Batteries

Volume 104, Number 10
Introduction: Batteries and Fuel Cells

This special issue of Chemical Reviews covers the cal and materials research has been focused during
electrochemical storage and generation of energy in the past three decades. The second paper, by Whit-
batteries and fuel cells. This area is gaining tremen- tingham, begins with a general historical background
dous importance for powering high technology devices to lithium batteries and then focuses on the next
and for enabling a greener and less energy-intensive generation of cathodes. The third, by Xu, gives an
transportation industry. Whether the demand is from in-depth review of the presently used and future
a cell phone, a computer, or an iPOD, consumers are electrolytes; this is followed by an extensive review
demanding a longer life in a smaller package and at by Arora and Zhang of the separators used in lithium
a lower cost with minimal if any wired connection. and related batteries. The following paper, by Long,
The consumer generally does not care whether the Dunn, Rolison, and White, addresses new three-
power source is a battery, a fuel cell, or something dimensional concepts for increasing the storage ca-
else, as long as it works. In the area of greener pacity. Critical to the development of new materials
transportation, there has been a surge of interest in are advanced characterization and modeling tech-
vehicles that are electrically powered, either totally, niques, and some of these are described by Grey and
as planned for the green Beijing Olympic Games, or Dupre and by Reed and Ceder in the last two papers
partially, as in hybrid electric vehicles. The present of the battery group. Several papers covering anodes,
generation of such vehicles uses a combination of an phosphate and nickel oxide cathodes, and nickel
internal combustion engine and a battery, today metal hydride batteries did not meet the publication
nickel metal hydride, as in the Toyota Prius, and deadline, and it is hoped that they will appear in
tomorrow lithium; a future generation is likely to be future issues.
a hybrid of a fuel cell and a battery.
Both batteries and fuel cells utilize controlled Fuel Cells
chemical reactions in which the desired process Although fuel cells were invented in the middle of
occurs electrochemically and all other reactions in- the 19th century, they didn’t find the first application
cluding corrosion are hopefully absent or severely until space exploration in the 1960s. Since then, the
kinetically suppressed. This desired selectivity de- development of fuel cell technology has gone through
mands careful selection of the chemical components several cycles of intense activity, each followed by a
including their morphology and structure. Nanosize period of reduced interest. However, during the past
is not necessarily good, and in present commercial two decades, a confluence of driving forces has
lithium batteries, particle sizes are intentionally created a sustained and significant world-wide effort
large. All batteries and fuel cells contain an electro- to develop fuel cell materials and fuel cell systems.
positive electrode (the anode or fuel) and an electro- These driving needs include the demand for efficient
negative electrode (the cathode or oxidant) between energy systems for transportation, the desire to
which resides the electrolyte. To ensure that the reduce CO2 emissions and other negative environ-
anode and cathode do not contact each other and mental impacts, and the demand for high energy
short out the cell, a separator is placed between the density power sources for portable electronic applica-
two electrodes. Most of these critical components are tions. Due to the high level of interest in fuel cells
discussed in this thematic issue. during the last decade or so, there have been numer-
The issue starts with a general introduction by ous summary articles and symposia focused on the
Brodd and Winter to batteries and fuel cells and the technology state of the art. In this thematic issue,
associated electrochemistry. It then continues first we present a series of summary articles that deal
with several papers discussing batteries and then with some of the fundamental scientific issues related
with papers discussing fuel cells. to fuel cell development.
Batteries A fuel cell that has desirable features for trans-
portation and portable power is the polymer electro-
Outside of the above introduction, the battery lyte membrane (PEM) system. The core of this
papers describe lithium batteries, where most chemi- technology is a polymer membrane that conducts
10.1021/cr020705e CCC: $48.50 © 2004 American Chemical Society
Published on Web 10/13/2004
4244 Chemical Reviews, 2004, Vol. 104, No. 10 Editorial
protons but separates the fuel from the oxidant. The of high energy density fuel must be considered. To
material used historically and most frequently in this end, Holladay, Wang, and Jones present a review
PEM fuel systems is Nafion, a perfluorocarbon-based of the developments of using microreactor technology
polymer carrying sulfonic acid residues. Nafion is a to convert liquid fuels into hydrogen for directly
commercial material and has received the most feeding into a PEM fuel cell.
extensive study of any PEM fuel cell membranes. Another fuel cell system undergoing intense re-
Mauritz and Moore prepared a summary of the search is the solid oxide type. Adler presents the
current understanding of the large volume of re- factors that govern the rate limiting oxygen reduction
search that has gone into optimizing and understand- reaction within the solid oxide fuel cell cathodes.
ing this membrane system. Other polymer systems McIntosh and Gorte, on the other hand, treat the
that would have even better performance than Nafion anode in the solid oxide fuel cell by examining
and/or have lower costs are being sought by research- catalytic direct hydrocarbon oxidation. Finally,
ers around the world. Hickner, Ghassemi, Kim, Calabrese Barton, Gallaway, and Atanossov take a
Einsla, and McGrath summarize work on such al- look at the future. In their article, they present a
ternative polymer systems for proton exchange mem- summary of some of the enzymatic biological fuel
branes. These types of materials have complex trans- cells that are being developed as implantable devices
port properties that involve not just proton movement and also to power microscale devices.
but also the movement of water. Theoretical treat- We hope this collection of papers will provide new
ments of the transport mechanisms and processes in researchers in the field with a starting point for
these proton conductors are given by Kreuer, Pad- advancing research. Furthermore, our hope is to
dison, Spohr, and Schuster and by Weber and New- stimulate the next generation of breakthroughs that
man. will lead to the success of fuel cell development.
In PEM fuel cells, catalyst activity and catalyst
efficiency are still significant issues. Russell and Rose
M. Stanley Whittingham
summarize fundamental work involving X-ray ab- Chemistry and Materials,
sorption spectroscopy on catalysts in low temperature State University of New York at Binghamton
fuel cell systems. These types of studies are very
useful for developing a detailed understanding of the
mechanisms of reactions at catalyst surfaces and Robert F. Savinell
could lead to the development of new improved Chemical Engineering,
Case Western Reserve University
efficient catalysts. Important in the development of
fuel cell technology are mathematical models of
engineering aspects of a fuel cell system. Wang writes Thomas Zawodzinski
about studies related to this topic. Chemical Engineering,
Finally, in order for PEM fuel cell systems to be Case Western Reserve University
affordable for portable power applications, a source CR020705E
Chem. Rev. 2004, 104, 4245−4269 4245
What Are Batteries, Fuel Cells, and Supercapacitors?
Contents
1. Introduction 4245
1.1. Batteries versus Fuel Cells versus 4245
Electrochemical Capacitors
1.2. Definitions 4247
1.3. Thermodynamics 4248
1.4. Kinetics 4249
1.5. Experimental Techniques 4250
1.6. Current Distribution and Porous Electrodes 4251
2. Batteries 4252
2.1. Introduction and Market Aspects 4252
2.2. Battery Operations 4253
2.3. Characteristics of Common Battery Systems 4254
2.4. Primary Batteries 4254 Dr. Martin Winter is currently University Professor for Applied Inorganic
2.5. Rechargeable Batteries 4257 Chemistry and Electrochemistry at the Institute for Chemistry and
Technology of Inorganic Materials, Graz University of Technology (Austria).
2.6. Selection Criteria for Commercial Battery 4258 His fields of specialization are applied electrochemistry, chemical technol-
Systems ogy and solid state electrochemistry with special emphasis on the
3. Fuel Cells 4259 development and characterization of novel materials for rechargeable
3.1. Introduction and Market Aspects 4259 lithium batteries.
3.2. Fuel Cell Operation 4261
3.3. Characteristics of Various Types of Fuel 4264
Cells
4. Electrochemical Capacitors (ECs) 4266
4.1. Introduction and Market Aspects 4266
4.2. Characteristics of the Electrical Double Layer 4267
4.3. EC Operation 4267
5. Summary 4269
1. Introduction
1.1. Batteries versus Fuel Cells versus
Electrochemical Capacitors
Energy consumption/production that rely on the
combustion of fossil fuels is forecast to have a severe Dr. Ralph J. Brodd is President of Broddarp of Nevada. He has over 40
years of experience in the technology and market aspects of the
future impact on world economics and ecology. Elec- electrochemical energy conversion business. His experience includes all
trochemical energy production is under serious con- major battery systems, fuel cells, and electrochemical capacitors. He is a
sideration as an alternative energy/power source, as Past President of the Electrochemical Society and was elected Honorary
long as this energy consumption is designed to be Member in 1987. He served as Vice President and National Secretary of
more sustainable and more environmentally friendly. the International Society of Electrochemistry as well as on technical
Systems for electrochemical energy storage and advisory committees for the National Research Council, the International
Electrotechnic Commission, and NEMA and on program review committees
conversion include batteries, fuel cells, and electro- for the Department of Energy and NASA.
chemical capacitors (ECs). Although the energy stor-
age and conversion mechanisms are different, there In batteries and fuel cells, electrical energy is
are “electrochemical similarities” of these three sys- generated by conversion of chemical energy via redox
tems. Common features are that the energy-providing reactions at the anode and cathode. As reactions at
processes take place at the phase boundary of the the anode usually take place at lower electrode
electrode/electrolyte interface and that electron and potentials than at the cathode, the terms negative
ion transport are separated. Figures 1 and 2 show and positive electrode (indicated as minus and plus
the basic operation mechanisms of the three systems. poles) are used. The more negative electrode is
Note that batteries, fuel cells, and supercapacitors designated the anode, whereas the cathode is the
all consist of two electrodes in contact with an more positive one. The difference between batteries
electrolyte solution. The requirements on electron and fuel cells is related to the locations of energy
and ion conduction in electrodes and the electrolyte storage and conversion. Batteries are closed systems,
are given in Figure 1 and are valid for all three with the anode and cathode being the charge-transfer
systems. medium and taking an active role in the redox
10.1021/cr020730k CCC: $48.50 © 2004 American Chemical Society
a tank, for example, fuels such as hydrogen and

hydrocarbons. Energy storage (in the tank) and
energy conversion (in the fuel cell) are thus locally
separated.1
In electrochemical capacitors (or supercapacitors),
energy may not be delivered via redox reactions and,
thus the use of the terms anode and cathode may not
be appropriate but are in common usage. By orienta-
tion of electrolyte ions at the electrolyte/electrolyte
interface, so-called electrical double layers (EDLs) are
formed and released, which results in a parallel
movement of electrons in the external wire, that is,
in the energy-delivering process.
In comparison to supercapacitors and fuel cells,
batteries have found by far the most application
markets and have an established market position.
Whereas supercapacitors have found niche markets
as memory protection in several electronic devices,
Figure 1. Representation of a battery (Daniell cell)
showing the key features of battery operation and the fuel cells are basically still in the development stage
requirements on electron and ion conduction. and are searching to find a “killer application” that
allows their penetration into the market. Fuel cells
established their usefulness in space applications
with the advent of the Gemini and Apollo space
programs. The most promising future markets for
fuel cells and supercapacitors are in the same ap-
plication sector as batteries. In other words, super-
capacitor and fuel cell development aim to compete
with, or even to replace, batteries in several applica-
tion areas. Thus, fuel cells, which originally were
intended to replace combustion engines and combus-
tion power sources due to possible higher energy
conversion efficiencies and lower environmental im-
pacts, are now under development to replace batter-
ies to power cellular telephones and notebook com-
puters and for stationary energy storage. The moti-
vation for fuel cells to enter the battery market is
simple. Fuel cells cannot compete today with com-
bustion engines and gas/steam turbines because of
much higher costs, inferior power and energy per-
formance, and insufficient durability and lifetime.
With operation times of typically <3000 h and, at
least to an order of magnitude, similar costs, batteries
are less strong competitors for fuel cells.
The terms “specific energy” [expressed in watt-
hours per kilogram (Wh/kg)] and “energy density” [in
watt-hours per liter (Wh/L)] are used to compare the
energy contents of a system, whereas the rate capa-
bility is expressed as “specific power” (in W/kg) and
“power density” (in W/L). Alternatively, the attributes
“gravimetric” (per kilogram) and “volumetric” (per
liter) are used. To compare the power and energy
Figure 2. Representation of (A, top) an electrochemical capabilities, a representation known as the Ragone
capacitor (supercapacitor), illustrating the energy storage plot or diagram has been developed. A simplified
in the electric double layers at the electrode-electrolyte Ragone plot (Figure 3) discloses that fuel cells can
interfaces, and (B, bottom) a fuel cell showing the continu- be considered to be high-energy systems, whereas
ous supply of reactants (hydrogen at the anode and oxygen supercapacitors are considered to be high-power
at the cathode) and redox reactions in the cell.
systems. Batteries have intermediate power and
energy characteristics. There is some overlap in
reaction as “active masses”. In other words, energy
energy and power of supercapacitors, or fuel cells,
storage and conversion occur in the same compart-
with batteries. Indeed, batteries with thin film
ment. Fuel cells are open systems where the anode
and cathode are just charge-transfer media and the
1 Strictly speaking, a single electrochemical power system is denoted
active masses undergoing the redox reaction are
a cell or element, whereas a series or parallel connection of cells is
delivered from outside the cell, either from the named a battery. The literature is confusing, as the terms fuel CELL
environment, for example, oxygen from air, or from and BATTERY are used independent of the number of cells described.
Editorial Chemical Reviews, 2004, Vol. 104, No. 10 4247
Figure 5. Comparison of the driving ranges for a vehicle

powered by various battery systems or a gasoline-powered
combustion engine.
Figure 3. Simplified Ragone plot of the energy storage cells, and not batteries, are considered for replace-
domains for the various electrochemical energy conversion ment of combustion engines. The theoretical values
systems compared to an internal combustion engine and in Figure 4 are an indication for the maximum energy
turbines and conventional capacitors. content of certain chemistries. However, the practical
values differ and are significantly lower than the
theoretical values. As a rule of thumb, the practical
energy content of a rechargeable battery is 25% of
its theoretical value, whereas a primary battery
system can yield >50% of its theoretical value in
delivered energy. In the future, fuel cells might be
able to convert the used fuels into electrical energy
with efficiencies of >70%. The difference between the
theoretical and practical energy storage capabilities
is related to several factors, including (1) inert parts
of the system such as conductive diluents, current
collectors, containers, etc., that are necessary for its
operation, (2) internal resistances within the elec-
trodes and electrolyte and between other cell/battery
components, resulting in internal losses, and (3)
limited utilization of the active masses, as, for
example, parts of the fuel in a fuel cell leave the cell
Figure 4. Theoretical specific energies [(kW h)/tonne] and without reaction or as, for example, passivation of
energy densities [(kW h)/m3] of various rechargeable bat- electrodes makes them (partially) electrochemically
tery systems compared to fuels, such as gasoline, natural inactive. However, as batteries and fuel cells are not
gas, and hydrogen. subject to the Carnot cycle limitations, they may
operate with much higher efficiencies than combus-
electrodes exhibit power characteristics similar to tion engines and related devices.
those of supercapacitors. Moreover, there are also
hybrids such as metal/air batteries (or, in other
words, metal/air fuel cells), which contain a battery
1.2. Definitions
electrode (metal anode) and a fuel cell electrode (air The following definitions are used during the
cathode). Finally, Figure 3 also shows that no single course of discussions on batteries, fuel cells, and
electrochemical power source can match the charac- electrochemical capacitors.
teristics of the internal combustion engine. High A battery is one or more electrically connected
power and high energy (and thus a competitive electrochemical cells having terminals/contacts to
behavior in comparison to combustion engines and supply electrical energy.
turbines) can best be achieved when the available A primary battery is a cell, or group of cells, for
electrochemical power systems are combined. In such the generation of electrical energy intended to be
hybrid electrochemical power schemes, batteries and/ used until exhausted and then discarded. Primary
or supercapacitors would provide high power and the batteries are assembled in the charged state; dis-
fuel cells would deliver high energy. charge is the primary process during operation.
Figure 4 shows the theoretical specific energies A secondary battery is a cell or group of cells for
[(kW h)/t] and energy densities [(kW h)/m3)] of the generation of electrical energy in which the cell,
various rechargeable battery systems in comparison after being discharged, may be restored to its original
to fuels, such as gasoline, natural gas, and hydrogen. charged condition by an electric current flowing in
The inferiority of batteries is evident. Figure 5, the direction opposite to the flow of current when the
showing driving ranges of battery-powered cars in cell was discharged. Other terms for this type of
comparison to a cars powered by a modern combus- battery are rechargeable battery or accumulator. As
tion engine, gives an impressive example of why fuel secondary batteries are ususally assembled in the
discharged state, they have to be charged first before 1.3. Thermodynamics

they can undergo discharge in a secondary process.
The energy storage and power characteristics of
A specialty battery is a primary battery that is in electrochemical energy conversion systems follow
limited production for a specific end-use. In this paper directly from the thermodynamic and kinetic formu-
specialty batteries will not be particularly addressed. lations for chemical reactions as adapted to electro-
The anode is the negative electrode of a cell chemical reactions. First, the basic thermodynamic
associated with oxidative chemical reactions that considerations are treated. The basic thermodynamic
release electrons into the external circuit. equations for a reversible electrochemical transfor-
The cathode is the positive electrode of a cell mation are given as
associated with reductive chemical reactions that
gain electrons from the external circuit. ∆G ) ∆H - T∆S (1)
Active mass is the material that generates electrical
current by means of a chemical reaction within the and
battery.
An electrolyte is a material that provides pure ionic ∆G° ) ∆H° - T∆S° (2)
conductivity between the positive and negative elec-
trodes of a cell. where ∆G is the Gibbs free energy, or the energy of
a reaction available () free) for useful work, ∆H is
A separator is a physical barrier between the the enthalpy, or the energy released by the reaction,
positive and negative electrodes incorporated into ∆S is the entropt, and T is the absolute temperature,
most cell designs to prevent electrical shorting. The with T∆S being the heat associated with the or-
separator can be a gelled electrolyte or a microporous ganization/disorganization of materials. The terms
plastic film or other porous inert material filled with ∆G, ∆H, and ∆S are state functions and depend only
electrolyte. Separators must be permeable to the ions on the identity of the materials and the initial and
and inert in the battery environment. final states of the reaction. The degree symbol is used
A fuel cell is an electrochemical conversion device to indicate that the value of the function is for the
that has a continuous supply of fuel such as hydro- material in the standard state at 25 °C and unit
gen, natural gas, or methanol and an oxidant such activity.
as oxygen, air, or hydrogen peroxide. It can have Because ∆G represents the net useful energy
auxiliary parts to feed the device with reactants as available from a given reaction, in electrical terms,
well as a battery to supply energy for start-up. the net available electrical energy from a reaction in
An electrochemical capacitor is a device that stores a cell is given by
electrical energy in the electrical double layer that
forms at the interface between an electrolytic solution ∆G ) -nFE (3)
and an electronic conductor. The term applies to
charged carbon-carbon systems as well as carbon- and
battery electrode and conducting polymer electrode
combinations sometimes called ultracapacitors, super- ∆G° ) -nFE° (4)
capacitors, or hybrid capacitors.
Open-circuit voltage is the voltage across the where n is the number of electrons transferred per
terminals of a cell or battery when no external mole of reactants, F is the Faraday constant, being
current flows. It is usually close to the thermody- equal to the charge of 1 equiv of electrons, and E is
namic voltage for the system. the voltage of the cell with the specific chemical
Closed-circuit voltage is the voltage of a cell or reaction; in other words, E is the electromotive force
battery when the battery is producing current into (emf) of the cell reaction. The voltage of the cell is
the external circuit. unique for each reaction couple. The amount of
electricity produced, nF, is determined by the total
Discharge is an operation in which a battery amount of materials available for reaction and can
delivers electrical energy to an external load. be thought of as a capacity factor; the cell voltage
Charge is an operation in which the battery is can be considered to be an intensity factor. The usual
restored to its original charged condition by reversal thermodynamic calculations on the effect of temper-
of the current flow. ature, pressure, etc., apply directly to electrochemical
Internal resistance or impedance is the resistance reactions. Spontaneous processes have a negative free
or impedance that a battery or cell offers to current energy and a positive emf with the reaction written
flow. in a reversible fashion, which goes in the forward
The Faraday constant, F, is the amount of charge direction. The van’t Hoff isotherm identifies the free
that transfers when one equivalent weight of active energy relationship for bulk chemical reactions as
mass reacts, 96 485.3 C/g-equiv, 26.8015 Ah/g-equiv.
Thermal runaway is an event that occurs when the ∆G ) ∆G° + RT ln(AP/AR) (5)
battery electrode’s reaction with the electrolyte be-
comes self-sustaining and the reactions enter an where R is the gas constant, T the absolute temper-
autocatalytic mode. This situation is responsible for ature, AP the activity product of the products and AR
many safety incidents and fires associated with the activity product of the reactants. Combining eqs
battery operations. 4 and 5 with the van’t Hoff isotherm, we have the
Nernst equation for electrochemical reactions: problem for low-rate applications; however, high-rate
batteries must make provisions for heat dissipation.
E ) E° + (RT/nF) ln(AP/AR) (6) Failure to accommodate/dissipate heat properly can
lead to thermal runaway and other catastrophic
Faraday’s laws, as summarized in eq 7, give the situations.
direct relationship between the amount of reaction
and the current flow. There are no known exceptions 1.4. Kinetics
to Faraday’s laws.
Thermodynamics describe reactions at equilibrium
It(MW) and the maximum energy release for a given reaction.
g) (7) Compared to the equilibrium voltage () open ciruit
nF
voltage, EOCV), the voltage drops off () “electrode
g is the grams of material transformed, I is the polarization” or “overvoltage”) when current is drawn
current flow (amps), t is the time of current flow from the battery because of kinetic limitations of
(seconds, hours), MW is the molecular or atomic reactions and of other processes must occur to
weight of the material being transformed, and n is produce current flow during operation. Electrochemi-
the number of electrons in the reaction. cal reaction kinetics follow the same general consid-
Assuming thermodynamic reversibility2 of the cell erations as those for bulk chemical reactions. How-
reaction and with the help of eqs 1 and 3, we can ever, electrode kinetics differs from chemical kinetics
obtain the reversible heat effect. in two important aspects: (1) the influence of the
potential drop in the electrical double layer at an
∆G ) -nFE ) ∆H - T∆S (8) electrode interface as it directly affects the activated
) ∆H - nFT(dE/dT) (9) comples and (2) the fact that reactions at electrode
interfaces proceed in a two-dimensional, not three-
By measuring the cell voltage as a function of dimensional, manner. The detailed mechanism of
temperature, the various thermodynamic quantities battery electrode reactions often involves a series of
for the materials in an electrode reaction can be physical, chemical, and electrochemical steps, includ-
determined experimentally. If dE/dT is positive, the ing charge-transfer and charge transport reactions.
cells will heat on charge and cool on discharge. Lead The rates of these individual steps determine the
acid is an example of a negative dE/dT, where the kinetics of the electrode and, thus, of the cell/battery.
cells cool on charge and heat on discharge. Ni-Cd is Basically, three different kinetics effects for polariza-
an example of a positive dE/dT, where the cells heat tion have to be considered: (1) activation polarization
on charge and cool on discharge. Heating and cooling is related to the kinetics of the electrochemical redox
of the cell can proceed with heat exchange with the (or charge-transfer) reactions taking place at the
environment. In general, the entropic heat is negli- electrode/electrolyte interfaces of anode and cathode;
gibly small compared to the irreversible heat re- (2) ohmic polarization is interconnected to the re-
leased, q, when a cell is in operation. Equation 10 sistance of individual cell components and to the
describes total heat release, including the reversible resistance due to contact problems between the cell
thermodynamic heat release along with the irrevers- components; (3) concentration polarization is due to
ible joule heat from operation of the cell in an mass transport limitations during cell operation. The
irreversible manner, during charge or discharge at polarization, η, is given by
finite current/rate. Irreversible behavior manifests
itself as a departure from the equilibrium or thermo- η ) EOCV - ET (12)
dynamic voltage. In this situation, the heat, q, given
off by the system is expressed by an equation in where EOCV is the voltage of the cell at open circuit
which ET is the practical cell terminal voltage and and ET is the terminal cell voltage with current, I,
EOCV is the voltage of the cell on open circuit. flowing.
Activation polarization arises from kinetics hin-
q ) T∆S + I(EOCV - ET) (10) drances of the charge-transfer reaction taking place
at the electrode/electrolyte interface. This type of
q ) heat given off by the system (11) kinetics is best understood using the absolute reac-
tion rate theory or the transition state theory. In
The total heat released during cell discharge is the
these treatments, the path followed by the reaction
sum of the thermodynamic entropy contribution plus
proceeds by a route involving an activated complex,
the irreversible contribution. This heat is released
where the rate-limiting step is the dissociation of the
inside the battery at the reaction site on the surface
activated complex. The rate, current flow, i (I ) I/A
of the electrode structures. Heat release is not a
and Io ) Io/A, where A is the electrode surface area),
2 A process is thermodynamically reversible when an infinitesimal
of a charge-transfer-controlled battery reaction can
reversal in a driving force causes the process to reverse its direction.
be given by the Butler-Volmer equation as
Since all actual processes occur at finite rates, they cannot proceed
with strict thermodynamic reversibility and thus additional nonrevers-
ible effects have to be regarded. In this case, under practical operation
i ) io exp(RFη/RT) - exp((1 - R)Fη)/RT (13)
conditions, voltage losses at internal resistances in the cell (these
kinetic effects are discussed below) lead to the irreversible heat
production (so-called Joule heat) in addition to the thermodynamic where the exchange current density, io ) koFA is the
reversible heat effect. exchange current density (ko is the reaction rate
constant for the electrode reaction, and A is the

activity product of the reactants), η is the polarization
or departure (overpotential) from equilibrium (η )
EOCV - ET), and R is the transfer coefficient, which
is best considered as the fraction of the change of
overpotential that leads to a change in the rate
constant for charge-transfer reaction. The exchange
current density is directly related to the reaction rate
constant, to the activities of reactants and products,
and to the potential drop across the double layer.
Reactions with larger io are more reversible and have
lower polarization for a given current flow. Electrode
reactions having high exchange currents (io in the Figure 6. Typical discharge curve of a battery, showing
range of 10-2 A/cm2) at room temperature are favored the influence of the various types of polarization.
for use in battery applications. The buildup and decay
of the activation polarization are fast and can be
identified by the voltage change on current interrup- by all three types of polarization. A variety of
tion in a time frame of 10-2-10-4 s. experimental techniques are used to study electro-
The activation polarization follows the Tafel equa- chemical and battery reactions. The most common
tion derived from eq 13 are the direct measurement of the instantaneous
current-voltage characteristics on discharge curve
η ) a - b log(I/Io) (14) shown in Figure 6. This curve can be used to deter-
mine the cell capacity, the effect of the dischage-
where a and b are constants. charge rate, and temperature and information on the
Ohmic polarization arises from the resistance of the state of health of the battery.
electrolyte, the conductive diluent, and materials of The impedance behavior of a battery is another
construction of the electrodes, current collectors, common technique that can reveal a significant
terminals, and contact between particles of the active amount of information about battery operation char-
mass and conductive diluent or from a resistive film acteristics. The impedance of an electrode or battery
on the surface of the electrode. Ohmic polarization is given by
appears and disappears instantaneously (e10-6 s)
when current flows and ceases. Under the effect of
Z ) R + jωX (17)
ohmic resistance, R, there is a linear Ohm’s Law
relationship between I and η.
where X ) ωL - 1/(ωC), j ) x-1, and ω is the
η ) IR (15) angular frequency (2πf); L is the inductance, and C
As the redox reactions proceed, the availability of is the capacitance. A schematic of a battery circuit
the active species at the electrode/electrolyte interface and the corresponding Argand diagram, illustrating
changes. Concentration polarization arises from lim- the behavior of the simple electrode processes, are
ited mass transport capabilities, for example, limited shown in Figure 7a. Activation processes exhibit a
diffusion of active species to and from the electrode semicircular behavior with frequency that is char-
surface to replace the reacted material to sustain the acteristic of relaxation processes; concentration
reaction. Diffusion limitations are relatively slow, processes exhibit a 45° behavior characteristic of
and the buildup and decay take g10-2 s to appear. diffusion processes often referred to as Warburg
For limited diffusion the electrolyte solution, the behavior; ohmic components are independent of
concentration polarization, can be expressed as frequency.
Each electrode reaction has a distinctive, charac-
η ) (RT/n) ln(C/Co) (16) teristic impedance signature. A schematic of a battery
circuit and the corresponding Argand diagram, il-
where C is the concentration at the electrode surface lustrating the behavior of the simple processes, are
and Co is the concentration in the bulk of the solution. shown in Figure 7b. In ideal behavior, activation
The movement or transport of reactants from the processes exhibit a semicircular behavior with fre-
bulk solution to the reaction site at the electrode quency that is characteristic of relaxation processes;
interface and vice versa is a common feature of all concentration processes exhibit a 45° behavior char-
electrode reactions. Most battery electrodes are po- acteristic of diffusion processes, and ohmic polariza-
rous structures in which an interconnected matrix tions have no capacitive character and are independ-
of small solid particles, consisting of both noncon- ent of frequency. The frequency of the maximum, fm,
ductive and electronically conductive materials, is of the semicircle gives the relaxation time, where τ
filled with electrolyte. Porous electrode structures are ) 1/fm ) RC. Here R is related to the exchange
used to extend the available surface area and lower current for the reaction and C is called the polariza-
the current density for more efficient operation. tion capacitance, CP. Typically, the CP is of the order
of 200 µF/cm, ∼10 times larger than the capacitance
1.5. Experimental Techniques of the EDL. Some electrochemical capacitors take
advantage of this capacitance to improve their per-
In practical batteries and fuel cells, the influence formance of the supercapacitors. Battery electrodes
of the current rate on the cell voltage is controlled have large surface areas and, therefore, exhibit large
Figure 7. (A, top) Simple battery circuit diagram, where

CDL represents the capacitance of the electrical double layer
at the electrode-solution interface (cf. discussion of super-
capacitors below), W depicts the Warburg impedance for
diffusion processes, Ri is the internal resistance, and Zanode
and Zcathode are the impedances of the electrode reactions.
These are sometimes represented as a series resistance
capacitance network with values derived from the Argand
diagram. This reaction capacitance can be 10 times the size Figure 8. Primary current distribution on the front
of the double-layer capacitance. The reaction resistance surface of the electrodes based on Kirkhof’s law calculation
component of Z is related to the exchange current for the for three different cell constructions: (A) Both connections
kinetics of the reaction. (B, bottom) Corresponding Argand to the cell are at the top. The higher resistance path at
diagram of the behavior of impedance with frequency, f, the bottom sections of the electrode reduces the current
for an idealized battery system, where the characteristic flow and results in a nonuniform current distribution. (B)
behaviors of ohmic, activation, and diffusion or concentra- All paths have equal resistance, and a uniform current
tion polarizations are depicted. distribution results. (C) The bipolar construction has equal
resistance from one end to the other.
capacitances. It is common for cells to have a capaci-
tance of farads and a resistance of milliohms.
Although the matrix may have a well-defined
The experimental techniques described above of planar surface, there is a complex reaction surface
charge-discharge and impedance are nondestructive. extending throughout the volume of the porous
“Tear-down” analysis or disassembly of spent cells electrode, and the effective active surface may be
and an examination of the various components using many times the geometric surface area. Ideally, when
experimental techniques such as Raman microscopy, a battery produces current, the sites of current
atomic force microscopy, NMR spectroscopy, trans- production extend uniformly throughout the electrode
mission electron microscopy, XAS, and the like can structure. A nonuniform current distribution intro-
be carried out on materials-spent battery electrodes duces an inefficiency and lowers the expected per-
to better understand the phenomena that lead to formance from a battery system. In some cases the
degradation during use. These techniques provide negative electrode is a metallic element, such as zinc
diagnostic techniques that identify materials proper- or lithium metal, of sufficient conductivity to require
ties and materials interactions that limit lifetime, only minimal supporting conductive structures.
performance, and thermal stabiity. The accelerated
Two types of current distribution, primary and
rate calorimeter finds use in identifying safety-
secondary, can be distinguished. The primary distri-
related situations that lead to thermal runaway and
bution is controlled by cell geometry. The placing of
destruction of the battery.
the current collectors strongly influences primary
current distribution on the geometric surface area of
1.6. Current Distribution and Porous Electrodes the electrodes. The monopolar construction is most
common. The differences in current distribution for
Most practical electrodes are a complex composite top connections and opposite end current collection
of powders composed of particles of the active mate- are shown in Figure 8A,B. With opposite end con-
rial, a conductive diluent (usually carbon or metal nections the current distribution is more uniform and
powder), and a polymer binder to hold the mix results in a more efficient use of the active material.
together and bond the mix to a conductive current The bipolar construction depicted in Figure 8C gives
collector. Typically, a composite battery electrode has uniform current distribution wherein the anode
∼30% porosity with a complex surface extending terminal or collector of one cell serves as the current
throughout the volume of the porous electrode. This collector and cathode of the next cell in pile config-
yields a much greater surface area for reaction than uration.
the geometric area and lowers polarization. The pores Secondary current distribution is related to current
of the electrode structures are filled with electrolyte. production sites inside the porous electrode itself. The
electrode structures from the older lead acid automo-

tive technology to the newest lithium ion (Li ion)
technology.
2. Batteries
2.1. Introduction and Market Aspects
Batteries are self-contained units that store chemi-
cal energy and, on demand, convert it directly into
electrical energy to power a variety of applications.
Batteries are divided into three general classes:
primary batteries that are discharged once and
discarded; secondary, rechargeable batteries that can
be discharged and then restored to their original
condition by reversing the current flow through the
cell; and specialty batteries that are designed to fulfill
Figure 9. Schematic porous electrode structure: (A) a specific purpose. The latter are mainly military and
Electrons from the external circuit flow in the current medical batteries that do not find wide commercial
collector which has contact to the conductive matrix in the use for various reasons of cost, environmental issues,
electrode structure. The redox reaction at the electrode
produces electrons that enter the external circuit and flow and limited market application. They generally do
through the load to the cathode, where the reduction not require time to start-up. At low drains, up to 95%
reaction at the cathode accepts the electron from the of the energy is available to do useful work.
external circuit and the reduction reaction. The ions in the
electrolyte carry the current through the device. (B) The
advantages disadvantages
reaction distribution in the porous electrode is shown for
the case where the conductivity of the electrode matrix is operate over a wide low energy content compared
higher than the conductivity of the electrolyte. temperature range to gasoline and other fuels
choice of chemical system expensive compared to coal
incorporation of porous electrode structures increases and voltage and gasoline
operate in any orientation no single general purpose
the surface area and shortens diffusion path lengths do not require pumps, system
to the reaction site. Current-producing reactions can filters, etc.
penetrate into a porous electrode structure to con- variable in size
siderable depth below the surface of the electrodes commonality of cell sizes,
as noted in Figure 9. The location of the reaction site worldwide
can deliver high current pulses
inside a porous electrode is strongly dependent on can choose best battery for a
the characteristics of the electrode structure and specific purpose (portable,
reactions themselves. The key parameters include the mobile, and stationary
conductivity of the electrode matrix, electrolyte con- applications)
ductivity, the exchange current, the diffusion char-
acteristics of reactants and products, and the total Success in the battery market depends largely on
current flow. In addition, the porosity, pore size, and four factors, noted in Figure 10. The market for
tortuosity of the electrode play a role. The effective- batteries in Table 1 is directly related to the applica-
ness of a porous electrode can be estimated from the tions they serve, such as automobiles, cellular phones,
active surface area, S, in cm2/cm3, and the penetra- notebook computers, and other portable electronic
tion depth LP of the reaction process into the porous devices. The growth in any particular segment fol-
electrode. Factors that influence the secondary cur- lows closely the introduction of new devices powered
rent distribution are the conductivity of the electro- by batteries. The introduction of new materials with
lyte and electrode matrix, the exchange current of higher performance parameters gives the various
the reactions, and the thickness of the porous layer. designers freedom to incorporate new functionality
Sophisticated mathematical models to describe and in present products or to create new products to
predict porous electrode performance of practical
systems have been developed. These formulations
based on models of primary and secondary battery
systems permit rapid optimization in the design of
new battery configurations. The high-rate perfor-
mance of the present SLI automotive batteries has
evolved directly from coupling current collector de-
signs with the porous electrode compositions identi-
fied from modeling studies.
Modeling has become an important tool in develop-
ing new battery technology as well as for improving
the performance of existing commercial systems.
Models based on engineering principles of current
distribution and fundamental electrochemical reac- Figure 10. How batteries are judged by users and the
tion parameters can predict the behavior of porous factors that control these criteria.
Table 1. Estimated Battery Market in 2003 ($ Millions

of Dollars)
system market size
primary
carbon-zinc 6500
alkaline 10000
lithium, military, medical, etc. 3400
subtotal 19900
secondary
lead acid 18400
small sealed rechargeable cells
lithium ion 3500
nickel metal hydride 1800
nickel cadmium 1500
othera 3100
subtotal 28300
total battery market 48200

a
Large vented and sealed Ni-Cd, Ni-Fe, Ag-Zn, etc.
expand the market scope. Batteries for notebook Figure 12. Voltage levels in the various sections of the
unit cell of a battery, fuel cell, or electrochemical capacitor.
computers have experienced double-digit growth, The structure and composition of the electrical double layer
whereas the automobile SLI market segment has differ at the anode and cathode.
grown with the gross national product. Batteries can
range in size from aspirin tablet (and even smaller)
with a few tens of mAh, for in-the-ear hearing aids, In practice, a porous electrically insulating material
to a building with 40 MWh for energy storage and containing the electrolyte is often placed between the
emergency power. anode and cathode to prevent the anode from directly
contacting the cathode. Should the anode and cathode
2.2. Battery Operations physically touch, the battery will be shorted and its
full energy released as heat inside the battery.
Figure 11 depicts the basic elements of a battery. Electrical conduction in electrolytic solutions follows
Figure 12 illustrates the operation of a battery, Ohm’s law: E ) IR.
showing the energy levels at the anode (negative) and Battery electrolytes are usually liquid solvent-
cathode (positive) poles and the electrolyte expressed based and can be subdivided into aqueous, nonaque-
in electronvolts. The negative electrode is a good ous, and solid electrolytes. Aqueous electrolytes are
reducing agent (electron donor) such as lithium, zinc, generally salts of strong acids and bases and are
or lead. The positive electrode is an electron acceptor completely dissociated in solution into positive and
such as lithium cobalt oxide, manganese dioxide, or negative ions. The electrolyte provides an ionic
lead oxide. The electrolyte is a pure ionic conductor conduction path as well as a physical separation of
that physically separates the anode from the cathode. the positive and negative electrodes needed for
electrochemical cell operation. Each electrolyte is
stable only within certain voltage ranges. Exceeding
the electrochemical stability window results in its
decomposition. The voltage stability range depends
on the electrolyte composition and its purity level.
In aqueous systems, conductivities of the order of 1
S/cm are common. The high conductivity of aqueous
solvent-based electrolytes is due to their dielectric
constants, which favor stable ionic species, and the
high solvating power, which favors formation of hy-
drogen bridge bonds and allows the unique Grotthus
conductivity mechanism for protons. Thermodynami-
cally, aqueous electrolytes show an electrochemical
stability window of 1.23 V. Kinetic effects may
expand the stability limit to ∼2 V.
In the nonaqueous organic solvent-based systems
used for lithium batteries, the conductivities are of
the order of 10-2-10-3 S/cm-1. Compared to water,
most organic solvents have a lower solvating power
and a lower dielectric constant. This favors ion pair
Figure 11. Block diagram of a cell or battery powering a
device. If a battery is recharged, the load is replaced with
formation, even at low salt concentration. Ion pair
an energy source that imposes a reverse voltage that is formation lowers the conductivity as the ions are no
larger than the battery voltage and the flow of electrons is longer free and bound to each other. Organic elec-
reversed. trolytes show lower conductivities and much higher
Table 2. Common Commercial Battery Systems

common name nominal voltage anode cathode electrolyte
primary
Leclanché (carbon-zinc) 1.5 zinc foil MnO2 (natural) aq ZnCl2-NH4Cl
zinc chloride (carbon-zinc) 1.5 zinc foil electrolytic MnO2 aq ZnCl2
alkaline 1.5 zinc powder electrolytic MnO2 aq KOH
zinc-air 1.2 zinc powder carbon (air) aq KOH
silver-zinc 1.6 zinc powder Ag2O aq KOH
lithium-manganese dioxide 3.0 lithium foil treated MnO2 LiCF3SO3 or LiClO4a
lithium-carbon monofluoride 3.0 lithium foil CFx LiCF3SO3 or LiClO4a
lithium-iron sulfide 1.6 lithium foil FeS2 LiCF3SO3 and/or LiClO4a
rechargeable
lead acid 2.0 lead PbO2 aq H2SO4
nickel-cadmium 1.2 cadmium NiOOH aq KOH
nickel-metal hydride 1.2 MH NiOOH aq KOH
lithium ion 4.0 Li(C) LiCoO2 LiPF6 in nonaqueous solventsa
specialty
nickel-hydrogen 1.2 H2 (Pt) NiOOH aq KOH
lithium-iodine 2.7 Li I2 LiI
lithium-silver-vanadium oxide 3.2 Li Ag2V4O11 LiAsFa
lithium-sulfur dioxide 2.8 Li SO2 (C) SO2-LiBr
lithium-thionyl chloride 3.6 Li SOCl2 (C) SOCl2-LiAlCl4
lithium-iron sulfide (thermal) 1.6 Li FeS2 LiCl-LiBr-LiF
magnesium-silver chloride 1.6 Mg AgCl seawater
a
In nonaqueous solvents. Exact composition depends on the manufacturer, usually propylene carbonate-dimethyl ether for
primary lithium batteries and ethylene carbonate with linear organic carbonates such as dimethyl carbonate, diethyl carbonate,
and ethylmethyl carbonate for lithium ion cells.
Figure 13. Energy storage capability of common com-

Figure 14. Energy storage capability of common recharge-
mercial primary battery systems.
able battery systems.
viscosities than aqueous electrolytes. Organic solvent- 2.4. Primary Batteries

based electrolytes (again with the help of kinetics)
are limited to ∼4.6 V. Exceeding the voltage limit in Figure 15A shows the discharge reaction of a CuS
the organic electrolytes results in polymerization or electrode in a Li-CuS cell. During the cell reaction,
decomposition of the solvent system. Solid electrolyte Cu is displaced by Li and segregates into a distinct
batteries have found limited use as the power source solid phase in the cathode. The products of this
for heart pacemakers and for use in military applica- displacement type of reaction, Li2S and Cu, are
tions. The basic principles described above apply to stable, and the reaction cannot be easily reversed.
fuel cells and electrochemical capacitors as well as Hence, the electrode reactions cannot be recharged
to batteries. and the cell is considered to be a primary cell, as the
discharge reaction is not reversible. The Li electrode
in Figure 15B is discharged by oxidation. The formed
2.3. Characteristics of Common Battery Systems Li+ cation is going into solution. The reaction is
reversible by redeposition of the lithium. However,
A list of common commercial systems is found in like many other metals in batteries, the redeposition
Table 2. A graphical representation of the energy of the Li is not smooth, but rough, mossy, and
storage capability of common types of primary and dendritic, which may result in serious safety prob-
secondary batteries is shown in Figures 13 and 14. lems. This is in contrast to the situation with a lead
It is beyond the scope of this paper to discuss all electrode in Figure 15C, which shows a similar
systems in detail. Instead, we want to review the solution electrode. Here, the formed Pb2+ cation is
most common electrode mechanisms for discharge only slightly soluble in sulfuric acid solution, and
and charge depicted in Figure 15. PbSO4 precipitates at the reaction site on the elec-
Figure 16. Two-step discharge curve of a MnO2 electrode

in aqueous solution showing the influence of one- and two-
phase discharge reaction mechanisms on the shape of the
discharge curve. The different shapes of the discharge
curves can be explained with the help of the Gibbs phase
rule.
Zinc manganese batteries consist of MnO2, a proton

insertion cathode (cf. Figure 15E), and a Zn anode of
the solution type. Depending on the pH of the
electrolyte solution, the Zn2+ cations dissolve in the
electrolyte (similar to the mechanism shown in
Figure 15B) or precipitate as Zn(OH)2 (cf. mechanism
in Figure 15C).
The discharge reaction of the MnO2 electrode
proceeds in two one-electron reduction steps as shown
in the discharge curve (Figure 16). Starting at cell
Figure 15. Schematics showing various discharge and voltages of 1.5 V, a coupled one-electron transfer and
charge mechanisms of battery electrodes, which serve as proton insertion reaction takes place. The transfor-
examples of the battery electrode charge/discharge mech- mation of MnO2 into MnOOH is a one-phase reaction.
anisms discussed in the text. Further reduction leads to a phase change as the
solid MnOOH turns into Mn2+ soluble in the solution,
trode surface. This solution-precipitation mechanism that is, a two-phase reaction. This is consistent with
is also working during the charge reaction, when the Gibbs phase rule that predicts the shape of the
PbSO4 dissolves and is retransformed into metallic discharge curve for one- and two-phase reactions
Pb. Figure 15D shows a typical electrochemical (Figure 16). When the number of phases, P, is equal
insertion reaction. The term “electrochemical inser- to the number of components, C, taking part in the
tion” refers to a solid-state redox reaction involving reaction as in the case of a two-phase reaction, the
electrochemical charge transfer, coupled with inser- number of degrees of freedom F () number of
tion of mobile guest ions (in this case Li+ cations) thermodynamic parameters that have to be specified
to define the system) is 2. If the values of two
from an electrolyte into the structure of a solid host,
parameters, usually pressure, p, and temperature,
which is a mixed, that is, electronic and ionic,
T, are specified, there is no degree of freedom left and
conductor (in this case graphite). Unlike displace-
other parameters of the system such as voltage have
ment type electrodes (Figure 15A) and solution type to be constant. Hence, the cell voltage stays constant
electrodes (Figure 15B), the insertion electrodes for a two-phase discharge reaction. If there is a
(Figure 15D) have the capability for high reversibil- degree of freedom left, as in the case of a one-phase
ity, due to a beneficial combination of structure and reaction, the cell voltage can be a variable and
shape stability. Many secondary batteries rely on changes (slopes-off) during discharge.
insertion electrodes for the anode and cathode. A Zinc-manganese batteries dominate the primary
prerequisite for a good insertion electrode is elec- battery market segment. Leclanché invented the
tronic and ionic conductivity. However, in those original carbon-zinc cell in 1860. He used a natural
materials with poor electronic conductivity, such manganese dioxide-carbon black core cathode with
as MnO2, good battery operation is possible. In this aqueous zinc chloride-ammonium chloride electro-
case, highly conductive additives such as carbon lyte, contained in a zinc can. An alternative version
are incorporated in the electrode matrix, as in employs a zinc chloride electrolyte and a synthetic
Figure 15E. The utilization of the MnO2 starts at the electrolytic manganese dioxide and has better per-
surface, which is in contact with the conductive formance than the original Leclanché cell. The car-
additive and continues from this site throughout the bon-zinc with zinc chloride electrolyte gives about
bulk of the MnO2 particle. Most electrodes in batter- the same performance, on lower radio-type drains,
ies follow one of the basic mechanisms discussed in as the alkaline cell and is strong in the Japanese and
Figure 15. European markets. The carbon-zinc cell still finds
system has captured the hearing aid market. Cells

are available in sizes smaller than an aspirin tablet
that fit into the ear to power the hearing aid.
The main applications of Zn-Ag2O cells are button
cells for watches, pocket calculators, and similar
devices. The cell operates with an alkaline electrolyte.
The Zn electrode operates as discussed, whereas the
Ag2O electrode follows a displacement reaction path
(cf. Figure 15A).
Primary lithium cells use a lithium metal anode,
the discharge reaction of which is depicted in Figure
15B. Due to the strong negative potential of metallic
lithium, cell voltages of 3.7 V or higher are possible.
As the lithium metal is very reactive, the key to
battery chemistry is the identification of a solvent
system that spontaneously forms a very thin protec-
tive layer on the surface of metallic lithium, called
the solid electrolyte interphase (SEI) layer. This
electronically insulating film selectively allows lithium
ion transport. Lithium batteries show higher energy
density than the alkaline cells but have a lower rate
capability because of the lower conductivity of the
Figure 17. Discharge mechanism of a Zn-MnO2 cell. nonaqueous electrolyte and the low lithium cation
From top to bottom, various stages of the discharge reaction transport rate through the SEI.
are depicted. On the Zn side, the local change of the pH Commercial lithium primary batteries use solid
alters the composition of the discharge product. and liquid cathodes. Solid cathodes include carbon
monofluoride, CFx, manganese dioxide (MnO2), FeS2,
wide use, and in 2003, worldwide it outsold the and CuS. Chemical (CMD) or electrolytic manganese
alkaline cell about 30 billion to 12 billion cells. dioxide (EMD) is used as the cathode with high-
In the Unites States, the alkaline electrolyte (KOH) temperature treatment to form a water-free active
version accounts for ∼80% of sales. The detailed material. The CFx is made from the elements, and
discharge reaction mechanism is shown in Figure 17. its cost is somewhat higher than that of the manga-
The influence of the pH and the change of the pH of nese dioxide cathode material. These cells are de-
the electrolyte solution during discharge on the signed for relatively low-rate applications. Both
formation/solubility of various zincate compounds chemistries are very stable, and cells can deliver
[e.g., Zn(OH)42+, etc.] to a change from a one phase g80% of their rated capacity after 10 years of storage.
to a two phase reaction at the point where Zn(OH)2 The lithium iron sulfide (FeS2) system has been
begins to precipitate. It should be noted that local developed for high-rate applications and gives supe-
pH changes occur also during discharge of the MnO2 rior high-rate performance compared to the alkaline
electrode (Figure 15E). zinc-manganese cells. Typically, the electrolyte is
Compared to the carbon-zinc cell, the alkaline cell propylene carbonate-dimethyl ether (PC-DME) with
is more reliable, has better performance, and is best lithium triflate (LiSO3CF3) or lithium perchlorate
for higher rate applications required for advanced (LiClO4) salt.
portable electronic devices. The current version of the The lithium thionyl chloride system employs a
alkaline cell is mercury free. Instead, it uses a soluble cathode construction. The thionyl chloride
combination of alloying agents and corrosion inhibi- acts as the solvent for the electrolyte and the cathode
tors to lower the hydrogen gas generation from active material. It has the highest energy density of
corrosion of the zinc anode and to compensate for the any lithium cell and is equal to that of the zinc-air
corrosion protection originally provided by the mer- aqueous cell. The reaction mechanism of the cell is
cury. A synthetic gel holds the zinc powder anode explained in Figure 18. In the inorganic electrolyte
together. The cathode is composed of an electrolytic (LiAlCl4 dissolved in SOCl2) the lithium metal and
manganese dioxide-graphite mixture with critical the electrolyte react chemically and form a SEI
impurities controlled to a e1 ppm level. mainly consisting of LiCl and S. LiCl and S are also
The zinc-air battery system has the highest en- the products of the electrochemical discharge reaction
ergy density of all aqueous batteries and equals that at the carbon positive electrode, where the liquid
of the lithium thionyl chloride battery (which is the cathode SOCl2 is reduced. The cell discharge stops
highest energy density lithium battery). The high when the electronically insulating discharge products
energy density results from the cell design, as only block the carbon electrode.
the zinc powder anode is contained in the cell. The Lithium-sulfur dioxide cells also use a liquid
other reactant, oxygen, is available from the sur- cathode construction. The SO2 is dissolved in an
rounding air. The air electrode is a polymer-bonded organic solvent such as PC or acetonitrile. Alterna-
carbon, sometimes catalyzed with manganese di- tively, SO2 is pressurized at several bars to use it in
oxide. The electrode has a construction similar to that the liquid state. The cell reaction is similar to that
of fuel cell electrodes (see section 3). The zinc-air depicted in Figure 18, with electronically insulating
sulfuric acid and water have to be provided for the

cell reaction and formation of the battery electrolyte.
For ionic conductivity in the charged and discharged
states, an excess of acid is necessary. Considering the
limited mass utilization and the necessity of inactive
components such as grids, separators, cell containers,
etc., the practical value of specific energy (Wh/kg) is
only ∼25% of the theoretical one (Figure 19) for
rechargeable batteries. Due to the heavy electrode
and electrolyte components used, the specific energy
is low.3 Nevertheless, the lead acid system serves a
variety of applications from automotive SLI and
motive power for forklift trucks and the like to
stationary energy storage for uninterruptible power
Figure 18. Discharge mechanism of a Li-SOCl2 cell. The supplies. Its low cost and established recycling pro-
cell can operate until the surface of the carbon cathode is cesses make it one of the “greenest” battery systems.
fully covered by electronically insulating LiCl and S According to Battery Council International, Inc.,
discharge products. The Li-SO2 cell is also a soluble ∼98% of the lead acid batteries in the United States
cathode system with a cell construction similar to that of are recycled.
the Li-SOCl2 cell. It follows a similar discharge reaction
where the reaction product is LiS2O4.
Nickel-cadmium (Ni-Cd) was the first small
sealed rechargeable cell. In alkaline (KOH) electro-
Li2S2O4 being the SEI component at the Li anode and lyte, the Cd negative electrode functions reversibly,
the solid discharge product at the carbon cathode. according to a solution-precipitation mechanism (cf.
The Li-SOCl2 and Li-SO2 systems have excellent Figure 15C) with Cd(OH)2 being the discharge prod-
operational characteristics in a temperature range uct. The Ni positive electrode is actually a Ni(OH)2
from -40 to 60 °C (SOCl2) or 80 °C (SO2). Typical electrode, which is able to reversibly de-insert/insert
applications are military, security, transponder, and protons during discharge/charge. It has excellent low-
car electronics. Primary lithium cells have also temperature and high-rate capabilities. For a long
various medical uses. The lithium-silver-vanadium time, it was the only battery available for power tools.
oxide system finds application in heart defibrillators. It powered the early cellular phones and portable
The lithium-iodine system with a lithium iodide computers. The availability of stable hydrogen stor-
solid electrolyte is the preferred pacemaker cell. age alloys provided the impetus for the creation of
the nickel-metal hydride (Ni-MH) cell. The hydro-
2.5. Rechargeable Batteries gen storage alloy is a proton-inserting negative
Rechargeable cells generally have lower energy electrode material that replaced the environmentally
storage capability than primary cells. The additional threatened cadmium negative electrode in the Ni-
requirements for rechargeability and long operation Cd. The positive electrode and the electrolyte stayed
limit the choice of chemical systems and construc- the same. Ni-MH quickly replaced the Ni-Cd for
tions to those that are more robust than for primary electronic applications because of its significantly
batteries. The lead acid battery dominates the re- higher energy storage capability and somewhat lighter
chargeable market. Both the Pb and PbO2 electrode weight. The Ni-MH has poor low-temperature ca-
reaction mechanisms follow the solution-precipita- pability and limited high-rate capability, but its
tion mechanism as depicted in Figure 15C and the higher energy density served to spur the development
cell reaction shown in Figure 19. In addition to the of the portable electronic device market. Today, it is
lead and lead oxide electrodes, sufficient amounts of the battery of choice for hybrid gasoline-electric
vehicles and is beginning to challenge the Ni-Cd for
power tool applications.
The Li ion battery, with significantly higher energy
density and lighter weight, replaced the Ni-MH as
soon as production capability was available. It is now
the battery of choice for portable electronic devices
and is challenging the Ni-MH for the hybrid vehicle
application. The Li ion cell has a carbon/graphite
anode, a lithium-cobalt oxide cathode, and an or-
ganic electrolyte of lithium hexafluorophosphate
(LiPF6) salt with ethylene carbonate-organic solvent
mixture. As in the Ni-MH battery, both the anode
and cathode in the Li ion cell follow an insertion
mechanism; however, instead of protons, lithium
cations are inserted and de-inserted (cf. Figure 15D).4
Figure 19. Depiction of the components of a lead acid 3 A specific energy of 30 Wh/kg literally means that 1 kg of a lead
battery showing the differences between theoretical and acid battery is able to power a 60 W lamp for only 0.5 h.
practical energy density of a lead acid battery and source 4 For layered host materials as used in the lithium ion cell, the term
of the differences. “intercalation” is used for the insertion of guests into the host structure.
materials but with high-rate performance for hybrid

electric vehicle and power tool applications. The new
materials that make possible this improvement in
performance are discussed elsewhere and are beyond
the scope of this overview.
Lithium metal rechargeable cells would have the
highest energy of all battery systems. Unfortunately,
on recharge, the lithium has a strong tendency to
form mossy deposits and dendrites in the usual liquid
organic solvents (cf. Figure 15B). This limits the cycle
life to ∼100-150 cycles (considerably lower that the
300 cycles required for a commercial cell), as well as
increasing the risk of a safety incident. Rechargeable
lithium metal-vanadium oxide cells (Li-VOx) with
poly(ethylene oxide) polymer electrolytes have been
developed for stationary energy storage applications.
Only a few of the thousands of proprosed battery
systems have been commercialized. A set of criteria
can be established to characterize reactions suitable
for use in selecting chemical systems for commercial
battery development. Very few combinations can
meet all of the criteria for a general purpose power
supply. The fact that two of the major battery
systems introduced more than 100 years ago, lead
acid (rechargeable) and zinc-manganese dioxide
(primary), are still the major systems in their cat-
egory is indicative of the selection process for chemi-
cal reactions that can serve the battery marketplace.
2.6. Selection Criteria for Commercial Battery

Systems
A set of criteria that illustrate the characteristics
of the materials and reactions for a commercial
battery system follow.
1. Mechanical and Chemical Stability. The materi-
als must maintain their mechanical properties and
their chemical structure, composition, and surface
over the course of time and temperature as much as
possible. This characteristic relates to the essential
reliability characteristic of energy on demand. Ini-
tially, commercial systems were derived from materi-
als as they are found in nature. Today, synthetic
materials can be produced with long life and excellent
stability. When placed in a battery, the reactants or
active masses and cell components must be stable
over time in the operating environment. In this
respect it should be noted that, typically, batteries
Figure 20. Construction of (A) cylindrical, (B) prismatic, reach the consumer ∼9 months after their original
and (C) polymer Li ion cells. (Reprinted with permission
from a brochure by Sony Corporation). assembly. Mechanical and chemical stability limita-
tions arise from reaction with the electrolyte, ir-
reversible phase changes and corrosion, isolation of
Sony introduced the Li ion in 1991. Since its intro- active materials, and local, poor conductivity of
duction, it has more than doubled in capacity in materials in the discharged state, etc.
response to the demand for higher performance 2. Energy Storage Capability. The reactants must
portable electronic devices, such as cellular phones have sufficient energy content to provide a useful
and notebook computers. The construction of the Li voltage and current level, measured in Wh/L or Wh/
ion cells is shown in Figure 20. Cells are available kg. In addition, the reactants must be capable of
in both liquid electrolyte and plasticized polymer delivering useful rates of electricity, measured in
electrolyte configurations. New anode and cathode terms of W/L or W/kg. This implies that the kinetics
materials hold promise to double the present perfor- of the cell reaction are fast and without significant
mance over the next 10 years. The Li ion market is kinetics hindrances. The carbon-zinc and Ni-Cd
poised to segment into the higher performance, systems set the lower limit of storage and release
higher cost segment, which continues the increase capability for primary and rechargeable batteries,
in energy density, and a segment with lower cost respectively.
3. Temperature Range of Operation. For military of being restored exactly to their original condition
applications, the operational temperature range is (crystal structure, chemical composition, etc.) on
from -50 to 85 °C. Essentially the same temperature reversal of the current flow (charging). After being
range applies to automotive applications. For a recharged by current reversal, the electrode materials
general purpose consumer battery, the operating must be able to deliver the same rate of discharge
temperature range is ∼0-40 °C, and the storage while maintaining their voltage level. Very few
temperatures range from -20 to 85 °C. These tem- chemical systems exhibit this characteristic.
peratures are encountered when using automobiles 9. Cycle Life. It is not enough for a chemical system
and hand-held devices in the winter in northern to be recharged and deliver power to qualify as a
areas and in the hot summer sun in southern areas. commercial rechargeable system. A commercial cell
4. Self-Discharge. Self-discharge is the loss of must be capable of completely discharging its energy
performance when a battery is not in use. An accept- and then fully recharging a minimum of 300 times
able rate of loss of energy in a battery depends and not lose >20% of its capacity. This requires a
somewhat on the application and the chemistry of very robust system and reversible electrode reactions.
the system. People expect a battery to perform its There can be no side reactions that result in the loss
intended task on demand. Li-MnO2 primary cells of the active materials during the charge-discharge
will deliver 90% of their energy even after 8 years cycle.
on the shelf; that is, their self-discharge is low. Some 10. Charge Time. The time it takes to recharge a
military batteries have a 20-year storage life and still battery completely relates to the use. For conven-
deliver their rated capacity. On the other hand, ience, recharging in 15 min is accepted for many
rechargeable batteries can be electrically restored to consumer applications. However, fast charging places
their operating condition and generally have more a stress on the robustness of the electrode reactions
rapid loss of capacity on storage. The rechargeable and may result in shortened cycle life. Most batteries
Ni-MH cell, for instance, will lose up to 30% of its require 3-8 h to recharge completely and maintain
capacity in a month. Usually, self-discharge increases their required cycle life. This slower charge rate
with temperature. allows time for the atoms and molecules to find their
5. Shape of the Discharge Curve. The issue of a correct positions in the charged material.
sloping versus a flat discharge depends on the 11. Overcharge/Overdischarge Protection. When a
intended use. For operation of an electronic device, battery is forced outside its thermodynamic voltage
a flat, unchanging, discharge voltage is preferred. A levels, the reaction path becomes unstable; irrevers-
sloping discharge is preferred for applications when ible new reactions can occur, and new compounds can
determining the state-of-charge is important. This form. These events harm the active material and
may be modified somewhat by the impact of cost. either reduce the capacity or render the system
Although a constant brightness is preferred in a inoperable. In addition, unsafe battery conditions
flashlight, the user may select carbon-zinc with a may occur under overcharge/overdischarge condi-
sloping discharge for its lower cost. The influence of tions. The Ni-Cd, Ni-MH, and lead acid have a
one- or two-phase reactions on the shape of the built-in overcharge and overdischarge characteristic
discharge curve was discussed previously (cf. Figure based on an oxygen recombination mechanism. Cell
16). designs often use the ratio of the capacities of each
6. Cost. The cost of the battery is determined by electrode (cell balance) to accomplish protection of the
the materials used in its fabrication and the manu- battery system. It is also possible to use electronic
facturing process. The manufacturer must be able to controls to control the charge and discharge voltage
make a profit on the sale to the customer. The selling limits within safe limits. The lithium-cobalt oxide
price must be in keeping with its perceived value cathode in the Li ion system is protected from
(tradeoff of the ability of the user to pay the price overvoltage and overdischarge by electronic means.
and the performance of the battery). Alkaline pri- Voltage excursions outside its operating range can
mary Zn-MnO2 is perceived to be the best value in cause irreversible changes in its crystal structure and
the United States. However, in Europe and Japan damage cell operations.
the zinc chloride battery still has a significant mar-
ket share. In developing countries, the lower cost 3. Fuel Cells
Leclanché carbon-zinc is preferred. Likewise, lead
acid batteries are preferred for automobile SLI over 3.1. Introduction and Market Aspects
Ni-Cd with superior low-temperature performance The chemical energy stored in hydrogen and sev-
but with a 10 times higher cost. eral hydrocarbon fuels is significantly higher than
7. Safety. All consumer and commercial batteries that found in common battery materials. This fact
must be safe in the normal operating environment provides the impetus to develop fuel cells for a variety
and not present any hazard under mild abuse condi- of applications. Fuel cells are an ideal primary energy
tions. The cell or battery should not leak, vent conversion device for remote site locations and find
hazardous materials, or explode. application where an assured electrical supply is
Added criteria for rechargeable batteries are as required for power generation, distributed power,
follows: remote, and uninterruptible power. Figure 21 depicts
8. Ability To Recharge and Deliver Power. The the operation of typical fuel cells. The various com-
rechargeable battery systems place a severe added ponents of a functioning fuel cell are shown in
requirement. The active materials must be capable Figures 22 and 23. Operating fuel cells are complex
Figure 21. Summary of the reactions and processes that

occur in the various fuel cell systems. Figure 22. Block diagram of the component parts of a
functioning fuel cell.
chemical plants that require sophisticated manufac-
turing techniques and control circuitry.
Some companies, hospitals, and buildings are
choosing to install fuel cell power in order to be free
from the outages experienced in utility supply lines.
Previously, emphasis had been placed on the devel-
opment of large fuel cells in the 200-300 kW range
for these applications. Several demonstration projects
of 1 MW and larger fuel cells have been undertaken,
usually composed of units of 250 kW output. Smaller
fuel cells in the range of 50-75 kW are now under
intense development for use in automobile and bus
propulsion, where their low emission characteristic Figure 23. Depiction of the components of a complete fuel
finds favor. The direct conversion methanol fuel cells cell system including the re-former and power conditioning
in the range of 5-25 W are proposed for portable unit.
electronics as a replacement for Li ion and Ni-MH
will continue to be used in space applications, where
batteries. The more promising commercial applica-
high energy densities and cryogenic storage of hy-
tions of fuel cells appear to be as a stationary power
drogen and oxygen represent a cost savings by
source for central and dispersed power stations
lowering the weight of the launch vehicle. They are
(megawatts) and as mobile power for portable elec-
also used for longer space missions, when the high
tronic devices and automobiles.
reliability and the production of drinking water are
important considerations. The high cost of qualifying
advantages disadvantages a new technology is a strong impediment for replac-
efficient energy conversion complex to operate ing the AFC in space applications.
modular construction best as primary energy source Fuel cells offer the cleanest power generation
nonpolluting impurities in gas stream shorten life possible. They are quiet in operation and can be
low maintenance pulse demands shorten cell life located close to the application. They produce much
silent expensive
safe limited availability less greenhouse emissions and can be more efficient
high energy density low durability in conversion of the energy in a fuel into power than
low power density per volume gasoline engines or utility thermal power plants. Fuel
cells are best utilized as a steady energy source and
Although several fuel cell technologies are reaching not as a power source to supply dynamic demands.
technical maturity, the economics of a fuel cell are For applications that require varying power de-
not clear. The commercial potential of fuel cells will mands, such as automotive propulsion, the use of the
depend on the ability to reduce catalyst and other fuel cell in a hybrid configuration with a battery or
expensive materials costs and to manufacture the EC will be required. The fuel cell provides steady
units at a competitive cost. Many uses of fuel cells power demand while the battery or EC handles the
place a premium on specific performance character- surge for regenerative breaking and acceleration as
istics. The relatively simple alkaline fuel cells (AFC) well as initial start-up.
Table 3. Types of Fuel Cells

advantages disadvantages comments
Alkaline(AFC)a
mechanically rechargeable limited activated life original development >30 years ago
low-cost KOH electrolyte intolerant of impurities in gas streams Apollo fuel cell
CO2 and CO operates at room temp to 80 °C
pure H2 only suitable fuel demo in vehicles in the 1970s
Polymer Electrolyte Membrane Fuel Cell (PEMFC)
nonvolatile electrolyte expensive catalysts required operates best at 60-90 °C
few materials problems CO a strong poison originally developed for space by GE
CO2 rejecting electrolyte H2O management essential hydrogen fuel (re-formed hydrocarbons, pure H2,
pressure differential between high-cost electrolyte MH storage)
anode and cathode pure H2 only suitable fuel main development efforts for automotive and
polymer electrolyte oxygen kinetics are slow stationary applications
intolerant of impurities
limited life
water management essential
Direct Methanol Fuel Cell (DMFC)
direct fuel conversion stable reaction intermediates operates best at 60 to 90 °C)
slow electrode kinetics high catalyst loadings same construction as PEMFC
improved wt and vol water management essential methanol fuel eliminates reformer
polymer electrolyte low overall efficiency lower current capability
methanol hazardous methanol crossover reduces efficiency
needs new membrane, higher efficiency
high catalyst loadings
main effort for portable electronic devices
Phosphoric Acid Fuel Cell (PAFC)
CO2 rejecting electrolyte H2 only suitable fuel operates best at ∼200 °C
high fuel efficiency anode CO catalyst poison stationary energy storage (nominal units, 250 kW)
O2 kinetics hindered available commercially
low conductivity electrolyte
high-cost catalysts
limited life
Molten Carbonate Fuel Cell (MCFC)
fast electrode kinetics materials problems and life operates best at 550 °C
high efficiency low sulfur tolerance nickel catalysts, ceramic separator membrane
CO/CH4 usable fuel high ohmic electrolyte hydrocarbon fuels re-formed in situ
direct reforming feasible low tolerance to sulfur several large demonstration units
high-grade heat available need to recycle CO2 significant government support
limited life
Solid Oxide Fuel Cell (SOFC)
high grade heat available high fabrication costs operates at 900 °C
fast electrode kinetics severe materials constraints conducting ceramic oxide electrodes
in situ reforming feasible high electrolyte conductivity hydrocarbon fuels re-formed in situ
no electrolyte management least sensitive to sulfur, etc.
high system efficiency
tolerant of impurities
a
Metal air batteries with replaceable anodes are often considered to be a fuel cell but are not considered here.
Fuel cells have been identified as a nearly ideal mance barrier in a small size as well as governmental
solution to power the requirements for motor vehicle regulations.
manufacturers, utility and nonutility generators, and It is estimated that the fuel cell market for dis-
portable electronic devices. Each segment has sig- tributed power and demonstration projects and con-
nificant incentives to develop alternate power sources. tracts amounted to about $100 million for 2003.
For automakers, the incentive is environmentally Research and development contracts to develop fuel
related, coupled with strong support from govern- cells for automotive propulsion and stationary energy
ment programs designed to move away from fossil storage are an order of magnitude larger.
hydrocarbon fuels. For the utilities, increasing envi-
ronmental pressure and power demand, coupled with 3.2. Fuel Cell Operation
limited generation capability, have created a need for
distributed power generation and storage. In a simi- Fuel cells, like batteries, convert the chemical
lar vein, those who need an uninterruptible power energy residing in a fuel into electrical energy on
supply, or freedom from the utility grid, find fuel cells demand. As in batteries and other electrochemical
to be an attractive option. Direct methanol fuel cells cells, fuel cells consist of an anode, where oxidation
for portable electronic devices such as notebook occurs, a cathode, where reduction occurs, and an
computers seem close to commercial reality and will electrolyte, where ions carry the current between the
compete with batteries for this market. The key electrodes. Fuel cells differ from batteries in that the
challenge for each will be to meet the cost-perfor- fuel and oxidant are not contained within the fuel
cell compartment but supplied continuously from an
ambient-90
ambient-90
operating
temp, °C
external source. In a real sense, fuel cells are like
internal combustion engines, as they operate only so
60-90
900a
long as the fuel is supplied. Fuel cells are not
550
200
electrically recharged, but after use, the tank is
refilled with fuel. From an operational point of view,
yttria-stabilized or yttria/calcia-
the fuel of choice is hydrogen gas, with the exhaust
phosphoric acid in SiC matrix
stabilized zirconia support

gas being water. Other fuels and hydrocarbons must
molten Li2CO3 in LiAlO2-

be converted to hydrogen for use in a fuel cell. The
direct conversion of fuels such as CH3OH and CH4
electrolyte
is possible under certain conditions. Each type of fuel
cell has a unique set of processes and reactions to
acidic polymer
acidic polymer
describe its operation. A summary of the character-
istics for the various fuel cell systems is given in
aq KOH
Tables 3 and 4.
The basic reaction of a H2-O2 fuel cell is
H2 + 1/2O2 ) H2O E° ) 1.229 V (18)
strontia-doped lanthanum-
oxygen (or air) carbon/platinum catalyst
carbon/platinum catalyst
manganite Perovskite
cathode composition
For this reaction ∆G° ) -235.76 kJ/mol and ∆H°
) -285.15 kJ/mol. Fuel cells follow the thermody-
and alternatives
namics, kinetics, and operational characteristics for
electrochemical systems outlined in sections 1 and
2. The chemical energy present in the combination
porous NiO
of hydrogen and oxygen is converted into electrical
energy by controlled electrochemical reactions at each
of the electrodes in the cell.
Fuel cells can be roughly divided into low-temper-
ature (ca. <200 °C) and high-temperature (ca. >450
cathode feed
porous cermet of Ni or Co oxygen or air

carbon/platinum catalyst oxygen or air
methanol or methanol-water carbon/platinum catalyst oxygen or air
carbon/platinum catalyst oxygen or air
oxygen or air
°C) fuel cells. Low-temperature fuel cells typically use
alkaline or acidic electrolytes. In acidic electrolytes
the electrode reactions are
anode: H2 - 2e- ) 2H+ (19) carbon/platinum catalyst
cathode: O2 + 2H+ + 2e- ) H2O2 (20)
and yttria-zirconia
anode composition
H2O2 + 2H+ + 2e- ) 2H2O (21)
overall: O2 + 4H+ + 4e- ) 2H2O (22)

porous Ni
Table 4. Typical Characteristics of Various Fuel Cell Systems
Oxygen undergoes a two-step indirect reduction

Mg- or Sr-doped lanthanum chromate as cell interconnects.
reaction. The stable H2O2 intermediate is undesir-

able, as it lowers the cell voltage and H2O2 attacks
and corrodes the carbonaceous electrode material
commonly used. Better catalysts are needed to speed
molten carbonate (MCFC) hydrogen or natural gas
the decomposition of H2O2 to reduce its impact on the

gasoline or natural gas
overall reaction. Similarly, a catalyst can enhance the

anode feed
H2 dissociation rate at the anode. Platinum or

platinum alloys are usually employed as catalysts to
pure hydrogen
pure hydrogen
pure hydrogen
speed the reactions on both electrodes in low-tem-

perature fuel cells. The hydrogen reaction kinetics
are fairly fast and do not require as high platinum
loading to deliver high currents. To suppress the two-
step oxygen reduction, significantly more catalyst is
required. Due to the very high price of nanostruc-
direct methanol (DMFC)
tured Pt, the Pt may be distributed on a porous high

phosphoric acid (PAFC)
surface area carbon. Typical catalyst loadings are

∼0.1 mg/cm2 of platinum or platinum alloy catalyst
solid oxide (SOFC)
on the anode and ∼0.5 mg/cm2 catalyst on the

type
PEM (PEMFC)
alkaline (AFC)
cathode. However, the catalyst loading varies, de-

pending on the type of fuel cell and its application.
A major thrust in research on low-temperature fuel
cells is to reduce the catalyst loading, to improve CO
a
tolerance, and to identify lower cost catalysts. As the

Figure 24. Typical power curve for a fuel cell. The voltage
drops quickly from the OCV due to the formation of the
peroxide intermediate. Operation of the fuel cell at the knee
of the curve where concentration is limiting performance
can damage the electrodes and lead to rapid deterioration
of cell operation.
O2 and H2 dissociation kinetics are better at higher

temperatures (>400 °C), low-cost electrode structures
of high surface area Ni and oxides such as spinels or
perovskites to replace the very effective, but costly,
Pt catalysts have been sought.
Fuel cell electrodes are more complex structures Figure 25. (A) Comparison of the energy storage capabil-
than battery electrodes. They serve three functions: ity of fuel cells and batteries. Only after several refueling
(1) to ensure a stable interface between the reactant operations are fuel cells more efficient energy storage
gas and the electrolyte, (2) to catalyze the electrode devices on a Wh/L and Wh/kg basis. (B) Fuel cells have a
reactions, and (3) to conduct the electrons from or to set volume and weight for the fuel cell stack and peripher-
the reaction sites. A significant problem is the control als to supply the reactants to the stack. The small
of the interface at the juncture of the reactant (gas) incremental fuel volume to continue operation supplying
energy makes them more efficient for longer operations.
phase, the electrolyte medium, and the catalyzed
conducting electrode, the so-called “three-phase bound-
in low-temperature fuel cells, the hydrocarbon fuels
ary”, where the electrolyte, electrode, and gas all
come together. A stable three-phase boundary is must be converted into hydrogen through a separate
critical to good performance and long operation. re-forming step placed between the tank and the fuel
Therefore, the porosity and the wetting behavior with cell. Any trace amounts of CO, CO2, and H2S found
electrolyte and the electrode surface must be pre- in the hydrogen-rich re-formate gas must be removed
cisely adjusted. before using the gas, as they irreversibly block the
Fuel cells can operate with very high electrical Pt catalyst and thus “poison” the electrode. To reduce
efficiencies approaching 60-70%. If the waste heat dependence on hydrocarbon fuels, the use of off-peak
of the fuel cell is also used, fuel efficiencies of 90% electricity from a series of new nuclear plants has
are possible. The performance of the fuel cell is been proposed, to produce the hydrogen for use in
judged by the voltage-current curve depicted in automotive applications.
Figure 24. The operating voltage is significantly less Apart from hydrocarbons and gasoline, other pos-
than the theoretical 1.23 V as the actual cell reaction sible fuels include hydrazine, ammonia, and metha-
involves H2, O2, and H2O and also the peroxide nol, to mention just a few. Fuel cells powered by
intermediate (cf. above), the latter influencing the direct conversion of liquid methanol have promise as
Nernst voltage and the OCV. The activation, ohmic, a possible alternative to batteries for portable elec-
and diffusion polarization characteristics determine tronic devices (cf. below). These considerations al-
the shape of the fuel cell discharge curve as they do ready indicate that fuel cells are not stand-alone
for batteries (cf. Figure 7). devices, but need many supporting accessories, which
As for the other electrochemical storage/conversion consume current produced by the cell and thus lower
devices, the fuel cell electrolyte must be a pure ionic the overall electrical efficiencies. The schematic of the
conductor to prevent an internal short circuit of the major components of a so-called “fuel cell system” is
cell. It may have an inert matrix that serves to shown in Figure 22. Fuel cell systems require so-
physically separate the two electrodes. Fuel cells may phisticated control systems to provide accurate me-
contain all kinds of electrolytes including liquid, tering of the fuel and air and to exhaust the reaction
polymer, molten salt, or ceramic. products. Important operational factors include stoi-
Hydrogen gas is the preferred fuel for low-temper- chiometry of the reactants, pressure balance across
ature fuel cells. The main obstacle in the use of the separator membrane, and freedom from impuri-
hydrogen as energy carrier is that hydrogen is not a ties that shorten life (i.e., poison the catalysts).
readily available fuel. In high-temperature fuel cells, Depending on the application, a power-conditioning
a hydrocarbon fuel, for example, methane or gasoline, unit may be added to convert the direct current from
can be fed directly into the cell. To use hydrocarbons the fuel cell into alternating current.
Fuel cell systems are not very volume efficient. The

ancillary equipment to operate the device can be two
to three times larger than the fuel cells themselves,
where energy conversion actually occurs. Everything
considered, the fuel cell stack5 is much less than 50%
of the total volume of the unit. In addition, fuel cell
stacks themselves have a low energy density in Wh/L
and Wh/kg compared to batteries as the gas distribu-
tion system requires significant extra volume and
weight. Only by taking into account the continuous
supply of fuel to the system during long-term opera-
tion is the energy density of the fuel cell greater than
that of advanced battery systems, as shown in Figure
25. However, due to the slow kinetics, especially at
the oxygen cathode side, the power capability of fuel
cells is lower than that of most batteries and inferior
to combustion engines and gas turbines (cf. Figure Figure 26. Schematic of a polymer electrolyte membrane
3). The detrimental impact of the weight and volume (PEM) fuel cell. The fuel cell stacks operate at 30-180 °C
of the fuel cell may be hidden by giving the “power with 30-60% efficiency. Fuel options include pure hydro-
capability” in A/cm2 and not in W/L or W/kg. gen, methanol, natural gas, and gasoline.
electrolytes, as they are soluble in acidic media. The

3.3. Characteristics of Various Types of Fuel AFC is susceptible to CO2 contamination of the
Cells electrolyte (clogs the pores) when air is used as well
as to poisoning of the Pt and Ni catalyst by sulfide
Fuel cells are typically classified by the type of and CO impurities in the feedstock. In other words,
electrolyte. Apart from certain specialty types, the the AFC requires pure hydrogen and oxygen as
five major types of fuel cells are alkaline fuel cell reactants. Because of the requirement of pure fuels
(AFC), polymer electrolyte fuel cell (PEMFC), phos- and elimination of CO2 for long life, terrestrial
phoric acid fuel cell (PAFC), molten carbonate fuel applications are limited. Both electrodes are fabri-
cell (MCFC), and solid oxide fuel cell (SOFC). cated with an active layer of platinum catalyst on
The AFC is one of the oldest fuel cell types. The carbon support and binder that is backed by a wet-
cell reactions are as follows (the existence of the proofed Teflon [PTFE, poly(tetrafluoroethylene)]
peroxide intermediate HO2- has been already dis- bonded carbon layer to control the wetting of the
cussed): electrodes by the electrolyte and thus the location of
the three-phase boundary. The carbons are generally
anode: H2 + 2OH- - 2e- ) 2H2O (23) treated to remove active entities on the surface before
being catalyzed. Graphite bipolar endplates contain
cathode: O2 + H2O + 2e- ) HO2- + OH- (24) the gas flow channels and serve to provide waste heat
removal. On shutdown, it is common to have the
HO2- + H2O + 2e- ) 3OH- (25) electrolyte empty into a sump on earthbound ap-
plications, as the cathode needs to be protected from
or: O2 + 2H2O + 4e- ) 4OH- (26) both absorption of CO2 by the electrolyte and corro-
sion by wet O2. The AFC operates at up to 1 A/cm2
The AFC was first developed for the Apollo mis- at 0.7 V.
sions. An updated version has been developed and is The PEMFC was first developed for the Gemini
still in use to provide electrical power for shuttle space vehicle. The electrodes are formed on a thin
missions. These power plants reach efficiencies of layer on each side of a proton-conducting polymer
60% in space applications. The electrolyte is potas- membrane, used as electrolyte. In a sense, the
sium hydroxide based. Noble metal catalysts are very electrolyte is composed similarly to plasticized elec-
active in the AFC for both the hydrogen and oxygen trolyte in a Li ion cell, where a liquid electrolyte
electrodes. The hydrogen and oxygen kinetics are component is immobilized in a polymer matrix. It
more facile in alkaline than acid electrolytes, thus consists of a solid polymer PTFE backbone with a
resulting in higher cell voltages.6 This permits the perfluorinated side chain that is terminated with a
use of non-noble metal catalysts, such as Raney sulfonic acid group. Hydration of the membrane
nickel, for the fuel electrode. Silver and spinel-type yields dissociation and solvation of the proton of the
oxides along with iron phthalocyanines and other acid group. The solvated protons are mobile within
porphyrins are good catalysts for the oxygen (air) the polymer and provide electrolyte conductivity. The
electrodes. These catalysts cannot be used in acidic membrane has low permeability to oxygen and
hydrogen (crossover) for high coulombic efficiency. A
5 A fuel stack is a series connection of fuel cells. Strictly speaking, typical PEMFC is depicted in Figure 26.
it is fuel cell battery (cf. footnote 1). The composition and design of A graphite (or metal) plate serves as the plenum
the fuel cell stack differ for the implementation of each type of cell. for the gas supply and for heat removal. A catalyzed
6 The higher voltage is not only due to better kinetics but also due
to the fact that oxygen reaction via the intermediate peroxide (HO2- carbon layer is applied to the membrane surface. A
in alkaline electrolytes) is more facile. thicker gas diffusion layer or porous carbon paper
provides gas transport to the reaction zone. The from the anode side to the cathode side of the cell.
composition and amount of catalyst differ for each On reaching the cathode, the methanol is oxidized.
electrode. The anode has a lower catalyst loading This significantly lowers the cathode voltage and the
than the cathode. The platinum-based catalysts are overall efficiency of cell operation. The typical DMFC
sensitive to H2S and CO impurities among others so yields ∼0.5 V at 400 mA/cm2 at 60 °C.
they must be eliminated from the feedstock for long The PAFC is another fuel cell operating in acidic
operation. The catalyst content of the anode is ∼0.1 media. It has been frequently used in energy storage
mg/cm2, and that of the cathode is ∼0.5 mg/cm2. applications. The cell operates at ∼200 °C. Below 150
Water management in the membrane is critical to °C, its conductivity is reduced, and above 220 °C, the
long-term performance. The proton transport carries phosphoric acid is too volatile and tends to decom-
water along (water drag) while carrying the current. pose. A SiC matrix separator holds the acid. The acid
The water concentration gradient results in back- rejects the water reaction product. Above 150 °C,
diffusion. However, for operation at high current, the some CO can be tolerated in the anode feedstock. The
anode side of the membrane must be humidified or kinetic hindrance at the oxygen cathode is the major
it will dehydrate. By the same token, product water source of losses. The active layer of platinum catalyst
is removed from the cathode side in the air stream on carbon black support and polymer binder is
to prevent flooding of the active layer. For reliable backed by a carbon paper with 90% porosity, reduced
operation, a water content of 30-60% in the mem- somewhat by PTFE binder. The active layer consists
brane is preferred for reliable operation. of pores in the range of 3-50 µm with micropores in
For higher temperature operation, a polybenzimid- the range of 0.0035 µm. The anode operates at nearly
azole-based polymer electrolyte may be preferred. reversible voltage with ∼0.1 mg/cm2 catalyst loading.
The PEMFC structures have good mechanical integ- The cathode requires a higher catalyst loading of ∼1
rity under compression and expansion from dif- mg/cm2 of catalyst. Graphite bipolar endplates con-
ferential temperature and pressure gradients that tain the gas flow channels and serve to provide heat
occur during operation. This system has minimal removal by liquid flow. On shutdown, the cathode
materials problems, except for the cost and operation needs to be protected from corrosion by wet O2. The
characteristics of the membrane. The PEMFC oper- electrode reactions are
ates at ∼1 A/cm2 at 0.7 V. The electrode reactions in
acidic media have been discussed above.
anode: H2 - 2e- ) 2H+ (30)
The DMFC uses the same basic cell construction
as for the PEMFC. It has the advantage of a liquid cathode: O2 + 4H+ + 4e- ) 2H2O (31)
fuel in that is easy to store and transport. There is
no need for the re-former to convert the hydrocarbon
fuel into hydrogen gas. The anode feedstock is a The hot H3PO4 electrolyte rejects water, the reaction
methanol and water mixture or neat methanol, product. The high temperature favors H2O2 decom-
depending on cell configuration. The DMFC is under position, and peroxide buildup is less pronounced
development as a power source for portable electronic than for the aqueous electrolyte systems.
devices such as notebook computers and cellular The MCFC finds application in energy storage
phones. The pure methanol or a methanol-water applications. It operates best at ∼560 °C, and the
mixture would be stored in a cartridge similar to that waste heat can be used in cogeneration. The system
used for fountain pens. Refueling would involve the does not use any noble metal catalysts and has a
quick replacement of the cartridge. The platinum- higher efficiency than the PEMFC and the PAFC.
ruthenium catalyst loadings for the anode are higher The separator is a LiAlO2 ceramic tile separator filled
than for the PEMFC and are in the range of 1-3 mg/ with molten carbonates to prevent crossover of the
cm2. Without the ruthenium, that is, with neat Pt, reactants and aid in CO3-2 transport. Lithium-rich
the anode reaction produces a stable formic acid carbonate electrolytes have higher conductivity but
intermediate. The reaction for the direct conversion lower gas solubility and higher corrosion rates. At
of methanol has a similar voltage as for hydrogen. the operating temperature, problem areas include
The overall cell reaction is NiO dissolution and structural stability of the anodes
and cathodes, changes in pore size distribution, and
CH3OH + 3/2O2 ) CO2 + 2H2O E ) 1.186 V distortion of the electrode structures. Alternate LiFeO2
(27) cathode materials have low conductivity, and Ni-
Cr anode materials have creep and stability issues.
The reaction at the anode is Control of the pore diameter is critical in the separa-
tor tiles. Increased pressure raises voltage but in-
anode: CH3OH + H2O - 6e- ) 6H+ + CO2 (28) creases gas solubility and cathode dissolution and
lowers operating life. The performance is limited by
the electrolyte resistivity and the removal of the heat
and generated by the losses in electrode polarization. The
cells are sensitive to sulfur contamination because
cathode: O2 + 4H+ + 4e- ) 2 H2O (29) of its effect on the nickel electrode materials and
catalyst. Other problem areas include seal stability
Due to the chemical similarity of water and metha- during thermal cycling and electrolyte creep. Typical
nol, the methanol has considerable solubility in the operating parameters are 150 mA/cm2 at 0.8 V at 600
polymer membrane, leading to significant crossover °C. The anode reactions using the fuel methane (or
CO and H2 stemming from CH4 conversion) are

anode: CH4 + 2H2O ) CO2 + 4H2 or (32)
CH4 + H2O ) CO + 3H2 (33)
then: CO + H2O ) CO2 + H2 (34)
then: H2 + CO32- - 2e- ) CO2 + H2O

(principal reaction) or (35)
CO + CO32- ) 2CO2
(minor reaction) (36)
cathode: O2 + 2CO2 + 4e- ) 2CO32- (37)

The SOFC operates at ∼800-1000 °C with O2-
conduction in the solid phase. Limitations and prob-
lems arise from the high operating temperatures and
place severe restrictions on the choice of materials.
The materials must have very similar coefficients of
expansion and be chemically stable in oxidizing and
reducing conditions. At the operating temperatures,
the hydrocarbon fuels (plus water, if needed) are
quickly and completely re-formed in situ. The system
operates at close to 96% thermodynamic efficiency
and is tolerant to most impurities. It can deliver high-
quality heat for cogeneration applications. The SOFC
can operate on most any hydrocarbon or hydrogen
fuel and is tolerant of short circuiting and to over-
loads. No noble metal catalysts are required. The
anode consists of a porous cermet of Ni or Co catalyst Figure 27. Schematic view of SOFC cylindrical and flat
plate cell constructions.
on yttria-stabilized zirconia. The zirconia acts to
inhibit grain growth of the catalyst particles of nickel
or cobalt and protects against thermal expansion. The an energy storage device based on the charge storage
electrolyte itself consists of yttria-stabilized zirconia, in the electrical double layer (EDL) of a high-surface-
which can be additionally deposited onto a calcia- area carbon in aqueous electrolytes. The market
stabilized zirconia. Yttria and calcia doping provide for these devices used for memory protection in
oxygen defects for better conductivity. The cathode electronic circuitry is about $150-$200 million an-
is generally a strontia-doped lanthanum-manganite nually. New potential applications for ECs include
perovskite. The Sr dopant provides for oxygen trans- the portable electronic device market, the power
fer to the cathode-electrolyte interface. A Mg- or Sr- quality market, due particularly to distributed gen-
doped lanthanum chromate is used for the current eration, and low-emission hybrid cars, buses, and
collector and the intercell connection. It is impervious trucks.
to the fuel and oxygen gases and is chemically and There is no commonly accepted nomenclature for
structurally stable in thin, dense, layered configura- ECs except that the definitions of anode, cathode,
tions. The cells operate at ∼1 A/cm2 at 0.7 V. Cells etc., carry over from battery and fuel cell usage, such
are constructed in cylindrical form or in the flat plate as the anode as the negative terminal and cathode
format shown in Figure 27. The cell reactions are as the positive terminal. In general, ultracapacitors
have referred to capacitors with two high-surface-
anode: H2 + O2- - 2e- ) H2O or (38) area carbon electrodes for the anode and cathode.
This arrangement where both electrodes have the
CO + O-2 - 2e- ) CO2 or (39) same configuration will be referred to as a symmetric
capacitor. The term supercapacitor has also been
CH4 + 4O-2 - 8e- ) 2H2O + CO2 (40) used to refer to the symmetric combination of two
carbon electrodes that are catalyzed with ruthenium
cathode: O2 + 4e- ) 2O-2 (41) dioxide (RuO2). The RuO2 introduces a redox couple
between two valence states of ruthenium, to result
4. Electrochemical Capacitors (ECs) in higher capacitance for the carbon electrodes, but
with a penalty in slower time constant for reacting
4.1. Introduction and Market Aspects to a pulse demand. A second type of EC combines a
battery or redox electrode with a carbon electrode,
ECs are sometimes called supercapacitors, ultra- such as nickel hydroxide cathode with a carbon
capacitors, or hybrid capacitors. The term ultra- anode. These supercapacitor or hybrid capacitors will
capacitor or supercapacitor is usually used to describe be referred to as an asymmetric EC.
reactivity with the surface. The outer Helmholtz

plane (OHP) refers to the distance of closest approach
of nonspecifically adsorbed ions (generally cations)
in solution. Cations that populate the OHP are
usually solvated and thus are generally larger than
the less solvated anions. The interactions of the ions
of the OHP with the surface charge have the char-
acter of coulombic interactions. The inner Helmholtz
plane (IHP) refers to the distance of closest approach
of specifically adsorbed ions (generally anions) and/
or adsorbed solvent molecules to the electrode sur-
face. These adsorption processes are determined by
chemical affinities of the ions to the electrode surface
and the field strength in the EDL. In practice, the
structure of the EDL is much more complex than the
models discussed above.
The double layer at the electrode surface forms and
relaxes almost instantaneously. It has a time con-
stant, or time of formation, of ∼10-8 s. Therefore, the
structure of the double layer has the capability to
respond rapidly to potential changes in the same time
frame. The process involves only a charge rearrange-
ment, not a chemical reaction. This rapid response
to change is in contrast to the situation with the
redox electrode reactions in batteries and fuel cells.
The time constant for the redox reactions is much
slower and in the range of 10-2-10-4 s related to the
Figure 28. (A, top) Simple Helmholtz model of the impedance of the reaction. The redox reactions
electrical double layer. It is essentially a picture of a
conventional capacitor. (B, bottom) Depiction of the electri-
contribute to “polarization capacitance” associated
cal double layer at the surface of the negative electrode with the electrode reactions. The other main differ-
showing the outer Helmholtz plane (OHP) and the inner ence between supercapacitors and batteries and fuel
Helmholtz plane (IHP). The inner Helmholtz plane (IHP) cells is the reversibility (short time constant) of the
refers to the distance of closest approach of specifically EDL process compared to the longer time constant
adsorbed ions and solvent molecules to the electrode of the redox reactions and the stress from detrimental
surface. The outer Helmholtz plane (OHP) refers to the side reactions, which reduce the cycle life and long-
distance of ions, which are oriented at the interface by
coulomb forces. term stability of the device. Whereas cycle life and
stability of the double layer electrochemical capacitor
can easily exceed 1 million cycles, battery electrodes
4.2. Characteristics of the Electrical Double Layer can reach this level only if charged and discharged
at a low depth of discharge.
When an electrode, that is, an electronic conductor, When carbons are placed in an electrolyte, they
is immersed into an electrolyte solution, that is, an generally assume a voltage near the zero point of
ionic conductor, there is a spontaneous organization charge of the EDL. In aqueous solutions, this is near
of charges at the surface of the electrode and in the 0 V versus hydrogen. By applying an external volt-
electrolyte facing the electrode. This EDL forms at age, many more additional ions and electrons can be
the electrode-electrolyte interface with one layer at moved to the double layer, increasing the capacitance
the surface inside the conductor and the other layer C () charge per applied voltage, C ) Q/U). As a rule
in the electrolyte as depicted in Figure 28. The two of thumb, carbons and metals typically have a double-
charged layers are considered to behave as a physical layer capacitance in the range of 10-40 µF/cm2. The
capacitor, with the charges in the solution and in the exact values depend mainly on the voltage and the
conductor separated by a distance of the order of extent of participation of the IHP in the electrical
molecular dimensions. The characteristics of the EDL double layer. A high-surface-area carbon electrode
depend on the electrode surface structure, the com- can yield a capacitance of ∼4 F/g.
position of the electrolyte, and the potential field
between the charges at the interface. Depending on 4.3. EC Operation
the surface charge of the electrode materials, positive
or negative ions from the electrolyte form the solution As noted above, electrochemical capacitors are close
part of the EDL at the interface between the elec- cousins to batteries. The simple circuit shown il-
trode and the electrolyte. A simplified structure is lustrates their basic operation.
shown in Figure 28A for the case of a negatively
charged electrode surface. According to this simple
Helmholtz model, the charges are concentrated on
each side of the electrode surface. A more complex
model of the EDL structure in Figure 28B takes into Here, Ca and Cc are the double-layer capacitances of
account the different sizes of the ions and their the anode and cathode, respectively. Ri is the internal
Figure 29. Depiction of the charging process of a symmetric capacitor.
resistance of the cell. For capacitors in series

1/C ) 1/Ca + 1/Cc (42)
If Ca ) Cc, as would be expected for an ultracapacitor,

then
C ) 1/2Ca (43)
High-surface-area carbon is the material of choice,

as it combines a large surface area wetted by the
electrolyte, high electronic conductivity, and chemical Figure 30. Operation principle of an EC in the discharged
and electrochemical stabilities with low cost. The state, during charging, and in the charged state: (A) for a
symmetric construction and (B) for an asymmetric con-
capacitance of these devices can be orders of magni- struction.
tude larger than those of conventional dry and
electrolytic capacitors. The voltage for electrochemi-
cal capacitors with aqueous electrolytes is ∼1 V, battery reaction of ∼10 times the capacitance of the
limited by the voltage stability of the electrolyte. By electrical double layer. For instance, if the nickel
switching to an organic-based electrolyte, voltages of battery cathode is substituted for the cathode in a
up to 2.7 V can be found in practice. However, the symmetric capacitor, for example, NiOOH for carbon,
organic electrolytes have lower double-layer capaci- then substituting Cc ) 10Ca in eq 42, the capacitance
tance and poorer conductivity. Because the energy of the EC is essentially doubled.
storage is given by energy ) 1/2QV2, the higher
voltage permitted by an organic electrolyte signifi- 1/C ) 1/Ca + 1/10Ca (44)
cantly increases the energy storage capability of the
EC. Because the resistivity is ∼100 times larger than C ≈ Ca (45)
for aqueous electrolytes, the time constant for re-
sponse to a large pulse is slower for the nonaqueous This asymmetric type of EC is often termed a
electrolyte-based ECs. “hybrid” capacitor. The typical discharge curve for
The charge-discharge of a symmetric EC com- this hybrid EC is shown in Figure 30. Because the
posed of two carbon electrodes with approximately battery electrode has a capacity of 3-10 times that
the same mass immersed in an aqueous or nonaque- of the double-layer electrode, it remains at an invari-
ous electrolyte is shown in Figure 29. With zero ant voltage during charge and discharge. As a result,
applied charge Q, both electrodes of a cell are at the the discharge voltage of the hybrid capacitor falls
same voltage. The potential of the electrodes in- more slowly than that of the carbon-carbon EDL
creases in opposite directions during charge, as each capacitor.
has approximately the same capacitance. Maximum In some cases, the kinetics of the redox charge-
cell operating voltage is reached when one of the discharge reactions can proceed almost as quickly
electrodes reaches the stability limit of the electro- and reversibly as EDL charging. Thin film redox
lyte. electrodes, based on the lithium intercalation/inser-
The asymmetric type of EC incorporates a battery tion principle such as Li4Ti5O12, exhibit high revers-
electrode as one of the electrodes. The battery elec- ibility and fast kinetics. The RuO2 materials depos-
trode has a capacitance associated with the redox ited on carbon show “pseudo-capacitive” charge-
discharge behavior as do polymeric materials such batteries, fuel cells nor electrochemical capacitors, by
as polyaniline, polypyrrole, and polydiaminoanthra- themselves, can serve all applications.
quinone (DAAQ). These have facile kinetics and have
shown high capacitance and long life. The insertion
of anions and cations into their structure can yield Martin Winter
capacitances of up to 200 µF/cm2 and, moreover, they Institute for Chemistry and Technology of
can be easily fabricated as thin films. Inorganic Materials, Graz University of
Technology, Stremayrgasse 16,
5. Summary A-8010 Graz, Austria
Electrochemical energy conversion devices are
pervasive in our daily lives. Batteries, fuel cells and
supercapacitors belong to the same family of energy Ralph J. Brodd*
conversion devices. They are all based on the funda- Broddarp of Nevada, Inc.,
mentals of electrochemical thermodynamics and ki- 2161 Fountain Springs Drive,
netics. All three are needed to service the wide energy Henderson, Nevada 89074
requirements of various devices and systems. Neither CR020730K
Chem. Rev. 2004, 104, 4271−4301 4271
Lithium Batteries and Cathode Materials
M. Stanley Whittingham*
Department of Chemistry and Materials Science, State University of New York, Binghamton, New York 13902-6000
Received June 16, 2004
Contents 6. Conclusions and What Does the Future Hold 4297

7. Abbreviations and Specialized Terms 4297
8. Acknowledgments 4297
2. Origins of the Lithium Battery 4273
9. References 4297
2.1. Early Concepts 4273
2.2. Molten Salt Systems at Argonne National 4273
Laboratory and General Motors
2.3. Concept of Mixed Conductors 4273 1. Introduction
2.4. Early Intercalation Concept 4274 In the previous paper Ralph Brodd and Martin
3. 1972−1980: Birth of the Rechargeable Lithium 4274 Winter described the different kinds of batteries and
Battery fuel cells. In this paper I will describe lithium
3.1. Intercalation in the Layered Dichalcogenides 4274 batteries in more detail, building an overall founda-
3.2. Trichalcogenides and Related Materials 4277 tion for the papers that follow which describe specific
3.3. Movement into Oxides 4277 components in some depth and usually with an
3.3.1. Layered Oxides 4277 emphasis on the materials behavior. The lithium
3.3.2. High-Valent Oxides of Vanadium and 4277 battery industry is undergoing rapid expansion, now
Molybdenum representing the largest segment of the portable
3.3.3. Mixed-Valent Oxides of VanadiumsV6O13 4278 battery industry and dominating the computer, cell
and LiV3O8 phone, and camera power source industry. However,
3.3.4. Double-Sheet Structures: Xerogels, 4278 the present secondary batteries use expensive com-
δ-Vanadium Oxides, and Nanotubes ponents, which are not in sufficient supply to allow
4. 1980−1990: Era of Layered Oxides and First 4279 the industry to grow at the same rate in the next
Large Commercialization decade. Moreover, the safety of the system is ques-
4.1. Early Studies of Layered Oxides 4279 tionable for the large-scale batteries needed for
hybrid electric vehicles (HEV). Another battery need
4.2. Lithium Cobalt Oxide, LiCoO2 4280
is for a high-power system that can be used for power
4.3. Lithium Nickel Oxide, LiNiO2 4281 tools, where only the environmentally hazardous Ni/
5. 1990−Present: Second-Generation Lithium 4282 Cd battery presently meets the requirements.
Batteries
A battery is a transducer that converts chemical
5.1. Spinels 4282 energy into electrical energy and vice versa. It
5.2. Other Layered Oxides 4283 contains an anode, a cathode, and an electrolyte. The
5.2.1. Mixed Nickel−Cobalt Dioxide, 4283 anode, in the case of a lithium battery, is the source
LiNi1-yCoyO2 of lithium ions. The cathode is the sink for the
5.2.2. Lithium Manganese Dioxide, LiMnO2 4283 lithium ions and is chosen to optimize a number of
5.2.3. Mixed Manganese−Cobalt Dioxide, 4284 parameters, discussed below. The electrolyte provides
LiMn1-yCoyO2 for the separation of ionic transport and electronic
5.2.4. Mixed Nickel−Manganese Dioxide, 4285 transport, and in a perfect battery the lithium ion
LiNi1-yMnyO2sMultielectron Redox transport number will be unity in the electrolyte. The
Systems cell potential is determined by the difference between
5.2.5. Mixed Nickel−Manganese−Cobalt Dioxide, 4287 the chemical potential of the lithium in the anode and
LiNi1-y-zMnyCozO2 cathode, ∆G ) -EF.
5.2.6. Lithium-Rich Mixed-Metal Dioxides, 4292 As noted above, the lithium ions flow through the
Li1+xM1-xO2 electrolyte whereas the electrons generated from the
5.3. Iron Compounds Including Oxides and 4293 reaction, Li ) Li+ + e-, go through the external
Phosphates circuit to do work. Thus, the electrode system must
5.3.1. Olivine Phase 4293 allow for the flow of both lithium ions and electrons.
5.3.2. Other Iron Phosphate Phases 4295 That is, it must be both a good ionic conductor and
5.3.3. Vanadium Phosphate Phases 4296 an electronic conductor. As discussed below, many
electrochemically active materials are not good elec-
* To whom correspondence should be addressed. Phone and fax: tronic conductors, so it is necessary to add an
(607) 777-4623. E-mail: stanwhit@binghamton.edu. electronically conductive material such as carbon
10.1021/cr020731c CCC: $48.50 © 2004 American Chemical Society
4272 Chemical Reviews, 2004, Vol. 104, No. 10 Whittingham
example, in cell phones, laptop computers, and MP3

players such as the iPOD. For some larger applica-
tions, such as the battery in hybrid electric vehicles
(HEV), power is most important as the materials
must be able to charge sufficiently fast to take
advantage of regenerative braking; otherwise, much
of the gas savings are lost. The all-electric consumer
electric vehicle is probably dead, at least in the
United States, because of its limited range and the
user’s desire for instant heating and air-conditioning
among other high-power consumption devices. How-
ever, there is still much interest in Asia, particularly
for scooters as well as for all electric vehicles.
The key requirements for a material to be success-
fully used as a cathode in a rechargeable lithium
M. Stanley Whittingham was born in Nottingham, England, and received battery are as follows.
his B.A. and D.Phil. degrees in Chemistry from Oxford University working
with Peter Dickens. In 1968 he went to Professor Robert A. Huggins (1) The material contain a readily reducible/
research group in the Materials Science Department at Stanford University oxidizable ion, for example a transition metal.
as a Postdoctoral Research Associate to study fast-ion transport in solids. (2) The material react with lithium in a revers-
In 1972 he joined Exxon Research and Engineering Company to initiate
a program in alternative energy production and storage. After 16 years in ible manner.
industry he joined the Binghamton campus of the State University of New (a) This dictates an intercalation-type reac-
York as Professor of Chemistry to initiate an academic program in Materials tion in which the host structure essentially
Chemistry. Presently he is also Professor of Materials Science and Director does not change as lithium is added.
of the Materials Science Program and Institute for Materials Research.
He was awarded the Young Author Award of the Electrochemical Society (3) The material react with lithium with a high
in 1971, a JSPS Fellowship in the Physics Department of the University free energy of reaction.
of Tokyo in 1993, and the Battery Research Award of the Electrochemical (a) High capacity, preferably at least one
Society in 2002 and was elected a Fellow of the Electrochemical Society lithium per transition metal.
in 2004. He was Principal Editor of the journal Solid State Ionics for 20 (b) High voltage, preferably around 4 V (as
years. His recent work focuses on the synthesis and characterization of limited by stability of electrolyte).
novel microporous and nano-oxides and phosphates for possible
electrochemical and sensor applications.
(c) This leads to a high-energy storage.
(4) The material react with lithium very rapidly
both on insertion and removal.
black. To physically hold the electrode together, a
(a) This leads to high power density, which
binder is also added. In these cases the electrochemi-
is needed to replace the Ni/Cd battery or for
cal reaction can only occur at those points where the
batteries that can be recharged using HEV
active material, the conductive diluent, and electro-
regenerative braking.
lyte meet. Thus, most electrodes are complex porous
(5) The material be a good electronic conductor,
composites.
preferably a metal.
This review will be mainly concerned with the (a) This allows for the easy addition or re-
cathode. The anode, the source of lithium, is normally moval of electrons during the electrochemi-
a graphitic carbon and will be discussed in detail by cal reaction.
Rachid Yazami in a future issue. The electrolyte (b) This allows for reaction at all contact
solution commonly comprises a lithium salt dissolved points between the cathode active material
in a mixture of organic solvents, examples include and the electrolyte rather than at ternary
LiPF6 or LiBOB (the BOB is the anion with the boron contact points between the cathode active
coordinated by two oxalate groups) in an ethylene material, the electrolyte, and the electronic
carbonate/dimethyl carbonate solvent; Kang Xu dis- conductor (such as carbon black).
cusses electrolytes later in this issue. Although (c) This minimizes the need for inactive
organic polymers containing lithium ions have been conductive diluents, which take away from
studied for several decades as possible electrolytes, the overall energy density.
their conductivity is still too low, and so they are only (6) The material be stable, i.e., not change
used when a liquid is added to give a plasticized structure or otherwise degrade, to overdis-
state. In a few cases, solid electrolytes have been used charge and overcharge.
for a few specialized applications such as for oil well (7) The material be low cost.
logging where elevated temperatures are found. (8) The material be environmentally benign.
Though the ions flow through the electrolyte, the Almost all of the research and commercialization
anode and cathode must be physically separated to of cathode materials has centered on two classes of
prevent an electrical short. This is accomplished by materials. The first contains layered compounds with
using a porous separator material, which allows an anion close-packed or almost close-packed lattice
wetting by the electrolyte and the flow of lithium ions in which alternate layers between the anion sheets
through it. P. Aurora discusses these materials in are occupied by a redox-active transition metal and
detail. lithium then inserts itself into the essentially empty
For most consumer devices, energy storage is key: remaining layers. This structure is depicted in Figure
operating time is key, so the more the better, as, for 1. This group is exemplified by first LiTiS2, followed
Lithium Batteries and Cathode Materials Chemical Reviews, 2004, Vol. 104, No. 10 4273
lithium iodine cell. The majority of the implantable

cardiac defribrillators use in the last 20 years have
used, as active cathode material, silver vanadium
oxide, Ag2V4O11.7-9 During discharge the silver is
reduced to silver metal, and in addition, more than
one lithium per vanadium can be reacted, giving it a
capacity over 300 mAh/g. The presence of metallic
silver greatly improves the electronic conductivity
and thus the rate capability. Future medical devices,
such as heart-assist devices, will require rechargeable
systems because the capacity of primary cells cannot
provide the power needed for active medical devices.
The copper analogue also reacts by exuding the
metal.
2.2. Molten Salt Systems at Argonne National

Laboratory and General Motors
Figure 1. Layered structure of LiTiS2, LiVSe2, LiCoO2, Most of the early work10-12 on lithium rechargeable
LiNiO2, and LiNiyMnyCo1-2yO2, showing the lithium ions
between the transition-metal oxide/sulfide sheets. The batteries used a molten salt as electrolyte and oper-
actual stacking of the metal oxide sheets depends on the ated at around 450 °C. The earliest cells used molten
transition metal and the anion. lithium and molten sulfur as the two electrodes, but
the problems with electrode containment proved
by LiCoO2, LiNi1-yCoyO2, and today LiNiyMnyCo1-2yO2. insurmountable. The anode used in the final versions
The spinels may be considered as a special case was the lithium aluminum alloy, LiAl, and iron
where the transition-metal cations are ordered in all sulfides, such as FeS and FeS2, which replaced the
the layers. The materials in the second group have molten sulfur of the early designs. Development
more open structures, like many of the vanadium ceased around 1990 when corrosion, temperature,
oxides, the tunnel compounds of manganese dioxide, and other issues overwhelmed the advantages of the
and most recently the transition-metal phosphates, system, the sodium sulfur system appeared more
such as the olivine LiFePO4. The first group, because promising, and early results on ambient lithium
of their more compact lattices, will have an inherent rechargeable systems began to show promise. These
advantage in energy stored per unit of volume, but low-cost iron sulfides were also considered in the
some in the second group, such as LiFePO4, are 1970s in ambient temperature cells. Iron pyrite was
potentially much lower cost. The following discussion found13 to react with an initial constant potential of
will center predominantly on these two classes of 1.5 V, allowing it to replace the more expensive Ag-
materials. Zn button cells and later making it a drop-in replace-
ment for the Zn/MnO2 alkaline cell. It is presently
2. Origins of the Lithium Battery marketed by Eveready as a primary high-energy cell.
On recharge the structure changes and the subse-
2.1. Early Concepts quent discharge has a two-step profile with an initial
discharge of around 2 V.13
Some of the earliest concepts came from Japan, It is still the dream of battery researchers to
where Matsuchita developed1 the Li/(CF)n battery develop a cell based on the lithium/sulfur couple
that was used, for example, in fishing floats. Lithium which on paper has a simple chemistry, has a much
fluoride and carbon are the final reaction products, higher energy density than most of the cathode
but the cell potential of 2.8-3.0 V suggests a different materials to be discussed below, and should be
electrochemical reaction. It was proposed that lithium capable of high rates if the sulfur is in solution.
initially intercalates the carbon monofluoride lattice Recent work14 has achieved high capacities at 2 V
and subsequently the lithium fluoride is formed:2 Li even at as low temperatures as -40 °C in electrolyte
+ (CF)n f Lix(CF)n f C + LiF. Although much work solvents of dioxolane and dimethoxyethane. These
has continued sporadically on the carbon fluorides, cells with their liquid polysufide cathode have gener-
by Exxon and others, the major challenge has been ated specific power exceeding 750 W/kg at 25 °C.
to make this reaction reversible even when lower However, such cells still have significant issues with
fluoride levels were used. Sanyo, the largest manu- self-discharge on standing, lithium recharging, and
facturer today of both lithium rechargeable and the highly resistive nature of the cathode. Whether
nickel metal hydride batteries, developed3,4 one of the all these issues will be overcome is still much
earliest lithium batteries with the Li/MnO2 system debated. A recent concept15 to protect the lithium
that they initially sold in solar rechargeable calcula- anode from the reactive polysulfide medium is to coat
tors.5 Some early work on ambient systems was also it with a single-ion conducting glass.
underway in the United States by 1970, for example,
by Dey et al.6 on the lithium reactivity with a range 2.3. Concept of Mixed Conductors
of metals, such as aluminum.
Many primary lithium batteries have been devel- In 1967 Yao and Kummer reported16 the remark-
oped for use in the medical field starting with the able electrolytic behavior of the β-alumina class of
materials, Na1+xAl11O17, which was proposed as the highest superconducting transition temperature, re-
electrolyte for a battery having a molten sodium sulted in the discovery of the very high free energy
anode and a molten sulfur cathode operating around of reaction of the alkali ions with these layered
300 °C. The measurement of the ionic conductivity materials. Thus, the stability of the hydrates, Kx(H2O)-
of these materials required a new approach, as their TaS2, could be explained30 as being due to their salt-
high ionic conductivity required the use of ionically like behavior, in contrast to the metallic-like behavior
reversible electrodes and the obvious electrode so- of the corresponding compounds of graphite, C8K. It
dium was too difficult to handle. To overcome this was also found that such intercalation reactions could
problem, the nonstoichiometric oxide bronzes of also be accomplished in an electrochemical cell either
tungsten and vanadium were used,17-19 for example, by electrolyzing dissolved salts or by inserting ions
NaxWO3 and LixV2O5; these bronzes have a wide from the anode.31-33
range of composition, are metallic conductors, and
Of all the layered dichalcogenides, titanium disul-
thus could readily serve as electrodes reversible to
fide was the most appealing for consideration as an
both alkali ions and electrons. Their mixed conduc-
tivity also led to the proposed use of the vanadium energy storage electrode:34-38 it was the lightest. It
oxide as electrodes for sodium or lithium batteries: was subsequently discovered that it was a semi-
Li-LixV2O5. metal,39 so no conductive diluent was needed in the
cathode structure, and any such addition was found
to be detrimental to the electrochemical behavior. It
2.4. Early Intercalation Concept was also found to form a single phase with lithium
Around 1970 two groups began studying the idea over the entire composition range of LixTiS2 for 0 e
of placing electrochemically active species inside an x e 1.40 This lack of phase change enables all the
electrically conductive host. At Bell Laboratories, lithium to be removed reversibly, without the need
Broadhead et al.20,21 conceived of intercalating iodine for energy wastage associated with the nucleation of
or sulfur between the layers of a dichalcogenide host new phases or the sluggish reactions when massive
material such as niobium diselenide. They believed rearrangement of the host must occur as the lithium
that the dichalcogenide host material itself was content is changed. This behavior may be contrasted
electrochemically inert.22 Such materials showed good with that observed in LiCoO2, to be discussed below,
cyclability at low depths of discharge. Subsequently, where phase changes result in only about one-half
they found that when selenium reacted with NbSe2, of the lithium being readily cycled in to and out of
it formed the triselenide NbSe3 which exhibited very the compound; today the capacity has been improved
good electrochemical behavior.23 At Stanford, Armand but is still well below the goal of one lithium per
et al.24-26 tried to incorporate oxides such as CrO3, transition-metal ion. Some specifics of the lithium
and subsequently halides, between the layers of titanium disulfide cell will now be described, as they
graphite. Again, electrochemical activity was found. are typical of what is desired in advanced lithium
However, subsequent work showed that in both these batteries. Although they will not be discussed here,
cases no intercalation had occurred and that the there was also interest41-44 in using the disulfides as
electrodes were mixtures of the guest and host the cathodes of sodium batteries; such systems are
materials. The large amount of host material re- seriously complicated by phase changes as the so-
quired also made this concept infeasible due to the dium content changes in NaxTiS2 or NaxTaS2 due to
resulting poor energy density. A similar concept,27 the sodium favoring trigonal prismatic coordination
still not tested, was to use as electrodes two known at intermediate x values and octahedral coordination
graphite materials C8K and C8Br where on discharge as x approaches one.
the potassium and bromine would react in a suitable
solvent and on charge would be intercalated into their Titanium disulfide has a hexagonal close-packed
respective graphite electrodes. sulfur lattice with the titanium ions in octahedral
sites between alternating sulfur sheets, as shown in
Figure 1. The TiS2 sheets are stacked directly on top
3. 1972−1980: Birth of the Rechargeable Lithium of one another, giving the sulfur anion stacking
Battery sequence ABAB. For nonstoichiometric sulfide Ti1+yS2
or for TiS2 prepared at high temperatures, some of
3.1. Intercalation in the Layered Dichalcogenides the titanium is found in the empty van der Waals
Around 1970 researchers at Stanford28 discovered layer. These disordered titanium ions prevent the
that a range of electron-donating molecules and ions intercalation of large molecular ions and impede the
could be intercalated into the layered dichalco- intercalation of even small ions such as lithium by
genides, in particular, tantalum disulfide, TaS2. pinning the TiS2 sheets together,45 thus reducing
These guest-host intercalation reactions modified their diffusion coefficients. Thus, material with the
the physical properties and, in particular, were found highest reactivity for lithium should have an ordered
to enhance the superconducting transition temper- structure, which dictates that it be prepared at
ature from 0.8 to over 3 K. It was also discovered that temperatures below around 600 °C.46-48 We will see
such compounds remained superconducting even that this is also important for the similarly structured
when the guest molecules were paramagnetic.29 layered oxides, discussed below, and even for tunnel
These studies continued at Exxon, where an inves- structures such as LiFePO4, where errant Fe ions in
tigation of the formation of the hydrated alkali-metal the Li sites can reduce reactivity49,50 and diffusion of
intercalates of tantalum disulfide, which showed the lithium.51
Figure 2. Discharge/charge curve of Li/TiS2 at 10 mA/

cm2 (reprinted with permission from ref 13, copyright 1978
Elsevier).
A typical insertion/removal galvanostatic cycle for

lithium in titanium disulfide is shown in Figure 2,
where the current used is 10 mA/cm2. This curve
shows the behavior typical of a single phase all the
way from TiS2 to LiTiS2, so that no energy is
expended in nucleating a new phase. However, closer
examination of the intercalation potential curve,
using the incremental capacity method for the first
time, showed localized ordering of the lithium ions.52
The electrolyte used was 2.5 M lithium perchlorate
in dioxolane, which was found to be an exceptional
electrolyte for effective lithium plating; this solvent
also does not co-intercalate with the lithium into the
sulfide. In contrast, when propylene carbonate was Figure 3. Large cells of LiTiS2 constructed for the Chicago
used as the solvent, any trace amount of moisture automotive show in 1977; vents were used for the LiB(CH3)4
resulted in the co-intercalation of the propylene salt in dioxolane.
carbonate with a concomitant large expansion of the
havior as shown in Figure 4. Initially, VSe2 is in
lattice.45 Such co-intercalation of solvent also pre-
equilibrium with LixVSe2, where x ≈ 0.25, then
vented the use of graphite as an anode for lithium
LixVSe2 is in equilibrium with LiVSe2, and finally
for many years, until appropriate nonintercalating
LiVSe2 is in equilibrium with Li2VSe2. The initial
solvents were found,53,54 initially dioxolane and then
two-phase behavior may be associated with the
a mixture of carbonate solvents. Dioxolane was also
unusual c/a ratio, which is presumably caused by the
found to be an effective electrolyte solvent for cells
desire of group VB elements to have trigonal pris-
using niobium triselenide.55,56 However, the electro-
matic coordination with sulfur and selenium, but in
lyte of LiClO4 in dioxolane is inherently unsafe.57
VSe2 the vanadium is octahedral (this c/a is almost
This clean electrolyte system allows these intercala-
that of TP coordination). When lithium is interca-
tion reactions to be readily followed by in-situ X-ray
lated, the structure becomes typically octahedral with
diffraction58 and in the optical microscope, which
a standard c/a ratio. The ready reversibility of lithium
show the microscopic details of the intercalation
in VSe2, even at 10 mA/cm2, shows that single-phase
process.59
behavior is not critical to effective use as a battery
The ready reversibility of lithium in titanium cathode; however, the phases formed only differ by
disulfide has permitted deep cycling for close to 1000 slight deformations of the octahedra, not wholesale
cycles with minimal capacity loss, less than 0.05% movement of anion sheets as in the conversion of
per cycle, with excess lithium anodes. Exxon mar- LixCoO2 to CoO2.
keted button cells with LiAl anodes60 and TiS2 Vanadium diselenide also showed the feasibility of
cathodes for watches and other small devices in intercalating a second lithium into the lattice. The
1977-1979; the LiAl anode improved the safety of LiVSe2/Li2VSe2 system must be two phase, as the
the cells. Some of the largest lithium single cells built lithium in LiVSe2 is in octahedral coordination
to date are those exhibited by Exxon at the Electric whereas in Li2VSe2 the lithium must move to tetra-
Vehicle Show in Chicago in 1977, shown in Figure hedral coordination and both sites cannot be occupied
3. at the same time. This two-lithium intercalation61,62
Most of the other dichalcogenides are also electro- can be accomplished either electrochemically or
chemically active,13 and many show a similar single- chemically, for example, by using butyllithium. Other
phase behavior with lithium intercalation. Vanadium dilithium layered materials such as Li2NiO2 have also
diselenide is an exception,61 showing two-phase be- been formed63 both electrochemically and chemically
Figure 5. Synthesis approaches for titanium disulfide,

after ref 46.
synthesized because the sulfides VS2 and CrS2 are

thermodynamically unstable at the usual tempera-
tures used for synthesis. Murphy et al.66,67 showed
that these compounds could be formed by the deinter-
calation of the lithium at ambient temperatures; this
work has led to a new route for the synthesis of
metastable compoundssthe deintercalation of stable
phases.
The metastable spinel form of TiS2, which has cubic
Figure 4. Electrochemical insertion of lithium into vana-
dium diselenide, showing reaction of two lithium (reprinted close-packing of the sulfide ions, was similarly formed
with permission from ref 61, copyright 1978 Elsevier):61 by the deintercalation of copper from CuTi2S4.68-70
(top) open-circuit potentials, (bottom) behavior on cycling This cubic structure can also be reversibly interca-
at 2 mA/cm2. lated with lithium, although the diffusion coefficient
is not as high as in the layered form.
using lithium benzophenone in tetrahydrofuran; it’s When a battery is being commercialized, the syn-
structure switches from the 3R-LiNiO2 phase to one thesis process used in the laboratory often cannot be
identical to those of Li2TiS2 and Li2VSe2, where the used because of the processing costs or cost of the
lithium ions sit in all the tetrahedral sites between starting materials. As an example, titanium disulfide
the NiO2 sheets forming a 1T structure. In a similar will be considered where in the laboratory bulk
manner, Li2Mn0.5Ni0.5O2 has also been synthesized titanium was used to provide electronic-grade tita-
in electrochemical cells, and both lithium ions can nium disulfide and sponge titanium provided re-
be cycled when part of the Mn is substituted by search battery-grade material. The latter could be
titanium.64 produced 1 lb at a time in silica tubes, had a surface
The group VI layered disulfides were originally not area of 5 m2/g, and allowed current densities of 10
thought to be of much interest because of poor mA/cm2 to be achieved. However, both involved
rechargeability. However, Haering et al.65 showed reaction with sulfur, which took from hours for the
that in MoS2, which occurs naturally as molybdenite, sponge to days for the bulk metal. Such material
if the molybdenum coordination could be changed would cost more than $100/lb. An inspection of the
from trigonal prismatic to octahedral, then the MoS2 commercial process, shown in Figure 5, for sponge
so formed could be effectively used as a cathode. They manufacture revealed that the precursor was tita-
accomplished this transformation by inserting one nium tetrachloride, a liquid at room temperature.
lithium per MoS2 into the lattice and then letting it This tetrachloride is available in tonnage quantities
convert to the new phase. This system formed the as it is used in the manufacture of titanium dioxide
foundation of a commercial battery developed by paint pigment. Therefore, a manufacturing process
MoliEnergy in British Columbia.65 was devised by two European companies that in-
Although the Li/TiS2 batteries were usually con- volved the formation of stoichiometric TiS2 by deposi-
structed in the charged state with pure lithium or tion from the gas-phase reaction of TiCl4 with H2S.
LiAl anodes,60 they were also built34 in the discharged This produced a sulfide with a morphology with many
state with LiTiS2 cathodes as now used in all LiCoO2 plates growing in three dimensions from a single
cells. In this scenario the cells must first be charged central point, which shows excellent electrochemical
by the deintercalation of the lithium ions. Whereas behavior.
LiVS2 and LiCrS2 were well known in the literature, The stoichiometry and ordering of the titanium is
the lithium-free compounds had not been successfully critical to the electrochemical behavior of TiS2. Stoi-
chiometric and ordered TiS2 has been shown to exist71

if the temperature is kept below 600 °C and to have
metallic conductivity.39 The titanium disorder can be
readily measured by attempting to intercalate weakly
bonding species such as ammonia or pyridine. In
practice, a slight excess of titanium, e1%, is benefi-
cial in that it reduces the corrosiveness of the sulfur
without significantly impacting the cell potential or
the lithium diffusion coefficient. It is preferable to
add this extra metal to the initial reaction medium.
3.2. Trichalcogenides and Related Materials

As noted above, Trumbore et al. at Bell Labs
discovered the electrochemical behavior of the tri-
selenide of niobium. NbSe3 was found23,72,73 to react
reversibly with three lithium ions to form Li3NbSe3
in a single phase. The other trichalcogenides also
readily react with lithium but not in such a reversible Figure 6. Reaction of n-butyllithium with titanium di-
manner. Thus, in TiS3, which is best represented as sulfide, vanadium pentoxide, and molybdenum trioxide
TiS(S-S), the polysulfide group reacts first with two (reprinted with permission from ref 77, copyright 1977 The
lithium, breaking the S-S bond to form Li2TiS3 in a Electrochemical Society).
two-phase reaction; this is followed by the reduction
of the titanium from Ti4+ to Ti3+ in a single-phase from the layered materials such as LiCoO2, which
reaction similar to TiS2 itself.40 Only this second step were of more interest for their magnetic properties.
is readily reversible, unlike Li3NbSe3, where all three
lithium ions are reversible.
3.3.2. High-Valent Oxides of Vanadium and Molybdenum
A number of other chalcogenide-rich materials Vanadium pentoxide, V2O5, and molybdenum tri-
have been studied, but although many of them have oxide, MoO3, were two of the earliest studied oxides.
a high capacity, their rates of reaction or conductivity Molybdenum oxide reacts readily with around 1.5
are low. This can be ameliorated to some extent by lithium/molybdenum,40,79 but is of little interest
admixing them with a high rate high-conductivity because of its low rate of reaction which is readily
material such as TiS2 or VSe2.74,75 At high rates of determined by the rate of temperature rise when the
lithium insertion the latter undergo reaction first, solid is added to the chemical lithiating agent n-
and then when the rate is reduced they are recharged butyllithium.77 Figure 6 shows the heat of reaction
by the other component, making them available for for three cathode materials, TiS2, V2O5, and MoO3;
the next high-current pulse. This mixed-cathode the higher the temperature increase, the greater the
approach might reemerge as batteries must meet power capability of the material. A high-power mate-
both high-power and high-energy demands. rial should be able to boil the hexane solution.
Vanadium pentoxide has been investigated for 30
In this time period a range of chemical reagents
years;40,80-84 it has a layered structure with weak
was devised to mimic the electrochemical reaction,
vanadium-oxygen bonds between the layers and is
either lithium reaction or lithium removal, so as to
now known to react by an intercalation mechanism:
speed-up the evaluation of new materials and get an 40,85
xLi + V2O5 ) LixV2O5.
idea of ease and depth of reaction. The most common
lithiating agent is n-butyllithium in hexane, a straw- The structural behavior on lithium insertion is
colored liquid, with clean clear easily identified fairly complex with the initial lithium merely inter-
reaction products such as octane, butane, and calating the structure, first forming the R-phase (x
butene.62,76,77 Although highly reactive with a poten- < 0.01) and then the -phase (0.35 < x <0.7), where
tial of around 1 V versus lithium metal, this reactant the layers are more puckered. At x ) 1, shifting of
gave much purer products than prior methods using one layer out of the two leads to the δ-phase.
naphthalene or liquid ammonia solvent. A series of However, if more than one lithium is discharged,
other chemical reagents with known redox potentials then significant permanent structural changes occur,
can be used to control the reductive or oxidative giving the γ-phase, which can be cycled in the range
intercalation of materials.78 0 e x e 2. In the R-, -, and δ-phases the VO5 square
pyramids that make up the structure of V2O5 are
arranged in rows which have the apexes ordered up,
3.3. Movement into Oxides up, down as shown schematically in Figure 7. In
3.3.1. Layered Oxides contrast, in the highly puckered γ-phase, these are
organized up, down, up, down. A rock-salt structure
Surprisingly the layered oxides with the same is formed when still more lithium is added; this
structures as the layered dichalcogenides were not compound is called the ω-Li3V2O5 phase. This ω-phase
studied in that time period. The thought was pre- cycles in a single solid-solution phase with the last
sumably that oxides toward the right of the periodic lithium coming out at over 4 V, clearly showing the
table would be of little interest, and it was not difference between this phase and the initial vana-
considered that lithium could be readily removed dium pentoxide phase, which has an open-circuit
potential of 3.5 V. This behavior is shown in Figure

7. The ω-material has a tetragonal structure, which
on extended cycling becomes a simple rock-salt
structure with a ) 4.1 Å with the formula Li0.6V0.4O.
Delmas82 reviewed these structural modifications and
attributes the high OCV at low lithium content to
the metastability of the defective rock-salt lattice
which can contain up to 60% vacancies on the cation
sublattice. The single sloping discharge plateau from
almost 4 to under 2 V makes it uninteresting for
practical applications where the change in potential
should preferably be under 0.5 V; moreover, this
phase also rapidly loses capacity on cycling.
3.3.3. Mixed-Valent Oxides of VanadiumsV6O13 and

LiV3O8
Murphy et al.86 made an extensive study of a
number of vanadium oxides and discovered the
excellent electrochemical behavior of the partially
reduced vanadium oxide, V6O13, which reacts with
up to 1 Li/V. They also recognized that the method
of preparation, which determines the V:O ratio,
critically controls the capacity for reaction with
lithium. The structure consists of alternating double
and single sheets of vanadium oxide sheets made up
of distorted VO6 octahedra. A variety of sites are
available for lithium intercalation, which if filled
sequentially would lead to the various steps seen in
the discharge curve. The lattice first expands along
the c-axis and then along the b-axis. Thomas et
al.87-91 made an in-depth study of the complex
intercalation process in single crystals of V6O13. This
phase was one of the leading candidates for polymer
electrolyte batteries, which require a lower voltage
system because of the limited thermodynamic stabil-
ity of the polymer.
Another vanadium oxide that has received much
attention is LiV3O8, which has a layer structure
composed of octahedral and trigonal bipyramidal
ribbons that can be swelled just like other layered
compounds and can intercalate lithium.92,93 Here
again, the method of preparation is important to its
electrochemical characteristics.94 West et al.95 made
a systematic study of the impact of synthesis tech-
nique on capacity and cycling and showed that
amorphous material increased the capacity above 2
V from 3-4 lithium per mole of LiV3O8 at low current
drains, 6-200 µA/cm2.
3.3.4. Double-Sheet Structures: Xerogels, δ-Vanadium

Oxides, and Nanotubes
There has been much interest in vanadium oxides
formed by sol-gel processes.96-98 These can be formed
by acidification of a sodium vanadate solution, for
Figure 7. Structure and discharge of vanadium pentox-
ide: (a) structure of V2O5 showing the square pyramids example, by passing it down an acidified ion-
sharing edges of the basal planes; (b) schematic showing exchange column. The resulting orange gel on drying
basal planes sharing edges and the orientation of the has the formula HxV2O5‚nH2O. About 1.1 mol of
apexes of the pyramids comparing the structure in the water may be removed under vacuum or on mild
perfect lattice of VO2 compared with the ordered defect heating, leading to H0.3V2O5‚0.5H2O, based on its high
lattice of V2O5 (reprinted with permission from ref 364, cation-exchange capability. The interlayer spacing is
copyright 1996 The Electrochemical Society); and (c) elec-
trochemical lithium intercalation into V2O5 showing the around 8.8 Å, which swells to 11.5 Å for the more
evolution of phases with degree of lithium intercalation and hydrated form with 1.8H2O. The protons and water
the cycling of ω-phase (reprinted with permission from ref can be readily exchanged for lithium and polar
82, copyright 1994 Elsevier). solvents. These vanadium oxide gels can be also
Figure 8. (a) Structure of the two-layer structure found

in xerogel, and (b) comparison of the electrochemical
behavior of the crystalline,82 xerogel,362 ω-phase,83 and
aerogel100 forms of vanadium pentoxide.
made99 by the hydrogen peroxide treatment of V2O5.

These xerogel vanadium pentoxides contain sheets
comprised of two vanadium oxide layers, with all the
vanadyl bonds on the outside leading to a distorted
octahedral coordination around the lithium instead
of the square pyramidal coordination found in crys-
talline V2O5. Recently, it was reported100 that making
V2O5 through an aerogel process with supercritical
drying in CO2/acetone gives a much more electro-
chemically active product, HyV2O5‚0.4H2O‚carbon; the
amount of carbon was 3.9 wt %. The dried material,
which had a 12.5 Å lattice spacing, reacts with
lithium in a single continuous discharge curve with Figure 9. Electrochemical behavior of (a) δ-MnyV2O5
(reprinted with permission from ref 105, copyright 2000
a midpoint emf of around 3.1 V and a total lithium Elsevier), (b) δ-NH4V4O10, and (c) Mn vanadium oxide
uptake of 4.1 Li by 2.8 V, thus giving a much higher nanotubes.112
capacity than crystalline V2O5 as shown in Figure 8.
The double layers of vanadium oxide found in the
xerogel have been described in a number of other electrochemical behavior.110 In some cases the capac-
vanadium oxides by Galy101 and Oka;102 they also ity increases on cycling;111 the electrochemical be-
form the double sheets described above for V6O13. havior of the manganese ion-exchanged compound is
These oxides, in which the vanadium is found in also shown112 in Figure 9.
distorted VO6 octahedra, show particularly attractive
electrochemical capacities103-107 exceeding 200 mAh/g 4. 1980−1990: Era of Layered Oxides and First
in some cases, as shown in Figure 9. However, at the
present time their rate capability appears somewhat
Large Commercialization
limited. More recently vanadium oxide nanotubes 4.1. Early Studies of Layered Oxides
have been synthesized, first by Spahr et al.;108,109
these compounds also contain double sheets of va- Although the heavier alkali-metal compounds of
nadium oxide and again have interesting but complex the oxides of manganese, cobalt, chromium, and
others had been extensively studied by a number of

groups, in particular, at Bordeaux by Hagenmuller
and Delmas113-119 in the 1970s, there was essentially
no in-depth study of the corresponding lithium
compounds. This was not dissimilar to the situation
in Nantes, where Rouxel et al.119-122 were researching
the layered dichalcogenides at the same time. This
can be in part associated with the lack of a good
lithiating agent, which was solved with the discovery
of the synthetic prowess of n-butyllithium and elec-
trochemical synthesis.
In the studies in the 1970s the structures of these
layered oxides and chalcogenides were extensively
studied. Whereas lithium only occupies octahedral
sites in the strictly layered materials when the Li/ Figure 10. Schematic of stacking of building blocks for
transition-metal ratio is unity or less, the larger one-block, two-block, and three-block structures.
alkali ions often occupy trigonal prismatic sites.
described above may be represented for the dioxide
Which site is occupied depends on the alkali cation
MO2 and lithiated LiMnO2 as follows.
size and their concentration, with some ions such as
sodium exhibiting a range of structures. The stacking
slab
of the MO2 slabs in the overall structure depends on stacking MO2 LiMO2
the alkali site occupancy. As significant reorganiza-
tion of the lattice must occur in switching from one single block ..(AcB)(AcB).. ..(AcB)c(AcB)..
structure to another, such changes will impede double block ..(AcBCaB)(AcBCaB).. ..(AcBaCaB)c(AcBaCaB)..
triple block ..(AcBCbABaC)- ..(AcBaCbAcBaC)b-
electrochemical reactions. Thus, it is important to (AcBCbABaC).. (AcBaCbAcBaC)..
know the phases formed as the alkali concentration
changes. In the case of the small lithium ion, one It can thus be seen that long-range atomic motion
needs only be concerned with the stacking sequence (slab-sliding) is required to convert one of these
and whether it changes as the lithium concentration structures into another one. These arrangements are
changes as the lithium ion normally occupies only shown schematically in Figure 10. It can also be
the octahedral sites for x e 1 in LixMO2. For most noted that in the single block structures all the
layered oxides the transition metal also only occupies cations are stacked directly above each other, while
octahedral sites. There are at least three different in the triple block structure all the cations are
stacking sequences of the MO2 slabs (building blocks) staggered to minimize any ionic interactions. Thus,
that can form the unit cell; the lithium ions reside the former tends to be formed by more covalent
between the slabs. These three stacking arrange- lattices and the latter by the more ionic lattices. In
ments are as follows. addition to these idealized hexagonal lattices, there
(a) Single blocks stacked upon one another, as are a number of instances in which the lattice is
in, for example, LiTiS2, CoO2, and TiS2. This distorted and where cations migrate between the
is the CdI2 structure. This is a hexagonal slabs and the interslab region. This, often unintended
close-packed lattice. A special case of the migration, can play a significant role in the reactivity
single block structure is given by the di- and electrochemical behavior of the material. The
lithium compounds, such as Li2VSe2, where special case of the dilithium compounds, Li2MO2,
the lithium resides on all the tetrahedral where the lithiums are in tetrahedral sites, can be
sites. represented by the slab-stacking sequence ..(AcB)-
(b) Double blocks stacked upon one another, ba(AcB)... There is only one realistic stacking se-
normally formed by ion exchange from a quence because an octahedral site in the transition-
sodium-containing compound. metal layer cannot be occupied directly opposite a
(c) Triple blocks, each block displaced by one- tetrahedral site. Where some octahedral sites are
third in the basal plane, stacked on top of one occupied in the lithium layer, the total lithium
another. This sequence is very common in the content will be less than two and/or the lattice will
lithiated oxides, such as LiCoO2 (the R-NaFeO2 become distorted or even amorphous.
structure). This is a cubic close-packed lat-
tice. 4.2. Lithium Cobalt Oxide, LiCoO2
In addition, at very low lithium contents not every Goodenough recognized that LiCoO2 had a struc-
layer between the slabs need be occupied by lithium ture similar to the layered structures of the dichal-
ions but, for example, just every other layer. This is cogenides and showed that the lithium could be
known as a second-stage compound. The above three removed electrochemically, thus making it a viable
stacking arrangements can also be described in terms cathode material.123 LiCoO2 has the R-NaFeO2 struc-
of the positions of the atoms. Thus, if a slab is ture with the oxygens in a cubic close-packed ar-
described by the designation AcB, where A, B, and rangement. On complete removal of the lithium, the
C represent the three possible positions of the anions oxygen layers rearrange themselves to give hexago-
in the hexagonal lattice and a, b, and c represent the nal close packing of the oxygen in CoO2.124 Be-
cation positions, then the three stacking sequences tween these composition limits several phases are
formed with varying degrees of distortion of the ccp 2.75 and 4.4 V without capacity fade over 70 cycles.
oxygen lattice. The composition Li0.5CoO2 can also be Other researchers139-141 quickly confirmed the posi-
formed in the spinel form,125 though it appears to be tive results of surface coatings. The mechanism of
metastable and is not normally formed during the protection is related to minimizing the reactivity of
cycling of the LixCoO2 electrode. However, a recent Co4+ on charge with the acidic HF in the electrolyte
TEM study126 has identified the spinel phase origi- coming from the interaction of moisture with the
nating on the surface of heavily cycled LiCoO2 electrolyte salt LiPF6.142 Removing the source of the
cathodes. HF should eliminate the capacity loss as found143,144
SONY combined the LiCoO2 cathode with a carbon for the spinel LiMn2O4, where switching to the
anode to make the first successful Li-ION bat- LiBOB salt essentially eliminated manganese dis-
tery,127,128 which now dominates the lithium battery solution and capacity loss. This was also the case for
market. The carbon anode, which forms the com- LiCoO2,145 where replacing the LiPF6 salt by LiBOB
pound LiC6 on reaction with lithium, makes a much or by completely drying the LiCoO2 by heating to over
safer battery than if pure lithium is used as there is 550 °C, improved the capacity retention at 180 mAh/g
much less chance of the formation of dendritic at a 4.5 V cutoff. Above this 4.5 V level, the three-
lithium, which can lead to cell shorting. The use of block cubic close-packed LixCoO2 converts to the 1T
graphitic carbon can result in the loss of 100-300 one-block hexagonal close-packed structure of CoO2;
mV in cell potential, which is feasible with the higher this requires substantial movement of the oxygen
potential LiCoO2 cathode but not with the lower layers in going from ABCA to ABA stacking sequence
potential of the TiS2 cathode. The commercial cell is which will eventually significantly disrupt/disorder
built in the discharged state: C-LiCoO2. It thus must the structure. Thus, one cannot expect capacities for
be charged before use. The theoretical capacity of the LiCoO2 much above 180 mAh/g to be attainable over
LiCoO2 cell is relatively low at around 130 mAh/g hundreds of cycles. A more recent study146 suggests
because only around 0.5 Li/Co can be reversibly that CoO2 has a monoclinically distorted CdCl2
cycled without causing cell capacity loss due to structure, whereas NiO2 has a monoclinically dis-
changes in the LiCoO2 structure. This can be associ- torted CdI2 structure, but little detail is given and
ated with two factors: phase changes that cause low no mention is made of the nickel content or distribu-
reaction rates and the poor stability of the electrode tion.
at low lithium contents. The message here may be Although the LiCoO2 cathode dominates the re-
that reactive nanosize materials/components may not chargeable lithium battery market, there is a limited
be viable for commercial batteries because of safety availability of cobalt, which causes it to have a high
and life considerations. price. This price limits it use to small cells, such as
The diffusion of lithium in LiCoO2 is 5 × 10-9 cm2/ those used in computers, cell phones, and cameras.
s, which compares with 10-8 cm2/s for LiTiS2.129 These An alternative cathode will be needed for large-scale
high diffusivity values are consistent with the ability applications, as envisioned in HEV or for load level-
to cycle these two cathodes at 4129 and 10 mA/cm2,13 ing. The LiCoO2 patent147 covered more than this one
respectively. However, the conductivity of LixCoO2 cathode, describing all layered transition-metal ox-
remains a challenge as it is reported130 to change ides with the R-NaFeO2 structure where the transi-
dramatically with composition, behaving like a metal tion metal is vanadium through nickel. In addition,
at x ) 0.6 and a typical semiconductor at x ) 1.1 (the combinations of the transition metals were described,
typical lithium-rich material used in commercial such as LiCo1-yNiyO2. In addition, another patent148
cells), changing by 2131 (for the x ) 1.1 compound) to covers the electrodeintercalation process for forming
4130 (for the x ) 1.0 compound) orders of magnitude the alkali-metal compounds AxMO2 with the R-NaFeO2
at ambient temperatures and up to 6 orders of structure where A is Li, Na, or K and x < 1.
magnitude at lower temperatures.131
The energy density of commercial cells has almost 4.3. Lithium Nickel Oxide, LiNiO2
doubled since their introduction in 1991; since 1999 Lithium nickel oxide, LiNiO2, is isostructural with
the volumetric energy density has increased from 250 lithium cobalt oxide but has not been pursued in the
to over 400 Wh/l.132 Details of the original commercial pure state as a battery cathode for a variety of
cells have been reviewed by Nishi,133 where key reasons, even though nickel is more readily available
aspects are discussed concerning the need for large than cobalt. First, it is not clear that stoichiometric
particle size, 15-20 µm, to increase safety and the LiNiO2 exists. Most reports suggest excess nickel as
intentional incorporation of lithium carbonate into in Li1-yNi1+yO2; thus, nickel is always found in the
the cathode to provide a safety valve. The lithium lithium layer, which pins the NiO2 layers together,
carbonate decomposes, releasing carbon dioxide when thereby reducing the lithium diffusion coefficient and
the charging exceeds 4.8 V, which breaks electrical the power capability of the electrode. Second, com-
flow in the cell. These lithium cobalt oxides also pounds with low lithium contents appear to be
contain excess lithium and can be best represented unstable due to the high effective equilibrium oxygen
by the formula Li1+xCo1-xO2. partial pressure, so that such cells are inherently
In 1999 Cho et al. reported134-138 in a pioneering unstable and therefore dangerous in contact with
series of papers that the capacity could be improved organic solvents. A second lithium can be inserted
by coating a metal oxide or phosphate on the surface either chemically or electrochemically, as in Li1.8-
of the LiCoO2 particles. They found that the capacity Ni1+yO2, which is a mixture as expected of “LiNiO2”
could be increased to 170 mAh/g when cycled between and “Li2NiO2”.63
We will discuss in section 5.2.1 of this review the

modification of this material by replacing a part of
the nickel by other elements such as cobalt and
aluminum. The former assists in ordering the struc-
ture, that is keeping the nickel in the nickel layer,
and the latter, being redox inactive, prevents the
complete removal of all the lithium, thus additionally
stabilizing the structure and preventing any phase
changes that might occur at very low or zero lithium
content. Unlike cobalt and nickel, manganese does
not form a stable LiMnO2 phase with the LiCoO2
structure, with the spinel structure being the stable
phase at the composition Li0.5MnO2. As there are a
myriad of structures149 with the 1:2 Mn:O ratio, other
structures may be stable at different lithium con-
tents.
5. 1990−Present: Second-Generation Lithium

