), the relaxation of 15 N appears even electrocatalysis and in the theory of
more efficient than that of the ‘‘metallic’’ the electrical double layer. To illustrate
13 C in CH 13 CN. This suggests that both
3
such effects via NMR, we show in
C and N in the CN group are directly Fig. 14, the electrode potential–induced
13 C line shifts for CO (circles) [8] and
attached to the platinum surface, forming a
side-on parallel orientation with respect to CN (squares) [6] on polycrystalline Pt.
the Pt–Pt surface bonds. The clear absence These results were obtained under active
of a Korringa relationship for 13 CH3 CN external potentiostatic control, and at room
(solid circles) indicates that the carbon temperature, and the inset shows typical
atom in the methyl group does not acquire 13 C NMR spectra of 13 CN, recorded at
Fig. 14 Electrostatic potential 500

dependence of 13 C shifts for
13 CO (circles) and 13 CN
450
(squares) chemisorbed on a
10-nm Pt electrocatalyst.
Shift
(Adapted from Ref. [4].) 400
13CO, 13CN
[ppm]
13
350 CO
13CN
200
150
−0.8 −0.4 0.0 0.4

Potential
[V]
different applied potentials. Apparently, potential has been found to be due

for both adsorbates, there is a linear essentially to the Knight shift [8]. The
relation between the electrode potential 13 C chemical shift response of CO
and the line shift – the more negative to electric fields (in biomolecules) has
the potential, the larger the frequency been estimated to be at most about
shift. 13 C becomes more deshielded as the −15 ppm V−1 , much smaller than the
potential goes to more negative values, and values found at the electrochemical
the slope is about −71 ppm V−1 for 13 CO interface (see above). In addition, for
and −50 ppm V−1 for 13 CN. For CO on CO on Pd, a corresponding change in
Pd, the slope was found to be even larger, the nuclear spin–lattice relaxation rate
that is, about −136 ppm V−1 [8]. as a function of applied field has also
A similar response in C–O vibrational been observed [8], indicating the electronic
frequency, νCO , to variations in the elec- nature of the change in NMR shift.
trode potential has been reported [39]. Second, the magnitude of the clean-
There, νCO decreased when the electrode surface EF -LDOS of the adsorbent is
potential became more cathodic. However, responsible for the extent of the Knight
the origin of this so-called vibrational Stark shift as well as the vibrational stretch
effect has been controversial for some frequency of chemisorbed CO [25, 36].
time, since a clear picture of how the That is to say, the higher the clean-
short-range (electronic) bonding and the surface EF -LDOS, the larger (lower) will
longer-range (electrostatic) field effects at be the 13 C Knight shift [25] (C–O stretch
surfaces are controlled by varying either frequency [36]) of CO, after chemisorption.
the electrode or the surface potential, has Both relationships are linear, with
not been available. slopes of about 12 ppm/Ry−1 ·atom−1
EC-NMR provides an additional and −4 cm−1 /Ry−1 ·atom−1 , respectively,
technique with which to probe these giving a ratio of −3 ppm/cm−1 . This
questions, for several reasons. First, the is to be compared with the ratio of
response of 13 C NMR to the electrode −2.8 ppm/cm−1 , obtained independently
from the slopes (–71 ppm V−1 and Acknowledgments

25 cm−1 V−1 , respectively) of δ(13 C)
and νCO versus electrode potential This work was supported by the United
relationships for CO on Pt black. Third, States National Science Foundation (grant
as shown in Sect. 3.1.3.4, variations in νCO CTS 97-26419), by an equipment grant
can be directly correlated with changes in from the United States Defense Advanced
2π ∗ backdonation – that is, the higher the Research Projects Agency (grant DAAH
backdonation, the lower the CO stretch 04-95-1-0581), and by the US Depart-
frequency. When taken together, all these ment of Energy under grant DEFG02-
observations indicate that the potential 96ER45439.
dependence of both the 13 C NMR shift
and the vibrational stretch frequency of References
adsorbed CO are primarily electronic in
nature, and originate from changes in 1. A. Zangwill, Physics at Surfaces, Cambridge
the EF -LDOS at the metal surface and
2. G. A. Somorjai, Introduction to Surface Chem-
at the adsorbate, induced by electrode istry and Catalysis, John Wiley & Sons, New
polarization. York, 1994.
3. A. Wieckowski, (Ed.), Interfacial Electrochem-
3.1.4 istry – Theory, Experiment, and Applications,
Future Perspectives Marcel Dekker, New York, 1999.
4. Y. Y. Tong, E. Oldfield, A. Wieckowski, Anal.
Chem. 1998, 70, 518A–527A.
EC-NMR has made considerable progress 5. J. B. Day, P.-A. Vuissoz, E. Oldfield et al., J.
during the past few years. It is now pos- Am. Chem. Soc. 1996, 118, 13 046–13 050.
sible to investigate in detail metal–liquid 6. J. J. Wu, J. B. Day, K. Franaszczuk et al.,
interfaces under potential control, to de- J. Chem. Soc., Faraday Trans. 1997, 93,
1017–1026.
duce electronic properties of electrodes
7. Y. Y. Tong, C. Belrose, A. Wieckowski et al.,
(platinum) and of adsorbates (CO), and J. Am. Chem. Soc. 1997, 119, 11 709, 11 710.
to study the surface diffusion of ad- 8. P.-A. Vuissoz, J.-P. Ansermet, A. Wiecko-
sorbates. The method can also provide wski, Phys. Rev. Lett. 1999, 83, 2457–2460.
information on the dispersion of com- 9. Y. Y. Tong, C. Rice, E. Oldfield et al. in The-
oretical Modeling of the Solid/Liquid Interface:
mercial carbon-supported platinum fuel
Electronic Perspective and Comparison with Ex-
cell electrocatalysts and on electrochem- periment (Ed.: W. Halley), American Chemi-
ically generated sintering effects. Such cal Society, Washington, DC, 2001.
progress has opened up many new re- 10. Y. Y. Tong, C. Rice, A. Wieckowski et al., J.
search opportunities since we are now Am. Chem. Soc. 2000, 122, 1123–1129,
11 921–11 924.
in the position to harness the wealth of
11. M. E. Lacey, R. Subramanian, D. L. Olson
electronic, EF -LDOS as well as dynamic et al., Chem. Rev. 1999, 99, 3133–3152.
and thermodynamic information that can 12. C. P. Slichter, Principles of Magnetic Reso-
be obtained from NMR experiments. As nance, 3rd Enlarged and Updated Edition,
such, it is to be expected that EC-NMR Springer-Verlag, Berlin, Germany, 1990.
13. J.-P. Ansermet, C. P. Slichter, J. H. Sinfelt,
will continue to thrive and may even-
Prog. NMR Spectrosc. 1990, 22, 401–421.
tually become a major characterization 14. T. M. Duncan, Colloids Surf. 1990, 49, 11–31.
technique in the field of interfacial elec- 15. M. Pruski in Encyclopedia of Magnetic Res-
trochemistry. onance (Eds.: D. M. Grant, R. K. Harris),
John Wiley & Sons, New York, 1996, pp. 29. J. J. van der Klink, J. Buttet, M. Graetzel,
4638–4638, Vol. 7. Phys. Rev. B 1984, 29, 6352–6355.
16. J. J. van der Klink, Adv. Catal. 2000, 44, 30. C. Rice, Y. Y. Tong, E. Oldfield et al., Elec-
1–117. trochim. Acta 1998, 43, 2825–2830.
17. P.-K. Wang, J.-P. Ansermet, S. L. Rudaz 31. J. P. Bucher, J. Buttet, J. J. van der Klink
et al., Science 1986, 234, 35–41. et al., Surf. Sci. 1989, 214, 347–357.
18. P. J. Slezak, A. Wieckowski, J. Magn. Reson., 32. J. Friedel, Adv. Phys. 1954, 3, 446–507.
Ser. A 1993, 102, 166–172. 33. V. Heine in Solid State Physics (Eds.:
19. W. D. Knight in Solid State Physics (Eds.: H. Ehrenreich, F. Seitz, D. Turnbull), Aca-
F. Seitz, D. Turnbull), Academic Press, New demic Press, New York, 1980, pp. 1–127,
York, 1956, pp. 93–136, Vol. 2. Vol. 35.
20. J. J. van der Klink, H. B. Brom, Prog. NMR 34. Y. Y. Tong, G. A. Martin, J. J. van der Klink,
Spectrosc. 2000, 36, 89–201. J. Phys. Condens. Matter 1994, 6, L533–L538.
21. J. Korringa, Physica 1950, 16, 601–610. 35. G. Blyholder, J. Phys. Chem. 1964, 68,
22. J.-P. Bucher, J. J. van der Klink, Phys. Rev. B 2772–2778.
1988, 38, 11 038–11 047. 36. Y. Y. Tong, P. Mériaudeau, A. J. Renouprez
23. R. Hoffmann, Rev. Mod. Phys. 1988, 60, et al., J. Phys. Chem. B 1997, 101,
601–628. 10 155–10 158.
24. P.-A. Vuissoz, J.-P. Ansermet, A. Wiecko- 37. C. Rice, Y. Y. Tong, E. Oldfield et al., J. Phys.
wski, Electrochim. Acta 1998, 44, 2457–2460. Chem. B 2000, 104, 5803–5807.
25. Y. Y. Tong, C. Rice, N. Godbout et al., J. Am. 38. J. Wu, C. Coretsopoulos, A. Wieckowski in
Chem. Soc. 1999, 121, 2996–3003. Electrochemical Society Proceedings, The Elec-
26. M. S. Yahnke, B. M. Rush, J. A. Reimer et al., trochemical Society, Pennington, NJ, 1997,
J. Am. Chem. Soc. 1996, 118, 12 250, 12 251. pp. 426–442, Vol. 97-17.
27. B. M. Rush, J. A. Reimer, E. J. Cairns, J. Elec- 39. D. K. Lambert, Electrochim. Acta 1996, 41,
trochem. Soc. 2001, 148, A137–A148. 623–630.
28. H. E. Rhodes, P.-K. Wang, H. T. Stokes et al., 40. M. Weinert, A. J. Freeman, Phys. Rev. B 1983,
Phys. Rev. B 1982, 26, 3559–3568. 28, 6262–6269.
3.2 EPR spectroscopy was first applied to

EPR Spectroscopy in Electrochemistry electrochemistry, albeit at low tempera-
ture, in 1958 by Ingram and coworkers [3],
Jay D. Wadhawan, Richard G. Compton,∗ who demonstrated the formation of radical
Oxford University, Oxford, United Kingdom ion species from the potentiostatic elec-
3.2.1 trolysis of various aromatic compounds in
Introduction N ,N -dimethylformamide (DMF). In these
experiments, aliquots of sample were with-
Electron paramagnetic resonance (EPR), drawn during the course of electrolysis,
or its equivalent term electron spin res- frozen in liquid nitrogen, and subse-
onance (ESR), is an ideal spectroscopic quently examined by EPR. It was Maki and
tool to study the formation of para- Geske [4, 5, 7, 8] who developed a comple-
mentary concept of electrochemical EPR
magnetic species (such as radicals) dur-
by studying the solution-phase spectra of
ing the course of electrolysis, as species
radicals that had been electrogenerated
with concentrations as low as 10−8 M
within the EPR cavity, thereby paving the
may be detected [1]. Further, electroly-
way for two forms of electrochemical EPR
sis has consistently proved to be one of
methodology, via in situ techniques and
the most convenient and attractive tech-
ex situ techniques. In the former, param-
niques for the production of interesting
agnetic species are generated within the
radicals [2–5]. Additional advantages of
EPR cavity, whilst in the latter, radicals
using a spectroscopic approach in the in-
that are externally generated are subse-
terrogation of complex electrode reaction
quently moved into the EPR cavity. It will
mechanisms are that it does not require be appreciated that the former technique
the destruction (nor activation) of the sam- is more advantageous than the latter, in
ple, and that it permits an inference of that it may be used in the study of radi-
the identity of paramagnetic species that cals with a shorter lifetime range. Whist
are formed as intermediates in a plethora the ex situ technique has many merits,
of electron-transfer processes. An example notably in Albery’s use of a tube electrode
that illustrates this is the electrochemi- with a mathematically well-defined flow of
cal reduction of 2,4-dichloronitrobenzene, electrolytes in the study of the kinetics
studied by Lawless and Hawley [6]. In this and mechanisms of a variety of elec-
reaction, the 2,4-dichloronitrobenzene rad- trode processes [9–14], this overview will
ical anion undergoes a thermally-induced be exclusively concerned with the more
C−Cl bond cleavage to form the neu- sensitive in situ approach to electrochemi-
tral chloronitrobenzene radical, which, cal EPR; the interested reader is referred to
after H atom abstraction from the sol- two earlier reviews by Waller and Comp-
vent/electrolytic system, is subsequently ton [15, 16] that describe the historical
further reduced. The question that arises development of electrochemical EPR.
is: which of the two C−Cl bonds is The ability of the in situ electrochemi-
cleaved? Analysis of the EPR spectrum cal EPR technique to provide information
of the intermediates revealed that the on kinetics and mechanisms of elec-
ortho-chlorine is removed, leaving the para- trode processes is the major reason for
chloronitrobenzene radical anion that is its adoption by electrochemists although
detected. its major limitation is that it requires
3.2 EPR Spectroscopy in Electrochemistry 171
the presence of a paramagnetic reac- upon passage through an inhomogeneous

tion intermediate. Following Maki and magnetic field.
Geske, Adams [17], Cauquis [18–22], Kas- The electronic charge (e) together with
tening [23, 24], and Dohrmann [25–28] all the angular momentum (s) confer a mag-
developed cells based upon a variety of cell netic moment (µ) upon the electron, ac-
geometries, but these did not provide ad- cording to Eq. (1),
vances in the study of reaction kinetics;
Sect. 3.2.3 covers the in situ electrochem- µ = ge γe s (1)
ical EPR cell designs that aim to remedy
this, and Sect. 3.2.4 gives examples of the in which γe is the magnetogyric ratio of
use of such cells in the analysis of kinetics the electron, given by the ratio −e/2me ,
and mechanisms. First, however, the un- me representing the mass of the electron.
derlying theoretical principles of electron The constant ge is referred to as the
paramagnetic resonance are covered, and electron g-factor and takes the value
from which, the similarity between this 2.002319314 for a free electron. The origin
and its sister technique, nuclear magnetic of this factor may be inferred from Dirac
resonance (NMR, see Chapter 3.1 of this theory; the deviation from exactly 2 reflects
volume), may be noted. the existence of a zero-point energy in
the harmonic oscillating electromagnetic
3.2.2 field. The interested reader is referred
Principles of EPR Spectroscopy to Ref. [31] for further details regarding
these phenomena.
This section outlines the theoretical con- The magnitude of the spin angular
cepts of EPR. The theory is similar to momentum vector (s) is given by
the related technique of NMR described
in Volume 2 Chapter 3.1. In the fol- |s| = h̄ s(s + 1) (2)
lowing, a basic knowledge of quantum
in which s refers to the spin angular mo-
mechanics will be assumed. Throughout
mentum quantum number. Experiment
the following, the symbol B refers to the
shows that s = 1/2, and the spin angular
magnetic flux density (or magnetic induc-
momentum vector is perceived to ‘‘pre-
tion field); the symbol H being reserved
cess’’ around the direction of the applied
for the magnetic field strength. In SI units,
magnetic field (conventionally labeled the
H = 4π × 10−7 B.
z-direction); the component of the spin
angular momentum in this direction be-
3.2.2.1 EPR Theory ing ± h̄/2 (Fig. 1). It is useful to introduce
The origin of EPR spectroscopy lies in the the quantum number Ms to label the al-
fact that electrons have both electrostatic lowed eigenvalues of sz ,
charge and spin angular momentum. The
former is observed by electron-deflection sz = Ms h̄ (3)
from negatively charged surfaces, in for
example, a cathode ray oscilloscope; the Evidently, Ms = ±1/2. This space quanti-
latter is easily verifiable in the clas- zation of s and µ will necessarily cause
sic Stern–Gerlach experiment [29, 30] in the two possible electron energy states to
which a beam of S-state silver atoms is split under the application of a magnetic
observed to split into two separate beams field (the Zeeman Effect); the energy, E
z Fig. 1 The precession of the

spin angular momentum vector
about a magnetic field applied
in the z-direction.
1h
Ms = + 1 2
2
Ms = − 1
2
of a magnetic dipole of moment µ in the ge γe Ms Bz

E= (5)
presence of a magnetic induction, B, is h̄
given by
E = −µ • B (4) Since Ms = ±1/2, only two energy states
are possible as given above. In the absence
The negative sign indicates that if the of a magnetic field, the two energy states
magnetic dipole is aligned parallel to are necessarily degenerate; application of
the applied field, the interaction energy a magnetic field causes the spin states
will be at a minimum; an antiparallel to adopt different energies, illustrated in
alignment corresponds to a maximum Fig. 2. Note that the lower state (E1 ) has
energy interaction. If the field is applied in Ms = −1/2, whilst that antiparallel, upper
the z-direction, Eq. (4) becomes state (E2 ), has Ms = +1/2. If the energy
Ms = + 1
2
2phn = gegeBZ
B=0
BZ Ms = − 1
2
B
dA
dB
Fig. 2 Splitting of the spin levels of a

free electron in a magnetic field, of
induction Bz . Also shown is the first
derivative of the absorption intensity, A,
∆Bpp
with respect to B, as a function of B.
of the interaction is large when compared of recording the spectrum depends upon
to kT, the alignment of the great majority the modulation used to amplify the absorp-
of dipoles would correspond to the lower tion signal (vide infra). The peak-to-peak
energy state. However, the energy of linewidth, Bpp , is generally taken as the
interaction in the EPR experiment is abscissa distance between the maximum
considerably smaller than kT. and the minimum of the dA/dB curve.
Transitions between the two spin states
may be induced by the application of elec- 3.2.2.2 The g-value
tromagnetic radiation of the appropriate Until now only free electrons have been
frequency, ν, such that the applied energy, discussed. In reality, the electron respon-
2πν h̄ exactly matches that of the Zeeman- sible for the resonant absorption is nor-
level separation, ge γe Bz , that is mally an unpaired one within a molecule,
ge γe Bz thereby experiencing a local magnetic field
ν= (6) different from the applied field, since the
2π h̄
applied field will induce currents in the
This phenomenon is known as magnetic electrons of the molecule. The resonance
resonance; the frequency at which these condition is consequently modified to
transitions occur is called the resonant
ge γe (1 − σ )Bz
frequency. Resonance can be observed in ν= (7)
two different ways; the frequency may be 2π h̄
varied at a fixed magnetic field strength, where σ represents a shielding constant
or mutatis mutandis the converse. In prac- similar to that in NMR spectroscopy. The
tice, it is more beneficial to utilize the g-factor of the molecule is then given by
latter of these two methods. Equation (6) the quantity g = ge (1 − σ ).
suggests that the resonant frequency of In addition to the magnetism arising
a free electron is 9.4 GHz for an applied from the electron spin, there may also
magnetic field of 3400 G (1 G = 10−4 T). be a contribution from the electronic
These approximate conditions are em- orbital motion since atomic orbitals have
ployed to experimentally observe EPR associated
√ angular momenta of magnitude
transitions; the frequency of 9.4 GHz is h̄ ( + 1), in which = 0,1, and 2 for
in the microwave region, corresponding to s-, p-, and d-orbitals, respectively. Hence,
a wavelength ∼3 cm (so-called ‘‘X-band’’). occupied orbitals that are nondegenerate
Occasionally, measurements are under- lead to nonzero angular momenta. An
taken under other conditions, such as orbital angular momentum, l, has an
Q-band (corresponding to ν = 35 GHz, associated magnetic moment
Bz = 12 500 G, λ = 0.86 cm) or S-band
µL = γe l (8)
(ν = 3.2 GHz, Bz = 1140 G, λ = 9.4 cm).
In contrast to NMR spectra, EPR spectra For light atoms and certain linear radicals,
are recorded as the first derivative of the in- s and l combine via the Russell–Saunders
tensity, A, of the absorption with respect to Coupling scheme to give a total angu-
the field induction, B, and as a function of lar momentum, j. EPR spectroscopy then
B, that is dA/dB versus B, as indicated in interrogates transitions between the differ-
Fig. 2. The reasons for doing this lie in the ent j levels. Hence, since the Hamiltonian
phase-sensitive methods of detection em- operator for an electron spin in a molecule
ployed in EPR, and consequently this way contains a term representing the effect of
the magnetic field on the electron spin homogeneity of the magnetic field and
magnetic moment, another for spin-orbit the frequency of the magnetic induction
coupling, and another for the interaction modulation. Several other processes affect
of the applied field with the orbital angular the linewidth and these are considered in
momentum, the g-factor can absorb the turn. It should be noted that the two most
effects of the orbital angular momentum frequently encountered lineshapes ob-
operators, giving rise to an effective Spin served are Gaussian and Lorentzian. For a
Hamiltonian. line centered at frequency ν ∗ , a Lorentzian
In most molecular radicals, the orbital lineshape is given by the equation
contribution is lost (‘‘quenched’’), as the
2T2
existence of covalent bonds lifts the f (ν) = (10)
degeneracy of the orbitals, and providing 1 + 4π 2 T22 (ν − ν ∗ )2
the electron is in an orbital sufficiently and a Gaussian lineshape by the equation
removed in energy from other levels, the √
effective g-value will be close to the ‘‘spin- f (ν) = T2 2π exp[−2π 2 T22 (ν − ν ∗ )2 ]
only’’ value, and the resonance condition (11)
is given by where the symbol T2 is defined later. The
gγe Bz Gaussian lineshape is observed when the
ν= (9)
2π h̄ line is the superposition of a large num-
in which g is now a unique property of the ber of unresolved individual components,
whole molecule. The value of the g-factor such a line being referred to as inhomo-
will be independent of the magnetic field geneously broadened, and is commonly
direction only in isotropic systems such as encountered in solid-state EPR spectra.
organic radicals in low-viscosity solutions More important is the Lorenztian line-
in which there is rapid and random shape, which is often observed for radicals
tumbling of the molecules so as to permit in solution, when complications due to un-
an average g-value to be observed. For the resolved hyperfine splittings do not occur.
purposes of understanding liquid phase
EPR spectra generated in electrochemical 3.2.2.3.1 Spin-lattice Relaxation The tr-
EPR experiments, this average value is ansitions between eigenstates induced by
sufficient. However, in a large number of the electromagnetic radiation necessarily
other systems, the g-factor is anisotropic, involves a perturbation of the Boltzmann
varying markedly with the orientation of Distribution Law,
the sample (this is a consequence of its
dependence upon the spin-orbit coupling). n2 −(E2 − E1 )
= exp (12)
In these cases, especially for solid-state n1 kT
spectra, the g-factor must be regarded as In this equation, k and T refer to the Boltz-
a tensor [32]. Furthermore, the anisotropy mann constant (1.380662 × 10−23 J K−1 )
of the g-tensor may contribute in some and the absolute temperature, respectively,
cases to the linewidth of the spectra from and ni refers to the electron population of
high-viscosity solution-phase radicals. the ith non-degenerate energy level of en-
ergy Ei , as given in Fig. 2. At equilibrium,
3.2.2.3 Linewidths the lower state is more filled, and reso-
The linewidth of an EPR spectrum de- nance initially induces more transitions
pends upon two instrumental factors, the from E1 to state E2 than the reverse. It is
expected that the populations will rapidly absorption lines occur. The former sit-
equalize, after which, there would be no uation is often encountered in the EPR
resultant energy absorption, even though spectra of transition metal ions, whilst the
resonant transitions would still be occur- latter is usually found for organic radicals
ring. This phenomenon is termed satura- in solution.
tion and is observed when high microwave Another form of spin–lattice relaxation
powers are applied. When low-power mi- occurs when there is a quadrupolar nu-
crowaves are employed, magnetic reso- cleus (vide infra) present in the paramag-
nance transitions are observed and non- netic species. Associated with this nucleus
radiative ‘‘relaxation’’ processes occur. is an electric quadrupolar field, which
These processes oppose saturation and will interact with gradients in the electric
help restore the natural order of the ther- component of the microwave radiation.
mal Boltzmann Distribution, and they are However, as this electric quadrupolar in-
termed spin-lattice relaxation phenomena. teraction is generally weak, it does not
The mechanism for spin-lattice relax- contribute much towards the relaxation.
ation is as follows. All paramagnetic
species in the sample have an associ- 3.2.2.3.2 Spin–spin Relaxation The sec-
ated magnetic field surrounding them with ond major nonradiative phenomenon is
which each of the other paramagnetic termed spin–spin relaxation and is char-
species may interact. In liquids, the ran- acterized by a relaxation time T2 . The
dom molecular collisions that constitute processes that contribute to T1 also in-
Brownian motion permit these local mag- fluence T2 , which controls the linewidth
netic fields to fluctuate; a fluctuation that (vide supra). However, the random forces
occurs at the resonant frequency will in- that modulate the electron spin energy
duce a ‘‘radiationless’’ transition. The spin- levels at low frequencies (below 1010 Hz)
lattice relaxation is characterized by a spin- contribute to T2 , whereas they have no
lattice relaxation time, T1 , which thus ef- influence upon T1 . Again, the source of
fectively controls the degree of saturation. the fluctuating magnetic field lies in the
Since the relaxation determines the life- anisotropic magnetic interaction within
time, t, of a spin state, the Heisenberg the molecule, mainly due to the anisotropy
Uncertainty Principle relates it to the in the g-tensor and to the dipolar hyper-
uncertainty of the Zeeman eigenvalues, fine coupling with magnetic nuclei. This
E1 and E2 , thereby allowing this phe- has assumed that the radical (or parent
nomenon to affect the linewidth of EPR species) concentration is not too high;
signals, as these depend inversely on T1 . otherwise, spin exchange occurs as a re-
E • t ≈ h̄ (13) sult of radical–radical collisions, thereby
causing a broadening of the EPR lines.
Clearly, if the spin–lattice interaction is The spin–spin process involves the inter-
strong, T1 is small, t is small, and E change of the spins of the colliding radi-
must be large, leading to a spread in cals. At very high radical concentrations,
the energy states between which transi- the spin ‘‘sees’’ an average environment as
tions occur. Hence, absorption effectively it moves from radical to radical, and so the
occurs over a small range of magnetic hyperfine splitting is lost and a single line
field strengths. Mutatis mutandis, if the results. This is the reason solid state EPR
spin–lattice interaction is weak, narrow spectra are generally of a single line, as
Fig. 3 Typical EPR signal

obtained upon reduction of
poly(nitrostyrene) immobilized
5G on an electrode.
shown in Fig. 3 for the poly(nitrostyrene) Tab. 1 Characteristics of some nuclei

radical anion.
At high parent concentration, electron Natural Spin g-value
exchange may occur between radical and abundance angular
parent species, thereby inducing a line % momentum
broadening; for example, the extent of such
Electron – 1/2 2.0023
broadening has been used to calculate a 1H 99.98 1/2 5.585
rate constant for the chemical exchange 2D 0.02 1 0.857
between benzonitrile and the benzonitrile 6 Li 7.43 1 0.8219
7 Li
radical anion generated electrochemically 92.5 3/2 2.1707
12 C 98.9 0 0
in situ [33] yielding a rate constant of 13 C 1.1 1/2 1.405
2 × 108 dm3 mol−1 s−1 . 14 N 99.63 1 0.404
15 N 0.37 1/2 −0.567
19 F 100 1/2 5.257
3.2.2.4 Hyperfine Structure 23 Na 100 3/2 1.477
The most useful characteristic of EPR 31 P 100 1/2 2.263
spectra in the identification of radical
species is the splitting due to hyperfine
coupling. This arises from the electron the magnetic field and is integral or
spin interacting with any magnetic nuclei half-odd integral in value. This range of
in the molecule. Magnetic nuclei, with (2I + 1) values can be expressed as the
nuclear spin I, behave analogously to Clebsch–Gordon series, MI = I , (I − 1),
electrons when placed into a magnetic (I − 2), . . . − (I − 1), −I .
field. Hence the nuclear magnetic moment It is instructive to consider the mech-
may be written in the form of Eq. 14, anism by which the hyperfine coupling
µN = gN γN I (14) arises. In a strong magnetic field, the inter-
action between an unpaired electron and
The gN values are dimensionless and a magnetic nucleus acts as a small pertur-
characteristic of the type of nucleus. γN bation δE to the Zeeman levels E1 and E2
is the nuclear gyromagnetic ratio. Some of the electron spin. This perturbation is
values are listed in Table 1, together with composed of two components,
the corresponding quantum numbers I
and MI , in which the quantum number δE = (δE)isotropic + (δE)anisotropic
MI allows for the space-quantization of (15)
The anisotropic term, (δE)anisotropic , repre- This interaction arises because of the pres-
sents the dipolar interaction that depends ence of a small, but nevertheless finite
upon the relative positions of the mag- spin density at the position of the nucleus.
netic moments of the unpaired electron In a strong magnetic field applied in the
and the nucleus (µe and µN ). In sin- z-direction, this term, referred to as the
gle crystals, this gives helpful information coupling constant, a, is proportional to the
on the crystal geometry; in polycrystalline square of the electronic wave function at
substances, it causes line broadening. the nucleus,
In liquid samples, however, since the 8π
a= ge gN γe γN |ψ(0)|2 (16)
molecules are constantly tumbling, the 3
magnetic dipole–dipole interactions aver- Since only s-orbitals have a finite prob-
age to zero, except for a small amount ability density at the nucleus (other or-
that is dependent upon the liquid viscosity bitals all have nodes in their wave func-
and hence gives rise to line broadening. tions there), the contact interaction can
The hyperfine structure due to radicals only occur when the electron occupies a
in solution, therefore, results from an molecular orbital in which there is some
isotropic term, (δE)isotropic , which is a s-orbital character.
consequence of a nonclassical interaction, Thus, for each nuclear spin state, a sep-
known as the Fermi-Contact interaction. arate transition in the EPR spectrum will
I = 1/2 MI I=1 MI
+1/2 +1
0
MS −1/2 MS −1
+1/2 +1/2
2phn
−1/2 −1/2
−1/2 +1
0
+1/2 −1
a a
B B
(a) MI +1/2 −1/2 (b) MI +1 0 −1
Fig. 4 Splitting of an EPR signal by hyperfine interaction

between the unpaired electron and a nucleus with spin
quantum number (a) I = 1/2 or (b) I = 1. Also shown are the
EPR spectra anticipated for both cases.
be observed. This is illustrated in Fig. 4 of doublets’’. When, however, a1 = a2 , the

in which the unpaired electron interacts nuclei 1 and 2 are referred to as being equiv-
with only one nucleus having a spin quan- alent. Typical EPR spectra are illustrated in
tum number, I , of 1/2 or 1, corresponding Fig. 5 for the case of two equivalent spin-
to a 1 H or a 14 N nucleus, respectively. 1/2 nuclei and equivalent spin-1 nuclei. In
The hyperfine coupling constant is the these cases, since three are certain spin
separation between the two lines, and is configurations that are degenerate, some
field-independent, and a characteristic of lines coincide.
the nucleus–electron interaction. During In general, n equivalent nuclei with spin
the electronic transitions, the nuclear spin quantum number I give rise to (2nI + 1)
quantum number, MI , does not change, equidistant hyperfine lines in the EPR
and we may formulate the ‘‘first-order’’ spectrum. In particular, for an electron
selection rules, interacting with n equivalent protons (I =
1/2), (n + 1) lines are produced whose
Ms = ±1; MI = 0 relative intensities may be predicted from
The above rules are valid provided the the coefficients of the binomial expansion
magnitude of the hyperfine coupling is not of (1 + x)n . These can be recalled most
too large; otherwise, the spins are partially readily with the help of Pascal’s triangle,
coupled and the behavior is more com- as shown in Fig. 6. Also shown in Fig. 6
plicated [32]. Hence, two hyperfine lines are the relative line intensities for the
(a doublet) are detected in the spectrum coupling of n equivalent spin-1 nuclei
resulting from the interaction of the un- with an electron (giving rise to (2n + 1)
paired electron with a proton, and three hyperfine lines). To illustrate this, Fig. 7
hyperfine lines (a triplet) when the electron shows the EPR spectrum of the benzene
interacts with a 14 N nucleus. Quantita- anion. This species has six equivalent
tively, under first-order conditions, the protons, and the EPR spectrum is clearly
electronic energy levels of a one-electron, a 1 : 6 : 15 : 20 : 15 : 6 : 1 septet, as would be
one-nucleus system are given by anticipated from the above ideas.
It is not immediately obvious how
E = Ms (−ge γe h̄Bz + aMI ) (17) hyperfine splitting arises in the case of
The above can be extended for radicals the benzene anion in which the electron
containing more than one nucleus, by occupies a π-orbital that has a nodal
splitting the levels E1 and E2 in succession plane coincident with the plane of the
depending upon interaction with the aromatic ring. Similarly, it is not obvious
nucleus in question. For a radical coupling as to why in the ethyl radical (CH3 CH2 • )
to two spin-1/2 nuclei, the energy states the methyl (or β) protons show a larger
are given by coupling constant that the α-protons. This
is rationalized below.
E = Ms (−ge γe h̄Bz + a1 M1 + a2 M2 ) Consider and isolated >C• −H unit with
(18) one electron in the π-orbital on the C-atom.
in which a1 and a2 are the two coupling This orbital is perpendicular to the plane
constants, and M1 and M2 are the nu- of the three trigonal σ -bonds. Electrons
clear spin quantum numbers. The EPR in the C−H bond will keep apart, as
spectrum has four lines of equal inten- they will mutually repel each other. This
sity, and is described as being a ‘‘doublet is represented in Fig. 8 in which two
‘‘structures’’ are possible, depending upon the spin of the electron in the sigma
whether the electron spin nearest the orbital on the C-atom takes on slightly
C-atom is parallel or antiparallel to the more of the character of the π-orbital. This
electron spin in the π-orbital. Exchange means that the electron at the H atom
forces [32] cause the structure with the must have a small excess of the opposite
parallel spins at the C-atom to have slightly spin and this resides in the 1s orbital
lower energy. Hence, the spins in the used to form the C−H bond. The result
C−H bond are slightly polarized so that is a finite spin density at the H-nucleus
M1 M2 ∑M 1 I=1 ∑M 1
I = 1/2
+1/2 +1/2 +1 +2
+1/2 −1/2 +1
−1/2 +1/2
0 0
−1/2 −1/2 −1 −1
MS MS −2
+1/2 +1/2
2phn
−1/2 −1/2
−1/2 −1/2 +2
−1 +1
−1/2 +1/2
+1/2 −1/2
0 0
a
+1/2 +1/2 +1 a +1
+2
B B
∑M1 +1 0 −1 ∑M1 +2 +1 0 −1 −2
Fig. 5 Splitting of an EPR signal by the hyperfine interaction between the unpaired electron
and two equivalent nuclei.
I = 1/2 1 n=0 I = 1: 1
1 1 1 1 1 1
1 2 1 2 1 2 3 2 1
1 3 3 1 3 1 3 6 7 6 3 1
1 4 6 4 1 4 1 4 10 16 19 16 10 4 1
1 5 10 10 5 1 5
1 6 15 20 15 6 1 6
Fig. 6 EPR line intensities for the coupling between two equivalent nuclei when I = 1/2
(Pascal’s Triangle) and I = 1.
5G
Fig. 7 The EPR spectrum for the benzene anion in

1,2-dimethoxyethane at −80 ◦ C.
Fig. 8Canonical structures of

the C−H fragment.
C H C H
and, hence, a hyperfine splitting is seen. since exchange forces are capable of
Quantitatively [32], the hyperfine coupling transmitting spin density through more
constant should be proportional to the than one bond. A more dominant ef-
π-electron density, ρ, on the carbon fect, however, is due to hyperconjuga-
atom. This is expressed mathematically tion. This mechanism involves the di-
in McConnell’s equation [34], rect overlap of the orbital containing
the unpaired electron with the C−H
a(H) = Qρ (19)
σ −bonds formed by the β-hydrogens,
where Q is a proportionality constant and as illustrated in Fig. 9. The magnitude
takes a value of about −23 G. The negative of this interaction depends upon the
coupling constant that is thus anticipated relative orientation of the C−H bond
is because the unpaired spin density at the to the orbital containing the unpaired
H atom is of opposite sign to that in the electron. If the angle between the two
π-orbital on the C-atom. is θ, then
The above can be easily extended to
account for EPR signals at β-protons, a(H) = A + B cos2 θ (20)
R
H
R′ C q
C Fig. 9 The magnitude of the
R ′′ hyperfine coupling between an
R ′′′ electron and a β-proton
R ′′ q H depends on the dihedral
angle, θ, between the C−H
R C R′ bond and the p-orbital
containing the
R ′′′ unpaired electron.
where A is close to zero and B is about upon the wavelength of the radiation that
46 G. In the ethyl radical, the methyl group is reflected off them – to the sample in the
can freely rotate and so an average value EPR cavity. The resonant cavity is located
is seen. Since the averaged value of cos2 θ in the homogeneous part of the magnetic
is 1/2, a value of ca. 23 G would be pre- field. Within this, standing waves are set
dicted for the coupling to these protons, in up by reflections off the end walls when the
excellent agreement with the experimen- radiation is fed in through a small coupling
tally determined value (22.38 G). hole from the waveguide; Fig. 11 shows the
type of cavity most frequently encountered,
3.2.2.5 EPR Instrumentation a rectangular TE102 cavity. The standing
Figure 10 shows a schematic illustration wave permits the spatial separation of the
of an EPR spectrometer. An electromagnet electric and magnetic components of the
supplies the adjustable applied magnetic microwave radiation. It can be seen that the
field. Between the poles, the field is ho- magnetic component, which is responsible
mogeneous to one part in 105 –106 . The for inducing EPR transitions, has maxi-
stability of this uniform field is achieved mum intensity at the cavity center and
by using a regulated power supply, any is therefore concentrated upon the sam-
variation in the field being detected by a ple, when placed there. The distribution
Hall effect device. Klystron oscillators gen- of the microwave field within the cavity
erate microwaves, the frequency of which is such that the sensitivity profile along
is determined by the voltage applied to the the vertical axis of the cavity obeys a sin2
klystron. The stability of the microwave relationship, with zero sensitivity at the
frequency is achieved using an automatic cavity edges and maximum sensitivity in
frequency control system. The microwaves the cavity center [35]. The electric compo-
travel through waveguides – hollow metal nent has a node at the cavity center. This
pipes of rectangular cross section with is advantageous, since aqueous samples
inside surfaces that are gold plated and in particular exhibit appreciable dielec-
whose dimensions are critically dependent tric loss at microwave frequencies, which
Detector
crystal
Waveguide
Klystron Sample Phase-sensitive
tube detection
Modulation coils
Signal out
Electromagnet Magnetic field

modulation (100 kHz)
Fig. 10 The basic features of an EPR spectrometer.

Sample tube Fig. 11 A TE102 EPR cavity.
Magnetic vector
of radiation
Electric vector
prevents the establishment of a standing the power input, whilst the latter protects
wave; by containing these samples within the klystron from reflected radiation.
a thin flat cell in which the solution is This radiation reaches the detector, a
held on the nodal plane, thereby escaping crystal diode via a ‘‘magic tee’’ bridge, as
interaction with the electric component of illustrated in Fig. 10. The bridge can be
the microwave field, this problem is over- adjusted (‘‘balanced’’) so that no radiation
come. The cells used are fabricated from reaches the detector if no absorption of
silica in preference to glass, since the latter microwaves occurs. At resonance, energy
contains paramagnetic impurities observ- absorption by the bridge unbalances the
able in the EPR spectrum. As well as the bridge and microwave radiation is detected
rectangular cavity described above, cylin- by the crystal diode detector.
drical cavities and microwave helices [36] One final point to note about the ex-
have found application in electrochemi- perimental EPR arrangement is that a
cal EPR. second magnetic field is set up around
The quality of the resonant EPR cavity is the cavity using Helmholtz coils through
measured by its efficiency for integrating which an alternating current is applied,
microwave energy. A measure of this thereby producing a sinusoidally vary-
efficiency is described by the quality factor ing magnetic field with time. This is
(Q-factor), known as field modulation, and, with a
feedback mechanism acts to prevent satu-
Energy Stored ration from arising. Further advantages
Q = 2π ×
Energy Dissipated Per Cycle include improvements in the signal-to-
(21) noise ratio made by increasing the modu-
A higher Q implies a greater capacity lation amplitude, although caution must
for energy storage, and only a narrower be duly exercised, since if the modu-
bandwidth of frequencies will be admitted lation amplitude is approximately equal
into the cavity. to the linewidth, the resolution dete-
An attenuator and a ferrite insulator are riorates. Hence, modulation amplitude
placed between the klystron and the wave- should generally not exceed about half the
guide. The former allows regulation of linewidth.
The ideas outlined in this section are A full discussion of approximations to the
treated in depth in the numerous text- double integral may be found in the review
books on EPR [1, 32, 37–41]. In particular, by Goldberg [42].
the book by Symons [37] provides an ex-
cellent and stimulating introduction for
3.2.3.1 The Bard–Goldberg Cell
chemists. All aspects of EPR instrumenta-
tion are thoroughly covered in the work by
3.2.3.1.1 Cell Design This in situ sta-
Poole [35].
tionary solution cell developed by Bard
and coworkers [43–45], is illustrated in
3.2.3
Fig. 12. It is similar to the in situ flat cell
In Situ Electrochemical EPR: Cell Design
pioneered by Adams and coworkers [17].
It consists of a platinum mesh working
In this section, contemporary methods for
electrode (1 mm grid size) placed in a
in situ electrochemical EPR are described,
with particular emphasis given to the use flat cell, with a U-shaped tungsten rod
of such techniques for the investigation counter electrode along the cell edges, and
of electrogeneration kinetics of short-lived a silver wire quasi-reference electrode lo-
intermediates and the mechanisms of cated in the flat cell above the working
their decay. The EPR-observable is the electrode. Unlike the Adams cell design,
EPR signal intensity. Since there are ohmic drop problems were removed by
many lines in each EPR spectrum, the the location of all three electrodes within
EPR signal intensity is strictly the double the EPR cavity for electrolyte solutions in
integral of the EPR signal with respect DMF. However, a careful choice of sup-
to the magnetic field [42]. However, this porting electrolyte was necessary; ClO2 •
can often be approximated by the peak- radicals may be produced at the counter
to-peak amplitude of the largest line in electrode when reductions occurred at the
the spectrum, provided the peak-to-peak working electrode in the presence of per-
linewidth (Bpp ) is sufficiently small. chlorate background electrolytes. The use
Pt wire
Reference
Ag wire
Teflon sleeve
Counter electrode
Working
electrode
Glass wall
Fig. 12 Schematic diagram of

the Bard–Goldberg cell.
of tetrabutlyammonium iodide as sup- dimerization [45]. Comparison of these

porting electrolyte mitigated this problem, data with the rate constants deduced
since nonparamagnetic iodine is formed from conventional cyclic voltammograms,
at the counter electrode during reduc- 160 ± 40 dm3 mol−1 s−1 [45], suggesting
tions [43]. It was concluded that where that the mechanism is, in fact, a radical-
radical formation is possible in this way, anion dimerization,
it might be necessary to add a nonradical
producing electroactive substance so that R + e− −−−→ R −•
the production of secondary radicals is pre- R −• + R −• −−−→ R2 2−
vented. Later work [45] confirmed these
problems did indeed arise from radicals
3.2.3.2 The Allendoerfer Cells
produced at the counter electrode.
3.2.3.2.1 Cell Design Allendoerfer and
3.2.3.1.2 Typical Applications Experi- coworkers [46–49] pioneered the electro-
ments performed with stable radical ions chemical EPR cell design to exhibit max-
in DMF, such as those produced from imum sensitivity to unstable radicals.
the reduction of nitrobenzene, azoben- These state-of-the-art in situ cells are prob-
zene, and 9,10-anthraquinone, indicated ably the most sensitive for the detection
that the observed signal was stable for of short-lived (∼10−5 s) paramagnetic in-
20–30 s after generation by a current step, termediates. Two designs have emerged,
and after which natural convection set in, one using a TE011 cylindrical EPR cav-
suggesting that kinetic measurements are ity [46–48], the basis of which is illustrated
possible from the magnitude of this con- in Fig. 13, and a cell based on a loop-
stant steady state EPR signal, within the gap resonator, Fig. 14 [49]. These designs
20 s time limit [43]. and their applications to the study of radi-
It was further shown that the ho- cals with small lifetimes are discussed in
mogeneous kinetics of the radical ions turn.
produced electrochemically could be in-
ferred from a current pulse experi-
ment in which the EPR signal is mea- 3.2.3.2.2 The Coaxial Cell The cylindri-
sured during and following the cur- cal TE011 cavity is modified by centering a
rent pulse. The signal transients were metallic rod along the axis of the cavity. A
then analyzed in terms of working quartz cell contains this cylindrical metal-
curves for first-order radical decomposi- lic conductor, and the liquid sample is
tion, second-order radical ion dimeriza- contained within the unfilled volume of the
tion, and second-order radical ion-parent cell. For conventional EPR spectroscopy,
dimerization [44]. In this manner, the the sensitivity of such a coaxial cavity was
mechanism of the reduction of olefins was shown to be at least as good as that of a
deduced, illustrated best by the dimeriza- rectangular TE102 cavity for lossy samples;
tion of dimethyl fumarate radical anions. but as an in situ cell for electrochemistry, a
Analysis of the EPR signal transients cell based upon the coaxial design exhibits
generated by a current pulse yielded greater sensitivity compared with in situ
rate constants of 160 ± 26 dm3 mol−1 s−1 cells based upon the conventional EPR flat
for radical-anion dimerization and 50 ± cell due to the significantly larger electrode
30 dm3 mol−1 s−1 for radical anion-parent area that can be accommodated when the
Fig. 13 A schematic diagram of the

Allendoerfer coaxial cell cavity design. A
cross section of the TE011 cylindrical
cavity (A) is shown. A metal cylinder
(B) is located along the axis of the cavity
within a silica sample tube (C).
central conductor of the coaxial cavity acts

as the working electrode.
Initial experiments [47] involved a cen-
tral conductor of 6 mm diameter and
50 mm length. One problem that arose was
that the resonant frequency of the coaxial
cavity was greater that the 9.0–9.8 GHz A B
range of the employed klystron. To
overcome this, the resonant frequency was
lowered by surrounding the central metal
rod with a material of suitable relative per-
mittivity, such as Teflon(R) . It was shown
that for a solid central conductor of good
conductivity, the cavity Q-factor (a direct C
measurement of the EPR sensitivity) was
not significantly different from that of the
TE011 cavity employed.
To overcome some problems associated
with using a solid central conductor, diameter 6 mm). In this form, the standing
the inner part of the coaxial cavity was microwave field ‘‘sees’’ only the portion
constructed from a finely wound shallow of solution between the helix and the
pitched helix, which stands freely against inner wall of the quartz tube and does not
the inner wall of a quartz test tube (internal penetrate the inside of the helix, as shown
0.4 mm od
0.3 mm id
Working
electrode
Loop-gap
resonator
5 mm
0.25 mm
0.8 mm od
0.6 mm id
Fig. 14 A schematic illustration 1 mm

of the active region of
Allendoerfer’s electrochemical
loop-gap resonator geometry.
Fig. 15 The EPR-visible part of the

Allendoerfer coaxial cell is that between
Quartz the wire helix and the quartz sample
tube (shaded region). The inner part of
the helix is free to house the reference
and counter electrodes without
interference with the EPR.
Wire
Inactive volume
1 mm
in Fig. 15. Since there is no penetration of the in situ cell were indistinguishable
microwaves within the helix, any material from those obtained from normal po-
located here does not affect the EPR signal larographic cells, except for the large
(or sensitivity), allowing the counter and increase in current due to the large
reference electrodes to be placed here. The surface area. It was estimated that the
counter electrode is a platinum cylinder uncompensated solution resistance was
placed centrally along the cavity axis such less than 1 -, even when highly resistive
that the current flow is radial, and so nonaqueous solvents were employed [47].
ensuring that there is a uniform current Radicals as short-lived as 10−5 s were
density over the whole cylindrical working- anticipated to be amenable to study,
electrode surface. The uniform current providing a high concentration (approx-
density meant that potential control of imately 0.1 M) of electroactive material
the whole electrode can be achieved was used; in agreement with previous es-
by monitoring a single point with a timates [26] of the shortest-lived radical,
Luggin capillary. it would be possible to study using in
situ techniques.
3.2.3.2.3 Application The surface area of A modification of the Allendoerfer cell
the working electrode was estimated to has been described by Ohya–Nishiguchi
be about 22 cm2 , significantly larger than and coworkers [50–52], whereby low-
in previous electrochemical EPR cell de- temperature studies were undertaken on
signs. The cell resistance was found to the electrochemistry of twenty aromatic
be 13 - when filled with 0.1 M tetrabuty- compounds. The low temperature was
lammonium perchlorate in DMF [47], a achieved by placing the electrolysis cell
value that compares favorably with the within a temperature control Dewar. How-
5–10 k- estimated by Bard and Gold- ever, it should be noted that this low-
berg [43] for an in situ cell based upon temperature in situ cell did not use a
a conventional EPR flat cell. Electro- reference electrode.
chemical measurements made with this In the study of short-lived radicals, the
cell design were shown to be free from presence of efficient hydrodynamic flow
ohmic distortions, even when used with is essential to sustain a constant supply
poorly conducting electrolytes, and that of electroactive material to the working-
the cyclic voltammograms observed from electrode surface, and hence ensures a
Fig. 16 The Allendoerfer– Reference electrode

Carroll cell as modified to Luggin capillary
provide the capability for Quartz tube
solution flow.
Auxiliary electrode
Working electrode
Teflon flow
baffles
Flow Flow
steady flux of radicals. Carroll adapted the higher frequencies. Loop-gap resonators
Alendoerfer cell to this end [46, 48]. The have a number of advantages over con-
changes to the helical arrangement are ventional resonant cavities [53], notably
illustrated in Fig. 16. Solution is prevented their smaller size and greater observed
from flowing into the central EPR inactive signal intensity. For electrochemical pur-
volume of the cell by a complex series poses, these are beneficial; the former lead
of baffles. The value of flow coupled to smaller electrolysis currents and con-
to the Allendoerfer cell was shown by comitantly less potential control problems,
the electroreduction of nitromethane in whilst the latter allows enhanced sensitiv-
aqueous solutions; the radical anion of ity on the basis of the total number of
nitromethane was observed and this was spins (that is spin concentration multiplied
shown to have a lifetime in the order of by sample volume), and is illustrated in
10 ms. To date, this is the shortest radical Fig. 17 for the case of a small sample of
lifetime observed by electrochemical EPR, α,α -diphenyl-β-picryl hydrazyl (DPPH).
although it has been calculated [50] that Allendoerfer’s loop-gap resonator cell
such a cell should be capable of observing design [49] is illustrated in Fig. 13. The
radicals with lifetimes of 10−5 s. working electrode comprises a gold wire
(although a variety of other materials
3.2.3.2.4 The Loop-gap Resonator Cell were also found to work equally as well)
The loop-gap resonator was introduced placed inside the active volume of the
into EPR spectroscopy by Froncisz and resonator; insertion into the resonator in
Hyde [53], who used an arrangement this manner was found to have no dele-
shown in Fig. 17. Its design and appli- terious effect, and little change in the
cations have been reviewed [54]. The loop resonator Q-factor. However, since the
behaves as an inductive element, whilst EPR active volume decreases, the energy
the gaps act as capacitative elements. It density increases; hence, to avoid satura-
has been found [53] that increasing the tion, the microwave power must decrease.
number of gaps permits EPR spectroscopy A platinum needle acted as the counter-
to be undertaken on larger samples and at electrode. For the purpose of generating
iii
iv i
ii R
ii
iii r t
ii
Z ii
i
W
(a)
X40 TE102
(b)
Loop gap
4G
(c)
Fig. 17 (a) The loop-gap resonator as used by Froncisz and Hyde. The
principle components, (i) loop, (ii) gaps, (iii) shield, and (iv) inductive
coupler are shown. The critical dimensions, Z, resonator length, r, resonator
radius, R, shield radius, t, gap separation, and W, gap width are also
indicated. The sample is inserted into the loop (i) through the coupler (iv).
The microwave magnetic field in the loop is parallel to the axis of the loop
and (b, c) Comparison of EPR spectra obtained of a point sample of DPPH
measured at X-band in a rectangular TE102 cavity and in a loop-gap resonator
of dimensions r = 0.6 mm, Z = 5 mm. The incident power was held
constant at a low, nonsaturating level. The loop-gap resonator yields 37 times
greater signal than the TE102 cavity.
radicals for EPR spectroscopy, this two- configuration employing a Ag|AgCl refer-
electrode arrangement was adequate, with ence electrode.
the platinum electrode acting as a quasi-
reference electrode in aqueous chloride.
However, for the purposes of in situ elec- 3.2.3.2.5 Application Allendoerfer and
trochemical EPR, this arrangement was coworkers [49] observed the time-depen-
refined to accommodate a three-electrode dent formation and post electrolysis decay
Fig. 18 Schematic illustration of the in

situ channel flow cell.
of the dianion radical of para-nitrobenzoic

acid in water at pH 11. It was found that
the radical decay kinetics were not a simple
function of pH or time.
3.2.3.3 The Channel Electrode Cell
3.2.3.3.1 Cell Design Compton and

Coles designed a channel flow cell that
could be used for in situ electrochemical
EPR [55]. The design of this demountable
flow cell, constructed in synthetic silica, is
given in Fig. 18. A rectangular metal foil
(such as platinum), cemented onto a sil-
ica plate, comprises the working electrode,
with a fine lead out wire at the rear making
the connection to the electrode. The whole
cell is sealed together with a low-melting
wax, which was found to be non-lossy. The
cell is connected to Polytetrafluoroethane acceptably low, whilst the Q-value when
(PTFE) tubing for connection into a flow this cell is filled as before is 1200 – a value
system, and is held within a TE102 res- that is typically observed for conventional
onant cavity, accomplished by placing it aqueous sample EPR methods.
inside a silica tube that runs right through The flow system used in conjunction
the cavity and fixing its position with with the channel unit is capable of
PTFE spacers. Movement of this support- a variable laminar flow, established by
ing tube, which also serves to protect the having a Reynolds Number, Re < 102
cavity from any leakage, may finely adjust (see Volume 2, chapter 2.2 (by Fisher), or
the cell position within the EPR cavity. Op- equally Volume 3, chapter 3.5 (by Mount).),
timally, the cell is positioned at the point employing flow rates in the range 10−4 to
of maximum EPR sensitivity, which cor- 10−1 cm3 s−1 , achieved using a gravity-fed
responds to the center of the EPR cavity, system (so as to avoid variation in pumping
with the plane of the electrode parallel to pressure [56]) together with capillaries
the electric component of the microwaves of different diameters and changes in
standing wave. For an empty cell with- reservoir height. A platinum gauze counter
out electrode, the cavity Q-factor is 6400, electrode is placed downstream of the
which is lowered to 2500 when filled with working electrode and a saturated calomel
an aqueous 0.1 M electrolytic solution. An or Ag|AgCl reference electrode upstream,
empty cell with an electrode present has both outside the cavity. The flow within
a Q-value of 4600, confirming that mi- the channel cell is parabolic, provided
crowave losses due to the lead out wire are that there is a sufficiently long lead-in,

that is if the lead-in length is given by (h − y)2
0.1 × h × Re, where h is the half-height νx = ν0 1− (23)
h2
of the channel [57]. This was verified
where ν0 is the mean solution velocity
experimentally by the observation of plug
at the center of the channel. The ve-
flow using a colored solution [55]. Edge
locity profile may be simplified for the
effects were found to be negligible if the
case of macroelectrodes by employing the
electrode size is less than 4 mm wide
Lévêque Approximation [64], (y/2h) 1,
in a 6-mm channel. Further, because of
yielding,
the electrolyte resistance in the narrow y
νx ≈ 2ν0 (24)
gap through which solution flows [58], the h
potential may be slightly non-uniform [23, Using this linearized form of the velocity
59]. This effect may be anticipated to parabola, the Levich equation [57] can be
perturb the current density unless the deduced for the case of a chemically
distance over which radicals are formed and kinetically uncomplicated electron-
is small compared with the total distance transfer process,
over which radicals are detected, thereby 1/3
encouraging the use of small electrodes. ilim = 0.925nF cbulk D 2/3 Vf
× (h2 d)−1/3 wxe

2/3
(25)
3.2.3.3.2 Application Channel electro- where d is the channel width, w is
des have well-defined hydrodynamic prop- the electrode width, xe is the electrode
erties [57], and the laminar flow convective- length, and Vf (= 4ν0 dh/3) is the volume
diffusion equation describing the mass flow rate.
transport regime (see Chapter 2.2 of this Two electrochemical EPR methodolo-
volume) within this cell at steady state for gies may be envisaged using a channel
large electrodes is given by [60, 61] flow cell for the determination of kinetic
and mechanism of radical decay. First is
∂ 2 c(x, y) ∂c(x, y) the measurement of the steady state EPR
D − νx =0 (22)
∂y 2 ∂x signal as a function of the electrode cur-
rent and flow rate, and using variable flow
where D is the diffusion coefficient of the rates to probe the radical lifetimes. Alter-
electroactive species, c(x, y) its concentra- natively, if the electrogenerated radical is
tion, x, the distance along the channel unstable, transient EPR signals may be
starting from the upstream edge of the recorded (at constant magnetic field in-
electrode, y the distance normal to the elec- duction) when the working electrode is
trode starting from the electrode surface, open-circuited, after a steady state has
and νx , the velocity of the solution in the x- been achieved. These are considered in
direction. Equation (22) reflects the effects turn for the case of a radical reacting
of convection in the direction of flow and via first-order kinetics. Physical applica-
diffusion in a direction (y) perpendicular tion of these methods is reserved until
to this. Diffusion in the direction of flow later, Sect. 3.2.4.
can be shown [62] to be negligible for the For the steady state case, the EPR
case of macroelectrodes, although this is signal intensity, S, can be calculated by
not the case for microelectrodes [63]. Since convoluting the sin2 sensitivity of the
the velocity profile is parabolic, cavity with the number of spins. Figure 19
Fig. 19 The coordinate system Cavity length

defining the location of the l
channel electrode in the y = 2h
EPR cavity. Flow
y=0
−xu 0 xc xe xd
shows the coordinate system defining the state to yield a current|EPR signal|flow rate
position of the electrode relative to the relationship for two cases. First, where
EPR cavity. xc denotes the center of the there is uniform concentration of radi-
cavity and x = 0 is the position of the cal at the electrode surface, and second,
upstream edge of the electrode. The cavity where there is a uniform current den-
length, l = xu + xd , where −xu and xd sity at the electrode. The first case holds
denote the x-coordinates of the upstream at all points on a reversible voltammet-
and downstream edges of the cavity, ric wave and for an irreversible wave if
respectively. Thus, xc = xd − l, where l is the current is diffusion-limited; the second
typically 24 mm and xc is 15 mm. The case holds near the foot of an irreversible
sin2 sensitivity profile allows maximum wave. In both cases, it was found that the
sensitivity at the center of the cavity steady state EPR signal for a stable radical
(x = xc ), whilst the sensitivity at the cavity followed the following dependence upon
edges (−xu and xd ) is effectively zero. the mass transport-limiting current and
Thus, integrating the spins throughout flow rate,
the cavity, i
S ∝ 2/3 (28)
xd
2 (x − xc )π
Vf
S = S0 sin
−xu This dependence was experimentally ver-
2h ified for the case of the reduction of
× c(x, y)dy dx (26) fluoroscein to the semifluorescein radical
0
anion [55]. For electrogenerated radicals
where S0 is the signal due to one mole of that subsequently decay by first-order ki-
the EPR active species located at the center netics, the dependence of S upon the rate
of the cavity, and the concentration of constant of the decay, k, is most con-
the paramagnetic species can be deduced veniently expressed in terms of an EPR
from Eq. (22). Since the radical species is detection efficiency, MK , given by
generated at the electrode, the integral can
2/3
be shortened to SVf
xd MK = (29)
(x − xc )π i
S = S0 sin2
0 where K is the normalized rate constant
2h for the radical decay,
× c(x, y)dy dx (27) 1/3
0
4h4 2 d 2
This equation can be solved either analyt- K=k (30)
9Vf2 D
ically [55, 58] or numerically [65–68] using
the Backward Implicit approach [69–71] For a stable radical, M0 is constant.
for the case of macroelectrodes at steady Figure 20 shows how the ratio MK /M0
1.0
MK
M0
0.5
0.0
0.0 2.5 5.0 7.5 10.0
K
Fig. 20 The variation of the EPR detection efficiency, MK , with the
normalized rate constant, K, for different sized electrodes (+, 2 mm;
×, 3 mm; O, 4 mm) located centrally in the cavity.
varies with K for an electrode positioned at EPR cavity, since all the radicals are found
the center of the EPR cavity for electrodes adjacent to the electrode surface and de-
of practical dimensions. In these cases, cay before they can be transported away
the EPR signal and the Faradaic current from the electrode; in contrast, at lower
are measured experimentally, for a range rates of decay, the curve deviates from
of flow rates, permitting the inference of sin2 behavior, and a maximum is observed
K from the appropriate value of MK in the upstream of the cavity center. Hence, al-
working curve (Fig. 20). It follows from the though the highest radical concentration
definition of the normalized rate constant is at the electrode surface, by generating
−2/3
that a plot of K against Vf allows the radicals above the maximum sensitivity, a
deduction of the rate constant for the decay greater number of radicals occupy the EPR
process, k. active area.
In Fig. 21, the quantity MK is normal- Attention is now turned to the use of
ized with respect to the value MKref obtained EPR signal transients in the deduction
when the electrode is located at the cavity of radical decay kinetics. The transient
center. This figure describes the EPR sig- signals may be experimentally observed
nal dependence upon location of a 5 mm (at constant magnetic field induction) by
electrode within the EPR cavity. For high either open-circuiting an electrode that
rate constants (K ≥ 5.0), the signal fol- had been previously held at a potential
lows the sin2 sensitivity profile of the at which Faradaic processes occur, or
1.5
1.0
M 0ref
MKref
0.5
0.0
0.0 0.5 1.0
xu/I
Fig. 21 The variation of Mref ref
k /M0 for a 5 mm electrode with position of the
electrode in the cavity for different rate constants (K = 0.0, ×; K = 1.0, O;
K = 5.0, ; K = 7.5, ).
by stepping the potential between two S(τ ) = S(τ → ∞)

defined values. The former method is nor- ∞
1 bn (K + p)n Jn
mally adopted, since in the latter method,
p 2n 5
contributions to the radical decay may oc- n=0 ? +
3 3
cur from, for example, the reoxidation × L−1 ∞ (32)
of a radical-anion if it is generated by bn K n J n

an electrochemically reversible reduction. 2n 5
n=0 ? +
Theory describing the transient EPR sig- 3 3
nals associated with the growth (when the
current is stepped up from zero) or de- in which τ is the normalized time variable,
cay (when the current is stepped down to  1/3
2/3
zero) of the electrogenerated radical upon 9Vf D
τ = t  4 2 2 (33)
a galvanostatic step at the working elec- 4h d
trode, for first-order kinetics is given else-
where [72]. For macroscopic electrodes, L−1 represents the inverse Laplace Trans-
the relevant convective-diffusion equa- formation (variable p), and ?(x) is the
tion is Gamma function [73]. The coefficients bn
∂C(x, y) ∂ 2 c(x, y) ∂c(x, y) are given by
=D − νx
∂t ∂y 2 ∂x 1 1
∞ ? n +
− kc(x, y) (31) 3(n−2)/3 3 3
bn =
π n!
where k is the first-order rate constant. The n=0

temporal EPR signal for a growth transient 2
× sin (n + 1)π (34)
can be shown to be, 3
and Jn are given by the expression for a decay transient is

xd
x − xu S(t) = S(t = 0) exp(−kt) (36)
Jn = sin2 π
0
This can be rationalized on the basis of the

x 2(n+1)/3 x − xe 2(n+1)/3 decay within a convection-free ‘‘reaction
× − layer’’ immediately adjacent to the elec-

trode surface: the paramagnetic species
x − xe x
×U d (35) are assumed to decay before they can dif-
fuse sufficiently far from the electrode for
them to experience a significant amount
where U(x) is the unit step function.
of transport by convection. Experimental
Theoretical growth transients for various
transients obtained from the reduction
values of normalized rate constant are
of fluorescein to the semifluorescein rad-
shown in Fig. 22 for an electrode located at
ical anion, and from the reduction of
the cavity center. For fast mass transport 2-nitropropane verified these results [58,
(or sluggish kinetics), the transient is 72, 74].
governed by both the mass transport of This cell has been applied successfully
material throughout the cavity and by for in situ electrochemical EPR, deducing
the kinetic decay or the growth of the both kinetics and reaction mechanisms,
radical. As the normalized rate constant, for radicals with lifetimes typically in the
K, increases (or as the mass transport range 10 to 100 ms: Sect. 3.2.4 will exam-
decreases), the time taken for the transient ine its application to ECE [58], DISP1 [74],
to reach a steady state value decreases, EC [68], and comproportionation reaction
and the transient changes shape until they mechanisms [75–77]. Other applications
attain a simple exponential form, which include the study of an ECEEE mechanism
1.0
S(t)/S(t ⫺ ∞)
0.5
0.0
0.5 1.0 1.5
t
Fig. 22 Calculated transients for a 4 mm electrode located at the cavity center
for different values of K (K = 0.0, O; K = 0.1, ; K = 1.0, ; K = 10.0, ;
K = 20.0, ).
for the reduction of nitromethane in aque- tube, x is the distance down the tube, D
ous alkaline solution [78]. One last point to is the diffusion coefficient of the species
note is that since the cell is constructed of a generated on the electrode, and c is its
transparent substance, silica, illumination concentration. However, recognizing that
of the electrode is possible, permitting pho- an ‘‘ideal’’ electrochemical cell for both
toelectrochemical EPR [55, 65–67, 79–81]. kinetic and mechanistic studies is of an
in situ design with the electrode located
3.2.3.4 The Albery In Situ Tube Electrode within the EPR cavity for maximum sen-
Cell sitivity, with well-defined and calculable
flow, experiments were undertaken using
3.2.3.4.1 Cell Design Albery and cowork- an in situ tubular electrode [83]. Figure 24
ers [9–14] used tubular electrodes for ex shows how the sensitivity of a standard
situ electrochemical EPR experiments. rectangular TE102 cavity to radicals varies
The tubular electrode is equivalent to the with distance through the EPR cavity ().
channel electrode in all respects, except With the tube electrode outside the cavity
that the cross section is circular rather (Curve a), the expected sin2 dependence
than rectangular [82, 137]. Like the later- is observed [35]; moving the tube elec-
developed channel flow cell, this setup trode into the cavity (Curve b) effects not
(shown in Fig. 23) permits the interroga- only a lower sensitivity, but also a grossly
tion of electrode reaction mechanisms of distorted sin2 curve. Furthermore, the sen-
relatively long-lived radical species, [9–14] sitivity is negligible in the vicinity of the
since the convective-diffusion equations electrode, irrespective of the placement of
are mathematically well defined, which at the latter within the cavity. It was found
steady state are given by Eq. (37) that the sensitivity of the EPR cavity is not
2 distorted if the tube electrode is replaced
r ∂c ∂ 2 c 1 ∂c by a semiannular tube, as shown in Fig. 24
ν0 1 − =D +
r0 ∂x ∂r 2 r ∂r (Curve c). With the semiannular tube elec-
(37) trode positioned centrally within the EPR
where ν0 is the velocity of flow at the center cavity, the expected sin2 sensitivity pro-
of the tube, r0 is the radius of the tube, r is file is observed with the electrode centered
the radial distance from the center of the at the maximum in the sensitivity curve,
r0
Fig. 23 The semiannular tube

electrode as used by Albery and
coworkers. The fine-hatched
segment is made of platinum xe
and the course-hatched
segment of Teflon . The
parabolic velocity profile is
shown. Typical dimensions are
radius, r0 = 0.5 mm and the
electrode length, xe = 2.0 mm.
d
50
S/%Mn2+
25
a
c
b
0
0.00 0.25 0.50 0.75 1.00
x/l
Fig. 24 Variation of the sensitivity of the cavity with distance through
the cavity. Curve a is obtained with no electrode inside the cavity;
Curve b is obtained with a tube electrode consisting of a complete
annulus at the cavity center. Curves c and d are obtained with an
electrode at the center of the cavity consisting of a half annulus and an
annulus with a missing sector, respectively.
albeit reduced by a factor of two, but never- equation [10],

theless adequate for EPR measurements.
2/3 1/3
Curve d in Fig. 24 shows the sensitivity ilim = 2.75nF D 2/3 xe Vf cbulk (38)
when an annular electrode with a sector
subtending 10◦ has been removed. The in which xe is the electrode length and Vf
sin2 function is now distorted, but the sen- the volume flow rate. Equation (38) was
sitivity at the electrode is enhanced by a verified by experimental observations on
factor of three when compared to Curve a. the oxidation of N ,N ,N ,N -tetramethyl-
The difference between a complete annu- para-phenylene diamine (TMPD) in aque-
lar electrode and a part-annular one is ous sulfate solution. The diffusion coef-
that in the former, the magnetic com- ficient calculated from these data was in
ponent of the standing microwave field agreement with literature values, showing
induces eddy currents in the complete cir- the semiannular tube electrode to be a
cle and therefore is unable to penetrate to satisfactory hydrodynamic electrode.
the electrolyte solution. These eddy cur- As with the channel electrode cell,
rents are minimized using an incomplete and following from the previous work
annulus. In the case of the semiannular undertaken using the ex situ tube electrode
electrode, there is a greater loss of sensi- arrangement, the EPR signal (S) is a
tivity, but the standing microwave field is function of the diffusion-limited current
preserved. (i) and the volume flow rate,
Application For a semiannular i

3.2.3.4.2 S = S0 Ix/ ( − xe )2/3 r02 2/3
tubular electrode, the limiting current is nF D 1/3 Vf
given by the appropriate form of the Levich (39)
where S0 is the EPR signal from one The cell itself (see Fig. 25) comprises a
mole of spins at the center of the cavity, TE011 cylindrical cavity that is converted
and Ix/ is a numerical factor that arises into a coaxial cavity by the addition of
from convoluting the concentration profile a smooth, polished copper rod that is
with the sensitivity of the cavity; for the positioned centrally within the cavity. The
electrode just outside the cavity, I0 = 0.53, rod itself (diameter 9 mm) is located within
while when the electrode is at the cavity a precision-bore silica tube, so that there
center, I1/2 = 0.47. Experimental results, is an annulus (of size 2h) around the rod,
for the study of radicals in the coats constrained to be of uniform thickness by
of modified electrodes [83, 84], verified a fine nylon thread running the length
the predicted S ∝ i/Vf
2/3
dependence, of the cavity. Typically 2h ∼ 100 µm, but
suggesting that like the channel electrode this is adjustable by polishing the rod.
EPR flow cell, this in situ cell is easily The central 4 mm of the copper rod is
amenable to the study of electrode reaction insulated from the rest of the rod, as this
mechanisms. However, it suffers from part is the working electrode. To ensure
several disadvantages when compared to a large negative potential window, this
the channel cell. First, the Albery cell is not central part of the rod was plated with
readily demounted, making examination mercury. As in the case of the channel cell,
and polishing of the electrode difficult. this cell was connected to a reservoir-fed
Second, it is uncertain whether eddy gravity flow system, with a platinum gauze
currents are induced in the electrode counter electrode located outside the cavity
by the 100 kHz modulation, rendering and sufficiently away downstream so as to
areas of the electrode EPR-insensitive. preclude counter electrode products from
diffusing into the cavity, and an upstream
Third, the channel electrode does not
reference electrode. Electrolyte solutions
suffer as large an ohmic drop. Fourth,
flow through the annular gap between the
irradiation of the electrode is not possible,
copper rod and the silica tube. Flow rates
rendering this in situ cell inadequate
within the ranges 10−4 to 10−1 cm3 s−1
in the study of photoelectrochemical
were obtained in this manner. Since the
phenomena.
flow pathway is narrow and resistive,
the potentiostat was modified so as to
3.2.3.5 The In Situ Coaxial Flow Cell provide high voltages to drive the current
through the channel. Electrolyte resistance
3.2.3.5.1 Cell Design Waller and Comp- in the narrow gap through which the
ton [85] effectively mimicked the Allendo- solution flows may lead to a non-uniform
erfer coaxial design (vide supra), whilst potential difference (vide supra), affecting
simultaneously maintaining the mathe- the voltammetry only in poorly conducting
matically well-defined laminar flow of the nonaqueous solutions using low currents.
channel flow cell. This improved cell for Like the Allendoerfer coaxial cells (vide
electrochemical EPR [85] allowed an im- supra), it was found that the resonant
provement in the channel cell regarding frequency of the cavity shifted above
lifetimes of radicals amenable to study, that of the empty cavity and outside
whilst retaining the hydrodynamic flow the klystron tuning range as a result of
that is essential for the investigation of the reduced effective cavity size produced
electrode reaction mechanisms. by the addition of the copper rod. As
35
A
B
115 40
C 4
D
E
40
35 55
F
15
G
11
9
Fig. 25 The in situ coaxial electrochemical EPR cell. A, Teflon

annulus, B, Teflon insulator, C, mercury-plated copper
electrode, D, copper, E, TE011 cylindrical cavity, F, Teflon
sheath, G, precision-bore silica tubing. The numbers shown
represent the dimensions in mm.
above, this problem was eliminated by around the copper rod at a fixed distance
partially filling the cavity with a material of from the cavity. Since the cylindrical cavity
appropriate relative permittivity (such as a sets up cylindrically symmetric standing
PTFE shielding around the silica tubing), microwave field patterns, this behavior can
shifting the resonant frequency back to only be rationalized by a perturbation of
typically 9.4 GHz. the component of the magnetic field that
The sensitivity of this modified cavity, as is modulated at 100 kHz, caused by eddy
observed with a small crystal of DPPH, was currents induced in the copper rod by the
found to follow a sin2 function along the 100 kHz component. Lenz’s Law requires
length of the cavity and at a fixed distance the induced magnetic field to oppose the
from the copper rod, as anticipated theo- applied field, and this effect is maximally
retically [35]. However, the EPR sensitivity observed when the applied field is per-
was observed to vary in a cos2 manner pendicular to the copper surface, and zero
when it is parallel to the copper surface. xe is the electrode length, D is the dif-
This cos2 variation in sensitivity has also fusion coefficient of the electrochemically
been observed in the Allendoerfer coaxial active species, and n is the number of elec-
cell [46, 48]. This effect inevitably means trons transferred per mole reactant during
that paramagnetic species will experience the electrode reaction. The applicability of
different amplitudes of the modulated field this equation was tested using the known
depending upon their location within the one-electron reduction of fluorescein in
cell. However, it was shown that this will aqueous alkali, to yield the semifluores-
reduce the EPR signal by only a factor cein radical anion, and it was found that
of one half from that which would be the well-defined current|potential curves
otherwise seen. observed gave rise to linear Tafel plots,
with the expected room temperature value
of 59 mV decade−1 slope. A linear Levich
3.2.3.5.2 Application Laminar flow was plot allowed inference of a diffusion coeffi-
established using flow rates such that the cient of 3.0 × 10−6 cm2 s−1 , in agreement
Reynolds Number, Re < 10. Thus, after with literature values [87]. Thus, the hy-
sufficient lead-in (specifically 0.1 × Re × drodynamics of this particular flow cell are
h), which in this case is negligible, a such that there is the expected parabolic,
parabolic velocity profile develops across laminar flow.
the flow-path. In this manner, the hydro- Since the hydrodynamics are analogous
dynamics of this electrochemical setup are to those experienced in the channel flow
equivalent to that of the channel electrode cell, it follows that steady state EPR signals
flow system; the mass transport-limiting should behave analogously, that is,
current is therefore given by the Levich
i
equation [86], S∝ 2/3
(41)
Vf
ilim = 0.7767nF cbulk D 2/3 ν0 h−1/3 σ xe

1/3 2/3
This equation should hold, providing the
(40) diffusion-layer thickness at the electrode
in which σ is the circumference of the is small when compared to the thickness
electrode, ν0 is the mean solution velocity, of the flow-path. Figure 26 shows data
0.0
−0.5
log10 s/i
[a.u.]
−1.0
−3.0 −2.0
Fig. 26 Steady state EPR
log10(Vf /cm3 s−1)
signal|current|flow rate data.
obtained from the fluorescein system 3.2.3.6 The Bond In Situ Microelectrode
plotted according to Eq. (41). A straight line Cell
is observed for a wide range of flow rates;
the line drawn has a gradient of −2/3, 3.2.3.6.1 Cell Design Bond and cowork-
as predicted by the assumed theory. The ers have designed various cells suitable
deviation at low flow rates is entirely due to for in situ electrochemical EPR [90–94].
the diffusion layer becoming comparable These cells generally have a small vol-
in size to h; less radical is produced than is ume and permit the recording of variable
predicted by assuming a wide channel [70, temperature EPR. In this section, only
89]. Hence at sufficiently slow flow rates, the most recently designed cell [94] will
there is near-exhaustive electrolysis, and be discussed.
the cell behaves as a thin-layer cell [88], in Figure 27 illustrates the all-glass cell
which the current is given by used for electrochemical EPR over a wide
temperature range. The reference elec-
4
i= nF ν0 hσ cbulk (42) trode (either an Ag|AgCl or platinum)
3 is fixed in the center of a Pt working-
The full transition between the two electrode coil, the latter extending to the
limits has been calculated [70, 89] and bottom of the quartz tube. These elec-
corresponds to a flow rate of 10−3 cm3 s−1 trodes have diameters of less than 70 µm,
for xe = 4 mm and 2h = 100 µm. The EPR so as to minimize the amount of metal
signal will be more sensitive to these present in the EPR cavity; the counter elec-
effects than the observed current, since trode is situated 15 mm above the working
the EPR signal reflects the concentration electrode outside the cavity, also permit-
of radicals throughout the channel, not ting interference-free measurements from
merely the concentration gradient at the species diffusing from the counter elec-
electrode surface. trode during electrolysis. With this design,
Cu rods
Soldered connections
Nylon support
PVC sleeves
B7 joint (quartz socket,

pyrex cone)
180
150
149
125
‘O’ ring
Average filling level
(approx. 0.25 ml)
Counter electrode
Sensitive region of cavity
Reference electrode
(70-µm Pt coated with Ag/Agcl) Fig. 27 Schematic diagram of
Working electrode the in situ Bond cell for
2 (70-µm Pt) microelectrodes. All lengths are
4 given in mm.
only a small volume of solution may be transport is the flow due to a jet of
placed into the bottom of the cell, typically electrolyte striking a planar electrode
0.2 cm3 , allowing even lossy solvents to be surface at right angles, and rapidly spreads
used in electrochemical EPR work. Fur- out over the electrode surface, the flow
thermore, by allowing complete sample outside the jet being at rest [97]. Like the
handling under an inert atmosphere, the channel electrode, the wall-jet electrode
cell facilitates the study of electrochem- is a non-uniformly accessible electrode,
ically generated radicals that are highly that is, the diffusion layer is not constant
oxygen and/or moisture sensitive. Further- over the working electrode surface, making
more, it was observed that this cell has a this electrode geometry successful in
low RC time constant, making fast scan mechanistic discrimination [56].
cyclic voltammetry viable [94]. The wall-jet electrode for in situ elec-
trochemical EPR [98] is shown in Fig. 28.
This arrangement is in fact a refinement
3.2.3.6.2 Application This cell has been
to an in situ cell designed by Scholz and
used for a variety of studies such as the oxi-
coworkers Fig. 29 [93], in which a jet of
dation of Cr(CO)2 (Ph2 PCH2 CH2 PPh2 )2
electrolyte impinges on a wire or mer-
[94], the reduction of [Cp2 Ti(acac)]ClO4
cury drop electrode. The modifications by
(Cp = η5 –C5 H5 , acac = acetylacetone)
Compton and coworkers [98] allow for con-
[94], dipropylpyridine-3,4-dicarboxylate
trolled convection and enhancement of its
[95], and S,S-dipropylbenzene-1,3-dicarbo-
sensitivity towards unstable radicals. The
thioate [95]. Generally, good voltammetric
cell itself consists of a silica tube with
curves were obtained at room temperature,
a mercury working electrode placed at
but these were often distorted at low tem-
one end. Solution flows into the cell via
peratures (typically 180 K), possibly due to
the steel jet so that it impinges normally
iR drop or due to the onset of thin-layer
and centrally on the electrode surface.
electrochemistry [94].
The system was connected to a gravity-
This cell has been recently compared
fed flow system, with a silver or platinum
with the channel flow cell [96]. Although
pseudo-reference electrode positioned up-
the Bond cell has many advantages, its ma-
stream of the mercury electrode and a
jor limitation is that it does not permit the
platinum gauze counterelectrode located
direct interrogation of electrode reaction
downstream of the cell.
mechanisms through the EPR signal in
The sensitivity profile of the cavity to rad-
an analogous way to that by Compton and
icals with the wall-jet cell placed along the
Coles (vide supra). Furthermore, the chan- vertical axis of the cavity, corresponding to
nel electrode cell is much more sensitive in a node in the electric component of the
detecting paramagnetic species than this standing microwave field, with the work-
cell by some orders of magnitude [96]. ing electrode surface at the cavity center,
exhibited little distortion from the theoret-
3.2.3.7 The In Situ Wall-jet Electrode Cell ically predicted sin2 behavior, even though
large amounts of metal (mercury and cop-
3.2.3.7.1 Cell Design As described in per) are present. Measurement of the cavity
Chapter 2.2 of this volume, a wall-jet Q-factor when the wall-jet cell containing
electrode is a well-defined hydrodynamic acetonitrile and 0.5 M tetrabutylammo-
electrode in which the convective mass nium perchlorate supporting electrolyte
Solution from Fig. 28 Schematic illustration

reservior of the in situ wall-jet EPR cell.
Reference
electrode
“Altex” T-piece
To
counter electrode
od = 0.4 cm
id = 0.2 cm
Glass
tubing
Jet 17.5 cm
Mercury working
electrode
Copper
wire
Solution were present in the EPR cavity, revealed

intlet Q = 1000, which is a comparable value to
that for the channel cell (vide supra).
RE
AE 3.2.3.7.2 Application Albery [99] and Ma-

tsuda [100] have derived an approximate
Solution equation for the mass transport-limited
outlet
current for an n-electron transfer reaction
at a wall-jet electrode,
ilim = 1.59kc nF D 2/3 ν −5/12 Vf

3/4
× a −1/2 R 3/4 cbulk (43)
in which kc is a constant determined by

experiment to be close to 0.90 [99–101],
ν is the kinematic viscosity, a is the
diameter of the jet, and R is the radius
of the electrode. Analysis of the well-
defined voltammograms for the reduction
Fig. 29 Schematic of the Scholz design

electrochemical EPR cell for use with
WE flowing solutions.
of 1,4-benzoquinone in acetonitrile in to 0.15 cm. Although satisfactory hydrody-

terms of Eq. (43), by plotting ilim against namic behavior was observed, when the
3/4
Vf revealed a straight line, and gave a cell dimensions were reduced for EPR
diffusion coefficient in good agreement compatibility and using a cell of the exact
with literature values, indicating that the type to that described by Scholz [93], this
hydrodynamics of this cell are well defined was no longer the case – significant devi-
and understood. ations from wall-jet behavior were found
It has been seen that the other in whatever the separation between the work-
situ hydrodynamic cells discussed above, ing electrode and the jet. Reproducibility
the channel electrode cell and the tube problems were additionally encountered,
electrode cell, both exhibit EPR signal (S) in part due to the trapping of air bubbles
characteristics dependent upon the ratio within the cell, and due to a greater sensi-
2/3 tivity of the cell to the exact location of the
i/Vf . However, in the case of the wall-
jet electrode under steady state conditions, jet, pointing to loss of the wall-jet behav-
in which the entire cell downstream of the ior and to the existence of ‘‘edge effects’’
working electrode is filled with radicals, caused by the walls of the cell.
the rate of loss of radicals from the cell In the Scholz design, the separation be-
is proportional to Vf , while the rate of tween the mercury working electrode and
their formation is proportional to i, thereby the jet is minimized so as to give rise to
anticipating a ‘‘thin-layer cell’’ as shown in Fig. 30.
i Near-exhaustive electrolysis of the elec-
S∝ (44) troactive material is thus likely to occur,
Vf
leading to large currents and a high sensi-
This relationship was verified experimen- tivity towards the detection of radicals that
tally for the benzoquinone/acetonitrile sys- are stable on the experimental timescale;
tem. The EPR signals were observed with the shape of the mercury electrode is
excellent signal-to-noise ratio. such that it may screen the working-
Analogous experiments in aqueous so- electrode|electrolyte interface from the
lutions gave similar responses. These 100 kHz magnetic field modulation used
present a more stringent test in terms in conventional EPR spectrometers, so that
of EPR compatibility since they are notori- only radicals downstream of the shielded
ously ‘‘lossy’’; in order to reduce dielectric area will be EPR active; the wall-jet ap-
loss to acceptable levels, the internal di- proach may be more suited in not only
ameter of the glass tube was reduced observing mechanistic pathways, but also
Flow
Fig. 30 The flow pattern

between the jet and the mercury Hg
working electrode under
‘‘thin-layer’’ conditions.
in the determination of EPR spectra from in homogeneous follow-on reactions. Fur-

short-lived radicals. Furthermore, since thermore, EPR alone cannot give struc-
the hydrodynamics of the Scholz flow- tural information concerning diamag-
electrode design are not mathematically netic reaction participants. Neudeck and
well described, it limits the use of its de- coworkers [102–108] combined UV-VIS
sign for the study of electrode reaction and EPR spectroscopies to give insights
mechanisms and kinetics. into whether primary or non-primary rad-
icals are formed during electrolysis.
3.2.3.8 The Dunsch–Petr–Neudeck Cell
for In Situ UV-VIS EPR Spectroelectro- 3.2.3.8.1 Cell Design The in situ elec-
chemistry trochemical UV-VIS-EPR cell is shown
In situ electrochemical EPR is a versa- schematically in Fig. 31. A TE102 cavity
tile spectroelectrochemical technique as was modified to permit optical transmis-
it allows quantitative inference of both sion. Fiber-optic light guides were used
the structure and concentration of para- for the illumination of the EPR cavity as
magnetic species formed during electro- well as to connect the cavity to a pho-
chemical processes. However, this it is todiode array detector, the latter so as
often difficult to distinguish between para- to improve UV-VIS measurements. The
magnetic species that have been formed light guides were arranged near the walls
by the heterogeneous transfer of an of the cavity so that there was no loss
electron and those formed subsequently in EPR sensitivity. Further, since the
Working Reference
electrode electrode
Second Gas inlet

counter electrode
Adjustable
cell holder
ESR-cavity
To photodiode From light source

array detector
Modulation
coils
Fig. 31 Schematic of the cell

used for simultaneous in situ
First electrochemical EPR and
counter electrode UV-VIS measurements.
distance between the two light guides is formed. This was found to be so in the case
just the cavity width, the sensitivity of the of TMPD. The absorbance is also related
UV-VIS measurement is maximal. In or- to the concentration of the absorbing
der to minimize photochemical reactions species. Measurements were conducted
from taking place, the light intensity used using 560-nm light, corresponding to the
was kept to a minimum. absorption of light by TMPD+• . Since
The electrochemical cell itself consisted the concentration of this species is not
of essentially a flat cell, fixed inside in uniform, the Beer–Lambert law is replaced
the optimal position within the EPR cavity by a differential equation relating the
so as to ensure maximum sensitivity. intensity of light transmitted (I ) to the
The working electrode employed was concentration of radicals (c),
either a platinum mesh [102] or later,
a partly laminated gold mesh [103, 104]. ∂I
= −εc(x)I (45)
An Ag|AgCl electrode was used as the ∂x
reference, whilst two counter electrodes
were employed to minimize oscillation of where ε is the extinction coefficient at
the cell voltage. The first counter electrode, a fixed wavelength, and x is the thick-
a palladium sheet, is situated below the ness of the solution layer. Since the
flat part of the cell, whilst the second, radical species is formed in the diffu-
connected by a resistor to the potentiostat, sion layer at the electrode, changing the
is located above it. scan rate of the cyclic voltammetric experi-
ment perturbs the concentration gradient.
3.2.3.8.2 Application The cell was tes-
It was found that in the range 20 to
ted with two electroactive compounds, 200 mV s−1 , the EPR signal|potential and
N ,N ,N ,N -tetramethyl-para-phenylene- the UV-VIS absorption|potential curves
diamine (TMPD) and para-aminodiphenyl- match up exactly, thereby allowing a
amine (PAD) in acetonitrile solutions. semi-quantitative measurement of radi-
The former compound undergoes a one- cals, or UV-VIS absorbing compounds.
electron oxidation to yield a dark blue At high scan rates (500 mV s−1 ), the
cation radical, TMPD+• whilst the lat- EPR signal|potential curve matches the
ter is thought to give rise to non- integrated charge|potential curve, but de-
primary radicals upon oxidation, by com- viates from the absorption curve, re-
proportionation of the oxidation prod- flecting a greater concentration gradient
uct, N -phenylquinonediimine with para- at the electrode surface. At low scan
aminodiphenylamine. In this manner, rates (10 mV s−1 ), further deviation due
absorbance|potential data (at a fixed wave- to diffusive processes occurs, permitting
length) can be obtained together with the distinction between primary and non-
EPR signal|potential and cyclic voltammet- primary radicals; at these scan rates, if
ric (current|potential) data. non-primary radicals are formed, neither
For long-lived primary radicals, the the EPR signal|potential curve nor the
EPR signal|potential curves should directly absorption|potential curve coincides with
overlay a suitable normalized integrated the charge|potential curve.
charge|potential curve, since the former This cell design has been used to study
reflects the number of spins in the cavity conducting polymer films at electrode
and the latter, the number of radicals surfaces [104–108].
3.2.3.9 The Dryfe–Webster Cell for Charge and is essentially a conventional silica
Transfer Across a Liquid|Liquid Phase flat cell [17] modified so that counter and
Boundary reference electrodes can be inserted into
both ends. It is assembled by first fill-
3.2.3.9.1 Cell Design In situ EPR spec- ing the higher density solvent through
troscopy is an ideal method to detect the capillary at the base of the cell us-
and identify paramagnetic participants in ing silicone tubing and a syringe, so
liquid|liquid processes, but to our knowl- that the solvent reaches the midpoint of
edge, has only been undertaken by Dryfe the flat portion of the cell, and then
and coworkers [109, 110]. The cell used sealing the silicone tubing using a Hoff-
to perform the interfacial polarization- mann clamp. The upper, less dense phase
EPR experiments is shown in Fig. 32, is then introduced into the cell from
the top using a Pasteur pipette. Using
this filling method, a liquid|liquid inter-
face can be generated and maintained
in the midportion of the flat part of the
cell.
The EPR cell has two coiled platinum
wire counter electrodes located outside the
thin-layer section, and two silver wire ref-
erence electrodes flattened at one end to
a thickness that enables their insertion
(b)
into the flat part of the cell, positioned
within 5 mm of the liquid|liquid inter-
(h) face. To obtain optimal control of the
potential difference between the liquid
phases, the portions of the reference
(c)
(a) electrodes outside the flat cell were jack-
eted in PTFE tubing and sealed so that
5 cm they are only in contact with the solu-
(g)
tions inside the flat part of the cell. The
(e)
(h) Fig. 32 Silica cell for performing

liquid|liquid electrochemical EPR
experiments: (a) interface between two
immiscible liquids, (b), platinum wire
counter electrode 1, (c) silver wire
reference electrode 1, (d) platinum wire
counter electrode 2, (e) silver wire
(i) reference electrode 2, (f) capillary to fill
lower portion of cell, (g) thin-layer
portion of cell, (h) Teflon /silicone
(d) rubber sleeves surrounding lower
portions of reference electrodes,
(j) (i) electrical contact to reference
electrode 2, and (j) electrical contact to
(f) counterelectrode 2.
inherent high resistivity of a flat cell com- generated in the appropriate single-
bined with low dielectric solvents (such phase three-electrode cell [17] using
as 1,2-dichloroethane, DCE) can be par- DCE and the same supporting electro-
tially overcome using this arrangement, lyte, bis(triphenylphosphoranylidene)am-
and reasonable linear sweep voltammetry monium tetrakis(4-chlorophenyl)borate
was observed using a four-electrode poten- (BTPPA+ TCPB− ). The broad EPR spectra
tiostat. All experiments were undertaken were thus explained by the strong ion-
in a TE102 cavity. pairs formed between the product radical
ions and the organic phase cation
3.2.3.9.2 Typical Applications EPR spec- or anion, respectively undergoing fast
tra were obtained for both the interfacial electron exchange with neutral TCNQ or
reduction of 7,7,8,8,-tetracyanoquinodi- TTF [113].
methane (TCNQ) and oxidation of tetrathi- EPR time-sweep experiments (see
afulvalene (TTF), when dissolved in DCE Fig. 33) for the oxidation of TTF,
by the aqueous phase ferri/ferrocyanide in which the magnetic field is held
redox couple, following the application constant, corresponding to the maximum
of a potential difference directly to the adsorption, by applying a potential
liquid|liquid interface. Previous work [111, difference (φ1 ) between the silver quasi-
112] suggested that a charge-transfer pro- reference electrodes located in each
cess occurs at the liquid|liquid interface, phase sufficient to cause an electron
due to the heterogeneous reduction of to be transferred across the water|DCE
TCNQ by the aqueous couple, interface from TTF to Fe(CN)6 3− , showed
increases in the EPR signal intensity,
TCNQ(DCE) + Fe(CN)6 4− (aq) corresponding to the formation of
−• TTF+• , exhibiting a square-root temporal
−
D−−−
−
3−
−TCNQ (DCE) + Fe(CN)6 (aq)
C
dependence as expected for a potential-
(46) step occurring under planar diffusion
A similar mechanism can be written for (Volume 3 Chapter 2.3). Reversing, to a
the oxidation of TTF, potential sufficiently negative to drive the
back reaction (φ2 ), immediately causes
TTF(DCE) + Fe(CN)6 3− (aq) a decrease in the EPR signal intensity,
−− +•
−−
4− since the radical cation is reduced to
D −
C−TTF (DCE) + Fe(CN)6 (aq) (47)
neutral TTF. Similar observations were
The presence of the organic ion radicals, made with TCNQ. Figure 33 also shows
and hence the occurrence of an electron- that when the potentiostatic circuit is
transfer process, was verified using EPR broken after the formation of TTF+• , the
spectroscopy. The EPR spectra for TCNQ−• EPR signal decreases, as TTF+• , a stable
and TTF+• in DCE at room temperature cation radical in DCE, diffuses away from
consisted of single broad lines with the central, most sensitive part of the
Bpp of 5.2 and 3.0 G, respectively. The EPR cavity. Further application of this
broad lines were thought not to arise work to benzoquinones [110] permitted
by interaction of the radical ions with the deduction that at the aqueous|DCE
the aqueous phase iron complexes, since interface, reduction of the parent
broad EPR spectra were also observed from compounds to yield semiquinones occurs
TCNQ−• and TTF+• electrochemically without their immediate protonation.
Fig. 33 Time-sweep first derivative EPR

∆f1 ∆f2 ∆f1 ∆f2 ∆f1 spectrum of TTF+• recorded at constant
Open Open Signal
magnetic induction using liquid|liquid
circuit circuit intensity electrochemical EPR cell. The organic
[a.u.] phase (DCE) contained 5 mM TTF,
36 mM BTPPA+ TCPB− . The aqueous
phase contained 25 mM K4 [Fe(CN)6 ],
0 t 500 25 mM K3 [Fe(CN)6 ], and 0.1 M Li2 SO4 .
[s] EPR sweep time = 500 s, EPR sweep
width = 10 G, modulation
amplitude = 2.0 G, and time
constant = 10 × 10−2 ms.
φ1 = oxidizing potential,
φ2 = reducing potential. The potential
difference imposed to induce oxidation of
the TTF was +0.7 V, whilst this process
was reversed using a potential difference
of −0.1 V (versus the silver
pseudo-reference electrodes used).
3.2.4 like two one-electron processes, suggest-

Illustrative Examples of EPR in ing the presence of a radical intermediate.
Electrochemistry An EPR study followed, confirming the
presence of a radical, identified as being
Applications of in situ electrochemical essentially PAD+• . A typical EPR spectrum
EPR are presented in this section. of this radical is illustrated in Fig. 34. Since
the pKa ’s of PAD are 4.4 and −0.072,
3.2.4.1 Radical and Radical Ion a mechanism consistent with these ob-
Identification servations suggested by Allendoerfer and
Male is
3.2.4.1.1 The Anodic Oxidation of para-
− +
Aminodiphenylamine (PAD) Allendoerfer PADH−
D−
−−−
C−NPQH + 2e + 2H (48)
−− +•
and coworker [114] studied the oxidation PADH + NPQHD−−−
C−2PAD (49)
of PAD at platinum electrodes in aqueous
solution, in the pH range 1.2–4.8. Cyclic where NPQ is to N -phenylquinonedi-
voltammetry in buffered solution produces imine, and the extra H refers to the fact
a single peak, with Emid (= 0.5(Epa + that the single protonated states are in-
Epc ) at 0.40 V versus SCE. This potential volved. The comproportionation reaction
parameter was found to linearly shift with is postulated as the EPR spectrum corre-
pH in the range pH 1–3, with a gradient sponds to the unprotonated form of the
corresponding to a proton : electron ratio radical. Further, since only a single re-
of 1 : 1. Further, from the variation in versible two-electron process is observed
peak potential separation with pH, this electrochemically, the comproportionation
process was estimated to correspond to constant must be less than one.
two electrons. Compton and coworkers [13] observed
In unbuffered solutions at pH ∼5.0, the a similar comproportionation reaction for
cyclic voltammograms appeared to look the case of para-phenylenediamine.
Fig. 34 Experimental EPR 10G

spectrum of PAD+• .
3.2.4.1.2 The Reduction of Dicyanobenzene stable paramagnetic species. This tech-

Gennaro and coworkers [115] studied the nique was first introduced by Janzen and
reduction of 1,2-dicyanobenzene in aceto- Blackburn [121, 122], using N-tert-butyl-
nitrile at mercury electrodes. Two re- α-phenylnitrone (PBN) to spin-trap alkyl
duction signals were observed; the first and alkoxy radicals. In a similar manner,
(at −1.32 V vs. SCE) was assigned to the Wadhawan and coworkers [123] were able
1,2-dicyanobenzene radical anion, the sec- to propose the existence of alkyl peroxides
ond (at −2.35 V vs. SCE) attributed to during sono-emulsion Kolbe electrosyn-
further reduction of the radical anion, pro- theses. It must be stressed that case should
ducing the 1,2-dicyanobenzene dianion, be taken to avoid electroactive spin traps;
which was thought to immediately react Bard and coworkers [124] showed that
with the solvent, producing benzonitrile, spin trapping is possible during electro-
the reduction of which is thus observed. chemical experiments, and that PBN is
This reaction pathway was later confirmed electroinactive in the potential range 1.5
by Compton and Waller [16] using in situ to −2.5 V versus SCE in acetonitrile. The
electrochemical EPR: Fig. 35 shows the interested reader is referred to Kemp’s
EPR spectra observed by potentiostatting review on spin trapping [125] for further
at the appropriate voltages. information.
3.2.4.2 Electrode Reaction Kinetics and

3.2.4.1.3 The Reduction of C60 This com-
Mechanism
pound undergoes a series of reduc- This section shows the versatility of the
tions [116, 117]. Electrochemical EPR has channel flow EPR cell for mechanistic
been undertaken in toluene|acetonitrile discrimination.
mixtures [118], but it is notoriously dif-
ficult to unambiguously interpret these
spectra [119, 120]. 3.2.4.2.1EC Versus ECE These reactions
can be characterized by the general reac-
tion scheme [126]
3.2.4.1.4 Spin Trapping If the radical to
be investigated is extremely short-lived, it A ± e− D
−−
−−−−B
•
Electrode C (i)
may not be possible to observe it directly. In
k
these cases, the radical is ‘‘spin trapped’’ Solution B• −−−→ C (ii)
by chemically reacting it with a diamag-
netic compound to produce a relatively Electrode C ± e− D
−−
−−−
C− products (iii)
−•
CN CN
a a
b b
5G
(a)
CN −•
5G
(b)
Fig. 35 EPR spectrum obtained by in situ electrolysis of 1,2-dicyanobenzene at
(a) −1.32 V versus SCE, where the spectrum was shown to be that of
1,2-dicyanobenzene radical anion and (b) −2.35 V versus SCE, where the spectrum
corresponds to the benzonitrile anion radical.
Steps (i) and (ii) define a pure EC mecha- ECE type, whereas if C is electrochemically
nism, while Steps (i), (ii), and (iii) corres- inert, the process is designated EC. Ana-
pond to an ECE mechanism. In effect, the lytical theory describing these processes in
mechanisms differ in terms of the elec- channel electrodes is established [127].
troactivity of the product, C, of reaction Compton and coworkers [58] deduced
(ii) so that if C is more readily oxidized that the EPR signal for an ECE reaction fol-
or reduced than A, the process is of the lows that for a simple E reaction (Eq. 27),
2.0
1.8
1.6
Neff
1.4
1.2
0.0
−3 −2 −1 0 1 2
log10KECE
Fig. 36 Working curve showing the relationship of Neff and K for an
ECE process.
provided the current, i, is replaced by of simple first-order kinetics (Eq. 36) per-
that giving rise to unreduced B• , namely, mitted the inference of a rate constant,
2/3
(iA − iC ). Further, a working curve, shown k. Plotting this parameter against Vf ,
in Fig. 36 was described showing how a gave a straight line graph that allowed
dimensionless rate constant, K, defined extrapolation to zero flow rate to give
by a value of 0.36 ± 0.02 s−1 for the first-
1/3
4h4 xe2 d 2 order decay of the radical anion. The
K=k (50) steady state EPR signal was found to
9Vf2 D
be directly proportional to the generat-
ing current, until after the onset of the
varies with ‘‘the effective number of
second voltammetric wave, when a dra-
electrons’’, Neff , passed,
matic decrease of EPR signal occurred.
Total Limiting Current These features are both consistent with
EC and ECE mechanisms. Conclusive ev-
Neff = Observed
Limiting Current For A Single idence of the mechanistic discrimination
between EC and ECE reactions came from
Electron Transfer Process
(51) analysis of the reduction current|flow rate
The electrochemical reduction of 2-nitro- data, shown in Fig. 37. It can be seen
propane at mercury-plated copper elec- that at fast flow rates, the data tend to-
trodes was investigated in aqueous solu- wards one-electron behavior, whilst at slow
tion, using Triton-X as a stabilizer. Two flow rates, they tend towards two-electron
waves were observed, the first corres- behavior, indicative of an ECE reaction.
ponding to the [(CH3 )2 CHNO2 ]−• radical Further analysis of the EPR signal|flow
anion, as evidenced by electrochemical rates data permitted the determination of
EPR. Upon removing the applied potential EPR detection efficiencies, from which a
from the electrode, EPR signal transients rate constant of 0.28 s−1 was obtained for
were obtained, analysis of which in terms radical decay.
1500
1000
[µA]
Iobs
500
0
0.0 0.2 0.3
(Volume flow rate)1/3
[cm s−1/3]
Fig. 37 The limiting current|flow rate behavior for the reduction of
2-nitropropane in aqueous solution. The solid lines are calculated
behavior for one- and two-electron processes.
3.2.4.2.2 ECE Versus DISP1 DISP1 re- F + e− D

−−
−−−
C−S
•
actions are described by the following

k
general kinetic scheme. S• + H+ −−−→ SH+•
Electrode A ± e− −−−→ B• (i)
SH+• + S• −
D−
−−−
C−F + L
k
Solution B• −−−→ C (ii) where it was assumed that the protonation
of S• is rate determining and irreversible,
Solution −−
B + CD−−−
C−A + D (iii)
as it is accompanied by a change in
In this scheme, Step (ii) is rate limiting. hybridization from sp2 to sp3 at the central
Compton and coworkers [74] deduced the carbon atom. In a similar manner to that
analytical theory for the solution of the above, the current|flow rate data showed
relevant convective-diffusion equation at one-electron behavior at fast flow rates,
steady state for the above homogeneous but two-electron behavior at slow flow
kinetic scheme, in terms of a working rates, indicative of either an ECE or a
curve, see Fig. 38. DISP process.
The reduction of fluorescein (F) to give Analysis of EPR signal transients, ob-
the semifluorescein (S• ) radical anion in tained by open-circuiting the electrode,
aqueous solution at a silver channel elec- using simple first-order kinetics, was
trode was studied [74]. Cyclic voltammetric found to give an apparent rate constant of
experiments gave rise to linear Levich 1.05 s−1 , ruling out a DISP2 mechanism.
plots, but the size of the reduction wave Using experimentally determined values
increased with decreasing pH, suggesting for the effective number of electrons trans-
that S• becomes unstable with respect to ferred together with the working curves,
disproportionation to yield F and leuco- values of K, were obtained for different
fluorescein (L), flow rates. This is plotted in Fig. 39, where
1.8
1.4
Neff
1.4
1.2
0.0
−3 −2 −1 0 1 2
log10 KDISP1
Fig. 38 Working curve showing the relationship of Neff and K for a
DISP1 process.
Fig. 39 Analysis of the limiting

current|flow rate data at
pH 9.68 assuming an ECE 6
mechanism (O) or a DISP1
mechanism (×).
K
0 100 200
(Volume flow rate)−2/3
[cm−2 s−2/3]
it is apparent that a rather better straight which would be that anticipated for the
line is obtained assuming a DISP1 rather pseudo-first-order rate constant for the
than ECE mechanism. Furthermore, the protonation of S• . Clearly, the analy-
variation in the decadic rate constant ob- sis in terms of a DISP1 mechanism is
served from the extrapolated EPR transient in better agreement with experimental
data (kapp ) with pH using the relationship observations.
1
k= 2 Vlim kapp (52) 3.2.4.2.3 EC Waller and coworkers em-
f →0
ployed the channel flow cell for the study
which arises since two S• radicals are of a catalytic mechanism [68]. The (pre-
lost for each time that S• is proto- equilibrium) EC mechanism can be de-
nated, is linear with a slope of −1, fined by the following kinetic scheme.
A ± e− D
−−
−−−
C−B (i) by the integral concentration of B (an
K anion radical) over the whole volume of the
B+P−
←−−
−−
→−A+Q (ii) EPR cavity weighted by a sin2 sensitivity
profile,
k
Q −−−→ Products (iii) 2h xoffset +xe /2
πx
S∝ b sin2
0 xoffset −xe /2 2xL
Solution of the appropriate convective- xL
diffusion equations at steady state per- πx
+ b sin2 dxdz
mits the behavior of this mechanism to xoffset +xe /2 2xL
be described in terms of a normalized (54)
rate constant: where b is the concentration of B, found
1/3 by numerical analysis of the pertinent
∗ [P ]bulk h2 xe2 convective-diffusion equation, 2xL is the
K = kK (53) length of the EPR cavity, and xoffset is the
[A]bulk 4ν02
distance (offset) from the center of the
where [P ]bulk and [A]bulk are the bulk con- electrode to the center of the cavity.
centrations of P and A. Under conditions The behavior of the steady state EPR
in which the Lévêque approximation is signal (S1 ) of B at a potential corresponding
valid [63, 71], the current is a unique func- to steady state limiting current for the
tion of K ∗ . The observed behavior is best reduction of A to B is related to the
expressed in terms of the effective num- generating current (provided B is stable),
ber of electrons transferred, Neff . The EPR as indicated in Sect. 3.2.3,
signal behavior may be deduced from nu-
i
merical simulation of the concentration of S1 ∝ 2/3
(55)
the ion-radical species, B, as indicated in Vf
Sect. 3.2.3.
Stepping the potential so that the steady
state limiting current for the reduction
3.2.4.2.4 Comproportionation Reactions of B to C is established, gives rise to a
Comproportionation reactions are of new EPR signal (S2 ). Figures 40 and 41
the form Y + Y2− → 2Y− . Spackman show the dependence of the dimensionless
and coworkers [75] studied the effect parameter S2 /S1 upon the rate constant
of a homogeneous comproportionation for step (iii) for the respective cases of
reaction on an electrode reaction that equal and mixed diffusion coefficients. As
proceeds stepwise via two single- kIII is initially increased from zero, the
electron transfers,
EPR signal ratio rises rapidly, but this
Electrode A + e− D
−−
−−−
C−B (i) tails off when the kinetics become mass
transport-controlled. S2 /S1 never reaches
Solution B + e− D
−−
−−−
C−C (ii) the limiting value of two, as this is what
Electrode A+C −−−→ 2B (iii) would be expected if all C produced by
the electroreduction of A is converted into
0 is more positive than E 0 . For
where EA|B B|C B by comproportionation. Rather, for in-
the Compton–Coles channel cell, the EPR finitely fast comproportionation kinetics,
signal from the regions adjacent to and two separate spatial zones are established,
downstream from the electrode is given one close to the electrode containing C (but
Fig. 40 Working curve showing the 1.4

relationship between S2 /S1 and log(kIII )
for the channel electrode flow cell 1.2
geometry of d = 0.591 cm,
2h = 0.036 cm, w = 0.402 cm and 1.0
xe = 0.425 cm, and
DA = DB = DC = 1.75 × 10−5 cm2 s−1 0.8
S2/S1
and a volume flow rate of 10−2 cm3 s−1 .
0.6
0.4
0.2
50 60 70 80 90
log10(kIII/ mol−1 cm3 s−1)
1.4 DB = 1.75 × 10−5 cm2 s−1

DB = 1.20 × 10−5 cm2 s−1
1.2
DB = 0.875 × 10−5 cm2 s−1
1.0
S2/S1
0.8
0.6
0.4
0.2
50 60 70 80 90
log10(kIII/ mol−1 cm3 s−1)
Fig. 41 Working curve showing the relationship between S2 /S1
and log(kIII ) for the channel electrode flow cell geometry of
d = 0.591 cm, 2h = 0.036 cm, w = 0.402 cm and xe = 0.425 cm,
and DA = 1.75 × 10−5 cm2 s−1 , DC = 0.875 × 10−5 cm2 s−1 , flow
rate = 10−2 cm3 s−1 with DB = 0.875 × 10−5 , DB = 1.20 × 10−5 ,
and DB = 1.75 × 10−5 cm2 s−1 .
not A), and the other extending into the increased values S1 and S2 can be explained
center of the channel containing A (but by noting that decreasing DB narrows the
not C), with a sharp interface between the diffusion layer of B, consequently reducing
two zones. the flow rate of B out of the channel,
The ratio of the diffusion coefficients of leading to an accumulation of radicals in
A and B has an effect upon the EPR signal the channel flow cell.
ratio S2 /S1 . If DB is decreased relative Spackman and coworkers [75] looked
to DA , assuming an initial condition at the case in which A is p-chloranil.
DA = DB , both S1 and S2 increase but at First, with the electrode positioned at the
different rates, so that S2 /S1 decreases. The center of the cavity, it was found that
at a potential (E1 ) corresponding to the phenomena as it is constructed using sil-

formation of the anion radical (B), a plot ica – a transparent material. Further, the
of log10 S1 /ilim versus log10 Vf gave a electrode material can be made to be
straight line with slope −2/3, as would a semiconducting material. Irradiation is
be expected for a stable radical. EPR signal simply achieved by illuminating the work-
(S2 )|flow rate data were then obtained by ing electrode through the irradiation port
potentiostatting the platinum foil working machined on one face of the cavity, allow-
electrode at a potential (E2 ) corresponding ing the transmission of a high proportion
to the reduction of the anion radical. of the interrogating light beam to the
Repositioning the electrode so that it was a working electrode and uniform irradia-
few millimeters downstream of the cavity tion of the latter so that photochemical
center, allowed a second set of S|Vf channel flow experiments can be rou-
data at both potentials. In both cases, tinely undertaken in an in situ mode [55,
good agreement was obtained between 65–67, 79–81]. In this manner, a vari-
experiment and theory. ety of mechanisms have been studied,
The dependence of S2 /S1 on the location notably photo-ECE [67, 79], photo-CE [65],
of the electrode within the cavity may be and photo-DISP2 [66].
rationalized as follows. When the electrode
is at E1 , B is generated by the electroreduc-
tion of A and then flows unperturbed out of 3.2.4.3 Modified Electrodes
the channel, so that a concentration profile In situ electrochemical EPR has been
of B within the channel is established, and widely applied in the study of radi-
so that the concentration of B decreases in cal intermediates within polymer films
both x and z directions. However, when on electrode surfaces. Examples include
the electrode potential is stepped to E2 , the the electrochemical doping in polypyr-
concentration of B no longer varies mono- role [129, 130] and the anodic oxidation
tonically with the x and z coordinates, but of poly(N -vinylcarbazole) films [131].
the precise distribution depends is deter-
mined by the electroreduction of B to C 3.2.4.4 Electrochemical EPR in Ionic
and the comproportionation reaction of A Liquids
with C. In particular, at the downstream Ionic liquids are ideal solvents for conduct-
edge of the electrode, [B] rises rapidly with ing electrochemical experiments, notably
x and z, since B is no longer consumed due to their (1) high thermal stability
electrochemically, but then passes through and wide liquid range, (2) ability to act
a maximum when C fully reacts. It is the as solvents for many organic and inor-
position of this maximum within the EPR ganic materials, whilst still being highly
cavity that determines the ratio S2 /S1 . polar, noncoordinating media, (3) high
In situ electrochemical EPR has been conductivity, and (4) negligible vapor pres-
employed in the study of the compropor- sure in contrast with conventional or-
tionation of methyl viologen [76, 77] and ganic solvents, permitting their study
cyanophenyl paraquat [128]. under conditions of high vacuum. Sur-
prisingly, little work on electrochemical
3.2.4.2.5 Photoelectrochemical EPR The EPR has been undertaken in these media.
channel flow cell is ideally suited for Allendoerfer and Osteryoung have pio-
the interrogation of photoelectrochemical neered this area [132], primarily focusing
upon measurements on polymers (such as Andy Waller, Rob Dryfe, Richard Webster,
polypyrrole [133] and polyaniline [134]) in Nathan Lawrence, Frank Marken, and An-
1-methyl-3-ethylimidazolium chloride and dreas Neudeck. We thank John Freeman
aluminum chloride. for the preparation of all of the figures in
Recently, Marken and coworkers [135] this manuscript.
reported a novel method for the electro-
generation of ionic liquids in which mi- References
crodroplets of N ,N ,N ,N -tetrahexyl-para-
phenylene diamine (THPD) are deposited 1. A. Carrington, A. D. McLachlan, Introduc-
upon the surface of a basal plane py- tion to Magnetic Resonance, Chapman &
rolytic graphite (or gold) electrode and are Hall, London, 1979.
2. G. K. Frenkel, P. H. Rieger, J. Chem. Phys.
immersed into an aqueous electrolyte so-
1963, 39, 309–318.
lution. The ionic liquid is formed as a con- 3. D. E. G. Austin, M. F. Peover, M. H. Ingram
sequence of the requirement to maintain et al., Nature 1958, 181, 1784–1785.
electroneutrality upon oxidizing THPD, 4. A. H. Maki, D. H. Geske, J. Chem. Phys.
1959, 30, 1356–1357.
THPD(oil) + X− (aq) −−−→ 5. A. H. Maki, D. H. Geske, J. Am. Chem. Soc.
1960, 82, 267–269.
[THPD+• X− ](oil) + e− (56) 6. J. G. Lawless, M. D. Hawley, J. Electroanal.
Chem. 1969, 23, App. 1–5.
It has been shown that this pro- 7. A. H. Maki, D. H. Geske, J. Chem. Phys.
cess occurs at the three-phase boundary 1960, 33, 825–830.
of THPD|aqueous electrolyte|solid elec- 8. A. H. Maki, D. H. Geske, J. Am. Chem. Soc.
trode [135, 136]. In situ EPR transients for 1961, 83, 1852–1856.
9. W. J. Albery, B. A. Coles, A. M. Couper
the oxidation and rereduction of THPD et al., J. Chem. Soc., Chem. Commun. 1974,
were seen to be consistent with the anion 198–199.
insertion process occurring at the triple 10. W. J. Albery, B. A. Coles, A. M. Couper, J.
phase junction [135]. Electroanal. Chem. 1975, 65, 901–910.
11. W. J. Albery, A. T. Chadwick, B. A. Coles
et al., J. Electroanal. Chem. 1977, 75,
3.2.5 229–235.
Conclusion 12. W. J. Albery, R. G. Compton, A. T. Chad-
wick et al., J. Chem. Soc., Faraday Trans. 1
In situ electrochemical EPR has been 1980, 76, 1391–1398.
shown to be a useful technique for probing 13. W. J. Albery, R. G. Compton, I. S. Kerr, J.
Chem. Soc., Perkin Trans. 2 1981, 823–826.
both the kinetics and mechanisms of elec- 14. W. J. Albery, R. G. Compton, J. Chem. Soc.,
trode processes that produce or consume Faraday Trans. 1 1982, 78, 1561–1568.
paramagnetic species. A variety of cell 15. R. G. Compton, A. M. Waller in Spectroelec-
designs have been presented, illustrating trochemistry (Ed.: R. J. Gale), Plenum Press,
the scope of this spectroelectrochemical New York, 1988, pp. 319–366.
16. A. M. Waller, R. G. Compton, Comprehen-
technique.
sive Chemical Kinetics, Elsevier, Amsterdam,
1989, pp. 297–348, Vol. 29.
Acknowledgments 17. L. H. Piette, R. Ludwig, R. N. Adams, Anal.
Chem. 1962, 34, 916–919, 1587–1592.
18. G. Cauquis, J. P. Billon, J. Cambrisson, Bull.
We are gratefully indebted to all those who Soc. Chim. Fr. 1960, 2062–2067.
have worked with us in the pursuit of 19. G. Cauquis, J. P. Billon, J. Raisson, Bull.
electrochemical EPR, notably Barry Coles, Soc. Chim. Fr. 1967, 199–205.
20. G. Cauquis, M. Genies, Bull. Soc. Chim. Fr. 42. I. B. Goldberg, Electron Spin Resonance:
1967, 3220–3227. Royal Society of Chemistry Specialist Periodi-
21. G. Cauquis, M. Genies, H. Lemaire et al., J. cal Reports, The Royal Society of Chemistry,
Chem. Phys. 1967, 42, 4642–4646. London, 1981, pp. 1–20, Vol. 6.
22. G. Cauquis, C. Berry, M. Maivey, Bull. Soc. 43. I. B. Goldberg, A. J. Bard, J. Phys. Chem.
Chim. Fr. 1968, 2510–2515. 1971, 75, 3281–3285.
23. B. Kastening, J. Divisek, B. Costissa- 44. I. B. Goldberg, A. J. Bard, J. Phys. Chem.
Mihelcic, Faraday Discuss. Chem. Soc. 1973, 1974, 78, 290–294.
56, 341–348. 45. I. B. Goldberg, D. Boyd, R. Hirasawa et al.,
24. B. Kastening, J. Divisek, B. Costissa- J. Phys. Chem. 1974, 78, 295–301.
Mihelcic et al., Z. Phys. Chem. 1973, 87, 46. J. B. Carroll, Ph. D. thesis, State University
125–132. of New York at Buffalo, 1983.
25. J. K. Dohrmann, F. Galluser, H. Wittchen, 47. R. D. Allendoerfer, G. A. Martinchek, S.
Faraday Discuss. Chem. Soc. 1973, 56, Bruckenstein, Anal. Chem. 1975, 47,
350–358. 890–894.
26. J. K. Dohrmann, K. J. Vetter, J. Electroanal. 48. R. D. Allendoerfer, J. B. Carroll, J. Magn.
Chem. 1969, 20, 23–29. Reson. 1980, 37, 497–501.
27. J. K. Dohrmann, F. Galluser, Ber. Bunsen- 49. R. D. Allendoerfer, W. Froncisz, C. C. Felix
Ges. Phys. Chem. 1971, 75, 432–437. et al., J. Magn. Reson. 1988, 76, 100–106.
28. J. K. Dohrmann, Ber. Bunsen-Ges. Phys. 50. H. Ohya-Nishiguchi, Bull. Chem. Soc. Jpn.
Chem. 1970, 74, 575–580. 1979, 52, 2064–2068.
29. O. Stern, Z. Phys. 1921, 7, 249–252. 51. F. Gerson, H. Ohya-Nishiguchi, G. Wydler,
30. W. Gerlach, O. Stern, Ann. Phys. (Leipzig) Angew. Chem., Int. Ed. Engl. 1976, 15,
552–556.
1924, 74, 673–675.
52. J. Bruken, F. Gerson, H. Ohya-Nishiguchi,
31. A. I. Akhiezer, V. B. Berestetski, Quantum
Helv. Chim. Acta 1977, 60, 1220–1227.
Electrodynamics, Monographs in Physics
53. W. Froncisz, J. S. Hyde, J. Magn. Reson.
and Astronomy, Wiley Interscience, New
1982, 47, 515–519.
York, 1965, Vol. XI, p. 1–50.
54. J. S. Hyde, W. Froncisz, Electron Spin Res-
32. N. M. Atherton, Electron Spin Resonance,
onance: Specialist Periodical Reports, The
Ellis Horwood, Chichester, UK, 1973.
Royal Society of Chemistry, London, 1986,
33. P. Ludwig, R. N. Adams, J. Chem. Phys. pp. 175–201, Vol. 10.
1962, 37, 828–831. 55. B. A. Coles, R. G. Compton, J. Electroanal.
34. H. M. McConnell, D. B. Chestnut, J. Chem. Chem. 1983, 144, 87–94.
Phys. 1958, 28, 778–781. 56. C. M. A. Brett, A. M. O. Brett, Comprehen-
35. C. P. Poole, Electron Spin Resonance: A Com- sive Chemical Kinetics, Elsevier, Amsterdam,
prehensive Treatise on Experimental Tech- 1988, pp. 355–376, Vol. 26.
niques, 2nd ed., Wiley, New York, 1983. 57. V. Levich, Physicochemical Hydrodynamics,
36. P. Boyer, J. Dericbourg, C. R. Acad. Sci. Prentice Hall, Englewood Cliffs, N.J., 1962.
Paris 1967, 429–437. 58. R. G. Compton, D. J. Page, G. R. Sealy, J.
37. M. C. R. Symons, Electron Spin Resonance Electroanal. Chem. 1984, 161, 129–138.
Spectroscopy, Van Nostrand Reinhold, Lon- 59. I. B. Goldberg, A. J. Bard, S. W. Feldberg, J.
don, 1978. Phys. Chem. 1972, 76, 2550–2559.
38. F. Gerson, High Resolution E.S.R. Spec- 60. P. R. Unwin, R. G. Compton, Comprehen-
troscopy, John Wiley & Sons, New York, sive Chemical Kinetics, Elsevier, Amsterdam,
1970. 1989, pp. 190–296, Vol. 29.
39. K. A. McLauchlan, Magnetic Resonance, Ox- 61. J. A. Cooper, R. G. Compton, Electroanalysis
ford University Press, Oxford, 1972. 1998, 10, 141–152.
40. P. W. Atkins, Molecular Quantum Mechan- 62. J. B. Flanagan, L. Marcoux, J. Phys. Chem.
ics, 3rd ed., Oxford University Press, Oxford, 1974, 78, 718–723.
1997. 63. J. A. Alden, R. G. Compton, J. Electroanal.
41. B. J. Tabner in Spectroscopy (Eds.: B. P. Chem. 1996, 404, 27–34.
Straughan, S. Walker), Chapman & Hall, 64. M. A. Lévêque, Ann. Mines Mem. Ser. 1928,
London, 1976, pp. 209–260, Vol. 1. 12, 201–208.
65. R. G. Compton, B. A. Coles, G. M. Stearn 86. H. Matsuda, J. Electroanal. Chem. 1967, 15,
et al., J. Chem. Soc., Faraday Trans. I 1988, 325–328.
84, 2357–2363. 87. P. J. Daly, D. J. Page, R. G. Compton, Anal.
66. R. G. Compton, B. A. Coles, M. B. G. Chem. 1983, 55, 1191–1197.
Pilkington, J. Chem. Soc., Faraday Trans. 88. B. H. Vassos, G. W. Ewing, Electroanalytical
I 1988, 84, 4347–4355. Chemistry, Wiley, New York, 1983, p. 130.
67. R. G. Compton, B. A. Coles, M. B. G. 89. S. G. Weber, W. C. Purdy, Anal. Chim. Acta
Pilkington et al., J. Chem. Soc., Faraday 1976, 100, 531–535.
Trans. 1990, 86, 663–671. 90. R. N. Bagchi, A. M. Bond, R. Colton, J.
68. A. M. Waller, R. J. Northing, R. G. Comp- Electroanal. Chem. 1986, 199, 297–303.
ton, J. Chem. Soc., Faraday Trans. 1990, 86, 91. R. N. Bagchi, A. M. Bond, C. L. Heggie
335–340. et al., Inorg. Chem. 1983, 22, 3007–3013.
69. S. Moldoveanu, J. L. Anderson, J. Elec- 92. R. N. Bagchi, A. M. Bond, F. Scholz et al., J.
troanal. Chem. 1984, 178, 45–49. Electroanal. Chem. 1988, 245, 105–111.
70. J. L. Anderson, S. Moldoveanu, J. Electro- 93. R. N. Bagchi, A. M. Bond, F. Scholz, J. Elec-
anal. Chem. 1984, 179, 107–118, 119–128. troanal. Chem. 1988, 252, 259–264.
71. R. G. Compton, M. B. G. Pilkington, G. M. 94. D. A. Fieldler, M. Koppenol, A. M. Bond, J.
Stearn, J. Chem. Soc., Faraday Trans. I 1988, Electrochem. Soc. 1995, 142, 862–868.
84, 2155–2164. 95. R. D. Webster, A. M. Bond, R. G. Compton,
72. R. G. Compton, D. J. Page, G. R. Sealy, J. J. Phys. Chem. 1996, 100, 10 288–10 296.
Electroanal. Chem. 1984, 163, 65–69. 96. R. D. Webster, A. M. Bond, B. A. Coles
73. M. Abramowitz, I. A. Stegun, Handbook of et al., J. Electroanal. Chem. 1996, 404,
Mathematical Functions, Dover Publica- 303–311.
tions, New York, 1970, p. 283–285. 97. M. B. Glauert, J. Fluid Mech. 1956, 1,
74. R. G. Compton, P. J. Daly, P. R. Unwin 625–627.
et al., J. Electroanal. Chem. 1985, 191, 15–21. 98. R. G. Compton, C. R. Greaves, A. M.
75. R. G. Compton, B. A. Coles, R. A. Spack- Waller, J. Electroanal. Chem. 1990, 277,
man, J. Phys. Chem. 1991, 95, 4741–4748. 83–88.
76. R. D. Webster, R. A. W. Dryfe, J. C. Eklund 99. W. J. Albery, C. M. A. Brett, J. Electroanal.
et al., J. Electroanal. Chem. 1996, 402, Chem. 1983, 148, 201–207.
167–172. 100. J. Yamada, H. Matsuda, J. Electroanal.
77. C.-W. Lee, J. C. Eklund, R. A. W. Dryfe Chem. 1973, 44, 189–194.
et al., Bull. Korean Chem. Soc. 1996, 17, 101. R. G. Compton, C. R. Greaves, A. M.
162–167. Waller, J. Appl. Electrochem. 1990, 20,
78. F. Prieto, R. D. Webster, J. A. Alden et al., J. 575–585.
Electroanal. Chem. 1997, 437, 183–188. 102. A. Petr, L. Dunsch, A. Neudeck, J. Elec-
79. R. G. Compton, R. G. Harland, M. B. G. troanal. Chem. 1996, 412, 153–159.
Pilkington et al., Port. Electrochim. Acta 103. A. Neudeck, L. Kress, J. Electroanal. Chem.
1987, 5, 271–275. 1997, 437, 141–148.
80. R. G. Compton, R. A. W. Dryfe, J. C. Ek- 104. A. Neudeck, A. Petr, L. Dunsch, J. Phys.
lund, Research in Chemical Kinetics (Eds.: Chem. B 1999, 103, 912–922.
R. G. Compton, G. Hancock), Elsevier, Am- 105. P.-H. Aubert, A. Neudeck, L. Dunsch et al.,
sterdam, 1993, pp. 239–315, Vol. 1. J. Electroanal. Chem. 1999, 470, 77–88.
81. R. G. Compton, R. A. W. Dryfe, Prog. React. 106. A. Neudeck, A. Petr, L. Dunsch, Synth. Met.
Kinet. 1995, 20, 245–297. 1999, 107, 143–158.
82. R. G. Compton, P. R. Unwin, J. Electroanal. 107. P. Rapta, R. Faber, L. Dunsch et al., Spec-
Chem. 1986, 205, 1–20. trochim. Acta A 2000, 56, 357–362.
83. W. J. Albery, R. G. Compton, C. C. Jones, J. 108. P. Rapta, A. Neudeck, A. Petr et al., J. Chem.
Am. Chem. Soc. 1984, 106, 469–475. Soc., Faraday Trans. 1998, 94, 3625–3630.
84. W. J. Albery, C. C. Jones, Faraday Discuss. 109. R. A. W. Dryfe, R. D. Webster, B. A. Coles
Chem. Soc. 1984, 78, 193–199. et al., Chem. Commun. 1997, 779–780.
85. R. G. Compton, A. M. Waller, J. Electroanal. 110. R. D. Webster, R. A. W. Dryfe, B. A. Coles
Chem. 1985, 195, 289–296. et al., Anal. Chem. 1998, 70, 792–800.
111. Y. Cheng, D. J. Schiffrin, J. Chem. Soc., 125. T. J. Kemp, Prog. React. Kinet. 1999, 24,
Faraday Trans. 1994, 90, 2517–2521. 287–345.
112. T. Solomon, A. J. Bard, J. Phys. Chem. 1995, 126. A. C. Testa, W. H. Reinmuth, Anal. Chem.
99, 17 487–17 492. 1961, 19, 1320–1322.
113. N. Haren, Z. Luz, M. Shporer, J. Am. Chem. 127. P. D. Moorland, R. G. Compton, J. Phys.
Soc. 1974, 96, 4788–4793. Chem. B 1999, 103, 8951–8959.
114. R. Male, R. D. Allendoerfer, J. Phys. Chem. 128. R. G. Compton, A. M. Waller, P. M. S.
1988, 92, 6237–6243. Monk et al., J. Chem. Soc., Faraday Trans.
115. A. Gennaro, F. Marron, A. Maye et al., J. 1990, 86, 2583–2587.
Electroanal. Chem. 1985, 185, 353–359. 129. A. M. Waller, A. N. S. Hampton, R. G.
116. R. G. Compton, R. A. Spackman, R. G. Compton, J. Chem. Soc., Faraday Trans. I
Wellington et al., J. Electroanal. Chem. 1992, 1989, 85, 773–778.
327, 337–341. 130. A. M. Waller, R. G. Compton, J. Chem. Soc.,
117. R. G. Compton, R. A. Spackman, R. G. Faraday Trans. I 1989, 85, 977–982.
Wellington et al., J. Electroanal. Chem. 1993, 131. R. G. Compton, F. J. Davis, S. C. Grant, J.
344, 235–247. Appl. Electrochem. 1986, 16, 239–243.
118. S. A. Olsen, A. M. Bond, R. G. Compton 132. M. A. M. Noel, R. D. Allendoerfer, R. A.
et al., J. Phys. Chem. A 1998, 102, Osteryoung, J. Phys. Chem. 1992, 96,
2641–2652. 239–245.
119. C. A. Reed, R. D. Bolskar, Chem. Rev. 2000, 133. J. F. Oudard, R. D. Allendoerfer, R. A.
1000, 1075–1189. Osteryoung, J. Electroanal. Chem. 1988, 241,
120. I. Noviandri, R. D. Bolskar, P. A. Lay et al., 231–238.
J. Phys. Chem. B 1997, 101, 6350–6359. 134. J. Tang, R. D. Allendoerfer, R. A. Oster-
121. E. G. Janzen, B. J. Blackburn, J. Am. Chem. young, J. Phys. Chem. 1992, 96, 3531–3537.
Soc. 1968, 90, 5909, 5910. 135. F. Marken, R. D. Webster, S. D. Bull
122. E. G. Janzen, B. J. Blackburn, J. Am. Chem. et al., J. Electroanal. Chem. 1997, 437,
Soc. 1969, 91, 4481–4484. 209–218.
123. J. D. Wadhawan, F. J. Del Campo, F. Mar- 136. J. C. Ball, F. Marken, Q. Fulian et al., Elec-
ken et al., J. Electroanal. Chem. 2001, 507, troanalysis 2000, 12, 1017–1025.
135–143. 137. B. A. Coles, R. G. Compton, J. Electroanal.
124. A. J. Bard, J. C. Gilbert, R. D. Godwin, J. Chem. 1981, 127, 37–43.
Am. Chem. Soc. 1974, 96, 620–621.
3.3 Spectroscopic Applications of STM and AFM in Electrochemistry 221
3.3 to quantum mechanics, the electrons can

Spectroscopic Applications of STM and tunnel across the gap and give rise to
AFM in Electrochemistry a measurable tunneling current when a
bias voltage is applied between the tip and
Nongjian Tao
the sample. The tunneling probability is
Arizona State University, Tempe, AZ
proportional to the overlap of the electron
3.3.1 clouds (wave functions, ψsub and ψtip ) that
Introduction spill out of the tip and the sample surfaces.
The overlap decreases exponentially as
Scanning Tunneling Microscopy (STM) [1] does the gap width, which leads to a simple
and the related Atomic Force Microscopy relation between the tunneling current (I )
(AFM) [2] are dream techniques that allow and the sample-tip distance (s) by
us to visualize atomic-scale phenomena
on surfaces. Shortly after the inventions, I ∝ exp(−1.025 φs) (1)
the capability of the techniques for imag-
ing electrode surfaces in solutions was where φ is the work function or the
recognized and applied to electrochem- tunneling barrier height in eV and s
istry [3–6]. Since the pioneering works, is in Å. The work function is typically
various electrochemical phenomena have ∼4 eV, so the tunneling current decays
been imaged and studied with atomic or about e2 ∼ 7.4 times per Å, according to
molecular resolution. Examples of such Eq. (1). The rapid decay means that only
studies can be found in Volume 3 of this the atom in the closest proximity to the
series and in many reviews [7–12]. In ad- sample surface dominates the measured
dition to high-resolution images, STM and tunneling current, which is the basis of the
AFM can provide local spectroscopy in- atomic resolution imaging of STM. When
formation of the surfaces and adsorbates. scanning the STM tip across the sample
This chapter provides an introduction to surface with a piezoelectric transducer, the
the spectroscopy applications in electro- local tunneling current can be mapped.
chemistry. We start with a brief outline A more common approach is to keep a
of STM and AFM fundamentals, and then constant current by adjusting the vertical
illustrate the spectroscopy applications us- position of the tip with a feedback loop.
ing selective examples. The vertical position of the tip recorded
as a function of the lateral position
3.3.2 provides a three-dimensional image of the
Electrochemical STM – Elementary sample surface.
Concepts The simple exponential decay given by
Eq. (1) was verified experimentally in vac-
STM consists of a sharp metal tip that is uum over a wide range of tunneling
placed over a conductive sample surface current despite that the simple model
(Fig. 1). The conduction electrons are ignores factors, such as imaging force ex-
free to propagate within the tip and the perienced by the tunneling electrons [13].
sample, but classically forbidden in the gap Replacing the vacuum gap with an elec-
between the tip and the sample in which trolyte, Eq. (1) holds approximately for a
the electron total energy is smaller than limited range of the gap with an effective
the potential energy. However, according barrier height (φ) somewhat smaller than
Fig. 1 Electron tunneling

between an STM tip and a
conductive substrate.
Substrate Tip
electrode electrode
f ysub
ytip
Ef + eVbias
Ef
the vacuum value [14–20]. However, when electronic states of the tip and the sam-
measuring the tunneling current over a ple surface. The tunneling current across
wide range of the gap, Lindsay and cowork- a vacuum or insulation gap between the
ers [18, 19] observed a significant deviation tip and the sample surface is given by [26]
from the simple exponential dependence.
4πe +∞
Hong and coworkers [20] reported that the I = [f (EF − eVbias + ε)
tunneling barrier for the electron tunnel- h̄ −∞
ing from tip to substrate was remarkably − f (EF − ε)]ρs (EF − eVbias + ε)
different from the reverse direction. Both
× ρ T (EF + ε)|M|2 dε (2)
the nonexponential decay and asymmet-
ric tunneling demand a better under- where f (E) is the Fermi–Dirac distribu-
standing of electron tunneling across an tion function, EF is the Fermi energy of
electrolyte gap. Schmickler and Hender- the tip, Vbias is the bias voltage, ρT (ε) and
son [21] treated the water molecules in ρS (ε) are the local density of states (DOS)
the tunneling gap as a dielectric medium of the tip and sample, respectively, and
and predicted a smaller tunneling barrier |M|2 is related to the overlap between the
than in vacuum. Kuznetsov and Ulstrups tip and the sample wave functions that
considered the effects of solvent fluctu- decays according to Eq. (1). When the tem-
ations on the tunneling barrier height perature is low, (kB T EF ) and |M|2 is
and tunneling rate [22]. They found that independent of energy, the tunneling cur-
the solvent fluctuations could enhance rent is a convolution of the tip and the
the tunneling rate, thus leading also to sample DOS over an energy range eVbias ,
a lower effective barrier height. More elab- eVbias
orate models involving resonant tunneling
I ≈ ρS (EF − eVbias + ε)
through solvent states [18, 23] and ab initio 0
simulations [24, 25] were also developed. × ρT (EF + ε) dε (3)
The STM image cannot always be lit-
erally interpreted as a geometrical to- In the presence of a molecular adsorbate,
pography of the sample surface because one may treat the combined adsorbate and
the tunneling current depends on the the substrate as the sample. However,
many models have been proposed to poorly conductive materials on electrodes.

treat the adsorbate more explicitly both The tip-sample force in AFM is typically
in vacuum [27–35] and in aqueous solu- detected by measuring the bending of a
tions [21, 36–40]. For example, Kuznetsov flexible cantilever on which the tip is at-
and Ulstrup [36, 41] pointed out that a tached (Fig. 2) [43]. As the tip approaches
strong coupling between the electronic the sample surface, it first experiences
states of a molecule and the environmen- long-range forces, such as the van der
tal fluctuations could affect the tunneling Waals and electrostatic interactions, then
current by lowering the effective tunnel- specific chemical bonds between the tip
ing barrier and by momentarily bringing and the sample surface, followed by a
a molecular state to align with the sub- rapid increase in the hard-core repulsive
strate/tip Fermi levels. In addition to force [44]. In the electrochemical environ-
the adsorbate-solvent coupling, Schmick- ment, additional forces such as ionic and
ler and coworkers [38, 42] took into ac- hydration forces arise [45]. These forces
count the adsorbate-substrate coupling. will be discussed in greater details later.
They considered explicitly a resonant tun- AFM can be operated in either the
neling process when the molecular state attractive or repulsive-force modes. In the
was brought to align with the substrate/tip repulsive mode, or the so-called contact-
Fermi levels. We will return to these theo- mode AFM, the force is highly stable
ries later. and images that show the periodic lattice
structures of the samples are routinely
3.3.3 obtainable. However, most of the reported
Electrochemical AFM – Elementary images show either perfectly ordered
Concepts atomic structures or defects that are
much greater than an atom. Ohnesorge
AFM is similar to STM in many aspects, and Binnig [46] investigated this issue by
but a fundamental difference between imaging calcite in water. They observed
them is that STM measures the tunnel- atomic-scale defects when the net repulsive
ing current, while AFM probes the force. force applied on the sample surface by
So unlike STM, which requires a con- the AFM tip was 0.1 nN or smaller.
ductive surface or small adsorbates on Increasing the repulsive force, they found
a conductive surface, AFM can provide that sharp monoatomic steps were slowly
high-resolution images of both conductive wiped away because the large contact force
and insulating surfaces. This is important resulted in a contact area between the
for imaging large biological molecules or tip and the sample much greater than
F = k∆z
∆z
D
(a) (b) (c)
Fig. 2 Tip–sample interactions in AFM.

an atom. The subsequent AFM image, much greater damping in solutions, which
nevertheless, still showed well-ordered broadens the resonance. Faster response
two-dimensional lattice structures owing time can be achieved by directly following
to the periodic modulation of the force by the resonance frequency [48]. This later
the periodic lattice of the sample surface. technique has been used to obtain atomic
The experiment shows that most AFM resolution images of Si(111) −7 × 7 struc-
experiments operated with a contact force ture [49, 50]. We note that even if tip-
much greater than 0.1 nN are probably not sample contact is involved, the AC mode
truly atomic resolution. is still helpful for imaging soft materials,
The contact force can be avoided by such as biological molecules adsorbed on
operating in the attractive force mode or electrodes, because the contact is intermit-
the noncontact mode. For high-sensitivity tent. Commercial AFM, such as Tapping
measurement of the attractive force, an AC Mode AFM [51–53] from Digital Instru-
technique was introduced. The technique ments and MAC Mode AFM [54–56] from
measures the change in the resonance Molecular Imaging Inc. are two examples.
frequency of the AFM cantilever. When
the tip is far away from the sample surface, 3.3.4
the resonant frequency f0 is given by, Selective Examples of Applications

k STM and AFM applications in electro-
f0 = 2π (4)
m chemistry as imaging tools are numer-
where k is the spring constant of the ous. Novel nonimaging applications are
cantilever and m is the effective mass. also abundant in the literature [57, 58].
When the tip approaches the sample, Because of space limitations, here we fo-
the tip–sample interaction changes the cus on tunneling spectroscopy and force
spring constant by k = ∂Ftip – sample /∂z. spectroscopy applications using selective
Consequently the resonance frequency examples that are directly related to electro-
changes by an amount of chemistry. Other important nonimaging
applications that are relevant to elec-
f k trochemistry include fabricating nanos-
≈ (5)
f0 2k tructures [59–68] and probing fast kinet-
assuming that both the changes in the ics [69–75].
spring constant and oscillation amplitude
of the cantilever are small [47]. One way to 3.3.4.1 Tunneling Spectroscopy
implement the idea for AFM imaging is The electronic states of a sample can
to drive the AFM cantilever at or near the be determined with STM by performing
resonance frequency and measure the os- Scanning Tunneling Spectroscopy (STS).
cillation amplitude as the feedback signal. The concept of tunneling spectroscopy
When the tip–sample interaction changes, was developed as early as in the sixties
the amplitude changes because of the using metal-oxide-metal junctions [26, 76].
change of the resonance frequency. This The original idea of building the STM by
method has a low response time when the Binnig and Rohrer was actually to perform
resonance is very sharp. The response time tunneling spectroscopy locally on a small
is usually acceptable for imaging in so- area. According to Eq. (3), the tunneling
lutions because the cantilever experiences current is a convolution of the tip and
the sample DOS. In order to determine probing the electronic states of a single
the sample DOS, one has to know the molecule or even a particular group within
tip DOS, which is not always possible. a molecule.
However, for a given voltage range, the tip Schmickler and coworkers [38, 42] sug-
DOS is often weakly dependent on energy, gested that lowest unoccupied molecular
ε, then Eq. (3) becomes, orbital (LUMO) and highest occupied
molecular orbital (HOMO) of molecules
dI
≈ ρS (EF − eVbias + ε) (6) adsorbed on an electrode could be probed
dV
by STM. The LUMO (HOMO) is broad-
Because the tunneling current is very ened and shifted as a result of the
sensitive to the tip-sample separation, coupling of the molecule to the envi-
the measured spectroscopic data are often ronment, which is described as density
normalized by I /V and given by [(V /I ) of oxidized (reduced) states. They have
(dI /dV )] = [(d ln(I ))/(d ln(V ))] [77]. considered a coherent one-step electron
In vacuum, STS is usually obtained by transfer from the tip electrode to the sub-
fixing the tip over the region of interest strate electrode or vice versa via a resonant
and then measuring tunneling current by transition through the oxidized states. An
sweeping the bias voltage. The method is, electron transfers from one electrode to
however, difficult to apply to STM in the the molecule and then to the second elec-
electrochemical environment because the trode before the molecule relaxes to the
sweeping bias voltage induces a large po- reduced state. This implies that the elec-
larization current that can overwhelm the tron tunneling time is much faster than
tunneling current. Furthermore, faradic the vibration time of the molecule, which
current at large bias voltage limits the is on the order of 10−14 s. If assuming the
spectroscopy to a relatively small range tunneling matrices, the tip, and the sub-
of bias voltage. Despite the difficulties, strate DOS are independent of energy, ε,
useful electronic information has been
the model gives,
obtained by performing tunneling spec-
troscopy outside the electrochemical cell eVbias
or by indirect methods. I≈ Dox (ε − eη + eVbias ) dε
0
(7)
3.3.4.1.1 Redox States An early attempt where η is the overpotential and Dox is the
to extract DOS of redox molecules from density of the oxidized states. When the
tunneling current was made by Mori- molecule couples weakly to the substrate
aki and coworkers [78]. They used thin and the tip, Dox takes the Gaussian form
SiO2 tunnel barriers on Pt silicide sub-
strates for tunneling spectroscopy of com- π (ε − λ)2
Dox (ε) = exp −
plex ions, such as Fe(CN)6 3−/4− and kB T λ 4kB T λ
Fe(EDTA)1−/2− . The energy dependent (8)
DOS extracted from the tunneling spec- where λ is the reorganization energy. The
tra was in good agreement with the tunneling spectrum, dI /dVbias versus Vbias
Classical Marcus–Gerischer model [79]. gives approximately the density of states
The experiment was macroscopic because (Dox ). In electrochemical STM, the over-
it measured a large number of molecules. potential (η) can be varied independent of
In contrast, STS has the capability of the tip-substrate bias voltage (Vbias ). So by
measuring the tunneling current as a func- where αeVbias is the energy shift in the
tion of the overpotential with a fixed Vbias , redox levels caused by the tip-substrate bias
one has a new way to perform tunneling voltage, and α is between 0 and 1, reflecting
spectroscopy. It follows from Eqs. (7 and the distribution of the bias voltage across
8) that I versus η gives also the density of the tunneling gap. This two-step electron
states for a small Vbias . transfer can also give rise to a maximum
Kuznetsov and Ulstrup [22, 36, 41] have at a potential depending on the relative
considered another limit, a sequential contributions from the two steps.
two-step process, in which the electron Friis and coworkers [80] considered an
transfers from the tip (substrate) to intermediate mechanism, in which an
the molecule and reduces the molecule. electron first transfers to the molecule and
The reduced states of the molecule are the molecule begins to relax towards the
located below the Fermi level of the reduced state. However, before it is fully
substrate, but a thermal fluctuation can relaxed to the reduced state, an electron
shift the states up and then allow a second transfer from the molecule to the second
electron transfer from the molecule to the electrode occurs when the temporarily oc-
substrate (tip), returning the molecule to cupied level passes the Fermi level of the
the oxidized state. Neglecting backward second electrode. They called this pro-
flowing current, the current caused by this cess coherent two-step electron transfer.
two-step process is More recently, Kuznetsov and Ulstrup [81]
have developed a systematic theory for
Rsm Rmt the sequential two-step process. The elec-
I≈ (9)
Rsm + Rmt tron transfers from one electrode to the
molecule and reduces the molecule, and
where Rsm and Rms are proportional to the
then transfers to the second electrode and
electron transfer rates from the substrate
reoxidizes the molecule, so the process is
to the redox molecule, and from the redox
reviewed as a cycle of consecutive molecu-
molecule to the tip, respectively. Again, if
lar reduction and reoxidation. A particular
we assume energy independent DOS for
interesting feature in the theory is that each
both the tip and the substrate, then Rsm
reduction-reoxidation cycle is composed
and Rms are given by,
of a large number of individual electron
+∞ transfer events between the molecule and
Rsm ≈ f (ε) both the tip and the substrate. This sig-
−∞
nificantly enhances the tunneling current
(ε − E0 − λ + αeVbias )2 compared to a single electron transfer.
× exp − dε
4kB T λ Experimentally extracting redox state
(10) information from STS has been also
pursued. Snyder and White measured
and STS of Fe-protoporphyrin IX multi-
+∞ layer films adsorbed on graphite elec-
Rmt ≈ [1 − f (ε + eVbias )] trode in air [82]. They observed an en-
−∞
hancement that was identical for both

(ε − E0 + λ + αeVbias )2 signs of the tip-substrate bias voltage.
× exp − dε
4kB T λ The data were interpreted in terms of
(11) tunneling via the adsorbate metal centers.
Tao studied electroactive molecule, Fe- were modified with thiol groups at one
protoporphyrins IX, coadsorbed with end so that they were tethered to the
nonelectroactive protoporphyrins IX on surface via the strong S-Au bonds. The
graphite substrate in electrolyte [83]. Be- derivative of these curves (dI /dV ) had a
cause of the difficulty in obtaining reliable Gaussian-shaped peak at a voltage charac-
STS in electrolyte, the apparent height teristic of the compound (Fig. 3). These
of Fe-protoporphyrins IX relative to that bias-dependent features were not ob-
of protoporphyrins IX was measured as served in the less electroactive molecules,
a function of the electrochemical poten- so the STM was shown to be capable
tial. A large increase in the tunneling of distinguishing electroactive molecules
current occurred when the potential was from nonelectroactive molecules. They at-
tuned to the redox potential. This ob- tempted to fit their data with both the
servation was originally interpreted as a coherent one-step and sequential two-
resonant tunneling via the oxidized state step models outlined above. Both mod-
of Fe-protoporphyrins IX in terms of the els yielded Gaussian-like peaks in the
coherent one-step model [83, 84], but the dI /dVbias versus Vbias plot. The Gaussian
coherent one-step model predicts a maxi- predicted by the first model (Eqs. 7 and
mum shifted away from the redox potential 8) was located at a lower potential than
by λ, the reorganization energy. This, that by the second model, but they could
although, could be due to a small reorgani- not distinguish the two models within the
zation energy or experimental uncertainty experimental error.
in the potential, the sequential two-step Hipps and coworkers obtained
electron transfer offers an alternative ex- STS metal-phthalocyanines adsorbed
planation [81]. on Au(111) surfaces with STM
Han and coworkers [85] studied STS of in conjunction with the tunneling
a series of metal-porphyrin compounds spectroscopy obtained with the metal-
adsorbed onto a Au(111) electrode in an or- insulator junctions [86–88]. STM images
ganic solvent (mesitylene). The molecules showed that the molecules lay flat on
0.03
FeBr
0.025
FeCl
0.02
Mn
0.015
dI/dV
2H
0.01
0.005
0
Zn Bare electrode
−0.005
−1000 −800 −600 −400 −200 0
Bias
[mV]
Fig. 3 Derivatives (dI/dVbias ) of the tunneling current versus bias voltage
curves of Fe, Mn, Zn, and H embedded in tetraphenol porphyrin (TPP).
(From Ref. [85] with permission.)
the surface, and the brightness of the identification of other species has also
central part of the molecule depended been demonstrated. Schott and White [98]
on the metal ion located there [86, 89]. studied a mixed monolayer of halogen
The measured resonant tunneling via the atoms (F, Cl, Br, and 1) on Ag(111),
molecular orbital shown in the STS was in addition to the different radii and
in good agreement with the macroscopic they observed differences in the tunneling
tunneling spectroscopy. Although the spectra of the halogen species. Another
experiments were measured in vacuum, interesting molecular identification with
where reliable tunneling spectroscopy STM was carried in monolayer
can be more easily obtained than in films of several primary substitute
air or in water, they demonstrated a hydrocarbons CH3 (CH2 )n CH2 X (X=CH3 ,
correlation between the orbital-mediated OH, NH2 , SH, Cl, Br, and I) on
tunneling spectroscopy data and the redox graphite–solution interfaces by Flynn
potentials of the metal-phthalocyanines in and coworkers [99, 100] The hydrocarbons
solution [87, 90]. formed well-ordered arrays on the surface.
In addition to the relative simple redox The NH2 , SH, Br, and I groups appeared as
molecules, electrochemical STM has been bright spots in the STM images while OH
applied to study redox proteins, such as cy- and Cl groups were not distinguishable
tochrome c [91–93], blue copper protein from the alkyl chain, so it is possible
Pseudomonas aeruginosa azurin [94–96], to identify the molecules with different
and PS I system [97]. An interesting ob- substitutes according to their ‘‘brightness’’
servation by Ulstrup and coworkers is a in the STM images. Claypool and
‘‘spike’’ localized near the center of each coworkers imaged a series of substituted
azurin molecule, which means a much alkanes and alkanols on graphite [101].
higher electron tunneling through the re- They found that all the functional
gion of the ‘‘spike’’ [94]. The ‘‘spike’’ was groups (halides, amines, alcohols, nitriles,
assigned to the electron transfer via the alkenes, alkynes, ethers, thioethers, and
redox center, copper, in azurin, which disulfides) studied in their experiments
enhanced the tunneling current. The tun- could be distinguished from each other
neling current, however, depended only except for Cl and OH groups. Their results
weakly on the bias voltage and the electro- were consistent with quantum chemistry
chemical overpotential, which could be be- calculations [101]. Recently, Claypool and
cause of the relative large bias voltage used coworkers extended the study to MoS2
in the experiment. This study provides surface and the images of tetradecanol
the first evidence of local ‘‘spectroscopic’’ on MoS2 were nearly identical to those
information of redox proteins with STM. obtained on graphite [102].
3.3.4.1.2 Molecular Identifications Be- 3.3.4.1.3 Molecular Electronics As sili-

cause STM probes the electronic con-based microelectronics is heading to-
properties of the molecules, it can identify wards the increasingly difficult road of
structurally similar molecules on the nanometer scale, building electronic de-
basis of the difference in the electronic vices with individual molecules (molecular
states of the molecules. The works on electronics) becomes a promising alter-
redox molecules described above are good native [103, 104]. Indeed, many molecules
examples of such applications. Molecular possess wonderful electronic properties
that could be used as basic elements in The separation between the two surfaces
a functional electronic device. A major is controlled by a piezoelectric transducer
challenge is to find a practical method and measured by an optical interferomet-
to wire individual molecules to the outside ric technique. Because SFA uses surfaces
world. STM can ‘‘wire’’ a single molecule with known geometries, precise measure-
with the sharp tip and probe electron ments of surface forces and energies are
transfer through individual molecules. A possible. However, SFA measures the
large number of works have already been force of a relative large surface area and
performed [86, 88, 105–112]. Gimzewski is limited to atomically flat surfaces. That
and coworkers have published an excellent is why most of the SFA studies have been
review recently [113], we describe here an carried out on mica surfaces. AFM mea-
interesting example that is based on the sure the force between a tip and a sample
redox properties of molecules. surface. A drawback is that the AFM tip
Gittins and coworkers [114] explored geometry is not always precisely known,
molecular electronics applications using although attempts to control the tip geom-
STS and electrochemistry (Fig. 4). Their etry have been made by attaching an object
system consisted of gold nanoparticles with known geometry onto the regular
attached to a gold electrode surface via AFM tip. Examples include colloidal par-
the redox-active bipyridinum moiety. They ticle [118, 119] and carbon nanotube [120].
controlled the electron transport between However, being a local probe, AFM does
the nanoparticles and the electrode by con- not require atomically flat sample surfaces.
trolling the redox state of the bipyridinum Furthermore, AFM allows one to image the
moiety. They measured the decay constant sample first and then place the tip over the
(k in Eq. 1) that dropped sharply on reduc- region of interest.
•
tion of bipy2+ to bipy + , from ∼16 nm−1 As mentioned before, AFM can measure
to ∼7 nm−1 . They attributed the observed surface forces using two different opera-
change to the injection of charge into the tion modes, DC and AC. In the DC mode,
redox group, and thus showed that the one measures the deflection (Z) of a
electronic properties of the junction could cantilever as a function of the tip-sample
be switched by the redox state. They also distance (D) that is varied usually by a
measured STS in mesitylene and observed piezoelectric transducer (Fig. 3). The tip-
a sharp peak in the dI /dVbias versus Vbias sample force is given by Hooke’s law in
plot (Fig. 4). Because of the stability of terms of Z, F = kZ, where k is the
•
the bipy + state, the peak was interpreted spring constant of the cantilever. This force
as a two-step electron transfer through clearly depends on the tip-sample distance,
a temporary population of the molecular D, which is given by D = Z − Z, in the
redox state. absence of sample deformation. Z in the
expression is the displacement of the piezo
3.3.4.2 Force Spectroscopy and the one that can be controlled in the
AFM has been applied to study various experiment. When the tip is far away from
surface forces [44, 115–117]. The principle the sample (large D), the force is zero.
is similar to the surface force appara- When the tip approaches the sample, it ex-
tus (SFA) that was developed some years periences various forces, electrostatic, van
ago [45]. SFA measures the force between der Waals, double-layer, solvation forces,
two curved atomically smooth surfaces. and so on. This is the regime of interest in
Molecule (1) used in redox gate
STM tip
N N
S S
8 8
2 Br
Au nanoparticle s
6 nm
Counter
electrode
Redox
e− gate 3 nm
Au substrate
30
Current
5
[nA]
20
[nA V−1]
0
dI/dV
−2 0
10 Substrate bias
[V]
−2.0 −1.5 −1.0 −0.5 0.0

Substrate bias
[V]
Fig. 4 (a) Schematic representation of a nanoscopic device. Electrons
can be injected into the redox gate by applying a suitable potential
between the substrate and the counterelectrode. (b) A current–voltage
curve (inset) and its derivative (dI/dVbias ) in mesitylene. The dotted line
corresponds to the control experiment, in which gold nanoparticles were
deposited on a hexanedithiol monolayer. (From Ref. [114] with
permission.)
terms of measuring surface forces. Finally, As we have already mentioned, the AC

when D approaches zero, the tip is in con- mode involves an oscillating AFM tip.
tact with the sample surface. The plots When the tip is driven towards a sample
of Z versus Z provide the tip-sample surface, the oscillation is altered by the
force, F (D), and are often referred to as tip–sample interactions. If the cantilever
force-distance curves. is driven to oscillate at a fixed frequency,
the changes in the oscillation amplitude force is given by −Ha R/6D 2 [45], where
can be used to obtain the tip-sample force ‘‘–’’ sign reflects attractive force, Ha is
information. An alternative approach is Hamaker constant.
to use a feedback to constantly monitor Assuming two semi-infinite surfaces
the resonance frequency that is changed with fixed surface charge density, Parse-
by the tip–sample interactions. The AC gian and Gingell [121] obtained an an-
force-distance curves are often referred alytical expression for the double-layer
to as dynamic force-distance curves. We force in the limit of small surface poten-
note that for small oscillation amplitude,
tials (<25 mV). Butt [122] calculated the
the change in the resonance frequency is
double-layer force between a flat sample
proportional to the derivative of the force
and a spherical tip with radius R using the
with respect to the tip-sample distance or
Derjaguin approximation. If KD R 1, his
differential stiffness (Eq. 5).
results can be simplified to
3.3.4.2.1 Double-layer Forces Double-

2πR
layer forces arise when two charged Fdl = [(σ 2 + σS2 )e−2KD D
ε0 εKD T
surfaces separated with an electrolyte
approach each other. The charge on each + 2σT σS e−KD D ] (12)
surface is balanced by an equal but
oppositely charged region of counterions
where σT , σS are the surface charge
in the electrolyte. Some of the counterions
densities of the tip and sample, and
are adsorbed on the surface, either
KD is the inverse of the Debye length
transiently or permanently, within the
(1/KD ). The Debye length depends on the
so-called Helmholtz layer. Beyond the
Helmholtz layer is a so-called diffuse concentration and the valence of ions and
electric double layer in which the ions is given by
are in rapid thermal motion. When
bringing two such charged surfaces 1 c
= nm (13)
together, a double-layer force appears that KD [X]
can be traced to an electrostatic and
entropic contribution. The electrostatic
where [X] is the electrolyte concentration
contribution favors an attraction between
in moles, c = 0.304, 0.176, and 0.152 for
the surfaces because they are pulled
1 : 1, 1 : 2 (or 2 : 1), and 2 : 2 electrolytes,
together by the oppositely charged
counterions. The entropic contribution respectively.
is, however, repulsive because the If σT < σS , that is, the tip is neutral,
configurational entropy favors more space the first term in Eq. (6) dominates and
between the surfaces. Because the van der the double-layer force decays exponentially
Waals force always presents between two with a decay length, 1/2KD . This term is
surfaces regardless the surface charge, it always positive and thus gives a repulsive
needs to be considered together with the force, while the second term can be either
double-layer force. Such a treatment was repulsive or attractive. The second term de-
given by the well-known DLVO theory [45]. cays with 1/KD , slower than the first term,
For a spherical tip geometry with a radius so the second term dominates the solvation
R and a flat substrate, the van der Waals force at a large tip-sample separation.
The above models assume constant to the metal electrode relative to a refer-
surface charge densities. Assuming con- ence electrode in the electrolyte [134–137].
stant surface potentials, Lin and cowork- Hillier and coworkers [137] studied the
ers [123] obtained force between a silica probe and a gold
electrode (Fig. 5). The force was attractive
Fdl = 4πεε0 RKD ϕT ϕS e−KD D (14)
at open circuit, but it depended strongly
where ϕT and ϕS are surface potentials on the potential applied to the gold and the
of the tip and the sample, respectively. type of the electrolyte. It was attractive at
In most practical cases, it is likely that positive potentials and became repulsive
neither the surface charge nor the potential at negative potentials. This was because
stays constant. the silica tip was negatively charged at a
AFM has been used to measure double- pH of 5.5. They carried out a systematic
layer forces between many insulator sur- study of the force as function of the ap-
faces, including Si3 N4 tips on mica, dia- plied potential and found that the repulsive
mond tips on mica, glass tips on glass, ZnS force decreased as the potential increased
colloidal particle on mica, and polystyrene and passed through a minimum, which
colloidal particle on polystyrene colloidal they referred to as the potential of zero
particle [119, 124, 125]. In these papers, force (pzf). They measured the pzf values
the experimental data were fit with the in solutions of NaF, KCl, and KI, which
second term of Eq. (6) or Eq. (8). The showed a close correspondence between
fitting allowed Weisenhorn and cowork- pzf and pzc (potential of zero charge).
ers [124] to extract the Debye length, which At large tip-surface separations, theoreti-
was found to agree nicely with Eq. (7). In cal fits of the measured force to solutions
addition to ionic concentration, other ex- of the Poisson–Boltzmann equation were
perimental parameters, such as pH, were excellent. At small separations, however,
also varied, which allows the determina-
neither constant surface charge (Eq. 6) nor
tion of the point of zero charge (isoelectric
constant surface potential boundary con-
point) [126].
ditions (Eq. 8) provided good fits. The poor
The double-layer forces between a metal
fits at small separations were attributed
and an insulator as well as between two
to an overestimate of the Hamaker con-
metal surfaces have also been studied. For
stant, an ill-defined position for the plane
example, Biggs and coworkers [127, 128]
have measured forces between a gold of surface charge, and the presence of
colloidal sphere and a flat gold substrate short-range solvation forces.
as a function of ionic concentration and Raiteri and coworkers [135] carried out a
pH. Several groups have carried out similar measurement of the force between
measurements of the double-layer forces a silicon nitride tip and a platinum or
with molecular adsorptions taking place gold substrate and between two conductive
on the metal surface [129–133]. surfaces, a gold sphere glued to an AFM
Most of the studies were focused on rel- cantilever and a gold electrode in 0.1 M
ative low surface potentials because of the KCl. They observed a large decrease in a
difficulty of making clean surfaces with repulsive force around the pzc of the gold
high potentials. In the case of metal sur- electrode, which agreed qualitatively with
faces, one can easily obtain a high surface the Poisson–Boltzmann theory. Arai and
potential by applying an external potential Fujihira [138] observed a similar repulsive
1.0
−700 1.0
−500 0.8 −700
0.8 −400
V −300 −400
0.6 V −200
−200
[mN m−1]
[mV]
−100 0.4 [mV] −100
F/R
0.6 0 0
+100 0.2 +100
0.0
[mN m−1]
0.4
−0.2
F/R
−0.4
0.2 0 2 4 6 8 10 12 14 16 18 20
d
[nm]
0.0
−0.2
−0.4
0 10 20 30 40 50 60
d
[nm]
Fig. 5 Force between silica sphere and gold electrode in an aqueous solution of 1 mM KCl at
25 ◦ C and pH = 5.5 as a function of the applied potential at gold electrode. The curves
correspond to, from top to bottom, electrode potentials of −700, −500, −400, −300, −200,
−100, 0, and 100 mV (vs SCE). Electrostatic repulsion decreases as the electrode potential
increases from −700 to 100 mV. Inset: force data for silica sphere and gold substrate in
10 mM KCl solution. (From Ref. [137] with permission.)
force between a gold-coated AFM tip and a When the tip approaches the substrate, it
gold substrate at pH = 10.9. has to break the order or hydrogen-bonded
network, so the measurement of solvation
3.3.4.2.2 Solvation Forces When the tip- forces provides ordering information and
sample distance is smaller than a few nm, discrete nature of solvent molecules near
the DLVO theory, that is, an attractive van the solid surfaces. The most general type
der Waals force and a repulsive double- of solvation force is oscillatory, arising
layer force, is often not adequate. One from the ordering of the solvent molecules
possible reason is the break down of into quasi-discrete layers near the surface.
the approximation that both the tip and This solvation force can be approximately
the substrate are treated as continuous described by a cosine function with an
media. Another reason is because other exponentially decaying amplitude [45],
forces arise at the small distances. One
of such forces is the so-called solvation 2πD
∼ cos e−D/d (15)
force that arises when solvent molecules d
between the tip and substrate are ordered
or form hydrogen-bonded network under where d is the diameter of the solvent
the influence of the tip and the substrate. molecule. Both the oscillation period and
the characteristic decay length are close oscillations reflect the layered structure
to d. of the solvents near the surface, and the
AFM has been used to study the sol- oscillation periods give the spacing be-
vation forces due to interfacial water tween two adjacent layers. For OMCTS,
and organic liquids [139–142]. Hoh and the period was found to be ∼0.82 nm,
coworkers [140] observed a discrete (os- in good agreement with the oblate shape
cillating) adhesive interaction between an of OMCTS with a major diameter of
AFM tip and a solid surface. They at- 1.0 to 1.1 nm and a minor diameter of
tributed the oscillations to either individual 0.7 to 0.8 nm. The data for mesitylene
hydrogen bonds between the tip and sur- was similar to that of OMCTS, except
face, or ordered water layers near the that the average period of the layers was
surface. O’Shea and coworkers used an measured to be ∼0.45 nm. This spac-
AFM in the DC mode to measure the ing is consistent in that the mesitylene
force on the AFM tip in ocatamethyl- rings lie flat with respect to the graphite
cyclotetrasiloxane (OMCTS), n-dodecanol, surface.
and water during the tip approaching the
graphite surface [141]. For both OMCTS 3.3.4.2.3 Surface Stress Surface stress is
and n-dodecanol, they observed an oscilla- closely related to another quantity, surface
tory force that was related to the layered tension (γ ), the energy cost per unit area
structure of the solvents. But the oscillatory to create a surface, according to
force did not appear in the case of water.

Later they used a magnetically oscillated ∂γ
σij = γ δij + (16)
tip microscope to directly measure the lo- ∂εij T
cal compliance of ordered liquid layers of
OMCTS and n-dodecanol on graphite and where σij , and εij are components of
on mica [141]. However, the spacing be- the surface stress and strain tensors, and
tween layers of OMCTS was substantially δij = 1 when i = j and δij = 0 when
smaller than the known smallest dimen- i = j . So, in general, the surface tension
sion of the molecule so that an orientation and the surface stress are not identical.
of the molecules in the layers could not However, the two quantities are identical
be determined. when the surface tension is indepen-
Using an AFM operated in the mag- dent of small strains (the second term
netic AC mode, Han and Lindsay [142] in Eq. 9 vanishes). This is true only for
determined the layered structure of sol- liquid or liquidlike surfaces that are free
vents at interfaces with high accuracy to rearrange in response to small pertur-
and reproducibility (Fig. 6). They found bations. Surface stress is an important
that the AFM cantilever amplitude oscil- thermodynamic quantity in electrochem-
lated in OMCTS and mesitylene as the tip istry and often known as ‘‘electrocapil-
approached the surface graphite. Similar larity’’. The name originates from the
oscillations were also found when the tip measurements of the interfacial tension
retracted away from the surface. The os- between a mercury electrode and an elec-
cillations became smeared out gradually trolyte in the earliest days of electrochem-
when increasing the tip-surface distance, istry [143]. Since the surface tension of
showing a transition from highly ordered a liquid electrode is identical to the sur-
solidlike structure to liquid structure. The face stress, the Lippmann equation can
(a)
Oscillation amplitude of cantilever 2
(b)
[Å]
2 ∆x
0 1 2 3 4 5 6 7 8 9 10
Z position
[nm]
Fig. 6 Two independent measurements of force spectra of the MAC mode
AFM at OMCTS-graphite interface. The oscillation amplitude of the magnetic
cantilever driven by an external magnetic field oscillates in both approaching
(solid line) and retracting (dotted line) curves in the region of a few nanometers
away from the surface due to ordered layers of OMCTS molecules at the
interface. The arrows on this and subsequent plots correspond to
repulsive-force maxima. (From Ref. [142] with permission.)
be derived for the mercury electrode that to be a useful quantity to measure in

relates the electrode potential, the sur- today’s research. For example, an elec-
face charge, and the interfacial stress. trochemical/thermodynamic procedure to
The Lippman equation predicts that the determine surface tension, excess, and
interfacial tension goes through a max- pressure has been developed and applied
imum at pzc, which has been used to to study molecular adsorptions on well-
determine pzc. Surface tension continues defined crystals [144].
Surface stress can be measured by metals on one of the two faces are used to
a number of experiments, including satisfy both requirements.
piezoelectric elements [145], ribbon elec- The microfabricated cantilever-based
trodes [146], and more recently devel- surface stress measurements have re-
oped micromechanical sensors [147–149]. cently found many applications: changes
It can be also measured with AFM in mass [154], heat dissipation [154, 155],
and STM, as demonstrated by several magnetic field [156], infrared light [157],
groups [150, 151]. The AFM and STM- and polarized light [158], to name a few.
based surface stress measurements usu- There is a recent surge of interest in using
ally detect the bending of a thin plate the cantilevers for chemical and biological
(cantilever) as a result of a surface stress sensors in ambient air or aqueous environ-
change at one of the two surfaces. If the ments [159–161]. For example, if one side
plate is isotropic and freely standing, the of the cantilever with receptor molecules
change in the surface stress can be calcu- is coated and the other side is left ‘‘inert’’,
lated from the amount of bending by [152] then a specific binding of chemical or bio-
logical species onto the receptor molecules
Et 2 results in a surface stress change in the
σ = z (17)
4L2 (1 − ν) cantilever. Fritz and coworkers [162] stud-
ied DNA hybridization and receptor-ligand
where E, t, L, and ν are the thick- binding using microfabricated cantilevers
ness, length, Young’s modulus, and Pois- that were functionalized with a selection
son ratio of the cantilever, respectively. of biomolecules, and they were able to de-
This equation has been widely used to tect a single base mismatch between two
determine changes in surface stress al- 12-base pair DNA oligomers.
though one end of the cantilever is often The microfabricated cantilevers have
clamped in experiments. Dahmen and been used to study surface stress induced
coworkers [153] recently performed a finite by electrochemical processes [163–166].
element analysis of single crystal plates Brunt and coworkers [165, 166] monitored
(anisotropic) with one end clamped. They surface stress changes due to the electro-
concluded that Eq. (10) derived for the un- chemical deposition and dissolution of a
constrained bending could still be used in Pb and Ag on an Au(111)-coated AFM can-
most cases with appropriate parameters. tilevers. The changes corresponded nicely
It is clear from Eq. (10) that the bending with the peaks in the cyclic voltammo-
of a cantilever because of a surface stress grams. Raiteri and coworkers [164] de-
is ∼(L/t)2 σ , which requires long and posited the two sides of a gold-coated
thin cantilevers for high sensitivity. When cantilever with different thiols, octade-
trying to improve the sensitivity, one has to canethiol, and 3-mercaptoprionic acid, re-
consider also the noise caused by environ- spectively. The cantilever was exposed to
mental vibration in the measurement. In a constant flow of aqueous electrolyte so-
order to minimize the environmental vi- lution with varying pH and the surface
bration, one requires a high-resonance fre- stress-induced deflection in the cantilever
√
quency, f = 0.163(t/L2 ) E/ρ, in which was constantly measured. They found that
ρ is the density of the cantilever material. the change in the surface stress had a
In AFM, microfabricated silicon or sili- minimum around pH 4 to 5. Increas-
con nitride cantilevers coated with various ing pH above 5 or decreasing it below 4
resulted in a bending of the cantilever suitable if one wants to study the sur-
toward the mercaptoprionic acid side. They face stress of a well-defined single crystal.
attributed the observation to the pro- This task can be performed with an STM
ton dissociation of the carboxyl group, setup [151, 167–169]. The sample in STM
which lowered the free enthalpy of the can be either a single crystal plate or ther-
carboxylate group. However, the simple mally evaporated metal films on glass plate
mechanism was not adequate to explain with one end clamped. One side of the sin-
the small increase in the surface stress gle crystal plate must be covered with, for
when reducing the solution pH below 4. example, nail polish. An STM tip is placed
When replacing the mercaptoprionic acid over the other end to detect the surface
monolayer with 2-aminoethanethiol and stress-induced bending in the plate. The
2-mercaptoethanol, the surface stress in- bending causes a change in the tunnel-
creased monotonically with increasing pH. ing gap between the tip and sample and
Another difference was that the surface is reflected in a change of the tunneling
stress change was much smaller than current. The STM feedback maintains a
that of mercaptoprionic acid, in consis- constant current by adjusting the vertical
tence with the smaller degrees of proton position of the STM via a z-piezoelectric
dissociation of 2-aminoethanethiol and 2- transducer. The amount of the adjustment
mercaptoethanol. gives the deflection, z, from which the
Measuring the surface stress of the AFM surface stress is extracted using Eq. (10).
cantilever has a number of advantages over A nice feature is that the same STM tip
other approaches, but the approach is not allows one also to image the surface.
3.2 4.6
Au(100) 1 × 1 (100) (1 × 1)
(100) (hex)
Au(100) hex
4.4
Surface stress (111) t (s)
2.8 Surface stress (100) t (s)

Au(111) 1 × 1
4.2
[N m−1]
[N m−1]
2.4
Au(111) rec
4.0
2.0
(111) (1 × 1) 3.8
(111) (rec)
1.6 3.6
−400 −200 0 200 400 600 800 1000
Potential vs. SCE
[mV]
Fig. 7 Surface stress versus the potential for the initially reconstructed and
unreconstructed surfaces of Au(100) and Au(111). All measurements are
from negative to positive potentials. (From Ref. [151] with permission.)
Using STM, Haiss and coworkers [168] One important example is scanning tun-
studied the surface stress in the presence neling spectroscopy that allows extrac-
of anion adsorption on Au(111) as a function of electronic states of the probed
tion of the electrode potential. They found a molecules. The electronic information can
linear dependence of the surface stress on be used to identify structurally similar
the surface charge, with the slope depend- molecules, to obtain chemical reactivity
ing on the type of electrolyte. However, of the molecules, and is directly relevant to
the finding does not appear to be uni- the current effort of developing molecular
versal since the surface stress of Au(100) electronics. However, in order to reliably
deviates significantly from the linear be- extract the information, improvements in
havior at negative potentials [170]. Bach both the experimental method for measur-
and coworkers [151] measured the change ing tunneling spectroscopy and theoretical
of the surface stress in the reconstruction description of the STM imaging mecha-
of the Au(111) and the Au(100) surfaces nism in the electrochemical environment
in 0.1 M HClO4 (Fig. 7). By controlling the are needed. Another important example is
electrochemical potential of the metal elec- the local force spectroscopy with AFM,
trodes, the reconstruction was reversibly which reveals not only various surface
lifted. For both surfaces, the reconstruc- forces but also mechanical properties of
tion relaxed the tensile stress by 22% and individual molecules. The microfabricated
5%, respectively, which can be understood cantilevers used as force sensors in AFM
since the average bond distances between have found broad applications in chem-
the surface atoms are shorter in the recon- ical and biological sensors, in detecting
structed phases. In the case of the Au(111) infrared light, magnetic field, and other
surface, the energy gain due to the relax- physical quantities and properties. Other
ation of the surface stress as estimated nonimaging applications of STM and AFM
from the Frenkel–Kontorova model was in conjunction with the electrochemical
close but not quite large enough to make techniques include nanofabrication and
the reconstruction energetically favored nanomanipulation, which will continue to
without the formation of the secondary play important roles in the emerging field
herringbone structure of the solitons. For of nanoscience and nanotechnology.
the Au(100) surface, the energy gain was
found to be too small to cause the forma- Acknowledgments
tion of the reconstruction.
I am grateful to Shaopeng Wang, Salah
3.3.5 Boussaad, Huixin He, and Yuwen Zhao
Conclusion for their comments and discussions. The
lab has been supported by grants from
Since the first applications of STM and NSF (CHE-9818073), PRF (33516-AC5,
AFM in electrochemistry more than a administered by the ACS), and AFOSR
decade ago, both techniques have matured (F49620-99-1-0112).
into powerful tools for imaging various
electrochemical phenomena with high res- References
olution. In addition to structural studies,
many nonimaging applications based on 1. G. Binnig, H. Rohrer, C. Gerber et al., Phys.
STM and AFM have been demonstrated. Rev. Lett. 1982, 49, 57–61.
2. G. Binnig, C. F. Quate, C. Gerber, Phys. 28. S. M. Lindsay, O. F. Sankey, Y. Li et al., J.