Batteries
5.1. Spinels
The spinel cathode LiMn2O4, originally proposed
by Thackeray et al.,150 has been extensively developed
by the Bellcore labs,151-153 and has recently been
reviewed by Thackeray154 with the key structural
aspects by Yonemura et al.155 The anion lattice again
contains cubic close-packed oxygen ions and is closely
related to the R-NaFeO2 layer structure, differing
only in the distribution of the cations among the
available octahedral and tetrahedral sites. The dis-
charge proceeds in predominantly two steps, one
around 4 V and the other around 3 V as shown in
Figure 11. Usually only the 4 V plateau is used, so
that the cell is constructed in the discharged state
and must be charged before use just as for LiCoO2.
charging: LiMn2O4 f Mn2O4 + Li (inserted into

anode host, such as graphitic carbon)
Figure 12. Correlation of the lattice parameter of the
spinel Li1+xMn2-xO4 with (a) the lithium content, (b)
It has been reported153,156
that the value of the cubic manganese oxidation state, and (c) capacity loss of the cell
lattice parameter, which is directly related to the over the first 120 cycles, after ref 157.
average oxidation state of the manganese, is crit-
ical to obtain effective cycling. The lattice param- Jahn-Teller distortion associated with the Mn3+ ion.
eter should preferably be 8.23Å or less, and such The variation in lattice parameter, ao, with chemical
values are associated with lithium-rich materials, composition in Li1+xMn2-xO2 is shown in Figure 12a
Li1+xMn2-xO4, where the average manganese oxida- (after ref 157); its variation is given by ao ) 8.4560
tion state is 3.58 or higher; this value minimizes - 0.21746x.
dissolution of manganese and also the impact of the The lattice parameter may also be used as an
indirect measurement of the manganese oxidation
state, shown in Figure 12b. Figure 12c, in addition,
shows clearly the impact of lattice parameter on the
percent capacity loss over the first 120 cycles. The
retention of cycling capacity at elevated temperatures
can be helped by the simultaneous doping with
aluminum and fluoride ions, as in Li1+xMn1-x-yAly-
O4-zFz, where y and z are around 0.2.157 Moreover, if
the potential on the surface of the spinel is kept above
that for the formation of the Li2Mn2O4 phase, then
the formation of Mn2+ by the disproportionation of
surface Mn3+ ions is minimized: 2Mn3+ ) Mn2+ +
Mn4+. It is the divalent manganese ions that are
soluble in the acidic electrolyte, and so every attempt
Figure 11. Potential profile of LiMnO2 and spinel LiMn2O4 must be made to minimize their formation. Once
(data from Bruce and Whittingham). dissolved into the electrolyte, the manganese ions can
diffuse across to the anode and be reduced there to effects on the layeredness of the structure.172 Thus,
manganese metal, thereby using up the lithium and for y ) 0.10, 0.20, and 0.30 the amount of 3d metal
reducing the electrochemical capacity of the cell. in the lithium layer is 6.1%, 8.4%, and 7.4%, respec-
This spinel is presently the center of much interest tively, for samples formed at 750 °C. Although a
as the cathode of a high-power lithium battery for LiFeO2 compound would be ideal for a low-cost
hybrid electric vehicles, even though under high battery, it has been reported173 that the lithium
drain rates its capacity is only around 80 mA/g. This cannot be deintercalated within the normal potential
material has been plagued by self-discharge when left ranges used in lithium batteries; this is explained174
under full charge, particularly at elevated tempera- by the lack of compression of the FeO6 octahedra
tures; however, this problem may have been solved which makes the reduction of one electron from Fe3+
by switching from the fluoride-containing LiPF6 salt, very difficult.
which in the presence of traces of moisture can An issue with all these layered oxides is their
generate HF, to salts such as LiBOB.143,144 The spinel electronic conductivity, which is not uniformly high
used is in a stabilized form in which a part of the across the lithium composition range, or nickel
manganese has been substituted as in Li1.06Mn1.95- substitution. Thus, cobalt substitution in LiNiO2, as
Al0.05O4, and there have been several hundred studies in LiNi0.8Co0.2O2, reduces the conductivity.169 In
on the impact of this substitution.157-159 These studies addition, as lithium is removed from the phase
will not be discussed here. An alternate solution LixNi0.1Co0.9O2168 or from LixCoO2130 the conductivity
pioneered by the Korean school is to coat the surface was found to increase dramatically by some 6 orders
of the spinel particles with materials such as zirco- of magnitude to around 1 S/cm from x ) 1 to 0.6.
nium dioxide or AlPO4, which are believed to act as These dramatic changes demand that a conductive
getters for any HF. diluent be added to the cathode-active material,
One spinel, Li4Ti5O12, is under serious consider- which reduces both the energy storage and the power
ation as an anode in high-power cells as its charging capabilities.
potential (lithium insertion) is around 1.55 V,160-162 Studies have shown that the cobalt-substituted
so there is no danger of lithium metal deposition as nickel oxides are more stable and thus are less likely
might occur on graphitic carbon at high rates. Rates to lose oxygen than the pure nickel oxide. The
as high as 12C have been claimed (total reaction in addition of a little of a redox-inactive element such
60/12 ) 5 min), both with nano- and microstructured as magnesium reduces the capacity fading on cy-
materials at 60 °C.163 This electrode is being consid- cling,175 as in LiNi1-yMgyO2;176 this inert element
ered in combination with a high-rate cathode such prevents the complete removal of all the lithium and
as a mixed layered oxide or with the spinel LiMn2O4 thus minimizes possibly structural collapse and
to give a lower cost and safe 2.5 V cell. It has also reaching such a high level of oxygen partial pres-
been coupled with the olivine LiFePO4 (see later in sure-NiO2 itself is thermodynamically unstable at
this paper) and cycles well at a potential of 1.8 V with 25 °C, as the equilibrium oxygen partial pressure
no capacity loss over a 100 cycles.164 If coupled with exceeds 1 atm.
a high-voltage spinel, then the cell potential might Substituted nickel oxides, such as LiNi1-y-zCoyAlzO2,
attain 3.5-4 V.165 are prime candidates for the cathode of advanced
lithium batteries for use in large-scale systems as
required for hybrid electric vehicles. On charging
5.2. Other Layered Oxides these mixed oxides the nickel is oxidized first to Ni4+
then the cobalt to Co4+.177 SAFT has constructed cells
5.2.1. Mixed Nickel−Cobalt Dioxide, LiNi1-yCoyO2 with these substituted nickel oxides that have been
Many different elements can be substituted into cycled 1000 times at 80% depth of discharge with an
the R-NaFeO2-type structure, and they impact the energy density of 120-130 Wh/kg.178
layeredness of the structure, its stability on lithium 5.2.2. Lithium Manganese Dioxide, LiMnO2
removal, and the retention of capacity on cycling. In
a series of papers the Delmas group166-169 and Much interest has been placed on the layered
Zhecheva et al.170 determined the structural details LiMnO2 compound for its prospects of providing not
and physical properties of the LiNi1-yCoyO2 system only a low-cost but also an environmentally benign
and showed that there is an increased ordering as cathode material.179-181 However, it is not thermo-
the cobalt concentration increases; they found that dynamically stable at elevated temperature and thus
the c/3a ratio increases monotonically from 1.643 to cannot be synthesized by the same methods as used
1.652 as y increases from 0 to 0.4 and that there is for materials like NaMnO2. Instead, other approaches
no nickel content on the lithium sites for y g 0.3. must be used. One approach is to ion exchange the
Thus, cobalt suppresses the migration of nickel to the sodium compound, giving LiMnO2, which was ac-
lithium site in the mixed Li nickel/cobalt compounds complished independently by Bruce181 and Delmas;182
and one would expect and finds the same behavior starting with different layered structures maintains
in the Li nickel/manganese/cobalt oxides. Cobalt is the stacking order of the parent manganese oxide,
also reported to facilitate the oxidation of iron atoms thus allowing study of the impact of stacking se-
in the structure.171 Other ions, such as iron, do not quence on electrochemical performance. An alterna-
have the same positive effect as cobalt; thus, in the tive preparative approach is to synthesize the struc-
compound LiNi1-yFeyO2 the capacity is reduced with ture at low temperatures, as, for example, by
increasing iron content and the iron has no positive hydrothermal synthesis/decomposition of alkali
permanganates,180,183-185 which in the case of lithium

results in the composition Li0.5MnO2‚nH2O. Mild
warming causes the loss of water to give the desired
layered LixMnO2; overheating to 150 °C results in the
formation of the spinel LiMn2O4. Birnessite-type
phases have also been made by acid treatment of
manganese oxides.186-189
However, LixMnO2 easily converts to the thermo-
dynamically stable spinel structure upon cycling
lithium in and out but apparently not on acid
delithiation;190 this conversion requires no oxygen
diffusion as both structures have ccp oxygen lattices.
This ccp lattice has oxygen layers in the sequence
AcB|aCbA|cBaC|bA, i.e., there are three building
blockssMnO2 blocks (upper case is oxygen, lower
case is manganese, italic lower case is lithium). Two
approaches to stabilize the layered LiMnO2 have been
taken. In the geometric stabilization approach, non-
ccp structures are proposed such as tunnel struc-
tures,191-195 two-block (see Figure 10) or other non
CCP close-packed structures,196 or “pillars” are placed
between the layers to provide the stabilization. We
reported on compounds KMnO2 180,184 and (VO)y-
MnO2,197,198 which are examples of such pillared
structures. The former is stable to spinel formation Figure 13. Electronic conductivity of pillared KMnO2 with
at low current densities, and the latter shows excel- 10% Co, Ni, and Fe substitution of the Mn.
lent stability but poor rate capability. The groups of
Dahn and Doeff among others have pursued non-ccp theoretically238 and found to go in a two-step process;
structures by looking at tunnel structures such as in the first step a fraction of the lithium and
Li0.44MnO2191,193,194 and also by ion-exchanging lay- manganese ions move rapidly into tetrahedral sites,
ered sodium manganese oxide compounds with non- and in the second stage these order into the spinel
ccp stacking of the oxygen sheets;199-205 these sheets arrangement of cations.
cannot readily reorganize after ion exchange to give
ccp stacking. This results in, for example, the two-
5.2.3. Mixed Manganese−Cobalt Dioxide, LiMn1-yCoyO2
block rather than the three-block structures such as Bruce et al.206,239,240 investigated the synthesis
O3, which is that of ccp LiMnO2. Such compounds and electrochemical behavior of cobalt-substituted
have also been studied with partial substitution of LiMn1-yCoyO2. They synthesized this material from
the manganese by, for example, Co.205 This ion- the sodium analogue by ion exchange and achieved
exchange process also results in much faulting in the values of y up to 0.5. These substituted materials
stacking of the layers, which impedes the layers from have the R-NaFeO2 structure. These materials, just
reordering into the thermodynamically stable O3 like the unsubstituted LiMnO2, convert to the spinel
phase. However, these phases intercalate lithium structure on cycling even at the low cycling rate of
over a rather wide range of potential and in some 0.1 mA/cm2; for y ) 0.1, this begins to occur on the
cases over two potential steps.201 first cycle and is not apparent until the thirtieth cycle
In the electronic stabilization approach the goal is for y ) 0.3. However, they cycle over the 3 V spinel
to make the electronic properties of Mn to be more plateau very well. The conversion to spinel is ex-
cobalt-like by substitution of the Mn with more pected to occur very rapidly at elevated temperatures.
electron rich elements such as Ni.196 The successful The partial substitution of the manganese ions by
substitution of Mn by Co190,206,207 and Ni208-214 has cobalt, iron, or nickel was found to significantly
been reported. The first studies on LiNi1-yMnyO2, for increase the electronic conductivity of the manganese
0 < y e 0.5, indicated low capacities and poor oxide, which in the pure state as in LiMnO2 or
reversibility.212 However, Spahr et al. later dem- KMnO2 is around 10-5 S/cm, too low to allow rapid
onstrated a high capacity and reversibility for reaction without addition of a conductive diluent. To
LiNi0.5Mn0.5O2213 with a discharge curve typical obtain sufficiently dense material for conductivity
of that of LiNiO2. More recently, the compounds measurement, the potassium analogues were syn-
LiNi1-y-zMnyCozO2 have been extensively investi- thesized at elevated temperatures with 10% of the
gated in the last 3 years and found to have properties manganese substituted. The data,241 shown in Figure
that qualify them as possible candidates for the 13, clearly show the advantage of adding cobalt,
replacement of LiCoO2.198,215-237 In addition to their which enhanced the conductivity by almost 2 orders
high lithiation capacities and reversibility, these of magnitude. Nickel had the least effect, and we will
compounds show higher thermal stabilities compared discuss the impact of joint additions of cobalt and
to the Co-free compounds. These compounds are nickel as in LiNi0.4Mn0.4Co0.2O2.219
discussed in the next sections. The layered-to-spinel These cobalt-substituted materials can also be
phase transition in LixMnO2 has also been considered prepared hydrothermally, and their cycling behavior
is much improved over the cobalt-free compounds.242 et al.213 repeated the work, also showing a maxi-
Conversion to spinel is observed on the first charge mum solubility of 0.5 Mn. They however found
cycle at rates of 1 mA/cm2, even when larger cations optimum electrochemical behavior for the composi-
such as potassium are incorporated into the structure tion LiNi0.5Mn0.5O2. They reported XPS and magnetic
as pillaring agents. The layered-to-spinel phase data that are consistent with the present inter-
transition in LixMnO2 has subsequently been ex- pretation of Ni2+ and Mn4+ ions rather than Ni3+ and
plained through modeling.238 Mn3+ and showed electrochemical cycling curves very
Work on these cobalt-substituted R-NaFeO2 struc- reminiscent of LiNiO2. This compound, which we will
ture materials, where manganese is the redox-active call the 550 material (0.5 Ni, 0.5 Mn, 0.0 Co) was
ion, has essentially ceased because of the inability rediscovered by Ohzuku214 in 2001, who reported very
to maintain the structure relative to conversion to good electrochemical data, and this was recon-
the spinel structure under realistic cycling conditions. firmed almost immediately by the Dahn group.252
Doping elements other than cobalt have also been These papers ignited a substantial amount of work
investigated, but substitution by nickel leads to a on this composition and on the cobalt-substituted
system where the manganese becomes the structure compounds, in particular those with the formula
stabilizer and nickel the electrochemically active ion. LiNiyMnyCo1-2yO2, where 0.5 e y e 0.33, which can
These compounds are thus best described as substi- be considered as a solid solution of LiNi0.5Mn0.5O2 and
tuted nickel oxides in which the manganese remains LiCoO2.
in the tetravalent state and the nickel is redox active Manganese shows no significant layer stabilization
between the +2 and +4 oxidation states; the man- capability as up to 10% nickel is found on the lithium
ganese helps in reducing the cost and stabilizing the sites, 9.3%253 for a sample formed at 1000 °C, and
lattice. Cobalt, as will be discussed below, plays a 11.2% for a sample formed at 900 °C;254 this nickel
critical role in controlling the ordering of the 3d ions is expected to reduce the rate capability of the
in the structure. electrode. The phase LiNi0.5Mn0.5O2 has the expected
hexagonal lattice with a ) 2.894 Å and c ) 14.277Å,213
5.2.4. Mixed Nickel−Manganese Dioxide, a ) 2.892 Å and c ) 14.301 Å,214 a ) 2.891 Å and c
LiNi1-yMnyO2sMultielectron Redox Systems ) 14.297 Å,253 a ) 2.888 Å and c ) 14.269 Å,254 a )
The groups of Ammundsen and Davidson an- 2.887 Å and c ) 14.262 Å,255 a ) 2.895 Å and c )
nounced results on the Li-Mn-Cr-O2 system243,244 14.311 Å,256 and mean c/3a ) 1.647. There is some
at the 2000 Lithium Battery meeting in Como. They disagreement about the structure formed on lithium
found245-250 that these mixed-metal compounds had removal. Venkatraman et al. reported254 a single
the NaFeO2 structure and cycled well when a subphase for 0 e Li e 1 with continuously varying lattice
stantial part of the transition metals were substi- parameters; the data reported for Li0.2Ni0.5Mn0.5O2 is
tuted by lithium as in Li3CrMnO5 or described in the consistent with single-phase behavior. However,
layered form as Li[Li0.2Mn0.4Cr0.4]O2. The discharge Yang et al.256 reported that on delithiation a second
curve showed the typical behavior of a single phase hexagonal phase with a ) 2.839 Å, c ) 14.428 Å was
rather than the two-phase, two-step behavior of the formed; this is in contrast to the pure LiNiO2, where
spinel. The manganese has the 4+ oxidation state, a third hexagonal phase is also found. Arachi et al.
and all the lithium can be removed giving Cr(VI). reported253 the formation of a monoclinic phase for
What was surprising was the high mobility of the Li0.5Ni0.5Mn0.5O2with a ) 4.924 Å, b ) 2.852 Å, c )
chromium, which during the redox process must 5.0875 Å, β ) 108.81°. Yang et al.256 also reported
move from an octahedral to a tetrahedral site; that excess lithium could be intercalated with a slight
increasing the temperature from 20 to 55 °C in- expansion of the hexagonal cell to a ) 2.908 Å and c
creased the cell capacity from around 125 to 165 ) 14.368 Å; this is possibly the phase Li2Ni0.5Mn0.5O2
mAh/g, consistent with having to move the highly and if so raises the question of which site the nickel
charged chromium ion as well as the lithium ion. in the lithium layer occupies if the lithium takes up
Capacities as high as 220 and 200 mAh/g were the tetrahedral sites. One would not expect the
obtained at low rates, 3248 and 10 mA/g,247 respec- rhombohedral structure reported but rather a single
tively, from 2.5 to 4.5 V. The lithium ions in the block structure with c around 4.8 Å if the lithiums
transition-metal layer were found clustered around are in the tetrahedral sites; the lithium ions in
the manganese ions as in Li2MnO3, and the system tetrahedral sites would be much too close to the
can be considered as a solid solution of Li2MnO3 and transition-metal ions in the transition-metal layer.
LiCrO2. Although of little commercial interest, be- There is no data reported on the stability of this
cause of the toxicity of Cr(VI), this pioneering re- phase, which might be expected to be metastable like
search provoked much thinking about other multi- Li2NiO2, which converts63 at 400 °C to the ortho-
electron redox reactions which are discussed below. rhombic form of Li2NiO2.257
The corresponding solution between Li2MnO3 and Spahr et al. reported213 in 1998 a capacity of 150
LiCoO2 did not show similar behavior, with the mAh/g falling to 125 mAh/g after 25 cycles and to 75
capacity decreasing with increasing manganese con- mAh/g after 50 cycles for a 550 sample prepared at
tent when cycled between 3.0 and 4.2 or 4.3 V.207,251 700 °C; they also showed that the capacity and
The LiNi1-yMnyO2 phase system was studied by the capacity retention increased as the synthesis tem-
Dahn group212 in 1992. They reported a solid solution perature was increased from 450 to 700 °C, which
for y e 0.5 but a deterioration of the electrochemical we now know to be too low a temperature for
behavior with increasing manganese content. Spahr optimum electrochemical behavior. Ohzuku et al.214
Figure 14. Lithium cell potentials of the two lithium

nickel manganese dioxides with layered and spinel struc-
tures.260
Figure 15. Ragone plot for LiNi0.5Mn0.5O2 showing the
capacity of the cathode as a function of the discharge
prepared the 550 material at 1000 °C and reported current density (data from ref 258).
a constant capacity of 150 mAh/g at 0.1 mA/cm2 for
30 cycles using top-up charging at 4.3 V. The cell
those reported above. The rate capability has been
potential214,258,259 varies from around 4.6 to 3.6 V as
determined in another study,258 where the capacity
shown in Figure 14;260 it shows a single phase for all
approaches 200 mAh/g at 0.17 mA/cm2, falling off to
x values in LixNi0.5Mn0.5O2. The composition LiNi0.25-
130 mAh/g at 6 mA/cm2; all cells were charged at 0.17
Mn0.75O4, which has the spinel rather than the
mA/cm2 and then held at 4.5 V for 19 h; cathode
layered structure, shows the typical two-phase, two-
loading was around 15 mg/cm2. These data are shown
step spinel discharge potential and is very similar
in Figure 15 as a Ragone plot and strongly suggest
to that of LiMnO2 shown in Figure 11 except for the
that pulse discharge rates in excess of 10 mA/cm2
higher potentials in this case. The 550 material,
should be achievable. Rate data will also be needed
synthesized at 900 °C and quenched to room tem-
on charging if such cells are to be considered for HEV
perature, also showed a capacity exceeding 150
applications.
mAh/g for over 50 cycles in thin film configuration.252
Material cycled261 with a 4.4 V top-up constant The electrochemically active element in this com-
potential charge after a constant current charge to pound is nickel, which cycles between the +2 and +4
4.4 V showed an initial capacity exceeding 170 mAh/g valence states, while the manganese remains as +4
but decayed over 20 cycles to less than 150 mAh/g, throughout independent of the lithium content. First-
whereas a sample charged to just 4.2 V only had a principal quantum mechanical calculations264 as well
capacity of 130 mAh/g but maintained this capacity as structural measurements confirm this redox as-
much better; increasing the temperature to 50 °C signment. As the manganese is always 4+, there is
increased the fade rate except where the maximum no concern with the Jahn-Teller distortion associ-
charge potential was limited to 4.2 V. The 550 ated with the Mn3+ ion. Consistent with this model
composition showed a lithium diffusion coefficient of it should be possible to replace the manganese by an
around 3 × 10-10 cm2/s for most of the lithium element such as titanium; Kang et al.265 synthesized
composition range. A contradictory report255 shows the compound Li0.9Ni0.45Ti0.55O2 with the R-NaFeO2
good cycling stability at 0.1 mA/cm2 (10 mA/g) even structure by ion exchange of the sodium analogue,
when charged to 4.6 V, with the capacity increasing as high temperatures lead to complete cation disorder
from 150 to 190 mAh/g as the charging cutoff and a rock-salt structure. About one-half of the
potential is raised from 4.3 to 4.6 V. A material lithium could be removed in an electrochemical cell,
formed at 1000 °C showed a lower capacity of around with only about one-half of that being re-intercalated
120 mAh/g at 0.1 mA/cm2; the capacity was increased on discharge; this is believed to be due to cation,
to about 140 mAh/g by addition of 5% cobalt, alumi- probably titanium, migration into the lithium layer.
num, or titanium.262 This suggests that the synthesis Clearly, the manganese ion plays a key stabilizing
temperature of 1000 °C may be too high, leading influence on the R-NaFeO2 structure.
possibly to excess nickel in the lithium layer. This Little is known about the electrical conductivity of
550 compound can intercalate a second lithium, the 550 material. One measurement shows a con-
particularly when some titanium is added, yLiNi0.5- ductivity of 6 × 10-5 S/cm at 25 °C219 for the fully
Mn0.5O2‚(1-y)Li2TiO3, which results from reduction lithiated LiNi0.5Mn0.5O2. The magnetic susceptibility
of Mn(IV) to Mn(II);64 no structural data was given exhibits Curie-Weiss behavior at elevated temper-
for this new phase. Lu et al.252 showed that the atures.262
capacity could be increased from around 160 to 200 The precise details of the structure of LixNi0.5-
mAh/g at 30 °C by substituting a part of the transi- Mn0.5O2 are complex, with a superstructure being
tion metals by lithium, Li[Ni1/3Mn5/9Li1/9]O2. There is observed in the X-ray diffraction pattern.252,266 Long-
one report263 of conductive carbon coatings raising the range order has been detected by transmission
capacity, but even then the capacity was lower than electron microscopy,267 and the domain size of this
ordering increases with the lithium content from 1

to 2 nm in the 550 compound to complete ordering
in Li2MnO3. As noted above, there is apparently
always 8-10% nickel in the lithium layer and a
corresponding amount of lithium in the transition-
metal layer. Grey et al.,268 using NMR studies, have
shown that the lithium in the transition-metal layer
is surrounded by six manganese ions, as in Li2MnO3.
Ceder and Grey thus propose,269 supported by experi-
ments and calculations, that the composition of the
transition-metal layer requires 0.5/6 lithium (∼8%);
the manganese ions in turn are surrounded by nickel
ions on the hexagonal lattice, leading to a 2x3 × 2x3
superlattice. On charging, the lithium is initially
removed from the lithium layer, but when two
adjacent lithium sites become vacant in that layer,
then the lithium ion in the transition-metal layer can
drop down from its octahedral site into the vacant
tetrahedral site. This is consistent with the NMR
observation of the reversible removal of the lithium
ion from the transition-metal layer on charging the Figure 16. Neutron powder diffraction pattern of a
material. This tetrahedral lithium is only removed layered substituted nickelate, LiNi0.4Mn0.4Co0.2O2.
at the highest potential, i.e., only after all the
octahedral lithium is removed. This model is consist- tween 2.5 and 4.2 V at 0.17 mA/cm2 at 30 °C; raising
ent with the presence of the tetrahedral lithium in the charge cutoff potential to 5.0 V increased the
Li0.5Ni0.5Ni0.5O2 proposed by Kobayashi et al.270,271 capacity to over 220 mAh/g, but capacity fade was
In conclusion, the 550 material has the following evident. This compound will be referred to as the 333
cathode characteristics. material.
The synthesis of these LiNi1-y-zMnyCozO2 com-
(1) It has a capacity of around 180 mAh/g for at pounds is typically accomplished using a modified
least 50 cycles under mild cycling conditions. mixed-hydroxide approach by reacting Ni1-y-zMny-
(a) Overcharging increases capacity fade, but Coz(OH)2 with a lithium salt in air or oxygen as
a protective coating might help. described in Liu’s first synthesis272 at 750 °C, which
(b) The synthesis temperature should be in is now known to be below the optimum temperature
excess of 700 °C and less than 1000 °C, of 800-900 °C.219 These conditions result in a single
probably optimally around 900 °C. phase with the layered O3 structure. A typical
(2) There are always nickel ions in the lithium diffraction pattern, as given in Figure 16, conforms
layer, up to 10%, which will restrict the rate to the R3h m symmetry of the R-NaFeO2 structure. The
capability and compromise the energy den- structure consists of a cubic close-packed arrange-
sity. ment of the oxide ions. The transition-metal ions in
(a) Cobalt additions can reduce the level of the structure occupy alternating layers in octahedral
nickel in the lithium layer, as in LiNi1-yCoyO2. sites. The structure and properties of the precursor
(b) The lithium in the transition-metal layer hydroxide compound have been studied;273 it has the
may be a necessary structural component. CdI2 structure like TiS2 but with some turbostratic
(3) The structure of the lithium poor phase is disorder and on heating can convert to a spinel phase.
unclear. The cell parameters219,228,233,236 of this tri-transition-
(4) Nickel is the electrochemically active ion. metal compound are slightly dependent on the tran-
(5) The electronic conductivity needs increasing. sition metals as shown in Figure 17. Both the
in-plane a parameter and the interlayer spacing c
5.2.5. Mixed Nickel−Manganese−Cobalt Dioxide, increases with the Ni content and decreases with the
LiNi1-y-zMnyCozO2 Co content for constant Mn content.219,228 For com-
Consideration of the above leads logically to a positions LiNiyMnyCo1-2yO2, the a and c parameters
mixing of the three transition metals, and the reports obey Vegard’s law, decreasing linearly with increas-
of such compounds were first published in 1999 by ing cobalt content.233 For constant nickel content, the
Liu et al.272 and in 2000 by Yoshio et al.,228 and the a parameter is directly proportional to [Mn] and
latter hypothesized that the addition of cobalt to inversely proportional to [Co], which is indicative of
LiMn1-yNiyO2 would stabilize the structure in a a larger Mn ion compared to the Co ion. This
strictly two-dimensional fashion. They found that the observation is contrary to the suggestion236 that Mn
transition-metal content in the lithium layer fell from does not have any effect on the parameters.
7.2% for LiMn0.2Ni0.8O2 to 2.4% for LiMn0.2Ni0.5Co0.3O2 The ratio c/3a of the lattice constants is a direct
and that the lithium insertion capacities exceeded measure of the deviation of the lattice from a perfect
150 mAh/g for the cobalt-substituted compounds. cubic close-packed lattice, which is it measures the
Ohzuku et al.,215 studying the symmetric compound layeredness of the lattice. An ideal ccp lattice has a
LiNi0.33Mn0.33Co0.33O2, synthesized at 1000 °C, also c/3a ratio of 1.633, whereas a pure layered lattice
found a capacity of around 150 mAh/g cycling be- with no transition metal in the lithium layer has a
the value is to 1.633, the greater the transition-metal

content in the lithium layer; thus, LiNiO2 has a c/3a
ratio of 1.639,213 almost the same as in the spinel
LiNi2O4,274 and that for LiNi0.5Mn0.5O2 is 1.644-1.649
depending on the investigator;213,252,270 addition of a
second lithium as in Li2NiO263 and Li2Ni0.5Mn0.5O2256
changes the c/a ratio very little, 1.648 and 1.647,
respectively. Figure 17 shows that this ratio becomes
much larger as soon as any cobalt is added to the
structure, indicating that cobalt confers layer-like
behavior to the lattice. Whereas the plot for [Mn] )
0.3 shows a closer approach to the cubic ideal value
of 1.633, as the Co content decreases (that is as the
Ni content increases), that for [Ni] ) 0.4 shows little
change. Thus, the c/3a ratio is strongly determined
by the nickel concentration, with the cobalt presence
lessening the amount of nickel in the lithium layer.
Figure 17d shows that the c/3a ratio increases
continuously with the cobalt content when [Ni] )
[Mn]. An analysis of the published data213,270 on
LiNi1-yMnyO2 indicates essentially no change of the
c/3a ratio of 1.644 ( 0.005 with manganese content
for 0.1 e y e 0.5. The mean of all the data for the
composition LiNi0.33Mn0.33Co0.33O2 is 1.657, which is
more layered than the mean 1.647 for the cobalt-free
LiNi0.5Mn0.5O2. For the 333 composition, the c/3a ratio
decreases with increase of formation temperature
from 900 to 1100 °C according to the equation c/3a
) 1.680 - 2.35 × 10-5 T, indicating an increasing
nickel content in the lithium layer with increasing
temperature.
Rietveld refinement was used276 to determine the
distribution of the transition-metal ions in the lay-
ered structure between the 3b and 3a sites; in a fully
ordered layered structure, these sites should be fully
occupied by transition metal and lithium, respec-
tively. The average scattering factor of the transition-
metal ions in the metal layer, 3b sites, was taken to
be equal to that of Co, whereas that in the lithium
layer occupying the interlayer site, 3a, was taken to
be equal to that of Ni. Figure 16 shows the neutron
diffraction pattern of the 442 material, and Rietveld
comparison with the X-ray powder diffraction clearly
showed that the transition metal in the lithium layer
is nickel, not cobalt or manganese. A similar conclu-
sion234 was reached in a recent neutron study on the
333 composition. Figure 18 shows the occupancy on
the Li site (3a site) as a function of overall composi-
tion and synthesis temperature. The data clearly
indicates that the transition-metal disorder is sup-
pressed by increasing cobalt content but not to the
same degree as in LiNi1-yCoyO2, where nickel disor-
der is only observed for y e 0.3, and increased by
increasing the nickel content. However, the synthesis
Figure 17. Cell parameters and c/3a ratio of layered temperature has as profound an effect as composi-
LiNiyMnzCo1-y-zO2, and c/3a ratio of the symmetric tion, as also indicated in Figure 18 for the composi-
LiNi1-yMn1-yCo2yO2. (Parts a-c are reproduced with per- tion LiNi0.4Mn0.4Co0.2O2, where the sample prepared
mission from ref 219, copyright 2004 The Royal Society of at 1000 °C then rapidly cooled to ambient tempera-
Chemistry.)
tures has almost 10% Ni occupancy in the lithium
c/3a ratio of 1.672 for TiS2 and when intercalated layer. Kim et al. also reported234 a high Ni content
increases to 1.793 for LiTiS2.45 The unusually low of 5.9% on the Li site for samples of the 333
value for CoO2 of 1.52124 increases to 1.664 for composition prepared at 950 °C. Only at 800 °C does
LiCoO2. ZrS2 also has an anomalously low value of the nickel disorder drop to zero with increasing cobalt
1.592, which increases to 1.734 in LiZrS2. The closer content. At 900 °C even with more cobalt than nickel
Figure 18. Lattice disorder, percent of lithium sites

occupied by nickel ions in LiNiyMnzCo1-y-zO2. Materials
synthesized at 800 °C for Whittingham219 and 900 °C for
Oh,278 Dahn,216,236 and Kim234 unless otherwise stated.
in the material there is still considerable nickel

disorder, almost 2% more Ni in the lithium layer at
900 °C than at 800 °C for all compositions. Clearly
high temperature increases the disorder of the nickel
ions just as earlier observed for TiS2; this effect may
be reduced by a slow cooling of the sample in an
oxidizing environment or a hold at 800 °C or slightly
lower if the original synthesis is carried out at higher
temperatures. This will allow the partial reordering
of the ions.235
Although these materials show good electrochemi-
cal behavior, their electronic conductivity is still low
for a high-rate cathode, and a means needs to be
found to increase the conductivity without the addi-
tion of excessive amounts of a conductor such as Figure 19. Magnetic behavior of LiNi0.5Mn0.5O2, LiNi0.4-
carbon black, which will reduce the volumetric energy Mn0.4Co0.2O2, and LiNi0.33Mn0.33Co0.33O2: (a) susceptibility
storage capacity. There has also been a report237 of and (b) field effect (after Ma et al.).276
the low packing density of powders which will also active redox role only in the later stages of lithium
severely reduce the volumetric energy density. The removal. The manganese is merely a spectator ion
conductivity of LiNi0.5Mn0.5O2 was 6.2 × 10-5 S/cm; but a critical one as at a minimum it reduces the cost
this increased on cobalt addition to 1.4 × 10-4 S/cm of the cathode. Studies of the magnetic behavior of
for LiNi0.4Mn0.4Co0.2O2 at 21 °C and 6.8 × 10-4 S/cm the compounds gives information about the location
at 100 °C.219 The value of the cobalt-free compound of the nickel ions, but studies of the magnetic
is similar to that observed previously241 for KMnO2 moment from Curie-Weiss behavior do not give
and LiMnO2, but in that case 2-10% Co substitution much key information as the combination Ni3+ +
led to a 100-fold increase in the conductivity to Mn3+ + Co3+ gives an almost identical moment to the
around 10-3 S/cm. This rather low effect of cobalt combination Ni2+ + Mn4+ + Co3+.219 However, the
substitution is not totally unexpected as cobalt presence of nickel ions in the lithium layer results
substitution in LiNiO2, as in LiNi0.8Co0.2O2, reduces in a hysteresis loop in the magnetic moment both
the conductivity.169 Sun et al.233 reported conductivity for the mixed material Li(NiMnCo)O2 and in
values of 2-5 × 10-4 S/cm essentially independent LiNi1-yAlyO2.275 The magnetic behavior of several of
of composition for cobalt contents up to 0.5, which these phases is shown in Figure 19 and shows that
suggests that changes in the electrochemical behavior at the higher temperatures they obey the Curie-
with composition is not a function of changes in Weiss law; as the addition of cobalt increases from
conductivity. However, they found233 that increasing 0.0 to 0.2 to 0.33, the hysteresis loop decreases,
the cobalt content increased the rate capability, indicating a reduction in the Ni2+ content in the
which might be associated with the lack of pinning lithium layer.276
Ni2+ ions in the lithium layers, which would reduce XPS studies have been made on a number of
the diffusivity of the lithium. compositions of these transition-metal oxides, and all
There have been a number of studies of the indicate predominantly divalent nickel. Thus, for the
physical and bonding behavior of these mixed transi- 442 compound the Co spectrum is clearly Co3+, and
tion-metal oxides, which conclude that in the fully the Mn spectrum can be assigned to 80% Mn4+ with
lithiated compounds the cobalt is trivalent, the nickel 20% Mn3+. The Ni spectrum is characterized by an
predominantly divalent, and the manganese tetra- intense and complicated satellite structure and
valent. Thus, the electrochemically active species is consistent with 80% Ni2+ and 20 Ni3+. Studies
predominantly nickel with the cobalt playing an on LiNi0.33Mn0.33Co0.33O2,224 LiNi0.5Mn0.5O2,277 and
within the potential window of 2.8-4.4 V; this

capacity dropped slightly to 170, 165, and 162 mAh/g
as the current density increased to 40, 80, and 160
mA/g (1.6 mA/cm2 or C rate). The Dahn group has
found similar good cycling ability for the close com-
position LiNi0.375Mn0.375Co0.25O2, where one sample216
with 5.5% Ni on the lithium sites showed at 30 °C a
capacity of 160 mAh/g dropping to 140 mAh/g after
50 cycles at a rate of 40 mA/g; raising the tempera-
ture to 55 °C increased the capacity to 170 mAh/g,
and it dropped only to 160 mAh/g after 50 cycles. A
second sample236 which had only 3.2% Ni on the Li
site had a lower capacity of 135-130 mAh/g over 50
cycles at 30 mA/g due to only being charged to 4.2 V;
increasing the charging to 4.4 V increases the capac-
ity by 20-30 mAh/g. This shows the critical effect of
charging potential. As noted for the LiCoO2 cathode
above, without doubt an appropriate coating would
allow the charging potential and hence the capacity
to be increased. It is not clear yet how critical the
nickel level in the lithium layer is as most studies
have been made on compounds with 2-5% nickel
levels with no obvious difference in behavior. This is
an area that merits further evaluation and under-
standing.
There have been the largest number of studies on
the 333 composition, which have been made by a
number of synthetic techniques over a wide range of
temperatures. Most of these studies show similar
behavior with the capacity increasing with increasing
charging potential. Increase of synthesis temperature
Figure 20. Electrochemical behavior of LiNi0.4Mn0.4- from 800 to 900 °C increases the initial capacity from
Co0.2O2: (a) capacity as a function of a cycle at 1 and 2
mA/cm2 (reproduced with permission from ref 219, copy-
173 to 190 mAh/g and the capacity after 16 cycles to
right 2004 The Royal Society of Chemistry) and (b) capacity 180 from 160 at the 0.3C rate in the potential window
as a function of discharge rate, average capacity of first 3.0-4.5 V.226 Ohzuku similarly reported215,231 capaci-
five cycles. Cutoff on charge, 4.3 V at 22 °C Ngala,219 4.4 V ties of 150 mAh/g for a 4.2 V cutoff and 200 mAh/g
at 30 °C Oh,278 4.4 V at 30 °C Lu,216 and 4.2 V at 30 °C for 4.6 and 5.0 V charging; others have reported224,225
MacNeil.236 similar increases in capacity on increasing the charg-
LiNiyCo1-2yMnyO2,216 with y ) 1/4 and 3/8, also ing potential, but in some cases there is marked
suggested 2+ and 4+ as the predominant oxidation capacity fade.234 Spray drying the powder has been
states for nickel and manganese, respectively, which reported279 to increase the capacity, 195 mAh/g at 0.2
then designates the nickel as the key electrochemi- mA/cm2 (20 mA/g) within the potential window of
cally active species. 3-4.5 V. Replacing one-half of the cobalt ions by iron
The electrochemical behavior of a number of dif- in the 333 composition with the goal of increasing
ferent compositions over a range of current densities the cell capacity below 4.5 V resulted in a higher cell
have been studied, and two of these studies identi- polarization, lower capacity, and increased capacity
fied198,219,220,278 the 442 material as having the highest fade.280
capacity and maintaining its capacity on cycling. The structural changes accompanying the removal
Pure Li1-yNi1+yO2 had the lowest capacity. The of Li from a number of these materials have been
electrochemical behavior of a sample of the 442 investigated. In the case of LiNi0.4Mn0.4Co0.2O2, the
composition,220 synthesized at 900 °C, at 22 °C is change in cell volume is less than 2%, much less than
shown in Figure 20, for current densities of 1 and 2 the 5% reported for LixNi0.75Co0.25O2281 and LixTiS2,40
mA/cm2 within the potential window 2.5-4.3 V; these making this compound much less liable to mechani-
are equivalent to rates of 44 and 104 mA/g. All the cally fracture on cycling. The 333 compound also
samples had an initial potential around 3.8 V. The shows a volume change of less than 2% as the lithium
temperature of synthesis was found to be important is removed.234 This small change of volume is associ-
with the optimum capacity and capacity retention ated with compensating changes in the a and c
being found at 800-900 °C, with a much lower parameters. When the c parameter increases, the a
capacity being found for samples prepared at 1000 parameter contracts and vice versa. These results are
°C.220 The rate capability of the 442 composition is shown in Figure 21 for the 442 compound,276 where
also shown in Figure 20 and shows the excellent the X-ray pattern was forced to fit the simple
reproducibility between the different research groups. hexagonal lattice of the LiMO2. In reality the situa-
Oh et al.278 found a constant capacity for 442 of 175 tion is probably more complex when x is less than
mAh/g at 0.2 mA/cm2 (20 mA/g or C/8) for 30 cycles around 0.3, and a more in-depth interpretation is
ditional nickel ions into the interslab region. Thus,

every effort should be made to prevent formation of
the 1T structure in nickel-rich compounds to avoid
capacity fade on deep charging. In the case of MnO2
itself, the manganese ions migrate at high potentials
to form a spinel-like phase242 and do not migrate back
when lithium is re-intercalated.
Inspecting Figure 21 shows that as the electron
density on the oxygen ions decreases, their effective
size also decreases, leading to the smaller a param-
eter (oxygen diameter). Reducing the lithium content
causes the c parameter to increase, presumably as
the electrostatic attractive forces between the nega-
tive layers and the positive lithium ions decrease.
Although a cursory study of the X-ray diffraction
patterns suggests a continuous change in the struc-
ture as the lithium is removed, more in-depth studies
suggest otherwise. Thus, one report217 proposes that
four hexagonal phases are formed as the lithium is
removed from the 333 compound, and another229
suggests that a monoclinic phase is formed. Clearly
additional studies are required ideally on single
crystals if they can be synthesized, much as was
accomplished with V6O13.
Are these cathodes ready and able to replace the
stabilized LiCoO2 used in most Li-Ion batteries
today? A recent joint academic/industrial study283
suggests that the 333 compound is at least equal to
or superior to the LiCoO2 cathode. In a prismatic cell
configuration, it showed a constant capacity of 600
mAh over 30 cycles at C rate. The cell can also be
charged to a higher potential cutoff without the
capacity loss associated with CoO2 structure changes
(O3 to 1Tsccp to hcp) in LiCoO2, which results in a
10-20 mAh/g higher storage capability. An issue
with all these layered oxides is their inherent ther-
modynamic instability upon lithium removal. Al-
though MnO2 is stable in air at room temperature,
neither CoO2 or NiO2 is, both having effective oxygen
partial pressures in excess of 1 atm. The stable
oxidation states of Mn, Co, and Ni in their simple
binary oxides are 4, 2, and 2, respectively; on heating
to over 500 °C, Mn2O3 becomes the stable oxide
followed by Mn3O4 at still higher temperatures. Thus,
Figure 21. Lattice parameters of LixNi0.4Mn0.4Co0.2O2: (a) their kinetic stability can create problems if there is
a and c in Å and (b) volume in Å3 (after Ma et al.).276 any thermal excursion in the cell. For the four oxides
LixNi1.02O2, LixNi0.89Al0.16O2, LixNi0.70Co0.15O2, and
underway. At the lowest lithium content Li0.05Ni0.4- LixNi0.90Mn0.10O2, a structural transformation first to
Mn0.4Co0.2O2 there is a trace amount of the one-block a spinel phase and then to a rock-salt phase was
structure (1T) of TiS2 or CoO2, in contrast to LiCoO2 found284,285 for lithium x values of 0.5 or less. The
itself where large amounts of this structure would second transformation is accompanied by a loss of
be seen. Thus, it appears that the 2-3% Ni on the oxygen, and the first may be depending on the
Li sites is impeding the structural reorganization; composition but usually when x is less than 0.5; the
this might be very advantageous for capacity reten- latter oxygen release occurs at lower temperature as
tion on deep cycling of the material and speaks the lithium content decreases and as low as 190 °C
toward leaving some nickel disorder in the structure. for Li0.3Ni1.02O2. The stability is improved on alumi-
This is consistent with an earlier study282 on the NiO2 num or cobalt substitution. The compound Li0.1NiO2
phases, which found that the 1T phase is formed for is reported286 to lose weight at 200 °C forming a rock-
Ni1.02O2 but not for nickel contents above 1.07, salt structure. The substitution of manganese for
suggesting that the NiO2 slabs are pinned by the nickel appears to move the transition to the spinel
extra nickel and thus cannot slide to form the 1T to higher temperatures; thus, Li0.5Ni0.5Mn0.5O2 even
form. That study also found that the 1T phase slowly after 3 days at 200 °C is still layered,287 but a spinel
and irreversibly transformed at high potentials into phase is formed above 400 °C and is stable to much
a rhombohedral phase due to the migration of ad- higher temperatures for the 1:1 Ni:Mn lithium-free
compound, eventually giving a mixture of spinel and or Co. The transition-metal cation can also be nickel
nickel oxide in air and NiO + Mn3O4 in nitrogen.288 or manganese including mixtures such as LiNi1-yCoyO2
The compounds Li0.5Ni0.4Mn0.4Co0.2O2 and Li0.5Ni0.33- as pointed out by Yoshio,228 Thackeray,289,290 and
Mn0.33Co0.33O2 both begin to lose weight above 300 Dahn,252 and the Li2MnO3 can be replaced by related
°C with major weight loss, 7-8%, only above 450 materials such as Li2TiO3 and Li2ZrO3. Li2MnO3 can
°C,276 which corresponds to reduction of Co(III) to be represented in the normal layered notation as
Co(II) and any Ni(IV) to (Ni(II); the manganese Li[Li1/3Mn2/3]O2. These solid solutions can thus be
remains Mn(IV), and the structure begins to change represented as LiM1-y[Li1/3Mn2/3]yO2, where M can be,
to spinel by 350 °C, and the spinel phase is still for example, Cr, Mn, Fe, Co, Ni, or mixtures thereof.
present at 600 °C. Addition of extra lithium will tend to push the
In conclusion the solid solutions of the 550 material manganese away from trivalent to tetravalent, thus
and LiCoO2 have the following cathode characteris- minimizing the impact of any Jahn-Teller distortion
tics. coming from Mn3+.
(1) They have a capacity of around 170 mAh/g Of particular curiosity is the end-member Li2MnO3
for at least 50 cycles under mild cycling which has been shown291,292 to exhibit unexpected
conditions to 4.4 V and over 150 mAh/g at 2 electrochemical activity on charging as the manga-
mA/cm2. nese is already in the 4+ oxidation state. This
(a) A higher charge potential cutoff increases “overcharging” can be associated with two phenom-
capacity. ena, removal of lithium with the concomitant loss of
(b) The synthesis temperature should be in oxygen giving a defective oxygen lattice and the
excess of 700 °C and less than 1000 °C, removal of lithium by decomposition of the electrolyte
probably optimally around 900 °C. giving protons which can ion exchange for the lithium.
(2) The cobalt reduces the number of nickel ions Which mechanism predominates depends on the
in the lithium layer. temperature and chemical composition of the oxide
(a) The final heating temperature needs to be lattice.293 In both cases the manganese oxidation
no higher than 800 °C. state remains unchanged. When significant amounts
(i) The ratio of Co/Ni needs to be greater of hydrogen are ion exchanged, the MO2 slabs slide
than 1 to eliminate all Ni in the lithium to give prismatic coordination between the layers as
layer. this, combined with a contraction in the interlayer
(b) For a final anneal temperature of 900 °C, spacing of around 0.3 Å, leads to optimum hydrogen
there will always be nickel ions in the bonding.292 These protons are lost as water on heat-
lithium layer. ing the oxide to around 150 °C.292 Acid leaching of
(c) A certain level of nickel ions will deter the Li2MnO3 also results in the removal of lithium, and
formation of the one-block structure at low here again both mechanisms of Li2O removal294,295
lithium concentrations. and proton exchange296 are believed to be operative.
(i) Minimization of formation of the one- Acid leaching of the lithium stoichiometric com-
block structure on charging will help main- pounds, such as LiNi0.4Mn0.4Co0.2O2, also results in
tain capacity on cycling. removal of lithium and a small amount of proton
(ii) There is a need for a determination of exchange.276
the acceptable or desirable level of nickel The ease of oxygen removal from the close-packed
in the lithium layer, as it may well not be lattice when lithium rich had been demonstrated297
zero. by its ready reduction by ammonia gas at 200 °C in
(3) There is probably not a single phase for all the case of the spinel Li[Li1/3Mn5/3]O4. It was also
lithium values from 0 to 1 in Lix(NiMnCo)O2. shown297 that this oxygen could be removed by
(a) The structure needs determining at low x electrochemically charging above around 4.3 V; the
values. material then showed the 4 V discharge behavior
(i) The 442 compound only forms small typical of a spinel. These reduced materials can best
amounts of the one-block structure by x ) be represented as Li[Li1/3Mn5/3]O4-δ.
0.05. Dahn et al. studied252 the solid solution Li2MnO3-
(4) Nickel is the electrochemically active ion at LiNiO2, which can be written as Li[NiyLi(1/3-2y/3)-
low potentials. Mn(2/3-y/3)]O2 ) yLiNiO2 + (1-y)Li[Li1/3Mn2/3]O2. The
(a) Cobalt is only active at the higher poten- a and c hexagonal lattice parameters increased
tials. linearly with the nickel content, y, from 0.08 to 0.5
(5) The electronic conductivity needs increas- with the c/3a ratio decreasing linearly, showing that
ing. the layeredness decreased with increasing nickel
(6) The optimum composition is still to be content as expected. Cells of these materials showed
determined for energy storage, power capabil- an irreversible charging plateau around 4.5 V, which
ity, lifetime, and cost considerations. is believed to be due to loss of oxygen as described
above. Before the plateau all the nickel is oxidized
5.2.6. Lithium-Rich Mixed-Metal Dioxides, Li1+xM1-xO2 to Ni4+. After this “overcharging” the electrodes were
As discussed earlier in the case of the chro- found to cycle well at 30 °C between 2.0 and 4.6 V
mium243,244 and cobalt251 systems, excess lithium can with the capacity increasing inversely with the value
be incorporated into the layered structure through a of y: 160, 180, and 200 mAh/g, respectively, for y )
solid solution of Li2MnO3 and LMO2, where M ) Cr 1/2, 5/12, and 1/3 (this trend has been confirmed298).
The first has no excess lithium and the lowest olivine phase, in particular LiFePO4, by Padhi et
capacity, showing the advantage of excess lithium. al.309 This is the first cathode material with poten-
The capacity of the y ) 1/3 material increased to 220 tially low cost and plentiful elements and also
mAh/g when the cycling temperature was increased environmentally benign that could have a major
to 55 °C. However, this high charging level results impact in electrochemical energy storage. For LiFe-
in a lower thermal stability of the material.252 The PO4, the discharge potential is about 3.4 V vs lithium
addition of excess lithium to the 550 composition, and no obvious capacity fading was observed even
Li1+x(Ni0.5Mn0.5)1-xO2 was reported299 to increase the after several hundred cycles. Its capacity approaches
stability of the material. The system Li2MnO3- 170 Ah/kg, higher than that obtained by LiCoO2 and
LiNiO2-LiMnO2 has been studied,300 and it shows comparable to stabilized LiNiO2, and moreover, it is
complete solubility along the Li2MnO3-LiNiO2 line; very stable during discharge/recharge. Since its
the electrochemical capacity was found to fall rapidly discovery, many research groups have tried to im-
as the nickel content decreased when the charging prove the performance of this material.49,50,198,315-326
potential was limited to 4.3 V. LiFePO4 can be synthesized by high-temperature
Thackeray et al.301 showed that Li2TiO3 forms a reactions,309 under hydrothermal conditions,50 or by
solid solution with LiNi0.5Mn0.5O2 and that the tita- sol-gel methods.327 Although the olivine phase can
nium helped allow the intercalation of a second be very easily synthesized hydrothermally within just
lithium into the structure.64 The advantages of add- a few minutes and its X-ray pattern looks good, it
ing Li2MnO3 to the layered cathode material have gives poor electrochemical properties; a close exami-
even been reported for manganese-rich materials, nation of the structure showed that there are about
which would be unstable otherwise relative to spinel 7% iron atoms in the lithium site, and this is reflected
formation. Thus, Li[Li0.2Ni0.2Mn0.6]O2 showed a steady- in the lattice parameters of a ) 10.381 Å, b ) 6.013
state capacity of around 200 mAh/g between 2.0 and Å, and c ) 4.716 Å compared with those for ordered
4.6 V at 0.1 mA/cm2, after gaining capacity for the LiFePO4 of a ) 10.333 Å, b ) 6.011 Å, and c ) 4.696
first 10 cycles.302 The behavior and stability of this Å.198 These iron atoms essentially block diffusion of
material at higher rates was not reported. Addition the lithium ions, as the diffusion is fast only along
of some cobalt to these manganese-rich compounds the tunnel and not between them;51 thus, it will be
was reported to help retain the capacity at higher critical in the use of this material to ensure ordering
discharge rates.222 Magnesium has also been proof the lithium and iron atoms. Firing the hydrother-
posed as a stabilizing agent for manganese-rich mal material to 700 °C resolved the disorder. Recent
materials.303 studies suggest that the hydrothermal approach can
Thus, there is one more parameter, the lithium be improved by modifying the synthesis conditions,
excess content, to be considered in addition to the for example, by adding a reducing agent such as
nickel, cobalt, and manganese ratios in designing the ascorbic acid328 to prevent surface ferric films; the
optimum composition for the ideal cathode. Each of hydrothermal method can also produce material with
these elements has its own role to play, whether it excellent electrochemical behavior even without a
is stabilizing the lattice (the manganese), being the carbon coating,329 as is necessary in most instances
electrochemically active member (the nickel), order- as discussed below.
ing the transition metals and perhaps increasing the
rate capability and the conductivity (the cobalt), or As this material has a very low conductivity at
increasing the capacity (the lithium). Whether other room temperature, it could achieve the theoretical
elements will play a critical role is yet to be deter- capacity only at a very low current density315 or at
mined, but some elements such as titanium have elevated temperatures,317 as suggested by Padhi309
been found to decrease the rate capability and due to the low lithium diffusion at the interface.
migrate to the lithium layer. Ravet et al.318 showed that a carbon coating signifi-
cantly improves the electrochemical performance of
5.3. Iron Compounds Including Oxides and this material; sucrose was proposed319 as one carbon
Phosphates precursor, and it was used on the initial hydrother-
mal samples.50 Many other studies have been made
A number of researchers,173,304,305 particularly in on finding means to improve the electronic conduc-
Japan, have been pursuing the oxides of iron as tivity of the LiFePO4 particles.49,198,320,321,323-326 Very
potential cathode materials for lithium cells. How- pure LiFePO4 samples are reported to have an
ever, materials of the type LiFeO2 have shown little electronic conductivity of 10-9 S/cm,324 whereas
ability for lithium removal. A number of other iron samples made from reagent-grade carbon-containing
compounds have been studied over the years, includ- materials have a conductivity of around 10-5-10-6
ing FeOCl,306 FePS3,307 KFeS2,308 and FeS2,13 but none S/cm.198 Huang et al.320 proposed coating the material
showed much reversibility. Although metal phos- with carbon-gel during the synthesis step and found
phates have been studied for more than 20 years capacities approaching 100% at very low cathode
since the discovery of fast ion transport in NASICON, loadings, 5 mg/cm2, and rather high carbon contents,
it is only recently that they have been considered as 20%. They obtained 800 cycles at around 120 mAh/g
cathodes309-312 or anodes312-314 of lithium batteries. at high rates. Masquelier proposed321 extensive mill-
ing of the material with carbon and then found high
5.3.1. Olivine Phase capacities at elevated temperatures. A 2002 paper by
Emphasis changed radically in 1997 with the Chang et al.324 showed excellent electrochemical
discovery of the electrochemical properties of the behavior when the LiFePO4 was “doped” with parts
Figure 22. Structures of orthorhombic LiFePO4 and trigonal quartz-like FePO4.
per million of elements such as niobium, which

increased the conductivity by 8 orders of magnitude.
This paper caused a surge of interest in this com-
pound and in an understanding of what causes the
conductivity changes; a recent paper by Nazar et
al.325 has shown that the conductivity increase is
related to the formation of a highly conductive iron
phosphide, Fe2P, surface film; this film is formed at
high temperature, particularly in the presence of a
reducing agent such as carbon.
The olivine structure is shown in Figure 22, and
on lithium removal the phase FePO4 is formed;
LiFePO4 and FePO4 have essentially the same struc-
ture. This FePO4 is isostructural with heterosite,
Fe0.65Mn0.35PO4. This is a two-phase system with
LiFePO4 being in equilibrium with FePO4 as shown
in the cycling plot in Figure 23a.198 This figure shows
that 100% capacity of the lithium is cyclable at 1 mA/
cm2 at 60 °C even at electrode loadings of 80 mg/cm2;
at room temperature, around 70% can be cycled at 1
mA/cm2 and 100% at 0.1 mA/cm2. Even at 10 mA/
cm2, almost 70% of the capacity is obtained at low
loadings, as shown in Figure 23b.
Critical to the use of LiFePO4 is its reactivity and
thermal stability and that of its charged product,
FePO4. Thomas reported317 that there were no ther-
mal excursions observed in cells or in a DSC experi-
ment in the range 25-85 °C. However, the olivine
structure is inherently unstable because of the edge
sharing between octahedra and tetrahedra, and
under pressure it converts to the spinel phase as
observed in the earth’s mantle. Recently, the trans-
formation of orthorhombic LiFePO4 to an olivine-like
LiFePO4 with the lithium in tetrahedral sites has
been reported;330 this form is electrochemically inac-
tive. In addition, more than two crystalline forms of
FePO4 are known: the orthorhombic form isostruc-
tural with LiFePO4 where the iron is found in FeO6 Figure 23. Electrochemical behavior of LiFePO4: (a)
octahedra and the trigonal form in which the iron is cycling at 1 mA/cm2 at 21 and 60 °C (reprinted with
found in FeO4 tetrahedra. These are shown in Figure permission from ref 198, copyright 2003 Elsevier) and (b)
22. The orthorhombic structure has been discussed Ragone plot for LiFePO4 from four groups, Chen332 (8 mg/
extensively.309 The trigonal form is composed of FeO4 cm2 loading and about 10 wt % carbon), Ngala219 (20-80
and PO4 tetrahedra, each FeO4 tetrahedron shares mg/cm2 and 10% carbon), Zane363 (10 mg/cm2 and 20%
carbon), and Huang320 (5 mg/cm2 and 20% carbon).
its four corners with four PO4 tetrahedra and vice
versa, giving a quartz-like structure. The all-tetra- over, surface glassy films tend to form at high
hedral form is electrochemically unreactive as Fe(II) temperatures.49 Thus, care must be taken to ensure
is not stable in tetrahedral configuration, and more- that it is not formed during the synthesis of LiFePO4
Table 1. Chemical Reactivity of Iron Phosphates49 gelatin,333 did not appear to be important. However,
BuLi the temperature of firing the reaction mixture ap-
reactivity pears to be much more important as that determines
compound synthesis method (mol/Fe) the amount of graphitic compound on the LiFePO4
LiFePO4 high temperature 1.85 surface; sp2 carbon was found to be much more
LiFePO4 hydrothermal 0.29 effective than sp3 carbon.334 The particle size is
FePO4 orthorhombic 3.25 almost certainly controlled by the carbon in the
FePO4 trigonal (700 °C-tetrahedral iron) 2.90
reaction mixture, even when only coming from carbon
FePO4‚2H2O amorphous phase 7.20
LiFePO4(OH) hydrothermal 3.24 in the reagent materials, leading to the positive
behavior of several samples.324 The temperature of
preparation was found to be important, with that
or its precursors. At low temperatures, when lithium prepared at 675 °C showing the best behavior in one
ions are absent, amorphous FePO4‚2H2O can be study.335
formed (see below); although it is electrochemically Much of the above discussion has centered on the
active, dehydration at high temperatures leads to the iron phase of the LiMPO4 olivine structure, but the
tetrahedral form of FePO4, which is electrochemically transition metal may also be manganese, nickel, and
inactive.331 cobalt. The naturally occurring form contains both
It is essential to understand the long-term stability iron and manganese. None of these other forms has
of electrode materials to ensure the extended life of yet showed superior electrochemically behavior to the
any battery. Thus, it is important to better under- iron compound, even though they have higher dis-
stand the reactivity of both the LiFePO4 and the charge potentials.336 The mixed Fe-Mn compound
FePO4 orthorhombic phases. For example, what discharges in two distinct steps, which can be associ-
happens when LiFePO4 is over discharged or does ated, respectively, with the iron and manganese
the orthorhombic FePO4 slowly switch to the quartz redox reactions. Several research groups have studied
form over time? When iron phosphates are reacted the cycling behavior of the pure LiMnPO4 com-
with an excess of n-butyllithium it was reported49 pound.337,338 The results are controversial: Yamada
that all but one reacted with sufficient butyllithium et al.337 related the inability to extract lithium from
to form lithium phosphate and iron as shown in Table LiMnPO4 electrochemically to the thermodynamic
1. Thus, it appears that these phosphate lattices are instability of olivine-type MnPO4 due to the Jahn-
destroyed at low lithium potentials; n-butyllithium Teller effect of Mn3+; Li et al.338 reported a reversible
is about 1 V versus pure lithium. An electrochemical capacity of about 140 mAh/g for the reaction between
study49 where the LiFePO4 cathode was subjected to LiMnPO4 and MnPO4. Recently, Delacourt et al.339
a discharge down to 1.0 V at 0.4 mA/cm2 confirmed studied the LiMnPO4 formed by direct precipitation
that lithium reacts with destruction of the LiFePO4 and obtained a reversible capacity of 70 mAh/g after
lattice and considerable loss of capacity. After five carbon coating by ball milling the materials with
cycles the capacity had dropped by 80%. Thus, carbon black. This suggests that MnPO4 is thermo-
lithium iron phosphate cells will require overdis- dynamically stable. Song et al.340 studied LiMnPO4,
charge protection in commercial applications. On the formed from the thermal decomposition of LiMnPO4-
other hand, no evidence has been found for the (OH), and found the Mn3+/Mn2+ transformation at
conversion of the metastable orthorhombic FePO4 about 4.1 V with a high polarization and low capacity;
phase to the quartz-like trigonal phase under normal heating with carbon black did not result in any
electrochemical conditions. Moreover, unlike the dramatic enhancement of the capacity. Ceder et al.,341
layered oxides described earlier, there is no tendency using an advanced theoretical treatment, have cal-
to lose oxygen on lithium removal from the lattice. culated the open-circuit voltages and the band gap
The low density of LiFePO4 results in a low in these phosphates, and found that as expected they
volumetric density, and so it is critical that the are very high. The open-circuit voltages are 3.5 V for
minimum volume of carbon and organic binder be LiFePO4, 4.1 V for LiMnPO4, 4.8 V for LiCoPO4, and
used in the electrode. Taking the density of LiFePO4, 5.1 V for LiNiPO4, thus explaining the lack of
Teflon, and carbon black as 3.6, 2.2, and 1.8 g/cm2, electrochemical activity for LiNiPO4 within the nor-
respectively, then an electrode containing 10 wt % mal cycling potential range. The very high calculated
carbon and 5 wt % Teflon will have a volumetric band gaps of 3.7 and 3.8 eV for LiFePO4 and Li-
energy density 25% less than the theoretical value. MnPO4 are consistent with their color and diffuse
This assumes that all the particles pack equally reflectance spectra and suggest that band-gap dif-
efficiently, which is unlikely to be the case with the ferences do not explain the different electrochemical
poor packing associated with carbon, particularly behavior; the electronic conductivity is likely due to
with decomposed sugar. A tap density study sug- a polaron mechanism.342 Some other recent theoreti-
gested that this carbon packs poorly, so the volumet- cal calculations343,344 suggesting that these materials
ric density penalty may be very significant.332 A are semi-metals are inconsistent with their white
number of carbons have been studied to determine color and are almost certainly in error due to prob-
how much carbon is needed for optimum electro- lems with the theory used.
chemical behavior. For carbon black little difference
was observed for carbon loadings from 6 to 15 wt %, 5.3.2. Other Iron Phosphate Phases
except that the polarization observed was slightly
higher at 6 wt %; the method of carbon addition, Several other iron phosphate structures have been
whether carbon black, carbon gel, sugar, or aqueous described. The phase Li3Fe2(PO4)3 has been stud-
ied,50,311,312 and it discharges below 3 V and carries

more phosphate deadweight than LiFePO4, so has
been of less interest since the discovery of LiFePO4.
Three groups345-347 have studied the electrochemical
behavior of the hydrated and anhydrous phase
FePO4‚nH2O and found essentially identical results.
Amorphous and crystalline FePO4‚nH2O were pre-
pared hydrothermally and then heat-treated over a
range of temperatures. The material obtained from
the amorphous dihydrate shows more than double
the capacity of the material obtained from the
crystalline dihydrate, 0.75 vs 0.3 Li/FePO4, respec-
tively. This might be associated with the amorphous
nature of the former relative to the more crystalline
structure of the latter; also, the latter has iron in an
essentially tetrahedral environment. The capacity of
the former is similar to that reported by Prosini et
al.348 for a biphasic mixture of phosphate and iron
oxide. Both the amorphous and crystalline materials
show behavior more typical of a single-phase reac-
tion, LixFePO4, as opposed to the two-phase behavior
of the LiFePO4-FePO4 system. The discharge curves
of amorphous FePO4 obtained at different tempera-
tures at 0.2 mA/cm2, between 4 and 2 V, are shown
in Figure 24 as are also the cycling curves and the
rate capability of the amorphous FePO4 sample
heated at 350 °C at different current densities, Ic )
I d.
Another class of iron phosphates is that related to
the minerals Giniite and Lipscombite. The structure
in this case consists of rods containing face-sharing
FeO6 octahedra, stacked orthogonally to one another,
giving nonintersecting tunnels through which lithium
ions can diffuse. The discharge potentials in this case
are sloping and lower than those of the LiFePO4/
FePO4 system.349,350
5.3.3. Vanadium Phosphate Phases

A number of vanadium phosphates have also been
studied as potential cathodes, including those of
general formula Li3V2(PO4)3351-356 and VOPO4.357
Li3V2(PO4)3 exists in two forms: the thermodynami-
cally stable monoclinic form and a rhombohedral
form that can be formed by ion exchange from the
stable sodium analogue with the NASICON struc-
ture. Two lithium ions can be removed from the Figure 24. Electrochemical behavior of FePO4‚nH2O
rhombohedral form at 3.77 V, and only 1.3 can be formed by the thermal decomposition of FePO4‚nH2O: (a)
reinserted.353 The monoclinic form is of more interest discharge of FePO4 as a function of annealing temperature,
as a cathode as all three lithium ions can be readily (b) cycling of FePO4 annealed at 350 °C, and (c) rate
removed and reversibly intercalated at high rate.351,352 capability of FePO4 annealed at 350 °C (reprinted with
However, the electrochemistry of this cathode is permission from refs 49 and 345, copyright 2002 Elsevier).
complex, showing a series of steps on charging but a
solid solution on lithium insertion from 0 to 2 lithium In the latter, after the H2VOPO4 is electrochemically
followed by two-phase behavior.355 This cathode is oxidized in a LiPF6/EC-DMC solution to give -
being commercialized by Valence Technology. VOPO4, the -VOPO4 can be reduced by reversing the
The -VOPO4 compound has particularly interest- current flow to give first LiVOPO4 and then Li2VOPO4,
ing properties,358 with an approximately 4 V flat as shown in Figure 25a.340 As expected from the ease
discharge potential some 0.5 V higher than LiFePO4 of these reactions, the structures of all four com-
and with higher electronic conductivity, leading to pounds are closely related, as shown in Figure 25b;360
the possibility of attaining higher power systems the building block for all these structures, which
than for LiFePO4 but at the expense of higher cost. consists of VO6 octahedra and PO4 tetrahedra, is
The -VOPO4 can be synthesized by removal of the shown at the lower right.
hydrogen atoms from VPO4‚2H2O(dH2VOPO4) either If a still higher redox potential is desired, then
thermally359 or by electrochemical deintercalation.340 heating -VOPO4 with LiF and carbon black at 550
will only be achieved if the costs of the other cell

components are also reduced. These include the
electrolyte, a lower cost anode with added safety
features including the use of natural graphite, and
probably the use of thicker cathodes to reduce the
cost of electrode support materials, separators, etc.
It can be anticipated that totally new materials with
unexpected properties will be discovered, so that the
goal of at least one lithium cycling per transition
metal at a rate of 10 mA/cm2 can be achieved with
pulse charging and discharging at still higher rates.
7. Abbreviations and Specialized Terms

anode electropositive electrode
cathode electronegative electrode
C rate measure of the time for cell discharge in
reciprocal hours, time of discharge ) 1/C
in hours
HEV hybrid electric vehicle
LiBOB lithium bis(oxalato)borate LiB(C2O4)2
LiPF6 lithium hexafluorophosphate, LiPF6
NASICON sodium superionic conductor
Ragone plot a plot of the electrochemical cell capacity as
a function of the magnitude of the dis-
charge or charge current
Vegard’s for a solid solution the lattice parameter
Law varies linearly with composition
8. Acknowledgments
This work was supported by the US Department
of Energy, Office of FreedomCAR and Vehicle Tech-
nologies, through the BATT program at Lawrence
Berkeley National Laboratory and by the National
Science Foundation through grant DMR0313963. I
Figure 25. (top) Electrochemical intercalation of lithium thank Drs. Marca Doeff, Kang Xu, and Michael
into -VOPO4,340 and (bottom) relationship between the Thackeray for many constructive suggestions for
various structures in the VOPO4 system; the building block improving the first draft of this manuscript. I also
for all these structures is shown at the lower right. thank Professor T. Ohzuku for providing Figure 14
and Miaomiao Ma, Natasha Chernova and Peter
°C for 15 min and then rapidly cooling to room Zavalij for help and much data.
temperature leads to the formation of LiVPO4F.361
This compound is isostructural with LiMPO4(OH) (M 9. References
) Fe, Mn). It has a potential of 4.2 V and a capacity
(1) Watanabe, N.; Fukuba, M. U.S. Patent 3,536,532, 1970.
of 0.55 Li per formula unit or 156 mAh/g.340 (2) Whittingham, M. S. J. Electrochem. Soc. 1975, 122, 526.
(3) Ikeda, H.; Saito, T.; Tamaru, H. Denki Kagaku 1977, 45, 314.
(4) Ikeda, H.; Uena, S.; Saito, T.; Nakaido, S.; Tamaru, H. Denki
6. Conclusions and What Does the Future Hold Kagaku 1977, 45, 391.
(5) Sanyo, Lithium Battery Calculator, Model CS-8176L.
Lithium batteries have made substantial and sig- (6) Dey, A. N. J. Electrochem. Soc. 1971, 118, 1547.
nificant gains in the last 30 years from becoming a (7) Liang, C. C.; Bolster, M. E.; Murphy, R. M. U.S. Patents
curiosity to becoming the dominant rechargeable 4,310,609 and 4,391,729, 1982.
(8) Takeuchi, E. S.; Thiebolt, W. C. J. Electrochem. Soc. 1988, 135,
battery for consumer portable applications. However, 2691.
the next market opportunities will be much tougher (9) Crespi, A.; Schmidt, C.; Norton, J.; Chen, K.; Skarstad, P. J.
Electrochem. Soc. 2001, A30.
to conquer as they mostly demand higher power (10) Vissers, D. R.; Tomczuk, Z.; Steunenberg, R. K. J. Electrochem.
capabilities at lower costs and often in larger systems Soc. 1974, 121, 665.
with enhanced safety. The prime markets are the (11) Gay, E. C.; Vissers, D. R.; Martino, F. J.; Anderson, K. E. J.
Electrochem. Soc. 1976, 123, 1591.
high-power segment presently dominated by the (12) Kaun, T. D.; Nelson, P. A.; Redey, L.; Vissers, D. R.; Henriksen,
environmentally unfriendly Ni/Cd battery and the G. L. Electrochim. Acta 1993, 38, 1269.
HEV segment presently occupied by the Ni/metal (13) Whittingham, M. S. Prog. Solid State Chem. 1978, 12, 41.
(14) Mikhaylik, Y. V.; Akridge, J. R. J. Electrochem. Soc. 2003, 150,
hydride battery. The layered oxides will, without A306.
doubt, continue their ongoing improvement with (15) Visco, S. J. (PolyPlus-Battery) U.S. Patent 6,214,061, 2004.
(16) Yao, Y. Y.; Kummer, J. T. J. Inorg. Nucl. Chem. 1967, 29, 2453.
mixed transition metals slowly displacing the pure (17) Whittingham, M. S.; Huggins, R. A. J. Chem. Phys. 1971, 54,
cobalt system; they appear to offer enhanced safety 414.
with enhanced capacity at a lower cost and are drop- (18) Whittingham, M. S.; Huggins, R. A. NBS Spec. Pub. 1972, 364,
139.
in technology. The lithium iron phosphate cathode (19) Whittingham, M. S. In Fast Ion Transport in Solids; van Gool,
offers the first potentially low-cost cathode, but this W., Ed.; North-Holland: Amsterdam, 1973.
(20) Trumbore, F. A.; Broadhead, J.; Putvinski, T. M. Electrochem. (75) Whittingham, M. S.; Jacobson, A. J. J. Electrochem. Soc. 1981,
Soc. Abstr. 1973, 1973-2, 61. 128, 485.
(21) Broadhead, J.; Trumbore, F. A. Electrochem. Soc. Abstr. 1973, (76) Dines, M. B. Mater. Res. Bull. 1975, 10, 287.
1977-1, 178/179. (77) Whittingham, M. S.; Dines, M. B. J. Electrochem. Soc. 1977, 124,
(22) Broadhead, J. U.S. Patent 3,791,867, 1973. 1387.
(23) Murphy, D. W.; Trumbore, F. A. J. Electrochem. Soc. 1976, 123, (78) Murphy, D. W.; Christian, P. A. Science 1979, 205, 651.
960. (79) Dampier, F. W. J. Electrochem. Soc. 1974, 121, 656.
(24) Armand, M. B. In Fast Ion Transport in Solids; van Gool, W., (80) Walk, C. R.; Gore, J. S. J. Electrochem. Soc. 1975, 122, 68C.
Ed.; North-Holland: Amsterdam, 1973. (81) Walk, C. R.; Margalit, N. J. Power Sources 1997, 68, 723.
(25) Armand, M. B.; Touzan, P. Mater. Sci. Eng. 1977, 31, 319. (82) Delmas, C.; Cognac-Auradou, H.; Cocciantelli, J. M.; Ménétrier,
(26) Armand, M. B. U.S. Patent 4,041,220, 1977. M.; Doumerc, J. P. Solid State Ionics 1994, 69, 257.
(27) Whittingham, M. S. Layered Compounds; Asilomar: Monterey, (83) Delmas, C.; Cognac-Auradou, H.; Coociatelli, J. M.; Ménétrier,
CA, 1972. M.; Doumerc, J. P. Solid State Ionics 1994, 69, 257.
(28) Gamble, F. R.; Osiecki, J. H.; Cais, M.; Pishardy, R.; Disalvo, F. (84) Dickens, P. G.; French, S. J.; Hight, A. T.; Pye, M. F. Mater.
J.; Geballe, T. H. Science 1971, 174, 493. Res. Bull. 1979, 14, 1295.
(29) Gamble, F. R.; Thompson, A. H. Solid State Commun. 1978, 27, (85) Whittingham, M. S. Electrochem. Soc. Abstr. 1975, 1975-1, 40.
379. (86) Murphy, D. W.; Christian, P. A.; Disalvo, F. J.; Carides, J. N. J.
(30) Whittingham, M. S. Mater. Res. Bull. 1974, 9, 1681. Electrochem. Soc. 1979, 126, 497.
(31) Whittingham, M. S. J. Chem. Soc., Chem. Commun. 1974, 328. (87) Bergström, Ö.; Gustafsson, T.; Thomas, J. Abstracts of the XVII
(32) Whittingham, M. S. U.S. Patent 4040017 and U.K. Patent International Union of Crystallography Congress, Seattle,
1468244, 1973. WA; International Union of Crystallography
(33) SubbaRao, G. V.; Tsang, J. C. Mater. Res. Bull. 1974, 9, 921. (http://www.iucr.ac.uk/): Cambridge, U.K., 1996.
(34) Whittingham, M. S. U.S. Patent 4009052 and U.K. Patent (88) Bergström, Ö.; Gustafsson, T.; Thomas, J. O. Acta Crystallogr.
1468416, 1973. 1997, C53, 528.
(35) Whittingham, M. S. Science 1976, 192, 1126. (89) Bergström, Ö.; Gustafsson, T.; Thomas, J. O. Acta Crystallogr.
(36) Whittingham, M. S. Chemtech 1979, 9, 766. 1998, C54, 1204.
(37) Whittingham, M. S.; Chianelli, R. R. J. Chem. Educ. 1980, 57, (90) Howing, J.; Gustafsson, T.; Thomas, J. O. Acta Crystallogr. 2003,
569. B59, 747.
(38) Holleck, G. L.; Driscoll, J. P. Electrochim. Acta 1977, 22, 647. (91) Bjork, H.; Lidin, S.; Gustafsson, T.; Thomas, J. O. Acta Crys-
(39) Thompson, A. H. Phys. Rev. Lett. 1975, 35, 1786. tallogr. 2001, B57, 759.
(40) Whittingham, M. S. J. Electrochem. Soc. 1976, 123, 315. (92) Besenhard, J. O.; Schöllhorn, R. J. Power Sources 1976, 1, 267.
(41) Steele, B. C. H. In Fast Ion Transport in Solids; van Gool, W., (93) Schöllhorn, R.; Klein-Reesink, F.; Reimold, R. J. Chem. Soc.,
Ed.; North-Holland: Amsterdam, 1973. Chem. Commun. 1979, 398.
(42) Basu, S.; Worrell, W. In Fast Ion Transport in Solids; Vashishta, (94) Nassau, K.; Murphy, D. W. J. Non-Cryst. Solids 1981, 44, 297.
P., Mundy, J. N., Shenoy, G. K., Eds.; Elsevier-North-Holland: (95) West, K.; Zachau-Christiansen, B.; Skaarup, S.; Saidi, Y.; Barker,
New York, 1979. J.; Olsen, I. I.; Pynenburg, R.; Koksbang, R. J. Electrochem. Soc.
(43) Newman, G. H.; Klemann, L. P. J. Electrochem. Soc. 1980, 127, 1996, 143, 820.
2097. (96) Livage, J. Mater. Res. Bull. 1991, 26, 1173.
(44) Winn, D. A.; Shemilt, J. M.; Steele, B. C. H. Mater. Res. Bull. (97) Livage, J. Chem. Mater. 1991, 3, 578.
1976, 11, 559.
(98) Livage, J.; Baffier, N.; Pereira-Ramos, J. P.; Davidson, P. Mater.
(45) Whittingham, M. S.; Gamble, F. R. Mater. Res. Bull. 1975, 10,
Res. Soc. Proc. 1995, 369, 179.
363.
(99) Chandrappa, G. T.; Steunou, N.; Livage, J. Nature 2002, 416,
(46) Whittingham, M. S.; Panella, J. A. Mater. Res. Bull. 1981, 16,
702.
37.
(100) Le, D. B.; Passerini, S.; Guo, J.; Ressler, J.; Owens, B. B.; Smyrl,
(47) Whittingham, M. S. U.S. Patent 4,007,055, 1975.
W. H. J. Electrochem. Soc. 1996, 143, 2099.
(48) Whittingham, M. S. U.S. Patent 4,084,046, 1975.
(49) Yang, S.; Song, Y.; Zavalij, P. Y.; Whittingham, M. S. Electro- (101) Galy, J. J. Solid State Chem. 1992, 100, 229.
chem. Commun. 2002, 4, 239. (102) Oka, Y.; Yao, T.; Yamamoto, N. J. Solid State Chem. 1997, 132,
(50) Yang, S.; Zavalij, P. Y.; Whittingham, M. S. Electrochem. 323.
Commun. 2001, 3, 505. (103) Zhang, F.; Zavalij, P. Y.; Whittingham, M. S. Mater. Res. Bull.
(51) Morgan, D.; VanderVen, A.; Ceder, G. Electrochem. Solid State 1997, 32, 701.
Lett. 2004, 7, A30. (104) Zhang, F.; Zavalij, P. Y.; Whittingham, M. S. Mater. Res. Soc.
(52) Thompson, A. H. Phys. Rev. Lett. 1978, 40, 489. Proc. 1998, 496, 367.
(53) Yazami, R.; Touzain, P. J. Power Sources 1983, 9, 365. (105) Zhang, F.; Whittingham, M. S. Electrochem. Commun. 2000, 2,
(54) Basu, S. U.S. Patent 4,423,125, 1982. 69.
(55) Whittingham, M. S.; Newman, G. H. J. Electrochem. Soc. 1981, (106) Torardi, C. C.; Miao, C. R.; Lewittes, M. E.; Li, Z. Electrochem.
128, 706. Soc. Proc. 2000, 2000-21, 68.
(56) Whittingham, M. S.; Newman, G. H. U.S. Patent 4,086,403, (107) Torardi, C. C.; Miao, C. R.; Lewittes, M. E.; Li, Z. J. Solid State
1976. Chem. 2002, 163, 93.
(57) Newman, G. H.; Francis, R. W.; Gaines, L. H.; Rao, B. M. L. J. (108) Spahr, M. E.; Stoschitzki-Bitterli, P.; Nesper, R.; Müller, M.;
Electrochem. Soc. 1980, 127, 2025. Krumeich, F.; Nissen, H. U. Angew. Chem., Int. Ed. Engl. 1998,
(58) Chianelli, R. R.; Scanlon, J. C.; Rao, B. M. L. J. Electrochem. 37, 1263.
Soc. 1978, 125, 409. (109) Spahr, M. E.; Stoschitzki-Bitterli, P.; Nesper, R.; Haas, O.;
(59) Chianelli, R. R. J. Cryst. Growth 1976, 34, 239. Novak, P. J. Electrochem. Soc. 1999, 146, 2780.
(60) Rao, B. M. L.; Francis, R. W.; Christopher, H. A. J. Electrochem. (110) Edström, K.; Gustafsson, T.; Nordliner, S. Electrochem. Solid
Soc. 1977, 124, 1490. State Lett. 2001, 4, A129.
(61) Whittingham, M. S. Mater. Res. Bull. 1978, 13, 959. (111) Nordlinder, S.; Edström, K.; Gustafsson, T. Electrochem. Solid-
(62) Murphy, D. W.; DiSalvo, F. J.; Hull, G. W.; Waszczak, J. V. Inorg. State Lett. 2001, 4, A129.
Chem. 1976, 15, 17. (112) Dobley, A.; Ngala, K.; Yang, S.; Zavalij, P. Y.; Whittingham, M.
(63) Dahn, J. R.; Sacken, U. v.; Michal, C. A. Solid State Ionics 1990, S. Chem. Mater. 2001, 13, 4382.
44, 87. (113) Delmas, C.; Devalette, M.; Fouassier, C.; Hagenmuller, P. Mater.
(64) Johnson, C. S.; Kim, J.-S.; Kropf, A. J.; Kahaian, A. J.; Vaughey, Res. Bull. 1975, 10, 393.
J. T.; Thackeray, M. M. Electrochem. Commun. 2002, 4, 492. (114) Delmas, C.; Fouassier, C.; Hagenmuller, P. J. Solid State Chem.
(65) Haering, R. R.; Stiles, J. A. R.; Brandt, K. U.S. Patent 4,224,390, 1975, 13, 165.
1980. (115) Delmas, C.; Fouassier, C.; Réau, J.-M.; Hagenmuller, P. Mater.
(66) Murphy, D. W.; Carides, J. N.; DiSalvo, F. J.; Cros, C.; Waszczak, Res. Bull. 1976, 11, 1081.
J. V. Mater. Res. Bull. 1977, 12, 825. (116) Delmas, C.; Fouassier, C. Z. Anorg. Allg. Chem. 1976, 420, 184.
(67) Murphy, D. W.; Cros, C.; DiSalvo, F. J.; Waszczak, J. V. Inorg. (117) Delmas, C.; Fouassier, C.; Hagenmuller, P. Physica 1980, 99B,
Chem. 1977. 81.
(68) Schöllhorn, R.; Payer, A. Angew. Chem., Int. Ed. Engl. 1985, (118) Delmas, C.; Braconnier, J.-J.; Fouassier, C.; Hagenmuller, P.
24, 67. Solid State Ionics 1981, 3/4, 165.
(69) Sinha, S.; Murphy, D. W. Solid State Ionics 1986, 20, 81. (119) LeBlanc, A.; Danot, M.; Trichet, L.; Rouxel, J. Mater. Res. Bull.
(70) Goodenough, J. B.; Manthiram, A.; Wnetrzewski, B. J. Power 1974, 9, 191.
Sources 1993, 43-44, 269. (120) Rouxel, J.; Danot, M.; Bichon, J. Bull. Soc. Chim. Fr. 1971, 3930.
(71) Thompson, A. H.; Gamble, F. R.; Symon, C. R. Mater. Res. Bull. (121) Danot, M.; Blanc, A. L.; Rouxel, J. Bull. Soc. Chim. Fr. 1969,
1975, 10, 915. 2670.
(72) Trumbore, F. J. Power Sources 1989, 26, 65. (122) Danot, M.; Rouxel, J. C. R. Acad. Sci. Paris 1973, C276, 1283.
(73) Murphy, D. J.; Trumbore, F. A. J. Cryst. Growth 1977, 39, 185. (123) Mitzushima, K.; Jones, P. C.; Wiseman, P. J.; Goodenough, J.
(74) Whittingham, M. S.; Jacobson, A. J. U.S. Patent 4,233,375, 1979. B. Mater. Res. Bull. 1980, 15, 783.
(124) Amatucci, G. G.; Tarascon, J. M.; Klein, L. C. J. Electrochem. (173) Kanno, R.; Shirane, T.; Inaba, Y.; Kawamoto, Y. J. Power Sources
Soc. 1996, 143, 1114. 1997, 68, 145.
(125) Rossen, E.; Reimers, J. N.; Dahn, J. R. Solid State Ionics 1993, (174) Delmas, C.; Menetrier, M.; Croguennec, L.; Levasseur, S.; Peres,
62, 53. J. P.; Pouillerie, C.; Prado, G.; Fournes, L.; Weill, F. Int. J. Inorg.
(126) Gabrisch, H.; Yazimi, R.; Fultz, B. J. Electrochem. Soc. 2004, Mater. 1999, 1, 11.
151, A891. (175) Pouillerie, C.; Perton, F.; Biensan, P.; Peres, J. P.; Broussely,
(127) Nagaura, T.; Tozawa, K. Prog. Batteries Solar Cells 1990, 9, 209. M.; Delmas, C. J. Power Sources 2001, 96, 293.
(128) Ozawa, K. Solid State Ionics 1994, 69, 212. (176) Pouillerie, C.; Croguennec, L.; Delmas, C. Solid State Ionics
(129) Goodenough, J. B.; Mizuchima, K. U.S. Patent 4,302,518, 1981. 2000, 132, 15.
(130) Imanishi, N.; Fujiyoshi, M.; Takeda, Y.; Yamamoto, O.; Tabuchi, (177) Nakai, I.; Nakagome, T. Electrochem. Solid State Lett. 1998, 1,
M. Solid State Ionics 1999, 118, 121. 259.
(131) Levasseur, S.; Menetrier, M.; Suard, E.; Delmas, C. Solid State (178) Brouselly, M. Lithium Battery Discusssion, Bordeaux-Arcachon
Ionics 2000, 128, 11. 2001.
(132) Yonezu, I.; Tarui, H.; Yoshimura, S.; Fujitani, S.; Nohma, T. (179) Delmas, C.; Capitaine, F. Abstracts of the 8th International
Abstracts of the International Meeting On Lithium Batteries; Meeting Lithium Batteries; Electrochemical Society: Penning-
Electrochemical Society: Pennington, NJ, 2004; Vol. 12, abstract ton, NJ, 1996; Vol. 8, abstract 470.
58. (180) Chen, R.; Whittingham, M. S. J. Electrochem. Soc. 1997, 144,
(133) Nishi, Y. In Lithium Ion Batteries; Wakihara, M., Yamamoto, L64.
O., Eds.; Kodansha: Tokyo, 1998. (181) Armstrong, A. R.; Bruce, P. G. Nature 1996, 381, 499.
(134) Cho, J.; Kim, G. Electrochem. Solid State Lett. 1999, 2, 253. (182) Capitaine, F.; Gravereau, P.; Delmas, C. Solid State Ionics 1996,
(135) Cho, J.; Kim, C.; Yoo, S. I. Electrochem. Solid State Lett. 2000, 89, 197.
3, 362. (183) Chen, R.; Zavalij, P. Y.; Whittingham, M. S. Mater. Res. Soc.
(136) Cho, J.; Kim, Y. J.; Park, B. Chem. Mater. 2000, 12, 3788. Proc. 1997, 453, 653.
(137) Cho, J.; Kim, Y. J.; Park, B. J. Electrochem. Soc. 2001, 148, (184) Chen, R.; Zavalij, P.; Whittingham, M. S. Chem. Mater. 1996,
A1110. 8, 1275.
(138) Cho, J.; Kim, Y. J.; Kim, J. T.; Park, B. Angew. Chem., Int. Ed. (185) Chen, R.; Chirayil, T.; Whittingham, M. S. Solid State Ionics
2001, 40, 3367. 1996, 86-88, 1.
(139) Wang, Z.; Wu, C.; Liu, L.; Chen, L.; Huang, X. Solid State Ionics (186) Bach, S.; Henry, M.; Baffier, N.; Livage, J. J. Solid State Chem.
2002, 148, 335. 1990, 88, 325.
(140) Wang, Z.; Wu, C.; Liu, L.; Wu, F.; Chen, L.; Huang, X. J. (187) Bach, S.; Pereiraramos, J. P.; Cachet, C.; Bode, M.; Yu, L. T.
Electrochem. Soc. 2002, 149, A466. Electrochim. Acta 1995, 40, 785.
(141) Liu, L.; Wang, Z.; Li, H.; Chen, L.; Huang, X. Solid State Ionics (188) Bach, S.; Pereira-Ramos, J. P.; Baffier, N. J. Solid State Chem.
2002, 152, 341. 1995, 120, 70.
(142) Cho, J.; Park, B. Lithium Battery Discussion Electrode Materials, (189) Pereira-Ramos, J. P.; Baddour, R.; Bach, S.; Baffier, N. Solid
Bordeaux, 2003; Abstract 1. State Ionics 1992, 53-56, 701.
(143) Liu, J.; Xu, K.; Jow, T. R.; Amine, K. Electrochem. Soc. Abstr. (190) Stoyanova, R.; Zhecheva, E.; Zarkova, L. Solid State Ionics 1994,
2002, 2002-2, 135. 73, 233.
(144) Amine, K.; Liu, J.; Kang, S.; Belharouak, I.; Hyung, Y.; Vissers, (191) Doeff, M. M.; Richardson, T. J.; Kepley, L. J. Electrochem. Soc.
D.; Henriksen, G. J. Power Sources 2004, 129, 14. 1996, 143, 2507.
(145) Chen, Z.; Dahn, J. R. Electrochim. Acta 2004, 49, 1079. (192) Whittingham, M. S. 1996 USDOE BATT Annual Report;
(146) Seguin, L.; Amatucci, G.; Anne, M.; Chabre, Y.; Strobel, P.; USDOE: Washington, D.C., 1997.
Tarascon, J. M.; Vaughan, G. J. Power Sources 1999, 81-82, (193) Doeff, M. M.; Peng, M. Y.; Ma, Y.; Visco, S. J.; DeJonghe, L. C.
604. U.S. Patent 5,558,961, 1996.
(147) Goodenough, J. B.; Mizuchima, K. U.S. Patent 4,302,518, 1980.
(194) Doeff, M. M.; Richardson, T. J.; Hwang, K.-T.; Anapolsky, A.
(148) Goodenough, J. B.; Mizushima, K. U.S. Patent 4,357,215, 1981.
ITE Battery Lett. 2001, 2, B.
(149) Whittingham, M. S.; Zavalij, P. Y. Solid State Ionics 2000, 131,
(195) Armstrong, A. R.; Huang, H.; Jennings, R. A.; Bruce, P. G. J.
109.
Mater. Chem. 1998, 8, 255.
(150) Thackeray, M. M.; David, W. I. F.; Bruce, P. G.; Goodenough, J.
(196) Chen, R.; Whittingham, M. S. Proceedings of the Annual
B. Mater. Res. Bull. 1983, 18, 461.
Automotive Technology Development Customers Coordination
(151) Tarascon, J.-M.; Guyomard, D. Solid State Ionics 1994, 69, 222.
Meeting, Dearborn, MI, Oct. 27-30, 1997; USDOE: Washington,
(152) Tarascon, J. M.; McKinnon, W. R.; Coowar, F.; Bowmer, T. N.;
DC, 1997; Vol. III, p 301.
Amatucci, G.; Guyomard, D. J. Electrochem. Soc. 1994, 141,
1421. (197) Zhang, F.; Ngala, K.; Whittingham, M. S. Electrochem. Commun.
(153) Tarascon, J. M.; Wang, E.; Shokoohi, F. K.; McKinnon, W. R.; 2000, 2, 445.
Colson, S. J. Electrochem. Soc. 1991, 138, 2859. (198) Yang, S.; Song, Y.; Ngala, K.; Zavalij, P. Y.; Whittingham, M. S.
(154) Thackeray, M. M. Prog. Solid State Chem 1997, 25, 1. J. Power Sources 2003, 119, 239.
(155) Yonemura, M.; Yamada, A.; Kobayashi, H.; Tabuchi, M.; Ka- (199) Lu, Z.; Dahn, J. R. Chem. Mater. 2001, 13, 1252.
miyama, T.; Kawamoto, Y.; Kanno, R. J. Mater. Chem. 2004, (200) Lu, Z.; Dahn, J. R. Chem. Mater. 2001, 13, 2078.
14, 1948. (201) Lu, Z.; Dahn, J. R. J. Electrochem. Soc. 2001, 148, A237.
(156) Tarascon, J. M. U.S. Patent 5,135,732, 1992. (202) Lu, Z.; Dahn, J. R. Chem. Mater. 2000, 12, 3583.
(157) Amatucci, G. G.; Pereira, N.; Zheng, T.; Tarascon, J.-M. J. (203) Eriksson, T. A.; Lee, Y. J.; Hollingsworth, J.; Reimer, J. A.;
Electrochem. Soc. 2001, 148, A171. Cairns, E. J.; Zhang, X.-f.; Doeff, M. M. Chem. Mater. 2003, 15,
(158) Amatucci, G. G.; Pereira, N.; Zheng, T.; Plitz, I.; Tarascon, J.- 4456.
M. J. Power Sources 1999, 81-82, 39. (204) Dolle, M.; Hollingsworth, J.; Richardson, T. J.; Doeff, M. M. Solid
(159) Shin, Y.; Manthiram, A. J. Electrochem. Soc. 2004, 151, A204. State Ionics 2004, in press.
(160) Ferg, E.; Gummow, R. J.; Kock, A. D.; Thackeray, M. M. J. (205) Shaju, K. M.; SubbaRao, G. V.; Chowdari, B. V. R. Electrochim.
Electrochem. Soc. 1994, 141, L147. Acta 2003, 48, 2691.
(161) Ohzuku, T.; Ueda, A.; Yamamoto, N. J. Electrochem. Soc. 1995, (206) Armstrong, A. R.; Gitzendanner, R.; Robertson, A. D.; Bruce, P.
142, 1431. G. Chem. Commun. 1998, 1833.
(162) Zaghib, K.; Simoneau, M.; Armand, M.; Gauthier, M. J. Power (207) Numata, K.; Yamanaka, S. Solid State Ionics 1999, 118, 117.
Sources 1999, 81-82, 300. (208) Fujiwara, M.; Yamada, S.; Kanda, M. Extended Abstracts of 34th
(163) Tessier, C.; Fachetti, O.; Siret, C.; Castaing, F.; Jordy, C.; Boeuve, Battery Symposium. Nagoya, Japan; The Electrochemical So-
J. P.; Biensan, P. Lithium Battery Discussion Electrode Materi- ciety of Japan: Tokyo, Japan, 1993; Vol. 34, p 135.
als, Bordeaux, 2003, Abstract 29. (209) Yoshio, M.; Yamato, K.; Itoh, J.; Noguchi, H.; Okada, M.; Mouri,
(164) Franger, S.; Bourbon, C.; LeCras, F. J. Electrochem. Soc. 2004, T. Electrochem. Soc. Proc. 1994, 94-28, 251.
151, A1024. (210) Nitta, Y.; Okamura, K.; Haraguchi, K.; Kobayashi, S.; Ohta, A.
(165) Panero, S.; Satolli, D.; Salamon, M.; Scrosati, B. Electrochem. J. Power Sources 1995, 54, 511.
Commun. 2000, 2, 810. (211) Caurant, D.; Baffier, N.; Bianchi, V.; Grégoire, G.; Bach, S. J.
(166) Rougier, A.; Saadouane, I.; Gravereau, P.; Willmann, P.; Delmas, Mater. Chem. 1996, 6, 1149.
C. Solid State Ionics 1996, 90, 83. (212) Rossen, E.; Jones, C. D. W.; Dahn, J. R. Solid State Ionics 1992,
(167) Saardoune, I.; Delmas, C. J. Solid State Chem. 1998, 136, 8. 57, 311.
(168) Saadoune, I.; Menetrier, M.; Delmas, C. J. Mater. Chem. 1997, (213) Spahr, M. E.; Novák, P.; Schnyder, B.; Haas, O.; Nesper, R. J.
7, 2505. Electrochem. Soc. 1998, 145, 1113.
(169) Saadoune, I.; Delmas, C. J. Mater. Chem. 1996, 6, 193. (214) Ohzuku, T.; Makimura, Y. Chem. Lett. 2001, 744.
(170) Zhecheva, E.; Stoyanova, R. Solid State Ionics 1993, 66, 143. (215) Ohzuku, T.; Makimura, Y. Chem. Lett. 2001, 642.
(171) Prado, G.; Fournes, L.; Delmas, C. J. Solid State Chem. 2001, (216) Lu, Z.; MacNeil, D. D.; Dahn, J. R. Electrochem. Solid State Lett.
159, 103. 2001, 4, A200.
(172) Prado, G.; Rougier, A.; Fournes, L.; Delmas, C. J. Electrochem. (217) Wang, Z.; Sun, Y.; Chen, L.; Huang, X. J. Electrochem. Soc. 2004,
Soc. 2000, 147, 2880. 151, A914.
(218) Tsai, Y. W.; Lee, J. F.; Liu, D. G.; Hwamg, B. J. J. Mater. Chem. (262) Kang, S.-H.; Kim, J.; Stoll, M. E.; Abraham, D.; Sun, Y. K.;
2004, 14, 958. Amine, K. J. Power Sources 2002, 112, 41.
(219) Ngala, J. K.; Chernova, N. A.; Ma, M.; Mamak, M.; Zavalij, P. (263) Cushing, B. L.; Goodenough, J. B. Solid State Sci. 2002, 4, 1487.
Y.; Whittingham, M. S. J. Mater. Chem. 2004, 14, 214. (264) Reed, J.; Ceder, G. Electrochem. Solid State Lett. 2002, 5, A145.
(220) Ngala, J. K.; Chernova, N.; Matienzo, L.; Zavalij, P. Y.; Whit- (265) Kang, K.; Carlier, D.; Reed, J.; Arroyo, E. M.; Ceder, G.;
tingham, M. S. Mater. Res. Soc. Symp. 2003, 756, 231. Croguennec, L.; Delmas, C. Chem. Mater. 2003, 15, 4503.
(221) Hwang, B. J.; Tsai, Y. W.; Chen, C. H.; Santhanam, R. J. Mater. (266) Ohzuku, T.; Makimura, Y. Electrochem. Soc. Abstr. 2003, 2003-
Chem. 2003, 13, 1962. 1, 1079.
(222) Kim, J.-H.; Park, C. W.; Sun, Y.-K. Solid State Ionics 2003, 164, (267) Meng, Y. S.; Ceder, G.; Grey, C. P.; Yoon, W.-S.; Shao-Horn, Y.
43. Electrochem. Solid State Lett. 2004, 7, A155.
(223) Jiang, J.; Dahn, J. R. Electrochem. Commun. 2004, 6, 39. (268) Yoon, W.-S.; Iannopollo, S.; Grey, C. P.; Carlier, D.; Gorman, J.;
(224) Shaju, K. M.; Rao, G. V. S.; Chowdari, B. V. R. Electrochim. Acta Reed, J.; Ceder, G. Electrochem. Solid State Lett. 2004, 7, A167.
2002, 48, 145. (269) Ceder, G.; Meng, Y.-S.; Shao-Horn, Y.; Grey, C. P. 12th Inter-
(225) Park, S. H.; Yoon, C. S.; Kang, S. G.; Kim, H.-S.; Moon, S.-I.; national Meeting on Lithium Batteries; Electrochemical Soci-
Sun, Y.-K. Electrochim. Acta 2004, 49, 557. ety: Pennington, NJ, 2004; Vol. 12, abstract 22.
(226) Hwang, B. J.; Tsai, Y. W.; Carlier, D.; Ceder, G. Chem. Mater. (270) Kobayashi, H.; Sakaebe, H.; Kageyama, H.; Tatsumi, K.; Arachi,
2003, 15, 3676. Y.; Kamiyama, T. J. Mater. Chem. 2003, 13, 590.
(227) Koyama, Y.; Tanaka, I.; Adachi, H.; Makimura, Y.; Ohzuku, T. (271) Kobayashi, H.; Arachi, Y.; Kageyama, H.; Tatsumi, K. J. Mater.
J. Power Sources 2003, 119-121, 644. Chem. 2004, 14, 40.
(228) Yoshio, M.; Noguchi, H.; Itoh, J.-i.; Okada, M.; Mouri, T. J. Power (272) Liu, Z.; Yu, A.; Lee, J. Y. J. Power Sources 1999, 81-82, 416.
Sources 2000, 90, 176. (273) Jouanneau, S.; Dahn, J. R. Chem. Mater. 2003, 15, 495.
(229) Belharouak, I.; Sun, Y.-K.; Liu, J.; Amine, K. J. Power Sources (274) Thomas, M. G. S. R.; David, W. I. F.; Goodenough, J. B. Mater.
2003, 123, 247. Res. Bull. 1985, 20, 1137.
(230) Chen, Y.; Wang, G. X.; Konstantinov, K.; Liu, H. K.; Dou, S. X. (275) Guilmard, M.; Pouillerie, C.; Croguennec, L.; Delmas, C. Solid
J. Power Sources 2003, 119-121, 184. State Ionics 2003, 160, 39.
(231) Yabuuchi, N.; Ohzuku, T. J. Power Sources 2003, 119-121, 171. (276) Ma, M.; Chernova, N. A.; Zavalij, P. Y.; Whittingham, M. S. J
(232) Yoon, W.-S.; Grey, C. P.; Balasubramanian, M.; Yang, X.-Q.; Power Sources 2004, in press.
Fischer, D. A.; McBreen, J. Electrochem. Solid State Lett. 2004, (277) Cushing, B. L.; Goodenough, J. B. Solid State Sci. 2002, 4, 1487.
7, A53. (278) Oh, S. W.; Park, S. H.; Park, C.-W.; Sun, Y.-K. Solid State Ionics
(233) Sun, Y.; Ouyang, C.; Wang, Z.; Huang, X.; Chen, L. J. Electro- 2004, 171, 167.
chem. Soc. 2004, 151, A504. (279) Li, D.-C.; Muta, T.; Zhang, L.-Q.; Yoshio, M.; Noguchi, H. J.
(234) Kim, J.-M.; Chung, H.-T. Electrochim. Acta 2004, 49, 937. Power Sources 2004, 132, 150.
(235) Jouanneau, S.; Eberman, K. W.; Krause, L. J.; Dahn, J. R. J. (280) Meng, Y. S.; Wu, Y. W.; Hwang, B. J.; Li, Y.; Ceder, G. J.
Electrochem. Soc. 2003, 150, A1637. Electrochem. Soc. 2004, 151, A1134.
(236) MacNeil, D. D.; Lu, Z.; Dahn, J. R. J. Electrochem. Soc. 2002, (281) Ohzuku, T.; Nakura, K.; Aoki, T. Electrochim. Acta 1999, 45,
149, A1332. 151.
(237) Jouanneau, S.; Macneil, D. D.; Lu, Z.; Beattie, S. D.; Murphy, (282) Croguennec, L.; Pouillerie, C.; Delmas, C. Solid State Ionics
G.; Dahn, J. R. J. Electrochem. Soc. 2003, 150, A1299. 2000, 135, 259.
(238) Reed, J.; Ceder, G.; VanDerVen, A. Electrochem. Solid State Lett. (283) Yoshizawa, H.; Ohzuku, T. Denki Kagaku 2003, 71, 1177.
2001, 4, A78. (284) Guilmard, M.; Croguennec, L.; Delmas, C. Chem. Mater. 2003,
15, 4476.
(239) Armstrong, A. R.; Paterson, A. J.; Robertson, A. D.; Bruce, P. G.
(285) Guilmard, M.; Croguennec, L.; Delmas, C. Chem. Mater. 2003,
Chem. Mater. 2002, 14, 710.
15, 4484.
(240) Armstrong, A. R.; Robertson, A. D.; Bruce, P. G. Electrochim.
(286) Arai, H.; Sakurai, Y. J. Power Sources 1999, 80-81, 401.
Acta 1999, 45, 285.
(287) Manthiram, A.; Chu, S. Electrochem. Soc. Abstr. 2000, 2000-2,
(241) Sharma, P.; Moore, G.; Zhang, F.; Zavalij, P. Y.; Whittingham,
72.
M. S. Electrochem. Solid-State Lett. 1999, 2, 494.
(288) Chen, R.; Zavalij, P. Y.; Whittingham, M. S.; Greedan, J. E.; Raju,
(242) Zhang, F.; Whittingham, M. S. Electrochem. Solid State Lett. N. P.; Bieringer, M. J. Mater. Chem. 1999, 9, 93.
2000, 3, 309. (289) Thackeray, M. M.; Johnson, C. S.; Amine, K.; Kim, J. U.S. Patent
(243) Ammundsen, B.; Desilvestro, H.; Paulson, J. M.; Steiner, R.; 6,677,082, 2004.
Pickering, P. J. 10th International Meeting on Lithium Batter- (290) Thackeray, M. M.; Johnson, C. S.; Amine, K.; Kim, J. U.S. Patent
ies, Como, Italy, May 28-June 2, 2000; Electrochemical Soci- 6,680,143, 2004.
ety: Pennington, NJ, 2000. (291) Robertson, A. D.; Bruce, P. G. Chem. Commun. 2002, 2790.
(244) Storey, C.; Kargina, I.; Grincourt, Y.; Davidson, I. J.; Yoo, Y.; (292) Robertson, A. D.; Bruce, P. G. Chem. Mater. 2003, 15, 1984.
Seung, D. Y. 10th International Meeting on Lithium Batteries, (293) Armstrong, A. R.; Bruce, P. G. Electrochem. Solid State Lett.
Como, Italy, May 28-June 2, 2000; Electrochemical Society: 2004, 7, A1.
Pennington, NJ, 2000; Vol. 10, abstract 234. (294) Russouw, M. H.; Thackeray, M. M. Mater. Res. Bull. 1991, 26,
(245) Paulsen, J. M.; Ammundsen, B.; Desilvestro, H.; Steiner, R.; 463.
Hassell, D. Electrochem. Soc. Abstr. 2000, 2002-2, 71. (295) Russouw, M. H.; Liles, D. C.; Thackeray, M. M. J. Solid State
(246) Whitfield, P. S.; Davidson, I. J.; Kargina, I.; Grincourt, Y.; Chem. 1993, 104, 464.
Ammundsen, B.; Steiner, R.; Suprun, A. Electrochem. Soc. Abstr. (296) Paik, Y.; Grey, C. P.; Johnson, C. S.; Kim, J.-S.; Thackeray, M.
2000, 2000-2, 90. M. Chem. Mater. 2002, 14, 5106.
(247) Grincourt, Y.; Storey, C.; Davidson, I. J. J. Power Sources 2001, (297) Richard, M. N.; Fuller, E. W.; Dahn, J. R. Solid State Ionics 1994,
97-98, 711. 73, 81.
(248) Storey, C.; Kargina, I.; Grincourt, Y.; Davidson, I. J.; Yoo, Y. C.; (298) Shin, S.-S.; Sun, Y.-K.; Amine, K. J. Power Sources 2002, 112,
Seung, D. Y. J. Power Sources 2001, 97-98, 541. 634.
(249) Balasubramanian, M.; McBreen, J.; Davidson, I. J.; Whitfield, (299) Myung, S.-T.; Komaba, S.; Kumagai, N. Solid State Ionics 2004,
P. S.; Kargina, I. J. Electrochem. Soc. 2002, 149, A176. 170, 139.
(250) Ammundsen, B.; Paulsen, J.; Davidson, I.; Liu, R.-S.; Shen, C.- (300) Zhang, L.; Noguchi, H.; Yoshio, M. J. Power Sources 2002, 110,
H.; Chen, J.-M.; Jang, L.-Y.; Lee, J.-F. J. Electrochem. Soc. 2002, 57.
149, A431. (301) Kim, J.-S.; Johnson, C. S.; Thackeray, M. M. Electrochem.
(251) Numata, K.; Sakaki, C.; Yamanaka, S. Solid State Ionics 1999, Commun. 2002, 4, 205.
117, 257. (302) Kang, S.-H.; Sun, Y. K.; Amine, K. Electrochem. Solid State Lett.
(252) Lu, Z.; MacNeil, D. D.; Dahn, J. R. Electrochem. Solid State Lett. 2003, 6, A183.
2001, 4, A191. (303) Lee, C. W.; Sun, Y.-K.; Prakash, J. Electrochim. Acta 2004, 49,
(253) Arachi, Y.; Kobayashi, H.; Emura, S.; Nakata, Y.; Tanaka, M.; 4425.
Asai, T. Chem. Lett. 2003, 32, 60. (304) Kanno, R.; Shirane, T.; Kawamoto, Y.; Takeda, Y.; Takano, M.;
(254) Venkatraman, S.; Manthiram, A. Chem. Mater. 2003, 15, 5003. Ohashi, M.; Yamaguchi, Y. J. Electrochem. Soc. 1996, 143, 2435.
(255) Sun, Y.-K.; Yoon, C. S.; Lee, Y. S. Electrochim. Acta 2003, 48, (305) Kanno, R.; Shirane, T.; Kawamoto, Y. Abstracts of the 8th
2589. International Meeting on Lithium Batteries; Electrochemical
(256) Yang, X.-Q.; McBreen, J.; Yoon, W.-S.; Grey, C. P. Electrochem. Society: Pennington, NJ, 1996; Vol. 8, abstract 133.
Commun. 2002, 4, 649. (306) Whittingham, M. S. U.S. Patent 4,049,887, 1996.
(257) Rieck, H.; Hoppe, R. Z. Anorg. Allg. Chem. 1972, 392, 193. (307) Thompson, A. H.; Whittingham, M. S. Mater. Res. Bull. 1977,
(258) Makimura, Y.; Ohzuku, T. J. Power Sources 2003, 119-121, 156. 12, 741.
(259) Ariyoshi, K.; Iwakoshi, Y.; Nakayama, N.; Ohzuku, T. J. Elec- (308) Jacobson, A. J.; Whittingham, M. S. U.S. Patent 4,143,213, 1978.
trochem. Soc. 2004, 151, A296. (309) Padhi, A. K.; Nanjundaswamy, K. S.; Goodenough, J. B. J.
(260) Ohzuku, T. Personal Communication, 2004. Electrochem. Soc. 1997, 144, 1188.
(261) Shaju, K. M.; SubbaRao, G. V.; Chowdari, B. V. R. Electrochim. (310) Masquelier, C.; Padhi, A. K.; Nanjundaswany, K. S.; Good-
Acta 2004, 49, 1565. enough, J. B. J. Solid State Chem. 1998, 135, 228.
(311) Morcrette, M.; Wurm, C.; Masquelier, C. Solid State Sci. 2002, (337) Yamada, A.; Chung, S.-C. J. Electrochem. Soc. 2001, 148, A960.
4, 239. (338) Li, G.; Azuma, H.; Tohda, M. Electrochem. Solid State Lett. 2002,
(312) Masquelier, C.; Padhi, A. K.; Nanjundaswamy, K. S.; Good- 5, A135.
enough, J. B. J. Solid State Chem. 1998, 135, 228. (339) Delacourt, C.; Poizot, P.; Morcrette, M.; Tarascon, J.-M.; Mas-
(313) Chen, J.-M.; Li, Y. J.; Hurng, W.-M.; Whittingham, M. S. quelier, C. Chem. Mater. 2004, 16, 93.
(Industrial Technology Research Institute, Chutung, Taiwan) (340) Song, Y.; Zavalij, P. Y.; Chernova, N. A.; Whittingham, M. S. J.
U.S. Patent 5,514,490, 1996. Electrochem. Soc., in press.
(314) Li, Y. J.; Whittingham, M. S. Solid State Ionics 1993, 63, 391. (341) Zhou, F.; Cococcionic, M.; Marianetti, C.; Morgan, D.; Chen, M.;
(315) Yamada, A.; Chung, S. C.; Hinokuma, K. J. Electrochem. Soc. Ceder, G. 12th International Meeting on Lithium Batteries;
2001, 148, A224. Electrochemical Society: Pennington, NJ, 2004; Vol. 12, abstract
(316) Andersson, A. S.; Kalska, B.; Häggström, L.; Thomas, J. O. Solid 283.
State Ionics 2000, 130, 41. (342) Morgan, D.; Maxisch, T.; Zhou, F.; Cococcioni, M.; Kang, K.;
(317) Andersson, A. S.; Thomas, J. O.; Kalska, B.; Häggström, L. Ceder, G. 12th International Meeting on Lithium Batteries;
Electrochem. Solid-State Lett. 2000, 3, 66. Electrochemical Society: Pennington, NJ, 2004; Vol. 12, abstract
(318) Ravet, N.; Goodenough, J. B.; Besner, S.; Simoneau, M.; Hov- 282.
ington, P.; Armand, M. Electrochem. Soc. Abstr. 1999, 99-2, 127. (343) Osorio-Guillen, J. M.; Holm, B.; Ahuja, R.; Johansson, B. Solid
(319) Ravet, N.; Besner, S.; Simoneau, M.; Vallée, A.; Armand, M.; State Ionics 2004, 167, 221.
Magnan, J.-F. (Hydro-Quebec) European Patent 1049182A2, (344) Xu, Y.-N.; Chung, S.-Y.; Bloking, J. T.; Chiang, Y.-M.; Ching,
2000. W. Y. Electrochem. Solid State Lett. 2004, 7, A131.
(320) Huang, H.; Yin, S.-C.; Nazar, L. F. Electrochem. Solid State Lett. (345) Song, Y.; Yang, S.; Zavalij, P. Y.; Whittingham, M. S. Mater.
2001, 4, A170. Res. Bull. 2002, 37, 1249.
(321) Masquelier, C.; Wurm, C.; Morcrette, M.; Gaubicher, J. Inter- (346) Masquelier, C.; Reale, P.; Wurm, C.; Morcrette, M.; Dupont, L.;
antional Meeting on Solid State Ionics, Cairns, Australia, July Larchera, D. J. Electrochem. Soc. 2002, 149, A1037.
9-13, 2001; The International Society of Solid State Ionics: (347) Hong, Y.-S.; Ryu, K. S.; Park, Y. J.; Kim, M. G.; Leeb, J. M.;
2001; paper A-IN-06. Chang, S. H. J. Mater. Chem. 2002, 12, 1870.
(322) Prosini, P. P.; Zane, D.; Pasquali, M. Electrochim. Acta 2001,
(348) Prosini, P. P.; Cianchi, L.; Spina, G.; Lisi, M.; Scaccia, S.;
46, 3517.
Carewska, M.; Minarini, C.; Pasquali, M. J. Electrochem. Soc.
(323) Yang, S.; Song, Y.; Zavalij, P. Y.; Whittingham, M. S. Mater.
2001, 148, A1125.
Res. Soc. Proc. 2002, 703, V7.9.
(349) Song, Y.; Zavalij, P.; Whittingham, M. S. Mater. Res. Soc. Proc.
(324) Chung, S.-Y.; Bloking, J. T.; Chiang, Y.-M. Nat. Mater. 2002, 1,
2003, 756, 249.
123.
(325) Herle, P. S.; Ellis, B.; Coombs, N.; Nazar, L. F. Nat. Mater. 2004, (350) Song, Y.; Zavalij, P. Y.; Chernova, N. A.; Whittingham, M. S.
3, 147. Chem. Mater. 2004, in press.
(326) Croce, F.; Epifanio, A. D.; Hassoun, J.; Deptula, A.; Olczac, T.; (351) Barker, J.; Saidi, M. Y. U.S. Patent 5,871,866, 1999.
Scrosati, B. Electrochem. Solid State Lett. 2002, 5, A47. (352) Saidi, M. Y.; Barker, J.; Huang, H.; Adamson, G. Electrochem.
(327) Hu, Y.; Doeff, M. M.; Kostecki, R.; Fiñones, R. J. Electrochem. Solid State Lett. 2002, 5, A149.
Soc. 2004, 151, A1279. (353) Gaubicher, J.; Goward, G.; Wurm, C.; Masquelier, C.; Nazar, L.
(328) Shiraishi, K.; Dokko, K.; Kanamura, K. 12th International F. Chem. Mater. 2000, 12.
Meeting on Lithium Batteries; Electrochemical Society: Pen- (354) Yin, S. C.; Grondley, H.; Strobel, P.; Nazar, L. F. J. Am. Chem.
nington, NJ, 2004; Vol. 12, abstract 297. Soc. 2003, 125, 326.
(329) Nuspi, G.; Vogler, C.; Eisgruber, M.; Wimmer, L.; Schall, N.; (355) Yin, S. C.; Strobel, P.; Anne, M.; Nazar, L. F. J. Am. Chem. Soc.
Fietzek, C.; Weydanz, W. 12th International Meeting on Lithium 2003, 125, 10402.
Batteries; Electrochemical Society: Pennington, NJ, 2004; Vol. (356) Huang, H.; Yin, S.-C.; Kerr, T.; Nazar, L. F. Adv. Mater. 2002,
12, abstract 293. 14, 1525.
(330) Garcı́a-Martin, O.; Alvarez-Vega, M.; Garcia-Alvarado, F.; (357) Azmi, B. M.; Ishihara, T.; Nishiguchi, H.; Takita, Y. J. Power
Garcia-Jaca, J.; Gallardo-Amores, J. M.; Sanjuán, M. L.; Amador, Sources 2003, 119-121, 273.
U. Chem. Mater. 2001, 13, 1570. (358) Kerr, T. A.; Gaubischer, J.; Nazar, L. F. Electrochem. Solid State
(331) Whittingham, M. S.; Yang, S.; Ngala, K.; Dobley, A.; Zavalij, P. Lett. 2000, 3, 460.
Y. International Meeting on Solid State Ionics, Cairns, Australia, (359) Lim, S. C.; Vaughey, J. T.; Harrison, W. T. A.; Dussak, L. L.;
July 9-13, 2001; The International Society of Solid State Jacobson, A. J.; Johnson, J. W. Solid State Ionics 1996, 84, 219.
Ionics: 2001; paper A-KN-06. (360) Zavalij, P. Y.; Whittingham, M. S. Rigaku J. 2004, 21, 2.
(332) Chen, Z.; Dahn, J. R. J. Electrochem. Soc. 2002, 149, A1184. (361) Barker, J.; Saidi, M. Y.; Swoyer, J. L. J. Electrochem. Soc. 2003,
(333) Dominko, R.; Gaberscek, M.; Drofenik, J.; Bele, M.; Pejovnik, S. 150, A1394.
Electrochem. Solid State Lett. 2001, 4, A187. (362) Pereira-Ramos, J. P.; Baffier, N.; Pistoia, G. In Lithium Batteries;
(334) Doeff, M. M.; Hu, Y.; McLarnon, F.; Kostecki, R. Electrochem. Pistoia, G., Ed.; Elsevier: Amsterdam, 1994; Vol. 5.
Solid State Lett. 2003, 6, A207. (363) Zane, D.; Carewska, M.; Scaccia, S.; Cardellini, F.; Prosini, P.
(335) Takahashi, M.; Tobishima, S.; Takei, K.; Sakurai, Y. J. Power P. Electrochim. Acta 2004, 49, 4259.
Sources 2001, 97-98, 508. (364) Chirayil, T. A.; Zavalij, P. Y.; Whittingham, M. S. J. Electrochem.
(336) Okada, S.; Sawa, S.; Egashira, M.; Yamaki, J.-i.; Tabuchi, M.; Soc. 1996, 143, L193.
Kageyama, H.; Konishi, T.; Yoshino, A. J. Power Sources 2001,
97-98, 430. CR020731C
Chem. Rev. 2004, 104, 4303−4417 4303
Nonaqueous Liquid Electrolytes for Lithium-Based Rechargeable Batteries

Kang Xu
Electrochemistry Branch, Sensor and Electron Devices Directorate, U.S. Army Research Laboratory, Adelphi, Maryland 20783-1197
Received November 3, 2003
Contents 8. Novel Electrolyte Systems 4362

8.1. Problems Facing State-of-the-Art Electrolytes 4362
1. Introduction and Scope 4303 8.2. Functional Electrolytes: Additives 4363
1.1. Fundamentals of Battery Electrolytes 4303 8.2.1. Bulk Electrolyte: Ion Transport 4364
1.2. The Attraction of “Lithium” and Its Challenge 4304 8.2.2. Anode: SEI Modification 4366
1.3. From “Lithium” to “Lithium Ion” 4305 8.2.3. Cathode: Overcharge Protection 4372
1.4. Scope of This Review 4306 8.3. New Electrolyte Components 4378
2. Electrolyte Components: History and State of the 4306 8.3.1. Nonaqueous Solvents 4378
Art
8.3.2. Lithium Salts 4383
2.1. Solvents 4307
8.4. Novel Electrolytes with a Wide Temperature 4390
2.1.1. Propylene Carbonate (PC) 4308 Range
2.1.2. Ethers 4308 8.4.1. Low-Temperature Performance 4390
2.1.3. Ethylene Carbonate (EC) 4309 8.4.2. High-Temperature Performance 4399
2.1.4. Linear Dialkyl Carbonates 4310 8.5. Electrolytes of Low Flammability 4400
2.2. Lithium Salts 4310 8.6. Polymer and Polymer Gel Electrolytes 4405
2.2.1. Lithium Perchlorate (LiClO4) 4311 8.6.1. Solid Polymer Electrolyte 4406
2.2.2. Lithium Hexafluoroarsenate (LiAsF6) 4312 8.6.2. Gel Polymer Electrolyte 4408
2.2.3. Lithium Tetrafluoroborate (LiBF4) 4312 9. Concluding Remarks 4410
2.2.4. Lithium Trifluoromethanesulfonate (LiTf) 4312 10. Acknowledgments 4410
2.2.5. Lithium Bis(trifluoromethanesulfonyl)imide 4313 11. References and Notes 4411
(LiIm) and Its Derivatives
2.2.6. Lithium Hexafluorophosphate (LiPF6) 4314
2.3. Brief Summary 4315 1. Introduction and Scope
3. Liquid Range of Electrolyte Solutions 4315
4. Ion Transport Properties 4318 1.1. Fundamentals of Battery Electrolytes
5. Electrochemical Stability: on Inert Electrodes 4322 Electrolytes are ubiquitous and indispensable in all
5.1. Anion of Lithium Salts 4323 electrochemical devices, and their basic function is
5.2. Solvents 4325 independent of the much diversified chemistries and
6. Electrochemical Stability: on Active Electrodes 4326 applications of these devices. In this sense, the role
6.1. Passivation on Lithium Anode 4326 of electrolytes in electrolytic cells, capacitors, fuel
6.2. Electrolyte/Carbonaceous Anode Interface: 4329 cells, or batteries would remain the same: to serve
SEI as the medium for the transfer of charges, which are
6.2.1. Exfoliation and Irreversible Capacities on 4329 in the form of ions, between a pair of electrodes. The
a Carbonaceous Anode vast majority of the electrolytes are electrolytic
6.2.2. Mechanism of SEI Formation 4331 solution-types that consist of salts (also called “elec-
6.2.3. Characterization of Surface Chemistry 4337 trolyte solutes”) dissolved in solvents, either water
(aqueous) or organic molecules (nonaqueous), and are
6.3. Electrolyte/Cathode Interface 4341
in a liquid state in the service-temperature range.
6.3.1. Passivation Film on a Cathode 4342 [Although “nonaqueous” has been used overwhelm-
6.3.2. Characterization of Surface Chemistry 4344 ingly in the literature, “aprotic” would be a more
6.3.3. Breakdown of Surface Layer 4346 precise term. Either anhydrous ammonia or ethanol
6.3.4. Passivation of Current Collector 4347 qualifies as a “nonaqueous solvent” but is unstable
6.4. A Few Words on Surface Characterizations 4351 with lithium because of the active protons. Neverthe-
7. Chemical and Thermal Stability/Safety of 4352 less, this review will conform to the convention and
Electrolytes use “nonaqueous” in place of “aprotic”.]
7.1. Long-Term Stability of Electrolytes at 4352 Because of its physical location in the electrochemi-
Elevated Temperatures cal devices, that is, being sandwiched between posi-
7.2. Stability of the SEI or Surface Layer at 4354 tive and negative electrodes, the electrolyte is in close
Elevated Temperatures interaction with both electrodes; therefore, when new
7.3. Thermal Safety of Electrolytes against Abuse 4357 electrode materials come into use, the need for
10.1021/cr030203g CCC: $48.50 © 2004 American Chemical Society
4304 Chemical Reviews, 2004, Vol. 104, No. 10 Xu
candidate electrode materials, and thus constantly

requests improvements in electrolyte stability. Ad hoc
surface chemistry is often necessary for the kinetic
stability of these new electrolyte/electrode interfaces.
While the potencies of electrode materials are usually
quantified by the redox potential in volts against
some certain reference potential,3 the stability of an
electrolyte can also be quantified by the range in volts
between its oxidative and reductive decomposition
limits, which is known as the “electrochemical win-
dow”. Obviously, the redox potential of both electrode
materials must fall within this electrochemical win-
dow to enable a rechargeable battery operation.
Certainly, electrochemical stability is only one of
the requirements that an electrolyte should meet. A
Kang Xu was born in Chengdu, China, and received his B.S. degree in generalized list of these minimal requirements should
Chemistry from Southwest Normal University in Chongqing, China, in 1985
and M.S. in Polymer Chemistry from Lanzhou Institute of Chemical Physics, include the following: (1) It should be a good ionic
Academy of Sciences, in 1988. After working on polymer electrolyte conductor and electronic insulator, so that ion trans-
materials from 1988 to 1992 at Chengdu Institute of Organic Chemistry, port can be facile and self-discharge can be kept to a
Academy of Sciences, he went to Arizona State University and received minimum. (2) It should have a wide electrochemical
a Ph.D. degree in Chemistry in 1996 under the tutelage of C. Austen window, so that electrolyte degradation would not
Angell. From 1997 to 2002, he was awarded the National Research Council occur within the range of the working potentials of
Research Associate Fellowship and the American Society for Engineer
Education Postdoctoral Fellowship, respectively, and he served during both the cathode and the anode. (3) It should also be
the tenures as a guest researcher at U.S. Army Research Laboratory inert to other cell components such as cell separators,
with T. Richard Jow as academic advisor. He was employed by the U.S. electrode substrates, and cell packaging materials.
Army Research Laboratory in 2002. His research interests concern (4) It should be robust against various abuses, such
materials development for electrochemical energy storage applications, as electrical, mechanical, or thermal ones. (5) Its
which include lithium or lithium ion batteries and electrochemical capacitors. components should be environmentally friendly.
He won R&D Achievement Awards from the Department of the Army in
1999, 2001, and 2002 for his work on electrolyte materials. He authored
over 60 research publications and 11 patents and is a member of the 1.2. The Attraction of “Lithium” and Its Challenge
Electrochemical Society.
Lithium has long received much attention as a
promising anode material. The interest in this alkali
compatible electrolytes usually arises. The interfaces metal has arisen from the combination of its two
between the electrolyte and the two electrodes often unique properties: (1) it is the most electronegative
dictate the performance of the devices. In fact, these metal (∼ -3.0 V vs SHE), and (2) it is the lightest
electrified interfaces have been the focus of interest metal (0.534 g cm-3).3 The former confers upon it a
since the dawn of modern electrochemistry1 and negative potential that translates into high cell
remain so in the contemporary lithium-based re- voltage when matched with certain cathodes, and the
chargeable battery technologies. latter makes it an anode of high specific capacity
In a battery,2 the chemical nature of positive and (3.86 A h g-1). In the 1950s lithium metal was found
negative electrodes (also called cathode and anode, to be stable in a number of nonaqueous solvents
respectively, although by a more strict definition, this despite its reactivity,4 and this stabilization was
convention is only correct during discharge) decides attributed to the formation of a passivation film on
the energy output, while the electrolyte, in most the lithium surface, which prevents it from having a
situations, defines how fast the energy could be sustained reaction with electrolytes. Intensified re-
released by controlling the rate of mass flow within search activities resulted in the commercialization
the battery. Conceptually, the electrolyte should of a series of lithium-based primary cells in the 1960s
undergo no net chemical changes during the opera- and 1970s, and the electrolyte solvents ranged from
tion of the battery, and all Faradaic processes are organic (propylene carbonate) to inorganic (thionyl
expected to occur within the electrodes. Therefore, chloride and sulfur dioxide).5-8
in an oversimplified expression, an electrolyte could The continued efforts to expand lithium chemistry
be viewed as the inert component in the battery, and into rechargeable technology, however, encountered
it must demonstrate stability against both cathode severe difficulties in terms of the cycle life and
and anode surfaces. This electrochemical stability of safety.9,10 Soon it was realized that the source of the
the electrolyte, which in actual devices is usually problems was the morphology of the lithium crystals
realized in a kinetic (passivation) rather than ther- newly deposited from the electrolytes upon re-
modynamic manner, is of especial importance to charge.11,12 Needlelike lithium crystals (called “den-
rechargeable battery systems, but it is often chal- drite”) grow on the anode upon charge and, during
lenged by the strong oxidizing and reducing nature the subsequent discharge, become electrically isolated
of the cathode and the anode, respectively. The from the substrate due to nonuniform dissolution
severity of this challenge is ever increasing with the rates at different sites of the dendrite. The direct
pursuit of new battery systems with higher energy victim of such lithium loss is energy density, because
densities, which drives the exploration of a more excessive lithium has to be used in the cell to make
oxidizing cathode and a more reducing anode as up for the loss.13 But more seriously, a hazard could
Electrolytes for Lithium-Based Rechargeable Batteries Chemical Reviews, 2004, Vol. 104, No. 10 4305
be caused by such “dead lithium” crystals, which are

electrochemically inactive but chemically hyper-reac-
tive due to their high surface area. When dendrite
growth pierces the separator and results in an
internal short, thermal runaway and explosion ensue.
The work on rechargeable lithium batteries from
the 1970s to 1980s overwhelmingly concentrated on
the electrolyte formulation, in the hope that proper
choices of electrolyte solvents or salts would suppress
or even eliminate the dendritic deposition of lithium.
Among the numerous electrolyte systems investi-
gated, an ether-based solution developed by an Israeli
company seems to have achieved the apex for lithium
metal-based rechargeable batteries.14 A lithium elec-
trode is highly stable in this solution up to 120 °C
because of excellent surface passivation,15 and over
300 depth discharge cycles have been reported.16 A
novel redox mechanism between the salt, LiAsF6, and
solvent, 1,3-dioxolane, shuts down the cell chemistry
at temperatures higher than 125 °C, thus preventing
thermal runaway.16,17 However, dendrite formation,
especially at high charge rates, still results in capac- Figure 1. Schematic description of a “(lithium ion) rock-
ity fade.18 In 1989 incidents of fire due to lithium ing-chair” cell that employs graphitic carbon as anode and
rechargeable batteries in the electronic devices, fol- transition metal oxide as cathode. The undergoing electro-
lowed by the manufacturer recalls, highlighted the chemical process is lithium ion deintercalation from the
end of general enthusiasm in lithium metal as an graphene structure of the anode and simultaneous inter-
anode.19 calation into the layered structure of the metal oxide
cathode. For the cell, this process is discharge, since the
reaction is spontaneous.
1.3. From “Lithium” to “Lithium Ion”
The failure of lithium as an anode due to dendrite electrolytes.25,26 However, the enhanced safety and
formation prompted the search for a way to circum- extended cycle life were not sufficient to offset the
vent the drastic morphological change of the anode penalty in energy density caused by the replacement
during cell cycling. As a result, “host-guest” chem- of lithium metal; hence, these systems were never
istry was considered. Also known as “intercalation”- commercialized.27,28
or “insertion”-type electrodes, this concept of revers- A breakthrough was made when Japanese research-
ible chemistry had been applied earlier to cathode ers exploited an old concept of using carbonaceous
materials for lithium batteries, as represented by the materials as anode intercalation host.29-31 The term
trail-blazing work of Whittingham20,21 and the sig- “lithium ion battery” that was introduced by those
nificant improvements by Goodenough et al. and researchers eventually prevailed and replaced the
others.22,23 Most of the host materials are transition other aliases such as “rocking-chair”,24 “shuttle-
metal oxides or chalcogenides with stable crystal cock”,32 or “swing” batteries.33 In the charged state
lattices, and their layer or tunnel structures provide of these carbonaceous anodes, lithium exists in its
the pathways for guest ions such as the lithium ion ionic rather than metallic state, thus eliminating any
to diffuse. By injecting or extracting electrons, the possibility of dendrite lithium. The advantage of this
redox reactions occur on the host lattice while mobile new host is highlighted by the low cost of carbon and
guest ions intercalate into or deintercalate from the the high lithium ion activity in the intercalation
host matrix to compensate for regional electroneu- compound; the latter renders an anode potential close
trality. During the whole intercalation/deintercala- to that of lithium metal and minimizes the energetic
tion cycle, there are no Faradaic changes in the penalty. In 1990 both Sony34 and Moli35 announced
“guest ion”. If a similar intercalation host could be the commercialization of cells based on petroleum
found and used as an anode material, then a battery coke and LiCoO2, though Sony was generally credited
employing such intercalation cathodes and anodes for making this technology a commercial reality. In
would only require the lithium ion to shuttle back the same year Dahn and co-workers published their
and forth between the electrodes without the pres- seminal report on the principle of lithium intercala-
ence of lithium metal. The nickname “rocking-chair tion chemistry with graphitic anodes and the effect
battery” was given to such a device that uses dual of electrolyte solvent in the process.36 In fact, the
intercalation electrodes,24 the working principle of conclusions drawn therein constitute the foundation
which is schematically depicted in Figure 1, using for the current lithium ion battery industry:
the example of the state-of-the-art lithium ion chem- (1) Electrolyte solvents decompose reductively on
istry. the carbonaceous anode, and the decomposition prod-
The concept of rocking-chair lithium batteries was uct forms a protective film. When the surface of the
confirmed experimentally by using lithiated oxides anode is covered, the film prevents further decom-
(Li6Fe2O3, LiWO2) as interaction anodes and other position of the electrolyte components. This film is
oxides (WO3, TiS2, V2O5) as cathodes in nonaqueous an ionic conductor but an electronic insulator.
(2) This reductive decomposition process occurs almost every one of these reviews. On the other hand,
only during the first charge and is absent in the attention has always been focused on electrode
following cycles so that the carbonaceous anode can materials, especially the anodes, and electrolytes as
be cycled many times in the electrolyte, yielding an important component of the cell have not been
stable capacity. comprehensively treated in any dedicated reviews.
(3) The chemical structure of the electrolyte sol- This review intends to fill this deficit by summariz-
vents critically influences the nature of the protective ing the progress made during the last 10 years in
film, and ethylene carbonate was found to be an the research and development of electrolytes for
essential component of the solvents that protects the lithium-based batteries. Since lithium ion chemistry
highly crystalline structure of graphite. is by far the only successfully commercialized re-
Obviously, the film formed on a carbonaceous chargeable lithium-based technology, emphasis will
anode plays a critical role in enabling a lithium ion be placed on the electrolytes developed for this
device to work reversibly. Presuming that the surface system. Liquid electrolytes will take the central
nature of the carbonaceous anode at low potentials stage, and the scope of the review will include their
is similar to that of lithium metal in nonaqueous ionics, phase diagrams, interfaces with cathode and
electrolytes, Dahn and co-workers adopted a model anode materials, long-term chemical stability in the
developed earlier by Peled to describe the passivation device, thermal properties and performance at ex-
on lithium metal37 and named this surface film on treme temperatures, and safety characterizations.
carbonaceous anodes a “solid electrolyte interface” Whenever an interdisciplinary topic involving both
(SEI). This term soon became the most frequently electrolyte and other cell components is encountered
used key word in publications concerning lithium ion (i.e., electrolyte/electrode interface and passivation
technology in the following decade. Although it will of electrodes), emphasis will be placed on the role and
turn out later that the exact mechanism involved in effect of electrolyte components.
the formation is far more complicated and remains For the convenience of this discussion, a somewhat
a controversial topic even today, it has been generally arbitrary demarcation was drawn between “state-of-
agreed that the electrolyte reduction products are the the-art” (SOA) and “novel” electrolyte systems, with
main components of an SEI and dictate the chemical the former referring to the ones currently used in
as well as thermal properties of the electrode. commercialized lithium ion cells and the latter to the
The decade following Dahn’s publication witnessed ones improved over the SOA systems but still under
an explosive growth in lithium ion technology re- development. It should be pointed out that the exact
search, and essentially all aspects of lithium ion electrolyte compositions in commercialized devices
technology were explored with state-of-the-art tech- are usually proprietary knowledge, but publications
niques, while the main excitement revolved around from the affiliated researchers normally disclose
developing new materials such as carbonaceous sufficient information to reveal the skeletal electro-
anode and metal oxide cathode materials and the lyte components employed. The distinction made in
electrolyte solvents and salts compatible with them. this review concerning the previously mentioned
The result of those intensified efforts was the suc- demarcation is based on such open literature.
cessful commercialization and the rapid thriving of This review will focus on the literature published
this youngest battery chemistry. By 2000, the quan- from 1990 to the middle of 2003. Meanwhile, a
tity of lithium ion cells manufactured reached ∼620 certain amount of attention will also be allocated to
million units with a market value of ∼1 billion the electrolytes for lithium batteries to avoid omitting
dollars,38 accounting for more than a 90% share of the important progress made in these closely related
the rechargeable battery market39 or 63% of total fields. When selecting references, efforts were made
sales in portable batteries. The employment of new to ensure academic quality as well as ready public
materials and novel engineering designs has pushed accessibility. For this reason, patents, various techni-
the cycle life, energy, and power density of this cal reports, and conference/workshop presentations/
technology to more than 2000 cycles, 160 W h kg-1, abstracts were avoided to the extent possible. There
and 5000 W kg-1, respectively.40 The major driving were exceptions, though, when there was no alterna-
force of this market remains the so-called “small tive reference source. Finally, although comprehen-
formula batteries” with capacities smaller than 1 A sive coverage was attempted, it is essentially impos-
h; however, industry-size lithium ion cells are in- sible to cover every aspect in an exhaustive manner.
creasingly being used in space, military, and other The choice of the references and the organization of
special applications, especially as traction power the content reflect the personal view of the author
sources for electric or hybrid electric vehicle (EV/ only.
HEV) applications.40
1.4. Scope of This Review 2. Electrolyte Components: History and State of

the Art
Since the inception of lithium ion technology, there
have been several reviews summarizing the knowl- Most compositions of lithium electrolytes are based
edge accumulated about this new technology from on solutions of one or more lithium salts in mixtures
various perspectives, with the latest being in 2003.41-48 of two or more solvents, and single-solvent formula-
Because electrolytes interact closely with both cath- tions are very rare, if there are any. The rationale
ode and anode materials during the operation, their behind this mixed solvent formulation is that the
effect on cell performance has been discussed in diverse and often contradicting requirements of bat-
Table 1. Organic Carbonates and Esters as Electrolyte Solvents
a The mp of DEC recorded in various literature sources (books, papers, commercial catalogs) has been -43 °C, which was
corrected by a very recent measurement (ref 50e). This widespread error of 30° seems to stem from a single source in 1921, which
was then registered by Beilstein Handbuch and escaped detection for approximately eight decades.
tery applications can hardly be met by any individual the following minimal criteria: (1) It should be able
compound, for example, high fluidity versus high to dissolve salts to sufficient concentration. In other
dielectric constant; therefore, solvents of very differ- words, it should have a high dielectric constant ().
ent physical and chemical natures are often used (2) It should be fluid (low viscosity η), so that facile
together to perform various functions simultaneously. ion transport can occur. (3) It should remain inert to
A mixture of salts, on the other hand, is usually not all cell components, especially the charged surfaces
used, because anion choice is usually limited, and of the cathode and the anode, during cell operation.
performance advantages or improvements are not (4) It should remain liquid in a wide temperature
readily demonstrated. range. In other words, its melting point (Tm) should
Solid polymer and gel polymer electrolytes could be low and its boiling point (Tb) high. (5) It should
be viewed as the special variation of the solution-type also be safe (high flash point Tf), nontoxic, and
electrolyte. In the former, the solvents are polar economical.
macromolecules that dissolve salts, while, in the For lithium-based batteries, the active nature of
latter, only a small portion of high polymer is the strongly reducing anodes (lithium metal or the
employed as the mechanical matrix, which is either highly lithiated carbon) and the strongly oxidizing
soaked with or swollen by essentially the same liquid cathodes (transition metal based oxides) rules out the
electrolytes. One exception exists: molten salt (ionic use of any solvents that have active protons despite
liquid) electrolytes where no solvent is present and their excellent power in solvating salts, because the
the dissociation of opposite ions is solely achieved by reduction of such protons and/or the oxidation of the
the thermal disintegration of the salt lattice (melt- corresponding anions generally occurs within 2.0-
ing). Polymer electrolyte will be reviewed in section 4.0 V versus Li,49 while the charged potentials of the
8 (“Novel Electrolyte Systems”), although lithium ion anode and the cathode in the current rechargeable
technology based on gel polymer electrolytes has in lithium devices average 0.0-0.2 V and 3.0-4.5 V,
fact entered the market and accounted for 4% of respectively. On the other hand, the nonaqueous
lithium ion cells manufactured in 2000.38 On the compounds that qualify as electrolyte solvents must
other hand, ionic liquid electrolytes will be omitted, be able to dissolve sufficient amounts of lithium salt;
due to both the limited literature concerning this therefore, only those with polar groups such as
topic and the fact that the application of ionic liquid carbonyl (CdO), nitrile (CtN), sulfonyl (SdO), and
electrolytes in lithium ion devices remains dubious. ether-linkage (-O-) merit consideration.
Since most of the ionic liquid systems are still in a
Since the inception of nonaqueous electrolytes, a
supercooled state at ambient temperature, it is
wide spectrum of polar solvents has been investi-
unlikely that the metastable liquid state could be
gated, and the majority of them fall into either one
maintained in an actual electrochemical device,
of the following families: organic esters and ethers.
wherein electrode materials would serve as effective
The most commonly used solvents from these fami-
nucleation sites for crystallization.
lies, along with their physical properties, are listed
in Tables 1 and 2, respectively,50 where the melting
2.1. Solvents temperature of diethyl carbonate (DEC) deserves
In accordance with the basic requirements for special attention because a significant correction has
electrolytes, an ideal electrolyte solvent should meet been made recently.50e
Table 2. Organic Ethers as Electrolyte Solvents
Scheme 1. Reduction of PC on a Lithium Surface:

An interesting observation should be made con-
One-Electron Process
cerning the dependence of the physical properties on
molecular cyclicity, since it will have a significant
effect on the formulation of electrolytes for lithium
ion cells. While all of the ethers, cyclic or acyclic,
demonstrate similar moderate dielectric constants
(2-7) and low viscosities (0.3-0.6 cP), cyclic and
acyclic esters behave like two entirely different kinds
of compounds in terms of dielectric constant and
viscosity; that is, all cyclic esters are uniformly polar
( ) 40-90) and rather viscous (η ) 1.7-2.0 cP), and
all acyclic esters are weakly polar ( ) 3-6) and fluid surface, and a one-electron reduction process has
(η ) 0.4-0.7 cP). The origin for the effect of molecular been proposed (see Scheme 1).55
cyclicity on the dielectric constant has been at- The overall capacity fading of lithium cells using
tributed to the intramolecular strain of the cyclic PC-based electrolytes, however, is a more complicated
structures that favors the conformation of better mechanism, although Scheme 1 plays a part in it.
alignment of molecular dipoles, while the more flex- The static stability of PC against a lithium surface
ible and open structure of linear carbonates results had been attributed earlier to the existence of a
in the mutual cancellation of these dipoles. protective layer,37,56 which consists of the decomposi-
tion products shown in Scheme 1, and prevents the
2.1.1. Propylene Carbonate (PC) sustained reaction of PC with lithium.55 On the other
hand, it was the dynamic reactivity that results in
Among these solvents, cyclic diesters of carbonic the lithium loss during cycling,9 for which the main
acid have undoubtedly attracted the main research cause is related to the nonuniform morphological
attention throughout the entire history of lithium change of the lithium surface rather than chemical
batteries, especially in the past decade, when their corrosion. Figure 2 schematically shows this nonuni-
role in forming an SEI on carbonaceous anodes formity of the lithium surface during the cycling
became recognized. However, the early interest in process, where uneven growth of the electrodeposited
these compounds arose solely from their high dielec- lithium crystals results in dendrites that in subse-
tric constant and, hence, their ability to dissolve a quent discharge processes (lithium dissolution) pro-
wide variety of lithium salts. In 1958, it was observed duce lithium particles that are electrically isolated
that lithium could be electrodeposited from a solution from the lithium anode. Microscopic studies have
of LiClO4 in PC,51 and PC became the immediate confirmed the existence of dendrites (Figure 3) and
focus of investigation.4,9,10 Its wide liquid range have attributed their formation to the presence of a
(defined by the difference between Tm and Tb, Table passivation film.37,57 Serious safety hazards are often
1), high dielectric constant, and static stability with caused by the generation of both dendrites and
lithium made it a preferred solvent, and considerable isolated lithium crystals.58,59 The former creates
efforts were made to purify it52 when a less-than-ideal internal shorts, and the latter is chemically active
stripping/plating efficiency (e 85%) for lithium was with the electrolyte solvents due to their huge surface
observed during cycling. The capacity of lithium cells areas.
with PC electrolytes also fades accordingly.4,9,10,54
However, it was soon realized that this poor cycling 2.1.2. Ethers
efficiency was more intrinsic to the electrolyte solvent
than contaminant-originated, and the reaction be- In view of the poor cycling efficiency and the
tween the PC and the newly deposited lithium potential hazards associated with PC, people turned
particles was thought to be the cause.9 More recent to ethers for improved lithium morphology. In the
studies employing spectroscopic means have con- 1980s, ethers were widely preferred by researchers
firmed the PC reduction on a newly formed lithium as an alternative candidate, because of their low
(>100 cycles) of the cells still produced dendrite

deposition,71 which terminated the cells by creating
shorts,72 despite the improved lithium morphology
observed in the short term.
In addition to the problem with the lithium anode,
a new factor contributing to the capacity fade sur-
faced as the oxidative decomposition of ether-based
compounds on the cathode surface.61,67,73 Electro-
chemical studies on these ether-based electrolytes
placed the potential for the oxidative breakdown of
the ether functionality at relatively low potentials.
On a platinum surface, for example, THF was found
to be oxidized at 4.0 V vs Li, while PC remained
stable up to 5.0 V.74 In an actual cell, ethers are more
readily decomposed at even lower potentials, because
of the highly catalytic surface of cathode materials.76
With the application of more potent 4.0 V cathode
materials (LixMO2, M ) Mn, Ni, or Co) in lithium or
lithium ion cells, the possibility of using ether
compounds as solvents or cosolvents diminished.75,77
During the 1990s, various ethers were gradually
phased out in most of the electrolyte systems under
Figure 2. (a) Schematic description for the growth of investigation. The failure of ether-based electrolytes
dendrite crystals on a Li surface. The film consisting of served as a perfect example to illustrate that, in a
decomposition products as shown in Scheme 1 prevents the battery, the electrolyte (solvents and salts) must cope
growth of large granular crystals but rather promotes the
formation of treelike dendrites. (b) Schematic description
with challenges from both the anode and the cathode.
for the formation of isolated lithium particles from Li On the other hand, the advantage of organic esters,
dendrites. The uneven dissolution of the dendrites leaves especially cyclic alkyl carbonates, was rediscovered
lithium crystals detached from the lithium substrate. The because of their excellent stability against oxidation
isolated lithium crystals become electrochemically “dead” on cathode surfaces.74,78,79
but chemically reactive due to their high surface area.
2.1.3. Ethylene Carbonate (EC)
The interest in alkyl carbonates was renewed with
the emergence of the lithium ion “shuttle” concept.24-31
Gone with the lithium anode is the difficult issue of
lithium morphology; subsequently, the higher anodic
stability of PC made it once again a promising
candidate. In the first generation of the commercial
lithium ion cells, a PC-based electrolyte was used by
Sony along with LixCoO2 as the cathode and petro-
leum coke as the anode.31 However, the real renais-
sance for using alkyl carbonates as lithium electrolyte
solvents was not brought about by PC but, quite
unexpectedly, by its high melting cousin EC.
Compared with PC, EC has comparable viscosity
and slightly higher dielectric constant, which are
Figure 3. Micrograph of a single dendrite lithium grown favorable merits for a solvent candidate (Table 1). In
in PC. (Reproduced with permission from ref 57 (Figure fact, its dielectric constant is even higher than that
6a). Copyright 1989 The Electrochemical Society.) of the the most common electrolyte solvent on the
earth: water ( ∼ 79).3 However, because of its high
viscosity and resultant high ionic conductivity, but, melting point (∼36 °C), it was never favored as an
most of all, the better lithium morphology during ambient-temperature electrolyte solvent in the early
cycling.60 The cycling efficiency of lithium was re- days of lithium battery research: the liquid range of
ported to be 88% in tetrahydrofuran (THF),60-62 an the electrolytes based on it would be too restricted.
average of 96% in 2-methyltetrahydrofuran (2-Me- Its higher melting point than those of other members
THF),62-65 97% in polymethoxy ethers66 and dimethoxy of the carbonate family (Table 1) is believed to arise
propane,67 and 98% in diethyl ether,62,68 although the from its high molecular symmetry, which renders it
safety concern over the high vapor pressure of diethyl a better stabilized crystalline lattice.80, 81
ether renders it an impractical candidate. The forma- EC was considered as an electrolyte cosolvent for
tion of dendritic lithium seemed to be sufficiently the first time by Elliot in 1964, who noted that, due
suppressed in these solvents even at high charge to the high dielectric constant and low viscosity of
rates.62 However, efforts to incorporate ether-based EC, the addition of it to electrolyte solutions would
electrolytes in lithium cells were still troubled by the favor ion conductivity.82 The findings did not attract
poor capacity retention,68-70 and prolonged cycling particular attention from the battery community
until the early 1970s, when Scrosati and Pistoia a lithium ion cell that employs graphite as an anode
exploited it to advantages for lithium battery elec- and 4.0 V metal oxide (LiMO2, M ) Co, Ni) as a
trolytes. They reported that, owing to the suppression cathode, the electrolyte must have oxidative stability
of the melting point by the presence of the solute, a up to ∼5 V vs Li.95-97
room-temperature melt would form, and extra sup- In 1994 a formulation that successfully met such
pression could be obtained when a small percentage a standard was first described in open literature by
(9%) of PC was added.83 Further investigation found Tarascon and Guyomard, who used a linear carbon-
that electrolytes based on EC as compared with PC ate, dimethyl carbonate (DMC), as a cosolvent with
demonstrated improvements, not only in bulk ion EC.98,99 As it has been pointed out, linear carbonates
conductivity but also in interfacial properties such differ from their cyclic cousins by their low boiling
as lower polarization on various cathode surfaces.84 points, low viscosity, and low dielectric constant.
Following these reports, EC began to appear as an They can form homogeneous mixtures with EC at any
electrolyte cosolvent in a number of new electrolyte ratio, and the resultant mixed electrolytes benefit not
systems under investigation, many of which still only from the melting-temperature suppression of EC
contained ethers.72,73,85-88 However, the first com- but also from the low viscosity (higher ion conductiv-
mercialized rechargeable lithium battery used an ity) of DMC. But what surprises researchers is the
ether-free composition, an EC/PC mixture, as the wide electrochemical stability window of this mixture
electrolyte solvent.89,90 Despite the melting-point electrolyte: it remains stable on a spinel cathode
suppression by the solute and other cosolvents, the surface up to 5.0 V. Considering that these linear
higher liquidus temperatures of the electrolyte due carbonates, in the absence of EC, are readily liable
to EC remained a factor limiting the low-temperature to oxidation on cathode surfaces at ∼4.0 V vs Li,76
applications of the lithium cell. the origin for the above improvement in the electro-
The unique position of EC as a lithium battery chemical window remains unclear, because the an-
electrolyte was established in 1990 when Dahn and odic stabilities of the ether-based electrolytes were
co-workers reported the fundamental difference be- hardly raised by their mixing with EC93,94 or PC.95,96
tween EC and PC in their effects on the reversibility It seems that a synergistic effect is achieved when
of lithium ion intercalation/deintercalation with gra- EC and DMC (or other linear carbonates) are mixed
phitic anodes.36 Despite the seemingly minute dif- because the merits of each individual solvent are
ference in molecular structure between the two, EC imparted on to the resultant mixture: high anodic
was found to form an effective protective film (SEI) stability of EC on cathode surfaces, high solvation
on a graphitic anode that prevented any sustained power of EC toward lithium salts, and low viscosity
electrolyte decomposition on the anode, while this of DMC to promote ion transport.
protection could not be realized with PC and the This new formulation of electrolytes based on a
graphene structure eventually disintegrated in a mixture of EC with a linear carbonate set the main
process termed “exfoliation” because of PC cointer- theme for the state-of-the-art lithium ion electrolytes
calation. The reason for the effectiveness of the SEI and was quickly adopted by the researchers and
has incited a lot of research interest in the past manufacturers.97,100-103 Other linear carbonates were
decade but remains an unsolved mystery, although also explored, including DEC,104-106 ethylmethyl
it is generally believed that EC undergoes a reduction carbonate (EMC),107 and propylmethyl carbonate
process on the carbonaceous anode via a similar path (PMC),108,109 and no significant differences were found
to that shown in Scheme 1. Because of the important between them and DMC in terms of electrochemical
role this SEI plays in lithium ion chemistry, the characteristics. The direct impact of this electrolyte
research efforts on this topic will be reviewed in a innovation is that the first generation carbonaceous
dedicated section (section 6). anode petroleum coke was soon replaced by graphitic
anode materials in essentially all of the lithium ion
2.1.4. Linear Carbonates cells manufactured after 1993. At present, the elec-
trolyte solvents used in the over one billion lithium
After Sony successfully marketed the first genera- ion cells manufactured each year are almost exclu-
tion lithium ion cells, numerous competitors emerged sively based on the mixture of EC with one or more
and a pursuit for higher energy density started. With of these linear carbonates, although each individual
the energetic advantage of highly crystalline carbon manufacture may have its own proprietary electro-
(graphitic) over disordered carbon being recognized, lyte formulation.
EC became the core and indispensable component of
the electrolyte formulation. 2.2. Lithium Salts
During the early 1990s, efforts were made to An ideal electrolyte solute for ambient rechargeable
expand the limited liquid range of EC-based electro- lithium batteries should meet the following minimal
lytes by using different cosolvents, including PC,91,92 requirements: (1) It should be able to completely
THF and 2-Me-THF,73,85-88 diethoxyethane (DEE),93,94 dissolve and dissociate in the nonaqueous media, and
and dimethoxyethane (DME).45,95-97 None of these the solvated ions (especially lithium cation) should
cosolvents performed satisfactorily though, because be able to move in the media with high mobility. (2)
the presence of PC usually caused a large irreversible The anion should be stable against oxidative decom-
capacity in the initial cycle of the lithium ion cell,36,91,92 position at the cathode. (3) The anion should be inert
while the ethers were found to be unstable against to electrolyte solvents. (4) Both the anion and the
the oxidation catalyzed by the surface of the charged cation should remain inert toward the other cell
cathode.93,94 Thus, it was generally realized that, for components such as separator, electrode substrate,
Table 3. Lithium Salts as Electrolyte Solutes
a Reference 111. b Reference 146. c Reference 114. d Reference 115. e Reference 116.
and cell packaging materials. (5) The anion should bases such as Cl- does not fully neutralize their
be nontoxic and remain stable against thermally activity, and as a result, they would attack most of
induced reactions with electrolyte solvents and other the nonaqueous solvents, especially ethers. The AlX4-
cell components. anions also cause severe corrosion to other cell
The available choice of lithium salts for electrolyte components such as the separators, usually made of
application is rather limited when compared to the polypropylene, and the insulating sealant, as well as
wide spectrum of aprotic organic compounds that the metallic packaging materials. On the other hand,
could make possible electrolyte solvents. This differ- anions based on milder Lewis acids can remain stable
ence could be more clearly reflected in a comprehen- with organic solvents under normal conditions (e.g.,
sive report summarizing nonaqueous electrolytes ambient temperature) and have been preferentially
developed for rechargeable lithium cells, in which investigated by researchers. These salts include
Dahn and co-workers described over 150 electrolyte lithium perchlorate (LiClO4) and various lithium
solvent compositions that were formulated based on borates, arsenates, phosphates, and antimonates,
27 basic solvents but only 5 lithium salts.50b LiMXn (where M ) B or As, P, and Sb and n ) 4 or
Because of the small ionic radius of lithium ion, 6, respectively). Table 3 lists some examples of these
most simple salts of lithium fail to meet the minimum salts along with some basic physical properties,50c
solubility requirement in low dielectric media. Ex- including ion conductivity data at room temperature
amples are halides, LiX (where X ) Cl and F), or the in PC or EC/DMC (1:1), respectively.50d,111-116 A brief
oxides Li2O. Although solubility in nonaqueous sol- summary of a few selected lithium salts of signifi-
vents would increase if the anion is replaced by a so- cance during the development of lithium cell electro-
called “soft Lewis base” such as Br-, I-, S2-, or lytes is given below.
carboxylates (R-CO2-), the improvement is usually
realized at the expense of the anodic stability of the 2.2.1. Lithium Perchlorate (LiClO4)
salt because these anions are readily oxidized on the LiClO4 has been a popular electrolyte solute owing
charged surfaces of cathode materials at <4.0 V vs to its satisfactory solubility and high conductivity
Li. (∼9.0 mS cm-1 in EC/DMC at 20 °C) as well as its
Most of the lithium salts that are qualified for the high anodic stability (up to 5.1 V on a spinel cathode
minimal solubility standard are based on complex surface in EC/DMC).99 Recent studies found that SEI
anions that are composed of a simple anion core films formed in LiClO4 electrolytes, on both lithium
stabilized by a Lewis acid agent. For example, the and carbonaceous anode surfaces, are of lower im-
anion of lithium hexafluorophosphate (LiPF6) could pedance than those formed in LiPF6 or lithium
be viewed as F- complexed by the Lewis acid PF5. tetrafluoroborate (LiBF4) electrolytes, because the HF
Such anions, also known as anions of superacids, is absent in the former.104,117 It is believed that HF,
have a structure in which the formal negative charge generated as the hydrolysis product of LiPF6 and
is well distributed by the strongly electron-withdraw- LiBF4 by trace moisture in the electrolyte solvents,
ing Lewis acid ligands, and the corresponding com- reacts with either alkyl carbonate or Li2CO3 and
plex salts are usually lower melting and better forms the highly resistive LiF.117,118 Compared with
soluble in low dielectric media than their parent other lithium salts, LiClO4 also has the merits of
salts. being relatively less hygroscopic and is stable to
The requirement for chemical inertness further ambient moisture.
excluded a family of lithium salts that have been However, the high oxidation state of chlorine (VII)
widely used in primary lithium batteries: LiAlX4 in perchlorate makes it a strong oxidant, which
(X ) halides).110 Since the Lewis acidities of the AlX3 readily reacts with most organic species in violent
are so strong, their complexation with the moderate ways under certain conditions such as high temper-
Scheme 2. Reaction between LiAsF6 and Ether The anodic stability of the AsF6- anion proved to
Solvents be high. In proper solvents, such as esters rather
than ethers, the electrolyte based on this salt can
remain stable up to 4.5 V on various cathode sur-
faces.78,99 The combination of cathodic and anodic
stability would have made LiAsF6 a very promising
candidate salt for both lithium and lithium ion
batteries had the toxicity not been a source of
concern. Instead, it was never used in any com-
mercialized cells but is still frequently used in
ature and high current charge.58,119 Actually, back in laboratory tests even today.14-18,123-125
the 1970s it had already been realized that LiClO4
was impractical as an electrolyte solute for industry 2.2.3. Lithium Tetrafluoroborate (LiBF4)
purposes;119 nevertheless, it is still frequently used
as a salt of convenience in various laboratory tests Like LiAsF6, LiBF4 is a salt based on an inorganic
because it is easy to handle and economical.43,79,91,94 superacid anion and has moderate ion conductivity
in nonaqueous solvents (Table 3). It was out of favor
2.2.2. Lithium Hexafluoroarsenate (LiAsF6) in the early days of lithium battery research because
the ether-based electrolytes containing it were found
While researchers focused on the morphology of to result in poor lithium cycling efficiencies, which
lithium cycling in nonaqueous electrolytes during the decayed rapidly with cycle number.60,126-128 The
late 1970s, it was found that the salt plays an reactivity of LiBF4 with lithium was suspected as
important role beside the solvents, and in general, discoloration occurred with time or heating.128
LiAsF6 was a superior salt to LiClO4 as an electrolyte The researchers who initiated the study on LiBF4
solute for lithium batteries.54 For a long period, the mentioned the multiple advantages of LiBF4 as
combination of LiAsF6 with various ethers became compared with other salts (e.g., less toxicity than
the most popular system under investigation.60,62,68,69,78 LiAsF6 and higher safety than LiClO4),128 but its
On average, lithium cycling efficiencies could reach moderate ion conductivity has been a major obstacle
>95% in these systems,54,62,66 although long-term to its application. More recent studies on its ionics
cycling in these electrolytes still promoted the growth and limiting properties in various nonaqueous sys-
of lithium dendrites.71 Chemical deterioration was tems established that, among the most common
also detected, as indicated by the discoloration of the anions encountered, BF4- has the highest mobility,
LiAsF6/2-Me-THF solution with time, and a reaction but its dissociation constant is significantly smaller
between LiAsF6 and the solvent was suspected.65,68 than those of LiAsF6 and LiPF6.111,129 The unfavorable
A mechanism was proposed based on the Lewis balance of these two properties results in the moder-
acidity of As(V), which cleaves the ether linkage and ate ion conductivity.
produces a series of gaseous and polymeric products Electrochemically, the BF4- anion was found to be
(see Scheme 2).68 stable against oxidation on a glassy carbon (GC)
The cathodic stability of the AsF6- anion was surface up to 3.6 V vs a standard calomel electrode
studied on a glassy carbon surface, and a reduction (SCE), which translates into ∼5.0 V vs lithium.130,131
process was found at ∼1.15 V vs Li:120 When a distinction is made, this stability limit is
somehow lower than those of AsF6- and PF6- anions;
AsF6- + 2e h AsF3 + 3F- however, caution must be exercised here, as these
data were measured on GC with quaternary am-
The above process was observed only in the initial monium as supporting electrolyte, instead of on a
cycles. Nevertheless, any electrochemical reduction surface of cathode materials. This could result in
of As(V) would raise concern about the safety of using substantial difference.76
LiAsF6 in a commercial battery, because, while The use of LiBF4 in lithium-based cells has been
arsenate in its high oxidation state (V) is not par- rare because of its inferior ion conductivity until
ticularly toxic, As(III) and As(0) species are.68,120-122 recently, when the thermal instability of LiPF6 and
From the electrochemical point of view, however, the the moisture sensitivity became recognized. Attempts
above reduction could be a benefit, especially for to replace LiPF6 in lithium ion cells have been made,
lithium ion cells, since an SEI formed on an anode and the cells based on LiBF4 electrolytes showed
at >1.0 V vs lithium would be very stable during the improved performance, not only at elevated temper-
operation of a lithium ion cell according to a semi- atures up to 50 °C132 but, surprisingly, also at low
empirical rule,106 which will be discussed in more temperatures as well.132-135 These observations could
detail in section 6. bring this salt back to research favor.
Very similar to the case of LiClO4, an SEI formed
from LiAsF6-based electrolytes, either on a lithium 2.2.4. Lithium Trifluoromethanesulfonate (LiTf)
or carbonaceous anode, mainly consists of alkyl
carbonates or Li2CO3 rather than LiF, as one would Another family of lithium salts is based on the
expect from the behavior of its close structural conjugate bases of the organic superacids, where acid
brothers LiPF6 or LiBF4.104,117,118 This can be at- strength is increased because of the stabilization of
tributed to the much less labile As-F bond that is anions by the strongly electron-withdrawing groups,
resistive to hydrolysis.120 usually perfluorinated alkyls. In these anions, the
delocalization of the formal negative charge is practi- Scheme 3. Resonance States of the Imide (Im)
cally realized by a combination of the inductive effect Anion
of the electron-withdrawing groups and the conju-
gated structures:
The attempt to use these salts originated from the

hope that their dissociation constants would be high
even in low dielectric media, and the organic nature
of perfluorinated alkyls would always assist the anode (carbon fiber) was used, and LiTf-based elec-
solubility of the salts in nonaqueous solvents. Be- trolytes in various solvent mixtures (EC/DMC, PC/
cause of the requirement for electrochemical stability, DMC, or EC/DME) showed better Coulombic effi-
lithium carboxylates (RF-CO2Li, where RF- ) perflu- ciency (∼98%) and discharge capacity.140
orinated alkyls) are excluded from consideration, The anodic stability of the Tf- anion, as measured
because their oxidation still occurs at ∼3.5 V vs on a GC surface, was not found to be particularly
lithium, which is similar to the cases of their non- high:130 inferior to BF4- and PF6- but better than
fluorinated counterparts.22 Obviously, the electron- ClO4-. Ab initio calculations yielded similar conclu-
withdrawing groups do not stabilize the carboxylate sions,131 and results measured on porous carbon
anions sufficiently to alter their oxidative stability. electrodes were consistent with those measured on
On the other hand, sulfonate (-SO3Li) became the GC.81
anion of choice because it is highly resistant to The real obstacle that eventually eliminated LiTf
oxidation, thermally stable, nontoxic, and insensitive as a candidate for lithium ion battery application is
to ambient moisture as compared with LiPF6 or the serious aluminum corrosion that occurred in LiTf-
LiBF4.122 As the simplest member of this category based electrolytes. Used as the substrate material to
(RF ) CF3), lithium triflate (LiTf) received extensive carry cathode active material in the cell, Al is
research as a candidate for lithium/lithium ion cells. constantly subject to high potentials during cell
Other similar salts studied include perfluoroethyl operation and, thus, must maintain inertness against
sulfonate (RF ) C2F5), perfluorobutylsulfonate (RF ) anodic dissolution. However, because of a special
C4F9),129-131 and the oligomeric versions that are interaction between the Tf- anion and Al, the latter
based on polyether linkages.136,137 dissolves anodically at ∼2.7 V and starts to pit at
One major drawback of these sulfonate salts is ∼3.0 V.141At 4.0 V vs lithium, the normal working
their poor ion conductivity in nonaqueous solvents potential of a charged cell, the anodic dissolution
as compared with other salts. In fact, among all the current is ∼20 mA/cm2 in PC. This high corrosion
salts listed in Table 3, LiTf affords the lowest rate leaves LiTf virtually little possibility of being
conducting solution. This is believed to be caused by used as an electrolyte solute for any high voltage
the combination of its low dissociation constant in cells, either lithium or lithium ion.
low dielectric media122,138 and its moderate ion mobil-
ity129 as compared with those of other salts. Serious
2.2.5. Lithium Bis(trifluoromethanesulfonyl)imide (LiIm)
ion pairing in LiTf-based electrolytes is expected,
and Its Derivatives
especially when solvents of low dielectric constant In 1984, Foropoulos and DesMarteau reported a
such as ethers are used.111,122 new acid based on an imide anion stabilized by two
The stability of the Tf- anion toward lithium has trifluoromethanesulfonyl (triflic) groups.142 Because
been studied by both electrochemical and spectro- of the strong electron-withdrawing nature of the
scopic means.139 While the surface analysis using triflic groups and the conjugation between them and
X-ray photoelectron spectra (XPS) found only a trace the lone electron pair on the nitrogen, the formal
of LiF, in addition to Li2CO3, present on the lithium negative charge in the anion is well delocalized, as
surface, indicating that this anion was not reactive shown by the resonance structures in Scheme 3.
with lithium, a poor lithium morphology was identi- As a result, the acid strength of the proton is
fied in LiTf-based electrolytes. Electrochemical quartz approximately equivalent to that of sulfuric acid in
crystal microbalance (EQCM) detected a very rough nonaqueous media.142 In view of the excellent mis-
surface of the cycled lithium, most likely related to cibility of this anion with organic nonpolar materials,
the formation of dendritic lithium. A simultaneous Armand et al. proposed using its lithium salt (later
impedance measurement confirmed the poor mor- nicknamed “lithium imide”, or LiIm) in solid polymer
phology with ever-increasing interfacial impedance electrolytes, based mainly on oligomeric or macro-
due to the incessant reactions between the rough molecular ethers.143 In no time, researchers adopted
lithium surface and the electrolyte. This sustained its use in liquid electrolytes as well, and initial
reaction resulted in a thick and resistive SEI film results with the carbonaceous anode materials seemed
that was not favorable to battery operation. However, promising.95 The commercialization of this new salt
in at least one case, LiTf was found to outperform by 3M Corporation in the early 1990s sparked
the state-of-the-art solute LiPF6, where an atypical considerable hope that it might replace the poorly
conducting LiTf, the hazardous LiClO4, the thermally Scheme 4. Decomposition of Carbonates by PF5
unstable LiBF4 and LiPF6, and the toxic LiAsF6 in
lithium battery applications.122 Extensive studies of
this salt were carried out to determine its ionics in
nonaqueous solutions111,116,129,130,131,139 and its applica-
tions in lithium or lithium ion cells.45,96,144,145
LiIm proved to be safe, thermally stable, and highly
conducting: it melts at 236 °C without decomposition Scheme 5. Hydrolysis of LiPF6 Salts by Moisture
(a rarity among lithium salts) and does not decom-
pose until 360 °C.146 Its ion conductivity in THF is
an order of magnitude higher than that of LiTf,
although lower than those of LiAsF6 and LiPF6,111,116,122
and when no Al is used, a lithium ion cell based on of well-balanced properties with concomitant com-
a LiNiO2 cathode and a petroleum coke anode can promises and restrictions. For example, in the com-
yield up to 1000 deep discharge cycles with LiIm in monly used carbonate solvent mixtures it has a lower
an EC/DMC solution.45 conductivity than LiAsF6 (Table 3),99,111,116 a lower
Ionics studies by Webber and Ue revealed that this dissociation constant than LiIm,129 a lower ionic
salt dissociates very well even in low dielectric mobility than LiBF4,129 a lower thermal stability than
solvents, although its large anion size usually results most of the other salts,149 a lower anodic stability
in a higher solution viscosity than those of other salts than LiAsF6 and LiSbF6,130,131 and a lower chemical
in a given solvent system. Thus, its good ion conduc- stability toward ambient moisture than LiClO4, LiIm,
tivity should be the result of a compromise between and LiTf. However, none of these other salts could
a high degree of dissociation and low mobility.122,129 meet all these multifaceted requirements simulta-
In this sense, LiIm favors solvents with a low neously as well as LiPF6 does.
dielectric constant. Electrochemical stability tests LiPF6 was proposed as an electrolyte solute for
were carried out on a GC electrode, and Im- was lithium-based batteries in the late 1960s,149 and soon
found to be stable against oxidation in EC/DMC up its chemical and thermal instabilities were known.150
to 2.5 V vs a Ag+/Ag reference, which translates to Even at room temperature, an equilibrium exists:
∼5.0 V vs Li, an oxidation limit lower than those for
LiBF4 and LiPF6,130,131 but still high enough to be LiPF6(s) h LiF(s) + PF5(g)
practical. The morphology of cycling lithium in LiIm-
based electrolytes is apparently superior to that in The generation of the gaseous product, PF5, drives
other salt-based electrolytes.139 the equilibrium to the right, and this process is
favored by elevated temperatures. In the presence of
Despite all of these merits, the application of LiIm
nonaqueous solvents, the strong Lewis acid PF5 tends
in lithium ion cells never materialized because it
to initiate a series of reactions, such as ring-opening
caused severe Al corrosion in electrolytes based on
polymerization or cleavage of ether linkages (Schemes
it.141 In situ surface studies using EQCM established
2 and 4).68,151,152
a reaction between the Im- anion and the Al sub-
On the other hand, the P-F bond is rather labile
strate in which Al(Im)3 is produced and adsorbed on
toward hydrolysis by even trace amounts of moisture
the Al surface.147 Undoubtedly, this corrosion of a key
in nonaqueous solvents, producing a series of cor-
component of the cell by Im- greatly restricts the
rosive products (Scheme 5). Thermal gravimetric
possible application of LiIm, because the role of Al
analysis (TGA) reveals that, in a dry state, LiPF6
as a cathode substrate in the lithium-based battery
loses 50% of its weight at >200 °C141 but that, in
industry is hard to replace, due to its light weight,
nonaqueous solutions, the deterioration occurs at
resistance to oxidation at high potential, excellent
substantially lower temperatures, for example, as low
processability, and low cost.
as 70 °C.
Efforts were made to reduce the reactivity of Im- The sensitivity of LiPF6 toward ambient moisture,
toward Al. Other salts that could passivate Al were solvents, and high temperature not only restricts its
used as additives with LiIm, and the results were range of applications, especially in nonaqueous bat-
encouraging.147 Also, structural modification of the teries, but also causes tremendous difficulty in its
imide anion was made by extending the perfluori- preparation and purification.150 Before the 1990s,
nated alkyl chain and was found to be effective, most of the commercially available LiPF6 had high
although at the price of lower ion conductivity.141,148 amounts of LiF and HF, and the purity issue became
Although LiIm has never been used in any compart of the reason the potential of LiPF6 was not fully
mercial lithium ion devices, it remains an interesting realized until recently. The manufacture of high-
salt to be investigated, especially for the polymer- purity LiPF6 (HF < 10 ppm) in industrial scale was
based electrolytes.47 achieved by Japanese companies in the late 1980s,
2.2.6. Lithium Hexafluorophosphate (LiPF6) finally leading to the commercialization of lith-
ium ion technology and the ensuing extensive re-
Among the numerous salts vying for lithium/ search.86,120,122
lithium ion batteries, LiPF6 was the obvious winner In nonaqueous solvents based on mixed alkyl
and was eventually commercialized. The success of carbonates, LiPF6 remains one of the most conducting
LiPF6 was not achieved by any single outstanding salts. For example, in EC/DMC (1:1) the conductivity
property but, rather, by the combination of a series is 10.7 mS cm-1, only fractionally lower than that of
LiAsF6. According to the ionics studies on the limiting only deliver their rated capacity and power in the
properties in various solvents, this excellent conduc- temperature range -20 to 50 °C. Below temperatures
tivity results from the combination of its ionic mobil- from -20 to -30 °C, both the capacity utilized and
ity and dissociation constant, although in neither the rate at which it is delivered (power) are deci-
category does LiPF6 stand at the most outstanding mated. This reduction in performance at low tem-
position:111,129 peratures is usually temporary, and rated capacity
can be recovered once the battery is brought back to
Average ion mobility: LiBF4 > LiClO4 > >20 °C. However, at temperatures higher than 60
LiPF6 > LiAsF6 > LiTf > LiIm °C, the performance deterioration is permanent,
because the reactions between the electrolyte solute
Dissociation constant: LiTf < LiBF4 < LiClO4 < and solvents are irreversible. Furthermore, the de-
composition products of these irreversible processes,
LiPF6 < LiAsF6 < LiIm which are often gaseous, can lead to hazardous
pressure build-up within the batteries.
The reversed order in the above two properties clearly
demonstrates the conflicting nature of the require-
ments and the advantage of the well-balanced prop- 3. Liquid Range of Electrolyte Solutions
erties of LiPF6.
Electrochemical studies on a GC electrode and The liquid range of a nonaqueous electrolyte sys-
various metal oxide-based cathode surfaces confirm tem is defined at the upper limit by the temperature
that the solution of LiPF6 in mixed carbonates can at which one of its components begins to vaporize
effectively resist oxidation up to 5.1 V,99,130 thus (also called the bubble temperature, θb) and at the
making it one of the few salts that can actually lower limit by the temperature at which one of its
support the operation of 4.0 V cathode materials. components begins to crystallize (liquidus tempera-
LiPF6 also effectively passivates an Al substrate at ture, θl).158 Apparently, along with restrictions im-
high potentials.141,153,154 It is generally accepted that posed by other factors, this range could serve as the
the anion participated in forming a protective film main basis for estimating the operating limits of
on Al, although there is no concensus as yet concern- lithium-based devices that employ such an electrolyte
ing the mechanism of the process.141,153-156 system. Surprisingly, despite the significance of this
The above merits made LiPF6 the salt of choice issue to practical applications, there have been rather
when lithium ion technology leaped from concept into few studies dedicated to the thermophysical proper-
product. In 1990, it was used by Sony in the first ties of the electrolytes, especially considering the
generation lithium ion cell,157 and since then, its large amount of effort spent on studying ion conduc-
position in the lithium ion industry has remained tion and electrochemical properties.
unchallenged. Like EC as an indispensable solvent Tarascon and Guyomard were perhaps the first
component, LiPF6 has become the indispensable researchers to try to delineate a temperature range
electrolyte solute for almost all lithium ion devices for lithium ion electrolytes.99 Following their seminal
manufactured in the past decade. formulation of new electrolytes based on cyclic and
linear carbonates, they measured boiling (bp) and
melting points (mp) for LiPF6/EC/DMC solutions as
2.3. Brief Summary the function of EC/DMC compositions. They found
After more than a decade of exploration, the that the bp of the electrolytes as a function of
skeletal components of the electrolyte for the com- composition exhibits a monotonic decrease from that
mercialized lithium ion devices have been identified. of EC (248 °C) to that of DMC (91 °C) with such a
Within the various brands of lithium ion cells, the severe curvature that, for most of the compositions,
exact electrolyte composition differs from manufac- it is dominated by the lower boiling component. The
turer to manufacturer, and the formulas remain same trend was observed in a similar system (PC/
proprietary information; however, the overwhelming DEC) by Ding recently,158 who made the observation
majority of these are apparently based on two indis- that enhancing the bp for a mixed solvent system by
pensable components: EC as the solvent and LiPF6 adding a higher boiling component would have little
as the solute. In most cases, one or more linear effect because of the previously mentioned depen-
carbonates, selected from DMC, DEC, or EMC, are dence. In other words, the upper limit of the liquid
also used as cosolvents to increase the fluidity and range for a binary system is mainly determined by
reduce the melting point of the electrolyte, thus the bp of the lower boiling component. There is
forming the popular composition consisting of LiPF6/ sufficient reason to believe that this rule would hold
EC/linear carbonate(s). true for ternary or higher order systems.
However, certain restrictions on battery perfor- It must be pointed out that, in the state-of-the-art
mance arise from these state-of-the-art electrolytes, electrolytes, the actual high-temperature limits for
for which these two indispensable components are application in cells are usually not set by the upper
mainly responsible: (1) a low-temperature limit (-20 boundary of the liquid range because other factors
°C) set by EC due to the high melting point and the might push the limits far lower. For example, the
high liquidus temperature it confers upon the solvent upper boundary of the liquid range is ∼90 °C for
mixture, and (2) a high-temperature limit (50 °C) set DMC-, 110 °C for EMC-, and 120 °C for DEC-based
by LiPF6 due to its reactivity with solvents. As a electrolytes, all of which are far above the high-
result, the commercialized lithium ion batteries can temperature limit set by the salt LiPF6 (70 °C).152
Even if LiPF6 is replaced by more thermally stable

salts, the thermal stability of passivation films on
both the anode and the cathode would still keep the
high-temperature limits lower than 90 °C, as do the
thermal stability of the separator (<90 °C for polypro-
pylene), the chemical stability of the insulating
coatings/sealants used in the cell packaging, and the
polymeric binder agents used in both cathode and
anode composites.
The lower boundary of the liquid range, on the
other hand, does usually serve as the low-tempera-
ture limit for the electrolytes. The mp of LiPF6/EC/
DMC was determined as a function of solvent com-
position by Tarascon and Guyomard, who concluded
that LiPF6/EC/DMC could be used in the solvent
compositions between 3:7 and 8:2 at temperatures
down to -25 °C.99 The mp-dependence on solvent
composition that was reported in the work, however,
does not seem to be rational, since it shows a long
plateau at ∼ -10 °C in 30-90% DMC after an initial
drop from 20 °C at 20% DMC,99 whereas a typical
simple eutectic feature would be expected instead.
When a closer comparison is made between the
results by Tarascon and Guyomard and the more Figure 4. Liquid-solid phase diagrams of EC/DMC, EC/
recent studies by Ding et al.,50e,159,160 it becomes EMC, and PC/EC. (Reproduced with permission from ref
159 (Figure 9). Copyright 2000 The Electrochemical Soci-
obvious that in the former work no distinction was ety.)
made between the liquidus and solidus temperatures;
that is, the mp’s at 20% and 90% DMC are liquidus
insoluble in their solid states; therefore, solid solu-
temperatures, while the rest (the plateau section)
tions do not form below solidus temperatures, and
seem to be solidus temperatures as measured by Ding
the binary system exists in the form of a heteroge-
et al. This confusion could have arisen from the fact
neous mixture of both solids.
that, for the intermediate compositions between 30%
Typically, the liquidus lines of a binary system
and 80% DMC, the liquidus transition usually ap-
curve down and intersect with the solidus line at the
pears as a broad peak of relatively negligible thermal
eutectic point, where a liquid coexists with the solid
effect that could evade notice, while the solidus
phases of both components. In this sense, the mixture
transition usually appears as a sharp and conspicu-
of two solvents should have an expanded liquid range
ous peak.
with a lower melting temperature than that of either
Systematic construction of carbonate mixture phase solvent individually. As Figure 4 shows, the most
diagrams was performed by Ding et al.50e,159,160 for a popular solvent combination used for lithium ion
series of binary carbonate solvent systems that are technology, LiPF6/EC/DMC, has liquidus lines below
frequently used as lithium ion electrolyte solvents. the mp of either EC or DMC, and the eutectic point
It was found that all of the binary combinations lies at -7.6 °C with molar fractions of ∼0.30 EC and
between EC, PC, DMC, EMC, and DEC yield the ∼0.70 DMC. This composition corresponds to volume
same basic feature of simple eutectic-type phase fractions of 0.24 EC and 0.76 DMC or weight frac-
diagrams, characterized by the V-shaped liquidus tions of 0.28 EC and 0.71 DMC. Due to the high mp
lines intersecting at the eutectic composition with a of both EC (36 °C) and DMC (4.6 °C), this low-
horizontal solidus line, although the details of each temperature limit is rather high and needs improve-
individual diagram vary greatly depending on their ment if applications in cold environments are to be
mp and cyclicity. Figure 4 shows a collection of such considered.
phase diagrams mapped for some EC-based binary A rather counterintuitive conclusion that can be
solvent systems. extracted from these phase diagrams, however, is
In these phase diagrams, the liquidus line repre- that simply introducing a low-melting component
sents the temperature at which one of the compo- does not necessarily extend the liquid range, as
nents crystallizes, while, below the solidus line, the evidenced by the replacement of DMC by the lower
whole system solidifies. Between the solidus and melting EMC (-53 °C) in an EC-based binary solvent
liquidus lines are the regions where solid and liquid system (Figure 4). The mismatch between the EC and
coexist. Since there is no solid phase above the EMC mp’s creates a liquidus line that approaches the
liquidus lines and the liquid is thermodynamically mp of EC for most of the compositions; therefore, the
stable, Ding et al. suggested that the liquidus tem- liquid range toward the low-temperature end actually
peratures should be adopted as the lower boundary shrinks as compared with that of the EC/DMC binary
of the liquid phase, instead of the solidus tempera- system. The replacement of DMC by another low-
tures.159,160 The patterns of these phase diagrams are melting linear carbonate, DEC (-74 °C), produces a
typical of binary systems in which the two compo- similar effect.23d It was proposed later that this poor
nents are mutually soluble in their liquid states but compatibility between EC and linear carbonates
originates from the intrinsic differences found at the below their liquidus temperatures without pro-
molecular level, that is, in the number of structural nounced deterioration.164,165
conformations that was allowed for by their freely However, an electrolyte is only kinetically stable
rotatable single bonds.158 below its liquidus line, and the presence of any
The comparison between the phase diagrams has nucleating agent could trigger the formation of a solid
led to the conclusion that the compatibility between phase. In an electrochemical device, the rough sur-
different solvents is determined by two factors: (1) face of the electrode composite, the fibers of the
molecular structure similarity and (2) melting point separator, or the edges of current collectors and tabs
proximity.50e,158-160 Thus, from the viewpoint of phase can all act as such a nucleating agent; therefore, the
diagram study alone, the effort to advance the liquid extended liquid range of electrolytes by supercooling
range toward lower temperature would be far more seems rather unreliable. Ding et al. studied the
effective if cyclic-cyclic or linear-linear binary sys- supercooling behavior of the EC/EMC system and the
tems were formulated rather than the cyclic-linear effect of various carbon particles on its diminution.166
combination that is currently used as the state-of- They found that, in this particular carbonate mix-
the-art lithium ion electrolytes. Unfortunately, most ture, only the liquidus temperature could be success-
of the possible formulations under these categories fully bypassed at a given cooling rate (10 °C/min),
would be vetoed by other considerations such as ion while, at the solidus temperature, binary crystalliza-
conductivity, electrochemical stability, and SEI for- tion occurs with negligible supercooling. Carbon
mation on electrodes, and the cyclic-linear combina- particles effectively diminish the supercooling by
tion remains the formulation of choice. Viewed from providing nucleation seeds and effectively promoting
another angle, the low-temperature limit on the the primary crystallization of EC at or near its
device will remain a challenge as long as the high- liquidus temperature. The effectiveness of these
melting EC remains an indispensable solvent com- carbon particles lies in the order
ponent for the lithium ion electrolytes, because it will
dominate the lower boundary of the liquid range no MCMB > activated carbon > carbon black
matter what linear cosolvents were used.
where MCMB stands for mesocarbon microbeads, a
The phase diagrams shown in Figure 4 only popular carbonaceous anode material used in the
describe the thermal properties of the binary solvents, lithium ion batteries.
while in practical applications the salt solutions of On the other hand, the presence of the salt, LiPF6,
these solvent mixtures would be of greater interest. assists the occurrence of supercooling by increasing
Ding et al. studied the effect of salt on the liquidus the solution viscosity and by depressing the liquidus
and solidus line locations at various concentrations temperature. At practical concentrations of LiPF6
up to 1.0 M and concluded that both lines would be (∼1.0 M), even the solidus temperature can be
depressed on temperature axes by a limited number circumvented, since there is no crystallization process
of degrees.159 At the salt concentrations investigated, observed for LiPF6/EC/EMC solution down to -120
it seems that the basic shape of the phase diagram °C, while the glass transition occurs at -103 °C. In
remains unchanged, while the liquidus and solidus such concentrated solutions, even the presence of
lines parallel those of the corresponding solvents. MCMB cannot initiate crystallization, and the su-
Thus, one can reliably estimate the phase transition percooling is completely suppressed at the cooling
temperatures of an electrolyte to be formed from a rate of 10 °C/min.
solvent with a known phase diagram. Thus, in actual applications, supercooling might
In the actual application of the electrolytes at low well exist when electrolytes are used at low temper-
temperatures, another thermal property that is closely atures and result in an extended liquid range.
related to the phase diagram and should not be However, it should be emphasized that such an
ignored is the so-called “supercooling” behavior,161-163 extended range remains fragile, since the supercool-
which occurs when the liquid circumvents crystal- ing as a kinetically stable behavior is highly condi-
lization at its mp or liquidus/solidus temperature, tional and unpredictable during the operation of the
usually because the high viscosity of the solvent device. The diminution or even elimination of super-
system at this temperature prevents the timely cooling is entirely possible if a slower cooling rate is
reorganization of molecules to ordered conformations. employed or long-term storage at low temperature
As a result, the solvent remains fluid at the temper- is exercised, but almost certainly, any prolonged
atures below their thermodynamic freezing point, and operation of an electrolyte below the liquidus line
this metastable liquid is called a “supercooled liquid”. would eventually encounter precipitation of solvent
The supercooled state ends when the glass transition components and result in performance deterioration.
occurs at some lower temperature and turns the Therefore, the reliable low-temperature limit should
system into a disordered (noncrystalline) solid phase. still be set by the liquidus lines as depicted in the
Since supercooling actually delays or even eliminates phase diagram.
crystallization of solvent components, the liquid Following the seminal formulation of electrolytes
range of the solvent system could be substantially based on carbonates by Tarascon and Guyomard,98,99
extended from the liquidus down to the glass transi- a lot of effort had been made to modify those binary
tion temperature (Tg). This extended range could lithium ion electrolytes. Often, ternary, quaternary,
potentially benefit the operation of the electrolyte at and even higher order solvent systems were pro-
low temperatures, and there are numerous examples posed, and the difficulty in constructing the phase
in which electrolytes were tested at temperatures far diagrams for these systems increased dramatically
of ionic compounds (usually crystalline salts) by polar

solvent molecules and (2) the migration of these
solvated ions through the solvent media.
During the solvation, the stability of the salt crystal
lattice is energetically compensated by the coordina-
tion of solvent dipoles with the naked ions (especially
cations); therefore, these ions should always migrate
with a “solvation sheath” around them, which con-
sists of a certain number of oriented solvent mol-
ecules. According to the results obtained from various
modeling approaches including ab initio quantum
mechanics, the small ionic radius of lithium usually
allows no more than four solvent molecules in its
solvation sheath.129,169-171 Using a new mass spec-
Figure 5. Calculated ternary phase diagram for EC/PC/
trum (MS) technique, a more recent determination
DMC as expressed in the form of a composition triangle of the coordination number (CN) for lithium ions
plane. The dotted lines represent the isotherms with 10 K seems to support these computational results: among
intervals with 300 K marked. (Reproduced with permission the various solvated lithium ion species, the peaks
from ref 167 (Figure 12). Copyright 2003 The Electrochemi- corresponding to Li(solv)2∼3+ are the most abundant
cal Society.) no matter whether the solvent is EC, PC, or γ-buty-
rolactone (γBL).172,173 This latter experimental ob-
due to the amount of experimental work required. servation also revealed the stability of the solvation
Consequently, the possibility of using computer sheath, which obviously remains intact even during
modeling to circumvent the laborious experimental the electrospraying under vacuum and the subse-
mapping was considered.167 Because a phase diagram quent ionization process. Therefore, there should be
is merely a graphic representation of thermodynam- sufficient confidence in the general belief that the
ics through minimizations in the free energy of the composition of solvated lithium ion remains un-
system under certain constraints, free-energy model- changed during its migration in an electrolyte solu-
ing of individual phases in a multicomponent system tion.1 Considering that both cation and anion could
should represent the phase boundary through phase be coordinated by solvents, ion conduction actually
equilibrium calculations. Since the electrolyte sys- consists of the oriented movement of ion/solvent
tems are usually used at constant pressure condi- complexes of both charges.
tions, the Gibbs free energy is the quantity minimized Ionic conductivity σ, which quantifies the ion
in calculating phase equilibria, and the computa- conduction ability, reflects the influence of these two
tional approach commonly known as the CALPHAD aspects, that is, solvation/dissociation and the sub-
method was used by Liu. The initial attempt seemed sequent migration, in terms of the free ion number
to be successful because phase diagrams of unary ni and the ionic mobility µi,:
systems for neat DMC, EC, and PC, and of binary
systems of EC/DMC, PC/DMC, and EC/PC were σ ) ∑niµiZie (1)
reproduced with satisfactory accuracy when com- i
pared with the work of Ding et al. Furthermore, by
combining the three binary systems, a ternary phase where Zi is the valence order of ionic species i, and e
diagram for EC/PC/DMC was predicted as shown in is the unit charge of electrons. For a single salt
Figure 5. As one would expect, the eutectic composi- solution, the cations and anions are the only two
tion was heavily PC-rich because PC was the lowest charged species present.174
melting of the three components. Although the eu- Ion conductivity has essentially become the quan-
tectic composition as predicted in the ternary phase tity used as the field-trial standard for any prospec-
diagram (0.10 EC/0.15 DMC/0.75 PC) was impracti- tive electrolytes, because it can be easily measured
cal for cell applications because of PC’s destructive with simple instrumentation, and the results are
effect on most carbonaceous anodes, the effectiveness highly accurate and reproducible. The methodology
of PC, when used in smaller amounts (<30%), in and the fundamental principles involved with the
improving low-temperature performance had been measurement have been summarized in a detailed
confirmed.168 Undoubtedly, continued efforts to pre- review.175 On the other hand, no reliable method has
dict the thermal properties of ternary or higher order been available so far for the exact determination of
systems should be encouraged for prospective elec- ion mobility (or a related property, diffusivity Di) and
trolytes, as they would provide useful knowledge for ionization degree, especially in electrolyte solutions
estimating their operating temperature range. in the concentration ranges of practical interest.1
The lack of ionic mobility data causes a serious
4. Ion Transport Properties inconvenience when the ion conduction ability of an
electrolyte is evaluated, because the measured con-
The ability to conduct ions is the basic function of ductivity is the result of the overall migration of both
electrolytes, which would determine how fast the anions and cations, while for lithium batteries only
energy stored in electrodes can be delivered. In liquid the portion of the current that was carried by the
electrolytes, the transport of ions is realized via a lithium cation matters. This portion of the current
two-step process: (1) the solvation and dissociation from lithium ion movement, which determines the
rate at which the battery operates, is quantified by facilitated if the anion is well stabilized by electron-
the lithium ion transference number (tLi): withdrawing functionalities. Successful examples of
such anions include PF6- or Im-, whose lithium salts
µLi dissociate readily as compared to those based on the
tLi ) (2) parental anions (LiF and lithium alkylamide, respec-
∑µi tively).
On the other hand, the mobility of an ion is known
i
to vary inversely with its solvation radius ri according
There have been numerous efforts aimed at esti- to the Stokes-Einstein relation:1
mating the lithium ion transference numbers in
nonaqueous solutions, and the data obtained via 1
different approaches vary appreciably. Nevertheless, µi ) (3)
6πηri
it is generally accepted that, in much diluted non-
aqueous solutions, lithium ion transference numbers
range from 0.20 to 0.40, depending on the properties where η is the viscosity of the media. With the cation
of salts and solvents.1,111,129,138,176-178 In other words, species fixed, this approach seems to be of little use
the anions are much more mobile than the lithium to increase cation mobility. However, a larger anion
ions in nonaqueous electrolytes. The small cation with lower anion mobility shows the application of
current portion in nonaqueous electrolytes is believed this approach in another way, which results in a
to be caused by the high surface charge density on higher cation transference number, although the
the cations (especially lithium ion) due to their small overall conductivity could decrease because of the
ionic radii, so they are much more favorably solvated reduction in the anion contribution. This effect was
and must move at slower speed with the solvation observed in imide and its derivatives.138 The extrem-
sheath, while high populations of anions could re- ity of this approach was represented by the salts with
main relatively “naked”. Solvation enthalpy calcula- oligomeric or polymeric anions, where t+ approaches
tions for cations and anions support this argu- 1.0 but the overall ion conductivities suffer
ment:169,176,178 in typical carbonate solvents, the former drastically.136,137,179-182 Hence, the approach of using
range between 20 and 50 kcal mol-1 while the latter large anions to enhance tLi is not widely pursued in
are below 10 kcal mol-1. liquid electrolytes.
A lithium ion transference number significantly So far, very few attempts at improving ion conduc-
less than 1 is certainly an undesired property, tivity have been realized via the salt approach,
because the resultant overwhelming anion movement because the choice of anions suitable for lithium
and enrichment near electrode surfaces would cause electrolyte solute is limited. Instead, solvent composi-
concentration polarization during battery operation, tion tailoring has been the main tool for manipulating
especially when the local viscosity is high (such as electrolyte ion conductivity due to the availability of
in polymer electrolytes), and extra impedance to the a vast number of candidate solvents. Considerable
ion transport would occur as a consequence at the knowledge has been accumulated on the correlation
interfaces. Fortunately, in liquid electrolytes, this between solvent properties and ion conductivity,1 and
polarization factor is not seriously pronounced. the most important are the two bulk properties of
The unavailability of data on dissociation degree the solvents, dielectric constant and viscosity η,
and mobility has thus made ion conductivity an which determine the charge carrier number (ni) and
alternative metric that has been universally adopted ion mobility (µi), respectively.
by the battery research and development community In order for a solvated ion to migrate under an
to evaluate the transport ability of electrolytes. electric field, it must be prevented from forming close
However, it should always be remembered that such ion pairs with its counterions by the solvating
a metric of convenience is based on an unstated solvent. The effectiveness of the solvent molecule in
assumption; that is, the increase in the overall shielding the interionic Coulombic attraction is closely
conductivity should originate, at least partially, from related with its dielectric constant. The critical
the improvement in the cation conductivity. Quali- distance for the ion pair formation q is given by eq 4
tatively, this assumption holds true, since a correla- according to Bjerrum’s treatment, with the hypoth-
tion does usually exist between ion conductivity and esis that ion-pair formation occurs if the interionic
power performance in batteries, although quantita- distance is smaller than q:1,183
tively the distribution of this increase between anions
and cations is unknown. |zizj|e2
The efforts to improve ion conductivity have re- q) (4)
volved around eq 1, that is, aiming at increasing 8π0kT
either the salt dissociation degree (ni) or the ionic
mobility (µi). Since these two factors are decided where z, 0, k, and T are the valence orders of ions,
simultaneously by the physicochemical natures of the the dielectric constant of vacuum, Boltzmann’s con-
salt and solvents, different approaches involving stant, and temperature, respectively. Apparently, in
either of these electrolyte components have been a solvent with a higher dielectric constant, ions would
adopted. have a higher probability of staying free at a given
For lithium electrolytes, the only variable in salt salt concentration and ion association would be less
structure is the anion. In a given nonaqueous solvent likely to occur. Most of these solvents are of high
system, the dissociation of a lithium salt would be boiling temperature and high viscosity (Table 1).
of the solvent mixing effect on these two properties

were given:195,196
s ) (1 - x2)1 + x22 (5)

ηs ) η1(1-x2)η2x2 (6)
where s, ηs, i, ηi, and xi are the dielectric constant
and viscosity of the mixture or the pure solvent
components and the volume fraction of the individual
solvent component, respectively.
Either s or ηs varies with solvent composition in a
monotonic way, and the additive effect of each term
seems to be able to account for the manner in which
ion conductivity varies with solvent composition. At
low DME concentration in the PC/DME system, the
mixture has a high dielectric constant, so that the
salt dissociates more completely. However, in this
region, the high viscosity, which impedes ionic move-
ment, dominates the ion conduction (Figure 6a). With
Figure 6. (a) Optimization of ion conductivity in mixed increasing DME content, the dielectric constant
solvents: 1.0 M LiClO4 in PC/DME. (b) Dependence of remains relatively high, but the system viscosity falls
dielectric constant () and fluidity (η-1) on solvent composi- drastically and the solvated ions migrate with higher
tion. These plots are reconstructed based on the data mobility. As a result, a net increase in ion conductiv-
reported in refs 187 and 194, respectively.
ity is achieved. Further increases in DME content
When solvated ions migrate within the electrolyte, (or low PC content) result in a very low dielectric
the drag force applied by the surrounding solvent medium where the effect of ion pair formation
molecules is measured by solvent viscosity η. Thus, outweighs that of low viscosity; thus, ion conductivity
in a solvent of lower viscosity, the solvated ions would drops with the increase in DME content. Therefore,
move more easily in response to an applied electric the maximum in ion conductivity versus solvent
field, as expressed by the Einstein-Stokes relation composition as shown in Figure 6a is actually the
(eq 3). Solvents of low viscosity have always been result of the compromise between the effects of the
considered the ideal candidates for electrolyte ap- dielectric constant and viscosity. Such a compromise
plication; however, their actual use was restricted illustrates the superiority of mixed solvent versus
because most of these solvents have low dielectric single solvent electrolytes.
constants (Tables 1 and 2) and cannot dissociate ions A simple mathematical treatment based on this
effectively enough to prevent ion pairing. model successfully reproduced the variation in ion
Since a high dielectric constant and low viscosity conductivity with solvent composition as observed in
usually cannot be integrated into a single solvent,184 experiments on numerous different systems. It also
a solvent mixture, usually binary with one of the proposed that, in an ideal situation where no ion pair
components selected for and the other for η, was formation is present, the change in ion conductivity
used to formulate electrolytes for lithium batteries should follow a linear relation as predicted by the
with the hope that a balance between these two semiempirical Walden’s law:194
properties could be arrived at via such mixing.185-187
The concept was rapidly accepted by researchers of Ληs ) constant (7)
the 1980s, usually using cyclic carbonates for their
high and linear or cyclic ethers for their low where Λ is the ion conductivity as normalized against
η.72,73,78,87,89,127,188-191 In almost all the cases, ion salt concentration (molar conductivity). The Walden
conductivity in mixed solvents is superior to that in product (Ληs) can be viewed as an ion conduction
single solvents, as Figure 6a shows for LiClO4 in PC/ metric that is normalized against both the solvent
DME, and a maximum in conductivity is usually viscosity and the salt concentration (i.e., the free ion
realized in medium compositions. number if no ion pairing occurs); therefore, its value
A series of works by Matsuda et al. composed serves to the first approximation as a quantification
perhaps the first systematic study to explore the of ion dissociation degree in electrolytes with either
physical foundation for such a mixing effect.127,187,193,194 a given solvent or a given salt.111,122
Using PC/DME as a model system, they investigated Although the above findings came from studies on
the dependence of vapor pressure, dielectric constant, mixtures of the cyclic carbonate PC with ethers, they
and viscosity on solvent composition, and they cor- remain qualitatively true for mixtures of cyclic and
related these variations with ion conductions. It was linear carbonates, that is, compositions of the state-
found that the dielectric constant varied with solvent of-the-art lithium electrolytes. Most likely, it was the
composition by following an almost linear relation, work by Matsuda et al. that delineated the basic
with slight positive deviations, while viscosity always guidelines for electrolyte formulation, which eventu-
showed a pronounced negative deviation from what ally led to the formulations by Tarascon and Guyo-
a linear relation would predict (Figure 6b). For such mard using cyclic (high ) and linear (low η) carbon-
binary solvent systems, approximate quantifications ate mixtures.93,98,99
Further analysis of related studies seems to argue of surface plots based on the close fit of experimental
that the success of the high-/low-η combination data to a fourth degree trivariant polynomial func-
might not be due to the simple additive effect of these tion:
two properties but, rather, a synergistic action of
these two variables through a mechanism that in- σ ) f(m,x,T) (8)
volves the solvent’s preference for cations in its
solvation sheath. During the dissolution process of a where σ, m, x, and T are ion conductivity in mS cm-1,
certain salt lattice in mixed solvents, the solvation salt concentration in mol kg-1, the mole fraction of
of the ions by a solvent molecule of a higher dielectric EC, and temperature in °C, respectively. Figure 7
constant would be energetically favored over that by summarizes the changes in ion conductivity with
a molecule of a lower dielectric constant. Conse- these variables and exhibits a general trend which
quently, it would be reasonable to expect that, after has been observed repeatedly for various electrolyte
equilibrium is established in the solution, the ions systems. This trend allows the tailoring of salt
would have solvation sheaths that are mainly com- concentration and solvent composition to maximize
posed of the high- solvents. Modeling results using ion conductivity at a given temperature for practical
molecular quantum mechanics support this hypoth- interests:
esis by showing that the less favored solvent mol- (1) Salt Concentration (m). At low salt concen-
ecules (low ) actually could be readily replaced by trations (<1.0 m), the number of free ions increases
the favored ones.197 Thus, in the EC/EMC system, the with salt concentration; consequently, ion conductiv-
solvation shell should be predominantly composed of ity also increases until it peaks at a higher concen-
EC.171 The most forceful evidence of this comes from tration. After this conductivity maximum, any in-
the experimental observation, where a new MS crease in salt concentration results in higher ion
technique employing low-energy ionization was ap- aggregation and higher viscosity of the solution,
plied to electrolytes based on a series of binary which reduces both the free-ion number and the ionic
compositions, including EC/DEC, EC/DMC, PC/DEC, mobility simultaneously. The location of this maxi-
PC/DMC, γBL/DEC, and γBL/DMC. In each of these mum conductivity on the salt concentration axes
systems, the overwhelmingly abundant species de- (mmax) is decided by the dielectric constant of the
tected were the solvation complexes of lithium with solvents as well as temperature. Generally speaking,
the cyclic solvents (EC, PC, or γBL), with coordina- a higher dielectric constant would shift the occur-
tion numbers between 2 and 3.173 rence of ion pairing to higher salt concentrations,
This selective solvation of lithium ions by high- while a higher temperature reduces the solution
solvent molecules would exclude the solvents of low-η viscosity. The common result of both scenarios is the
from the solvation sheath and leave the latter as free, shift of mmax to higher salt concentrations.
noncoordinating solvent molecules. As a result, the (2) Solvent Composition (xEC). At a given tem-
media in which the solvated ions migrate are mainly perature, the solvent composition determines the
composed of these free solvent molecules, which outcome of the interplay between dielectric constant
impart their low-η to benefit the movement of the and viscosity; hence, a similar relation between σ and
solvated ions. In this way, a synergistic participation xEC as shown for the PC/DME system in Figure 6a
from both high- and low-η solvents contributes to should be expected, as is indeed the case. However,
the optimization of ion conduction. temperature and salt concentration have such a
pronounced effect on this dependence of conductivity
The implication of such a picture of the solution on solvent composition that sometimes this relation
structure on the microscopic level not only concerns will appear as monotonic in the given range of solvent
ion transport but also further relates to the electro- compositions.
chemical stability of the electrolytes in lithium ion For example, at a given salt concentration of 1.6
cells, because these solvent molecules in the solvation m, solvents with a higher xEC are favored at high
sheath, such as EC or PC, migrate with the ions to temperatures (>50 °C) because the influence of
electrode surfaces and are probably more involved in viscosity is less pronounced and σ increases mono-
the oxidative or reductive processes than the non- tonically with xEC. At low temperatures (<10 °C), this
coordinating, low-η solvent molecules, such as the relation is reversed because of the predominate role
linear carbonates. This could have a profound impact of viscosity. At intermediate temperatures between
on the chemical nature of the electrolyte/electrode 20 and 40 °C, σ peaks versus xEC, indicating that at
interfaces (section 6). neither high nor low xEC is the compromise between
Knowing how ion conduction is determined by the and η able to optimize ion conduction. Similarly,
interplay between the dielectric constant and viscos- salt concentration also affects the dependence of
ity, the dependence of ion conductivity on different conductivity on solvent composition and produces the
variables that are of practical interest can be ex- various shapes in σ-xEC relations shown in Figure
plained consistently. Extensive studies have been 7, including single maximum curves and monotonic
carried out on the effects of salt concentration, solvent increases or decreases at different salt concentrations
composition, and temperature on ion conductivity in and temperatures.
different electrolyte systems,113,195,196,198,199 among (3) Temperature (T). With other variables being
which the most representative is the meticulous work the same, ion conductivity increases with tempera-
by Ding et al. on a series of binary systems pertinent ture monotonically until at very high temperatures
to the state-of-the-art lithium ion cells.195,196 Figure the dielectric constant outweighs the viscosity in
7 shows the LiPF6/EC/DMC system as an example affecting ion conduction. Such high temperatures,
Figure 7. Detailed mapping of the dependence of ion conductivity on salt concentration (m), solvent composition (x), and
temperature (T) in a commonly used binary solvent system EC/DMC by surface plots. The orientations, titles, and units
of the axes used in the plots are shown in the bottom portion of the figure, and the temperatures for these plots are, in
order of their appearance, from 60 to -30 °C in 10 °C increments. The σ ranges for the surface plots are, in order of their
appearance, (8.49, 16.7), (7.78, 14.6), (7.03, 12.6), (6.23, 10.6), (5.44, 8.7), (4.63, 7.00), (3.14, 5.47), (1.93, 4.14), (1.04, 3.00),
and (0.46, 2.06) mS cm-1. (Reproduced with permission from ref 195 (Figure 4). Copyright 2001 The Electrochemical Society.)
however, are usually beyond the range of practical general knowledge that ion transport in liquids or
interest.195,200 Conversely, ion conductivity at sub- any noncrystalline polymer media is coupled with
ambient temperatures is predominately determined solvent media.162,163,200
by the increase in solution viscosity, although the In summary, these trends in the change of conduc-
dielectric constant becomes higher simultaneously. tivity with m, xEC, and T can be consistently inter-
A higher salt concentration accelerates the drop of preted in terms of the change of and η with these
ion conductivity with decreasing temperature, be- same variables. Since these factors and their effect
cause it contributes to a higher viscosity. The com- on ion conductivity are not unique to the system
bined effect of higher viscosity and low temperature illustrated, LiPF6/EC/DMC, these trends should pro-
is shown by the steeper σ-T curves of higher m. vide general guidance as to how ion conductivities of
Solvent composition, on the other hand, also has a other electrolyte systems with similar compositions
definite though mild influence on the temperature- would change with these same variables, and they
dependence of ion conductivity. The surface plots in should constitute a useful database for the under-
Figure 7 reveal that the change of conductivity with standing of more complex systems, such as ternary
temperature speeds up as the solvent becomes EC- or quaternary mixtures.
richer,195 though different mechanisms involving
either predominant or η at low or high salt
concentrations, respectively, are believed to be at 5. Electrochemical Stability: on Inert Electrodes
work.
To see more clearly the temperature effect on ion The cycle life of a rechargeable battery depends on
conduction, the logarithmic molal conductivity was the long-term reversibility of cell chemistries, and the
plotted against the inverse of temperature, and the electrochemical stability of the electrolyte plays a
resultant plots showed apparent non-Arrhenius be- crucial role in maintaining this reversibility. In
havior, which can be nicely fitted to the Vogel- electrochemistry, there have been numerous tech-
Tamman-Fulcher (VTF) equation: niques developed to measure and quantify the elec-
trochemical stability of electrolyte components, and
σ ) AT-1/2e-B/R(T-T0) (9) the most frequently used technique is cyclic voltam-
metry (CV) in its many variations.201,202
where A and B are constants characteristic of the In voltammetric experiments, the oxidative or
conduction process and T0 is the vanishing conduc- reductive decompositions of the investigated electro-
tivity temperature, which can be determined through lyte components (solvents or salts) are made to occur
fitting. The T0 values obtained were found to be on an electrode whose potential is controlled, and the
closely related to the glass transition temperatures corresponding decomposition current recorded as the
of the solution systems, which agreed with the function of the potential is used as the criterion for
the stability limits. However, in contrast to the reflected.75,76,93,94,98,99 However, since neat electrolyte
simpler task of measuring ion conduction or deter- solutions are used as both the targets of study and
mining phase boundaries, the electrochemical de- the supporting electrolyte, the observed stability is
composition is often a very complicated process, the result of the possible contributions from all
determined not only by thermodynamic factors but, components, and it is often difficult to distinguish
more importantly, also by kinetic factors such as the whether the stability limits are set by the solvent or
electrode surface, scan rate, and concentration of the the salt decompositions, especially in the case of
species under study. Therefore, electrochemical sta- anodic (oxidation) limits. Because of the high con-
bility data for a given substance depend heavily on centration of electrolytes used, it is also difficult to
the conditions under which they are measured and apply a thermodynamic interpretation to the stability
defined, and the electrochemical stability limits as data obtained in this way. Furthermore, the coexist-
reported in the literature are not always consistent ence of the reversible redox processes that may occur
with each other. Depending on the concentration of simultaneously on these active electrode materials
the target component in the solution under study and can make the determination of electrochemical sta-
the working electrode used, the voltammetry tech- bility limits difficult, and frequently even the defini-
niques favored by the researchers of the battery tion of electrochemical stability becomes ambiguous
community fall into three major categories. because of the passivation on active electrodes. The
The first approach is the standard voltammetry stability data for various lithium ion cell electrolytes
technique used in conventional analytical electro- determined in this way are therefore still scarce
chemistry.202 Normally the target components are despite their importance.
dissolved at much diluted concentrations in a sup- As a compromise between the above two ap-
porting electrolyte, and the electrodes used are made proaches, the third approach adopts nonactive (inert)
of inert materials such as glassy carbon (GC), nickel materials as working electrodes with neat electrolyte
(Ni), and noble metals such as gold (Au), silver (Ag), solutions and is the most widely used voltammetry
and platinum (Pt). The advantages of this approach technique for the characterization of electrolytes for
include the following: (1) the diffusion pattern of the batteries, capacitors, and fuel cells. Its advantage is
investigated species is well-defined in the diluted the absence of the reversible redox processes and
solution so that the redox properties thus determined passivations that occur with active electrode materi-
can be more reliably linked to the thermodynamic als, and therefore, a well-defined onset or threshold
properties;201 (2) it is possible to study the decomposi- current can usually be determined. However, there
tion behavior of solvents and salts separately so that is still a certain arbitrariness involved in this ap-
their contributions to the overall stability of the proach in the definition of onset of decomposition, and
electrolyte can be distinguished. However, the wide disparities often occur for a given electrolyte system
application of this approach is restricted by the when reported by different authors50d,75,76,130,131,203
availability of supporting electrolytes, which are Therefore, caution should be taken when electro-
supposed to be stable in the potential ranges where chemical stability data from different sources are
compared.
the target components decompose. In the case of
This section will discuss the electrochemical sta-
lithium-based electrolytes, the wide electrochemical
bilities of different solvents and salts used in state-
stability window of the electrolyte solvents and salts
of-the-art electrolytes that were determined with
makes it hardly possible to find a supporting medium
nonactive electrodes (i.e., in the first and the third
that is more resistant against decomposition. An
approaches). When active rather than inert electrodes
additional disadvantage of this approach also in-
are used as working surfaces, many complicated
cludes the fact that the surface of the inert electrode,
processes, including the reversible electrochemical
usually nonporous, is very different in catalytic
redox chemistries as well as surface passivation,
activity from the porous surfaces of the composite
occur simultaneously. These related materials will
electrode materials used in real electrochemical
be dealt with in a dedicated section (section 6).
devices. As a consequence, this approach could over-
estimate the electrochemical stability of the compo-
nents.76 5.1. Anion of Lithium Salts
The second approach is an adaptation of the Table 4 lists selected electrochemical stability data
voltammetry technique to the working environment for various lithium salt anions that are commonly
of electrolytes in an operational electrochemical used in lithium-based electrolytes, with the measure-
device. Therefore, neat electrolyte solutions are used ment approaches indicated. Although it has been
and the working electrodes are made of active known that the reduction of anions does occur,
electrode materials that would be used in an actual sometimes at high potentials, the corresponding
electrochemical device. The stability limits thus processes are usually sluggish and a definite poten-
determined should more reliably describe the actual tial for such reductions is often hard to determine.
electrochemical behavior of the investigated electro- The reduction of solvents, occurring simultaneously
lytes in real life operations, because the possible with that of anions on the electrode, further compli-
extension or contraction of the stability window, due cates the interpretation efforts. For this reason, only
to either various passivation processes of the elec- the anodic stability of salt anions is of interest, while
trode surface by electrolyte components or electro- the cathodic limit of the salt in most cases is set by
chemical decomposition of these components cata- the reduction of its cation (i.e., lithium deposition
lyzed by the electrode surfaces, would have been potential).
Table 4. Anodic Stability of Electrolyte Solutes: Nonactive Electrodes

working
salt solvent conc/M electrode Eaa (i/(mA cm-2)) approach ref
- 130, 131b
ClO4 PC 0.65 GC 6.1 (1.0) 3
PC Pt 4.6 3 204
BF4- PC 0.65 GC 6.6 (1.0) 3 130, 131, 81c
EC/DMC 1.0 AC 4.78 2
PF6- THF 0.001 GC 4.4 (0.1) 1 120d
SL 0.001 GC 4.8 (0.1) 1 120
PC 0.65 GC 6.8 (1.0) 3 130, 131
nonee none GC 4.94 (1.0) 3 206
none none Pt 5.00 (1.0) 3 206
EC/DMC 1.0 AC 4.55 2 81
AsF6- PC 0.65 GC 6.8 (1.0) 3 130, 131
none none GC 5.05 (1.0) 3 206
none none Pt 5.10 (1.0) 2 206
EC/DMC 1.0 AC 4.96 3 81
THF 1.0 GC 4.25 (0.1) 1 120
THF 0.009 GC 1 120
SL 0.8 GC 4.69 (0.1) 1 120
SbF6- THF 1.0 GC 4.10 (0.1) 1 120
PC 0.65 GC 7.1 (1.0) 3 130, 131
Tf- PC 0.65 GC 6.0 (1.0) 3 130, 131
PC 0.10 Pt 5.0 (0.5) 3 207
EC/DMC 1.0 AC 4.29 2 81
Beti- PC 0.65 GC 6.3 (1.0) 3 130, 131
PC 0.1 GC 6.2 (0.5) 3 207
Im- PC 0.65 GC 6.3 (1.0) 3 130, 131
PC 0.1 Pt 5.3 (0.5) 3 130, 131
none none GC 5.06 (1.0) 3 206
none none Pt 5.13 (1.0) 3 206
EC/DMC 1.0 AC 4.33 2 81
a Anodic limit, potential referred to Li+/Li, cutoff current density in parentheses. b Scan rate: 5 mV s-1. c Activated carbon as
working surface. Scan rate: 10 mV s-1. d Supporting electrolyte 0.1 M Bu4NBF4. Scan rate: 100 mV s-1. e The solvent-free condition
was realized by using an ionic liquid based on imidazolium cation, at 80 °C. Scan rate: 20 mV s-1.
The most noteworthy observation among the en- (GLN), the stability order of these anionic species was
tries of Table 4 is that these anodic stability data are determined as130
widely diversified depending on the conditions under
which they were obtained. For example, the anodic SbF6- > AsF6- g PF6- > BF4- > Im- > Tf >
stability limits of PF6-, AsF6-, and SbF6- determined
ClO4-
in THF solutions are obviously lower than those
determined in carbonate solutions, and a possible It should be pointed out that the above anodic
explanation lies in the fact that THF itself is not an stability order is highly conditional, since the order
oxidation-resistant solvent; therefore, its own de- had been determined by approach 2 above, where the
composition is most likely responsible for these stability limits are defined as the potential at which
limits.120 the decomposition current density reaches an arbi-
trary value (1.0 mA cm-2 in this case). Any change
The rather systematic work was performed by Ue
in this criterion could result in a reversal of the order.
et al., who used tetraalkylammonium salts as sup-
The coupling effects of solvent/salt on electrolyte
porting electrolytes and measured the oxidation stability can also be observed when mixture solvents
potential of a number of anions, most of which were are used, and the stability of the electrolyte can be
considered promising candidates for lithium-based much improved when a stable solvent/salt is selected.
batteries.50d,130,131 Figure 8 shows the linear sweep For example, the room-temperature breakdown volt-
voltammograms obtained from these solutions with age of electrolytes LiX/EC/DEE lies in the order205
a GC electrode.50d The reduction limits seem to be
caused by the decomposition of the ammonium ClO4- > Im- > Tf- > AsF6- > PF6- > BF4-
cations, as evidenced by the similar cathodic current
response for all of these solutions in Figure 8 and However, when a linear carbonate replaces the
further confirmed by the analysis conducted on the unstable ether DEE, the order changes into
decomposition products.204 On the other hand, the
ClO4- ∼ PF6- ∼ BF4- > AsF6- > Im- > Tf-
oxidation limits were determined simultaneously by
both salt and solvent. For example, the anodic limits It is believed that the reactivity of DEE toward Lewis
were set by the unstable anions (Tf - and ClO4-), acids, PF6- and BF4-, contributes to the early de-
whereas solvent PC would be oxidized first if stable composition of the electrolytes.
anions such as BF4- and PF6- are present. By using To exclude the solvent’s effect on the anodic stabil-
a more oxidition-resistant solvent, glutaronitrile ity of salt anions, Koch et al. made a series of salts
Scheme 7. Proposed Single-Electron Mechanism

for the Reduction of EC and DEC
stability limits for some nonaqueous solvents com-

monly used in lithium-based battery research. De-
spite the inconsistency created by the varying mea-
surement conditions, these data express a general
trend that we have discussed in section 2; that is,
carbonates and esters are more anodically stable,
while ethers are more resistant to cathodic decom-
positions.
Among cyclic carbonates and esters, EC seems to
be more liable toward reduction on nonactive elec-
trodes,210-214 which agrees with molecular orbital ab
initio calculations of the corresponding radical an-
ions. The difference between EC and PC in reductive
reactivity is attributed to the steric effect introduced
by the methyl on PC rather than the electronic effect,
as supported by the molecular orbital ab initio
Figure 8. Determination of anodic stability for various calculations for the corresponding radical anions.212
anions in PC solution on GC. (Reproduced with permission On the other hand, the reduction of PC seems to be
from ref 130 (Figure 1a and b). Copyright 1997 The
Electrochemical Society.)
a rather sluggish process, as evidenced by the high
background current level distributed in a broad
potential range during voltammetric scans. The slow
Scheme 6. Electrochemical Oxidation of the PF6- kinetics of PC toward cathodic reduction is correlated
Anion: the Solvents or the Trace Moisture in the with its readiness to cointercalate into a graphene
Electrolyte Serves as the Proton Source of structure of carbonaceous anodes and then decom-
H-Abstraction pose, causing exfoliation of the latter.214
Aurbach and co-workers performed a series of ex
situ as well as in situ spectroscopic analyses on the
based on a low-melting organic cation and measured surface of the working electrode upon which the cyclic
the oxidation limits of these solvent-free molten salts voltammetry of electrolytes was carried out.210-213 On
on the surfaces of GC, tungsten, and Pt, respec- the basis of the functionalities detected in FT-IR,
tively.206 The “intrinsic anodic stability” measured by X-ray microanalysis, and nuclear magnetic resonance
them follows an order completely different from the (NMR) studies, they were able to investigate the
one determined in solvents: mechanisms involved in the reduction process of
carbonate solvents and proposed that, upon reduc-
tion, these solvents mainly form lithium alkyl car-
Im- ∼ AsF6- > PF6- bonates (RCO3Li), which are sensitive to various
contaminants in the electrolyte system. For example,
A mechanism was also proposed for the instability the presence of CO2 or trace moisture would cause
of the PF6- anion (see Scheme 6). the formation of Li2CO3. This peculiar reduction
The above stability order is supported by the product has been observed on all occasions when
results of ab initio calculations, where a correlation cyclic carbonates are present, and it seems to be
has been established between the oxidation limits of independent of the nature of the working electrodes.
these anions and the energy level of their highest A single electron mechanism has been shown for PC
occupied molecular orbital (HOMO). However, the reduction in Scheme 1, while those of EC and linear
gapping discrepancies that exist between “intrinsic” carbonates are shown in Scheme 7.214
and solution stability limits remain to be explained. The radical anion is considered to be the common
intermediate formed for these carbonate molecules,
5.2. Solvents and its existence in the electrochemical reductions
has been confirmed experimentally.178 As will be
A solvent with an ideal electrochemical stability for revealed later, this mechanism seems to be universal
a high-energy-density battery purpose should possess for carbonates, whether on inert or active electrodes,
high oxidation and low reduction potentials at the and alkyl carbonates have been widely believed to
same time. Table 5 lists selected electrochemical be a key component in forming a protective interface
Table 5. Electrochemical Stability of Electrolyte Solvents: Nonactive Electrodes

working
solvent salt/conc (M) electrode Eaa Ecb ref
PC Et4NBF4/0.65 GC 6.6 130, 131
nonec Pt 5.0 ∼1.0 74
Bu4NPF6 Ni 0.5 212
LiClO4/0.1 Au, Pt 1.0-1.2 214
LiClO4/0.5 porous Pt 4.0 305
LiClO4 Pt 4.7 177
LiClO4 Au 5.5 306
LiAsF6 Pt 4.8 177
EC Et4NBF4/0.65 GC 6.2 130, 131
Bu4NPF6 Ni 0.9 212
LiClO4/0.1 Au, Pt 1.36 214
DMC Et4NBF4/0.65 GC 6.7 130, 131
LiClO4/0.1 Au, Pt 1.32 214
LiPF6/1.0 GC 6.3 76
LiF GC 5.0 312
DEC Et4NBF4/0.65 GC 6.7 130, 131
LiClO4/0.1 Au, Pt 1.32 214
EMC Et4NBF4/0.65 GC 6.7 130, 131
LiPF6/1.0 GC 6.7 76
γBL LiAsF6/0.5 Au, Ag 1.25 208
THF Et4NBF4/0.65 GC 5.2 130, 131
LiClO4 Pt 4.2 177
LiAsF6/1.0 GC 4.25 (0.1) 214
nonec Pt 4.0 < -2.0 74
LiAsF6/1.0 GC 4.2 64
LiAsF6 Pt 4.2 177
2-Me-THF LiClO4 Pt 4.1 177
LiAsF6/1.0 GC 4.15 (0.1) 214
LiAsF6/1.0 GC 4.2 64
LiAsF6 Pt 4.1 177
DME Et4NBF4/0.65 GC 5.1 130, 131
LiClO4 Pt 4.5 177
LiAsF6 Pt 4.5 177
a,b
Anodic and cathodic limits, potential referred to Li+/Li. c The salt-free condition was realized via an ultramicroelectrode
technique.
between active electrodes and electrolytes. In a more mining influence on the physicochemical nature of
general context, since the surface chemistries on both the new electrode surface, such as thermal and
inert and active electrodes have been found to be very chemical stability, as well as its impedance to ionic
similar in most of the nonaqueous electrolytes when conduction. The occurrence of passivation constitutes
these electrodes are polarized to low potentials, the the foundation upon which many high-energy-density
early analytical work, mainly performed by Aurbach battery chemistries are built, including lithium-based
and co-workers, should be recognized as the founda- chemistry. For this reason, the electrolyte/electrode
tion for the later understandings of the SEI on interface has been the focus of research interest for
carbonaceous anodes. both lithium and lithium ion chemistries.
On the other hand, it should be pointed out that,
6. Electrochemical Stability: on Active Electrodes in addition to the protective effect of passivation, the
passivated interface also acts as a barrier to the facile
The requirement that electrolytes be inert toward ion transport that occurs between the electrode and
both anode and cathode materials is usually realized electrolyte. More often than not, the bottleneck for
by the kinetic rather than thermodynamic stability the overall battery chemistry is constituted by pas-
of the former against reductions and oxidations. sivation. Excessive passivation is especially undes-
Given the strong reducing and oxidizing potency of ired because it reduces the power performance of the
the electrode materials employed in the high-energy- cell. For lithium ion cells, this power reduction
density battery chemistries, the possibility of a usually happens on the cathode surface.
thermodynamically stable electrolyte is usually non-
existent, and it is the chemical passivation of these 6.1. Passivation on Lithium Anode
reducing or oxidizing surfaces that ensures the
inertness of the bulk electrolytes during cell chem- Almost immediately after lithium metal was found
istry. to be stable in nonaqueous electrolytes, researchers
Passivation is a process where the products from suggested that the passivation of the lithium surface
the initial decomposition of electrolyte form a dense, by electrolytes is the origin of this unexpected stabil-
protective film that covers up the pristine surface of ity, because the reduction potentials of these organic
the electrode and prevents any sustained decomposi- solvents are far above that of lithium.4,215-217 Peled
tion. The electrolyte components that are sacrificed was the first author to formally introduce the concept
to form such a protective film would have a deter- of a protective interface between lithium and elec-
Scheme 8. Proposed Two-Electron Mechanism for SEI on lithium is not Li2CO3 but the lithium salt of
the Reduction of PC alkyl carbonate, and it was most probably formed via
a one-electron mechanism followed by radical termi-
nation paths as shown in Scheme 1.55 This compound
has an FT-IR signature at ∼1650 cm-1 for the
carbonyl stretching, which was confirmed by the
characterization of a synthesized authentic sample
trolytes and describe the fundamental physicochem- of a lithium alkyl carbonate, CH3CH2CH2CO3Li.55
ical properties of such an interface.37 He argued that, Since XPS also detected the decomposition products
because of the high electronegativity of lithium, the from lithium salts, which are in the form of simple
free contact between it and the electrolyte compo- halides, alkoxides, or oxides, a competition between
nents never actually exists; instead, the reaction solvents and salts obviously exists. However, the
between the lithium electrode and electrolyte com- formation of alkyl carbonate seems to be dominant
ponents occurs instantaneously, and the solid prod- when EC is present because of the more reactive
ucts from this reaction compose a thin film that grows nature of EC toward cathodic reductions.117,209 The
on the lithium with the reaction. The decomposition formation of lithium alkyl carbonate was also con-
could only stop when this film, which is nonconduc- firmed in an independent diagnosis work, where the
tive to electrons but conductive to ions, covers the reduction products of EC in a supporting electrolyte
whole surface of the lithium and attains a certain were hydrolyzed by D2O and then subject to NMR
thickness. Once formed, the film stays on the lithium analysis, which identified ethylene glycol as the main
surface at all times, and it cannot be completely species formed, as indicated by the singlet at δ ) 3.7
removed even by the stripping/deposition of lithium. ppm vs TMS.208,212 Thus, Aurbach and co-workers
Since this film acts like an electrolyte in its conduc- concluded that PC and EC were reduced to the
tive preference, Peled named it the “solid electrolyte following lithium alkyl dicarbonates, respectively:
interface (or interphase)” (SEI). Considering the low
ionic conductivity of solid-state electrolytes, he also
proposed that the rate determining step for a redox
process on such surfaces would be the diffusion of
lithium ions through the SEI,218 instead of the
electron charge-transfer between electrode and solu- Further studies demonstrated that alkyl carbonate
tion species, as was believed before. is very sensitive to the common contaminants in
Using a parallel capacitor model, Peled and Straze electrolytes, which is probably the reason that early
calculated the apparent thickness of the SEI for a work identified Li2CO3 as the main component in the
series of active metal electrodes, including lithium, SEI.118 For example, trace moisture produces Li2CO3
calcium, and magnesium, with the equation219 when the salt anion is stable (such as ClO4- or
AsF6-):249
A
L) (10)
Cπ(3.6 × 1012)
where A, L, C, and are the electrode area, the
thickness, the capacitance, and the dielectric constant However, when the salt anion is susceptible to
of SEI, respectively. They estimated that the average hydrolysis by trace moisture (such as BF4- or PF6-),
SEI thickness of lithium in nonaqueous electrolytes the generated HF eliminates alkyl carbonate as a
is 25-100 Å.218,219 Electrons tunneling through the surface species and only LiF can be observed as the
film of this thickness are believed to be minimal. overwhelming species in the SEI:117
The chemical composition of the SEI is believed to
be closely related to the electrolyte composition. For
example, in thionyl chloride-based electrolytes, the
SEI is proposed to be mainly composed of LiCl,218
while it becomes Li2S2O4 in sulfur dioxide-based Alkyl carbonate is also unstable upon storage on the
electrolytes218 or Li2O in ether-based electrolytes.56 lithium electrode, probably because of its continuous
After performing microscopic observations of lithium electrochemical reduction.118 Thus, Aurbach et al.
surfaces treated with PC-based electrolytes, Dey further proposed that the SEI might have a multi-
believed the main component of this interface to be layer structure within which the simple inorganic
Li2CO3, which is the decomposition product of PC species such as Li2CO3 and Li2O are more stable and
through a two-electron mechanism (see Scheme 8).56 closer to lithium, while alkyl carbonate is more likely
This conclusion seems to be supported by results from to be distributed in the outer layers.117,118
Auger electron spectroscopy (AES) and X-ray photo- Kanamura et al. carried out detailed XPS studies
electron spectroscopy (XPS).220,221 However, more on lithium electrodes that were either statically
recent studies suggest that the above process might treated with or cycled in LiBF4-based electrolytes.222,223
be oversimplified, and the actual process involves a By sputtering the surface of the lithium electrode,
complex competition between a number of reductive they were able to record the depth profile of the
reactions of different components. related chemical species. Their conclusions are in
With surface-sensitive spectroscopic means, Aur- good agreement with the hypothesis by Aurbach et
bach et al. argued that the main component in the al.;117,118 that is, while alkyl carbonate can be detected
Figure 10. Impedance complex plane (Nyquist plots) of

lithium electrode in (A) 1.0 M LiPF6/EC/PC and (B) 1.0 M
LiClO4/EC/PC at initial time (0.0 h) and after 24 h. Re and
Im stand for the real and imaginary parts of the impedance
measured, respectively. Frequency was indicated in the
figure for selected data points. Note that the first semi-
circle corresponds to SEI impedance. (Reproduced with
Figure 9. Schematic illustrations of the surface film permission from ref 86 (Figure 2). Copyright 1992 The
formed on lithium in nonaqueous electrolytes based on Electrochemical Society.)
LiBF4 solutions and the subsequent reactions. (Reproduced
with permission from ref 222 (Figure 12). Copyright 1995 Aside from voltammetric techniques, ac impedance
The Electrochemical Society.) is also a powerful tool widely used to study the
interfacial properties of lithium in nonaqueous elec-
in the outlaying layer of the SEI, as indicated by its trolytes. It is one of the few in situ techniques and
signature binding energy of 289.0 eV at the C 1s therefore is often used in combination with voltam-
region, its abundance rapidly decreases with sput- metry, known as electrochemical impedance spec-
tering time. On the other hand, O 1s spectra clearly troscopy (EIS). As an example, Figure 10 shows the
reveal the increasing abundance of Li2O species. LiF impedance response of a symmetrical cell, lithium|1.0
exists throughout the SEI and is relatively indepen- M LiX in PC/EC|lithium, drawn in the Nyquist plot,
dent of sputtering, a result of the sensitivity of BF4- where LiX is LiPF6 or LiClO4. Typically, two semi-
anion toward trace moisture in electrolytes. Kana- circles would be observed for such cells at high and
mura et al. proposed two possible paths for the medium frequencies, if the time constants for each
formation of LiF: (1) simple acid-base reaction component are sufficiently separated, along with a
between HF and alkyl carbonate or Li2CO3, or (2) spike at the low-frequency end.175 It is generally
direct reduction of BF4- anion by lithium.222,223 Ad- accepted that the semicircle at medium frequency
ditional reactions between the solvents and the corresponds to the ionic migration process in the SEI
lithium electrode also seem possible after the initial and the one at lower frequency to the charge-transfer
formation of the SEI, since the abundance of organic process on lithium, whereas the intercept at the high
species increases in the inner layer with storage time frequency end with the real axis represents bulk
increases, according to C 1s spectra. These organic electrolyte resistance. Examination of the interfacial
species might be some polymeric products that re- resistance in various electrolyte solutions reveals that
sulted from PC or other carbonates instead of alkyl the SEI on lithium grows with time of exposure to
carbonates alone, as evidenced by the C 1s signal electrolytes, and the chemical nature of both solvent
around 286 eV and by an earlier XPS work.224 This and salt anion seems to relate closely to the semi-
latter process is attributed to the permeation of circle for the interfacial film.86,225-227 For example, the
solvent through the SEI and its subsequent reaction resistance of the SEI that formed in the LiPF6-based
with lithium. The resulting polymeric films, most electrolyte is smaller than the one formed in the
likely polyether moieties, are embedded with LiF LiClO4-based electrolyte, and the presence of EC also
crystals. Figure 9 schematically shows the lithium renders a more conductive SEI on lithium.86,225-227
surface structure and these subsequent reactions of An empirical rule (with frequent exceptions though)
the SEI.222 might be stated here concerning the resistance of the
SEI: an electrolyte with higher bulk ion conductivity electrolyte system that can suppress or even elimi-
usually results in an SEI of lower impedance, either nate lithium dendrite formation.
on a lithium or carbonaceous electrode, as will be
discuss in later sections. 6.2. Electrolyte/Carbonaceous Anode Interface:
As a mass sensor with nanogram sensitivity, a SEI
quartz crystal microbalance (QCM) was used in
combination with voltammetry by Naoi et al. to 6.2.1. Exfoliation and Irreversible Capacities on a
monitor the change occurring on the lithium surface Carbonaceous Anode
during SEI formation.228 Like EIS, it is one of the It has been known since the mid 1950s that
few analytical tools that can reveal in situ informa- graphite can form intercalation compounds with
tion on the interfacial process, which includes the lithium ions, which are accommodated in the
change in mass as well as the surface morphology of interstitial region between the planar graphene
the lithium electrode. It was found that, during the sheets.230-232,233 The most lithium-enriched intercala-
cycling process, the already-formed SEI was repeat- tion compound of this family has a stoichiometry of
edly destroyed and rebuilt, as indicated by the LiC6, and its chemical reactivity is very similar to
consistent mass increase with cycle numbers. Among that of lithium metal. There have been a number of
the various combinations of salts and solvents tested, different chemical approaches to the preparation of
the LiPF6-based electrolyte seems to have the most these compounds, for example, by direct reactions of
rapid reaction kinetics with lithium, since the lithium graphite with molten lithium at 350 °C,232 with
electrode in it was observed to gain net mass even lithium vapor at >400 °C,234 or with lithium powder
during the stripping step, suggesting that the reac- under high pressure,235,236 and so forth.
tion kinetics between the fresh lithium surface and On the other hand, the electrochemical synthesis
the electrolyte is fast enough to compensate for the of these lithium graphite intercalation compounds
mass loss caused by the lithium dissolution. On the (Li-GIC) has been proven difficult. In earlier work,
other hand, the net mass accumulated on lithium is it had been found that the most commonly used
much higher in LiClO4- and LiTf-based electrolytes electrolyte solvent, PC, decomposed reductively on
than in the LiPF6-based electrolyte. The conclusion the graphite electrode at a potential of ∼0.80 V, and
from the above two observations seems to point to a the irreversible process led to the physical disinte-
more efficient and protective SEI formed by the gration of graphite.237 The occurrence of this irrevers-
LiPF6-based electrolyte. The author ascribed this ible reduction apparently prohibits any possibility of
result to the trace amount of LiF in the SEI, which the lithium ion intercalating into graphite, which
is absent in LiClO4- and LiTf-based electrolytes.118 should happen at a much lower (and therefore more
The measurement of lithium surface roughness also reductive) potential. The destruction of graphite by
reveals LiPF6 as the favored salt in electrolytes PC was repeatedly observed in different electrolytes
because it forms a smoother and more uniform SEI, based on PC, and this disintegrating process of the
thus minimizing the probability of dendrite growth graphite structure was named “exfoliation”.238-243
on a relative scale as compared with the cases of the Besenhard et al. proposed that the exfoliation was
other salts studied. caused by the cointercalation of PC molecules with
The significance of an SEI on lithium stability lithium ions into the interplanar structure of the
should be evaluated from two different angles: one graphite and the subsequent decomposition there-
is the static stability that relates to standing storage, in.239-242 As a result, the multilayer structure of
and the other is the dynamic stability that relates to graphite, which is only held together by weak van
reversibility. It is the SEI formation on the lithium de Waals forces, falls apart because of the strain
surface that leaves lithium statically stable in a introduced by the gaseous products, believed to be
nonaqueous electrolyte; conversely, the SEI also mostly propylene.234-237,243
renders a nonuniform surface morphology for depos- Realizing that the solvent must be the key to the
ited lithium, so that the current density across the exfoliation, later researchers explored different polar
surface is unevenly distributed during the lithium organic molecules such as dimethyl sulfoxide (DMSO)
stripping/deposition, with the direct consequence and DME as candidates to replace PC, in the hope
being dendrite growth. that they would not cointercalate or decompose; but
The roughness of the SEI depends heavily on the most of these efforts failed to endorse the usefulness
chemical nature of the electrolyte. For example, it of Li-GIC as a negative electrode to replace lith-
was argued that an SEI consisting of LiF/Li2O would ium.239-241 In the 1980s, the only successful example
provide a much more uniform current distribution,223 of electrochemical intercalation of lithium into graph-
while in numerous earlier works it was also observed ite was reported by Yazami and Touzain in 1983 with
that trace moisture has a positive effect on the a polymer electrolyte based on poly(ethylene oxide)
lithium cycling efficiency in nonaqueous electrolytes (PEO).244 As it is essentially impossible for the
by assisting in forming a compact and uniform macromolecular solvent PEO to cointercalate, this
SEI.57,229 Nevertheless, the dendrite issue is a major electrolyte supported the reversible lithium ion in-
challenge to lithium metal-based chemistry that still tercalation into and deintercalation from natural
remains unresolved. The prospects for this battery graphite. Using electrochemical titration techniques,
technology, still attractive because of its high energy the potential of the stage I and II Li-GIC was
density as compared with the state-of-the-art lithium determined to be between 0.50 and 0.20 V vs Li, thus
ion technology, rely on the discovery of a new confirming the conceptual feasibility that Li-GIC
could be used as a negative electrode to replace

lithium with minimal energetic penalty. Thus, a
negative electrode based on the stage I Li-GIC would
have the following reversible chemistry:
discharge
LixC6 y\
charge
z xLi+ + 6C + xe
In the ideal situation of 100% utilization (x ) 1.0),

the capacity corresponding to the above anode half
reaction is 372 mA h g-1. However, due to the low
ion conductivity of the polymer electrolyte and the
high interfacial impedance between it and the graph-
ite electrode, this elegant example of electrochemical
preparation of lithiated graphite is of limited practi-
cal significance.
Considering the vulnerability of the highly crystal-
line structure of graphite, some researchers in the
late 1980s focused their attention on carbon structure
instead of electrolyte formulation. Successes were
seen with various amorphous carbon electrodes,
including carbon fiber,245 pyrolytic carbons,246 and
petroleum coke.247 The insensitivity of these carbon- Figure 11. (a) Initial 11/2 cycles of a Li/petroleum coke
aceous materials to solvent cointercalation seems to cell. The cell was cycled at a rate of 12.5 h for ∆x ) 0.5 in
be related to their amorphous portion, which has LixC6. (b) Initial 11/2 cycles of a Li/graphite cell. The cell
stacking defects that coexist with the tiny crystalline was cycled at a rate of 40 h for ∆x ) 0.5 in LixC6. F denotes
sectors and serve to pin the graphene layers together the irreversible capacity associated with SEI formation, E
to suppress the lattice expansion that occurs during the irreversible capacity due to exfoliation, and I the
reversible capacity due to lithium intercalation into carbon.
the solvent cointercalation. Thus, the indefinite “ex- 1.0 M LiAsF6 in EC/PC was used as electrolyte. (Repro-
foliation plateau” at 0.80 V is completely absent from duced with permission from ref 36 (Figure 2). Copyright
these amorphous carbon anodes in PC-based electro- 1990 The Electrochemical Society.)
lytes,36 and reversible lithium ion intercalation is
realized up to >103 cycles in various nonaqueous attractive as an anode candidate), it also becomes
electrolytes, although the initial cycles are always more liable to solvent co-intercalation.43,96,249,250 It
accompanied by a portion of irreversible capacity that was only after Dahn and co-workers revealed the role
depends on the nature of the carbon and the compo- of the SEI on the reversibility of carbonaceous
sition of the electrolytes.36,157 These efforts eventually electrodes and the effect of EC therein that the
led to the commercialization of the first generation energetic advantage of highly graphitic carbonaceous
lithium ion cells.29-31,34-35 materials regained practical significance for the
However, two penalties, both associated with the lithium ion industry.36 As a direct consequence of
energy density, arise from the disordered anode improvements in electrolyte formulations aided by
structure: (1) a smaller Coulombic capacity than the the knowledge gained in SEI chemistry, amorphous
theoretical value for LiC6 and (2) a sloping potential carbonaceous anode materials were gradually phased
profile during both charging and discharging.43,157 out from commercial lithium ion technology, starting
The former is caused by the small crystallinity of in the 1990s.38
these amorphous carbons, because it is in the highly In retrospect, the significance of Dahn’s seminal
ordered graphene sheets where lithium ion could be work lies in two aspects: (1) the fundamental un-
accommodated, while the latter is caused by a broad derstanding of how carbonaceous materials operate
distribution of adsorption site energies in the disor- in nonaqueous electrolytes and (2) the more practical
dered carbons, which leads to a broad potential range side of how the above dilemma concerning energy
within which lithium intercalation occurs.125 Figure density and reversibility can be overcome. This
11 compares the cycling behavior of an amorphous knowledge dictates the development of electrolytes
carbon anode with that of graphite in a half anode for state-of-the-art lithium ion chemistry.
cell248 and clearly demonstrates the above two penal- On the fundamental front, Dahn et al. successfully
ties.36 While the latter is certainly disadvantageous accounted for the irreversible capacity that accom-
because of the unstable cell voltage, the combination panies all carbonaceous anodes in the first cycling.
of these two results in a much lower energy density They observed that the irreversible capacity around
that Li-GIC can offer theoretically. 1.2 V follows an almost linear relation with the
Hence, a dilemma was encountered between energy surface area of the carbonaceous anodes and that this
density and stability for these various forms of irreversible process is essentially absent in the fol-
carbonaceous materials; that is, as the carbonaceous lowing cycles.36 Therefore, they speculated that a
anode is more graphitic in structure, the degree of passivation film that resembles the one formed on
lithium ion intercalation may be closer to the ideal lithium electrode in nonaqueous electrolyte37,218,219
(x ) 1.0), and its potential profile may be closer to must also be formed on a carbonaceous electrode via
that of Li+/Li and remain relatively flat (therefore similar electrolyte decompositions, and only because
of the more porous surface of the latter would the mogeneous current distribution, (4) good adhesion to
capacity associated with the reduction process be- the carbonaceous anode surface, (5) good mechanical
come noticeable. Once formed, the physicochemical strength and flexibility so that it allows the expan-
property of this film should be similar to the solid sion and contraction of the graphene lattice during
electrolyte model that Peled had proposed for the the reversible intercalation/deintercalation process,
lithium surface; that is, it is an ionic conductor and and (6) low solubility in electrolytes so that continu-
an electronic insulator,37 and therefore, the sustained ous dissolution of SEI would not occur, resulting in
reductive decomposition could be prevented. Hence, persistent decomposition of electrolyte and consum-
the term SEI that Peled had invented for the passi- mation of the limited source of lithium from the
vation of lithium electrode was transplanted on the cathode.
carbonaceous anode materials. There has been considerable controversy concern-
On the practical side, Dahn’s work demonstrated ing the mechanism of SEI formation on a carbon-
that, by altering the electrolyte composition, the aceous anode, but it is generally agreed that the
exfoliation of graphitic materials could be eliminated, initial electrolyte decomposition is responsible and
since the chemical nature of the SEI is dictated by that a competition among a variety of reactions
electrolyte composition, especially by the solvents. involving the solvent as well as the salt components
The “magic ingredient” identified in this case is EC, is also present.
whose structural difference with PC is merely a 6.2.2.1. Peled’s Model: Anode/Electrolyte Interface
methyl substituent. As Figure 11 shows, the presence Film. In their proposal of SEI formation on a car-
of EC not only prevents the physical disintegration bonaceous electrode in nonaqueous electrolytes, Dahn
of graphite that occurs at 0.8 V but also supports the actually adopted Peled’s model for lithium’s surface
reversible intercalation (during the discharge of the and extended it to carbonaceous electrodes. By this
half-cell) and deintercalation (during the correspond- model, a two-dimensional passivation film is estab-
ing charge) of lithium ions at a very low potential, lished via a surface reaction.
<0.20 V, with Coulombic capacity approximately
Because of the similar potentials between fully
approaching the theoretical value of LiC6. Because
lithiated graphite and lithium metal, it has been
PC is present in the electrolyte, a certain degree of
suggested that the chemical nature of the SEIs in
graphite exfoliation can still be observed, but obvi-
both cases should be similar.36 On the other hand, it
ously EC plays an effective role in suppressing the
has also been realized that for carbonaceous anodes
destruction of the graphite structure. The latter
this formation process is not expected to start until
development in electrolyte formulation made by
the potential of this anode is cathodically polarized
Tarascon and Guyomard effectively suppressed this
(the discharge process in Figure 11) to a certain level,
side reaction to a negligible level,44,98,99 although a
because the intrinsic potentials of such anode materi-
certain irreversible capacity was always present,
als are much higher than the reduction potential for
accounting for the material supply required by the
most of the solvents and salts. Indeed, this potential
SEI formation. The irreversible capacity, or the ratio
polarization process causes one of the most funda-
of this capacity to the total reversible capacity (called
mental differences between the SEI on lithium metal
Coulombic inefficiency), has become one of the pa-
and that on a carbonaceous anode. For lithium metal,
rameters for measuring the performance of a certain
the SEI forms instantaneously upon its contact with
electrolyte on carbonaceous anodes.
electrolytes, and the reduction of electrolyte compo-
The commercial lithium ion cells on the consumer nents should be indiscriminate to all species pos-
market nowadays have already undergone a so-called sible,37,218,219 while, on a carbonaceous anode, the
“forming process” at the manufacture sites, during formation of the SEI should be stepwise and prefer-
which a stable SEI was formed to ensure that no ential reduction of certain electrolyte components is
more irreversible process was left. The Coulombic possible.
efficiency of these cells ought to be 100% under the
Endo et al. investigated the reductive decomposi-
conditions specified for their application. On the other
tion of various electrolytes on graphite anode materi-
hand, any accidental misuse such as overcharge, high
als by electron spin resonance (ESR).178 In all of the
temperature exposure, and mechanic impact might
electrolyte compositions investigated, which included
damage the already formed SEI, resulting in more
LiClO4, LiBF4, and LiPF6 as salts and PC, DMC, and
irreversible reactions during charging and conse-
other esters or ethers as solvents, the solvent-related
quently a loss from the rated capacity.
radical species, which were considered to be the
6.2.2. Mechanism of SEI Formation intermediates of reductive decomposition,55 were
detected only after prolonged cathodic electrolysis.
According to Peled’s model, the existence of an SEI With the aid of molecular orbital calculation, they
constitutes the foundation on which lithium ion found that the reduction of salt anion species is very
chemistry could operate reversibly. Therefore, an difficult, as indicated by their positive reduction
ideal SEI should meet the following requirements: enthalpy and that of free solvent (∆Hr ≈ -1 kcal
(1) electron transference number te ) 0 (otherwise, mol-1). However, the coordination of lithium ions
electron tunneling would occur and enable continuous with these solvents dramatically reduces the corre-
electrolyte decomposition), (2) high ion conductivity sponding reduction enthalpy (∆Hr ≈ -102 kcal mol-1)
so that lithium ions can readily migrate to intercalate and renders the reaction thermodynamically favored.
into or deintercalate from graphene layers, (3) uni- In other words, if no kinetic factors were to be
form morphology and chemical composition for ho- considered, the SEI formed on carbonaceous anodes
would predominantly consist of the decomposition

products of those solvents in the solvation sheath of
lithium ion, which migrates toward the negatively
charged anode surface. For state-of-the-art electro-
lytes, the reduction of cyclic carbonates such as EC
and PC should provide the major species to build up
the SEI while the participation of linear carbonates
should be relatively inconsequential. A similar con-
clusion was drawn by Wang et al., who employed a
high-level density functional theory to investigate the
reductive decomposition mechanism for EC molecules
in electrolyte solution and found that, while the
reduction of a free EC molecule is very unlikely, the
coordination of lithium ion to an EC molecule renders
the one- or two-electron reduction processes thermo-
dynamically possible in a supermolecular structure
such as Li+(EC)n (n ) 1-5).170
In view of the possibility that certain electrolyte
components could be preferentially reduced on car-
bonaceous anode, Peled and co-workers explored a
means to manipulate the chemical nature of the SEI
by deliberately using unstable electrolyte ingredi-
ents.106 They argued that, since these components
would be reduced at a higher potential, the SEI would
be completed far before solvent cointercalation oc- Figure 12. Schematic illustration of the SEI formation
curred, and the probability of cointercalation and its mechanism via the decomposition of Li(solv)xCy. Recon-
following exfoliation could be minimized. EC seems structed based on ref 251.
to be such a component due to its reactivity.212-214
As a measure of the ease with which electrolyte The early studies have identified the existence of
components could be reduced, Peled and co-workers graphite intercalation compounds with solvent mole-
proposed using the vast data bank of the rate cules.239-243 On the basis of the knowledge about
constant (ke) of reduction in aqueous media, and a these compounds and their reactions, a mechanism
fair correlation was established between this constant for SEI formation was proposed later by Besenhard
and the SEI formation potential.106 Thus, ke could be that involves the initial formation of a ternary GIC
used as a tool for the selection of electrolyte solvents [Li(solvent)xCy] and its subsequent decomposition
and salts. According to this model, ideal electrolytes near the edge sites of the graphene planes to form
should be formulated with the candidates of ke > 109 the SEI.251 Figure 12 schematically depicts the SEI
M-1 s-1. By this standard, AsF6-, EC, vinylene formation process according to this mechanism. Upon
carbonate (VC), and CO2 are favored, while BF4- and cathodic polarization of the graphite anode, the
ClO4- are not. solvated lithium ion migrates to the negatively
This model conceptually forms the theoretical basis charged surface of graphite and is intercalated into
for the later development of electrolyte additives, as graphene layers at ∼1.0-0.80 V before any reduction
evidenced by the success of CO2 and VC in suppress- occurs. The ternary GIC thus formed, for example,
ing the irreversible capacity in the initial cyclings. Li(EC/DME)xCy, has a short lifetime and decomposes
But so far as major electrolyte components are within the time scale of slow scan CV (∼104 s), as
concerned, this model is not widely applied, since indicated by the irreversible peak observed when the
scanning rate is low; therefore, according to Besen-
many other properties such as ion conduction and
hard et al., this process might easily be mistaken as
phase diagrams must also be taken into account if
an ordinary irreversible reduction of the electrolyte.
the component is present in high concentrations.
However, at certain faster scan rates (e.g., 10 mV
6.2.2.2. Besenhard Model: Ternary Graphite In- s-1), part of the solvated ion could still be reversibly
tercalation Compound (GIC). In addition to the removed from graphene interlayer sites.251 The re-
indiscriminate versus selective reductions, another ductive decomposition of these cointercalated solvents
fundamental difference between graphite and lithium then renders an SEI that extends from the graphite
electrodes is the presence of the interlayer voids of surface at the edge sites into the interior of the
the former that could accommodate both lithium ions interlayer voids.
and solvent molecules. Therefore, some researchers The direct evidence that Besenhard et al. presented
argued that the reductive decomposition of electro- for the formation of a ternary GIC is the dilatometric
lytes in contact with the former might not be only a measurement of the graphite electrode, which indi-
simple surface reaction as suggested in Peled’s model. cates a crystal expansion of 150% at the cointerca-
Instead, the solvent could cointercalate into graphene lation potential.251 However, this expansion due to
layers before they decompose therein, and the pas- solvent cointercalation was never confirmed on the
sivation film thus formed could penetrate into the microscopic level. All of the in situ X-ray diffraction
structure of graphite. (XRD) measurements conducted by different research-
ers on the cycling graphite failed to offer any evidence

that a substantial change in the interlayer distance
occurred around the cointercalation potential.96,124,252,255
In these experiments, the maximal shift in 2θ of the
(002) diffraction peak during graphite cycling, which
can reflect the size of the guest species, only corre-
sponds to an expansion of 0.35 Å in the c-axis; hence,
bare lithium ions seem to be the only species that
can be accommodated.252-254 It should be pointed out
that the existence of ternary GICs is beyond question,
and the doubt here is about whether they could be
Figure 13. Correlation of gas evolution on a graphite
electrochemically formed under a similar situation electrode in 1.0 M LiClO4/PC/EC (50:50) with the irrevers-
in the forming of lithium ion cells.256-260 So far, the ible process at 0.80 V during the first discharge. Note the
chemical or electrochemical synthesis of ternary GICs level off of gas volume as soon as reversible lithium ion
has failed to produce any compositions based on intercalation starts. (Reproduced with permission from ref
carbonate solvents, despite the success of Ogumi and 261 (Figure 2). Copyright 1993 The Electrochemical Soci-
co-workers with various solvents such as ether or ety).
alkylsulfoxide,257-260 casting more doubt on the Be- Scheme 9. Electrochemical Reduction of PC on
senhard model. Graphite: Ternary GIC Mechanism
In defense of the Besenhard model, Chung et al.
argued that the lack of XRD evidence for ternary
GICs might be simply due to either their rapid
decomposition or the localized presence of them near
the graphite edges.255 Since XRD probing of the
material is based on the averaged diffraction re- Scheme 10. Electrochemical Reduction of PC on
sponse of the sample lattice, these wider spacings of Graphite: Modified Ternary GIC Mechanism
the ternary GICs might not be detected as an
averaged bulk property.
In situ EQCM studies of graphite in various
electrolytes also challenge the formation of ternary
GICs with the real-time monitoring of the graphite
electrode mass increase during cathodic polariza-
proposed a surface mechanism involving a two-
tion.101 It was found that, between 0.8 and 0.5 V,
electron process as shown in Scheme 8. Arakawa and
where such GICs are supposed to be stable, the mass
Yamaki quantitatively analyzed the gas volume
change per quantity of electricity (∆m/∆Q) was 27-
generated during the electrochemical decomposition
35 g F-1, corresponding well to Li2CO3 that has a ∆m/
of PC on a graphite electrode and found a mismatch
∆Q of 36.9 g F-1. If solvents such as EC (∆m/∆Q )
between the Coulombic quantity and the equivalents
88.07 g F-1), PC (∆m/∆Q ) 102.1 g F-1), or the
of propylene gas generated, with an efficiency be-
solvated lithium ion [Li+(PC)n] (∆m/∆Q > 300 g F-1
tween 50% and 70%, depending on current density.243
assuming a coordination number of 3) cointercalate
Apparently this result conflicts with Scheme 8, and
into the graphene structure, the corresponding mass
other reaction processes must also exist simulta-
gain on the graphite anode, which is too conspicuous
neously. Using a kinetic treatment, they suggested
to miss, should have been well recorded by the quartz
a mechanism (see Scheme 9) where a ternary GIC is
crystal sensor.
the intermediate, which underwent two parallel but
The thermodynamic stability of a ternary GIC is competitive paths to form either the gaseous product
also questionable. Obviously, between a bare lithium propylene and Li2CO3 or lithiated binary GICs. Using
ion and one solvated by molecular dipoles, the this mechanism, Arakawa and Yamaki successfully
intercalation of the former between two giant explained the relation between gas volume rate and
graphene anions is far more favored thermodynami- time.243
cally than that of the latter. The fully lithiated GIC Following a similar approach, Shu et al. used an
LiC6, for example, does not solvate in nonaqueous EC/PC mixture instead of neat PC as electrolyte
electrolyte solvents, and the tendency of lithium to solvent, and their analysis of propylene gas volume
prefer binary (i.e., without solvent cointercalation) corroborates the observations of Arakawa and Ya-
instead of ternary GICs has also been noticed in the maki.261 Furthermore, because EC was present in
solution syntheses.257 their electrolyte, the reversible lithium intercalation
Despite the concerns raised by XRD, EQCM, and could occur after a long plateau at ∼0.8 V (represent-
thermodynamics, the Besenhard model still received ing PC decomposition), therefore a correlation be-
extensive support from various experimental obser- tween the gas volume and this irreversible process
vations as summarized below and soon became the was able to be established, as shown in Figure 13.
prevalent model used by researchers in the lithium Considering Aurbach’s spectroscopic observations (to
ion battery community. be discussed later), a modified mechanism (see
It had been discovered earlier that when electrolyte Scheme 10) was proposed by Shu et al., wherein a
solvents decompose reductively on graphite, one of competition exists between the surface reaction lead-
the products is gaseous propylene.237 Dey et al. ing to radical anions and the formation of ternary
GICs, via the one-electron process in both cases.

According to Shu et al., these intermediate species
underwent further single-electron reduction and
produce Li2CO3 and propylene gas, while alkyl car-
bonates are generated via radical termination as
shown in Schemes 1 and 755,214 to become the major
ingredients in the surface film.
In separate research, Matsumura et al. performed
quantitative analyses of lithium content in graphite
electrodes with plasma spectroscopy and correlated
the results with the quantities of electricity associ-
ated with the reversible and irreversible processes.262
With XPS, they found that after the graphite is
delithiated, there is a certain amount of lithium
remaining in the graphite that cannot be removed
electrochemically. Depth profiles established by sput-
tering the graphite sample with O2+ confirmed that
these lithium species are distributed rather evenly
in the bulk of the graphite. The author ascribed the
existence of these electrochemically nonremovable Figure 14. Schematic diagram summarizing the stepwise
lithium ions as the result of lithium reaction with formation of the SEI on a graphite surface. (Reproduced
with permission from ref 266 (Figure 10). Copyright 1997
active sites on the carbon surface.262 However, there The Electrochemical Society.)
is also the possibility that these lithium signals are
from the lithium-containing ingredients in the three- the corresponding shift of this band in the EC/DMC-
dimensional SEI that exists in the graphite matrix. based electrolyte. But in the presence of DME, the
Kim and Park investigated the mechanism of graphite surface structure was detected to be ir-
lithium ion intercalation in graphite anodes employ- reversibly altered in the range between 0.9 and 0.5
ing solid-state NMR.263 Their results perhaps offer V, as indicated by a shoulder on the E2g2 band. Since
the most direct evidence in support of Besenhard’s no lithium ion intercalation is supposed to occur in
GIC model. By adding strongly coordinating additives this potential range, the authors attributed the
for lithium ions, 12-crown-4 and 18-crown-6 ethers, Raman spectral changes to the extensive DME coint-
into the electrolyte solution, they were able to observe ercalation. Interestingly enough, a DME-based ter-
an obvious Knight shift in the 7Li signals of the nary GIC was indeed electrochemically obtained and
graphite powder that was caused by the coordination. identified by Abe and Ogumi and co-workers with
Separate 13C NMR tests conducted on the same XRD.257-260 As a matter of fact, the results from this
graphite sample also identified the signals of crown Raman study support the Besenhard model but also
ether as well as carbonate (more likely its decomposi- cast doubt upon it simultaneously because no such
tion product) in the graphite powder following lithia- irreversible E2g2 band shift had been observed in the
tion. Assuming that the adsorbed additives and EC/DMC electrolytes, although obviously the SEI
solvents on the graphite surface have all been was formed in that case too.
thoroughly removed during the washing procedure Various microscopic means were also applied to
that preceded NMR measurements, the above obser- study the SEI formation process, but the reproduc-
vation should be considered as the first confirmation ibility of the results is highly dependent on the
that solvent molecules indeed are found in the bulk condition under which the observations were made
of the graphite, and their cointercalation with lithium and the pretreatment history of the samples. Even
ions during the lithiation process would most likely for the same observation, the interpretations could
be the path. However, since the author did not vary from author to author. For example, with a
present any blank test to prove the effectiveness of scanning tunnel microscope (STM), Inaba et al.
the washing procedure, the possibility of surface observed the formation of some “blisters” on the
contamination due to the remnant solvent molecules graphite surface during its cathodic polarization and
being trapped in the porous structure of the graphite described them as the swelling of the graphene layer
electrode could not be completely excluded. Never- due to solvent cointercalation;265 however, Farrington
theless, solid-state NMR proved to be an effective tool and co-workers, after observing the same phenom-
in studying the bulk structure of graphite anodes, enon with an atomic force microscope (AFM), ascribed
and more efforts on SEI mechanisms should be done these island structures to the depositions of the
with this technique. decomposition products from the solvent.266 One
In situ Raman spectra studies performed on graph- common phenomenon that was observed by all of
ite anodes also seem to reveal a cointercalation these microscopic experiments is the stepwise forma-
occurrence that leads to exfoliation. Huang and Frech tion of the surface species,265-269 which appear first
used solutions of LiClO4 in EC/EMC and EC/DME near the edge sites of the highly ordered graphite
as electrolytes and monitored the E2g2 band at ∼1580 surface at potentials as high as 1.6 V and then grow
cm-1 in the Raman spectra of the graphite that was and cover the whole electrode at potentials below 0.80
cycled between 2.0 and 0.07 V.264 Reversible lithium V, as shown in Figure 14. Since the intrinsic reduc-
intercalation and deintercalation was indicated by tion potentials of the related solvents are much lower
natures, ranging from thermodynamically stable

anions such as O2-, S2-, and halides located in the
immediate vicinity of the anode surface to partially
reduced organic species such as alkyl carbonates and
polyolefins located closer to the electrolyte solution.
They suggested that, in the SEI of such a structure,
the impedance to the intergrain lithium ion transport
would be the rate-determining step. Apparently the
pattern in which these “mosaic” pieces are arranged
is decided by the order in which the electrolyte
components are reduced. The reductions are consid-
ered to be surface processes, and the structural
difference of the anode (lithium or graphite) is
considered inconsequential.
Figure 15. Schematic drawings of various models: (a, left)
“mosaic” SEI model by Peled et al. (Reproduced with
In an attempt to explain why a linear methyl alkyl
permission from ref 270 (Figure 1). Copyright 1997 The carbonate such as DMC and EMC can form a stable
Electrochemical Society.) (b, right) Surface double layer SEI on a graphite surface while ethyl or higher alkyl
capacitor model by Ein-Eli (Reproduced with permission carbonates such as DEC cannot,107,108,271 Ein-Eli
from ref 272 (Figure 1). Copyright 1999 The Electrochemi- proposed another surface model in which the decom-
cal Society.) posed electrolyte solvents are adhered to the nega-
tively charged graphite anode surface through Cou-
(Table 5), the edge-site process at the higher poten-
lombic interactions.272 He argued that, during the
tials might have been preceded by solvent cointer-
cathodic polarization of the anode, the decomposition
calation.
products, in the form of lithium organic salts, would
Recently, Chung et al. conducted EIS studies on a
graphite electrode that had been previously exfoliated be lined up in a model similar to that of a double-
in a PC-based electrolyte and then reassembled in layer capacitor, as Figure 15b shows, and would
an EC-based electrolyte after washing.255 They found attach themselves to the existing film via the posi-
that, compared with the fresh graphite electrode or tively charged lithium ion “head” and the partially
the graphite anode that had been precycled in an EC- positively charged carbons. According to this model,
based electrolyte, the exfoliated graphite sample the effectiveness as well as the stability of the SEI
exhibited a much higher (>300%) double-layer ca- would depend on these points of adhesion. The
pacitance, which was roughly proportional to the schematic drawing in Figure 15b shows that the
surface area that was accessible to the electrolyte. decomposition products from methyl alkyl carbonate
On the basis of this observation, Chung et al. can form a tight adhesion without “loose ends”, while
postulated that the previous history exfoliation had a loose alkyl tail from DEC or higher alkyl carbonates
resulted in a significant rupture of the graphite prevents the formation of a compact film. Like Peled’s
surface, as confirmed by the STM results obtained original model37,218,219 and its modification,270 this
by Inaba et al. on the graphite exfoliation process.269 model is still based on a two-dimensional surface
Apparently, between a simple surface reaction model reaction. It must be pointed out, though, that the
and Besenhard’s model, the latter would more easily graphic representation as shown in Figure 15b might
account for this surface area increase, because the be oversimplified and should only be understood as
direct consequence of ternary GIC formation is the an empirical rule rather than an accurate picture on
creation of new surfaces. the molecular level, considering that the carbonyl
6.2.2.3. Other Models. In addition to Besenhard’s bonds in the carbonate molecules have to be distorted
model, the other models were mainly modifications out of the sp2 plane of the carbonyl carbon to meet
developed from the original Peled’s concept for lithium the contact points.
electrode passivation, with surface reaction as the To investigate the failure mechanism of graphite
major process, and emphasis was placed upon the electrodes in nonaqueous electrolytes, Aurbach and
composition and structure of the precipitated film or co-workers used various electrolyte solvents, salts,
the interaction between the precipitated products and and graphite electrodes with varying structures and
the bulk electrolyte components. morphologies and attempted to correlate the SEI
On the basis of the results from XPS studies by effectiveness with these parameters.273-275 Their
Kanamura and co-workers that the SEI has a mul- extensive studies led to the conclusions that the
tilayered structure,222,223 Peled and co-workers modi- failure mechanism is solvent-dependent and exfolia-
fied their lithium electrode passivation model to tion does not always happen to graphite anodes even
include carbonaceous anodes and proposed a so-called if they physically disintegrate. For the reductively
“mosaic model” to describe the SEI structure on the stable ether-based solvents (see Table 5), no massive
anode, as Figure 15a shows.270 According to this surface film was observed above 0.30 V, and these
model, multiple reductive decompositions occur be- solvents, coordinating with lithium ions, cointerca-
tween the negatively charged anode surface and the lated into graphene and caused complete exfoliation
various electrolyte components simultaneously, de- of the latter. In this case, the evidence of destroyed
positing a mixture of insoluble products on the anode. crystallinity was obtained by XRD. However, for
This “heteropolymicrophase” SEI consists of many carbonates that are reductively active, such as EC
microregions that are of entirely different chemical or PC (Table 5), surface precipitations were observed
at potentials well above that of lithium ion intercala-

tion, which prevented the complete exfoliation of the
graphite anode, as indicated by its mainly intact
crystallinity.255,275 Rather, the physical disintegration
in this case could only be caused by the electric
isolation of graphite fragments on a macroscopic
level, which is the result of both the solvent’s chemi-
cal nature and the graphite’s surface morphology.
On the basis of these conclusions, Aurbach and co-
workers proposed a model for SEI formation that
makes a compromise between solvent cointercalation
(for ether-based solvents) and the surface reaction
mechanism (for carbonates). According to the model,
the graphite surface in electrolytes based on carbon-
ate solvents including PC was covered with a surface Figure 16. Voltage profiles for the first two lithium
film, and the balance between the kinetics of pre- intercalation/deintercalation cycles realized on graphite
cipitation of highly cohesive surface films and the anode in t-BC/EMC and c-BC/EMC solutions of 1.0 M
kinetics of overall surface reactions producing gas- LiPF6. (Reproduced with permission from ref 255 (Figure
eous products determines whether the graphite is 7). Copyright 2000 The Electrochemical Society.)
stabilized. In deep crevices that exist on the graphite Scheme 11. Nondifferentiation of Surface
surface, such gaseous products created sufficient Reactions toward Diastereomers t-BC and c-BCa
strain and electrically disconnected the graphite
fragments. Solvent cointercalation seems less prob-
able with the existence of surface films. Obviously,
this model would be able to account for the observed
fact that surface area increases after the graphite is
cycled in PC-based electrolytes, while the graphene
crystal structure remains unchanged on a micro-
scopic scale.255
The drastic difference between EC and PC con-
cerning the stability of the graphite in them, despite
their close structural similarity, was ascribed by
Aurbach and co-workers to the effect of the methyl
group from PC, which intervenes with good adhesion
and cohesion of the formed product to the graphite
surface,275 in a very similar way, as described by Ein-
Eli, to the “alkyl loose tails”.272 Thus, the gas evolu- a An Identical Radical Anion Was Produced When the Ring-
tion kinetics is faster than the buildup of the SEI on Opening Step Eliminated the Chirality of the Ether Carbon
the graphite surface, and the electric disconnection
and regional disintegration occur, starting at the that reversible lithium intercalation/deintercalation
deep crevices driven by the gas pressure therein. occurred with t- but not c-BC. The difference between
6.2.2.4. The Mystery of Exfoliation. However, all of t-BC and c-BC in terms of the graphite performance
the above models that recognize surface reactions as is a replica of the difference between EC and PC, as
the film formation path were strongly challenged by shown in Figure 16. Such different behaviors from
the results of the comparative studies carried out by geometric isomers t-BC and c-BC would not be
Chung et al.255,276 To explore the origin of graphite expected if surface reactions dominated the formation
exfoliation, they used a series of model carbonates of the SEI, as proposed in the models by Peled,37,270
to study their cointercalation behavior with graphites Ein-Eli,272 and Aurbach,273-275 because the chirality
and found a correlation between the stereo hindrance of the carbons would be eliminated once the ring
of the solvent molecules and the behavior of the opened to form the radical anion as the predecessor
graphite anodes in the electrolytes based on them. to lithium alkyl carbonates (Scheme 11). Thus, any
These model carbonates are all structural analogues reasonable model for SEI formation has to involve a
of PC and include a pair of geometric isomer trans- mechanism in which the stereo difference between
and cis-butylene carbonates (t-BC and c-BC) as well t-BC and c-BC could be recognized and play a role in
as trans-2-pentylene carbonate (t-PC): determining the physicochemical nature of the re-
sultant SEI. Chung et al. argued that the fundamen-
tal concept of the Besenhard model (i.e., the coint-
ercalation of the solvent to form a ternary GIC before
any reductive decomposition like the one in Scheme
11) would be able to explain the above stereo effect,
since the graphene layer structure would serve as a
sub-nanoscopic sieve (graphene interlayer distance
∼ 0.35 nm) that could tell the difference in the strain
When the graphite anode was cycled in the electro- introduced by these geometric isomers, as shown in
lytes based on these carbonate solvents, it was found a schematic drawing in Figure 17. Thus, PC and c-BC
decomposition of electrolyte components leads to the

formation of a protective film on the anode. However,
these models differ in the mechanism by which the
SEI is formed, especially concerning the issue of
whether a ternary GIC is formed before the reductive
decomposition occurs. Although each of these models
can elegantly account for certain experimental ob-
servations, the Besenhard model that evolves around
solvent cointercalation seems to be supported by the
most experimental evidences, despite the fact that
the electrochemical formation of a key species of this
model, a metastable ternary GIC intermediate, has
not been experimentally confirmed. Nevertheless, it
is generally agreed nowadays that to a certain extent
the solvent cointercalation does occur and is at least
a part of the process related to the formation of the
SEI. The complete clarification of the above contro-
versy relies on obtaining more experimental evidence
at a microscopic level from further studies.
6.2.3. Characterization of Surface Chemistry
Relative to the controversy associated with the
mechanism of SEI formation, there is less uncer-
tainty in the knowledge about the chemical composi-
tion of the SEI, due mainly to the exhaustive surface
spectroscopic studies carried out by Aurbach and co-
workers on carbonaceous anodes in various nonaque-
Figure 17. Schematic drawing of the GIC-exfoliation ous electrolytes, adopting both in situ and ex situ
model. Differentiation of the stereo difference among EC, approaches.104,108,123,124,249,250 Table 6 lists the chemical
PC, and related carbonates by graphene structure. compounds as identified by these spectroscopic means
and the proposed chemical reactions leading to those
have similar stereo hindrances and would exert species.27-284 As it has been pointed out, the solvents,
similar strains into the graphene structure and cause especially the cyclic carbonates, play a more impor-
exfoliation, while t-BC may experience severe geo- tant role in the surface chemistry of the anode than
metric constraints for cointercalation. Hence, it can the salt anions.178
exhibit a reaction behavior on a graphite surface that Compared with the surface chemistry of nonactive
is qualitatively different from that of PC. electrodes212-214 or lithium electrode,55,117,118,209 simi-
Again, caution must be taken when using the lar chemical species were identified despite the
graphic representation as shown in Figure 17 be- differences in the electrode surfaces. A major modi-
cause such oversimplification might be misleading fication of the previously accepted two-electron re-
when a complex process such as the formation of an ductive pathway as suggested by Dey and Sullivan237
SEI on graphite is handled. For example, the graphic was proposed by Aurbach and co-workers based on
representation in Figure 17 suggests that cointerca- the identification of lithium alkyl carbonate by FT-
lation of solvent occurs without the assistance of IR.108,124,249 They suggested that the surface reductive
lithium ions, while, in actual cointercalation, the process for most carbonate molecules proceeds via a
solvents that coordinated with lithium ions in the single-electron path leading to the intermediate, as
solvation sheath would preferentially cointercalate. shown in Schemes 1 and 7, and that Li2CO3 and
With the supermolecular structure of the solvation alkylenes were formed through either the continued
sheath in consideration, it would be more difficult to reduction of this intermediate or the secondary
predict the effect of diastereomers on cointercalation reaction between it and trace moisture in the system.
by the analogue of Figure 17. Specifically, the following structures were assigned
On the other hand, with an average solvation to the decomposition products from EC and PC,
number of four, it would be hard to imagine that which supposedly constitute the main composition of
cointercalation would occur without breaking the the SEI layer:102,104,117
solvation structure of the lithium ion, considering the
required expansion in the graphite structure to
accommodate such a gigantic guest. Most probably,
at the edge sites of the graphite where cointercalation
occurs, the solvated lithium ion is progressively
stripped of its solvation sheath, and the free solvents
would then insert into the graphite interior in a close
way, as shown by Figure 17. The predecessor of alkyl carbonate, a radical anion,
To summarize, various models have been proposed has been experimentally observed by ESR, the life
to depict the formation of an SEI on a graphite anode, span of which depends on the carbonate structure
based on the common knowledge that the reductive and ranges from minutes to hours but is independent
of the salt species used.178 Moreover, this one-electron

reduction mechanism does seem to be strongly fa-
vored by the fact that the intermediate does not
convert to alkylenes in 100% yield, reminiscent of the
work by and Arakawa and Yamaki243 and Shu et
al.261 who have already reported that the generation
of gaseous products does not match the electric
quantity injected into the anode.
A more quantitative confirmation of alkyl carbon-
ate formation came from transmission electron mi-
croscopy (TEM), by which electron diffraction (ED),
electron energy loss spectra (EELS), and imaging
were conducted on disordered and graphitic carbon-
aceous anodes cycled in LiClO4/EC, respectively.280
Although the ED does not identify any crystalline
phase other than the hexagonal structured graphite
and Li2CO3 on the surface (probably suggesting that
alkyl carbonates are noncrystalline), the atomic
concentration ratio of oxygen and carbon (O/C) as
determined by EELS ranges between 1 and 1.5,
unequivocally indicating that the proposed alkyl
carbonate formed from EC reduction. Interestingly,
this O/C ratio was also found to vary with the
potential to which the carbonaceous anodes were
cathodically polarized: above 0.90 V an O/C ratio )
3.0 was obtained for the presence of Li2CO3, while
below 0.80 V an O/C ratio ) 1.0-1.5 was obtained.
Thus, the authors proposed that the SEI was formed
in a two-step process: (1) formation of Li2CO3 occurs
at potentials between 1.0 and 0.80 V, and (2) forma-
tion of alkyl carbonates is favored at lower potentials Figure 18. Depth profile of various chemical elements in
below 0.80 V. the SEI formed on HOPG: (a) basal plane and (b) edge
In retrospect, probably a more reasonable explana- section. (Reproduced with permission from ref 281 (Figure
tion could hereby be proposed concerning this poten- 4). Copyright 1999 The Electrochemical Society.)
tial-dependent reductive decomposition. Combining
the observations from Shu et al. about the correlation previously observed on a lithium surface and identi-
of gas products with a 0.80 V plateau,261 Farrington fied as alkyl carbonates by Kanamura et al.222,223,282,283
and co-workers about the deposition of the film onto after referencing with the C 1s signal of Li2CO3 at
the surface,266 Aurbach and co-workers about the 290.5 eV. Surface sputtering with Ar+ reduces the
competition between single- and two-electron reduc- abundance of this species rapidly, suggesting that
tive paths,108,124 and Kanamura et al. about the this species might only be stable on the top layer of
multilayered structure of the SEI,222,223 one can the SEI.
conclude that, at potentials above 0.80 V, where the Besides lithium alkyl carbonates, XPS also identi-
film has not completely covered the basal surface of fied a wide variety of decomposition products from
graphite, the two-electron process as proposed in other electrolyte components, including polyether
Scheme 8 is likely the predominant process because moieties as well as the lower valence As species. The
of the good electronic conductivity of the graphite depth profile of the SEI established by prolonged
surface, leading to Li2CO3 and ethylene; below 0.80 sputtering by Ar+ reveals the multilayered structure
V the single-electron process prevails because of the of the SEI, as shown in Figure 18a, in which the
much slower electron hopping kinetics, leaving the organic species such as alkyl carbonate and Li2CO3
surface with alkyl carbonate depositions and render- are present in predominant percentages on the solu-
ing negligible gas evolution simultaneously. There- tion side of the SEI while simple inorganic species
fore, the potential-dependence of the decomposition such as Li2O or As(0) are more stable on the graphite
compounds as observed in a TEM study by Naji et side of the SEI probably because of the more complete
al.280 is actually the potential-dependence of the reduction facilitated by faster electron-tunneling
competition between one- and two-electron paths, as kinetics.
shown in Schemes 1 and 7. Differences in chemical composition were also
In addition to FT-IR, XPS experiments performed observed in the SEIs formed on basal planes and edge
on a graphite anode that had been cycled in various sites. The former were more enriched with organic
carbonate-based electrolytes also identified an alkyl species and the latter with inorganic species, espe-
carbonate species. Bar-Tow et al. characterized the cially with the decomposition products that obviously
surface of a highly oriented pyrolytic graphite (HOPG) originated from salt anions. The authors thus con-
that had been cycled in LiAsF6/EC/DEC and found cluded that, on the basal plane, the major contribu-
the C 1s signal located at 289 eV,281 which had been tion to SEI formation is from solvent reductive
Scheme 12. Formation of Polymeric Species in the were formed in the SEI instead of an alkyl carbonate
SEI such as C2H5OCO2Li.271,279
In the FT-IR analysis of an SEI on an anode
surface, Ross and co-workers also disagreed with
Aurbach and co-workers on the interpretation of the
spectra.285 The difference is mainly concentrated on
the assignments of two absorption peaks at 838 and
decomposition, whereas, at edge sites, the salt anion 1650 cm-1, respectively, which were ascribed to the
decomposition occurs on a larger scale.281 This loca- bending and stretching of carbonyls in lithium alkyl
tion-dependence of the SEI chemical composition carbonate.55,118 While the former was indeed observed
strongly implies that the edge sites of the graphite by Ross and co-workers in their spectra, they ex-
are the main entrances for lithium ion intercalation. cluded the possibility of it arising from lithium alkyl
XPS studies, carried out with LiBF4- and LiPF6- carbonate in view of its strong intensity. By compar-
based electrolytes,277 or with carbonaceous anodes ing the IR spectra of an authentic sample of lithium
with varying graphitic degrees,125 or at different monoethyl carbonate and propylene oxide, they sug-
potentials after being fully cycled,125 rendered quite gested that the signal might originate from the
similar results, except that the abundance of LiF asymmetric stretching of the O-C-O linkage from
increased in the SEI because of the higher sensitivity an epoxy structure, which was generated from the
of LiBF4 and LiPF6 to moisture. Andersson and reductive decomposition of EC via an electrochemi-
Edström observed a large amount of polymeric carbon cal-chemical process that involved the reduction of
on the cycled graphite surfaces that was linked with trace moisture in the electrolyte.285 Considering the
an ether-like oxygen, as evidenced by the signals low bp of this compound (10.7 °C) and its strong
located between 285.5 and 286.5 eV.277 They at- tendency toward ring-opening polymerization by the
tributed the formation of these polymeric species to catalysis of acids or cations, it is very unlikely that
the ring-opening polymerization of cyclic carbonates it would be a persistent ingredient of the SEI.
initiated by strong Lewis acids such as PF5 (Scheme Perhaps ethylene oxide was only detected as a
12). transient species in the IR studies, which eventually
Aside from cyclic carbonates, the decomposition polymerized into an oligoether species. On the other
products from linear carbonates were also identified hand, the signature stretching mode of alkyl carbon-
in the forms of either lithium alkyl carbonates or ate at 1650 cm-1 was completely absent in IR spectra
alkoxides, as shown by Scheme 7 and also in Table obtained by these authors.278 By comparison with
6a.125,271,279 Relatively, the reduction of linear carbon- synthesized spectra based on authentic compound
ates was thought to be less consequential as com- samples, the carbonyl stretching of high intensity
pared to their cyclic counterparts, apparently due to was assigned to lithium oxalate (Li2C2O4) and suc-
their smaller presence in the solvation sheath of cinate (LiO2CCH2CH2CO2Li) (Table 6a). While the
lithium cations.170,171 former was obviously formed from CO2 through some
In accordance with these experimental results, reductive process, the source for the latter remains
Wang et al. employed density functional theory unclear. A likely mechanism is the recombination of
calculations to comprehensively examine the possible radicals, which would explain the formation of new
reduction pathways for EC molecules in super- C-C bonds. At the same time, the presence of trace
molecular structures Li+-(EC)n (n ) 1-5) and found moisture complicates the surface chemistries by
that, thermodynamically, both one- and two-electron reacting with lithium alkoxides and renders the SEI
reductive processes are possible.170 A complete array with species such as LiOH, CH3OH, and LiHCO3.
of the possible reduction products from EC was listed
in their paper considering the various competitive Apparently, the above controversy regarding the
pathways, and they concluded that both (CH2OCO2- identification of chemical species in the anode SEI
Li)2 and (CH2CH2OCO2Li)2 are the leading species will require further analytical work to resolve.
in SEI, while minority species such as lithium alkox- In addition to chemical compositions, another
ide, lithium carbide, and the inorganic Li2CO3 co- important aspect about SEI formation that is of
exist.170 practical significance to the forming of lithium ion
A more recent XPS study by Ross and co-workers cells is the potential range in which the above
challenged the above general conclusion that lithium reactions occur leading to the formation of the SEI.
alkyl carbonates constitute the main composition of Because of the earlier observation of the plateau near
the SEI.284 Using ultrahigh vacuum surface analyti- 0.80 V36,237 and the concurrent gas evolution,243,261 it
cal methodology, they meticulously examined the was generally believed that the formation of the SEI
reaction between a fully lithiated HOPG and DEC occurred at this potential. An in situ analysis using
as a function of temperature and found that all a differential electrochemical mass spectrum (DEMS)
carbonate signatures above 289 eV rapidly disappear performed on graphite as well as other electrode
in the room-temperature range, giving way to Li2O, surfaces confirmed the above process by simulta-
although alkyl carbonate as a metastable species did neously recording voltammetry and analyzing the
exist in the subambient temperature range briefly. gaseous products from the electrode being cathodi-
Between 0 and 100 °C, the C/O stoichiometry ratio cally polarized. It was found that hydrogen and
correspondingly changes from the known 1.6 for DEC ethylene were produced during the first cycle at about
to 1.0. Combining these observations, they suggested 1.50 and 0.80 V, respectively, as indicated by their
that oxalate species along with lithium methoxide molecular peaks as well as fragmentation patterns
Table 6. Reductive Decompositions on Carbonaceous Anodes

A. Electrolyte Solvents
B. Electrolyte Salts and Other Components
in the MS.286 No sign of CO or CO2 had been found, contamination; thus, the SEI formation process
contradicting Schemes 1, 7, and 8. The production of seemed to start below ∼0.80 V. However, since the
H2 was ascribed to the reduction of trace moisture MS can only detect gaseous species, another surface
spectroscopic means based on FT-IR was used as a

complimentary analysis. It was concluded that, since
the salt anion concentration near the electrode
decreases rapidly at the same potential as H2 is
formed, the first electrolyte decomposition should
occur at potentials near 1.50 V. Microscopic observa-
tions on the graphite surface cycling in nonaqueous
electrolytes were in agreement with this conclusion;
that is, surface deposits were visualized at potentials
as high as 2.0-1.50 V near the edge sites, while basal
planes would be completely covered at potentials
below 0.80 V.266,267 The combination of DEMS286 and
AFM266,267 results perfectly echoed the conclusions
drawn by Bar-Tow et al.281 that the film at the edge Figure 19. Dependence of the resistance (Rf) of the SEI
film on the cell voltage for the lithium|graphite half-cell
sites was mainly formed by the reductive process of using LiPF6/EC/EMC (3:7 wt) as electrolyte. Shown are the
salt anions at higher potentials (∼1.50 V), while the data for the 1st and the 10th cycles under galvonostatic
film on basal planes that is formed at lower potentials conditions with the rate of 0.05 mA h cm-2. (Reproduced
(∼0.80 V) has the contribution mainly from solvent with permission from ref 287 (Figure 4). Copyright 2001
reductions. The Electrochemical Society.)
A few other studies on the SEI placed the occur-
rence of SEI formation in lower potential ranges that SEI film when the graphite anode is repeatedly
overlap with those of the lithium intercalation pro- cycled in LiPF6/EC/EMC (3:7 wt). In the potential
cess, that is, below 0.25 V. Using a plasma spectrom- range above 0.25 V, Rf remains relatively stable at a
eter, Matsumura et al. conducted quantitative chemi- high level, where solvent decomposition is believed
cal analyses on the lithium content in carbonaceous to account for half of the irreversible lithium. The
anodes that were electrochemically lithiated at dif- SEI film at this stage is incomplete and porous and,
ferent stages and correlated the irreversible lithium hence, unable to shield the graphite surface from
content (i.e., the electrochemically irremovable lithium further solvent reactions. A drastic decrease in Rf
that was consumed to form the SEI and then re- within a narrow range of 0.20-0.04 V, which ac-
mained therein) with the cathodic polarization po- counts for more than half of the irreversible capacity,
tential.262 Unexpectedly, they found that a rather probably represents a process in which the compact
high portion (>80%) of the lithium was distributed and conductive SEI film is formed and starts to
in the potential range 0.50-0.01 V, with a relatively protect the graphite surface.287 Thus, the authors
small percentage in the potential range above 0.80 proposed that SEI formation on a graphite surface
V, where the formation of the SEI was thought to actually goes through two stages: (1) at potentials
have occurred. above 0.25 V, where solvent decomposition leaves a
primitive surface deposition that is porous and also
Similar results were obtained by Zhang et al., who electronically conducting, and (2) at potentials below
analyzed the galvanostic cycling data of a graphite 0.25 V, where an insulating and compact SEI is
anode in LiPF6/EC/PC/EMC and calculated the dis- formed. Once formed, this SEI film is protective of
tribution of irreversibility in the first lithiation the graphite anode, and a reversible impedance
process in different potential ranges.287 They found response is expected, as shown by Figure 19 for all
that, of the total 7% irreversible capacity in the first of the lithiation/delithiation processes after the first
forming cycle, a majority (∼6%) was generated within lithiation. The above potential-dependence of SEI
the potential range between 0.25 and 0.04 V, where resistance was reproducible when LiPF6/PC/EC/EMC
different stages of LiGIC were formed. The quantita- (1:1:3 wt) was used despite the higher irreversible
tive difference between the results of Zhang et al.287 capacity due to the presence of PC.287
and Matsumura et al.262 on the irreversible lithium
The origin of these reductive processes in the
at low potential ranges might be due to the different
lithiation potential range remains unclear, but it does
graphitic degrees of the carbon anode materials that
not necessarily involve new components from bulk
were used in their investigations, but the common
electrolytes. One possibility is that the radical inter-
observation of these two experiments is that, concur-
mediates, as shown in Schemes 1 and 7, undergo
rent with lithium intercalation, some other irrevers-
further single-electron reduction, leading to lithium
ible processes are also underway. A reasonable
salts with simple anions such as CO32-, O2-, or
explanation would be that SEI formation is not
F-,222,223 while reactions between intercalating lithium
completed above 0.80 V but continued deep within
ions and the active sites of the graphite interior such
the potential range where lithium ion intercalation
as hydroxyls or carbene radicals are also possible.262
proceeds.
The electrochemical impedance analysis carried out 6.3. Electrolyte/Cathode Interface
in the same study by Zhang et al. seemed to confirm
the above speculation with the change in the resis- Compared with the research interest in the passi-
tance of the SEI film as a function of the lithiation vation of anodes, especially materials with carbon-
potential and corresponded well with the irreversible aceous origins, there have been relatively few studies
capacity analysis. Figure 19 shows the potential- dedicated to the understanding of the interface
dependence of the resistance for lithium ions in the formed between electrolytes and cathode surfaces. A
probable reason might be the stability of these

cathode hosts against solvent cointercalation and
exfoliation due to the layered structure being held
together by Coulombic interactions between op-
positely charged slabs composed of metal cations and
oxide anions.288 Nevertheless, irreversible decomposi-
tions, although oxidative in nature, do occur on
cathode surfaces, and the decomposition products
very likely form a passivation film that prevents any
sustained solvent decomposition. In this sense, the
interface between electrolyte and cathode should
possess the same physicochemical fundaments of the
SEI on anodes, that is, electronic insulator and
lithium ion conductor. A number of authors have
used the term “SEI” as well to describe the passiva-
tion of cathode surfaces in lithium/lithium ion cells;
however, by the currently accepted convention the
term is usually reserved for carbonaceous anodes.
Instead, a more general reference of “passivation Figure 20. Equivalent circuit based on surface layer
film” or “surface layer” has been used for electrolyte/ formation on cathode materials (a, top) and the electrolyte/
cathode interfaces. cathode interface (b, bottom). (Reconstructed based on ref
295.)
6.3.1. Passivation Film on a Cathode
the solid to occur. This equivalent circuit and its
The detection of protective films on cathode sur- numerous variations have become a universally
faces has not been as straightforward as that of the adopted model that simulates the behavior of both
SEI on anodes, partially because a native surface anode and cathode in nonaqueous electrolytes.
film, mainly composed of Li2CO3, already exists on In formulating new electrolyte compositions that
all transition metal oxide cathode materials based on can withstand the high potentials of the cathode
manganese, cobalt, and nickel.289-294 This surface materials, Guyomard and Tarascon also realized that
component could arise from the precursors used to oxidative decomposition of electrolyte components
synthesize these metal oxides or, more likely, from occurred on cathode surfaces, and passivation of the
the reaction between the metal oxides and the CO2 surface prevented the bulk electrolytes from further
in the atmosphere during the processing of these decompositions.93,98 Using LiPF6/EC/DMC and the
strongly oxidizing materials.290 Upon contact with manganese spinel LiMn2O4, they systematically ex-
electrolytes, this native film is usually eroded by the plored the origin of the oxidative decomposition by
acidic electrolyte salts currently employed by the quantitatively analyzing the irreversible capacity
lithium ion industry, and to make things more associated with electrolyte oxidation and established
complicated, the active materials are usually involved a correlation between it and the surface area of the
in the subsequent oxidation of the electrolyte solvents composite cathode. Thus, they concluded that the
on the exposed cathode surface. Proving the existence oxidative decomposition of electrolytes on cathodes
of a surface film on the cathodes has been difficult, is surface-catalyzed, and very likely the cathode
and spectroscopic identification, which has been surface would be covered evenly with a surface layer
proven to be an effective tool in studying the SEI on that is composed of the decomposition products.
anodes, often yields ambiguous results. Almost during the entire 1990s, the main interest
Goodenough and co-workers were perhaps the of the lithium ion research community was focused
earliest authors to suggest that a film exists on the on electrolyte/anode interfaces while its cathode
cathode/electrolyte interface.295 In an attempt to counterpart was overlooked until various lithium ion
simulate the ac impedance responses of an intercala- systems, especially those based on manganese spinel
tion-type cathode in liquid electrolytes, they discov- cathodes, were found to suffer power loss and capac-
ered that Li1-xCoO2 in PC cannot be described as a ity fade upon prolonged cycling or storage at elevated
simple intracompound lithium ion diffusion; instead, temperatures.296-302 Preliminary diagnostic studies
a more complex electrochemical process, including pointed to electrolyte/cathode interfaces as the source
the formation of a surface layer on the electrode due of the degradation, and intensive research has been
to the oxidation of the electrolyte, must be considered. carried out to address this issue since the late 1990s.
They proposed an equivalent circuit, shown in Figure Aurbach et al. studied the interfacial behavior of
20, for the lithium intercalation process into such various cathode materials in LiAsF6/EC/DMC using
cathodes, wherein Re, Rsl and Csl, Rct and Cct, and Zw EIS and found that, for all of these cathodes, that is,
represent the bulk resistance of electrolytes, the LiNiO2, LiCoO2, and LixMn2O4, the impedance spec-
resistance and capacitance of the surface layer, the tra obtained reflected several processes in series.290
resistance and capacitance of the charge-transfer In other words, the overall lithium ion intercalation
process, and the Warburg impedance, respectively. in and deintercalation from a variety of LiMO2 bulk
In such a model, lithium migration has to go through materials included the inevitable step of lithium ion
the surface layer because of its serial nature in order migration through a certain surface layer in a man-
for the charge-transfer and further diffusion within ner very similar to that of the reversible lithium ion
intercalation/deintercalation process on carbonaceous Scheme 13. Self-Discharge of Spinel Cathode in

materials. Electrolyte
Complementary to the EIS results, wide-line 7Li
NMR research conducted by Greenbaum and co-
workers also supported the existence of such a contact with these cathodes.292 They speculated that
surface layer.299 They reported that the lithiated the hydrolysis products from the salts such as HF,
(discharged) cathode LiNi0.8Co0.2O2 showed two iden- AsF3, or AsF5 reacted with the native surface layer
tifiable 7Li signals with a negligible difference in and led to its breakdown. Thus, it is this heavy
chemical shift, and these signals were characterized involvement of salt anion that generates the detected
by vastly different spin-lattice relaxation times (T1). LiF in the surface layer. By correlating with the EIS
The authors assigned the one with T1 < 10 ms to the study, the effect of LiF on cell impedance was
intercalated lithium ions in the bulk cathode and the confirmed. During prolonged cycling, the surface
other on the order of seconds to the lithium ions layer on the cathode seemed to continue to grow;
trapped in the surface layer. Corresponding 19F NMR however, the chemical compositions did not experi-
also revealed the existence of a different F other than ence qualitative changes. Through comparative stud-
the binder PVdF; hence, the authors suggested that ies, the latter authors confirmed Aurbach’s observa-
LiF constitutes the surface layer. However, the tion that the pristine Li2CO3 surface layer on
double-resonance (7Li-19F) measurements that would LiNi0.80Co0.20O2 was changed, as evidenced by a
be sensitive to the presence of LiF were inconclusive. wealth of new chemical components corresponding
Visual detection of surface layers on cathodes using to the oxidations of both solvents and salt anions,
microscopy techniques such as SEM seems to be with LiF being the dominant species.293,294 In view
supportive of the existence of LiF as a particulate- of the highly oxidative nature of metal oxides, these
type deposition.292,294 The current sensing atomic authors suggested a spontaneous formation process
force microscope (CSAFM) technique was used by for the new surface layer, in which the direct (hence,
McLarnon and co-workers to observe the thin-film nonelectrochemical) redox reactions occur between
spinel cathode surface, and a thin, electronically the active materials of cathode and electrolyte com-
insulating surface layer was detected when the ponents, leading to the lithiation of the former and
electrode was exposed to either DMC or the mixture the oxidation of the latter (Scheme 13). On a man-
EC/DMC. The experiments were carried out at an ganese spinel cathode, this spontaneous process was
elevated temperature (70 °C) to simulate the poor complicated by the coupling of the surface oxidations
storage performance of manganese spinel-based cath- from electrolyte components with the concomitant
odes, and degradation of the cathode in the form of Mn2O4 disproportionation and Mn2+ dissolution.293,300
disproportionation and Mn2+ dissolution was ob- Controversial results from the surface chemistries
served.300 This confirms the previous report by Taras- were obtained.
con and co-workers that the Mn2+ dissolution is acid- A few authors argued that the existence of surface
induced and the electrolyte solute (LiPF6) is mainly layers on cathodes could also be confirmed by the
responsible.301 irreversible capacity that only occurs in the first
The speculation about LiF in the surface layer was charge process.303,304 By correlating this irreversible
supported by McBreen and co-workers, who detected capacity with cell voltages during the forming process
the formation of a surface layer on well-cycled of a LiNiO2-based cathode, they concluded that the
LiNi0.85Co0.15O2 in LiPF6/EC/DMC with soft X-ray surface layer was formed via a two-step process: (1)
absorption spectroscopy (XAS).302 In the energy re- below 3.40 V, where a primitive and resistive film
gion between 500 and 950 eV, the cycled cathode was formed, and (2) between 3.4 and 3.8 V, where a
demonstrated major changes at O and F K-edges as highly conductive film was built up through a further
compared with the pristine powder or as-prepared oxidative breakdown of the primitive layer. The EIS
cathode samples. By comparing with standard LiF study that was carried out simultaneously showed
XAS, the authors believed the signals to be arising an obvious potential-dependence of the resistance
from the mixture of PVdF (as binder) and LiF on the associated with the surface layer (Figure 20), which
cathode surface. The source of LiF was believed to is characterized by a dramatic drop in the corre-
be generated by the decomposition of LiPF6. Consid- sponding voltage regions.303 The possibility of cathode
ering that the organic decomposition products are structure reorganization should not be excluded for
unstable on highly charged cathodes, the authors the irreversible stage between 3.6 and 3.8 V, though,
suggested that the surface layer on the cathode might and the authors suggested that probably both elec-
be composed of a dense coating of LiF. This very trolyte oxidation and cathode structural transition
resistive surface layer was believed to be responsible are responsible.
for the power loss of the lithium ion cell after Similar irreversible processes have been previously
prolonged storage. observed by Aurbach et al. on LiNiO2290 and Croce
Aurbach et al. and Andersson et al. also detected et al. on LiNi0.75Co0.25O2305 by slow scan cyclic volta-
the presence of surface layers on cathodes with XPS. mmetric (SSCV) measurements. However, in their
The former authors noted that the presence of salts EIS studies the former authors did not observe any
(LiPF6 and LiAsF6) played a crucial role in changing obvious potential-dependence for the resistance as-
the surface state of cathode materials due to their sociated with the surface layer290 but rather an
acidic nature, because pure solvents do not change invariant resistance in the range between 3.6 and 4.2
the native surface layer, Li2CO3, when brought into V. This latter observation is in direct disagreement
Scheme 14. Possible Electrochemical Oxidation

Path for PC
with the results of Zhang et al.303,304 and Croce et

al.,305 where the change of surface-layer resistance
with potential is obvious.
Some authors pointed out that the surface layer
on the metal oxide cathodes might not be as stable
as that on the carbonaceous anode, and the cell power
loss and capacity fade are mainly caused by degrada-
tion of this electrolyte/cathode interface.293 Their
argument was supported by the increase in imped- Figure 21. FTIR spectra (diffuse reflectance mode) mea-
ance with cycling, which very likely suggests a sured from a pristine LiNiO2 composite cathode and with
continued growth of surface layers on the cathode.292 a cathode after galvanostatic cycling in Li salt/EC/DMC
There was also one report revealing the complete solutions. (Reproduced with permission from ref 292 (Fig-
absence of surface layers on cathodes after prolonged ure 1). Copyright 2000 The Electrochemical Society.)
cycling, probably indicating that the electrolyte com- Scheme 15. Possible Mechanism for the
ponents or impurities react with the surface layer Formation of Polycarbonates
and leave the cathode surface with the IR-insensitive
LiF.298
6.3.2. Characterization of Surface Chemistry
The overwhelming majority of the studies on the
oxidative decompositions of solvents and salt
anions were carried out on nonactive elec-
trodes.74,81,120,130,131,204,206,207,306-308 On the basis of the
detection of ring-opening reactions by in situ FT-IR,
Scheme 16. Surface Nucleophilic Reactions
a mechanism involving a single-electron process between LiNiO2 and Electrolyte Solvents
producing a radical cation was proposed for carbonate
solvents (Scheme 14).307 Subsequent decomposition
of the intermediates leads to gaseous as well as solid
products, which form a solid film on the electrode.
The existence of such radical species during the
oxidative decompositions of electrolyte components
has been confirmed for a wide variety of electrolytes
by ESR, and the surface of a charged cathode (LiCoO2
at 4.3 V) is identified as the source of its generation, formed on the cathode surface upon storage or cycling
because a parallel blank test with electrolytes in the but emphasized that, because of the complicated
absence of a cathode produces no radical species.309 nature of the surface oxidation processes, it is impos-
The authors depict a picture of delocalized radical sible to obtain an unambiguous picture of the surface
cation structures that are coordinated by the neigh- chemistry.292 Since the native Li2CO3 does not change
boring solvent molecules, whose half-life is on the when in contact with pure (salt-free) solvents, the salt
order of minutes depending on the electrolytes. Thus, anions obviously played a crucial role in forming
it is reasonable to believe that Scheme 14 represents these new surface species. Cycling seems to facilitate
the initial oxidative cleavage of the carbonate sol- the surface chemistry, as Figure 21 shows, since a
vents. variety of absorptions corresponding to CsH, CdO,
The systematic surface characterization was con- and CsO bonds were identified. The carbonyl func-
ducted by Aurbach and co-workers using FT-IR tionality around 1800-1700 cm-1 was believed to be
spectroscopy for LiCoO2, LiNiO2, and LiMn2O4 spinel polycarbonate species, a possible source of which
cathodes.290 After cycling in electrolyte LiAsF6/EC/ could be the oxidation of EC or the ring-opening
DMC between 3.0 and 4.4 V, all of the cathode polymerization of EC catalyzed by a nucleophilic
surfaces were found to be covered with a wealth of mechanism (Scheme 15). The nucleophilic initiator
new chemical species, while the signals correspond- RO-, as evidenced by the absorptions near 1100 cm-1,
ing to the native Li2CO3 diminished. Preliminary could be generated on the surface of the cathode
interpretation of these spectra led the authors to because of the possible reactions between the active
conclude that the species very much resemble the mass and the solvents (Scheme 16). Meanwhile, the
lithium alkyl carbonates observed on various anode signals around 1650 cm-1 (asymmetric CdO), 1350-
surfaces that have been formed, as supported by the 1300 cm-1 (symmetric CdO), 1100 cm-1 (CsO), and
signals around 1650 cm-1. With further study, they 850 cm-1 (OCO2-) still suggest species with similar
confirmed that the electrolyte-related species were structure to alkyl carbonates. Since the mechanism
Figure 22. C 1s, O 1s, F 1s, P 2p, Li 1s, and Ni 2p XPS spectra for LiNi0.8Co0.2O2 cycled in LiPF6/EC/DEC. (Reproduced
with permission from refs 294 (Figure 6). Copyright 2002 The Electrochemical Society.)
in Scheme 16 involves direct electron transfer from surface species.294 Figure 22 lists the XPS spectra
electrolyte solvents to electrode active mass, these generated from the cathode sample after the forma-
reactions are spontaneous and do not require cell tion cycle. In C 1s spectra, three new peaks were
charging to initiate. ascribed to a polymeric moiety (285.1 eV), a carbon
XPS and energy-dispersive X-ray analysis (EDAX) in ether linkage (286.8 eV), and a semi-carbonate
of the cathode surface chemistry clearly revealed such as an alkyl carbonate (290.0 eV). The polymeric
that, in various electrolytes using different salts, the moiety could be attributed to hydrocarbon species or
salt anions or the impurities related to them were polymers based on an ether linkage. On the other
also heavily involved in the surface layer forma- hand, quantitative analyses through peak devolution
tion.292 EDAX detected a pronounced increase in the showed that a majority of the surface lithium content
F content of the surface after the cathode was cycled is accounted for by the presence of LiF, LiPF6, and
in LiPF6 or LiAsF6 solutions, and appreciable amounts Li2CO3, and alkyl carbonates as well as polycarbon-
of As or P were also present on the surface. The ates are also likely. The F 1s spectrum corroborates
surface chemistries of a cathode cycled in LiAsF6 the dominant presence of LiF by the peak at 684.9
solution as detected by XPS correlated well with FT- eV, while the other F signal arises from the binder
IR, which confirmed the formation of organic and PVdF. P 2p spectra could be deconvoluted into a
inorganic carbonates with signals at 285 and 290 eV series of contributing components, which should
as well as RO-. The peak at 294 eV was ascribed to include the oxidation products from LiPF6 such as
polycarbonate species, in accordance with the peaks LixPFy.
around 1800-1700 cm-1 in FT-IR. However, the Obviously, the above compounds might result from
cathode surface that had been cycled in LiPF6 solu- several possible reactions. Since they did not find the
tions lacked the signs of all carbonate or polycarbon- transformation of the metastable alkyl carbonate into
ate species, and the authors attributed the absence Li2CO3 upon high-temperature storage, as they had
of these carbonyl-containing compounds to the role observed on carbonaceous anodes, Andersson et al.
of HF, which reacts with carbonates and leaves LiF challenged the spontaneous reaction mechanism sug-
on the surface. This is supported by the F 1s spectra gested in Scheme 16 and proposed that the carbonate
with a strong 685 eV signal, which is smaller for species on cathode surfaces would be polycarbonates
cathode surfaces cycled in LiAsF6 solution. Further instead of simple alkyl carbonates.294 Furthermore,
examination of F 1s spectra also points to other self-discharge that leads to lithiation of the cathode
F-containing species with P-F and As-F bonds, host was observed during storage; hence, the authors
which are evident in As 3d and P 2p spectra as well. argued that a mechanism similar to Scheme 13 would
Interestingly, Ni 2p signals of oxidation states for be more likely. Thus, the initiation for EC polymer-
various Ni species were also detected in all of the ization would be more likely realized by the electro-
cathode surfaces that were cycled in either LiPF6 or philic attack from the Lewis acid, PF5, a decomposi-
LiAsF6 solutions, as expected from the surface reaction product of LiPF6, instead of a nucleophilic attack
tions involving LiNiO2 as a nucleophilic agent (Scheme from RO-. The fact that PF5 reacts with cyclic
16). carbonates, leading to polymerization of the latter,
More recent XPS studies were conducted by Ander- seems to favor this rationale.298
sson et al. on a similar LiNi0.8Co0.2O2 cathode cycled In a parallel study performed on the spinel cathode
in LiPF6/EC/DEC, and more detailed efforts were LiMn2O4, Eriksson et al. reported both XPS and FT-
made to assign the signature signals for various IR results293 that conflicted with those by Aurbach
Scheme 17. Mechanism for Electrolyte Oxidation tions of the electrolytes in actual lithium ion cells,
Coupled with Spinel Disproportionation and Mn2+ because a number of factors related to the operation
Dissolution of the cell rely on it, such as the potential limit within
which the cell can be cycled, the overcharge margin
that a certain electrolyte formulation can provide,
Scheme 18. Mechanism for Electrolyte Oxidation and the design and selection of cathode materials.
on a Fully Charged Cathode Surface Nevertheless, inadequate studies have been con-
ducted so far on this issue, and most of the studies
on the oxidative decomposition of electrolytes were
et al.290,292 Besides polymeric species that were most carried out on nonactive electrodes on which the
likely polyethers, there was no sign of any carbonate- oxidative processes for carbonate-based solvents were
like species detected when the spinel was cycled in a generally believed to occur at potentials higher than
LiPF6-based electrolyte, as evidenced by both C 1s 4.0 V despite the widely scattered data from the
and O 1s XPS and FT-IR spectra. F 1s still identified different electrode surfaces, salt species, and cutoff
LiF as the major surface species along with partially criteria employed in those investigations (Tables 4
decomposed products from LiPF6 as well as the MnF2
and 5). However, considering the vast differences in
species. When LiPF6 was replaced by LiBF4, semi-
surface areas as well as surface functionalities be-
carbonate moieties began to appear in C 1s; therefore,
tween the nonactive electrodes and composite cathode
the authors concluded that HF, which was more
materials, there is sufficient reason to question
abundant in LiPF6 solutions, removed the initially
whether the decomposition limits of electrolytes as
formed carbonate species. Meanwhile, a wealth of IR
determined on these nonactive electrodes would serve
absorptions also indicated the formation of polyether
as reliable references when cathode composites of
structures for which the mechanism of formation
higher surface area and hence higher surface cata-
from EC or linear carbonates remains unclear.
lytic activity are in contact with electrolytes.
The formation of surface species on a spinel cathode
surface is complicated by the coupling of electrolyte The discrepancies in stability limits caused by the
oxidation with the cathode structural disproportion- differences in the surface areas of working electrodes
ation and the subsequent Mn2+ dissolution.296,300,310 had been noticed by Guyomard and Tarascon early
Combining Scheme 13 with the latter considerations, on.93 The solvent oxidation obviously proceeded with
a modified mechanism was proposed by Eriksson et much slower kinetics on Pt electrodes than on
al. (Scheme 17).293 This mechanism only seems to composite cathode materials based on spinel man-
apply when the cathode is not fully charged (<4.3 ganese oxide. Cattaneo and Ruch311 and Imhof and
V). Since disproportionation can no longer occur at Novák312 also recognized the importance of using the
the fully delithiated potentials (4.3 V) where almost actual battery electrode materials instead of nonac-
all manganese species are in the Mn4+ state and tive electrodes to obtain more relevant decomposition
λ-MnO2 exists as the stable lattice, an alternative potentials of electrolyte components in lithium ion
mechanism was also proposed by the same authors cells. They used manganese dioxide (MnO2), LiNiO2,
to account for the surface oxidation of electrolytes in LiCoO2, and spinel LiMn2O4 as working electrodes,
this case (Scheme 18). Since similar surface layers and they investigated the potential of cathode surface
have been observed on spinel cathodes in partially layer breakdown, which was found to vary in a
and fully charged states, their thickness and chemical pronounced way with each cathode surface: 4.2 V for
composition do not seem to be dependent on the LiClO4/PC on MnO2,311 4.2 V for LiIm/PC on LiN-
cathode potential, while thickness does increase with iO2,312 and 4.8 V for LiIm/PC on both LiCoO2 and
both storage time and temperature. LiMn2O4.312 By means of an in situ MS technique,
these authors confirmed that the breakdown process
6.3.3. Breakdown of Surface Layer was accompanied by the evolution of gas that con-
While the low potential limit an electrolyte faces sisted mainly of CO2 and other species related to salt
in lithium ion cells is set by the lithium metal anions, an indication of bulk decomposition of elec-
deposition on carbonaceous anodes, the high limit is trolyte components that involve both salt and sol-
decided by the breakdown of the surface layer on vent.311,312
cathode materials. At potentials above this limit, the A more conspicuous example is the anodic stability
protection against electron tunneling is lost and the of linear carbonate (Figure 23).76 While the anodic
decomposition of bulk electrolyte components occurs. limits for both DMC and EMC are above 5.0 V on
Because of the limited amount of research carried out GC, as observed by similar measurements,313 the
on cathode surfaces, there is no microscopic picture major decomposition current starts before 4.5 V for
available for this breakdown mechanism and it both solvents when a composite cathode based on
remains unknown whether a global dissolution or spinel LiMn2O4 is used as working electrode.76 Ap-
physical disintegration of the surface layer is involved parently, the enormous difference between GC and
or, rather, following a similar mechanism of dielectric a composite cathode could not be explained solely by
capacitor “breakdown”, the surface layer remains the larger surface area of the latter. It seems more
largely intact but the high potential of the cathode likely that the irreversible oxidative processes were
“burns through” pathways for electron conduction at catalyzed by the active mass of the spinel surface,
local sites on the surface. possibly involving spinel lattice disproportionation.
The knowledge about this decomposition upper Therefore, the stability limits as measured on all of
limit of potential is significant for practical applica- these nonactive and nonporous electrodes can be
Figure 24. Voltammogram at 55 °C for the electrolytes

(a) 1.0 M LiClO4/EC/DEE and (b) 1.0 M LiPF6/EC/DMC
on spinel cathode LiMn2O4. (Reproduced with permission
from ref 98 (Figure 1). Copyright 1993 The Electrochemical
Society.)
Figure 23. Anodic stability of linear dialkyl carbonates
on GC and a composite cathode surface. Electrolytes: 1.0 sponds to the delithiation of the spinel structure,
LiPF6 in DMC or EMC, respectively. (Reproduced with
permission from ref 76 (Figure 2). Copyright 1999 The which is reversible. The oxidative decomposition of
Electrochemical Society.) DEE occurs below 4.4 V almost immediately after the
delithiation is completed. This proximity to the
Table 7. Electrochemical Stability of Electrolyte potential of the cathode electrochemistry leaves little
Solvents: Active Electrodes tolerance for overcharge. On the other hand, the EC/
working DMC-based electrolyte only shows negligible back-
electrolyte electrode Eaa ref ground current before 5.1 V. After 4.3 V, the spinel
1.0 M LiClO4/PC/DME (1:1) LiV3O8 4.6 307 lattice has actually turned into λ-MnO2. On the
1.0 M LiAsF6/PC/DME (1:1) LiV3O8 4.7 307 surface of the λ-MnO2, the low current of the EC/
1.0 M LiClO4/EC/DEE (1:1) LiMn2O4 4.55 93 DMC exhibits the extreme stability of this mixed
1.0 M LiIm/EC/DEE (1:1) LiMn2O4 4.4 99
1.0 M LiIm/EC/DME (1:1) LiMn2O4 4.35 99
carbonate formulation. As already pointed out in
1.0 M LiPF6/EC/DEE (1:1) LiMn2O4 3.8 93 earlier sections, the replacement of ether compounds
1.0 M LiBF4/EC/DEE (1:1) LiMn2O4 3.4 99 by linear carbonate results in a significant expansion
1.0 M LiClO4/PC LiMn2O4 >5.1 99 of the oxidative potential limits by ∼0.80 V.
1.0 M LiBF4/EC/DMC (2:1) LiMn2O4 >5.1 98 The anodic limit for the electrochemical stability
1.0 M LiClO4/EC/DMC (2:1) LiMn2O4 >5.1 99
1.0 M LiPF6/EC/DMC (2:1) LiMn2O4 >5.1 44, 98
of these carbonate mixtures has been determined to
1.0 M LiClO4/EMS LiMn2O4 5.8 75, 314 be around 5.5 V in numerous studies.314-317 Thus,
1.0 M LiIm/EMS LiMn2O4 5.8 75, 314 new electrolyte formulations are needed for any
1.0 M LiPF6/EMS LiMn2O4 5.8 314 applications requiring >5.0 V potentials. For most
1.0 M LiPF6/EiBS LiMn2O4 5.8 314 of the state-of-the-art cathode materials based on the
1.0 M LiPF6/DMC LiMn2O4 ∼4.0 314 oxides of Ni, Mn, and Co, however, these carbonate
1.0 M LiPF6/EMC LiMn2O4 4.5 314
1.0 M LiPF6/FPMS/DMC LiMn2O4 5.55 314 mixtures can provide a sufficiently wide electrochemi-
1.0 M LiPF6/FPMS/EMC LiMn2O4 5.55 314 cal stability window such that the reversible lithium
1.0 M LiPF6/EC/PC/DMC (1:1:2) LiNiVO4 >4.9 314 ion chemistry with an upper potential limit of 4.30
a
Anodic limits, potential referred to Li+/Li.
V is practical.
When the poor anodic stability of DMC or EMC
alone on a similar cathode surface is considered,76 the
misleading, and the authentic anodic stability limits role of EC in stabilizing the solvent system becomes
of any electrolyte component in an actual lithium ion obvious. A conclusion that could be extracted from
device can only be evaluated on the surface of the these studies is that the existence of EC not only
electrode that is used in the device. renders the electrolyte system with superior cathodic
Unfortunately, this approach for electrochemical stability by forming an effective SEI on the carbon-
stability determination has not been widely adopted. aceous anode but also acts as a key component in
The few exceptions include the seminal electrolyte forming a surface layer on the cathode surface that
work by Guyomard and Tarascon.98,99 In the formula- is of high breakdown potential. It is for its unique
tion of new electrolytes for lithium ion technology, abilities at both electrodes that EC has become an
the spinel composite electrode was used as the indispensable cosolvent for the electrolyte used in
standard working surface in all of the voltammetric lithium ion cells.
measurements. The oxidative decomposition limits
of the new electrolytes thus determined are sum- 6.3.4. Passivation of Current Collector
marized in Table 7 along with a handful of stability Another issue closely related with the anodic
data that were determined in a similar approach for stability of electrolytes is the interaction between
other electrolyte systems. electrolyte components and the commonly used cath-
Figure 24 shows the result of the voltammetry for ode substrate Al in lithium ion cells.
EC/DMC- and EC/DEE-based electrolytes. The typi- Al and Cu foils are used as current collectors for
cal two-step process between 3.9 and 4.3 V corre- cathode and anode materials, respectively, in lithium
ion cells. In a working lithium ion cell that uses liquid

electrolytes, the thin coating of active electrode
materials is soaked with the liquid electrolytes and
the interaction between the electrolyte components
and the current collectors strongly affects the per-
formance stability of the cell, especially at the
cathode side, where Al is constantly held at high
potentials. It has been confirmed by numerous ob-
servations that, in well-cycled lithium ion cells, the
Al substrate suffers severe pitting-like corrosion at
a fully charged state, which leads to a shortened
calendar life and fading capacity.298,318,319 Because
very thin Al foils are used (down to 10 µm) in
consideration of the energy density, the pitting could
cause complete disintegration and fragmentation in
some cases. Figure 25. Al corrosion current density at 4.20 V (vs Li+/
Few significant efforts were made on this issue Li) in 1.0 M PC solutions of various lithium salts showing
a difference between Tf-, Im-, and other anions. (Repro-
before a new lithium salt (LiIm) was found to cause duced with permission from ref 141 (Figure 1). Copyright
serious Al corrosion in nonaqueous electrolytes dur- 1997 Elsevier.)
ing the early 1990s.141 Only in recent years has an
in-depth understanding of this phenomenon been usually caused severe corrosion on the Al surface, as
obtained, a direct result from the increased research indicated by the high anodic current at potentials as
interest driven by the expectation that this new salt low as 3.0 V.141,147,155,156 Figure 25 compares the
may replace the thermally unstable LiPF6.153,154,320-323 ability of selected lithium salts to stabilize the Al
Al is thermodynamically unstable, with an oxida- surface in the electrolytes containing them. The
tion potential at 1.39 V.3 Its stability in various passivating ability of the salt anion does not appear
applications comes from the formation of a native to depend on the solvent used, and in one solvent
passivation film, which is composed of Al2O3 or mixture, EC/DME, the following stability order is
oxyhydroxide and hydroxide.320 This protective layer, established:156
with a thickness of ∼50 nm,322 not only stabilizes Al
in various nonaqueous electrolytes at high potentials LiBF4 > LiPF6 > LiClO4 > LiIm > LiTf
but also renders the Al surface coating-friendly by
enabling excellent adhesion of the electrode materi- New synthetic efforts generated a few structural
als. It has been reported that with the native film derivatives of LiIm with longer perfluorinated alkyls
intact Al could maintain anodic stability up to 5.0 V such as LiBeti and a lithium salt of an asymmetric
even in LiIm-based electrolytes.147 Similar stability imide anion, and the stability of Al in the electrolytes
has also been observed with Al pretreated at 480 °C based on these salts was found to be much improved
in air, which remains corrosion-free in LiClO4/EC/ when compared with those in electrolytes based on
DME up to 4.2 V.156 However, since mechanical LiIm and LiTf.141,148
damage of the native film (e.g., accidental abrasion)
is often inevitable during electrode processing and
cell assembly, the in situ re-forming of a protective
surface layer in lithium ion cells depends strongly
on the electrolytes.
In comparison with the surface layer chemistry on
active cathode materials where both salt anions and
solvents are involved, a general perception extracted Krause et al. thus believed that a correlation existed
from various studies is that the salt species has the between anion molecular weight and their protective
determining influence on the stabilization of the Al abilities, although the complete mechanism is not
substrate while the role of solvents does not seem to clear.141
be pronounced, although individual reports have A distinctive dissimilarity between the passivation
mentioned that EC/DMC seems to be more corrosive of Al and the formation of an SEI on a graphitic
than PC/DEC.321 Considering the fact that pitting anode is the effectiveness of surface pre-formation:
corrosion occurs on Al in the polymer electrolytes while it has been reported in numerous studies that
LiIm/PEO322 or LiTf/PEO,323 where the reactivity of an SEI formed in one electrolyte can continue to
these macromolecular solvents is negligible at the protect the graphite structure when the anode is
potentials where the pitting appears, the salt appears transferred to another exfoliating electrolyte (such
to play the dominant role in Al corrosion. as PC),250,255,324 Al that has been pretreated in LiPF6-
Among the various lithium salts investigated, or LiBF4-based electrolytes still suffers severe anodic
LiPF6, LiAsF6, and LiBF4 were found to effectively corrosion when switched to LiTf- or LiIm-based
passivate and stabilize Al at higher potentials (>4.2 electrolytes.141,156 This observation provides addi-
V) than required by the operation range of common tional evidence supporting the hypothesis that an
cathode materials based on LiNiO2, LiCoO2, and oxidative decomposition of anions occurs at the Al
LiMn2O4, while LiIm, LiTf, and their derivatives surface.141
Figure 26. Micrographs of a corroded Al surface in various electrolytes. (a, top left) Micrograph in LiPF6/EC/DMC
electrolyte, after 150 electrical cycles. The light areas are mounds or nodules. (Reproduced with permission from ref 321
(Figure 7). Copyright 1999 The Electrochemical Society.) (b, bottom left) SEI image of the cross-sectional view of the mounds
as shown in part a. (Reproduced with permission from ref 321 (Figure 8a). Copyright 1999 The Electrochemical Society.)
(c, right) Micrograph in LiIm/PEO after 1 h of galvanostatic polarization at 100 µA cm-2. (Reproduced with permission
from ref 322 (Figure 14). Copyright 1999 The Electrochemical Society.)
Microscopic examination of the aged Al retrieved Scheme 19. Al Corrosion in Nonaqueous

from well-cycled cells found that, even in electrolytes Electrolytes
based on LiPF6 or LiBF4, localized pitting occurred
on the Al (Figure 26a).298,318,319 Closer examination
of these corrosion sites under an SEM revealed that
what had optically appeared to be a “pit” was actually
a “mound” or “nodule” (Figure 26b) where both Al(0)
and Al2O3 existed, according to XPS analysis.321 Due of prolonged cycling in an actual rechargeable cell.
to the anodic state of the Al during the electric The damage to Al in these cases was on a macroscopic
cyclings in these cells, the presence of Al(0) practi- scale.
cally suggests that the mounds were somehow elec- Impedance analyses of the Al under corrosion were
trically isolated from the foil. The authors presented conducted via EIS. On the basis of the models
two possible explanations for this situation: (1) previously established for the corrosion of other
corrosion and its associated reaction products under- metals in both aqueous and nonaqueous electro-
mined the surface of a developed pit and caused the lytes,325,326 the corrosion process was proposed as a
overlying metal to bulge or break away or (2) a two-step adsorption/oxidation/desorption process
soluble Al(III) corrosion product was electrodeposited (Scheme 19).147,156
onto the poorly conductive solid corrosion products The impedance response with frequency can be
during the cathodic process (discharge) so that reduc- closely simulated by the equivalent circuit shown in
tion actually occurred. The fact that the open circuit Figure 27a, where Re, Rct, Cdl, Rad, and Cad represent
voltage (OCV) of Al lies between 3.2 and 3.6 V makes the resistance or capacitance for the electrolyte
the latter reductive step possible in normal battery solution, charge-transfer, double layer, and adsorbed
cyclings. layer, respectively. An interesting correlation was
On the other hand, when the electrolyte salts were found between the passivating ability of various
LiTf or LiIm, serious corrosions in the form of high- anions and the resistances of the two impedance
density pits were visible before any extensive cycling components Rct and Rad, which are high for LiPF6-
was conducted,141,322 as shown by Figure 26c, where and LiBF4-based electrolytes and low for LiTf- or
Al was merely polarized anodically in a LiIm-based LiIm-based electrolytes.156 Using the rationale pro-
electrolyte for a much smaller time scale than that posed by the authors, the former component (Rct) is
occurs, probably by an Al+-mediation mechanism.