Rev. Lett. 1986, 56, 930–933. Phys. Chem. 1990, 94, 4655–4660.
3. R. Sonnenfeld, P. K. Hansma, Science 1986, 29. N. D. Lang, Phys. Rev. Lett. 1986, 56,
232, 211–213. 1164–1167.
4. H. Y. Liu, F. R. F. Fan, C. W. Lin et al., J. 30. V. Mujica, M. Kemp, M. A. Ratner, J. Chem.
Am. Chem. Soc. 1986, 108, 3838, 3839. Phys. 1994, 101, 6849–6864.
5. S. Manne, H. J. Butt, S. A. C. Gould et al., 31. H. Ou-Yang, B. Kallebring, R. A. Marcus, J.
Appl. Phys. Lett. 1990, 56, 1758, 1759. Chem. Phys. 1993, 98, 7565–7824.
6. S. Manne, P. K. Hansma, J. Massie et al., 32. A. J. Fisher, P. E. Blöchl, Phys. Rev. Lett.
Science 1991, 251, 183–186. 1993, 70, 3263–3266.
7. J. Lipkowski, P. N. Ross, (Eds.), Imaging of 33. D. N. Futaba, S. Chiang, J. Vac. Sci. Tech-
Surfaces and Interfaces, Wiley-VCH, New nol., A 1997, 15, 1295–1298.
York, 1999. 34. V. M. Hallmark, S. Chiang, Surf. Sci. 1995,
8. A. A. Gewirth, B. K. Niece, Chem. Rev. 1997, 329, 255–268.
97, 1129–1162. 35. X. W. Wang, N. J. Tao, F. Cunha, J. Chem.
9. A. J. Bard, H. D. Abruna, C. E. Chidsey Phys. 1996, 105, 3747–3752.
et al., J. Phys. Chem. 1993, 97, 7147–7173. 36. A. M. Kuznetsov, P. Sommer-Larsen, J. Uls-
10. T. P. Moffat, Electroanal. Chem. 1999, 21, trup, Surf. Sci. 1992, 275, 52–64.
211–316. 37. J. E. T. Andersen, A. A. Kornyshev, A. M.
11. D. M. Kolb, Electrochem. Acta 2000, 45, Kuzenetsov et al., Electrochim. Acta 1997,
2387–2402. 42, 819–831.
12. J. E. T. Andersen, J.-D. Zhang, Q. Chi et al., 38. W. Schmickler, C. Widrig, J. Electroanal.
Trend Anal. Chem. 1999, 18, 665–674. Chem. 1992, 336, 213–221.
13. C. J. Chen, Introduction to Scanning Tun- 39. H. Sumi, J. Phys. Chem. B 1998, 102,
neling Microscopy, Oxford University Press, 1833–1844.
New York, 1993. 40. A. K. Mishra, S. K. Rangarajan, Theochem:
14. J. H. Coombs, J. B. Pethica, IBM J. Res. Dev. J. Mol. Struct. 1996, 361, 101–109.
1986, 20, 455–459. 41. A. M. Kuznetsov, J. Ulstrup, Mol. Phys.
15. R. Christoph, H. Siegenthaler, H. Rohrer 1992, 87, 1189.
et al., Electrochim. Acta 1989, 34, 1011. 42. W. Schmickler, J. Electroanal. Chem. 1990,
16. J. Wiechers, T. Twomey, D. M. Kolb et al., 296, 283–289.
J. Electroanal. Chem. 1988, 248, 451. 43. M. Tortonese, R. C. Barrett, C. F. Quate,
17. M. Bingelli, D. Carnal, R. Nyffenegger et al., Appl. Phys. Lett. 1993, 62, 834–836.
J. Vac. Sci. Technol., B 1991, 9, 1985–1992. 44. S. M. Lindsay in Scanning Tunneling Mi-
18. J. Pan, T. W. Jing, S. M. Lindsay, J. Phys. croscopy and Related Techniques (Ed.:
Chem. 1993, 98, 4205–4208. D. Bonnell), John Wiley & Sons, New York,
19. A. Vaught, T. W. Jing, S. M. Lindsay, Chem. 2000.
Phys. Lett. 1995, 236, 306–310. 45. J. N. Israelachvili, Intermolecular and Sur-
20. Y. A. Hong, J. R. Hahn, H. Kang, J. Chem. face Forces, Academic Press, London, 1992.
Phys. 1998, 108, 4367–4370. 46. F. Ohnesorge, G. Binnig, Science 1993, 260,
21. W. Schmickler, D. Henderson, J. Elec- 1451–1456.
troanal. Chem. 1990, 290, 283–291. 47. F. J. Giessibl, Phys. Rev. B 1997, 56,
22. M. V. Vigdorovich, A. M. Kuznetsov, J. Uls- 16 010–16 015.
trup, Chem. Phys. 1993, 176, 539–554. 48. T. R. Albrecht, P. Grutter, D. Horne et al.,
23. G. Repphun, J. Halbritter, J. Vac. Sci. J. Appl. Phys. 1991, 69, 668–673.
Technol., A 1995, 13, 1693–1698. 49. F. J. Giessibl, Science 1995, 267, 68–71.
24. W. Schmickler, Surf. Sci. 1995, 335, 50. F. J. Giessibl, Appl. Phys. Lett. 2000, 76,
416–421. 1470–1472.
25. I. Benjamin, D. Evans, A. Nitzan, J. Chem. 51. Q. Zhong, D. Inniss, K. Kjoller et al., Surf.
Phys. 1997, 106, 1291–1293. Sci. 290, L688–L690.
26. J. Bardeen, Phys. Rev. Lett. 1960, 6, 57–59. 52. P. K. Hansma, J. P. Cleveland, M. Rad-
27. J. Tersoff, D. R. Hamann, Phys. Rev. B 1985, macher et al., Appl. Phys. Lett. 1994,
31, 805–814. 1738–1741.
53. C. A. J. Putman, K. O. Vanderwerf, B. G. 77. R. M. Feenstra, J. A. Stroscio, A. P. Fein,

Degrooth et al., Appl. Phys. Lett. 1994, 64, Surf. Sci. 1987, 181, 295–306.
2454–2456. 78. H. Morisaki, H. Nishkawa, H. Ono et al., J.
54. S. M. Lindsay, Y. I. Lyubchenko, N. J. Tao Electrochem. Soc. 1990, 137, 2759–2763.
et al., J. Vac. Sci. Technol. 1993, 11, 808–815. 79. H. Gerischer, Z. Phys. Chem. N. F. 1961, 27,
55. E. I. Florin, M. Radmacher, B. Fleck et al., 48.
Rev. Sci. Instrum. 1994, 65, 639–643. 80. E. P. Friis, Y. I. Kharkats, A. M. Kuznetsov
56. W. Han, S. M. Lindsay, T. Jing, Appl. Phys. et al., J. Phys. Chem. A 1998, 102,
Lett. 1996, 69, 4111–4113. 7851–7859.
57. A. A. Kornyshev, M. Sumetskii in Electro- 81. A. M. Kuznetsov, J. Ulstrup, J. Phys. Chem.
chemical Nanotechnology: In situ Local Probe A 2000, 104, 11 531–11 540.
Techniques of Electrochemical Interfaces (Ed.: 82. S. R. Snyder, H. S. White, J. Electroanal.
W. J. L. A. W. Plieth), Wiley-VCH, Wein- Chem. 1995, 394, 177–185.
heim, Germany, 1998, pp. 45–55. 83. N. J. Tao, Phys. Rev. Lett. 1996, 76,
58. N. J. Tao, C. Z. Li, H. X. He, J. Electroanal. 4066–4069.
Chem. 2000, 492, 81–93. 84. W. Schmickler, N. J. Tao, Electrochim. Acta
59. R. M. Nyffenegger, R. M. Penner, Chem. 1997, 432, 2809–2815.
Rev. 1997, 97, 1195–1230. 85. W. Han, E. N. Durantini, T. A. Moore et al.,
60. L. A. Nagahara, T. Thundat, S. M. Lindsay, J. Phys. Chem. 1997, 10, 10 719–10 725.
Appl. Phys. Lett. 1990, 57, 270–272. 86. X. Lu, K. W. Hipps, X. D. Wang et al., J.
61. D. M. Kolb, R. Ullmann, T. Will, Science Am. Chem. Soc. 1996, 118, 7197–7202.
87. U. Mazur, K. W. Hipps, J. Phys. Chem. B
1997, 275, 1097–1099.
1999, 103, 9721–9727.
62. C. B. Ross, L. Sun, R. M. Crooks, Langmuir
88. L. Scudiero, D. E. Barlow, K. W. Hipps, J.
1993, 9, 632–636.
Phys. Chem. B 2000, 104, 11 899–11 905.
63. W. J. Li, J. A. Nirtanen, R. M. Penner, Appl.
89. X. Lu, K. W. Hipps, J. Phys. Chem. B 1997,
Phys. Lett. 1992, 60, 1181–1183.
101, 5391–5396.
64. R. T. Potzshke, G. Staikov, W. J. Lorentz
90. U. Mazur, K. W. Hipps, J. Phys. Chem. 1995,
et al., J. Electrochem. Soc. 1999, 146,
99, 6684–6688.
141–149.
91. J. E. T. Andersen, P. P. Moller, M. V. Peder-
65. U. Stimming, R. Vogel, D. M. Kolb et al., J. sen et al., Surf. Sci. 1995, 325, 193–205.
Power Sources 1993, 43, 169–180. 92. B. Zhang, E. Wang, Probe Microsc. 1997, 1,
66. D. Hofmann, W. Schindler, J. Kirchner, 57–60.
Appl. Phys. Lett. 1998, 73, 3279–3281. 93. J. D. Zhang, Q. Chi, S. J. Dong et al., J.
67. R. Schuster, V. Kirchner, X. H. Xia et al., Electroanal. Chem. 1996, 379, 535–539.
Phys. Rev. Lett. 1998, 80, 5599–5602. 94. E. P. Friis, J. E. T. Andersen, Y. I. Kharkats
68. C. Z. Li, N. J. Tao, Appl. Phys. Lett. 1998, 72, et al., Proc. Natl. Acad. Sci. USA 1999, 96,
894–897. 1379–1384.
69. G. Binnig, H. Fuchs, E. Stoll, Surf. Sci. 95. E. P. Friis, J. E. T. Andersen, P. Moller et al.,
1986, 169, L295–L298. J. Electroanal. Chem. 1997, 431, 35–38.
70. M. L. Lozano, M. C. Tringides, Europhys. 96. J. J. Davis, C. M. Halliswell, H. O. A. Hill
Lett. 1995, 30, 537–542. et al., New J. Chem. 1998, 22, 1119–1123.
71. M. Sumetskii, A. A. Kornyshev, Phys. Rev. 97. I. Lee, J. W. Lee, R. J. Warmack et al., Proc.
B 1993, 48, 17 493–17 506. Natl. Acad. Sci. USA 1995, 92, 1965–1969.
72. M. Sumetskij, A. A. Kornyshev, U. Stimm- 98. J. H. Schott, H. S. White, Langmuir 1994,
ing, Surf. Sci. 1994, 307, 23–27. 10, 486–491.
73. O. M. Magnussen, M. R. Vogt, Phys. Rev. 99. D. M. Cyr, B. Venkataraman, G. W.
Lett. 2000, 85, 357–360. Flynn et al., J. Phys. Chem. 1996, 100,
74. M. Giesen-Seibert, R. Jentjens, M. Poens- 13 748–13 759.
gen et al., Phys. Rev. Lett. 1993, 71, 100. H. B. Fang, L. C. Giancarlo, G. W. Flynn, J.
3521–3524. Phys. Chem. B 1999, 103, 5712–5715.
75. M. Giesen, R. Randler, S. Baier et al., Elec- 101. C. L. Claypool, F. Faglioni, W. A. Goddard
trochim. Acta 1999, 45, 527–536. et al., J. Phys. Chem. B 1997, 101,
76. I. Giaever, Phys. Rev. Lett. 1960, 5, 147, 148. 5978–5995.
102. C. L. Claypool, F. Faglioni, W. A. Goddard 129. V. Kane, P. Mulvaney, Langmuir 1998, 14,
et al., J. Phys. Chem. B 1999, 103, 3303–3311.
7077–7080. 130. B. A. Hu K, Langmuir 1997, 13, 5418–5425.
103. C. Joachim, J. K. Gimzewski, A. Aviram, 131. C. Z. Hu K, J. K. Whitesell, A. J. Bard, Lang-
Nature 2000, 408, 541–548. muir 1999, 15, 3343–3347.
104. P. Ball, Nature 2000, 406, 118–120. 132. S. Manne, J. P. Cleveland, H. E. Gaub et al.,
105. L. A. Bumm, J. J. Arnold, M. T. Cygan et al., Langmuir 1994, 10, 4409–4413.
Science 1996, 271, 1705–1707. 133. I. Larson, D. Y. C. Chan, C. J. Drummond
106. G. S. McCarty, P. S. Weiss, Chem. Rev. et al., Langmuir 1997, 13, 2429–2431.
1999, 99, 1983–1990. 134. R. Raiteri, M. Preuss, M. Grattarola et al.,
107. G. Leatherman, E. N. Durantini, D. Gust Colloids Surf., A 1998, 136, 191–197.
et al., J. Phys. Chem. B 1999, 103, 135. R. Raiteri, M. Grattarola, H.-J. Butt, J. Phys.
4006–4010. Chem. 1996, 100, 16 700–16 705.
108. Y. Q. Xue, S. Datta, S. Hong et al., Phys. 136. T. Ishino, H. Hieda, K. Tanaka et al., Jpn. J.
Rev. B 1999, 59, R7852–R7855. Phys. 1994, 33, L1552–L1554.
109. D. J. Wold, C. D. Frisbie, J. Am. Chem. Soc. 137. A. C. Hillier, S. Kim, A. J. Bard, J. Phys.
2000, 122, 2970, 2971. Chem. 1996, 100, 18 808–18 817.
110. S. Datta, W. D. Tian, S. H. Hong et al., 138. T. Arai, M. Fujihira, J. Vac. Sci. Technol., B
Phys. Rev. Lett. 1997, 79, 2530–2533. 1996, 14, 1378–1382.
111. C. Joachim, J. K. Gimzewski, Chem. Phys. 139. J. B. Pathica, W. C. Oliver, Phys. Scr. 1987,
Lett. 1997, 265, 353–357. T19, 61–66.
112. V. J. Langlais, R. R. Schlittler, H. Tang et al., 140. J. H. Hoh, J. P. Cleveland, C. B. Prater et al.,
Phys. Rev. Lett. 1999, 83, 2809–2812. J. Am. Chem. Soc. 1992, 114, 4917, 4918.
113. J. K. Gimzewski, C. Joachim, Science 1999,
141. S. J. O’Shea, M. E. Welland, T. Rayment,
283, 1683–1688.
Appl. Phys. Lett. 1992, 60, 2356–2358.
114. D. I. Gittins, D. Bethell, D. J. Schiffrin et al.,
142. W. Han, S. M. Lindsay, Appl. Phys. Lett.
Nature 2000, 408, 67–69.
1998, 72, 1656–1658.
115. E. Meyer, H. Heinzelmann, P. Grutter et al.,
143. A. J. Bard, L. R. Faulkner, Electrochemical
J. Microsc. 1988, 152, 269.
Methods, John Wiley & Sons, New York,
116. B. Cappela, G. Dietler, Surf. Sci. Rep. 1999,
1980.
34, 1–104.
117. S. Manne, H. E. Gaub, Curr. Opin. Colloid 144. J. Lipkowski, L. Stolberg in Adsorption
Interface Sci. 1997, 2, 145–152. of Molecules at Metal Electrodes (Eds.:
118. W. A. Ducker, T. J. Senden, R. M. Pashley, J. Lipkowski, P. N. Ross), VCH Publishers,
Nature 1991, 353, 239–241. New York, 1992, pp. 171–237.
119. Y. Q. Li, N. J. Tao, J. Pan et al., Langmuir 145. R. E. Malpas, R. A. Fredlein, A. J. Bard, J.
1993, 9, 637–641. Electroanal. Chem. 1979, 98, 171.
120. H. J. Dai, J. H. Hafner, A. G. Rinzler et al., 146. K. F. Lin, T. R. Beck, J. Electrochem. Soc.
Nature 1996, 384, 147–151. 1976, 123, 1145.
121. V. A. Parsegian, D. Gingell, Biophys. J. 147. R. E. Martinez, W. M. Augustyniak, J. A.
1972, 12, 1192. Golovchenko, Phys. Rev. Lett. 1990, 64,
122. H.-J. Butt, Biophys. J. 1991, 60, 1438–1444. 1035–1038.
123. X. Y. Li, F. Creuzef, H. Arribart, J. Phys. 148. A. J. Schell-Sorokin, R. M. Tromp, Phys.
Chem. 1993, 97, 7272–7276. Rev. Lett. 1990, 64, 1039–1042.
124. A. L. Weisenhorn, P. Maivald, H.-J. Butt 149. D. Sader, H. Ibach, Phys. Rev. B 1991, 43,
et al., Phys. Rev. B 1992, 45, 11 226–11 232. 4263–4267.
125. D. T. Atkins, R. M. Pashley, Langmuir 1993, 150. R. Berger, E. Delamarche, H. P. Lang et al.,
9, 2232–2236. Science 1997, 276, 2021–2023.
126. R. Raiteri, S. Martinoia, M. Grattarola, Bio- 151. C. E. Bach, M. Giesen, H. Ibach et al., Phys.
sens. Bioelectron. 1996, 11, 1009–1017. Rev. Lett. 1997, 78, 4225–4228.
127. S. Biggs, P. Mulvaney, J. Chem. Phys. 1994, 152. G. G. Stoney, Proc. R. Soc. London, Ser. A
100, 8501–8505. 1909, 82, 172–183.
128. S. Biggs, P. Mulvaney, C. Zukoski et al., J. 153. K. Dahmen, S. Lehwald, H. Ibach, Surf. Sci.
Am. Chem. Soc. 1994, 116, 9150–9157. 2000, 446, 161–173.
154. R. Berger, H. P. Lang, C. Gerber et al., 163. H.-J. Butt, J. Colloid Interface Sci. 1996, 180,
Chem. Phys. Lett. 1998, 294, 363–369. 251–260.
155. J. R. Barnes, R. J. Stephenson, M. E. Well- 164. R. Raiteri, H.-J. Butt, M. Grattarola, Elec-
and et al., Nature 1994, 372, 79–81. trochem. Acta 2000, 46, 157–163.
156. R. P. Cowburn, A. M. Moulin, M. E. Well- 165. T. A. Brunt, T. Rayment, S. J. O’Shea et al.,
and, Appl. Phys. Lett. 1997, 71, 2202–2204. Langmuir 1996, 12, 5942–5946.
157. P. I. Oden, P. G. Datskos, T. Thundat et al., 166. T. A. Brunt, E. D. Chabala, T. Rayment
Appl. Phys. Lett. 1996, 69, 3277–3279. et al., J. Chem. Soc., Faraday Trans. 1996,
158. P. Kremmer, A. M. Moulin, R. J. Stephen- 92, 3807–3812.
son et al., Science 1997, 277, 1799–1802. 167. W. Haiss, J. K. Sass, J. Electroanal. Chem.
159. H. P. Lang, M. K. Baller, R. Berger et al., 1995, 386, 267–270.
Anal. Chem. Acta 1999, 393, 59–65. 168. W. Haiss, R. J. Nichols, J. K. Sass et al., J.
160. T. Thundat, E. Finot, Z. Hu et al., Appl. Electroanal. Chem. 1998, 452, 199–202.
Phys. Lett. 2000, 77, 4061–4063. 169. H. Ibach, C. E. Bach, M. Giesen et al., Surf.
161. H. F. Ji, E. Finot, R. Dabestani et al., Chem. Sci. 1997, 375, 107–119.
Commun. 2000, 6, 457, 458. 170. H. Ibach, Electrochem. Acta 1999, 45,
162. J. Fritz, M. K. Baller, H. P. Lang et al., Sci- 575–581.
ence 2000, 288, 316–318.
3.4 In situ FTIR as a tool for mechanistic studies. Fundamentals and applications 243
3.4 made using the so-called attenuated total

In situ FTIR as a tool for mechanistic reflection (ATR), a technique largely used in
studies. Fundamentals and applications infrared spectroscopy for liquid samples.
In this technique, the infrared beam is
Francisco C. Nart and Teresa Iwasita introduced into a high refraction index
Universidade de São Paulo, São Paulo, Brazil
infrared transparent crystal and the crystal
3.4.1 put in contact with the sample. When
Historical Overview the light hits the edge of the crystal in
contact with a medium of lower refraction
Electrochemistry experienced in the recent index, the light is totally reflected. If
years the introduction of many in situ there is some absorbing species in the
techniques for the purpose of improving sample, then the refraction index will have
the understanding of the electrode process an imaginary component that changes
at a molecular level. Among these tech- the refraction indexes at the wavelength
niques, the infrared spectroscopy plays where absorption exists. The light is then
a relevant role as a tool to study, in transmitted to the liquid sample and it is
situ, the electrode surface. The molecular therefore attenuated at the detector.
specificity permits the identification of ad- The mostly used crystal as an ATR el-
sorbates and reaction products and allows ement and substrate for the electrode
the study of physicochemical properties of has been germanium [4, 5]. Germanium
adsorbed molecules. crystals have good transparency and good
The infrared spectroscopy has been used electrical conductivity, but are easily dis-
largely in Surface Science to study ad- solved at positive potentials in aqueous
sorbed molecules on surfaces in vacuum. solutions. The electrode can be deposited
The infrared beam is reflected at the crys- onto the crystal or, in some cases, the ger-
tal surface and the adsorbed molecules manium can be used as an electrode for
interact with the incident light [1, 2]. some processes in which the nature of the
Greenler proposed the basic principles electrode does not affect the reaction, such
of the reflection-absorption spectroscopy as in outer sphere reactions.
in 1966 [3]. From his work, it becomes The theory of internal reflection applied
clear that the light interacting with the to electrochemical studies has been dis-
adsorbed molecules on a metal surface is cussed in detail by Hansen [6] for both the
only the parallel mode of the electric field infrared and the visible region of the elec-
of the incident radiation. This leads to the tromagnetic spectrum. However, the ATR
so-called surface selection rule, a central confor electrochemical studies has been aban-
cept in reflection-absorption spectroscopy doned rapidly, because of the difficulties
(see Sect. 3.4.2.1.2). in obtaining an electrode with good opti-
In situ infrared spectroscopy has been cal transparency in the infrared region and
used in electrochemistry as early as good electrical conductivity. In general, the
1970 [4]. Obviously, to apply the infrared ATR crystal must be a semiconductor or
spectroscopy to the study of the electrode insulating material. In cases in which the
surface, the strong absorption of light electrode is deposited on the surface of
by the solvent must be minimized. One the element, the condition of optical trans-
of the first attempts to use infrared parency requires a very thin film in which
spectroscopy as an in situ tool was the conductivity is very low. Moreover,
these very thin films are not enough to detector (lock-in amplifier). This technique
protect the crystal from corrosion in the has been called Electrochemically Modulated
usual electrochemical solutions. Infrared Spectroscopy (EMIRS) [11]. The po-
Although ATR infrared spectroscopy tential modulation, which is a square wave,
for in situ studies of electrochemical and the spectra were collected at the two
systems presents many limitations, one potentials of the step modulation. The
interesting property of this technique is the spectra of each potential are added and
amplification of the IR signal by excitation then ratioed to obtain the R/R.
of surface plasmons. Osawa and coworkers EMIRS has been successfully applied to
used this technique to study the adsorption many systems. Briefly it can be mentioned
of water [7] and other molecules [8] on the study of adsorbates at the electrode
thin-film gold electrodes. However, the surface [10], the detection of adsorbed re-
amplification of the IR signal through action intermediates for the oxidation of
plasmon excitation is limited to metals small organic molecules [12], and the de-
such as gold, silver, and other sp metals. termination of the water structure in the
Another important application of the double layer [13]. However, the potential
ATR infrared spectroscopy is for the modulation in EMIRS is its drawback,
study of conducting polymers [9]. Most of, since it prevents the study of irreversible
these studies are conducted in nonaqueous processes as the system must return to the
media in which the corrosion of the ATR same conditions each time the potential
element is not important. is changed. Other important limitations of
Though the internal reflection mode EMIRS are related to both the electrical and
(ATR) encounters many applications, the chemical relaxation effects caused by the
external reflection mode is much more potential modulation at 12 Hz. The elec-
versatile to study metallic electrode sur- trical relaxation is due to the high ohmic
faces. The external reflection mode has drop of the electrolyte confined in the thin
been introduced to study metallic electrode solution layer required for the in situ mea-
surfaces by Bewick and coworkers [10]. To surements. The chemical relaxation is due
minimize the strong attenuation of the to ion migration induced by the change in
radiation by the presence of a solvent in solution composition caused by the elec-
electrochemical systems, the electrode was trode potential change. These aspects have
placed very close to the infrared transpar- been discussed in detail in the following
ent window. Practically, the electrode is text [14–16] (see Sect. 3.4.2.3).
slightly pressed against the flat window The introduction of Fourier Trans-
surface leaving a very thin electrolyte layer formed Infrared Spectrometers to the
between the electrode and the window. electrochemical experiments eliminated
At the beginning, a dispersive instru- the need of potential modulation to in-
ment was used. The signal-to-noise ratio crease the signal-to-noise ratio. Fourier
was very poor, since the amount of species Transform Spectrometers present two ad-
sampled was very low and the dispersive vantages compared to the dispersive equip-
instrument has a low throughput. The low ment [17], which have improved the possi-
signal-to-noise ratio was then improved bilities of the in situ measurements. One
by applying to the electrode, a potential advantage originates from the fact that
modulation of low frequency (typically be- the beam hits the sample without pass-
tween 7 and 10 Hz) and using a phase ing through a monochromator (Jacquinot
advantage), thus avoiding the losses caused has been used associated to the FTIRS.
by the monochromator components like In theory this technique allows the acqui-
slit and gratings. The net result is a better sition of only integral bands, since this
signal-to-noise ratio, eliminating the need is based on the surface selection rule
of a phase detector. The other advantage (see below Sect. 3.4.2.1.2), in which the s-
is known as multiplex (or Felget) advan- polarized light is only active away from the
tage, in which all the wave numbers can electrode surface. This technique has been
be sampled simultaneously, increasing the called Fourier transform-infrared reflection-
speed of spectral acquisition. However, the absorption spectroscopy FT-IRRAS [28], al-
spectral acquisition at different potentials though this acronym has been used also
is necessary, since the Fourier transform for the integral bands obtained without po-
equipments are single-beam devices and a larization modulation. The advantages and
reference spectrum is always needed. The limitations of this technique have been
use of Fourier Transform IR spectroscopy analyzed in detail by Seki et al. [29]. The
in in situ studies was introduced by Pons main limitation is that the reflectivity is
and coworkers [18, 19]. The technique was different for the p- and s-polarized radi-
called Subtractively Normalized Interfacial ation, and it changes with the radiation
Fourier Transform-Infrared Spectroscopy energy, the incident angle, and thickness
(SNIFTIRS) due to the way the spectra of the thin electrolyte layer. Thus the
were acquired. presence of water and ions in the so-
The use of Fourier Transform instru- lution can cause absorption bands even
ments eliminates much of the limitations for species not adsorbed on the electrode.
of the EMIRS, since no more potential Therefore, compensating mechanisms are
modulation is needed. The signal-to-noise needed and even so, it is very difficult to
ratio is far less than the dispersive instru- prevent artifacts. In the case in which the
ment and can be improved statistically by bands of the adsorbed species are very
adding more scans, since the spectral ac- close to the bands in solution, it is possi-
quisition time is much lower. With the ble to mix the signals coming from each
Fourier Transform equipment, also the ir- one [16, 30]. The polarization modulation
reversible processes can be studied, since has been used to study adsorbed CO [28]
it is no longer required to return to the and CN− [28, 29], since the bands of these
same potentials as for modulation (see species are far away from the bands of the
Sect. 3.4.4). This not only allows the ac- water and ions used for the electrochemical
quisition of derivative or bipolar bands but experiments.
also the acquisition of integral bands, as in
the case of the adsorbed CO on platinum 3.4.2
electrodes [20–25], which was impossible The External Reflectance Spectroscopy
with EMIRS. The speed of spectral acquisi-
tion of Fourier Transform Infrared (FTIR) 3.4.2.1 Fundamentals
instruments allows also the follow-up of Let us consider the reflection on a flat
a reaction during a dynamic polarization mirror-finished surface as depicted in
curve [26, 27]. Fig. 1. The complex electric field of the
To avoid the spectral acquisition at two standing wave above the surface is the re-
different potentials in order to have a ref- sult of the vectorial sum of the incident
erence spectrum, polarization modulation and the reflected beam [1]. The magnitude
Fig. 1 Reflection of the s-(⊥) and p-() polarized light on a flat mirror surface. Note the
direction of the electric field upon reflection.
of the field is a function of the incidence (2) the introduction of the surface selection
angle and of the light polarization. This rule.
is a consequence of the phase shift upon
reflection. The phase shift for the elec- 3.4.2.1.1 The Band Intensity The absorp-
tric field perpendicular to the incident tion of infrared radiation by a submono-
plane (s-polarized radiation) is almost 180◦ layer of an adsorbed species is the result
for all the incident angles. For the elec- of the interaction of the electric field of
tric field parallel to the incident plane the radiation with the electric dipoles of
(p-polarization), the phase shift is angle de- the adsorbed species and with the elec-
pendent and changes drastically between trons of the reflecting metallic surface.
60◦ and 90◦ , as can be seen in Fig. 2. Therefore, the intensity of the absorption
These results have two practical con- is determined by the electric properties of
sequences for the reflection-absorption the metal. The absorption is the nor-
spectroscopy: (1) the dependence of the malized difference between the reflectivity
band intensity on the incident angle and of the surface free from the absorbing
Angle
0 45 90
0
Phase shift
−90
−180
Fig. 2 Phase shift for the reflected s-(δ⊥ ) and p-(δ ) polarized light upon
reflection on a metallic surface. (Data from Ref. [3] with permission.)
species (R0 ) and the surface covered with Considering that for a metallic surface
an absorbing layer (R): RS ≈ 1, the electric field at the surface
is practically zero.
R − R0 For the p-polarized light, the mean
= (1)
R0 quadratic field is given by [6]
The intensity of a band in the spectrum Ep1

2
1/2
is the result of the dissipation of the = cos2 θ1 (1 + Rp ) − 2Rp
Ep1
02
electromagnetic radiation in the absorbing
z
media at a given energy. Under these
× cos δpr + 4π ξ1
circumstances, the change in the energy λ

flux is proportional to the mean quadratic 1/2
+ sin2 θ1 (1 + Rp ) + 2Rp
field (see Appendix A). Therefore, it is
z
important to understand the behavior of
× cos δpr + 4π ξ1 (5)
the electric field on a metallic surface λ
to rationalize the factors affecting the
intensity of light absorption. The change in the intensity of the electric
The mean quadratic field, referred to the field for the p-polarized radiation is shown
incident mean quadratic field in the same in Fig. 3. The intensity of the field
direction for the s-polarized radiation, is [6] increases for higher incidence angle up
to ca. 90◦ . The best angle is around 85◦ ,
ES1
2 known as the Brewster angle.
1/2
= (1 + RS ) + 2RS From this discussion, two important
ES1
02
consequences can be derived: (1) the s-
z
× cos δSr + 4π ξ1 (2) polarized light does not interact with the
λ species very close to the metallic surface
(see Sect. 3.4.2.1.2). The immediate con-
where RS is the s-polarized light reflec-
sequence is that adsorbed species cannot
tivity, δSr the phase angle of the reflected
be sampled with s-polarized light but only
radiation, λ is the wavelength of the radia-
with p-polarized light. (2) The intensity for
tion, z is the position above the reflecting
the absorption of radiation is angle depen-
surface, and ξ = n cos θ, where n is the
dent and maximum for large angles in the
refraction index and θ the incidence angle.
case of metal/air or metal/vacuum surface.
On the surface plane (z = 0), the electric
For electrochemical applications, a very
field is given by [6]
thin liquid electrolyte on the surface is
needed, and for this a transparent window
ES1
2
1/2
02
= (1 + RS ) + 2RS cos δSr (3) must be used to keep the electrolyte film
ES1 in front of the surface. This makes the
whole situation more complicated. The
since for a reflection on a metallic surface, introduction of a very thin liquid film
the phase shift is almost π, cos δSr ≈ −1 along with the transparent window (see
and the mean quadratic field for this case is below) changes the optimum incidence
angle, as can be seen in Fig. 4, in which it is
ES1
2
1/2 2 shown how the different media change the
= 1 − RS (4)
ES1
02 direction of the beam. It is worth noting the
2.0
n3 = 3.0
k3 = 30.0
n1 = 1.0
〈E z20〉
〈E z2〉
1.0
45 90
Angle
Fig. 3 Change of the mean square electric field for the air/metallic surface with
the incidence angle. The simulation parameters are shown in the figure.

qualitative differences if a flat or prismatic n1 (1 − rP )
window is used. + n1 sin θ1 S
n22
The equations for a system consist-
ing of metallic surface/thin-film elec-
2
(1 − rP )
trolyte/prismatic window were developed + in1 sin θ1 cos θ1 S
ξ2
by Seki et al. [29]. Briefly, the equations are
(7)
where
ES2
2
z
= [(1 + rS )C + i(1 − rS )S]2 (6)
ES2
02 C = cos 2πξ2 (8)
λ
z
S = sin 2πξ2 (9)
and λ
and
2
EP2
= cos θ1 (1 − rP )C rS12 + rS23 e−2iβ
EP2
02
rS = (10)
2 1 + rS12 rS23 e−2iβ
n1 (1 + rP ) rP12 + rP23 e−2iβ
+ iξ2 S rP = (11)
n22 1 + rP12 rP23 e−2iβ
Electrode
Absorbate
Electrolyte
Window
q2
Air
q1
(a)
Electrode
Absorbate
Electrolyte
Prismatic window
Air
q
Air
(b)
Fig. 4 The reflection of the light beam for an electrochemical interface, including
the window [(a) flat or (b) prismatic], showing the different internal reflections of
the beam.
and r12 and r23 are the Fresnel co- dispersion of the energy – was described
efficients for the window/electrolyte and in detail, but not the band intensity itself.
electrolyte/metal interfaces, respectively. β The band intensity obviously depends on
is the energy attenuation in the solution the magnitude of the electric field (see
phase and is given by Appendix A), but the intensity of light ab-

sorption depends on the dynamic electric
d
β = 2π ξ2 (12) dipole moment of the species interacting
λ
with the electric field. For the band inten-
where d is the thickness of the thin sity in reflection-absorption spectra, one
electrolyte film. elegant and relatively simple equation has
So far the intensity of the elec- been presented by Persson [31]. Accord-
tric field – which will be related to the ing to this model, the adsorbed adlayer
is constituted by an ensemble of electric the intensity in the reflection-absorption

dipoles and the integrated band intensity infrared spectroscopy. More specifically,
is given by both the dynamic dipole moment, which
will depend on a specific vibration, com-
16π 2 N 2 bines with the surface properties through
dν(ν) = µ G(α) (13)
h̄c A the G(α) function. Therefore, quantitative
studies from vibrational intensities must
where N/A is the density of adsorbed
take into account the incidence angle at
particles, µ is the dynamic dipole moment
the electrode surface.
operator for the dipole perpendicular to the
surface, is the vibrational frequency, and
3.4.2.1.2 Surface Selection Rule The
ν is the frequency of the incident radiation.
most important consequence of the behav-
G(α) is a function of the incident angle and
ior of the electric field upon reflection on a
of the dielectric constant ε of the metal and
metal surface is the limitation of the dipole
is defined [31] as
that can interact with the electric field at the
√
sin2 α ε cos α surface. The phase shift of the electric field
G(α) = √ (14)
cos α ε cos α − 1 upon reflection for the s-polarized light on
a metal surface leads to a vanishing electric
Equation (13) is useful in understand- field and therefore there is no interaction
ing the different components affecting with the dipoles of the adsorbed molecules.
2.0
n3 = 3.0
k3 = 30.0
n1 = 1.0
〈E 2〉
〈E 20〉
1.0
5.0 10.0
Z
[µm]
Fig. 5 Change of the intensity of the reflected mean square electric field
with the distance from the reflecting surface for the s- and
p-polarized radiation.
The s-polarized radiation becomes active molecule. This will tell which dipoles will
only for distances of ca. 5 nm above the be active for a given adsorption geom-
surface. On the other hand, the electric etry thus affording for the ascertion of
field for the p-polarized light is a maxi- the bonding of the adsorbed species to
mum at the surface and decreases when the surface. Obviously very simple linear
going away from the surface (see Fig. 5). molecules, like CO and CN− will have
The electric field for the p-polarized light practically only two bonding possibilities
is almost the same as the s-polarized light and the surface selection rule is not useful
at ca. 2.5 µm for the metal/air interface to determine the kind of bonding. How-
for 60◦ of incidence angle at 2000 cm−1 . ever, for more complex species, the surface
Therefore, the p-polarized light interacts selection rule is of great help.
with the dipoles at the surface and away One interesting example is the applica-
from it. Since in order to interact with tion of the surface selection rule to assign
the electric field the dipole must present the kind of bonding to the surface of the
the same direction, it turns out that only nitrate ion (NO3 − ). This ion when isolated
the species with a nonzero dipole moment presents a D3h symmetry (see Fig. 7).
perpendicular to the surface will interact In principle, the NO3 − ions could
with the infrared beam. This is the surface adsorb with the plane of the molecule
selection rule and is an important tool in either parallel or perpendicular to the
reflection-absorption spectroscopy applied surface. For a parallel adsorption, the
to electrochemical systems in order to dis- surface selection rule will allow only
tinguish adsorbed species from species in the out-of-plane bending modes, since
solution and to assign specific adsorption the stretching modes present always the
geometry for the adsorbed species. displacement parallel to the molecular
In Fig. 6, the s- and p-polarized spec- plane. The perpendicular adsorption has
tra for adsorbed sulfate species on two possibilities, that is, the nitrate
Au(111) [32] are presented to illustrate the ions can be adsorbed by coordination of
use of the surface selection rule to differ- either one or two oxygen to the surface.
entiate adsorbed species from species in Experimentally, the stretching mode of the
solution. In these spectra, the up band at adsorbed nitrate ions [33, 34] is detected.
1100 cm−1 is visible in both spectra and Therefore it can be concluded that the
this band is assigned to the dissolved sul- ion is adsorbed with the molecular plane
fate. The down bands are seen only in the perpendicular to the surface, giving rise to
p-polarized spectrum, since the s-polarized a C2v symmetry (see Fig. 8). The breakup
light cannot sample the surface. of symmetry occurs because the oxygen
Another application of the surface se- in contact with the surface is no longer
lection rule is related to the fact that the equivalent to the noncoordinating oxygen.
p-polarized light has a specific direction The stretching vibrational modes of the
on the surface, that is, the electric field of adsorbed NO3 − are presented in Fig. 8,
the p-polarized light will be perpendicular in which the symmetry and direction
to the surface. Therefore, the assignment of displacement of each mode are in-
of the bonding of adsorbed species us- dicated according to the character table
ing the surface selection rule demands the for the C2v point group (see Table 1).
information on the directions of the differ- The symbols in the character table have
ent dipole displacements of the adsorbed the usual meaning [35]. Briefly, C is the
s-pol. light
1100 cm−1
R/R0 = 0.05%
p-pol. light
950 cm−1
1196
1500 1400 1300 1200 1100 1000 900 800

Wave number
[cm−1]
Fig. 6 s- and p-polarized light spectra for the adsorbed sulfate species on
Au(111) in 0.69 mol L−1 HF, 0.5 mol L−1 KF, and 10−2 mol L−1 K2 SO4
solution. Ref. Pd/H2 . The s-polarized spectrum is extracted. (From Ref. [32]
with permission.)
symmetry axis, the subscript 2 stands in the indicated direction and stands for
for an order two symmetry, and v is the infrared activity. αij are the compo-
for vertical. σ is the reflection symme- nents of the polarizability and accounts
try plane, I is the identity operation, T for the Raman activity. A and B are the
is the translator vector of the molecule nondegenerated species (one-dimensional
in the indicated direction (x, y of z), and representation). A represents a symmet-
R is the rotation vector of the molecule ric species with respect to the rotation
C3
sv
sv C2
sv
sh
O
N
C2
C2
Fig. 7 Sketch of the D3h symmetry elements for the free nitrate ion.
Fig. 8 Proposed N−O

vibrations for the different
modes of the adsorbed NO3 − O O
ions under a C2v symmetry.
N N
O O O O
A1 A1
N
O O
B1
about the symmetry axis (represented by represents the antisymmetric species with
the character +1). The meaning of A is respect to the rotation about the sym-
that upon the rotation operation the vi- metry axis (represented by the character
brational mode remains unchanged. B −1), meaning that upon the symmetry
Tab. 1 Character table for the C2v point group
C2v I C2 (z) σ (xz) σ (yz)
A1 +1 +1 +1 +1 Tz αxx , αyy , αzz

A2 +1 +1 −1 −1 Rz αxy
B1 +1 −1 +1 −1 Tx , Ry αxz
B2 +1 −1 −1 +1 Ty , Rx αyz
operation the vibrational mode has the adsorbed forming a C2v symmetry and the
direction inverted. 1024 cm−1 is the other A1 mode, corre-
The character table allows identifying sponding to the activation of the totally
the active infrared modes. These modes symmetric NO3 stretching vibration [34].
are those with nonzero translation repre- This 1436 cm−1 mode would be the ν3
sented by T in the character table. For the stretching vibration and the 1024 cm−1
C2v point group, the A and B modes are would be the ν1 stretching vibration de-
active in the infrared spectroscopy. Adding picted in Fig. 8.
to this analysis the surface selection rule, In the case of in situ infrared in elec-
only those modes with translation in the di- trochemical systems, strong limitations
rection perpendicular to the surface will be in the spectral range, due to the use of
active. Considering that the surface is in- chemically inert windows, impede the ob-
cluded in the xy plane, only the A1 modes servation of modes usually in the range
can be active, since only the A1 have trans- of stretching vibrations for the inorganic
lation in the z direction. The A1 mode is molecules. Organic molecules usually dis-
the symmetric species with respect to the play also bending modes in the same
rotation about the symmetry axis, as de- spectral range, but the kind of analysis
scribed above. Therefore, for the adsorbed is practically the same.
NO3 − ion, only the symmetric modes will
be active in the infrared spectroscopy.
A detailed assignment of vibration 3.4.2.2 Experimental Setup
to a given molecule requires careful The experimental setup for in situ infrared
calculations, but for practical purposes, in for electrochemical studies requires the
the case of adsorbed species, it is possible construction of an electrochemical cell
to use the character table to assign the equipped with a window transparent to the
vibrations observed, once the vibrational infrared radiation and stable in aqueous
modes of the free molecule (or ion) are electrolytes as used for the electrochemical
known. In Fig. 8, the possible allowed studies. The window can be either flat
vibrations are proposed for the adsorbed or prismatic and is usually placed at the
NO3 − ion. bottom of the cell. The prismatic window
The in situ reflection-absorption spec- allows a more favorable angle of incidence.
trum of nitrate ions on a Au(100) electrode The most commonly used window
is shown in Fig. 9, in which two bands at materials are CaF2 (900–77 000 cm−1 )
1436 and 1024 cm−1 can be observed. The and ZnSe (500–20 000 cm−1 ). The BaF2
1436 cm−1 band has been assigned to the (770–66 000 cm−1 ) can be used only
A1 symmetric mode for the nitrate ions in some solutions, since barium forms
1.2 V
1.1 V
1V
1024
R/R0 = 0.025%
1436
1600 1400 1200 1000

Wave number
[cm−1]
Fig. 9 In situ FTIR p-polarized light spectra of adsorbed NO3 − ions on
Au(100) in 1 mol L−1 and 0.05 mol L−1 HNO3 . Ref. Pd/H2 . (From Ref. [33]
with permission.)
insoluble salts with sulfate, for example, measurements is shown in Fig. 10. In
and interferes with the spectral data. With the upper part is the entrance for the
prismatic windows with angles of ca. 65◦ , electrodes and gases. This kind of cell is
it is possible to obtain an incidence angle to be placed vertically, which makes the
up to 77◦ . For a flat window, even for an optical setup slightly more complicated,
incidence angle of 60◦ at the window/air but with clear advantages over the cell
interface, the maximum angle of incidence placed horizontally, since in the vertical
attainable is of ca. 47◦ [36]. cell solution leaks are rare, while in the
A sketch of a typical electrochemical horizontal cell the leaks are more common.
cell for in situ infrared spectroscopic Placing the cell vertically allows also
Screw for Connection

fixation of the electrode
Guide for the electrode
PTFE support
Gas entrance
Position for voltammetry
Reference electrode
Auxiliary electrode
Position for the in situ

Window FTIR
support
PTFE ring
Window
Fig. 10 Sketch of an electrochemical cell used for in situ measurements specially
designed to operate vertically.
the study of organic reactions, in which for the setup using a prismatic and flat
replacement of solution is needed in some window, respectively [38]. The simulation
cases. Other cells with distinct design can shows a dramatic decrease in the reflec-
be seen in the literature [35–37]. tivity when a flat window is used. The
The cell contains the three electrodes differences in s- and p-reflectance are not
as any usual electrochemical cell for the as big as in the case of a prismatic window.
control of potential. The working electrode Other important effects are the interfer-
is a reflecting surface and for this it must ence fringes for the flat window, which
be polished to a mirror finish. are more pronounced for the s-polarized
An important detail is the difference light. The fringes are caused mainly by the
between the reflectivity using a flat win- flat window and not by the thin electrolyte
dow or a prismatic window for the in layer, since the wavelength of the infrared
situ spectra. Simulating a change in ab- radiation is larger than the thickness of the
sorbance in the thin-film layer, that is, a thin electrolyte film. Finally, it is important
change in the solution composition, data to stress that the reflectance is almost inde-
of reflectivity are shown in Fig. 11(a, b) pendent of the incidence angle up to 40◦ .
There is a maximum for 70◦ of incidence angle between 40◦ and 90◦ and there is a
angle at the window. When a prismatic pronounced maximum around 70◦ .
window is used, the radiation reflectivity The consequence of this behavior is that
is strongly dependent on the incidence when a prismatic window is used, the
1.02
1.00
0.98
0.96
∆R/R
0.94
0.92
0.90
0.88
0 20 40 60 80
Angle
(a) [degree]
Fig. 11 Simulated change in reflectivity for the system: (a) flat window/electrolyte/metal for a
change in the absorption in the electrolyte layer (note the interference fringes caused by the
internal reflections) and (b) prismatic window/electrolyte/metal. The parameters used in the
simulation are: n1 = 1.0 (air); n2 = 1.4 (CaF2 window); n3 = 1.32 + i0.015 (electrolyte); and
n4 = 10.9 + 15.6 (Pt). Electrolyte film thickness: 2.0 × 10−6 m; flat window thickness:
2.0 × 10−3 m. The change in absorbance in the electrolyte has been simulated by changing k
by 0.01.
1.0
0.9
0.8
R /R 0
0.7
0.6
0.5
0.4
0 20 40 60 80
Angle
(b) [degree]
Fig. 11 (Continued)
large difference in reflectivity between s- advantage for the in situ studies, since
and p-polarized light makes it difficult to the overall signal-to-noise ratio will be very
distinguish between adsorbed species and high compared with that obtained with a
species in solution, since the s-polarized flat window.
signal will be much lower than the p-
polarized signal. In summary, even if 3.4.2.3 Transport Effects in the Thin
the s-polarized light spectral signal is too Electrolyte Film
weak, this does not mean that the spectral In one typical in situ FTIR experiment, it is
feature is only due to adsorbed species. necessary to measure a reference spectrum
On the other hand, the use of a prismatic at a potential in which the electrochemical
window will maximize the spectral signal process does not take place and to measure
from the surface and it will be of great a sample spectrum in which the desired
process takes place. The two spectra are the observation of the electrochemical
ratioed to obtain the final spectrum as process in study. The pH change effect
in Eq. (1). This kind of experiment was was observed incidentally by Christensen
originally called Subtractively Normalized et al. [39] during CO2 reduction in a phos-
Interfacial Fourier Transform-Infrared Spec- phate buffer solution. The phosphate ions
troscopy (SNIFTIRS). equilibria are very well-known and are eas-
In the process, of changing the potential ily observed in the in situ FTIR spectra.
to acquire the two spectra, the components Corrigan and Weaver [15] demonstrated
of the solution contained in the thin that the double-layer charging during the
electrolyte film between the electrode and potential step used is already enough
the window can diffuse either to or from to produce ion migration inside the so-
the thin layer. The diffusion inside the lution film as shown for the study of
thin electrolyte layer is relatively slow [5]. N3 − ions adsorption on a polycrystalline
Therefore the compensation of species that gold electrode.
are consumed by reaction or adsorption One example of ion equilibrium dis-
is also slow. This is the reason that location in a phosphate buffer solution
in SNIFTIRS spectra usually there are caused by pH changes [40] is shown in
bands in both up and down direction Fig. 12. The spectra were obtained us-
with respect to the baseline. Taking R0 ing a potential step between −0.25 V
as the reference spectrum and R as the versus a reversible hydrogen electrode
sample spectrum, down bands (R0 > R) (RHE) and the potentials indicated in
correspond to the formation of species and the figure. The spectra were taken with
up bands (R0 < R) to the consumption of s- and p-polarized radiation to confirm
new species (see Eq. 1). that the changes in the spectra occur
These effects are particularly dramatic in the solution and not at the elec-
for platinum electrodes. Usually, a poten- trode surface.
tial of 0.05 V in acidic solutions is used The up bands are then those corre-
as reference both in studies of ion ad- sponding to the consumption of species
sorption and organic molecules reaction. after the potential step and the down
At this potential, the platinum electrode bands are those corresponding to the
is covered with a monolayer of hydro- species formed after the potential step.
gen. Changing the potential to higher The bands are the 1162 and 1075 cm−1 ,
values will cause hydrogen desorption leav- corresponding to the antisymmetric and
ing H+ ions in the thin layer causing symmetric modes of the PO2 group, and
pH changes. the band at 942 cm−1 , corresponding to
Different groups [14–16] have addressed the symmetric stretching of the P(OH)2
this problem in detail. It has been shown group of the H2 PO4 − . On the other hand,
that two different kinds of problems can the up bands at 1091 and 991 cm−1 are
occur. Both double-layer charging and assigned to the doubly degenerated stretch-
consumption of species can cause migra- ing mode of the PO3 and the P−OH
tion or diffusion of species into or from stretching vibration, respectively. Actually
the thin solution layer, or pH changes the HPO4 2− species present the dou-
can shift acid-base equilibria that can be bly degenerated mode at 1072 cm−1 , but
sampled by the in situ FTIR spectra. the down band at 1075 cm−1 are over-
In both cases, these effects can mask lapped and disturb the band at 1072
s-pol. light p-pol. light
−50
400
400
[mV]
[mV]
E
E
600 600
R/R0 0.2% R/R0 0.5%
1000
1000
1200 1100 1000 1200 1100 1000

Wave number
[cm−1]
Fig. 12 In situ FTIR spectra of the changes in Up bands at 1162 cm−1 , 1018 cm−1 , and
phosphate protonation in solution for different 942 cm−1 are assigned to H2 PO4 − and the
applied potential step between −0.25 V and the down bands at 1091 cm−1 and 991 cm−1 to
potentials indicated on a platinum electrode in a HPO4 2− . (Data reproduced from Ref. [41] with
phosphate buffer pH = 7.4; ionic strength 0.2. permission.)
giving a peak at 1091 cm−1 . In summary, these effects, an appropriate supporting

from the spectral signals observed in the electrolyte must be used to compensate
spectrum of Fig. 12, a phosphate buffer the acid-base equilibrium changes. Other
equilibrium has been sampled showing more complicated approaches, like a flow
the reaction: cell with an electrode with a small hole in
the center to allow electrolyte flow, have
HPO4 2− + H3 O+ ⇐ ⇒ H2 PO4 − + H2 O been proposed [37]. The other version is a
(15) window with a hole that allows electrolyte
This is an example of how pH changes flow [39]. However, the operation of such
caused by H+ desorption during the electrodes or windows is not easy and
potential step used for the measurement the fluctuations in the electrode-window
cause equilibrium changes in solution, distance preclude the compensation of
and any possible band due to adsorbed the full spectrum of the optical path. So
phosphate is completely buried by the far, the use of a supporting electrolyte
solution spectral signals. To minimize permitted the distinction of the solution
bands from the bands arising from the The simplest anion studied is the
adsorbates [41]. CN− . The most recent studies on CN−
adsorption have been done by Tadjeddine
3.4.3 and coworkers [44] using sum frequency
Infrared Spectroscopy under Equilibrium generation, SFG a spectroscopic technique
Conditions. Study of Adsorbed Sulfate Ions based on nonlinear optics. This technique
on Single-crystal Electrode Surfaces is surface specific, since it is based on the
property that when two laser beams meet
In this section, some selected applications
at a nonsymmetric place, the frequency
of in situ infrared spectroscopy to study
of the two beams is summed. Then,
the reversible adsorption of anions on the
with a laser in the visible region and
electrode surface will be discussed. The
an infrared tunable laser, the vibrational
relevant contribution of the in situ FTIR
bands can be sampled. According to
spectroscopy to this field was to permit
the SFG measurements for the CN−
the study of the structure of the adsorbed
adsorption, two bands are observed and
anions in the double-layer of the elec- have been assigned to the C- and N-down
trochemical interface for solid electrodes cyanide ion adsorbed on the gold surface.
without a very well-defined double-layer Although many interesting studies on the
region, as in the case of platinum elec- CN− and other anions adsorption can be
trodes. Moreover, the identification of the found, in this chapter only the adsorption
nature of adsorbed species using in situ in- of sulfate ions will be addressed in detail
frared spectroscopy has been the object of to show the potentialities of in situ FTIR
intense debate and relevant contributions in the study of electrochemical systems.
have been given. The importance of sulfate ions is related
to the fact that it is a largely used
3.4.3.1 The Adsorption of Anions (The anion in supporting electrolytes both for
Electrical Double Layer) fundamental studies and applications.
The use of in situ FTIR spectroscopy to
study ions adsorption is related to the need 3.4.3.1.1 The Adsorption of Sulfate on
of experimental tools to access the double- M(111) Single-crystal Electrodes One co-
layer structure with molecular specificity. mmon feature of adsorbed sulfate ions on
This is especially true for the solid metals, the M(111) faces of different fcc metals is
where the study of the double layer using that the vibrational features observed in
the classical capacitance studies is not easy. the in situ FTIR spectra do not depend on
Spectroscopy combined with in situ scan- the degree of dissociation of the species in
ning tunneling microscopy (STM) [42, 43] solution. When sulfate ions are adsorbed
constitutes a very powerful approach, al- from strongly acid solutions containing
lowing the determination of the molecular mainly bisulfate ions or from mildly acid
identity and organization of the double- solutions containing mainly sulfate ions,
layer components. However, it is not the the vibrational feature observed is basically
purpose of this chapter to review in de- the same (see Fig. 13) [42, 45–47]. In the
tail the double-layer structure, but rather, spectra of Fig. 13, it is possible to see the
to show how in situ infrared spectroscopy up band at 1100 cm−1 for the solution with
can contribute to the understanding of the pH = 2.8. This band is the F mode of the
double layer at molecular specificity. free sulfate ions in solution. The absence of
262
Pt(111) 1100
Au(111)
1100
(ii) 1100
Rh(111)
3 New Experimental Evidences
(ii)
(i)
1050
(i)
1254 952 960
1050
1185 1150
(i)
(ii)
R/R0 = 0.1% 1240
R/R0 = 0.01% R/R0 = 0.05%

1192 1247
1272
1350 1250 1150 1050 950 1300 1200 1100 1000 900 1400 1300 1200 1100 1000
Wave number Wave number Wave number
(a) [cm−1] (b) [cm−1] (c) [cm−1]
Fig. 13 In situ FTIR spectra of adsorbed sulfate species on (a) Pt(111). (Data extracted from Ref. [45] with permission.) (b) Au(111). (Data extracted
from Ref. [32] with permission.) (c) Rh(111) in (i) 0.69 mol L−1 HF, 0.5 mol L−1 KF, and 10−2 mol L−1 K2 SO4 and (ii) 7.3 mol L−1 HF and 10−2 mol L−1
K2 SO4 . (Data extracted from Ref. [46] with permission.)
up bands at 950 cm−1 reveals that sulfate neutral SO4 2− solutions [42] (see Fig. 14).
ions are unperturbed when dissolved, re- On the Au(111) electrode, a weak band
taining the Td symmetry. The up band, at 958 cm−1 , corresponding to the totally
as discussed above (see Sect. 3.4.2.3), is symmetric vibration has been observed
due to the consumption of sulfate ions (see Figs. 13(b) and 14). The activation of
from the thin electrolyte film upon adsorp- this band reveals the breakup of symmetry
tion. For the same spectrum (pH = 2.8), a caused by the anion adsorption. This band
down band can be observed between 1200 has been assigned either to the activation
and 1250 cm−1 for saturation coverage for of the totally symmetric mode of the SO4 2−
the three metals, corresponding to the ad- ion due to the breakup of symmetry or to
sorbed ions. This band was not observed the S−OH stretching vibration of adsorbed
with s-polarized light, which confirms it bisulfate [48]. Measuring the vibrational
is due to the adsorbed sulfate species (see features of the adsorbed ion from a deuter-
Fig. 6 for the Au(111) case). Comparing the ated solution (D2 SO4 in D2 O), Shingaya
spectra obtained from the different solu- and Ito observed two bands, depending
tions, it is clear that this feature is the same, on the applied potential. For potentials
independent of the species present in solu- between 0.7 and 0.8 V, only one band at
tion, indicating that either bisulfate is dis- 1153 cm−1 is found. At 0.9 V, this band
sociated when adsorbed, or sulfate is proto- shifts to 1170 cm−1 and a new band at ca.
nated. Similar results are observed by other 1200 cm−1 appears with a very weak inten-
authors for Au(111) electrodes even for sity. The 1170 cm−1 band looses intensity
4 × 10−4 a.u.
1193
958
1187
1103
Fig. 14 In situ FTIR spectra of

adsorbed sulfate species on
Au(111) in 0.5 mol L−1 H2 SO4
1200 900
(lower) and 0.1 mol L−1
Na2 SO4 (upper). (Reproduced
n
from Ref. [42] with permission.) [cm−1]
for more positive potentials and the band asymmetric mode, since this would have
at 1200 cm−1 starts to increase intensity a dipole parallel to the surface, but when
and shifts strongly with the potential up to sulfate is adsorbed a C3v symmetry could
1230 cm−1 for 1.5 V. produce a second symmetric mode that
An important observation is the intensity can account for this band. The main
decrease of the low-energy band followed argument used is that calculations (no
by the increase of the higher-energy band. detail has been given) show that the
There is also a concurrent change in SO3 symmetric mode couples with the
the intensity of the 950 cm−1 band. The S−OH bending, but not with the S−OD
authors [48] attributed this behavior to a and this would be the reason for the
mixture of sulfate and bisulfate on the band splits at a certain potential. The
surface in a given range of potential. main problem in this interpretation is
The band at 1153 cm−1 was assigned to the 956 cm−1 band (compare Figs. 14
the asymmetric stretching of the sulfate and 15) that has been observed also for
ions, while the 1220 cm−1 band is due nondeuterated solutions. It is therefore
to the bisulfate. The 950 cm−1 band was difficult to explain the absence of isotopic
assigned to the S−OD stretching. Probably shift upon changing H for D in the
the 1153 cm−1 band is not due to an S−OH stretching.
E
vs SHE
[mV]
1500
1234 0.5 M D2SO4 in D2O

1300
956 0.5 M H2SO4 in H2O

1100 1250
∆R/R
900
1225
Wave numbers
800
1170
[cm−1]
1200
700
957 1175
0.001
1153
1150
1400 1300 1200 1100 1000 900 600 800 1000 1200 1400 1600
Wave numbers E vs SHE
(a) [cm−1] (b) [mV]
Fig. 15 (a) In situ FTIR spectra of adsorbed sulfate species on Au(111) in 0.5 mol L−1 D2 SO4 in D2 O
and (b) band center energy as a function of the applied potential. (From Ref. [48] with permission.)
More detailed results from the calcula- SO3 to a Pt(111) surface to compare it with
tions are necessary to account for these in situ spectra [49]. They concluded that
differences. The controversy on the nature depending on the amount of water, the two
of the adsorbed sulfate species (sulfate or species can be identified. Bands at 1230
bisulfate) can be attributed to the limita- and 953 cm−1 were (arbitrarily) assigned
tions of the in situ FTIR spectroscopy in to the bisulfate. This species would be
this case. Associated with the difficulty in converted into sulfuric acid (bands at
identifying the nature of the adsorbed an- 1276, 1043, and 941 cm−1 ) when water
ion (sulfate or bisulfate) is the difficulty evaporates. On the basis of these results,
in determining the symmetry of the ad- they concluded that there are two adsorbed
sorbed species, which would give the clue species on Pt(111) electrodes, depending
in determining the coordination to the sur- on the applied potential. Bisulfate species
face. The problem of the adsorbed sulfate are adsorbed for potentials between 0.4
is that only the symmetric modes will be and 0.75 V (identified by the band at
active, according to the surface selection 1230 cm−1 ) and sulfuric acid for potentials
rule (see Sect. 3.4.2.1.2). It is assumed that between 0.75 and 1.10 V (identified by
the Td symmetry is broken upon adsorp- the bands at 1276 cm−1 ). Although this
tion, since the oxygen atoms are no longer is an interesting approach to identify the
equivalent. The final symmetry will de-
fingerprints of adsorbed sulfate species, a
pend on the coordination to the surface.
definitive proof for bisulfate and sulfuric
Coordinating one or three oxygen to the
acid could be easily obtained in UHV by
surface will lead to a C3v symmetry, while
using deuterated water to confirm the
coordinating two oxygen will lead to a C2v
presence of sulfuric acid. A mixture of
symmetry. The C2v symmetry has no de-
different species at a given potential or
generated modes, while the C3v has two
surface condition cannot be ruled out
degenerated modes, one for the stretching
and this is a further complication to the
modes. Therefore, for the free species in
solution, the C2v symmetry must present interpretation of in situ spectra.
one more stretching mode than the C3v . Other explanations for the adsorbed
Because of the fact that the surface selec- species have also been given. Faguy and
tion rule limits the active modes to those coworkers [50] suggested that the adsorbed
having the dynamic dipole moment per- species forms ion pairs with the hydro-
pendicular to the surface, two different nium ions close to the surface, and they
coordinations will present the same num- even suggested that the ion is coordi-
ber of bands. Considering only the number nated by three protonated oxygen [51]. This
of bands, it is not possible to identify the model tries to rationalize the discrepancy
symmetry of the adsorbed sulfate species; in the assignment of vibrational bands of
and consequently, it is not possible to as- the in situ FTIR spectra to sulfate or bisul-
sign unequivocally either the coordination fate. Actually, the close proximity of the
or the nature of the adsorbed species. coadsorbed hydronium ions can mimic
Other strategies have been used to clear the bisulfate situation and this could be
these points. also an explanation of the Shingaya and
Shingaya and coworkers have simulated Ito results. A change in the hydronium ion
the electrochemical environment in ultra concentration with potential may happen,
high vacuum (UHV) by adding water and producing a different number of hydrogen
bonding between the sulfate ions and the interactions of the adsorbed ions, since the
coadsorbed water. ion–ion distance can be measured. How-
Other strategy to identify the coordina- ever, there are evidences that hydronium
tion and therefore to deduce the symmetry ions or water molecules are adsorbed √ be-
of adsorbed sulfate was presented by tween two neighboring ions in the 7
Lennartz and coworkers [43], using simul- direction. This obviously can break the
taneously in situ infrared spectroscopy and chemical interaction, but probably not the
in situ STM. In this study, the authors used dipole–dipole coupling since water and
in situ STM to obtain the registry or the sulfate frequencies are well separated.
√ The
sulfate adlayer to the Cu(111) substrate by two ions adsorbed in the 3 direction
changing the bias current and obtaining an do not have water molecules in between
image containing the contrast due to both and therefore chemical interactions can be
the adsorbed sulfate and the underlying operative as well.
Cu(111). Using Fourier Transform filtra- Although the final symmetry cannot
tion, they were able to separate both signals be deduced only from the spectroscopic
and, since these were obtained from the data, it is important to stress that the
same experiment, the overlap of the two spectroscopic data supply information on
images gave the registry between the ad- the strength of the chemical bonding [48]
sorbate and the underlying surface. They and on the effect of the applied electric
observed for Cu(111) a bridging type ad- field [55, 56]. Moreover, in situ STM
sorption of the sulfate using two oxygen data have been reported only for a full
resulting in a C2v symmetry. Knowing the monolayer, but not for submonolayer
coordination, it is easier to obtain the sym- coverage. The in situ FTIR is sensitive
metry and therefore the identity of the also in the submonolayer regime. By
adsorbed species. For example, two-fold appropriate calibration (through the full
coordinated adsorbed bisulfate will have a monolayer as determined by in situ STM),
Cs symmetry. Under this symmetry, the it is possible to determine the adsorption
noncoordinated S−O and S−OH will pro- isotherm, even for metals for which it is
duce two vibrations, other than the S−O2 very difficult to obtain capacitance data.
symmetric vibrations, with the component
in the direction perpendicular to the sur- 3.4.3.1.2 The Sulfate Ion Adsorbed on the
face. In the particular case of sulfate ions M(100) and (110) Single-crystal Electrodes
adsorbed on Cu(111), it is clear that the The adsorbed sulfate species on other
adsorbed ion is the dissociated sulfate and low-index planes of platinum and gold
not the bisulfate ions, since only the strong present a more complex spectroscopic
symmetric band at 1205 to 1220 cm−1 has structure [32, 46, 50, 57, 58]. Fig. 16
been observed [43]. shows the spectra of adsorbed sulfate at
Another important result from in situ saturation coverage for the three low-index
STM imaging of adsorbed sulfate on planes of the platinum electrode in a
M(111), single-crystal metal electrodes is solution of pH = 2.8. Clearly, a second
the observation of the adsorbate structure. band at 1008 cm−1 can be easily identified
It is found for Au(111) [42], Pt(111) [52], for Pt(100) (weak) [57] and for Pt(110)
Rh(111)
√ [53], Ir(111) [54], and Cu(111) [43]
√ (strong) [58]. These two planes do not
a 3 × 7 structure. This kind of infor- match a C3v symmetry for the adsorbed
mation can be used in determining lateral sulfate ions, and therefore we should
Pt(110)
1106
1234
Pt(100)
1108
1208 1100
Pt(111)
R/R0 = 0.1%
1254
1300 1200 1100 1000

Wave number
[cm−1]
Fig. 16 In situ FTIR spectra for the adsorbed sulfate species
on platinum single-crystal electrodes in 0.69 mol L−1 HF,
0.5 mol L−1 KF, and 10−2 mol L−1 K2 SO4 . Ref. Pd/H2 . (Data
extracted from Ref. [45] (Pt(111)), Ref. [57] (Pt(100)), and
Ref. [58] (Pt(110)) with permission.)
expect a C2v symmetry if undissociated band indicates that either the adsorbed
bisulfate ions are adsorbed. However, a C2v sulfate ion is tilted on the surface or that
symmetry has only two allowed bands, that a second species is also adsorbed. A tilted
is, the symmetric modes with respect to the adsorbed sulfate through a single oxygen
symmetry axis (the A1 modes). A second will permit other vibrations to be active,
since the symmetry is no longer the C2v when bisulfate is the major species in
and practically all modes will have some solution, a band at ca. 1100 cm−1 is
component of the dynamic dipole moment observed [59]. A similar situation has been
perpendicular to the surface. observed for adsorbed sulfate on Au(100)
The other possibility is that on Pt(100) as shown in Fig. 17, in which two strong
and Pt(110), sulfate and bisulfate are bands are observed at 1165 and 1091 cm−1
coadsorbed on the surface. Comparing for an applied potential of 0.8 V. The
the bands for sulfate species adsorbed 1165 cm−1 band shifts to 1185 cm−1 ,
on polycrystalline platinum electrodes in but the 1091 cm−1 band is practically
strongly acid solutions, it is clear that insensitive to the applied potential. A weak
R/R0 = 0.033%
1.2 V
956
1091
1.0 V 1181
0.8 V
1165 1091
0.7 V
1300 1200 1100 1000 900

Wave number
[cm−1]
Fig. 17 In situ FTIR spectra for the adsorbed sulfate species on Au(100)
single-crystal electrode in 7.3 mol L−1 HF and 10−2 mol L−1 K2 SO4 .
Ref. Pd/H2 . (Data reproduced from Ref. [34] with permission.)
band at 956 cm−1 can be viewed in some barriers for surface reactions to start on
spectra, which has been observed also the electrode surfaces.
for the sulfate adsorbed on Au(111). The
presence of two strong bands suggests 3.4.4
that probably at this solution pH, there Infrared Spectroscopy under
is a mixture of sulfate and bisulfate on Nonequilibrium Conditions. Study of
the surface. Indeed, the sulfate species Systems Relevant in Electrocatalysis
adsorbed on Au(100) surfaces in solutions
in which the undissociated sulfate is the One of the major successes of in situ
major species, reveals only one strong FTIR spectroscopy concerns the study
band at 1161 to 1190 cm−1 , indicating that of complex electrode reactions. In this
when only sulfate is present in solution, chapter, we present typical applications
the 1091 cm−1 band is absent for the related to research in electrocatalysis of
adsorbed species. alcohol oxidation. With the aim to study
Interestingly the relative intensities of electrocatalysis of reactions involving can-
the 1165 to 1185 cm−1 and 1091 cm−1 didate substances for fuel cells applica-
features change with the applied potential. tions, aliphatic alcohols of low molecular
The 1165 to 1185 cm−1 band is stronger weight have been intensively investigated.
than the 1091 cm−1 at more positive In particular, remarkable progress in this
applied potential, which could suggest direction was achieved in the last years in
that bisulfate can be further dissociated, the case of methanol oxidation. In fact,
since the applied potential induces a OH the interest in establishing the nature of
adsorption and therefore the acidity on the the adsorbed species of methanol was the
surface is changed favoring the bisulfate motivation of one of the first papers using
dissociation. in situ infrared spectroscopy [60]. Because
Summarizing, the adsorption of sulfate of the fact that adsorbed carbon monox-
is of major interest in electrochemistry, ide is typically formed when small organic
since sulfuric acid solutions are largely molecules interact with platinum, we start
used as a supporting electrolyte for many this section presenting data on the electro-
applications, including kinetic studies. So, catalysis of CO oxidation.
the molecular structure of the double layer As explained in Sects. 3.4.2.1 and 3.4.2.2,
is an important piece of information in application of the external reflectance tech-
order to understand the effect of the dou- nique requires the use of a thin layer
ble layer on the reaction kinetics. The of solution between the working elec-
infrared spectroscopy is an interesting tool trode and the IR window. Because of this
to access this kind of data, providing in- experimental approach, there has been
formation on the nature of the species, some reticence to use the IR method
and with the help of in situ STM, the under conditions of current flow. Criti-
structure and coordination of the adsorbed cism was centered on IR-drop effects and
ions. It is likely that this structure will reactant-depletion problems in the solu-
not survive intact when faradaic reactions tion thin layer. However, although IR
take place on the electrode surface, but drop and reactant depletion indeed occur,
certainly the structure formed will affect these problems can be experimentally min-
the threshold of the reaction, since dif- imized and should not be overestimated.
ferent structures can account for different On one side, usually employed solutions
have a good conductivity and, for instance, monoxide as a model molecule in both
for a potential step of 100 mV within the electrochemical and UHV environments.
H-region of platinum, charge compensa- UHV data have been used to compare
tion via ion migration can occur within the behavior of adsorbates at the elec-
tens of a second [30]. On the other side, trochemical interface with that observed
since the Fourier transform method al- at metal/gas interfaces [23]. Also, spectro-
lows a fast rate of data acquisition, reactant electrochemical data on adsorbed CO are
depletion can be reduced to a negligible interesting in the study of the Stark effect,
extent by using a reasonable experiment which can produce changes in band cen-
design, minimizing polarization time at ter frequency [55, 61] and integrated band
each potential. Very valuable information intensity [56]. Spectroscopic studies of CO
can then be obtained under conditions of within this context have been reported and
current flow. discussed elsewhere [30, 41]. In electro-
catalysis, carbon monoxide can be consid-
3.4.4.1 Carbon Monoxide ered as an intermediate (or poison) during
A significant number of studies using in the oxidation of small organic molecules
situ FTIR spectroscopy have been devoted (methanol, formic acid, ethanol, among
to adsorbed carbon monoxide. This is not others). In the present section, we have
surprising, in view of the supporting ex- selected some studies concerning mainly
perimental and theoretical data on the CO this issue, using spectroscopic data to com-
system in UHV and the interest on carbon pare the CO behavior on different catalysts:
2075
Pt
2061
PtRu
2010
Ru
R /R0 = 0.005 Fig. 18In situ FTIRS of a saturated CO

monolayer adsorbed at 300 mV on Pt,
2250 2100 1950 1800 Ru, and PtRu (50 : 50) electrodes in
0.1 M HClO4 . Sample spectra measured
Wave number at 300 mV, reference spectra obtained at
[cm−1] 800 mV.
polycrystalline Pt, Ru, and (50 : 50) PtRu Potential dependence of the band frequency
alloy [62], and single-crystal electrodes as As the potential is stepped in the positive
Pt(111)Ru [63, 64] and (0001)Ru [65, 66]. direction, the band center for adsorbed
CO shifts to higher wave numbers for all
3.4.4.1.1 Adsorption and Oxidation of CO electrode materials (Fig. 21). The following
on Polycrystalline Substrates In situ FTIR characteristics in the ν − E relationship
spectra in Fig. 18 show spectra for ad- are noteworthy:
sorbed CO at polycrystalline Pt, PtRu,
• For pure platinum, the band center
and Ru, respectively. The surfaces were
exhibits first a linear dependence on
saturated with CO at 0.3 V in CO-saturated
potential with dν/dE = 28 cm−1 V−1 .
0.1 M HClO4 solutions. After elimination
As CO oxidation starts, νCO remains
of bulk CO, sample spectra were collected
approximately constant in the potential
at 0.3 V and corresponding reference
interval between 0.45 and 0.55 V and
spectra were measured at 0.8 V, a potential
then decreases as the oxidation to
at which CO is completely oxidized.
CO2 proceeds.
Only linear-bonded CO is observed on
• For pure ruthenium, the band center fre-
the three surfaces. The ranges of observed
quency shifts linearly up to a potential
frequencies are 2065 to 2075 cm−1 for Pt,
of 0.3 V with dν/dE = 52 cm−1 V−1 .
2055 to 2065 cm−1 for PtRu(50 : 50), and
Then a more pronounced shift (dν/dE
2000 to 2020 cm−1 for Ru.
77 cm−1 V−1 ) is observed up to 0.45 V.
Of major importance for the electrocatal-
The onset of oxidation can be extrapo-
ysis, is the comparison of CO2 production
lated to 0.4 V, as judged from the CO2
on the different materials as a function
band intensity.
of potential. For this purpose, series of
• For the PtRu alloy, the frequency shifts
spectra were taken using the three cata-
linearly with dν/dE = 32 cm−1 V−1 at
lysts in the potential range between 0.1
low potentials and, as oxidation begins,
and 0.6 V versus RHE [62], as shown in
the frequency drops immediately to
Fig. 19 for the PtRu (50 : 50) alloy. The
lower values.
left side of Fig. 19, showing the bands
for CO2 , was calculated against a refer- At constant (saturation) coverage, the
ence spectrum at 0.1 V. The increasing marked effect of potential in pure Ru
negative-going band represents the forma- (highest dν/dE value) can be related
tion of CO2 as the potential is stepped to a stronger backbonding effect. This
to higher values. The right side of the is also reflected in the lower frequency
figure was calculated taking as a refer- values of the COL band at this metal
ence a spectrum at 0.8 V, a potential at as compared to the other two. Figure 21
which CO is completely oxidized. Here, gives also information on the properties
the intensity of the negative-going bands of the adlayer as oxidation sets on. Thus,
is proportional to the CO coverage at the faster red shift (to lower frequencies)
the respective potentials. The integrated of the band center frequency for PtRu
band intensities for CO2 , at 2341 cm−1 beginning at the onset of CO oxidation
are plotted in Fig. 20. Accordingly, the indicates a rapid decrease of the lateral
catalytic activity for COad oxidation de- interactions, that is, oxidation of the
creases in the order PtRu(50 : 50) > CO adlayer probably occurs within a
Ru > Pt. relatively loose structure. Contrasting, the
0.60 V 0.55 V
0.50 V
0.55 V
0.45 V
CO(L)
0.50 V
0.40 V
0.45 V
0.30 V
CO2
0.20 V
0.40 V
0.10 V
0.35 V
∆R/R = 5 10−3
2400 2300 2200 2100 2000 1900

Wavenumber
[cm−1]
Fig. 19 In situ FTIR spectra for saturated CO adsorbates on a
PtRu(50 : 50) alloy electrode in 0.1 M HClO4 from 100 mV to
600 mV at 50 mV interval. CO was adsorbed at 300 mV. For each
spectrum, 100 interferograms were collected at 8 cm−1 resolution,
acquisition time ca. 44 s. Bands on the left side, corresponding to
CO2 produced during COads oxidation, were calculated against a
reference spectrum taken at 100 mV (CO2 was absent). Bands on
the right side, corresponding to adsorbed CO, were calculated
against a reference spectrum taken at 800 mV after complete
oxidation of CO. (From Ref. [62] with permission.)
red shift observed on the pure metals Ru(001) under UHV conditions [67] show
is somewhat slower indicating that CO a decrease of CO desorption energy from
oxidation probably occurs only at the Ed = 140 kJ mol−1 for the pure metal
border of islands, the local CO coverage to Ed = 107 kJ mol−1 in the presence of
remaining relatively high. coadsorbed oxygen (θO = 0.5)) [62, 67].
Two different slopes observed for pure Such a behavior is expected in the presence
Ru at potentials prior to CO oxidation have of electronegative coadsorbates, which
been related to the possible interaction cause a withdrawal of metal electrons
with adsorbed oxygen (Thermal desorption and diminish the metal-π* back donation.
data from coadsorption of CO and O at The consequences of this effect are
1.0
CO2
0.8
Integrated band intensity

0.6
[a.u.]
0.4
0.2 PtRu Ru Pt
0.0
0.2 0.4 0.6

Potential (RHE)
[V]
Fig. 20 Integrated band intensity for CO2 as a function of potential.
The intensities were calculated from series of spectra obtained at each
catalyst as shown in Fig. 20 for the PtRu alloy. (From Ref. [62] with
permission.)
a lowering of the metal-CO binding A comparison of the absorption co-

energy as observed in UHV [67] and a efficients for CO on different materials
stronger C−O bond, which is reflected requires a normalization, taking into ac-
in the increase of νCO shift between 0.35 count differences in electrode roughness.
and 0.45 V. With the aim of normalization, one can
use, again, the band intensity of CO2
The relative absorption coefficients of CO produced upon total oxidation of the
on Pt, Ru, and PtRu alloys Spectra for CO respective adlayer. The (COads )/(CO2 ) in-
submonolayers on the three materials are tensity ratios are 0.87 (for Pt), 0.60 (for
shown in Fig. 22. The degree of coverage PtRu), and 0.25 (for Ru) with an esti-
θCO indicated in the spectra was calcu- mated error of ca. 8% [62]. These nor-
lated using the ratio of the band intensity malized intensities of COad reflect, in a
for CO2 to the maximum CO2 intensity, first approximation, the relative values of
obtained after complete oxidation of a the absorption coefficient of CO on the three
CO-saturated layer. Figure 22 shows that substrates.
for comparable band intensities of CO2 , On the basis of the above discussion,
the corresponding intensity of the COad the lower intensity of the CO band at Ru
feature largely differs for the three elec- may really reflect a lower value for the
trodes, the largest signal being observed at dynamic dipole moment of the transition
Pt and the smallest one at Ru. for CO on this metal as compared to the
2085 Fig. 21 Dependence of the band center

frequency on potential for saturated CO
Pt adlayers on Pt, Ru, and PtRu(50 : 50)
alloy. The frequencies were taken from
2070 series of spectra measured at each
catalyst as shown in Fig. 20 for the PtRu
alloy. (From Ref. [62] with permission.)
2055
Wave number
PtRu
[cm−1]
2040
2025
Ru
2010
1995
0.0 0.2 0.4 0.6 0.8

E vs RHE
[V]
other materials. In a study of the effects of a screening effect caused by the electronic
coverage on the band intensity and band polarizability of the adsorbed molecules.
center of COad at Ru(001), H. Pfnür and In such a case, the reduced band inten-
coworkers [68] have suggested that strong sity for CO could be caused by strong
dipole–dipole coupling can reduce the dy- dipole–dipole coupling and/or large island
namic dipole moment. Furthermore, these domains; this would be an alternative (or
authors show that this effect markedly additional) explanation for the lower rate
depends on the size of the CO islands. In- of CO oxidation observed during oxidative
deed, the band intensity should be affected stripping at pure ruthenium as compared
by lateral interactions to a larger extent, to the alloy.
when adsorption takes place in the form The dependence of νCO on θCO for Pt,
of large islands, since in such a case PtRu, and Ru is shown in Fig. 23. A linear
most of CO molecules are surrounded by dependence of νCO values is observed
neighbors of an identical nature. Using for all three materials, the respective
a different theoretical approach, Persson dνCO /dθCO values being 58 cm−1 (for Pt),
and Rydberg [69] confirmed the predom- 64 cm−1 (for Ru), and 50 cm−1 (for PtRu).
inance of dipole–dipole coupling in the It can be stated that, from the spectroscopic
coverage effects observed by Pfnür and point of view, the PtRu (50 : 50) alloy
coworkers [68], and explained the reduc- approaches more closely the behavior of
tion of the band intensity as being due to pure Pt than that of pure Ru.
q = 0.12
q = 0.41
q = 0.43
q = 0.45
q = 0.69
q = 0.69 q = 0.75
q=1
q=1
q=1
R/R0 = 5 × 10−3 R/R0 = 5 × 10−3 R/R0 = 5 × 10−3
Pt PtRu (50:50) Ru
2400 2200 2000 2400 2200 2000 2400 2200 2000

Wave number
[cm−1]
Fig. 22 Comparison of in situ FTIR spectra for submonolayer CO adsorbates on Pt,
PtRu(50 : 50), and Ru electrodes in 0.1 M HClO4 . CO was adsorbed at 300 mV. Sample
spectra were acquired at 300 mV and computed against a reference spectrum taken at
800 mV; 100 interferograms were collected at 8 cm−1 resolution, acquisition time ca. 44 s.
The CO coverages θco indicated on each spectrum were calculated from the ratio of the
band intensity for CO2 at each coverage to the CO2 band obtained at saturation coverage.
(From Ref. [62] with permission.)
3.4.4.1.2 Adsorption and Oxidation of CO forming islands [70, 71]. The catalytic
on Single-crystal Substrates action is thus expected to occur at the
border of Ru domains.
CO adsorbed on Pt(111)/Ru Attempts In situ FTIR spectra for adsorbed CO
to study the influence of the surface on this material are shown in Fig. 24(a, b)
structure on the rate of oxidation of for two Ru coverages. At low Ru coverage
CO, methanol, and ethanol motivated the (θRu = 0.2), two bands related to CO
preparation of catalysts using adsorbed bonded to Pt atoms in on-top and bridge
Ru on well-defined Pt(111) surfaces [63]. positions are observed near 2058 cm−1
However, the resulting structures, when and 1816 cm−1 , respectively. A third band
investigated via STM, show that Ru at 1982 cm−1 corresponds to CO linearly
segregates on the surface of Pt(111) bonded to Ru atoms. This feature is better
2080
Pt
PtRu
2040
Wave number
Ru
[cm−1]
2000
1960
0.0 0.2 0.4 0.6 0.8 1.0

qCO
Fig. 23 Dependence of vCO on θCO for submonolayer CO adsorbates
on Pt, PtRu(50 : 50), and Ru electrodes in 0.1 M HClO4 at a constant
potential of 300 mV: CO adsorption at 300 mV. Other details as in
Fig. 23. (From Ref. [62] with permission.)
developed at the electrode with higher as a function of the electrode potential;

concentration of Ru, appearing at near pure polycrystalline Ru is included for
2008 cm−1 . This and the band for COL at Pt comparison. The Pt(111)/Ru samples
sites (2047 cm−1 ) are the only prominent exhibit values near to pure Pt(111) (at low
features observed. Ru concentration) or pure Ru (at high Ru
It is noteworthy that for both surface concentration).
coverages with Ru, Pt(111)/Ru exhibits Comparing the binary catalyst with the
features corresponding to CO frequencies pure metals, it can be stated that the pres-
near those of pure ruthenium and pure Pt ence of Ru shifts the CO−Pt feature to
as well. This is not the case for the PtRu lower wave numbers. Conversely, the pres-
(50 : 50) alloy, which exhibits only one band ence of Pt shifts the CO−Ru feature to
at frequencies close to those for pure Pt higher wave numbers. This observation
(Figs. 18 and 23). It can be thus stated that can be related to the degree of backbond-
opposite to the Pt(111)/Ru electrode, the ing. Since the work function of Pt(111)
behavior of the alloy surface is that of a (5.93 eV) [72] is substantially higher than
more homogeneous surface, with respect that of Ru (4.71 eV) [73], a significant elec-
to the distribution of individual Pt and tron transfer from Ru to neighboring Pt
Ru atoms. atoms is expected. Such an electronic
In Fig. 25, the band center frequencies effect should weaken the Ru−CO bond
for COL on different materials are plotted (because of decreased back donation) and
600 600
550 550
500 500
550
550
450 450
400
400
500 500
350
350
300 300
450
250
450 250
400
200
400 200
150
150
R/R0 = 0.01 1982 R/R0 = 0.01 1982
1816
2047
2058
qRu = 0.2 qRu = 0.75 2008
2400 2200 2000 1800 2400 2200 2000 1800

Wave number Wave number
(a) [cm−1] (b) [cm−1]
Fig. 24 In situ FTIR spectra for adsorbed CO and for the CO2 produced during the CO
stripping in 0.1 M HClO4 on Pt(111) modified with Ru. Ru coverage: (a) 0.2 mL and
(b) 0.75 mL. After the formation of the CO adlayer at 0.1 V, CO was eliminated from the
solution by N2 bubbling. The potential was then changed from 0.1 V onwards in 50 mV
steps. The sample potentials are indicated in the corresponding spectra. The CO2 region
(band at 2341 cm−1 ) was calculated with a reference spectrum taken at 0.1 V (a potential
in which CO2 is not formed). For the CO region, the reference spectrum was one taken at
0.8 V, that is, a potential in which COads was completely oxidized. (From Ref. [63] with
permission.)
strengthen the Pt−CO bond (increased during several minutes via the changes
back donation). Thus, the relative values in both CO and CO2 signals. The result
of the IR frequencies can be taken as a is plotted in Fig. 26. Surprisingly, the
measure of the CO−Metal bond strength. Pt(111)/Ru electrode containing 75% Ru
The rate of CO oxidation at 0.45 V on presents the highest rate of CO oxidative
different PtRu materials was monitored stripping. The result can be rationalized in
Pt(111)
2060
Wave number
2040 Pt(111)/Ru
[cm−1]
2020
Ru
2000
0.0 0.2 0.4 0.6

Potential vs RHE
[V]
Fig. 25 Comparison of the C−O stretch wave number for the COL on the
Pt(111), Ru-modified Pt (111), and pure polycrystalline Ru as a function
of potential. Data were derived from IR spectra as those in the figure.
Experimental conditions as in Fig. 25. (From Ref. [63] with permission.)
0.8
1.0 Pt(111)
Band intensity for COL
Band intensity for CO2
0.8 0.6
Pt(111)/Ru
[a.u.]
[a.u.]
0.6 Ru 0.4
0.4 PtRu alloy PtRu alloy

0.2
0.2 Ru
Pt(111)/Ru 0.0 Pt(111)
0.0
0 2 4 6 8 0 2 4 6 8
Time Time
(a) [min] (b) [min]
Fig. 26 Change of the IR band intensity for The respective surface was saturated with CO at
adsorbed: (a) CO and (b) for CO2 during CO 0.1 V after which, CO was eliminated from the
stripping at a constant potential of 0.45 V on electrolyte by nitrogen bubbling. (From Ref. [63]
Pt(111), Pt(111)/Ru with θRu = 0.75, with permission.)
PtRu(50 : 50) alloy, and pure polycrystalline Ru.
terms of a reaction mechanism as follows: and for the oxidation product, CO2 , is
plotted in Fig. 27(b). The initial values of
H2 O + * −−−→ OHad + H+ + e− the band intensity for COL and COH are
(16) constant and start decreasing concomitant
with the appearance of CO2 at ca. 200 mV
OHad + COad −−−→ 2* + CO2 + H+ + e−
versus Ag/AgCl. The intensity of the CO2
(17)
feature passes through a maximum at
Where * represents a free site on
ca. 600 mV. From this potential onwards,
the surface.
the signal decreases, in spite of the fact
On pure Pt, the first step is rate
that oxidation of COL continues up to a
determining, that is, the overall process
potential of 1100 mV. The latter process,
starts as soon as OH adsorption sets in.
however, is relatively slow and CO2
The oxidation of the CO adsorbed layer
diffuses out of the thin layer, causing the
is considered to initiate from sites that
observed decrease of the band intensity.
allow OH adsorption. On pure Ru, on
When analyzing the potential-dependent
the other hand, the second step seems
behavior of the band frequency, additional
to be rate determining. The adsorption of
information on the surface structure and
O-containing species has been reported
composition is useful. Thus, data from
already for potentials down to 0.2 V
Auger and low-energy electron diffraction
versus RHE [74], that is, well below the
Spectroscopy (LEEDS) experiments were
CO oxidation peak. This means that
performed on emersed Ru(0001) elec-
OH and CO can coexist on the surface
trodes [65]. In the range of potentials
within a certain potential region. The
between −80 mV and +200 mV versus
second step sets in only at a higher
Ag/AgCl, a (2 × 2) phase of oxygen exists
potential. In this model, the faster CO
on the surface [65]. Furthermore, oxygen
oxidation on the Pt(111)/Ru electrode can was observed up to an emersion poten-
be explained by a decrease in the bond tial of 1100 mV, though forming another
energy CO−Ru, reflected in the decrease phase, which was suggested to be (1 × 1).
of the backbonding effect. Four different potential regions are dis-
tinguished in Fig. 27(c) for the potential-
CO adsorbed on Ru(0001) In situ FTIRS dependent frequency of linear-bonded
was used to study CO adlayers on CO. The initial slope of the plot up to
Ru(0001) [65, 66]. Contrasting the behav- 175 mV, (dνCO /dE) = 37 cm−1 V−1 , was
ior at polycrystalline Ru, where only the interpreted as due to either a Stark effect or
band for linear-bonded CO is observed to electron back donation from the metal
(Fig. 27a), a saturated adlayer of CO ad- into the 2π ∗ orbital of CO [30]. As the po-
sorbed on Ru(0001) exhibits two different tential increases, between 200 and 425 mV
geometries. The spectra in Fig. 27(a) show versus Ag/AgCl, the slow oxidation of CO
two bands assigned to CO in linear commences and the slope (dνCO /dE) de-
and hollow position, COL and COH , creases slightly to 29 cm−1 V−1 . The third
respectively. This assignment is supported region, with (dνCO /dE) = 4 cm−1 V−1 , is
by STM data [66] showing this adsorption accompanied with a significant increase
geometry in a c(2 × 2)–2CO structure. of CO oxidation. At potentials between
The potential dependence of the band 625 mV and 1100 mV versus Ag/AgCl,
intensity for both forms of adsorbed CO the frequency changes again with a slope
1100 mV (a) Ru(0001) (b)

×5 16
1000 mV
Band intensity
800 mV
12 COL
[a.u.]
650 mV
550 mV 8
CO2
500 mV COH
4
450 mV
0
400 mV
2040
250 mV (c)
2030
200 mV
2020
150 mV
2010
COL
[cm−1]
0 mV
nCO
1800 (d)
−100 mV
1790
−200 mV
1780
COH
R/R0 = 1770
0.005 COL COH
1760
2100 2000 1900 1800 −0.2 0.0 0.2 0.4 0.6 0.8 1.0
Wave number E vs Ag/AgCl
[cm−1] [mV]
Fig. 27 (a) In situ FTIR spectra for CO adsorbed at Ru(0001) at different potentials (vs
Ag/AgCl/Cl− ) as indicated. CO was adsorbed at −100 mV during 5 min, then the solution was
replaced with pure base electrolyte (0.1 M HClO4 ). Spectra were normalized versus a
spectrum taken at 1100 mV, after waiting for 2 min. in order to completely oxidize CO.
(b) Integrated band intensities for COL , COH from spectra is shown in (a). CO2 band intensity
from the same experiments. (c), (d) Potential dependence of the band center frequency for
COL and COH , respectively, from spectra shown in (a). (From Ref. [65] with permission.)
similar to the initial value for the saturated At submonolayer coverage, a splitting of
layer. Observing that CO oxidation slowly the COL feature was reported [65], which
continues, it is obvious that the effect of indicates different CO-adlayer structures
potential on the band center frequency that can be tuned by potential. It has been
prevails over the lowering of the total CO suggested that coadsorbed O, forming dif-
coverage. That means that compact island ferent structures depending on potential,
domains of CO on the surface of Ru(0001) plays an important role in the compression
remain on the surface even at very high processes of the COL adlayer [65]. Con-
potentials [65]. firming this interpretation, no splitting
of the COL adlayer was observed for originate through a reaction of HCOOH
electrodes pretreated by annealing at with excess CH3 OH.). Other bands result
1200 K in a hydrogen stream [66]. Under from adsorbed species produced during
these conditions, the Ru oxide layer seems methanol adsorption. Thus, the bands at
to be completely eliminated. The band for ca. 2050 cm−1 and 1820 cm−1 are due,
COH , Fig. 27(d), shows several transitions respectively, to linear and bridge adsorbed
whose origin needs to be investigated. A se- CO [60].
ries of spectra with different CO coverages The spectra reveal that from all three
exhibits this band only for the saturated materials, the alloy shows up (a) the largest
adlayer [66]. production of CO2 and (b) the largest
intensities for linear-bonded CO at low
3.4.4.2 Methanol potentials. Thus, the gained information
The electrooxidation of methanol occupies, via IR spectroscopy is, not only the better
since many years, the center of interest performance of the alloy in relation to CO2
in electrocatalysis of fuel cell reactions. production but also its ability to adsorb
Some of the most important findings on methanol at lower potentials (observe
this issue are described in Chapter 5.2. the band intensity for CO at 0.15 mV).
We do not intend to review here all of Besides this, for both PtRu materials,
the spectroscopic work done on methanol the CO band reaches at 0.30 to 0.35 V
but to illustrate some highlights of the a stationary value and, in spite of a
FTIRS method when applied to study the strong production of CO2 at 0.5 V, the
electrocatalysis of methanol oxidation. band intensity for CO practically does
It is well established that the rate of not diminish. This indicates a high-rate
methanol electrooxidation at platinum can methanol of readsorption at PtRu catalysts.
be enhanced in the presence of Ru as The catalytic mechanism of PtRu has
a cocatalyst. In Fig. 28 are represented been interpreted in terms of a so-called
in situ FTIR spectra obtained during bifunctional effect of the surface in
oxidation of methanol on three different which Pt sites adsorb and dissociate
materials: Pt(111), Pt(111)/Ru(39%), and a methanol-forming CO and Ru atoms
PtRu alloy (85 : 15); the second electrode adsorb and dissociate water molecules,
was prepared by adsorbing Ru from a thus providing, at low potentials, oxygen
RuCl3 solution onto a well prepared atoms needed to complete the oxidation of
Pt(111) electrode [70]. For each material adsorbed CO to CO2 [75]. The facts above,
sample, spectra (R) were obtained at the showing an increased rate of adsorption of
indicated potential and calculated as the methanol in the presence of Ru, indicate
ratio R/R0 , with R0 being a respective that the bifunctional mechanism alone
reference spectrum obtained at 0.05 V, a does not fully describe the catalytic action
potential in which methanol adsorption of ruthenium.
is minimized.
For all electrodes, the band at 2341 cm−1 3.4.4.3 Ethanol
is due to CO2 formed in the thin-layer The electrocatalysis of ethanol oxidation
cavity. Another feature due to a soluble using in situ FTIR spectroscopy was stud-
species is the band at ca. 1720 cm−1 due to ied in several papers [76–80]. Some facts
a carbonyl group of the product formic acid concerning the soluble reaction products
c/o methyl formate (Methyl formate can at the low-index faces of Pt single-crystal
0.20 V 0.20 V 0.15 V
0.30 V 0.35 V
0.30 V
0.40 V
0.45 V
0.40 V 0.50 V
0.50 V
0.55 V 0.55 V
0.50 V
2056
2046
0.55 V
1721
1826
2059
1719
R/R0 = 0.02
R/R0 = 0.03
R/R0 = 0.01
Pt(111) 2341 Pt(111)/Ruads 2341 PtRu alloy

2341
2500 2000 2500 2000 2500 2000

Wavenumber
[cm−1]
Fig. 28 In situ FTIR spectra for Pt(111), Pt(111)/Ru 39%, and PtRu alloy
(85 : 15) in 0.5 M CH3 OH + 0.1 M HClO4 . Potentials as indicated on each
spectrum; reference spectrum taken at 0.05 V. (From Ref. [70] with permission.)
electrodes have been discussed by Chang C2 H5 OH + 0.1M HClO4 , ethanol adsorp-

et al. [77]. In a more or less quantitative tion is minimized. Therefore, this proce-
manner, these authors used FTIR spec- dure allows the monitoring of the complete
tra to estimate the amount of adsorbed process of ethanol adsorption and oxida-
and soluble species formed during a po- tion as the potential is increased in the
tential scan. Some disagreement with later positive direction. Because of the well-
results in Ref. [80] concerning the inten- known problem of surface poisoning by
sity of CO bands may be explained in strongly adsorbed organic residues, data
terms of a difference in the experimental obtained under these conditions may not
approach. be observable in the second and following
As in the case of methanol, when potential scans or when the electrode
Pt electrodes are contacted at 0.05 V comes in contact with ethanol without po-
versus RHE, with the solution containing tential control.
Spectral sequences during the elec- These anions migrate into the thin layer
trochemical adsorption and oxidation of of electrolyte to keep electric neutrality
0.1 M ethanol in 0.1 M HClO4 on Pt(111) (note, that on positively increasing the
and Pt(100) are displayed in Fig. 29. In potential, in addition to the double-layer
Table 2, the band assignments for the charge changes, the hydrogen ions con-
bands marked in the spectra are given. Ac- centration in the thin layer of electrolyte
etaldehyde, acetic acid, and CO2 have been increases. Both processes requiring the
earlier reported from in situ FTIR results, migration of negatively charged ions into
as products of ethanol oxidation [76]. In the thin layer).
addition to these soluble products, a Adsorbed species observed are linear-
gain of bulk perchlorate ions (band at bonded CO at ca. 2050 cm−1 and ac-
1110 cm−1 ) can be observed in the spec- etate ions at 1404 to 1412 cm−1 [82, 83].
tra with increasing positive potentials. Adsorbed acetate ions originate from
0.15
0.15
0.35
0.25
0.40 0.30
0.50
0.50
0.55 0.60
[V]
[V]
1866 0.65
E
0.60
0.70
2983
1713
2061 0.70 0.80

2055
2906
1712
1412
1407
1370
1282
2343
1282
R/R0 = 1.5%
2632
2341
R/R0 = 1.5%
Pt(111)
2625
Pt(100)
3000 2500 2000 1500 3000 2500 2000 1500

Wave number Wave number
[cm−1] [cm−1]
Fig. 29 In situ FTIR spectra (256 scans, 8 cm−1 resolution) at Pt(111) (left side) and Pt(100)
(right side) in 0.1 M C2 H5 OH + 0.1M HClO4 . Reference spectrum collected at 50 mV, and
sample spectra collected at the indicated potential in 50 mV steps (some spectra were
omitted for clarity). (From Ref. [80] with permission.)
Tab. 2 Assignment of some of the fundamental bands in the spectra of Fig. 29
Wave number Functional group or Mode, comments,

[cm−1 ] chemical species references
2983, 2906 CH3 , CH2 C−H str., 81

2632 (broad) COOH O−H str., 81
2341 CO2 C−O asym. str., 81
2055–2060 Adsorbed CO Linearly bonded, 60
1713 COOH or CHO C=O str., carbonyl, 81
1402 Adsorbed CH3 COO− C−O sym. str., 82
1370/1281 COOH Coupl. C−O str. −OH def., 81
1100 ClO4 − Cl−O str. (F), 35
dissociation of acetic acid in the ad- in Fig. 30(a, b) attest for the surface
sorbed state. Bridge-bonded CO, at about sensitivity of the reaction. At Pt(100) both
1860 cm−1 is observed at Pt(100) and linear and bridge-bonded CO start being
to a minor extent at Pt(111). In the formed short above 0.2 V. In the low-
presence of bulk ethanol, as in the exper- potential region, before oxidation begins,
iments in Fig. 29, positive-going features the site interconversion, CO-bridge →
as those near 3000 cm−1 (CH stretching CO-linear, is observed. This phenomenon
region [83]) are due to consumption of bulk is well-known for CO formed at Pt(100)
ethanol. either from dissolved CO [23, 25] or from
the adsorption of other fuels [84, 85]. At
3.4.4.3.1 Potential Dependence of the Band around 0.5 V, oxidation to CO2 begins;
Intensities The dependence of the band maximum production of CO2 is observed
intensities for COad and CO2 plotted at 0.8 V. At this potential, all adsorbed
Pt(111) Pt(100)
20
45
COL
CO2
Band intensity
Band intensity
15
CO2
30
[a.u.]
[a.u.]
10
COL COB
15
5
0
0
0.2 0.4 0.6 0.8 1.0 1.2 0.2 0.4 0.6 0.8 1.0
(a) [V] (b) [V]
Fig. 30 Band intensity for CO and CO2 from spectra collected during the
oxidation of 0.1 M ethanol in 0.1 M HClO4 at (a) Pt(111) and (b) Pt(100).
Experimental conditions as in Fig. 29. (From Ref. [80] with permission.)
CO is consumed and the CO2 signal CO and CO2 . Formation of acetaldehyde

decays as CO2 diffuses out of the thin- can be followed without interference
layer cavity. via a feature at 933 cm−1 [77]. However,
At Pt (111), the formation of linear- because of the cut off of CaF2 (around
bonded CO starts around 0.15 V. The 1000 cm−1 ), a flat ZnSe window had to
integrated band intensity in Fig. 30(a), in- be used to monitor the band intensity
dicates a maximum for CO formation for acetaldehyde (933 cm−1 ). Data for
at 0.55 V. The arrest observed at in- acetic acid (1280 cm−1 ) in Fig. 32 were
termediate potentials coincides with the taken from the spectra obtained with the
beginning of the formation of acetalde- ZnSe window.
hyde (band at 1713 in Fig. 29), thus The changes for acetaldehyde and acetic
indicating that this reaction competes with acid resemble those reported by Chang
that producing CO (i.e. cleavage of the et al. [77]. The onset of acetaldehyde for-
C−C bond). mation, at about 0.4 V, is accompanied by
A general phenomenon of small organic a sluggish appearance of acetic acid. Both
molecules is the presence of anodic signals exhibit a pronounced increase in
currents during the negative-going sweep intensity at 0.6 V, coinciding with the on-
of the voltammogram as shown in Fig. 31. set of CO oxidation, that is, with a higher
For the interpretation of this reactivation availability of Pt sites. It is interesting to
phenomenon, the formation of reaction note, during the positive-going polariza-
products during a cyclic scan as shown tion, that CO2 grows up to 1.0 V, that
in Fig. 32 should be considered. For is, beyond the potential at which the sur-
these experiments, a 60◦ CaF2 prism was face becomes free from CO. This being
used to monitor the bands for adsorbed an indication of the presence of other
1.0
0.8
Current density
0.6
[mA cm−2]
0.4
0.2
0.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4

Potential vs RHE
[V]
Fig. 31Cyclic voltammogram for a Pt(111) electrode in 1 M C2 H5 OH + 0.1 M
HClO4 solution. Sweep rate 0.05 V/s.
(a) 8
9
COB
6
8
4 7
CH3COOH
2 6
Band intensity
Band intensity
0 COL 5
[a.u.]
[a.u.]
4
1.5 CH3CHO × 3
3
1.0 2
CO2
1
0.5
0
0.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2
(b) [V] (c) [V]
Fig. 32 Potential dependence of the band intensities for CO, CO2 , acetaldehyde, and acetic
acid, from spectra obtained at Pt(111) during the oxidation of ethanol in 1 M C2 H5 OH + 0.1 M
HClO4 . After taking a reference spectrum at 0.05 V, potential steps were applied towards higher
positive potentials and then back to the initial value. All spectra are referred to the first one taken
at 0.05 V. A 60◦ prismatic CaF2 window was used for CO and CO2 and a flat ZnSe window for
acetaldehyde and acetic acid. (From Ref. [80] with permission.)
strongly adsorbed species not observed in strength of water adsorption increases as

the spectra. As the applied polarization the applied potential grows in the posi-
is reversed, both acetaldehyde and acetic tive direction. During the negative-going
acid increase slowly. However, at ca. 0.9 V, scan, a decrease of water–platinum inter-
which is the potential at which the cur- action favors the access of ethanol to the
rent in the voltammogram increases again, surface as indicated by the increase of the
a pronounced increase of acetic acid is CO feature.
observed. Also, CO2 and CO grow during
the reverse scan from 0.8 V downwards. 3.4.5
The recovering in activity has been re- Final remarks
lated to a competition between water and
the organic substance for Pt sites [86, 87]. The examples presented in this chap-
Infrared spectroscopy data show that the ter show the power of in situ FTIR
spectroscopy to study electrochemical sys- 11. A. Bewick, K. Kunimatsu, B. S. Pons et al., J.

tems. Under equilibrium conditions the Electroanal. Chem. 1984, 160, 47–61.
12. B. Beden, A. Bewick, C. Lamy, J. Electroanal.
infrared spectroscopy provides molecular
Chem. 1983, 148, 147–160.
level information on the double layer, 13. A. Bewick, J. Electroanal. Chem. 1983, 150,
contributing to the understanding of the 481–493.
physics and chemistry of the charged in- 14. T. Bae, D. A. Scherson, E. Yeager, Anal.
terface. Thus, the potential and structure Chem. 1990, 62, 45–49.
15. D. S. Corrigan, M. J. Weaver, J. Electroanal.
dependence of ion adsorption and the ge-
Chem. 1988, 239, 55–66.
ometry of the adsorbed ions as well, can 16. T. Iwasita, F. C. Nart, J. Electroanal. Chem.
be inferred from the inspection of the IR 1990, 295, 215–224.
spectra. 17. P. R. Griffiths, J. A. de Haseth in Chemi-
On the other hand, under conditions cal Analysis: A Series of Monographs on
of current flow, IR spectra can be used Analytical Chemistry and Its Applications
(Eds.: P. J. Elving, J. D. Winefordner), John
as a tool to obtain analytical informa- Wiley & Sons, New York, 1986, Vol. 83,
tion on electrochemical processes. This pp. 274–283.
approach has provided valuable informa- 18. S. Pons, J. Electroanal. Chem. 1983, 150,
tion on the electrocatalysis of the oxi- 495–504.
dation of organic substances, which is 19. S. Pons, T. Davison, A. Bewick, J. Electroanal.
Chem. 1984, 160, 63–71.
currently used for mechanistic interpre- 20. S. C. Chang, M. J. Weaver, J. Chem. Phys.
tations and for the evaluation of catalytic 1990, 92, 4582–4594.
materials. 21. S. C. Chang, M. J. Weaver, J. Phys. Chem.
1990, 94, 5095–5102.
22. S. C. Chang, M. J. Weaver, Surf. Sci. 1990,
Acknowledgments 230, 222–236.
23. S. C. Chang, M. J. Weaver, Surf. Sci. 1990,
FAPESP and CNPq are gratefully acknowl- 238, 142–162.
edged for their financial support. 24. F. Kitamura, M. Takahashi, M. Ito, Surf. Sci.
1989, 223, 493–508.
25. F. Kitamura, M. Takahashi, M. Ito, J. Phys.
References Chem. 1988, 92, 3320–3323.
26. D. S. Corrigan, M. J. Weaver, J. Electroanal.
1. L. H. Little, Infrared Spectra of Adsorbed Chem. 1988, 241, 143–162.
Species, Academic Press, London, 1966. 27. L. H. Leung, M. J. Weaver, J. Phys. Chem.
2. S. A. Francis, A. H. Ellison, J. Opt. Soc. Am. 1988, 92, 4019–4022.
1959, 49, 131–138. 28. K. Kunimatsu, H. Seki, W. G. Golden et al.,
3. R. G. Greenler, J. Chem. Phys. 1966, 50, Surf. Sci. 1985, 158, 596–608.
310–315. 29. H. Seki, K. Kunimatsu, W. G. Golden, Appl.
4. A. H. Reed, E. Yeager, Electrochim. Acta Spectrosc. 1985, 39, 437–443.
1970, 15, 1345–1354. 30. T. Iwasita, F. C. Nart in Advances in Elec-
5. H. Neff, P. Lange, D. K. Roe et al., J. Elec- trochemical Science and Engineering (Eds.:
troanal. Chem. 1983, 150, 513–519. H. Gerischer, C. W. Tobias), Verlag Chemie,
6. W. N. Hansen, J. Opt. Soc. Am. 1968, 58, 380. Weinheim, 1995, Vol. 4, pp. 123–216.
7. K. Ataka, M. Osawa, Langmuir 1998, 14, 31. B. N. J. Persson, Solid State Commun. 1979,
951–959. 30, 163–166.
8. M. Osawa, K. Ataka, K. Yoshii et al., J. Elec- 32. I. R. Moraes, F. C. Nart, J. Electroanal. Chem.
tron. Spectrosc. 1993, 64, 371–379. 1998, 461, 110–120.
9. Z. Ping, G. E. Nauer, H. Neugebauer et al., 33. M. C. P. M. da Cunha, M. Weber, F. C. Nart,
Electrochim. Acta 1997, 42, 1693–1700. J. Electroanal. Chem. 1996, 414, 163–170.
10. A. Bewick, K. Kunimatsu, B. S. Pons, Elec- 34. I. R. Moraes, M. C. P. M. da Cunha, F. C.
trochim. Acta 1980, 25, 465–480. Nart, J. Bras. Chem. Soc. 1996, 7, 453–460.
35. K. Nakamoto, Infrared and Raman Spectra of 60. Beden, C. Lamy, A. Bewick et al., J. Elec-
Inorganic and Coordination Compounds, 4th troanal. Chem. 1981, 121, 343.
edition, John Wiley & Sons, New York, 1986. 61. D. K. Lambert, Electrochim. Acta 1996, 41,
36. P. Christensen, A. Hamnett in Comprehen- 623–630.
sive Chemical Kinetics (Eds.: R. G. Compton, 62. W. F. Lin, T. Iwasita, W. Vielstich, J. Phys.
A. Hamnett), Elsevier, Amsterdam, 1989, Chem. B 1999, 103, 3250–3257.
Vol. 29, pp. 1–104. 63. W. F. Lin, M. S. Zei, M. Eiswirth et al., J.
37. R. J. Nichols, A. Bewick, Electrochim. Acta Phys. ChemB 1999, 103, 6968–6977.
1988, 33, 1691–1694. 64. S. Cram, K. A. Friedrich, K. P. Geyzers et al.,
38. E. L. de Sa, Ph. D. Thesis, Universidad de Fresenius’ J. Anal. Chem. 1997, 358, 189–192.
São Paulo, São Carlos, 1997. 65. W. F. Lin, P. A. Christensen, A. Hamnett
39. P. Christensen, A. Hamnett, A. V. G. Muir, et al., J. Phys. Chem. 2000, 104, 6642–6652.
J. Electroanal. Chem. 1986, 241, 361–371. 66. N. Ikemiya, T. Senna, M. Ito, Surf. Sci. 2000,
40. A. C. Chapman, L. E. Thirwell, Spectrochim. 464, L681–L685.
Acta 1964, 20, 937. 67. T. E. Madey, C. Benndorf, Surf. Sci. 1985,
41. T. Iwasita, F. C. Nart, Prog. Surf. Sci. 1997, 164, 602–624.
55, 271–340. 68. H. Pfnür, D. Menzel, F. M. Hoffmann et al.,
42. G. J. Edens, X. Gao, M. J. Weaver, J. Elec- Surf. Sci. 1980, 93, 431–452.
troanal. Chem. 1994, 375, 357–366. 69. B. N. J. Persson, R. Rydberg, Phys. Rev. B
43. M. Lennartz, P. Broekmann, M. Arenz et al., 1981, 24, 6954–6970.
Surf. Sci. 1999, 442, 215–222. 70. T. Iwasita, H. Hoster, A. John-Anacker et al.,
44. A. Tadjeddine, A. Le Rille, Electrochim. Acta Langmuir 2000, 16, 522–529.
1999, 45, 601–609. 71. H. Hoster, T. Iwasita, H. Baumgärtner et al.,
45. F. C. Nart, T. Iwasita, M. Weber, Electrochim. PCCP 2001, 3, 337–346.
Acta 1994, 39, 2093–2096. 72. B. E. Nieuwenhuys, W. H. M. Sachtler, Surf.
46. I. R. Moraes, F. C. Nart, J. Braz. Chem. Soc.; Sci. 1973, 34, 225.
2001, 12, 138–143. 73. B. E. Nieuwenhuys, R. Bouwman, W. H. M.
47. P. W. Faguy, N. Markovic, R. R. Adzic et al., Sachtler, Thin Solid Films 1973, 21, 5.
J. Electroanal. Chem. 1990, 289, 245–262. 74. E. Ticianelli, J. G. Beery, M. T. Paffett et al.,
48. Y. Shingaya, M. Ito, J. Electroanal. Chem. J. Electroanal. Chem. 1989, 258, 61–77.
1999, 467, 299–306. 75. M. Watanabe, S. Motoo, J. Electroanal. Chem.
49. Y. Shingaya, M. Ito, Surf. Sci. 1997, 386, 1975, 60, 267–273.
34–47. 76. T. Iwasita, B. Rasch, E. Cattaneo et al., Elec-
50. P. W. Faguy, N. S. Marinkovic, R. R. Adzic, trochim. Acta 1989, 34, 1073–1079.
Langmuir, 1996, 12, 243–247. 77. S.-C. Chang, L.-W. H. Leung, M. J. Weaver,
51. P. W. Faguy, N. S. Marinkovic, R. R. Adzic, J. Phys. Chem. 1990, 94, 6013–6021.
J. Electroanal. Chem. 1996, 407, 209–218. 78. F. Cases, M. López-Atalaya, J. L. Vázquez
52. M. Funtikov, U. Stimming, R. Vogel, J. Elec- et al., J. Electroanal. Chem. 1999, 278,
troanal. Chem. 1997, 428, 147–153. 433–440.
53. L. J. Wan, S. L. Yau, K. Itaya, J. Phys. Chem. 79. T. Iwasita, E. Pastor, Electrochim. Acta 1994,
1995, 99, 9507–9513. 39, 531–537.
54. L. J. Wan, M. Hara, J. Inukai et al., J. Phys. 80. X. H. Xia, H.-D. Liess, T. Iwasita, J. Elec-
Chem. B 1999, 103, 6978–6983. troanal. Chem. 1997, 437, 233–240.
55. F. C. Nart, T. Iwasita, Electrochim. Acta 1992, 81. G. Socrates, Infrared Characteristic Group
37, 2179–2184. Frequencies, Wiley, New York, 1966.
56. F. C. Nart, T. Iwasita, Electrochim. Acta 1996, 82. A. Rodes, E. Pastor, T. Iwasita, J. Electroanal.
41, 631–636. Chem. 1994, 376, 109–118.
57. F. C. Nart, T. Iwasita, M. Weber, Electrochim. 83. J. Shin, W. J. Tornquist, C. Korzeniewski
Acta 1994, 39, 2093–2096. et al., Surf. Sci. 1996, 364, 122–130.
58. T. Iwasita, F. C. Nart, A. Rodes et al., Elec- 84. X. H. Xia, T. Iwasita, F. Y. Ge et al., Elec-
trochim. Acta 1995, 40, 53–59. trochim. Acta 1996, 41, 711–718.
59. F. C. Nart, T. Iwasita, J. Electroanal. Chem. 85. T. Iwasita, X. H. Xia, E. Herrero et al., Lang-
1992, 322, 289–300. muir 1996, 12, 4260–4265.
86. T. Iwasita, X. H. Xia, J. Electroanal. Chem. of the Poynting vector is Eq. (4)
1996, 411, 95–102.
87. T. Iwasita, X. H. Xia, H.-D. Liess et al., J. 1

Phys. Chem. B 1997, 101, 7542–7547. S = Re(E × H∗ ) (A3)
2
Therefore, the rate of the radiation dissi-
Appendix
pation will be Eq. (2)

If S is the vector representing the elec- 1

−divS = σ E × E∗ (A4)
tromagnetic energy flux intensity, the 2
conservation equation will be
where σ is the effective conductivity of the
dw absorbing medium at the field frequency
+ div S = 0 (A1)
dt and the quantity E × E∗ is the time average
of the E2 , therefore
where dw/dt is the speed of energy
dissipation. S is the Poynting vector and is 1

−divS = σ E2 (A5)
defined by Eq. (4) 2
S=E×H (A2) Hence the rate of energy dissipation

is proportional to the square of the
Where E is the electric field and H is mean electric field of the incident ra-
the magnetic field of the electromagnetic diation and it is the result of the
radiation. Giving that the electromagnetic spectroscopic measurement. The inten-
radiation is a high-frequency radiation, the sity of the E2 is directly related to the
observed signal will be a time average of amount of dissipated energy by the ab-
the radiation dissipation. The time average sorbing medium.
295
4.1 methods have provided remarkable insight

Structure Relationships in Electrochemical into the surface electrochemistry and the
Reactions ‘‘structure sensitivity’’ of many important
electrochemical processes.
C. A. Lucas
In this chapter, we review the progress in
University of Liverpool, Liverpool, United
the surface science of electrochemical re-
Kingdom
actions. The methods of the surface X-ray
N. M. Marković diffraction and rotating ring disk experi-
University of California, Berkeley, California ments are described in Sect. 4.1.2, for sur-
face X-ray diffraction – the focus being on
4.1.1 the application of the technique to single-
Introduction crystal electrodes in the electrochemical
environment. Section 4.1.3 then describes
In this report, we review recent studies the nature of the surface structures that are
of single-crystal transition metal electrode formed on ‘‘bare’’ single-crystal transition
surfaces by combining in situ surface X-ray metal surfaces in electrolyte. The phenom-
diffraction measurements with more tradi- ena of surface reconstruction and surface
tional electrochemical techniques in order relaxation are illustrated by detailed ex-
to probe the influence of the surface struc- amples, namely Au(001) and Pt(110), in
ture on the electrochemical reactivity. This which the dependence of the surface struc-
functionality is generally termed structure ture on the applied electrode potential is
sensitivity. In the last decade, the in situ sur- described. Section 4.1.4 first describes the
face X-ray diffraction technique has been surface structures and energetics of an-
a critical tool for determining the potential ion adsorption onto different single-crystal
stability of specific surface structures in electrode surfaces, that is, Pt and Au, and
electrolyte under reaction conditions. On then the underpotential deposition (UPD)
the other hand, the rotating ring disk elec- of metals onto the same substrates. Re-
trode (RRDE) has been routinely used for sults from the different metal surfaces
determining the kinetics of electrochemi- are presented in an attempt to under-
cal reactions on single-crystal surfaces and stand the influence of the substrate on
evaluating the potential-dependent surface the structure of the overlayer and the effect
coverage by an adsorbed species. In com- of anions on the UPD process. The re-
bination, the X-ray diffraction and RRDE sults highlight the strength of combining
296 2 Interfacial Structure and Kinetics
RRDE measurements with X-ray diffrac- cells. This gives

tion results to investigate the detailed
sin2 (N1 Q·a1 )
nature of the surface structures that are I (Q) ∝ F (Q)F ∗ (Q)
formed, particularly in coadsorption stud- sin2 (Q·a1 )
ies, for example, the influence of anion sin2 (N2 Q·a2 ) sin2 (N3 Q·a3 )
species on the UPD process. In Sect. 4.1.5, × (1)
sin2 (Q·a2 ) sin2 (Q·a3 )
the oxygen reduction reaction (ORR) is
used as a model electrochemical reaction where Q is the momentum transfer,
to demonstrate the relation between the defined by the scattering angle 2θ and
metal–O2 energetics and reaction path- the unit cell structure factor, F , is given by
way/kinetics as well as the importance of

n
the local symmetry of surface atoms in de- F (Q) = fj exp(iQ·rj ) (2)
termining the electrocatalytic properties of j =1
metal surfaces.
where rj is the relative atomic position
4.1.2 in the unit cell and the sum is over all
Methods of the atoms in the unit cell each having
an atomic scattering factor fj (fj is Q-
4.1.2.1 X-ray Diffraction dependent and proportional to the atomic
number of the relevant atom). When the
4.1.2.1.1 Theory It is beyond the scope crystal is large, Eq. (1) produces a series
of this review article to provide a compre- of very sharp peaks in which Q·ai = π. If
hensive description of basic X-ray diffrac- the scattering is from a single monolayer
tion from surfaces. Instead, readers are (ML), that is, N3 = 1 (the a3 vector is along
referred to the excellent reviews by Fei- the surface normal), then the scattering
denhans’l [1], Fuoss and Brennan [2], and normal to the crystal surface becomes
Robinson and Tweet [3] for explicit details. completely diffuse.
In this section, we will focus on some ba- Considering the case of a perfectly
sic ideas that pertain to X-ray diffraction terminated crystal surface and assuming
studies of surfaces in an electrochemi- that the Bragg conditions are met in the
cal environment, in particular with regard plane of the crystal (i.e. Qx ·a1 = Qy ·a2 =
to crystal-truncation rod (CTR) measure- nπ), then for a semi-infinite crystal, Eq. (1)
ments. reduces to

Since the interaction of hard X-rays N 2 N 2 F (Q)F ∗ (Q) Qz ·a3
with matter is weak, provided that the I (Q) ∝ 1 2 sin2
2 2
exact Bragg conditions for a perfect crystal (3)
are not met, the kinematical scattering The Bragg points of the finite crystal are
approximation, in which scattering is connected by streaks of intensity along the
treated as a single event, can be used. For surface normal direction in the reciprocal
a crystal volume defined by N1 , N2 , and lattice. At the minimum position, midway
N3 unit cells along the crystal axes defined between Bragg reflections, the scattered
by the vectors a1 , a2 , and a3 , the scattered intensity is similar to the scattering
intensity can be written as the product of expected from a single monolayer. The
two scattering amplitudes, one from the streaks of intensity are known as crystal-
unit cell and one from the lattice of unit truncation rods (CTRs) as they arise from
4.1 Structure Relationships in Electrochemical Reactions 297
the truncation of the crystal lattice at the the atoms in the surface layer present.
surface [3, 4]. Similar effects are observed when adsor-
Figure 1 illustrates some of the informa- bates are formed on crystal surfaces as a
tion that can be obtained from the analysis result of the difference in the scattering
of CTR data. The intensity variation along factors between the adsorbed atoms and
a CTR as a function of the surface nor- the substrate atoms. Hence for a mono-
mal phase shift, Q·a3 = 2πl, is shown for layer of adsorbed atoms, which have half
the case of a perfect surface (solid line), of the scattering power of the substrate
a surface with a reduced density in the atoms, a similar result to the dotted line
topmost atomic layer (dotted line), and in Fig. 1 would be observed. Similarly, an
a surface in which the topmost atomic adlayer with an increased surface density
layer is expanded away from the bulk crys- would cause an increase in the intensity
tal (dashed line). When l is an integer, at the anti-Bragg conditions. For CTRs
the Bragg conditions are satisfied and the measured with an in-plane contribution
kinematical approximation breaks down. to the momentum transfer, the changes
The minimum of the scattering is ob- to the CTR profile are dependent on the
served at the ‘‘anti-Bragg’’ positions when relative phase of the scattering from the
l has half-integer values. Surface rough- substrate and the adlayer, which can cause
ness causes a decrease in the scattering either an increase or decrease in the in-
at the anti-Bragg positions as shown by tensity at the anti-Bragg positions. Finally,
the dotted line in the figure, which is a a change in the lattice spacing at the sur-
calculation for a surface with only half of face causes asymmetry in the intensity
103
Calculated intensity
102
10
0 p 2p 3p 4p
Fig. 1 The intensity variation along a CTR as a function of the
surface normal phase shift, calculated for a perfect surface (solid
line), a surface with a lower density topmost atomic layer (dotted
line) and a surface in which the topmost atomic layer is
expanded by 10% of the layer spacing (dashed line).
around the Bragg reflections, as shown surface atoms themselves). By combining

by the dashed line in Fig. 1, which is specular CTR results (in which Q is en-
for a 10% expansion of the surface layer tirely along the surface normal direction)
relative to the bulk. Again, similar ef- with nonspecular CTR results (in which Q
fects are produced by adsorbates, as the has an additional in-plane contribution), it
adsorbate-substrate layer spacing is very is possible to build up a three-dimensional
often different from that of the bulk sub- picture of the atomic structure at the elec-
strate layer spacing. trode surface. If the surface or adlayer
By careful modeling of CTR data, it adopts a different symmetry from that of
is possible to extract structural infor- the underlying bulk crystal lattice, then
mation, such as surface coverage, sur- the scattering from the surface becomes
face roughness, and layer spacings (both separate from that of the bulk in recip-
adsorbate–substrate distances and the rocal space and it is possible to measure
expansion/contraction of the substrate the surface scattering independently. This
(0 1 l)
(0 1 l) (1 1 l)
(015)
(114)
(0 0 l)
(013)
(1 0 l)
(0 k 0) (112)
(012) (0 0 l) (1 0 l)
(0 k 0) (004)
(003) (104) (011)
(103)
(110)
(002)
60°
(101) (101)
(a) (h 0 0) (b) (h 0 0)
Fig. 2 The real-space surface structures and corresponding reciprocal space lattices for:
(a) (111); (b) (001); and (c) (110) surface terminations of an fcc crystal. The unit cells are shown
in the real-space structures and the corresponding Bragg reflections indicated in the reciprocal
space lattices.
Fig. 2 (Continued)
(0 1 l) (1 1 l)
(0 0 l) (1 0 l) (112)
(002)
(011 )
(0 k 0)
(101)
(110)
(c)
(h 0 0)
independent structural information can 4.1.4 can be quoted without reference

be combined with CTR analysis to give to the experimental details in each case.
the registry of the surface layers with re- Figure 2 shows the reciprocal space lattices
spect to the bulk lattice [1–3]. CTR and for the (111), (001), and (110) crystal
surface structure analysis are presented in faces with the corresponding real-space
Sects. 4.1.3 and 4.1.4 of this report for a surface structures shown schematically in
variety of electrochemical surfaces. each case. In surface X-ray diffraction
experiments, it is always desirable for the
4.1.2.1.2 Reciprocal Lattices for Low-index surface normal direction to be defined as
fcc Metal Surfaces This report describes the (0, 0, l) reciprocal space direction.
results from the three low-index surfaces This is also the convention adopted for
of metal crystals with the face centered low-energy electron diffraction (LEED)
cubic (fcc) crystal structure. As such, it experiments.
is instructive to describe the reciprocal The close-packed (111) surface is shown
space lattice for each surface orientation so in Fig. 2(a). For this surface, a hexagonal
that the results from the X-ray diffraction unit cell is defined, such that the surface
experiments presented in Sects. 4.1.3 and normal is along the (0, 0, l)hex direction and
the (h, 0, 0)hex and (0, k, 0)hex vectors lie in the form
the plane of the surface and subtend 60◦ .
The units for h, k, and l are a ∗ = b∗ = I (h, k, l) ∝
√ √ 2
4π/ 3aNN and c∗ = 2π/ 6aNN where f

aNN is the nearest-neighbor distance in the 1 − exp[−2πi(h/2 + k/2 + l/2)] (6)
crystal. Because of the ABC stacking along
the surface normal direction, the unit cell Equations (4–6) are by no means rigorous
contains three monolayers and the Bragg but they essentially form the basis for
reflections are spaced apart by multiples analyzing CTR data from the three low-
of three in l. The basic scattering equation index surfaces of fcc transition metals.
for an unrelaxed, unmodified (111) surface The equations are easily modified to
along the CTRs shown in Fig. 2(a) is then include surface relaxation effects, surface
given by roughness, or adsorbed adlayer species
and are used in the analysis of the results
I (h, k, l) ∝ presented in Sects. 4.1.3 and 4.1.4 of
2 this report.
f

1 − exp[2πi(h/3 − k/3 − l/3)] (4)
4.1.2.1.3 Experimental Details The gen-
where f is the atomic form factor eral experimental procedures used in X-ray
(including the Debye–Waller temperature diffraction measurements of electrochem-
factor [5]). ical systems have been described in detail
The (001) surface (Fig. 2b) is more open in previous articles [6–8]. After surface
than the (111) surface and is indexed to a preparation of the electrode to be studied,
surface tetragonal unit cell. This is related the crystal is transferred (preferably whilst
to the conventional cubic unit cell by the being protected from the atmosphere by a
transformations (1, 0, 0)t = 1/2(2, 2, 0)c , drop of electrolyte on the surface) into an
(0, 1, 0)t = 1/2(2, −2, 0)c , and (0, 0, 1)t = X-ray electrochemical cell. The design of
(0, 0, 1)c . The units for h, k, and l are the electrochemical cell that we have used
√
a ∗ = b∗ = 2π/aNN and c∗ = 4π/ 2aNN , in our experiments is shown schematically
where aNN is the nearest-neighbor distance in Fig. 3. In this design, the crystal is held
in the crystal. The CTR scattering, by in place by tightening down the central col-
analogy with Eq. (4), is given by umn of the cell, which is screw-threaded
2 into the main body (all parts are made of
f Kel-F material, which can be chemically
I (h, k, l) ∝
1 − exp[−πi(h + k + l)] cleaned and is relatively easy to machine).
(5) The crystal surface then forms the highest
Finally, the (110) reciprocal surface unit part of the cell and so is easily accessi-
cell is rectangular as shown in Fig. 2(c) and ble to the incident X-ray beam. Electrical
the reciprocal lattice notation is such that contact to the working electrode and to a
h is along [1 −1 0], k along [0 1 0], and l is counter and reference electrode are made
along the [1 1 0] surface normal. The units through feedthrough connections on the
for h, k, and l are a ∗ = c∗ = 2π/aNN and side of the cell. In our experiments, a pal-
√
b∗ = 4π/ 2aNN , where aNN is the nearest- ladium/hydrogen reference electrode was
neighbor distance in the crystal. For this used, which was in direct contact with
surface, the CTR scattering reduces to the electrolyte in the main body of the
Polypropylene film
Sample
Reference electrode
Electrolyte in Electrolyte out
Counterelectrode
Working electrode
Fig. 3 A schematic diagram of the electrochemical X-ray scattering cell.
cell. Care has to be taken in this situa- some results by other authors are used,
tion as, depending on the nature of the which were obtained on beamlines at the
electrochemical reactions that are being Brookhaven National Synchrotron Light
studied, there can be shifts in the refer- Source (NSLS). For experimental details,
ence potential. Two further connections readers are referred to the referenced work
form electrolyte feedthroughs to the cell in each case. For a general description of
and the liquid is contained within the cell the procedures used in synchrotron sur-
by a thin (∼10 µm) polypropylene window face X-ray diffraction experiments, readers
sealed with an ‘‘o’’ ring. During the X-ray are referred to the excellent review by Fei-
experiment, the electrochemical cell can be denhans’l [1]. A typical procedure for any
deflated to trap a thin layer of electrolyte given electrochemical system used in our
(∼10 µm) between the electrode surface experiments is roughly as follows:
and the X-ray window to shorten the so-
lution pathlength of the X-ray beam. Note (1) In the X-ray cell, immediately after
that this cell design allows the sample to be mounting in the X-ray goniometer and
oriented in any direction as the electrolyte being put under potential control, the
is contained, that is, there is no need for cyclic voltammetry (CV) was measured
any specialized diffractometer equipment for comparison with literature results.
in order to maintain a horizontal surface After the initial cycling, the electrode
geometry. In order to minimize contam- potential was transferred to computer
ination from the atmosphere, the cell is control for the duration of the X-ray
surrounded by an outer shell containing a experiment. Although with time the
nitrogen atmosphere. features in the CV became distorted
Most of the X-ray measurements de- as the electrolyte made contact with
scribed in this report were carried out the polycrystalline back and sides of
at room temperature on beamline 7-2 the crystal, the general features were
at the Stanford Synchrotron Radiation always observable for the duration
Laboratory (SSRL) utilizing a focused of the experiment (typically up to a
monochromatic X-ray beam. Additionally, maximum of 24 h).
(2) The X-ray spectrometer would be 4.1.2.1.4 New Techniques Surface X-ray
aligned with two or more Bragg diffraction is now a well-established tech-
reflections to relate the spectrometer nique for probing the atomic structure at
angles to the reciprocal space lattice the electrochemical interface and, since
of the sample. Subsequently all data the first in situ synchrotron X-ray study in
were measured in units of the sample 1988 [6], several groups have used the tech-
reciprocal space (see previous section). nique to probe a variety of electrochemical
(3) For a particular electrochemical sys- systems. Most analysis has followed the
tem, the dependence of the X-ray methodology outlined in Sects. 4.1.2.1.1 to
scattering intensity at various key re- 4.1.2.1.3, whereby structural information
ciprocal space positions (usually at at a fixed electrode potential is obtained
points on the CTRs) would be mon- by detailed measurement of the scatter-
itored as a function of the applied elec- ing from surface structures and modeling
trode potential over the potential range of the CTR profiles. In this section, a
of interest. This potentiodynamic tech- couple of recent applications of X-ray scat-
nique, which we have termed as X-ray tering are described that are particularly
voltammetry (XRV) [9], is described in useful in studies of the electrochemical
more detail in Sect. 4.1.2.1.4. interface.
(4) The XRV measurement is used to
(1) X-ray Voltammetry (XRV): XRV is a
identify structural changes that are
potentiodynamic measurement whereby
occurring at different electrode poten-
the scattered X-ray intensity at a particular
tials. These would then be probed with
reciprocal lattice point is monitored as the
the electrode potential held at the rel-
potential is cycled. As can be seen from
evant value. Typically, in searching for
Fig. 1, positions on the CTRs are sensitive
in-plane structures, scans along the
to surface expansion or to adsorption
(h, 0, 0), (0, k, 0), and (h, k, 0) in-
plane reciprocal space directions at effects. By measuring the XRV at a number
small out-of-plane momentum trans- of different CTR positions, an insight
fer (l = 0.1–0.15) were performed. into the nature of the structural changes
From these scans, it would be hoped at the surface can be obtained without
to locate scattering that had a dif- recourse to the detailed measurement
ferent symmetry to that of the bulk of the CTR profiles [11, 12]. If scattering
CTR scattering. Data for CTR analy- as a result of an ordered surface layer
sis would be obtained by performing with a different symmetry to that of
rocking scans at sequential l values the underlying bulk crystal is obtained,
along a particular CTR. The rocking then the measurement of the potential
scans can then be integrated (either dependence of this scattering directly
numerically or by fitting a functional indicates the potential range of stability
form to the lineshape), which results of the structure. This information is
in the CTR profile plotted as l versus particularly useful for comparison with CV
integrated intensity. This data is suit- results as features in the CV can be directly
able for analysis using the equations correlated with structural changes at the
given in the previous section, modi- electrode surface. Several examples of this
fied by the appropriate instrumental measurement are presented in Sects. 4.1.3
resolution corrections [10]. and 4.1.4 of this report.
(2) Anomalous scattering techniques: In of electrochemical reactions. Although

Sect. 4.1.2.1.1, the atomic form factor, f , several experimental methods are now
gives the scattering power of each atom routinely used in the surface electrochem-
within a particular unit cell structure. istry, the rotating disk electrode (RDE)
When the incident X-ray energy is away and rotating ring disk electrode (RRDE)
from atomic adsorption edges, f is Q- techniques have become standard tools in
dependent and related to the number of determining kinetic parameters, reaction
electrons that the atom has. However, products, and potential-dependent surface
if the X-ray energy is tuned near to an coverage by adsorbates [15, 16]. The RDE
adsorption edge then f must be modified is a solid disk embedded in an isolating
by the anomalous dispersion corrections material attached to the shaft that ro-
according to tates about the electrode surface normal.
The Navier–Stokes equation can be solved
f (Q, E) = f0 (Q) − f (E) − if (E) (7) exactly for this system, providing exact
correction for reactant mass transport lim-
Close to an adsorption edge f (E) can in- itations. As the concentration of dissolved
crease to ∼10 electrons and so can have gasses, viz. H2 , O2 , CO, and so on, in aque-
a dramatic effect on the atomic form fac- ous solutions is relatively low, for example,
tor and hence the scattered X-ray signal. 10−5 –10−6 mol cm−2 , the mass-transfer
In the complex electrochemical environ- limitation in the hydrogen oxidation re-
ment, this effect can be extremely helpful action (HOR), oxygen-reduction reaction
in identifying structural models, especially (ORR), and CO oxidation reaction are very
in coadsorbate systems. For example, the significant, and the use of the RDE is
contribution to the X-ray scattering from absolutely essential for definitive kinetic
a particular atom in a structure may be studies. The most useful equation to study
effectively isolated by tuning the energy electrode kinetics that are under mixed
of the incident X-ray beam through an control, that is, the case when the activa-
atomic adsorption edge. This can help tion (iac ) and mass transport–controlled
the interpretation of X-ray diffraction data (iL = Bω1/2 ) current densities combine to
(examples are given in Sect. 4.1.4.2 for an- yield the total current density as the sum
ion/metal coadsorption) and, additionally, of reciprocal, is
may be used to probe oxidation states
at surfaces and in adsorbates [13]. The 1 1 1
= + (8)
combination of diffraction measurements i iac iL
with traditional extended X-ray absorption where the diffusion limiting current is pro-
fine-structure (EXAFS) techniques (so- portional to angular velocity ω1/2 [15, 16].
called diffraction anomalous fine-structure The success of the RDE method has
(DAFS) [14]) may lead to studies in which stimulated the development of several
the local environment of atoms selected by other rotating configurations. The RRDE
the diffraction conditions is probed spec- is perhaps the most useful extension of
troscopically. the idea of the RDE. The RRDE was first
developed by Frumkin and Nekrasov to
4.1.2.2 Rotating Ring Disk Methods detect unstable intermediates in electrode
A large array of experimental techniques reactions. The RRDE, Fig. 4, consists of
has been developed to probe the behavior a central disk electrode surrounded by
Fig. 4 The RRDE consisting of a central disk electrode surrounded

by a concentric insulating annulus and a concentric ring electrode.
a concentric insulating annulus (Teflon) interference from kinetic and double-

and then a concentric ring electrode. layer effects. The best example is UPD
To operate an RRDE, a bipotentiostat is of copper in solution containing Br− , in
needed, which can control the potential which the amount of Cu UPD and/or
of the disk and the ring independently, bromide anion deposited/adsorbed on the
in order to measure the current going disk electrode can be estimated using
through each of them separately. The the two ring-electrode properties – ring
reaction intermediates formed on the disk shielding and ring collection [16]. For both
electrode can be monitored using the modes, the UPD species is reduced to
ring-collection properties. In the ring- metal at the ring, and electroactive species
collection mode, intermediates produced in solution must undergo convective-
on the disk electrode can be ‘‘collected’’ diffusion controlled reduction at the ring
by the ring electrode [16]. In the RRDE electrode. The ring-shielding property of
configuration, however, only a fraction the RRDE is illustrated in Sect. 4.1.4.2.3 in
of intermediates produced on the disk which it is used to assess the mass flux of
electrode will reach the ring electrode. This both the Cu2+ and the Br− from and to the
fraction is determined by the collection Pt(111) disk electrode. If either Br or Cu
efficiency, N , which is defined as the is adsorbed at the Pt(111) disk, the surface
ratio between the ring and disk currents, coverages θBr,Cu can be assessed from the
N = ir /id . In Sect. 4.1.5, the collection ring currents in either potentiodynamic or
mode of the RRDE is used to monitor potentiostatic experiments:
the formation of intermediates during the

ORR on the disk electrode.
The RRDE can also be useful to obtain (1/v) (ir − ir∞ ) dE
1
accurate thermodynamic data without θBr,Cu = (9)
Q AnN
∞ the preparation of Au(hkl) is to use the
(ir(t) − ir∞ ) dτ
1 0 flame-annealing technique [17]. After pre-
θBr,Cu = (10) treatment to produce a flat, well-oriented
Q AnN
surface, the crystal is heated in an hy-
where ir∞ and ir are unshielded (no drogen or butane flame and then allowed
UPD process on the disk electrode) and to cool in air before transfer to the cell.
shielded ring currents, respectively, N is This method, or variations of it, can pro-
the collection efficiency, v refers to the duce clean surfaces as evidenced by the
sweep rate, n is the number of electrons subsequent observation of the Au surface
(n = 2 for Cu2+ reduction to Cu0 and reconstructions in electrolyte.
n = 1 for Br− oxidation to 1/2 Br2 on UHV preparation of crystal surfaces
the ring electrode), and Q is the charge is time consuming and can be difficult
corresponding to monolayer formation for some metals. Furthermore, for more
of ad-atoms on Pt(111) based on the reactive metals than Au, it is transfer
surface atomic density of Pt(111)-(1 × of the crystal after surface preparation
1) (1.53 × 1015 atoms cm−2 ), assuming that is vital, as it is important to mini-
one completely discharged ad-atom per mize the interaction of the surface with
platinum atom. (e.g. 208 µC cm−2 for the atmosphere. The development of the
Brad and 480 µC cm−2 for Cuupd . By flame-annealing method by Clavilier and
combining the powerful in situ X-ray coworkers [18] has become of critical im-
scattering technique with the elegant portance in surface electrochemistry. The
RRDE method, remarkable insight into the hydrogen flame-annealing method has
structural nature of the Pt(hkl)-UPD-anion been developed to incorporate a cooling
electrochemical interface can be obtained procedure whereby the crystal is cooled in
as is demonstrated in Sect. 4.1.4. a hydrogen or argon atmosphere before a
drop of electrolyte is placed on the surface
4.1.2.3 Surface Preparation for transfer to the electrochemical cell. By
A key aspect in the study of single-crystal using purified gases, the contact with oxy-
metal electrodes is the preparation of the gen is minimized during preparation and
surface prior to the experiment and the this procedure has been used successfully
transfer of the crystal into the electro- for the preparation of Pt(hkl) surfaces.
chemical environment. One approach is to Unfortunately other metal surfaces,
prepare the surface in an ultrahigh vacuum such as Cu, Ag, and Ni, cannot be prepared
(UHV) environment by cycles of ion sput- by flame-annealing methods. The most
tering and annealing. This methodology successful method for these surfaces has
has the advantage that the surface qual- been to use a long preanneal of the
ity can be checked during preparation by surface in a forming gas, such as hydrogen,
standard surface science techniques such followed by a short electrochemical etch,
as LEED and Auger electron spectroscopy. rinsing with water and then direct transfer
UHV preparation methods have been par- into the electrochemical cell [19–21]. The
ticularly useful in the study of Au(hkl) exact conditions for the etching procedure
electrodes as the surfaces are relatively in- vary for different metals and even for
ert and the crystal can be transferred in different surface orientations of the same
air from the UHV system to the electro- metal. For the results included in this
chemical cell. An alternative method for report, readers are referred to the reference
material for precise descriptions of the of the fcc metals, Pt, Ir, and Au, have
surface preparation that was followed in all been shown to reconstruct under cer-
each case. tain conditions of sample temperature and
surface preparation [24]. Given that the
4.1.3 hexagonal (111) lattice planes form the
Surface Structures most dense surfaces they have the least ten-
dency to reconstruct and only the Au(111)
Modern surface crystallographic studies surface exhibits a reconstruction at room
have shown that on the atomic scale, temperature. The reconstruction is rather
most clean metals tend to minimize complex and involves a small increase in
their surface energy by two kinds of
the surface density (23 surface atoms in-
surface atom rearrangements – relaxation
stead of 22), which leads to a large unit cell,
and reconstruction [22–26]. In this review, √
the (23 × 3) structure, which is known
the term surface reconstruction applies to
as the ‘‘chevron’’ or ‘‘herringbone’’ recon-
the case in which there is lateral (i.e. in the
struction [27.] In contrast, the most open
surface plane) movement of surface atoms
(110) surfaces have a strong tendency to
such that the surface layer has a symmetry
reconstruct, exhibiting (1 × n) periodici-
that is different from that of the underlying
bulk of the crystal. Hence, the surface ties [24]. The (110) surfaces of Pt, Ir, and
layer has a two-dimensional unit cell Au usually reconstruct to form a (1 × 2)
that is different from the corresponding or (1 × 3) periodicity. After some initial
two-dimensional unit cell of a layer in controversy, it is now well established
the bulk. The periodicity of the surface that the (1 × 2) structure is characteris-
can be defined by Woods’ notation; tic of clean surfaces. These structures are
for example, an unreconstructed surface called ‘‘missing-row’’ structures since ev-
would be termed as (1 × 1), whereas if ery other row in the topmost atomic layer
the surface unit cell size was doubled is missing. Hence, in the transition from
in one of the primary vector directions, the (1 × 1) to the (1 × 2) state, there is a
it would be termed as (2 × 1), and so 50% decrease in the density of the sur-
on. On the other hand, surface relaxation face layer and significant mass transport
applies to the case in which the surface must occur. Theoretical calculations have
layer is in a (1 × 1) state but the layer indicated that the (1 × 2) reconstruction is
is displaced along the surface normal stabilized on the 5d transition metal (110)
direction from the position expected for surfaces due to the participation of the d
bulk termination of the crystal lattice. In electrons in bonding in the solid and the
this section, both surface reconstruction decrease of the kinetic energy of delocal-
and surface relaxation effects are described ized electrons at the surface [28]. For Au
with specific examples chosen to illustrate and Pt, contaminants seem to play the
the phenomena as they are observed in the dominant role in the appearance of the
electrochemical environment. (1 × 3) structure, that is, it seems that the
(1 × 3) structures obtained in UHV are
4.1.3.1 Surface Reconstruction impurity-stabilized [29].
Perhaps the most intriguing recon-
4.1.3.1.1 Gas–solid Interface Under struction occurs on the (001) surfaces.
UHV conditions, the clean low-index faces For example, under UHV conditions, the
Au(001) surface exhibits a hexagonal re- due to the difficulty in preparing clean sur-
construction, that consists of a single, faces and transferring them to the electro-
buckled, slightly distorted overlayer that chemical environment (see Sect. 4.1.2.3).
is aligned close to the [110] bulk direc- Ex situ techniques, whereby the electrode
tion and is often referred to as a ‘‘5 × 20’’ is emersed from solution and transferred
reconstruction although, strictly speaking, back into UHV for surface analysis, can-
the overlayer is incommensurate [30, 31]. not give definitive results due to the loss of
Theoretical studies of the Au(001) sur- potential control and/or solution species.
face suggest that the reconstruction forms The advances in the application of in situ
because the energy gained from the forma- methods, for example, scanning tunneling
tion of the more compact reconstructed microscopy (STM) [17, 21] and X-ray scat-
top layer is greater than the energy lost tering, have alleviated this ‘‘emersion gap’’
as a result of the loss of registry with problem, and provided definitive structure
the substrate. The energy difference be- determination of metal surfaces in elec-
tween the relaxed (1 × 1) surface and the trolyte and under potential control.
reconstructed surface is calculated to be Studies of the reconstruction of (111)
∼0.2 eV atom−1 [32]. X-ray measurements metal surfaces date back to the early
of the specular CTR at room temperature 1970s, when it was reported that UHV-
have shown that the reconstructed Au layer prepared Pt(111)-(1 × 1) surface remained
is 1.25 times denser than a bulk Au mono- intact after contact with solution [33, 34].
layer and that it is expanded away from More recently systematic, in situ X-ray scat-
the expected bulk position by ∼20% of tering [35] and STM [21, 28] studies have
the bulk lattice spacing [30]. Grazing inci- demonstrated that well-ordered P(111)-
dence X-ray studies have shown that the (1 × 1) structures, prepared either in UHV
overlayer is aligned approximately with the or by the flame-annealing procedure, are
[110] direction although the detailed in- stable in aqueous solutions. As for Pt(111),
plane structure is dependent on the rate the (111) surfaces of Ag and Cu exhibit
of cooling of the surface [31]. For slowly a (1 × 1) termination in the absence of
cooled samples, the overlayer is almost ex- any strongly adsorbing species, such as
clusively rotated by ±0.8◦ from the [110] anion adsorption or UPD metals. In con-
direction with a very small fraction of the trast, the Au(111) single-crystal surface
domains unrotated. The in-plane surface exhibits a potential-dependent reconstruc-
diffraction pattern for this structure cor- tion that is unique among the fcc metal
responds to a hexagon of spots centered surfaces. As in UHV, the surface forms
around the origin of reciprocal space as √
the (23 × 3) structure at negative elec-
shown in Fig. 5. In the [110] direction, trode potentials whereas at positive poten-
the peaks are centered at (1.206, 1.206, 0). tials the unreconstructed (1 × 1) surface
The corresponding real-space structure is is observed. The driving force for the
shown schematically in the lower part of √
(1 × 1) ↔ (23 × 3) phase transition was
the figure. proposed to be due to an induced surface
charge density of 0.07 ± 0.02 electrons per
4.1.3.1.2 Solid–solution Interface Estab- atom, that is, it was postulated that the
lishing the relationship between a surface formation of the surface reconstruction is
structure in electrolyte and that observed in driven by the induced surface charge [36.]
UHV has always been problematic, largely Alternatively, it has been proposed that
d = 0.206 (1, 0)h

(1, 1, 0)c
d
1.206
d
(0, 0, 0) (1, 1, 0)c
(0, 1)h
(a) (b)
1.6°
1.6°
1.6°
a g
b
(c) (d)
Fig. 5 The in-plane diffraction pattern for the the underlying Au(001) face with square
Au(001) surface. (a) The unreconstructed phase; symmetry and the open symbols the scattering
(b) the distorted hexagonal phase; (c) the due to the hexagonal surface reconstruction.
rotated, distorted hexagonal phase. In each case, (d) A real-space schematic picture of a rotated
the filled symbols represent the scattering from domain of the reconstruction.
adsorption/desorption of solution species the real driving force for the potential-
is responsible for the changes in the sur- induced reconstruction of Au(111) is. At
face structure [37]. Considering that the present, we consider the issue to be un-
adsorption of ions from supporting elec- resolved; for further details the readers
trolyte is closely related to the charge are refered to Kolb’s review on recon-
redistribution at the interface, it is very struction phenomena at metal–electrolyte
difficult to conclude unambiguously what interfaces [17].
The structural behavior of the (001) sur- the reciprocal lattice positions indicated
faces in aqueous electrolytes has been in Fig. 5. Figure 6(c) shows a scan through
more controversial. For example, the first the (1, 1, 0.4) position that corresponds
ex situ experiments indicated that the to a point on the nonspecular (1, 1,
Pt(001) – ‘‘5 × 20’’ structure transforms l) CTR. This scan is representative of
into the (1 × 1) phase upon contact with the in-plane mosaic spread of the crys-
aqueous solution or even when in con- tal surface, which is ∼0.5◦ full-width at
tact with water vapor [38]. In contrast, half-maximum (FWHM). Figures 6(a, b)
Zei and coworkers found that the UHV- show scans at the (1.206, 1.206, 0.4) po-
prepared reconstructed Pt(001) surface sition in 0.1 M KOH (−0.35 V) and 0.1 M
remained stable in contact with elec- HClO4 (−0.05 V) solutions, respectively;
trolyte, although no convincing evidence of the electrode potential chosen so that the
potential-induced reconstruction was ever surface is in the reconstructed state. In
detected [39]. X-ray scattering measure- 0.1 M KOH, the scan shows a single broad
ments have unambiguously demonstrated peak aligned with the Au(110) direction. In
that the Pt(001)-(1 × 1) surface is stable in 0.1 M HClO4 , however, there is no longer
the potential range between the adsorp- a central peak but the scan displays two
tion of hydrogen and hydroxyl species and peaks rotated by ∼0.75◦ from the [110]
the reconstructed surface was never ob-
direction. These results indicate good gen-
served [40]. In contrast with Pt(001), the
eral agreement with UHV studies of the
Au(001) surface in electrolyte exhibits a
Au(001) surface [30, 31], although the ro-
potential-induced reconstruction that is
tated peaks observed in 0.1 M HClO4 are
very similar to the temperature-induced
unaccompanied by the strong central peak
reconstruction in UHV [7]. To illustrate
observed in the UHV studies, whereas in
the general phenomenon of surface re-
0.1 M KOH no rotated domains are ob-
construction in electrolyte, representative
served at all. We should note that the
results for this system are now presented
to demonstrate the power of the X-ray results shown in Fig. 6 were obtained after
diffraction technique in elucidating the numerous cycles of the electrode potential
potential-induced structural transforma- in both electrolytes during which the re-
tion of the Au(001) surface. construction was lifted and reformed and
The potential stability of the gold sur- so the results shown are representative of
face reconstruction in the electrochemical true equilibrium conditions.
environment has been studied by CV, Figure 7 shows some XRV measure-
ex situ emersion LEED experiments [17], ments, which demonstrate the potential
STM, and X-ray diffraction [7]. In both range of stability of the ‘‘5 × 20’’ re-
acid and alkaline solutions, there is good construction in the different electrolytes.
agreement between the experimental tech- Figure 7(a) shows the intensity depen-
niques that the ‘‘5 × 20’’ reconstruction dence at the position of one of the rotated
(see Fig. 5) is formed at cathodic poten- peaks in 0.1 M HClO4 , Fig. 7(b) shows
tials and that it can be reversibly lifted the intensity dependence at the unrotated
and formed upon cycling the applied peak position in 0.1 M KOH, and Fig. 7(c)
potential anodically. Figure 6 shows repre- shows data for the same solution obtained
sentative in-plane X-ray diffraction results at (0, 0, 1.3), a position close to the anti-
in the form of rocking scans through Bragg position along the specular CTR
0.1 M KOH
0.16
0.12
(a)
0.12
0.1 M HClO4
0.11
Intensity
[a.u.]
0.10
0.09
(b)
0.30
(1, 1, 0.4)
0.20
0.10
−3 −2 −1 0 1 2 3
∆f
(c) [degrees]
Fig. 6 Rocking scans through: (a) (1.206, 1.206, 0.4) in 0.1 M KOH (at
E = −0.35 V); (b) (1.206, 1,206, 0.4) in 0.1 M HClO4 (E = −0.05 V);
and (c) (1, 1, 0.4), a CTR position for a Au(001) electrode. These
positions are illustrated in Fig. 5. The closed symbols in (a) show the
measured profile in both solutions after the lifting of the reconstruction
at positive potentials.
in which the reconstruction of the sur- Because of the slow kinetics in the for-
face causes an increase in the scattered mation of the reconstructed surface, at
intensity due to the increased density of the sweep rate used in these measure-
the surface layer. The results show the ments, 2.5 mV s−1 , there is considerable
reversible lifting and formation of the re- hysteresis as the potential is cycled. Pre-
construction in both electrolyte solutions. vious kinetic studies [8] have shown that
0.19
0.1 M HClO4
0.18
0.17
0.16
(a)
0.15
0.14
0.1 M KOH
Intensity
[a.u.]
0.12
0.10
(b)
(0, 0, 1.3)
0.40
0.20
0.0 0.4 0.8 1.2

E
(c) [V]
Fig. 7 XRV measurements at: (a) the rotated peak position in 0.1 M HClO4 (see
Fig. 6b); (b) (1.206, 1.206, 0.4) in 0.1 M KOH; and (c) at (0, 0, 1.3), a position on
the specular CTR, in the same solution as (b).
the reconstruction forms over a timescale same hysteresis implies that the kinetics
of 15–100 s, which is consistent with the are controlled by the large-scale movement
data shown in Fig. 7. The fact that the data of the Au surface atoms rather than any
at the CTR position, Fig. 7(c), shows the local ordering effects. As with the Au(111)
surface, it is difficult to propose an unique of a (1 × 2) reconstruction were observed.

driving force for the surface reconstruction Figures 8(a, b) show longitudinal (along
although it is interesting that the details of [0 k 0]) and rocking scans (approximately
the reconstructed state (i.e. rotated vs. un- along [h 0 0]) through the (0, 1.5, 0.1) super-
rotated domains) appear to depend on the lattice Bragg reflection. The solid lines are
electrolyte. fits of a Lorentzian lineshape to the data,
The first insight into the structure of which allow the correlation length of the
the (110) surfaces in electrolyte was also (1 × 2) unit cell to be derived. From these
obtained from ex situ studies [17]. For ex- measurements, and from similar scans
ample, the UHV-prepared Pt(110)-(1 × 2) through other superlattice reflections, a
surface was found to be stable upon con- correlation length of ∼350 Å along the
tact with several solutions, if the potential [1 0 0] direction and ∼250 Å along the
cycling was restricted to certain potential [0 1 0] direction is calculated, in which
regions. More recently, the stability and these correspond to the parallel and per-
structure of the Pt(110) surface in sev- pendicular directions to the rows in the
eral electrolytes has been examined by in missing-row model of the (1 × 2) recon-
situ X-ray diffraction [41, 42]. The follow- struction. The larger correlation length
ing results were obtained in 0.1 M NaOH corresponds to the limit of the instru-
solution, chosen so as to avoid any strongly mental resolution and so this only puts
adsorbing anions. The potential was cy- a lower limit on the domain size. Interest-
cled over the range 0.1–0.7, V, which ingly, there is no evidence of any shifts in
corresponds to the hydrogen adsorption the (0, 1.5) or (0, 2.5) in-plane reflections
region (<0.3 V) and the region in which re- along the [0 1 0] direction, in contrast with
versible adsorption of OH species occurs. the previously reported measurements of
Over this entire potential range, diffrac- the UHV-prepared surface [43]. This shift
tion peaks consistent with the presence was linked to the presence of randomly
3.0
2.5
Intensity
[a.u.]
2.0
1.5
1.0
1.48 1.52 −0.4 0.0 0.4
(h, 0, 0.1) ∆f
(a) (b) [degrees]
Fig. 8 The measured X-ray intensity at (0, 1.5, 0.1): (a) along the [0 1 0] direction; and
(b) along the [1 0 0] direction for a Pt(110) electrode in 0.1 M NaOH (E = 0.1 V). The
solid lines are fits of a Lorentzian lineshape to the data.
distributed single-height steps on the sur- the fact that the phase transition was ob-
face. The Lorentzian lineshape does not served at the pzc, it was proposed that the
perfectly produce the data in the tails of the structural transformation of the Au(110)
Bragg reflection (Fig. 8a) and this could be surface was governed by the specific (con-
due to a residual step distribution. It is tact) adsorption of anions. In contrast, in
important to emphasize that the (1 × 2) an X-ray diffraction experiment, a (1 × 3)
surface reconstruction was stable over a reconstruction was observed in salt solu-
wide potential range, for example, even at tions (0.1 M NaCl) [46]. As for Au(111),
1.2 V there is still a strong (110)-(1 × 2) Ocko and coworkers postulated that the
diffraction pattern. Upon sweeping the formation of the surface reconstruction
potential negatively, however, the (1 × 2) was driven by the induced negative sur-
reconstruction is finally lifted as the surface charge. It was also suggested that
face atoms move to accommodate the the (1 × 3) reconstruction was stabilized
oxide reduction. It should also be noted over the (1 × 2) structure due to the neg-
that the Pt(110)-(1 × 2) structure was so ative surface charge, by analogy to UHV
stable in aqueous electrolytes that ad- results showing that Ag(110) underwent
sorption of CO on this surface did not a (1 × 1) → (1 × 2) transition after the
induce the (1 × 2) ↔ (1 × 1) transition adsorption of K on the surface [47]. As
that is observed in UHV upon adsorp- for the other Au surfaces, it is very diffi-
tion of CO [44]. The high stability of cult to resolve unambiguously the driving
the Pt(110)-(1 × 2) structure implies that force for the potential-induced structural
the potential-induced mobility of Pt sur- transformation, especially given that sur-
face atoms at the solid–liquid interface face charge and anion adsorption are so
is significantly reduced compared to the strongly correlated.
temperature-induced mobility in UHV.
In contrast with Pt(110), the reconstruc- 4.1.3.2 Surface Relaxation
tion at the Au(110)–liquid interface can The relaxation of metal surfaces under
be reversibly lifted and reformed, im- UHV conditions has been well studied
plying that the mobility of gold surface by utilizing LEED I –V analysis and a
atoms is appreciable even in contact with significant database now exists [23]. For
electrolyte. Depending on the nature of unreconstructed low-index single-crystal
the supporting electrolyte, however, either surfaces, it is often found that the
the (1 × 2) or (1 × 3) reconstruction was outermost layer of atoms is contracted
found to be stable at negative potentials. In toward the second layer, compared to the
particular, Magnussen and coworkers [45] bulk layer spacing. Adsorption on the clean
demonstrated that in perchloric acid the surface in UHV, however, usually reverses
(1 × 2) phase was stable at the potential this trend and results in the outermost
of zero charge (denoted hereafter as the layer of metal atoms being expanded away
pzc). Positive of the pzc, the reconstruction from the second layer. The experimental
is lifted and the Au(110)-(1 × 1) structure data has led to interesting theoretical
is clearly observed in STM experiments. work aimed at establishing a fundamental
The observed structural transformation understanding of surface relaxation [48].
is highly reversible, the (1 × 2) structure The same concepts can be applied to the
being fully restored when the applied po- M(hkl)–electrolyte interface, and indeed
tential is negative of the pzc. Because of relaxation has also been observed at the
Ag(hkl), Au(hkl), and Pt(hkl) surfaces in states of the Au surface atoms at the Fermi
electrochemical environments [40, 46, 49]. level.
As opposed to UHV systems, in the Of the three low-index Pt surfaces that
electrochemical environment, it is not we have studied, the Pt(110)-(1 × 2) has
possible to perform LEED experiments the most dramatic dependence of the sur-
and surface relaxation effects can only be face relaxation on the applied electrode
probed using X-ray diffraction techniques. potential [41, 42]. In the following discus-
As shown in Sect. 4.1.2.1, the analysis sion, the surface relaxation effects are
of CTR data is relatively simple to interpreted as being caused by the spe-
perform and can provide accurate values cific nature of the adsorption sites on the
of surface relaxation in both the UHV Pt(110)-(1 × 2) surface and the strength of
environment and when the surface is in the adsorbate–metal bonding, the two of
contact with the electrolyte. Additionally, course being interrelated. The results were
the potential dependence of the surface obtained in alkaline solution (Fig. 9), in
relaxation can be probed by performing which the surface relaxation is attributed
XRV measurements at suitable reciprocal to the adsorption of Hupd and OHad [41].
lattice positions (see Sect. 4.1.2.1). The To confirm the dominant role of these ad-
results can give good insight into the sorption processes, the charge associated
nature of the interaction between species with hydrogen adsorption and hydroxyl ad-
that can be adsorbed from the electrolyte sorption is also shown in Fig. 9. As shown
solution and the metal surface. in the previous section, the (1 × 2) surface
Toney and coworkers used X-ray scat- reconstruction was stable over the entire
tering to probe the distribution of water potential range. However, changes in the
molecules at the Ag(111)–electrolyte in- diffracted intensities were observed, par-
terface [49, 50]. In addition to observing a ticularly along the CTRs. Such changes,
large increase in the density of the wa- in this case at (0, 0, 1.55) on the specu-
ter layer adjacent to the Ag surface, they lar CTR, are shown in the lower part of
also observed an electrorestrictive effect Fig. 9. The changes in intensity indicate a
in which the topmost Ag atomic layer potential dependence of the relaxation at
was contracted into the surface (by 0.03 Å) the Pt(110) surface. Measurements at (0,
relative to the bulk Ag(111) spacing by step- 0, 2.45) showed the opposite behavior to
ping from −0.23 to 0.52 V (vs. an Ag/AgCl those at (0, 0, 1.55), that is, the curve was
reference electrode) in 0.002 M NaF solu- inverted. These results suggest that addi-
tion. This effect was obtained by modeling tional expansion of the surface is caused
of nonspecular CTR data as the scatter- both by hydrogen adsorption and hydroxyl
ing contained no contribution from the adsorption. As discussed in Sect. 4.1.2.1.1,
incommensurate water adlayers. In a re- relaxation in the surface normal direction
cent X-ray scattering study [51], Lucas and induced by Hupd at 0.05 V can be probed
coworkers have used similar techniques by modeling CTR data. The specular CTR,
to probe the relaxation of the Au(111) (0, 0, l) and nonspecular CTR, (0, 1, l),
surface in its unreconstructed state. The data are shown in Fig. 10. In order to
results have been correlated with surface model the data, the structural parameters
stress measurements and density func- derived by Vlieg and coworkers [52] for the
tional theory (DFT) calculations that show Pt(110)-(1 × 2) surface in UHV were used,
corresponding changes in the density of allowing the vertical displacements in the
200
Change density
100
[µc cm−2]
0
10 µA
QH
QOH
(a)
Side view
5.2
(0, 0, 1.55)
5.0
X X
Intensity
4.8
[a.u.]
4.6
4.4
Top view
4.2
0.0 0.2 0.4 0.6 0.8 1.0
E
(b) [V]
Fig. 9 Left: A schematic illustration of the Pt(110)-(1 × 2) surface showing the
directions of the atomic relaxations derived from fits to the X-ray scattering data. In
the top view, the letter X indicates the most likely site for hydrogen adsorption.
Right: (a) The CV of a Pt(110) disk electrode measured in an RRDE configuration
(20 mV s−1 ). The symbols correspond to the charge associated with hydrogen and
hydroxyl adsorption. (b) The XRV measured at (0, 0, 1.55) for the positive
(symbols) and negative (dashed line) sweep directions (2 mV s−1 ).
top two Pt layers, the buckling in the third The magnitude of the relaxations derived
layer and an increased Debye–Waller fac- from the best fit are as follows: 0.35 Å ex-
tor (or roughness) of the first and second pansion in the topmost layer (25% of the
layers to vary. The best fit to the data (ob- layer spacing), an inward displacement of
tained by a least-squares method) is shown 0.08 Å in the second layer (6%), a buckling
by the solid lines in Fig. 10. The dashed of 0.08 Å in the third layer, and root-mean-
line represents a calculation of the CTRs square roughness of 0.41 and 0.16 Å in the
according to the structural model proposed first and second layers, respectively (the
by Vlieg and coworkers [52]. The biggest bulk thermal roughness is 0.05 Å). The
discrepancy between the calculation for relaxation directions are indicated in the
the model derived in previous work and schematic model of the structure in Fig. 9.
the measured data occurs at the anti-Bragg The most striking difference between
position on the specular CTR. It was only the structural model for the Pt(110)-
possible to reproduce the data by including (1 × 2) reconstruction in electrolyte and
an expansion of the topmost layer spacing. the model derived from UHV studies is
10−1
Intensity
[a.u.]
10−2
10−3
(a)
10−1
Intensity
[a.u.]
10−2
10−3
0.5 1.0 1.5 2.0 2.5

l
(b) [rlu]
Fig. 10 The measured CTR data: (a) (0, 0, l); and (b) (1, 0, l) for the
Pt(110) electrode at an electrode potential of 0.1 V. The dashed lines are
calculated for a contracted (1 × 2) missing-row model that was found in
UHV studies [52] and the solid lines are fits to the data indicating a 20%
expansion of the surface [41].
in the relaxation of the topmost Pt atom missing row with a resultant electrostatic
in the unit cell. The contraction of the force that pulls the ion cores into the
surface observed in vacuum studies has surface [53]. In electrolyte, the situation
been explained on the basis of electrostatic is inherently more complex because of the
considerations in which electrons are presence of water, ions, and the electric
transferred from the top atom towards the field at the interface. In addition, the data
described above were taken in a potential selected examples of coadsorption between

region in which ∼1 monolayer of hydrogen metal UPD layers and anions.
is adsorbed onto the surface. Theoretical
studies of hydrogen adsorption onto Pt 4.1.4.1 Anion Adsorption
surfaces have indicated that the most The adsorption of anions on metal elec-
likely site for adsorption is the threefold trodes has been one of the major topics
coordinated site (indicated by the letter X in surface electrochemistry. Anion ad-
in the top view of the surface in Fig. 9), sorption has an important, and generally
with the d orbitals of the neighboring adverse, effect on the kinetics of electro-
Pt atoms making a major contribution chemical reactions and thus needs to be
to the bonding. Full occupation of these
understood in some detail. Most of the
sites agrees with the hydrogen coverage
recent progress has come from the com-
measured electrochemically. Interestingly,
bination of conventional electrochemical
adsorption of hydrogen into these sites
methods with ex situ and in situ surface-
was proposed as the mechanism for
sensitive probes. Of particular interest are
an increased corrugation of the Pt(110)
chemisorbed (also called contact adsorbed
surface in UHV, in which a 20% expansion
or specifically adsorbed) anions, whose ad-
was observed by He diffraction [54].
sorption is controlled by both electronic
and chemical forces. In this report, we aim
4.1.4
to highlight some of the general themes of
Adsorbate-modified Surfaces
this work by focusing on the adsorption of
bromide anions onto Au and Pt electrodes.
It is well established that the adsorption
For the Au(001) and Au(111) electrodes,
of atoms and/or molecules on clean
the focus is on the structural transitions
metal surfaces has a dramatic effect on
that are observed as a result of atomic size
the surface structure. The presence of
effects. For Pt(111) and Pt(001), RRDE data
the adsorbed layer usually alters the
are combined with X-ray scattering mea-
metal surface relaxation (see previous
surements to correlate structural changes
section) and can cause clean metal surfaces
with the potential-dependent surface cov-
either to reconstruct or, more commonly,
erage of the anion species.
to deconstruct, that is, return from a
reconstructed state to the (1 × 1) phase.
The thermodynamic driving force for 4.1.4.1.1 Au(111)/Br X-ray diffraction
adsorbate-induced restructuring is the and STM studies of the adsorption
formation of strong adsorbate–substrate of bromide (and other halides) onto
bonds that are comparable to or stronger Au single-crystal electrodes have been
than the bonds between the substrate reported in a series of papers from the
atoms. Within the framework of this Brookhaven group [55–57]. Figure 11(a)
article, we present a selected review shows a CV for a Au(111) electrode
of some recent studies of adsorption in deaerated 0.1 M HClO4 + 0.1 M NaBr
phenomena on single-crystal transition solution. As discussed by Kolb [9],
metals. The section is split into two parts: the peaks at −0.12 V (anodic sweep)
the first dealing with the adsorption of and −0.18 V (cathodic sweep) are
anions, in particular bromide, and the related to the lifting/forming of the
second describes metal UPD and gives Au surface reconstruction, which is
2 E0
1
Current
[µA]
−1
−2
−3
−0.4 −0.2 0.0 0.2 0.4 0.6 0.8

(a) EAg/AgCl
[V]
(0, 1)
<T, 1>
(0, 0) (i) (1, 1) (X, X)

(ii)
(1, 0) (3X, 0)
<1, 1>
(b)
4.25
4.20
Br–Br spacing OBr
4.15
[Å]
4.10
4.05
(c) 4.00
Fig. 11 (a) The CV for a Au electrode in 0.1 M HClO4 + 0.1 M NaBr. (b) The
in-plane X-ray diffraction pattern observed at potentials positive of E0 .
Positions A, B, and C correspond to the first-, second-, and third-order
diffraction peaks from the bromide adlattice, respectively (filled circles). Also
shown is a schematic illustration of the corresponding bromide structure.
(c) The bromide–bromide nearest-neighbor spacing (obtained from the X-ray
measurements) as a function of the applied potential and solution
concentration; 0.1 M (circles), 0.033 M (squares), 0.01 M (triangles), and
0.001 M (diamonds). The lines are polynomial fits to the data. (This figure is
taken from Ref. [55].)
induced by a small amount of bromide with the change in the bromide–bromide

adsorption/desorption. The major change nearest-neighbor spacing that is shown
in the bromide surface coverage occurs in Fig. 11(c) for four different solution
in the potential range −0.2 to +0.2 V, concentrations of NaBr. At the potential at
that is, from the reconstruction peak which the adlayer is formed (0.42 V), the
across the broad shoulder feature. At Br–Br spacing is 4.24 Å and it decreases
higher potentials, there is a small down to a saturation value of 4.03 Å as
current associated with additional specific the potential is increased. This saturation
adsorption of bromide. At the position value is close to the van der Waals diameter
marked by E0 (0.42 V), there is a of bromide (3.70–4.00 Å). The fact that the
small reversible peak, which is correlated electrocompression is continuous and no
with an order–disorder transition in discontinuities are observed in Fig. 11(c)
the bromide adlayer. The ordering in implies that the adlayer does not lock-in to
the bromide adlayer was probed by any high-order commensurate structures
X-ray diffraction measurements [55]. At as this would give rise to a fixed Br–Br
potentials above 0.2 V, the Au surface is spacing in a certain potential range. Fits
unreconstructed, however, no scattering to the 0.1 M NaBr data to a third-order
from an ordered bromide structure was polynomial function are shown by the
detected when the potential was below solid line in Fig. 11(c). The dashed lines
E0 . When the potential was raised above were obtained by successive potential
E0 , the in-plane diffraction pattern shown shifts of ∼65 mV per logarithm of NaBr
schematically in Fig. 11(b) was obtained. concentration. Such a shift is consistent
This pattern exhibits six hexagonally with the 59 mV shift expected for a
arranged pairs of diffraction peaks at the nominal charge transfer of one electron
first-, second-, and third-order positions from the bromide to the Au electrode, that
(marked by A, B, and C, respectively, is, the electrosorption valency, γ = −1.
in the figure). Rocking scans at these
positions showed the presence of two 4.1.4.1.2 Au(001)/Br Unlike the contin-
rotated peaks (i.e. equidistant from the uous electrocompression of the bromide
principal Au reciprocal lattice vector, as for adlayer observed on Au(111), the bro-
the surface reconstruction of Au(001) in mide adlayer on Au(001), a substrate
alkaline electrolyte – Sect. 4.1.3.1.2). This with square symmetry, has been ob-
diffraction pattern indicates the presence served to undergo a commensurate-
of a rotated-hexagonal adlayer structure, incommensurate transition in which a
√ √
shown schematically in Fig. 11(b). commensurate c( 2 × 2)R45◦ structure
The rotation angle and the radial peak transforms continuously to an incommen-
√
positions for the bromide structure depend surate c( 2 × p)R45◦ structure [57]. The
on the applied electrode potential over the CV for Au(001) in 0.05 M NaBr is shown
range of which the structure was stable. In in Fig. 12. Three sharp peaks (labeled as
fact, the rotation angle ranged from ±3.2◦ P1, P2, and P3) are observed as the po-
to ±4.7◦ , whereas the position of the first- tential is swept between −0.25 and 0.6 V;
order peak varied from 0.683 to 0.716 [in P1 corresponds to the lifting of the Au
units of Au(1,0,0)] as the potential was reconstruction to leave the surface in the
changed from 0.42 to 0.76 V. The change (1 × 1) state, P2 corresponds to the forma-
√ √
in the radial peak position can be correlated tion of the c( 2 × 2)R45◦ structure (as
15 0.10
P1
0.08
10
0.06
c(√2 × p)
c(√2 × 2√2)
[µA cm−2]
0.04
Current
Intensity
[a.u.]
5
P2 0.02
P3
0.00
0
P3′
P2′
−5
−0.3 0.0 0.3 0.6

ESCE
[V]
Fig. 12 The CV for a Au(001) electrode in 0.05 M NaBr solution and the
corresponding X-ray intensity at the position (0, 1, 0.1) in which scattering from a
bromide adlayer was observed. (Taken from Ref. [57].)
K
(2, 2)
(c) H
2√2a pa
K (2, 2)
√2a √2a
(a) (b) (d) H
√ √
Fig. 13 Real-space model (a, b) of the commensurate c( 2 × 2)R45◦ and
√ ◦
incommensurate c( 2 × p)R45 bromide structures observed on Au(001). The
open circles correspond to bromide atoms. (c, d) show the corresponding
in-plane reciprocal space patterns in which the squares are scattering from the
Au substrate and the circles are Br reflections. In (d), the peaks move outward
along K with increasing potential. (Taken from Ref. [57].)
shown in Fig. 12 by the intensity changes decreases from 4.56 Å (ε = 0, commen-

at the (1, 0, 0.1) reciprocal space posi- surate) to 4.14 Å (ε = 0.13). This latter
tion which is where scattering from such a spacing is close to the minimum value
structure would arise), and P3 corresponds observed on the Au(111) surface (4.02 Å).
to the commensurate-incommensurate
√ √
phase transition. The c( 2 × 2)R45◦
4.1.4.1.3 Pt(111)/Br In contrast to anion
structure is shown schematically in
adsorption onto gold electrodes, where a
Fig. 13(a), which indicates that the surface
stable double-layer region occurs over a
coverage by bromide is θ = 1/2 and that
wide potential range, on platinum surfaces
the structure is close to an hexagonal ar-
there is always strong competition between
rangement, despite the square symmetry
anions and either H or OH adsorption.
of the underlying substrate. This implies
This competition tends to inhibit the
that the elastic interactions between the
formation of ordered anion structures and
relatively large Br ad-atoms (which would
can make the adlayers difficult to observe
favor hexagonal packing) dominate over
by X-ray scattering measurements. The
the adsorbate–substrate interaction. The
interaction of bromide ions with Pt(hkl)
observed in-plane diffraction pattern is
surfaces has been studied extensively at the
shown in Fig. 13(c) in which the squares
solid–electrolyte interface, perhaps even
correspond to substrate reflections and the
more so than at the vacuum interface.
circles to Br reflections. A similar Br struc-
The modern in situ structural probes of
ture was also observed for vapor-deposited
STM and X-ray diffraction have played a
Br on Au(001) [58].
significant role in understanding bromide
At potentials positive of P3 (0.38 V), the
ion adsorption on Pt(hkl) surfaces. What
bromide adlayer undergoes a commensu-
we will emphasize in this review is the
rate-incommensurate transition that was
importance of the surface geometry in
signified by the movement of the low-
the Pt-bromide energetics, and thus in the
order diffraction features continuously and
ordering of the adlayers.
uniaxially outward with increasing poten-
Quantitative measurements of the cov-
tial. As shown in the diffraction pattern
erage by Brad on Pt(111) surface were ob-
(Fig. 13d), reflections were only observed
tained by purely electrochemical methods,
at (1, 0.5 + ε/2), (0, 1 + ε), and (2, 1 + ε),
as described in detail in Ref. [59]. Briefly,
along with symmetry equivalents (ε de-
by utilizing the ring-shielding properties
notes the incommensurability). The real-
of the RRDE, Sect. 4.1.2.2, it was possi-
space model of the structure is shown in
ble to determine the potential-dependent
Fig. 13(b). For the domain shown, the bro-
surface coverage by bromide and its elec-
mide lattice is commensurate along H (the
√ trosorption valence (γ ) on Pt(111). The
2 direction) and incommensurate along
electrosorption valence
K. As the potential increases above 0.38 V,
the value of ε increases continuously and id∞
this corresponds to movement of the bro- γ = = 0.99 (11)
(ir∞ − irfs )/N
mide atoms along the directions shown
by the dashed lines in Fig. 13(b). In the is a measure of the degree of discharge of
incommensurate phase, the Br nearest- the ion upon adsorption, that is γ = 1 for
neighbor spacing is fixed at 4.078 Å, a univalent ion corresponds to complete
whereas the next-nearest-neighbor spacing discharge. Figure 14 shows that as the
Pt(111) + 8.10−5 M Br−

(900 rpm: 50 mV/s: 0.1 M HClO4)
Disk current
0
[µA]
0.4
10 0.3
[ML]
q Br
0.2
(a)
0.1
0.0
−0.2 0.0 0.2 0.4
(c) E
Ring current
[V]
+1.08 V ½ Br + e−
[µA]
Br− 2
ir∞
−0.2 0.0 0.2 0.4

E
(b) [V]
Fig. 14 RRDE results for a Pt(111) electrode in 0.1 M
HClO4 + 10−4 M Br− : (a) Disk current; (b) corresponding ring
current; and (c) the potential-dependent bromide coverage
obtained from the results using Eq. (9).
Pt(111) disk potential is swept across the manifesting the continuous adsorption of
first voltammetric peak, the associated ad- bromide on the Pt(111) disk. The qualita-
sorption of bromide is demonstrated by the tive correspondence between ring and disk
concomitant decrease of the ring current currents may be evaluated quantitatively in
below its unshielded value. The same ob- terms of θBr according to Eq. (9); to avoid
servation is made for the subsequent, more the mass transport resistances addressed
positive peak, which clearly relates this pro- in Sect. 4.1.2.2, the bromide adsorption
cess to further bromide adsorption on the isotherm is extracted from the negative-
Pt(111) disk. Following these two char- going sweep. The resulting %Br versus E
acteristic voltammetric peaks, the Pt(111) is shown in Fig. 14(c). As evident from
disk current diminishes to a double-layer- the shape of curve, the adsorption of Br
like structure above ≈0.2 V. At the same (γBr ≈ 1) is a two-step process, the first
time, the ring current remains below ir∞ being displacement of the Hupd state in
until the positive potential limit is reached, a narrow interval of potential, the second
a continuous compression until a close- as a function of the electrode potential

packed adlayer with a maximum surface indicated that the hexagonal adlayer un-
coverage of 0.44 Br per Pt surface atom derwent a continuous electrocompression.
is reached at ≈0.6 V. As demonstrated be- By repeating the Br− /Pt(111) experiments
low, combining the purely electrochemical a number of times, however, it became
measurements of coverage with the struc- apparent that the behavior of this struc-
tures observed by X-ray diffraction enables ture was more complex than originally
a reasonable physical picture of the adsorp- proposed [61]. Figure 15 is typical of the
tion process to emerge. results obtained, showing scans along (h,
In a preliminary report of the re- 0, 0.1) through the first-order diffraction
sults for bromide adsorption onto Pt(111), peak from the hexagonal bromide over-
it was shown that an incommensurate layer at different electrode potentials. At
‘‘(3 × 3)’’ hexagonal bromide adlayer is 0.2 V, there is a very strong peak due to
present on the Pt(111) surface in the po- the (3 × 3) bromide adlayer located exactly
tential range 0.05–0.7 V [60]. From fits at the commensurate (3 × 3) position (do-
to the scattering profile at the lowest- main size, D ∼ 100 Å) and an additional
order diffraction peak, the peak position weak peak at slightly higher wave vector
0.65 V
0.05
0.55 V
0.04
0.35 V
0.03
Intensity
[a.u.]
0.25 V
0.02
0.01
0.2 V
Fig. 15 X-ray diffraction scans along (h,
0, 0.1) for Br− /Pt(111) at different
electrode potentials. The data were
obtained by subtraction of a background 0.00
scan performed at −0.2 V. The solid
lines are fits to the data and the dashed
0.66 0.68 0.70 0.72
lines mark the positions h = 0.67 and
h = 0.71. (h, 0, 0.1)
(h = 0.68). Upon stepping the electrode Both of these unit cells are relatively small
potential to 0.25 V, the intensity distribu- (8.33 and 19.39 Å for the (3 × 3) and (7 × 7)
tion shows only a very weak double peak structures, respectively) and so the ad-
lineshape. At 0.35 V, the scattered inten- layer is well ordered. In between these two
sity is further reduced and it is almost phases, we note that HOC structures with
impossible to distinguish any measurable intermediate Br–Br atomic spacings must
feature from the background scattering. have much larger unit cells (>30 Å) [62].
Upon stepping to higher potential, a weak Given that the maximum domain size that
peak gradually emerged at h = 0.715, be- is observed is ∼60 Å, this could explain
coming stronger as the potential was raised the decreased peak intensity (or disorder)
to a positive potential limit of 0.65 V. In in the intermediate potential region as the
Fig. 15, rocking scans through the peak, at Br ad-atoms do not have enough mobility
any potential, showed the intensity to be to form well-ordered structures with large
centered along the Pt1, 0, 0 lattice vector unit cells.
and there was no evidence of any rota-
tion of the adlayer. Stepping the potential 4.1.4.1.4 Pt(001)/Br System As for the
to −0.2 V always caused the peak to dis- Pt(111)-Br system, the surface structure
appear, consistent with the voltammetry and potential-dependent surface coverages
results, which indicate complete desorp- by Br on Pt(001) were obtained from the
tion of bromide at this potential (see the combination of X-ray scattering and RRDE
isotherm in Fig. 14). measurements [63]. The effect of adsorbed
The isotherm for bromide adsorption bromide on the voltammetric features of
(Fig. 14c) shows that increasing the poten- Pt(001) recorded in 0.1 M HClO4 is easily
tial in the ‘‘double-layer’’ region causes a observed by comparing the voltammetry
continuous increase in the bromide cov- with and without Br− , as shown in Fig. 16.
erage. The data shown in Fig. 15 and in Close inspection of the voltammetry in
other experiments, however, are not con- Fig. 16(a) reveals that formation of a mono-
sistent with the notion that the additionally layer of hydrogen in the solution contain-
adsorbed bromide leads to a continuous ing bromide occurs through two distinctive
compression of the incommensurate bro- voltammetric features: a main sharp peak
mide adlayer, as this would cause the sharp at −0.1 V and a small peak at −0.175 V.
peak at h = 0.67 to shift gradually to higher These two peaks correspond to simul-
wave vector. An alternative explanation of taneous hydrogen adsorption/desorption
our results is that bromide forms a series with bromide desorption/adsorption on
of high-order commensurate (HOC) struc- (001) terrace sites and most probably on
tures on the Pt(111) surface, that is, at all (001) × (111) terrace-step sites, respec-
potentials the structure corresponds to a tively [63]. Further inspection of the RRDE
close-packed monolayer but that the peri- results reveals that, the major adsorption
odicity depends critically on the ratio of peak on Pt(001) is not followed with a
the bromide and Pt lattice parameters. At second sharp peak, as was found for the
0.2 V, the unit cell corresponds to a (3 × 3) adsorption of Br− on Pt(111). The qualita-
structure with a basis of four Br atoms. At tive correspondence between ring and disk
0.6 V, the unit cell is a (7 × 7) structure currents may be evaluated quantitatively
containing 25 Br atoms, that is, with a 7 : 5 in terms of θBr as shown in the previous
ratio of the Br and Pt lattice parameters. discussion for Br/Pt(111). The resulting
Pt(001) + ≈10−4 M Br−

(900 rpm; 50 mV/s; 0.1 M HClO4)
Disk current
0
[µA] 300
(c)
Qd
200
[µC cm−2]
20 Qr/N
Q
100
(a)
0
−0.2 0.0 0.2 0.4
2
Ring current
E
+1.08 V [V]
Br∼ ½ Br2 + e−
[µA]
i∞
r
−0.2 0.0 0.2 0.4