According to this mechanism, upon polarization of Al
at high potentials, highly reductive intermediates of
Al+ were generated,328,329 and they migrated to the
vicinity of the passivation layer/electrolyte interface,
where they subsequently reacted with adsorbed
anions or solvents. Thus, the reaction kinetics depend
on how readily the anions or solvents are adsorbed
on the Al surface or surface layer.
The depth profile established by Ar+ sputtering
detects the distinct difference between the Al surfaces
that are aged in various electrolytes, which suggests
that a much thicker surface layer is formed in LiIm
than in LiPF6 or LiBeti electrolytes.141 It appears that
the high corrosion of Al in a LiIm-based electrolyte
stems from a relatively porous and diffuse passiva-
tion layer where the high levels of LiF contribute to
the disorder and porosity of the otherwise dense and
protective Al2O3 or carbonate layer. The latter is
found in Al surfaces passivated in LiBeti or LiPF6.141
Combining the above observations, Krause et al.
explained the correlation between anion size and the
Figure 27. (a) Equivalent circuit used to express the
corrosion process of Al. (b) Schematic representation of a
passivating ability by referring to the ability to
possible mechanism for Al corrosion in LiIm-based elec- adsorb on the Al surface. Thus, small anions such as
trolytes. (Reproduced with permission from ref 147 (Figure Tf- and Im- tend to adsorb to a high packing density,
14). Copyright 2000 The Electrochemical Society.) and their reduction by Al+ predominates the surface
process, with the product of LiF rendering the film
related to the exchange current (i0) at the electrolyte/ nonprotective. Beti- anions adsorb poorly, and there-
Al surface: fore, the reduction of solvent molecules prevails, with
RT the surface layer consisting of a dense protective
Rct ) (11) carbonate species. It must be pointed out that this
nFi0 mechanism remains to be confirmed with more direct
evidence of the presence of Al+ because the reductive
while Rad indicates whether an adsorbed (passivation) process of solvents at a positively charged cathode
layer could adhere to the surface of Al. Therefore, the does sound very unusual.
physicochemical nature of the passivation layer is
dictated by its chemical composition, which is in turn On the other hand, an XPS study by Braithwaite
strongly affected by the anion oxidation chemistry. et al. on aged Al in LiPF6/EC/DMC or LiPF6/PC/DEC
Surface chemical characterization of the passiva- identified Li, F, and P as the major surface species,
tion layer on the Al surface has been performed and the authors believed that the signals for these
mainly via XPS, and the interpretation of results species were not simply from the adsorbed electro-
generated by various researchers still remains con- lytes because twice as much F was observed on the
troversial. Because salt anions with active fluorine surface than would be associated with the PF6-
(LiPF6, LiAsF6, and LiBF4) are able to form stable anion.321 The existence of these nonstoichiometric
surface layers on Al and protect it from corrosion, compositions suggested that the oxidative decomposi-
early studies had suggested that fluoride species such tion of the salt anion had occurred. The high binding
as LiF and AlF3 are crucial to the protection.141,153,154 energy of Li as observed in Li 1s spectra indicates
The XPS study by Kanamura et al. seemed to support that such Li species exist in a very electron-deficient
this speculation, having observed the F 1s signal at environment, as one similar to LiF. However, the
686.5 eV on Al that had been passivated in LiPF6/ observed binding energy for F is typical for the
PC electrolyte.148,327 Thus, they concluded that the -CHF2 species and distinctively different from that
surface species on Al after prolonged exposure to the of a metal fluoride. Further attempts to analyze these
above electrolyte consisted of a mixture of Al2O3 and pit nodules by XPS analysis proved to be inconclusive
AlF3. because of the high inhomogeneity of these areas,
Krause et al. challenged this assignment of the F which exhibited complex chemistries that varied from
1s signal.141 In their XPS study of Al cycled in nodule to nodule. In general, high concentrations of
electrolytes based on LiIm, LiBeti, and LiPF6, they Li and F as well as Al2O3 were confirmed. On the
also observed this F signal that is typical of metal basis of these seemingly conflicting results, the
fluoride, but they believed that LiF instead of AlF3 authors concluded that the role of these surface
would be responsible for it because of the detection species formed in nonaqueous electrolytes could not
of Li and the relative abundance of these two ele- be determined, and the protection against corrosion
ments. Carbonate species as indicated by both C 1s could simply have come from the native passivation
and O 1s spectra were also present, which may exist film Al2O3.
in the forms of either Al2(CO3)3 or Li2CO3. Thus, the More recently, Kanamura et al. studied the corro-
authors postulated that oxidation of solvent also sion of an Al electrode that was anodically polarized
at a series of potentials in nonaqueous electrolytes (Im)3 (∆m/∆Q ) 280.2) or a reaction intermediate of

based on LiIm, LiTf, and a new asymmetric imide, Al+ or of a solution species, 141 but these compounds
Li(CF3SO2)(C4F9SO2)N, using both XPS and FT-IR.148 rapidly desorb from the Al surface, reducing the
These spectra confirmed the presence of organic numerator in ∆m/∆Q.
species on the Al surface, and by monitoring the IR On the basis of the EQCM observations, the
signature absorption for the carbonate functionality authors proposed an adsorption/oxidation/desorption
at 1710 cm-1, they determined that these species mechanism for the severe pitting corrosion of Al in
appeared at ∼3.6 V in LiIm- and LiTf-based electro- LiIm- and LiTf-based electrolytes, which is schemati-
lytes; however, when using LiPF6, LiBF4, and LiClO4, cally shown in Scheme 19 and Figure 27b.147 Accord-
this absorption was not found till 5.2, 4.4, and 4.6 ing to this mechanism, Al oxidizes to form adsorbed
V,327 respectively. Hence, they argued that, while Al Al(Im)3 that eventually desorbs from the surface
was well passivated by the native Li2CO3 film, such because these species are soluble in the electrolyte
an oxidation by electrochemical means would not solvents. It is the desorption of these oxidized prod-
occur. In other words, the difference in the effect of ucts that leaves the otherwise smooth Al surface with
the electrolyte components on Al stability is not in pits. The possibility also exists that, before desorption
how they decompose to form a new protection layers occurs, the adsorbed species undergoes further oxida-
such as through the familiar mechanism of SEI tion; however, since the oxidation of Im- is insignifi-
formationsbut in how inert they are to the existing cant below 4.5 V according to studies carried out on
native layer. This inertness depends on the cathode nonactive electrodes similar to Al,81,130,131,206 it seems
potential, and the occurrence of solvent oxidation is unlikely that further oxidation of the adsorbed Al-
indicative of the breakdown of the native passivation (Im)3 would occur.
film. Apparently, Tf- and Im- anion species are On the other hand, the oxidized products in LiPF4
corrosive to the native film, so the breakdown poten- and LiBF4 electrolytes could consist of less soluble
tial of it is much lower in the electrolytes based on species such as AlF3 and Al2(CO3)3, as suggested by
these salts. Therefore, it would be reasonable to ∆m/∆Q ) 26-27, indicating negligible desorption
believe that a reaction occurring between Al2O3 and from the Al surface and therefore less corrosion of
electrolyte solutes at high potentials is responsible the Al substrate.
for the onset of major corrosion, during which the
coordination ability of Im- and Tf- anions toward 6.4. A Few Words on Surface Characterizations
Al3+ acts as the driving force that accelerates the
anodic dissolution of Al substrate. Beyond any doubt, the electrode/electrolyte inter-
faces constitute the foundations for the state-of-the-
Large anion size was also found to favor Al stabi- art lithium ion chemistry and naturally have become
lization, since the 1710 cm-1 signature did not appear the most active research topic during the past decade.
till a potential higher than 5.0 V was reached for the However, the characterization of the key attributes
new, asymmetric imide. Unlike the interpretation of of the corresponding surface chemistries proved
Krause et al.,141 Kanamura et al. attributed its rather difficult, and significant controversy has been
inertness to the well-distributed negative charge and generated. The elusive nature of these interfaces is
the lower probability of these anions forming an ion believed to arise from the sensitivity of the major
pair with Al3+.148 chemical compounds that originated from the decom-
The only in situ and semiquantitative study on Al position of electrolyte components.
corrosion was carried out by Yang et al. using EQCM Hence, the presence of trace impurities, which
to monitor the mass change of an Al electrode in either pre-exist in pristine electrode and bulk elec-
various electrolytes based on LiIm, LiTf, LiPF6, trolyte or are introduced during the handling of the
LiBF4, and LiClO4.147 The most interesting discovery sample, could profoundly affect the spectroscopic
was that, during the process of corrosion in all of the images obtained after or during certain electrochemi-
electrolytes, the mass change of Al was positive (i.e., cal experiments. This complication due to the impu-
the Al electrode gained net weight) instead of losing rities is especially serious when ex situ analytic
weight, as one would expect from a simple anodic means were employed, with moisture as the main
dissolution of Al. perpetrator. For cathode/electrolyte interfaces, an
Furthermore, normalization of the weight gain additional complication comes from the structural
against the corresponding charge quantity offered a degradation of the active mass, especially when over-
semiquantitative method to identify the possible delithiation occurs, wherein the decomposition of
species formed, and in the case of LiBF4- and LiPF6- electrolyte components is so closely entangled with
based electrolytes, ∆m/∆Q ) 26-27 was obtained, the phase transition of the active mass that dif-
suggesting that essentially identical species were ferentiation is impossible. In such cases, caution
adsorbed on Al. Compounds with similar ∆m/∆Q should always be exercised when interpreting the
values include AlF3 (∆m/∆Q ) 19) and Al2(CO3)3 (∆m/ conclusions presented.
∆Q ) 30). On the other hand, LiTf- and LiIm-based On the other hand, when the chemical composition
electrolytes are characterized by very high ∆m/∆Q of the surface layer is discussed, distinction should
values. For example, in chronoamperometry stepped be made regarding the conditions under which such
from 2.0 to 4.5 V, an Al electrode in LiIm/PC is characterizations were carried out and the history
accompanied by a ∆m/∆Q ) 174. It appears that the of the electrode surface. For example, an electrode
initial oxidation could be due to the formation of an surface that was subjected to long-term cycling is
Al compound of large molecular weight such as Al- certainly different from the surface state of the same
electrode that only experienced the initial forming Scheme 20. Bulk Reaction between Electrolyte
process. Unfortunately, this distinction was seldom Components at Elevated Temperatures
made in the literature. Most work also failed to
characterize the pre-existing film on the electrode
surface prior to electrochemical studies, and this
neglect could lead to significant misinterpretation of
the surface chemistry, because organic species could
pre-exist on cathode as well as anode surfaces as the
result of various contaminations or processing. In this
sense, to obtain insightful understanding about the
interfacial chemistries in a lithium ion cell, close peratures are also often used as an accelerating tool
collaboration is still needed between electrochemcal to study the otherwise slow reactions at ambient
and spectroscopic characterizations. conditions.
The intrinsic instability of the state-of-the-art
7. Chemical and Thermal Stability/Safety of electrolyte systems seems to arise mainly from the
interaction between the cyclic carbonate and the
Electrolytes Lewis acids, HF, PF5, or POxFy,150,330 which are
generated by the decomposition or the hydrolysis of
7.1. Long-Term Stability of Electrolytes at LiPF6 by trace moisture. While typical electrolyte
Elevated Temperatures compositions such as LiPF6/EC/DMC could seem to
In most cases, the electrochemical stability of be indefinitely stable at room temperature in the
electrolyte components was studied at room temper- proper containers,331 Kinoshita and co-workers found
ature with various galvanostatic or voltammetric that significant decomposition occurred during stor-
experimental techniques that were conducted within age under He atmosphere at 85 °C, as indicated by
a certain time scale, usually <105 s. The stability discoloration, gas production, and solid precipita-
thus determined should be considered relatively tion.152 They estimated that at 85 °C the reactions
“transient” where the application in a longer time were accelerated 10-fold as compared with the case
frame comparable to the cell lifetime is concerned, of room temperature. Chemical analysis conducted
since the effect of those irreversible reactions of slow on the stored electrolyte solution with a gas chro-
kinetics at room temperature, occurring either be- matograph detected the formation of a series of new
tween the bulk electrolyte components (chemical) or species, among which the major component had a
between the electrodes and the electrolytes (electro- higher retention time (thus lower volatility) than EC.
chemical), cannot be factored in by such time-limited Through comparison with an authentic sample, the
measurements. Only when stability of electrolytes is authors identified this species as dimethyl-2,5-dioxa-
considered in a longer time frame (t ∼ 106 s or up to hexane carboxylate (DMDOHC), which was obviously
months) can the impact of these reactions be visible. formed from the ring-opening of EC and its subse-
As a consequence of the slow irreversible processes, quent transesterification with DMC (Scheme 20).
the performances of the lithium cells are often Electrolytes based on an EC/DEC mixture rendered
adversely affected, as typically evidenced by the the corresponding product, DEDOHC, as shown in
persistent fade in capacity, loss in power rate, and Scheme 20.48a,298 The effect of such a transesterifica-
increasing internal pressure within the cell, which tion species on the performance of electrolytes in
can be attributed respectively to the irreversible lithium ion cells has not been described.
consumption of the limited lithium ion sources, the Other species with higher retention times were
depositions of high-impedance products on the elec- suggested to be the products formed through the
trode surface, and the generation of gaseous products coupling of two ring-opened EC or oligoether com-
by slow reactions.152,277,297,309 pounds. On the other hand, the 1H NMR conducted
On the other hand, since most of these reactions on the reacted electrolyte showed broad singlet bands
are thermally activated, their kinetics are accelerated of equal intensity at 3.746 and 4.322 ppm vs TMS,
by the rise in temperature in an Arrhenius-like which resembled those reported for the poly(ether
manner. Therefore, within a much shorter time scale, carbonate) as polymerization products of EC.332,333
the adverse effect of these reactions could become The authors concluded that the polymerization and
rather significant during the storage or operation of the transesterification were possibly initiated by PF5.
the cells at elevated temperatures. In this sense, the To confirm this hypothesis, they introduced newly
long-term and the thermal stability of electrolytes can generated PF5 gas into the solvent mixture EC/DMC.
actually be considered as two independent issues that After 10 h at room temperature, similar reaction
are closely intertwined. The study of temperature patterns were observed by chromatography, with the
effects on electrolyte stability is made necessary by above mixed carbonate as the major decomposition
the concerns over the aging of electrolytes in lithium- product, although its abundance was much lower.
based devices, which in practical applications are Minor differences unaccounted for between the two
expected to tolerate certain high-temperature envi- chromatographs were attributed to the presence of
ronments. The ability of an electrolyte to remain Li+ as well as the high ambient temperature in the
operative at elevated temperatures is especially thermal reaction.
important for applications that are military/space- By monitoring the change in the ratio of EC/DMC
related or traction-related (e.g., electric or hybrid over time, Kinoshita and co-workers also noticed that
electric vehicles). On the other hand, elevated tem- EC decreased at a more rapid rate than DMC,
Scheme 21. Ester Exchange between Electrolyte Scheme 22. Ester Exchange Reaction Mechanism
Solvents
probably because PF5 preferentially attacks the cyclic

structure of EC. This hypothesis is consistent with
the well-studied ring-opening polymerization of EC
that is usually initiated by cationic species, as shown generation of alkoxide anions is followed by the
in Schemes 4, 12, and 15. The polymeric product addition of them to the carbonyls of linear carbonates
could be a mixture of oligoether and polycarbonate (Scheme 22). In other words, the ester-exchange
structures.332,333 reaction should start concurrently with the formation
Besides salt/solvent interactions, the ester ex- of the SEI, as shown in Scheme 7. According to the
change between linear carbonates was also observed above mechanism, the alkoxide anions, once gener-
during the cycling of the cells. Terasaki et al. were ated in the first stage, would continue to be produced
the first to observe that a significant amount of EMC in a chain reaction manner, and the above ester
was formed in a solvent mixture of DMC and DEC exchange would continue through the cell cyclings
and proposed a reductively initiated mechanism.334 and be independent of the SEI formation. However,
Interestingly, a corresponding oxidatively initiated the authors found that the extent of the ester-
mechanism was suggested by Ohta et al., when they exchange reaction was dependent on the anode
observed that EMC, ethyl propionate, and DMC were surface passivation, and the complete formation of
generated from the original solvent mixture of DEC the SEI eventually prevented it. Therefore, a termi-
and methyl propionate (Scheme 21).335 In both cases, nation process had to exist at some stage, and the
EC was used as a cosolvent and LiPF6 as an electro- authors suggested that a cathode surface in a charged
lyte salt, but no identification of any products that state would serve as such a terminator by quenching
involved the participation of these two components (oxidizing) the alkoxide anions. In view of the acidic
was reported, probably because of the slow kinetics nature of the electrolytes based on LiPF6, neutraliza-
at room temperature or the insensitivity of the tion of alkoxide anions by trace acid is another
analytical method used to detect the new products. possible termination means.
Takeuchi et al. thoroughly investigated the ester- The absence of EC in the discussions of all these
exchange reaction between DMC and DEC in the reports334-336 is interesting, since one would suspect
presence of PC and LiPF6. By exposing the solvent that, based on the knowledge about EC reactivity,
mixture to individual cell components, including alkoxide could also attack its cyclic structure and
charged electrode materials at different potentials, cause irreversible reactions.152,285,332,333 One possibility
salts, and acid impurities (HF), and monitoring the would be that, using gas chromatography, the iden-
generated EMC content by means of a gas chromato- tification of a high-volatility species such as linear
graph, they conclusively proved that the ester- carbonate is easy and reliable, while the formation
exchange process was reductively initiated by the of the polymeric species, which EC would most likely
lithiated carbon anode with potentials below 1.50 V decompose into, is more difficult to detect.
and essentially negated the oxidative mechanism Endo et al. studied the gradual degradation of
proposed by Ohta et al.336 Since DMC and DEC could electrolyte solutions in lithium or lithium ion cells
also be detected from the original PC/EMC mixture by monitoring the production of active radical species
in the presence of the lithiated carbon, the exchange with ESR.337 To elongate the lifetime of the radicals
process should be reversible. so that their hyperfine structure could be analyzed,
In a further effort to identify the active intermedi- a spin-trapping technique was adopted to stabilize
ate that initiates the reaction, they tested the effect the original radicals generated from the reductive
of a few possible ingredients on the production of decompositions of electrolyte solvents on the anode
EMC based on the knowledge about the chemical surface. In this way, the authors were able to gain
composition of the SEI on carbonaceous anodes. an insight into the structure of the radical species.
These model compounds included Li2CO3, LiOCH3, The authors found that, even in the presence of a
and LiOH, while lithium alkyl carbonate was not good, protective SEI, generation of active alkyl radi-
tested due to its instability and therefore rare avail- cals, rather than radical anions, continued in the
ability.337 The results unequivocally showed that electrolyte and that a chain reaction initiated by
LiOCH3 effectively catalyzes the ester exchange. cathodic polarization of the electrode was very likely
A two-stage mechanism was proposed by Takeuchi responsible. In other words, once the radicals were
et al. based on these observations, wherein the produced in the initial cathodic polarization, the
they concluded that the components of the SEI layer

are soluble in DMC.338 Because of the similar
physicochemical properties between various linear
carbonate solvents, this dissolution of the SEI might
very likely be universal in all state-of-the-art elec-
trolytes.
This corrosion of the SEI by linear carbonate
solvents would undoubtedly produce adverse effects
on the performance of lithium ion cells. During long-
term cycling, the damaged SEI has to be repaired
constantly by the same electrochemical reactions that
occurred in the initial formation process, which
consumes the limited lithium ion source in the cell
and increases the impedance at the electrode/
electrolyte interface.
Figure 28. DSC trace of the reactions occurring between A more detailed study on the thermal stability of
a fully lithiated graphitic anode and electrolyte. Anode the SEI on lithiated and delithiated graphitic anodes
surfaces both rinsed with DMC and unrinsed were studied. was carried out by Amatucci and co-workers.339 By
(Reproduced with permission from ref 338 (Figure 1). storing the cells at 70 °C for 4 days and then
Copyright 1998 The Electrochemical Society.)
resuming cycling at room temperature, they pre-
sented direct evidence that the SEI was damaged at
gradual decomposition of electrolyte solvents could
the elevated storage temperature by successfully
occur in the bulk of the solution without involving
detecting the reappearance of the irreversible process
electrodes. This propagation of radicals inevitably led
at 0.75 V, which has been previously determined to
to the polymerization of solvents. Using liquid chro-
be the formation of the SEI by reductive decomposi-
matography, which is more sensitive to the detection
of polymeric species, reductively induced polymeri- tion of the electrolyte. Furthermore, a correlation was
zation of ester-based solvents was confirmed. also established between the capacity loss after such
In the gradual degradation mechanism proposed storages and the rebuilding of the SEI, with the latter
by those authors, radical anions are first generated being proportional to such factors as the storage
during SEI formation through a single-electron pro- temperature, the storage time, and the surface area
cess, as proposed by Aurbach et al.,108,124,249 and then of the anode. An Arrhenius behavior was actually
polymerization proceeded gradually from these radi- observed for the dependence of capacity loss on the
cal anions with subsequent and continuous produc- storage temperature, with an activation energy of
tion of more active alkyl radicals. In this sense, even 39.8 kJ mol-1.339
a perfect SEI that insulates any electron transfer As expected, the state-of-charge (SOC) of the anode
through it cannot prevent the continuous degradation also influences this rebuilding process, since the
of electrolyte solvents. capacity loss due to the storage is much higher for
the fully lithiated carbon anode than for the fully
7.2. Stability of the SEI or Surface Layer at delithiated one. This fact was explained by the
Elevated Temperatures authors in terms of the reactivity of the carbon
surfaces, which are partially exposed because of SEI
The presence of a protective SEI or surface layer
corrosion, toward electrolyte components. Thus, in-
prevents those irreversible reactions of electrolytes
tercalated lithium ions continuously diffuse from the
on anode/cathode surfaces that are otherwise favored
interior of the graphitic structure through the im-
by thermodynamics. Like the chemical process in the
perfect SEI coverage and participate in the reaction
bulk electrolyte, the reactivity of the surface films
with electrolyte solvents to re-form the SEI. Under
formed in state-of-the-art electrolytes is negligible at
stationary storage, an equilibrium would be reached
room temperature. However, during long-term stor-
age and cycling, their stability is still under question. between the SEI dissolution rate and the lithium ion
When conducting a differential scanning calorim- diffusion rate. However, the net effect would be the
etry (DSC) study on the stability of carbonaceous irreversible consumption of electrolyte solvents as
anodes in electrolytes, Tarascon and co-workers well as lithium ions.
found that, before the major reaction between lithi- The long-term deterioration of the SEI at elevated
ated carbon and fluorinated polymers in the cell, temperatures also was observed by McLarnon and
there was a transition of smaller thermal effect at co-workers, who conducted postmortem analyses on
120 °C, marked peak (a) in Figure 28.338 They the electrodes from consumer lithium ion cells that
ascribed this process to the decomposition of SEI into were stored at various temperatures between 40 and
Li2CO3, based on the previous understanding about 70 °C for weeks.298 Using the IR absorption at 838
the SEI chemical composition and the thermal stabil- cm-1 as the indicator of the SEI component, they
ity of lithium alkyl carbonates.102,117,249,281 Interest- quantitatively correlated the stability of the SEI on
ingly, those authors noticed that the above transition a carbonaceous anode against storage temperature
would disappear if the carbonaceous anode was and SOC. In an extreme situation where the cell was
rinsed in DMC before DSC was performed, while the stored at 70 °C with 9% swing of SOC, no remaining
other major processes remained (Figure 28). Thus, SEI was found on the anode surface.
Figure 29. C 1s and O 1s XPS spectra for fresh, precycled, and stored graphite anodes. (Reproduced with permission
from ref 294 (Figure 17). Copyright 2002 The Electrochemical Society.)
Scheme 23. Thermal Decomposition of Lithium indicated by the decreasing C 1s signal at 291 eV
Alkyl Carbonate in the SEI (corresponding to ROCO2Li), but the new species did
not seem to be Li2CO3, since the C 1s signal appeared
at 290-290.5 eV instead of at 292 eV. These new
species, still of a carbonyl nature, were thought by
the authors to be polycarbonate formed from the ring-
Tarascon and co-workers proposed that the SEI on opening polymerization of EC. The discrepancy be-
a carbonaceous anode is subject not only to simple tween the works of Amine et al. and Edström et al.
dissolution but also to decomposition at elevated could have arisen from the difference in the storage
temperatures (>120 °C) into Li2CO3.338 While the duration, which was much longer in the former case.
existence of some chemical process was generally One common conclusion shared by both groups of
agreed upon by various authors, they differed vastly researchers was the dramatic increase in LiF content
on what the original SEI components had turned into as a result of the storage, which became the major
through the process.277,278,284,340 species on the anode surface.277,294 Because of the
In agreement with Tarascon and co-workers,338 relatively low abundance of alkyl carbonate species
Dahn and co-workers proposed that the thermal in the original SEI and the large amount of LiF after
decomposition of the SEI occurred via the transfor- storage, the authors disagreed with Kanamura et al.,
mation of the metastable lithium alkyl carbonate into who attributed the formation of LiF to the reaction
stable Li2CO3 (Scheme 23).340-342 This suggestion was of lithium alkyl carbonate or Li2CO3 with HF.282
supported by Amine and co-workers, who studied the Rather, they suggested that, the direct decomposition
change of surface chemistry on well-cycled carbon- of fluorinated salt anions (PF6- or BF4-) occurred at
aceous anodes with elevated temperature storage by the elevated temperature and formed LiF through a
using XPS and observed the conspicuous presence of variety of possible reactions.
Li2CO3 on the graphite surface that was stored at 70 A distinct difference between the behaviors of
°C for prolonged durations (> weeks), as indicated LiBF4- and LiPF6-based electrolytes must be pointed
by the C 1s signal at 290.4 eV in Figure 29.294 out. The recycling after storage at 60-80 °C revealed
Additional evidence came from the decreasing ether/ that lithium/graphite half-anode cells with LiBF4-
alkoxide signal at 287.4 eV in C 1s spectra and 534 based electrolyte suffered serious capacity loss, while
eV in O 1s spectra. Furthermore, the elemental with LiPF6-based electrolyte they maintained stable
carbon signal at 284.5 eV was also clearly visible, performance, as Figure 30 shows. A similar trend was
indicating that the original SEI layer had been found for a graphite anode with a different SOC, and
damaged, which left parts of the bulk graphene layers a threshold value (60 °C) for the storage temperature
exposed. seemed to exist for this performance degradation.
However, Li2CO3 was not observed by Andersson Qualitatively the same surface species were detected
and Edström in their XPS analysis of the graphitic by XPS analyses on graphitic anodes that were
anode that had been precycled and then stored at 60 formed in LiPF6- and LiBF4-based electrolytes; there-
°C for 7 days.277 They found that, during the storage, fore, the above difference could not be explained by
the original SEI consisting of lithium alkyl carbonate chemistries. Andersson et al. speculated that the
was indeed unstable and disappeared with time, as morphology of the major surface species, LiF, played
Figure 31. Effect of lithium salt on Mn2+ dissolution from

Li1.05Mn1.95O4 stored in EC/DMC (2:1) electrolytes at 55 °C.
The Mn2+ concentrations in the electrolytes were deter-
mined by atomic absorption spectroscopy. (Reproduced with
permission from ref 301 (Figure 4). Copyright 1999 The
Electrochemical Society.)
decomposition and rebuilding of surface layers are

responsible for the fade in capacity and loss in
power.296,297,343-346
Using thermogravimetric analysis (TGA), Du Pas-
quier et al. studied the thermal breakdown of the
surface layer on a spinel cathode that was aged in
LiPF6/EC/DMC at 100 °C for 8 h,301 and they detected
the presence of two distinct weight loss processes at
Figure 30. Capacity loss due to storage at elevated
temperatures for Li/graphite half-cells. All cells were 140 and 500 °C. While assigning the latter to the
precycled at room temperature (cycles 1-3) prior to storage transformation of spinel into Mn2O3, they believed
at indicated temperatures for 1 week, followed by continued that the former process corresponded to the burning
cycling at room temperature. The electrolytes used were of the organic coating on the cathode surface and the
EC/DMC (2:1) and (a) 1.0 M LiPF6 or (b) 1.0 M LiBF4. The concomitant release of CO2, similar to the thermal
cells were stored in delithiated states. (Reproduced with transition that had been observed on the graphitic
permission from ref 277 (Figure 2). Copyright 2001 The
Electrochemical Society.) anode in the same temperature range.338 Naturally,
at room temperature, the above destructive process
a critical role in defining the surface state of the might only proceed with negligible kinetics.
anode. A highly speculative mechanism was proposed For a spinel manganese cathode, the decomposition
in accordance with this hypothesis in which the LiF of the SEI is often entangled with the instability of
formed at 60 °C from a LiBF4-based electrolyte blocks the cathode structure in electrolytes at elevated
the intercalation edge sites of the graphitic anode temperatures, and the participation of electrolyte
more effectively than the LiPF4-based counter- components in the Mn2+ dissolution or the dispro-
parts.277 portionation of the spinel is already widely ac-
The work by Ross and co-workers on the thermal cepted.293,301,343 The acidic nature of the electrolyte,
stability of DEC on a lithiated graphitic anode further originated from the moisture-sensitivity of PF6- or
challenged the proposal that Li2CO3 was formed BF4- anions, is obviously the main cause, since a
during the decomposition of the SEI.284 By the well- close correlation has been established by Du Pasquier
controlled exposure of the fully lithiated graphite et al. between the HF content and the Mn2+ content
LiC6 to DEC vapor in a vacuum and gradually in the electrolyte301 and LiPF6 has been identified as
heating the electrode, they used XPS to monitor the the salt species that causes the most severe Mn2+
transformation of the SEI as the function of temper- dissolution in state-of-the-art electrolytes301,343 (Fig-
ature. From 0 to 110 °C, the major surface species ure 31), apparently due to its higher tendency to
were believed to be oxalate anions as well as alkox- hydrolyze. Elevated temperatures promote this dis-
ides, while, from 120 to 140 °C, obvious chemical solution process in a pronounced way, as Figure 32
reactions were observed. On the basis of the C/O shows. The continuous trend in Mn2+ concentration
stoichiometric change, it seemed that organic anions implies that the dissolution process is kinetically
such as oxalate anions were converted into simple governed, and there is no threshold temperature at
inorganic anions. The authors thus proposed that, in which this process is switched on. However, accelera-
the exothermic process that was observed by Taras- tion does occur above 40 °C.301 XRD analyses on the
con and co-workers in the DSC experiment,338 the SEI recovered spinel powder stored at 100 °C revealed
most probably decomposed into simple inorganic the formation of a new phase, which the authors
species such as Li2O and CO, instead of Li2CO3.284 identified to be a protonated λ-MnO2 that is partially
On a cathode surface, the instability of electrolytes inert with respect to electrochemistry. To account for
at elevated temperatures is also well studied, and it the Mn2+ concentration and the formation of a
has been generally established that the continuous protonated λ-MnO2 phase, the authors believed that
Figure 32. Effect of elevated temperatures on Mn2+

dissolution from Li1.05Mn1.95O4 stored in LiPF6/EC/DMC
(2:1). The Mn2+ concentrations in the electrolytes were
determined by atomic absorption spectroscopy. (Reproduced Figure 33. Relative amount (atom %) of surface species
with permission from ref 301 (Figure 1). Copyright 1999 C, O, and F on cathode samples as a function of storage
The Electrochemical Society.) temperature. Numbers in parentheses indicate the storage
duration in weeks. (Reproduced with permission from ref
Scheme 24. Structural Transformation of a Spinel 294 (Figure 8). Copyright 2002 The Electrochemical Soci-
Cathode at Elevated Temperatures Assisted by the ety.)
Acidic Species in a LiPF6-Based Electrolyte
time, as Figure 33 shows. This unexpected trend in
F abundance contradicts the knowledge about the
instability of LiPF6 salt at elevated temperatures as
well as the results of LiF content on graphite277,294
a proton-lithium ion-exchange process occurred in- and spinel materials301,343 during similar storage
stead of simple disproportionation (Scheme 24).310 tests. To account for this abnormality, the authors
The structural transformation of spinel cathode argued that, while it would be impossible for the F
materials is mainly responsible for the capacity content to actually decrease with temperature, the
fading of lithium ion cells based on them and LiPF6- screening of the F 1s signal by an overlaying surface
containing electrolytes. Although other authors differ compound that is non-fluorine in nature could occur
on the mechanism by which the capacity loss is and would reduce the opportunity for the F 1s
caused, they generally agree that the acidic nature electron to escape.294 In other words, the polymeric
of the electrolyte plays an influential role.296,300,346 and inorganic moieties grown on a cathode surface
Thus, to make the advantages of spinel cathode camouflage the increased LiF content.
materials feasible, the acidic nature of the electro-
lytes has to be mitigated, and the replacement of 7.3. Thermal Safety of Electrolytes against Abuse
lithium salts based on anions with labile fluorine
such as PF6-, BF4-, and AsF6- seems to be a logical The catastrophic scenario for the thermal reaction
solution. between electrolytes and electrode materials in a
On the other hand, the adverse effect of HF on lithium-based cell is called “thermal runaway”,347-351
capacity seems to be much less significant to other whose onset temperature determines the safety limit
cathode materials based on LiNiO2 and LiCoO2, of lithium-based cells. Thermal runaway occurs when
which exhibit higher stability during storage at the amplitude of an accidental temperature hike in
elevated temperatures, although an abundance of LiF the cell, usually initiated by local overheating, reaches
found on cathode surfaces still indicates the reaction a certain threshold value, and the heat thus gener-
of a salt anion, most probably after its hydrolysis by ated can no longer be thermally dissipated but
trace moisture.294,302 In the XPS analysis conducted instead triggers a series of exothermic side reactions.
on a LiNi0.8Co0.2O2 cathode that had been stored at Since the rate of heat generation is higher than that
various temperatures up to 70 °C, Andersson et al. of heat dissipation, these exothermic processes pro-
reported that the surface chemistry did not show any ceed under an adiabatic-like condition and the global
apparent dependence on storage temperature, stor- cell temperature experiences an uncontrollable rise.
age duration, or SOC (i.e., the polycarbonates and Figure 34a shows an example of a thermal runaway
various F-containing species in the surface layer that was initiated in a well-cycled (25 times) lithium
remain stable regardless of storage temperature).294 cell by a 140 °C hot box test.348 Once initiated,
However, quantitative changes in the relative ratio thermal runaway apparently does not terminate until
of these species did occur during the storage, and it the reactive agents are completely consumed or, in
was believed that a thicker surface layer was formed most cases, the cell ruptures, venting with flames and
as a result of longer storage at higher temperature. smoke and sometimes even explodes.348,351,352 For
This conclusion was drawn from the simultaneous lithium or lithium ion batteries of industry sizes (i.e.,
increase of C and O species during the storage despite a capacity of >10 A h), serious safety hazards would
the fact that F species actually decreased at the same ensue from such a thermal breakdown.
above the onset threshold. Instead, thermal runaway

is often caused by various abuses, which could be
thermal (overheating), electrical (accidental over-
charge, high pulse power extraction), or mechanical
(crushing, internal or external short circuit).349,351,352
Various techniques have been employed to evaluate
the safety of lithium ion cells, including hotbox348,360
and accelerating rate calorimetry (ARC)348,360,361 as
well as abuse tests such as overcharging, shorting,
crushing, and nail-penetration.351,352 Profound differ-
Figure 34. Temperature response of (a, left) cycled lithium ences in the abuse tolerance were found among the
(Li/MnO2) and (b, right) lithium ion (coke/LiCoO2) cells lithium ion cells from various manufacturers despite
subjected to a 140 °C hot box test. These cells contain the common onset temperature of ∼160 °C for nearly
LiAsF6/EC/PC and LiPF6/PC/DEC as electrolytes, respec- all these cells. In addition to the cell materials
tively, and were cycled at room temperature to the indi- (electrolyte and electrode) that determine the amount
cated cycle numbers. Note that the thermal runaway occurs of heat generation and the rate of heat release, the
for a lithium cell that has been cycled 25 times due to the
formation of high surface area lithium, while extended cell design and engineering was also found to be of
cycling actually offers better safety to lithium ion cells vital importance for abuse tolerance by affecting the
(Reproduced with permission from ref 348 (Figures 1 and efficiency of heat dissipation.349,355,360,361 The presence
4). Copyright 1994 Elsevier.) of excess electrolytes can increase the overall thermal
conductivity of the cell by 55-70% and eliminate its
The concern over the safety of lithium-based cells dependence on OCV but cannot alter the anisotropy
drew a great deal of attention after the unsuccessful of heat flow within the cell. For example, in an
commercialization of lithium cells.353-358 The surface electrolyte-filled cathode/separator/anode stack that
morphology of the well-cycled lithium electrode was simulates an actual lithium ion cell, the cross-plane
believed to be the main source for thermal runaway, thermal conductivity is ∼10 times lower than the in-
because the presence of high surface area dendritic plane one.360 In agreement with the above results,
lithium would greatly promote the exothermic reac- Leising et al. compared the internal and external
tions between lithium and electrolyte, thus lowering temperatures of a prismatic 1500 mA h lithium ion
the threshold temperature at which self-heating cell during an external short, and they found a
occurred. As Figure 34a shows, the temperature of difference of 40 °C between the two.352 The conse-
the fresh cell (cycle 0) under the same testing quence of this anisotropy in thermal conductivity is
condition does not show any temperature overshoot especially serious for the popular spiral-wound cell
but instead approaches 140 °C in an asymptotical design because it creates an adiabatic-like environ-
manner, as would be expected from a thermally inert ment for any thermal perturbation occurring near the
object. Temperature overshoot, a sign of self-heating core of the cell and increases the opportunity for
within the cell, increases monotonically with cycle thermal runaway.
number until, after 25 cycles, the cell possesses a At elevated temperatures, there are five possible
highly reactive lithium surface and undergoes ther- reactions that can occur between the cell compo-
mal runaway, as characterized by the dramatic rise nents: (1) thermal decomposition of bulk electrolyte;
in temperature.348 At the time, the role played by (2) chemical reduction of electrolyte by the anode; (3)
cathode materials in the safety of lithium cells seems
chemical oxidation of electrolyte by the cathode; (4)
to be overshadowed by the presence of lithium anodes
thermal decomposition of the cathode and anode; or
because of the much higher reactivity of the latter.
(5) melting of the separator and the consequent
The replacement of lithium by carbonaceous anode
internal short. To identify which of these contributes
materials eliminated the highly reactive anode sur-
the decisive amount of heat that is critical in initiat-
face as a possible trigger for thermal runaway and
ing the thermal runaway, it is necessary to study the
led to the successful commercialization of lithium ion
thermal responses of these individual components or
technology in the 1990s. However, safety remains an
component couples separately.
issue, since thermal runaway can be caused by
mechanisms other than the formation of lithium The bulk electrolyte components do not seem to be
dendrites. A general opinion is that the replacement such an initiation, since the thermal decomposition
of the metallic lithium by a carbonaceous anode does of the salts and the interactions between the salts
improve safety, as Figure 34b shows.348,351,352,359 In and the solvents start as low as 70 °C,150,152,330,337 and
contrast to a metallic lithium anode, a well-cycled the thermal effect of these reactions is generally
carbonaceous anode offers additional self-heat sup- negligible as compared with the processes involving
pression, as evidence by the absence of the temper- the fully charged anode and cathode materials.361
ature overshoot which was originally observed in a In early research efforts, attention was concen-
fresh cell. For most commercial lithium ion cells, the trated on carbonaceous anodes because of the earlier
onset temperature that would trigger a thermal experiences with metallic lithium. Dahn and co-
runaway is considered to be >150 °C.351,352,360 workers studied the thermal response of carbon-
According to mathematic modeling and abuse tests, aceous materials in the presence of electrolytes in an
the normal operation for most lithium or lithium ion adiabatic environment created in a thermal analysis
cells that use nonaqueous electrolytes should not technique known as accelerating rate calorimetry
induce sufficient heat to raise the cell temperature (ARC). By choosing an arbitrary threshold value for
the self-heating rate (SHR) of 0.02 °C min-1, they

found that, besides the lithiation state of the anode,
electrolyte composition had a critical effect on the
onset temperature of thermal runaway, with the SHR
increasing in the order
EC < DME < DEC < DMC

While the rationale for the above order remains to
be well-understood, the authors speculated that the
relative effectiveness and solubility of the SEI as well
as the reactivity of these bulk solvents might be
responsible.348 Overall, this study showed that the
reaction between the electrolyte and the lithiated
carbon could trigger thermal runaway, except at
much higher onset temperatures than those of lithium
electrodes.
In their DSC studies on the SEI of carbonaceous
anodes, Du Pasquier et al. identified the exothermic
process at 120-140 °C as the transformation of the
alkyl carbonate-dominated SEI into the more stable
Li2CO3, and the heat generation associated with this
process strongly depended on the surface area of the
electrode. The reduction of electrolytes by lithiated Figure 35. Temperature-time (a) and SHR (b) profiles
carbon occurred between 210 and 230 °C, with measured for lithiated MCMB in LiPF6/EC/DEC (33:67).
comparable reaction heat to that of the transforma- The MCMB electrodes were lithated to three composi-
tion of an SEI.338 According to the 140 °C hotbox test tions: Li0.75C6 (0.0 V), Li0.45C6 (0.089 V), and Li0.25C6 (0.127
V). (Reproduced with permission from ref 340 (Figure 8).
that was conducted with a lithium ion cell, it seemed Copyright 1999 The Electrochemical Society.)
that the breakdown of the SEI could be excluded as
the cause for thermal runaway.348 Works by other
which appears to be relatively independent of the
authors confirmed the occurrence of this SEI trans-
lithiation degree of the carbonaceous anode, should
formation process at >100 °C and the dependence of
represent a reaction that proceeds until one of the
its heat generation on the carbonaceous anode sur-
reactants is depleted. The decomposition of bulk
face,359,362,363 but they disagreed on its role in initiat-
electrolyte does not contribute to the SHR until 190
ing thermal runaway. For example, Joho et al.
°C; therefore, there must be another exothermic
systematically measured the heat evolution of vari-
process between 150 and 190 °C. During this tem-
ous graphitic anodes in the presence of LiPF6/EC/
perature range, the SHR does depend on the lithia-
DMC between 80 and 220 °C and evaluated the
tion degree of the anode, with the more lithiated
possibility that the graphitic anode was the source
anode generating a higher SHR. XRD analysis con-
for thermal runaway in lithium ion cells. On the basis
firmed that this process does consume the interca-
of the overall heat capacity data of the Sony 18650
lated lithium in the graphite interior.
lithium cell, they estimated that the temperature rise
in the cell as the result of heat generation of those On the basis of the above observation, Dahn and
processes ranged between 30 and 90 °C, depending co-workers proposed a thermal reaction scheme for
on the surface area as well as the irreversible the coupling of carbonaceous anodes and electrolytes.
capacity of the anode, and concluded that, while the The initial peak, which was almost identical for all
transformation occurred at >100 °C, a temperature of the anode samples and independent of lithiation
hike of this amplitude could place the cell well in the degrees, should arise from the decomposition of the
range of thermal runaway onset temperatures be- SEI because the amount of SEI chemicals was only
tween 160 and 180 °C; therefore, the thermal decom- proportional to the surface area. This could have been
position of the SEI and the subsequent reduction of due to the transformation of the metastable lithium
electrolyte components by lithiated carbon could be alkyl carbonate into the stable Li2CO3. After the
the source that initiates thermal runaway.359 depletion of the SEI, a second process between 150
In their further study on the thermal reaction and 190 °C was caused by the reduction of electrolyte
between carbonaceous anodes and electrolytes, Ri- components by the lithiated carbon to form a new
chard and Dahn improved the ARC technique and SEI, and the autocatalyzed reaction proceeded until
miniaturized it for small sample amounts (<1.0 g) all of the intercalated lithium was consumed or the
while retaining the sensitivity.340,341 The self-heating thickness of this new SEI was sufficient to suppress
was measured for MCMB in various electrolytes with further reductions. The corresponding decrease in
different initial heating temperatures, and the SHR SHR created the dip in the least lithiated sample in
profile, characterized by a common initial peak, was Figure 35. Above 200 °C (beyond the ARC test range
obtained for nearly all of the tests, as shown in Figure as shown in Figure 35), electrolyte decomposition
35. The authors speculated that the SHR profile occurred, which was also an exothermic process.
represented in Figure 35 featured characteristics of The tests with various electrolytes revealed that
three distinct reaction processes. The initial peak, both electrolyte salts and solvent compositions affect
the initial SHR below 100 °C, which is of critical

importance to the initiation of thermal runaway. For
example, using the threshold value of the SHR at
0.02 °C min-1 as the sign of onset, the self-heating
for MCMB in LiPF6/EC/DEC (33:67) starts at 80 °C,
whereas it begins at 50 °C in LiBF4/EC/DEC (50:50)
and at 70 °C in LiPF6/EC/DEC (50:50). The SHR
profiles obtained for the MCMB anode in LiBF4/EC/
DEC did not show the initial peak that corresponds
to the transformation of the SEI, suggesting that the
SEI formed on the carbonaceous anode in LiBF4-
based electrolyte did not contain high amounts of
metastable components such as lithium alkyl carbon-
ate, as the SEI formed in LiPF6- or LiAsF6-based
electrolytes would have. This inference was in agree-
ment with the work by Aurbach et al.,104,117 Kana- Figure 36. Comparison of DSC and TGA profiles for a
mura et al.,222,223 and Andersson and Edström.277 On cathode (PE), anode (NE), and separator (SP) composite
the basis of these ARC results, these authors believed with those of cathode/electrolyte and anode/electrolyte.
that if a lithiated MCMB containing LiPF6/EC/DEC Note the close match between the DSC/TGA profiles of the
(33:67) was heated directly to 150 °C, the known stack and the cathode/electrolyte couple. The OCV of the
stack is 4.15 V. (Reproduced with permission from ref 361
threshold for thermal runaway for most commercial (Figure 4). Copyright 1999 The Electrochemical Society.)
lithium ion cells, the anode would self-heat at a rate
of 100 °C min-1. Hence, the universal self-heating cathode/electrolyte and the anode/electrolyte, and a
behavior of a lithiated carbonaceous anode in the close match was identified between the DSC profile
presence of electrolytes is very useful for predicting of the cathode/electrolyte couple and that of the cell
the initial self-heating in a practical lithium ion cell stack, which consisted of cathode/separator/anode in
when combined with the overall heat capacity of the the presence of electrolytes in a simulation of an
cell.340 actual lithium ion cell (Figure 36).361 Therefore, the
The seminal work by Maleki et al. in 1999 seemed authors concluded that the thermal runaway was
to provide a direct answer to the question about mainly caused by the heat generation from the
which of the five possible processes was responsible cathode material, which decomposed to release oxy-
for the thermal runaway of a lithium ion cell.361 Using gen, and the electrolyte components were oxidized
ARC, they first determined the thermal runaway subsequently. While the reaction of the anode/binder
onset temperature in a lithium ion cell based on occurred at a relatively higher temperature, its
LiCoO2/graphite with LiPF6/EC/DMC/DEC to be 167 substantial heat certainly contributed to the ultimate
°C. The thermal reaction, however, was found to start consequences of the thermal runaway. In a more
at 123 °C and continued to self-heat the system to comprehensive conclusion, the safety of a given
the above onset temperature. Using DSC and TGA, lithium ion cell is dominated by the electrolyte
they further determined the heat evolution as well reactions with the cathode and with the anode
as the thermal profile for the individual components simultaneously.
of the cell in the presence of electrolytes, which The thermal decomposition of various cathode
included cathode, anode, and anode binder (PVdF). materials based on transition metal oxides has been
known since the potential of these materials as
Thus, the total exothermic heat generation of 697 rechargeable lithium-based batteries became real-
J g-1 was recorded for the anode-related reactions, ized, and the decomposition chemistry releasing
which included the major heat generated from anode/ oxygen has usually been thought to proceed at
binder reactions starting at 210 °C and the anode/ temperatures above 200 °C.362,364-366 Interestingly,
electrolyte reactions starting at 125 °C. The break- Maleki et al. found that the presence of a surface
down of the SEI on the anode occurred at 120 °C, as layer on the cathode, which was formed through the
characterized by the distinct peak, which followed an decomposition of electrolytes on the cathode during
endothermic process that was attributed by the the cycling of the cell, actually promoted the thermal
authors to the bulk electrolyte decomposition below decomposition of the cathodes to occur at a lower
100 °C. While the heat effect of the SEI breakdown temperature with higher heat generation and mass
seemed negligible as compared to those of the cathode/ loss. The decomposition of the surface layer served
electrolyte and anode/electrolyte reactions, it con- as the thermal trigger for the major oxygen-releasing
tributed to the self-heating of the cell. For cathode- process.
related processes, the total exothermic heat generation A direct result of the work by Maleki et al. is the
was 407 J g-1, with the major thermal reaction gradual shift of safety research from carbonaceous
starting at 167 °C. In other words, although the heats anode to cathode materials. In the following years,
associated with anode-related reactions are higher, Dahn and co-workers used an improved ARC tech-
the heat from the cathode processes was released at nique to conduct a series of systematic investigations
a relatively lower temperature and was more likely on the reaction between electrolyte and cathode
responsible for the triggering of thermal runaway. materials based on LiCoO2367-369,371 and spinel
A comparison was made between the profiles of LiMn2O4.368,370 These studies have depicted a clear
heat evolution associated with the reactions of the picture regarding the safety of lithium ion devices.
Scheme 25. Thermal Decomposition of LiCoO2 in Scheme 27. Thermal Combustion of EC by a Fully
the Absence of Electrolyte Delithiated Spinel Cathode
Scheme 26. Thermal Combustion of EC by the

Cathode LiCoO2
material could offer improved thermal safety as
compared to LiCoO2, but the SHR in this case was
very sensitive to the relative amount of electrolyte
compared to that of the cathode, while the corre-
In a standard electrolyte, LiPF6/EC/DEC, the fully sponding dependence was not found with LiCoO2/
delithiated LiCoO2 (OCV ) 4.2 V) showed the thres- electrolyte.368 Thus, only when the electrolyte/cathode
hold SHR of 0.02 °C min-1 at ∼130 °C that eventually ratio was in the vicinity of 1:6, which is close to the
led to thermal runaway at 160 °C. It had been found actual electrolyte/cathode ratio in lithium ion cells,
that the thermal decomposition of LiCoO2 occurred would the LiMn2O4-based cathode show a lower SHR
at a much higher temperature (>200 °C) with oxygen than LiCoO2-based cells, in consistence with the
generation (Scheme 25). The existence of Co3O4 had hotbox data.368
been confirmed by XRD analysis conducted on a Similar to the results obtained on the thermal
cathode sample that was heated to 320 °C. However, reaction of LiCoO2, the SHR of LixMn2O4 in the
MacNeil and Dahn found that the presence of elec- presence of LiPF6/EC/DEC reached the threshold
trolyte solvents catalyzed the decomposition of LiCoO2 value of 0.02 °C min-1 at ∼150 °C, indicating the
by reacting with the oxygen, thus greatly lowering beginning of exothermic events; then the system
the onset temperature at which oxygen was released. gradually self-heated to 275 °C. In comparative tests,
XRD analysis performed on the cathode after its LiMn2O4 in the absence of electrolytes showed a
reaction with EC/DEC showed that, when solvents detectable thermal event at ∼160 °C and converted
were present, a new phase “CoO” was generated in from the λ-MnO2 structure to β-MnO2 one, while the
addition to Co3O4 and the unreacted LiCoO2.369 The oxidation of electrolyte solvent led to MnO with
authors ascribed the formation of this Co(II) species increased exothermic effect. Since MnCO3 and Mn2O3
to the reducing power of the organic solvents. In an were also detected by XRD analysis, CO2 and H2O
ideal situation with full combustion of EC, the were suspected to be the combustion products of the
authors proposed the reaction scheme shown in solvents.370 Assuming EC as the fuel to be fully
Scheme 26. As a result, the presence of organic combusted, 1 mol of delithiated spinel material can
species as fuel not only reduced the onset tempera- release 2 mol of oxygen and generate heat that is
ture for the thermal decomposition of LiCoO2 but also sufficient to raise the cell temperature over 300 °C,
rendered the reaction more exothermic. This solvent based on the heats of combustion for EC (1161.4 kJ
effect, that is, lower onset temperature and higher mol-1) and DEC (2715 kJ mol-1) (see Scheme 27). In
reaction heat for thermal decomposition of cathode experiments, the actual rise in the ARC sample
materials due to the participation of electrolyte temperature was only 80 °C, most probably due to
solvents, has actually been observed previously by incomplete combustion or depletion of solvent.370
Maleki et al., who ascribed it to a complex synergistic Therefore, the presence of the electrolyte solute,
mechanism among the electrolyte components and LiPF6, added complexity to the thermal decomposi-
LiCoO2 that catalyzed the reaction between electro- tion of the LiMn2O4-based cathode. Contrary to the
lytes and the liberated oxygen.361 Undoubtedly, as salt effect found with the LiCoO2 cathode,369 the onset
fuel in the adiabatic system, the presence of organic temperature of exothermic activity as represented by
solvents in the electrolyte increases the chance of SHR > 0.02 °C min-1 decreased as the concentration
thermal disaster for lithium ion cells. of LiPF6 increased.370 Apparently, LiCoO2 and
In contrast to that of solvents, the effect of the LiMn2O4 have fundamental differences in the way
electrolyte solute, LiPF6, on the thermal decomposi- they react with solvent in the presence of salt. It
tion of the cathode, LiCoO2, was found to be suppres- seemed that the salt mainly contributes to an initial
sion instead of catalyzation.369 The SHR of a partially thermal instability, which increases with LiPF6-
delithiated cathode was measured in a series of concentration.370
electrolytes with various salt concentrations, and a Pointing out that the acidic nature of LiPF6 assists
strong suppression of the self-heating behavior was the dissolution of Mn2+ into electrolyte solution,301,343
found as the concentration of LiPF6 increased above the authors speculated that HF in the electrolyte
0.50 M. The mechanistic rationale behind this salt solution efficiently cleans up the spinel surface of
effect is still not well understood, but the authors MnO; therefore, the bulk electrolyte solvents can be
speculated that the salt decomposition coated the continuously exposed to the fresh surface of Mn2O4
cathode with a protective layer that acted as a and be oxidized.370 As a result, solvent oxidation
combustion retardant. On the basis of these results, would proceed more rapidly as compared with the
the authors recommended a higher salt concentration case of an electrolyte that is less acidic. Therefore,
(>1.50 M) for LiCoO2-based lithium ion cells is for a spinel manganese-based lithium ion cell, a
preferred in terms of thermal safety. higher thermal safety would be obtained with lower
Parallel studies carried out on a manganese spinel salt concentration, and the authors suggested 0.5 M
LiMn2O4-based cathode showed that this cathode as the optimum concentration at which the ion
conductivity and thermal reactivity could be best moisture and lower tendency to react with cyclic
balanced.370 carbonates in a manner similar to that shown by
A recent publication discussing the salt effect on Schemes 4, 12, and 15.132,133 However, a similar
the thermal safety of a LiCoO2 cathode was entitled paradoxical conclusion was drawn about the thermal
with an intriguing but seemingly paradoxical ques- safety of LiCoO2 in the presence of the electrolyte
tion: “Can an electrolyte for lithium ion batteries be based on it, whose inability to produce polymeric
too stable?”, which revealed the relative importance coatings on the cathode is held responsible.371 The
of the interfacial stability arising from the passiva- above results from Dahn and co-workers suggested
tion efficiency against that of the “intrinsic stability” that the salts that were traditionally thought to be
with respect to the safety of cathode materials in thermally stable, such as LiBF4, LiIm, and LiBeti,
lithium ion cells.371 Using an improved DSC tech- should not be used in lithium ion cells if the thermal
nique that enabled the direct analysis of electrolytes safety is the top concern in the application environ-
containing volatile components, the authors investi- ment, which includes large size cells working under
gated the thermal stability of the LiCoO2 cathode in high-rate discharge, at elevated temperatures or at
the presence of the electrolytes based on LiPF6, LiIm, the risk of mechanic abuses and so forth, and that
and LiBeti. The latter two salts, which were based LiPF6 remains the electrolyte solute of choice in
on an imide anion stabilized by two perfluorinated terms of thermal safety.
sulfonyl groups, had been developed in the early
1990s to replace the thermally unstable LiPF6 and 8. Novel Electrolyte Systems
have been reported to be thermally stable till >300
°C.146 8.1. Problems Facing State-of-the-Art Electrolytes
Surprisingly, when the cathode material, LiCoO2,
was in the presence of these “thermally stable” salts, Summarizing the materials reviewed in sections
LiIm and LiMe, much higher reactivity was detected 2-7, one can immediately conclude that the current
than that in the presence of LiPF6, as indicated by state-of-the-art electrolyte systems for lithium ion
the total absence of any combustion suppression on batteries are far from perfect and that, at least in
SHR that had been observed with LiPF6.371 DSC the following four aspects, there is still room for
results of LiCoO2 in the presence of LiIm- or LiBeti- possible improvement. Therefore, research and de-
based electrolytes confirmed the above observation, velopment efforts are continued in an attempt to
which showed the onset thermal decomposition of reformulate new electrolyte systems or to modify the
LiCoO2 to be at ∼280 °C, whereas in LiPF6-based current state-of-the-art electrolyte systems.
electrolytes the same thermal event was much sup- (1) Irreversible Capacity. Because an SEI and
pressed in terms of heat evolution as the concentra- surface film form on both the anode and cathode, a
tion of LiPF6 increased. In other words, the presence certain amount of electrolyte is permanently con-
of LiIm and LiBeti did not introduce any increase in sumed. As has been shown in section 6, this irrevers-
the thermal stability of the electrode, while LiPF6, ible process of SEI or surface layer formation is
although believed to be thermally unstable, efficiently accompanied by the quantitative loss of lithium ions,
suppressed the thermal decomposition of the cathode. which are immobilized in the form of insoluble salts
The authors ascribed the above “abnormality” to such as Li2O or lithium alkyl carbonate.262 Since most
the passivation of the cathode surface by the reaction lithium ion cells are built as cathode-limited in order
products of the electrolyte solvent and salt. Since to avoid the occurrence of lithium metal deposition
LiPF6 readily decomposes organic solvents such as on a carbonaceous anode at the end of charging, this
EC through a ring-opening mechanism at relatively consumption of the limited lithium ion source during
low temperatures, the decomposition products, which the initial cycles results in permanent capacity loss
were believed to consist of a wide variety of com- of the cell. Eventually the cell energy density as well
pounds, including a PEO-like polymer shown by as the corresponding cost is compromised because of
Schemes 4, 12, and 15, deposited on the cathode the irreversible capacities during the initial cycles.
surface and formed a protective layer between the The extent of the irreversible capacity depends on
highly oxidizing cathode materials and the bulk both the anode material and the electrolyte composi-
electrolyte solvents. According to the hypothesis of tion. Empirical knowledge indicates that the PC
the authors, this polymeric coating on the cathode presence, which is well-known for its tendency to
particles strongly delayed the thermal combustion of cause the exfoliation of the graphene structures, is
the solvents by hindering the release of oxygen, especially apt to induce such irreversible capacities.
resulting in a more controlled thermal decomposition On the other hand, reformulation of the electrolyte
of LiCoO2. On the other hand, similar polymer species may lead to significant reduction in the irreversible
were not formed from LiIm- or LiBeti-based electro- capacity for given electrode materials.
lytes because of the stability of these salts, and the (2) Temperature Limits. The two indispensable
thermal combustion of electrolyte solvents in the components of the present lithium ion electrolyte
presence of charged cathode materials proceeded systems are LiPF6 as salt and EC as solvent. Unfor-
unhindered, releasing heat that is sufficient to trigger tunately, these two components also impart their
thermal runaway. sensitivity to extreme temperatures to the lithium
Another well-studied salt, LiBF4, was also believed ion technology, thus imposing temperature limits to
to be more thermally and chemically stable than the operation of lithium ion cells. In a somewhat
LiPF6 in terms of its higher tolerance against trace oversimplified account, one can hold EC responsible
for the lower, and LiPF6 for the higher, temperature approaches, including the development of new elec-
instabilities. trolyte solvents and salts and the application of
Thus, at temperatures lower than the liquidus functional additives. This section is dedicated to cover
temperature (usually above -20 °C for most electro- these efforts, most of which, although, have not been
lyte compositions),50e,159,160 EC precipitates and dras- adopted in actual lithium ion electrolytes.
tically reduces the conductivity of lithium ions both
in the bulk electrolyte and through the interfacial
films in the system. During discharge, this increase 8.2. Functional Electrolytes: Additives
of cell impedance at low temperature leads to lower
Instead of entirely replacing the major components
capacity utilization, which is normally recoverable
of the current state-of-the-art electrolyte systems that
when the temperature rises. However, permanent
cause problems, an efficient and economical alterna-
damage occurs if the cell is being charged at low
tive is to modify certain targeted functions of the
temperatures because lithium deposition occurs,
electrolytes by incorporating a new component at
caused by the high interfacial impedance, and results
small concentrations, known as an additive, so that
in irreversible loss of lithium ions. An even worse
its potential impact on the existing electrolyte can
possibility is the safety hazard if the lithium deposi-
be minimized. In this way, the bulk properties of the
tion continues to accumulate on the carbonaceous
electrolyte system can be maintained with the al-
surface.
ready proven merits such as cost and environmental
At temperatures higher than 60 °C, various de- concerns barely changed, since the presence of the
compositions occur among the electrolyte compo- new component in the bulk is negligible. On the other
nents, electrode materials, and SEI or surface layers, hand, the additive could significantly change the
while LiPF6 acts as a major initiator or catalyst for targeted property. This is especially pronounced in
most of these reactions.152,310,332,333 The damage caused terms of interfacial properties because these additives
by high-temperature operation is permanent. Be- are usually preferentially involved in interfacial
cause gaseous products accumulate, a safety hazard redox processes before the main components of the
is also likely. Therefore, the specified temperature bulk electrolyte are.
range for the normal operation of most commercial The additive approach has been used for lithium
lithium ion cells is -20 °C to +50 °C. While sufficient batteries to improve the surface morphology of a
for most consumer purposes, the above range severely lithium electrode so that dendrite growth can be
restricts the applications of lithium ion technology avoided.10, 372-375 Since the concept of the SEI was
for special areas such as military, space, and vehicle proposed by Peled et al.,37 the emphasis has been
traction uses. placed on the reductive decomposition of these addi-
(3) Safety and Hazards. The linear carbonate tives and the effect of the decomposition products on
solvents are highly flammable with flash points the physicochemical properties of the SEI. Obviously,
usually below 30 °C. When the lithium ion cell is when carbonaceous materials replaced lithium metal
subject to various abuses, thermal runaway occurs as anodes, the same line of thought led to the
and causes safety hazards. Although electrode ma- attempts at controlling the SEI chemistry by means
terials and their state-of-charge play a more impor- of using various additives.376 During the past decade,
tant role in deciding the consequences of the hazard, this approach has been thoroughly studied, the focus
the flammable electrolyte solvents are most certainly of which has been placed at the SEI on the anodes,
responsible for the fire when a lithium ion cell vents. although additives targeting other cell components
The seriousness of the hazard is proportional to the have also been developed. However, because of direct
size of the cell, so flame-retarded or nonflammable commercial interest, most of the work on additives
lithium ion electrolytes are of special interest for has never been published in technical journals,
vehicle traction batteries. especially the work that has eventually been accepted
(4) Better Ion Transport. In most nonaqueous for use in commercial lithium ion cells. As an
electrolytes, the ion conductivity is much lower as alternative, patent disclosures and conference ab-
compared with aqueous solutions, and the part of the stracts did reveal scattered information about this
current that is carried by the lithium ions is always aspect, although fundamental insight is usually
less than half.176 Although it has been found that, unavailable in these forms of literature. In recent
for actual cell operation, the impedances at the years, electrolytes containing additives have been
interfaces of anode/electrolyte and cathode/electrolyte named “functional electrolytes” by some research-
weigh far more than the bulk ion conductivity does, ers.377
there is a semiempirical rule with very few excep- According to the functions targeted, the numerous
tions: the higher the bulk ion conductivity of an chemicals tested as electrolyte additives can be
electrolyte is, the more conductive the SEI or surface tentatively divided into the following three distinct
films formed in this electrolyte can be. On the other categories: (1) those used for improving the ion
hand, the improvement in lithium ion transference conduction properties in the bulk electrolytes; (2)
number is certainly welcome, although its signifi- those used for SEI chemistry modifications; and (3)
cance in liquid electrolytes might not be as high as those used for preventing overcharging of the cells.
in polymer electrolytes. Since the additives designed for the last purpose are
Since the inception of lithium ion technology, there usually compounds with oxidation potentials close to
has been a great deal of research aimed at improving the operating potential of the cathode materials, the
the state-of-the-art electrolyte systems via various coverage of additive studies has included essentially
Table 8. Performance-Enhancing Additives for Bulk Properties
every major component of the lithium ion cell that transference number is lowered as a result of its
interacts with the electrolytes. presence. In other words, the addition of crown ethers
in nonaqueous electrolytes actually promotes the
8.2.1. Bulk Electrolyte: Ion Transport undesired anion transport. The main barrier for the
application of crown ethers in electrolyte, however,
The ability of crown ether to coordinate with is their toxicity. The environmental concern over the
lithium ion has long been recognized, and in terms processing and disposal of any materials containing
of the cavity size, 12-crown-4 and 15-crown-5 have these crown ethers makes it impossible for industry
been identified as the most efficient ligands for to adopt them in large-scale applications.
lithium ion.378 The idea of using these cyclic polyether To develop an additive that selectively coordinates
compounds to promote the solvation of lithium salts with salt anions and frees lithium ion for conduction,
in nonaqueous electrolytes was actively pursued McBreen and co-workers pursued a molecular design
when rechargeable cells based on lithium metal and tailor-synthesis approach that yielded several
electrodes were still the commercial objective. It has families of novel compounds based on nitrogen or
been found that the presence of both 12-crown-4 and boron centers with strongly electron-withdrawing
15-crown-5 can effectively improve the solubility of substituents.
the lithium salts and increase the ionic conductivity The first family of the so-called anion receptors was
in the resultant electrolytes, especially when the aza-ethers that were based on cyclic or linear amides,
solvents have low dielectric constants.379-383 This where the nitrogen core was made electron-deficient
improvement in bulk ion conductivity is also reflected by the perfluoroalkylsulfonyl substituents so that
in the interface properties, as the charge-transfer these amides would preferentially interact with the
resistance on the LiCoO2 cathode is also reported to electron-rich anions through Coulombic attraction,
be reduced because of the presence of 12-crown- contrary to how their unsubstituted counterparts
4.380,382 The electrochemical stability limits are not would act.384,385 Two selected representatives from the
obviously influenced by crown ethers, but considering aza-ether family are shown in Table 8. When used
their ether-like structure, one should be concerned as additives in solutions of various lithium halides
with their stability on the fully charged cathode LiX in THF, these novel compounds were found to
surface in the long term. In polymer-based electro- increase both the solubility and the ion conductivity
lytes, 12-crown-4 was also found to decrease the glass of these solutions. For example, the ion conductivity
transition temperature of the system.382 of the LiCl/THF solution was 0.0016 mS cm-1, while
On the other hand, since the increase of ion the LiCl/THF solution with one of the linear aza-
conductivity is realized through the coordination of ethers containing eight perfluoroalkylsulfonyl sub-
lithium ions by crown ether molecules, the lithium stituents (n ) 5 for the linear aza-ether shown in
been achieved thus far in nonaqueous lithium elec-

trolytes.
Unfortunately, these aza-ethers showed limited
solubility in the polar solvents that are typically
preferred in nonaqueous electrolytes, and the elec-
trochemical stability window of the LiCl-based elec-
trolytes is not sufficient at the 4.0 V operation range
required by the current state-of-the-art cathode
materials. They were also found to be unstable with
LiPF6.390 Hence, the significance of these aza-ether
compounds in practical applications is rather limited,
although their synthesis successfully proved that the
concept of the anion receptor is achievable by means
of substituting an appropriate core atom with strong
electron-withdrawing moieties.
In their continued efforts, McBreen and co-workers
selected boron, an electron-deficient atom, as the core
to build a series of new anion receptors using the
same tactics with electron-withdrawing substituents.
These new additives can be classified roughly into
Figure 37. NEXAFS spectra at the K-edge of chloride for thesethreesubcategories: borate,386-388borane,313,386,387,389-391
(a) LiCl crystal, (b) 0.2 M LiCl/THF, (c) 0.2 M LiCl/THF +
0.1 M aza-ether that does not have electron-withdrawing and boronate.392 Selected representatives from each
substituents on N, and (d) 0.2 M LiCl/THF + 0.1 M linear category are also listed in Table 8.
aza-ether with n ) 3 in Table 8. Note the white line peak Basically, these boron-based anion receptors are
split when the electron-withdrawing substituents perfluo- much more efficient in coordinating anions, perhaps
romethylsulfonyl are present. (Reproduced with permission because of the electron-deficient nature of boron,
from ref 384 (Figure 2). Copyright 1996 The Electrochemi-
cal Society.) since most of them can even effectively dissociate LiF
up to 1.0 M, which is virtually insoluble in most
organic solvents, and yield ion conductivities as high
Table 8) exhibited an increase of ∼900 times at 1.4 as 6.8 mS cm-1 in DME.313 Considering that the
mS cm-1.384 There was an obvious relationship be- electrochemical oxidation potential for F- is ∼5.9 V,
tween the ability of the aza-ether to coordinate the this new electrolyte does indeed seem attractive in
lithium cations and the extent by which ion conduc- providing a wide electrochemical stability window.
tivity was improved, since the latter proportionally On a GC electrode, the electrochemical stability
increased with both the number of the electron- limits were found to be in the range 4.05-5.50 V,
withdrawing substituents in the molecule and the usually set by the oxidation of electrolyte sol-
electron-withdrawing ability of the substituents.384,385 vents,388,392 while, on various cathode materials,
For example, tosylsulfonyl, a weaker electron-with- stable cycling performance was observed with the
drawing group as compared with perfluoroalkylsul- upper voltage limit as high as 4.30-4.50 V.313,387,390
fonyl, proved to be a much less efficient additive.385 Cycling tests at elevated temperature (55 °C) further
On the basis of these dependences, it seemed that showed that the electrolytes based on LiF coordinated
these aza-ethers did act as anion receptors in the by these additives were stable when compared with
nonaqueous solutions. the state-of-the-art electrolytes based on LiPF6.389
To further confirm that these aza-ethers actually Similar stability was also found on the carbonaceous
coordinate with the salt anions, the authors used anode surface, where the authors concluded that the
near X-ray absorption fine structure spectroscopy presence of these anion receptors did not interfere
(NEXAFS) to study the coordination symmetry of the with the formation of the SEI film, and the dissolu-
chloride anion in LiCl/THF solutions, in the absence tion of the SEI was not detected even after heat
or presence of aza-ether additives, and detected that treatment that would dissolve LiF salt.391
the presence of the additives created a clear split in Among the three subcategories, boronate com-
the Cl K-edge white line peak, as shown in Figure pounds seemed to be the most efficient in coordinat-
37, indicative of the interaction between chloride and ing with anions and enhancing lithium ion stability,
the electron-deficient nitrogen.384 Further XRD stud- although the number of electron-withdrawing sub-
ies conducted on complex crystals grown from the stituents in boronate is only two. The authors thus
cosolutions of the cyclic aza-ether and lithium halide inferred that the ability of these anion receptors to
salts supported the above NEXAFS observations with capture an anion depends not only on the electron-
new Bragg peaks representing a larger d spacing deficiency of the core atom but perhaps also on the
(∼15 Å) in the crystal. Thus, the authors concluded steric hindrance presented by these substituents on
that these new aza-ethers are indeed anion receptors, the core. With only two substituents, the core of the
whose preference in coordinating ions was an exact boronates is obviously more exposed and therefore
opposite to that of the conventional crown ethers. more easily accessible for an anion. The higher ion
With the addition of these molecules, the ion con- conductivity achieved by boronate additive therefore
ductivity and the lithium transference number were comes from the better balance between the electron-
enhanced simultaneously, a benefit that has rarely deficiency and steric openness of this compound as
compared with boranes and borates.392 On the other value for bulk electrolyte solvents, especially PC.
hand, in terms of ion conductivity enhancement, Naturally, certain requirements have to be met by
borates and boranes behave similarly,386 though the decomposition products of the additives, such as
borates show less stability electrochemically and insolubility in electrolytes, film compactness, and low
thermally than boranes.387 impedance to lithium ion transport.
The solubility of these boron-based additives is also The search for potential SEI additives in the past
much higher than that of aza-ethers, and their effects decade has been carried out mainly on a trial-and-
on ion conductivity were studied in popular electro- error basis until, recently, a semiempirical rule was
lyte solvent mixtures such as PC/EC/DMC387 or EC/ developed to facilitate the screening process, which
DMC.388,389,392 Although for the salts such as LiPF6, judges the readiness of a certain compound to be
which are already well dissociated in these solvent reduced on the anode by the energy level of its lowest
mixtures, the enhancement in ion conductivity by the unoccupied molecular orbital (LUMO).377,394 The basis
anion coordination might not be as pronounced as of this rule is the assumption that a molecule with a
that for lithium halides, there is one advantage for lower energy level of its LUMO should be a better
LiPF6 being used with these additives: the thermally electron acceptor and, therefore, more reactive on the
unstable LiPF6 is effectively stabilized by the pres- negatively charged surface of the anode. Sharing the
ence of tris(pentafluorophenyl)borane (TPFPB), as view that interfacial stability originates from the
indicated by the slow scan cyclic voltammetry experi- reactivity, this rule is conceptually related to the
ments conducted on a spinel cathode surface after selecting tool for the SEI-forming electrolyte compo-
storage at 55 °C for 1 week.390 The authors attributed nents as proposed by Peled et al.,106 despite the fact
this stabilization to the coordination of PF6 anion, that the latter was based on a kinetic consideration
which elongated the interatomic distance between Li (exchange current) of the molecule instead of a
and F. This hypothesis was supported by the conclu- thermodynamic one (LUMO energy level). Quantum
sion of a molecular dynamic simulation published calculations have yielded LUMO energy levels for
earlier, wherein Tasaki and Nakamura postulated most of the commonly used electrolyte solvents as
that the reactivity of lithium cation and F would be well as some of the tested SEI additives, which are
suppressed by the elongated distance between them compiled in Table 10.255,377,394,408 Apparently, the
if an effective anion coordination was achieved.393 LUMO levels for all aliphatic cyclic carbonates are
Since these anion receptors based on boron offer a virtually identical, whereas structural modifications
wide variety of merits, their application seems prom- with an aromatic bond, double bond, or substitution
ising if the cost and toxicology do not constitute any of halogen atoms cause large decreases in the LUMO
serious barrier for industry-scale productions. level.255 This trend is in good agreement with the
experimental findings, since most of the additives
8.2.2. Anode: SEI Modification that demonstrated promising performances in elec-
Due to the core importance of the SEI formation trolytes indeed possess one of the active functional-
on carbonaceous anodes, the majority of the research ities. A fairly linear relationship has been established
activities on additives thus far aim at controlling the between the LUMO energy level and the reduction
chemistry of the anode/electrolyte interface, although potential for these additives.377
the number of publications related to this topic is The current efforts with SEI additives usually aim
rather limited as compared with the actual scale of at one of these two distinct but closely interrelated
interest by the industry. Table 9 summarizes the goals: (1) to minimize the irreversible capacity
additives that have been described in the open needed in the first cycle to form the SEI and (2) to
literature. In most cases, the concentration of these enable the use of PC at higher concentrations in the
interface-targeted additives is expected to be kept at electrolyte. While the former would generally benefit
a minimum so that the bulk properties of the a lithium ion cell in terms of a more stable SEI, the
electrolytes such as ion conduction and liquid ranges latter is of special significance to the purpose of
would not be discernibly affected. In other words, for adapting the electrolytes based on PC to the highly
an ideal anode additive, its trace presence should be graphitic anode materials so that the operating
sufficient to decouple the interfacial from bulk prop- temperature can be expanded toward the lower limit
erties. Since there is no official standard available without sacrificing energy density.
concerning the upper limit on the additive concentra- In the early era of lithium ion cell research,
tion, the current review will use an arbitrary criterion Aurbach et al. noticed that the presence of CO2 in
of 10% by weight or volume, above which the added the electrolyte had pronounced effects on the lithia-
component will be treated as a cosolvent instead of tion behavior of graphitic anodes. A number of
an additive. electrolytes, which were thought to be incompatible
With few exceptions, the additives that are in- with graphite because they are based on solvents
tended for the modification of the SEI usually have such as methyl formate or THF, delivered much
high reduction potentials, which ensure that these improved performance under 3-6 atm of CO2.250 They
additives are reduced on the anode surface before the proposed that CO2 participated in the formation of
bulk electrolyte components are involved. In other the SEI by a two-electron process, yielding Li2CO3,
words, during the first charging of a lithium ion cell, which assisted in the buildup of the protective surface
an SEI with the chemical signature from an additive film.249 However, in PC-based electrolytes, CO2 pres-
would have been formed before the potential of the ence proved to be ineffective,123 while, in electrolytes
carbonaceous anode reached the onset reduction based on carbonate mixtures such as EC/DMC, the
Table 9. Anode SEI Additives
effect of CO2 seemed to be negligible.104 These efforts 20% effectively suppressed the cointercalation of PC
could be viewed as the first attempts of modifying and supported the reversible lithium intercalation/
the SEI with electrolyte additives. deintercalation of graphitic anodes at low poten-
SO2, on the other hand, was found to be a much tials.395,396 The reduction potential of SO2 was found
more efficient additive, as its presence at less than to be at ∼2.7 V, far above that corresponding to the
Table 10. LUMO Energy Level and Reduction Potentials of Solvents and Additives
cointercalation/decomposition process of PC.251,395,396 alkyl carbonate, the SEI also contained reduction
Thus, Ein-Eli et al. believed that the SEI formation products originated from SO2 such as Li2S and
was initiated by the predominant decomposition of lithium oxysulfur species. An extra merit of SO2 as
SO2.395,396 On the basis of the FT-IR analysis of the an additive in electrolytes is the increase in ion
graphitic anode surface, they suggested that, in conductivity, which is caused by its high dielectric
addition to the solvent decomposition products, lithium constant and low viscosity.396 However, the obvious
Figure 38. Effect of crown ethers on the irreversible Figure 39. Stabilization of a graphitic anode in 1.0 M
capacity. Voltage profiles of a graphite anode in 1.0 M LiClO4/PC electrolyte by 5% ES. For comparison, the
LiClO4/EC/PC (1:1) electrolyte containing (a) 1.0 M 12- voltammograms of a graphitic anode in 1.0 M LiClO4/PC
crown-4, (b) 1.0 M 18-crown-6, and (c) no crown ethers. electrolyte with 10% SO2 were also shown. (Reproduced
(Reproduced with permission from ref 261 (Figure 3). with permission from ref 399 (Figure 3). Copyright 1999
Copyright 1993 The Electrochemical Society.) The Electrochemical Society.)
disadvantages of such an additive are the concerns was cycled in electrolytes containing 12-crown-4.250
over corrosion and the environmental and safety Since the peaks corresponding to the possible reduc-
hazard issues related to the chemical nature of SO2. tion products of 12-crown-4 were absent in the
Its instability on cathode materials has not been well spectra, they concluded that the effect of crown ether
addressed either, since most of the published results was not due to its participation in the buildup of the
were obtained on lithium/graphite half-cells. SEI. Instead, the prevention of graphene exfoliation
was very likely due to the absence of PC molecules
Besides CO2 and SO2, other gaseous species such in graphene interlayers, a direct result of its expul-
as N2O were also considered as additives,376 but none sion from the lithium ion solvation sheath. This was
of these ideas seemed practical for the lithium ion even so when the concentration of 12-crown-4 was
industry, because the application of such approaches only one tenth of the salt concentration. The ef-
would certainly introduce an additional cost as well fectiveness of 12-crown-4 in suppressing PC cointer-
as safety concerns for the commercial cells, especially calation/decomposition is further demonstrated by
when increased internal pressure due to the gas the cycling of the graphitic anode in electrolytes
production during prolonged cycling had already based on neat PC.250 Considering that the main
become a persistent problem for the manufacturers signals collected from the anode surface during the
of industry-size lithium ion cells. experiment were at ∼1661 cm-1, which is character-
In the early 1990s, Wilkinson and Dahn realized istic of the carbonyl stretching in lithium alkyl
that, in addition to enhancing ion conductivity, crown carbonate, the authors suggested that PC was forced
ethers could also reduce the irreversible capacity to be reduced on the surface via a single-electron
associated with the reduction process at 0.80 V when process. This seemed to be in good agreement with
the electrolyte was based on high concentrations of the previous report that only within graphene struc-
PC.397 Shu et al. studied the phenomenon by using tures would PC likely undergo reduction via a two-
12-crown-4 and 18-crown-6 as additives in LiClO4/ electron process and result in exfoliation.
EC/PC and established the direct connection between Despite the highly efficient performance of crown
the decrease in gas evolution associated with the 0.80 ethers in reducing the irreversible capacity at the
V irreversible process and the addition of 12-crown- anode side, especially when PC-based electrolytes
4.398 As Figure 38 shows, the effectiveness of these were used, the toxic nature of these compounds still
crown ethers in suppressing the PC cointercalation/ prevented their application in commercial lithium ion
decomposition seems to be related to the cavity cells.
size,261 knowing that 12-crown-4 serves as a better A sulfur analogue of EC, ethylene sulfite (ES), was
chelating agent for lithium ion than 18-crown-6.378 proposed as an additive for PC-based electrolytes by
The authors suggested that, unlike most of the Winter and co-workers,399 apparently because of its
additives that obey the empirical rule of the LUMO structural similarity to EC and its potential, under
energy level, a crown ether might not be chemically reductive conditions, to release SO2, a known additive
involved in the formation of the SEI but rather affect that effectively suppresses PC decomposition.395,396 As
this process indirectly by means of preferential the voltammetry in Figure 39 shows, ES in only 5%
solvation of lithium ions. Thus, the exclusion of PC presence successfully eliminated the exfoliation of the
molecules from the solvation sheath rather than the graphite anode, whereas 10% SO2 failed. The ir-
reductive decomposition of crown ethers is respon- reversible process corresponding to the reduction of
sible for the reduced irreversible process at 0.80 V.398 ES occurred at ∼2.0 V, lower than that of SO2 by
Their hypothesis was confirmed by the work of ∼0.80 V; however, the quantity of charge associated
Aurbach and co-workers, who performed detailed FT- was much lower. According to the authors, the above
IR investigations on the graphitic anode surface that apparent gap between the reduction potentials of ES
and SO2 actually excluded the self-suggesting as- reduction process could be attributed either to the
sumption that the SEI-forming ability of ES should possibility that the passivation of the graphitic anode
stem from its tendency to release SO2, which in turn by the decomposition products was so efficient that
participates chemically in the buildup of SEI in a the reduction of VC only occurred to a trace extent
mechanism suggested by Ein-Eli et al.395,396 Further- or to the possibility that the scan rate employed in
more, the reduction of SO2 was reversible, whereas the experiments (1.0 mV s-1) was too fast for any
the reduction of ES did not seem to be, as evidenced quasi-equilibrium to be established between the
by the absence of any anodic current in Figure 39. additive VC and the anode surface, the result of
On the other hand, the attempt to use ES as a bulk which was the total drowning of the VC reduction
solvent proved to be unsuccessful because of the high peak by the main event occurring at the lower
irreversible capacity caused by the reduction of ES, potentials. According to a separate voltammetry
although it looked like the reversible capacity associ- study carried out on inert electrodes (Au and GC),
ated with lithium intercalation was not affected by VC could be reduced irreversibly at 1.4 V.214
the irreversible process at 2.0 V. Considering that EIS results indicated that the impedance on a
these results were obtained in an anode half-cell graphite surface was much lower in the presence of
where lithium was the excess material, one should VC; however, at elevated temperatures (60 °C), it
realize that the irreversible reduction of ES would became higher than those of VC-free electrolytes.
cost the capacity of a full lithium ion cell. Therefore, EQCM using a gold-plated quartz crystal detected
ES should only be used as an additive at small that the mass accumulation in VC-containing elec-
concentrations. trolytes was higher by 50% as compared with the
The anodic stability of the electrolyte was not cases of VC-free electrolytes during voltammetric
negatively affected by the presence of ES in LiClO4/ scanning between 3.0 and 0.5 V, indicating that the
PC, as evidenced by the cyclic voltammetry carried reactive VC might be profoundly involved in the
out on the LiMn2O4 surface. The onset of major formation of the SEI. The comparison between the
electrolyte decomposition, seemingly contributed by FT-IR spectra collected on both gold mirror and
the oxidation of PC, occurs at ∼4.8 V, which would graphite electrodes for VC-free and VC-containing
allow the safe use of most state-of-the-art cathode electrolytes further confirmed the participation of VC
materials based on LiCoO2, LiMn2O4, or LiNiO2. The with characteristic absorptions at ∼3000 cm-1, rep-
first cycling of the cathode half-cell, Li/LiMn2O4, was resenting the possible existence of polymeric moieties
shown in the report, but without the extended cycling in addition to the alkyl carbonates observed com-
of this cell or full lithium ion cells.399 monly on a graphitic anode that was cycled in
Among the various additives tested, vinylene car- carbonate-based solvents.
bonate (VC) might be the most famous in the lithium The surface chemistry of VC on graphite was also
ion research and development community, although
semiquantitatively investigated by C 1s and F 1s
the number of publications related to it seems to be
spectra in XPS analyses,404 as shown by Figure 40.
rather small. Its importance can be evidenced by the
Apparently, the presence of VC drastically reduced
number of companies that vied for the patent rights
the content of LiF in the SEI, as indicated by the LiF
for it.400-402
signal at 685 eV, which only appeared as a shoulder
The reactivity of VC apparently arises from its for that of PVdF (anode binder) at 687 eV while it
polymerizable vinyl functionality and its structure was the predominant species in VC-free electrolyte.
of high strain, which is caused by the sp2-hybridized On the other hand, the abundance of lithium alkyl
carbons on the ring (Tables 9 and 10).403 The small carbonate in the SEI as represented by the signal at
concentration of VC can effectively reduce the ir- 289 eV was increased with VC concentration. This
reversible capacity associated with the 0.80 V process improvement of the SEI chemistry with higher
in almost any PC-based electrolyte,400 and its pres- lithium alkyl carbonate content and lower LiF con-
ence, even in bulk concentrations, does not constitute tent was reflected in the increased stability of a VC-
any instability on the charged surface of either anode
containing electrolyte, which demonstrated a more
or cathode materials. In the latter case, an electrolyte
stable performance in a lithium/graphite anode half-
of 1.0 M LiPF6/PC/VC performed reversibly up to 4.3
cell at an elevated temperature of 60 °C, while the
V with a LiMn2O4 cathode.401 In a recent report
identical cell based on VC-free electrolyte suffered
Aurbach and co-workers thoroughly described the
constant capacity loss.404
effect of VC as additive in electrolytes for lithium ion
cells based on the graphitic anode and LiMn2O4 or The possible impact of VC on the stability of an
LiNiO2 cathode, using various techniques including electrolyte at a cathode surface was also investigated
EIS, EQCM, FT-IR, and XPS.404 According to the and no obvious negative effect of VC on either LiNiO2
voltammetry results of their work, there was no (Figure 41) or LiMn2O4 was observed at up to 4.2 V.
discernible characteristic reductive process in volta- A slight kinetic improvement due to the presence of
mmetry that could be unambiguously assigned to the VC, as indicated by the better resolved peaks and
reduction of VC on the graphitic anode, in sharp higher current response, was observed in the slow
contrast to the cases of other additives that showed scan voltammetry, as shown in Figure 41.404 FT-IR
distinctive reduction processes at high potentials. The studies on the cycled cathode surface indicated the
only difference between VC-containing and VC-free formation of surface species that were probably
electrolytes was the much-reduced irreversible ca- oligomeric VC, originating from the cationic polym-
pacity at 0.90-0.80 V. This visual absence of the VC erization on the delithiated cathode surfaces.
Figure 41. Cyclic voltammograms of a LiNiO2 cathode in

VC-free and VC-containing 1.0 M LiAsF6/EC/DMC elec-
trolyte at scan rates of (a) 0.50 mV s-1 and (b) 0.1 mV s-1.
(Reproduced with permission from ref 404 (Figure 11).
Copyright 2002 Elsevier.)
Figure 40. XPS C 1s and F 1s spectra of graphitic anodes
cycled in 1.0 M LiAsF6/EC/DMC electrolyte. 5% VC was
Another highly efficient additive for the anode SEI
used as additive in the left spectra. Note the different scales is catechole carbonate (CC), reported by Wang et al.,
for the two F 1s spectra. (Reproduced with permission from which has a benzene ring fused with a functionality
ref 404 (Figure 10). Copyright 2002 Elsevier.) structurally similar to EC (Tables 9 and 10).377,409
Like VC, in small percentage (∼2%), it suppresses
the irreversible capacity due to PC decomposition/
In conclusion, many authors believed that VC was cointercalation at 0.80 V and supports extended
a rather effective additive for chemical modification cycling of the graphitic anode in LiPF6/PC/DEC. A
of the anode SEI. It not only reduced the irreversible study on the concentration-dependence of this ir-
capacity during the initial charging process of lithium reversible capacity indicated that the optimum con-
ion cells, but it also improved the stability of the SEI tent of CC lies between 0.5% and 2%. The authors
at elevated temperatures. Because of VC participa- postulated that the mechanism through which CC
tion, the new SEI chemistry contained polymeric stabilized the graphitic anode against exfoliation by
species that stemmed from the reductive polymeri- PC should involve the direct participation of CC in
zation of VC, which was characterized by a high the formation of the SEI, since its small concentration
content of alkyl carbonate functionalities. Compara- does not allow for preferential solvation as crown
tive studies of this additive in commercial lithium ether does. They further proposed that in electrolyte
ion cells proved that VC presence improved cycle life solutions CC served as a radical scavenger and
performance.405-407 quenched PC radical anions so that the continued
It should be mentioned that the commercially reduction of PC via a two-electron process was
available VC usually contained a small percentage inhibited.
(<2%) of prohibitors that stabilized the reactive VC The anodic stability of the electrolyte containing
from polymerizing upon storage. These compounds 2% CC was tested on a LiCoO2 cathode, and no
were usually radical scavengers such as 2,6-di-tert- obvious oxidative decomposition was observed. It
butyl-p-cresol (DBC) or butylated hydroxytoluene must be noted, though, that the above conclusion was
(BHT). Recently, reports pointed out that the pres- based on a voltammetry experiment conducted with
ence of these prohibitors actually had a negative a relatively high scanning rate (15 mV s-1). The
impact on the anodic stability of VC on various performance test of the electrolyte LiPF6/PC/DEC
cathode materials, and VC of high purity was of vital with 2% CC in a full lithium ion cell was shown to
importance in obtaining reproducible performance in deliver a slightly fading capacity when cycling be-
lithium ion cells.404,408 tween 2.75 and 4.10 V.
The continued efforts in this area in recent years

generated a series of reactive compounds as potential
candidates, which include halogenated species such
as bromo-γ-γBL,410 other γBL derivatives,411 com-
pounds containing vinyl groups,412,413 and compounds
that belong to the succinimide family377 (Tables 9 and
10). Most of these additives were reported to be
effective in reducing the irreversible capacities in the
first charge process, while some also successfully
eliminated the cointercalation of PC and avoided the
exfoliation of graphitic anodes. Among these works,
one interesting mechanistic study was carried out by
Matsuo et al., who employed a 13C NMR technique
to investigate the effect of γBL derivatives on lithium
solvation and found that PC molecule was excluded
from the solvation sheath of lithium ions by these
additives.411 Thus, these authors suggested that γBL
derivatives might suppress PC decomposition via a
mechanism that combines both preferential solvation
as crown ethers do and direct chemical participation
as most of the other additives do, although the
preferential solvation effect would be negligible when
the concentration of these additives is far below 1.0
M. Figure 42. Schematic representation of the shuttling
The work on SEI-modification additives is cur- occurring in an overcharged cell that is based on electro-
rently carried out throughout the lithium ion re- lytes containing redox additive as protection.
search community, and candidates of new structure
are being tested in large numbers. As a result, the injected into the cell when overcharge occurs.414,415
major lithium ion manufacturers have applied vari- The mechanism, intended to prevent the oxidative
ous additives in their electrolyte formulations. How- decomposition of electrolyte solvents at the cathode
ever, the lack of information in the open literature surface, was based on a redox additive that has an
makes the in depth and comprehensive review of this oxidation potential slightly lower than that of the
new branch of electrolyte chemistry difficult. electrolyte anodic decomposition and would serve as
a current shuttle under the condition of overcharge,
8.2.3. Cathode: Overcharge Protection as Figure 42 shows. On an overcharged cathode
surface, the redox additive would be transformed into
Unlike the anode-targeted additives discussed in its oxidized form [O], which, via diffusion across the
the preceding part, the additives intended for cathode cell electrolyte, would be reduced back to its original
protection have a much longer history than lithium (reduced) state [R] on the surface of a negatively
ion technology itself and were originally developed charged anode. Thus, the reversible nature of the
for rechargeable cells based on lithium metal anodes redox couple [R]/[O] would maintain the above cycle
and various 3.0 V class cathode materials. of “oxidation-diffusion-reduction-diffusion” indefi-
It has been long recognized that the accidental nitely and hence lock the cathode potential at the
overcharge of lithium cells would lead to irreversible oxidation potential of [R] until termination of the
decomposition of electrolyte components, which is charge. Ideally, no permanent damage would be done
mainly the oxidative decomposition of solvents on a to the cell capacity during the redox shuttling, since
cathode surface and whose reaction heat and gaseous the Faradaic current was only carried by the revers-
products are responsible for the hazardous thermal ible redox reactions, while, energetically, the excess
runaway. Since the overwhelming majority of the charge injected would be neither stored in the
charging protocols are based on cell voltage as the cathode nor consumed in reversible decompositions
indicator for the end-of-charge, such accidents are but rather dissipated in the form of heat associated
especially likely to occur for multicell battery packs with the shuttling current. Therefore, the net result
in which cells with mismatched capacities are put in of the above redox shuttling mechanism during
series. To avoid the cell degradation or hazards overcharging could be viewed as a controlled internal
related to overcharging, cells for battery packs must short, and the effectiveness of a certain redox additive
be handpicked to match a nominal capacity or each could be evaluated in terms of the maximum over-
individual cell within the pack must be regulated charge current it can shuttle.
externally with electronic devices, such as positive Narayanan et al. conducted a theoretical analysis
temperature coefficient (PTC) resistors or integrated on the cell parameters that determined this maxi-
circuits (IC). However, both remedies add to the mum shuttling current.416 By assuming that the mass
manufacturing cost and lead to a lower energy transport of the redox couple [R]/[O] is mainly real-
density of the cells. ized by means of diffusionswhich is reasonable
As an alternative, Abraham and co-workers pro- because the low concentration of [R]/[O] at the
posed an electrochemical mechanism built in the additive level makes the field-assisted migration
electrolyte that is able to shunt the excess charge negligiblesthey applied the finite linear diffusion
approximation to gain the diffusion-limited steady- position of electrolyte solvents but high enough to
state overcharge current i as ensure a full utilization of cathode active mass; (2)
{ ( )}
the electrochemical reversibility of the redox reac-
Ec - E0c tions should not degrade discernibly within the time
exp nF frame of cell life; (3) the additive should be inert
nFADC RT
{ ( )}
i) (12) before and after activation by overcharging (i.e., both
l Ec - E0c oxidized and reduced forms of the additive must not
1 + exp nF react with any electrolyte components).
RT
During the past two decades, the redox shuttle
where F, A, D, l, C, Ec, and E°c are the Faraday mechanism has been influencing researchers as the
constant, electrode area, diffusion coefficient for [R] most promising solution to the challenge of cathode
or [O], interelectrode spacing, total concentration of overcharge, and among the limited number of pub-
the additive ([R] + [O]), single electrode potential of lications, most of the additives were selected based
the cathode, and formal potential of the redox couple on their redox potentials.
[R]/[O], respectively. The cathode materials employed for the early
Obviously, when no overcharging exists (i.e., Ec - lithium-based systems were 3.0 V class oxides or
E°c , 0), the shuttle current is negligible, while the sulfides; thus, the redox potential for the additive
redox reaction remains not activated. should be located in the neighborhood of 3.2-3.5 V.
On the other hand, when overcharge does occur, Accordingly, the first generation redox additive pro-
eq 12 could be much simplified on the condition that posed by Abraham et al. was based on the iodine/
Ec - E°c . RT/nF, and eq 12 is transformed into iodide couple,415 which could be oxidatively activated
at the cathode surface at 3.20 V and then reduced at
nFADC
i) ) id (13) the lithium surface.414,415,417-420 For most of the ether-
l based solvents such as THF or DME that were used
The applied condition represents a relatively large at the time, the oxidation potential of iodide or
positive deviation of the single-electrode potential for triiodide occurred below that of their major decom-
a cathode from the oxidation potential of the redox positions, while the high diffusion coefficients of both
couple [R]/[O]. For a single-electron reaction at room iodine and iodide in these electrolyte systems (∼3 ×
temperature, the above criterion for the deviation Ec 10-6 cm2 s-1) offered rapid kinetics to shuttle the
- E°c corresponds to RT/nF ) 0.026 V, and one overcharge current. Similarly, bromides were also
would therefore expect the simplification that leads proposed.418-420 However, this class of halide-based
to eq 13 to hold true for most of the overcharge additives were deemed impractical due to the volatil-
situations encountered in practical applications. ity and reactivity of their oxidized forms (halogen).
Thus, eq 13 expresses the limiting overcharge The well-known redox compounds based on the
current that can be sustained by a particular choice metallocene family, which are much more amicable
of redox additive and electrochemical conditions, and than halide/halogen couples, were also first investi-
it suggests that this maximum diffusion current id gated by Abraham and co-workers as redox shut-
is independent of the cell potential or the formal tles.421,422-426 These organometallic complexes consist
potential of the redox couple but is solely determined of a metal coresusually Fe or Coscoordinated by two
by its concentration and diffusion coefficient in the cyclopentadieneyl ligands, as Table 11 shows, and
electrolyte. In the case that the overcharge current they are able to undergo reversible redox reactions
becomes too high for the redox couple to shuttle, as in the neighborhood of 3.2-3.5 V.421 The electron
would happen when a large terminal voltage is exchange occurs through the change of valence orders
mistakenly set during constant potential charge, any of the metal core; however, the formal potential of
excess current over the maximal limit specified by this redox process not only varies with the metal core
eq 13 would continue to delithiate the cathode but also is sensitive to the substitution on the
material and cause irreversible decompositions, and cyclopentadieneyl rings.424 The number as well as the
safety hazards would eventually be caused by this chemical nature of the substituents (electron-donat-
overflowing charge. ing or -withdrawing) could affect the potential by as
To maximize the current limit that could be shunted much as (0.30 V.427 Theoretically, this substitution-
by redox additives so that the occurrence of such sensitivity offers a convenient tool to tailor the redox
irreversible processes due to overflowing current potential of these metallocenes for various cathode
could be more efficiently suppressed, the redox ad- materials.422,425
ditive apparently should be present in the electrolyte Abraham et al. tested the performance of an
at high concentrations, and both its oxidized and n-butyl-substituted ferrocene in a lithium/TiS2 cell,
reduced forms should be very mobile species. Where where LiAsF6/THF or LiAsF6/2MeTHF was used as
the criteria for selecting potential redox additives are an electrolyte. They found that, under overcharge
concerned, these requirements can be translated into conditions, the voltage of the cell containing ferrocene
higher solubility in nonaqueous media and lower as an additive leveled off at 3.25 V, corresponding
molecular weight. In addition to solubility and dif- well to its redox potential of 3.18-3.50 V, while the
fusion coefficients, the following requirements should reference cell without additive was overcharged up
also be met by the potential redox additives: (1) the to 5.0 V, corresponding to the decomposition of
formal potential of the redox couple [R]/[O] should THF.422 No capacity loss was observed in successive
be lower than the onset potential for major decom- cyclings after the overcharge test, thus confirming
Table 11. Cathode Surface Layer Additives: Overcharge Protection
CID: current interrupter device activated by internal pressure.
that the additive based on ferrocene had little del- trolyte on a GC working electrode, where the peaks
eterious effect on the stability of the cell components. indicated as O2 and O3 represent the oxidation of
These promising results initiated a new round of cyclopentadieneyl rings and electrolyte solvents,
research activities based on structural modifications respectively, while R2 and R3 stand for the reductive
on the cyclopentadieneyl rings.425-429 decomposition of electrolyte solvents and the deposi-
In a systematical study, Golovin et al. investigated tion of lithium from solution. Obviously, on the anode
a series of metallocene derivatives in terms of their side, the limit was set by the lithium deposition, as
redox potentials, mass transport properties, and the reduction of solvents only occurs in the first
chemical and electrochemical stabilities in both elec- charging cycle in a lithium-based cell. The anodic
trochemical test cells and commercial-size AA re- limit, on the other hand, was imposed by the O2. The
chargeable cells.429 Figure 43 shows the complete redox potential of the shuttle agent ferrocene was
voltammetric scan of the ferrocene-containing elec- indicated by a pair of closely located peaks [R1]/[O1]
indefinitely. The successive cyclings of these over-

charged cells showed that no adverse effect was
caused by these ferrocene additives on cycle life.
Furthermore, the reversibility of the redox reactions
and the long-term stability of them with respect to
other cell components were also tested. The authors
had observed over a hundred “turnovers” of these
ferrocene shuttles in the cell while demonstrating a
reversible shuttle effect, and prolonged cyclings of
these cells showed excellent cycle life, indicating the
good compatibility between ferrocene additives and
the bulk electrolytes.
The practical importance of overcharge protection
by these ferrocene additives was further confirmed
with a battery pack in which cells with mismatched
capacity were intentionally connected in series and
subjected to overcharge.427 As expected, the added
ferrocene acted to prevent the cell with low capacity
from being overcharged as the other cell continued
Figure 43. Cyclic voltammograms of 0.08 M ferrocene in to charge. At the end of the charge, all cells had
1.0 M LiAsF6/EC/PC conducted at 20 mV s-1. (Reproduced
with permission from ref 429 (Figure 2). Copyright 1992
attained essentially the same state-of-charge, and
The Electrochemical Society.) successive discharge and recharge showed excellent
cycling characteristics, despite the mismatched ca-
pacities in the pack. If no redox additives were used,
such mismatch in capacity would result in poor cycle
life or even hazards.
The ability of these ferrocenes as carriers for
shuttling current through the cell was determined
by charging the cell with a constant current at the
fully charged state (3.5 V). When the interelectrode
spacing was between 25 and 50 µm, the limiting
shuttle current of these ferrocenes was found to be
as high as 2.0 mA cm-2.429
One adverse effect of these additives on cell per-
formance seemed to be related to their blocking of
ionic paths on the surface of cathode materials, as
indicated by the reduced power capabilities in the
presence of ferrocenes. Analysis on the concentration
Figure 44. Voltage profile of overcharged Li/LixMnO2 AA changes of ferrocene additives in the electrolyte
cells containing different substituted ferrocenes as redox solutions before and after their exposure to cathode
additives in 1.0 M LiAsF6/EC/PC: (A) reference; (B) fer- materials established that an adsorption of ferrocene
rocenyl ketone; (C) dimethylaminomethylferrocene; (D) species occurred on the cathode surface, 93% of which
ferrocene; (E) n-butylferrocene. (Reproduced with permis- would be covered when as low as 0.3 M ferrocenes
sion from ref 429 (Figure 6). Copyright 1992 The Electro-
chemical Society.) were present in the electrolyte solution.429 This
surface deactivation resulted in the loss of both rate
at ∼3.25 V, which occurred reversibly. With various capability and capacity.
substitutions on the cyclopentadieneyl rings (Table Redox shuttles based on aromatic species were also
11), the redox potential could be adjusted to occur in tested. Halpert et al. reported the use of tetracyano-
the range 3.09-3.55 V. For cathode materials based ethylene and tetramethylphenylenediamine as shuttle
on LixMnO2, whose potential at full charge is 3.50 V, additives to prevent overcharge in TiS2-based lithium
these substituted ferrocenes could make feasible cells and stated that the concept of these built-in
candidates as redox additives, since their redox overcharge prevention mechanisms was feasible.430
potentials can be adjusted to be between those of the Richardson and Ross investigated a series of substi-
major oxidative decomposition of electrolytes (>4.0 tuted aromatic or heterocyclic compounds as redox
V) and the cathode full utilization (3.50 V). shuttle additives (Table 11) for polymer electrolytes
The effect of these ferrocene-based additives on that operated on a Li2Mn4O9 cathode at elevated
overcharge protection is shown in Figure 44, where temperatures (85 °C).431 The redox potentials of these
AA cells based on lithium, LixMnO2, and electrolytes compounds ranged between 2.8 and 3.5 V, and like
with or without additives were overcharged. In the ferrocene-based additives, they are only suitable for
absence of these redox shuttles (A), the cell voltage cathode materials of low voltage.
continues to rise, indicating the occurrence of major Accompanying the commercialization of lithium ion
irreversible decompositions within the cell whereas technology, the emergence of 4.0 V class cathode
the presence of shuttle agents (B-E) locks the cell materials based on spinel, LiCoO2, and LiNiO2 pre-
potential in the vicinity of their redox potentials sents a more stringent requirement for the selection
of shuttle additives, since few redox couples could

undergo reversible turnover at such high potentials
near 4.0 V. Among those few were some organome-
tallic complex compounds with Fe, Ru, Ir, or Ce as
metal cores and phenanthroline or bipyridine as
ligands. Their redox potentials were found to be in
the vicinity of 4.0 V, and like ferrocenes, fine adjust-
ment in the redox potential could be realized by
varying the chemical nature, number, and relative
positions of the substituents on the aromatic rings.432
However, these complexes had very limited solubili-
ties (<50 mM) in nonaqueous media, which, in
combination with their large molecular weight and
size (hence, low diffusion coefficient), contributed to
a low shuttle current limit. Tests in a cathode half-
cell using LiPF6/PC/DMC showed that under over-
charge the hike in cell voltage was only delayed, but
not locked, at the level of the redox potential, appar-
ently due to the low limiting diffusion current that
these organometallic redox additives can carry.432
Inspired by earlier reports that some substituted
aromatic compounds can be reversibly reduced and
oxidized in the neighborhood of 4.0 V,433 Adachi et
al. focued their attention on a series of anisole-based
compounds, which have two methoxy substituents
and one halogen substituent on the benzene ring
(Table 11).432 The high solubility of this class of
organic compounds in carbonate-based electrolytes
makes them a family of more promising candidates
than the metal complexes. The cyclic voltammetry
of one of those potential redox additives, 4-fluoro-1,2-
dimethoxybenzene, was shown in Figure 45a. Of the
two electrochemical processes, as represented by the
two pairs of peaks located between 4.0 and 4.75 V,
the one on the positive side is similar to the charac-
teristic profile of a reversible redox couple as judged
by the separation of the anodic and cathodic peaks
as well as their shape. Similar behavior was also
observed for bromo-substituted counterparts with
higher shuttle voltage.
A correlation was found between the reversibility
of these redox couples and the relative positions of
the two methoxy groups. Among the various anisole
compounds investigated, only those with the two
methoxy groups at 1,2- (ortho-) and 1,4- (para-)
positions showed reversibility at the 4.0 V region,
while 1,3- (meta-) counterparts were oxidized ir-
reversibly, as indicated by the absence of the corre-
sponding reduction process in Figure 45b. The au-
thors proposed that the stabilization of the oxidized
product by methoxy substituents plays an important
role in determining whether the reaction is revers-
ible. Similar to the rationale for the aromatic elec-
trophilic substitution that is familiar to organic
chemists, the canonical structures of the possible Figure 45. Cyclic voltammetry of (a) 4-halo-1,2-dimethoxy-
oxidized states with the largest contribution are used benzenes (Reproduced with permission from ref 432 (Figure
8). Copyright 1999 The Electrochemical Society.) and (b)
to explain this stabilization effect by substitutes, as 1,3-dimethoxybenzenes in LiPF6/PC/DMC on Pt (Repro-
shown in Scheme 28. duced with permission from ref 432 (Figure 9). Copyright
It was assumed that, during an oxidation of these 1999 The Electrochemical Society.)
aromatic moieties, an electron would be removed
from the aromatic ring at the location where the other hand, the stability of this cationic product relies
electron-donating methoxy is positioned. Whether on the effectiveness of the delocalization of the formal
this oxidation is reversible critically depends on the charge. Apparently, the lone pair of electrons on the
stability of the produced cationic product. On the methoxy oxygen can effectively contribute to this
Scheme 28. Effect of Methoxy Relative Position on the Stabilization of the Oxidized State of Anisole Redox
Additives: (a) 1,2-Methoxybenzene, Whose Oxidized Product Was Stabilized by the Neighboring Methoxy;
(b) 1,3-Methoxybenzene, in Which the Meta Methoxy Fails To Stabilize the Dianion
delocalization if the methoxy is at either ortho- benzene as soon as the cell voltage leveled off because
(Scheme 28a) or para-positions (not shown in Scheme redox shuttling started. This suggests that the cur-
28), while the methoxy at the meta-position (Scheme rent supplied over 4.3 V (or after 60 h) is neither
28b) cannot. In this sense, the unstable oxidized stored in the cathode nor consumed by irreversible
product from the meta-substituted compounds would decompositions. Rather, the shuttling redox couples
act like an intermediate and readily undergo further convert this part of the current into heat during their
side reactions, thus rendering the original oxidation repeated movement across the cell. In other words,
irreversible. The position of the halogen does not during the course of shuttling, no work is done by
seem to be important, as long as the two methoxy the current to change the state-of-charge in the
groups are either para or ortho to each other; cathode, and all of the current supplied is dissipated
however, its presence is critical, since 1,4-dimethoxy- by the diffusion of [O] and [R] species of the redox
benzene itself shows no reversible redox behavior. additive.
Thus, the role of the halogen in affecting the stability
of the cationic species remains unclear. Nevertheless,
the empirical structure-property relationship dis-
covered by the authors for the anisole family could
serve as a useful guideline for the future selection of
redox additives.
Overcharge tests were carried out in LiCoO2 cath-
ode half-cells that contained these additives, and a
redox shuttle effect was observed between 4.20 and
4.30 V, close to the redox potentials of these additives.
The same shuttling effect was observed even after 2
months of storage for these cells, indicating the
stability and redox reversibility of these additives.
A closer examination of the capacity retention re-
vealed that 4-bromo-1,2-dimethoxybenzene seemed
to have the best shuttle-voltage performance for the
4.0 V lithium cell used.432 The stability of these
additives against reductive decomposition was also Figure 46. Correlation between the voltage profiles and
tested by the authors on metallic lithium as well as heat flow of a cell under overcharge: cells with 100 mM
4-bromo-1,2-dimethoxybenzene (solid lines) and reference
on carbonaceous anodes, and no deterioration was cell (dashed lines) (Reproduced with permission from ref
detected. 432 (Figure 17). Copyright 1999 The Electrochemical
A convincing analysis carried out in this work Society.)
established a direct connection between the shuttle
current and the thermal effect in the cell, and it helps Following Adachi et al., aromatic compounds with
us to understand energetically the mechanism by similar functionalities were proposed for polymer
which redox additives worked as shuttle carriers. As electrolytes as redox shuttle additives, which in-
Figure 46 shows, calorimetric measurement was cluded bipyridyl and biphenyl carbonates and di-
performed on cells that were being overcharged fluoroanisoles.48b,434 All these additives could protect
simultaneously, while the voltage profiles and heat the cathode from overcharging in the vicinity of 4.1
flows were plotted together to reveal the correlation. V.434
In the reference cell, where no additive was present, Other less prominent types of additives, also in-
there was no heat flow until the very end of the tended for overcharge protection, were termed “shut-
charge, where electrochemical decomposition of the down additives” in the battery industry based on
electrolyte solvents and the thermal reaction between their tendency at high potentials to release gas,
cathode and electrolyte would occur. On the other which in turn would activate a current interrupter
hand, the obvious thermal effect could be detected device (CID), or to polymerize and block the ion
for the test cell containing 4-bromo-1,2-dimethoxy- passage in the electrolyte. The former included such
compounds as pyrocarbonates,48b biphenyl,48b,435,436 (4) wettability toward porous separators as well as

cyclohexylbenzene,48b and phenyl-tert-butyl carbon- electrode materials, (5) relatively low toxicity, and
ate,437 while the latter include biphenyl and other (6) relatively low cost.
substituted aromatic compounds.438,439 Related pub- This section reviews these research efforts in the
lications on these additives have been absent except past decade on developing new solvents and lithium
for patent disclosures and conference presentations. salts for nonaqueous electrolytes of lithium ion cells,
However, according to Yoshino et al., some of these but the cosolvents or additives developed for non-
additives have been used in commercial lithium ion flammable electrolytes, most of which are phosphorus
cells for several years.48b,440 The major difference or fluorinated molecules, are not included, since their
between these additives and the redox shuttles is presence is intended for improvement in safety rather
that, once activated by the cathode potential, the than performance. They will be reviewed in section
gassing or polymerizing of these additives terminates 8.5.
the cell permanently, while the operation of redox With few exceptions, these new electrolyte solvents
additives is nondestructive, at least theoretically. focus on possible improvements in low-temperature
Since the redox additives have a maximal current performance, while new salts are intended to offer
limit defined by their diffusion coefficient and con- higher thermal stability. This divided directions of
centration, the presence of the gassing- and polymer- pursuit after the targeted improvements is appar-
izing-type additives would serve as a more reliable ently created by the fact that solvent and salt,
line of defense against catastrophic failure from respectively, impose the upper and lower tempera-
overcharging. Therefore, the integration of multiple ture limits of the current state-of-the-art electrolytes.
additives of different types into a single electrolyte
seems to be a feasible approach on the condition that 8.3.1. Nonaqueous Solvents
the destructive additives should have higher activa-
tion potential. The state-of-the-art electrolytes use mixtures of
cyclic and acyclic carbonates as solvents, whose
functions are to solvate lithium salts and to facilitate
8.3. New Electrolyte Components lithium ion transport, respectively. The key cyclic
The modification of electrolytes via additives is solvent, EC, is also responsible for forming a protec-
attractive to industry as an economical approach; tive SEI on graphitic anodes and probably a similar
however, its impact on electrolyte performance is surface layer on metal oxide cathodes. However, this
mainly restricted to tuning interfacial-related prop- indispensable solvent simultaneously sets the narrow
erties because of their small concentration in the range of service temperature for these electrolytes
electrolyte, while other challenges for the state-of- with its high melting point, while its replacement by
the-art electrolytes such as temperature limits, ion other low-melting solvents such as its structural
conductivity, and inflammability are still determined analogue PC is often rendered difficult by the re-
by the physical properties of the bulk components. quirement for SEI-forming ability. The attempts at
Improvements in these bulk-related properties can solving this dilemma have been directed at the
only be realized by replacing the bulk components of structural modification of either EC or PC so that a
the electrolytes with new solvents and salts, but such balance between low melting point and favorable
efforts have been met with difficulty, since more often interfacial chemistry could be reached.
than not the improvement in the individually tar- On the other hand, the linear carbonates used in
geted properties is achieved at the expense of other the state-of-the art electrolytes, DMC, DEC, and
properties that are also of vital importance to the EMC, serve as diluents to the high-melting and
performance of electrolytes. Such “collateral damage” viscous EC. They have been known to be unsuitable
undermines the significance of the improvements as single solvents because of their inability to solvate
achieved and, in some cases, even renders the entire lithium salts as well as their instability on the
effort unworthy. oxidizing surface of cathode materials, while the
Nevertheless, research activities in this arena gassing of lithium ion cells during long-term cycling
continue to be driven by the potential commercial is also believed to arise from them. However, any
interest that might arise from any possible replace- possible replacements for linear carbonates to serve
ment of the state-of-the-art electrolyte components. as cosolvents with EC or PC should at least possess
Realizing that the probability of success is rather the major prerequisites of lower viscosity and lower
limited with any radical change of the entire elec- melting point.
trolyte system, an increasing number of researchers Preferably, the new solvents are also expected to
on novel electrolytes are adopting the current state- possess better stability or ability in interfacial chem-
of-the-art electrolytes as the platforms for innovation istry on both anode and cathode materials so that
and attempting to approach the targeted improve- the new electrolyte formulation can rely less on EC;
ments without serious sacrifices in the well-estab- or they are expected to be less inflammable, as a
lished merits, which at least should include (1) facile major shortcoming of the linear carbonates is their
ion transport as characterized by ion conductivities low flash points (Tf) (Table 1). In the search for new
above 5 mS cm-1 at room temperature, (2) electro- solvents, fluorination has been adopted as a favorable
chemical stability on both carbonaceous anode and approach to achieve improvements in these two
metal oxide cathode materials in the range 0-5 V, aspects because the presence of C-F bonds in organic
(3) inertness to other cell parts such as packaging molecules is found to affect interfacial chemistry on
materials and anode and cathode current collectors, carbonaceous anodes in a positive manner,441-443 and
Table 12. Novel Nonaqueous Solvents and Their Major Properties
many organofluorine compounds act as flame-retar- Coulombic efficiency in the first charging cycle com-
dants.444,445 Table 12 summarizes the novel electro- parable to that of the commercial electrolytes for
lyte solvents according to their structural cyclicity. lithium ion cells.446,447 The potential plateau at 0.80
8.3.1.1. Cyclic Solvents. Halogen substitution on V, characteristic of the reductive decomposition of PC,
the carbonate ring of EC and PC is postulated to was completely eliminated due to the presence of
serve the dual purpose of lowering melting temper- ClEC, while a new process was observed at 1.70 V.
ature by breaking the molecular symmetry and When taking the irreversible capacity in the first
improving the SEI-forming ability. Shu et al. used cycle as a metric, the optimum concentration of ClEC
chloroethylene carbonate (ClEC) as a cosolvent for was determined to be 30 vol %,448 although in a
PC and found that an effective and protective SEI ternary solvent system containing EC, its concentra-
could be formed on a graphitic anode, with the tion could be minimized to 5%.447 Further electro-
chemical studies confirmed that ClEC forms a com- achieved by these cyclic-only electrolytes is 6.6 mS
pact and uniform SEI on the surface of the graphitic cm-1 (for LiPF6 in TFPC/EC 1:3) and in neat TFPC
anode that can prevent the cointercalation of PC and it is only ∼3.0 mS cm-1. On the other hand, EIS
the subsequent exfoliation of graphene even in PC- studies carried on anode half-cells found that the
rich electrolytes,449,450,451 and the origin of this effec- ClEC/TFPC couple forms the least resistive SEI film.
tive interfacial chemistry seems to be the reductive While the electrochemical behavior of neat TFPC was
decomposition of ClEC at 1.70 V, which generates not described, the cathodic stability of the electrolytes
CO2 as a major intermediate, as indicated by the based on mixed solvents, TFPC/ClEC and TFPC/EC,
conspicuous absorption at 2341 cm-1 in in situ FT- was tested with a graphitic anode, and no reductive
IR spectra.449 The fate of CO2 is yet unclear, but a processes characteristic of TFPC were visible, in
strong possibility is its further reduction into lithium addition to those of ClEC (1.70 V) and EC (0.60 V).
alkyl carbonate as a major component of the new SEI. Similar to the cases of ClEC and FEC, its presence
Since this process occurs at a potential far above that did prevent PC cointercalation, but the irreversible
of PC cointercalation, graphitic materials could suc- capacity associated with PC decomposition still ac-
cessfully be used in ClEC/PC mixed solvents. Natu- counted for ∼132 mA h g-1 of the irreversible capacity
rally, such a formulation without EC should offer in the anode half-cell, which was ∼40% of the total
much better performance at subzero temperatures. charge. Spectroscopic means, including XPS and FT-
The tests in full lithium ion cells, however, yielded IR, were used to analyze the SEI chemistry on the
less than satisfactory results.447 Despite the stable anode, and the presence of a C-F bond was identi-
cell life observed during the test of >800 cycles, only fied, indicating the decomposition products of TFPC
92% Coulombic efficiency was achieved, as compared do constitute part of the SEI. Cycling performance
with the 98+% Coulombic efficiency for the corre- of these electrolytes was tested in anode and cathode
sponding anode half-cells. Initial suspicion about the half-cells, but no data on full lithium ion cells were
anodic instability of ClEC was excluded, because this shown.453
less-than-ideal efficiency did not lead to obvious Among the limited choices of candidates intended
capacity fading, and cyclic voltammetry carried at to replace cyclic carbonates, γ-butyrolactone (γBL)
high potentials up to 4.2 V showed no discernible seemed to be the most promising and therefore
oxidation of ClEC, suggesting that ClEC is at least closest to practical application. γBL has long been
as stable as EC on the cathode surface. Therefore, considered a possible component for electrolytes in
the above loss in charge efficiency should come from lithium/lithium ion cells due to its moderately high
the combined presence of both anode and cathode dielectric constant, relatively low viscosity, EC-like
materials. Shu et al. thus proposed a shuttle-like structure, and excellent solvating ability.454-456 Its
mechanism in which ClEC was reductively cleaved reductive behavior has been studied on inert elec-
on a carbonaceous anode, yielding the organic prod- trode surfaces,208,209 lithium surfaces,222,457 and graph-
ucts related to alkyl carbonates and LiCl. Since LiCl ite surfaces.123 Unfortunately, in the early era of
is weakly soluble in nonaqueous media, chloride lithium ion technology, Aurbach and co-workers had
would diffuse to the cathode surface and be oxidized found that γBL/LiAsF6 solution failed to perform
into Cl2. This internal chemical shuttling process satisfactorily in lithium ion cells based on the graph-
between the anode and cathode surfaces consumes ite anode unless under an atmosphere of CO2, thus
charges, and its reversibility accounts for the con- diminishing its possible application.123 Detailed mecha-
stantly low Coulombic efficiency during long-term nistic studies attributed the failure to metallic lithium
cycling tests. This internal self-discharge mechanism deposition on the carbonaceous anode surface and the
is obviously undesirable and eventually rendered subsequent reaction with γBL, which was made
ClEC an impractical cosolvent. possible by the high resistance of the SEI formed in
Considering the virtually zero solubility of LiF in γBL-based electrolytes.177
nonaqueous media, McMillan et al. synthesized the The publications related to the use of γBL as an
fluorinated counterpart of ClEC.442 As expected, the electrolyte solvent for lithium ion cells have been
shuttle phenomenon was eliminated because of the scarce until recently, when a Japanese group from
fluorination, as evidenced by the quantitative Cou- Toshiba reported the electrolyte based on a solution
lombic efficiency, while a similar SEI effect was of LiBF4 in γBL/EC mixtures for laminated thin
maintained, since FEC/PC mixed solvent can support lithium ion cells.133,458 This new formulation, when
reversible lithiation/delithiation of the graphitic an- used in combination with a graphitic carbon fiber,
ode materials. However, capacity was observed to exhibited stable cell performance with 94% Coulom-
fade by 37% in 200 cycles. Follow-up work on this bic efficiency in the first cycle. Most importantly, it
solvent has been minimal.452 afforded much higher thermal stability than the
Similar structural modification was also performed electrolytes based on LiPF6 solutions, since storage
on PC. Trifluoropropylene carbonate (TFPC) was of the cell at full state-of-charge (4.20 V) at 85 °C only
synthesized in the hope that a novel electrolyte free led to negligible capacity loss. The low-temperature
of linear carbonates could be formulated with im- performance was also much improved as compared
proved safety in the case of fire.443,453 Like ClEC and to the cases of state-of-the-art electrolytes, with 88%
FEC, it is liquid at room temperature with a high of the nominal capacity retained at -20 °C at the
flash point (134 °C), but its high viscosity results in drain rate of 0.5 C. Since the new formulation is free
slower ion transport within the electrolytes, because of linear carbonates, its tolerance against thermal
the maximum ion conductivity at room temperature abuse should be superior to that of the current
with cyclic components including EC and PC. Binary,

ternary, and even quaternary compositions were
formulated with these esters, and improvements of
low-temperature performance were described.406,461-463
According to Herreyre et al., ternary compositions
based on ethyl acetate (EA) and methyl butyrate
(MB) demonstrated electrochemical stability up to
4.85 V on the cathode surface at 60 °C, while the
compositions containing lower molecular weight meth-
yl acetate (MA) deteriorated.406 Similar conclusions
were drawn by Smart et al., who investigated a series
of linear esters and carbonates as low-temperature
electrolyte components and found that a desirable
SEI (i.e., less resistive and protective) could only be
Figure 47. Voltage profile of graphitic anode in LiPF6 and formed when esters of higher molecular weight were
LiBF4 solutions in γBL/EC (1:1). (Reproduced with permis- used, although esters of lower molecular weight could
sion from ref 460 (Figure 1). Copyright 2003 The Electro- afford higher bulk ion conductivity at low tempera-
chemical Society.) tures because of their lower viscosities.462 Judged on
the basis of the reversible capacity in the first charge
commercial lithium ion cells. Preliminary results
cycle achieved by a graphite anode in quaternary
showed that the gassing of this new electrolyte is
electrolytes, LiPF6/EC/DEC/DMC/ester, the esters
much less severe, and an independent study employ-
with longer alkyl chains were obviously superior to
ing the in situ DEMS technique confirmed that the
the ones of lower molecular weight:
dominant participation of γBL in SEI formation
reduced the gas evolution by forming a γ-alkoxy-β-
keto ester:459 ethyl propionate (EP, 340.75 mA h g-1) >
ethyl butyrate (EB, 309.46 mA h g-1) >
MA (236.5 mA h g-1) > EA (214.2 mA h g-1)
The above cell performance order is almost the

reversal of that for the low-temperature ion conduc-
The spectroscopic identification of the β-keto ester tivity, which clearly shows the decoupling of inter-
and the reduction mechanism leading to it have been facial properties from bulk properties. Thus, Smart
previously described by Aurbach and co-workers et al. concluded that the nature of the SEI instead of
when studying the surface chemistry of γBL-based bulk ion conductivity plays the key role when new
electrolytes on various electrodes polarized to low solvents are selected.
potentials.123,208,209 To improve the thermal stability of electrolytes
Interestingly, all the above improvements seemed against lithium metal electrodes, various fluorinated
to be unexpectedly achieved from components (LiBF4 organic solvents were tested as electrolyte solvents
and γBL) that were already tested individually and (Table 12).441,464-466 For example, Yamaki et al.
considered inferior in performance to the commonly introduced a series of partially fluorinated esters in
used LiPF6 and various carbonates. According to a which the fluorination was conducted on the carboxy-
more recent publication, this unique match between lic acid section and used them as single solvents for
LiBF4 and γBL originates from some interactions electrolytes,465 while Smart et al. synthesized a series
between the two that are not yet understood.460 For of partially fluorinated carbonates for similar pur-
example, among the five salts investigated that poses.466 When compared with their nonfluorinated
included LiBF4, LiPF6, LiAsF6, LiIm, and LiClO4, counterparts, the presence of fluorination in these
only LiBF4 in γBL/EC supported reversible lithium molecules was reported to result in lower melting
intercalation/deintercalation with capacities close to points, higher anodic stability, increased safety, and
the theoretical value, as Figure 47 shows. Other salts favorable SEI-forming characteristics on metallic
seemed to suffer from a highly resistive surface film lithium or graphitic anodes. For fluorinated carbon-
on the graphite surface, which worsened with cycling ates a number of electrochemical techniques such as
number. Although extensive characterizations were Tafel polarization and micropolarization confirmed
carried out with techniques such as XRD, SEM, and that facile kinetics of lithium ion intercalation were
XPS, the nature of this interaction remains unclear. made possible because of the new SEI chemistry
8.3.1.2. Linear Solvents. For novel solvents that imparted by those fluorinated solvents. In terms of
are intended to replace linear carbonates, the em- interfacial resistance and utilized capacity, fluori-
phasis was placed on such properties as melting nated carbonates are generally superior to the flu-
temperature, viscosity, and inflammability of the orinated carbamates studied (Table 12).466
solvents. As Table 12 shows, these solvents can be 8.3.1.3. Other Novel Solvents. In addition to
roughly classified into four groups: (1) linear esters molecules based on carboxylic and carbonic esters,
or carbonates,406,461-463 (2) fluorinated esters,464,465 (3) other types of organic polar compounds were also
fluorinated carbonates,466 and (4) fluorinated ethers.441 explored as possible solvent candidates but with
Like linear carbonates, none of these linear esters limited success. Zhang and Angell synthesized a
were used as single solvents but rather as cosolvents dimeric glycol ester of boric ester (BEG, Table 12) in
the hope that the electron-deficient boron of the

molecule would assist in dissociating lithium salts
and coordinating with anions.467 The solvent, despite
its high viscosity, can dissolve various lithium salts
to appreciable concentrations, and it exhibited par-
ticular stabilization of metallic lithium at elevated
temperatures up to 100 °C. With EC as cosolvent,
the electrolytic solution of LiClO4 and LiIm showed
moderate ion conductivities on the level of 10-3 S
cm-1. The anodic stability of the electrolytes based
on this borate ester was measured on a Pt electrode,
and a wide electrochemical stability window of ∼5.8
V was reported. The cathodic stability of the borate
ester on graphitic anode material was also confirmed
by cycling an MCMB electrode in an electrolyte
LiClO4 in BEG/EC, which yielded stable cycles.
Considering the high viscosity of both solvent com-
ponents used in the study, the above cell test results
appear to be promising, but better results could be Figure 48. Anodic stability as measured on a spinel Lix-
achieved if conventional diluents such as linear Mn2O4 cathode surface for EMS-based electrolytes: (a)
carbonate were used. However, no further work has LiIm; (b) LiClO4; (c) LiTf. In all cases, 1.0 m lithium salt
solutions were used, and slow scan voltammetry was
been reported on this interesting solvent. conducted at 0.1 mV s-1, with lithium as counter and
Choquette et al. investigated the possibilities of reference electrodes and spinel LixMn2O4 as working
using a series of substituted sulfamides as possible electrode. (Reproduced with permission from ref 75 (Figure
electrolyte solvents (Table 12).468 These compounds 3). Copyright 1998 The Electrochemical Society.)
are polar but viscous liquids at ambient temperature, room temperature has been measured for LiPF6 in a
with viscosities and dielectric constants ranging sulfolane/DME mixture, which should obviously be
between 3 and 5 mPa s and 30 and 60, respectively, attributed to the combined effect of the high dielectric
depending on the alkyl substituents on amide nitro- constant of sulfolane and the low viscosity of DME.
gens. The ion conductivities that could be achieved On the other hand, the higher melting temperatures
from the neat solutions of LiIm in these sulfamides of the linear counterparts of sulfolane excluded them
are similar to that for BEG, that is, in the vicinity of from being considered as room-temperature electro-
10-3 S cm-1. Like BEG, it should be suitable as a lyte solvents. For example, the lithium cell based on
polar cosolvent used in a mixed solvent system, the simplest member of the linear sulfone family,
though the less-than-satisfactory anodic stability of dimethyl sulfone (DMS, mp 110 °C), has to operate
the sulfamide family might become a drawback that at 150 °C, although, surprisingly, high Coulombic
prevents their application as electrolyte solvents, efficiency and stable performance could be achieved
because usually the polar components in an electro- even at that high temperature.471
lyte system are responsible for the stabilization of With the introduction of structural asymmetry, Xu
the cathode material surface. As measured on a GC and Angell attempted to reduce the lattice energy of
electrode, the oxidative decomposition of these com- the sulfone molecules so that the melting tempera-
pounds occurs around 4.3-4.6 V when 100 µA cm-2 ture of linear sulfones could be lowered to the level
was used as the cutoff criterion, far below that for that allowed room-temperature application.75 They
cyclic carbonate-based solvents. synthesized five linear sulfones with asymmetric
Another class of organic compounds that could alkyl substituents and found that all of these com-
make promising electrolyte solvents for lithium-based pounds were liquids or solids that melt in the vicinity
rechargeable cells is based on alkyl sulfones. Since of room temperature. They were able to dissolve
most of these compounds are solids with melting various lithium salts up to 3.0 M and yield moderate
points above ambient temperature, interest was ion conductivity, and the most interesting solvent
rarely invested in them when possible candidates for candidate from the family was the simplest member,
electrolyte solvents are considered. One exception is ethylmethyl sulfone (EMS), which has an extraordi-
the cyclic member, tetramethylene sulfone (also narily high dielectric constant (95 at 20 °C, as
known as sulfolane), whose melting temperature is compared with 89 of EC).314 When tested on the
close to room temperature (27 °C) and whose dielec- surface of a spinel-based composite cathode material,
tric constant is close to that of PC (60 at 25 °C). The this solvent showed an electrochemical stability
electrolyte based on the salt solutions of this cyclic window of approximately 5.80 V, as Figure 48 shows.
sulfone remains liquid till subzero temperatures This high anodic stability limit, which remains the
because of the freezing-depression effect of the solute. highest ever known on the surface of a composite
Matsuda et al.469 and Morita et al.470 have individu- cathode material (versus an inert electrode), is sig-
ally tested it as a cosolvent for lithium cells, and focus nificant, since it would allow the various applications
was placed on the lithium cycling efficiencies. Gener- of high potential requirements such as a energy
ally decent ion conductivity could be achieved either storage device with 5.0 V class cathode materials or
with sulfolane as a single solvent or with mixed electrolytic and electroanalytic devices that need an
solvents based on it. For example, 15 mS cm-1 at inert supporting medium. The usefulness of this
Figure 50. Galvonostatic cycling of anode (Li/graphite)

Figure 49. Effect of cosolvent FPMS on the anodic and cathode (Li/LixMn2O4) half-cells using 1.0 m LiPF6 in
stability of the mixed solvents. Also shown for comparison FPMS/EMC 1:1 and 1:2 mixture solvents, respectively.
are the neat linear carbonates. In all cases, 1.0 m LiPF6 i ) 0.001 mA cm-2. (Reproduced with permission from ref
solutions were used, and slow scan voltammetry was 314 (Figure 8). Copyright 2002 The Electrochemical Soci-
conducted at 0.1 mV s-1, with lithium as counter and ety.)
reference electrodes and spinel LixMn2O4 as working
electrode. (Reproduced with permission from ref 314 (Fig-
ure 6). Copyright 2002 The Electrochemical Society.) candidate for the lithium ion cell if the manufactur-
ing cost of it could be reduced.
anodically stable solvent has been confirmed by the 8.3.2. Lithium Salts
various applications that followed the report of Xu The pursuit of new lithium salts has been driven
and Angell.316,317,472,473 For example, Seel and Dahn by the thermal instability of the current state-of-the-
have successfully used EMS as the nonaqueous art lithium salts that were based on perfluorinated
electrolyte for an anion intercalation cell, which anions, and the thermal as well as chemical inertness
enables the otherwise impossible staging of graphene has been taken as the main metric to evaluate any
layers with PF6- anions at 5.60 V;316 and the elec- potential candidates to replace those salts, although,
trolyte based on the same sulfone also supported the similar to the new solvent efforts, often the improve-
complete delithiation of a new cathode material at ment in these properties comes at the price of other
∼5.4 V.317 properties that are equally important for the opera-
Unfortunately, EMS cannot form an effective SEI tion of a lithium ion cell.
on graphitic anode materials, thus undermining the 8.3.2.1. Lithium Tris(trifluoromethanesulfon-
possibility of its use in lithium ion cells. In their yl)methide (LiMe). Following the development of
further work, Xu and Angell partially fluorinated the LiIm, Dominey invented a new lithium salt based on
alkyl of an asymmetric linear sulfone, inspired by the a carbanion that is stabilized by three perfluorinated
reports that fluorinated alkyl could improve the SEI methanesulfonyl groups.474 Because of the effective
chemistry on a graphitic anode.441-443 The fluorinated delocalization of the formal charge on the anion, the
sulfone, 3,3,3-trifluoropropylmethyl sulfone (FPMS), salt LiMe (Table 13) could be dissolved in various
has a melting point at 56 °C and could only be used nonaqueous media and showed better ion conductivi-
as cosolvent with other diluents such as DMC and ties than LiIm.146 Its stability at high temperature
EMC.314 The anodic stability of sulfone compounds is confirmed by TGA studies, which detected no sign
seemed to be maintained, despite the presence of of decomposition before 340 °C, while accordingly the
fluorination and the mixing with linear carbonates, electrolyte solution based on the salt remained stable
as evidenced by the oxidative limits at ∼5.70 V at 100 °C. The electrochemical stability of the salt
observed in Figure 49 for the mixed solvents. The was studied in THF solution, and the cyclic voltam-
SEI-forming ability, however, was indeed improved metry conducted on GC showed major anodic decom-
due to the fluorination, as shown by Figure 50, position process at ∼4.0 V. Although, on the basis of
wherein the electrolytes based on these mixed sol- the previously published data this decomposition
vents supported reversible lithium ion intercalation/ seemed to be caused by the solvent THF rather than
deintercalation of the graphite anodes. The cycling by the salt anion,64,74,177 the authors did not report
tests in the longer term confirmed that the surface further electrochemical measurements in other more
of a graphitic anode was well protected because close- stable solvents.146 Initially the salt was reported to
to-unity Coulombic efficiency was obtained, and the be inert toward an Al current collector,116,153 but a
authors suggested that FPMS might be a promising more detailed study later found that corrosion of Al
Table 13. Novel Lithium Salts and Their Major Properties
a Thermal decomposition temperature determined by TGA. b Anodic stability limit determined by cyclic voltammetry. c Stainless
steel working electrode. d Number of fluorine substituents on the aromatic ring.
does occur in various LiMe-based electrolytes at seems unlikely because Al is universally adopted as
potentials above 4.5 V, though somehow it is less the current collector for the cathode materials.
severe than that for the case of LiIm.147 Therefore, 8.3.2.2. Lithium Borates with Aromatic Li-
its application in high-potential lithium ion cells gands. A new class of lithium salts was developed
by Barthel et al. in the mid 1990s based on a borate and act to extend the potential range of application
anion chelated by various aromatic ligands.475-480 for these salts remains to be tested. Al was reported
Table 13 summarizes the representative members to be stable in the electrolytes based on at least two
from this class along with their selected physical and of these salts.479,480
electrochemical properties. The authors described the The cycling test for one of the salts was conducted
synthesis and chemical or physical characterizations by Handa et al. in lithium cells using a low-potential
of these salts in detail but only provided limited cathode material, V2O5.482 OCV and initial discharg-
electrochemical data. Generally speaking, this class ing behavior similar to those of other nonaqueous
of salts is rather stable thermally and decomposes electrolytes were reported, although no data concern-
only at very high temperatures without melting, ing extended performance were given. The key prop-
although moisture can still decompose them through erty that would decide whether these salts could be
hydrolysis.481 The solubilities of most of these salts used in lithium ion technology (i.e., the ability of
in nonaqueous media seemed to be dependent on the forming a protective SEI on the surface of graphitic
substitution of aromatic ligands, and moderate to materials) has not been reported for any of these
good ion conductivities have been reported, ranging salts.
from 0.6 to 11.1 mS cm-1 depending on the solvents
used. 8.3.2.3. Lithium Borates with Nonaromatic
The stabilities of these salts against oxidation were Ligands. The presence of aromatic ligands in
studied with cyclic voltammetry on various inert Barthel’s salts was believed to be responsible for the
electrodes, and an interesting correlation was estab- high melting points and basicity of the borate anions,
lished between the number of electron-withdrawing which in turn translate into moderate or poor solu-
substituents on aromatic rings and the anodic stabilities and ion conductivities as well as low anodic
bilities (i.e., higher oxidation potentials were found stabilities. To avoid use of these bulky aromatic
for the better stabilized borate anions with more substituents, Xu and Angell synthesized a series of
electron-withdrawing groups). This correlation could borate anions that are chelated by various alkyl-
be well explained by the order of the HOMO energy based bidentate ligands, which serve as electron-
levels obtained by quantum chemical calculations.477,479 withdrawing moieties by the presence of fluorine or
Thus, the anodic decomposition potential ranges from carbonyl functionalities.113,483,484 Table 13 lists the
3.6 V for the unsubstituted borate475 to 4.60 V for the selected members of this aromatic-free borate family.
borate with fluorinated and sulfonated aromatic Compared with their aromatic counterparts, these
ligands.479 A similar relationship was reported by novel salts showed much higher ion conductivity and
Sasaki et al., who listed the following order of anodic stability, while maintaining comparable ther-
oxidation potential limits according to the voltam- mal stability. Detailed studies of ionics indicated that
metry results obtained on a Pt electrode, and re- these salts could well dissociate in the media of
vealed the obvious dependence of the anodic stability moderate dielectric constants and yield ion conduc-
of these anions on the electron density of the aromatic tivities slightly lower than those for state-of-the-art
rings:481 electrolyte solvents.485,486 As an example, in EC/DMC
solutions of lithium bis(oxalato)borate (LiBOB) and
LiPF6, the ion conductivities are 7.5 and 10 mS cm-1,
respectively. For at least one of these salts, the
lithium ion transport number seemed to be higher
than 0.50 because of the large anions size.485,486 The
dependence of ion conductivity on salt concentration
is also different from the familiar bell-shaped de-
pendences observed for LiPF6- or LiBF4-based solu-
tions: the isothermal ion conductivity of these lithium
borate solutions remains almost independent of salt
concentration in the range 0.5-1.0 M, which could
be advantageous for practical applications.
Among these new borates, particular attention
should be paid to a salt based on oxalato ligands,
which has aroused intense interest recently in the
lithium ion research and development community.
In addition to the above thermodynamic consider- This salt was invented by Lischka et al.487 and
ation, kinetics also play an important role in deter- independently synthesized and investigated by Xu
mining the anodic stability of these salts. For ex- and Angell, who also gave it the popular name
ample, some salts whose decomposition products are LiBOB. Following these extensive physical charac-
polymeric moieties were found to passivate the terizations, a rather comprehensive electrochemical
electrode surface effectively.478 Therefore, although evaluation was conducted on this salt by Xu et
the intrinsic oxidation potentials for these anions al.,155,324,488,489 who found that the solutions of LiBOB
were not as high (∼4.0 V), they showed stability up in mixed carbonate solvents met the complete set of
to 4.50 V in subsequent scans. It should be cautioned stringent requirements for electrolyte solute intended
here, though, as the passivation was only observed for lithium ion cell applications: (1) it is anodically
on an inert electrode surface, whether similar pas- stable on the surface of composite cathode materials
sivations would occur on an actual cathode surface up to 4.3 V, (2) it can form a protective SEI on the
Figure 53. Stabilization of graphite in PC by LiBOB.

Voltage profiles of lithium/graphite half-cells containing 1.0
m lithium salts in neat PC as electrolytes. Only for LiBOB/
PC was the complete lithiation/delithiation cycle achieved.
Figure 51. Cathodic and anodic stability of LiBOB-based (Reproduced with permission from ref 324 (Figure 1).
electrolytes on metal oxide cathode and graphitic anode Copyright 2002 The Electrochemical Society.)
materials: Slow scan cyclic voltammetry of these electrode
materials in LiBOB/EC/EMC electrolyte. The scan number even superior to those of LiBF4, LiClO4, and LiPF6.
and Coulombic efficiency (CE) for each scan are indicated Especially, compared with the industry standard,
in the graph. (Reproduced with permission from ref 155 LiPF6, this salt also offers the additional advantages
(Figure 2). Copyright 2002 The Electrochemical Society.) of being thermally stable, being economical in terms
of manufacturing cost, being more environmentally
friendly by decomposing into less corrosive products
in the presence of moisture. The combination of these
merits has not been found in other novel salts thus
far investigated and has certainly made LiBOB a
hopeful contender for lithium ion technology applica-
tions. The stable performance of LiBOB-based elec-
trolytes was confirmed by the extended cycling of the
full lithium ion cells using such electrolytes, where
no capacity fading was detected during the operation
of ∼200 cycles.155
An unexpected but certainly welcomed discovery
on LiBOB is its peculiar cathodic chemistry.324 As has
been pointed out in the preceding sections, the
cornerstone of lithium ion technology is the formation
of a protective SEI on the carbonaceous materials,
and the conventional wisdom in this technology is
that PC cannot be used in combination with graphitic
anode materials because of its strong tendency to
cointercalate and exfoliate the graphene structure.
Prior to LiBOB, the presence of electrolyte solutes
alone has never been able to challenge this wisdom,
as Figure 53 shows. As an exception, however, LiBOB
Figure 52. Passivation of Al substrate in LiBOB-based in neat PC solution successfully enables the revers-
electrolytes: Time-decaying current observed on an Al
electrode at various potentials containing 1.0 M LiBOB in ible lithium ion intercalation/deintercalation on vari-
EC/EMC. Inset: the dependence of steady-state current ous graphitic anode materials, with capacity utiliza-
density (at t ) 103 s) on applied potential as obtained on tion and Coulombic efficiencies comparable with
an Al electrode in electrolytes based on various salts in the those of the state-of-the-art electrolytes. Considering
same mixed solvent. (Reproduced with permission from ref that similar stabilization of graphene structure in PC
155 (Figure 1). Copyright 2002 The Electrochemical Soci- was only able to be achieved by molecular additives
ety.)
before, the authors postulated that the BOB anion
surface of graphitic anode materials that supports must have participated in the SEI formation during
reversible lithium ion intercalation/deintercalation, the initial lithiation process of the graphite, most
and (3) it stabilizes an Al current-collector at high likely through a single-electron reductive process.324
potentials up to 6.0 V. Figures 51 and 52 present a The durability of such a protective SEI was put to a
brief summary of these qualifications, by all of which stringent test by cycling a full lithium ion cell based
LiBOB showed performances comparable with or on LiBOB/PC or other EC-free formulations as elec-
trolytes. As the result, stable capacity and quantita-

tive Coulombic efficiency were obtained for up to 100
cycles, similar to the case of the state-of-art electro-
lytes.324 Since these formulations are EC-free, the
lower temperature limit that has been plaguing EC-
centered electrolytes should be much improved.
Xu et al. emphasized that the significance of the
above serendipity lies not only in the fact that
PC-rich solvents could be used with graphitic anode
materials because of LiBOB but more in the perspec-
tive that the unofficial law governing the selection
of electrolyte solvents could be revised, and unprec-
edented flexibility is offered by LiBOB to the refor-
mulating of lithium ion electrolytes, where the dis-
placement of the high-melting-point component EC
is no longer restricted by the concern over the
stability of the graphitic anode.
In the subsequent investigations, these authors
adopted a preformation technique to differentiate the
steps for the formation of an effectively protective SEI
by LiBOB.488 They systematically interrupted the
lithiation of a graphite anode in LiBOB/PC at a series
of potentials and then tested the survivability of these
preformed anodes in LiPF6/PC, which would have
severely exfoliated the graphene structure had there Figure 54. Peculiar surface chemistry of BOB anion on
been no protective SEI present. The results confirmed graphitic anode material: XPS C 1s spectra for a graphitic
the critical role of BOB anion in the protection of anode surface cycled in LiBOB- and LiPF6-based electro-
graphitic anodes and, perhaps more importantly, lytes. The peaks were resolved into three major contribu-
found that an SEI sufficiently effective in protecting tions representing (1) hydrocarbon at 284.5 eV, (2) oligo-
graphite against PC cointercalation and exfoliation ether linkages at 286.5 eV, and (3) lithium alkyl carbonates
at 289.37 eV, respectively. (Reproduced with permission
was not completely formed till ∼0.60 V. They thus from ref 489 (Figure 3). Copyright 2003 The Electrochemi-
concluded that a 1.60 V reduction process that was cal Society.)
universally observed for all LiBOB-based electrolytes
might not have a direct relation with the formation Scheme 29. Proposed Mechanism for BOB
of this effective SEI.488 Reductive Decomposition on a Graphitic Surface
at Low Potentials
XPS analysis was also conducted by Xu et al. to
characterize the surface chemistry of a graphitic
anode resulted from LiBOB electrolyte, which yielded
an entirely different image of the SEI in terms of the
relative abundances of the chemical species, as
Figure 54 shows.489 With BOB-anion participation,
the content of chemical species similar to lithium
alkyl carbonate became the major components of the
SEI, as evidenced by the prominent peak at ∼289 eV.
The authors believed that these alkyl carbonate-like
species are originated from the oxalate moiety of the
BOB anion and are mainly responsible for the effec-
tive protection of graphitic anodes. A tentative mech-
anism was proposed accordingly in Scheme 29.489
Qualitatively, the chemical species present in the SEI
formed by a LiBOB-based electrolyte seemed to be weakened acidity of the LiBOB electrolyte, cathode
identical to those formed by LiPF6 counterparts, since materials based on spinel LiMn2O4 were well stabi-
the comparison between the XPS C 1s spectra lized, and the dissolution of Mn2+ species, which had
indicated that essentially the same peak locations for been considered a main cause of the capacity fading
hydrocarbon, ether-containing species, and semi- of this class of cathode materials, was nearly elimi-
carbonates were found in both spectra (Figure 54). nated.490 The thermal safety of LiBOB-based elec-
This fact could suggest that similar active agents trolytes against reductive decompositions was also
originated from the state-of-the art electrolytes are compared with that of the LiPF6 counterpart by Jiang
present in the LiBOB-originated SEI, despite the fact and Dahn, who concluded that LiBOB is a much safer
that these species are much more populated due to electrolyte solute as compared with LiPF6 on a
the involvement of BOB anion in SEI formation. graphitic anode because its exothermic reaction with
Further studies on LiBOB as a potential replace- the fully lithiated MCMB was delayed by 90 °C due
ment for LiPF6 were carried out by Liu et al., who to the presence of LiBOB.491 However, considering
reported that, due to the absence of HF and the much the conclusions by Maleki et al.361 and especially
MacNeil and Dahn,371 whether the LiBOB-based On the basis of their previous experiences with
electrolytes are safer against thermal runaway would lithium borates coordinated by substituted ligands,
still depend on their interaction with the cathode Barthel and co-workers modified the chelatophos-
materials. phate anion by placing various numbers of fluorines
On the basis of the findings on LiBOB performance on the aromatic ligands.495 Table 13 lists these
in nonaqueous solvents and other advances made to modified salts and their major physical properties.
improve the low-temperature performance of lithium As expected, the introduction of the electron-with-
ion electrolytes, Jow and co-workers proposed that drawing fluorines did promote the salt dissociation
an electrolyte with a much wider temperature range and reduce the basicity of phosphate anion, resulting
could be formulated using LiBOB alone or in combi- in increased ion conductivity and anodic stability.
nation with other salts.492 The following section (8.4) The phosphate with the perfluorinated aromatic
will be dedicated to this topic. ligands showed an anodic decomposition limit of 4.3
Another series of lithium borates with nonchelating V on Pt in EC/DEC solution. So far, these modified
alkyl ligands were briefly reported by Yamaguchi et lithium phosphates have attracted only academic
al. recently, where perhalogenated carbonyls were interest, and their future in lithium ion cell applica-
used to make the ligands electron-withdrawing.493 tions remains to be determined by more detailed
Their solubilities in linear carbonates are obviously studies.
superior to that of LiBOB, as solutions of these salts 8.3.2.5. Lithium Fluoroalkyl Phosphates
could be made in neat DMC or DEC with concentra- (LiFAP). Since the P-F bonds in PF6- anion are
tions higher than 1.0 m. The anodic stabilities of the labile toward the attack of moisture and other
electrolytes based on these salts were measured on nucleophiles such as the organic solvents with high
Pt and found to be comparable with those of electro- electron-donicity, they are believed to be the source
lytes based on LiBOB, while Al substrates were not for the instability of the state-of-the-art electrolytes.
corroded in these electrolytes at potentials up to 4.8 Recent structural modification of PF6- anion was
V. However, the thermal stability of these salts conducted by Satori et al. via the partial replacement
seemed to be inferior to that of the chelated borates of fluorine substituents with perfluoroalkyls in the
with either aromatic or alkyl ligands, although their hope that chemical as well as thermal stability of the
decomposition temperatures, ranging between 94 and resultant salts (LiFAP) could be improved due to the
135 °C, were still higher than that of LiPF6. Among inert nature of the P-C bonds.496,497 Since the new
these salts, the one with the perfluorinated carbonyl
substituents are also strong electron-withdrawing
ligands was considered to be the best-performing, and
groups, the main merits of LiPF6, such as good
the cycling of a graphitic anode half-cell containing
solubility and ion conductivity, are expected to be
a solution of this salt in EC/EMC showed that stable
maintained.
capacity could be obtained during 50 cycles. With the
exceptions of the initial cycles, quantitative Coulom- In their patent disclosure, Satori et al. described
bic efficiency was observed, indicating this salt is eight such salts with different perfluorinated alkyls,497
stable against cathodic reductions on graphitic anode the representatives of which are listed in Table 13
materials. However, no data of full lithium ion cells along with the main physical and electrochemical
were reported. properties. Among these salts, the most thoroughly
8.3.2.4. Lithium Chelatophosphates. If the work investigated is the lithium salt based on an anion
of Barthel et al. and Angell et al. on various borate- with three pentafluoroethyl groups. Hydrolysis stud-
based salts could be viewed as the structural modi- ies confirmed that, due to the replacement of fluorine
fication of a perhalogenated borate salt (LiBF4) that by the much bulkier perfluorinated alkyls, the sen-
was already used in commercial lithium ion cells, sitivity of the anion toward moisture is much re-
then similar modifications were also carried out on duced, as evidenced by the almost negligible amount
the more popular salt LiPF6, the industry standard of H2O consumption during 70 h of storage (Figure
of lithium ion cell electrolytes. 55) and, correspondingly, the negligible HF level in
Handa et al. reported the synthesis of a phosphorus the electrolyte.498 The much lower reactivity of LiFAP
equivalent of Barthel’s salts in which the hexavalent should be the result of the combined effects of the
phosphorus(V) was coordinated by three bidentate lower chemical activity of the P-C bond and the
ligands, 1,2-benzenediolato-O,O′.494 Its thermal sta- spatial hindrance to the attack of H2O. Ion conduc-
bility is similar to that of its boron counterparts, and tivities of LiFAP in mixed carbonate solvents were
moderate ion conductivity was achieved in nonaque- slightly lower as compared with LiPF6, obviously due
ous media. The authors attributed the less-than- to the larger anion size; however, comparable anodic
satisfactory ion conduction to the large size of the stability was observed for LiFAP in the voltammetry
anions, which increased the viscosity of the resultant scan conducted on a Pt electrode; that is, the major
electrolyte solutions. The anodic stability limit, as oxidative event occurred at ∼5.0 V, although a higher
measured by voltammetry on a Ni electrode, was background current level at potentials above 4.0 V
below 3.7 V. A preliminary test of this salt in EC/ was observed. Preliminary cycling tests were con-
THF was conducted in a lithium cell using the low ducted in cathode half-cells based on LiMn2O4 using
potential cathode, V2O5, and the authors believed LiFAP/EC/DMC as electrolyte, and higher capacity
that this salt could be a superior electrolyte solute, was obtained than with LiPF6 despite the fact that
judging from the utilized cell capacity that was close the capacity of the cells based on both electrolytes
to the theoretical value. faded in a similar manner.
Figure 55. Presence of fluoroalkyls enhancing the anion

stability against moisture. Consumption of H2O and the
generation of HF in 1.0 M salt solutions of EC/DMC with
added H2O at 500 ppm (for LiPF6) and 1000 ppm (for
LiFAP), respectively: triangle, H2O concentration; circle,
HF concentration. (Reproduced with permission from ref
498 (Figure 2). Copyright 2001 Elsevier.)
More stringent electrochemical characterizations

were carried by Aurbach and co-workers, who com-
paratively investigated the interfacial properties of
the electrolytes based on LiFAP on anode and cath-
ode materials against the benchmark salts LiPF6 and
LiBeti through various instrumental means, includ-
ing voltammetry, EIS, FT-IR, and XPS.499,500 They Figure 56. Slow scan (10 µV s-1) voltammetry on a
found that, while on both the graphitic anode and graphite working electrode: (a) 1.0 M LiFAP in EC/DEC/
DMC; (b) LiPF6 in EC/DEC/DMC. Solid line: pristine
the spinel cathode the LiFAP-based electrolyte showed graphite. Dashed line: after 1 week of cycling. (Reproduced
higher capacity utilization and better capacity reten- with permission from ref 499a (Figure 4). Copyright 2003
tion, slower kinetics for lithium ion intercalation/ The Electrochemical Society.)
deintercalation were found in both cases. The cause
for the slow kinetics seemed to be related to the obvious gap for LiPF6 in the anodic section were
resistive SEI and surface films on each electrode, observed. In other words, there were long-term
which was unexpected because there should have secondary processes in the LiPF6-based electrolyte
been little LiF present in the SEI, due to the stability that affected the electrode kinetics, while the gra-
of LiFAP, and it was concluded that LiF was mainly phitic anode in the LiFAP-based electrolyte was
responsible for the high-impedance interfaces on under better protection from such undesirable pro-
either the anode or the cathode with LiPF6 as the cesses.
electrolyte solute. FT-IR and XPS indicated that Therefore, the authors concluded that, although
LiFAP-based electrolytes generated higher amounts direct identification was not available through spec-
of lithium alkyl carbonate on the inert electrode (Pt troscopic means, the FAP anion must have partici-
and Au) surfaces that were cathodically polarized to pated in the formation of surface layers, which served
simulate the surface chemistry on a cycled graphitic as protection against sustained decompositions on
anode, but these spectroscopic means failed to iden- one hand but were also responsible for the high
tify any direct involvement of FAP anion in SEI impedance across the interfaces on the other. These
formation. Nevertheless, it is apparent that in LiFAP- robust surface films might exist on both anode and
based electrolytes the surface species are mainly cathode surfaces and consist mainly of lithium alkyl
reduction products from solvents, while in LiPF6- carbonates because of the low level of HF in the
based electrolytes the anion is apparently more solution.
involved in the reductive decompositions, as evi- The thermal stability of LiFAP was also studied
denced by the XPS C 1s and F 1s spectra.499 The by Aurbach and co-workers in EC/DEC/DMC solution
higher stability of LiFAP is shown clearly by Figure using ARC. As compared with LiPF6, LiFAP delayed
56, wherein two sets of CV were measured for LiFAP- the onset thermal decomposition of the electrolyte by
and LiPF6-based electrolytes after stabilization of the ∼10 °C; however, the self-heating became much more
electrode and after 1 week of cyclings, respectively. severe once the reactions started.500 The above ARC
In all the voltammetric profiles, four distinct stages was conducted in the absence of electrode materials.
of lithium intercalation are visible, but a very sig- In brief summary, LiFAP as a potential replace-
nificant difference exists between LiFAP and LiPF6: ment for LiPF6 would result in a stable performance
nearly identical current responses for LiFAP but an in lithium ion cells with possibly increased thermal
stability, but the power rate of such cells would survey of the commercial lithium ion cells made by
probably be inferior to that of the state-of-the-art major manufacturers in the late 1990s, the power
electrolytes. If the manufacturing cost of producing rate and energy density that can be delivered at -40
these perfluorinated anions could be reduced, LiFAP °C averaged only 1.25% and 5% of those achievable
or its derivatives might be a competitive electrolyte at room temperatures, respectively.506 A similar
solute for lithium ion electrolytes. survey conducted a few years later showed that
8.3.2.6. Lithium Salts Based on Heterocyclic significant progress had been made, where as much
Anions. Lithium salts based on organic anions where as 30% of the rated capacity was deliverable at -40
the formal charge is delocalized throughout substi- °C by the best cells tested.507 However, the deterio-
tuted heterocyclic moieties were also reported spo- rated performance still plagued the reliable opera-
radically, which included, for example, lithium 4,5- tions of this technology. Despite the sharp differences
dicyano-1,2,3-triazolate501 and lithium bis(trifluoro- among the cells from various manufacturers, a com-
borane)imidazolide (LiId).502 The former was devel- mon observation concerning their low-temperature
oped as a salt to be used for polymer electrolytes such behavior is that the major deterioration in perfor-
as PEO, and no detailed data with respect to elec- mance occurs below -20 °C, while above this thres-
trochemistry were provided, while the latter, which hold temperature at least 80% of the rated capacity
could be viewed as a Lewis acid-base adduct be- is deliverable.505,507
tween LiBF4 and a weak organic base, was intended This sharp decline in cell output at subzero tem-
for lithium ion applications (Table 13). peratures is the combined consequence of the de-
Because of the organic presence in these anions, creased capacity utilization and depressed cell po-
the corresponding lithium salts are highly soluble in tential at a given drain rate, and the possible causes
media of low or medium dielectric constants and have been attributed so far, under various conditions,
could provide ion conductivities comparable to those to the retarded ion transport in bulk electrolyte
of LiPF6. LiId was reported to have high anodic solutions,503,508,509,510 the increased resistance of the
stability as compared with LiPF6, although the surface films at either the cathode/electrolyte inter-
background current of the electrolytes based on it was face506,507 or the anode/electrolyte interface,461,504,511
higher than those for LiPF6 counterparts at poten- the resistance associated with charge-transfer pro-
tials above 4.0 V, as measured by cyclic voltammetry cesses at both cathode and anode interfaces,134,135,512
on Pt. The compatibility of this salt with LiNi0.8- and the retarded diffusion coefficients of lithium ion
Co0.2O2 cathode material was investigated by testing in lithiated graphite anodes.165,513 The efforts by
the corresponding half-cells, and performances com- different research teams have targeted those indi-
parable with or better than those of LiPF6-based vidual electrolyte-related properties to widen the
electrolytes were reported. However, in the case of temperature range of service for lithium ion cells.
MCMB, it took longer for LiId-based electrolyte to 8.4.1.1. Solvent Approach. Since the high-melt-
reach full utilization of the capacity, and it was only ing EC constitutes 30-50% of the bulk electrolyte in
after ∼10 cycles when the difference between LiId most of the commercial lithium ion cells, it has been
and LiPF6 vanished. A higher irreversible capacity the primary blame for the poor performances of such
was also associated with LiId in the first cycle, where electrolytes at low temperatures, and its mixing with
the Coulombic efficiency was 10.2% versus 6.2% for other lower melting cosolvents has been the most
LiPF6, suggesting that more charges were consumed favorable approach so far adopted by the researchers
in building up an SEI when LiId was the electrolyte with the aim to develop an electrolyte for subambient
solute. On the other hand, the authors argued that applications. As a pioneering attempt to develop a
this capacity loss was negligible in terms of the total low temperature for lithium ion chemistry, Ein-Eli
capacity. Considering the presence of B-F bonds in et al. used methyl formate (MF, mp -99 °C) and
the anion of LiId, its thermal and chemical stability reported the ion conductivity at -40 °C for LiMe in
should be similar to that of LiBF4, and its potential EC/MF (1:3) to be 5.4 mS cm-1.102 The authors
applications to replace LiPF6 remain to be confirmed concluded that the depression effect of MF rendered
by further studies. the electrolyte liquid at such low temperatures.
However, the performance of the anode half-cell was
8.4. Novel Electrolytes with a Wide Temperature not improved ideally as expected, since only 50%
Range deliverable capacity was achieved at -2 °C, which
far exceeded the extent by which the bulk ion
8.4.1. Low-Temperature Performance conductivity dropped. The possible origin for this
discrepancy might arise from the elimination of
Following its rapid rise to dominance in the con- supercooling at temperatures below -20 °C due to
sumer cell market intended for portable electronics, the presence of electrode particles in the anode half-
lithium ion technology was actively considered for cell, which are absent in conductivity cells, but more
special applications such as those in military and likely, it indicates that the bulk ion conductivity may
space missions.503-505 However, the poor performance not dictate the performance of the electrolytes in an
of the state-of-the-art lithium ion cells at tempera- actual lithium ion device. Since electrolytes based on
tures below -20 °C remained a major obstacle to neat MF failed to form an effective SEI on graphite
enabling the normal operations in harsh environ- and the stability of a graphite anode in the above EC/
ments that are frequently encountered in those MF mixture solvents was clearly dependent on the
missions. For example, according to a comprehensive EC presence, it appears that reductive decomposition
Figure 57. Ion conductivity of electrolytes containing 1.0

M LiPF6 in (1) EC/DMC (3:7), (2) EC/DEC (3:7), and (3)
Figure 58. Comparison of the discharge profiles of graphite-
EC/DMC/DEC (1:1:1). (Reproduced with permission from
based AA-size lithium ion cells at -20 °C among the
ref 508 (Figure 1). Copyright 1999 The Electrochemical
various electrolytes. (Reproduced with permission from ref
Society.)
508 (Figure 7). Copyright 1999 The Electrochemical Soci-
ety.)
of MF made it an unfavorable candidate despite its
low melting temperature. thresdhold value.461,462 In other words, the reformula-
An important contribution to the search for a low- tion of these low-temperature electrolytes has to
temperature electrolyte was made by Smart et al. in consider the surface chemistry of the new cosolvents
late 1990s, apparently driven by the needs of NASA in addition to their positive effects on ion conductiv-
for the application of lithium ion devices in space ity. This conclusion, now well accepted as a basic
missions.461,462,466,508,515 The basic strategy of the team guideline, has been confirmed by the numerous
was to minimize the concentration of EC in the examples found later, where the most conductive
ternary or quaternary electrolyte formulations so that electrolytes do not necessarily provide the best im-
a synergistic effect could be reached by the various provement at low temperatures.134,168,511,512
components based on either linear carbonates or Following the initial work by Smart et al., similar
esters. In an ideal scenario of role assignment, less efforts were reported based on various binary and
than 30% EC would suffice for the effective solvation ternary compositions, such as EC/DMC/EMC
of lithium salts and ensure the formation of an (1:1:1),503,505,509 EC/DMC/DEC (2:2:1),505 EC/EMC
effective SEI on electrode surfaces, while the majority (1:3),505 EC/DMC/methylpropyl carbonate (MPC) and
components such as the low-melting DEC and the EC/DMC/isopropyl carbonate (iPC) at a series of
less viscous DMC would confer upon the mixture varying compositions.514 Unfortunately, below -30 °C
solvents a low liquidus temperature and facile ion the above compositions still failed to function well.515
transport at low temperatures. Figure 57 shows this In view of the limited improvements, Smart et al.
synergistic effect achieved in a ternary electrolyte sought to use linear alkyl esters, which are lower
through the comparison of the temperature-depen- melting than the common alkyl carbonates by an
dence of ion conductivities against the corresponding average of 20-30 °C (Table 1). On the basis of the
binary electrolytes.508 The discharge characteristics experience that MF was incompatible with graphitic
of the cells based on these electrolytes were also anodes, they chose alkyl esters of higher carboxylic
compared at -20 °C, as Figure 58 shows. An appar- acids so that carboxylate connected with an alkyl
ent gap exists among the electrolytes in terms of the instead of H as in MF. Thus, acetates such as MA
delivered cell capacity, although at room temperature and EA and higher carboxylates such as EP and EB
all of the electrolyte compositions behave similarly. were used to formulate quaternary electrolytes on the
In an extensive characterization of the irreversible basis of the ternary compositions EC/DMC/DEC
capacity during the initial cycle, the durability and (T).461,462 As expected, due to the introduction of these
resistance of the SEI, the self-discharge rate, and so esters, the ion conductivity was obviously improved,
forth, the ternary electrolyte was shown to possess a while both anodic and cathodic stability as measured
series of intermediary qualities as compared with the on a Pt surface were negatively affected. In terms of
binary baselines. The authors believed that it was compatibility with a graphite anode, there exists an
these well-balanced properties of the ternary com- apparent distinction between the acetates (MA and
position that made it an excellent candidate for EA) and the esters of higher carboxylic acids (EP and
subzero applications. EB), since the reversible capacities that can be
An important conclusion that these authors drew utilized by the electrolytes containing the acetates
based on their polarization and EIS studies is that are much lower, and EP- or EB-based electrolytes
the properties of the SEI film on a graphite anode yield comparable performances with the ternary
surface play a far more decisive role in determining baseline. Therefore, the authors speculated that the
the kinetics of the lithiation/delithiation at low tem- SEI formed in the presence of low molecular weight
peratures than does the bulk ion conductivity, al- esters appeared to be resistive and inadequately
though it is necessary for the latter to achieve a protective, whereas, in the presence of esters of
Figure 59. Effect of ester-containing quaternary electro- Figure 60. Discharge capacities of LiNiCoO2/MCMB
lytes on the discharge capacity of AA-size lithium ion cells lithium ion cells at -20 °C with different carbonate-based
(0.4-0.5 A h) at -40 °C under the drain rate of 25 mA. electrolytes. Cells are charged at room temperature and
The electrolytes are (1) 1.0 M LiPF6/EC/DEC/DMC discharged using a C/15 rate. (Reproduced with permission
(1:1:1), (2) 0.75 M LiPF6/EC/DEC/DMC/MA (1:1:1:1), (3) from ref 515 (Figure 4). Copyright 2003 Elsevier.)
0.75 M LiPF6/EC/DEC/DMC/EA (1:1:1:1), and (4) 0.75 M
LiPF6/EC/DMC/MA (1:1:1). (Reproduced with permission cycle). Thus, the authors concluded that these new
from ref 462 (Figure 12). Copyright 2002 The Electrochemi- electrolytes could be used as cosolvents for low-
cal Society.)
temperature electrolytes. Considering that Smart et
higher molecular weight, the SEI could be formed al. and Herreyre et al. used entirely different cells
with more desirable attributes. Combining the ob- as their testing vehicles, the above discrepancy might
servations of Ein-Eli et al.,102 Smart et al.,461,462 and not be too incomprehensible, since similar discrep-
Herreyre et al.,406 it could be tentatively concluded ancies had been encountered when the trends ob-
that longer alkyl chains in the carboxylic acid section served in anode half-cells did not correlate well with
of the esters play a critical role in determining the that of prototype cells with a different cell design.462
cathodic stability of this component on a graphite It actually indicated the complexity in the operation
anode. The tests in AA-size full lithium ion cells were of lithium ion chemistry, during which various factors
only reported for EA- and MA-based quaternary including chemistry as well as engineering exert their
electrolytes, and Figure 59 shows the discharging influences.
profiles of these cells at -40 °C. Despite their Besides the Coulombic capacity, Herreyre et al.
negative effect on anode capacity utilization at room also pointed out that the depression in cell voltage
temperature, MA and EA still improved the capacity at low temperatures could be mitigated by EA and
significantly. MB as cosolvents because of the reduced resistance
On the other hand, the presence of these esters in in both bulk electrolyte and electrolyte/electrode
the electrolyte solutions raised concern over the long- interfaces. Moreover, they found that this cell voltage
term performance at room temperatures, because EIS depression was also related to the salt concentration;
studies indicated that the resistance associated with using 1.5 M instead of 1.0 M LiPF6 in the electrolyte
the SEI film increased at a much higher rate for enabled the increase of the cell working range from
ester-based electrolytes as compared with the com- -30 to -40 °C, as, in most cases, the cell cycling is
positions that were merely based on carbonates. The regulated by a preset cutoff potential.
authors attributed this rising cell impedance to the Out of the belief that the alkyl esters are reactive
reactivity of these esters toward the electrode active in a lithium ion cell during the long-term operation,
material, which resulted in the continued growth of Smart et al. proposed a “carbonate-only” guideline
the SEI film in the long term and suggested that and formulated a series of quaternary compositions
alkyl esters, especially those of acetic acid, might not consisting of EC/DEC/DMC/EMC, in which the EC
be appropriate cosolvents for low-temperature ap- concentration remained under 25%.515 The solution
plication electrolytes.461 of 1.0 M LiPF6 in EC/DEC/DMC/EMC (1:1:1:3) was
The work of Herreyre et al., however, took a much reported to be the best composition, whose ion
more optimistic tone on the use of linear esters EA conductivity at -40 °C was 1.32 mS cm-1, and its
and MB.406 In LiCoO2/graphite cells, the ternary cycling in LiNiCoO2/MCMB lithium ion cells was
electrolyte compositions such as 1.0 M LiPF6 in EC/ comparable to that of the baseline electrolyte 1.0 M
DMC/EA, EC/DMC/MB, and PC/EC/MB were re- LiPF6 in EC/DMC (3:7) or (1:1) at room temperatures.
ported to be able to deliver as much as 88-95% of At -20 °C, the superiority of these quaternary
the rated capacity at -30 °C with a C/2 rate or 81- electrolytes became pronounced, as compared with
87% at -40 °C, while at room temperature the the ternary and binary baselines shown in Figure 60.
capacity fading rate (0.05% per cycle for MB and More importantly, these electrolytes also allowed
0.09% per cycle for EA) and capacity retention with the lithium ion cells based on them to be charged at
high-temperature storage (60 °C, 14 days at full low temperatures at reasonable rates, as shown by
state-of-charge) of these cells were comparable with Figure 61, which had been impossible for most of the
those of the state-of-the-art electrolytes (0.05% per lithium ion cells because the high impedance of the
Figure 61. Discharge capacities of LiNiCoO2/MCMB

lithium ion cells at -40 °C with different carbonate-based
electrolytes. Cells are charged at -40 °C and discharged Figure 62. Cycle life performance of SAFT DD-size
using a C/4 rate. (Reproduced with permission from ref 515 lithium ion cells containing 1.0 M LiPF6/EC/DEC/DMC/
(Figure 6). Copyright 2003 Elsevier.) EMC (1:1:1:3) at various temperatures. Cutoff voltages for
low-temperature charge were indicated in the graph.
cells at a discharged state would result in lithium Copyright 2003 Elsevier.)
deposition on a graphite surface due to the high
overpotential.515 Using a three-electrode cell, Smart
et al. monitored the potential of the cathode and
anode of a LiNiCoO2/MCMB lithium ion cell during
its cycling at -20 °C, and they confirmed that,
although the potential of the graphite anode was
indeed driven to negative regions in the charge
process, no lithium deposition occurred, as evidenced
by the absence of a potential plateau at 0.0 V
corresponding to lithium stripping in the following
discharge process, which had been observed previ-
ously in a similar cell using LiPF6/EC/EMC (1:3) as
the electrolyte.165 They attributed this to the facile
kinetics of lithium ions in the bulk electrolyte as well
as in the SEI film on the graphite anode. The cycling
tests of SAFT prototype lithium ion cells carried out
in a wide temperature range between -70 and 40 °C
confirmed that the above quaternary composition is
an excellent low-temperature electrolyte. As illus-
trated in Figure 62, when the cell is continuously Figure 63. Delithiation capacity of an MCMB anode at
cycled at -20 °C using a C/10 charge rate and a C/5 -20 °C in various electrolytes following charge (lithiation)
discharge rate, stable capacity retention can be at -20 °C. The drain rate is 50 mA (∼C/12). (Reproduced
obtained with 80% of the rated capacity delivered. with permission from ref 466 (Figure 4). Copyright 2003
Elsevier.)
Using lower rates (C/15 charge and C/10 discharge),
70% of the rated capacity can be accessed even when
cycled at -40 °C. achieved in the best case, however. EIS and polariza-
Fluorinated carbonates were also used by Smart tion studies on the interface between the electrolyte
et al. as low-temperature cosolvents (Table 12), in the and the graphitic anode showed that these fluori-
hope that better low-temperature performances could nated cosolvents in general created less resistive
be imparted by their lower melting points and favor- surface films and more rapid charge-transfer kinetics
able effects on SEI chemistry.466 Cycling tests with when compared with nonfluorinated carbonates, and
anode half-cells showed that, compared with the the difference increased with decreasing tempera-
ternary composition with nonfluorinated carbonates, ture. Full lithium ion cells based on LiNixCo1-xO2 and
these fluorinated solvents showed comparable and MCMB were also assembled with one of these novel
slightly better capacity utilizations at room temper- solvents, and good reversibility was observed, al-
ature or -20 °C, if the cells were charged at room though no data on subzero temperature tests were
temperature; however, pronounced differences in reported.
discharge (delithiation) capacity could be observed if According to the authors, additional merits of such
the cells were charged (lithiated) at -20 °C, where fluorinated carbonates would include their lower
one of these solvents, ethyl-2,2,2-trifluoroethyl car- flammability and higher stability against storage at
bonate (ETFEC), allowed the cell to deliver far elevated temperatures as compared with their non-
superior capacity, as Figure 63 shows. Only 50% of fluorinated counterparts; therefore, the incorporation
the capacity deliverable at room temperature was of these novel solvents into the current commercial
lithium ion cells seems promising if the manufactur-

ing costs and environmental requirements do not
constitute any obstacle.
In addition to linear carbonates, PC was also
considered as a cosolvent that could help to improve
the low-temperature performance of the electrolytes,
mainly due to its wide liquid range and solvation
ability to lithium salts. This latter property seems
to be a merit relative to the linear carbonates, whose
dielectric constants are generally below 10 and whose
displacement of EC usually causes the solubility of
lithium salts to decrease in such mixed solvents,
especially at low temperatures.
To characterize the effect of PC addition on the
related properties of the electrolytes, Zhang et al. Figure 64. Arrhenius plots of the ion conductivities for
carried out a comparative study on PC/EC/EMC the electrolytes composed of 1.0 m lithium salts in PC/EC/
(1:1:3) and EC/EMC (3:7).168 They found that, because EMC (1:1:3). (Reproduced with permission from ref 134
of the higher viscosity of PC, the ion conductivity of (Figure 1). Copyright 2002 Elsevier.)
the former formulation was lower at subzero tem-
peratures, and its declination accelerated with de-
creasing temperature. Nevertheless, due to the large
portion of linear carbonates in the formulation, the
compromise in ion conductivity was marginal. Dis-
charge tests in LiNiO2/graphite lithium ion cells were
carried out down to -30 °C, and PC-containing
electrolytes was found to outperform the baseline.
According to the authors, this improvement could not
be attributed to the bulk ion conductivity of the PC-
based electrolyte but to the lower cell impedances
associated with the films on both cathode and anode
surfaces. No cycling tests below -20 °C were de-
scribed. Considering that PC would induce higher
irreversible capacities in the initial cycles and even-
tually result in lower capacity in cathode-limit lithium
ion cells, the use of it as a low-temperature electrolyte
cosolvent is possible only if one of the following
measures is adopted to suppress the cointercalation
and decomposition of PC: (1) the use of SEI-modify-
ing additives, (2) the use of special salts that can help
form a more protective SEI (such as LiBOB), or (3)
the surface modification of the graphite anode.
8.4.1.2. Salt Approach. Another less expected
approach aimed at low-temperature electrolytes was
also proposed by Zhang et al., who reported that the
replacement of LiPF6 by LiBF4 would result in
improved discharge capacity of lithium ion cells.134,135 Figure 65. Comparison between the discharge capacities
LiBF4 had been considered an inferior electrolyte for lithium ion cells at various temperatures. Relative
capacity was defined as the ratio of the capacity at the
solute to LiPF6 due to its lower ion conductivity in specified temperature to the one obtained at 20 °C: (a, top)
nonaqueous solvents and the less protective SEI LiPF6/EC/PC/EMC (1:1:3); (b, bottom) LiBF4/EC/PC/EMC
formed by these electrolytes, and its application in (1:1:3). (Reproduced with permission from ref 134 (Figure
commercial lithium ion cells seemed to be possible 2). Copyright 2002 Elsevier.)
only in the unique solutions containing γBL. As
Figure 64 shows, in the entire temperature range the bulk electrolytes does not dictate low-temperature
from 60 to -60 °C, LiBF4 exhibited a lower ion performance.
conductivity than LiPF6 in PC/EC/EMC (1:1:3) solu- By EIS analysis of the corresponding lithium ion
tions,134 while a similar trend was found for EC/DMC/ cells, Zhang et al. showed that the impact of SEI
DEC solutions as well.135 However, all the LiBF4- resistance on total cell impedance was rather negli-
containing lithium ion cells obviously outperformed gible, and hence, they attributed the superior low-
their LiPF6 counterparts in discharge capacity at temperature behavior of LiBF4-based electrolytes to
temperatures below 0 °C, and as Figure 65 shows, the lower resistance associated with the so-called
at -40 °C, its PC/EC/EMC solution was still able to “charge-transfer processes”, which are usually rep-
deliver up to 86% of the rated capacity, while the resented in impedance spectra by the semicircle at
LiPF6 counterpart could only access 72%. These the lower frequency region.516 This suggestion could
results once again confirmed that ion conductivity in be viewed as a further extension of the conclusion
electrolyte.517 This observation indirectly corrobo-

rated the conclusions of Zhang et al. and seemed to
indicate that LiBF4 possessed certain qualities of an
electrolyte solute for low-temperature-oriented ap-
plications. In combination with the advances ob-
tained in the solvents approach, further improvement
in the performance at extreme temperatures might
be a probable perspective.
8.4.1.3. Limiting Factors for Low-Tempera-
ture Operation. One controversial topic that has
raised wide attention relates to the limiting factors
of the low temperature of lithium ion cells. The
researchers not only debated about whether the
anode or cathode controls the overall low-tempera-
ture performance of a full lithium ion cell but also
disagree upon the rate-determining steps that govern
Figure 66. Nyquist plots of the impedance spectra as
measured for the fully charged lithium ion cells at -30 °C the low-temperature kinetics of lithium ion intercala-
in which the inset shows the magnified view of the high- tion at the graphitic anode.
frequency part. Electrolytes are 1.0 m LiPF6 (hollow) and Due to the contributions from Smart et al.,462 the
LiBF4 (solid) in PC/EC/EMC (1:1:3). Note that the semi- emphasis of low-temperature study on electrolytes
circles in the inset are almost invisible in the scale of the
whole spectra. (Reproduced with permission from ref 134
has been placed on the anode side, and the SEI film
(Figure 4). Copyright 2002 Elsevier.) on the graphitic surface has generally been recog-
nized as the most resistive component in the journey
drawn by Smart et al., who had identified RSEI as the of the lithium ions during the cell operation, which
limiting resistance in the lithium cell operation at must travel across the electrolyte and intercalate into
low temperatures.462 In fact, Smart et al. had already or deintercalate from the bulk graphite structure.
realized the effect of Rct on low-temperature behavior, This hypothesis was mainly established on the basis
since they also used an effective exchange current of two observations made in the EIS studies of the
density to represent the lithium ion intercalation lithium/graphite half-cells: (1) the resistance corre-
kinetics. This latter quantity should reflect the sponding to the surface film component (RSEI) far
overall effect of RSEI and Rct.462,508 outweighs the component representing electrolyte
Thus, by comparing the impedance spectra of the bulk resistance, and (2) the steep temperature-
lithium ion cells, Zhang et al. showed that the dependence of RSEI in the low-temperature ranges
resistance corresponding to the charge-transfer pro- below -20 °C matches the rapid deterioration of the
cess (Rct) dominates the total cell resistance at low half-cell performance, while the ion conductivities do
temperatures. For example, at -30 °C (Figure 66), not suffer any dramatic drop in this range.164,168
the semicircle and the interception corresponding to These two phenomena have been repeatedly observed
Rf and Rb have to be magnified in the inset to be in various electrolyte systems in which novel cosol-
visible. Considering the huge difference between the vents were added to depress the liquidus temperature
magnitudes of Rb or Rf and Rct, the authors concluded and to improve solution transport properties,
that the poor low-temperature performance of lithium and complementary evidence was also obtained
ion cells is caused by the slow kinetics of charge- from various electrochemical polarization tech-
transfer at these temperatures. The decoupling of the niques.466,511,512
bulk resistance (Rb, related to the ion conductivity The above hypothesis was questioned by Huang et
in the bulk electrolytes) from the charge-transfer al., who suggested that the critical factor that limits
process and film resistances (Rct and Rf, related to the anode capacity accessible at the low temperatures
the ion transport in the surface film and the Faradaic is the kinetics of the lithium ion in the bulk carbon-
process, respectively) can also be clearly seen in aceous anode instead of the surface film.513 Their
Figure 66 (at -30 °C): that is, although the cell based argument was based on the universally observed
on LiBF4 has higher Rb (the inset), its Rct is much asymmetric behavior of graphitic anodes toward
lower than that of the cell with LiPF6.134 Similar lithiation and delithiation at low temperatures; that
observations were also made with EC/DMC/DEC is, while the fully charged graphite can release the
formulations.135 It should be noted that, since full intercalated lithium ions at temperatures below -20
lithium ion cells were used in these EIS studies, Rct °C with relative readiness, the attempt to lithiate a
and Rf mentioned above should reflect the combina- fully discharged graphite anode at the same temper-
tion of these corresponding processes on both anode/ ature is severely hampered by the high resistance at
electrolyte and cathode/electrolyte interfaces. the surface. Considering the electrolyte nature of the
In another work unrelated to the low-temperature SEI, which serves as an electronic insulator but an
electrolytes, Mohamedi et al. characterized the spray- ionic conductor, the authors pointed out that it should
deposited thin film of spinel cathode material by neither behave like a diode that only impedes lithium
means of EIS and studied the correlation between insertion but allows extraction nor vary in its resis-
electrolyte composition and the impedance compo- tance depending on the state-of-charge of the graph-
nents. Among the three lithium salts investigated, ite. Therefore, the above asymmetric behavior should
the lowest Rct and Rf were obtained in a LiBF4-based not be attributed to the SEI, whose resistance should
graphite surface and the subsequent permanent

capacity loss when an unusual cycling regulation was
applied to deal with the reduced capacity accessible
at low temperatures.165 These authors attributed the
lithium deposition to the excessive concentration
polarization on the graphite surface.
However, more recent work by Wang et al.511 and
Zhang et al.512 seemed to present convincing evidence
challenging the suggestion by Huang et al. that
electrolyte and the SEI resistances do not affect the
low-temperature performance of graphitic anodes.
The former authors used the combination of the
galvanostatic intermittent titration (GITT) technique
and EIS to analyze the individual impedance com-
ponents that simulate graphite in the quaternary
electrolyte composition LiPF6/EC/PC/DMC/EMC
(4:1:3:2). Since the reaction resistance as measured
by GITT is the sum of the electrolyte bulk resistance,
RSEI, Rct, stage transformation, and the resistance
Figure 67. Effect of coke particle size on the charge that corresponds to lithium diffusion in graphite, it
capacity at various temperatures. (Reproduced with per- is possible to distinguish the impact of their temper-
mission from ref 513 (Figure 8). Copyright 2000 The
Electrochemical Society.) ature-dependences on the overall electrode kinetics.
They found that, during the low-temperature cy-
clings, the diffusional resistances were similar during
be independent of the lithiation degree in the graph- charge and discharge, thus contradicting the hypoth-
ite, but rather to the lithium ion diffusion within the esis by Huang et al. that the concentration-depen-
carbonaceous anode structure, which has direction- dence of lithium diffusion would result in higher
ality due to the concentration polarization. Thus, the concentration polarization to lithiation than to delithi-
diffusion coefficient of lithium ion within the carbon- ation. Furthermore, in a quantitative manner, they
aceous anode (DLi) would always be higher for lithium also found that the magnitude of RSEI makes it the
deintercalation but lower for intercalation, and the dominating component at the low temperatures when
gap between the two processes becomes more promi- compared with the other impedance components and,
nent as temperature goes down.513 hence, concluded that at -30 °C the limiting factor
To further support their argument that low-tem- for the lithium intercalation is RSEI.
perature lithium ion intercalation is governed by The latter authors used anode and cathode sym-
lithium diffusion instead of the SEI, the authors metrical cells in EIS analysis in order to simplify the
investigated the effect of carbon particle size on its complication that often arises from asymmetrical
specific capacity by choosing two carbonaceous an- half-cells so that the contributions from anode/
odes of the same capacity at room temperature but electrolyte and cathode/electrolyte interfaces could be
different particle sizes (6 and 25 µm). They found isolated, and consequently, the temperature-depend-
that, when cycling at temperatures below -20 °C, the ences of these components could be established. This
anode with fine particles outperformed the one with is an extension of their earlier work, in which the
coarse particles in the charge capacity, as Figure 67 overall impedances of full lithium ion cells were
shows. The authors believed that this particle size studied and Rct was identified as the controlling
effect was consistent with their contention that factor.134,135 As Figure 68 shows, for each of the two
lithium diffusion is the rate-determining step that interfaces, Rct dominates the overall impedance in the
limits the low-temperature performance of carbon- symmetrical cells as in a full lithium ion cell, indicat-
aceous anodes, because the large diameter of the 25 ing that, even at room temperature, the electrodic
µm particle would require longer lithium diffusion reaction kinetics at both the cathode and anode
lengths and, consequently, a lower lithiation degree surfaces dictate the overall lithium ion chemistry. At
would be realized in this case, with the assumption lower temperature, this determining role of Rct
that DLi is the same for both anode materials. becomes more pronounced, as Figure 69c shows, in
Combining the above evidences, Huang et al. which “relative resistance”, defined as the ratio of a
further concluded that the improvement in the low- certain resistance at a specific temperature to that
temperature performances of lithium ion cells would at 20 °C, is used to compare the temperature-
eventually rely on the effort to develop anode materi- dependences of bulk resistance (Rb), surface layer
als of high lithium diffusion coefficients instead of resistance (Rsl), and Rct. For the convenience of
the electrolytes and SEI that were less resistant. comparison, the temperature-dependence of the ion
The conclusion of Huang et al. was supported by conductivity measured for the bulk electrolyte is also
Lin et al., who used a three-electrode cell design to included in Figure 69 as a benchmark. Apparently,
monitor the voltage profile of both the anode and both Rb and Rsl vary with temperature at a similar
cathode in a full lithium ion cell during cycling at pace to what ion conductivity adopts, as expected, but
low temperatures and found that these cyclings a significant deviation was observed in the temper-
resulted in the deposition of metallic lithium on the ature dependence of Rct below -10 °C. Thus, one
Figure 68. Nyquist plots of a charged lithium ion cell, a

lithiated graphite/graphite cell, and a delithiated cathode/
cathode symmetrical cell. The inset is an equivalent circuit
used for the interpretation of the impedance spectra.
could conclude that the rate-determining factor that

limits the capacity utilization of the graphitic anode
as well as the cathode should most likely be Rct.
Perhaps a more important finding by these authors
is the “concentration-dependence” of Rct. As Huang
et al. has pointed out, RSEI should be independent of
state-of-charge,513 and the authors proved that RSEI
remains relatively constant on both charged and
discharged anode and cathode surfaces, as shown by
the insets of Figure 70. However, Rct showed a strong
dependence on the lithiation degree of both the anode
and cathode. Two extreme situations were simulated
in Figure 70, which correspond to a fully charged
lithium ion cell, wherein the cathode is delithiated
and the anode is lithiated, and a fully discharged
lithium ion cell, wherein the cathode is lithiated and
the anode is delithiated. Apparently, in the former
case, Rct on both the anode and cathode is substan- Figure 69. Comparison for temperature-dependence of the
tially high, indicating a high polarization for the relative resistances of a charged lithium ion cell, a lithiated
lithiation of the anode and delithiation of the cathode, graphite/graphite cell, and a delithiated cathode/cathode
while the reverse processes for each electrode are cell. The dashed curves show the “relative resistance” of
much easier in a fully charged lithium ion cell. Hence, the electrolyte, which was taken as the ratio of the
electrolytic conductivity at a specific temperature to the
these authors proposed that the asymmetrical be- conductivity at 20 °C: (a) Rb; (b) Rsl; (c) Rct. (Reproduced
havior of lithium ion cells at low temperature most with permission from ref 512 (Figure 4). Copyright 2003
likely originated from the concentration-dependence Elsevier.)
of Rct instead of the directionality of lithium diffusion
within the bulk graphite anode.512 Since in numerous
reports Rct has been shown to be closely related to as long as the temperature-dependence and “concen-
SEI or surface film properties, although its physical tration-dependence” of Rct remain unchanged, the
significance remains ambiguous,516 a fact beyond relative magnitude of Rct versus RSEI should not
doubt is that this rate-determining component is reverse the conclusion by Zhang et al. Nevertheless,
under the influence of electrolyte composition, as in such situations it seems to be necessary to define
evidenced by Smart et al.’s success.515 On the other a new quantity that would more accurately describe
hand, the suggestion by Huang et al. that the the overall kinetics of lithium ion intercalation/
electrolyte and SEI do not control the low-tempera- deintercalation in the surface film. The interface
ture performance of lithium ion cells seems to be exchange current derived from various polarization
challenged in these electrode/electrolyte systems. techniques that has been used by Smart et al. could
It must be noted here that the observation made be such a quantity, because both RSEI and Rct are
by Zhang et al. about the dominating magnitude of taken into account.
Rct could be electrode-specific because Rct is not In most of the situations discussed above, the
always higher than RSEI in all systems and at all graphite anode was investigated as the single and
temperatures. At least in the works by Smart et isolated component, while no or little consideration
al.461,462 and Wang et al.,511 Rct in magnitude is was given to the other components where simulta-
comparable with or even smaller than RSEI. Certainly, neous electrochemical processes occurred. In a full
Nyquist plots obtained in the EIS analysis of the first

generation of Sony 18650 cells, he assigned the larger
semicircle at lower frequency to the anode/electrolyte
interface and the smaller at medium frequency to the
cathode/electrolyte interface on the basis that the
diameter of the former relies on the state-of-charge
of the anode, while the latter becomes larger in
diameter when the surface area of the LiCoO2
cathode decreases. Thus, he concluded that the
charge-transfer process occurring at the anode/
electrolyte interface is the slowest step for the whole
cell chemistry. However, since the EIS study was
conducted on the lithium ion cell without an inde-
pendent reference electrode, the attempt to separate
the contribution from each individual electrode to the
overall cell impedance is deemed unreliable. His way
of assigning the two semicircles to individual elec-
trodes was also questionable, since it is well accepted
that these semicircles correspond to different pro-
cesses rather than different electrode interfaces.134,462
As has been shown in Figure 68, since the time
constants for these two electrochemical components,
RSEI and Rct, are comparable at anode/electrolyte and
cathode/electrolyte interfaces, respectively, the im-
pedance spectra of a full lithium ion could have
similar features in which the higher frequency semi-
circle corresponds to the surface films on both the
anode and the cathode, and the other at lower
frequency corresponds to the charge-transfer pro-
cesses occurring at both the anode and the cathode.512
By incorporating an independent lithium reference
Figure 70. Asymmetrical behavior of Rct toward lithiation electrode in a commercial lithium ion cell from A&T,
and delithiation. Comparison between the Nyquist plots Nagasubramanian managed to separate the contri-
of the anode and cathode symmetrical cells at different butions from the anode and cathode to cell impedance
states-of-charge: (a) graphite/graphite; (b) cathode/cathode. and established their individual temperature pro-
Copyright 2003 Elsevier.) files.506 He found that the increase in cell impedance
with decreasing temperature mostly came from the
lithium ion cell, which integrates all of those com- cathode/electrolyte interface, while the contributions
ponents, including electrolyte, separator, cathode, from the anode/electrolyte interface or the bulk
and anode, the identification of the limiting factor resistance were negligible. Similar impedance be-
for low-temperature performance becomes more com- havior was also observed in commercial cells from
plicated, and conflicting conclusions have been re- Moli and Panasonic.506 Thus, he concluded that the
ported from various authors. Clearly, these conclu- interfacial resistance at the cathode, which should
sions are highly conditional and most likely system- include both RSEI and Rct, is mainly responsible for
specific; therefore, caution should always be taken the poor cell performance at low temperatures. This
when interpreting the results of these studies. conclusion was supported by Chen et al., who inves-
Conceptually, besides SEI and charge-transfer tigated the 18650 lithium ion cell assembled by
resistance in the interfaces as well as lithium diffu- Polystor using EIS and identified charge-transfer
sion coefficients in the electrodes, the possible limit- resistance at the LiNi0.8Co0.2O2 cathode/electrolyte
ing factors that might affect the kinetics of lithium interface as the main contributor of overall cell
ion chemistry could also be electrode surface area and impedance, although the study was only carried out
porosity, electrode density and loading, affinity of at room temperatures.518 An indirect evidence for this
binder toward electrolyte, and separator porosity and conclusion came from the XPS studies conducted by
lipophilicity. Any material or engineering flaw could Andersson et al. on the cathode and anode surface
make any of these factors the rate-determining step from the same lithium ion cell, which showed that
of the kinetics; hence, it is not strange that different the surface film on the cathode is much thicker
limiting factors have been identified for different compared with the anode SEI.294
lithium ion systems when low-temperature perfor- The opposite conclusion was reported by Lin et
mance was investigated. al.,165 who used a three-electrode configuration to
Ozawa was perhaps the first author who tried to study the electrode polarization of the MCMB anode
determine whether the anode or cathode acts as the and LiCoO2 cathode under galvanostatic conditions.
kinetic bottleneck in a full lithium ion cell, although They found that in all cases the polarization at the
the purpose at the time was not intended for low- MCMB anode surface far outweighs that at the
temperature considerations.157 In interpreting the cathode to such an extent that the potential profile
determining step in the temperature ranges below

-40 °C if other factors such as the precipitation of
bulk electrolyte components do not intervene.
The EIS studies using symmetrical cells by Zhang
et al. presented a third answer to the question about
whether the cathode or the anode is the limiting
factor.512 They showed that the controlling factor for
each electrode, Rct, is comparable in magnitude at
room temperature for a graphite as well as a LiNiO2
electrode (Figure 68); and with decreasing tempera-
tures down to -20 °C, these charge-transfer quanti-
ties also decline following a matching profile (Figure
69c). Therefore, merely on the basis of the impedance
measurement, it would be difficult to tell which
interface is rate-determining. Thus, the authors
Figure 71. Tafel polarization measurement at different concluded that charge-transfer processes in both the
temperatures performed on MCMB and LiNiCoO2 elec- anode and the cathode limit the capacity utilization
trodes in 1.0 M LiPF6/EC/DEC/DMC/EMC (1:1:1:3). (Re- at subambient temperatures.
produced with permission from ref 515 (Figure 8). Copy-
right 2003 Elsevier.) 8.4.2. High-Temperature Performance
of the full lithium ion cell actually mirrors that of Compared with the efforts spent on the low-
the anode. This dominance of anode polarization temperature performance, less attention has been
becomes even more severe at low temperatures, and paid to the applications of lithium ion technology at
the logical conclusion should be that the kinetics at elevated temperatures, with perhaps storage stability
the graphitic anode are the rate-determining step. as the only exception. Cycling tests at temperatures
In a more recent survey of the commercial lithium above 50 °C have been rarely reported in the litera-
ion cells, Fan believed that the resistance of the ture, most likely owing to the chemical instability of
cathode surface layer was the factor limiting the low- LiPF6 in the organic solvents at elevated temperature
temperature performance of the cell chemistry.507 and the difficulty of replacing it with new lithium
Unlike the previous researchers who also believed salts.
that the cathode acted as the limiting factor,462,506,512 Using the thermally stable salt LiBOB, Xu et al.
Fan specifically excluded the role of Rct when he showed that a full lithium ion cell can operate at
made the identification, and his major arguments temperatures up to 70 °C with limited capacity
were based on the summary and interpretation of the fading, while LiPF6-based cells suffer obvious per-
previously published data on lithium diffusion coef- manent capacity loss, as shown in Figure 72.155,492
ficients in both the cathode and anode and the fact The authors believed that the chemical stability of
that the surface area of the cathode is normally only BOB anion and the absence of reactive decomposition
a fraction of the anode. Without data from direct products such as HF and PF5 confer upon the
measurement, the above speculation seems to be electrolyte stable performance at elevated tempera-
premature; hence, further experimental confirmation tures. Liu et al. also reported the improved cycling
is needed. performance of lithium ion cells based on the spinel
With the aim of gaining insight into the kinetics cathode and LiBOB-containing electrolyte at elevated
of lithium intercalation/deintercalation of both an- temperatures.490 Since Mn2+-dissolution caused by
odes and cathodes in novel low-temperature electro- the HF from LiPF6 has resulted in a severe capacity
lytes, Smart et al. carried out the Tafel polarization fading problem for spinel-based cathode materials,
experiments on MCMB and LiNiCoO2 electrodes as especially at elevated temperatures, the application
a function of temperature in a three-electrode cell, of this promising cathode material has been pre-
and the results are summarized in Figure 71. In the vented; therefore, the thermal and chemical stability
temperature range from ambient to -40 °C, the of it in the presence of LiBOB might have special
limiting current densities observed on the anode significance.
remained higher than those observed on the cathode. Another salt that is less sensitive to moisture than
In other words, under the condition of galvanostatic LiPF6, LiBF4, was also tested as an electrolyte solute
cycling at low temperatures, the cathode would be intended for high-temperature applications. Zhang
preferentially polarized and very likely serves as the et al. reported that electrolytes based on this salt
bottleneck for the kinetics of lithium ion chemistry. could allow the lithium ion cells to cycle at temper-
As pointed out earlier, these polarization current atures up to 70 °C.132 Irreversible reactions occurred
densities reflect the resistances of both the surface at temperatures above 80 °C, and the cells lost
film and charge-transfer on anodes and cathodes, capacity rapidly, which was accompanied by the rise
respectively. Interestingly, the temperature-depend- of cell impedance simultaneously.
ences established for the limiting current densities As the structurally modified version of LiPF6,
on anodes and cathodes showed that, as temperature LiFAP has also been reported to be less chemically
decreases, the gap between the polarizations at sensitive due to the partial replacement of fluorine
anode/electrolyte and cathode/electrolyte interfaces with the more stable perfluorinated alkyls.496,497
rapidly closes (Figure 71), predicting a switch of rate- Aurbach and co-workers investigated the stability of
Figure 73. Cycling of graphite anode half-cells at 60 °C

after extended cycling at room temperatures. 1.0 M LiFAP,
0.50 M LiFAP/0.50 M LiPF6, and 1.0 M LiPF6 (inset) in
EC/DEC/DMC (2:1:2) were used as electrolytes. (Repro-
duced with permission from ref 499a (Figure 6). Copyright
2003 The Electrochemical Society.)
8.5. Electrolytes of Low Flammability

As diluents of low viscosity for the electrolyte
solutions, the linear carbonates or esters were used
at high concentrations along with EC, and their high
flammability, as indicated by the low flash points
(Tables 1 and 2), has imparted the state-of-the-art
electrolyte systems with potential fire hazards in the
situations of various types of accidents and abuses.
Hence, these organic solvents serve as fuel for the
Figure 72. Capacity and capacity retention of the full various combustion processes, whether in open air
lithium ion cells based on graphite as anode, LiNiO2 as when cells are ruptured and accidental sparks ignite
cathode, and LiPF6 or LiBOB in EC/EMC as electrolytes: their vapors351 or under hermetic conditions where
(a) room temperature; (b) elevated temperatures. (Repro- overcharge or overheating trigger thermal run-
duced with permission from ref 155 (Figure 5). Copyright
away.340,341,360
The efforts aiming to improve the thermal safety
of electrolytes under these abuse conditions have
the electrolytes based on this salt at elevated tem- intensified in recent years, partially driven by the
peratures in anode half-cells.499a To ensure the com- more stringent safety requirements for large industry
plete formation of the SEI on the graphitic anode, lithium ion cells intended for electric vehicle (EV) or
all cells subjected to high-temperature testing were hybrid electric vehicle (HEV) applications. Unlike
preconditioned by cycling at room temperature. As consumer-size lithium ion cells with capacities below
the inset of Figure 73 shows, the half-cell containing 2 A h, the battery packs for the above applications
baseline electrolyte with LiPF6 as salt suffers an are usually based on individual cells of 10 A h or
immediate capacity loss at 60 °C, while the half-cells higher in capacity, and the dangerous combination
based on LiFAP solutions could be cycled at this of flammable solvents with the highly energetic
temperature for 50 times before deterioration hap- electrodes in such scaled-up systems would be se-
pened. Surprisingly, the most stable behavior was verely magnified by the amount of materials con-
demonstrated by the electrolyte based on the mixed tained in the cell. For example, a recent test, which
salt 0.50 M LiFAP/0.50 M LiPF6, which not only was designed as a close simulation of a car accident
delivered stable performance at 60 °C (Figure 73) but in real life, showed that violent fire and explosion
also showed higher capacity and lower fading rate ensued from the ignition of electrolyte vapors by an
even at 80 °C. As a rationale for the mixed salt effect, externally generated spark when a fully charged
these authors in their more recent work proposed lithium ion cell vented.519
that some possible bulk reactions occurred between Since any compromise in cell energy density for the
the two salts, one of which is the nucleophilic sake of safety would be undesired, most of the
substitution of the fluorinated alkyl by F-, forming research efforts were concentrated on the reformula-
new P-F bonds and HCF2CF3.499b Since the FAP- tion of the electrolytes by using a flame-retarding
anion in this reaction actually serves as HF-scaven- additive or cosolvent, with the goal that its presence,
ger, it should have a positive effect on both the kept at a minimum, could result in nonflammability
performance of the Li-graphite electrodes and the or at least retarded flammability of the whole elec-
thermal behavior of the solutions. trolyte system.
Figure 74. Improved thermal stability of an electrolyte by flame retardant HMPN: (a, left) DSC traces for baseline
electrolyte with (1.68%) and without HMPN in the presence of a fully lithiated graphite anode (Reproduced with permission
from ref 523 (Figure 5). Copyright 2000 The Electrochemical Society.); (b, right) SHR of baseline electrolyte with (10.0%)
and without HMPN in the presence of metallic lithium. (Reproduced with permission from ref 523 (Figure 6). Copyright
The early patent disclosures have claimed the to the nonflammability of HMPN with a mechanism
application of a wide spectrum of gas-evolving ingre- of passivating the surface of lithium. Because of its
dients and phosphorus-based organic molecules as small presence, HMPN does not show any negative
flame retarding additives in the electrolytes.520 impact on the electrochemical stabilities of the base-
Pyrocarbonates and phosphate esters were typical line electrolyte, as indicated by the cyclic voltamme-
examples of such compounds. The former have a try experiments as well as the preliminary cycling
strong tendency to release CO2, which hopefully could results in a cathode half-cell based on LiNi0.8Co0.2O2.
serve as both flame suppressant and SEI formation Similar improvements in thermal stability were
additive, while the latter represent the major candi- also made with a series of partially fluorinated esters
dates that have been well-known to the polymer by Yamaki et al.,465 who used these fluorinated esters
material and fireproofing industries.521,522 The elec- shown in Table 12 as the single solvents for electro-
trochemical properties of these flame retardants in lytes intended for the application of rechargeable
lithium ion environments were not described in these metallic lithium cells. Compared with their nonflu-
disclosures, but a close correlation was established orinated counterparts, the thermal stability of these
between the low flammability and low reactivity electrolytes against both metallic lithium and LiCoO2
toward metallic lithium electrodes for some of these cathode materials was significantly improved, and in
compounds. Further research published later con- both cases, the onset temperatures for the major
firmed that any reduction of flammability almost exothermic reactions were postponed to above 300 °C.
always leads to an improvement in thermal stability The authors believed that the higher stability of the
on a graphitic anode or metal oxide cathode. fluorinated esters against metallic lithium was re-
Prakash and co-workers described the first ap- lated to their chemical reactivity at room tempera-
plication of a flame retardant additive in an electro- ture with lithium, the reaction products of which
lyte solution that has been commonly used for lithium coated the surface of lithium with a thick SEI film
ion technology.523 By using a substituted cyclic phos- that delayed the reaction of the molten lithium at
phorimide, hexamethoxycyclophosphazene (HMPN), higher temperatures. On the other hand, since the
at additive levels (<10%), they investigated the intrinsic decomposition of LiCoO2 cathode materials
improvement in thermal stability of the baseline occurred at ∼300 °C, the authors attributed the
electrolyte LiPF6/EC/DMC by means of DSC and thermal event observed in the temperature range
ARC, respectively. As Figure 74a shows, the presence between 300 and 350 °C to the combination of the
of HMPN at concentrations below 2% resulted in a individual thermal decompositions of LiCoO2 and
dramatic reduction in the thermal reactivity of the electrolyte. The cycling efficiencies of metallic lithium
electrolyte toward the fully lithiated graphite anode, in those fluorinated esters were also studied, based
as indicated by the much lower heat generation as on which the authors proposed that rechargeable
well as the much delayed onset temperature of the cells using metallic lithium as anode might be practi-
major exothermic process. The improved stability was cal due to the stabilization effects of those solvents
also found in the self-heating of the electrolyte under on both the lithium anode and LiCoO2 cathode. No
adiabatic conditions in the presence of metallic data were reported on the reduction behavior of those
lithium, as Figure 74b shows. The maximum SHR solvents on graphitic anodes; therefore, the applica-
of the baseline electrolyte at 178 °C was apparently tion of these novel solvents in lithium ion cells
triggered by the melting of lithium (mp 180 °C), remains to be investigated.
which induced a high rate of self-heating at 0.68 °C Following the pioneering work by Prakash and co-
min-1, while the presence of 10% HMPN effectively workers,523 the researchers of the electrolytes for
suppressed this self-heating process, with a maxi- lithium ion cells started to concentrate their attention
mum SHR of 0.16 °C min-1 at 170 °C. The authors on the organophosphorus(V) compounds, and a num-
attributed these improvements in thermal stability ber of publications since 2001 marked this renewed
Table 14. Flame-Retarding Additives or Solvents
interest in nonflammability of electrolytes.524-532 The in that the effectiveness of a certain flame retardant
overwhelming choice of these compounds has cer- is proportional to the percentage of P in its molecule
tainly originated from the in-depth knowledge about and inversely to its bp, while, for the baseline
the combustion of organic materials accumulated in components, their flammability is proportional to
the polymer industry, where halogenated and their bp and inversely to the H content in their
organophosphorus compounds have been identified molecule. The lower flammability of the electrolyte
as the most effective flame retardants.521,522,533 Tra- formulated with TMP was also confirmed to yield
ditionally, two major models have been proposed to higher thermal stability by calorimetry tests, in
explain the flame retardation achieved by these which the thermal reaction between LiPF6/EC/DEC
compounds: (1) char-formation, which builds up a and the LiCoO2 cathode was apparently suppressed
thermal barrier between the condensed and gaseous due to the presence of 20% TMP.
phases, and (2) radical-scavenging, by which the Unfortunately, TMP was found to be cathodically
chain reaction is inhibited in the gaseous phase due unstable on a graphitic anode surface, where, in a
to the radical traps formed by the decomposition manner very similar to PC, it cointercalated into the
products of these additives.533 graphene structure at 1.20 V and then decomposed
Wang et al. seemed to favor the second mechanism to exfoliate the latter, although its anodic stability
when the effectiveness of organophosphate in flame did not seem to be a problem. For this reason, TMP
retardation was discussed in their work.524 Using has to be used in amounts less than 10% with EC
trimethyl phosphate (TMP, R1-3 ) CH3 in Table 14) and other carbonates in high concentration in order
as the flame retardant additive/cosolvent, they con- to achieve decent performance in lithium ion cells.
firmed that when TMP content was higher than a However, capacity fading caused by the increase of
certain threshold value, which depends on the flam- cell impedance cast doubt on the application of this
mability of the baseline solvents, the electrolyte could flame retardant in a lithium ion cell.524 To avoid the
be rendered nonflammable. On the basis of the poor cathodic stability of TMP on graphitic anodes,
previous findings that the radical species containing the possibility of using it with other amorphous
phosphate have been detected in the MS, they carbon electrodes was also explored by the authors.525
proposed that such radicals act like a trap to scav-
enge the main active agent for flame propagation, H• The above flame retardants, HMPN and TMP,
radicals. To estimate the minimum amount (Nlimit) along with another commercially available alkyl
of TMP needed in any binary electrolyte composition phosphate, triethyl phosphate (TEP), were system-
to achieve nonflammability, they even derived an atically characterized by Xu et al.526 To quantify the
empirical equation: flammability of the electrolytes so that the effective-
ness of these flame retardants could be compared on
[ ]
CPTH a more reliable basis, these authors modified a
log Nlimit ) 2.6 - 9.3 (14) standard test UL 94 HB, intended for solid polymer
CHTP samples, and measured the self-extinguishing time
(SET) instead of the universally used flame propaga-
The unitless quantity in eq 14, CPTH/CHTP, is the so- tion rate.520a Compared with the UL 94 HB, this new
called “nonflammability index” defined by the authors quantity is more appropriate for the evaluation of the
using the atom content of H or P in the two electro- electrolytes of low flammability, since the electrolytes
lyte components and their boiling points, respectively. that are determined to be “retarded” or “nonflam-
Qualitatively, this equation is of general significance mable” by this method all showed zero flame propa-
Figure 75. Flammability of the electrolytes containing

various phosphorus-based flame retardants (FR). All elec-
trolytes are composed of these flame retardants in 1.0 m
LiPF6/EC/EMC. (Reproduced with permission from ref 529
(Figure 1). Copyright 2003 The Electrochemical Society.)
gation rate. Since the SET thus obtained could be

normalized against the amount of electrolytes, the
reproducibility of flammability data was also im-
proved, and comparison between these different
flame retardants was made possible.
Figure 75 shows the dependence of the normalized
SET on the concentration of different additives/
cosolvents in a baseline electrolyte 1.0 M LiPF6/EC/
EMC (1:1). Apparently, the flammability decreases
steadily with the concentration of TMP, TEP, and
HMPN, but even in high concentrations, they fail to
render the electrolytes completely nonflammable.
In their further work, Xu et al. combined the effects
of the two different groups of flame retardants,
halogens and phosphates, into one molecule in the
hope that the integration would result in the im-
provement of flame retarding efficiency.527-530 For
this purpose, they synthesized a series of organo-
phosphates with partially fluorinated alkyls, which
included TFP, BMP, and TDP.
A brief summary of their physical properties is

provided in Table 14. Their effects on the flammabil- Figure 76. Cathodic stability on graphitic anode: Dif-
ity of the baseline electrolyte are also compared in ferential capacity vs voltage plots for anode and cathode
half-cells containing different concentrations of (a) TMP,
Figure 75. Apparently, the combination of fluorina- (b) TEP, and (c) HMPN. Only the first cycles are shown.
tion with phosphate structure has brought the ex- Concentrations are indicated in the graph with the Cou-
pected higher efficiency in suppressing burning. For lombic efficiency for each cycle in the parentheses. (Repro-
example, with 20% TFP the electrolyte was rendered duced with permission from ref 526 (Figure 2). Copyright
nonflammable. 2002 The Electrochemical Society.)
While all these phosphate-based cosolvents were
shown to be rather stable on various cathode materi- While confirming the results reported by Wang et al.
als, Xu et al. concentrated the evaluation effort on that TMP is cathodically instable, the authors also
the reduction behavior of these flame retardants at showed that longer alkyl substituents (as in TEP)
the surface of graphitic anode materials. Figure 76 improved cathodic stability, although, in long-term
shows the results obtained with electrolytes contain- cycling or storage at elevated temperatures, the
ing high concentrations of TMP, TEP, and HMPN. reductive decomposition of TEP still caused poor cell
performances. Among the three nonfluorinated co-

solvents, HMPN was the only exception: its presence
up to 40% did not have any negative impact on the
cathodic stability of the electrolyte on the graphitic
anode. Whether it also participated in the formation
of the SEI remained to be investigated.
The fluorinated phosphates, on the other hand,
showed generally improved cathodic stability, the
extent of which seemed to be proportional to the
content of fluorine in the molecule. The stability was
further confirmed by the storage test of these elec-
trolytes in a fully charged lithium ion cell at 60 °C.
The cycling of the lithium ion cells with these
electrolytes containing up to 40% TFP and BMP
showed little difference from the baseline electrolyte
at room temperature and at moderate drain rate,
while the extended cycling up to 300 times delivered
stable capacity with negligible fading. Like HMPN,
these flame retardants also seemed to be involved in
the formation of a more effective SEI on carbonaceous
anode materials because higher concentrations of
them always resulted in higher capacity retention in
the extended cycle life tests.
However, as Figure 77 shows, the presence of TFP
did render the lithium ion cells with lower rate
capability, although reformulation of the baseline
electrolyte with a higher content of the low viscosity
solvent could alleviate the loss in rate performance.
A similar trend was found with HMPN, which was
more viscous than TFP. Studies on ion conductivity
and EIS corresponded well with the above findings;
that is, the resistances associated with the ion
transport in both bulk electrolyte528 and the electrode/
electrolyte interfaces530 increased with the presence
of these flame retardants. In other words, even with
the electrochemically stable TFP and HMPN, the
improvement in low flammability has to be realized
at the expense of cell performance because the
flammability of a system is mainly a bulk property
and the reduction of it requires the use of flame
retardants in high concentrations. On the other hand,
Figure 77. Effect of TFP concentration in different
if the stringent standard of “nonflammability” set by baseline electrolytes on the rate performance of the lithium
Xu et al. is abandoned and thermal stability with ion cells: (a) 1.0 m LiPF6/EC/EMC (1:1); (b) 1.0 m LiPF6/
respect to electrode materials as measured by DSC PC/EC/EMC (3:3:4). (Reproduced with permission from ref
or ARC is adopted as the criterion, then these flame 530 (Figure 5). Copyright 2003 The Electrochemical Soci-
retardants could be used at much lower concentra- ety.)
tions, that is, as additives instead of cosolvents. This
Amine and co-workers reported similar organo-
difference has been shown by the work of Wang et
phosphates with larger alkyl (TBP) or aromatic (TPP)
al. and Xu et al.
substituents, as Table 14 shows.531 They adopted an
While the consideration of nonflammability and improved UL 94 vertical burning technique to evalu-
SEI stability favors a high concentration of these ate the efficiency of these flame retardants in the
organophosphorus compounds in electrolytes, the baseline 1.0 M LiPF6/EC/DEC (1:1) and tried to
capacity utilization, rate capabilities, and low-tem- correlate the flammability of the resultant electro-
perature operation require that they be used at lytes with their thermal stability against a lithiated
minimal concentrations. A compromise would be graphite anode under adiabatic conditions. They
reached between 15 and 20% TFP or BMP in a binary found that the flame propagation rate was signifi-
1.0 M LiPF6 in EC/EMC (1:1) system or at higher cantly reduced even with only 1% TPP, while the
than 30% in a ternary 1.0 M LiPF6 in PC/EC/EMC optimum content of it was 5%. The exothermic
(1:1:3) system. Such electrolytes are completely or at reactions of the electrolytes with the fully lithiated
least nearly nonflammable. To further alleviate the graphite were effectively suppressed, as indicated by
above tradeoff, Xu et al. suggested that new cosol- the lower SHR, lower heat generation, and higher
vents of higher flame retarding ability should be onset temperature, similar to the observations of
tailor-made.530 Prakash et al. and Wang et al. These authors
believed that the char-formation mechanism was

responsible for the delayed thermal decomposition.
Like all the phosphates investigated as cosolvents,
TBP and TPP showed higher anodic stability, as
confirmed by their cycling in lithium ion cells based
on a LiNi0.8Co0.2O2 cathode up to 4.2 V, and separate
cyclic voltammetry tests also showed that they would
not decompose anodically below 5.0 V on an inert
working electrode. Little capacity fading was detected
during the extended tests of TPP or TBP in full
lithium ion cells up to 150 cycles.
In view of the poor cathodic stability of TMP on a
graphitic anode, Ota et al. recently described a cyclic
phosphate ethylene ethyl phosphate (EEP) (Table 14)
to be used with TMP as the additive for the modifica-
tion of the SEI.532 They found that 5% EEP could Figure 78. Rate capability of graphite/LiCoO2 lithium ion
render the electrolyte 1.0 M LiPF6/EC/DEC/TMP (6: cells using nonflammable MFE-based electrolytes and EC/
2:2) stable on the graphite surface down to low EMC baseline. (Reproduced with permission from ref 534
potentials and support the reversible intercalation/ (Figure 6). Copyright 2003 The Electrochemical Society.)
deintercalation of lithium ions, while the electrolyte
1.0 M LiPF6/EC/DEC/TMP (6:2:2) without EEP of EC and LiPF6 led to the significant reductions in
steadily decomposed at 0.40 V in a manner similar not only Rct but also RSEI, while XPS and FT-IR
to that for PC. The decomposition plateau at a much confirmed that an increase in the amount of lithium
lower potential as compared with the case of neat alkyl carbonate was simultaneously observed on the
TMP solution, as reported by Wang et al.,524 should graphite surface. In other words, the newly formu-
be attributed to the high EC content used by Ota et lated nonflammable electrolyte based on MFE re-
al. in their electrolyte composition. The evidence from mained dependent on the use of EC and LiPF6. This
XPS and FT-IR seemed to confirm that EEP was probably showed from another angle why EC and
directly involved in the formation of the SEI, as both LiPF6 are indispensable components in the state-of-
spectra detected the signature stretching of P-O the-art electrolytes.
bonds that did not exist in EEP-free electrolyte. The One surprise in Arai’s results is perhaps the anodic
authors suggested that EEP underwent a ring- stability of the ether-based solvent MFE, which
opening decomposition while participating in the remained stable anodically at up to 4.2 V on the
surface chemistry on the graphite surface during LiCoO2 surface. Considering the numerous previous
charge. In temperature-programmed decomposition reports that the ether linkage is oxidatively decom-
mass spectra (TPD-MS), such an SEI was shown to posed near 4.0 V,74,75,78,93 one might attribute this
be more thermally stable even than the SEI formed additional stability of MFE to the fluorination of the
by the baseline electrolyte 1.0 M LiPF6/EC/DEC (1: molecule.
1): the former decomposes at 125 °C and the latter The MFE-based nonflammable electrolyte was also
at 110 °C. The authors summarized that EEP should tested in lithium ion cells using a spinel cathode.535
be a promising additive for lithium ion electrolyte The effects of EC and LiPF6 seemed to be distin-
because it showed the simultaneous merits of being guished: while the presence of EC improved the rate
flame retardant and an SEI modifier. capability of these lithium ion cells, longer cycle life
with higher capacity retention could only be achieved
Another non-phosphorus cosolvent was recently
by further addition of LiPF6. On the basis of the
described as the bulk component for a nonflammable
above results, the authors believed that MFE-based
electrolyte by Arai, who used a partially fluorinated
bulk electrolyte provided a better baseline in terms
ether, methyl nonafluorobutyl ether (MFE), to form
of safety and nonflammability, although further
a binary mixture solvent with the linear carbonate
optimization should be carried out on this new
EMC and reported the complete elimination of the
baseline.
flash point of the electrolyte thus-based (Table
14).534,535 The 18650 lithium ion cells using graphite/
LiCoO2 and 1.0 M LiBeti/MFE/EMC (8:2) delivered
8.6. Polymer and Polymer Gel Electrolytes
the same capacity at low drain rates (0.1 C) as the Polymer electrolytes intended for applications in
baseline electrolyte 1.0 M LiPF6/EC/EMC (3:7) but lithium-based cells could be roughly divided into two
showed no thermal runaway under the abusive major classifications: (1) those based on neat high
conditions of nail-penetration of an overcharged cell; polymers, which serve as both solvent to dissolve
however, the rate capability of this nonflammable lithium salts and mechanical matrix to support
electrolyte is very low. Galvanostatic and EIS studies processability,536-540 and (2) those based on polymers
on the corresponding cathode and anode half-cells gelled by conventional electrolyte solutions, wherein
identified that the highly resistive anode/electrolyte the small organic molecules serve as the major
interface was the source for this problem. With 0.5 solvents, while the low percentage of high polymer,
M EC and 0.1 M LiPF6 salt as additives in the MFE- fully swollen by these solvents, acts only to provide
based bulk electrolyte, the polarization was signifi- dimensional stability.541,542 In recent literature the
cantly reduced, as shown in Figure 78. The presence former is usually referred to as solid polymer elec-
trolytes (SPEs), and the latter as gel polymer elec- Scheme 30. Schematic Illustration of Lithium Ion
trolytes (GPEs). Due to the poor ion conductivities, Transport in Polyether Media (Redrawn from Ref
SPEs remain materials of mere academic interest, 536)
with only remote prospects for their applications. On
the other hand, GPEs have shown much higher
feasibility, and second generation lithium ion cells
from the major manufacturers have incorporated
these novel electrolytes. Apparently the proximity of
GPEs to the conventional nonaqueous electrolytic
solutions, in physical chemistry and electrochemistry,
is the main reason. associated with certain local structural relaxations
The current section does not intend to comprehen- related to the glass transition of the polymer.538 A
sively review SPE and GPE materials, as the preced- generally accepted model described a microscopic
ing sections do to liquid electrolytes, mainly because sequence in which lithium ions were coordinated by
of the fact that a number of recent reviews have the ether oxygen atoms on the segments of a poly-
covered this field thoroughly,41-48,536-542 the latest of meric chain in a similar way to their complexation
which was published in 2003,48b and also the fact that by crown ethers or other oligoether-based solvents.
the most promising systems, that is, GPEs, are A continuous segmental rearrangement accompanied
similar in many ways to the conventional liquid by the gradual replacement of the ligands in the
electrolytes, including their electrochemical stabili- solvation sheath of lithium ions, hence, resulted in
ties on various electrodes and the corresponding the long-range net displacement of lithium ions, as
interfacial chemistries. Rather, a more general over- Scheme 30 shows.539 Thus, these polymer-salt com-
view will be given on both systems with little detailed plexes may exhibit mechanical properties which are
discussion on any individual electrolyte, although similar in most ways to those of true solids due to
exceptions apply to a few selected systems that are chain entanglement of the polymer host, while the
of special significance. microscopic environment that a lithium ion sees
remains liquidlike and the ion conductivity is “coupled”
8.6.1. Solid Polymer Electrolyte to the local segmental motion of the polymer. The
latter property is often characterized by the glass
While Wright and co-workers were the first group transition temperature (Tg) of the polymer, which to
of researchers to discover that the ether-based poly- a great extent determines the mechanical strength
mer poly(ethylene oxide) (PEO) was able to dissolve and hence processability of a polymer material.
inorganic salts and exhibit ion conduction at room The above coupling between ion transport and
temperature,543 it was the suggestion from Armand mechanical properties has created a perpetual di-
et al. that placed these novel materials at the center lemma concerning the prospect of using these poly-
stage of lithium electrolyte research for more than a mer ion conductors, since the efforts to improve their
decade.544 The number of comprehensive reviews on ion conductivitiesstypically between 0.01 and 0.001
this subject could serve as an indicator of the general mS cm-1 at room temperatures and far short of
enthusiasm for these materials during the supporting normal operation in lithium (ion) cellss
period.41,47,536-542 were always met by sacrificing dimensional stability.
At the time, the targeted device was the recharge- Various new host polymers with low Tg were inves-
able cell using a metallic lithium anode, and the tigated, which included the more flexible polypropy-
worldwide interest was mainly stimulated by the lene oxide (PPO) hosts that remain completely amor-
following projected advantages that these polymer phousatroomtemperatures,thesiloxane-orphosphazene-
electrolytes could potentially offer when compared to based polymers whose low Tg renders the polymer
their liquid counterparts: (1) excellent processability host more flexible and facilitates faster ion conduc-
and flexibility that could enable the fabrication of tion, the comb-type polyethers in which ether link-
ultrathin lithium cells of various geometric shapes ages are attached as side-chains to the backbone so
so that high energy and power density could be that more flexible segmental motion is possible, the
achieved for versatile applications, (2) higher safety polymer composite electrolytes that have incorpo-
due to the absence of flammable organic solvents and rated inorganic additives to eliminate the crystallin-
the much lower reactivity of macromolecules toward ity of polyether segments, and so forth. However,
lithium, (3) the possible prevention of the growth of almost always before the ion conductivities reach the
lithium dendrite crystals upon cycling, and (4) the coveted goal of the “liquidlike” level (i.e., >1.0 mS
high dimensional stability that could lead to the cm-1), the mechanical strength of these polymers
elimination of a separator, so further improvement approaches “liquidlike” state, thus conflicting with
in both energy density and manufacturing cost could the fundamental reason polymer electrolytes are
be achieved due to the simplified cell configuration desired. The introduction of chemical cross-linking
and enhanced packing efficiency. or the preparation of various interpenetrating net-
Despite the initial misinterpretation in the ion works (IPNs) indeed improves the dimensional sta-
conduction mechanism, it was soon realized that the bility of polymer electrolytes, but the ion conductivity
ion conduction in PEO and other similar polyether- swings back to levels < 0.01 mS cm-1.
based media mainly occurred in the amorphous During the two decades following the initial pro-
phases. Increasing evidences were obtained that ionic posal by Armand, essentially little progress was made
motion in these polymer ion conductors was closely concerning the actual application of SPEs, where the
poor ion conduction exists as a seemingly insurmount- He argued that (1) the flexibility that a polymer
able barrier. Due to the low level ion conductivity, electrolyte could offer presents no true advantage to
little has been known about the electrochemical cell design, since the current cell components (i.e.,
stability of these polymer electrolytes on various electrodes coated on substrates and a separator based
intercalation-type electrode materials, and even less on polyolefin films) already possess sufficient flex-
about their cycling in actual lithium and lithium ion ibility, and limitations on the geometric shapes of
cells. On the basis of a general survey of the most lithium cells do not come from the rigidity of elec-
solid polymer electrolyte systems developed so far, trolytes but rather from the terminals/connectors as
an upper limit in ambient-temperature ion conduc- well as economical considerations; (2) the low reactiv-
tivity seems to exist at 0.1 mS cm-1 for any polymer ity of a polymer toward an electrode would very likely
host whose mechanical strength could afford the be counteracted by the micrometer thickness of such
formation of a free-standing film.538-541 Applications polymer materials, since the chances of an internal
at high temperature seem to be a solution to the poor short increase significantly (the necessity for thin
ion conductivities, but a high rate of degradation films arises from the poor ion conductivities of such
often deters the efforts, as indicated by the fading electrolyte materials and the need to provide suf-
capacities of such cells. ficient ion flux for the normal cell operation); (3) the
Various attempts have been made to break the safety of polymer electrolyte-based cells would be
coupling of ion transport from the polymeric segmen- further challenged by technical difficulties, since it
tal relaxation. Combining the projected merits of would be almost impossible to manufacture a poly-
SPEs and superionic glass electrolytes that have been mer film of micrometer thickness in large scale and
the subject of intensive study in solid-state ionics pinhole free; and (4) there would be hardly any
since the late 1970s, Angell and co-workers proposed improvement in the tolerance against mechanical
an innovative approach of “polymer-in-salt” to bypass abuse for the polymer electrolytes, since they do not
the poor ion conductivity of the “salt-in-polymer” offer superior mechanical strength to that of the
SPEs.545 In the proposed “ionic rubber”, a polymer presently used polyolefin separators.
material of high molecular weight is used only at In addition to the criticisms from Anderman, a
additive levels as a sheer provider of mechanical further challenge to the application of SPEs comes
strength and filming processability, whereas the bulk from their interfacial contact with the electrode
of the electrolytes consist of the highly conductive materials, which presents a far more severe problem
ionic liquid. Ideally, these materials should have a to the ion transport than the bulk ion conduction
Tg low enough to remain rubbery at room tempera- does. In liquid electrolytes, the electrodes are well
ture while preserving the lithium ion conductivities wetted and soaked, so that the electrode/electrolyte
similar to those of the typical ionic liquid systems interface is well extended into the porosity structure
because most of the ions would be free of coordination of the electrode; hence, the ion path is little affected
with polymer segments due to the small presence of by the tortuosity of the electrode materials. However,
the latter. Unfortunately, although the initial results the solid nature of the polymer would make it
with simple lithium salts confirmed the conceptual impossible to fill these voids with SPEs that would
feasibility of the approach, exceeding difficulties were have been accessible to the liquid electrolytes, even
encountered in the attempts to formulate a practical if the polymer film is cast on the electrode surface
system, which should be based on an ionic liquid or from a solution. Hence, the actual area of the
such mixtures that not only have low melting and interface could be close to the geometric area of the
glass-transition temperatures but also are inert to electrode, that is, only a fraction of the actual surface
the polymer additives at these temperatures as well area. The high interfacial impedance frequently
as retain electrochemical stability in a wide potential encountered in the electrochemical characterization
range.546,547 of SPEs should originate at least partially from this
More recently, Ingram et al.548 and Wright et al.549 reduced surface contact between electrode and elec-
independently tried to develop new polymer hosts trolyte. Since the porous structure is present in both
with secondary structures similar to that of a liquid electrodes in a lithium ion cell, the effect of interfacial
crystalline state, so that ion transport could occur impedances associated with SPEs would become
with a higher degree of freedom in the highly oriented more pronounced as compared with the case of
environments and become at least partially decoupled lithium cells in which only the cathode material is
from the polymer segmental relaxations. Ion conduc- porous.
tivities approaching “liquidlike” values have been Even after all the above issues, that is, mechanical
obtained on the condition that the liquid crystalline strength, ion conductivity, and interfacial resistance,
state could be maintained.549 However, the incorpora- have been resolved, SPEs still have to face the crucial
tion of these novel polymer materials in electrochemi- issue of surface chemistry on each electrode if the
cal devices remains to be tested. application is intended for lithium ion technology,
On the other hand, ion conductivity is not the only and there is no reason to be optimistic about their
obstacle that prevents the application of SPEs. In prospects.
1994, Anderman published a review highly critical Despite the wide variety of polymer hosts that have
of the prospects for the application of SPEs in been synthesized and tested, the fundamental chemi-
electrochemical devices, in which he questioned cal structures adopted for SPEs remain strictly ether-
almost all of the previously projected advantages based and are variations of the original oligo(ethylene
from the viewpoint of cell design and engineering.550 oxide) structure, primarily due to the fact that no
other chemical linkages are able to possess sufficient the oxidative or reductive degradations of polymer
solvation power for inorganic salts while remaining segments in a lithium or lithium ion cell environ-
amorphous and flexible at room temperature. The ment. Certainly, the concomitant consequence of the
inability of ether-based solvents, such as DME and low polymer presence in GPEs is the poorer mechan-
DEE, to form an effective SEI on carbonaceous ical strength as compared with the cases of the pure
anodes has been reported in numerous reports, where SPEs, and either chemical or physical cross-linking
porous and highly resistive SEIs originating from is frequently necessary for the dimensional stability
these solvents have been described. Moreover, the of such gel materials.
anodic stability of these ethers on cathode materials The recent advances of GPE materials have been
does not offer relief, since their oxidative decomposi- exhaustively covered in two general review articles
tion potentials have been reported to be well below up to 2000.541,542 Especially in the review by Song et
4.0 V, which falls in the reversible lithiation/delithia- al., a rather systematical coverage was given on the
tion range for most of the metal oxide cathode most popular GPE systems that were based on PEO,
materials and far below the oxidation limits of ∼5.5 poly(acrylonitrile) (PAN), poly(methyl methacrylate)
V set by carbonate-based solvents. Although the (PMMA), and poly(vinylidene fluoride) (PVdF), along
reduction in the reactivity of ether linkages could be with an excellent discussion of the fundamental
expected because of the giant molecular size of the aspects concerning the formation, morphological
polymer solvent, the long-term stability of such structure, and physical stability of GPE.541 Because
polyether materials sandwiched between a graphite of the wide variety of polymer and copolymer hosts
anode and a metal oxide cathode does raise concern. available in the market, the number of new GPE
This concern is indirectly supported by the fact that, systems reported in the literature is ever increasing,
among the few rare publications wherein the cycling especially in recent years, while the bulk electrolyte
behavior of SPEs was tested under either galvano- compositions and hence the fundamental electro-
static or voltammertic conditions, decent cycle life chemistry associated remain relatively constant. It
was obtained only when low-potential cathodes such is not the objective of the current section to review
as V2O5, TiS2, or polymeric electroactive materials those individual systems in detail.
were used, while the industry standard LiCoO2 or On the application front, GPE technology has been
other 4.0 V class cathode materials always rendered used by most of the major manufacturers of lithium
poor cycling performances. ion cells following the third generation lithium ion
In a brief summary, although the studies on SPEs cells of Sony, and an energy density of 180 W h K
remain of academic interest in the areas of materials g-1 was reported.2 The specific polymer hosts used
science and solid state ionics, their immediate ap- in commercial lithium ion cells were seldom reported
plication in any commercial electrochemical devices, in the open literature for obvious reasons. It should
especially in the state-of-the-art lithium ion industry, be mentioned here that, in marketing terms, these
seems to be remote. Considering the multilateral cells are often confusingly referred to as “polymer
challenges that SPE materials are facing, any iso- lithium ion cells”, giving the wrong impression that
lated breakthrough in an individual property might solvent-free SPEs are in use.
not be sufficient to justify replacement of the current The only commercial GPE cell that had been
electrolytes used in lithium ion cells. However, it described in the open literature was perhaps the
must be cautioned that the above judgment is strictly Bellcore/Telcordia technology based on a fluorinated
based on the status quo of lithium ion technology, polymer, PVdF, from which one could readily sense
which could change with the improvement of elec- that the key factor controlling the success of certain
trode materials. Hence, one should not draw prema- polymer hosts in lithium ion cells is no longer
ture conclusions about the fate of SPEs because, after material chemistry only, and that more often than
all, science development is full of unexpected. Recent not the processing and fabrication of the GPE plays
advances made by Scrosati and co-workers have shed the decisive role.556
a light of hope on the SPE applications in lithium/
lithium ion cells intended for elevated-temperature As pointed out by Song et al.,541 the significance of
environments, where the disadvantage of low ion the Bellcore technology rested with its technical
conductivity could be avoided.501,551-555 innovation rather than scientific discovery because
PVdF-based polymers as a host for GPE had been
8.6.2. Gel Polymer Electrolyte investigated by Feuillade et al.557 and Tsuchida et
al.558,559 as early as two decades ago. However, what
Compared with SPEs, GPEs are much closer to distinguished Bellcore technology not only from those
actual applications because they inherited the major early studies but more generally from all of the
properties from the bulk liquid electrolytes, including previous research in polymer electrolytes was that
ion conduction, electrochemical stability on both the traditional approach of preparing a homogeneous
carbonaceous anode and various metal oxide cathode GPE film was abandoned and replaced by a two-step
materials, safety, and tolerance against mechanical process, in which PVdF-hexafluoropropylene (HFP)
and electric abuses. Moreover, since the polymer at copolymer was first processed into a microporous film
additive levels only serves as the skeleton providing that could be assembled with cathode and anode films
dimensional integrity, the ether linkage is no longer in an ambient atmosphere, and then the whole cell
the sole choice of building block for its chemical assembly could be activated by liquid electrolytes in
structure, thus eliminating potential concerns over a similar manner, where the traditional polyolefin
the electrochemical stability that might arise from separator was wetted by liquid electrolyte. After
activation, the liquid electrolyte swells the originally

microporous film and eventually forms a GPE. Since
the only step that must be conducted in a moisture-
controlled environment is the electrolyte injection,
the advantages of this technology in terms of manu-
facturing cost and operational ease are apparent.
Perhaps more important than cost is the solution
to the crucial problem of interfacial contacts that
always plagues homogeneous GPE films prepared
from traditional approaches. Since both cathode and
anode composite materials are coated on their sub-
strates with the same PVdF-HFP copolymer as the
binder, the “in situ gellification” following the elec-
trolyte activation effectively fuses the three cell
components into an integrated multilayer wafer
without physical boundaries, so that the interfaces
between anode and electrolyte or cathode and elec-
trolyte are well extended into the porous structures
of these electrodes, with close similarity to the
interfaces that a liquid electrolyte would access.
Another important merit of the “in situ gellifica-
tion”, rarely mentioned by various authors in the
literature, is that the limitation on electrolyte com-
position can be relaxed. In the traditional process of
making a GPE, the liquid electrolyte has to be heated
with the polymer host to form the gel, during which
the thermal instability of the lithium salt (LiPF6 or
LiBF4) and the volatility of the solvents (DMC, EMC,
etc) could possibly cause the resultant GPE to deviate
from the desired composition or even to degrade. It
is for this reason that in most of the literature on
GPE the liquid electrolytes have to be based on LiIm,
LiBeti as salts, and EC/PC as solvents. In Bellcore
technology, on the contrary, the state-of-the-art
electrolytes, the typical of which is LiPF6/EC/DMC,
could be used, since gellification occurs only after the
cells are assembled.556
According to Tarascon and co-workers, the swelling
of PVdF-HFP by liquid electrolytes was never com-
plete due to the semicrystalline nature of the copoly-
mer, which tends to microphase-separate after the
activation by electrolyte. On the other hand, it is
those crystalline domains in the gelled PVdF-HFP
that provide mechanical integrity for the resultant
GPE.556 Thus, a dual phase structure was proposed
for the Bellcore GPE by some authors, wherein the
amorphous domain swollen by a liquid electrolyte
serves as the ion conduction phase, while tiny crys-
tallites act as dimensional stabilizer. Figure 79. SEM of methanol-extracted separators pre-
pared in the weight ratio of plasticizer/PVdF-HFP/SiO2 of
Apparently, the formation of the microporous struc- 5:3:2. The extracted plasticizer in parts a and c is DBP,
ture within the PVdF-HFP copolymer was of critical and that in part b is an oligomer with molecular weight
importance to the success of Bellcore technology, and 450. Note that the scale in parts a and b is 5.0 µm and
the ion conductivity was proportional to the uptake that in part c is 100 nm. (Reproduced with permission from
of the liquid electrolyte. To achieve the desired ref 561 (Figure 2). Copyright 2000 The Electrochemical
Society.)
porosity of PVdF film, Bellcore researchers prepared
the initial polymer blend of PVdF with a plasticizer the distribution of pore sizes, their interconnectivity,
dibutylphthalate (DBP), which was then extracted by and the affinity of the inner wall of pores toward
low boiling solvents after film formation. Thus, a liquid electrolytes would all influence the final uptake
pore-memory would be left by the voids that were of electrolytes.560,561
previously occupied by DBP. However, due to the Improvements based on Bellcore technology were
incomplete dissolution of such high-melting DBP reported recently by Wunder and co-workers.561,562
during the extraction process, the pore-memory could Using PEO oligomers instead of DBP, they obtained
never be restored at 100% efficiency.556,559 Beside the the PVdF-HFP microporous films with the pore size
total volume of pores thus created by the plasticizer, increased from nanoscale to microscale, as shown in
atures so that it is not responsible for the thermal

runaway induced by overcharge. According to the
investigations by Maleki et al.360,361 and Dahn et
al.,369-371 cathode/electrolyte interaction, instead of
anode/electrolyte interaction, should be the main
source of the hazard to safe operation.
9. Concluding Remarks
The traditional role of electrolytes in batteries is
of less consequence than the electrode materials.
While the recent understandings about electrolytes
in lithium-based rechargeable cells revealed that the
Figure 80. Cycle life of 2 A h cells (A and B) and a 25 A chemical composition of electrolytes profoundly af-
h cell (C) at 100% depth-of-charge at ambient temperature. fects the cell performance in many aspectssfrom
(Reproduced with permission from ref 563 (Figure 7). lithium cycling efficiency to rate capability, and from
Copyright 2001 Elsevier.) capacity retention at various temperatures to toler-
ance against abusessthe fact remains that the choice
Figure 79. As a result, the ion conductivity rose by of electrolyte components is dictated by the electrode
∼70% from 0.2 mS cm-1 of DBP-extracted GPE to materials in use. Thus, the current state-of-the-art
1.2 mS cm-1 of the oligomer-extracted counterpart. electrolyte systems in lithium ion cells are tailor-
The cycling tests in a lithium ion cell based on made for the specific cell chemistry in which the
MCMB/LiCoO2 showed an improvement in capacity various electrolyte components yield ad hoc surface
by 40% and in rate capability by 70%. In their further chemistries on metal oxide cathodes and especially
work, Wunder and co-workers used polystyrene (PS) on graphitic anodes. Radical change in electrolyte
to form a blend with PVdF-HFP copolymer, which, composition is not expected to be likely, as long as
via a “phase-inversion” technique, yielded a mi- the cell chemistry remains the reversible shuttling
croporous film that could possess an ion conductivity of lithium ions between those metal oxides and
as high as 4.0 mS cm-1.562 Characterization using graphitic carbons, but innovations will continue to
various analytic means led the authors to conclude be carried out concerning the rate performance,
that the presence of PS increases pore volume, temperature range of service, safety of scaled-up cells,
wherein the liquid electrolyte forms a highly conduc- and “solidification” of the electrolyte.
tive path, while the swollen PVdF-HFP region is On the other hand, the emergence of new cell
relatively resistive, corresponding to the low ion chemistries would call for reformulation of the elec-
conductivities found in the GPE based on homoge- trolytes, as has always been the case during the
neous PVdF by previous investigators.557-559 At this history of lithium battery development. There have
stage, the demarcation between GPE and the tradi- been numerous such efforts in seeking more energetic
tional microporous separators based on polyolefin or safer cathode materials, which include the novel
materials becomes quite ambiguous. 5.0 V class mixed metal oxides and the more recent
The Bellcore GPE technology, bearing the brand olivine iron phosphate family, while metallic lithium
name of “plastic lithium ion cells” or “PLion cells”, remains a main attraction as a potential anode
has been licensed worldwide to various battery considering its tremendous energy density. In the
manufacturers, and a recent report by Han et al. latter case, electrolyte reformulation has been used
described the adaptation of it to large lithium ion as a main factor in controlling the undesired forma-
cells designed as traction power in electric vehicle tion of dendritic lithium during the long-term cy-
applications.563 Figure 80 shows the cycle life of such clings, an example of which is the work of Aurbach
cells at room temperatures, while tests of an im- et al., who found ad hoc surface chemistry on metallic
proved 28 A h cell carried out between 60 and -20 lithium provided by an ether-based electrolyte similar
°C yielded 95% and 60% of the rated capacity, to that of a carbonaceous anode.14,18 Polymer-based
respectively. These cells also passed all the safety electrolytes are also expected to be effective in
tests, including external short circuit, high current regulating the surface morphology of metallic lithium.
charge, impact, nail-penetration, and heating, with Upon the successful solution of the safety issue of the
the exception of overcharge, where the cells became lithium electrode, the revival of the lithium cell is a
swollen and burst into flames between 105% and possibility, along with new electrolyte formulations
150% of overcharged capacities. Since there is always that cater to the new cell chemistries.
a concern over the stability of the C-F bond in PVdF
polymers with respect to the fully lithiated carbon- 10. Acknowledgments
aceous anode,338 there is doubt once again about the
appropriateness of fluorinated polymers as GPE The author wants to thank Dr. Wu Xu (Arizona
hosts for lithium-based cell applications. However, State University) and Ms. Deborah Funk (Army
it should be kept in mind that the lithium ion cells Research Laboratory) for the meticulous and critical
based on liquid electrolytes behave similarly under reading of the manuscript, and Dr. Vera Zhuang
the same abusive conditions. A more likely situation (Lawrence Berkeley National Laboratory) for the
is that the reaction of the fluorinated polymers with invaluable input concerning the spectroscopic char-
carbonaceous anodes occurs at much higher temper- acterization of surface chemistry; Dr. Shengshui
Zhang (U.S. Army Research Laboratory) offered (35) Dahn, J. R.; von Sacken, U.; Fong, R. Extended Abstracts of the
178th Electrochemical Society Meeting, Seattle, WA, Oct. 14-
assistance in locating numerous literature sources, 19, 1990; Abstract No. 42; Electrochemical Society: Pennington,
and the discussions with Dr. Zhang, Dr. Jan Allen NJ.
(U.S. Army Research Laboratory), and Professor (36) Fong, R.; von Sacken, U.; Dahn, J. R. J. Electrochem. Soc. 1990,
137, 2009.
Takeshi Abe (Kyoto University) were helpful to the (37) Peled, E. J. Electrochem. Soc. 1979, 126, 2047.
author. Special acknowledgments should be given to (38) Brodd, J. Extended Abstracts of the 201st Electrochemical Society
Dr. Richard Jow (U.S. Army Research Laboratory) Meeting, Philadelphia, PA, May 12-17, 2002; Abstract No. 259;
Electrochemical Society: Pennington, NJ.
and Professor C. Austen Angell (Arizona State Uni- (39) Hannon, D. Purchasing Magazine 2002, Feb. 21.
versity) for the encouragement and discussions, and (40) Broussely, M. 11th International Meeting on Lithium Batteries,
Monterey, CA, June 23-28, 2002; Abstract No. 415; Electro-
Dr. Wishvender Behl (U.S. Army Research Labora- chemical Society: Pennington, NJ.
tory) for granting me access to his complete collection (41) Megahed, S.; Scrosati, B. J. Power Sources 1994, 51, 79.
of electrochemistry literature. (42) Brandt, K. Solid State Ionics 1994, 69, 173.
(43) Sawai, K.; Iwakoshi, Y.; Ohzuhu, T. Solid State Ionics 1994, 69,
273.
11. References (44) Guyomard, D.; Tarascon, J. M. Solid State Ionics 1994, 69, 222.
(45) Ebner, W.; Fouchard, D.; Xie, L. Solid State Ionics 1994, 69, 238.
(1) For fundamentals of modern electrodics, see: Bockris, J. O’M.; (46) Arora, P.; White, R. E.; Doyle, M. J. Electrochem. Soc. 1998, 145,
Reddy, A. K. N. Modern Electrochemistry, 2nd ed.; Plenum 3647.
Press: New York, 2000; Vol. 2. (47) Tarascon, J. M.; Armand, M. Nature 2001, 414, 359.
(2) For elementary readings concerning battery technology and (48) (a) Blomgren, G. E. J. Power Sources 1999, 81/82, 112. (b)
chemistry, see: Linden, D., Ed. Handbook of Batteries, 2nd ed.; Blomgren, G. E. J. Power Sources 2003, 119/121, 326.
McGraw-Hill: New York, 1995; 3rd ed.; McGraw-Hill: New (49) Fry, A. J. Synthetic Organic Electrochemistry, 2nd ed.; John
York, 2001. Wiley: London, 1989.
(3) The redox potentials against the standard hydrogen electrode (50) The physical data were extracted and compiled from the follow-
(SHE) for various reactions, usually called “electrochemical ing literature sources: (a) Jang, G. J.; Tomkins, R. P. T.
series”, are listed in the annually updated CRC Handbook of Nonaqueous Electrolytes Handbook; Academic Press: New York,
Chemistry and Physics; CRC Press: Boca Raton, FL. On the 1972; Vol. 1. (b) Dudley, J. T.; Wilkinson, D. P.; Thomas, G.;
other hand, unless otherwise indicated, the potential values used LeVae, R.; Woo, S.; Blom, H.; Horvath, C.; Juzkow, M. W.; Denis,
in this review are all referred to the Li+/Li reference electrode. B.; Juric, P.; Aghakian, P.; Dahn, J. R. J. Power Sources 1991,
(4) Jasinski, R. High Energy Batteries; Plenum Press: New York, 35, 59. (c) Aldrich Handbook of Fine Chemicals and Laboratory
1967. Equipment; Aldrich Chemical Co.: 2003-2004. (d) Ue, M.; Ida,
(5) Fraioly, A. V.; Barber, W. A.; Feldman, A. M. U.S. Patent K.; Mori, S. J. Electrochem. Soc. 1994, 141, 2989. (e) Ding, M.
3,551,205, 1970. S.; Xu, K.; Zhang, S.; Jow, T. R. J. Electrochem. Soc. 2001, 148,
(6) Schneider, A. A.; Moser, J. U.S. Patent 3,674,562, 1972. A299.
(7) Moser, J. U.S. Patent 3,660,163, 1972. (51) Harris, W. S. Electrochemical Studies in Cyclic Esters. Ph.D.
(8) Fukuda, M.; Ijima, T. In Power Sources; Collins, D. H., Ed.; Thesis, University of California, Berkeley, CA, 1958.
Academic Press: London, 1975; Vol. 5, p 713. (52) Jasinski, R.; Kirkland, S. Anal. Chem. 1967, 39, 1663.
(9) Selim, R.; Bro, P. J. Electrochem. Soc. 1974, 121, 1457. (53) Jasinski, R.; Carroll, S. Anal. Chem. 1968, 40, 1908.
(10) Raul, R. D.; Brummer, S. B. Electrochim. Acta 1977, 22, 75. (54) Rauh, R. D.; Reise, T. F.; Brummer, S. B. J. Electrochem. Soc.
(11) Koch, V. R.; Young, J. H. J. Electrochem. Soc. 1978, 125, 1371. 1978, 125, 186.
(12) Yoshimatsu, I.; Hirai, T.; Yamaki, J. J. Electrochem. Soc. 1988, (55) Aurbach, D.; Daroux, M. L.; Faguy, P. W.; Yeager, E. J.
135, 2422. Electrochem. Soc. 1987, 134, 1611.
(13) Broadhead, J.; Trumbore, F. A. In Power Sources; Collins, D. (56) Dey, A. N. Thin Solid Films 1977, 43, 131.
H., Ed.; Academic Press: London, 1975; Vol. 5, p 661. (57) Aurbach, D.; Gofer, Y.; Langsam, J. J. Electrochem. Soc. 1989,
(14) Dan, P.; Mengeritsky, E.; Geronov, Y.; Aurbach, D.; Weissman, 136, 3198.
I. J. Power Sources 1995, 54, 143. (58) Newman, G. H.; Francis, R. W.; Gaines, L. H.; Rao, B. M. J.
(15) Aurbach, D.; Gofer, Y.; Ben-Zion, M. J. Power Sources 1992, 39, Electrochem. Soc. 1980, 127, 2025.
163. (59) Rauh, R. D.; Brummer, S. B. Electrochim. Acta 1977, 22, 85.
(16) Aurbach, D.; Weissman, I.; Zaban, A.; Ein-Eli, Y.; Mengeritsky, (60) Koch, V. R.; Young, J. H. J. Electrochem. Soc. 1978, 125, 1371.
E.; Dan, P. J. Electrochem. Soc. 1996, 143, 2110. (61) Koch, V. R. J. Electrochem. Soc. 1979, 126, 181
(17) Aurbach, D.; Zaban, A.; Ein-Eli, Y.; Weissman, I.; Chusid, O.; (62) Koch, V. R.; Goldman, J. L.; Mattos, C. J.; Mulvaney, M. J.
Markovski, B.; Levi, M. D.; Levi, E.; Schechter, A.; Granot, E. Electrochem. Soc. 1982, 129, 1.
J. Power Sources 1997, 86, 91. (63) Koch, V. R.; Young, J. H. Science 1979, 204, 499.
(18) Aurbach, D.; Zinigrad, E.; Teller, H.; Dan, P. J. Electrochem. (64) Goldman, J. L.; Mank, R. M.; Young, J. H.; Koch, V. R.; Young,
Soc. 2000, 147, 1274. J. H. J. Electrochem. Soc. 1980, 127, 1461.
(19) Dominey, L. A. In Nonaqueous Electrochemistry; Aurbach, D., (65) Desjardins, C. D.; Cadger, T. G.; Salter, R. S.; Donaldson, G.;
Ed.; Marcel-Dekker: New York, 1999; Chapter 8. Casey, E. J. J. Electrochem. Soc. 1985, 132, 529.
(20) Whittingham, M. S. Science 1976, 192, 1126. (66) Foos, J. S.; McVeigh, J. J. Electrochem. Soc. 1983, 130, 628.
(21) Whittingham, M. S. Prog. Solid State Chem. 1978, 12, 1. (67) Foos, J. S.; Stolki, T. J. J. Electrochem. Soc. 1988, 135, 2769.
(22) (a) Mizushima, K.; Jones, P. C.; Wiseman, P. J.; Goodenough, (68) Abraham, K. M.; Goldman, J. L.; Natwig, D. L. J. Electrochem.
J. B. Mater. Res. Bull. 1980, 15, 783. (b) Mizushima, K.; Jones, Soc. 1982, 129, 2404.
P. C.; Wiseman, P. J.; Goodenough, J. B. Solid State Ionics 1981, (69) Dampier, F. W. J. Electrochem. Soc. 1981, 128, 2501.
3/4, 171. (70) Abraham, K. M.; Pasquariello, D. M.; Martin, F. J. J. Electro-
(23) Murphy, D. W.; Christian, P. Science 1979, 205, 651. chem. Soc. 1986, 133, 661.
(24) Armand, M. In Materials for Advanced Batteries; Murphy, D. (71) Yoshimatsu, I.; Hirai, T.; Yamaki, J. J. Electrochem. Soc. 1988,
W., Broadhead, J., Steele, B. C. H., Eds.; Plenum Press: New 135, 2422.
York, 1980; p 145. (72) Abraham, K. M.; Foos, J. S.; Goldman, J. L. J. Electrochem. Soc.
(25) Lazzari, M.; Scrosati, B. J. Electrochem. Soc. 1980, 127, 773. 1984, 131, 2197.
(26) Auborn, J.; Barberio, Y. L. J. Electrochem. Soc. 1987, 134, 638. (73) Geronov, Y.; Puresheva, B.; Moshtev, R. V.; Zlatilova, P.; Kosev,
(27) Semkov, K. W.; Sammels, A. F. J. Electrochem. Soc. 1987, 134, T.; Staynov, Z.; Pistoia, G.; Pasquali, M. J. Electrochem. Soc.
766. 1990, 137, 3338.
(28) Scrosati, B. J. Electrochem. Soc. 1992, 139, 2776. (74) Campbell, S. A.; Bowes, C.; McMillan, R. S. J. Electroanal. Chem.
(29) Nagaura, T.; Nagamine, M.; Tanabe, I.; Miyamoto, N. Prog. 1990, 284, 195.
Batteries Sol. Cells 1989, 8, 84. (75) Xu, K.; Angell, C. A. J. Electrochem. Soc. 1998, 145, L70.
(30) Nagaura, T.; Ozawa, K. Prog. Batteries Sol. Cells 1990, 9, 209. (76) Xu, K.; Ding, S.; Jow, T. R. J. Electrochem. Soc. 1999, 146, 4172.
(31) Nishi, Y.; Azuma, H.; Omaru, A. U.S. Patent 4,959,281, 1990. (77) Simon, B.; Boeuve, J. P.; Broussely, M. J. Power Sources 1993,
(32) Ohzhku, T.; Ueda, A.; Nagayama, M. J. Electrochem. Soc. 1993, 43/44, 65.
140, 1862. (78) Plichta, E.; Slane, S.; Uchiyama, M.; Salomon, M.; Chua, D.;
(33) Bittihn, R.; Herr, R.; Hoge, D. J. Power Sources 1993, 43/44, Ebner, W. B.; Lin, H. W. J. Electrochem. Soc. 1989, 136, 1865.
223. (79) Uchida, I.; Sato, H. J. Electrochem. Soc. 1995, 142, L139.
(34) Nagaura, T. Proceedings of the 5th International Seminar on (80) Angell, C. A.; Choi, Y. J. Microsc. 1986, 141, 251.
Lithium Battery Technology and Applications, Deerfield Beach, (81) Xu, K.; Ding, S.; Jow, T. R. J. Electrochem. Soc. 2001, 148, A267.
FL, March 5-7, 1990; Florida Educational Seminars, Inc.: Boca (82) Elliott, W. Report No. 1, Contract NAS 3-6015 (N 65-11518), Sept
Raton, FL. 1964.
(83) Pistoia, G.; De Rossi, M.; Scrosati, B. J. Electrochem. Soc. 1970, (130) Ue, M.; Takeda, M.; Takehara, M.; Mori, S. J. Electrochem. Soc.
117, 500. 1997, 144, 2684.
(84) Pistoia, G. J. Electrochem. Soc. 1971, 118, 153. (131) Ue, M.; Murakami, A.; Nakamura, S. J. Electrochem. Soc. 2002,
(85) Subbarao, S.; Shen, D. H.; Deligiannis, F.; Huang, C. K.; Halpert, 149, A1572.
G. J. Power Sources 1990, 29, 579. (132) Zhang, S. S.; Xu, K.; Jow, T. R. J. Electrochem. Soc. 2002, 149,
(86) Takami, N.; Ohsaki, T.; Inada, K. J. Electrochem. Soc. 1992, 139, A586.
1849. (133) Takami, N.; Ohsaki, T.; Hasebe, H.; Yamamoto, M. J. Electro-
(87) Arakawa, M.; Tobishima, S.; Hirai, T.; Yamaki, J. J. Electrochem. chem. Soc. 2002, 149, A9.
Soc. 1986, 133, 1527. (134) Zhang, S. S.; Xu, K.; Jow, T. R. Electrochem. Commun. 2002, 4,
(88) Surampudi, S.; Shen, D. H.; Huang, C. K.; Narayanan, S. R.; 928.
Attia, A.; Halpert, G.; Peled, E. J. Power Sources 1993, 43/44, (135) Zhang, S. S.; Xu, K.; Jow, T. R. J. Solid State Electrochem. 2003,
21. 7, 147.
(89) Tobishima, S.; Arakawa, M.; Hirai, T.; Yamaki, J. J. Power (136) Xu, K.; Angell, C. A. Electrochim. Acta 1995, 40, 2401.
Sources 1989, 26, 449. (137) Ito, K.; Ohno, H. Electrochim. Acta 1998, 43, 1247.
(90) McMillan, R. S.; Juskow, M. W. J. Electrochem. Soc. 1991, 138, (138) Croce, F.; D’Aprano, A.; Nanjundiah, C.; Koch, V. R.; Walker,
1556. C.; Salomon, M. J. Electrochem. Soc. 1996, 143, 154.
(91) Yamaura, J.; Ozaki, Y.; Morita, A.; Ohta, A. J. Power Sources (139) Naoi, K.; Mori, M.; Naruoka, Y.; Lamanna, W. M.; Atanasoski,
1993, 43/44, 233. R. J. Electrochem. Soc. 1999, 146, 462.
(92) Zhang, S. S.; Liu, Q. G.; Yang, L. L. J. Electrochem. Soc. 1993, (140) Ishikawa, M.; Morita, M.; Asao, M.; Matsuda, Y. J. Electrochem.
140, L107. Soc. 1994, 141, 1105.
(93) Guyomard, D.; Tarascon, J. M. J. Electrochem. Soc. 1992, 139, (141) Krause, L. J.; Lamanna, W.; Summerfield, J.; Engle, M.; Korba,
937. G.; Loch, R.; Atanasoski, R. J. Power Sources 1997, 68, 320.
(94) Tarascon, J. M.; Guyomard, D.; Baker, G. L. J. Power Sources (142) Foropoulos, J.; DesMarteau, D. D. Inorg. Chem. 1984, 23, 3720.
1993, 43/44, 689. (143) Armand, M.; Gorecki, W.; Andreani, R. Proceedings of the 2nd
(95) Dahn, J.; von Sacken, U.; Juzkow, M. W.; al-Janaby, H. J. International Meeting on Polymer Electrolytes; Scrosati, B., Ed.;
Electrochem. Soc. 1991, 138, 2207. Elsevier: London, 1989; p 91.
(96) Ohzuku, T.; Iwakoshi, Y.; Sawai, K. J. Electrochem. Soc. 1993, (144) Xu, K.; Day, N. D.; Angell, C. A. J. Electrochem. Soc. 1996, 143,
140, 2490. L209.
(97) Huang, S. Y.; Kavan, L.; Exnar, I.; Grätzel, M. J. Electrochem. (145) Wang, X.; Yasukawa, E.; Kasuya, S. J. Electrochem. Soc. 2000,
Soc. 1995, 142, L142. 147, 2421.
(98) Guyomard, D.; Tarascon, J. M. J. Electrochem. Soc. 1993, 140, (146) Dominey, L. A.; Koch, V. R.; Blakley, T. Electrochim. Acta 1992,
3071. 37, 1551.
(99) Tarascon, J. M.; Guyomard, D. Solid State Ionics 1994, 69, 293. (147) Yang, H.; Kwon, K.; Devine, T. M.; Evans, J. W. J. Electrochem.
(100) Zheng, T.; Liu, Y.; Fuller, E. W.; Tseng, S.; von Sacken, U.; Dahn, Soc. 2000, 147, 4399.
J. J. Electrochem. Soc. 1995, 142, 2581. (148) Kanamura, K.; Umegaki, T.; Shiraishi, S.; Ohashi, M.; Takehara,
(101) Morita, M.; Ichimura, T.; Ishikawa, M.; Matsuda, Y. J. Electro- Z. J. Electrochem. Soc. 2002, 149, A185.
chem. Soc. 1996, 143, L26. (149) Methlie, G. J. U.S. Patent 3,415,687, 1968.
(102) Ein-Eli, Y.; Thomas, S. R.; Chadha, R.; Blakley, T. J.; Koch, V. (150) Wiesboeck, R. A. U.S. Patent 3,654,330, 1972.
R. J. Electrochem. Soc. 1997, 144, 823. (151) Kawamura, T.; Kimura, A.; Egashira, M.; Okada, S.; Yamaki,
(103) Chu, A. C.; Josefowicz, J. Y.; Farrington, G. C. J. Electrochem. J. J. Power Sources 2001, 104, 260.
Soc. 1997, 144, 4161. (152) Sloop, S. E.; Pugh, J. K.; Wang, S.; Kerr, J. B.; Kinoshita, K.
(104) Aurbach, D.; Ein-Eli, Y.; Markovsky, B.; Zaban, A.; Luski, S.; Electrochem. Solid-State Lett. 2001, 4, A42.
Carmeli, Y.; Yamin, H. J. Electrochem. Soc. 1995, 142, 2882. (153) Behl, W. K.; Plichta, E. J. J. Power Sources 1998, 72, 132.
(105) Koetschau, I.; Richard, M. N.; Dahn, J. R.; Soupart, J. B.; (154) Behl, W. K.; Plichta, E. J. J. Power Sources 2000, 88, 192.
Rousche, J. C. J. Electrochem. Soc. 1995, 142, 2906. (155) Xu, K.; Zhang, S.; Jow, T. R.; Xu, W.; Angell, C. A. Electrochem.
Solid-State Lett. 2002, 4, A26.
(106) Peled, E.; Golodnitsky, D.; Menachem, C.; Bar-Tow, D. J.
(156) Zhang, S.; Jow, T. R. J. Power Sources 2002, 109, 458.
(157) Ozawa, K. Solid State Ionics 1994, 69, 212.
(107) Ein-Eli, Y.; Thomas, S. R.; Koch, V. R.; Aurbach, D.; Schecheter,
(158) Ding, M. S. J. Electrochem. Soc. 2003, 150, A455.
A.; Markovsky, B. J. Electrochem. Soc. 1996, 143, L273.
(159) Ding, M. S.; Xu, K.; Jow, T. R. J. Electrochem. Soc. 2000, 147,
(108) Ein-Eli, Y.; McDevitt, S. F.; Aurbach, D.; Markovsky, B.; 1688.
Schecheter, A. J. Electrochem. Soc. 1997, 144, L180. (160) Ding, M. S.; Xu, K.; Jow, T. R. J. Therm. Anal. Calorim. 2000,
(109) Sekai, K.; Azuma, H.; Omaru, A.; Fujita, S.; Imoto, H.; Endo, 62, 177.
T.; Yamaura, K.; Nishi, Y. J. Power Sources 1993, 43/44, 241. (161) Ding, M. S. J. Electrochem. Soc. 2002, 149, A1063.
(110) Linden, D. In Handbook of Batteries, 2nd ed.; Linden, D., Ed.; (162) Angell, C. A. Science 1995, 267, 1924.
McGraw-Hill: New York, 1995; Chapters 14 and 36. (163) Angell, C. A. Chem. Rev. 2002, 102, 2627.
(111) Ue, M. J. Electrochem. Soc. 1995, 142, 2577. (164) Plichta, E. J.; Behl, W. K. Proceedings of the 38th Power Sources
(112) Ravdel, B.; Abraham, K. M.; Gitzendanner, R.; Marsh, C. Conference, Cherry Hill, NJ, 1998; p 444.
Extended Abstracts of the 200th Electrochemical Society Meeting, (165) Lin, H. P.; Chua, D.; Salomon, M.; Shiao, H. C.; Hendrickson,
San Francisco, CA, Sept 2-7, 2001; Abstract No. 97; Electro- M.; Plichta, E.; Slane, S. Electrochem. Solid-State Lett. 2001, 4,
chemical Society: Pennington, NJ. A71.
(113) Xu, W.; Angell, C. A. Electrochem. Solid State Lett. 2001, 4, E1. (166) Ding, S. P.; Xu, K.; Zhang, S. S.; Jow, T. R.; Amine, K.;
(114) Hossain, S. Handbook of Batteries, 2nd ed.; Linden, D., Ed.; Henriksen, G. L. J. Electrochem. Soc. 1999, 146, 3974.
McGraw-Hill: New York, 1995; Chapter 36. (167) Liu, Z. K. J. Electrochem. Soc. 2003, 150, A359.
(115) Schmidt, M.; Heider, U.; Kuehner, A.; Oesten, R.; Jungnitz, M.; (168) Zhang, S. S.; Xu, K.; Allen, J. L.; Jow, T. R. J. Power Sources
Ignat’ev, N.; Sartori, P. J. Power Sources 2001, 97/98, 557. 2002, 110, 217.
(116) Walker, C. W.; Cox, J. D.; Salomon, M. J. Electrochem. Soc. 1996, (169) Blint, J. R. J. Electrochem. Soc. 1995, 142, 696.
143, L80. (170) Wang, Y.; Nakamura, S.; Ue, M.; Balbuena, P. B. J. Am. Chem.
(117) Aurbach, D.; Zaban, A.; Schecheter, A.; Ein-Eli, Y.; Zinigrad, E.; Soc. 2001, 123, 11708.
Markovsky, B. J. Electrochem. Soc. 1995, 142, 2873. (171) Yanase, S.; Oi, T. J. Nucl. Sci. Technol. 2002, 39, 1060.
(118) Aurbach, D.; Ein-Ely, Y.; Zaban, A. J. Electrochem. Soc. 1994, (172) Fukushima, T.; Matsuda, Y.; Hashimoto, H.; Arakawa, R.
141, L1. Electrochem. Solid-State Lett. 2001, 4, A127.
(119) Jasinski, R.; Carroll, S. J. Electrochem. Soc. 1970, 117, 218. (173) Matsuda, Y.; Fukushima, T.; Hashimoto, H.; Arakawa, R. J.
(120) Nanjundiah, C.; Goldman, J. L.; Dominey, L. A.; Koch, V. R. J. Electrochem. Soc. 2002, 149, A1045.
Electrochem. Soc. 1988, 135, 2914. (174) It should be pointed out that the simple summation of the
(121) Koch, V. R. J. Electrochem. Soc. 1979, 126, 181. contribution from each individual ionic species comes from the
(122) Webber, A. J. Electrochem. Soc. 1991, 138, 2586. assumption that the Kohlrausch independent ion rule, estab-
(123) Ein-Eli, Y.; Markovsky, B.; Aurbach, D.; Carmeli, Y.; Yamin, H.; lished for infinitely diluted solutions, still applies in the elec-
Luski, S. Electrochim. Acta 1994, 39, 2559. trolytes of practical salt concentrations (∼1.0 M), which is
(124) Aurbach, D.; Ein-Eli, Y. J. Electrochem. Soc. 1995, 142, 1746. incorrect due to the intensive interionic couplings in the latter.
(125) Gnanaraj, J. S.; Levi, M. D.; Levi, E.; Salitra, G.; Aurbach, D.; However, eq 1 remains qualitatively useful (see ref 1).
Fischer, J. E.; Claye, A. J. Electrochem. Soc. 2001, 148, A525. (175) For an excellent introductory reading on ac impedance tech-
(126) Matsuda, Y.; Morita, M.; Takada, K. J. Electrochem. Soc. 1984, niques for the purpose of ion conductivity measurement or study
131, 1991. of interfacial properties, please see: Linford, R. G. In Electro-
(127) Matsuda, Y.; Morita, M.; Yamashita, T. J. Electrochem. Soc. chemical Science and Technology of Polymers, 2nd ed.; Linford,
1984, 131, 2821. R. G., Ed.; Elsevier Applied Science: London, 1990; p 281.
(128) Takata, K.; Morita, M.; Matsuda, Y. J. Electrochem. Soc. 1985, (176) Blomgren, G. E. In Lithium Batteries; Gabano, J. P., Ed.;
132, 126. Academic Press: London, 1983.
(129) Ue, M. J. Electrochem. Soc. 1994, 141, 3336. (177) Tobishima, S.; Okada, T. Electrochim. Acta 1985, 30, 1715.
(178) Endo, E.; Ata, M.; Tanaka, K.; Sekai, K. J. Electrochem. Soc. (227) Gaberšček, M.; Pejovnik, S. J. Electrochem. Soc. 1999, 146, 933.
1998, 145, 3757. (228) Naoi, K.; Mori, M.; Shinagawa, Y. J. Electrochem. Soc. 1996,
(179) Tsuchida, E.; Kobayashi, N.; Ohno, H. Macromolecules 1988, 21, 143, 2517.
96. (229) Geronov, Y.; Schwager, F.; Muller, R. H. J. Electrochem. Soc.
(180) Zhou, G.; Khan, I. M.; Smid, J. Polym. Commun. 1989, 30, 52. 1982, 129, 1422.
(181) Zhang, S. S.; Wan, G. X. J. Appl. Polym. Sci. 1993, 48, 405. (230) Hérold, A. Bull. Soc. Chim. Fr. 1955, 187, 999.
(182) Tada, Y.; Sato, M.; Takeno, N.; Nakacho, Y.; Shigehara, K. Chem. (231) Guérard, D.; Hérold, A. Carbon 1975, 13, 337.
Mater. 1994, 6, 27. (232) Billaud, D.; Hérold, A. Carbon 1979, 17, 183.
(183) Robinson, R. A.; Stokes, R. H. Electrolyte Solutions; Butter- (233) Graphite Intercalation Compounds, Vol. I and II; Zabel, H., Solin,
worths: London, 1959; p 392. S. A., Eds.; Springer-Verlag: New York, 1990 and 1992.
(184) Some organic nitrogen-containing compounds (nitriles, ni- (234) Bagouin, M.; Guérard, D.; Hérold, A. C. R. Acad. Sci., Ser. C
troalkyls, amides, etc.) are the rare exceptions to this rule. For 1966, 262, 557.
example, for acetonitrile, ) 36 and η ) 0.3 cP at room (235) Guérard, D.; Hérold, A. C. R. Acad. Sci., Ser. C 1972, 275, 571.
temperature. However, the poor electrochemical stability of these (236) Avdeev, V. V.; Nalimova, V. A.; Semenenko, K. N. High-Pressure
compounds prevented them from being used in batteries. Res. 1990, 6, 11.
(185) Jasinski, R. Electrochem. Technol. 1968, 6, 28. (237) Dey, A. N.; Sullivan, B. P. J. Electrochem. Soc. 1970, 117, 222.
(186) Saito, T.; Ikeda, H.; Matsuda, Y.; Tamura, H. J. Appl. Electro- (238) Eichinger, G. J. Electroanal. Chem. 1976, 74, 183.
chem. 1976, 6, 85. (239) Besenhard, J. O. Carbon 1976, 14, 111.
(187) Matsuda, Y.; Satake, H. J. Electrochem. Soc. 1980, 127, 877. (240) Besenhard, J. O.; Möhwald, H.; Nickl, J. J. Carbon 1980, 18,
(188) Tobishima, S.; Yamaji, A. Electrochim. Acta 1984, 29, 267. 399.
(189) Tobishima, S.; Yamaji, A.; Okada, T. Electrochim. Acta 1984, (241) Besenhard, J. O.; Fritz, H. P Angew. Chem., Int. Ed. Engl. 1983,
29, 1471. 22, 950.
(190) Arakawa, M.; Tobishima, S.; Hirai, T.; Yamaki, J. J. Electrochem. (242) Besenhard, J. O.; Fritz, H. P. J. Electroanal. Chem. 1974, 53,
Soc. 1986, 133, 1527. 329.
(191) Tobishima, S.; Arakawa, M.; Hirai, T.; Yamaki, J. J. Power (243) Arakawa, M.; Yamaki, J. J. Electroanal. Chem. 1987, 219, 273.
Sources 1987, 20, 293. (244) Yazami, R.; Touzain, P. J. Power Sources 1983, 9, 365.
(192) Tobishima, S.; Arakawa, M.; Yamaki, J. Electrochim. Acta 1990, (245) Yazami, R.; Guérard, D. Power Sources 1993, 43/44, 39.
35, 383. (246) Kanno, R.; Takeda, Y.; Ichikawa, T.; Nakanishi, K.; Yamamoto,
(193) Matsuda, Y.; Nakashima, H.; Morita, M.; Takasu, Y. J. Electro- O. J. Power Sources 1989, 26, 535.
chem. Soc. 1981, 128, 2552. (247) Mohri, M.; Yanagisawa, N.; Tajima, Y.; Tanaka, H.; Mitate, T.;
(194) Matsuda, Y.; Morita, M.; Kosaka, K. J. Electrochem. Soc. 1983, Nakajima, S.; Yoshida, M.; Yoshimoto, Y.; Suzuki, T.; Wada, H.
130, 101. J. Power Sources 1989, 26, 545.
(195) Ding, M. S.; Xu, K.; Zhang, S. S.; Amine, K.; Henriksen, G. L.; (248) The configuration of the “half anode cell” enables the separate
Jow, T. R. J. Electrochem. Soc. 2001, 148, A1196. evaluation of the anode material alone; thus, it is widely used
(196) Ding, M. S.; Jow, T. R. J. Electrochem. Soc. 2003, 150, A620. as a convenient tool. However, it should be kept in mind that
(197) Blint, R. J. Electrochem. Soc. 1997, 144, 787. the “anode materials” under study are actually “cathode materi-
(198) Choquette, Y.; Brisard, G.; Parent, M.; Brouillette, D.; Perron, als” in this half-cell: Li|electrolyte|carbonaceous “anode”. To
G.; Desnoyers, J. E.; Armand, M.; Gravel, D.; Lougui, N. J. make things more confusing, different researchers use both
Electrochem. Soc. 1998, 145, 3500. “charge” (referring to this carbonaceous material in an imaginary
(199) Chen, H. P.; Fergus, J. W.; Jang, B. Z. J. Electrochem. Soc. 2000, full lithium ion cell, therefore, conceptually correct) and “dis-
147, 399. charge” (referring to this half-cell, therefore, operationally
(200) Smedley, S. I. The Interpretation of Ionic Conductivity in Liquids; correct) to describe the same process of lithiating the carbon
Plenum Press: New York, 1980; Chapter 3. electrode in this half-cell. Since there is no standard convention,
(201) Evans, D. H.; O’Connell, K. M.; Petersen, R. A.; Kelly, M. J. J. caution must be taken when electrochemical literature concern-
Chem. Educ. 1983, 60, 290. ing this topic is being studied. This review will use less
(202) Bard, A. J.; Faulkner, L. R. Electrochemical Methodss ambiguous terms such as “lithiation” or “delithiation” for the
Fundamentals and Applications, 2nd ed.; John Wiley & Sons: cycling of anode half-cells.
New York, 2000. (249) Chusid, O.; Ein Ely, Y.; Aurbach, D.; Babai, M.; Carmeli, Y. J.
(203) Ue, M.; Murakami, A.; Nakamura, S. J. Electrochem. Soc. 2002, Power Sources 1993, 43/44, 47.
149, A1572. (250) Aurbach, D.; Ein-Eli, Y.; Chusid, O.; Carmeli, Y.; Babai, M.;
(204) Eggert, G.; Heitbaum, J. Electrochim. Acta 1986, 31, 1443. Yamin, H. J. Electrochem. Soc. 1994, 141, 603.
(205) Guyomard, D.; Tarascon, J. M. J. Power Sources 1995, 54, 92. (251) Besenhard, J. O.; Winter, M.; Yang, J.; Biberacher, W. J. Power
(206) Koch, V. R.; Dominey, L. A.; Nanjundiah, C.; Ondrechen, M. J. Sources 1993, 54, 228.
J. Electrochem. Soc. 1996, 143, 798. (252) Dahn, J. R. Phys. Rev. B 1991, 44, 9170.
(207) Kita, F.; Kawakami, A.; Sonoda, T.; Kobayashi, H. In New Sealed (253) Jiang, Z.; Alamgier, M.; Abraham, K. M. J. Electrochem. Soc.
Rechargeable Batteries and Supercapacitors; Barnett, B. M., 1995, 142, 333.
Dowgiallo, E., Halpert, G., Matsuda, Y., Takehara, Z., Eds.; The (254) Aurbach, D.; Levi, M. D.; Levi, E.; Schechter, A. J. Phys. Chem.
Electrochemical Society Proceedings Series; Pennington, NJ, B 1997, 101, 2195.
1993; PV 93-23, p 321. (255) Chung, G.; Kim, H.; Yu, S.; Jun, S.; Choi, J.; Kim, M. J.
(208) Aurbach, D.; Gottlieb, H. Electrochim. Acta 1989, 34, 141. Electrochem. Soc. 2000, 147, 4391.
(209) Aurbach, D. J. Electrochem. Soc. 1989, 136, 906. (256) Although by the solution method the ternary GICs composed of
(210) Aurbach, D.; Gofer, Y. J. Electrochem. Soc. 1991, 138, 3529. [Li+(solv)-graphite] were indeed synthesized and identified by
(211) Goren, E.; Chusid, O.; Aurbach, D. J. Electrochem. Soc. 1991, XRD, the solvents used were mostly ether compounds that have
138, L6. high donicity toward lithium ion. Similar Li GICs with carbonate
(212) Aurbach, D.; Gofer, Y.; Ben-Zion, M.; Aped, P. J. Electroanal. molecules were never obtained. On the contrary, even via the
Chem. 1992, 339, 451. solution method, binary GICs (i.e., bare lithium ion intercalation
(213) Aurbach, D.; Chusid, O. J. Electrochem. Soc. 1993, 140, L155. without solvent) instead of ternary ones were often preferentially
(214) Zhang, X.; Kostecki, R.; Richardson, T. J.; Pugh, J. K.; Ross, P. formed in these ether solvents, casting more doubt about the
N. J. Electrochem. Soc. 2001, 148, A1341. possibility of their formation via lithium ion cells. The large
(215) Tiedemann, W. H.; Bennion, D. N. J. Electrochem. Soc. 1973, c-axis distance measured by XRD for these ternary GICs ranged
120, 1624. between 7 and 11 Å and confirmed from another angle that in
(216) Jorne, J.; Tobias, C. W. J. Electrochem. Soc. 1974, 121, 994. situ XRD carried out on a graphite anode during its cycling never
(217) Jorne, J.; Tobias, C. W. J. Appl. Electrochem. 1975, 5, 279. detected such GICs. See refs 257-260 for the preparation and
(218) Peled, E. In Lithium Batteries; Gabano, J. P., Eds.; Academic characterization of various ternary GICs based on ether com-
Press: London, 1983; p 43. pounds and lithium ion.
(219) Peled, E.; Straze, H. J. Electrochem. Soc. 1977, 124, 1030. (257) Mizutani, Y.; Abe, T.; Ikeda, K.; Ihara, E.; Asano, M.; Harada,
(220) Dousek, F. P.; Jansta, J.; Rihaz, J. J. Electroanal. Chem. 1973, T.; Inaba, M.; Ogumi, Z. Carbon 1997, 35, 61.
46, 281. (258) Abe, T.; Mizutani, Y.; Tabuchi, T.; Ikeda, K.; Asano, M.; Harada,
(221) Nazri, G.; Muller, R. H. J. Electrochem. Soc. 1985, 132, 1385. T.; Inaba, M.; Ogumi, Z. J. Power Sources 1997, 68, 216.
(222) Kanamura, K.; Tamura, H.; Shiraishi, S.; Takehara, Z. J. (259) Mizutani, Y.; Abe, T.; Inaba, M.; Ogumi, Z. Synth. Met. 2002,
Electrochem. Soc. 1995, 142, 340. 125, 153.
(223) Kanamura, K.; Shiraishi, S.; Takehara, Z. J. Electrochem. Soc. (260) Abe, T.; Kawabata, N.; Mizutani, Y.; Inaba, M.; Ogumi, Z. J.
1996, 143, 2187. Electrochem. Soc. 2003, 150, A257.
(224) Epelboin, I.; Froment, M.; Garreau, M.; Thevenin, J.; Warin, D. (261) Shu, Z. X.; McMillan, R. S.; Murray, J. J. J. Electrochem. Soc.
J. Electrochem. Soc. 1980, 127, 2100. 1993, 140, 922.
(225) Thevenin, J. G.; Muller, R. H. J. Electrochem. Soc. 1987, 134, (262) Matsumura, Y.; Wang, S.; Mondori, J. J. Electrochem. Soc. 1995,
273. 142, 2914.
(226) Gaberšček, M.; Jamnik, J.; Pejovnik, S. J. Electrochem. Soc. (263) Kim, Y.; Park, S. J. Electrochem. Soc. 2001, 148, A194.
1993, 140, 308. (264) Huang, W.; Frech, R. J. Electrochem. Soc. 1998, 145, 765.
(265) Inaba, M.; Siroma, Z.; Funabiki, A.; Ogumi, Z.; Abe, T.; Mizutani, (306) Rasch, B.; Cattaneo, E.; Novak, P.; Vielstich, W. Electrochim.
Y.; Asano, M. Langmuir 1996, 12, 1535. Acta 1991, 36, 1397.
(266) Chu, A. C.; Josefowicz, J. Y.; Farrington, G. C. J. Electrochem. (307) Kanamura, K.; Toriyama, S.; Shiraishi, S.; Takehara, Z. J.
Soc. 1997, 144, 4161. Electrochem. Soc. 1995, 142, 1383.
(267) Hirasawa, K. A.; Sato, T.; Asahina, H.; Yamaguchi, S.; Mori, S. (308) Ossola, F.; Pistoia, G.; Seeber, R.; Ugo, P. Electrochim. Acta 1988,
J. Electrochem. Soc. 1997, 144, L81. 33, 47.
(268) Alliata, D.; Kötz, R.; Novák, P.; Siegenthaler, H. Electrochem. (309) Matsuta, S.; Kato, Y.; Ota, T.; Kurokawa, H.; Yoshimura, S.;
Commun. 2000, 2, 436. Fujitani, S. J. Electrochem. Soc. 2001, 148, A7.
(269) Inaba, M.; Siroma, Z.; Kawatate, Y.; Funabiki, A.; Ogumi, Z. J. (310) Blyr, A.; Sigala, C.; Amatucci, G.; Guyomard, D.; Chabre, Y.;
Power Sources 1997, 68, 221. Tarascon, T. M. J. Electrochem. Soc. 1998, 145, 194.
(270) Peled, E.; Golodnitsky, D.; Ardel, G. J. Electrochem. Soc. 1997, (311) Cattaneo, E.; Ruch, J. J. Power Sources 1993, 43/44, 341.
144, L208. (312) Imhof, R.; Novák, P. J. Electrochem. Soc. 1999, 146, 1702.
(271) Ein-Eli, Y.; McDevitt, S. F.; Laura, R. J. Electrochem. Soc. 1998, (313) Sun, X.; Lee, H. S.; Lee, S.; Yang, X. Q.; McBreen, J. Electrochem.
145, L1. Solid-State Lett. 1998, 1, 239.
(272) Ein-Eli, Y. Electrochem. Solid-State Lett. 1999, 2, 212. (314) Xu, K.; Angell, C. A. J. Electrochem. Soc. 2002, 149, A920.
(273) Aurbach, D.; Markovsky, B.; Gamolsky, K.; Levi, E.; Ein-Eli, Y. (315) Fey, G. T.; Li, W.; Dahn, J. R. J. Electrochem. Soc. 1994, 141,
Electrochim. Acta 1999, 45, 67. 2279.
(274) Aurbach, D.; Levi, M. D.; Levi, E.; Schechter, A. J. Phys. Chem. (316) Seel, J. A.; Dahn, J. R. J. Electrochem. Soc. 2000, 147, 892.
B 1997, 101, 2195. (317) Lu, Z. H.; Dahn, J. R. J. Electrochem. Soc. 2001, 148, A710.
(275) Aurbach, D.; Teller, H.; Levi, E. J. Electrochem. Soc. 2002, 149, (318) Shembel, E. M.; Apostolova, R. D.; Strizhko, A. S.; Belosokhov,
A1255. A. I.; Naumenko, A. F.; Rozhkov, V. V. J. Power Sources 1995,
(276) Chung, G.; Kim, H.; Jun, S.; Kim, M. Electrochem. Commun. 54, 421.
1999, 1, 493. (319) Zhang, S. S.; Ding, M. S.; Jow, T. R. J. Power Sources 2001,
(277) Andersson, A. M.; Edström, K. J. Electrochem. Soc. 2001, 148, 102, 16.
A1100. (320) Lopez, S.; Petit, J. P.; Dunlop, H. M.; Butruille, J. R.; Tourillon,
(278) Zhuang, G.; Ross, P. Electrochem. Solid-State Lett. 2003, 6, A136. G. J. Electrochem. Soc. 1998, 145, 823.
(279) Aurbach, D.; Markovsky, B.; Schechter, A.; Ein Eli, Y. J. (321) Braithwaite, J. W.; Gonzales, A.; Nagasubramanian, G.; Lucero,
Electrochem. Soc. 1996, 143, 3809. S. J.; Peebles, D. E.; Ohlhausen, J. A.; Cieslak, W. R. J.
(280) Naji, A.; Ghanbaja, J.; Humbert, B.; Willmann, P.; Billaud, D. Electrochem. Soc. 1999, 146, 448.
J. Power Sources 1996, 63, 33. (322) Chen, Y.; Devine, T. M.; Evans, J. W.; Monteiro, O. R.; Brown,
(281) Bar-Tow, D.; Peled, E.; Burstein, L. J. Electrochem. Soc. 1999, I. G. J. Electrochem. Soc. 1999, 146, 1310.
146, 824. (323) Munshi, M. Z. A.; Gopalienger, R.; Owens, B. B. Solid State
(282) Kanamura, K.; Tamura, H.; Takehara, Z. J. Electroanal. Chem. Ionics 1988, 27, 259.
1992, 333, 127. (324) Xu, K.; Zhang, S.; Poese, B. A.; Jow, T. R. Electrochem. Solid-
(283) Kanamura, K.; Tamura, H.; Shiraishi, S.; Takehara, Z. J. State Lett. 2002, 5, A259.
Electroanal. Chem. 1995, 394, 49. (325) Diard, J. P.; LeCanut, J. M.; LeGorrec, B.; Montella, C. Elec-
(284) Lee, C.; Mun, B.; Ross, P. N. J. Electrochem. Soc. 2002, 149, trochim. Acta 1998, 43, 2469.
A1286. (326) Casparac, R.; Martin, C. R.; Lisac, E. S.; Mandic, Z. J. Electro-
(285) Zhang, X.; Pugh, J. K.; Ross, P. N. Electrochem. Solid-State Lett. chem. Soc. 2000, 147, 991.
2001, 4, A82. (327) Kanamura, K.; Okagawa, T.; Takehara, Z. J. Power Sources
(286) Imhof, R.; Novák, P. J. Electrochem. Soc. 1998, 145, 1081. 1995, 57, 119.
(287) Zhang, S.; Ding, M. S.; Xu, K.; Allen, J.; Jow, T. R. Electrochem. (328) Atanasoski, R. T.; Law, H. H.; Tobias, C. W. Electrochim. Acta
Solid-State Lett. 2001, 4, A206. 1987, 32, 877.
(288) Some metal oxide structures are unstable when over-delithiated, (329) Atanasoski, R. T.; Serb, J. J. Chem. Soc. 1992, 57, 935.
and as a consequence, the crystal lattice collapses to form a new (330) Christe, K. O.; Dixon, D. A.; McLemore, D.; Wilson, W. W.;
phase that is electrochemically inactive. Examples are the so- Sheehy, J. A.; Boatz, J. J. Fluorine Chem. 2000, 101, 151.
called “Jahn-Teller effect” for spinel cathodes and similar (331) Polypropylene or stainless steel containers are normally used,
behavior for LiNiO2 and LiCoO2 materials as well. These but glass containers can be severely corroded by LiPF6-contain-
irreversible processes are considered to be caused by the intrinsic ing electrolytes, because the formation of gaseous SiF4 serves
properties of the crystalline materials instead of electrolytes and as the driving force for the reaction between the fluorinated
are, therefore, beyond the scope of the current review. See ref Lewis acids PF5 or HF and the main composition of glass Na2-
46 for a detailed review. SiO3.
(289) Mansour, A. Surf. Sci. Spectrosc. 1996, 3, 279. (332) Vogdanis, L.; Heitz, W. Macromol. Rapid Commun. 1986, 7, 543.
(290) Aurbach, D.; Levi, M. D.; Levi, E.; Teller, H.; Markovsky, B.; (333) Vogdanis, L.; Martens, B.; Uchtmann, H.; Hensel, F.; Heitz, W.
Salitra, G.; Heider, U.; Heider, L. J. Electrochem. Soc. 1998, 145, Macromol. Chem. 1990, 191, 465.
3024. (334) Terasaki, M.; Yosjida, H.; Tukamoto, H.; Mizutani, M.; Yamachi,
(291) Aurbach, D. J. Power Sources 2000, 89, 206. M. Denki Kagaku 1993, 61, 1417.
(292) Aurbach, D.; Gamolsky, K.; Markovsky, B.; Salitra, G.; Gofer, (335) Ohta, A.; Koshina, H.; Okuno, H.; Murai, H. J. Power Sources
Y.; Heider, U.; Oesten, R.; Schmidt, M. J. Electrochem. Soc. 2000, 1995, 54, 6.
147, 1322. (336) Takeuchi, E. S.; Gan, H.; Palazzo, M.; Leising, R. A.; Davis, S.
(293) Eriksson, T.; Andersson, A. M.; Bishop, A. G.; Gejke, C.; M. J. Electrochem. Soc. 1997, 144, 1944.
Gustafsson, T.; Thomas, J. O. J. Electrochem. Soc. 2002, 149, (337) Endo, E.; Ata, M.; Sekai, K.; Tanaka, K. J. Electrochem. Soc.
A69. 1999, 146, 49.
(294) Andersson, A. M.; Abraham, D. P.; Haasch, R.; MacLaren, S.; (338) Du Pasquier, A.; Disma, F.; Bowmer, T.; Gozdz, A. S.; Amatucci,
Liu, J.; Amine, K. J. Electrochem. Soc. 2002, 149, A1358. G.; Tarascon, J. M. J. Electrochem. Soc. 1998, 145, 472.
(295) Thomas, M. G. S. R.; Bruce, P. G.; Goodenough, J. B. J. (339) Zheng, T.; Gozdz, A. S.; Amatucci, G. G. J. Electrochem. Soc.
Electrochem. Soc. 1985, 132, 1521. 1999, 146, 4014.
(296) Jang, D. H.; Shin, Y. J.; Oh, S. M. J. Electrochem. Soc. 1996, (340) Richard, M. N.; Dahn, J. R. J. Electrochem. Soc. 1999, 146, 2068.
143, 2204. (341) Richard, M. N.; Dahn, J. R. J. Electrochem. Soc. 1999, 146, 2078.
(297) Xia, Y.; Zhou, Y.; Yoshio, M. J. Electrochem. Soc. 1997, 144, 2593. (342) MacNeil, D. D.; Larcher, D.; Dahn, J. R. J. Electrochem. Soc.
(298) Zhang, X.; Ross, P. N.; Kostecki, R.; Kong, F.; Sloop, S.; Kerr, J. 1999, 146, 3596.
B.; Striebel, K.; Cairns, E. J.; McLarnon, F. J. Electrochem. Soc. (343) Pistoia, G.; Antonini, A.; Rosati, R.; Zane, D. Electrochim. Acta
2001, 148, A463. 1996, 41, 2683.
(299) Wang, Y.; Guo, X.; Greenbaum, S.; Liu, J.; Amine, K. Electro- (344) Amatucci, G. G.; Schmutz, C. N.; Blyr, A.; Sigala, C.; Gozdz, A.
chem. Solid-State Lett. 2001, 4, A68. S.; Larcher, D.; Tarascon, J. M. J. Power Sources 1997, 69, 11.
(300) Matsuo, Y.; Kostecki, R.; McLarnon, F. J. Electrochem. Soc. 2001, (345) Antonini, A.; Bellito, C.; Pasquali, M.; Pistoia, G. J. Electrochem.
148, A687. Soc. 1998, 145, 2726.
(301) Du Pasquier, A.; Blyr, A.; Courjal, P.; Larcher, D.; Amatucci, (346) Tsunekawa, H.; Tanimoto, S.; Marubayashi, R.; Fujita, M.;
G.; Gérand, B.; Tarascon, J. M. J. Electrochem. Soc. 1999, 146, Kifune, K.; Sano, M. J. Electrochem. Soc. 2002, 149, A1326.
428. (347) Levy, S. C.; Bro, P. Battery Hazards and Accident Prevention;
(302) Balasubramanian, M.; Lee, H. S.; Sun, X.; Yang, X. Q.; Mood- Plenum Press: New York, 1994; p 115.
enbaugh, A. R.; McBreen, J.; Fischer, D. A.; Fu, Z. Electrochem. (348) von Sacken, U.; Nodwell, E.; Sundher, A.; Dahn, J. Solid State
Solid-State Lett. 2002, 5, A22. Ionics 1994, 69, 284.
(303) Zhang, S. S.; Xu, K.; Jow, T. R. Electrochem. Solid-State Lett. (349) Chen, Y.; Evans, J. W. J. Electrochem. Soc. 1996, 143, 2708.
2002, 5, A92. (350) Song, L.; Evans, J. W. J. Electrochem. Soc. 1998, 145, 2327.
(304) Zhang, S. S.; Xu, K.; Jow, T. R. J. Electrochem. Soc. 2002, 149, (351) Tobishima, S.; Yamaki, J. J. Power Sources 1999, 81/82, 882.
A1521. (352) Leising, R. A.; Palazzo, M. J.; Takeuchi, E. S.; Takeuchi, K. J.
(305) Croce, F.; Nobili, F.; Deptula, A.; Lada, W.; Tossici, R.; D’Epifanio, J. Electrochem. Soc. 2001, 148, A838.
A.; Scrosati, B.; Marassi. Electrochem. Commun. 1999, 1, 605. (353) Hasegawa, K.; Arakawa, Y. J. Power Sources 1993, 43/44, 523.
(354) Gee, M. A.; Laman, F. C. J. Electrochem. Soc. 1993, 140, L53. (403) Jehoulet, C.; Biensan, P.; Bodet, J. M.; Broussely, M.; Moteau,
(355) Chen, Y.; Evans, J. W. J. Electrochem. Soc. 1994, 141, 2947. C.; Tessier-Lescourret, C. In Batteries for Portable Applications
(356) Newman, J.; Tiedemann, W. J. Electrochem. Soc. 1995, 142, and Electric Vehicles; Holmes, C. F., Landgrebe, A. R., Eds.; The
1054. Electrochemical Society Proceeding Series; Pennington, NJ,
(357) Pals, C. R.; Newman, J. J. Electrochem. Soc. 1995, 142, 3274. 1997; P. V. 97-18, P974.
(358) Pals, C. R.; Newman, J. J. Electrochem. Soc. 1995, 142, 3282. (404) Aurbach, D.; Gamolsky, K.; Markovsky, B.; Gofer, Y.; Schmidt,
(359) Joho, F.; Novák, P.; Spahr, M. E. J. Electrochem. Soc. 2002, 149, M.; Heider, U. Electrochim. Acta 2002, 47, 1423.
A1020. (405) Contestabile, M.; Morselli, M.; Paraventi, R.; Neat, R. J. J. Power
(360) Maleki, H.; Al Hallaj, S.; Selman, J. R.; Dinwiddie, R. B.; Wang, Sources 2003, 119/121, 943.
H. J. Electrochem. Soc. 1999, 146, 947. (406) Herreyre, S.; Huchet, O.; Barusseau, S.; Perton, F.; Bodet, J.
(361) Maleki, H.; Deng, G.; Anani, A.; Howard, J. J. Electrochem. Soc. M.; Biensan, Ph. J. Power Sources 2001, 97/98, 576.
1999, 146, 3224. (407) Broussely, M.; Blanchard, P.; Biensan, Ph.; Planchat, J. P.;
(362) Zhang, Z.; Fouchard, D.; Rea, J. R. J. Power Sources 1998, 70, Nechev, K.; Staniewicz, R. J. J. Power Sources 2003, 119/121,
16. 859.
(363) Biensan, P.; Simon, B.; Pérès, J. P.; de Guibert, A.; Brousely, (408) Oesten, R.; Heider, U.; Schmid, M. Solid State Ionics 2002, 148,
M.; Bodet, J. M.; Perton, F. J. Power Sources 1999, 81/82, 906. 391.
(364) Dahn, J. R.; Fuller, E. W.; Obrovac, M.; von Sacken, U. Solid (409) Wang, C.; Nakamura, H.; Komatsu, H.; Yoshio, M.; Yoshitake,
State Ionics 1994, 69, 265. H. J. Power Sources 1998, 74, 142.
(365) Arai, H.; Okada, S.; Sakurai, Y.; Yamaki, J. J. Electrochem. Soc. (410) Naji, A.; Ghanbaja, J.; Willmann, P.; Billaud, D. Electrochim.
1997, 144, 3117. Acta 2000, 45, 1893.
(366) Schilling, O.; Dahn, J. R. J. Electrochem. Soc. 1998, 145, 569. (411) Matsuo, Y.; Fumita, K.; Fukutsuka, T.; Sugie, Y.; Koyama, H.;
(367) MacNeil, D. D.; Christensen, L.; Landucci, J.; Paulsen, J. M.; Inoue, K. J. Power Sources 2003, 119/121, 373.
Dahn, J. R. J. Electrochem. Soc. 2000, 147, 970. (412) Santer, H. J.; Möller, K.-C.; Ivanco, J.; Ramsey, M. G.; Netzer,
(368) MacNeil, D. D.; Hatchard, T. D.; Dahn, J. R. J. Electrochem. F. P.; Yamaguchi, S.; Besenhard, J. O.; Winter, M. J. Power
Soc. 2001, 148, A663. Sources 2003, 119/121, 368.
(369) MacNeil, D. D.; Dahn, J. R. J. Electrochem. Soc. 2001, 148, (413) Möller, K.-C.; Santer, H. J.; Kern, W.; Yamaguchi, S.; Besenhard,
A1205. J. O.; Winter, M. J. Power Sources 2003, 119/121, 561.
(370) MacNeil, D. D.; Dahn, J. R. J. Electrochem. Soc. 2001, 148, (414) Holleck, G. L.; Harris, P. B.; Abraham, K. M.; Buzby, J.;
A1211. Brummer, S. B. Technical Report No. 6, Contract N00014-77-
(371) MacNeil, D. D.; Dahn, J. R. J. Electrochem. Soc. 2003, 150, A21. 0155; EIC Laboratories: Newton, MA, 1982.
(372) Matsuda, Y. J. Power Sources 1993, 43, 1. (415) Abraham, K. M.; Brummer, S. B. In Lithium Batteries; Gabano,
(373) Hirai, T.; Yoshimatsu, I.; Yamaki, J. J. Electrochem. Soc. 1994, J., Ed.; Academic Press: New York, 1983.
141, 2300. (416) Narayanan, S. R.; Surampudi, S.; Attia, A. I.; Bankston, C. P.
(374) Kovać, M.; Milićev, S.; Kovać, A.; Pejovnik, S. J. Electrochem. J. Electrochem. Soc. 1991, 138, 2224.
Soc. 1995, 142, 1390. (417) Abraham, K. M.; Pasquariello, D. M.; Willstädt, E. B. Proceedings
(375) Mogi, R.; Inaba, M.; Jeong, S.; Iriyama, Y.; Abe, T.; Ogumi, Z. of the 33rd International Power Sources Symposium, Cherry Hill,
J. Electrochem. Soc. 2002, 149, 2002. NJ, June 13-16, 1988; The Electrochemical Society, Inc.: 1988;
(376) Besenhard, J. O.; Wagner, M. W.; Winter, M.; Jannakoudakis, p 38.
A. D.; Jannakoudakis, P. D.; Theodoridou, E. J. Power Sources (418) Behl, W. K.; Chin, D. T. J. Electrochem. Soc. 1988, 135, 16.
1993, 43/44, 413. (419) Behl, W. K.; Chin, D. T. J. Electrochem. Soc. 1988, 135, 21.
(377) Gong, J. B.; Tsumura, T.; Nakamura, H.; Yoshio, M.; Yoshitake, (420) Behl, W. K. J. Electrochem. Soc. 1989, 136, 2305.
H.; Abe, T. 202nd Meeting of the Electrochemical Society, Salt (421) Abraham, K. M.; Pasquariello, D. M. U.S. Patent, 4,857,423,
Lake City, UT, Oct 20-24, 2002; Abstract No. 200; Electro- 1989.
chemical Society: Pennington, NJ. (422) Abraham, K. M.; Pasquariello, D. M.; Willstaedt, E. B. J.
(378) Izatt, R. M.; Bradshaw, J. S.; Nielsen, S. A.; Lamb, J. D.; Electrochem. Soc. 1990, 137, 1856.
Christensen, J. J. Chem. Rev. 1985, 85, 271. (423) Abraham, K. M. Electrochim. Acta 1993, 38, 1233.
(379) Morita, M.; Hayashida, H.; Matsuda, Y. J. Electrochem. Soc. (424) Mason, J. G.; Rosenblum, M. J. Am. Chem. Soc. 1960, 82, 4206.
1987, 134, 2107. (425) Wilkinson, D. P.; Dudley, J.; Golovin, N. In Rechargeable Lithium
(380) Nagasubramanian, G.; Di Stefano, S. J. Electrochem. Soc. 1990, Batteries; Subbarao, S., Koch, V. R., Owens, B. B., Smyrl, W.
137, 3830. H., Eds.; The Electrochemical Society: Pennington, NJ, 1990;
(381) Salomom, M. J. Solution Chem. 1990, 19, 1225. PV 90-5, p 77.
(382) Nagasubramanian, G.; Attia, A. I.; Halpert, G. J. Electrochem. (426) Abraham, K. M. In Rechargeable Lithium Batteries; Subbarao,
Soc. 1992, 139, 3043. S., Koch, V. R., Owens, B. B., Smyrl, W. H., Eds.; The Electro-
(383) Lonergan, M. C.; Ratner, M. A.; Shriver, D. J. Am. Chem. Soc. chemical Society: Pennington, NJ, 1990; PV 90-5, p 1.
1995, 117, 2344. (427) Cabrera, C. R.; Bard, A. J. J. Electroanal. Chem. 1989, 273, 147.
(384) Lee, H. S.; Yang, X. Q.; McBreen, J.; Choi, L. S.; Okamoto, Y. J. (428) Cha, C. S.; Ai, X. P.; Yang, H. X. J. Power Sources 1995, 54,
Electrochem. Soc. 1996, 143, 3825. 255.
(385) Lee, H. S.; Sun, X.; Yang, X. Q.; McBreen, J.; Callahan, J. H.; (429) Golovin, M. N.; Wilkinson, D. P.; Dudley, J. T.; Holonko, D.; Woo,
Choi, L. S. J. Electrochem. Soc. 2000, 146, 9. S. J. Electrochem. Soc. 1992, 139, 5.
(386) Lee, H. S.; Yang, X. Q.; Xiang, C. L.; McBreen, J.; Choi, L. S. J. (430) Halpert, G.; Surampudi, S.; Shen, D.; Huang, C. K.; Narayanan,
Electrochem. Soc. 1998, 145, 2813. S.; Vamos, E.; Perrone, D. J. Power Sources 1994, 47, 287.
(387) Sun, X.; Lee, H. S.; Yang, X. Q.; McBreen, J. J. Electrochem. (431) Richardson, T. J.; Ross, P. N. J. Electrochem. Soc. 1996, 143,
Soc. 1999, 146, 3655. 3992.
(388) Sun, X.; Lee, H. S.; Yang, X. Q.; McBreen, J. J. Electrochem. (432) Adachi, M.; Tanaka, K.; Sekai, K. J. Electrochem. Soc. 1999, 146,
Soc. 2002, 149, A355. 1256.
(389) Sun, X.; Lee, H. S.; Yang, X. Q.; McBreen, J. Electrochem. Solid- (433) Hammerich, O.; Parker, V. D. Electrochim. Acta 1973, 18, 537.
State Lett. 2001, 4, A184. (434) Kerr, J. B.; Tian, M. U.S. Patent 6,045,952, 2000.
(390) Sun, X.; Lee, H. S.; Yang, X. Q.; McBreen, J. Electrochem. Solid- (435) Mao, H.; Wainwright, D. S. Canadian Patent 2,205,683, 1997.
State Lett. 2002, 5, A248. (436) Yoshio, M.; Yoshitake, H.; Abe, K. 204th Meeting of Electrochemi-
(391) Sun, X.; Lee, H. S.; Yang, X. Q.; McBreen, J. Electrochem. Solid- cal Society, Orlando, FL, Oct 12-16, 2003; Abstract No. 280.
State Lett. 2003, 6, A43. (437) Kuboki, T.; Ohsaki, T. U.S. Patent 6,413,679, 2002.
(392) Lee, H. S.; Sun, X.; Yang, X. Q.; McBreen, J. J. Electrochem. (438) Reimers, J. N.; Way, B. M. U.S. Patent 6,074,777, 2000.
Soc. 2002, 149, A1460. (439) Mao, H. Canadian Patent 2,163,187, 1995.
(393) Tasaki, K.; Nakamura, S. J. Electrochem. Soc. 2001, 148, A984. (440) Yoshino, A. In Proceedings of the 4th Hawaii Battery Conference,
(394) Morita, M.; Ishikawa, M.; Matsuda, Y. In Lithium Ion Batter- ARAD Enterprises, Hilo, HI, Jan 8, 2002; ARAD Enterprises:
ies: Fundamentals and Performance; Wakihara, M., Yamamoto, Hilo, HI; p 102.
O., Eds.; Wiley-VCH: New York, 1999; p 156. (441) Besenhard, J. O.; von Werner, K.; Winter, M. US. Patent 5,-
(395) Ein-Eli, Y.; Thomas, S. R.; Koch, V. R. J. Electrochem. Soc. 1996, 916,708, 1999.
143, L195. (442) McMillan, R.; Slegr, H.; Shu, Z. X.; Wang, W. J. Power Sources
(396) Ein-Eli, Y.; Thomas, S. R.; Koch, V. R. J. Electrochem. Soc. 1997, 1999, 81/82, 20.
144, 1159. (443) Inaba, M.; Kawatate, Y.; Funabiki, A.; Jeong, S. K.; Abe, T.;
(397) Wilkinson, D.; Dahn, J. R. U.S. Patent 5,130,211, 1992. Ogumi, Z. Electrochim. Acta 1999, 45, 99.
(398) Shu, Z. X.; McMillan, R. S.; Murray, J. J. J. Electrochem. Soc. (444) Petrella, R. V. In Flame Retardant Polymeric Materials; Lewin,
1993, 140, L101. M., Atlas, S. M., Pearce, E. M., Eds.; Plenum Publishing: New
(399) Wrodnigg, G. H.; Besenhard, J. O.; Winter, M. J. Electrochem. York, 1978; Vol. 2, p 159.
Soc. 1999, 146, 470. (445) Cullis, C. F.; Hirshler, M. M. The Combustion of Organic
(400) Simon, B.; Boeuve, J.-P. U.S. Patent 5,626,981, 1997. Polymers; Clarendon Press: Oxford, 1981.
(401) Barker, J.; Gao, F. U.S. Patent 5,712,059, 1998. (446) Shu, Z. X.; McMillan, R. S.; Murray, J. J.; Davidson, I. J. J.
(402) Naruse, Y.; Fujita, S.; Omaru, A. U.S. Patent 5,714,281, 1998. Electrochem. Soc. 1995, 142, L161.
(447) Shu, Z. X.; McMillan, R. S.; Murray, J. J.; Davidson, I. J. J. (494) Handa, M.; Suzuki, M.; Suzuki, J.; Kanematsu, H.; Sasaki, Y.
Electrochem. Soc. 1996, 143, 2230. Electrochem. Solid-State Lett. 1999, 2, 60.
(448) For this reason ClEC and FEC were treated as cosolvents in (495) Eberwein, M.; Schmidt, A.; Schmidt, M.; Zabel, M.; Burgemeis-
this review, instead of as additives as they were in most ter, T.; Barthel, J.; Kunz, W.; Gores, H. J. J. Electrochem. Soc.
literature sources. 2003, 150, A994.
(449) Winter, M.; Novák, P. J. Electrochem. Soc. 1998, 145, L27. (496) Sartori, P.; Ignatyev, N. The 194th Meeting of the Electrochemical
(450) Naji, A.; Ghanbaja, J.; Willmann, P.; Billaud, D. J. Power Sources Society, Boston, MA, Nov 1-6, 1998; Abstract 160.
1999, 81/82, 207. (497) Sartori, P.; Ignatyev, N. U.S. Patent 6,210,830, 2001.
(451) Winter, M.; Imhof, R.; Joho, F.; Noák, P. J. Power Sources 1999, (498) Schmidt, M.; Heider, U.; Kuehner, A.; Oesten, R.; Jungnitz, M.;
81/82, 818. Ignat’ev, N.; Sartori, P. J. Power Sources 2001, 97/98, 557.
(452) Nagasubramanian, G. The 1999 Joint International Meeting of (499) (a) Gnanaraj, J. S.; Levi, M. D.; Gofer, Y.; Aurbach, D.; Schmidt,
the Electrochemical Society, Honolulu, HI, Oct 17-22, 1999; M. J. Electrochem. Soc. 2003, 150, A445. (b) Gnanaraj, J. S.;
Abstract 334; Electrochemical Society: Pennington, NJ. Zinigrad, E.; Asraf, L.; Sprecher, M.; Gottlieb, H. E.; Geissler,
(453) Arai, J.; Katayama, H.; Akahoshi, H. J. Electrochem. Soc. 2002, W.; Schmidt, M.; Aurbach, D. Electrochem. Commun. 2003, 5,
149, A217. 946.
(454) Salomon, M.; Plichta, E. J. Electrochim. Acta 1985, 30, 113. (500) Gnanaraj, J. S.; Zinigrad, E.; Levi, M. D.; Aurbach, D.; Schmidt,
(455) Chagnes, A.; Carré, B.; Willmann, P.; Lemordant, D. J. Power M. J. Power Sources 2003, 119/121, 799.
Sources 2002, 109, 203. (501) Egashira, M.; Scrosati, B.; Armand, M.; Béranger, S.; Michot,
(456) Fukushima, T.; Matsuda, Y.; Hashimoto, H.; Arakawa, R. J. C. Electrochem. Solid-State Lett. 2003, 6, A71.
Power Sources 2002, 110, 34. (502) Barbarich, T. J.; Driscoll, P. F. Electrochem. Solid-State Lett.
(457) Rendek, L. J.; Chottiner, G. S.; Scherson, D. A. J. Electrochem. 2003, 6, A113.
Soc. 2003, 150, A326. (503) Plichta, E. J.; Behl, W. K. Proceedings of the 38th Power Source
(458) Takami, N.; Sekino, M.; Ohsaki, T.; Kanda, M.; Yamamoto, M. Conference, Cherry Hill, NJ, 1998; p 444.
J. Power Sources 2001, 97/98, 677. (504) Plichta, E. J.; Hendrickson, M.; Thompson, R.; Au, G.; Behl, W.
(459) Lanz, M.; Novák, P. J. Power Sources 2001, 102, 277. K.; Smart, M. C.; Ratnakumar, B. V.; Surampudi, S. J. Power
(460) Chagnes, A.; Carré, B.; Willmann, P.; Dedryvére, R.; Gonbeau, Sources 2001, 94, 160.
D.; Lemordant, D. J. Electrochem. Soc. 2003, 150, A1255. (505) Hamlen, R.; Au, G.; Brundage, M.; Hendrickson, M.; Plichta, E.;
(461) Smart, M. C.; Ratnakumar, B. V.; Surampudi, S.; Wang, Y.; Slane, S.; Barbarello, J. J. Power Sources 2001, 97/98, 22.
Zhang, X.; Greenbaum, S. G.; Hightower, A.; Ahn, C. C.; Fultz, (506) Nagasubramanian, G. J. Appl. Electrochem. 2001, 31, 99.
B. J. Electrochem. Soc. 1999, 146, 3963. (507) Fan, J. J. Power Sources 2003, 117, 170.
(462) Smart, M. C.; Ratnakumar, B. V.; Surampudi, S. J. Electrochem. (508) Smart, M. C.; Ratnakumar, B. V.; Surampudi, S. J. Electrochem.
Soc. 2002, 149, A361. Soc. 1999, 146, 486.
(463) Vetter, J.; Novák, P. J. Power Sources 2003, 119/121, 338. (509) Shiao, H. C.; Chua, D.; Lin, H.; Slane, S.; Salomon, M. J. Power
(464) Nakajima, T.; Dan, K.; Koh, M. J. Fluorine Chem. 1998, 87, 221. Sources 2000, 87, 167.
(465) Yamaki, J.-I.; Yamazaki, I.; Egashira, M.; Okada, S. J. Power (510) Sawai, K.; Ohzuku, T. J. Electrochem. Soc. 2003, 150, A674.
Sources 2001, 102, 288. (511) Wang, C.; Appleby, A. J.; Little, F. E. J. Electrochem. Soc. 2002,
(466) Smart, M. C.; Ratnakumar, B. V.; Ryan-Mowrey, V. S.; Suram- 149, A754.
pudi, S.; Prakash, G. K. S.; Hu, J.; Cheung, I. J. Power Sources (512) Zhang, S. S.; Xu, K.; Jow, T. R. J. Power Sources 2003, 115, 137.
2003, 119/121, 359. (513) Huang, C.-K.; Sakamoto, J. S.; Wolfenstine, J.; Surampudi, S.
(467) Zhang, S. S.; Angell, C. A. J. Electrochem. Soc. 1996, 143, 4047. J. Electrochem. Soc. 2000, 147, 2893.
(468) Choquette, Y.; Brisard, G.; Parent, M.; Brouillette, D.; Perron, (514) Gitzendanner, R.; Ehrlich, G.; Marsh, C.; Marsh, R. The 194th
G.; Desnoyers, J. E.; Armand, M.; Gravel, D.; Slougui, N. J. Meeting of the Electrochemical Society, Boston, MA, Nov 1-6,
Electrochem. Soc. 1998, 145, 3500. 1998; Abstract 157; Electrochemical Society: Pennington, NJ.
(469) Matsuda, Y.; Morita, M.; Yamada, K.; Hirai, K. J. Electrochem. (515) Smart, M. C.; Ratnakumar, B. V.; Whitcanack, L. D.; Chin, K.
Soc. 1985, 132, 2538. B.; Surampudi, S.; Croft, H.; Tice, D.; Staniewicz, R. J. Power
(470) Morita, M.; Okada, Y.; Matsuda, Y. J. Electrochem. Soc. 1987, Sources 2003, 119/121, 349.
134, 2665. (516) The physical significance of the impedance response correspond-
(471) Bach, S.; Baffier, N.; Pereira-Ramos, J. P.; Messina, R. J. Power ing to the semicircle at lower frequency has been ambiguous.
Sources 1993, 43/44, 569. Conventionally it was assigned to the “charge-transfer” for the
(472) Yan, W.; Lerner, M. M. J. Electrochem. Soc. 2001, 148, D83. electrochemical reactions; hence, conceptually, it should not be
(473) Park, S. H.; Winnick, J.; Kohl, P. A. J. Electrochem. Soc. 2002, influenced by electrolytes due to the presence of the electronically
149, A1196. insulating surface films. However, experiences show that it is
(474) Dominey, L. A. U.S. Patent, 5,273,840, 1993. in fact readily affected by the chemical composition of the
(475) Barthel, J.; Wühr, M.; Buestrich, R.; Gores, H. J. J. Electrochem. electrolytes; therefore, it should involve an interfacial process
Soc. 1995, 142, 2527. that is sensitive to the surface film formation. There was a
(476) Barthel, J.; Buestrich, R.; Carl, E.; Gores, H. J. J. Electrochem. suggestion (see ref 518) that the SEI is a mixed conductor for
Soc. 1996, 143, 3565. both ions and electrons, so that the electrons can transport
(477) Barthel, J.; Buestrich, R.; Carl, E.; Gores, H. J. J. Electrochem. across it to reach the redox reaction sites. This hypothesis is
Soc. 1996, 143, 3572. well able to explain why Rct is affected by SEI chemical
(478) Barthel, J.; Buestrich, R.; Gores, H. J.; Schmidt, M.; Wühr, M. composition, however, at the risk of overturning the fundamen-
J. Electrochem. Soc. 1997, 144, 3866. tals of lithium ion chemistry.
(479) Barthel, J.; Schmidt, M.; Gores, H. J. J. Electrochem. Soc. 1998, (517) Mohamedi, M.; Takahashi, D.; Itoh, T.; Uchida, I. Electrochim.
145, L17. Acta 2002, 47, 3483.
(480) Barthel, J.; Schmidt, M.; Gores, H. J. J. Electrochem. Soc. 2000, (518) Chen, C. H.; Liu, J.; Amine, K. J. Power Sources 2001, 96, 321.
147, 21. (519) Roth, E. P.; Crafts, C. C.; Doughty, D. H. The 202nd Meeting of
(481) Sasaki, Y.; Handa, M.; Kurashima, K.; Tonuma, T.; Usami, K. the Electrochemical Society, Salt Lake City, UT, Oct 20-24,
J. Electrochem. Soc. 2001, 148, A999. 2002; Abstract 216.
(482) Handa, M.; Fukuda, S.; Sasaki, Y.; Usami, K. J. Electrochem. (520) (a) Yokoyama, K.; Fujita, S.; Hiwara, A.; Naruse, Y.; Toriida,
Soc. 1997, 144, L235. M.; Omaru, A. U.S. Patent 5,580,684, 1996. (b) Narang, S. C.;
(483) Xu, W.; Angell, C. A. Electrochem. Solid-State Lett. 2000, 3, 366. Ventura, S. C.; Dougherty, B. J.; Zhao, M.; Smedley, S.; Koolpe,
(484) Xu, W.; Angell, C. A. Solid State Ionics 2002, 147, 295. G. U.S. Patent 5,830,600, 1998. (c) Narang, S. C.; Ventura, S.
(485) Videa, M.; Xu, W.; Geil, B.; Marzke, R.; Angell, C. A. J. C.; Cox, P. U.S. Patent 6,168,885, 2001.
Electrochem. Soc. 2001, 148, A1352. (521) Lyons, J. W. The Chemistry and Uses of Flame Retardants; John
(486) Xu, W.; Shusterman, A.; Videa, M.; Velikov, V.; Marzke, R. L.; Wiley & Sons: New York, 1970.
Angell, C. A. J. Electrochem. Soc. 2003, 150, E74. (522) Cullis, C. F.; Hirshler, M. M. The Combustion of Organic
(487) Lischka, U.; Wietelmann, U.; Wegner, M. DE 19829030 C1, 1999. Polymers; Clarendon Press: Oxford, 1981.
(488) Xu, K.; Zhang, S.; Jow, T. R. Electrochem. Solid-State Lett. 2003, (523) Lee, C. W.; Venkatachalapathy, R.; Prakash, J. Electrochem.
6, A117. Solid-State Lett. 2000, 3, 63.
(489) Xu, K.; Lee, U.; Zhang, S.; Wood, M.; Jow, T. R. Electrochem. (524) Wang, X.; Yasukawa, E.; Kasuya, S. J. Electrochem. Soc. 2001,
Solid-State Lett. 2003, 6, A144. 148, A1058.
(490) Amine, K.; Liu, J.; Kang, S.; Belharouak, I.; Hyung, Y.; Vissers, (525) Wang, X.; Yasukawa, E.; Kasuya, S. J. Electrochem. Soc. 2001,
D.; Henriksen, G. J. Power Sources 2004, 129, 14. 148, A1066.
(491) Jiang, J.; Dahn, J. R. Electrochem. Solid-State Lett. 2003, 6, (526) Xu, K.; Ding, M. S.; Zhang, S.; Allen, J. L.; Jow, T. R. J.
A180. Electrochem. Soc. 2002, 149, A622.
(492) Jow, T. R.; Ding, M. S.; Xu, K.; Zhang, S. S.; Allen, J. L.; Amine, (527) Xu, K.; Zhang, S.; Allen, J. L.; Jow, T. R. J. Electrochem. Soc.
K.; Henriksen, G. L. J. Power Sources 2003, 119/121, 343. 2002, 149, A1079.
(493) Yamaguchi, H.; Takahashi, H.; Kato, M.; Arai, J. J. Electrochem. (528) Ding, M. S.; Xu, K.; Jow, T. R. J. Electrochem. Soc. 2002, 149,
Soc. 2003, 150, A312. A1489.
(529) Xu, K.; Ding, M. S.; Zhang, S.; Allen, J. L.; Jow, T. R. J. (546) Angell, C. A.; Fan, J.; Liu, C.; Lu, Q.; Sanchez, E.; Xu, K. Solid
Electrochem. Soc. 2003, 150, A161. State Ionics 1994, 69, 343.
(530) Xu, K.; Zhang, S.; Allen, J. L.; Jow, T. R. J. Electrochem. Soc. (547) Angell, C. A.; Xu, K.; Zhang, S.-S.; Videa, M. Solid State Ionics
2003, 150, A170. 1996, 86/88, 17.
(531) Hyung, Y. E.; Vissers, D. R.; Amine, K. J. Power Sources 2003, (548) Imrie, C. T.; Ingram, M. D.; McHattie, G. S. J. Phys. Chem. B
119/121, 383. 1999, 103, 4132.
(532) Ota, H.; Kominato, A.; Chun, W.-J.; Yasukawa, E.; Kasuya, S. (549) Zheng, Y.; Chia, F.; Ungar, G.; Wright, P. V. Chem. Commun.
J. Power Sources 2003, 119/121, 393. 2000, 16, 1459.
(533) Granzow, A. Acc. Chem. Res. 1978, 11, 177. (550) Anderman, M. Solid State Ionics 1994, 69, 336.
(534) Arai, J. J. Electrochem. Soc. 2003, 150, A219. (551) Scrosati, B.; Croce, F.; Persi, L. J. Electrochem. Soc. 2000, 147,
(535) Arai, J. J. Power Sources 2003, 119/121, 388. 1718.
(536) Ratner, M. A.; Shriver, D. F. Chem. Rev. 1988, 88, 109. (552) Persi, L.; Croce, F.; Scrosati, B.; Plichta, E.; Hendrickson, M. A.
(537) Watanabe, M. In Solid State Ionics: Materials and Applications; J. Electrochem. Soc. 2002, 149, A212.
Chowdari, B. V. R., Ed.; World Scientific Publishing Co.: (553) Pamero, S.; Satolli, D.; D’Epifano, A.; Scrosati, B. J. Electrochem.
Singapore, 1992; p 149. Soc. 2002, 149, A414.
(538) Sanchez, J.-Y. In Solid State Ionics: Materials and Applications; (554) Aihara, Y.; Appetecchi, G. B.; Scrosati, B. J. Electrochem. Soc.
Chowdari, B. V. R., Ed.; World Scientific Publishing Co.: 1992; 2002, 149, A849.
p 159. (555) Appetecchi, G. B.; Aihara, Y.; Scrosati, B. J. Electrochem. Soc.
(539) Armand, M. Solid State Ionics 1994, 69, 309. 2003, 150, A301.
(540) Shriver, D. F.; Bruce, P. G. In Solid State Electrochemistry; (556) Tarascon, J.-M.; Gozdz, A. S.; Schmutz, C.; Shokoohi, F.; Warren,
Bruce, P. G., Ed.; Cambridge University Press: Cambridge, P. C. Solid State Ionics 1996, 86/88, 49.
U.K., 1995; p 95. (557) Feuillade, G.; Perche, P. J. Appl. Electrochem. 1975, 63, 5.
(541) Song, J. Y.; Wang, Y. Y.; Wan, C. C. J. Power Sources 1999, 77, (558) Tsuchida, E.; Ohno, H.; Tsunemi, K. Electrochim. Acta 1983,
183. 28, 591.
(542) Dias, F. B.; Plomp, L.; Veldhuis, J. B. J. J. Power Sources 2000, (559) Tsunemi, K.; Ohno, H.; Tsuchida, E. Electrochim. Acta 1983,
88, 169. 28, 833.
(543) (a) Fenton, D. E.; Parker, J. M.; Wright, P. V. Polymer 1973, (560) Du Pasquier, A.; Zheng, T.; Amatucci, G. G.; Gozdz, A. S. J.
14, 589. (b) Wright, P. V. Br. Polym. J. 1975, 7, 319. (c) Wright, Power Sources 2001, 97/98, 758.
P. V. J. Polym. Sci., Polym. Phys. Ed. 1976, 14, 955. (561) Wang, H.; Huang, H.; Wunder, S. L. J. Electrochem. Soc. 2000,
(544) (a) Armand, M. B.; Chabagno, J. M.; Duclot, M. The 2nd 147, 2853.
International Conference on Solid Electrolytes, St. Andrews, (562) Huang, H.; Wunder, S. L. J. Electrochem. Soc. 2001, 148, A279.
Scotland, 1978; Abstract No. 65. (b) Armand, M. B.; Chabagno, (563) Han, K. N.; Seo, H. M.; Kim, J. K.; Kim, Y. S.; Shin, D. Y.; Jung,
J. M.; Duclot, M. In Fast Ion Transport in Solids; Duclot, M. J., B. H.; Lim, H. S.; Eom, S. W.; Moon, S. I. J. Power Sources 2001,
Vashishta, P., Mundy, J. N., Shenoy, G. K., Eds.; North-Holland, 101, 196.
Amsterdam, 1979.
(545) Angell, C. A.; Liu, C.; Sanchez, E. Nature 1993, 362, 137. CR030203G
Chem. Rev. 2004, 104, 4419−4462 4419
Battery Separators
Pankaj Arora* and Zhengming (John) Zhang
Celgard, LLC, 13800 South Lakes Dr., Charlotte, North Carolina 28273
Received March 30, 2004
Contents
1. Introduction and Scope 4419
2. Battery and Separator Market 4420
3. Separator and Batteries 4421
4. Separator Requirements 4422
5. Separator Types 4422
5.1. Microporous Separators 4422
5.2. Nonwovens 4422
5.3. Ion Exchange Membranes 4423
5.4. Supported Liquid Membranes 4423
5.5. Polymer Electrolyte 4423
5.6. Solid Ion Conductors 4423
6. Separator for Nonaqueous Batteries 4423
6.1. Lithium Ion 4424 Pankaj Arora is a Senior Research Engineer at Celgard LLC in Charlotte,
6.1.1. Separator Development 4424 NC. He specializes in the design and modeling of electrochemical power
6.1.2. Separator Requirements 4427 sources and is currently working in the Battery Applications Laboratory
of Celgard, where he helps guide separator development work for lithium
6.1.3. Separator Properties/Characterization 4429 batteries. He has a B.Tech. degree in Electrochemical Engineering from
6.1.4. Effect of Separator on Cell Performance 4436 the Central Electrochemical Research Institute in Karaikudi, India, and a
and Safety Ph.D. degree in Chemical Engineering from the University of South
6.2. Lithium Polymer 4440 Carolina, Columbia, SC. Pankaj can be reached by email at pankajarora@
celgard.com.
6.3. Lithium-Ion Gel Polymer 4441
6.4. Lithium Primary Systems 4443
6.4.1. Separator Requirements 4443
6.4.2. Chemistries 4444
7. Separator for Aqueous Batteries 4445
7.1. Leclanche (Zinc Carbon) 4446
7.2. Alkaline Zinc MnO2 4446
7.3. Lead-Acid Batteries 4447
7.3.1. Flooded Electrolyte Lead Acid 4447
7.3.2. Valve Regulated Lead Acid (VRLA) 4449
7.4. Nickel Systems 4450
7.4.1. Nickel−Cadmium 4450
7.4.2. Nickel−Metal Hydride 4451
7.4.3. Nickel−Hydrogen 4452
7.5. Zinc Systems 4452 Zhengming (John) Zhang is Vice President of New Technology at Celgard
7.5.1. Silver−Zinc 4452 LLC in Charlotte, NC. He has been working in Solid State Ionics, Batteries,
and Battery Separators since 1984. He has published more than 50 papers
7.5.2. Nickel−Zinc 4454 and patents and has co-edited a book on battery. He has been a invited
7.5.3. Zinc−Air 4455 speaker for many professional conferences, invited lecturer for United
7.5.4. Zinc−Bromine 4456 Nations Development Program, and is a Visiting Professor at Xiamen
7.6. Redox Flow Batteries 4456 University, Fujian, China. He has a B.S. in Mechanical Engineering from
Shanghai University, Shanghai, China, an M.S. in Electrochemistry from
8. Mathematical Modeling of Batteries/Separators 4457 Shandong University, Jinan City, China, and a Ph.D. in Materials Chemistry
9. Summary 4458 from the University of California at Santa Barbara, Santa Barbara, CA.
10. Future Directions 4458 John can be reached by email at zjzhang@celgard.com.
11. Acknowledgments 4459 1. Introduction and Scope
12. References 4459
Many advances have been made in battery tech-
nology in recent years, both through continued
* Corresponding author. E-mail: pankajarora@celgard.com. Tele- improvement of specific electrochemical systems and
phone: 704 587 8478. Fax: 704 588 7393 through the development and introduction of new
10.1021/cr020738u CCC: $48.50 © 2004 American Chemical Society
4420 Chemical Reviews, 2004, Vol. 104, No. 10 Arora and Zhang
battery chemistries. Nevertheless, there is still no one requirements, properties, and characterization tech-
“ideal” battery that gives optimum performance niques are described with respect to lithium-ion
under all operating conditions. Similarly, there is no batteries. The separators used in other batteries are
one separator that can be considered “ideal” for all only discussed briefly. Despite the widespread use
battery chemistries and geometries. of separators, a great need still exists for improving
A separator is a porous membrane placed between the performance, increasing the life, and reducing the
electrodes of opposite polarity, permeable to ionic flow cost of separators. In the following sections, an
but preventing electric contact of the electrodes. A attempt is made to discuss key issues in various
variety of separators have been used in batteries over separators with the hope of bringing into focus
the years. Starting with cedar shingles and sausage present and future directions of research and devel-
casing, separators have been manufactured from opment in separator technologies.
cellulosic papers and cellophane to nonwoven fabrics,
foams, ion exchange membranes, and microporous 2. Battery and Separator Market
flat sheet membranes made from polymeric materi-
als. As batteries have become more sophisticated, The battery industry has seen enormous growth
separator function has also become more demanding over the past few years in portable, rechargeable
and complex. battery packs. The majority of this surge can be
attributed to the widespread use of cell phones,
Separators play a key role in all batteries. Their personal digital assistants (PDA’s), laptop computers,
main function is to keep the positive and negative and other wireless electronics. Batteries remained
electrodes apart to prevent electrical short circuits the mainstream source of power for systems ranging
and at the same time allow rapid transport of ionic from mobile phones and PDA’s to electric and hybrid
charge carriers that are needed to complete the electric vehicles. The world market for batteries was
circuit during the passage of current in an electro- approximately $41 billion in 2000, which included
chemical cell.1,2 They should be very good electronic $16.2 billion primary and $24.9 billion secondary
insulators and have the capability of conducting ions cells.21
by either intrinsic ionic conductor or by soaking
A recent study from Freedonia has predicted ag-
electrolyte. They should minimize any processes that
gregate U.S. demand for primary and secondary
adversely affect the electrochemical energy efficiency
batteries to climb 5.5% annually through 2007 to $14
of the batteries.
billion. The growth will be driven by strong demand
Very little work (relative to research of electrode for battery-powered electronic devices like digital
materials and electrolytes) is directed toward char- cameras and 3G wireless phones, and increasing
acterizing and developing new separators. Similarly, production of electrical and electronic equipment. The
not much attention has been given to separators in secondary battery demand is expected to outpace the
publications reviewing batteries.1-10 A number of primary battery market gains through 2007 benefit-
reviews on the on cell fabrication, their performance, ing from strong growth in the use of high-drain
and application in real life have appeared in recent portable electronic devices. The lead-acid batteries
years, but none have discussed separators in detail. will account for over half of all rechargeable demand
Recently a few reviews have been published in both in 2007, although lithium-ion and NiMH batteries
English and Japanese which discuss different types will record the strongest growth. Alkaline batteries
of separators for various batteries.11-20 A detailed could remain the dominant type, accounting for more
review of lead-acid and lithium-ion (li-ion) battery than two thirds of all primary battery sales in 2007.22
separators was published by Boehnstedt13 and Spot- The rechargeable battery (NiCd, NiMH, and
nitz,14 respectively, in the Handbook of Battery lithium-ion) market for 2003 for portable electronics
Materials. Earlier Kinoshita et al. had done a survey was around $5.24 billion, around 20% more then
of different types of membranes/separators used in 2002. The lithium-ion battery market was around
different electrochemical systems, including batter- $3.8 billion (∼73%). They are now used in more than
ies.11 90% of cellphones, camcorders, and portable comput-
The majority of the separators currently used in ers, worldwide, and have also been adopted in power
batteries were typically developed as spin-offs of tools recently.23
existing technologies. They were usually not devel- The tremendous progress in lithium-ion cells is
oped specifically for those batteries and thus are not clearly visible with as much as a 2-fold increase in
completely optimized for systems in which they are the volumetric and gravimetric energy density for
used. One positive result of adapting existing tech- both 18650 and prismatic cells between 1994 and
nologies is that they are produced in high volume at 2002. In past few years the lithium-ion production
relatively low cost. The availability of low cost has expanded to South Korea (Samsung, LG, etc.)
separators is an important consideration in the and China (BYD, B&K, Lishen, etc.) from Japan.
commercialization of batteries, because the battery Several Japanese (Sanyo, Sony, MBI, NEC, etc.) and
industry traditionally operates with thin profit mar- Korean (LG Chemical) manufacturers have also
gins and relatively small research budgets. moved their manufacturing plants to China.23 Japan,
The purpose of this paper is to describe the various which controlled 94% of the global rechargeable
types of separators based on their applications in battery market in 2000, has seen its market share
batteries and their chemical, mechanical and elec- drop to about 65% of the global market.23-25 The
trochemical properties, with particular emphasis on continued growth in lithium-ion battery market has
separators for lithium-ion batteries. The separator led to a strong demand for battery separators. All the
Battery Separators Chemical Reviews, 2004, Vol. 104, No. 10 4421
Figure 1. Typical battery configurations: (a) button cell; (b) stack lead-acid; (c) spiral wound cylindrical lithium-ion; (d)
spiral wound prismatic lithium-ion.
major separator manufacturers (Celgard, Asahi, and accommodate a particular cell shape and design. The
Tonen) have either increased their capacity in 2003 separators are either stacked between the electrodes
or are planning to increase it in 2004.26-28 or wound together with electrodes to form jellyrolls
There is not too much information available on the as shown in Figure 1. Stacked cells are generally held
battery separator market in the literature. It is together by pressure from the cell container. The
estimated that about 30% of the rechargeable lithium lithium-ion gel polymer stacked cells are prepared
battery market or $1.5 billion is the size of the by bonding/laminating layers of electrodes and sepa-
battery materials or components market. Battery rators together. The separator properties should not
separators for lithium batteries are about a $330 change significantly during the bonding process. In
million market within the total battery components some cases, the separators are coated to help in
market.29,30 Recently, the Freedonia Group has re- bonding process, thus reducing the interfacial resis-
ported that the U.S. demand for battery separators tance.33-35
will increase to $410 million in 2007 from $237 In the conventional way of making spirally wound
million in 1977 and $300 million in 2002, respec- cells, two layers of separators are wound along with
tively.31,32 the positive and negative electrodes, resulting in a
positive/separator/negative/separator configuration.
3. Separator and Batteries They are wound as tightly as possible to ensure good
interfacial contact. This requires the separators to
Batteries are built in many different shapes and be strong to avoid any contact between the electrodes
configurationssbutton, flat, prismatic (rectangular), through the separator. The separator also must not
and cylindrical (AA, AAA, C, D, 18650, etc.). The cell yield and reduce in width, or else the electrodes may
components (including separators) are designed to contact each other. Once wound, the jellyroll is
inserted into a can, and filled with electrolyte. The In most batteries, the separators are either made
separator must be wetted quickly by the electrolyte of nonwoven fabrics or microporous polymeric films.
to reduce the electrolyte filling time. A header is then Batteries that operate near ambient temperatures
crimped into the cell to cover the can from top. In usually use separators fabricated from organic ma-
some prismatic cells, the jellyroll is pressed at high terials such as cellulosic papers, polymers, and other
temperatures and pressures and then inserted into fabrics, as well as inorganic materials such as
thin prismatic (rectangular) cans. A typical 18650 asbestos, glass wool, and SiO2. In alkaline batteries,
lithium-ion cell uses around 0.07-0.09 m2 of separa- the separators used are either regenerated cellulose
tor, which is approximately 4-5% of the total cell or microporous polymer films. The lithium batteries
weight.36 with organic electrolytes mostly use microporous
films.
4. Separator Requirements For the sake of discussion, we have divided the
separators into six typessmicroporous films, non-
A number of factors must be considered in selecting wovens, ion exchange membranes, supported liquid
the best separator for a particular battery and membranes, solid polymer electrolytes, and solid ion
application. The characteristics of each available conductors. A brief description of each type of separa-
separator must be weighed against the requirements tor and their application in batteries are discussed
and one selected that best fulfills these needs. A wide below.
variety of properties are required of separators used
in batteries. The considerations that are important 5.1. Microporous Separators
and influence the selection of the separator include They are fabricated from a variety of inorganic,
the following: organic, and naturally occurring materials and gen-
•Electronic insulator erally contain pores that are greater than 50-100 Å
•Minimal electrolyte (ionic) resistance in diameter. Materials such as nonwoven fibers (e.g.
•Mechanical and dimensional stability nylon, cotton, polyesters, glass), polymer films (e.g.
•Sufficient physical strength to allow easy handling polyethylene (PE), polypropylene (PP), poly(tetrafluo-
•Chemical resistance to degradation by electrolyte, roethylene) (PTFE), poly(vinyl chloride) (PVC)), and
impurities, and electrode reactants and products naturally occurring substances (e.g. rubber, asbestos,
•Effective in preventing migration of particles or wood) have been used for microporous separators in
colloidal or soluble species between the two electrodes batteries that operate at ambient and low tempera-
•Readily wetted by electrolyte tures (<100 °C). The microporous polyolefins (PP, PE,
•Uniform in thickness and other properties or laminates of PP and PE) are widely used in lithium
The order of importance of the various criteria based nonaqueous batteries (section 6.1), and filled
varies, depending on the battery applications. The polyethylene separators in lead-acid batteries (section
above list presents a broad spectrum of requirements 7.3), respectively.
for separators in batteries. In many applications, a
compromise in requirements for the separator must 5.2. Nonwovens
generally be made to optimize performance, safety, Nonwovens are textile products that are manufac-
cost, etc. For example, batteries that are character- tured directly from fibers. They are defined as a
ized by small internal resistance and consume little manufactured sheet, web, or matt of directionally or
power require separators that are highly porous and randomly oriented fibers, bonded by friction, and/or
thin, but the need for adequate physical strength may cohesion, and/or adhesion excluding paper and prod-
require that they be thick. ucts which are woven, tufted, stichbounded incorpo-
In addition to the above general requirements each rating binding yarns or filaments, or felted by
battery type has other requirements essential for wetmilling whether or not additionally needed. The
good performance and safety. For example, separa- fibers may be of natural or manmade origin. They
tors in sealed nickel-cadmium (NiCd) and nickel- may be staple or continuous filaments or maybe
metal hydride (NiMH) batteries should be highly formed in situ.37
permeable to gas molecules for overcharge protection, The macroporous fibrous matrix is either dry laid,
the separator in lithium-ion cells should have a meltblown, or wet laid. The wet laid process is very
shutdown feature for good safety, separators for similar to the papermaking process. The fibers are
alkaline batteries should be flexible enough to be bonded together by chemical or thermal bonding. The
wrapped around the electrodes, and the separator for meltblown process is a binderless process and there
an SLI (starting, lighting and ignition) battery could the polymer fiber web is extruded. Typical pore sizes
also serve as a mechanical-shock cushion. of the fibrous matrix vary from 1 to 100 µm.
Nonwovens are widely utilized as separators for
several types of batteries. Lightweight, wet laid
5. Separator Types nonwovens made from cellulose, poly(vinyl alcohol),
Separators for batteries can be divided into differ- and other fibers have achieved considerable success
ent types, depending on their physical and chemical as separators for popular primary alkaline cells of
characteristics. They can be molded, woven, non- various sizes. The key nonwoven attributes include
woven, microporous, bonded, papers, or laminates. consistently uniform basis weight, thickness, porosity
In recent years, there has been a trend to develop and resistance to degradation by electrolytes. Non-
solid and gelled electrolytes that combine the elec- wovens are also successfully employed as separators
trolyte and separator into a single component. in NiCd’s.
Table 1. Types of Separators Used in Different Type of Secondary Lithium Batteries

battery system type of separator composition
lithium-ion (liquid electrolyte) microporous polyolefins (PE, PP, PP/PE/PP)
lithium-ion gel polymer microporous PVdF
microporous polyolefins (PE, PP, PP/PE/PP) coated with
PVdF or other gelling agents
lithium-polymer (e.g. Li-V6O13) polymer electrolyte poly(ethylene oxide) with lithium salt
The materials used in nonwoven fabrics include a electrolytes/membranes are discussed briefly in sec-
single polyolefin, or a combination of polyolefins, such tion 6.3.
as polyethylene (PE), polypropylene (PP), polyamide
(PA), poly(tetrafluoroethylene) (PTFE), polyvinylidine 5.5. Polymer Electrolyte
fluoride (PVdF), and poly(vinyl chloride) (PVC).
Nonwoven fabrics have not, however, been able to Polymer electrolytes (e.g., poly(ethylene oxide),
compete with microporous films in lithium-ion cells. poly(propylene oxide)) have attracted considerable
This is most probably because of the inadequate pore attention for batteries in recent years. These poly-
structure and difficulty in making thin (<25 µm) mers form complexes with a variety of alkali metal
nonwoven fabrics with acceptable physical properties. salts to produce ionic conductors that serve as solid
electrolytes. Their use in batteries is still limited due
5.3. Ion Exchange Membranes to poor electrode/electrolyte interface and poor room
temperature ionic conductivity. Because of the rigid
These membranes are generally fabricated from structure, they can also serve as the separator.
polymeric materials containing pores with diameters Polymer electrolytes are discussed briefly in section
of less than 20 Å. The transport properties of ions in 6.2.
these membranes are characterized by strong inter-
actions between the permeating species and the 5.6. Solid Ion Conductors
molecular structure of the polymer. This interaction
is due to the presence of ion-exchange groups in the They serve as both separator and electrolyte. These
membrane, which allows the membrane to discrimi- are generally inorganic materials that are impervious
nate between permeating or migrating ions by virtue barriers to gases and liquids. They allow one or more
of their specific charge. kinds of ions to migrate through their lattice when a
Radiation grafted membranes such as Permion potential gradient or a chemical gradient is present.
manufactured by RAI Research Corporation are ion- These types of separators are beyond the scope of this
exchange membranes. Such membranes are used as article.
battery separators in alkaline batteries. They are
made from PE, PP, or Teflon-based films, which have 6. Separator for Nonaqueous Batteries
excellent oxidation resistance and superior chemical
resistance to alkali. However, they are totally im- All lithium based batteries use nonaqueous elec-
pervious to electrolyte flow, and therefore, they have trolytes because of the reactivity of lithium in aque-
almost infinite resistance as a separator in this form. ous solution and because of the electrolyte’s stability
By using radiation grafting and cross-linking tech- at high voltage. The majority of these cells use
niques, however, selected chemical species are grafted microporous membranes made of polyolefins. In some
as pendant chains to the base structure of the linear cases, nonwovens made of polyolefins are either used
polymer without altering the inert backbone. This alone or with microporous separators. This section
modification imparts desirable hydrophilic properties will mainly focus on separators used in secondary
to the films without materially impairing their excel- lithium batteries followed by a brief summary of
lent chemical resistance. This paper provides a very separators used in lithium primary batteries.
limited discussion on ion exchange membranes, as Lithium secondary batteries can be classified into
their application in batteries is very limited. three types, a liquid type battery using liquid elec-
trolytes, a gel type battery using gel electrolytes
5.4. Supported Liquid Membranes mixed with polymer and liquid, and a solid type
battery using polymer electrolytes. The types of
These types of separators consist of a solid matrix separators used in different types of secondary
and a liquid phase, which is retained in the mi- lithium batteries are shown in Table 1. The liquid
croporous structure by capillary forces. To be effective lithium-ion cell uses microporous polyolefin separa-
for batteries, the liquid in the microporous separator, tors while the gel polymer lithium-ion cells either use
which generally contains an organic phase, must be a PVdF separator (e.g. PLION cells) or PVdF coated
insoluble in the electrolyte, chemically stable, and microporous polyolefin separators. The PLION cells
still provide adequate ionic conductivity. Several use PVdF loaded with silica and plasticizer as
types of polymers, such as polypropylene, polysulfone, separator. The microporous structure is formed by
poly(tetrafluoroethylene), and cellulose acetate, have removing the plasticizer and then filling with liquid
been used for porous substrates for supported-liquid electrolyte. They are also characterized as plasticized
membranes. The PVdF coated polyolefin-based mi- electrolyte. In solid polymer lithium-ion cells, the
croporous membranes used in gel-polymer lithium- solid electrolyte acts as both electrolyte and separa-
ion battery fall into this category. Gel polymer tor.
Figure 2. Polyolefin separators used in lithium-ion batteries.