E
(b) [V]
Fig. 16 RRDE results for a Pt(001) electrode in 0.1M HClO4 + 10−4 M
Br− : (a) Disk current; (b) corresponding ring current; and (c) the
potential-dependent charge due to bromide adsorption obtained from
(a) and (b).
bromide adsorption isotherm (charge) ver- surface plane, information about the sur-
sus disk potential is shown in Fig. 16(c) face normal structure (the Pt–Br spacing)
yielding a maximum coverage of ≈0.42 ML and local bonding sites for specifically
at the positive potential limit of 0.5 V. adsorbed anions can be obtained from
In the X-ray scattering experiments, no analysis of the specular and nonspecular
superlattice peaks were found at any po- CTRs [63]. Figures 17(a, b) show the spec-
tential for Br/Pt(001) in agreement with ular (0, 0, l) CTR and nonspecular (1, 0, l)
the STM results of Bittner and cowork- CTR measured at an electrode potential of
ers [64]. Despite the lack of adsorbate −0.2 V, where no bromide (or based on the
structures with long-range order in the adsorption isotherm for Br− , a negligible
10
Intensity
[a.u.]
0.1
(a) (b)
1.4
Ratio I0.2 V/I−0.15 V
1.2
1.0
0.8
0.6
1 2 3 1 2 3
(c) (0, 0, l) (d) (1, 0, l)
Fig. 17 CTR data for the Br− /Pt(001) system: (a) (0, 0, l); and
(b) (1, 0, l) measured at −0.2 V where no Br is adsorbed on the
surface. The solid lines are a fit to the data including relaxation
of the Pt surface. (c, d) show the changes in intensity measured
at 0.2 V as referenced to the data at −0.2 V. The fits to the data
are described in the text.
amount) is expected to be present on the ratio of intensities along the CTR at 0.2 V
surface, but the coverage by hydrogen is (Figs. 17(a, b)) and −0.2 V. Systematic
essentially unity. The solid lines represent changes in the data are apparent, which
a simultaneous fit to the data in which the must be accounted for by changes in the
layer spacing between the first and second Pt surface expansion and/or the presence
Pt layers and the occupancy and rough- of the adsorbed Brads . In calculating a fit
ness (in the form of a static Debye–Waller to the ratio data, the total Br coverage was
factor) of the topmost Pt layer were varied. fixed to 0.35 (to be consistent with the
The best fit gave an occupancy of 0.81 in the RRDE results) and the partial Brads occu-
surface Pt layer, a root-mean-square (r.m.s) pation of the three low-energy adsorption
roughness, σ = 0.09 Å and a surface ex- sites (hollow, bridge, and on-top) was var-
pansion of the Pt lattice by 1.4% of the bulk ied. The only other parameters were the
lattice spacing (∼0.03 Å). Figures 17(c, d) Br–Pt vertical separation and the Pt sur-
show the changes in the CTR data that face expansion. The best fit is shown by the
occur upon the initial stage of bromide solid lines in Figs. 17(c, d). The expansion
adsorption at ca. −0.2 V, displayed as a of the topmost Pt layer is reduced to 1% and
the occupations (and heights) of bromide that copper UPD onto Pt(111) in the pres-
in the hollow, bridge, and on-top sites are ence of halide anions also causes the
0.17 (1.8 Å), 0.08 (2.0 Å), and 0.09 (2.2 Å), formation of an aligned, incommensurate
respectively, with an r.m.s roughness of hexagonal CuBr (or CuCl) bilayer struc-
0.19 Å. With such a limited data set there ture prior to formation of a full copper
is obviously some ambiguity involved in monolayer.
assigning occupancies to the adsorption The absence of any ordered structures
sites. However, a fit to the specular CTR for the Brad adlayer on Pt(001) contrasts
data with Brads located at a single height with the ordered structures observed on
above the surface yielded an unreasonably Au(001), Au(111), and Pt(111). The or-
large r.m.s roughness (σ = 0.7 Å) and this dered Brads adlayer structures observed
suggests a distribution of heights consis- on Pt(111) and Au(111) are incommen-
tent with the different adsorption sites. surate, which means that there is no
The covalent radii of Pt and Brads are 1.14 preference for any particular site. How-
and 1.30 Å, respectively, which gives cal- ever, the energy minimum for adsorption
culated vertical separations of 1.60, 2.08, at the hollow sites on (001) surfaces is
and 2.44 Å for the hollow, bridge, and on- considerably deeper than for the hol-
low sites on the (111) surfaces. This
top sites, respectively. These values are in
explains the appearance of the com-
good agreement with the CTR results and √ √
mensurate c( 2 × 2)R45◦ phase on the
certainly indicate that the Pt–Br bond is
Au(001) surface, which is observed un-
covalent in nature.
til increased bromide coverage drives the
commensurate–incommensurate transi-
4.1.4.1.5 Summary of Anion Adsorption on tion [57]. Initial adsorption of bromide into
Au(hkl) and Pt(hkl) Surfaces As evident random hollow sites on Pt(001), there-
from the above discussion, the differences fore, seems likely, up to a maximum
between the different anion structures are coverage determined by the lateral re-
a consequence of two fundamental prop- pulsion between bromide anions. Given
erties: the strength of the metal–anion that the second–near-neighbor distance
interaction, and the symmetry of the an- for hollow sites is approximately 3.9 Å,
ions with respect to the atomic geometry this maximum coverage is quite low and
of the surface. The adsorption of bromide further adsorption probably leads to occu-
onto Au(111) leads to the formation of pation of other sites, as suggested by the
a rotated, hexagonal incommensurate ad- CTR results. As the potential is increased,
layer, which undergoes compression and the Pt–Br bond becomes stronger, thus
rotation as the potential is increased [55]. further reducing the mobility of the ad-
Bromide adsorption onto Pt(111) also leads sorbed atoms. This may be the mechanism
to the formation of a hexagonal, incom- that prevents the formation of any Brads
mensurate adlayer, however, the adlayer structure with long-range order on the
is aligned with the Pt lattice at all po- Pt(001) surface.
tentials and appears to form a series of
HOC phases rather than exhibit contin- 4.1.4.2 Underpotential Deposition
uous electrocompression [61]. In relation There are many different methods for pro-
to the presence of HOC phases on the ducing bimetallic catalytic surfaces. Until
Pt(111) surface, in Sect. 4.1.4.2, we show recently, UHV-prepared surfaces had an
enormous advantage over the surfaces pre- here and to follow on from Sect. 4.1.4.1,
pared by electrochemical methods because we focus on the influence of anions on Cu
an UHV system equipped with modern UPD onto Au(111), Pt(111), and Pt(001).
surface-sensitive techniques provided mi-
croscopic structural information in a rel- 4.1.4.2.1 Cu UPD on Au(111) in Sulfuric
atively direct fashion. Recently, however, Acid Solution Cu UPD onto Au(111) in
the emergence of surface characterization sulfuric acid has been an archetypal system
techniques such as X-ray scattering and for the study of anion coadsorption during
STM/atomic force microscopy (AFM), op- metal UPD. The UPD of Cu occurs in
erating under electrochemical conditions, two stages and there has been good
has allowed the electrodeposition method agreement that in the second stage a full
to become equally important in the syn- monolayer of Cu is formed. The first stage,
thesis of model bimetallic structures. In however, has been somewhat controversial
this section, we will focus on bimetallic and a unique structural model was not
Au and Pt single-crystal surfaces produced confirmed until a detailed X-ray diffraction
by the UPD solution phase method. UPD study by Toney and coworkers [67]. This is
corresponds to the electrochemical adsorp- partly because the Cu2+ ion is not fully
tion, often of one monolayer, that occurs at discharged to neutral Cu0 but carries an
electrode potentials positive of the Nernst unknown charge that makes the surface
potential below which bulk metal adsorp- coverage by Cu difficult to determine by
tion occurs [16]. Numerous experiments cyclic voltammetry. The X-ray scattering
have shown that the UPD layer can dra- experiment followed on from an ex situ
matically alter the chemical and electronic LEED and Auger spectroscopy study in
properties of the interface. The UPD layer which it was proposed that the Cu atoms
√
is also the first stage of bulk metal deformed a honeycomb structure with ( 3 ×
√
position and its structure, therefore, can 3)R30◦ symmetry and a surface coverage
strongly influence the structure of the of 2/3 ML per Au surface atom [68].
bulk deposit. Early studies of UPD us- Toney and coworkers confirmed these
ing polycrystalline substrates have, more results by performing a detailed surface
recently, been extended to single-crystal X-ray diffraction study of the structure
substrates, which not only allows the role involving the measurement of the in-
of surface atomic structure to be explored, plane structure factors, the l dependence
but also permits the study of the interface of some of the fractional-order reflections
structures by diffraction-based techniques, and the bulk CTRs passing through the Au
such as surface X-ray scattering. Quite a Bragg reflections [67]. Figure 18(a) shows
few systems have now been studied using the in-plane diffraction pattern for the
√ √
this technique and this has led to a greater ( 3 × 3)R30◦ structure together with
understanding of the physics determining some of the fractional-order rod data. The
the structure of the UPD layer, in particu- data in Fig. 18 immediately ruled out the
√ √
lar with regards to the role of the electrode ( 3 × 3)R30◦ structure proposed from
potential and of various other adsorbing STM and AFM experiments (a triangular
species that can be present in solution. Cu adlayer) [69] as such a flat adlayer
Many relevant papers in UPD studies would give rise to fractional-order profiles
are referenced in the review articles by that were far more intense and fall off
Kolb [65] and Adzic [66]. For our purposes slowly with L. The data in Fig. 18 is
(02) (1/3 1/3) & (4/3 4/3) rods

1000 (1/3 1/3) rod, data
(1/3 1/3) rod, flt
( 23 53 ) 800 (4/3 4/3) rod, data
(4/3 4/3) rod, flt
(− 23 43 ) ( 43 43 )
(−11) (01) (11) 600
F2
[eu]
(− 13 23 ) ( 23 23 ) 400
1 1
(3 3) ( 43 13 )
(00) (10) (20) 200
(23 − 13 )
0
( 43 − 23 )
0 1 2 3 4 5 6
L
(a) (b) [r/u]
(−2/3 4/3) & (2/3 5/3) rods (4/3 1/3) rods

500 (−2/3 4/3) rod, data 700 (4/3 1/3) rod, data
(−2/3 4/3) rod, flt (4/3 1/3) rod, flt
400 (2/3 5/3) rod, data 600
(2/3 5/3) rod, flt
500
300
F2
F2
400
[eu]
[eu]
200 300
200
100
100
0 0
0 1 2 3 4 0.5 1.0 1.5 2.0 2.5 3.0
L L
(c) [r/u] (d) [r/u]
Fig. 18 X-ray diffraction data for Cu/Au(111) in l) and (2/3, 5/3, l); (d) (4/3, 1.3, l). The solid lines
sulfuric acid solution at the first stage of Cu are fits to the data according to the structural
deposition: (a) In-plane diffraction pattern; model shown in Fig. 19. (Taken from Ref. [67].)
(b) (1/3, 1/3, l) and (4/3, 4/3, l); (c) (−2/3, 4/3,
only consistent with a model consisting oxygen atom points away from the surface.
of atoms at different heights above the From the detailed analysis of the data, it
Au(111) surface. was possible to determine several bond
Toney and coworkers considered a range lengths in the structure. The Cu−O bond
of structural models containing both length was 2.15 Å, a distance comparable
ordered Cu and sulfates and proposed to the Cu−O bond lengths in organic
the model shown in Fig. 19. In this compounds, which implied that the sulfate
model, Cu atoms form a honeycomb lattice molecules were chemically bonded to the
(2/3 ML coverage) and sulfate molecules Cu ad-atoms hence stabilizing the partial
are adsorbed in the honeycomb centers charge of the Cu. It should be noted that
above the Cu atoms. Both Cu atoms this open Cu structure is not found in UPD
and sulfate molecules occupy fcc threefold studies of Cu on Au(111) in perchloric
hollow sites on the Au(111) lattice. Three acid solution thus confirming the crucial
oxygen atoms of each sulfate molecule are role of the sulfate anions in the UPD
bonded to Cu atoms and the remaining process.
Fig. 19 Interfacial structure of

Cu/Au(111) derived from the data in
Fig. 18. (a) Top view and (b) side view.
The large light gray, medium gray, small
light gray, and filled spheres represent
Au, Cu, S, and O atoms, respectively.
(Taken from Ref. [67].)
(a)
(b)
4.1.4.2.2 Cu UPD on Pt(111) in Sulfuric can show single peaks or double peaks in
Acid Solution The interpretation of pro- both sweep directions. The filled circles in
cesses associated with the formation of the Fig. 20(a) correspond to the Cu coverage
Cu monolayer on Pt(111), and the nature obtained by potentiostatic measurements,
of the Pt(111)–Cu structure, have been in which the amount of Cu deposited on
the subject of considerable controversy. the disk electrode in a potential step from
Overviews with some different perspec- Ei (indicated in the figure) is assessed
tives can be found in Refs. [70–80]. The from the corresponding change in the con-
RRDE method was successfully applied to stant Cu2+ flux to the ring electrode. The
investigate the kinetics of Cu2+ deposition guide to the eye (solid line) shows that in
and to determine the potential-dependent the potential region in between the two
surface coverage by Cuupd (%Cu(upd) ). The peaks, the Cu coverage is constant in the
lower curve in Fig. 20(a) shows a cathodic range 0.6–0.7 ML (i.e. Cu atoms per Pt
sweep of the cyclic voltammogram (CV) surface atom).
measured on the disk electrode in an The structural details pertaining to Cu
RRDE for UPD of Cu onto Pt(111) in UPD have been derived from in situ
sulfuric acid solution. The presence of X-ray scattering measurements [81]. The
sulfate anions in solution causes a split- specular CTR results in Fig. 20(b) clearly
ting of the voltammetric Cu deposition indicate that UPD occurs in a two-stage
peak. This result is in agreement with process and the sharpness of the peak
previous studies that, depending on the implies that the structure formed in the
solution concentrations and sweep rates, middle potential region is present over
Fig. 20 (a) The lower curve shows the

cathodic sweep of the cyclic Pt(111)
0.8
voltammogram (CV) for a Pt (111) disk 0.05 M H2SO4
electrode in an RRDE configuration in 5 × 10−5 M Cu2+
0.05 M H2 SO4 + 10−5 M Cu. The w = 900 rpm
0.6
closed circles represent the Cu coverage
(atoms per surface Pt atom) assessed
ΘCu
from changes in the ring current during 0
Cu deposition onto the disk electrode as 0.4
Ei
the potential is stepped (for each point)
from Ei . (b) Changes in the X-ray
[µA]
−2
i
scattering at (0, 0, 3.9), lower panel, (0, 0.2
1, 2.5), middle panel, and (1/3 4/3, 1.5),
top panel, as the electrode potential is
swept at 2 mV s−1 . The solid lines and 0.0 −4
dashed lines are for the positive-going 0.0 0.1 0.2 0.3 0.4
(anodic) and negative-going (cathodic) E
sweep directions, respectively. (a) [V]
(1/3, 4/3, 1.5)

0.30
0.20
× 100
0.10
0.40
(0, 1, 2.5)
0.35
Intensity
[a.u.]
0.30
0.25
0.20
(0, 0, 3.9)
0.25
0.20
0.0 0.1 0.2 0.3
E
(b) [V]
a very narrow potential range. Indeed, range 30–60 Å. This correlation length is
at 0.15 V weak peaks, which could be at the limit of detection by X-ray scattering
√ √
indexed to a ( 3 × 3)R30◦ superlattice, and the weak intensities coupled with the
diffraction patterns were observed. For narrow stable potential range precludes a
example, rocking scans through the (2/3, detailed structural analysis as performed
2/3, 2), (1/3, 1/3, 1.5), and (1/3, 4/3, 1.5) for the Au/Cu/H2 SO4 system [67]. It
positions showed broad peaks, FWHM should be noted, however, that there are a
of 0.84◦ , 0.49◦ , and 1.72◦ , respectively, number of similarities, most noticeably
which gives an ordered domain size in the that the superlattice structure factors
are not measurable at small out-of-plane one sulfur and two oxygen atoms (Z = 64).
momentum transfer. This implies that Additional sensitivity to the coverage and
a similar structure may be responsible location of the Cu atoms was obtained by
for the observed diffraction pattern, an performing the CTR measurements with
assumption that can be tested by analysis incident X-ray energies of 8.78 keV (200 eV
of CTR data. In the top part of Fig. 20(b) below the Cu K adsorption edge) and
is shown the X-ray scattering intensity at 8.94 keV (5 eV below the Cu K adsorption
(1/3, 4/3, 1.5) as the potential is swept at the edge). In simulating the CTR data, the
same rate as the data in the lower figures. Cu atomic form factor in the scattering
Although the signal to noise is poor, it can equations was replaced by the form of
clearly be seen that the scattering from the Eq. (7). At each electrode potential, a
√ √
( 3 × 3)R30◦ structure is only present simultaneous fit to the CTR data (E =
in the same narrow region highlighted by 8.78 keV) and the ratio data set (R =
the changes at (0, 0, 3.9). I8.78 keV /I8.94 keV ) was performed. The
To gain more insight into the nature structural parameters allowed to vary were
of the Cu UPD structures, the (0, 0, the Cu coverage, Pt–Cu surface normal
l), (1, 0, l) and (0, 1, l) CTRs were spacing, sulfate coverage, and Cu-sulfate
measured at 0.0 and 0.15 V in order to surface normal spacing, along with their
derive structural models. Such results are respective enhanced Debye–Waller-type
relatively insensitive to the orientation of roughness. Results of fits to the specular
any adsorbed sulfate molecules and so the CTR data and ratio data at electrode
sulfate anion was approximated to a single potentials of 0.0 and 0.15 V are shown
molecule (i.e. single ion core) consisting of in Fig. 21 by the solid lines.
1.3
E = 0.15 V E = 0.0 V
I8.78 keV/I8.34 keV
1.2
1.1
1.0
0.9
1.0
Fig. 21 The measured specular CTR
Intensity
[a.u.]
scattering for an incident X-ray energy of

8.78 keV (lower panels) and the ratio
between a similar data set measured at
0.1 8.94 keV (upper panels) at electrode
potentials of 0.15 V (left) and 0.0 V
(right). The solid lines are fits to the
data to a single structural model and the
2 4 2 4
ratio data set is calculated by fixed
(0, 0, l ) (units of c*) changes in the Cu atomic form factor.
Use of the anomalous scattering tech- be responsible for the weak ordering of
nique enhances the accuracy of the Cu the intermediate Cu-sulfate bilayer phase
coverage determination. At 0.15 V, the Cu and for the inability to complete the Cu
coverage is 0.60 with an adsorbed sulfate monolayer at 0.0 V.
layer of coverage 0.22 on top of the Cu
layer. By comparison with the structural
√ √ 4.1.4.2.3 Cu UPD onto Pt(111) in the Pres-
model derived for the ( 3 × 3)R30◦ Cu-
ence of Bromide Anions The RRDE re-
sulfate structure on Au(111) (Fig. 19), it
sults for Cu UPD on Pt(111) in solution
can be seen that the CTR results are in
containing Cu2+ and Br− ions are illus-
good agreement with that study. The bond
trated in Fig. 22. As the disk potential is
lengths obtained for the Cu-sulfate bilayer
swept across the Cu UPD region, both Cu
are in good agreement with the Au study
UPD peaks recorded on the disk electrode,
in which it was proposed that the sulfate
first at 0.25 V and the second at 0.15 V, are
is chemically bonded to the hexagonal Cu
mirrored by the decrease in the ring cur-
layer. At 0.0 V, the results give a Cu cov-
rent from its ir∞ value. Following these two
erage of 0.81 ML per Pt surface atom with
characteristic Cu UPD peaks, the Pt(111)-
0.21 ML of sulfate anions still adsorbed on
disk current diminishes to a capacitive
top of the Cu layer. Because of the increase
current below 0.1 V, manifesting the com-
in Cu coverage, the sulfate anions are lo-
cated at an increased height above the Cu pletion of the Cu monolayer on the Pt(111)
adlayer presumably as the hollow sites in disk electrode; consistent with the return
the hexagonal Cu adlayer have been filled of ir to ir∞ in the same potential region.
by Cu ad-atoms. It is interesting that the The fact that the ring-shielding current, ir ,
Cu coverage derived by the CTR measure- mirrors the current on the disk electrode
ment are in good agreement with those clearly indicates that both cathodic pro-
obtained from the RRDE experiments and cesses are primarily related with Cu UPD.
it appears, therefore, that it is not possi- The same observation is made for the sub-
ble to complete a full Cu monolayer on sequent positive sweep, at which there are
the Pt(111) surface. The lower limit of two distinctive processes seen from the
potential (0.0 V) was chosen to avoid the ring electrode; the first as a shoulder at
formation of Cu+ via one-electron reduc- ca. 0.375 V and second as a sharp peak
tion of Cu2+ without further reduction at 0.425 V, both clearly related to strip-
to metallic copper. Fits to the nonspecu- ping of Cu from the Pt(111) disk electrode.
lar CTR data at both potentials (0.0 and The total amount of Cu deposited by UPD
0.15 V), again to both raw data and ratio was evaluated quantitatively by integration
data sets, gave occupation of both types of of the ring current, yielding a maximum
threefold hollow sites, fcc and hexagonal Cu coverage of 0.95 ML (almost identical
close packed (hcp), although the fcc sites with the value inferred from disk measure-
are favored (fcc/hcp = 0.51/0.09 at 0.15 V ments, the insert in Fig. 22). This rather
and 0.67/0.16 at 0.0 V). In contrast with the small difference in the charge assessed
Au(111) system in which the Cu ad-atoms from the disk and the ring measurements
occupied only the fcc sites, it appears that (ca. 5 µC cm−2 ) suggests that the anion co-
on Pt(111) the energy difference between adsorption contributes insignificantly to
these sites may be smaller. Occupation of the Coulombic charge passing the inter-
both types of threefold hollow site may face during Cu deposition.
20
0.0
qdisk
900 rpm
5 mVs−1
15 qring
[ML]
qCu
0.5
Disk current
10
[µA]
1.0
5 0.0 0.1 0.2 0.3 0.4
E
[V]
0.1 M HClO4
(a) 5 × 10−5 M Cu2+
1 × 10−4 M Br−
i∞r
Ring current
−0.275 V
[µA]
Cu2+ + 2e Cu
2 µA
0.0 0.2 0.4

E
(b) [V]
Fig. 22 Top: CV for Cu UPD on a Pt(111) disk electrode in solution
containing Br− . Bottom: ring-electrode currents recorded with the ring
being potentiostated at −0.275 V. Insert: copper surface coverages
assessed from the disk electrode (θd ) and the ring electrode (θr ) during
the positive-going sweep.
The ring-shielding properties of the as indicated by the bromide adsorption

RRDE can also be used to assess the isotherm inferred from solution free of
mass flux of Br− from and to the Pt(111) copper, insert of Fig. 23. In the potential
disk electrode during Cu UPD [74, 79]. region between 0.45 and 0.3 V, a small
Figure 23 shows that starting from the desorption of Br− (≈2 µC cm−2 ) just
positive potential limit, the ring current at preceding Cu deposition is demonstrated
≈0.45 V equals ir∞ indicating the absence by the concomitant increase of the ring
of either Br− adsorption or desorption on current above its unshielded value. As
the disk. In fact, at this potential, the the disk potential is swept across the
surface is almost fully covered by Brad , first UPD peak, the ring current remains
0.50
900 rpm
20 mVs−1
[ML]
qBr
Pt(111)
0.25
Pt(111)/Cu
Disk current
0.0
0.0 0.2 0.4
[µA]
E
[V]
0.1 M HClO4
1 × 10−4 M Cu2+
5 µA 1 × 10−4 M Br−
0
Ring current
I R∞ = 16.2 µA
[µA]
1 µA
+1.06 V
Br− ½ Br2 + e−
0.0 0.2 0.4

E
[V]
Fig. 23 Top: Potentiodynamic curve for Cu UPD on a Pt(111) disk
electrode. Bottom: Ring current corresponding to the change in the
Br− flux during copper deposition. Insert: Surface coverage of
bromide at the disk electrode with and without copper present in
solution (assessed from ring transients).
almost constant, manifesting an absence in between the two Cu UPD peaks is equal
of either Br− adsorption or desorption; to the coverage following the complete
the second Cu UPD peak is accompanied stripping of Cu from the Pt(111) surface.
by the increase in the ring current Note also that after complete stripping of
from ir∞ , with completion of the Cu Cu, the Br coverage is exactly the same as
monolayer causing the ring current to in the case when the adsorption of Br−
return back to its unshielded value. The on Pt(111) was monitored in a solution
resulting θBr from integration of the ring free of Cu. Remarkably, the total change in
currents versus the potential of the Pt(111) adsorption/desorption of Br− during Cu
surface in a solution containing Cu2+ , and UPD is rather low, ca. 4 µC cm−2 .
the corresponding values for the Pt(111) X-ray scattering measurements of the
surface in the solution free of Cu2+ , are Cu UPD layer on Pt(111) in solution con-
shown in the insert of Fig. 23. Note that θBr taining bromide anions are summarized
in Fig. 24 [77, 80]. With the potential at of the fundamental peak (Fig. 24c) show
0.225 V, four symmetry-independent in- that the structure responsible for this peak
plane peaks were observed at (0, 0.735), is potential dependent and present only
(0.735, 0.735), (0, 1.47), (0.735, 1.47), as in the potential region between the two
shown by the filled circles in the in-plane Cu UPD peaks. The location of the fun-
diffraction pattern illustrated in Fig. 24. damental peak at 0.735 (and higher-order
The changes in the scattering intensity peaks) shifted to 0.742 upon increasing the
0.14
2.0
0.12
Intensity
Intensity
1.5
0.10
1.0
0.08
0.5
0.06
(O k) 0.0
−2 0 2 −0.4 0.0 0.4
∆f ∆f
(a) [°] (b) [°]
(h O)
0.77
6
Integrated intensity
0.76
Peak position (h)
CuBr
4 0.75 CuCl
2 0.74
0 0.73
0.2 0.3 0.4 0.2 0.3 0.4
E E
(c) [V] (d) [V]
Fig. 24 The in-plane X-ray diffraction pattern observed at 0.225 V for Cu
UPD onto Pt(111) in solution containing Br− . The Pt CTRs are represented
by filled circles and the open circles correspond to scattering from the
‘‘4 × 4’’ structure. Rocking scans through (a) (0, 0.741 0.1) and (b) (1, 0,
0.1). (c) Integrated intensities obtained from scans such as the one shown
in (a) for the CuBr bilayer (filled symbols) and CuCl bilayer (open circles)
phases. (d) Corresponding position of the first-order diffraction peak from
the CuBr and CuCl bilayer phases as a function of the electrode potential.
potential stepwise, as shown in Fig. 24(d). atom arranged in two hexagonal layers,
The comparable data for chloride solution one of Cu and one of Br, each with the
is also shown. The location and symme- coverage of ≈0.53 ML.
try of the in-plane peaks indicates that
the adsorbed species form a hexagonal su- 4.1.4.2.4 Cu UPD onto Pt(001) in the
perstructure that is incommensurate with Presence of Bromide Anions In pure
the underlying platinum lattice; the near- perchloric acid, Cu is deposited as p(1 × 1)
est commensurate unit cell is (4 × 4), islands on Pt(001) in a one-step adsorption
which would have a fundamental peak process as evidenced by a single reversible
at 0.750. Note that the chloride struc- change in the scattered X-ray intensity
ture has a unit cell that is smaller than at (1, 0, 0.1), an ‘‘anti-Bragg’’ point
(4 × 4), and the bromide structure has one on the nonspecular CTR [82]. The effect
that is larger than (4 × 4), illustrating in of bromide adsorbed onto the surface
a rather dramatic way the dominant role (i.e. from solution) is to broaden the
played by the anion and Cu-anion bond- potential range of Cu deposition, although
ing in determining the structure of this there is no evidence of a stagewise
intermediate phase. The inserts (a) and deposition process. This contrasts with
(b) in Fig. 24 show two rocking scans results obtained on the Pt(111) surface
(φ scans) through the (0, 0.735, 0.1) and in which two distinct steps in Cu UPD
(1, 0, 0.1) peaks, respectively. Both peaks were observed (see previous section). With
have Lorentzian lineshapes with the width the electrode potential held at 0.43, 0.12,
of the (0, 0.735) peak corresponding to a and −0.13 V, in each case allowing enough
domain size of ≈200 Å, somewhat smaller time for the system to reach an equilibrium
than the estimated terrace size of the state, a search was performed for in-
Pt(111) surface, ≈1000 Å. Additional infor- plane X-ray scattering peaks. In agreement
mation about the composition of the unit with the results in Sect. 4.1.4.1.4, no
cell of this phase was obtained by utiliz- peaks were found at 0.43 V, where only
ing ‘‘anomalous scattering’’ methods, that Br is adsorbed onto the surface. In
is, by measuring the relative intensities addition, no peaks were found at 0.12 V,
of the four in-plane peak intensities close an intermediate potential in which on the
to the Cu K adsorption edge (8979 eV). Pt(111) surface under identical conditions
In general, the intensities of the relatively the (4 × 4) structure was observed. At
strong (0.735, 0.735) and (0, 1.47) peaks −0.13 V, however, peaks at reciprocal
were weaker close to the Cu edge, indi- lattice positions, which could be indexed
cating that the adsorbed superstructure to a c(2 × 2) structure, were found. The
does not consist of only Cu atoms (which in-plane scattering in reciprocal space is
would result in the reduction of the in- represented in Fig. 25 together with some
tensity of all in-plane peaks near the Cu rocking scans at the various reciprocal
edge) but must also include Br (for more lattice points that are indicated, (a) (1/2,
details, see Ref. [80]). The simplest real- 1/2, 0.1), (b) (0, 1, 0.1), and (c) (1/2, 3/2,
space structure consistent with the data 0.1). As can be seen from Fig. 25, all of
has an hexagonal unit cell with the lat- the measured c(2 × 2) reflections were
tice spacing (0.735)−1 times the Pt lattice relatively broad compared to the peak at
spacing, or 3.36 Å, and a two atom basis (1, 0, 0.1) on the Pt CTR and, from
containing one copper and one bromide Lorentzian fits to the data (solid lines
Pt(bragg)
Pt(CTR)
2
c(2 × 2)
h
0 1 2
k
Intensity
[a.u.]
−2 0 2 −0.4 0.0 0.4 −2 0 2

∆f
[degrees]
(a) (b) (c)
Fig. 25 A representation of the in-plane X-ray scattering measured at
−0.13 V in solution containing 0.1 M HClO4 + 10−3 M
Cu2+ + 10−2 M KBr. The solid circles correspond to the measured
c(2 × 2) reflections and the squares to the location of bulk Pt CTRs
and Pt Bragg reflections. The lower figures show scans through the
indicated reciprocal lattice points at: (a) (1/2, 1/2, 0.1); (b) (0, 1, 0.1);
and (c) (1/2, 3/2, 0.1). In each case, the solid lines are fits of a
Lorentzian lineshape to the data.
in Fig. 25), a domain size in the range that the (4 × 4) structure observed in the
30–60 Å for the c(2 × 2) structure was first stage of Cu deposition on Pt(111)
calculated. The c(2 × 2) diffraction pattern contained both Cu and Br in the unit cell.
was only formed if the electrode potential Given that the c(2 × 2) diffraction pattern
was held in the region below −0.12 V in this study was obtained at −0.13 V, a
and even then, only if the potential was potential in which a significant amount of
approached from more negative potential. Cu is adsorbed onto the surface, similar
Once formed, the c(2 × 2) structure was measurements were made at several l
stable at negative potential until the onset values along the (1/2, 1/2, l) and (1/2, 3/2,
of bulk Cu deposition. l) rods. Measurements of the integrated
In the previous section, anomalous intensities at 8779 eV, 200 eV below the
scattering techniques were used to show Cu K adsorption edge, and 8974 eV, 5 eV
below the Cu edge, were performed but Tab. 1 Structural parameters to the fit to
showed absolutely no dependence on the the data in Fig. 26. The coverages, θCu and
θBr , are with respect to a full copper
incident X-ray energy. Given that the
monolayer (one Cu atom per surface Pt
dispersion corrections to the Cu atomic atom) and a full c(2 × 2) Br adlayer (0.5 Br
form factor change significantly over this atoms per surface Pt atom)
energy range, this result implies that no
Cu is contained in the c(2 × 2) unit cell. θCu 1.0 ± 0.06
It seems likely, therefore, that the c(2 × 2) dPt – Cu 1.75 ± 0.05 Å
structure consists of an ordered Br lattice σCu 0.13 ± 0.05 Å
that is formed on top of a pseudomorphic θBr [c(2 × 2)] 0.9 ± 0.1
Cu layer. dCu – Br 1.79 ± 0.08 Å
σBr 0.3 ± 0.2 Å
To obtain detailed structural informa-
tion CTR measurements were modeled by
a c(2 × 2) Br adlayer on top of a p(1 × 1)
pseudomorphic Cu monolayer, where the schematic of the structure is shown in
respective coverages, surface normal spac- Fig. 26.
ings, and roughnesses were allowed to vary Although the presence of Br anions
in order to fit the data. The CTRs passing considerably slows the kinetics of ordering,
through the bulk Pt Bragg reflections in- at negative potential (−0.13 V) a complete
clude contributions to the scattering both pseudomorphic p(1 × 1) Cu monolayer
from the Cu and Br adlayers and the Pt is formed on the Pt(001) surface. When
lattice. To further test the model, the CTRs this monolayer is completed, Br forms
were measured with incident X-ray ener- a c(2 × 2) overlayer on top of the Cu
gies of 8779 and 8974 eV and the structural monolayer. The Cu–Br spacing implies
parameters were simultaneously refined that, as was the case for the Pt–Br bond for
to fit both the raw CTR data set (mea- Br adsorption onto Pt(001) (Sect. 4.1.4.1.4),
sured at 10 keV) and the intensity ratio the metal–halide bond is covalent in
I8779 eV /I8974 eV . This method increases nature. The near-neighbor spacing in the
the sensitivity of the fitting procedure to Br adlayer is 3.92 Å, which is similar to
the Cu layer as shown in Sect. 4.1.4.2.2. the near-neighbor spacing of Br adsorbed
CTR data taken with the potential held at onto Pt(111) at positive electrode potential
−0.13 V, where the c(2 × 2) structure was (Sect. 4.1.4.1.3). The Br coverage on the
present, is shown in the lower panels of pseudomorphic Cu monolayer is nearly
Fig. 26. The top panels in Fig. 26 show identical to the coverage measured on the
the ratio data sets, I8779 eV /I8974 eV , which Pt surface at positive potential and implies
clearly indicates the sensitivity of the CTR that close to a full Br monolayer remains
measurements to the Cu adlayer. The solid on the electrode surface during Cu UPD. It
lines in Fig. 26 correspond to the results is interesting that the bromide layer forms
of a simultaneous fit to these data sets in a structure with long-range order on the Cu
which the Cu coverage, θCu , Pt–Cu spac- monolayer and not on the Pt(001) surface
ing, dPt – Cu , coverage of Br in a c(2 × 2) in solution free of Cu.
adlayer, that is, two Br atoms per c(2 × 2)
unit cell, θBr , Cu–Br spacing, dCu – Br , and 4.1.4.2.5 A Physical Model for Cu UPD
the Cu and Br roughnesses were var- The results in this section have demon-
ied. The results are listed in Table 1. A strated that Cu UPD is a multistage
Fig. 26 The measured CTRs at an

1.6 electrode potential of −0.13 V in 0.1 M
HClO4 + 10−3 M Cu2+ + 10−2 M KBr
I8779 keV/I8974 keV
1.4 where the c(2 × 2) structure is present.

1.2
Upper panels: The ratio between CTR
data sets taken with incident X-ray
1.0 photon energies of 8779 and 8974 eV,
that is, I8779 eV /I8974 eV . The solid lines
0.8
are results of a simultaneous fit to all of
0.6 the data according to the structural
model that is schematically illustrated.
The open circles are Pt substrate atoms,
10−1 the shaded circles are Cu atoms, and
Intensity
the black circles are Br atoms that form

[a.u.]
the c(2 × 2) structure. The side view

10−2
indicates the surface normal spacings
listed in Table 1.
10−3
1 2 3 1 2 3
(0, 0, l) (1, 0, l)
dCu-Br
dPt-Br
process that, irrespective of the orientation a true bimetallic surface exists at which
of surface and/or the nature of specif- both Pt and Cu atoms can do cataly-
ically adsorbing anions, is governed by sis. In either sulfate or halide-containing
a complex interplay of Cuupd –substrate, solutions, a multistep deposition occurs
Cuupd –anion, and the substrate–anion with the formation of ordered anion adlat-
energetics. The structure of the adlayer tices. Possible Pt(111)–Cuupd –anion and
depends strongly on the anion in the Pt(001)–Cuupd –anion structural models
supporting electrolyte. In perchloric acid, as a function of Cuupd coverage are shown
the UPD layer on Pt(111) grows as pro- in Fig. 27. This model is very similar to
gressively larger patches of Cu having previously proposed models for Cu UPD
the Pt lattice constant, that is, a pseudo- on Pt(111) [75] and Au(111) [83], and is
morphic adlayer. From the electrocatalysis based on the enhanced adsorption of halide
standpoint, the Cuupd –ClO4 − system is and other anions on Cuupd –modified Pt
potentially the most interesting, because sites. Clearly, Cuupd is either sandwiched
Pt(111)
200 3.
[µA cm−2]
- Anions
(4 × 4)
jd
100 - Cu
4.
O2
2.
1.
(a)
3.
Pt(001)
100
(2 × 2)
[µA cm−2]
jd
4. 0.2 0.4 0.6 0.8 1.0

E vs RHE
(b)
[V]
Fig. 27 A cartoon schematic of the Cu UPD process on Pt in the presence of anion

species. States 1–4 represent increasing Cu coverage, that is, consistent with a
negative sweep of the electrode potential.
between the Pt surface and anions or is by Cuupd (0.5 < %Cu < 1), anions are ei-
in contact with anions adsorbed on the ther entirely displaced from the surface
adjacent Pt sites [77]. States 1a and 1b show by Cuupd (state 3a) or Cuupd and anions
a close-packed bisulfate/halide layer that is form two-layer structures in which an-
present on the Pt(111) and Pt(001) surfaces ions are adsorbed on both Pt as well as
at potentials positive of Cu UPD. States Cuupd sites (state 3b in Fig. 27). Note that
2a and 2b refer to the initial deposition state 3a is representative of the ordered
of Cuupd on Pt(111) and Pt(001) surfaces (4 × 4) Clupd –Clad (Brad ) bilayer structure.
on which the Cu2+ ion is only partially In contrast with the Pt(111)–Cuupd –anion
discharged. At higher surface coverages system, on Pt(001), in the same potential
range (0.1–0.4 V), a mixture of Cu and Br platinum group metals measurable cur-
atoms is present on the Pt(001) surface rents are obtained only below 1 V. It is
in a disordered structure. Presumably the customary, therefore, in kinetic studies of
strong affinity of adsorbates for the Pt the ORR to use the current density at a
fourfold hollow site prevents the forma- fixed potential, for example, at 0.9 V, as
tion of a long-range ordered bilayer phase a measure of the reaction rate instead of
(for example, a (001) plane of Cu(I)Br) as the exchange current density, for example,
adsorbate mobility is reduced on the (001) at the reversible potential. Given that the
substrate. The final step in Cu UPD is reversible potential of the ORR is 1.23 V,
the ‘‘filling-in’’ of the Cuupd monolayer virtually all reversible electrode potentials
to form a bilayer phase: that is, a pseu- for solid elements are less positive, and
domorphic (1 × 1) Cuupd monolayer with consequently not many substrates may
either a disordered anion adlayer on the top even be considered as oxygen electrodes.
of Pt(111)–Cuupd [80, 81], state 4a, or an Besides the stability of electrode materi-
ordered c(2 × 2) bilayer Cuupd –Brad struc- als, the ORR is accompanied with many
ture on the Pt(001) surface [82], state 4b. other complications, such as the neces-
In both Cl− and Br− solutions, Cu UPD sity of cleaving the strong O−O bond (ca.
onto Pt(111) proceeds via the same inter- 497 kJ mol−1 ) to produce the maximum
mediate phase. In sulfuric acid solution, efficiency in the 4e− process. Complete
however, the equivalent intermediate state evaluation of electrocatalysis for the ORR
on Pt(111) corresponds to the forma- must, therefore, accommodate many con-
√ √
tion of the honeycomb ( 3 × 3)R30◦ tributing factors, including variations of
Cuupd –HSO4 structure [81] that consists the free energy of activation, the heats
of both 0.22 ML bisulfate anions and of adsorption of reactive intermediates,
0.66 ML of Cuupd in the unit cell. The same and the preexponential surface coverage
structure was proposed for the Au(111)- term, (1 − %ad ). In this section, closely
sulfate system [67, 68]. The formation of following the microscopic description of
the intermediate state (state 2) occurs over clean and modified metal surfaces (pre-
a very narrow region of potential. As will sented in Sects. 4.1.3 and 4.1.4), we discuss
recent efforts towards a fundamental un-
be shown in Sect. 4.1.5, the existence of
derstanding of electrocatalysis of the ORR
bare platinum sites in state 2 (Fig. 27) is
on these surfaces.
essential for the adsorption of O2 , and this
first-stage formation plays the key role in
the activity of the Pt–Cuupd interface for 4.1.5.1 Reaction Pathway
the ORR. The ORR is a multielectron reaction that
may include a number of elementary steps
4.1.5 involving different reaction intermediates.
The Oxygen-reduction Reaction (ORR) Of the various reaction schemes proposed
for the ORR [84], the simplified version
The main problem in the development of of the scheme given by Wroblowa and
an efficient electrochemical energy con- coworkers [85] appears to be the most
version device (fuel cell) is in the sluggish effective one to describe the complicated
ORR kinetics even on the most catalytically reaction pathway by which O2 is reduced
active electrode materials, for example, on at metal surfaces:
k1 where n is the number of electrons, K is

the rate constant, cO2 is the concentration
k2 k3
O2 O2, ad H2O2, ad H2O of O2 in the solution, %ad is the total
surface coverage by anions (%ad) and OHad
k4 k5 (%OHad ), x is either 1 or 2 depending on
the site requirements of the adsorbates,
H2O2
i is the observed current, E is the
applied potential, 1G∗θ is the standard
On the basis of this reaction scheme, O2
free energy (‘‘chemical’’) of activation
can be electrochemically reduced either
(1G∗θ = 1G∗0 + γ 1G0% %ad ), 1G∗0 is the
directly to water with the rate constant
standard free energy of activation at
k1 without intermediate formation of
zero coverage, 1G0% is the difference
H2 O2,ad (so-called ‘‘direct’’ 4e− reduction)
in the heats of adsorption of reactance
or to H2 O2,ad with the rate constant k2 and products in the rate-controlling step
(‘‘series’’ 2e− reduction). The adsorbed (1G0% = 1G00 + r%ad ), β and γ are the
peroxide can be electrochemically reduced symmetry factors (assumed to be 1/2),
to water with the rate constant k3 (‘‘series’’ and r is a parameter characterizing the
4e− pathway), catalytically (chemically) rate of change of the apparent standard
decomposed on the electrode surface free energy of adsorption of reaction
(k4 ), or desorbed into the bulk of the intermediates with the surface coverage by
solution (k5 ). Although a number of adsorbing species. In deriving Eq. (12), it
important problems pertaining to the is assumed that the reactive intermediates,
interpretation of the reaction pathway (O2 − )ad and (HO2 − )ad , are adsorbed only
for the ORR on metal surfaces have yet to low coverage, that is, they are not a
to be resolved, it appears that a series significant part of %ad . In the following
pathway via an (H2 O2 )ad intermediate sections, this reaction pathway and rate
may be operative on pure Pt, Au, and expression are used to analyze the effects of
Cu and their bimetallic surfaces. This various factors on the kinetics of the ORR
can be considered as a special case on Pt(hkl), Au(hkl), and Cu(hkl) surfaces.
of the general mechanism in which k1
is essentially zero, that is, there is no 4.1.5.2 ORR on Pt(hkl) Surfaces
splitting of the O−O bond before a During the past two decades, a great deal
peroxide species is formed. Peroxide, on of work in surface electrochemistry has
the other hand, may (k5 = 0) or may been aimed at elucidating the role of the
not (k5 = 0) be further reduced to water. local symmetry of surface atoms in electro-
In either case, the rate-determining step catalysis, particularly for the kinetics of the
appears to be the addition of the first ORR on platinum single-crystal surfaces.
electron to O2,ad . The rate expression is It is now well established that the kinetics
then [86, 87], of the ORR on Pt(hkl) surfaces are sensitive
to the surface structure [88, 89] and arise
because of structure-sensitive adsorption
−βF E
i = nF KcO2 (1 − %ad )x exp of spectator species, such as Hupd [89, 90],
RT OHad [91], HSO4(ad) [90], Clad [92], and

−1G∗θ Brad [87]. Within the limited scope of this
× exp (12)
RT report, it will not be possible to review all
of these results. Rather, we will show, us- Pt(111), the ORR on Pt(110) (Fig. 28b) and
ing representative examples, the kind of Pt(001) (Fig. 28c) in Cl− -containing solu-
information that can improve our under- tion is accompanied quantitatively with
standing of the role of the local symmetry peroxide formation, implying that within
of platinum surface atoms in O2 reduc- the same potential limits H2 O2 is formed
tion electrocatalysis. There are two general as an intermediate on the Pt(110) and
observations concerning the structure sen- Pt(001) electrodes covered with Clad . On
sitivity of the ORR: (i) the same activation Pt(001) the percentage of H2 O2 pro-
energy in both acid (at the reversible po- duced between 0.25 < E < 0.6 V during
tential ca. 42 kJ mol−1 [93]) and alkaline the ORR is almost constant and is close
solution (at 0.8 Vca. 40 kJ mol−1 [94]) has to 25%. Note that, at potentials more
been found for all three Pt(hkl) surfaces; negative than E = 0.3 V, the surface cov-
(ii) the structural sensitivity is most pro- erage with Clad is reduced substantially,
nounced in electrolytes in which there is but the O2 -reduction currents sharply de-
strong adsorption of anions. As a con- crease to diffusion-limiting currents for
sequence, the structure sensitivity of the ca. 2e− per O2 molecule at the negative
ORR on Pt(hkl) surfaces is mainly de- potential limit. In the same potential re-
termined by the preexponential (1 − %ad ) gion, the O2 -reduction currents on the disk
coverage term. electrode are accompanied quantitatively
by the H2 O2 -oxidation currents on the
4.1.5.2.1 ORR at the Pt(hkl)–halide Inter- ring electrode, reaching maximum per-
face For the purpose of demonstrating oxide production of ca. 90% at 0.075 V.
the importance of the (1 − %ad ) term in In Cl− -free solution, however, a 3e− re-
the kinetics of the ORR on Pt(hkl), two duction is observed at the same potential
representative sets of polarization curves limit [90]. This may suggest that Clad is
are shown in Figs. 28 and 29. Figure 28 always present on Pt(001) covered by Hupd
shows that in the presence of Cl− , the (as also observed for Brad on Pt(001) – see
variation in activity increases in the or- Sect. 4.1.4.1.4) thus, the availability of Pt
der (001) < (110) < (111). Figure 28(a) atoms required for the breaking of the
shows that the ORR on the Pt(111) sur- O−O bond are significantly reduced and
face in 0.1 M HClO4 + 10−3 M Cl− is the ORR proceeds entirely through the 2e−
strongly inhibited, with an activity in a peroxide pathway on the Pt(001)-Hupd -Clad
solution containing Cl− being several or- surface. The amount of peroxide formed
ders of magnitude lower than in 0.1 M on the Pt(110) surface is substantially
HClO4 [88, 92]. Figure 28(a) also shows smaller, Fig. 28(b), ca. 2.5%, and no in-
that between 0.3 < E < 0.9 V, the ring crease in peroxide formation was observed
currents were a small fraction of the disk in the Hupd region.
currents, implying that on the surface As demonstrated recently [92], the key to
highly covered with Clad , oxygen reduc- resolving the structural sensitivity of the
tion proceeds almost entirely through the ORR and the HOR on Pt(hkl) surfaces in
4e− reduction pathway. The appearance Cl− -containing solution is to be found in
of peroxide oxidation currents on the ring understanding the interplay between the
electrode begins at potentials negative of electronic properties (the work function)
0.25 V, and parallels with the adsorption of the substrate and the match between
of hydrogen on Pt(111). In contrast with the Pt surface atomic geometry and the
[µA]
50
IR
0
−2
[mA cm−2]
−4
iD
−6 Pt(111)
(a)
IR H2O2
[mA cm−2] [µA] [%]
2
;
−2
−4
iD
Pt(110)
−6
(b)
0.05 M H2SO4 + 10−3 M Cl−

IR H 2 O2
[µA] [%]
0.05 M H2SO4
50
;
−2
[mA cm−2]
−4
iD
Pt(001)
−6
0.2 0.4 0.6 0.8

E
(c) [VRHE]
Fig. 28 Polarization curves for Pt single-crystal electrodes in
0.05 M H2 SO4 with and without 10−3 M Cl− : (a) Pt(111); (b) Pt(110);
and (c) Pt(001).
adsorbing anions. The high activity of the consequence of this positive value of the
Pt(111)-Clad surface for the ORR can be pzc is that the onset of the desorption of
attributed to the highest work function Clad occurs at more positive potentials on
(the pzc) of the Pt(111) surface. An obvious this surface. The fact that Clad inhibits
the initial adsorption of O2 molecules, insight into this relationship, the ORR has
but does not affect the pathway of the recently been studied on the Pt(111) sur-
reaction implies that the interaction of face in solution containing Br− anions.
Clad with the close-packed (111) surface As demonstrated in Sect. 4.1.4.1.3, the
is not very strong. On the other hand, potential dependence of bromide adsorp-
the lower pzc of Pt(001) would require a tion/surface structure was obtained from a
larger shift in the thermodynamic driving combination of RRDE and X-ray scattering
force to displace Clad from the surface, measurements.
accounting for a strong deactivation of Figure 29(b) summarizes a family of po-
the ORR on the Pt(001)-Clad surface. The larization curves for the ORR on Pt(111) in
substantial production of peroxide on the 0.1 M HClO4 + 10−4 M Br− along with a
Pt(001)-Clad surface is related to strong representative polarization curve recorded
adsorption of Clad on the fourfold hollow in solution free of Br− anions. In all ex-
(001) sites, (similar to the mechanism periments, the ring was potentiostated at
that frustrates the ordering of Br on the 0.95 V, where oxidation of peroxide arriv-
Pt(001) surface – see Sect. 4.1.4.1.4) with a ing at the ring is under diffusion control,
corresponding reduction in the available Sect. 4.1.2.2. In pure perchloric acid, at
pairs of platinum sites required for the low overpotentials, the ring currents were
breaking of the O−O bond. This model a very small fraction of the disk currents,
may seems to conflict with the case of implying that on the surface covered with
Pt(110), in which the kinetics of the ORR reversible adsorbed OHad (the so-called
is higher than on the Pt(001) surface even ‘‘butterfly region’’) oxygen reduction pro-
though the pzc of the (110) surface has a ceeds almost entirely through the direct
lower value and interaction of Clad with low 4e− pathway. The appearance of peroxide
coordinated (110) sites is as strong as on oxidation currents on the ring electrode
Pt(001). The key to resolving this seeming begins at potentials negative of 0.0 V, and
contradiction is, however, in the structure parallels the adsorption of hydrogen on
of the Pt(110) surface itself (as described in Pt(111). At the negative potential limit,
Sect. 4.1.3). Clad is most probably adsorbed the limiting current corresponds to exactly
in the ‘‘troughs’’ or ‘‘missing rows’’ of the two-electron reduction of O2 . Figure 29(b)
reconstructed (110) surface and, hence, O2 shows that the ORR on the Pt(111) sur-
adsorption and the breaking of the O−O face is strongly inhibited by Br− , with
bond may take place on the top rows of an activity in the solution containing Br−
platinum sites (the first atomic layer in the being several orders of magnitude lower
structural model shown in Fig. 9). than in pure perchloric acid. Figure 29(b)
Clad is not observed to form any ordered shows that O2 reduction on the Pt(111)-Br
structures on Pt(hkl) surfaces, in contrast surface starts at ca. 0.3 V (%Br ≈ 0.4 ML
with Brad on Pt(111) – see Sect. 4.1.4.1.3). and %OH ≈ 0.0 ML in Fig. 29a) and is un-
Furthermore, the potential dependence der combined kinetic-diffusion control be-
surface coverage by Clad cannot be dis- tween 0.1 < E < 0.3 V. Figure 29(b) also
cerned from RRDE measurements. As a shows that peroxide was detected on the
consequence, it is difficult to find the ring electrode in the same potential re-
relationship between the activity of the gion, indicating that H2 O2 is formed as
ORR and the potential dependence of an intermediate on the electrode cov-
Clad surface coverage/structures. To get ered with θBr ≈ 0.4–0.35 ML. At potentials
900 RPM
10 µA
0.4
0.3
[Ml]
qBr
0.2
Pt(111)
0.1 M HClO4
0.1
0.1 M HClO4 + 8 × 10−3 M Br
H ads/des on Pt(111)/Br
(a) 0.0
Pt(111)/Br
150 2500
1600
100 900
[µA]
IR
400
50
RPM
0
400
2500
[mA]
−ID
1.0
2.0
−0.2 0.0 0.2 0.4 0.6
E
(b) [V]
Fig. 29 (a) CV for Pt(111) in 0.1 M HClO4 (dashed line) showing
the effect of Br− in solution (solid line) and the effect on the
adsorption/desorption of hydrogen (dotted line). The Br coverage is
shown by the dot-dashed line. (b) Polarization curves for the ORR
on Pt(111) in 0.1 M HClO4 + 10−4 M Br− .
more negative than E = 0.1 V, the sur- currents increase to a diffusion-limiting

face coverage with Br decreases substan- current for 3e− per O2 molecule. In the
tially, for example, from θBr ≈ 0.35 ML same potential region, the O2 -reduction
to θBr ≈ 0.25 ML, and the O2 -reduction currents are accompanied quantitatively by
the H2 O2 -oxidation currents on the ring which is usually from ca. 60 mV dec−1
electrode, implying that in the presence at low current densities to 120 mV dec−1
of Br− anions O2 reduction does not pro- at high current densities [86]. The change
ceed entirely through the 4e− reduction, in the slope has been attributed to the
as found for the Pt(111)-Clad system [92]. change from Temkin to Langmuirian
It was proposed that strongly adsorbed conditions for the adsorption of reaction
Br can simultaneously suppress the initial intermediates [95], or as being due to a
adsorption of the O2 molecule and the for- change in the surface coverage by PtOHad ,
mation of pairs of platinum sites needed which effects the adsorption of molecular
for the breaking of the O−O bond [87]. O2 [96]. In both theories, the first charge-
On the basis of this proposition, the phys- transfer step has been proposed as the
ical model that appears to rationalize the rate-determining step. It has, however,
ORR on the Pt(111)-Brad electrode is one been very difficult to obtain definitive
in which the active sites for the adsorp- evidence that will support either theory,
tion of molecular O2 are the small number since both the nature of intermediates
of ‘‘bare’’ platinum sites that are created in the ORR have never been probed
when desorption of Brad takes place from experimentally and the surface coverage
the Pt(111) surface. Finally, by compar- by OHad species was poorly defined. By
ing the ORR in Fig. 29 with the potential eliminating the surface heterogeneity with
stability of the Brad structure in Fig. 15, single-crystal electrodes, it was proposed
one can find no correlation between the that there are two modes of action of
ordered surface structure and the kinetics the OHad state on the kinetics of the
of the ORR. ORR on the Pt(111) surface [97]: (i) OHad
The analysis for the ORR on Pt(111) can block the adsorption of O2 on active
has shown that the Tafel plot of mass platinum sites, that is, they compete for
transport corrected currents for the ORR at the same sites; and (ii) OHad can alter the
the Pt(111) surface in solution containing adsorption energy of intermediates (the
Br− is ca. 118 mV dec−1 . The fact that r%ad follows the Temkin conditions) that
the Pt(111)-Brad surface has a single slope are formed during the ORR on the bare Pt
of 118 mV dec−1 strongly suggests that sites neighbored by the OHad .
the apparent standard free energy of
adsorption of the reaction intermediates is 4.1.5.2.2 ORR at the Pt(hkl)–Cuupd Inter-
not affected by the neighboring Brad , that faces Studies of the kinetics of the ORR
is, the adsorption of reaction intermediates on platinum single-crystal electrodes mod-
that are formed during the ORR on ified by UPD metal ad-atoms are of interest
the OHad -free Pt(111) surface obeys the primarily from a fundamental science per-
Langmuirian conditions. In contrast, the spective. Since it is possible to begin the
Tafel plot in 0.1 M HClO4 solution experiment with a clean Pt(hkl) surface in
free of Br− deviates significantly from which the ORR is well known, the change
120 mV dec−1 , similarly to what is found in both rate and reaction pathway with the
in the literature for the O2 reduction addition of another metal to the surface,
on a polycrystalline Pt electrode [95]. A usually in a highly structured manner, can
number of theories have been proposed provide some insight into the ORR on met-
to explain the transition in the Tafel als that are otherwise difficult to study, for
slope on polycrystalline Pt electrodes, example, Cu, Pb, and Bi. Only the results
on Cuupd -modified surfaces are reviewed process, the ORR currents at the Cuupd -
here because trace levels of copper may be modified surfaces can also be measured
present in real systems, mostly as a corro- under steady state conditions, see the
sion product, thus the effects of Cu UPD closed circles in Fig. 30. In these exper-
on Pt will serve as a model system to un- iments, the ORR is more inhibited than
derstand why a small amount of Cu UPD in the potentiodynamic experiments. As
has a devastatingly inhibiting effect on the shown in Fig. 30, the deactivation is more
ORR. The first result for the ORR on the pronounced in the presence of Cl− ions,
Pt(111)-Cuupd electrode was published by presumably due to stronger interaction of
Abe and coworkers [76]. By using STM and specifically adsorbed chloride (Clad ) than
hanging meniscus rotating disk electrode bisulfate (H2 SO4(ad) ) anions with the Pt-
(HMRDE) methods, the authors demon- Cuupd surface. Analysis of the RRDE data
strated a close correlation between the (Fig. 30) also revealed that the mechanism
inhibition of the ORR with the microstruc- for the ORR on Cuupd -modified electrodes
ture of Cuupd adlayer. In sulfuric acid is the same as on unmodified Pt(111),
solution, threefold hollow platinum sites that is, proceeds mostly as a 4e− re-
are still accessible for O2 adsorption inside action pathway with negligible solution
√ √
the honeycomb ( 3 × 3)R30◦ structure phase peroxide formation, (ca. 3%). It
that forms at ca. 0.5 V, but the reduced is surprising, however, that a relatively
geometry favors 2e− reduction to peroxide small amount of Cuupd has such a dev-
versus 4e− reduction to water. The cor- astating effect on the rate of ORR. This
relation between the surface structure of anomalously large inhibition by a very
Cuupd and the kinetics of the ORR can small amount of Cuupd is attributed to
be examined by using a combination of enhanced anion adsorption on platinum
RRDE and X-ray scattering measurements atoms adjacent to Cuupd atoms [98], see
(the X-ray diffraction measurements are Sect. 4.1.4.2 and Fig. 27. The predominant
described in Sect. 4.1.4.2). These results re- role of anions in determining the rate of the
veal that there is no correlation between the ORR on Cuupd -modified platinum single-
ORR kinetics and the ordered Cuupd sur- crystal surfaces is supported by the fact
face structures irrespective of the nature that the activity of the ORR decreases in
of the specifically adsorbing anions [98]. the same sequence as the Cu-anion bond
The effect of Cuupd on the ORR at the strength increases: Pt–Cuupd –HSO4,ad
Pt(111) interface in 0.05 M H2 SO4 and Pt–Cuupd –Br− . The model that rational-
0.05 M H2 SO4 + 10−3 M Cl− are shown izes these results is one in which the active
in Figs. 30(a, b). For comparison, the po- sites for the adsorption of molecular O2
larization curves for the ORR in Cu2+ -free are the small number of platinum ‘‘is-
solutions under the same experimental lands’’ created in state 2 in Fig. 27. It is
conditions are shown as dotted curves. important to note that in the Cu UPD
Clearly, in a solution containing Cu2+ potential region, Cuupd is always covered
ions, the ORR is strongly inhibited in both with anions, even at the most negative
Cl− -free and Cl− -containing solutions. potentials at which Pt(001) and Pt(111)
Very similar behavior was also observed are covered by nominally 1 ML of Cuupd .
on a polycrystalline electrode [99, 100] and It is obvious, therefore, that is impossi-
Pt(111) [76, 98] in different acid solutions. ble to determine the true catalytic activity
Given that Cu UPD is a rather slow of an anion-free pseudomorphic (1 × 1)
350
1
1.0
Pt(111)
Pt(111)
0.5
ΘCu
ΘCu
100 µA cm−2
50 µA cm−2
0 0.0
Ei
2 Interfacial Structure and Kinetics
20 0.05 M H2SO4 + 8.10−5 M Cu2+ 20 0.05 M H2SO4 + 8.10−5 M

0.05 M H2SO4 Cu2+ + 1.10−3 M Cl−
10 10
IR
IR
0.05 M H2SO4 + Cl−
[µA]
[µA]
×5
0 0
0 0
−2 −2
iD
iD
−4 −4
[mA cm−2]
[mA cm−2]
−6 −6
0.2 0.4 0.6 0.8 1.0 0.2 0.4 0.6 0.8 1.0
E E
(a) [VRHE] (b) [VRHE]
Fig. 30 The effect of Cuupd on the ORR at the Pt(111) interface in (a) 0.05 M H2 SO4 and (b) 0.05 M H2 SO4 + 10−3 M Cl− , represented at the bottom of
(a) and (b). In both cases, the solution contains 10−4 M Cu2+ . The corresponding Cu coverages (open circles) obtained from RRDE experiments (disk
currents are shown by solid lines) are represented at the top.
100 100
0 0
Idisk
Idisk
[µA]
[µA]
−10
−0.8 −0.4 0.0 0.0
−−2 0.8 −0.4
50 50
Iring
Iring
Edisk vs SCE Edisk vs SCE
[µA]
[µA]
[V] [V]
(b) 0 0 (d)
0.0 0
Cu(111) Cu(001)
@ 1600 rpm @ 1600 rpm
−0.5
−1
−1.0
−2
Idisk
Idisk
[mA]
[mA]
−1.5
0.5 M H2SO4 0.5 M H2SO4 −3

−2.0 0.1 M HClO4 0.1 M HClO4
−2.5 −4
−0.8 −0.6 −0.4 −0.2 0.0 −0.8 −0.6 −0.4 −0.2 0.0
Edisk vs SCE Edisk vs SCE
(a) [V] [V] (c)
Fig. 31Polarization curves (positive sweeps) of the ORR on (a, b) Cu(111) and (c, d) Cu(001) electrodes in 0.5 M H2 SO4 (dashed lines) and 0.1 M
HClO4 (solid lines). Inserts: The corresponding CVs obtained on the disk electrodes in 0.1 M HClO4 in the absence of O2 .
4.1 Structure Relationships in Electrochemical Reactions
351
Cuupd monolayer. One way to overcome forms a particularly stable adlayer between
this problem is to study the ORR on Cu −0.3 < E < −0.1 V. In addition, by uti-
single-crystal surfaces. lizing a potentiostatic STM image, it was
shown that disordering of this structure
4.1.5.3 ORR on Cu(hkl) Surfaces coincides with the onset of formation of
In contrast with the extensive studies of the the cathodic peak, for example, with the
ORR on Pt(hkl) surfaces, there has been desorption of (bi)sulfate anions at −0.35 V
no substantial fundamental study of the in Fig. 31(a). Given that the onset of the
effects of anion adsorption on the kinetics ORR on the Cu(111) surface in Fig. 31(a)
of the ORR on Cu(hkl) surfaces. Very is observed at the same potential, it was
recently, by utilizing the RRDE technique, suggested that the kinetics and reaction
Brisard and coworkers [101] have shown pathway of the ORR is controlled by the
that the ORR on Cu(111) and Cu(001) number of Cu ‘‘bare’’ sites on which the
surfaces in sulfuric acid solution is a adsorption of O2 molecules may take place.
structure-sensitive process, see Fig. 31. Although structural information regarding
As for Pt(hkl), an interpretation of the the surface coverage of (bi)sulfate anions
variation in the activity of this process with on Cu(001) is lacking, it is reasonable to
the different low-index crystal surfaces suggest that the ORR pathway on Cu(001)
of Cu can be presented on the basis is the same as on Cu(111).
of the premise of the structure-sensitive The overall inhibiting effect of anion
adsorption of sulfuric acid anions on adsorption on the kinetics of the ORR
Cu(hkl) surfaces, for example, as for Pt(hkl) was determined by comparing the ORR on
with the (1 − %ad ) term. Cu(111) and Cu(001) in 0.5 M H2 SO4 and
A comparison of the polarization curves 0.1 M HClO4 [101]. Figure 31(a) shows
(positive sweeps) of the ORR on Cu(111) that in 0.1 M HClO4 the initial reduction of
and Cu(001) at the same rotation rate O2 molecules is associated with production
(1600 rpm), Fig. 31, shows relatively small, of peroxide on the ring electrode (Fig. 31b).
for example, less than a factor of 2, differ- A comparison of the voltammetric curve of
ences in activity, but very different reaction Cu(111) in 0.1 M HClO4 solution free of
pathways below −0.4 V. Since very little O2 (inset of Fig. 31b) with the polarization
is known about the nature of adsorbed curves for the ORR in Fig. 31(a), sug-
anions on Cu(hkl) in sulfuric acid solu- gests that the amount of peroxide formed
tion, or how the adsorption sites might on the disk electrode is determined by
differ between the different surfaces, it is the potential-induced change in the sur-
difficult to rationalize the structural sen- face coverage of OHad on the Cu(111)
sitivity of the ORR on Cu(hkl). For the surface [101]. Thus, the decrease in the
Cu(111)-SO4,ad interface, both Wilms and surface coverage by OHad on Cu(111) re-
coworkers [102] and Li and Nichols [103] sults in an increase in the rate of the ORR
found by STM that a complex incommen- on the Cu(111) surface, as observed for
surate ordered structure of sulfate anions the ORR on Pt(hkl) in the potential re-
formed at the Cu(111) surface through- gion of the structure-sensitive adsorption
out the −0.1 to −0.4 V potential region in of OHad [97]. Further inspection of Fig. 31
H2 SO4 . This ordered structure does not shows that the combined kinetic-diffusion
appear to lie along any high symmetry control of the ORR on Cu(111) and Cu(001)
directions of the Cu(111) substrate but it is followed with the smooth transition to a
well-defined diffusion-limiting current for the reason for this remarkable effect re-
the 4e− reduction (−0.65 < E < −0.3 V), mains unclear. It was reported [104] that
consistent with an absence of peroxide on the unique catalytic activity of the Au(001)
the ring in this potential region. The fact surface only appears in the potential re-
that the ORR on Cu(111) and Cu(001) is gion in which OHad is adsorbed on the
orders of magnitude lower in sulfuric acid surface, see Fig. 32. Figure 32 also shows
solution than in pure perchloric acid, in- that at potentials negative of the potential
dicates that the compact layer of strongly for Au(001)-OHad formation, the surface
adsorbed bisulfate anions suppresses the provides only 2e− reduction. On the ba-
initial adsorption of the O2 molecule on sis of this evidence, it was proposed that
these surfaces, for example, the kinetics of the change in the catalytic activity of the
the ORR on Cu(hkl) is determined by the Au(001) surface from 4e− to 2e− reduc-
(1 − %ad ) term in Eq. (12). tion was due to the existence of a AuOH
layer. On the other hand, it was also pro-
4.1.5.4 ORR on Au(hkl) Surfaces posed that the higher activity of Au(001)
The ORR on gold single-crystal surfaces, for the ORR is related to the dissociative
was also found to be structure sensitive adsorption of HO2 − anions on the four-
in both alkaline and acid solutions. In fold symmetry sites that occur only on the
fact, the first observation that the rate of (001)-(1 × 1) surface [105]. Consequently,
the ORR is structure sensitive was made the change in the oxygen-reduction path-
from measurements of the ORR on gold way from 4e− to 2e− reduction is due
single-crystal surfaces in alkaline solu- to potential-induced reconstruction of the
tion [104, 105]. In acid solutions [105, 106], (1 × 1) surface to a structure similar to the
the order of activity increases in the ‘‘hex’’ reconstruction found in UHV. The
sequence: Au(111) < Au(110) < Au(001). results in Sect. 4.1.3.2.1 (Fig. 7) indicate
In contrast with Pt(hkl) and Cu(hkl) sur- that the (1 × 1) ↔ ‘‘hex’’ reconstruction
faces, the ORR on Au(hkl) surfaces in is not the dominant mechanism for this
acid solution is a very slow process. Be- change in reaction pathway since the tran-
cause of the very weak interaction of O2 sition in reduction kinetics occurs in a
molecules with gold, two-electron reduc- much narrower potential range than the
tion with 100% production of peroxide (1 × 1) ↔ ‘‘hex’’ transition. Furthermore,
was commonly observed on all Au(hkl) the time constant for the two transitions
surfaces. The ORR on Au(hkl) in alkaline differs by approximately two orders of mag-
solution represents one particularly in- nitude, with the structural transition being
teresting example of a structure-sensitive the slower.
reaction. In independent studies by two
research groups [104, 105], the Au(001) 4.1.5.5 Summary of the ORR
surface in alkaline solution was found Just as in heterogeneous catalysis, the
to be the most active of the low-index ultimate challenge in electrocatalysis is
surfaces (by nearly four orders of mag- to relate the microscopic details of ad-
nitude) and was reported as even more sorbed states of reaction intermediates
active than polycrystalline platinum under to the macroscopic measurement of ki-
the same experimental conditions. In spite netic rates. There are many strategies
of efforts by many research groups to de- that may be employed in this endeavor.
termine the origin of this catalytic activity, Here we develop the relation between
0
Current
100 µA cm−2
900
1600 2 mA cm−2
2500 rpm
1.4
Reconstruction fraction
Intensity (normalized)
1.2
1.0 0.00
0.8 0.25
0.50
0.6
0.75
0.4
1.00
−1.2 −0.8 −0.4 0.0 0.4 0.8
E
[V]
Fig. 32 Top: Polarization curves for the ORR on Au(001) in 0.1 M
KOH at different rotation rates. The dashed line shows the CV.
Bottom: Corresponding changes in the X-ray scattering signal at (0,
0, 2.3) where the scattered intensity depends on the presence of the
surface reconstruction, that is, deconstruction leads to an increase
in the intensity. The fraction of the surface that is reconstructed is
shown on the right-hand axis.
the kinetics of the ORR and two fac- (the work function, for example, the pzc)
tors: a chemical-dependent part, 1G∗θ = of the substrate and the match between
1G∗0 + γ 1G0% %ad , and the preexponen- the surface atomic geometry of the sin-
tial (1 − %ad ) term. The first term is mainly gle crystals and the geometry of adsorbing
determined by the nature of interaction spectator species (Hupd , OHad , and an-
of O2 molecules and reaction intermedi- ions), that is, by the ability of the surface
ates with the electrode surface, that is, to adsorb molecular O2 . In the following
by the M–O2 energetics and the ability discussion, these two terms are used to
of the substrate (M) to break the O−O analyze the difference in the kinetics of
bond. The preexponential (1 − %ad ) term the ORR on Pt(hkl), Au(hkl), and Cu(hkl)
is, however, mainly determined by the in- and the structure sensitivity of the ORR on
terplay between the electronic properties M(hkl) surfaces. One should keep in mind,
however, that a compensation effect exists between surface atomic geometry of the
between the preexponential factor and the single crystals and the geometry of adsorb-
heat of activation. ing spectator species. The importance of
Given that the energetics of O2 mole- the pzc in controlling the (1 − %ad ) term
cules and intermediates on Pt(hkl), Au(hkl), is obvious from the Pt(hkl)–Clad system
and Cu(hkl) in solution is not known, the in which, due to the high anodic pzc of
1G∗θ = 1G∗0 + γ 1G0% %ad term cannot be Pt(111) and the ability of this surface to
determined on these electrodes. However, desorb Clad at the most positive potentials,
one can use the values of thermodynamic this surface is the most active for the ORR.
functions that are obtained from UHV In sulfuric acid solution, however, an ex-
measurements and test them in the elec- ceptionally large deactivation is observed
trochemical system for consistency. Two at the Pt(111) surface. For this system, the
general features in the energetics of the (1 − %ad ) term is affected more with the
O2 interaction clearly emerge from the symmetry factors than with the pzc of the
UHV studies: (i) although the heat of surface; for example, due the strong ad-
adsorption of oxygen increases in the sorption of the (bi)sulfate anion, which is
order Pt(hkl) < Cu(hkl)<<<Au(hkl), the induced by the geometrical match between
heats of O2 adsorption are relatively in- the hexagonal (111) surface and the C3v ge-
sensitive to the surface structure of the ometry of the oxygen atoms of the sulfate
substrate [107, 108]; and (ii) the energet- anions, the ORR is strongly deactivated on
ics of interaction are strongly coverage the Pt(111) surface. The exceptionally low
dependent [23]. The coverage dependence activity of Cu(111) and Au(111) in sulfuric
on single-crystal surfaces arises primar- acid solution is also related to the strong in-
ily from adsorbate–adsorbate repulsive teraction of bisulfate anions with the (111)
interactions rather than surface hetero- surface atoms.
geneity [109]. As we demonstrate below, The most direct probe of the pzc
on the basis of ‘‘UHV-like’’ arguments effects in the kinetics of the ORR can
for the M-O2 energetics and the potential- be obtained by comparing the catalytic
dependent surface coverage of spectator activities of Pt(hkl) versus Cu(hkl). On
species (from Sects. 4.1.3 and 4.1.4), one both surfaces, in solution free of strongly
can develop a very reasonable relation be- adsorbing anions, the reaction mechanism
tween the kinetics of the ORR and the is the same (the 4e− reduction), the
parameters 1G∗% , and (1 − %ad ). formation of peroxide appears to precede
We first discuss the structure sensitivity the breaking of the O−O bond [91]. Owing
on the same metal substrates. The fact that to the relatively high energy of adsorption
the kinetics of the ORR on Pt(hkl) (Figs. 28 of oxygen on Cu (ca. 200 kJ mol−1 on
and 29), Cu(hkl) (Fig. 31), and Au(hkl) Cu vs. 350 kJ mol−1 on Pt [23]), the
(Fig. 32) are orders of magnitude lower exceptionally large difference in catalytic
in acid solutions containing strongly ad- activity between Cu and Pt cannot be
sorbing anions indicates that the structure explained by a difference in the isosteric
sensitivity is due primarily to the structure- heat of adsorption of O2 on these two
sensitive adsorption of spectator species, surfaces. We propose, therefore, that the
for example, by the (1 − %ad ) term. The main contribution to the deactivation of
(1 − %ad ) term is determined by both the the ORR on Cu(hkl) is due to the (1 − %ad )
pzc of the substrate as well as the match term, which is controlled by the low
pzc of Cu, leading to high coverages of attain surface sensitivity. We have high-
oxide/anions at low overpotentials and the lighted the applicability of this technique to
inability of Cu to adsorb O2 molecules. studies of the electrochemical interface, in
It is interesting that although the pzc of particular, with the ability to probe changes
Au is much higher than on Cu, Au(hkl) is in the surface atomic structure as the
catalytically inactive in acid solution and electrode potential is changed and differ-
the ORR proceeds as a 2e− process [105]. ent reaction conditions prevail. The X-ray
This can be attributed to the very low heat technique is particularly powerful when
of adsorption of O2 on Au, which is well combined with RRDE measurements be-
established in UHV [23]. Therefore, in acid cause RRDE can supply element-specific
solutions, the main difference in the ORR adsorption isotherms, which are key to
pathway on Pt(hkl) and Cu(hkl) (ca. 4e− the interpretation of the X-ray data. In
reduction) versus Au(hkl) (2e− reduction) turn, structural results obtained from X-
lies rather in the chemical 1G∗θ term ray diffraction measurements can give
than the (1 − %ad ) term. While the 1G∗θ vital insight into the kinetics of electro-
term is determining the reaction pathway, chemical reactions, such as the oxida-
the (1 − %ad ) term is determining the tion–reduction reaction, which are cru-
availability of Au sites for O2 adsorption. cially dependent on the exact nature of
Because of the high pzc of Au, desorption the available adsorption sites on an elec-
of anions/oxides occurs at more positive trode surface.
potentials than on Cu and, in turn, the Several future developments of the
onset potential for the ORR is more positive experimental techniques and materials are
on Au(hkl) than on Cu(hkl) surfaces. What anticipated. For example, the study of
is influencing the electrocatalytic activity of bimetallic crystals relevant to applications
Au(001) in alkaline solution is, however, in electrocatalysis, nanoparticle systems,
still a mystery. Further work is needed and further applications of spectroscopic
in order to understand the ability of the techniques in combination with X-ray
Au(001) surface to break the O−O bond at scattering studies. The work highlighted in
such low overpotentials. this chapter lays the foundation for these
developments.
4.6
Conclusion Acknowledgment
In this chapter, we have attempted to show This work was supported by the Direc-
that in situ structural analysis techniques tor, Office of Energy Research, Materials
and traditional electrochemical measure- Sciences Division (MSD) of the U.S.
ments can be combined to obtain detailed Department of Energy (DOE) under con-
insight into electrochemical reactions on tract no. DE-AC03-76SF00098. Research
single-crystal electrode surfaces. Thus, it was carried out in part at SSRL, which
is possible to establish a link between is funded by the Division of Chemical
macroscopic electrochemical phenomena Sciences (DCS), U.S. DOE and at the
and atomic-scale surface structures. X-ray XMaS UK-CRG beamline at the ESRF,
diffraction is a relatively new technique Grenoble. CAL is grateful for the sup-
that takes advantage of the high bright- port of an EPSRC Advanced Research
ness of synchrotron radiation sources to Fellowship.
References 25. R. I. Masel, Principles of Adsorption and

Reaction on Solid Surfaces, John Wiley &
1. R. Feidenhans’l, Surf. Sci. Rep. 1989, 10, Sons, New York, 1996.
105. 26. M. A. Van Hove, K. Hermann, P. R. Watson
2. P. H. Fuoss, S. Brennan, Annu. Rev. Mater. et al., Surf. Rev. Lett. 1999, 6, 805.
Sci. 1990, 20, 365. 27. A. R. Sandy, S. G. J. Mochrie, D. M. Zehner
3. I. K. Robinson, D. J. Tweet, Rep. Prog. Phys. et al., Phys. Rev. B 1991, 43, 4667.
1992, 55, 599. 28. K.-M. Ho, K. P. Bohnen, Phys. Rev. Lett.
4. I. K. Robinson, Phys. Rev. B 1986, 33, 3830. 1987, 59, 1833.
5. B. E. Warren, X-ray Diffraction, Dover Pub- 29. I. K. Robinson, P. J. Eng, C. Romainczyk
lications, New York, 1990. et al., Phys. Rev. B 1993, 47, 10 700.
6. M. G. Samant, M. F. Toney, G. L. Borges 30. B. M. Ocko, D. Gibbs, K. G. Huang et al.,
et al., J. Phys. Chem. 1988, 92, 220. Phys. Rev. B 1991, 44, 6429.
7. B. M. Ocko, J. Wang, A. Davenport et al., 31. D. Gibbs, B. M. Ocko, D. M. Zehner et al.,
Phys. Rev. Lett. 1990, 65, 1466. Phys. Rev. B 1990, 42, 7330.
8. I. M. Tidswell, N. M. Markovic, C. A. Lucas 32. N. Takeuchi, C. T. Chan, K. M. Ho, Phys.
et al., Phys. Rev. B 1993, 47, 16 542. Rev. Lett. 1989, 63, 1273.
9. C. A. Lucas, J. Phys. D: Appl. Phys. 1999, 32, 33. R. M. Ishikawa, A. T. Hubbard, J. Elec-
A198. troanal. Chem. 1976, 69, 317.
10. M. F. Toney, D. G. Wiesler, Acta Crystal- 34. E. Yeager, W. E. O’Grady, M. Woo, J. Elec-
logr., Sect. A 1993, 49, 624. trochem. Soc. 1978, 125, 348.
11. C. A. Lucas, N. M. Markovic, P. N. Ross, 35. I. M. Tidswell, N. M. Markovic, P. N. Ross,
Surf. Sci. 2000, 448, 65. J. Electroanal. Chem. 1994, 376, 119.
12. C. A. Lucas, N. M. Markovic, B. N. Grgur 36. J. Wang, B. M. Ocko, A. J. Davenport et al.,
et al., Surf. Sci. 2000, 448, 77. Science 1992, 255, 1416.
13. E. D. Specht, F. J. Walker, Phys. Rev. B 1993, 37. K. P. Bohnen, D. M. Kolb, Surf. Sci. 1998,
47, 13 743. 407, L629.
14. H. Stragier, J. O. Cross, J. J. Rehr et al., 38. F. T. Wagner, P. N. Ross, J. Electroanal.
Phys. Rev. Lett. 1992, 69, 3064. Chem. 1983, 150, 141.
15. W. J. Albery, M. L. Hitchman, Ring-Disc 39. M. S. Zei, N. Batina, D. M. Kolb, Surf. Sci.
Electrodes, Clarendon Press, Oxford, 1971. 1994, 306, L519.
16. A. J. Bard, L. R. Faulkner, Electrochemical 40. I. M. Tidswell, N. M. Markovic, P. N. Ross,
Methods, John Wiley & Sons, New York, Phys. Rev. Lett. 1993, 71, 1601.
1980. 41. C. A. Lucas, N. M. Markovic, P. N. Ross,
17. D. Kolb, Prog. Surf. Sci. 1996, 51, 109. Phys. Rev. Lett. 1996, 77, 4922.
18. J. Clavilier, J. Electroanal. Chem. 1980, 107, 42. N. M. Markovic, B. N. Grgur, C. A. Lucas
211. et al., Surf. Sci. 1997, 384, L805.
19. M. Samant, M. Toney, G. Borges et al., Surf. 43. I. K. Robinson, Y. Ku, L. C. Feldman, Phys.
Sci. 1988, 193, L29. Rev. B 1984, 29, 4762.
20. Y. S. Chu, I. K. Robinson, A. A. Gewirth, 44. T. Gritsch, D. Coulman, R. J. Behm, G. Ertl,
Phys. Rev. B 1997, 55, 7945. Phys. Rev. Lett. 1989, 63, 1086.
21. O. M. Magnussen, J. Scherer, B. M. Ocko 45. O. M. Magnussen, J. Wiechers, R. J. Behm,
et al., J. Phys. Chem. 2000, 104, 1222. Surf. Sci. 1993, 289, 139.
22. G. A. Somorjai, M. A. Van Hove, Prog. Surf. 46. B. M. Ocko, G. Helgeson, B. Schardt et al.,
Sci. 1989, 30, 201. Phys. Rev. Lett. 1992, 69, 3350.
23. G. A. Somorjai, Introduction to Surface 47. C. L. Fu, K. M. Ho, Phys. Rev. Lett. 1989, 63,
Chemistry and Catalysis, John Wiley & Sons, 1617.
New York, 1993. 48. P. J. Feibelman, Surf. Sci. 1996, 360, 297.
24. P. A. Thiel, P. J. Estrup in CRC Handbook 49. M. F. Toney, J. N. Howard, J. Richer et al.,
of Surface Imaging and Visualization (Ed.: Nature 1994, 368, 444.
A. T. Hubbard), CRC Press, Boca Raton, 50. M. F. Toney, J. N. Howard, J. Richer et al.,
Fla., 1995. Surf. Sci. 1995, 335, 326.
51. R. J. Nichols, T. Nouar, C. A. Lucas, 76. T. Abe, G. M. Swain, K. Sashikata et al., J.

W. Haiss, W. A. Hofer, Surf. Sci. 2002, 513, Electroanal. Chem. 1995, 382, 73.
263–271. 77. N. M. Markovic, C. Lucas, H. A. Gasteiger
52. E. Vlieg, I. K. Robinson and K. Kern, Surf. et al., Surf. Sci. 1997, 372, 239.
Sci. 1990, 233, 248. 78. A. C. Finnefrock, L. J. Buller, K. L. Ringland
53. M. W. Finnis, V. Heine, J. Phys. F 1974, 4, et al., J. Am. Chem. Soc. 1997, 119,
L37. 11 703.
54. E. Kirsten et al., Surf. Sci. 1990, 231, L183. 79. N. M. Markovic, H. Gasteiger, C. Lucas
55. O. M. Magnussen, B. M. Ocko, J. X. Wang et al., Surf. Sci. 1995, 335, 91.
et al., J. Phys. Chem. 1996, 100, 5500. 80. I. M. Tidswell, C. A. Lucas, N. M. Markovic
56. B. M. Ocko, O. M. Magnussen, J. X. Wang et al., Phys. Rev. B 1995, 51, 10 205.
et al., Physica B 1996, 221, 238. 81. C. A. Lucas, N. M. Markovic, P. N. Ross,
57. B. M. Ocko, O. M. Magnussen, J. X. Wang Phys. Rev. B 1997, 56, 3651.
et al., Phys. Rev. B 1996, 53, R7654. 82. C. A. Lucas, N. M. Markovic, P. N. Ross,
58. E. Bertel, E. P. Netzer, Surf. Sci. 1980, 97, Phys. Rev. B 1998, 57, 13 184.
409. 83. S. Wu, Z. Shi, J. Lipkowski et al., J. Phys.
59. H. A. Gasteiger, N. M. Markovic, P. N. Chem. 1997, 101, 10 310.
Ross, Langmuir 1996, 12, 1414. 84. K. Kinoshita, Electrochemical Oxygen Tech-
60. C. A. Lucas, N. M. Markovic, P. N. Ross, nology, John Wiley & Sons, New York,
Surf. Sci. 1996, 340, L949. 1992.
61. C. A. Lucas, N. M. Markovic, P. N. Ross, 85. H. Wroblowa, Y. C. Pan, J. Razumney, J.
Phys. Rev. B 1997, 55, 7964. Electroanal. Chem. 1976, 69, 195.
62. P. Zeppenfeld, U. Becher, K. Kern et al., 86. M. R. Tarasevich, A. Sadkowski, E. Yeager
Phys. Rev. B 1992, 45, 5179.
in Comprehensive Treatise in Electrochem-
63. N. M. Markovic, C. A. Lucas, H. A. Gastei-
istry (Eds.: J. O. M. Bockris, B. E. Conway,
ger et al., Surf. Sci. 1996, 365, 229.
E. Yeager et al.), Plenum Press, New York,
64. A. M. Bittner, B. Wintterline, B. Beran et al.,
1983.
Surf. Sci. 2001, 335, 291.
87. N. M. Markovic, H. A. Gasteiger, B. N.
65. D. M. Kolb in Advances in Electrochem-
Grgur et al., J. Electroanal. Chem. 1999, 467,
istry and Electrochemical Engineering (Eds.:
157.
H. Gerischer, C. W. Tobias), John Wiley &
Sons, New York, 1978. 88. F. El Kadiri, R. Faure, R. Durand, J. Elec-
66. R. R. Adzic in Advances in Electrochem- troanal. Chem. 1991, 301, 177.
istry and Electrochemical Engineering (Eds.: 89. N. M. Markovic, R. R. Adzic, B. D. Ca-
H. Gerischer, C. W. Tobias), John Wiley & han et al., J. Electroanal. Chem. 1994, 377,
Sons, New York, 1984. 249.
67. M. F. Toney, J. N. Howard, J. Richer et al., 90. N. M. Markovic, H. A. Gasteiger, P. N.
Phys. Rev. Lett. 1995, 75, 4472. Ross, J. Phys. Chem. 1995, 99, 3411.
68. Z. Shi, J. Lipkowski, J. Electroanal. Chem. 91. N. M. Markovic, H. A. Gasteiger, P. N.
1994, 364, 289. Ross, J. Phys. Chem. 1996, 100, 6715.
69. O. M. Magnussen, J. Hotlos, R. J. Nichols 92. V. Stamenkovic, N. M. Markovic, P. N.
et al., Phys. Rev. Lett. 1990, 64, 2929. Ross Jr, J. Electroanal. Chem. 2000, 500,
70. J. H. White, H. D. Abruna, J. Phys. Chem. 44.
1990, 94, 894. 93. B. N. Grgur, N. M. Markovic, P. N. Ross Jr,
71. R. Michaelis, M. S. Zei, R. S. Zhai et al., J. Can. J. Chem. 1997, 75, 1465.
Electroanal. Chem. 1992, 339, 299. 94. T. J. Schmidt, B. N. Grgur, N. M. Marko-
72. R. Durand, R. Faure, D. Aberdam et al., vic et al., J. Electroanal. Chem. 2001, 500,
Electrochim. Acta 1992, 37, 1977. 36.
73. N. Markovic, P. N. Ross, Langmuir 1993, 9, 95. A. Damjanovic in Comprehensive Treatise
580. in Electrochemistry (Eds.: J. O. M. Bockris,
74. N. M. Markovic, H. A. Gasteiger, P. N. B. E. Conway), Plenum Press, New York,
Ross, Langmuir 1995, 11, 4098. 1969.
75. R. Gomez, H. S. Yee, G. M. Bommarito 96. M. R. Tarasevich, Electrochimiya 1973, 9,
et al., Surf. Sci. 1995, 335, 101. 578.
97. N. M. Markovic, P. N. Ross Jr in Interfa- 104. R. R. Adzic, N. M. Markovic, V. B. Vesovic,

cial Electrochemistry-Theory, Experiments and J. Electroanal. Chem. 1984, 165, 105, 121.
Applicatoions (Ed.: A. Wieckowski), Marcal 105. J. D. E. McIntyre, Peck in The Physics
Dekker, New York, 1999. and Chemistry of Electrocatalysis (Eds.:
98. N. M. Markovic, B. N. Grgur, C. A. Lucas, J. D. E. McIntyre, M. J. Weaver, E. Yeager),
P. N. Ross, Electrochim. Acta. 1998, 44, The Electrochemical Society, Pennington,
1009–1017. N.J., 1984.
99. G. Kokkinidis, D. Jannakoudakis, J. Elec- 106. R. R. Adzic in Electrocatalysis (Eds.: J. Lip-
troanal. Chem. 1984, 162, 163. kowski, P. N. Ross), Wiley-VCH, New York,
100. S. A. S. Machado, A. A. Tanaka, E. R. Gon- 1998.
zales, Electrochim. Acta 1991, 36, 1325. 107. G. Ertl, N. Neumann, K. M. Streit, Surf. Sci.
101. G. M. Brisard, N. Bertrand, P. N. Ross et al., 1977, 64, 393.
J. Electroanal. Chem. 2000, 480, 219. 108. P. A. Thiel, R. J. Behm, P. R. Norton et al.,
102. M. Wilms, P. Broekmann, C. Stuhlmann J. Chem. Phys. 1983, 78, 7448.
et al., Surf. Sci. 1998, 416, 121. 109. W. T. Lee, L. Ford, P. Blowers et al., Surf.
103. W. H. Li, R. J. Nichols, J. Electroanal. Chem. Sci. 1998, 416, 141.
1998, 456, 153.
4.2 on models within the framework of the

Theoretical Aspects Associated with transition state theory. On the other hand,
Charge-transfer Kinetics across Interfaces ion transfer will be examined not only from
between Two Immiscible Electrolyte the viewpoint of charge transfer in polar
Solutions media but also in terms of ion transport in
heterogeneous systems.
David J. Fermı́n, Henrik Jensen, and Hubert
H. Girault
4.2.2
Ecole Polytechnique Fédérale de Lausanne,
Heterogeneous Electron Transfer across
Lausanne, Switzerland
ITIES
4.2.1
Introduction and Scope ET reactions across a liquid|liquid junc-
tion are undergoing a renaissance as a
Charge-transfer reactions at interfaces be- result of the increasing interest in molecu-
tween two immiscible electrolyte solu- lar electrochemistry. The pioneering work
tions (ITIES) have been the subject of by Samec [6, 7], Kharkats [8], Girault and
a tremendous amount of research from Schiffrin [9], and later by Marcus [10–12]
an experimental and theoretical point of established fundamental aspects that ex-
view in the last thirty years. The relevance perimental evaluation is yet to ascertain.
of this kind of molecular interfaces has For instance, the dependence of the bi-
been addressed in various review articles molecular ET rate constant on the Galvani
and books, expanding from fundamental potential difference and the distance sep-
charge transfer at interfaces to drug de- arating the redox species are among the
livery in biological systems, ion-selective most controversial issues in this area.
electrodes, sensors, artificial photosynthe- Throughout the 1980s and early 1990s,
sis, and so forth [1–5]. The scope of this few experimental works were reported
chapter is to review the theoretical models on the dynamics of ET as probed by
developed in recent years for rationalizing cyclic voltammetry and electrochemical
the kinetic behavior of electron and ion impedance spectroscopy employing four-
transfer across ITIES. This chapter does electrode potentiostats [13–21]. However,
not intend to provide an exhaustive ac- the analysis of these data was rather limited
count of experimental developments and because of the primitive knowledge of the
specific models but an overview of the structure and potential distribution across
more general theoretical descriptions in a ITIES. More recently, the introduction of
critical fashion. In doing so, we have made spectroelectrochemical techniques, Scan-
an attempt to present the basic description ning Electrochemical Microscopy (SECM),
of each model employing a homogenized and dynamic photoelectrochemistry has
terminology. provided new insights into the correlation
The discussions will be mostly centered between structure and reactivity at these
on ideal systems in which the reactants interfaces. As a result, not only has our un-
are considered as particles with given derstanding of the fundamental properties
redox properties and solvation energies of ITIES significantly increased but we are
present at the interfacial region between also able to envisage novel potential ap-
two dielectric media. In the case of plications as well as rationalize important
electron transfer (ET), we shall concentrate aspects concerning life-sustaining redox
4.2 Charge-transfer Kinetics across Interfaces between Two Immiscible Electrolyte Solutions 361
biological processes [1]. In this section, we ferricenium/ferrocene (Fc+ /Fc) has been
shall highlight fundamental aspects in re- employed in numerous studies in organic
lation to the energetic and dynamics of solvents as a reference potential [22]. In
heterogeneous ET bringing about some the case of the water|DCE junction, the
relevant examples. following reaction can be envisaged:
4.2.2.1 Thermodynamic Aspects in

1
2 H2(w) + Fc+ +
(DCE) −−−→ H(w) + Fc(DCE)
Connection to Heterogeneous ET (4)
The thermodynamic driving force for an where the subscript ‘‘w’’ and ‘‘DCE’’ corre-
ET process across liquid|liquid junctions is spond to the water and 1,2-dichloroethane
determined by the difference in the redox electrolyte phases, respectively. In order to
potential of both species as well as the estimate the formal Gibbs energy for Re-
Galvani potential difference. Considering action (4), the following thermodynamic
the general reaction cycle can be introduced:
−→
kobs Fc+ +
(DCE) −−−→ Fc(w) (5)
◦ −−−
−−
→ ◦
Ow + R2 ← − Rw + O2 (1)
1
←−
1 1
2 H2(w) + Fc+ +
(w) −−−→ H(w) + Fc(w) (6)
kobs
Fc(w) −−−→ Fc(DCE) (7)
the equality in the electrochemical po-
tential determines that the ratio of the From Eqs. (4 to 7), it follows that
activities of the redox species is linked to
◦ ◦ ◦ ◦ ,w
the Galvani potential difference by EFc,+ /Fc = w o φFc+ + EFc+ /Fc
SHE SHE
w ◦
RT cO1 cR2 w→◦
w ◦ GFc
w o φ = φ
o et − ln w c◦ (2) − (8)
nF cR1 O2 F
◦
where the formal potential, w o φet , is
where the three terms correspond to the
determined by formal transfer potential of ferricinium,
the formal redox potential of Fc+ /Fc in wa-
◦ ◦ ,◦ ◦ ,w
wo φet = EO2/R2 − EO1/R1 ter versus the SHE, and the Gibbs energy
SHE SHE
(3) associated with the partition of ferrocene
By definition, both redox potentials in between the two phases, respectively. In-
Eq. (3) have to be referred to a common dependent evaluation of the three con-
◦ ◦
reference potential, in this case that of the stants provided EFc,+ /Fc = 0.638 ± 0.020 V
standard hydrogen electrode (SHE). Obvi- versus SHE [5]. Taking this value as a refer-
ously, some thermodynamic assumptions ence, the redox potential of several couples
have to be introduced in order to evalu- in DCE often found in the literature can
◦ ,◦
ate [EO2/R2 ]SHE , which denotes the formal be summarized as exemplified in Table 1.
redox potential of a redox couple in the These data allow predicting whether the
organic phase versus the SHE in water. formal ET potential as defined by Eq. (3)
A simple approach to this problem is effectively located within the ideally
◦ ,◦
involves the estimation of [EO2/R2 ]SHE for polarizable window. This condition, also
a given redox couple that can be further referred to as redox potential matching,
used as a reference potential in the organic is essential for studying ET processes at
phase. For instance, the redox couple ITIES under potentiostatic control.
Tab. 1 Formal redox potential versus SHE in The condition of redox potential match-
water for species often employed in ing is illustrated in Fig. 1. Considering the
electrochemical studies at the
concentration ratio of ferri/ferrocyanide
water|1,2-dichloroethane interface
indicated in the cell representation and
◦ ,◦
EFe(CN) 3− /Fe(CN) 4− as 0.36 ± 0.01 V, the
[E◦ ,DCE ]SHE /V

Redox couple 6 6
◦
value of w o φet in the presence of Fc
Decamethylferrocene+/0 0.07 should be located close to 0.16 V. The
Dimethylferrocene+/0 0.55 cyclic voltammogram features a response
Diferrocenylethane+/0 0.55 with a half-wave transfer potential that co-
Butylferrocene+/0 0.56 ◦
Ferrocene+/0 0.64
incides with w o φet . The signal observed
Trianysilamine+/0 0.76 at more negative potentials corresponds
Tri(bromophenyl)amine+/0 1.10 to the transfer of tetrapropylammonium
Tetraphenylborate0/− 0.93 employed as internal reference for the esti-
Tetrakis(4-chlorophenyl)- 1.15 mation of the Galvani potential difference
borate0/− ◦
Tetracyanoquinodimethane0/− −0.29
(wo φTPA+ = −0.093 V [23]). It should be
Chloranil0/− −0.17 noted that the half-wave potential asso-
Benzoquinone0/− −0.43 ciated with the ET across the ITIES is
not necessarily correlated to the formal
Fig. 1 Cyclic voltammogram at the

water|1,2-dichloroethane interface in the presence of
ferricyanide/ferrocyanide couple and ferrocene (Fc) at
10 mV s−1 . This experiment was performed with a
four-electrode potentiostat, while the electrochemical
cell can be represented as
Li2 SO4 1.5 M

3−
Fe(CN)6 0.1 M
Ag Ag2 SO4 Fe(CN)6 4− 0.01 M
TPA+ 10−4 M
(aq)
40
TPA+ BTPPATPBCl BTPPACl
20 0.01 M 0.001 M
0 Fc 4 × 10−3 M LiCl AgCl Ag
0.01 M
−20
106 i
[A]
ET
(DCE) (aq)
−40
−60 where BTPPA and TPBCl stand for

bis(triphenylphosphoranylidene)-ammonium and
−80 tetrakis-(4-chlorophenyl) borate, respectively. The
surface area of the liquid|liquid junction was 1.53 cm2 .
+
−0.2 0.0 0.2 0.4 The cation tetrapropylammonium (TPA ) was
w introduced in the aqueous as a reference for the
∆of Galvani potential difference. (Reprinted from Ref. [5]
[V] with permission from Marcel Dekker, INC.)
ET potential. This process involves the voltammograms obtained in the presence

diffusion of four species, consequently, of an excess of ferri/ferrocyanide are ex-
if the concentration of the reactants in emplified in Fig. 2(a). As expected for a
both phases is similar, then rather large quasi-reversible ET process, the peak cur-
peak-to-peak separations could be ob- rent increases linearly with the square root
served [24]. In the particular case of Fig. 1, of the scan rate. The spectral changes
the redox couple in the aqueous phase is obtained in situ by Total Internal Re-
largely in excess with respect to the or- flection from the DCE phase are also
ganic phase, therefore its concentration contrasted in Fig. 2(b) as a function of the
can be taken as constant and the ET Galvani potential difference. The spectra
rate is only controlled by the diffusion of evolution shows the bands associated with
tetracyanoquinodimethane (TCNQ). Un- the radical TCNQ− generated from the
der these conditions, commonly referred heterogeneous ET reaction. Indeed, the
to as the constant composition approxi- differential spectroscopic signal at 670 nm
mation, the aqueous phase behaves as a exhibits the same potential dependence of
‘‘solid electrode’’ with the Fermi level de- the current [26]. This close correlation be-
fined by the electrochemical potential of tween electrochemical and spectroscopic
the ferri/ferrocyanide couple. behavior indicates that the dynamic of
Another well-studied redox system at ET can be evaluated by both approaches.
liquid|liquid interfaces involves the elec- Furthermore, consistent data for this sys-
tron acceptor TCNQ [14, 15, 25, 26]. Cyclic tem has also been gathered employing
180
0
106 i
[A]
Fig. 2 Cyclic voltammograms of the

−180 v: 49, 36, 25, 16, 9 mV/s
heterogeneous reduction of
tetracyanoquinodimethane (TCNQ) by
the ferricyanide/ferrocyanide couple −0.24 −0.14 −0.04 0.06 0.16 0.26
across the water|DCE interface at w
various scan rates (a). The composition ∆o f
of the electrochemical cell is as in Fig. 1 (a) [V]
but with a ratio 0.01/0.4 M
Fe(CN)6 3− /Fe(CN)6 4− and 6 10−4 M
TCNQ in DCE. Absorption spectra
2.0
obtained in total internal reflection from
the DCE phase as the potential is swept
Absorption
1.5
from 0.190 to 0.260 V, with a difference
40 mV between each spectrum (b). The 1.0
spectra show the main features of the
radical TCNQ•− as the potential is 0.5
swept toward the reduction potential.
(Fig. 2a was reprinted from Ref. [25] 0.0
with permission from Royal Society of 600 700 800 900 1000
Chemistry. Fig. 2b was reprinted from
Ref. [26] with permission from Wavelength
Elsevier Science.) (b) [nm]
techniques based on SECM at nonpolariz- and

able interfaces [27, 28]. Consequently, we

shall adopt this system as an example for ←− ←−◦ (1 − αobs )nFwoφ
some of the discussions in connection to kobs = kobs exp −
RT
fundamental aspects on ET. (11)
where the preexponential parameters can
4.2.2.2 Electron-transfer Kinetics be evaluated at the standard potential
As in any interfacial redox process, the rate
◦
of heterogenous ET across ITIES is deter- ◦ −→◦ αobs nFwoφ
k = kobs exp
mined by the interfacial concentration of RT
the reactants and the phenomenological ◦ ◦
←− (1 − αobs )nFwoφ
bimolecular ET rate constant (kobs ). Ex- = kobs exp −
pressions concerning the mass transport RT
of species involved in the redox processes (12)
have been addressed for a variety of geome- From Eqs. (10 to 12), the faradaic current
tries, for example, planar diffusion [24], as a function of the applied potential can
spherical diffusion at microinterfaces [29], be expressed in terms of the concentration
as well as for ‘‘approach curves’’ familiar of each of the reactants at the respective
from techniques based on SECM [30–35]. interfacial reaction planes (position along
In this section, we shall concentrate in- the z-axis),
stead on the physical aspects underlying

the rate constant of heterogeneous ET. In ◦ w ◦
particular, the dependence of kobs on the I = nFAkobs cO1 (zO1 )cR2 (zR2 )
thermodynamic driving force as well as w ◦
αobs nF (wo φ − o φ )
the nature of the activation energy will be × exp
highlighted. RT

◦ w ◦
− nFAkobs cR1 (zR1 )cO2 (zO2 )
4.2.2.2.1 The Phenomenological Butler–
Volmer Expression Considering symmet- w ◦
−(1 − αobs )nF (wo φ − o φ )
ric energy surfaces for the initial and × exp
RT
final states of the reaction described in (13)
Eq. (1), the activation energy will be af- Equation (13) can be rewritten in terms of
fected by a fraction αobs of the Galvani the overpotential η = wo φ − o φeq , as
w
potential difference
w (z )c◦ (z )
Gact = Gact,eq − αobs (w w
o φ − o φeq ) cO1 O1 R2 R2
I = I0 w c◦
(9) cO1 R2
where Gact,eq and w φ
o eq correspond to
αobs nF η
the Gibbs activation energy and the Gal- × exp
vani potential difference at the equilibrium RT
state. The effect on the ET rate constants w (z )c◦ (z )
cR1 R1 O2 O2
can be phenomenologically described as − w c◦
cR1 O2

−→ −→◦ αobs nFw oφ −(1 − αobs )nF η
kobs = kobs exp (10) × exp (14)
RT RT
where substantial deviations from Eq. (16). An-

other parameter to be dealt with from first
◦ w ◦ ◦
I0 = nFAkobs [cO1 cR2 ]1−αobs [cO1
w
cR2 ]αobs basis is αobs . Does this parameter only re-
(15) flect the fraction of the Galvani potential
If we assume that the concentrations at difference acting on the Gact as indicated
the reaction planes are equal to that in in Eq. (9)?
the bulk, Eq. (14) is simplified to the
corresponding Butler–Volmer expression 4.2.2.2.2 The Rate Constant of Electron
for ET at ITIES, Transfer We shall commence by assum-
ing that the liquid|liquid interface can be
αobs nF η described in terms of a sharp bound-
I = I0 exp
RT ary between water and a lower dielectric
medium. The single ET process of Eq. (1)
−(1 − αobs )nF η
− exp (16) involves the approaching of the reactants
RT
to the interface as schematically repre-
This simplistic derivation opens up some sented in Fig. 3. The observed rate of ET is
important questions that we shall address given by
in the next sections. Firstly, the observed w ◦
jobs = kobs cO1 cR2 (17)
ET rate constant needs to be considered in
terms of microscopic parameters such as where kobs is the phenomenological bi-
the distance separating the redox species, molecular ET rate constant (units of
the adiabatic properties of the systems, cm4 mol−1 s−1 ). This rate constant effec-
and the thermodynamic driving force. If tively corresponds to an integral overall
the reactants feature ionic species, the con- distance l and the characteristic angles
centration profiles at the interfaces arising defining the position of the reactants at
from polarization effects can introduce the interface (Fig. 3). Following a Marcus
z
z
aR2 aR2
zR2 zR2
l
l qmax
z=0 z=0
q
a O1
zO1
a O1
(a) (b)
Fig. 3 (a) Schematic representation of the coordinates for the calculations of the hypervolume
as described by Marcus in Ref. [10]. (b) The maximum value of the angle θ for a given zR2 and l
arrangement.
type of analysis [7, 9–12], a distance close to aO1 + aR2 , and ex-
ponentially decrease as l increases for
−Gact larger distances. Consequently, taking
kobs = κνυ exp (18)
kB T k(l)/k(lmax ) = exp[−β(l − aO1 − aR2 )], it
follows that the integration in Eq. (19) re-
where κ is the transmission coefficient,
sults in
ν is the frequency of molecular motion,
and υ is the ‘‘volume’’ per unit of area
υ = 2π(aO1 + aR2 )β −3 (21)
in which the reactants are separated by
a distance (l + l). l is defined as the Equation (21) is associated with the lim-
region where the contribution to the total iting case for a sharp phase boundary
ET is maximized. As discussed later, the between both liquids. Marcus also esti-
activation free energy Gact is dependent mated the hypervolume for the case in
on the reorganization energy term as which both reactants are allowed to pene-
well as the work terms associated with trate the interfacial region in such a way
approaching reactants and separating the that their centers can effectively reach the
products. If we consider the arrangement liquid boundary. For such configuration,
depicted in Fig. 3, it follows that [10] the limits of the integrations in Eq. (19)
∞ l−aO1 θmax (zo ,l) 2π are different and υ approaches to
υ= l2
l=aR2 +aO1 zo =aR2 θ=0 =0 υ ≈ π(aO1 + aR2 )3 β −1 (22)
k(l)
× sin θ d dθ dzo dl If we consider that the sum of the ionic
k(lmax )
∞ radii is of the order of 10 Å and β = 1 Å−1 ,
=π (l − aR2 − aO1 )2 l it can be estimated that the hypervolume,
l=aR2 +aO1 and consequently the rate constant, is
k(l)dl approximately two orders of magnitude
× (19)
k(lmax ) larger for the case in which the ions are
allowed to penetrate the interfacial region.
where corresponds to the angle along the Marcus proposed that the analysis in-
plane of the interface. Assuming that the volving Eqs. (18 to 22) was also valid in
ions cannot penetrate the phase boundary, the adiabatic limit. Smith and cowork-
θ max can be defined as the maximum angle ers criticized this point on the basis of
θ obtained for a given value of z1 and l the fact that the transmission coefficient
as depicted in Fig. 3(b). In Eq. (19), the exhibits a complex dependent on the dis-
parameter k(l) is given by tance separating the redox species, and
therefore should be considered in the in-
−Gact (l)
k(l) = κ(l) exp (20) tegration step [36]. Let us consider the
kB T
Landau–Zener transmission coefficient
and lmax corresponds to the character- 2ς
istic distance at which the weighting κ= (23)
(1 + ς)
factor k(l) is maximized. Drawing analo-
gies with the behavior observed for ET where
in bulk liquids, Marcus suggested that
k(l) can go through a maximum at ς = 1 − exp(−2π) (24)
and nonadiabatic limit (i.e. κ = 4π) can be

|H |2 π 3/2 approximated to
= (25)
2πhν(kB T λ)1/2
where the parameter H corresponds to kobs ≈ 2πκν(aO1 + aR2 )β −3

electronic coupling matrix element
−Gact
× exp (28)
−β(l − aO1 − aR2 ) kB T
H = H0 exp (26)
2
which is identical in form to the previous
and λ is the local reorganization energy. Marcus derivation (Eq. 18) taking the
From classical transition state theory, the
hypervolume as in Eq. (21). A numerical
Landau–Zener approximation accounts
comparison between the limiting case
for the probability of ET upon multiple
given by Eq. (28) and the full integration of
passages through the transition state
Eq. (27) is displayed in Fig. 4 as a function
region of the energy surface. Equation (23)
of the parameter . It is observed that
is equal to 1 for adiabatic ET processes and
decreases toward the nonadiabatic limit. away from the nonadiabatic limit, Eq. (28)
On the basis of this spatial dependence provides bimolecular rate constants one or
of the transmission coefficient as well as two orders of magnitude smaller than for
the Gibbs activation energy (as discussed the ‘‘complete’’ model.
in the next section), it is clear that a A further refinement introduced by
more complete model will include the Smith and coworkers involves the reso-
integration not only of the hypervolume lution of the general integrand (27) taking
as implicitly done for Eqs. (21 and 22) but into account the distance dependence of
also of the local ET rate constant and even the reorganization energy. As discussed in
the concentration profile of the reactants the next section, the solvent reorganization
at the interface [36]. term is not only dependent on the distance
The meticulous analysis by Smith and separating the redox couples but also on
coworkers allows illustrating the effect of their relative positions with respect to the z-
each of the spatially dependent parameters axis. Fig. 5 shows the value of the integrand
on the integrand of Eq. (19) [36]. Including calculated from the nonadiabatic limit of
all terms in the integrand, it follows the Marcus expression, as well as from
∞ l−aO1 θmax (zo ,l) 2π the general model excluding concentra-
kobs = νκ tion profiles. In general, these simulations
l=aR2 +aO1 zo =aR2 θ=0 =0 show that the integrand is more sharply

−Gact (l) 2 dependent on the distance l in the gen-
× exp l gR1 (z)gO2 (z)
kB T eralized model. In order to illustrate the
× sin θ d dθ dz dl (27) contribution of the various physical ele-
ments involved in the ET dynamics, we
where g is the corresponding concentra- shall look closer at the reorganization en-
tion profile along the z-axis. Neglecting ergy as well as the interfacial distribution
distance dependence on the Gibbs en- of reactants (which is connected to the
ergy of activation and the concentration characteristic work terms for approaching
profiles, the solution of Eq. (27) in the reactants to the interface).
10−17
10−18
[cm4s−1]
2
ks
10−19 1
10−20
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
(a) Γo
10−15
3
10−16
2
[cm4s−1]
10−17
ks
10−18
10−19
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4

(b) Γo
Fig. 4 Calculated electron-transfer rate constants as a function
of the adiabatic parameter for aO1 = aR2 = 3 Å and
λ = 1 eV. The parameter β was evaluated as 0.25 and 1 Å−1 for
(a) and (b), respectively. Curve 1 was estimated from Eq. (28),
employing the approximation κ = κLZ (l = aO1 + aR2 ). Curve 2
is obtained from the full numerical integration of Eq. (27).
Curve 3 was calculated as in curve 2, but assuming a
concentration profile of the two reactants that increases toward
the interface. (Reprinted from Ref. [36] with permission from
Elsevier Science.)
4.2.2.2.3 The Solvent Reorganization En- (λ), the concentration independent term of
the Gibbs energy of ET (G◦et ), as well as

ergy The Gibbs energy of activation al-
ready included in Eq. (18) can be described work terms for approaching reactants (wp )
in terms of the total reorganization energy and separating products (ws ) to and from
1.0
z=3Å
all curves divided by 3Å peak height

Full integrand (C(z) assumed bulk)
0.8
z = 3.25 Å
0.6
z = 3.5 Å Integrand curves

0.4 for z = 3 to 4.75 Å
in 0.25 Å increments
Ho = 0.01 eV
0.2
0.0
0 1 2 3 4 5 6
(a) R −a1 −a2
1.2
z=3Å
Marcus integrand (C(z) assumed bulk)
all curves divided by 3Å peak height
1.0
z = 3.25 Å
0.8
z = 3.5 Å
Integrand curves
0.6 for z = 3 to 4.75 Å
in 0.25 Å increments
Ho = 0.01 eV
0.4
0.2
0.0
0 1 2 3 4 5 6
(b) R −a1 −a2
Fig. 5 Evaluation of the full integrand of Eq. (27) as a function of

the distance l employing Eq. (29) for Gact , Eq. (32) for λs , and
Eq. (26) for H (a). Other parameters were evaluated as
op op
εws = εos = 30, εw = εo = 2, H0 = 0.01 eV, β = 1 Å−1 , and
◦
Get = 0. The integrand corresponding to Eq. (19) under identical
condition is shown in (b). The dependence of the integrand on the
distance is sharper in (a) than in (b) owing to the inclusion of the
distance dependent of Gact and λs . (Reprinted from Ref. [36] with
the interface dielectric media and for the general geo-

2 metrical arrangement depicted in Fig. 3(a),
G◦et + ws − wp

λ provides an expression for λs of the form
Gact = 1+
4 λ
(ne)2 1 1 1
(29) λs = − s
where G◦et is given by

4πε0 εoop εo 2aR2

◦ ◦
(ne)2 1 1 1
Get = −nF (w w + op − s
o φ − o φet ) (30) 4πε0 εw εw 2aO1

The term in Eq. (30) corresponds to the (ne)2 1
+
so-called driving force for the ET process. 4πε0 4zR2
Considering that reactants and products
op op

involve ionic species, the work terms (wp ) εo −εw εos − εws
× op op op −
and (ws ) feature a chemical part associated εo (εo + εw ) εos (εos + εw s)
with their solvation energies and an elec-

(ne)2 1
trostatic term that is determined by the +
4πε0 4zO1
local potential distribution. This electro-

op op
static term is rather important as it may εw −εo εws − εs
o
account for part of the potential depen- × op op op − s s
εw (εw + εo ) εw (εw + εos )
dence of the phenomenological ET rate
constant. We shall give a closer look to this (ne)2 2 1 1
− op op − s
point in the next section. 4πε0 l εo + εw εo + εw s
The contribution of the solvent reorgani- (32)
zation energy to the total λ as considered by where εop and εs are the optical and static
Kharkats [8] and later by Marcus [11]. The dielectric constants, n is effective number
expressions obtained appeared not entirely of electrons transferred, and e is the
op op
consistent, but successive revision estab- elementary charge. Taking εw = εo =
lished that the key aspect lies in separating εop and εw s = ε s = ε s , Eq. (32) reduces to
o
the static and optical terms of the integral the well-known expression obtained for
of the electric displacement vectors over homogeneous ET,
the volume of the two liquids system [37]
  n2 1 1 1
λhm = + −
s
4πε0 2aO1 2aR2 l
1  1 − → − → 
λs =  (Df − Di )2 dv 1 1
2ε0 εst × − s (33)
∞−Vw −Vo static ε op ε
 
If we take 1,2-dichloroethane as the organic
1  1 − → − → 
−  (Df − Di )2 dv solvent, Eq. (32) can be simplified to
2ε0 εop
∞−Vw −Vo optical
w|DCE n2 0.189 0.275
(31) λs = +
−→ −
→ 4πε0 aR2 aO1
where Di and Df are the displacement
vectors for the initial and final states, 0.029 0.014 0.494
+ − −
respectively. Integration of Eq. (31), con- zR2 zO1 l
sidering a sharp boundary between two (34)
It can be noticed that the dominant factors possible locations of the reactants at the
in Eq. (34) are related to the ionic size and interface, including the case in which
the distance separating the redox species. one of them is allowed to cross from
Assuming 0.5 nm for the ionic radii of both one phase to the other [38]. This anal-
reactants in a face-to-face configuration ysis provides a quantitative comparison
w|DCE
and a total distance l of 1 nm, λs can between the homogeneous limit (Eq. 33)
be estimated to be of the order of 0.7 eV. and the corresponding heterogeneous case
This magnitude is considerably larger than (Eq. 32).
the common internal reorganization terms
for species in which no substantial changes 4.2.2.2.4 Potential Dependence of the Het-
are seen in the structures during the redox erogeneous Electron-transfer Rate Constant
process. Therefore, it is expected that the Some of the earlier works devoted to ratio-
solvent reorganization energy will provide nalize the dependence of the observed ET
the most important contribution to the rate constant on the applied potential were
total λ. based on basic models for charge transfer
Marcus also calculated the chemical part in polar media. Two of the main issues
of the work term for the formation of the arising from these considerations are the
precursor complex [11] structure of the liquid|liquid boundary and
the distribution of the electrostatic poten-
◦ 1 (zR2 )2 (zO1 )2
wp = − − tial. In the pioneering work by Samec in
4πε0 4do εos 4dw εws
s 1979, the potential dependence of kobs was
ε − εos 2 zO1 zR2 considered assuming a model involving
× w s + εs
+ s + εs
(35)
εw o l εw o two diffuse layers separated by an ion-free
compact layer [7]. The region delimited by
It should be noticed that this expression is
the inner layer establishes the distance for
obtained for a sharp boundary between
maximum probability of ET (R ∗ ). In this
the dielectric; therefore it may not be
case, the element δR ∗ is estimated by the
directly applicable in the mixed-solvent re-
space volume of the ion of radius aO1
gion model. The equivalent expression for
located in one of the reaction plane and
the successor complex (ws◦ ) can be easily
certain molecular volume adjacent to the
obtained by replacing the charges of the
phase boundary,
reactants by the corresponding products.
For the sake of comparison, if we intro- δR ∗ = 2aO1 AVm (36)
duce the parameters employed earlier for
w|DCE
estimating λs , it can be obtained that The density probability of the precursor
they are of the order of 10 and 80 meV, complex, (R ∗ ), can be obtained by a
respectively. For this calculation we have distribution of the type
also assumed that zO1 = zO2 = +1 and
zR1 = zR2 = 0. It can be clearly seen that (R ∗ ) = aO1 aR2

wp◦ and ws◦ are expected to be rather small −zO1 zwO1 φ + zR2 zoR2 φ
w|DCE × exp
in comparison to λs , therefore the kT
activation energy (see Eq. 29) is mostly (37)
controlled by the latter. It should also be where ‘‘zO1 ’’ and ‘‘zR2 ’’ correspond to the
mentioned that Benjamin and Kharkats position of the edges of the inner layer.
have developed expressions for λs for all From Eqs. (36 and 37), and considering
that the ET probability at the precursor potential independent parameters,

complex is determined by Gact (Eq. 29), ◦
the rate constant of ET can be evaluated as wp = wp + zO1 FzwO1 φ + zR2 FzoR2 φ
(41)
◦
ws = ws + zO2 FzoO2 φ + zR1 FzwR1 φ
pzc −zO1 zwO1 φ − zR2 zoR2 φ
kobs = kobs exp (42)
RT
zO1
The terms wp◦ and ws◦ are associated with
−αnzR2 φ the changes in solvation energy and ionic
× exp (38)
RT atmosphere experienced by the reactants
and products in the mixed-solvent region.
where The second and third terms in Eqs (41
and 42) reflect the work involved to

pzc ∗ −λ displace the ions in and out of the
kobs = 2aO1 Vm B exp
4RT electrified interfaces. Effectively, zwO1 φ

and zoR2 φ are the potential drop developed
◦
−αnFo φet
w
between the reaction plane and the bulk
× exp (39)
RT of the corresponding electrolyte phase.
1 nF zO1 In the original analysis by Samec [7],
◦
α= + (zR2 φ − w
o φet ) (40) these electrostatic components provide
2 4λ
the most important contributions to the
work terms. This is simply because of
and B ∗ involved the overlap integral of
the assumption that chemical changes in
the wave function associated with the
the ionic environment can be neglected
initial and final states. Although the
outside the ‘‘inner compact layer.’’
preexponential term is not fully developed
The approach followed by Girault and
as discussed in Sect. 4.2.2.2.2, the whole
Schiffrin [9] was based on the ‘‘encounter
quantity can be experimentally accessed by preequilibrium’’ model, previously em-
measurements of the phenomenological ployed for ET in condensed phase [39] and
rate constant at the potential of zero metal|electrolyte interfaces [40]. It follows
charge (pzc). On the other hand, a very from this model,
important implication from Eq. (38) is that
the dependence of the kobs on the Galvani kobs = Kpe ket (43)
potential difference may only arise from
changes in the interfacial concentration of where Kpe is the equilibrium constant for
the reactants and not by the Gibbs energy forming the precursor complex, which is
of activation. given by
Girault and Schiffrin proposed a model
e ∗ −wp
in which the interface features a mixed- Kp = δR exp (44)
RT
solvent region [9]. To approach this re-
gion, the reactants and products undergo Similar to the previous consideration, δR∗
changes in their solvation energies and corresponds to the ‘‘reaction zone’’ where
ionic atmosphere. Consequently, the work the molecules feature the appropriate
terms associated with the formation of configuration for the ET step. Considering
the precursor complex and dissociation of Eqs. (20, 29, 30, and 41 to 44), an
the successor complex also feature some expression for kobs identical in form
to Eq. (38) can be obtained but with a far more complex than what is described
preexponential factor of the form by current modeling.

−λ −wp◦
kobs = δR ∗ νeff exp
pzc 4.2.2.3 Selected Experimental Results
exp
4RT RT The first quantitative evaluations of hetero-
◦ geneous electron transfer rate constants
α(wp◦ − ws◦ + nFw

o φet )
× exp were performed employing cyclic voltam-
RT metry and electrochemical impedance
(45) spectroscopy at polarizable interfaces [13,
here νeff is the effective electron hopping 14, 17, 21]. Careful choice of redox couples
frequency. It should also be noticed that if and electrolyte composition was essential
the potential independent components of in order to avoid interference from cou-
the work terms are negligible with respect pled ion-transfer processes. From these
to the reorganization energy and the ther- early studies, rate constants typically of the
modynamic driving force, Eq. (45) reduces order of 10−22 cm4 s−1 were measured for
to an expression similar to Eq. (39). Appar- the ET between hexacyanoferrate and var-
ently, this might be physically plausible if ious redox couples in the organic phase. A
, wp◦ , and
w|DCE common aspect to all these measurements
we compare the values of λs
◦ is the redox potential matching, which al-
ws as derived from Eqs. (34 and 35).
As we mentioned previously, the po- lows to onset the ET reaction within the
tential dependence of kobs can arise from polarizable window commonly obtained
changes in the Gibbs energy of ET or for liquid|liquid junctions. Obviously, this
from concentration polarization effects. boundary condition does not permit study-
ing redox processes featuring large values
Schmickler approached this problem by
of G◦et . As discussed later, dynamic stud-

simulating the potential dependence of

ies far from equilibrium provide very
the distribution of ionic species using a
useful information on fundamental pa-
lattice-gas model [41]. These simulations
rameters such as the preexponential factor
showed that as far as the distance separat-
in Eq. (28) and the reorganization energy
ing the redox species is smaller than the
(Eq. 32).
Debye length of each space charge region,
The particular case of hexacyanoferrate
the potential dependence would be mostly and TCNQ (see Fig. 2) have been stud-
controlled by the concentration polariza- ied not only by electrochemical impedance
tion phenomena. This conclusion appears spectroscopy [14] but also by Poten-
more consistent with the ‘‘mixed-solvent tial Modulated Reflectance Spectroscopy
region’’ model, which allows a close ap- (PMR) [26], SECM [42], and Microelec-
proach between the two reactants. Some trochemical Measurements at Expanding
experimental reports have shown devia- Droplets (MEMED) [27, 28]. Considering
tions from the Butler–Volmer behavior, that the composition of the electrolyte
which could reflect concentration polariza- varied for all these approaches, rate con-
tion phenomena. However, as reviewed in stants between 10−23 and 10−22 cm4 s−1
the next section, a clear dependence of the were consistently obtained for this sys-
ET rate constant and the Galvani potential tem. Cheng and Schiffrin demonstrated
difference have been observed, suggesting that these figures are consistent with the
that the potential and ionic distribution are basic expressions derived by Marcus for a
3.0 Potential dependence of the

Fig. 6
electron-transfer rate constant (kobs )
pzc
2.5 normalized to the value at the pzc (kobs )
for the TCNQ reduction by
ln (kobs/kobs)
pzc
2.0 hexacyanoferrate at the water|DCE

interface. Full line was estimated from
1.5 Experimental Eq. (38), assuming that the potential
drop across the inner layer is negligible
1.0
(zzO1
R2
φ ≈ 0 V). Dependence of the
0.5 potential drop across the ET reaction
Model plane with the Galvani potential
0.0 difference, as obtained from
(a) −0.20 −0.15 −0.10 −0.05 0.00 recalculation of the two trends in (a),
7 assuming a transfer coefficient of
0.5 (b). These results suggest that
6 approximately 30% of the Galvani
potential difference is developed
5
between the reacting species under the
O1
102 ∆zR2 f
4 experimental conditions defined in

z
[V]
Fig. 2. (Reprinted from Ref. [26] with

3 permission from Elsevier Science.)
2
1
0
−0.20 −0.15 −0.10 −0.05 0.00
∆wof
(b) [V]
sharp liquid|liquid boundary Eqs. (28 and was considered. The interfacial potential
32) [14]. was estimated using a Poisson–Boltzman
The potential dependence of the ob- distribution of the ionic species. The
served ET rate constant from hexacyanofer- substantial underestimation of the experi-
rate (II) to TCNQ is shown in Fig. 6 [26]. mental points can be taken as evidence that
The rate constants were estimated from a finite fraction of the applied potential is
chronoabsorptometric and PMR measure- developed between the reaction planes zO1
ments under the conditions described in and zR2 . Assuming a transfer coefficient
Fig. 2. Deviations from the Butler–Volmer α = 0.5, Fig. 6(b) shows that the local po-
behavior are evident at potentials away tential could be up to 30% of the total
from the pzc, which were also pointed Galvani potential difference.
out by Cheng and Schiffrin in a previous The dependence of the observed ET rate
paper [14]. The theoretical curve shown in constant on the thermodynamic driving
Fig. 6 illustrates a simulation of the po- force indeed remains rather controversial.
tential dependence of the observed rate Unlike electrochemical and spectroelectro-
constant normalized to the value at the chemical approaches at polarizable inter-
pzc employing Eq. (38). The potential drop faces, techniques based on SECM and
between the plane at zO1 and zR2 was ini- dynamic photoelectrochemical measure-
tially neglected and only the contribution ments have allowed probing rate constants
associated with concentration polarization far away from equilibrium conditions. In
SECM, redox-active species are generated separating the redox couple, therefore in-
in the proximity of the interface by an ap- creasing the solvent reorganization energy
proaching microelectrode. The feedback (see Eq. 32). In this case, a decrease in
current at the microelectrode tip contain kobs with increasing potentials at high val-
information on the rate of ET at the ues of driving force has been observed.
liquid|liquid boundary. On the other hand, This phenomenon was taken as first evi-
the photoinduced ET involving dyes and dence of the Inverted Marcus Region at a
redox species at either side of the inter- polarized interface. However, the limited
face manifest itself as photocurrent re- number of experimental points does not
sponses [5, 43–47]. These photoresponses allow quantifying parameters such as the
are proportional to the ratio between the reorganization energy term.
rate constants of excited state decay and Barker and coworkers employed essen-
heterogeneous ET. The common point tially the same approach but reducing the
between these two approaches is that concentration of the redox couple in the
highly reactive species are generated at aqueous phase. Although the analysis of
the interface, either by light or by an ap- the approaching curves is further compli-
proaching microelectrode; therefore large cated by the coupling of diffusion regimes
driving forces for the ET can be attained. in either side of the interface, the most
Tsionsky and coworkers studied the important aspect is that the overall rate
dynamics of ET from various redox cou- of ET decreases. Consequently, rate con-
ples located in the aqueous phase to the stants of fast reactions, which previously
oxidized zinc tetraphenylporphyrin in ben- appeared diffusion-controlled, were accu-
zene by SECM. The driving force was rately measured. This analysis provided a
varied by changing the concentration ratio dependence of the rate constant on the
of a common ion (ClO4 − ) and by using driving force as illustrated in Fig. 7(b). The
different redox couples in the aqueous authors rationalized the limited number of
phase. It was observed that for driving experimental points employing Eqs. (18,
forces smaller than 0.5 eV, the potential 29, and 30), providing values for the reor-
dependence of the observed rate constant ganization energy and the preexponential
exhibits a Butler–Volmer behavior. Ignor- factor of 0.55 eV and 8.3 × 10−20 cm4 s−1 ,
ing concentration polarization effects, this respectively. In principle, the value for λs is
dependence is consistent with Eq. (38). For consistent with Eq. (32), assuming a face-
larger driving forces, the observed rate to-face approach and for l ∼ aO1 + aR2 .
constant becomes potential independent The previous works based on SECM have
suggesting a diffusion control regime. It not addressed the specific contributions of
should be mentioned that the redox cou- the Galvani potential difference and the
ples in the aqueous phase are in large formal ET potential to the thermodynamic
excess with respect to the zinc porphyrin driving force. Our recent work based on
in the organic phase in order to avoid cou- photoinduced ET at externally polariz-
pling of diffusion profiles in both phases able interfaces quantitatively tackles this
(the constant composition approximation). problem employing the thermodynamic
Interestingly, a different behavior is ob- relations discussed in Sect. 4.2.2.1 [48].
served in the presence of a C-10 surfactant The advantage of this approach is that
as depicted in Fig. 7(a). The adsorbed the driving force can be changed potentio-
lipid layer effectively increases the distance statically as well as by changing the redox
−1.5 Fig. 7 ‘‘Driving force’’ dependence of

3 kobs obtained for the reduction of
1 ZnPor+ in benzene by various redox
couples in the aqueous phase as probed
−2.0
by SECM in the presence of a full
[cm s−1]
logk f
2 monolayer of C10-lipid (a). As the

driving force increases, kobs increases in
−2.5 the presence of Fe(CN)6 4− (curve 1),
but decreases for Co (II) sepalchrate
(curve 2) and V2+ (curve 3). A similar
−3.0 analysis is presented in (b) but in the
1500 1000 500 absence of the C-10 lipids and for
o
∆E + ∆w j
o substantially smaller concentration of
(a) [mV] the aqueous redox couple. The curve
2.5
in (c) was obtained from photocurrent
measurements at the polarizable
2.0 Fe(CN)63-/4- water|DCE interface in the presence of
1.5 water-soluble porphyrin dimer
[cm s−1 M−1]
(ZnTPPS : ZnTMPyP) and ferrocene

log (k12)
1.0
(Fc), dimethylferrocene (DMFc),
FeEDTA-/2-
0.5 butylferrocene (ButylFc),
Mo(CN)63-/4-
0.0 diferrocenylethane (DfcEt), and
decamethylferrocene (DCMFc). (Figs. a,
−0.5 Ru(CN)63-/4- b and c were reprinted from Refs. [32,
−1.0 34], respectively, with permission from
0 100 200 300 400 500 600 700 800 900 the American Chemical Society.)
∆E1/2
(b) [mV]
10.0
1.0
[M−1 s−1 cm]
ket
0.1 : Fc
: DMFc
: ButylFc
: DFCEt
0.01 : DCMFc
0.001
0.0 0.5 1.0 1.5 2.0
−∆Geto'
(c) [eV]
couple acting as quencher in the organic energy, which shift the inverted region be-
phase. From the photocurrent analysis, the yond 1 eV. The number of experimental
phenomenological rate constant of ET as a points obtained from a family of ferrocene
function of the driving force is illustrated derivatives allows a confident analysis in
in Fig. 7(c). It can be observed that the terms of Eq. (29). This rather high reorga-
ET from the redox couple in the organic nization energy can also be explained in
phase to the excited state of the porphyrin terms of Eq. (34), assuming aFc ∼ 0.2 nm,
dimer exhibits a rather high reorganization apor ∼ 0.7 nm, and l ∼ 1 nm.
Finally, it is rather interesting to no- activated process and as a pure transport

tice that Figs. 7(b and c) provide pre- process. Usually, the ion transport across a
exponential factors of the order of 5 × liquid|liquid interface has been described
10−20 cm4 s−1 . Comparing with the sim- using the Nernst–Planck equation [60, 61]
ulations in Fig. 4, it can be concluded that or the Langevin equation [57–59, 65]. The
the experimental data is consistent with basic ideas of these independent theories
a highly nonadiabatic limit ( → 0), in will be reviewed starting from first princi-
which the expressions derived by Marcus ples and a more generalized approach will
and Smith and coworkers do converge. be outlined. In a recent theoretical work the
This low electronic coupling between the transport description has been combined
redox species may arise from the differ- with a microscopic theory on the mecha-
ence in the solvation properties, which nism of ion transfer [68]. Other models for
induces interfacial separation distances of ion transfer will be dealt with briefly in a
the order of 1 nm. separate paragraph [55, 64–66, 79].
4.2.3 4.2.3.1 Kinetics of Ion-transfer Reactions

Ion-transfer Reactions Based on the Assumption of an Activated
Ion-transfer Event
Ion-transfer reaction at interfaces between The simplest kinetic approach is to con-
immiscible electrolyte solutions (ITIES) sider an ion-transfer reaction as an acti-
has been an area of active research for vated process as schematically shown in
more than two decades [49–52]. Several Fig. 8, but we shall consider the transfer
research groups have developed theories of the ion across the complete interface
dealing with the fundamental features at which the Galvani potential differ-
of ion-transfer processes at (ITIES) [7, ence between the two phases varies. This
53–68]. The theoretical work has often derivation may be termed the global ap-
been initiated by advances in experimental proach.
techniques used to obtain kinetic infor- For this simple kinetic model, the
mation on the ion-transfer kinetics [53, Gibbs energy in the adjacent phases is
69–74]. More recently, computer simu- defined by the standard electrochemical
lations have given important information potential, which can be defined as the
on the microscopic surface structure [51, concentration independent part of the
57, 75–78]. The microscopic nature of electrochemical potential.
the ion-transfer event has been discussed The current is defined by
in the theoretical works on ion trans- −→ ←− ◦
fer, and different factors that are poten- I = zFA k cw − k c (46)
tially important for the ion-transfer event
−→ ←−
at ITIES have been suggested: elastic- where k and k are the forward and
ity of the inner layer, surface structure backward global rate constants with units
in general, the electric potential energy of cm s−1 . The bulk concentrations are
(dependent on the Galvani potential), given by cw and c◦ . The charge of the
hydrodynamic friction, and changes in ion is z, A is the surface area, and F is
solvation. the Faraday constant. At equilibrium, the
There are two main approaches to de- Galvani potential difference is equal to the
scribe ion-transfer reactions, that is, as an standard transfer potential defined by
Interface
energy
Gibbs
m0, w + zFfw
RT ln(co/cw) m0, o + zFfo
Reaction coordinate
Fig. 8 Standard Gibbs energy profile for global ion-transfer
reactions.
o,w→ ◦
◦ Gtr,i the entire interfacial region composed of
w
o φi = (47) two back-to-back diffuse layers. The only
zi F
merit of this global approach is to yield
◦
where Go,w→
tr,i is the standard Gibbs an expression that in essence can be
energy of transfer from water to the considered as a Butler–Volmer equation
organic phase. At the standard transfer for an ion-transfer reaction. Many groups
potential the activation energy barrier is have observed this Butler–Volmer behav-
regarded as symmetrical. In the standard ior, as originally suggested by Gavach [53].
case at equilibrium we have cw = c◦ = c Figure 9 illustrates the results obtained by
w ◦
and w o φeq = o φ . When we apply a means of chronocoulometry for the trans-
Galvani potential difference different from fer of acetylcholine.
the standard transfer value, only a fraction,
α, of the driving force is active on lowering 4.2.3.2 Interfacial Reaction Independent of
the energy barrier: the Local Electric Field
w ◦
Another way to treat an ion-transfer
◦ w
I = zFAk c ezF α(o φ−o φ ) reaction is to consider the reaction as
w ◦
an elementary ‘‘jump’’ across a thin
− e−zF (1−α)(o φ−o φ )
w
(48) interfacial layer, which we will call the
reaction plane, but it may also be called
In the general case, this equation becomes ‘‘inner layer’’ or ‘‘mixed-solvent layer’’
w ◦ depending on the interfacial model used.
◦ w
I = zFAk cw ezF α(o φ−o φ ) The interface (about 1 nm thick) comprises
w ◦
back-to-back diffuse layers.
◦
− c e−zF (1−α)(o φ−o φ ) (49)
w
With the notation of Fig. 10, the current
now reads
If this simple kinetic approach can be −→ ← −
used as a primary model, it is difficult I = zFA k ca − k cb (50)
to ascribe a physical meaning to the rate
−
→ ←−
constants. Indeed, these rate constants where k and k represent the rate
correspond to the transfer of an ion across constants of the ion-transfer reaction
−0.5
a = 0.48
−1.0
[cm s−1]
a = 0.53
log k
−1.5
−150 −100 −50 0 50

(∆wof − ∆wof½)
[mV]
Fig. 9 Tafel plots for ion transfer of acetylcholine. From water to DCE (◦) and from DCE to water (•).
Reaction plane
energy
Gibbs
m0,w + zFfa
a
RT ln(c b/c a)
m0,o + zFfb
b
Reaction coordinate
Fig. 10 Standard Gibbs energy profile for an elementary
ion-transfer reaction.
Preequilibrium Ion at Transfer Ion at Preequilibrium

Ion Ion in
interface interface
in water organic phase
in position a in position b
Scheme 1 Elementary steps for kinetically controlled ion transfer.

across the reaction plane and where If we further assume that α does not
the concentrations ca and cb are the vary with w o φ (which is a very rough
concentrations on both sides of the approximation), the following expression
reaction plane. In a first approximation, for the current ensues:
we can assume that the rate constants are w ◦
◦ w
potential-independent and that the only I = zFAk cw eαzF (o φ−o φ )/RT
consequence of the interfacial polarization w ◦

◦
− c e−(1−α)zF (o φ−o φ )/RT (56)
w
is to vary the interfacial ion concentration
as described by Boltzmann statistics:
This result is similar to what was obtained
ca = cw e−zFw φ/RT
a
(51) in the simple approach (Eq. 49). The
◦ fundamental difference is that now it is
cb = c e−zFo φ/RT
b
(52)
possible to relate k 0 to the microscopic
In the application of these two equations it features of the ion transfer.
is assumed that there is a preequilibrium
between the reaction plane and the 4.2.3.3 Interfacial Reaction Dependent on
bulk regions. the Electric Field
It is important to realize that the applica- Still considering an ion-transfer reaction
tion of the above scheme and Eqs. (51 and as an interfacial process, Eq. (50) remains.
52) are based on the assumption that the However, we will now assume that the rate
ion transfer is slow compared to diffusion constants depend on the applied Galvani
of ions to a and b from bulk solutions. By potential difference according to
substitution in Eq. (50), we obtain
−
→ − → a
− k = k0 eβzFb φ/RT (57)
→
I = zFA k cw e−zFw φ/RT
a
←− ← −
k = k0 e−(1−β)zFb φ/RT
a
(58)
←− ◦
− k c e−zFo φ/RT
b
(53)
In these equations it is assumed that
Furthermore, if the positions a and b are transition state theory can be applied
close, we can assume that the electrical in the description of the ion transfer.
potential difference between these two By substitution in Eq. (50), the following
points is negligible. In this case, Eq. (53) equation is obtained:
reduces to −
→ a
− I = zFA k0 ca eβzFb φ/RT
→ w
I = zFA k cw eαzFo φ/RT
←−
− k0 cb e−(1−β)zFb φ/RT (59)
a
←− ◦
− k c e−(1−α)zFo φ/RT (54)
w
If we still assume that there are equilibri-

The proportion of the Galvani potential ums between the interfacial position and
difference occurring in the aqueous phase the bulk, we have
is termed α. At equilibrium, at the standard −→
I = zFA k0 cw e−βzFa φ/RT e−zFw φ/RT
b a
transfer potential, we have
−
→ αzFwo φ ◦ ←− w ◦
← − ◦
ke RT = k e−(1−α)zFo φ /RT − k0 c e(1−β)zFa φ/RT e−zFo φ/RT
b b
◦
=k (55) (60)
We can define the potential φ # such that the random force caused by fluctuations in
the condensed medium:
#a φ = βba φ (61)
dv 1 dU (x, t)
b# φ = (1 − β)ba φ (62) = −βv − + A(x, t)
dt m dx
Equation (60) then reads
The derivative of U with respect to po-
−
→ sition is the regular force acting on the
I = zFA k0 cw e−zFw φ/RT
#
particle. However, this so-called stochastic

←− ◦ ◦
approach results in similar Nernst–Planck
− k0 c e−zF# φ/RT (63) type equations albeit with microscopic
diffusion coefficient differing from the tra-
The coefficient α can be defined as ditional macroscopic diffusion coefficient.
◦ The microscopic diffusion coefficient, D,
#w φ = αw φ (64)
is related to β, through D = (kB T )/(βm).)
◦ ◦
# φ = (1 − α)w φ (65)
4.2.3.4 Kinetics of Ion-transfer Reactions
Equation (63) can now be written as Based on a Transport Description
−
→ w
Another way to treat ion-transfer reactions
I = zFA k0 cw eαzFo φ/RT is to consider them as a special case of ionic
← − ◦ conductivity across an interface. From a
− k0 c e−(1−α)zFo φ/RT (66)
w
phenomenological viewpoint, anionic flux
can be written [80]:
At the formal standard transfer potential,
we have Ji = −ci ũi gradµ̃i (68)
−
→ αzFwo φ ◦ /RT ←− w ◦
k0 e = k0 e−(1−α)zFo φ /RT
By taking into account that the electro-
◦
=k (67) chemical potential can be considered as
the sum of a standard term, which is sol-
By defining k ◦ in this way, we recover vent dependent, a concentration/activity
an expression of the current similar to term, and an electrical term, we have
Eq. (49). This approach may be compared
◦
with the of work Koryta [54] and Samec [7]. Ji (x) = −ci ũi (x)[gradµi (x)
In the latter, an activation energy and a
solvent reorganization energy are explic- + RT grad ln ci (x)
itly associated with the rate constant of + zi F gradφ(x)] (69)
ion transfer. (The microscopic transport
of ions across the ITIES may also be which is identified as the Nernst–Planck
described using the Langevin equation, equation:
see Refs. [58, 59], which relates the mi-
croscopic rate of particle motion, v, of a Ji (x) = −RT ũi (x)gradci (x) − zi F ci (x)
particle of mass m with coordinate x and
◦
time t, to the regular potential of the field × ũi (x)grad[φ(x) − φi (x)] (70)
U (x, t), a phenomenological friction coef-
ficient, β, and a function A(x, t) describing with φi◦ (x) = −µ◦i /zi F .
To integrate this equation, we can We then obtain an equation for the

rewrite it in the following way: ion flux:

Ji (x) = −RT ũi (x) ũi RT cia eya − cib eyb
Ji = (ya − yb )
◦
L eya − eyb
× grad ci ezi F [φ(x)−φi (x)]/RT (78)
◦ The parameter y is defined as
× e−zi F [φ(x)−φi (x)]/RT (71)
ya − yb zi F ◦
y= = (a φ − ab φi )
If we consider the ion-transfer reaction as 2 2RT b
the crossing of the inner layer, we have to (79)
integrate this equation between a in the The ion flux can now be expressed as

water phase and b in the oil phase. 2ũi RT y cia e2y − cib
◦ Ji =
a ezi F [φ(x)−φi (x)]/RT L e2y − 1
Ji dx = −RT
b ũi (x) ũi RT y
a = (ca ey − cib e−y ) (80)
◦ L sinh y i
× d ci ezi F [φ(x)−φi (x)]/RT (72)
b The approach outlined above is essentially
If φ(x) − φi◦ (x) varies linearly through the equivalent to that suggested by Kaki-
interface, by defining uchi [60], which again relies on the work
of Goldman for ion transfer across mem-
zi F ◦ branes. Expressions for the ion-transfer
yb = [φ(b) − φi (b)] (73)
RT rate constants may be obtained as below
zi F ◦
if we recall that the ionic flux across the
ya = [φ(a) − φi (a)] (74) interface can be described according to
RT
Eq. (50).
we can write
−→ ũi RT yey
zi F [φ(x) − φi◦ (x)] x k =
sinh y
(81)
= yb + (ya − yb ) L
RT L −y
(75) ←− ũi RT ye
k = (82)
If it is assumed that the ionic mobility is L sinh y
constant through the interface, we have
Since it may be noted that
L
Ji ye−y
eyb +(ya −yb )x/L dx = −RT lim
yey
= lim =1 (83)
ũi 0 y→0 sinh y y→0 sinh y
a
◦
× d ci ezi F [φ(x)−φi (x)]/RT (76) It immediately follows that a useful
b definition for ki◦ is
Solving this integral leads to
◦ ũi RT
L ki = (84)
Ji LJi L
eyb e(ya −yb )x/L dx =
ũi RT 0 ũ i RT −
→ ←−
ya The equations describing k and k can
e −e yb
easily be incorporated into Eq. (50) to
× = −RT [cia eya − cib eyb ]
(ya − yb ) describe the entire current potential char-
(77) acteristic. Note that in this treatment the
In general, 1/2ab φi◦ can be approximated

ion transfer is treated as a transport pro-
cess and ũi is a phenomenological mobility ◦
by 1/2wo φi and we can therefore write
of the ion in the inner layer. This model
has successfully accounted for experimen- ũi RT y
tal data [71] and the experimental values Ji =
L sinh y
of α and k ◦ can be given phenomeno- w ◦
× ciw e−zi F (w φ+1/2o φi )/RT
#
logical meanings. Within the framework
of the model suggested by Kakiuchi, the
w ◦

◦
− ci e−zi F (o φ−1/2o φi )/RT
#
ion-transfer rate constant can be related to (88)
the microscopic structural features of the
liquid|liquid reaction plane describing the Finally, by defining α as the fraction of the
ionic mobility and interface width. total potential drop between the two bulk
By analogy to the derivation of Eqs. (56 phases occurring from bulk water to the
and 66), we can obtain an expression for Ji middle point of the potential drop between
on the basis of the bulk concentration of a and b, we arrive at the following result:
the ion in water and oil. The concentration
of the ion at a and b may be expressed ũi RT y
according to Eqs. (51 and 52) if we again Ji =
L sinh y
assume the formation of a preequilibrium w w ◦
according to Sch. 1, or in other words × ciw ezi F (αo φ−1/2o φi )/RT
that the ion transfer/transport across the
w ◦ ]/RT

◦
− ci e−zi F [(1−α)o φ−1/2o φi
w
liquid|liquid interface is slow compared to
the diffusion/migration of the ion in the (89)
diffuse layers. Under such conditions the It should be noted that in order to apply
following expression for Ji ensues: this refined model in general it is necessary
ũi RT y w −zi Faw φ/RT to know the potential drop across the inner
Ji = c e layer. We can recover Eq. (56) if we make
L sinh y i
a a ◦ the assumption that α is independent
× ezi F (b φ−b φi )/2RT ◦
of w o φ and define k according to the
◦
− ci e−zi Fo φ/RT following equation valid at equilibrium at
b
◦
the formal standard transfer potential:
× e−zi F (b φ−b φi )/2RT
a a
(85)
ũi RT y w ◦ w ◦
ezi F (αo φi −1/2o φi )/RT =
We can define a potential φ # such that L sinh y
ũi RT y
a# φ = #b φ = 12 ab φ (86)
L sinh y
leading to the following expression: w ◦ −1/2w φ ◦ )/RT ◦
× e−zi F ((1−α)o φi o i =k
ũi RT y (90)
Ji =
L sinh y Note, however, that this approach leads to
◦
a ◦ a different dependence of k ◦ on w o φi .
× ciw e−zi F (w φ+1/2b φi )/RT
#
An interesting point concerns the fact that
a ◦
when α is close to 0.5, the dependence of
◦
− ci e−zi F (o φ−1/2b φi )/RT
#
◦
(87) k ◦ on w
o φi vanishes.
4.2.3.5 General Treatment of Ion-transfer where Aw and A◦ are given by

Reactions at ITIES
a
In recent theoretical work on ion transfer x
at ITIES, the ion transfer through the dif- Aw = ezi Fw φ/RT dx (95)
−∞
fuse layer has been analyzed [62, 63, 67]. ∞
◦ x
In the following we will assume that the A =− ezi Fo φ/RT dx
ion transport from bulk solution to the b
interface can be described according to b x
the Nernst–Planck equation. No specific = ezi Fo φ/RT dx (96)
−∞
assumptions regarding the ion distribu-
tion or the potential profile will be made. The concentrations of the ion at a and b
The ion flux across the diffuse layers can are then obtained as
then be obtained by solving the follow-
ing equations:
Ji Aw
cia = e−zi Fw φ −
a w
a + c (97)
RT ũwi
i
Ji ezi F φ(x)/RT dx = −RT
−∞ Ji A◦ ◦
cib = e−zi Fo φ −
b
+ c (98)
a RT ũ◦i i
× ũw
i d(ezi F φ(x)/RT ) (91)
−∞
∞ In the case of small Aw and A◦ , the tradi-
Ji e zi F φ(x)/RT
dx = −RT tional Boltzmann distribution is recovered.
b
This situation corresponds to the case in
∞
◦ which the potential drops between a and
× ũi d(ezi F φ(x)/RT ) (92)
b ‘‘w’’ and between b and ‘‘o’’ are small.
These expressions can be inserted into
Note that in these equations the integration
limits ∞ and −∞ represent the bulk Eq. (50) to obtain the following result:
phases of the oil and water phase,
−→ w −zi Faw φ/RT
respectively, which is identified as the k ci e
position from the interface in which the ←−
− k ci◦ e−zi Fo φ/RT
b
Galvani potentials equal the bulk phase Ji = (99)
values. Upon integration, the following −→ Aw −zi Faw φ/RT
1+ k e
expressions are obtained: RT ũw i
◦
a z F φ a /RT ←− A
e−zi Fo φ/RT
b
ci e i
w
− ciw ezi F φ /RT −k
Ji = −RT ũw RT ũ◦i
i a
ezi F φ(x)/RT dx
−∞ Note that a potential dependence may
−RT ũw a
easily be incorporated into an expression
= w
i
cia ezi Fw φ/RT − ciw (93) −→ ←−
A for k and k as in Eqs. (57 and

◦ 58). A Butler–Volmer type expression
ci◦ ezi F φ /RT − cib ezi F φ /RT
b
◦
Ji = −RT ũi ∞ is obtained (Eq. 53) when Aw and A◦
ezi F φ(x)/RT dx can be approximated to zero in Eqs. (97
b and 98). Alternatively, Eqs. (97 and 98)
RT ũ◦i can be incorporated into Eq. (80) if the
◦ b
= ◦ ci − cib ezi Fo φ/RT (94) ion transfer is described as a pure
A
transport process: potential distribution when realistic pa-

rameters of the ion and diffuse layers
(ciw ey e−zi Fw φ/RT
a
are used [62]. From these considerations
ũi y it may thus be concluded that traditional
−ci◦ e−y e−zi Fo φ/RT )
b
steady state current voltage curves gener-
L sinh y
Ji = w ally do not give much information about
A
ey e−zi Fw φ/RT
a
1+ the inner layer processes.
RT ũw In general, experimental studies on ion-
i
A◦ −y −z Fb φ/RT ũi y transfer reactions lead to values of ‘‘appar-
− e e i o
RT ũ◦i L sinh y ent’’ rate constants. In this connection it is
(100) interesting to note that the nominator in
As in Eq. (80), we have here specified a Eq. (99) is essentially similar to the expres-
linear potential profile across the interface. sion for Ji in Eq. (66) obtained from the
−
→ ←− simple Butler–Volmer approach. It thus
If we define k and k as in Eq. (81
and 82), we recover Eq. (99). The potential immediately follows that the correction
dependence on the rate constant is, of an experimentally determined apparent
however, somewhat different from what rate constant based on Eq. (66) (equivalent
is obtained from expressions based on to the Frumkin correction used for ET at
solid electrodes) is not in direct agreement
transition state theory (Eqs. 57 and 58).
with the more general treatment leading to
In a practical situation it can be assumed
Eq. (99). This point was first recognized by
that a and b are approximately equal. Then
Senda [63], who termed the denominator
the potential drop in the aqueous phase is
of Eq. (99) the Levich correction.
simply obtained as w a φ ≈ αo φ.
w
Regardless of whether the transfer or

4.2.3.6 The Marcus Model for Ion Transfer
transport approach is followed, the flux
at Liquid|Liquid Interfaces
equation can, in principle, be solved for any
Recently, Marcus has attempted to treat
given potential profile in the diffuse layers.
the ion transfer as a combined activation-
Usually, the potential profile is quite well
controlled and transport-controlled pro-
described using the Poisson–Boltzmann cess [68]. In the Marcus theory for ion
equation, which in the case of a z : z transfer between two immiscible solu-
supporting electrolyte can be solved ana- tions, the ion transport is described using
lytically. More advanced theories may also a Nernst–Planck type equation [57]. This
be applied [81, 82]. In a recent theoretical equation leads to an expression of k ◦ sim-
study based on the idea of determining ilar to that obtained by Kakiuchi [60],
the permeabilities of the diffuse double
layers and the interface independently, ◦ ũRT D
k = = (101)
different potential profiles were analyzed. L L
The Poisson–Boltzmann potential profile, The experimentally determined k ◦ leads to
a stepwise linear potential profile, and a a phenomenological diffusion coefficient
potential step profile were compared. In- that is in the order of 10−8 cm2 s−1 ,
terestingly, even a very simple stepwise assuming an interface thickness of 10 Å.
potential profile is hardly distinguishable A central point in the Marcus theory
from the linear potential profile and the rel- is the experimentally and theoretically
atively complicated Poisson–Boltzmann verified presence of protrusions at the
liquid|liquid interface. It is assumed that The electrochemical potential of the ion

the ion transfer occurs via attachment of at zb and za is given by µ(zb ) and µ(za ),
the ion to a protrusion in one phase. respectively. Within the framework of a
The probability of forming a protrusion transition state model, the (x, y) oscillation
of height h is given by P (h). The of the ion transverse to the protrusion
probability of forming a protrusion is is considered. The thermally averaged
related to the Gibbs energy of formation of mean square displacement or the (x, y)
the protrusion. oscillator is given by r 2 . The surface
The overall process of transfer of an ion density of protrusions is termed σ and
I from a solvent A to a solvent B can be g is a structural factor. It is instructive
shown schematically as to consider one of the limiting situations
in which Keq Ak
diff is the rate-controlling
AnA I + nB B −−−→ nA A + IBnB (102) step. In this case the expression for krate
reduces to
Here nA and nB denotes a measure of the
A
solvation of I in A and B. krate = Keq kdiff
The attachment and detachment of an
ion to a protrusion in A is described by = 12 gπr 2 σ P (ha )ha kdiff (109)
the rate constants k A and k−
A . Using these
The expression for kdiff is essentially the
rate constants as well as the rate constant
same as that derived by Kakiuchi for an
of detachment of the ion in solvent B, it is
ion-transfer reaction given in Eq. (82). It
possible to describe the ion flux across the
thus immediately follows that the apparent
inner layer.
good agreement between experimentally
J = [k A ci (za ) − k−
A
ci (za )]P (ha )ha (103) determined ion-transfer rate constants
and Kakiuchi’s model is also consistent
B
J = k− ci (zb )P (hb )hb (104) with the Marcus model. Furthermore,
the small phenomenological diffusion
The overall rate constant of ion transfer is coefficient needed to fit the data to Eqs. (80
obtained as to 84) may simply be explained by a small
1 1 1 1 A.
Keq
= + +
krate A
kassn Ak
Keq diff
B kB
Keq The Marcus theory is difficult to apply
diss
(105) directly in a practical situation, owing to
In this equation the rate constants and lack of knowledge of the probability of
equilibrium constants may be estimated the formation of protrusions. One way
using the transition state theory. In this to overcome this problem could be to
treatment an activation energy can easily employ capillary wave theory for the in-
be incorporated in the expressions for k A terface between two immiscible electrolyte
B using the transition state theory.
and kdiss solutions. Recently, theoretical efforts have
been made to describe capillary waves at
A
kassn = k A P (ha )ha (106) soft electrified interfaces [83]. It may be
A 1 2 possible to use such theories to quantify
Keq = 2g πr σ P (ha )ha (107)
the value of P (h)h. One of the major
B
= 1 2 complications is related to the fact that the
Keq 2g πr σ P (hb )hb
surface tension is dependent on the Gal-
e−[µ(zb )−µ(za )]/kB T (108) vani potential difference between the two
phases, which means that P (ha )ha is In a model put forward by Indenbom,
potential-dependent. Theoretical [77] and the ITIES is considered as an elastic
experimental [84, 85] studies have been film in which forces of interfacial tension
undertaken to clarify the dependence of counteract the forces of an external
the surface structure on an electric field, electric field [64]. This causes a potential-
but more studies are clearly needed be- dependent compressed grooved structure
fore quantitative predictions can be made. of the interface at which ion transfer takes
In particular, a clear-cut experimental ev- place when the compression reaches a
idence of a link between protrusions at critical state. The validity of this picture
liquid|liquid interfaces and ion transfer is will rely on detailed molecular dynamics
still lacking. studies of the microscopic features of ion
transfer across ITIES. In this connection it
4.2.3.7 Other Models is relevant to mention the work of Cunnane
Another model that treats the ion transfer and coworkers in which the kinetics of
across ITIES as an activated transport pro- ion transfer across a monolayer adsorbed
cess by applying classical transition state at a liquid|liquid interface is discussed in
theory has been suggested by Girault and terms of the energy required to form a pore
Schiffrin [55]. Also in this case the equa- of a size similar to that of the transferring
tion describing the ion flux across the ion [56].
interface is similar to that obtained by Schmickler has treated the ion-transfer
the Butler–Volmer approach. Naturally, process within the framework of a lattice-
the interpretation of the α values and k ◦ gas model [66]. In this model the interface
is somewhat different in that case. Experi- is treated as an extended region in which
mentally [85] and theoretically [79], the ef- the two solvents mix. The reaction rate
fect of solvent viscosity on the ion-transfer is generally obtained using the transition
process has been confirmed indicating the state theory in combination with Kramers
importance of this parameter. theory depending on whether frictional
Aoki has modeled ion transfer as a com- forces are important.
bined transport and activated process [65]. The central idea in Schmicklers model
In this approach the activation energy for is to treat the liquid|liquid system as
ion transfer is a combination of an elec- an infinite three-dimensional lattice. The
tric potential energy, an activation energy fraction of sites occupied by the two kinds
due to solvation changes, and an energy of solvent molecules at a position z are
required to form a pore into the interface. termed f1 (z) and f2 (z).

The microscopic ion transport is described  1
1 − for z ≤ 0
using the Langevin equation. In certain 2e z/δ
f1 (z) =
cases, such as when viscous effects are 
 1 for z ≥ 0
pronounced, the theory is in good agree- 2e−z/δ
ment with the Butler–Volmer equation. (110)
A limiting factor in this theory is re- f2 (z) = 1 − f1 (z) (111)
lated to the fact that it cannot successfully
account for the experimentally observed In these equations, δ represents the width
concave variation of the logarithm to the of the interface.
ion-transfer rate constant with the Galvani The final result is shown to be in
potential difference. good agreement with a Butler–Volmer
behavior. Schmickler’s treatment may be these processes are strongly nonadiabatic.

useful in a theoretical modeling of the The maximum values of rate constants
transfer process, in particular to determine measured either by SECM or photoelec-
the potential of mean force acting on the trochemical techniques are of the order
ion. Qualitative experimental predictions of 10−19 cm4 s−1 , which coincide with
may also be made. This model can, the nonadiabatic limit obtained from the
for instance, successfully account for the models outlined in Sect. 4.2.2. The low
dependence of the solvation energy on electronic coupling between reactants is a
the ion-transfer rate constant. It may be result of the typical separation distances
noted that no assumptions regarding the at the interface. These distances are also
microscopic structure of the liquid|liquid responsible for reorganization energies
interface are made in this description close to 1 eV as estimated from experi-
and, for instance, surface tension is mental data.
not considered explicitly. An advantage Undoubtedly, a key development that
of this relatively simple model is that will clarify basic aspects in connection to
additional complications such as ion ET dynamic is the effective control on the
pairing of supporting electrolyte ions can distance separating the redox species. It
be described [86]. can be envisaged that redox couples linked
by functionalized spacers, which allow
4.2.4 highly ordered self-assembling structures,
Conclusion and Outlook could provide very valuable information on
the rate of interfacial ET and the effect of
The analysis presented in this review the interfacial potential. In this case, elec-
appears to show that while classical transi- tronic coupling between reactants could
tion state theory provides a good account be quite substantial and sophisticated
for experimental observation concerning modeling will be required to fully rational-
ET kinetics, the picture is less clear for ize these findings. Ultrafast photoinduced
ion-transfer processes. Our understand- processes can be effectively approached by
ing remains hindered by the primitive spectroscopic Pump-Probe methods down
knowledge of the ITIES structure at a to the femtosecond timescale.
microscopic level, despite current efforts The ion-transfer theories outlined in this
based on computer modeling [41, 51, 66, chapter are all based on idealized models.
87]. The potential distribution across the In practice, complications in the measure-
interface is also an essential piece of in- ments related to adsorption of the transfer-
formation that is yet to be clarified in the ring ion or chemical reactions associated
liquid|liquid context. Furthermore, these with the ion transfer arise. Detailed the-
analyses have excluded any specific inter- oretical descriptions of the microscopic
action of ions at the interface, which can nature of such complications are still lack-
also exert a profound effect on the reac- ing. One of the most interesting feature
tivity of the system. In order to address of simple ion transfers concerns the fact
all these aspects, it is crucial to develop that they always appear fast, and indeed
novel spectroscopic techniques coupled to as the experimental techniques have been
electrochemical systems. improved, the measured values have in-
Dynamic studies of heterogeneous ET creased even for similar ion transfers. One
across ITIES consistently suggest that way to clarify the microscopic features of
ion-transfer reactions could be to perform 20. M. Lher, R. Rousseau, E. Lhostis et al., C. R.

photoinduced ion-transfer reactions in a Acad. Sci. Paris, t. 322, Série IIb 1996, 322,
55–62.
controlled way. Such an experiment could,
21. Q. Z. Chen, K. Iwamoto, M. Seno, Elec-
for instance, be carried out using a surfac- trochim. Acta 1991, 36, 291–296.
tant that upon illumination generates an 22. S. Daniele, M. A. Baldo, C. Bragato, Elec-
ion that can transfer across the interface. trochem. Commun. 1999, 1, 37–41.
23. H. H. Girault, see the web site dcwww.epfl.ch
for a comprehensive list of free energy of ion
References transfer at various liquid|liquid interfaces.
24. A. A. Stewart, J. A. Campbell, H. H. Girault
1. F. Reymond, D. J. Fermı́n, H. J. Lee et al., et al., Ber. Bunsen-Ges. Phys. Chem. 1990, 94,
Electrochim. Acta 2000, 45, 2647–2662. 83–87.
2. A. G. Volkov, D. W. Deamer, Liquid-Liquid 25. Z. Ding, P.-F. Brevet, H. H. Girault, Chem.
Interfaces: Theory and Methods, CRC Press, Commun. 1997, 2059–2060.
Boca Raton, Fla., 1996. 26. Z. F. Ding, D. J. Fermin, P. F. Brevet et al., J.
3. A. G. Volkov, D. W. Deamer, D. I. Tanelian Electroanal. Chem. 1998, 458, 139–148.
et al., Liquid Interfaces in Chemistry and 27. J. Zhang, C. J. Slevin, L. Murtomaki et al.,
Biology, John Wiley & Sons, New York, 1998. Langmuir 2001, 17, 821–827.
4. R. Lahtinen, D. J. Fermı́n, K. Kontturi et al., 28. J. Zhang, P. R. Unwin, J. Phys. Chem. B 2000,
J. Electroanal. Chem. 2000, 483, 81–87. 104, 2341–2347.
5. D. J. Fermı́n, R. Lahtinen in Dynamic Aspects 29. B. Quinn, R. Lahtinen, L. Murtomaki et al.,
of Heterogeneous Electron Transfer Reactions Electrochim. Acta 1998, 44, 47–57.
at Liquid|Liquid Interfaces (Ed.: A. Volkov), 30. A. J. Bard, D. E. Cliffel, C. Demaille et al.,
Marcel Dekker, Boca Raton, Fla., 2001, Ann. Chim. 1997, 87, 15–31.
pp. 179–228. 31. A. L. Barker, J. V. Macpherson, C. J. Slevin
6. Z. Samec, V. Marecek, J. Weber, J. Elec- et al., J. Phys. Chem. B 1998, 102, 1586–1598.
troanal. Chem. 1978, 96, 245–247. 32. A. L. Barker, P. R. Unwin, S. Amemiya et al.,
7. Z. Samec, J. Electroanal. Chem. 1979, 99, J. Phys. Chem. B 1999, 103, 7260–7269.
197–205. 33. M. Tsionsky, A. J. Bard, M. V. Mirkin, J.
8. Y. I. Kharkats, Sov. Electrochem. 1976, 12, Phys. Chem. 1996, 100, 17 881–17 888.
1257. 34. M. Tsionsky, A. J. Bard, M. V. Mirkin, J. Am.
9. H. H. J. Girault, D. J. Schiffrin, J. Electroanal. Chem. Soc. 1997, 119, 10 785–10 792.
Chem. 1988, 244, 15–26. 35. C. Wei, A. J. Bard, M. V. Mirkin, J. Phys.
10. R. A. Marcus, J. Phys. Chem. 1990, 94, Chem. 1995, 99, 16 033–16 042.
4152–4155, 7742. 36. B. B. Smith, J. W. Halley, A. J. Nozik, Chem.
11. R. A. Marcus, J. Phys. Chem. 1990, 94, Phys. 1996, 205, 245–267.
1050–1055. 37. H. H. Girault, J. Electroanal. Chem. 1995,
12. R. A. Marcus, J. Phys. Chem. 1991, 95, 388, 93–100.
2010–2013. 38. I. Benjamin, Y. I. Kharkats, Electrochim. Acta
13. Y. Cheng, D. J. Schiffrin, J. Electroanal. 1998, 44, 133–138.
Chem. 1991, 314, 153–163. 39. N. Sutin, Acc. Chem. Res. 1982, 15, 275.
14. Y. Cheng, D. J. Schiffrin, J. Chem. Soc., 40. J. T. Hupp, M. J. Weaver, J. Electroanal.
Faraday Trans. 1993, 89, 199–205. Chem. 1983, 152, 1–14.
15. Y. F. Cheng, D. J. Schiffrin, J. Chem. Soc., 41. W. Schmickler, J. Electroanal. Chem. 1997,
Faraday Trans. 1994, 90, 2517–2523. 428, 123–127.
16. V. J. Cunnane, G. Geblewicz, D. J. Schiffrin, 42. B. Liu, M. V. Mirkin, J. Am. Chem. Soc. 1999,
Electrochim. Acta 1995, 40, 3005–3014. 121, 8352–8355.
17. G. Geblewicz, D. J. Schiffrin, J. Electroanal. 43. D. J. Fermı́n, Z. Ding, H. Duong et al., J.
Chem. 1988, 244, 27–37. Phys. Chem. B 1998, 102, 10 334–10 341.
18. S. Kihara, M. Suzuki, K. Maeda et al., J. 44. D. J. Fermı́n, H. Doung, Z. Ding et al., Elec-
Electroanal. Chem. 1989, 271, 107–125. trochem. Comm. 1999, 1, 29–32.
19. E. Makrlik, Z. Phys. Chem. 1987, 268, 45. D. J. Fermı́n, H. Duong, Z. Ding et al., Phys.
212–214. Chem. Chem. Phys. 1999, 1, 1461–1467.
46. D. J. Fermı́n, H. Duong, Z. Ding et al., J. 67. L. Murtomäki, A. K. Kontturi, D. J. Schiffrin,

Am. Chem. Soc. 1999, 121, 10 203–10 210. J. Electroanal. Chem. 1999, 474, 89–93.
47. H. Jensen, J. J. Kakkassery, H. Nagatani 68. R. A. Marcus, J. Chem. Phys. 2000, 113,
et al., J. Am. Chem. Soc. 2000, 122, 1618–1629.
10 943–10 948. 69. Z. Samec, V. Marecek, J. Weber et al., J.
48. N. Eugster, D. J. Fermı́n, H. H. Girault, J. Electroanal. Chem. 1981, 126, 105–119.
Phys. Chem. B 2002, 106, 3428–3433. 70. P. D. Beattie, A. Delay, H. H. Girault, Elec-
49. P. Vanysek, Electrochim. Acta 1995, 40, trochim. Acta 1995, 40, 2961–2969.
2841–2847. 71. T. Kakiuchi, Y. Takasu, J. Phys. Chem. B
50. H. H. Girault, Mod. Aspects Electrochem. 1997, 101, 5963.
1993, 25, 1–62. 72. Z. F. Ding, F. Reymond, P. Baumgartner
51. I. Benjamin, Annu. Rev. Phys. Chem. 1997, et al., Electrochim. Acta 1998, 44, 3–13.
48, 407. 73. Z. F. Ding, R. G. Wellington, P. F. Brevet
52. Z. Samec, T. Kakiuchi, Adv. Electrochem. Sci. et al., J. Electroanal. Chem. 1997, 420, 35–41.
Eng. 1995, 4, 297–361. 74. D. J. Fermı́n, Z. Ding, P. F. Brevet et al., J.
53. C. Gavach, B. D’Epenoux, F. Henry, J. Elec- Electroanal. Chem. 1998, 447, 125–133.
troanal. Chem. 1975, 64, 107–115. 75. I. Benjamin, Chem. Rev. 1996, 96, 1449–1475.
54. J. Koryta, P. Vanysek, M. Brezina, J. Elec- 76. I. Benjamin, Science 1993, 261(5128),
troanal. Chem. 1977, 75, 211–228.
1558–1560.
55. H. H. J. Girault, D. J. Schiffrin, J. Electroanal.
77. K. J. Schweighofer, I. Benjamin, J. Elec-
Chem. 1985, 195, 213–227.
troanal. Chem. 1995, 391, 1–10.
56. V. J. Cunnane, D. J. Schiffrin, M. Fleisch-
78. P. A. Fernandes, M. N. D. S. Cordeiro, J. A.
mann et al., J. Electroanal. Chem. 1988, 243,
N. F. Gomes, J. Phys. Chem. B 1999, 103,
455–464.
6290–6299.
57. I. Benjamin, J. Chem. Phys. 1992, 96,
79. R. Ferrigno, H. H. Girault, J. Electroanal.
577–585.
58. Y. Y. Gurevich, Y. I. Kharkats, J. Electroanal. Chem. 2001, 496, 131–136.
Chem. 1986, 200, 3–16. 80. Z. Samec, V. Marecek, D. Homolka, J. Elec-
59. Z. Samec, Y. I. Kharkats, Y. Y. Gurevich, J. troanal. Chem. 1985, 187, 31–51.
Electroanal. Chem. 1986, 204, 257–266. 81. Q. Cui, G. Zhu, E. Wang, J. Electroanal.
60. T. Kakiuchi, J. Electroanal. Chem. 1992, 322, Chem. 1995, 383, 7–12.
55–61. 82. L. I. Daikhin, A. A. Kornyshev, M. Urbakh, J.
61. T. Kakiuchi, J. Noguchi, M. Senda, J. Elec- Electroanal. Chem. 2000, 483, 68–80.
troanal. Chem. 1992, 327, 63–71. 83. Z. H. Zhang, I. Tsuyumoto, S. Takahashi
62. K. Kontturi, J. A. Manzanares, L. Murtomae- et al., J. Phys. Chem. A 1997, 101, 4163–4166.
ki, Electrochim. Acta 1995, 40, 2979–2984. 84. Z. H. H. Zhang, I. Tsuyumoto, T. Kitamori
63. M. Senda, Electrochim. Acta 1995, 40, et al., J. Phys. Chem. B 1998, 102,
2993–2997. 10 284–10 287.
64. A. V. Indenbom, Electrochim. Acta 1995, 40, 85. Y. Shao, H. H. Girault, J. Electroanal. Chem.
2985–2991. 1990, 282, 59–72.
65. K. Aoki, Electrochim. Acta 1996, 41, 86. C. M. Pereira, W. Schmickler, F. Silva et al.,
66. W. Schmickler, J. Electroanal. Chem. 1997, 87. I. Benjamin, ACS Symp. Ser. 1994, 568,
426, 5–9. 409–422.
4.3 The Influence of Ad-atom Adsorption on Reaction Rates and Mechanisms 391
4.3 This equation makes it clear that electro-

The Influence of Ad-atom Adsorption on catalysis consists of two terms: an electrical
Reaction Rates and Mechanisms one (the term depending on the poten-
Georgios Kokkinidis tial difference across the interface) and
Aristotle University of Thessaloniki, Thessa- a chemical one (the Arrhenius term),
loniki, Greece which depends on the nature of the elec-
trode material.
4.3.1
Although by applying high overpoten-
Introduction
tials one can increase the rate of an
It is well known that many chemical re- electrochemical reaction on a given elec-
actions do not by themselves occur at a trode/catalyst even by more than 10 orders
significant rate, although thermodynam- of magnitude, the activity of the electrode,
ics predict that they are very favorable. The the Arrhenius term, still remains the most
rate of such reactions, however, may be important fact in electrocatalysis. This is
increased by several orders of magnitude obvious, since the energy efficiency of any
by a catalyst. The catalyst (homogeneous cell is determined mainly by the overpoten-
or heterogeneous) is a substance, which al- tials necessary at the anode and cathode.
ters the rate of a chemical reaction without The lower the overpotentials, the better
itself being consumed or generated in the is the energy efficiency. Furthermore, the
process. Similarly, there are many electro- application of high overpotentials usually
chemical reactions that also do not proceed introduces problems due to the inter-
with a sufficient rate because of poor ki- ference of undesired reactions (e.g. the
netics. For these reactions, it is again hydrogen or the oxygen evolution reaction,
necessary to find an active electrode/catalyst the electrodissolution of the electrode,
or, better, an electrocatalyst, otherwise high etc.).
overpotentials should be applied in or- The catalytic activity of an electrode
der to promote the reaction. Thus, the depends on its electronic – and perhaps
term electrocatalysis was introduced for the crystallographic – properties. When the re-
heterogeneous charge-transfer reactions actants or intermediates are adsorbed on
that take place at the electrode/electrolyte the electrode surface, the reaction rate will
solution interface. The objective of electro- depend additionally on the heat of adsorp-
catalysis is to discover electrocatalysts that tion, on the strength of bond-forming or
allow electrode reactions to occur with a bond-breaking, and on the properties of
sufficient rate at potentials very close to the the adsorption isotherm. All these factors
equilibrium potential and, in some cases, may be changed by the modification of the
to improve the selectivity with respect to electrode surface from adsorbed species
the reaction products. existing in the solution, not participating
The catalytic role of an electrode is in the reaction. These species can some-
revealed by writing the rate of an elec- times improve the catalytic activity of the
trochemical reaction as follows [1]: bare material. For example, heavy metal
atoms (e.g. Pb, Tl, and Bi) deposited in the
j kT −
Go# /RT
ν= = c e underpotential region have been shown
nF h
to increase the catalytic activity of no-
× e−αF
/RT (for a cathodic process) ble metal electrodes (e.g. Pt, Au, Ag, Pd,
(1) etc.) for several technologically interesting
electrochemical reactions. Such reactions fundamental and practical significance in

are the reduction of oxygen, the oxida- electrocatalysis.
tion of organic fuels, the reduction of The UPD process was studied for a large
aromatic nitro compounds, and so forth. number of systems using both polycrys-
Although noble metals are considered to talline and monocrystalline electrodes as
be the best electrocatalysts in contempo- substrates and the progress of understand-
rary electrocatalysis, their modification by ing all related parameters was covered in
foreign metal ad-atoms deposited in the depth in several review articles [2–5]. Here
underpotential region may improve their we will give only an outline of some ther-
catalytic activity for certain electrochemi- modynamic and structural aspects, which
cal reactions. will be useful for the better understanding
of the role of the UPD ad-atoms as reaction
4.3.2 modifiers in electrocatalysis.
Metal Ad-atoms Deposited in the The equilibrium potential for the UPD
Underpotential Region process (Reaction 3) may be formulated
as [3]
Foreign metal ad-atoms can be deposited
on an electrode electrochemically in the RT fM z+ cM z+
Eads (θ) = E o + ln
so-called underpotential region, that is, zF fads θ
the potential region positive to the re- (4)
versible Nernst potential for the bulk where fM z+ and fads are the activity
metal deposition. coefficients of the metal ion in solution
and the bulk deposit (fads = 1 for θ ≈ 1),

E = Eads − ENernst ≥ 0 (2) cM z+ is the metal ion concentration
in solution, and θ the UPD ad-atom
The underpotential deposition (abbrevi- coverage. Since θ takes values between
ated to UPD) of a metal M on a foreign 0 and 1, Eq. (4) yields an equilibrium
metal substrate M is a process that takes potential that is always more positive
place according to the reaction than the corresponding Nernst potential
for the bulk metal deposition. Moreover,
M z+ + ze− ←
−−−
−−
→ Eq. (4) implies that the submonolayer
− Mads (3)
equilibrium potential shifts toward more
This Faradaic adsorption of M z+ ions dis- positive potentials by 60/z mV dec−1 for
solved in the solution of the supporting constant θ with increasing the metal ion
electrolyte leads to the formation of sub- concentration in solution.
and monolayer coverages of discharged M There are two excess parameters, the
ad-atoms and constitutes the first stage charge σ = (∂γ /∂E)µ and the surface con-
in electrocrystallization. The UPD process centration = −(∂γ /∂µ)E from which
was studied rather extensively because of the electrosorption valency may be defined
its importance in investigating a wide vari- and experimentally determined:
ety of electrochemical phenomena, such
1 ∂σ
as adsorption on electrified interfaces, = (5)
F ∂ E
charge-transfer reactions, surface diffu-
sion, nucleation and growth, double-layer The symbols in all the above expressions
changes, and, of course, because of its have their usual meaning.
The electrosorption valency expresses From theoretical considerations of bond

the charge flow during the electrosorption formation in electrosorbates it was con-
process at a constant potential. It takes cluded that the underpotentially deposited
values ranging between zero and the metal ad-atoms can approximately be
ionic charge z. Values very close to z considered to be covalently bonded and
show that the adsorbed species are almost nearly completely discharged if the differ-
completely discharged, while values of ence in the Pauling’s electronegativities
lower than z indicate adsorption with of the substrate and adsorbed metal is
partial charge transfer. The adsorbed |
χ| ≤ 0.5 [6]. That means these systems
species are recognized as ad-atoms when have electrosorption valency /z ≈ 1. Sys-
takes values very close or equal to z. tems in which the electrosorption valency
Saturate coverages with ad-atoms are ob- is equal to the ionic charge are those re-
tained at potentials close to the potential for sulting from the UPD of heavy metal ions
the bulk metal deposition. The maximum on noble and transition metal electrodes.
surface concentration is about max ≈ There are a number of parameters that
2 × 10−9 mol cm−2 . This surface concen- can actually determine the structure of the
tration corresponds to a maximum amount UPD monolayers [4]. The most outstand-
of charge σmax ≈ z 200 µC cm−2 [3, 4]. ing ones are certainly the structure of the
The extent of the UPD region depends substrate, the ratio between the diameter
on the binding energies between the of the adsorbed species and the distance
adsorbed atoms and the atoms of the of two adjacent adsorption sites, and of
substrate metal. There exists a linear course the degree of the UPD ad-atom
relationship between
Ep and
, where coverage. At low surface coverages a UPD
is the difference in the work functions metal monolayer consists of isolated M

between the substrate and the adsorbed ad-atoms or partially discharged and desol-
metal. This linear correlation is expressed vated ions, which are randomly distributed
by a semiempirical Eq. (3): on the substrate owing to the lateral re-
pulsion. With the increase in the degree

Ep = 0.5
(6) of coverage some ordering occurs and at
high surface coverages the UPD layer may
Ep is defined as the difference between form a sequence of two-dimensional peri-

the bulk and the most positive adsorp- odic superlattice structures with electronic
tion peak of the corresponding cyclic properties closely approaching those of the
voltammogram for the UPD. The lin- bulk metal.
ear correlation holds for systems with On single-crystal electrodes the UPD
polycrystalline substrate electrodes, while deposits often form well-defined, or-
for systems with single-crystal electrodes dered monolayers. Hypothetical superlat-
some deviations were observed [3]. This tice structures have been proposed, which
behavior suggests that the binding ener- rely on the charges determined for the var-
gies of the UPD are not only determined ious peaks observed in the cyclic voltam-
by the difference in the electronegativities mograms. The anions of the supporting
expressed as
but they are also influ- electrolytes influence the shape and the
enced by the lateral interactions that are position of the UPD peaks in the cyclic
governed by the crystallographic orienta- voltammograms, although the superlat-
tion of the substrate. tice structures are not always affected by
interactions between the ad-atoms and layers to their influence on the catalytic
the anions adsorbed on the UPD mono- activity of the substrate. The reason is that
layer [7]. the superlattice structures of the ad-atom
In the beginning, conventional electro- layers are influenced by the strong ad-
chemical techniques had been used to sorption of intermediates involved in the
study the UPD, both on polycrystalline mechanism of many electrocatalytic reac-
and single-crystal electrodes. The work tions. For example, the rate of formic acid
done on single-crystal electrodes included oxidation on Pt modified by Pb ad-atoms is
both low index planes and stepped sur- different if lead is adsorbed before or after
faces. The technique developed by Clavilier formic acid [25].
et al. [8, 9] (the so-called flame annealing
technique) has led to a revolution in elec- 4.3.3
trochemical studies, since single crystals Electrocatalysis by Ad-atoms
of platinum and gold can be relatively eas-
ily formed and precisely oriented using an Modified electrodes with ad-atoms (usually
optical technique. Single-crystal surfaces deposited in the underpotential region)
were previously prepared in an ultrahigh exhibit enhanced electrocatalytic activity
vacuum (UHV) chamber by sputtering for several categories of electrochemical
and annealing, and their structure was reactions. The most extensively studied
defined by low energy electron diffraction
reactions are those related to the de-
(LEED) [10–12]. The results of the various
velopment of low-temperature fuel cell
techniques were compared mainly in con-
technology, namely, the reduction of oxy-
nection with the hydrogen adsorption on
gen and the oxidation of organic fuels. The
platinum [13, 14].
ad-atoms may influence the rate and the
The electrochemical methods provide
mechanism of electrochemical reactions
only indirect information about the struc-
through [26–29]
ture of the UPD ad-atom layers in an
atomic scale. In order to obtain more di-
rect information about the structure of the 1. the third-body effect,
UPD ad-atom layers, many investigators 2. the prevention of poison formation,
have adopted the use of in situ techniques, 3. the bifunctional mechanism, and
in which the electrode surface is exam- 4. the modification of the electronic prop-
ined with surface science methods. The erties of the electrode surface.
methods mainly used are in situ X-ray
diffraction [7, 15–17], in situ scanning tun- In the third-body mode of action, the
neling microscopy (STM) [18–23], and in ad-atoms are considered as inert species,
situ atomic force microscopy (AFM) [24]. actually not participating in the reaction.
These methods have provided detailed in- Their presence acts only to diminish the
formation on the atomic structure, the places where the poison formation reac-
thermodynamic stability, and the depen- tion can take place. The poisoning reaction
dence of the structure on the potential requires greater number of neighboring
for several UPD systems. However, little sites (say Pt surface atoms) not occu-
attempt has been made (and this mainly pied by ad-atoms than the main reaction
concerns the reduction of oxygen) to cor- that proceeds in parallel through reactive
relate directly the structure of the ad-atom (not strongly adsorbed) intermediates. The
enhanced reactivity in the presence of ad- substrate atom and an adjacent ad-atom
atoms results from decreased poisoning are involved.
and, thereby, an increase in the rate of the Finally, the ad-atoms may cause lo-
reaction branch, which requires smaller cal electronic changes on the substrate,
number of neighboring sites. thus affecting the reactant–substrate in-
Blocking of poison formation through teractions. The role of ad-atoms on the
the third-body mechanism is physical. Ad- modification of the catalytic properties of
atom coverages at around 0.5, although metal surfaces has been extensively stud-
permitting the adsorption/dehydrogena- ied [30–32]. The behavior of these systems
tion step (which is necessary for initiating has been rationalized in terms of changes
the oxidation of an organic molecule), in adsorption bond strength of the reac-
prevent the strong adsorption of inter- tant species caused by changes in catalyst
mediates, which poison the electrocatalyst (metal) work function, which results from
surface. In some cases, it is believed the electronic interactions of the ad-atoms
that adsorbed hydrogen is involved in the with the active metal surface [30–32].
poison formation reaction. The ad-atoms
block the adsorption of hydrogen, thus in- 4.3.4
hibiting the poison reaction. The required Reactions Catalyzed by Ad-atoms
coverage by the ad-atoms should be greater
than 0.5 in the hydrogen potential region 4.3.4.1 Reduction of Oxygen
to avoid poison formation. The reduction of oxygen is one of the
The bifunctional mechanism refers not most extensively studied electrochemical
only to catalysts modified by ad-atoms reactions. There are two reasons for the
but also to electrocatalysis on bimetallic increased interest for this reaction, firstly,
surfaces. The oxidation of methanol on its fundamental significance with respect
to electrode kinetics and, secondly, its role
Pt−Ru alloy surfaces is a typical exam-
in the area of electrochemical energy
ple of bifunctional mechanism. The role
conversion and corrosion. The kinetics and
of the modifier is to supply oxygen for
mechanism of oxygen electroreduction
the oxidation of either reactive or poison-
greatly depends on the electrode material.
ing intermediates. The potential in which
Oxygen reduction is also a very sensitive
increased reaction rates are observed is
reaction to the surface structure, as shown
governed by the potential in which the
by studies on single-crystal electrodes
adjacent modifier or ad-atom adsorbs oxy-
of different metals. There are several
gen, which should be less positive than
extended reviews and book chapters,
that of platinum. The bifunctional mecha-
which cover all these aspects of oxygen
nism reduces the oxidation overpotential,
electroreduction [33–37].
which is particularly important for fuel
In aqueous solutions, oxygen reduction
cells. The bifunctional mechanism refers
is considered to proceed by two overall
also to catalysts on which chemisorption reaction pathways [38].
of reducible groups and reductive cleavage
of bonds might increase the electrocatalyst The direct four-electron pathway:
selectivity. A bridge adsorption for the re-
active intermediate is assumed to explain O2 + 4H+ + 4e− −−−→ 2H2 O
the selective bond cleavage, in which a (acid solutions) (7)
O2 + 2H2 O + 4e− −−−→ 4OH− O O O O O O

(alkaline solutions) (8) M M M M
The peroxide pathway:

The end-on adsorption (Pauling’s model)
O2 + 2H+ + 2e− −−−→ H2 O2
favors the two-electron reduction of O2 to
(2H+ + 2e− ) H2 O2 . The bridge model, with the two
−−−−−−−−→ 2H2 O
bonds with two metal sites, favors the
(acid solutions) direct four-electron reduction of O2 to H2 O
(9) through rupture of the O−O bond.
−OH− Metal ad-atoms deposited in the under-
O2 + H2 O + 2e− −−−−→ HO2 −
potential region strongly influence the rate
(H2 O + 2e− ) and mechanism of oxygen reduction. Pos-
−−−−−−−−→ 3OH− itive and/or negative catalytic effects were
(alkaline solutions) observed depending on the substrate elec-
(10) trode. The work done was reviewed by
Depending on the electrode material, the Adzic [26] and Kokkinidis [27]. It was re-
reduction of oxygen occurs by one or the ported that Pb, Tl, and Bi ad-atoms catalyze
other reaction pathway. The two reaction the reduction of oxygen on Au and car-
pathways may also take place in parallel. bon electrodes. The half-wave potential
The most powerful technique for the is shifted toward less negative poten-
quantitative determination of the extent tials and the two-electron reduction to
of these reactions was proved to be the hydrogen peroxide is transformed to the
rotating ring-disc electrode voltammetry, four-electron reduction to water. The cat-
which allows the detection of hydrogen alytic action is mainly observed in alkaline
peroxide on the ring electrode. electrolytes and at potentials of low UPD
On some electrodes, hydrogen perox- ad-atom coverages. Similar behavior was
ide cannot undergo further reduction and also observed on ruthenium [40]. Ruthe-
is obtained as the final reaction prod- nium partially oxidized, which is inactive
uct. Electrodes on which the two-electron for O2 reduction in alkaline solutions, can
reduction to hydrogen peroxide predom- be made very active by Pb and Tl ad-atoms.
inates are Au, Hg, and various carbon Traces of Pb and Tl are sufficient to ac-
materials. Electrodes on which the four- tivate the four-electron reduction on this
electron reduction to water is predominant electrode.
include Ag, Pt, and the Pt family metals. In contrast to positive catalysis on gold
Platinum is considered to be the best elec- and carbon electrodes, Pb, Tl, and Bi ad-
trocatalyst for oxygen reduction, since the atoms cause a negative catalytic effect on
reduction on this electrode occurs with the Pt and Ag electrodes. The four-electron
lowest overpotential. The reaction path- reduction on the bare metals changes to
way that is more favorable depends on the two-electron reduction on the electrodes
way molecular oxygen is adsorbed on the covered by the modifiers. This partial
metal surface. Three different models for inhibition occurs at negative potentials, in
adsorption have been proposed, the Grif- the mass-transport control region, where
fith’s model, the Pauling’s model, and the almost complete UPD ad-atom coverages
bridge model [38, 39]: are obtained. Finally, the reduction of
oxygen on Pt [41] and Au [42] is completely reflects the concentration-dependence of

inhibited by Cu ad-atoms deposited at the UPD process according to Eq. (4).
underpotentials. Figure 2 shows ring-disc measurements
Figure 1 shows, as an example, the for O2 reduction on Au and Au/Tlads
rotating-disc voltammograms of oxygen in an alkaline solution, obtained from
reduction on Ag(100) in the absence Ref. [44] (Fig. 2a), and on Pt and Pt/Pbads
and in the presence of Pb2+ ions in in an acid solution, obtained from Ref. [45]
solution [43]. The cyclic voltammogram for (Fig. 2b). The ring currents indicate the
the UPD of Pb on the Ag(100) single- production of hydrogen peroxide as a
crystal electrode is given for comparison. function of the disc potential. The positive
It should be noted that the observed catalytic effect observed on Au and Au(hkl)
potential shift of the inhibition effect electrodes was attributed by Jüttner [46] to
4
[µA cm−2]
0
j
−4
(b) −8
0.0
−0.1 (c)
E vs NHE
[V]
−0.4
[mA cm−2]
−0.2 −5
5.10 5.10−3
cPb2+/M
j
−0.8 4
5
2 3
1
−1.2
−0.2 0.0 0.2
E vs NHE
(a) [V]
Fig. 1 (a) Current–potential curves of O2 reduction on Ag(100) in
aerated solution of 0.5 M HClO4 + xM Pb(ClO4 )2 . (ν = 1 mV s−1 ;
f = 25 Hz). x/M: (1) 0; (2) 5 × 10−5 ; (3) 5 × 10−4 ; (4) 5 × 10−3 ; and
(5) 5 × 10−2 . (b) Cyclic voltammogram for the UPD of Pb on
Ag(100) in 0.5 M HClO4 + 5 × 10−3 M Pb(ClO4 )2 . (ν = 10 mV s−1 ).
(c) Potential variation of the descending part of curves 2 to 5 versus
cPb 2+ . (Reproduced from Ref. [43] with permission.)
0.8
0.2
[µA]
iR
[mA]
iR
0.1 0.0
Pt
0.0 Pt/Pb
−0.8
0.0
Au 400
Au/TI
[mA]
iD
−0.4
−1.6
[mA]
iD
−0.8
−2.4 1400 rpm
−1.2
−0.6 −0.4 −0.2 0.0 0.0 0.4 0.8
ED vs Hg/HgO ED vs RHE
(a) [V] (b) [V]
Fig. 2 (a) Current–potential curves of O2 Pt−Pt rrde in 0.2 M HClO4 without (full lines)
reduction on a Au−Au rrde in 0.1 M NaOH and with (dashed lines) 2 × 10−5 M Pb2+
without (full line) and with (dashed line) (ν = 40 mV s−1 ). ER = 1.2 V. (Reproduced from
6.6 × 10−6 M Tl(I) (ν = 10 mV s−1 ). ER = 0.1 V. Refs. [44, 45] with permission.)
(b) Current-potential curves of O2 reduction on a
the change of the adsorption configuration AFM observations, Chen and Gewirth [24]
from end-on on bare Au to bridge reported that the maximum catalytic activ-
adsorption on Au partially covered by ity for H2 O2 electroreduction on Au(111)
the modifiers Au−O−O−M (M = Pb, Tl, modified by Bi UPD monolayers is ob-
Bi). As mentioned earlier, this mode of served in a narrow potential region, where
adsorption favors the O−O bond cleavage. the (2 × 2)Bi structure is obtained. At
At high ad-atom coverages the bridge potentials negative to this region, the rect-
adsorption does not exist any more and angular Bi structure is formed and the
oxygen is reduced to H2 O2 as on bulk Pb, reduction of H2 O2 is almost completely
Tl, and Bi electrodes. suppressed. The (2 × 2)Bi structure en-
The influence of the structure of the ables the heterobimetallic bridge adsorp-
ad-atom layers on the reduction of oxy- tion (Au−O−O−Bi), which, according to
gen was examined by using in situ optical the views of Jüttner [46], may explain the
and scanning probe microscopies. With enhanced reactivity.
these techniques reliable information on More recent studies combined the in
the structure of the ad-atom layers can be situ techniques with the use of single-
obtained during the course of oxygen or crystal electrodes in a rotating geometry,
hydrogen peroxide reduction. In most sys- the so-called rotating hanging menis-
tems studied, the UPD adlattice structures cus electrode (RHME) [47–49] or in the
are not affected by the presence of molec- most desired geometry of a rotating ring-
ular oxygen in solution. On the basis of disc electrode. This combination provides
unique possibilities for studying struc- 4.3.4.2 Reduction of Nitro and Nitroso
tural effects in O2 electrocatalysis. Adzic Compounds
et al. [50] studied the effect of Tl adlayers The reduction of nitro compounds has
on the reduction of oxygen on Au(111) in played an important role in the devel-
acid solutions by using X-ray diffraction opment of organic electrochemistry. The
and the RHME. They found that the re- nitro group is one of the most easily re-
duction mechanism changes from a four- ducible groups. In aqueous solutions the
electron reduction at a low Tl coverage to a general reduction mechanism, first pro-
two-electron reduction at a higher Tl cov- posed by Haber, is
erage. The (2 × 2)Tl structure is conducive
(2e− + 2H+ )
to the four-electron reduction, while the RNO2 −−−−−−−−→ RNO
close-packed rotated hexagonal Tl phase, −H2 O
which exists at more negative potentials, is (2e− + 2H+ )
responsible for the two-electron reduction. −
←−−
−−−−
−−−−
−−
→− RNHOH
The reduction of oxygen does not change
the Tl coverage, although it decreases the (2e− + 2H+ )
in-plane diffraction intensity. This indi- −−−−−−−−→ RNH2 (11)
−H2 O
cates that O2 molecules interact with the
Tl ad-atoms and provide direct evidence According to this mechanism, nitro com-
that oxygen reduction is, indeed, an inner- pounds are reduced to the corresponding
sphere and not an outer-sphere reaction. hydroxylamines or amines by an irre-
Finally, it is worth mentioning the versible four- or six-electron process that
pronounced inhibition effect of oxygen re- proceeds through nitroso intermediates.
duction on Au(100) and Pt(111) electrodes Nitroso intermediates are not stable on the
caused by bromide adsorption (Fig. 3). It electrode surface because their reduction
is well known that the specific adsorp- potential is more positive than that of the
tion of halide anions inhibit the reduction initial nitro compounds. Azoxy, azo, and
of oxygen, but information of the inhibi- hydrazo derivatives can also be obtained
tion mechanism in an atomic level was by electrolysis, particularly in alkaline so-
attempted for the first time by Adzic lutions, as a result of chemical follow-up
and
√ Wang √ [51]. As shown in Fig. 3, the reactions rather than electrochemical reac-
c( 2 × 2 2)R45◦ adlayer structure on tions. The electrochemistry of all these
Au(100) and the (3 × 3)Br structure on compounds has been discussed in re-
Pt(111) cause a complete inhibition of views [52–54]. Among the different nitro
the two-electron reduction on Au(100) and compounds, nitrobenzene has played a
the four-electron reduction on Pt(111) sur- central role and its reduction is one of the
faces, respectively. The reduction can take most thoroughly studied electrochemical
place only at potentials more negative than reaction. The reduction involves adsorp-
the low potential limit of the existence tion of reactants and intermediates and
of these structures. The stability of the follows direct electron-exchange mecha-
structure of the Br adlayers during oxygen nism on most cathodes except of platinum,
reduction was checked by measuring the in which the reduction proceeds by an elec-
X-ray intensity as a function of the poten- trocatalytic hydrogenation mechanism.
tial in the absence and in the presence of The influence of ad-atoms on the reduc-
oxygen in solution. tion of nitro and nitroso compounds on
4 (i)
Au(100), Br− (i)
Pt(111), Br−
[µA cm−2]
50
0
j
[µA]
0
i
−4 (3 × 3)Br
−50
(ii)
Intensity
0.8
0.4 0.0 (ii)
c(√2 × 2√2)
0.0
−0.4
Pt(111), Br−
[mA]
(iii)
i
0.0
[mA cm−2]
−0.8
−
Au(100), Br Pt(111)
j
−0.2
Au(100)
−1.2
−1.4
0.0 0.4 0.8 0.0 0.4 0.8
E vs NHE E vs NHE
(a) [V] (b) [V]
Fig. 3 (a) (i) Voltammetry curve of the Au(100) electrode in 0.1 M HClO4 + 20 mM Br−
(ν = 10 mV s−1 ). (ii) X-ray intensity at (1/2, 1/2, 0.1) position in the absence (dashed line) and
in the presence (full line) of oxygen. (iii) O2 reduction on Au(100) without and with 20 mM Br−
(ν = 20 mV s−1 ). (b) (i) Voltammetry curve of the Pt(111) electrode in 0.1 M HClO4 + 20 mM
Br− (ν = 20 mV s−1 ). (ii) O2 reduction on a hanging meniscus rotating Pt(111) disc electrode
in 0.1 M HClO4 without and with 20 mM Br− (ν = 20 mV s−1 ; f = 625 rpm). (Reproduced
from Ref. [51] with permission.)
Ag, Au, and Pt electrodes has been studied solutions. On the bare metals, nitroben-
during the last two decades [55–61]. It was zene is reduced to aniline. The UPD of
found that ad-atom layers of Pb, Tl, and Bi Pb on Ag and Au causes a partial inhibi-
deposited at underpotentials have a strong tion of nitrobenzene reduction. The step of
influence on the performance of the above the reduction of the intermediate hydrox-
substrates for the nitro and nitroso group ylamine to aniline is completely inhibited
electroreduction, both in terms of the im- when the silver and gold substrates are
provement of the catalytic activity and the screened off by a lead monolayer. Figure 4
catalytic selectivity of these electrodes. shows the polarization curves for the ni-
trobenzene and nitrosobenzene reduction
Reduction of nitrobenzene on Ag/Pb(upd) on Ag(111) and Ag(111)/Pb(upd) elec-
and Au/Pb(upd) The electrocatalytic in- trodes. The limiting current of nitroben-
fluence of Pb(upd) adlayers on the zene falls from the six-electron level
reduction of nitrobenzene on Ag [55] and to the four-electron level at potentials
Au [56] electrodes was studied in acidic more negative than the third UPD peak
0.0
[mA cm−2]
−0.1
j
2 34 5
−0.2
1
(a)
0.0
[mA cm−2]
−0.1
j
1
(b) −0.2
0.1
[mA cm−2]
0.0
j
−0.1
−0.2 0.0 0.2 0.4

E vs NHE
(c) [V]
Fig. 4 Steady state Current–potential curves for (a) Ph−NO2
(5 × 10−4 M) and (b) Ph-NO (5 × 10−4 M) reduction on Ag(111)
in 0.2 M LiClO4 + 7.5 × 10−3 M HClO4 in methanol without
Pb(ClO4 )2 (1) and with Pb(ClO4 )2 ; (2) 5 × 10−4 ; (3) 10−3 ;
(4) 2.5 × 10−3 ; and (5) 5 × 10−3 M. (c) Cyclic voltammogram for
the UPD of Pb on Ag(111) in the same electrolyte with
2.5 × 10−3 M Pb(ClO4 )2 (ν = 10 mV s−1 ). (Reproduced from
Ref. [55] with permission.)
at which complete ad-atom coverage is X-ray diffraction [15] and STM [20] stud-
obtained. In this potential region, lead ies. Also, the second reduction wave
forms on Ag(111) a rotated hexagonal of nitrosobenze, which corresponds to
close-packed structure as revealed by the reduction of phenylhydroxylamine to
aniline, is completely suppressed in the NO2

same potential region. PhCH + 2e− + 2H+
NO2
Reduction of aromatic nitro and nitroso com-

pounds on Pt/M(upd) In contrast to silver
and gold electrodes, interesting catalytic NOH
PhC + H2O
effects were found for the reduction of NO2
aromatic nitro compounds on Pt surfaces (13)
modified by Pb, Tl, and Bi UPD adlayers Reaction (12), producing phenylnitro-
in aqueous solutions [57–59]. On clean Pt methane, predominates on Pt/Tl(upd),
the reduction of nitro compounds occurs while on Pt/Pb(upd) the major reac-
in the potential region where hydrogen tion is the reduction to ω-nitrobenzald-
is evolved and no separate well-formed re- oxime.
duction waves are obtained. On Pt/M(upd) Also, in the case of the reduction of
the reduction begins at more positive po- benzofuroxan [59] the UPD adlayers of Pb,
tentials and diffusion-controlled limiting Tl, and Bi strongly modify the catalytic
currents are obtained. Generally, the catal- activity and selectivity of the platinum
ysis of the reduction of nitro compounds electrode. Figure 5 shows the rotating-
was attributed [57] to the change of the disc voltammograms for benzofuroxan
nitro group reduction mechanism from reduction on Pt and Pt/M(upd), M = Pb,
catalytic hydrogenation on clean Pt to di- Tl, Bi.
rect electron-transfer mechanism on Pt On bare platinum the reduction pro-
surfaces covered by the UPD adlayers, as ceeds via catalytic hydrogenation to
on metals with high hydrogen overvoltage o-nitroaniline (Scheme 2), which is re-
(Scheme 1). duced further to o-phenylenediamine. On
In the case of phenyldinitromethane [58] the other hand, on UPD-modified plat-
the UPD-modified surfaces exhibit also
inum surfaces, the reduction occurs at
remarkable electrocatalytic selectivity as
much more positive potentials and follows
regards the final reduction products under
the electronation mechanism, that is, the
electrolysis conditions. From the two
direct exchange of electrons between the
competitive reaction routes
dinitroso tautomeric form and the modi-
NO2 fied electrode surface. The first step is now
PhCH + 2e− + 2H+
NO2
the reduction to o-benzoquinone dioxime
that appears as a stable intermediate over
a wide potential range (0.60 to 0.45 V)
PhCH2NO2 + HNO2 before it is reduced further to the final
(12) products.
*
O
** 2Hads *** * mHads
(Pt): Ph-NO 2 + 4Pt Ph-N Ph-N-O Products
* −H2O r.d.s
O
2e− + 2H+ 2e− + 2H+ 2e− + 2H+

(Pt/Mads): Ph-NO 2 Ph-NO Ph-NHOH Ph-NH2
−H2O −H2O
Scheme 1 Reduction of nitrobenzene on Pt and Pt/Mads (M = Pb, Tl).