6.1. Lithium Ion polyethylene (PE), polypropylene (PP), or laminates40
of polyethylene and polypropylene. They are made
The past decade has seen significant advances in up of polyolefins materials because they provide
the ambient temperature lithium battery technology. excellent mechanical properties, chemical stability
Lithium-ion batteries are the preferred power source and acceptable cost.41,42 They have been found to be
for most portable electronics because of their higher compatible with the cell chemistry and can be cycled
energy density, longer cycle life, and higher opera- for several hundred cycles without significant deg-
tional voltage as compared to NiCd and NiMH radation in chemical or physical properties.
systems. In 2002, 66% of the total rechargeable
Commercial membranes offer pore size in the range
battery market for mobile IT and communication
0.03-0.1 µm, and 30-50% porosity. The low melting
devices used lithium-based batteries and the rest
point of PE enables their use as a thermal fuse. As
used nickel-based batteries.38,39
the temperature approaches the melting point of the
A typical lithium-ion cell consists of a positive
polymer, 135 °C for PE and 165 °C for PP, porosity
electrode composed of a thin layer of powdered metal
is lost. The trilayer material (PP/PE/PP)43 has been
oxide (e.g., LiCoO2) mounted on aluminum foil and
developed by Celgard where a PP layer is designed
a negative electrode formed from a thin layer of
to maintain the integrity of the film, while the low
powdered graphite, or certain other carbons, mounted
melting point of PE layer is intended to shutdown
on a copper foil. The two electrodes are separated by
the cell if an over-temperature condition is reached.108
a porous plastic film soaked typically in LiPF6 dis-
Asahi Kasai’s flat-film membrane “Hipore” is avail-
solved in a mixture of organic solvents such as
able in thicknesses ranging from 20 µm to several
ethylene carbonate (EC), ethyl methyl carbonate
hundred micrometers, and with highly uniform pore
(EMC), or diethyl carbonate (DEC). In the charge/
sizes ranging from 0.05 to 0.5 µm.44 The major
discharge process, lithium ions are inserted or ex-
manufacturers of lithium-ion battery separators along
tracted from the interstitial space between atomic
with their typical products are listed in Table 2.
layers within the active materials.
Sony’s introduction of the rechargeable lithium-ion In recent years there have been a strong demand
battery in the early 1990s precipitated a need for new for higher capacity lithium-ion cells because of the
separators that provided not only good mechanical strong growth in portable electronics. One way to
and electrical properties but also added safety through achieve higher capacity is by reducing the thickness
a thermal shutdown mechanism. Although a variety of separators. The battery manufacturers have started
of separators (e.g., cellulose, nonwoven fabric, etc.) using 20 and 16 µm separators in higher capacity
have been used in different type of batteries, various (>2.0 A h) cylindrical cells, and 9 µm separators in
studies on separators for lithium-ion batteries have lithium-ion gel polymer cells.
been pursued in past few years as separators for Nonwoven materials have also been developed for
lithium-ion batteries require different characteristics lithium-ion cells but have not been widely accepted,
than separators used in conventional batteries. in part due to the difficulty in fabricating thin
A novel microporous separator using polyolefins materials with good uniformity and high strength.14
has been developed and used extensively in lithium- Nonwoven separators have been used in button cells
ion batteries since it is difficult for conventional and bobbin cells when thicker separators and low
separator materials to satisfy the characteristics discharge rates are acceptable.
required in lithium-ion batteries. In lithium-ion
6.1.1. Separator Development
batteries two layers of separators are sandwiched
between positive and negative electrodes and then The process for making lithium-ion battery separa-
spirally wound together in cylindrical and prismatic tors can be broadly divided into dry45,46 and wet47
configurations. The pores of the separator are filled processes. Both processes usually employ one or more
with ionically conductive liquid electrolyte. orientation steps to impart porosity and/or increase
Microporous polyolefin membranes (see Figure 2) tensile strength. The dry process involves melting a
in current use are thin (<30 µm) and are made of polyolefin resin, extruding it into a film, thermally
Table 2. Major Manufacturers of Lithium-Ion Battery Separators along with Their Typical Products
manufacturer structure composition process trade name
Asahi Kasai single layer PE wet HiPore
Celgard LLC single layer PP, PE dry Celgard
multilayer PP/PE/PP dry Celgard
PVdF coated PVdF, PP, PE, PP/PE/PP dry Celgard
Entek Membranes single layer PE wet Teklon
Mitsui Chemical single layer PE wet
Nitto Denko single layer PE wet
DSM single layer PE wet Solupur
Tonen single layer PE wet Setela
Ube Industries multi layer PP/PE/PP dry U-Pore
Table 3. Manufacturing Process of Typical Microporous Film

process mechanism raw material properties typical membranes manufacturers
dry process drawing polymer simple process PP, PE, PP/PE/PP Celgard, Ube
anisotropic film
wet process phase separation polymer + solvent isotropic film PE Asahi, Tonen
polymer + solvent + filler large pore size PE Asahi
high porosity
annealing it to increase the size and amount of

lamella crystallites, and precisely stretching it to
form tightly ordered micropores.48-52 In this process,
a row lamellar crystal structure is generated in the
polymer in the initial extrusion step. This nonporous
structure is highly oriented as a result of extrusion
and annealing conditions. The films are then stretched
to form micropores. This microporous structure is
continuous throughout the bulk interior of the mem-
brane.53
Polypropylene and polyethylene microporous films
obtained by this method are available from Cel-
gard48,50,54,55 and Ube.56 The dry process is technologi-
cally convenient because no solvents are required.
However, only a uniaxial stretching method has been Figure 3. Generalized process for lithium-ion separator
successful to date, and as a result, the pores are manufacturing.66 Each step of the separator manufacturing
slitlike in shape and the mechanical properties of process has online detection systems to monitor the quality
films are anisotropic. The tensile strength in the of the separator.
lateral direction is relatively low.
Wet process (phase inversion process)57,58 involves polyolefin blending solution. The manufacturing pro-
mixing of hydrocarbon liquid or some other low- cess of typical microporous film by dry and wet
molecular weight substance generally with a poly- process is compared in Table 3.
olefin resin, heating and melting the mixture, ex- A simplified flowchart for separator manufacturing
truding the melt into a sheet, orientating the sheet process is shown in Figure 3.66 The virgin polymer
either in the machine direction (MD) or biaxially, and is prepared and mixed with processing aids (e.g.,
then extracting the liquid with a volatile solvent.45,59 antioxidants, plasticizer, etc.) and then extruded. The
Separators made by the wet process are available extruded polymer then goes through different steps,
from Asahi Kasei,60 Tonen,61-63 and Mitsui Chemi- which vary from process to process. For the dry
cals64 and more recently from Polypore/Membrana process, it can involve film annealing and stretching,
and Entek.65 The structure and properties of the while for the wet process, it can involve solvent
membranes can be changed by controlling the com- extraction and stretching. The finished film is then
position of the solutions and the evaporation or slit into required widths and packed into boxes and
subtractions of solvents in the gelation and solidifica- shipped to the battery manufacturers. With the
tion processes. The separators made by wet process advent of thinner separators, the film handling
use ultrahigh-molecular-weight polyethylene (UH- during manufacturing steps has become very impor-
MWPE). The use of UHMWPE gives good mechanical tant for the final quality of the film. Each step of the
properties as well as some degree of melt integrity. separator manufacturing process has online detection
Ihm et al. have given a nice overview of the wet systems to monitor the quality of the separators.
process by preparing a separator with polymer blends Uniaxially oriented films generally have high
of high-density polyethylene (HDPE) and ultrahigh strength in only one direction, whereas biaxially
molecular weight polyethylene (UHMWPE).58 They oriented films are relatively strong in both machine
showed that the mechanical strength and drawing direction (MD) and transverse direction (TD). Al-
characteristics are influenced by the content and the though intuitively one might expect biaxially oriented
molecular weight of the UHMWPE contained in a films to be preferred over uniaxially oriented films,
Table 4. Typical Properties of Some Commercial Microporous Membranes

separator/properties Celgard 2730 Celgard 2400 Celgard 2320 Celgard 2325 Asahi Hipore Tonen Setela
structure single layer single layer trilayer trilayer single layer single layer
composition PE PP PP/PE/PP PP/PE/PP PE PE
thickness (um) 20 25 20 25 25 25
gurley (s) 22 24 20 23 21 26
ionic resistivitya (Ω cm2) 2.23 2.55 1.36 1.85 2.66 2.56
porosity (%) 43 40 42 42 40 41
melt temp (°C) 135 165 135/165 135/165 138 137
a
In 1 M LiPF6 EC:EMC (30:70 by volume).
in practice biaxial orientation provides no perfor- Microporous polyethylene separator material com-
mance advantage. In fact, biaxial orientation tends posed of a combination of randomly oriented thick
to introduce TD shrinkage. This shrinkage, at el- and thin fibrils of ultrahigh molecular weight poly-
evated temperatures, can allow electrodes to contact ethylene (UHMWPE), Solupur, manufactured by
each other. The separator must have sufficient DSM Solutech, is also an interesting separator mate-
strength in the machine direction so that it does not rial for lithium-ion batteries. Solupur is fabricated
decrease in width or break under the stress of in standard grades with base weights ranging from
winding. The strength in the transverse direction is 7 to 16 g/m2 and mean pore size ranging from 0.1 to
not as important as that in the machine direction 2.0 µm and a porosity of 80-90%.72 Ooms et al.
during the process of making spirally wound batter- carried out a study on a series of DSM Solupur
ies. The minimum generally practical requirement materials with different permeability. Rate capability
for the mechanical strength of the 25-µm separator and cycling tests of these materials were compared
is 1000 kg/mm2.58 with commercial available separators in CR2320 type
The typical properties of some commercial mi- coin cells. Solupur materials showed low tortuosity,
croporous membranes are summarized in Table 4. high strength and puncture resistance, excellent
Celgard 2730 and Celgard 2400 are single layer PE wettability, and good high rate capability and low-
and PP separators, respectively, while Celgard 2320 temperature performance because of its high porosity
and 2325 are trilayer separators of 20 and 25 µm and UHMWPE structure.73
thickness. Asahi and Tonen separators are single Recently Nitto Denko has developed a battery
layer PE separators made by the wet process. Basic separator made by a wet process that had high
properties, such as thickness, gurley, porosity, melt puncture strength and high heat rupture resistance.74
temperature, and ionic resistivity are reported in They used a polyolefin resin with a high molecular
Table 4. These properties are defined in section 6.1.3. weight rubber as its main component materials and
Efforts have been made to find a new route for dry cross-linked through oxidation in air. The melt
process using biaxial stretching techniques for pre- rupture temperature, as measured by thermomech-
paring polypropylene microporous films, which may nical analysis was over 200 °C in this material. They
have submicrometer pore sizes and narrow pore size also tried cross-linking ultrahigh molecular weight
distributions and high permeability to gases and polyethylene with electron-beam and ultraviolet ir-
liquids combined with good mechanical properties. radiation, but this had the side effect of causing
The biaxially stretched polypropylene microporous deterioration in the polyolefin including rupture of
films (Micpor) were made by using nonporous polypro- the main chains and therefore resulted in reduced
pylene films of high β-crystal content.67 The porosity strength.
of these films can be as high as 30-40%, with an ENTEK Membranes LLC has developed Teklons
average pore size of approximately 0.05 µm. The a highly porous, ultrahigh molecular weight polyeth-
pores on the surface were almost circular in shape ylene separator for lithium-ion batteries. At the
compared to slitlike pores observed in uniaxial writing of this publication, the separator is available
stretched samples and exhibited high permeability in small quantities. Pekala et al. characterized Cel-
to fluids with good mechanical properties and almost gard, Setela, and Teklon separators in terms of their
circular pore shape with narrow pore size distribu- physical, mechanical, and electrical properties.75
tion.68-70 Celgard’s separators are by far the best-character-
The PP/PE bilayers40 and PP/PE/PP trilayer sepa- ized battery separators in the literature as they have
rators were developed by Celgard. Multilayer separa- been widely used in numerous battery systems.
tors offer advantages by combining the lower melting Bierenbam et al.45 has described the process, physical
temperature of PE with the high-temperature strength and chemical properties, and end-use applications.
of PP. Nitto Denko has also patented a single-layer Fleming and Taskier76 described the use of Celgard
separator made from a blend of PE/PP by the dry microporous membranes as battery separators. Hoff-
stretch process.71 According to the patent, the sepa- man et al.77 presented a comparison of PP and PE
rator has microporous regions of PE and PP. On Celgard microporous materials. Callahan discussed
heating in an oven, the impedance of the separator a number of novel uses of Celgard membranes.
increases near the melting point of PE and the Callahan and co-workers98 also characterized Celgard
impedance remains high until beyond the melting membranes by SEM image analysis, mercury poro-
point of PP. However, battery performance data have simetry, air permeability, and electrical resistivity,
not been presented. and later they characterized the puncture strength
and temperature/impedance data for Celgard mem- especially at low temperatures. They tested the
branes.40 Spotnitz et al. reported short-circuit behav- performance and safety behavior of Separion separa-
ior in simulated, spirally wound cells, as well as tor in 18650 cells and found the performance to be
impedance/temperature behavior and thermomechan- comparable to that of polyolefin-based separators.83
ical properties.108 Yu78 found that a trilayer structure The potential use of polymeric ion-exchange mem-
of PP/PE/PP Celgard microporous membranes pro- branes in the next generation single-ion secondary
vided exceptional puncture strength. lithium polymer batteries was shown by Sachan et
Nonwoven materials such as cellulosic fibers have al.84,85 Conductivities exceeding 10-4 S/cm with trans-
never been successfully used in lithium batteries. ference numbers of unity were achieved for Nafion
This lack of interest is related to the hygroscopic converted to the Li+ salt form.
nature of cellulosic papers and films, their tendency To obtain a thin (less than 15 µm) separator for
to degrade in contact with lithium metal, and their lithium batteries, Optodot has taken a different
susceptibility to pinhole formation at thickness of less approach of high-speed coating of a metal oxide sol-
than 100 µm. For future applications, such as electric gel coating on a smooth surface followed by a delami-
vehicles and load leveling systems at electric power nation step to provide the free-standing separator.
plants, cellulosic separators may find a place because Using this approach, separator with thicknesses from
of their stability at higher temperatures when com- 6 to 11 µm was made on large-scale production
pared to polyolefins. They may be laminated with coating equipment.86 They found that the sol-gel
polyolefin separators to provide high-temperature
separators with a thickness in the middle of this
melt integrity.
range of 8-9 µm have the preferred combination of
Asahi Chemical Industry carried out an explor- thinness and strength. The metal oxide sol-gel
atory investigation to determine the requirements for coating is water-based with no organic solvents
cellulose based separators for lithium-ion batteries.79 present. The coating formulations include a polymer
In an attempt to obtain an acceptable balance of and a surfactant. The polymer provides improved
lithium-ion conductivity, mechanical strength, and coating rheology, mechanical strength, and other
resistance to pinhole formation, they fabricated a properties. The surfactant provides improved wetting
composite separator (39-85 µm) that consists of properties on the substrate. The films prepared were
fibrilliform cellulosic fibers (diameter 0.5-5.0 µm) around 11 µm thick, with 45% porosity, and were
embedded in a microporous cellulosic (pore diam- completely wettable in nonaqueous electrolyte and
eter: 10-200 nm) film. The fibers can reduce the had a melt temperature greater than 180 °C. These
possibility of separator meltdown under exposure to films are relatively thin and should help in increasing
heat generated by overcharging or internal short- the capacity but may not be strong enough for tightly
circuiting. The resistance of these films was equal to wound cells. Moreover, the shutdown temperature of
or lower than the conventional polyolefin-based mi- the separator seems to be very high and thus not
croporous separators. The long-term cycling perfor- suitable for lithium-ion batteries.
mance was also very comparable.
Gineste et al. carried out the grafting of hydrophilic
Pasquier et al.80 used paper based separators in flat monomers onto PP or PE separators to improve the
pouch type lithium-ion batteries and compared the wettability of separators used in secondary lithium
performance with cells made with Celgard type batteries with a lower content of wetting agents.87,88
polyolefin based separators. The paper separators They used a PP film (Celgard 2505) of 50 µm
had good wetting properties and good mechanical thickness after irradiating in air by electron beams
properties but did not provide the shutdown effect with a dose ranging from 0.5 to 4 Mrad. The irradi-
essential for large lithium-ion batteries. Their resis- ated film was grafted by a monofunctional monomer
tance was similar to polyolefin separators, and when (acrylic acid, AA), in the presence of difunctional
all water traces were removed from paper, their
cross-linking agent (diethylene glycol dimethacrylate,
cycling performance was similar to that of Celgard
DEGDM). The separators start loosing mechanical
separators. The paper-based separators can be used
properties, when the grafting ratio is higher than
in small flat pouch type cells where high strength
50%.
and shutdown behavior is not required. For larger
spherically wound cells, which require strong separa-
6.1.2. Separator Requirements
tors with a shutdown feature, one can never use
paper-based separators. In lithium-based cells, the essential function of
Recently Degussa announced that they have de- battery separator is to prevent electronic contact,
veloped Separion separators for lithium batteries by while enabling ionic transport between the positive
combining the characteristics of flexible polymeric and negative electrodes. It should be usable on high-
separators with the advantages of chemical and speed winding machines and possess good shutdown
thermally resistant and hydrophilic ceramic materi- properties. The most commonly used separators for
als. Separion is produced in a continuous coating primary lithium batteries are microporous polypro-
process. Ceramic materials, e.g., alumina, silica, and/ pylene membranes. Microporous polyethylene and
or zirconia are slip coated and hardened onto a laminates of polypropylene and polyethylene are
support.81,82 According to Degussa, Separion separa- widely used in lithium-ion batteries.89 These materi-
tors have an excellent high temperature stability, als are chemically and electrochemically stable in
superior chemical resistance, and good wettability, secondary lithium batteries.
The general requirements90 for lithium-ion battery erties of trilayer (PP/PE/PP) separators with PP as
separators are given below. the outside layer and PE as inner layer are superior.
6.1.2.1. Thickness. The lithium-ion cells used in 6.1.2.7. Dimensional Stability. The separator
consumer applications use thin microporous separa- should lay flat and should not curl at the edges when
tors (<25 µm). The separators being developed for unrolled, as this can greatly complicate cell assembly.
EV/HEV applications will require thicker (∼40 µm) The separator should also not shrink when exposed
separators. The thicker the separator, the greater the to electrolyte. The cell winding should not affect the
mechanical strength and the lower the probability porous structure in any adverse way.
of punctures during cell assembly but the smaller the
amount of active materials that can be placed in the 6.1.2.8. Puncture Strength. The separators used
can. The thinner separators take up less space and in wound cells require a high puncture strength to
permit the use of longer electrodes. This increased avoid penetration of electrode material through the
both capacity and, by increasing the interfacial area, separator. If particulate material from the electrodes
rate capability. The thinness also makes it a low penetrates the separator, an electrical short will
resistance separator. result and the battery will be rejected. The separators
used in lithium-ion batteries require more strength
6.1.2.2. Permeability. The separators should not
then the one used in lithium primary batteries. The
limit the electrical performance of the battery under
normal conditions. Typically the presence of separa- primary lithium batteries have only one rough elec-
tor increases the effective resistivity of the electrolyte trode, and thus it requires less strength. As empiri-
by a factor of 6-7. The ratio of the resistivity of the cally observed, for most applications, the puncture
separator filled with electrolyte divided by the resis- strength should be at least 300 g/mil for separators
tivity of the electrolyte alone is called MacMullin used in lithium-ion cells. Mix penetration strength
number. MacMullin numbers are as high as 10-12 is a better measure of separator strength in a battery
have been used in consumer cells. compared to puncture strength.
6.1.2.3. Gurley (Air Permeability). Air perme- 6.1.2.9. Mix Penetration Strength. The suscep-
ability is proportional to electrical resistivity, for a tibility of separators to particle penetration is char-
given separator morphology. It can be used in place acterized by mix penetration strength.49 During the
of electrical resistance (ER) measurements once the winding of the spiral wrap construction considerable
relationship between gurley and ER is established. mechanical pressure is applied to the cathode-
The separator should have low gurley values for good separator-anode interface. Any loose particle could
electrical performance. be forced through the separator and short the cell.
6.1.2.4. Porosity. It is implicit in the permeability The mix penetration strength should be at least 100
requirement; typically lithium-ion battery separators kgf/mil for separators used in lithium-ion cells.
have a porosity of 40%. Control of porosity is very 6.1.2.10. Thermal Stability. Lithium-ion batter-
important for battery separators. Specification of ies can be poisoned by water, and so materials going
percent porosity is commonly an integral part of into the cell are typically dried at 80 °C under
separator acceptance criteria. vacuum. Under these conditions, the separator must
6.1.2.5. Wettability. The separators should wet not shrink significantly and definitely must not
out quickly and completely in typical battery elec- wrinkle. Each battery manufacturer has specific
trolytes. drying procedures. The requirement of less than 5%
6.1.2.6. Electrolyte Absorption & Retention. A shrinkage after 60 min at 90 °C (in a vacuum) in both
separator should be able to absorb and retain elec- MD and TD direction is a reasonable generalization.
trolyte. Electrolyte absorption is needed for ion 6.1.2.11. Pore Size. A key requirement of separa-
transport. The microporous membranes usually do tors for lithium batteries is that their pores be small
not swell on electrolyte absorption. enough to prevent dendritic lithium penetration
6.1.2.7. Chemical Stability. The separators should through them. Membranes with submicrometer pore
be stable in the battery for a long period of time. They sizes have proven adequate for lithium batteries.
should be inert to both strong reducing and strong
6.1.2.12. Tensile Strength. The separator is
oxidizing conditions and should not degrade or loose
wound with the electrodes under tension. The sepa-
mechanical strength or produce impurities, which can
rator must not elongate significantly under tension
interfere with the function of the battery. The sepa-
rator must be able to withstand the strong oxidizing in order to avoid contraction of the width. A tensile
positive electrode and the corrosive nature of the strength specification is sometimes given, but the key
electrolyte at temperatures as high as 75 °C. The parameter is Young’s modulus in the machine direc-
greater the oxidation resistance, the longer the tion. Since Young’s modulus is difficult to measure,
separator will survive in a cell. Polyolefins (e.g., a 2% offset yield is a good measure; less than 2%
polypropylene, polyethylene) exhibit high resistance offset at 1000 psi is acceptable for most winding
to most of the conventional chemicals, good mechan- machines.
ical properties, and a moderate temperature range 6.1.2.13. Skew. Ideally, when a strip of separator
for application making them ideal polymers for is laid out, the separator should be straight and not
lithium-ion battery separators. Polypropylene sepa- bow or skew. In practice, however, some skew is often
rators exhibit better oxidation resistance properties observed. If sufficiently extreme, the skew can cause
when in contact with the positive electrode in a misalignment between the electrodes and separator.
lithium-ion cell. Thus, the oxidation resistance prop- Skew can be measured by laying the separator flat
Table 5. General Requirements for Lithium-ion Battery Separator90

parameter goal
thicknessa,b (µm) <25
electrical resistance (MacMullin no.,c dimensionless) <8
electrical resistance (ohms cm2) <2
gurleyd (s) ∼ 25/mil
pore sizee (µm) <1
porosity (%) ∼ 40
puncture strengthf (g/mil) >300
mix penetration strength (kgf/mil) >100
shrinkageg (%) <5% in both MD and TD
tensile strengthh (%) <2% offset at 1000 psi
shutdown temp (°C) ∼130
high-temp melt integrity (°C) >150
wettability complete wet out in typical battery electrolytes
chemical stability stable in battery for long period of time
dimensional stability separator should lay flat; be stable in electrolyte
skew (mm/m) <0.2
a
ASTM D5947-96. Standard Test Methods for Physical Dimensions of Solid Plastics Specimens. ASTM International. b ASTM
D2103. Standard Specification for Polyethylene Film and Sheeting”, ASTM International. c Caldwell, D. L.; Poush, K. A. U.S.
Patent, 4,464,238, 1984. d ASTM D726. Standard Test Methods for Identification of Fibers in Textiles. ASTM International. e ASTM
E128-99. Standard test method for Maximum Pore Diameter and Permeability of Rigid Porous Filters for Laboratory Use. ASTM
International. f ASTM D3763. Standard Test Method for High-Speed Puncture Properties of Plastics using Load and Displacement
Sensors. ASTM International. g ASTM D1204. Standard Test methods for Linear Dimensional Changes of Nonrigid Thermoplastic
Sheeting or Film at Elevated Temperatures. ASTM International. h ASTM D882. Standard Test Method for Tensile Properties of
Thin Plastic Sheeting. ASTM International.
on a table parallel with a straight meter stick. The battery. The general requirements for lithium-ion
skew should be less than 0.2 mm/m of separator. battery separators are also summarized in Table 5.
6.1.2.14. Shutdown. Lithium-ion batteries sepa-
rators provide some margin of protection against 6.1.3. Separator Properties/Characterization
short circuit and overcharge in lithium-ion cells. The Separators are characterized by structural and
separators exhibit a large increase in impedance at functional properties; the former describes what they
temperature about 130 °C that effectively stops ionic are and the latter how they perform. The structural
transport between the electrodes.91,92 The greater the properties include chemical (molecular) and microc-
mechanical integrity of the separator above 130 °C, rystalline nature, thickness, pore size, pore size
the greater the margin of safety the separator can distribution, porosity, and various chemical and
provide. If the separator loses mechanical integrity, physical properties such as chemical stability, and
then the electrodes can come into direct contact, react electrolyte uptake. The functional properties of inter-
chemically, and result in thermal runaway. The est are electrical resistivity, permeability, and trans-
shutdown behavior of a separator can be character- port number. It is useful to characterize separator
ized by heating the separator (saturated with elec- materials in terms of their structural and functional
trolyte) to high temperatures and simultaneously properties and to establish a correlation of these
monitoring the electrical resistance of the separa- properties with their performance in batteries. A
tor.92,108 variety of techniques are used to evaluate separators.
6.1.2.15. High-Temperature Stability. A separa- Some of these techniques are discussed in this
tor might provide an extra margin of safety if it can section.
prevent the electrodes from contacting one another 6.1.3.1. Gurley. Separator permeability is usually
at high temperatures. Separators with good mechan- characterized by air permeability. The gurley number
ical integrity at high temperatures can provide a expresses the time required for a specific amount of
greater margin of safety for lithium-ion cells. Ther- air to pass through a specific area of separator under
mal mechanical analysis (TMA) can be used to a specific pressure. The standard test method is
characterize the high-temperature stability of sepa- described in ASTM-D726 (B).
rators. Utilizing TMA, the separator is held under The gurley number is used to characterize separa-
constant load and the degree of elongation vs. tem- tors because the measurement is accurate and easy
perature is measured; at the temperature where the to make, and deviations from specific values are a
separator loses mechanical integrity, the elongation good indication of problems. Air permeability (gurley)
increases dramatically. is proportional to electrical resistance (ER), for a
6.1.2.16. Electrode Interface. The separator given separator morphology.98 Gurley can be used in
should form a good interface with the electrodes to place of ER measurements once the relationship
provide sufficient electrolyte flow. between gurley and ER is established. A lower gurley
In addition to the above properties, the separator value means higher porosity, lower tortuosity, and
must be essentially free of any type of defects accordingly lower ER.
(pinholes, gels, wrinkles, contaminants, etc.). All of 6.1.3.2. Electrical Resistance. The measurement
the above properties have to be optimized before a of separator resistance is very important to the art
membrane qualifies as a separator for a lithium-ion of battery manufacture because of the influence the
separator has on electrical performance. Electrical Ω, A is the electrode area in cm2, and l is the
resistance is a more comprehensive measure of thickness of membrane in cm. Similarly, the specific
permeability then the gurley number, in that the resistivity of the electrolyte, Fe (Ω cm), is given by
measurement is carried out in the actual electrolyte
solution. The ionic resistivity of the porous membrane ReA
is essentially the resistivity of the electrolyte that is
Fe ) (2)
l
embedded in the pores of the separator. Typically, a
microporous separator, immersed in an electrolyte where Re is the measured resistance of electrolyte in
has an electrical resistivity about 6-7 times that of Ω. The ratio of the resistivity of a separator mem-
a comparable volume of electrolyte, which it dis- brane to that of the electrolyte is called the Mac-
places. It is a function of the membrane’s porosity Mullin number, Nm, which can be used to predict the
and tortuosity, the resistivity of the electrolyte, the influence of the separator on battery performance.97
thickness of the membrane, and the extent to which
the electrolyte wets the pores of the membrane.93 The Fs τ2
Nm ) ) (3)
electrical resistance of the separator is the true Fe
performance indicator of the cell. It describes a
predictable voltage loss within the cell during dis- where τ is the tortuosity and is the porosity of the
charge and allows one to estimate rate limitations. separator. The MacMullin number describes the
Classical techniques for measuring electrical re- relative contribution of a separator to cell resistance.
sistivity of microporous separators have been de- It is almost independent of electrolyte used and also
scribed by Falk and Salkind5 and by Robinson and factors out the thickness of the material. It assumes
Walker.94 The resistivity of an electrolyte is more that the separator wets completely in the electrolyte
accurately determined by ac methods since dc can used for the test. From eqs 1 and 3, the electrical
polarize the electrodes and cause electrolysis of the resistance of a microporous membrane is given by the
solution. Modern ac impedance measuring systems following5,114
( )
allow rapid measurements of cell resistance over a
wide range or frequencies from which resistance can τ2l
Rm ) Fe (4)
be calculated free of capacitance effects. Compared A
to the dc techniques, the equipment required and the
theory necessary to interpret the ac techniques are It has been shown for Celgard membranes that the
more complex; however, ac measurements yield membrane resistance can be related to the gurley
information about long-range migration of ions and number by98
polarization phenomena occurring within the cell. In
Fe
an ac measurement, a sinusoidal voltage is applied RmA ) tgurd (5)
to a cell, and the sinusoidal current passing through 5 × 18 × 10-3
the cell as a result of this perturbation is determined.
A four-electrode cell is usually used for resistivity where Rm is the membrane resistance (Ω), A is the
measurements. The outer two electrodes serve to membrane area (cm2), Fe is the specific electrolyte
apply a sinusoidal potential, and the resulting cur- resistance (Ω cm), tgur is the gurley number (10 cm3
rent passing through the inner two electrodes is air, 2.3 mmHg), d is the pore size, and 5 × 18 × 10-3
measured. This technique is employed to avoid the is a scaling factor.
complications arising from a nonuniform potential The usual procedure for characterizing battery
field near the outer two electrodes. An excellent separators is to cut several test samples from the
review of experimental techniques for measuring finished material. Thus, only a small portion of the
electrical resistivity in aqueous solution is avail- separator is actually examined. Ionov et al. has
able.95,96 proposed an alternative technique to measure the
The separator resistance is usually characterized resistance of a separator over a large separator
by cutting small pieces of separators from the fin- area.99 In this technique, the separator material is
ished material and then placing them between two passed through an electrolyte bath between electrical
blocking electrodes. The separators are completely resistance measuring transducers. The set of trans-
saturated with the electrolyte. The resistance (Ω) of ducers installed in the bath transverse to the moving
the separator is measured at a certain frequency by sheet of separator material examines the whole
ac impedance techniques. The frequency is chosen so surface of the material. If the production process
that the separator impedance is equal to the separa- ensures good uniformity in the physicochemical
tor resistance. To reduce the measurement error, it properties of the separator material over the whole
is best to do multiple measurements by adding extra surface, the transducer outputs will clearly be close
layers. The average resistance of single layer is to one another. A nonuniform separator will cause
determined from multiple measurements. The spe- significant deviations from the average value at
cific resistivity, Fs (Ω cm), of the separator saturated various sections of the material. In this case, the
with electrolyte is given by sections having lower or higher resistance compared
with the average value should be regarded as flawed.
RsA 6.1.3.3. Porosity. The porosity is important for
Fs ) (1) high permeability and also for providing a reservoir
l
of electrolyte in the cell. Higher and uniform porosity
where Rs is the measured resistance of separator in is desirable for unhindered ionic current flow. Non-
uniform porosity leads to nonuniform current density This feature will be increasingly important as battery
and can further lead to reduced activity of the manufacturers continue to increase the cell capacity
electrodes. Cell failure can result because during with thinner separators. The pore structure is usually
discharge some areas of the electrodes work harder influenced by polymer composition, and stretching
then other. conditions, such as drawing temperature, drawing
Porosity of a separator is defined as the ratio of speed, and draw ratio. In the wet process, the
void volume to apparent geometric volume. It is separators produced by the process of drawing after
usually calculated (eq 6) from the skeletal density, extraction (as claimed by Asahi Chemical and Mitsui
basis weight, and dimensions of the material and so Chemical) are found to have much larger pore size
may not reflect the accessible porosity of the material. (0.24-0.34 µm) and wider pore size distribution than
those produced by the process of extraction (0.1-0.13
porosity (%) ) µm) after drawing (as claimed by Tonen).58
[1-
(sample weight/sample volume)
polymer density ]
× 100 (6)
The testing of battery separators and control of
their pore characteristics are important requirements
for proper functioning of batteries. Mercury porosim-
The standard test method is described in ASTM etry has been historically used to characterize the
D-2873. The actual or accessible porosity can also be separators in terms of percentage porosity, mean pore
determined by the weight of liquid (e.g., hexadecane) size and pore size distribution.100 In this method, the
absorbed in the pores of the separator. In this size and volume of pores in a material are measured
method, the separator weight is measured before and by determining the quantity of mercury, which can
after dipping in hexadecane solvent, and the porosity be forced into the pores at increasing pressure.
is calculated (eq 7) by assuming that volume occupied Mercury does not wet most materials, and a force
by hexadecane is equal to the porous volume of the must be applied to overcome the surface tension
separator. forces opposing entry into the pores.
The hydrophobic (e.g. polyolefins) separators are
porosity (%) ) also characterized with Aquapore (non-mercury po-
volume occupied by hexadecane rosimetry) technique, where water is used in place
× 100 (7) of mercury. This is a very useful technique for
(volume of polymer + characterizing polyolefin-based separators used in
volume occupied by hexadecane) lithium batteries.101 Porosimetry gives pore volume,
surface area, mean pore diameter, and pore size
6.1.3.4. Tortuosity. Tortuosity is the ratio of mean distribution. In a typical experiment, the sample is
effective capillary length to separator thickness. The placed in the instrument and evacuated. As the
tortuosity factor, τ of a separator can be expressed pressure increase, the quantity of water forced into
by the pores increases in proportion to the differential
pore volume, the size of the pores corresponding to
ls the instantaneous pressure. Thus, increasing the
τ) (8)
d pressure on a membrane having a given pore size
distribution results in a unique volume vs pressure
where ls is the ion path through the separator and d or pore diameter curve. The pressure required for
is the thickness of the separating layer. intrusion of water in to a pore of diameter, D, is given
Tortuosity is a long-range property of a porous by following equation
medium, which qualitatively describes the average
pore conductivity of the solid. It is usual to define τ 4γ cos θ
by electrical conductivity measurements. With knowl- D) (9)
p
edge of the specific resistance of the electrolyte and
from a measurement of the sample membrane resis- where D is the diameter of the pore assuming the
tance, thickness, area, and porosity, the membrane pore to be cylindrical, p is the differential pressure,
tortuosity can be calculated from eq 3. γ is the surface tension of the nonwetting liquid,
This parameter is widely used to describe the ionic water, and θ is the contact angle of water. The pores
transport by providing information on the effect of generally are not of spherical shape of a constant
pore blockage. A tortuosity factor τ ) 1, therefore, diameter. They usually vary in their form and size.
describes an ideal porous body with cylindrical and Thus, statements of any pore diameter are always
parallel pores, whereas values of τ > 1 refer to more to be viewed with the above in mind.
hindered systems. Higher tortuosity is good for Another technique, capillary flow porometry has
dendrite resistance but can lead to higher separator been developed by Porous Materials Inc.102 to char-
resistance. acterize battery separators.103,104 The instrument can
6.1.3.5. Pore Size and Pore Size Distribution. measure a number of characteristics of battery
For any battery applications, the separator should separators such as size of the pore at its most
have uniform pore distribution to avoid performance constricted part, the largest pore size, pore size
losses arising from nonuniform current densities. The distribution, permeability, and envelope surface
submicrometer pore dimensions are critical for pre- area.109
venting internal shorts between the anode and the Scanning electron microscopy (SEM) is also used
cathode of the lithium-ion cell, particularly since to examine separator morphology. SEM pictures of
these separators tend to be as thin as 25 µm or less. some commercial membranes are shown in Figures
Figure 4. Scanning electron micrographs of the surface of single layer Celgard separators used in lithium batteries: (a)
2400 (PP), (b) 2500 (PP), and (c) 2730 (PE).
Figure 5. Scanning electron micrographs of Celgard 2325 (PP/PE/PP) separator used in lithium-ion batteries: (a) surface
SEM and (b) cross-section SEM.
4-6. The surface SEM of Celgard 2400, 2500, and Figure 5 shows the surface SEM and cross-section
2730 are shown in Figure 4. It is clear from the SEM of Celgard 2325. The surface SEM only shows
images that the pores are uniformly distributed. Both the PP pores while the PE pores are visible in the
Celgard 2400 and 2500 are single layer PP separa- cross-section. It is clear from the image that all three
tors, but the pore size of Celgard 2500 is substantially layers are of equal thickness. The SEM of separators
larger than Celgard 2400. Thus, it has lower resis- made by wet process are shown in Figure 6. The pore
tance and is more suited for high rate applications. structure of all of these membranes is very similar.
Figure 6. Scanning electron micrographs of separators made by wet process and used in lithium-ion batteries: (a) Setela
(Tonen), (b) Hipore-1 (Asahi), (c) Hipore-2 (Asahi), and (d) Teklon (Entek).
Asahi-1 (Figure 6b) separator has significantly larger available puncture strength machines made for tex-
pores compared to the other membranes. tiles tend to give meaningless results when testing
Image analysis has been used to characterize the battery separator membranes. More reproducible
pore structure of synthetic membrane materials.105 results can be obtained with a load frame (such as
The Celgard films have also been characterized by an Instron Machine). The mix penetration strength
scanning tunneling microscopy, atomic force micros- is a better measure of mechanical strength for battery
copy, and field emission scanning electron micros- separators as it measures the force required to create
copy.53,106 The pore size of the Celgard membranes a short through the separator when electrode mix is
can also be calculated from eq 5, once the MacMullin pushed through it.
number and gurley values are known. The strength of the separator depends greatly on
6.1.3.6. Puncture Strength. A separator is re- the materials used and the manufacturing method.
quired to have sufficient physical strength to endure The wet-biaxial method simultaneously stretches in
the rigors of cell assembly and day-to-day charge- the MD and TD directions and thus achieves a
discharge cycling. Physical strength is required to material that has tensile modulus and rupture
withstand basic handling, cell blocking/assembly, strength in both directions. Both high polymer en-
physical shock, punctures, abrasion, and compres- tanglement and stretching help increase the physical
sion. strength of the separator.
The puncture strength (PS) is the weight that must 6.1.3.7. Mix Penetration Strength. The force
be applied to a needle to force it completely through required to create a short through a separator due
a separator.45,107 It has been used to indicate the to mix (electrode material) penetration defines mix
tendency of separators to allow short-circuits in a cell penetration strength. In this test force (with a 1/2 in.
that may occur due to holes generated in the separa- diameter ball) is applied on the positive electrode/
tor by the rough surface of an electrode during the separator/negative electrode sandwich, and the force
battery assembly and charge-discharge cycle. The at which the mix penetrates through the separator
PS requirement for lithium-ion batteries is higher and creates an electronic short is called mix penetra-
them lithium-foil batteries, because the separator tion force. Mix penetration strength is used to
must contend with two rough surfaces. Commercially indicate the tendency of separators to allow short-
Figure 7. Internal impedance (at 1 kHz) of Celgard 2325 (PP/PE/PP) separator as a function of temperature. Heating
rate: 60 °C/min.
circuits during battery assembly. The mix penetra- ature and preventing venting in short-circuited cells.108
tion resistance test is more closely related to particle It usually takes place close to the melting tempera-
penetration resistance compared to puncture resis- ture of the polymer when the pores collapse turning
tance.49 the porous ionically conductive polymer film into a
6.1.3.8. Tensile Strength. The tensile strength nonporous insulating layer between the electrodes.
measurements (e.g., Young’s modulus, percent offset At this temperature there is a significant increase
strength, elongation at break, and stress at break) in cell impedance and passage of current through the
can be made by utilizing widely known standard cell is restricted. This prevents further electrochemi-
procedures. These tests are carried out in both MD cal activity in the cell, thereby shutting the cell down
and TD directions. The tensile properties are depend- before an explosion can occur.
ent on the manufacturing process. The uniaxially The ability of the PE based separator to shutdown
oriented films have high strength in only one direc- the battery is determined by its molecular weight,
tion, whereas biaxially oriented films are more percent crystallinity (density) and process history.
uniformly strong in both MD and TD directions. Material properties and processing methods might
ASTM test method D88-00, “Standard test method need to be tailored so that the shutdown response is
for tensile properties of thin plastic sheeting”, is an spontaneous and complete. The optimization needs
appropriate test. to be done without affecting the mechanical proper-
The separator should be strong enough to with- ties of the material in the temperature range of
stand mechanical handling during cell winding and interest. This is easier to do with the trilayer separa-
assembly. It should be dimensionally stable and tors manufactured by Celgard since one material is
should not neck down during winding. The decrease utilized for the shutdown response and another for
in width will allow the electrodes to touch each other the mechanical properties. Polyethylene containing
and create a short. Thus, the tensile property of the separators, in particular trilayer laminates of polypro-
separator should be very strong in MD direction pylene, polyethylene, and polypropylene, appear to
compared to TD direction. have the most attractive properties for preventing
6.1.3.9. Shrinkage. Shrinkage test is carried out thermal runaway in lithium-ion cells.109,110 The shut-
on both MD and TD directions. In this test, the down temperature of 130 °C is usually enough to
dimensions of separators are measured and then control the cell heating and avoid thermal runaway
stored at 90 °C for a fixed time. The shrinkage is then in lithium-ion cells. A lower temperature shutdown
calculated from the change in dimensions as shown will be desirable if it does not affect the separator
in eq 10. mechanical properties or high-temperature cell per-
formance in any adverse way.
Li - Lf
shrinkage (%) ) × 100 (10) The shutdown property of separators is measured
Li by measuring the impedance of a separator while the
temperature is linearly increased.92,108 Figure 7 shows
where Li is the initial length and Lf is the final length the actual measurement for Celgard 2325 membrane.
of separator after high temperature storage. The The heating rate was around 60 °C/min, and the
uniaxially stretched separators tend to shrink in the impedance was measured at 1 kHz. The rise in
MD direction only, while the biaxially stretched impedance corresponds to a collapse in pore structure
separators shrink in both MD and TD directions. The due to melting of the separator. A 1000-fold increase
shrinkage of separators can also be compared by in impedance is necessary for the separator to stop
carrying out the thermal mechanical analysis (TMA) thermal runaway in the battery. The drop in imped-
test at a constant load and rate. ance corresponds to opening of the separator due to
6.1.3.10. Shutdown. Separator shutdown is a coalescence of the polymer and/or to penetration of
useful and essential mechanism for limiting temper- the separator by the electrodes; this phenomenon is
Figure 8. DSC of Celgard 2730 (PE), 2400 (PP), and 2325 (PP/PE/PP).
referred to as a loss in “melt integrity”. This test is allowed temperature scan rates of 5 °C/min and
fairly reliable in indicating the temperature at which higher and obtained results similar to those of Laman
the impedance rises but shows some variability in et al.
characterizing the subsequent drop in impedance. Prior work related with shutdown separators also
In Figure 7, the shutdown behavior of a multilayer involved application of waxes on membranes.115,116 In
(PP/PE/PP) separator (Celgard 2325) is shown. The these cases, the wax or low melting polymers were
impedance rise occurred near the melting point of coated on the polyolefin separator. The disadvantage
polyethylene (130 °C) and remained high until such of this technique is that the coating can block the
time as the melting point of polypropylene (165 °C) pores of the separator and thus can affect the
is attained. The shutdown temperature of the sepa- performance by increasing separator resistance. More-
rator is governed by the melting point of the separa- over, the coating level has to be very high to get
tor material. At the melting point the pores in the complete shutdown.
separator collapse to the form a relatively nonporous The shutdown characteristic provides protection
film between the anode and the cathode. This was from external short-circuit and during cell over-
confirmed by DSC as shown in Figure 8. The DSC charge. It provides little protection from internal
scan in Figure 8 gives a peak melting temperature shorts should they occur. Should the electrodes touch
of 135 °C for Celgard 2730, 168 °C for Celgard 2400, each other or become shorted from a dendritic growth
and 135/165 °C for Celgard 2325. The shutdown of soluble impurity or other dendrite forming soluble
behavior of thinner separators (<20 µm) is very material, the separator only helps in avoiding delayed
similar to thicker separators. The battery manufac- failures. In the case of an instant failure during an
turers have been very successful in using the thinner internal short circuit, the heating rate is too high and
separators without compromising on the shutdown the separator shutdown is not fast enough to control
behavior of the separators. the heating rate.
Laman et al. introduced the use of impedance 6.1.3.11. Melt Integrity. The separators used in
measurements as a function of temperature to char- lithium-ion batteries should have high-temperature
acterize shutdown separators.92 Using a temperature melt integrity. The separator should maintain its
scan rate of 1 °C/min, they found that the impedance melt integrity after shutdown so that the electrodes
increased several orders of magnitude near the do not touch and create a short. This helps in
melting point of the separator. They verified the avoiding the thermal runaway even when the cell is
patent claims of Lundquist et al.111 that bilayer exposed to high temperatures. Thermal mechanical
separators of PE and PP gave a temperature window analysis (TMA) is a very good technique to measure
of high impedance extending approximately between the high-temperature melt integrity of separators.
the melting point of the polymers. The concept of TMA involves measuring the shape change of a
using separators consisting of distinct layers, one of separator under load while the temperature is lin-
which could act as a fuse, was developed by Lun- early increased. Typically, separators show some
dquist et al.112,113 Laman’s results have been cor- shrinkage and then start to elongate and finally
roborated by Geiger et al.40 and Spotnitz et al.108,114 break as shown in Figure 9. This test utilizes a small
Spotnitz et al. developed a thin layer cell which separator samples (about 5-10 mm length (MD) and
Figure 9. TMA of Celgard 2400 (PP) and 2325 (PP/PE/PP). A constant load (2 g) is applied while the temperature is
ramped at 5 °C/min.
Table 6. TMA Data for Typical Celgard Separators 6.1.3.12. Wettability and Wetting Speed. Two
Celgard 2400 Celgard 2325 physical properties of separators, which are impor-
shrinkage onset temp (°C) 121 106
tant to the operating characteristics of a battery, are
deformation temp (°C) 156 135, 154 electrolyte absorption and electrolyte retention. Any
rupture temp (°C) 183 192 good separator should be able to absorb a significant
amount of electrolyte and also retain the absorbed
electrolyte when the cell is in operation. These are
about 5 mm width), which is held in mini-instron type more important in sealed cells where no free elec-
grips. The sample is held with a constant 2 g load trolyte is present. A maximum amount of electrolyte
while the temperature is ramped at 5 °C/min past in the separator is desirable to achieve minimum cell
the melting point until the tension ruptures the film. internal resistance.
Three parameters are reported from TMA tests
shrinkage onset temperature, melt temperature, and The separator wettability can limit the perfor-
melt rupture temperature. It has proved to be a more mance of batteries by increasing the separator and
reproducible measure of melt integrity of the separa- cell resistance. Separator wetting speed can be cor-
tor.108 related with electrolyte filling time in real cells. The
wetting speed is determined by the type of polymer
Figure 9 shows the TMA data for two different (surface energy), pore size, porosity, and tortuosity
Celgard membranes. The shrinkage onset tempera- of the separators. There is no generally accepted test
ture, deformation temperature, and rupture temper- for separator wettability. However, simply placing a
ature are summarized in Table 6. The single layer drop of electrolyte on the separator and observing
PP membrane (Celgard 2400) showed a higher soft- whether the droplet quickly wicks into the separator
ening temperature (∼121 °C), a deformation temper- is a good indication of wettability. The contact angle
ature around 160 °C, and a very high rupture is also a good measure of wettability.
temperature around 180 °C. The multilayer polypro-
pylene/polyethylene/polypropylene separator (Cel- The uptake of electrolyte by many hydrophobic
gard 2325) combined the low-temperature shutdown polymer separators can be enhanced either by wet-
property of polyethylene with the high-temperature ting agents or ionic-functional groups (e.g. ion-
melt integrity of polypropylene, resulting in a sepa- exchange membranes).
rator with softening (∼105 °C) and melt temperature
(∼135 °C) very similar to that of PE and rupture
6.1.4. Effect of Separator on Cell Performance and Safety
temperature (∼190 °C) very similar to that of PP. Although the material of a battery separator is
Separators with melt integrity greater than 150 °C inert and does not influence electrical energy storage
are desirable for lithium-ion cells. The trilayer sepa- or output, its physical properties greatly influence the
rators with polypropylene on the outside help in performance and safety of the battery. This is espe-
maintaining the melt integrity of the separators at cially true for lithium-ion cells, and thus the battery
higher temperatures compared to single layer PE manufacturers have started paying more attention
separators. This is especially important for bigger to separators while designing the cells. The cells are
lithium-ion cells being developed for hybrid and designed in such a way that separators do not limit
electric vehicles. the performance, but if the separator properties are
Table 7. Safety and Performance Tests for Lithium-Ion Batteries and the Corresponding Important Separator
Property and Its Effect on the Cell Performance and/or Safety
cell property separator property comments
cell capacity thickness cell capacity can be increased by making the separator thinner
cell internal resistance resistance separator resistance is a function of thickness, pore size, porosity,
and tortuosity
high rate performance resistance separator resistance is a function of thickness, pore size, porosity,
and tortuosity
fast charging resistance low separator resistance will aid in overall faster charging by
allowing higher and/or longer constant current charging
high-temp storage oxidation resistance oxidation of separators can lead to poor storage performance and
reduce performance life
high-temp cycling oxidation resistance oxidation of separators can lead to poor cycling performance
self-discharge weak areas, pinholes soft shorts during cell formation and testing can lead to
internal current leakage
long-term cycling resistance, shrinkage, pore size high resistance, high shrinkage and very small pore size can
lead to poor cycling performance
overcharge shutdown behavior; separator should completely shutdown and then maintain its
high-temp melt integrity melt integrity at high temp
external short circuit shutdown behavior separator shutdown stops the cells from overheating
hotbox high-temp melt integrity separator should be able to keep the two electrodes apart at
high temp
nail crush shutdown (to stop delayed failure) in the case of internal shorts, the separator may be the only
safety device to stop the cell from overheating
bar crush shutdown (to stop delayed failure) in the case of internal shorts, the separator may be the only
safety device to stop the cell from overheating
not uniform, or if there are other issues, it can affect trolyte and their mechanical strength is not as high
the performance and safety of cells. This section will as thicker separators. Thus, appropriate changes
focus on the effect of the separator properties on cell should be made in cell design to keep the cell safe.
performance and safety. Table 7 shows different The handling and manufacturing of thinner separa-
types of safety and performance tests for lithium-ion tors is also a challenge for the separator manufactur-
batteries and the corresponding important separator ers. They are required to maintain the same electrical
property and how it affects performance and/or and mechanical properties and to have better quality
safety. for thinner separators. The separator manufacturers
To achieve good performance of lithium-ion cells, have installed better controls and quality standards
the separators should have low resistance, low shrink- and have started offering 16-µm separators. A lot of
age and uniform pore structure. The separator with battery experts are of the opinion that the 16-µm is
high resistance will perform poorly during high rate the thinnest they can use and still maintain the
discharge and will also increase the cell charging stringent performance and safety requirements of
time. Low shrinkage is a very important character- lithium-ion cells.
istic for separators, especially for higher capacity The separators inside the lithium-ion batteries
cells. These cells are used in high-speed laptop experience extreme oxidizing environment on the
computers, which can experience higher tempera- side facing the positive electrode and extreme re-
tures (∼70-75 °C) under certain conditions.117 This ducing environment on the side facing the negative
can lead to shrinkage of separators and ultimately electrode. The separators should be stable in these
higher cell resistance and poor long-term cycling. The conditions during long-term cycling especially at
shrinkage in TD direction can lead to safety issues high temperatures. Separators with poor oxidation
because of an internal short between the electrodes. resistance can lead to poor high-temperature stor-
Larger pores can lead to shorts during cell manufac- age performance and poor long-term cycling be-
turing or can fail during hipot testing. Larger pores havior. The oxidation resistance properties of trilayer
will allow more soft shorts and higher self-discharge, (PP/PE/PP) separators with PP as the outside layer
especially during high-temperature storage. Very and PE as inner layer is superior compared to
small pore size can lead to higher resistance and polyethylene separators. This is because of the better
poor cycle life during high-temperature cycling and oxidation resistance properties of polypropylene in
storage. Thus, the pore size of the separator should contact with the positive electrode in a lithium-ion
be optimized to achieve good strength and perfor- cell.
mance. The products formed by the decomposition of the
One of the ways to increase cell capacity is by electrolyte can also block the pores of the separator,
decreasing the thickness of separators. The newer leading to increase in cell resistance. The separators
high capacity cells (>2.0 A h) generally use 20 and with lower resistance also helps in better low tem-
16 µm separators as compared to 25 µm separators perature performance. At very low temperatures, the
used in cells with 1.6-1.8 A h capacity. The thinner resistance of the electrolytes is very high and thus
separators offer lower resistance and help in increas- smaller contribution from separator helps in keeping
ing the capacity. However, they can hold less elec- the cell resistance lower.
Zeng et al.118 has shown that small amounts of temperature, which results from cell abuse (e.g.,
active lithium metal could be added to a lithium-ion short circuit, overcharge). The separator should not
battery via the separator by using vacuum deposition only shutdown around 130 °C, but it should also
techniques. The lithium films (4-8 µm) were depos- maintain its mechanical integrity at higher temper-
ited onto a microporous PP film and showed that the atures, preferably at temperatures. If the separator
lithium electrochemically reacted with either elec- does not shutdown properly then the cell will con-
trode and thus the intrinsic irrereversible capacity tinue to heat during an overcharge test and can lead
of the negative electrode was compensated for using to thermal runaway. The high-temperature melt
volumetrically efficient lithium metal. This may be integrity of separators is also a very important
a novel idea to allow higher capacity designs but is property to keep the cell safe during extended
likely impractical and uneconomical due to issues overcharge or during extended exposure to higher
involved with lithium plating on polymer films and temperatures.
handling the resulting films.
Figure 10 shows a typical short-circuit curve for
The lithium-ion cells have demonstrated power loss an 18650 lithium-ion cell with shutdown separator,
when aged and/or cycled at high temperatures. Norin LiCoO2 positive electrode, and MCMB carbon nega-
et al.119 demonstrated that the separator is at least tive electrode. The cell does not have other safety
partly responsible for the power loss due to the devices (e.g., CID, PTC), which usually work before
intrinsic increase in its ionic resistance. They showed separator shutdown. As soon as the cell is short-
that impedance increased significantly upon cycling circuited externally through a very small shunt
and/or aging of lithium-ion cells at elevated temper- resistor, the cell starts heating because of the large
atures and that separators accounts for ∼15% of the current drained through the cell. The shutdown of
total cell impedance rise. They later reported that the the separator, which occurs around 130 °C, stops the
loss in ionic conductivity of the separator was due to cell from heating further. The current decrease is
blocking of the separator pores with the products caused by increase of battery internal resistance due
formed due to electrolyte decomposition, which was
to separator shutdown. The separator shutdown
significantly accelerated at elevated temperatures.120
helps in avoiding the thermal runaway of the cell.
The U.S. Department of Transportation (DOT)
classifies all lithium-ion batteries as hazardous ma- Cells can be overcharged when the cell voltage is
terials for shipping in the same category as lithium incorrectly detected by the charging control system
metal primary batteries.121 It grants exceptions based or when the charger breaks down. When this hap-
on the cell capacity and ability of the cells to pass pens, the lithium ions remaining in the cathode are
specified tests. There are several groups that regu- removed and more lithium ions are inserted into the
late, or provide testing, to verify safe operation of anode then under standard charging conditions. If
lithium-ion cells under abuse conditions. In addition, the lithium insertion ability of the carbon anode is
the UL Laboratories,122,123 the International Electro- small, lithium metal in the form of dendrites may be
technic Commission,124 and the United Nations125 deposited on the carbon and this causes a drastic
have developed standardized safety testing proce- reduction in thermal stability. At higher charging
dures. These tests are designed to ensure that cells rates, the heat output increases greatly because the
are safe to ship and are resistant to typical abuse joule heat output is proportional to I2R. Several
conditions such as internal shorting, overcharge, exothermic reactions (e.g., reaction between lithium
overdischarge, vibration, shock, and temperature and electrolyte, thermal decomposition of anode and
variations that may be encountered in normal trans- cathode, thermal decomposition of electrolyte, etc.)
portation environments. occur inside the cell as its temperature increases.
Underwriters Laboratories (UL) requires that con- Separator shutdown happens when the cell temper-
sumer batteries pass a number of safety tests (UL ature reaches the melting point of polyethylene as
1642126 and UL-2054127). There are similar recom- shown in Figure 11. The CID and PTC of the 18650
mendations from UN for transport of dangerous cells was removed to test the performance of separa-
goods,128 and from the International Electrotechnical tor alone. The current decrease is caused by increase
Commission (IEC) and Japan Battery Association. 129 of battery internal resistance due to separator shut-
An abnormal increase in cell temperature can occur down. Once the pores of the separator have closed
from internal heating caused by either electrical due to softening, the battery cannot continue to be
abusesovercharge or short circuitsor mechanical charged or discharged, and thus thermal runaway
abusesnail penetration or crush. A higher cell tem- is prevented. During continued overcharge, the sepa-
perature could also be a result of external heating. rator should maintain its shutdown feature and
For this reason, battery packs containing lithium- should not allow the cell to heat again. It should also
ion cells are designed with safety control circuits that maintain its melt integrity and should not allow the
have redundant safety features (PTC, CID, vent, two electrodes to touch each other.
thermal fuse, etc.). Shutdown separators are one of The separator should also not allow any dendrite
the safety devices inside the cell and act as a last to penetrate through the separator to avoid internal
line of defense. The separator shutdown is irrevers- shorts. During an internal short, the separator is the
ible, which is fine for polyethylene-based separators, only safety device that can stop the thermal runaway,
which melt around 130 °C. if the failure is not instant. If the heating rate is too
The impedance of the separator increases by 2-3 high, then instant failure will occur which cannot be
orders of magnitude due to an increase in cell stopped by separator shutdown. If the heating rate
Figure 10. Typical short-circuit behavior of a 18650 lithium-ion cell with shutdown separator and without PTC (positive
temperature coefficient) and CID (current interrupt device). This test simulates an external short circuit of a cell.
Figure 11. Typical overcharge behavior of a 18650 lithium-ion cell with shutdown separator. The PTC (positive temperature
coefficient) and CID (current interrupt device) were removed from the cell header.
is not too high then the separator shutdown can help duction. Thermal runaway is likely to take place as
in controlling the heating rate and stop thermal local heat generation induces electrolyte and elec-
runaway. trode materials to decompose. On the other hand, if
Generally in a nail penetration test, an instanta- the battery is fully penetrated, the increased contact
neous internal short would result the moment the area would lower the current density, and conse-
nail is tucked into the battery. Enormous heat is quently all tests would pass the nail penetration test.
produced from current flow (double layer discharge Internal short-circuit tests are more difficult to pass
and electrochemical reactions) in the circuit by the then the external short-circuit tests described earlier,
metal nail and electrodes. Contact area varies ac- because contact area between metal nail contact is
cording to depth of penetration. The shallower the smaller than the contact area between current col-
depth, the smaller the contact area and therefore lectors, where the current density would therefore be
the greater the local current density and heat pro- larger.
Figure 12. Typical nail penetration behavior of a 18650 lithium-ion cell with shutdown separator. This test simulates
internal short circuit of a cell. Key: (a) cell passed nail penetration test; (b) cell failed nail penetration test.
Figure 12 shows the typical nail penetration be- demonstrated a meltdown effect. Trilayer separators
havior of a 18650 lithium-ion cell with shutdown had meltdown temperatures as high as 160 °C
separator, LiCoO2 positive electrode, and MCMB because of the presence of additional layers of higher
carbon negative electrode. Clearly, there was a volt- melting PP. It was found that the separators are not
age drop from 4.2 to 0.0 V, instantaneously, as the able to shutdown the cell completely. In the case of
nail penetrates through (when an internal short an overcharged test, the cell could continue to charge
circuit occurs), and the temperature rose. When the at lower currents even after the shutdown event,
heating rate is low, the cell stops heating when the rendering the cell a potential hazard if not disposed
temperature is close to separator shutdown temper- immediately and safely. This usually does not become
ature as shown in Figure 12a. If the heating rate is an issue in commercial cells because the cell manu-
very high, then the cell continues to heat and fails facturers have addressed this issue by including
the nail penetration test as shown in Figure 12b. In multiple cutoff devices within a single cell.
this case, the separator shutdown is not fast enough Development efforts are under way to displace the
to stop cell from thermal runaway. Thus, the separa- use of microporous membranes as battery separators
tor only helps in avoiding delayed failures in the case and instead use gel electrolytes or polymer electro-
of internal short circuit as simulated by nail and bar lytes. Polymer electrolytes, in particular, promise
crush tests. Separators with high-temperature melt enhanced safety by eliminating organic volatile sol-
integrity and good shutdown feature (to avoid de- vents. The next two sections are devoted to solid
layed failures) are needed to pass the internal short- polymer and gel polymer type lithium-ion cells with
circuit test. Thinner separators (<20 µm) used in focus on their separator/electrolyte requirements.
high capacity cells should offer shutdown and high-
temperature melt integrity properties similar to those 6.2. Lithium Polymer
of thicker separators. The decrease in separator
strength should be balanced with changes in cell Because of their high theoretical capacity, lithium
design. The separator properties across the length polymer batteries have long been identified as a very
and width should be very uniform to keep the cell promising technology to meet the requirements of
safe during abnormal use. upcoming applications such as standby power and
The mechanism and characteristics of thermal electric vehicles. Research and development of poly-
cutoff devices in several prismatic lithium-ion cells mer electrolytes for ambient-temperature recharge-
was studied by Venugopal130 by monitoring the able lithium batteries has always been very active.
impedance at 1 kHz and the open circuit voltage Rapid progress for the past 2 decades in this field
(OCV) of the cells as a function of temperature. All has led to numerous monographs and reviews.131-137
the cells studied contained PE-based separators with These polymers are generally polyethers, poly-
a shutdown temperature between 130 and 135 °C. (ethylene oxide) (PEO), or poly(propylene oxide)
Within this narrow temperature range, the shutdown (PPO).
separators caused a sharp and irreversible rise in Solid polymer electrolytes serve two principal roles
impedance of the cell. Single layer PE separators in rechargeable lithium batteries. Not only do they
were effective up to around 145 °C, above which they function as the traditional electrolyte, i.e., the me-
dium for ionic transport but also they function as the swollen with a liquid electrolyte. The gel polymer
separator which insulates the cathode from the electrolyte approach to the lithium-ion technology
anode. Consequently, the polymer electrolyte must combines the positive attributes of both the liquid
have sufficient mechanical integrity to withstand (high ionic conductivity) and solid polymer electro-
electrode stack pressure and stresses caused by lytes (elimination of leakage problems).
dimensional changes, which the rechargeable elec- Gel polymer lithium-ion batteries replace the con-
trodes undergo during charge/discharge cycling. ventional liquid electrolytes with an advanced poly-
Lithium polymer electrolytes formed by dissolving mer electrolyte membrane. These cells can be packed
a lithium salt LiX (where X is preferably a large soft in lightweight plastic packages as they do not have
anion) in poly(ethylene oxide) PEO can find useful any free electrolytes and they can be fabricated in
application as separators in lithium rechargeable any desired shape and size. They are now increas-
polymer batteries.138-140 Thin films must be used due ingly becoming an alternative to liquid-electrolyte
to the relatively high ionic resistivity of these poly- lithium-ion batteries, and several battery manufac-
mers. For example, the lithium-ion conductivity of turers, such as Sanyo, Sony, and Panasonic have
PEO-Li salt complexes at 100 °C is still only about started commercial production.147,148 Song et al.149
1
/100th the conductivity of a typical aqueous solution. have recently reviewed the present state of gel-type
A polymer electrolyte with acceptable conductivity, polymer electrolyte technology for lithium-ion bat-
mechanical properties and electrochemical stability teries. They focused on four plasticized systems,
has yet to be developed and commercialized on a which have received particular attention from a
large scale. The main issues which are still to be practical viewpoint, i.e., poly(ethylene oxide) (PEO),
resolved for a completely successful operation of these poly(acrylonitrile) (PAN),150 poly(methyl methacry-
materials are the reactivity of their interface with late) (PMMA),151,152 and poly(vinylidene fluoride)
the lithium metal electrode and the decay of their (PVdF) based electrolytes.153-157
conductivity at temperatures below 70 °C. Croce et
al. found an effective approach for reaching both of One particular version of the lithium-ion gel poly-
these goals by dispersing low particle size ceramic mer cells, also known as plastic lithium-ion cell
powders in the polymer electrolyte bulk.141,142 They (PLION), was developed by Bellcore (now Telcordia
claimed that this new “nanocomposite polymer elec- Technologies).158-160 In this case, Gozdz et al. devel-
trolytes” had a very stable lithium electrode interface oped a microporous plasticized PVdF-HFP based
and an enhanced ionic conductivity at low tempera- polymer electrolyte which served both as separator
ture, combined with good mechanical properties. Fan and electrolyte. In PLION cells, the anode and
et al.143 has also developed a new type of composite cathode are laminated onto either side of the gellable
electrolyte by dispersing fumed silica into low to membrane. Good adhesion between the electrodes
moderate molecular weight PEO. and the membranes is possible because all three
The gel type polymer electrolyte prepared by sheets contain significant amounts of a PVdF copoly-
dispersing ceramic powders (e.g., Al2O3) into a matrix mer that can be melted and bonded during the
formed by a lithium salt solution contained in a poly- lamination step.
(acrylonitrile) (PAN) network was reported by Ap- The PVdF-HFP separators used in PLION cells
petecchi et al.144 These new types of composite gel were around 3 mil thick, and had poor mechanical
electrolytes had high ionic conductivity, wide elec- properties. It has been reported that the major source
trochemical stability, and particularly, high chemical of rate limitation in PLION cells was the separator
integrity even at temperatures above ambient. Kim thickness.161 The rate capability of these cells can be
et al.145 used a blend of PVdF-HFP and PAN as a significantly improved by decreasing the separator
matrix polymer to attain high ionic conductivity and thickness to that typically used in liquid electrolyte
good mechanical strength. The PAN can give me- system. Moreover, in the absence of shutdown func-
chanical integrity and structural rigidity to a porous tion, the separator does not contribute to cell safety
membrane without inorganic fillers. The high ionic in any way. Park et al. reported that the HFP content
conductivity was due to the high volume of pores and in separators did not have any significant impact on
a high affinity of the membrane for electrolyte cell performance.162 The Bellcore process has proven
solution.146 to be an elegant laboratory process but is difficult to
implement in large-scale production.
6.3. Lithium-Ion Gel Polymer To overcome the poor mechanical properties of
The solid polymer electrolyte approach provides polymer and gel polymer type electrolytes, mi-
enhanced safety, but the poor ambient temperature croporous membranes impregnated with gel polymer
conductivity excludes their use for battery applica- electrolytes, such as PVdF, PVdF-HFP, and other
tions, which require good ambient temperature per- gelling agents, have been developed as an electrolyte
formance. In contrast, the liquid lithium-ion tech- material for lithium batteries.163-173 Gel coated and/
nology provides better performance over a wider or gel-filled separators have some characteristics that
temperature range, but electrolyte leakage remains may be harder to achieve in the separator-free gel
a constant risk. Midway between the solid polymer electrolytes. For example, they can offer much better
electrolyte and the liquid electrolyte is the “hybrid protection against internal shorts when compared to
polymer” electrolyte concept leading to the so-called gel electrolytes and can therefore help in reducing
gel polymer lithium-ion batteries. Gel electrolyte is the overall thickness of the electrolyte layer. In
a two-component system, viz., a polymer matrix addition the ability of some separators to shutdown
Figure 13. Internal impedance (at 1 kHz) of PVdF coated Celgard trilayer separators as a function of temperature. Heating
rate: 60 °C/min.
at a particular temperature allows safe deactivation layer directly applied either to the electrode or the
of the cell under overcharge conditions. separator surfaces as an effective ion-conductive
The shutdown behavior of PVdF coated Celgard adhesive. Sanyo181,182 investigators, on the other
trilayer membranes is shown in Figure 13. The hand, used thermally polyerizable additives to gel,
shutdown is defined by the sharp increase in resis- or solidify, liquid electrolyte solutions in a wound,
tance around 130 °C. The PVdF coating should be packaged battery.
porous and should not block the pores to maintain The ceramic fillers (e.g. Al2O3, SiO2, TiO2) can
similar ionic conductivity. The scanning electron greatly influence the characteristics and properties
micrographs of the PVdF coated membrane is shown of polymer electrolyte by enhancing the mechanical
in Figure 14. The cross section SEM of Celgard 3300 stability and the conductivity.141,183-186 γ-LiAlO2,
provides visual evidence that the coating is porous Al2O3, and MgO were used as fillers by Prosini et
and is not blocking the pores of the top PP layer. al.187 in a PVdF-HFP polymer matrix to form self-
Abraham et al.174 were the first ones to propose standing, intrinsically porous separators for lithium-
saturating commercially available microporous poly- ion batteries. The MgO based separators showed the
olefin separators (e.g., Celgard) with a solution of best anode and cathode compatibilities.
lithium salt in a photopolymerizable monomer and
a nonvolatile electrolyte solvent. The resulting bat- Liu et al.188 has successfully prepared a PVdF-
teries exhibited a low discharge rate capability due HFP/PE composite gel electrolyte by cast method.
to the significant occlusion of the pores with the They showed that when the PE content was over 23
polymer binder and the low ionic conductivity of this wt %, the electrical impedance of the composite gel
plasticized electrolyte system. Dasgupta and Ja- electrolyte increased rapidly by several orders of
cobs163,175 patented several variants of the process for magnitude, around the melting point of PE. The SEM
the fabrication of bonded-electrode lithium-ion bat- pictures showed that the PE particles were fused and
teries, in which a microporous separator and elec- formed into a continuous film at or near the PE
trode were coated with a liquid electrolyte solution, melting point, which cuts off the ion diffusion. This
such as ethylene-propylenediene (EPDM) copolymer, shutdown feature of the composite gel electrolyte can
and then bonded under elevated temperature and help in preventing the cell runaway under abusive
pressure conditions. This method required that the usage. Similarly, poly(ethylene oxide) (PEO) coated
whole cell assembling process be carried out under separators were prepared by Kim et al.189 by coating
scrupulously anhydrous conditions, which made it PEO onto a microporous PE separators. The ionic
very difficult and expensive. conductivity of PEO coated membranes was higher
More recent methods proposed by Motorola176,177 then the base film. Kim et al. prepared the polymer
and Mitsubishi Electric178 researchers differ in imple- electrolytes by coating poly(ethylene oxide) (PEO) and
mentation details, but they share a common feature poly(ethylene glycol dimethacrylate) (PEGDMA) onto
in that a separate adhesive layer (PVdF) is applied a microporous polyethylene membrane (Asahi Kasei,
to the separator and used to bond the electrode and 25 um, 40% porosity).190 They showed that the
the separator films, using in the first case the hot, relative weight ratio of PEO and PEGDMA coated
liquid electrolyte as an in situ PVdF plasticizer. on to the microporous membrane played a critical role
Recently, Sony179,180 researchers described the use of in determining the uptake of electrolyte solution and
a thin, liquid electrolyte-plasticized polyacrylonitrile ionic conductivity.
Figure 14. Scanning electron micrographs of Celgard PVdF coated separators used in lithium gel polymer batteries: (a)
surface SEM, (b) cross-section SEM of coated trilayer, and (c) cross section of PVdF coating.
6.4. Lithium Primary Systems size batteries, three chemistries are currently avail-
able for consumer applications: Li-MnO2, Li-CFx,
Lithium primary batteries, with their outstanding and Li-FeS2. Li-MnO2 and Li-CFx cells are 3 V
performance and characteristics, are being used in cells that come in special sizes and are typically used
increasing quantities in a variety of applications, in cameras, safety devices, industrial applications,
including cameras, memory backup circuits, security etc. Li-FeS2 is a 1.5 V battery that is a drop-in
devices, calculators, watches, etc. They were first replacement for primary alkaline and C/Zn batteries.
used in early 1970s in selected military applications,
As the separator requirements for most of the
but since then they have been designed using a
above batteries are very similar, they will not be dealt
number of different chemistries, in a variety of sizes
with in detail. In this section, we will generally
and configurations, and have been used in different
describe the separator requirements followed by a
consumer applications. They are attractive because
brief discussion on few selected systems.
of their high voltage, high specific energy, good power
density, flat discharge characteristic, good low tem- 6.4.1. Separator Requirements
perature performance, and superior shelf life.
The term “lithium battery”, applies to many dif- The requirements for the separator properties are
ferent types of chemistries, each using lithium as the generally lower in lithium primary cells compared to
anode but differing in cathode material, electrolyte, lithium secondary cells. The lack of a charging
and chemistry as well as in design and other physical process helps avoid undesirable electrochemical de-
and mechanical features. Lithium primary batteries posits (e.g., dendrites) as well as the generation of
can be classified into several categories, based on the oxidizing substances. The mechanical strength re-
type of electrolyte (or solvent) and cathode material quirements are less severe because the negative
usedssoluble cathode cells (e.g., SO2, SOCl2, SO2Cl2), electrode is made of lithium foil, which is relatively
solid cathode cells (e.g., V2O5, MnO2, (CF)n, FeSx, smooth. Further, the shutdown behavior of separator
CuO), and solid electrolyte cells (I2 (P2VP)). The Li- is not mandatory because the thermal runaway
I2 system is unique as the cell reaction product, LiI, occurs only when the cell temperatures goes beyond
forms the separator in situ. In the large cylindrical 180 °C (melting point of lithium). The basic require-
Table 8. Separators and Their Manufacturers for Major Primary Lithium Battery Systems
battery system type of separator compositions (manufacturer)
Li/SO2 microporous glass fiber; polypropylene;b Tefzel
a
Li/SOCl2 microporous glass fiber;c glass mat or filter paper

Li/MnO2 microporous polypropylened (Celgard); polyethylene (Asahi Kasei)
Li/(CF)n microporous polypropylenee
Li/CuO microporous nonwoven glass
Li/V6O13 microporous polypropylene (Celgard 2400)f
solid-ion conductor poly(ethylene oxide)-lithium saltg
Li/V2O5 microporous polypropyleneh
Li/FeS2 microporous polypropylenei
Li/CuCl2 microporous Tefzel
solid electrolyte (Li-I2 (P2VP)) solid ion conductor LiI film (self-healing)j
a
Walk, C. R. In Lithium Batteries; Gabano, J. P., Ed.; Academic Press: New York, 1983, p 281. b Linden, D.; McDonald, B. J.
Power Sources 1980, 5, 35. c Klinedinst, K. A. J. Electrochem. Soc. 1984, 131, 492. d Moses, P. R.; Turchan, M. J.; Taylor, A. H.;
Mank, R. M. In Lithium Batteries; Gabano, J. P., Ed.; Academic Press: New York, 1983, p 333. e Morita, A.; Iijima, T.; Fujii, T.;
Ogawa, H. J. Power Sources 1980, 5, 111. f Abraham, K. M.; Goldman, J. L.; Holleck, G. L. Lithium Batteries; Venkatasetty, H.
V., Ed.; The Electrochemical Society, Inc.: Pennington, NJ, 1981; p 271. g Armand, M. B. Second International Meeting on Lithium
Batteries, Extended Abstracts No. 2, Paris, France. Solid State Ionics 1983, 9/10, 745. h Walk, C. R. In Lithium Batteries; Gabano,
J. P., Ed.; Academic Press: New York, 1983, p 265. i Uetani, Y.; Yokoyama, K.; Okamoto, O. J. Power Sources 1980, 5, 89. j Liang,
C. C.; Holmes, C. F. J. Power Sources 1980, 5, 3.
ments are very genericslow electrical resistance, designed for high-current pulse applications as well
thin, high strength, low shrinkage, relatively small as continuous high-rate operation.
pore size, and no defects. As is true for separators The lithium MnO2 system has been preferred over
for use in lithium-ion batteries, all the properties of lithium CFx for the primary lithium cell market for
separators for primary lithium batteries should be its lower cost and essentially equivalent performance
very uniform. for camera (cylindrical cells), watch (coin cells), and
The most common separator by far used in these memory applications. Dan et al. published results for
cells is the single layer PP separator made by Celgard an AA size rechargeable size cell based on Li/MnO2
(Celgard 2400) as shown in Table 8. chemistry with energy density of 125-140 W h/kg
and 280-315 W h/L. The cell used a polypropylene
6.4.2. Chemistries separator and was able to pass all safety tests (short
circuit, overcharge, deep discharge, crush, and nail
The separators used in major lithium primary penetration).191 Ikeda192 and co-workers reported the
systems are listed in Table 8. The majority of the development of “long life and high reliability Li/MnO2
lithium primary systems shown in Table 8 use cell for back up” and demonstrated how the mi-
microporous membranes (single layer PP or PE) croporous structure of Celgard flat sheet membranes
as separators. Some of the systems are discussed reduced self-discharge to provide reliable operation
below. for up to 10 years.
6.4.2.1. Lithium MnO2. The Li/MnO2 battery is 6.4.2.2. Lithium CFx. The Li/CFx battery consists
the most widely used primary lithium battery. It is of a lithium anode, polycarbon monofluoride cathode,
used in a wide variety of applications such as long- and microporous polypropylene separator saturated
term memory backup, safety and security devices, with organic electrolyte. These batteries are used as
cameras, many consumer devices, and military elec- power sources for watches, portable calculators,
tronics. It uses a lithium anode, microporous polypro- memory applications, and so on.
pylene separator impregnated with electrolyte, and 6.4.2.3. Lithium Iron Sulfide. Energizer has
MnO2 cathode. It is available in many configurations marketed Li-FeS2 (L91) AA-size cells since 1990.193
(including coin, bobbin, spirally wound cylindrical, These cells usually have a high surface area jellyroll
and prismatic configurations in multicell batteries) construction, using pyrite (FeS2) coated onto alumi-
and in designs for low, moderate, and moderately num foil as the cathode, a polyolefin (polypropylene)
high drain application. microporous separator, and a lithium foil anode. This
The Li/MnO2 coin cells use a lithium anode disk, system is particularly well suited for high drain and
which is separated from the manganese dioxide pellet constant power applications, where its higher operat-
cathode by a nonwoven polypropylene separator, ing voltage extends service time.
impregnated with the electrolyte. The bobbin-type 6.4.2.4. Lithium SO2. The lithium SO2 systems are
design contains a central lithium anode core sur- mainly used in military and some industrial and
rounded by the manganese oxide cathode, separated space applications. This system is particularly known
by a polypropylene separator impregnated with the for its capability to handle high current and high
electrolyte. The bobbin design maximizes the energy power requirements, for its excellent low-tempera-
density due to the use of thick electrodes and the ture performance, and for its long shelf life. They are
maximum amount of active materials. In the spirally typically fabricated in cylindrical structure by spi-
wound cells, the lithium anode and the cathode are rally winding rectangular strips of lithium foil, a
wound together with a microporous polypropylene microporous polypropylene separator, the cathode
separator to form the jelly roll. These cells are electrode, and a second separator layer.
6.4.2.5. Lithium SOCl2. The lithium SOCl2 bat- Martino et al. have demonstrated the use of BN
tery has one of the highest cell voltages (∼3.6 V) and felt separators in engineering tests.195 They have high
energy densities of the practical battery systems. porosity (∼90%), and hence, low ionic resistance, in
These batteries have been fabricated in a variety of addition to excellent compatibility with other cell
sizes and designs, ranging from wafer or coin cells materials at the operating temperature of 470 °C.
with capacities as low as 400 mA h to cylindrical cells However, this separator is too expensive and has poor
in bobbin and spirally wound electrode structures mechanical properties and so cannot prevent elec-
and to large 10 000 A h prismatic cells. Li/SOCl2 trode shape change during cell operation.196
bobbin batteries are manufactured in a cylindrical MgO, Y2O3, and AIN powder-based separators are
configuration. The anode is made of lithium foil, substantially less expensive.197,198 Cell assembly with
which is swaged against the inner wall of the can, such powder separators is simple and amenable to
the separator is made of nonwoven glass fibers, and mass production. Although powder separators were
the cathode is made of Teflon-bonded acetylene black. shown to maintain their integrity and conform to
6.4.2.6. Lithium TiS2. In 1990, W. R Grace & Co small dimensional changes of electrodes, their poros-
developed rechargeable Li-TiS2 cells with high en- ity (∼50%) is considerably lower than that of BN felt
ergy density, and excellent cycle life. To improve the and this drawback can limit their application to
safety, they developed a fusible separator (SafeTsep) stationary load-leveling systems operating at low
for use in both primary and secondary lithium current densities.197
batteries.112 The separator consists of two or more Shimizu et al. has developed a low-cost MgO porous
microporous layers with at least one of the layers particle separator, which possessed both the high
capable of transforming to a nonporous membrane porosity (∼85%) of the BN felt and the high mechan-
at a temperature between 100 and 150 °C. The ical resistance of powder separators.199 The separator
additional layers maintain the dimensional stability was prepared by loosely sintering fine MgO powder
of the separator during this transformation.113 with Mg(NO3)2 as the binder. The particles thus
formed showed excellent performance as separators
6.4.2.7. Lithium Bromine. Ultrafine porous poly- in Li-Al/FeSx cells.
mer membranes were prepared by Weininger et al.194
by adding sodium benzoate or other salts of benzoic The state of development of the BN-felt separator
acid to the melted polymer in a weight ratio of 70- is more advanced than that of the MgO powder
85 parts of benzoate to 30-15 parts of polymer. The separator. However, the ceramic powder separator
salt does not dissolve in the polymer but forms a has the potential for lower cost than the felt separa-
tor.
dispersion or colloidal suspension. Processing the
polymer involved milling on differential rolls with the
benzoate salt at 140-160 °C for PE or 170-180 °C 7. Separator for Aqueous Batteries
for PP. The mixtures are then cooled close to the
The aqueous batteries use water based electrolytes
softening point of the polymer and sheeted into
(e.g., KOH electrolyte for NiCd and NiMH and H2-
0.0005 in thick films. The final leaching of the sheets
SO4 electrolyte for lead acid), which are less resistive
is done in water at 20-55 °C. The salt was generally
then nonaqueous electrolytes. Polyolefin materials
extracted over a period of 1-16 h. However, leaching
are generally suitable for use in the manufacture of
is almost complete within 5-10 min. The polymers
separators for these batteries, but they are not
are also irradiated with high-energy electrons of 20 inherently wettable by aqueous electrolytes. Such
Mrad at a dose rate of 10 Mrad/min before or after electrolytes are therefore unable to penetrate the
extraction of the salt. This irradiation step is em- pores of a separator formed from such a material, so
ployed to improve the thermal and mechanical prop- that ion migration through the pores in solution will
erties of the porous polymer. Irradiation after leach- not occur without modification. This problem is
ing is more effective in strengthening mechanical and sometimes overcome by treating the polyolefin mate-
thermal properties, but by this treatment, the small- rial with a surfactant, which allows an aqueous
est pores are closed by cross-linking. These mem- electrolyte to wet the material. However, such sur-
branes are flexible, about 50% porous, have ex- factant can be removed from the surfaces of the
tremely fine pore sizes (40-120 Å), and are used as polyolefin material when electrolyte is lost from the
separators in rechargeable lithium bromine batteries. device, for example during charging and discharging
The lithium bromine system is not a commercial cycles, and it is not subsequently replaced on the
system. A lot of developmental work has gone into material when the electrolyte is replenished.
developing the system, but it still is not commercial- This problem has also been more durably ad-
ized. dressed by modifying the surface properties of the
6.4.2.8. Lithium Iron Sulfide (High Tempera- polyolefin materials used to form polymeric sheets,
ture). High-temperature molten salt Li-Al/LiCl- by graft-copolymerizing to those surfaces a mono-
KCl/FeSx cells are known for their high energy meric substance which, after copolymerization, con-
density and superior safety. At one point they were fers hydrophilic properties and, in some cases, ion
being actively pursued for electric vehicle and pulse- exchange properties. This technique has been found
power applications. Historically, boron nitride (BN) to be practical when the porous substrate is formed
cloth or felt has been used as the separator in flooded- from PE, which has been found to lend itself well to
electrolyte cells, while MgO pressed-powder plaques a graft-copolymerization reaction of this kind. How-
have been used in starved-electrolyte cells. ever, it has been found that when such a reaction is
attempted using polyolefin materials other than PE, or other gelling agent on one or both sides.203 The
the rate of the grafting reaction is reduced signifi- paper selected must be of low metallic impurity. A
cantly. kraft paper separator provides thinner separator
Graft polymerization is a convenient method for the spacing, lower internal resistance, and increased
modification of the physical and chemical properties active materials volume.204 One of the critical func-
of polymer materials and is of particular interest for tions of the separator is to keep the zinc can surface
synthesis of the hydrophilic membranes. Graft copo- properly wetted.
lymerization can be achieved by various methods As the paste separator is relatively thick compared
such as an exposure to ionizing radiation or ultra- with the paper liner, about 10% or more manganese
violet light and the use of chemical initiators. Ionizing dioxide can be accommodated in a paper-lined cell,
radiation is one of the most promising methods resulting in a proportional increase in capacity.205
because of its rapid and uniform formation of active The pasted separator carbon-zinc cells were phased
sites for initiating grafting throughout the matrix. out more then a decade ago. All the cells made since
Under appropriate experimental conditions, modifi- then are paper-lined constructions.
cations of polymer properties can be accomplished not The performance and capacity advantages of alka-
only on the surface but also throughout the polymer. line batteries vs carbon-zinc is resulting in the
There have been several reports on radiation continuous decline of this battery. The low cost of the
grafting of acrylic and methacrylic acid onto various carbon zinc cell is a major reason for its continued
substrates. These include both the direct grafting use. Thus, cost is a major consideration in the
method and the preirradiation method to synthesize development and selection of separators for this
ion-exchange membranes. Two cation exchange mem- system.
branes modified with the carboxylic acid group for a
battery separator were prepared by radiation-induced 7.2. Alkaline Zinc MnO2
grafting of acrylic acid (AA) and methacrylic acid
(MA) onto a polyethylene film by Choi et al.200 They Since its introduction in the early 1960s, the
found that KOH diffusion flux of AA-grafted PE alkaline manganese dioxide (Zn/MnO2) cell has ad-
membrane and MA-grafted PE membrane increased vanced to a dominant position in the portable battery
with an increase in the degree of grafting. AA-grafted market. It gets its name from the use of alkaline
PE membrane had a higher diffusion flux then MA- solutions as electrolyte. Alkaline cells have many
grafted PE membrane. Electrical resistance of both acknowledged advantages over zinc-carbon, includ-
membranes decreased rapidly with an increase in the ing a higher energy density, longer shelf life, superior
degree of grafting up to 120% and then leveled off. leakage resistance, better performance in both con-
tinuous and intermittent duty cycles, and lower
Battery separators having carboxylic acid group
internal resistance, which allows them to operate at
were prepared by radiation induced grafting of acrylic
high discharge rates over a wider temperature range.
acid onto a polyolefin nonwoven fabric (PNF). The
PNF comprised approximately 60% polyethylene and In general, alkaline MnO2 cells use “macroporous”
40% polypropylene. It was found that the wetting nonwoven separators made from fibrous materials
speed, electrolyte retention, thickness, and ion- and binders. Various blends of fibers including cel-
exchange capacity increased, whereas the electrical lulose, synthetic fiber, such as PVA fiber, nylon
resistance decreased with increasing grafting yield.201 rayon, etc. are used in producing this type of separa-
The surface characteristics of the separators can also tor. Other types such as gelled, inorganic, and radia-
be modified by plasma discharge. tion-grafted separators have been tried, but they
have not gained much practical use. Some of the more
This subsequent section discusses different types
advanced nonwoven separators are made by coating
of aqueous batteries and the separators used in those
fibers made of an acrylonitride polymer with a
batteries. The recent work in the development of new
surface layer of an acrylic acid which has carboxyl
separators for aqueous batteries will also be dis-
groups, that are then mixed with polypropylene fiber
cussed in detail.
and bound by a PVA fiber. The carboxyl group is
partially cross-linked, resulting in swelling, but
7.1. Leclanche (Zinc Carbon) remains insoluble in the electrolyte. This demon-
Dry cells have been well-known for over 100 years strates the complexity of today’s separators in con-
and form the technical basis of today’s modern dry trast to earlier separator materials, which were
cell industry. Zinc carbon cells are the most widely primarily cellulosic.
used of all the primary batteries worldwide because Techniques commonly used to produce fibrous webs
of their low cost, availability, and acceptability in include the wet laid, dry laid carded, and meltblown
various situations. The two major separator types processes. The wet laid or paper making process is
ever used or in use are gelled paste and paper coated the predominant method for several reasons. The wet
with cereal or other gelling agents such as methyl- laid process, configured properly, allows for the
cellulose.202 The paste type is dispensed into the zinc blending of cellulosic and polymeric components.
can, and the preformed bobbin is inserted, pushing Also, the ability to use short cut length and fine
the paste up the can walls between the zinc and the denier fiber provides for consistent blending, uniform
bobbin. A typical paste electrolyte uses zinc chloride, formation, and controlled pore structure.12
ammonium chloride, water, and starch or flour as the Because of the removal of mercury from the alka-
gelling agents. The coated-paper type uses a special line MnO2 cell, there has been a re-formulation in
paper coated with flour, starch, regenerated cellulose, the anode paste. Because of this change, theses cells
now use a nonwoven laminated to cellophane as the and properties of separators used for lead-acid bat-
separator to ensure against shorting due to zinc teries are reviewed. The reader is referred to recent
solubility and migration. This can lead to higher reviews published by Boehnstedt13,207,208 and
separator resistance. The key to success of such a others209-211 for detailed descriptions of lead-acid
lamination is to use an adhesive (e.g., PVA) that will separators.
not interfere with the ionic movement of the elec-
trode. Some manufacturers of the mercury free cells 7.3.1. Flooded Electrolyte Lead Acid
use nonwoven mat of much smaller pore size made
possible by the availability of 0.1-0.3 denier man- Separators currently used in lead-acid batteries can
made fiber. The smaller diameter fiber allows for be classified based on their materials of construction
more uniform web formation at low ream weight and into four major types: plastic (PE/silica, PVC/silica,
thickness resulting in improved high rate perfor- sintered PVC), paper (phenolic resin impregnated
mance. Often a compromise must be struck between cellulose), glass (glass fiber mat), and rubber (hard
wicking rate and average pore size. rubber/silica, flexible rubber/silica, coated rubber/
silica) separators. Table 9 shows the different types
The rechargeable Zn-MnO2 battery is an out-
of separators used in batteries along with their
growth of the primary battery. In these type of
manufacturing process and main features. Glass,
batteries, a “microporous” type separator is required
paper, and sintered PVC separators can be classified
to resist dendritic growth more likely to evolve as a
as macroporous separators having an average pore
result of numerous recharges. Materials like cel-
diameter greater than 10 µm while all other separa-
lophane, grafted membrane, or polymeric films can
tors can be classified as microporous separators
be used as laminates to absorptive substrates. An-
having an average pore diameter smaller than 1 µm.
other approach providing a similar effect is to coat a
All of these separators can be utilized as leaf separa-
fibrous base with a layer of viskose. Both cellophane
tors in battery construction. Polyethylene can be used
and viskose materials have remarkable properties in
also for enveloped separators around either the
that they are electropermeable when wetted with
positive or the negative plate. The use of “envelope”
KOH. To produce the functional laminate, the base
separators is popular in small, sealed cells, SLI,
nonwoven is designed for wicking and absorbency
motive power, and standby batteries to facilitate
while the cellophane or membrane provides the
production and to control lead contamination during
short resistance. The adhesive used in laminating the
manufacturing.
separator must not adversely affect electrical resis-
tance and should not interfere with the electrochemi- The environment concerns surrounding the lead-
cal behavior of the cell. On the basis of sales, the acid battery (e.g., the automotive battery) have been
alkaline Zn-MnO2 market is about 3-4 times larger increasing in severity in recent years. The improve-
than the lithium-ion battery market.2 As this par- ments and development of the separators have
ticular article is mainly focused on separators used proceeded in accordance with the changes in the
in lithium-based batteries, we will not be discussing specifications for the batteries which were first made
Zn-MnO2 separators in more detail. with wooden separators (the preferred wood was
Oregon cedar as it contained small amounts of lignin
7.3. Lead-Acid Batteries that enhanced the performance of lead negative),
then progressed through microporous rubber separa-
It has been a long time since the invention of the tors, cellulose separators, and synthetic pulp separa-
lead-acid battery, but it still represents the most tors (SPG) with glass mats, to PVC separators, and
important secondary chemical power sourcesboth in now polyethylene separators have evolved. This
number of types and diversity of application. The sequential change in separator technology has pro-
lead-acid battery has maintained its leading role for vided continuous improvements in the charge and
so many decades due to its competitive electrical discharge efficiency of batteries and has given high
characteristics and price and due to its adaptability vehicle-starting capability and reliability. Moreover,
to new applications. It is manufactured in a variety short circuits (caused by particles of active material
of sizes and designs, ranging from less than 1 to over dislodged from the battery plates) are prevented due
10 000 A h.206 to the smaller pores and excellent electrochemical
Lead-acid batteries can be classified into three oxidation resistance of the PE separator. These
major types or categories, namely, automotive (SLI), features contributed greatly to the improvement in
stationary, and motive power (industrial). In addi- battery life.
tion, there are many special batteries that cannot be Rubber separators have good voltage characteris-
easily categorized as either of the above types. As tics, the ability to retard antimony transfer, proper-
these types of batteries are constructed with different ties to retard dendrite growth, and good electrochemi-
materials and design to meet the requirements of cal compatibility.212 Due to the hydrophilic properties
their intended end uses, each requires a particular of the rubber composition, the separators are highly
separator with specific material composition, me- wettable and renewable for the dry-charging process.
chanical design, and physical, chemical, and electro- Paik et al. showed that ACE-SIL (sulfur cured, hard
chemical properties that are tailored for the battery rubber) separators performed well in industrial sta-
and its relevant specific uses. These batteries are tionary or traction batteries, FLEX-SIL (electron-
generally available in flooded electrolyte or valve beam-cured, flexible rubber separator) separators are
regulated (sealed) versions. In this section the types suited for deep-cycling batteries, and MICROPOR-
Table 9. Typical Separators Used in Lead-Acid Battery Systems

separator class manufacturing process properties
wood paper cellulosic separators made from comparatively large pore size and
cotton linters or craft pulp relatively high electrical resistance
and generally coated with phenolic
resin for acid resistance
and strength
hard-rubber213 rubber made by mixing natural rubber, finer pore diameter (0.2 µm average),
rehydrated precipitated silica, relatively lower electrical resistance,
and sulfur, which is then extruded excellent oxidation resistance, retards
and calendared, vulcanized under antimony transfer
water, and dried
flexible rubber213 rubber made by mixing natural rubber, flexible, fine pore structure (0.06 µm
rehydrated precipitated silica, average), retards antimony transfer
which is then extruded and
calendared, irradiated with an
ionizing electron beam, and dried
glass-mat rubber213 rubber/glass mat made by mixing a polymeric finer pore diameter (<0.2 µm average),
emulsion, precipitated silica, and high porosity, excellent thermal
rubber, which is then coated on dimensional stability
a fiberglass mat and finally cured
and dried
sintered PVC plastic made by sintering PVC powder of medium pore size (10-20 µm), generally
a particle size ranging between good chemical resistance
10 and 20 µm
synthetic PVC plastic made from mixture of PVC, silica small to medium average pore size and
fine powder, and a solvent, and relatively low electrical resistance
then extruded, calandered, and
extracted
synthetic pulp with plastic/glass made from blending PE synthetic medium pore size, low electrical
glass mat pulp, synthetic fiber, and fine silica resistance, and long service life at
powder and then heat treated high temp; more difficult to process
and assemble
polyethylene (PE) plastic made from a mixture of UHMW PE fine pore size, low electrical resistance,
powder, fine silica powder and high puncture resisatance, and strongly
mineral oil, which is extruded as a resistant to oxidation
film, calendered, and made porous
by extraction
glass fiber mat glass deposition on a single sheet, excellent wettability, durable in an acid
a mixture of fibers dispersed environment, good resiliency, high
in an aqueous solution temp stability, more difficult to
process and assemble
SIL (a coated, glass mat and rubber separator) “polyethylene separators” is somewhat misleading,
separators have been found to be good choice for high since such a separator consists mainly of agglomer-
rate discharging or cranking applications and for ates of precipitated silica, being held within a net-
various types of gel cells.213 Recently Daramic DC work of extremely long-chain UHMW PE.215 A typical
UHMW PE has demonstrated excellent performance PE separator formulation comprises precipitated
in these applications, as well. silica (∼60 wt %), UHMW PE (∼20 wt %), and
Poly(vinyl chloride) (PVC) and polyethylene (PE) mineral process oil (∼15 wt %), as well as some
separators have been the most commonly used sepa- processing aids, like antioxidants and/or proprietary
rators in automotive batteries for the last 20 years. surface tension modifiers.216,217
Polyethylene separators have a narrow pore size The microporous PE separator is commercially
distribution. The PVC separator is built up by sin- manufactured by passing the ingredients through a
tering PVC powder in general of a particle size heated extruder, passing the extrudate generated by
ranging between 10 and 20 µm. The decrease of the extruder through a die and into the nip formed
particle size in the sintered product is negligible by two heated calendar rolls to form a continuous
compared to the particle size of the raw materials. web, extracting a substantial amount of the process-
The pores are dispersed homogeneously with a me- ing oil from the web by use of a solvent, drying the
dium size ranging between 10 and 20 µm. Since a extracted web, slitting the web into lanes of prede-
PVC separator exclusively consists of PVC, it exhibits termined width, and winding the lanes into rolls.216
advantageously good chemical resistance against acid The PE separators have excellent microporous
and alkaline solutions. Unlike PE, PVC is disadvan- structure for electrolyte flow with minimal lead
taged due to its brittleness. The decline in PVC particle deposits; excellent ductility, strength, and
separators in recent years is in part due to their toughness for envelopability and plate puncture
tendency to yield chloride ions from chemical attack. resistance; excellent oxidation, chemical and thermal
The battery separator currently used by most resistance to resist premature deterioration; and good
flooded cell type lead-acid battery manufacturers are manufacturability with high production efficiency
of the microporous PE type. It was invented in the and relatively low raw material cost, which reduces
late 1960s by W. R. Grace & Co.214 The term overall manufacturing costs.218 The PE pocket sepa-
Table 10. Comparison of Properties of Different Separators Used in Lead-Acid Batteries

property rubber cellulose PVC PE glass fiber
year available 1930 1945 1950 1970 1980
electrical resistance very poor poor poor very good very good
porosity sufficient good poor good very good
battery performance (cold crank) poor sufficient sufficient very good very good
maximum pore size good poor sufficient very good poor
mean pore diameter good poor poor very good poor
purity good fair good good good
resistance to shorting good poor poor very good poor
corrosion resistance very good poor good very good good
oxidation resistance fair poor good very good very good
envelopable (sealability) very poor very poor sufficient very good very poor
flexibility brittle brittle brittle excellent good
ration is in almost all aspects significantly superior are thinner than “industrial” lead-acid batteries used
to leaf separation. Only PE separators can be envel- for standby power sources and traction devices.
oped and can develop good sealability. It has low Endoh220 has reported that one reason for the
electrical resistance, sufficient porosity, small pore occasionally shortened life of batteries assembled
size, and great resistance to both shorting and with PE pocket separators is the development of
corrosion. The PE separator, by virtue of its low internal short circuits at the bottom part of the PE
electrical resistance, generally provides better cold separator due to anodic corrosion causing active
cranking performance. These separators are very material to shed from the positive plates and leak
flexible and offer excellent oxidation resistance if the through the separators. He also found that when
residual oil content is controlled and/or proprietary synthetic pulp (SP) separators with glass mats are
chemical modifiers have been incorporated. A com- used, it is possible not only to restrain the shedding
parison of the properties of different types of separa- from positive plates but also to protect the separators
tors is given in Table 10. against intensive oxidation so that no internal short
PE separators have contributed to improved bat- circuits take place. He concluded that the use of SP
tery specific energy and specific power, increased separators with a glass mat is required to produce
battery cycle life, and higher temperature operating long service-life batteries, especially in tropical re-
capabilities. PE separators have gained in popularity gions.
and have generally replaced PVC, cellulose, glass Higashi et al.221 carried out endurance test under
fiber, and other conventional separators. The transi- high-temperature conditions on automotive batteries
tion to microporous PE envelope separators started made with three different types of separators. One
in the U.S. in the 1970s, followed by Europe in 1980s. group was assembled with PE pocket separators for
Today, PE separators have captured almost 100% of the negative plates, another with PE pocket separa-
the U.S. market and more than 70% of the remaining tors with glass mats for the positive plates, and a
worldwide automotive markets.219 third with leaf-type synthetic pulp separators with
In a flooded cell type lead-acid battery, the battery glass mats. They concluded that battery assembly
separator typically has “ribs” or protrusions extend- with PE pocket separators with a glass mat is an
ing from at least one planar face of the separator. effective way to achieve good endurance (i.e. life
Such ribs are either formed integrally with the extension at high temperature), and leaf-type syn-
backweb of the separator or can be subsequently thetic pulp separators with glass mats are the best
applied to the backweb as a bead of same or different approach for hot climatic conditions.
material as the backweb, or they can be formed by
7.3.2. Valve Regulated Lead Acid (VRLA)
embossing the backweb. The ribs function is to
provide proper spacing between the plates and to An important development in lead-acid battery
provide a space wherein free electrolyte resides. The technology is the valve regulated lead-acid battery.
ribs also provides pressure to hold the electrodes in These batteries operate on the principle of oxygen
contact with the separator. This reduces the need for recombination, using a “starved” or immobilized
precise dimensional control on the cell components. electrolyte. The oxygen generated at the positive
Microporous PE separators typically have a config- electrode during charge can, in these battery designs,
uration comprising a backweb having a predeter- diffuse to the negative electrode, where it can react,
mined thickness and a plurality of parallel or pat- in the presence of sulfuric acid, with freshly formed
terned ribs spaced apart by a predetermined distance lead. The separator material should provide in-
and extending outwardly from one planar surface of numerable gas channels between the plates through
the backweb. The ribs extend continuously in a which oxygen can migrate from the positive to the
longitudinal direction parallel to the edges of the negative electrode. These batteries differ from its
separator material. The thickness of the backweb and flooded electrolyte precursor in a number of impor-
height and spacing of the ribs is specified to the tant ways.222 They have been manufactured for many
separator manufacturer by the battery manufacturer; years with microfiber glass separators, also called
based on specifications designed to maximize certain absorptive glass mat (AGM). They are inherently
battery characteristics desired by the battery manu- resistant to acid stratification and have the ad-
facturer. SLI batteries tend to have separators that ditional important advantage of being essentially
Table 11. Separators and Their Manufacturers for Nickel and Zinc Based Battery Systems
battery system type of separator manufacturer and composition
nickel-cadmium microporous nylon, Pellon 2505, Polypropylene
nickel-metal hydride microporous polypropylene (Scimat, Freudenberg)
nickel-hydrogen microporous polymer-inorganic compounds, asbestos, zirconium oxide cloth,
Celgard 3501, polypropylene
nickel-zinc supported liquid membrane, polypropylene (Celgard 3400, 3501), PVA
microporous, ion-exchange
silver-zinc ion exchange Permion (RAI); Silvergard (RAI); cross-linked PVA
microporous pellon, cellophane, Celgard 3400
zinc-HgO microporous cellulose, poly(vinyl alcohol), polyolefins, poly(vinyl chloride)
zinc-air ion-exchange PVA cross-linked with dimethylolurea
microporous asbestos, nonwoven/polypropylene laminate
zinc/bromine microporous polyolefins (Daramic)
ion-exchange nafion (DuPont)
maintenance free. The separator is a crucial compo- in shorts. The fundamental cause for zinc electrode
nent in determining the useful life of a VRLA cell. shape change is nonuniform current distribution
While a prime function of the VRLA cell separator across the electrode surface. This leads to concentra-
is to hold the cell’s electrolyte in place, it must also tion gradients, which in turn give rise to densification
offer characteristics that prevent major failure mech- and dendrite formation. In the other systems, solu-
anisms occurring in the cell’s positive and negative bility of the negative materials is minimal, and
plates. therefore, failure by shorting is nonexistent or de-
The microglass separator, since its discovery by layed for a large number of cycles. The vented Ni/
McClelland and Devitt, has been the material of Fe, Ni/Cd, and Ni/Zn cells are mostly constructed in
choice for VRLA designs.223,224 It is a wet laid a prismatic configuration. Systems built in a sealed
nonwoven (glass fiber) “paper” and is manufactured cylindrical pressurized configuration are NiCd, NiMH,
on a paper machine. The type of paper machine used and Ni-H2. The pressure configuration in the NiCd
by the manufacturer can influence the separator and NiMH differs from the Ni-H2. In the former,
properties. Three propertiessporosity, uniformity, pressure is used to hold the electrodes and separator
and fiber directionalitysare important attributes in contact. In the latter pressure is the result of
that can be influenced by the type of fiber used. The storage of a gaseous reactant. The type of separators
glass fiber, which has a zero contact angle with the and their manufacturers for nickel-based systems are
acid, is durable in the acid environment, and the fine listed in Table 11
fiber structure also has good resiliency to allow for a
sustained pressure against the plate. The microglass 7.4.1. Nickel−Cadmium
separator has a high porosity in the 90-95% range The NiCd secondary battery is the most popular
and is very conformable. The separator can adapt to alkaline secondary battery, and it is available in
imperfections in the plate surface. The separator also several cell designs and in a wide range of sizes. It
has high-temperature stability. Recent studies have is a very reliable, sturdy, long-life battery, which can
shown that higher levels of fine fiber and higher be operated effectively at relatively high discharge
separator compression provide improved cycle per- rates and over a wide temperature range.
formance in VRLA batteries.225-228
7.4.1.1. Vented NiCd. The vented NiCd battery
On the other hand, AGM separators offer little is one of the best-known power sources in the
control over the oxygen transport rate or the recom- commercial and military fields, particularly for air-
bination process. The arrival of too much oxygen to craft and communication applications. It has excel-
the negative plate could result in overheating, hin- lent high rate and low temperature performance
drance of the battery’s ability to recharge, or even a capabilities. It also has a long useful life capability
loss of capacity. Furthermore, AGM separators ex- and is both physically and electrochemically rugged.
hibit low puncture resistance.
Early separators used in NiCd cells for space
applications consisted of materials as ordinary as
7.4. Nickel Systems Whatman filter paper5 and eventually nonwoven
The nickel-based systems include the flowing sys- nylon felt (Pellon 2505). Pellon 2505 was more stable
tems: nickel-iron (Ni/Fe), nickel-cadmium (NiCd), than cellulosic separators, but it too had stability
nickel-metal hydrides (NiMH), nickel-hydrogen (Ni/ limitations. In the late 1960s and early 1970s,
H2), and nickel-zinc (Ni/Zn). All nickel systems are nonwoven polypropylene saw limited use in NiCd
based on the use of a nickel oxide active material cells. The PP separators were quite useful where
(undergoing one valence change from charge to sterilization (heating the cell to 135 °C) was re-
discharge or vice versa). The electrodes can be pocket quired.229,230 The major difficulty with nonwoven
type, sintered type, fibrous type, foam type, pasted polypropylene was that it was only marginally wet-
type, or plastic roll-bonded type. All systems use an table, and the amount of electrolyte that can be
alkaline electrolyte, KOH. placed in the cell was limited.
The separators may be simple absorbent material In the early 1980s, Pellon Corporation (now
except in Ni/Zn where zinc solubility creates zinc Freudenberg) discontinued the manufacturing of
electrode shape change and zinc dendrites, resulting Pellon 2505. This led to the qualification of Pellon
2536 in 1984 as the new separator material for the Mechanical strength is an important criterion for
major portion of NiCd applications in the U.S.231,232 sealed NiCd cells, because wound cells (spiral type
Hughes Aircraft took a different approach by evalu- construction) are manufactured automatically at very
ating Freudenberg FS2117 (a nonwoven nylon fabric), high speeds. Melt blown PP fleeces, with their
nonwoven polypropylene impregnated with polyben- excellent tensile properties, are good for these cells.
zimidazole (PBI) and zirconium oxide material (Zir- Frequently two layers of the same or different
car) impregnated with either PBI or polysulfone materials are used, to gain increased protection
(PS).233 Their objective was to develop a separator against shorts. For higher temperature applications
that is chemically stable for a minimum of 10 years (up to about 60 °C) PP fleeces are preferred since they
and has a better electrolyte retention and gas perme- offer better chemical stability, though at lower elec-
ability than the state-of-the-art Pellon 2505 separa- trolyte absorption.12
tor. The main purpose of the polymer impregnation The nylon-based separators tend to degrade by
of the zirconia material was to improve the mechan- undergoing hydrolysis in the KOH electrolyte. The
ical strength of the fabric, which was too fragile for products formed by the hydrolysis reaction are be-
easy handling in the fabrication of the NiCd cells. lieved to cause a charge imbalance between positive
The purpose of the PBI impregnation of the polypro- and negative electrodes in a sealed NiCd cell leading
pylene felt was to improve the wettability in the to a loss of overcharge protection. The harmful
electrolyte. They found that Zircar impregnated with products include carbonates, which accumulate in the
PBI was the best separator for geosynchronous orbit electrolyte. The nylon degradation also gives rise to
applications while Zircar impregnated with PS was ammonia and hydroxylamine redox couple, which can
best for low earth orbit applications. The PP impreg- increase the self-discharge rate. The rate of the
hydrolysis of Nylon-66 separator material was ap-
nated with PBI found to be as good as Pellon 2505
proximately one-half of that of the nylon-6 mate-
separator in some cell designs but had longer life
rial.237 It is well-known that in sealed NiCd cells a
because of excellent chemical stability and was
separator weight loss of 10% is often sufficient to
additionally less expensive than zirconia material.
cause internal shorting.238
The separator used in vented NiCd’s is a thin,
multiple ply combination. It consists of a cloth (woven 7.4.2. Nickel−Metal Hydride
or felted nylon) that electrically separates the positive The sealed nickel-metal hydride battery has char-
and negative plates and an ion-permeable plastic acteristics very similar to those of the sealed NiCd
membrane (microporous polypropylene, Celgard 3400, battery. The main difference is that the NiMH
a wettable variant of Celgard 2400) that serves as battery uses hydrogen, absorbed in a metal alloy, for
the gas barrier. This thin gas barrier, which becomes the active negative material in place of the cadmium
relatively soft when wetted, is frequently placed used in the NiCd battery. The NiMH batteries have
between two layers of cloth separators and receives a higher energy density and are considered more
significant mechanical support from them. Previ- environmentally friendly than the NiCd battery. The
ously, cellophane has also been used as the gas sealed NiMH battery, however, does not have the
barrier layer. Studies done by Hill234 and Feldman very high rate capability of NiCd battery, and is less
and Verville235 showed superior performance of NiCd tolerant of overcharge.
batteries with Celgard microporous polypropylene The self-discharge rate of a Ni/MH battery is
separators as compared to cellophane. The authors generally much higher than that of NiCd battery,
attribute this improvement to the better chemical especially when conventional nylon separators are
and thermal stability of the polyolefin sheet mem- used.239 The presence of oxygen and hydrogen gases
brane. The outer nylon separator degrades with use cause the polyamide materials to decompose, produc-
by undergoing hydrolysis in the KOH electrolyte, ing corrosion products which poison the nickel hy-
thus seriously degrading the battery life.236 It in- droxide, promoting premature oxygen evolution and
volves a slow hydrolysis reaction followed by fast also forming compounds capable of a redox shuttle
electrochemical oxidations of the hydrolysis reaction between the two electrodes which further increases
products. the rate of self-discharge.240 In the case of the NiMH
battery, ammonia and amine participate in the
7.4.1.2. Sealed NiCd. In sealed cells, the positive shuttle reaction in the same way that the nitrate ion
and negative electrodes are wound together in a jelly does in the NiCd battery.
roll with a separator material between them. The Ikoma et al. carried out a detailed investigation to
working principle of sealed NiCd batteries is based study the self-discharge mechanism and contribution
on internal oxygen consumption. As a prerequisite of separators.241 They used nonwoven fabric made of
the separator has to be permeable to gaseous oxygen, conventional polyamide (PA), PP (with surfactant),
which is achieved by separator pores being of a and a nonwoven fabric whose main material was
specific minimum size and not all of them being filled sulfonated-PP (hydrophilic) as separators. When
with electrolyte at the same time, so as to leave some nonwoven fabric made of chemically stable sul-
gas channels. Nonwovens (fleeces) made of polya- fonated-PP is used as a separator instead of a
mide, PE, and PP fibers have proven themselves in conventional polyamide separator, the self-discharge
these cells. With their porosity they can absorb rate of the NiMH battery was strongly depressed, to
sufficient electrolyte, and due to their pore size the same level as that of NiCd battery.240,242,243
distribution, they can simultaneously bind electrolyte Recently Nagarajan et al.244 characterized three
and allow oxygen transfer. different commercial AA cells and compared the
materials and cell performance. They used a dif- high temperatures. Although asbestos is still in use,
ferential scanning calorimeter (DSC) to determine there is a trend to replace it with Zircar (zirconium
the materials used as separators. Toshiba cells oxide cloth). Zircar fibrous ceramic separators are
contained both polyethylene and polypropylene, while available in textile product forms (Zircar Products,
the Sanyo and Matsushita cells separators were Inc.). These textiles are composed of zirconium oxide
identified as nonwoven fabrics fabricated from con- fibers stabilized with approximately 10% yttrium
ventional polyamide. Since the Sanyo and Matsushita oxide.251 These materials offer the extreme temper-
cells used the conventional polyamide separator, the ature and chemical resistance of ceramic zirconia.
self-discharge in these cells was substantially higher They are constructed of essentially continuous indi-
due to the shuttle reaction of the ammonia and vidual filaments fabricated in flexible textile forms.
amine, which expedite self-discharge. They also Zircar has the drawback of being expensive and
characterized the commercial cells and determined delicate from an assembly perspective. Aerospace
that separators contribution to the overall weight of cells have shown excellent medium term perfor-
the AA cell was around 3%. mances with their nylon separators, but it is known
Scimat Ltd. has shown that acrylic acid grafted that nylon is weak in terms of its stability over time
nonwoven polyolefin separators have the ability to and its temperature sensitivity.249
absorb chemical impurities, for example ammonia, Zirfon separator is a new alternative for Ni-H2
from the alkaline environments. It has been shown batteries. It is a porous composite separator material
that by using a grafted polyolefin separator, the free composed of a polysulfone matrix and ZrO2, which is
ammonia present inside a NiMH cell is trapped by present in a powder form. The manufacturing is
the separator, resulting in a reduction in self- based on the film-casting technique. It is very stable
discharge to levels normally associated with NiCd in concentrated KOH solutions at elevated temper-
cells.245 In October 2002, Scimat Ltd. announced the atures. These films are around 300 µm thick.252
launch of the next generation of separators for NiMH SORAPEC has tested Zirfon in Ni-H2 cells and has
and NiCd cells using its second-generation grafting indicated that it is one of the best separators.249
technology.246
The commonly used separator material now is the 7.5. Zinc Systems
surface treated polypropylene. The surface treatment Zinc is a nontoxic, relatively inexpensive, and
helps in making the polypropylene permanently abundant material. It is the most electropositive
wettable. Surface treatments involve the grafting of metal which is fully compatible with aqueous elec-
a chemical such as acrylic acid to the base fibers to trolytes. Its low (negative) electrode potential and its
impart wettability and is accomplished using a high hydrogen overpotential make it a very suitable
variety of techniques such as UV or cobalt radiation. negative electrode material for use in aqueous elec-
Another method of imparting wettability to the trolytes.253
polypropylene is a sulfonation treatment where the Although primary cells with zinc electrodes are
base fiber material is exposed to fuming sulfuric acid. widespread, secondary systems based on the same
The separator surface is designed to be made hydro- materials still suffer from a short lifetime arising
philic to the electrolyte. from problems caused by zinc mass redistribution,
Cheng et al.247 carried out the impedance study on undesirable changes in zinc morphology, and the
a foam-type NiMH battery with nonwoven PP sepa- influence of zinc ions on the opposite electrodes.
rator to determine the main causes of early cycle Because nickel-zinc and manganese dioxide systems
deterioration. Their data indicated that the decrease are otherwise suitable for electric vehicle propulsion,
in the voltage characteristic of the battery was due an effort to develop a long-lasting secondary storage
to drying out of the separator that increases the battery is very desirable. One avenue would be to find
ohmic resistance of the battery and that decay of the a better separator for such systems. A lot of work has
total discharge capacity is due to an inactive surface been done by different groups in developing new
that increases the charge-transfer of the battery. separators for zinc-based secondary systems, but
none of them have been very successful in developing
7.4.3. Nickel−Hydrogen long cycle life zinc systems. The type of separators
Nickel-hydrogen batteries offer long cycle life that and their manufacturers for zinc-based systems are
exceeds that of other maintenance-free secondary listed in Table 11.
battery systems and accordingly makes it suitable for
many space applications. Three types of separator 7.5.1. Silver−Zinc
materials have been used for aerospace Ni-H2 cellss The first practical silver-zinc battery was devel-
asbestos (fuel-cell-grade asbestos paper), Zircar (un- oped more then 60 years ago. Since then, primary
treated knit ZYK-15 Zircar cloth),248 and nylon.1,249 and secondary silver-zinc batteries have attracted
In year 2000, Japan supplied over 90% of the nickel- a variety of applications due to their high specific
hydrogen battery separators globally.250 energy, proven reliability and safety, and the highest
Fuel-cell-grade asbestos is a nonwoven fabric with power output per unit weight and volume of all
a thickness of 10-15 mil. The asbestos fibers are commercially available batteries. However, they find
made into a long roll of nonwoven cloth by a carding very limited use in commercial applications, because
or paper-making process. Asbestos is a controversial of their high price and limited cycle life. Development
material with respect to health considerations in of a battery separator which will improve the per-
addition to potentially being a source of corrosion at formance and life of zinc based alkaline cells has been
and continues to be the focus of concerted efforts in

many battery groups, both in industry and academia.
The separator used in silver-zinc cells should be
permeable to water and hydroxyl ions, stable in
strong alkaline solutions, and not oxidized by the
solid silver oxide or dissolved silver ions and should
retard the migration of dissolved ions to the anode.203
A sophisticated separator system is needed to avoid
short-circuits resulting from growth of dendrites.5
Because of the slight solubility of silver oxides in
alkaline electrolyte, little work had been done with
zinc/silver oxide cells until 1941 when Andre254
suggested the use of a cellophane barrier. Cellophane
prevents migrating silver ions from reaching the
anode255 by reducing them to insoluble silver metal.
The cellophane is oxidized and destroyed in the
process, making it less effective for long-life cells.
Cycle life is very limited, even with the best separa-
tors known to date.
In general, secondary silver-zinc cells require a Figure 15. Viscous process for regenerated cellulose.263
minimum of three different separators. The inner
separator, or positive interseparator, serves both as The tendency for cellulose to degrade in alkaline
electrolyte reservoir and as a barrier to minimize solutions has been addressed by using multiple layers
oxidation of the main separator by the highly oxida- of the separator in the battery. Lewis et al.262 showed
tive silver electrode. This separator is usually made that double-ply UCB cellophane films were less
of a relatively inert fiber such as nylon or polypro- subject to dendritic hard shorting by zinc and re-
pylene. The outer separator, or negative intersepa- tained some degree of polymerization “DP” after
rator also serves as an electrolyte reservoir and can extended cycling. Another way to minimize cellulose
also, ideally, serve to stabilize the zinc electrode and degradation in alkaline solutions has been to keep
retard zinc penetration of the main separator, thus the KOH concentration high (around 45%). The rate
minimizing the dendrite growth. Improvement in cell of degradation reaction slows down due to the KOH
life has been reported with improved inorganic nega- tying up the water used in the hydrolysis reaction.263
tive electrode interseparators utilizing such materials The silver-treated cellophane gives improved life
as asbestos and potassium titanate.256 Asbestos helps performance, reduction in rates of silver migration,
in preventing or reducing the magnitude of short and promotes a longer performance life, compared to
circuits and also acts as a silver “stopper”, but it is plain cellophane.264 However, cellophane still has
bulky and reacts with silver oxide. It may also significant disadvantages in terms of poor resistance
contaminate the cell with iron and posts a health risk to attack by the alkaline electrolyte and by active
during fabrication. A potassium titanate mat helps materials in the electrodes. These limit the wet life
to reduce zinc shape changes as well as the magni- of cells, even when used in multiple layers. Cel-
tude of short circuits. It is also bulky and more lophane is also ineffective in preventing shape change,
expensive.1 which curtails cycle life.
The primary ion exchange separator remains the Thomas Danko has described the manufacturing
key to the life of the secondary silver cell. The process and properties of cellulose in detail.263 The
cellulosics (cellophane, treated cellophane, and fi- term regenerated cellulose refers to cellulose that has
brous sausage casing) are usually employed in mul- been converted from a solubilized cellulose derivative
tiple layers as the primary separators for these back into a pure cellulose derivative. The most
cells.257 Cellophane has to be specifically treated as notable of these processes is the viscous process in
battery grade, meaning a very low content of glyc- which cellulose of fairly high DP can be dissolved and
erine (or deglycerinated), since glycerine is an organic maintained at viscosities that are easy to handle (up
material that causes fast degradation of the cell to 300000 cps). Figure 15 describes the viscous
performance by weakening the texture of the cel- process in detail.263 The molecular weight of cellulose
lulosic material under the influence of the strong is described by the degree of polymerization (DP),
oxidants (silver peroxide and silver monoxide) present which is the number of glucose units in the molecular
in colloidal form.258 Cellophane prevents migrating chain. The final DP is selected based on desired
silver ions from reaching the anode by reducing them properties and a balance of the regeneration process
to insoluble silver metal.259 The cellophane is oxidized conditions. Purity of the cellulose plays an important
and destroyed in the process, making it less effective role in the production of regenerated cellulose. The
for long-life cells. It also physically inhibits the zinc impurities (hemicellulose which is a low DP polysac-
dendrite formation.260 Cellophane film is typically charide) in the cellulose reduces the strength of the
drawn to a thickness of 1 mil, and usually five or six final film. The properties of the regenerated cellulose
layers are used to wrap the cathode (silver plate) in can vary depending on the source of wood pulp used
the cell.261 (purity) and the regeneration conditions used to
convert the xanthate derivative back in to cellulose. to as much as 4-5 years. RAI Research Corporation
By using a high purity wood pulp, a stronger and less also developed an improved ion-exchange membrane
porous film can be made. (Silvergard), which has better selectivity toward
For battery separators, regenerated cellulose is silver-ion diffusivity than other ion-exchange mem-
placed on the surface of nonwoven so that the branes.
nonwoven is available to promote the wicking of the
electrolyte. The nonwovens should not allow the 7.5.2. Nickel−Zinc
penetration of viscose into itself. Suitable nonwovens The nickel-zinc (NiZn) system is attractive as a
are made from polypropylene, poly(vinyl alcohol), and secondary cell because of its high energy density and
hardwood hemps. Regenerated cellulose films are low material cost and the low level of potential
commonly used in alkaline manganese cells, both pollutants contained. The widespread use of nickel-
primary and secondary, in NiCd industrial batteries, zinc batteries, particularly as electric vehicle power
as well as in silver-zinc batteries. sources, would be strongly enhanced by significantly
A variety of other separators have been proposed extending the deep-discharge cycle life beyond the
and developed to replace or complement the cellulo- current level of 100-300 cycles.270 Considerable work
sics, such as poly(vinyl alcohol) (PVA),265 radiation- has been done in the past to develop a suitable
grafted polyethylene,260 microporous polypropylene, separator for nickel- and silver-zinc batteries.271,272
inorganic separators,256 and other synthetic polymer An excellent discussion of separator development is
membranes. The inorganic separators are stable at contained in a comprehensive review.273
higher temperatures (∼150 °C), and also resist attack These cells generally fail because of zinc dendrite
by silver oxides and electrolyte, but it leads to higher shorting or zinc electrode shape changes.274 To miti-
cell resistance and they are more expensive. Modified gate the problem of dendrite growth and subsequent
polyethylene films are thinner then standard materi- shorting, battery separators should contain submi-
als leading to higher energy density. They also crometer-size pores, longer pore paths, increased
provide a greater resistance to attachment by silver thickness, and multiple layers.275 As these solutions
oxides and electrolyte but are more expensive, more all lead to increased resistance losses in batteries, a
difficult to handle and provide lower resistance to compromise in the properties of separators is neces-
zinc penetration.1 It requires a coating of the ap- sary. The separator can also influence shape change.
propriate material(s) to improve the resistance to zinc Separators with rapid mass transfer properties will
and silver. reduce the electro-osmosis effects and concentration
The degradation in mechanical properties of non- gradients within the battery, thus slowing the rate
woven battery separator was studied by Danko266 by of shape change. A lot of developmental work has
storing the separators in 40% KOH for long period been done on the nickel-zinc system, but it is still
of time. The test showed that among the nonwovens, not a commercial system.
the polyamide retained about 93% of its initial tensile The most common material used is cellophane,
strength whereas poly(vinyl alcohol) and cellulosic which is a cellulose film, which acts as a membrane
nonwovens retained only 55% and 35%, respectively. and is capable of resisting zinc penetration. The cycle
Adding the viscose coating to the nonwovens im- life of cells utilizing this material is severely limited
proved tensile strength retention by 20-25% for the due to the hydrolysis of the cellophane in alkaline
poly(vinyl alcohol) and cellulosic materials. The solution. Various methods have been tried to stabilize
viscose-coated polyamide retained more than 98% of cellulose materials, such as chemical treatment and
its initial tensile strength. radiation grafting to other polymers, but none have,
Recently Serenyl used a flexible alkaline separator as of now proved economically feasible. The most
(FAS) in Silver-Zinc cells, which consists of a mi- successful zinc migration barrier material yet devel-
croporous polyolefin film, with inorganic filler.267 This oped for the nickel-zinc battery is Celgard mi-
can be folded around the silver and/or zinc electrodes croporous polypropylene film.1 It is inherently hy-
to form conventional “U” wraps or heat sealed bags. drophobic so it is typically treated with a wetting
They showed that the FAS was not attacked by the agent for aqueous applications.
electrolyte and helps in inhibiting the “shape change” There are two broad classes of separators employed
of zinc electrode. in nickel-zinc batteries; a main separator, which
Another commonly used alternate separator mate- exhibits resistance to dendrite penetration, and an
rial is prepared from a radiation graft of methacrylic interseparator, which principally acts as an electro-
acid onto a polyethylene membrane such as Per- lyte reservoir and wicking layer. Both main and
mion.259 The graft makes the film wettable and interseparator should be resistant to chemical attack
permeable to the electrolyte. The low resistance by the alkaline electrolyte and resistant to oxidative
Permion 2192 barrier membrane is suitable for high- attack by nascent oxygen, permanently wettable by
rate KOH cells while higher resistance Permion 2193 the electrolyte, flexible, heat sealable, tear resistant,
is suitable for low rate NaOH cells. Cellophane is still and inexpensive.
often used in conjunction with the grafted separator The two basic kinds of nickel-zinc main separators
as a sacrificial barrier. The lamination of cellophane are the membranes and the microporous separators.
to either side of the Permion membrane results in a Membrane separators are those in which ionic trans-
synergistic action for stopping silver migration.268 Lee port occurs through the interaction of the hydrophilic
and co-workers269 reported that Permion extended groups attached to the polymer with the ionic groups
the service life of Zn/AgO button cells from 2 to up in the electrolyte. Ionic transport through mi-
croporous separators, on the other hand, occurs by Castle Technology Corporation had developed a
diffusion through discrete pores. Other than wetting, supported liquid membrane battery separator to
there is essentially no interaction of this type of main prevent zinc dendrite growth and decrease the shape
separator with the electrolyte. changes on zinc electrode.281 These membranes use
An important advantage of the microporous sepa- a highly specific organic transport reagent contained
rator is that it has high water and electrolyte in a microporous membrane support to transport
permeability leading to less electrode shape change. hydroxyl ions while rejecting zincate ions. Several
They can be prepared from chemically inert poly- materials (polyvinyl chloride (PVC), PVC copolymer,
mers, and hence have good chemical and oxidation nylon, nonwoven PP, and microporous PP) were
resistance. The disadvantages of the microporous tested, and Celgard 2500 was found to be the best
membranes are risk of dendrites shorting through material because of its larger pore size and higher
the discrete pores and greater scope for zinc oxide to porosity.
deposit within the separator pores. Membrane sepa- The nickel zinc system would be a suitable power
rators should provide greater resistance to zinc source for many portable devices if the cells were
dendrite penetration, much lower rates of active tightly sealed and had a prolonged life. In sealed
material migration away from the negative electrode, nickel-zinc batteries, oxygen, which is formed on the
and probably no deposition of zinc oxide within the positive electrode, has to recombine with metallic zinc
separator. Their main disadvantages are the higher at the negative electrode. The microporous membrane
rates of negative electrode shape change resulting must have some degree of permeability to oxygen.282
from poor mass transfer properties and the greater Sato et al. developed a separator which consisted of
vulnerability to chemical attack.276 nylon nonwoven cloth coated with poly(vinyl alcohol)
(PVA) containing boric acid. The separator has a high
Lundquist276 compared the performance of mem- ionic conductivity and is resistant to zincate ion
brane separators with microporous separators and penetration.283
found that the rate of zinc electrode shape change is To address the zinc dendrite problem in nickel-
greater in batteries with membrane separators al- zinc cells, eVionyx claims to have developed a pro-
though in either case shape change was the primary prietary membrane system that is nonporous, has
failure mode. Batteries containing separators with very high ionic conductivity, is of low cost, and can
the lowest electrolytic resistivity and highest water block zinc dendrite penetration even in high concen-
permeability gave the longest cycle life. The results trations of KOH.284-286 The polymeric membrane has
favored the use of microporous separators since its an ionic species contained in a solution phase thereof.
superior mass transfer characteristics minimize the The ionic species behaves like a liquid electrolyte,
rate of shape change. while at the same time the polymer-based solid gel
Studies made by several authors, such as Poa,277 membrane provides a smooth impenetrable surface
Wagner,278 and Brown,279 have shown significant that allows the exchange of ions for both discharging
improvement in the performance of NiZn recharge- and charging of the cell.
able batteries, when microporous polyolefin flat sheet
membranes were used as separators. Poa and co- 7.5.3. Zinc−Air
workers examined the diffusion of the OH- electro- For many applications zinc-air technology offers
lyte through candidate separator materials, including the highest available energy density of any primary
microporous polypropylene, Celgard 3501 (Celgard battery system. It also offers a flat discharge voltage,
2500 coated with a proprietary surfactant). The OH- long shelf life, safety and ecological benefits, and low
diffusional flow for Celgard 3501 was found to be energy cost. Since these cells are open to the ambient
greater than either the traditional swollen, nonpo- atmosphere, a factor limiting universal applications
rous cellophane separator used in the battery or a of zinc-air technology is the trade off between long
nonporous methacrylic acid grafted and cross-linked service life and maximum power capability.
polyethylene.277
The separator materials used in zinc-air batteries
Krejci et al. studied the transport of zincate ions should be inert to oxidation, stable in cells at high
through modified microporous PP separators (Cel- potentials, chemically inert to KOH at high temper-
gard 3401, Celgard 3501) using polarography and atures, have a fine porous capillary structure to hold
conductometry.280 Celgard 3401 and 3501 were cho- the electrolyte in the pores (and limit diffusion of
sen because they are proven to have slow redistribu- oxygen gas), have a high bubble pressure, and exhibit
tion of zinc material, good mechanical properties, and low electrical resistance. Usually a polypropylene
low electrical resistivity. Soluble Nafion as an ion membrane such as treated Celgard 2500 or Celgard
exchange modifying agent was applied to the PP 3501 is used as separator in these cells. The function
microporous separators. They studied the influence of the separator is to prevent cross mixing of the
of Nafion and surfactant (coated on 3401 and 3501) positive and negative electrolytes but to still allow
on transport of zinc ions through the membrane. It the transport of ions. Because of its porosity, the
was found that the zincate ions transport through cationic compound produced in the anode reaction
the membrane was higher for higher conductivity can easily pass through the separator to cathode that
membranes and vice versa. Higher amounts of sur- leads to an increased polarization and decreased
factant and Nafion coating increase both the mem- capacity and battery life. The separator should be
brane conductivity and the membrane’s ability to able to stop the passage of these cationic compounds
allow zinc ions to pass. to increase the cell life. The commercial Zn-air cells
also use laminated separators (e.g. Celgard 4560 and port of bromine through microporous separators and
Celgard 5550), which have a microporous membrane accumulation of a bromine-oil complex in the sepa-
laminated with a nonwoven. rator lead to efficiency losses. Thus, improved sepa-
Dewi et al. used a cationic polysulfonium mem- rators with lower bromine mass transport and higher
brane as a separator in a zinc-air cell. They showed electrolytic conductivity would be beneficial for im-
that the polysulfonium membranes were highly ef- proving the energy efficiency of zinc-bromine bat-
fective in preventing zinc cation permeation from the teries.
anode to the cathode compared to the commercially
available PP separators.287 7.6. Redox Flow Batteries
A new polymeric separator membrane has been The redox flow battery (RFB) concept was first
developed for alkaline electrochemical systems (e.g. proposed by L. H. Thaller at the NASA Lewis
metal/air, Zn/Ni, Zn/MnO2, Zn/HgO etc.) by Schubert Research Center, Cleveland, OH.293,294 Since then, it
et al.288,289 It is a water-soluble copolymer of acrylic has been investigated as one of the energy storage
acid and sodium styrenesulfonate monomers (PAASS). systems for solar power, nuclear power, load leveling
The monomer ratio is varied to manipulate electro- purposes, and electric vehicles.295
lyte absorption and ionic resistivity of the copolymer. The all-vanadium RFB developed by the University
Processes have been developed to coat PAASS onto of New South Wales has the advantage of a higher
conventional nonwoven separator papers to produce electromotive force (1.4 V in a vanadium system
composite membranes. The monomer ratio and the compared to 1.1 V in an Fe/Cr system) and a higher
coating weight can be adjusted to balance ionic energy density compared with other RFB systems.296
resistivity and shorting protection to provide ad- It employs V(II)/V(III) and V(IV)/V(V) redox couples
equate overall performance. PAASS coated separa- in the negative and positive half-cell electrolytes. The
tors demonstrate good barrier properties to soluble ideal membrane for this system should exhibit low
zincate and silver species. Preliminary use of PAASS permeation rates of the active species to minimize
separator materials in a rechargeable Zn/air system self-discharge and also have low area resistivity to
showed a significant improvement in shorting protec- minimize losses in internal energy. Further, the
tion over microporous materials such as Celgard membrane should also exhibit good chemical stability
3401. for long cycle life. The energy losses in a V-RFB are
The primary zinc-air cells have captured the partly caused by the undesired transport of vana-
hearing aid market, but the rechargeable zinc-air dium ions of different oxidation states across the
battery is still in the developmental stage. membrane, which separates the anolytic from the
catholytic solution in the battery system.297
7.5.4. Zinc−Bromine Ion exchange membranes of cation and anion
selectivity are usually used as the membrane separa-
The zinc-bromine battery is an attractive technol- tors for V-RFB. Commercially available cation selec-
ogy for both utility-energy storage and electric-vehicle tive CMV membrane (Asahi Glass) degrades by the
applications. The technology has been limited in use fully charged anolyte,298 but the Nafion membrane
and not commercialized due to two major draw- (Dupont) does not appear to degrade. However, the
backs, the tendency of zinc to form dendrites and current efficiency of the V-RFB using Nafion de-
the high solubility of bromine in the aqueous zinc creases due to self-discharge of the vanadium ions,
bromide electrolytes. Dendritic zinc deposits could which permeated through the membrane.298 Further-
easily short circuit the cell, and the high solubility more, the cost of Nafion is exceptionally high. Chlo-
of bromine allows diffusion and direct reaction with rosulfonated polyolefin membrane shows a good
the zinc electrode, resulting in self-discharge of the resistivity to oxidation and its resistivity in the RFB,
cell. almost equal to Nafion.299 It has also been reported
Various materials have been used as separators in that a cross-linked chlorosulfonated homogeneous
zinc-bromine cells. Ideally a material is needed membrane formed by accelerated electron radiation
which allows the transport of zinc and bromide ions at 2.5 Mrad/pass had a longer lifetime than the non
but does not allow the transport of aqueous bromine, cross-linked membrane because of the increase in the
polybromide ions, or complex phase structures. Ion chemical stability.300,301
selective membranes are more efficient at blocking Microporous membranes perform poorly in V-RFB
transport then nonselective membranes.290 These compared to ion exchange membranes due to their
membranes, however, are more expensive, less du- poor selectivity. Modification of low cost microporous
rable, and more difficult to handle then microporous separators to impart perm-selective properties could
membranes (e.g., Daramic membranes).291 The use lead to significant cost reduction. Chieng et al.
of ion selective membranes can also produce problems developed a composite membrane from Daramic (a
with the balance of water between the positive and microporous separator), treated with Amberlite 400CG
negative electrolyte flow loops. Thus, battery devel- (an ion-exchange resin), and cross-linked using divi-
opers have only used nonselective microporous ma- nylbenzene.302,303 The separator selectivity was im-
terials for the separator.292 proved by partially blocking or by reducing the pore
Microporous separators have the advantage of size or by incorporating some ion exchange capability
being relatively low-cost and adequately stable in the to the separator. The chemical stability of the mem-
electrolyte, but unfortunately they contribute to brane was found to be excellent, and the membrane
energy-efficiency losses in the battery. Rapid trans- was not susceptible to fouling.304 Sulfonation of the
cross-linked Daramic separators was done to impart fouling decreased when the degree of cross-linking
some cation-exchange capacity, and it was successful of the anion-exchange membrane was increased.310
in reducing the amount of water transport across the Hruska et al.311 evaluated various microporous
membrane in V-RBC.305 The scale-up of sulfonation materials (Celgard 5511 composed of porous polypro-
process is both difficult and expensive due to the very pylene and Daramic W0733 composed of porous
corrosive nature of concentrated sulfuric acid. Incor- polyethylene) to reduce the columbic efficiency loss
poration of cation-exchange groups using poly(sodium caused by the oxidation of the iron plate. The chemi-
4-styrenesulfonate) (PSSS) has been studied, and cal resistances of both Celgard and Daramic mem-
results show that the method reduces the water branes in iron electrolytes were excellent at pH ) 0.
transport across the membrane. The PSSS treatment Celgard separators with very low resistivity offered
step can be easily scaled up.306 some improvement in columbic efficiency but the
Diffusion coefficients of vanadium ions in CMS Daramic separators gave very high columbic efficien-
(Neosepta made by Tokuyama Soda), CMV (Selemion cies in a laboratory size redox cell. Daramic separa-
made by Asahi Glass), and CMX (Neosepta made by tors also limited the ferric ion diffusional loss to about
Tokuyama Soda) cation exchange membranes were 3% for both charge and discharge modes of operation.
determined by measuring the ion-exchange fluxes of Two commercial ion exchange membranes were also
the vanadium ions with H3O+ ions using a dialysis evaluated in this study: Neosepta ACH-45T (Tokuya-
cell.297 The lowest diffusion coefficients were observed ma Soda Company Ltd.) and Selemion DMV (Asahi
in the CMS membrane for all vanadium ions. CMS Glass Company Limited). In laboratory cells, the
membranes were found to be most suitable for membranes discolored, giving evidence of deteriora-
V-RFB, as it was expected to prevent cross contami- tion, and the resistivity of the membrane increased
nation of vanadium ions.307 significantly. The increased resistivity was attributed
One of the most important requirements that must to FeCl4- fouling. Thus, attempts at lowering the
be met is the membrane’s ability to prevent excessive polarization by maintaining a pH differential across
transfer of water from one half cell to the other. The the cell with commercial anion membranes were
preferential transfer of water can be a problem in the unsuccessful because of membrane fouling. In spite
vanadium battery as one half-cell (the negative half of the poor voltaic performance of the membrane, the
cell in the case of cation exchange membranes) is overall Coulombic efficiency was found to be excep-
flooded and becomes diluted, while the other becomes tionally good.
more concentrated, adversely affecting the overall
operation of the cell. Most of the membranes show 8. Mathematical Modeling of Batteries/Separators
good initial water transfer properties, but their
performance deteriorates with exposure to the vana- Computer simulations have been used as a tool for
dium solutions. Sukkar et al.308 evaluated various understanding and optimizing battery performance
polyelectrolytes to determine whether they could since the 1970s.312-314 Continued progress in compu-
improve the selectivity and stability of the mem- tational tools has enabled ever-increasing sophistica-
branes in the vanadium redox cell solutions. Both the tion in battery modeling and a steady increase in the
cationic and anionic polyelectrolytes evaluated im- number of systems to which modeling has been
proved the water transfer properties of the mem- applied. Today it is possible to obtain simulation
branes, although upon extended exposure to the codes for all of the major rechargeable batteries, some
vanadium electrolyte the modified membranes did of which are available for free.315
not maintain their improved water transfer proper- The mathematical models of different types of
ties. The solvent based Nuosperse 657 modified batteries (lead-acid,316,317 NiMH,318 lithium-ion319,320)
membrane displayed exceptional properties initially have been developed during the past few years.315
but also failed to maintain its performance with This has led to a better understanding of those
extended exposure to the vanadium solutions. systems. The present models consider only the thick-
Another system under investigation is the iron/ ness and porosity of the separators. Not much work
chromium redox flow battery (Fe/Cr RFB) developed has been done in incorporating the effect of physical
by NASA. The performance requirements of the and chemical properties of separators on the perfor-
membrane for Fe/Cr RFB are severe. The membrane mance and safety of batteries. This is also because
must readily permit the passage of chloride ions, but the microstructure of separators and their effect on
should not allow any mixing of the chromium and transport properties in batteries are generally known
iron ions. An anionic permselective membrane CDIL- only qualitatively. A thorough understanding of the
AA5-LC-397, developed by Ionics, Inc., performed microstructure of separators would be beneficial for
well in this system.309 It was prepared by a free modeling studies and optimization of electrochemical
radical polymerization of vinylbenzyl chloride and systems. This will help in making the battery model
dimethylaminoethyl methacrylate in a 1:1 molar predictions more practical and reliable.
ratio. One major issue with the anionic membranes The separator pore structure is usually very com-
was its increase in resistance during the time it was plex. It consists of a porous network of interconnected
exposed to a ferric chloride solution. The resistance pores, which are filled with liquid electrolyte. A
increase termed “fouling” is related to the ability of complete description of the pore structure would
the ferric ion to form ferric chloride complexes, which require a very intricate model. Simulations are only
are not electrically repelled by the anionic membrane. practically possible if the structure is represented by
An experiment by Arnold and Assink indicated that a simplified quasi-continuum involving a few param-
eters. In such an approach, the “effective” electrolytic batteries. The model shows that, as the cell potential
conductivity, σeff, is often defined by97 exceeds the oxidation potential of the polymer, the
cell is transformed, over a time scale of a few
σeff ) Rσ0, R ≈ 1.5 (11) minutes, from a battery into a resistor, after which
a steady-state overcharge condition is attained.
where σ0 is the bulk ionic conductivity of the elec-
trolyte, is the void volume fraction of separator filled 9. Summary
with electrolyte, and R is the Bruggeman exponent. The ideal battery separator would be infinitesi-
The general applicability of R ∼ 1.5, appears ques- mally thin, offer no resistance to ionic transport in
tionable because separator pores are never of an ideal electrolytes, provide infinite resistance to electronic
shape. Fan and White321 chose a R value of 2.5 for conductivity for isolation of electrodes, be highly
separators in NiCd batteries, and Doyle et al.322 used tortuous to prevent dendritic growths, and be inert
3.3 for lithium-ion batteries. Arora et al.323 measured to chemical reactions. Unfortunately, in the real
the value as 2.4 for PVdF based separators by world the ideal case does not exist. Real world
measuring the separator and electrolyte conductivity separators are electronically insulating membranes
at different salt concentrations. Doyle et al. used an whose ionic resistivity is brought to the desired range
even higher Bruggeman exponent of 4.5 for quantify- by manipulating the membranes thickness and po-
ing the ionic conductivity of their plasticized electro- rosity.
lyte membrane.322 It is clear that no single separator satisfies all the
Patel et al. showed that a Bruggeman exponent of needs of battery designers, and compromises have to
1.5 is often not valid for real separator materials, be made. It is ultimately the application that decides
which do not have uniform spherical shape.324 Porous which separator is most suitable. We hope that this
networks based on other morphologies such as oblate paper will be a useful tool and will help the battery
(disk-type) ellipsoids or lameller increase the tortuous manufacturers in selecting the most appropriate
path for ionic conductivity and result either in a separators for their batteries and respective applica-
significant increase of the exponent R, or in a tions. The information provided is purely technical
complete deviation from the power law. They showed and does not include other very important param-
that spherical or slightly prolate ellipsoidal pores eters, such as cost of production, availability, and
should be preferred for separators, as they lead to long-term stability.
higher ionic conductivity separators. There has been a continued demand for thinner
Tye325 explained that separator tortuosity is a key battery separators to increase battery power and
property determining transient response of a separa- capacity. This has been especially true for lithium-
tor and steady-state electrical measurements do not ion batteries used in portable electronics. However,
reflect the influence of tortuosity. He recommended it is very important to ensure the continued safety
that the distribution of tortuosity in separators be of batteries, and this is where the role of the separa-
considered; some pores may have less tortuous paths tor is greatest. Thus, it is essential to optimize all
than others. He showed mathematically that separa- the components of battery to improve the perfor-
tors with identical average tortuosities and porosities mance while maintaining the safety of these cells.
could be distinguished by their unsteady-state be- Separator manufacturers should work along with the
havior if they have different distributions of tortu- battery manufacturers to create the next generation
osity. of batteries with increased reliability and perfor-
Doyle et al.322 used a mathematical model to mance, but always keeping safety in mind.
examine the effect of separator thickness for the This paper has attempted to present a comprehen-
PVdF:HFP gel electrolyte system and found that sive review of literature on separators used in various
decreasing separator thickness below 52 µm caused batteries. It is evident that a wide variety of separa-
only a minor decrease in ohmic drop across the cell. tors are available and that they are critical compo-
The voltage drops in the electrodes were much more nents in batteries. In many cases, the separator is
significant. Mao and White326 developed a math- one of the major factors limiting the life and/or
ematical model for discharge of a Li/TiS2 cell. Their performance of batteries. Consequently, development
model predicted that increasing the thickness of the of new improved separators would be very beneficial
separator from 25 to 100 µm decreased the discharge for the advanced high capacity batteries.
capacity from 95% to about 90%; further increasing
separator thickness to 200 µm reduced discharge
capacity to 75%. These theoretical results indicate
10. Future Directions
that conventional separators (25-37 µm thick) do not Up until more recent history, most of the separa-
significantly limit mass transfer of lithium. tors and membranes historically used had not been
The use of electroactive polymers for overcharge specifically developed for battery applications. Thus,
protection has been recently reported for lithium-ion future research should be aimed at developing sepa-
batteries.327,328 The electroactive polymer incorpo- rators that are specifically tailored for battery ap-
rated into a battery’s separator is an attractive new plications. The general objectives of separator re-
option for overcharge protection. Thomas et al.329 search should be as follows: (a) to find new and cost-
developed a mathematical model to explain how effective separators, (b) to understand the separator
electroactive polymers such as polythiophene can be properties in batteries, and (c) to optimize separator
used to provide overcharge protection for lithium-ion properties related to specific cell performance, life,
and safety. One way to achieve some of these goals (34) Sun, L.; Chen, G.; Xu, D.; Abe, T. Presented at the 204th Meeting
of The Electrochemical Society, 2003; Abstract 423.
will be to develop mathematical models that reflect (35) Sun, L. U.S. Patent 2003/0152828A1, 2003.
the effects of separator resistance, thickness, pore (36) Johnson, B. A.; White, R. E. J. Power Sources 1998, 70, 48.
size, shrinkage, tortuosity, and mechanical strength (37) Hoffman, H. G. Proceedings of the Tenth Annual Battery Confer-
ence on Applications and Advances; IEEE: New York, 1995; p
on the final performance and safety of batteries. The 253.
battery separators for tomorrow will demand more (38) Bradford, S. M. Battery Power Prod. Technol. 2004, (Mar), 17.
than just good insulation and mechanical filtration; (39) Frost & Sullivan Research World Rechargeable Battery Markets
for Mobile IT and Communication Devices (A575-27); 2002.
they will require unique electrochemical properties. (40) Geiger, M.; Callahan, R. W.; Diwiggins, C. F.; Fisher, H. M.;
Hoffman, D. K.; Yu, W. C.; Abraham, K. M.; Jillson, M. H.;
Nguyen, T. H. Presented at The Eleventh International Seminar
11. Acknowledgments on Primary and Secondary Battery Technology and Application,
Fort Lauderdale, FL, Florida Educational Seminars Inc., Feb
The authors would like to acknowledge numerous 28-Mar 3, 1994.
colleagues and staff members at Celgard, LLC., and (41) Tanba, H. Molding Process 1999, 11, 759.
(42) Adachi, A.; Spotnitz, R. M.; et al. Osaka Chem. Marketing Center
Polypore, International, Inc., for their help with the 1997, 69-80.
preparation and publication of this paper. (43) Yu, W. C.; Hux, S. E. U.S. Patent, 5,952,120, 1999.
(44) Hipore, Asahi Kasai, http://www.asahi-kasai.co.jp/memrbane/
english/tradenm/t07.html.
12. References (45) Bierenbaum, H. S.; Isaacson, R. B.; Druin, M. L.; Plovan, S. G.
Ind. Eng. Chem. Prod. Res. Dev. 1974, 13, 2.
(1) Linden, D.; Reddy, T. B. Handbook of Batteries, 3rd ed.; McGraw- (46) Jpn. Ind. News 1996, 91.
Hill: New York, 2002. (47) Kim, S. S.; Lloyd, D. R. J. Membr. Sci. 1991, 64, 13.
(2) Besenhard, J. O. Editor, Handbook of Battery Materials; Wiley- (48) Druin, M. L.; Loft, J. T.; Plovan, S. G. U.S. Patent 3,801,404.
VCH: Weinheim, Germany, 1999. 1974.
(3) Berndt, D. Maintenance Free Batteries, 3rd ed., Research Studies (49) Schell, W. J.; Zhang, Z. The Fourteenth Annual Battery Confer-
Press Ltd.: Taunton, Somerset, England, 2003. ence on Applications and Advances, Long Beach, CA; IEEE: New
(4) Bode, H. Lead-Acid Batteries; John Wiley: New York, 1977. York, 1999; p 161.
(5) Falk, S. U.; Salkind, A. J. Alkaline Storage Batteries; John (50) Isaacson, R. B.; Bierenbaum, H. S. U.S. Patent 3,558,764, 1971.
Wiley: New York, 1969. (51) Kamei, E.; Shimomura, Y. U.S. Patent 4,563,317, 1986.
(6) Fleischer, A.; Lander, J. J. Zinc-Silver Oxide Batteries; John (52) Yu, T. H. Processing and Structure-Property Behavior of
Wiley: New York, 1971. Microporous Polyethylene-From Resin to Final Film Ph.D.
(7) Brodd, R. J., Friend, H. M., Nardi, J. C., Eds. Lithium Ion Battery Dissertation, Virginia Polytechnic Institute and State Univer-
Technology: ITE-JEC Press: Brunswick, OH, 1995. sity, Blacksburg, VA, 1996.
(8) Wakihara, M., Yamamoto, O., Eds. Lithium Ion Batteries, (53) Sarada, T.; Sawyer, L. C.; Ostler, M. I. J. Membr. Sci. 1983, 15,
Fundamentals and Performance; Wiley-VCH: New York, 1998. 97.
(9) Yoshino, A. Chem. Ind. 1995, 146, 870. (54) Bierenbaum, H. S.; Daley, L. R.; Zimmerman, D.; Hay, I. L. U.S.
(10) Schalkwijk, W. A. V., Ed. Advances in Lithium Ion Batteries; Patent, 3,843,761, 1974.
Kluwer Academic: New York, 2002. (55) Hamer, E. A. G. U.S. Patent, 4,620,956, 1986.
(11) Kinoshita, K.; Yeo, R. Survey on Separators for Electrochemical (56) Hiroshi, K.; Tetuo, A.; Akira, K. U.S. Patent 5,691,047, 1997.
Systems; LBNL: January 1985. (57) Kesting, R. E. Synthetic Polymeric Membranes, 2nd ed.; John
(12) Benett, J.; Choi, W. M. Developments in small cell separators. Wiley & Sons Inc.: New York, 1985; Chapter 2.
In Proceedings of the 10th Annual Battery Conference on Ap- (58) Ihm, D. W.; Noh, J. G.; Kim, J. Y. J. Power Sources 2002, 109,
plications & Advances; IEEE: New York, 1995; p 265. 388.
(13) Boehnstedt, W. In Handbook of Battery Materials; Besenhard, (59) Takita, K.; kono, K.; Takashima, T.; Okamoto, K. U.S. Patent,
J. O., Ed.; VCH Wiley: Amsterdam and New York, 1999. 5,051,183, 1991.
(14) Spotnitz, R. In Handbook of Battery Materials; Besenhard, J. (60) Michiyuki, A.; Jpn. Patent 8064194, 1996.
O., Ed.; VCH Wiley: Amsterdam and New York, 1999. (61) Kotaro, T.; Koichi, K.; Tatsuya, T.; Kenkichi, O. U.S. Patent 5,-
(15) Shirai, H.; Spotnitz, R. Lithium Ion Secondary Battery-Materi- 051,183, 1991.
als and Applications; Yoshio, K, Ed.; Nikkan Kogyo Shin-bun: (62) Koichi, K.; Kotaro, T.; Mamoru, T.; Tatsuya, T. Jpn. Patent
Tokyo, 1996; p 91 (in Japanese). 8012799, 1996.
(16) Shirai, H.; Spotnitz, R.; Atsushi, A. Chem. Ind. 1997, 48, 47 (in (63) Norimitsu, K.; Kotaru, T.; Koichi, K.; Hidehiko, F. U.S. Patent
Japanese). 6,153,133, 2000.
(17) Hiroshi, T. In The Latest Technologies of the New Secondary (64) Akinao, H.; Kazuo, Y.; Hitoshi, M. U.S. Patent 6,048,607, 2000.
Battery Material; Ogumi, Z., Ed.; CMC: Tokyo, 200X; p 99 (in (65) Pekala, R. W.; Khavari, M. U.S. Patent 6,586,138, 2003.
Japanese). (66) Userguide, FreedomCar Separator Costing Document, February
(18) Hiroshi, T. In Advanced Technologies for Polymer Battery; 2003.
Oyama, N., Ed.; CMC: Tokyo, 200X; p 165 (in Japanese). (67) Xu, M.; Hu, S.; Guan, J.; Sun, X.; Wu, W.; Zhu, W.; Zhang, X.;
(19) Koichi, K. In Advanced Technologies for Polymer Battery; Oyama, Ma, Z.; Han, Q.; Liu, S. U.S. Patent 5,134,174, 1992.
N., Ed.; CMC: Tokyo, 200X; p 174. (In Japanese). (68) Fisher, H. M.; Wensley, C. G. U.S. Patent 6,368,742, 2002.
(20) Kiyoshi, K. In Lithium Secondary Battery Technology for the 21st (69) Zhu, W.; Zhang, X.; Zhao, C.; Zu, W.; Hou, J.; Xu, M. Polym.
Century, Kanamura, K., Ed.; CMC: Tokyo, 200X; p 116. Adv. Technol. 1996, 7, 743.
(21) Brodd, R. J.; Bullock, K. R.; Leising, R. A.; Middaugh, R. L.; (70) Sadamitsu, K.; Ikeda, N.; Hoki, Manabu, Nagata, K.; Ogino, K.
Miller, J. R.; Takeuchi, E. J. Electrochem. Soc. 2004, 151, K1. World Patent Application 02066233A1, 2002.
(22) Battery Power Prod. Technol. 2003, 7 (May), 15. (71) Higuchi, H.; Matsushita, K.; Ezoe, M.; Shinomura, T. U.S. Patent
(23) Takeshita, H. Presented at The 21st International Seminar & 5,385,777, 1995.
Exhibit on Primary & Secondary Batteries, Fort Lauderdale, FL, (72) Calis, G. H. M.; Daemen, A. P. M.; Gerrits, N. S. J. A.; Smedinga,
Florida Educational Seminars Inc., March 8, 2004. J. T. J. Power Sources 1997, 65, 275.
(24) Battery EV Technol. 2004, 28 (Jan), 2. (73) Ooms, F. G. B.; Kelder, E. M.; Schoonman, J.; Gerrits, N.;
(25) Pilot, C. The Worldwide Rechargeable Battery Market. Pre- Smedinga, J.; Calis, G. J. Power Sources 2001, 97-98, 598.
sented at Batteries 2004, 6th ed.; Paris, June 2-4, 2004. (74) Yamamura, Y.; Ooizumi, S.; Yamamoto, K. Separator for re-
(26) Celgard LLC http://www.celgard.com. chargeable lithium ion batteries with high puncture strength
(27) Celgard completes capacity expansion, Global Sources, http:// and high melt rupture temperature. Nitto Denko Technol. Rep.
www.globalsources.com, February 04, 2004. (http://www.nitto.com/rd/rd6_1.html), 2001, 39, 39.
(28) Asahi Kasei adding battery membrane capacity. Nikkei (75) Pekala, R. W.; Khavari, M.; Dobbie, G.; Lee, D.; Fraser-Bell, G.
Net Interact. http://www.nni.nikkei.co.jp/AC/TNKS/ Presented at the 17th International Seminar & Exhibit on
Nni20030806D06JFA23.htm, August 7, 2003. Primary and Secondary Batteries, Fort Lauderdale, FL, Florida
(29) Advanced Rechargeable Battery Industry 2001/2002, Nomura Educational Seminars, Inc., March 6-9, 2000.
Research Institute Limited, 2002. (76) Fleming, R.; Taskier, H. Prog. Batteries Solar Cells 1990, 9, 58.
(30) About Edison Batteries, Inc., http://www.optodot.com/sys-tmpl/ (77) Hoffman, D.; Fisher, H.; Langford, E.; Diwiggins, C. Prog.
htmlpage/. Batteries Solar Cells 1990, 9, 48.
(31) Battery and Fuel Cell Components; The Fredonia Group: 2003. (78) Yu, W. C.; Callahan, R. W.; Diwiggins, C. F.; Fischer, H. M.;
(32) Adv. Battery Technol. 2004, 40 (Feb), 22. Geiger, M. W.; Schell, W. J. Presented at the North America
(33) Hamano, K.; Yoshida, Y.; Shiota, H.; Shiraga, S.; Aihara, S.; Membrane Society Conference, Breckenridge, CO, 1994.
Murai, M.; Inuzuka, T. U.S. Patent 6,664,007, B2, 2003. (79) Kuribayashi, I. J. Power Sources 1996, 63, 87.
(80) Pasquier, A. D.; Gozdz, A.; Plitz, I.; Shelburne, J. Presented at (124) Secondary Lithium Cells and Batteries for Portable Applications.
the 201st meeting of The Electrochemical Society, Philadelphia, International Electrotechnic Commission, IEC 61960-1 and IEC
PA, May 12-17, 2002. 61960-2.
(81) Augustin, S.; Volker, H.; Gerhard, H.; Christian, H. Desalination (125) Recommendations on the Transport of Dangerous Goods, Manual
2002, 146, 23. of Tests and Criteria. United Nations: New York, 1999.
(82) http://www.separion.com. (126) Safety Standard for Lithium Batteries, UL 1642, Underwriters
(83) Hying, C. Separion separators for lithium batteries-safety & Laboratories Inc, Third Edition, 1995.
performance. Presented at Batteries 2004, 6th Ed.; Paris, June (127) Standard for Household and Commercial Batteries, UL 2054,
2-4, 2004. Underwriter Laboratories, Inc., 1993.
(84) Sachan, S.; Ray, C. A.; Perusich, S. A. Polym. Eng. Sci. 2002, (128) UN Recommendations on the Transport of Dangerous Goods,
42, 1469. December 2000.
(85) Sachan, S.; Perusich, S. Electrochemical Society Meeting, Se- (129) A Guideline for the Safety Evaluation of Secondary Lithium
attle, 1999. Cells. Japan Battery Association, 1997.
(86) Carlson, S. A. Membr. Sep. Technol. News 2004, 22, 8. (130) Venugopal, G. J. Power Sources 2001, 101, 231.
(87) Abraham, K. M. Electrochim. Acta 1993, 38, 1233. (131) Alamgir, M.; Abraham, K. M. In Lithium Batteries: New
(88) Gineste, J. L.; Pourcell, G. J. Membr. Sci. 1995, 107, 155. Materials, Developments and Perspectives; Industrial Chemistry
(89) Hoffman, D. K.; Abraham, K. M. In Proceedings of the Fifth Library 5; Pistoia, G., Ed.; Elsevier: New York, 1994; Chapter
International Seminar on Lithium Battery Technology and 3.
Applications; Florida Educational Seminars, Inc.: Deerfield (132) Gray, F. M. Polymer Electrolytes; RSC Materials Monograph;
Beach, FL, 1991. The Royal Society of Chemistry: Cambridge, U.K., 1997.
(90) USABC “Development of low cost separators for lithium-ion (133) Fauteux, D.; Massucco, A.; McLin, M.; Van Buren, M.; Shi, J.
batteries”, RFPI 2001. Electrochim. Acta 1995, 40, 2185.
(91) Laman, F. C.; Sakutai, Y.; Hirai, T.; Yamaki, J.; Tobishima, S. (134) North, M.; Markin, T. L.; Hooper, A.; Tofield, B. C. Second
Ext. Abstr., 6th Int. Meet. Lithium Batteries 1992, 298-300. International Meeting on Lithium Batteries, Paris, France, April
(92) Laman, F. C.; Gee, M. A.; Denovan, J. J. Electrochem. Soc. 1993, 25-27, 1984;Extended Abstracts # 19.
140, L51. (135) Chem. Int. 1982, (June, No. 3), 14.
(93) Abraham, K. M. Electrochim. Acta 1993, 38, 1233. (136) Appetecchi, G. B.; Dautzenberg, G.; Scrosati, B. J. Electrochem.
(94) Robinson, R. G.; Walker, R. L. In Batteries; Collins, D. H., Ed.; Soc. 1996, 143, 6.
The MacMillan Company: New York, 1963; p 15. (137) Armand, M. Solid State Ionics 1983, 9 & 10, 745.
(95) Lander, J. J.; Weaver, R. D.; Salkind, A. J.; Kelley, J. J. In (138) Lightfoot, P.; Mehta, M. A.; Bruce, P. G. Science 1993, 262, 883.
Characteristics of Separators for Alkaline Siver Oxide Zinc (139) Vincent, C. A.; Scrosati, B. Modern Batteries. An Introduction
Secondary Batteries. Screening Methods; Cooper, J. E., Fleischer, to Electrochemical Power Sources; Arnold: London, 1993.
A., Eds.; NASA Technical Report NAS 5-2860, 1964. (140) Appetecchi, G. B.; Passerini, S. J. Electrochem. Soc. 2002, 149,
(96) Kilroy, W. P.; Moynihan, C. T. J. Electrochem. Soc. 1978, 125, A891.
520. (141) Croce, F.; Appetecchi, G. B.; Persi, L.; Scrosati, B. Nature
(97) MacMullin, R. B.; Muccini, G. A. AIChE J. 1956, 2, 393. (London) 1998, 394, 4496.
(98) Callahan, R. W.; Nguyen, K. V.; McLean, J. G.; Propost, J.; (142) Croce, F.; Persi, L.; Ronci, F.; Scrosati, B. Solid State Ionics 2000,
Hoffman, D. K. Proceedings of the 10th International Seminar 135, 47.
on Primary and Secondary Battery Technology and Application; (143) Fan, J.; Fedkiw, P. S. J. Electrochem. Soc. 1997, 144, 399.
Florida Educational Seminars, Inc.: Fort Lauderdale, FL, 1993. (144) Appetecchi, G. B.; Romagnoli, P.; Scrosati, B. Electrochem.
(99) Ionov, V. V.; Isakevitch, V. V.; Katalevsky, E. E.; Chernokoz, A. Commun. 2001, 3, 281.
J. J. Power Sources 1990, 30, 321.
(145) Kim, D. W.; Sun, Y. K. J. Power Sources 2001, 102, 41.
(100) Lowell, S.; Shields, E. Powder Surface Area and Porosity, 3rd
(146) Chojnacka, J.; Acosta, J. L.; Morales, E. J. Power Sources 2001,
ed.; Chapman and Hall: New York, 1991.
97-98, 819.
(101) PMI Conference 2000 Proceedings, PMI short course, Ithaca, NY,
(147) Nishi, Y. Advances in Lithium Ion Batteries; van Schalkwijk,
Oct 16-19, 2000.
W., Scrosati, B., Eds.; Kluwer Academic/Plenum Publishers:
(102) Porous Materials Inc, http://www.pmiapp.com.
New York, 2002.
(103) Jena, A. K.; Gupta, K. M. J. Power Sources 1999, 80, 46.
(104) Jena, A. K.; Gupta, K. M. J. Power Sources 2001, 96, 214. (148) EE Times.com: http://www.eet.com/story/OEG19990121S0013.
(105) Zeman, L.; Denault, L. J. Membr. Sci. 1992, 71, 221. (149) Song, J. Y.; Wang, Y. Y.; Wan, C. C. J. Power Sources 1999, 77,
(106) Chen, R. T.; Saw, C. K.; Jamieson, M. G.; Aversa, T. R.; Callahan, 183.
R. W. J. Appl. Polym. Sci. 1994, 53, 471. (150) Min, H. S.; Ko, J. M.; Kim, D. W. J. Power Sources 2003, 119-
(107) Fujii, T.; Mochizuki, T. U.S. Patent, 5, 759, 678, 1998. 121, 461.
(108) Spotnitz, R.; Ferebee, M.; Callahan, R. W.; Nguyen, K.; Yu, W. (151) Jo, S. I.; Sohn, H. J.; Kang, D. W.; Kim, D. W. J. Power Sources
C.; Geiger, M.; Dwiggens, C.; Fischer, H.; Hoffman, D. Proceed- 2003, 119-121, 478.
ings of the 12th International Seminar on Primary and Second- (152) Kim, H. S.; Kum, K. S.; Cho, W. I.; Cho, B. W.; Rhee, H. W. J.
ary Battery Technology and Applications; Fort Lauderdale, FL, Power Sources 2003, 124, 221.
Florida Educational Seminars, Inc.: Fort Lauderdale, FL, 1995. (153) Abraham, K. M.; Alamgir, M. Solid State Ionics 1994, 70-71,
(109) Venugopal, G.; Moore, J.; Howard, J.; Pendalwar, S. J. Power 20.
Sources 1999, 77, 34. (154) Schmutz, C.; Tarascon, J. M.; Gozdz, A. S.; Schumtz, C. N.;
(110) Venugopal, G. The role of plastics in lithium-ion batteries. Warren, P. C.; Shokoohi, F. K. Proc. Electrochem. Soc. 1995, 94
Proceedings of the 3rd Annual Conference on Plastics for Portable (28), 330.
and Wireless Electronics; Philadelphia, PA, 1997; p 11. (155) Murata, K.; Izuchi, S.; Yoshihisa, Y. Electrochim. Acta 2000, 45,
(111) Lundquist, J. T.; Lundsager, C. B.; Palmer, N. L.; Troffkin, H. 1501.
J.; Howard, J. U.S. Patent 4,731,304, 1998. (156) Jiang, Z.; Carroll, B.; Abraham, K. M. Electrochim. Acta 1997,
(112) Lundquist, J. T.; Lundsager, C. B.; Palmer, N. I.; Troffkin, H. 42, 2667.
J. U.S. Patent 4,650,730, 1987. (157) Song, J. Y.; Cheng, C. L.; Wang, Y. Y.; Wan, C. C. J. Electrochem.
(113) Zuckerbrod, D.; Giovannoni, R. T.; Grossman, K. R. Proceedings Soc. 2002, 149, A1230.
of the 34th International Power Sources Symposium; Cherry Hill, (158) Gozdz, A. S.; Schmutz, C. N.; Tarascon, J. M.; Warren, P. C.
NJ, 1990; p 172. U.S. Patent, 5,456,000, 1995.
(114) Spotnitz, R.; Ferebee, M. W. Meeting Abstracts, The Electro- (159) Gozdz, A. S.; Tarascon, J. M.; Schmutz, C. N.; Warren, P. C.;
chemical Society Inc., Volume 96-2, Fall Meeting, San Antonio, Gebizlioglu, O. S.; Shokoohi, F. Tenth Annual Battery Conference
TX, October 6-11; The Electrochemical Society: New York, on Advances and Applications, Long Beach, CA, January 10-
1996. 13; IEEE: New York, 1995; 301.
(115) Faust, M. A.; Suchanski, M. R.; Osterhoudt, H. W. U.S. Patent (160) Tarascon, J. M.; Gozdz, A. S.; Schumtz, C. N.; Shokoohi, F. K.;
No. 4,741,979, 1988. Warren, P. C. Solid State Ionics 1996, 86-88, 49.
(116) Matthias, U.; Dieter, B.; Heinrich, R.; Thomas, B.; U.S. Patent, (161) Pasquier, A. D.; Waren, P. C.; Culver, D.; Gozdz, A. S.; Amatucci,
6,511,517, 2003. G.; Tarascon, J. M. Proc. Electrochem. Soc. 1999, 99 (24), 360.
(117) Maleki, H.; Shamsuri, A. K. J. Power Sources 2003, 115, 131. (162) Park, C. K.; Kakirde, A.; Ebner, W.; Manivannan, V.; Chai, C.;
(118) Zeng, S.; Moses, P. R. J. Power Sources 2000, 90, 39. Ihm, D. J.; Shim, Y. J. J. Power Sources 2001, 97-98, 775.
(119) Norin, L.; Kostecki, R.; McLarnon, F. Electrochem. Solid State (163) Dasgupta, S.; Jacobs, J. K. U.S. Patent, 5,437,692, 1995.
Lett. 2002, 5, A67. (164) Abraham, K. M.; Alamgir, M.; Hoffman, D. K. J. Electrochem.
(120) Kostecki, R.; Norin, L.; Song, X.; McLarnon, F. J. Electrochem. Soc. 1995, 142, 683.
Soc. 2004, 151, A522. (165) Pendalwar, S. L.; Howard, J. N.; Venugopal, G.; Oliver, M. U.S.
(121) Hazardous Materials Regulations; Code of Federal Regulations, Patent, 5,716,421, 1998.
CFR49 173.185. (166) Gozdz, A. S.; Plitz, I.; Du Pasquier, A.; Zheng, T. Proceedings of
(122) UL1640, Lithium Batteries. Underwriters Laboratories, Inc. the 200th ECS Meeting, Fall 2001, Phoenix, AZ, 2001; Vol. 2000-
(123) UL2054, Household and Commercial Batteries. Underwriters 2001, pp 336-351.
Laboratories, Inc. (167) Gozdz, A. S. U.S. Patent 6,328,770, 2001.
(168) Kim, D. W.; Oh, B.; Park, J. H.; Sun, Y. K. Solid State Ionics (215) Wang. L. C.; Harvey, M. K.; Stein, H. L.; Scheunemann, U. The
2000, 138, 41. Role of UHMW-PE in Microporous PE Separators. Proceedings
(169) Wang, Y.; Sejdic, J. T.; Steiner, R. Solid State Ionics, 2002, 148, of the 12th Annual Battery Conference on Applications &
443. Advances; IEEE: New York, 1997; p 69.
(170) Gozdz, A. S.; Plitz, I.; DuPasquier, A.; Zheng, T. Presented at (216) Larsen, D. W.; Kehr, C. L. U.S. Patent 3,351,495, 1996.
the 198th Meeting of the Electrochemical Society, Phoenix, AZ, (217) Boehnstedt, W. J. Power Sources 2001, 95, 234.
Oct 22-27, 2000. (218) Wang, L. C.; Harvey, M. K.; Ng, J. C.; Scheunemann, U. J. Power
(171) Spotnitz, R. M.; Wensley, C. G. U.S. Patent 6,322,923, 2000. Sources 1998, 73, 74.
(172) Fabrice, C.; Bradford, R. WO 02/50929 A2, 2002. (219) Boehnstedt, W. J. Power Sources 1996, 59, 45.
(173) Jeong, Y. B.; Kim, D. W. J. Power Sources 2004, 128, 256. (220) Endoh, H. J. Power Sources 1996, 599, 51.
(174) Abraham, K. M.; Alamgir, M.; Hoffman, D. K. J. Electrochem. (221) Higashi, T.; Endoh, H. J. Power Sources 1998, 73, 110.
Soc. 1995, 142, 683. (222) Rand, D. A. J.; Woods, R.; Dell, R. M. Batteries for Electric
(175) Dasgupta, S.; Jacobs, J. K. U.S. Patent 5,498,489, 1996. Vehicles; Research Studies Press: 1998; ISBN 0-86380-205-0.
(223) McClelland, D. H.; Devitt, J. L. U.S. Patent 3,862,861, 1975.
(176) Gies, P.; Charles, R. A.; Dieken, J. World Patent 00/13248, 2000.
(224) Zguris, G. C. J. Power Sources 1998, 73, 60.
(177) Eschbach, F. O.; Oliver, M. U.S. Patent, 5,681,357, 1997. (225) Zguris, G. C. J. Power Sources 1997, 67, 307.
(178) Hamano, K.; Shiota, H.; Shiraga, S.; Aihara, S.; Yoshida, Y.; (226) Zguris, G. C. J. Power Sources 1996, 59, 131.
Murai, M,; Inuzuka, T. U.S. Patent, 5, 981, 107, 1999. (227) Pavlov, D.; Ruevski, S.; Naidenov, V.; Sheytanov, G. J. Power
(179) Akashi, H. U.S. Patent, 5,658,686, 1997. Sources 2000, 85, 164.
(180) Akashi, H. Paper presented at the International Symposium on (228) Ferreira, A. L. J. Power Sources 1999, 78, 41.
Polymer Electrolytes, ISPE-5, Uppsala, Sweden, Aug 11-16, (229) Britting, A. O. Proceedings of the 1981 Goddard Space Flight
1996. Center Battery Workshop; NASA Conference Publication 2217;
(181) Fujii, T. Proceedings of the 17th International Seminar & Exhibit NASA: Washington, DC, 1981; p 327.
on Primary and Secondary Batteries; Florida Educational Semi- (230) Britting, A. O. J. Power Sources 1984, 12, 305-316.
nars, Inc.: Boca Raton, FL, 2000. (231) Morrow, G. W. J. Power Sources 1987, 21, 313.
(182) Nakane, I.; Narukawa, S. Presented at Power 2000, The 8th (232) Francis, R. W.; Haag, R. L. J. Power Sources 1986, 18, 147.
Annual International Conference on Power Requirements for (233) Lim, H. S.; Pickett, D. F. Separator Evaluation in NiCd Cells.
Mobile Computing and Wireless Communications, San Diego, Proceedings of the Intersociety Energy Conversion Engineering;
CA, Sept 24-27, 2000. IEEE: New York, 1990; Vol. 3, p 61.
(183) Kim, K. M.; Ryu, K. S.; Kang, S. G.; Chang, S. H.; Chung, I. J. (234) Hill, J. M. U.S. Army ERADCOM, Devet-TR-75-F, 1979.
Macromol. Chem. Phys. 2001, 202, 866. (235) Feldman, K.; Verville, G. DREO Technical Note, 78-6, 1978.
(184) Kim, K. M.; Park, N. G.; Ryu, K. S.; Chang, S. H. Polymer 2002, (236) Lim, H. S.; Margerum, J. D.; Verzwyvelt, S. A.; Lackner, A. M.
43, 3951. Proceedings of the 27th Power Sources Conference, Atlantic City,
(185) Kim, K. M.; Ko, J. M.; Park, N. G.; Ryu, K. S.; Chang, S. H. NJ, June 21-24; 1976; p 83.
Solid State Ionics 2003, 161, 121. (237) Lim, H. S.; Margerum, J. D.; Verzwyvelt, S. A.; Lackner, A. M.;
(186) Scrosati, B.; Croce, F.; Persi, L. J. Electrochem. Soc. 2000, 147, Knechtli, R. C. J. Electrochem. Soc. 1989, 136, 605.
1718. (238) Scott, W. R.; Rusta, D. W. NASA RP 1052, 1979.
(187) Prosini, P. P.; Villano, P.; Carewska, M. Electrochim. Acta 2002, (239) Fetcenko, M. A.; Venkatesan, S.; Ovshinsky, S. Proceedings of
48, 227. the Symposium on Hydrogen Storage Materials, Batteries &
(188) Liu, X.; Kusawake, H.; Kuwajima, S. J. Power Sources 2001, Electrochemistry; 1992; p 141.
97-98, 661. (240) Wada, M. Polym. Adv. Technol. 1994, 5, 645.
(189) Kim, D. W.; Ko, J. M.; Chun, J. H.; Kim, S. H.; Park, J. K. (241) Ikoma, M.; Hoshina, Y.; Matsumoto, L.; Iwakura, C. J. Electro-
Electrochem. Commun. 2001, 3, 535. chem. Soc. 1996, 143, 1904-1907.
(190) Kim, D. W.; Noh, A.; Chun, J. H.; Kim, S. H.; Ko, J. M. Solid (242) Ikoma, M.; Takahashi, O.; Tsuboi, R.; Matsumoto, L. Denki
State Ionics 2001, 144, 329. Kagaku 1993, 61, 997.
(243) Furukawa, N. J. Power Sources 1994, 51, 45.
(191) Dan, P.; Mengeritski, E.; Geronov, Y.; Aurbach, D.; Weisman, I.
(244) Nagarajan, G. S.; Van Zee, J. W. J. Power Sources 1998, 70,
J. Power Sources 1995, 54, 143.
173-180.
(192) Ikeda, H.; et al. Proceedings of the Symposium on Lithium (245) Cook, J. A.; Lancaster, I. M. Proc. Electrochem. Soc. 1998, 98
Batteries; The Electrochemical Society: Pennington, NJ, 1984; (15), 55.
Vol. 84, 311. (246) Scimat’s latest separators. Batteries Int. 2002 (Oct).
(193) Grady, J. P. Presented at the 20th International Seminar and (247) Cheng, S.; Zhang, J.; Liu, H.; Leng, Y.; Yuan, A.; Cao, C. J. Power
Exhibit on Primary and Secondary Batteries, Fort Lauderdale, Sources 1998, 74, 155-157.
Florida, Florida Educational Seminars, Inc., Mar 17-20, 2003. (248) Hamling, P.; Hamling, B. H. Zirconia Battery SeparatorssAn
(194) Weininger, J. L.; Holub, F. F. J. Electrochem. Soc. 1970, 117, historical perspective and development update. Proceedings of
340. the 11th Annual Battery Conference on Applications & Advances;
(195) Martino, F. J.; Gay, E. C.; Moore, W. E. J. Electrochem. Soc. IEEE: New York, 1996; p 15.
1982, 129, 2701. (249) Vermeiren, P. H.; Adriansens, W.; Moreels, J. P.; Leysen, R. Int.
(196) Swaroop, R. B.; Battles, J. E. J. Electrochem. Soc. 1981, 128, J. Hydrogen Energy 1998, 23, 5, 321-324.
1873. (250) Chem. Business Newsbase 2001 (July 25).
(197) Mathers, J. P.; Boquist, C. W.; Olszanski, T. W. J. Electrochem. (251) http://www.zircarzirconia.com.
Soc. 1978, 125, 1913. (252) Vermeiren, P. H.; Adriansens, W.; Leysen, R. Int. J. Hydrogen
(198) Bandyopadhyay, G.; Swaroop, R. B.; Battles, J. E. J. Electrochem. Energy 1996, 21, 8, 679-684.
Soc. 1982, 129, 2187. (253) Barak, M., Ed. Electrochemical Power Sources; Peter Peregri-
(199) Shimizu, Y.; Terasaki, M.; Kashihara, S. J. Power Sources 1984, nus: London, 1980.
13, 235. (254) Andre, H. Bull. Soc. Fr. Elect. 1941, 6 (1), 132.
(200) Choi, S. H.; Park, S. Y.; Nho, Y. C. Radiat. Phys. Chem. 2000, (255) Agostino, V. D.; Lee, J.; Orban, G. In Grafted Membranes in
57, 179. Zinc-Silver Oxide Batteries; Fleisher, A., Lander, J. J., Eds.;
(201) Choi, S. H.; Kang, H. J.; Ryu, E. N.; Lee, K. P. Radiat. Phys. Wiley: New York, 1971; p 271.
Chem. 2001, 60, 495. (256) Charkey, A. Proc. 26th Annual Power Sources Symp. 1976, 87.
(202) Navy Primary and Secondary Batteries, Design and Manufac- (257) Karpinski, A. P.; Makovetski, B.; Russell, S. J.; Serenyi, J. R.;
turing Guidelines, Department of the Navy, Sept 1991; p 85. Williams, D. C. J. Power Sources 1999, 80, 53-60.
(203) Linden, D. Handbook of Batteries, 2nd ed.; McGraw-Hill: New (258) Himy, A. Silver Zinc BatterysPhenomena & Design Principles,
York, 1994. 1st ed.; Vantage Press: New York, 1986.
(204) Vinal, G. Primary Batteries; Wiley: New York, 1950. (259) Agostino, V. D.; Lee, J.; Orban, G. In Grafted Membranes in
(205) Mantell, C. L. Batteries & Energy Systems, 2nd ed. McGraw- Zinc-Silver Oxide Batteries; Fleisher, A., Lander, J. J., Eds.;
Hill: New York, 1983. Wiley: New York, 1971; p 271.
(260) Himy, A. Silver-Zinc Batteries: Phenomena and Design Prin-
(206) Mardegain, S. B. Battery Power Prod. Technol. 2003, (Jan).
ciples, 1st ed.; Vantage Press: New York, 1983; p 31.
(207) Boehnstedt, W. J. Power Sources 1996, 59, 45. (261) Lewis, H.; Grun, C.; Salkind, A. J. Power Sources 1997, 65, 29.
(208) Boehnstedt, W. J. Power Sources 2004, 133, 59. (262) Lewis, H.; Jackson, P.; Salkind, A.; Danko, T.; Bell, R. J. Power
(209) Lander, J. J. Proceedings of the Symposium on Battery Separa- Sources 2001, 96, 128.
tors; The Electrochemical Society: Columbus, OH, 1974; p 4. (263) Danko, T. Properties of cellulose separators for alkaline second-
(210) Prout, L. J. Power Sources 1993, 46, 117. ary batteries. Proceedings of the 10th Annual Battery Conference
(211) Vinal, G. W. Storage Batteries; Wiley: New York, 1945. on Applications & Advances; IEEE: New York, 1995; p 261.
(212) Butherus, A. D.; Lindenberger, W. S.; Vaccaro, F. J. Bell Syst. (264) Lewis, H. L.; Hammersley, V. L.; Wharton, S. P. NSWC:
Technol. J. 1970, 1377. evaluation of cellophane separation in model rechargeable
(213) Paik, S. L.; Terzaghi, G. J. Power Sources 1995, 53, 283. silver-zinc cell. Presented at the 38th Power Sources Confer-
(214) Kung, J. J. Power Sources 1994, 48, 129. ence, Cherry Hill, NJ, 1998.
(265) Lewis, H.; Henderson, S.; Danko, T. Separator composition (293) Thaller, L. H. U.S. Patent 3,996,064, 1976.
evaluation in model rechargeable silver zinc cells. Proceedings (294) Jorne, J. Am. Sci. 1983, 71 507.
of the 16th Annual Battery Conference on Applications & (295) Sum, E.; Skyllas-Kazacos, M. J. Power Sources 1985, 15, 179.
Advances; IEEE: New York, 2001. (296) Skyllas-Kazacos, M.; Grossmith, F. J. Electrochem. Soc. 1987,
(266) Danko, T. Strength Properties of Separators in Alkaline Solu- 134, 2950.
tions. Proceedings of the 11th Annual Battery Conference on (297) Wiedemann, E.; Heintz, A.; Lichtenthaler, R. N. J. Membr. Sci.
Applications & Advances; IEEE: New York, 1996; p 283. 1998, 141, 207.
(267) Serenyl, R. Proceedings of the 41st Power Sources Conference, (298) Skyllas-Kazacos, M.; Kasherman, D.; Hong, D. R.; Kazacos, M.
Philadelphia, June 14-17, 2004; 2004; p 465. J. Power Sources 1991, 35, 399.
(268) Megahed, E. A.; Davig, D. C. Power Sources; Academic: London,
(299) Ohya, H.; Kawahara, T.; Kang, A. S.; Aihara, M.; Negishi, Y.
1981; Vol. 8.
Denki kagaku 1994, 62, 863.
(269) Lee, J.; Agostino, V. D.; Zapisek, S.; Freid, R.; Cannizzaro, J.
Membranes and Ionic and Electronic Conducting Polymers; (300) Hwang, G. H.; Ohya, H. J. Membr. Sci. 1997, 132, 55.
Yeager, E. B., Ed.; The Electrochemical Society Inc.: Pennington, (301) Hwang, G. J.; Ohya, H. J. Membr. Sci. 1996, 120, 55.
NJ, 1983; p 245. (302) Chieng, S. C.; Kazacos, M.; Skyllas-Kazacos, M. J. Power Sources
(270) Drumm, J. J. British Patent 365,125, 1930. 1992, 39, 11.
(271) Kishimoto, T.; Yamane, M.; Bogauchi, T.; Eguchi, Y. U.S. Patent (303) Chieng, S. C.; Kazacos, M.; Skyllas-Kazacos, M. J. Membr. Sci.
5,320,916, 1994. 1992, 75, 81.
(272) Kishimoto, T.; Yamane, M.; Bogauchi, T.; Eguchi, Y. U.S. Patent (304) Mohammadi, T.; Skyllas-Kazacos, M. J. Membr. Sci. 1995, 98,
5,547,779, 1996. 77.
(273) McLarnon, F. R.; Cairns, E. J. J. Electrochem. Soc. 1991, 138 (305) Mohammadi, T.; Skyllas-Kazacos, M. J. Membr. Sci. 1995, 107,
(Feb). 35.
(274) McBreen, J.; Cairns, E. J. In Advances in Electrochemistry & (306) Mohammadi, T.; Skyllas-Kazacos, M. J. Power Sources 1995, 56,
Electrochemical Engineering; Gerischer, H., Tobias, C. W., Eds.; 91.
Wiley: New York, 1978; Vol. 11; pp 273-351. (307) Wiedemann, E.; Heintz, A.; Lichtenthaler, R. N. J. Membr. Sci.
(275) Bennion, D. N. A review of membrane separators and Zinc- 1998, 141, 215.
Nickel oxide battery development. Prepared for Argonne Na- (308) Sukkar, T.; Kazacos, M. S. J. Membr. Sci. 2003, 222, 249.
tional Laboratory under Contract No. 31-109-38-5455, October (309) Assink, R. A. J. Membr. Sci. 1984, 17, 205.
1980.
(310) Arnold, C.; Assink, R. A. J. Appl. Polym. Sci. 1984, 29, 2318.
(276) Lundquist, J. T. J. Membr. Sci. 1983, 13, 337.
(277) Poa, D. S.; Cook, G. M.; Yao, N. P. ANL/OEPM-83-4. National (311) Hruska, L. W.; Savinell, R. F. J. Electrochem. Soc. 1981, 128,
Technical Information Service: 1983. 18.
(278) Wagner, C. A.; Almerini, A.; Smith, R. L. Proceedings of the 29th (312) Newman, J.; Tiedemann, W. AIChE J. 1975, 21, 25.
International Power Sources Symposium; The Electrochemical (313) Newman, J. S. Electrochemical Systems, 2nd ed.; Prentice Hall
Society Inc.: Pennington, NJ, 1980. Inc.: Englewood Cliffs, NJ, 1991.
(279) Brown, R. A.; Cloyd, J. S. Proceedings of the 29th International (314) Vidts, P. D.; White, R. E. J. Electrochem. Soc. 1997, 144, 1343.
Power Sources Symposium; The Electrochemical Society Inc.: (315) Ceder, G.; Doyle, M.; Arora, P.; Fuentes, Y. MRS Bull. 2002,
Pennington, NJ, 1980. 27, 619.
(280) Krejci, I.; Vanysek, P.; Trojanek, A. J. Electrochem. Soc. 1993, (316) Newman, J.; Tiedemann, W. J. Electrochem. Soc. 1997, 144,
140, 2279. 3081.
(281) J. Power Sources 1984, 11, 242-243. (317) Gu, H.; Nguyen, T. V.; White, R. E. J. Electrochem. Soc. 1987,
(282) Bugnet, B.; Doniat, D.; Rouget, R. Proc. 40th Power Sources Conf. 134, 2953.
2002, 10 (13), 535. (318) Vidts, P. D.; Delgado, J.; White, R. E. J. Electrochem. Soc. 1995,
(283) Sato, Y.; Kanda, M.; Niki, H.; Ueno, M.; Murata, K.; Shirogami, 142, 4006.
T.; Takamura, T. J. Power Sources 1983, 9, 147. (319) Doyle, M.; Fuller, T. F.; Newman, J. J. Electrochem. Soc. 1993,
(284) Chen, M.; Tsai, T. U.S. Patent 6,605,391, 2003. 140, 1526.
(285) Chen, M.; Li, L. F.; Tsai, T. U.S. Patent, 6,358,651, 2002. (320) Fuller, T. F.; Doyle, M.; Newman, J. J. Electrochem. Soc. 1994,
(286) Chen, M.; Cao, F.; Liang, L.; Tsai, T.; Faris, S. Proc. 41st Power 141, 1.
Sources Conf. 2004, 473. (321) Fan, D.; White, R. E. J. Electrochem. Soc. 1991, 138, 17.
(287) Dewi, E. L.; Oyaizu, K.; Nishide, H.; Tsuchida, E. J. Power
Sources 2003, 115, 149-152. (322) Doyle, M.; Newman, J.; Gozdz, A. S.; Schumtz, C. N.; Tarascon,
(288) Schubert, M. A.; Myers, J. P.; Thrasher, G. T. Acrylic Acid- J. M. J. Electrochem. Soc. 1996, 143, 1890.
Sodium Styrene Sulfonated Copolymer Separator Coatings for (323) Arora, P.; Doyle, M.; Gozdz, A. S.; White, R. E.; Newman, J. J.
Alkaline Electrolytes. Presented at the 202nd Meeting of the Power Sources 2000, 88, 219.
Electrochemical Society, Salt Lake City, UT. (324) Patel, K. K.; Paulsen, J. M.; Desilvestro, J. J. Power Sources
(289) Schubert, M. A.; Myers, J. P. Proc. 41st Power Sources Conf. 2003, 122, 144.
2004, 122. (325) Tye, F. L. J. Power Sources 1983, 9, 89.
(290) Will, F. G. J. Electrochem. Soc. 1979, 126, 36. (326) Mao, Z.; White, R. E. J. Power Sources 1993, 43-44, 181.
(291) Bellows, R.; Einstein, H.; Grimes, P.; Kantner, E.; Malachesky, (327) Denton, F.; Howard, J. N.; Anani, A. A.; Fernandez, J. M. U.S.
P.; Newby, K.; Tsein, H. Development of a Circulating Zn-Br2 Patent 6,228,516, 2001.
Battery Phase I. Final Report, Exxon Research and Engineering (328) Mao, H.; Wainwright, U.S. Patent 6,074,766, 1990.
Company and Sandia National Laboratories, SAND82-7022, Jan (329) Thomas-Alyea, K. E.; Newman, J.; Chen. G.; Richardson, T. J.
1983. J. Electrochem. Soc. 2004, 151, A509.
(292) Cathro, K. J.; Constable, D. C.; Hoobin, P. M. J. Power Sources
1988, 22, 29-57. CR020738U
Chem. Rev. 2004, 104, 4463−4492 4463
Three-Dimensional Battery Architectures

Jeffrey W. Long,*,† Bruce Dunn,*,‡ Debra R. Rolison,*,† and Henry S. White*,§
Surface Chemistry Branch, Code 6170, Naval Research Laboratory, Washington, DC 20375, Department of Materials Science and Engineering,
UCLA, Los Angeles, California 90095, and Department of Chemistry, University of Utah, Salt Lake City, Utah 84112
Received December 4, 2003
Contents 5. Structural Characterization of 3-D Electrode 4478

Architectures
1. Introduction 4463 5.1. Characterization of the Pore−Solid 4478
1.1. Drivers for 3-D Power 4463 Architecture
1.2. What Do We Mean by 3-D? 4465 5.1.1. Physisorption 4479
2. Quantitative Advantages of 3-D Cell Architectures 4467 5.1.2. Small-Angle Scattering Techniques 4479
2.1. Overview of Length Scales in 3-D Battery 4467 5.1.3. Electron Microscopy 4479
Designs 5.1.4. Scanning Probe Microscopy 4480
2.2. Quantifying the Advantages of 3-D 4467 5.2. Corroboration of Electrochemical Processes 4481
Architectures with Spectroscopic Techniques
2.3. Current Distribution in 3-D Batteries 4469 6. Fabrication En Route to 3-D Integration of 4483
2.3.1. Interdigitated Electrode Arrays 4469 Architectural Components
2.3.2. Plate, Tubule, and Aperiodic 3-D 4470 6.1. Micromachining 4483
Architectures 6.1.1. Carbon MEMS (C-MEMS) 4483
2.4. Overall Design Issues 4470 6.1.2. Micromachining of Silicon Molds 4484
2.5. Cathodes and Anodes Separated by 4470 6.2. 3-D Fabrication Based on 2-D Structures: 4485
Nanometer-Wide Electrolytes Origami
3. Prior Examples of Small Power 4471 7. Advances toward Integration of Active 4486
3.1. Scope of Small Power 4471 Components into a 3-D Battery
3.2. Miniaturizing Microreactors and Fuel Cells 4471 7.1. Present Status 4486
3.2.1. Compact Mixed Reactors 4471 7.2. Electrodeposition of DielectricssSeparators 4486
3.2.2. Microfabrication Approaches for Small 4472 and Cation-Conductive Electrolytes
Power 7.3. Hybrid Polymer−Aerogel Nanoarchitectures 4487
3.2.3. Membraneless Opportunities with Laminar 4472 as Electrochemical Capacitors
Flow in Microfluidic Reactors 7.4. Solid-State Electrical Characterization of 4487
3.2.4. Power in Vivo and Biologically Derived 4472 Conformal, Ultrathin Polymer Dielectrics
3.3. Thin-Film Batteries 4473 7.5. Ultrathin Polymer ElectrolytesLeakage 4488
3.4. Generating Power from Ambient Sources 4474 Currents
3.5. Hybrid Micropower 4474 7.6. Last Step: Add an Interpenetrating Anode 4488
4. Component Design for 3-D Battery Structures 4474 7.7. Packaging of 3-D Batteries 4489
4.1. Thinking Like Architects 4474 8. Conclusions 4489
4.2. Porous Membranes as Structure-Directing 4475 9. Acknowledgments 4489
Templates 10. References 4489
4.3. Template Synthesis of Ordered Macroporous 4476
Solids with Colloidal Crystals
4.4. Template Synthesis of Mesoporous Solids 4476 1. Introduction
with Supramolecular Assemblies
4.5. Hierarchical Designs Based on Templating 4477 1.1. Drivers for 3-D Power
Strategies
4.6. Aerogels and Related 4478 The worldwide thirst for portable consumer elec-
Nanostructures;Aperiodic Pore−Solid tronics in the 1990s had an enormous impact on
Architectures portable power. Lithium ion batteries, in which
lithium ions shuttle between an insertion cathode
(e.g., LiCoO2) and an insertion anode (e.g., carbon),
* To whom correspondence should be addressed. J.W.L: e-mail, emerged as the power source of choice for the high-
jwlong@ccs.nrl.navy.mil; telephone, (+1)202-404-8697. B.D.:
e-mail, bdunn@ucla.edu; telephone, (+1)310-825-1519. D.R.R.: performance rechargeable-battery market. The per-
e-mail, rolison@nrl.navy.mil; telephone, (+1)202-767-3617. H.S.W.: formance advantages were so significant that lithium
e-mail, white@chem.utah.edu; telephone, (+1)810-585-6256. ion batteries not only replaced Ni-Cd batteries but
† Naval Research Laboratory.
‡ UCLA. left the purported successor technology, nickel-metal
§ University of Utah. hydride, in its wake.1 The thick metal plates of
10.1021/cr020740l CCC: $48.50 © 2004 American Chemical Society
4464 Chemical Reviews, 2004, Vol. 104, No. 10 Long et al.
Jeffrey Long (left) was born in Great Falls, Montana in 1970, but he spent Bruce Dunn holds the Nippon Sheet Glass Chair in Materials Science &
most of his early years in Winston-Salem, NC. He received a B.S. in Engineering at UCLA. He received his B.S. at Rutgers University and his
Chemistry with Honors from Wake Forest University in 1992. Working M.S. and Ph.D. degrees at UCLA. He was a staff scientist at the General
with Prof. Royce Murray, he earned a Ph.D. in Chemistry from the Electric Corporate Research and Development Center before joining the
University of North Carolina at Chapel Hill in 1997. His research focuses UCLA faculty in 1980. His research interests concern the synthesis of
on nanostructured materials, particularly hybrid nanoarchitectures for ceramics and inorganic materials and characterization of their electrical,
applications in sensing, separations, and electrochemical energy storage electrochemical, and optical properties. A continuing theme in his research
and conversion. is the use of sol−gel methods to synthesize materials that incorporate
Debra Rolison (right) was born in Sioux City, Iowa in 1954. She received specific dopants and are capable of developing unique microstructures
a B.S. in Chemistry from Florida Atlantic University in 1975 and a Ph.D. and properties. The areas presently being studied in his group include
in Chemistry from the University of North Carolina at Chapel Hill in 1980 biosensors, intercalation compounds, aerogels, and organic/inorganic hybrid
under the direction of Prof. Royce W. Murray. She joined the Naval materials.
Research Laboratory as a research chemist in 1980 and currently heads
the Advanced Electrochemical Materials section. She is also an Adjunct
Professor of Chemistry at the University of Utah. Her research at the
NRL focuses on multifunctional nanoarchitectures, with special emphasis
on new nanostructured materials for catalytic chemistries, energy storage
and conversion, biomolecular composites, porous magnets, and sensors.
traditional batteries gave way to lithium ion cells in

which the anode, separator/electrolyte, and cathode
were stacked, spiral wound, or folded.2 Although the
materials in today’s batteries differ from those of the
displaced technologies and the energy density of the
electrochemical cell is decidedly larger, the basic two-
dimensional (2-D) character and layer-by-layer con-
struction of the cell remains intact. The next gen-
eration of lithium ion-polymer electrolyte batteries
will offer improvements in gravimetric and volumet- Henry White was born in Chapel Hill, North Carolina in 1956. He received
ric energy densities (W h g-1 and W h L-1, respec- a B.S. in Chemistry from the University of North Carolina in 1978 and a
tively), but the same configurations will prevail.3 Ph.D. in Chemistry from the University of Texas at Austin in 1982. He
Another massive market force is expected to de- held a postdoctoral appointment at the Massachusetts Institute of
velop over the next decade in the area generally Technology from 1983 to 1984 and was on the faculty of the Department
of Chemical Engineering and Materials Science at the University of
known as microelectromechanical systems or MEMS.4 Minnesota from 1984 to 1993. He is currently a Professor of Chemistry
This field grew out of the integrated-circuits (IC) at the University of Utah. His research interests include magnetic-field-
industry, and in less than a decade, MEMS devices induced transport, oxide films and corrosion, iontophoretic transdermal
successfully established high-volume commercial mar- drug delivery, and electrochemical phenomena at electrodes of nanometer
kets for accelerometers and pressure sensors in the dimensions.
automotive industry, ink-jet print heads, and digital
micromirrors for image projection. By 2001, the field. In particular, the issue of how to power MEMS-
global market for MEMS was nearly $14 billion.5 based sensors and actuators has been largely ne-
Developments in the MEMS field seem to be con- glected.11 While a conventional macroscopic power
tinual, and MEMS devices are beginning to make supply could be employed for these devices, the
significant contributions in new arenas, including prospect of powering arrays of micromachined sen-
fluidics, wireless communications, sensors, and sors and actuators is likely to lead to interconnection
optics.6-9 The rapid expansion of MEMS into new problems, cross-talk, noise, and difficulty in control-
areas is due to the development of surface microma- ling the power delivered. In contrast, the complexity
chining techniques, which involve photolithographic of power delivery is reduced if one were to design the
patterning, deposition, and selective etching of mul- power supply to be of the same dimensional scale as
tilayered films to form device structures.10 the sensors and actuators. Specifically, site-specific
There is tacit recognition that power is a vital issue power can be realized, and improvements in noise
for the continued development of the MEMS device and power efficiency may be achieved.
Three-Dimensional Battery Architectures Chemical Reviews, 2004, Vol. 104, No. 10 4465
battery is able to supply 2 J mm-3. The reality,

however, is that thin-film batteries are 2-D devices.
Such batteries are necessarily thin in order to
prevent power losses typically associated with the
slow transport of ions, but a practical limitation
arises from the fracture of cathode films that get too
thick (>2 µm), which leads to low capacity per square
millimeter. With a typical device thickness on the
order of 15 µm, a thin-film battery requires a signifi-
cant amount of area in order to supply 2 J mm-3.
The real question becomes whether a thin-film
Figure 1. Smart dust mote and its components: Micro- battery can supply what the smart dust mote re-
fabricated sensors, optical receiver, signal-processing and quires within the constraints of the areal “footprint”
control circuitry; the power source consists of a solar cell available for the battery on the smart dust motes1
and a thick-film battery. (Derived with permission from J of energy over 1 mm2 of “real estate.” The tradi-
ref 16. Copyright 2001 IEEE)
tional yardsticks by which battery energy and power
are normalized are weight and volume [i.e., gravi-
The power needs for MEMS devices are diverses metric and volumetric energy (and power) densities].
and batteries may not be the best choice to provide These metrics are no longer sufficient when one
power to systems based on various types of MEMS considers portable power for small devices: such
drives. For example, magnetic drives operate at less devices have limited area available to integrate
than 1 V, but they require generating hundreds of components into a system. For this reason it is
milliamperes, which becomes a difficult challenge for important to establish the effective area of the power
batteries sized on the subcentimeter scale. The source, its footprint, and to normalize the available
required micro- to nanoampere current levels for energy (and/or power) to this area.
electrostatic and piezoelectric MEMS are feasible for
The energy per unit area as reported for several
batteries, but the tens to hundreds of volts that are
lithium thin-film batteries ranges from 0.25 to ∼2 ×
needed will present difficulties for batteries with
10-2 J mm-2.15 Thus, thin-film batteries, despite their
nominal voltages of 3 V. However, there may be a
excellent energy per unit volume, fall far short of
niche for batteries that would be used to power 10-
being able to power a smart dust mote for 1 day. If
15 V drives.
the areal footprint were made 100 times larger (at 1
A more reasonable application for batteries is in cm2), the thin-film approach would be acceptable. The
the area of MEMS-based sensing. In one particularly consequences of the 2-D nature of thin-film batteries
exciting direction, batteries supply power to autono- are easily overlooked. The calculation by Koeneman
mous nodes, with each node containing one or more et al. ignored the 2-D character of thin-film batteries
sensors, along with powering the computation and when they concluded that these batteries could carry
communication capabilities necessary to relay the out some 60 000 actuations of a “smart bearing”.11
findings of each node.12 One approach, “smart dust”, When one considers the actual area available for the
represents an excellent example where integrated power source on the device, only about 1200 actua-
batteries may contribute; the basic device is shown tions are possible.
in Figure 1. The device integrates into a single
Smart dust motes are not just intriguing lab toys
package, MEMS components (sensors, beam steer-
but successfully demonstrated devices.16 However,
ing), optical communication components (optical re-
because of the inability to obtain scale-appropriate
ceiver, semiconductor laser diode), signal-processing
lithium ion batteries, the first devices tested were
and control circuitry, and a power source based on
powered with hearing-aid batteries. The example of
batteries and solar cells, in a volume on the order of
the smart dust mote illustrates a critical issue
a cubic millimeter.
concerning all small power designs and their effective
A calculation of the power requirements of the integration on-board the device. To power devices
smart dust mote underscores our point that the with limited real estate and maintain a small areal
present generation of batteries cannot effectively footprint, batteries must somehow make good use of
power this device. Thin-film batteries are among the thickness. Three-dimensional configurations offer a
most advanced of the lithium battery systems, with means to keep transport distances short and yet
a capability to scale down to dimensions on the same provide enough material such that the batteries can
order of magnitude as the cubic millimeter of the dust power MEMS devices for extended periods of time.
mote.13 The energy density for the thin-film system The detailed calculations discussed in section 2 show
is ∼2 J mm-3,14 which matches or exceeds standard that batteries configured as 3-D structures can
lithium ion systems, such as those that power laptop readily achieve the 1 J mm-2 areal footprint required
computers. A key design requirement for the smart to power the smart dust mote.
dust mote is that the power consumption cannot
exceed 10 µW. If the dust mote uses this power
continuously over a day, it will consume ∼1 J.
1.2. What Do We Mean by 3-D?
Can a thin-film battery supply the 1 J per day of Lithium ion batteries use insertion processes for
energy necessary to power a smart dust mote? At first both the positive and negative electrodes, leading to
glance, it would appear that there should be no the term “rocking chair” battery.1 The resulting
problem; the device consumes 1 J mm-3 and the transport of Li ions between the electrodes, usually
arranged in a parallel-plate configuration, is 1-D in

nature. To minimize power losses resulting from slow
transport of ions, the thickness of the insertion
electrodes, as well as the separation distance between
them, is kept as small as possible. This approach may
appear counterintuitive in the effort to produce a
useful battery, because reducing the thickness of the
electrode results in lower energy capacity and shorter
operating time. Thus, battery design always trades
off between available energy and the ability to release
this energy without internal power losses.
In recent years there has been the realization that
improved battery performance can be achieved by
reconfiguring the electrode materials currently em-
ployed in 2-D batteries into 3-D architectures. Some
of the envisioned approaches are reviewed in more
detail in section 2. The general strategy of this
approach is to design cell structures that maximize
power and energy density yet maintain short ion
transport distances. While many possible architec-
tures can achieve this goal, a defining characteristic
of 3-D batteries is that transport between electrodes
remains one-dimensional (or nearly so) at the micro-
scopic level, while the electrodes are configured in
Figure 2. Examples of prospective 3-D architectures for
complex geometries (i.e., nonplanar) in order to charge-insertion batteries: (a) array of interdigitated
increase the energy density of the cell within the cylindrical cathodes and anodes; (b) interdigitated plate
footprint area. A 3-D matrix of electrodes (in a array of cathodes and anodes; (c) rod array of cylindrical
periodic array or an aperiodic ensemble) is necessary anodes coated with a thin layer of ion-conducting dielectric
to meet both the requirements of short transport (electrolyte) with the remaining free volume filled with the
lengths and large energy capacity. Improvements in cathode material; (d) aperiodic “sponge” architectures in
which the solid network of the sponge serves as the charge-
energy per unit area and high-rate discharge capa- insertion cathode, which is coated with an ultrathin layer
bilities are two of the benefits that may be realized of ion-conducting dielectric (electrolyte), and the remaining
for these 3-D cells. free volume is filled with an interpenetrating, continuous
Perhaps the most obvious 3-D design is that anode.
consisting of interdigitated electrodes shown in Fig- important consideration, because prospective syn-
ure 2a. This prototype configuration is used in thetic techniques, such as templating through me-
modeling studies, the results of which are reviewed soporous membranes, do not necessarily produce
in section 2. The anode and cathode consist of arrays periodic arrays.
of rods separated by a continuous electrolyte phase. A completely aperiodic 3-D battery configuration
The spatial arrangement of the anode and cathode is the “sponge” approach, where the electrolyte layer
arrays determines the current-potential distribu- is formed around a random 3-D network of electrode
tion. Clearly, the short transport distances lead to a material (Figure 2d). This design strategy also rep-
much lower interelectrode ohmic resistance as com- resents a concentric configuration in that the elec-
pared to traditional planar battery configurations. trolyte envelops the electrode material while the
A central feature for the interdigitated configura- other electrode material fills the mesoporous and
tion is that the arrays need to be periodic. A variation macroporous spaces and surrounds the electrolyte.
on this approach is to use interdigitated plates rather Short transport-path characteristics between the
than rods (Figure 2b), which is analogous to stacking insertion electrodes are preserved with this arrange-
2-D batteries with parallel connection. ment. In contrast to the other 3-D designs, all battery
There is no particular reason that the electrolyte componentssanode, cathode, and electrolytesare
must serve as the continuous phase in 3-D battery continuous throughout the sponge structure.
architectures. Another 3-D design utilizes a concen- The above discussion provides the context for 3-D
tric arrangement where the rod array is composed batteries. That is, there are a variety of small power
of one of the active electrode materials and is then applications, typified by MEMS devices, which the
coated by an electrolyte layer. The other electrode most advanced, 2-D lithium battery systems are
material then fills the remaining free volume and unable to satisfy. The inability to provide sufficient
serves as the continuous phase, as shown in Figure power is because of configuration and not because of
2c. As was discussed for the above battery configura- intrinsic energy density. Three-dimensional designs
tions, short transport distances between the insertion offer the opportunity to achieve milliwatt-hour ener-
electrodes lead to low ohmic resistance. Although a gies in cubic millimeter packages and, more impor-
periodic arrangement is shown in Figure 2c, it is not tantly, with square millimeter footprints. While such
clear at the present time whether this is beneficial power sources may not influence the enormous com-
or not with respect to the current-potential distribu- mercial markets in cell phones and laptop computers,
tion. Relaxing the periodicity condition may be an they are certain to impact emerging markets where
integrated power is required for communication,

sensing, and networking.
This paper discusses the prospects for creating 3-D
architectures for batteries. We next introduce calcu-
lations for the 3-D interdigitated array battery
(Figure 2a), which will illustrate the design consid-
erations under which the 3-D configuration leads to
better performance than the conventional 2-D one.
Other sections then discuss the prior art in small
Figure 3. 2-D parallel-plate and 3-D interdigitated-array
power, followed by the materials, synthetic, and batteries.
fabrication approaches required to achieve 3-D de-
signs. Finally, we review the present status of the tion is nonuniform.19 This nonuniformity can be
systems most likely to demonstrate true 3-D battery readily recognized by inspection of Figure 2a and
operation. realizing that the distance between the anodes and
cathodes is variable along the circumference on any
2. Quantitative Advantages of 3-D Cell individual electrode. A quantitative description of
Architectures this issue is discussed below.
The concentric tubule arrays and aperiodic con-
2.1. Overview of Length Scales in 3-D Battery tinuous sponge architectures are based on conformal
Designs deposition methods, in which layers of electrolyte and/
or electrode materials are sequentially20-23 or simul-
The four architectures introduced in Figure 2 have
taneously assembled within an array of pores in a
wide-ranging length scales and geometries that are
membrane (Figure 2c) or in an aperiodic array
largely dictated by their respective methods of fab-
(Figure 2d). Film deposition methods (see section 7)
rication. These length scales and geometries will
can yield electrode and electrolyte films of nanometer
determine the performance characteristics of 3-D
thicknesses, and in fact, a prototype of the aperiodic
batteries based on these architectures. Thus, before
architecture has been recently reported.20-23 Thus it
attempting to quantify some of the advantages (as
is possible to envision batteries soon in which the
well as the disadvantages) of 3-D cells, a brief
cathode and anode are separated by 10 nm. However,
overview is presented of the methods of fabrication
at these distances, the physical structure of the
currently employed, with emphasis on the resulting
interface must be exceedingly stable to prevent
geometrical and scale differences. Detailed descrip-
shorting of the electrodes. In addition, electrostatic
tions of the fabrication procedures are presented in
interactions between the cathode and anode (i.e.,
section 4.
“overlapping” double layers) and exceedingly fast
At present, only the individual component arrays
transport rates within the separator will undoubtedly
(i.e., individual cathode or anode arrays) of the
influence the current-voltage curves in complex
periodic interdigitated electrode cell (Figure 2a) have
ways that are just now being investigated.24 One key
been fabricated using lithographic methods. The
advantage of these 3-D architectures (and unlike the
results of these limited efforts, however, provide a
periodic interdigitated array) is that the current
rough guide to what is possible in battery design
density across the electrode surfaces is uniform.
within the microlithographic laboratory. Typically,
the electrodes are cylindrical rods with diameters
ranging from 5 to 100 µm and lengths from 10 to 200
2.2. Quantifying the Advantages of 3-D
µm. For instance, Madou and co-workers have fab-
Architectures
ricated an array of cylindrical carbon electrodes with As discussed in section 1, the key advantage
∼10-µm diameter and ∼200-µm length.17 Litho- associated with the proposed 3-D battery structures
graphic fabrication methods also allow other geom- is the ability to achieve large areal energy capacities
etries beyond cylinders to be considered (e.g., trian- without making sacrifices in power density that may
gular rods). It is most probable that the use of current result from slow interfacial kinetics (associated with
lithographic methods will continue to yield structures a small electrode area-to-volume ratio) and/or ohmic
with lateral dimensions (i.e., electrode diameter and potential losses (associated with long transport dis-
electrode-to-electrode spacing) on the micrometer tances). This argument is now quantitatively dem-
length scales, with more emphasis placed on increas- onstrated in the following paragraphs for the 3-D
ing the length of the electrode, L, to achieve high interdigitated electrode array shown in Figure 2a.
aspect ratios. This strategy yields the high areal To facilitate a demonstration of the advantages of
energy capacity that drives the interest in 3-D the 3-D architecture, we quantitatively compare
designs (vide infra). Other periodic arrays, such as metrics related to performance (e.g., areal energy
those based on the use of origami to “unfold” an array capacity, active surface area) of a conventional 2-D
of lithographically designed cathodes and anodes on parallel-plate design with the 3-D interdigitated
a base support18 and of arrays of plates (Figure 2b), array cell (Figure 3). We assume a thin-film 2-D
are also envisioned at the micrometer length scale. battery that comprises a 1-cm2-area anode and
Smaller lateral dimensions will likely be obtained in cathode, each 22.5-µm-thick and separated by a 5-µm-
the future with advances in lithographic methods. thick electrolyte. The total volume of electrodes and
One disadvantage of all periodic interdigitated separator is 5 × 10-3 cm3 (the cell housing is ignored
electrode cells is that the primary current distribu- for simplicity, but is expected to be a comparable
Figure 5. An increase in the height of the interdigitated

Figure 4. Square array of interdigitated anodes and 3-D battery, L, results in increased areal energy capacity
cathodes. and electrode area, without an increase in transport
distances.
percentage of the total volume in both 2-D and 3-D
designs).
It is relatively straightforward to show that a
corresponding 3-D square-array battery (Figure 4)
with the same total volume (i.e., 5 × 10-3 cm3),
constructed from 5-µm-radius cathode and anodes
that are separated (center to center) by 10 µm,
contains ∼39% of the energy capacity of the thin-film
design. Clearly, the lower energy capacity is due to
a higher percentage of the total volume being oc-
cupied by the electrolyte. While the energy capacity Figure 6. Dependence of electrode utilization on electrode
of the 3-D design varies with the shape and arrange- conductivity (σ) and ion diffusivity (D). L and r are the
length and radius of the electrodes, respectively.
ment of the electrodes, it will always be lower than
that of the 2-D design for equal total volumes.
There are several intriguing advantages of the 3-D conductivity of the electrode materials (σ) and the
design that are not reflected in the above numbers. ionic conductivity of the electrodes and electrolyte,
For instance, the active cathode and anode surface in addition to the electrode geometry, as indicated
areas in the 3-D design are 3.5 cm2 each, significantly in Figure 6.
larger than the 2-D design (1 cm2). This difference The dimensionless number U ) (r2/L2)(µ/σ)(1/C),
favors the 3-D design through a reduction in inter- where r is the radius of the electrode, µ is ionic
facial kinetic overpotential, which is inversely pro- mobility of cations (Li+), and C is the volumetric
portional to the electrode area. In addition, the energy capacity (C/cm3), is a quantitative measure
distance that ions must be transported in discharging of uniformity of current across the 3-D electrode
the 2-D battery is 350% larger than in the 3-D design. surfaces and, thus, whether the electrode material
Thus, in principle, the 3-D design is significantly less is uniformly utilized during cell charging and dis-
susceptible to ohmic losses and other transport charging. Decreasing U corresponds to a more uni-
limitations. To achieve equal transport length scales form current distribution along the length of the
in the 2-D design (i.e., by decreasing the electrode electrode. This scenario is a desired one, as long as
thickness to 5 µm) would require a 330% increase in the decrease in U results from high electronic con-
the areal footprint in order to maintain equal cell ductivity in the electrodes (i.e., large σ), rather than
volume, a significant disadvantage in employing low ionic conductivity in the electrolyte (i.e., small
these devices in MEMS and microelectronic applica- µ). Increasing U corresponds to a more nonuniform
tions. discharge of the electrodes, which may result in
While the above comparison of 2-D and 3-D designs underutilization of the electrode materials during
indicates that the 3-D battery has inherently lower rapid discharge, as well as increasing stress along
energy capacity per total cell volume, in fact, the the length of the electrodes. Similarly, U ) (w2/h2)-
capacity of the 3-D design can be increased without (µ/σ)(1/C) for a rectangular electrode in the interdigi-
limit by increasing L, without sacrificing the small tated plate design (Figure 2b), where w is the
areal footprint or high power density. As shown in electrode thickness and h is the electrode height
Figure 5, for the same areal footprint, i.e., 1 cm2, the measured from the base. Clearly, numerical values
above square array 3-D design with L ) 500 µm has of U will depend on electrode geometry and materials.
a capacity that is 350% larger that the 2-D design. The arguments developed in this section apply to
Such a microbattery would contain 222 222 cathodes all of the potential 3-D designs shown in Figure 2.
and 222 222 anodes, with ∼35 cm2 each of active In each design, an increase in areal energy capacity
cathode and anode area! can be obtained by simply increasing L, without any
Clearly L cannot be increased without limit, as the loss of power density due to interfacial kinetics or
ohmic resistance of the electrodes will become suf- slow ion transport. Similar to the interdigitated array
ficiently large to offset the advantages of increased cell, this advantage is eventually limited by the
areal capacity. While this problem has not been electronic conductivity of the electrode materials.
treated systematically, it is clear that the optimized It is worth noting that the 3-D interdigitated plate
value of L will be determined by the electronic design of Figure 2b can be realized by stacking 2-D
Figure 7. Top panels: Schematic diagram of 3-D cylindri-

cal battery arrays in parallel row (left) and alternating Figure 8. (Top panel, left) Schematic diagram of hexago-
anode/cathode (right) configurations. Middle panels: Iso- nal 2:1 cathode/anode battery array. (Top panel, right)
potential lines between cathode (C) and anode (A) for unit Schematic diagram of triangular battery array. (Middle
battery cells. Bottom panel: Current densities (in arbitrary panel) Isopotential lines between cathode (C) and anode
units, a.u.) at the electrode surfaces as a function of the (A) for unit battery cell. (Bottom panel) Current density
angle θ (see middle panel for definition of θ). The area of at the cathode and anode surfaces, plotted on the same a.u.
the cathodes and anodes is equal throughout the diagram. scale used in Figure 7. (Reprinted with permission from
(Reprinted with permission from ref 19. Copyright 2003 ref 19. Copyright 2003 Elsevier.)
Elsevier.)
anodes are placed on a rectangular grid. This ar-
thin-film batteries. However, to maintain the key rangement of electrodes results in high current flow
advantage of low ohmic potential losses, it is neces- between each neighboring cathode/anode pair (i.e.,
sary for the thickness of the film electrodes to be at θ ) 0°), with a relatively steep decrease (∼40%)
greatly reduced. Currently, low-cost manufacturing in current between adjacent cathodes or anodes (θ )
techniques limit film electrodes to thicknesses greater 90°).
than about ∼20 µm. A significantly better design based on the rectan-
gular grid is the alternating cathode/anode configu-
2.3. Current Distribution in 3-D Batteries ration in which each anode is surrounded by four
2.3.1. Interdigitated Electrode Arrays nearest neighbor cathodes (and vice versa). In this
geometry, the higher number of nearest neighbor
All interdigitated electrode arrays suffer from a electrodes of opposite polarity permits a significantly
nonuniform primary current distribution. Figures 7 more uniform primary current density at each elec-
and 8 show examples of the potential and primary trode. However, even in this improved geometry, the
current distributions for representative interdigitated current varies by 20% along the electrode circumfer-
microbattery designs, computed using finite element ence, a limitation that may not be tolerable in some
simulation,25 with the assumption of a uniform cells. The primary current uniformity may be im-
electrolyte conductivity. To allow comparison of the proved by increasing the ratio of the electrode grid
current densities between different battery designs, spacing to the electrode radius in this cell, but at the
all simulations assume identical values of the voltage sake of reducing power density.
between cathode and anode. Current densities are Unlike the conventional battery, a 3-D interdigi-
plotted in the same arbitrary units in each figure, tated array need not contain equal numbers of anodes
allowing direct comparison of the relative power and cathodes. Indeed, there may be situations where
output of different interdigitated geometries. Isopo- battery design is optimized by using unequal number
tential lines within the “unit cell” of each interdigi- densities of cathodes and anodes in order to balance
tated design are also presented. The lengths of the the capacities of the active materials and the kinetics
cathodes and anodes, L, are assumed to be suf- of the charge-transfer reactions. The left-hand side
ficiently long to ignore end effects.19 of Figure 8 shows an example of an interdigitated
Figure 7 illustrates the sensitivity of the current design utilizing twice as many cathodes as anodes.
distribution to electrode placement for two similar Here, each anode is surrounded by six cathodes,
interdigitated designs. In the parallel row design, providing a relatively uniform current density on the
alternating rows of cylindrically shaped cathodes and anode while sacrificing current uniformity at the
cathodes. This design might be useful in a 3-D

microbattery where a uniform current density is
critical at one electrode (e.g., an insertion electrode).
Modern microlithography allows for the fabrication
of essentially any envisioned electrode and cell
geometry. Specifically, electrodes need not have the
cylindrical shape considered in the previous ex-
amples. For example, the close-packed array of
triangular cathodes and anodes shown on the right-
hand side of Figure 8 would be expected to yield
greater cell capacity and increased power. The tradeoff
of this design, obviously, is the highly nonuniform
primary current. Such a geometry may be appropri-
ate in a situation where the net current is limited
by electron-transfer kinetics, and thus the current
distribution is uniform across the electrode surface
regardless of the electrode geometry.
It is clear that while significant increases in both
power and areal energy capacity are obtainable from
3-D interdigitated arrays relative to conventional 2-D
batteries, the inherent difficulty in achieving a
uniform current distribution may limit some devices.
However, interdigitated electrode geometries and Figure 9. (a) Schematic of a closed cell containing LiCl.
configurations not considered above are likely to yield The number of ions (Li+ and Cl-) is assumed to be constant.
current distributions significantly better than the (b) The dependence of the potential profile across the two-
examples described here. ion cell as a function of the electrolyte thickness, d, and
ion concentration.
2.3.2. Plate, Tubule, and Aperiodic 3-D Architectures
will be operative when the electrodes are placed in
The issue of the nonuniform current density is not such close proximity to one another.
important for the plate, tubule, and aperiodic archi- For simplicity, we consider a closed cell such as
tectures shown in Figure 2. The current density in that shown in Figure 9a, which contains LiCl as the
these architectures will be perfectly uniform due to electrolyte. Assuming that the cathode and anode are
the 1-D nature of transport between anode and separated by 10 nm (an experimentally achieved
cathode. Radial transport occurs in the tubular and separation distance), it is instructive to first consider
aperiodic architecture, while planar transport is any direct physical or chemical interactions between
operative in the plate architecture (neglecting end the electrodes. Clearly at 10 nm, electron tunneling
effects). between the surfaces will be negligibly slow, the
decay length of tunneling being ca. 1 Å.26 Thus, the
2.4. Overall Design Issues electrodes are not electronically coupled, and normal
Modeling the overall current-voltage behavior of redox reactions at the cathode and anode are still
3-D cells is a complex function that depends on many necessary for battery operation. At shorter distances,
factors (ionic and electronic conductances, interfacial e.g., 1 nm, the electrodes would spontaneously dis-
reaction kinetics, charge capacity, etc.). Because of charge via direct tunneling between the cathode and
the wide range of possible (and mostly unexplored) anode, a process equivalent to shorting the battery.
architectures, this will be a rich field of study for Thus, there is a theoretical lower limit on the
electrochemical engineers and will likely lead to new thickness of the battery separator.
design concepts in battery technology. Interdigitated Although the cathode and anode are not in elec-
electrode arrays are highly susceptible to stress due tronic communication at 10 nm, their presence is
to both the nonuniform current density and finite known to each other through the electrical fields that
electronic conductivity of the electrodes. Thus, a originate from their respective surface charges. As
delicate balance of the geometrical length scales shown in Figure 9b, the overlap of the electrical
(electrode radius and length, and the spacing be- double layers is more significant as either the dis-
tween electrodes) and materials properties will be tance between the electrodes, d, is reduced or the
required in optimizing any 3-D battery. concentration of the electrolyte is decreased. The role
of the electrolyte concentration is quantitatively
2.5. Cathodes and Anodes Separated by expressed through the Debye length, κ-1,26 which can
Nanometer-Wide Electrolytes be estimated for a 1:1 electrolyte at room temperature
using the expression, κ-1 (nm) ∼0.3/[LiCl]1/2, where
As noted in section 3.1, the fabrication of cells in [LiCl] is the molar concentration (M) of the electro-
which the cathode and anode are separated by a lyte. In essence, κ-1 represents how well the electro-
conformal thin-film electrolyte, perhaps a few tens lyte shields the solution interior from the electrical
of nanometers in thickness, is on the near horizon. charge of the electrode surfaces. A 1 M LiCl solution
Thus, it is interesting to consider potential phenom- corresponds to κ-1 ) 0.3 nm, while a 10 mM solution
ena, not normally considered in battery design, that corresponds to a κ-1 of 3 nm. As a rule of thumb, the
electric potential drops to about 5% of its surface Table 1. Types of Electron-Producing Power Sources
value at a distance from the surface of ∼3κ-1 (i.e., thermal/nuclear/
∼9 nm in a 10 mM LiCl solution). Thus, two elec- electrochemical mechanical-to-electric
trodes separated by 10 nm in a 10 mM LiCl solution batteries thermoelectrics
will clearly be in electrostatic contact through their fuel cells, biofuel cells, pyroelectrics
overlapping double layers. semi-fuel cells
The consequence of overlapping double layers on supercapacitors, thermionics
ultracapacitors
battery operation has only been recently considered, photovoltaics β-cells (nuclear “solar” cells)
and our understanding of this effect is rudimentary.24 harvesting adventitious harvesting adventitious
The electric fields will almost certainly have an effect energy/fuel energy/work
on the transport of ions through the electrolyte, as radioisotope thermoelectric
migration will be important in the strong fields of generators (RTGs)s
R- or β-emitting nuclear sources
the double layers, which can approach or exceed 106
V cm-1. Whether transport of Li+ is impeded or
enhanced by the field will depend on signs of the sources, not all of which are electrochemical. The
electrode charges (determined by the potentials of importance of integrating nanotechnology to improve
zero charge, an ill-defined quantity for most battery these power sources has been discussed, especially
electrode materials) and whether the battery is being for portable applications.30 Some of these sources are
charged or discharged. Modeling of this phenomenon described in greater detail in this section and em-
requires a simultaneous consideration of transport phasize those approaches that have demonstrated
(e.g., Nernst-Planck equation), electrostatics (Pois- power production in centimeter dimensions or smaller
son equation), and statistical thermodynamics (Boltz- sizes. The hybrid micropower supplies discussed at
mann equation), similar to recent modeling of the the end of this section begin to address the issue of
influence of double-layer structure on transport at achieving energy and power capabilities in small
individual nanometer-scale electrodes27 and Levich’s footprint areas.
treatment of the dynamic diffuse layer at macroscopic
planar electrodes.28 3.2. Miniaturizing Microreactors and Fuel Cells
At small electrode separations, the number of ions
in the electrolyte between the electrodes is quite Miniaturization of electrochemical power sources,
small. For instance, the average nearest neighbor in particular batteries and fuel cells, has been
distance between cations in 10 mM LiCl solution is described as a criticalsbut missingscomponent in
∼2 nm. Thus, on average, only a few Li+ ions are transitioning from in-lab capability to the freedom
located between cathode and anode at any position of autonomous devices and systems.29,30 In top-down
on the electrode surfaces, and these few ions carry approaches, macroscopic power sources are scaled to
the current at that surface position. The number of the microlevel usually by the use of fabrication
ions is so small in this situation that the battery can methods, often in combination with new materials.
be well described as a capacitor, with the separator Power generation schemes that can themselves be
being a nonionic dielectric material, resulting in a microfabricated are particularly appealing, as they
linear potential drop between the electrodes (Figure can lead to a one-stop fabrication of device/machine
9b). In addition, even if migration from the electrical function with an integrated power source.
double layers is neglected, the small separation 3.2.1. Compact Mixed Reactors
distance between the electrodes will result in enor-
mous diffusional fluxes. If sufficiently high, these One miniaturization challenge common to fuel cells
fluxes can lead to breakdown of electroneutrality and and batteriessand of critical relevance in designing
formation of a space-charge layer in the electrolyte.27 microscopically and nanoscopically featured 3-D
Much effort in the future will be required to explore batteriessis scaling down the phase that separates
and quantify these ideas. the anode from the cathode. The separator must
prevent direct electronic contact yet permit electrical
contact (ion flux) between the active electrodes.
3. Prior Examples of Small Power Electrodes in rechargeable batteries can undergo
morphological changes that compromise the integrity
3.1. Scope of Small Power of the whole cell, e.g., by establishing a direct
The development of small power sources for por- electronic path (hard shorts), such as the metal
table electronics goes well beyond miniaturization of needles that grow off of a charged-discharged zinc
batteries and fuel cells.4 Other approaches under electrode and may pierce the separator (glass, poly-
development include miniaturizing combustion en- mer, paper) with potentially pyrophoric consequences
gines and thermoelectrics and harvesting energy (one reason the alkaline cell is a primary, use-once,
from ambient sources such as vibration and temper- cell).
ature differences. Most of these approaches are Fabrication issues can arise when thinning the
directed at consumer products and are not designed separator material or in trying to ensure high-
to save real estate. Nonetheless, it is apparent that quality, preferably pinhole-free coverage of dimen-
decreasing the size of power sources and moving to sionally scaled-down anodes and cathodes. One strat-
micro- and even nanoscale power sources offers a egy that can be used to avoid fabricating a separator
number of opportunities.29 at all, at least in electrolyzers and fuel cells, takes
Table 1 lists examples of electron-producing power advantage of decades of work in compact mixed
reactors (CMR), where the specificity of electrocata- demand of 100 mA cm-2.37 The electrode footprint,
lyzed active electrodes permits intermixing of reac- which was smaller than the silicon wafers, was 22.5
tants (or fuel/oxidant).31 The catalytic specificity mm × 22.5 mm. The same authors have also adapted
bypasses the need for a physical barrier, as long as silicon micromachining to fabricate a twin fuel-cell
the electrodes are electronically isolated. Another stack38 using a previously described flip-flop structure
engineering design feature of these single-chamber sandwiched between silicon wafers again acting as
systems is the hydrodynamic flow of the molecular current collectors and flow distributors.34
reactants. Wainwright et al. recently microfabricated a poly-
In 1990, Dyer applied the CMR concept of hydro- mer-based fuel cell with on-board hydrogen storage.
dynamic control of mixed reactant flow to a thin-film One of the key goals of this work is to provide
planar fuel cell.32 The electrical (i.e., ionic) contact independent sizing of power and energy capacity in
between the fuel cell anode and cathode was main- a passive system requiring no fans, pumps, or exter-
tained by diffusion across/through a hydrous alumi- nal humidification.39 The first performance reported
num oxide film (e50-nm-thick) sandwiched between was for a device built on a ceramic substrate (alu-
the electrocatalytic electrodes, with the top Pt elec- mina), but devices built on silicon and polymeric
trode deposited so as to be sufficiently porous for substrates are also mentioned. The microfabricated
gases to pass through to the alumina membrane. The fuel-cell block is coupled to a metal-hydride storage
specific areal power that was generated with a mixed block, printed with inks of LaAl0.3Ni4.7, which gener-
H2/O2 flow was 1-5 mW cm-2 at a cell voltage of 950 ates hydrogen by desorption when placing a 20 mA
mV. By using a lightweight substrate (Kapton) onto load on the fuel cell.39
which the Pt electrodes and hydrous alumina inter-
layer were deposited as thin films, a power density 3.2.3. Membraneless Opportunities with Laminar Flow in
of 100 W kg-1 was reported. Similar results were Microfluidic Reactors
obtained upon replacing the hydrous alumina mem- One way to ease any difficulties that may arise in
brane with the proton-exchange polymer membrane fabricating a membrane, especially in design con-
Nafion, raising the possibility of fabricating confor- figurations that are not planar, is to go membrane-
mal, “open-face” fuel cells on a flexible substrate. less. Recent reports take advantage of the laminar
flow innate to microfluidic reactors40-42 to develop
3.2.2. Microfabrication Approaches for Small Power membraneless fuel cells. The potential of the fuel cell
The nickel-zinc battery provides a good example is established at the boundary between parallel
of how macroscopic power sources can be miniatur- (channel) flows of the two fluids customarily com-
ized via microfabrication methods. A patterned, side- partmentalized in the fuel cell as fuel (anolyte) and
by-side configuration was fabricated with an electro- oxidant (catholyte). Adapting prior redox fuel cell
plated zinc anode and a NiOOH cathode (total cell chemistry using a catholyte of VV/VIV and an anolyte
area of 1 to 5 mm2).33 A photosensitive epoxy served of VIII/VII,43 Ferrigno et al. obtained 35 mA cm-2 at
as the separator and was also used to define the side 1.1 V (for a power density of 38 mW cm-2) at a linear
walls of the cell. The open-circuit voltage of the cells flow rate of 12.5 cm s-1,44,45 which was comparable
was 1.7-1.8 V. The energy density reported for these to that seen with the standard redox fuel cell reported
Ni-Zn microbatteries (2 J cm-2 at 50 mA discharge) by Kummer and Oei;43 the fuel utilization was,
is similar to that available with thin-film batteries, however, much lower. Luo et al. adapted the enzyme-
although the performance does not achieve the catalyzed redox that powers biofuel cells to create a
targeted energy density of 1 J mm-2.13 The cells do membraneless microfluidic fuel cell using a molecular
generate reasonable power levels: at a 1 mA dis- fuel (1,4-dihydrobenzoquinone) and a laccase-cata-
charge rate, the 2-mm2 cells generate 1-2 mW. As lyzed cathode to reduce molecular oxygen. The maxi-
discussed in section 3.5, the Ni-Zn battery is feasible mum power density reported was 80 µA cm-2 at 150
as the energy-storage component in a hybrid mi- mV (8.3 µW cm-2) and a linear flow rate of 1 cm s-1,
cropower system. but the cell could sustain 0.35 V at lower current
Fuel cells incorporating lithographic methods and densities.46
masking/deposition/etching protocols have been fab-
ricated on Si wafers and thereby satisfy two critical
3.2.4. Power in Vivo and Biologically Derived
needs in a standard fuel cell: collection of electrons One area in which miniaturization has progressed
(current collectors) and controlling the flow field of beyond the size regime of interest to the laptop and
fuel and oxidant.34 Kelley et al. produced a miniature cellular telephone industry is to provide in vivo power
direct methanol fuel cell (DMFC) with a current- in biomedical applications.47 An implantable defibril-
voltage and fuel utilization performance that matched lator battery, which must provide pulse power (on
standard-sized DMFCs prepared in-lab.35,36 A work- demand) with high energy density, was achieved by
ing volume for the miniature DMFC of 12 mm3 was coupling two smaller batteries (Li/MnO2 primary
reported, with an operational performance of 822 W battery plus a Li/iodine cell) with a packaged volume
h kg-1 at 70 °C.35 on the order of 10 mL.48 Such power sources must
In more recent reports, a thin-film fuel cell sand- have low self-discharge rates.
wiched between two silicon wafers that had been In an effort to use biological energy transduction
anisotropically etched to form feed holes and chan- to miniaturize a biofuel cell for in vivo applications,
nels for the reactants (H2 and O2) demonstrated a Heller and co-workers have created membraneless,
stable voltage of 0.75 V over 300 h at a current caseless cells that can function under physiological
conditions.49-54 The current collectors are microfibers the thickness of the actual battery is on the order of
of carbon that are catalyzed with enzymes to oxidize 100 µm because of packaging and the critical need
glucose (via glucose oxidase) and reduce O2 (laccase to protect the components from moisture.
or bilirubin oxidase) and wired to their respective The thin-film secondary lithium battery system
carbon current collectors with osmium-based redox that has progressed the furthest is that developed
polymers. The active area of the most recent minia- at the Oak Ridge National Laboratory.13 The elec-
ture biofuel cell is 0.44 mm2, and this glucose oxidase/ trolyte in this case is a lithium phosphorus oxynitride
laccase-catalyzed biofuel cell produced 0.78 V in a pH (“Lipon”) which is deposited by rf magnetron sput-
5 buffer at a power density of 2.68 µW mm-2.53 The tering, as is the cathode, followed by thermal evapo-
improved cell voltage was obtained by using an ration of the lithium. A number of different cathode
osmium redox polymer with a more reducing poten- materials, including layered transition metal ox-
tial, which lowered the overvoltage necessary to drive ides60,61 and vanadium oxides,62 have been reported
the coupled enzyme/cofactor redox to convert the and, for the most part, excellent results have been
glucose fuel to gluconolactone. obtained. The thin-film batteries achieve operating
Recently a new hybrid power source has been lifetimes of thousands of cycles and routinely operate
reported that couples oxidation at a dye-photosensi- at current densities above 1 mA cm-2. In tailoring
tized nanocrystalline semiconducting SnO2 photo- these batteries for integrated circuit (IC) applications,
anode with the enzyme-catalyzed reduction of O2.55 which involve solder reflow at temperatures of 250
Although miniaturization has not yet been reported °C, this group developed a lithium-free approach
for this new hybrid, the developments already where the anode is formed by electroplating lithium
achieved to miniaturize biofuel cells coupled to those on a copper current collector.63
being developed for charge-insertion oxides should be
The research on thin-film batteries has led to
technically transferable to this system.
considerable understanding of the kinetics and struc-
tural changes occurring during lithium insertion and
3.3. Thin-Film Batteries deinsertion in amorphous and nanocrystalline films.13
Miniaturization of batteries has been an active However, the fact remains that these thin-film bat-
area of industrial research because of the need for teries have limited capacity, because of the thickness
portable power sources for such items as watches, of the cathode, while the overall cell resistance can
hearing aids, and cameras. These miniaturization be larger than desired, because of the cathode/
approaches are based, for the most part, on tradi- electrolyte interface. As a result, cell energies are in
tional battery manufacturing, and it is only within the range of 100-500 µW h cm-2. As discussed
the past decade that researchers have begun to previously, these systems are considerably below the
employ fabrication methods that rely on more ad- 1 J mm-2 values required for powering MEMS
vanced material-processing approaches such as vapor devices.
deposition and solution processing. Lithium batteries, An interesting direction to thin-film battery fabri-
in particular, have received considerable attention cation is the use of forward-transfer methods based
because of their high energy density. Two comple- on laser-assisted direct-write processes. This process-
mentary lithium batteries have emerged: one is ing method is a soft transfer that minimally affects
based on polymer electrolytes; the other is based on the materials being laid down, so it is compatible
inorganic electrolytes.15 The former, which is actually with the transfer of metals, oxides, polymers, and
a thick-film system, is the basis for commercial even liquid and gel electrolytes. The variant known
products that provide milliwatt-hour of energy. Pri- as MAPLE-DW (matrix-assisted pulsed-laser evapo-
mary batteries of the type Li/LixMnO2, marketed by ration-direct write)64 has recently been used to direct-
Panasonic among other companies, are on the order write a lithium ion battery.65 The MAPLE-fabricated
of 0.3-0.5-mm-thick and offer between 10 and 20 mA LiCoO2-carbon/carbon microbattery prototype was
h at 3 V. Secondary (i.e., rechargeable) lithium sealed within trilayer metal-polymer laminate and
polymer batteries are also being developed.56 The cycled in air. A typical microbattery with a footprint
polymer electrolyte in most cases is a gel electrolyte of 4 mm × 4 mm (patterns can be written to 10 µm)
formed by polymers and organic solutions of lithium has a capacity of 155 µA h (at a C/5 rate, where a 1
salts or is a plasticized polymer electrolyte.2,57,58 C rate completely discharges the full capacity of the
Inorganic electrolytes lend themselves much more battery in 1 h) and 100 mA h g-1 of LiCoO2. Because
readily to thin-film batteries. Over a decade ago, carbon was included in the transfer ink with the
Eveready Battery developed thin-film Li/TiS2 second- charge-insertion oxide, the electrode could be depos-
ary batteries, which used a sputtered lithium oxysul- ited in thicker layers than a sputter-deposited thin-
fide glass as the electrolyte,59 while other thin-film film battery without creating severe ohmic losses.
batteries using glassy oxides as the electrolyte con- The resulting prototype approaches the power that
tinue to be developed.15 The fact that these inorganic other thin-film batteries provide but did so with a
electrolytes have low conductivity at room tempera- footprint smaller than 1 cm2. An advantage of such
ture is compensated by the short diffusion length of direct-write fabrication is that it can ultimately use
only a few micrometers. The fabricated batteries are the electronics substrate as part of the power-source
truly thin film as various vapor deposition methods packaging to save weight in the battery system.
are used to fabricate the cathode and anode as well Current work is aimed at coupling Li ion microbat-
as the electrolyte. The total thickness of the active teries with energy-harvesting devices (solar cells, RF
components is in the range of 10-20 µm; however, antennae, etc.) for hybrid power supply applications.
Kushida et al. recognized the limitation of the low establish some operating parameters and fabrication
capacity of thin-film lithium batteries for IC applica- methods. For the most part, these methods are
tions when they demonstrated the operation of arrays expected to provide low power, as the first devices
of thin-film lithium batteries.66 Rather than develop produced an output of 3 mW.
a power generator for an entire chip, this work was
directed at providing local power to individual (or a 3.5. Hybrid Micropower
few) VLSI circuits. The battery took the form of a
3 × 1 array in which three parallel 100-µm-wide In the use of 3-D batteries to power MEMS devices,
stripes of cathode material were crossed by a 200- the tacit assumption is that the battery serves as the
µm-wide stripe of lithium. A spin-on phosphosilicate electrochemical power source. That is, the battery
glass was used as the separator, while the cathode directly supplies all the energy for the device. A
material (LiMn2O4) was deposited by solution pro- hybrid power supply is an alternative concept based
cessing in a trench etched in the silicon wafer to on combining energy conversion and energy storage.11
provide better device definition. The prospect of a In this case, the battery is a component of the power
battery array is an interesting one, and although the supply rather than the sole power source. The bat-
energy per unit area was not very high (∼35 µW h teries are expected to provide power when the energy
cm-2), the concept of supplying localized power at the conversion device is not working or when additional
individual circuit level is intriguing. power is required, i.e., pulse power for communica-
tion. The energy conversion device, which obtains
3.4. Generating Power from Ambient Sources energy from ambient sources, is used to power the
microsystem and charge the batteries.
One important area of MEMS sensor technology Koeneman et al. described a micropower supply
is the development of distributed sensors, some of that was designed to power a “smart bearing”.11 The
which are embedded, remote sensors in buildings and ambient energy, a rotating shaft, was converted using
other structures. Battery replacement is difficulty or a wire coil. Energy storage was accomplished using
impractical in this case and approaches for generat- a thin-film lithium battery. Because of the device
ing power from ambient sources are being pursued. application, a suitable actuation driver was required
The most familiar ambient energy source is solar and to deflect the membrane. Although the device was
the use of photovoltaics is widely used. Solar energy, not built, the analysis established that a hybrid
however, is not always available for embedded sen- power supply could be integrated with a MEMS
sors and for this reason another ambient energy device to provide on-board power. One interesting
source of interest for powering small devices is the point in the analysis is that the authors did not
use of mechanical vibration. Beeby and co-workers consider the 2-D nature of the battery and overesti-
reported the use of an electromagnetic transducer mated the number of actuations by 50 times.
based on the movement of a magnetic pole with A more recent paper considered the use of a hybrid
respect to a coil.67 This electromechanical power power supply for powering autonomous microsen-
generator uses neodymium iron boride permanent sors.70 Such devices are similar to the dust mote
magnets mounted on a spring board inside an induc- pictured in Figure 1 and described in the Introduc-
tive coil. As the mass deflects from the vibration, tion. This hybrid power supply combined a solar cell
there is a varying amount of magnetic flux passing to meet standby requirements and to charge the
through the coil. An electromotive force is induced battery, which was the microfabricated Ni-Zn bat-
(Faraday’s law), the magnitude of which is propor- tery described in section 3.2. The authors also showed
tional to the rate of change of the coil position. that if the battery alone was to serve as the power
Shaking the device causes the magnets to vibrate at source, the footprint would be substantially larger,
322 Hz. The device occupies a volume of 240 mm3 over 12 cm2. The hybrid device was constructed and
and generated a maximum useful power of 0.53 mW, some feasibility experiments were carried out, but no
corresponding to a 25-µm amplitude at an excitation actual devices were powered.
frequency of 322 Hz.
Another power generator that converts vibrational
energy to electrical energy is based on a mass-
4. Component Design for 3-D Battery Structures
spring-resonator structure.68 Laser micromachined
copper springs were fabricated into different patterns
4.1. Thinking Like Architects
to optimize device performance. In this device, gen- Jumping out of Flatland71 to a true third dimension
erators with a volume of 1 cm3 were able to produce (rather than stacking 2-D elements, which seems to
up to 4.4 V peak-to-peak with maximum root-mean- be the microlithography community’s perception of
square power of over 800 µW. The vibration required 3-D) permits the use of hierarchical designs: nano-
to generate this power had frequencies ranging from structured electrode materials, organized into larger
60 to 110 Hz with 200-µm amplitude. The fabricated macroscopic features. One of the most critical com-
generator was able to drive an infrared transmitter, ponents of these new battery designs is an initial
and the authors contend that the device is capable degree of “nothingness”, that is a continuous phase
of driving low-power integrated circuits. of porosity. This open volume is then available to be
Another approach to converting vibrations in the filled with a second phase, for example an infiltrated
environment is the use of an inertial generator based electrolyte or even an opposing electrode structure,
on thick-film piezoelectric materials.69 A prototype forming a true 3-D, intermingled battery assembly.
generator was produced enabling the authors to Following IUPAC conventions, pores are identified
by their respective sizes: macropores (>50 nm),

mesopores (2-50 nm), and micropores (<2 nm).72
Nanopore and nanoporous are not internationally
defined terms, so they are an unknown and often
misleading class of pores.
Chemical methods are ideally suited to construct
textured materials with features of solid and void on
the meso- or nanoscale. The chemistry can include
such techniques as sol-gel synthesis, chemical vapor
deposition (CVD), and electrodeposition, where mo-
lecular or ionic precursors are reacted under con-
trolled conditions to assemble the desired solid
structures. Combining such techniques with structure- Figure 10. Schematic showing the conversion of a porous
membrane into a template in which the pores are filled or
directing templates (porous membranes, colloidal coated to form random ensembles or ordered arrays of
crystals, micelles, etc.) permits the design of hierar- nanometer- and micrometer-scale cylinders or tubes. (De-
chical structures with pore sizes that cover a wide rived with permission from ref 87. Copyright 1997 Elec-
range from mesopores to large macropores. In an trochemical Society.)
alternative to template synthesis, sol-gel methods
are used to assemble 3-D networks of nanoscale structures (Figure 10a). The pores in such mem-
particles into a monolithic wet gel. With careful branes may be either random, as in “track-etched”
processing of these gels to remove the pore-filling polycarbonate, or an ordered array, as in electro-
fluid, aerogels and related structures are rendered chemically prepared alumina membranes. The pores
that exhibit aperiodic, through-connected void net- in these membranes can be as small as 10 nm and
works of mesopores and/or macropores. range upward into the hundreds of nanometers and
An extensive research effort is currently underway micrometers.
to produce common battery materials, including Various strategies are used to produce electrode
carbons and transition-metal oxides, in various meso- structures within the membrane pores, including
structured and nanostructured forms. The basic sol-gel synthesis, CVD, electrodeposition, and elec-
functional requirements for these new materials are troless deposition. With careful control of the syn-
largely the same as for conventional secondary bat- thetic conditions, the pores are either filled com-
tery electrodes: a combination of electrical conductiv- pletely or preferentially coated at the pore walls,
ity and the ability to undergo reversible ion-insertion producing hollow tubes (see Figure 10b). Following
reactions. However, the inherent characteristics of infiltration with the desired electrode material, the
nanostructured electrode materials, an extensive membrane is subsequently removed under conditions
electrode/electrolyte interface and facile transport of that do not disturb the active material, leaving an
ions through the void volume, yield superior electro- array of either solid nanofibers or nanotubes attached
chemical performance relative to conventional bat- to a current collector like the bristles of a brush
tery materials. In practically all of the cases dis- (Figure 11). In this case there is very limited inter-
cussed below, electrode materials designed with 3-D connectedness between the nanofibers, except at the
nanostructured interfaces exhibit significantly better current collector base.
charge-storage properties, particularly under high Using this approach, template-synthesized elec-
rates of charge/discharge, than do conventional forms trodes have been prepared not only from metals, but
of the same composition, and in some instances, the from TiO2,75 V2O5,80,81 LiMn2O4,82,83 SnO2,84-86 TiS2,87
nanostructured materials provide unexpectedly high carbon,88-90 and various conducting polymers.91 The
overall capacity due to the defective nature of nano- final templated feature need not consist of a single
scale solids. material or phase. Martin and co-workers prepared
The following section focuses specifically on those striped metal nanorods of gold and platinum,92 which
design strategies that produce either monoliths or was a concept ultimately extended to the preparation
supported films that may then serve as a solid but of nanoscopic barcodes93 and optical tags.94 Recently
textured platform in the first design step toward a Dewan and Teeters infiltrated alumina membranes
3-D battery architecture. to template a device: a V2O5 xerogel-carbon battery
using a poly(ethylene oxide)-lithium triflate electro-
4.2. Porous Membranes as Structure-Directing lyte.95
Many of the templated electrode ensembles de-
Templates scribed above function as high-performance battery
The use of porous membranes as templates for electrodes in lithium-containing electrolytes.76 The
electrode structures was pioneered by Martin and co- fibrous morphology of these templated electrodes
workers nearly 20 years ago,73 and this approach has reduces the solid-state transport distances for lithium
since been extended to include numerous electrode ions participating in the charge-storage reaction. This
compositions and geometries73-78 and applications feature facilitates not only good cyclability but also
beyond energy storage, including sensing and sepa- rapid charging and discharging. For example, tem-
rations.79 In this approach, chemical and electro- plate-synthesized SnO2 electrodes deliver gravimetric
chemical routes are used to fill in the cylindrical, capacities that are orders of magnitude higher than
uniform, unidirectional pores of a free-standing mem- a thin-film control electrode when discharged at
brane with electrochemically active materials and greater than a 50 C rate. The void spaces between
Figure 12. Inverse opal of vanadium oxide ambigel. The

pores are formed by packing 1-µm styrene beads and
infiltrating a vanadium sol. (Reproduced with permission
from ref 100. Copyright 2002 The Royal Society of Chem-
istry.)
CVD, and electrodeposition, depending on the desired

composition. Removal of the colloidal templating
spheres renders a negative replica (the inverse opal)
structure of the active material, with an intercon-
nected, 3-D array of pores, typically sized in the
hundreds of nanometers.
Figure 11. Electron micrographs of membrane-templated
electrodes: (a) ensemble of rods (Reprinted with permission This general procedure for producing macroporous
from ref 80. Copyright 1999 Electrochemical Society.); (b) solids has recently been exploited to synthesize
array of nanotubes (Reprinted with permission from ref 79. electrode architectures that are targeted for lithium
Copyright 2001 American Chemical Society.) battery applications. Sakamoto and Dunn synthe-
sized inverted opal structures of V2O5 by infiltrating
the individual electrode fibers also provide room for a colloidal crystal template with a vanadia sol-gel
expansion during electrode cycling, which is a par- formulation.100 The resulting macroporous structure
ticular problem for SnO2, because a Sn metal phase is shown in Figure 12. The walls are composed of
alloys with electrogenerated Li metal, with accom- vanadium oxide ambigel, leading to a hierarchical
panying large volume changes. architecture that demonstrated high capacity for
In a related approach, these same porous alumina lithium at high discharge rates. Stein and co-workers
membranes serve as a mask through which O2 have expanded this concept further to produce
plasmas are used to etch underlying carbon films.96,97 macroporous ordered structures of V2O5, SnO2, and
This etching process produces honeycomb carbon LiNiO2.98,101 Hierarchical electrode structures are
structures that are positive replicas of the alumina- generated by this method as well, with the inorganic
membrane mask. This process has successfully pro- skeleton of the 3-D ordered macroporous structure
duced honeycomb structures of both diamond96 and composed of fused nanoscale grains of electrode ma-
graphitic carbon,97 with pore sizes in the carbon terial that are themselves mesoporous. Macroporous
replica around 70 nm. carbon structures have also been produced by infil-
4.3. Template Synthesis of Ordered Macroporous trating organic precursors into the interstitial void
Solids with Colloidal Crystals of silica sphere colloidal crystals, followed by carbon-
ization and removal of the silica template.102-106
The porous membrane templates described above
do exhibit three-dimensionality, but with limited 4.4. Template Synthesis of Mesoporous Solids
interconnectedness between the discrete tubelike with Supramolecular Assemblies
structures. Porous structures with more integrated
pore-solid architectures can be designed using tem- These same general templating strategies are
plates assembled from discrete solid objects or su- extended into the mesopore size regime with the use
pramolecular structures. One class of such structures of self-assembled supramolecular arrays, a topic
are three-dimensionally ordered macroporous (or which has been extensively reviewed elsewhere.107-121
3-DOM) solids,98 which are a class of inverse opal In this approach, ionic surfactants or block copoly-
structures.99 The design of 3-DOM structures is based mers are assembled into micellar aggregates or
on the initial formation of a colloidal crystal composed liquid-crystalline phases. With careful tuning of the
of monodisperse polymer or silica spheres assembled reaction conditions, these assemblies can be orga-
in a close-packed arrangement. The interconnected nized into either highly ordered hexagonal, lamellar,
void spaces of the template, ∼26 vol % for a face- or cubic structures, or more disordered wormlike
centered-cubic array, are subsequently infiltrated structures. The microscale phase segregation in these
with the desired material. systems acts to restrict and direct the growth of an
Strategies for filling the void space of the colloidal inorganic guest phase, which can be incorporated by
crystal utilize sol-gel chemistry, salt precipitation, various methods. With transcriptive synthesis the
The synergistic synthesis method for supramolecu-

lar templating was initially developed for silica solids,
but this chemistry is being expanded to include a
number of transition-metal oxides, including some of
interest for electrochemical applications.112,123 Among
these, mesoporous TiO2 has been the most widely
investigated, primarily due to its potential employ-
ment in photovoltaic cells.124 Combining liquid-phase
titania precursor chemistry and templating strategies
Figure 13. Self-assembled organic template (micellar or has produced mesoporous TiO2 structures, with both
liquid crystalline) that directs the assembly of the inorganic ordered and wormlike pore structures.125-133 Grätzel
phase about the structure-directing organic phase. (Cour-
tesy of Prof. Sarah Tolbert, UCLA.)
and co-workers have reported on the reversible
lithium ion insertion reactions of mesoporous anatase
TiO2.127,134,135 Similar strategies have been used to
generate mesoporous WO3, an important electrochro-
mic oxide, where the high surface areas and meso-
porous pathways facilitate rapid coloration/decolor-
ation responses.136,137
Another route to mesoporous, electrochemically
active phases utilizes electrodeposition methods,
where the presence of surfactants or a liquid-crystal-
line phase at the electrode surface directs the forma-
tion of the growing film. The templated electrodepo-
sition approach was initially developed by Attard et
al. to produce mesoporous platinum films.138 This
protocol was followed to develop mesoporous tin as a
potential anode for Li ion batteries and nickel/nickel
oxide electrodes for aqueous-based batteries.139,140
Related strategies have also been used to prepare
mesoporous V2O5141 and conducting polyaniline (as
nanowire arrays).142 In addition to driving the elec-
trodeposition process, the electrochemical interface
can be controlled to promote the formation of hybrid
assemblies of surfactant-inorganic in dilute surfac-
tant solutions en route to forming mesoporous oxide
films, as demonstrated by Stucky et al.143,144
Figure 14. Nanocast carbon. (Reprinted with permission Templated mesoporous carbon structures are gen-
from ref 122. Copyright 1999 American Chemical Society.) erally produced by first incorporating organic precur-
sors into the void space of a preformed mesoporous
supramolecular assembly is first formed separately silica or aluminosilicate template, either by liquid
followed by infiltration of the guest phase. Synergistic infiltration or CVD methods. Following pyrolysis of
synthesis encompasses the process whereby template organic components, the inorganic templates are
self-assembly and inorganic phase synthesis are removed, rendering an inverse carbon replica. This
combined in a cooperative reaction scheme to produce approach was first reported by Ryoo and co-work-
the mesostructured hybrid (Figure 13). ers122 and has subsequently been extensively adapted
In either the transcriptive or synergistic strategy, to generate various mesoporous carbon struc-
removal of the organic template by extraction or tures.104,145-161 The pore-solid architectures of these
calcination renders the inorganic mesoporous struc- carbon solids can be designed on the basis of the
ture. For synthetic schemes that are not compatible choice of inorganic template from among the many
with the formation of stable template assemblies, an available and range from ordered 2-D structures to
alternative approach is to use a preformed, templated more disordered 3-D structures. These mesoporous
inorganic host, such as mesoporous silica, as a mold carbons are promising candidates as electrochemical
to “nanocast” the desired material as an inverse materials in double-layer capacitors. The mesoporous
replica of the host, such as that seen in Figure 14.122 pathways within these electrode structures promote
rapid transport of electrolyte ions during cycling,
In the following section, we restrict our discussion resulting in electrodes that can be charged and
to templated mesoporous solids that are of potential discharged rapidly with only minimal loss of capac-
interest as battery electrodes, including many transi- ity.150,162-164
tion-metal oxides and carbon. This slice of the
literature still points the interested reader to many 4.5. Hierarchical Designs Based on Templating
articles on the synthesis and physical characteriza-
tion of relevant mesoporous materials. A much
Strategies
smaller number of electrochemical studies with tem- Hierarchical structures with pore-solid features
plated mesoporous electrodes have been published, that cover a wide size range within one structure, as
and these studies in particular will be noted. seen in Figure 15, are achieved by carefully combin-
forms than those for comparable aerogels. A fourth

scheme of pore-fluid removal utilizes a freeze-drying
process to render highly porous “cryogels”.176-180
Aerogels and related structures have the important
characteristics of extremely high surface area and a
through-connected, aperiodic network of porosity.
Additionally, these solids can be cast and molded into
a range of shapes and forms or cast as thin films on
planar supports. One attribute that seems to set
these aperiodic architectures apart from the meso-
porous ordered solids, even the nominally 3-D porous
ones, is proven performance in rate-critical applica-
tions. The rapid and facile flux of molecules through
the aerogel architecture avoids the potentially dev-
astating loss of throughput through 1-D channels if
Figure 15. Hierarchical structure formed by combining a blockage occurs.181 Aerogels have already demon-
liquid-phase templating with micromolding. (Reprinted
with permission from ref 166. Copyright 1998 American
strated orders of magnitude faster response for
Association for the Advancement of Science.) sensing, energy storage, and energy conversion than
other pore-solid architectures.167,181
ing the liquid-phase templating approaches described Electrically conductive aerogels have been pre-
above: surfactant templating (up to 10 nm), block pared from a range of transition metal oxides, includ-
copolymer templating (up to 30 nm), and colloidal ing vanadium oxide,172,182-189 manganese oxide,173,174,190
crystal templating (100 nm-1 µm).165,166 This strat- molybdenum oxide,191-193 and tin oxide.194 All of these
egy can be used directly to fabricate inorganic nanostructured oxide forms exhibit improved perfor-
structures or indirectly for hierarchical, porous car- mance as Li ion insertion electrodes in conventional
bon structures that are produced by inverse replica- electrochemical cells.167,195 For example, vanadium
tion of a corresponding hierarchical inorganic mold.161 pentoxide aerogel electrodes accommodate the re-
These synthetic processes can be further combined versible electrochemical insertion of up to four Li ions
with micromolding techniques to create features up per V2O5 unit. In addition to higher overall capacities,
to 1 mm.166 aerogel electrodes deliver more energy (capacity)
under high charge-discharge rates than do conven-
4.6. Aerogels and Related tional materials.
Nanostructures;Aperiodic Pore−Solid Carbon aerogels are typically derived from the
Architectures reaction of resorcinol (or related aryl precursors) and
formaldehyde, which leads to a polymer gel. The
An ordered porous network is not a prerequisite postprocessed polymeric aerogels are then pyrolyzed
in designing high-performance 3-D battery struc- in inert atmospheres at high temperature to form
tures.167 Sol-gel methods can be used to synthesize highly conductive, monolithic carbon structures that
monolithic gels that are composed of a 3-D intercon- retain the characteristics of an aerogel.196-198 The
nected network of nanoscale particles, co-continu- synthesis of carbon aerogels offers considerable flex-
ously intermingled with a phase of fluid-filled pores. ibility as the pore-solid organic architecture can be
The processing of these wet gels to render the dried adjusted by simple changes in the initial reaction
forms must be carefully controlled so that the highly conditions of the polymer gel. Through such varia-
porous architecture is retained. The direct evapora- tions, carbon aerogels can be prepared in either
tion of typical pore-filling fluids (high-surface-tension predominantly mesoporous or macroporous forms.
liquids such as water and alcohols) creates capillary Carbon aerogels are being exploited as electrodes for
forces at the liquid-vapor interface, resulting in pore electrochemical capacitors, where the high surface
collapse and a partially densified porous solid known area, through-connected porosity, and electronic con-
as a xerogel. This structural collapse is avoided if the ductivity of the aerogel promote rapid charge-
pore fluid is removed under supercritical conditions discharge reactions.199-201 Commercial supercapaci-
where no liquid-vapor interface is ever established. tors based on carbon aerogels are available from
Gels processed by supercritical fluid extraction, usu- several companies.
ally with CO2 as the pore fluid, are denoted as
aerogels.167-170 5. Structural Characterization of 3-D Electrode
In an alternative to supercritical drying, the pore
fluids in a wet gel are replaced with a low-surface-
Architectures
tension, nonpolar liquid, such as an alkane, which is
then allowed to evaporate under subcritical or ambi-
5.1. Characterization of the Pore−Solid
ent-pressure conditions.170-172 The resulting solids,
Architecture
denoted as “ambigels” (for ambient-pressure-dried The common feature of the 3-D electrode structures
gels),167 do exhibit a moderate degree of densification described above is the intimate intermingling of solid
but retain a large fraction of pore volume, with pores and void space, in either ordered or aperiodic ar-
usually in the mesopore size range.173-175 The den- rangements, with textures covering an extensive
sification that accompanies ambient-pressure drying range of length scales. This section reviews the
provides for more mechanically rugged monolithic experimental methods used to characterize the pore
size and structure and to determine the electrochemi-

cal properties of the pore-solid nanoarchitectures.
The quality and type of pore structure in such an
architecture often determines the electrochemical
performance of the resulting battery material. For
instance, when the pore structure is infiltrated with
a liquid electrolyte, interconnected mesopores and
macropores provide an open pathway for the diffusion
of electrolyte ions that participate in the charge-
storage mechanisms. The size and interconnected-
ness of the pore network will be crucial in achieving
true intermingled cathode/electrolyte/anode designs.
From the standpoint of electrochemical perfor-
mance, micropores (less than 2 nm) are a far less Figure 16. The pore-size distribution for sol-gel-derived
desirable design component for 3-D power. Even birnessite NaδMnO2·xH2O as processed into three pore-solid
when employed in traditional electrochemical cells, nanoarchitectures: xerogel, ambigel, and aerogel. Distribu-
the electrode area within a micropore is not fully tions are derived from N2 physisorption measurements and
utilized in energy-storage mechanisms, due to the calculated on the basis of a cylindrical pore model. (Re-
lower effective electrolyte conductivity within the printed with permission from ref 175. Copyright 2001
American Chemical Society.)
micropore and sieving effects for larger electrolyte
ions.202,203 Although templated microporous materi-
neities in electron density within porous architec-
als, especially in carbon, have been fabricated and
tures.204,206 Small-angle scattering measurements
characterized for catalytic and adsorbent applica-
extend the range of length scales that can be ana-
tions, we are not including this class of materials for
lyzed, revealing features from micropores to the large
further 3-D design consideration because of the
inability to electrify the surface area within mi- macropores that cannot be effectively probed by gas
cropores on a practical time scale. adsorption. Another distinguishing feature of SAS
techniques is that both open and isolated pores
5.1.1. Physisorption contribute to the experimentally observed structure.
Gas adsorption techniques are often used to char- The results of gas adsorption analysis and such SAS
acterize the pore structure of materials possessing techniques as small-angle X-ray scattering (SAXS)
micropores, mesopores, and small macropores.204 can be combined to provide the most thorough
Adsorption measurements with probes such as N2 characterization of complex pore-solid architectures.
provide information on accessible surface area and A key recent advancement for SAS methods has been
pore volume. In addition, careful analysis of the the development of the chord-length distribution
experimental data also yields information about pore (CLD) approach, which yields much more accurate
sizes, shapes, and the degree of interconnectedness.205 characterizations of pore sizes and distributions than
Various mathematical models can be used to fit the with previous SAS approaches.207,208
experimental data from adsorption isotherms, yield-
ing pore-size distribution (PSD) plots describing the 5.1.3. Electron Microscopy
pore structure in the mesopore and small macropore
regime. The generation of PSD plots is critical in the Electron microscopy techniques are routinely ap-
characterization of structured materials with aperi- plied to structured nanomaterials to assess the solid
odic pore-solid architectures, such as aerogels and morphology for features from tens of micrometers to
ambigels, where it is difficult to directly image the nanometers in size. Traditional microscopy methods
pore structure with microscopic techniques. In Figure provide useful information in the form of 2-D projec-
16, the pore-size distributions are contrasted for tion images, but they do not fully describe the 3-D
birnessite, a Na+-templated, sol-gel-derived lamellar structures discussed in the preceding section. How-
polymorph of MnO2, expressed in three types of ever, the recent advancements in computer-aided
pore-solid nanoarchitectures (xerogel, ambigel, and electron microscopy techniques do provide 3-D infor-
aerogel). The PSD plots make it clear that the pore mation for structured solids.209-212 One such approach
structure in xerogels comprises micropores and small based on high-resolution electron microscopy (HREM)
mesopores, while the distribution in the aerogel is was developed by Terasaki, Stucky, Ryoo and others
weighted toward the larger mesopores and macro- specifically for porous structures exhibiting long-
pores. The ambigel lives up to its name with few range periodicity on the mesoscale (i.e., templated
micropores and macropores and most of its pore mesoporous solids).212-216 In this method, 2-D pro-
volume represented in mesopores sized from 15 to jected HREM images of the sample taken at different
30 nm.175 Reporting just the average pore diameter symmetry orientations are subjected to Fourier trans-
does not begin to capture the richness of the nanoar- form analysis to determine the crystal structure
chitectures: xerogel, 9 nm; ambigel (hexane), 12 nm; factors and ultimately to obtain a 3-D representation
ambigel (cyclohexane), 17 nm; and aerogel, 19 nm. of the ordered structure. The measurements provide
information on the sizes, shapes, and connectivity of
5.1.2. Small-Angle Scattering Techniques the ordered pore structures in such materials as
A second approach to pore-structure analysis, templated mesoporous silica and carbon (see Figure
small-angle scattering (SAS), probes the heteroge- 17). Unfortunately, these methods are not applicable
Figure 18. Schematic drawing depicting SECM measure-

ment of (a) molecular transport within a porous material
and (b) electrochemical activity on one electrode in a
battery array.
5.1.4. Scanning Probe Microscopy

Scanned probe microscopies (SPM) that are capable
of measuring either current or electrical potential are
promising for in situ characterization of nanoscale
energy storage cells. Mass transfer, electrical con-
ductivity, and the electrochemical activity of anode
and cathode materials can be directly quantified by
Figure 17. High-resolution electron micrographic deter- these techniques. Two examples of this class of SPM
mination of cage and bimodal pore structure in the meso- are scanning electrochemical microscopy (SECM) and
porous silicates SBA-6 and SBA-6: (a) 3-D structure of
SBA-6 derived from the electrostatic potential map showing
current-sensing atomic force microscopy (CAFM),
large (A) and small (B) cages with associated large and both of which are commercially available.
small pores; (b) schematic of the large and small cages in In the SECM measurement (Figure 18), a small
the SBA family; (c) pore sizes for calcined SBA-1 deter- microelectrode (typically a metal or carbon electrode)
mined by Horvath-Kawazoe analysis of the argon adsorp- is rastered across the surface of interest, and the
tion isotherm branch. (Reprinted with permission from ref current resulting from a Faradaic reaction is mea-
213. Copyright 2000 Macmillan Magazines, Ltd.)
sured.218-220 The experiment is arranged such that
the tip current is proportional to the local concentra-
for materials with irregular or hierarchical pore- tion of a redox species, which in turn may reflect
solid architectures. molecular transport rates within a porous matrix
A more general approach to 3-D structure deter- (Figure 18a)221 or the electron-transfer activity at an
mination is electron tomography, where the 3-D electrode (Figure 18b).
image of a solid (in a thin slice) is reconstructed from Quantitative theories of various SECM experi-
a series of 2-D images taken at regular tilt angle ments are well-developed and have been applied in
intervals.209 Although not yet a common technique investigations of both electrochemical and nonelec-
for materials science, electron tomography is emerg- trochemical systems,222 including, for example, in-
ing as a powerful tool for imaging complex solids with vestigations of electro-osmotic flow and transport
features approaching the nanometer scale.217 This rates in fuel-cell membranes.223 Because one detects
technique is also more generally applicable to the molecules that are undergoing random diffusion, the
textured materials described in this review, as it does spatial resolution of SECM is not as great as that of
not rely on any long-range order within the sample. scanning tunneling microscopy. However, the nano-
In addition to providing insights into sample scale pore structure in Nafion membranes has been
morphology, electron tomography can also be ex- imaged by this technique, demonstrating better than
ploited to derive 3-D compositional information. For 5-nm spatial resolution.224 And in contrast to STM
instance, energy-filtered transmission electron mi- or AFM, the SECM signal is very specific to local
croscopy (EFTEM) or Z-contrast scanning transmis- chemistry, providing a wealth of information about
sion electron microscopy (STEM) can be combined local concentrations of ions and molecules, chemical
with tomography to provide 3-D elemental mapping kinetics, and transport rates.
capabilities, within some limitations, for light and A complementary microscopy to SECM is CAFM,
heavy elements, respectively.209 In the future, elec- in which the AFM tip is made of a metal such that
tron tomography will be highly beneficial for char- current can be measured as the tip is scanned, in
acterizing the multifunctional architectures that are contact, across the surface.225 Conducting-AFM offers
currently being designed for 3-D power sources. the advantage of directly measuring local electrical
5.2. Corroboration of Electrochemical Processes

with Spectroscopic Techniques
Three-dimensional electrode nanoarchitectures ex-
hibit unique structural features, in the guise of
amplified surface area and the extensive intermin-
gling of electrode and electrolyte phases over small
length scales. The physical consequences of this type
of electrode architecture have already been dis-
cussed,167 and the key components include (i) mini-
mized solid-state transport distances; (ii) effective
mass transport of necessary electroreactants to the
large surface-to-volume electrode; and (iii) magnified
surfacesand surface defectscharacter of the electro-
chemical behavior. This new terrain demands a more
deliberate evaluation of the electrochemical proper-
ties inherent therein.
Many of the traditional electroanalytical methods
can be coupled to spectroscopic measurements, either
ex situ or in situ, to elucidate the electronic and
structural changes that accompany the electrochemi-
cal charge-storage process. The simplest approach
takes advantage of the electrochromic properties of
the various transition-metal oxides and the changes
Figure 19. SECM-AFM images of a polycarbonate mem- that occur upon charge insertion and de-insertion.
brane containing 100-nm-radius pores. (Top) AFM topog- The electrochromic response of thin oxide films
raphy image; (Bottom) SECM image of Ru(NH3)63+ trans-
port in membrane pores. (Reprinted with permission from supported on conductive, transparent substrates
ref 227. Copyright 2001 American Chemical Society.) provides an in situ, temporal indicator of electronic
state changes that occur during ion-insertion reac-
conductivity, which can be inferred only from the tions. The power of spectroelectrochemistry arises
SECM measurement. More importantly, both topog- from the fact that the spectral information, obtained
raphy and surface electronic conductivity can be concurrently with the electrochemically driven elec-
measured simultaneously, allowing the experimen- tronic state changes, decouples from other electro-
talist to correlate structure and reactivity. An ex- chemical phenomena, such as double-layer capaci-
ample of this is the CAFM imaging of electronic tance and electrolyte decomposition, which can
defects in the native oxide on titanium electrodes, potentially interfere with the electrochemical-only
which shows that high conductivity is associated with characterization.228
grain boundaries of the underlying metal.25 Ad- Spectroelectrochemical analysis of charge-insertion
ditional CAFM studies have focused on the investiga- nanostructured materials already offers important
tion of the electrical properties of electrocatalytic Pt insight into these systems. These methods were
nanoparticles deposited on and in a porous Ti/TiO2 recently exploited to characterize the electrochemical
film.226 processes of nanostructured manganese oxide ambi-
A recent development in SPM technology is the gel and xerogel films.175,229 Spectroelectrochemical
combination of SECM and AFM to produce a hybrid measurements were used to corroborate electronic
high-resolution microscope that allows simultaneous state changes with the observed electrochemical
topographic and electrochemical imaging.227 Figure response for the insertion of small cations (Li+, Mg2+)
19 shows an example of this measurement in which and the unexpected insertion of a bulky organic
pore structure and molecular transport of a redox- cation (tetrabutylammonium). Vanadium pentoxide
active molecule (Ru(NH3)63+) were simultaneously exhibits two distinct electrochromic features that can
imaged at ∼1-nm resolution. Inspection of this image be assigned to the VV/VIV transition at either sto-
clearly shows a correlation between transport rates ichiometric V2O5 sites or V sites adjacent to oxygen
and pore structure. vacancy defects.230 Exploiting this phenomenon,
Potential applications of SECM and CAFM in Rhodes et al. were able to assign these different
fundamental studies of nanoscale batteries and bat- physical environments in the voltammetry for sol-
tery materials are readily envisioned. For instance, gel-derived V2O5 (Figure 20) and to track the relative
a small ion-selective electrode can be used as the distributions of these two energetically distinct inser-
SECM tip, providing a means to image and quantify tion sites for sol-gel-derived V2O5 films as a function
Li+ distribution and transport rates. A second ap- of temperature-atmosphere treatments.231,232
plication is mapping the electrostatic potential dis- For more sophisticated spectroelectrochemical analy-
tribution between cathode and anode using a well- sis, electrochemical cycling is coupled with X-ray
poised reference electrode as the scanning tip. absorption spectroscopy (XAS) methods, which probe
Conducting-AFM imaging of battery electrode ma- electronic and local atomic structures with element
trixes comprising electronically insulating electrolyte specificity.233,234 Because XAS techniques are compat-
and a conducting material would allow rapid visu- ible with in situ investigations and do not require
alization of electrical connectivity. long-range structural order for analysis, they are
Figure 20. Spectroelectrochemical analysis of thin films of V2O5 ambigels supported on conductive glass (indium-tin
oxide). The current response is given by the continuous line, and the change in absorbance monitored at 400 and 800 nm
as a function of potential (and time) is shown as individual data points. The V2O5 ambigel was prepared by gelation of
aqueous metavanadate, dried from cyclohexane, and calcined in air at 170 °C. (Printed with permission from ref 232.)
often used to characterize complex battery materials

as a function of electrochemical state.235-239 Sol-gel-
derived V2O5 electrodes are prime candidates for XAS
analysis due to their disordered, nanoscale nature
and exceptional electrochemical properties.240-245 As
noted earlier, V2O5 ambigels and aerogels exhibit
anomalously high capacities for Li ion insertion, with
insertion of up to four Li per V2O5 for electrochemical
lithiation and up to six Li per V2O5 with chemical
lithiation.245 On the basis of the conventional inser-
tion mechanisms, any lithiation beyond x ) 2 for the
formula LixV2O5 should result in the formation of VIII
sites in the oxide lattice. Passerini et al. reported that
for chemically lithiated V2O5, the formal vanadium
oxidation state unexpectedly did not go below the VIV
state, even with lithiation up to Li5.8V2O5, suggesting
charge delocalization beyond the V sites.245 But Figure 21. Visible absorption spectra for a 1.6-µm-thick
NaδMnO2·xH2O birnessite ambigel film in 1 M LiClO4/
Mansour et al. recently described in situ XAS analy- propylene carbonate as a function of electrode potential.
sis of V2O5 aerogels and ambigels, demonstrating that (Reprinted with permission from ref 175. Copyright 2001
with electrochemical cycling the high degree of Li ion American Chemical Society.)
insertion is accompanied by the formation of VIV and
subsequently VIII sites, when the degree of lithiation storage materials is a key area of future research in
surpasses two Li per V2O5.243,244 the drive to improve the performance of batteries and
One concern that might arise in using electronic ultracapacitors.167 As discussed above, the aerogel
spectroscopy to quantify highly defective structures forms of disordered V2O5·0.5H2O afford Li-to-V2O5
is hinted at by results with nanostructured birnessite stoichiometries of 4-6,183,195 while those for the bulk,
NaδMnO2·xH2O. The color center in MnO2 is the Mn4+ crystalline material are 1-2.246 One clue to this
center, so optical density is lost in the visible spec- impressive improvement in the Li ion capacity of
trum as Mn(III) sites are electrogenerated (see Figure V2O5 can be found by inducing deliberate defects in
21). The as-prepared and calcined birnessite ambigel polycrystalline V2O5:247 Creation of proton-stabilized
is mixed valent but predominantly Mn(IV), which can cation-vacancies increases the capacity 23% relative
be quantitatively assayed by chemical redox titra- to the ∼170 mA h g-1 obtained with the as-received,
tion.174,175 The initial visible spectrum, which mea- micrometer-sized polycrystalline V2O5.
sures the electronic state of the Mn centers, has a While XAS techniques focus on direct characteriza-
lower absorbance than the material after electro- tions of the host electrode structure, nuclear magnetic
oxidation, in agreement with the chemically deter- resonance (NMR) spectroscopy is used to probe local
mined mixed-valence state of the initial material. The chemical environments via the interactions of inser-
spectrochemical determination of the Mn(III)/Mn(IV) tion cations that are NMR-active nuclei, for example
ratio, however, does not match that obtained by lithium-6 or -7, within the insertion electrode. As
chemical titration and undercounts the amount of with XAS, NMR techniques are element specific (and
Mn(IV). Because of the high surface-to-volume ratio nuclear specific) and do not require any long-range
and greater defect nature of these nanostructured structural order in the host material for analysis.
materials, we question whether any spectroscopic Solid-state NMR methods are now routinely em-
electronic state measurement truly measures a local- ployed to characterize the various chemical compo-
ized site that remains unaffected by the vacancies nents of Li ion batteries: metal oxide cathodes, Li
that arise in these materials. ion-conducting electrolytes, and carbonaceous an-
Characterizing (experimentally and computation- odes.248,249 Coupled to controlled electrochemical in-
ally) and exploiting the nature of defects in charge- sertion/deinsertion of the NMR-active cations, the
chemical shifts of the 6Li or 7Li are used to distin-

guish between chemically distinct ion insertion sites
within the host electrode structure, to differentiate
electrochemically inactive Li ion sites, and to assess
the mobilities of Li ions. Nuclear magnetic resonance
studies of battery electrodes have recently been
extended to in situ investigations of lithium insertion
in carbon electrodes by Chevallier et al.250
6. Fabrication En Route to 3-D Integration of

Architectural Components
6.1. Micromachining
The three-dimensional electrode arrays that are
the basis for the interdigitated 3-D battery design
(Figure 2a) have been fabricated successfully using
different micromachining methods. Micromachining
has emerged as an indispensable approach for fab-
ricating materials into complex geometries.251 These
methods evolved from the realm of integrated circuits Figure 22. Array of carbon electrodes prepared by the
as researchers developed techniques for deposition, C-MEMS process. The aspect ratio is approximately 20:1.
photolithographic patterning, and selective etching, (Courtesy of Prof. Marc Madou, University of Californias
among other processes. Distinctions can be drawn Irvine.)
between bulk machining, where three-dimensional
features are etched in bulk materials, and surface these pyrolyzed electrodes behave similarly to those
micromachining, where features are built up, layer- of glassy carbon, a standard carbon electrode used
by-layer, on the surface of a substrate. Surface in electrochemical studies. It is evident that good
micromachining also involves the use of sacrificial electrochemical responses are obtained for materials
films that are later dissolved to release the free- pyrolyzed at temperatures above 800 °C, although
standing components. In both types of micromachin- the best properties seem to occur at pyrolysis tem-
ing, photolithography serves as the basis for fabri- peratures above 1000 °C. Voltammetric measure-
cating three-dimensional features. ments show that the electron-transfer kinetics mea-
Three-dimensional electrode arrays have been fab- sured for Fe(CN)63-/4- and Ru(NH3)63+/2+ are compar-
ricated using two very different micromachining able to those determined for other carbon films and
methods. One approach, named carbon MEMS or glassy carbon.254 One interesting difference with
C-MEMS, is based on the pyrolysis of photoresists. glassy carbon is that it is possible to prepare pyro-
The use of photoresist as the precursor material is a lyzed carbon electrodes with lower oxygen content.
key consideration, since photolithography can be used The low oxygen content in the carbon film and its
to pattern these materials into appropriate struc- smooth surface are believed to be responsible for low
tures. The second approach involves the microma- capacitance (8 µF/cm2), a feature that may be impor-
chining of silicon molds that are then filled with tant for analytical applications.
electrode material. Construction of both anode and Although these first studies were directed at 2-D
cathode electrode arrays has been demonstrated interdigitated electrodes (L/d , 1) whose structures
using these microfabrication methods. are not very useful for the proposed 3-D designs, it
is apparent, nonetheless, that this microfabrication
6.1.1. Carbon MEMS (C-MEMS) approach can be adapted for 3-D array electrodes.
The synthesis and electrochemical properties of The recent work reported by Madou and colleagues
carbon films prepared from positive photoresist have demonstrated the types of electrode arrays that
been reported.252-254 The initial direction for this constitute the key design element for the interdigi-
work was the fabrication of carbon interdigitated tated 3-D battery.17 By using a process similar to
electrodes. In this work, positive photoresist was spin their initial work,254 but with a negative photoresist,
coated on a silicon substrate, patterned by photoli- carbon arrays with much higher L/d ratios were
thography, and pyrolyzed to form the carbon elec- prepared, as shown in Figure 22.
trode. In more recent work, laser excitation has been The results by Madou et al. are especially signifi-
used to both pyrolyze the film and to write the cant, as they clearly demonstrate the ability to
electrode pattern.255 fabricate array electrodes with aspect ratios that are
The influence of pyrolysis conditions on the struc- on the order of 20:1. At these aspect ratios, the
ture, morphology, electrical properties, and electroanalytical models indicate that the capacity of 3-D
chemical behavior has been investigated. Raman batteries exceeds that of 2-D geometries. Two key
spectroscopy shows that characteristic sp2 carbon questions remain to be answered for C-MEMS: are
bands form from the pyrolysis treatments. The these electrodes electrochemically reversible to lithium,
electrochemical properties for a few of the electrode and what is the resistance of rods with such high
systems have been reported and, for the most part, aspect ratios?
Figure 23. Processing flow for 3-D electrode array fabrication using silicon micromachining with colloidal filling of the
electrode material. The six steps are identified as the following: (i) patterned photoresist (PR) on silicon substrate, (ii) PR
removal after DRIE micromachining, (iii) insulate silicon mold by oxidation, (iv) colloidal electrode filling material centrifuged
into the mold, (v) silver epoxy added to provide mechanical stability and electrical contact, (vi) the electrode flipped over
and released from the mold by immersion in a TEAOH solution.
This first question concerning electrochemical prop- 6.1.2. Micromachining of Silicon Molds
erties has been answered in a series of half-cell
experiments using the C-MEMS array as the working A second approach for fabricating electrode arrays
electrode and lithium as both the counter and refer- has involved micromachining of silicon molds,257
which are filled with electrode material by colloidal
ence electrode.256 In these experiments, reversible
processing methods. In contrast to C-MEMS, this
intercalation of lithium was demonstrated with C-
fabrication approach is suitable for both anodes and
MEMS array electrodes with an aspect ratio as high cathodes, as one merely alters the composition of the
as 6:1. Voltammetric sweeps indicate that the elec- powders. The process flow for electrode array fabrica-
trochemical behavior is similar to that of coke elec- tion is depicted in Figure 23.
trodes as most of the lithium intercalation occurs
The silicon molds are prepared using photolithog-
below 0.5 V, with a broad deintercalation peak at raphy and deep-reactive ion etching (DRIE). A 12-
0.3 V. Galvanostatic measurements on the C-MEMS µm layer of photoresist is spin-coated onto a silicon
arrays show a large irreversible capacity loss on first wafer and patterned with an array of circles, ranging
discharge followed by good cycling properties, which from 30-120 µm in diameter, 50 µm apart. The UV-
is also consistent with the behavior of coke electrodes. exposed sections are dissolved away with developer
The lithium capacity normalized to the footprint area and 40-120-µm-deep holes in the exposed areas of
of the electrode array is 0.125 mA h cm-2. This value silicon are created by DRIE. From these dimensions,
is nearly twice that of an unpatterned pyrolyzed film it was possible to explore the fabrication of electrode
of SU-8 photoresist.256 The reason for the higher arrays of different L/d ratios. The mold is cleaned in
capacity is due to the greater active volume, contrib- a Piranha bath (H2SO4/H2O2 solution) and a 1 µm
uted by the carbon posts, over the footprint area. thick thermal oxide is grown on the silicon using wet
Gravimetrically, these lithium capacities are within oxidation at 1100 °C. The presence of the oxide layer
the range of values reported for coke electrodes. helps in releasing the infiltrated electrode array.
However, as discussed in the Introduction, the more The second step involves filling the silicon mold
relevant parameter to use in characterizing array with the electrode powders. Several different active
electrodes is the lithium capacity per unit area of the electrode powders were investigated including LiCoO2,
array footprint. carbon black, single-wall carbon nanotubes (SWNT),
and vanadium oxide nanorolls (VONR). The active
The second key questionsthe resistance of carbon
material is mixed with a binder (polyvinylidene
rods with high aspect ratiossis just beginning to be fluoride, PVdF) in order to keep the array intact. The
addressed. The first measurements made on indi- addition of binder is a standard procedure in the
vidual pyrolyzed carbon rods with L/d ) 6 and a fabrication of battery electrodes, as is the addition
diameter of ∼25 µm indicated resistance values on of carbon black to improve the electronic conductivity
the order of 90-100 ohms. These measured values of cathodes.1 The silicon mold was placed in a vial
are reasonably consistent with calculated resistances containing a suspension of the electrode powders in
based on the resistivity values reported for pyrolyzed ethylene glycol and then centrifuged, during which
photoresist. Future work that combines experiments time the powders filled the silicon mold. The mold
with simulations will be able to establish how lithium was removed from the vial and heated to 200 °C to
capacity varies with aspect ratio. evaporate the ethylene glycol and melt the PVdF,
Figure 25. Discharge curves for the first three cycles for
an electrode array composed of vanadium oxide nanorolls
(see Figure 24). The current was 20 µA and the footprint
for the electrode array was 10 mm2.
The micromachining of silicon molds is one key

Figure 24. Electrode array prepared by the powder factor that determines the dimensions of the elec-
processing method shown in Figure 23. The electrode array trode array and, therefore, the energy of the 3-D
structure is intact, although the individual rods contain battery. The use of electrochemical etching in HF to
defects. fabricate high-aspect-ratio trenches in silicon was
introduced in 1990.259 Several groups have applied
thus binding the active electrode powders. Silver this technique in the fabrication of microstructures.260
epoxy was spread over the back of the mold to Recently, Chamran et al. reported the use of this
maintain the mechanical integrity of the array as technique for 3-D microbattery fabricatrion.261 The
well as to provide electrical contact. process flow is shown in Figure 26a. High intensity
The final step in the process is to remove the from the illumination source ensures that the elec-
electrode array from the silicon mold. This releasing tronic hole concentration is sufficiently high to sus-
function is accomplished by immersing the filled mold tain the etching process. The locations of the etched
in an aqueous solution of tetraethylammonium hy- holes are determined by the initiation tips litho-
droxide (TEAOH) heated to 80 °C. As the TEAOH graphically formed on top of the wafer before the
begins to dissolve the silicon, the electrode array photoelectrochemical etching starts. The narrow
separates from the mold, usually as an intact piece. dimension concentrates the local electric field and
An SEM image of an array of vanadium oxide ensures that etching occurs primarily at the tip. The
nanorolls is shown in Figure 24. It is evident from resulting mold has etched holes with L/d ratios >10.
this photomicrograph that the fabrication process is The importance of high, localized fields for electro-
far from perfect. The individual rods have defects chemical micromachining has also been exploited to
including pores and cracks, and the arrays occasion- create precision features below 100 nm by using a
ally have missing rods. sequence of 500-ps pulses.262
The electrochemical properties of the microma- Calculations indicate that 3-D batteries con-
chined electrode arrays have been determined for structed from electrode arrays with the dimensions
several of the electrode materials. In these half-cell shown in Figure 26b will be capable of providing over
experiments, the electrode array served as the work- 2 mW h in a 5-mm3 package, an appropriate size and
ing electrode and lithium as the counter and refer- energy for powering MEMS devices. The holes are 5
ence electrodes. The first experiments were carried µm in diameter by 100-µm-long and the center-to-
out on carbon arrays composed of powders of Ketjen center distance between holes is 10 µm. For such
Black. Reversible intercalation and deintercalation dimensions, it is not known whether the colloidal
of lithium were obtained, and reversible capacities processing approach described above will be effective,
in the range of 0.4-0.5 mA h cm-2 were reported.257 and it may be necessary to develop alternative
Recent results with vanadium oxide nanorolls powder-filling methods.
indicate that array electrodes prepared with these
materials exhibit reversible intercalation of lithium 6.2. 3-D Fabrication Based on 2-D Structures:
and behave in much the same fashion as conventional Origami
VONR electrodes. Discharge curves for the first three
cycles of a VONR array for L/d ) 3 and a footprint Another approach to forming 3-D electrode struc-
area of 10 mm2 are shown in Figure 25. The plateaus tures is based on various folding strategies. This
observed in these experiments are consistent with the “origami” approach was first reported for electronic
results we obtained for well-ordered VONR pow- and optoelectronic systems, as there is great interest
ders.258 Figure 25 shows that the arrays exhibit good in developing three-dimensional electronic networks
cycling properties and the lithium capacity normal- and microelectronic devices. The folding of both
ized to the footprint area is ∼1.5 mA h cm-2. This planar263 and linear264 structures into 3-D configura-
electrode-area-normalized capacity is about 10 times tions has been demonstrated; the adaptation of this
higher than the value reported for thin-film cath- work to electrochemical devices is just starting. Shao-
odes.13 Once again the higher capacity for the foot- Horn et al. reported on the assembly of 3-D electro-
print area is due to the greater volume contributed chemical structures using patterned 2-D electrode
by the VONR posts. sheets.18 The electrodes were designed to fold at
because the anode and cathode must be electrically

isolated. Nonetheless, the use of capillary force to
cause folding is attractive, and it would seem that
there is an opportunity to use other materials with
more compatible electrical properties. Much of the
work done with conducting polymers that serve as
actuators under various applied forces may also be
adaptable to the origami effort.265
7. Advances toward Integration of Active

Components into a 3-D Battery
7.1. Present Status
The construction of an operational, fully 3-D bat-
tery has yet to be achieved. However, as the previous
sections have shown, on an individual basis, most of
the active components for the 3-D designs presented
in Figure 2 have been demonstrated successfully.
Both anode and cathode array electrodes, the design
elements that form the basis of the interdigitated
battery (Figure 2a), have been fabricated and their
electrochemical properties determined. The next key
step is to carefully align these arrays so that inter-
digitation is achieved. The concentric tubule ap-
proach is also well along as carbon honeycomb
electrode structures have recently been prepared.97
These structures represent the continuous electrode
phase (Figure 2c) and can serve as the porous
membrane template, instead of alumina, for growing
electrode materials.74 One critical feature that needs
to be resolved is the presence of an electrolyte that
separates the two electrodes.
Of all the battery designs, it is the continuous
sponge that is furthest along toward integrating
components into a 3-D architecture (Figure 2d). This
Figure 26. (a) Process flow for the fabrication of high-
aspect-ratio holes in silicon as based on photoassisted approach involves the design and fabrication of a
etching: (i) thermal oxide is grown on n-Si and patterned; three-dimensional network from the appropriate
(ii) KOH etching through the holes; (iii) removal of oxide; nanoscale building blocks, including the use of “noth-
and (iv) photoassisted anodic etching of silicon in HF. (b) ing” (void space) and deliberate disorder as design
Cross-section of silicon etched by the photoassisted process components. A recent accomplishment in this area
yielding an array of holes of 5-µm diameter by ∼100-µm- has been the electrodeposition of a conformal elec-
long; the center-to-center distance between holes is 10 µm.
trolyte layer on an aerogel substrate: complete
integration of the electrode and electrolyte in a
specific creases based on the application of Lorentz random 3-D network. The last step in the fabrication
forces. The 3-D structure is then assembled by the of this 3-D battery is to integrate an interpenetrating
sequential folding of the 2-D electrode sheets. This anode.
work is clearly at its inception, as current efforts have
The importance of developing pinhole-free, electro-
only been able to achieve limited folding. Efforts are
lyte films of nanometer thickness is potentially useful
underway to demonstrate a 3-D capacitor structure.
for all 3-D battery designs. For this reason, most of
The attractive feature of the origami approach is this section reviews the synthesis and characteriza-
the prospect of using readily manufacturable 2-D tion of this ultrathin polymer electrolyte with an
patterning to form the electrochemical components. emphasis on topics such as leakage currents and
By proper folding, these 2-D components are then dielectric strength, which become critically important
assembled into 3-D structures. The difficulty to date at the nanoscale. A few comments concerning the
seems to be in the actual folding process. A significant packaging of 3-D batteries are made at the end.
difference between the 3-D electronic structures and
the electrochemical structures is that the former rely 7.2. Electrodeposition of DielectricssSeparators
on capillary forces from molten solder to achieve and Cation-Conductive Electrolytes
folding. As the liquid solder tries to minimize surface
area, it self-assembles the 1-D or 2-D precursors into One of the key steps en route to a 3-D nanoscopic
a 3-D structure. Upon cooling, the solder serves to battery requires fabricating an ultrathin film of a
provide both electrical connection and mechanical polymer separator/electrolyte over chemically stable,
stability to the 3-D electronic network. The solder physically rugged, cation-insertion oxide scaffolds,
approach is not viable for electrochemical devices, such as supported films of MnOx ambigels.20 In
keeping with our architectural metaphor, the fabri-

cation protocol should be capable of painting the
polymer as a conformal, ultrathin coating onto the
walls of the battery building without protruding
deeply into the rooms (and clogging the continuous
porous network). The experiments underway create
ultrathin, electronically insulating, cation-permeable
films on charge-insertion oxide nanoarchitectures via
electropolymerization of aryl monomers. The advan-
tages of electropolymerizing aryl monomers include
the ability to self-limit the growth of the electrogen-
erated polymer, usually by depositing from neutral Figure 27. Schematic for solid-state electrical measure-
or basic electrolytes, to yield thin (usually ∼10 nm ments with ultrathin polymer separators electrodeposited
thick), poorly electron-conductive polymer layers that onto planar indium-tin oxide (ITO) or Au substrates. The
conformally coat all electrifiable surface area. Espe- top electrode makes a soft contact with the polymer by slow
cially appealing in the design of interpenetrating 3-D evaporation of Au or direct contact with a liquid metal
(either Hg or GaIn eutectic) using a micrometer-controlled
batteries, electropolymerization provides a fabrica- syringe to control the approach to contact. Measurements
tion protocol that should be feasible within nanoar- are made in an argon-filled glovebox to minimize effects of
chitectures (unlike line-of-sight chemical vapor depo- O2 and H2O.
sition approaches) and should ensure intimate contact
at the electrode-polymer-electrolyte interface. electrode structures. As mentioned previously, carbon
Such conformal, ultrathin polymer separators must aerogels are attractive electrode structures for elec-
satisfy a range of physical and chemical requirements trochemical capacitors, because their open pore struc-
in order to perform at the level necessary for charge tures, high surface areas, and high electronic con-
insertion on the nanometer scale. These attributes ductivities facilitate rapid charge-discharge reactions.
include (i) highly electronically insulating, preferably However, carbon aerogels ultimately have very lim-
>10-15 S cm-1; (ii) pinhole-free to prevent electronic ited energy densities as electrochemical capacitors,
shorts between the charge-insertion cathode and the since energy storage occurs primarily by the double-
anode; (iii) ionically conducting to the charge-inser- layer capacitance mechanism. This limitation has
tion ion (which in some designs may be a cation other been addressed recently at the Naval Research
than Li+), preferably without the need for solvent or Laboratory by the application of conformal, ultrathin
plasticizing molecules dissolved or partitioned into poly(o-methoxyaniline) (POMA) coatings to conven-
the polymer film; (iv) chemically nondegradable and tional carbon aerogel electrodes. Preliminary experi-
electrochemically stable against Li metal, i.e., over ments indicate that the addition of the pseudocapac-
a 3-4 V (vs Li) range; (v) forms a stable interface itance of the POMA coatings increases the volumetric
with metallic lithium, preferably without forming a capacitance of the hybrid electrode structure by
highly resistive solid-electrolyte interphase layer; and g200%, even at high rates of charge-discharge.21
(vi) exhibits a high dielectric strength, preferably
>106 V cm-1. 7.4. Solid-State Electrical Characterization of
The use of nanoscale (5-30-nm-thick) solid poly- Conformal, Ultrathin Polymer Dielectrics
mer electrolytes should significantly improve rate
capabilities for batteries and other solid-state ionic Although the literature on electrodeposited elec-
devices and represents a considerable improvement troactive and passivating polymers is vast, surpris-
over current micrometer-thick solid electrolytes used ingly few studies exist on the solid-state electrical
in thin-film batteries.13,266 The electro-oxidation of properties of such films, with a focus on systems
o-phenylenediamine in aqueous pH 9 electrolyte onto derived from phenolic monomers,268,269 and appar-
manganese oxide (MnO2) ambigels produces a con- ently none exist on the use of such films as solid
formal polymer film that retains the mesoporosity of polymer electrolytes. To characterize the nature of
the original oxide and is so defect-free over the >200 ultrathin electrodeposited polymers as dielectrics and
m2/g nanoscopic MnO2 network that it prevents electrolytes, solid-state electrical measurements are
reductive dissolution of the MnO2 nanoarchitecture made by electrodeposition of poly(phenylene oxide)
upon immersion in aqueous acid.20 Beyond serving and related polymers onto planar ITO or Au sub-
as a protective coating, the poly(o-phenylenediamine) strates and then using a two-electrode configuration
acts as a selective ion-transport membrane, where with a soft ohmic contact as the top electrode (see
proton transport from the external electrolyte through Figure 27).23 Both dc and ac measurements are taken
the polymer coating is determined by the electro- to determine the electrical and ionic conductivities
chemical state of the polymer, a phenomenon known and the breakdown voltage of the film.
as “electrochemical ion gating”.267 The electronic insulation of these electrodeposited
polymer layers must hold to a two-terminal voltage
7.3. Hybrid Polymer−Aerogel Nanoarchitectures of 4 V if lithium (or lithium ion) anodes are to be used
as Electrochemical Capacitors in the 3-D nanobattery. Because the polymers must
also be thin, high dielectric strengths are critical. As
The recognition that self-limited arylamine-based seen in Table 2, diminishing the thickness of the
polymers could be electrochemically activated in dielectric to the nanoscale exacts a higher standard
aqueous acids has led to a new class of 3-D hybrid in terms of the quality of the dielectric. For example,
Table 2. Required Dielectric Strengths for

Nanometer-Thick Dielectrics
field across dielectric at
dielectric thickness/nm 3 V/(MV cm-1)
5 6
10 3
Figure 29. Schematic of the 1-D transport between an
20 1.5 anode and cathode separated by an ultrathin, conformal
50 0.6 polymer electrolyte.
Table 3. Self-Discharge Times for a 3-D Nanobattery
as a Function of Electrolyte Thickness and Electrical
Conductivity
σelec/S cm-1 thickness, d/nm discharge time/ha
10-9 5 10-5
10-16 5 102
10-16 50 103
a The calculation assumes a cathode structure with 50 µg
of MnO2, 100 cm2 of real electrode surface area, an electrode

capacity of 120 mA h g-1, and a 3 V operating voltage. Leakage
current is calculated from the equation, Ileak ) VσelecA/d, where
V is the voltage and A is the electrode area. Discharge time is
calculated by dividing capacity by the leakage current.
junctions formed at monomolecular layers between

Au contacts.272
Ions can be introduced into these conformal, ul-
trathin polymers by exposure to nonaqueous electro-
lytes. Electrochemical cycling of the MnO2||PPO
hybrid in 1 M LiClO4/CH3CN shows that Li ions are
transported through the ∼30-nm-thick polymer film
to the underlying metal oxide.22 Spectroelectrochemi-
cal measurements verify that the polymer-encapsu-
lated mesoporous MnO2 nanoarchitecture undergoes
the normal cation-insertion reactions with Li ions
Figure 28. Solid-state dc electrical measurements of
polyphenylene oxide in an ITO||PPO||GaIn sandwich. The
from the external electrolyte; the charge under the
thickness of the polymer was measured by tapping-mode voltammetric curves is comparable with and without
atomic force microscopy. (Reprinted with permission from the polymer.
ref 23. Copyright 2004 American Chemical Society.)
7.5. Ultrathin Polymer ElectrolytesLeakage
placing 3 V across a 10 nm thick dielectric requires Currents
a material with a dielectric strength of 3 × 106 V
As discussed in section 2.5, one must be cognizant
cm-1. Dielectrics do exist with that degree of field
of the thickness of the dielectric/electrolyte that is
strength, e.g., quartz at 7 × 106 V cm-1 270 as well as
sandwiched by the 3-D interpenetrating anode and
polymers, e.g., poly(2,6-dimethylphenylene oxide)
cathode. Even with highly electronically insulating
(PDPO) at 2.3 × 106 V cm-1.271
polymers, nanoscale distances that surpass those
With an ITO||PPO||GaIn sandwich and the two-
necessary to sustain quantum tunneling can lead to
electrode configuration shown in Figure 27, a dielec-
leakage currents, arising from small electrical con-
tric strength of g1.4 × 106 V cm-1 is obtained for 32-
ductivities at such short transport path lengths. For
nm-thick electrodeposited poly(phenylene oxide).23
the 1-D transport characteristic of the aperiodic 3-D
This value agrees with an estimated dielectric strength
nanoarchitecture, as shown schematically in Figure
derived from previously reported current-voltage
29, and with a dielectric thickness that falls below
electrical data on 180-nm-thick poly(phenylene ox-
0.1 µm, self-discharge can render 3-D nanobatteries
ide).268 At voltages <4.5 V dc, the ohmic response in
useless, as shown in Table 3. The importance of
Figure 28 is stable; at applied voltages > 4.5 V dc,
negligible electronic conductivity for the dielectric
the current increases 4 orders of magnitude, from a
cannot be overemphasized when the length scales are
few nanoamperes to tens of microamperes. Thus, 4.5
nanoscopic.
V dc is used to calculate an apparent dielectric
strength.
The 10 000-fold increased currents at >4.5 V dc are
7.6. Last Step: Add an Interpenetrating Anode
not due to irreversible breakdown of the polymer Two-thirds of the assembly necessary to form the
dielectric, however, because an ohmic response is sponge architecture for a 3-D battery have now been
again obtained when Vappl < 4.5 V dc.23 It appears demonstrated. By having a suitable pinhole-free,
that a reversible ion or atom migration, derived from thin-to-ultrathin polymer barrier formed over the
the soft contact electrode, may be occurring. This “walls” of the battery architecture, the remaining free
possibility has recently been invoked to explain volume can be filled with a counter electrode, as
phenomena reported for certain molecular electronics described above and depicted schematically in Figure
2d. The resulting architecture will be a solid-state, how power sources of comparable scale will be
fully integrated 3-D battery. The use of mesoporous, fabricated. The 2-D configurations of traditional
high surface area aerogels and ambigels as sub- batteries may not be effective here, despite their high
strates maximizes the interface between the cathode energy density. Instead, energy conversion and har-
and anode, while the electrodeposited polymer mini- vesting approaches may be more suitable for power-
mizes the separation between the cathode and anode. ing microdevices, simply because of the ability to
In contrast to other 3-D designs, this nanostructure provide on-board power. These directions are being
collapses to a uniform 2-D current density distribu- actively investigated.
tion and more effectively utilizes available volume. Three-dimensional batteries offer a different ap-
proach to the portable power field. In this paper we
7.7. Packaging of 3-D Batteries have presented 3-D designs that emphasize power
sources with small areal footprints but do not com-
A key issue that must be addressed in order for promise power and energy density. While this ap-
3-D batteries to become viable for small power proach may not help solve the power needs for cell
applications is packaging. The need to protect the phones and laptop computers, it will have a signifi-
battery in a volumetrically efficient package design cant impact on current and future generations of
without affecting battery materials or chemistry is microdevices. In particular, distributed sensor net-
a significant challenge. As discussed throughout this works and wireless communication systems are
paper, MEMS devices represent an appropriate ap- representative areas where 3-D batteries would be
plication where 3-D batteries could be well-utilized, welcomed enthusiastically because the power sup-
and the prospect of developing a conveniently pack- plies currently in use are many times the size of the
aged power source, especially one attached to the device.
backside of the device so that the real estate available This paper has established some of the design rules
for the micromachine is not reduced (see Figure 1), for 3-D batteries and the accompanying materials
would be an attractive direction. It would seem, and fabrication strategies. The battery materials field
therefore, that packaging approaches for 3-D batter- is already beginning to explore 3-D concepts. Hier-
ies should be those that lead to independently pack- archical designs based on nanostructured materials,
aged components that can be readily integrated into including the deliberate management of void space,
portable power applications. have been organized into larger macroscopic struc-
The packaging of 3-D batteries is likely to extend tures and the first results are impressive, with larger
beyond those designs developed for lithium polymer gravimetric capacities for lithium and higher dis-
and thin-film batteries. The packaging protocols charge rates. It is because of this activity in materials
developed for integrated circuit technologies appear that most of the necessary components for 3-D
to be extremely useful, since many of the objectives batteries are already in hand and the demonstration
of IC packaging are identical to those of batteries.251 of the first operational 3-D batteries is imminent.
Two approaches which are particularly attractive are
the methods developed for single chip packages: chip 9. Acknowledgments
carriers (ceramic or polymer) and TO (transistor This work has been made possible by the support
outline) headers. Both approaches lead to hermetic of the U. S. Office of Naval Research (ONR) and the
sealing and conveniently interface the component to Office of the Secretary of Defense through the Mul-
the “outside world”. One advantage with batteries is tidisciplinary University Research Initiative (MURI)
that the wire bonding requirements for attaching on 3-D Microbatteries (under Grant N00014-01-1-
leads is much less demanding than that of integrated 0757). In particular, the authors would like to thank
circuits. The extensive experience with these ap- three visionary scientific officers at the ONR who
proaches and the wide variety of packaging materials during their various tenures advocated programs to
available ensure that the vital issue of process explore electrochemical power from an integrated 3-D
compatibility with battery chemistry and materials perspective: Robert Nowak, Richard Carlin, and
can be addressed successfully. The resulting pack- Michele Anderson. We also wish to acknowledge the
aged device, a “battery chip” designed to provide contributions of our insightful and productive col-
milliwatt-hour energies in cubic millimeter packages, leagues: Christopher Rhodes [ONR-MURI postdoc-
could be readily integrated on microdevices for por- toral associate (2002-2004)], Michael Doescher
table power. [NRC-NRL postdoctoral associate (2002-2005)],
Rhonda Stroud (NRL), Christopher Smith, Ryan
8. Conclusions Hart, George Baure, Fardad Chamran, Jimmy Lim,
Chai-Won Kwon, and Tim Yeh. We also appreciate
One of the lessons learned in the 1990s was that the efforts of our colleagues who provided personal
the enormous need for high-performance portable communications and allowed us to include some of
power is not diminishing. Consumer electronics their work: Prof. Marc Madou (University of Cali-
continues to be a vibrant, worldwide market force, fornia-Irvine), Prof. Kris Pister (University of Cali-
leading to ever-increasing demands for portable fornia-Berkeley), Prof. C.-J. Kim (UCLA), and Prof.
power. The inability of lithium ion batteries to fully Sarah Tolbert (UCLA).
satisfy consumer electronics has been one of the
principal motivations for the dramatic rise in fuel- 10. References
cell research and development. As the dimensions of (1) Wakihara, M. Mater. Sci. Eng. R-Rep. 2001, 33, 109.
devices continue to shrink, the question arises as to (2) Tarascon, J.-M.; Armand, M. Nature 2001, 414, 359.
(3) Sadoway, D. R.; Mayes, A. M. MRS Bull. 2002, 27, 590. (48) Drews, J.; Wolf, R.; Fehrmann, G.; Staub, R. J. Power Sources
(4) Pescovitz, D. Small Times 2002, 2, 20. 1999, 80, 107.
(5) Mullins, J. IEEE Spectrum 2001, 23. (49) Chen, T.; Barton, S. C.; Binyamin, G.; Gao, Z. Q.; Zhang, Y. C.;
(6) Lin, L. Y.; Wu, M. C.; Sawada, R.; Mohr, J. A. J. Lightwave Kim, H.-H.; Heller, A. J. Am. Chem. Soc. 2001, 123, 8630.
Technol. 2003, 21, 582. (50) Mano, N.; Heller, A. J. Electrochem. Soc. 2003, 150, A1136.
(7) Niarchos, D. Sensors Actuators A-Phys. 2003, 106, 255. (51) Mano, N.; Mao, F.; Heller, A. J. Am. Chem. Soc. 2002, 124,
(8) Tilmans, H. A. C.; De Raedt, W.; Beyne, E. J. Micromech. 12962.
Microeng. 2003, 13, S139. (52) Mano, N.; Mao, F.; Heller, A. J. Am. Chem. Soc. 2003, 125, 6588.
(9) Hartnagel, H. L.; Rodriguez-Girones, M. IEICE Trans. Electron. (53) Mano, N.; Mao, F.; Shin, W.; Chen, T.; Heller, A. Chem.
2003, E86C, 1911. Commun. 2003, 518.
(10) Mehregany, M.; Zorman, C. A. MRS Bull. 2001, 26, 289. (54) Kim, H.-H.; Mano, N.; Zhang, X. C.; Heller, A. J. Electrochem.
(11) Koeneman, P. B.; Busch-Vishniac, I. J.; Wood, K. L. J. Micro- Soc. 2003, 150, A209.
electromech. Sys. 1997, 6, 355. (55) de la Garza, L.; Jeong, G.; Liddell, P. A.; Sotomura, T.; Moore,
(12) Kahn, J. M.; Katz, R. H.; Pister, K. S. J. J. Commun. Networks T. A.; Moore, A. L.; Gust, D. J. Phys. Chem. B 2003, 107, 10252.
2000, 2, 188. (56) Murata, K.; Izuchi, S.; Yoshihisa, Y. Electrochim. Acta 2000, 45,
(13) Bates, J. B.; Dudney, N. J.; Neudecker, B.; Ueda, A.; Evans, C. 1501.
D. Solid State Ionics 2000, 135, 33. (57) Meyer, W. H. Adv. Mater. 1998, 10, 439.
(14) Note that 1 W h ) 3600 J, and a 2 J mm-3 energy density (58) Song, J. Y.; Wang, Y. Y.; Wan, C. C. J. Power Sources 1999, 77,
corresponds to 0.56 mW h mm-3. The MEMS community reports 183.
energy units as J, whereas the battery community typically uses (59) Jones, S. D.; Akridge, J. R. J. Power Sources 1993, 44, 505.
W h, and we will use the established conventions for energy units (60) Neudecker, B. J.; Zuhr, R. A.; Kwak, B. S.; Bates, J. B.;
when discussing the work from these respective communities. Robertson, J. D. J. Electrochem. Soc. 1998, 145, 4148.
(15) Souquet, J. L.; Duclot, M. Solid State Ionics 2002, 148, 375. (61) Wang, B.; Bates, J. B.; Hart, F. X.; Sales, B. C.; Zuhr, R. A.;
(16) Warneke, B.; Last, M.; Liebowitz, B.; Pister, K. S. J. IEEE Robertson, J. D. J. Electrochem. Soc. 1996, 143, 3203.
Comput. Magazine 2001, 34, 44. (62) Bates, J. B.; Dudney, N. J.; Lubben, D. C.; Gruzalski, G. R.;
(17) Wang, C.; Taherabadi, L.; Randhawa, A.; Madou, M.; Kwon, C.- Kwak, B. S.; Yu, X.; Zuhr, R. A. J. Power Sources 1995, 54, 58.
W.; Dunn, B. In Abstracts of the 204th Meeting of the Electro- (63) Neudecker, B. J.; Dudney, N. J.; Bates, J. B. J. Electrochem.
chemical Society; Electrochemical Society: Pennington, NJ, Soc. 2000, 147, 517.
2003; Abstr. #1276. (64) Piqué, A.; Chrisey, D. B.; Auyeung, R. C. Y.; Fitz-Gerald, J.; Wu,
(18) Shao-Horn, Y.; Hidrovo, C.; Jurga, S. M.; Smith, H. I.; Barba- H. D.; McGill, R. A.; Lakeou, S.; Wu, P. K.; Nguyen, V.; Duignan,
stathis, G. In Abstracts of the 204th Meeting of the Electrochemi- M. Appl. Phys. A 1999, 69, S279.
cal Society; Electrochemical Society: Pennington, NJ, 2003; (65) Wartena, R.; Curtright, A. E.; Arnold, C. B.; Piqué, A.; Swider-
Abstr. #1279. Lyons, K. E. J. Power Sources 2004, 126, 193.
(19) Hart, R. W.; White, H. S.; Dunn, B.; Rolison, D. R. Electrochem. (66) Kushida, K.; Kuriyama, K.; Nozaki, T. Appl. Phys. Lett. 2002,
Commun. 2003, 5, 120. 81, 5066.
(20) Long, J. W.; Rhodes, C. P.; Young, A. L.; Rolison, D. R. Nano (67) El-hami, M.; Glynne-Jones, P.; White, N. M.; Hill, M.; Beeby,
Lett. 2003, 3, 1155. S.; James, E.; Brown, A. D.; Ross, J. N. Sensors Actuators A 2001,
(21) Long, J. W.; Dening, B. M.; Rolison, D. R. J. Non-Cryst. Solids 92, 335.
2004, in press. (68) Ching, N. N. H.; Wong, H. Y.; Li, W. J.; Leong, P. H. W.; Wen,
(22) Rhodes, C. P.; Long, J. W.; Doescher, M. S.; Dening, B. M.; Z. Sensors Actuators A 2002, 97-98, 685.
Rolison, D. R. J. Non-Cryst. Solids 2004, in press. (69) Glynne-Jones, P.; Beeby, S. P.; White, N. M. IEEE Proc.-Sci.
(23) Rhodes, C. P.; Long, J. W.; Doescher, M. S.; Rolison, D. R. J. Meas. Technol. 2001, 148, 68.
Phys. Chem. B 2004, 108, in press. (70) Harb, J. N.; LaFollette, R. M.; Selfridge, R. H.; Howell, L. L. J.
(24) Smith, C. P.; White, H. S. Submitted for publication. Power Sources 2002, 104, 46.
(25) Boxley, C. J.; White, H. S.; Gardner, C. E.; Macpherson, J. V. J. (71) Abbott, E. A. The Annotated FlatlandsA Romance of Many
Phys. Chem. B 2003, 107, 9677. Dimensions; Perseus Publishing: Cambridge, MA, 2002.
(26) Bard, A. J.; Faulkner, L. R. Electrochemical Methods, 2nd ed.; (72) Sing, K. S. W.; Everett, D. H.; Haul, R. A. W.; Moscou, L.;
John Wiley: New York, 2001. Pierotti, R. A.; Rouquerol, J.; Siemieniewska, T. Pure Appl.
(27) Smith, C. P.; White, H. S. Anal. Chem. 1993, 65, 3343. Chem. 1985, 57, 603.
(28) Levich, V. G. Physicochemical Hydrodyamics; Prentice Hall: (73) Penner, R. M.; Martin, C. R. J. Electrochem. Soc. 1986, 133, 310.
New York, 1962. (74) Martin, C. R. Chem. Mater. 1996, 8, 1739.
(29) Curtright, A. E.; Bouwman, P. J.; Wartena, R. C.; Swider-Lyons, (75) Lakshmi, B. B.; Patrissi, C. J.; Martin, C. R. Chem. Mater. 1997,
K. E. Int. J. Nanotechnol. 2004, 1, 149. 9, 2544.
(30) Approaches to Combat Terrorism (ACT): Opportunities for Basic (76) Sides, C. R.; Li, N. C.; Patrissi, C. J.; Scrosati, B.; Martin, C. R.
Research, Joint Workshop by The Directorate of the Mathematical MRS Bull. 2002, 27, 604.
and Physical Sciences, NSF and the Intelligence Community; (77) Hulteen, J. C.; Martin, C. R. J. Mater. Chem. 1997, 7, 1075.
National Science Foundation (http://www.mitre.org/public/act/ (78) Martin, C. R.; Mitchell, D. T. In Electroanalytical Chemistry;
10_22_final.pdf), 2003. Bard, A. J., Rubinstein, I., Eds.; Marcel Dekker: New York,
(31) Priestnall, M. A.; Kotzeva, V. P.; Fish, D. J.; Nilsson, E. M. J. 1999; Vol. 21.
Power Sources 2002, 106, 21. (79) Rouhi, A. M. Chem. Eng. News 2001, 79, 29.
(32) Dyer, C. K. Nature 1990, 343, 547. (80) Patrissi, C. J.; Martin, C. R. J. Electrochem. Soc. 1999, 146, 3176.
(33) Humble, P. H.; Harb, J. N.; LaFollette, R. J. Electrochem. Soc. (81) Patrissi, C. J.; Martin, C. R. J. Electrochem. Soc. 2001, 148,
2001, 148, A1357. A1247.
(34) Lee, S. J.; Chang-Chien, A.; Cha, S. W.; O’Hayre, R.; Park, Y. (82) Li, N. C.; Patrissi, C. J.; Che, G. L.; Martin, C. R. J. Electrochem.
I.; Saito, Y.; Prinz, F. B. J. Power Sources 2002, 112, 410. Soc. 2000, 147, 2044.
(35) Kelley, S. C.; Deluga, G. A.; Smyrl, W. H. AICHE J. 2002, 48, (83) Nishizawa, M.; Mukai, K.; Kuwabata, S.; Martin, C. R.; Yone-
1071. yama, H. J. Electrochem. Soc. 1997, 144, 1923.
(36) Kelley, S. C.; Deluga, G. A.; Smyrl, W. H. Electrochem. Solid- (84) Li, N. C.; Martin, C. R. J. Electrochem. Soc. 2001, 148, A164.
State Lett. 2000, 3, 407. (85) Li, N. C.; Martin, C. R.; Scrosati, B. Electrochem. Solid-State
(37) Yu, J.; Cheng, P.; Ma, Z.; Yi, B. J. Power Sources 2003, 124, 40. Lett. 2000, 3, 316.
(38) Yu, J.; Cheng, P.; Ma, Z.; Yi, B. Electrochim. Acta 2003, 48, 1537. (86) Li, N. C.; Martin, C. R.; Scrosati, B. J. Power Sources 2001, 97-
(39) Wainright, J. S.; Savinell, R. F.; Liu, C. C.; Litt, M. Electrochim. 8, 240.
Acta 2003, 48, 2869. (87) Che, G.; Jirage, K. B.; Fisher, E. R.; Martin, C. R.; Yoneyama,
(40) Beebe, D. J.; Mensing, G. A.; Walker, G. M. Annu. Rev. Biomed. H. J. Electrochem. Soc. 1997, 144, 4296.
Eng. 2002, 4, 261. (88) Che, G.; Lakshmi, B. B.; Martin, C. R.; Fisher, E. R.; Ruoff, R.
(41) Quake, S. R.; Scherer, A. Science 2000, 290, 1536. S. Chem. Mater. 1998, 10, 260.
(42) Choban, E. R.; Markoski, L. J.; Stoltzfus, J.; Moore, J. S.; Kenis, (89) Che, G. L.; Lakshmi, B. B.; Fisher, E. R.; Martin, C. R. Nature
P. A. Power Sources Proc. 2002, 40, 317. 1998, 393, 346.
(43) Kummer, J. T.; Oei, D.-G. J. Appl. Electrochem. 1985, 15, 619. (90) Che, G. L.; Lakshmi, B. B.; Martin, C. R.; Fisher, E. R. Langmuir
(44) Ferrigno, R.; Stroock, A. D.; Clark, T. D.; Mayer, M.; Whitesides, 1999, 15, 750.
G. M. J. Am. Chem. Soc. 2002, 124, 12930. (91) Martin, C. R. Acc. Chem. Res. 1995, 28, 61.
(45) Ferrigno, R.; Stroock, A. D.; Clark, T. D.; Mayer, M.; Whitesides, (92) Martin, B. R.; Dermody, D. J.; Reiss, B. D.; Fang, M. M.; Lyon,
G. M. J. Am. Chem. Soc. 2003, 125, 2014. L. A.; Natan, M. J.; Mallouk, T. E. Adv. Mater. 1999, 11, 1021.
(46) Luo, T.-J. M.; Fei, J.; Lim, K. G.; Palmore, G. T. R. In (93) Nicewarner-Peña, S. R.; Freeman, R. G.; Reiss, B. D.; He, L.;
Nanotechnology and the Environment; Karn, B., Masciangioli, Peña, D. J.; Walton, I. D.; Cromer, R.; Keating, C. D.; Natan,
T., Colvin, V. L., Alivisatos, A. P., Eds.; American Chemical M. J. Science 2001, 294, 137.
Society: Washington, DC, 2004; in press. (94) Keating, C. D.; Natan, M. J. Adv. Mater. 2003, 15, 451.
(47) Takeuchi, E. S.; Leising, R. A. MRS Bull. 2002, 27, 624. (95) Dewan, C.; Teeters, D. J. Power Sources 2003, 119, 310.
(96) Masuda, H.; Watanabe, M.; Yasui, K.; Tryk, D.; Rao, T.; (146) Ryoo, R.; Joo, S. H.; Kruk, M.; Jaroniec, M. Adv. Mater. 2001,
Fujishima, A. Adv. Mater. 2000, 12, 444. 13, 677.
(97) Li, N. C.; Mitchell, D. T.; Lee, K. P.; Martin, C. R. J. Electrochem. (147) Joo, S. H.; Jun, S.; Ryoo, R. Microporous Mesoporous Mater.
Soc. 2003, 150, A979. 2001, 44, 153.
(98) Stein, A.; Schroden, R. C. Curr. Opin. Solid State Mater. Sci. (148) Kruk, M.; Jaroniec, M.; Kim, T. W.; Ryoo, R. Chem. Mater. 2003,
2001, 5, 553. 15, 2815.
(99) John, S.; Busch, K. J. Lightwave Technol. 1999, 17, 1931. (149) Kim, T. W.; Park, I. S.; Ryoo, R. Angew. Chem., Int. Ed. 2003,
(100) Sakamoto, J. S.; Dunn, B. J. Mater. Chem. 2002, 12, 2859. 42, 4375.
(101) Stein, A. Adv. Mater. 2003, 15, 763. (150) Lee, J.; Yoon, S.; Oh, S. M.; Shin, C. H.; Hyeon, T. Adv. Mater.
(102) Kang, S.; Yu, J. S.; Kruk, M.; Jaroniec, M. Chem. Commun. 2002, 2000, 12, 359.
1670. (151) Fuertes, A. B.; Nevskaia, D. M. J. Mater. Chem. 2003, 13, 1843.
(103) Gundiah, G.; Govindaraj, A.; Rao, C. N. R. MRS Bull. 2001, 36, (152) Fuertes, A. B.; Nevskaia, D. M. Microporous Mesoporous Mater.
1751. 2003, 62, 177.
(104) Yoon, S. B.; Kim, J. Y.; Yu, J.-S. Chem. Commun. 2001, 559. (153) Yu, J. S.; Kang, S.; Yoon, S. B.; Chai, G. J. Am. Chem. Soc. 2002,
(105) Zakhidov, A. A.; Baughman, R. H.; Iqbal, Z.; Cui, C. X.; 124, 9382.
Khayrullin, I.; Dantas, S. O.; Marti, I.; Ralchenko, V. G. Science (154) Lu, A. H.; Schmidt, W.; Schüth, F. New Carbon Mater. 2003,
1998, 282, 897. 18, 181.
(106) Zakhidov, A. A.; Khayrullin, I. I.; Baughman, R. H.; Iqbal, Z.; (155) Lu, A. H.; Schmidt, W.; Spliethoff, B.; Schüth, F. Adv. Mater.
Yoshino, K.; Kawagishi, Y.; Tatsuhara, S. Nanostructured Mater. 2003, 15, 1602.
1999, 12, 1089. (156) Yoon, S. B.; Kim, J. Y.; Yu, J.-S. Chem. Commun. 2002, 1536.
(107) Lindén, M.; Schacht, S.; Schüth, F.; Steel, A.; Unger, K. K. J. (157) Yang, H. F.; Shi, Q. H.; Liu, X. Y.; Xie, S. H.; Jiang, D. C.; Zhang,
Porous Mater. 1998, 5, 177. F. Q.; Yu, C. Z.; Tu, B.; Zhao, D. Y. Chem. Commun. 2002, 2842.
(108) Liu, T. B.; Burger, C.; Chu, B. Prog. Polym. Sci. 2003, 28, 5. (158) Kim, S. S.; Pinnavaia, T. J. Chem. Commun. 2001, 2418.
(109) Soler-Illia, G. J. d. A. A.; Crepaldi, E. L.; Grosso, D.; Sanchez, (159) Vix-Guterl, C.; Boulard, S.; Parmentier, J.; Werckmann, J.;
C. Curr. Opin. Coll. Interface Sci. 2003, 8, 109. Patarin, J. Chem. Lett. 2002, 1062.
(110) Soler-Illia, G. J. d. A. A.; Sanchez, C.; Lebeau, B.; Patarin, J. (160) Han, B. H.; Zhou, W. Z.; Sayari, A. J. Am. Chem. Soc. 2003,
Chem. Rev. 2002, 102, 4093. 125, 3444.
(111) Schüth, F. Chem. Mater. 2001, 13, 3184. (161) Taguchi, A.; Smått, J. H.; Lindén, M. Adv. Mater. 2003, 15, 1209.
(112) Yang, P. D.; Zhao, D. Y.; Margolese, D. I.; Chmelka, B. F.; Stucky, (162) Yoon, S.; Lee, J. W.; Hyeon, T.; Oh, S. M. J. Electrochem. Soc.
G. D. Chem. Mater. 1999, 11, 2813. 2000, 147, 2507.
(113) Mann, S.; Burkett, S. L.; Davis, S. A.; Fowler, C. E.; Mendelson, (163) Lee, J.; Yoon, S.; Hyeon, T.; Oh, S. M.; Kim, K. B. Chem.
N. H.; Sims, S. D.; Walsh, D.; Whilton, N. T. Chem. Mater. 1997, Commun. 1999, 2177.
9, 2300. (164) Zhou, H. S.; Zhu, S. M.; Hibino, M.; Honma, I. J. Power Sources
(114) Palmqvist, A. E. C. Curr. Opin. Coll. Interface Sci. 2003, 8, 145. 2003, 122, 219.
(115) Gin, D. L.; Gu, W. Q.; Pindzola, B. A.; Zhou, W. J. Acc. Chem. (165) Soten, I.; Ozin, G. A. Curr. Opin. Coll. Interface Sci. 1999, 4,
Res. 2001, 34, 973. 325.
(116) Lin, H. P.; Mou, C. Y. Acc. Chem. Res. 2002, 35, 927. (166) Yang, P. D.; Deng, T.; Zhao, D. Y.; Feng, P. Y.; Pine, D.; Chmelka,
(117) Patarin, J.; Lebeau, B.; Zana, R. Curr. Opin. Coll. Interface Sci. B. F.; Whitesides, G. M.; Stucky, G. D. Science 1998, 282, 2244.
2002, 7, 107. (167) Rolison, D. R.; Dunn, B. J. Mater. Chem. 2001, 11, 963.
(118) Forster, S. In Colloid Chemistry 1; Topics in Current Chemistry; (168) Kistler, S. S. Nature 1931, 127, 741.
Antonietti, M., Ed.; Springer: Berlin, 2003; Vol. 226. (169) Hüsing, N.; Schubert, U. Angew. Chem., Int. Ed. Engl. 1998,
(119) Schüth, F. Angew. Chem., Int. Ed. 2003, 42, 3604. 37, 23.
(120) Ying, J. Y.; Mehnert, C. P.; Wong, M. S. Angew. Chem., Int. Ed. (170) Pierre, A. C.; Pajonk, G. M. Chem. Rev. 2002, 102, 4243.
1999, 38, 56. (171) Harreld, J.; Wong, H. P.; Dave, B. C.; Dunn, B.; Nazar, L. F. J.
(121) Polarz, S.; Antonietti, M. Chem. Commun. 2002, 2593. Non-Cryst. Solids 1998, 225, 319.
(122) Ryoo, R.; Joo, S. H.; Jun, S. J. Phys. Chem. B 1999, 103, 7743. (172) Coustier, F.; Lee, J. M.; Passerini, S.; Smyrl, W. H. Solid State
(123) Yang, P. D.; Zhao, D. Y.; Margolese, D. I.; Chmelka, B. F.; Stucky, Ionics 1999, 116, 279.
G. D. Nature 1998, 396, 152. (173) Long, J. W.; Swider-Lyons, K. E.; Stroud, R. M.; Rolison, D. R.
(124) Grätzel, M. J. Sol-Gel Sci. Technol. 2001, 22, 7. Electrochem. Solid-State Lett. 2000, 3, 453.
(125) Wang, Y. D.; Ma, C. L.; Sun, X. D.; Li, H. D. Appl. Catal. A-Gen. (174) Long, J. W.; Stroud, R. M.; Rolison, D. R. J. Non-Cryst. Solids
2003, 246, 161. 2001, 285, 288.
(126) Wang, Y. D.; Ma, C. L.; Sun, X. D.; Li, H. D. J. Non-Cryst. Solids (175) Long, J. W.; Qadir, L. R.; Stroud, R. M.; Rolison, D. R. J. Phys.
2003, 319, 109. Chem. B 2001, 105, 8712.
(127) Kavan, L.; Rathousky, J.; Grätzel, M.; Shklover, V.; Zukal, A. (176) Yamamoto, T.; Nishimura, T.; Suzuki, T.; Tamon, H. Carbon
Microporous Mesoporous Mater. 2001, 44, 653. 2001, 39, 2374.
(128) Froba, M.; Muth, O.; Reller, A. Solid State Ionics 1997, 101, 249. (177) Xu, J. J.; Yang, J. S. Electrochem. Commun. 2003, 5, 230.
(129) Kluson, P.; Kacer, P.; Cajthaml, T.; Kalaji, M. J. Mater. Chem. (178) Yang, J. S.; Xu, J. J. J. Power Sources 2003, 122, 181.
2001, 11, 644. (179) Tamon, H.; Ishizaka, H.; Yamamoto, T.; Suzuki, T. Drying
(130) Yun, H.-S.; Miyazawa, K.; Zhou, H. S.; Honma, I.; Kuwabara, Technol. 2001, 19, 313.
M. Adv. Mater. 2001, 13, 1377. (180) Tamon, H.; Ishizaka, H.; Yamamoto, T.; Suzuki, T. Carbon 1999,
(131) Yoshitake, H.; Sugihara, T.; Tatsumi, T. Chem. Mater. 2002, 14, 37, 2049.
1023. (181) Rolison, D. R. Science 2003, 299, 1698.
(132) Zheng, J.-Y.; Pang, J.-B.; Qiu, K.-Y.; Wei, Y. J. Mater. Chem. (182) Passerini, S.; Ressler, J. J.; Le, D. B.; Owens, B. B.; Smyrl, W.
2001, 11, 3367. H. Electrochim. Acta 1999, 44, 2209.
(133) Nishimura, S.; Shishido, A.; Abrams, N.; Mallouk, T. E. Appl. (183) Salloux, K.; Chaput, F.; Wong, H. P.; Dunn, B.; Breiter, M. W.
Phys. Lett. 2002, 81, 4532. J. Electrochem. Soc. 1995, 142, L191.
(134) Krtil, P.; Fattakhova, D.; Kavan, L.; Burnside, S.; Grätzel, M. (184) Zhang, F.; Passerini, S.; Owens, B. B.; Smyrl, W. H. Electrochem.
Solid State Ionics 2000, 135, 101. Solid-State Lett. 2001, 4, A221.
(135) Kavan, L.; Attia, A.; Lenzmann, F.; Elder, S. H.; Grätzel, M. J. (185) Chaput, F.; Dunn, B.; Fuqua, P.; Salloux, K. J. Non-Cryst. Solids
Electrochem. Soc. 2000, 147, 2897. 1995, 188, 11.
(136) Ozkan, E.; Lee, S. H.; Liu, P.; Tracy, C. E.; Tepehan, F. Z.; Pitts, (186) Coustier, F.; Passerini, S.; Smyrl, W. H. J. Electrochem. Soc.
J. R.; Deb, S. K. Solid State Ionics 2002, 149, 139. 1998, 145, L73.
(137) Cheng, W.; Baudrin, E.; Dunn, B.; Zink, J. I. J. Mater. Chem. (187) Dong, W.; Rolison, D. R.; Dunn, B. Electrochem. Solid-State Lett.
2001, 11, 92. 2000, 3, 457.
(138) Attard, G. S.; Bartlett, P. N.; Coleman, N. R. B.; Elliott, J. M.; (188) Harreld, J. H.; Dong, W.; Dunn, B. MRS Bull. 1998, 33, 561.
Owen, J. R.; Wang, J. H. Science 1997, 278, 838. (189) Le, D. B.; Passerini, S.; Guo, J.; Ressler, J.; Owens, B. B.; Smyrl,
(139) Nelson, P. A.; Owen, J. R. J. Electrochem. Soc. 2003, 150, A1313. W. H. J. Electrochem. Soc. 1996, 143, 2099.
(140) Nelson, P. A.; Elliott, J. M.; Attard, G. S.; Owen, J. R. Chem. (190) Passerini, S.; Coustier, F.; Giorgetti, M.; Smyrl, W. H. Electro-
Mater. 2002, 14, 524. chem. Solid-State Lett. 1999, 2, 483.
(141) Liu, P.; Lee, S. H.; Tracy, C. E.; Yan, Y. F.; Turner, J. A. Adv. (191) Dong, W.; Dunn, B. J. Non-Cryst. Solids 1998, 225, 135.
Mater. 2002, 14, 27. (192) Dong, W.; Dunn, B. J. Mater. Chem. 1998, 8, 665.
(142) Huang, L. M.; Wang, Z. B.; Wang, H. T.; Cheng, X. L.; Mitra, (193) Dong, W.; Mansour, A. N.; Dunn, B. Solid State Ionics 2001,
A.; Yan, Y. X. J. Mater. Chem. 2002, 12, 388. 144, 31.
(143) Baeck, S. H.; Choi, K. S.; Jaramillo, T. F.; Stucky, G. D.; (194) Harreld, J. H.; Sakamoto, J.; Dunn, B. J. Power Sources 2003,
McFarland, E. W. Adv. Mater. 2003, 15, 1269. 115, 19.
(144) Choi, K. S.; Lichtenegger, H. C.; Stucky, G. D.; McFarland, E. (195) Owens, B. B.; Passerini, S.; Smyrl, W. H. Electrochim. Acta 1999,
W. J. Am. Chem. Soc. 2002, 124, 12402. 45, 215.
(145) Jun, S.; Joo, S. H.; Ryoo, R.; Kruk, M.; Jaroniec, M.; Liu, Z.; (196) Bock, V.; Emmerling, A.; Fricke, J. J. Non-Cryst. Solids 1998,
Ohsuna, T.; Terasaki, O. J. Am. Chem. Soc. 2000, 122, 10712. 225, 69.
(197) Schaefer, D. W.; Pekala, R.; Beaucage, G. J. Non-Cryst. Solids (236) Balasubramanian, M.; Sun, X.; Yang, X. Q.; McBreen, J. J. Power
1995, 186, 159. Sources 2001, 92, 1.
(198) Tamon, H.; Ishizaka, H.; Mikami, M.; Okazaki, M. Carbon 1997, (237) Yoon, W. S.; Grey, C. P.; Balasubramanian, M.; Yang, X. Q.;
35, 791. McBreen, J. Chem. Mater. 2003, 15, 3161.
(199) Frackowiak, E.; Beguin, F. Carbon 2001, 39, 937. (238) Yoon, W. S.; Kim, N.; Yang, X. Q.; McBreen, J.; Grey, C. P. J.
(200) Li, W. C.; Reichenauer, G.; Fricke, J. Carbon 2002, 40, 2955. Power Sources 2003, 119, 649.
(201) Pekala, R. W.; Farmer, J. C.; Alviso, C. T.; Tran, T. D.; Mayer, (239) Hwang, B. J.; Tsai, Y. W.; Santhanam, R.; Liu, D. G.; Lee, J. F.
S. T.; Miller, J. M.; Dunn, B. J. Non-Cryst. Solids 1998, 225, J. Electrochem. Soc. 2003, 150, A335.
74. (240) Giorgetti, M.; Berrettoni, M.; Passerini, S.; Smyrl, W. H.
(202) Koresh, J.; Soffer, A. J. Electrochem. Soc. 1977, 124, 1379. Electrochim. Acta 2002, 47, 3163.
(203) Salitra, G.; Soffer, A.; Eliad, L.; Cohen, Y.; Aurbach, D. J. (241) Giorgetti, M.; Passerini, S.; Smyrl, W. H.; Mukerjee, S.; Yang,
Electrochem. Soc. 2000, 147, 2486. X. Q.; McBreen, J. J. Electrochem. Soc. 1999, 146, 2387.
(204) Kaneko, K. J. Membr. Sci. 1994, 96, 59. (242) Giorgetti, M.; Mukerjee, S.; Passerini, S.; McBreen, J.; Smyrl,
(205) Webb, P. A.; Orr, C. Analytical Methods in Fine Particle W. H. J. Electrochem. Soc. 2001, 148, A768.
Technology; Micromeritics Instrument Corp.: Norcross, GA, (243) Mansour, A. N.; Smith, P. H.; Baker, W. M.; Balasubramanian,
1997. M.; McBreen, J. Electrochim. Acta 2002, 47, 3151.
(206) Fratzl, P. J. Appl. Crystallogr. 2003, 36, 397.
(244) Mansour, A. N.; Smith, P. H.; Baker, W. M.; Balasubramanian,
(207) Göltner, C. G.; Smarsly, B.; Berton, B.; Antonietti, M. Chem.
M.; McBreen, J. J. Electrochem. Soc. 2003, 150, A403.
Mater. 2001, 13, 1617.
(208) Smarsly, B.; Antonietti, M.; Wolff, T. J. Chem. Phys. 2002, 116, (245) Passerini, S.; Le, D. B.; Smyrl, W. H.; Berrettoni, M.; Tossici,
2618. R.; Marassi, R.; Giorgetti, M. Solid State Ionics 1997, 104, 195.
(209) Midgley, P. A.; Weyland, M. Ultramicroscopy 2003, 96, 413. (246) Livage, J. Solid State Ionics 1996, 86-88, 935.
(210) Terasaki, O.; Ohsuna, T.; Liu, Z.; Kaneda, M.; Kamiya, S.; (247) Swider Lyons, K. E.; Love, C. T.; Rolison, D. R. Solid State Ionics
Carlsson, A.; Tsubakiyama, T.; Sakamoto, Y.; Inagaki, S.; Che, 2002, 152-153, 99.
S.; Tatsumi, T.; Camblor, M. A.; Ryoo, R.; Zhao, D.; Stucky, G.; (248) Grey, C. P.; Greenbaum, S. G. MRS Bull. 2002, 27, 613.
Shindo, D.; Hiraga, K. In Nanoporous Materials III, Studies in (249) Grey, C. P.; Lee, Y. J. Solid State Sci. 2003, 5, 883.
Surface Science and Catalysis; Sayari, A., Jaroniec, M., Eds.; (250) Chevallier, F.; Letellier, M.; Morcrette, M.; Tarascon, J.-M.;
Elsevier: Amsterdam, 2002; Vol. 141. Frackowiak, E.; Rouzaud, J. N.; Beguin, F. Electrochem. Solid-
(211) de Jong, K. P.; Koster, A. J. Chem. Phys. Chem. 2002, 3, 776. State Lett. 2003, 6, A225.
(212) Thomas, J. M.; Terasaki, O.; Gai, P. L.; Zhou, W. Z.; Gonzalez- (251) Madou, M. Fundamentals of Microfabrication; CRC Press: Baton
Calbet, J. Acc. Chem. Res. 2001, 34, 583. Rouge, LA, 1997.
(213) Sakamoto, Y. H.; Kaneda, M.; Terasaki, O.; Zhao, D. Y.; Kim, J. (252) Kim, J.; Song, X.; Kinoshita, K.; Madou, M.; White, B. J.
M.; Stucky, G.; Shim, H. J.; Ryoo, R. Nature 2000, 408, 449. Electrochem. Soc. 1998, 145, 2314.
(214) Sakamoto, Y.; Dı́az, I.; Terasaki, O.; Zhao, D. Y.; Pérez-Pariente, (253) Kostecki, R.; Song, X. Y.; Kinoshita, K. J. Electrochem. Soc. 2000,
J.; Kim, J. M.; Stucky, G. D. J. Phys. Chem. B 2002, 106, 3118. 147, 1878.
(215) Kaneda, M.; Tsubakiyama, T.; Carlsson, A.; Sakamoto, Y.; (254) Ranganathan, S.; McCreery, R.; Majji, S. M.; Madou, M. J.
Ohsuna, T.; Terasaki, O.; Joo, S. H.; Ryoo, R. J. Phys. Chem. B Electrochem. Soc. 2000, 147, 277.
2002, 106, 1256. (255) Kostecki, R.; Song, X. Y.; Kinoshita, K. Electrochem. Solid-State
(216) Carlsson, A.; Kaneda, M.; Sakamoto, Y.; Terasaki, O.; Ryoo, R.; Lett. 2002, 5, E29.
Joo, S. H. J. Electron Microsc. 1999, 48, 795. (256) Wang, C.; Taherabadi, L.; Jia, G.; Madou, M.; Yeh, Y.; Dunn, B.
(217) Fernandez, J. J.; Lawrence, A. F.; Roca, J.; Garcia, I.; Ellisman, Electrtochem. Solid-State Lett. 2004, in press.
M. H.; Carazo, J. M. J. Struct. Biol. 2002, 138, 6. (257) Baure, G.; Kwon, C.-W.; Lee, G. G.; Chamran, F.; Kim, C.-J.;
(218) Engstrom, R. C. Anal. Chem. 1984, 56, 890. Dunn, B. In Micropower and Microdevices; Brandon, E. J., Ryan,
(219) Bard, A. J.; Fan, F.-R. F.; Mirkin, M. V. Anal. Chem. 1989, 61, A., Harb, J., Ulrich, R., Eds.; Electrochemical Society: Penning-
132. ton, NJ, 2002; Vol. PV2002-25.
(220) Bard, A. J.; Fan, F.-R. F.; Mirkin, M. V. In Electroanalytical (258) Sun, D.; Kwon, C.-W.; Baure, G.; Richman, E.; MacLean, J.;
Chemistry; Bard, A. J., Ed.; Marcel Dekker: New York, 1994; Dunn, B.; Tolbert, S. H. Submitted for publication.
Vol. 18. (259) Lehmann, V.; Foll, H. J. Electrochem. Soc. 1990, 137, 653.
(221) Bath, B. D.; White, H. S.; Scott, E. R. Imaging molecular (260) Foll, H.; Christophersen, M.; Carstensen, J.; Hasse, G. Mater.
transport across membranes; John Wiley: New York, 2001. Sci. Eng. R-Rep. 2002, 39, 93.
(222) Scanning Electrochemical Microscopy; Bard, A. J.; Mirkin, M.
(261) Chamran, F.; Christophersen, M.; Kim, C.-J. In Abstracts of the
V., Eds.; John Wiley: New York, 2001.
204th Meeting of the Electrochemical Society; Electrochemical
(223) Bath, B. D.; White, H. S.; Scott, E. R. Anal. Chem. 2000, 72,
Society: Pennington, NJ, 2003; Abstr. #1292.
433.
(224) Fan, F.-R. F.; Bard, A. J. Science 1995, 270, 1849. (262) Kock, M.; Kirchner, V.; Schuster, R. Electrochim. Acta 2003, 48,
(225) Miki, T.; Yanagi, H. Langmuir 1998, 14, 3405. 3213.
(226) Macpherson, J. V.; Gueneau de Mussy, J.-P.; Delplancke, J.-L. (263) Gracias, D. H.; Kavthekar, V.; Love, J. C.; Paul, K. E.; White-
Electrochem. Solid-State Lett. 2001, 4, E33. sides, G. M. Adv. Mater. 2002, 14, 235.
(227) Macpherson, J. V.; Jones, C. E.; Barker, A. L.; Unwin, P. R. Anal. (264) Whitesides, G. M.; Boncheva, M. Proc. Natl. Acad. Sci. U.S.A.
Chem. 2002, 74, 1841. 2002, 99, 4769.
(228) Heineman, W. R. J. Chem. Educ. 1983, 60, 305. (265) Smela, E. Adv. Mater. 2003, 15, 481.
(229) Long, J. W.; Young, A. L.; Rolison, D. R. J. Electrochem. Soc. (266) Jang, Y. I.; Dudney, N. J.; Blom, D. A.; Allard, L. F. J.
2003, 150, A1161. Electrochem. Soc. 2002, 149, A1442.
(230) Talledo, A.; Granqvist, C. G. J. Appl. Phys. 1995, 77, 4655. (267) Burgmayer, P.; Murray, R. W. J. Am. Chem. Soc. 1982, 104,
(231) Rhodes, C. P.; Dong, W.; Long, J. W.; Rolison, D. R. In Solid 6139.
State Ionics IV; Wachsman, E. D., Swider Lyons, K. E., Carolan, (268) Dubois, J.-E.; Tourillon, G.; Pham, M.-C.; Lacase, P.-C. Thin
M. F., Garzon, F. H., Liu, M., Stetter, J. R., Eds.; Electrochemical Solid Films 1980, 69, 141.
Society: Pennington, NJ, 2003; Vol. PV2002-26. (269) McCarley, R. L.; Thomas, R. E.; Irene, E. A.; Murray, R. W. J.
(232) Dong, W.; Long, J. W.; Rolison, D. R. Unpublished data, 2001, Electroanal. Chem. 1990, 290, 79.
Naval Research Laboratory. (270) Whitehead, S. Dielectric Breakdown of Solids; Oxford University
(233) Parsons, J. G.; Aldrich, M. V.; Gardea-Torresdey, J. L. Appl. Press: Oxford, UK, 1953.
Spec. Rev. 2002, 37, 187. (271) Adohi, B.; Gosse, J. P.; Gosse, B. J. Phys. III 1991, 1, 1623.
(234) de Groot, F. Chem. Rev. 2001, 101, 1779. (272) Service, R. F. Science 2003, 302, 556.
(235) McBreen, J.; Balasubramanian, M. JOM-J. Miner. Met. Mater.
Soc. 2002, 54, 25. CR020740L
Chem. Rev. 2004, 104, 4463−4492 4463
Three-Dimensional Battery Architectures