0.0
(a)
0
−1.5
(b)
0
1 2 3 4
−3.0
[mA cm−2]
[mA cm−2]
(c)
j
0
j
−4.5
(d)
0
0.2
−6.0
0.0 0.4 0.8 1.2
E vs NHE
[V]
−7.5
0.0 0.4 0.8
E vs NHE
[V]
Fig. 5 Current-potential curves for benzofuroxan (10−3 M) reduction on a
Pt rotating-disc electrode in 0.5 M HClO4 without (1) and with (2) 10−3 M
TlClO4 ; (3) 10−3 M Pb(ClO4 )2 ; (4) 10−3 M Bi(ClO4 )3 (ν = 10 mV s−1 ;
f = 25 Hz). Inset: Cyclic voltammograms for the UPD of Pb, Tl, and Bi on
Pt in 0.5 M HClO4 . (a) Without ions; (b) with 5 × 10−4 M Pb2+ ; (c) with
5 × 10−4 M Tl+ ; and (d) with 5 × 10−4 M Bi3+ . Scan rate ν = 50 mV s−1 .
(Reproduced from Ref. [59] with permission.)
Pt
O
O
N NO2
N
O* 2Hads
O + 2Pt
N
N* NH2
O*
NO2 **
N
O* mHads
+ 4Pt Products
NH2 NH2
Pt/M(upd)
O NOH
N NO
2e− + 2H+
O
N NO NOH
Scheme 2 Reduction of benzofuroxan on Pt and Pt/M(upd)
(M = Pb, Tl, Bi).
NO2
O2N NO2
N N N O2N
N
N NO2 N N N N
H H
H H O
Reduction of heterocyclic nitro compounds The behavior of these compounds differs

on Au/M(upd) Overlayers of Pb and from that of nitrobenzene. The strong
Tl ad-atoms, deposited on Au in the electron-attracting heterocyclic nuclei, on
underpotential region, exert pronounced the one hand, renders the nitro group more
catalytic effects on the reduction of easily reducible and, on the other hand,
some N -heterocyclic nitro compounds [56, helps to stabilize the hydrated form of the
60, 61]. nitroso intermediates.
0 160
(i)
−1 0
1 3
[mA cm−2]
n=2
[mA cm−2]
−2 2 −2
[µA]
2
iR
80
jL
j
3 1
−3 −4
1
2
4 n=4
0 10 20
−4 Au ER = 1.6 V
Au Pb w1/2
[s−1/2] 0
−5 ER = 0.5 V
−10
[µA]
iR
0.2 −20 1
(ii) −1 3
−30
[mA cm−2]
0.1 1
[mA cm−2]
−2 2
jD
0.0
j
3
−0.1 −3
−0.2
−4
0.0 0.4 0.8 0.0 0.4
E vs SHE ED vs SHE
(a) [V] (b) [V]
Fig. 6 (a) (i) Current-potential curves for the UPD of Pb on Au in 0.2 M
reduction of 3-nitro-1,2,4-triazole (10−3 M) on a HClO4 ·cPb 2+ = 5 × 10−4 M; ν = 50 mV s−1 .
Au disc electrode in 0.2 M HClO4 without (b) Au−Au rotating ring-disc (N = 0.18)
(dashed lines) and with (full lines) 5 × 10−4 M measurements for 3-nitro-1,2,4-triazole
Pb(ClO4 )2 (ν = 20 mV s−1 ). f/Hz: (1) 12.5; (10−3 M) in 0.2 M HClO4 + 5 × 10−4 M
(2) 25; (3) 50; and (4) 83.3. The inset shows the Pb(ClO4 )2 (ν = 20 mV s−1 ). f/Hz: (1) 12.5;
jL versus ω1/2 (ω = 2π f) plots at (1) −0.1 V and (2) 25; and (3) 50. (Reproduced from Ref. [56]
(2) +0.21 V. (ii) Cyclic voltammogram for the with permission.)
The effect of covering a gold disc with reaction, the limiting current exhibits ki-
Pb ad-atom layers on the reduction of netic character.
3-nitro-1,2,4-triazole is shown in Fig. 6(a). On Au partially covered by Pb ad-atoms
Nitrotriazole undergoes a four-electron re- (E = 0.4 to 0.2 V), an electrocatalytic
duction both on Au and Au/Pb(upd) [56]. mechanism appears to be operative. The
However, the reduction mechanism de- reduction proceeds through chemisorp-
pends on the UPD ad-atom coverage. tion of the nitro group and reductive
On bare Au and Au surfaces with al- cleavage of one of the two N−O bonds,
most complete ad-atom layer (E < 0.15 V), which leads directly to the hydroxylamino
the four-electron reduction follows the compound and gives diffusion-controlled
ECE mechanism (Scheme 3). The reaction four-electron limiting currents. As in the
proceeds through a N ,N -dihydroxyamine case of oxygen reduction, a bridge ad-
derivative (the hydrated form of the ni- sorption of the nitro group with a bridge
troso intermediate) that is reduced further complex involving a substrate (Au) and an
after an irreversible loss of water. Be- adsorbed atom (M = Pb, Tl) was assumed
cause of the slow rate of the dehydration to explain the strong chemisorption on Au
Au and Au/M(upd) ( high coverage)

OH OH
NO2 N OH N
N N N O H
− +
2e + 2H kfast kslow
N N N H − H2O
N N N
H H
NO NHOH
N N
2e− + 2H+
N N H
N N
Au/M(upd) (low coverage)

O Ar HO Ar
N O 2e− + 2H+ 2e− + 2H+

N OH
ArNHOH + H2O
r.d.s fast
O Ar
N O (M = Pb, Tl)
Au M
Scheme 3 Reduction mechanism of 3-nitro-1,2,4-triazole on Au and Au/M(upd)
(M = Pb, Tl).
partially covered by ad-atoms. Analogous first two fuels to CO2 requires the do-
behavior was also observed with the other nation of oxygen, which comes either
heterocyclic nitro compounds [60, 61]. from water or hydroxyl ions depending
The ring-disc measurements presented on the pH of the solution. In aqueous
in Fig. 6(b) confirm the above mecha- acid solutions the overall reaction may be
nisms. The final product of the reduc- written as
tion of nitrotriazole both at potentials of
diffusion-controlled and kinetically con- CHx Oy + (2 − y)H2 O −−−→
trolled limiting currents is hydroxylamino- CO2 + [2(2 − y) + x](H+ + e− ) (14)
triazole, which is collected by the ring
electrode at ER = 1.6 V. In the potential Since all hydrogen atoms must be ab-
region, where the current is kinetically stracted from the fuel, the electrocatalyst
controlled, the intermediate nitrosotria- must support the dehydrogenation. Elec-
zole was detected at ER = 0.5 V. In fact, trodes that can be used as efficient anodes
the nitroso compound detected on the ring in C1 electrocatalysis are those with great
electrode is not that formed on the disc, affinity toward dehydrogenation reactions,
because it is rapidly reduced further, but namely, Pt and some other metals of the
that produced from the dihydroxyamine Pt-Group. Unfortunately, the current den-
intermediate upon its dehydration across sities obtained with these electrodes are
the gap from disc to ring. Dihydroxyamine very low owing to poisoning effects. The
itself is not oxidized back to nitrotriazole main effect of poisoning is to block sites
on the ring, as one would expect, because on the electrode surface.
it is rapidly stabilized by intramolecular Many efforts have been made to identify
hydrogen bonds. the nature of the poisoning species. The
The reduction of the above N -hetero- most successful techniques are the in situ
cyclic nitro compounds on Au modified by IR reflectance spectroscopies. Intensive
ad-atoms may by used as model reactions work has been done both on polycrystalline
for studying the surface geometric and and single-crystal electrodes. The work
electronic structure in electrocatalysis. done has been reviewed by Bewick and
By using Au, Ag, or Pt single-crystal Pons [62], Beden and Lamy [63], and more
electrodes, it could be possible to correlate recently by Sun [64]. It is now recognized
the catalytic activity with the superlattice that CO (either linearly or bridge-bonded
structures of the ad-atoms. to the surface) is the poison and not formyl
species (COH), as believed previously [65,
4.3.4.3 Oxidation of Small Organic 66]. Carbon monoxide is formed by de-
Molecules hydration in the case of HCOOH and by
The electrooxidation of simple one- dehydrogenation in the case of HCHO and
carbon molecules (C1 molecules), such CH3 OH. The CO formation reaction con-
as methanol, formaldehyde, and formic stitutes the one branch of the so-called
acid, became the subject of considerable ‘‘dual mechanism,’’ originally suggested
and continuous interest. This is quite by Capon and Parsons for the electrochem-
reasonable since these compounds, par- ical oxidation of C1 molecules [65, 66]. The
ticularly methanol, are potential fuels for other branch is the reaction through reac-
fuel cells. The complete oxidation of the tive intermediates.
[P] the electrode to poisoning, the more im-

C1 molecules CO2 portant is the enhancement caused by
[RI] ad-atoms.
[P] is the poison and [RI] the reactive Formic acid electrocatalysis by ad-atoms
intermediates. All species involved in the The electrooxidation of formic acid on
mechanisms may be considered as reactive noble metals, such as Pt, Pd, Rh, Ir, is
intermediates, if they are not strongly a classical example of a self-poisoning
bonded to the surface and are oxidized reaction. The currents obtained are very
to CO2 . low and depend strongly on the potential
The importance of the oxidation of C1 sweep direction (anodic or cathodic) and
molecules in electrochemical energy con- the crystallographic orientation of the
version has prompted intense research electrode surface. The Pt(111) face with
efforts to find new more active electro- the hexagonal lattice structure was found
catalysts or to improve the activity of to be the surface that gives the higher
existing catalysts, for instance, platinum. currents and is less sensitive to poisoning
Efforts to enhance the catalytic proper- effects [69, 70]. Foreign metal ad-atoms
ties of metal electrodes toward C1 elec- cause striking catalytic effects on the
trooxidation include both bulk alloys and oxidation of formic acid on polycrystalline
pure metal surfaces modified by ad-atoms and single-crystal electrodes. Ad-atoms
(modifiers), usually deposited in the un- may be deposited either by UPD or
derpotential region. irreversibly by immersing the electrode
Generally, the ad-atoms cause positive into the metal solution at open circuit.
catalytic effects with significant enhance- The early work was reviewed by Adzic [26],
ment of the electrocatalytic activity of plat- while the most recent work by Jarvi and
inum in several cases. Several reviews have Stuve [29].
been published on this subject [26–28, Modifiers that can enhance the elec-
67]. Very recently Ross [68] and Jarvi and trocatalytic activity of Pt, Pd, Rh, and
Stuve [29] have discussed the more re- Ir electrodes are Pb, Tl, Bi, Cd (see
cent advances in our understanding of Ref. [26] and references therein). Among
the fundamentals of C1 electrocatalysis these modifiers, lead appears to be the
by ad-atoms. The different types of en- stronger promoter in formic acid electro-
hancement (third-body effect, bifunctional catalysis [26, 71–74]. Figure 7 shows the
mechanism, poison destabilization, and effect of the UPD of Pb on formic acid
electronic modification) are well docu- electrooxidation on Pt. As seen, very high
mented. The new information obtained currents are obtained on Pt/Pb(upd) in
from the in situ spectroscopic studies the range between 0.3 and 0.7 V. Another
about the nature of poisons and the de- important observation is the coincidence
pendence of their coverages on potential, of the currents during the positive and
as well as the use of single-crystal elec- the reverse negative potential scans, which
trodes with defined surface structure and is considered as an evidence of reduced
specific reactivity, enables a deeper in- poisoning effects.
sight in the electrocatalysis by ad-atoms. The more recent studies on formic acid
As a general rule, one can say that, ex- electrooxidation refer to Pt single-crystal
cept for methanol, the more susceptible electrodes modified by ad-atoms with
80
Pt
Pt/Pb
60
[mA cm−2]
40
j
20
0
0.0 0.4 0.8 1.2 1.6
E vs NHE
[V]
Fig. 7 Voltammetric curves for formic acid (0.265 M)
oxidation on a Pt electrode in 1 M HClO4 without (dashed
line) and with (full line) 10−3 M Pb2+ ions. Scan rate
ν = 50 mV s−1 . (Reproduced from Ref. [71] with permission.)
controlled coverage, which are adsorbed compared to that of massive Pd single-

irreversibly on the Pt surface. Ad-atoms crystal electrodes. Palladium overlayers
deposited in this way include Pd [75, 76], show higher resistivity against CO poi-
Pb [77], Sb [78], As [79], and Bi [78, 80]. soning than bare Pd, which was attributed
Palladium is the stronger promoter for to electronic effects.
formic acid oxidation. Indeed, adsorbed
Pd drastically enhances the electrocatalytic Formaldehyde electrocatalysis by ad-atoms
activity of the Pt(100) electrode toward Poisoning effects during the electroox-
formic acid oxidation (Fig. 8). High cur- idation of formaldehyde were mainly
rent densities are obtained at low potentials observed in acid solutions. Ad-atoms de-
(0.22 V vs. RHE), where bare Pt is not posited at underpotentials enhance the
active. The catalytic activity reaches a max- electrocatalytic activity of the Pt electrode
imum at θPd = 0.6 [75]. Multilayers of toward HCHO oxidation [81–83]. Accord-
Pd, epitaxially grown on Pt single-crystal ing to Motoo and Shibata [82], ad-atoms
electrodes, also exhibit enhanced catalytic that do not adsorb oxygen (e.g. Pb, Tl,
activity toward formic acid oxidation [76] Bi) have only a small influence resulting
20
0.1
[mA cm−2]
[mA cm−2]
j
0.0
0.5
E
j
10
−0.1 [VRHE]
0
(a) 0.1 0.5 0.9
20
0.2
[mA cm−2]
[mA cm−2]
j
10 0.0
j
0.5
E
−0.2 [VRHE]
0
0.1 0.5 0.9
E vs RHE
(b) [V]
Fig. 8 (a) Cyclic voltammograms of Pt(100) (dashed line) and
Pt(100)/Pdads (θ = 0.32) (full line) in 0.25 M HCOOH + 0.5 M
H2 SO4 . (b) Cyclic voltammograms of Pd (dashed line) and
Pt(100)/Pdads (θ = 1.06) (full line) in the same electrolyte. The
insets show blank voltammograms in 0.5 M H2 SO4 . (Reproduced
from Ref. [75] with permission.)
from geometric factors (third-body mecha- the potential region of Pt-oxide formation.
nism), while oxygen adsorbing ad-atoms Many efforts have been made to promote
(e.g. Ru, Ge, Sn, Sb) give rise to a the activity of Pt by adding a second metal,
greater enhancement. The ad-atoms fa- either as ad-atoms or by alloying it with Pt.
cilitate the donation of oxygen through These studies are extensive and overviews
the bifunctional mechanism, thus lower- of the literature may be found in Refs. [26,
ing the oxidation overpotential. 28, 29, 68].
Here, we will restrict ourselves to point-
Methanol electrocatalysis by ad-atoms The ing out that ad-atoms, such as Pb and
oxidation of methanol takes place at po- Bi, which are strong promoters for formic
tentials more positive than that needed acid electrocatalysis, only slightly increase
for formic acid. Even on platinum, which the activity of Pt for methanol oxida-
is the most active electrocatalyst, the ox- tion [83–86]. The most successful pro-
idation occurs at high overpotentials and moter for methanol electrocatalysis is Ru
significant currents are obtained only in and perhaps Sn, albeit not as ad-atoms.
Pt-Ru alloys are now recognized as the pronounced in alkaline solutions, in which
best electrocatalysts for methanol oxida- the currents on the modified surfaces are
tion exhibiting lower overpotentials and much higher compared to those obtained
long-term stability, both important factors on platinum without ad-atoms. It should
for direct methanol fuel cells. In a series of be noted that in alkaline media the
excellent papers, Gasteiger et al. [87–90] catalytic effect is observed only at very
have provided valuable information re- low concentrations (around 10−6 M) of
garding the reactivity of Pt−Ru alloys. The heavy metal ions in the solution. At higher
effect of Ru on the kinetics of CH3 OH concentrations, negative catalytic effects
electrooxidation was rationalized in terms are observed rather than positive catalysis.
of the bifunctional action and a series re- In general, the main products formed
action mechanism. The Pt sites act as cen- during the oxidation of primary and
ters for the adsorption/dehydrogenation secondary alcohols are the corresponding
of methanol-producing poison and car- aldehydes and ketones, while the main
bonaqueous intermediates, and Ru sites product of the oxidation of glucose is
act as centers for adsorbing oxygen- gluconic acid.
containing species at less positive potential In acid solutions the catalytic effect
than Pt. The oxygen species on Ru sites ox- caused by ad-atoms is characterized by
idize both CO and the reactive carbonaque- the shift of the onset of oxidation toward
ous intermediates by surface diffusion re- less positive potentials and an increase
actions. The surface composition of Ru al- of the peak current density by a factor
ters the balance between the reaction rates of 2 to 5, depending on the kind of ad-
of CO formation and oxidation with the 7 atom and the organic substance being
to 10% Ru surface being the optimum sur- oxidized. In alkaline solutions the catalytic
face composition for enhanced reactivity. effect is mainly recognized by an immense
increase of the oxidation current, especially
4.3.4.4 Oxidation of Bigger Organic with Pb and Bi ad-atoms. In alkaline
Molecules solutions the poisoning effect is not
Other organic molecules, whose oxidation so significant, since strongly adsorbed
can be catalyzed by ad-atoms, include pri- intermediates are not involved in the
mary and secondary alcohols [85, 91–93], oxidation mechanism.
diols and polyols [94–100], and differ- On the contrary, in acid solutions
ent monosaccharides [101–105]. Aldehy- strongly adsorbed species are supposed
des are oxidized at more positive potentials to be formed during the oxidation of al-
compared to the oxidation potentials of cohols and sugars. The catalytic effect
primary alcohols and their oxidation is was generally interpreted in terms of a
catalyzed only by oxygen-adsorbing ad- decrease of the electrode poisoning. The
atoms [106, 107]. nature of the poisoning species produced
Generally, submonolayers of ad-atoms during the dissociative chemisorption of
of heavy metals (such as Pb, Tl, Bi, aliphatic alcohols [108–111] and ethylene
Cd, etc.) deposited in the underpotential glycol [112] on Pt has been studied by
region increase the catalytic activity of Pt in situ reflectance spectroscopies. It was
toward the electrooxidation of alcohols and demonstrated that CO is the main poison-
sugars both in aqueous, acid, and alkaline ing species and that the coverage of the
solutions. The catalytic effect is more surface by adsorbed CO decreases with
the increase of the carbon chain. The for- on Pt produces CO, which should be the
mation of CO results from the dissociative major adsorbate responsible for poisoning
chemisorption and the C1 –C2 bond break- of the electrode surface.
ing. In the case of D-glucose oxidation on The structure of the organic molecules
Pt, the poisoning species is believed to be should also be considered in interpret-
a gluconolactone type intermediate [113, ing the catalytic effects caused by the
114], which is also considered as the reac- modifiers. This can be deduced from the
tive intermediate producing gluconic acid comparative studies (performed in acid so-
upon hydrolysis. lutions) of the UPD effect on the oxidation
of two different series of organic com-
OH O Pt
O −(2H+ + 2e−) O pounds with the same functional group,
C1 + 2Pt C1 Pt such as aliphatic primary alcohols [85] and
R R
monosaccharides [103]. The lack of uni-
form catalytic behavior (based on current
densities) leads to the conclusion that be-
O sides the third-body effect of the ad-atoms
C1 O + 2Pt that is physical in nature the specific inter-
R
(15a) acting forces of the reaction intermediates
with the modified electrode surface must
OH
also play a significant role in the electro-
O OH catalytic process.
+ 2Pt
HO Finally, ad-atom-modified electrodes
OH may also exhibit selectivity regarding the
OH
products obtained after electrolysis. Selec-
tive oxidation is particularly important in
OH
organic electrosynthesis. Two examples of
O O Pt selective oxidation of organic compounds
HO Pt will be discussed: the oxidation of gly-
OH
OH colaldehyde in acid solutions on Pt and
Pt/Sbads [117] and the oxidation of glu-
conic acid in alkaline solutions on Pt and
Pt/Pbads [118].
OH On smooth Pt the main product of gly-
O colaldehyde oxidation is glyoxal, while on
O + 2Pt Hydrolysis Pt/Sbads it is glycolic acid [117] (Scheme 4).
HO
OH OH The Sb ad-atoms increase the reaction
(15b) rate of both the oxidation of the CHO
However, more recent studies by in situ IR group to COOH group and the oxidation of
spectroscopy [115, 116] have demonstrated the CH2 OH group to CHO group, but the
that the catalytic decomposition of glucose enhancement of the first reaction is much
Pt
CHO–CHO (glyoxal)
Scheme 4 Selective oxidation
CH2OH–CHO
of glycolaldehyde on Pt and Pt/Sbads
Pt/Sbads . CH2OH–COOH (glycolic acid)
more significant than that of the second oxidation of some dienolic compounds,
one. Thus, the CHO group is selectively such as ascorbic acid [124, 125] and rh-
oxidized on the Pt electrode modified by odizonic acid [122]. Figure 9 presents an
the Sb ad-atoms. example of how the UPD of Tl affects the
In the case of D-gluconic acid [118], anal- cyclic voltammetric behavior of the sys-
ysis of the oxidation products by ionic chro- tem pyrocatechol/o-benzoquinone on the
matography has shown that on smooth Pt electrode.
Pt the main products are oxalic, tartaric, The improvement of the reversibility
5-keto-D-gluconic, and D-glucuronic acids. of the redox systems was attributed by
On Pt modified by Pb ad-atoms, D-gluconic Kokkinidis et al. [120, 122] to the change
acid is selectively oxidized to 2-keto-D- of the reaction mechanism. The reaction
gluconic acid, which is obtained with mechanism changes from an inner-sphere
relatively high yield (33%). The reaction mechanism involving adsorbed interme-
rate and the orientation of D-gluconic acid diates on bare Pt to an outer-sphere one
depend on the nature of the electrode. On without complications from adsorption
bare Pt the reaction rate is very low and of the reacting molecules on Pt/Mads
the oxidation occurs preferably at C6 yield- (M = Pb, Tl, Bi).
ing D-glucuronic acid. On Pt/Pbads the
rate of reaction is significantly increased 4.3.4.6 Inorganic Electrochemical
and the oxidation occurs at C2 produc- Reactions
ing 2-keto-D-gluconic acid. On Pt/Pbads Inorganic electrochemical reactions influ-
no D-glucuronic acid was obtained. This enced by ad-atoms include the oxidation
means that oxidation at C6 does not oc- of hydrazine and the reduction of some
cur at all, indicating that the presence of oxo-anions.
ad-atoms changes the way that the react- The influence of the UPD of heavy met-
ing molecules link and interact with the als on the oxidation of hydrazine on Pt was
electrode surface. studied in aqueous acid and alkaline so-
lutions [126]. Adlayers of Pb, Tl, Bi, Cu,
4.3.4.5 Reversible Redox Reactions Cd, Ag cause a complete inhibition of
It is known that some redox systems show the four-electron oxidation of hydrazine
poor reversibility on platinum. This is due to nitrogen. On the Pt-modified surfaces
to the fact that a monolayer of chemisorbed the oxidation commences at sufficiently
species from the redox components is positive potentials where appreciable dis-
formed that hinders the exchange of solution of ad-atoms occurs.
electrons between the rest of the diffusing On the contrary, UPD monolayers of
molecules and the electrode. Pb, Tl, and Bi markedly catalyze the re-
Underpotentially deposited ad-atom lay- duction of chromates (Cr2 O7 2− ) [127] and
ers of Pb, Tl, and Bi improve markedly persulfates (S2 O8 2− ) [127, 128] on Pt. The
the reversibility of the redox syst- E1/2 of the reduction waves shifts toward
ems: hydroquinone/p-benzoquinone [119, more positive potentials and diffusion-
120], pyrocatechol/o-benzoquinone [120], controlled limiting currents are attained
adrenaline/adrenalinequinone [121], hexa- within the UPD region. The catalytic
hydroxybenzene/tetrahydroxy-1,4-benzoq- activity was interpreted in terms of pre-
uinone [122], p-benzoquinone dioxime/ venting the covering of the Pt surface
p-dinitrosobenzene [123], as well as the by species resulting from the dissociative
3 5
3 2 4
[mA cm−2]
1
2 3
2
jp
1 2
0 1
1 0 20 10
√v
[mV s−1]
[mA cm−2]
0
j
−1
0.93 0.81
Pt
2
[V]
Ep
0.91 0.79 Pt/TI

−2 1
0.89 0.77
−1.4 −1.0 −0.6 −0.2
log (v )
[V s−1]
−3
0.2 0.6 1.0
E vs NHE
[V]
Fig. 9 Cyclic voltammograms for pyrocatechol (2.5 × 10−3 M) on Pt in
0.5 M HClO4 without (dashed lines) and with (full lines) 10−3 M TlClO4 .
Scan rate ν/mV s−1 : (1) 50; (2) 100; (3) 200; (4) 300; and (5) 400. The
insets show plots of jp versus ν 1/2 and Ep versus log ν without (1) and
with (2) TlClO4 . (Reproduced from Ref. [120] with permission.)
chemisorption of Cr2 O7 2− and S2 O8 2− inhibition effect was correlated to the su-

anions, thus leading to faster electron perlattice structures of the UPD ad-atoms
exchange for the charge-transfer step in- on the different Ag single-crystal surfaces.
volved in the reaction mechanisms. In the
case of dichromate reduction in acid so- 4.3.5
lutions, a remarkable discontinuity in the Conclusion
voltammetric behavior was observed when
the potential was scanned in the positive The influence of ad-atoms on reaction rates
direction. This behavior is related to the an- and mechanisms has attracted the interest
odic desorption of the UPD monolayers, of many researchers over the last 25 years.
which in the presence of Cr2 O7 2− ions oc- The most extensively studied reactions are
curs very abruptly owing to the competitive the reduction of oxygen and the oxidation
adsorption of dichromates. of organic fuels because of the importance
Finally, the reduction of NO3 − anions of these reactions in electrochemical en-
on Ag(hkl) electrodes is strongly inhibited ergy conversion. The challenge was to
by Pb and Tl ad-atom layers [129]. The improve the catalytic activity of Pt and
to develop new, more efficient electrocat- H. Gerischer, C. W. Tobias), Wiley, New

alysts for these reactions. On the other York, 1978, pp. 125–271, Vol. 11.
4. K. Jüttner, W. J. Lorenz, Z. Phys. Chem. N.F.
hand, studies on ad-atom-modified elec-
1980, 122, 163–185.
trodes have allowed us to understand how 5. S. Szabó, Int. Rev. Phys. Chem. 1991, 10,
surface geometric and electronic structure 207–248.
influences catalytic activity and reactivity. 6. J. W. Schultze, F. D. Koppitz, Electrochim.
Significant progress has been made to- Acta 1976, 21, 327–343.
ward this direction by using single-crystal 7. M. F. Toney, J. G. Gordon, M. G. Samant
et al., Langmuir 1991, 7, 796–802
electrodes with well-defined structures of 8. J. Clavilier, R. Faure, G. Guinet et al., J.
substrate and ad-atom layers in combi- Electroanal. Chem. 1980, 107, 205–209.
nation with in situ spectroscopies and 9. J. Clavilier, D. Armand, S. C. Sun et al., J.
scanning probes. These techniques have Electroanal. Chem. 1986, 205, 267–277.
provided information on reaction interme- 10. A. T. Hubbard, R. M. Ishikawa, J. Katekaru,
diates (reactive or poisons) and surface
11. P. N. Ross, J. Electrochem. Soc. 1979, 126,
properties under operating conditions that 67–77.
can help us to directly correlate the sur- 12. P. N. Ross, F. T. Wagner in Advances in Elec-
face structure and composition with the trochemistry and Electrochemical Engineering
catalytic activity. We already know enough (Eds.: H. Gerischer, C. W. Tobias), Wiley,
about oxygen reduction, but there is still New York, 1984, pp. 69–112, Vol. 13.
13. R. R. Adzic in Modern Aspects of Electrochem-
much to learn about the oxidation of or- istry (Eds.: R. E. White, J. O. ’M. Bockris,
ganic fuels. B. E. Conway), Plenum Press, New York,
Except the enhancement of the elec- 1990, pp. 163–236, Vol. 21.
trocatalytic activity, ad-atoms may also 14. R. Parsons, G. Ritzoulis, J. Electroanal.
improve the selectivity of the surface of Chem. 1991, 318, 1–24.
15. M. G. Samant, M. F. Toney, G. L. Borges
the electrocatalysts regarding the final re-
et al., J. Phys. Chem. 1988, 92, 220–225.
action products. Little work has been done 16. M. F. Toney, J. G. Gordon, M. G. Samant
in this direction. Probably in the future, et al., Phys. Rev. 1992, B45, 9362–9374.
electrodes modified by ad-atoms will be 17. J. Wang, B. M. Ocko, A. J. Davenport et al.,
more attractive as potential catalysts in se- Phys. Rev. 1992, B46, 10 321–10 338.
18. M. P. Green, K. J. Hanson, D. A. Scherson
lective electrocatalysis. The improvement
et al., J. Phys. Chem. 1989, 93, 2184–2187.
of the selectivity would require thorough 19. M. Binggeli, D. Carnal, R. Nyffenegger
mechanistic studies in order to get a et al., J. Vac. Sci. Technol. 1991, B9,
better understanding of the various fac- 1985–1992.
tors that may influence the selectivity of 20. U. Müller, D. Carnal, H. Siegenthaler et al.,
these electrodes. Phys. Rev. 1992, B46, 12 899–12 901.
21. X. Gao, A. Hamelin, M. J. Weaver, Phys.
Rev. 1992, B46, 7096–7102.
References 22. K. Sashikata, N. Furuya, K. Itaya, J. Elec-
troanal. Chem. 1991, 316, 361–368.
1. J. O.’M. Bockris, A. K. N. Reddy, Modern 23. D. M. Kolb, Electrochim. Acta 2000, 45,
Electrochemistry, Plenum Press, New York, 2387–2402.
1977, Vol. 2. 24. C.-H. Chen, A. A. Gerwirth, J. Am. Chem.
2. W. J. Lorenz, H. D. Herrman, N. Wüthrich Soc. 1992, 114, 5439, 5440.
et al., J. Electrochem. Soc. 1974, 121, 25. X. Xia, T. Iwasita, J. Electrochem. Soc. 1993,
1167–1177. 140, 2559–2565.
3. D. M. Kolb in Advances in Electrochem- 26. R. R. Adzic in Advances in Electrochem-
istry and Electrochemical Engineering (Eds.: istry and Electrochemical Engineering (Eds.:
H. Gerischer, C. W. Tobias), Wiley, New 48. N. M. Markovic, R. R. Adzic, B. Cahan

York, 1984, pp. 159–260, Vol. 13. et al., J. Electroanal. Chem. 1994, 377,
27. G. Kokkinidis, J. Electroanal. Chem. 1986, 249–259.
201, 217–236. 49. N. M. Markovic, H. A. Gasteiger, P. N.
28. R. Parsons, T. VanderNoot, J. Electroanal. Ross, J. Phys. Chem. 1995, 99, 3411–3415.
Chem. 1988, 257, 9–45. 50. R. R. Adzic, J. Wang, B. M. Ocko, Elec-
29. T. D. Jarvi, E. M. Stuve in Electrocatalysis trochim. Acta 1995, 40, 83–89.
(Eds.: J. Lipkowski, P. N. Ross), Wiley, New 51. R. R. Adzic, J. X. Wang, Electrochim. Acta
York, 1998, pp. 75–153. 2000, 45, 4203–4210.
30. N. D. Lang, S. Holloway, J. K. Norskov, 52. W. Kemula, T. M. Krygowski in Encyclope-
Surf. Sci. 1985, 150, 24–28. dia of Electrochemistry of the Elements (Eds.:
31. M. P. Kiskinova, Stud. Surf. Sci. Catal. 1992,
A. J. Bard, H. Lund), Marcel Dekker, New
70, 1–137.
York, 1979, pp. 77–208, Vol. 13.
32. C. G. Vayenas, I. V. Yentekakis in Hand-
53. H. Lund in Organic Electrochemistry (Eds.:
book of Heterogeneous Catalysis (Eds.: G. Ertl,
M. M. Baizer, H. Lund), 3rd ed., Marcel
H. Knözinger, J. Weitkamp), Wiley-VCH,
Weinheim, New York, 1997, Dekker, New York, 1991, pp. 401–432.
pp. 1310–1325, Vol. 3. 54. A. Cyr, P. Huot, J.-F. Marcoux et al., Elec-
33. A. Damjanovic in Modern Aspects of Elec- trochim. Acta 1989, 34, 439–445.
trochemistry (Eds.: J. O.’M. Bockris, B. E. 55. G. Kokkinidis, K. Jüttner, Electrochim. Acta
Conway), Plenum Press, New York, 1969, 1981, 26, 971–977.
pp. 369–483, Vol. 5. 56. G. Kokkinidis, A. Papoutsis, G. Papanasta-
34. M. R. Tarasevich, A. Sadkowski, E. Yeager siou, J. Electroanal. Chem. 1993, 359,
in Comprehensive Treatise of Electrochem- 253–271.
istry (Eds.: B. Conway, J. O.’M. Bockris, 57. G. Kokkinidis, P. Jannakoudakis, Electro-
E. Yeager et al.), Plenum Press, New York, chim. Acta 1984, 29, 821–828.
1983, pp. 301–332, Vol. 7. 58. G. Kokkinidis, E. Coutouli-Argyropoulou,
35. K. Kinoshita, Electrochemical Oxygen Tech- Electrochim. Acta 1985, 30, 493–499.
nology, Wiley, New York, 1990. 59. G. Kokkinidis, N. Argyropoulos, Electro-
36. A. J. Appleby, J. Electroanal. Chem. 1993, chim. Acta 1985, 30, 1611–1620.
357, 117–179. 60. A. Papoutsis, G. Kokkinidis, J. Electroanal.
37. R. Adzic in Electrocatalysis (Eds.: J. Lip- Chem. 1994, 371, 231–239.
kowski, P. N. Ross), Wiley, New York, 1998, 61. G. Kokkinidis, A. Kelaidopoulou, J. Elec-
pp. 197–242. troanal. Chem. 1996, 414, 197–208.
38. E. Yeager, Electrochim. Acta 1984, 29, 62. A. Bewick, B. S. Pons in Advances in In-
1527–1537. frared and Raman Spectroscopy (Eds.:
39. P. Fischer, J. Heitbaum, J. Electroanal. R. J. H. Clark, R. E. Hester), Wiley Heyden,
Chem. 1980, 112, 231–238. London, 1985, pp. 1–63, Vol. XII.
40. N. Anastasijevic, Z. M. Dimitrijevic, R. R.
63. B. Beden, C. Lamy in Spectroelectrochemis-
Adzic, Electrochim. Acta 1992, 37, 457–464.
try-Theory and Practice (Ed.: R. G. Gale),
41. G. Kokkinidis, D. Jannakoudakis, J. Elec-
Plenum Press, New York, 1988, pp.
troanal. Chem. 1984, 162, 163–173.
189–261, Chap. 5.
42. T. Abe, G. M. Swain, K. Sachikata et al., J.
Electroanal. Chem. 1995, 382, 73–83. 64. S.-G. Sun in Electrocatalysis (Eds.: J. Lipkow-
43. A. Swetanova, K. Jüttner, J. Electroanal. ski, P. N. Ross), Wiley, New York, 1998,
Chem. 1981, 119, 149–164. pp. 243–290.
44. R. Amadelli, N. Markovic, R. Adzic et al., J. 65. A. Capon, R. Parsons, J. Electroanal. Chem.
Electroanal. Chem. 1983, 159, 391–412. 1973, 44, 1–7.
45. S. A. S. Machado, A. A. Tanaka, E. R. Gonza- 66. A. Capon, R. Parsons, J. Electroanal. Chem.
lez, Electrochim. Acta 1994, 39, 2591–2597. 1973, 45, 205–231.
46. K. Jüttner, Electrochim. Acta 1984, 29, 67. B. Beden, J.-M. Léger, C. Lamy in Modern
1597–1604. Aspects of Electrochemistry (Eds.: J. O.’M.
47. F. El Kadiri, R. Faure, R. Durand, J. Elec- Bockris, B. E. Conway, R. E. White), Plenum
troanal. Chem. 1991, 301, 177–188. Press, New York, 1992, pp. 97–264, Vol. 22.
68. P. N. Ross in Electrocatalysis (Eds.: J. Lip- 92. D. Sazou, N. Xonoglou, G. Kokkinidis, Col-
kowski, P. N. Ross), Wiley, New York, 1998, lect. Czech. Chem. Commun. 1986, 51,
pp. 43–74. 2444–2454.
69. J. Clavilier, R. Parsons, R. Durand et al., J. 93. C. T. Hable, M. S. Wrighton, Langmuir
Electroanal. Chem. 1981, 124, 321–326. 1993, 9, 3284–3290.
70. R. R. Adzic, A. V. Tripkovich, W. O’Grady, 94. G. Kokkinidis, D. Jannakoudakis, J. Elec-
Nature 1982, 296, 137, 138. troanal. Chem. 1982, 133, 307–315.
71. R. R. Adzic, D. N. Simic, A. R. Despic et al., 95. F. Kadirgan, B. Beden, C. Lamy, J. Elec-
J. Electroanal. Chem. 1975, 65, 587–601. troanal. Chem. 1982, 136, 119–138.
72. R. R. Adzic, D. N. Simic, A. R. Despic et al., 96. F. Kadirgan, B. Beden, C. Lamy, J. Elec-
J. Electroanal. Chem. 1977, 80, 81–99. troanal. Chem. 1983, 143, 135–152.
97. M. Shibata, N. Furuya, M. Watanabe, J.
73. D. Pletcher, V. Solis, J. Electroanal. Chem.
1982, 131, 309–323.
98. G. Pierre, A. Ziode, M. L. Kordi, Electrochim
74. M. Shibata, S. Motoo, J. Electroanal. Chem.
Acta 1987, 32, 601–606.
1985, 188, 111–120.
99. P. Ocon, B. Beden, H. Huser et al., Elec-
75. M. J. Llorca, J. M. Feliu, A. Aldaz et al., J. trochim. Acta 1987, 32, 387–394.
Electroanal. Chem. 1994, 376, 151–160. 100. P. Ocon, B. Beden, C. Lamy, Electrochim.
76. M. Baldauf, D. M. Kolb, J. Phys. Chem. 1996, Acta 1987, 32, 1095–1101.
100, 11 375–11 381. 101. M. Sakamoto, K. Takamura, Bioelectrochem.
77. S. A. Campbell, R. Parsons, J. Chem. Soc., Bioenerg. 1982, 9, 571–582.
Faraday Trans. 1992, 88, 833–841. 102. N. Xonoglou, G. Kokkinidis, Bioelectrochem.
78. E. Herrero, J. M. Feliu, A. Aldaz, J. Elec- Bioenerg. 1984, 12, 485–498.
troanal. Chem. 1994, 368, 101–108. 103. G. Kokkinidis, N. Xonoglou, Bioelectrochem.
79. A. Fernandez-Vega, J. M. Feliu, A. Aldaz Bioenerg. 1985, 14, 375–387.
et al., J. Electroanal. Chem. 1991, 305, 104. N. Xonoglou, I. Moumtzis, G. Kokkinidis,
80. E. Herrero, A. Fernandez-Vega, J. M. Feliu 105. G. Kokkinidis, J.-M. Léger, C. Lamy, J. Elec-
et al., J. Electroanal. Chem. 1993, 350, 73–88. troanal. Chem. 1988, 242, 221–242.
81. D. M. Spasojevic, R. R. Adzic, A. R. Despic, 106. M. Shibata, S. Motoo, J. Electroanal. Chem.
J. Electroanal. Chem. 1980, 109, 261–269. 1985, 187, 151–159.
82. S. Motoo, M. Shibata, J. Electroanal. Chem. 107. M. Shibata, S. Motoo, J. Electroanal. Chem.
1982, 139, 119–130. 1986, 201, 23–32.
83. M. Beltowska-Brzezinska, J. Heitbaum, 108. B. Beden, M. C. Morin, F. Hahn, C. Lamy,
W. Vielstich, Electrochim. Acta 1985, 30, J. Electroanal. Chem. 1987, 229, 353–366.
1465–1471. 109. H. Hitmi, E. M. Belgsir, J.-M. Léger et al.,
84. B. Beden, F. Kadirgan, C. Lamy et al., J. Electrochim. Acta 1994, 39, 407–415.
Electroanal. Chem. 1981, 127, 75–85. 110. S.-G. Sun, D.-F. Yang, Z.-W. Tian, J. Elec-
troanal. Chem. 1990, 289, 177–187.
85. G. Kokkinidis, D. Jannakoudakis, J. Elec-
111. E. Pastor, S. Wasmus, T. Iwasita et al., J.
troanal. Chem. 1983, 153, 185–200.
86. M. Shibata, S. Motoo, J. Electroanal. Chem.
112. F. Hahn, B. Beden, F. Kadirgan et al., J.
1987, 229, 385–394.
87. H. A. Gasteiger, N. Markovic, P. N. Ross 113. S. Ernst, J. Heitbaum, C. H. Hamann, Ber.
et al., J. Phys. Chem. 1993, 97, 12 020–12 029. Bunsen-Ges. Phys. Chem. 1980, 84, 50–55.
88. H. A. Gasteiger, N. Markovic, P. N. 114. Yu. B. Vassiliev, O. A. Khazova, N. N. Niko-
Ross et al., J. Electrochem. Soc. 1994, 141, laeva, J. Electroanal. Chem. 1985, 196,
1795–1803. 105–125, 127–144.
89. H. A. Gasteiger, N. Markovic, P. N. Ross 115. T. T. Bae, E. Yeager, J. Electroanal. Chem.
et al., Electrochim. Acta 1994, 39, 1825–1832. 1991, 309, 131–145.
90. N. M. Markovic, H. A. Gasteiger, P. N. Ross 116. F. Largeaud, K. B. Kokoh, B. Beden et al., J.
et al., Electrochim. Acta 1995, 40, 91–98. Electroanal. Chem. 1995, 397, 261–269.
91. G. Kokkinidis, A. Papoutsis, J. Electroanal. 117. M. Shibata, N. Furuya, M. Watanabe, J.
Chem. 1989, 271, 233–247. Electroanal. Chem. 1993, 344, 389–393.
118. K. B. Kokoh, P. Parrot, E. M. Belgsir et al., 124. K. Takamura, M. Sakamoto, J. Electroanal.

Electrochim Acta 1993, 38, 1359–1365. Chem. 1980, 113, 273–283.
119. R. Wetzel, L. Müller, P. Göttlich, Z. Phys. 125. M. J. Walls, J. M. Feliu, A. Aldaz et al., J.
Chem. 1977, 258, 528–532. Electroanal. Chem. 1989, 260, 237–244.
120. G. Kokkinidis, J. Electroanal. Chem. 1984, 126. G. Kokkinidis. P. Jannakoudakis, J. Elec-
172, 265–279. troanal. Chem. 1981, 130, 153–162.
121. M. Sakamoto, K. Takamura, Bioelectrochem. 127. G. Kokkinidis, D. Sazou, J. Electroanal.
Bioenerg. 1983, 10, 251–260. Chem. 1987, 237, 137–144.
122. G. Kokkinidis, D. Sazou, I. Moumtzis, J. 128. G. Kokkinidis, D. G. Zatkas, D. Sazou, J.
Electroanal. Chem. 1986, 213, 135–147. Electroanal. Chem. 1988, 256, 137–148.
123. G. Kokkinidis, G. Papanastasiou, J. Elec- 129. C. Mayer, K. Jüttner, W. J. Lorenz, J. Appl.
troanal. Chem. 1988, 257, 239–255. Electrochem. 1979, 9, 161–169.
4.4 can be tuned for specific applications by

Electrochemistry in Micelles and adjusting the relative amounts of the com-
Microemulsions ponents [1–3].
The aim of this chapter is to de-
James F. Rusling
scribe electrochemistry in micelles and
Departments of Chemistry and Pharmacology
microemulsions in fundamental and prac-
University of Connecticut, Storrs, CT, USA
tical terms. A major focus is on the use of
4.4.1 these media to purposely influence the
Introduction desired outcome of electrochemical re-
actions. The chapter also describes how
Analytical and synthetic electrochemistry electrochemical methods can be used for
is usually done in homogeneous aque- structural characterization of these fluids.
ous solutions or organic solvents. How- In the next section (Sect. 4.4.2), we discuss
ever, there are advantages in some cases structures, properties, and dynamics of mi-
in using microheterogeneous fluids such celles and microemulsions. Subsequent
as micellar solutions and microemul- sections present relevant aspects of di-
sions [1–3]. These include solubilization rect electrochemistry and electrochemical
of organics and biomolecules, manip- catalysis in micelles and microemulsions.
ulation of electrode surface properties,
stabilization or destabilization of reactive 4.4.2
intermediates [4, 5], control of reaction Surfactant-stabilized Microheterogeneous
pathways, and enhancement of rates of Fluids
mediated reactions [6, 7]. Advantages also
accrue from the low cost [8] and low toxicity 4.4.2.1Structures of Micelles and
of micellar solutions and microemulsions Microemulsions
compared to pure aprotic solvents often
used for organic electrochemistry. Elec- 4.4.2.1.1 Surfactants and Micellar Structure
trochemical methods can also be used to The formation of micellar aggregates from
characterize micelles and microemulsions ionic surfactants in water causes sharp
via the measurement of diffusion coeffi- discontinuities in conductivity and sur-
cients of electroactive probes [4, 5, 9]. face tension. Water-soluble surfactants
Micelles and microemulsions depend form spherical or globular micelles at
on surfactants for their internal structure. concentrations near the CMC (Fig. 2).
Surfactants (surface-active agents or de- Micelle formation is thermodynamically
tergents) are amphiphilic molecules with favored in water for water-soluble sur-
charged or polar head groups attached to factants such as cetyltrimethylammonium
one or more hydrocarbon tails (Fig. 1). Mi- bromide (CTAB), sodium dodecyl sul-
celles are aggregates that form in water fate (SDS), and Brij 35, and is driven
when concentrations of surfactants exceed by hydrophobic interactions between the
a critical micelle concentration (CMC). Mi- tails [10]. The charged or polar head groups
croemulsions are microheterogeneous flu- face the water phase and the hydrocarbon
ids made from oil, water, and surfactants, tails reside in the interior of the micelle.
and sometimes cosurfactants. They appear For ionic surfactants in water, the head
homogeneous to the eye but contain dy- group region is only partly neutralized
namic heterogeneous nanostructures that by counterions, setting up an interfacial
4.4 Electrochemistry in Micelles and Microemulsions 419
CH3 Br−
+
N
CH3
DDAB-didodecyldimethylammonium bromide
O− O
P
O O
DHP-dihexadecylphosphate
CH3
Br−
+
CH3 N
CH3
CTAB-cetyltrimethylammonium bromide
−
Na+ SO4
SDS-sodium dodecylsulfate
CH2 OOC
CH OOC
–
CH2 O3POCH2CH2N+(CH3)3
PC-phosphatidylcholine (lecithin)
Br−
N
+
Cetylpyridinium bromide
HO(CH2CH2O)23
Brij-35 polyoxyethylene(23)dodecyl ether
Fig. 1 Amphiphilic molecules (surfactants) used to make
micellar solutions and microemulsions.
potential on the order of 100 mV [3]. This larger. Micelles having a single narrow
interfacial region consists of an electri- size distribution are called monodisperse.
cal double layer (i.e. head groups and Hydrophobic cores of globular micelles
counterions) called the Stern layer and typically have diameters of 1 to 3 nm.
provides binding sites for solutes of the Micellar core volume (Vc ) can be esti-
opposite charge. If salt is added to a mi- mated from the average number of sur-
cellar solution, the surface potential is factant molecules per micelle, that is, the
partly neutralized and the micelle grows aggregation number (Nag , ca. 30–200),
Water
Water
Water
Water
Globular micelle
Oil
Oil
Water
Oil
Oil Rod-shaped micelle
Inverse micelle
Fig. 2 Oversimplified representations of micellar structures.
Surfactants are represented with dark circles for head groups and
curved lines for hydrocarbon tails.
via [10] parameter, v/a0 lc , can be used to predict

the shape of the aggregate formed [11]. To
Vc = m (27.4 + 26.9n )Å (1) a first approximation, a0 is the area of
the head group, and the other terms are
for a single chain surfactant m = Nag for found from
a double chain surfactant m = 2Nag and
n is the number of carbon atoms in the lc = 1.5 + 1.26nÅ (2)
tail less one. Solubilization of nonpolar
v = 27.4 + 26.9n Å (3)
molecules occurs in hydrophobic regions
of the micelles but near the Stern layer [3]. where n is the number of carbons in
At intermediate concentrations consid- the tail of length lc , and n = n − 1. For
erably above the CMC, rod-shaped micelles a double chain molecule, n is multi-
often form. At very high concentrations, plied by 2. Values of v/a0 lc < 1/3 suggest
lamellar aggregates and other condensed spherical micelles, 1/3 > v/a0 lc > 1/2 pre-
phases form. In organic solvents, some dicts rod-shaped micelles, 1/2 > v/a0 lc >
surfactants form spherical inverse micelles 1 suggests lamellar structures including bi-
with the head groups facing small pools layers, and v/ao lc > 1 implies that reverse
of trapped water, with hydrocarbon tails micelles will form in organic solvents.
facing the bulk solvent phase (Fig. 2). The packing parameter is influenced to
Structures of micelles depend on the some extent by added salt, which decreases
structures and concentrations of the sur- average head group area, and by other ex-
factant molecules. The surfactant packing perimental variables [12].
Water
Water
Oil
Oil
Water Oil
Water Water
o/w microemulsion
Oil
Oil
Oil
Water Bicontinuous microemulsion
Oil
Oil
w/o microemulsion
Fig. 3 Oversimplified representations of microemulsion structures.
4.4.2.1.2 Microemulsion Structure Sur- leads to interfacial potentials on the order

factant molecules residing at oil–water of 100 mV.
interfaces lower the interfacial free energy A third type of microemulsion does
to nearly zero and drive the formation of not contain droplets but features contin-
optically clear, thermodynamically stable uous oil and water phases intertwined in
microemulsions [1–5]. Charged or polar dynamic extended networks (Fig. 3). Con-
head groups of the surfactants (Fig. 1) face ductivity is imparted to microemulsions
the water phase and hydrocarbon tails ex- by ionic surfactants or salt in a continuous
tend into the oil phase. Microemulsions water phase. Thus, o/w and bicontinuous
can contain droplets of oil in a contin- microemulsions can be made in conduc-
uous water phase (oil-in-water or o/w tive forms suitable for electrochemical
microemulsions) or droplets of water in experiments.
a continuous oil phase (water-in-oil or w/o The surfactant packing parameter v/a0 lc
microemulsions). These droplets have di- can be used to predict what types of
ameters in the range 10 to 100 nm, larger nanostructures in microemulsions can
than most micelles but much smaller form using specific surfactants. Am-
than droplets in emulsions. As with mi- phiphilic molecules with 0.5 < v/a0 lc < 1
celles, the surface charge on the droplets tend to form three-component o/w and
bicontinuous microemulsions. Surfactants they are bound to micelles, microemulsion

with v/a0 lc < 0.5 usually require cosurfac- droplets, or interfacial surfactant layers
tants such as long-chain alcohols to form in bicontinuous microemulsions. Binding
microemulsions and those with v/a0 lc > 1 can occur by coulombic or hydrophobic
tend to form w/o microemulsions. These interactions. An understanding of the in-
rules provide guidelines, but the type of fluence of solute binding on electrochem-
microemulsion obtained also depends on ical experiments relies on the knowledge
ratios of oil, water, cosurfactant, and sur- that micelles and microemulsion nanos-
factant [2]. tructures exist in mobile equilibria with
solutes and surfactant monomers (Fig. 4).
4.4.2.2 Dynamics Ejection of a monomer or solute from glob-
A major influence on the mass trans- ular aggregates occurs on the microsec-
port of electroactive solutes ensues when ond timescale, and recapture is diffusion
Solute
Solute
Solute
+
Fig. 4 Dynamic equilibria involving micelles, solutes and surfactant monomers.

controlled, that is, on the nanosecond bilayers, other regions of the surface begin
timescale. Recapture rates do not depend filling in.
strongly on solute structure, but exit rate Atomic force microscopy (AFM) [19–21]
constants range from 103 to 107 s−1 , and provides a more detailed molecular pic-
decrease with the size and solubility of the ture of adsorbates at concentrations above
solute [3, 4]. the CMC. Adsorbed cationic surfac-
Multiply charged ions have a strong tant C14 TAB (tetradecyltrimethylammo-
coulombic attraction to oppositely charged nium bromide) was arranged in surface
micelles; for example, Ru(bpy)3 2+ [bpy = micelles on silica and in long cylin-
2, 2 -bipyridine] binds strongly to SDS ders on mica. Double-chained DDAB was
micelles. Such solutes reside on the mi- organized in bilayers on mica. These
celle surface, although the ligands can structures result from the balance of adsor-
contribute hydrophobic interactions to bate–surface interactions and aggregate
the binding. Nonpolar solutes like aro- curvature controlled by intermolecular in-
matic hydrocarbons reside in hydropho- teractions.
bic regions close to the Stern layer of Ionic surfactants above the CMC form
the micelles. Spectroscopic studies show full coverage aggregates on metal and
that their microenvironments are more carbon electrodes [4, 22, 23]. Multilayers
polar in ionic micelles than in hy- may form at extreme potentials of the
drocarbon oils but less polar than in opposite sign of the surfactant head group.
water [13]. Developing a definitive molecular picture
of the aggregate structure based only
4.4.2.3 Surfactant Aggregates on Solid on electrochemical results is difficult.
Surfaces Spectroscopy of the electrode–adsorbate
Adsorption of surfactants on charged solution interface has helped elucidate
metal oxides is understood better than on structural features [4, 14].
solid electrodes and instructive parallels A few general features of the surfac-
can be drawn [14]. For ionic surfactants tant adsorbate structure can be extracted
of opposite charge to the metal oxide from the information available. At po-
surface, the first stage of adsorption tentials negative of the point of zero
from very dilute aqueous solutions is charge (PZC) on hydrophilic metal and
thought to result from neutralization of carbon electrodes, cationic surfactants ad-
the surface charge [15–17]. Well below sorb head down. Positive of the PZC,
the CMC, surfactants adsorb individually anionic surfactants adsorb head down.
with little intermolecular interaction. As Nonionic surfactants may adsorb head
the concentration increases, adsorption down on hydrophilic electrodes on either
increases sharply and surface aggregates side of the PZC. Adsorption of cationic
begin to form. It was suggested that at surfactants may be head down even at
this point the surface is characterized potentials positive of the PZC. This may in-
by patches of bilayers [15, 18]. In the volve adsorbed anions such as chloride on
inner adsorbate layer, head groups face the electrode. Surface aggregate structures
the solid surface. The outer layer of the above the CMC may include bilayers, sur-
surfactant has its charged head groups face micelles, or cylinders depending on
facing the water phase. Once the high the nature of the surfactant, the electrode
charge density patches are covered with surface, and the applied potential.
4.4.3 as obstruction effects. D for water-

Electrochemistry in Micellar and soluble, unbound electroactive ions can
Microemulsion Systems be interpreted in terms of obstruction by
oil droplets in o/w microemulsions [5, 9]
4.4.3.1 Mass Transport – Diffusion as in Eq. (7):
Mass transport rates of electroactive so-
lutes can be affected greatly by micelles D = D0 (1 − φc )p+1 (7)
and microemulsions. In cyclic voltamme-
try, for the reaction of a reversibly reduced
where φc is the volume fraction of the oil
or oxidized solute (O or R),
in the microemulsion, D0 is the diffusion
coefficient of the free ion in water, and p is
O + ne− = R (4)
a fitted coefficient. D values of unbound
the peak current (Ip ) in a quiet solution water-soluble ions in microemulsions are
depends on the apparent diffusion coeffi- typically 40 to 80% of those in water.
cient D as given by the Randles–Sevcik Diffusion of electroactive species residing
equation [24]: in a single microphase of bicontinuous
microemulsions is somewhat obstructed
1/2 by the o/w network.
nF
Ip = 0.4463nFAC ∗ ν 1/2 D 1/2 Solutes bound to micelles, droplets, or
RT
(5) interfacial regions can have D values ≥
where n is the number of electrons tenfold smaller than their D values in the
transferred per molecule, F is Faraday’s corresponding homogeneous solvent. For
constant, A is the electrode area, C* is the aggregates with a single size distribution,
solute concentration, R is the gas constant, where the equilibrium between the bound
T is the temperature in Kelvin, and ν is and the free solute is fast with respect
the scan rate. Equation (5) predicts a linear to the timescale of the electrochemical
plot of Ip versus ν for a reversible reduction experiment, D is given by a two-state
or oxidation. The slope of this plot can be model [4, 5]:
used to estimate D in a micellar solution
or microemulsion. D = fa D 0 + fb D 1 (8)
Linear sweep voltammetry at ultramicro-
electrode disks of radius r < 10 µm under where fa is the mole fraction of the
mass transport control, usually achieved unbound solute, fb is the fraction of
at scan rates <50 mV s−1 , provides a lim- the bound solute, D0 is the diffusion
iting current IL that depends directly on coefficient of the free solute, and D1
D [25]: is that of the micelle or droplet in the
IL = 4nF D C ∗ r (6) fluid. Equation (8) represents the usual
case in micelles and microemulsions
A number of other electrochemical meth- when D is estimated by cyclic, rotating
ods can also be employed to ob- disk, or ultramicroelectrode voltammetry
tain D [4, 9]. at relatively low scan rates [4].
D is influenced by binding to micelles, D1 can be used to estimate the
microemulsion droplets, or surfactant hydrodynamic radius (rh ) of a mi-
monolayers at o/w interfaces, as well celle or microemulsion droplet via the
Stokes–Einstein equation: reaction:
D1 =
kT
(9) M + mX = MXm (10)
6πηrh [MXm ]
Km = (11)
where k is Boltzmann’s constant, η is [M][X]m
the viscosity, and T is the temperature
Assuming [X] CX , that is, that the probe
in Kelvin. D1 can be obtained by using
is almost entirely bound, Eq. (8) can be
equilibrium expansions of Eq. (7) describ-
expressed as
ing D versus total concentration of an
electroactive probe, CX [4, 5]. At typical D0 CM K m CXm−1 D1
surfactant and probe concentrations, more D = +
than one electroactive probe (X) can bind 1 + CM K m CXm−1 1 + CM K m CXm−1
to a micelle or droplet (M) in the overall (12)
10
7
[cm 2 s−1]
107D ′
2
0 0.5 1 1.5 2 2.5 3 3.5
Fc
[mM]
Fig. 5 Influence of the concentration of ferrocene (Fc) on D measured by cyclic voltammetry
at a glassy carbon electrode in 0.15 M CTAB/0.1 M tetraethylammonium bromide. Points are
experimental: line is best fit to Eq. (12) with n = 3 by nonlinear regression, giving
D1 = 3.3 × 10−7 cm2 s−1 for the rod-shaped micelles. (Reproduced with permission from
Ref. [4], Copyright 1994 by Marcel Dekker.)
12
[cm2 s−1]
106D ′
0.00 0.50 1.00 1.50

B12
[mM]
Fig. 6 Influence of concentration of cobalt(II) corrin complex vitamin B12r
on D measured by linear sweep voltammetry at a 10-µm radius carbon
ultramicroelectrode in w/o microemulsion made with 0.2 M aerosol AT,
4 M water, and isooctane. Points are experimental: line is best fit to
Eq. (12) with n = 3 by nonlinear regression giving
D1 = 0.63 × 10−6 cm2 s−1 for the water droplets. Using this value with
Eq. (9) gave a hydrodynamic radius of 7.5 nm. (Adapted with permission
from Ref. [27], Copyright by American Chemical Society.)
Equation (12) holds for micelles and w/o micelles using micelle-soluble ferrocene
or o/w microemulsions with single size (Fc) as the electroactive probe. Fitting
distributions. It describes the dependence Eq. (12) to the data is typically done with
of D on CM at constant CX and on CX D0 , D1 , and K as parameters. A series
at constant CM . K is an apparent binding of nonlinear regressions are done, each
constant for a single probe molecule. with a different fixed integer value m =
Plots of D versus CX at constant CM 2, 3, . . . , q, until the minimum standard
are useful for determining D1 because deviation of the regression in the series is
the aggregate size is likely to remain found [26].
constant for the full set of CX . When Figure 6 illustrates the use of Eq. (12)
CM is increased, aggregate size and shape for a w/o microemulsion. Ultramicroelec-
may change [3, 4]. If K is too large, trodes were used to obtain D for the water
no dependence of D versus CX will be droplets using a water-soluble electroactive
observed. probe. Results of the regression analysis
Figure 5 illustrates agreement of D allow an estimate of the hydrodynamic
versus CX data with Eq. (12) for rod-shaped radius of the spherical droplet.
Inspection of Figs 5 and 6 shows that at the reversible midpoint or formal potential
high probe concentration, limiting values in the microheterogeneous fluid (EM r ) is
of D tend toward D1 . Thus, a reasonable

r ◦ RT K0 (1 + Kr )
estimate of D1 for monodisperse systems EM = EW + ln
can be obtained by measuring D at nF Kr (1 + K0 )
relatively large probe concentrations, for RT D
example, 2 to 5 mM. For polydisperse − ln 0 (13)
2nF Dr
systems, additional equilibria must be
considered. A three-state model was used where D s are measured diffusion coef-
to develop a version of Eq. (12) for two ficients in the microheterogeneous solu-
◦
micellar size distributions, and applied to tions, EW is the formal potential of A/B
systems with coexisting globular and rod- in the continuous phase, and the other
shaped micelles [4]. constants have their usual electrochemical
meanings. There are several limiting cases
allowing simplifications:
4.4.3.2 Formal Potentials
Case 1. A and B exist mainly in the
Binding electroactive solutes to micelles or
bound form, and Kr , K0 1.
microemulsion droplets influence formal

potentials (E o ) estimated by voltamme- r ◦ RT K0
EM = EW + ln (14)
try. The pseudophase equilibrium approach nF Kr

can predict E o in the microheterogeneous
Examples include multiply charged metal
media. Equilibria of micelles or droplets
complexes ML3+ /ML2+ in SDS mi-
with both forms of the redox couple
celles [28].
must be considered via a square scheme
Case 2. B is entirely bound but A is
(Scheme 1) for the reversible redox cou-
not bound.
ple A/B.
The pseudophase equilibrium model con- ◦ RT 1 + Kr RT D
r
EM = EW + ln − ln 0
siders solutes to be dissolved in contin- nF Kr 2nF Dr
uous and discrete (micelles or droplets) (15)
‘‘phases’’ in a way similar to the partition An example is ferrocene bound to
in conventional two-phase systems of oil positively charged alkyltrimethylammo-
and water [3, 4]. The equilibrium distribu- nium micelles, where oxidation prod-
tion constants are uct ferricinium ion dissociates when
formed [29].
[AW ] [BW ]
K0 = and Kr =
[AM ] [BM ]
4.4.3.3 Electrode Kinetic Effects
Subscript W refers to the water or Attempts were made several decades ago
continuous phase, and subscript M refers to describe the influence of adsorbed
to the micellar or droplet phase. In general, surfactants on the apparent heterogeneous
AM + ne− BM
K0 Kr
M + AW + ne− M + BM
Scheme 1
electron transfer rate constant k ◦ using

MXn −
$−
−−−
#− MXn−1 + X (17)
Eq. (16) [4, 9]: −− + −
X $−−−
#− X + e (at electrode) (18)
◦
k = k0 (1 − θ) + θk1 (16) When MXn reaches the electrode, either
an electron is donated directly from one
where k0 is the rate constant on a bare of the bound Xs or dissociation (Eq. 18)
electrode, k1 is the rate constant on a yields a free X that donates the electron.
fully coated electrode, and θ is fractional Both coulombic and hydrophobic in-
surface coverage of surfactant. While many teractions of reactants with adsorbed
systems deviate from Eq. (16) at large θ, surfactant on electrodes are important
a decreased k ◦ in micellar solutions and

in determining electron transfer kinet-
microemulsions is often found. ics. Reactants in micellar solutions and
For a reactant molecule or ion in microemulsions can be preconcentrated
a micellar solution or microemulsion, into adsorbed surfactant films on elec-
predictions of electron transfer kinetics trodes [30], yielding mixed layers of re-
at electrodes need to consider [14]: (1) the actants and nonelectroactive surfactants.
distance between the electrode and the Coulombic effects in micellar solutions
reactant, (2) the environment surrounding may result in small kinetic enhancements
the reactant at the time of electron when ionic reactants interact with op-
transfer, (3) structure and dynamics of positely charged surfactant adsorbed on
surfactant aggregates on the electrode, and electrodes. Partial inhibition of electron
(4) dynamics of interactions of the reactant transfer can occur by coulombic repul-
with surfactant structures on the electrode sion if the charge sign on the reactant
and with micelles. A molecular picture and adsorbed surfactant are the same. Hy-
of these events during electron transfer drophobic molecules and ions may show
is by no means clear, and quantitative a small amount of preconcentration on
predictions are not possible at this time. the electrode.
A qualitative view of the above factors is Studies of oxidations in micellar CTAB
given in the following paragraphs. and SDS media [31] suggested average
When small electroactive ions or mole- microenvironments for ferrocene during
cules are bound to larger aggregates like electron transfer in micellar solutions that
micelles or microemulsion droplets, the are more polar than in apolar solvents
reactant (probe) is transported to the elec- but less polar than in water. Similar
trode along with the larger, slower diffus- microenvironmental polarity is found for
ing aggregate. Equation (12) describes the hydrophobic solutes in micelles [13].
influence of concentration of surfactant Studies of electroactive surfactants 1
or reactant on electrochemically measured to 4 (Fig. 7) in dilute or micellar solu-
diffusion coefficients. At [X] >2 mM, the tions have provided insight into adsorbate
measured apparent diffusion coefficient structure, interactions, and dynamics at
D approaches the diffusion coefficient of electrodes [14]. Ordered films were formed
the micelle D1 . This implies electrolysis on glassy carbon electrodes from dilute
of one reactant X per micelle. This elec- solutions of 3. Single-molecule residence
trolyzed X could reside within MXn , or be times on the electrode in micromolar
released by dissociation, as illustrated in solutions were 4.5 s for Fc-C8, 14 s for Fc-
Eqs. (17 and 18) for an oxidation: C12, and 66 s for Fc-C16. Bilayers of 3 were
adsorbed on electrodes from microemul- CH3

sions made with 3, oil, water, and alcohol.
Comparisons of electrochemical k ◦ val-

CH3 N +
ues of surfactants 1 and 2 in microemul- OOCFc
CH3
sions and micellar solutions helped es- (1) 2-Fc
tablish qualitative pictures of dynamics
at relevant electrode-fluid interfaces. Rate CH3
constants for oxidations of ferrocene (Fc)
2-Fc (1) and 5-Fc (2) were similar in homo- CH3 N+
geneous DMF and DMSO on Pt and glassy
CH3 OOCFc
carbon electrodes [32, 33]. However, in
aqueous CTAB micelles, electron transfer (2) 5-Fc
rates were in the order Fc > 2-Fc > 5-Fc,

with tenfold differences in successive val- CH3
ues. This was attributed to 2-Fc and +
CH3 N (CH2)n−1CH3
5-Fc achieving head down orientations on
the electrode prior to electron transfer. CH2 Fc-Cn, n = 8,12,16
Adsorbed CTA+ on the electrode seems to (3)
help order 1 and 2 on the electrode prior Fc
to electron transfer.
In a bicontinuous CTAC microemul- CH3
sion, ferrocene had a k ◦ twice as large

+
HO N (CH2)n−1CH3
as 1 and 2 [33], but values for 2-Fc and
5-Fc were similar. These relatively small Fc (4)
differences in electron transfer rates in
FcOH-Cn, n = 10,14,18
the microemulsion cannot be explained by
head down–tail up orientation of 1 and Fig. 7 Electroactive
ferrocene-containing surfactants.
2 at the time of electron transfer as pro-
posed for micellar solutions. The results
suggest an increased disorder and mobil- In microemulsions, the alcohols are dis-
ity in the electrode–fluid interface in the tributed between the oil–water interfaces
CTAC microemulsion compared to micel- and the oil phase of the microemulsions,
lar CTAB solutions. with increases in chain length favoring
Alcohols are often used as cosurfac- binding at the oil–water interface. Results
tants in microemulsions, and insight has of these and other studies [35–37] suggest
been obtained from the electrochemistry that the presence of sufficient cosurfac-
of ferrocene alcohols (4). Oxidations of tant can improve interfacial fluidity and
FcOHC10 , FcOHC14 , and FcOHC18 [34] facilitate electron transfer at electrodes.
were nearly reversible and controlled by In general, then, surfactant aggregates
diffusion in microemulsions of DDAB, on electrode surfaces can control electron
CTAC, or SDS. In micellar solutions, transfer kinetics. Aggregates such as
electrode reactions were more complex bilayers, cylinders, or surface micelles
and reflected strong adsorption of the adsorb onto electrodes in solutions with
ferrocene alcohols onto the electrode. surfactant concentrations above the CMC.
On hydrophilic electrodes, head down surface, bringing the reactant close enough
adsorption of surfactants seems to be to the electrode for electron transfer,
preferred. analogous to the joining of micelles and
The electroactive reactant may replace submicellar aggregates on the millisecond
adsorbed surfactant at a site on the elec- timescale. This process may be indis-
trode and approach the electrode closely. tinguishable from Eqs. (17 and 18) by
In micellar CTAB, this may occur with voltammetry [14].
reactants 1, 2, and 3. However, adsorbed A picture of the microemulsion–elec-
reactant is unlikely to replace all of the trode interface emerges through compar-
adsorbed nonelectroactive surfactant in ison with the micellar case. Although
micellar solutions well above the CMC, and the surfactant is adsorbed onto electrodes
mixed adsorbate layers form on the elec- from both types of fluids, the adsorbate
trode. Alternatively, if surfactants remain in microemulsions is likely to be more
strongly adsorbed, hydrophobic reactants disordered and more fluid. Results for fer-
could enter the surface film and approach rocene alcohols 4 suggest that there is no
the electrode to within roughly one head long-chain alcohol cosurfactant strongly
group diameter prior to electron transfer. adsorbed onto electrodes. However, some
Both possibilities can be inferred for spe- alcohol may be distributed into a dynamic
cific experimental systems. Comparisons surfactant layer on the electrode. Oils in
of ferrocene, 1 and 2 suggested that un- some microemulsions could also pene-
derivatized ferrocene may approach the trate the surfactant layer on the electrode,
electrode within the distance of one head perhaps inducing disorder and increased
group. Small reactants typically have k ◦

fluidity. However, much more informa-
values in micellar solutions consistent tion is needed before a comprehensive
with this view [4]. Thus, depending on understanding of electron transfer pro-
the ability of the solute to compete with cesses at electrodes in microemulsion is
nonelectroactive surfactant for sites on the reached.
electrode, electron transfer can take place
to adsorbed reactant or to reactant that ap- 4.4.4
proaches the electrode to within a distance Pathway Control of Organic Reactions
of roughly one surfactant head group.
There are several molecular scenarios 4.4.4.1 Direct Electrolyses
for the delivery of micelle-bound reac- Control of organic reaction pathways can
tants into adsorbed surfactant films on be achieved by using micellar media and
electrodes. One possibility is dissociation microemulsions. In the following sections,
(Eq. 17) followed by entry of the reactant selected cases are presented rather than a
into the aggregate film on the electrode, comprehensive review.
orientation near the surface, and elec- Products of the reduction of nitrobenzene
tron transfer. Making the analogy between in water depend on pH [38]. In neutral and
these latter processes and the adsorption weakly acid media, phenylhydroxylamine
rates of 3, entry into the films and orienta- is always the product. In stronger acid
tion is expected to occur on a millisecond (pH< 4), phenylhydroxylamine is reduced
timescale. to aniline. In alkaline media, dimeric
Alternatively, a micelle in solution could products azoxybenzene (5), azobenzene
join with aggregates on the electrode (6), or hydrazobenzene (7) are formed.
O− recovery of lesser-chlorinated products by

breaking the microemulsion with acid [42].
N+ N
N N Micelles can also facilitate one-electron
dimerizations over two-electron reduc-
5 6 tions. An example is electrohydrodimer-
ization of activated olefins, including the
commercially important conversion of
H acrylonitrile to adiponitrile in the Mon-
N
N santo process, which employs an emulsion
H
of acrylonitrile, water, and short-chain
7 tetraalkylammonium salts [43]. High se-
lectivity for the dimer adiponitrile can also
Under conditions where phenylhy- be achieved by using millimolar concen-
droxylamine is the product, a single trations of nonionic surfactants such as
four-electron voltammetric peak is found Triton X-100 [44]. Surfactant adsorbed on
for nitrobenzene. However, at pH > 10 the electrode is thought to create reaction
containing surface-active camphor or sites of low proton availability, shutting
gelatin, a single one-electron wave for re- down the competing two-electron reduc-
duction of nitrobenzene to its anion radical tion that requires protons. Other organic
is found, with a subsequent three-electron dimerizations also benefit from using mi-
wave at more negative potentials [4]. Sim- cellar solutions [4].
ilar behavior occurs in SDS micelles A variety of electrochemical oxidations
and was attributed to hydrophobic-based and reductions have been examined in mi-
stabilization of the nitrobenzene anion cellar solutions. For further information,
radical by the micelles [39]. Surfactant so- the reader is directed to comprehensive
lutions allow electrochemical generation reviews [4, 22, 30, 41, 45].
of relatively stable radical anions of ni- Only a few other direct electrochem-
trobenzene and its derivatives, and have ical reductions have been studied in
been used for electron spin resonance microemulsions. Reductions of naphtha-
(ESR) studies. lene and biphenyl resulted in selective
Electrolysis of nitrobenzene at −1.5 V reduction of a single benzene ring in
versus Ag/AgNO3 in DDAB/hexane/water the polyaromatic hydrocarbon [40], as in
microemulsions, even when acidified, 8 to 12. Products 8 and 9, in addition to
gave dimers 5 or 6 as products [40]. No biphenyl, were also found from catalytic
hydroxylamine or aniline was detected, reduction of polychlorinated biphenyls in
suggesting that the reaction takes place microemulsions [46].
at sites of low proton availability.
Direct reductions of organohalide pol- 4.4.4.2 Electrochemical Polymerization in
lutants have been done in solutions con- Microemulsions
taining ionic and nonionic surfactants [41], Microemulsions may have advantages over
but often with low current efficiencies. conventional solvents for synthesizing
One interesting approach involved the conducting polymers. W/o microemulsion
use of an acid-labile nonionic surfactant, droplets can serve as microreactors to con-
1% oil, and water for the dechlorination trol polymer growth kinetics and particle
of hexachlorobenzene. This allowed facile size [47].
which pyrrole was electrochemically poly-

merized. This porous composite had a
8 9 large surface area for electrochemical re-
actions and allowed greater permeation by
reactants.
Nonconductive w/o microemulsions
10 11 were used for electrochemically induced
polymerization at a specially designed
solid polymer working electrode. Us-
12 ing a microemulsion of water, toluene,
and sodium bis(2-ethylhexyl) sulfosucci-
nate (AOT), acrylamide was polymerized
W/o microemulsions of water, aniline, via persulfate reduction [51]. An applied
light petroleum, and nonionic surfactant potential initiated polymerization, which
Empilan NP-5 have been utilized for elec- continued for several hours after the power
trochemical polymerization of aniline to was off. This resulted in a latex suspen-
polyaniline [48]. Improved homogeneity sion with comparable molecular weight
and conductivity was achieved compared and particle size to those obtained by ther-
with polyaniline grown in water. The het- mal or UV initiation. The degree of stirring
erogeneous nature of the microemulsion controlled the particle size (7–130 nm).
directed the mode of polymer growth The influence of structure and concen-
and improved conductivity and struc- tration of cosurfactant amides in the above
tural features. w/o microemulsions was investigated by
Poly(p-phenylene) was made by elec- ultramicroelectrode voltammetry [36]. Ox-
trochemical polymerization in o/w mi- idation of amides was detected only above
croemulsions [49] of benzene, sulfuric a threshold concentration. The threshold
acid, and anionic, cationic, or neutral decreased as the chain length of amide in-
surfactant. Benzene radical cation was sta- creased because longer chain amides were
bilized by the anionic surfactant, resulting more extensively adsorbed at the oil–water
in polymer with less cross-linking, smaller interface. Increasing the amount of acry-
particle size, and a relatively narrow size lamide makes the interfacial region more
distribution. With cationic surfactants, fluid and increases the permeability to elec-
the radical cation destabilized the water troactive species so that electron transfer
droplets and led to a broader size distribu- can occur more easily [37, 52].
tion of polymer particles.
Stepwise polymer film formation in 4.4.4.3 Mediated Reactions in Micellar
a microemulsion has been explored [50]. Solutions
The first step was the partial polymer- Micellar rate enhancement of thermal and
ization of a microemulsion containing light-initiated biomolecular reactions of-
acrylamide, styrene, pentanol, water, and ten occurs via preconcentration of the two
SDS. Addition of potassium persulfate reactants in the micelles [3]. In electro-
and azo-bis-isobutyronitrile gave a vis- chemical catalysis (Scheme 2), the analo-
cous mixture that was used to coat the gous situation can occur for participants
electrode surface. Evaporation of pen- in biomolecular reactions coadsorbed into
tanol left a highly porous surface onto surfactant films on electrodes. Interfacial
Scheme 2 Simple pathway for

P + e− Q
electrochemical catalysis.
Q+A P+B
preconcentration in micellar solutions can kM

kobs = 2
(20)
be exploited in electrochemical catalysis in φM
which a catalyst (mediator, P/Q) is added
to the system to shuttle electrons between Equation (20) shows that kobs is enhanced
the electrode and a reactant that is oth- by compartmentalization of all the reac-
erwise difficult to oxidize or reduce [30]. tants into micellar volume Vt φM . In elec-
Rate-determining steps (r.d.s.) in electro- trochemical catalysis in micellar solutions,
chemical catalysis are often biomolecular, large rate enhancements are observed
for example, the reaction between Q and when the reaction occurs in surfactant
A in Scheme 2. aggregates on the electrode surface [4,
The observed rate of a chemical reaction 30]. In this case, the compartmentaliza-
in micellar solutions is the sum of rates tion volume where the biomolecular r.d.s.
occurs is that of the surfactant film on
in the continuous and micellar phases [3].
the electrode.
Consider a bimolecular chemical reaction
Catalytic reductions of organohalides
between Q and A in which the reactants
are examples of electrochemical cataly-
are totally bound to micellar aggregates.
sis that can give large rate enhancements
The rate constant kobs , obtained on the
in micellar solutions [4, 30]. Anthracene
basis of amounts of Q and A in the total
derivatives and metal complexes catalyze
system volume (Vt ), is given by
these two-electron carbon–halogen cleav-
age reactions. The pathway is shown in
kM [Q]M [A]M Scheme 3. Catalyst P is dissolved in the
kobs = (19)
[Q][A] micellar solution or immobilized on the
electrode. At applied potentials near E ◦ ,
where kM is the rate constant in the P is reduced to Q, which reacts with aryl
micellar phase, denoted by subscript M. halide ArX, regenerating P. This latter re-
The concentration of the reactant in the action is often the r.d.s. Subsequent steps
micellar phase is the total concentration yield hydrocarbon ArH. Since catalyst P
divided by the volume fraction of the is regenerated in the pathway, the voltam-
micelles, φM , so that metric peak current of P is larger in the
P + e− Q E ° (at electrode)
k1
ArX + Q ArX• + P
k2
k
ArX• Ar• + P
Scheme 3 Electrochemical
catalytic reduction of aryl
halides. Ar• + H+ + e− ArH
presence of ArX. This ‘‘catalytic current’’ dechlorination of polychlorinated biphenyl

can be used to estimate rate constants. (PCB) mixtures [46, 54–56] and the pesti-
Mediator 9-phenylanthracene (9-PA) cide DDT [56, 57] and were also employed
was used to reduce 4-bromobiphenyl in for mediated electrolytic dechlorination of
CTAB solutions on Hg electrodes [4, 30]. PCBs adsorbed onto soils and clays [55].
Voltammetry showed a large preconcentra- The reactions were most successful with
tion of 9-PA in a thick film of CTAB on the a lead cathode in a simple, undivided,
Hg surface. Reduction of 9-PA at −2.2 V two-electrode electrochemical reactor op-
versus SCE gave the anion radical that was erated under constant-current using zinc
stabilized by the positively charged CTAB phthalocyanine as a mediator.
film. The observed k1 for reaction of 9-PA The catalytic conversion of trans-1,2-
anion radical with 4-bromobiphenyl was dibromocyclohexane (DBCH) to cyclohex-
107 M−1 s−1 in 0.1 M CTAB compared ene with macrocyclic cobalt complexes [6,
to 300 M−1 s−1 in DMF, demonstrating 58] was used as a probe to investigate the
a large rate enhancement in the micel- kinetics of mediated electrochemical reac-
lar solution. Electrolysis using this system tions in microemulsions. The P/Q catalyst
showed a clean, high-yield conversion of couples were macrocyclic cobalt complexes
4-bromobiphenyl to biphenyl. CoII L/CoI L such as cobalt corrins, salen,
Other catalysts with E ◦ values more porphyrins, and phthalocyanines.
positive than −2 V showed smaller rate
enhancements because a thick CTAB film Br
Br Co(II)L, 2e−
apparently does not form on Hg electrodes
at these potentials. Negatively charged + 2 Br−
clays on electrode surfaces have been (21)
used to form films of cationic micelles In bicontinuous DDAB microemulsions,
on electrodes for organohalide reductions CoII L mediators reside in the water
positive of −2 V. Details of these and other phase and DBCH in the oil phase, and
mediated reactions in micellar solutions the reaction (Eq. 21) probably occurs at
are found in reviews [4, 30]. the o/w interface. Cyclic voltammetry
(Fig. 8) shows the reversible CoII /CoI
4.4.4.4Mediated Reactions in reduction–oxidation peaks of the mediator
Microemulsions and an increase in reduction current when
DBCH is added. Direct reduction of DBCH
4.4.4.4.1 Reductive Dehalogenations Mi- occurs at a much more negative potential.
croemulsions are usually more useful than The catalytic current was used to es-
micelles for electrochemical synthetic ap- timate the apparent rate constant k1 for
plications because larger amounts of polar the DBCH reaction with CoI L [58]. A
linear plot of log k1 versus E ◦ of the

and nonpolar reactants can be solubi-
lized. Electrochemical catalysis has been mediator (Fig. 9) in DDAB microemul-
used in microemulsions for the elec- sions and organic solvents was obtained.
trolytic conversion of organohalide pollu- A similar linear plot was found for the
tants to hydrocarbons [53] using mediators catalytic reduction of benzyl bromide in
such as metal phthalocyanines and cobalt microemulsions [59]. These plots suggest
complexes. Microemulsions were used that the bimolecular reactions in these
for the complete electrochemical catalytic microemulsions are controlled mainly by
50
b
c
[µA]
25
I
−0.60 −0.80 −1.00 −1.20 −1.40 −1.60 −1.80

E vs SCE
[V]
Fig. 8Cyclic voltammograms at 0.1 V s−1 on glassy carbon electrodes in a DDAB
microemulsion: a) 0.4 mM Co(salen) alone; b) 0.4 mM Co(salen) +1.5 mM DBCH; and c)
1.5 mM DBCH alone without catalyst. (Adapted with permission from Ref. [58], Copyright
by American Chemical Society.)
the intrinsic activation free energy of the DBCH to cyclohexene in microemulsions

reaction via E ◦ , and not by partition be-

was achieved with turnover numbers
tween oil and water phases. This situation much higher than with dissolved cata-
requires sufficiently large o/w interfacial lysts. The catalytic properties of these
area of the fluid. PLL–Co films could be controlled by mi-
As with micelles, incorporation of reac- croemulsion composition [61], and anionic
tants into an adsorbed surfactant layer on micelles formed within the cationic films
an electrode [6, 7] can lead to high reactant in SDS microemulsions.
concentrations in a restricted reaction vol- Turnover rates for the reduction of
ume and enhanced rates of bimolecular re- DBCH to cyclohexene mediated by the
actions. Conversely, when reactants in a bi- PLL–Co film in bicontinuous SDS mi-
molecular r.d.s. reside separately in oil and croemulsions were controlled by the dif-
water phases in microemulsions with inference between the reduction potential
of the reactant and E ◦ of the catalyst in

sufficient interfacial area or slow partition
dynamics, the reaction rate may be slower the film, similar to dissolved cobalt com-
than in a homogeneous solution [27]. plex catalysts. High conductivity and low
The reduction of DBCH was also ef- viscosity of the bulk microemulsion also
fected by using carbon electrodes coated facilitated fast catalyst turnover.
with catalytic films made by covalently link-
ing poly-L-lysine (PLL) onto oxidized car- 4.4.4.4.2 Bond Forming Reactions Me-
bon electrodes, then attaching a reversible diated electrochemical synthesis in mi-
cobalt corrin catalyst [60]. Covalent linkage croemulsions can be used to construct
was necessary for stability of the catalytic bonds. Carbon–carbon linkages between
films in microemulsions. Conversion of an alkyl halide RX and an activated olefin
6
Co(salen)
B12
CoOEP
4
[M−1 s−1]
log k1
CoTPP
CoPcTS
0
−1.2 −1.0 −0.80 −0.60 −0.40
E °' vs SCE
[V]
Influence of catalyst formal potential (E◦ ) on log k1 for reaction of CoI L with DBCH

Fig. 9
in a bicontinuous microemulsion (%) of DDAB/water/dodecane (21/39/40) and in
dimethylformamide (•) for dissolved catalysts vitamin B12, Co(salen), cobalt
phthalocyaninetetrasulfonate (CoPCTS), cobalt tetraphenylporphyrin (CoTPP), and cobalt
octaethylporphyrin (CoOEP). Points from reactions in the microemulsion (%) represent
apparent k1 values for 0.4, 0.5, 1.0, and 2.0 mM catalyst in order of decreasing log k1 .
(Adapted with permission from Ref. [58], Copyright by American Chemical Society.)
CoIIL + e− CoIL (at electrode)

k1
RX + CoIL R–CoIIIL + X−
R–CoIIIL + (H+, e− or hn) + CH2 = CHZ R– CH2CH2Z + CoIIL
Scheme 4 Mediated electrochemical bond formation.
are made by using CoII L complexes as light. The resulting alkyl anion or radical
mediators as in Scheme 4. is trapped by an activated olefin to form a
Here, X represents a halogen atom carbon–carbon bond.
and Z is an electron-withdrawing group. Several examples of mediated bond
Electrochemically reduced mediator CoI L formation are given below. Conjugated
reacts with alkyl halide RX to yield additions of alkyl iodides to 2-cyclohexen-
intermediate R–CoIII L, which is cleaved 1-one gave 3-alkyl cyclohexanones 13
at a more negative potential or by visible (Scheme 5) [62] in 70 to 80% yields in
Scheme 5
Co(I)L + RI R–Co(III)L + I−
O O
hn
R–Co(III)L + + Co(II)L
R
13
microemulsions using −0.85 V + light. Catalysis using vitamin B12 at −1.5 V

Poorer yields were obtained at −1.45 V in versus SCE at carbon electrodes in
the dark because of competing reductions. CTAB and SDS bicontinuous microemul-
Cyclization of 4-bromobutyl-2-cyclohex- sions converted 2-(3-bromopropyl)-2-cyclo-
en-1-one (14) to 1-decalone (15) in mi- hexen-1-one (16) to the 5-endo-trig cy-
croemulsions was also mediated by vi- clization product 4-hydrindanone (17) in
tamin B12 (Scheme 6) [62, 63]. Cis- and 62 to 70% yields (Scheme 7) [64]. Similar
trans-1-decalone were obtained in 90% conditions gave only 7 to 19% of 17
yields with either light or electrochemical in DMF, MeOH, or MeOH/water. The
cleavage of the alkyl–cobalt bond in mi- microemulsions seem to provide reaction
croemulsions and in DMF. Microemul- sites of low proton availability to inhibit
sions gave a stereoselective ratio of 93 : 7 protonation of a key carbanion interme-
trans:cis, but poor stereoselectivity was ob- diate whose protonation would prevent
tained in homogeneous DMF. Preferential cyclization.
formation of thermodynamically favored Finally, phase transfer catalysis
trans-1-decalone involved equilibration of was achieved in DDAB microemul-
isomers via keto–enol tautomerism cat- sions. Dichlorocarbene was gener-
alyzed by hydroxide ions formed by the ated at a carbon cathode from
coelectrolysis of water. dichlorodibromomethane used as the
O O
Br Co(III)L
+ Co(I)L
+ Br−
14
O O
Co(III)L
hn
+ Co(II)L
15
Scheme 6
O O
Co(II)L
Br
−1.5 V
Scheme 7 16 17
oil constituent of a microemulsion [65]. 3. J. H. Fendler, Membrane Mimetic Chemistry,

Tetrabutylammonium bromide helped to John Wiley & Sons, New York, 1982.
4. J. F. Rusling in Electroanalytical Chemistry
catalyze this reaction in the oil phase. The
(Eds.: A. J. Bard), Marcel Dekker, New York,
carbene reacted with cyclohexene to gen- 1994, pp. 1–88, Vol. 19.
erate 7,7-dichloro-bicycloheptane. The best 5. J. F. Rusling in Modern Aspects of Electrochem-
current efficiency was 80%. istry (Eds.: B. E. Conway, J. O’M. Bockris),
Plenum Press, New York, 1994, pp. 49–104,
No. 26.
4.4.5
6. J. F. Rusling, D.-L. Zhou, J. Electroanal.
Summary Chem. 1997, 439, 89–96.
7. J. F. Rusling in Reactions and Synthesis in
This chapter describes the electrochem- Surfactant Systems (Eds.: J. Texter), Marcel
istry of small reactants dissolved in mi- Dekker, New York, 2001; in press.
cellar solutions and microemulsions. A 8. S. Friberg, Adv. Colloid Interface Sci. 1990, 32,
167–182.
major influence of these microheteroge- 9. R. A. Mackay, Colloids Surf. 1994, 82, 1–23.
neous fluids on reversible reactants is 10. C. Tanford, The Hydrophobic Effect, 2nd ed.,
slowing down mass transport. These phe- John Wiley & Sons, New York, 1980.
nomena enable electrochemical probes to 11. J. Israelachvili, Intermolecular and Surface
be used to characterize aggregate mass Forces, 2nd ed., Academic Press, San Diego,
Calif., 1992.
transport and size in the fluids. Tuning the
12. D. F. Evans, D. J. Mitchell, B. W. Ninham, J.
compositions of micelles and microemul- Phys. Chem. 1986, 90, 2817–2825.
sions can control pathways and kinetics of 13. K. Kalyanasundarum, Photochemistry in Mi-
direct organic reactions, polymerizations, croheterogeneous Systems, Academic Press,
and mediated electrochemical reactions. New York, 1987.
14. J. F. Rusling, Colloids Surf. 1997, 123–124,
81–88.
Acknowledgments 15. J. H. Harwell, J. C. Hoskins, R. S. Schecter
et al., Langmuir 1985, 1, 251.
The author’s work on micelles and mi- 16. M. A. Yeskie, J. H. Harwell, J. Phys. Chem.
1988, 92, 2346.
croemulsions described in this chapter was 17. P. Chandar, P. Somasundaran, N. Turro, J.
initially supported by US PHS Grant No. Coll. Interface Sci. 1987, 117, 31.
ES03154 from the National Institute of En- 18. J. F. Scamehorn, R. S. Schecter, W. H. Wade,
vironmental Health Sciences, NIH. More J. Colloid Interface Sci. 1982, 85, 463.
recent synthetic aspects were supported by 19. S. Manne, H. E. Gaub, Science 1995, 270,
1480, 1481.
Grants nos. CTS-9306961, CTS-9632391, 20. S. Manne, J. P. Cleveland, H. E. Gaub et al.,
and CTS-9982854 from NSF. The contents Langmuir 1994, 10, 4409.
of this chapter are solely the responsibil- 21. E. J. Wanless, W. A. Ducker, J. Phys. Chem.
ity of the author and do not necessarily 1996, 100, 3207.
represent official views of NIH or NSF. 22. T. C. Franklin, S. Mathew in Surfactants
in Solution (Eds.: K. L. Mittall), Plenum
Publishing, New York, 1989, pp. 267–286,
References Vol. 10.
23. N. Shinozuka, S. Hayano in Solution Chem-
1. P. L. Luisi, L. Magid, CRC Crit. Rev. Biochem. istry of Surfactants (Eds.: K. L. Mitall), Plenum
1987, 20, 409–474. Publishing, New York, 1979, pp. 599–623,
2. M. Bourrel, R. S. Schechter, Microemulsions Vol. 2.
and Related Systems, Marcel Dekker, New 24. A. J. Bard, L. R. Faulkner, Electrochemical
York, 1988. Methods, John Wiley & Sons, New York, 1980.
25. M. Fleishmann, S. Pons, D. R. Rolison, P. P. 46. S. Zhang, J. F. Rusling, Environ. Sci. Technol.
Schmidt, Ultramicroelectrodes, Datatech, 1993, 27, 1375–1380.
North Carolina, 1987. 47. J. F. Rusling, C. J. Campbell in Encyclope-
26. J. F. Rusling, T. F. Kumosinski, Nonlinear dia of Surface and Colloid Science (Ed.:
Computer Modeling of Chemical and Bio- A. Hubbard), Marcel Dekker, New York,
chemical Data, Academic Press, San Diego, 2002, 4143–4161.
1996. 48. H. S. O. Chan, L. M. Gan, C. H. Chow et al.,
27. A. Owlia, Z. Wang, J. F. Rusling, J. Am. J. Mater. Chem. 1993, 3, 1109–1115.
Chem. Soc. 1989, 111, 5091–5098. 49. A. Manna, K. Bandyopadhyay, K. Vijaya-
28. Y. Ohsawa, Y. Shimazaki, S. Aoyagui, J. Elec- mohanan et al., Langmuir 1998, 14, 84–90.
troanal. Chem. 1980, 114, 235. 50. D. A. Kaplin, S. Qutubuddin, Synth. Met.
29. Y. Ohsawa, S. Aoyagui, J. Electroanal. Chem. 1994, 63, 187–194.
1982, 136, 353. 51. E. Garcia, L. E. Oppenheimer, J. Texter in
30. J. F. Rusling, Acc. Chem. Res. 1991, 24, Electrochemistry of Colloids and Dispersions
75–81. (Eds.: R. A. Mackay, J. Texter), VCH Publish-
31. A. P. Abbott, C. L. Miaw, J. F. Rusling, J. ers, New York, 1992.
Electroanal. Chem. 1992, 327, 31–46. 52. E. Garcia, S. Song, L. E. Oppenheimer et al.,
32. A. P. Abbott, G. Gounili, J. M. Bobbitt et al., Colloids Surf., A 1995, 94, 131–136.
J. Phys. Chem. 96, 11 091–11 095. 53. J. P. Zelina, J. F. Rusling in Environmental
33. G. Gounili, J. M. Bobbitt, J. F. Rusling, Lang- Analysis and Remediation (Eds.: R. A. Mey-
muir 1995, 11, 2800–2805. ers), John Wiley & Sons, New York, 1998,
34. X. Zu, J. F. Rusling, Langmuir 1997, 13, pp. 1567–1583.
3693–3699. 54. E. Couture, J. F. Rusling, S. Zhang, Trans.
35. E. Garcia, S. Song, L. E. Oppenheimer et al., Inst. Chem. Eng.(U.K.) 1992, 70:B, 153–157.
Langmuir 1993, 9, 2782. 55. S. Zhang, J. F. Rusling, Environ. Sci. Technol.
36. E. Garcia, J. Texter, J. Colloid Interface Sci. 1995, 29, 1195–1199.
1994, 162, 262. 56. J. F. Rusling, S. Schweizer, S. Zhang et al.,
37. B. Antalek, A. J. Williams, E. Garcia et al., Colloids Surf. 1994, 88, 41–49.
Langmuir 1994, 10, 4459. 57. S. Schweizer, Q. Huang, J. F. Rusling,
38. A. J. Fry, Synthetic Organic Electrochemistry, Chemosphere 1994, 28, 961–970.
2nd ed., Wiley, New York, 1989. 58. D.-L. Zhou, J. Gao, J. F. Rusling, J. Am.
39. G. L. McIntire, D. M. Chiappardi, R. L. Cas- Chem. Soc. 1995, 117, 1127–1134.
selberry et al., J. Phys. Chem. 1982, 86, 59. D.-L. Zhou, H. Carrero, J. F. Rusling, Lang-
2632. muir 1996, 12, 3067–3074.
40. H. Carrero, J. Gao, J. F. Rusling et al., Elec- 60. D.-L. Zhou, C. K. Njue, J. F. Rusling, J. Am.
trochem. Acta 1999, 45, 503–512. Chem. Soc. 1999, 121, 2909–2914.
41. N. J. Bunce, S. G. Merica, J. Lipkowski, 61. C. K. Njue, J. F. Rusling, J. Am. Chem. Soc.
Chemosphere 1997, 35, 2719–2726. 2000, 122, 6459–6463.
42. S. G. Merica, C. E. Banceu, W. Jedral et al., 62. J. Gao, J. F. Rusling, D.-L. Zhou, J. Org.
Environ. Sci. Technol. 1998, 32, 1509–1514. Chem. 1996, 61, 5972–5977.
43. D. Pletcher, F. C. Walsh, Industrial Electro- 63. J. Gao, C. K. Njue, J. K. N. Mbindyo et al., J.
chemistry, 2nd ed., Blackie Academic, Lon- Electroanal. Chem. 1999, 464/1, 31–38.
don, 1993. 64. J. Gao, J. F. Rusling, J. Org. Chem. 1998, 63,
44. M. R. Moncelli, F. Pergola, G. Aloisi et al., J. 218, 219.
Electroanal. Chem. 1983, 172, 233. 65. C. J. Campbell, J. F. Rusling, Langmuir 1999,
45. R. A. Mackay, J. Texter, (Eds.), Electrochem- 15, 7416, 7417.
istry in Colloids and Dispersions, VCH Pub-
lishers, New York, 1992.
443
5.1 process. However, the right definition of

Electrocatalysis an electrocatalyst requires the comparison
of rates in the presence and the absence of
Enrique Herrero, Juan M. Feliu, Antonio Aldaz the electrode material. The problem arises
Universidad de Alicante, Alicante, Spain just in the possibility of performing this
5.1.1 comparison since, in most cases, there is
Introduction: Concept of Electrocatalysis no homogeneous electrochemical reaction
and Electrocatalyst with which to compare.
Although the definition of catalyst can-
The concept of a catalyst is quite clear in not be directly extrapolated for that cor-
chemistry. It was first defined by Ostwald responding to electrocatalyst, the concepts
in the nineteenth century as a substance underlying the definition can be effectively
that only modifies the velocity of a chemi- used. In a catalyzed reaction, there is a spe-
cal reaction without suffering any chemical cific interaction between the catalyst and
change in the process. This definition is some species involved in the chemical re-
based on a comparison between the rates action. The specific interaction is then the
of the reaction in the presence and in the key point in the catalysis definition and,
absence of the catalyst. Two different types thus, the electrocatalyst can be defined as
of catalysis can be defined: homogeneous an electrode material that interacts specif-
catalysis, in which the catalyst and all the ically with some species involved in the
species involved in the reaction are in the reaction and remains unaltered after the
same phase and heterogeneous catalysis, reaction. This definition allows a good dis-
when the catalyst constitutes a different tinction between electrocatalyzed and non-
phase than that containing the reaction electrocatalyzed reactions: the existence of
species and the reaction takes place on the specific surface interactions is characteris-
surface of the catalyst. An inaccurate ex- tic of an electrocatalyzed reaction.
trapolation of the definition of the catalyst It is implicit in the definition of the
to electrocatalyst would indicate that all the electrocatalyst that this specific alters the
electrode materials are electrocatalyst since energetics of the reaction path, that is, it
the electrode reactions are heterogeneous changes the activation energy, and thus
reactions in which an inert material (the the velocity of reaction. The lack of experi-
electrode) is always present and normally mental or calculated rates for the reaction
does not suffer any chemical change in the in absence of an electrocatalyst is then
444 5 Kinetics and Mechanism of Selected Electrochemical Processes
not necessary to define an electrocatalytic parameters can be measured. For instance,

reaction. Moreover, if several electrode ma- when using a stationary technique in
terials are used and the reaction rates for which the diffusion layer is constant, the
a given reaction depends on the electrode limiting diffusion current for any species
material (and/or surface structure), it can can be written as
be said that these electrodes are acting as
a true electrocatalyst with different effi- jlim,i = ki ci∞ (1)
ciency. These different reaction rates can
where ci∞ is the bulk concentration of i
only be justified if there is a specific interac-
and ki is the mass-transfer coefficient,
tion between the electrode and the species.
which is a function of the diffusion
On the other hand, if the reaction is in-
coefficient and can be calculated for
sensitive to the electrode material, these
the different electrochemical techniques.
electrodes are not interacting with the re-
For a general reaction Ox + ne− = R,
action species and are not electrocatalysts.
the relationship between the stationary
When comparing the catalytic activity of
current and the potential is
several electrodes, the comparisons have
to be made either at constant overpotential ◦ RT kR
E=E + ln
and measuring the current density (value nF kOx
that is proportional to the reaction rate) RT (j − jlim,Ox )
or at constant current density and mea- + ln (2)
nF (jlim,R − j )
suring the overpotential. Of course, the
final goal is always higher rates at lower Whenever a process studied with a station-
overpotentials. ary technique follows Eq. (2), the process
As a first step, it is important to is mass controlled and improvement in the
define which reactions are susceptible to reaction rates will not increase the overall
be catalyzed. In principle, the reaction reaction rate. If not, it may be interesting
rates of any process can be increased. to find a better electrocatalyst.
In a heterogeneous process, such as Improvement of the reaction rates is
electrode reactions, the diffusion of the not the only attribute sought in an
active species to the electrode may be electrocatalyst. Often, a higher selectivity
the rate-determining step of the whole versus any competitive reaction that may
process. In that case, any improvement take place is also required. For that reason,
in the rate of the electron-transfer step catalysts given higher reaction rates but
would not produce any change in the lower material yields are normally not
overall rate of the process, since the mass- used. Thus, a good electrocatalyst is the
transfer process is still the limiting step. one that gives high current densities at
The electrode reaction will behave then as low overpotential and with high selectivity
a reversible diffusion-controlled reaction. for the desired product.
Whenever the reaction in the actual When studying the electrocatalytic be-
experimental conditions is not diffusion havior of a given material, several proper-
controlled, it may be interesting to find a ties may influence its overall performance:
better electrocatalyst for it. The criteria for the electronic properties, the surface struc-
defining a reaction as diffusion controlled ture, and the different types of sites present
depends on the technique employed for the on it, the particle size, surface compo-
study. According to the technique, several sition, and so on. The variety of these
5.1 Electrocatalysis 445
properties makes it difficult to understand reactions the energy of the transition state
what the role of the electrocatalyst in a is affected greatly by the proximity of the
reaction is. This way any interpretation electrode and the reacting species. A strong
may be proposed. In this short chapter, overlap takes place when specific interac-
we will focus on the most fundamentals tion between the surface and some of the
aspects of the electrocatalysis, that is, how reactant species exists.
the different properties of a given mate- These situations are depicted in Fig. 1
rial relate to its electrocatalytic behavior. for a multistep reaction Ox + ne− = R.
In order to understand the fundamen- In this figure, A denotes the transition
tals of electrocatalysis, it is important to state and P and S, respectively, denote the
limit and control the number of possible precursor and successor states that precede
variables for a given experiment. Single- and follow the transition state. The full line
crystal metal electrodes are normally used represents the energetics of the process for
in these studies, since they possess a very a large distance between reactive species
well-defined structure and allow to estab- and electrode. If the reacting species come
lish a correlation between surface structure closer to the electrode surface and these
and its catalytic activity. For that reason, species have attractive interactions with the
we will emphasize the results obtained on electrode surface, the dashed and dotted
well-defined metal electrode surfaces. De- lines are obtained, where the dotted line
spite the use of single-crystal electrodes in represents a stronger overlap than the
the last decades, the number of studies is dashed line. In this simplest model, the
rather limited and only a few systems have transition state corresponds to the same
been investigated systematically. However, reaction coordinate.
it is worth noting that the results shown These curves can be related to the inner-
here will have to be later applied to the and outer-sphere reactions pathways. The
real surfaces employed in more applied full line represents the outer-sphere re-
processes. Several general articles about action, whereas the inner-sphere reactions
electrocatalysis can be found in the litera- are depicted by the dashed and dotted lines.
ture [1–8]. As can be seen, outer-sphere reactions
are always weak overlapping reactions,
5.1.2 that is, reactions in which the electrode
Electronic Effects in Electrocatalysis is not acting as an electrocatalyst. Inner-
sphere reactions can be either strong or
5.1.2.1 Specific Interaction and Transfer weak overlap reactions, depending on the
Rates properties of the reactant species and elec-
The specific interaction between the electrode material. From a chemical point of
trode material and some of reacting species view, a chemical bond between the reac-
has to be defined formally by the overlap of tant species and the electrode is formed
the electronic states of the reactant and the in the strong overlap cases, that is, the
electrode material at the transition state. reactant species are chemisorbed on the
A weak overlap electron-transfer reaction electrode surface.
is defined as that in which the energy of From a practical point of view, it is im-
the transition state is not affected by the portant to distinguish between electrocat-
proximity of the reacting species to the alyzed and nonelectrocatalyzed reactions.
electrode. Conversely, in strong overlap In addition to the criteria given above, in
1
Energy
Ox 3
P
∆G
R
Reaction coordinate
Fig. 1 Schematic free energy profile scheme for an electron

electroreduction. O, oxidized species; R, reduced species; A, transition
state; and P and S, the precursor and successor states, respectively, that
precede and follow the transition state. ( ) for a large distance
between the reactive species and the electrode. (-----) Energetics of the
process in which the reacting species come closer to the electrode surface
and these species have attractive interactions with the electrode surface.
( . . . . ) Same as a dashed line but with a stronger overlap.
situ infrared and Raman spectroscopies material and its surface structure, it can
allow the detection of adsorbed species be said that the electrode is acting as
on the electrochemical interface. These an electrocatalyst. However, different reac-
species present vibrational frequencies tion rates on different electrode materials
that depend on the electrode material and may be the consequence of changes in
that are different than those found for the specific reactant solvation at the in-
the species in solution, a clear indication terface or electrostatic double-layer effects,
that the species are interacting with the and not a consequence of a strong overlap
electrode surface. In other cases, the de- transfer [9].
tection of intermediate adsorbed species In the simplest cases, it is possible
is not possible and indirect methods have to compare rates for homogeneous and
to be used to determine if the electrode heterogeneous electron-transfer reactions.
is acting as an electrocatalyst. Normally, if According to the Marcus theory for
the reaction rates depend on the electrode homogeneous outer-sphere reactions [10],
the reaction rate is given by step (Reaction (1)) takes place through
Langmuir kinetics and k1 , k−1 > k2 and
−λ using the steady state approximation, the
k1 = A1 exp (3)
4RT reaction rate ν can be written as
where λ is the reorganization factor and k2 K1 CA

ν= (7)
A1 , the preexponential factor [10]. Analo- 1 + K1 CA
gously, the reaction rate for heterogeneous
reactions is where K1 is the equilibrium constant for
Reaction (1), and k2 the rate constant for
−λ step 2. Now, it is desirable to express both
k2 = A2 exp (4)
2RT the constants as a function of the Gibbs
energy of adsorption (G◦1 ). K1 can be
Dividing Eqs. (3 and 4) leads to readily written as
1/2
k1 k2 G◦1
= (5) K1 = exp − (8)
A1 A2 RT
When the reaction rates for the homoge- k2 can be expressed as a function of the
neous and heterogeneous reaction rates activation energy (EA2 ) and the electrode
fulfill Eq. (3), it is clear that the reaction potential, E:
is taking place through an outer-sphere
mechanism, and the electrode material is ◦ EA2
k2 = k2 exp −
not acting as a true electrocatalyst. RT

β2 F (E − E ◦ )
× exp − (9)
5.1.2.2 Adsorption and Electrocatalysis RT
As stated previously, adsorption of some
species involved in the electrochemical re- where β is the symmetry factor for
action is the key parameter to consider the electron-transfer step and E ◦ is the
an electrode material as an electrocatalyst. standard potential for the overall reaction.
Therefore, it is important to relate the ac- For a given series of homologous reactions
tivity of such a material with its adsorption (in this case, the reaction S − A + e− → B
properties. Let us consider first the sim- using different surfaces), the activation
plest reaction A + e− → B, in which A is energy can be written as a function of
the only species adsorbed on the catalyst, the Gibbs energy change of the reaction
and try to relate the reaction rate to the using the Brønsted relationship [11]:
adsorption properties of the surface S. A
◦ ◦
simple reaction mechanism is EA = EA − γ G (10)
1 where γ is the transfer coefficient (gener-

S + A ←−→ S − A
−1 ally close to 0.5) and EA◦ is the intrinsic
2 activation energy for the surface in which
S − A + e− −−−→ S + B (6) the Gibbs energy of the reaction of the
adsorbed species is zero. Substituting
where S stands for an adsorption Eq. (10) into Eq. (9), the following equation
site. Assuming that the adsorption can be obtained:

◦ EA2 As can be seen, the reaction rate ν
k2 = k2 exp −
RT is a function of the Gibbs energy of
adsorption, the electrode potential, and
β2 F (E − E ◦ )
× exp − the concentration of A. In fact, the reaction
RT rate can be divided into two separate terms:

◦ E + γ2 G◦2
◦
one containing the dependence of G◦1
= k2 exp − A2 and CA and the other one that depends
RT
on the electrode potential. If the catalytic
β2 F (E − E ◦ )
× exp − (11) activity of several surfaces is compared at
RT constant potential, the reaction rate will
depend only on G◦1 and CA . In Fig. 2,
The Gibbs energy of the second step (G◦2 )
the term
of the overall process is related to the
overall Gibbs energy (G◦T ) and the Gibbs (1 − γ2 )G◦1
k2 exp −
a CA
energy of adsorption (G◦1 ) according to = RT
k2 = (16)
G◦1
◦ ◦
G2 = GT − G1
◦
(12) 1 + exp − CA
RT
where G◦1 depends on the electrode has been plotted versus the Gibbs energy
=
material. Substituting Eq. (12) into Eq. (11) of adsorption. It follows that k2 , and ob-
viously the reaction rate, increases with
◦ E ◦ + γ2 G◦T − γ2 G◦1 the Gibbs energy of adsorption, reaches
k2 = k2 exp − A2
RT a maximum, and then decreases, giving
a curve known as the volcano curve. This
β2 F (E − E ◦ )
× exp − (13) kind of curve is very common in catalyzed
RT
processes, not only in the simple process
and grouping all the constant terms for the that has been considered here, since most
series of homologous reactions (G◦T and reaction kinetics can be described by math-
◦
EA2 ) into k2a , the following expression can ematical expressions similar to Eq. (14),
and therefore have the same qualitative
be obtained:
behavior. This behavior demonstrates that

γ2 G◦1 adsorption increases reaction rates pro-
k2 = k2a exp
RT vided that its energy is not too high. In such
case, the adsorbed intermediate is too sta-
β2 F (E − E ◦ )
× exp − (14) ble to react on to the products. This idea is
RT called, in heterogeneous catalysis, the Prin-
ciple of Sabatier [12]. The comparisons in
Substituting Eqs. (8 and 14) into Eq. (7)
electrochemistry are performed using the
yields
exchange current density (j0 ) instead of the

(1 − γ2 )G◦1 reaction rate. It can be easily demonstrated
k2a exp − that j0 follows the same dependence on
RT
ν= G◦1 that the reaction rate, and there-
G◦1
1 + exp − CA fore, volcano curves are obtained. In fact,
RT such curves have been obtained for several

βF (E − E ◦ ) electrochemical processes, such as hydro-
× exp − CA (15)
RT gen evolution [13], ethylene oxidation [14],
log (k#2)
∆G °1
=
Fig. 2 Generalized representation of k2 versus G◦1 .
formic acid oxidation [15, 16], and oxygen In other processes that involve two
reduction [17]. In all these cases, the reac- adsorbed species that react (for instance,
tions proceed through mechanisms much Langmuir–Hinselwood mechanisms), the
more complicated than those studied here dependence of the reaction rate with the
and for instance, the nature of the relevant adsorption energy is more complicated
adsorbed species for hydrogen evolution is since two adsorption energies are involved.
unclear [18]. An additional complication of For this case, the ideal electrocatalyst is
the electrochemical processes is the possi- the one that has intermediate adsorption
ble dependence of G◦1 with the applied energy for both adsorbed species. If the
potential. Even in the simplest case, this chemistry of the two adsorbed species is
dependence arises either from the specific different, which is normally the case, it
interaction of the reacting species with the is difficult to find a surface that has the
electrode material, or from the specific in- right combination of adsorption energies
teraction of any competing species (water, for both species.
anions) present in solution. In fact, the CO oxidation (among other oxidation
concept of the adsorption isotherm of a reactions of interest in electrocatalysis)
molecule in solution is more complicated belongs to this type of reactions. It has
than that in the gas phase and has to take been normally proposed that adsorbed CO
into account the presence of other species. reacts through a Langmuir–Hinselwood
Additional comments on this problem will mechanism with adsorbed OH from
be made in the next paragraph. water dissociation. Therefore, this reaction
500 500
400 400
300 300
200 200
100 100
[µA cm−2]
[µA cm−2]
0 0
j
j
−100 −100
−200 −200
−300 −300
−400 −400
−500 −500
0.0 0.5 1.0 1.5 0.0 0.5 1.0 1.5
E vs. RHE E vs. RHE
(a) [V] (b) [V]
Fig. 3 Voltammetric profiles of CO oxidation on CO-saturated 0.1 M HClO4
solutions. (a) Au(110) electrode. The dashed line corresponds to the voltammetric
profile obtained in the supporting electrolyte alone. Scan rate 100 mV cm−1 .
(Reproduced with permission from Ref. [24]). (b) Pt(111) electrode. Scan rate
20 mV cm−1 . (Reproduced with permission from Ref. [25].)
requires a metal that adsorbs OH and in the differences of both electrocatalytic

CO with moderate energies. We can then activities is the CO-adsorption energies.
compare, in this case, the performance In spite of the fact that the catalytic
of platinum and gold for the oxidation. performance of these metals towards CO
To assess the interactions between the oxidation is far from optimum (G values
CO molecule and the electrode surface, for the reaction indicates that solution
in situ Fourier Transform Infrared (FTIR) CO should be oxidized at 0.0 V), the
spectroscopy has been used. For platinum oxidation of CO from solution on gold
electrodes, strong absorption bands from takes place at less positive potentials than
adsorbed CO have been detected [19–21], on platinum (Fig. 3) [24, 25]. The low
even after CO has been eliminated from activity of the platinum surface is related
solution, whereas, in the case of gold, very to the strong adsorption of CO, which
weak absorption bands are only observed hinders its oxidation. In this case, the
when CO is present in solution [22, 23]. strong adsorption not only stabilizes the
Therefore, it can be concluded that CO is adsorbed CO but also prevents OH from
strongly adsorbed on the platinum surface adsorption since it completely covers the
in comparison to gold surfaces. Although platinum surface.
the adsorption energy for OH on both Other reaction mechanisms present
surfaces is different, being lower for the additional complications. That is the
more noble metal (gold), the main factor case of the parallel-path mechanisms,
especially when an intermediate is strongly also been used in combination with a linear
adsorbed on the surface poisoning the potential sweep in the technique known as
active sites and thus preventing further pulsed voltammetry [28, 29] (Fig. 4).
reaction of the molecule. These parallel- Another important characteristic of
path mechanisms are often found for the these reactions is the different behavior
oxidation of organic molecules in which of the molecules when adsorbed on elec-
the poisoning intermediate is normally trochemical and ultrahigh vacuum (UHV)
the adsorbed CO. The simplest parallel- environments. In UHV, formic acid dis-
path mechanism is that of the oxidation sociates on a platinum surface to yield
of formic acid. Schematically, it can only H2 and CO2 , and CO is not a reac-
written that tion intermediate [30, 31]. The differences
E1 can be attributed to the presence of water
Active intermediate CO2 molecules on the electrochemical environ-
HCOOH E2 > E1 ments. The presence of water, a polar
E2
CO + H2O CO2 molecule, guides the adsorption process
(17) of the reacting molecule. Methanol ad-
The simplicity of this reaction scheme sorption on Pt(111) surfaces provides a
makes formic acid oxidation a model clear example of such changes induced by
reaction in electrocatalysis. For these the presence of water (or a solvent). In
reactions, it is important to assess the UHV, methanol adsorbs on the Pt(111)
catalytic activity of the surface for both surface through the oxygen atom after
paths (specially the determination of losing the hydrogen bonded to it. In elec-
the reaction path through the active trochemical environments, the first step
intermediate) in order to investigate the in the methanol oxidation is the breaking
catalytic effects of a given surface. The of one out of the three C−H bonds, and
problem is the reciprocal interference the molecule adsorbs on the surface with
of the paths that makes difficult the the OH group pointing outwards [26]. The
evaluation of the reaction rates of both polar nature of the OH group and its in-
paths separately. The reaction rate for the teractions with the water molecule (also
active intermediate path can be obtained polar) prevent the adsorption through the
only in the absence of surface poison. In oxygen group.
order to achieve that condition, several
strategies have been used. The simplest 5.1.3
strategy is to measure the current density. Surface Structure Effects
After cleaning the surface of the poison
(that normally can be achieved at potentials 5.1.3.1 Role of the Surface Symmetry in
above 0.75 V), the electrode potential is Electrocatalysis
then stepped to a lower value at which It is clear that adsorption processes are
the oxidation of the molecule takes place. structure sensitive, that is, the adsorption
As the electrode surface at time zero (just of species depends on the structure
after the step) is free from the poison, the of the surface site. This change in
current extrapolated at t = 0 is the reaction structure may involve several changes:
rate through the active intermediate path. first, the surface will have a different
This method has been used for formic acid energy, as revealed by the dependence
and methanol [26, 27]. This strategy has of the work function with the surface
[mA cm−2] 20
i
a b
0.2 0.5
E RHE
[V]
Fig. 4 (a) Pulsed votammogram; and (b) stationary voltammogram for a
Pt(100) electrode in 0.5 M H2 SO4 + 0.1 M HCOOH solution ( ),
pulsed voltammogram (-----). Pulse voltammetry conditions: sampling
window time: 0.30 s; poison oxidation pulse length: 0.360 s; potential of
pulse: 0.95 V; time between the end of the poison oxidation pulse and the
opening of the sampling window: 0.015 s. Sweep rate: 10 mV s−1 .
(Reproduced with permission from Ref. [28].)
structure. As generally accepted, there is PZC, altering the typical Butler–Volmer

a linear relationship between the work kinetics [34].
function and the potential of zero charge Second, the changes in surface geometry
(PZC) of the electrode [32, 33], which will lead to changes in the geometry of the
in turn plays an important role in the adsorption sites and the distance between
adsorption process of the different species them. The effect of the changes in the
and in the properties of the double site geometry can be studied easily using
layer. In this way, the changes in the low-index single-crystal electrodes, since
surface energy plays a double role in they have a defined surface structure and
the electrocatalysis, one linked to the few different types of sites. For instance,
changes in the adsorption energy of the (111) surfaces of face centered cubic (fcc)
molecules and the second role related to metals have on top, bridged-bonded and
the changes in the double-layer structure. two different types of three hollow sites
The interfacial structure has been proved (tetrahedral and octahedral), whereas (100)
to affect the oxidation/reduction of the surfaces have on top, bridged-bonded and
charges species at potentials close to the four hollow sites. If the most favorable site
for the adsorption of a given species is (Fig. 5)). Moreover, multibonded CO, in
the hollow sites, there will be significant which CO is coordinated to three platinum
differences in the energy of adsorption atoms, is only observed on the Pt(111)
in the (111) and (100) surfaces, owing surface. Additionally, the formation of
to the different geometry of the sites, multibonded CO not only requires a three-
and thus an effect in the electrocatalysis fold symmetry surface but also long-range
of the species will be observed. In other order [35]. Experiments performed using
cases, the species may bond to the surface stepped surfaces show that the multi-
through more than one surface site. In bonded CO does not appear on surfaces
this case, the right surface requires not having short terraces with (111) symme-
only the right adsorption sites, but also try. This effect suggests that long-range
these sites have to be at a distance order domains will have a different re-
compatible with the geometry of the activity. Additionally, the IR frequencies
adsorbing molecule. for on-top CO are also dependent on the
In most cases, it is almost impossible surface symmetry [19–21]. Changes in the
to decouple the effects of the surface IR frequency are associated to changes in
energy changes and the changes in the site the bonding energy of the CO molecule
geometry. Moreover, in electrochemistry, to the surface. If the adsorption energy
adsorption is always a competitive process of the surface is different for the same
since there are other species, like the adsorption site in surfaces with different
solvent or the supporting electrolyte, which geometry, the energy of the surface is play-
can be adsorbed on the electrode surface ing a role in the adsorption process of
apart from the electroactive species. In CO. This example clearly illustrates the
some cases, these competing adsorption interrelation between the different prop-
processes govern the adsorption of the erties (symmetry, energy, etc.) of a given
electrocatalytically active species. Owing surface regarding its adsorption properties
to the intrinsic difficulty in assigning and the difficulty to assign the changes in
the changes of the adsorption to any its adsorption properties to a given struc-
specific property of the surface, we will tural factor.
term the effect related to the surface The changes in the adsorption properties
properties as surface structure effects. It is are not only restricted to changes in the
clear that the interpretation becomes more binding geometry for the adsorption or
difficult when polycrystalline materials are surface structure. When the species has
to be considered. several groups that are liable to bond to the
As aforementioned, surface structure ef- surface, changes in the surface structure
fects on adsorption are well known in may lead to changes in the binding
electrochemistry. A typical example is group. That is the case for urea adsorbed
CO adsorption on low-index single-crystal on platinum single-crystal electrodes. On
platinum surfaces. For CO, three differ- Pt(100), urea is always bonded to the
ent adsorption geometries exist: on-top, surface through the two nitrogen atoms of
bridge-bonded, and multibonded. The rel- the molecule [36]. On the other hand, the
ative population of each adsorption species adsorption is coverage sensitive on Pt(111).
and the surface structure of adsorbate layer For low urea coverages, the adsorption
are dependent on the surface structure takes place trough only one nitrogen atom,
and metal (see for instance Refs. [19–21] whereas for high urea coverages, the
Pt (111) Pt (100) Pt (110)

qCO = 0.16 qCO = 0.16 qCO = 0.2
4 × 10−3 a.u.
qCO = 0.6 qCO = 0.85 qCO = 1.0
Rh (111) Rh (100) Rh (110)

qCO = 0.25 qCO = 0.24
qCO ∼ 0.2
qCO = 0.75 qCO = 0.75 qCO = 1.0
2000 1800 2000 1800 2000 1800

n
[cm−1]
Fig. 5 FTIR spectra for CO irreversibly adsorbed on platinum and rhodium low-index
surfaces in 0.1 M HClO4 at 0.25 V versus SCE at saturation and low CO coverages. The
reference spectrum was acquired at 0.45–0.5 V after CO oxidation. (Reproduced with
permission from Ref. [19].)
O adsorption is through the oxygen of the

carbonyl group [37] (Fig. 6).
C
HN NH Pt (100) The identification of the different ad-
sorbed species and its relation with the
catalytic activity of the surface is much
more troublesome. The problem generally
O arises from the short life of the adsorbed
C
NH 2 Pt (111) intermediates, which makes difficult its
HN Low coverage detection with the usual techniques. In
spite of this fact, the effects of the surface
structure on electrocatalysis are begin-
ning to be well documented. The typical
HN NH 2
example is again formic acid oxidation
C
Pt (111)
O High coverage Fig. 6 Proposed adsorption modes of
urea on Pt(111) at low and high urea
coverages and on Pt(100). (Reproduced
with permission from Ref. [37].)
on platinum single-crystal electrodes. The the first step in the adsorption process of
effect of the surface structure affects both the species of interest is the replacement
the direct and the poisoning paths for of the adsorbed anions. The stronger the
the formic acid oxidation [38]. Among the anion adsorption, the more difficult the
three basal planes, the lowest poison- displacement of the anion by the incoming
ing rate corresponds to Pt(111), which molecule is, resulting in a lower cur-
also shows the lower catalytic activity [39, rent. The voltammetric characteristics for
40]. When comparing the intrinsic activ- formic acid or methanol oxidation remain
ity of Pt(111) and Pt(100) in sulfuric acid essentially the same, but lower currents
medium, it can be seen that the current are obtained when anions are strongly ad-
densities for the Pt(100) surface at 0.5 V sorbed on the electrode surface.
versus RHE are 10 times higher than Until this point, the use of low-index
those obtained on Pt(111) [28, 29]. Simi- crystal surface electrodes has allowed
lar behavior is obtained for methanol on to get insight into the electrocatalytic
platinum electrodes (Table 1) [27]. These phenomena. However, real surfaces are far
results clearly indicate that the reactivity from being perfect and comprise several
of the molecule depends on the electrode types of sites, including step and kink sites,
surface structure. and the size of the ordered domains is
Methanol oxidation also illustrates the rather limited. For that reason, it is also
effect of the competing adsorption on important to rationalize the effect of such
the electrocatalysis. The presence of a sites in the reaction, using stepped and
strong adsorbing anion on the support- kinked surfaces. Because of its complexity,
ing electrolyte hinders the oxidation of there are few examples of these effects
the molecule (Fig. 7). Thus, the current in the literature, but the research on the
densities for methanol oxidation in phos- subject is expected to become more and
phoric acid solution are lower than those more important.
obtained in sulfuric acid, being the high- The presence of defects in a given sur-
est values those obtained in perchloric acid face structure may also play an important
solution [26, 27]. This is because the oxi- role in the electrocatalytic process. Nor-
dation of methanol molecules takes place mally, long-range ordered domains are
at potentials in which the anions are ad- separated from each other by step or kink
sorbed on the electrode surface. Therefore, sites. The substrate atoms that belong to
kink or step sites have special adsorption
properties since they have a lower coordi-
Tab. 1 Maximum voltammetric peak current
densities (jCV ), and instantaneous current nation number, that is, they have a lower
densities at 0.2 V (jE=0.2 ) and 0.4 V (jE=0.4 ) (vs. number of nearest neighbor substrate
Ag/AgCl) for methanol oxidation in 0.1 M atoms than the surface atoms in a well-
H2 SO4 + 0.1 M HClO4 for Pt(111), Pt(110), and ordered domain. These properties may re-
Pt(100) electrodes. Data taken from Ref. [27] sult in an increased catalytic activity. In or-
der to model the real surfaces, stepped and
Pt(111) Pt(110) Pt(100) kinked surfaces are used since they have a
well-defined distribution of these sites and
jCV (mA cm−2 ) 0.74 24.5 4.2
they can be prepared reproducibly.
jE=0.2 (mA cm−2 ) 0.57 6.5 5.2
jE=0.4 (mA cm−2 ) 4.8 17.0 6.1 A clear example of the role of defect
sites in electrocatalysis can be found in
−0.4 −0.2 0.0 0.2 0.4 0.6 0.8
2.0
1.5
[mA cm−2]
i
1.0
0.5
(a) 0.0
1.00
0.75
[mA cm−2]
i
0.50
0.25
0.00
(b)
0.50
0.40
[mA cm−2]
0.30
a3
i
0.20
0.10
a1
0.00 a2
−0.4 −0.2 0.0 0.2 0.4 0.6 0.8

(c) E (Ag/AgCl)
[V]
Fig. 7Voltammetric profiles for the Pt(111) electrode in: (a) 0.1 M
HClO4 + 0.2 M CH3 CH2 OH; (b) 0.1 M H2 SO4 + 0.2 M
CH3 CH2 OH; and (c) 0.1 M H3 PO4 + 0.2 M CH3 CH2 OH. Scan rate
50 mV s−1 . (Reproduced with permission from Ref. [27].)
the oxidation of adsorbed CO on Rh(111) This fact clearly indicates the necessity of
electrodes [41]. For well-ordered surfaces, defect sites in the surface structure of the
the stripping of the adsorbed CO and Rh(111) surface in order to oxidize the
the recovery of the clean surface is adsorbed layer.
quite difficult. However, this process is Glucose oxidation is a clear example
considerably easier in a stepped surface. of the effects of the domain size of the
electrode in the electrocatalysis [42–45]. additional property – they are chiral. This
This molecule is relatively big and has means that it is possible to prepare two
several groups that can be attached to the nonoverlapping specular surfaces having
electrode surface. Consequently, it prefers the same type of kink site. Analogously to
relatively flat areas to adsorb and oxidize. the chiral molecules, the kink sites have
For this reason, the maximum catalytic been termed as R or S depending on its
activity is found for the Pt(111) [42–44] symmetry [46, 47]. If an enantiomer inter-
and Pt(100) electrodes [44]. If stepped acts with a chiral site, the reactivity can be
surfaces are used, the oxidation current is dependent on the symmetry of the chiral
proportional to the terrace size [44], a clear site, leading to the formation of differ-
indication of the influence of the domain ent reaction products and/or to different
sizes in the reaction (Fig. 8). Moreover, reaction rates. Therefore, chiral surfaces
an additional oxidation process at low exhibit stereoselective activity.
potentials appears that corresponds to the In order to observe the stereoselective
oxidation on the step sites at which glucose behavior of the surface, the kink site and a
adsorption would be more reactive. chiral center of the enantiomeric molecule
have to be involved in the adsorption
5.1.3.2 Chiral Surfaces process that yields the reaction products.
As aforementioned, defect sites (kink Moreover, owing to the symmetry rules,
sites) may have an important role crossed activity has to be observed, that
in electrocatalysis. Kink sites have an is, the behavior of an R surface with an S
[mA cm−2]
j
1.0 (100)
(27, 1, 1)
(19, 1, 1)
0.5 (11, 1, 1)
(711)
(511)
0 (311)
0.5 1.0
E RHE
[V]
Fig. 8Voltammetric profiles for the oxidation of 0.01 M glucose in 0.1 M
HClO4 on different Pt(s):[n(100) × (111)] stepped surfaces. Scan rate
50 mV s−1 . (Reproduced with permission from Ref. [44].)
Current density 500 500
Current density
[µA cm−2] 400 400
[µA cm−2]
300 300
200 200
100 100
0 0
−100 −100
0 .2 .4 .6 .8 .9 0 .2 .4 .6 .8 .9
E (Pd/H ref) E (Pd/H ref)
(a) [V] (b) [V]
500 500
Current density
Current density
400 400
[µA cm−2]
[µA cm−2]
300 300
200 200
100 100
0 0
−100 −100
0 .2 .4 .6 .8 .9 0 .2 .4 .6 .8 .9
E (Pd/H ref) E (Pd/H ref)
(c) [V] (d) [V]
Fig. 9 Voltammograms for glucose oxidation on the two chiral Pt(643) electrodes in 0.05 M
H2 SO4 supporting electrolyte. (a) Pt(643)S electrode in 5 mM D-glucose; (b) Pt(643)S electrode
in 5 mM L-glucose; (c) Pt(643)R electrode in 5 mM D-glucose; and (d) Pt(643)R electrode in
5 mM L-glucose. Scan rate: 50 mV s−1 . (Reproduced with permission from Ref. [47].)
molecule should be the same as that of an S activity for the reaction considered. In
surface with an R molecule. Such behavior those cases, an alternative is to try to
has been observed with L- and D-glucose modify the surface properties of the metal
and the R and S Pt(643) surfaces [47, 48]. in order to increase its activity. In most
As can be seen in Fig. 9, the voltammetric cases, this can be achieved by depositing
profile for the oxidation of the molecule a foreign ad-atom on the surface. This
depends on the surface symmetry and the particular case can be compared to alloys
glucose isomer. The main difference is but is should be remembered that bulk
the apparition of a new peak at 0.31 V in properties are different, that is, the bulk
the S – D and R – L combinations, which is phase is a pure metal and only the surface
absent in the other two cases. The current has a different composition in the case
density at 0.31 V increases with the kink of a modified electrode. It is assumed
density, thus proving its relationship with that the foreign ad-atom will maintain
the site symmetry. the surface structure of the substrate,
leading to the growth of a well-defined
5.1.4 surface layer. The new species on the
Surface Composition Effects surface can affect the desired reaction in
three different ways. (1) It may change the
In many cases, the most active metals electronic properties of the surface in such
available have only a limited catalytic a way that the adsorption of the reactant
100
[µA cm−2]
0.5
j
[mA cm−2]
E (RHE)
[V]
j
10
0.5
E (RHE)
[V]
Fig. 10 Quasi-stationary voltammetric profiles of Pt(100) (-----) and
Pd–Pt(100) (θPd = 0.32) ( ) electrodes in 0.25 M
HCOOH + 0.5 M H2 SO4 solution. Inset: blank voltammograms of the
same electrodes in 0.5 M H2 SO4 solution. Scan rate 50 mV s−1 .
(Reproduced with permission from Ref. [49].)
is modified and the catalytic properties oxidation is shown in Fig. 10 (dashed line).
of the surface are enhanced. (2) It also As can be seen, no current is recorded in
may affect the availability of adsorption the positive-going scan, since the surface
sites required for a given reaction, that is completely covered by the poison (CO).
is, blocking the possibility of a poisoning In the negative-going scan, the maximum
reaction. (3) It may provide new reaction current is round 17 mA cm−2 at 0.45 V
sites for some of the reactive species, acting (close to the intrinsic activity of Pt(100)
as a bifunctional catalyst. We will explain electrodes). When palladium is deposited
these effects with some examples. on the electrode surface, the oxidation peak
shifts to 0.21 V, both in the positive- and
5.1.4.1 Changes in the Electronic negative-going scans (Fig. 10, full line).
Properties of the Metal This peak reaches its maximum when the
When a foreign ad-atom is deposited on a palladium coverage is ca. 0.5, with an in-
surface, the electronic properties of the en- tensity of ca. 55 mA cm−2 . The current
semble ad-atom–surface atom are differ- density at 0.21 V for the Pd-modified sur-
ent from that of the original surface atoms. face is much higher than the intrinsic
One of the clearest examples is that of activity at this potential obtained for the
formic acid oxidation on Pt(100) + Pd [49]. unmodified Pt(100) surface and for a pal-
On the unmodified Pt(100) electrode, ladium electrode (negligible currents are
the voltammetric profile for formic acid recorded in both scans for the unmodified
electrode at this potential, see Fig. 4), and In the case of formic acid oxidation,
the oxidation potential has diminished this effect is well documented. It has been
ca. 0.3 V. proposed that the dissociation of formic
It is worth noting that the catalytic acid to yield water and CO (the poisoning
effect of the ad-atom-modified electrodes intermediate) requires two adjacent sites,
can be different from that found for whereas the direct oxidation through the
an alloy of these metals, since even in active intermediate demands only one
the case of solid solutions in the whole site [50]. If the ad-atom deposits randomly
composition range, the surface structure on the electrode surface, there will be
is not the same. In the alloy, there is a a coverage at which the distribution of
microscopic mixture of the two species the ad-atoms does not leave two adjacent
participating in the alloy, whereas in sites. In that case, the dissociation reaction
the ad-atom-modified electrodes, the ad- to yield the CO molecule is completely
atoms are deposited on top of the surface. inhibited. Of course, the reaction through
Moreover, the surface concentration of the the active intermediate exhibits lower
different metallic atoms in the alloy can be currents, since the number of available
completely different from that obtained adsorption sites on the surface are lower,
in the bulk, making the interpretation but the electrode presents the advantage
of the results obtained with the alloy of a very low poisoning rate and stable
electrodes more difficult. In this respect, currents with time. A good example of this
much more research is needed in order to behavior is found for selenium-modified
fully understand the different behavior of Pt(111) surfaces [50]. As the selenium
the alloys. coverage is increased, the currents at 0.5 V
and the total poison accumulation decrease
5.1.4.2 Ensemble Effects linearly, as expected for ad-atoms acting
In the previous section, we have discussed only as a third body. At a selenium coverage
the cases in which the ad-atom plays of 0.28 (coverage at which 84% of the
a positive influence on the catalytic surface platinum sites have been blocked
properties of the surface. In other cases, by the ad-atom) the poison formation is
the ad-atom may not exert any effect. The completely inhibited whereas the surface
only consequence of the presence of the still oxidizes formic acid [50].
ad-atom is that some of the active catalytic
sites of the surface become blocked. It 5.1.4.3 Bifunctional Catalysts
is said that the ad-atom is acting like a For some reactions, two or more species
third body. In some cases, a third-body have to be adsorbed on the surface for
effect may result in a surface that has the reaction to proceed. That is the case,
some interesting properties. That is the for instance, of CO oxidation on metal
case when the reaction mechanism has electrodes. As aforementioned, CO and
parallel paths, or there is a competing an oxygen-containing species have to be
reaction. If the site requirements for the adsorbed on the electrode surface. For
different paths are different, the catalytic such cases, electrocatalysis is much more
activity of the ad-atom-modified surface difficult, since the adsorption energy for
may be altered. These effects are known the different species has to be the adequate
as ensemble effects, since they require a for the reaction to proceed. In some cases,
given structure of the adlayer to appear. the adsorption energy is good for one of
the reactant species but not for the other, activity than the platinum surface [56].
resulting in low current densities for the This kind of bifunctional mechanism is
process. If it is possible to find an ad- also operative for CO oxidation on the
atom that has more adequate adsorption same electrode surface.
energy for the second adsorbing species, It is important to establish if the
the surface will exhibit better catalytic mechanism of the observed catalytic effect
properties. In this case, the reaction played by an ad-atom is electronic or if
will take place preferentially between one the ad-atom is acting as a bifunctional
species adsorbed on the original surface catalyst. The problem arises from the fact
and the second species adsorbed on the that both effects can be considered as
ad-atom. That kind of catalyst is called a short-range effects, that is, the catalytic
bifunctional catalyst. enhancement is only observed in the
The typical example of a bifunctional surface substrate atoms closest to the ad-
catalyst is the ruthenium-modified plat- atom. For the electronic effects, it is known
inum surfaces for the oxidation of that the effect induced by an ad-atom
methanol [51–58]. Of all the metals stud- extends to the first and second row or
ied, platinum is the one that displays the neighboring atoms to the ad-atoms [60].
best catalytic activity for methanol oxi- For the bifunctional catalyst, the pair
dation. However, its catalytic activity is responsible for the catalytic effect is the
still too low for practical purposes. The combination of an active site in the ad-
problem arises from the necessity of the atom and an active site in the surface
adsorption of a second species able to close to it. This way, the determination
transfer the oxygen group required for its of the electrocatalytic mechanism requires
complete oxidation to CO2 . Ruthenium is a detailed knowledge of the distribution
known to have a better affinity for OH than of the ad-atom on the surface and
platinum but its interaction with methanol the condition under which the catalytic
is very poor [59]. Thus, a platinum sur- enhancement is found. These questions
face, which has good interaction energy are currently under investigation.
with methanol, modified with ruthenium, That is the case, for instance, of CO ox-
whose interaction with OH is better than idation in surfaces modified with arsenic
that of platinum, exhibits better catalytic and bismuth (Fig. 11) [61, 62]. Both ad-
properties than platinum. atoms catalyze the oxidation of adsorbed
The observed catalytic enhancement is CO layers. From the measured IR frequen-
potential dependent, since the adsorption cies for CO, it is clear that arsenic and
energy for OH is also potential dependent. bismuth alter the surface energy of the
The energy of OH adsorption increases electrode. However, the frequency shifts
with the potential. Ruthenium adsorbs OH are in opposite directions, the effect of
at less positive potentials than platinum, arsenic is a blue shift of the frequencies
and therefore the onset for methanol oxi- whereas the presence of bismuth causes a
dation takes place at lower potentials [56]. red shift [62]. If the effective mechanism
If the potential is increased, the adsorption of these ad-atoms were an electronic effect,
energy for OH will increase. At a specific the effects of bismuth and arsenic would be
potential, the energy of OH adsorption will the opposite. Owing to the fact that the elec-
be too high for the reaction, and the mod- trocatalytic effect is only observed at poten-
ified surface will display lower catalytic tials at which the ad-atoms adsorb OH, the
30
20
[µA cm−2]
j
10
0 25 50 75
t
(a) [s]
100
[µA cm−2]
j
50
0 5 10 15 20 25
t
(b) [s]
Fig. 11 CO-oxidation transients for Pt(111) at 0.70 V for
(a) As = 0.04 and CO = 0.68; (b) Bi = 0.10 and
CO = 0.56 ( ), As = 0.00 and CO = 0.56 (- - - - - -).
(Reprinted with permission from Ref. [62].)
main electrocatalytic mechanism of these effect and to be able to predict its catalytic
ad-atoms is a bifunctional mechanism. properties regarding other molecules or
other surface ad-atoms systems. Owing
to the complexity of the simulations, it
5.1.4.4 Modeling of the Catalytic Effects
is important to have a well-defined sys-
The last point in electrocatalytic studies is tem in which the experimental data are
trying to model the catalytic effect of the well understood. The most comprehensive
ad-atoms to achieve a full knowledge of its simulations use clusters to model the
surface and tried to determine the most platinum electrodes [65]. In this model, the
favorable adsorption sites and the activa- ad-atoms can play an effective catalysis on
tion energy for the process. However, as the process or act as a simple third body.
a result of the long computational times In the first case, the activity of the surface
required to calculate the different parame- is directly proportional to the number of
ters of the process, the size of the cluster pair ad-atom–surface sites, whereas in the
are limited to a small number of metal latter case, the activity is proportional to
atoms and to one or two different species the number of unoccupied surface sites.
adsorbed on the cluster surface in most Since oxidation of formic acid takes place
of the cases. Therefore, the results ob- though a parallel-path mechanism, the
tained with these simulations represent a effects of different levels of poisoning are
semiquantitative approach to the process. also considered. For the cases in which the
Moreover, the fact that the solvent is ab- surface is completely covered by poison,
sent from most of the simulations and both types of ad-atoms (the catalytically
the lateral interactions between adsorbed effective ad-atoms and the third-body ad-
species are not taken into account, the atoms) produce similar qualitative effects,
applications to electrochemical problems that is, both types increases the current for
are more limited. However, some stud- the oxidation of formic acid [65]. Of course,
ies have tried to calculate the activation the catalytic enhancement is higher in the
energy for the CO-oxidation process in a case of the ad-atoms that modifies the
platinum surface with and without ruthe- electronic properties of the surface, since
nium, obtaining a potential dependence of the global effect will be the combination
the activation energy [63, 64]. of the electronic enhancement and the
In other cases, a simpler simulation third-body effect (any ad-atom always acts
based on statistical models is preformed. as a third body). This is the case, for
This kind of simulation was performed to instance, for the Pt(100) surfaces modified
determine the effect of the different ad- with ad-atoms [65–67]. For the surfaces
atoms in the oxidation of formic acid on with low poisoning, that is, the Pt(111)
50
40
30
[mA cm−2]
j
20
10
Fig. 12 Comparison between
experimental (•) and theoretical
currents versus coverage curves for the 0
electrooxidation of formic acid at
Bi–Pt(111) electrode in 0.25 M 0.0 0.2 0.4 0.6 0.8 1.0
HCOOH + 0.5 M H2 SO4 . (Reproduced
with permission from Ref. [65].)
qBi
electrode, the distinction between the 14. A. T. Kuhn, H. Wroblowa, Trans. Faraday.
effects of both types of ad-atoms is clear. Soc. 1967, 63, 1458.
15. A. Capon, R. Parsons, J. Electroanal. Chem.
For the electronically effective ad-atoms, 1973, 44, 239.
the catalytic activity of the surface increases 16. R. R. Adzic, D. N. Simic, A. R. Despicm
with the ad-atom coverage until it reaches a et al., J. Electroanal. Chem. 1977, 80, 81.
maximum when 75% of the initial surface 17. A. J. Appleby, Catal. Rev. 1970, 4, 221.
sites have been covered, that is, bismuth- 18a. S. Schuldiner, J. Electrochem. Soc. 1963, 110,
332.
modified Pt(111) surfaces (Fig. 12) [65, 68]. 18b. S. Schuldiner, J. Electrochem. Soc. 1968, 115,
For the ad-atoms acting only as a third 362.
body, the current of the oxidation of formic 19. S.-C. Chang, M. J. Waver, Surf. Sci. 1990,
acid diminishes with the ad-atom coverage, 238, 142.
20. S.-C. Chang, J. D. Roth, Y. Ho et al., J. Elec-
as observed with the selenium-modified
tron Spectrosc. Relat. Phenom. 1990, 54/55,
Pt(111) surfaces [65]. 1185.
21. S.-C. Chang, J. D. Roth, M. J. Waver, Surf.
References Sci. 1991, 244, 113.
22. S.-C. Chang, A. Hamelin, M. J. Weaver, J.
Phys. Chem. 1991, 95, 5560.
1. A. J. Appleby in Comprehensive Treatise
23. S.-C. Chang, A. Hamelin, M. J. Weaver, Surf.
of Electrochemistry (Eds.: B. Conway,
Sci. 1990, 239, L543.
J. O’M. Bockris, E. Yeager et al.), Plenum
24. G. J. Edens, A. Hamelin. M. J. Weaver, J.
Press, New York, 1983, pp. 173–239, Vol. 7.
Phys. Chem. 1996, 100, 2322.
2. L. I. Kristalik in Advances in Electrochem-
25. A. Wieckowski, M. Rubel, C. Gutierrez, J.
istry and Electrochemical Engineering (Eds.:
Electroanal. Chem. 1995, 382, 97.
P. Delahay, C. Tobias), John Wiley & Sons, 26. K. Franaszczuk, E. Herrero, P. Zelenay et al.,
New York, 1970, pp. 283–339, Vol. 7. J. Phys. Chem. 1992, 96, 8509.
3. R. Parsons, Proc. Conf. 75th Anniv. Real 27. E. Herrero, K. Franaszczuk, A. Wieckowski,
Soc. Esp. de Fis. Y Quim, Madrid, 1981, J. Phys. Chem. 1993, 97, 9730.
pp. 352–358. 28. J. Clavilier, J. Electroanal. Chem. 1987, 236,
4. J. O’M. Bockris, A. Reddy, Modern Electro- 87.
chemistry, Plenum Press, New York, 1970. 29. A. Fernández-Vega, J. M. Feliu, A. Aldaz
5. J. O’M. Bockris, S. U. M. Khan, Surface Elec- et al., J. Electroanal. Chem. 1991, 305, 229.
trochemistry, Plenum Press, New York, 1973. 30. M. R. Columbia, P. A. Thiel, Surf. Sci. 1990,
6. R. Adzic in Modern Aspects of Electro- 235, 53.
chemistry (Eds.: R. E. White, J. O’M. Bockris, 31. M. R. Columbia, A. M. Crabtree, P. A. Thiel,
B. E. Conway), Plenum Press, New York, J. Am. Chem. Soc. 1992, 114, 1231.
1990, pp. 163–236, Vol. 21. 32. S. Trasatti in Advances in Electrochemistry
7. J. Lipkowski, P. N. Ross, (Eds.), Electrocataly- and Electrochemical Engineering (Eds.:
sis, Wiley-VCH, New York, 1998. H. Gerischer, C. W. Tobias), Wiley Inter-
8. A. Wieckowski, (Ed.), Interfacial Electrochem- science, New York, 1977, pp. 213–321,
istry, Marcel Dekker, New York, 1999. Vol. 10.
9. M. J. Weaver in Chemical Kinetics (Ed.: 33. W. Schmickler, Interfacial Electrochemistry,
R. G. Compton), Elsevier, Amsterdam, 1987, Oxford University Press, New York, 1996.
pp. 1–60, Vol. 27. 34. A. N. Frumkin, Z. Electrochem. 1955, 59, 809.
10. R. A. Marcus, J. Electroanal. Chem. 2000, 483, 35. A. Rodes, R. Gómez, J. M. Feliu et al., Lang-
2. muir 2000, 16, 811.
11. J. N. Brønsted, K. J. Pedersen, Z. Phys. Chem. 36. V. Climent, A. Rodes, J. M. Orts et al., Lang-
1923, 108, 185. muir 1997, 13, 2380.
12. R. I. Masel, Principles of Adsorption and Re- 37. V. Climent, A. Rodes, J. M. Orts et al., J.
action on Surfaces, John Wiley & Sons, New Electroanal. Chem. 1999, 461, 65.
York, 1996. 38. C. Lamy, J. M. Leger, C. Clavilier et al., J.
13. S. Trasatti, J. Electroanal. Chem. 1972 39, 163. Electroanal. Chem. 1983, 150, 71.
39. S. Motoo, N. Furuya, Ber. Bunsen-Ges. Phys. 54. H. A. Gasteiger, N. Markovic, P. N. Ross
Chem. 1987, 91, 457. et al., J. Phys. Chem. 1993, 97, 12 020.
40. S. Motoo, N. Furuya, J. Electroanal. Chem. 55. H. A. Gasteiger, N. Markovic, P. N. Ross
1985, 184, 303. et al., Electrochim. Acta 1994, 39, 1825.
41. R. Gómez, J. M. Orts, J. M. Feliu et al., J. 56. E. Herrero, K. Franaszczuk, A. Wieckowski,
Electroanal. Chem. 1997, 432, 1. J. Electroanal. Chem. 1993, 361, 269.
42. K. Popovic, A. Tripkovic, N. Markovic et al., 57. W. Chrzanowski, A. Wieckowski, Langmuir
J. Electroanal. Chem. 1990, 295, 79. 1997, 13, 5974.
43. K. Popovic, A. Tripkovic, N. Markovic et al., 58. T. Frelink, V. Visscher, J. A. R. van Veen,
J. Electroanal. Chem. 1991, 313, 181. Surf. Sci. 1995, 335, 353.
44. M. J. Llorca, J. M. Feliu, A. Aldaz et al., J. 59. B. J. Kennedy, A. W. Smith, J. Electroanal.
Electroanal. Chem. 1991, 316, 175. Chem. 1990, 293, 103.
45. A. Rodes, M. J. Llorca, J. M. Feliu et al., Anal. 60. P. J. Fiebelman, D. R. Ramman, Phys. Rev.
Quim. 1996, 92, 118. Lett. 1984, 52, 61.
46. C. F. McFaden, P. S. Cremer, A. J. Gellman, 61. E. Herrero, J. M. Feliu, A. Aldaz, J. Catal.
Langmuir 1996, 12, 2483. 1995, 152, 264.
47. G. A. Attard, A. Ahmadi, J. M. Feliu et al., 62. E. Herreo, A. Rodes, J. M. Pérez et al., J.
Langmuir 1999, 15, 1420. Electroanal. Chem. 1995, 393, 87.
48. G. A. Attard, A. Ahmadi, J. M. Feliu et al., J. 63. A. B. Anderson, E. Grantscharova, J. Phys.
Phys. Chem. B 1999, 103, 1381. Chem. 1995, 99, 9143.
49. M. J. Llorca, J. M. Feliu, A. Aldaz et al., J. 64. A. B. Anderson, E. Grantscharova, S. Seong,
Electroanal. Chem. 1994, 376, 151. J. Electrochem. Soc. 1996, 143, 2075.
50. M. J. Llorca, E. Herrero, J. M. Feliu et al., J. 65. E. Leiva, T. Iwasita, E. Herrero et al., Lang-
Electroanal. Chem. 1994, 373, 217. muir 1997, 13, 6287.
51. M. Watanabe, S. Motoo, J. Electroanal. Chem. 66. A. Fernández-Vega, J. M. Feliu, A. Aldaz
1975, 60, 267. et al., J. Electroanal. Chem. 1988, 258, 101.
52. K. Franaszczuk, J. Sobkowski, J. Electroanal. 67. J. Clavilier, A. Fernández-Vega, J. M. Feliu
Chem. 1992, 327, 235. et al., J. Electroanal. Chem. 1989, 261, 113.
53. T. Iwasita, F. C. Nart, W. Vielstich, Ber. 68. J. Clavilier, A. Fernández-Vega, J. M. Feliu
Bunsen-Ges. Phys. Chem. 1990, 94, 1030. et al., J. Electroanal. Chem. 1988, 258, 89.
5.3 will play dominant roles in preserving the

Electrochemical Nucleation and Growth excess energy and, in some cases, this
might result in a long lasting metastable
Benjamin R. Scharifker and Jorge Mostany equilibrium, with indefinite preservation
Universidad Simón Bolı́var, Caracas, of the supersaturated state. Thus, in terms
Venezuela of classical thermodynamics, the total free
5.3.1 energy of a spherical drop in equilibrium
Phase Formation on Electrodes with its vapor is obtained from the prop-
erties of the condensed bulk phase as
The onset of stable structures, known as well as the interface. The Kelvin equa-
nuclei, is characteristic of the phase for- tion [2] ln(p/p∞ ) = 2σ L /rkT states that
mation phenomena [1]. These structures the excess vapor pressure of a drop varies
are aggregates of atoms or molecules, and with the surface tension or free energy of
they become centers for the propagation of the surface, σ , and the molar volume of
the two-dimensional or three-dimensional the condensed phase L , and also with the
new phase. The phase formation process inverse of the radius of curvature of the
is supported by material originated from surface, 1/r; p∞ is the equilibrium vapor
the bulk of the mother phase and in- pressure of the flat surface between phases
variably, during electrochemically driven with the same chemical composition. Ac-
phase formation processes, occurs also cording to this expression, small clusters
with chemical reactions taking place at the will not be stable unless the pressure of the
interface between the developing phase vapor surpasses their equilibrium vapor
and the electrolyte solution. A sequence of pressure, and this will invariably occur at a
events lead to the formation and growth of value higher than the equilibrium pressure
the new phase; these include adsorption of of the bulk phase. Under such conditions,
precursors on the electrode substrate, fre- the increasing volume of the new phase
quently accompanied also by charge trans- reduces the free energy, whereas the free
fer, formation of two- or three-dimensional energy change due to the expansion of the
nuclei, and the growth of the new phase. surface is always positive. Therefore, the
The supersaturation (or excess Gibbs en- combined contributions due to formation
ergy) of the system decreases with the of a bulk phase and a surface produce a
formation and growth of the new phase; maximum of
G, occurring under given
hence the overall process is categorized as conditions at a certain size of the aggregate
a first-order phase transformation. of the new phase, thus defining the critical
An appropriate description of the en- radius r ∗ . The critical radius must be ex-
ergetics and kinetics of the nucleation ceeded for a nucleus to grow irreversibly,
process requires consideration of the equi- and this may happen through fluctuations
librium properties of the bulk phases in temperature, pressure or composition,
involved as well as contributions that arise or, in electrochemical systems, by changes
from the formation and growth of the inter- of the electrical potential.
face between them. Clusters that constitute For a simple electrochemical pro-
the seeds of the phase transformation are cess, Mz+ (sol)
+ ze− = M(ads) , the equi-
usually composed of a small number of librium potential Erev is given by
atoms or molecules. For such small en- the Nernst equation Erev = E ◦ +
tities, the surface or interfacial properties RT /zF ln(aMads /aMn+,sol ), where E ◦ is the
eq
5.3 Electrochemical Nucleation and Growth 513
standard potential of the redox equilibrium Döring [4] provided an analytical solution
between bulk metal and a solution of Mn+ , considering a system of constant compo-
aMads is the equilibrium surface activity of sition in which supercritical clusters are
eq
ad-atoms, and aMn+,sol is the activity of Mn+ reintroduced into the system as the equiv-
in solution. This quantity may be modified alent amount of discrete units. A steady
at will varying the electrode potential to state expression of the nucleation rate fol-
a new value E = Erev + η, where η is the lows, Js = Zαm∗ cm∗ , where αm∗ is the net
overpotential at which the surface activity probability of addition of an atom per unit
of M satisfies the relation aMads /aMads = time from a critical cluster of size m∗ , and
eq
exp(−zF η/RT ). It can be shown that both cm∗ , the equilibrium concentration of criti-
the energy of formation of the critical nu- cal clusters, is related to the monomer con-
cleus,
G∗ = 16πσ 3 M 2 φ(θ)/3ρ 2 (zF η)2 centration through the Bolztmann equa-
(the function φ(θ) of the contact angle tion, cm∗ = c1 exp(−
Gm∗ /kT ) and the
θ between the nucleus and the substrate nondimensional Zeldovich factor Z, which
will be defined later), as well as its radius, accounts for the fact that the steady state
r ∗ = −2σ/zF η, depend on the overpo- concentration at m∗ is only 1/2 of the con-
tential [3]. This illustrates the possibility centration at equilibrium, and that critical
of controlling, rapidly, precisely, and re- clusters may still decay [5].
versibly, the conditions that determine the Classical nucleation theory uses macro-
occurrence of electrochemical nucleation scopic properties characteristic of bulk
processes. phases, like free energies and surface
A critical cluster of atoms is at a max- tensions, for the description of small clus-
imum of free energy. Considering then ters These macroscopic concepts may lack
that the appearance of critical nuclei is physical significance for typical nucleus
the limiting step of the nucleation process sizes of often a few atoms as found
and that the probability of their formation from experimental studies of heteroge-
would be proportional to exp(−
G∗ /kT ), neous nucleation. This has prompted the
the nucleation rate will be proportional to development of microscopic models of the
exp(−16πσ 3 M 2 φ(θ)/3ρ 2 z2 η2 kT ), where kinetics of nucleation in terms of atomic
M is the molar mass of the deposit and interactions, attachment and detachment
ρ is its mass density. The preexponential frequencies to clusters composed of a
factor derives from kinetic arguments, con- few atoms and with different structural
sidering that the populations of clusters of configurations, as part of a general nu-
different sizes are in a steady state, deter- cleation theory based on the steady state
mined by a sequence of reactions such as nucleation model [6]. The size of the crit-
ical nucleus follows straightforwardly in
αn−1 αn
−−→ −−−→ −−−→ −−→ the atomistic description from the loga-
m−1 m m+1 rithmic relation between the steady state
←−− ←−−− ←−−− ←−− nucleation rate and the overpotential. It
βn βn+1
(1) has been shown that at small supersat-
Thus, consecutive addition (and detach- urations, the atomistic description corre-
ment) of atoms leads eventually to the sponds to that of the classical theory of
formation of supercritical nuclei that grow nucleation [7].
irreversibly and are not further considered Given that in electrochemical sys-
in the steady state equations. Becker and tems the supersaturation of a developing
phase is determined by the overpotential, occurs at very early stages, the above ex-
experimental studies of the kinetics of pression soon reduces to N = N0 , and
phase formation processes are gener- nucleation is ‘‘instantaneous’’. Conversely,
ally conducted under potentiostatic confor very small A, the number density of
ditions. At constant potential then, it nuclei increases initially linearly with time,
is generally considered that the steady N = N0 At, and nucleation is said to be
state nucleation rate Js maintains a con- ‘‘progressive’’.
stant value A, which depends on the In multiple nucleation conditions, the
applied overpotential. The nucleation fre- current density may be expressed as a
quency A may be determined from convolution of the growth current of in-
the analysis of the current that repre- dividual nuclei and their birth rate on the
sents the overall rate of electrodeposition. surface. In the very initial stages, nuclei
Analyses of current transients, together are small in both size and number, and
with direct observation of the electrode they grow independently of each other.
surface, constitute the most common The current transient is then given √ by
methods for the study of nucleation i(t) = πzF (2Dc)
√
3/2 (Mt/ρ)1/2 N %( At)
0
processes. [9], where %( At) is the Dawson √ integral,
√ −At
√ At λ2
In many cases of electrochemical phase %( At) = 1 − (e / At) 0 e dλ.
√
formation reactions, the charge-transfer For, At ≥ 20, %( At) → 1 and the equa-
step is fast and the rate of growth of tion above describes a current density
mature supercritical nuclei is well de- proportional to N0 t 1/2 , corresponding to
scribed by mass transport of electrode- instantaneous nucleation; for √ At ≤ 0.2
positing species to hemispherical growth nucleation is progressive, %( At) →
centers [8]. The current attending the (2/3)At and the current density is pro-
growth of an isolated nucleus is proportional to (2/3)N0 At 3/2 [10, 11].
portional to the square root of its age, An overall description of the nucleation
I (t) = πzF (2Dc)3/2 (Mt/ρ)1/2 , where c is and growth process valid for all times
the molar concentration and D is the is in general complicated by interactions
diffusion coefficient of electrodepositing among growing centers. When growth
species in solution. On the other hand, is diffusion controlled, then overlap of
the rate of appearance of nuclei on a the spherical diffusional fields supporting
surface with a number density N0 of ac- the growth of individual nuclei must be
tive sites for nucleation may be expressed considered, eventually collapsing into a
as dN/dt = A(N0 − N ), where A is the single planar field to the plane of the elec-
nucleation rate per site, from which the trode. The many-body problem that ensues
number density of nuclei on the sur- has been approached following several
face as a function of time will be given strategies. General models not restricted
by N = N0 [1 − exp(−At)]. Depending on to the instantaneous case, and allowing
the frequency at which nuclei appear and determination of A and N0 from cur-
the corresponding rate at which active rent transients, consider two-dimensional
sites are depleted, the nucleation pro- projections of the hemispherical diffu-
cess may be classified as ‘‘instantaneous’’ sional fields around individual nuclei onto
or ‘‘progressive’’. When A is very large the plane of the electrode [12] and es-
(A 1/t), depletion of nucleation sites timate their overlap using the Avrami
theorem [13]. The current density is then and more recently in microelectronics
given by [11]: and the manufacture of nanostructures.
The experimental conditions of the elec-
zF Dc
i(t) = trochemical processes (including current
(πDt)1/2 density, potential, electrolyte composition,
(At)1/2
e−At 2 or the use of complexing agents) have
1− eλ dλ substantial effects on the morphology
(At)1/2 0
× and properties of the electrodeposits [14].
1 − e−At
1− In particular, interest in developing effi-
At
cient and environmentally friendly energy

3/2 cM 1/2 sources has motivated the study of the
× 1 − exp −(2π) D catalytic properties of micro- and nano-
ρ
metallic clusters of noble and transition

1 − e−At metals. These clusters may be deposited
× N0 t 1 − (2) on metal surfaces or dispersed in three-
At
dimensional structures, such as porous
The expression above describes appro- matrices, polymeric membranes, or thin
priately experimental current transients conductive polymer films, to catalyze reac-
obtained in studies of electrodeposition tions such as oxygen reduction or the oxi-
processes for numerous systems. The elec- dation of hydrogen or organic molecules.
trodeposition of Hg may be considered Because of the strong effects of surface
as a model system for the study of the structure on reactivity, available theories
fundamentals of electrochemical phase of phase formation have become especially
formation, and is briefly discussed here useful for controlling the size, morphol-
(see Sect. 5.3.4.4). Analysis of current tran- ogy, density, and spatial distribution of
sients allows determining the values of metallic electrodeposits on electrodes.
the nucleation rate and the number den-
sity of active sites. The experimentally 5.3.2
observed dependence of the nucleation Thermodynamics of Electrochemical Phase
rates on overpotential and the charge Formation
transferred during the electrodeposition
reaction are in agreement with funda- The thermodynamic treatment of phase
mental theoretical postulates. The number transitions is usually based upon the
density of nuclei observed microscopically equilibrium properties of the bulk phases
on the surface, on the other hand, are in involved. A description of the formation
good agreement with those predicted from of a new phase requires consideration
the nucleation rates obtained. Overall, ex- of at least two additional factors, the
perimental studies show the dominant contribution of changes in the surface
role of deposit|substrate interactions dur- states to the total energy, and the influence
ing the electrochemical phase formation of these changes on the rate of formation
phenomena. of the new phase.
Detailed knowledge of the mechanisms Transformation of one phase into an-
and kinetics of phase formation are of prac- other will occur only when both phases
tical interest for applications in surface are not at mutual equilibrium; the greater
finish, electrometallurgy, electrocatalysis, the deviation from equilibrium, the greater
the driving force for the transformation. If large ratio of surface to volume, which
both phases are well developed, that is, if for macroscopic bodies tends to zero. Lord
they are sufficiently large for the contribu- Kelvin [2] demonstrated that the larger the
tions of surface effects to be ignored, then vapor pressure in equilibrium with a small
the slightest deviation from equilibrium drop of liquid at a given temperature, the
will cause a displacement of the bound- smaller the radius r of the drop. If the drop
ary between phases, to counterbalance the is not sufficiently small, then its surface
deviation. However, if the phases are not energy may be expressed as the product of
well developed, that is, when one of them the surface area 4πr 2 , and the macroscopic
does not even exist, then the initial sur- surface tension, σ , corresponding to r →
face contributions to the total energy can ∞. The thermodynamic potential of the
be sufficiently important so as to impede total system, composed of vapor V and
the formation of a new phase. Such sys- liquid L, is expressed by the equation,
tems are in metastable equilibrium, and
may remain in such condition for consid- δG = µV dnV + µL dnL + σ dS (3)
erable periods of time. Common examples
where nV and nL represent the number of
of metastable equilibria are supersaturated
molecules on V and L, µV and µL are the
vapors (whether overcooled or overcom-
chemical potentials referred to a molecule
pressed) and supercooled liquids, some of
in the bulk phase at the given temperature
which, as in the case of glasses, remain
and pressure, and S is the surface area
indefinitely as stable amorphous solids.
of the drop. Thermodynamic equilibrium
The new phase generates from the determined by the condition δG = 0 and
metastable system necessarily in the form for a closed system,
of nuclei. These are small clusters of
atoms or molecules that in the prevail- d(r 2 )
ing conditions have developed into a size µV − µL + 4πσ =0 (4)
dnL
sufficiently large to grow spontaneously,
ensuring their own viability and, eventu- Designating the volume occupied by a
ally, the stability of the new phase. The molecule in the liquid phase as L , then
intensive properties of nuclei differ from nL = 4πr 3 /3L , and,
the bulk phase only because of their small 2σ L
size. This point of view is not necessar- µV − µL + =0 (5)
r
ily correct since the properties, structure,
and even the composition of small clusters In the limiting case of r → ∞, this
may not be identical to those of the cor- expression reduces to the ordinary equi-
responding bulk phase, but the notion of librium condition between macroscopic
a ‘‘nucleus’’ determined essentially by its phases, µV = µL . Differentiating Eq. (5)
size is useful to relate the macroscopic and and considering that dµV = V dp and
microscopic descriptions of phase forma- dµL = L dp, where V is the volume per
tion, as described below. molecule in the gaseous phase, then
Leaving aside for the moment the
1
problem of the origin of the nucleus (V − L )dp = 2σ L d (6)
r
and considering only the conditions for
its further growth, the effect of its small Neglecting L in comparison with V
size is manifested by an overwhelmingly and considering the vapor as an ideal
gas, for which V = kT /p, where k is reactions such as

Boltzmann’s constant and T the tempera-
−
ture, then Mz+
(solution) + ze(electrode) −−−→

p 2σ L M(new phase)
ln = (7)
p∞ rkT (8)
for metal ion reduction to the correspond-
where p/p∞ is the ratio between the
ing metallic phase, or
pressures exerted over the surface of
radius of curvature r and a flat surface,
of infinite radius of curvature. This is M(solid) + aXz−
(solution) −−−→
the starting point of the classical theory MXa(new phase)
of nucleation. According to Eq. (7), small
clusters of the new phase are characterized + aze−(electrode)
by an excess vapor pressure. If the (9)
supersaturation pressure is larger than p, for anodic deposit formation.
then the cluster will continue growing; According to Erdey–Grüz and Vol-
if it were less than p, it would tend mer [15], the supersaturation c/c∞ , where
to evaporate. Since every cluster had c is the concentration of ions in solution
to be initially of molecular size (and in equilibrium with a surface of radius of
therefore p p∞ ), it still remains to curvature r, and c∞ is the corresponding
be explained how a nucleus of a new concentration of ions in equilibrium with
phase ever appears. The reason is that a planar surface, is directly determined by
all systems at finite temperature undergo the overpotential, η = E − Erev , where E
fluctuations, which eventually produce is the electrode potential and Erev is the
a cluster with an equilibrium vapor equilibrium potential, in the absence of net
pressure less than or equal to that of current flow for a single electrode reaction.
the supersaturated state. Hence a new For the metal electrode|solution inter-
phase will be formed, and according to face,
classical theory, will continue growing. But
RT c
before entering into a detailed discussion η= ln (10)
zF c∞
of the matter, we will describe briefly some
fundamental concepts necessary to discuss where F is Faraday’s constant. This
equilibrium and phase transformations in relation expresses that through external
electrochemical systems.
control of the electrical potential, the
conditions governing the electrochemical
5.3.3
nucleation process can be precisely and
Electrochemical Nucleation Kinetics
reversibly controlled.
Electrochemical nucleation can be consid-
ered as a series of partial reactions in 5.3.3.1 The Classical Nucleation Model
which at least one step involves charge From Eq. (3), the Gibbs energy of forma-
transfer by electrons or ions through the tion of a nucleus may be formally described
solution|electrode interface. The rate of the as [3]
process is determined by the potential dif-
ference existing across the double layer in
G = δGvolume + δGsurface (11)
where cluster with radius r ∗ is designated as

a ‘‘critical nucleus’’ an entity in precise
4
δGvolume = πr 3
GV (12) equilibrium with the initial phase as
3 implied by (∂
G/∂r)r ∗ = 0, with equal
where
GV is the Gibbs energy of probabilities of growing or decaying.
formation of the bulk phase per unit If r → r + δr, the cluster will grow
volume, and irreversibly; alternatively, if r → r − δr, it
will disappear. Equation (15) is a direct
δGsurface = 4πr 2 σ (13) consequence of Eq. (7), and is known as
the Gibbs–Kelvin equation, relating the
δGvolume is always a negative quantity, Gibbs energy of an interface with it radius
the thermodynamic way of expressing the of curvature. The reversible work required
spontaneity of the process, arising from to form a critically sized cluster from the
the greater stability of the new phase supersaturated phase can be obtained from
with respect to the initial state. On the Eq. (14), integrating between r = 0 and
other hand, the interfacial contribution, r = r ∗ [3],
δGsurface , is necessarily positive; otherwise
the new phase would immediately become ∗ 16π σ3

Ghomo =
dispersed in the initial phase. For small 3 (
GV )2
σ
values of r, the second term predominates,

= 4π(r ∗ )2 (16)
and the total change in Gibbs energy 3
involved in the formation of a nucleus is which has been designated here as
also positive, that is, (∂
G/∂r)small r > 0.
G∗homo (in homogeneous phase) to em-
However, for larger values of r, the phasize that Eq. (16) refers to the work
first term of Eq. (11) predominates and of formation of a nucleus from the
(∂
G/∂r)large r < 0. It is then evident bulk phase.
that a particular value of the radius A comparison of Eqs. (15 and 7) reveals
r = r ∗ exists, in which the Gibbs energy that both the radius of the critical nucleus
change is maximum, (∂
G/∂r)r ∗ = 0. and the work needed to form it are
Thus equating the first derivative of functions of the ratio c/c∞ . When c = c∞ ,
Eq. (11) to zero, that is, when the solution is saturated
in relation to an interface of infinite
∂
G
= 4π(r ∗ )2
GV radius of curvature, the work required to
∂r r ∗
form a nucleus tends to infinite, and the
+ 8π(r ∗ )σ = 0 (14) phase transition does not occur. As the
supersaturation ratio c/c∞ increases, the
from which r ∗ , known as the critical radius of the critical nucleus, and hence its
radius, is obtained as reversible work of formation, diminishes.
2σ Supersaturation is one of the principal
r∗ = − (15) factors determining the work required to

GV
form a critical nucleus, but there are
Clusters with radius of curvature less other factors worth considering too. In
than r ∗ will minimize their energy by the majority of cases and especially under
disappearing in the initial phase, while electrochemical conditions, formation of
larger ones will grow spontaneously. A a new phase occurs on a surface. The
interaction between the substrate and the A second relation between the three
new phase has a strong influence on forces is obtained using Dupre’s equation,
the reversible work of formation of the defining the reversible work W2,3 needed
critical nucleus, thus nucleation occurs to separate a unit area of the solid|liquid
onto those sites on the surface with the interface. Conservation of energy requires
minimal work of formation of critical that [17]
clusters, the so-called active sites, to be
treated further below. σ2,3 + W2,3 = σ1,3 + σ1,2 (18)
Consider the deposition of a liquid
subtracting Eq. (17) from Eq (18),
phase in the solid|solution interface. The
liquid surface makes contact with the solid W2,3 = σ1,2 (1 + cos θ) (19)
at a characteristic angle θ, the contact
angle [16], defined as the angle between This quantity is maximum when θ = 0◦ ,
the tangent planes to the solid and liquid that is, when the liquid completely wets
phases on the contact line, as shown in the solid and W2,3 is zero when θ =
Fig. 1. 180◦ and the liquid evades the solid,
At any point along the equilibrium a characteristic behavior that constitutes
line between the contacting phases, there an effective method to determine the
are three surface tension forces σi,j : σ1,2 interaction energy between surfaces.
between the liquid drop and the solution, The work of formation of a liquid drop
at the contact angle θ with respect onto a surface may be written as
to the surface; σ1,3 between the solid

G =
GV V + σ1,2 S1,2
and the solution, directed along the
plane of the surface, and σ2,3 due to + σ2,3 S2,3 + σ1,3 S1,3 (20)
the solid|liquid interface, also along the
plane but opposing σ1,3 . To maintain where V is the volume of the drop and
equilibrium, the solid reacts with an S1,2 , S2,3 and S1,3 are the surface areas
additional force perpendicular to the of the different interfaces involved. To
surface and directed towards the bulk of calculate the work of formation of the
the solid. The equilibrium condition is critical nucleus, it is sufficient to know the
expressed by the Young equation relation φ(θ) between the volume of the
spherical shell in contact with the solid, v,
σ2,3 − σ1,2 cos θ − σ1,3 = 0 (17) and the volume of a sphere with an equal
s1, 2
s1,3 s2, 3
q
Fig. 1 Model of a liquid drop on a solid surface.

radius of curvature, Ve , in equilibrium where, to simplify the notation, σ1,3 has

with the supersaturated solution. The been written as σ . From this expression,
heterogeneous work of formation of the
2σ M
critical nucleus on the surface would be r∗ = (25)
given by the corresponding work to form a ρzF |η|
sphere with the same radius of curvature,
Figure 2 shows the free energy of hemi-
Eq. (16), multiplied by φ(θ) spherical clusters (θ = 90◦ , φ(θ) = 1/2) of
mercury as a function of the radius, for

G∗hetero =
G∗homo φ(θ) different values of the overpotential.

16π(σ1,3 )3
= φ(θ) (21)
3(
GV )2 5.3.3.2 Nucleation on Electrode Surfaces
The probability of formation of a
The volume of the spherical shell is v = critical nucleus is proportional to
(2 − 3 cos θ + cos3 θ)πr 3 /3 whereas that exp(−
G∗ /kT ) [18], hence the rate
of the sphere is Ve = 4πr 3 /3, thus, of nucleation will be proportional
to exp(−4πσ (r ∗ )2 φ(θ)/3kT ) and this
v (2 − 3 cos θ + cos3 θ) exponential term is central to the theory
φ(θ) = = of nucleation; a full description of the
Ve 4
(22) nucleation rate however requires also
0 ≤ φ(θ) ≤ 1, nucleation at an interface determining the preexponential factors.
requires always less energy than within According to Farkas [19], each cluster is
the bulk phase. characterized by its number of molecules,
To describe electrochemical phase for- n, and through kinetic arguments it
mation processes, it is necessary to es- is possible to evaluate the evolution of
tablish the role of the overpotential on n. Since n is an integer, it can only
the formation of nuclei. Under electro- vary in ±1 due to random addition or
chemical conditions, the Gibbs energy of loss of molecules. n may in principle
interfacial phase formation
GV , appears change in 2 or even more units due
from the charge transfer of ionic species to attachment of dimers, trimers, and
across the double layer, and can be defined so forth, but these events will be
in terms of the product of the net charge infrequent and may be neglected. This
of formation of a cluster and the applied argument leads to an expression for
overpotential: the nucleation rate that contains an
indeterminate integration constant, that
−zF ηρ Becker and Döring managed to eliminate

GV = (23)
M reintroducing clusters attaining a given
size, bigger than critical, as an equivalent
where ρ is the density of the deposit and M number of monomers [4]. In this way, a
is the molar mass of the electrodepositing steady state is maintained even though
species. Then, the free energy of critical the system is closed, and under these
nucleus formation may be expressed as conditions the number of clusters of a
given size remain constant. Thus, a steady
16πσ 3 M 2 φ(θ) state distribution of cluster sizes may be

G∗hetero = (24)
3(ρzF η)2 defined, and the nucleation rate is then
30 mV
3.0
45 mV
2.0
1018 ∆G*
60 mV
[J]
1.0
90 mV
0.0
−1.0
0 5 10 15 20 25
108
[r cm−1]
Fig. 2 Free energy of formation of hemispherical mercury clusters from
Hg2 2+ aqueous solution at different overpotentials, as a function of
nucleus size. σ = 300 dyn cm−1 , obtained from electrocapillary curves of
◦
Hg in KNO3 solution at EHg(II)/Hg = 0.8 V.
αn−1 αn
expressed as the rate of aggregation of −−−→ −−−→
molecules to the critical nucleus, times the n−1 n n+1 (26)
steady state concentration of clusters of ←−−− ←−−−
βn βn+1
critical size.
The treatment of Zeldovich [5] does The change in concentration of clus-
not differ in essence from the already ters of n molecules may be writ-
stated argument and allows for a more ten as dCn(t)/dt = αn−1 Cn−1 (t) − (αn +
precise evaluation of critical parameters βn )Cn (t) + βn+1 Cn+1 (t), which has the
form of Kolmogorov differential equation
with a minimal of assumptions. It fol-
for Markov processes in discrete number
lows considerations from the Kramers
space and continuous time [21]. αn and
theory of transition states, according to
βn are respectively the net probabilities
which the passage through the activated
of incorporation or loss of molecules by
state is subject to random external in- a cluster per unit time, and these may
fluences, similar to those affecting the be defined formally as the aggregation or
Brownian movement of particles in in- detachment frequencies times the surface
tense fields [20]. Thus, the system contains area of the cluster of n molecules. Given
a constant number of molecules suffering the small size of the clusters, αn and βn
statistical fluctuations at temperature T . are not simple functions of n and in gen-
If Cn (t) is the concentration of clusters eral they are unknown. However, if αn
of n molecules at time t, then, allow- and βn are not functions of time, then
ing only transitions between contiguous an equilibrium distribution Cn◦ of clus-
neighbors, that is, from n to n − 1 or ter sizes exists, such that dCn◦ /dt = 0 for
n + 1, Cn (t) = Cn◦ , and the following differential
equation may be stated if we consider that Eq. (29) in Eq. (27),

in equilibrium, the number of clusters
∂Cn (t) ∂ ∂Cn (t)
that progress from n to n + 1 equals those = αn
decaying from n + 1 to n per unit time: ∂t ∂n ∂n

αn Cn (t) ∂(
Gθn )
∂Cn (t) ∂ ◦ ∂ Cn (t) + (30)
= αn Cn (27) kT ∂n
∂t ∂n ∂n Cn◦
To solve this equation, it is convenient
This expression holds if the time during to define a ‘‘critical region’’ of values
which the macroscopic phase transition of n in which the Gibbs energies of
occurs is large compared to the charac- clusters differ from the maximum
G∗
teristic times of the accretion and decay less than kT, the energy of thermal
processes, αn−1 and βn−1 . fluctuations (cf. Fig. 3), extending a cluster
Equation (27) is the Fokker–Planck size range of (2πkT /γ )1/2 , where γ =
equation and describes diffusion in the −(∂ 2
Gθn /∂n2 )n∗ .
presence of an external force, αn plays a Since the Gibbs energy of activa-
role similar to that of the diffusion coeffi- tion is maximal at the critical nucleus,
cient in the Fick equation; for the special (∂
Gθn /∂n)n∗ = 0 and to a first approx-
case of Cn◦ = C ◦ = constant, substituting imation the variation of
G within the
αn for D, we obtain Fick’s law, critical region may be neglected; the sec-
ond term in the right-hand side of Eq. (30)
∂Cn (t) ∂ 2 Cn (t) will be small compared to the first and
=D (28)
∂t ∂n2 a steady state solution is readily found at
fixed supersaturation and when clusters
However, the field of force or gradi-
attaining a given supercritical size are in-
ent in the Gibbs energy introduced by
stantaneously extracted from the system
Eqs. (11–13) must be, in general, retained
and replaced by an equivalent amount
to describe the random diffusion of clus-
of monomers. Under these conditions,
ters through the distribution of their sizes.
∂Cn (t)/∂t = 0 and the steady state nucle-
The equilibrium concentration of n-mers
ation rate Js may be expressed as
is given by

∂Cn
−
Gθn Js = −αn ∗ (31)
◦ ∂n n∗
Cn = C1 exp (29)
kT
which is clearly a diffusion equation
where
Gθn is the standard Gibbs energy that may be solved approximately with
of formation of a n-mer from n monomers the limiting layer model. Thus replacing
and C1 is the concentration of monomers −(∂Cn /∂n)n∗ by Cn∗ /
n, where
n is the
in the initial phase. Referring to Eq. (16) width of the critical region, (2πkT /γ )1/2
describing the Gibbs energy of a critical and taking account that diffusion occurs at
cluster (cf. Fig. 2), it is clear that Cn◦ is both sides of the critical region, then [22]
minimal at n = n∗ (n∗ is the number 2αn∗ Cn∗
of molecules constituting the critical Js = (32)
(2πkT /γ )1/2
nucleus) or r = r ∗ and that
G∗ =

Gθn∗ therefore constitutes the activation Thus the steady state nucleation rate is
Gibbs energy for nucleation. Substituting expressed as the product of the equilibrium
0.655
0.650 kT
0.645
1018 ∆G q
[J]
0.640
0.635
0.630
0.625
30 35 40 45 50 55 60
n
[atoms]
Fig. 3 Critical region around n∗ within which the energies of hemispherical
mercury clusters at η = −90 mV differ from
G∗ less than kT.
concentration of critical clusters, Cn◦∗ , the For small clusters, this procedure is
rate of attachment of monomers to critical prone to criticism on several grounds.
clusters, αn∗ , and a ‘‘nonequilibrium The thermodynamic properties of small
factor’’, known as the Zeldovich factor Z, clusters, for instance, surface tension or
which takes into account that the steady composition, are supposed to be identical
state concentration of n∗ is only 1/2 of the to those of the equilibrium bulk phase. In
equilibrium value and that supercritical particular, it is considered that clusters are
clusters may still decay. The Zeldovich of homogeneous chemical composition,
factor is thus defined as but this is hardly so in clusters that are
2 only a few nm in diameter, given that for
1 ∂
Gθn them the width of the interfacial bound-
Z= − (33)
2πkT ∂n2 n∗
ary would be comparable to the radius
of curvature. It is then inappropriate to
The treatment as outlined above provides extrapolate the normal notion of surface
a good approximation of the steady state tension to such small clusters [23]; since
nucleation rate under conditions in which the surface tension arises from the steep
G∗ may be calculated. The principal discontinuity between phases, it should be

difficulty in finding the nucleation rate is in lower at high curvatures [24]. The surface
evaluating the equilibrium concentration width of the liquid–vapor transition region
of clusters in terms of their formation free has been calculated, and for simple liquids
energies. The classical way of realizing this is about seven or eight molecular diame-
is to consider all clusters as chemically ters [25]. This means that the thermody-
equivalent,
Gθn is obtained from the namic properties of a system composed
Gibbs energy change for their formation of, for example, argon vapor and argon
from monomers, as a sum of ‘‘volume’’ clusters smaller than ∼300 molecules will
and ‘‘surface’’ terms (cf. Eq. 11). differ from those of a system composed of
bulk liquid and vapor phases at the same data and the kinetic equations derived
temperature. This is the basis of the dif- from the classical theory of nucleation.
fuse drop model [26], according to which
the nucleus presents uniform properties at 5.3.3.3 Microscopic Approach to Phase
low supersaturations, with interfacial en- Formation: The Atomistic Model
ergies independent of radius of curvature. As discussed in the previous section, the
At increasing supersaturation, the work liquid drop model confronts serious dif-
of formation decays progressively, the ficulties in describing the nucleation rate
interface becomes diffuse, and the compo- when the critical size is of just a few atoms,
sition of the nucleus becomes nonuniform as is commonly found in experimental
throughout. At even higher supersatura- studies of electrochemical nucleation. For
tion, the composition at the center of the such small clusters, not only is it inap-
nucleus approaches that of the external propriate to use the macroscopic surface
phase; the radius first decreases but then energy to calculate the nucleation rates,
passes through a minimum and then di- the concept of interfacial energy itself is
verges, as the work of formation tends to open to question. For instance, all atoms
zero at very high supersaturation. Hence, are on the surface in clusters smaller than
six atoms. On this basis, Walton [29] devel-
unless the interfacial energy can be de-
oped an expression for the concentration
termined, the Gibbs energy of formation
of clusters using statistical mechanics,
of small clusters, needed for the estima-
avoiding the concept of surface energy
tion of the nucleation rate, will remain
and deriving the nucleation rate from
unknown. In an experimental study of the
purely kinetic considerations. Walton’s tre-
electrochemical nucleation of metals from
atment follows closely the classical work of
molten salts, Hills and coworkers [27] ob- Volmer [30] but using partition functions
served that the radius of critical nuclei instead of classical thermodynamics to de-
passed through a minimum as the super- termine the equilibrium concentration of
saturation increased. The diffuse liquid clusters. The expression obtained for Cn◦
drop model predicts this behavior, but is [29]
its adoption shifts the problem from de- n
termining the surface tension of small Cn◦ C1 En
= exp (34)
clusters to the no simpler task of calcu- N0 N0 kT
lating the Gibbs energy of formation of where N0 is the number density of
a nonhomogeneous phase. An alternative adsorption sites on the surface, which is
description is to introduce in the classical assumed to be much higher than C1 , the
expression of the energy of formation of number density of adsorbed atoms, and
the surface, Eq. (11), the surface tensions En = εn − nε1 is the energy of formation
between the liquid and the supersaturated of the cluster, taking as zero the potential
phase at the different overpotentials [28], energy of the adsorbed atom. En is found
instead of the equilibrium surface tension minimizing the formation energies of all
between the liquid and the supersaturated possible configurations of n atoms in
phase. This is a remarkable result; ex- all possible orientations, which may be
trapolation of macroscopic properties of accomplished, for example, by trial and
materials to small clusters conducts to rea- error. The nucleation rate is calculated
sonable agreement between experimental assuming that all clusters larger than
critical are immediately extracted from both the notion of critical nucleus as well
the system; the critical nucleus is defined as that of equilibrium concentration of
as that cluster for which the probability clusters, starting from the expression for
of growing is less than or equal to the nucleation rate obtained by Becker and
1/2 but, upon incorporating one atom, Döring [4], that is,
attains a probability of growing larger
than 1/2. The probability of supercritical ω+1 C1
Js = (36)
clusters decaying and turning subcritical
m
ω−2 ω−3 · · · ω−i
is considered as very slim and is therefore 1+
ω+2 ω+3 · · · ω+i
neglected. The nucleation rate is then i=2
expressed as
where ω+i and ω−i are the frequencies
n ∗
of attachment and detachment of atoms,
σn∗ Ra2
J =R respectively, from a cluster of i atoms,
a ν
∗ given by
[(n + 1)Qad + En∗ − QD ]
× exp
kT Ci QD
ω+i = αi ν exp −
(35) N0 kT
where R is the incidence rate of atoms
to the surface, σn∗ is the capture cross and
section of critical nuclei, a is the distance
between adsorption sites, ν is an attempt (Ei − Ei−1 + QD )
ω−i = βi ν exp −
frequency, Qad is the bond energy with the kT
surface, and QD is the activation energy for (37)
surface diffusion. Since n∗ can only adopt where αi is the number of ways of forming
integer values, the same size of critical a cluster of size i + 1 incorporating one
nucleus should operate over a range of ad-atom to a cluster of size i and βi is the
temperatures, and discontinuities in the number of ways an atom can detach from
temperature-dependence of the nucleation a cluster of size i. Given that for each birth,
rate are to be expected upon changes i → i + 1, there is a corresponding death,
of the critical size along temperature i + 1 → i, αi and βi are related by
intervals. Regarding the clustering process
as an infinite chain of intermediates αi = βi+1 (38)
with relative concentrations determined
by their respective rates of growth and The sum in the denominator in Eq. (36) is
decay [31] avoids altogether having to taken to an arbitrary size m that may be
consider the notion of critical nucleus but, defined as, for example, the minimal size
unfortunately, does not lead to analytical experimentally observable. Thus Js turns
expressions for the nucleation rate without into the rate of conversion of the largest
a number of additional assumptions, invisible clusters into detectable clusters
although it is possible to obtain explicit and, since the rate in the steady state is
expressions for certain limiting cases [32]. equal for clusters of all sizes, the size of
Stoyanov [6] reexamined Walton’s atom- the critical nucleus does not need to be
istic theory of heterogeneous nucleation defined. Substituting Eqs. (37) and (38) in
and derived the nucleation rate avoiding Eq. (36), the steady state nucleation rate is
obtained as electrodeposition reaction. According to

Eq. (41), U−n∗ − Un∗ −1 = ϕn∗ represents

m
i
Js = ω+1 C1 1+ (βj N0 the separation work of an atom from the
i=2 j =2 cluster of n∗ atoms. Furthermore, if the
number density N0 of sites available for
× exp(−Ej /kT )
adsorption of atoms on the surface is in

large excess with respect to the number
× exp(−Ej −1 /kT )/βj +1 C1 (39)
density N1 of adsorbed atoms, then this
latter is given by
Since E1 = 0 and α1 = β2 , the product in
N0 − N1 ze η
the denominator may be expanded, and N1 = exp

0
(43)
N0 kT
ω+1 C1
Js = i−1 so that
m
β2 N0 −Ei
1+ exp ω−n∗ k−n∗ 1 ϕ ∗
n
i=1 i
α C1 kT = exp −
ω+n∗ −1 k+n∗ −1 c kT
(40)
Equation (40) reduces into Eq. (35) if any ze0 (E0 + η)
× exp − (44)
of the terms in the sum in the denominator kT
is much larger than others, in which case
and the nucleation rate is expressed as
the notion of critical nucleus is recovered.
The rate of electrochemical phase for- Un − (1 − αe )ze0 E0
Js = ZcN0 exp −
mation according to the atomistic model kT
was derived by Milchev and coworkers [7] ∗
%(n )
considering that the frequencies of attach- × exp −
kT
ment of atoms to form stable clusters of n∗ ∗
atoms and detachment to form unstable (n + 1 − αe )ze0 η
× exp (45)
clusters of n∗ − 1 atoms, is given by kT
ω+n∗ −1 = k+n∗ −1 c where Z contains the ratio of products of
frequencies of attachment and detachment
Un∗ −1 − (1 − αe )ze0 (E0 + η)
× exp − from clusters up to the critical and %(n) =
kT nϕ1/2 − ;ϕi represents the difference in
(41) energy of n atoms when they are part of an
and infinitely large crystal on the one hand, and
ω−n∗ = when they form an independent cluster on
the electrode surface on the other, that is,
U−n∗ + αe ze0 (E0 + η) %(n) represents the surface energy of the
k−n∗ exp −
kT cluster of n atoms.
(42)
where Un∗ −1 and U−n∗ are the energies 5.3.4
of transfer of electrodepositing ions from Experimental Studies of Electrochemical
solution to clusters and vice versa, c is Phase Formation
the concentration of electrodepositing ions
in solution, αe is the electronic transfer Metal electrodeposition occurs at the in-
coefficient, e0 is the charge of the electron, terface between an electronically conduct-
and E0 is the equilibrium potential of the ing substrate and an ionically conducting
electrolyte. The sequence of events leading of growth of mature, supercritical nuclei,

to the formation of a new phase may in- are well described by the mass transfer
clude the adsorption of metal ad-atoms, of the electrodepositing species to growth
as well as two-dimensional or three- centers. Nuclei are small and they may
dimensional nucleation, followed by the be regarded as ultramicroelectrodes [38];
growth of the bulk deposit. The par- the diffusion-controlled current flux to an
ticular mechanism for the growth of isolated hemisphere of radius r0 on the
the new phase on the conductive subsurface is [8, 39]
strate is determined by a balance of the
interactions between ad-atoms and the I = 2πzF Dcr0 (46)
substrate, <(Meads −S), and the crystal-
lographic compatibility of the substrate Where zF is the molar charge transferred
with the bulk metal deposit [1]. Under during the electrodeposition process and
quasi-equilibrium conditions, in which D is the diffusion coefficient of the
low supersaturations and the absence of electrodepositing species in the bulk of
kinetic limitations might be assumed, the solution. The same current also causes
phase formation occurs through two fun- the growth of the deposit, thus, following
damental mechanisms: if the energy of Faraday’s law:
metal|substrate interactions dominates, zFρ dV
<(Meads –S) < <(Meads –Me), then the I = (47)
M dt
concentration of ad-atoms Meads is small
and the formation of three-dimensional and accordingly,
clusters is favored. On the other hand,
zFρ dr
if <(Meads –S) > <(Meads –Me), then the 2πzF Dcr0 = 2πr02 (48)
degree of commensurability between the M dt
crystalline habits of the substrate and the from which the radius of an isolated
adsorbate determines if the new phase nucleus of age t is given by
grows as epitaxial bidimensional layers or
as three-dimensional nuclei. 2DcMt
r0 = (49)
ρ
5.3.4.1 Three-dimensional Nucleation
Experimental studies of electrochemical and the growth current of an isolated
phase formation may be carried out by di- nucleus as a function of time is
rect observation of the number of nuclei πzF (2Dc)3/2 M 1/2 t 1/2
on the electrode surface at different stages I (t) = (50)
ρ 1/2
of the growth process [33, 34], or by relat-
ing the electrical current with the number In general, experimental studies of the
density of nuclei on the electrode [8, 35], kinetics of formation of metallic phases
in which case it is necessary to establish on electrodes are performed potentiostati-
also the rate of growth of the deposit. In cally, stepping the potential from a positive
many cases of electrochemical phase for- to a negative value with respect to the re-
mation reactions, particularly during the versible potential of the electrodeposition
electrodeposition of metals from molten reaction. Under potentiostatic conditions,
salts [36], or aqueous solutions, [37], the it is generally assumed that both the steady
charge-transfer step is fast and the rate state nucleation rate Js and the growth rate
of nuclei will remain a constant, with rates The solution of Eq. (53) may be ex-
depending on the applied overpotential. pressed as
The value A of the steady state nucleation
rate can be obtained from analysis of the πzF (2Dc)3/2 M 1/2 N0 t 1/2
i(t) = % (54)
current, which represents the overall rate ρ 1/2
of the electrodeposition process. For a real where
surface at constant overpotential, the rate

of formation of nuclei over a given num- e−At (At)1/2 2
%=1− eλ dλ (55)
ber of active sites N0 may be expressed (At)1/2 0
as [12]
%(At)1/2 is Dawson’s integral, a tabu-
dN
= (N0 − N ) (51) lated function [42]. For, At ≥ 20, % → 1
dt and Eq. (54) describes the current density
corresponding to ‘‘instantaneous’’ nucle-
Integration with initial condition N = 0 at
ation, proportional to N0 t 1/2 ; for At ≤ 0.2,
t = 0 gives the number density of nuclei
% → (2/3)At and the current density is
as a function of time:
proportional to (2/3)N0 At 3/2 , correspond-
ing to the limiting case of ‘‘progressive’’
N = N0 [1 − exp(−At)] (52) nucleation.
Equation (54) and the limiting cases of
Typical values of N0 are 104 cm−2 < ‘‘instantaneous’’ and ‘‘progressive’’ nucle-
N0 < 1010 cm−2 , frequently dependent of ation are valid only during the initial stages
the potential but always smaller than of the electrocrystallization process, when
the number density of atoms on the the number and size of the nuclei are suffi-
surface, of ca. 1015 cm−2 [40, 41]. For ciently small so as for growth to occur inde-
small values of A ( 1/t), Eq. (52) reduces pendently from each other. The general de-
to N = N0 At and the nucleation process scription of the current transient requires
is said to be progressive, while for large consideration of interactions among nu-
values of A, N = N0 and the saturation clei [9–12, 43–48]. Initially, as discussed
number density of nuclei is achieved above, nuclei grow independently of each
immediately after applying the potential other and the process is controlled by
step, and nucleation is instantaneous. hemispherical diffusion to isolated nu-
Under conditions of multiple nucleation, clei – the ‘‘short time’’ behavior. After
the current density can be expressed some time, the diffusion fields around nu-
as the convolution of the growth rates clei overlap and eventually the growth of
of individual nuclei at time t after the deposit is controlled by linear diffusion
having born at time u (Eq. 50) and towards the planar electrode surface – the
their appearance on the surface, dN/du = ‘‘long time’’ behavior. To overcome the
difficulties encountered in describing this
AN0 exp(−Au), Eqs. (51 and 52) [9]:
many-body problem, some approaches
have taken into account the overlap of
πzF (2Dc)3/2 M 1/2
i(t) = three-dimensional diffusional fields using
ρ 1/2 Avrami’s theorem [13], either considering
t
dN slices [45] or two-dimensional projections
× (t − u)1/2 du (53)
0 dt on the electrode surface in the form of
circular ‘‘diffusion zones’’ [9, 11, 12, 43, where the term at the left is the current to
44]. Other treatments have avoided the a single hemispherical nucleus, Eq. (52),
Avrami approach altogether, considering and the right-hand side term is the Cottrell
mean field approximations [46] or using current for a plane electrode of equivalent
statistical mechanics to calculate nucleus area πrd2 , generated at the moment of birth
growth [48], but restricting consideration of the nucleus. Accordingly, the radial
to only the limiting case of instantaneous flux (flux density × area) equated to the
nucleation. Other developments [49, 50] planar flux to the equivalent diffusion
have considered the changes in concentra- zone, defines its area as [12]
tion occurring around nuclei during their 1/2
cM
diffusion-control growth and the effect of πrd2 = (2π) 3/2
D (t − u) (57)
these changes on the nucleation rates and ρ
on the spatial distribution of nuclei, but However, nuclei do not grow indepen-
do not discuss how to determine the nu- dently of each other and their interactions
cleation rate A and the number density of results in overlap of the diffusion fields
active sites N0 from experimental data. around them. Overlap may be accounted
Models allowing to determine A and for using Avrami’s theorem, [13]
N0 obtain the current transient from
the material flux to ‘‘free’’, noninteract- θd = 1 − exp(−θex ) (58)
ing, growth centers, considering circular
diffusion zones around them, with time- where θd is the fraction of electrode surface
dependent radii rd . As shown in Fig. 4, covered by planar diffusion zones and θex
these are two-dimensional projections of is the ‘‘extended’’ coverage, that is, the
three-dimensional fields that define, for fraction of surface that would be covered
a hemispherical nuclei of radius r0 , an in absence of overlap,
equivalent area toward which the same π 3/2 (2Dc)3/2 M 1/2
amount of matter that diffuses spherically θex = AN0
ρ 1/2
to a three-dimensional nucleus diffuses by t
planar diffusion. × (t − u) exp(−Au) du (59)
The mass balance results in 0
that after integration yields

πzF (2Dc)3/2 M 1/2 (t − u)1/2
= 1/2
ρ 1/2 2MDc
θex = 2π N0 (πDt)1/2 t 1/2 >
zF Dc(πrd2 ) ρ
× (56) = αAt>
πD(t − u)1/2 (60)
(a) (b) (c)

Fig. 4 (a) Hemispherical diffusion fields; (b) projection to the
electrode surface as planar diffusion zones; and (c) their
overlap on the plane of the surface.

where α = (2π)3/2 D(cM/ρ)1/2 N0 /A and 3/2 cM 1/2
× 1 − exp − (2π) D
ρ
(1 − e−At )
>=1− (61) 1 − e−At
At × N0 t 1 − (66)
At
The coverage of the electrode surface with
diffusion zones, considering overlap, is
5.3.4.2 Two-dimensional
then given by the following expression:
Electrocrystallization
θd = 1 − exp[−α(At − 1 + e−At )] (62) Two-dimensional phase formation occurs
preferentially when a strong substrate-
The ‘‘instantaneous’’ limit corresponds to metal interaction exists, a process that
the α → 0 limit, high nucleation rate A on typically involves the formation of growth
small number density of active sites N0 , centers a few atoms thick, that expand and
while the ‘‘progressive’’ situation results coalesce to form a monolayer that serves
when the value of A is low and/or N0 as a precursor deposit to subsequent two-
is elevated, that is, when α → ∞. The dimensional metal layers.
current density is obtained as the planar As discussed before, for a determinate
diffusive flux to an electrode of fractional supersaturation level
µ, ion transfer
area θd : from solution to electrode across the elec-
zF Dc trochemical double layer occurs accom-
i(t) = θd (63) panied by the corresponding free energy
δ
change,
GV (n) = −nze|η|. A fraction
The length of the diffusion layer is ob- φ(n) of this energy is consumed in the
tained from the balance between Eqs. (54 formation of the new nucleus|solution
and 63) [11] and nucleus|surface interfaces, giving
for the total change of Gibbs free en-
πzF (2Dc)3/2 M 1/2 N0 t 1/2 zF Dc
1/2
%= θex ergy associated with the formation of a
ρ δ cluster comprising n atoms,
GV (n) =
(64)
−nze|η| + φ(n). This expression has a
as
> minimum
G∗ for the equilibrium form
δ = (πDt)1/2 (65) of the crystal with the lower surface en-
%
ergy σi among the i crystalline faces, as
According to this expression [11], the determined by the Gibbs–Curie expres-
rate of expansion of the diffusion layer sion [51], φ = ;i σi Ai = min (for constant
depends on the nucleation rate A, but volume V ).
not on the density of active sites. For In the case of epitaxial grow over
instantaneous nucleation, δ = (πDt)1/2 , an ideal substrate, the surface energies
whereas for progressive nucleation, δ = of the two-dimensional clusters reduce
(3/4)(πDt)1/2 . From Eqs. (54, 60 and 64), to border energies εi (J cm−1 ), as the
the current density is described as [11, 12] emergence of the new phase is balanced
(At)1/2 by disappearance of substrate area. Under
e−At 2
these conditions, the 2D crystal satisfies
1− 1/2
eλ dλ
zF Dc (At) 0 the Gibbs–Curie condition for the excess
i(t) =
(πDt)1/2 1 − e−At free energy, in analogy to the three-
1−
At dimensional case; φ = ;i εi Li = min (at
constant area), where Li is the border rate of attachment of atoms to the growth
length of the 2D crystal. By appropriate center, the radial expansion rate would
considerations for simple geometrical be dr(t)/dt = Mk/ρ and the temporal
forms [52], the number of atoms of the dependence of the radius r(t) = Mkt/ρ.
critical 2D nucleus may be obtained from These expressions lead to the following
equation for the current associated with the
bε2
n∗ = (67) growth of an isolated disc-shaped nucleus:
(ze|η|)2
2πnF k 2 hM
where b is a geometrical factor, the area i2D (t) = t (70)
ρ
occupied by an atom in the cluster surface,
and ε the mean value of the border energies As previously considered with respect
for the i crystalline faces, from which to the three-dimensional nucleation sit-
the free energy of the critical 2D nuclei uation, the expression for the current
is
G∗ = ze|η|n∗ The Volmer–Weber transient corresponding to multiple two-
expression applied to the two-dimensional dimensional nucleation requires consider-
case gives ation of the coalescence of growth centers,
which diminishes the edge length avail-
bε2 able for attachment of new atoms or
J = A2D exp − (68)
ze|η|kT molecules, and produces a decay of the cur-
rent. By means of the Avrami theorem (see
where A2D may be considered constant, al-
Sect. 5.3.4.1), the following expressions are
though it encloses the potential-dependent
obtained for the limiting cases of instan-
contributions coming from the Zeldovich
taneous and progressive two-dimensional
factor Z and the probability of attachment
nucleation:
to critical nuclei ωn∗ n∗ .
Analysis of ln(J ) versus 1/|η| relation- 2πzF MhN0 k 2 t
ships from experimental data allows to I2D,inst =
ρ
determine the nucleation parameters n∗
and
G∗ for the critical nucleus, through πN0 M 2 k 2 t 2
× exp − (71)
the relation [52] ρ2
d ln J 1 d
G∗ πzF MhN0 k 2 t 2
=− = n∗
zF
(69) I2D,prog =
d|η| KT d|η| RT ρ

πAN0 M 2 k 2 t 3
The description of the current transient × exp −
arising from nucleation and growth of 3ρ 2
two-dimensional monolayers, considers as (72)
limiting step the attachment of atoms or
molecules to the edge of the growing two- 5.3.4.3 Spatial Distribution of Nuclei
dimensional centers. The required charge The fundamental reason to study nu-
to form a disc-shaped nuclei of radius r cleation kinetics under potentiostatic
and height h is Q(r) = zF πr(t)2 hρ/M, conditions is to maintain constant super-
from which the instantaneous current saturation and hence a constant nucleation
i(r, t) = zF 2πr(t)hρ/M)dr(t)/dt may be rate, except for nonstationary effects oc-
obtained differentiating with respect to curring at the onset of the potential step.
time. If k (in moles cm−2 s−1 ) is the However, the irreversible growth of the
nuclei diminishes the concentration of Thus, the probability of forming a

electrodepositing species and inhibits the nucleus somewhere in the electrode sur-
nucleation rate in their vicinity. Thus the face at a given moment t is affected by
number of nuclei that ultimately attain the presence of a nucleus i located at
stable growth over the surface depends on a distance ri ; this may be expressed as
the local variations of the nucleation rate, pi (ri , t) = A(ri , t)/A0 . If the effects of
as a result of changes in concentration several nuclei in the vicinity of the site
occurring during the growth process. At were independent, then the joint nucle-
sufficiently high overpotentials and in the ation probability would be expressed as
presence of excess supporting electrolyte, p(ri , t) = Dpi (ri , t), but the radial sym-
the local nucleation rate at a given mo- metry of the diffusional hemispherical
ment t and at a distance r from the center fields generated by the growth of nuclei
of a nucleus of age τ = (t − u) and radius is robust [54], thus the concentration of
r0 = (2DcM/ρ)1/2 τ 1/2 is given by [53] electrodepositing species is determined by
the nucleus that exerts the more intense
A(r, t) diffusional field on the particular site, act-
n∗ +1 ing then as the most influential neighbor,
(2DcM/ρ)1/2 τ 1/2
= A0 1 − (73) and superseding the effects induced by
r
other smaller or more distant neighbors.
where A0 is the noninhibited nucleation Then the local nucleation rate is not de-
rate at a sufficiently large distance from a termined by the combined effects of all
growing nucleus. The local variation of the the nuclei surrounding the particular site,
nucleation rate due to the growth of nuclei but only by the ‘‘most influential neigh-
is shown in Fig. 5. bor’’ [50]. This most influential neighbor
−rd/r0 rd /r0
1.0
n* = 0
0.8
2
A(r, t)/A0
0.6
5
0.4
10
0.2
0.0
−20 −10 0 10 20
r/r0(t)
Fig. 5 Local nucleation rate around a growing nucleus of radius r0 ,
for different critical nucleus sizes. Also shown is the exclusion zone of
radius rd , defining regions inside (A = 0) and outside (A = A0 ) the
exclusion zone.
and its effects on the nucleation process Evaluation of NS from Eq. (74) requires
may be identified defining a circular zone numerical methods, but for the par-
around each nucleus, inside which for- ticular case of ‘‘progressive’’ nucle-
mation of further nuclei is excluded. The ation, that is, when nucleation is slow
overall rate of nucleation will diminish as and occurs on a large number den-
a result of the decreased supersaturation sity of active sites, then [55] NS =
caused by the growth of nuclei, and will 1/2
AN0 /[2(8πcM/ρ)1/2 D] . Values of NS
eventually vanish as the exclusion zones obtained from this expression agree favor-
expand and overlap, covering all the area
ably with experimental number densities
available. In Fig. 5, the size of the exclusion
of Pb, Ag, and Hg nuclei obtained by di-
zone has been identified with that of the
rect microscopic observation of electrode
diffusion zones considered on the current
surfaces [56–58].
transient analysis. Under this assumption,
Local variations of the nucleation rate af-
the saturation number density of nuclei on
the electrode surface is given by fect not only the saturation number density
∞ of nuclei, but also their spatial distribu-
tion [59, 60], and this can be studied ana-
NS = AN0 exp − (2π)3/2 D
0 lyzing images of the electrode surface. The
1/2 spatial distribution of nuclei is represented
cM (Au − 1 − e−Au )
× N0 in Fig. 6 by the probability density of the
ρ A
distances between nearest neighbors, in
× e−Au du (74) nondimensional, normalized form. For
1.0
0.8
(2pNs)−1/2p(r)
0.6
0.4
0.2
0.0
0 1 2 3 4
(2pNs)1/2r
Fig. 6 Normalized probability density of nearest neighbor distances
for 2.2 × 106 drops/cm2 of mercury, electrodeposited onto vitreous
carbon at 220 mV from 0.01 mol dm−3 Hg2+ aqueous solution
(
) [58]. Also shown are the probability densities corresponding to
uniformly distributed drops (thin line) and excluded nucleation for
distances less than rd = [(8π cM/ρ)1/2 D(t − u)]1/2 from each drop
(heavy line), obtained from digital simulations [59].
uniformly distributed nuclei, the nearest distances as the ordering of the nuclei on
neighbor distribution can be obtained as the plane increases. As is shown in Fig. 6,
follows [56]. The probability of finding the inhibition of nucleation in the vicinity of
nearest neighbor at a distance r from a growing nuclei induces a certain degree of
given nucleus, P (r), can be expressed as ordering, which may be advantageous, for
the product of the probability of not find- example, in the preparation of catalysts to
ing one at a distance less than r, and the reduce sintering and to assist accessibility
probability of finding one between r + dr: of the catalytic sites by reacting species.
r In general [50, 56–58], electrodeposited
dP (r) = p(r)dr = 1 − p(u) du nuclei at low overpotentials are uniformly
0 distributed, reflecting a random distribu-
× 2πNS r dr (75) tion of active sites on the electrode surface,
while at higher overpotentials the spatial
This may be now expressed as dP (r)/[1 − distribution of nuclei is affected by inhibi-
P (r)] = 2πNS rdr, and after integration tion of the nucleation rate in their vicinity.
becomes P (r) = 1 − exp(−πNS r 2 ). Thus,
p(r)dr = 2πNS r exp(−πNS r 2 ) (76) 5.3.4.4 Experimental Examples

Mercury electrodeposition is a model sys-
The probability density in Eq. (76) has max- tem for experimental studies of electro-
imum value for rm = 1/(2πNS )1/2 , and we chemical phase formation. On the one
may express it in nondimensional form in hand, the product obtained is a liquid drop,
terms of this characteristic distance, as corresponding very well with the liquid
drop model of classical nucleation theory.
(2πNS )−1/2 p(r) = (2πNS )1/2 r Besides, electron transfer is fast [61] and
× exp(− 12 [(2πNS )1/2 r 2 ]) (77) therefore the growth of nuclei is controlled
by mass transport to the electrode sur-
This normalized distribution does not de- face [44]. On the other hand, the properties
pend on the number density of nuclei on of the mercury|aqueous solution interface
the electrode surface, and only denotes have been the object of study for over a
the ordering pattern of nuclei on the elec- century and hence are fairly well under-
trode surface. If nuclei were, for example, stood. The high overpotential for proton
ordered on a square array, then the dis- reduction onto both mercury and vitreous
tance between nearest neighbors would carbon favor the study of the process over
1/2
be 1/NS and the probability of find- a wide range of overpotentials. In spite
ing the nearest neighbor would change of the complications introduced by the
from 0 to 1 at a nondimensional distance equilibrium between the Hg2+ , Hg2 2+ ,
−1/2
NS /(2πNS )−1/2 = (2π)1/2 = 2.507. In and Hg species, this system offers an
a hexagonal array, all nearest neighbors are excellent opportunity to verify the funda-
separated by a nondimensional distance mental postulates of the electrochemical
√
(4π/ 3)1/2 = 2.694. For a random uni- nucleation theory. In fact, the dependence
form distribution, the probability density of the nucleation rate on the oxidation
is maximum at the characteristic distance, state of the electrodepositing species is
that is, (2πNS )−1/2 /(2πNS )−1/2 = 1. In fully consistent with theory: critical nu-
nondimensional coordinates, the maxi- clei appear with similar sizes and onto
mum of the distribution shifts to higher similar number densities of active sites
either from Hg2 2+ or Hg2+ solutions, as for sufficiently long times, at which linear
expected from the identical interactions be- dependence is observed in accordance
tween deposit and substrate in both cases. with Cottrell behavior, or by hydrodynamic
This demonstrates the dominant role of methods using rotating electrodes. In the
the interaction between the surface and particular case of mercury, nucleation may
the deposit in the phase formation probe realized also from Hg(II) solutions, and
cess, as discussed before. Fig. 7 shows a Fig. 9 shows the nucleation rates as a func-
family of current transients obtained for tion of the overpotential from both Hg(I)
mercury electrodeposition from Hg(I) so- and Hg(II) solutions, obtained from the
lution onto vitreous carbon at different analysis of current transients.
potentials. According to Eq. (45), the atomistic
A similar family of theoretical current theory predicts a linear dependence of
transients, obtained from evaluation of the nucleation rate with overpotential for
Eq. (66) with the known values of z, F , c, constant size of the critical nucleus, as
M, and ρ corresponding to the reduction shown in Fig. 7. Also, in accordance with
of Hg(I) to Hg(0), is shown in Fig. 8. Eq. (45), the value of the slope doubles
As shown in both the experimental and for mercury deposition from Hg(II) as
theoretical transients, the currents first compared with deposition from Hg(I).
increase as a result of the formation of mer- The slopes of logarithmic plots of the
cury drops and their growth controlled by nucleation rate as a function of the
radial diffusion, decaying at longer times overpotential, such as that shown in
due to the onset of semi-infinite planar dif- Fig. 9, confirm the small sizes of critical
fusion. The diffusion coefficients may be nuclei in electrochemical phase formation
obtained either from i versus t −1/2 plots processes. Frequently, as suggested by the
2.5
250
240
2.0 230
220
[mA cm−2]
1.5 210
i
200
190
1.0
0.5
0.0
0 5 10 15 20 25 30
t
[s]
Fig. 7 Current transients recorded during Hg electrodeposition onto
vitreous carbon from 0.01 mol dm−3 Hg2 2+ solutions in KNO3
1 mol dm−3 at the indicated overpotentials (in mV) [58].
2.5
g
f
2.0 e
d
[mA cm−2]
c
1.5
b
i
a
1.0
0.5
0.0
0 5 10 15 20 25 30
t
[s]
Fig. 8 Current transients according to Eq. (66), for mercury
electrodeposition from 10 mM Hg2 2+ solutions, with D = 10−5 cm2 s−1 ,
at nucleation rates A and number densities of active sites N0 ,
respectively, of: (a) 0.02 s−1 and 0.4 × 106 cm−2 ; (b) 0.04 s−1 and
0.6 × 106 cm−2 ; (c) 0.06 s−1 and 0.8 × 106 cm−2 ; (d) 0.08 s−1 and
1.0 × 106 cm−2 ; (e) 0.10 s−1 and 1.2 × 106 cm−2 ; (f) 0.12 s−1 and
1.4 × 106 cm−2 ; and (g) 0.14 s−1 and 1.6 × 106 cm−2 . The broken line
indicates the Cottrell diffusive limiting current to the electrode surface.
102
101
100
[s−1]
A
10−1
10−2
10−3
0.15 0.20 0.25 0.30
h
[V]
Fig. 9 Nucleation rates of Hg onto vitreous carbon from 0.01 mol dm−3
Hg2 2+ (◦) and 0.01 mol dm−3 Hg2+ (
) solutions, as a function of the
overpotential [58].
slopes in Fig. 9, single atoms adsorbed on phase formation processes, and some
the surface grow irreversibly and play the illustration has been given by way of
role of critical nuclei. a specific example, the electrodeposi-
Usually, only a small fraction of the sur- tion of mercury. Many other theoreti-
face is active for nucleation, the number cal and experimental studies show that
density of active sites is potential depen- electrochemical methods may be advan-
dent, and much less than the atomic tageously used for the electrodeposition
density of the surface [40]. Analysis of of a wide variety of materials. Electro-
transients by means of Eq. (66) allows chemical phase formation offers possibil-
also obtaining the number densities of ities for controlling the chemical nature,
sites on the electrode surface. As shown size, morphology, and spatial distribu-
in Fig. 10, nucleation of mercury either tion of dispersed and nanostructured
from Hg2 2+ or Hg2+ solutions takes deposits of a variety of materials, includ-
place at a given overpotential onto a ing metals [62, 63], semiconductors [64],
similar number of sites, as would be or conducting polymers [65]. Controlled
expected from the identical interactions monodisperse deposition of materials at
between the electrodeposited atoms and nanoscopic scales, which has been shown
the substrate in both cases. This obser- to be achievable with electrochemical
vation reaffirms the dominant role of the methods, is necessary for optimal per-
surface|deposit interactions in phase for- formance of a variety of devices and
mation phenomena. processes, such as, for example, quan-
We have presented here some of the tum dots or clean electrochemical energy
fundamental aspects of electrochemical converters.
107
[cm−2]
N0
106
105
0.15 0.20 0.25 0.30

h
[V]
Fig. 10 Number densities of nucleation sites on the surface of vitreous
carbon as a function of the overpotential, obtained from analysis of current
transients obtained during mercury deposition from 0.01 mol dm−3 Hg2 2+
(◦) and 0.01 mol dm−3 Hg2+ (
) aqueous solutions [58].
References 27. G. J. Hills, D. J. Schiffrin, J. Thompson, Elec-

trochim. Acta 1974, 19, 671.
1. E. Budevski, G. Staikov, W. J. Lorenz, Elec- 28. R. McGraw, H. Reiss, J. Stat. Phys. 1979, 20,
trochemical Phase Formation and Growth, 385.
John Wiley & Sons, New York, 1996. 29. D. Walton, J. Chem. Phys. 1962, 37, 2182.
2. W. Thomson, Proc. R. Soc. Edinb. 1870, 7, 63. 30. M. Volmer, Z. Elektrochem. 1929, 35, 555.
3. J. W. Gibbs, Collected Works (Thermodynam- 31. G. Zinsmeister, Vacuum 1966, 16, 529.
ics), Yale University Press, New Haven, 1948, 32. R. M. Logan, Thin Solid Films 1969, 3, 59.
Vol. I. 33. R. Kaischew, B. Mutaftchiew, Electrochim.
4. R. Becker, W. Döring, Ann. Phys. 1935, 24, Acta 1965, 10, 643.
719. 34. S. Toschev, A. Milchev, K. Popova et al., C.R.
5. J. B. Zeldovich, Acta Phys. Chem. USSR 1943, Acad. Bulg. Sci. 1969, 22, 1413.
18, 1. 35. M. Fleischmann, H. R. Thirsk, Electrochim.
6. S. Stoyanov, Thin Solid Films 1973, 18, 91. Acta 1960, 2, 22.
7. A. Milchev, S. Stoyanov, R. Kaischev, Thin 36. F. Lantelme, J. Chevalet, J. Electroanal. Chem.
Solid Films 1974, 22, 255. 1981, 121, 311.
8. G. J. Hills, D. J. Schiffrin, J. Thompson, Elec- 37. B. R. Scharifker, G. J. Hills, J. Electroanal.
trochim. Acta 1974, 19, 657. Chem. 1981, 130, 81.
9. M. Sluyters-Rehbach, J. H. O. J. Wijenberg, 38. B. R. Scharifker in Modern Aspects of Electro-
E. Bosco et al., J. Electroanal. Chem. 1987, chemistry (Eds.: J. O. ’M. Bockris, B. E. Con-
236, 1. way, R. E. White), Plenum Press, New York,
10. B. R. Scharifker, J. Mostany, M. Palomar- 1992, p. 467, Vol. 22.
Pardavé et al., J. Electrochem. Soc. 1999, 146, 39. S. Fletcher, J. Chem. Soc., Faraday Trans. 1
1005. 1983, 79, 467.
11. L. Heerman, A. Tarallo, J. Electroanal. Chem. 40. J. Mostany, J. Mozota, B. R. Scharifker, J.
1999, 470, 70. Electroanal. Chem. 1984, 177, 25.
12. B. R. Scharifker, J. Mostany, J. Electroanal. 41. E. Michailova, A. Milchev, J. Appl. Elec-
Chem. 1984, 177, 13. trochem. 1990, 21, 170.
13. M. Avrami, J. Chem. Phys. 1939, 7, 1103. 42. M. Abramowitz, I. A. Stegun, Handbook of
14. W. J. Basirun, D. Pletcher, A. Saraby- Mathematical Functions, Dover Publications,
Reintjes, J. Appl. Electrochem. 1996, 26, 873. New York, 1965, p. 298.
15. T. Erdey-Grüz, M. Volmer, Z. Phys. Chem. 43. G. Gunawardena, G. Hills, I. Montenegro
1931, 157, 165. et al., J. Electroanal. Chem. 1982, 138, 225.
16. T. Young, Philos. Trans. R. Soc. (Londres) 44. B. Scharifker, G. Hills, Electrochim. Acta
1805, 1, 65. 1983, 28, 879.
17. R. Aveyard, D. A. Haydon, An Introduction to 45. E. Bosco, S. K. Rangarajan, J. Electroanal.
the Principles of Surface Chemistry, Cambridge Chem. 1982, 134, 213.
University Press, Cambridge, 1973, p. 74. 46. P. A. Bobbert, M. M. Wind, J. Vlieger, Phys-
18. M. Volmer, A. Weber, Z. Phys. Chem. 1926, ica A 1987, 146, 69.
119, 277. 47. M. V. Mirkin, A. P. Nilov, J. Electroanal.
19. L. Farkas, Z. Phys. Chem. 1927, 125, 236. Chem. 1990, 283, 35.
20. H. A. Kramers, Physica 1940, 7, 284. 48. M. Tokuyama, Physica A 1990, 169, 147.
21. E. Parzen, Stochastic Processes, Holden-Day, 49. A. Milchev, W. S. Kruijt, M. Sluyters-
San Francisco, 1962. Rehbach et al., J. Electroanal. Chem. 1993,
22. J. Frenkel, Kinetic Theory of Liquids, Claren- 362, 21.
don Press, Oxford, 1946. 50. E. Garcı́a-Pastoriza, J. Mostany, B. R. Schari-
23. E. A. Guggenheim, Trans. Faraday Soc. 1940, fker, J. Electroanal. Chem. 1998, 441, 13.
36, 408. 51. P. Curie, Bull. Soc. Mineral. 1885, 8, 145.
24. R. C. Tolman, J. Chem. Phys. 1949, 17, 333. 52. R. Kaischev, Bull. Acad. Bulg. Sci. Phys. 1951,
25. K. S. C. Freeman, I. R. McDonald, Mol. Phys. 2, 191.
1973, 26, 529. 53. W. S. Kruijt, M. Sluyters-Rehbach, J. H. Slu-
26. J. W. Cahn, J. E. Hilliard, J. Chem. Phys. yters et al., J. Electroanal. Chem. 1994, 371,
1959, 31, 688. 13.
54. L. C. R. Alfred, K. B. Oldham, J. Electroanal. 60. J. Mostany, A. Serruya, B. R. Scharifker, J.

Chem. 1995, 396, 257. Electroanal. Chem. 1995, 383, 37.
55. B. R. Scharifker, Acta Cient. Venez. 1984, 35, 61. P. Bindra, A. P. Brown, M. Fleischmann
211. et al., J. Electroanal. Chem. 1975, 58, 39.
56. A. Serruya, J. Mostany, B. R. Scharifker, J. 62. J. L. Fransaer, R. M. Penner, J. Phys. Chem.
Chem. Soc. Faraday Trans. 1993, 89, 255. B 1999, 103, 7643.
57. A. Serruya, B. R. Scharifker, I. González 63. G. Oskam, P. C. Searson, J. Electrochem. Soc.
et al., J. Appl. Electrochem. 1996, 26, 451. 2000, 147, 2199.
58. A. Serruya, J. Mostany, B. R. Scharifker, J. 64. R. M. Penner, Acc. Chem. Res. 2000, 33, 78.
Electroanal. Chem. 1999, 464, 39. 65. M. F. Suárez, R. G. Compton, J. Electroanal.
59. B. R. Scharifker, J. Mostany, A. Serruya, Elec- Chem. 1999, 462, 211.
trochim. Acta 1992, 37, 2503.
3
1.1 interest in potentiometric ion-selective

Introduction to Electroanalytical electrodes (ISEs) following the massive
Techniques and Instrumentation breakthrough in detection limits to the
pM level [1, 2]. Moreover, potentiomet-
Patrick R. Unwin ric methods are used in some scan-
University of Warwick, Coventry,
ning electrochemical microscopy (SECM)
United Kingdom
experiments, discussed in Chapter 3.3
1.1.1 and some biosensor formats (Chap-
Context ter 2.11), and so it is useful to provide
a brief summary of the basic princi-
Many advances in understanding electro- ples here.
chemical processes, described throughout Conductivity measurements are much
this Encyclopedia, have come about as a less chemically specific and so are not
result of continued developments in tech- considered in detail in this volume,
niques and instrumentation for studying although Chapter 2.11 describes some
electrode reactions and interfaces. This vol- biosensors that operate on conductimetric
ume is concerned with some of the most principles. It is also important to note,
important general electrochemical meth- in passing, the use of conductimetric
ods, and also highlights some of the more methods in analysis [3], particularly for
specialized techniques that are of wide capillary electrophoresis [4, 5] and in ion
applicability in chemistry and its borders chromatography [6].
with other fields. Our understanding of electrochemical
Simple electroanalytical techniques fall methods has also been greatly enhanced
into three groups: potentiometry, con- by the introduction of spectroscopic, struc-
ductivity, and voltammetric/amperometric tural and other probes of electrochemical
techniques. This volume is mostly con- surfaces and systems [7]. This continues
cerned with the third type of method, to be a developing area and consequently
often referred to under the general head- there is little scope within a single volume
ing dynamic electrochemistry. Given the to review all of the techniques avail-
importance of potentiometric methods able, rather the focus is on some of
of analysis, these will be introduced the most commonly encountered struc-
and discussed in this chapter. As men- tural and chemical probes of interfaces
tioned in Sect. 1.1.3, there is renewed and solutions.
4 1 Introduction and Basic Principles
As discussed in more detail in Sect. 1.1.5, phases. Electron transfer from a metal
this volume of the Encyclopedia is divided electrode to a redox active species in an
into three broad sections. The first section, electrolyte solution, and the transfer of
of which this chapter is an element, is con- an ion from an aqueous phase to an im-
cerned with introducing some of the basic miscible organic phase are examples of
concepts of electroanalytical chemistry, in- electrochemical processes. The first case is
strumentation – particularly electronic cir- the most common class of electrochemical
cuits for control and measurements with process and will be discussed in detail in
electrochemical cells – and an overview of most of the chapters in this volume. Ion
numerical methods. Computational tech- transfer processes are, however, of signif-
niques are of considerable importance in icant interest in many areas, for example
treating electrochemical systems quanti- in bioenergetic and cellular processes [9],
tatively, so that experimental data can and in the switching behavior of polymer
be analyzed appropriately under realistic modified electrodes [10], to name but two.
conditions [8]. Although analytical solu- These areas are visited in Volumes 9 and
tions are available for many common 10 of the Encyclopedia and some aspects
electrochemical techniques and processes, are highlighted in the treatments of scan-
extensions to more complex chemical ning electrochemical microscopy (SECM)
systems and experimental configurations and the electrochemical quartz crystal mi-
requires the availability of computational crobalance (EQCM), in Chapters 3.3 and
methods to treat coupled reaction-mass 2.7 of this volume.
transport problems. Although the interest in an electro-
The second section of this volume is con- chemical system is generally on a single
cerned with electroanalytical techniques, interface, it is important to recognize that
starting with the principles of standard an individual interface cannot be studied
voltammetric and amperometric meth- experimentally. It is necessary to investi-
ods, then progressing to more specialized, gate electrochemical cells that involve several
but equally important, experimental ap- types of interface coupled together in se-
proaches that can provide major insights ries. In the simplest form, electrochemical
into electrochemical processes. Finally, the cells involve two electrodes in a single
last section of this volume focuses on electrolyte phase. The difference in elec-
spectroelectrochemistry and surface mi- trical potential between the two electrodes
croscopy techniques. in such a cell can readily be measured,
usually with a high impedance voltmeter,
1.1.2 to ensure that no current is drawn when
Basic Concepts and Terminology the measurement is made. The cell poten-
tial is a measure of the energy inherent
In order to set the scene for some of the in the system to transfer charge between
chapters that follow, it is useful to provide a the two electrodes via an external circuit.
brief overview of some of the terminology The overall cell potential comprises the
and basic concepts needed to understand individual changes in potential that occur
electroanalytical techniques. Electrochem- when crossing each of the interfaces in the
istry is a broad field, encompassing those cell. It is important to recognize that the
processes that involve the passage of change in electric potential when moving
charge across the interface between two a test charge from one conducting phase
1.1 Introduction to Electroanalytical Techniques and Instrumentation 5
to an adjacent conducting phase usually electronic instrumentation are explained

occurs over a short length scale so that in detail in Chapter 1.2 of this volume.
huge electric fields may predominate at the For a two-electrode cell, the net reac-
interfaces in electrochemical cells. There tion comprises two half-reactions, involving
are consequently significant effects on the the processes at the two electrodes. As
structure of solvents in the interfacial re- already mentioned, usually only one of
gion and the ion distribution, as discussed these processes is of interest, and this oc-
further in several of the chapters in Sect. 2 curs at the working electrode in dynamic
of Volume 1 of the Encyclopedia. More gen- electrochemistry experiments or indicator
erally, the magnitude of the potential dif- electrode in equilibrium (potentiometric)
ference at an interface controls both the di- experiments. The other electrode is made
rection and kinetics of electron transfer. It up such that it maintains a constant com-
follows that the control and measurement position throughout the measurement,
of cell potential is an important aspect thus providing a reference potential. The
of dynamic electrochemistry experiments most common reference for aqueous so-
and methods for achieving this through lution is the saturated calomel electrode
the appropriate design and operation of (SCE), depicted in Fig. 1, and this provides
Caps
KCl crystals
Calomel
paste
Saturated
KCl solution
Porous
vycor Hg
Pt wire
connection
Fig. 1 Schematic of a saturated calomel electrode (SCE). Contact

between this reference electrode and the solution is made via the
porous vycor.
a useful example with which to introduce with this. For comparative purposes, tables
some electrochemical notation, terminol- of electrode (or half-cell) potentials have
ogy, and convention. The half-cell reaction been constructed, with the half-cell reac-
for this electrode is tions at unit activity. Following convention,
they are written as reductions. Such in-
−− −
Hg2 Cl2 + 2e−−−
−2Hg + 2Cl (1) dexes [11, 12] are essentially a measure
of the free energy of redox couples, as
The use of saturated (and solid) KCl highlighted below, so allowing one to pre-
in this electrode ensures that the Cl− dict whether a redox reaction of interest
concentration is maintained at a constant will occur spontaneously. Moreover, these
level. Any reduction of Hg2 Cl2 (driving tables provide valuable data to permit the
Eq. (1) from left to right) produces Cl− , but calculation of equilibrium constants and
this cannot exceed the saturated value. A
complexation constants.
shift in Eq. (1) from right to left (oxidation
When constructing a cell, by the com-
of Hg and formation of Hg2 Cl2 ) will tend
bination of two half-cells, the convention
to deplete Cl− , but this is replenished via
adopted is that the right-hand electrode
the dissolution of solid KCl. Note that both
is always considered as a reduction (ca-
Hg2 Cl2 and Hg, as solid and pure liquid,
thodic) process and the left-hand electrode
respectively, have an activity of unity.
is an oxidation (anodic) process. This for-
It is convenient to develop a notation
malism can be illustrated by considering
for electrochemical cells and half-cells. In
the following cell:
the case of Eq. (1), the half-cell would be
represented as Pt|H2 (a = 1)|H+ (aq, a = 1)
Hg|Hg2 Cl2 |KCl(aq, saturated) (2) ||Ag+ (a = 1)|Ag (4)
Here, a vertical line signifies a boundary for which the half-cell reactions are
between two phases. If a phase contains
two or more components, a comma is used right-hand electrode: Ag+ + e −−−→ Ag
to separate them, while a double vertical (5)
line denotes a phase boundary in which the
left-hand electrode: 1
2 H2 − e −−−→ H+
interfacial potential drop is negligible, as
(6)
usually occurs in the case of a salt bridge.
In this case, the cell potential is +0.799 V at
It should be noted that the international
standard reference electrode is the stan- 298 K and the silver electrode is positive.
dard hydrogen electrode (SHE), sometimes This means that the standard potential
referred to as normal hydrogen electrode of the Ag+ /Ag couple has this value, in
(NHE), in which all the components have other words the standard potential for the
unit activity (a = 1): reduction of Ag+ to Ag is +0.799 V versus
SHE. On the other hand, if the cell had
Pt|H2 (a = 1)|H+ (aq, a = 1) (3) been written with the Ag electrode on
the left-hand side, the cell would have a
This electrode is not easy to use practically, potential of −0.799 V, implying that the
but the potential of the NHE electrode is Ag electrode is negative and the oxidation
assigned the value zero and, strictly, all potential of Ag to Ag+ is −0.799 V versus
other electrode potentials are compared SHE. Strictly then, the potential of any cell,
Ecell , is alters the energy of the Fermi level in

the electrode compared with the energy
Ecell = ERHE − ELHE (7) levels of the chemical moieties in the elec-
trolyte (or immobilized on the electrode
where ERHE and ELHE represent the poten-
surface). For example, taking the electrode
tials of the right- and left-hand electrodes,
to negative potentials increases the en-
respectively, written as reduction pro-
ergy of electrons and ultimately a level
cesses in the particular schematic of a cell.
will be attained where the electrons flow
The terminology adopted to describe
from the electrode to vacant states on the
electrochemical cells implies that a sponta-
chemical species, as shown in Fig. 2(a).
neous process has a positive electrode poten-
This causes a reduction current to flow.
tial. Ecell defines the maximum work (free
Conversely, if the electrode is taken to-
energy), G, that a cell can provide:
ward positive potentials, a situation will
G = −nF Ecell (8) be achieved where electrons will transfer
from chemical species in solution to the
where n is the number of electrons in the electrode, causing an oxidation current to
redox process and F is Faraday’s constant flow, as illustrated in Fig. 2(b). The elec-
(96485 C mol−1 ). Under standard condi- trode potentials at which these processes
tions, with all substances at unit activity: occur will clearly be related to the standard
potential of the redox process of interest,
G0 = −nF Ecell
0
(9) while the current flow is governed by the
kinetics of the reaction. These aspects of
0 is the standard cell potential. A
where Ecell dynamic electrochemistry are considered
detailed treatment of electrode potentials further in Sect. 1.1.4. In the next section,
is given in Chapter 1 of Volume 1 of the we first consider the analytical applica-
Encyclopedia. tions of electrochemical cells and potential
In contrast with potentiometry, where measurements at equilibrium.
an electrode potential is measured (with
no current flowing), in most dynamic elec- 1.1.3
trochemistry experiments, the potential of Potentiometry
the working electrode is controlled with
respect to a reference electrode and the In contrast with amperometric electrodes,
resulting current is measured. There are potentiometric electrodes are passive in
exceptions to the latter class of measure- that they do not convert the analyte. The
ments, notably chronopotentiometry [13] main use of potentiometry is to measure
in which a constant current is generated the concentration (strictly activity) of a
and the potential of the working elec- target analyte and so it is useful to
trode is measured, but for the purposes revisit the Nernst equation, relating the
of this introduction it is appropriate to cell potential to activity. For this purpose,
consider simple potential control of the consider a simple redox reaction at the
working electrode. Chronopotentiometry right-hand electrode of a cell, with SHE as
is discussed later in Chapter 2.2 of this the left-hand electrode. The process at the
volume (Sect. 2.2.5). Since the reference right-hand electrode is
electrode potential is fixed, changing the
potential applied to the working electrode −−
υO O + ne−−−
− υR R (10)
Reduction process Fig. 2 Schematic of heterogeneous

Electrode Solution electron transfer for (a) reduction
processes and (b) oxidation processes.
e
− of concentrations:
RT γ νR RT [R]νR
Potential E = E0 − ln RνO − ln
nF γO nF [O]νO
(14)
+ where γi (i = O or R) denotes the
activity coefficient. Under conditions in
(a) which the activities are effectively constant
(for example in systems with a large
Oxidation process
excess of supporting electrolyte), we may
Electrode Solution
introduce the formal potential, E 0 , of the
O/R couple:
−
RT γ νR
E0 = E0 − ln RνO (15)
Potential
nF γO
e allowing us to simplify the Nernst relation:
+
RT [R]νR
E = E0 − ln (16)
(b)
nF [O]νO
It is convenient to express Eq. (16) on a
Here, νi (i = O or R) represents the log10 scale so that
stoichiometric coefficients. The overall cell
2.303RT [R]νR
reaction for the case defined is E = E0 − log10 (17)
n nF [O]νO
υO O + H2 −−−→ νR R + nH+ (11) Evaluating the prelogarithmic term, this
2
shows that the potential of the cell changes
The free energy of the process is [14]
by ±59/n mV (at 298 K) for a decade
aRνR aH
n
+
change in concentration of either the O
G = G0 + RT ln νO n/2
(12) or R species (for νO = νR = 1), which is
aO aH2 the basis of analytical potentiometry.
A practical example of an O/R couple
Since the activities of H+ and H2 in the
would be Fe3+ /Fe2+ in aqueous solution.
SHE are unity, it follows from Eqs. (8 and
To record the potential of such a system,
9) that
one would typically use an inert metal,
RT a νR such as Pt, as the indicator electrode.
E = E0 − ln RνO (13) There are many other examples of
nF aO
potentiometric electrodes, including a
This is the Nernst equation, describing the metal in contact with a solution containing
potential of the O/R couple as a function the corresponding metal ions, as in
of the activities of O and R. We may the case of the Ag|Ag+ electrode. This
also write the Nernst equation in terms electrode, which comes under the classical
term electrode of the first kind, responds to membranes, tailored to respond selectively
the concentration of Ag+ ions in solution. to a specific primary ion [22–28]. Poly-
An electrode of the second kind involves meric membranes are frequently based on
a metal coated in a sparingly soluble plasticized polyvinyl chloride or silicone
salt of the metal, which responds to the rubber. As with the other membrane-based
concentration of the anion of the salt in electrodes described, these act to separate
solution. The most well-known example is an internal solution containing the ion of
Ag|AgCl|Cl− , which provides a simple, but interest at a fixed concentration from an
effective method for chloride ion analysis external solution containing the target ion
in aqueous solution. The SCE, already at an unknown concentration, which is to
mentioned, is another example of this type be determined. Membrane materials have
of electrode. been developed that have appropriate prop-
The pH electrode is perhaps the best erties so that they interact selectively with
known ISE [15, 16], in which a thin glass the ion of interest. The response of these
membrane (50 µm dimension) separates electrodes is governed by two-phase host
an internal solution of fixed pH from an (ionophore)–guest (ion) binding processes
external test solution in which the pH and the associated kinetics and thermo-
is to be determined [17]. The potential dynamics of the binding process [25–28].
across the glass-barrier membrane is The membrane-active recognition compo-
measured with reference electrodes in nent of the membrane can either be an
each phase. The functioning of the glass ion exchanger or a neutral carrier, such
electrode is complicated, with the majority as a macrocyclic compound. A difference
of the glass membrane in a dry state in activity across the membrane produces
and a layer about 50 nm into each a free energy difference and hence poten-
side of the membrane involved in an tial difference:
exchange process of protons for Na+ ,
RT a ext
K+ and Li+ [18, 19]. It follows that these Em = ln iint (18)
ions are the major interferents for pH zi F ai
measurements with this type of electrode,
where aiext and aiint are the activities of
but the effect of each ion depends on the
the ion, i, of interest (charge zi ) external
type of glass used. Conduction through the
and internal to the membrane. The ion-
resistive membrane is thought to occur via
recognition event generates an interfacial
the movement of mobile cations in the
potential at each side of the membrane
silicate glass network.
and it is the difference in these potentials,
The selective membrane may also be
Em , that is measured with two reference
a single crystal, as in the F− ISE [20, 21],
electrodes, one on either side of the mem-
which utilizes LaF3 doped with EuF2 to cre-
brane, as shown schematically in Fig. 3.
ate fluoride vacancies that facilitate ionic
In practice, the activity of the target ion
conduction via F− . This electrode shows
in the filling solution of an ISE is constant,
excellent selectivity with µM detection lim-
so the potentiometric response of a cell
its, governed by the sparingly soluble
comprising an ISE and reference electrode
nature of LaF3 in aqueous solution. The
should obey the relation
only interferent to consider is OH− ion.
The most versatile class of potentio- RT
metric ISEs employ liquid and polymer Ecell = constant + ln aiext (19)
zi F
Fig. 3 Principles of membrane-based

V ISEs. The potential difference between
reference electrodes in the solutions on
each side of the membrane is
determined. In the case shown, the ion
of interest is a cation that interacts
selectively with the membrane.
M
e
m
b
Reference r Reference
electrode a electrode
n
(External) e (Internal)
Mn+ Mn+
External Internal
solution solution
The constant in Eq. (19) includes all of of interest, ai , in the case of one interfering
the contributions to the cell potential that ion type, zi = zj . Clearly, as the product
are invariant and the value is unique Kij aj increases, there is a deviation from
to a particular cell. Equation (19) repre- Nernstian behavior at higher ai and the
sents the ideal situation in which the detection limit of the ISE method suffers
ISE responds selectively to only one type increasingly.
of ion, i. In reality, interfering ions, j The selectivity coefficient of an ISE can
(charge zj ) may have an effect on the po- be measured in several ways [28], the most
tential. Assuming a linear concentration popular of which are as follows:
gradient within the membrane, the Nicol-
1. The fixed interference method: The cell
sky–Eisenman equation [29, 30] is obtained
potential is determined using solutions
for the potential difference:
in which the primary ion activity is var-
RT ied, with a constant level of interferent
Ecell = constant + ion present, until a Nernstian response
zi F
  to the primary ion is seen. The corre-
sponding linear response in this region
× ln aiext +  (20)
z /zj
Kij aj i
is extrapolated to the potential for the
j
background interferent alone and Eq. (20),
where Kij is the potentiometric selectivity in rearranged form, is used to determine
coefficient. Figure 4 shows how an inter- Kij .
fering ion changes the relation between 2. The separate solution method: The cell
the cell potential and the activity of the ion potential is measured for two solutions,
0.00
−0.05
−0.10
−0.15 0.001
−0.20 0.0001
[ V]
−0.25
E
0.00001
−0.30
−0.35 0
−0.40
−0.45
−0.50
−8 −7 −6 −5 −4 −3 −2 −1 0
log10(ai)
Fig. 4 Effect of an interferent ion, activity, aj , on the response of an ISE to an
ion of interest, activity, ai . The charges on the two ions are the same (unity),
for this example, and the curves are labeled with values of the product, Kij aj .
The cell potential is assumed to have a value of zero for unit activity of the ion
of interest.
one containing the primary ion only at a semilogarithmic form. Moreover, if the
known activity and one with the interferent ionic strength of solutions is maintained
only at the same activity. As with the at a fixed level, the Nernst equation can
fixed interference method, the responses be written in terms of concentrations, as
obtained are analyzed using Eq. (20). already outlined in the case of cell poten-
tials, in Eqs. (16 and 17). The logarithmic
Given the reliance of these methods on response of ISEs means that the stabil-
Eq. (20), it is important to note that ity and reproducibility of the cell potential
deviations from this equation have been are of critical importance. Small uncertain-
reported, particularly for mixtures of ties in cell potential can have a significant
ions of different charge [31]. Moreover, it impact on concentration measurements,
has been pointed out that many biased even for an ideally responsive electrode.
values have been reported using these It follows from Eq. (19) that a small un-
methods [28]. This has led to a resurgence certainty, dEcell , in a cell potential results
of interest in the matched potential in a corresponding error in the measured
method [32], which provides a measure of activity (or concentration):
the selectivity of an ISE. Details of the
relative merits of each of these methods RT da
dEcell = (21)
have been summarized in an excellent zi F a
review [28].
In other words, there is a relative error
For analytical purposes, concentration (a/a) in the determination of activity that
rather than activity is usually of inter- depends on the error in the cell potential
est. This can readily be obtained through Ecell :
the use of calibration plots. Given the

logarithmic dependence of cell poten- a zi FEcell
= (22)
tial on activity, calibration plots are in a RT
Thus, an error, Ecell = ±1 mV results in ions at the external membrane surface,

an error of ±4% in activity for a monova- thereby preventing the detection of sub-
lent ion and ±8% for a divalent ion. µM concentrations. By simply lowering
Conventional ISEs typically exhibit a the concentration of the inner electrolyte
Nernstian response over the concentration solution, it has been shown that detec-
range 10−1 –10−5 M, with deviation from tion limits may be extended to the 10 nM
linearity occurring at lower concentrations, level [2].
usually in the range 10−5 –10−6 M. A sig- To achieve pM detection, the leaching
nificant breakthrough in detection limits process was further minimized by using
by up to six orders of magnitude has been an ion exchange (buffering) process in the
achieved by recognizing that a major factor filling solution, involving the primary ion
affecting detection limits is the diffusion and a second ion [1], thereby lowering the
of ions from the inner filling solution, concentration of the primary ion in the
across the membrane into the external ISE. This created a flux of primary ion
analyte solution [1, 2]. Internal filling solu- toward the inner solution, but the ion
tions have typically been of concentration buffering process ensured that the con-
about 1–10 mM to ensure that a constant centration of primary ion within the filling
concentration is maintained in the body solution remained constant. To establish
of the ISE. However, with a much lower that the response of the ISE was not biased
concentration external to the membrane, in the opposite direction, it was important
there is a large driving force for diffu- that the flux in the inward direction was not
sion from the electrode filling solution too high, requiring careful experimental
into the analyte solution. Such leakage design. Figure 5 shows typical potentio-
appears to cause µM levels of primary metric data for a Pb2+ selective electrode of
150
100
50
0
[mV]
EMF
−50
−100
−150
−14 −12 −10 −8 −6 −4 −2

log aM
Fig. 5 Response of a Pb2+ ISE of conventional design (open symbols) and
with an ion buffer as the filling solution (solid symbols); see text for details.
(Data are taken from Ref. [1], with permission from the American
Chemical Society.)
conventional composition (1 : 1 mixture of A basic question when introducing

10−3 M PbCl2 and 0.1 M MgCl2 ) compared dynamic electrochemistry is: what does
to the response for the same membrane, the current, i, that flows represent?
but with an ion buffer filling solution By definition
comprising 1 cm3 of 0.1 M Pb(NO3 )2 in
dq
100 cm3 of 0.05 M EDTA-Na2 (pH 4.34). i= (23)
The calculated activity of Pb2+ in the latter dt
solution was 10−12 M and this clearly ex- where q is the charge passed in unit time, t.
tends the detection limits of the electrode Charge is simply related to the number of
toward the pM regime, compared to the moles of reagent converted by electrolysis,
µM limits of a conventional ISE. Several ex- N , via Faraday’s law:
cellent reviews covering developments in q
this area are available [25–27]. The use of N= (24)
nF
potentiometry in biosensors is considered
further in Chapter 2.11 of this volume. It follows that the rate of reaction is
therefore directly related to the current:
1.1.4 dN 1 dq i
Dynamic Electrochemistry = = (25)
dt nF dt nF
As already introduced in Sect. 1.1.2, dy- Since the electrochemical process occurs
namic electrochemistry involves the appli- at an electrode of area, A, it is appropriate
cation of a potential to a working electrode, to express the rate as a surface flux, j0
with respect to a reference electrode, to (mol m−2 s−1 ):
promote electron transfer at the elec-
1 dN ī
trode–solution interface. We also note, in j0 = = (26)
passing, that a variant on this theme is the A dt nF
use of an interface between two immiscible In this last equation, the current density is
electrolyte solutions as a working electrode ī = i/A.
to promote either a redox reaction or an It follows that in dynamic electrochem-
ion transfer current [33, 34]. Investigations istry the applied potential represents the
of liquid–liquid interfaces have extended driving force for charge transfer (usu-
the range of species that are detectable ally electron transfer) and the current
with dynamic electrochemistry, but will that flows is a measure of the rate of
not be discussed further as charge transfer the reaction. Electrochemical experiments
at liquid–liquid interfaces is considered in of this type are classified as voltam-
detail in Chapter 4.2 in Volume 2 of the metry or amperometry, and several of
Encyclopedia. Further information on the the most important techniques are con-
structure and nature of liquid–liquid in- sidered in subsequent chapters of this
terfaces can also be found in Chapter 2.3 in volume. These techniques differ in the
Volume 1. The use of SECM to investigate form of the potential signal applied to the
liquid–liquid interfaces is highlighted in working electrode, the type of mass trans-
Sect. 3.3.4.2 of Chapter 3.3 in this volume. port regime employed and the current
There are also excellent texts available that response measured. For example, cyclic
cover theoretical and practical aspects of voltammetry, discussed in Chapter 2.1, uti-
liquid–liquid interfaces [35, 36]. lizes a triangular potential waveform with
respect to time and the signal is the may write

current as a function of applied poten-
tial. Step and pulse techniques, considered Eapp = (φWE − φsol ) + (φsol − φref )
in Chapter 2.2, involve sudden jumps in + iRsol (27)
the potential. For large potential pertur-
Eapp = (φWE − φref ) + iRsol (28)
bations, the resulting current is usually
measured as a function of time, but where φWE , φsol and φref are, respectively,
for small potential perturbations superim- the potentials of the working electrode,
posed on a potential ramp, as in differential solution and reference electrode. The
and normal pulse voltammetry, the sig- product, iRsol , is an ohmic term associated
nal is current versus potential. With the with the passage of current through the
advent of digital instrumentation (Chap- finite resistance of the solution.
ter 1.2), the type of waveform applied to the The type of set up shown in Fig. 6 is
working electrode is essentially unlimited only used for situations where low cur-
and there are consequently a huge num- rents are measured (100 nA and smaller
ber of possible dynamic electrochemistry for typical electrolyte solutions), notably for
techniques. experiments with microelectrodes (or ul-
In the simplest situation, a voltammetric tramicroelectrodes), discussed in detail in
circuit involves a measurement cell such Chapter 2.5. The requirement of a small
as that shown in Fig. 6. A potential is current is both to make the ohmic term
applied to the working electrode, with negligible and to ensure that there are
respect to a reference electrode, and minimal changes to the reference elec-
the current is measured. Further details trode composition that would, otherwise,
on the instrumentation for this type of lead to an unstable reference electrode po-
measurement are given in the following tential. The ohmic term may, of course,
chapter (Sect. 1.2.1). The applied potential, also be minimized by making Rsol as small
Eapp , falls sharply across the various as possible, for example, by working with
phase boundaries in the cell and we added supporting electrolyte to increase
Apply E [V] Measure i [A]
Reference Working
electrode electrode
Fig. 6 Simple two-electrode arrangement for dynamic electrochemistry

experiments.
the conductivity of the solution. This also electrolyte or solvent, so that current flows
ensures that migration of any charged an- without the need for a large overpoten-
alytes of interest is negligible, and that tial. A consequence is that the auxiliary
the full potential drop at the working electrode should have a much larger area
electrode/solution boundary (φWE − φsol ) than the working electrode. Gauze mate-
occurs over sufficiently short distances rials, with high surface area, have proved
commensurate with electron tunneling to be particularly popular as auxiliary elec-
(10 Å). trode materials.
For experiments in which the current is To minimize errors in the applied po-
larger, for example, when larger electrodes tential due to uncompensated solution
are used, it is necessary to use a three- resistance, the reference electrode can be
electrode mode (Fig. 7) and make use of a connected to the solution via a tapered
potentiostat. A wide range of potentiostats glass Luggin probe that may be placed
is available commercially [37]. In addition close to the working electrode surface [38].
to the working and reference electrodes, The optimal position is about twice the di-
the set up utilizes an auxiliary electrode. ameter of the probe, which decreases the
The control potential is applied between potential drop due to uncompensated re-
the working and reference electrodes, but sistance seen between the working and ref-
no current flows through the reference, erence electrodes [39]. Although moving
rather this is the role of the auxiliary elec- the probe closer still would decrease this
trode, which provides the current required term further, there are detrimental conse-
by the working electrode without in any quences in terms of shielding the current
way limiting the response. The associated path and perturbing the current distribu-
process at the auxiliary (counter) electrode tion. It is also important that the Luggin
is usually electrolysis of the supporting probe does not perturb mass transport,
Apply E [V] Measure i [A]
i=0
Working
electrode
Reference Counter
electrode electrode
Fig. 7 Three-electrode arrangement for dynamic electrochemistry

experiments.
particularly when convective transport is 3. Chemical reactions, either preceding or

used to deliver material to the surface of the following electron transfer.
working electrode. Even with these precau- 4. Surface reactions such as adsorp-
tions, it is important to note that the ohmic tion, desorption and electrodeposi-
term, due to uncompensated resistance, tion–dissolution.
may not be completely negligible, espe-
cially when the solution comprises resis- The simplest electrode processes involve
tive nonaqueous media. In this case, there only Steps 1 and 2, for example, in the case
are several instrumental methods that can of the redox reaction of the Ru(NH3 )6 2+/3+
be used to compensate, some of which are couple. As already alluded to in Sect. 1.1.2,
discussed in the next chapter (Sect. 1.2.6). electron transfer rates are strongly poten-
Further details on cell and electrode design tial dependent and this topic is discussed
are available [40] and discussed in many of in several chapters in Volume 2 (Chap-
the chapters in this volume. ters 1.1, 1.2 and 1.3). Mass transfer may
We saw from Eq. (26) that the current occur by diffusion, migration and convec-
in a dynamic electrochemistry experiment tion. As discussed above, most dynamic
is a measure of the rate of the electrode electrochemistry experiments are carried
process and to complete this section a out with a large excess of supporting elec-
brief overview is given of the factors that trolyte so that if the reactants and products
can limit the rate of such reactions. The are charged, migration of these species
textbook view [41] of some of the steps that can be neglected [42]. Diffusion is the
might be involved in an electrode reaction key transport step because concentration
is given in Fig. 8. The steps involved are gradients are always involved in elec-
trochemical processes. This is the mass
1. Mass transfer of species between bulk transfer mode assumed for the treatments
solution and the electrode surface. of cyclic voltammetry in Chapter 2.1 and
2. Heterogeneous electron transfer at the step and pulse techniques in Chapter 2.2.
electrode/solution interface. An overview of transport by diffusion and
Electrode surface region Bulk solution
Chemical
Electrode reactions Mass transfer
Oxsurf Oxbulk
Ox′
Desorption
Adsorption
Ox′ads
ne−
Red′ads
Desorption
Chemical
Adsorption reactions Mass transfer
Red′ Redsurf Redbulk
Fig. 8 Typical steps involved in an electrode reaction.

migration is given in Chapter 2.1 in Vol- scanning tunneling microscopy (STM) and
ume 2. Convection may occur naturally atomic force microscopy (AFM) have also
on long timescales and under conditions become important tools for investigating
in which there are significant density these processes, as discussed later in this
and temperature gradients, but such ef- volume (Chapters 2.7, 3.1 and 3.2).
fects are usually undesirable because mass
transport is difficult to treat quantitatively 1.1.5
under such conditions. On the other hand, Outline of This Volume
well-characterized forced convection can
considerably enhance mass transport to Modern electroanalytical chemistry is a di-
electrode surfaces, as in hydrodynamic verse subject covering instrumentation,
electrodes [43], discussed in Chapter 2.4. technique development, modeling, sur-
Such electrodes are beneficial in the study face and interfacial science, sensor de-
of fast heterogeneous electron transfer pro- sign, spectroscopy and microscopy, among
cesses or coupled chemical reactions be- other topics. Consequently, the chapters
cause the slowest process in the scheme of and themes selected for this volume re-
Fig. 8 limits the rate of the electrode reac- flect the broad scope of electroanalytical
tion and the associated current flow. When chemistry. The scene is set in the first
mass transport (or diffusion) limits the rate section, which covers some of the core
of the reaction, the electrode process is said aspects of electroanalytical chemistry. Fol-
to be transport (or diffusion) limited. Con- lowing this chapter is a treatment of
versely, if heterogeneous electron transfer analog and digital instrumentation by Wipf
is the sluggish step in the process, one may (Chapter 1.2), describing the basic princi-
term the reaction to be in a surface-limited ples of control/measurement protocols in
regime. Microelectrodes (Chapter 2.5) and electrochemical systems, including poten-
SECM (Chapter 3.3) have found appli- tiostatic and galvanostatic circuits. Signal
cation in the characterization of rapid transduction, ohmic drop compensation
heterogeneous and homogeneous kinet- and data acquisition are among the topics
ics due to the very high mass transfer that receive particular attention. Chap-
rates available, which has expanded the ter 1.3 by Britz considers the application
timescale of dynamic measurements be- of digital simulation to solve coupled mass
fore transport effects dominate. transport-reaction problems in electroana-
Electrode reactions are often compli- lytical chemistry. This chapter covers the
cated by additional chemical reactions basics of digital simulation, by exam-
(step 3), which may occur either homo- ple, and introduces a flavor for some of
geneously in solution or on the electrode the more advanced simulation techniques
surface. In addition to the techniques that are becoming increasingly popular.
mentioned, the use of cyclic voltammetry An overview of packages available com-
to investigate such processes is consid- mercially and from academic groups is
ered in Chapter 2.1 (Sect. 2.1.2.4) in this also provided.
volume together with several examples Section 2 of this volume is concerned
of these reactions (Sect. 2.1.5). Adsorp- with electroanalytical methods, predomi-
tion and deposition processes (step 4) are nantly connected to dynamic electrochem-
considered extensively in Sects. 3 and 4, istry. The most widely used techniques
respectively, of Volume 1. The EQCM, in electrochemistry are linear sweep and
cyclic voltammetry (LSV and CV), and potential waveform in which there is a
consequently Chapter 2.1, by Speiser, cov- preconcentration step, prior to voltam-
ers the basics of these techniques. After an metric analysis (usually by LSV). The
introduction to voltammetric terminology, preconcentration step needs to be car-
this chapter presents the theoretical back- ried out under conditions of well-defined
ground to LSV and CV, covering mass and efficient mass transport, such as the
transport, heterogeneous electron trans- use of hydrodynamic control (the sub-
fer, coupled chemical reactions in solution ject of Chapter 2.4) or microelectrodes
and surface-bound redox systems. The ef- (Chapter 2.5). Chapter 2.3 covers the main
fects of nonlinear and restricted diffusion, stripping techniques, that is, anodic strip-
compared to planar diffusion, are also ping voltammetry, which is predominantly
highlighted. Practical considerations, in- used for the detection of heavy met-
cluding electrodes, cells, supporting elec- als, stripping potentiometry (introduced
trolyte and solvents, are discussed along in Chapter 2.2, Sect. 2.2.5) and adsorptive
with some of the more common pitfalls stripping voltammetry. The latter method
that might be encountered when making allows the detection of a wide range of
voltammetric measurements. Advances in species, including metals and organic com-
LSV and CV, including the use of micro- pounds that cannot be readily deposited
electrodes (or ultramicroelectrodes), are (preconcentrated) by electrolysis. Practi-
given special attention. This treatment cal considerations when using stripping
complements Chapter 2.5 on microelec- methods are highlighted, including elec-
trodes, summarized below. Applications trode materials, cells and instrumentation.
of LSV and CV are described, including Many of the methods discussed so far
both classical examples and more recent rely on diffusion as the mode of mass
developments. transport. The addition of mechanical con-
Closely related to LSV and CV are vection, as well as diffusion, through the
step and pulse techniques, described in forced movement of the solution or elec-
Chapter 2.2 by Brett and Oliveira Brett. trode can be beneficial for enhancing the
These techniques are widely used in elec- reproducibility of mass transport and de-
troanalytical chemistry and a treatment livering a steady state regime. Chapter 2.4
is provided of discrete step techniques, by Mount is concerned with hydrodynamic
such as single and double potential step electrodes, which make particular use of
chronoamperometry, staircase voltamme- forced convection. The focus is on solid
try and pulse techniques such as nor- electrodes such that the dropping mer-
mal pulse voltammetry, differential pulse cury electrode (and polarography), which
voltammetry and square wave voltamme- is of historical interest in the develop-
try. Current steps are also considered, ment of electroanalytical chemistry as a
including applications in potentiometric subject, is not considered. The interested
stripping analysis, which serves as a link reader should consult Refs. [44, 45], while
to Chapter 2.3 on stripping methods by Ref. [46] provides an excellent historical
Wang. Although more specialized than overview of the development of electro-
the other voltammetric techniques dis- analytical chemistry, including the key
cussed above, stripping analysis is an role of polarography. After an introduc-
extremely sensitive method for trace-level tion to the basic principles, Chapter 2.4
detection. This is achieved by using a reviews the main hydrodynamic methods,
including the rotating disc electrode (RDE) principles and theory together with some
and its ring-disc and optical versions. In of the applications. The instrumentation
the RDE and related devices, solution flow for impedance is described and experimen-
to the electrode surface occurs because of tal details are discussed thoroughly. The
the rotating motion of a disc-shaped elec- main impedance techniques make mea-
trode, which establishes a well-defined and surements averaged over a macroscopic
calculable flow pattern in the solution. Al- sample, but it is also possible to make spa-
ternatively, the solution can be flowed past tially resolved measurements, which are
a stationary electrode, as in the wall-tube, discussed in this chapter.
wall-jet and channel (or tube) electrodes, Several aspects of impedance spec-
which are also reviewed. A recent trend troscopy interface well with the EQCM
has been the development of microhydro- described in the following chapter by
dynamic systems, which are highlighted Hillman (Chapter 2.7). The EQCM is a
briefly. A detailed review of such systems powerful technique for investigating the
is also available [47]. formation and properties of thin films on
As already mentioned several times electrode surfaces. For thin, rigid films,
in this chapter, and elsewhere in this the EQCM serves as a gravimetric sensor
volume, microelectrodes (ultramicroelec- of surface processes. For thicker films, the
trodes) have, in many ways, revolutionized associated viscoelastic properties can be in-
the practice of voltammetry. In Chap- vestigated under electrochemical control.
ter 2.5, Forster describes how microelec- Chapter 2.7 provides a detailed status re-
trodes are pushing the boundaries of port on EQCM, covering the history of
electrochemistry in many ways, includ- the technique, the principles and exper-
ing basic electroanalysis, in applications imental details, including the integration
in biological systems, in artificial noses of EQCM with complementary techniques.
and array devices, and for investigations The diverse range of applications of EQCM
in low conductivity media and the solid are organized by system and discussed
state. The exploitation of microelectrodes in detail.
in the study of ultrafast processes, under An aim of this volume is to highlight
both steady state conditions and with fast rapidly developing areas of electroanalyt-
scan/transient methods, is given particular ical chemistry and electrochemistry. In
attention. The earlier parts of this chapter this context, the application of ultrasound
set the scene by reviewing the properties on electrochemical processes is a topic
of microelectrodes, including mass trans- of particular interest. In a series of three
port effects, while the concluding part of chapters, Compton and coworkers pro-
the chapter speculates on where the field vide a treatment of the underlying phys-
might develop further. ical aspects connected with the coupling
Impedance spectroscopy differs from of ultrasound to electrochemical systems
many of the voltammetric techniques dis- (Chapter 2.8) and applications in electro-
cussed above, such as linear sweep and analysis (Chapter 2.9). The first of these
potential step methods, in that only a chapters considers the effect of ultrasound
small perturbation (usually of sinusoidal on mass transport, on the electrode sur-
form) is applied to the electrochemi- face and on chemical reactions in solution,
cal cell. In Chapter 2.6, Krause discusses while the second chapter looks at the
impedance methods, introducing the basic use of sonoelectrochemical methods in
analysis, notably stripping voltammetries, in passing, that a great deal of ex situ work
developing some of the ideas in Wang’s has been carried out on electrode surfaces,
earlier chapter. The investigation of bipha- but these studies are beyond the scope of
sic systems also receives attention. Tra- this volume. Many of these techniques are
ditionally, one of the major interests in based on ultrahigh vacuum electron and
sonochemistry has been the possibility of ion spectroscopies [49], and consequently
changing the course of chemical reactions the electrode must be transferred from
and synthetic pathways. For completeness, the electrochemical environment to an
Chapter 2.10 therefore considers the im- evacuated chamber prior to analysis.
pact of ultrasound on electrosynthesis, Although special apparatus has been
with a discussion of cell design, and a designed to minimize the exposure of
range of examples of sonoelectrosynthetic the electrode to the atmosphere [50], there
reactions, which mainly originate from are the questions of possible artefacts
the organic chemistry arena. Further in- due to the transfer process [51] and how
formation on organic electrochemistry is realistically the conditions of the solid
provided in Volume 8 of the Encyclopedia, surface in contact with the liquid are
while Volume 7 is concerned with inor- maintained after transfer.
ganic systems. The advent of STM has had a signifi-
Perhaps the area of analysis in which cant impact on surface electrochemistry,
electrochemistry has had the biggest im- providing unprecedented information on
pact on society is in biosensors, no- the structure and dynamics of electrode
tably the glucose biosensor [48]. Although surfaces. STM is a natural technique for
Volume 9 is concerned with bioelectro- electrochemical surface science, since it
chemistry, it is important that this area allows the electron density of metal and
of electroanalytical chemistry is repre- semiconductor surfaces to be imaged with
sented appropriately in Volume 3. Conse- subatomic resolution, while also allow-
quently, Schuhmann and Bonsen provide ing surface processes to be imaged in a
an overview of the physical principles and wide variety of environments, including
applications of biosensors in Chapter 2.11. the solid–liquid interface. Chapter 3.1 by
A comprehensive overview is given of Moffat focuses on the use of in situ STM as
amperometric, potentiometric, conducti- a probe of electrode surfaces. This includes
metric and impedimetric formats for the principles of the STM technique, ex-
biosensors, and the relative merits of each perimental details and a glimpse of some
are fully assessed. Potential new directions of the many applications, ranging from
are highlighted, particularly connected to imaging to surface modification.
miniaturization and multisensor array de- The related technique of AFM uses
tection strategies. the force between a probe and a surface
The final section of this volume is to investigate either structure on the
entitled In situ structural and spectroscopic nanometer scale or interfacial forces. As
probes of electrochemical systems. The six with STM, AFM is much more than
chapters in this section describe some a technique for electrochemistry, but it
of the most important microscopy and has found considerable application in this
spectroscopy techniques that have found area, given that so many electrochemical
application in the study of electrode processes and electroanalytical methods
surfaces and systems. It should be noted, require an understanding of interfacial
structure and properties at high resolution. in Chapter 3.5. This chapter provides a
In Chapter 3.2, Macpherson introduces detailed insight into how in situ in-
the principles of AFM and its application frared spectroscopy may be implemented
to electrified interfaces, covering both and used most effectively to investigate
measurements of surface structure and electrode surfaces. The wide-ranging appli-
interfacial forces. A recent direction has cations of the various infrared techniques
been the integration of microscopic and are highlighted through a series of se-
nanoscopic electrodes into AFM tips, lected examples.
so allowing simultaneous topographical Raman spectroscopy is attracting in-
(force) and electrochemical imaging of creasing interest from electrochemists
surfaces, with high spatial resolution [52]. interested in areas as diverse as elec-
Developments in this area of combined trocatalysis, corrosion, electrodeposition,
SECM-AFM are highlighted. battery technology, and sensor develop-
The last section of Chapter 3.2 leads ment. In Chapter 3.6, Tian and Ren lay
neatly into Chapter 3.3 by Horrocks, which the foundations of Raman spectroscopy
is concerned with SECM. This chapter pro- and describe how it may be used effec-
vides a comprehensive overview of the tively to investigate electrode surfaces in
principles and applications of this tech-
situ. This chapter covers the basic prin-
nique, which continues to attract consid-
ciples and theory at a readily accessible
erable interest from the electrochemistry
level and describes the various experimen-
community and beyond. The main SECM
tal configurations that can be used for
techniques are described along with the
both macroscopic and microscopic inves-
theoretical methods that can be used to
tigations. A wide range of applications
treat the SECM response. Selected appli-
from several areas of electrochemistry,
cations cover solid–liquid, liquid–liquid
and liquid–gas interfaces, with examples and some of the current developments
spanning the life sciences and materi- and future prospects for the technique are
als science. highlighted.
Traditionally, UV-visible spectroscopy
has been the main spectroscopic probe Acknowledgment
of electrochemical systems, covered in
Chapter 3.4 by Crayston. This chapter The author thanks Mr. I. D. Macklam
describes the main techniques, theory for preparing the illustrations for this
and applications in the study of solution chapter.
species (inorganic and organic), thin films
and modified electrodes, along with other References
topics. A wide variety of cells have
been described, to allow the use of UV- 1. T. Sokalski, A. Ceresa, T. Zwickl, E. Pretsch,
visible spectroscopy in electrochemistry, J. Am. Chem. Soc. 1997, 119, 11 347.
and consequently these are described in 2. S. Mathison, E. Bakker, Anal. Chem. 1998,
some detail. 70, 303.
3. F. J. Holler, C. G. Enke in Laboratory
In the past two decades, infrared spec-
Techniques in Electroanalytical Chemistry
troscopy has become a powerful tool (Eds.: P. T. Kissinger, W. R. Heinemann),
for the investigation of electrochemical 2nd ed., Marcel Dekker, New York, 1996,
systems, as described by Christensen pp. 237–265, Chap. 8.
4. K. Mayrhoff, A. J. Zemann, E. Schnell, G. K. 26. P. Bühlmann, E. Pretsch, E. Bakker, Chem.

Bonn, Anal. Chem. 1999, 71, 3828. Rev. 1998, 98, 1593.
5. T. Kappes, P. C. Hauser, Electroanalysis 2000, 27. E. Bakker, P. Bühlmann, E. Pretsch, Electro-
12, 165. analysis 1999, 11, 915.
6. J. Weiss in Ion Chromatography (Eds.: 28. E. Bakker, E. Pretsch, P. Bühlmann, Anal.
C. Dyllick, B. Böck), Wiley-VCH, Weinheim, Chem. 2000, 72, 1127.
Germany, 1995. 29. B. P. Nicolsky, Zh. Fiz. Khim. 1937, 10, 495.
7. G. Jerkiewicz, M. P. Soriaga, K. Uosaki et al., 30. J. W. Ross Jr., Science (Washington DC) 1967,
(Eds.), Solid-Liquid Electrochemical Interfaces, 156, 1378.
ACS Symposium Series, ACS, Washington, 31. E. Bakker, R. K. Meruva, E. Pretsch et al.,
DC, 1996, Vol. 656. Anal. Chem. 1994, 66, 3021.
8. D. Britz, Digital Simulation in Electrochem- 32. V. P. Y. Gadzekpo, G. D. Christian, Anal.
istry, 2nd ed., Springer, Berlin, Germany Chim. Acta 1984, 164, 279.
1988. 33. D. J. Fermin, R. Lahtinen in Liquid Inter-
9. R. B. Gennis, Biomembranes: Molecular Struc- faces in Chemical, Biological and Pharmaceu-
ture and Function, Springer, New York, 1989. tical Applications (Ed.: A. G. Volkov), Mar-
10. R. W. Murray, (Ed.), Molecular Design of cel Dekker, New York, 2001, pp. 179–227,
Electrode Surfaces in Techniques in Chem- Chap. 8.
istry Founding (Ed.: A. Weissberger), Wiley- 34. H. H. Girault, D. J. Schiffrin in Electroan-
Interscience, New York, 1992, Vol. XXII. alytical Chemistry (Ed.: A. J. Bard), Mar-
11. A. J. Bard, J. Jordan, R. Parsons, (Eds.), Stan- cel Dekker, New York, 1988, pp. 1–141,
dard Potentials in Aqueous Solutions, Marcel Vol. 15.
Dekker, New York, 1985. 35. A. G. Volkov, D. W. Deamer, (Eds.), Liquid-
12. G. Milazzo, S. Caroli, Tables of Standard
Liquid Interfaces: Theory and Methods, CRC
Electrode Potentials, Wiley-Interscience, New
Press, Boca Raton, Fla., 1996.
York, 1977.
36. A. G. Volkov, D. W. Deamer, D. L. Tanelian
et al., Liquid Interfaces in Chemistry and
Methods, 2nd ed., Wiley, New York, 2001,
Biology, John Wiley, New York, 1998.
pp. 305–330, Chap. 8.
37. J. Smith, V. Hinson-Smith, Anal. Chem.
14. E. B. Smith, Basic Chemical Thermodynamics,
2002, 74, 593A.
4th ed., Oxford University Press, Oxford,
1991. 38. A. J. Bard, L. R. Faulkner, Electrochemical
15. A. Evans, Potentiometry and Ion Selective Methods, 2nd ed., Wiley, New York, 2001,
Electrodes, Wiley, New York, 1987. p. 27.
16. D. Amman, Ion Selective Microelectrodes: 39. J. Wang, Analytical Electrochemistry, 2nd ed.,
Principles, Design and Application, Springer, Wiley-VCH, New York, 2000, p. 105.
Berlin, Germany, 1986. 40. F. M. Hawkridge in Laboratory Techniques in
17. R. G. Bates, Determination of pH, 2nd ed., Electroanalytical Chemistry (Eds.: P. T. Kiss-
Wiley-Interscience, New York, 1973. inger, W. R. Heinemann), 2nd ed., Marcel
18. F. G. K. Baucke, Ber. Bunsen-Ges. Phys. Dekker, New York, 1996, pp. 267–291,
Chem. 1996, 100, 1466. Chap. 9.
19. F. G. K. Baucke, J. Phys. Chem. B 1998, 102, 41. A. J. Bard, L. R. Faulkner, Electrochemical
4835. Methods, 2nd ed., Wiley, New York, 2001,
20. M. S. Frant, J. W. Ross Jr., Science (Washing- p. 23.
ton, DC) 1966, 154, 1553. 42. A. J. Bard, L. R. Faulkner, Electrochemical
21. J. J. Lingane, Anal. Chem. 1967, 39, 881. Methods, 2nd ed., Wiley, New York, 2001,
22. A. Guindilis, P. Atanasov, E. Wilkins, Elec- pp. 137–155, Chap. 4.
troanalysis 1997, 9, 661. 43. C. M. A. Brett, A. M. C. F. Oliveira Brett,
23. C. Henry, Anal. Chem. 1998, 70, 594A. Comp. Chem. Kinet. 1986, 26, 355.
24. W. E. Morf, The Principles of Ion-Selective 44. J. J. Lingane, Electroanalytical Chemistry, 2nd
Membranes and of Membrane Transport, ed., Wiley-Interscience, New York, 1958.
Elsevier, New York, 1981. 45. A. M. Bond, Modern Polarographic Methods
25. E. Bakker, P. Bühlmann, E. Pretsch, Chem. in Analytical Chemistry, Marcel Dekker, New
Rev. 1997, 97, 3083. York, 1980.
46. A. J. Bard, C. G. Zoski, Anal. Chem. 2000, 72, 50. J. Lipkowski, P. N. Ross, (Eds.), Structure of
346A. Electrified Interfaces, Wiley-VCH, Weinheim,
47. J. V. Macpherson, N. Simjee, P. R. Unwin, Germany, 1993.
Electrochim. Acta 2001, 47, 29. 51. A. J. Bard, L. R. Faulkner, Electrochemical
48. A. E. G. Cass, G. Davis, G. D. Francis et al., Methods, 2nd ed., Wiley, New York, 2001,
J. Am. Chem. Soc. 1984, 56, 667. p. 709.
49. D. P. Woodruff, T. A. Delchar, Modern Tech- 52. C. E. Gardener, J. V. Macpherson, Anal.
niques of Surface Science, 2nd ed., Cambridge Chem. 2002, 74, 576A.
1.2 reduces current flowing through the cell.

Analog and Digital Instrumentation
Icell = I − Iin (2)
David Wipf
Mississippi State University, Starkville, Making Rin much larger than the resis-
Mississippi, U.S.A. tance in the electrochemical cell allows
the source to produce nearly the desired
current. Adequate current sources can
1.2.1 be constructed from high-voltage batter-
Electrical Control ies and a large series resistor. In Fig. 2(a),
a 10-µA current source is made from a
Controlling the flow of electricity in an 100-V battery and a 10-M series resis-
electrochemical cell is fundamental to tor. The rated current is available when
faradaic electrochemistry. Experiments are the load, RL , is a short circuit. This cir-
arranged to control the potential (voltage) cuit can be drawn in an equivalent way
applied across an electrochemical cell or (Fig. 2b), with a 10-µA current source in
to regulate current flow through a cell. A parallel with a 10-M parallel resistor.
DC voltage source can be used to apply These two representations illustrate the
a voltage between the counterelectrode presence of a compliance voltage, that is,
(CE) and the working electrode (WE) as the maximum voltage available for forc-
illustrated in Fig. 1(a). This, and any other ing current through a load. The current
DC voltage source, is considered to consist is only approximately constant for this
of an ideal internal voltage source (Vin ) and source and RL must be 100 times smaller
an internal series resistor (Rin ). Current (I ) than RP to maintain a 1% accuracy on
passing through the cell alters the voltage supplied current.
applied to the cell The limitations of these voltage and
current sources have led to the use
Vcell = Vin − (I × Rin ) (1) of electronic circuitry that allows more
efficient and accurate electrical control. A
brief discussion of the relevant electronic
Rin in battery voltage sources varies from
circuits is required to understand basic
about 35 in a 9-V zinc–carbon battery
electrochemical instrumentation.
to less than 6 m in a lead-acid cell.
Electronically controlled sources produce 1.2.2
very low Rin , an important point to which Electronic Principles
we will return.
IR drop generated by the Rin of a voltage An essential component of modern elec-
source compromises accurate potential trochemical instrumentation is the opera-
control, particularly for high cell currents. tional amplifier (OA or op amp) [1]. An OA
This point led many early investigators to is a high-gain differential amplifier with
prefer controlled current methods. A real an output voltage (Vo ) proportional to the
current source can be considered to consist difference between noninverting (V+ ) and
of an ideal current source in parallel with inverting (V− ) inputs
an internal resistance (Fig. 1b). Current
passing through the parallel resistance Vo = A(V+ − V− ) (3)
1.2 Analog and Digital Instrumentation 25
Icell
Rin Vcell
Rin
Voltage Iin
Vin Current
source
source
(a) (b)
Fig. 1 (a) Voltage and (b) current sources connected to a
two-electrode electrochemical cell.
Fig. 2 Current sources applied to a 10 MΩ

load resistance. (a) Source constructed
from a battery and resistor and (b) an
100 V 10 µA RL RP 10 MΩ RL
equivalent representation of an ideal 10 µA
current source and a parallel source
resistance.
(a) (b)
where A is the open-loop DC voltage gain. inverting and noninverting inputs. This
Typically, an OA has an open-loop DC volt- produces a restoring voltage that forces
age gain greater than 100 000 and input the inverting input voltage to approach the
impedances of 1011 to 1014 . OAs are rep- value of the noninverting input, eventually
resented schematically as a three-terminal stabilizing at equal input and output volt-
device with two inputs and one output ages. In contrast, if the feedback path is
(Fig. 3a). OAs require a power source; how- reversed, so that the output is connected
ever, power connections are often omitted to the noninverting input, unstable posi-
for clarity. The very high open-loop gain tive feedback forces the output away from
of the OA is useful in negative feedback the input.
(closed-loop) configurations. In fact, OAs Under negative feedback, a portion, β,
are rarely operated in an ‘‘open-loop’’ con- of the output is subtracted from Vin
figuration since A varies significantly with (i.e. V+ − V− ). Thus, the output is Vo =
signal frequency and temperature. The A(Vin − βVo ) and the closed-loop voltage
voltage-follower configuration (Fig. 3b) is gain, G = Vo /Vin is
a typical OA circuit. Connecting the output A
to the inverting input forces the output to G= (4)
1 + Aβ
match the noninverting input. This is un-
derstood by considering the consequence For the voltage follower of Fig. 3(b), β is 1,
of an unequal input and output voltage. so the voltage gain is
The high differential gain greatly ampli- A
fies the voltage difference between the G= (5)
1+A
Fig. 3 (a) An operational V− − V− −

amplifier (OA) symbol. (b) An Vo Vo
OA connected as a V+ + V+ +
voltage-follower amplifier. (a) (b)
The high open-loop gain of the OA the output resistance (or impedance) by
increases gain stability. As long as A the factor 1/(1 + Aβ).
is much larger than unity, it can vary Voltage followers are used to ‘‘buffer’’
considerably with little change in G. the output of voltage sources. An un-
Negative feedback also causes the output buffered voltage source is ‘‘loaded’’ when
resistance (Ro ) to be reduced a load causes current to flow from the
source. In Fig. 5(a), a battery and a poten-
Ro
Ro = (6) tiometer are used to construct a variable
1 + Aβ voltage source. In the presence of a load
where Ro is the open-loop output resis- resistance, the load parallels a portion of
tance. Figure 4 illustrates this important the potentiometer, changing the parallel
consequence of negative feedback. A sim- portion of the potentiometer resistance
ple model of an OA is made with an to Rp = Rb RL /(Rb + RL ). This causes the
internal voltage gain block having inter- output voltage to depend on the load, that
nal input and output resistances, Rin and is, Vo = Vin (Rp /(Rp + Ra ) when Rb is not
Ro . In an open-loop configuration, a volt- smaller than RL . Incorporating a voltage
age divider is formed by Ro and the load follower improves the situation (Fig. 5b).
resistance, RL . The voltage divider reduces The follower has a high-input resistance
the output by the ratio RL /(Ro + RL ). and thus does not significantly load the
In closed-loop configuration, the negative potentiometer. The low-output impedance
feedback loop compensates the voltage of the follower permits small RL values
drop caused by current flowing at the (i.e. heavy load currents). Thus, the volt-
output. Essentially, the voltage difference age follower is an amplifier. Allowing a
between inputs, Vd , is adjusted to make very small input current to generate a
Vo equal to Vin , within the limits of the large current through a load produces a
open-loop gain. The net effect is to reduce net power gain.
Rin
Ro Vo
Vd AVd
RL Fig. 4 A simplified model of an

OA-based voltage-follower
Vin amplifier showing input and
output resistance and
internal gain.
− Vo Fig. 5 (a) A variable voltage source

Ra Ra constructed with a battery and
Vo
Vin + potentiometer. (b) An improved variable
Vin
Rb RL voltage source that employs a
Rb RL
voltage-follower buffer to minimize
(a) (b) loading by the load resistance.
Three simplifying assumptions often The circuit in Fig. 6(b) is an invert-

suffice to analyze most OA circuits. These ing amplifier and has an overall gain of
are: (1) the open-loop gain is infinite; Vo /Vin = −Rin /Rf . This circuit is ana-
(2) the input impedance of the OA is infi- lyzed by assuming that negative feedback
nite; and (3) the output impedance of the maintains the inputs at the same potential.
OA is zero. These assumptions are un- With the noninverting input connected to
realistic but do provide good results for circuit common, the inverting input poten-
low-frequency applications. An implica- tial is also maintained at circuit common.
tion of these assumptions is that, under Since the inverting input is maintained at
negative feedback, the output voltage will the common potential but is not connected
swing to precisely the value required to to common, the inverting input is at vir-
make the inverting and noninverting in- tual common (or virtual ground). Circuit
puts equal. common (the inverted triangular symbol),
Applying these assumptions to the volt- rather than earth ground, is used because
age follower predicts that Vo = Vin , the earth ground is not always necessary or
input impedance is infinite, and the out- desirable. For example, there is no direct
put impedance is zero. With large A ground connection in a floating or isolated
values, these predictions are very close circuit. Circuit common is the designated
to the analysis above. A simple varia- voltage reference point and is the ultimate
tion of the voltage-follower circuit pro- sink or source of current of the circuit.
duces the voltage-follower-with-gain circuit Tracking the current flow at the inverting
of Fig. 6(a). R1 and R2 produce a voltage input simplifies analysis of the inverting
divider that sets V− = Vo R1 /(R1 + R2 ). amplifier. A voltage applied to Rin is
With the noninverting and inverting in- dropped across virtual ground, producing
puts equal, Vo = Vin (R1 + R2 )/R1 . Thus, an input current Iin = Vin /Rin . Since no
the follower-with-gain circuit has a voltage current is allowed to enter the inverting
gain, G, equal to 1 + R2 /R1 and an input input, an equal and opposite current,
impedance equal to the input impedance If = Vo /Rf , must flow from the output to
of the noninverting input (i.e. ‘‘infinite’’). the virtual ground of the inverting input.
R1 R2 Rf
Rin
V− − Vin −
Vo Vo
Vin + +
(a) (b)
R1
V1 Rf Rf
R2
V2
R3
V3 − Iin −
Vo Vo
+ +
(c) (d)
Fig. 6 Various OA amplifier circuits: (a) follower-with-gain; (b) inverting;
(c) summing (adding); and (d) current follower.
With Iin = −If , Vo /Vin = −Rin /Rf . The apparent with low-level inputs or high-gain
inverting amplifier has a linear gain, circuits. OAs also do not have infinite input
easily adjusted by Rf , and can deamplify impedances, resulting in nanoampere to
or amplify input signals. The output femtoampere currents flowing in or out of
inversion is rarely a problem because a the inputs. These currents produce voltage
second inverting amplifier can restore the drops across input resistors or offset the
signal polarity. The main difficulty with input current in a current follower. In most
the inverting circuit is that the input cases, circuits can be designed to minimize
impedance is equal to Rin , which may the effects of voltage and current offsets
cause loading of some signals. but these methods will not be discussed
The presence of a virtual ground in here. The most serious deviation from
the inverting amplifier configuration per- the simplifying assumptions made above
mits a number of other useful amplifier occurs when OAs are applied to high-
circuits. The adder circuit of Fig. 6(c) al- frequency signals. Most OAs are designed
lows a scaled summation of a number for low-frequency operation and their use
of voltage sources with Vo = −(V1 R1 + with signals above 100 kHz requires care
V2 R2 + V3 R3 + · · ·)/Rf . Analysis of this and careful component selection. The
circuit follows that of the inverting ampli- effect of frequency on OA response will
fier; each input produces a current flow be discussed in more detail below.
to virtual ground. A current flowing from
the output to virtual ground cancels the 1.2.3
net input current. The current follower of Electronic Electrical Control
Fig. 6(d) produces Vo = −Iin /Rf , where
Iin is from a current source (such as an Using OAs allows construction of potential
electrochemical cell). Note that the mul- and current control circuitry that operates
tiple current sources can be connected more ideally than the simple voltage and
to the inverting input to produce the current sources of Fig. 1 [2]. In voltage
equivalent of the adder circuit, where control applications, the use of an OA
Vo = −(I1 + I2 + I3 + · · ·)/Rf . buffer between a voltage source and the
The simple analysis described above cell replaces the internal resistance of
breaks down in the absence of negative the voltage source with the low-output
feedback. For example, exceeding the impedance of the OA. However, a simple
electrical specifications of the OA prevents extension of the OA buffer permits
negative feedback. A common problem elimination of a large fraction of the
is exceeding the OAs current or voltage solution resistance of the electrochemical
capabilities. Most OAs are low-power cell. This OA circuit is called a potentiostat
devices capable of supplying no more than or voltage clamp.
about 20 mA. In addition, the maximum
output voltage is usually limited to values 1.2.3.1 Potentiostat Circuits
somewhat less than the supply voltages A simple model of an electrochemical cell
(typically ±15 V). Low signal levels also is made with an electrical resistor, con-
challenge the analysis. Offset or bias necting the counter and working electrodes,
voltages at the OA inputs range from representing the electrolyte solution. Flow
several millivolts to tens of microvolts. This of current between these electrodes pro-
produces output errors that are particularly duces an ohmic voltage (I × R), opposing
the applied voltage. As described in Eq. (6) a potential difference (equal to Ec ) between
above, introducing a negative feedback RE and WE.
connection between the amplifier output A potentiostat is a significant improve-
to its inverting input reduces the output re- ment over the simple voltage source of
sistance of an OA. Extending the negative Fig. 1(a). A reference electrode diminishes
feedback loop further, to include a portion the total resistance between counter and
of the solution resistance, reduces ohmic working electrodes by sensing (and thus
drop. A model of this process is illustrated correcting) the voltage near the working
in Fig. 7, where the solution resistance, electrode. This lowers the IR drop in the
Rs , is modeled as two components: the cell. Another advantage is that the voltage
compensated, Rc , and the uncompensated controlling the cell potential, Ec , is not
resistance, Ru . A probe, reference elec- required to be a low-impedance source,
trode (RE), placed between the CE, and allowing a number of different control-
the WE, completes a negative feedback ling sources to be used without regard to
loop between RE and the inverting OA their ability to drive the necessary current
input. This connection reduces Rc by the through the cell.
following amount: The potentiostat requires a third elec-
Rc trode, RE, to act as a sensing electrode.
Rc = (7)
1+A Although the use of a third electrode may
where Rc is the new compensated resis- seem to be a disadvantage, in fact, the
tance and A is the open-loop OA gain. For third electrode is important beyond its
example, if Rc is 1 k, Rc is 0.010 , with role in reducing the cell resistance. A two-
A = 105 . As Rc becomes negligible, Rs is electrode voltage source requires that the
reduced to Ru . The potential at point RE CE serve both as an RE and as a source or
is now equal to Ein (inverting and nonin- sink of the electrical current flowing at the
verting inputs have equal potentials under WE. In order to supply the current with-
negative feedback) and thus the potential out unduly perturbing their equilibrium
applied across Ru is Ein . Drawing this half-cell potential, CEs must be physically
same circuit connected to an electrochem- large. In a three-electrode cell, controlled
ical cell illustrates a simple potentiostat by a potentiostat, the CE function is split
circuit (Fig. 8). The potentiostat maintains between two electrodes: a large surface
Ein + CE Rc Ru WE
Fig. 7 Using a voltage follower
to eliminate the effect of Rc in a −
simple electrical model of an RE
electrochemical cell.
Ec +
− AE
I
Eout = IRm
RE WE
I Rm
Fig. 8 A potentiostat based on
a voltage-follower circuit.
area auxiliary electrode, AE, which supplies inputs are required when compared with
the cell current, and a reference electrode, the follower potentiostat. As in the follower
RE, which maintains a cell potential refer- potentiostat, negative feedback is used to
ence. The RE is not required to supply any reduce cell resistance and the buffered
significant current to the high-impedance RE voltage again allows small reference
input of the OA. In consequence, the RE is electrodes to be used.
smaller, conserving space and cost. Thus, Using either the follower or adder po-
three-electrode potentiostatic cell control tentiostats does permit a very serious fault
is advantageous even when ohmic drop is condition to exist. Negative feedback is
not an issue. only present when AE and RE are in elec-
However, potentiostatic circuits are not trolyte solution. Removing either of the
always necessary. A common situation is electrodes from solution or disconnecting
found in the use of ultramicroelectrodes, them from the potentiostat prevents the
UMEs, in which the cell current is in stabilizing effect of negative feedback. An
nanoamperes and smaller due to the electrocution hazard exists when the AE
small size of the WE [3]. With such small connection is not present. The lack of neg-
currents, even small reference electrodes ative feedback will force the potentiostat
are used as counterelectrodes without fear into saturation and voltages from ten to
of significant perturbation of their half-cell several hundred volts will appear on the
potential. The potentiostat of Fig. 8 can AE lead (depending on the power output
accommodate this case by connecting both capability of the potentiostat). In addition,
the AE and RE leads together. Note that the absence of potential control allows sig-
this converts the potentiostat into a simple nificant current flow between RE and WE.
voltage follower. UMEs are considered fur- RE can thus be greatly perturbed from
ther in Chapter 2.5 and Sect. 2.1.4.1 of equilibrium perhaps ruining the reference
Chapter 2.1. electrode. Alternately, disconnecting or re-
A different type of potentiostat circuit is moving the RE from solution, produces
constructed by an adder-type OA circuit large potentials between AE and WE as
(Fig. 9) rather than the follower type the potentiostat goes into saturation. At
of Fig. 8. The adder (or summing) OA the least, this can ruin an experiment or
input allows multiple control inputs; for damage the WE. To avoid this, most po-
example, an offset potential may be added tentiostats are equipped with a switch to
to a separate voltage ramp. Since the adder turn the cell potential control off between
inputs are not high impedance, a voltage experiments or during setup. However, se-
follower buffers the output of the reference curely fastening cell leads and electrodes
electrode. Also, note that inverted control is a safe practice, as well as disconnecting
R
Ec −
C
R + I
Eout = IRm
+
R I Rm
ERE = −EWE −
Fig. 9 A potentiostat based on
a summing amplifier circuit.
and connecting the WE first and last, the supply voltages (about 13–15 V for the
respectively, in any experiment. typical ±15 V supplies). The voltage out-
put of the OA is Icell × RLoad and this can
1.2.3.2 Galvanostatic Circuits be used to verify correct operation of the
Galvanostats are instruments that provide galvanostat; outputs near the compliance
a controlled current through an electro- voltage indicate impending loss of negative
chemical cell. As in the potentiostatic cir- feedback control. RLoad is not necessarily
cuits, negative feedback produces superior ‘‘ideal’’. Such is often the case when the
current control. A galvanostat constructed generalized load is replaced with an elec-
from an OA is illustrated in Fig. 10. The trochemical cell, which presents a load that
current flowing through the cell, Icell , is varies nonlinearly with potential and time.
The load terminals are the working and
Ein auxiliary electrodes. By connecting the WE
Icell = −Iin = − (8)
Rin to the inverting input (virtual ground), the
Conveniently, the input voltage programs WE potential can be monitored by measur-
the magnitude and polarity of the cur- ing the potential of RE versus ground. The
rent. Changing the value of Rin scales reference electrode is not necessary for op-
the current range. Analyzing the operation eration of the galvanostat but, since the
of the circuit is straightforward. Negative potential of the working electrode is often
feedback holds the inverting input at vir- desired in constant-current experiments,
tual ground by matching the current flow is often included. Controlled current ex-
across the load to the opposite current periments are not affected by IR drop (as
produced by dropping Ein across Rin . As long as the galvanostat has sufficient com-
long as negative feedback conditions can pliance voltage to overcome the IR drop)
be maintained, the current will be main- but the measurement of WE voltage will
tained at a value programmed by Ein and include an IRu drop component. Proper
Rin . Physical limitations of the OA prevent placement of the reference electrode will
certain current flows through the load. An minimize this but an advantage of gal-
open-circuit condition cannot support a vanostatic operation is that the measured
current flow and the OA itself can only potential is corrected by simply subtracting
supply a limited amount of current (about the IRu voltage.
10–20 mA for typical devices). The voltage A difficulty with the galvanostatic circuit
required to produce the current (the com- of Fig. 10 is that the programming voltage
pliance voltage) is limited to values near Ein must supply the identical magnitude
−EWE
ILOAD
LOAD
Fig. 10 A programmable
Iin
current source. Replacing LOAD Rin
−
with an electrochemical cell
Ein
converts the current source to a +
galvanostat.
R
+ − RS
Ein R D B −EWE
− +
R +
S
−
Fig. 11 An improved galvanostat circuit.
of current flowing through the load (cell). 1.2.4

This can be a problem for heavy current Signal Transduction
applications. The galvanostatic circuit in
Fig. 11 avoids this problem. The output A potentiostat’s function is to maintain a
current is programmed by Ein and a constant potential between the WE and the
‘‘sense’’ resistor Rs in which RE. In most cases, a potentiostat includes,
as a separate function, a means to con-
Ein vert the current flowing through the cell
Icell = (9) into a signal measurable by an external
Rs
recorder. The current is transduced into
In operation, the current flow through a voltage and often further amplified and
filtered before presentation at the output
the cell (or load) produces a voltage drop
terminal. A classical transduction method
across Rs , which is buffered by the sense
is to simply measure the voltage developed
amplifier (S). The output of S is applied to
across a resistor. A measuring resistor,
the inputs of the differential amplifier (D).
Rm , inserted between WE and common
A ‘‘buffer’’ amplifier placed in the negative
(cf. Figs. 8 and 9) generates a voltage ac-
feedback loop provides greater current
cording to Ohm’s law, Vo = I Rm . Note
capacity. Buffer amplifiers, of the type used
that Rm is in series with the solution re-
here, can produce high-output currents, as sistance and the total ohmic drop in the
much as several amperes. Since negative circuit is I × (Ru + Rm ). Thus, a disad-
feedback is operating in the differential vantage of this circuit is that Rm must
amplifier feedback loop, the inverting and be carefully matched to the expected cell
noninverting inputs are equal, and the current to avoid excessive ohmic drop. An
output of the buffer amplifier is precisely alternative procedure is to insert Rm in
that required to maintain the programmed the CE as illustrated in Fig. 12. Rm is in
current through Rs and, thus, through the series with the cell solution resistance but,
cell. For good results, the voltage drop in this case, Rm is included in the com-
across Rs should be small compared to the pensated part of the solution resistance
compliance voltage of the galvanostat. By and does not contribute to the overall
using different values of the sense resistor, cell ohmic drop. In this method, neither
currents of nanoamperes to amperes terminal of Rm is connected to circuit com-
can be controlled and programmed by mon (i.e. it is ‘‘floating’’). A differential
Ein . OA is used in this situation to measure
Fig. 12 Current measurement

by use of a differential amplifier Rm
to measure the voltage drop
across Rm connected to the
R R
auxiliary electrode. Other
control circuitry is not shown. R
+ Eout = IRm
R
a potential difference across a floating configuration, the WE potential is main-

source. However, a differential OA is not tained at ‘‘virtual ground’’. A disadvantage
suitable when the resistor terminals are at of this circuit is that loss of negative feed-
several tens of volts, or higher, above cir- back and potential control occurs when the
cuit common (as can occur in high-power amplifier’s electrical capacity is exceeded.
potentiostats). In such cases, measuring This is especially a problem for heavy cell
the small differential voltage across the currents, where the OA must supply an
resistor in the presence of a much larger equal current to maintain virtual ground.
‘‘common-mode’’ voltage requires use of The addition of a buffer amplifier will allow
an instrumentation amplifier. Instrumenta- operation with cell currents larger than the
tion amplifiers are designed to accurately OA capabilities. Potential control is also
extract the differential voltage while reject- lost when the combination of cell current
ing common-mode voltages five orders of and Rf would require an output voltage
magnitude larger. The figure of merit is the greater than the power supply voltage of
common-mode rejection ratio (CMRR), often the OA. In addition, virtual ground is lost at
specified as decibels (i.e. 100 dB is a 105 high frequency because of the inadequate
rejection). Although most instrumentation high-frequency gain of OAs. High-gain cir-
amplifiers do not allow the common-mode cuits can lose control at frequencies as low
voltage to be higher than the amplifier sup- as 100 Hz. A more detailed discussion of
ply voltage, several allow common-mode the frequency response of potentiostatic
voltages of up to 200 V, permitting use on circuitry is presented in Sect. 1.2.7.
high-power potentiostats.
The current-follower OA configuration 1.2.5
(Fig. 6d) is often used as a current trans- Alternate Control Circuits
ducer in potentiostat circuits (Fig. 13). The
output voltage is proportional to the feed- In some experiments, such as a rotat-
back resistor, Rf , making amplification of ing ring-disk electrode (see Section 2.4.2.2
the current signal straightforward. In this of Chapter 2.4) or liquid chromatography
Rf
−
Fig. 13 Current measurement + Eout = −IRf
using a current-follower circuit.
detection, it is desirable to control two or amplifier that generates this potential

more working electrodes in the same elec- difference, which, when applied to the
trochemical cell. Although separate poten- noninverting input of current-transducer
tiostats can be used to control the working amplifier OA4, shifts the potential of WE2
electrodes, several problems occur. The to a value equal to EW2 − EW1 . Amplifier
control and measurement functions of the OA4 transduces the WE2 current. Since
potentiostats must be electrically isolated. the voltage output of OA4 is not referenced
This requires isolated power supplies (e.g. to circuit common, differential amplifier
batteries), control signals, and outputs. In OA7 is used to restore the output to a
addition, each working electrode requires a circuit common reference. In the bipoten-
separate auxiliary and reference electrode. tiostat, the auxiliary electrode supplies the
An alternate strategy is to use a bipoten- total current required at WE1 and WE2;
tiostat; a potentiostatic circuit designed however, OA2 and OA1 act to maintain
to allow simultaneous potential control the potential difference between the first
of two working electrodes. A simplified working and RE at EW1 . If desired, addi-
schematic of a bipotentiostat circuit is il- tional working electrodes can be added to
lustrated in Fig. 14. The circuit is based this circuit by duplicating the circuits of
on the adder-type potentiostat (Fig. 9). The OA4–OA7 as many times as necessary.
addition of amplifiers OA4–OA7 provides An alternate method for construction of
control of the second working electrode. multipotentiostats is to use a grounded ref-
The bipotentiostat maintains working elec- erence configuration (Fig. 15). Here the RE
trode 1 (WE1) at circuit common, while is maintained at virtual ground and the po-
WE2 is lifted off circuit common by tential of the working electrode is changed
an amount equal to the potential differ- at the noninverting input of the current
ence between the two working electrodes transducer. Since the output of the current
(i.e. EW2 − EW1 ). OA6 is the differential transducer floats above circuit common by
Rf1
Ec R
−
OA1 −
R + AE OA3
+ Eout = −IRf1
RE WE1
+
OA2 WE2 Rf2
ERE = −EWE1 − R
− R
R OA4 −
R + R OA7
R + Eout = −IRf2
R −
− R OA6 R
EWE2 in OA5 +
+
R
Fig. 14 A bipotentiostat circuit. Circuitry enclosed by the dashed line is that added to support a
second working electrode.
WEx = Rf
R
−
WE1 − R
+ −
WE2 + R
+ Eout = −IRf
WE3 R
...
Ew in
WEn
Fig. 15 A multipotentiostat circuit based on a grounded reference configuration. The

dashed line surrounds the circuitry required for each working electrode.
Ein , the output is referenced to common by 1.2.6

a differential amplifier. Additional work- Ohmic Drop Compensation
ing electrodes are added by replicating
the current-transducer/differential ampli- In galvanostatic or amperometric ex-
fier combination. All working electrodes periments, the solution, electrode, and
share the auxiliary and reference electrodes connection resistance produce an ohmic
in this configuration. voltage in proportion to the cell current. In
In bipotentiostats employing shared ref- galvanostatic measurements, ohmic drop
erence and auxiliary electrodes, the degree distorts the measured cell voltage. How-
to which the solution resistance can be ever, a more serious effect occurs in
eliminated by RE placement is limited. voltammetry in which the ohmic drop op-
In such cases, the RE should be located poses the applied potential. Voltage sweeps
nearest the WE with the heaviest cur- become nonlinear and potential steps have
rent flow. Obviously, the operation of prolonged rise times producing a distorted
these bi- or multipotentiostats will be voltammetric response. There are many
compromised under conditions in which reports of methods to alleviate ohmic drop
ohmic drop is large. Experimental arti- in the literature [4, 5]. The use of a po-
facts are introduced when the loop does tentiostat provides a partial solution by
not instantaneously respond to changes in employing a negative feedback loop to
current at the working electrode. In par- eliminate a fraction of the solution resis-
ticular, potential steps at either working tance. Numerous methods to minimize
electrode will introduce spurious current the remaining ‘‘uncompensated’’ resis-
transients at the other electrode due to tance (Ru ) have been proposed [6–9], but
the finite response time of the feedback the two most commonly used compensa-
loop. The transients not only obscure tion procedures are positive feedback and
the current signal but also indicate a current-interruption compensation.
brief loss of potential control. Using dual
potentiostats is not necessarily a solu- 1.2.6.1 Current Interruption
tion to this problem, since large currents In the current-interruption method, the
flowing from either of the two auxil- flow of current through the cell is periodi-
iary electrodes to closely spaced working cally stopped. During this time, there is no
electrodes will cause a similar loss of po- ohmic drop in the cell (I R = 0); therefore,
tential control. the true potential of the working electrode
CST
R Cf
−
Rf
R +
SC
ERE + −
S/H
− +
SHC
SC
SHC
ERE
Fig. 16 A simplified schematic and timing diagram of a

current-interruption potentiostat used for IR drop elimination. The
FET switch is closed and the S/H is in sample mode when digital
control signals SC and SHC are high.
can be measured [10]. Figure 16 illustrates feedback loop maintains the potential be-
the basic instrument. An electronic switch tween the WE and reference at the control
[typically a field-effect transistor (FET) potential (Ec ). Unlike the simple follower
switch] at the auxiliary electrode is dis- potentiostat, however, the potential is
connected for several microseconds at a controlled without ohmic drop distortion.
frequency of 10 to 50 kHz. At the moment It should be noted that implementation of a
of disconnection, the ohmic potential properly performing current-interruption
rapidly drops to zero (the drop is not in- potentiostat is significantly more complex
stantaneous because of the cell inductance than that suggested by the illustration in
and the finite response time of the ampli- Fig. 16 [11–13]. Since the potential is not
fiers). After a delay to allow the ohmic drop controlled continuously, the potentiostat
potential to collapse, a pulse signal is sent feedback loop must operate more slowly
to a sample-and-hold (S/H) circuit, which than otherwise. By adding a compensation
monitors the RE potential. The double- capacitor (CST ) in the feedback loop, the
layer capacitance maintains the working frequency response is reduced to a value
electrode potential until it eventually de- lower than the pulse frequency. Without
cays through faradaic reactions. An S/H this, the interruption pulses will destabi-
circuit collects a sample of the potential lize the feedback loop. In addition, the
using a set of switches to charge a low- current follower filters the cell current by
leakage capacitor during a sampling pulse. the combination of Cf and Rf , which elim-
At the pulse end, the capacitor value is out- inates transients caused by the current
put through a buffer amplifier to ‘‘hold’’ switch. Although a higher pulse frequency
the sample potential until another sample is desirable, increasing the pulse frequency
pulse occurs. The S/H output is applied to beyond 100 kHz is practically difficult.
the inverting input of the potentiostat OA. This limits this technique to relatively slow
As in the follower potentiostat, the negative measurements. The current-interruption
technique is well suited for corrosion, offsets the ohmic drop. It is important
battery, and electrosynthesis applications to realize that the correction occurs con-
that often do not require rapid measure- tinuously. A one-time correction will not
ments and in which heavy current flows suffice since the current flowing is that
introduce unacceptable ohmic drop in observed with ohmic drop present. For full
even highly conducting electrolytes. correction, the current used for correction
must have no ohmic drop. A continuous
1.2.6.2 Positive Feedback correction is able to approach this ideal but,
This method introduces a positive feed- out of necessity, the correction signal will
back path into the potential control circuit. always lag the cell current, preventing com-
Normally, positive feedback is not desir- plete correction. For this reason, any delay
able but is used here in conjunction with in the correction loop reduces the amount
negative feedback such that the overall of correction possible and, as discussed
amount of feedback remains negative. An below, can destabilize potential control.
illustration of an adder-type potentiostat The compensation fraction amount is
incorporating positive feedback is shown given by f × Rf /Ru . Although complete
in Fig. 17. A fraction, f , of the voltage compensation is not possible for the
output (created by the potentiometer) is reasons given above, the amount of com-
summed with the RE potential. (Note that pensation actually achieved in practice
the potentiometer resistance should be is surprisingly low – 80% compensation
much less than the resistance, R, at the is typical. Delays introduced in the cur-
inputs of the control amplifier to avoid rent transduction and in measuring the
loading the potentiometer output. Alter- RE potential are the main reason for in-
nately, a voltage follower can by inserted complete compensation. Positive feedback
after the potentiometer.) Because of the in- also requires knowledge of Ru . Although,
verted current-follower output, the signal Ru can be determined before an experi-
is fed back as positive feedback. Positive ment in many cases, some experiments,
feedback reduces the uncompensated re- notably corrosion, present situations in
sistance since the amount of ohmic drop which the solution or electrode resistance
is proportional to the cell current. Thus, varies with time. A number of authors
adding a voltage proportional to the cell have presented methods to automatically
current to the auxiliary electrode potential measure Ru during an experiment [6,
R −
Rf
Ein +
R
−
− Eout = −IRf
R +
+
f Rf/Ru
Fig. 17 Potentiostat circuit with positive feedback IR drop correction.

7, 9]. Because of the relatively simple potential or current control. To illustrate

implementation, positive feedback is a this, a test circuit (Fig. 18) was modeled
popular method for reducing ohmic drop. using a SPICE-based electronic simula-
Positive feedback can also operate more tor [17].
rapidly than current-interruption methods Initially, the simulation uses the circuit
and so is useful in analytical or physical components noted in the figure and this
measurements. will serve as a comparison point. A triangle
wave generator produces a cyclic voltam-
1.2.7 metry (CV) wave at 2000 V s−1 as the input
Control Loop Frequency Response for the potentiostat. OA1 is a common low-
cost LF356 (National Semiconductor) de-
The behavior of the control loop was vice and stabilizing capacitor CST is 10 pF.
previously considered at low frequencies; The electrochemical cell is simulated as a
however, the ability to control the poten- dummy cell with 1200 of solution re-
tial or current at higher frequencies is sistance, a double-layer capacitance, Cdl ,
important in many electrochemical meth- of 0.2 µF, and faradaic impedance, ZF , of
ods, including ac impedance and short- 106 . The reference electrode is situated
timescale voltammetry. Although the be- to provide Rc = 1000 and an uncom-
havior of a potentiostat or galvanostat at all pensated resistance, Ru of 200 . RRE and
frequencies can be determined through ex- CRE are included to simulate the behavior
periment or calculation for a well-defined of a reference electrode. A Luggin–Haber
system (such as when connected to a capillary, used to provide for close RE–WE
dummy cell) [14–16], in practice real elec- spacing, often exhibits high resistance and
trochemical cells are not only electrically capacitance. Initially, RRE and CRE are set
complex but also extremely variable. Deter- to 10 k and 1 pF, respectively. OA2 is
mining the behavior of a general-purpose simulated as an ideal OA. The current
instrument is impractical over the wide follower, OA3, is an LT1007 (Linear Tech-
range of conditions that might occur nology) OA, a higher performance OA
under normal operation. However, it is than the LF356 OA. At 2000 V s−1 , the
possible to determine with some confi- maximum double-layer charging current
dence what conditions might cause loss of is 400 µA and the 200 Rf value produces
Stabilizing Cf Filter &

CST network gain
WE
Rf
interface
Ein 10 kΩ Cdl
− Rc Ru
OA1 − −IRf
10 kΩ
10 kΩ
+ ZF OA3
RRE +
− CRE
OA2 iR compensation
+ Rpf
Positive
RE
feedback
Fig. 18 Model potentiostat and electrochemical cell circuit used in stability calculations.
0.20
0.15
0.10
0.05
[A] 0.00
i
−0.05
−0.10
−0.15
−0.20
0.0 0.2 0.4 0.6 0.8 1.0
E
[V]
Fig. 19 Calculated CV curves with a 2000 V s−1 scan rate applied to the
model potentiostat of Fig. 18. ( ) Rc = 1000 , Ru = 200 ,
CST = 10 pF, Cdl = 0.2 µF, RRE = 10 k, CRE = 1 pF,
Rf = 200 , Cf = 1 nF; ( . . . . ) Rc = 1198 , Ru = 2 , CST = 10 pF;
(-.-.-.) Rc = 1198 , Ru = 2 , CST = 1000 pF; (- - - - ) Rc = 1000 ,
Ru = 200 , CST = 1000 pF.
an output voltage of 80 mV. A connection because OA1 is unable to supply sufficient

for positive feedback ohmic compensation current to charge Cdl at the rate demanded
is included. Initially, the potentiometer, by the negative feedback loop. Since the
Rpf , is set to give zero positive feedback. OA is unable to supply current at the
required rate, there is a delay introduced
1.2.7.1 Effect of Reference Position into the control loop. The bode and phase
For Ru of 200 , the CV current in Fig. 19 shift plots of Fig. 20(a, b) plot the behavior
shows that the charging current does not of a series of sine waves applied to the input
level off to the expected 400 µA until of the circuit. The result is a log–log plot of
about 200 µs into the wave. This slow rise loop gain and phase shift versus frequency
is caused by the ohmic drop or, more for this circuit. The amount of phase shift,
precisely, the Ru Cdl time constant of 40 µs. with respect to the low-frequency behavior,
Moving the reference electrode closer to is a measure of delay in the feedback loop.
the working electrode decreases Ru . A At low frequencies, the Bode plot indicates
simulation with Ru = 2 (Rc = 1198 ) that the magnitude of the gain is 1 (0 dB).
might be expected to show a 100-fold At higher frequencies, the voltage gain
speed-up in the rise time. In fact, a drops to about 70% (−3 dB) of its DC value
set of damped oscillations (ringing) is for Ru = 200 at a frequency of 270 kHz.
observed on the rising and falling edge of This drop is mainly due to the inherent
the charging current. The oscillations are frequency response of the OA. For clarity,
evidence of destabilization of the potential only the phase shift in excess of the 180◦
control by decreasing Ru . The control loop shift produced by negative feedback is
is unable to maintain the desired potential plotted in Fig. 20(b). For Ru = 200 , the
20
0
dB
−20
−40
(a)
30
60
Phase shift
90
120
150
180
1000 10000 100000
Log frequency
(b) [Hz]
Fig. 20 Calculated (a) bode and (b) phase shift curves for the
circuit of Fig. 18. Conditions are given in Fig. 19.
phase shift only becomes appreciable in decreases in Ru cause increased ringing

the region in which the loop gain is already duration and amplitude.
decreasing. However, in the case in which Stabilizing the control loop to avoid
Ru is 2 , the phase shift rapidly grows oscillation is possible in several ways.
to 90◦ near a frequency of 32 kHz. The A recommended procedure to increasing
Bode plot also shows a peak in the gain the stability of amplifiers driving capaci-
at the same frequency. The nature of the tive loads is the introduction of a small
ringing in the CV in Fig. 19 is now clearer: resistance in series with the load. The
at the frequency at which the phase shift benefit of this is clear from the figures
increases to 90◦ and the gain is larger above: a too-small Ru value destabilizes
than unity, oscillations (at 32 kHz) caused the loop. Some amplifiers are specified
by positive feedback are possible. Further to drive larger capacitive loads and would
be preferred here. An alternate method low-pass filter. This adds an additional

is to reduce the loop gain below unity at delay in the loop, producing oscillations
the oscillation frequency. Changing CST at 50 kHz, as indicated by the peak in
to 1000 pF stabilizes the potentiostat with the Bode plot at the 90◦ phase shift
Ru = 2 . Ringing is reduced in the CV (Fig. 22).
(Fig. 19). The Bode and phase shift plots
(Fig. 20) show that a CST of 1000 pF re- 1.2.7.2 Positive Feedback Stability
duces the gain below unity for frequencies Addition of positive feedback for IRu com-
at which the phase shift is greater than pensation produces a control situation
90◦ . Note that this reduces the overall similar, but not identical, to condition
frequency response of the potentiostat to in which Ru is reduced by moving the
below 16 kHz. This is clearer in the CV RE closer to the WE. The bode plot and
and bode plots in which Ru is 200 and CV curve of Fig. 23(a, b) illustrate the
CST = 1000 pF; the CV response is clearly potentiostat response when Ru = 200
affected by the slower rise time of the (Rc = 1000 ) and the compensation frac-
potentiostat circuitry. tion is set to 0, 90, and 99%, that is,
The RE itself can also destabilize the an effective Ru value of 200, 20, and
control loop. Figure 21 shows the CV 2 , respectively. Compare these curves
acquired under the initial conditions but to Figs. 19 and 20, which have actual
with RRE = 50 k, CRE = 300 pF. These values of Ru of 200, 20, and 2 . In
values are reasonable for a long, thin addition, the effect of a 1000 pF stabi-
Luggin–Haber capillary. The effect is to lizing capacitor (CST ) is plotted for an
reduce the frequency response of the RE effective Ru of 2 . Except for one ex-
connection, that is RRE and CRE for a ception, the CV curves of Figs. 23(b) and
0.20
0.15
0.10
0.05
[A]
0.00
i
−0.05
−0.10
−0.15
−0.20
0.0 0.2 0.4 0.6 0.8 1.0
E
[V]
Fig. 21 Calculated CV curves with a 2000 V s−1 scan rate applied to the
model potentiostat of Fig. 18. ( ) Rc = 1000 , Ru = 200 ,
CST = 10 pF, Cdl = 0.2 µF, RRE = 10 k; CRE = 1 pF; Rf = 200 ;
Cf = 1 nF; ( . . . . ) RRE = 50 k; CRE = 300 pF.
20
0
dB
−20
−40
(a)
0
30
60
Phase shift
90
120
150
180
1000 10000 100000
Log frequency
(b) [Hz]
Fig. 22 Calculated (a) bode and (b) phase shift curves for the
circuit of Fig. 18. Conditions are given in Fig. 21.
19 are essentially identical. Note the ex- 1.2.7.3 Other Destabilizing Conditions
ception that the stabilizing capacitor is The simple Ru Cdl model of the elec-
much less effective in reducing ringing trochemical cell provides a challenging
in the positive feedback case. The Bode control situation. The presence of dif-
plots of Figs. 23(a) and 20(a) are quite diffusional faradaic current reduces the
ferent. Gain rises above unity over a larger reactance of the working electrode in-
range of frequencies in the positive feed- terface by adding a parallel noncapaci-
back case. This reflects the fact that the tive current path across Cdl . However,
frequencies necessary to reduce IRu drop some electrode processes can transiently
must be supplied at increased gain. Note increase the reactance of the interface,
also that the ‘‘job’’ of reducing IRu drop thus decreasing the control loop sta-
is shifted to the current transducer, OA3, bility. For example, potential-dependent
in positive feedback, which partly explains adsorption or desorption of ions at the
the reduced effectiveness of CST in mini- interface or passivation/depassivation phe-
mizing ringing. nomena can destabilize an otherwise
40
20
dB
(a)
−20
Log frequency
0.20 [Hz]
0.15
0.10
0.05
[A]
0.00
i
−0.05
−0.10
−0.15
−0.20
0.0 0.2 0.4 0.6 0.8 1.0
E
(b) [V]
Fig. 23 Calculated (a) and (b) CV curves with a 2000 V s−1 scan rate
applied to the model potentiostat of Fig. 18 using different amount of
positive feedback IR compensation. ( ) 0% compensation,
Rc = 1000 , Ru = 200 , CST = 10 pF, Cdl = 0.2 µF, RRE = 10 k;
CRE = 1 pF; Rf = 200 ; Cf = 1 nF; (-.-.-.) 90% compensation,
CST = 10 pF; ( . . . . ) 99% compensation, CST = 10 pF; (- - - - ) 99%
compensation, CST = 1000 pF.
stable control loop. In such cases, the requirements for a 1-V step applied to the
control loop must be further stabilized, test circuit of Fig. 18. With Rc = 1000 ,
often at the cost of a decreased frequency Ru = 200 , and CST = 10 pF, OA1 is re-
response. quired to supply a transient voltage and
Maintaining control is more difficult current greater than 35 V and 30 mA, re-
with certain potential excitation wave- spectively. This is excess of what most
forms. Step functions (as in chronoam- simple OAs can supply and a buffer ampli-
perometry) are particularly severe tests for fier (as in Fig. 11) would be required. Re-
potentiostats. As an example, consider the ducing Ru to 2 (Rc = 1198 ) increases
the voltage and current required to greater noise, and various combinations of these.
than 350 V and 300 mA, respectively. How- Applications of some of these methods are
ever, this transient occurs during the first discussed in the following chapters (e.g.
microsecond after a step. Degrading the Chapter 2.1 considers cyclic voltammetry
frequency response of OA1 (e.g. by in- and Chapter 2.2 describes chronoamper-
creasing CST ) reduces the transient power ometry. Although traditional analog elec-
requirements. tronic methods were used to generate
any of these excitation waveforms, digi-
1.2.8 tal methods are much more efficient and
Excitation Signals and Data Acquisition versatile and can replace nearly all tradi-
tional analog signal generation methods.
A variable voltage source is often used to The use of high-accuracy and very fast
program the magnitude and time course digital-to-analog converters (DACs) in con-
of the potential or current excitation sig- sumer electronics, such as in compact disc
nal applied to an electrochemical cell. The players and in high-resolution PC graphics
controlling voltage can be as simple as cards, has made DAC-based signal sources
a battery connected to a manually ad- widely available and, in general, signif-
justed potentiometer. The output signal icantly less expensive and more precise
of a potentiostatic or galvanostatic circuit than equivalent analog sources.
is usually a voltage proportional to the A DAC converts a set of binary digits
current at the WE or cell voltage. Tradi- into an analog current or voltage. The
tionally, the output of an electrochemical number of the binary digits used in each
experiment would be recorded by hand in a conversion sets the output resolution of a
laboratory notebook or with an electrome- DAC. A typical DAC might have 12 bits of
chanical recorder using a pen on graph resolution, corresponding to 212 (i.e. 4096)
paper. Oscillographic recorders, storage possible analog outputs. An internally or
oscilloscopes, or magnetic tape are used for externally supplied reference sets the DAC
transient experiments. However, the avail- output range. This allows a DAC to provide
ability of inexpensive personal computers a wide range of possible outputs. For
(PCs) has essentially eliminated these tra- example, the DAC reference might be
ditional analog recording methods in favor used to set an output range of ±10 V
of more versatile digital recording. Sim- or, alternately, 0 to −1 V. In either case,
ilarly, analog methods of programming the DAC resolution (e.g. 12 bits) fixes
excitation signals have been supplanted by the digital resolution but the accuracy
digital means. This section discusses digi- of the analog output depends on the
tal methods of experiment control and data output range. The number of bits of
acquisition. resolution, n, sets the full-scale accuracy
according to this equation 100%/(2n − 1).
1.2.8.1 Excitation Signals Thus, a 12-bit DAC has an accuracy of
Signals for programming a controlling 0.024%, which corresponds to 4.9 mV or
potential or current in an electrochemi- 0.24 mV accuracy on a ±10 or 0–1 V
cal cell include large amplitude potential range, respectively. Toggling the value
sweeps or steps (as in cyclic voltammetry of the least significant bit (lsb) produces
or chronoamperometry), sine and square the smallest possible output change.
waves, pulse sequences, pseudorandom Increasing the output accuracy of a DAC
requires increasing the number of bits signal should be no smaller than one-half
or decreasing the full-scale output range. of the full-scale range. For example, an
The rate at which the digital codes are input signal of between ±1 V presented to
sent to the DAC is governed by a clock, a 12-bit ADC with a ±10 V input range will
which also, necessarily, sets the maximum be digitized with only 9 bits of resolution,
rate at which the analog output can vary. limiting the ultimate precision to 0.2% of
The clock is simply a sequence of digital the full scale. Most ADC systems provide
pulses occurring at regular intervals. The on-board amplification that can be used
maximum rate at which a DAC can to better match the input signal to the
produce an output is typically limited by full-scale range of the data. More recently,
the ability of the analog output to settle to a inexpensive 16-bit and higher-resolution
fraction of the lsb value. In general, a DAC ADC systems have become available. A 16-
with greater resolution will not be able to bit ADC retains a precision of 0.01% over
generate output as rapidly as a DAC with a ±10 V range. The higher resolution also
lower resolution. allows a ‘‘zoom’’ into the data to extract
A concern with using a DAC-based sig- small signals from a large background.
nal source is the limited output resolution. Increased accuracy is possible with
Commonly, the DAC resolution is 12 bits sample averaging. Rather than acquire a
on PC-based instrumentation. With a ±10- single sample at sample rate f, n samples
V scale, the lsb change is 4.9 mV, which are acquired at sample rate n × f . The
means that a voltage sweep generated with samples are then averaged to retain the
this DAC has a staircase appearance. How- effective sampling rate f . Averaging has a
ever, changing the output range to ±2 V number of advantages. If a signal contains
improves the resolution to less than 1 mV, only white noise, the signal-to-noise ratio,
which makes the sweep essentially identi- SNR, is improved by a factor of n1/2 .
cal to the ideal linear response. The ±2 V In addition, the effective resolution in
range still permits most voltammetry ex- the digitized sample is increased by a
periments, but if necessary, using a second factor of 0.5 × log2 (n). Averaging 16 12-
DAC provides an additional offset voltage. bit samples, produces one sample with
Using a higher-resolution DAC is an even an effective 14-bit resolution. A point
better option. For example, a 16-bit DAC of diminishing returns occurs with this
allows better than 0.3 mV accuracy with a method since all ADC cards show a loss in
full-scale range of ±10 V. their effective number of bits (ENOB) as
the sampling rate increases. The sample
1.2.8.2 Data Acquisition rate must be proportionately increased to
In most cases, data from electrochemical allow for averaging and the input circuitry
experiments are acquired using analog- may not completely settle at the highest
to-digital converters (ADCs) [18]. An ADC acquisition rate, decreasing resolution.
converts an analog signal (usually voltage) However, the ENOB often does not drop
to a binary code. As in DACs, ADCs use significantly until sample rates of greater
a digital clock to set the rate at which than 10 kHz are used.
digital codes are generated. The number The Nyquist criterion states that an in-
of bits in the digital output is related put signal can be recovered if it is sampled
to the accuracy with which the signal is at least two times the bandwidth of the
encoded. For highest accuracy, the input signal. Sampling slower than the Nyquist
rate leads to aliasing, where the undersam- to purchase an instrument with limited ex-
pled signal appears as an artifact at lower perimental capabilities and then enable a
frequencies. Aliasing is avoided by using set of experiments if needed at a future
a low-pass filter on the input to the ADC date. The new feature set will be per-
and sampling at twice the filter cutoff fre- manently enabled or leased for a specific
quency or higher. Note that the filter will time interval or number of experiment
attenuate the signal power at frequencies repetitions.
higher than the filter cutoff but strong In other instruments, a general-purpose
noise sources above the cutoff frequency computer communicates with separate de-
may produce aliased signals. Failure to pre- vices containing ADC and DAC circuitry.
vent aliasing will place the noise present at Often the ADC and DAC circuitry is on a
frequencies higher than the Nyquist rate computer board that plugs into an inter-
into the digitized data samples. nal expansion slot of the PC. Alternately,
the ADC and DAC might be part of a
1.2.8.3 Microcomputer Control and general-purpose test instrument such as
Storage an oscilloscope, high-resolution voltam-
ADCs and DACs rely on a microcomputer meter, or an arbitrary waveform generator.
or microcontroller to acquire and provide, In such cases, data and control signals are
respectively, the digital codes needed for exchanged between the PC and the test in-
their operation [19]. In some commercial strument via an external bus. Serial busses,
instrumentation, a microcomputer is em- such as RS-232, are included as standard
bedded into the instrument along with equipment on most PCs and have been
memory and other circuitry to provide for a popular choice. The general-purpose in-
timing and external communication. Such terface bus (GPIB or standardized as the
integrated devices are referred to as micro- IEEE-488) is widely used on many types
controllers. In addition to acquiring data of test and measurement equipment. It
and generating excitation signals, the mi- has the advantage that up to 15 devices
crocontroller accepts user input from the can be attached to one GPIB controller.
front panel (e.g. pushbuttons or switches) GPIB busses are not standard equipment
or from a PC over a digital communi- on PCs but can be added with a plug-in
cation channel. The microcontroller also controller card.
maintains displays and data output. The A recent trend in PCs is towards ‘‘legacy-
computer program for the microcontroller free’’ systems, which ultimately will lead
is highly specific for the hardware found to a loss of internal expansion slots.
in the instrument and is typically stored in Introduction of standardized, high-speed
read-only memory (ROM) where it is pro- communication ports compensates this
tected against user modification for both loss. Importantly, test equipment will not
reliability and security reasons. Some in- be restricted for use with specific hard-
struments employ rewriteable ROM that ware or operating systems (such as plug-in
allows the manufacturer to send out up- data-acquisition boards). The Universal Se-
dated command sets or fix program bugs. rial Bus, USB, and IEEE-1394 (FireWire)
Newer instruments will probably allow are examples of low- (1.5–12 Mb s−1 ) and
the manufacturer or vendor to enable fea- high-speed (up to 400 Mb s−1 ), respec-
tures remotely via the Internet or special tively, serial communication busses that
computer commands. This allows a user are available on modern PCs [20]. These
are superior to RS-232 and GPIB busses efficient coding of programs is necessary
since either is faster and both support to minimize the expense in purchasing
more that 60 devices per controller while and preparing the ROM storage chips, es-
also supplying power to the device. Electro- pecially for commercial instrumentation.
chemical equipment employing these or For this reason, programming in assem-
similar communication busses are likely bly or in an efficient compiled language
to appear since the busses are now ubiq- such as C is common. In contrast, PCs are
uitous and commercialization has made general-purpose computers and, as such,
interface circuitry inexpensive. Communi- are required to run an operating system
cation links can also be established using as wells as other programs that run si-
Internet or similar protocols (e.g. TCP/IP). multaneously with the instrument control
Some test equipment is now available with software. This requires the various pro-
Ethernet ports, allowing remote data col- grams to cooperate with each other and,
lection. The use of wireless connections to thus, share hardware and processor re-
test equipment is likely in the future given sources. Modern operating systems, such
the rapid commercialization of comput- as Window, Macintosh, or Linux, enforce
ing in this area. An emerging technology limits on the ability of a single program
is Bluetooth, which uses a 2.4 GHz RF to directly control a piece of hardware or
link to provide a low-speed (1 Mb s−1 ), respond in a guaranteed time to a hard-
short-range (10 m) data link to other Blue- ware signal. Thus, programming at a low
tooth devices [21]. Current Internet and level is inconvenient. The solution is to
wireless data connections are probably use a higher-level programming language
too unreliable and slow for many appli- that allows the user to interact with the
cations. For example, the 2.4 GHz band hardware through vendor-supplied drivers.
of Bluetooth and other RF links is sus- One example is LabView (National In-
ceptible to interference from microwave struments), a platform-independent data-
ovens, which operate at the same fre- flow language programmed by connecting
quency and, ironically, are the reason for iconic representations of functions, exe-
the availability of the 2.4 GHz commu- cution blocks, hardware, and so forth.
nication band. However, several reports Vendors also supply drivers for conven-
of remote or ‘‘teleoperated’’ experiments tional programming languages such as
over Internet links are available and it is Basic, C, and Fortran.
likely that these applications will rapidly Use of a general-purpose computer
increase in number [22–25]. for controlling an electrochemical instru-
Microcontroller systems and PCs use ment allows using computing power to
software programs to manage the hard- perform many data manipulation and dis-
ware in electrochemical instruments. Mi- play functions. An example is the use
crocontroller systems are designed to op- of the computer to perform coulomet-
erate in ‘‘real time’’. Data must be acquired ric experiments by digital integration of
and stored without interruption from other the cell current. This task would require
tasks, such as updating a display or ac- a separate circuit in a completely ana-
cepting user input. For this reason, the log instrument but incorporation into a
control programs are usually designed computer-controlled instrument requires
to be the sole program running on the only a few lines of computer code. The
instrument’s microcontroller. In addition, amount of analog circuitry in modern
and future instruments will be reduced to The section of Fig. 24 to the left is
the bare minimum required to maintain marked ‘‘Polarization Control’’ and is in-
electrical control with all other functions volved in electrical control of the cell
assigned to increasingly faster computers. potential or current. Note that the instru-
ment can operate as either a potentiostat
1.2.9 or galvanostat and the potential or current
Example Instrument applied to the cell is programmed by sum-
ming the internal sources, marked DC Ref
There are many examples of electro- and Sweep, with an external polarity in-
chemical instrumentation in the literature. put. The box marked Feedback/Bandwidth
Unlike the basic ideas presented here, control represents actual control circuitry
practical instruments often contain a great similar to that in Figs. 8 and 11. Re-
deal of additional circuitry. A concern, lays and electronic switches, controlled by
especially with commercial instruments, the internal microprocessor, allow switch-
is that the circuit operates reliably and ing between potentiostatic or galvanostatic
within specification. Circuitry to provide mode. The microprocessor also sets the
protection from input and output over- control loop bandwidth, which allows the
voltage/overcurrent is present. Often, ad- experimenter to trade bandwidth for in-
dition of calibration circuitry allows the creased control loop stability. Other inputs
instrument to self-calibrate and report to the feedback control circuitry are the
out-of-specification or fault conditions. De- RE potential and IR compensation (if nec-
spite this additional complexity, the basic essary). A power amplifier is inserted at
operation of the instrument follows the the counter (auxiliary) electrode connec-
electronic principles outlined above. A tion. This allows currents of up to 2 A to
block schematic diagram of the Solartron be applied to the cell.
Instruments Model 1287 ‘‘Electrochemi- Current flow at the working electrode is
cal Interface’’ illustrates a typical high- determined from the voltage drop across
end commercial instrument (Fig. 24). The a measuring resistor. Unlike in Fig. 9,
SI1287 is similar to other commercial in- a differential amplifier (cf. Fig. 12) mea-
strumentation in that it is designed to sures the voltage drop, which permits a
provide an experimenter with a wide range floating WE configuration. An optional
of possible experiments. There is no direct offset, ×10 gain, and low-pass filter is
control of the internal circuitry by the user; applied to the current signal before it
the internal microprocessor mediates all is sent to the current output terminal.
aspects of the instrument. Although man- Voltage measurement with the SI1287 em-
ual operation of the SI1287 is possible ploys a somewhat unusual configuration.
with the front-panel controls, the designers Two REs are provided and a differential
clearly intended control of the instrument amplifier measures the voltage difference
by a personal computer. To that end, a between them (RE). The use of two REs
RS-232 or GPIB interface provides a com- has a number of significant advantages. As
munication bus. The block diagram is not discussed previously, using a measuring
intended to be a complete description and, resistor at the working electrode intro-
for example, does not indicate the cir- duces an additional voltage drop caused
cuitry involved with the digital control or by the product of cell current and Rm .
front-panel display. Connecting RE2 to the WE and RE1 to
∆RE
POL
Summing
DC ref
amp ΣPOL
∆RE-Bi
P stat Differential Reject Sum
Sweep G stat amp volts Volts
Feedback/ × 1 output
Sum
bandwidth IR × 10
control comp
Voltage measurement
×1
× 0.01 RE1 RE2 I
Polarity Power
input amplifier Current
Current to output
Reject ×1 Filter/
voltage
amps Sum × 10 buffer
Polarization control converter
CE Current measurement I-Bi
RE1 RE2
CE WE
Fig. 24 Simplified schematic diagram of the Solartron model SI1287 instrument.

1.2 Analog and Digital Instrumentation
49
the RE avoids this problem. The voltage 2. P. T. Kissinger in Laboratory Techniques in

difference between references does not in- Electroanalytical Chemistry (Eds.: P. T. Kissin-
ger, W. R. Heineman), Marcel Dekker, New
clude the voltage drop produced by Rm . A York, 1996, pp. 165–194.
second use for dual reference electrodes 3. R. M. Wightman, D. O. Wipf in Electroan-
is to monitor the voltage dropped across alytical Chemistry (Ed.: A. J. Bard), Mar-
a membrane or phase boundary. An op- cel Dekker, New York, 1989, pp. 267–353,
tional offset, ×10 gain, and low-pass filter Vol. 15.
4. D. Britz, J. Electroanal. Chem. 1978, 88,
is applied to the current signal before it is 309–352.
sent to the voltage output terminal. Either 5. R. M. Souto, Electroanalysis 1994, 6, 531–542.
positive feedback or current-interruption 6. R. L. Deutscher, S. Fletcher, J. A. Hamilton,
methods is used for IR compensation. Electrochim. Acta 1986, 31, 585–589.
The SI1287 uses two internal ADCs 7. H. Yamagishi, J. Electroanal. Chem. 1992,
326, 129–137.
with accuracy greater than 16 bits for all 8. D. O. Wipf, Anal. Chem. 1996, 68, 1871–1876.
measurements. The nine different mea- 9. G. S. Popkirov, J. Electroanal. Chem. 1993,
surement points are marked on the dia- 359, 97–103.
gram by oval-shaped labels, that is, POL 10. J. D. McIntyre, W. F. Peck Jr, J. Electrochem.
Soc. 1970, 117, 747–751.
or RE. Measurements are made by us-
11. R. Bezman, Anal. Chem. 1972, 44,
ing the internal microprocessor to switch 1781–1785.
measurement points as required. The mea- 12. D. Britz, W. A. Brocke, J. Electroanal. Chem.
surements are sent to the PC as they are 1975, 58, 301–311.
collected or a small amount of internal 13. W. J. Wruck, R. M. Machado, T. W. Chap-
man, J. Electrochem. Soc. 1987, 134, 539–546.
memory is used to store measurements. 14. D. K. Roe in Laboratory Techniques in Elec-
troanalytical Chemistry (Eds.: P. T. Kissinger,
1.2.10 W. R. Heineman), Marcel Dekker, New
Conclusion York, 1986, pp. 195–236.
15. D. Garreau, P. Hapiot, J.-M. Savéant, J. Elec-
It was once said that there were more troanal. Chem. 1989, 272, 1–16.
different types of electrochemical instru- 16. C. Amatore, E. Maisonhaute, G. Simonneau,
ments than electrochemists to operate
17. P. R. Gray, P. J. Hurst, S. H. Lewis et al.,
them. The nature of an electrochemist is to Analysis and Design of Analog Integrated
tinker in the desire to squeeze a bit more Circuits, John Wiley & Sons, New York, 2001.
signal-to-noise or a few more ohms of IR 18. M. Tyler, Sci. Comp. Instrum. 2001, 18,
drop out of an experiment. This section 32–34.
19. C. G. Masi, R&D 2001, 43, 14–19.
only presents a small portion of the in-
20. B. Weaver, Sci. Comp. Instrum. 2001, 18,
formation required to fully comprehend 27–30.
or construct instrumentation for electro- 21. J. C. Haartsen, IEEE Personal Commun. 2000,
chemistry. It is hoped that the information 7, 28–36.
given will allow the reader to appreciate the 22. J. W. Overstreet, A. Tzes, IEEE Control Syst.
Mag. 1999, 19, 19–34.
design and operation of electrochemical 23. O. Frederik, J. R. Danaraj, B. Fleet et al.,
instrumentation. Electroanalysis 1999, 11, 1027–1032.
24. C. L. Enloe, W. A. Pakula, G. A. Finney et al.,
References IEEE Trans. Educ. 1999, 42, 174–179.
25. M. Hadidahassan, S. J. Young, S. T. Peltier
1. P. Horowitz, W. Hill, The Art of Electronics, et al., J. Struct. Biol. 1999, 125, 235–245.
Cambridge University Press, Cambridge,
1989, pp. 175–261.
1.3 Digital Simulation in Electroanalytical Chemistry 51
1.3 techniques referred to without citation;

Digital Simulation in Electroanalytical these can all be found in the Refs. [1–3].
Chemistry
1.3.2
Dieter Britz
Historical
Aarhus Universitet, Århus C, Denmark
Electrochemists consider Feldberg as the
person who introduced digital simulation
1.3.1
to the field. The extensive chapter de-
Introduction
scribing the method was written between
the years 1964 [4] to 1969. This was the
In this chapter, digital simulation, as ap-
standard text for a long time. It is not
plied in electroanalytical chemistry, will
always appreciated that Randles, in fact,
be described. The description must nec- used finite differences to solve the LSV
essarily be brief, but references are pro- problem in the year 1948 [5], thus mak-
vided for more detail. In particular, the ing him the first electrochemist to do
classical chapter by Feldberg [1] and the so. Finite differences (the technical term
more detailed (and more up-to-date) mono- mathematicians use for digital simulation
graph [2], as well as the excellent review of or, more precisely, for the implementa-
the field by Speiser [3], going up to 1996, tion of one method of digital simulation),
are recommended reading. go back to about 1911 (Richardson) and
Very broadly speaking, digital simu- the technique was formalized in 1924.
lation in electrochemistry is the nu- Feldberg used what the present author
merical solution of Fick’s second diffu- calls the box method (see Sect. 1.3.4), yet
sion equation, the standard procedure in the mathemat-
ics/computer field is the point method.
∂c ∂ 2c The box method will get brief treatment
=D 2 (1)
∂t ∂x here but the accent is now on points. They
make advanced methods (see Sect. 1.3.9),
where c is the concentration of a diffusing such as the implicit methods that have
species, t is the time, x is the distance from made possible great advances in efficiency,
the electrode, and D is the diffusion coeffi- easier to implement.
cient of the diffusing species. The equation Up to the writing of the present au-
as given is one-dimensional and in Carte- thor’s monograph [2], in 1988, there was
sian coordinates, and can take other forms a number of unsolved problems, mainly
according to the number of species consid- concerned with homogeneous chemical
ered and other coordinate systems. Except reactions, which introduce several prob-
in some simple (model) cases, numerical lems such as thin reaction layers and
solution is the only way to get results. The coupled sets of equations. These prob-
most popular electroanalytical techniques lems have now been solved, at the cost
are linear sweep or cyclic voltammetry of programming complexity. Bieniasz [6]
[LSV, CV, cf Chapter 1.1 and 2.1 in this introduced variable grids to enable dy-
volume] and potential step voltamme- namic gridding, thereby overcoming the
try [Chapter 2.2 in this volume]. In the reaction layer problem (and problems
sections to follow, there will be names and with sharp transients); and Rudolph [7]
solved the problem of coupled equations This, classical Cottrell, system has an
by introducing the block-tridiagonal nota- analytical solution that can be used to test
tion (see Sect. 1.3.9). The next push is now methods and programs. That solution is in
into the area of ultramicroelectrodes (see terms of concentrations (see Chapter 2.2
Sect. 1.3.11). in this volume for the solution). What
one normally wants is the current i at
1.3.3 a given time, and this is derived from
Some Mathematics the concentration profile by using Fick’s
first diffusion equation, relating flux to
In order to simulate an electroanalytical concentration gradients:
problem, it is necessary first to express
it in a suitable mathematical form. For ∂c
i = nFADc (3)
one-dimensional problems, Eq. (1) must ∂x x=0
be augmented by the boundary conditions,
including initial conditions. There are where n is the number of electrons
some typical systems (experiments) that transferred, F the Faraday constant, and
illustrate the procedure: potential step and A the electrode area. Differentiating the
controlled current experiments, and LSV solution c(x, t) as given in Chapter 2.2
of a reversible electrochemical reaction. in this volume and inserting this into
These will suffice, although in later Eq. (3), gives the well-known solution for
sections, other more complicated systems the current. Equation (3) is also the basis
will be mentioned. for the calculation of current in digital
simulation.
The constant current (chronopotentio-
1.3.3.1 Three Model Systems metric) experiment has a different set of
In a potential step experiment such as the boundary conditions, the main difference
Cottrell experiment, we must know the being the imposition of a constant current
starting conditions, that is, the concen- or, via Eq. (3), a constant concentration
trations for the whole cell; since we are gradient at the electrode. The set is
dealing (for the moment) in only one di-
mension, this means concentrations along t ≤ 0, all x: c = c∗
the space variable x. As the experiment
proceeds, the concentrations at a range of ∂c
t > 0, x = 0: = const (4)
distances from the electrode will change, ∂x x=0
but at the electrode (x = 0), it will be fixed t > 0, x −−−→ ∞: c = c∗
at zero, and at some sufficient distance
from the electrode, it will retain the value (note that it is only the middle one of the
c∗ , the initial bulk concentration. All this three equations that is different).
is expressed formally as a set of bound- The above two cases involve only a single
ary conditions: species of concern. In LSV or CV, however,
one must follow the development in time
t ≤ 0, all x: c = c∗ and space of both species (if not more) of
the electrochemical reaction, for example,
t > 0, x = 0: c = 0 (2)
t > 0, x −−−→ ∞: c = c∗ A + e− ⇐⇒ B (5)
and we designate the respective concen- normalizing relations:

trations by cA and cB . There are now two
diffusion equations and more boundary c
C=
conditions. If the system is reversible and c∗
there is no B present initially, and if the po- t
T = (7)
tential scan starts at E1 and proceeds with τ
the scan rate −v (see Chapter 2.1, this vol- x x
X= =√
ume) then the combined set of boundary δ Dτ
conditions is
thus forming the new dimensionless
t ≤ 0, all x: cA = c∗ ; cB = 0 variables C, T , and X. Equation (1) then
transforms to the simpler form,
t > 0, x −−−→ ∞: cA = c∗ , cB = 0
t ≤ 0: E = E1 ∂C ∂ 2C
= (8)
t > 0: E = E − vt ∂T ∂X 2

cA nF
t > 0, x = 0: = exp (E − E 0 ) and the boundary conditions for the
cB RT
Cottrell experiment also normalize to
(6) the set
in which we note a form of the Nernst
equation and the standard potential E 0 , T ≤ 0, all X: C = 1
and the bulk concentration c∗ refers to
species A (as is usual for a set of species T > 0, X = 0: C = 0 (9)
involved in a reaction system). The two T > 0, X −−−→ ∞: C = 1
species each have some of their own
boundary conditions and are also coupled
The solution of this Eq. (8) with (9)
by the Nernst condition.
now applies to all sets of experimental
parameters c∗ , D, and τ and is thus much
1.3.3.2 Normalization more useful. In all cases, the solution
Normally, the equations as presented will be within the bounds 0 ≤ T ≤ 1
above, are rendered dimensionless (nor- and 0 ≤ X ≤ mδ, in which m must be
malized) by expressing the variables as chosen suitably (this will be described
multiples of reference values. For the Cot- in Sect. 1.3.7). Current is also expressed
trell system, the reference time value is the without dimensions, simply (in view
observation time, or duration of the exper- of Eq.
3) as the dimensionless gradient
iment, τ ; concentrations are referred to c∗ ∂C
, which will be given the symbol
and distance x to a suitable length scale δ. ∂X X=0
In view of the solution of Eqs. (1 and 2) (see G in this chapter.
Chapter 2.2, this volume and Ref. [2]),√ this For chronopotentiometry, a good choice
scale is conveniently defined as δ = Dτ , for the observation time is the transition
which means that the concentration profile time itself, and this has the interesting
will extend into the solution bulk by only consequence that the constant current
√
a few δ-units. We then have the following reduces to constant G = 1/2 π, the
boundary conditions (4) becoming boundary conditions
T ≤ 0, all X: CA = 1; CB = 0
T ≤ 0, all X, C = 1
1√
T > 0, X −−−→ ∞: CA = 1, CB = 0
T > 0, X = 0: G = 2 π (10)
T ≤ 0: p = p1
T > 0, X −−−→ ∞: C = 1
T > 0 : p = p1 − T
and the solution must be that C(X = 0) CA
T > 0, X = 0: = exp(p)
falls to zero at T = 1. CB
(14)
The LSV system is also normalized,
(note the much simpler form of the Nernst
eliminating the scan rate as well. It was
condition in the last line).
noted above (Equation set 6) that the swept
potential changes at the rate −v, where v
has units of Vs−1 . First, we refer voltages 1.3.3.3 Denormalization
to units of RT /nF and shift it relative to Having obtained a solution to a given
E0: problem in dimensionless terms, we
might wish to go back to dimensioned
nF
p= (E − E 0 ) (11) variables, in order to compare with an
RT experiment. This simply reverses the
normalizing equations. Thus, we have
which makes the dimensionless unit p
equal to about 25.69n mV and it is centered c = c∗ C
on the standard potential E 0 . The scan rate t = τT
too is now expressed in terms of changes
(15)
in p-units per s: x =δ·X
RT
E = E0 + p
p = p1 − at (12) nF
and the current through (computed) G:
with the new unit a defined as the c∗
inverse time i = nFAD 1/2 √ G (16)
τ
nF The groundwork is now laid for the

a= v (13) simulation methods themselves.
RT
1.3.4
We can now define a reference time τ , The Box Method
being the time taken for the potential to
change by one p-unit. Normalizing time t In this section, a very brief description
as above (Eq. 8), that is T = t/τ (τ = a −1 ), of the ‘‘box method’’ is given. This
we now have a sweep starting at p1 and is the method as originally devised by
moving, for all experimental parameter Feldberg [1]. It is, in the opinion of the
sets, with the dimensionless sweep rate 1, present author, awkward and outdated,
as is seen in the new dimensionless although many prefer it because it appears
to provide a closer definition for what is a time interval δt,

being done, in contrast with the ‘‘point
method’’, which is the way numerical δt
ci (t + δt) = ci (t) + D (ci−1 (t)
mathematicians have always solved partial δx 2
differential equations. − 2ci (t) + ci+1 (t)) (18)
Consider Fig. 1, showing three boxes
along a narrow tube representing the (for details of the derivation of this
diffusion path in one dimension. The formula, see Refs. [1, 2]). It will be seen
tube has been sectioned into short lengths in Sect. 1.3.5 that this is identical with
(boxes) of equal length δx, and they are the expression derived there, using points.
numbered consecutively, starting from 1. However, there are problems with the box
The essence of the box method is to use
method. First of all, consider the end
Fick’s first diffusion equation, relating the
elements close to the electrode, Fig. 2.
unit area flux f , in moles per second
At the electrode, how does one express
per m2 , of the diffusing substance at a
the concentration gradient between the
given point (or plane) to the concentration
electrode and the first box? The way this
gradient there:
has normally been handled is to assume
dc a distance of half δx between these. As
f = −D (17)
dx explained in [2], however, this gives rise
to an incorrect formula that happens to
This is very intuitive and perhaps explains
the appeal of the method. The problems work surprisingly well by accident. When
begin with finding approximations to the dealing with geometries other than the
derivative term. The boxes are considered simple ones depicted here – for example,
to have uniform concentration within each boxes that do not all have the same
box, and the gradient between two boxes is length, or curved shells as one gets
expressed by the concentration difference with a cylindrical or spherical diffusion
divided by the distance between adjacent space – the question of how far adjacent
boxes. For example, the flux of substance boxes are from each other becomes
from box i into box i + 1 is approximated vexing. Feldberg and his coworkers have
as −D(ci+1 − ci )/δx and this, combined devised corrections for curved elements, to
with the analogous expression for the flux overcome such problems. It is the strong
from box i − 1 into box i and converting view of the present author that all these
to concentration change in the center box, problems vanish with the use of points,
gives the discrete explicit formula for the and this method will be used in the rest of
new concentration in the center box i after this chapter.
Fig. 1 Three boxes in the ci−1 ci ci+1

solution bulk.
c1 c2
Fig. 2 End-boxes at the
c0
electrode.
... Fig. 3 Points distribution.

c0 c1 c2 ci−1 ci ci+1
1.3.5 right-hand side of the diffusion equation,

Point Discretizations and produced a set of ordinary differential
equations. There are N of these:
The point method proceeds from the
discrete approximation of Fick’s second dC1 1
= 2 (C0 − 2C1 + C2 )
diffusion equation, for example, Eq. (1) or, dt H
after normalization, Eq. (8) or variants of dC2 1
that equation according to the chosen ge- = 2 (C1 − 2C2 + C3 )
dt H
ometry. Both time and space are sampled
at a set of points, and approximations for ···
the left-hand and the right-hand sides of dCi 1
the equation are sought. If the spacing of = 2 (Ci−1 − 2Ci + Ci+1 )
dt H
the points is even throughout time (inter-
···
vals δt or, dimensionless, δT ) and space (h,
or H ), then it is always straightforward to dCN 1
= 2 (CN −1 − 2CN + CN +1 )
develop good approximations in terms of dt H
concentration values at these points. Con- (20)
sider Fig. 3, showing a number of points This is a useful approach, leading to a large
in space along the single dimension X variety of methods. Note that nothing has
with the corresponding concentration val- yet been said about how to handle the
ues, and intervals H between the points. boundary values that are not subject to
The usual discrete approximation of the diffusion in themselves. The outer value
right-hand side of Eq. (8) is is normally constant at the bulk value.
The value at the electrode, C0 , is the
∂ 2C 1 problem, and will be dealt with in the
≈ 2 (Ci−1 − 2Ci + Ci+1 ) (19)
∂X 2 H next section in detail. For the moment,
applying at the point X = iH . If we let us dwell on the Set (20). We need to
represent (sample) the whole diffusion find discrete expressions for the left-hand
space by N points, placing the electrode sides, the time derivatives. Here, we can
at X = 0 and including one more point draw on the practice in the field of ordinary
outside the diffusion space at X = (N + differential equations (ode’s). Taking the
1)H (which will be the convention here), ith equation out of (20) and expressing it
we then have N expressions of the in the simple form
type (19), for which the first (i = 1) uses
the boundary point at the electrode and the dCi
= fi (t) (21)
last (i = N ) uses the outer boundary, or dt
the bulk of the solution. The latter will not
there are several common ways to dis-
normally change during an experiment.
cretize it. The Euler formula, or forward
The approach described up to this point
difference, is
is called the method of lines, or MOL. It is
also called the semidiscretization approach, Ci (t + δt) − Ci (t)
because so far, we have only discretized the = fi (t) (22)
δt
where the right-hand side indicates that and substituting the symbol λ, defined as
the expression is intended to pertain to
δt
time t. This, as will be seen in Sect. 1.3.7, λ= (25)
is the easiest – but least efficient – method H2
to follow. By using Taylor expansions (see this becomes
Ref. [2] for details), it can be shown that this
scheme is first order with respect to time. Ci
= Ci + λ(Ci−1 − 2Ci + Ci+1 ) (26)
This means that if we try to reduce the
errors in a given computation by making Now, without going into details, it is well
known that the calculation will deviate
the time interval δt smaller, the error will
from the proper values more and more
get smaller proportionally with δt. This is
within a few time steps, unless λ ≤ 0.5
not very good. The method has another
for this explicit method. Such a deviation
limitation, of stability, and this may be the
(which might take the form of increasingly
point to mention this problem.
wild swings, or wide one-sided divergence)
is called instability. Thus, the stable range
1.3.5.1 Stability for λ is 0 − 0.5. This is inconvenient, as
If we combine the above forward difference it makes for inefficient computation. If
in time (22) with the discretization (19), we one attempts to lower errors by decreasing
can get an explicit expression for the new the spacing along X (that is, the value of
concentration at the point (iH, t + δt): H ), one must also reduce δt such that λ
remains within the stable range.
Ci (t + δt) − Ci (t) 1 The explicit (Euler) method described
= 2 (Ci−1 (t)
δt H above has this stability limitation. There
− 2Ci (t) + Ci+1 (t)) (23) are other methods that do not. One of
them (reverting again to the ode 21) is the
We now introduce some simplifying no- backward difference (or backward implicit,
tation. Unless certain advanced methods BI) formula:
are used, the situation is always that one Ci (t + δt) − Ci (t)
has, at a given time, a row of known con- = fi (t + δt) (27)
δt
centrations, at time t, and develops a new
row of concentrations at time (t + δt). It Note that the left-hand side is identical
is necessary to mark the X-positions (the with (22), which is the forward difference
subscripts), but time can be marked more formula. The difference is in the right-
conveniently. If there is no time marked, hand side, which is a function of values at
this implies time t; and time (t + δt) will the next time level, to which the same left-
be indicated by a superscript dash on hand-side approximation now pertains.
the concentration, for example C
. Using Intuitively, one might expect this form
to be about as inefficient as the forward
this notation and now also rearranging
difference and this is in fact true; this
Eq. (23), we get the explicit expression for
formula, too, is first order with respect to
the new concentration
time, as is the forward difference formula.
δt The big difference is that this formula,
Ci
= Ci + (Ci−1 − 2Ci + Ci+1 ) applied to the diffusion equation, yields
H2
(24) a method that is stable for all λ. This is
the formula then, taking a shell i at some radius, the

explicit discretization of the equation is
Ci
= Ci + λ(Ci−1

− 2Ci
+ Ci+1

) (28)
δt
and is known as the BI method, or Ci
= λ(Ci−1 − 2Ci + Ci+1 ) +
the Laasonen method (see Ref. [2]). The 2H Ri
reader will realize, on perusal of the × (Ci+1 − Ci−1 ) − KδtCi (31)
system (20) that, if this formula is applied
to that system, the result is a system of where the second term on the right-hand
N unknowns, no longer explicit in each side of (30) has been replaced by a central
equation, and the system needs to be difference (hence the division by 2H ). If
solved as a whole. The method is one we take Ri = iH , then this becomes
of several implicit methods that will be
λ
discussed in Sect. 1.3.9. Despite its poor Ci
= λ(Ci−1 − 2Ci + Ci+1 ) + (Ci+1
performance in its simple form, the BI 2i
method forms the basis for some highly − Ci−1 ) − KδtCi (32)
efficient methods.
The last way to be mentioned here to 1.3.5.2 Gradients at the Electrode
discretize the ode (21) is what is called the Equation (3) shows how the current is
trapezium method: related to the gradient of concentration
at the electrode, and in dimensionless
Ci (t + δt) − Ci (t) δt
= fi t + (29) computations, the gradient G, as defined
δt 2
above, is often required, either as an output
where the same left-hand-side expression (expression of current), or as a boundary
now pertains to the midpoint (in time), condition (see the next section). Thus,
and becomes a central difference. This a way of expressing it is needed. The
one is second order with respect to
simplest approximation is
time or O(δt 2 ), meaning that the error
is proportional to δt 2 . This is much C1 − C0
more efficient than first-order methods. G≈ (33)
H
As with the BI (Laasonen) method, the
application of this formula to system (20) This is a forward difference, applied at
(see Sect. 1.3.9) yields an implicit system of one end of the interval. It is rather a poor
equations that must be solved as a whole. approximation, first order with respect to
This is the Crank–Nicolson method; and H . It is nevertheless favored by many
it, like the Laasonen method, is stable for and may be justified, especially if that
all values of λ. In fact, this statement can first interval is very small, in which case
be misleading; there are problems with the approximation is not too bad (this
this method, to be discussed later. argument comes in with unequal intervals,
One last example should suffice to illus- see Sect. 1.3.9). It is however rather easy to
trate the way the point method can be used. use better approximations. Formula (33)
If we take the diffusion equation in cylin- uses just two concentration points; if one
drical coordinates and add a homogeneous uses more, the approximation becomes
first-order chemical reaction, better, of higher order. The present author
∂C ∂ 2C 1 ∂C tends to use a six-point approximation, but
= 2
+ − KC (30) there is reasonable argument for using the
∂T ∂R R ∂R
three-point formula: is an example of this, the value C0 being

fixed at zero (see Equation sets 2 and 9).
− 32 C0 + 2C1 − 12 C2 A different boundary condition is the
G≈ (34)
H Neumann condition or derivative bound-
ary condition. An example is seen with
which is O(H 2 ) and thus matches the
chronopotentiometry, Equation sets (4 and
order of the approximation (19) used in
10). The procedure here is that a value of C0
simulations. In general, the formula for is computed such that it fits with the con-
an n-point approximation is centration profile (set of points C1 . . . CN ),
so as to satisfy the gradient specification.
1
n−1
G≈ βi Ci (35) Using the simple two-point approxima-
H tion (33) and given a G value, this yields
i=0
an explicit expression for C0 ,
where the β-coefficients depend on n and
are tabulated in Ref. [2]. The values, for C0 = C1 − H G (36)
example, for n = 3 (Eq. 34), are (−1.5, 2, Extensions to higher-order formulae, us-
−0.5). ing (35), are obvious. Note however, that
The reader should now have a good in more complex contexts, an equation
idea of how the various forms of the like (36) might not be explicit; it might be
diffusion equation are discretized, using one of a set of such equations for a number
the point method. of species and the variables on the right-
hand side might themselves be unknowns
1.3.6 (see Sect. 1.3.9).
Boundary Conditions A third kind of boundary conditions
lies in between the two presented above;
In the last section, discretization of the par- they are the mixed boundary conditions
tial differentials away from the boundaries and normally involve two species. Take
was described; those boundaries must now Reaction (5), involving the two species A
be added to complete the picture. In the last and B. In Sect. 1.3.3 (boundary conditions
section, the convention was established set 14), the reversible case was mentioned,
that the diffusion space is represented and a boundary condition was
by a number of points, counting from
CA,0
zero to N + 1, where the points numbered = exp(p) (37)
1 . . . N are those at which concentrations CB,0
will change as a result of diffusion, and better expressed as
the other two are boundaries, subject to
CA,0 − CB,0 exp(p) = 0 (38)
other relationships. Thus, point N + 1 is
normally held constant, at the initial bulk where p was the dimensionless electrode
value for the species concerned (there are potential. This condition couples the two
exceptions to this). species. It is not enough, however, there
For the point at the electrode, there are being the two unknowns, CA,0 and CB,0 .
technically three classes of conditions. The The other required condition comes from
Dirichlet condition is that in which the con- the fact that the fluxes at the electrode of
centration there has a certain value (per- the two substances must be equal and op-
haps time variant); the Cottrell experiment posite (B is generated at the same rate as
A is used up): is normalized to the dimensionless K0 by

fA + fB = 0 (39) τ
K0 = k0 (44)
D
Flux was defined in (17) and we must here
take note of the fact that the two species’ where τ is the designated reference time
diffusion coefficients might not be the and D the diffusion coefficient of the
same. For this reason, we define diffusion reference species. If we further define two
coefficient ratios, all referred to that of the convenient dimensionless rate constants,
main species, whose bulk concentration is the forward and backward rates,
also the reference, c∗ . Thus, for species B, Kf = K0 exp(−αp)
we have
DB Kb = K0 exp((1 − α)p) (45)
dB = (40)
DA then, the Butler–Volmer equation be-
The flux now being equal to the diffusion comes
coefficient multiplied by the concentra-
GA = Kf CA,0 − Kb CB,0 (46)
tion gradient, Condition (39) becomes, in
terms of dimensionless gradients. and if we again, for example, use the two-
point approximation for GA and rearrange,
GA + dB GB = 0 (41) this gives the discrete expression
where we now have two different gradients (1 + Kf H )CA,0 − Kb H CB,0 = CA,1
G, both expressed in discrete terms by a (47)
form of Eq. (35). If, for illustration, the sim- This must be paired with the flux condi-
ple two-point formula (33) is chosen, the tion (43) to yield the two unknowns.
discretized boundary condition becomes
CA,1 − CA,0 CB,1 − CB,0 1.3.7
+ dB = 0 (42) The Explicit Method
H H
and eliminating the H terms and rear- Here, the simple explicit method is briefly
ranging the unknowns on the left-hand gone through, in practical detail. The
side, finally, example taken is the one-dimensional Cot-
trell system, diffusion Eq. (8) and bound-
CA,0 + dB CB,0 = CA,1 + dB CB,1 (43)
ary conditions (9). In writing a simulation
This now, taken together with Condi- program, some practical decisions have to
tion (38), determines the two unknowns. be made. Normally, the time interval for
Note again, as with Eq. (36), that (43) need a complete simulation is 0 ≤ T ≤ 1, and
not be explicit but may, in complex situa- that leaves only the number of steps NT
tions, be merely one of several statements or its inverse δT to set. Then there is the
involving a larger set of unknowns. interval H in X. This is related to the time
Lastly, the system described by Reac- interval by the important parameter λ (see
tion (5) might be quasireversible or even Eq. 25), and it might be better to decide on
irreversible, in which case the boundary its value, which then provides that for H .
condition is given by the Butler–Volmer Another quantity is the distance along X
equation. It is preferable to express it in to which the simulation is to extend. Here,
dimensionless form, using potential p, experience (and the solution of the Cottrell
and the heterogeneous rate constant k0 equation, see Chapter 2.2 in this volume)
tells that a value of 6 units of δ as defined 4. For all NT values of T , recalculate

in Equation set (7) is sufficient. In case one the concentration array using Eq. (26),
wants to change it later for some reason, it outputting the desired quantities at
is good practice to make it a program pa- chosen times; these might be G(T )
rameter, Xlim , and set that in the program values and/or the errors in G.
preamble (or, if using Fortran 90, a mod-
ule). This then determines the number of There is a small point to watch in
points N along X, and the required ar- point 4. When moving along the points
ray(s) can be allocated. The present author in X, it is important to use old concen-
prefers these steps: trations only in applying Eq. (26). This
is easily achieved using a moving trio
1. Read in NT and λ; of concentration scalars equal to the old
2. Calculate H (needed for the calculation values, while replacing the actual concen-
of G) and thus N ; tration array with the newly computed
3. Allocate the concentration array values. A short program in Fortran 90,
C0 . . . CN +1 and initialize it (all Ci = 1, which carries out the above steps, is given
C0 = 0); below.
module STUFF
integer,parameter :: dbl=selected real kind(14)
real(kind=dbl),parameter :: Xlim=6.0, pi=3.14159265358979
end module
program COTTSIM
! To simulate the Cottrell experiment, outputting the current
! at doubling intervals. Current G is calculated as 3-pt approx.
use STUFF; implicit none
integer :: NT, N, iT, iX, next out
real(kind=dbl) :: dT, H, lambda, G, C1, C2, C3
real(kind=dbl),allocatable :: C(:)
print ‘("NT, lambda?")’; read*, NT, lambda
dT = 1.0 dbl / NT; H = SQRT(dT/lambda); N = CEILING (Xlim/H)
ALLOCATE (C(0:N+1)); C = 1; C(0) = 0; next out = 1 ! C set up
print ‘(4x, "iT", 5x, "T", 7x, "G", 3x, "G(analyt)")’ ! Header
do iT = 1, NT ! Grand T-loop
C1 = C(0); C2 = C(1) ! Running scalars
do iX = 1, N ! X-loop
C3 = C(iX+1) ! Next scalar grabbed
C(iX) = C2 + lambda * (C1 - 2*C2 + C3) ! new C in array
C1 = C2; C2 = C3 ! Scalars shifted along
enddo
if (iT == next out. OR. iT == NT) then
G = (-3*C(0) + 4*C(1) - C(2)) / (2*H)
print ’(i6, 3f8.3)’, iT, iT*dT, G, 1.0 dbl/SQRT(pi*iT*dT)
next out = 2 * next out
endif
enddo
DEALLOCATE (C)
end program COTTSIM
This rather reduced program produces do iX = 1, N ! X-loop

the following output (excluding the I/O C3 = C(iX+1) ! Next scalar grabbed
C(iX) = C2 + lambda*(C1 - 2*C2 + C3)
dialogue) when run with the input param-
! new C in array
eters NT = 100, λ = 0.4: C1 = C2; C2 = C3
! Scalars shifted along
iT T G G(analyt) enddo
1 0.010 4.427 5.642 if (iT == next out .OR. iT == NT)
2 0.020 3.921 3.989 then
4 0.040 2.808 2.821 T = iT * dT
8 0.080 1.996 1.995 C(0) = (4*C(1) - C(2) - 2*H*G)/3
16 0.160 1.414 1.410 ! C(0) again
32 0.320 0.999 0.997 print ’(i6, 3f8.3)’, iT, T, C(0),
64 0.640 0.706 0.705 1-SQRT(T)
100 1.000 0.565 0.564 next out = 2 * next out
endif
enddo
Clearly, we can do better, introducing
...
error calculations and so forth, but this will
suffice as a brief example. The program can
Using the same parameters as for the
be easily converted to doing a simulation of
Cottrell program, this produces the output
chronopotentiometry. Here, the quantity
of interest is C0 , which should fall to exactly
zero at T = 1. Its value there, then, is the iT T C(0) C(0)(analyt)
1 0.010 0.857 0.900
simulation error. The boundary conditions
2 0.020 0.825 0.859
set (10) applies and, rather than Eq. (36), 4 0.040 0.774 0.800
the three-point equivalent, obtained from 8 0.080 0.698 0.717
Eq. (34) is used to set C0 at the beginning of 16 0.160 0.586 0.600
every iteration in T . This is a nontrivial 32 0.320 0.424 0.434
point. The explicit method calculates 64 0.640 0.193 0.200
100 1.000 -0.006 0.000
a new array of concentration values
from an old array. That old array must
include the appropriate C0 , whose value showing that the error is −0.006 at the
satisfies the boundary condition, such end. The explicit method is first order
that Eq. (36) (or another, perhaps higher- with respect to the time interval δt. A
point approximation of it) is satisfied. This run using the input NT = 200 (double the
causes some apparent redundancy in the previous) yielded the final error of −0.003,
program, arising from the fact that C0 is confirming the order. As will be seen,
set before and, whenever there is output, there are better methods. Here, these two
also after, an iteration. Here is the middle examples will have to suffice.
part of the program, in which it differs
from the Cottrell one: 1.3.7.1 Fudging
There is, especially among adherents of
...
the box method, a rather widespread
do iT = 1, NT ! Grand T-loop
C(0) = (4*C(1) - C(2) - 2*H*G)/3
custom of adjusting the time value, after
! New C(0), 3-pt approx a number of iterations, by subtracting
C1 = C(0); C2 = C(1) half a time interval, and assigning that
! Running scalars adjusted time to the calculated current.
Thus, after j iterations, when T = j δT , (in which the product of the decay of
the current is presented as belonging species B is of no interest), the concen-
to (j − 0.5)δT . The only argument for tration profiles of the two species A and B
this is that it seems to work, providing will have different extents into the solution,
more accurate current values for systems depending on the (often first-order) rate
like the Cottrell experiment. However, constant of the decay reaction. The faster
there is no formal justification for the this reaction runs, the more species B will
trick. This is a fudge and should not be confined to a thin ‘‘reaction layer’’ near
be used. The trick is also applied in the the electrode. This has been quantified; if
case of backward differentiation formula the diffusion coefficient be D and the rate
(BDF) (see Sect. 1.3.9) and there it is fully constant of the decay reaction (assumed ir-
justified, as will be seen later. reversible) k, then the characteristic thick-
ness µ of the reaction layer, (analogous to
1.3.8 the diffusion layer) is given by
Homogeneous Chemical Reactions
D
µ= (49)
The major reason for doing electrochem- k
ical digital simulations is the occurrence
steady state is often established. For first-
of chemical reactions taking place in the
order reactions, the rate constant can
solution, involving either the starting sub-
be normalized by the chosen time ref-
stances or products of the electrochemical
erence τ :
reactions at the electrode, called hetero-
geneous reactions. In contrast, reactions K = kτ (50)
taking place in the solution are called ho- and then the normalized reaction layer
mogeneous. These are of interest in several thickness, in terms of the unit δ, is
ways, either for themselves in mechanistic
studies or, if they are known, their rates. µ∗ = K −1/2 (51)
These can in principle be determined from
comparison of experimentally measured This was, until recent years, a major prob-
curves with those calculated. These calcu- lem for simulations, because large values
lations almost always involve simulation, of the rate constant can mean that there are
since analytical solutions are known only few, or even no, points placed within the
for rather trivial and therefore uninterest- concentration profile of the species form-
ing cases. ing a reaction layer. This led to the use
A systematic classification of homoge- of unequal intervals (see Sect. 1.3.9) but
neous chemical reactions (hcr) in electro- this, too, is limited if a fixed distribution of
chemical cells can be read in Chapter 1 the unequally placed points is used. That
in Volume 8 and here, only some exam- problem has been solved by using dynamic
ples that impinge on simulations will be grids, also described briefly in Sect. 1.3.9.
mentioned. Another problem, not solved until
The following problems arise with hcrs. recently, comes from so-called coupled sys-
In a system such as tems. Take the catalytic (EC’) system
A + e− ⇐⇒ B A + e− ←−→ B
B −−−→ prod (48) B −−−→ A (52)
in which the product B reverts, in below zero, it can never recover from that
a (pseudo-) first-order reaction, to A. erroneous condition and the simulation
There will be two transport equations deviates more and more from reality. This
to describe this system (assume equal happens especially with methods that have
diffusion coefficients for simplicity, and oscillatory responses.
a normalized rate constant K for the
catalytic decay), 1.3.9
Advanced Methods
∂CA ∂ 2 CA
= + KCB
∂T ∂X 2 In this limited space, it is not pos-
∂CB ∂ 2 CB sible to provide an exhaustive survey
= − KCB (53) of the advanced techniques for simula-
∂T ∂X 2
tions, for which see Ref. [2], especially
Note that the first of the two equations has the 3rd edition. Here, only those ad-
terms in both variables, CA and CB . This vances that have helped to overcome the
is not a problem when using an explicit major problems of thin reaction layers,
method, but these are practically useless coupled systems, and efficiency, will be
when hcrs are involved and implicit meth- mentioned. Such excellent methods as or-
ods are normally chosen for their greater thogonal collocation, finite elements, or
efficiency. However, while these produce other sophisticated methods are left out of
tridiagonal systems of equations for sin- necessity.
gle species or uncoupled multireactions,
they produce more complicated systems
for coupled reaction systems, which are 1.3.9.1 Unequal Intervals
not amenable to the usual solution tech- In Sect. 1.3.8, the reaction layers often
niques. This problem was solved recently formed when homogeneous chemical
by Rudolph [7] and references therein, who reactions are involved, were mentioned.
introduced the technique known as block- The higher the homogeneous reaction
tridiagonal solution into electrochemical rates, the thinner the layers become, and
simulations, (see Ref. [2], 3rd edn. only) it is harder and harder to ensure that
even a few sample points are placed
and Sect. 1.3.9.
in that thin region. In practice, first-
Second-order reactions present a few
order reaction rates have an upper limit
small problems. They introduce nonlinear
of (dimensionless) about 1010 , so that
terms into the transport equations and one
reaction layers can get as thin as 10−5 .
wishes to avoid these in the discretized
Using equal intervals in the usual interval
equation systems. There are some tricks
0 ≤ X ≤ 6, this would mean of the order
that work rather well, see Ref. [2]. Another
of 106 points along X and, in order not
problem associated with second-order
to have too high a λ value (Eq. 25), a
reactions, such as
correspondingly large number of time
2B −−−→ C (54) intervals. However, the changes are most
marked near the electrode and clearly it
is that the associated discrete equation will is of advantage to crowd more points
approximate in some linear way the term into that region, loosening up further out.
−KCB2 into the transport equation. The There are two ways to implement unequal
result is that if CB at some stage falls intervals.
A simple transformation of the X-axis Feldberg and Rudolph [7] handle this by
can do the trick, for example, the formula always choosing a very unequal spacing
(in effect, large β) that hopefully will be
Y = ln(1 + aX) (55) able to cope with the highest possible
reaction rates. This means an extremely
with the adjustable parameter a; the
small first interval near the electrode, and
greater its value, the more compression
then also justifies these authors’ use of the
there is of points near the electrode. Hav-
simple two-point approximation for the
ing chosen the value, it is then a simple
current (G).
matter to divide the Y -range into a small
A better way is, clearly, to use a grid
number of (equally spaced) points. Using
of points whose spacing dynamically ad-
this transformation, the diffusion equa-
justs itself suitably. This was introduced
tion must also be transformed, replacing
by Bieniasz [6], and see references therein
derivatives with respect to X with those
who has concluded that his latest scheme,
to Y , the new space variable. For de-
which he calls the patch-adaptive strat-
tails, see Ref. [2]; Eq. (8) transforms to
egy [6], is to be preferred. It is apparently
a slightly different form, the details of
quite simple. One starts with a number
which will not be gone into here, for
of equally spaced points along X. The
the following reason. Numerical special-
program calculates a new concentration
ists regard transformation, and discretiza-
array, and then it exactly halves all the
tion on the (equally spaced) transformed
intervals and calculates another new con-
grid, as the best method to use. How-
centration array. The two solutions are
ever, Rudolph [10] has recently shown
then compared, enabling an error esti-
that direct discretization on the unequally
mate. For those regions in which the
spaced grid in fact yields better results.
error estimate exceeds some threshold, the
The transformation (55) is mathematically
points in that region (and in that region
equivalent to a sequence of points xi ,
only) are doubled once more. Thus, locally,
i = 1, N + 1, spaced according to the
one always deals with equal intervals, but
expression
the intervals themselves vary with X. This
βi − 1 algorithm will always adapt the point dis-
xi = x1 (56) tribution to conditions. If a thin reaction
β −1
layer is formed, points will be doubled
where x1 is the first point away from the up repeatedly in that region until satis-
electrode and β a stretching parameter factory solutions are found there. Note
>1. However, the method does not require that this will work, even when there are
this specific placing of the points, and any sharp concentration changes away from
suitable placing can be used. How this is the electrode, whereas unequal intervals
implemented is explained below. as normally used, will not. Clearly, this
Both the transformation and direct is not an easy programming job, but one
techniques suffer from the fact that the must pay a price for progress. This is one
placement of the points (or boxes) is reason that electrochemists will increas-
static during a given simulation. If one ingly make use of commercial packages
is attempting, for example, to fit a reaction (see Sect. 1.3.10).
rate to an experimental curve, one cannot In general, Rudolph’s recent findings
choose the transform parameter suitably. point to an arbitrary grid of points,
set by some unspecified method, at the will have to suffice to illustrate how
positions xi , i = 1, N + 1. The second Rudolph solved this problem [7]. Take the
spatial derivative at each point xi then catalytic system (52) in Sect. 1.3.8, which
deviates from the form in (19) and is a yields the two coupled pdes shown in (53).
weighted sum of the three concentrations On an unequally spaced grid, these equa-
centered on that point, tions are discretized to the form at the
point xi ,
∂ 2 Ci Ci−1
≈2

∂X 2 (xi − xi−1 )(xi+1 − xi−1 ) CA,i−1 + a1 (i)CA,i + ak (i)CB,i

Ci
− + a2 CA,i+1 = bA,i
(xi − xi−1 )(xi+1 − xi )

CB,i−1 + [a1 (i) − ak (i)]CB,i

Ci+1
+
+ a2 CB,i+1 = bB,i (58)

(xi+1 − xi )(xi+1 − xi−1 )
(57) in which the a-coefficients arise from
which is a first-order expression. the discretization process (and depend
on the algorithm used, for example,
1.3.9.2 The Rudolph Method for Coupled Crank–Nicolson or Laasonen, (see Ref. [2],
Systems 3rd edn. only for details). The point is
If the single-species diffusion Eq. (8) is that the first of the two equations in-
discretized in one of the implicit schemes cludes terms for both species and the
such as that of Crank & Nicolson or whole system of these equations is thus
of Laasonen (BI), the result is in ef- not amenable to the usual Thomas al-
fect a tridiagonal system of equations in gorithm for a tridiagonal system (this is
the N unknown concentrations at the in fact not quite true, since a related
next time level. This can be solved us- strategy can be devised in this partic-
ing the Thomas algorithm, by attacking ular case but the statement is true in
at the ends at which either values are general for coupled systems). Rudolph,
known, or certain boundary conditions however, recognized that the two equa-
are given (see Ref. [2] for details). If an tions can be lumped into a single
electrochemical reaction involves several one in terms of vectors and a ma-
species but the diffusion equation for each trix, becoming
species contains terms in concentrations
only of that one species, the Thomas al- Ci−1 + Ai Ci + a2 Ci+1 = Bi (59)
gorithm can still be applied; the species
will then be coupled by their boundary in which the vector Ci now stands for both
conditions (for example, the Nernst con- concentration terms, that is
dition (37) and flux equality (39)), which
is easy to handle. However, certain sys- CA,i
Ci ≡ (60)
tems result in diffusion equations with CB,i
terms in more than one species, with the
and similarly for the other three vectors,
result that the system of discrete equa-
and the coefficient matrix is
tions has a bandwidth greater than three

and is not amenable to the convenient a (i) ak (i)
Ai ≡ 1 (61)
Thomas algorithm. A standard example 0 (a1 (i) − ak (i))
from equation pair (58). Clearly, Eq. (59) is in this direction, notably by the Compton
now back to the form of the single-species group, using Krylov methods. The reader
discrete equation, with the difference that is referred to Ref. [8] for details.
instead of a row of concentration scalars,
we now have a row of concentration vectors 1.3.9.3 BDF(FIRM)
and a coefficient matrix. This forms a For many years, it was considered that if
system called block-tridiagonal, since the one wanted to use an implicit method,
system has (vector, matrix) elements only the method of Crank–Nicolson was the
on the three middle diagonals. It is thus choice (see Ref. [2] for a description).
amenable to a kind of extended Thomas However, although stable for all λ and
algorithm, and even boundary conditions having second-order accuracy, this method
can be incorporated as vector/matrix has the disadvantage that it reacts to initial
expressions. This then solves completely transients, such as from a potential step,
the problem of coupled equation systems, in an oscillatory manner; and the larger
at the price of a little extra complexity in the value of λ, the longer these (error)
the programming. oscillations persist. Oscillating errors are
The Rudolph method has its limitations. not necessarily a bad thing – they provide
Firstly, as mentioned above, it is harder a visible sign of convergence as they
to program. Secondly, it cannot be im- are damped out – but if they are large
plemented effectively for two-dimensional and persist over a large fraction of
systems. Both problems are overcome by the total simulation time, they render
what the present author calls the brute force the simulation useless. So a method
method. This recognizes that computers that has similar accuracy and stability
are very fast these days and we can some- as Crank–Nicolson, and has a smooth
times afford to be wasteful of computer response to transients, is desirable. The
time, especially if this makes program- Laasonen method almost does the trick; it
ming easier (and thus, safer from program- fulfills all but one of these requirements,
ming errors). Take the discrete equation that is, it lacks accuracy being, like the
pair (58). If we simply write this for all N explicit method, first order with respect
pairs and include the particular discrete to time. There are two tricks that can be
expressions for the boundary conditions, employed to improve its accuracy, one of
the result is a possibly large linear equation them being the BDF. This is the term
system. There are rather efficient subrou- used in the area of ordinary differential
tines for solving such systems directly, and equations. The method was suggested for
this is the easiest option. In its simplest use with pdes and was recently revived by
implementation, this ignores the fact that Feldberg and coworkers (see Ref. [2], 3rd
the coefficient matrix is mostly empty and edn. only, for references), who prefer to call
the calculation is wasteful. By appropri- it the ‘‘fully implicit Richtmyer method’’
ate ordering of the concentration arrays (‘‘FIRM’’) for historical reasons.
(specifically, in this case, using the order Consider Eq. (27), which expresses the
CA,1 , CB,1 , CA,2 , CB,2 , . . . , CA,N , CB,N ), time derivative as a backward difference,
the matrix can be made tightly banded, that is, takes the simple two-point expres-
which allows more efficient subroutines sion to pertain to the next point in time
to be used that bypass all the zero el- at t + δt. This has an error that is dom-
ements. Some progress has been made inated by the first-order term in δt. It is
possible to increase this order and thus to calculate the very first new point at time δt
get better accuracy for the derivative, by using, say, the three-point scheme above,
bringing in more points lying at earlier we not only need the initial concentration
times; the more points that are brought values but also values at time −δt, which
in, the higher the order. The expressions do not exist. The problem is worse for
for this are very similar to the multipoint higher-point schemes. There are several
expressions for G, as seen in Eqs. (34 and strategies to handle this. Feldberg and
35)-in fact, the coefficients are the same coworkers simply set all past values to
but in reverse order and sign, as we are the initial values and start from there.
now looking backwards. In all cases, what From the world of numerical solution
we get is a derivative that applies to the of odes, we have a more rational start.
next time level. The simple three-point Here, the first step is simply a BI step,
scheme serves to illustrate the method. using the two-point scheme, the same
The diffusion equation discretizes to as the Laasonen method. This yields two
concentration rows, the initial row and the
3
1
C − 2Ci +
Ci new one at δt. These can now be used
2 i 2 with a three-point scheme to generate

= λ(Ci−1 − 2Ci
+ Ci+1

) (62) the next row at 2δt, and the resulting

three rows with a four-point scheme and
in which
Ci denotes the value at the so forth. When the required number of
previous time level t − δt. Note that the rows has been generated, one goes on
right-hand side is the Laasonen form. This with that number. This method produces
again gives rise to a system of equations reasonable results, which are better than
of the unknowns C
, which can be solved those from the rough start used by the
once the appropriate boundary conditions Feldberg school, in itself (but see below).
are applied. For coupled systems, the The result is still not as good as one
Rudolph method can be used. The result might expect, however, as can be seen
for the three-point scheme as seen above is when one uses a better approximation
a smooth stable response and second-order than (19) for the derivative along X. It
accuracy, the same as Crank–Nicolson. is then seen that the values do not, as
The three-point scheme is in fact a good expected, improve with the number of
match for the three-point expression (19) points used for the BDF scheme, beyond
normally used, and experiments indeed the three-point scheme.
show that using more points in time There is a surprise here, however.
for BDF does not gain much accuracy, The Feldberg school uses a trick that
the limit being then set by the three- renders the simple start highly accurate.
point derivative along X. Feldberg and At each step n, the time corresponding
coworkers nevertheless prefer a five-point to that step, which ought perhaps to be
BDF scheme. At about this number of nδt, is instead corrected by half a time
points, the solution shows signs of small interval to (n − 1/2)δt. This is the same
oscillations, which become worse for the device previously used for the explicit
six- and seven-point schemes; the scheme box method, and described as a fudge in
is unstable for more than seven points. Sect. 1.3.7. In this case, however, there
There is a problem with this method, is rational basis for doing this, as was
however, in how to start it. In order to shown recently [9]. Briefly, it turns out
that the simple start, for any order BDF, smooth response to transients. Note that
shifts the time by precisely 1/2δt and thus there are no starting problems to over-
the correction brings the simulated values come here. The method can be driven to
exactly into line with the time. BDF might higher orders, by taking combinations of
thus be regarded as a useful simulation even smaller steps and using more com-
method, if one is willing to put up plicated formulae than (64). As with BDF,
with having to keep several concentration however, given the second-order nature
vectors in memory. of (19), there is no appreciable gain in
going to more than the two-step scheme
1.3.9.4 Extrapolation described above. In fact, the efficiency falls
The alternative to BDF, that still makes use off, because there is more calculation to do
of the pleasant properties of the Laasonen but no accuracy gained.
or BI method, is extrapolation. When one Extrapolation is an old technique going
says that discretization (27) is first order back to about 1927 and was suggested
with respect to δt, one means that the error for general pdes in the 1980s, and revived
ei can be expressed as a polynomial like for use with electrochemical simulations
in recent years (see Ref. [2], 3rd edn. for
ei = a1 δt + a2 δt 2 + · · · (63) details and references). It appears to be the
best method at the time of writing.
where, most often, the lowest order (here,
the term in δt) dominates. If this could
1.3.9.5 Method of Lines
be eliminated, then a smaller error would
Increasingly, as the quest for efficiency ac-
result, dominated by the (much smaller)
celerates and more sophisticated methods
next order term. Briefly, this can be done
are used, the electrochemist who wants to
with the following. First calculate the new
do experiments and not spend all his/her
value, with the time interval δt, getting
time writing and debugging computer
an error approximately equal to a1 δt;
programs, will use program or subrou-
call this error e(1, 1), meaning that it is
tine packages. In the next section, some
the error from one step of one whole
whole simulation packages will be men-
δt interval. Then repeat the calculation,
tioned. A cheaper solution however is to
now using two steps of half a δt interval
use professional subroutines as parts of
each. The resulting error, e(2, 0.5), will
the programs. One such approach is the
be approximately equal to 0.5a1 δt, since
MOL. This has already been mentioned
we have taken half-δt steps. Note that we
in Sect. 1.3.5, in which only the deriva-
do not know what these errors are, we
tive in X was discretized, leaving a set
only know their interrelationships. Now, if
of odes (20). If the boundary conditions
we subtract twice the error e(2, 0.5) from
are added to this system, the result is a
e(1, 1), the result should be close to zero.
differential-algebraic equation (DAE) sys-
That is, using the same notation as for the
tem, and there exist subroutines to solve
errors, the calculation
such systems, written by experts. The prob-

Ci = Ci (1, 1) − 2Ci (2, 0.5) (64) lem then reduces to presenting the data in
the form required by the subroutine, and
the lowest-order error term will be elim- it does the rest. Usually, these routines can
inated. The result is Crank–Nicolson-like either do a single step or a number of steps,
accuracy, together with stability and a perhaps even determine the best step size
or sizes itself. Ultimately, MOL may be- it is a little demanding to learn to use.
come the method of choice, supported by It presupposes some knowledge of such
such professional routines. programs on the part of the user, who
can choose the algorithm to be used, for
1.3.10 example. It is Bieniasz’ contention that
Commercial Packages different algorithms are best suited to
different problems. This might be a bit
A step further than using professional sub- of a challenge to the nonmathematical
routines as part of one’s own program is electrochemist, who just wants simulation
to use the several available commercial results. At this time, too, ELSIM has only
whole packages for electrochemical sim- rather crude facilities for parameter fitting.
ulation. A reasonably up-to-date listing of This is taken to a high level in the commer-
available simulation programs (including cial package DigiSim, written by Rudolph
some online ones) can be seen at the and Feldberg (in C++) and distributed by
website ftp://electrochem.cwru.edu, then BAS (at a considerable price). This is a
going on to the directories pub and es- PC Windows program and rather easy to
tir. Speiser [3] has described the main learn to use. It now uses the BDF (or, as
contenders in detail and compared them they prefer, FIRM) method and strongly
by computations. The reader is referred unequal intervals that are considered to
to his article for details of these pro- enable computations of even the fastest ho-
grams. Speiser has himself produced such mogeneous reactions. The program is able
a program, and was one of the first to to compare simulations with experimental
do so; it is the package EASI, and he data files and adjust a number of parame-
is still working on it sporadically (avail- ters to best match the data. The program
able from the author, Bernd.Speiser@uni- has suffered from bugs; these may have
tuebingen.de), now using the language been removed by now. Another program is
C++. It uses a ‘‘menu’’ of reaction mech- POLAR by Huang (info@electroanal.com
anisms to choose from. One of the next of or http://www.electroanal.com); its de-
such packages was that of Gosser, CVSIM, scription promises many features. Nervi
which comes with Gosser’s book on cyclic has a program, ESP, to be seen at
voltammetry. It is written in Pascal and http://chpc06.ch.unito.it/chemistry/esp
uses the explicit method. Bieniasz pro- manual.html. There is an online service
duces the more general and sophisticated for simulation and data fitting at Oxford,
ELSIM (also written in C++), which allows http://physchem.ox.ac.uk:8000/wwwda,
the user to enter a symbolic form of the provided by the Compton group.
differential equations to be solved as well With all these packages, as pointed out
as the boundary conditions; it now also by Speiser [3], there is an element of
allows entering mechanisms as one writes uncertainty as to the output. The best
them normally. Bieniasz is developing, strategy is probably to run a given package
at the time of this writing, a PC Win- (or two or more of them, if possible)
dows version of his program. His program not only on the problem at hand, with
is available through him, nbbienia@cyf- its (presumably) unknown result, but also
kr.edu.pl. ELSIM is very general and can on a similar problem, simplified to the
handle practically any problem one is likely point at which an analytical solution exists
to encounter, including adsorption, but for comparison.
1.3.11 steep gradients at the electrode edges.

The Future Predictably, this is where new research
efforts will be directed in the next decade,
As was written in the foregoing, the
and one expects greater use of professional
major problems in electrochemical digital
subroutines, at the least.
simulation in one dimension have now
been solved. The new frontier is in two-
and more-dimensional systems. During References
the last 20 years or so, ultramicroelectrodes
have more or less replaced the mercury 1. S. W. Feldberg in Electroanalytical Chemistry
(Ed.: A. J. Bard), Marcel Dekker, New York,
drop, and these form a two-dimensional 1969, pp. 199–296, Vol. 3.
diffusion space, whether they be single 2. D. Britz, Digital Simulation in Electrochem-
disks, or disk arrays, or (arrays of) strips istry, 2nd ed., Springer, Berlin, Germany,
or generator-collector strips, and so forth. 1988. 3rd revised edition is expected to appear
Here, the problems include the fact of in 2003.
3. B. Speiser in Electroanalytical Chemistry
the large numbers of nodes required for (Eds.: A. J. Bard, I. Rubinstein), Marcel
reasonably accurate computations, and Dekker, New York, 1996, pp. 1–108, Vol. 19.
thus, long computation times and extreme 4. S. W. Feldberg, C. Auerbach, Anal. Chem.
computer memory needs, at the least; 1964, 36, 505–509.
as well as the fact that discretization 5. J. E. B. Randles, Trans. Faraday Soc. 1948, 44,
327–338.
usually produces (widely) banded systems 6. L. K. Bieniasz, J. Electroanal. Chem. 2000,
of equations, so that sophisticated methods 481, 115–133.
of solution need to be used in order 7. M. Rudolph in Physical Electrochemistry (Ed.:
to have sufficient memory and realistic I. Rubinstein), Marcel Dekker, New York,
computation times. There is also a lack of 1995, pp. 81–129.
8. J. A. Alden, R. G. Compton, J. Phys. Chem. B
theoretical work on the numerical methods 1997, 101, 8941–8954, 9606–9616.
used. It is by no means certain that 9. D. Britz, J. Electroanal. Chem. 2001, 515, 1–7.
the familiar stability criteria will apply 10. M. Rudolph, J. Electroanal. Chem. 2002, 529,
with these systems, which often have 97–108.
81
2.1 was called ‘‘potentiostatische Dreieckspan-

Linear Sweep and Cyclic Voltammetry nungsmethode’’ (potentiostatic triangular
voltage method) [10, 11] owing to the char-
Bernd Speiser acteristic triangular development of the
.. ..
Institut fur Organische Chemie, Tubingen,
working electrode potential when perform-
Germany
ing ‘‘cyclic’’ experiments. Stationary rather
2.1.1 than dropping electrodes were used, and
Introduction consequently the term ‘‘stationary elec-
trode polarography’’ [12] became popular.
2.1.1.1 Historical Background Already in the mid-1960s, there was rich
On the basis of polarography [1, 2], a potential of applying such experiments to
variety of widely used instrumental elec- the determination of concentrations but
troanalytical techniques has been elabo- even more to the elucidation of reac-
rated and used. Some of them are called tion mechanisms and kinetics coupled to
voltammetries [3]. The basic experiments electron transfer at an electrode was rec-
were initially performed by Matheson and ognized. Today the resulting linear sweep
Nichols [4] and independently advanced by or cyclic voltammetries are employed as
Randles [5, 6] and Sevčik [7]. simple, flexible routine techniques in par-
While a more physico-chemical defini- ticular as sophisticated means to solve
tion will be provided in Sect. 2.1.1.2, we chemical and mechanistic problems. The
observe here, that a linearly changing combination with computer control, ul-
(scanned or sweeped) potential is imposed tramicroelectrodes, and digital simulation
on a working electrode, and the result- has further contributed to their success.
ing current is recorded and analyzed.
Hence, advances were related to the de- 2.1.1.2 Potential Scan
velopment of fast and accurate control Voltammetries – Definitions
of the electrode potential with poten- The voltammetric [13] (volt-ampero-metric)
tiostats [8]. Since already in early work techniques discussed here record the
the potential was changed so fast that current i through a working electrode as
electrical recording devices had to be a function of time t, while the potential
used, the term ‘‘cathode ray polarogra- E applied to this electrode is controlled
phy’’ [5, 6, 9] or ‘‘oscillographic polarog- as a linear function of t. During the
raphy’’ [7] appeared. Later, the technique experiment the electrode is stationary, and
82 2 Electroanalytical Methods
usually the electrolyte is not agitated, but or ‘‘cyclic triangular wave voltammetry’’
at rest. The excitation signal can consist of (‘‘chronoamperometry with multiple tri-
a single linear sweep (often called a scan) angular potential sweeps’’) is used for the
from a starting (Estart ) to an end potential multicycle variant. It seems, however, that
(Fig. 1a), of two such segments (two scans, the latter distinction is neither observed in
often called a cycle) between a starting and common usage nor necessary.
a ‘‘switching’’ potential (Eλ ; Fig. 1b; the The rate at which E changes is the
end potential is not necessarily the same potential (sometimes also: voltage) scan rate
as the starting potential), of three segments
dE
(Fig. 1c shows an often encountered case) v= (1)
or of several cycles (multicycle experiment; dt
Fig. 1d). (often simply termed ‘‘scan rate’’). Al-
If E is changed according to Fig. 1(a), though in most cases, the absolute value of
the experiment is called linear scan or v is the same for all scans of an experiment,
linear sweep voltammetry (LSV), while a change of |v| at Eλ (or other switching
variants according to Figs. 1(b to d) are points) has been suggested [15]. This varia-
commonly called cyclic voltammetry (CV). tion does not seem to have attracted much
Although in particular the latter term attention, however.
is widely used, IUPAC [14] recommends Owing to the linear relationship between
the more formal names ‘‘chronoamper- E and t as given in Eq. (1), the current can
ometry with linear potential sweep’’ for be plotted versus time (as directly mea-
LSV, while ‘‘triangular wave voltamme- sured in the experiment; typical curves in
try’’ (‘‘chronoamperometry with triangular Figs. 2a and 2b) or versus potential (typical
potential sweep’’) is reserved for an ex- results in Figs. 2c and 2d). The latter case is
periment according to Fig. 1(b), and CV most common (‘‘cyclic voltammogram’’).
0.6 0.6
0.5 Eend 0.5 El
Potential, E
Potential, E
0.4 0.4
0.3 E 0.3
[V]
[V]
start
0.2 0.2
0.1 0.1 Estart
0.0 0.0
−0.1 −0.1
0 5 10 15 20 0 5 10 15 20 25 30 35 40
(a) Time, τ (b) Time, τ
0.6 0.6
0.4 0.5
Potential, E
Potential, E
0.2 0.4
0.3
[V]
[V]
0.0
0.2
−0.2 0.1
−0.4 0.0
−0.6 −0.1
−0.1 0.1 0.3 0.5 0.7 0.9 1.1 −10 0 10 20 30 40 50 60 70 80 90
Time, t
(c) [s] (d) Time, τ
Fig. 1 Excitation signals for potential scan voltammetries: (a) linear sweep voltammetry
(LSV); (b–d) cyclic voltammetry (CV).
2.1 Linear Sweep and Cyclic Voltammetry 83
1.8 1.7
1.6
Current, i·106
Current, i·106
1.4 1.2
1.2 0.7
1.0
[A]
[A]
0.8 0.2
0.6 −0.3
0.4
0.2 −0.8
0.0 −1.3
−10 0 10 20 30 40 50 60 0.00 0.02 0.04 0.06 0.08 0.10 0.12
Time, t·103 Time, t
(a) [s] (b) [s]
1.8 1.7
1.6
Current, i·106
Current, i·106
1.4 1.2
1.2 0.7
1.0
[A]
[A]
0.8 0.2
0.6 −0.3
0.4
0.2 −0.8
0.0 −1.3
−0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 −0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6
Potential, E Potential, E
(c) [V] (d) [V]
Fig. 2 Typical results of LSV (left) and CV (right) experiments plotted versus time (a, b) or
potential (c, d).
When plotting i versus E, it has conven- these techniques. Some state-of-the-art de-
tionally to be observed that the scan is velopments will be described and finally
presented clockwise as indicated by the applications will be demonstrated.
arrows in Fig. 2(d). Still, however, two con- Space limitations do only allow a rather
ventions for plotting potentials exist (either selective discussion. For further infor-
positive or negative potentials to the right). mation, the reader is referred to review
Likewise, the sign of currents (anodic cur- articles [13, 18–24].
rents defined positive or negative) is not
unequivocal [16, 17]. Consequently, when 2.1.2
interpreting a cyclic voltammogram one Theory of Potential Scan Voltammetries
has to be careful as regards the exact direc-
tion of the individual scans. 2.1.2.1 General
As already highlighted in Chapter 1.1, elec-
trode processes are complex (see, e.g.
2.1.1.3 Scope of the Chapter Volume 8, Chapter 1), and usually consist
The present chapter will first deal with of several coupled steps, among these at
the theoretical background of LSV and least transport to and from the electrode
CV experiments. The basic equations de- and electron transfer, often linked to chem-
scribing the most important processes will ical transformations. Each of these steps
be discussed. A short account of practical may influence the voltammetric response.
problems will lead to a discussion of the ex- In order to theoretically describe a voltam-
perimental limitations and the most com- metric experiment, all pertinent processes
mon caveats to be observed when applying have to be included.
Although it is possible to derive the nec- where

essary mathematical equations in ‘‘real’’,

dimensioned quantities, it is often advan- nF 0
θ = exp (Estart − E ) (4)
tageous to use a more general, ‘‘dimen- RT
sionless’’ formulation. This may involve
both the current, and the time/potential contains all time-independent quantities,
variables. Often, however, in the literature and
only the current is normalized, while the
potential is retained as a dimensioned vari-
exp[at − (m − 1)atλ ]
for odd m
able, only referred to the formal potential S=
exp[matλ − at] for even m
E 0 of the electrode process. (5)
We assume a voltammetric experiment (with a = nF /RT × v) encapsulates the
with equal scan rates for all scans and variation with t. Note the sign change
an excitation waveform as in Fig. 1(b). in Eq. (5) as compared to Nicholson and
Furthermore, we assume that the substrate Shain’s [12] work, where reductions were
is present totally in its reduced form considered.
at Estart , i.e. the primary process is an
oxidation. Conversion to the case of an
2.1.2.2 Reversible Electron Transfer
initial reduction is easily possible.
We will first consider a simple system, and
The temporal development of the elec-
then include further complications.
trode potential will then be defined as [12]
Basic starting assumptions are the
following:
E = Estart + vt − (m − 1) × vtλ
• The electrode is a solid stationary disk
for odd m
of area A.
E = Estart − vt + m × vtλ • The electrolyte is not in motion and
for even m (2) extends virtually infinitely from the
electrode. It contains an excess of
supporting electrolyte.
where m numbers successive scans and tλ • The (reduced) substrate of the electrode
is the time needed to scan the potential reaction is present in the electrolyte
from Estart to Eλ . Note that v is always with homogeneous distribution in a
taken positive for the oxidation case concentration c0 . It undergoes electron
considered. transfer (assumed to be in equilibrium
We calculate the concentration ratio for at all times) with a formal potential E 0
the reduced and oxidized redox partners at and has a diffusion coefficient D.
the electrode, as defined by the Nernst
equation (n, F , R, and T have their Under such assumptions (‘‘reversible’’ sys-
usual meaning; x is the distance from tem), diffusion is the main mode of
the electrode surface) transport and semiinfinite linear diffusion
conditions prevail. Thus, Fick’s 2nd law

cox (x = 0) nF 0
describes the variation of concentrations
= exp (E − E ) = θS with t and the space coordinate x (extend-
cred (x = 0) RT
(3) ing from the electrode surface, x = 0, into
the electrolyte): the electrode (see also Volume 8, Chap-

ter 1) by
∂cred ∂ 2 cred
= Dred (6) ∂cred
∂t ∂x 2 i = nFAD = nFAfred,x=0
∂x x=0
∂cox ∂ 2 cox (11)
= Dox (7)
∂t ∂x 2 (fred,x=0 is the flux of species red at x = 0).
A particular form of Eq. (11) is
At the beginning of the experiment, t = 0,
0
i = nFADcred πDred aχ(at) (12)
0
cred (x, 0) = cred , cox (x, 0) −−−→ 0 (8)
as derived by Nicholson and Shain [12]
i.e. only the reduced form is present. It is where χ(at) is a dimensionless function
assumed that during the experiment there obtained from the integration procedure
is no change of concentrations at a large (current function).
distance from the electrode (bulk solution): The result for a single scan is shown as
a current–potential curve in Fig. 3(a). It
0
cred (x → ∞, t) = cred , exhibits the characteristic peak shape with
a peak potential Ep and a peak current
cox (x → ∞, t) −−−→ 0 (9) ip . The potential where the current attains
50% of the peak value (ip /2) on the rising
while Eq. (3) applies at x = 0. Often it flank is the half-peak potential Ep/2 .
is further assumed that Dox = Dred for The form of the curve is qualitatively
simplification. explained as follows: at the beginning
Finally, Eq. (10) defines mass conserva- of the experiment, essentially no current
tion at x = 0: flows since we start from an equilibrium
situation (see Sect. 2.1.3.4). Scanning E
∂cox ∂cred induces a change in cox /cred at x = 0
Dox = −Dred
∂x x=0 ∂x x=0 according to Eq. (3): Red is converted into
(10) ox at the electrode, resulting in a current.
Equations (6) to (10) and (3) form a sys- The concentration of red is depleted in the
tem of second order partial differential immediate vicinity of the electrode surface,
equations with initial [Eq. (8)] and bound- while the concentration of ox increases.
ary conditions [Eqs. (3), (9) and (10)]. The These changes of c(x = 0) with respect to
presence of the time dependence in Eq. (3) the bulk solution values induce diffusional
prevents a closed form solution of this sys- transport. Thus, red diffuses towards the
tem. Several approaches have been used to electrode, while ox diffuses to the bulk.
calculate the concentrations c as a function The resulting flux f of material sustains
of x and t, including Laplace transfor- the current flow.
mation [7, 12] and digital simulation [25], Further increase of E causes a decrease
all relying on numerical integration steps. of cred (x = 0) and an increase of fred (x =
Chapter 1.3 provides further information 0) with an increase of i. Simultaneously,
on digital simulation. according to Eq. (12) a diffusion layer
From concentration profiles c = f (x, t) develops
√ that extends over a thickness
it is possible to calculate the tempo- δ ∼ t. The increase in δ tends to
ral development of the current through decrease fred (x = 0). The two effects are
1.8 13
1.6 ip
11
Current, i·106
Current, i·106
1.4
1.2 9
1.0 Ep 7
[A]
[A]
0.8 ip/2 5
0.6 3
0.4 Ep/2
0.2 1
0.0 −1
−0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 −0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6
(a) [V] (b) [V]
1.5
Current, i·106
1.0 E fp
0.5
[A]
0.0
−0.5 E rp
−1.0
−1.5
0.0 0.2 0.4 0.6 0.8 1.0 1.2
Potential, E
(c) [V]
Fig. 3 Current–potential curves for a reversible electron transfer: (a) shape and important
features of a linear sweep voltammogram; (b) linear sweep voltammograms as a function of the
scan rate v = 0.1, 0.2, 0.5, 1., 2., 5. V s−1 with increasing ip ; (c) cyclic voltammograms and their
dependence on Eλ .
superimposed. If cred (x = 0) decreases to diffusion-controlled current is increased

very small values, further changes of E over that for longer timescales (smaller v).
have no more effect, and the current is This is due to the fact that the concen-
mainly determined by the expansion of tration gradient and the flux of educt to
δ. About 180 mV after the peak, the the electrode increase with increasing v.
decay of the voltammetric i/t curve can Figure 3(b) demonstrates this relationship
essentially be described by potential-step that is often used to prove diffusion con-
Cottrell behavior (see Chapter 2.2). trol of the current as opposed to currents
Quantitatively, the peak current is given due to surface-bound or adsorbed redox
by Ref. [12] species (see Sect. 2.1.2.5).
√ The position of the peak on the potential
πχp = 0.4463 or
axis is also of interest. For the reversible
0
ip = nFAcred aDred × 0.4463 (13) case,
28
At T = 298 K the so-called Randles–Sevčik Ep = E 0 + mV (15)
n
equation
for T = 298 o C (oxidation) and
ip = (2.69 × 105 )n3/2 Acred
0
vDred (14)
28
follows. Note, that ip in both Eqs. (13) Ep/2 = E 0 − mV (16)
n
and (14) is proportional to v 1/2 . The scan
rate defines the timescale of the experi- hold. Both relationships only apply if
ment. For short timescales (high v), the Dred = Dox .
Obviously, the peak on the first scan of a transfer step have to be included, most
cyclic voltammogram is identical to the one often by a Butler–Volmer relationship
of a single scan experiment. The reverse
(‘‘back’’) scan curve and the details of the −αnF 0
i = i0 exp (E − E )
reverse peak, however, depend on the exact RT
position of Eλ (Fig. 3c). In this peak the
(1 − α)nF
product of the electrode reaction, which − exp (E − E 0 ) (18)
RT
has accumulated in the diffusion layer
during the forward scan, is transformed which is substituted for Eq. (3). Here,
back to the starting species. For practical 0
i0 = nFAks cred is the exchange current
purposes, the peak shape is independent (defined at E = E 0 ) and α is the transfer
of Eλ for n|Eλ − E 0 | > 35 mV [12]. For coefficient.
n|Eλ − Ep | > 180 mV, it is also true that Depending on the relative magnitudes
|Epf − Epb | ≈ 58 mV at T = 298 K. Still, of D and ks , we move from a situa-
the reverse peak current is a function of Eλ . tion in which diffusion control predom-
An important characteristic of the cyclic inates (small D, large ks ; reversible case,
voltammetric i/E curve is the current see Sect. 2.1.2.1) through a mixed-control
ratio [26] regime (both diffusion and kinetics are im-
b portant; quasi-reversible case [9, 27, 28])
ipb ip,0 iλ0
= + 0.485 + 0.086 (17) to a situation in which the rate of elec-
ipf f
ip,0 f
ip,0 tron transfer controls the overall reaction
(large D, small ks ; irreversible case [12]).
where currents with subscript 0 are This continuum of conditions is character-
measured relative to the zero current line ized in LSV or CV by the dimensionless
and iλ,0 is the current at Eλ . This ratio is quantity
unity for the simple case considered here. ks
Note that the area under the curves, ψ= √ (19)
aDred
i.e. the charges transferred in the forward
and reverse scans, are not equal, since not The dependence of voltammograms on
all starting material is regenerated during ψ has been analyzed by Matsuda and
the reverse scan and some product (ox in Ayabe [9] (LSV) as well as Nicholson
the present discussion) escapes from the and Shain [27] (CV) and characteristic
electrode. voltammograms can be found in Volume 8
If many cycles are performed, the succes- Chapter 1.
sively recorded voltammograms gradually
change until a steady state is reached since 2.1.2.4 Coupled Homogeneous Chemical
the concentrations at Estart shift between Kinetics
cycles [12]. However, only changes in peak To consider coupled chemical reactions
current are essential while peak potentials in the homogeneous electrolyte phase,
are affected to a minor extent. the diffusion equations such as Eqs. (6)
or (7) are augmented by the kinetic
2.1.2.3 Kinetics of Electron Transfer terms corresponding to the reaction of the
As a first complication, we consider a finite species whose diffusion it describes (re-
rate constant ks of the electron transfer. In action–diffusion equations). For example,
such a case, the kinetics of the electron for a reaction
±e− resulting current depends on the homoge-

−−
A
−−−
− B −−−→ C (20) neous reaction.
Many types of coupled chemical reac-
where the reduced form A of the redox tions have been considered under LSV or
couple is oxidized to B, which then reacts CV conditions, and the resulting effects on
in a homogeneous chemical reaction, the voltammetric curves have been analyzed.
governing equations become This forms the basis for both qualitative
derivation of mechanisms and quantitative
∂cA ∂ 2 cA determination of kinetic constants for elec-
= DA 2 (21)
∂t ∂x tron transfer coupled reaction steps from
∂cB ∂ 2 cB LSV and CV. Some examples are discussed
= DB 2 − kcB (22) in Volume 8, Chapter 1.
∂t ∂x
∂cC ∂ 2 cC
= DC 2 + kcB (23) 2.1.2.5 Adsorption and Surface-Bound
∂t ∂x Redox Species
In these equations, we assume that the If the redox species is confined to the im-
chemical reaction is of first order and char- mediate vicinity of the electrode surface,
acterized by a rate constant k. B disappears either by adsorption forces or by covalent
as determined by kcB , and each react- bonds (modified electrodes), there is no
ing molecule B generates one molecule need for transport from the bulk of the
C. Consequently, the concentration pro- electrolyte in order to induce the elec-
files of all species are modified, and the trochemical reaction. No concentration
profile develops. On the other hand, the
number of molecules subject to electron
Current function, √p × c⭈10−3
transfer is limited to those attached to

300 the surface. These facts lead to a char-
200 acteristic voltammetric response, see e.g.
100 Fig. 4.
0 The current–potential curve in the
−100 simplest case (Langmuir adsorption iso-
−200 therm) [29, 30] is mirror-symmetric to
−300 the line normal to the potential axis
−0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6
at E 0 with a peak potential Ep = E 0 .
Potential, E
[V]
Furthermore, the current returns to zero
(a)
after the peak. This is explained by the
fact that all surface-bound material is
Current function, √p × c
2
converted according to Eq.(3). Since no
1
diffusional loss is involved, all product
0
can be converted back to the starting
−1
−2
Fig. 4 Cyclic voltammograms for
−3 surface attached redox active molecules:
−4 (a) Langmuirian behavior;
−0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6
(b) Frumkin-type behavior with
Potential, E attractive interaction between oxidation
(b) [V] product molecules.

material in the reverse scan. Thus, the ∂c ∂ 2 c 2 ∂c
areas under the two peaks will ideally =D + (24)
∂t ∂r 2 r ∂r
be identical. The peak current for such
a system will be proportional to v rather where r is a coordinate parallel to the
than v 1/2 . radius of the electrode disk. Thus, the flux
The exact shape and quantitative fea- and consequently the current increases
tures of the voltammograms will depend over the linear case. At intermediate
on a variety of parameters, e.g. the situations a mixture of linear and spherical
adsorption isotherm followed, the sur- diffusion applies:
face concentration of the redox species, 2
or the presence or absence of inter- ∂c ∂ c 1 ∂c ∂ 2 c
=D + + (25)
molecular forces between the adsorbed ∂t ∂r 2 r ∂r ∂z2
molecules (see Fig. 4b for an exam- Here, z is the coordinate perpendicular
ple with attractive forces between prod- to the disk. The form of linear sweep
uct molecules) [31–33]. Additional effects
voltammograms changes from peak to S-
such as slow transport of counterions be-
shaped, indicating that the diffusion layer
tween bulk and the surface layer may
quickly attains a steady state with a fixed
further change the characteristics of the
thickness (Fig. 5).
voltammograms [34].
In even more complex cases, three-
dimensional forms of Fick’s law such as
2.1.2.6 Nonlinear Diffusion Eq. (26) must be used
At long timescales or for small dimensions 2
of the electrode, the assumption of linear ∂c ∂ c ∂ 2c ∂ 2c
=D + + (26)
diffusion (see Sect. 2.1.2.2) no longer ∂t ∂x 2 ∂y 2 ∂z2
holds. The form of the partial differential
equations [e.g. Eq. (6)] or the diffusion 2.1.2.7Restricted Diffusion
term therein, changes to, for example, The assumption of semiinfinite diffusion
spherical diffusion: becomes invalid if the diffusion space is
2.4
1.9
Current, i·109
1.4
[A]
0.9
0.4
−0.1
−0.5 −0.3 −0.1 0.1 0.3 0.5 0.7
Potential, E
[V]
Fig. 5 Cyclic voltammogram of ferrocene (c = 4 × 10−4 M) at an
ultramicroelectrode disk (10 µm diameter, Pt, v = 10 mV s−1 ) with
spherical diffusion.
limited, i.e. the electrolyte does not extend nanometer region [36], where, however, re-
to virtual infinity from the electrode. In producibility may be a problem and the
such situations, the boundary condition exact determination of the electroactive
at large distances from the electrode is area is difficult [37].
no longer constant with t. Rather, the The cells for voltammetric experiments
diffusing material is being exhausted. usually comprise a three-electrode ar-
Restricted diffusion is found for example rangement, with working and counter-
in thin layer cells [35] under potential scan electrodes sufficiently spaced, while the
conditions. reference electrode is brought close to
the working electrode surface with a
2.1.2.8 Simulation Haber–Luggin capillary to minimize iR
Solutions of the differential equation loss (see Sect. 2.1.3.4). The conventional
systems including initial and boundary cell volume is up to ≈20 ml, but only
conditions as discussed in the previous a small part of this total volume is af-
sections are generally provided by sim- fected during the experiment (diffusion
ulation. Mathematical models describing layer). With sufficient mixing after a scan
the basic physico-chemical processes are or cycle, a large number of voltammo-
numerically resolved [25]. LSV and CV grams can be recorded in the same
are among the main methods treated in solution without interference in most
electrochemical simulations. Further in- cases. Smaller cells for microelectrode
formation on this subject can be found use [13] or cells for high-purity appli-
in Chapter 1.3 of this Volume and (with cations in organic solvents have been
mechanistic background) in Chapter 1 of described [38].
Volume 8.
2.1.3.2 Solvents and Supporting
2.1.3 Electrolytes
Practical Considerations The whole range of solvents (aqueous
and nonaqueous) and supporting elec-
2.1.3.1 Electrodes and Cells trolytes (usually 0.1 M in order to avoid
While initially [4–7] often hanging mer- migration effects) including the popular
cury drop electrodes were employed, today tetra-alkyl ammonium salts are used in
the entire range of electrode materials, in voltammetric studies. There are no par-
particular Pt, Au, and C (glassy carbon, ticular restrictions from the voltammetric
graphite) is used for LSV and CV. Conse- conditions. It must be noted, however, that
quently, not only spherical (drop) but also for mechanistic studies high-purity elec-
planar (disk) and cylindrical (wire) electrolyte components are necessary. Since
trode geometries became important. This the substrates are often present in con-
was already recognized early [12] by si- centrations of 10−4 M or lower, even
multaneously treating planar and spherical traces of impurities may have detri-
systems. The choice of electrode material mental effects on the current–potential
strongly depends on the application. curves (see also Sect. 2.1.3.3). The choice
The characteristic size of electrodes is of solvent and supporting electrolyte
conventionally a few millimeters (radius clearly depends on the application and
for a planar disk) as a maximum. The cur- factors such as solubility of the sub-
rent minimum for disk electrodes is in the strate.
2.1.3.3 Experimental Limitations computer-controlled instruments allow

Applicability of voltammetric techniques is scan rates up to several hundred volts
in some cases limited due to several factors per second. With electrodes in the mm
which will be discussed in the following. radius range, however, often iR drop (see
In a real system, we will observe Sect. 2.1.3.4) reaches critical values mak-
the oxidation or reduction of electrolyte ing useful interpretation impossible, at
components, the electrode material itself, least in organic electrolytes. Smaller elec-
or of impurities in the electrolyte at trode sizes may be used for larger scan
certain potentials. These processes define rates (see Sect. 2.1.4.1).
the accessible potential window (Fig. 6), and Another inherent limitation becomes
the observable electrode reactions should important if the size of the diffusion
yield voltammetric signals well inside layer becomes comparable to that of the
this window to allow analysis without electrical double layer [39]. Thus, scan
interference of background contributions. rates are limited to a practical maximum
At low scan rates, the tacit assump- of ≈106 V s−1 if such effects are not taken
tion that no convection takes place in
into account [39] (see, however, Ref. [40]).
the cell breaks down. For example, in
a dichloromethane/0.1 M NBu4 PF6 elec-
trolyte at room temperature, convection 2.1.3.4 Artifacts and How to Avoid Them
effects are already observable at v ≈ Although the voltammetric experiments
10 mV s−1 . This defines a lower limit of discussed in this Chapter are rather sim-
the applicable scan rates (see also edge ple to perform and reasonable protocols
effects below). are incorporated into the control software
At high scan rates, in most cases sam- of most modern computer controlled elec-
pling rates or potentiostat rise times, limit trochemical instruments, there are several
applicability of voltammetric experiments important points to observe in order to
(see Chapter 1.2). Modern commercial avoid artifacts.
15
10
Current, i·106
5
[A]
−5
−10
−2.3 −1.8 −1.3 −0.8 −0.3 0.2 0.7 1.2 1.7
Potential, E
[V]
Fig. 6 Potential window in a dichloromethane/0.1 M NBu4 PF6 electrolyte,
potentials vs. a Ag/0.01 M Ag+ reference electrode.
Selection of starting and switching po- selection means to choose Estart well
tentials. The end points of the potential (usually several hundred millivolts) be-
scans have to be selected inside the po- fore the first peak in the voltammogram.
tential window of the electrode/electrolyte • On the other hand, the selection of Eλ
combination. Furthermore, however, two will only influence the reverse peaks.
guidelines should be observed if at all Depending on the actual problem,
possible: several choices of Eλ are meaningful
(if e.g. several peaks are present in the
• The starting potential should be located current/potential curves). In general,
at a potential where no current flows however, for quantitative analyses of
through the working electrode. Often reverse peak features, Eλ should be at
this potential can be determined as least 180 mV past the forward peak.
the ‘‘open circuit’’ or ‘‘rest’’ potential
of the solution. This ensures that the Selection of scan parameters. As men-
current/potential curve starts at i = 0. tioned earlier, the scan rate significantly
The initial condition of homogeneous influences the shape and quantitative
distribution of the redox active species features of voltammograms. Usually, a
throughout the electrolyte is fulfilled. If variation of v gives important informa-
this precaution is not taken, the product tion. There are, however, other scan-
of the electron transfer is produced at related parameters that may affect the
the electrode before the start of the scan current/potential curves. In computer
and diffuses towards the bulk electrolyte. controlled instruments, the ideal linear
Consequently, some distortion of the variation of E is often simulated us-
voltammetric curve may be observed. ing a staircase function (Fig. 7). The re-
In practical terms, the recommended sponse from such an excitation function
0.6
0.5
0.4
Potential, E
0.3
[V]
0.2
0.1
0.0
−0.1
−0.01 0.01 0.03 0.05 0.07 0.09 0.11
Time, t
[s]
Fig. 7 Staircase type potential scans as substitute for linear scans; step size
exaggerated.
is equivalent for practical purposes if the some cases. Background subtraction may
change of E during a step is small enough. thus additionally be required.
What exactly is small enough, is a complex Nonlinear diffusion. The voltammetric
function of several experimental variables behavior related to linear (e.g. at short
and the electrode reaction itself [41]. times) and spherical (e.g. at large times
Selection of potential cycles. It has been or small electrodes) diffusion has been
mentioned earlier that multicycle ex- discussed in Sect. 2.1.2. Of course, there
periments lead to a steady state cur- are intermediate situations, in which
rent/potential curve eventually (to be mixed behavior is observed, which may
distinguished from steady state micro- be regarded as a distortion of either of the
electrode curves!). Often such curves are extreme types of transport. In particular,
presented in the literature. However, quan- use of conventionally sized electrodes at
titative analysis, in particular comparison slow scan rates causes the increase of
to simulations, is greatly complicated with peak currents (normalized to v 1/2 ) with
such steady state data and pristine curves decreasing v since additional nonlinear
should be preferred. Furthermore, prod- transport of material across the edge of
uct molecules produced after several cycles the electrode occurs (‘‘edge diffusion’’ or
may induce additional chemical reactions ‘‘edge effect’’ [47]). Consequently, too slow
or show additional peaks. If it is desired scan rates should be avoided.
to study such effects, the variation of
2.1.4
the voltammogram with the cycle number
Present Methodological Developments
must be analyzed in greater detail.
Uncompensated resistance and double layer 2.1.4.1 Voltammetry at
effects. In particular, in organic solvents Ultramicroelectrodes
and at high scan rates uncompensated re- Ultramicroelectrodes (see Chapter 2.5)
sistance with the resulting iR drop and have attracted considerable attention for
double layer effects may affect the voltam- voltammetric experiments recently [39,
mograms [42]. The iR drop distorts the 48]. Such devices are defined conven-
linear E/t curve, usually assumed in LSV tionally by size (disks with diameter
or CV. Of course, in turn, the current is ≤20 µm [48]) or by physico-chemical be-
affected. While this can be controlled to havior [39]. Owing to their size, at ultrami-
some extent by the use of ultramicroelec- croelectrodes the diffusion layer attains a
trodes, it does not entirely vanish. Thus, steady state shortly after the beginning of
iR compensation [43, 44] or correction [45, the experiment. Thus, if the timescale is
46] is strongly recommended. Further de- long enough, the i/E curves under LSV
tails on this topic are given in Chapter 1.2 or CV conditions show a plateau type
(Sect. 1.2.6). Double layer charging adds shape (see Sect. 2.1.2.6). On the other
in the most simple case a constant current hand, in faster experiments, a peak still
id at constant v to the i/E curve. Since id develops. The transport rate and, conse-
is proportional to v, while the peak cur- quently, the current per unit area (current
rent ip increases with v 1/2 , id increasingly density) become very large because of
dominates the current signal at high v. the change from linear to spherical dif-
Furthermore, remaining electrolyte impu- fusion. Other processes than diffusion
rities will distort the voltammograms in may become rate controlling under these
conditions. Consequently, ultramicroelec- In such timescales, it is possible to study

trode voltammetry may be used to deter- very fast heterogeneous electron transfer
mine heterogeneous electron transfer rate rate constants [48]. Diffusion layers as thin
constants. Furthermore, the high trans- as a few nanometers are characteristic for
port rates render homogeneous chemical such fast scan rates [50]. Coupled homoge-
reactions less important and decrease their neous chemical reaction steps become less
influence on the voltammograms. important, and highly reactive intermedi-
The diffusion-controlled steady state ates can be detected [48]. The chemical re-
current at an ultramicroelectrode of radius versibility of electrode reactions increases
r0 is given by and thus redox potentials of electron trans-
fer reactions involving extremely unstable
id = 4nF Dc0 r0 (27) species become available [48].
Note that the functional relationship be- 2.1.4.3 Voltammetry Without Supporting
tween current, c0 and D in Eq. (27) is Electrolyte
different from that in Eqs. (13) and (14). The decreased iR drop in voltammetric
This has been used to determine c0 and experiments at ultramicroelectrodes has
D simultaneously from steady state and been exploited to perform electrochem-
non–steady state curves at the same ultra- istry under conditions in which no or only
microelectrode, e.g. in gel monoliths [49]. a small concentration of supporting elec-
In the following two sections, further trolyte is added and allows measurements
applications of LSV or CV at ultramicro- in low-polarity solvents (e.g. hydrocar-
electrodes are discussed in brief. bons), without the presence of excess
ions, or even in the gas phase [51]. This
topic is discussed further in Chapter 2.5
2.1.4.2 Fast Scan Techniques (Sect. 2.5.5.6). In these cases, the transport
As mentioned already (Sect. 2.1.4.1), at of charge in the electrolyte is realized by
short times scales, i.e. fast scan rates, LSV small amounts of impurities [48], by ions
or CV curves at ultramicroelectrodes still of the substrate material itself [52], or those
attain the conventional peak shape related generated in the electrode reaction [39].
to linear diffusion. Under these conditions, Thus, migration has to be considered as an
another advantage of such electrodes additional mode of transport, in particular
becomes apparent: the small electrode for multiply charged species [52]. A recent
area, resulting in a small double layer modeling study [53] has provided evidence
capacity Cd (hence a small time constant that LSV should be best suited to deal with
RCd ), and a small current i (hence a small situations of high uncompensated resis-
iR drop). Artifacts due to double layer tance as compared to chronopotentiometry
charging and uncompensated iR drop and chronoamperometry.
become less prominent, and consequently Since nonpolar solvent based elec-
high scan rates up to 106 V s−1 can be trolytes cannot be used in conventional
used [39] as compared to the limit of a voltammetric experiments, solubility prob-
few 10 V s−1 at conventional electrodes. lems may arise for some organic com-
Furthermore, specific techniques allow the pounds. Ultramicroelectrodes allow the
recording of iR-drop free voltammograms use of benzene, toluene, and even hexane
even at v > 106 V s−1 [50]. or supercritical CO2 as solvents and the
study of electrode reactions or analytical both instrumental approaches were com-

determinations in such environments is bined in the form of in situ spectroelec-
now actively studied [48]. trochemistry [55, 56] under potential scan
Furthermore, the polarity range for conditions.
mechanistic or analytical studies is con- Thus, a large number of spectra is
siderably broadened. The stability of in- recorded during the potential scan of an
termediates under such unconventional LSV or CV experiment. Each spectrum can
conditions can be investigated. be assigned to a situation of defined poten-
tial (i.e. concentration ratio at the electrode)
2.1.4.4 Voltammetry Under Hydrodynamic and to a certain time since the start of
Conditions the experiment (kinetics). If recorded fast
Usual conditions for LSV or CV experi- enough, the spectral properties of even
ments require a quiet solution in order to short lived intermediates are accessible.
Further details on spectroelectrochemical
allow undisturbed development of the dif-
measurements of this type are given in
fusion layer at the electrode. Some groups,
Chapter 3.4.
however, have purposely used the inter-
play between diffusion and convection in
electrolytes flowing in a channel or similar 2.1.4.6 Derivative Voltammetry
devices [23]. In these experiments (see also The current signal in LSV or CV is rather
Chapter 2.4), mass transport to the elec- broad even for the reversible electron trans-
trode surface is dramatically enhanced. A fer case. The width of the reversible peak at
steady state develops [54] with a diffusion half height is about 96 mV. Thus, resolu-
layer of constant thickness. Thus, such tion is relatively low and exact determina-
conditions are in some way similar to tion of the peak potential is often difficult.
It is well known that features of sig-
the use of ultramicroelectrodes. Hydro-
nals can be enhanced by examining their
dynamic voltammetry is advantageous in
derivative. Consequently, derivative cyclic
studying processes (heterogeneous elec-
voltammetry (DCV) was developed [57]. The
tron transfer, homogeneous kinetics) that
derivative is usually calculated numerically
are faster than mass transport under usual
by simple differencing (i/t) if the time
CV or LSV conditions. A recent review
step increments are small enough [58], by
provides several examples [22].
a Savitzky-Golay polynomial least-squares
procedure [59], or by Fourier transforma-
2.1.4.5 Voltammetric tion [60]. Also, hardware based differentia-
Spectroelectrochemistry tion is possible [60].
Although, in particular CV is ideally suited Figure 8 shows simulated cyclic voltam-
to detect the presence of intermediates mograms and their derivatives. The peak
in an electrode reaction and to define potentials can easily be extracted with high
the kinetics and thermodynamics of the accuracy from the derivative curve as the
reaction, it is impossible to derive di- points at which the baseline is crossed.
rect structural information about inter- It has been emphasized [61] that dis-
mediates or products from voltammetric tortions by double layer charging (see
data. On the other hand, such informa- Sect. 2.1.3.4) are eliminated in DCV
tion advantageously can be provided by curves, since their constant contribution
spectroscopic experiments. Consequently, disappears in the derivative signal.
Current differential 106

1.7 200
150
Current, i ·106
1.2
100
0.7
[A s−1]
50
[A]
0.2 0
−50
−0.3 −100
−0.8 −150
−200
−1.3 −250
−0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.00 0.02 0.04 0.06 0.08 0.10 0.12
(a) (b)
Current differential 106

1.9 240
190
Current, i ·106
1.4
140
[A s−1]
0.9 90
[A]
0.4 40
−10
−0.1
−60
−0.6 −110
−1.0 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.00 0.02 0.04 0.06 0.08 0.10 0.12
Potential, E Time, t
(c) [V] (d) [s]
Fig. 8 Cyclic voltammograms of a reversible electron transfer without (a) and with a homogeneous
follow-up reaction (c) and their derivatives (b, d).
Furthermore, DCV was employed for I (t) [63–65]. Recently, extended semiinte-
kinetic and mechanistic analyses of elec- grals with a general convoluting function
trode reactions [57, 59, 61]. The ratio of g(t) have been defined:
the derivative peak intensities (see Fig. 8d) t
is a convenient measure of the origi- m(t) = i(τ )g(t − τ )dτ (29)
nal i/E curve shape and sensitive to 0
homogeneous [57, 61] and heterogeneous where g(t) depends on the electrode
kinetics [59]. geometry and kinetics [66].
Transformation (Eq. 28) can be accom-
2.1.4.7 Semiintegral (‘‘Convolution’’) and plished by analog circuits [67], but also
Semiderivative Voltammetry several numerical algorithms have been
Another transformation of high interest provided [16, 17, 60, 68, 69].
in LSV or CV is the calculation of the The semiintegral of an LSV or CV
semiintegral [62] of the current, m(t) = voltammogram (Fig. 9) has some impor-
d−1/2 i(t)/dt −1/2 . It is characterized by the tant properties. The sigmoidal branches of
fact that applying it twice in succession, the forward and reverse part of the m(t)
the integral of the original function will be curve for a reversible redox process exactly
generated. This function is given by coincide [16, 17]. Moreover, the semiinte-
gral and in particular its maximum value
t
i(τ )1 are independent of the exact E/t rela-
m(t) = dτ = I (t) tionship under ideal conditions [46, 62].
0 (t − τ )
π 1/21/2
(28) Thus, effects of iR drop on the linearity
with the dummy variable τ , and is of the potential scan excitation function in
thus identical to the convoluted current LSV or CV disappear, when considering
1.7 200
Semiintegral, I ·109
Current, i·106
1.2
150
0.7
[As1/2]
[A]
0.2 100
−0.3
50
−0.8
−1.3 0
−0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 −0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6
(a) (b)
200
Semiintegral, I ·109
1.4
Current, i ·106
0.9 150
[As1/2]
[A]
0.4
100
−0.1
−0.6 50
−1.1 0
−0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 −0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6
(c) [V] (d) [V]
Fig. 9 Cyclic voltammograms of an electron transfer with fast (a) and slow (b) heterogeneous
kinetics and their semiintegrals (b,d).
m(t) [70]. Quasireversibility of the redox thus been used to improve resolution
process or iR drop contributions cause of closely spaced peaks with severe over-
the two branches of the semiintegral lap [72].
to split (Fig. 9d) providing an extremely
sensitive test for reversibility and the 2.1.5
absence of iR effects [70]. From the extent Applications
of separation of the two branches calcula-
tion of the uncompensated resistance was
In this section, a few examples have been
attempted [46].
selected from the literature to demonstrate
Further applications of the m(t) or
the possibilities of the experimental tech-
I (t) transform of LSV or CV curves are
niques described in the present chapter. In
found in mechanistic analysis [63–65, 71].
the examples mostly CV will be used. Apart
In particular, the fact that no a-priori
from the impossibility to detect intermedi-
assumption of the rate law for the het-
ates and products during the reverse scan,
erogeneous electron transfer is necessary
for the calculations has been stressed. similar arguments, however, apply to LSV.
Thus, deviations from the simple But-
ler–Volmer behavior can be identified [64]. 2.1.5.1 Typical Voltammograms
A sophisticated example for the use – Classical Applications
of such techniques will be discussed Figure 10 shows three selected cyclic
later. voltammograms that demonstrate the
Semiderivative or semidifferential voltam- scope of the CV technique as a tool for the
mograms appear sharper than their non- characterization of redox active chemical
transformed counterparts [60] and have compounds.
25 III
I 100
Current, i
Current, i
[µA]
[µA]
0 0
IV
II
−100
−1.0 −0.5 0.0 0.5 1.0 −1.0 −0.75 −0.5 −0.25 0.0
(a) [V] (b) [V]
0
Current, i
[µA]
−50
−100
−3.0 −2.5 −2.0 −1.5 −1.0
Potential, E
(c) [V]
Fig. 10 Typical cyclic voltammograms of (b) 3,6-bis(dimethylamino)durene, v = 0.5 V s−1
organic molecules. (a) N, N, N , (replotted from Ref. [73], in which further
N -tetramethyl-p-phenylenediamine, TMPD, at Pt experimental details can be found);
electrode, c = 0.27 mM in CH2 Cl2 /CH3 CN (c) benzylchloride (replotted from Ref. [74], in
(1 : 1)/0.1 M NBu4 PF6 , v = 0.5 V s−1 ; which further experimental details can be found).
N, N, N , N -Tetramethyl-p-phenylene- potentials. The difference in Ep for two

diamine 1 is a well-known example of a peaks related to the same redox process is
system with three stable redox states. The close to 58 mV in both cases, as expected
neutral starting species can be oxidized to for a diffusion-controlled one-electron pro-
a radical cation (also known as ‘‘Wurster’s cess (see Sect. 2.1.2.2). Equation 13 allows
Blue’’, 1•+ ) [75] and a dication 12+ (see e.g. the calculation of D = 1.6 × 10−5 cm2 s−1 .
Refs. [76 and 77]). This two-step redox pro- There are some interesting cases,
cess results in a cyclic voltammogram as in in which two successive one-electron
Fig. 10(a) with two oxidation peaks I and steps give rise to only a single
III and two corresponding reduction sig- peak couple [79, 80] (see Fig. 10b for
nals II and IV. The two peak couples I/II 3,6-bis(dimethylamino)durene). This was
and III/IV are well separated, and mirror called ‘‘inverse’’ potential ordering [79],
the increasing potential needed to gener- since in such a case oxidation of the
ate increasingly positively charged cationic intermediate is thermodynamically more
species. The E 0 of the two steps (E10 = easy as compared to that of the starting
−0.29 V, E20 = +0.34 V; both values ver- compound. In several examples, [73, 80,
sus the ferrocene/ferricinium ion standard 81] this was attributed to substantial
redox couple [78]) are simply calculated as structural changes during the redox
mean values of the corresponding peak process.
An extreme form of structural reorga- Under such conditions very thin diffusion
nization is met in dissociative electron layers (nanometers and below) are gen-
transfer [82], in which the electron trans- erated, offering the study of correspond-
fer is accompanied by bond cleavage. Thus, ingly small objects in detail. Thus, fourth
the overall reaction, e.g. generation polyamidoamide (PAMAM)
dendrimers with 64 pendant [Ru(tpy)2 ]2+
R–X + e− −−−→ R• + X− (30)
(tpy = terpyridyl) centers (radius ≈5 nm)
appears totally irreversible, as presented were adsorbed on a Pt electrode immersed
in Fig. 10(c) for benzylchloride, R = into an acetonitrile electrolyte [40, 83]. At
C6 H5 CH2 [74]. The passage from stepwise (relatively) slow scan rates (v = 36 kV s−1 )
to concerted (dissociative) mechanisms cyclic voltammograms are close to Lang-
was explored [82]. muirian shape (Sect. 2.1.2.5) with ip ∼
v, and the dendrimers act as an ad-
2.1.5.2 Selected Recent Examples sorbed monolayer on the electrode surface
This section will selectively present some (Fig. 11, top left). At high scan rates (v ≤
recent examples of LSV or CV use for 2.52 MV s−1 ) the shape of the i/E curves
the solution of chemical problems. The changes to that approaching diffusion con-
selection is by no means comprehensive trol (Fig. 11, bottom) and ip becomes
and necessarily limited to only a small proportional to v 1/2 . Hence, by adjustment
number of papers. of the diffusion layer thickness through v
The first example shows the possibilities in molecular dimensions, the ‘‘squatted’’
of ultramicroelectrodes (Sect. 2.1.4.1) at spheres of the dendrimer molecules on
extremely fast scan rates (Sect. 2.1.4.2). the surface were studied (‘‘electrochemical
10 20
2
0 0
0
−2 −10 −20
0.036 MV s−1 0.14 MV s−1 0.54 MV s−1
40
[µA]
50
50
I
0 0 0
−50
−40 −50
1.08 MV s−1 1.8 MV s−1 2.52 MV s−1
0.5 1 1.5 2 0.5 1 1.5 2 0.5 1 1.5 2
E
[V]
Fig. 11 Cyclic voltammograms of PAMAM dendrimer with 64 pendant Ru(tpy)2+ moieties
(CH3 CN/0.6 M NEt4 BF4 , Pt electrode with radius 5 µm, scan rates as indicated); (with kind
permission of Wiley-VCH from Ref. [40]).
microtome’’). The rate of electron transfer One important question in the light of
between the redox centers in the den- current electron transfer theories [85–87]
drimer and the mobility of the Ru moieties is that of the transition between step-
within the polymer were determined. wise (electron transfer and bond cleavage
The detailed analysis of LSV data using as separate elementary steps) or con-
the convolution techniques (Sect. 2.1.4.7) certed (dissociative electron transfer [88])
is demonstrated by studies concerned mechanisms. For the two extremes, one
with the electrochemical reduction of per- expects largely different activation parame-
benzoate tert-butylesters [84]. The overall ters for the electron transfer at an electrode.
chemically irreversible two-electron reduc- In particular, in contrast to the simple
tion leads, by cleavage of the O–O bond, Butler–Volmer relationship (Eq. 18) with
to tert-butyloxy anion and benzoate anion: a constant transfer coefficient, potential
dependent α values become evident. The
tBuOOCOAr + 2e− −−−→ tBuO−
experimentally accessible apparent trans-
+ ArCOO− (31) fer coefficient
150 150
100 100
[µA V1/2 s1/2]
[µA V1/2 s1/2]

i / v1/2
i / v1/2
50 50
0 0
−1.0 −1.5 −2.0 −0.5 −1.0 −1.5
E vs SCE E vs SCE
(a) [V] (b) [V]
60 60
[µA s1/2]
[µA s1/2]
I
30 30
0 0
−1.0 −1.5 −2.0 −0.5 −1.0 −1.5

(c) E vs SCE (d) E vs SCE
[V] [V]
Fig. 12 Cyclic voltammograms (curves in a and b) and the corresponding convoluted
semiintegrals (curves in c and d) for the reduction of tert-butyl-perbenzoate (a and c)
and its p-acetyl derivative (b and d) in DMF/0.1 M NBu4 ClO4 (currents in a and b
normalized to v1/2 ) at various scan rates. (Reprinted with permission from Ref. [84],
Copyright 1999 American Chemical Society.)

RT d ln k and semiintegral curves for perbenzoic
αapp = − (32)
F dE acid tert-butyl ester and the p-acetyl
derivative [84]. Calculation of αapp = f (E)
is related to the potential dependence of and analysis of this quantity revealed
the heterogeneous electron transfer rate transition from a concerted mechanism
constant k [89]. The latter in turn can be in the unsubstituted parent molecule to a
determined from semiintegrated I (t) and stepwise reaction in the p-nitro derivative.
original i(t) LSV data as [63, 64] Figure 13 shows an example in which
Ilim − I (t) UV/Vis-spectroscopy was coupled to CV
ln k = ln D 1/2 − ln (33) in order to investigate the anodic oxidation
i(t)
of 1,1-dimethyl-3-aryltriazenes. While the
with Ilim being the limiting value of p-dimethylamino derivative is reversibly
the semiintegral. Figure 12 shows LSV oxidized to a radical cation in a first step,
30
20
Current, i
[µA]
10
−10
−1.0 −0.5 0.0 0.5 1.0

Potential, E
(a) [V]
C1
B1
C2
A1 A2
0.3 4
B2
Absorbance
3
0.2
2
t
0.1 1 [s]
0
0.0
300 350 400 450 500 550 600
λ
(b) [nm]
Fig. 13 Cyclic voltammogram (a) and temporal development of
UV/Vis spectra during potential scan (b) of 3-(p-ethoxy-
phenyl)-1,1-dimethyl-triazene (with kind permission of
Wiley-VCH from Ref. [91]); further details see text and Ref. [91].
less electron-donating substituents lead to 3. J. C. Eklund, A. M. Bond, J. A. Alden et al.,

more reactive radical cations as primary ox- Adv. Phys. Org. Chem. 1999, 32, 1–120.
4. L. A. Matheson, N. Nichols, Trans. Elec-
idation products. These have been detected
trochem. Soc. 1938, 73, 193–210.
by ESR spectroscopy [90–91]. A detailed 5. J. E. B. Randles, Trans. Faraday Soc. 1948, 44,
picture of the follow-up reaction is de- 322–327.
rived from the temporal development of 6. J. E. B. Randles, Trans. Faraday Soc. 1948, 44,
UV/Vis-spectra during CV: maxima A1 327–338.
7. A. Sevčik, Collect. Czech. Chem. Commun.
and A2 correspond to the starting tri- 1948, 13, 349–377.
azene, while B1 and B2 are assigned to 8. R. S. Rodgers, Today Chem. Work 1995, 4(6),
the transient radical cation. The final re- 30–34.
action product exhibits signals C1 and 9. H. Matsuda, Y. Ayabe, Z. Electrochem. 1955,
C2 , which remain stable during the re- 59, 494–503.
10. F. G. Will, C. A. Knorr, Z. Elektrochem. 1960,
verse scan of the CV. A comparison to the 64, 258–269.
UV/Vis-spectrum of an authentic sample 11. H. Binder, A. Köhling, G. Sandstede, Chem.-
of the correspondingly substituted phenyl- Ing.-Tech. 1968, 40, 543–548.
diazonium ion proves the formation of this 12. R. S. Nicholson, I. Shain, Anal. Chem. 1964,
compound by cleavage of an N–N-bond in 36, 706–723.
13. D. H. Evans in Microelectrodes: Theory and
the radical cation. Applications, (Eds.: M. I. Montenegro et al.),
NATO ASI Ser., Ser. E, Kluwer Academic
2.1.6 Publishers, Dordrecht, 1991, pp. 17–32,
Conclusion Vol. 197.
14. J. Inczédy, T. Lengyel, A. M. Ure, A. Gelen-
Linear sweep and cyclic voltammetry are csér and A. Hulanicki, Compendium of
among the most widely used electroana- Analytical Nomenclature, 3rd ed., Blackwell
Science, Oxford, 1998, pp. 8–45.
lytical techniques for analysis of electron
15. J. M. Savéant, Electrochim. Acta 1967, 12,
transfer related reactions. They are simple 999–1030.
to apply, available in modern computer 16. A. J. Bard, L. R. Faulkner, Electrochemical
based electrochemical instruments and Methods. Fundamentals and Applications, 2nd
backed by extensive theoretical treatment. ed., Wiley, New York, 2001, p. 6.
Besides classical applications in mechanis-
Methods. Fundamentals and Applications, 2nd
tic analysis (see also Volume 8, Chapter 1), ed., Wiley, New York, 2001, p. 248ff.
advances in data treatment, ultramicro- 18. B. Speiser, Chem. in uns. Zeit 1981, 15,
electrode use, and combination with other 62–67.
techniques allow the study of molecular 19. J. Heinze, Angew. Chem. 1984, 96, 823–840;
Angew. Chem., Int. Ed. Engl. 1984, 23,
electrochemical systems in great detail. 831–847.
20. D. H. Evans, Acc. Chem. Res. 1977, 10,
Acknowledgments 313–319.
21. D. H. Evans, Chem. Rev. 1990, 90, 739–751.
I thank Silke Buchmann and Steffen Heiss 22. J. C. Eklund, A. M. Bond, R. Colton et al.,
Inorg. Chem. 1999, 38, 2005–2011.
for technical help with Figs. 1, 5 and 6. 23. J. A. Alden, R. G. Compton, Anal. Chem.
2000, 72, 199A–203A.
References 24. A. W. Bott, Curr. Sep. 1999, 18, 9–16.
25. B. Speiser, in Electroanalytical Chemistry,
1. J. Heyrovský, Chem. Listy 1922, 16, 256–264. (Eds.: A. J. Bard, I. Rubinstein), Marcel
2. J. Heyrovský, Recl. Trav. Chim. 1925, 44, Dekker, New York, 1996, pp. 1–108,
488–495. Vol. 19.
26. R. S. Nicholson, Anal. Chem. 1966, 38, 1406. 50. C. Amatore, E. Maisonhaute, G. Simonneau,
27. R. S. Nicholson, I. Shain, Anal. Chem. 1965, Electrochem. Commun. 2000, 2, 81–84.
37, 178–190. 51. J. Ghoroghchian, S. Pons, M. Fleischmann,
28. T. W. Rosanske, D. H. Evans, J. Electroanal. J. Electroanal. Chem. 1991, 317, 101–108.
Chem. 1976, 72, 277–285. 52. A. M. Bond, D. C. Coomber, S. W. Feldberg
29. R. H. Wopschall, I. Shain, Anal. Chem. 1967, et al., Anal. Chem. 2001, 73, 352–359.
39, 1514–1527. 53. J. C. Myland, K. B. Oldham, Anal. Chem.
30. E. Laviron, in Electroanalytical Chemistry, 2000, 72, 3210–3217.
(Ed.: A. J. Bard), Marcel Dekker, New York 54. R. G. Compton, P. R. Unwin, J. Electroanal.
1982, pp. 53–157, Vol. 12. Chem. 1986, 206, 57–67.
31. C. Schulz, B. Speiser, J. Electroanal. Chem. 55. W. R. Heineman, Anal. Chem. 1978, 50,
1993, 354, 255–271. 390A–402A.
32. E. Laviron, J. Electroanal. Chem. 1974, 52, 56. W. R. Heineman, J. Chem. Educ. 1983, 60,
395–402. 305–308.
33. E. Laviron, L. Roullier, J. Electroanal. Chem. 57. E. Ahlberg, V. D. Parker, Acta Chem. Scand.
1980, 115, 65–74. 1981, B35, 117–121.
34. R. W. Murray, (Ed.), Molecular Design of 58. E. Ahlberg, V. D. Parker, J. Electroanal.
Electrode Surfaces, Techniques of Chemistry, Chem. 1981, 121, 73–84.
Wiley & Sons, New York, 1992, Vol. 22. 59. H. Wang, O. Hammerich, Acta Chem. Scand.
35. A. J. Bard, L. R. Faulkner, Electrochemical 1992, 46, 563–573.
Methods. Fundamentals and Applications, 2nd 60. J.-S. Yu, Z.-X. Zhang, J. Electroanal. Chem.
ed., Wiley & Sons, New York, 2001, p. 455ff. 1996, 403, 1–9.
36. R. M. Penner, M. J. Heben, T. L. Longin 61. M. Tilset, in Energ. Organometallic Species,
(Ed.: J. A. Martinho Simões), NATO ASI
et al., Science 1990, 250, 1118–1121.
Ser., Ser. C, Kluwer Academic Publishers,
37. W. R. Fawcett, M. Opallo, Angew. Chem.
Dordrecht 1992, pp. 109–129, Vol. 367.
1994, 106, 2239–2252; Angew. Chem., Int.
62. M. Goto, K. B. Oldham, Anal. Chem. 1973,
Ed. Engl. 1994, 33, 2131–2143.
45, 2043–2050.
38. H. Kiesele, Anal. Chem. 1981, 53, 1952–1954.
63. J. C. Imbeaux, J. M. Savéant, J. Electroanal.
39. C. Amatore, in Physical Electrochemistry. Prin-
Chem. 1973, 44, 169–187.
ciples, Methods, and Applications, Monographs
64. J. M. Savéant, D. Tessier, J. Electroanal.
in Electroanalytical Chemistry and Electrochem- Chem. 1975, 65, 57–66.
istry, (Ed.: I. Rubinstein), Marcel Dekker, 65. J. M. Savéant, D. Tessier, J. Electroanal.
New York, 1995, pp. 131–208. Chem. 1977, 77, 225–235.
40. C. Amatore, Y. Bouret, E. Maisonhaute et al., 66. P. J. Mahon, K. B. Oldham, J. Electroanal.
Chem. Phys. Chem. 2001, 2, 130–134. Chem. 1998, 445, 179–195.
41. A. W. Bott, Curr. Sep. 1997, 16, 23–26. 67. K. B. Oldham, Anal. Chem. 1973, 45, 39–47.
42. J. C. Myland, K. B. Oldham, Anal. Chem. 68. J. M. Rodrı́guez Mellado, Comput. Chem.
2000, 72, 3972–3980. 1991, 15, 235–236.
43. D. Britz, J. Electroanal. Chem. 1978, 88, 69. I. Prieto, J. M. Pedrosa, M. T. Martı́n et al., J.
44. A. W. Bott, J. O. Howell, Curr. Sep. 1992, 11, 70. P. E. Whitson, H. W. VandenBorn, D. H.
21–24. Evans, Anal. Chem. 1973, 45, 1298–1306.
45. E. Eichhorn, A. Rieker, B. Speiser, Anal. 71. R. S. Rodgers, Anal. Chem. 1975, 47,
Chim. Acta 1992, 256, 243–249. 281–285.
46. A. M. Bond, K. B. Oldham, G. A. Snook, 72. A. Bobrowski, G. Kasprzyk, J. Mocák, Collect.
Anal. Chem. 2000, 72, 3492–3496. Czech. Chem. Commun. 2000, 65, 979–994.
47. J. B. Flanagan, L. Marcoux, J. Phys. Chem. 73. K. Hu, D. H. Evans, J. Electroanal. Chem.
1973, 77, 1051–1055. 1997, 423, 29–35.
48. J. Heinze, Angew. Chem. 1993, 105, 74. C. P. Andrieux, A. Le Gorande, J.-M. Savéant,
1327–1349; Angew. Chem., Int. Ed. Engl. J. Am. Chem. Soc. 1992, 114, 6892–6904.
1993, 32, 1268–1288. 75. H. D. Roth, Tetrahedron 1986, 42, 6097–6100.
49. M. M. Collinson, P. J. Zambrano, H. Wang 76. T. Yao, S. Musha, M. Munemori, Chem. Lett.
et al., Langmuir 1999, 15, 662–668. 1974, 939–944.
77. M. A. Zón, H. Fernández, J. Electroanal. 85. R. A. Marcus, Angew. Chem. 1993, 105,
Chem. 1990, 295, 41–58. 1161–1172; Angew. Chem., Int. Ed. Engl.
78. G. Gritzner, J. Kůta, Pure Appl. Chem. 1984, 1993, 32, 1111–1121.
56, 461–466. 86. R. A. Marcus, J. Electroanal. Chem. 1997, 438,
79. D. H. Evans, K. Hu, J. Chem. Soc., Faraday 251–259.
Trans. 1996, 92, 3983–3990. 87. J.-M. Savéant, Pure Appl. Chem. 1997, 69,
80. B. Speiser, M. Würde, C. Maichle-Mössmer, 269–271.
Chem. -Eur. J. 1998, 4, 222–233. 88. J.-M. Savéant, Adv. Electron Transfer Chem.
81. K. Hu, D. H. Evans, J. Phys. Chem. 1996, 100, 1994, 4, 53–116.
3030–3036. 89. S. Antonello, F. Maran, J. Am. Chem. Soc.
82. J.-M. Savéant, Adv. Phys. Org. Chem. 1990, 1997, 119, 12 595–12 600.
26, 1–130. 90. B. Gollas, B. Speiser, Angew. Chem. 1992,
83. C. Amatore, Y. Bouret, E. Maisonhaute et al., 104, 336–338; Angew. Chem. Int. Ed. Engl.
Chem.-Eur. J. 2001, 7, 2206–2226. 1992, 31, 332–334.
84. S. Antonello, F. Maran, J. Am. Chem. Soc. 91. L. Dunsch, B. Gollas, A. Neudeck et al.,
1999, 121, 9668–9676. Chem. Ber. 1994, 127, 2423–2429.
2.2 Step and Pulse Techniques 105
2.2 is an appropriate theoretical description.

Step and Pulse Techniques Nevertheless, two possible drawbacks of
digital instruments should be pointed out.
Christopher M. A. Brett and Ana Maria First, potential and current ramps have
Oliveira Brett been replaced by staircases, which may or
Universidade de Coimbra, Coimbra, Portugal may not give an equivalent response de-
2.2.1 pending on the step size. Secondly, the
Introduction response is sampled, which may lead to an
increase in noise. Both these problems and
Step and pulse techniques are now an the strategy adopted in order to circumvent
essential part of the armory of tech- them will be discussed during the course
niques employed by the electrochemist. of the chapter. Further details on analog
When first introduced, applications were and digital instrumentation can be found
essentially limited to discrete single or in Chapter 1.2.
double steps in applied potential (or The chapter begins with a theoretical
current), the response to the perturba- description of the potential step and
tion being monitored as a function of the resulting current response. This is
time – chronoamperometry (or chronopo- then extended to a double potential step.
tentiometry). There were instrumental dif- Following this, pulse and square wave
ficulties in implementing anything more techniques are introduced and described,
complex, as has happened with the first together with applications. Current step
version of square wave polarography in the techniques are then discussed together
1950s [1]. A number of current sampling with their application.
strategies were developed in connection
with dropping mercury electrodes and syn- 2.2.2
chronization with drop growth, and this Potential Step
led to the advent of normal pulse and dif-
ferential pulse polarography in the 1970s, All potential pulse techniques are based
which were the first widely available pulse on combinations of potential steps. It is
techniques [2]. Nevertheless, the instru- therefore of importance to understand the
mentation permitted very little flexibility current response to potential steps as a
in experimental conditions. function of time. Further details can be
Developments in microprocessor-based found in Ref. [3].
instrumentation completely changed the
situation. Nowadays, nearly all commercial 2.2.2.1 Single Potential Step
instruments – potentiostats and galvanos- We begin with the derivation of the re-
tats – are digitally based, which means that sponse to a potential step from a value at
the programming of an almost infinite which no current flows, to one at which the
variety of step and pulse waveforms has diffusion-limited current passes, Fig. 1(a).
become relatively easy to carry out. This The response of the current to this pertur-
opens up a multitude of exciting possibil- bation will be a sharp change from zero
ities to the experimentalist, since he can current followed by relaxation to a value
adapt the applied waveforms to the ki- close to zero, the final steady state magni-
netics and mechanism of the electrode tude of which is determined by the flow
reaction under study, so long as there of species to the electrode surface. As will
Reaction of all species

reaching the electrode
|E |
No electrode
reaction t
(a) t =0
[t ]
I
(b) t
Fig. 1 Potential step chronoamperometry: (a) schematic
application of potential step and (b) chronoamperometric
response.
be seen, this varies according to electrode gradient changes with time, it is necessary
geometry and solution convection. In this to solve Fick’s second law to obtain the
section, we will consider only diffusion in- chronoamperometric response
duced by concentration gradients of the
electroactive species. ∂c ∂ 2c
=D 2 (2)
For a uniformly accessible planar elec- ∂t ∂x
trode, the diffusion process is known as
semi-infinite linear diffusion, since it can The boundary conditions are
be assumed to occur only in one dimension
t = 0 No electrode reaction c0 = c∞ (3a)
perpendicular to the electrode surface. The
observed current depends directly on the t ≥0 Bulk solution lim c = c∞ (3b)
x→∞
concentration gradient at the electrode sur-
face, (∂c/∂x)0 . According to Fick’s first law t > 0 Diffusion-limited current, Id ,
at electrode surface (3c)
∂c
I = nFAD (1)
∂x 0
The mathematical solution of this equation
where x is the perpendicular distance is most easily carried out using Laplace
from the electrode. Since the concentration transforms and leads to the expressions
• for the variation of concentration with that if the sphere is large (correspond-
distance from the electrode ing to small 1/r0 ), then the effects of
sphericity are small and there is essen-
x
c = c∞ 1 − erfc (4) tially only semiinfinite linear diffusion.
x(Dt)1/2
The other extreme represents the situa-
• for the concentration gradient at the tion at very small electrodes and a steady
electrode surface state current is predicted. The reason
for this is that a very steep concentra-
∂c 1
= (5) tion gradient is created, which sucks in
∂x 0 (πDt)1/2 electroactive species; this situation occurs
for microelectrodes.
The calculated current is therefore
Below the transport-limited current,
nFAD 1/2 c∞ kinetics have to be taken into account.
I (t) = Id (t) = (6)
(πt)1/2 The relevant expressions for the current
are given in Table 1 for both planar and
which is known as the Cottrell equation [4]. spherical electrodes, considering that only
Note that the current decays with the the reduced species is initially present
square root of time, as illustrated in in solution. This is what is usually
Fig. 1(b). encountered in practice. Details of the
At a spherical electrode of radius r0 , derivation of these expressions may be
which corresponds to a first approximation
found in Ref. [3].
to a mercury drop electrode, the form of
A point that has not been considered so
Fick’s second law to solve is
far, but which becomes crucial on using
2
∂c ∂ c 2 ∂c pulse techniques that are combinations
=D + (7) of potential steps, is that of capacitive
∂t ∂r 2 r ∂r
currents. These are predominantly due to
and the boundary conditions are changes in the charge of the electrolyte
double layer on applying a potential step
t = 0 r ≥ r0 No electrode reaction c0 = c∞
E. If a constant double layer differential
(8a) capacity, Cd , is assumed, then the charging
t ≥0 Bulk solution lim c = c∞ current is given by
r→∞
(8b)
E −t
t > 0 r = r0 Diffusion-limited current, IC = exp (10)
R RCd
Id , at electrode surface (8c)
where R is the solution resistance.
The calculated diffusion-limited current is Whereas R is independent of electrode
area, Cd is directly dependent and so
1 1
Id (t) = nFADc∞ + (9) IC is proportional exponentially to the
(πDt)1/2 r0
electrode area. Since the voltammetric cur-
It can be seen by comparison with rent is directly proportional to electrode
Eq. (6) that this is the Cottrell equa- area, the ratio If /IC increases with de-
tion together with a correction term for creasing electrode area. This is a good
spherical diffusion. The reason for these reason for using microelectrodes (Chap-
two terms can be easily understood in ter 2.5).
Tab. 1 Current responses following application of a potential step to the system R → O + ne− with
only R initially present in solution
Current [I]
nFADc∞
Reversible Plane
(1 + θ)(π Dt)1/2

nFADc∞ 1 1
Sphere +
(1 + θ) (π Dt)1/2 r0
1/2
k2 t kt
Irreversible Plane nFAkc∞ exp erfc
D D1/2
1/2
D k2 t kt D
Sphere nFAkc∞ 1 + exp erfc 1/2
−
kr0 D D kr0
1/2
2kt
Irreversible Plane and sphere nFAkc∞ 1 −
(small t) (π D)1/2

θ = exp[(nF/RT)(E − Eθ )] k rate constant of electrode reaction
2.2.2.2 Double Potential Step There are situations in which interfer-

The second step toward the development ence effects can be removed by the use
of a pulse technique is that of a dou- of a double potential step. For example,
ble potential step in which the potential if two electroactive species are oxidized at
is altered between two values. The sim- the same potential and only one of them
plest case is to return to a value at can be reduced, then more information
which no electrode reaction occurs, the can be gained.
step being inverted after a time τ . The
expressions that are obtained for initial 2.2.2.3 Potential Step Chronocoulometry
oxidation of a species of concentration c In the previous subsections, the chrono-
to the region of the diffusion-limited cur- amperometric response to a potential step
rent are or steps has been evaluated. However,
experimentally it may be advantageous
nFAD 1/2 c∞
t <τ I= (11) to study the variation of charge with
(πt)1/2
time – chronocoulometry [6]. Although it
t >τ I = nFAD 1/2 c∞ may appear that integration of a chronoam-

1 1 perometric signal after the experiment,
× − which is easy to do in a modern dig-
(πt)1/2 π(t − τ )1/2
(12) itally based potentiostat, is equivalent,
showing that the backward response is there are subtle differences that can be
superimposed on the continuation of the useful and concern the reduction of sig-
forward response (Fig. 2). There is thus a nal noise if the charge is registered
memory effect that also appears in square directly. In particular, the recording of
wave voltammetry (SWV). Expressions for the charge leads to improved signal-to-
kinetic control in one of the two steps and noise ratios toward the end of the tran-
in both steps exist in the literature [5]. sients when the current is close to zero.
Fig. 2 Double potential step:

(a) potential-time waveform and
(b) current response.
E
0 t t
(a)
I
0 t t
(b)
Additionally, integration reduces signal A double potential step, following from

noise in general. Eq. (12), leads to
If a potential step is applied from a
region of zero current, the expressions 2nFAD 1/2 c∞ 1/2 1/2

Qt>τ = t − (t − τ )
given in Table 1 may be integrated to give π 1/2
the corresponding charge, Q. (14)
In the case of a step to the diffusion- This can be particularly useful because
limited current region, the integrated when t > τ , there is no net capacitive
Cottrell equation is obtained: contribution to the charge, that is, the
difference between the charges at time
(t − τ ) and time t. Thus, the adsorption
2nFAD 1/2 c∞ t 1/2
Qt = (13) phenomena can be probed. For this, the
π 1/2 charge difference, Qdif , between Qτ and
Qt>τ is necessary and is given by
When capacitive charges – and possi-
bly oxidation or reduction of adsorbed 2nFAD 1/2 c∞
species – are taken into account, there may Qdif =
π 1/2
be a nonzero intercept (see upper line in
Fig. 3). × τ 1/2 − t 1/2 + (t − τ )1/2 (15)
Qt (t < t)
nFAΓ0
t 1/2
Qdif(t > t)
Fig. 3 Double potential step chronocoulometry, showing deduction of

surface concentration of adsorbed species. See text for details.
Figure 3 shows plots of Qt vs t 1/2 and of equal height is applied to the electrode,
of Qdif vs t 1/2 and how the surface and the current is sampled toward the
concentrations, 0 , of adsorbed species end of each step of width τ (Fig. 4). It
may be deduced. was originally devised to work in conjunc-
tion with the dropping mercury electrode
2.2.3 (tast polarography), and here the periodic
Staircase Voltammetry renewal of the electrode at the end of
each step precluded any memory effects
The main importance of staircase voltam- between current sampling in the consecu-
metry, in the context of this chapter, is that tive steps.
it forms an integral part of many widely However, at a nonrenewed electrode,
practiced pulse techniques. In staircase the response to a staircase waveform of
voltammetry, a sequence of potential steps this kind is similar to that from a linear
∆Es = 10/n mV
E
t
Fig. 4 Staircase voltammetry waveform.
ramp. In modern digital potentiostats, steps of varying height and in forward or

linear sweep, and cyclic voltammetry are reverse direction, with current sampling.
usually implemented as a staircase wave- Much of the important work in this
form (Fig. 4). In this context, it is therefore area has been summarized in literature
important to quantify the conditions under reviews [11–13].
which staircase and linear sweep/cyclic re- As seen in previous sections, the re-
sponses are equivalent. It has been shown sponse to a potential step is a pulse of
in Ref. [7] for a simple electrode process current, which decreases with time as
that these criteria are that the current the electroactive species near the elec-
is sampled after 0.25 of the step width trode surface is consumed and consists
(τ /4) and for more complicated cases of a faradaic, If , and a capacitive contri-
from τ /4 up to τ /2 [8]. Obviously there bution, IC . The advantage of most pulse
is also a dependence on step height, Es . techniques results from the measurement
Whatever the mechanism of the reaction, of the current flow near the end of the
theory shows that the differences between pulse when the faradaic current has de-
sampling times become insignificant if cayed, often to a diffusion-limited value
nEs < 0.26 mV [9], a situation that is not but when the capacitive current is in-
easy to accomplish in practice, since a sin- significant. Pulse widths, tp , are adjusted
gle linear sweep scan of 500 mV would to satisfy this condition and the addi-
require over 1900 current sampling points. tional condition that time has not been
In fact, the error is usually not signifi- allowed for natural convection effects to
cant for nEs < 5 mV and with sampling influence the response. There is a greatly
times close to the end of the step, this later improved signal-to-noise ratio (sensitivity)
sampling time ensuring a lesser, hopefully compared to steady state techniques and in
zero, capacitive contribution. Often, higher many cases, greater selectivity. Detection
values are permitted by the instruments, limits are of the order of 10−7 M. Fur-
although the justification for treating the thermore, for analytical purposes, most
experiment as linear scan in such cases, current-voltage profiles from the pulse
that is, whether the results should be an- techniques are faster to interpret than
alyzed as staircase voltammetry, should those of dc voltammograms, because they
be addressed. Either a true linear ramp are peak-shaped rather than the typical
must be used or the current-time profile step curve of conventional voltammet-
properly analyzed as a staircase voltam- ric methods.
mogram (see Sect. 2.2.2). Particular care Pulse voltammetric techniques, most
must be exercised when studying adsorp- used in electrochemistry, are normal
tion phenomena, although recognizing the pulse voltammetry (NPV) and differen-
existence of a pulsed response can be used tial pulse voltammetry (DPV). In square
to good advantage for measuring rate con- wave voltammetry (SWV), there may be
stants of adsorbed species [10]. a non-faradaic contribution to the in-
dividual currents but the current sam-
2.2.4 pling strategy essentially eliminates this
Pulse Techniques through subtraction, as will be seen
in Sect. 2.2.4.3. SWV was pioneered by
Many pulse techniques have been devised Barker [1] in the 1950s, but due to in-
on the basis of a succession of potential strumentation development only 40 years
later did its use by electrochemists for reaction occurs. The potential increment,
analytical, kinetic, and mechanistic work E, between successive pulses is constant,
become widespread. and sufficient time should be left between
pulses for the current on the reverse step to
2.2.4.1 Normal Pulse Voltammetry (NPV) die away to zero and for all concentration
NPV involves the imposition of square gradients to disappear. The current is
wave voltage pulses (i.e. potential steps of measured near the end of the pulse and
fixed time duration) of increasing height a plot made of the successive current
upon a constant value of applied potential sampled points against the potential of the
(Fig. 5a). This base potential, Ebase , is applied pulses. This gives a voltammetric
chosen where I = 0, that is, no faradaic profile of the same form as a steady state
∆E ~ 10/n mV
E
Ebase
t′
(a) t
I
(b) E
Fig. 5 Normal pulse voltammetry: (a) potential-time waveform and (b) schematic voltammogram.
voltammogram, Fig. 5(b), as can also be may be useful. For this, the base poten-
seen from the expressions in Table 1, tial is fixed at the potential corresponding
which are directly applicable. Effective to the diffusion-limited reaction and the
sweep rates, determined by the staircase pulses are applied in the reverse direc-
period, τ , are generally of the order of 1 tion. This is called reverse pulse voltammetry
to 10 mV s−1 , corresponding to intervals (RPV).
between pulses of 2 to 4 s; pulse widths The influence of reactant adsorption
usually vary between 5 and 100 ms. on the shape and height of the NP
The maximum obtainable current is voltammogram has been investigated and
thus given by the Cottrell equation is conveniently summarized in Ref. [14].
nFAD 1/2 c∞
Imax = (16) 2.2.4.2 Differential Pulse Voltammetry
(πtm )1/2 (DPV)
where tm is the sampling time after In DPV, the difference between two
application of the pulse. If tm is constant sampled currents is measured, registered
throughout the NP scan for a given just before the end of the pulse and
electroactive species, then the current is just before pulse application. In the first
directly proportional to concentration. instruments to offer this technique, the
A pulse leads to higher mass transport pulses were superimposed on a linear
than a hydrodynamic electrode (Chap- ramp of potential. However, in digital
ter 2.4) owing to the sharply induced potentiostats, it is simpler to superimpose
concentration gradient, so that kinetic ef- the pulses on a staircase waveform. Thus,
fects may be more visible. In particular, an the base potential is incremented in a
electrode reaction that appears reversible staircase (Fig. 4) and the pulse, of constant
at a hydrodynamic electrode may become height, is of width 10 or more times smaller
quasi-reversible under NPV conditions. than the period of the staircase waveform
Another important advantage of the use (Fig. 6a). If these conditions are followed,
of pulses starting at the base potential is the staircase width is generally the same
that irreversible adsorption of the product as for NPV, so that once again the effective
of the electrode reaction will be much re- sweep rate is of the order of 1 to 10 mV s−1 .
duced and the short timescale may mean The difference between the two sampled
that effects of coupled homogeneous reac- currents (I (2) − I (1)) is plotted against
tions are not observed. the staircase potential and leads to a peak-
The NPV technique was first developed shaped waveform, as shown in Fig. 6(b).
as NP polarography (NPP) at dropping The peak for a reversible system occurs
mercury electrodes. In these cases, the at a potential
pulses are synchronized with the drops,
E
the pulse being applied near the end Ep = E1/2 − (17)
of drop life when the increase in the 2
surface area of the drop is least. Current where E is the pulse amplitude (with
sampling in NPP also serves to remove the sign included). The reason for the differ-
oscillations resulting from drop formation. ence between Ep and E1/2 for fast kinetics
When there are parallel electrode reac- is that the current is plotted as a function of
tions of the initial electroactive species or Ebase and not as a function of the potential
interfering species, a modification of NPV halfway up the pulse.
∆Es = 10/n mV
t
+
50/n mV
2
E
1 t′ t ~ 0.5 → 5 s
(t − t′) ~ 5 → 50 ms
t
(a) t
1.0
I /Ip
0.5
W1/2
0.0
−0.2 −0.1 0.0 0.1 0.2

n ( E − E p)
(b) [V]
Fig. 6 Differential pulse voltammetry: (a) potential-time waveform: sum of
staircase and synchronized pulses and (b) schematic voltammogram.
The corresponding current, Ip , is given decreases. Values of E greater than

by the expression 100 mV are not useful because the peak
width increases with the pulse amplitude.
nFAD 1/2 c∞ 1 − σ Many applications have been described.
Ip = −
(πtm )1/2 1+σ An interesting example is the oxidation

1−σ of the bases of single strand DNA, in
= ICot (18) which an oxidation scan followed by a
1+σ
reduction differential pulse scan clearly
where show the adsorption and blocking of the
nF E
σ = exp (19) electrode surface [15]; the relatively slow
RT 2
timescale of the DP scan is appropriate for
and ICot is the current in the Cottrell this purpose.
equation at the sampling time tm (Eq. 16). The effect of adsorption of reactants on
Two extreme cases of this expression can peak shape and peak potential has been
be considered: thoroughly studied and is summarized in
Ref. [14].
• Large |E|, σ 1. The term (1 −
σ )/(1 + σ ) becomes ±1 (reduction or
oxidation, respectively), equal to the 2.2.4.3 Square Wave Voltammetry (SWV)
value for NPV (Eq. 16), and Ip = ICot . The form of SWV most electrochemists
• Small |E|. By expansion of use today is based on work dating from
(1 − σ )/(1 + σ ), one finds σ → 1969 [16] and principally developed by
(nFE/4RT). Osteryoung and coworkers using large
amplitude pulses [11, 17], such that a steady
state is not achieved, and it has taken over
n2 F 2 AD 1/2 c∞ a large part of the traditional domain of
Ip = − E
4RT (πtm )1/2 application of DPV.
The potential waveform consists of a
nFE
= ICot (20) square wave superimposed on a staircase
4RT
(Fig. 7). The current at the end of the
This latter equation is valid if E is less forward pulse, If , and the current at
than (2RT/nF) but appears to be applicable the end of the reverse pulse, Ib , are
for larger pulse heights. both registered as a function of staircase
The expression for the half-width at half- potential, which is midway between the
height, W1/2 potentials corresponding to the forward
and backward potential steps. Usually the
3.52RT
W1/2 = (21) net current, Inet , that is (If − Ib ), is plotted
nF against E. Since If and Ib usually have
is equal to a value of 90.4/n mV at 25 ◦ C, opposite signs, the difference, Inet , is
showing that peaks separated by 50/n mV larger than each individual component in
can be resolved. the region of the peak that is centered
As the system becomes more irre- on the half-wave potential. At potentials
versible, the difference between E1/2 for corresponding to the diffusion-limited
a reversible system and Ep increases, current, Inet = 0. Examples of square wave
the peak becomes wider and its height responses are shown in Fig. 8.
∆E s
t
+
∆Esw
∆E s
E
2
tp
t
Fig. 7 Square wave voltammetry potential-time waveform: sum of staircase and synchronized
square wave.
Over the small potential range between until IC dies away, which is a disadvantage
forward and reverse pulses, the interfacial of techniques such as fast scan cyclic
capacity is normally constant, and thus voltammetry. Consequently, higher scan
calculation of the difference between rates can be used in SWV because
responses from backward and forward it effectively removes the background
pulses effectively annuls the capacitive current from the measurements. This
contributions. It is not necessary to wait means that there is the possibility of
1.0 1.0
0.8 0.8
0.6
0.6
0.4
0.4
0.2
I /Ip, net
I /Ip, net
0.2
0.0
0.0
−0.2
−0.2
−0.4
−0.4 −0.2 0.0 0.2 0.4 −0.4 −0.2 0.0 0.2 0.4
(E − E 0′) (E − E 0′)
(a) [V] (b) [V]
1.0 1.2
0.8 1.0
0.6 0.8
0.6
0.4
I /Ip, net
I /Ip, net
0.4
0.2
0.2
0.0
0.0
−0.2
−0.4 −0.2 0.0 0.2 0.4 −0.4 −0.2 0.0 0.2 0.4
(E − E 0′) (E − E 0′)
(c) [V] (d) [V]
Fig. 8 Square wave voltammogram profiles for R → O + e− . (a) Reversible reaction and k0 = (b) 0.3, (c) 0.1, and (d) 0.03 cm s−1 . Esw = 50 mV,
2.2 Step and Pulse Techniques
f = 50 Hz, αc = 0.5. Dotted lines show forward and backward currents and solid line the net current.
117
investigating kinetic systems, in particular Because there is a memory effect

with fast reactions. The three different between each potential step, the current
plots of If , Ib , and Inet may also be overlaid, is given for each value of m at time tp , after
the resulting combination plot, such as application of the step by
those in Fig. 8 having been suggested as
the most useful form for investigating
m
Qi − Qi−1
kinetics and mechanisms [18]. Im = ICot = ICot
p
(m − i + 1)1/2
SWV is characterized by four parame- i=1
(24)
ters: square wave period, τ , pulse width,
in which Qi = θi /(1 + θi ), assuming equal
tp = τ /2, step height, Es , and pulse
diffusion coefficients for oxidized and
height, Esw . The pulse width is related
reduced species, and is defined as in
to the square wave frequency, f = 1/(2tp )
Table 1 for potential step. It can be seen
and as the staircase step at the beginning
that
p is a dimensionless current relating
of each cycle is Es , it means that the ef-
the observed current to that which would
fective scan rate is υ = Es /2tp = f Es .
be obtained from the Cottrell equation.
A square wave can be described by
Subtraction of neighboring values of Im
the equation
leads to the difference current. The peak
E = −(−1)m Esw (22) current is given by
and in combination with a staircase Ip = ICot

p (25)
waveform, this leads to
For a reversible reaction, the peak current
m+1 is symmetrical about the peak potential,
Em = Ei ± Int −1
2 which is identical to the reversible half-
× Es µ(−1)m Esw (23) wave potential. In this case, the form of
the peak is the exact differential of the
where the upper signs refer to a positive- steady state voltammetric curve, as would
going scan and the lower to a negative- be expected.
going scan, and ‘‘Int’’ refers to the integer Some representative values of
p are
part of the expression. Thus, odd values shown in Table 2. It can be seen from the
of m refer to the forward pulse and even form of this equation that a peak shape
values of m to the backward pulse. will be obtained.
Tab. 2 Peak current dimensionless scaling factors,

p , for
square wave voltammetry (see Eq. 25) [10]
nEsw [mV] nEs [mV]

1 5 10 20
0 0.0053 0.0238 0.0437 0.0774

10 0.2376 0.2549 0.2726 0.2998
20 0.4531 0.4686 0.4845 0.5077
50 0.9098 0.9186 0.9281 0.9432
100 1.1619 1.1643 1.1675 1.1745
The pulse width, tp , is usually varied over approaches for more complex cases have
a large range 1 to 500 ms, that is, f = 1 been developed [24–26].
to 500 Hz, and defines the experimental An important advantage of SWV for
timescale, whereas Es is much smaller many electroanalytical applications is that
than Esw . Recommended values are in the negative potential region, oxygen
Es = 10/n mV for staircase step height, does not have to be excluded from solu-
and Esw = 50/n mV for pulse height. tion when using SWV, unless it directly
Thus, the pulses can be shorter than in interferes with the electrode reaction un-
DPV or NPV and a broader spectrum of der study for two reasons. First, in the
scan rates can be applied. Since effective limiting-current region for oxygen reduc-
sweep rates of 1 to 10 mV s−1 , as in DPV, tion, the forward and reverse currents
to scan rates of 1 V s−1 or higher can be are equal leading to a zero net current.
employed, this enables the achievement Second, during a scan from negative po-
of kinetic and analytical data with the tentials in the positive direction, the fast
same technique. Because of the lower effective scan rate means that no electroac-
consumption of electroactive species by tive oxygen species have time to diffuse to
reaction in relation to DPV, there are less the electrode surface from bulk solution.
problems of blocking the electrode surface. Thus, the experimental time is less and
However, care must be taken with the procedures are simplified because prior
interpretation of SWV data, as illustrated bubbling of nitrogen or argon in the solu-
in Fig. 8. Figure 8(a) shows profiles for tion is avoided. This is used to good effect
the forward and reverse and net currents in anodic stripping voltammetry, in par-
for a reversible reaction. If the electrode ticular in combination with square wave
kinetics are slow, then the forward current injection techniques [27, 28] or with ultra-
is shifted to higher overpotentials than the sound [29] in which simplifying protocols
reverse current, which becomes relatively is a useful advantage.
smaller, at the same time as the peak width SWV also allows the possibility of
becomes greater. This is demonstrated in measurements over a wide dynamic range,
Figs. 8(b) to (d). defined as its ability to respond to both high
In a multistep reaction with a first and low concentrations of electroactive
rate-determining quasi-reversible step, two species. Additionally, the rapidity means
peaks may appear. Unless the individual that successive voltammograms can be
forward and backward current profiles are recorded and a voltammetric profile with
monitored, the erroneous conclusion that time registered, for example, for analysis
two species are reacting could be reached. of eluents from a high-pressure liquid
Similarly, if there is adsorption, many chromatography column.
different nuances of which have been
investigated [19–22], then effects on the 2.2.4.4 Other Pulse Techniques
peaks become apparent, and the reverse There are several other pulse techniques
peak can entirely disappear. The response that have been developed for particular
at mercury film electrodes has also been situations, but are not widely employed.
deduced and tested [23]. Fortunately, fit- They include:
ting and simulation software packages
exist that can aid in the task of unravelling • Reverse pulse voltammetry. The base po-
mechanisms and kinetics and theoretical tential is held at a value in which
all electroactive species react at the is that the time during which electrode
diffusion-limited rate. Pulses of increas- reactions can occur is much larger and
ing height are then made, in the back- this may cause complications, especially
ward sense, so this has similarities with from adsorption and electrode blocking.
NPV. This can reduce problems of inter- SWV seeks to address the disadvantages
ference or parallel electrode reactions. of both NPV and DPV by being ‘‘differen-
• Differential normal pulse voltammetry. tial’’ and also by being fast to minimize
As presently used, in this waveform adsorption. It cannot always reach this
each pulse corresponding to NPV (see goal. Indeed, there are situations in which
Fig. 5a) is divided into two equal parts SWV is too fast, given the slow kinetics of
with a small increase in potential in the some reactions, to be able to register any
second half of the pulse. The difference electrode response. In these cases, DPV
between the currents sampled halfway must be employed.
through and at the end of the main pulse Many instruments allow the user com-
is displayed as a function of the potential plete freedom to choose the waveform pa-
of the first half of the pulse. Advantages rameters and to cover the whole transition
relative to DPV are that less time is spent range from DPV, as originally conceived,
at potentials in which adsorption and so to SWV by reducing the interval between
on can occur. DPV pulses until it becomes equal to the
pulse width. In this sense, SWV can be
For a specific reaction scheme, it is now rel- thought of as a special case of a DPV wave-
atively simple to program a potentiostat in form when the interval between pulses
the desired fashion and many waveforms is reduced until it is equal to the pulse
have appeared in the literature. width. However, although the waveform
construction is very similar, the reasoning
2.2.4.5 Comparison between Potential behind the conception of the techniques
Pulse Techniques was rather different as has been previously
It is useful to compare the relative discussed, particularly regarding effective
advantages of the most used potential scan rate and elimination of capacitive con-
pulse techniques. tributions. Another important difference is
The advantage of NPV is short electroly- that in SWV the potential, at which the cur-
sis time, the main disadvantage being that rent is registered on SWV, is the staircase
the charging current becomes larger as the potential shifted in the scan direction by
pulse becomes of larger amplitude. |E/2| relative to DPV (see consequences
Concerning DPV, residual capacitive in Eq. 17, for example).
contributions will be subtracted out more
effectively, since the pulses are of equal 2.2.5
height, which means that DPV is better Current Step and Chronopotentiometry
than NPV in the majority of situations,
although the faradaic sensitivity is theoret- In current step techniques, a step in ap-
ically the same. Also, the peak response of plied current at an electrode causes a
DPV is more useful for resolving the sig- change in the potential. After charging
nals due to two species with close half-wave the double layer, faradaic reactions will
potentials than wave shape of NPV re- begin to occur, giving rise to a chronopo-
sponse. The disadvantage relative to NPV tentiogram.
2.2.5.1 Variation of Potential with Time and in which

At a planar electrode, the equation to be
RT DR 1/2
solved under conditions of linear semi- Eτ/4 = E θ + ln (31)
infinite diffusion is the same as for nF DO
a potential step. The difference is the The parameter Eτ/4 is identifiable with
third boundary condition, which instead the reversible half-wave potential in a
of defining the diffusion-limited current voltammogram and is thus usable as an
expresses the concentration gradient re- identification parameter for the electroac-
sulting from the applied current at the tive species in mixtures (see Fig. 9a). A
electrode surface. plot of E vs lg [(t 1/2 –τ 1/2 )/t 1/2 ] gives a
Thus the equation to solve is straight line of slope 59.2 mV at 25 ◦ C
for a reversible system (Fig. 9b). Another
∂c ∂ 2c simple and useful diagnostic parame-
=D 2 (26)
∂t ∂x ter for a reversible system at 25 ◦ C is
|E3τ/4 – Eτ/4 | = 47.9/n mV.
with the boundary conditions Table 3 shows these expressions and
those obtained for quasi-reversible and
t =0 No electrode reaction c0 = c∞
irreversible systems. Further details are
(27a) given in Ref. [3].
t ≥0 Bulk solution lim c = c∞
x→∞
(27b) 2.2.5.2 Current Steps and Potentiometric

∂c Stripping Analysis
t >0 I = nFAD (28) The derivatives of the expressions in
∂x 0
Table 3 are easily calculated and are
The solution to the equation is
dE RT τ 1/2
=±
I τ 1/2 nFAD 1/2 π 1/2 dt r 2nF t (τ 1/2 − t 1/2 )
= (29)
c∞ 2 dE RT t 1/2
=± (32)
dt irr 2αnF t (τ 1/2 − t 1/2 )
which is known as the Sand equation [30].
The transition time, τ , corresponds to which correspond to minimum values of
the situation when c0 = 0 when all the
dE 27 RT
electroactive species in the zone of the =± t = 4τ/9
electrode have been consumed. If this is dt r,min 8 nF τ

obeyed (i.e. a simple electrode process), dE 2RT
then the quantity (I τ 1/2 /c∞ ) is constant. =± t = τ/4 (33)
dt irr,min αnF τ
It can be shown, for a reversible system,
by substituting the concentrations of O Thus the transition time can be calculated
and R at the electrode surface together with from the derivatives.
the Sand equation in the relevant Nernst A plot of (dt/dE) vs E, that is, the inverse
equation that the variation of potential with expressions, leads to peaks that identify
time for an oxidation is given by the species and the height of which are di-
rectly concentration dependent, since the
RT τ 1/2 − t 1/2 transition time depends linearly on the
E = Eτ/4 − ln (30) concentration of electroactive species. This
nF t 1/2
0.2
0.1
n(E − Et/4)
[V]
0.0
−0.1
0.00 0.25 0.50 0.75 1.00
t
(a) [τ]
0.1
n(E − Et/4)
[V]
0.0
−1 0 1 2
(b) lg(t1/2 − t 1/2)/t 1/2
Fig. 9 Response to a current step for a reversible system, illustrated as an
oxidation: (a) chronopotentiogram and (b) plot of E vs lg [(t1/2 –τ 1/2 )/t1/2 ].
Tab. 3Variation of potential following application of a current step to the system

R → O + ne− with only R initially present in solution
Potential (E) |E3τ/4 − Eτ/4 |[mV]
RT τ 1/2 − t1/2
Reversible Eτ/4 − ln 47.9/n
nF t1/2

RT 2k0
Irreversible Eτ/4 − ln (τ 1/2 − t1/2 ) 33.8/αn
αa nF (π D)1/2
k0 standard rate constant of electrode reaction
Fig. 10 Plot of dt/dE vs E from

chronopotentiograms as
appears in potentiometric
stripping analysis.
dt/dE
−0.1 0.0 0.1

n (E − Et/4)
[V]
concept is used in potentiometric stripping useful electroanalytical data. The number

analysis (PSA) [31]. After accumulation of of publications on these themes using
the trace species to be measured on the pulse techniques is very large so it is
electrode, reoxidation is carried out either not possible here to refer to the many
by an oxidant in solution or by an ap- studies undertaken in this area, and the
plied current. Originally the results were interested reader is directed to the special-
presented in the form of chronopoten- ist literature.
tiograms such as those in Fig. 9(a), but
now are commonly displayed as a plot References
of (dt/dE) vs E (Fig. 10). This makes
identification and quantition simpler and 1. G. C. Barker, I. L. Jenkins, Analyst 1952, 77,
is more similar to results obtained from 685–696.
differential pulse and square wave voltam- 2. J. B. Flato, Anal. Chem. 1972, 44, 75A–87A.
3. C. M. A. Brett, A. M. Oliveira Brett, Electro-
metry. Further details on PSA, sometimes chemistry. Principles, Methods, and Applica-
termed stripping potentiometry, are given tions, Oxford University Press, Oxford, 1993,
in Chapter 2.3 (Sect. 2.3.3). Chap. 10, pp. 199–223.
4. F. G. Cottrell, Z. Phys. Chem. 1902, 42, 385.
5. F. Kimmerle, J. Chevalet, J. Electroanal.
2.2.6
Chem. 1969, 21, 237–255.
Concluding Remarks 6. F. C. Anson, Anal. Chem. 1966, 38, 54–57.
7. M. Serelathan, R. A. Osteryoung, J. G. Oster-
The purpose of this chapter was to show young, J. Electroanal. Chem. 1986, 214,
how the potential and current step and 141–156.
the various pulse techniques can be used 8. M. Serelathan, R. A. Osteryoung, J. G. Oster-
young, J. Electroanal. Chem. 1987, 222,
to gain kinetic and mechanistic infor- 69–100.
mation concerning simple and complex 9. M. Penczek, Z. Stojek, J. Buffle, J. Elec-
electrode processes as well as to provide troanal. Chem. 1989, 270, 1–6.
10. H. A. Heering, M. S. Mondal, F. A. Arm- 20. J. J. O’Dea, A. Ribes, J. Osteryoung, J. Elec-

strong, Anal. Chem. 1999, 71, 174–182. troanal. Chem. 1993, 345, 287–301.
11. J. Osteryoung, J. J. O’Dea, in Electroanalyt- 21. J. H. Reeves, S. Song, E. F. Bowden, Anal.
ical Chemistry: A Series of Advances (Ed.: Chem. 1993, 65, 683–688.
A. J. Bard), Marcel Dekker, New York, 1986, 22. S. Komorsky-Lovrić, M. Lovrić, J. Electroanal.
pp. 209–308, Vol. 14. Chem. 1995, 384, 115–122.
12. J. G. Osteryoung, M. M. Schreiner, Crit. Rev. 23. S. P. Kounaves, J. J. O’Dea, P. Chandreskhar
Anal. Chem. 1988, 19, S1–S27. et al., Anal. Chem. 1986, 58, 3199–3202.
13. J. Osteryoung, Acc. Chem. Res. 1993, 26, 24. B. A. Brookes, R. G. Compton, J. Phys. Chem.
77–83. B 1999, 102, 5289–5295, 9020–9028.
14. H. P. van Leeuwen, J. Buffle, M. Lovrić, Pure 25. A. B. Miles, R. G. Compton, J. Phys. Chem. B
Appl. Chem. 1992, 64, 1015–1028. 2000, 104, 5331–5342.
15. C. M. A. Brett, A. M. Oliveira Brett, S. H. 26. A. B. Miles, R. G. Compton, J. Electroanal.
P. Serrano, J. Electroanal. Chem. 1994, 366, Chem. 2000, 487, 75–89.
225–231. 27. C. M. A. Brett, A. M. Oliveira Brett, L. C.
16. L. Ramaley, M. S. Krause Jr., Anal. Chem. Mitoseriu, Anal. Chem. 1994, 66, 3145–3150.
1969, 41, 1362–1365. 28. C. M. A. Brett, A. M. Oliveira Brett, F.-M.
17. J. J. O’Dea, J. Osteryoung, R. A. Osteryoung, Matysik et al., Talanta 1996, 43, 2015–2022.
Anal. Chem. 1981, 53, 695–701. 29. F.-M. Matysik, S. Matysik, A. M. Oliveira
18. J. J. O’Dea, J. Osteryoung, T. Lane, J. Phys. Brett et al., Anal. Chem. 1997, 69, 1651–1656.
Chem. 1986, 90, 2761–2764. 30. H. J. S. Sand, Philos. Mag. 1901, 1, 45.
19. M. Lovrić, S. Komorsky Lovrić, R. W. Murray, 31. D. Jagner, Trends Anal. Chem. 1983, 2, 53–56.
Electrochim. Acta 1988, 33, 739–744.
2.3 Stripping Analysis 125
I
2.3 that can be deposited at an electrode,
Stripping Analysis and particularly for those metals that
dissolve in mercury. Here, the metals are
Joseph Wang accumulated by potentiostatic deposition
New Mexico State University, Las Cruces,
onto a small-volume mercury electrode
New Mexico, U.S.A.
(a thin mercury film or a hanging
mercury drop), or onto a nonmercury
surface (Fig. 1). This is accomplished
2.3.1
by cathodic deposition at a controlled
Introduction
time and potential. The amalgam-forming
deposition step is given by Eq. (1):
Stripping analysis is a powerful electro-
analytical technique for trace metal mea- Mn+ + ne- + Hg - M(Q) (1)
surements [l]. Its remarkable sensitivity is
attributed to the “built-in” preconcentra- The deposition potential should be ca. 0.3
tion step, during which the target metals to 0.4 V more negative than the reduction
are accumulated onto the working elec- potential of the metal analyte. Since the
trode. The combination of such an effec- sensitivity of the stripping operation is de-
tive preconcentration step with advanced pendent on the deposition time, the latter
electrochemical measurements of the ac- should be selected according to the concen-
cumulated analytes generates favorable tration of the target metals (from around
signal-to-background characteristics, and 0.5 min at the 10h7 M level to 10 min
hence results in extremely low detection for lo-’ - lo-lo M concentrations). The
limits (down to the picomolar level). Other deposition step is usually facilitated by
advantages of stripping analysis include convective transport of the analyte to the
its multielement and speciation capabil- surface of the working electrode. This can
ities, low cost, and suitability to on-site, be accomplished by stirring or flowing the
in situ, and on-line applications. Different solution or by rotating the electrode. Qui-
versions of stripping analysis, which differ escent solutions can be used in connection
mainly in the nature of the preconcentra- to ultramicorelectrodes (described below).
tion step (electrolytic vs. adsorptive) and Only a small, and yet reproducible, frac-
in the method of detection (voltammetric tion of the metal in the solution is being
vs. potentiometric), are available [2]. Fun- deposited.
damental and practical aspects of these Following the preselected deposition
different stripping protocols are discussed period, the forced convection is stopped,
in the following sections. and a positive-going potential scan is
initiated, during which the amalgamated
metals are reoxidized and “stripped” away
2.3.2
from the surface:
Anodic Stripping Voltammetry
M(Hg) w Mnf + ne- + Hg (2)
2.3.2.1 Principles
Anodic stripping voltammetry (ASV) is The excitation signals used during the
the oldest, and still the most widely “stripping” step are usually pulse-voltam-
used version of stripping analysis [3]. The metric waveforms (that discriminate
technique is applicable to metal ions against the charging-current background
126\ 2 Electroanalytical Methods
Accumulation step Fig. 1 The deposition and

8 stripping steps of anodic
stripping voltammetry.
P ----------- I Stripping step

Cd
I
I
I
I
I
E~u -----------~---
Time
contribution) or a linear potential ramp. and stripping steps. The exact behavior
The stripping voltammogram, recorded governing the shape of the stripping
during this measurement step, consists voltammogram depends on the type of
of multiple current peaks, corresponding electrode and stripping waveform em-
to the reoxidation of the amalgamated ployed. Various theoretical treatments are
metals, and their "stripping" out of thus available [4, 5]. In all cases, the peak
the electrode (Fig. 2). Such an output current is proportional to the deposition
provides the qualitative and quantitative period and metal concentration:
information through measurements of the
(3)
peak potential and current, respectively.
The ASV operation thus allows the with K being a constant (including the
simultaneous measurements of four to six electrode area (A), scan rate (v), number
trace metals. of electrons (n), diffusion coefficient (D),
The peak current depends upon vari- and convection rate). For example, the
ous parameters of the preconcentration stripping peak current for a mercury film
Cd
Cu
-1.0 -0.5
Potential
Fig.2 Typical stripping voltammogram for a mixture containing 10-8 M
zinc, cadmium, lead, and copper.
electrode (MFE) is given by analyst should also be aware that he is
working with trace levels, and should thus
ip = 1.1157 x 106 n 2 ACRlv (4) employ a clean work practice for mini-
mizing contamination risks or adsorption
where I is the film thickness, and CR is the
losses. The reagents and solutions used for
concentration of the metal in the electrode
preparing supporting electrolytes or stan-
(which is proportional to C and tdep).
dards should thus be of the highest purity
A "mirror image" of ASV is cathodic
possible.
stripping voltammetry (CSV), which in-
volves an anodic deposition of the analyte,
2.3.3
followed by stripping in a negative-going
Stripping Potentiometry
potential scan.
Stripping potentiometry (SP) is another
2.3.2.2 Interferences attractive version of stripping analy-
As with other voltammetric techniques, sis [8, 9]. This was introduced in Chap-
stripping voltammetry may suffer from inter 2.2 (Sect. 2.2.5). Like other stripping
sufficient resolution of neighboring peaks. protocols, stripping potentiometry con-
Within a potential window of about 1.5 V, sists of two steps. The first (preconcen-
more than 15 metals may yield stripping tration) step, in which the metals are elec-
peaks. Thus, it is often difficult to resolve trodeposited onto the working electrode
metals that have peak potentials in close (usually a mercury film), is followed by a
proximity to one another. The problem stripping step in which the amalgamated
of overlapping peaks can be addressed metals are brought back to the solution
by adding a masking agent or using a with the aid of an oxidant (e.g. oxygen,
"medium-exchange" protocol. Stripping mercury ion) or a constant anodic current.
voltammetry can also be complicated by The resulting potentiogram (E vs. t plot)
the formation of intermetallic compounds consists of stripping plateaus, as in a redox
by two metals (e.g. Cu-Zn) that codeposit titration curve (Fig. 3a), with sharp poten-
into the mercury electrode [6]. Such in- tial changes accompanying the depletion
termetallic interaction usually causes one of each metal from the surface. Such a po-
of the stripping peaks to be depressed or tentiogram provides both the quantitative
shifted. This can be circumvented by plat- and qualitative information. The transition
ing only one of the metals (at a lower time needed for the oxidation of a given
potential) or via a preferential formation metal, fM, is a quantitative measure of the
of another intermetallic compound (e.g., sample concentration of the metal:
eu-Ga). Another problem is the presence
of surfactants that adsorb on the work-
ing electrode and block the deposition [7]. (5)
Such adsorption may affect both the de-
position and stripping steps, leading to where Cox is the concentration of the ox-
lower or broader peaks, and shifts in the idant. The qualitative identification relies
peak potential. The surfactant problem can on potential measurements (in accordance
be minimized through coverage with a with the Nemst equation for the amalga-
permselective/protective film (e.g. Nafion, mated metal). Modem SP instruments use
Agar), or using ultraviolet irradiation. The microcomputers to register fast stripping
E dt! dE
-1.2 3
2 Cd
-1.0
-0.8
o
-0.6 -1.4 -1.0 -1.6
E
(a) 25 ms (b)
Fig. 3 Stripping potentiometric curves for 90 ~g L-1 tin, cadmium,

and lead following 90 s deposition at -1.4 V.
events and to convert the wave-shaped re- Adsorptive stripping protocols commonly
sponse to a more convenient peak over rely on the formation of surface-active
a flat baseline (i.e. to dt jdE versus E complexes of the target metal (in the
signals; Fig. 3b). Here, the peak area is pro- presence of a suitable complexing agent)
portional to the bulk concentration, with followed by the interfacial accumulation
the peak potential providing the qualitative of the complex onto a mercury drop or
identification. Because of the short (mil- film electrode (Fig. 4). Most procedures
lisecond) transition times, a fast rate of involve the reduction of the metal in the
data acquisition (kilohertz) is required for adsorbed complex during a negatively go-
obtaining a sufficient number of "counts" ing potential scan. In few cases, it is
for a defined peak. possible to exploit the reduction of the
Stripping potentiometry offers several ligand in the complex. This is usually
important advantages (over analogous accomplished by scanning the potential
ASV schemes), including reduced suscep- in the negative direction (using linear or
tibility to organic surfactants and reliable pulse-scan modes) following the adsorp-
operation in nondeaerated samples. tive accumulation step. The quantitation
of extremely low levels of the target metals
2.3.4 (10- 10 M) can be realized in connec-
Adsorptive Stripping Voltammetry tion to short (3-5 min) adsorption times.
Even lower concentrations, down to the
The power and scope of stripping analysis 10- 12 M level, can be achieved by combin-
have been greatly enhanced during the past ing the adsorptive stripping operation with
two decades owing to the introduction of catalytic effects [13, 14]. Such adsorptive-
adsorptive stripping procedures [10-12]. catalytic stripping voltammetric protocols
Such use of adsorptive accumulation per- commonly involve a catalytic cycle in the
mits convenient measurements of over presence of a chemical oxidant (such as
two dozen trace metals that cannot be read- bromate, nitrite, or hydrogen peroxide).
ily deposited (e.g. Cr, AI, U, Fe, Ti, V, Mo). Different kinds of catalytic systems can be
8
Accumulation Stripping
MI1+ + nL ~ MLn11+
Eacc~----------------------~
Time
-
Fig.4 Steps in adsorptive stripping voltammetry: formation of a metal complex, its
adsorptive accumulation, and the reduction of the adsorbed complex.
considered depending upon the reaction used for detecting trace levels of organic
mechanism [14]. The remarkably low de- compounds (e.g. pharmaceuticals) and
tection limits of such operations are thus biologically important compounds (includ-
attributed to the dual current amplification ing hormones and nucleic acids). Non-
effect (associated with the adsorption and electroactive macromolecules may also
catalytic processes). be determined following their interfacial
The adsorptive stripping response (of accumulation from tensiometric peaks (re-
the surface-confined species) is directly resulting from their adsorption-desorption
lated to its surface concentration, with the mechanism).
adsorption isotherm - commonly that of
Langmuir - providing the relationship be- 2.3.5
tween the surface and bulk concentrations Practical Considerations
of the adsorbate. As a result, calibration
curves display an initial linear portion, fol- 2.3.5.1 Electrodes
lowed by a curvature, and a leveling off at Proper choice of the working electrode is
high concentrations. The response is de- crucial for the success of the stripping
pendent upon numerous other variables operation. The ideal working electrode
that affect the extent of adsorption, in- should offer effective preconcentration,
cluding the electrolyte, pH, accumulation a favorable redox reaction of the target
potential and time, convection rate, and metal, reproducible and renewable sur-
ligand concentration. face, and a low background current over
Besides expanding the scope of strip- a wide potential range. Mercury is the
ping analysis to numerous trace elements electrode of choice for many stripping
that cannot be electroplated (e.g. Table 1), applications. Two types of mercury elec-
adsorptive stripping procedures have been trodes are commonly used, the hanging
130 I 2 Electroanalytical Methods
Tab. 1 Common adsorptive stripping procedures for trace metals
Metal Complexing Supporting Detection References

agent electrolyte limit (M)
AI Dihydroxyanthraqui none- BES buffer 1 x 10- 9 15

sulfonic acid
Co Nioxime H EPES buffer 6xlO- 12 16
Cr Diethylenetriamine- Acetate buffer 4xlO- 1O 17
pentaacetic acid
Fe Solochrome violet RS Acetate buffer 7 x 10- 10 18
Mo Oxine Hydrochloric 1 x 10- 10 19
acid
Ni Dimethylglyoxime Ammonia buffer 1 x 10- 10 20
Pt Formazone Sulfuric acid 1 x 10- 12 21
Sn Tropolone Acetate buffer 2xlO- 1O 22
Ti Mandelic acid Potassium 7 x 10- 12 23
chlorate
U Oxine PI PES buffer 2xlO- 1O 24
V Catechol PI PES buffer 1 x 10- 10 25
mercury drop electrode (HMDE) and mer- Ultramicroelectrodes have been shown
cury film electrode (MFE)_ The HMDE to be useful for eliminating the need
relies on reproducible drop formation at for forced convection during the depo-
the tip of a glass capillary. The MFE sition step and for obviating the need
may be preplated (prior to the measure- for supporting electrolyte [28). The lat-
ment) or codeposited with the target metal ter is particularly attractive for speciation
analytes_ Glassy carbon, carbon fiber, irid- work in low ionic-strength natural waters
ium microdisks, and screen-printed car- and for minimizing the risk of contam-
bon are the most common supports for ination. Arrays of microelectrodes have
the mercury film_ MFEs possess higher been shown to be useful for measur-
surface-to-volume ratio (compare to the ing metal concentration profiles in high
HMDE), and consequently offer a more
spatial resolution [29). Further informa-
efficient preconcentration [26). Alternative
tion is provided in Chapter 2.5. Microfab-
"nonmercury" electrodes, including gold,
ricated (mass-produced) electrodes, par-
carbon, silver, or bismuth ones, have been
developed for addressing concerns regard- ticularly screen-printed electrode strips,
ing the toxicity and handling of mercury. have been developed as low-cost dis-
Particularly attractive are the bismuth-film posable ("one-shot") metal sensors [30).
electrodes that offer an attractive stripping The deliberate modification of conven-
performance, similar to that of mercury tional electrodes can benefit various strip-
electrodes, while eliminating toxicity prob- ping applications, by imparting higher
lems [27)- Gold surfaces are characterized selectivity, sensitivity, and stability. Var-
by a wider potential window (compare to ious surface coatings have thus been
mercury ones) to allow convenient detec- used for protection against surface-active
tion of important metals, such as arsenic, foulants or for preconcentration of target
selenium, or mercury_ metals.
2.3.5.2 Cells and Instrumentation submersible stripping-based sensors [34]
Electrochemical stripping analysis is car- and submersible ("lab-on-cable") stripping
ried out in a conventional ("beaker-type") probes [35, 36]. Such devices offer a fast re-
three-electrode cell (of 5-50-ml volume). turn of the chemical information (i.e. a di-
The three electrodes, as well as the tube rect and reliable assessment of the fate and
used for bubbling the deoxygenating gas, gradient of metal contaminants) in a safe,
are supported in five holes in the cell timely, and cost-effective manner. These,
cover. Various microcells with 20 to 500 III along with the introduction of single-use
volumes can be used when the sample (disposable) microfabricated metal-sensor
volume is limited. Particularly attractive strips [30] or micromachined handheld to-
are thin-layer cells in which the entire tal stripping analyzers [35] would allow
sample is confined within a thin layer (of moving the measurements of trace met-
less than 10-llm thickness) at the elec- als to the field, and performing them more
trode surface [6]. Smaller sample volumes rapidly, reliably, and inexpensively.
can be accommodated in connection with Portable (handheld), battery-powered,
ultramicroelectrodes and advanced micro- easy-to-use stripping analyzers have also
fabrication processes. The latter result in been developed, particularly in connec-
planar strips that can be considered as dis- tion to decentralized screening of blood
posable electrochemical cells onto which lead in children [37]. The operation of
sample droplets are placed. these compact stripping meters has been
The adaptation of electrochemical strip- combined with disposable screen-printed
ping techniques to on-line flow systems electrodes, similar to those used by dia-
provides several advantages, including betes patients for monitoring their blood
high sample throughput, continuous mon- glucose [30].
itoring capability, improved precision,
lower risk of contamination, and conve- 2.3.6
nient medium exchange. A wide range Representative Examples
of flow-through stripping detectors has
been developed for on-line monitoring Numerous applications of stripping anal-
of metals [1]. Fully automated flow sys- ysis to many environmental, clinical, or
tems, based on such detectors, have been industrial problems have been reported.
described [31]. High-throughput stripping The technique has been widely used
assays of small samples volumes have for the measurement of trace metals in
been realized in connection to flow- natural waters, including pristine and pol-
injection [32] and batch-injection [33] op- luted inland and marine waters [1, 36,
erations. Further information on several 38]. The technique is particularly suit-
types of hydrodynamic electrodes is given able for the study of chemical specia-
in Chapter 2.4. tion, that is, the determination of dif-
The portable instrumentation and low ferent physical-chemical forms of the
power demands of stripping analysis sat- element [39]. Measurements of oxidation
isfy many of the requirements for on-site states, complexation capacity, stability con-
and in situ monitoring of trace metals. stants, or "labile metal" concentration,
Efforts in this direction have led to the represent typical stripping speciation ex-
replacement of conventional stripping pro- periments. Because of its fundamen-
tocols and systems with remotely deployed tally different detection principles, the
Tab.2 Representative applications of stripping analysis
Metal Sample Stripping Working References

matrix mode electrode
Antimony Gunshot residue ASV MFE 40

Cadmium Rain ASV MFE 41
Chromium Soil AdSV HMDE 42
Cobalt Seawater AdSV HMDE 16
Copper Steel ASV HMDE 43
Iodide Seawater CSV HMDE 44
Iron Wine AdSV HMDE 45
Lead Blood SP MFE 46
Manganese Wastewater SP MFE 47
Mercury Fish ASV Au 48
Nickel Aerosols AdSV HMDE 49
Platinum Gasoline AdSV HMDE 50
Selenium Soil CSV HMDE 51
Thallium Urine ASV HMDE 52
Titanium Seawater AdSV HMDE 23
Uranium Groundwater AdSV HMDE 53
Zinc Eye tissue ASV HMDE 54
Note: AdSV: Adsorptive stripping voltammetry.
adsorptive stripping approach provides a and other applications, are illustrated in

different speciation information (based Table 2.
on ligand competition) compared to
conventional ASV procedures. Recent ad- References
vances in miniaturization and remote
sensing have facilitated the application of 1. J. Wang, Stripping Analysis: Principles, Instru-
stripping analysis for on-site monitoring mentation and Applications, VCH Publishers,
of toxic metals, including in situ (real- DeerfieId Beach, FIa., 1985, p. 160.
2. A. Fogg, J. Wang, Pure Appl. Chem. 1999, 71,
time) profiling [35] Other environmental
891.
applications of stripping analysis include 3. T. R. Copeland, R. K. Skogerboe, Anal. Chem.
assays of sediments, fly ash, rain, or air- 1974,46,1257A.
borne particulate matter (e.g. Table 2). 4. D. Roe, J. Toni, Anal. Chem. 1965, 37, 1503.
Stripping analysis has been widely used 5. W. T. de Vries, J. Electroanal. Chem. 1965, 9,
for measuring trace metals in biologi- 448.
6. T. DeAngelis, R. Bond, E. Brook et aI., Anal.
cal fluids. The use of stripping anal- Chem. 1977, 49, 1792.
ysis for decentralized testing of blood 7. J. Wang, D. Luo, Talanta 1984,31,703.
lead in children has been particularly 8. D. Jagner, Trends Anal. Chem. 1983, 2(3), 53.
successful [46]. Measurements of mer- 9. D. Jagner, Analyst 1982, 107, 593.
cury in fish [48], of antimony in gunshot 10. C. M. G. van den Berg, Anal. Chim. Acta
1991, 250, 265.
residues [40], or of cadmium impurities in
11. J. Wang, Voltammetry after nonelectrolytic
zinc plant electrolyte [55], represent typical proconcentration in Electroanalytical Chem-
food, forensic, and industrial applications, istry (Ed.: A. J. Bard), Marcel Dekker, New
respectively, of stripping analysis. These, York, 1989, pp. 1-88, Vol. 16.
12. M. Paneli, A. Voulgaropoulos, Electroanalysis 34. J. Wang, N. Foster, S. Armalis, D. Larson,
1993, 5, 355. A. Zirino, K. O\sen, Anal. Chim. Acta 1995,
13. M. Czae, J. Wang, Talanta 1999,50,921. 310,223.
14. A. Bobrowski, J. Zarebski, Electroanalysis 35. J. Wang, B. Tian, J.
Luet aI., Anal. Chim.Acta
2000, 12,1777. 1999, 385,429.
15. C. M. G. van den Berg, K. Murphy, J. Riley, 36. M. Tercier, J. Bume, Electroanalysis 1993, 5,
Anal. Chim. Acta 1986, 188,177. 187.
16. J. Donat, K. Bruland, Anal. Chem. 1988, 60, 37. C. Yarnitzky, J. Wang, B. Tian, Talanta 2000,
240. 51,333.
17. J. Golimowski, P. Valenta, H. W. Nurnberg, 38. E. P. Acheterberg, C. Braungradt, Anal.
Fresenius Z. Anal. Chem. 1985, 322, 315. Chem. Acta 1999,400,381.
18. J. Wang, J. Mahmoud, Fresenius Z. Anal. 39. T. M. Florence, Talanta 1982,29,345.
Chem. 1987, 327, 789. 40. N. Komanur, G. van Loon, Talanta 1977, 24,
19. C. M. C. van den Berg, Anal. Chem. 1985, 57, 184.
1532. 41. H. Emons, A. Baede, M. Schoning, Electro-
20. B. Pihlar, P. Valenta, H. W. Nurnberg, Frese- analysis 2000, 12,1171.
nius Z. Anal. Chem. 1981, 307, 337. 42. J. Wang, J. Lu, J. K. Olsen, Analyst 1992,117,
21. J. Wang, J. Zadeii, M. S. Lin, J. Electroanal. 1913.
Chem. 1987, 237, 281. 43. S. Gottesfeld, M. Ariel, J. Eleetroanal. Chem.
22. J. Wang, J. Zadeii, Talanta 1987, 34, 909. 1965, 9, 112.
23. K. Yokoi, C. M. C. van den Berg, Anal. Chim. 44. G. Luther, C. Swartz, W. Ullman, Anal.
Acta 1991, 245, 167. Chem. 1988, 60, 1721.
24. C. M. G. van den Berg, N. Nimmo, Anal. 45. J. Wang, S. Mannino, Analyst 1989,114,643.
Chem. 1987, 59, 269. 46. P. Ostapczuk, Clin. Chem. 1992, 38, 195.
25. C. M. G. van den Berg, Anal. Chem. 1984, 56, 47. G. Scollary, G. Chen, T. Cardwell et aI., Elee-
2383. troanalysis 1995, 7, 386.
26. T. M. Florence, J. Electroanal. Chem. 1970, 48. J. Golimoski, I. Gustavsson, Fresenius Z.
27,273. Anal. Chem. 1984, 317, 484.
27. J. Wang, J. Lu, S. Hocevar et aI., Anal. Chem. 49. M. Nimmo, G. Fones, Anal. Chim. Acta 1994,
2000, 72, 3218. 291, 321.
28. J. Wang, P. Tuzhi. Anal. Chim. Acta 1987, 50. K. Hoppstock, M. Michulitz, Anal. Chim.
197,367. Acta 1997, 350, 135.
29. M. L. Tercier-Waeber, J. Pei, J. Buffle et al., 51. S. Porbes, G. Bound, T. West, Talanta 1979,
Electroanalysis 2000, 12,27. 26,473.
30. J. Wang, B. Tian, Anal. Chem.1992, 64, 1706. 52. D. L. Levit, Anal. Chem. 1973, 45, 1291.
31. A. Zirino, S. Lieberman, C. Clavell, Environ. 53. J. Wang, R. Setiadji, L. Chen et aI., Electro-
Sci. Technol. 1978, 12, 73. analysis 1992, 4, 161.
32. J. Wang, H. Dewald, Anal. Chem. 1994, 56, 54. T. Williams, O. Foy, C. Benson et al., Anal.
156. Chim. Aeta 1975, 75, 250.
33. C. M. Brett, A. M. Brett, L. Tugelea, Electro- 55. A. M. Bond, Anal. Chim. Acta 1999, 400, 333.
analysis 1996, 8, 639.
2.4 relatively slow and rate-limiting and the

Hydrodynamic Electrodes overall flux is given by
Andrew R. Mount jmt = kmt c∞ (3)

The University of Edinburgh, Edinburgh,
United Kingdom At the electrode, electrochemical reaction
then occurs, governed by a characteristic
2.4.1 first-order rate constant ke . Under steady
Introduction state conditions, the electrochemical reac-
tion flux is also equal to the overall flux j ,
A hydrodynamic electrode is an elec- that is, je = j = jmt . The electrochemical
trode system in which the solution near flux is given by
the electrode undergoes forced mechani-
cal convection (stirring). This chapter is je = ke c0 (4)
concerned with controlled hydrodynamic
electrodes in which this mechanical con- When the reaction is under electrochem-
vection is reproducible and leads to steady ical control, the electrochemical reaction
state currents. There are several advan- is relatively slow and rate-limiting and
tages of using electrodes with controlled c∞ = c0 . Then the overall flux is given by
hydrodynamics. These are most easily seen je = ke c∞ (5)
by considering a general scheme for elec-
trochemical reaction In general, the overall flux j for the
kmt ke , ±ne−
reaction is determined by the following
Rbulk −−−→ Rsurf −−−−−→ P n−/n+ (1) equation:

1 1 1 1
where a reactant R is electrochemically = + (6)
j kmt ke c∞
converted to a product P at an electrode
surface by accepting or donating n elec- with the smallest of the rate constants
trons (i.e. reduction or oxidation). The determining the overall reaction rate.
overall reaction requires the mass trans- The overall rate of reaction is measured
port of R from the bulk of solution to the experimentally as the current i where
electrode surface, the rate of which can be
considered to be governed by a first-order i = ±nFAj (7)
rate constant kmt in cm s−1 . The rate of the
where A is the electrode area (cm2 ), F
mass transport reaction is then given by
is Faraday’s constant (C mol−1 ), and the
the flux of species (mol cm−2 s−1 ) to the
positive and negative signs correspond
electrode surface
to oxidation and reduction, respectively.
jmt = kmt (c∞ − c0 ) (2) From Eqs. (6 and 7), it can therefore be
seen that reproducible currents can only
where c is the concentration of R in the be obtained in systems in which both kmt
bulk (at a near infinite distance ∞ from and ke can be controlled.
the electrode) or at the electrode surface (at ke is easily controlled by controlling the
zero distance, 0, from the electrode). The electrode potential. However, in stagnant
maximum flux for this process is when solutions, or solutions with irreproducible
c0 = 0, when the rate of mass transfer is or unstable convection, kmt is not constant
2.4 Hydrodynamic Electrodes 135
and is time-dependent, as the extent of detection. Measurement of the decrease in

depletion of reactant near the electrode the amount of product detected as the flux
governs the rate of mass transfer, and (and hence the transit rate) is decreased
this changes with time. This often leads experimentally therefore enables the reac-
to irregular or unpredictable currents. The tivity of any unstable products generated
first advantage of controlled hydrodynamic upstream to be determined.
electrodes is that they generate repro- This chapter considers each of the major
ducible and predictable solution convec- hydrodynamic electrode systems in turn.
tion through mechanical stirring, which Only those systems at which steady state
leads to control of kmt and hence i, pro- measurements can be achieved are con-
ducing steady state (or time-independent) sidered, therefore excluding those systems
currents. Variation of kmt is easily achieved with hydrodynamic regimes that inher-
by varying the convection rate, allowing ently produce time-dependent oscillations
the quantitative determination of the role superimposed on steady state behavior,
of mass transport in the overall kinetics such as vibrating electrodes [1] and son-
of electrode reaction. The second advan- icated electrode systems [2] in which ul-
tage is that the reproducible flow results trasound is used to produce fluid motion.
in a reproducible and controllable flux of
The latter are discussed further in Chap-
product away from the electrode surface.
ters 2.8, 2.9 and 2.10 of this volume. For
This allows the downstream detection and
each system, the hydrodynamic regime is
identification of these products, for exam-
described, along with representative appli-
ple, by means of a second electrode [as
cations that demonstrate the applicability
in the rotating ring disc electrode (RRDE,
of the system to particular measurements
see Sect. 2.4.2.2], which can be set up (by
of electrochemical kinetics and mecha-
control of electrode type and potential)
to detect and identify particular products, nism. The system characteristics are also
or by a suitable spectroscopic technique discussed, to highlight the suitability of the
[such as electron spin resonance (esr)]. system to particular applications, giving in-
Applications of esr in electrochemistry sight into the strengths and weaknesses of
are discussed further in Volume 2 of the each system.
Encyclopedia. The third advantage is that
the controlled hydrodynamics allow the 2.4.2
fraction of products that can be detected Rotating Disc Hydrodynamics
(or collected) downstream to be theo-
retically determined (e.g. the collection 2.4.2.1 Rotating Disc Electrode(RDE)
efficiency of the RRDE, see Sect. 2.4.2.2). The RDE (Fig. 1) is a disc electrode that
Comparison of the amount of product is surrounded by a coplanar insulating
detected experimentally and expected the- mantle (typically made of poly(tetrafluoro-
oretically therefore allows quantitation of ethene), PTFE). This is placed into a
the amount of product produced by the solution and rotated at a controlled rotation
reaction process (giving the efficiency of speed, W (Hz), which induces stirring in
the reaction). Finally, control of product the solution [3].
flux through stirring rate can be used The RDE acts as a pump (Fig. 2),
to control the transit time from product drawing the solution up toward the
generation to downstream collection or electrode along the z-axis (Fig. 2a), then
Fig. 1 The rotating disc

electrode.
Rotation
Speed
W
PTFE sleeve
Pt disc
z
Electrode surface Fig. 2 Solution flow for RDE

(a) along z-axis and (b) near
disc surface.
Bulk solution
(a) (b)
zD Fig. 3 Concentration profile for the

RDE normal to the electrode. Reaction
[O]∞ of O at the RDE produces a
concentration [O]0 at the electrode
surface. Mass transport to the electrode
occurs by diffusion across zD . At z > zD ,
the solution is well-stirred and the
[O]
Bulk solution concentration of O is the bulk

concentration[O]∞ .
[O]0
flinging it out radially across the electrode Viewed along the z-axis, there is a diffusion
surface (Fig. 2b). layer, zD , across which the reactant diffuses
This electrode is uniformly accessi- to the electrode (Fig. 3).
ble [4], in that during a reaction the flux It must be remembered that when
(and hence the current density) is the z < zD , although transport along the z-axis
same across the entire disc surface. Also, is by diffusion only, the diffusion layer is
the system possesses circular symmetry not totally stagnant, as there is significant
about the z-axis. This greatly simplifies the radial convection (Fig. 2b) that sweeps
mathematical description of the hydrody- products away from the electrode. This is
namics, and allows an analytical solution utilized in product detection in the RRDE
of the convective-diffusion equation [5]. system (see Sect. 2.4.2.2).
The thickness of the diffusion layer is couple. This is because the uniform
rotation speed–dependent, and is given by accessibility of the RDE ensures that the
flux of substrate to the electrode is constant
zD = 0.643D 1/3 ν 1/6 W −1/2 (8) across the surface, and the controlled
hydrodynamics ensures a constant and
where D is the diffusion coefficient of
reproducible substrate concentration at the
the reacting species (cm2 s−1 ) and ν is
surface of the electrocatalytic film. Both of
the kinematic viscosity of the solvent
these effects simplify kinetic modeling.
(cm2 s−1 ). Since in the z-direction trans-
In these experiments a general form of
port is only by diffusion,
the Koutecky–Levich equation generally
∂c (c∞ − c0 ) applies,
j = −D =D (9) 1 1 1
∂z zD = + (13)
i iL i∞
with
D with the second term in Eq. (12) general-
kmt = (10)
zD ized to be a mass transfer–independent
The mass transport limiting current, iL , on current i∞ (independent of W ). A value
the electrode (which is the current when ke of i∞ can generally be obtained from the
is extremely large, the electrode is highly y-intercept of a straight-line plot of 1/i
reactive, and [O]0 = 0 i.e. c0 = 0) is given against W −1/2 (termed a Koutecky–Levich
by a combination of Eqs. (3, 7, 8, and 10). plot). The infinity subscript is used to indi-
This is termed the Levich equation [5], cate that this is the current that would be
obtained at infinite rotation speed, when
iL = ±1.554nFAD 2/3 ν −1/6 W 1/2 c∞ the first term is unimportant and c0 = c∞ .
(11) In each case, the theoretical parameters
and more generally, the current is given by that make up i∞ reflect the nature of
a combination of Eqs. (6, 7, 8, and 10) as the rate-determining step in the reaction
the Koutecky–Levich equation [6], at the surface, between the substrate and
1 1 1 the immobilized redox species. For a sim-
= ± (12) ple catalytic reaction between a substrate
i iL nFAke c∞
and an electrocatalytic monolayer or thin
From Eq. (11), the first term in Eq. (12) film, the rate-determining step is often the
can be varied by changing W , while the kinetics of the electron transfer reaction be-
second term can be varied by changing the tween the substrate and the redox species
electrode potential and hence ke . Hence and the rate constant for this reaction can
this relationship allows the relative effects be obtained from i∞ . However, for thicker
of the rate of mass transport (the first term) surface films (e.g. polymer modified elec-
and the rate of electrochemical reaction trodes), the rate of transport of electrons or
(the second term) on the overall reaction substrate across the film can also become
rate to be separated and quantified. rate-determining, and terms for these pro-
The RDE is a particularly good system cesses are then included in i∞ . Albery
for the measurement of the electrode and Hillman [7, 8] and Andrieux and
kinetics of electrocatalytic films (modified Saveant [9] have each produced general
electrodes), in which mediated electron theoretical treatments that enable the rate-
transfer to substrate species in solution determining step for these electrocatalytic
occurs via a surface-immobilized redox reactions to be identified and quantified.
This approach has also been extended to RDE (the ring electrode) to detect products
other modified electrode reactions and this from the RDE (Fig. 4). In this case,
has been the subject of a comprehensive re- products are ejected outwards radially
view [10]; a good example is the reaction be- from the disc electrode, and so the
tween an immobilized enzyme and its sub- ring electrode is positioned outside of
strate, in which complexation between the and coplanar with the disc electrode,
immobilized enzyme (the redox species) with a relatively narrow insulating gap
and the substrate occurs [11, 12]. We have separating them.
also shown [13] that the electropolymeriza- The crucial parameter for the rotating
tion and trimer film formation of a range disc electrode is the collection efficiency
of 5-substituted indoles at the RDE N . This determines the fraction of stable
R R product generated with 100% efficiency
at the disc electrode that will be col-
lected by the ring, under mass transport
R control, when there is a mass trans-
−7e −, −6H+ N + N
3 port–limiting ring current. In general,
N H H
N under these conditions,
H H
R jR = −NjD (15)
(14)
produce straight-line Koutecky–Levich
where jR and jD are the ring and disc
plots, from which the rate constant for
coupling of the indole monomer radical fluxes, respectively. The collection effi-
cations produced on the electrode surface ciency can either be determined experi-
(which form the redox-active film of indole mentally or theoretically. It is independent
trimers) can be obtained. Furthermore, of rotation speed and is merely a function
use of the RDE ensures controlled flux of of electrode geometry. Under most condi-
indole monomer to the whole of the elec- tions it is generally denoted as N0 , given
trode surface, which ensures a much more by [14, 15]
reproducible and homogenous film struc-
ture across the disc than in experiments α
N0 = 1 − F + β 2/3 [1 − F (α)]
with stationary electrodes. β
− (1 + α + β)2/3
2.4.2.2 The Rotating Ring Disc Electrode
α
(RRDE) × 1−F (1 + α + β)
The RRDE utilizes a second independent β
electrode positioned downstream of the (16)
Fig. 4 The positioning of a ring

Ring
electrode downstream of the
electrode
rotating disc electrode.
with to selectively detect the bromide ion pro-

√ duced by the electrochemical reduction
3 (1 + θ 1/3 )3
F (θ) = ln of the inhalation anaesthetic, halothane
4π 1+θ (CF3 CHClBr), at an RDE [16]. The elec-
1/3
3 2θ − 1 1 trochemical reduction of halothane at an
+ tan−1 √ + RDE is one potential method for sensing
2π 3 4
this inhalation anaesthetic. However, this
(17)
reduction occurs at a very similar poten-
and tial to the reduction of oxygen gas, which
3 precludes deconvolution of the proportion
r2 of the reduction current due to oxygen, and
α= −1 (18)
r1 that due to halothane. At a silver ring elec-
3 3 trode, the selective mass transport–limited
r3 r2
β= − (19) oxidation of bromide
r1 r1
−e−
α and β are therefore merely functions of Ag + Br− (aq) −−−→ AgBr(s) (21)
the disc radius r1 , the inner ring radius r2 ,
and the outer ring radius r3 . N0 typically was shown to occur [16]. This produced
has a value around 0.2 for most RRDE solid silver bromide, which deposited
systems. on the ring surface at a characteristic
In our work on the polymerization of potential. The oxidative current for this
5-substituted indoles [13], we have used process and the reduction charge obtained
the ring electrode of the RRDE to detect from reducing the silver bromide formed
and quantify the amount of soluble trimer on the ring electrode after a given period
formed during electropolymerization (vide of time (by the reverse reaction of Eq. 21)
supra). Significant soluble fractions were were both shown to be a good measure
obtained with indole monomers that con- of the flux of bromide produced at the
tained solubilizing R groups, such as disc. Comparison of the ring bromide flux
amino groups or hydroxy groups. Fur- and the disc electron flux established the
thermore, judicious selection of the ring stoichiometry of the halothane reduction
potential enabled the protons produced by reaction to be [16]
the reaction in Eq. (14) to be detected on
+2e− , +H+
the ring as CF3 CHClBr −−−−−−−→ CF3 CH2 Cl + Br−
H+ + e− −−−→ 12 H2 (20) (22)
and the amount of bromide detected at
and the measured fluxes of protons on the the silver ring was shown to be a good
ring were compared to the flux of electrons measure of the proportion of halothane
on the disc to establish the stoichiometry in the halothane/oxygen mixtures [17].
of Eq. (14). This, in combination with the overall
The ring electrode can also be made from disc current, allowed deconvolution of the
a different material to the disc, designed relative amounts of halothane and oxygen
to selectively detect a particular species in the mixture.
produced by the disc reaction. An exam- The ring electrode of an RRDE has
ple is the use of a silver ring electrode also been used to selectively detect the
products from the redox reaction of a ND , must be used [18], where

rotating disc–modified electrode. In this
work, a thionine-coated electrode was 1 2 r 2 + r3 3

ND = 1 − F (α) − F
produced by the electropolymerization of 6 3 2r1
thionine. Thionine is a phenothiazine 1
dye, and the resulting film is redox- − F (α + β) (24)
6
active, with the thionine redox centers
being able to be redox-cycled between The exact form of the proton transients
their oxidized (thionine, Th) and reduced seen at the ring is determined by the
(leucothionine, L) forms, according to the degree of protonation of the L centers and
their ability to act as a buffer, mopping
up the protons produced in the redox
+
reaction. This is determined by the pKa s
N for deprotonation of the various protonated
forms of L. Quantitative analysis of the
H2N S NH2 proton transients observed at the bismuth
oxide ring electrode enabled these pKa
Th+
values to be determined as
2+
H pKa = 5.5
+ +
N LH3 2+ −
−
−−−−
−−−−
−− LH2 + H

pKa = 7.8
LH2 + − +
S
H3N NH3 −
−−−−
−−−−
−− LH + H

2+
LH3
pKa = 7.8
− +
LH −
−
−−−−
−−−−
−−L +H
(25)
general equation
and the rate-determining step for the
Th+ + 2e− + mH+ − −−−−
− LHm
(m−1)+
film redox reaction was shown to be the
(23) transport of electrons through the film.
The value of m is dependent upon When m = 2, then the redox reaction
the pH of the solution. If m = 0, then involves incorporation into the film or
redox reaction involves the incorporation expulsion from the film of solution coun-
and expulsion of protons during the teranions (and/or countercations) to main-
redox reaction and this can be measured tain overall film electroneutrality during
by a pH-sensitive bismuth oxide ring redox cycling. In unbuffered solution near
electrode [18, 19]. In the special case when pH 7, the reduction of Th+ rapidly causes
the ring electrode is a potentiometric the film pH to rise above 7.8, and the film
sensor with a low exchange current redox reaction is
density, such as the bismuth oxide sensor
Th+ X− − +
(s) + 2e + M(aq) −−−→
for pH [18, 19], the ring electrode reaction
does not significantly perturb the radial L− M+ −
(s) + X(aq) (26)
concentration profile of the disc product
across the ring, and a different collection This flux of X− can be measured by
efficiency, termed the detection efficiency, an anion-sensitive ring electrode such as
the Ag/AgX potentiometric ring electrode, for each of these steps (Fig. 5e), and
which shows a selective and quantifiable the responses added to give the overall
change in potential due to the local concen- ring flux transient (Fig. 5f ). A similar
tration of X− anion [19]. Measurements procedure can often be applied in reverse
have been made with X− as Cl− , using to generate the disc flux that gave rise to
a silver/silver chloride ring electrode in an observed ring flux [22]. In this case,
an electrolyte solution of 50 mM NaClO4 , the parameters of a number of general
1 mM NaCl [19]. The magnitude of the disc current functions (usually tan h and
ring transient, when compared to that ex- exponential functions) are adjusted so
pected from the disc current transient, that when added, their calculated time-
indicated that, as expected, reaction oc- dependent ring flux response matches
curred according to Eq. (26) and that the the experimentally observed ring flux.
film was permselective to Cl− over ClO4 − The sum of the disc flux functions then
by a factor of 20. Further general infor- corresponds to the disc flux transient.
mation on potentiometric measurements One final application of the RRDE is in
is given in Sect. 1.1.3 of Chapter 1 in this the measurement of the kinetics of the
volume. following chemical reactions of species
Such ring current measurements are electrogenerated at the disc with other
of necessity not steady state, as the solution species (often termed titration
reaction involves the reduction of a redox- reactions). In one of the simplest exam-
active film of finite thickness coated on ples of these reactions, reactant A is
the disc electrode. However, the ring generated electrochemically on the disc
response to a transient disc current has electrode, which then undergoes reaction
also been calculated. The time-dependent with reactant B, present in solution, dur-
collection efficiency Nτ has been derived ing transit between disc and ring. The
analytically for a galvanostatic (constant ring electrode is set to detect reagent A,
current and hence constant flux) step on and because of reaction, the observed ring
the disc electrode [20] as a function of the flux is typically less than that predicted
dimensionless time from the disc flux and Eq. (15). A good
1/3 example [23] is the use of the RRDE to
D study the kinetics of bromination of the
τ = 4.0 Wt (27)
ν fatty acids: oleic (or cis-9-octadecanoic) acid
[CH3 (CH2 )7 CH=CH(CH2 )7 COOH], eru-
where t is the time, which allows the cic (or cis-13-docosenoic) acid [CH3 CH2
corresponding time-dependent variation CH=CH(CH2 )11 COOH], and linolic (or
of the ring flux to be calculated from cis,cis-9,12-octadecanoic) acid [CH3 (CH2 )3
Eqs. (7 and 15) (Fig. 5a and b) for a (CH2 CH=CH)2 (CH2 )7 COOH] in a mixed
constant disc flux. This approach has been solvent system of acetic acid and ace-
generalized to produce a computational tonitrile. Titration with bromine is one
method for predicting the ring flux for potential analytical method for measuring
any form of disc flux transient [21], by the amount of double bonds in an oil con-
considering any disc current transient as taining these unsaturated hydrocarbons
consisting of the sum of many discrete in solution, but the kinetics of reaction
galvanostatic steps (Fig. 5d). The time- must be sufficiently fast for this approach
dependent ring flux can then be calculated to be viable. In this work, bromine was
|jD|
|jR|
(a) Time (b) Time
|jD|
|jD|
Time
Time
(c) (d)
|jR|
|jR|
Time
Time
(e) (f)
Fig. 5 The time-dependent ring flux (b) produced from a galvanostatic disc flux step (a)
and Nτ (d) the deconvolution of a disc flux transient (c) into a series of galvanostatic disc
flux steps (e) the calculation of the time-dependent ring fluxes from these steps and their
combination to produce (f), the theoretical ring flux for (c).
electrogenerated at the disc by the oxida- with the reverse reaction of being used at
tion of Br− according to the reaction the ring to detect the amount of bromine.
−e− In the gap between disc and ring, reaction
Br− −−−→ 12 Br2 (28) between bromine and the unsaturated
hydrocarbon R occurs disc, enabling the electrochemistry of

solution photoactive species excited by
k2
R + Br2 −−−→ RBr2 (29) the light to be studied by measuring
the steady state photocurrents obtained
which reduces the amount of Br2 at the disc electrode. For stable species,
available for detection at the ring, these photocurrents tend to vary with
and hence the measured collection W −1/2 , as only photoexcited species that
efficiency Nk , such that Nk < N0 . The are produced in the diffusion layer
concentration of unsaturated hydrocarbon react at the electrode without being
was chosen to be greatly in excess [23] swept away, and the thickness of this
compared to the amount of Br2 produced; diffusion layer varies with W −1/2 (Eq. 8).
therefore the reaction followed pseudo- Systems studied include photogalvanic
first-order kinetics, with the first-order rate cells [25] and photoactive colloids [26].
constant given by More recently, analytical expressions have
been derived for analysis of the transient
k1 = k2 corg (30) and steady state photocurrents for the
PE process [27], where photoexcitation (P)
where corg is the bulk concentration of a sensitizer molecule (S) leads to
of organic hydrocarbon. For such a its electrochemical oxidation (E) at an
first-order homogeneous reaction, the electrode surface
relationship of Nk to N0 has been derived
analytically [24] as hν
S −−−→ S∗ photoexcitation
N0 0.20 ν 1/3
= k1 (31) k0
Nk W D S∗ −−−→ S quenching
Measurements of Nk from the measured
−e−
disc and ring currents S∗ −−−→ S+ electrode oxidation
jR = −Nk jD (32)
and these have been used to analyze the
as a function of W were therefore used to response of colloidal CdS particles [27].
determine k1 and then k2 from Eqs. (16 to This approach has been extended to the
19, 30 and 31). study of PCE processes [27], in which the
photoexcited S∗ chemically (C) reacts with
2.4.2.3 The Optical Rotating Disc a scavenger A to give A− and S+ , and it
Electrode (ORDE) is A− that is then electrochemically (E)
The ORDE consists of a quartz rod, oxidized on the disc.
polished at both ends, one of which is A related experimental setup, the
coated with a quasi-metallic antimony- rotating optical ring-disc electrode
doped tin oxide film to form a (semi-) (RORDE), has also been developed [28,
transparent disc electrode. Light is shone 29]. Illumination in this case occurs
down the rod and through the disc through a quartz disc, with detection of
electrode as it is rotated in solution. the photosensitized products occurring on
The rotation of the ORDE imposes a platinum ring. This ring electrode has the
RDE hydrodynamics at the transparent advantage of being more chemically inert
than tin oxide, allowing it to be used to This section will focus on the major
study a wider variety of photoactive species. solution flow hydrodynamic systems,
initially looking at those in which
2.4.3 fluid exits a tube via a nozzle and
Solution Flow Hydrodynamic Systems impinges on a stationary electrode
(or electrodes), and then at those in
Although the RDE and RRDE systems which an electrode (or electrodes) is
are extremely widely utilized, there are (are) incorporated into the tube itself.
some problems associated with them for It will be seen that many of these
some applications. One of these is that systems are not uniformly accessible;
the disc electrode rotates, which makes it although this can be a disadvantage
difficult to attach ancillary sensing devices when studying systems such as modified
such as spectrophotometers in close electrodes (see Sect. 2.4.2.1), in which
proximity to the electrode to monitor disc uniform mass transport simplifies analysis
products. To overcome this, hydrodynamic and ensures equivalence of reaction
systems have been developed in which across the entire electrode, it has been
the electrode remains stationary and the argued that nonuniformly accessible
solution moves. These also have the systems should have greater inherent
advantage of being compatible detection kinetic discrimination [31] when studying
systems for the flowing eluent from electrochemical reactions with coupled
dynamic solvent separation systems such chemical reactions. This is because in
as chromatographic columns. Unlike the many cases, analysis of these systems
RDE and RRDE, a packed bed electrode generally involves fitting experimental
can also easily be incorporated upstream data to theoretical curves. These curves
of the system through which the solution often predict the variation in the
must flow. This enables electrochemical effective number of electrons, neff , passed
redox reaction of reagents in the during reaction with the velocity of the
column, producing new species with 100% hydrodynamic solution [controlled by W
conversion in solution for subsequent for the RDE (see Sect. 2.4.2.1) and the
analysis in the system. An interesting volume flow rate Vf for the solution flow
example of the use of the Packed Bed systems below]. Nonuniformly accessible
Wall-jet Electrode (PBWJE) [30] is the electrodes show a variation in the mass
electrochemical detection of nitrate ion by transfer flux (and hence the balance of
means of the WJE (see Sect. 2.4.3.2). This importance of kinetic and hydrodynamic
involves the electrochemical production terms) across the electrode surface, which
of Cu2+ ion from a copper packed results in a more gradual change in neff
bed, which is then deposited by with Vf than for neff with W for the RDE.
reduction on the disc electrode of This more gradual change should enhance
a wall-jet electrode (WJE), producing the distinction between differing coupled
a fresh copper disc electrode. This chemical reaction mechanisms and lead to
fresh copper film is necessary for the greater accuracy in kinetic analysis. This
electrochemical reduction and detection has been demonstrated for measurements
of nitrate at a WJE, as reduction leads of the ECE reaction (electron transfer,
to progressive poisoning of the copper chemical reaction, electron transfer) at
electrode. both a WJE and an RDE [32]. Although
Fig. 6The wall-tube electrode d

configuration. Nozzle
H Flow
Electrode r
Unwin and Compton [33] have shown where Vf is the volume flow rate of
no significant distinction for first-order fluid (cm3 s−1 ) through the nozzle. When
chemical processes, they have postulated comparing this equation to Eq. (11), it is
that this will not be the case for second- clear that Vf is the equivalent variable for
and higher-order processes. the WTE to W for the RDE, as is the case
for all solution hydrodynamic systems.
2.4.3.1 The Wall-tube Electrode (WTE)
The wall-tube electrode system involves 2.4.3.2 The Wall-jet Electrode (WJE)
pumping the electrolyte solution through The WJE has a similar configuration to
a nozzle onto a disc electrode as a jet the WTE, but in this case the nozzle
of fluid (Fig. 6). In this arrangement, the diameter d is very much smaller than
nozzle diameter d is larger than the disc the electrode diameter 2r. This produces
electrode diameter 2r, and electrode and the hydrodynamic flow pattern shown in
jet are arranged coaxially. Under these Fig. 7.
conditions, the factors controlling the The limiting current for the WJE
mass transport hydrodynamics of a WTE is (under conditions in which the
are well established [34]. The mass transfer electrode–cell wall separation is very
characteristics are analogous to the RDE, much larger than the electrode–nozzle
with essentially uniform accessibility separation) [15, 36]
across the electrode surface [35]. There
iL = 1.38nF D 2/3 ν −5/12 d −1/2 r 3/4 c∞ Vf
3/4
is, however, a slight dependence of the
(34)
mass transport–limited current on the
As can be seen from Fig. 7, the electrode
nozzle–electrode separation H , with [34]
is not uniformly accessible; in fact the
thickness of the diffusion layer varies with
iL = 12.08nFAD 2/3 ν −1/6 c∞
r 5/4 and hence from Eq. (9), the mass

Vf 1/2 H −0.054 transport–limited flux at the electrode
× (33)
πd 3 d varies with r −5/4 . Figure 7 also shows
Diffusion layer
Fig. 7 The flow pattern at the

WJE.
that away from the center of the jet in and the EQCM allowed simultaneous
the diffusion layer, flow is away from the measurement of the mass loss from the
electrode, ensuring that a fresh electrolyte copper disc electrode and the current
from the jet is efficiently brought to due to copper oxidation and dissolution.
the electrode surface. Also, the relatively Correlation of these two measurements
narrow jet decreases the volume flow allowed direct determination of the
rate of solution required for controlled variation of the overall copper dissolution
hydrodynamics, decreasing the volume valency (between 1 and 2) as a
required for analysis compared with the function of the chloride concentration and
WTE, and the jet spreads radially across the the hydrodynamic conditions employed.
electrode surface, ensuring that a relatively Although there has been one study of
large proportion of the reagent in the jet the deposition of clays using an EQCM
can be detected on the electrode, thereby disc system constructed in the RDE
increasing detection sensitivity. As a result, configuration [42], this is more of a
the WJE has been used as a sensitive technological and engineering challenge
analytical electrochemical detector in flow- than using the WJE configuration. An
through chromatographic systems [37, 38]. overview of the EQCM and its applications
Two further applications demonstrate is the subject of Chapter 2.7 of this volume.
the relative ease of incorporation of
ancillary electronic measurement systems 2.4.3.3 The Wall-jet Ring-disc Electrode
at a stationary disc in the WJE (WJRDE)
compared with a rotating disc in the The WJRDE system has also been
RDE. The first involves using the WJE developed [43], and this has enabled
configuration to measure the rate of still greater sensitivity in analytical
deposition of colloidal particles on a detection. Fundamentally, the WJRDE is
(semi-) transparent doped tin oxide disc a WJE in which the disc electrode is
electrode as a function of electrode replaced by a ring-disc electrode (see
potential, by measuring the degree of Sect. 2.4.2.2). The hydrodynamic regime
light scattering of the adsorbed particles of the WJE leads to lower collection
on the transparent electrode using total efficiencies N for the WJRDE (from
internal reflection microscopy [39, 40]. Eq. 15) compared to an RRDE of similar
In this work, the predicted radial dimensions, as significant flow in the
variation of the mass transport–limited diffusion layer is away from the electrode
flux across the electrode proved extremely surface. Also, unlike the RDE, this
useful in assessing whether deposition collection efficiency N varies (albeit
occurred under mass transport or surface predictably and reproducibly) with Vf .
electrochemical control. The second Despite this, the WJRDE configuration
concerns measurement of the kinetics is particularly applicable to analytical
and mechanism of copper dissolution detection of species in flow injection
in acid chloride media [41]. This work analysis and liquid chromatography [44],
involves the use of an electrochemical and particularly those species that take
quartz crystal microbalance (EQCM) as the part in plating–stripping reactions [45].
disc electrode in the WJE configuration. This configuration enables reagents in
Controlled dissolution was achieved the jet to be electrochemically deposited
because of the WJE hydrodynamics, (plated) and accumulated on the surface
Fig. 8 The tube electrode

system.
Vf r
Electrode
x
of the disc by maintaining a suitable therefore varies with x −1/3 across the
disc potential. The reaction can then be electrode surface, and hence like the WJE,
reversed at the disc by changing potential, this electrode is not uniformly accessible.
stripping the reagents off the electrode If long electrodes or slow flow conditions
in a concentrated burst, to be detected are avoided, so that the diffusion layer
by the downstream ring electrode. This thickness does not become comparable to
preconcentration of reagents on the disc the tube radius (resulting in the depletion
markedly increases the overall detection of reagents extending into the center of the
sensitivity of the system when compared tube), the limiting current is given by the
to the WJE. Levich equation for the TE [5]
1/3
2.4.3.4 The Tube or Tubular Electrode (TE) iL = 5.43nF (Dl)2/3 Vf c∞ (36)
The TE system consists of an annular
electrode that is inserted into the length of where l is the length of the electrode
a cylindrical tube down which the solution (i.e. it extends from x = 0 to x = l).
under study flows (Fig. 8). This equation is derived assuming flow
In this simple configuration, the volume is sufficiently fast such that diffusion in
flow rate Vf is chosen to ensure laminar the direction of convective flow can be
flow of solution in the cylindrical tube. considered to have a negligible effect,
Because of friction with the tube walls, the convection is sufficient to ensure that the
velocity of liquid v in the tube varies in a concentration gradient of reactive species
parabolic manner with the distance r from near the electrode can be linearized and
the tube center [46, 47] the diffusion layer thickness is much
2
smaller than the tube dimensions across
r the whole electrode surface [48]. Under
v = v0 1− (35)
r0 these conditions, the limiting current does
not depend upon either r0 or (unlike the
where v0 is the velocity at the center of RDE and RRDE) the kinematic viscosity
the tube and r0 is the radius of the tube, of the solvent. At the slow flow or long
so that r ≤ r0 . As the flow is laminar, the electrode limit (often termed the thin layer
diffusion layer thickness (the equivalent cell configuration), in which complete
of zD for the RDE, see Sect. 2.4.2.1) conversion of the reagents in the tube
increases with the distance x, downstream occurs by the end of the electrode, the
from the upstream edge of the tube simple expression
electrode, actually as x 1/3 . The mass
transport–controlled flux (from Eq. 9) iL = nF Vf c∞ (37)
applies. A transition therefore occurs one-electron oxidation of N, N, N , N -

between these two limiting analytical tetramethyl-p-phenylenediamine (TMPD)
expressions (Eqs. 36 and 37) at
NMe2 NMe2
intermediate values of Vf , but the complete
theoretical description of the response
−e −
of the TE can only be determined
numerically (see Sect. 2.4.4) [48]. This
does mean that the hydrodynamic NMe2 NMe2
conditions of the TE can be varied to
control the degree of electrochemical confirming both the stability of the
conversion of reagents in solution and radical and the viability of this approach.
that unlike the systems considered Theoretical and experimental studies
previously, essentially 100% conversion were subsequently extended to unstable
of reagents to products can be achieved radicals that decompose by simple
in the thin electrode configuration first-order kinetics [50] and those that
by judicious choice of electrode decompose by second-order kinetics [47],
length and flow. This can often be for which the TE esr system allowed
useful in electroanalysis, for example, direct measurements of the first-
in the design of electrochemical order and second-order rate constants,
chromatographic detectors. respectively.
The detection and generation efficiency The analysis of steady state voltammetric
and the controlled hydrodynamics of curves at a TE has also been achieved
this system have also been utilized in for electrochemically reversible [51], quasi-
the electrochemical generation, detection, reversible [52], and irreversible [52] elec-
and characterization of radicals. In this tron transfer reactions by numerical
methods (see Sect. 2.4.4), which enables
work, a TE was positioned just upstream
experimental voltammetric curves to be
of an esr spectrometer cavity, so that
analyzed to give electrochemical kinetic
the radicals generated were transported
information.
into the cavity without the electrode
destabilizing the cavity resonance. Not
only does esr allow identification and 2.4.3.5 The Channel Electrode (CE)
characterization of the radical generated The CE system is shown in Fig. 9. In
but the reproducible hydrodynamics allow hydrodynamic terms it is closely related
the theoretical variation of esr signal to the TE system.
intensity with Vf to be calculated. Initial It consists of an electrode embedded in
work concentrated on stable radical the wall of a tube down which solution
systems, in which both steady state flows. Generally, h d to ensure edge
effects on the flow are negligible. As with
esr signal intensities [46] and the time-
the TE (see Sect. 2.4.3.4), the solution is
dependent response of the esr signal
under laminar flow, with the parabolic
intensity to a step change in tube electrode
variation in velocity with height given
current [49] were measured and compared
in this case by
with theory. In this work, good agreement
y 2
was found between the theoretical and
v = v0 1 − (38)
experimental signal intensities for the h
a 2h Flow
Flow w
x
l
(a) (b)
Fig. 9 The channel electrode system seen from (a) above and (b) the side. The
electrode dimensions are width w and length l. The channel width and height are d and
2h, respectively, with y, the height variable, ranging from −h (at the electrode surface)
to h and with y = 0 corresponding to the center of the channel.
The diffusion layer thickness also varies current shows a smooth transition be-
with x 1/3 , which from Eq. (9) means that tween Eqs. (38 and 39) with decreasing
the mass transport–limited flux across flow and increasing electrode length. Thus,
the CE varies with x −1/3 , so that the like the TE, the CE system can be used in
electrode is nonuniformly accessible. An the thin layer cell configuration for the
analogous Levich equation to the TE has complete detection of electrochemically
been derived analytically for the CE. This active species in chromatographic sepa-
gives the mass transport–limited current rations [55]. The CE, like the TE, can also
for fast flow (where as with the TE, be employed for the efficient generation
diffusion in the direction of convective and detection of radicals, by combining
flow can be considered to be negligible and the CE with radical detection and quantita-
where convection is so efficient that the tion using an esr spectrometer. However,
concentration gradient of reactive species unlike the TE, the CE system can in
near the electrode can be linearized [53]) fact be placed directly in the center of
and for short electrodes (in which the the cavity of the esr spectrometer without
diffusion layer can be considered to be thin significant perturbation of the resonant
compared to the height of the channel) as sensitivity, as long as the cell is constructed
2/3 1/3 of synthetic silica [56]. This enables detec-
Dl Vf tion of radical intermediates and products
iL = 0.925nF w c∞
h d with much shorter lifetimes than in the
(39) TE configuration. Concentration distribu-
As with the TE, at slow flow and for long tions of electrogenerated radicals have
electrodes, complete conversion of the therefore been calculated under steady
electroactive species can occur, and under state conditions for model electrode re-
these conditions the TE thin layer cell equa- action mechanisms and have been used
tion (Eq. 37) applies. Like the TE, these are to produce the theoretical esr signal inten-
limiting cases, and the complete rigorous sity [57–59]. Analysis of experimental data
description of the response of the CE at using this theory has then allowed elec-
all flows and electrode lengths requires trode reaction mechanisms to be verified
numerical methods (see Sect. 2.4.4) [48]. and radical lifetimes to be measured.
However, for both these systems this is rel- A further advantage of the CE system
atively straightforward [54] (see Sect. 2.4.4) is that unlike the TE, a window of trans-
and the resulting mass transport–limited parent material can be incorporated into
the channel unit in the region of the elec- AQH + e− −−−→ [AQH]•−
trode [60–62]. Incorporation of this win-
dow either directly above or downstream E step (42)
of the electrode allows spectroscopic inter-
Identification of the esr spectrum and
rogation and/or photoexcitation of electro-
quantitative analysis of the esr signal
generated products or intermediates with
intensity as a function of electrogeneration
light, while a window either above or
current and channel flow conditions
upstream of the electrode allows spectro-
established that the reaction given in
scopic interrogation and/or photoexcita-
Eq. (40) occurred in the absence of light,
tion of reagents prior to electrochemical
producing the stable [AQBr]•− radical.
reaction. A good example of the use of
Upon irradiation, the esr spectrum of
such a system is in the electrochemical re-
[AQH]•− was observed at slow flow; the
duction of 1-bromoanthroquinone (AQBr),
loss of the esr intensity of the [AQBr]•−
O Br radical was then quantitatively analyzed
as Vf was varied, confirming the reaction
mechanism and producing a value for the
rate constant of the C reaction (Eq. 41)
as k = 0.047 s−1 , in good agreement with
O other independent measurements [64, 65].
which is simultaneously irradiated with
488-nm light [63]. Detection of the radical 2.4.3.6 The Double Channel Electrode
products of the reaction was by means of (DCE) and Double Tube Electrode (DTE)
an esr spectrometer. This work established The DCE system (Fig. 10) is the equivalent
that reaction occurred by a photo-ECE configuration for the CE as the RRDE
mechanism (see Sect. 2.4.2.2) is for the RDE (see
Sect. 2.4.2.1).
AQBr + e− −−−→ [AQBr]•− In this configuration, a second or
E step (40) collector channel electrode is placed
downstream of the first or generator CE to
488 nm, k
[AQBr]•− + hν −−−−−−→ Br− + AQH detect products from the electrogeneration
reaction [66]. An analogous system can
Photo-assisted C step (41) also be constructed for the TE, with a
x3
a 2h Flow
w
Flow
x1
x1
x2 x2
(a) (b) x3
Fig. 10 The DCE system: (a) top view and (b) side view. x1 , x2 , and x3 are the first or
generating channel length, the distance to the inner edge of the second or collecting
channel electrode and the distance to the outer edge of the collecting channel
electrode, respectively, along the flow path.
second tube electrode positioned with its for N :

inner and outer edges at distances of
x2 and x3 downstream of a generating N = 1 + λ2/3 [1 − F (θ)] − (1 + λ + θ)2/3

electrode of length x1 (Fig. 8), producing a θ
× 1−F (1 + λ + θ)
DTE [67, 68]. As the hydrodynamics of the λ
TE and CE are essentially the same, the
θ
DCE and DTE have similar characteristics; −F (44)
λ
hence, in this section, only the DCE will
be discussed. which has been experimentally veri-
The collector channel electrode of the fied for the reversible ferri/ferrocyanide
DCE performs a similar function to couple [72], whereas the second treat-
the ring electrode of the RRDE (see ment [71], which assumes a uniform cur-
Sect. 2.4.2.2), electrochemically detecting rent distribution on the generator elec-
products electrogenerated on the upstream trode, applicable to electrochemically irre-
electrode, in this case the generator versible systems, gives
channel electrode. Usually, the generator
1+θ
electrode is held at a potential that ensures N = 1 − (1 + λ + θ)F
λ
mass transport–limited electrogeneration
of product and the collector electrode θ
+ (λ + θ)F
is held at a potential to ensure λ
33/2 λ2/3
mass transport–limited electrochemical
reaction of this product. [Like the RRDE + (1 + θ)1/3 − λ1/3
2π
(see Sect. 2.4.2.2), when more than one
product is formed, the collector electrode (45)
material and the potential can be chosen In both cases
to selectively detect the product of choice.] x2 − x1
θ= (46)
As with the RRDE, the theoretical x1
collection efficiency N of the collector
and
electrode (see Eq. 15) has been determined x3 − x2
λ= (47)
for the simplest case in which one stable x1
electroactive product is formed at 100%
and F (θ) is the function defined in
efficiency
Eq. (17). It should be noted that as
for the RRDE, both treatments predict
jcoll = −Njgen (43) N to be dependent only on electrode
geometry and not on flow conditions.
where jcoll and jgen , are the collector and However, the nonuniform accessibility
generator electrode fluxes, respectively, of the channel configuration results in
determined from the currents by a nonuniform current distribution of
Eq. (7) [66, 69–71]. In the first treatment the channel electrode, which leads to
[66, 69, 70], a uniform surface different values of N for reversible
concentration was assumed across and irreversible electrochemistry at the
the generator electrode, applicable to generator electrode. This is because the
electrochemically reversible systems. This reversible system (the first treatment)
produces the following relationship assumes rapid electrode kinetics and a
constant concentration of reactant and NH NH

product across the electrode surface,
which results in more current being k
+NHMe2
passed at the upstream end of the OH−
generator electrode than in the irreversible
+NMe2 O
case. Products have more chance of
escaping from the upstream part of the (49)
generator electrode into bulk solution The aromatic products of both the
and evading the collector electrode. As first step (i.e. oxidized ADMA) and
a consequence, the collection efficiency the second step (following deamination)
is greater for the irreversible system. are electrochemically active, and can
This variation in collection efficiency can be detected under mass transport
in principle be used to determine the control at 0.0 V and −0.2 V, respectively,
electrochemical kinetics on the generator versus the saturated calomel electrode
electrode for those reagents in which on the collector electrode. Setting the
the electrochemical and mass transfer potential at −0.2 V therefore detects
rate constants are comparable at the the current due to both the products.
generator electrode. Under slow flow As expected, when this current was
conditions, the rate of mass transfer is rate- used to calculate the experimental
limiting and the system can be considered collection efficiency N , it closely
reversible, with N given by Eq. (44); coincided with that expected for a
however, for fast flow, if the surface stable reversible electrochemical reaction
electrochemical reaction becomes rate- (Eq. 44). However, measurement of the
limiting, the reaction can be considered current at 0.0 V also enabled the
irreversible, with N given by Eq. (45). The relative fluxes at the collector electrode
flow condition at which the transition due to oxidized ADMA and the
between these two limiting values occurs product of the chemical reaction to be
is therefore a useful measure of the deconvoluted. Increasing Vf allowed the
electrochemical kinetics [48]. transit time from generator to collector
Like the RRDE (see Sect. 2.4.2.2), the to be decreased, which decreased the
DCE has also been extensively used to importance of the following chemical
study the kinetics and mechanisms of reaction. Comparison of the experimental
electrode reactions involving following variation of N with Vf for each
chemical reactions. A good example is product with the theoretical response
the electrooxidation of 4-amino-N, N - solved through numerical methods [73]
dimethylaniline (ADMA), which proceeds allowed a value of k to be determined
by an EC reaction in basic aqueous as k = 1.8 × 104 dm3 mol−1 s−1 . This
solution [71]. was in good agreement with previous
spectrophotometric measurements [72],
NH2 NH
demonstrating the applicability of this
approach.
−2e− − H+ As with the RRDE (see Sect. 2.4.2.2),
transient collection measurements are
NMe2 +NMe2 also possible with the DCE. Transient
(48) data collection and analysis is extremely
important for measuring fluxes due to and θ, λ, and N are given by Eqs. (46, 47
the redox reaction of finite amounts and 44), respectively.
of species adsorbed on the generator Transient current DCE studies have
electrode (e.g. for modified electrodes, see been utilized in the investigation of the
Sect. 2.4.2.2) or for reactions in which mechanism of anodic metal dissolution.
generation is accompanied by adsorption This reaction may be simply written as
of products on the generator electrode.
−ne−
The time-dependent collection efficiency n+
M(s) −−−→ M(aq) (55)
Nτ , has been calculated for a stable
species generated under electrochemically but for many metals the actual mechanism
reversible conditions and collected under is much more complex, often involving
mass transport–controlled conditions adsorbed intermediates. Such processes
with 100% efficiency. This allows the have been studied using transient
time-dependent variation of the collector measurements on the RRDE ([75], see
electrode current to be calculated Sect. 2.4.2.2), but transient measurements
from Eqs. (7 and 15) for a constant have also successfully been carried out
generator electrode current. Nτ is the using the DCE [76, 77]. In this work,
collection efficiency as a function of the a galvanic current step on the metal
dimensionless time τ , which is related to generator electrode was used to induce
the time t, by oxidation and dissolution of the metal
1/3 according to Eq. (55). The collector
4v02 D electrode was then set to a potential to
τ = t (50)
h2 l 2 cause mass transport–limited reduction
and detection of the metal ions, and
Analytical expressions have been derived the transient collection current recorded.
both at short τ and long τ [74], and
Some of these reactions take place via
these have been shown to correlate closely
adsorbed intermediates
with the general Nτ function determined
numerically (see Sect. 2.4.4), applicable at −ne− n+
M(s) −−−→ M(ads)
all τ [74]. This general function can be
reproduced to within 4.5% by the simple n+
M(ads) n+
−−−→ M(aq)
analytical expression
Nτ 1 + tanh in which case adsorption on the generator
= (51) electrode causes the experimental collec-
N 2
tor current to rise more slowly with time
in the range 0.1 ≤ θ ≥ 1.0 and 0.5 ≤ λ ≥ to its steady state value, after the onset of
1.5, where metal oxidation, than would be expected
2 in the absence of adsorption. The ex-
1
= E− τ − (E)(B)2/3 pected collector current, in the absence of
B adsorption, can be calculated from the gen-
1.5 1 erator electrode current and Eqs. (7, 15, 44,
+ − (52) and 51) (or, more accurately, the current
B 4/3 2τ
predicted using the general Nτ function
E = 1.44 − 0.386λ (53)
calculated numerically, rather than Eq. 51).
B = 2.53θ + 0.38λ + 0.9 (54) The difference in charge (produced by
integrating the current with time) between Luminescence was measured through
the expected collection in the absence of a transparent window (see Sect. 2.4.3.5)
adsorption and the experimental collec- using a charge coupled device (CCD)
tion has therefore been used not only as a detector. The variation of ECL intensity
method for determining whether adsorp- as a function of Vf permitted kinetic
tion takes place but also as a quantitative and mechanistic modeling of the ECL
measure of the total number of moles of reaction.
adsorbed intermediate on the generator
electrode [76, 77]. 2.4.4
An interesting application of the DCE is Numerical Calculations
in the study of electrochemiluminescent
reactions [78]. In this work, the generator Although analytical solutions can be
and collector electrodes were used to derived for many reactions using the
generate the reduced and oxidized forms RRDE (see Sect. 2.4.2.2), owing to the
of diphenylaniline (DPA) according to simplifications afforded by its circular
the equations symmetry [3], there are many reactions,
particularly those with complicated
gen, +e− homogeneous kinetics, for which the
DPA −−−−−→ DPA− theoretical response of the RRDE requires
a numerical solution. An introduction
coll, −e−
DPA −−−−−→ DPA+ (56) and overview of numerical methods was
given in Chapter 1.3 of this volume. Bard
and Prater [79] and Feldberg [80, 81] were
These species were then mixed under
the first to develop suitable methodology.
the controlled hydrodynamics of the DCE.
Their approach was to divide the solution
This gave the excited state, DPA∗ , which
in the vicinity of the RRDE into a grid
luminesces because of the following
(Fig. 11). The circular symmetry of the
reactions:
system ensures that all variables are
circularly symmetric, which reduces the
DPA− + DPA+ −−−→ DPA∗ + DPA
numerical analysis to a two-dimensional
−−−→ 2DPA + hν (57) problem, requiring a two-dimensional grid
Disc r Ring
Fig. 11 The division of the

solution near the RRDE into a
two-dimensional grid.
as shown. As long as one chooses a where 1/2

grid with enough elements, any smooth W3
C = −8.0 (59)
function, f (r, z), (such as concentration) ν
can then be approximated to a series of
The final term in Eq. (58) is present when
discrete values, one of which is assigned
analyzing electron transfer processes that
to each box location r, z. The derivatives
include a coupled homogeneous chemical
of these functions can then simply be
reaction. For example, for first-order decay
obtained from the differences between the
of i, the term is negative, m = 1, n =
neighboring boxes.
0, and k is a first-order rate constant
Simulation occurs with successive time
(s−1 ), whereas for bimolecular decay of
steps, t from t = 0. The initial conditions
i by reaction with j , m = 1, n = 1,
(at t = 0) define the starting configuration
and k is a second-order rate constant
of the system. Boundary conditions at the (cm3 mol−1 s−1 ). A similar convective-
disc and the ring and in the gap ensure diffusion equation exists for each species,
that the simulation remains physically i, j , and so on; these can be solved
realistic (typical boundary conditions to give ∂c/∂t for all boxes and all
are that the concentration gradient of species. These ∂c/∂t terms then enable
electrochemically active species at the the overall changes in concentration in
disc is set by the current at the disc, each box over a short time step t
that this concentration gradient must to be determined, which gives the new
be zero at the insulator between disc concentrations of each species in each box
and ring and that the concentrations at time t.
in the bulk solution must remain This process can be successively re-
constant at the bulk concentration value peated, generating concentration profiles
at all t). Numerical calculation then for all species at times 2t, 3t, and so
typically involves the calculation of on until steady state conditions have been
the concentration differences between reached. The flux (and hence the current
neighboring boxes at the initial time, from Eq. 7) at the ring electrode at any time
t =0, to give the concentration gradients at any point on the ring electrode surface
∂ci /∂z, ∂ci /∂r, ∂ 2 ci /∂z2 , and ∂ 2 ci /∂r 2 is then simply determined by Fick’s first
for the redox-active species i. The law at the electrode surface (z = 0) for the
convective-diffusion equation for the redox-active species i
system is then used to calculate dci /dt
∂ci
for each box, with the controlled j = −D (60)
hydrodynamics of each system ensuring ∂z z=0
reproducible, known, time-independent To produce the most accurate values of
solution convection. For example, for the ring current, the concentration gradients
RDE, the convective-diffusion equation for must be most accurately determined at
species i is and near the ring electrode, where these
gradients are usually the steepest. A finer
dci ∂ 2 ci ∂ci grid is often used on the innermost box,
= Di 2 + Cz2
dt ∂z ∂z near the ring electrode [82], to achieve this.
∂c Also, non-Cartesian space has been used,
− Crz ± kcim cjm . . . (58) replacing the Cartesian coordinate z with
∂r
w, where [83] variable r in Fig. 11 and the y-component,

which is the channel height variable
x −x 3 (Fig. 9 and see Sect. 2.4.3.5), replacing z
exp dx
3 in Fig. 11 [55, 85, 86]. The generator and
w = 0∞ 3 (61)
−x collector channel electrodes then occupy
exp dx similar positions in this grid as the ring
0 3
and disc, respectively, in Fig. 11. In order
where to make the simulation two-dimensional
and hence simple, the electrode width
W 1/2
x=2 z (62) (w in Fig. 10) is usually assumed to be
D 1/3 ν 1/6 sufficiently large to ensure uniformity
This has two effects. Firstly, instead in this dimension (although if this is
of the z variable extending from 0 to not the case, simulation, although more
infinity, the w variable extends from complicated, is still perfectly feasible).
0 to 1, thereby reducing the number As long as the system has laminar
of boxes required in the calculation. flow, the convective-diffusion equation for
Secondly, this transformation linearizes this system is
the concentration gradients, which makes
dci ∂ 2 ci y 2 ∂c
the contribution of each box to the total = Di 2 − ν0 1 − 2
dt ∂y h ∂r
calculation more equal. These effects
combine to greatly enhance the efficiency ± kcim cjm (64)
of the calculation.
as the convective flow in the system is
Similar numerical calculations can
determined by Eq. (38). The backwards
also be carried out on the WJRDE
implicit method has been found to be the
and the WTE because of their similar
best method of calculation for the DCE
circularly symmetric geometry. These
system [55].
involve using the appropriate convective-
diffusion equation for the hydrodynamic
2.4.5
system of study. For the WJRDE this
Microhydrodynamic Systems
is [84]
dci ∂ 2 ci ∂ci ∂c With recent improvements in microma-

= D i 2 + vz − vr ± kcim cjm chining and microfabrication, solution
dt ∂z ∂z ∂r
(63) flow hydrodynamic systems have been
where vz and vr are the solution velocity produced on the micrometer scale. A mi-
components in the z- and r-directions, crometer scale WTE (MWTE) system has
respectively. As for all these hydrodynamic recently been developed [87], as has a mi-
electrodes, the variation of vz and vr with crojet electrode (MJE) [88–90]. This has
r and z are well established. similar dimensions to the MWTE, but in
For the DCE (and the DTE), such circular the MJE the nozzle has been placed off-axis
symmetry does not apply. In this case, a to enhance mass transport to the electrode.
two-dimensional grid in x and y is again A micromachined WJRDE (MWJRDE) has
constructed as in Fig. 11, but with the x- also been produced and characterized [91]
component, which is the channel length and microchannel electrode (MCE) and
variable (see Sect. 2.4.3.6), replacing the microchannel double electrode (MCDE)
systems have been made both by a determined (see Sect. 2.4.3.6). There is
sandwiching method [92] and by lithog- a viable solution velocity range for any
raphy [93]. The fundamental question is hydrodynamic system; for example, if the
‘‘why is there this drive toward fabrication velocity is too low, irreproducible stirring
on the microscale?’’ occurs, whereas if it is too high, turbulent
The first and most obvious answer rather than radial flow occurs, which again
is the reduction in volume inherent causes instability and irreproducibility.
in reducing the hydrodynamic electrode Reduction of system dimension therefore
scale. It has been shown (see Sect. 2.4.3) allows shorter transit times and faster
that all solution flow hydrodynamic reaction kinetics to be accessed within
systems produce fluxes that are this viable solution velocity range [94]. The
directly proportional to Vf (usually to shortest response time for any system can
some fractional power) and inversely also be determined in transient studies
proportional to electrode scale. This is by the electrode double layer charging
because it is the solution velocity that time, and this is also reduced with
determines the hydrodynamic regime. electrode dimension. It is to be expected
Reduction in electrode scale therefore that these advantages will ensure an
leads to a comparable solution velocity increase in the use of microhydrodynamic
at lower Vf , which allows experimental electrode systems. Further applications
measurement and theoretical analysis of microelectrochemical systems are the
of lower volumes of solution. This subject of the next chapter.
is compatible with microanalysis in,
for example, biological systems and References
microchromatography. Furthermore, the
lower Vf values required enable less 1. D. E. Williams, K. Ellis, A. Colville et al., J.
technologically complex pumping systems Electroanal. Chem. 1997, 432, 159.
to be employed. The second answer is 2. G. Macfie, R. G. Compton, J. Electroanal.
Chem. 2001, 503, 125.
that the collection efficiency of detector 3. W. J. Albery, M. L. Hitchman, Ring-Disc Elec-
electrodes (such as in the DCE, the trodes, Clarendon Press, Oxford, 1971.
DTE, and the WJRDE systems) is 4. W. J. Albery, S. Bruckenstein, J. Electroanal.
increased as the gap between generator Chem. 1983, 144, 105.
and collector electrodes is reduced. 5. V. G. Levich, Physicochemical Hydrodynam-
ics, Prentice Hall, Englewood Cliffs, N.J.,
Thus, micromachined systems offer the 1962, pp. 60–78.
possibility of enhanced detection. The 6. J. Koutecky, V. G. Levich, Zh. Fiz. Khim.
final answer is that as the distance 1956, 32, 1565.
between generator and collector electrodes 7. W. J. Albery, A. R. Hillman, Annu. Rep. C
is reduced, the transit time between 1981, 78, 377.
8. W. J. Albery, A. R. Hillman, J. Electroanal.
electrodes at a given solution velocity Chem. 1984, 170, 27.
is decreased. This transit time often 9. C. P. Andrieux, J. M. Saveant, Molecular De-
determines the fastest kinetic rate constant sign of Electrode Surfaces, Wiley Interscience,
that can be measured by the system for New York, pp. 207–270.
unstable species, for if the kinetics are 10. M. E. G. Lyons, Electroactive Polymer Elec-
trochemistry Part 1; Fundamentals, Plenum
so fast that the reaction is completed in Press, New York, 1994, pp. 237–371.
the zone between generator and collector, 11. W. J. Albery, P. N. Bartlett, J. Electroanal.
a rate constant cannot accurately be Chem. 1985, 194, 211.
12. M. Sharp, B. Lindholm, E. Lotta-Lind, J. 39. W. J. Albery, R. A. Fredlein, G. R. Kneebone

Electroanal. Chem. 1989, 274, 35. et al., Colloids Surf. 1990, 44, 337.
13. P. Jennings, A. C. Jones, A. R. Mount et al., 40. M. Itagaki, M. Tagaki, K. Watanabe, J. Elec-
J. Chem. Soc., Faraday Trans. 1997, 93, 3791. troanal. Chem. 1997, 440, 139.
14. W. J. Albery, S. Bruckenstein, Trans. Faraday 41. A. Göransson, C. Trägårdh, J. Colloid Inter-
Soc. 1965, 62, 1920. face Sci. 2000, 231, 228.
15. W. J. Albery, C. M. A. Brett, J. Electroanal. 42. N. Shirtcliffe, Colloids Surf. 1999, 155, 277.
Chem. 1983, 148, 201. 43. W. J. Albery, C. M. A. Brett, J. Electroanal.
16. A. R. Mount, D. Clark, M. S. Appleton et al., Chem. 1983, 148, 201, 211.
J. Electroanal. Chem. 1992, 334, 155. 44. J. C. Hoogvliet, F. Elferink, C. J. Van Der
17. A. R. Mount, D. Clark, M. S. Appleton et al., Pool et al., Anal. Chim. Acta 1983, 153, 149.
J. Electroanal. Chem. 1992, 340, 287. 45. C. M. A. Brett, M. M. P. M. Neto, J. Chem.
18. W. J. Albery, E. J. Calvo, J. Chem. Soc., Fara- Soc., Faraday Trans. 1 1986, 82, 1071.
day Trans. 1 1983, 79, 2583. 46. W. J. Albery, B. A. Coles, A. M. Couper, J.
19. W. J. Albery, A. R. Mount, J. Chem. Soc., Electroanal. Chem. 1975, 65, 901.
Faraday Trans. 1 1989, 85, 1181, 1189, 3717. 47. W. J. Albery, A. T. Chadwick, B. A. Coles
20. W. J. Albery, J. S. Drury, M. L. Hitchman, et al., J. Electroanal. Chem. 1977, 75, 229.
Trans. Faraday Soc. 1971, 67, 2162. 48. P. R. Unwin, R. G. Compton, Comp. Chem.
21. P. J. Colby, Ph.D. thesis, University of Lon- Kin. 1989, 29, 290.
don, London, 1983. 49. W. J. Albery, R. G. Compton, A. T. Chadwick
22. W. J. Albery, M. G. Boutelle, P. J. Colby et al., et al., J. Chem. Soc., Faraday Trans. 1 1980,
J. Chem. Soc., Faraday Trans. 1 1982, 78, 2757. 76, 1391.
23. J. Malyyszko, E. Malyszko, E. Rutkowska- 50. W. J. Albery, R. G. Compton, J. Chem. Soc.,
Ferchichi et al., Anal. Chim. Acta 1998, 376, Faraday Trans. 1 1982, 78, 1561.
357. 51. W. J. Blaedel, L. N. Klatt, Anal. Chem. 1966,
24. W. J. Albery, S. Buckenstein, Trans. Faraday 38, 879.
Soc. 1966, 62, 1946. 52. W. J. Blaedel, L. N. Klatt, Anal. Chem. 1966,
25. W. J. Albery, Acc. Chem. Res. 1982, 15, 142. 39, 1065.
26. W. J. Albery, P. N. Bartlett, J. D. Porter, J. 53. M. A. Lèvêque, Ann. Mines Mem. Ser. 12
Electrochem. Soc. 1984, 131, 2892, 2896. 1928, 13, 201.
27. C. Boxall, W. J. Albery, Phys. Chem. Chem. 54. J. Lankelma, H. Poppe, J. Chromatogr. 1976,
Phys. 2000, 2, 3631, 3641, 3651. 125, 375.
28. D. C. Johnson, E. W. Resnic, Anal. Chem. 55. J. L. Anderson, S. Moldoveanu, J. Electroanal.
1972, 44, 637. Chem. 1984, 179, 107.
29. P. R. Gaines, V. E. Peacock, D. C. Johnson, 56. B. A. Coles, R. G. Compton, J. Electroanal.
Anal. Chem. 1975, 47, 1393. Chem. 1983, 144, 87.
30. D. Pletcher, Z. Poorabedi, Electrochim. Acta 57. R. G. Compton, B. A. Coles, G. M. Stearn
1979, 24, 1253. et al., J. Chem. Soc., Faraday Trans. 1 1988,
31. R. G. Compton, P. R. Unwin, J. Electroanal. 84, 2357.
Chem. 1986, 205, 1. 58. R. G. Compton, B. A. Coles, M. B. G. Pilking-
32. R. G. Compton, A. C. Fisher, G. P. Tyley, J. ton, J. Chem. Soc., Faraday Trans. 1, 1988, 84,
Appl. Electrochem. 1991, 21, 295. 4347.
33. R. G. Compton, P. R. Unwin, J. Electroanal. 59. A. M. Waller, R. J. Northing, R. G. Compton,
Chem. 1988, 245, 287. J. Chem. Soc., Faraday Trans. 1 1990, 86, 335.
34. D. T. Chin, C. H. Tsang, J. Electrochem. Soc. 60. T. R. Beck, Corrosion 1977, 33, 9.
1978, 125, 1461. 61. T. R. Beck, S. G. Chan, Corrosion 1981, 37,
35. W. J. Albery, S. Bruckenstein, J. Electroanal. 665.
Chem. 1983, 144, 105. 62. J. N. Harb, J. Electrochem. Soc. 1986, 133,
36. M. B. Glauert, J. Fluid Mech. 1956, 1, 625. 439C.
37. B. Soucaze-Guillous, W. Kutner, Electroanal- 63. R. D. Webster, R. A. W. Dryfe, J. C. Eklund
ysis 1997, 9, 31. et al., J. Electroanal. Chem. 1996, 402, 167.
38. T. McCreedy, P. R. Fielden, Analyst 1995, 64. R. G. Compton, J. C. Eklund, L. Nei, J. Elec-
120, 2343. troanal. Chem. 1995, 381, 87.
65. R. G. Compton, B. A. Coles, M. B. G. Pilking- 81. S. W. Feldberg, M. L. Bowers, F. C. Anson, J.

ton, J. Chem Soc., Faraday Trans. 1 1990, 86, Electroanal. Chem. 1986, 215, 11.
663. 82. W. J. Albery, C. C. Jones, A. R. Mount, Comp.
66. H. Gerischer, I. Mattes, R. Braun, J. Elec- Chem. Kin. 1989, 29, 129.
troanal. Chem. 1965, 10, 553. 83. J. M. Hale, Batteries Vol. 2, Pergamon Press,
67. G. W. Schieffer, W. J. Blaedel, Anal. Chem. Oxford, 1965, p. 147.
1977, 49, 49. 84. P. Laevers, A. Hubin, H.Terryn et al., J. Appl.
68. J. Hermann, H. Schmidt, W. Vielstich, Z. Electrochem. 1995, 25, 1023.
Phys. Chem. 1984, 139, 83. 85. S. Moldoveanu, J. L. Anderson, J. Electroanal.
69. H. Matsuda, J. Electroanal. Chem. 1968, 16, Chem. 1984, 175, 67.
153. 86. J. L. Anderson, S. Moldoveanu, J. Electroanal.
70. R. Braun, J. Electroanal. Chem. 1968, 19, 23. Chem. 1984, 179, 119.
71. R. N. Cook, D. Phil. thesis, University of 87. F. Bjrefors, J. Gadomska, M. Donten et al.,
Oxford, 1973, 313. Anal. Chem. 1999, 71, 4926.
72. K. Aoki, H. Matsuda, J. Electroanal. Chem. 88. J. V. Macpherson, M. A. Beaston, P. R.
1978, 94, 157. Unwin, J. Chem. Soc., Faraday Trans. 1995,
73. K. Aoki, K. Tokuda, H. Matsuda, J. Elec- 91, 889.
troanal. Chem. 1977, 79, 49. 89. R. D. Martin, P. R. Unwin, J. Electroanal.
74. K. Aoki, K. Tokuda, H. Matsuda, J. Elec- Chem. 1995, 397, 325.
troanal. Chem. 1986, 195, 229. 90. J. V. Macpherson, S. Marcar, P. R. Unwin,
75. S. Bruckenstein, D. T. Knapp, J. Am. Chem. Anal. Chem. 1994, 66, 2175.
Soc. 1968, 90, 6303. 91. K. Toda, S. Oguni, Y. Takamatsu et al., J.
76. T. Tsuru, T. Nishimura, S. Haruyama, Denki Electroanal. Chem. 1999, 479, 57.
Kagaku 1984, 52, 532. 92. N. V. Rees, J. A. Alden, R. A. W. Dryfe et al.,
77. T. Tsuru, T. Nishimura, K. Aoki et al., Denki J. Phys. Chem. 1995, 99, 14 813.
Kagaku 1982, 50, 712. 93. R. G. Compton, A. C. Fisher, R. G. Welling-
78. Q. Fulian, A. C. Fisher, D. J. Riley, Electro- ton et al., J. Phys. Chem. 1993, 97, 10 410.
analysis 2000, 12, 503. 94. J. A. Alden, S. Hakoura, R. G. Compton,
79. A. J. Bard, K. B. Prater, J. Electrochem. Soc. Anal. Chem. 1999, 71, 806.
1970, 117, 207.
80. S. W. Feldberg, J. Electroanal. Chem. 1969, 3,
199.
2.5 can be investigated in high resistance sol-

Microelectrodes – Retrospect and Prospect vents, in solids, in supercritical fluids, and
in gases. Moreover, they facilitate mea-
Robert J. Forster surements at short timescales and at low
Dublin City University, Dublin, Ireland temperatures and allow redox mapping
2.5.1 to be performed with extraordinary spa-
Background and Introduction tial resolution. Microscopic sensors and
arrays impact the development of areas
Microelectrodes, also commonly known as as diverse as the manipulation of indi-
ultramicroelectrodes, may be defined as vidual nanoparticles and the mapping of
electrodes whose critical dimension is in brain functions.
the micrometer range [1–5]. Strictly, all
2.5.2
electrode dimensions should be smaller Microelectrode Fabrication and
than the thickness of the diffusion layer Characterization
on the timescale of the electrochemical
experiment. Microelectrodes have several We define microelectrodes as miniature
attractive features, including small cur- electrodes in which the critical electrode
rents, steady state responses in quiescent dimension is of the order of 10 µm yet it
solutions, and short response times [6]. remains much greater than the thickness
Microelectrodes have been used since the of the electrical double layer, which is
1940s, for example, to measure oxygen typically 10 to 100 Å.
concentrations within tissue [7]. However,
it is only since the 1980s that the wide
2.5.2.1 Microelectrodes
availability of microscopic wires and fibers,
Figure 1 illustrates the five common mi-
as well as sensitive instrumentation, has
croelectrode geometries. The microdisc
allowed them to be used extensively. Tradi-
is the most popular geometry and is
tionally, electrochemistry suffered relative employed in approximately 50% of all in-
to spectroscopy because electrochemical vestigations. Other common geometries
measurements could be made only in include cylinders (20%) and arrays (20%),
highly conducting media such as aque- with the remaining 10% comprising bands
ous solutions containing a salt as sup- and rings and less frequently spheres,
porting electrolyte. This restriction arose hemispheres, and more unusual assem-
because resistance between the working blies. The most popular materials include
or sensing electrode and the reference platinum, carbon fibers, and gold, al-
electrode limited the precision with which though mercury, iridium, nickel, silver,
the applied potential could be controlled. and superconducting ceramics have also
However, the currents observed at micro- been used. Microdisc electrodes predomi-
electrodes typically lie in the picoamp to nate because of their ease of construction,
nanoamp range, which is several orders of and because the sensing surface of the
magnitude smaller than those observed electrode can be mechanically polished.
at conventional millimeter-dimensioned Microelectrodes in the form of discs,
macroelectrodes. These small electrolysis cylinders, and bands are commonly fab-
currents often completely eliminate ohmic ricated by sealing a fine wire or foil
effects. Thus, electrochemical processes into a nonconducting electrode body such
2.5 Microelectrodes – Retrospect and Prospect 161
Fig. 1 The most common

microelectrode geometries and their
associated diffusion fields.
Microdisk Microband
as glass. Spherical and hemispherical
microelectrodes are typically formed by
electrodepositing mercury onto platinum
or iridium microdiscs.
2.5.2.2 Nanodes
The term nanode has been introduced to
describe electrodes with at least one di-
mension in the nanometer size range [8]. Microring
Although their fabrication and charac- Microcylinder
terization remain challenging, these ul-
trasmall probes with critical dimensions
approaching those of individual molecules
are now an integral part of the field.
A procedure commonly used to fabricate
nanodes uses the electrochemical etch- 3-Element microdisk array
ing approaches originally developed for
the production of STM tips [9]. For exam-
ple, applying a 2 V root-mean-square AC Approaches based on micropipette
waveform to a platinum wire immersed in pullers allow the height-to-radius ratio of
a solution of saturated CaCl2 (60% v/v), the conical shaped nanode to be controlled
H2 O (36%), and HCl (4%) produces an over a much wider range than is accessible
atomically sharp tip [10]. These tips can using electrochemical sharpening. Wong
then be sealed in glass and the tip ex- and Xu [12] have described an elegant ap-
posed by careful polishing. However, a proach to the fabrication of carbon disk
disadvantage of this approach is that the nanodes using a laser-based micropipette
height-to-radius ratio of the tip cannot be puller. First, quartz capillaries are pulled
varied over a wide range causing the elec- using a laser-based micropipette puller to
trode radius to increase significantly as one form a small tip. A carbon deposit is then
polishes down through the cone. formed by pyrolyzing methane, and finally
Unwin and coworkers [11] have devel- electrical contact is made with the disk us-
oped a straightforward approach for fabri- ing mercury. The carbon nanodes exhibit a
cating platinum microelectrodes with tip well-defined response for dopamine with a
sizes as small as a few nanometers. Their low background charging current. Mirkin,
innovation is to use an electrophoretic Lewis and coworkers [13] have also em-
paint to insulate the electrode walls while ployed a laser-based micropipette puller
leaving an electrochemically active tip ex- but to fabricate platinum nanodes. In their
posed. These electrodes have excellent approach, the laser simultaneously heats
properties for use in scanning electro- both the glass electrode body and the plat-
chemical microscopy (SECM), discussed inum microwire while it is mechanically
further in Chapter 3.3. stretched to a nanometer-dimensioned tip.
The tip is exposed either by chemical etch- because the double-layer charging current
ing or micropolishing and ranges from depends simply on the area exposed to the
2 to 500 nm in radius. The voltammetric electrolytic solution. In contrast, at long
performance of these nanodes is excellent, experimental timescales, the diffusion lay-
with close to theoretical resistances and ers associated with the Faradaic reaction
interfacial capacitances being observed. coalesce and the total Faradaic current is
Carbon fiber nanodes have been fabri- given by the area of the microelectrodes
cated by first etching a single fiber using an plus the insulator. In this way, the contri-
argon ion beam thinner [14]. Then a single bution from capacitance to the total current
fiber with a tip diameter of approximately is reduced compared to a uniformly acces-
50 nm is inserted into a glass capillary and sible surface.
sealed by heating under a vacuum. The The first approach to fabricating an ar-
analytical performance of these electrodes ray of coupled microelectrodes consists of
is excellent with wide dynamic ranges and sealing a bundle of individually conduct-
low limits of detection being observed for ing microwires or fibers in an insulating
dopamine and 5-hydroxytryptamine. material and making a single electrical
Although the voltammetric performance connection to the assembly. Alternatively,
of some nanodes is excellent, it is impor- a conventional macroelectrode may be
tant to note that when the thicknesses modified with an insulating layer that has
of the depletion and double layers become microscopic faults through which analytes
comparable, classical theories are not likely can diffuse to the electrode surface. Elec-
to be applicable [15]. trodes of this kind are known as random-
array microelectrodes or RAM electrodes.
2.5.2.3 Microelectrode Arrays Although both approaches give robust de-
A microelectrode array consists of a se- vices, they suffer from the problem of
ries of microelectrodes separated by an ill-defined spacing between the individual
insulating material. The small size of mi- elements. This irregular spacing causes
croelectrodes facilitates the use of arrays, the diffusion layers to overlap at different
because several elements can fit into a times making correlations with theory dif-
small surface area. Within these arrays, ficult. However, Amatore and coworkers
individual electrodes may all be connected developed a theory to allow the size and in-
in parallel, that is, coupled electrodes, or tersite separation of these microelectrodes
they may be independently addressable. to be estimated from transient electro-
chemical data. Self-assembled monolayers
2.5.2.3.1 Arrays of Coupled Electrodes have been used to create RAM elec-
Two significant objectives in using mi- trodes, for example, Crooks and Baker [16]
croarrays are to increase the magnitude formed arrays of nanometer scale elec-
of the observed current and to improve trodes by cyanide etching hexadecanethiol
the ratio of the Faradaic to charging cur- monolayers confined to Au(111) that was
rents. Arrays of microelectrodes achieve modified by underpotential deposition of
the first objective simply by increasing copper. The electrode radii ranged from
the area at which Faradaic reactions can 6 to 80 nm. While the theoretical limiting
occur compared to a single microelectrode. current correlated with the experimental
Microarrays achieve the second objective value, there was significant scatter in the
data due to the wide distribution of elec- using anisotropically etched silicon mas-
trode size and pitch. Other approaches ters [21]. Arrays of interdigitated electrodes
to partially blocking an electrode surface with interelectrode spacings as small as
include membrane-covered electrodes in 3 nm can be fabricated using a shadow
which the pores are filled with carbon paste evaporation technique [22].
or other conducting material using me-
chanical or electrochemical approaches.
2.5.2.3.2 Independently Addressable Arrays
There is considerable interest in us-
Microfabrication technology, which takes
ing microlithography, microcontact print-
ing, and micromachining to fabricate advantage of the thin- and thick-film tech-
arrays. For example, Whitesides and niques developed for the integrated circuit
coworkers [17] described a laser direct (IC) industry, allows the microelectrode
write method for generating micrometer- size, shape, material, and pitch to be
dimensioned patterns for microanalytical precisely controlled [23, 24]. Moreover, it
applications. An elastomeric stamp was allows a separate electrical contact to be
used to fabricate a microelectrode array made to each microelectrode so that it can
by printing on a silicon wafer. Other be addressed independently. Table 1 de-
approaches to fabricating well-defined cou- tails the main microfabrication techniques
pled arrays include the use of the atomic that are commonly used for microelec-
force microscope [18] or electron beam trode and microsensor fabrication [25]. As
lithography [19] to pattern spontaneously illustrated in Fig. 2, photolithography and
adsorbed or self-assembled monolayers chemical etching or liftoff are indispens-
followed by electrodeposition. For exam- able for this purpose [26]. A key distin-
ple, microelectrode arrays with electrode guishing feature of microelectrode arrays
widths and spacings of 500 and 250 nm compared to conventional integrated cir-
have been created using deep UV pho- cuits is the need to package the device
tolithography [20]. Comparable resolution in a way that protects it from the sample
can be obtained using hot embossing and the chemical environment [27]. The
lithography or nanoimprinting with liftoff production of packaging approaches that
Tab. 1 Microfabrication techniques that can be used to produce microelectrode arrays and
microchemical sensors and biosensors
Component Techniques or materials
Electrode pattern Photolithography etching, liftoff, metal masking, screen printing

Insulation Chemical vapor deposition, hydrophobic polymer
Electrolyte layer pattern Screen printing, photosensitive polymer
Container Anisotropic etching, field-assisted bonding, photosensitive polymer
Flow cell Anisotropic etching, isotropic etching, field-assisted bonding,
photosensitive polymer, dicing-saw
Liquid junction Anisotropic etching, porous silicon, hydrophilic polymer
Immobilization of Liftoff, ink-jet printing, screen printing
enzyme
Needle structure Anisotropic etching, etch-stop
Thin film Thin film

Photoresist
Photoresist
(a) (b)
Evaporated metal
Former Mask
Thin film
Paste
Mask
(c) (d)
Fig. 2 Schematic representation of the approaches used to fabricate
microelectrode arrays and sensors. (a) Etching; (b) liftoff; (c) metal
masking; and (d) screen printing.
yield highly stable devices with long life- typically six orders of magnitude smaller
times, but avoid problems with sample than those observed at macroelectrodes.
contamination because of leaching from These small currents often completely
the encapsulant, remains a technological eliminate ohmic drop effects even when
challenge [28]. working in organic solvents. For example,
the steady state current observed at a 5-µm-
2.5.3 radius microdisc is approximately 2 nA for
Properties of Microelectrodes a 1.0 mM solution of ferrocene. Taking a
reasonable value of 0.01 −1 cm−1 as the
2.5.3.1 Reduced Ohmic Effects specific conductivity, then Eq. (1) indicates
When Faradaic and charging currents that the resistance will be of the order
flow through a solution, they generate a of 50 000 . This analysis suggests that
potential that acts to diminish the applied the iR drop in this organic solvent is
potential by an amount iR, where i is the a negligible 0.09 mV. In contrast, for a
total current and R is the cell resistance. conventional macroelectrode the iR drop
This is an undesirable process that leads would be of the order of 5 to 10 mV. Under
to distorted voltammetric responses. these circumstances, distorted current
As discussed in detail by Newman [29], responses and shifted peak potentials
the solution resistance for a disc-shaped ul- would be observed in cyclic voltammetry.
tramicroelectrode is inversely proportional It is useful to investigate the effect of
to the electrode radius, an experimental timescale on the iR drop
observed at microelectrodes. In a subse-
1
R= (1) quent section, we discuss in more detail
4κr
how the diffusion field at microelectrodes
where κ is the conductivity of the solution depends on the characteristic time of the
and r is the radius of the microdisc. experiment. At short times, the dominant
Equation (1) shows that R increases as the mass transport mechanism is planar diffu-
electrode radius decreases. However, the sion and the microelectrode behaves like a
currents observed at microelectrodes are macroelectrode. Therefore, at short times,
the current i decreases with decreasing lower timescale limit that is imposed by
electrode area (∝ r 2 ). Since the resistance the RC cell time constant, that is, the prod-
increases with decreasing electrode radius uct of the solution resistance, R, and the
rather than area, the product iR decreases double-layer capacitance, C, of the work-
with decreasing electrode radius in short ing electrode. Meaningful electrochemical
timescale experiments. In contrast, at long data can only be extracted at timescales
experimental timescales, the Faradaic cur- that are typically five to ten times longer
rent depends directly of the radius making than the RC time constant [30]. Therefore,
the product iR independent of the elec- an important objective when seeking to
trode radius. make high-speed-transient measurements
is to minimize the cell time constant.
2.5.3.2 Interfacial Capacitance The existence of the double-layer capaci-
Altering the potential that is applied to tance at the working electrode complicates
an electrode causes the charge on the electrochemical measurements at short
metal side of the interface to change timescales. Specifically, the double-layer
and some reorganization of the ions and capacitance must be charged through the
solvent dipoles in the double layer on solution resistance in order to change the
the solution side of the interface will potential across the Faradaic impedance,
occur. This process causes electrons to and this process cannot be achieved in-
flow into or out of the surface, giving stantaneously.
rise to a charging or capacitive response. The time constant for this charging
The double-layer capacitance for a disc- process is given by Eq. (3),
shaped ultramicroelectrode is proportional πrCo
to the area of the electrode surface and is RC = (3)
4κ
given by:
C = πr 2 Co (2) and is typically hundreds of microseconds
for a conventional millimeter-sized elec-
where Co is the specific double-layer capac- trode placing a lower limit on the useful
itance of the electrode. Thus, shrinking the timescale of the order of several millisec-
size of the electrode causes the interfacial onds. In contrast, as illustrated in Fig. 3,
capacitance to decrease with decreasing r 2 . the cell time constant decreases linearly
These low capacitive currents are particu- from approximately 2 µs to 80 ns as the
larly important for analytical applications electrode radius is reduced from 25 to
in which the ability to discriminate a 1 µm. The RC time constant can also
Faradaic signal above background charg- be decreased by making measurements
ing often dictates the limit of detection that in highly conducting solutions, for exam-
can be achieved. ple, response times as short as 5 ns can
be achieved for an ideal 5-µm-radius mi-
2.5.3.3 Electrode Response Times croelectrode by making measurements in
Beyond chemical analysis, a major applica- highly concentrated acidic solutions [31].
tion of microelectrodes is the investigation However, there are a number of practical
of chemical reactivity. Therefore, it is im- problems associated with the design and
portant to consider the effect of shrinking fabrication of microelectrodes that cause
the electrode size on its response time. them to have RC time constants that
Every electrochemical measurement has a greatly exceed those predicted by Eq. (3).
2500 Fig. 3 Relationship between the RC cell

y = 80.139x + 4.3159 time constant and the radius of
R 2 = 0.9991 platinum microdiscs in which the
2000
RC time constant
supporting electrolyte is 0.1 M HCl. Cell

time constants were measured using
1500
chronoamperometry conducted on a
[ns]
microsecond to submicrosecond
1000 timescale by stepping the potential from
0.200 to 0.250 V versus Ag/AgCl.
500
0
0 10 20 30
Radius
[µm]
An important cause of nonideal re- aqueous electrolyte solution, the double-

sponses is stray capacitance within the layer capacitance of a 1-mm-radius disc will
electrochemical system that may arise be approximately 1 µF. This value is very
from the electrode itself, the leads, or elec- significantly larger than the pF stray capac-
trical connections. Stray capacitance will itance found in a typical electrochemical
increase the cell time constant as described experiment. However, for a 1-µm-radius
by Eq. (4). microdisc, the interfacial capacitance will
decrease by six orders of magnitude to
1
RC = (πr 2 Co + CStray ) (4) approximately 1 pF. Therefore, stray ca-
4κr pacitance of even a few pF will cause
where CStray is the stray capacitance. Al- the observed RC time constant to increase
though it depends on the microelectrode significantly beyond the minimum value
design and the experimental setup, this dictated by double-layer charging alone.
stray capacitance is typically between a Thus, in seeking to implement ultrafast
few pF and several tens of pF. The cell transient techniques, it is vital to minimize
time constant observed in these circum- the stray capacitance.
stances depends strongly on the relative There are two major sources of stray ca-
magnitudes of the double-layer and stray pacitance. First, the capacitance of the po-
capacitances. At a normal size electrode, tentiostat and leads. By using high-quality
the stray capacitance is negligible com- cable of minimum length, for example,
pared to the double-layer capacitance, and by mounting the current-to-voltage con-
therefore, does not significantly affect the verter directly over the electrochemical cell,
observed cell time constant. However, and by avoiding the use of switches as
when the magnitude of the double-layer ca- far as possible, stray capacitance from the
pacitance is reduced by shrinking the size electrochemical system can be minimized.
of the electrode to micron and submicron Second, the microelectrode itself. For ex-
dimensions, the stray and double-layer ca- ample, if there is a small imperfection in
pacitances can become comparable. For ex- the seal between the insulator and the elec-
ample, taking a typical value of 40 µF cm−2 trode material, then solution leakage will
as the specific double-layer capacitance for cause the RC cell time constant to increase
a platinum electrode in contact with 1 M massively and the Faradaic response may
2
become obscured by charging/discharging ∂C(r, t) ∂ C(r, t) 2 ∂C(r, t)
processes. Moreover, as shown by Wight- =D +
∂t ∂r 2 r ∂r
man [32], as well as Faulkner and cowork- (5)
ers [33], using silver epoxy or mercury to The boundary conditions for the potential
make the electrical connection between step experiments described above are
the microwire and a larger hook-up wire
can cause the RC cell time constant to lim C(r, t) = C ∞
r→∞
increase dramatically. This increase arises C(r, 0) = C ∞
because the electronically conducting mer-
cury/glass insulator/ionically conducting C(r, t) = 0 for t > 0
solution junctions cause significant stray
where r is the distance from the center of
capacitance. It is important to note that
the sphere, D is the diffusion coefficient
these effects may only become apparent in
for the redox-active species, and C is the
high-frequency measurements.
concentration as a function of distance r
and time t.
2.5.4
Equation (5) can be solved using Laplace
Mass Transport
transform techniques to give the time
evolution of the current, i(t), subject to the
Oxidation or reduction of a redox-active
boundary conditions described resulting
species at an electrode surface generates in Eq. (6),
a concentration gradient between the in-
terface and the bulk solution. This redox nFADC ∞ nFAD 1/2 C ∞
process requires electron transfer across i(t) = + (6)
rs π 1/2 t 1/2
the electrode–solution interface. The rate
at which electron transfer takes place where n is the number of electrons
across the interface is described by the transferred in the redox reaction, F is
heterogeneous electron transfer rate con- Faraday’s constant, and A is the geometric
stant, k. If this rate constant is large, electrode area.
Equation (6) shows that the current
then diffusional mass transport will con-
response following a potential step con-
trol the current observed. Our objective
tains both time-independent and time-
is to describe how the diffusion field
dependent terms. The differences in the
evolves in time. The experiment of inter-
electrochemical responses observed at
est involves stepping the potential from
macroscopic and microscopic electrodes
an initial value at which no electrode
arise because of the relative importance
reaction occurs, to one at which elec-
of these terms at conventional electro-
trolysis proceeds at a diffusion-controlled chemical timescales. It is possible to
rate. We consider the case of a spher- distinguish two limiting regimes depend-
ical electrode of radius rs placed in a ing on whether the experimental timescale
solution that contains only supporting is short or long.
electrolyte and a redox-active species of
concentration C. The concentration gra- 1. Short times. At sufficiently short times,
dient at the electrode surface is obtained the thickness of the diffusion layer
by solving Fick’s second law in spherical that is depleted of reactant is much
coordinates. smaller than the electrode radius and
Planar diffusion
field
Spherical
diffusion
field
Electrode Insulator Electrode Insulator

(a) (b)
Fig. 4 Diffusion fields observed at microelectrodes. (a) Linear
diffusion observed at short times and (b) radial (convergent)
diffusion observed at long times.
the spherical electrode appears to be larger than the first and the current
planar to a molecule at the edge response induced by the potential step
of this diffusion layer. Under these initially decays in time according to the
conditions, the electrode behaves like Cottrell equation.
a macroelectrode and mass transport
is dominated by linear diffusion to nFAD 1/2 C ∞
the electrode surface as illustrated i(t) = (7)
π 1/2 t 1/2
in Fig. 4(a). At these short times,
the t −1/2 dependence of the second Figure 5 shows the relationship that
term in Eq. (6) makes it significantly exists between the range of useable scan
2
log (ro)
0 Fig. 5 Theoretical limitations on

[µm]
ultrafast cyclic voltammetry. The shaded

−2 area between the slanted lines
represents the radius that a microdisc
−4 must have if the ohmic drop is to be less
than 15 mV and distortions due to
3 4 5 6 7 8
nonplanar diffusion account for less
log (v) than 10% of the peak current.
(a) [V s−1] (a) Without iR drop compensation by
positive feedback and (b) with 90% and
99% ohmic drop compensation. The
2 9.9% dotted areas in (a) and (b) represent the
9.0% regions in which transport within the
log (ro)
0 double layer affects the voltammetric

[µm]
response. Limits are indicative and

−2 correspond approximately to a 5 mM
anthracene solution in acetonitrile,
−4 0.3 M tetrafluoroborate as supporting
3 4 5 6 7 8 electrolyte. (Reproduced by permission
of Marcel Dekker from C. Amatore,
log (v) Electrochemistry at Microelectrodes, Ed.
(b) [V s−1] I. Rubenstein, Chap. 4, 1995.)
rates and electrode radius subject to the radius, the experimental timescale must
condition that ohmic drop is negligible be longer than 80 s. Therefore, steady state
and that the dominant mass transport is not observed for macroelectrodes at the
regime is linear diffusion [34]. tens of mV s−1 timescale typical of con-
2. Long times. At long times, the transient ventional cyclic voltammetry experiments.
contribution given by the second term However, reducing the electrode radius by
of Eq. (6) has decayed to the point at a factor of a thousand to 5 µm, means
which its contribution to the overall that a steady state response can be ob-
current is negligible. At these long served for times longer than 80 µs. Since
times, the spherical character of the the steady state current becomes more
electrode becomes important and the dominant with increasing time, steady
mass transport process is dominated by state responses are easily observed for
radial (spherical) diffusion as illustrated microelectrodes in electrochemical exper-
in Fig. 4(b). iments run at conventional timescales.
Figure 6(a) shows the sigmoidal-shaped re-
The current attains a time-independent sponses that characterize steady state mass
steady state value given by Eq. (8). transfer in slow scan-rate cyclic voltamme-
nFADC ∞ try. In contrast, as illustrated in Fig. 6(b), at
iss = (8) short experimental timescales (high scan
rs
rates), peaked responses similar to those
The steady state response arises because observed at conventional macroelectrodes
the electrolysis rate is equal to the are seen. It is important to emphasize
rate at which molecules diffuse to the that there is no theoretical barrier to ob-
electrode surface. serving steady state behavior for any size
Since short and long times are relative of electrode. However, for conventional
terms, it is useful to determine the millimeter-dimensioned macroelectrodes,
times over which transient and steady the time taken is too long to be experimen-
state behaviors will predominate and tally practical. Moreover, unintentional
how this time regime is affected by the convection, for example, caused by build-
electrode radius. ing vibration, will play an important role
This objective can be achieved by con- in the mass transport process at these
sidering the ratio of the transient to steady long timescales.
state current contributions (Eqs. 7 and 8, The preceding analysis considered a
respectively). This analysis gives a dimen- spherical electrode because its surface is
sionless parameter (πDt)1/2 /rs that can uniformly accessible, and a simple closed-
be used to calculate a lower time limit at form solution to the diffusion equation
which the steady state contribution will exists [35]. The microdisc is the most
dominate the total current to a specified widely used geometry, but the deriva-
extent. For example, the time required for tion of rigorous expressions describing
the steady state current contribution, iss , their experimental response is complicated
to be ten times larger than the transient because the surface is not uniformly ac-
component, it , can be calculated. Taking cessible. For discs, electrolysis at the outer
a typical value of D as 1 × 10−5 cm2 s−1 circumference of the disc diminishes the
for a redox couple dissolved in an aqueous flux of the electroactive material to the cen-
solution, then for an electrode of 5-mm ter of the electrode. However, microdisc
(a) (b)
i red
i ox
4 nA 4 nA
0.5 0 0.5 0
E vs Ag/AgClO4
[V]
Fig. 6 Effect of scan rate on the cyclic voltammetry of 1.0 mM
ferrocene at a 6.5-µm gold microdisc where the supporting
electrolyte is 0.1 M tetrabutyl ammonium perchlorate in acetonitrile.
(a) Scan rate is 0.1 V s−1 and (b) scan rate is 10 V s−1 . (Reproduced
with the permission of the American Chemical Society from
J. O. Howell and R. M. Wightman, Anal. Chem. 1984, 56, 524.)
and microring geometries share the advan- 2.5.4.1 Instrumental Challenges

tage of spherical microelectrodes in that Taking typical values of 1 mM and
quasi-spherical diffusion fields are estab- 1 × 10−5 cm2 s−1 for the concentration
lished in relatively short periods of time. and diffusion coefficient, respectively, the
The steady state current is given by, steady state current is of the order of
2 nA for a 5-µm-radius microdisc. As de-
iss = γ nF DCr (9) scribed by Eq. (9), the magnitude of this
current will decrease with decreasing elec-
where γ is 4 and 2π for disk and hemi-
trode radius and picoamp currents will
spherical shaped electrodes, respectively.
be observed for a 10-nm-radius electrode.
Observing a steady state response de-
pends on all the electrode dimensions These are certainly small currents but
being small, not just the radius, and is can be measured with relatively simple
therefore not achieved for every geome- circuits based on operational amplifiers.
try at the timescales considered above. For This success can only be achieved be-
example, band electrodes whose thickness cause the measurement is performed at
is in the micrometer range, but whose long timescales. The ability to measure nA
length is several millimeters, do not ex- to pA currents at nanosecond timescales
hibit true steady state responses. However, remains a challenge.
a high analyte flux to the ends of the
band often makes it possible to observe 2.5.5
a pseudo–steady state condition in a prac- Applications
tical sense. Radial diffusion gives very high
rates of mass transport to the electrode sur- 2.5.5.1 Electroanalysis
face with a mass transport coefficient of the The advantages of microelectrodes are
order of D/r. Therefore, even at rotation exploited in many different areas of elec-
rates of 104 rpm, convective transport to troanalysis, with the environmental, food
a rotating macroelectrode is smaller than quality assurance, and biomedical appli-
diffusion to a 1-µm microdisc. cations being the most active. Portability
represents one of the key advantages of potential at the other. This approach gives
electrochemistry over spectroscopy, espe- enhanced performance under steady state
cially for on-site analysis. conditions with improved baselines due to
Bond and coworkers [36] have probed reduced charging currents. Square-wave
the ability of microelectrodes to determine ASV has already been conducted using
low concentrations of electroactive species microfabricated mercury microelectrode
using flow injection analysis. Ferrocene arrays for the detection of Cd2+ , Pb2+ ,
was chosen as a test system to avoid any Cu2+ , and Zn2+ [40, 41].
complications associated with irreversible When dealing with complex matrices
reactions. Measuring concentrations of the such as those encountered in food and
order of 10 nM proved challenging and medical samples, it is often difficult to de-
required the use of a battery operated two- velop a sensor that exhibits a sufficiently
electrode potentiostat because of 50-Hz selective response for the target analyte.
noise coming from the mains power Under these circumstances, arrays and
supply. Bond has also shown that it mathematical modeling, for example, prin-
may be easier to realize low limits cipal components analysis, artificial neural
of detection using macro- rather than networks, or other pattern recognition ap-
microelectrodes [37]. For example, the proaches, may be required [42].
electrochemical detection of As(III) at a Recently, a number of papers have ap-
platinum electrode in an HPLC system peared describing novel approaches to the
becomes less favorable as the electrode detection of analytes based on principles
radius decreases. Thus, while 10 nM traditionally associated with biological sys-
As(III) could be detected at a 50-µm-radius tems or semiconductors. For example,
microelectrode, the limit of detection Amatore and coworkers [43] described as-
increased to 500 nM when a 2.5-µm- semblies of paired microband electrodes
radius electrode was used. This falloff in that behave like neuronal synapses. The
performance appears to arise because of generator electrode mimics a synaptic ter-
imperfect seals and high stray capacitance minal while the collector functions as a
for the smaller electrodes. postsynaptic membrane. These artificial
Electrochemical-stripping techniques synapses can be designed in several config-
promise not only portability but also urations allowing Boolean functions such
high sensitivity with disposable one- as AND or OR operations to be performed.
shot systems, as already highlighted Bartlett and coworkers have reported a
in Chapter 2.3. Wang and Tian [38] molecular electronic device that is based on
have used screen-printed electrodes direct electrical communication between
for stripping measurements of trace an enzyme (horseradish Peroxidase, HRP)
mercury. This research group has and a conducting polymer [44]. The sys-
also demonstrated the significant tem uses two carbon microband electrodes
benefits of performing anodic stripping separated by a 20-µm gap that is bridged
voltammetry (ASV) of trace metals at by poly(aniline) coated with an enzyme-
lithographically fabricated pyrolitic carbon containing layer. In the presence of hy-
interdigitated microelectrode arrays [39]. drogen peroxide, the polymer is driven
Their innovation was to collect the into its nonconducting state. This analyte-
soluble metal ions generated at one set induced switching from conducting to in-
of electrodes by applying a reducing sulating states is reminiscent of transistor
operation and represents a novel molecular by the two electrodes, it is repeatedly

electronics approach to chemical sensing. electrolyzed so as to generate a measur-
able current. Transport of the electroactive
2.5.5.2 Small Sample Volumes species is dictated by Brownian motion
Microelectrodes open up the possibility of causing the events to be infrequent.
probing redox processes in small sample Rather than rely on controlling the ex-
volumes or physically small spaces. For perimental timescale to probe only a tiny
example, the redox properties of sample volume, Wightman and coworkers have
volumes as small as a few picoliters have demonstrated that microcapillaries can be
used to create vials with a volume of
been interrogated. Experiments of this
100 pL [47]. Evaporation of the sample
kind are possible because, as described
can be a particular problem when work-
by Eq. (10), for solution phase reactants,
ing with such small volumes, but it can
the depletion layer thickness, δ, depends
be minimized by placing a drop of min-
on the experimental timescale.
eral oil on top of the aqueous solution.
δ = (πDt)1/2 (10) Another difficulty with measurements in
ultrasmall volumes is depletion of the reac-
For a 5-µm-radius microelectrode and a tants causing the responses to deviate from
1-µs electrolysis time, the volume that is those found in conventional bulk experi-
depleted of reactant will be less than 5 pL! ments [48]. Recently, lithography [49–51]
Wightman and coworkers have exploited and UV laser photoablation [52] have been
this method to probe electrochemilumi- used to create vials with picoliter vol-
nescent reactions involving individual re- umes. For example, as illustrated in Fig. 7,
actant pairs [45]. Their approach is to use Ewing and coworkers [53] have used a
50-µs potential steps at a 5-µm-radius mi- template produced using lithography to
croelectrode to electrolyze a few femtoliters imprint microvials with volumes between
of a 9,10-diphenylanthracene solution so as 300 and 0.4 pL. Steady state voltammo-
to generate a small population of radical grams for ferrocenecarboxylic acid show
anions. By allowing these radical anions to close to ideal behavior even for vials with
diffuse into a solution containing the rad- 1-pL volumes. Microvials of this kind are
ical cation of the same molecule, singlet- particularly important for biological appli-
excited states of 9,10-diphenylanthracene cations since they facilitate the detection of
were created. This electronically excited redox-active messengers from single cells
state then decays back to the ground state without significant dilution.
by emission. Because of the low concen-
trations involved, these reactions are seen 2.5.5.3 Biosystems
as individual light-producing events. Since the pioneering work of Davies and
Bard and Fan developed an approach Brink [7] in 1942 that measured the con-
to allow single redox-active molecules to centration of oxygen in animal muscle,
be detected by using a piezoelectric posi- microelectrodes have been instrumental in
tioner to move a microelectrode toward a providing information about the concen-
large counterelectrode so as to create a mi- tration and temporal release of redox-active
crovolume electrochemical cell [46]. When biomolecules. This research is becoming
a redox-active molecule moves from bulk more important in the light of evidence
solution into the microvolume defined that not only is the absolute concentration
Fig. 7 Scanning electron micrographs

of polystyrene microvials: (a) 75; (b) 20;
and (c) 0.4 pL. The microvials were
sputtered with Pd/Au prior to imaging.
(Reproduced with the permission of the
American Chemical Society from
R. A. Clark, P. B. Hietpas, A. G. Ewing,
Anal. Chem. 1997, 69, 259.)
(a) 10 µm
of a chemical messenger important in
dictating a cellular response, so too is
the time profile (frequency) of the output.
The chemical events of interest are of-
ten restricted to the interior or exterior
surfaces of single cells. Therefore, to pro-
vide useful information about in vivo
biochemistry, these measurements must
(b) 10 µm
be performed with a high degree of spatial
and temporal resolution, as well as a high
degree of sensitivity and selectivity [54].
Spectroscopic techniques such as fluo-
rescence microscopy, magnetic resonance
imaging, ion mass spectrometry, and even
X-ray emission imaging can provide use-
ful information about the two-dimensional
structure of biosystems. However, they (c) 5 µm
are typically restricted to high analyte
concentrations, millimeter rather than
micrometer resolution, and slow (> ms) 2.5.5.3.1 Spatial Sampling and Timescale
time responses. The first challenge is to ensure that
The mammalian brain has been the fo- measurements of the neurotransmitter
cus of a significant research effort over concentration are sufficiently spatially lo-
the last 25 years and represents an ex- calized to provide a meaningful insight
traordinarily challenging environment in into the brain’s structure. Of particular
which to perform analytical chemistry. importance is the volume of tissue that
At every level of organization, the brain is sampled, since this dictates the size of
is temporally and spatially heterogeneous structure that may be examined. For exam-
with neuronal structures of differing sizes ple, a microdialysis loop combining two
(from nanometer to millimeter) commu- 0.4-mm stainless steel cannulae cannot
nicating with each other at timescales provide information on a micron length
ranging from the microseconds to hours scale. In contrast, Eq. (10) indicates that by
or days. When the objective is to eluci- using microelectrodes and experimental
date the structure-function relationship of timescales less than 100 µs, the diffusion
these assemblies, the ability to make spa- layer thickness will be less than 1 µm and
tially resolved measurements across a wide a high degree of spatial resolution can be
range of timescales is paramount [55, 56]. achieved. For example, Lu and Gratzl [57]
investigated the drug resistance of indi- high, but the total number of molecules
vidual cancer cells by voltammetrically released will be very small. In this regard,
monitoring the efflux of the anticancer the high mass rather than concentration
drug, doxorubicin, from the cells. This ap- sensitivity of electrochemical techniques,
proach promises to provide information and the ability to routinely measure small
about the heterogeneity of individual can- currents, ≤1 pA, is important.
cer cell resistance on a length scale not
previously attainable. 2.5.5.4 Neuronal Networks for Drug
Wightman and coworkers, have demon- Evaluations
strated that the distance between the Beyond in vivo studies of fundamen-
microelectrode and the source cell can tal biochemistry, microelectrode arrays
dramatically affect the nature of the micro- have been used as platforms for growing
electrode response [58–60]. As illustrated active neuronal networks [62–65]. These
in Fig. 8, driving the electrode toward cultures continue to grow and remain ac-
the cell causes the amplitude of the cur- tive for extended periods, exceeding three
rent spikes associated with catecholamines months under some conditions. These in
release to increase in amplitude and vitro networks display complex spatiotem-
become narrower. This behavior arises poral spike and burst patterns that are
because the neurotransmitter rapidly dif- highly sensitive to their chemical environ-
fuses in the extracellular medium giving ment. For example, Gross and cowork-
rise to smaller, broader peaks when the ers [66] have grown networks of embryonic
microelectrode is far from the release spinal cord and auditory cortex tissues on
site. 64 element arrays and performed quan-
titative investigations into the effects of
2.5.5.3.2 Sensitivity A second key issue cannabinoid mimetics, for example, anan-
in bioelectrochemistry is sensitivity [61]. damide and methanandamide, on cell ac-
In the case of localized release, for ex- tivity. These investigations reveal that the
ample, neurotransmitter release through agonists generate specific, concentration-
exocytosis, the local concentration may be dependent spontaneous activity that can
(a) (d)
Fig. 8 Amperometric detection at

isolated bovine adrenal medullary cells
detected at glass encased (a–c) and
etched (d–f) carbon fiber electrodes at 1
(b) (e) (A, D), 5 (B, E) and 10 µm (C, F).
Measurements at each position were
20 pA made simultaneously with the large and
2.5 sec small electrode. Release of
catecholamines was induced by a 3-s,
(c) (f) 100-µM nicotine exposure applied at
1.25 s. (Reproduced with the permission
of the American Chemical Society from
K. T. Kawagoe, J. A. Jankowski,
R. M. Wightman, Anal. Chem. 1991,
63, 1589.)
be reversibly turned on and off. The Moreover, the ability to deposit a coat-
intra- and interculture response is highly ing on a microstructure without the need
reproducible suggesting that the cul- for precise positioning equipment, for ex-
tures provide reliable quantitative infor- ample, drop-on-demand technology, makes
mation about drug interactions with neu- electropolymerization especially attractive.
ronal tissues. Resistance is often the measurement of
While these investigations are likely choice, since reference and counterelec-
to play pivotal roles in identifying new trodes are not required and the measure-
chemical treatments for neuroconditions ment can be performed in both solution
ranging from Parkinson’s disease to and gas phases. Typically, two adjacent
schizophrenia and depression, microelec- microelectrodes are connected and poly-
trodes will continue to play important mer deposited from the solution across
roles in invasive surgery. For example, the gap separating the two elements. For
ablative procedures, deep brain stimu- analysis, the resistance between adjacent
lation, and cell transplantation are all
electrodes is measured as the device is
emerging treatments for Parkinson’s dis-
exposed to the gases of interest. Because
ease. Microelectrodes, especially devices
the responses are typically sparingly selec-
in which the conducting wire or fiber is
tive rather than specific, each electrode or
coated with a microscopically thin insulat-
collection of electrodes within the array is
ing layer, provide important information
functionalized in a slightly different way,
about the physical and chemical structure
of the brain making surgical intervention for example, chemically distinct polymers,
in the basal ganglia considerably more pre- or simply different morphologies induced
cise [67]. by the deposition rate.
This technology has spawned at least
five commercial instruments: the AromaS-
2.5.5.5 Artificial Noses
can (Alphatech International), the NOSE
Another area of substantial interest is the
(Neotronics Olfactory Sensing Equipment;
development of the electronic nose [68,
Neotronics), the Bloodhound (University
69]. Microarrays modified with conducting
of Leeds Innovations Ltd.), and the Elec-
polymers coupled with pattern recogni-
tronic Nose (Nordic Sensor Technologies).
tion or neural networks, represent a useful
The MOSE-nose (Metal Oxide Sensor Elec-
approach to analyzing or at least char-
acterizing gas mixtures [70]. The most tronic nose) system (Alpha) is based on
popular detection mechanism is based on metal oxide sensors. The principal tar-
the changes in electronic resistance when get of these arrays is quality control in
the conducting polymers are exposed to the food and beverage industries (no-
different volatile compounds. The sensi- tably, cheeses and wines), in which the
tivity to different volatiles can be altered by subjectivity of the human nose can some-
changing the counterion incorporated into times create problems with traditional
the polymer at the time of synthesis or by analytical techniques. However, medical
adding functional groups to the polymer applications are emerging in which body
backbone. The ease with which the physic- odors are used for the diagnosis of vari-
ochemical properties of these materials ous diseases and infections. The ability of
can be changed contrasts with conven- these devices to accurately mimic the hu-
tional pellistor or metal oxide technologies. man nose has been discussed in Ref. [71]
along with artificial intelligence approaches 2.5.5.6.2 Voltammetry in the Absence of

to data mining [72–74]. Deliberately Added Electrolyte Voltamme-
try in highly pure water without added
2.5.5.6 Low Conductivity Media electrolyte is now well established [80–83].
The ability to make electrochemical mea- The ability to perform stripping analysis
of metals without the need to add poten-
surements in low ionic strength media
tially contaminating supporting electrolyte
allows analyses to be performed in unusual
is a particularly important application [80].
media, for example, solids, frozen solu-
As illustrated in Fig. 9, Ciszkowska and
tions, low ionic strength media, and so on,
Osteryoung [84] have probed the voltam-
and facilitates comparison of spectroscopic
metric reduction of three metal cations:
and electrochemical information [75].
thallium, cadmium, and lead, in solu-
tions of various salts {TlNO3 , Tl2 SO4 ,
2.5.5.6.1 Theoretical Descriptions The Cd(NO3 )2 , Pb(NO3 )2 , and CdSO4 }, con-
first successful attempt to predict the taining either no supporting electrolyte
current for a redox reaction in the ab- or where LiClO4 or Ca(NO3 )2 were
sence of a supporting electrolyte was added at various concentrations. The in-
presented by Amatore and coworkers [76, fluence of migration was studied under
77]. Oldham also produced seminal work a wide range of concentrations of the
in this area and presented a rigorous treat- electroactive species and supporting elec-
ment, including an analysis of the ohmic trolyte [77, 85].
drop [78]. These mathematical treatments
reveal that the total solute concentration 2.5.5.6.3 Voltammetry in Pure Liquid Org-
is uniform throughout the solution if anics White and coworkers pioneered the
electroneutrality is maintained within the use of microelectrodes in neat organic
depletion layer. However, Feldberg, White, liquids and the theoretical interpretation
and coworkers [15, 79] have shown that of the voltammetric responses [86–93].
electroneutrality may not be maintained A very low level of electrolyte helps
when a significant part of the depletion to lower the ohmic drop and permits
layer lies within the double layer. This well-defined voltammetric waves to be
situation may arise when nanodes or elec- obtained. Voltammograms have been ob-
trodes with dimensions approaching that tained for redox processes of several
of the double layer (100 Å to 1 µm in organic compounds, including nitroben-
the absence of supporting electrolyte) are zene [86, 87], 4-cyanopyridine [88], aniline
used. It is perhaps important to note and pyrrole [94], acetonitrile [95], simple
that even in the absence of deliberately alcohols [96], dimethylsulfoxide [97], and
added electrolytes, the ion concentration DMF [98]. Significantly, because the an-
is typically of the order of 10−6 M and alyte concentration is so high in these
electroneutrality will be maintained even experiments, water and other impurities
for electrodes of micron dimension. More- do not influence the observed responses to
over, charged species are typically created any great extent.
within the depletion layer during elec- Perhaps the most specific feature of
trolysis. Thus, as the Faradaic reaction voltammetry of pure organics is the for-
proceeds, the ionic strength within the mation of a microlayer of an ionic liquid at
depletion layer increases. the electrode surface [88]. This thin layer of
a
20 b
c
15 d
Current
[nA]
10
−5
−0.6 −0.8 −1.0 −1.2 −1.4
E
[V]
Fig. 9 Staircase voltammograms for reduction of 0.65 mM TlSO4 at
a silver-based mercury film microelectrode. The concentration of
supporting electrolyte (LiClO4 ) is 0 (a), 0.1 (b), 1 (c) and 100 mM (d).
(Reproduced with the permission of the American Chemical Society
from K. M. Ciszkowska, J. G. Osteryoung, Anal. Chem. 1995,
67, 1125.)
ionic liquid also leads to large differences allowed self-association of the solution
in the viscosity of the media, which change components to be detected.
both the activities of all species close to
the electrode surface and their diffusion 2.5.5.6.4 Complexation Equilibria Tradi-
coefficients. These layers can be very sta-
tionally, voltammetric investigations of
ble, especially for the electrooxidation of
complexation suffered relative to spec-
simple alcohols and the electroreduction
troscopy because an excess of supporting
of nitrobenzene. In extreme situations,
electrolyte had to be added. A high con-
precipitation may occur at the electrode
centration of electrolyte can cause contam-
surface. White and coworkers have used
ination or can compete with the analyte
interferometry [90] and voltammetry at el-
evated pressure [91] to probe the interfacial of interest in the complexation reaction.
structure in these systems. These care- Moreover, the high ionic strength can
ful studies allow the pressure dependence cause the activity of the analyte to devi-
of molecular transport to be investigated ate significantly from its concentration.
and have demonstrated that the interfacial Also, the ability to perform electrochem-
layer can be compressed [91]. White and ical measurements without deliberately
Ragsdale [93] used the Cullinan–Vignes adding supporting electrolyte may extend
equation to predict the reduction wave the range of analyte : ligand concentrations
height of nitrobenzene in acetonitrile over that can be investigated.
a wide range of compositions. Signifi- Palys and coworkers [99] have extended
cantly, the deviations of experimental data the work of Myland and Oldham [100]
from the theoretical predictions for mole to develop a model that describes the
fractions of nitrobenzene less than 0.4 steady state microelectrode response for
a wide range of complexes that are inert authors demonstrate that the experimen-
on the experimental timescale. Theory tally observed current for the reduction
indicates that the steady state limiting of [Ru(2,2 -dipyridyl)3 ]2+ is approximately
current depends on the type of complex- 50% smaller than that expected for a di-
ation equilibrium, the type of the change rect two-electron reduction reaction. In
in the reactant charge number, and the contrast, the theory that includes repro-
complex formation constant, β. The the- portionation effects accurately models the
ory provides a good description of the experimental response.
experimental data found for the com- Measurement of reaction kinetics can
plexation of europium(III) with 1,4,8,11- also be facilitated by employing microelec-
tetraazacyclotetradecane (cyclam) in a 4 : 1 trodes and low ionic strength solutions.
mixture (v/v) of methanol and dimethyl- For example, in an elegant investigation,
sulfoxide. Unwin and Macpherson probed the kinet-
ics of silver chloride dissolution in aqueous
2.5.5.6.5 Electron Transfer Mechanisms solutions containing no supporting elec-
Beyond its relevance for understanding trolyte using the scanning electrochemical
reaction energetics and dynamics, eluci- microscope (SECM) [114]. They also used
dating the mechanism of electron transfer the alternating direction implicit finite dif-
reactions allows devices, materials, and ference method to solve the mass transport
synthetic pathways to be rationally de- equation for the system and demonstrated
signed. The effect of low ionic strength that steady state current-distance measure-
on the reaction mechanism of a wide ments allow the order of the reaction
variety of organic [101–103], organometal- to be determined. Significantly, this in-
lic [104], and inorganic species [105–110] vestigation suggests that rate constants,
has been explored. When attempting to ex- particularly in the fast kinetic limit, can be
tract accurate thermodynamic information measured with greater precision in the ab-
under these low ionic strength conditions, sence of an inert electrolyte. Further infor-
it is essential to accurately correct for mation on SECM is given in Chapter 3.3.
ohmic drop effects [111, 112].
Amatore and coworkers have developed 2.5.5.6.6 Analytical Applications ASV is a
a theory describing successive electron sensitive technique that often provides ex-
transfers in low ionic strength solutions cellent selectivity in the detection of metal
and explored the coupling of migration ions (see Chapter 2.3). In an elegant an-
and homogeneous electron transfer re- alytical study, Ewing and coworkers [80]
actions [113]. The analytical solutions ob- used ASV to determine Pb2+ concen-
tained allow the limiting currents to be trations in solution. The microelectrodes
calculated for any stepwise electron trans- were thin mercury films deposited on ul-
fer reaction as a function of the ratio of trasmall carbon-ring electrodes. Thin ring
the redox to supporting electrolyte concen- electrodes can have effective diffusional
trations. This work provides a dramatic areas that are more than one hundred
insight into the importance of consider- times larger than microdisks of the same
ing the homogeneous electron transfer geometrical area. This increase in acces-
reactions between species that differ by sibility to diffusing species gives a higher
+2 in oxidation state, that is, repro- current efficiency that can reduce the limit
portionation reactions. For example, the of detection by an order of magnitude.
The authors report the effects of several many desirable properties for analysis in
experimental variables on the anodic strip- low ionic strength media such as lake
ping current including, potential scan rate, waters. The determination of lead and cad-
preconcentration duration, deposition po- mium was also investigated by Daniele
tential, concentration of Hg+ during the and Mazzocchin [118] on a mercury film
in situ deposition step, and Pb2+ concen- deposited on a platinum disk electrode.
tration. Performing ASV in the absence of The influence of the solution resistance
deliberately added supporting electrolyte and migration on the peak position, peak
was investigated as a means of reducing width at half-height, and the peak current
impurity levels in the samples. were investigated. The results obtained
The high analyte flux at microelectrodes, reveal that even a small ohmic drop
and their short response times, can be leads to larger fwhm values. Moreover,
exploited to increase the speed and sensi- their investigations emphasize the impor-
tivity of stripping analysis. Baranski and tance of considering the impact of ionic
Harman [115] used fast cathodic stripping strength–dependent liquid junction po-
analysis at microelectrodes for the determi- tentials when attempting to determine
nation of various anions including iodide, peak potentials accurately [119].
bromide, sulfide, and cysteine. The results
obtained using electrodes of conventional 2.5.5.7 Solid State Investigations
size at slow scan rates (ca. 100 mV s−1 ) Voltammetry at microelectrodes without
were compared with those obtained using supporting electrolyte has proven to be
microelectrodes under fast linear scan con- a very useful technique for studying
ditions (ca. 700 V s−1 ). This study suggests the transport of electroactive ions and
that fast scan methods can simultaneously molecules in complex systems including
decrease the analysis time and improve polyelectrolytes, molten salts [120], solu-
sensitivity since larger currents are ob- tions of high molecular weight polymers
served at high scan rate. that contain ionic groups [121–123], col-
As discussed by Tercier and Buffle [116], loidal suspensions [124], and polymeric
when trying to carry out ASV in samples gels [125]. Beyond the traditional objec-
of low or variable ionic strength, it can be tive of understanding ionic conductivity,
difficult to correlate peak heights and an- microelectrodes also allow the dynamics
alyte concentration. This difficulty arises of segmental polymer chain motion to
because counterion transport is coupled be investigated. One approach to probing
to oxidation of the metal that is concen- these dynamics is to dissolve redox-active
trated within the mercury or polymeric molecules in the polymer and then to
film. Therefore, the peak height, width, measure their diffusion rates through the
and position can depend significantly on matrix. The high viscosity of these media
the identity and concentration of the coun- gives diffusion coefficients of the order of
terions in the sample. 10−7 cm2 s−1 , which is almost two orders
Buffle and coworkers [117] have in- of magnitude smaller than values typically
vestigated iridium-based mercury-plated observed in aqueous solution.
microelectrodes for the determination of In recent years, there has been grow-
trace metals including Pb2+ and Cd2+ ing interest in electrochemical studies
at nM levels. These electrodes give in the solid state, that is, in the ab-
analytically reproducible results and have sence of liquid electrolyte phase [126–128].
Suitable candidates for these investiga- bulk mixed-valence materials, such as sin-
tions have mobile charge compensating gle crystals of silicotungstic acid [130, 131].
counterions and contain mixed-valence These crystals contain highly mobile pro-
redox centers. Beyond its theoretical im- tons at a high concentration, which serve
portance, solid networks are important the same purpose as supporting elec-
for chemical sensing [128], as well as trolytes in conventional electrochemistry.
for the development of charge storage, Potential step and cyclic voltammetry have
electrochromic, and molecular electron- been performed using a three-electrode
ics devices [129]. Faulkner and Kulesza solid-state electrochemical cell of the type
have successfully applied microelectrodes illustrated in Fig. 10. As shown in Fig. 11,
to study solid-state redox transitions in the solid-state voltammetry of a single
Ag/AgCl reference
filled with gel + 3MNaCl
Glass or tygon
tubing
Ceramic tip
Fig. 10 Schematic representation of a

three-electrode cell for probing
Graphite ring counter
solid-state redox processes.
(Reproduced with the permission of the
Single crystal American Chemical Society from
P. J. Kulesza, L. R. Faulkner, J. Chen,
Working W. G. Klemperer, J. Am. Chem. Soc.
electrode 1991, 113, 379.)
100 nA
Current
0 0
0.0 −0.2 −0.4 −0.6

E
[V]
Fig. 11 Solid-state voltammetry of a silicotungstic acid single
crystal using a 5-µm-radius carbon fiber microdisk. The scan rate
is 5 mV s−1 . (Reproduced with the permission of the American
Chemical Society from P. J. Kulesza, L. R. Faulkner, J. Am. Chem.
Soc. 1994, 115, 11 878.)
crystal of silicotungstic acid exhibits three particles efficiently catalyze the discharge
well-defined redox transitions correspond- of intrastructural protons.
ing to the injection of one, two, or four Recently, Bond and coworkers developed
electrons to hexavalent tungsten atoms, a very versatile method of studying the
depending on the applied potential. Steady voltammetry of solids in which insoluble
state plateau currents are observed, which microcrystals are immobilized on an elec-
is consistent with predominantly radial trode surface that is then placed in an
mass transport to the microelectrode sur- electrolyte. Voltammetry reveals that nu-
face. Solid-state electrochemical measure- cleation and crystal growth phenomena
ments have allowed the mixed-valence can accompany redox switching [133]. Sys-
redox site concentration Co , the effective tems of this kind have been used to probe
diffusion coefficient Dapp. , the standard the dynamics of crystal growth, the effect
heterogeneous electron transfer rate con- of the charge, and structure of the com-
stant k o , and the formal potential E o , plex on crystallization [134] as well as the
to be measured for this fast redox con- dynamics of charge transport through the
ducting molecular solid. Recently, Kulesza films [135].
and coworkers [132] further developed this In collaboration with the Bond
approach and prepared a single crystal group [136], we have investigated the
of 12-tungstophosphoric acid, which con- voltammetric properties of microcrystals
tains traces of chloroplatinic acid in its of [Os(bpy)2 4-tet-Cl](ClO4 ) (Fig. 12)
secondary hydrated structure. Following (where bpy is 2,2 -bipyridyl and 4-
partial reduction of the system, metal- tet is 3,6-bis(4-pyridyl)-1,2,4,5-tetrazine)
lic Pt microcenters are generated. While that are mechanically attached to
these particles do not adversely affect the macro- and microdisk electrodes. The
voltammetry of the polytungstate, at suffi- complex is insoluble in water, allowing
ciently negative potentials, the dispersed Pt the solid-state redox properties to
E
[V]
0.7 0.5 0.3 0.1 −0.1
0
−1
N N −2
Fig. 12 Structure of [Os(bpy)2 4-tet N Os N
Cl]ClO4 used to form solid deposits on N
Cl −3
microelectrode. The voltammograms HClO4 −4
are steady state responses recorded at
[pA]
1 mV s−1 using a 2-µm-radius

N N
−5
i
N N
microelectrode. The supporting
−6
electrolyte is 1.0 M HClO4 and 1.0 M
N
NaClO4 for the smaller and larger NaClO4 −7
limiting currents, respectively.
(Reproduced with the permission of the −8
American Chemical Society from −9
R. J. Forster, T. E. Keyes, A. M. Bond, J.
Phys. Chem. B 2000, 104, 6389.) −10
be probed by placing the modified probed redox processes occurring on the

electrode in an aqueous electrolyte media. submicrosecond timescale [31, 138].
Scanning electron microscopy reveals Ultrafast electrochemical techniques
that repeated voltammetric cycling in provide information about the kinetics and
sodium perchlorate electrolyte can induce thermodynamics of redox processes that
crystallization of the material on the occur at microsecond or even nanosec-
electrode surface. In contrast, in perchloric ond timescales. This short timescale is
acid, the films remain amorphous even achieved either by making very rapid
after several thousand voltammetric scans. changes in the applied potential or by using
We have used voltammetry conducted ultrasmall probes to achieve very high rates
at slow (1 mV s−1 ) and fast (1 V s−1 ) of diffusion under steady state conditions.
scan rates to determine the absolute The former approach was introduced in
concentrations of redox centers within the the context of LSV and CV in Chapter 2.1.
film (1.8 and 1.6 M for NaClO4 and HClO4 Microelectrodes play pivotal roles in both
electrolytes, respectively) and the apparent approaches. Electrochemistry has several
diffusion coefficients for homogeneous advantages over spectroscopy in that it
charge transport (2.7 × 10−11 and 5.0 × provides direct information about electron
10−11 cm2 s−1 for NaClO4 and HClO4 transfer and coupled chemical reactions.
electrolytes, respectively). The dependence In transient measurements, decreasing
of the voltammetric peak current on the the lower accessible timescale depends
pH of the contacting electrolyte was used to critically on fabricating ultramicroelec-
estimate the pKa of the unbound pyridine trodes that continue to respond ideally as
moiety of the tetrazine ligand as 3.5 ± 0.1 their critical dimension, for example, the
in the solid state as compared to 2.7 ± 0.2 radius of a microdisk, decreases. The sec-
in essentially aqueous solution. ond difficulty is with ohmic drop since a
small diffusion layer corresponds to a large
2.5.5.8 Ultrafast Electrochemical concentration gradient, and hence large
Techniques currents. In steady state approaches, ultra-
Many significant electrochemical events, small probes are required to make short
such as electron and proton transfers, timescale measurements. As discussed in
ligand exchanges, isomerizations, and Sect. 2.5.2.2, various approaches that yield
ejection of leaving groups, occur on nanodes, that is, electrodes of nanometer
the low microsecond and nanosecond dimension, have been developed but the
time domains. To achieve a meaning- production of well-characterized nanodes
ful insight into these redox processes, it remains a challenge.
must be possible to measure large rate
constants. However, conventional electro- 2.5.5.8.1 Transient Techniques In tran-
chemical methods cannot fulfill this role sient electrochemical measurements in-
since they are restricted to millisecond, or volving a solution phase redox couple,
longer, timescales [137]. Thus, while mod- one seeks to create a competition between
ern pulsed laser spectroscopy has provided the reaction of interest, that is, electron
a powerful new insight into chemical pro- transfer at the electrode surface or cou-
cesses that occur at picosecond and even pled homogeneous steps, and diffusion
femtosecond timescales, it is only recently of the species to and from the electrode
that electrochemists have meaningfully surface [139].
Heterogeneous electron transfer dynamics. angles, that is, the Marcus inner sphere re-
Understanding the factors that impact the organization energy is small [154], causing
rate of heterogeneous electron transfer the heterogeneous electron transfer rate
across a metal–solution interface under- constant to be large. Wightman [155] has
pins the development of advanced sen- studied this reaction using cyclic voltam-
sors, batteries, and molecular electronic metry at scan rates up to 105 V s−1 and
devices. As shown in Table 2, the het- found good agreement between the pre-
erogeneous electron transfer dynamics dictions of the Nicholson and Shain theory
of a diverse range of organic and inor- and experiment after correcting for the
ganic species have been investigated us- nonspherical nature of the microdisc used.
ing transient techniques [140–153]. How- Despite the many elegant investigations
ever, cyclic voltammetry of the an- that have been conducted into the het-
thracene–anthracene anion radical has erogeneous electron transfer dynamics of
been used extensively as a reference sys- solution phase reactants, the magnitude
tem for characterizing new electrodes and of the diffusion-controlled current at short
instruments. This focus arises because the times ultimately places a lower limit on
aromatic nature of the molecule leads to the accessible timescale. As described by
very small changes in the bond lengths and Eq. (10), the thickness of the diffusion layer
Tab. 2 Rate constants for heterogeneous electron transfer as determined using transient methods
Analyte Electrodea ko [cm s−1 ] References
Anthracene Au, 6.5 µm 3.46 ± 0.55 155

Au, 5 µm 3.3 140
Au, 5 µm 2.6 141
Au, 3, 8.5 µm 3–4.8 142, 143
Au, 6 µm 10 144
Anthraquinone Pt cylinder, r = 25.4 µm, l < 0.25 cm 1.78 ± 0.35 155
Pt, 5 µm 1.5 141
Au, 6.5 µm 2.4 145
Benzoquinone Pt, 5 µm 0.14 141
Ferrocene Pt, 5 µm 1.1 141
Au, 5 µm 3.1 ± 1.1 146
Pt, 5 to 25 µm 1.4–3.6 145
Pt, 10 µm 147
Ferrocyanide C cylinder, r = 15 µm, l = 500 µm 0.0114 ± 0.0022 148
Pt, 10, 50 µm 0.42 ± 0.03 149
C, 5 µm 0.06 ± 0.05 150
Pt, 1 to 30 µm 0.64–0.79 151
Pt, 20 µm 0.2 152
9-fluorenone Pt, 6 µm 3 153
[Ru(bpy)3 ]2+ Hg, 5.5 µm 0.45 146
Au, 5 µm 2.5 146
a Dimension given is the radius of a microdisc electrode unless otherwise stated.

is proportional to the square root of the developed that exhibit close to ideal re-
polarization time. One can estimate that versible electrochemical behavior under a
the diffusion layer thickness is approx- wide variety of experimental conditions of
imately 50 Å if the diffusion coefficient timescale, temperature, solvent, and elec-
is 1 × 10−5 cm2 s−1 and the polarization trolyte.
time is 10 ns. Given a typical bulk con- We have investigated the redox proper-
centration of the electroactive species of ties of the ground and electronically ex-
1 mM, this analysis reveals that only 10 000 cited states of [Ru(bpy)2 Qbpy]2+ (Fig. 13)
molecules would be oxidized or reduced monolayers using a combination of a
at a 1-µm-radius microdisc under these nanosecond pulsed laser and cyclic voltam-
conditions. The average current for this metry at megavolt per second scan
experiment is only 170 nA, which is too rates [160]. The ground state voltamme-
small to be detected with low nanosecond try of these monolayers in acetonitrile is
time resolution. nearly ideal, and five redox states are acces-
Therefore, in order to probe the dy- sible over the potential range from +1.3 to
namics and energetics of ultrafast hetero- −2.0 V. Chronoamperometry conducted
geneous electron transfer dynamics this on a microsecond timescale was used
diffusion limitation must be eliminated. to measure the heterogeneous electron
One successful approach to achieving transfer rate constant, k, for both metal
this objective is to use self-assembled and ligand-based redox reactions. Standard
or spontaneously adsorbed monolayers. heterogeneous electron transfer rate con-
When immobilized on an electrode sur- stants, k o , were 5.1 ± 0.3 × 105 s−1 , 3.0 ±
face, the electroactive species no longer 0.1 × 106 s−1 and 3.4 ± 0.2 × 106 s−1 for
needs to diffuse to the electrode to un- the 3+/2+, 2+/1+, and 1+/0 couples, re-
dergo electron transfer. Moreover, the spectively. Significantly, the free energy of
electroactive species is preconcentrated activation is constant at 6.9 ± 0.6 kJ mol−1
on the electrode surface. For example, for all three redox couples investigated.
in the situation considered above, there Following photoexcitation using a laser
will be approximately 1.7 × 10−20 mol of pulse at 355 nm, emission is observed
electroactive material within the diffusion from the monolayers with an excited state
layer. Given that the area of a 1-µm disc is lifetime (6.2 µs) that exceeds that of the
approximately 3.1 × 10−8 cm2 , this trans- complex in solution (1.4 µs). It appears
lates into an equivalent surface coverage that weak electronic coupling between the
of about 5.4 × 10−13 mol cm−2 . In con- adsorbates and the electrode means that
trast, the surface coverage, , observed the excited states are not completely de-
for dense monolayers of adsorbates is activated by radiationless energy transfer
typically more than two orders of magni- to the metal. As illustrated in Fig. 13,
tude larger with coverages of the order of in the first report of its kind, we used
10−10 mol cm−2 typically being observed. voltammetry at megavolt per second scan
This higher concentration gives rise to rates to directly probe the redox poten-
much larger currents that are easier to de- tials and electron transfer characteristics
tect at short timescales. As exemplified by of electronically excited species.
the work of Chidsey [156], Abrûna [157],
Faulkner [158], and Finklea [159], elec- Homogeneous chemical kinetics. The de-
troactive adsorbed monolayers have been crease in the lower accessible time limit
13
1.5 1 0.5 0 −0.5 −1 −1.5 −2
[µA]
E
i
Scan 2
[V]
−7
N
N
N
Ru N −12
N N
Scan 1
−17
N N
−22
Fig. 13 Structure of the [Ru(bpy)2 Qbpy]2+ complex used to form luminescent

monolayers on platinum microelectrodes. The electrode radius is 5 µm and
the response shown is immediately following laser excitation at 355 nm. The
scan rate is 3 × 105 V s−1 , the surface coverage is 1.1 × 10−10 mol cm−2 and
the supporting electrolyte is 0.1 M TBABF4 in acetonitrile. The initial potential
is −1.2 V. (Reproduced with the permission of the American Chemical Society
from R. J. Forster, T. E. Keyes, J. Phys. Chem. B 1998, 102, 10 004.)
possible using microelectrodes also has response observed for the oxidation of
important implications for probing the dy- anthracene at slow scan rates becomes
namics of rapid homogeneous chemical fully reversible at a scan rate of 104 V s−1 .
reactions. For example, bimolecular reac- This behavior is opposite to that expected
tions in solution cannot proceed faster when heterogeneous electron transfer is
than the rate at which molecules come slow and suggests that the cation radical
into close contact. Thus, bimolecular rate undergoes a following chemical reaction.
constants cannot exceed the diffusion- The ability to make the voltammetric re-
limited rate constant that is of the order sponse reversible means that the formal
of 109 to 1010 M−1 s−1 in most organic potentials of highly reactive species can be
solvents. Since the characteristic time of accurately measured.
cyclic voltammetry is RT/Fν, where ν is The kinetics and mechanism of
the scan rate, experiments performed at the homogeneous reactions following
MV s−1 scan rates allow kinetic informa- the reduction of NAD+ and synthetic
tion, such as lifetimes that are close to the analogs continue to be extensively
diffusion limit, to be obtained [161]. For investigated. Several reaction mechanisms
example, as illustrated in Fig. 14, Wight- have been observed for NAD+ analogs,
man [162] has shown that the irreversible with molecules containing at least one
100 V s−1
20 nA
10,000 V s−1
100 nA
1.0 0.5
E vs Ag/Ag+
[V]
Fig. 14 Cyclic voltammograms obtained at a 50-µm-radius
platinum microelectrode for the oxidation of anthracene
(2.36 mM) in acetonitrile containing 0.6 M TEAP. (Reproduced
with the permission of the American Chemical Society from
J. O. Howell, R. M. Wightman, J. Phys. Chem. 1984, 88, 3915.)
hydrogen at the 4-position, that is, direct electrogenerated neutral radicals at

analogs of NAD+ , typically undergoing a rate of approximately 1.6 ± 0.1 ×
a dimerization reaction [163]. We have 107 M−1 s−1 in DMF. The 1-Methyl-
probed the effect of changing the 4-carbamidopyridinyl and 1-Methyl-3,4-
substitution pattern on the mechanism dicarbamidopyridinyl radicals react via a
and kinetics of the homogeneous solution pH-dependent ECE-DISP1 mechanism, E,
reactions of electrogenerated 1-Methyl- C and DISP denote electron transfer,
carbamidopyridinyl radicals. Fast scan following chemical and disproportionation
cyclic voltammetry and double potential reactions, respectively. These high-speed
step chronoamperometry conducted on investigations not only provide accurate
a microsecond timescale reveal that thermodynamic and kinetic information
1-Methyl-3-carbamidopyridinyl radicals but also allow the effect of substitution
react via a dimerization mechanism patterns on reaction mechanisms to
involving direct coupling of the be elucidated.
2.5.5.9 Steady State Electrochemistry of solution phase redox-active molecules

Under steady state conditions, the critical and a representative sample is given in
timescale for diffusion, tD , is not related to Table 3 [113, 164–173]. These data show
the actual duration of the experiment since that the rate of electron transfer across the
the diffusion layer thickness depends only electrode–solution interface varies signif-
on the electrode radius. Under these con- icantly, rate constants up to 220 cm s−1
ditions, tD is approximately equal to r 2 /D. being observed for ferrocene. These vari-
This result has a profound impact on the ations reflect differences in the reaction
size of microelectrodes required to make adiabaticity and activation barriers that ex-
measurements at steady state, for exam- ist for the individual systems.
ple, given a typical diffusion coefficient of
10−5 cm2 s−1 , microelectrodes with radii Homogeneous chemical kinetics. A sec-
of less than 30 nm are required to address ond important application of steady state
submicrosecond timescales. measurements is in studies of chemical re-
activity. Steady state measurements using
Heterogeneous electron transfer dynamics. electrodes of different radii can provide a
Steady state voltammetry has been widely powerful insight into the kinetics of ho-
used to probe the dynamics of heteroge- mogeneous reactions where the limiting
neous electron transfer for a wide range current density depends on the magnitude
Tab. 3 Rate constants for heterogeneous electron transfer determined using steady state methods
Analyte Electrodea ko [cms−1 ] References
Anthracene Au ring
r = 0.09 µm r = 5 µm 3.33 ± 0.05 164
(C6 H6 ) Cr (CO)3 + Pt, 25 µm ≥0.3 165
Cytochrome c C, 6.3 µm >0.4 166
9,10-diphenylanthracene Au ring
r = 90 µm r = 5 mm 5.7 ± 0.1 164
Ferrocene C, 6 µm 2.3 ± 0.8 167
Pt, 0.3 to 25 µm ≥6 168
Pt, 1 µm >2 169
Pt, 16 Å to 2.6 µm 220 ± 120
Fe(OEP)(N-Melm)2 + Pt, 1 to 25 µm 0.4 170
Pt, 0.5 to 12.5 µm 0.38 171
Pt, 1 to 25 µm 0.35 170
Fe(TPP) (Hlm)2 + Pt, 1 to 25 µm 0.5 170
Fe (TPP) py2 + Pt, 1 to 25 µm 0.6 170
MV2+(F) Pt, 22 Å to 0.21 µm 170 ± 90
Naphthalene Au ring
r = 0.2 µm r = 20.5 µm 0.88 ± 0.02 164
Oxygen Pt, 12.7 to 250 µm 0.63 ± 0.05 172
Ru (NH3 )6 3+ Au, 5 µm 0.076 173
Pt, 11 Å to 11.1 µm 79 ± 44
[Ru(bpy)3 ]2+ Pt, 11.2 µm 113
Tetracyanoquinodiethane Au ring
r = 0.2 µm 0.23 ± 0.01 164
r = 20.5 µm
Zn(TPP) Pt, 1 to 25 µm >1 170

Tab. 4 Rate constants for homogeneous chemical reactions determined using steady state methods
Analyte Electrodea k References
Anthracene oxidation Pt, 0.3 to 62.5 µm 190 ± 50 s−1 174

Ascorbic acid oxidation at Pt, 2.5 to 25 µm 1.3 × 105 M−1 s−1 175
a Prussian blue film
Ferrocyanide oxidation in Pt band pair, gap = 2 27 ± 4 M−1 s−1 176
the presence of ascorbic to 12 µm
acid
9,10-diphenylanthracene + C, 6 to 9 µm 3.9 ± 0.6 M−1 s−1 177
4,4-dibromodiphenyl
Anion radicals + alkyl Pt, 0.25 µm 9 × 10−4 to 1.7 × 104 178, 179
halides
[Fe(CN)6 ]3− + Pt, 0.3 to 25 µm 3.0 ± 0.6 × 103 M−1 s−1 180
aminopyridine
Pt, 2.5 to 432 µm 1.8 × 103 M−1 s−1 181
Pt band pair, gap = 2 8 ± 1 × 102 M−1 s−1 176
to 12 µm
Hexamethylbenzene Pt, 0.3 to 25 µm 720 ± 100 s−1 174
oxidation
H2 evolution on Pt from Pt, 0.3 to 25 µm 4.1 × 1010 M−1 s−1 180
acetic acid solution
1-napthylamine oxidation Pt, 0.5 to 12.5 µm 4.1 × 103 s−1 182
Triphenylamine oxidation Pt, 0.3 to 20 µm >3 × 104 M−1 s−1 174
Thioselenanthrene 8.87 ± 1.1 s−1 183
Dibenzo-1,2-diselenine 20.7 ± 2.8 s−1 184
of the homogeneous rate constant. Hence, reactions has been rigorously treated in
as described in Table 4 coupled chemical the literature. Osteryoung and cowork-
(C) and electron transfer (E) reactions, for ers [186] used steady state voltammetry
example, CE mechanisms, catalytic follow- and simulations in the presence and ab-
up processes, as well as reactions involv- sence of supporting electrolyte to describe
ing disproportionation (DISP) have been the reactions under diffusional and mi-
characterized [174–184]. As illustrated in grational transport conditions. The treat-
Fig. 15, the position of the wave on the po- ment covers cases ranging from strong
tential axis depends on the homogeneous acids (essentially an E mechanism) to
reaction rate and kinetic information can weak acids in which the dissociation step
be obtained by probing how E1/2 depends is important.
on the electrode radius [185]. It is important to note that SECM plays
The dissociation of protons from or- an important role not only in topographical
ganic acids followed by reduction of the imaging but also in probing chemical
protons to hydrogen is an archetypal ex- reactivity [187, 188]. A comprehensive
ample of a CE mechanism. However, description of the design, functioning,
despite their ubiquity and importance, it and application of the SECM is given in
is only recently that this family of CE Chapter 3.3.
Fig. 15 Dependence of E1/2 for

an EC reaction on the logarithm
of the electrode radius.
(Reproduced with the
F
permission of Elsevier RT
T/n
lnK2
publishers from K. B. Oldham, J. nF
=R
Electroanal. Chem. 1991, 313, 3.)
E 1/ 2
pe
slo
1/2 k D 1/2
D
ln ln k2
k2 −2
ln a
2.5.5.10 AC Electrokinetics As illustrated in Fig. 16, AC electrokinet-

One of the defining themes of the next ics uses an electric field created between
ten years will be molecular electronics, es- two or more microelectrodes to induce
pecially nanotechnology. Electrochemistry a dipole within a cell, particle, or macro-
has a key role to play in this emerg- molecule. With a suitable electrode design,
ing area. Electrochemistry can be used a variety of motions including pushing,
to produce and characterize clean sur- pulling, and rotation can be induced by
faces, for example, electrochemical clean- changing the nature of the dynamic field.
ing of metals. It can direct the assembly In many ways, these forces may be viewed
of supramolecular structures, for exam- as an electrostatic equivalent of optical
ple, by using self-assembled or spon- tweezers [189] and optical spanners [190]
taneously adsorbed monolayers. It can in that they exert translational and rota-
address molecular components, for exam- tional forces due to the interaction between
ple, by switching their oxidation state. It the particle and the applied field. AC
can read their state, for example, through electrokinetic techniques such as dielec-
measurements of the open circuit po- trophoresis [191] and electrorotation [192]
tential. Moreover, microelectrodes offer have been utilized for many years for the
the possibility of selecting and precisely manipulation, separation, and analysis of
moving nanometer-dimensioned objects cellular-scale particles. However, recent
using AC electrokinetics. AC electrokinet- advances in semiconductor manufactur-
ics offers advantages over scanning-probe ing technology have enabled researchers
methods of nanoparticle manipulation in to develop electrodes for manipulating
that the equipment used is simple, cheap, macromolecules as small as 9 kDa us-
and has no moving parts, relying entirely ing both attractive [193] and repulsive AC
on the electrostatic interactions between electrostatic forces, and to concentrate 14-
the particle and dynamic electric field. nm beads from solution [194]. Trapping
Furthermore, there is theoretical evidence of single particles such as viruses and
that as manufacturing technology further nanometer-diameter latex spheres in con-
improves, single nanoparticles may be ma- tactless potential energy wells [195] has
nipulated using this approach. also been demonstrated. Ultimately, this
Fig. 16 A schematic of typical array of

four microelectrodes used in
dielectrophoresis experiments. The gap
between opposing electrodes in the
center of the array is typically of the
(a) (b) order 10 to 50 µm across, but can be as
small as 500 nm or as large as 1 mm. To
induce dielectrophoretic motion in
(c) (d) particles suspended near the electrode
array, the electrodes are polarized such
that a and c are of the same phase,
while b and d are antiphase. To cause
electrorotation in particles within the
central gap, electrodes B, C, and D
would be phase-shifted by 90◦ , 180◦ ,
and 270◦ with respect to electrode A.
technology may be applied for the manip- 2.5.6.1 Electron Transfer Mechanisms
ulation of single molecules [196]. A profound understanding of those factors
that influence the dynamics of electron
2.5.6 transfer across electrode/solution and ho-
Future Directions mogeneous chemical reactions has been
achieved. However, a level of experimen-
This review has attempted to convey a tal and theoretical insight into coupled
sense of the many significant new in- chemical and electron transfer reactions,
sights and opportunities that have been for example, coupled proton and electron
made possible by microelectrodes. This transfer, remains elusive. It is probable
revolution in electrochemistry has greatly that new approaches and models will
extended the range of solvents, temper- emerge in this key area over the next
atures, and timescales under which it is five years.
now possible to obtain direct informa-
tion about redox processes. In particular, 2.5.6.2 Neurochemistry
today microelectrodes allow experiments, Major advances are likely in the area of
for example, voltammetry in oil or con- intrasynaptic detection of both electrical
crete, to be performed that would simply and chemical messengers. In particular,
have been impossible a few years ago. electrochemical measurements at micro-
This advance has not only revolution- electrodes and nanodes promise to provide
ized the field internally, it has broadened powerful new insights into cognitive brain
the impact of electrochemistry into new function including learning, memory, and
dimensions of space and time, for ex- perhaps even consciousness itself. How-
ample, microsecond monitoring of neu- ever, given that neurotransmitter release
rotransmitter release with single cell spa- can occur at frequencies of the order of
tial resolution. 10 kHz, success in this area will only be
There are significant opportunities both possible if ideally responding nanodes can
within the field and other disciplines for be produced. While approaches to cor-
electrochemists to make pivotal contribu- recting seal and other defects have been
tions. We consider some of them below. developed for metallic microelectrodes,
they have not been widely investigated for precisely delivered to their intended site
the carbon fiber microelectrodes that are of action.
often preferred for biochemical investiga-
tions. 2.5.6.4 Analysis
Many key problems in chemical sensing
2.5.6.3 Medicine can be solved by using microelectronics
and micromechanics to fabricate systems
2.5.6.3.1 Nonanimal-based Drug Testing that convert biological or chemical re-
Microelectrode arrays are proving to be sponses into an electrical signal. For
useful platforms on which collections of instance, a well-known DNA chip already
fully functioning cells can be grown and allows infectious diseases or genetic al-
manipulated. Moreover, there is increas- terations associated with many cancers
ing evidence that these cell cultures react to be detected. It seems likely that elec-
to chemical stimulation in a similar way trochemical detection of mutations in
to their parent tissue in fully functional human DNA using microelectrode arrays
animals. Therefore, it is likely that these will continue to develop so as to offer
microarray methods will play increasingly high-speed sequencing and detection of
important roles in determining the effi- pathogenic DNA.
cacy, toxicity, and therapeutic levels for
new drug treatments. 2.5.6.5 Nanodevices and Circuits
Molecular electronics represents a pow-
2.5.6.3.2 Prosthetic Devices To reach for erful approach to the continued minia-
an object of interest, primates typically use turization of electronic circuits down to
their visual system to determine where the lower nanometer scale. One signif-
the object is located and how to move icant challenge is the electrical connec-
the arm while avoiding obstacles. Plans tion of molecular devices by nanowires.
to move the arm are then formulated in In this regard, the ability of micro-
the posterior parietal cortex. Then the mo- electrodes, to both image, for example,
tor areas develop these plans into precise through SECM and fabricate, for example,
movement instructions that are normally through spatially controlled electrodepo-
conveyed down the spinal cord to the mus- sition, micro- and nano-structures will
cles in the arm. When the spinal cord is continue to be invaluable.
injured, neural commands can no longer
reach the arm. On the basis of the cur- References
rent state of the art in microelectrodes,
it seems likely that it will be possible to 1. A. M. Bond, Analyst 1994, 119, R1.
2. R. J. Forster, Chem. Soc. Rev. 1994, 289.
tap into these signals and reroute them 3. M. I. Montenegro, M. A. Queiros, J. L. Das-
around the damaged area to the natural chbach, (Eds.), Microelectrodes: Theory and
arm, by microstimulating muscles, or to Applications, Kluwer Academic Press, Dor-
a prosthetic arm. Microelectrodes have a drecht, The Netherlands, 1991.
key role to play in this process by provid- 4. J. Heinze, Angew. Chem., Int. Ed. Engl. 1993,
105, 1327.
ing a fundamental understanding of the 5. R. M. Wightman, D. O. Wipf, Electroanalyt-
underlying neurobiology of this process, ical Chemistry (Ed.: A. J. Bard), Marcel Dek-
and by allowing the nerve impulses to be ker, New York, 1989, pp. 267–353, Vol. 15.
6. J. Heinze, Angew. Chem., Int. Ed. Engl. 1993, 32. D. O. Wipf, A. C. Michael, R. M. Wight-
32, 1268. man, J. Electroanal. Chem. 1989, 15, 269.
7. P. W. Davies, F. Brink, Rev. Sci. Instrum. 33. L. R. Faulkner, M. R. Walsh, C. Xu in Con-
1942, 13, 524. temporary Electroanalytical Chemistry (Ed.:
8. R. M. Penner, M. J. Heben, T. L. Longin A. Ivaska), Plenum Press, New York, 1990,
et al., Science 1990, 250, 688. pp. 5–12.
9. P. A. Christensen, Chem. Rev. 1992, 197. 34. W. J. Bowyer, E. E. Engelman, D. H. Evans,
10. C. Lee, C. J. Miller, A. J. Bard, Anal. Chem. J. Electroanal. Chem. 1989, 67, 262.
1991, 63, 78. 35. C. Amatore, B. Fosset, Anal. Chem. 1996,
11. C. J. Slevin, N. J. Gray, J. V. MacPherson 68, 4377.
et al., Electrochem. Commun. 1999, 1, 282. 36. J. W. Bixler, A. M. Bond, Anal. Chem. 1986,
12. D. K. Y. Wong, L. Y. F. Xu, Anal. Chem. 58, 2859.
1995, 67, 4086. 37. R. S. Stojanovic, A. M. Bond, E. C. V. But-
13. Y. H. Shao, M. V. Mirkin, G. Fish et al., ler, Anal. Chem. 1990, 62, 2692.
Anal. Chem. 1997, 69, 1627. 38. J. Wang, B. Tian, Anal. Chim. Acta 1993,
14. X. J. Zhang, W. M. Zhang, X. Y. Zhou et al., 274, 1.
Anal. Chem. 1996, 68, 3338. 39. J. Wang, D.-B. Luo, T. Horiuchi, Electro-
15. C. P. Smith, H. S. White, Anal. Chem. 1993, analysis 1998, 10, 107.
65, 3343. 40. G. T. A. Kovacs, C. W. Storment, S. P.
16. W. S. Baker, R. M. Crooks, J. Phys. Chem. B Kounaves, Sens. Actuators, B 1995, 23, 41.
1998, 102, 10 041. 41. C. Belmont, M.-L. Tercier, J. Buffle et al.,
17. B. A. Grzybowski, R. Haag, N. Bowden Anal. Chim. Acta 1996, 329, 203.
et al., Anal. Chem. 1998, 70, 4645. 42. R. I. Stefan, J. F. van Staden, H. Y. Aboul-
Enien, Crit. Rev. Anal. Chem. 1999, 29, 133.
18. H. Sugimura, O. Takai, N. Nakagiri, J. Elec-
43. C. Amatore, L. Thouin, J. Warkocz, J. Chem.
troanal. Chem. 1999, 473, 230.
-Eur. J. 1999, 5, 456.
19. M. A. Guillorn, D. W. Carr, R. C. Tiberio
44. P. N. Bartlett, P. R. Birkin, J. H. Wang et al.,
et al., J. Vac. Sci. Technol., B 2000, 18, 1177.
Anal. Chem. 1998, 70, 3685.
20. P. Van Gerwen, W. Laureyn, W. Laureys
45. M. M. Collinson, R. M. Wightman, Science
et al., Sens. Actuators, B 1998, B48, 73.
1994, 268, 1883.
21. H. Schift, R. W. Jaszewski, C. David et al.,
46. F.-R. F. Fan, A. J. Bard, Science 1995, 267,
Microelectron. Eng. 1999, 46, 121. 871.
22. G. Phillipp, T. Weinmann, P. Hinze et al., 47. B. Giros, M. Jaber, S. Jones et al., Nature
Microelectron. Eng. 1999, 46, 157. 1996, 379, 606.
23. M. Madou, Fundamentals of Microfabrica- 48. R. Kashyap, M. Gratzl, Anal. Chem. 1998,
tion, CRC Press, Boca Raton, Fla., 1997. 70, 1468.
24. P. J. Holmes, R. G. Loasby, Handbook of 49. R. A. Clark, A. G. Ewing, Anal. Chem. 1998,
Thick Film Technology, Electrochemical 70, 1119.
Publications, Port Erin, Great Britain, 1976. 50. A. Vogel, J. W. Schulze, Electrochim. Acta
25. H. Suzuki, Mater. Sci. Eng., C 2000, 12, 55. 1999, 44, 3751.
26. M. Hatzakis, B. J. Canavello, J. M. Shaw, 51. C. S. Henry, I. Fritsch, J. Electrochem. Soc.
IBM J. Res. Dev. 1980, 24, 452. 1999, 146, 3367.
27. W. K. Ko, Mater. Chem. Phys. 1995, 42, 169. 52. J. S. Rossier, M. A. Roberts, R. Ferrigno
28. G. Schmitt, J. W. Schultze, F. Fassbender et al., Anal. Chem. 1999, 71, 4294.
et al., Electrochim. Acta 1999, 44, 3865. 53. R. A. Clark, P. B. Hietpas, A. G. Ewing,
29. J. Newman, J. Electrochem. Soc. 1968, 113, Anal. Chem. 1997, 69, 259.
501. 54. R. M. Wightman, P. Runnels, K. Troyer,
30. A. J. Bard, L. R. Faulkner, Electrochemical Anal. Chim. Acta 1999, 400, 5.
Methods: Fundamentals and Applications, 55. G. Chen, A. G. Ewing, Crit. Rev. Neurobiol.
John Wiley & Sons, New York, 1980. 1997, 11, 59.
31. R. J. Forster, Ultrafast electrochemical tech- 56. P. C. Garell, M. A. Granner, M. D. Noh,
niques in Encyclopedia of Analytical Chem- et al., Rev. Sci. Instrum. 1998, 69, 4027.
istry (Ed.: R. Meyers), John Wiley & Sons, 57. H. Lu, M. Gratzl, Anal. Chem. 1999, 71,
New York, 2000, pp. 10 142–10 171. 2821.
58. K. T. Kawagoe, J. A. Jankowski, R. M. 83. J. Wang, J. M. Zadeii, J. Electroanal. Chem.

Wightman, Anal. Chem. 1991, 63, 1589. 1988, 246, 297.
59. R. M. Wightman, T. J. Schroeder, J. F. Fin- 84. M. Ciszkowska, J. G. Osteryoung, Anal.
negan et al., Biophys. J. 1995, 68, 383. Chem. 1995, 67, 1125.
60. D. J. Michael, R. M. Wightman, J. Pharm. 85. J. B. Cooper, A. M. Bond, K. B. Oldham, J.
Biomed. Anal. 1999, 19, 33. Electroanal. Chem. 1992, 331, 877.
61. D. M. Cannon, N. Winograd, A. G. Ewing, 86. R. A. Malmsten, H. S. White, J. Electrochem.
Annu. Rev. Biophys. Biomol. Struct. 2000, 29, Soc. 1986, 133, 1067.
239. 87. R. A. Malmsten, C. P. Smith, H. S. White,
62. G. W. Gross in Enabling Technologies for Cul- J. Electroanal. Chem. 1986, 215, 223.
tured Neural Networks (Eds.: D. A. Stenger, 88. R. B. Morris, K. F. Fischer, H. S. White, J.
T. M. McKenna), Academic Press, New Phys. Chem. 1988, 92, 5306.
York, 1994, pp. 277–317. 89. J. D. Norton, S. A. Anderson, H. S. White,
63. K. S. Guillory, R. A. Normann, J. Neurosci. J. Phys. Chem. 1992, 96, 3.
Methods 1999, 9, 21. 90. Q. Li, H. S. White, Anal. Chem. 1995, 67,
64. G. W. Gross, F. U. Schwalm, J. Neurosci. 561.
Methods 1994, 52, 73. 91. K. J. Stevenson, H. S. White, J. Phys. Chem.
65. G. W. Gross, H. M. E. Azzazy, M.-C. Wu 1996, 100, 18 818.
et al., Biosens. Bioelectron. 1995, 10, 553. 92. S. C. Paulson, N. D. Okerlund, H. S. White,
66. S. I. Morefield, E. W. Keefer, K. D. Chap- Anal. Chem. 1996, 68, 581.
man et al., Biosens. Bioelectron. 2000, 15, 93. S. R. Ragsdale, H. S. White, J. Electroanal.
383. Chem. 1997, 432, 199.
67. P. A. Starr, J. L. Vitek, R. A. E. Bakay, Neu- 94. R. L. McCarley, M. Morita, K. O. Wilbourn
rosurgery 1998, 43, 989.
et al., J. Electroanal. Chem. 1988, 245, 321.
68. J. W. Gardner, P. N. Bartlett, Electronic
95. J. Cassidy, S. B. Khoo, S. Pons et al., J. Phys.
Noses, Principles and Applications, Oxford
Chem. 1985, 89, 3933.
University Press, Oxford, 1999.
96. M. Ciszkowska, Z. Stojek, J. Electroanal.
69. W. Göpel, Sens. Actuators, B 2000, 65, 70.
Chem. 1993, 344, 135.
70. M. S. Freund, N. S. Lewis, Proc. Natl. Acad.
97. M. Koncka, Z. Stojek, Electroanalysis 1995,
Sci. U.S.A. 1995, 92, 2652.
7, 1010.
71. J. W. Gardner, P. N. Bartlett, Sens. Actua-
tors, B 1994, 18, 21. 98. J. Gadomska, Z. Stojek, Electroanalysis 1998,
72. T. Bachinger, C. F. Mandenius, Trends Bio- 10, 307.
technol. 2000, 18, 494. 99. M. Palys, Z. Stojek, M. Bos et al., Anal.
73. L. B. Kish, R. Vajtai, C. G. Granqvist, Sens. Chim. Acta 1997, 337, 5.
Actuators, B 2000, 71, 55. 100. J. C. Myland, K. B. Oldham, J. Electroanal.
74. C. Distantea, P. Sicilianob, L. Vasanellib, Chem. 1993, 347, 49.
Sens. Actuators, B 2000, 69, 248. 101. M. F. Bento, M. J. Medeiros, M. I. Monte-
75. M. Ciszkowska, Z. Stojek, J. Electroanal. negro et al., J. Electroanal. Chem. 1993, 345,
Chem. 1999, 466, 129. 273.
76. C. Amatore, M. R. Deakin, R. M. Wight- 102. J. D. Norton, H. S. White, J. Electroanal.
man, J. Electroanal. Chem. 1987, 225, 49. Chem. 1992, 325, 341.
77. C. Amatore, B. Fosset, J. Bartelt et al., J. 103. R. T. Robertson, B. D. Pendley, J. Elec-
Electroanal. Chem. 1988, 256, 255. troanal. Chem. 1994, 374, 173.
78. K. B. Oldham, J. Electroanal. Chem. 1988, 104. F. Rourke, R. Gash, J. A. Crayston, J. Organ-
250, 1. omet. Chem. 1992, 423, 223.
79. J. D. Norton, H. S. White, S. W. Feldberg, J. 105. J. B. Cooper, A. M. Bond, Anal. Chem. 1993,
Phys. Chem. 1990, 94, 6772. 65, 2724.
80. D. K. Y. Wong, A. G. Ewing, Anal. Chem. 106. B. Soucaze-Guillous, W. Kutner, K. M.
1990, 62, 2697. Kadish, J. Electrochem. Soc. 1996, 143, 550.
81. A. M. Bond, F. G. Thomas, Langmuir 1988, 107. D. Pletcher, S. Sotiropoulos, J. Electroanal.
4, 341. Chem. 1993, 356, 109.
82. L. Nyholm, G. Wikmark, Anal. Chim. Acta 108. D. Pletcher, S. Sotiropoulos, J. Chem. Soc.,
1993, 273, 41. Faraday Trans. 1995, 91, 457.
109. D. R. Rolison, J. Z. Stemple, J. Chem. Soc., 135. P. G. Bruce, (Ed.), Solid State Electrochem-
Chem. Commun. 1993, 25. istry, Cambridge University Press, Cam-
110. Y. Takeoka, T. Aoki, K. Sanui et al., J. bridge, UK, 1995.
Electroanal. Chem. 1997, 438, 153. 136. R. J. Forster, T. E. Keyes, A. M. Bond, J.
111. M. F. Bento, L. Thouin, C. Amatore et al., J. Phys. Chem. B 2000, 104, 6389.
Electroanal. Chem. 1998, 446, 91. 137. C. Amatore in Electrochemistry at Ultramicro-
112. M. F. Bento, L. Thouin, I. Montenegro et al., electrodes (Ed.: I. Rubinstein), Marcel Dek-
J. Electroanal. Chem. 1998, 443, 137. ker, New York, 1995, Chap. 4, pp. 131–208.
113. C. Amatore, S. C. Paulson, H. S. White, J. 138. F. Magno, I. Lavagnini, Anal. Chim. Acta
Electroanal. Chem. 1997, 439, 173. 1995, 305, 96.
114. J. V. Macpherson, P. R. Unwin, J. Phys. 139. M. I. Montenegro, Applications of micro-
electrodes in kinetics in Research in
Chem. 1996, 100, 19 475.
Chemical Kinetics (Eds.: R. G. Compton,
115. A. R. Harman, A. S. Baranski, Anal. Chim.
G. Hancock), 1994, pp. 1–80, Vol. 2.
Acta 1990, 239, 35.
140. C. P. Andrieux, D. Garreau, P. Hapiot et al.,
116. M. L. Tercier, J. Buffle, Electroanalysis 1993,
5, 187. 141. M. I. Montenegro, D. Pletcher, J. Elec-
117. R. R. De Vitre, M. L. Tercier, M. Tsacopou- troanal. Chem. 1986, 200, 371.
los et al., Anal. Chim. Acta 1991, 249, 419. 142. C. Amatore, C. Lefrou, F. Pfluger, J. Elec-
118. S. Daniele, G. A. Mazzocchin, Anal. Chim. troanal. Chem. 1989, 270, 43.
Acta 1993, 273, 3. 143. C. Amatore, C. Lefrou, J. Electroanal. Chem.
119. T. T. Wooster, M. L. Longmire, H. Zhang 1992, 324, 33.
et al., Anal. Chem. 1992, 64, 1132. 144. D. O. Wipf, R. M. Wightman, J. Phys. Chem.
120. M. E. Williams, H. Masui, R. W. Murray, J. 1989, 93, 4286.
Phys. Chem. B 2000, 104, 10 699. 145. I. Lavagnini, P. Pastore, F. Magno, J. Elec-
121. S. E. Morris, M. Ciszkowska, J. G. Oster- troanal. Chem. 1992, 333, 1.
young, J. Phys. Chem. 1993, 97, 10 453. 146. D. O. Wipf, E. W. Kristensen, M. R. Deakin
122. C. Scordilis-Kelley, J. G. Osteryoung, J. et al., Anal. Chem. 1988, 60, 306.
Phys. Chem. 1996, 100, 797. 147. H. F. Zhou, N. Y. Gu, S. J. Dong, J. Elec-
123. M. Ciszkowska, J. G. Osteryoung, J. Phys. troanal. Chem. 1998, 441, 153.
Chem. 1998, 102, 291. 148. A. Neudeck, J. Dittrich, J. Electroanal. Chem.
124. J. M. Roberts, P. Linse, J. G. Osteryoung, 1991, 313, 37.
Langmuir 1998, 14, 204. 149. W. Huang, R. McCreery, J. Electroanal.
125. M. Ciszkowska, M. D. Guillaume, J. Phys. Chem. 1992, 326, 1.
Chem. A 1999, 103, 607. 150. N. Oyama, T. Ohsaka, N. Yamamoto et al.,
126. P. J. Kulesza, J. A. Cox, Electroanalysis, 1998, J. Electroanal. Chem. 1989, 265, 297.
10, 73. 151. C. Beriet, D. Pletcher, J. Electroanal. Chem.
127. E. F. Dalton, N. A. Surridge, J. C. Jernigan 1994, 375, 213.
152. K. Winkler, J. Electroanal. Chem. 1995, 388,
et al., Chem. Phys. 1990, 141, 143.
151.
128. J. A. Cox, K. S. Alber, C. A. Brockway et al.,
153. D. O. Wipf, M. R. Wightman, Anal. Chem.
Anal. Chem. 1995, 67, 993.
1988, 60, 2460.
129. C. E. D. Chidsey, R. W. Murray, Science
154. R. A. Marcus, J. Chem. Phys. 1965, 43, 679.
1986, 231, 25. 155. J. O. Howell, R. M. Wightman, Anal. Chem.
130. P. J. Kulesza, L. R. Faulkner, J. Chen et al., 1984, 56, 524.
J. Am. Chem. Soc. 1991, 113, 379. 156. C. E. D. Chidsey, Science 1991, 251, 919.
131. P. J. Kulesza, L. R. Faulkner, J. Am. Chem. 157. D. Acevedo, H. D. Abruña, J. Phys. Chem.
Soc. 1994, 115, 11 878. 1991, 95, 9590.
132. P. J. Kulesza, B. Karwowska, B. Grzybowska 158. R. J. Forster, L. R. Faulkner, J. Am. Chem.
et al., Electrochim. Acta 1998, 44, 1295. Soc. 1994, 116, 5444.
133. A. M. Bond, F. Scholtz, J. Phys. Chem. 1991, 159. H. O. Finklea, D. D. Hanshew, J. Am.
95, 7640. Chem. Soc. 1992, 114, 3173.
134. S. Fletcher, C. S. Halliday, D. Gates et al., J. 160. R. J. Forster, T. E. Keyes, J. Phys. Chem. B
Electroanal. Chem. 1983, 159, 267. 1998, 102, 10 004.
161. C. A. Amatore, A. Jutand, F. Pflüger, J. 180. M. Fleischmann, F. Lassere, J. Robinson

Electroanal. Chem. 1987, 218, 361. et al., J. Electroanal. Chem. 1984, 177, 97.
162. J. O. Howell, R. M. Wightman, J. Phys. 181. S. Dong, G. Che, Electrochim. Acta 1992, 37,
Chem. 1984, 88, 3915. 2701.
163. P. Hapiot, J. Moiroux, J. M. Savéant, J. Am. 182. S. Daniele, P. Ugo, G. A. Mazzocchin, J.
Chem. Soc. 1990, 112, 1337. Electroanal. Chem. 1989, 267, 129.
164. A. Russell, K. Repka, T. Dibble et al., Anal. 183. R. Muller, L. Lamberts, M. Evers, J. Elec-
Chem. 1986, 58, 2961. troanal. Chem. 1996, 417, 35.
165. C. G. Zoski, D. A. Sweigart, N. J. Stone 184. R. Muller, L. Lamberts, M. Evers, J. Elec-
et al., J. Am. Chem. Soc. 1988, 110, 2109. troanal. Chem. 1996, 401, 183.
166. F. N. Büchi, A. M. Bond, J. Electroanal. 185. K. B. Oldham, J. Electroanal. Chem. 1991,
Chem. 1991, 314, 191. 313, 3.
167. A. Owlia, J. F. Rusling, Electroanalysis 1989, 186. A. Jaworski, J. G. Osteryoung, M. Donten
1, 141. et al., Anal. Chem. 1999, 71, 3853.
168. A. M. Bond, T. L. E. Henderson, D. R. 187. P. T. Lillehei, L. A. Bottomley, Anal. Chem.
Mann et al., Anal. Chem. 1988, 60, 1878. 2000, 72, 189R.
169. J. Daschbach, D. Blackwood, J. W. Pons 188. M. V. Mirkin, B. R. Horrocks, Anal. Chim.
et al., J. Electroanal. Chem. 1987, 237, 269. Acta 2000, 406, 119.
170. Y. Zhang, C. D. Baer, C. Camaioni-Neto 189. A. Ashkin, J. M. Dziedzic, J. E. Bjorkholm
et al., Inorg. Chem. 1991, 30, 1682. et al., Opt. Lett. 1986, 11, 288.
171. K. B. Oldham, C. G. Zoski, A. M. Bond, J. 190. N. B. Simpson, K. Dholakia, L. Allen et al.,
Electroanal. Chem. 1988, 248, 467. Opt. Lett. 1997, 22, 52.
172. B. R. Scharifker, P. Zelenay, J. O. ’M. Bock- 191. T. B. Jones, Electromechanics of Particles,
ris, J. Electrochem. Soc. 1987, 134, 2714. Cambridge University Press, Cambridge,
173. A. J. Bard, M. V. Mirkin, P. R. Unwin et al., UK, 1995.
J. Phys. Chem. 1992, 96, 1861. 192. U. Zimmermann, G. A. Neil, Electromanip-
174. M. Fleischmann, F. Lassere, J. Robinson, J. ulation of Cells, CRC Press, Boca Raton, Fla.,
Electroanal. Chem. 1984, 177, 115. 1996.
175. S. Dong, G. Che, J. Electroanal. Chem. 1991, 193. M. Washizu, S. Suzuki, O. Kurosawa et al.,
315, 191. IEEE Trans. Ind. Appl. 1994, 30, 835.
176. T. V. Shea, A. J. Bard, Anal. Chem. 1987, 59, 194. T. Müller, A. Gerardino, T. Schnelle et al.,
2101. J. Phys. D: Appl. Phys. 1996, 29, 340.
177. C. L. Miaw, J. F. Rusling, A. Owlia, Anal. 195. M. P. Hughes, Phys. Med. Biol. 1998, 12,
Chem. 1990, 62, 268. 3639.
178. S. U. Pedersen, K. Daasbjerg, Acta Chem. 196. M. P. Hughes, H. Morgan, J. Phys. D: Appl.
Scand. 1989, 43, 301. Phys. 1998, 31, 2205.
179. K. Daasbjerg, S. U. Pedersen, H. Lund, Acta
Chem. Scand. 1989, 43, 876.
2.6 electrodes to a condition far from equi-

Impedance Methods librium. In contrast, impedance methods
are based on perturbation of the electro-
Steffi Krause chemical cell with an alternating signal of
University of Sheffield, Sheffield, United small magnitude allowing measurements
Kingdom at equilibrium or steady state. The pertur-
bation can be of a wide range of parameters
such as the applied potential, the applied
2.6.1 current, the convection rate at hydrody-
Introduction namic electrodes, or light intensity. The
main advantage of these techniques is that
Impedance spectroscopy is a powerful the response can be regarded as approxi-
technique for investigating electrochem- mately linear as long as the perturbations
ical systems and processes. Its main are sufficiently small.
strength lies in its ability to interro-
gate relaxation phenomena, whose time 2.6.2
constants range over several orders of Impedance – A Complex Quantity
magnitude. In contrast to other electro-
chemical techniques, it is noninvasive and 2.6.2.1 Definition
can be used for investigating bulk as well as The perturbations applied to an electro-
interfacial processes connected with time chemical system are usually sinusoidal.
constants ranging from minutes down to For simplification, we will only consider
microseconds. sinusoidal perturbations of the applied
There is a vast amount of literature on voltage
the subject of impedance measurements V (t) = V0 sin ωt (1)
comprising a large number of different
where V (t) is the voltage at time t, V0 is
applications, such as corrosion, characteri-
the voltage amplitude, and ω is the radial
zation of thin films and coatings, batteries,
frequency (in rad s−1 ). The relationship
semiconductor electrodes, sensors, bio-
between the radial frequency and the
logical systems, and many more. It is
frequency f (in Hz) is ω = 2πf . The
beyond the scope of this article to cover
current response I (t) will be a sinusoid
all of these applications comprehensively.
at the same frequency but shifted in phase
This chapter, therefore, concentrates on
the description of the main principles I (t) = I0 sin(ωt + ) (2)
and theories and selected applications of
where I (t) is the current at time t, I0 is
impedance methods. A more thorough the current amplitude, and is the phase
treatment of the subject from the point shift by which the voltage lags the current
of view of corrosion can be found in [1, 2], (see Fig. 1). Analogous to Ohm’s law for a
impedance spectroscopy of solid systems dc circuit, the impedance is defined as the
is described in [3]. The fundamentals of ratio of voltage and current
impedance spectroscopy of electrochemi-
cal systems are also explained in [4, 5]. V (t)
Z= (3)
Linear sweep or potential step tech- I (t)
niques described in previous chapters The impedance has a magnitude (Z0 =
(see Chapters 2.1 and 2.2) usually drive V0 /I0 ) and a phase () and is thus a
2.6 Impedance Methods 197
Fig. 1 Sinusoidal voltage

perturbation and current V
response.
I
V or I
t
Φ/w
Fig. 2 Impedance presentation

in the complex plane. Z and Z
are the imaginary and real parts
of the impedance, Z0 is the
magnitude of the impedance,
and is the phase angle.
Z0
Z"
Z'
vector quantity. It is therefore convenient This is shown in a plot of the imaginary

to present impedance in complex notation part versus the real part of the impedance
(see Fig. 2) (Fig. 3a), which is also called a complex
plane or Nyquist plot.
Z = Z0 (cos + j sin ) = Z + j Z
√ (4)
2.6.2.2.2 Capacitance If a sinusoidal vol-
where j = (−1), Z is the real part and
tage is applied across a pure capacitor, the
Z the imaginary part of the impedance.
impedance can be calculated according to
Using Euler’s relationship, we can
the relationship
also write
Z = Z0 ej (5) 1 j
Z= =− (6)
j ωC ωC
The complex plane diagram presented in
Fig. 2 shows the interrelationship between where C is the capacitance. The magnitude
the two different forms of presentation. of the impedance for a pure capacitor is
2.6.2.2 Impedance Response of Selected 1

Circuit Elements and Their Combinations Z0 = (7)
ωC
2.6.2.2.1 Resistance If a sinusoidal volt- and the phase angle is = −π/2, that is,
age is applied to a pure resistor of value R, the impedance depends on the frequency
then Z0 = R and = 0 for all frequencies. and is entirely imaginary (Fig. 3b).
R C
−Z"
−Z"
Z' Z'
(a) (b)
R C
R
−Z"
−Z"
Z' Z'
(c) (d)
Fig. 3 Nyquist plots for different equivalent circuits.
2.6.2.2.3 Resistance and Capacitance in and the phase angle is

Series For the serial combination of a
resistor and capacitor, the result is shown
1
in Fig. 3(c). According to Kirchhoff’s law, = arctan − (11)
RωC
the currents through both elements are the
same. The total potential equals the sum
2.6.2.2.4 Resistance and Capacitance in
of the potentials across the capacitor and
Parallel The impedance for a resistor
resistor
and a capacitor in parallel shows the
j I (t) shape of a semicircle in the complex
V (t) = I (t)R − (8)
ωC plane diagram (Fig. 3d). According to
and the total impedance can be calculated Kirchhoff’s law for a parallel circuit, the
according to potentials across both circuit elements
are equal, while the total current can
j be calculated from the sum of the
Z=R− (9)
ωC currents flowing through resistor and
capacitor
The magnitude of the impedance is

1 V (t) ωCV (t)
Z0 = R 2 + 2 2 (10) I (t) = − (12)
ω C R j
For the impedance, we can write electrochemical double layer that can be
expressed as the double-layer capacitance
1 1 ωC 1 ωC −1 Cdl , and finally the impedance of the
= − Z= −
Z R j R j charge-transfer process also called faradaic
(13)
impedance. The equivalent circuit consists
The magnitude of the impedance is
of a parallel combination of double-layer
−1/2
1 2 2
capacitance and faradaic impedance Zf
Z0 = + ω C (14) with the electrolyte resistance in series
R2
(Fig. 4a). In the absence of any elec-
and the phase angle is troactive species, the faradaic impedance
= arctan(−RωC) (15) becomes infinitely large, and the equiv-
alent circuit in Fig. 4(a) simplifies to a
2.6.2.3 Equivalent Circuit Elements and serial combination of electrolyte resistance
the Electrochemical Cell and double-layer capacitance. In the pres-
Many researchers take the view that the ence of electroactive species, the faradaic
transfer function for a given system should impedance can often be described by a
be derived from the equations governing simple resistor Rct (Fig. 4b).
the kinetics of the electrochemical reac- AC-impedance measurements at a se-
tions involved. This will be demonstrated ries of different frequencies can be used to
for a simple charge-transfer reaction in identify and separate the different circuit
Sect. 2.6.3. A second method for modeling components. This can be done graphically
electrochemical processes involves the use
of networks of electrical circuit elements,
so-called equivalent circuits, which can be Cdl
selected on the basis of an intuitive under-
standing of the electrochemical system.
It has been shown many times that for Re
simple systems, equivalent circuits can be (a) Zf
used to derive useful information from
impedance spectra as long as they are Cdl
based on the physical and chemical prop-
erties of the system and do not contain
arbitrarily chosen circuit elements. Re
In order to express a simple three- Rct
(b)
electrode electrochemical cell in terms of
an equivalent circuit, at least three quanti-
ties need to be taken into account. One High Low
frequencies frequencies
of them is the resistance of the elec-
−Z"
trolyte solution between the reference and

the working electrodes Re , another is the
Fig. 4 Equivalent circuits [(a) general

electrochemical cell; (b) simple
electrochemical cell without diffusion];
Re Rct
and (c) Nyquist plot for a simple
electrochemical cell without diffusion. (c) Z'
from the Nyquist plot (Fig. 4c). At high reaction. Rct can be related to the exchange
frequencies, the impedance of the double- current of a reversible electrode reaction as
layer capacitance is very small (Eq. 6) and well as to the corrosion rate in a steady state
the charge-transfer resistance is effectively system. Frequently however, the behavior
shorted out, that is, the intercept of the of electrochemical systems is governed by
semicircle with the real axis at high fre- more complicated reaction mechanisms.
quencies represents the resistance of the Particularly at lower frequencies, diffusion
electrolyte solution. With decreasing fre- and adsorption processes are picked up by
quency, the impedance of the double-layer ac-impedance measurements that require
capacitance becomes greater and the im- the derivation of more complex models.
pact of the charge-transfer resistance on Examples for this are given in Sect. 2.6.3.
the total impedance increases. At suffi-
ciently low frequencies, the impedance of 2.6.2.4 Forms of Data Presentation
the double-layer capacitance becomes sig- As described in Sect. 2.6.2.1, impedance
nificantly larger than the charge-transfer is a vector quantity and as such can be
resistance, and the impedance spectrum expressed either in terms of real and
is dominated by electrolyte and charge- imaginary parts or in terms of magni-
transfer resistance (low-frequency inter- tude and phase. From these two forms
cept of the semicircle with the real axis, of mathematical expression, the two most
Fig. 4c). The diameter of the semicircle is common forms of data presentation can
given by the charge-transfer resistance Rct . be derived. The complex plane or Nyquist
The frequency at the top of the semicircle, diagrams, which plot the real versus the
where the imaginary part of the impedance imaginary part of the impedance, have al-
reaches its maximum, is ready been shown for a variety of different
1 equivalent circuits in the previous para-
ωmax = (16) graphs (see Figs. 3 and 4c). Nyquist plots
Rct Cdl
have proven very useful for estimating
From this, the time constant τ for the impedance parameters from impedance
faradaic process can be defined as spectra appearing as single or multiple arcs
in the complex plane as explained above.
τ = Rct Cdl (17) However, the information they present
is not complete since there is no indi-
A more accurate method of determining cation of the frequency the impedance was
values for the equivalent circuit elements measured at. For the presentation of ex-
is to use one of the many commercial perimental data, the points in the Nyquist
software packages for fitting ac-impedance plot must be labeled with the frequencies
data to user-defined equivalent circuits. they correspond to.
Most fitting programs require very accu- Another form of data presentation is the
rate guesses of the initial parameters, that Bode plot. Most commonly, it shows a plot
is, the graphical approach is still very use- of the phase angle and the logarithm of
ful for initial estimates of equivalent circuit the magnitude of the impedance versus
parameters. the logarithm of the frequency (Fig. 5a).
The simple approach described above When resistive behavior dominates the
can be used for estimating the double-layer impedance behavior, a horizontal line is
capacitance and the rate of the electrode observed in the presentation of log Z0
Fig. 5 Bode plots for the equivalent

circuit shown in Fig. 4(b).
log Z0 ( )
versus log f . When the impedance re-
Φ( )
sponse is determined by a capacitance, a
straight line with a slope of – 1 is observed
(Fig. 5a) Sometimes alternative Bode plots
are presented, which show log Z and log
Z versus log f (Fig. 5b). In contrast to log f
(a)
Nyquist plots, both types of Bode plot in-
clude all of the information obtained by
ac-impedance measurements.
log Z" ( )
log Z' ( )
2.6.3
Description of Electrochemical Processes
2.6.3.1 Faradaic Processes and

Diffusion-related Phenomena
It was shown that an electrochemical (b) log f
cell could be described with a simple
equivalent circuit (Fig. 4a) containing the mass-transfer impedance ZW (Fig. 6b),
electrolyte resistance, the double-layer ca- also called the Warburg impedance.
pacitance, and the impedance of the For the circuit shown in Fig. 6(a), the
faradaic process. Previously, the faradaic total voltage drop is
process was described in terms of a simple
Q
charge-transfer resistance (Sect. 2.6.2.3) V (t) = I (t)Rs + (18)
neglecting the diffusion of electroac- Cs
tive species. where Q is the charge.
For the interpretation of the faradaic After differentiation, it can be writ-
impedance in the presence of diffusion- ten that
related phenomena, it is convenient to
dV (t) dI (t) I (t)
subdivide it into two circuit elements. This = Rs + (19)
can be done in two different ways. Zf can dt dt Cs
be presented by a resistance Rs in series For a sinusoidal perturbation of the current
with a pseudocapacitance Cs according according to
to Fig. 6(a), or it can be subdivided into
the charge-transfer resistance and the I (t) = I0 sin ωt (20)
(a)
Rs CS
Fig. 6 Two possible

presentations of the faradaic
impedance Zf as an equivalent R ct
(b) ZW
circuit [5].
Eq. (19) changes to Reaction (22) and the amplitude of the ac

perturbation is small, Eq. (25) can be lin-
dV (t) I0
= Rs I0 ω cos ωt + sin ωt (21) earized using the Taylor series expansion
dt Cs
RT cO (0, t) cR (0, t) I (t)
For the simple charge-transfer reaction η= ∗ − +
nF cO cR∗ i0
O + ne− −
←−−
−−
→−R (22) (26)
The first partial differential in Eq. (24)
where O and R are both soluble, it can be ∂V /∂I is the charge-transfer resistance
written that and can be calculated to
∂V (t) RT
V (t) = f [I (t), cO (0, t), cR (0, t)] (23) = Rct = (27)
∂I (t) nF i0
where cO (0,t) and cR (0,t) are the surface
The other partial differentials can be
concentrations of the oxidized and the
derived from Eq. (26) to
reduced species at times t.
Hence, ∂V (t) RT
= ∗
dV (t) ∂V (t) dI (t) ∂cO (0, t) nF cO
=
dt ∂I (t) dt ∂V (t) RT
=− (28)
∂V (t) dcO (0, t) ∂cR (0, t) nF cR∗
+
∂cO (0, t) dt The other three factors have to be evalu-
ated for the current–time relationship in
∂V (t) dcR (0, t)
+ (24) Eq. (20)
∂cR (0, t) dt
In order to obtain an analytic expression dI (t)
= I0 ω cos ωt (29)
for dV /dt, the six factors on the right-hand dt
side of Eq. (24) have to be evaluated. The Expressions for the other factors are de-
three partial differentials are determined rived considering mass transfer assum-
by the kinetics of the electrode reaction ing semi-infinite linear diffusion with
and can therefore be derived from the the initial conditions cO (x, 0) = cO ∗ and
current-overpotential equation ∗
cR (x, 0) = cR . Details of this derivation can
I (t) cO (0, t) − αnF η be found in [5], Chapter 9.2.
=− ∗ e RT The derivatives of the surface concentra-
i0 cO
tions are
cR (0, t) − (1−α)nF η
− e RT (25) dcO (0, t) I0 ω 1/2
cR∗ =
dt nFA 2DO
where cO ∗ and c∗ are the bulk concentra-
R × (sin ωt + cos ωt) (30)
tions of oxidized and reduced species, η is
the overpotential, i0 is the exchange cur- dcR (0, t) I0 ω 1/2
=−
rent, F is the Faraday constant, R is the dt nFA 2DR
gas constant, T is the temperature, and α × (sin ωt + cos ωt) (31)
is the transfer coefficient.
Provided that the time averaged po- where DO and DR are the diffusion coeffi-
tential is the equilibrium potential for cients of oxidized and reduced species.
Equation (24) then becomes and the imaginary part of the Warburg
impedance results from Cs
dV (t) σ
= Rct + 1/2 I0 ω cos ωt
dt ω 1 σ
1/2
ZW = − = − 1/2 (37)
+ I0 σ ω sin ωt (32) ωCs ω
where that is,

1 1 σ σ
σ =
RT
√ + ZW = 1/2
− j 1/2 (38)
1/2 ∗ ω ω
DR cR∗
1/2
n2 F 2 A 2 D O cO
(33) Since real and imaginary parts of the
By comparison of Eqs. (21 and 32), Warburg impedance ZW have the same

we obtain value, a plot of ZW versus ZW would
σ show a straight line with a phase an-
Rs = Rct + (34)
ω1/2 gle of 45◦ (Fig. 9a). The impedance of an
electrochemical cell in which the faradaic
and
1 impedance in Fig. 4(a) is expressed as
= σ ω1/2 (35) a serial combination of charge-transfer
Cs
resistance and Warburg impedance (see
From Eqs. (34 and 35), the Warburg Fig. 6b) is presented in Fig. 7. The result-
impedance can be identified. A comparing equivalent circuit is also known as
ison with the two forms of presentation of Randles circuit. The semicircle is caused
the faradaic impedance in Figs. 6(a and b) by a charge-transfer-controlled reaction
shows that the second term in Eq. (34) is and can be interpreted as discussed in
identical to the real part of the Warburg Sect. 2.6.2.3. The straight line with a slope
impedance of 1 is due to the Warburg impedance
σ and indicates a purely diffusion-controlled
ZW = (36) reaction at the low-frequency limit.
ω1/2
−Z"
Z'
Cdl
Fig. 7 Randles circuit and
Nyquist plot for reversible
electrochemical system with Re
diffusion-limited behavior at low R ct ZW
frequencies.
The analysis above has shown that it is electrolyte solution, which would cause
possible to derive kinetic parameters from the concentrations of solution species to
the impedance spectra of a redox system. be constant at a certain distance from the
The charge-transfer resistance is directly electrode surface (c(δ, t) = c∗ for all t, in
related to the exchange current (Eq. 27) which c* is the bulk concentration). In
and a medium diffusion coefficient for this case, the impedance is given by the
oxidized and reduced species can be expression
calculated from the coefficient σ of the √
Warburg impedance (Eq. 33). finite tanh(δ j ω/D)
ZW = R0 √ (39)
Because of the assumption of semi- δ j ω/D
infinite diffusion made by Warburg for
the derivation of the diffusion impedance, where δ is the diffusion layer thickness,
it predicts that the impedance diverges R0 is the diffusion resistance for ω → 0,
from the real axis at low frequencies, and D is the diffusion coefficient of
that is, according to the above analysis, the diffusing particle. This behavior is
the dc-impedance of the electrochemical analogous to a finite length transmission
cell would be infinitely large. It can be line terminated with a resistance. In
shown that the Warburg impedance is contrast to the Warburg impedance for
analogous to a semi-infinite transmission semi-infinite diffusion, the finite Warburg
line composed of capacitors and resistors impedance as described in Eq. (39) will
(Fig. 8) [3]. However, in many practical bend over to the real axis at low frequencies
cases, a finite diffusion layer thickness giving rise to a distorted semicircle
has to be taken into consideration. The (Fig. 9b). This concurs with the fact that
first case to be considered is that of electrochemical cells usually have a finite
enforced or natural convection in an dc-resistance.
Fig. 8 Infinite length transmission line that describes the behavior of

semi-infinite diffusion ([3] Chapter 2.1).
−Z"
−Z"
−Z"
Z' Z' Z'

(a) (b) (c)
Fig. 9 Impedance spectra for diffusion-limited behavior (a) semi-infinite
diffusion; (b) finite diffusion with unhindered ion transfer at the far end of
the diffusion region; and (c) finite diffusion with blocked ion transfer at the
far end of the diffusion layer.
A finite length diffusion layer thickness causality can arise when the response
cannot only be caused by constant con- is not caused by the input but for ex-
centrations of species in the bulk of the ample, by a concentration, current, or
solution but also by a reflective boundary, potential relaxation upon departure of the
that is, a boundary that cannot be pene- system from equilibrium. Causality can
trated by electroactive species (dc/dx = 0). also be disturbed as result of instrument
This can happen when blocking occurs at artifacts or noise. The stability of an electro-
the far end of the diffusion region and no chemical system is usually not guaranteed
dc current can flow through the system, when continually changing systems such
for example, a thin film of a conducting as corroding electrodes or batteries are
polymer sandwiched between a metal and investigated.
an electrolyte solution [6]. The impedance To assess whether meaningful data that
in this case can be described with the fulfill the above conditions have been
expression obtained, a range of techniques and
√ diagnostic tools have been developed to
finite δ 2 coth(δ j ω/D) validate impedance data. The problem of
ZD = √ (40)
DC0 δ j ω/D nonlinearity can generally be overcome by
making the amplitude of the perturbation
where C0 is the limiting capacitance for signal small enough to approach quasi-
ω → 0. This behavior is analogous to a linear conditions. This can be achieved by
finite length transmission line terminated measuring impedance spectra at different
with an open circuit, that is, at low perturbation amplitudes. The amplitude
frequencies capacitive behavior is observed can be regarded sufficiently small when it
(Fig. 9c). has no significant effect on the impedance
In this section, the behavior of a redox spectrum measured. In electrochemical
system at the equilibrium potential has systems, nonlinearity mainly affects the
been discussed. It should, however, be low-frequency part of the spectrum, which
noted that impedance spectroscopy of is determined by the faradaic current.
irreversible systems can also yield useful The high-frequency part of the spectrum
information. For example, the charge- determined by the electrolyte resistance
transfer resistance determined at the and the double-layer capacitance shows
corrosion potential corresponds to the approximately linear behavior [7].
slope of the current–potential curve (Rct = Depending on the technique used for
dV (t)/dI (t)) at that potential and allows data acquisition, different methods for data
calculation of the rate of corrosion [1]. validation can be applied. If a sinusoidal
signal were applied to a nonlinear system,
2.6.3.2 Validation of Experimental Data the response function would contain
Impedance is only properly defined as a multiples (harmonics) of the excitation
transfer function when the system un- signal. Popkirov and Schindler utilized
der investigation fulfills the conditions this effect to develop a technique for
of linearity, causality, and stability dur- assessing the linearity of an impedance
ing the measurement. However, as the measurement obtained using a time-
current voltage characteristic of any elec- domain technique (see Sect. 2.6.7.2) [8].
trode shows (see Eq. 25), electrochemical They compared the frequencies contained
systems are nonlinear. Deviation from in the perturbation and the response
signals by performing a Fast Fourier The imaginary part of the impedance

Transform (FFT) on the response signal. can be calculated from the real part of the
When the system behaves nonlinearly, impedance from
harmonics of the frequencies contained
2ω ∞ Z (x) − Z (ω)
in the perturbation signal appear in the Z (ω) = − dx
response spectrum. π 0 x 2 − ω2
(41)
Erroneous impedance data cannot only
If the high-frequency limit of the real part
be caused by nonlinearity but very fre-
of the impedance is known, the real part of
quently by a lack of stability of the system
the impedance can be obtained from the
under investigation. The easiest way to
imaginary part of the impedance using
establish whether a system has changed
during the course of an impedance mea- Z (ω) = Z (∞)
surement is to repeat the experiment and
2 ∞ xZ (x) − ωZ (ω)
compare both sets of data. For most ap- + dx
plications, instabilities could and should π 0 x 2 − ω2
be eliminated by adjusting the measuring (42)
conditions accordingly. However, it may If the zero frequency limit of the real part
also be desirable to characterize systems, of the impedance is known, the real part
which are known to change over time, or of the impedance can be obtained at any
even to measure changes of the complex frequency from the imaginary part using
impedance in real time [9, 10]. As long as 2ω
the measurement time is short compared Z (ω) = Z (0) +
π
to the time constant of relaxation of the
∞
system, valid impedance data can be ob- x/ωZ (x) − ωZ (ω)
× dx
tained. Since the lower the measurement 0 x 2 − ω2
frequency, the greater the measurement (43)
time, it is generally accepted that high- The relationship between phase angle and
frequency data can be valid even when modulus of the impedance is
low-frequency data are affected by non-
2ω ∞ ln Z0 (x)
stationary behavior of the electrochemical (ω) = dx (44)
π 0 x 2 − ω2
system.
Kramers–Kronig relations have been To assess whether experimental data fulfill
used by a large number of scientists as the Kramers–Kronig relations, one part of
a diagnostic tool for the validation of the impedance is calculated from the other
impedance data to establish which fre- part of the impedance, which has been
quency range contains consistent data [2]. experimentally determined. For example,
Originally developed for optical applica- the imaginary part can be calculated from
tions, they can also be applied to elec- the measured real part of the impedance
trochemical impedance spectroscopy. The using Eq. (41). The calculated imaginary
Kramers–Kronig relations are a series of part of the impedance can then be
integral equations that govern the relation- compared with the measured imaginary
ship between the real and imaginary parts part of the impedance. Any difference
of complex quantities for systems fulfilling between calculated and measured values
the conditions of linearity, causality, and would indicate a deviation from one of
stability. the conditions of linearity, causality, or
stability. This method, however, requires the form

that at least one part of the impedance
k
is known between the frequency limits of Z(ω) = Z0 + (45)
zero and infinity. (1 + j τk ω)
k
In most cases, the frequency range is
not sufficiently large to integrate over It is equivalent to a series of a number of
the frequency limits zero and infin- RC parallel combinations in series with the
ity. To overcome this problem, various solution resistance, which is also known
models for extrapolating the experimen- as the Voigt model. If a sufficient number
tal data sets to the frequency limits of terms is included, the model should be
have been suggested. These methods able to fit impedance data of any stationary
were reviewed by Agarwal and cowork- electrochemical cell. Since the model
ers in [2] and will only be summarized is consistent with the Kramers–Kronig
in this chapter. Kendig and coworkers relations, any failure to fit experimental
suggested extrapolating impedance data data using this model indicates a violation
to low frequencies assuming that the of the Kramers–Kronig relations.
imaginary impedance is symmetrical (see
for example, Fig. 5b) [11]. However, this 2.6.4
can only be applied to simple systems Distributed Circuit Elements
with one time constant. Macdonald and
coworkers extrapolated polynomial fits to Impedance data are frequently fitted with
the data into the unmeasured frequency an equivalent circuit made up of circuit
range [12]. Esteban and coworkers calcu- elements, which are related to the physical
lated impedance data below the lowest processes in the system under investiga-
measured frequency by using Eqs. 41 tion. In many cases, ideal circuit elements
and 42 (or 43) simultaneously forcing such as resistors and capacitors can be
the measured data set to satisfy the applied. Mostly, however, distributed cir-
Kramers–Kronig relations [13]. Any dis- cuit elements are required in addition
continuity between the measured and the to the ideal circuit elements to describe
calculated data set would then indicate the impedance response of real systems
erroneous impedance data. adequately. Various distributed circuit ele-
A method that does not require ex- ments and their applications are discussed
trapolating data to frequency ranges not in [3, 15].
accessible by experiment is the fit of the Historically, the Warburg impedance,
data with an equivalent circuit. Equiva- which models semi-infinite diffusion of
lent circuits consisting of passive circuit electroactive species, was the first dis-
elements satisfy the Kramers–Kronig re- tributed circuit element introduced to
lations. Any successful fit of such an describe the behavior of an electrochemical
equivalent circuit to a data set would cell. As described above (see Sect. 2.6.3.1),
confirm that the data is consistent with the Warburg impedance (Eq. 38) is also
the Kramers–Kronig relations. A general- analogous to a uniform, semi-infinite
ization of this approach, which suggests transmission line. In order to take account
the use of a so-called measurement model, of the finite character of a real electrochem-
was described by Agarwal and cowork- ical cell, which causes deviations from the
ers [14]. The measurement model is of Warburg impedance at low frequencies,
the finite length Warburg impedance was it is only valid over a limited frequency
introduced (Eq. 39). range. In any physically valid model, there
In all real systems, some deviation from should be a shortest and a longest relax-
ideal behavior can be observed. If a po- ation time. This is, however not true for the
tential is applied to a macroscopic system, CPE. Therefore, the CPE usually needs to
the total current is the sum of a large be modified at both ends of the frequency
number of microscopic current filaments, spectrum. In many cases, this does not
which originate and end at the electrodes. pose a problem, since impedance mea-
If the electrode surfaces are rough or surements are not carried out at extreme
one or more of the dielectric materials frequencies.
in the system are inhomogeneous, then all Other empirical distributed elements
these microscopic current filaments would have been described, which can be ex-
be different. In a response to a small- pressed as a combination of a CPE and one
amplitude excitation signal, this would or more ideal circuit elements. Cole and
lead to frequency-dependent effects that Cole found that frequency dispersion in
can often be modeled with simple dis- dielectrics results in an arc in the complex
tributed circuit elements. One of these ε plane (an alternative form of presenta-
elements, which have found widespread tion) with its center below the real axis
use in the modeling of impedance spec- (Fig. 10a) [16]. They suggested the equiv-
tra, is the so-called constant phase element alent circuit shown in Fig. 10(b), which
(CPE). A CPE is defined as includes a CPE and two capacitors. For
ω → 0, the model yields capacitance C0
ZCPE = A0 (j ω)−α (46)
and for ω → ∞ the model yields capac-
where A0 and α are frequency-independent itance C∞ . The model can be expressed
parameters and 0 ≤ α ≤ 1. The name con- with the following empirical formula for
stant phase element stems from the fact the complex dielectric constant
that the phase angle is frequency indepen-
ε0 − ε∞
dent. In the complex plane diagram, a CPE ε = ε∞ + (47)
1 + (j ωτ0 )1−αε
would appear as a straight line at a con-
stant phase angle of = −(α × 90◦ ). For where ε∞ and ε0 are the dielectric
α = 0, the CPE describes an ideal resistor, constants for ω → ∞ and ω → 0, and τ0
and for α = 1, it describes an ideal capac- is a generalized relaxation time. The CPE
itor. The CPE has been used to describe parameter A0 (Eq. 46) is correlated to τ0
different processes. For α = 0.5, the CPE through
represents homogeneous diffusion, that is,
Eqs. (38 and 46) are mathematically equiv- τ0 ≡ [A0 (C0 − C∞ )]1/1−α (48)
alent for α = 0.5. For other values of α,
the CPE can be used to describe inhomo- Analogous behavior has been described
geneous diffusion. for conducting systems, that is, systems
The CPE has also been interpreted as a that can pass a direct current. In this
distribution of relaxation times due to in- case, a semicircle with its center be-
homogeneities in dielectric materials, that low the real axis can be found in the
is, materials that can pass an alternating complex impedance plane (Fig. 11a). The
current but not a dc current. The main equivalent circuit (Fig. 11b) contains a
problem associated with the CPE is that combination of a CPE and two resistors,
Fig. 10 (a) Cole–Cole plot

showing a depressed arc in the
complex ε plane and
(b) equivalent circuit describing
the frequency dispersion in
e"
dielectrics [16].
(a) e'
C 0 − C∞ CPE
(b) C∞
Fig. 11 (a) Frequency dispersion in

electrochemical systems leads to a
depressed semicircle in the complex
impedance plane (shown for different
−Z"
values of the exponent α of the CPE

(Eq. 46), α = 0.9, α = 0.8, ◦α =
0.7), (b) equivalent circuit describing
the frequency dispersion in
electrochemical systems [3].
(a) Z'
CPE
R∞
(b) R0 − R∞
R0 and R∞ , which represent the low and Impedance spectra of real electrochemical
high-frequency limits. The impedance of cells will usually not show a perfect semi-
this circuit can be expressed as circle as the one presented in Fig. 4(c)
R0 − R∞ but depressed semicircles as shown in
Z = R∞ + Fig. 11(a).
1 + (R0 − R∞ )A−10 (j ω)
α
(49) This behavior is observed when the
A comparison between the circuit shown surface of an electrode is not perfectly
in Fig. 11(b) and the equivalent circuit flat. In the simple equivalent circuits
of a simple electrochemical cell (Fig. 4b) for electrochemical cells discussed in
shows that they are analogous, with R∞ ≡ Sects. 2.6.2 and 2.6.3, the double-layer
Re , R0 − R∞ ≡ Rct , and Cdl ≡ CP E. capacitance (see Figs. 3c, 4b, and 7) is
therefore frequently replaced by a CPE, the exponent α of the CPE and the frac-
although the parameters α and A0 in tal dimension are contradictory, and there
Eq. (46) can often not be accurately is some controversy about the validity of
correlated to physical properties of the these theories since there is little experi-
system under investigation. A few special mental evidence to support them [19, 20].
cases in which the parameters of the CPE More recently, models have been sug-
can be derived directly from the electrode gested that are based on the assumption
morphology are discussed in Sect. 2.6.5. that the frequency dependence of capac-
For further information regarding dis- itance on solid electrodes is due to the
tributed circuit elements, refer to [3, 15]. atomic scale inhomogeneities rather than
the geometric aspects of surface rough-
2.6.5 ness [21].
The Effect of Electrode Morphology
2.6.5.1 Porous Electrodes
As indicated in Sect. 2.6.4, modeling of real
De Levie was the first to develop a transmis-
electrochemical systems usually requires
sion line model to describe the frequency
the aid of distributed circuit elements.
In this section, the relationship between dispersion in porous electrodes [22]. Here,
the morphological properties of rough or only the effect of double-layer charging
porous electrodes and their impedance in the absence of a faradaic process will
behavior will be discussed. be discussed. De Levie’s model is based
As established by a large number of au- on the assumption that the pores are
thors, the impedance of solid electrodes cylindrical, of uniform diameter, and semi-
in the absence of faradaic processes usu- infinite length, not interconnected and
ally deviates from pure capacitive behavior. homogeneously filled with electrolyte. The
Only for smooth and clean surfaces like electrode material is assumed to have no
that of liquid mercury, a pure capacitance resistance. Under these conditions, a pore
can be used to describe the impedance behaves like a uniform RC transmission
behavior (Fig. 13, curve a). The devia- line as shown in Fig. 12. If a sinusoidal
tion from ideality has frequently been excitation is applied, the transmission line
attributed to surface roughness or poros- behavior causes the amplitude of the signal
ity. In the case of rough electrodes, the to decrease the further the distance from
impedance behavior is usually approxi- the opening of the pore, that is, only a frac-
mated using a CPE (see Eq. 46) with tion of the pore is effectively taking part in
0.5 < α < 1. Several attempts have been the double-layer charging process. From
made to derive a relationship between the that, a penetration depth λ of the signal
CPE and a fractal surface morphology [17, can be defined, which is inversely propor-
18]. However, the relationships between tional to the square root of frequency and
Fig. 12 Section of the

Rdz equivalent circuit of a pore, in
Cdz which z is the distance from the
pore entrance. Solution
resistance and double-layer
z z + dz capacitance are uniformly
distributed [22].
Fig. 13 Impedance behavior of

an electrode in the absence of
electroactive species for (a) a a b c
perfectly flat electrode; (b) a
porous electrode with uniformly
sized pores according to de
Levie’s model; and (c) an
−Z"
electrode with a pore size
distribution [23].
Z'
directly proportional to the square root of behavior since the electrode surface taking
the radius of the pores part in the double-layer charging process
is now frequency independent (Fig. 13,
1 κr
λ= (50) curve b).
2 Cdl ω Experimental results match the predic-
where r is the radius of a pore, κ is tions of de Levie’s model at high fre-
the electrolyte conductivity, and Cdl is the quencies. At low frequencies, however,
double-layer capacitance. deviations from ideal capacitive behavior
At very high frequencies, the penetration are frequently observed (Fig. 13, curve c).
depth is negligible and only the capaci- Song and coworkers suggested that the
tance effects of the flat external electrode nonideal behavior might be due to pore
surface are measured. At lower frequen- size distribution [23, 24]. This leads to
cies, the penetration depth and therefore a distribution of the penetration depth
the influence of the pores become pro- even at a single frequency since each
gressively more important since the inner pore would have a different penetration
surface of the electrode is very large com- depth depending on its dimensions. The
pared with the outer surface. Hence, a model developed by Song and coworkers
decrease of the frequency is accompanied integrates dispersion due to pore size dis-
by an increase in capacitance. The result tribution as well as in-the-pore dispersion
is a straight line in the complex plane di- described by de Levie. It successfully de-
agram at a phase angle of −45◦ , which is scribes the low-frequency branch of the
equivalent to a Warburg impedance or a impedance data for a porous electrode
CPE with α = 0.5. and can also be used to assess the aver-
If the model is extended to pores with age radius of pores, the width of the pore
finite and uniform length, the penetration size distribution, the average pore length,
depth will eventually reach the length and the surface area. Song and coworkers
of the pores when the frequency is also suggested that the model might al-
sufficiently low. A further decrease in low the use of impedance spectroscopy for
frequency would result in pure capacitive porosimetry.
2.6.6 the simple impedance behavior of the

Characterization of Coatings and electrolyte resistance in series with the
Membranes geometric capacitance of the polymer film
(Fig. 14b), which can be calculated accord-
In the previous sections, the impedance ing to
behavior of electrochemical cells was de- ε × ε0 × A
scribed, with a view of how kinetic pa- Cgeom = (51)
l
rameters of electrochemical reactions or
electrode properties such as the mor- where ε is the dielectric constant of the
phology might be extracted from an polymer, ε0 is the dielectric constant of
impedance spectrum. However, electro- vacuum, A is the surface area of the
chemical impedance spectroscopy has electrode, and l the thickness of the
been utilized for a vast number of applica- polymer film. Measured values of the
tions and is not limited to mechanistic in- geometric capacitance can be used to
vestigations of electrode reactions. A large calculate the dielectric constant or the
number of studies have been dedicated thickness of the polymer if one of both
to the investigation of coated electrode parameters is known.
surfaces. In this chapter, a few selected Impedance spectroscopy can be used to
examples will be given of how impedance distinguish between different processes of
spectra of coated electrodes can be evalu- degradation. A dissolution of a polymer
ated and what information can be gained film layer-by-layer would only lead to a
from them. reduction in the film thickness l in Eq. (51),
that is, the geometric capacitance of the
2.6.6.1 Insulating Coatings film would increase, but the principle
Insulating coatings have been of partic- shape of the impedance spectrum would
ular interest in corrosion protection. A stay the same (Fig. 14a) [28]. Frequently,
large number of papers deal with the however, the formation of pores is the
characterization of polymer layers and main reason for coating failure. This
their degradation [25–27]. An electrode behavior can be approximated with the
with a perfectly insulating coating shows equivalent circuit shown in Fig. 15(b) [25],
log Z0( )
Φ( )
Fig. 14 (a) Change of impedance

(a) log f during degradation of an insulating
coating layer-by-layer (simulated data,
solid symbols: before degradation, open
symbols: after partial degradation) and
Re (b) equivalent circuit of coated electrode
(b) Cgeom
before and after partial degradation.
Fig. 15 (a) Change of impedance

during degradation of an insulating
coating by pore formation (simulated
data, solid symbols: before degradation,
log Z0( )
open symbols: after partial degradation)
Φ( )
and (b) equivalent circuit of a partially
degraded film [25].
and the impedance spectrum changes

as shown in Fig. 15(a). If a coating
has electrolyte-filled pores that extend (a) log f
through the entire thickness of the coating,
then an apparent bulk resistance Rb can Cgeom
be observed, which corresponds to the
resistance of the electrolyte in the pores. Cdl
The ratio of the resistance Rbt of an
Re
electrolyte layer that occupies the same
Rb
space as the polymer layer to the bulk
(b) R ct
resistance Rb is equivalent to the fraction
of the film, which consists of electrolyte-
filled pores, and therefore to the porosity metal–polymer interface after penetrating
P of the coating the film. The relationship given in Eq. (53)
can only be regarded as a rough estimate of
Rbt the wetted metal area since the spreading
P = (52)
Rb of electrolyte along the electrode surface
leads to a nonuniform current distribution
Impedance spectra of partially degraded
that can best be described with a linear
coatings cannot only be used to esti-
transmission line [25].
mate the porosity of the film but also to
The degradation of coatings, although
estimate the area of the metal electrode
an unwanted effect in corrosion pro-
wetted by electrolyte. Since the inter-
tection, has been exploited for sen-
face metal–electrolyte is represented by
sor development. Disposable electrodes
the parallel combination of the charge-
were coated with biodegradable films.
transfer resistance and the double-layer
In the presence of either an enzyme
capacitance; both of these can be used
or the product of an enzymatic reac-
to estimate the fraction of the electrode
tion, the polymer degraded. The change
surface W wetted by electrolyte
of impedance observed during the en-
Cdl 0
Rct zyme catalyzed degradation of the poly-
W= 0
= (53) mer films has been used as a sensor
Cdl Rct
signal to detect analytes such as urea,
where Rct0 and C 0 are the charge-transfer creatinine, and a variety of different en-
dl
resistance and double-layer capacitance zymes [28–30].
of an uncoated electrode. Frequently, the
wetted surface area is considerable larger Noninsulating Films
2.6.6.2
than the porosity of the coating indicating There has been great interest in
that the electrolyte spreads along the the characterization of conducting
polymers using impedance spectroscopy. the low-frequency limit, capacitive be-

Johnson and coworkers described the havior can be observed indicating that
theoretical response of a metal- no dc current can flow through the
polythiophene-electrolyte system and a system.
metal-polythiophene-metal system [6]. For
the metal-conducting polymer-electrolyte The impedance spectrum of the metal-
arrangement, the following regions in conducting polymer-metal system was
the impedance spectrum were predicted predicted to show the following regions
(Fig. 16): (Fig. 17):
• A high-frequency semicircle corre- • The high-frequency semicircle is again

sponds to the bulk resistance Rb of caused by a parallel combination of
the polymer layer in parallel to its geo- the bulk resistance and geometric
metric capacitance. The bulk resistance capacitance of the film.
reflects the concentration and mobil- • Only one semicircle is predicted for the
ity of charge carriers in the conducting metal–polymer interfaces, since only
polymer. electron transfer can occur.
• Two charge-transfer semicircles are ex- • Since electron transfer occurs at both
pected, which correspond to two RC metal–polymer interfaces, a dc current
parallel combinations of the double- can flow through the system, that
layer capacitance and charge-transfer is, the low-frequency limit is not a
resistance at the electrode–polymer capacitance as in the case of the metal-
interface and the double-layer capac- conducting polymer-electrolyte system
itance and charge-transfer resistance but the impedance response bends over
at the polymer–electrolyte interface. to the real axis after the Warburg region.
At the metal–polymer interface, elec-
tron transfer would occur while at In principle, a large number of param-
the polymer–electrolyte interface anion eters such as conductivity, charge-transfer
transfer is expected. resistors, and diffusion coefficients can be
• At low frequencies, a diffusion-limited determined from the impedance spectra
region appears, which, if fitted to a discussed above. However, in practice, it
Warburg impedance, can be used to de- is only possible to resolve the different
termine diffusion coefficients of charge processes and their characteristic shapes
carriers in the conducting polymer. At in the impedance spectrum if their time
c
Z"
w
Fig. 16 Theoretical impedance
response of a metal-poly-
Rb Rct1 Rct2
thiophene-electrolyte
Z' structure [6].
Fig. 17 Theoretical impedance

response of a
metal-polythiophene-metal
structure [6].
Z"
w
Rb Rct
Z'
constants are sufficiently different. Fre- impedance and good current sensitivity.
quently, the time constants of interface and Further information analog and digital in-
bulk processes are not as clearly separated strumentation is given in Chapter 1.2.
as shown in Figs. 16 and 17, and mixed The techniques commonly utilized for
impedance behavior is observed making ac-impedance measurements in modern
an interpretation of the impedance spectra equipment can be subdivided into two
more difficult. main groups – single-sine and multiple-
Other examples for the characterization sine techniques. The lock-in technique
of films and membranes using impedance and frequency-response analysis will be
spectroscopy are the investigation of elec- described as representatives of the single-
trolyte uptake of polymer films [31–33] sine techniques and FFTs will be intro-
and the determination of exchange cur- duced as an example for multiple-sine
rents at ion-selective membranes for po- techniques.
tentiometric sensors [32, 34].
2.6.7 2.6.7.1 Single-sine Techniques

Instrumentation In single-sine techniques, a small-ampli-
tude sinusoidal signal with a fixed fre-
Impedance test equipment usually com- quency is applied to the test cell. The
prises an ac measurement unit and a po- response signal is then analyzed to extract
tentiostat or galvanostat. For many applica- the two components of the impedance (real
tions, such as biomedical investigations or and imaginary parts or magnitude and
the characterization of thin films in which phase). This experiment is then repeated
it is not essential to maintain a dc-voltage at a series of different test frequencies
level during the impedance measurement, usually starting at the highest frequency
a potentiostat is not required. In cases and finishing at the lowest in order to
like this, the ac-measurement unit can be minimize sample perturbation. The main
used more efficiently by itself since it usu- advantages of single-sine techniques are
ally allows more accurate measurements
and the use of higher frequencies. Po- • that they produce high-quality data,
tentiostats, however, are a useful addition • that fast measurements can be per-
to the impedance test equipment when formed at high frequencies since the
the system under investigation has a high measurements can be performed dur-
impedance, since they have high-input ing one cycle of the sine wave,
• that the instrumentation is simple since system under investigation according to

the sine wave can be produced with a
frequency generator, and V in = V0in cos(ωt + ) (55)
• that measurements can be carried out
at very high frequencies. then the product of these two signals
would be a sinusoid at twice the reference
frequency
2.6.7.1.1 Phase-sensitive Detectors – Lock-
in Amplifiers Lock-in amplifiers are de- V PSD = 12 V0ref V0in cos + 12 V0ref V0in
signed for signal detection in a high-noise
background and are therefore capable of × cos(2ωt + ) (56)
very accurate readings. A schematic of a
lock-in amplifier is shown in Fig. 18. An Figure 19 shows the case in which = 0.
oscillator produces a sinusoidal waveform, The mean level of V PSD , which is the dc
which is simultaneously applied to the component of the detector signal, can be
electrochemical cell and fed into the refer- calculated to
ence input. The response wave is applied PSD
Vmean = 12 V0ref V0in cos (57)
to the signal input and amplified includ-
ing noise. The bandwidth of the noise is In a lock-in amplifier, the reference signal
reduced by passing the amplified signal is passed through a phase shifter that
through a band pass filter [35, 36]. compensates for any phase shift between
A lock-in amplifier usually provides a input and reference signal introduced
dc output that is proportional to the ac by the system under investigation. Since
input signal. The rectifier, which performs there is no phase shift between input and
this ac to dc conversion, is called a Phase- reference signals ( = 0) after passing
Sensitive Detector (PSD). The detector through the phase shifter, the mean level of
multiplies input and reference signal with the detector signal assumes its maximum
each other. If the reference signal is a
possible value of 1/2V0ref V0in according to
voltage according to
Eq. (57). If the reference signal is fixed
V ref = V0ref cos ωt (54) at a constant level, the input signal and
therefore the impedance can be calculated
and the input signal is a voltage propor- from the mean level of the detector signal.
tional to the ac current flowing through the The dc component of the detector signal is
Mixer
Input (phase Output
amplifier Band pass sensitive Low pass amplifier
Signal filter detector) filter
input Output
Test
system
Oscillator Reference
trigger
Reference
input Phase shifter
Fig. 18 Schematic of a lock-in amplifier [35].

Fig. 19 Detection principle of a lock-in

amplifier [35].
Input signal
isolated from the signal by using a low pass
filter and by measuring the filtered output t
with a dc-voltmeter. Noise associated with
the input signal would also be multiplied
with the reference signal. However, since
noise does not have a fixed frequency or
phase relationship to the reference signal,
Reference signal
it does not cause a change of the mean dc
level, that is, lock-in amplifiers are highly
frequency selective. t
There are different types of PSDs, three
of which will be discussed briefly. The
first one is an analog multiplier in which
the input signal is multiplied with the
reference signal via an electronic circuit.
PSD output
In the presence of large noise, it is difficult Mean level

to guarantee linear operation of analog
multipliers that therefore often show poor t
noise rejection.
The simplest form of PSD is a digital
switching multiplier. Its operation is
shown in Fig. 20. During the first half
cycle, while the reference signal is positive,
the input signal remains unchanged. using DSPs is the dynamic range that
During the second half cycle, the reference is inherent to all digital instruments (see
signal becomes negative, and the input Chapter 1.2).
signal is inverted. This type of PSD usually Many modern lock-in amplifiers include
operates linearly. However, the main a microprocessor to perform simple math-
disadvantage of switching multipliers is, ematical manipulations, and can easily be
that they do not only detect signals at the incorporated into a computer-controlled
frequency of the reference signal but also at setup. Measurements using a lock-in am-
odd harmonics of the reference frequency. plifier provide high-quality data but are
The PSD found in digital lock-in am- usually more time consuming than those
plifiers is the digital multiplier. In this carried out with a frequency response an-
case, the input signal is first amplified alyzer (FRA).
and then digitized. This digitized signal
is multiplied with a digitized version of 2.6.7.1.2 Frequency Response Analysis
the reference signal using a digital signal Frequency Response Analysis is the most
processor (DSP). This technology provides widely used technique for impedance
perfect multiplication and excludes odd testing. Similar to the lock-in technique,
harmonics of the reference frequency. it can extract a small signal from a very
A problem associated with instruments high background of noise automatically
Fig. 20 The digital-switching multiplier

is the simplest form of the
phase-sensitive detector [36].
Input signal
used to reject these spurious components

t
is analogous to the digital multiplication of
the input signal with a reference waveform
performed by a phase-sensitive detector
(see Sect. 2.6.7.1.1). However, in the case
of an FRA, no phase shifter is needed
to compensate for the phase shift be-
− 0 +
tween input and reference signals, but

t the integrated waveforms are immediately
correlated by multiplying them by sine
and cosine reference waveforms, and the
resulting signals are integrated.
For a pure sinusoidal input signal
V in = r sin(ωt + ) = a sin ωt + b cos ωt

Mean level
0 +
(58)
t the results of the integrations would be

2π/ω
1 a
(V in sin ωt)dt = (59)
T 0 2
which corresponds to the real component

rejecting dc and harmonic responses. of the input signal and
An FRA correlates the input signal with

2π/ω
reference sine waves. To achieve fast 1 b
measurements, FRAs are usually equipped (V in cos ωt)dt = (60)
T 0 2
with separate analyzers for each input.
FRAs have a sine wave generator that which corresponds to the imaginary com-
outputs a small-amplitude voltage signal ponent of the input signal. Once real and
to the system under investigation. The imaginary parts of the input signals have
response signals, usually the voltage mea- been determined by correlation, the com-
sured between two reference points in plex impedance of the test object can
the electrochemical cell and a voltage be calculated. It can be mathematically
signal proportional to the current flow- proven that all the spurious components
ing through the cell, are fed into the are rejected by this technique of correlation
input channels, digitalized, and then in- provided that a sufficiently large number of
tegrated over several cycles in order to cycles have been used for the integration.
reject noise. Some FRAs are also capable of analyzing
In addition to the waveform at the fre- harmonics. This could be achieved by mul-
quency of interest, the measured signals tiplying the input signal with a waveform
usually contain a dc component, harmon- of the appropriate frequency or by carrying
ics, and noise (see Eq. 67). The process out an FFT on the input signal.
Fig. 21 FFT waveform composed of

three different frequencies [36]. sin(wt + Φ1)
The technique of digital correlation can

be applied to a large range of frequencies.
However, to allow easier analysis at
high frequencies, the measured signal
is shifted to lower frequencies using a
technique called heterodyning. The input
signal is multiplied with a reference
sin(3wt + Φ2)
signal of slightly different frequency. The
resulting signal has two components, one
at the sum of both frequencies and one
low-frequency component at the difference
of both frequencies. The high-frequency
signal can easily be filtered out, and the
low-frequency signal is then analyzed by
correlation as described above.
sin(5wt + Φ3)
2.6.7.2 Multiple-sine Techniques

Impedance spectroscopy is frequently
used to characterize systems that change
with time. In Sect. 2.6.3.2, it was shown
how inconsistent impedance data could be
eliminated using Kramers–Kronig trans-
forms. Another way to reduce the effects
of a system changing during the measure- Sum
ment is to reduce the total measurement
time by using a multisine technique, which
is also frequently called time-domain or
FFT technique.
In the case of multisine techniques,
a measurement is carried out at several
frequencies simultaneously. Analogous to
white noise, waveforms of typically 15 to
20 different frequencies and equal ampli- carried out during a full cycle of the
tudes are superimposed as demonstrated lowest frequency resulting in a reduced
in Fig. 21 [36]. The phases of the superim- measurement time. This advantage can
posed signals are randomized to minimize be utilized to investigate electrochemical
the amplitude of the composite signal. In processes in real time provided that time
contrast to single-sine techniques, mul- required for data acquisition is shorter
tisine techniques do not require waiting than the time constants of the system un-
for a full cycle to be completed for each der investigation.
of the frequencies used. For a multi- The voltage across the system under in-
sine technique, a full experiment can be vestigation and the current, which contains
information about the response of the sys- and responses at harmonics of the excita-
tem to each of the frequencies contained tion frequency. Therefore, the frequencies
in the perturbation signal, are measured superimposed for a multisine experiment
in the time domain. The time-domain sig- have to be chosen very carefully. To pre-
nals are digitized and transferred into the vent faulty results, the frequencies chosen
frequency domain by carrying out an FFT. are usually odd harmonics of the low-
The resulting data for each discrete fre- est frequency in order to eliminate the
quency can be treated the same way as second harmonics components that may
the impedance data obtained with a single- be caused by a nonlinear response of the
sine technique. Repeated application of system. Nonlinear behavior of the system
the waveform and averaging of the sig- would cause additional frequencies to ap-
nal before FFT is applied can improve pear in the response signal. The response
the signal-to-noise ratio of the multisine at these additional frequencies would ap-
technique, although it also increases the pear in the time-domain signal at places
measurement time required. not occupied by the frequencies of the ex-
It is worth considering the effect of citation waveform and at places already
a change of the system during the occupied by the excitation waveform. As
experiment. Because of the fact that the shown by Popkirov and coworkers, this
perturbation signal is a superposition of would cause a distortion of the response
different frequencies, data are acquired and data scattering [9]. Hence, not only
for each frequency throughout the entire nonstationary effects but also nonlinearity
experiment. The result for one frequency of the response appear in the FFT spectra
is then obtained by averaging all the data as additional scatter.
sampled for this particular frequency. If
the system under investigation changes 2.6.7.3 Experimental Setup
during the experiment, the data averaged For impedance measurements on electro-
for each frequency would vary noticeably chemical systems, frequently a potentio-
resulting in an impedance plot with stat of suitably high bandwidth is inte-
considerable scatter. This scatter can be grated into the impedance measurement
taken as a clear indication of unreliable system. The operation of a potentiostat
data, but it also makes the entire data has been described in Chapter 1.2 of this
set unusable. As described in Sect. 2.6.3.2, volume. In the case of electrochemical
the effect of a change of the system impedance measurements, the potentio-
during the measurement can frequently stat is not only responsible for maintaining
be eliminated by discarding low-frequency a defined dc level but also for applying the
data. For a multisine technique, this means correct ac voltage to the working electrode.
that the experiment has to be repeated with A typical measurement setup with an FRA
a higher low-frequency limit in order to and a potentiostat is shown in Fig. 22 [37].
further reduce the measuring time. The polarization potential and the ac per-
As discussed in Sect. 2.6.10, electro- turbation are added together and applied
chemical systems can show a nonlinear re- to the electrochemical cell at the counter
sponse, that is, the current response of an electrode (CE terminal). The voltage differ-
electrochemical system can be composed ence between the two reference electrodes
of a response at the excitation frequency RE1 and RE2 is measured and fed back to
FRA
x1 / x10
DC ref δRE
V out
Sweep + P/G stat Reject dc
Pol IR comp
x1 / x10
x1
I out
x0.01
ECI I/V convert Reject dc
CE RE1 RE2 WE
Fig. 22 Example for impedance measurement setup with an FRA [37]. (ECI: electrochemical
interface, P/G stat: potentiostat/galvanostat).
the control loop, which corrects the volt- reference electrode in the electrochemical
age applied to the counter electrode until cell, for example, in the characterization
the required potential difference between of batteries, or when it is important to
RE1 and RE2 is established. The voltage maintain a homogeneous current distri-
measured between RE1 and RE2 and the bution as, for example, in low-conductivity
current measured at the working electrode media, in which an inhomogeneous cur-
are amplified by the potentiostat and fed rent distribution can lead to measure-
into the FRA as voltage signals. The poten- ment artifacts [38]. If the properties of a
tiostat can provide dc rejection as well as single electrode (the working electrode)
compensate for the solution resistance if are to be investigated using a two-
required (IR compensation). electrode arrangement, the impedance of
In an experimental setup as described the counter electrode has to be mini-
above, there is usually a choice between a mized by giving it a considerably larger
two-, three-, or four-electrode cell. These surface area than that of the working
options determine which part of the electrode.
electrochemical cell is characterized by the For the characterization of electrode
impedance measurements. properties, a three-electrode cell is superior.
In a two-electrode cell, terminal RE1 is In this case, terminal RE1 is connected
connected to the counter electrode CE, and to a reference electrode placed close to
terminal RE2 is connected to the working the working electrode, which itself is con-
electrode (Fig. 23a). The impedance is al- nected to terminal RE2 (Fig. 23b). Since
ways measured between RE1 and RE2, the impedance is now measured be-
that is, the impedance measured includes tween reference and working electrodes,
that of the counter electrode, the elec- the impedance obtained will only be
trolyte solution, and the working electrode. influenced by the properties of the work-
This type of electrode arrangement is typ- ing electrode and the properties of the
ically used if it is impossible to position a electrolyte solution between working and
CE RE1 RE2 WE Fig. 23 (a) Two-electrode cell;

(b) three-electrode cell; and
(c) four-electrode cell.
(a)
CE RE1 RE2 WE
(b)
CE RE1 RE2 WE
(c)
reference electrodes, that is, the influence electrochemical systems and new materials
of the counter electrode on the impedance as it provides information over a large
spectrum is eliminated and that of the range of frequencies simultaneously.
resistance of the electrolyte solution is However, the results obtained using this
minimized. method are always surface averaged. This
The applications of impedance spec- can lead to difficulties when the proper-
troscopy are not limited to the characteri- ties of an electrode or a thin film are not
zation of electrode properties. Sometimes uniform over the entire sample area. Sev-
it is desirable to investigate the properties eral attempts have been made to develop
of membranes, solutions, or dielectrics. methods that can detect impedance of elec-
trochemical systems with local resolution.
For this kind of application, four-electrode
Two of these techniques will be described
cells provide the best results. Two reference
in this section.
electrodes are placed in the electrochem-
ical cell between counter and working
2.6.8.1 Local Electrochemical Impedance
electrodes (Fig. 23c). The impedance mea- Spectroscopy (LEIS)
sured depends purely on the properties of Lillard and coworkers developed a method
the electrolyte or membrane between the called Local Electrochemical Impedance
two reference electrodes, and the electrode Spectroscopy LEIS [39]. It relies on the fact
properties are completely eliminated from that ac current densities in the solution
the impedance spectrum. very near to the working electrode are pro-
portional to the local impedance properties
2.6.8 of the electrode. In order to determine
Spatially Resolved Impedance the current densities normal to the sur-
Measurements face, the ac potential drop was measured
between planes parallel to the electrode
As demonstrated repeatedly in this chap- surface employing a two-electrode micro-
ter, ac-impedance spectroscopy is a valu- probe. A schematic of a commercially
able tool in the characterization of available experimental setup for LEIS is
shown in Fig. 24. The ac current density to the surface of the working electrode. In
can be calculated from the potential drop reality, however, the ac currents spreads
V (ω)probe using the following relation- in the solution as a function of the

ship: distance from the electrode surface, that
is, the resolution of LEIS depends on

V (ω)probe σ
I (ω)local = (61) the distance of the two-electrode probe
l from the electrode surface. The spatial
where σ is the conductivity of the elec- resolution of LEIS is also limited by the
trolyte and l is the distance between the size of the two-electrode probe, since it
two microelectrodes normal to the sur- is not possible to resolve features smaller
face. The area-normalized magnitude of than the probe. A reduction in size of
the local impedance is then obtained from the probe is also accompanied by an
increase of its impedance, which makes
V0 (ω) l
Z0 (ω)local = (62) it more difficult to measure the current

V (ω)probe σ
density accurately. Lillard and coworkers
where V0 (ω) is the magnitude of the ac therefore indicated that these competing
potential applied between working and effects needed to be optimized in order to
reference electrodes. achieve maximum spatial resolution [39].
The assumption made in the derivation Bayet and coworkers modified LEIS by
of Eq. (62) is that the current density at replacing the two-electrode microprobe
the tip of the two-electrode probe and the with a single vibrating probe to improve
current density at the electrode surface the local resolution [40].
are the same, which means that all the LEIS has been applied successfully to
current measured at the two-electrode study defects in organic coatings [41, 42]
probe is assumed to be flowing normal and passivation of metal surfaces [43].
Computer
I in
Electrometer FRA
Scanning Counter electrode

head
controller Potentiostat
Reference
Dual
electrode
electrode
scanning
probe
l
Working electrode
Fig. 24 Schematic of an experimental setup for LEIS (provided by Uniscan

instruments).
2.6.8.2 Scanning Photo-induced The resulting current causes electrons and

Impedance Microscopy (SPIM) holes to collect on both sides of the in-
Another impedance-based imaging tech- sulator effectively charging the capacitor.
nique for laterally resolved characteri- When the illumination is turned off, the
zation of thin films or electrochemi- capacitor discharges and a current flows
cal systems is Scanning Photo-induced in the opposite direction. A modulated
Impedance Microscopy (SPIM) [44]. It is or chopped light beam focused on one
based on photocurrent measurements at region of the semiconductor substrate re-
field-effect structures. In their simplest ar- sults in a local ac current. When the space
rangement, field-effect structures consist charge region is biased towards accumu-
of a semiconductor substrate with a thin lation, no photocurrent flows. Figure 25
insulator, and a gate electrode. This gate (curve a) schematically shows the change
electrode can be a metal film resulting of the photocurrent with the dc voltage.
in the structure Metal Insulator Semicon- If the system under investigation, for
ductor (MIS) or, alternatively, Electrolyte example a thin film, is deposited on a semi-
Insulator Semiconductor structures are conductor insulator structure, the pho-
used, in which the electrolyte is in direct tocurrent decreases as shown in Fig. 25
contact with the insulator, and a reference (curve b). The difference in photocurrent
electrode is required to fulfill the function is a function of the complex impedance
of the gate electrode. of the layer added. Local differences in the
Different regions of the semiconductor impedance can be detected in the scanning
in a field-effect structure can be addressed mode. An additional variation of the mod-
with light. A dc voltage is applied be- ulation frequency of the light results in ac-
tween the semiconductor substrate and impedance spectra with lateral resolution.
the gate electrode. Because of the presence In classical ac-impedance spectroscopy,
of an insulator, this does not produce a a small-amplitude ac-voltage signal is
dc current but controls the space charge applied to the system under investigation,
region in the semiconductor at the in- and the resulting ac current is measured.
terface between the semiconductor and In the case of SPIM, a local ac voltage
the insulator. Electron hole pairs created is created as a result of the photo-
during illumination of the semiconductor effect within the semiconductor by a
separate in the field of the space charge focused, modulated light beam leading to
region if it is biased towards inversion. a localized ac current. In contrast to other
a
i photo
b
Fig. 25 Curve (a) photocurrent of a
field-effect structure as a function of the
dc voltage; curve (b) decrease of the
saturation current of the photocurrent
curve upon deposition of a thin film on
Vdc
the field-effect structure.
localized impedance techniques such as response ac current and its phase angle
LEIS, no probe or microelectrode needs are registered. The function of the dc
to be positioned above the sample. It is potential is to enforce and change the
however required that the system under surface concentrations of the electroac-
investigation be deposited on a field-effect tive species at the electrode surface. A
structure. This can be done in two different simplification of the theoretical treatment
arrangements: (1) The film is deposited can be achieved when the perturbation
onto a semiconductor insulator structure frequency is sufficiently high to sepa-
followed by the deposition of a thin metal rate the ac and dc contributions of the
film (Fig. 26a). This allows measurements response signal. Because of the differ-
in the gas phase. (2) The film is deposited ent timescales of the long term diffusion
onto a semiconductor insulator structure caused by the dc potential and the rapid
and is then exposed to an electrolyte diffusional fluctuations caused by the ac
solution (Fig. 26b). potential, the surface concentration set
up by the dc potential has the same ef-
fect as a bulk concentration on the ac
2.6.9
AC Voltammetry process.
In this chapter, only the response of a
reversible system will be described. Linear
AC voltammetry is an extension of classical
sweep ac voltammetry of a Nernstian
linear sweep techniques such as cyclic
system is considered.
voltammetry. Its main strength lies in
the quantitative characterization of elec- O + ne− −
−
−−−
−R (63)
trode processes, and it can also be used
for analytical purposes. A dc ramp with The mean surface concentrations of the
a comparatively slow sweep rate and electroactive species are exactly the same
an ac signal are superimposed and ap- as they would be in a linear sweep
plied to a working electrode, and the experiment without a superimposed ac
Electrode
Metal / solution
Thin film under

investigation
A A
SiO2
Silicon
Metal contact
Light
(a) (b)
Fig. 26 Experimental arrangements for SPIM for (a) gas phase measurements and
(b) measurements in electrolyte solutions.
potential (see Chapter 2.1 of this volume). the sensitivity of ac voltammetry as an

The amplitude of the ac current can be analytical technique [45, 46]. Although un-
derived according to der these conditions, the peak current is
not proportional to the ac-amplitude any-
∗E
n2 F 2 Aω1/2 DO cO
1/2
E0 0 more, it could be demonstrated that the
I0 = = 2
(64)
Zf 4RT cosh (a/2) area under the peak is proportional to
the perturbation amplitude and the an-
where a = nF /RT (Edc − E1/2 ), E0 is the alyte concentration [45]. Since the rules
amplitude of the ac potential, Zf is for of semi-infinite diffusion apply, a
the faradaic impedance, Edc is the dc phase angle of −45◦ can be observed
potential, and E1/2 is the reversible half for any reversible electrode reaction (see
wave potential [5]. The bell shaped ac Sect. 2.6.3.1).
voltammogram described by Eq. (64) is By extending ac linear sweep voltamme-
shown in Fig. 27. At the peak maximum, try by a reverse scan, ac-cyclic voltammetry
the applied dc potential equals the half is obtained. Surface concentrations of
wave potential. the electroactive species are the same at
The peak current can be expressed as the same potential for forward and re-
∗E
1/2 verse scans. Hence, the peaks for forward
n2 F 2 Aω1/2 DO cO 0
Ip = (65) and reverse scans are identical. In con-
4RT trast to classical cyclic voltammograms,
From Eq. (65), it can be seen that the ac-cyclic voltammograms have a clear base-
peak current is proportional to n2 , ω1/2 , line that is advantageous for quantitative
and cO * [5]. The latter is utilized for the measurements.
application of ac voltammetry for an- If the dc process is not fully re-
alytical purposes. The peak current is versible, the surface concentrations of the
also proportional to the amplitude of electroactive species are different at a given
the ac potential E0 . This is, however, dc potential for forward and reverse scans,
limited to small amplitudes, that is, to that is, for quasi-reversible systems a dis-
quasi-linear conditions. More recent sim- placement of the peaks for forward and
ulations have shown that ac voltamme- reverse scan can be observed. This dis-
try can be carried out at considerably placement can be used to derive kinetic
larger ac-potentials in order to increase parameters of the electrode reaction. For
I0
0 Fig. 27 Characteristic shape of

Edc an ac voltammogram for a
E1/2
reversible system.
a detailed discussion of the behavior of of techniques based on the nonlinearity of

quasi-reversible and irreversible systems, electrochemical systems is, that the second
see Bard and Faulkner [5]. and higher harmonic signals are relatively
free of charging currents. Since the
2.6.10 double-layer capacitance behaves in much
Nonlinearity of Electrochemical more linear fashion than the faradaic
Systems – Analysis of Higher Harmonics processes, the charging currents of the
double layer can mainly be found at
In previous chapters only the linearized the excitation frequency ω. Hence, the
form of the current–potential relation determination of kinetic parameters is
has been used. The other terms of the frequently easier at higher harmonics.
Taylor series expansion were dropped. Harmonic analysis has been used widely
However, in reality, current–potential re- for determining corrosion rates (see, for
lationships show a considerable amount of example, Jafar and coworkers in [2]).
curvature. The Taylor series expansion of Higher harmonics can be detected by
the Butler–Volmer equation predicts that FRAs or lock-in amplifiers, which can be
tuned to detect a multiple of the excitation
∂I

I =
E frequency. An alternative is to extract
∂E E0 the harmonic signals from the response
using FFT.
1 ∂ 2I
+
E 2 + · · · (66) Second- or higher-order effects have not
2 ∂E 2 E0
only been utilized for impedance spec-
When a pure sinusoidal voltage is applied troscopy but also for ac voltammetry [45,
to an electrochemical cell, the waveform of 47, 48]. While an ac voltammogram (shown
the resulting current is very often distorted
due to this nonlinear current–potential
relationship unless the excitation voltage I0
is sufficiently small. The response signal
can be described as
0
Edc
I = Idc
+ I0 sin(ωt + ) + I1 sin(2ωt + 1 )
+ I2 sin(3ωt + 2 ) + · · · + N oise E1/2
(67) (a)
where Idc is the dc component of the
current and I1 , I2 , I3 . . . are harmonic |I0|
distortion components.
It is useful to consider the second and
higher harmonics for the characterization
of electrochemical systems. The advantage
Fig. 28 Second harmonic ac

voltammogram for a reversible system 0
(a) phase-sensitive current detection E1/2 Edc
and (b) modulus of the current. (b)
in Fig. 27) corresponds to the first deriva- 6. B. W. Johnson, D. C. Read, P. Christensen

tive of the current–potential relationship et al., J. Electroanal. Chem. 1994, 364,
103–109.
(first term in Eq. 66), the higher harmon- 7. G. S. Popkirov, R. N. Schindler, Electrochim.
ics ac voltammograms that depend on the Acta 1995, 40, 2511–2517.
curvature of the current–potential rela- 8. G. S. Popkirov, R. N. Schindler, Electrochim.
tionship, correspond to that of the second Acta 1993, 38, 861–867.
and higher derivatives (Fig. 28a). The sign 9. G. S. Popkirov, Electrochim. Acta 1996, 41,
1023–1027.
inversion of the current at E1/2 can be de- 10. M. E. Orazem, P. T. Wojcik, M. Durbha et al.,
tected in phase-sensitive measurement as Mater. Sci. Forum 1998, 289-2, 813–828.
a phase shift of 180◦ . A measurement of 11. M. Kendig, F. Mansfeld, Corrosion 1983, 39,
the magnitude of the ac-current amplitude 466, 467.
12. D. D. Macdonald, M. Urquidimacdonald, J.
would result in a voltammogram of the
Electrochem. Soc. 1985, 132, 2316–2319.
shape shown in Fig. 28(b). 13. J. M. Esteban, M. E. Orazem, J. Electrochem.
Soc. 1991, 138, 67–76.
14. P. Agarwal, M. E. Orazem, L. H. Garciarubio,
Acknowledgments
J. Electrochem. Soc. 1992, 139, 1917–1927.
15. R. L. Hurt, J. R. Macdonald, Solid State Ionics
I would like to thank Werner Moritz, 1986, 20, 111–124.
Ron Armstrong, and Andrea Sabot for 16. K. S. Cole, R. H. Cole, J. Chem. Phys. 1941, 9,
their suggestions for improving this 341–351.
17. S. H. Liu, Phys. Rev. Lett. 1985, 55, 529–532.
chapter. I am also very grateful to 18. A. Lemehaute, G. Crepy, Solid State Ionics
Ed Holden (Perkin Elmer), Andrew 1983, 9–10, 17–30.
Hinton (Solartron Analytical), Graham 19. G. Kahanda, M. Tomkiewicz, J. Electrochem.
Johnson (Uniscan Instruments), and Soc. 1990, 137, 3423–3429.
20. T. Pajkossy, J. Electroanal. Chem. 1994, 364,
Timoer Frelink (Eco Chemie) for pro-
111–125.
viding me with up-to-date technical 21. Z. Kerner, T. Pajkossy, Electrochim. Acta
information. 2000, 46, 207–211.
22. R. de Levie, Electrochim. Acta 1963, 8,
751–780.
References
23. H. K. Song, H. Y. Hwang, K. H. Lee et al.,
Electrochim. Acta 2000, 45, 2241–2257.
1. B. Cottis, S. Turgoose, R. Newman, Corro- 24. H. K. Song, Y. H. Jung, K. H. Lee et al., Elec-
sion Testing Made Easy: Impedance and Noise trochim. Acta 1999, 44, 3513–3519.
Analysis, NACE International, Houston, Tex., 25. R. D. Armstrong, D. Wright, Electrochim.
2000. Acta 1993, 38, 1799–1801.
2. John R. Scully, David C. Silverman, M. Ken- 26. F. Mansfeld, C. H. Tsai, Corrosion 1991, 47,
dig, Electrochemical Impedance: Analysis and 958–963.
Interpretation, American Society for Testing 27. S. Krause, C. J. McNeil, R. D. Armstrong
and Materials, West Conshohocken, 1993. et al., J. Appl. Electrochem. 1997, 27, 291–298.
3. J. R. Macdonald, Impedance Spectroscopy Em- 28. C. Sumner, A. Sabot, K. Turner et al., Anal.
phasizing Solid Materials and Systems, 1st ed., Chem. 2000, 72, 5225–5232.
John Wiley & Sons, New York, 1987. 29. C. J. McNeil, D. Athey, M. Ball et al., Anal.
4. C. M. Brett, A. M. Oliviera Brett, Electrochem- Chem. 1995, 67, 3928–3935.
istry: Principles, Methods, and Applications, 30. W. O. Ho, S. Krause, C. J. McNeil et al.,
Oxford University Press, New York, 1993. Anal. Chem. 1999, 71, 1940–1946.
5. A. J. Bard, L. R. Faulkner, Electrochemical 31. B. Malmgren-Hansen, T. S. Sorensen, J. B.
Methods: Fundamentals and Applications, 1st Jensen et al., J. Colloid Interface Sci. 1989,
ed., John Wiley & Sons, New York, 1980. 130, 359–385.
32. R. D. Armstrong, A. K. Covington, G. Pts 1 and 2, Transtec Publications, Zurich,

P. Evans et al., Electrochim. Acta 1984, 29, Uetikon, 1998, pp. 57–68, Vol. 289-2.
1127–1131. 41. F. Zou, D. Thierry, Electrochim. Acta 1997,
33. A. Friebe, W. Moritz, J. Appl. Polym. Sci. 42, 3293–3301.
1994, 51, 625–634. 42. R. S. Lillard, J. Kruger, W. S. Tait et al., Cor-
34. J. Szeponik, W. Moritz, F. Sellam, Ber. Bun- rosion 1995, 51, 251–259.
sen-Ges. Phys. Chem. Chem. Phys. 1991, 95, 43. J. Kruger, R. S. Lillard, C. C. Streinz et al.,
1448–1453. Mater. Sci. Eng., A Struct. Mater. Properties
35. PerkinElmer Instruments, What is a lock-in Microstruct. Process. 1995, 198, 11–18.
amplifier?, Technical Note TN1000. 44. S. Krause, H. Talabani, M. Xu et al., Elec-
36. EG&G Princeton Applied Research Electro- trochim. Acta 2002, 47, 2143–2148.
chemical Impedance Measurements: Instru- 45. S. O. Engblom, J. C. Myland, K. B. Oldham,
mentation and Techniques, Application Note J. Electroanal. Chem. 2000, 480, 120–132.
AC-3. 46. D. J. Gavaghan, A. M. Bond, J. Electroanal.
37. Solartron, Materials Test Products, Version Chem. 2000, 480, 133–149.
1.6.4. 47. W. T. Yap, J. Electroanal. Chem. 1998, 454,
38. G. Hsieh, S. J. Ford, T. O. Mason et al., Solid 33–38.
State Ionics 1996, 91, 191–201. 48. D. A. Brevnov, H. O. Finklea, J. Electroanal.
39. R. S. Lillard, P. J. Moran, H. S. Isaacs, J. Chem. 2000, 488, 133–139.
Electrochem. Soc. 1992, 139, 1007–1012.
40. E. Bayet, F. Huet, M. Keddam et al., Electro-
chemical Methods in Corrosion Research Vi,
2.7 to the film redox state (controlled via po-

The Electrochemical Quartz Crystal tential or charge), temperature, timescale
Microbalance (frequency), and environment (both sol-
vent and electrolyte ions). These effects
A. Robert Hillman are reviewed with selected examples from
University of Leicester, Leicester, United
a diverse range of materials, encom-
Kingdom
passing metals, metal oxides, semicon-
ductors, metal coordination complexes,
Abstract
and redox, conducting, and insulating
polymers.
This chapter reviews the development and
use of the electrochemical quartz crys-
2.7.1
tal microbalance (EQCM) as an in situ
Introduction
probe of interfacial processes involving the
formation, properties, and redox switch-
2.7.1.1 Overview
ing of multilayer films. For films that The quartz crystal microbalance (QCM)
are sufficiently rigid and/or thin (termed is one of a large family of acoustic wave
acoustically thin), the EQCM functions as devices, whose utility is based on the sensi-
a quantitative gravimetric probe of surface tivity of resonant frequency and amplitude
population changes. This well-established responses to their environment [1]. The
methodology is now being used to explore common feature of all the devices is that
both thermodynamic and kinetic aspects they use excitation of a piezoelectric el-
of film compositional changes, for exam- ement (commonly, but not universally,
ple, in response to electrochemical stimuli. quartz) to launch an acoustic wave that
Correlation of mobile species (ion and interacts with the medium immediately
solvent) population changes – within the adjacent to the piezoelectric element. The
effectively static film matrix – with elec- nature of this interaction – generally char-
trochemical control and response func- acterized through phase and amplitude
tions (potential, current, and charge) now shifts (as functions of frequency) – is de-
provides mechanistic information. Com- termined by the local composition and
bination with other in situ (e.g. optical properties of the medium. Different de-
or spectroscopic) techniques is a devel- vices use different crystal cuts to generate
oping area and can facilitate separation different crystal modes that propagate in
of individual ion and solvent transfers. different ways. However, the common fea-
For films that are softer and/or thicker ture is that the acoustic waves generally
(termed acoustically thick), the acoustic only propagate a short distance into the
wave is attenuated during passage across contacting medium, so that the frequency
the film, which in turn undergoes acous- response of the device is predominantly
tic deformation. The EQCM now responds determined by interfacial properties; this
primarily to film viscoelastic properties, is ideal for electrochemical systems.
that is, film (rather than mobile species) In this review, we discuss the develop-
dynamics. This is a developing area, but it ment and application to electrochemical
is now possible to extract film shear modu- problems of a bulk acoustic wave device,
lus components (storage and loss moduli); the thickness shear mode (TSM) resonator,
these are found to respond significantly whose operation (in the simplest case of a
2.7 The Electrochemical Quartz Crystal Microbalance 231
Rigid overlayer
Electrodes
Quartz
Fig. 1 Schematic diagram of the shear motion of a TSM
resonator loaded with a rigid overlayer.
rigid overlayer) is schematically indicated EQCM has become accepted as a standard

in Fig. 1. Generically, the utility of the tool in the armory of the interfacial
device exploits the fact that almost any sur- electrochemist.
face process – whether chemically or phys- Although this is a relatively new field,
ically originated – will result in a change the rapid increases in capability and areas
in surface composition and/or species of application of the EQCM have prompted
population(s). The consequent change in a number of reviews in journals and
acoustically coupled mass causes a change books [2–6]. The reader entering the field
in the resonant frequency of the oscilla- will find these helpful, although it is
tor. Through suitable analysis, expounded noteworthy that over half of the literature
in this article, one can invert this process in this field has been published from
and attribute the frequency response to 1998 onwards, that is, after all but one
the underlying interfacial physicochemical of these reviews; the most significant
process(es) and materials properties. consequence of this is the necessarily low
coverage given to viscoelastic effects (see
below). There are also special issues of
2.7.1.2 History and Prior Reviews primary journals related to the topic [7, 8]
The growth of EQCM studies as an and those interested in keeping abreast
independent field has been remarkable. of recent developments should consult
Following the extension of the ex situ the outputs of two established series of
QCM method to in situ operation as biannual meetings [9, 10].
EQCM in 1985, annual publication rates
were at single figures for the first few 2.7.1.3 Organization and Coverage of
years, rising steadily through the 1990s Topics
to around 100 per year at present. This Since full listings of the numerous
does not include the huge number of EQCM-based publications can readily be
(nonelectrochemical) QCM publications, generated by computer-based literature
presently running at over 300 per year and searches, this article is selective with
totaling well over 2000 to date. There are the goal of clearly illustrating specific
also a considerable number of publications phenomena or processes. A companion
that report studies in which the (E)QCM article in this series [11] focuses on the
was used as a supporting (rather than more overtly interfacial aspects (such as
primary) technique, to the extent that it fluid coupling – including roughness and
is not always listed in abstracts; this is ‘‘slip’’ phenomena – and double-layer ef-
testament to the speed with which the fects) and monolayer films (including
simple adsorption, self-assembled mono- such as film displacement, that under-

layers, and underpotential deposition of lie viscoelastic phenomena (discussed in
metals). It provides important updates of Sects. 2.7.2.1.2 and 2.7.3.7).
topics that were the focus of previous re- When loaded with a thin film, a TSM
views [2–6]. In contrast, the present article resonator functions as a gravimetric probe
focuses on thicker (multilayer) interfacial of the populations of surface-attached
systems, which those reviews covered in species. In this context, ‘‘thin’’ implies a
less detail. film whose combination of thickness and
For sufficiently thin films, the general shear stiffness is such that the acoustic
concepts applied to monolayer systems wave does not deform it; consequently, the
apply; in particular, the EQCM functions film moves synchronously with the res-
as a simple gravimetric probe of interfa- onator. Such a film is termed acoustically
cial populations and processes. For thicker thin. Monolayer adsorbates can generally
be expected to conform to this require-
films, factors other than inertial mass can
ment. Since such systems have been
exert a significant influence on EQCM re-
described in a companion article in this
sponse. Prominent among these are film
series, we describe this simpler case (as it
viscoelastic properties, whose influence is
relates to multilayer films) first and in less
commonly dominant for thick polymer
detail.
films. Accordingly, the next section cov- In contrast, when loaded with a suf-
ers the principles and practice of operation ficiently thick film, a TSM resonator
of the EQCM under conditions in which it functions as a rheological probe of the
functions as a gravimetric probe and then properties of surface-attached species. An
as a viscoelastic probe. As will become ap- acoustically thick film is one whose com-
parent, a given system can show either type bination of thickness and shear stiffness
of response (according to such parameters is such that the acoustic wave significantly
as temperature, solvent, and potential). deforms it. Consequently, displacement of
Thus, the account of experimental appli- the outer regions of the film exhibits a
cation of the EQCM is organized on the phase delay with respect to displacement of
basis of the (electro)chemistry involved. the underlying resonator. Multilayer adsor-
bates may behave in this manner if they are
2.7.2 either intrinsically ‘‘soft’’ or are plasticized
The EQCM Technique by the permeation of a solvent or other
small molecules. This situation is more
2.7.2.1 Principles complicated and it is only recently that
Detailed explanations of the operation of the theoretical aspects have been treated
TSM resonator devices have been given in quantitative fashion for fluid-immersed
elsewhere [1, 12]; here, we summarize the films. Accordingly, this case is described
principles in only as much detail as re- subsequently to the simpler gravimetric
quired to understand their capabilities and case and in greater detail.
limitations in the electrochemical context.
As a complement to the above electron- 2.7.2.1.1 The EQCM as a Gravimetric Probe
ically oriented descriptions, Kanazawa’s Application of an AC voltage across an AT-
physical model [13] is extremely helpful cut [1] quartz crystal results in the crystal
in visualizing the mechanical aspects, undergoing the shear motion illustrated
in Fig. 1. Resonance occurs when there impedances) is generally not significant

is constructive interference from the wave for mass changes involving rigid films
fronts propagating from the upper and that are less than 1% of the mass of the
lower surfaces of the resonator, that is, quartz crystal; most problems encountered
the outgoing and reflected waves. This in interfacial studies of rigid films conform
criterion is satisfied when the thickness to this.
of the crystal corresponds to half the Implicit in the derivation of the Sauer-
acoustic wavelength. Thus the thickness brey equation is the idea that the medium
of the crystal defines the base operating into which the acoustic wave propagates
frequency. does not dissipate energy. It has long been
The central tenet behind all the appli- recognized that if the ambient medium
cations discussed in this review is that were a liquid, then this would no longer be
a change in surface composition and/or true – indeed it was commonly presumed
species populations leads to a change in that the energy dissipation would be so
the acoustic loading, which is detectable great that attenuation of the shear wave
via a change in resonator frequency re- would so severely damp the oscillation as
sponse. In the simplest case, one can to prevent in situ use of the QCM. In
readily envisage that a change in the mass the 1980s, it was shown that, although
of the mechanically oscillating system (the significant, fluid damping is not fatal to
quartz resonator plus associated surface- in situ operation of the QCM [15] – thus
attached material) will lead to a change in its electrochemical extension, the EQCM,
resonant frequency. For TSM resonators, became a reality [16]. Furthermore, the ef-
the frequency responds to the change in fective viscous coupling of the fluid (the
areal mass density (i.e. change in mass decay length, xL , of the wave in the fluid)
per unit area, (M/g cm−2 ) rather than was calculable [17] in terms of fluid prop-
the change in mass (m/g) per se. For a erties (viscosity (ηL ) and density (ρL )) and
change in elastically coupled (‘‘rigid’’) in- the crystal operating frequency:
ertial mass, the resonant frequency change
1/2 1/2
(f /Hz) of an AT-cut quartz crystal res- ηL 2vL
onator from its initial value (f0 ) is given by xL = = (2)
πρL f0 ω

2 where vL (= ηL /ρL ) is the kinematic
f = − f02 M (1)
ρq vq viscosity of the fluid and ω(= 2πf0 ) is
the angular frequency of the resonator.
where ρq and vq , respectively, are the
density and shear wave velocity within
the quartz. The simple form of Eq. (1), 2.7.2.1.2 The EQCM as a Viscoelastic Probe
known as the Sauerbrey equation [14], is a The Sauerbrey equation was developed to
result of making the approximation that describe films that are thin and elastic
the overlayer producing the added mass (in common parlance, ‘‘rigid’’). In this
can be treated as an extension of the situation, the addition of more of the same
quartz. This is strictly only true when the material or of a second rigid material
layer is thin, elastic, and has the same can be described in a linear additive
acoustic impedance as quartz. In practice, manner. This is because the ‘‘new’’ layer
the acoustic mismatch problem (differing experiences the same displacement as
film and quartz characteristic acoustic the existing layer, that is, there is no
decay or phase shift of the acoustic driving force) and the resultant current
wave across the inner layer. However, (the response). An analogous approach
if the material immediately adjacent to can be used for the surface mechanical
the resonator surface is nonrigid, the impedance of the resonator loading. For
effect(s) of additional loading(s) on the the system as a whole, the mechanical
resonator are not linearly additive. This is impedance (ZS ) is the quotient of the sur-
because the material immediately adjacent face stress (Txy , driving the motion) and
to the resonator now suffers a deformation the particle velocity (vx , the motion) at the
across its thickness. Thus, any additional resonator surface (y = 0):
material on the outer surface will not

experience the same displacement as when Txy
ZS = (3)
overlaid on a rigid inner layer. Instead, the vx y=0
acoustic wave it experiences from the inner
layer will be subject to a phase shift and
Pursuing this analogy, components in the
significant attenuation with respect to the
electrical equivalent circuit can be assigned
wave initially launched by the resonator.
direct mechanical significance: resistance
One common experimental approach to
represents energy loss, capacitance rep-
this situation (described in Sect. 2.7.2.2.2)
resents energy storage, and inductance
is based on impedance methods. An AC
represents inertial mass. Thus, one overall
electrical excitation drives the resonator,
strategy is to model the electrical response
whose mechanical response is determined
in terms of an equivalent circuit and then
by the viscoelastic characteristics of the
to transform the components into their
loading (in this context, surface films
and electrolyte solutions). The measured mechanical analogs.
electrical driving and response functions There are two electrical equivalent cir-
are related by the electrical impedance, cuits in common usage, the transmission
and analogous mechanical functions can line model (TLM) and a lumped element
be defined; the electrical and mechanical model (LEM) commonly referred to as the
analogs are linked through the piezoelec- Butterworth–van Dyke (BvD) model; these
tric effect, quantified by the electrome- are illustrated in Figs. 2(a and b), respec-
chanical coupling factor of the resonator tively. In the TLM, there are two acoustic
(usually quartz, for which K 2 = 7.74 × ports that represent the two crystal faces:
10−3 ). Here we summarize the general one is exposed to air (i.e. is stress-free,
concepts, developed by Martin and a se- indicated by the electrical short) and the
ries of coworkers [1, 18, 19], and illustrate other carries the mechanical loading (here,
them for surface loadings typical of elec- a film and the electrolyte solution, repre-
trochemical systems. sented below by the mechanical loading
In electrical/electrochemical impedance, ZS ). These acoustic ports are connected
it is common practice to represent the sys- by a transmission line, which is in turn
tem under study by an electrical equivalent connected to the electrical circuitry by a
circuit, in which a combination of resis- transformer representing the piezoelectric
tors, capacitors, and inductors represents coupling. For the TLM, one can show [18,
19] that the motional impedance (Zm 1 ) as-
the physical components of the system.
For the system as a whole, the impedance sociated with the surface loading can be
is the quotient of the applied potential (the related to the mechanical impedances of
Quartz resonator Nonpiezoelectric

(piezoelectric) load layers
Viscoelastic Newtonian
film fluid
jX
Z lm
Co
(a)
L1
Co C1
R1
L2
R2
(b)
Fig. 2 Electrical equivalent circuit models for a TSM resonator:
(a) transmission line model (TLM) and (b) Butterworth–vanDyke
lumped element model (LEM). Circuit elements are defined in the
main text.
quartz (Zq ) and the loading (ZS ) by unity. Hence,

1 π ZS
1 φq (ZS /Zq ) j (ZS /Zq ) Zm ≈ 2
= R2 + j ωL2
Zm = 1 − 4K ωS C0 Zq
4K 2 ωC0 2 tan(φq /2) (5)
(4) In Eq. (5), we have indicated that the sim-
In practice, one is always interested in plified expression for Zm1 can be expressed
measurements close to resonance (ω as a series combination of a resistor and an
ωs ), so that the phase shift, φq π. For inductor. In the mechanical context, these
most systems of interest, the loading represent energy loss and inertial mass,
is generally a small perturbation on respectively. Equation (5) in fact describes
the total impedance (ZS /Zq 1). Under the BvD LEM; in other words, the LEM
these conditions, the second term in the (Fig. 2b) is a low loading/near resonance
bracket in Eq. (4) is small compared to approximation of the more general TLM
(Fig. 2a). Simple considerations show that frequency, ρq is the density of quartz, and
the vast majority of electrochemical ex- µq is the shear elastic constant for quartz.
periments correspond to this low loading In fact, since the mass loading exerts only
case and that the error incurred by the a small perturbation on the resonant fre-
use of the simpler Eq. (5) [as opposed to quency (f0 fS ) and νq = (µq ρq )1/2 , it
the more cumbersome Eq. (4)] is less than can be seen that Eq. (6) is a restatement
1% – comparable to or better than the pre- of the Sauerbrey equation. Thus, an un-
cision of experimental data. equivocal test for film rigidity is a crystal
In the earlier literature, there was a impedance measurement that returns a
tendency to ascribe anomalous EQCM zero value for R2 ; in the event that R2 is
behavior to viscoelastic effects; in the nonzero but constant with changing condi-
absence of quantitative models for these tions (e.g. film redox state), then the film is
phenomena, this was not surprising. This nonrigid but frequency changes (e.g. with
problem no longer exists, since Eq. (5) redox state) are not viscoelastically driven.
suggests a simple test for ‘‘rigid’’ layer Aspects of loading with nonrigid films
behavior. An elastic film does not dissipate have been considered by several au-
energy, that is, will contribute nothing to thors [18–22]. The primary case of interest
the electrical equivalent circuit resistance. in the electrochemical context is load-
In this case, in Eq. (5), R2 = 0 and L2 is ing with a rigid layer (the electrode), a
proportional to the inertial mass of the viscoelastic film (commonly, though not
film. One can then show [19] that
necessarily, a polymer), and then a New-
L 2 fS 2f 2 M tonian fluid, schematically illustrated in
f = − = √S (6) Fig. 3. (The rigid layer component may
2L1 ρq µq
also include material entrapped within sur-
where L1 and L2 are represented in face features [23].) The characteristic me-
Fig. 2(b), fS is the series resonant chanical impedances of the two nonrigid
Liquid
Viscoelastic film
‘Rigid’ layer
Crystal Electrodes
Fig. 3 Schematic diagram of the propagation of an acoustic shear wave

launched by a TSM resonator loaded with a viscoelastic overlayer and
exposed to a fluid. Note the progressive zero, significant, and dramatic
attenuations of the wave on moving from the rigid layer (electrode plus
surface feature-entrapped material) to the viscoelastic solid to the fluid. The
acoustic decay lengths in these three regions are, respectively, infinity,
√
[2G/{1 − G /G}]1/2 /(ω ρ f ), and 1.4[G /ρf ]1/2 /ω; in the latter two
instances, typical values are 2 and 0.2 µm.
components are j ωhf ρf , which corresponds to additivity

ωρ η 1/2 of a rigid film and the fluid. In the
L L
ZL = (1 + j ) (7) latter case, the film is sufficiently thin
2 (acoustically thin) that there is negligible
for the liquid and acoustic deformation across it. Bandey and
coworkers have generalized Eq. (10) to the
Zf = (Gρf )1/2 (8) case of an arbitrary number of nonrigid
layers [19]. As will now be revealed, one
for the film, where ρf is the film density can use such an equation to predict
and G its shear modulus: the behavior of multiple nonpiezoelectric
layers of known properties, but not
G = G + j G (9) to determine (from experimental data)
unknown layer properties.
in which G is the shear storage modulus,
There is an additional phenomenon wor-
G is the shear loss modulus, and j =
√ thy of mention for viscoelastic films – film
−1.
resonance. As film thickness increases, an
One implication of the treatment of
acoustic phase delay (φ) develops across
Sect. 2.7.2.1.1 is that the impedances of
the film. One can show [1, 24] that the
serially deposited rigid layers are additive;
phase shift developed across a film of thick-
this is because each successive layer expe-
ness hf and shear modulus G is given by
riences the full acoustic motion for the res-

onator, transmitted without loss through ρf

each underlying layer. However, nonrigid φ = ωhf Re
G
layers do not behave in this fashion: as one
moves away from the resonator surface, √ 1 + G /G
= ωhf ρf (11)
each successive layer experiences only a 2|G|
fraction of the acoustic displacement en-
tering the underlying layer. Consequently, When the phase shift reaches a quarter
the component impedances of the sepa- of the acoustic wavelength (φ = π/2),
rate loadings (here, film and fluid) are no bearing in mind that reflection at the
longer simply additive. For a viscoelastic outer (film/solution) interface introduces
film of thickness hf immersed in a semi- a phase shift of π, the ‘‘round trip’’
infinite Newtonian fluid, the total surface for a wave launched at and returning to
mechanical impedance is given by [18, 19] the resonator surface corresponds to one
wavelength. This results in constructive
ZL cosh(γ hf ) + Zf sinh(γ hf ) interference between the returning wave
ZS = Zf
Zf cosh(γ hf ) + ZL sinh(γ hf ) and the next outgoing wave. Provided
(10) that the reflectivity coefficient at the
where the complex wave propagation film/solution interface is close to unity
constant γ is defined as j ω(ρf /G)1/2 . A (a condition fulfilled when the film and
little algebra reveals that Eq. (10) has the fluid characteristic impedances are very
‘‘required’’ limiting forms: (1) when the different), there will then be a dramatic
film is very thick the term in brackets change in coupling of energy from the
approaches unity and ZS → Zf , so the resonator into the film; this is termed
resonator sees only film and no fluid; and film resonance. In terms of the observed
(2) when the film is very thin ZS → ZL + frequency response, there is a sharp
increase in admittance and an increase first described by Bruckenstein [16]. Since

in resonant frequency. As the system then, several variants on this have been
passes through resonance, for example, described and commercial instrumenta-
with further increase in film thickness, the tion is now readily available. Significant
response returns back to the preresonance extensions of the original design include
trend of decreasing resonant frequency low noise circuitry [26], rapid data acqui-
and admittance. sition [27], a purpose-built cell [28], and a
The sequence of events as the system dual QCM in which a fluid-exposed ref-
passes through film resonance is illus- erence crystal compensates for changing
trated schematically in Fig. 4 [25], in which fluid properties [29].
the total (observed) impedance (full line) Beyond this, but still restricting attention
is the combination of the crystal (dotted to frequency measurement (i.e. gravimet-
line), and film (dashed line) components. ric mode), the concept of AC electro-
As film thickness increases (moving from gravimetry has been developed in a series
frame 1 to frame 5), the film component of papers by Gabrielli and coworkers.
moves rapidly downwards in frequency. Although applications of this methodol-
As it passes through the crystal compo- ogy to specific systems are discussed in
nent, the prediction is that the total res- Sect. 2.7.3, some of the papers provide spe-
onance will – transiently – split into two cific details on the generic principles [30]
peaks (frame 3). As the resonance con- and instrumentation [31, 32]. The essential
dition is passed (frame 5), the response concept is that, analogous to conven-
returns to a more ‘‘normal’’ appearance. tional electrochemical admittance, ĩ/Ẽ(ω),
More detailed theoretical exploration of one can define an electrogravimetric-
this phenomenon [25] shows that both the transfer function, m̃/Ẽ(ω), as a func-
absolute and relative values of the shear tion of frequency, ω. Other useful
modulus components have significant in- quantities are the mass/charge-transfer
fluence over the resonator response. In function, (m/q)(ω), and the partial
particular, the double-peak nature of the mass/charge-transfer functions [32]:
response shown in Fig. 4 is only predicted
miS m mi 1
for low loss films (i.e. films for which the (ω) = (ω) ∓
loss tangent, G /G < 1). E E j ω F ZF (ω)
(12)
for the anion (i = A; minus sign) and
2.7.2.2 Experimental Methodology the cation (i = C; plus sign), where
ZF (ω) is the conventional electrochemi-
2.7.2.2.1 Basic EQCM Technique The cal impedance. The mass/charge-transfer
usual configuration is for one of the elec- function can be regarded (with the in-
trodes on the crystal to be exposed to clusion of appropriate constants) as a
air and the other to be exposed to the generalized version of the normalized
electrolyte; the latter is then the working mass change, MF /Q, and the partial
electrode in a standard three-electrode elec- mass/charge-transfer functions are anal-
trochemical cell. The basic electronic cir- ogous to the quantity φi defined [33, 34]
cuitry suitable for driving a TSM resonator as the algebraic sum of the total mass
exposed to a liquid and maintained un- change and the contribution of either the
der electrochemical potential control, and anion (leaving the cation, i = C) or cation
measuring the resonant frequency, was (leaving the anion, i = A) contribution to
175 a1 b1 4.2 µm 0.175

Z1
125 Z2 0.125
[mS]
[kΩ]
Z
Y
Z1 + Z2
75 0.075
25 0.025
175 a2 b2 4.8 µm 0.175
125 0.125
[mS]
[kΩ]
Z
Y
75 0.075
25 0.025
175 a3 b3 0.175
5 µm
125 0.125
[mS]
[kΩ]
Z
Y
75 0.075
Peak 1 Peak 2
25 0.025
175 a4 b4 5.2 µm 0.175
125 0.125
[mS]
[kΩ]
Z
75 0.075 Y
25 0.025
175 a5 b5 0.175
5.8 µm
125 0.125
[mS]
[kΩ]
Z
Y
75 0.075
25 0.025
4.25 4.75 5.25 5.75 4.25 4.75 5.25 5.75
f f
[MHz] [MHz]
Fig. 4 Simulation of TSM resonator frequency be measured experimentally). Resonator base
response in the vicinity of film resonance. Panels frequency: 5 MHz. Film parameters:
1–5 represent increasing film thickness (as ρf = 1 g cm−3 , G = 108 dyn cm−2 ,
labeled). Panels a show impedance components G /G = 0.1. (Reproduced from Ref. [25] with
for the resonator (dotted line), the film (dashed permission from the American
line), and the composite resonator (full line). Chemical Society.)
Panels b show calculated admittance (as would
the mass change. In principle, the dynam- 30 MHz fundamental QCM for in situ
ics of different contributory processes (e.g. application has been described [35] and
ion and solvent transfers) can be resolved shown to have the appropriately increased
in the frequency domain, analogous to sensitivity. Although the sensitivity gain
their resolution in the time domain in, is clearly significant, in fact this is com-
for example, a chronoamperometric exper- monly not the limiting factor for studies
iment. Through the use of complex plane of multilayer films. The issue of making
representations of the admittance, one can measurements at higher harmonics will
in principle identify the individual mo- be returned to in the context of acous-
bile species contributions. Although this tically thick films (see Sect. 2.7.3.7.2), in
ac technique is unquestionably instrumen- which the intention is to use variations of
tally more complicated, it allows detailed frequency as a means of varying timescale.
exploration of timescale effects (via the Generally, on the basis of the above com-
frequency domain) with all the usual ad- promises, the operating frequency is in the
vantages of impedance methods. range 5 to 10 MHz. For resonators using
The resonant condition for the TSM AT-cut quartz crystals with f0 = 10 MHz,
device is that the crystal thickness be the Sauerbrey equation reduces to
an odd multiple of half the acoustic
wavelength, λ = vs /f . Since the shear
M(g cm−2 ) = −4.426 × 10−9 f (Hz)
wave velocity, vs , is equal to (µq /ρq )1/2
(14)
(presuming quartz, q, as the piezoelectric
Typically, the electrode area is on the order
material):
of 0.2 cm2 , so the mass sensitivity is on the

1 µq order of 1 ng (ca. 10 pmol for a ‘‘typical’’
hs = (13) adsorbate). In addition to the quantitative
2f ρq
aspects of the Sauerbrey equation (high
for the 5 and 10 MHz crystals most sensitivity), Eq. (14) highlights the gen-
commonly used in electrochemical ap- erality of detection and absolute nature
plications of the QCM, practical devices of the sensing process. These attractive
have nominal thicknesses of 0.0334 and features contributed to a long and success-
0.0167 mm, respectively. Selection of crys- ful history of TSM resonators being used
tal operating frequency is based upon a for a range of gas sensing and gas/solid
trade-off between sensitivity and mechani- interfacial applications, in which surface
cal robustness: decreasing thickness leads accumulation of material from the gas
to higher sensitivity (see Eq. 1) but at the phase was the underlying physicochemical
cost of increased fragility, which makes process [1, 12]. For comparison purposes,
mounting the crystal in the cell more diffi- we note that the gaseous environment is
cult. One way to gain increased sensitivity rather less demanding in two respects than
is to use higher (necessarily odd) harmon- the liquid environment inherent to electro-
ics, for example, 30 MHz for a 10 MHz chemical applications: if one wishes to use
fundamental frequency device, although a bulk shear acoustic wave, the more frag-
in this case the sensitivity only increases ile thinner crystals can be used in a gas;
linearly with frequency, rather than the the lower attenuation associated with a gas
quadratic dependence of the fundamen- allows one to use surface acoustic wave
tal mode (see Eq. 1). A chemically milled (SAW) devices.
For typical fluids (water and most has been considered at varying levels
organic solvents), xL (at 10 MHz) is on the of sophistication [4, 37] and we do not
order of 170 nm. The effectively coupled consider it further here, except where it
mass of this fluid layer, expressed as an impinges directly on the study of other
areal mass density, is simply the product interfacial processes, notably viscoelastic
of xL and the fluid density. For a smooth film dynamics (see below).
10-MHz resonator, immersion in an
aqueous solution couples ca. 17 µg cm−2 2.7.2.2.2 Crystal Impedance Technique
of fluid, resulting in a decrease in the Section 2.7.2.1.2 described the principles
resonant frequency of ca. 3.8 kHz. Beyond by which film viscoelastic properties
this relatively small (cf. 10 MHz operating (shear moduli) could be determined
frequency) baseline ‘‘shift’’, the EQCM can from the surface mechanical impedance.
then be used in situ to study interfacial Experimentally, the latter can be
processes under potential control. determined by measuring the electrical
Two practical issues arise at this stage. impedance of the resonator in the vicinity
First, it is important to maintain fluid of resonance, exploiting the piezoelectric
properties constant during the experiment; effect to transform the electrical driving
in practice, this generally amounts to good function and the mechanical response
temperature control. For example, if the function. A less well-explored variant
liquid density changed by 1% (due to on this steady state approach is to
temperature drift of ca. 1 K), then the drive the system with an ac function,
component of f due to the liquid (see then open the circuit and monitor the
above) would change by ca. 40 Hz: this transient decay. These complementary
would be fatal for monolayer adsorption approaches have recently been brought
studies and significant, but not fatal, for within an elegant general framework by
studies of thick (e.g. polymeric) layers. Kanazawa [38]. Comparison of the various
In some cases, notably when the time methods is made in Fig. 5 [38]; the reader
intervals are extended or when the surface is referred to the original paper for
mass changes are small, this – rather than a detailed exposition of the strategies
the sensitivity of frequency measurement summarized by this figure. Essentially,
per se – can become the limiting factor. what the figure shows is that there are
One obvious solution to this problem is to two transient solutions (depending upon
measure frequency shifts with respect to whether one goes to a condition of open
a reference crystal immersed in the same circuit or short circuit) and a steady
solution. In practice, electrical connection state solution. In the latter case, one can
between the ‘‘working’’ and ‘‘reference’’ either proceed via an electrically based
crystals makes this nontrivial, but a approach (to determine the equivalent
dual quartz crystal microbalance (DQCM) circuit parameters) or a physically based
achieving compensation for fluid-based model that leads to the displacements
changes has been demonstrated [29, 36]. in the system. To date, the predominant
Second, most crystal – and thus overlying source of film viscoelastic properties is
electrode – surfaces are not smooth at the impedance data, so coverage of this topic
atomic level. Simplistically, fluid entrained here will focus on steady state methods,
within surface features behaves as though but one can expect the transient methods
rigidly coupled to the resonator. This issue to assume increasing importance. As
Quartz Film Liquid
Boundary
conditions
Q4
Q4 = 0 Φ0 Q4 ≠ 0
J = −jwe22
Q4
1
=0 J(w)
Q4
Short Open Equiv ckt Displ's.
circuit circuit f, R, L, Cs A, B, ...H
Steady
state
L solutions
Cs
R
Equiv ckt Equiv, open

transient Equiv, short Transient
solutions
Physical, open (frequency
Physical, short and
decay times)
Fig. 5 Schematic routes to transient and steady state solutions for the frequency
response of a TSM resonator upon the application of different electrical control
functions. Symbols explained in main text. Q4 defines the current–voltage relationship
for the resonator. (Reproduced from Ref. [38] with permission from The Electrochemical
Society.)
this powerful methodology is developed, associated with the crystal (e.g. R1 , C0 , C1 ,

experience of analogous electrochemical and L1 in the LEM). The second experi-
control functions leads one to speculate ment will provide parameters associated
that the different approaches represented with the liquid ηL and ρL ). In princi-
in Fig. 5 will yield subtly different insights, ple, one would then hope that the third
so that they will in fact be complementary measurement would provide appropriate
rather than competitive. parameters for the electroactive film.
In a ‘‘typical’’ EQCM experiment, one In the case of a rigid (acoustically thin)
could envisage being able to character- film, the frequency shift provides film
ize: (1) the bare crystal in air; (2) the mass (Eq. 1); this will be the sum of
‘‘immersed’’ bare crystal (by which it is the masses of the electroactive material
implied that one face is exposed to the elec- (e.g. metal, polymer, or metal oxide) and
trolyte solution and the other to air, so that the liquid that permeates it. Coulomet-
the crystal is not electrically shorted); and ric assay – which may be based on film
(3) the immersed film-coated crystal. The deposition or subsequent redox chem-
first experiment will provide parameters istry – provides the amount of electroactive
material, via Faraday’s law. Thus Q and f regime, in which the Sauerbrey equation
provide the two unknowns – the amounts applies. With the additional input of the
of electroactive material and of solution electrochemical charge data (or indeed, any
within the film. If required, the amounts other direct measure of film thickness),
of the two components (of known individ- one can then determine film thickness and
ual density) can be used to deconvolute the density (as described above) in this regime
film mass into film thickness and density; in which viscoelastic parameters have no
this requires the reasonable approximation bearing on the response. Second, one
of additive volumes, that is, a zero volume operates in the acoustically thick regime,
of mixing. in which viscoelastic properties do govern
In the case of a nonrigid (acoustically the response. The tactic is to make the
thick) film, the problem is underdeter- assumption – which is in fact made for all
mined: there are four unknowns (film of the approaches described – that the film
thickness, film density, and the two shear is homogeneous at all coverages. Firstly,
moduli components) and less than four this allows one to carry the acoustically
measurands. Several approaches to this thin regime data-derived film density into
problem have been adopted. One argu- the acoustically thick regime. Secondly, it
ment is that viscoelastic behavior is due allows one to project the charge–thickness
to high solvent swelling, so the film is liq- relationship from the acoustically thin
uidlike, that is, ηf ηL and G G [22, regime into the acoustically thick regime,
39]. An alternative approach is to assume so that coulometric data – even in the
a value for ρf and to fix the loss tangent, acoustically thick regime – provide film
tan α = G /G [22, 39, 40]. Another ap- thickness. Now, there is a unique solution
proach is to constrain ρf and hf according in terms of film shear moduli for crystal
to additional information [23, 41, 42], for impedance responses.
example, ρf must lie between ρL and ρP A flow chart demonstrating the protocol
(density of pure polymer) and coulomet- is shown in Fig. 6. The procedure has been
ric assay provides a lower bound (h0f ) for demonstrated for poly(3-methylthiophene)
hf corresponding to a slab of pure poly- films, by analysis of frequency response as
mer. In various ways, all these approaches a function of time during film electropoly-
are attempts to reduce a four-parameter merization: short (long) time responses
problem (in ρf , hf , G , and G ) to a two- represent the acoustically thin (thick) film
parameter problem, which is amenable to scenario [24]. Film mass (whether or not
an ‘‘unambiguous’’ solution based on crys- directly accessible from f data) de-
tal impedance data (frequency response fines the product hf ρf , so (as shown in
of the loaded crystal in the vicinity of Fig. 6 [24]) a plot of hf versus ρf is a
resonance). hyperbola. As film mass (polymer cover-
Although each of these procedures has age) increases, a series of hyperbolae are
some merit, none of them can claim generated. The acoustically thin film data
to be mathematically rigorous or totally (f and Q) define the unique solution
physically defensible. A recent solution (of the infinity of solutions on the hyper-
to the problem [24] is to split this four- bola) for ρf as indicated in Fig. 6 [24]; this
parameter problem into two separate two- value is projected across all the hyperbolae.
parameter problems, as follows. First, The procedure is reliant upon two fac-
one operates in the acoustically thin tors: knowing Faraday’s law for the system
Input Diagnostics Operations Output

Acoustically
thin ∆m = rf /c1 = hfrf hf
Y rf
Zs R2 = 0 ?
Q (a) rf ∝ Q ?
Acoustically N Make thinner film

thick (b)
Y Q = khf
Zs = f (Q)? hf
Acoustically
thick
G', G"
N Fit Zs
Semi-
infinite rf
Fit Zs
G', G"
Fig. 6 Summary of data interpretation strategy (described in main text) for

extracting viscoelastic parameters from film-loaded TSM resonator
frequency response. Input parameters (at left of diagram) are resonator
impedance and any selected parameter representative of film thickness
(here, charge, interpreted using Faraday’s law). Upper part of the scheme
relates to acoustically thin films (yielding hf and ρf ). Lower part of diagram
relates to acoustically thick films (yielding, with the help of hf and ρf , G and
G ). (Reproduced from Ref. [24] with permission from the American
Chemical Society.)
and film homogeneity. It worked well for one implication is that the uniform film
the system used, poly(3-methylthiophene), model is inappropriate. Previous method-
suggesting that the uniform film model ologies (see above) would presumably
was physically reasonable and that the also have returned physically unreasonable
doping level was accurately known. It is (and nonunique?) data, but could provide
recommended that the approach be tried no insight into the underlying physical
for other systems; in cases in which it does reason, if indeed they gave any direct indi-
not yield physically reasonable solutions, cation of a problem.
2.7.2.2.3 Combination of the EQCM with frequency have been shown to be a means
Other In situ Techniques The QCM has of calibrating tip–surface separation [54].
now attained a status comparable with There have been relatively few combina-
many other in situ techniques: electro- tions of the EQCM with the spectroscopic
chemistry drives the interfacial chemistry and optical techniques that have revolu-
and the QCM probes the constituent tionized interfacial electrochemistry over
processes. Although the combination of the last two decades. Some of these
electrochemical and gravimetric data is are considered further in Sect. 3 of this
very powerful, it is sometimes inadequate volume. Of these, simultaneous use of vis-
for solving complicated problems. For ex- ible spectroscopy [55], Fourier transform
ample, if an electroactive film is nonperms- infrared (FTIR) spectroscopy [56], and el-
elective (e.g. at high electrolyte concentra- lipsometry [57] are significant examples
tion), one would generally expect redox that should be more generally exploited.
switching to be accompanied by the ex- A more specialized example is the combi-
change of anions, cations, and the solvent. nation, via a special dual thin layer cell, of
On the basis of electrochemical (charge, the EQCM and differential electrochemi-
Q) and gravimetric (mass change, M) cal mass spectrometry DEMS for real-time
data only, this three-parameter problem is determination of volatile products and de-
posited or dissolved species [58]. The com-
underdetermined. One approach is to vary
bination of the EQCM with UV–visible
the concentration [34] (see Sect. 2.7.3.7.1),
spectroscopy is particularly interesting:
but this is lengthy and vulnerable to film
electrochemistry (via charge) provides the
aging effects. Another approach is the si-
total change in electron population within
multaneous introduction of an additional
the film, electronic spectroscopy provides
technique to provide the third piece of
insight into the partition of this charge
information. One example of this is the
among the various source(s) or sink(s), and
combination of the EQCM with probe
the QCM provides the identities of the ions
beam deflection (PBD; sometimes called required (by electroneutrality) to move in
the mirage effect) [43–47]. response to this change in film charge. A
The advantages of controlled mass very nice example of this is given by a study
transport have long been recognized in of polypyrrole redox switching [55], in
electrochemistry, as discussed in Chap- which the electroactive film was deposited
ter 2.4. In the present context, the el- on an indium tin oxide (ITO) electrode
egant combination of the rotating disk supported on a quartz TSM resonator, with
electrode (RDE) [48–50] and wall-jet elec- an uncoated ITO electrode/quartz crystal
trode (WJE) [51, 52] solution hydrodynam- used to provided a ‘‘blank’’ for the spectro-
ics with EQCM interfacial detection has scopic data. This particular study was in
proved to be very effective. fact rather more sophisticated, since it also
Combination with scanning electro- explored the viscoelastic properties of the
chemical microscopy (SECM), the subject polymer film using simultaneous resonant
of Chapter 3.3, has also been explored and resistance measurements. The authors
the tip–surface separation was found to in- systematically explored all relationships
fluence the crystal impedance [53]. Viewed between the electrochemical, acoustic, and
from the opposite perspective, the obser- optical data, of which a representative ex-
vation of stress effects on crystal resonant ample is shown in Fig. 7 [55]. Through
1.0
0.9
0.8
Absorbance
[a.u.]
0.7
0.6
0.5
0.4
1.4 1.5 1.6 1.7
Resonant resistance
[kΩ]
1.0
0.9
0.8
Absorbance
[a.u.]
0.7
0.6
0.5
0.4
−1.2 −1.0 −0.8 −0.6 −0.4 −0.2 0.0
Frequency shift
[kHz]
Fig. 7 Correlation of absorbance (at 800 nm) with resonant
resistance (left-hand panel) and frequency shift (right-hand
panel) during redox cycling of a polypyrrole film on an ITO
electrode supported on a 9-MHz TSM resonator.
Voltammetric experiment: scan rate 50 mV s−1 . Solution:
aqueous 0.1 mol dm−3 KClO4 . (Reproduced from Ref. [55]
with permission from The Electrochemical Society.)
this, they were able to demonstrate a clear Low-resolution imaging of polymer

correlation between the film color change film profiles has been achieved using a
(represented by a specific electronic band) scanning-electrode quartz crystal analyzer
and its mass (represented by the crystal res- (SEQCA) method [59]. In this technique,
onant frequency); given the different pos- illustrated in Fig. 8 [59], one uses a small
sible mechanisms for maintaining elec- probe electrode, whose location can be
troneutrality and the very different kinetics effectively rastered over the surface (al-
for electron, cation, anion, and solvent though, in practice, this was achieved by
transfers, this is a significant outcome. keeping the probe stationary and moving
Probe
Resonating region
Monitored surface
Electrode Quartz crystal
Fig. 8 Schematic illustration of the principle of the

scanning-electrode quartz crystal analyzer (SEQCA) operating in
‘‘overscanning’’ mode, with a small probe electrode on the
polymer/solution loaded side of the resonator; in ‘‘underscanning’’
mode, the full electrode and the small probe locations are reversed.
(Reproduced from Ref. [59] with permission from the Royal Society
of Chemistry.)
the sample), to excite the crystal at se- in situ applicability, high sensitivity, and
lective locations. Figure 8 [59] illustrates a generality of detection; in the acoustically
mode in which a small probe scans over a thin regime, there is the additional analyti-
polymer-coated crystal upper surface, with cal advantage that quantitation is absolute.
a normal electrode on the lower surface. Consequently, the technique has been
Since this approach will fail when the applied to almost every class of electro-
electrolyte concentration is high, a sug- chemical system. This contrasts with many
gested variant of the mode illustrated is spectroscopic techniques, in which aspects
one in which the small probe scans under of selectivity can yield considerable detail,
a crystal with a bare lower surface and a but only for a restricted set of systems.
normal film-coated electrode on the upper As will become apparent, the use of the
surface. Although this method lacks the EQCM as a gravimetric probe – operating
high spatial resolution of some other (e.g. in the acoustically thin regime – is a ma-
scanning tunneling microscopy (STM) or ture area, providing a broad coverage
SECM) imaging methods, one could imag- of materials, properties, and chemistry.
ine it might in future provide different Viscoelastic studies are a newer avenue
information such as local film viscoelastic- of enquiry that provide new challenges
ity associated with local variations in film and novel insights; since this topic nec-
solvation or redox state. essarily has a lower profile in previous
reviews [2–6], it is given greater coverage
2.7.3 here. The facility to extract unequivocal
Electrochemical Systems physically meaningful materials param-
eters (shear modulus values) [24] (see
2.7.3.1 Strategy and Organization Sect. 2.7.2.2.2) should provide an impetus
Having considered the types of phenom- for rapid future development.
ena one might expect to see using the
EQCM, we now consider application of 2.7.3.2 Metals
the EQCM to interfacial electrochemistry Metal films are, at first sight, obvious can-
and see how material properties influence didates for the application of the EQCM as
the nature of the response. In broad terms, a gravimetric sensor: these materials are
the EQCM has three attractive attributes: rigid, have relatively high relative atomic
mass, and the application of Faraday’s law use of a rotating EQCM [66]. Another tech-
to coulometric data should be unequivocal. nologically important example of this is
Commonly, metal redox systems of known electroless deposition. The generic situa-
redox chemistry (Mz+/0 ) are used to cal- tion is that metal ions (whether free or
ibrate the EQCM. The EQCM ‘‘nano’’ complexed) are reduced to the elemen-
version of Faraday’s classic Ag+/0 exper- tal metal by a solution reducing agent,
iment [16, 28, 60] is an obvious example Red (e.g. formaldehyde). Under operat-
and the Cu2+/0 system has been simi- ing conditions, the net current flow is
larly exploited [53]. Figure 9 [43] illustrates zero, as a consequence of balanced cur-
this calibration experiment for an EQCM rents for Mz+ reduction and Red oxidation.
combined with a PBD instrument. Sim- Mechanistically, the common requirement
plistically, one anticipates that the current is to obtain the partial currents as a
represents the electron flux, the mass rep- function of potential (and other chem-
resents the Ag deposited/dissolved, and ical variables) from the measured i –E
the optical deflection (2) of the laser beam curve. In an EQCM experiment, the elec-
passing across the face of the electrode is trochemical data provides the total (net)
the response of the concentration gradi- current and application of Faraday’s law
ent in solution. Correlation of the three to the QCM (mass) response provides
responses (in either integral or differential the partial current for Mz+ reduction;
mode) unequivocally demonstrates that the Red partial oxidation current is then
Ag+/0 redox chemistry is the sole inter- obtained by difference. A classic exam-
facial process [43]. ple is electroless Cu deposition, illustrated
As is well known for the QCM [12], in Fig. 10 [67]. In the more complicated
the mass sensitivity of the resonant case of Au electroless deposition from
frequency is not uniform across the dimethylborane solutions, the EQCM has
resonator surface: it is radially dependent, been used to provide evidence for for-
with a maximum at the center of the mation and reduction of an oxide layer
electrode and falling to effectively zero (we and for adsorption of a reactive inter-
do not discuss fringing effects here) at mediate [68] and shown to be a powerful
the edge of the electrode. The carryover technique for screening plating bath for-
of this spatial distribution result to the mulations [69].
EQCM has been demonstrated using the More detailed examination of metal
deposition of Cu [61–63] or Ag [64] ‘‘dots’’ deposition, using crystal impedance mea-
across the electrode surface. In a mirror- surements, shows that there can also
image experiment, localized laser-induced be damping effects [70]. In the case of
dissolution of a Ni–Cu alloy has been used Cu and Ag, these were shown to be at-
to map the spatial distribution of EQCM tributable to film morphology through the
mass sensitivity [65]. associated increase in viscous coupling
In other cases, the EQCM can provide a to the liquid. An attractive goal is the
means of differentiating between metal de- manipulation of metal surface morphol-
position and other processes. This is clearly ogy, and crystal impedance measurements
critical in electrowinning and a good exam- have been shown to provide useful in-
ple is the deposition of Co, which has been sights during electrodeposition of nanos-
distinguished from the accompanying pro- tructured Pt from liquid crystalline plating
cess of hydrogen evolution by the elegant baths [71].
0.8
0.6
[mA cm−2]
0.4
j
0.2
0.0
−0.2
(a) −350 −250 −150 −50 50 150 250
3000
2500
2000
[ng cm−2]
1500
∆M
1000
500
−500
(b) −350 −250 −150 −50 50 150 250
250
150
50
[µrad]
−50
q
−150
−250
−350
−350 −250 −150 −50 50 150 250
E (vs Ag/Ag+)
(c) [mV]
Fig. 9 Calibration experiment for a combined EQCM/PBD
instrument, using the electrodeposition of Ag. Panels (a), (b), and (c),
respectively, show current, mass change, and beam deflection
responses to a cyclic voltammetric experiment (scan rate: 20 mV s−1 ).
Working electrode: Au (area 0.22 cm2 ) on a 10-MHz AT-cut quartz
crystal. Probe beam: 633 nm He–Ne laser parallel to Au electrode at a
distance of 155 µm. Solution: 1 mmol dm−3 AgNO3 /0.2 mol dm−3
HClO4 . (Reproduced from Ref. [43] with permission from Elsevier.)
400
iox
200
[µA]
inet
i
0
ired
−200
−0.9 −0.8 −0.7 −0.6 −0.5
E vs Ag/Agcl
[V]
Fig. 10 Mixed potential diagram for an electroless Cu deposition bath.
inet is the measured (total) current, ired is the determined from the rate of
Cu deposition (rate of QCM frequency change) using Faraday’s law, and
iox was then determined by the difference. Working electrode: Cu (area
0.34 cm2 ) on 5-MHz quartz crystal. Solution contained 5 mmol dm−3
CuSO4 , 10 mmol dm−3 Na2 EDTA, 50 mmol dm−3 Na2 SO4 , and
50 mmol dm−3 HCHO at pH 10.8. Temperature: 70 ◦ C. (Reproduced
from Ref. [67] with permission from The Electrochemical Society.)
2.7.3.3 Metal Oxides solution, consequent upon the interfacial

For broadly the same reasons as metals, pH change driven by electrochemical hy-
metal oxides/hydroxides are amenable to drogen generation. The data show how the
study by the EQCM. This is true for films deposition rate can be varied by current
that are electroactive or passivating, and control. The slope of the plot, in conjunc-
which may be formed by chemical, elec- tion with the value of the current value,
trochemical, or thermal routes. Given the was fitted to an empirical model for depo-
fraction of elements in the periodic table sition efficiency. More generally, one could
forming such materials and the breadth consider it as providing an effective molar
of applications – including corrosion, en- mass (defined as MF /Q) for the deposit,
ergy storage, and electronic devices – the for example (under different experimental
technological importance of this area is conditions to those employed in the cited
immense. study) providing film water content. In the
Prominent among the electroactive context of Ni(OH)2 film redox chemistry,
metal (hydr)oxides studied is nickel hy- a key mechanistic issue is the competi-
droxide. The EQCM has proved to be tion between the interfacial transfers of
extremely useful in exploring aspects of the metal cation (commonly K+ or Li+ ) and
electrochemically driven precipitation of hydroxide anion or protons in order to
Ni(OH)2 and its subsequent redox chem- satisfy electroneutrality and of the sol-
istry. Figure 11 [72] shows EQCM data vent in order to satisfy activity constraints.
for Ni(OH)2 deposition from Ni(NO3 )2 The EQCM [73, 74] has been extremely
110 0.50 mA
100 0.25 mA
90 0.10 mA
80 0.05 mA
70
60
Mass
[µg]
50
40
30
20
10
0
−10
0 1 2 3 4 5 6 7 8 9 10
Time
[min]
Fig. 11 Plots of electrode mass change versus time for the electrochemical
deposition of Ni(OH)2 from 0.2 mol dm−3 Ni(OH)2 solution at different
imposed currents (as indicated). Electrode: Au (area 0.2 cm2 ) on 10-MHz
quartz crystal. (Reproduced from Ref. [72] with permission from The
Electrochemical Society.)
informative in this respect, especially when film, like α-Ni(OH)2 , shows a mass in-
coupled with a laser-deflection technique crease upon oxidation, but redox cycling
to monitor stress effects [75] or probe beam progressively converts it to the β-form,
deflection to extract individual ion and sol- which shows a mass decrease upon oxida-
vent fluxes as functions of potential or tion. There is, however, a subtle contrast
timescale [76]. This system is also com- between the phase change behavior of
plicated by phase changes; the structural the Co-modified material and the parent
changes (predominantly between α- and β- Ni(OH)2 , studied earlier by Scherson and
Ni(OH)2 ) result in significant changes in coworkers [77], in that the inclusion of Co
internal volume, which dramatically influ- prevents the shift in oxidation potential
ence the gravimetrically observable solvent (i –E response) to more positive poten-
exchange signature [77, 78]. tials, even though it has virtually no effect
In the search for battery materials on the α- to β-phase transition that is
with better performance characteristics, so graphically seen through the redox-
the parent nickel hydroxide system has driven solvent-transfer process (seen in
been modified by the inclusion of other the M –E response of Fig. 12). The salu-
metal ions. The EQCM has been used tary lesson from this example is that the
to monitor redox-driven ion and solvent electrochemical goals may tempt one to
transfers in sol–gel derived nickel–cobalt interpret the gravimetric data in terms of
oxide films [79] and, through the solvent- the charge carrying species (ions), but in
transfer signature shown in Fig. 12, phase fact a neutral species (solvent) can be dom-
changes in electroprecipitated Co-Ni(OH)2 inant. The unexpected positive outcome is
films [78]. An as-prepared Co–Ni(OH)2 that one learns, indirectly from the solvent
500
60th cycle
400 50th cycle
40th cycle
Mass change
300
[ng (a.u.)]
30th cycle
20th cycle
200
10th cycle
6th cycle
100
2nd cycle
0
0.25 0.30 0.35 0.40 0.45 0.50 0.55
Potential
[VHg/HgO/1 M KOH]
Fig. 12 Mass change versus potential responses to multiple
redox cycling of an electroprecipitated Co/Ni(OH)2 film
(containing 10.2% Co) supported on an Au electrode. Solution:
1 mol dm−3 KOH. Potential scan rate: 10 mV s−1 .
(Reproduced from Ref. [78] with permission from The
population changes, as much about the and Yu and coworkers have used the
film structural as about the film redox EQCM to study both electrochemically [86]
changes; the former controls the dynamics and chemically [80] deposited films. The
of the latter. electrochemical performance is critically
The EQCM has also been used to dependent upon the ‘‘microtexture’’ of
study other metal oxide systems of rele- the film, that is, the presence of solvent-
vance to energy storage and other applica- filled pores and cavities. The EQCM, as
tions, including MnO2 [80], PbO2 [81, 82], illustrated by the data of Fig. 13 [80] for
and sol–gel derived films of V2 O5 [83], chemically deposited material, is able to
WO3 [84], and Nb2 O5 [85]. Although film separate out the charge and solvation ef-
deposition (e.g. nucleation and growth fects. By comparing the current and mass
of PbO2 [81]) can obviously be studied, responses shown in Fig. 13 [80], along with
the primary exploitation of the technique the analogous anodic half-cycle data, it was
has been probing the redox-driven sur- shown that micropores within the film
face composition changes, for example, are filled with solvent, electrolyte, and dis-
associated with lithium insertion: corre- solved manganese species when the film
lation of charge and mass data provides is oxidized (to MnO2 ), but that these pores
access to solvent effects that would be are emptied when the film is reduced (to
difficult to observe by other means. A sys- Mn(OH)2 ). The authors highlighted the
tem of long-standing interest is MnO2 , fact that expulsion of species from the
160
PIVC3
PIIIC3
NPC3
PIIC3
80 PIC3 PPC3
[µA cm−2]
IO2
Intensity 0
−80 I-E O2-evolution

profile 2Icap
(a) −160
−10
a1
3
[mC cm−2]
Q•3a1
Charge
−20
Q3a2
−30 Q-E
profile
−40
(b) −50
60
A3
50 M-E ϑ2-3
Mass variation
profile
[µg cm−2]
40 D3
E3 O2-evolution
30
20 ∆5 ∆4 ∆3 ∆2 ∆1
10 B3
0
−0.8 −0.4 0.0 0.4 0.8
Potential vs Hg/HgO
(c) [V]
Fig. 13 Responses of (a) current; (b) charge; and (c) film mass change to
a linear cathodic potential sweep of a chemically deposited (via MnSO4
reduction of KMnO4 ) MnO2 film. Electrodes: Au (electroactive area
1.37 cm2 , piezoactive area 0.34 cm2 ) on 5-MHz quartz crystal. Solution:
1 mol dm−3 KOH. Potential scan rate: 1 mV s−1 . Lines marked with filled
circle and diamond symbols, respectively, represent contributions due to
oxygen reduction and capacitive currents, respectively. Vertical dashed
lines separate different regimes, represented by the annotations and
described in detail in the original work. (Reproduced from Ref. [80] with
permission from The Electrochemical Society.)
oxide structure into the fluid phase does are effectively inertially coupled to the un-
not necessarily correspond to mass loss derlying resonator.
(as perceived by the EQCM), since these A rather different oxide system is the
species may be trapped within pores that Cu2 O film electrodeposited from alkaline
copper lactate solution. At constant cur- is shown in Fig. 15 [92], which depicts
rent, the potential undergoes periodic os- the frequency increase (mass loss) for a
cillations. These are reflected in changing zinc film (representative of the surface of
QCM f –t slopes (equivalent, under con- a galvanized steel) upon exposure to an
stant current conditions, to f –Q slopes) industrial alkaline cleaner, as part of a
that were interpreted in terms of the forma- cleaning and chromating procedure. The
tion of an alternating Cu/Cu2 O multilayer response clearly has two components. By
nanostructure [87]. examining their pH and oxygen sensitivi-
In other systems, oxide films can be pas- ties, it was possible to deduce that the faster
sivating. An elegant example of the power process is associated with chemical disso-
of the EQCM is the use of frequency lution of an oxide layer on the outer zinc
change data to calculate the fraction of surface and that the slower process is an-
the total current associated with oxide odic dissolution of zinc. As Fig. 15 shows,
film growth on Cr [88]. Analogous to the the first stage is unaffected by the age of the
approach in electroless deposition (see bath, but the second stage is dramatically
above), the EQCM can be used to provide slower in a ‘‘used’’ bath; this is attributable
selectivity in terms of the process of inter- to increased solution zincate concentration
est, here oxide film growth. Film thickness in the latter case. The EQCM has also been
data estimated using the EQCM were used to show that zinc corrosion is signifi-
in excellent agreement with X-ray photo- cantly affected by illumination [93]. (As an
electron spectroscopy measurements, but aside, it is worth mentioning that in both of
of course the EQCM has the advan- these studies the complications of Au/Zn
tage that the data are acquired in situ, alloy formation were avoided by deposition
in time-resolved fashion, and at a frac- of an intervening layer of Cu between the
tion of the instrumental cost. Potential Au electrode and the Zn overlayer.)
steps were executed in two stages across Corrosion is generally recognized to in-
the passive domain and the film growth volve spatially localized and temporally
responses compared with two models: discrete events. This is significant in the
the interface model (growth limited by context of using the EQCM as a probe,
growth/dissolution at the film/solution in- since the Sauerbrey equation represents
terface) and the high-field model (growth a spatially averaged sensitivity, use of
limited by ion conductivity through the metal deposition/dissolution to establish
film). As Fig. 14 [88] shows, either model the bell-shaped spatial sensitivity function
can explain growth at lower potentials across the resonator surface was discussed
(0.0–0.4 V) but only the interfacial model in Sect. 2.7.3.2. Consequently, mass loss
can describe growth at higher potentials (corrosion) will give rise to a frequency
(0.4–0.8 V). The method has been exchange (increase) that is dependent upon
tended to stainless steels containing Cr, the location of the event on the sur-
Mo, or W [89–91]. face. Furthermore, the appearance of pits
The closely related process of corrosion on the surface will change the coupling
is also amenable to study by the EQCM. to the fluid. This complicated situation
Since the species transferred across the has begun to be considered recently by
interface (metal ions) are generally rela- Gabrielli and coworkers [94], who attached
tively heavy, a gravimetric sensor is very thin foils of authentic 304 stainless steel
sensitive. An excellent example of this (rather than depositing model layers) to
Fig. 14 Passive film growth on Cr 0.25

following application of a potential step
from: (a) 0.0 to 0.4 V and (b) 0.4 to
0.8 V. Full line is film thickness 0.20
Thickness change
calculated from QCM frequency
response (with widely spaced dots 0.15
[nm]
indicating uncertainty, 2σ ). The
predictions of the interface and
0.10 EQCM
high-field models (explained in main
2s
text) are also indicated by the line-styles
0.05 IFM
marked IFM and HFM, respectively.
HFM
Electrodes: Cr film on Au supported on
10-MHz AT quartz crystal. Solution: 0.00
0.1 mol dm−3 H2 SO4 + 0.4 mol dm−3 0 20 40 60 80 100
Na2 SO4 . (Reproduced from Ref. [88] (a) Time
with permission from The [s]
0.25
0.20
Thickness change
IFM
0.15
[nm]
0.10
HFM
0.05
0.00
0 20 40 60 80 100
(b) Time
[s]
5
200
4
150
[mg m−2]
3
[kHz]
−∆m
∆f
100
2
1 50
A B
0 0
0 5 10 15 20 25
Time
[s]
Fig. 15 Mass changes of Zn films at open-circuit potential exposed
to new (full line) and used (dash-dot line) alkaline cleaning solution.
Electrodes: Zn film on Au (with Cu overlayer) supported on 10-MHz
AT-cut quartz crystal. Solution: RIDOLINE industrial alkaline cleaner,
pH 12.43. Temperature: 316 K. (Reproduced from Ref. [92] with
permission from the Royal Society of Chemistry.)
an EQCM, then observed transient fre- passage. Effectively, the slope of a M –Q

quency change events. Scanning electron plot is used to provide an effective molar
microscopy (SEM) images showed both mass that can be compared with the stoi-
large and small pits, with the latter pre- chiometries of various model processes. In
dominating to the extent that the spatial one case, coupling the EQCM with a flow
averaging intrinsic to use of the Sauerbrey injection stripping method was used to
equation was a reasonable approximation. obtain the composition of CdSe films [98].
Inhibition of corrosion (through deliber- The case of cathodic deposition of
ate addition of organic inhibitors) has been Cu(2−x) Se from a Cu(SCN)4 3− /HSeO3 −
studied on Ni [95] and Fe [50]; in many solution is interesting, in that deposi-
respects this aspect is an adsorption prob- tion at – 0.1 V gives a M –Q relation-
lem, and indeed Landolt and coworkers ship consistent with four-electron reduc-
used the EQCM to test different isotherms tion of HSeO3 − to Se, and at slightly
for the selected inhibitors. more negative potentials yields the de-
Chemically more complex oxide systems sired Cu(2−x) Se. At much more negative
studied by the EQCM include barium and potentials (E < −0.65 V) metallic Cu is
strontium tungstates and molybdates on deposited. In a voltammetric experiment,
W and Mo, in which dissolution and pre- illustrated in Fig. 16 [103], one thus sees a
cipitation processes result in the formation mass (frequency) versus charge signature
of surface ternary oxide layers [96]. that is attributable to the initial deposition
of Se, then a chemical process that con-
2.7.3.4 Semiconductors sumes surface Se to produce a Cu(2−x) Se
Electrochemical methods can be used for film, and eventually the codeposition of Cu
the preparation of semiconductor mate- and Se as Cu(2−x) Se. When Cu(2−x) Se was
rials in thin film form. This class of deposited, the composition was found to
preparation methods generally centers on be independent of potential, but morpho-
a redox-driven electrodeposition and is logical (simplistically, roughness) changes
thus amenable to study by the EQCM. occurred and (presumably through liquid
Chemically, the complications arise be- coupling effects) lead to changes in the
cause the stoichiometry of the materi- f –Q slope with potential.
als may be variable, according to the
particular conditions employed. For exam- 2.7.3.5 Metal Complexes
ple, Rajeshwar and coworkers and Kemell Most of the electroactive metal complexes
and coworkers have studied the com- studied using the EQCM are in fact
plexities of deposition (by oxidation or polymeric materials, which are discussed
reduction of solution precursors) of a in Sect. 2.7.3.7. A few examples that
range of chalcogenides, including CdS, highlight nonpolymeric metal complex
In2 S3 , and CdTe [97], CdSe [98], PbS [99], chemistry are presented here.
PbSe [100], PbTe [101], CuInSe2 [102], and Surface-attached microcrystals of Fe(η5 -
Cu(2−x) Se [103]. Generally, electrogravi- C5 Ph5 ) exposed to aqueous and wa-
metric data can provide insight into the ter/acetonitrile media can be made to
processes involved and the composition undergo Fe(III/II) redox chemistry. In
of the resultant material by the applica- addition to the expected ion and solvent
tion – at the nanogram level – of Faraday’s transfers accompanying the redox chem-
law to the frequency response to charge istry, the EQCM revealed some unexpected
4.0 × 10−5
2.0 × 10−5
0.0
[A]
−2.0 × 10−5
I
−4.0 × 10−5
−6.0 × 10−5
−8.0 × 10−5
−0.8 −0.6 −0.4 −0.2 0.0

E vs Ag/AgCl
(a) [V]
Scan direction
0
−50
−0.23 V
[Hz]
−100
∆f
−0.06 V −0.31 V
−0.36 V
−150
−200 −0.65 V
−0.22 V
−3 × 10−3 −2 × 10−3 −1 × 10−3 0

∆Q
(b) [C]
Fig. 16 Data from a voltammetric deposition study of the Cu2−x Se system:
(a) current versus potential and (b) frequency change versus charge plots.
Electrode: Au (area = 0.236 cm2 ) on 5-MHz AT-cut quartz crystal. Solution:
50 mmol dm−3 CuCl + 1 mmol dm−3 SeO2 + 4 mol dm−3 KSCN/pH 3.
Potential scan rate: 10 mV s−1 . (Reproduced from Ref. [103] with
permission from Elsevier.)
characteristics associated with phase be- the Co complex redox chemistry and for
havior [104]. Co(II)-ethylenediamine com- Cu electrode dissolution in the presence of
plexes have been explored as reducing halides, which are believed to act as bridges
agents in electroless deposition baths (see in the adsorption of the Co complex on the
also Sect. 2.7.3.2). As part of this process, Cu surface [52].
the EQCM, in a wall-jet configuration, has Electrocrystallization, as an example of
been used to extract partial currents for an electrochemically driven deposition
process, is amenable to study using redox switching, and solvation changes.

the EQCM. The technique has been Although other techniques have undoubt-
used to show that electrocrystalliza- edly contributed to our understanding
tion of the partially oxidized cyanoplat- of this material, recent EQCM stud-
inum complexes K1.75 Pt(CN)4 (H2 O)1.5 ies show in quantitative manner that
and K2 Pt(CN)4 Cl0.3 (H2 O)3 from aqueous initially formed Prussian Blue corre-
solutions of K2 Pt(CN)4 occurs by pro- sponds to Fe4 3+ [Fe(CN)6 ]·6H2 O (referred
gressive nucleation and one-dimensional to as ‘‘insoluble Prussian Blue’’) and
growth and that their dissolution occurs at that reduction (in potassium-containing
open circuit [105]. electrolytes) to Everitt’s salt generates
A rather clever use of the EQCM in K2 Fe2+ [Fe(CN)6 ] [107, 108]. The product
a molecular recognition context is the of reoxidation is KFe3+ [Fe(CN)6 ] (referred
binding of ferrocene species to surface- to as ‘‘soluble Prussian Blue’’). Subse-
bound β-cyclodextrin hosts [106]. By using quent redox cycling then corresponds
a carboxylate-substituted ferrocene and
to reversible switching between soluble
changing the solution pH and/or ap-
Prussian Blue and Everitt’s salt. One con-
plied electrode potential, it was possible
sequence of the high stability of this
to control the guest charge type (negative,
system is that EQCM data are generally of
neutral, or positive, visualized in a Pour-
extremely high-quality: M –(or f )–Q
baix diagram) and thereby manipulate the
host–guest interactions. plots are linear, show little hysteresis and
are very reproducible, allowing one to pur-
sue mechanistic issues in some detail.
2.7.3.6 Other Inorganic Materials Complete redox switching (after the
Although, in the context of electrode sur- initial ‘‘insoluble/soluble Prussian Blue’’
face modifiers, inorganic materials are conversion) in aqueous solutions of al-
sometimes perceived to lack some of the
kali metal and ammonium cations was
synthetic flexibility associated with poly-
found [108] to yield effective molar mass
meric materials (see Sect. 2.7.3.7), they
changes MF /Q close to, but distinguish-
commonly possess greater stability and
able from, the cation molar mass. The in-
robustness to physicochemical manipula-
ference was that there was a small amount
tion. This has provided the motivation to
study the interfacial properties of a number of solvent transfer (less than one water per
of electroactive inorganic materials using transferred cation), the direction of which
the EQCM. was dependent upon the electrolyte cation.
Prominent among these is Prussian (It is worth noting that the effective molar
Blue (simplistically, ‘‘ferric ferrocyanide’’), mass change for reduction in K+ solu-
which can undergo reduction to a ‘‘fer- tion was found to be essentially identical
rous ferrocyanide’’ material, known as in the two studies reported (30 [107] and
Everitt’s salt. There are three issues re- 31 [108] g mol−1 ), a level of agreement sel-
lating to the redox chemistry of this dom seen in studies of polymer films (see
material upon which the EQCM has shed Sect. 2.7.3.7) for which minor variations in
light: irreversible compositional changes film preparation, handling, and aging re-
accompanying redox cycling of freshly sult in apparent inconsistencies between
deposited Prussian Blue, ion transfer re- different studies.) In the case of Prussian
quired to maintain electroneutrality upon Blue films immersed in NaCl solution (but
protected by a Nafion layer), the more so- may result in a major change in the ob-
phisticated ac electrogravimetric method served (net) mass response.
was able to show that sodium ions are de- In an extension of their Prussian Blue
solvated prior to entry into the Prussian study, Ogura and coworkers [110] pre-
Blue film [109]. pared bilayers in which a polyaniline film
In contrast with reversible redox switch- was overlaid onto a Prussian Blue in-
ing of precycled films, the initial redox ner layer. The interesting feature here
cycle of Prussian Blue yielded more com- is that, on charge grounds, polyaniline
plex mass responses with much greater (see Sect. 2.7.3.7.2) is a barrier to the nor-
irreversible solvation changes. This is mal cation-transfer-based electroneutrality
dramatically illustrated by the data of mechanism exhibited by the Prussian
Fig. 17 [108], which shows the i –E and Blue/Everitt’s salt redox chemistry (see
M –E responses to the first redox cy- above). For thin polyaniline outer layers,
cle for each of two nominally identical the cation-transfer mechanism is achieved,
films, one exposed to H2 O and the other but thick polyaniline films preclude this. At
to D2 O. Although the difference in sol- intermediate outer layer thicknesses, there
vent molar masses is quite small (11%), is a more complex mechanism in which
solvent transfer is such a large compo- redox-driven K+ cation expulsion (on oxi-
nent of this initial redox-driven structural dation of the Everitt’s salt form of the inner
change that the overall responses are very layer) from the inner layer to the outer layer
different. Somewhat reminiscent of the α- is electrically balanced by Cl− anion entry
/β-Ni(OH)2 case, this underscores the fact into the outer layer. Further oxidation of
that when the observed response is a sum the bilayer (oxidizing the polyaniline) re-
of individual transfer components (com- sults in expulsion of the K+ (‘‘inherited’’
monly in opposing directions, for example, from the inner layer), the Cl− normally
in response to a volume constraint), a rel- incorporated in this polyaniline oxidation
atively small variation in one component process already being present to neutralize
1.0 3.0
H2O
0.5 2.0
[mA cm−2]
1.0
[mg cm−2]
0.0
∆M
0.0
I
−0.5
−1.0
D2O H2O D2O
−1.0 −2.0
−1.5 −3.0
0.6 0.4 0.2 0.0 −0.2 0.6 0.4 0.2 0.0 −0.2
E vs Ag/AgCl E vs Ag/AgCl
(a) [V] (b) [V]
Fig. 17 Current (left panel) and mass change (right panel) responses during the first redox
cycle of Prussian Blue films (deposition charge 10 mC cm−2 ). Electrodes: Au
(area = 0.32 cm2 ) on 6-MHz AT-cut quartz crystals. Solution: 0.5 mol dm−3 KNO3 in H2 O
(full lines) or D2 O (dotted lines). Potential scan rate: 10 mV s−1 . (Reproduced from Ref. [108]
with permission from Elsevier.)
the previously internally transferred K+ deposition of calcium carbonate scale

ions. Polymer/polymer bilayers will be dis- on Au [116], the electrodeposition of
cussed in Sect. 2.7.3.7.2, but this example clays [48], and the redox chemistry of elec-
neatly shows how spatial configuration of troactive metal coordination complexes
materials at the interface can alter the ([Ru(bpy)3 ]2+/3+ , [Ru(NH3 )6 ]2+/3+ , and
responses normally associated with the [Fe(CN)6 ]4−/3− ) ion-exchanged into clay
individual components. films [117]. The EQCM has also been ap-
The EQCM has also been used to study plied to the study of C60 [118, 119] and
redox-driven ion and solvent transfers in carbon nanotubes [120, 121], although the
Prussian Blue–related materials, in which data have not yet lead to clear or specific
one or other of the iron species is re- conclusions.
placed by another metal. These include
ferric ruthenocyanide (‘‘ruthenium pur- 2.7.3.7 Polymers
ple’’) [111], silver hexacyanoferrate [112], In terms of electrochemical systems, poly-
and nickel hexacyanoferrate [113, 114]. mer films are arguably the main class of
EQCM data for the latter case – the best
material to which the EQCM has been
characterized of this set – was shown to
applied. This section is split into a sep-
be consistent with a broadly electrolyte
arate discussion of redox, conducting,
cation-dominated electroneutrality mech-
and insulating polymer films; this is to
anism (as for the parent Prussian Blue
some extent a matter of organizational
material), but with some evidence for elec-
convenience and the recurrence of simi-
trolyte anion participation, in the redox
lar phenomena and properties should be
process.
noted. Some attention has been given to
An interesting example of the use of
film electrodeposition, but most studies
the EQCM to monitor both interfacial
relate to ion and solvent transfers ac-
electrochemical and subsequent chemi-
companying film redox switching. As will
cal processes is the deposition of Hg on
become clear from the examples selected,
thin TSM resonator-immobilized graphite
the EQCM has been central to a funda-
electrodes [115]. As the authors conceded,
mental change in the way electroactive
there are a variety of complicating fac-
polymer films are now viewed, particularly
tors – notably spatially inhomogeneous
with regard to the role of competitive (an-
deposition of Hg droplets and viscoelas-
ion versus cation) ion transfers, solvation
tic phenomena associated with these fluid
effects, and polymer dynamics (as man-
deposits. Nevertheless, in chloride media,
ifested through viscoelastic phenomena).
when electrodeposition of elemental Hg
It is unlikely that these insights would
was followed by setting the electrode to
open circuit, the electrode unequivocally have resulted from purely electrochemical
continued to gain mass; this was inter- measurements, such as the voltammetric
preted in terms of the comproportionation experiments that dominate the study of
of deposited Hg(0) and dissolved Hg(II), these systems.
to produce surface-immobilized Hg2 Cl2 .
There are a number of other inor- 2.7.3.7.1 Redox Polymers The term ‘‘re-
ganic systems that have been studied dox polymer’’ is used to denote a poly-
using the EQCM, although in somewhat mer that contains electroactive groups
less detail. These include electrochemical that are discrete, localized entities. The
electroactive groups may be pendant to a choice of mechanisms by which it can

the polymer backbone, as in polyvinylfer- satisfy electroneutrality upon redox switch-
rocene, or may be an integral part of it, ing: anion ejection or cation injection upon
as in poly(xylylviologen) [PXV]. In either reduction and vice versa upon oxidation.
case, they function as essentially inde- Quite generally, the balance of these com-
pendent entities, and charge transport is peting ion transfers may be dependent
by a hopping-type mechanism, involving upon applied potential, film charge state,
sequential self-exchange between groups experimental timescale, and solution ionic
in the oxidized and reduced forms of strength. The power of the EQCM is its
the couple that are in appropriately close ability to distinguish between them, under
proximity. either thermodynamically controlled (long
A wide range of electrochemical (i–V–t) timescale) or kinetically controlled (short
methods has been used to study the over- timescale) conditions. At the crudest level,
all extent and dynamics of redox polymer anion and cation transfers must be in op-
film switching, as manifested through posite directions, so the sign of the mass
the electron flux (current) or population change will be different. At a more sophis-
(charge). However, these methods can- ticated level, one can interpret the EQCM
not provide insight into the associated frequency response to extract quantitative
transfers of ions and solvent: this has population changes and fluxes of the ions
been one of the main contributions of involved as functions of applied potential
the EQCM to this area. The electroneu- or film charge.
trality constraint is generally appreciated: A very graphic example of this facility
the film must overall be electrically neu- to distinguish anion- versus cation-based
tral, so electron transfer (in this context, mechanisms for maintaining electroneu-
electrochemically driven) into or out of trality is given in Fig. 18 [122] for the
the film at the electrode/polymer interface poly(1-hydroxyphenazine) system. In this
must be balanced by an electrically equiv- example, the cation is proton so the
alent ion transfer at the polymer/solution competing ions have very different mo-
interface. Under permselective conditions, bilities – a feature that will be returned
the transferred ion will be the counterion to shortly. Beyond the fact that it clearly
(anion for a positively charged film and has two distinct components, the i − E
cation for a negatively charged film). How- curve is relatively featureless, that is, un-
ever, this is strictly only true at very low helpful. In this particular case, the film
ionic strength – a situation deliberately is sufficiently thin (acoustically thin) that
violated in most electrochemical experi- the EQCM frequency response can be in-
ments – and at equilibrium – a situation terpreted in gravimetric terms. During the
generally violated in potentiodynamic ex- first part of polymer oxidation (‘‘step I’’ in
periments. In concentrated electrolytes (in the figure), the frequency decreases (mass
principle, above ca. 0.1 mol dm−3 ), activ- increases): this is consistent with anion
ity arguments can be used to show that entry. During the second part of polymer
there may be significant amounts of coion oxidation (‘‘step II’’ in the figure), the fre-
(and an equivalent amount of counte- quency increases (mass decreases): this is
rion, commonly referred to as salt in the consistent with cation exit. This qualitative
ion-exchange literature) partitioned into conclusion is supported by the PBD data in
the film. Consequently, the system has panel (b) of Fig. 18. Quantitative analysis
0.10
0.00
Current
[mA]
Poly (1-hydroxyphenazine)
−0.10
1 M HClO4
(a) −0.20
0.24
Beam deflection
0.16
[mrad]
0.08
0.00
(b) −0.08 Step I Step II

Frequency
[Hz]
20 Hz
−0.4 −0.2 0.0 0.2 0.4 0.6 0.8 1.0

Electrode potential (RHE)
(c) [V]
Fig. 18 Current, beam deflection, and frequency change responses
(panels (a–c), respectively) of a poly(1-hydroxyphenazine) film to a
redox cycle under cyclic voltammetric conditions. Electrodes: glassy
carbon (probe beam experiment) and Au (area = 0.36 cm2 ) on 5-MHz
AT-cut quartz crystals (QCM experiment). Solution: 1 mol dm−3
HClO4 . Potential scan rate: 50 mV s−1 . (Reproduced from Ref. [122]
of the mass change data reveals that a the full potential range transiently drives
counterflux of solvent occurs (to which within the film an excess anion popula-
the probe beam is insensitive), to an ex- tion change, which then relaxes on longer
tent that broadly maintains a constant film timescales.
volume. Interestingly, shorter timescale Many organic redox species have
experiments than those of Fig. 18 reveal protonatable groups (e.g. amines), so in
that oxidation using a potential step across aqueous media one might expect proton
transfer to be a plausible means of acetate species. Kinetically, this promises

satisfying electroneutrality, even if only a very neat – albeit unexpected – situation
transiently. An interesting, and slightly un- since proton transfer in the aquated acidic
usual, example of coupled electron/proton medium of the film is about as fast an
transfers is provided by the polythionine ion transfer as one could expect in a redox
system. In strong acid media, there is a polymer, whereas transfer of the solvent (a
complicated electrolyte-dependent compe- larger species, whose motion is not field as-
tition between anion and proton trans- sisted) is expected to be much slower. The
fers [123]. However, in acetic acid media, mass–charge data shown in Fig. 19 [124]
the coordinating power of the carboxylic allow one to test this speculation. The
acid is such that it remains in the film, linear M –Q relationship during film
regardless of the polymer redox state, reduction indicates that solvent expulsion
and proton (nominally the coion) trans- is rapid enough to keep pace with the cou-
fers rather than acetate (nominally the pled e− /H+ transfers, but the dramatic
counterion) [124]. The system (under these curvature and hysteresis associated with
conditions) is further unusual in that vol- film oxidation indicates that solvent entry
ume constraints restrict the amount of is very slow; given the compactness of the
solvent that can enter the film, to the film, this is entirely plausible.
extent that a simple coordination model The behavior described above for poly-
with integer solvation stoichiometry ap- thionine is a simple and helpful introduc-
plies. The overall redox process is associ- tion to the concepts of redox-driven solva-
ated with an effective molar mass change tion change, participation of coion, kinetic
of 16 g mol−1 , interpreted as a coupled effects (in terms of entry vs. exit rates for
2e− /2H+ transfer in one direction and a given species and of charged vs. neutral
a single solvent molecule transfer in the species transfers), and temporal resolution
opposite direction, as represented by the of mobile species transfers. However, this
scheme-of-squares mechanism shown in level of simplicity is rare. Quite generally,
Scheme 1. Chemically, this is rationalized the extent of solvent transfer is governed
by considering that the solvent molecule by an activity constraint [125, 126]: at equi-
entering upon polymer oxidation takes up librium, the activity of the solvent within
a coordination site made available by an the film must equal the activity of the sol-
internal (acetate/acetic acid) proton trans- vent in the solution (an effectively constant
fer and, upon polymer reduction, the water value, given the huge excess present). A
must leave the film to make this site avail- change in film redox (charge) state results
able to the more coordinatively powerful in a change in solvent activity coefficient
Scheme 1 Scheme-of-squares
−2e − 2H+
representation of the redox-switching [(ThH42+(A−)2)•X] [(ThH+(A−))•X•(HA)]
mechanism of a polythionine film
exposed to acetic acid solution. Th
represents a monomeric thionine unit,
−H2O +H2O
A represents acetate, and X can be
either a water or acetic acid molecule.
(Reproduced from Ref. [124] with
permission from the American [(ThH4+(A−)2)•(H2O)•X] [(ThH+A−)•(H2O)•X•(HA)]
Chemical Society.) +2e + 2H+
250
200
150
[ng cm2]
∆m
100
50
−50
−500 0 500 1000 1500 2000 2500 3000
Q
[µC cm2]
Fig. 19 Mass change versus charge plots for redox cycling under
voltammetric conditions of a polythionine film ( = 13.8 nmol cm−2 ) in
aqueous acetate solution. Electrode: Au (area = 0.23 cm2 ) on 10-MHz
AT-cut quartz crystals. Solution: 0.1 mol dm−3 total acetate, buffered to pH
4.5. Potential scan rates: 5 (), 50 (), 100 () mV s−1 . (Reproduced from
Ref. [124] with permission from the American Chemical Society.)
within the film, thus requiring a change in study in this field [127] established this
solvent concentration (population) within to be the case for PVF films exposed
the film. At the present state of knowl- to aqueous ClO4 − and PF6 − electrolytes.
edge, there is no prospect of being able to Subsequent studies explored this system
calculate a priori values of solvent activity in more detail, showing that solvent entry
coefficients within an electroactive poly- accompanied anion entry (to the extent of a
mer film. Given the diversity of chemical few water molecules per anion) and that, at
species involved in this class of materials, higher electrolyte concentrations (typically
one must anticipate diverse behavior and at in excess of 1 mol dm−3 ), permselectivity
best hope to rationalize the experimental failure occurred, that is, salt also parti-
observations; EQCM measurements can tioned into the film. Somewhat similar to
provide the data to facilitate this approach. the case of polythionine discussed above,
One of the most widely studied electroac- the individual mobile species transfers
tive polymer film systems is polyvinylfer- (ClO4 − , H2 O, and Na+ ClO4 − ) consequent
rocene (PVF). PVF films can be deposited upon redox switching of PVF were re-
by electroprecipitation from organic sol- solved by monitoring the current and
vents (as discussed below) then studied frequency (mass) responses to a potential
as surface-immobilized films in aqueous step [34]. The current response is com-
media. Redox switching between the two plete relatively rapidly and is accompanied
ferrocene (Fc) states can be represented by a rapid mass change; on electroneu-
as a Fc+/0 process. On electroneutrality trality grounds, this is dominated by the
grounds, one would then expect anion transfer of a charged species, shown by
transfer into (out of) the film upon oxi- the M –Q relationship to be ClO4 − .
dation (reduction), and indeed a very early Thereafter, at essentially zero current,
there is continued substantial mass that one driver for solvent transfer is to
change; on electroneutrality grounds, this fill/vacate space within the film in order to
must be the transfer of a net neu- facilitate ion transfers. To the extent that
tral species – solvent under permselective the film is a kinetically frozen matrix – the
conditions and solvent + salt under non- dynamics and elasticity of polymer film
permselective conditions. Thus, it was matrices will be considered shortly – this
possible to separate out, using time and is a qualitatively reasonable notion. On
electrolyte concentration as variables, the the other hand, the exchange of ions and
individual contributions to the overall re- solvent between the film and the solution
dox process. phase are often considered without regard
It is intuitively obvious that while ther- to any film volume constraint. There is
modynamic studies (focused upon popu- thus a clear need to establish a means
lations, when the fluxes have decayed to to test whether the volume is fixed and
zero) will benefit from considerations of to explore the consequences of the result.
the integral quantities charge and mass, In a recent study, the effect of impos-
kinetic studies will benefit from consid- ing or removing a volume constraint on
erations of the corresponding differential the partition of the solvent and salt was
quantities current and mass flux. A recent studied [129]. Theoretically, this was repre-
study of PVF redox-switching dynamics sented through a permselectivity index (R),
has shown how the charge (or potential) defined as the ratio of the concentrations
dependence of the ratio of the fluxes of sol- of coion and polymer charge sites. Exper-
vent to counterion (ρ = jsolvent /janion ) can imentally, this was explored through the
be of particular benefit [128]. The varia- change in permselectivity index (R), de-
tions (or independence) of ρ with potential, rived from redox-driven film mass changes
charge, or scan direction as functions of under nonpermselective (high electrolyte
timescale (e.g. potential scan rate in a concentration) conditions. It was shown
voltammetric experiment) have mechanis- that R, under given conditions, can be
tic diagnostic value. Plots of ρ versus Q and expressed in terms of a concentration
E provide qualitative identification (even parameter (S) and a thermodynamic pa-
on immediate visual inspection) of the rameter (T ), where S is the ratio of
rate-limiting process, and more detailed the solution electrolyte concentration to
quantitative exploration can be expected to the film charge site concentration, and
yield kinetic parameters. Given that most T contains the salt partition coefficient
modern data acquisition systems store di- and activity coefficients. For the simple
rectly (or give facile access to) the relevant case of a ‘‘+1/0’’-type redox polymer (e.g.
fluxes, future utilization of this analytical PVF) exposed to a 1 : 1 electrolyte, one can
methodology is encouraged. show that
PVF is one of the few systems for which R = 0.5[−1 + (1 + 4S 2 TOx
2 1/2
) ] (15)
quantitative data have been obtained in
any detail; since this must be the ulti- Thus, a plot of (R + 1/2)2 (from the
mate aspiration for any useful study, it EQCM data) versus S 2 should yield a
is worth noting how far the present state straight line, from which the salt par-
of knowledge allows one to proceed. In tition coefficient (within the thermody-
rationalizing observed solvent transfers, namic parameter, T ) might be estimated.
there is a common tendency to consider More detailed criteria (described within
the original work [129] and based on the thereby automatically compensating for
sensitivity of the film concentrations in- changes in fluid properties, for exam-
volved in Eq. (15) to film volume changes) ple, viscosity and density. The value of
determine whether or not this simple plot this facility has been demonstrated for
will be linear. For the case of PVF exposed the case of PVF films exposed to the
to concentrated solutions of NaClO4 , it 1 : 2 electrolyte disodium naphthalene-1,5-
was deduced that the films shrink signifi- disulfonate (Na2 NDS) [130], a relatively
cantly (ca. 35%). The partition coefficients rare example of an EQCM study using
for the net neutral species into the ox- anything other than a 1 : 1 electrolyte. The
idized form of the polymer were found crucial added capability provided by the
to be Ksalt,Ox 1.0 and Ksolv,Ox 0.15. DQCM (cf. EQCM) is that one can expose
Given the concentration of water in the the modified electrode to a series of so-
bulk solution (55.5 mol dm−3 ), the latter lutions (here, of different concentration)
corresponds to a water concentration in and attribute the change in frequency re-
the film of ca. 9 mol dm−3 . When com- sponse to a change in film composition,
pared with the redox site concentration of free of any change in solution properties.
ca. 3 mol dm−3 , this indicates that even By this means, it was possible to demon-
the oxidized form of PVF is quite hy- strate film permselectivity for [Na2 NDS]
drophobic. <64 mmol dm−3 and progressive perms-
Even in the light of the above advances, electivity failure at higher concentrations.
if one were to identify a limitation in The salt permeation characteristics were
the EQCM methodology, it would prob- such that film oxidation resulted in elec-
ably be that it is a differential method, troneutrality maintenance by ingress of
that is, the measured frequency changes a mixture of NDS2− and NaNDS− in
can only lead directly to film population the high concentration regime; while one
changes (whatever the species involved). might have arrived at this qualitative no-
In many cases, this is adequate for the tion by other means, the DQCM has for
application. However, there are cases in the first time provided quantitative infor-
which absolute populations are helpful. mation.
For example, in the case of the PVF films A rather neat example of the manip-
discussed here, oxidation is usually asso- ulation of film electroneutrality mainte-
ciated with an increase within the film of nance mechanism is provided by the re-
a few (5–10, according to the electrolyte dox behavior of poly(xylylviologen) (PXV)
and other conditions) solvent molecules films, in which the nominal counterion is
per redox site. However, placing a phys- poly(styrenesulfonate) (PSS). In this case,
ical interpretation on this is somewhat the ‘‘counterion’’ is effectively immobile,
reliant upon knowing the ‘‘baseline’’ for and the question arises as to whether the
this change, that is, whether it represents PXV and PSS ion charges compensate
a relatively minor or major population each other or are separately compensated
change. The experimental problem is the by (other) electrolyte ions. This will clearly
absence of a suitable ‘‘reference’’ point. An have implications for the possible sources
instrumental approach designed to pro- and sinks of ionic charge available to satisfy
vide this is the DQCM [29, 36] discussed electroneutrality upon PXV redox switch-
in Sect. 2.7.2.2.1. This device uses a refer- ing, and the EQCM is ideally placed to
ence crystal exposed to the same solution, make the distinction. It was found that
the result was very dependent upon the changes occur. From an electrochemical
ratio of PXV : PSS in the film as cast [131]. perspective, the consequence of solvent
The f –E traces for voltammetric redox within the film is the more facile elec-
conversion of the viologen were inter- tron/ion motion, manifested by a less
estingly complex: in all cases, they were positive anodic peak potential on subse-
nonmonotonic within a given half-cycle quent redox cycling.
and showed substantial mass changes well All the above applications of the EQCM
beyond completion of the redox process to study redox polymer films have relied
as indicated by the cessation of current upon the TSM resonator as a gravimetric
passage. The short timescale (kinetically probe of surface populations and changes
controlled) mechanism must therefore in- therein, that is, have been for acoustically
volve both anion and cation transfers. thin films (see Sect. 2.7.2.2.1). The same
The long timescale (thermodynamically was true for the inorganic films discussed
controlled) situation was interpreted as in Sects. 2.7.3.2–2.7.3.6. However, the
predominant – but not exclusive – cation effect of solvent entry is very different for
transfer for PXV : PSS = 1 : 2 films and polymer films: in this case, the solvent can
vice versa for PXV : PSS = 2 : 1 films. act as a plasticizer, ultimately resulting in
All the ion and solvation changes dis- the film becoming sufficiently softened so
cussed so far in this section relate to that it is acoustically deformed by the TSM
chemically reversible processes. Anecdo- resonator motion. In other words, one
tally, it is widely accepted in the redox moves from the acoustically thin regime
polymer modified electrode literature that to the acoustically thick regime. Under
the first redox cycle of a newly deposited these circumstances, the TSM resonator
film is atypical. Although the reasons and no longer responds predominantly to film
processes are chemically rather different, mass; rather, the frequency response is
this is reminiscent of the ‘‘first cycle’’ ef- viscoelastically controlled. In the context
fect discussed in Sect. 2.7.3.6 for Prussian of film redox chemistry (cf. deposition),
Blue films. PVF provides a typical example this means that the response provides
of this first cycle, or ‘‘break-in’’, effect. The insight into the polymer matrix – often
initial EQCM response to PVF oxidation perceived to be the ‘‘fixed’’ component
in water (after its deposition from an or- of the system – rather than the mobile
ganic solvent, typically dichloromethane) species (ions and solvent) that permeate
is quite different from that of a previ- it. Consequently, measurements on the
ously cycled film [132]. The result is most same system in both the acoustically thin
clearly demonstrated by considering the and thick regimes provide complementary
mass flux – or, better still, the difference insights into film dynamics.
between the total mass flux and the elec- Theoretical treatments for the acousti-
tron/ion flux – as a function of applied cally thick film case have only recently
potential or film charge. Such plots show been developed (see Sect. 2.7.2.1.2); thus
a once-only pulse of solvent into a new this topic has not been reviewed previ-
film; this solvent (typically ca. five sol- ously. The field is at an interesting stage of
vent molecules per ferrocene redox site) development: the existence of viscoelastic
is retained (‘‘trapped’’) within the film phenomena in this context are generally
thereafter and provides the baseline upon accepted, the basic theoretical concepts
which subsequent redox-driven solvation have been delineated [18–20, 22, 23, 25],
and these analyses have been applied to equivalent circuit parameters [134], but
a small number of experimental studies. only more recently have the tools to extract
We now consider some representative ex- physically useful parameters (shear mod-
amples of redox polymer films that show ulus components) been developed. The
these effects. earlier discussions in this section on the
The TSM resonator frequency response use of the EQCM as a probe of mobile
shows a peak admittance (resonance), species transfers accompanying PVF redox
whose location on the frequency axis is chemistry in aqueous media were based on
determined by the capacitive and induc- the notion of film rigidity; for the studies
tive elements of the system and whose described, crystal impedance spectra sup-
magnitude is determined by the resistive port this idea. However, when exposed
elements in the system (in which these ele- to the dichloromethane solutions typically
ments can, for example, be represented by used for the electroprecipitation of PVF
the equivalent circuits in Fig. 2). An acous- films, the situation is very different. The
tically thin (‘‘rigid’’) film will contribute basis of the film deposition strategy is that
nothing to the resistance of the system, reduced PVF (PVF0 ) is soluble in CH2 Cl2 ,
that is, will not provide a mechanism but oxidized PVF (PVF+ ) is not. However,
for energy dissipation. Consequently, film although PVF+ A− (where A− represents
population (mass) changes will be man- the electrolyte anion, e.g. ClO4 − or PF6 − )
ifested by a translation at constant shape is insoluble, it is relatively lyophilic, that is,
and amplitude of the resonance peak along highly solvated. The consequences of this
the frequency axis; this translation will are seen in crystal impedance spectra ac-
be described by the Sauerbrey equation quired dynamically during deposition, as
(see Eq. 6 for parameterization in terms shown in Fig. 20 [134]. Progressive depo-
of the inductive element in the equivalent sition of polymer on the electrode surface
circuit of Fig. 2b). Contrastingly, an acous- results in an increase in inertial mass and
tically thick film will contribute additional thus the anticipated decrease in resonant
resistance to the system; that is, internal frequency. However, there is also a dra-
friction will be a mechanism for energy dis- matic decrease in admittance, which can be
sipation. Consequently, film population or parameterized as an increase in the resis-
composition changes will be manifested tance in the motional arm of the equivalent
by changes in shape and amplitude (as circuit of Fig. 2(b). When the film is suf-
well as location) of the resonance peak. ficiently thick, acoustic deformation can
These contrasting variations in peak shape also be seen in aqueous electrolytes. This
characteristics provide a diagnostic for film is nicely illustrated by a crystal impedance
(non)rigidity. study of PVF break-in (for thicker films
PVF provides a good example of how than those discussed above and in acid,
the EQCM can be used to explore solva- rather than salt, media), in which the ir-
tion processes via their influence over film reversibly incorporated solvent can lead to
viscoelastic properties. It also provides a viscoelastic behavior [135].
representative example of progress in this The sensitivity of film viscoelastic prop-
area: an early study [133] of PVF electrode- erties to the ambient medium applies
position clearly identified the presence of not only to the solvent but also to the
film viscoelastic phenomena, which were electrolyte. This is illustrated for the
subsequently parameterized in terms of case of poly[Os(bipy)2 Cl(PVP)10 ] (where
3.5
Bare crystal in solution
3.0
2.5 0.7 V
1h
2.0 0.0 V
2h
[mS]
G
1.5 3h
4h
5h
1.0 6h
0.5
0.0
9.92 9.94 9.96 9.98 10.00 10.02 10.04

f
[MHz]
Fig. 20 Crystal admittance spectra acquired during the deposition of a PVF
film. Electrode: Au (area = 0.23 cm2 ) on 10-MHz AT-cut quartz crystal.
Solution: PVF (2 mmol dm−3 in ferrocene units)/0.1 mol dm−3 TPA+
ClO4 − /CH2 Cl2 . At t = 0, potential stepped from 0.0 to 0.7 V. Numbers
adjacent to spectra indicate deposition times. (Reproduced from Ref. [134]
PVP represents polyvinylpyridine) films. concentration), the OsPVP-loaded TSM

These ‘‘OsPVP’’ films are rigid when device resonant frequency decreases upon
exposed to aqueous sodium perchlorate film oxidation, indicating a mass in-
solution, but viscoelastic when exposed crease consequent upon anion and a
to sodium p-toluenesulfonate (in cer- little solvent entry (essentially as for
tain concentrations) [136]. This is imme- PVF redox chemistry). Contrastingly, in
diately discerned by inspection of the concentrated p-toluenesulfonate solution,
crystal impedance spectra for films in the Au/OsPVP-loaded TSM device res-
the two media [136], but it is a salutary onant frequency increases upon film
warning that the variations of resonant oxidation: a gravimetric interpretation
frequency – as one would monitor in a would – inappropriately – suggest mobile
‘‘simple’’ EQCM experiment – show con- species expulsion.
trasting responses despite the essentially An additional driver of viscoelasticity
identical underlying physical phenom- changes is temperature, either for rea-
ena. In perchlorate solution (and for sons based on free volume at constant
solutions of moderate p-toluenesulfonate film composition (see the brief discussion
of Williams–Landel–Ferry (WLF) theory frequency and resistance with temperature

in Sect. 2.7.3.7.2) or through temperature- (Fig. 21). Starting with the high tempera-
dependent film composition (e.g. solva- ture (collapsed, and thus rigid) form and
tion). An example of the latter phe- progressively decreasing the temperature,
nomenon is the behavior of poly(vinylfer- entry of the solvent softens the film and
rocene-co-N -isopropylacrylamide) films as results in an increase in energy dissipation
a function of temperature, illustrated in and a decrease in resonant frequency; this
Fig. 21 [137]. There is a transition temper- is consistent with a shift from a rigid to
ature (Tt ) below which the film is highly a viscoelastic film. Further decrease of the
solvated (‘‘swollen’’) and above which it temperature, and entry of more solvent,
is poorly solvated (‘‘shrunken’’). In these eventually results in a drop in resonant
two solvation states, both film thermody- resistance and an increase in resonant fre-
namics (formal potential) and dynamics quency. Presumably, this is a consequence
(mobile species diffusion coefficients) are of decoupling the more distant regions of
very different. For each redox state, the the film from the underlying resonator; in
change in solvation state can be moni- terms of the model of Bandey and cowork-
tored through the change in TSM resonant ers [19], this might be interpreted as a shift
10000
Oxidized state
5000
[Hz]
∆f
Reduced
0 state
−5000
10 20 30 40
Temperature
(a) [°C]
1500 Fig. 21 Changes in resonant frequency

(panel A) and resonant resistance
Reduced state (panel B) for a poly(vinylferrocene-co-N-
1000 isopropylacrylamide) film
(VF :NIPAA = 1 : 13; dry film thickness
500 ca. 5 µm) as a function of temperature,
[Ω]
∆R
with the film maintained in the reduced

0 state (E = 0.0 V; dashed line) and in the
oxidized state (E = 0.5 V; full line).
Electrode: Pt (area = 0.2 cm2 ) on
−500
9-MHz AT-cut quartz crystal. Solution:
Oxidized state
0.1 mol dm−3 NaClO4 . Temperature
−1000 ◦
10 20 30 40 scanned from 35 to 10 C over the
course of ca. 1 h. (Reproduced from
Temperature Ref. [137] with permission from the
(b) [°C] American Chemical Society.)
from a finite to a semi-infinite viscoelastic for π-delocalization and create space for
film, of which only a portion is sampled by accompanying ion motion. In the first
the acoustic wave. As seen in Fig. 21, the case, the EQCM – as a gravimetric sen-
transition temperatures are different for sor – responds to ion transfer into/out of a
the two film redox states: Tt,Red < Tt,Ox . conducting polymer film and in the second
Thus, high temperatures (Tt,Red < Tt,Ox < case – as a viscoelastic probe – it responds
T ) result in the redox-driven intercon- to polymer dynamics, exactly as described
version of two rigid films (the simple in Sect. 2.7.3.7.1 for redox polymers.
case, interpretable in gravimetric terms), In the earlier electrochemical conduct-
low temperatures (T < Tt,Red < Tt,Ox ) re- ing polymer literature, the emphasis was
sult in the redox-driven interconversion of on coupled electron/ion motion. Accord-
two viscoelastic films, and intermediate ingly, within the EQCM literature in this
temperatures (Tt,Red < T < Tt,Ox ) result area, the gravimetric capability to probe
in the redox-driven interconversion of a ion transfers, for example, distinguishing
rigid (oxidized) and a viscoelastic (reduced) opposing anion versus cation motion, was
film. In summary, it is now possible predominant. This gravimetric approach
to achieve, and to monitor through the was consistent with the simpler instru-
EQCM, combined thermal and electro- mentation in use at that time, but more
chemical control of film viscoelasticity. recently, the capabilities of the EQCM with
regard to characterizing viscoelastic phe-
2.7.3.7.2 Conducting Polymers The term nomena have assumed a higher profile in
‘‘conducting polymer’’ is used here to this field.
refer to polymers with electronic charge Given that the generic issues are the
delocalization along the spine of the poly- same, it is not surprising that the EQCM
mer; the electroactive functionality is thus has been applied to all the key types of
a delocalized system, contrasting with materials in this category, including par-
the discrete electroactive functionalities of ent compounds and derivatives based on
‘‘redox’’ polymers. From an applications polypyrrole [32, 138–151], polyaniline [46,
viewpoint, the raison d’être for conduct- 57, 152–156], and polythiophene [41, 45,
ing polymers is that they show high (and 157–162]. In this review, selected exam-
controllable) electronic conductivity. It is ples (from over 100 papers on polypyrroles,
not the purpose of this review to dis- ca. 70 papers on polyanilines, and ca. 30
cuss these electronic aspects of conduction papers on polythiophenes) are chosen to
processes, but rather to explore two associ- provide specific illustrations of the types
ated phenomena into which the EQCM of phenomena observed and amenable to
brings insight. First, injection of elec- exploration using the EQCM.
tronic charge into a conducting polymer is An early example of the diagnostic value
intimately coupled to ion dynamics: injec- of the EQCM for exploring doping mech-
tion or removal of a charge-compensating anisms of conducting polymer films is
ion (‘‘dopant’’). Second, both the elec- the study by Naoi and coworkers [138]
tronic motion along the polymer spine of polypyrrole films prepared with an-
and the ionic motion through the polymer ions of different sizes. In the case of
matrix are functions of the polymer dy- films prepared with small anions, typified
namics: respectively, the rates at which the by perchlorate and tetrafluoroborate, elec-
polymer strands achieve internal planarity trochemical injection of positive charge
sites (oxidation) of the polymer had the and on long timescales p-toluenesulfonate
anticipated predominant effect of inter- transfer predominates.
calating anionic dopant ions (along with At low electrolyte concentration (typi-
some solvent). At the other extreme, film cally c < 0.1 mol dm−3 ), polypyrrole films
deposition in the presence of very large in both redox states are permselective at
polymeric anions, typified by polystyrene- equilibrium, that is, exclude ‘‘salt’’, but
sulfonate and polyvinylsulfonate, resulted the undoped film accumulates salt un-
in the anion being trapped within the film der kinetically controlled conditions [149].
in both polypyrrole oxidation states. Con- Thus, the first redox cycle from an equili-
sequently, polypyrrole redox chemistry brated reduced film must result in anion
resulted in cation transfer, but of course in entry, but the accumulation of some cation
the opposite sense to the transfer of small (salt) during the redox cycle allows a
anions. The use of intermediate-sized an- mixed anion/cation mechanism to occur
ions, represented by p-toluenesulfonate, in a subsequent redox cycle. This is a
was found to result in mixed redox-driven consequence of the typical experimental
anion and cation transfers. timescale (10–100 s) being much shorter
Qualitatively, this picture is relatively than the equilibration time for salt and/or
simple to understand in terms of the solvent transfer (up to 1000 s); extended
mobilities of the anions involved. Quan- holding of the film at a given potential
titatively, the picture is rather complex, eventually restores equilibrium. When one
particularly when the ion sizes are not too moves to thermodynamically nonpermse-
different. In the case of p-toluenesulfonate lective conditions (c < 0.1 mol dm−3 ), the
(with a small solution cation), anion trans- situation becomes much more compli-
fer is the thermodynamically ‘‘preferred’’ cated [150] since the presence of both
process, but the greater mobility of a small anions and cations in both redox states
cation makes the latter transfer kinetically opens up all the mechanistic possibilities.
more facile. Consequently, the dominant A significant difficulty in this type of
process is timescale dependent. Further- problem is visualizing film compositional
more, as indicated in Sect. 2.7.3.7.1, the changes. This is facilitated by the use of
presence of ‘‘salt’’ within the film is elec- the scheme-of-cubes [163], in which three
trolyte concentration dependent. Cations orthogonal axes can be used to repre-
(coions) can only be ejected from the film sent three elementary steps in film redox
if they are present in the first place, that switching, here film redox state (elec-
is, if the film is nonpermselective. The tron/anion population), solvation state
complexities of this process have only re- (solvent population), and permselectivity
cently been unraveled [149, 150] and the failure (salt population). The concept is il-
EQCM response (film composition) is de- lustrated for the case of polypyrrole films
pendent upon the experimental timescale exposed to sodium p-toluenesulfonate so-
(e.g. voltammetric scan rate), film history lution in Fig. 22 [150], in which there is
(first or subsequent redox cycle after equi- the added subtlety of representing cation
libration), and electrolyte concentration. transfer (a ‘‘fourth dimension’’) as the al-
Film mass changes during redox cycling gebraic sum of anion and salt transfers
are nonmonotonic: on short timescales in opposite direction, that is, a diag-
cation (sodium) transfer is the predomi- onal transfer across a cube face. The
nant mode of satisfying electroneutrality attractions of this model are its greater
+C
+C
+C +C −CA
−CA P+CA*
SP+ −CA
CA*
+S SP+
−CA CA*
+CA +CA
P 0CA P 0CA 0
P CA
−S +CA −S
−C −C −C
P+
(a) (b) (c)

Fig. 22 Scheme-of-cubes [163] representations of mechanistic pathways for polypyrrole film redox switching on: (a) long; (b) intermediate; and (c) short
timescale. The (x, y, z)-coordinate system represents (film redox state, solvent population, salt population): the left (right) hand cube faces represent
reduced (oxidized) polymer; the rear (front) faces represent less (more) solvated film; lower (higher) faces represent lower (higher) salt population.
Translations along the x, y, z-coordinates, respectively, represent E, C, and C processes in a traditional EC-type model; E is used to represent an
electrochemical step, but with electroneutrality maintained by cation transfer (cf. anion transfer for an E-type elementary step). A, C, and S annotations,
respectively, indicate transfers of anion, cation, and solvent species. The mechanisms indicated by the heavy lines are (a) E C C repetitively, (b) E C C in
the first half-cycle, subsequently reducing to E ,C as the C (solvation) step is kinetically frozen, (c) E ,C , progressively reducing to E as the C (salt
transfer) step is progressively kinetically frozen on decreasing timescales. (Reproduced from Ref. [150] with permission from the Royal Society
of Chemistry.)
2.7 The Electrochemical Quartz Crystal Microbalance
273
palatability rather than long sequences of this data, comparison of the observed
stoichiometric equations, and the fact that probe beam response with that calculated
the ‘‘axes’’ are directly related to selected for pure anion transfer.
components of the EQCM frequency re- The above applications were targeted on
sponse (which may be directly or only ion delivery. Complementary to this is
indirectly related to the charge). The par- targeted ion uptake. The EQCM has been
ticular feature highlighted in Fig. 22 is used to explore polypyrrole as a material
the effect of decreasing the experimental for facilitating removal from solution of
timescale (here, by increasing voltammet- heavy metal ions, such as lead, cadmium,
ric scan rate): this progressively ‘‘freezes cobalt, and nickel [140, 143].
out’’ each elementary step, starting with At a more sophisticated instrumenta-
the slowest (here, solvent entry), then the tional level than monitoring EQCM res-
next slowest (here, salt entry) until only onant frequency at constant or linearly
the fastest step (here, field-driven coupled swept potential (as used for the studies
electron/cation transfers) remains. Since discussed above) is the use of electro-
the effective timescale is also a function gravimetric impedance. The concept is
of film thickness, this provides a means essentially the same as conventional elec-
of correlating data for different polymer trochemical impedance, except that the
coverages.
response function is the film mass (TSM
It will be clear from the foregoing
device resonant frequency). This is a pow-
paragraphs that the ion transfer satisfy-
erful diagnostic technique, firstly through
ing electroneutrality upon redox switching
the location of different species’ responses
of polypyrrole films can be manipulated
in different quadrants of the traditional
by an appropriate choice of physical
complex plane plot and, secondly, through
and chemical parameters. This has lead
the facility to vary the timescale (see above)
to exploration of the ability to deliver
over a wide range. Using this approach,
ions from the film under electrochemical
control. Obvious candidates of interest rather detailed kinetic and isotherm data
are ions with therapeutic activity: since were obtained for various anion and cation
these tend to be relatively large – but and water (solvent) transfers into/out of
not polymeric – such systems are al- polypyrrole [32]. Interestingly, with the
most inevitably of the ‘‘mixed-transfer’’ benefit of the detailed information obtain-
type, so the EQCM has proved invaluable from this ac technique, one cannot
able in studying them. Examples in- only confirm that all species participate (as
clude salicylate [146, 147], naproxen [146], deduced from detailed analysis of voltam-
nicoside [146], chlorpromazine [141], and metric data [150]), but can also extract
heparin [142]. Figure 23 [147] provides an quantitative kinetic parameters and fur-
example of the time-dependent competi- ther show that they are film charge-state
tion between opposing anion and cation dependent. Kwak and coworkers have used
transfers, in this case with the additional this approach to study ion dynamics of
help of the PBD experiment. The goal polypyrrole containing small (nitrate) [151]
was to achieve dominant salicylate trans- and large (poly(styrenesulfonate)) [144] an-
fer, but the short time response is cation ions. These measurements again pro-
dominated, as graphically shown by the vide additional detail and sophistication
nonmonotonic mass transients and, using compared to the simpler voltammetric
0.0
E vs SCE
[mV]
−0.4
−0.8
(a)
[µA cm−2] 2000
1000
0
j
−1000
(b) −2000
12000
8000
[ng cm−2]
4000
∆M
0
−4000
(c) −8000
1500
1000
[µrad]
500
q
0
−500
−1000
0 50 100 150 200
t
(d) [s]
Fig. 23 Combined EQCM/PBD responses of a polypyrrole film to
redox switching in sodium salicylate solution. Electrode: Au
(area = 0.23 cm2 ) on 10-MHz AT-cut quartz crystal. Solution:
aqueous 0.5 mol dm−3 sodium salicylate. Potential step program as
shown in panel (a). Dotted line in panel (d) represents prediction,
based on current trace of panel (b), of salicylate but no sodium
transfer to satisfy electroneutrality. (Reproduced from Ref. [147] with
permission from The Electrochemical Society.)
measurements, for example, revealing dif- In Sect. 2.7.3.7.1, appropriate control of

ferent types of charge site (‘‘fast’’ and polymer composition of a redox polymer
‘‘slow’’) and associated ion population dy- (PVF) was shown to lead to the intro-
namics [151]. Detailed modeling of voltam- duction of viscoelastic phenomena and to
metric EQCM responses to ion transfers thermal sensitivity. For polypyrrole, de-
has also been used to deduce the presence position from micellar surfactant media
of ions in two environments, labeled ‘‘free’’ (dodecylsulfate and dodecylbenzenesul-
and ‘‘bound’’ [145], whose dynamics are fonate) also leads to changes in film mor-
different. phology and viscoelastic behavior [139].
Thermal sensitivity, based on temperature- that respond to film thickness, DC po-

dependent swelling as monitored using tential (film charge state), and electrolyte
the EQCM, resulted from deposition composition. By implementing the partial
in the presence of such polymers as electrogravimetric-transfer function ap-
2-acrylamido-2-methyl-1-propane sulfonic proach (see Eq. 12), this leads to isotherm
acid (AMPS) [148]. data for individual mobile species as func-
Polyaniline has proved an attractive tions of potential. This is a clear example
electroactive material in part because of detailed analysis of high-quality data
of its stability in aqueous media and providing a rich source of mechanistic
the possibility of (rapid) ionic charge information.
transport by protons. Thus, most of the The deposition of polyaniline has also
EQCM studies of polyaniline relate to film been studied using a rather elegant com-
redox switching, with the now familiar bination of the EQCM with ellipsometry,
goal of identifying the ion(s) responsi- as shown in Fig. 24 [57]. The data compare
ble for maintaining electroneutrality and the simultaneously determined gravimet-
the extent of solvent transfer. Although ric (f ) and optical (thickness) data during
transfer of protons, as the cations in the electrochemical deposition of polyani-
this system, is difficult to detect on a line from an aqueous solution of the
gravimetric basis, the near absence of a monomer. Since, in the acoustically thin
rapid mass response associated with rapid regime, f αM (via the Sauerbrey equa-
charge transfer is a signature – albeit near tion), the instantaneous slope of the plot
invisible – of proton transfer (for those provides the film density. The linearity
who read the Sherlock Holmes stories, of the plot shown clearly signals constant
it is like ‘‘the dog that did not bark in composition (polymer and solvent volume
the night’’). The situation is rather com- fractions) with film thickness and provides
plex, since there are two redox processes an alternative route to separating the areal
whose behaviors are dependent upon pH, density (M/g cm−2 into the volumetric
so one must be careful to restrict conclu- film density and thickness components, cf.
sions to the conditions under which they the method outlined in Sect. 2.7.2.2.2.
were deduced. In aqueous perchloric acid One of the other attractions of polyani-
medium, charge transfer in the first oxida- line is the ease of preparing (or even com-
tion process was found to be compensated mercial availability) of substituted deriva-
by anion injection and in the second oxida- tives. The simplest example is the methyl
tion process by cation (proton) ejection, in derivative poly(o-toluidine), for which the
each case accompanied by some solvent EQCM in conjunction with PBD has
transfer [152]. A recent ac electrogravi- been used to obtain the individual anion,
metric study [156] of polyaniline redox cation, and solvent fluxes/film popula-
switching in various aqueous acid media tions. By comparing the EQCM frequency
shows how far this technique has now ad- responses to a potential sweep [46] and a
vanced, by direct comparison of the rather potential step [154], it is possible to show
featureless mass–potential responses to that the competing ion transfers have
cyclic voltage scans with the complex plane intrinsic potential and timescale depen-
representation of the electrogravimetric- dencies; the contributions of the protons
transfer function, (m/E)(ω). In the and perchlorate ions are also pH depen-
latter case, one sees features (semicircles) dent. One interesting feature is that anion
4000
3200
[Hz] 2400
dapp = 1.45 g cm−3
−∆f
1600
800
0
0 100 200 300 400 500
Optical thickness
[nm]
Fig. 24 Variation of EQCM resonant frequency with simultaneously
ellipsometrically determined film thickness during deposition of
polyaniline. Electrode: Pt on 5-MHz AT-cut quartz crystal. Solution:
1 mol dm−3 aniline/2 mol dm−3 HCl. Galvanostatic control:
i = 77 µA cm−2 . Optical measurements at λ = 550 nm. (Reproduced
from Ref. [57] with permission from Elsevier.)
transfer is accompanied by solvent transfer with the larger cations, the contribution to
in the opposite direction, suggesting that electroneutrality maintenance (transport
in this case film rigidity imposes a volume number) decreased with ionic radius.
constraint. Another derivative of polyani- In most cases, stability and oxidation
line, this time with a fused aromatic ring, potential issues mean that polythiophenes
is poly(1,8-diaminonaphthalene) [153]. In are prepared, studied, and used in non-
this case, the EQCM was used to show aqueous media, and the influence of
how amine groups on the polymer can the solvent can be significant. The ex-
be used to extract Hg2+ ions from tent of solvent swelling is determined
solution; detailed considerations of the by polymer–solvent interactions, which
data suggest a 1 : 1 monomer : metal ion can be manipulated both by choice of
stoichiometry. A rather more exotic deriva- solvent and introduction of substituents,
tive is poly(aniline-co-N -propane sulfonic here in the 3- and 4-positions. Electropoly-
acid aniline), a so-called self-doped sys- merization of alkyl-substituted thiophenes
tem in which the parent aniline and provides good examples of the effects
sulfonic acid aniline comonomers carry one can observe. EQCM (with PBD) mea-
opposite charges [155]. EQCM and sup- surements of potentiodynamic deposition
porting crystal impedance measurements from acetonitrile : benzonitrile (4 : 1) of
were obtained for such films exposed to poly(didodecylterthiophene) (with the sub-
propylene carbonate and acetonitrile solu- stituents in the 3 ,4 - or 3,3 -positions)
tions, in which the electrolyte anion and showed that film deposition occurred pri-
cation sizes were varied. In all cases, a marily during the reductive half-cycle, in
mixed ion-transfer process was observed: the form of short chain oligomers of the
neutral polymer [45, 157]. These studies Fig. 25 [162], that complete redox conver-
were focused on the early stages of film sion of either layer does not occur before
deposition, that is, nucleation, so a gravi- commencement of redox conversion of
metric interpretation was appropriate, in the other layer. Rather, minor conver-
contrast to other studies discussed below. sion of the inner layer to its conducting
In terms of ion and solvent transfers, state occurs first (signaled by cation trans-
polythiophenes provide some of the more fer to balance PBT n-doping; see mass
complex examples of acoustic wave behav- flux at high scan rate), opening up con-
ior and thus more sophisticated examples ducting pathways to completely convert
of data interpretation. In the simpler, the outer layer (signaled by anion trans-
gravimetric regime early measurements fer to balance PXV redox conversion; see
of ion, solvent, and cation (‘‘salt’’) opposite mass flux at slow scan rate).
transfer at acoustically thin poly(2,2 - This ability of the EQCM to diagnose
bithiophene) [33] films have been shown ion-transfer mechanisms in bilayers has
to be typical; for example, the presence of been further demonstrated for [polypyr-
salt in reduced poly(1,4-ethylenedioxythio- role – poly(styrenesulfonate)]/poly(3-octyl-
phene) films is implied by a substantial thiophene) [164] and [polypyrrole–poly-
contribution from cation expulsion during (styrenesulfonate)]/poly(vinylferrocene)
[165] systems.
polymer oxidation [159].
As described in Sect. 2.7.2.1, there is
A more complicated scenario is pre-
no sharp distinction between ‘‘rigid’’ and
sented by an electroactive bilayer, in which
‘‘nonrigid’’ behavior, but rather a con-
a poly(2,2 -bithiophene) (PBT) film is over-
tinuous variation between films that are
laid with a PXV redox polymer film [162].
acoustically thin (hf < δ) and acoustically
Individually, the two layers (here, both
thick (hf ≥ δ). Since the decay length, δ, is
acoustically thin) behave exactly as one
a function of shear modulus and thereby of
would expect: the PBT film undergoes pre-
film solvent content, medium effects are
dominant anion (cation) exchange upon critical. The discussion of polythiophene
p-doping (n-doping), albeit with some systems above was deliberately restricted
coion participation, and the PXV film is a to examples of acoustically thin films, but
cation exchanger. The difference when one we now move to a consideration of acous-
moves to the bilayer is that the usual ion- tically thick polythiophene-based films.
transfer requirements of the inner PBT The transition between the acoustically
layer are not generally consistent with the thin and thick situations is illustrated
permselective properties of the outer PXV by the redox-switching (doping) behav-
layer. The result is a transient accumula- ior of poly(3-methylthiophene) films ex-
tion of electrolyte in the film upon redox posed to acetonitrile [158]. At low cover-
cycling. In this case, the EQCM is able ages, parameterized through deposition
to reveal the mechanism of bilayer recharge (Qdep ≤ 12 mC cm−2 ), the films
dox switching, by using gravimetrically are acoustically thin and the resonator
determined ion transfer – in conjunction frequency response can be interpreted in
with the electroneutrality condition – as gravimetric terms f converted to M,
the marker for locating the destination according to the Sauerbrey equation). In
of injected electronic charge. Variation this particular case, the film was found
of timescale (potential scan rate) shows, to contain two different regions (‘‘zones’’)
1.0 50 mV s−1 0.5

0.5
[µg cm−2 s−1]

[mA cm−2]
0.0 0.0
Current
dM dt
−0.5
−1.0 −0.5
−1.5
−1.0
−2.0
4 4
1 V s−1
2 2
[µg cm−2 s−1]

[mA cm−2]
0 0
Current
dM dt
−2 −2
−4 −4
−6
−6
−8
−8
−1.6−1.2−0.8−0.4 0.0 0.4 0.8 1.2 −1.6−1.2−0.8−0.4 0.0 0.4 0.8 1.2
Potential Potential
[V ] [V ]
Fig. 25 Current (left panels) and mass flux (right panels) responses for a
polybithiophene/polyxylylviologen (PBT/PXV) bilayer. Electrode: Au (area = 0.23 cm2 ) on
10-MHz AT-cut quartz crystal. Solution: 1 mol dm−3 tetraethylammonium
perchlorate/acetonitrile. Voltammetric experiment, scan rate 50 mV s−1 (upper panels),
1 V s−1 (lower panels). (Reproduced from Ref. [162] with permission from the Royal Society
of Chemistry.)
with different solvation characteristics and and were parameterized through equiva-
ion dynamics. For somewhat thicker films lent circuit parameters for the BvD model
(Qdep ca. 30–50 mC cm−2 ), viscoelastic of Fig. 2(b).
characteristics were visible in the crystal The absolute value of film thickness at
admittance spectra, but not dramatic. In which one sees the shift from acousti-
this regime, a gravimetric interpretation cally thin to thick behavior will depend
is inappropriate, but modeling in terms significantly on the polymer itself, the
of viscoelastic parameters is inaccurate, solvent, and other physicochemical pa-
since the acoustic deformation is not large. rameters, notably temperature, applied
Much thicker films (Qdep > 80 mC cm−2 ) potential, and timescale. An example of
showed dramatic viscoelastic effects, from the importance of the latter two con-
which shear moduli could be extracted. trol parameters is provided by a study
A similar pattern can be expected for of poly(3-hexylthiophene) films exposed
other systems, such as the more elaborate to propylene carbonate [161]. Storage and
trimeric fused ring dithioenothiophene- loss moduli, derived from the admittance
based polymers [160]. Admittance spectra spectra, for a film held at different ap-
for films of these materials deposited plied potentials (effectively, controlling
with 5 < Qdep /mC cm−2 < 120 showed charge) are shown in Fig. 26 [161]. Im-
the dramatic effects of film viscoelasticity mediate observations are that the film is
Fig. 26 Viscoelastic properties of

8 electropolymerized
[108 dynes cm−2]
poly(3-hexylthiophene). Fitted
6 (a) storage modulus (G ) and (b) loss
modulus (G ) as functions of charge
G'
4 and frequency. Frequency dependence

obtained by use of harmonics of
2 polished 10 MHz fundamental
Au-coated TSM resonator; open (filled)
0 symbols for increasing (decreasing)
0 2 4 6 8 potential measurement sequence (lines
Q are merely a guide to the eye).
[mC cm−2] Frequency (MHz): 10 (, ), 30 (, ),
50 (, ), 70 (♦, ). Electrode: Au
14 (area = 0.23 cm2 ) on polished 10-MHz
12 AT-cut quartz crystal. Solution:
[108 dynes cm−2]
0.1 mol dm−3 tetraethylammonium

10
hexafluorophosphate/propylene
8 carbonate. (Reproduced from Ref. [161]
G"
6 with permission from the Royal Society

4 of Chemistry.)
2
0
0 2 4 6 8
Q
[mC cm−2]
in the middle of the viscoelastic range by making admittance measurements at

(G ∼ 108 dyn cm−2 ) and is lossy (G < the fundamental frequency (10 MHz) and
G ). Although these were anticipated to higher harmonics (30, 50, and 70 MHz).
represent equilibrium measurements, by Over this frequency (timescale) range, the
holding the applied potential at a series of shear modulus components vary by ap-
constant values for times on the order of proximately an order of magnitude. In
minutes, there is clearly hysteresis. This polymer science, it is common to use
is not a matter of slow charge (coupled ‘‘spring and dashpot’’ mechanical models
electron/ion) transport, since the data are to explain the frequency dependence of
plotted as functions of resultant electro- viscoelastic parameters. The exchange be-
chemical charge (obtained by integrating tween the polymer and solution of ions and
the current response) not of applied poten- solvent might be expected to be associated,
tial. The conclusion reached was that this respectively, with electrostatic stiffening
is a consequence of slow solvent transfer. and plasticization effects. In the case of
Thus, mobile species dynamics – which poly(3-hexylthiophene) films [41, 161], the
one commonly monitors directly through increase in G with frequency is qualita-
gravimetric experiments – are here ob- tively consistent with a Maxwell model
served indirectly through polymer dynam- (spring and dashpot in series) and in-
ics in a viscoelastic experiment. consistent with a Voigt model (spring
Figure 26 also shows shear modulus and dashpot in parallel). Although nei-
data as functions of frequency, obtained ther model can quantitatively describe the
observed responses, this avenue of en- increases with increasing loading of the
quiry is at an early stage. Certainly, TSM TSM resonator; once resonance is passed,
resonators can provide viscoelastic param- the anticipated pattern of response is
eters in a frequency range inaccessible to resumed. In enquiring why more obser-
bulk mechanical techniques, which typi- vations of film resonance have not been
cally operate in the sub-kHz range. reported, three corollaries are highlighted.
Although temperature has not been First, an additional requirement is that
systematically explored as a variable in this the reflectivity coefficient for the acoustic
context, this will unquestionably prove a wave at the film/solution interface must be
fruitful future avenue of enquiry and the close to unity. This corresponds to a sharp
associated effects might be rationalized in gradient of acoustic properties, that is, a
terms of the concept of time–temperature sharp polymer/solution interface, a condi-
superposition and the WLF model that are tion that may commonly not be fulfilled.
commonly used in polymer science. Second, if the polymer is lossy, attenuation
In Sect. 2.7.2.1.2, the phenomenon of of the acoustic wave in passage twice
film resonance was discussed. In this across the required film thickness may be
special situation, the film thickness cor- too great for a significant effect. Third,
responds to one quarter of the acoustic there is an interesting effect in terms
wavelength, that is, the acoustic phase shift of the operating frequency. In the case
defined by Eq. (11) has the numerical value of poly(3-hexylthiophene), the film thick-
φ = π/2. For a film of given shear modu- ness required to achieve film resonance
lus, progressive increase in thickness will was essentially independent of frequency
eventually result in this condition being (comparing data for the fundamental and
satisfied. This phenomenon is illustrated third harmonic, 10 and 30 MHz, re-
in Fig. 27 [41] for a poly(3-hexylthiophene) spectively). Inspection of Eq. (11) shows
film as a function of the polymerization that this requires a frequency-dependent
charge during deposition. As can be seen, shear modulus (in the simplest situation,
the resonant frequency transiently moves increasing with the square of frequency),
sharply upwards and the peak amplitude consistent with a viscoelastic polymer in
2.5
Fig. 27 Crystal admittance spectra as a
2.0
function of charge density, acquired
Conductance
dynamically during potentiodynamic 1.5

electropolymerization of a
[mS]
poly(3-hexylthiophene) film. Solution: 1.0

3.7 mmol dm−3
3-hexylthiophene/0.1 mol dm−3 0.5
tetraethylammonium
hexafluorophosphate/propylene 0.0
carbonate. Electrode: Au
0
(area = 0.23 cm2 ) on polished 10-MHz 5
AT-cut quartz crystal. Spectra acquired 10
Charge 15
at E = 0.4 V following each potential [mC cm−2]
20
10.0 10.1
9.8 9.9
cycle. (Reproduced from Ref. [41] with
permission from the American Frequency
Chemical Society.) [MHz]
the transition region between rubbery and per se – at best, one can only determine
glassy regimes. film performance electrochemically using
Several other polymers with delocalized solution redox probes to seek out its imper-
systems have been studied. One example fections. Consequently, the determination
is poly(p-phenylenevinylene), for which of polymer coverage using a gravimetric
the EQCM was used to study ion mo- probe is attractive. The EQCM has been
tion during n-doping [166]. The case of used to study a number of acrylate-based
an electropolymerized nickel-salen-based insulating films, such as acrylonitrile [167],
complex, poly[Ni(saltMe)], studied using methylmethacrylate [167], methacryloni-
the combined EQCM/PBD technique, is trile [168], and N -vinyl-2-pyrrolidone [168].
rather interesting in terms of the chemistry The EQCM was used to explore the effi-
involved [47]. Although one might expect ciency of deposition during the polymer-
to classify the material as a ‘‘metal com- ization process [168] and, postdeposition,
plex’’ (and indeed the monomer behaves in issues of film solubility [167].
this fashion), spectroscopic studies show
the polymer redox chemistry to be ligand-, 2.7.3.8 Other Organic Materials
rather than metal-based, and to involve a The redox chemistry of solid-state lat-
delocalized system, that is, it should be tices based on TCNQ, prepared by oxi-
classed as a ‘‘conducting polymer’’. Crys- dizing crystallites of 9-aminoacridinium
tal impedance spectra showed the polymer (TCNQ)2 exposed to calcium or potas-
films to behave rigidly, so a gravimet- sium acetate solutions, has been studied
ric interpretation of EQCM data could be using the EQCM [169]. EQCM frequency
used; from this it was possible to demon- responses to redox conversion of the
strate dominance of anion (cf. cation) TCNQ0/− and TCNQ−/2− states were in-
transfer in maintaining electroneutrality terpreted in terms of permselective cation
during redox switching and to quantify transfer of CaOAc+ and K(H2 O)4 + . The
the associated transfer of solvent. Film first reduction half-cycle was associated
resonance has been observed in crystal with an anomalously large frequency
admittance spectra of poly(carbazole) [42]. change, suggesting a one-off solvation
In this case, the film thickness satisfying process or structural change. Chronoam-
the resonance condition was inversely perometric experiments (associated with
related to frequency (comparing data for shorter timescales) suggested that oxida-
the fundamental and the third harmonic). tion proceeds by anion insertion from the
According to Eq. (11), this is consistent electrolyte, then diffusional loss of neutral
with a frequency-independent shear mod- electrolyte from the solid-state lattice. Al-
ulus, contrasting with the behavior of poly- though the physical nature of the film is
(3-hexylthiophene). rather different from a redox polymer or a
metal oxide film, the parallels in behavior
2.7.3.7.3 Insulating Polymers Electro- are clear.
chemical polymerization has been used There has been considerable interest in
to generate insulating polymeric coatings the coupling to electrode surfaces of DNA
on electrode surfaces, usually for the pur- which, as a large electroinactive molecule,
poses of surface protection. In these cases, is amenable to gravimetric detection by
there can be no use of electrochemistry the QCM. Following immobilization onto
to characterize the deposited material Au electrodes, electrochemically triggered
release of DNA has been monitored us- Since the QCM sensitivity is spatially vari-
ing the EQCM [170]; the longer-term goal ant – maximum at the center and falling to
is on-demand gene delivery. Oligonu- effectively zero at the edges of the exciting
cleotides have also been immobilized electrode – this means that there is no er-
within polypyrrole films on electrode sur- ror incurred by using the usual spatially av-
faces, for example, by incorporating an eraged sensitivity function of frequency to
electropolymerizable pyrrole functionality mass change (the Sauerbrey equation). An-
onto the oligonucleotide [171–173]. (Al- other exception to this lateral uniformity
though one could categorize these as is the case of heterogeneous bubble evolu-
‘‘polypyrrole’’ systems, the entity of inter- tion. Simplistically, the concept is that nu-
est is the oligonucleotide, and the primary cleation of a bubble at the interface results
role of the polypyrrole is the provision of in the replacement of fluid with gas, so
an immobilization matrix.) These studies that the coupling of the ambient medium
are at a relatively early stage, but offer to the resonator is locally changed. Since
the promise of sophisticated biosensing the characteristic size of a bubble (typically
devices. in excess of 1 µm) is likely to be much
larger than the decay length of the acoustic
2.7.3.9 Miscellaneous wave in the fluid (typically 0.2 µm, for a 10-
The phenomenon of oscillating reactions MHz resonator in aqueous solution), there
is widely recognized in chemistry. In those is effectively complete (local) decoupling of
cases in which a surface-bound species or the resonator and fluid. This situation has
the surface itself is involved in the com- been studied for electrolytically generated
plex sequence of steps, the EQCM offers hydrogen [177–179] and chlorine [179]. In
the prospect of additional information. An the latter case, the situation is complicated
example of this is the role of an oxide layer by the formation/removal of oxide layers
in the oxidation of formaldehyde at Pt and and electrode dissolution, to both of which
Rh [174]. Similarly, in the oxidation of 2- the EQCM responds. In addition to the
propanol at Pt electrodes, it was found that lateral sensitivity issue, that is, the fact
the oscillations (in potential and mass, at that the frequency response is dependent
constant current) increased in amplitude upon the spatial location of the bubble,
until the positive extreme of the potential the QCM response is dependent upon in-
excursion reached a value consistent with terfacial energetics, which determine the
PtOH and/or PtO formation [175]. Oscilla- contact angle and thus the bubble shape.
tions in mass at open-circuit potential were Although this case has not subsequently
also observed during the dissolution of Cu been explored in great detail, spectral anal-
in sulfate media when the solution con-

Encyclopedia of Electrochemistry 10 Volume Set

Uploaded by

Copyright:

Available Formats

Encyclopedia of Electrochemistry 10 Volume Set

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Encyclopedia of Electrochemistry 10 Volume Set

Uploaded by

Copyright:

Available Formats

1

Oleg A. Petrii, Galina A. Tsirlina

1.1 Introductory Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

1.1 electrochemistry. Electrodes can be based

applied to the charged species does not 1.2

When the free electrostatic charge in

Fig. 1 A scheme of potential drops at

(Fig. 2), we obtain

ing only in their electric states (all the (10)

2 Fig. 2 Illustration of two

Tab. 1 Conventional reference electrodes [36]

Reference Potential versus Analogs Media

Calomel electrodes Mercurous Aqueous and

Note: NCE: Normal calomel electrodes; SCE: saturated calomel electrode.

A special comment should be given of 1.7

Special reference electrodes are intro- It should be mentioned that no standard

1.13 and the problem reduces to the determi-

EK = (EK )H+ ,H2 + E ≈ −4.44 + E The value of the Galvani potential M S ϕ

E 0(Li+/Li) −3.05 −1.39

E 0(Al3+/Al) −1.66 −2.78

Fig. 4 Comparison of SHE and ‘‘absolute’’ scales, as reported in Ref. [9].

of mutual contact: at their common interface, respectively.

0 0.2 0.4 0.6

scale, in which the potential of electrode 2. A. Frumkin, A. Gorodetzkaja, Z. Phys. Chem.

2.1 The interfaces dielectric–electrolyte are

nonequilibrium potential difference bet- being primarily dependent on the solvent

Traditionally, special attention in This equation enables one to completely

in which these complications are assumed anion-reversible electrodes in the case

m/s dσ dµ the others if the charge vs. potential de-

electronic charge density at the electrode, In electrochemical kinetics, this model

Gouy–Chapman (GC), that is, to consider whole interface, K ≡ σ/(E − Eσ =0 ), to

15 10 5 −5 −10 −15 −20

15 10 5 0 −5 −10 −15 −20

15 10 5 0 −5 −10 −15 −20

uncertainty in the CH value is enormous. It charge, since it is too sensitive even to

0 0.05 0.10 0.02 0.04 0.06

Fig. 4 Parsons–Zobel plots, C−1

veriﬁed incompletely. Besides, it does not Analysis of this ﬁgure leads to a

Thus-found compact-layer capacitance solvent molecules (see Sects. 2.1.7–2.1.9),

charge requires a microscopic model of attains signiﬁcant values, being depen-

2.1.6.1 ‘‘Uniform’’ Model of Solid

methods of analysis to establish the PC the linearized Poisson–Boltzmann ap-

For Gaussian height–height correla- roughness approximation, and an exact

by a speciﬁc statistical-mechanical scheme the nonlocal dielectric function. Referring

maximum centered at m . 2.1.8.2 Multistate Models

function of its center of mass,

− 0.004 − 0.002 0.000 0.002 0.004

εεb ε=1 εε ((σ)

In contrast to what was suggested in approximate the response function, spec-

Fig. 17 The geometry of electromagnetic wave reﬂection

The modulation technique is based where C is the capacitance of the

−1.5 −1.4 −1.3 −1.2 −1.1 −1.0 − 0.9 − 0.8 − 0.7

Fig. 18 The evaluated interfacial relaxation, extracted from the data on

p(cos ϑµ ) ρ(z) p(cos ϑµ )

−1 − 0.5 0 0.5 1 −1 − 0.6 − 0.2 0.2

◦ (zi e)2 LH ≡ (4πCH )−1 (80)

if the ion is located closer to the electrode determination of experimental values of 

from mixed electrolyte solutions in the Grahame–Parsons isotherm to interpret

Wads /kBT Wadso + Wimc/kBT

0.1 − 0.1 − 0.3

EK = (EK )H+ ,H2 + E ≈ −4.44 + E The value of the Galvani potential M S ϕ

maximum centered at m . 2.1.8.2 Multistate Models

if the ion is located closer to the electrode determination of experimental values of

G(h) is the electrostatic contribution to ﬁrst approach uses the Gouy–Chapman