Jeffrey W. Long,*,† Bruce Dunn,*,‡ Debra R. Rolison,*,† and Henry S. White*,§
Surface Chemistry Branch, Code 6170, Naval Research Laboratory, Washington, DC 20375, Department of Materials Science and Engineering,
UCLA, Los Angeles, California 90095, and Department of Chemistry, University of Utah, Salt Lake City, Utah 84112
Contents 5. Structural Characterization of 3-D Electrode 4478

Architectures
1. Introduction 4463 5.1. Characterization of the Pore−Solid 4478
1.1. Drivers for 3-D Power 4463 Architecture
1.2. What Do We Mean by 3-D? 4465 5.1.1. Physisorption 4479
2. Quantitative Advantages of 3-D Cell Architectures 4467 5.1.2. Small-Angle Scattering Techniques 4479
2.1. Overview of Length Scales in 3-D Battery 4467 5.1.3. Electron Microscopy 4479
Designs 5.1.4. Scanning Probe Microscopy 4480
2.2. Quantifying the Advantages of 3-D 4467 5.2. Corroboration of Electrochemical Processes 4481
Architectures with Spectroscopic Techniques
2.3. Current Distribution in 3-D Batteries 4469 6. Fabrication En Route to 3-D Integration of 4483
2.3.1. Interdigitated Electrode Arrays 4469 Architectural Components
2.3.2. Plate, Tubule, and Aperiodic 3-D 4470 6.1. Micromachining 4483
Architectures 6.1.1. Carbon MEMS (C-MEMS) 4483
2.4. Overall Design Issues 4470 6.1.2. Micromachining of Silicon Molds 4484
2.5. Cathodes and Anodes Separated by 4470 6.2. 3-D Fabrication Based on 2-D Structures: 4485
Nanometer-Wide Electrolytes Origami
3. Prior Examples of Small Power 4471 7. Advances toward Integration of Active 4486
3.1. Scope of Small Power 4471 Components into a 3-D Battery
3.2. Miniaturizing Microreactors and Fuel Cells 4471 7.1. Present Status 4486
3.2.1. Compact Mixed Reactors 4471 7.2. Electrodeposition of DielectricssSeparators 4486
3.2.2. Microfabrication Approaches for Small 4472 and Cation-Conductive Electrolytes
Power 7.3. Hybrid Polymer−Aerogel Nanoarchitectures 4487
3.2.3. Membraneless Opportunities with Laminar 4472 as Electrochemical Capacitors
Flow in Microfluidic Reactors 7.4. Solid-State Electrical Characterization of 4487
3.2.4. Power in Vivo and Biologically Derived 4472 Conformal, Ultrathin Polymer Dielectrics
3.3. Thin-Film Batteries 4473 7.5. Ultrathin Polymer ElectrolytesLeakage 4488
3.4. Generating Power from Ambient Sources 4474 Currents
3.5. Hybrid Micropower 4474 7.6. Last Step: Add an Interpenetrating Anode 4488
4. Component Design for 3-D Battery Structures 4474 7.7. Packaging of 3-D Batteries 4489
4.1. Thinking Like Architects 4474 8. Conclusions 4489
4.2. Porous Membranes as Structure-Directing 4475 9. Acknowledgments 4489
Templates 10. References 4489
4.3. Template Synthesis of Ordered Macroporous 4476
Solids with Colloidal Crystals
4.4. Template Synthesis of Mesoporous Solids 4476 1. Introduction
with Supramolecular Assemblies
4.5. Hierarchical Designs Based on Templating 4477 1.1. Drivers for 3-D Power
Strategies
4.6. Aerogels and Related 4478 The worldwide thirst for portable consumer elec-
Nanostructures;Aperiodic Pore−Solid tronics in the 1990s had an enormous impact on
Architectures portable power. Lithium ion batteries, in which
lithium ions shuttle between an insertion cathode
(e.g., LiCoO2) and an insertion anode (e.g., carbon),
* To whom correspondence should be addressed. J.W.L: e-mail, emerged as the power source of choice for the high-
jwlong@ccs.nrl.navy.mil; telephone, (+1)202-404-8697. B.D.:
e-mail, bdunn@ucla.edu; telephone, (+1)310-825-1519. D.R.R.: performance rechargeable-battery market. The per-
e-mail, rolison@nrl.navy.mil; telephone, (+1)202-767-3617. H.S.W.: formance advantages were so significant that lithium
e-mail, white@chem.utah.edu; telephone, (+1)810-585-6256. ion batteries not only replaced Ni-Cd batteries but
† Naval Research Laboratory.
‡ UCLA. left the purported successor technology, nickel-metal
§ University of Utah. hydride, in its wake.1 The thick metal plates of
Jeffrey Long (left) was born in Great Falls, Montana in 1970, but he spent Bruce Dunn holds the Nippon Sheet Glass Chair in Materials Science &
most of his early years in Winston-Salem, NC. He received a B.S. in Engineering at UCLA. He received his B.S. at Rutgers University and his
Chemistry with Honors from Wake Forest University in 1992. Working M.S. and Ph.D. degrees at UCLA. He was a staff scientist at the General
with Prof. Royce Murray, he earned a Ph.D. in Chemistry from the Electric Corporate Research and Development Center before joining the
University of North Carolina at Chapel Hill in 1997. His research focuses UCLA faculty in 1980. His research interests concern the synthesis of
on nanostructured materials, particularly hybrid nanoarchitectures for ceramics and inorganic materials and characterization of their electrical,
applications in sensing, separations, and electrochemical energy storage electrochemical, and optical properties. A continuing theme in his research
and conversion. is the use of sol−gel methods to synthesize materials that incorporate
Debra Rolison (right) was born in Sioux City, Iowa in 1954. She received specific dopants and are capable of developing unique microstructures
a B.S. in Chemistry from Florida Atlantic University in 1975 and a Ph.D. and properties. The areas presently being studied in his group include
in Chemistry from the University of North Carolina at Chapel Hill in 1980 biosensors, intercalation compounds, aerogels, and organic/inorganic hybrid
under the direction of Prof. Royce W. Murray. She joined the Naval materials.
Research Laboratory as a research chemist in 1980 and currently heads
the Advanced Electrochemical Materials section. She is also an Adjunct
Professor of Chemistry at the University of Utah. Her research at the
NRL focuses on multifunctional nanoarchitectures, with special emphasis
on new nanostructured materials for catalytic chemistries, energy storage
and conversion, biomolecular composites, porous magnets, and sensors.
traditional batteries gave way to lithium ion cells in

which the anode, separator/electrolyte, and cathode
were stacked, spiral wound, or folded.2 Although the
materials in today’s batteries differ from those of the
displaced technologies and the energy density of the
electrochemical cell is decidedly larger, the basic two-
dimensional (2-D) character and layer-by-layer con-
struction of the cell remains intact. The next gen-
eration of lithium ion-polymer electrolyte batteries
will offer improvements in gravimetric and volumet- Henry White was born in Chapel Hill, North Carolina in 1956. He received
ric energy densities (W h g-1 and W h L-1, respec- a B.S. in Chemistry from the University of North Carolina in 1978 and a
tively), but the same configurations will prevail.3 Ph.D. in Chemistry from the University of Texas at Austin in 1982. He
Another massive market force is expected to de- held a postdoctoral appointment at the Massachusetts Institute of
velop over the next decade in the area generally Technology from 1983 to 1984 and was on the faculty of the Department
of Chemical Engineering and Materials Science at the University of
known as microelectromechanical systems or MEMS.4 Minnesota from 1984 to 1993. He is currently a Professor of Chemistry
This field grew out of the integrated-circuits (IC) at the University of Utah. His research interests include magnetic-field-
industry, and in less than a decade, MEMS devices induced transport, oxide films and corrosion, iontophoretic transdermal
successfully established high-volume commercial mar- drug delivery, and electrochemical phenomena at electrodes of nanometer
kets for accelerometers and pressure sensors in the dimensions.
automotive industry, ink-jet print heads, and digital
micromirrors for image projection. By 2001, the field. In particular, the issue of how to power MEMS-
global market for MEMS was nearly $14 billion.5 based sensors and actuators has been largely ne-
Developments in the MEMS field seem to be con- glected.11 While a conventional macroscopic power
tinual, and MEMS devices are beginning to make supply could be employed for these devices, the
significant contributions in new arenas, including prospect of powering arrays of micromachined sen-
fluidics, wireless communications, sensors, and sors and actuators is likely to lead to interconnection
optics.6-9 The rapid expansion of MEMS into new problems, cross-talk, noise, and difficulty in control-
areas is due to the development of surface microma- ling the power delivered. In contrast, the complexity
chining techniques, which involve photolithographic of power delivery is reduced if one were to design the
patterning, deposition, and selective etching of mul- power supply to be of the same dimensional scale as
tilayered films to form device structures.10 the sensors and actuators. Specifically, site-specific
There is tacit recognition that power is a vital issue power can be realized, and improvements in noise
for the continued development of the MEMS device and power efficiency may be achieved.
battery is able to supply 2 J mm-3. The reality,

however, is that thin-film batteries are 2-D devices.
Such batteries are necessarily thin in order to
prevent power losses typically associated with the
slow transport of ions, but a practical limitation
arises from the fracture of cathode films that get too
thick (>2 µm), which leads to low capacity per square
millimeter. With a typical device thickness on the
order of 15 µm, a thin-film battery requires a signifi-
cant amount of area in order to supply 2 J mm-3.
The real question becomes whether a thin-film
Figure 1. Smart dust mote and its components: Micro- battery can supply what the smart dust mote re-
fabricated sensors, optical receiver, signal-processing and quires within the constraints of the areal “footprint”
control circuitry; the power source consists of a solar cell available for the battery on the smart dust motes1
and a thick-film battery. (Derived with permission from J of energy over 1 mm2 of “real estate.” The tradi-
ref 16. Copyright 2001 IEEE)
tional yardsticks by which battery energy and power
are normalized are weight and volume [i.e., gravi-
The power needs for MEMS devices are diverses metric and volumetric energy (and power) densities].
and batteries may not be the best choice to provide These metrics are no longer sufficient when one
power to systems based on various types of MEMS considers portable power for small devices: such
drives. For example, magnetic drives operate at less devices have limited area available to integrate
than 1 V, but they require generating hundreds of components into a system. For this reason it is
milliamperes, which becomes a difficult challenge for important to establish the effective area of the power
batteries sized on the subcentimeter scale. The source, its footprint, and to normalize the available
required micro- to nanoampere current levels for energy (and/or power) to this area.
electrostatic and piezoelectric MEMS are feasible for
The energy per unit area as reported for several
batteries, but the tens to hundreds of volts that are
lithium thin-film batteries ranges from 0.25 to ∼2 ×
needed will present difficulties for batteries with
10-2 J mm-2.15 Thus, thin-film batteries, despite their
nominal voltages of 3 V. However, there may be a
excellent energy per unit volume, fall far short of
niche for batteries that would be used to power 10-
being able to power a smart dust mote for 1 day. If
15 V drives.
the areal footprint were made 100 times larger (at 1
A more reasonable application for batteries is in cm2), the thin-film approach would be acceptable. The
the area of MEMS-based sensing. In one particularly consequences of the 2-D nature of thin-film batteries
exciting direction, batteries supply power to autono- are easily overlooked. The calculation by Koeneman
mous nodes, with each node containing one or more et al. ignored the 2-D character of thin-film batteries
sensors, along with powering the computation and when they concluded that these batteries could carry
communication capabilities necessary to relay the out some 60 000 actuations of a “smart bearing”.11
findings of each node.12 One approach, “smart dust”, When one considers the actual area available for the
represents an excellent example where integrated power source on the device, only about 1200 actua-
batteries may contribute; the basic device is shown tions are possible.
in Figure 1. The device integrates into a single
Smart dust motes are not just intriguing lab toys
package, MEMS components (sensors, beam steer-
but successfully demonstrated devices.16 However,
ing), optical communication components (optical re-
because of the inability to obtain scale-appropriate
ceiver, semiconductor laser diode), signal-processing
lithium ion batteries, the first devices tested were
and control circuitry, and a power source based on
powered with hearing-aid batteries. The example of
batteries and solar cells, in a volume on the order of
the smart dust mote illustrates a critical issue
a cubic millimeter.
concerning all small power designs and their effective
A calculation of the power requirements of the integration on-board the device. To power devices
smart dust mote underscores our point that the with limited real estate and maintain a small areal
present generation of batteries cannot effectively footprint, batteries must somehow make good use of
power this device. Thin-film batteries are among the thickness. Three-dimensional configurations offer a
most advanced of the lithium battery systems, with means to keep transport distances short and yet
a capability to scale down to dimensions on the same provide enough material such that the batteries can
order of magnitude as the cubic millimeter of the dust power MEMS devices for extended periods of time.
mote.13 The energy density for the thin-film system The detailed calculations discussed in section 2 show
is ∼2 J mm-3,14 which matches or exceeds standard that batteries configured as 3-D structures can
lithium ion systems, such as those that power laptop readily achieve the 1 J mm-2 areal footprint required
computers. A key design requirement for the smart to power the smart dust mote.
dust mote is that the power consumption cannot
exceed 10 µW. If the dust mote uses this power
continuously over a day, it will consume ∼1 J.
1.2. What Do We Mean by 3-D?
Can a thin-film battery supply the 1 J per day of Lithium ion batteries use insertion processes for
energy necessary to power a smart dust mote? At first both the positive and negative electrodes, leading to
glance, it would appear that there should be no the term “rocking chair” battery.1 The resulting
problem; the device consumes 1 J mm-3 and the transport of Li ions between the electrodes, usually
arranged in a parallel-plate configuration, is 1-D in

nature. To minimize power losses resulting from slow
transport of ions, the thickness of the insertion
electrodes, as well as the separation distance between
them, is kept as small as possible. This approach may
appear counterintuitive in the effort to produce a
useful battery, because reducing the thickness of the
electrode results in lower energy capacity and shorter
operating time. Thus, battery design always trades
off between available energy and the ability to release
this energy without internal power losses.
In recent years there has been the realization that
improved battery performance can be achieved by
reconfiguring the electrode materials currently em-
ployed in 2-D batteries into 3-D architectures. Some
of the envisioned approaches are reviewed in more
detail in section 2. The general strategy of this
approach is to design cell structures that maximize
power and energy density yet maintain short ion
transport distances. While many possible architec-
tures can achieve this goal, a defining characteristic
of 3-D batteries is that transport between electrodes
remains one-dimensional (or nearly so) at the micro-
scopic level, while the electrodes are configured in
Figure 2. Examples of prospective 3-D architectures for
complex geometries (i.e., nonplanar) in order to charge-insertion batteries: (a) array of interdigitated
increase the energy density of the cell within the cylindrical cathodes and anodes; (b) interdigitated plate
footprint area. A 3-D matrix of electrodes (in a array of cathodes and anodes; (c) rod array of cylindrical
periodic array or an aperiodic ensemble) is necessary anodes coated with a thin layer of ion-conducting dielectric
to meet both the requirements of short transport (electrolyte) with the remaining free volume filled with the
lengths and large energy capacity. Improvements in cathode material; (d) aperiodic “sponge” architectures in
which the solid network of the sponge serves as the charge-
energy per unit area and high-rate discharge capa- insertion cathode, which is coated with an ultrathin layer
bilities are two of the benefits that may be realized of ion-conducting dielectric (electrolyte), and the remaining
for these 3-D cells. free volume is filled with an interpenetrating, continuous
Perhaps the most obvious 3-D design is that anode.
consisting of interdigitated electrodes shown in Fig- important consideration, because prospective syn-
ure 2a. This prototype configuration is used in thetic techniques, such as templating through me-
modeling studies, the results of which are reviewed soporous membranes, do not necessarily produce
in section 2. The anode and cathode consist of arrays periodic arrays.
of rods separated by a continuous electrolyte phase. A completely aperiodic 3-D battery configuration
The spatial arrangement of the anode and cathode is the “sponge” approach, where the electrolyte layer
arrays determines the current-potential distribu- is formed around a random 3-D network of electrode
tion. Clearly, the short transport distances lead to a material (Figure 2d). This design strategy also rep-
much lower interelectrode ohmic resistance as com- resents a concentric configuration in that the elec-
pared to traditional planar battery configurations. trolyte envelops the electrode material while the
A central feature for the interdigitated configura- other electrode material fills the mesoporous and
tion is that the arrays need to be periodic. A variation macroporous spaces and surrounds the electrolyte.
on this approach is to use interdigitated plates rather Short transport-path characteristics between the
than rods (Figure 2b), which is analogous to stacking insertion electrodes are preserved with this arrange-
2-D batteries with parallel connection. ment. In contrast to the other 3-D designs, all battery
There is no particular reason that the electrolyte componentssanode, cathode, and electrolytesare
must serve as the continuous phase in 3-D battery continuous throughout the sponge structure.
architectures. Another 3-D design utilizes a concen- The above discussion provides the context for 3-D
tric arrangement where the rod array is composed batteries. That is, there are a variety of small power
of one of the active electrode materials and is then applications, typified by MEMS devices, which the
coated by an electrolyte layer. The other electrode most advanced, 2-D lithium battery systems are
material then fills the remaining free volume and unable to satisfy. The inability to provide sufficient
serves as the continuous phase, as shown in Figure power is because of configuration and not because of
2c. As was discussed for the above battery configura- intrinsic energy density. Three-dimensional designs
tions, short transport distances between the insertion offer the opportunity to achieve milliwatt-hour ener-
electrodes lead to low ohmic resistance. Although a gies in cubic millimeter packages and, more impor-
periodic arrangement is shown in Figure 2c, it is not tantly, with square millimeter footprints. While such
clear at the present time whether this is beneficial power sources may not influence the enormous com-
or not with respect to the current-potential distribu- mercial markets in cell phones and laptop computers,
tion. Relaxing the periodicity condition may be an they are certain to impact emerging markets where
integrated power is required for communication,

sensing, and networking.
This paper discusses the prospects for creating 3-D
architectures for batteries. We next introduce calcu-
lations for the 3-D interdigitated array battery
(Figure 2a), which will illustrate the design consid-
erations under which the 3-D configuration leads to
better performance than the conventional 2-D one.
Other sections then discuss the prior art in small
Figure 3. 2-D parallel-plate and 3-D interdigitated-array
power, followed by the materials, synthetic, and batteries.
fabrication approaches required to achieve 3-D de-
signs. Finally, we review the present status of the tion is nonuniform.19 This nonuniformity can be
systems most likely to demonstrate true 3-D battery readily recognized by inspection of Figure 2a and
operation. realizing that the distance between the anodes and
cathodes is variable along the circumference on any
2. Quantitative Advantages of 3-D Cell individual electrode. A quantitative description of
Architectures this issue is discussed below.
The concentric tubule arrays and aperiodic con-
2.1. Overview of Length Scales in 3-D Battery tinuous sponge architectures are based on conformal
Designs deposition methods, in which layers of electrolyte and/
or electrode materials are sequentially20-23 or simul-
The four architectures introduced in Figure 2 have
taneously assembled within an array of pores in a
wide-ranging length scales and geometries that are
membrane (Figure 2c) or in an aperiodic array
largely dictated by their respective methods of fab-
(Figure 2d). Film deposition methods (see section 7)
rication. These length scales and geometries will
can yield electrode and electrolyte films of nanometer
determine the performance characteristics of 3-D
thicknesses, and in fact, a prototype of the aperiodic
batteries based on these architectures. Thus, before
architecture has been recently reported.20-23 Thus it
attempting to quantify some of the advantages (as
is possible to envision batteries soon in which the
well as the disadvantages) of 3-D cells, a brief
cathode and anode are separated by 10 nm. However,
overview is presented of the methods of fabrication
at these distances, the physical structure of the
currently employed, with emphasis on the resulting
interface must be exceedingly stable to prevent
geometrical and scale differences. Detailed descrip-
shorting of the electrodes. In addition, electrostatic
tions of the fabrication procedures are presented in
interactions between the cathode and anode (i.e.,
section 4.
“overlapping” double layers) and exceedingly fast
At present, only the individual component arrays
transport rates within the separator will undoubtedly
(i.e., individual cathode or anode arrays) of the
influence the current-voltage curves in complex
periodic interdigitated electrode cell (Figure 2a) have
ways that are just now being investigated.24 One key
been fabricated using lithographic methods. The
advantage of these 3-D architectures (and unlike the
results of these limited efforts, however, provide a
periodic interdigitated array) is that the current
rough guide to what is possible in battery design
density across the electrode surfaces is uniform.
within the microlithographic laboratory. Typically,
the electrodes are cylindrical rods with diameters
ranging from 5 to 100 µm and lengths from 10 to 200
2.2. Quantifying the Advantages of 3-D
µm. For instance, Madou and co-workers have fab-
Architectures
ricated an array of cylindrical carbon electrodes with As discussed in section 1, the key advantage
∼10-µm diameter and ∼200-µm length.17 Litho- associated with the proposed 3-D battery structures
graphic fabrication methods also allow other geom- is the ability to achieve large areal energy capacities
etries beyond cylinders to be considered (e.g., trian- without making sacrifices in power density that may
gular rods). It is most probable that the use of current result from slow interfacial kinetics (associated with
lithographic methods will continue to yield structures a small electrode area-to-volume ratio) and/or ohmic
with lateral dimensions (i.e., electrode diameter and potential losses (associated with long transport dis-
electrode-to-electrode spacing) on the micrometer tances). This argument is now quantitatively dem-
length scales, with more emphasis placed on increas- onstrated in the following paragraphs for the 3-D
ing the length of the electrode, L, to achieve high interdigitated electrode array shown in Figure 2a.
aspect ratios. This strategy yields the high areal To facilitate a demonstration of the advantages of
energy capacity that drives the interest in 3-D the 3-D architecture, we quantitatively compare
designs (vide infra). Other periodic arrays, such as metrics related to performance (e.g., areal energy
those based on the use of origami to “unfold” an array capacity, active surface area) of a conventional 2-D
of lithographically designed cathodes and anodes on parallel-plate design with the 3-D interdigitated
a base support18 and of arrays of plates (Figure 2b), array cell (Figure 3). We assume a thin-film 2-D
are also envisioned at the micrometer length scale. battery that comprises a 1-cm2-area anode and
Smaller lateral dimensions will likely be obtained in cathode, each 22.5-µm-thick and separated by a 5-µm-
the future with advances in lithographic methods. thick electrolyte. The total volume of electrodes and
One disadvantage of all periodic interdigitated separator is 5 × 10-3 cm3 (the cell housing is ignored
electrode cells is that the primary current distribu- for simplicity, but is expected to be a comparable
Figure 5. An increase in the height of the interdigitated

Figure 4. Square array of interdigitated anodes and 3-D battery, L, results in increased areal energy capacity
cathodes. and electrode area, without an increase in transport
distances.
percentage of the total volume in both 2-D and 3-D
designs).
It is relatively straightforward to show that a
corresponding 3-D square-array battery (Figure 4)
with the same total volume (i.e., 5 × 10-3 cm3),
constructed from 5-µm-radius cathode and anodes
that are separated (center to center) by 10 µm,
contains ∼39% of the energy capacity of the thin-film
design. Clearly, the lower energy capacity is due to
a higher percentage of the total volume being oc-
cupied by the electrolyte. While the energy capacity Figure 6. Dependence of electrode utilization on electrode
of the 3-D design varies with the shape and arrange- conductivity (σ) and ion diffusivity (D). L and r are the
length and radius of the electrodes, respectively.
ment of the electrodes, it will always be lower than
that of the 2-D design for equal total volumes.
There are several intriguing advantages of the 3-D conductivity of the electrode materials (σ) and the
design that are not reflected in the above numbers. ionic conductivity of the electrodes and electrolyte,
For instance, the active cathode and anode surface in addition to the electrode geometry, as indicated
areas in the 3-D design are 3.5 cm2 each, significantly in Figure 6.
larger than the 2-D design (1 cm2). This difference The dimensionless number U ) (r2/L2)(µ/σ)(1/C),
favors the 3-D design through a reduction in inter- where r is the radius of the electrode, µ is ionic
facial kinetic overpotential, which is inversely pro- mobility of cations (Li+), and C is the volumetric
portional to the electrode area. In addition, the energy capacity (C/cm3), is a quantitative measure
distance that ions must be transported in discharging of uniformity of current across the 3-D electrode
the 2-D battery is 350% larger than in the 3-D design. surfaces and, thus, whether the electrode material
Thus, in principle, the 3-D design is significantly less is uniformly utilized during cell charging and dis-
susceptible to ohmic losses and other transport charging. Decreasing U corresponds to a more uni-
limitations. To achieve equal transport length scales form current distribution along the length of the
in the 2-D design (i.e., by decreasing the electrode electrode. This scenario is a desired one, as long as
thickness to 5 µm) would require a 330% increase in the decrease in U results from high electronic con-
the areal footprint in order to maintain equal cell ductivity in the electrodes (i.e., large σ), rather than
volume, a significant disadvantage in employing low ionic conductivity in the electrolyte (i.e., small
these devices in MEMS and microelectronic applica- µ). Increasing U corresponds to a more nonuniform
tions. discharge of the electrodes, which may result in
While the above comparison of 2-D and 3-D designs underutilization of the electrode materials during
indicates that the 3-D battery has inherently lower rapid discharge, as well as increasing stress along
energy capacity per total cell volume, in fact, the the length of the electrodes. Similarly, U ) (w2/h2)-
capacity of the 3-D design can be increased without (µ/σ)(1/C) for a rectangular electrode in the interdigi-
limit by increasing L, without sacrificing the small tated plate design (Figure 2b), where w is the
areal footprint or high power density. As shown in electrode thickness and h is the electrode height
Figure 5, for the same areal footprint, i.e., 1 cm2, the measured from the base. Clearly, numerical values
above square array 3-D design with L ) 500 µm has of U will depend on electrode geometry and materials.
a capacity that is 350% larger that the 2-D design. The arguments developed in this section apply to
Such a microbattery would contain 222 222 cathodes all of the potential 3-D designs shown in Figure 2.
and 222 222 anodes, with ∼35 cm2 each of active In each design, an increase in areal energy capacity
cathode and anode area! can be obtained by simply increasing L, without any
Clearly L cannot be increased without limit, as the loss of power density due to interfacial kinetics or
ohmic resistance of the electrodes will become suf- slow ion transport. Similar to the interdigitated array
ficiently large to offset the advantages of increased cell, this advantage is eventually limited by the
areal capacity. While this problem has not been electronic conductivity of the electrode materials.
treated systematically, it is clear that the optimized It is worth noting that the 3-D interdigitated plate
value of L will be determined by the electronic design of Figure 2b can be realized by stacking 2-D
Figure 7. Top panels: Schematic diagram of 3-D cylindri-

cal battery arrays in parallel row (left) and alternating Figure 8. (Top panel, left) Schematic diagram of hexago-
anode/cathode (right) configurations. Middle panels: Iso- nal 2:1 cathode/anode battery array. (Top panel, right)
potential lines between cathode (C) and anode (A) for unit Schematic diagram of triangular battery array. (Middle
battery cells. Bottom panel: Current densities (in arbitrary panel) Isopotential lines between cathode (C) and anode
units, a.u.) at the electrode surfaces as a function of the (A) for unit battery cell. (Bottom panel) Current density
angle θ (see middle panel for definition of θ). The area of at the cathode and anode surfaces, plotted on the same a.u.
the cathodes and anodes is equal throughout the diagram. scale used in Figure 7. (Reprinted with permission from
(Reprinted with permission from ref 19. Copyright 2003 ref 19. Copyright 2003 Elsevier.)
Elsevier.)
anodes are placed on a rectangular grid. This ar-
thin-film batteries. However, to maintain the key rangement of electrodes results in high current flow
advantage of low ohmic potential losses, it is neces- between each neighboring cathode/anode pair (i.e.,
sary for the thickness of the film electrodes to be at θ ) 0°), with a relatively steep decrease (∼40%)
greatly reduced. Currently, low-cost manufacturing in current between adjacent cathodes or anodes (θ )
techniques limit film electrodes to thicknesses greater 90°).
than about ∼20 µm. A significantly better design based on the rectan-
gular grid is the alternating cathode/anode configu-
2.3. Current Distribution in 3-D Batteries ration in which each anode is surrounded by four
2.3.1. Interdigitated Electrode Arrays nearest neighbor cathodes (and vice versa). In this
geometry, the higher number of nearest neighbor
All interdigitated electrode arrays suffer from a electrodes of opposite polarity permits a significantly
nonuniform primary current distribution. Figures 7 more uniform primary current density at each elec-
and 8 show examples of the potential and primary trode. However, even in this improved geometry, the
current distributions for representative interdigitated current varies by 20% along the electrode circumfer-
microbattery designs, computed using finite element ence, a limitation that may not be tolerable in some
simulation,25 with the assumption of a uniform cells. The primary current uniformity may be im-
electrolyte conductivity. To allow comparison of the proved by increasing the ratio of the electrode grid
current densities between different battery designs, spacing to the electrode radius in this cell, but at the
all simulations assume identical values of the voltage sake of reducing power density.
between cathode and anode. Current densities are Unlike the conventional battery, a 3-D interdigi-
plotted in the same arbitrary units in each figure, tated array need not contain equal numbers of anodes
allowing direct comparison of the relative power and cathodes. Indeed, there may be situations where
output of different interdigitated geometries. Isopo- battery design is optimized by using unequal number
tential lines within the “unit cell” of each interdigi- densities of cathodes and anodes in order to balance
tated design are also presented. The lengths of the the capacities of the active materials and the kinetics
cathodes and anodes, L, are assumed to be suf- of the charge-transfer reactions. The left-hand side
ficiently long to ignore end effects.19 of Figure 8 shows an example of an interdigitated
Figure 7 illustrates the sensitivity of the current design utilizing twice as many cathodes as anodes.
distribution to electrode placement for two similar Here, each anode is surrounded by six cathodes,
interdigitated designs. In the parallel row design, providing a relatively uniform current density on the
alternating rows of cylindrically shaped cathodes and anode while sacrificing current uniformity at the
cathodes. This design might be useful in a 3-D

microbattery where a uniform current density is
critical at one electrode (e.g., an insertion electrode).
Modern microlithography allows for the fabrication
of essentially any envisioned electrode and cell
geometry. Specifically, electrodes need not have the
cylindrical shape considered in the previous ex-
amples. For example, the close-packed array of
triangular cathodes and anodes shown on the right-
hand side of Figure 8 would be expected to yield
greater cell capacity and increased power. The tradeoff
of this design, obviously, is the highly nonuniform
primary current. Such a geometry may be appropri-
ate in a situation where the net current is limited
by electron-transfer kinetics, and thus the current
distribution is uniform across the electrode surface
regardless of the electrode geometry.
It is clear that while significant increases in both
power and areal energy capacity are obtainable from
3-D interdigitated arrays relative to conventional 2-D
batteries, the inherent difficulty in achieving a
uniform current distribution may limit some devices.
However, interdigitated electrode geometries and Figure 9. (a) Schematic of a closed cell containing LiCl.
configurations not considered above are likely to yield The number of ions (Li+ and Cl-) is assumed to be constant.
current distributions significantly better than the (b) The dependence of the potential profile across the two-
examples described here. ion cell as a function of the electrolyte thickness, d, and
ion concentration.
2.3.2. Plate, Tubule, and Aperiodic 3-D Architectures
will be operative when the electrodes are placed in
The issue of the nonuniform current density is not such close proximity to one another.
important for the plate, tubule, and aperiodic archi- For simplicity, we consider a closed cell such as
tectures shown in Figure 2. The current density in that shown in Figure 9a, which contains LiCl as the
these architectures will be perfectly uniform due to electrolyte. Assuming that the cathode and anode are
the 1-D nature of transport between anode and separated by 10 nm (an experimentally achieved
cathode. Radial transport occurs in the tubular and separation distance), it is instructive to first consider
aperiodic architecture, while planar transport is any direct physical or chemical interactions between
operative in the plate architecture (neglecting end the electrodes. Clearly at 10 nm, electron tunneling
effects). between the surfaces will be negligibly slow, the
decay length of tunneling being ca. 1 Å.26 Thus, the
2.4. Overall Design Issues electrodes are not electronically coupled, and normal
Modeling the overall current-voltage behavior of redox reactions at the cathode and anode are still
3-D cells is a complex function that depends on many necessary for battery operation. At shorter distances,
factors (ionic and electronic conductances, interfacial e.g., 1 nm, the electrodes would spontaneously dis-
reaction kinetics, charge capacity, etc.). Because of charge via direct tunneling between the cathode and
the wide range of possible (and mostly unexplored) anode, a process equivalent to shorting the battery.
architectures, this will be a rich field of study for Thus, there is a theoretical lower limit on the
electrochemical engineers and will likely lead to new thickness of the battery separator.
design concepts in battery technology. Interdigitated Although the cathode and anode are not in elec-
electrode arrays are highly susceptible to stress due tronic communication at 10 nm, their presence is
to both the nonuniform current density and finite known to each other through the electrical fields that
electronic conductivity of the electrodes. Thus, a originate from their respective surface charges. As
delicate balance of the geometrical length scales shown in Figure 9b, the overlap of the electrical
(electrode radius and length, and the spacing be- double layers is more significant as either the dis-
tween electrodes) and materials properties will be tance between the electrodes, d, is reduced or the
required in optimizing any 3-D battery. concentration of the electrolyte is decreased. The role
of the electrolyte concentration is quantitatively
2.5. Cathodes and Anodes Separated by expressed through the Debye length, κ-1,26 which can
Nanometer-Wide Electrolytes be estimated for a 1:1 electrolyte at room temperature
using the expression, κ-1 (nm) ∼0.3/[LiCl]1/2, where
As noted in section 3.1, the fabrication of cells in [LiCl] is the molar concentration (M) of the electro-
which the cathode and anode are separated by a lyte. In essence, κ-1 represents how well the electro-
conformal thin-film electrolyte, perhaps a few tens lyte shields the solution interior from the electrical
of nanometers in thickness, is on the near horizon. charge of the electrode surfaces. A 1 M LiCl solution
Thus, it is interesting to consider potential phenom- corresponds to κ-1 ) 0.3 nm, while a 10 mM solution
ena, not normally considered in battery design, that corresponds to a κ-1 of 3 nm. As a rule of thumb, the
electric potential drops to about 5% of its surface Table 1. Types of Electron-Producing Power Sources
value at a distance from the surface of ∼3κ-1 (i.e., thermal/nuclear/
∼9 nm in a 10 mM LiCl solution). Thus, two elec- electrochemical mechanical-to-electric
trodes separated by 10 nm in a 10 mM LiCl solution batteries thermoelectrics
will clearly be in electrostatic contact through their fuel cells, biofuel cells, pyroelectrics
overlapping double layers. semi-fuel cells
The consequence of overlapping double layers on supercapacitors, thermionics
ultracapacitors
battery operation has only been recently considered, photovoltaics β-cells (nuclear “solar” cells)
and our understanding of this effect is rudimentary.24 harvesting adventitious harvesting adventitious
The electric fields will almost certainly have an effect energy/fuel energy/work
on the transport of ions through the electrolyte, as radioisotope thermoelectric
migration will be important in the strong fields of generators (RTGs)s
R- or β-emitting nuclear sources
the double layers, which can approach or exceed 106
V cm-1. Whether transport of Li+ is impeded or
enhanced by the field will depend on signs of the sources, not all of which are electrochemical. The
electrode charges (determined by the potentials of importance of integrating nanotechnology to improve
zero charge, an ill-defined quantity for most battery these power sources has been discussed, especially
electrode materials) and whether the battery is being for portable applications.30 Some of these sources are
charged or discharged. Modeling of this phenomenon described in greater detail in this section and em-
requires a simultaneous consideration of transport phasize those approaches that have demonstrated
(e.g., Nernst-Planck equation), electrostatics (Pois- power production in centimeter dimensions or smaller
son equation), and statistical thermodynamics (Boltz- sizes. The hybrid micropower supplies discussed at
mann equation), similar to recent modeling of the the end of this section begin to address the issue of
influence of double-layer structure on transport at achieving energy and power capabilities in small
individual nanometer-scale electrodes27 and Levich’s footprint areas.
treatment of the dynamic diffuse layer at macroscopic
planar electrodes.28 3.2. Miniaturizing Microreactors and Fuel Cells
At small electrode separations, the number of ions
in the electrolyte between the electrodes is quite Miniaturization of electrochemical power sources,
small. For instance, the average nearest neighbor in particular batteries and fuel cells, has been
distance between cations in 10 mM LiCl solution is described as a criticalsbut missingscomponent in
∼2 nm. Thus, on average, only a few Li+ ions are transitioning from in-lab capability to the freedom
located between cathode and anode at any position of autonomous devices and systems.29,30 In top-down
on the electrode surfaces, and these few ions carry approaches, macroscopic power sources are scaled to
the current at that surface position. The number of the microlevel usually by the use of fabrication
ions is so small in this situation that the battery can methods, often in combination with new materials.
be well described as a capacitor, with the separator Power generation schemes that can themselves be
being a nonionic dielectric material, resulting in a microfabricated are particularly appealing, as they
linear potential drop between the electrodes (Figure can lead to a one-stop fabrication of device/machine
9b). In addition, even if migration from the electrical function with an integrated power source.
double layers is neglected, the small separation 3.2.1. Compact Mixed Reactors
distance between the electrodes will result in enor-
mous diffusional fluxes. If sufficiently high, these One miniaturization challenge common to fuel cells
fluxes can lead to breakdown of electroneutrality and and batteriessand of critical relevance in designing
formation of a space-charge layer in the electrolyte.27 microscopically and nanoscopically featured 3-D
Much effort in the future will be required to explore batteriessis scaling down the phase that separates
and quantify these ideas. the anode from the cathode. The separator must
prevent direct electronic contact yet permit electrical
contact (ion flux) between the active electrodes.
3. Prior Examples of Small Power Electrodes in rechargeable batteries can undergo
morphological changes that compromise the integrity
3.1. Scope of Small Power of the whole cell, e.g., by establishing a direct
The development of small power sources for por- electronic path (hard shorts), such as the metal
table electronics goes well beyond miniaturization of needles that grow off of a charged-discharged zinc
batteries and fuel cells.4 Other approaches under electrode and may pierce the separator (glass, poly-
development include miniaturizing combustion en- mer, paper) with potentially pyrophoric consequences
gines and thermoelectrics and harvesting energy (one reason the alkaline cell is a primary, use-once,
from ambient sources such as vibration and temper- cell).
ature differences. Most of these approaches are Fabrication issues can arise when thinning the
directed at consumer products and are not designed separator material or in trying to ensure high-
to save real estate. Nonetheless, it is apparent that quality, preferably pinhole-free coverage of dimen-
decreasing the size of power sources and moving to sionally scaled-down anodes and cathodes. One strat-
micro- and even nanoscale power sources offers a egy that can be used to avoid fabricating a separator
number of opportunities.29 at all, at least in electrolyzers and fuel cells, takes
Table 1 lists examples of electron-producing power advantage of decades of work in compact mixed
reactors (CMR), where the specificity of electrocata- demand of 100 mA cm-2.37 The electrode footprint,
lyzed active electrodes permits intermixing of reac- which was smaller than the silicon wafers, was 22.5
tants (or fuel/oxidant).31 The catalytic specificity mm × 22.5 mm. The same authors have also adapted
bypasses the need for a physical barrier, as long as silicon micromachining to fabricate a twin fuel-cell
the electrodes are electronically isolated. Another stack38 using a previously described flip-flop structure
engineering design feature of these single-chamber sandwiched between silicon wafers again acting as
systems is the hydrodynamic flow of the molecular current collectors and flow distributors.34
reactants. Wainwright et al. recently microfabricated a poly-
In 1990, Dyer applied the CMR concept of hydro- mer-based fuel cell with on-board hydrogen storage.
dynamic control of mixed reactant flow to a thin-film One of the key goals of this work is to provide
planar fuel cell.32 The electrical (i.e., ionic) contact independent sizing of power and energy capacity in
between the fuel cell anode and cathode was main- a passive system requiring no fans, pumps, or exter-
tained by diffusion across/through a hydrous alumi- nal humidification.39 The first performance reported
num oxide film (e50-nm-thick) sandwiched between was for a device built on a ceramic substrate (alu-
the electrocatalytic electrodes, with the top Pt elec- mina), but devices built on silicon and polymeric
trode deposited so as to be sufficiently porous for substrates are also mentioned. The microfabricated
gases to pass through to the alumina membrane. The fuel-cell block is coupled to a metal-hydride storage
specific areal power that was generated with a mixed block, printed with inks of LaAl0.3Ni4.7, which gener-
H2/O2 flow was 1-5 mW cm-2 at a cell voltage of 950 ates hydrogen by desorption when placing a 20 mA
mV. By using a lightweight substrate (Kapton) onto load on the fuel cell.39
which the Pt electrodes and hydrous alumina inter-
layer were deposited as thin films, a power density 3.2.3. Membraneless Opportunities with Laminar Flow in
of 100 W kg-1 was reported. Similar results were Microfluidic Reactors
obtained upon replacing the hydrous alumina mem- One way to ease any difficulties that may arise in
brane with the proton-exchange polymer membrane fabricating a membrane, especially in design con-
Nafion, raising the possibility of fabricating confor- figurations that are not planar, is to go membrane-
mal, “open-face” fuel cells on a flexible substrate. less. Recent reports take advantage of the laminar
flow innate to microfluidic reactors40-42 to develop
3.2.2. Microfabrication Approaches for Small Power membraneless fuel cells. The potential of the fuel cell
The nickel-zinc battery provides a good example is established at the boundary between parallel
of how macroscopic power sources can be miniatur- (channel) flows of the two fluids customarily com-
ized via microfabrication methods. A patterned, side- partmentalized in the fuel cell as fuel (anolyte) and
by-side configuration was fabricated with an electro- oxidant (catholyte). Adapting prior redox fuel cell
plated zinc anode and a NiOOH cathode (total cell chemistry using a catholyte of VV/VIV and an anolyte
area of 1 to 5 mm2).33 A photosensitive epoxy served of VIII/VII,43 Ferrigno et al. obtained 35 mA cm-2 at
as the separator and was also used to define the side 1.1 V (for a power density of 38 mW cm-2) at a linear
walls of the cell. The open-circuit voltage of the cells flow rate of 12.5 cm s-1,44,45 which was comparable
was 1.7-1.8 V. The energy density reported for these to that seen with the standard redox fuel cell reported
Ni-Zn microbatteries (2 J cm-2 at 50 mA discharge) by Kummer and Oei;43 the fuel utilization was,
is similar to that available with thin-film batteries, however, much lower. Luo et al. adapted the enzyme-
although the performance does not achieve the catalyzed redox that powers biofuel cells to create a
targeted energy density of 1 J mm-2.13 The cells do membraneless microfluidic fuel cell using a molecular
generate reasonable power levels: at a 1 mA dis- fuel (1,4-dihydrobenzoquinone) and a laccase-cata-
charge rate, the 2-mm2 cells generate 1-2 mW. As lyzed cathode to reduce molecular oxygen. The maxi-
discussed in section 3.5, the Ni-Zn battery is feasible mum power density reported was 80 µA cm-2 at 150
as the energy-storage component in a hybrid mi- mV (8.3 µW cm-2) and a linear flow rate of 1 cm s-1,
cropower system. but the cell could sustain 0.35 V at lower current
Fuel cells incorporating lithographic methods and densities.46
masking/deposition/etching protocols have been fab-
ricated on Si wafers and thereby satisfy two critical
3.2.4. Power in Vivo and Biologically Derived
needs in a standard fuel cell: collection of electrons One area in which miniaturization has progressed
(current collectors) and controlling the flow field of beyond the size regime of interest to the laptop and
fuel and oxidant.34 Kelley et al. produced a miniature cellular telephone industry is to provide in vivo power
direct methanol fuel cell (DMFC) with a current- in biomedical applications.47 An implantable defibril-
voltage and fuel utilization performance that matched lator battery, which must provide pulse power (on
standard-sized DMFCs prepared in-lab.35,36 A work- demand) with high energy density, was achieved by
ing volume for the miniature DMFC of 12 mm3 was coupling two smaller batteries (Li/MnO2 primary
reported, with an operational performance of 822 W battery plus a Li/iodine cell) with a packaged volume
h kg-1 at 70 °C.35 on the order of 10 mL.48 Such power sources must
In more recent reports, a thin-film fuel cell sand- have low self-discharge rates.
wiched between two silicon wafers that had been In an effort to use biological energy transduction
anisotropically etched to form feed holes and chan- to miniaturize a biofuel cell for in vivo applications,
nels for the reactants (H2 and O2) demonstrated a Heller and co-workers have created membraneless,
stable voltage of 0.75 V over 300 h at a current caseless cells that can function under physiological
conditions.49-54 The current collectors are microfibers the thickness of the actual battery is on the order of
of carbon that are catalyzed with enzymes to oxidize 100 µm because of packaging and the critical need
glucose (via glucose oxidase) and reduce O2 (laccase to protect the components from moisture.
or bilirubin oxidase) and wired to their respective The thin-film secondary lithium battery system
carbon current collectors with osmium-based redox that has progressed the furthest is that developed
polymers. The active area of the most recent minia- at the Oak Ridge National Laboratory.13 The elec-
ture biofuel cell is 0.44 mm2, and this glucose oxidase/ trolyte in this case is a lithium phosphorus oxynitride
laccase-catalyzed biofuel cell produced 0.78 V in a pH (“Lipon”) which is deposited by rf magnetron sput-
5 buffer at a power density of 2.68 µW mm-2.53 The tering, as is the cathode, followed by thermal evapo-
improved cell voltage was obtained by using an ration of the lithium. A number of different cathode
osmium redox polymer with a more reducing poten- materials, including layered transition metal ox-
tial, which lowered the overvoltage necessary to drive ides60,61 and vanadium oxides,62 have been reported
the coupled enzyme/cofactor redox to convert the and, for the most part, excellent results have been
glucose fuel to gluconolactone. obtained. The thin-film batteries achieve operating
Recently a new hybrid power source has been lifetimes of thousands of cycles and routinely operate
reported that couples oxidation at a dye-photosensi- at current densities above 1 mA cm-2. In tailoring
tized nanocrystalline semiconducting SnO2 photo- these batteries for integrated circuit (IC) applications,
anode with the enzyme-catalyzed reduction of O2.55 which involve solder reflow at temperatures of 250
Although miniaturization has not yet been reported °C, this group developed a lithium-free approach
for this new hybrid, the developments already where the anode is formed by electroplating lithium
achieved to miniaturize biofuel cells coupled to those on a copper current collector.63
being developed for charge-insertion oxides should be
The research on thin-film batteries has led to
technically transferable to this system.
considerable understanding of the kinetics and struc-
tural changes occurring during lithium insertion and
3.3. Thin-Film Batteries deinsertion in amorphous and nanocrystalline films.13
Miniaturization of batteries has been an active However, the fact remains that these thin-film bat-
area of industrial research because of the need for teries have limited capacity, because of the thickness
portable power sources for such items as watches, of the cathode, while the overall cell resistance can
hearing aids, and cameras. These miniaturization be larger than desired, because of the cathode/
approaches are based, for the most part, on tradi- electrolyte interface. As a result, cell energies are in
tional battery manufacturing, and it is only within the range of 100-500 µW h cm-2. As discussed
the past decade that researchers have begun to previously, these systems are considerably below the
employ fabrication methods that rely on more ad- 1 J mm-2 values required for powering MEMS
vanced material-processing approaches such as vapor devices.
deposition and solution processing. Lithium batteries, An interesting direction to thin-film battery fabri-
in particular, have received considerable attention cation is the use of forward-transfer methods based
because of their high energy density. Two comple- on laser-assisted direct-write processes. This process-
mentary lithium batteries have emerged: one is ing method is a soft transfer that minimally affects
based on polymer electrolytes; the other is based on the materials being laid down, so it is compatible
inorganic electrolytes.15 The former, which is actually with the transfer of metals, oxides, polymers, and
a thick-film system, is the basis for commercial even liquid and gel electrolytes. The variant known
products that provide milliwatt-hour of energy. Pri- as MAPLE-DW (matrix-assisted pulsed-laser evapo-
mary batteries of the type Li/LixMnO2, marketed by ration-direct write)64 has recently been used to direct-
Panasonic among other companies, are on the order write a lithium ion battery.65 The MAPLE-fabricated
of 0.3-0.5-mm-thick and offer between 10 and 20 mA LiCoO2-carbon/carbon microbattery prototype was
h at 3 V. Secondary (i.e., rechargeable) lithium sealed within trilayer metal-polymer laminate and
polymer batteries are also being developed.56 The cycled in air. A typical microbattery with a footprint
polymer electrolyte in most cases is a gel electrolyte of 4 mm × 4 mm (patterns can be written to 10 µm)
formed by polymers and organic solutions of lithium has a capacity of 155 µA h (at a C/5 rate, where a 1
salts or is a plasticized polymer electrolyte.2,57,58 C rate completely discharges the full capacity of the
Inorganic electrolytes lend themselves much more battery in 1 h) and 100 mA h g-1 of LiCoO2. Because
readily to thin-film batteries. Over a decade ago, carbon was included in the transfer ink with the
Eveready Battery developed thin-film Li/TiS2 second- charge-insertion oxide, the electrode could be depos-
ary batteries, which used a sputtered lithium oxysul- ited in thicker layers than a sputter-deposited thin-
fide glass as the electrolyte,59 while other thin-film film battery without creating severe ohmic losses.
batteries using glassy oxides as the electrolyte con- The resulting prototype approaches the power that
tinue to be developed.15 The fact that these inorganic other thin-film batteries provide but did so with a
electrolytes have low conductivity at room tempera- footprint smaller than 1 cm2. An advantage of such
ture is compensated by the short diffusion length of direct-write fabrication is that it can ultimately use
only a few micrometers. The fabricated batteries are the electronics substrate as part of the power-source
truly thin film as various vapor deposition methods packaging to save weight in the battery system.
are used to fabricate the cathode and anode as well Current work is aimed at coupling Li ion microbat-
as the electrolyte. The total thickness of the active teries with energy-harvesting devices (solar cells, RF
components is in the range of 10-20 µm; however, antennae, etc.) for hybrid power supply applications.
Kushida et al. recognized the limitation of the low establish some operating parameters and fabrication
capacity of thin-film lithium batteries for IC applica- methods. For the most part, these methods are
tions when they demonstrated the operation of arrays expected to provide low power, as the first devices
of thin-film lithium batteries.66 Rather than develop produced an output of 3 mW.
a power generator for an entire chip, this work was
directed at providing local power to individual (or a 3.5. Hybrid Micropower
few) VLSI circuits. The battery took the form of a
3 × 1 array in which three parallel 100-µm-wide In the use of 3-D batteries to power MEMS devices,
stripes of cathode material were crossed by a 200- the tacit assumption is that the battery serves as the
µm-wide stripe of lithium. A spin-on phosphosilicate electrochemical power source. That is, the battery
glass was used as the separator, while the cathode directly supplies all the energy for the device. A
material (LiMn2O4) was deposited by solution pro- hybrid power supply is an alternative concept based
cessing in a trench etched in the silicon wafer to on combining energy conversion and energy storage.11
provide better device definition. The prospect of a In this case, the battery is a component of the power
battery array is an interesting one, and although the supply rather than the sole power source. The bat-
energy per unit area was not very high (∼35 µW h teries are expected to provide power when the energy
cm-2), the concept of supplying localized power at the conversion device is not working or when additional
individual circuit level is intriguing. power is required, i.e., pulse power for communica-
tion. The energy conversion device, which obtains
3.4. Generating Power from Ambient Sources energy from ambient sources, is used to power the
microsystem and charge the batteries.
One important area of MEMS sensor technology Koeneman et al. described a micropower supply
is the development of distributed sensors, some of that was designed to power a “smart bearing”.11 The
which are embedded, remote sensors in buildings and ambient energy, a rotating shaft, was converted using
other structures. Battery replacement is difficulty or a wire coil. Energy storage was accomplished using
impractical in this case and approaches for generat- a thin-film lithium battery. Because of the device
ing power from ambient sources are being pursued. application, a suitable actuation driver was required
The most familiar ambient energy source is solar and to deflect the membrane. Although the device was
the use of photovoltaics is widely used. Solar energy, not built, the analysis established that a hybrid
however, is not always available for embedded sen- power supply could be integrated with a MEMS
sors and for this reason another ambient energy device to provide on-board power. One interesting
source of interest for powering small devices is the point in the analysis is that the authors did not
use of mechanical vibration. Beeby and co-workers consider the 2-D nature of the battery and overesti-
reported the use of an electromagnetic transducer mated the number of actuations by 50 times.
based on the movement of a magnetic pole with A more recent paper considered the use of a hybrid
respect to a coil.67 This electromechanical power power supply for powering autonomous microsen-
generator uses neodymium iron boride permanent sors.70 Such devices are similar to the dust mote
magnets mounted on a spring board inside an induc- pictured in Figure 1 and described in the Introduc-
tive coil. As the mass deflects from the vibration, tion. This hybrid power supply combined a solar cell
there is a varying amount of magnetic flux passing to meet standby requirements and to charge the
through the coil. An electromotive force is induced battery, which was the microfabricated Ni-Zn bat-
(Faraday’s law), the magnitude of which is propor- tery described in section 3.2. The authors also showed
tional to the rate of change of the coil position. that if the battery alone was to serve as the power
Shaking the device causes the magnets to vibrate at source, the footprint would be substantially larger,
322 Hz. The device occupies a volume of 240 mm3 over 12 cm2. The hybrid device was constructed and
and generated a maximum useful power of 0.53 mW, some feasibility experiments were carried out, but no
corresponding to a 25-µm amplitude at an excitation actual devices were powered.
frequency of 322 Hz.
Another power generator that converts vibrational
energy to electrical energy is based on a mass-
4. Component Design for 3-D Battery Structures
spring-resonator structure.68 Laser micromachined
copper springs were fabricated into different patterns
4.1. Thinking Like Architects
to optimize device performance. In this device, gen- Jumping out of Flatland71 to a true third dimension
erators with a volume of 1 cm3 were able to produce (rather than stacking 2-D elements, which seems to
up to 4.4 V peak-to-peak with maximum root-mean- be the microlithography community’s perception of
square power of over 800 µW. The vibration required 3-D) permits the use of hierarchical designs: nano-
to generate this power had frequencies ranging from structured electrode materials, organized into larger
60 to 110 Hz with 200-µm amplitude. The fabricated macroscopic features. One of the most critical com-
generator was able to drive an infrared transmitter, ponents of these new battery designs is an initial
and the authors contend that the device is capable degree of “nothingness”, that is a continuous phase
of driving low-power integrated circuits. of porosity. This open volume is then available to be
Another approach to converting vibrations in the filled with a second phase, for example an infiltrated
environment is the use of an inertial generator based electrolyte or even an opposing electrode structure,
on thick-film piezoelectric materials.69 A prototype forming a true 3-D, intermingled battery assembly.
generator was produced enabling the authors to Following IUPAC conventions, pores are identified
by their respective sizes: macropores (>50 nm),

mesopores (2-50 nm), and micropores (<2 nm).72
Nanopore and nanoporous are not internationally
defined terms, so they are an unknown and often
misleading class of pores.
Chemical methods are ideally suited to construct
textured materials with features of solid and void on
the meso- or nanoscale. The chemistry can include
such techniques as sol-gel synthesis, chemical vapor
deposition (CVD), and electrodeposition, where mo-
lecular or ionic precursors are reacted under con-
trolled conditions to assemble the desired solid
structures. Combining such techniques with structure- Figure 10. Schematic showing the conversion of a porous
membrane into a template in which the pores are filled or
directing templates (porous membranes, colloidal coated to form random ensembles or ordered arrays of
crystals, micelles, etc.) permits the design of hierar- nanometer- and micrometer-scale cylinders or tubes. (De-
chical structures with pore sizes that cover a wide rived with permission from ref 87. Copyright 1997 Elec-
range from mesopores to large macropores. In an trochemical Society.)
alternative to template synthesis, sol-gel methods
are used to assemble 3-D networks of nanoscale structures (Figure 10a). The pores in such mem-
particles into a monolithic wet gel. With careful branes may be either random, as in “track-etched”
processing of these gels to remove the pore-filling polycarbonate, or an ordered array, as in electro-
fluid, aerogels and related structures are rendered chemically prepared alumina membranes. The pores
that exhibit aperiodic, through-connected void net- in these membranes can be as small as 10 nm and
works of mesopores and/or macropores. range upward into the hundreds of nanometers and
An extensive research effort is currently underway micrometers.
to produce common battery materials, including Various strategies are used to produce electrode
carbons and transition-metal oxides, in various meso- structures within the membrane pores, including
structured and nanostructured forms. The basic sol-gel synthesis, CVD, electrodeposition, and elec-
functional requirements for these new materials are troless deposition. With careful control of the syn-
largely the same as for conventional secondary bat- thetic conditions, the pores are either filled com-
tery electrodes: a combination of electrical conductiv- pletely or preferentially coated at the pore walls,
ity and the ability to undergo reversible ion-insertion producing hollow tubes (see Figure 10b). Following
reactions. However, the inherent characteristics of infiltration with the desired electrode material, the
nanostructured electrode materials, an extensive membrane is subsequently removed under conditions
electrode/electrolyte interface and facile transport of that do not disturb the active material, leaving an
ions through the void volume, yield superior electro- array of either solid nanofibers or nanotubes attached
chemical performance relative to conventional bat- to a current collector like the bristles of a brush
tery materials. In practically all of the cases dis- (Figure 11). In this case there is very limited inter-
cussed below, electrode materials designed with 3-D connectedness between the nanofibers, except at the
nanostructured interfaces exhibit significantly better current collector base.
charge-storage properties, particularly under high Using this approach, template-synthesized elec-
rates of charge/discharge, than do conventional forms trodes have been prepared not only from metals, but
of the same composition, and in some instances, the from TiO2,75 V2O5,80,81 LiMn2O4,82,83 SnO2,84-86 TiS2,87
nanostructured materials provide unexpectedly high carbon,88-90 and various conducting polymers.91 The
overall capacity due to the defective nature of nano- final templated feature need not consist of a single
scale solids. material or phase. Martin and co-workers prepared
The following section focuses specifically on those striped metal nanorods of gold and platinum,92 which
design strategies that produce either monoliths or was a concept ultimately extended to the preparation
supported films that may then serve as a solid but of nanoscopic barcodes93 and optical tags.94 Recently
textured platform in the first design step toward a Dewan and Teeters infiltrated alumina membranes
3-D battery architecture. to template a device: a V2O5 xerogel-carbon battery
using a poly(ethylene oxide)-lithium triflate electro-
4.2. Porous Membranes as Structure-Directing lyte.95
Many of the templated electrode ensembles de-
Templates scribed above function as high-performance battery
The use of porous membranes as templates for electrodes in lithium-containing electrolytes.76 The
electrode structures was pioneered by Martin and co- fibrous morphology of these templated electrodes
workers nearly 20 years ago,73 and this approach has reduces the solid-state transport distances for lithium
since been extended to include numerous electrode ions participating in the charge-storage reaction. This
compositions and geometries73-78 and applications feature facilitates not only good cyclability but also
beyond energy storage, including sensing and sepa- rapid charging and discharging. For example, tem-
rations.79 In this approach, chemical and electro- plate-synthesized SnO2 electrodes deliver gravimetric
chemical routes are used to fill in the cylindrical, capacities that are orders of magnitude higher than
uniform, unidirectional pores of a free-standing mem- a thin-film control electrode when discharged at
brane with electrochemically active materials and greater than a 50 C rate. The void spaces between
Figure 12. Inverse opal of vanadium oxide ambigel. The

pores are formed by packing 1-µm styrene beads and
infiltrating a vanadium sol. (Reproduced with permission
from ref 100. Copyright 2002 The Royal Society of Chem-
istry.)
CVD, and electrodeposition, depending on the desired

composition. Removal of the colloidal templating
spheres renders a negative replica (the inverse opal)
structure of the active material, with an intercon-
nected, 3-D array of pores, typically sized in the
hundreds of nanometers.
Figure 11. Electron micrographs of membrane-templated
electrodes: (a) ensemble of rods (Reprinted with permission This general procedure for producing macroporous
from ref 80. Copyright 1999 Electrochemical Society.); (b) solids has recently been exploited to synthesize
array of nanotubes (Reprinted with permission from ref 79. electrode architectures that are targeted for lithium
Copyright 2001 American Chemical Society.) battery applications. Sakamoto and Dunn synthe-
sized inverted opal structures of V2O5 by infiltrating
the individual electrode fibers also provide room for a colloidal crystal template with a vanadia sol-gel
expansion during electrode cycling, which is a par- formulation.100 The resulting macroporous structure
ticular problem for SnO2, because a Sn metal phase is shown in Figure 12. The walls are composed of
alloys with electrogenerated Li metal, with accom- vanadium oxide ambigel, leading to a hierarchical
panying large volume changes. architecture that demonstrated high capacity for
In a related approach, these same porous alumina lithium at high discharge rates. Stein and co-workers
membranes serve as a mask through which O2 have expanded this concept further to produce
plasmas are used to etch underlying carbon films.96,97 macroporous ordered structures of V2O5, SnO2, and
This etching process produces honeycomb carbon LiNiO2.98,101 Hierarchical electrode structures are
structures that are positive replicas of the alumina- generated by this method as well, with the inorganic
membrane mask. This process has successfully pro- skeleton of the 3-D ordered macroporous structure
duced honeycomb structures of both diamond96 and composed of fused nanoscale grains of electrode ma-
graphitic carbon,97 with pore sizes in the carbon terial that are themselves mesoporous. Macroporous
replica around 70 nm. carbon structures have also been produced by infil-
4.3. Template Synthesis of Ordered Macroporous trating organic precursors into the interstitial void
Solids with Colloidal Crystals of silica sphere colloidal crystals, followed by carbon-
ization and removal of the silica template.102-106
The porous membrane templates described above
do exhibit three-dimensionality, but with limited 4.4. Template Synthesis of Mesoporous Solids
interconnectedness between the discrete tubelike with Supramolecular Assemblies
structures. Porous structures with more integrated
pore-solid architectures can be designed using tem- These same general templating strategies are
plates assembled from discrete solid objects or su- extended into the mesopore size regime with the use
pramolecular structures. One class of such structures of self-assembled supramolecular arrays, a topic
are three-dimensionally ordered macroporous (or which has been extensively reviewed elsewhere.107-121
3-DOM) solids,98 which are a class of inverse opal In this approach, ionic surfactants or block copoly-
structures.99 The design of 3-DOM structures is based mers are assembled into micellar aggregates or
on the initial formation of a colloidal crystal composed liquid-crystalline phases. With careful tuning of the
of monodisperse polymer or silica spheres assembled reaction conditions, these assemblies can be orga-
in a close-packed arrangement. The interconnected nized into either highly ordered hexagonal, lamellar,
void spaces of the template, ∼26 vol % for a face- or cubic structures, or more disordered wormlike
centered-cubic array, are subsequently infiltrated structures. The microscale phase segregation in these
with the desired material. systems acts to restrict and direct the growth of an
Strategies for filling the void space of the colloidal inorganic guest phase, which can be incorporated by
crystal utilize sol-gel chemistry, salt precipitation, various methods. With transcriptive synthesis the
The synergistic synthesis method for supramolecu-

lar templating was initially developed for silica solids,
but this chemistry is being expanded to include a
number of transition-metal oxides, including some of
interest for electrochemical applications.112,123 Among
these, mesoporous TiO2 has been the most widely
investigated, primarily due to its potential employ-
ment in photovoltaic cells.124 Combining liquid-phase
titania precursor chemistry and templating strategies
Figure 13. Self-assembled organic template (micellar or has produced mesoporous TiO2 structures, with both
liquid crystalline) that directs the assembly of the inorganic ordered and wormlike pore structures.125-133 Grätzel
phase about the structure-directing organic phase. (Cour-
tesy of Prof. Sarah Tolbert, UCLA.)
and co-workers have reported on the reversible
lithium ion insertion reactions of mesoporous anatase
TiO2.127,134,135 Similar strategies have been used to
generate mesoporous WO3, an important electrochro-
mic oxide, where the high surface areas and meso-
porous pathways facilitate rapid coloration/decolor-
ation responses.136,137
Another route to mesoporous, electrochemically
active phases utilizes electrodeposition methods,
where the presence of surfactants or a liquid-crystal-
line phase at the electrode surface directs the forma-
tion of the growing film. The templated electrodepo-
sition approach was initially developed by Attard et
al. to produce mesoporous platinum films.138 This
protocol was followed to develop mesoporous tin as a
potential anode for Li ion batteries and nickel/nickel
oxide electrodes for aqueous-based batteries.139,140
Related strategies have also been used to prepare
mesoporous V2O5141 and conducting polyaniline (as
nanowire arrays).142 In addition to driving the elec-
trodeposition process, the electrochemical interface
can be controlled to promote the formation of hybrid
assemblies of surfactant-inorganic in dilute surfac-
tant solutions en route to forming mesoporous oxide
films, as demonstrated by Stucky et al.143,144
Figure 14. Nanocast carbon. (Reprinted with permission Templated mesoporous carbon structures are gen-
from ref 122. Copyright 1999 American Chemical Society.) erally produced by first incorporating organic precur-
sors into the void space of a preformed mesoporous
supramolecular assembly is first formed separately silica or aluminosilicate template, either by liquid
followed by infiltration of the guest phase. Synergistic infiltration or CVD methods. Following pyrolysis of
synthesis encompasses the process whereby template organic components, the inorganic templates are
self-assembly and inorganic phase synthesis are removed, rendering an inverse carbon replica. This
combined in a cooperative reaction scheme to produce approach was first reported by Ryoo and co-work-
the mesostructured hybrid (Figure 13). ers122 and has subsequently been extensively adapted
In either the transcriptive or synergistic strategy, to generate various mesoporous carbon struc-
removal of the organic template by extraction or tures.104,145-161 The pore-solid architectures of these
calcination renders the inorganic mesoporous struc- carbon solids can be designed on the basis of the
ture. For synthetic schemes that are not compatible choice of inorganic template from among the many
with the formation of stable template assemblies, an available and range from ordered 2-D structures to
alternative approach is to use a preformed, templated more disordered 3-D structures. These mesoporous
inorganic host, such as mesoporous silica, as a mold carbons are promising candidates as electrochemical
to “nanocast” the desired material as an inverse materials in double-layer capacitors. The mesoporous
replica of the host, such as that seen in Figure 14.122 pathways within these electrode structures promote
rapid transport of electrolyte ions during cycling,
In the following section, we restrict our discussion resulting in electrodes that can be charged and
to templated mesoporous solids that are of potential discharged rapidly with only minimal loss of capac-
interest as battery electrodes, including many transi- ity.150,162-164
tion-metal oxides and carbon. This slice of the
literature still points the interested reader to many 4.5. Hierarchical Designs Based on Templating
articles on the synthesis and physical characteriza-
tion of relevant mesoporous materials. A much
Strategies
smaller number of electrochemical studies with tem- Hierarchical structures with pore-solid features
plated mesoporous electrodes have been published, that cover a wide size range within one structure, as
and these studies in particular will be noted. seen in Figure 15, are achieved by carefully combin-
forms than those for comparable aerogels. A fourth

scheme of pore-fluid removal utilizes a freeze-drying
process to render highly porous “cryogels”.176-180
Aerogels and related structures have the important
characteristics of extremely high surface area and a
through-connected, aperiodic network of porosity.
Additionally, these solids can be cast and molded into
a range of shapes and forms or cast as thin films on
planar supports. One attribute that seems to set
these aperiodic architectures apart from the meso-
porous ordered solids, even the nominally 3-D porous
ones, is proven performance in rate-critical applica-
tions. The rapid and facile flux of molecules through
the aerogel architecture avoids the potentially dev-
astating loss of throughput through 1-D channels if
Figure 15. Hierarchical structure formed by combining a blockage occurs.181 Aerogels have already demon-
liquid-phase templating with micromolding. (Reprinted
with permission from ref 166. Copyright 1998 American
strated orders of magnitude faster response for
Association for the Advancement of Science.) sensing, energy storage, and energy conversion than
other pore-solid architectures.167,181
ing the liquid-phase templating approaches described Electrically conductive aerogels have been pre-
above: surfactant templating (up to 10 nm), block pared from a range of transition metal oxides, includ-
copolymer templating (up to 30 nm), and colloidal ing vanadium oxide,172,182-189 manganese oxide,173,174,190
crystal templating (100 nm-1 µm).165,166 This strat- molybdenum oxide,191-193 and tin oxide.194 All of these
egy can be used directly to fabricate inorganic nanostructured oxide forms exhibit improved perfor-
structures or indirectly for hierarchical, porous car- mance as Li ion insertion electrodes in conventional
bon structures that are produced by inverse replica- electrochemical cells.167,195 For example, vanadium
tion of a corresponding hierarchical inorganic mold.161 pentoxide aerogel electrodes accommodate the re-
These synthetic processes can be further combined versible electrochemical insertion of up to four Li ions
with micromolding techniques to create features up per V2O5 unit. In addition to higher overall capacities,
to 1 mm.166 aerogel electrodes deliver more energy (capacity)
under high charge-discharge rates than do conven-
4.6. Aerogels and Related tional materials.
Nanostructures;Aperiodic Pore−Solid Carbon aerogels are typically derived from the
Architectures reaction of resorcinol (or related aryl precursors) and
formaldehyde, which leads to a polymer gel. The
An ordered porous network is not a prerequisite postprocessed polymeric aerogels are then pyrolyzed
in designing high-performance 3-D battery struc- in inert atmospheres at high temperature to form
tures.167 Sol-gel methods can be used to synthesize highly conductive, monolithic carbon structures that
monolithic gels that are composed of a 3-D intercon- retain the characteristics of an aerogel.196-198 The
nected network of nanoscale particles, co-continu- synthesis of carbon aerogels offers considerable flex-
ously intermingled with a phase of fluid-filled pores. ibility as the pore-solid organic architecture can be
The processing of these wet gels to render the dried adjusted by simple changes in the initial reaction
forms must be carefully controlled so that the highly conditions of the polymer gel. Through such varia-
porous architecture is retained. The direct evapora- tions, carbon aerogels can be prepared in either
tion of typical pore-filling fluids (high-surface-tension predominantly mesoporous or macroporous forms.
liquids such as water and alcohols) creates capillary Carbon aerogels are being exploited as electrodes for
forces at the liquid-vapor interface, resulting in pore electrochemical capacitors, where the high surface
collapse and a partially densified porous solid known area, through-connected porosity, and electronic con-
as a xerogel. This structural collapse is avoided if the ductivity of the aerogel promote rapid charge-
pore fluid is removed under supercritical conditions discharge reactions.199-201 Commercial supercapaci-
where no liquid-vapor interface is ever established. tors based on carbon aerogels are available from
Gels processed by supercritical fluid extraction, usu- several companies.
ally with CO2 as the pore fluid, are denoted as
aerogels.167-170 5. Structural Characterization of 3-D Electrode
In an alternative to supercritical drying, the pore
fluids in a wet gel are replaced with a low-surface-
Architectures
tension, nonpolar liquid, such as an alkane, which is
then allowed to evaporate under subcritical or ambi-
5.1. Characterization of the Pore−Solid
ent-pressure conditions.170-172 The resulting solids,
Architecture
denoted as “ambigels” (for ambient-pressure-dried The common feature of the 3-D electrode structures
gels),167 do exhibit a moderate degree of densification described above is the intimate intermingling of solid
but retain a large fraction of pore volume, with pores and void space, in either ordered or aperiodic ar-
usually in the mesopore size range.173-175 The den- rangements, with textures covering an extensive
sification that accompanies ambient-pressure drying range of length scales. This section reviews the
provides for more mechanically rugged monolithic experimental methods used to characterize the pore
size and structure and to determine the electrochemi-

cal properties of the pore-solid nanoarchitectures.
The quality and type of pore structure in such an
architecture often determines the electrochemical
performance of the resulting battery material. For
instance, when the pore structure is infiltrated with
a liquid electrolyte, interconnected mesopores and
macropores provide an open pathway for the diffusion
of electrolyte ions that participate in the charge-
storage mechanisms. The size and interconnected-
ness of the pore network will be crucial in achieving
true intermingled cathode/electrolyte/anode designs.
From the standpoint of electrochemical perfor-
mance, micropores (less than 2 nm) are a far less Figure 16. The pore-size distribution for sol-gel-derived
desirable design component for 3-D power. Even birnessite NaδMnO2·xH2O as processed into three pore-solid
when employed in traditional electrochemical cells, nanoarchitectures: xerogel, ambigel, and aerogel. Distribu-
the electrode area within a micropore is not fully tions are derived from N2 physisorption measurements and
utilized in energy-storage mechanisms, due to the calculated on the basis of a cylindrical pore model. (Re-
lower effective electrolyte conductivity within the printed with permission from ref 175. Copyright 2001
American Chemical Society.)
micropore and sieving effects for larger electrolyte
ions.202,203 Although templated microporous materi-
neities in electron density within porous architec-
als, especially in carbon, have been fabricated and
tures.204,206 Small-angle scattering measurements
characterized for catalytic and adsorbent applica-
extend the range of length scales that can be ana-
tions, we are not including this class of materials for
lyzed, revealing features from micropores to the large
further 3-D design consideration because of the
inability to electrify the surface area within mi- macropores that cannot be effectively probed by gas
cropores on a practical time scale. adsorption. Another distinguishing feature of SAS
techniques is that both open and isolated pores
5.1.1. Physisorption contribute to the experimentally observed structure.
Gas adsorption techniques are often used to char- The results of gas adsorption analysis and such SAS
acterize the pore structure of materials possessing techniques as small-angle X-ray scattering (SAXS)
micropores, mesopores, and small macropores.204 can be combined to provide the most thorough
Adsorption measurements with probes such as N2 characterization of complex pore-solid architectures.
provide information on accessible surface area and A key recent advancement for SAS methods has been
pore volume. In addition, careful analysis of the the development of the chord-length distribution
experimental data also yields information about pore (CLD) approach, which yields much more accurate
sizes, shapes, and the degree of interconnectedness.205 characterizations of pore sizes and distributions than
Various mathematical models can be used to fit the with previous SAS approaches.207,208
experimental data from adsorption isotherms, yield-
ing pore-size distribution (PSD) plots describing the 5.1.3. Electron Microscopy
pore structure in the mesopore and small macropore
regime. The generation of PSD plots is critical in the Electron microscopy techniques are routinely ap-
characterization of structured materials with aperi- plied to structured nanomaterials to assess the solid
odic pore-solid architectures, such as aerogels and morphology for features from tens of micrometers to
ambigels, where it is difficult to directly image the nanometers in size. Traditional microscopy methods
pore structure with microscopic techniques. In Figure provide useful information in the form of 2-D projec-
16, the pore-size distributions are contrasted for tion images, but they do not fully describe the 3-D
birnessite, a Na+-templated, sol-gel-derived lamellar structures discussed in the preceding section. How-
polymorph of MnO2, expressed in three types of ever, the recent advancements in computer-aided
pore-solid nanoarchitectures (xerogel, ambigel, and electron microscopy techniques do provide 3-D infor-
aerogel). The PSD plots make it clear that the pore mation for structured solids.209-212 One such approach
structure in xerogels comprises micropores and small based on high-resolution electron microscopy (HREM)
mesopores, while the distribution in the aerogel is was developed by Terasaki, Stucky, Ryoo and others
weighted toward the larger mesopores and macro- specifically for porous structures exhibiting long-
pores. The ambigel lives up to its name with few range periodicity on the mesoscale (i.e., templated
micropores and macropores and most of its pore mesoporous solids).212-216 In this method, 2-D pro-
volume represented in mesopores sized from 15 to jected HREM images of the sample taken at different
30 nm.175 Reporting just the average pore diameter symmetry orientations are subjected to Fourier trans-
does not begin to capture the richness of the nanoar- form analysis to determine the crystal structure
chitectures: xerogel, 9 nm; ambigel (hexane), 12 nm; factors and ultimately to obtain a 3-D representation
ambigel (cyclohexane), 17 nm; and aerogel, 19 nm. of the ordered structure. The measurements provide
information on the sizes, shapes, and connectivity of
5.1.2. Small-Angle Scattering Techniques the ordered pore structures in such materials as
A second approach to pore-structure analysis, templated mesoporous silica and carbon (see Figure
small-angle scattering (SAS), probes the heteroge- 17). Unfortunately, these methods are not applicable
Figure 18. Schematic drawing depicting SECM measure-

ment of (a) molecular transport within a porous material
and (b) electrochemical activity on one electrode in a
battery array.
5.1.4. Scanning Probe Microscopy

Scanned probe microscopies (SPM) that are capable
of measuring either current or electrical potential are
promising for in situ characterization of nanoscale
energy storage cells. Mass transfer, electrical con-
ductivity, and the electrochemical activity of anode
and cathode materials can be directly quantified by
Figure 17. High-resolution electron micrographic deter- these techniques. Two examples of this class of SPM
mination of cage and bimodal pore structure in the meso- are scanning electrochemical microscopy (SECM) and
porous silicates SBA-6 and SBA-6: (a) 3-D structure of
SBA-6 derived from the electrostatic potential map showing
current-sensing atomic force microscopy (CAFM),
large (A) and small (B) cages with associated large and both of which are commercially available.
small pores; (b) schematic of the large and small cages in In the SECM measurement (Figure 18), a small
the SBA family; (c) pore sizes for calcined SBA-1 deter- microelectrode (typically a metal or carbon electrode)
mined by Horvath-Kawazoe analysis of the argon adsorp- is rastered across the surface of interest, and the
tion isotherm branch. (Reprinted with permission from ref current resulting from a Faradaic reaction is mea-
213. Copyright 2000 Macmillan Magazines, Ltd.)
sured.218-220 The experiment is arranged such that
the tip current is proportional to the local concentra-
for materials with irregular or hierarchical pore- tion of a redox species, which in turn may reflect
solid architectures. molecular transport rates within a porous matrix
A more general approach to 3-D structure deter- (Figure 18a)221 or the electron-transfer activity at an
mination is electron tomography, where the 3-D electrode (Figure 18b).
image of a solid (in a thin slice) is reconstructed from Quantitative theories of various SECM experi-
a series of 2-D images taken at regular tilt angle ments are well-developed and have been applied in
intervals.209 Although not yet a common technique investigations of both electrochemical and nonelec-
for materials science, electron tomography is emerg- trochemical systems,222 including, for example, in-
ing as a powerful tool for imaging complex solids with vestigations of electro-osmotic flow and transport
features approaching the nanometer scale.217 This rates in fuel-cell membranes.223 Because one detects
technique is also more generally applicable to the molecules that are undergoing random diffusion, the
textured materials described in this review, as it does spatial resolution of SECM is not as great as that of
not rely on any long-range order within the sample. scanning tunneling microscopy. However, the nano-
In addition to providing insights into sample scale pore structure in Nafion membranes has been
morphology, electron tomography can also be ex- imaged by this technique, demonstrating better than
ploited to derive 3-D compositional information. For 5-nm spatial resolution.224 And in contrast to STM
instance, energy-filtered transmission electron mi- or AFM, the SECM signal is very specific to local
croscopy (EFTEM) or Z-contrast scanning transmis- chemistry, providing a wealth of information about
sion electron microscopy (STEM) can be combined local concentrations of ions and molecules, chemical
with tomography to provide 3-D elemental mapping kinetics, and transport rates.
capabilities, within some limitations, for light and A complementary microscopy to SECM is CAFM,
heavy elements, respectively.209 In the future, elec- in which the AFM tip is made of a metal such that
tron tomography will be highly beneficial for char- current can be measured as the tip is scanned, in
acterizing the multifunctional architectures that are contact, across the surface.225 Conducting-AFM offers
currently being designed for 3-D power sources. the advantage of directly measuring local electrical
5.2. Corroboration of Electrochemical Processes

with Spectroscopic Techniques
Three-dimensional electrode nanoarchitectures ex-
hibit unique structural features, in the guise of
amplified surface area and the extensive intermin-
gling of electrode and electrolyte phases over small
length scales. The physical consequences of this type
of electrode architecture have already been dis-
cussed,167 and the key components include (i) mini-
mized solid-state transport distances; (ii) effective
mass transport of necessary electroreactants to the
large surface-to-volume electrode; and (iii) magnified
surfacesand surface defectscharacter of the electro-
chemical behavior. This new terrain demands a more
deliberate evaluation of the electrochemical proper-
ties inherent therein.
Many of the traditional electroanalytical methods
can be coupled to spectroscopic measurements, either
ex situ or in situ, to elucidate the electronic and
structural changes that accompany the electrochemi-
cal charge-storage process. The simplest approach
takes advantage of the electrochromic properties of
the various transition-metal oxides and the changes
Figure 19. SECM-AFM images of a polycarbonate mem- that occur upon charge insertion and de-insertion.
brane containing 100-nm-radius pores. (Top) AFM topog- The electrochromic response of thin oxide films
raphy image; (Bottom) SECM image of Ru(NH3)63+ trans-
port in membrane pores. (Reprinted with permission from supported on conductive, transparent substrates
ref 227. Copyright 2001 American Chemical Society.) provides an in situ, temporal indicator of electronic
state changes that occur during ion-insertion reac-
conductivity, which can be inferred only from the tions. The power of spectroelectrochemistry arises
SECM measurement. More importantly, both topog- from the fact that the spectral information, obtained
raphy and surface electronic conductivity can be concurrently with the electrochemically driven elec-
measured simultaneously, allowing the experimen- tronic state changes, decouples from other electro-
talist to correlate structure and reactivity. An ex- chemical phenomena, such as double-layer capaci-
ample of this is the CAFM imaging of electronic tance and electrolyte decomposition, which can
defects in the native oxide on titanium electrodes, potentially interfere with the electrochemical-only
which shows that high conductivity is associated with characterization.228
grain boundaries of the underlying metal.25 Ad- Spectroelectrochemical analysis of charge-insertion
ditional CAFM studies have focused on the investiga- nanostructured materials already offers important
tion of the electrical properties of electrocatalytic Pt insight into these systems. These methods were
nanoparticles deposited on and in a porous Ti/TiO2 recently exploited to characterize the electrochemical
film.226 processes of nanostructured manganese oxide ambi-
A recent development in SPM technology is the gel and xerogel films.175,229 Spectroelectrochemical
combination of SECM and AFM to produce a hybrid measurements were used to corroborate electronic
high-resolution microscope that allows simultaneous state changes with the observed electrochemical
topographic and electrochemical imaging.227 Figure response for the insertion of small cations (Li+, Mg2+)
19 shows an example of this measurement in which and the unexpected insertion of a bulky organic
pore structure and molecular transport of a redox- cation (tetrabutylammonium). Vanadium pentoxide
active molecule (Ru(NH3)63+) were simultaneously exhibits two distinct electrochromic features that can
imaged at ∼1-nm resolution. Inspection of this image be assigned to the VV/VIV transition at either sto-
clearly shows a correlation between transport rates ichiometric V2O5 sites or V sites adjacent to oxygen
and pore structure. vacancy defects.230 Exploiting this phenomenon,
Potential applications of SECM and CAFM in Rhodes et al. were able to assign these different
fundamental studies of nanoscale batteries and bat- physical environments in the voltammetry for sol-
tery materials are readily envisioned. For instance, gel-derived V2O5 (Figure 20) and to track the relative
a small ion-selective electrode can be used as the distributions of these two energetically distinct inser-
SECM tip, providing a means to image and quantify tion sites for sol-gel-derived V2O5 films as a function
Li+ distribution and transport rates. A second ap- of temperature-atmosphere treatments.231,232
plication is mapping the electrostatic potential dis- For more sophisticated spectroelectrochemical analy-
tribution between cathode and anode using a well- sis, electrochemical cycling is coupled with X-ray
poised reference electrode as the scanning tip. absorption spectroscopy (XAS) methods, which probe
Conducting-AFM imaging of battery electrode ma- electronic and local atomic structures with element
trixes comprising electronically insulating electrolyte specificity.233,234 Because XAS techniques are compat-
and a conducting material would allow rapid visu- ible with in situ investigations and do not require
alization of electrical connectivity. long-range structural order for analysis, they are
Figure 20. Spectroelectrochemical analysis of thin films of V2O5 ambigels supported on conductive glass (indium-tin
oxide). The current response is given by the continuous line, and the change in absorbance monitored at 400 and 800 nm
as a function of potential (and time) is shown as individual data points. The V2O5 ambigel was prepared by gelation of
aqueous metavanadate, dried from cyclohexane, and calcined in air at 170 °C. (Printed with permission from ref 232.)
often used to characterize complex battery materials

as a function of electrochemical state.235-239 Sol-gel-
derived V2O5 electrodes are prime candidates for XAS
analysis due to their disordered, nanoscale nature
and exceptional electrochemical properties.240-245 As
noted earlier, V2O5 ambigels and aerogels exhibit
anomalously high capacities for Li ion insertion, with
insertion of up to four Li per V2O5 for electrochemical
lithiation and up to six Li per V2O5 with chemical
lithiation.245 On the basis of the conventional inser-
tion mechanisms, any lithiation beyond x ) 2 for the
formula LixV2O5 should result in the formation of VIII
sites in the oxide lattice. Passerini et al. reported that
for chemically lithiated V2O5, the formal vanadium
oxidation state unexpectedly did not go below the VIV
state, even with lithiation up to Li5.8V2O5, suggesting
charge delocalization beyond the V sites.245 But Figure 21. Visible absorption spectra for a 1.6-µm-thick
NaδMnO2·xH2O birnessite ambigel film in 1 M LiClO4/
Mansour et al. recently described in situ XAS analy- propylene carbonate as a function of electrode potential.
sis of V2O5 aerogels and ambigels, demonstrating that (Reprinted with permission from ref 175. Copyright 2001
with electrochemical cycling the high degree of Li ion American Chemical Society.)
insertion is accompanied by the formation of VIV and
subsequently VIII sites, when the degree of lithiation storage materials is a key area of future research in
surpasses two Li per V2O5.243,244 the drive to improve the performance of batteries and
One concern that might arise in using electronic ultracapacitors.167 As discussed above, the aerogel
spectroscopy to quantify highly defective structures forms of disordered V2O5·0.5H2O afford Li-to-V2O5
is hinted at by results with nanostructured birnessite stoichiometries of 4-6,183,195 while those for the bulk,
NaδMnO2·xH2O. The color center in MnO2 is the Mn4+ crystalline material are 1-2.246 One clue to this
center, so optical density is lost in the visible spec- impressive improvement in the Li ion capacity of
trum as Mn(III) sites are electrogenerated (see Figure V2O5 can be found by inducing deliberate defects in
21). The as-prepared and calcined birnessite ambigel polycrystalline V2O5:247 Creation of proton-stabilized
is mixed valent but predominantly Mn(IV), which can cation-vacancies increases the capacity 23% relative
be quantitatively assayed by chemical redox titra- to the ∼170 mA h g-1 obtained with the as-received,
tion.174,175 The initial visible spectrum, which mea- micrometer-sized polycrystalline V2O5.
sures the electronic state of the Mn centers, has a While XAS techniques focus on direct characteriza-
lower absorbance than the material after electro- tions of the host electrode structure, nuclear magnetic
oxidation, in agreement with the chemically deter- resonance (NMR) spectroscopy is used to probe local
mined mixed-valence state of the initial material. The chemical environments via the interactions of inser-
spectrochemical determination of the Mn(III)/Mn(IV) tion cations that are NMR-active nuclei, for example
ratio, however, does not match that obtained by lithium-6 or -7, within the insertion electrode. As
chemical titration and undercounts the amount of with XAS, NMR techniques are element specific (and
Mn(IV). Because of the high surface-to-volume ratio nuclear specific) and do not require any long-range
and greater defect nature of these nanostructured structural order in the host material for analysis.
materials, we question whether any spectroscopic Solid-state NMR methods are now routinely em-
electronic state measurement truly measures a local- ployed to characterize the various chemical compo-
ized site that remains unaffected by the vacancies nents of Li ion batteries: metal oxide cathodes, Li
that arise in these materials. ion-conducting electrolytes, and carbonaceous an-
Characterizing (experimentally and computation- odes.248,249 Coupled to controlled electrochemical in-
ally) and exploiting the nature of defects in charge- sertion/deinsertion of the NMR-active cations, the
chemical shifts of the 6Li or 7Li are used to distin-

guish between chemically distinct ion insertion sites
within the host electrode structure, to differentiate
electrochemically inactive Li ion sites, and to assess
the mobilities of Li ions. Nuclear magnetic resonance
studies of battery electrodes have recently been
extended to in situ investigations of lithium insertion
in carbon electrodes by Chevallier et al.250
6. Fabrication En Route to 3-D Integration of

Architectural Components
6.1. Micromachining
The three-dimensional electrode arrays that are
the basis for the interdigitated 3-D battery design
(Figure 2a) have been fabricated successfully using
different micromachining methods. Micromachining
has emerged as an indispensable approach for fab-
ricating materials into complex geometries.251 These
methods evolved from the realm of integrated circuits Figure 22. Array of carbon electrodes prepared by the
as researchers developed techniques for deposition, C-MEMS process. The aspect ratio is approximately 20:1.
photolithographic patterning, and selective etching, (Courtesy of Prof. Marc Madou, University of Californias
among other processes. Distinctions can be drawn Irvine.)
between bulk machining, where three-dimensional
features are etched in bulk materials, and surface these pyrolyzed electrodes behave similarly to those
micromachining, where features are built up, layer- of glassy carbon, a standard carbon electrode used
by-layer, on the surface of a substrate. Surface in electrochemical studies. It is evident that good
micromachining also involves the use of sacrificial electrochemical responses are obtained for materials
films that are later dissolved to release the free- pyrolyzed at temperatures above 800 °C, although
standing components. In both types of micromachin- the best properties seem to occur at pyrolysis tem-
ing, photolithography serves as the basis for fabri- peratures above 1000 °C. Voltammetric measure-
cating three-dimensional features. ments show that the electron-transfer kinetics mea-
Three-dimensional electrode arrays have been fab- sured for Fe(CN)63-/4- and Ru(NH3)63+/2+ are compar-
ricated using two very different micromachining able to those determined for other carbon films and
methods. One approach, named carbon MEMS or glassy carbon.254 One interesting difference with
C-MEMS, is based on the pyrolysis of photoresists. glassy carbon is that it is possible to prepare pyro-
The use of photoresist as the precursor material is a lyzed carbon electrodes with lower oxygen content.
key consideration, since photolithography can be used The low oxygen content in the carbon film and its
to pattern these materials into appropriate struc- smooth surface are believed to be responsible for low
tures. The second approach involves the microma- capacitance (8 µF/cm2), a feature that may be impor-
chining of silicon molds that are then filled with tant for analytical applications.
electrode material. Construction of both anode and Although these first studies were directed at 2-D
cathode electrode arrays has been demonstrated interdigitated electrodes (L/d , 1) whose structures
using these microfabrication methods. are not very useful for the proposed 3-D designs, it
is apparent, nonetheless, that this microfabrication
6.1.1. Carbon MEMS (C-MEMS) approach can be adapted for 3-D array electrodes.
The synthesis and electrochemical properties of The recent work reported by Madou and colleagues
carbon films prepared from positive photoresist have demonstrated the types of electrode arrays that
been reported.252-254 The initial direction for this constitute the key design element for the interdigi-
work was the fabrication of carbon interdigitated tated 3-D battery.17 By using a process similar to
electrodes. In this work, positive photoresist was spin their initial work,254 but with a negative photoresist,
coated on a silicon substrate, patterned by photoli- carbon arrays with much higher L/d ratios were
thography, and pyrolyzed to form the carbon elec- prepared, as shown in Figure 22.
trode. In more recent work, laser excitation has been The results by Madou et al. are especially signifi-
used to both pyrolyze the film and to write the cant, as they clearly demonstrate the ability to
electrode pattern.255 fabricate array electrodes with aspect ratios that are
The influence of pyrolysis conditions on the struc- on the order of 20:1. At these aspect ratios, the
ture, morphology, electrical properties, and electroanalytical models indicate that the capacity of 3-D
chemical behavior has been investigated. Raman batteries exceeds that of 2-D geometries. Two key
spectroscopy shows that characteristic sp2 carbon questions remain to be answered for C-MEMS: are
bands form from the pyrolysis treatments. The these electrodes electrochemically reversible to lithium,
electrochemical properties for a few of the electrode and what is the resistance of rods with such high
systems have been reported and, for the most part, aspect ratios?
Figure 23. Processing flow for 3-D electrode array fabrication using silicon micromachining with colloidal filling of the
electrode material. The six steps are identified as the following: (i) patterned photoresist (PR) on silicon substrate, (ii) PR
removal after DRIE micromachining, (iii) insulate silicon mold by oxidation, (iv) colloidal electrode filling material centrifuged
into the mold, (v) silver epoxy added to provide mechanical stability and electrical contact, (vi) the electrode flipped over
and released from the mold by immersion in a TEAOH solution.
This first question concerning electrochemical prop- 6.1.2. Micromachining of Silicon Molds
erties has been answered in a series of half-cell
experiments using the C-MEMS array as the working A second approach for fabricating electrode arrays
electrode and lithium as both the counter and refer- has involved micromachining of silicon molds,257
which are filled with electrode material by colloidal
ence electrode.256 In these experiments, reversible
processing methods. In contrast to C-MEMS, this
intercalation of lithium was demonstrated with C-
fabrication approach is suitable for both anodes and
MEMS array electrodes with an aspect ratio as high cathodes, as one merely alters the composition of the
as 6:1. Voltammetric sweeps indicate that the elec- powders. The process flow for electrode array fabrica-
trochemical behavior is similar to that of coke elec- tion is depicted in Figure 23.
trodes as most of the lithium intercalation occurs
The silicon molds are prepared using photolithog-
below 0.5 V, with a broad deintercalation peak at raphy and deep-reactive ion etching (DRIE). A 12-
0.3 V. Galvanostatic measurements on the C-MEMS µm layer of photoresist is spin-coated onto a silicon
arrays show a large irreversible capacity loss on first wafer and patterned with an array of circles, ranging
discharge followed by good cycling properties, which from 30-120 µm in diameter, 50 µm apart. The UV-
is also consistent with the behavior of coke electrodes. exposed sections are dissolved away with developer
The lithium capacity normalized to the footprint area and 40-120-µm-deep holes in the exposed areas of
of the electrode array is 0.125 mA h cm-2. This value silicon are created by DRIE. From these dimensions,
is nearly twice that of an unpatterned pyrolyzed film it was possible to explore the fabrication of electrode
of SU-8 photoresist.256 The reason for the higher arrays of different L/d ratios. The mold is cleaned in
capacity is due to the greater active volume, contrib- a Piranha bath (H2SO4/H2O2 solution) and a 1 µm
uted by the carbon posts, over the footprint area. thick thermal oxide is grown on the silicon using wet
Gravimetrically, these lithium capacities are within oxidation at 1100 °C. The presence of the oxide layer
the range of values reported for coke electrodes. helps in releasing the infiltrated electrode array.
However, as discussed in the Introduction, the more The second step involves filling the silicon mold
relevant parameter to use in characterizing array with the electrode powders. Several different active
electrodes is the lithium capacity per unit area of the electrode powders were investigated including LiCoO2,
array footprint. carbon black, single-wall carbon nanotubes (SWNT),
and vanadium oxide nanorolls (VONR). The active
The second key questionsthe resistance of carbon
material is mixed with a binder (polyvinylidene
rods with high aspect ratiossis just beginning to be fluoride, PVdF) in order to keep the array intact. The
addressed. The first measurements made on indi- addition of binder is a standard procedure in the
vidual pyrolyzed carbon rods with L/d ) 6 and a fabrication of battery electrodes, as is the addition
diameter of ∼25 µm indicated resistance values on of carbon black to improve the electronic conductivity
the order of 90-100 ohms. These measured values of cathodes.1 The silicon mold was placed in a vial
are reasonably consistent with calculated resistances containing a suspension of the electrode powders in
based on the resistivity values reported for pyrolyzed ethylene glycol and then centrifuged, during which
photoresist. Future work that combines experiments time the powders filled the silicon mold. The mold
with simulations will be able to establish how lithium was removed from the vial and heated to 200 °C to
capacity varies with aspect ratio. evaporate the ethylene glycol and melt the PVdF,
Figure 25. Discharge curves for the first three cycles for
an electrode array composed of vanadium oxide nanorolls
(see Figure 24). The current was 20 µA and the footprint
for the electrode array was 10 mm2.
The micromachining of silicon molds is one key

Figure 24. Electrode array prepared by the powder factor that determines the dimensions of the elec-
processing method shown in Figure 23. The electrode array trode array and, therefore, the energy of the 3-D
structure is intact, although the individual rods contain battery. The use of electrochemical etching in HF to
defects. fabricate high-aspect-ratio trenches in silicon was
introduced in 1990.259 Several groups have applied
thus binding the active electrode powders. Silver this technique in the fabrication of microstructures.260
epoxy was spread over the back of the mold to Recently, Chamran et al. reported the use of this
maintain the mechanical integrity of the array as technique for 3-D microbattery fabricatrion.261 The
well as to provide electrical contact. process flow is shown in Figure 26a. High intensity
The final step in the process is to remove the from the illumination source ensures that the elec-
electrode array from the silicon mold. This releasing tronic hole concentration is sufficiently high to sus-
function is accomplished by immersing the filled mold tain the etching process. The locations of the etched
in an aqueous solution of tetraethylammonium hy- holes are determined by the initiation tips litho-
droxide (TEAOH) heated to 80 °C. As the TEAOH graphically formed on top of the wafer before the
begins to dissolve the silicon, the electrode array photoelectrochemical etching starts. The narrow
separates from the mold, usually as an intact piece. dimension concentrates the local electric field and
An SEM image of an array of vanadium oxide ensures that etching occurs primarily at the tip. The
nanorolls is shown in Figure 24. It is evident from resulting mold has etched holes with L/d ratios >10.
this photomicrograph that the fabrication process is The importance of high, localized fields for electro-
far from perfect. The individual rods have defects chemical micromachining has also been exploited to
including pores and cracks, and the arrays occasion- create precision features below 100 nm by using a
ally have missing rods. sequence of 500-ps pulses.262
The electrochemical properties of the microma- Calculations indicate that 3-D batteries con-
chined electrode arrays have been determined for structed from electrode arrays with the dimensions
several of the electrode materials. In these half-cell shown in Figure 26b will be capable of providing over
experiments, the electrode array served as the work- 2 mW h in a 5-mm3 package, an appropriate size and
ing electrode and lithium as the counter and refer- energy for powering MEMS devices. The holes are 5
ence electrodes. The first experiments were carried µm in diameter by 100-µm-long and the center-to-
out on carbon arrays composed of powders of Ketjen center distance between holes is 10 µm. For such
Black. Reversible intercalation and deintercalation dimensions, it is not known whether the colloidal
of lithium were obtained, and reversible capacities processing approach described above will be effective,
in the range of 0.4-0.5 mA h cm-2 were reported.257 and it may be necessary to develop alternative
Recent results with vanadium oxide nanorolls powder-filling methods.
indicate that array electrodes prepared with these
materials exhibit reversible intercalation of lithium 6.2. 3-D Fabrication Based on 2-D Structures:
and behave in much the same fashion as conventional Origami
VONR electrodes. Discharge curves for the first three
cycles of a VONR array for L/d ) 3 and a footprint Another approach to forming 3-D electrode struc-
area of 10 mm2 are shown in Figure 25. The plateaus tures is based on various folding strategies. This
observed in these experiments are consistent with the “origami” approach was first reported for electronic
results we obtained for well-ordered VONR pow- and optoelectronic systems, as there is great interest
ders.258 Figure 25 shows that the arrays exhibit good in developing three-dimensional electronic networks
cycling properties and the lithium capacity normal- and microelectronic devices. The folding of both
ized to the footprint area is ∼1.5 mA h cm-2. This planar263 and linear264 structures into 3-D configura-
electrode-area-normalized capacity is about 10 times tions has been demonstrated; the adaptation of this
higher than the value reported for thin-film cath- work to electrochemical devices is just starting. Shao-
odes.13 Once again the higher capacity for the foot- Horn et al. reported on the assembly of 3-D electro-
print area is due to the greater volume contributed chemical structures using patterned 2-D electrode
by the VONR posts. sheets.18 The electrodes were designed to fold at
because the anode and cathode must be electrically

isolated. Nonetheless, the use of capillary force to
cause folding is attractive, and it would seem that
there is an opportunity to use other materials with
more compatible electrical properties. Much of the
work done with conducting polymers that serve as
actuators under various applied forces may also be
adaptable to the origami effort.265
7. Advances toward Integration of Active

Components into a 3-D Battery
7.1. Present Status
The construction of an operational, fully 3-D bat-
tery has yet to be achieved. However, as the previous
sections have shown, on an individual basis, most of
the active components for the 3-D designs presented
in Figure 2 have been demonstrated successfully.
Both anode and cathode array electrodes, the design
elements that form the basis of the interdigitated
battery (Figure 2a), have been fabricated and their
electrochemical properties determined. The next key
step is to carefully align these arrays so that inter-
digitation is achieved. The concentric tubule ap-
proach is also well along as carbon honeycomb
electrode structures have recently been prepared.97
These structures represent the continuous electrode
phase (Figure 2c) and can serve as the porous
membrane template, instead of alumina, for growing
electrode materials.74 One critical feature that needs
to be resolved is the presence of an electrolyte that
separates the two electrodes.
Of all the battery designs, it is the continuous
sponge that is furthest along toward integrating
components into a 3-D architecture (Figure 2d). This
Figure 26. (a) Process flow for the fabrication of high-
aspect-ratio holes in silicon as based on photoassisted approach involves the design and fabrication of a
etching: (i) thermal oxide is grown on n-Si and patterned; three-dimensional network from the appropriate
(ii) KOH etching through the holes; (iii) removal of oxide; nanoscale building blocks, including the use of “noth-
and (iv) photoassisted anodic etching of silicon in HF. (b) ing” (void space) and deliberate disorder as design
Cross-section of silicon etched by the photoassisted process components. A recent accomplishment in this area
yielding an array of holes of 5-µm diameter by ∼100-µm- has been the electrodeposition of a conformal elec-
long; the center-to-center distance between holes is 10 µm.
trolyte layer on an aerogel substrate: complete
integration of the electrode and electrolyte in a
specific creases based on the application of Lorentz random 3-D network. The last step in the fabrication
forces. The 3-D structure is then assembled by the of this 3-D battery is to integrate an interpenetrating
sequential folding of the 2-D electrode sheets. This anode.
work is clearly at its inception, as current efforts have
The importance of developing pinhole-free, electro-
only been able to achieve limited folding. Efforts are
lyte films of nanometer thickness is potentially useful
underway to demonstrate a 3-D capacitor structure.
for all 3-D battery designs. For this reason, most of
The attractive feature of the origami approach is this section reviews the synthesis and characteriza-
the prospect of using readily manufacturable 2-D tion of this ultrathin polymer electrolyte with an
patterning to form the electrochemical components. emphasis on topics such as leakage currents and
By proper folding, these 2-D components are then dielectric strength, which become critically important
assembled into 3-D structures. The difficulty to date at the nanoscale. A few comments concerning the
seems to be in the actual folding process. A significant packaging of 3-D batteries are made at the end.
difference between the 3-D electronic structures and
the electrochemical structures is that the former rely 7.2. Electrodeposition of DielectricssSeparators
on capillary forces from molten solder to achieve and Cation-Conductive Electrolytes
folding. As the liquid solder tries to minimize surface
area, it self-assembles the 1-D or 2-D precursors into One of the key steps en route to a 3-D nanoscopic
a 3-D structure. Upon cooling, the solder serves to battery requires fabricating an ultrathin film of a
provide both electrical connection and mechanical polymer separator/electrolyte over chemically stable,
stability to the 3-D electronic network. The solder physically rugged, cation-insertion oxide scaffolds,
approach is not viable for electrochemical devices, such as supported films of MnOx ambigels.20 In
keeping with our architectural metaphor, the fabri-

cation protocol should be capable of painting the
polymer as a conformal, ultrathin coating onto the
walls of the battery building without protruding
deeply into the rooms (and clogging the continuous
porous network). The experiments underway create
ultrathin, electronically insulating, cation-permeable
films on charge-insertion oxide nanoarchitectures via
electropolymerization of aryl monomers. The advan-
tages of electropolymerizing aryl monomers include
the ability to self-limit the growth of the electrogen-
erated polymer, usually by depositing from neutral Figure 27. Schematic for solid-state electrical measure-
or basic electrolytes, to yield thin (usually ∼10 nm ments with ultrathin polymer separators electrodeposited
thick), poorly electron-conductive polymer layers that onto planar indium-tin oxide (ITO) or Au substrates. The
conformally coat all electrifiable surface area. Espe- top electrode makes a soft contact with the polymer by slow
cially appealing in the design of interpenetrating 3-D evaporation of Au or direct contact with a liquid metal
(either Hg or GaIn eutectic) using a micrometer-controlled
batteries, electropolymerization provides a fabrica- syringe to control the approach to contact. Measurements
tion protocol that should be feasible within nanoar- are made in an argon-filled glovebox to minimize effects of
chitectures (unlike line-of-sight chemical vapor depo- O2 and H2O.
sition approaches) and should ensure intimate contact
at the electrode-polymer-electrolyte interface. electrode structures. As mentioned previously, carbon
Such conformal, ultrathin polymer separators must aerogels are attractive electrode structures for elec-
satisfy a range of physical and chemical requirements trochemical capacitors, because their open pore struc-
in order to perform at the level necessary for charge tures, high surface areas, and high electronic con-
insertion on the nanometer scale. These attributes ductivities facilitate rapid charge-discharge reactions.
include (i) highly electronically insulating, preferably However, carbon aerogels ultimately have very lim-
>10-15 S cm-1; (ii) pinhole-free to prevent electronic ited energy densities as electrochemical capacitors,
shorts between the charge-insertion cathode and the since energy storage occurs primarily by the double-
anode; (iii) ionically conducting to the charge-inser- layer capacitance mechanism. This limitation has
tion ion (which in some designs may be a cation other been addressed recently at the Naval Research
than Li+), preferably without the need for solvent or Laboratory by the application of conformal, ultrathin
plasticizing molecules dissolved or partitioned into poly(o-methoxyaniline) (POMA) coatings to conven-
the polymer film; (iv) chemically nondegradable and tional carbon aerogel electrodes. Preliminary experi-
electrochemically stable against Li metal, i.e., over ments indicate that the addition of the pseudocapac-
a 3-4 V (vs Li) range; (v) forms a stable interface itance of the POMA coatings increases the volumetric
with metallic lithium, preferably without forming a capacitance of the hybrid electrode structure by
highly resistive solid-electrolyte interphase layer; and g200%, even at high rates of charge-discharge.21
(vi) exhibits a high dielectric strength, preferably
>106 V cm-1. 7.4. Solid-State Electrical Characterization of
The use of nanoscale (5-30-nm-thick) solid poly- Conformal, Ultrathin Polymer Dielectrics
mer electrolytes should significantly improve rate
capabilities for batteries and other solid-state ionic Although the literature on electrodeposited elec-
devices and represents a considerable improvement troactive and passivating polymers is vast, surpris-
over current micrometer-thick solid electrolytes used ingly few studies exist on the solid-state electrical
in thin-film batteries.13,266 The electro-oxidation of properties of such films, with a focus on systems
o-phenylenediamine in aqueous pH 9 electrolyte onto derived from phenolic monomers,268,269 and appar-
manganese oxide (MnO2) ambigels produces a con- ently none exist on the use of such films as solid
formal polymer film that retains the mesoporosity of polymer electrolytes. To characterize the nature of
the original oxide and is so defect-free over the >200 ultrathin electrodeposited polymers as dielectrics and
m2/g nanoscopic MnO2 network that it prevents electrolytes, solid-state electrical measurements are
reductive dissolution of the MnO2 nanoarchitecture made by electrodeposition of poly(phenylene oxide)
upon immersion in aqueous acid.20 Beyond serving and related polymers onto planar ITO or Au sub-
as a protective coating, the poly(o-phenylenediamine) strates and then using a two-electrode configuration
acts as a selective ion-transport membrane, where with a soft ohmic contact as the top electrode (see
proton transport from the external electrolyte through Figure 27).23 Both dc and ac measurements are taken
the polymer coating is determined by the electro- to determine the electrical and ionic conductivities
chemical state of the polymer, a phenomenon known and the breakdown voltage of the film.
as “electrochemical ion gating”.267 The electronic insulation of these electrodeposited
polymer layers must hold to a two-terminal voltage
7.3. Hybrid Polymer−Aerogel Nanoarchitectures of 4 V if lithium (or lithium ion) anodes are to be used
as Electrochemical Capacitors in the 3-D nanobattery. Because the polymers must
also be thin, high dielectric strengths are critical. As
The recognition that self-limited arylamine-based seen in Table 2, diminishing the thickness of the
polymers could be electrochemically activated in dielectric to the nanoscale exacts a higher standard
aqueous acids has led to a new class of 3-D hybrid in terms of the quality of the dielectric. For example,
Table 2. Required Dielectric Strengths for

Nanometer-Thick Dielectrics
field across dielectric at
dielectric thickness/nm 3 V/(MV cm-1)
5 6
10 3
Figure 29. Schematic of the 1-D transport between an
20 1.5 anode and cathode separated by an ultrathin, conformal
50 0.6 polymer electrolyte.
Table 3. Self-Discharge Times for a 3-D Nanobattery
as a Function of Electrolyte Thickness and Electrical
Conductivity
σelec/S cm-1 thickness, d/nm discharge time/ha
10-9 5 10-5
10-16 5 102
10-16 50 103
a The calculation assumes a cathode structure with 50 µg
of MnO2, 100 cm2 of real electrode surface area, an electrode

capacity of 120 mA h g-1, and a 3 V operating voltage. Leakage
current is calculated from the equation, Ileak ) VσelecA/d, where
V is the voltage and A is the electrode area. Discharge time is
calculated by dividing capacity by the leakage current.
junctions formed at monomolecular layers between

Au contacts.272
Ions can be introduced into these conformal, ul-
trathin polymers by exposure to nonaqueous electro-
lytes. Electrochemical cycling of the MnO2||PPO
hybrid in 1 M LiClO4/CH3CN shows that Li ions are
transported through the ∼30-nm-thick polymer film
to the underlying metal oxide.22 Spectroelectrochemi-
cal measurements verify that the polymer-encapsu-
lated mesoporous MnO2 nanoarchitecture undergoes
the normal cation-insertion reactions with Li ions
Figure 28. Solid-state dc electrical measurements of
polyphenylene oxide in an ITO||PPO||GaIn sandwich. The
from the external electrolyte; the charge under the
thickness of the polymer was measured by tapping-mode voltammetric curves is comparable with and without
atomic force microscopy. (Reprinted with permission from the polymer.
ref 23. Copyright 2004 American Chemical Society.)
7.5. Ultrathin Polymer ElectrolytesLeakage
placing 3 V across a 10 nm thick dielectric requires Currents
a material with a dielectric strength of 3 × 106 V
As discussed in section 2.5, one must be cognizant
cm-1. Dielectrics do exist with that degree of field
of the thickness of the dielectric/electrolyte that is
strength, e.g., quartz at 7 × 106 V cm-1 270 as well as
sandwiched by the 3-D interpenetrating anode and
polymers, e.g., poly(2,6-dimethylphenylene oxide)
cathode. Even with highly electronically insulating
(PDPO) at 2.3 × 106 V cm-1.271
polymers, nanoscale distances that surpass those
With an ITO||PPO||GaIn sandwich and the two-
necessary to sustain quantum tunneling can lead to
electrode configuration shown in Figure 27, a dielec-
leakage currents, arising from small electrical con-
tric strength of g1.4 × 106 V cm-1 is obtained for 32-
ductivities at such short transport path lengths. For
nm-thick electrodeposited poly(phenylene oxide).23
the 1-D transport characteristic of the aperiodic 3-D
This value agrees with an estimated dielectric strength
nanoarchitecture, as shown schematically in Figure
derived from previously reported current-voltage
29, and with a dielectric thickness that falls below
electrical data on 180-nm-thick poly(phenylene ox-
0.1 µm, self-discharge can render 3-D nanobatteries
ide).268 At voltages <4.5 V dc, the ohmic response in
useless, as shown in Table 3. The importance of
Figure 28 is stable; at applied voltages > 4.5 V dc,
negligible electronic conductivity for the dielectric
the current increases 4 orders of magnitude, from a
cannot be overemphasized when the length scales are
few nanoamperes to tens of microamperes. Thus, 4.5
nanoscopic.
V dc is used to calculate an apparent dielectric
strength.
The 10 000-fold increased currents at >4.5 V dc are
7.6. Last Step: Add an Interpenetrating Anode
not due to irreversible breakdown of the polymer Two-thirds of the assembly necessary to form the
dielectric, however, because an ohmic response is sponge architecture for a 3-D battery have now been
again obtained when Vappl < 4.5 V dc.23 It appears demonstrated. By having a suitable pinhole-free,
that a reversible ion or atom migration, derived from thin-to-ultrathin polymer barrier formed over the
the soft contact electrode, may be occurring. This “walls” of the battery architecture, the remaining free
possibility has recently been invoked to explain volume can be filled with a counter electrode, as
phenomena reported for certain molecular electronics described above and depicted schematically in Figure
2d. The resulting architecture will be a solid-state, how power sources of comparable scale will be
fully integrated 3-D battery. The use of mesoporous, fabricated. The 2-D configurations of traditional
high surface area aerogels and ambigels as sub- batteries may not be effective here, despite their high
strates maximizes the interface between the cathode energy density. Instead, energy conversion and har-
and anode, while the electrodeposited polymer mini- vesting approaches may be more suitable for power-
mizes the separation between the cathode and anode. ing microdevices, simply because of the ability to
In contrast to other 3-D designs, this nanostructure provide on-board power. These directions are being
collapses to a uniform 2-D current density distribu- actively investigated.
tion and more effectively utilizes available volume. Three-dimensional batteries offer a different ap-
proach to the portable power field. In this paper we
7.7. Packaging of 3-D Batteries have presented 3-D designs that emphasize power
sources with small areal footprints but do not com-
A key issue that must be addressed in order for promise power and energy density. While this ap-
3-D batteries to become viable for small power proach may not help solve the power needs for cell
applications is packaging. The need to protect the phones and laptop computers, it will have a signifi-
battery in a volumetrically efficient package design cant impact on current and future generations of
without affecting battery materials or chemistry is microdevices. In particular, distributed sensor net-
a significant challenge. As discussed throughout this works and wireless communication systems are
paper, MEMS devices represent an appropriate ap- representative areas where 3-D batteries would be
plication where 3-D batteries could be well-utilized, welcomed enthusiastically because the power sup-
and the prospect of developing a conveniently pack- plies currently in use are many times the size of the
aged power source, especially one attached to the device.
backside of the device so that the real estate available This paper has established some of the design rules
for the micromachine is not reduced (see Figure 1), for 3-D batteries and the accompanying materials
would be an attractive direction. It would seem, and fabrication strategies. The battery materials field
therefore, that packaging approaches for 3-D batter- is already beginning to explore 3-D concepts. Hier-
ies should be those that lead to independently pack- archical designs based on nanostructured materials,
aged components that can be readily integrated into including the deliberate management of void space,
portable power applications. have been organized into larger macroscopic struc-
The packaging of 3-D batteries is likely to extend tures and the first results are impressive, with larger
beyond those designs developed for lithium polymer gravimetric capacities for lithium and higher dis-
and thin-film batteries. The packaging protocols charge rates. It is because of this activity in materials
developed for integrated circuit technologies appear that most of the necessary components for 3-D
to be extremely useful, since many of the objectives batteries are already in hand and the demonstration
of IC packaging are identical to those of batteries.251 of the first operational 3-D batteries is imminent.
Two approaches which are particularly attractive are
the methods developed for single chip packages: chip 9. Acknowledgments
carriers (ceramic or polymer) and TO (transistor This work has been made possible by the support
outline) headers. Both approaches lead to hermetic of the U. S. Office of Naval Research (ONR) and the
sealing and conveniently interface the component to Office of the Secretary of Defense through the Mul-
the “outside world”. One advantage with batteries is tidisciplinary University Research Initiative (MURI)
that the wire bonding requirements for attaching on 3-D Microbatteries (under Grant N00014-01-1-
leads is much less demanding than that of integrated 0757). In particular, the authors would like to thank
circuits. The extensive experience with these ap- three visionary scientific officers at the ONR who
proaches and the wide variety of packaging materials during their various tenures advocated programs to
available ensure that the vital issue of process explore electrochemical power from an integrated 3-D
compatibility with battery chemistry and materials perspective: Robert Nowak, Richard Carlin, and
can be addressed successfully. The resulting pack- Michele Anderson. We also wish to acknowledge the
aged device, a “battery chip” designed to provide contributions of our insightful and productive col-
milliwatt-hour energies in cubic millimeter packages, leagues: Christopher Rhodes [ONR-MURI postdoc-
could be readily integrated on microdevices for por- toral associate (2002-2004)], Michael Doescher
table power. [NRC-NRL postdoctoral associate (2002-2005)],
Rhonda Stroud (NRL), Christopher Smith, Ryan
8. Conclusions Hart, George Baure, Fardad Chamran, Jimmy Lim,
Chai-Won Kwon, and Tim Yeh. We also appreciate
One of the lessons learned in the 1990s was that the efforts of our colleagues who provided personal
the enormous need for high-performance portable communications and allowed us to include some of
power is not diminishing. Consumer electronics their work: Prof. Marc Madou (University of Cali-
continues to be a vibrant, worldwide market force, fornia-Irvine), Prof. Kris Pister (University of Cali-
leading to ever-increasing demands for portable fornia-Berkeley), Prof. C.-J. Kim (UCLA), and Prof.
power. The inability of lithium ion batteries to fully Sarah Tolbert (UCLA).
satisfy consumer electronics has been one of the
principal motivations for the dramatic rise in fuel- 10. References
cell research and development. As the dimensions of (1) Wakihara, M. Mater. Sci. Eng. R-Rep. 2001, 33, 109.
devices continue to shrink, the question arises as to (2) Tarascon, J.-M.; Armand, M. Nature 2001, 414, 359.
(3) Sadoway, D. R.; Mayes, A. M. MRS Bull. 2002, 27, 590. (48) Drews, J.; Wolf, R.; Fehrmann, G.; Staub, R. J. Power Sources
(4) Pescovitz, D. Small Times 2002, 2, 20. 1999, 80, 107.
(5) Mullins, J. IEEE Spectrum 2001, 23. (49) Chen, T.; Barton, S. C.; Binyamin, G.; Gao, Z. Q.; Zhang, Y. C.;
(6) Lin, L. Y.; Wu, M. C.; Sawada, R.; Mohr, J. A. J. Lightwave Kim, H.-H.; Heller, A. J. Am. Chem. Soc. 2001, 123, 8630.
Technol. 2003, 21, 582. (50) Mano, N.; Heller, A. J. Electrochem. Soc. 2003, 150, A1136.
(7) Niarchos, D. Sensors Actuators A-Phys. 2003, 106, 255. (51) Mano, N.; Mao, F.; Heller, A. J. Am. Chem. Soc. 2002, 124,
(8) Tilmans, H. A. C.; De Raedt, W.; Beyne, E. J. Micromech. 12962.
Microeng. 2003, 13, S139. (52) Mano, N.; Mao, F.; Heller, A. J. Am. Chem. Soc. 2003, 125, 6588.
(9) Hartnagel, H. L.; Rodriguez-Girones, M. IEICE Trans. Electron. (53) Mano, N.; Mao, F.; Shin, W.; Chen, T.; Heller, A. Chem.
2003, E86C, 1911. Commun. 2003, 518.
(10) Mehregany, M.; Zorman, C. A. MRS Bull. 2001, 26, 289. (54) Kim, H.-H.; Mano, N.; Zhang, X. C.; Heller, A. J. Electrochem.
(11) Koeneman, P. B.; Busch-Vishniac, I. J.; Wood, K. L. J. Micro- Soc. 2003, 150, A209.
electromech. Sys. 1997, 6, 355. (55) de la Garza, L.; Jeong, G.; Liddell, P. A.; Sotomura, T.; Moore,
(12) Kahn, J. M.; Katz, R. H.; Pister, K. S. J. J. Commun. Networks T. A.; Moore, A. L.; Gust, D. J. Phys. Chem. B 2003, 107, 10252.
2000, 2, 188. (56) Murata, K.; Izuchi, S.; Yoshihisa, Y. Electrochim. Acta 2000, 45,
(13) Bates, J. B.; Dudney, N. J.; Neudecker, B.; Ueda, A.; Evans, C. 1501.
D. Solid State Ionics 2000, 135, 33. (57) Meyer, W. H. Adv. Mater. 1998, 10, 439.
(14) Note that 1 W h ) 3600 J, and a 2 J mm-3 energy density (58) Song, J. Y.; Wang, Y. Y.; Wan, C. C. J. Power Sources 1999, 77,
corresponds to 0.56 mW h mm-3. The MEMS community reports 183.
energy units as J, whereas the battery community typically uses (59) Jones, S. D.; Akridge, J. R. J. Power Sources 1993, 44, 505.
W h, and we will use the established conventions for energy units (60) Neudecker, B. J.; Zuhr, R. A.; Kwak, B. S.; Bates, J. B.;
when discussing the work from these respective communities. Robertson, J. D. J. Electrochem. Soc. 1998, 145, 4148.
(15) Souquet, J. L.; Duclot, M. Solid State Ionics 2002, 148, 375. (61) Wang, B.; Bates, J. B.; Hart, F. X.; Sales, B. C.; Zuhr, R. A.;
(16) Warneke, B.; Last, M.; Liebowitz, B.; Pister, K. S. J. IEEE Robertson, J. D. J. Electrochem. Soc. 1996, 143, 3203.
Comput. Magazine 2001, 34, 44. (62) Bates, J. B.; Dudney, N. J.; Lubben, D. C.; Gruzalski, G. R.;
(17) Wang, C.; Taherabadi, L.; Randhawa, A.; Madou, M.; Kwon, C.- Kwak, B. S.; Yu, X.; Zuhr, R. A. J. Power Sources 1995, 54, 58.
W.; Dunn, B. In Abstracts of the 204th Meeting of the Electro- (63) Neudecker, B. J.; Dudney, N. J.; Bates, J. B. J. Electrochem.
chemical Society; Electrochemical Society: Pennington, NJ, Soc. 2000, 147, 517.
2003; Abstr. #1276. (64) Piqué, A.; Chrisey, D. B.; Auyeung, R. C. Y.; Fitz-Gerald, J.; Wu,
(18) Shao-Horn, Y.; Hidrovo, C.; Jurga, S. M.; Smith, H. I.; Barba- H. D.; McGill, R. A.; Lakeou, S.; Wu, P. K.; Nguyen, V.; Duignan,
stathis, G. In Abstracts of the 204th Meeting of the Electrochemi- M. Appl. Phys. A 1999, 69, S279.
cal Society; Electrochemical Society: Pennington, NJ, 2003; (65) Wartena, R.; Curtright, A. E.; Arnold, C. B.; Piqué, A.; Swider-
Abstr. #1279. Lyons, K. E. J. Power Sources 2004, 126, 193.
(19) Hart, R. W.; White, H. S.; Dunn, B.; Rolison, D. R. Electrochem. (66) Kushida, K.; Kuriyama, K.; Nozaki, T. Appl. Phys. Lett. 2002,
Commun. 2003, 5, 120. 81, 5066.
(20) Long, J. W.; Rhodes, C. P.; Young, A. L.; Rolison, D. R. Nano (67) El-hami, M.; Glynne-Jones, P.; White, N. M.; Hill, M.; Beeby,
Lett. 2003, 3, 1155. S.; James, E.; Brown, A. D.; Ross, J. N. Sensors Actuators A 2001,
(21) Long, J. W.; Dening, B. M.; Rolison, D. R. J. Non-Cryst. Solids 92, 335.
2004, in press. (68) Ching, N. N. H.; Wong, H. Y.; Li, W. J.; Leong, P. H. W.; Wen,
(22) Rhodes, C. P.; Long, J. W.; Doescher, M. S.; Dening, B. M.; Z. Sensors Actuators A 2002, 97-98, 685.
Rolison, D. R. J. Non-Cryst. Solids 2004, in press. (69) Glynne-Jones, P.; Beeby, S. P.; White, N. M. IEEE Proc.-Sci.
(23) Rhodes, C. P.; Long, J. W.; Doescher, M. S.; Rolison, D. R. J. Meas. Technol. 2001, 148, 68.
Phys. Chem. B 2004, 108, in press. (70) Harb, J. N.; LaFollette, R. M.; Selfridge, R. H.; Howell, L. L. J.
(24) Smith, C. P.; White, H. S. Submitted for publication. Power Sources 2002, 104, 46.
(25) Boxley, C. J.; White, H. S.; Gardner, C. E.; Macpherson, J. V. J. (71) Abbott, E. A. The Annotated FlatlandsA Romance of Many
Phys. Chem. B 2003, 107, 9677. Dimensions; Perseus Publishing: Cambridge, MA, 2002.
(26) Bard, A. J.; Faulkner, L. R. Electrochemical Methods, 2nd ed.; (72) Sing, K. S. W.; Everett, D. H.; Haul, R. A. W.; Moscou, L.;
John Wiley: New York, 2001. Pierotti, R. A.; Rouquerol, J.; Siemieniewska, T. Pure Appl.
(27) Smith, C. P.; White, H. S. Anal. Chem. 1993, 65, 3343. Chem. 1985, 57, 603.
(28) Levich, V. G. Physicochemical Hydrodyamics; Prentice Hall: (73) Penner, R. M.; Martin, C. R. J. Electrochem. Soc. 1986, 133, 310.
New York, 1962. (74) Martin, C. R. Chem. Mater. 1996, 8, 1739.
(29) Curtright, A. E.; Bouwman, P. J.; Wartena, R. C.; Swider-Lyons, (75) Lakshmi, B. B.; Patrissi, C. J.; Martin, C. R. Chem. Mater. 1997,
K. E. Int. J. Nanotechnol. 2004, 1, 149. 9, 2544.
(30) Approaches to Combat Terrorism (ACT): Opportunities for Basic (76) Sides, C. R.; Li, N. C.; Patrissi, C. J.; Scrosati, B.; Martin, C. R.
Research, Joint Workshop by The Directorate of the Mathematical MRS Bull. 2002, 27, 604.
and Physical Sciences, NSF and the Intelligence Community; (77) Hulteen, J. C.; Martin, C. R. J. Mater. Chem. 1997, 7, 1075.
National Science Foundation (http://www.mitre.org/public/act/ (78) Martin, C. R.; Mitchell, D. T. In Electroanalytical Chemistry;
10_22_final.pdf), 2003. Bard, A. J., Rubinstein, I., Eds.; Marcel Dekker: New York,
(31) Priestnall, M. A.; Kotzeva, V. P.; Fish, D. J.; Nilsson, E. M. J. 1999; Vol. 21.
Power Sources 2002, 106, 21. (79) Rouhi, A. M. Chem. Eng. News 2001, 79, 29.
(32) Dyer, C. K. Nature 1990, 343, 547. (80) Patrissi, C. J.; Martin, C. R. J. Electrochem. Soc. 1999, 146, 3176.
(33) Humble, P. H.; Harb, J. N.; LaFollette, R. J. Electrochem. Soc. (81) Patrissi, C. J.; Martin, C. R. J. Electrochem. Soc. 2001, 148,
2001, 148, A1357. A1247.
(34) Lee, S. J.; Chang-Chien, A.; Cha, S. W.; O’Hayre, R.; Park, Y. (82) Li, N. C.; Patrissi, C. J.; Che, G. L.; Martin, C. R. J. Electrochem.
I.; Saito, Y.; Prinz, F. B. J. Power Sources 2002, 112, 410. Soc. 2000, 147, 2044.
(35) Kelley, S. C.; Deluga, G. A.; Smyrl, W. H. AICHE J. 2002, 48, (83) Nishizawa, M.; Mukai, K.; Kuwabata, S.; Martin, C. R.; Yone-
1071. yama, H. J. Electrochem. Soc. 1997, 144, 1923.
(36) Kelley, S. C.; Deluga, G. A.; Smyrl, W. H. Electrochem. Solid- (84) Li, N. C.; Martin, C. R. J. Electrochem. Soc. 2001, 148, A164.
State Lett. 2000, 3, 407. (85) Li, N. C.; Martin, C. R.; Scrosati, B. Electrochem. Solid-State
(37) Yu, J.; Cheng, P.; Ma, Z.; Yi, B. J. Power Sources 2003, 124, 40. Lett. 2000, 3, 316.
(38) Yu, J.; Cheng, P.; Ma, Z.; Yi, B. Electrochim. Acta 2003, 48, 1537. (86) Li, N. C.; Martin, C. R.; Scrosati, B. J. Power Sources 2001, 97-
(39) Wainright, J. S.; Savinell, R. F.; Liu, C. C.; Litt, M. Electrochim. 8, 240.
Acta 2003, 48, 2869. (87) Che, G.; Jirage, K. B.; Fisher, E. R.; Martin, C. R.; Yoneyama,
(40) Beebe, D. J.; Mensing, G. A.; Walker, G. M. Annu. Rev. Biomed. H. J. Electrochem. Soc. 1997, 144, 4296.
Eng. 2002, 4, 261. (88) Che, G.; Lakshmi, B. B.; Martin, C. R.; Fisher, E. R.; Ruoff, R.
(41) Quake, S. R.; Scherer, A. Science 2000, 290, 1536. S. Chem. Mater. 1998, 10, 260.
(42) Choban, E. R.; Markoski, L. J.; Stoltzfus, J.; Moore, J. S.; Kenis, (89) Che, G. L.; Lakshmi, B. B.; Fisher, E. R.; Martin, C. R. Nature
P. A. Power Sources Proc. 2002, 40, 317. 1998, 393, 346.
(43) Kummer, J. T.; Oei, D.-G. J. Appl. Electrochem. 1985, 15, 619. (90) Che, G. L.; Lakshmi, B. B.; Martin, C. R.; Fisher, E. R. Langmuir
(44) Ferrigno, R.; Stroock, A. D.; Clark, T. D.; Mayer, M.; Whitesides, 1999, 15, 750.
G. M. J. Am. Chem. Soc. 2002, 124, 12930. (91) Martin, C. R. Acc. Chem. Res. 1995, 28, 61.
(45) Ferrigno, R.; Stroock, A. D.; Clark, T. D.; Mayer, M.; Whitesides, (92) Martin, B. R.; Dermody, D. J.; Reiss, B. D.; Fang, M. M.; Lyon,
G. M. J. Am. Chem. Soc. 2003, 125, 2014. L. A.; Natan, M. J.; Mallouk, T. E. Adv. Mater. 1999, 11, 1021.
(46) Luo, T.-J. M.; Fei, J.; Lim, K. G.; Palmore, G. T. R. In (93) Nicewarner-Peña, S. R.; Freeman, R. G.; Reiss, B. D.; He, L.;
Nanotechnology and the Environment; Karn, B., Masciangioli, Peña, D. J.; Walton, I. D.; Cromer, R.; Keating, C. D.; Natan,
T., Colvin, V. L., Alivisatos, A. P., Eds.; American Chemical M. J. Science 2001, 294, 137.
Society: Washington, DC, 2004; in press. (94) Keating, C. D.; Natan, M. J. Adv. Mater. 2003, 15, 451.
(47) Takeuchi, E. S.; Leising, R. A. MRS Bull. 2002, 27, 624. (95) Dewan, C.; Teeters, D. J. Power Sources 2003, 119, 310.
(96) Masuda, H.; Watanabe, M.; Yasui, K.; Tryk, D.; Rao, T.; (146) Ryoo, R.; Joo, S. H.; Kruk, M.; Jaroniec, M. Adv. Mater. 2001,
Fujishima, A. Adv. Mater. 2000, 12, 444. 13, 677.
(97) Li, N. C.; Mitchell, D. T.; Lee, K. P.; Martin, C. R. J. Electrochem. (147) Joo, S. H.; Jun, S.; Ryoo, R. Microporous Mesoporous Mater.
Soc. 2003, 150, A979. 2001, 44, 153.
(98) Stein, A.; Schroden, R. C. Curr. Opin. Solid State Mater. Sci. (148) Kruk, M.; Jaroniec, M.; Kim, T. W.; Ryoo, R. Chem. Mater. 2003,
2001, 5, 553. 15, 2815.
(99) John, S.; Busch, K. J. Lightwave Technol. 1999, 17, 1931. (149) Kim, T. W.; Park, I. S.; Ryoo, R. Angew. Chem., Int. Ed. 2003,
(100) Sakamoto, J. S.; Dunn, B. J. Mater. Chem. 2002, 12, 2859. 42, 4375.
(101) Stein, A. Adv. Mater. 2003, 15, 763. (150) Lee, J.; Yoon, S.; Oh, S. M.; Shin, C. H.; Hyeon, T. Adv. Mater.
(102) Kang, S.; Yu, J. S.; Kruk, M.; Jaroniec, M. Chem. Commun. 2002, 2000, 12, 359.
1670. (151) Fuertes, A. B.; Nevskaia, D. M. J. Mater. Chem. 2003, 13, 1843.
(103) Gundiah, G.; Govindaraj, A.; Rao, C. N. R. MRS Bull. 2001, 36, (152) Fuertes, A. B.; Nevskaia, D. M. Microporous Mesoporous Mater.
1751. 2003, 62, 177.
(104) Yoon, S. B.; Kim, J. Y.; Yu, J.-S. Chem. Commun. 2001, 559. (153) Yu, J. S.; Kang, S.; Yoon, S. B.; Chai, G. J. Am. Chem. Soc. 2002,
(105) Zakhidov, A. A.; Baughman, R. H.; Iqbal, Z.; Cui, C. X.; 124, 9382.
Khayrullin, I.; Dantas, S. O.; Marti, I.; Ralchenko, V. G. Science (154) Lu, A. H.; Schmidt, W.; Schüth, F. New Carbon Mater. 2003,
1998, 282, 897. 18, 181.
(106) Zakhidov, A. A.; Khayrullin, I. I.; Baughman, R. H.; Iqbal, Z.; (155) Lu, A. H.; Schmidt, W.; Spliethoff, B.; Schüth, F. Adv. Mater.
Yoshino, K.; Kawagishi, Y.; Tatsuhara, S. Nanostructured Mater. 2003, 15, 1602.
1999, 12, 1089. (156) Yoon, S. B.; Kim, J. Y.; Yu, J.-S. Chem. Commun. 2002, 1536.
(107) Lindén, M.; Schacht, S.; Schüth, F.; Steel, A.; Unger, K. K. J. (157) Yang, H. F.; Shi, Q. H.; Liu, X. Y.; Xie, S. H.; Jiang, D. C.; Zhang,
Porous Mater. 1998, 5, 177. F. Q.; Yu, C. Z.; Tu, B.; Zhao, D. Y. Chem. Commun. 2002, 2842.
(108) Liu, T. B.; Burger, C.; Chu, B. Prog. Polym. Sci. 2003, 28, 5. (158) Kim, S. S.; Pinnavaia, T. J. Chem. Commun. 2001, 2418.
(109) Soler-Illia, G. J. d. A. A.; Crepaldi, E. L.; Grosso, D.; Sanchez, (159) Vix-Guterl, C.; Boulard, S.; Parmentier, J.; Werckmann, J.;
C. Curr. Opin. Coll. Interface Sci. 2003, 8, 109. Patarin, J. Chem. Lett. 2002, 1062.
(110) Soler-Illia, G. J. d. A. A.; Sanchez, C.; Lebeau, B.; Patarin, J. (160) Han, B. H.; Zhou, W. Z.; Sayari, A. J. Am. Chem. Soc. 2003,
Chem. Rev. 2002, 102, 4093. 125, 3444.
(111) Schüth, F. Chem. Mater. 2001, 13, 3184. (161) Taguchi, A.; Smått, J. H.; Lindén, M. Adv. Mater. 2003, 15, 1209.
(112) Yang, P. D.; Zhao, D. Y.; Margolese, D. I.; Chmelka, B. F.; Stucky, (162) Yoon, S.; Lee, J. W.; Hyeon, T.; Oh, S. M. J. Electrochem. Soc.
G. D. Chem. Mater. 1999, 11, 2813. 2000, 147, 2507.
(113) Mann, S.; Burkett, S. L.; Davis, S. A.; Fowler, C. E.; Mendelson, (163) Lee, J.; Yoon, S.; Hyeon, T.; Oh, S. M.; Kim, K. B. Chem.
N. H.; Sims, S. D.; Walsh, D.; Whilton, N. T. Chem. Mater. 1997, Commun. 1999, 2177.
9, 2300. (164) Zhou, H. S.; Zhu, S. M.; Hibino, M.; Honma, I. J. Power Sources
(114) Palmqvist, A. E. C. Curr. Opin. Coll. Interface Sci. 2003, 8, 145. 2003, 122, 219.
(115) Gin, D. L.; Gu, W. Q.; Pindzola, B. A.; Zhou, W. J. Acc. Chem. (165) Soten, I.; Ozin, G. A. Curr. Opin. Coll. Interface Sci. 1999, 4,
Res. 2001, 34, 973. 325.
(116) Lin, H. P.; Mou, C. Y. Acc. Chem. Res. 2002, 35, 927. (166) Yang, P. D.; Deng, T.; Zhao, D. Y.; Feng, P. Y.; Pine, D.; Chmelka,
(117) Patarin, J.; Lebeau, B.; Zana, R. Curr. Opin. Coll. Interface Sci. B. F.; Whitesides, G. M.; Stucky, G. D. Science 1998, 282, 2244.
2002, 7, 107. (167) Rolison, D. R.; Dunn, B. J. Mater. Chem. 2001, 11, 963.
(118) Forster, S. In Colloid Chemistry 1; Topics in Current Chemistry; (168) Kistler, S. S. Nature 1931, 127, 741.
Antonietti, M., Ed.; Springer: Berlin, 2003; Vol. 226. (169) Hüsing, N.; Schubert, U. Angew. Chem., Int. Ed. Engl. 1998,
(119) Schüth, F. Angew. Chem., Int. Ed. 2003, 42, 3604. 37, 23.
(120) Ying, J. Y.; Mehnert, C. P.; Wong, M. S. Angew. Chem., Int. Ed. (170) Pierre, A. C.; Pajonk, G. M. Chem. Rev. 2002, 102, 4243.
1999, 38, 56. (171) Harreld, J.; Wong, H. P.; Dave, B. C.; Dunn, B.; Nazar, L. F. J.
(121) Polarz, S.; Antonietti, M. Chem. Commun. 2002, 2593. Non-Cryst. Solids 1998, 225, 319.
(122) Ryoo, R.; Joo, S. H.; Jun, S. J. Phys. Chem. B 1999, 103, 7743. (172) Coustier, F.; Lee, J. M.; Passerini, S.; Smyrl, W. H. Solid State
(123) Yang, P. D.; Zhao, D. Y.; Margolese, D. I.; Chmelka, B. F.; Stucky, Ionics 1999, 116, 279.
G. D. Nature 1998, 396, 152. (173) Long, J. W.; Swider-Lyons, K. E.; Stroud, R. M.; Rolison, D. R.
(124) Grätzel, M. J. Sol-Gel Sci. Technol. 2001, 22, 7. Electrochem. Solid-State Lett. 2000, 3, 453.
(125) Wang, Y. D.; Ma, C. L.; Sun, X. D.; Li, H. D. Appl. Catal. A-Gen. (174) Long, J. W.; Stroud, R. M.; Rolison, D. R. J. Non-Cryst. Solids
2003, 246, 161. 2001, 285, 288.
(126) Wang, Y. D.; Ma, C. L.; Sun, X. D.; Li, H. D. J. Non-Cryst. Solids (175) Long, J. W.; Qadir, L. R.; Stroud, R. M.; Rolison, D. R. J. Phys.
2003, 319, 109. Chem. B 2001, 105, 8712.
(127) Kavan, L.; Rathousky, J.; Grätzel, M.; Shklover, V.; Zukal, A. (176) Yamamoto, T.; Nishimura, T.; Suzuki, T.; Tamon, H. Carbon
Microporous Mesoporous Mater. 2001, 44, 653. 2001, 39, 2374.
(128) Froba, M.; Muth, O.; Reller, A. Solid State Ionics 1997, 101, 249. (177) Xu, J. J.; Yang, J. S. Electrochem. Commun. 2003, 5, 230.
(129) Kluson, P.; Kacer, P.; Cajthaml, T.; Kalaji, M. J. Mater. Chem. (178) Yang, J. S.; Xu, J. J. J. Power Sources 2003, 122, 181.
2001, 11, 644. (179) Tamon, H.; Ishizaka, H.; Yamamoto, T.; Suzuki, T. Drying
(130) Yun, H.-S.; Miyazawa, K.; Zhou, H. S.; Honma, I.; Kuwabara, Technol. 2001, 19, 313.
M. Adv. Mater. 2001, 13, 1377. (180) Tamon, H.; Ishizaka, H.; Yamamoto, T.; Suzuki, T. Carbon 1999,
(131) Yoshitake, H.; Sugihara, T.; Tatsumi, T. Chem. Mater. 2002, 14, 37, 2049.
1023. (181) Rolison, D. R. Science 2003, 299, 1698.
(132) Zheng, J.-Y.; Pang, J.-B.; Qiu, K.-Y.; Wei, Y. J. Mater. Chem. (182) Passerini, S.; Ressler, J. J.; Le, D. B.; Owens, B. B.; Smyrl, W.
2001, 11, 3367. H. Electrochim. Acta 1999, 44, 2209.
(133) Nishimura, S.; Shishido, A.; Abrams, N.; Mallouk, T. E. Appl. (183) Salloux, K.; Chaput, F.; Wong, H. P.; Dunn, B.; Breiter, M. W.
Phys. Lett. 2002, 81, 4532. J. Electrochem. Soc. 1995, 142, L191.
(134) Krtil, P.; Fattakhova, D.; Kavan, L.; Burnside, S.; Grätzel, M. (184) Zhang, F.; Passerini, S.; Owens, B. B.; Smyrl, W. H. Electrochem.
Solid State Ionics 2000, 135, 101. Solid-State Lett. 2001, 4, A221.
(135) Kavan, L.; Attia, A.; Lenzmann, F.; Elder, S. H.; Grätzel, M. J. (185) Chaput, F.; Dunn, B.; Fuqua, P.; Salloux, K. J. Non-Cryst. Solids
Electrochem. Soc. 2000, 147, 2897. 1995, 188, 11.
(136) Ozkan, E.; Lee, S. H.; Liu, P.; Tracy, C. E.; Tepehan, F. Z.; Pitts, (186) Coustier, F.; Passerini, S.; Smyrl, W. H. J. Electrochem. Soc.
J. R.; Deb, S. K. Solid State Ionics 2002, 149, 139. 1998, 145, L73.
(137) Cheng, W.; Baudrin, E.; Dunn, B.; Zink, J. I. J. Mater. Chem. (187) Dong, W.; Rolison, D. R.; Dunn, B. Electrochem. Solid-State Lett.
2001, 11, 92. 2000, 3, 457.
(138) Attard, G. S.; Bartlett, P. N.; Coleman, N. R. B.; Elliott, J. M.; (188) Harreld, J. H.; Dong, W.; Dunn, B. MRS Bull. 1998, 33, 561.
Owen, J. R.; Wang, J. H. Science 1997, 278, 838. (189) Le, D. B.; Passerini, S.; Guo, J.; Ressler, J.; Owens, B. B.; Smyrl,
(139) Nelson, P. A.; Owen, J. R. J. Electrochem. Soc. 2003, 150, A1313. W. H. J. Electrochem. Soc. 1996, 143, 2099.
(140) Nelson, P. A.; Elliott, J. M.; Attard, G. S.; Owen, J. R. Chem. (190) Passerini, S.; Coustier, F.; Giorgetti, M.; Smyrl, W. H. Electro-
Mater. 2002, 14, 524. chem. Solid-State Lett. 1999, 2, 483.
(141) Liu, P.; Lee, S. H.; Tracy, C. E.; Yan, Y. F.; Turner, J. A. Adv. (191) Dong, W.; Dunn, B. J. Non-Cryst. Solids 1998, 225, 135.
Mater. 2002, 14, 27. (192) Dong, W.; Dunn, B. J. Mater. Chem. 1998, 8, 665.
(142) Huang, L. M.; Wang, Z. B.; Wang, H. T.; Cheng, X. L.; Mitra, (193) Dong, W.; Mansour, A. N.; Dunn, B. Solid State Ionics 2001,
A.; Yan, Y. X. J. Mater. Chem. 2002, 12, 388. 144, 31.
(143) Baeck, S. H.; Choi, K. S.; Jaramillo, T. F.; Stucky, G. D.; (194) Harreld, J. H.; Sakamoto, J.; Dunn, B. J. Power Sources 2003,
McFarland, E. W. Adv. Mater. 2003, 15, 1269. 115, 19.
(144) Choi, K. S.; Lichtenegger, H. C.; Stucky, G. D.; McFarland, E. (195) Owens, B. B.; Passerini, S.; Smyrl, W. H. Electrochim. Acta 1999,
W. J. Am. Chem. Soc. 2002, 124, 12402. 45, 215.
(145) Jun, S.; Joo, S. H.; Ryoo, R.; Kruk, M.; Jaroniec, M.; Liu, Z.; (196) Bock, V.; Emmerling, A.; Fricke, J. J. Non-Cryst. Solids 1998,
Ohsuna, T.; Terasaki, O. J. Am. Chem. Soc. 2000, 122, 10712. 225, 69.
(197) Schaefer, D. W.; Pekala, R.; Beaucage, G. J. Non-Cryst. Solids (236) Balasubramanian, M.; Sun, X.; Yang, X. Q.; McBreen, J. J. Power
1995, 186, 159. Sources 2001, 92, 1.
(198) Tamon, H.; Ishizaka, H.; Mikami, M.; Okazaki, M. Carbon 1997, (237) Yoon, W. S.; Grey, C. P.; Balasubramanian, M.; Yang, X. Q.;
35, 791. McBreen, J. Chem. Mater. 2003, 15, 3161.
(199) Frackowiak, E.; Beguin, F. Carbon 2001, 39, 937. (238) Yoon, W. S.; Kim, N.; Yang, X. Q.; McBreen, J.; Grey, C. P. J.
(200) Li, W. C.; Reichenauer, G.; Fricke, J. Carbon 2002, 40, 2955. Power Sources 2003, 119, 649.
(201) Pekala, R. W.; Farmer, J. C.; Alviso, C. T.; Tran, T. D.; Mayer, (239) Hwang, B. J.; Tsai, Y. W.; Santhanam, R.; Liu, D. G.; Lee, J. F.
S. T.; Miller, J. M.; Dunn, B. J. Non-Cryst. Solids 1998, 225, J. Electrochem. Soc. 2003, 150, A335.
74. (240) Giorgetti, M.; Berrettoni, M.; Passerini, S.; Smyrl, W. H.
(202) Koresh, J.; Soffer, A. J. Electrochem. Soc. 1977, 124, 1379. Electrochim. Acta 2002, 47, 3163.
(203) Salitra, G.; Soffer, A.; Eliad, L.; Cohen, Y.; Aurbach, D. J. (241) Giorgetti, M.; Passerini, S.; Smyrl, W. H.; Mukerjee, S.; Yang,
Electrochem. Soc. 2000, 147, 2486. X. Q.; McBreen, J. J. Electrochem. Soc. 1999, 146, 2387.
(204) Kaneko, K. J. Membr. Sci. 1994, 96, 59. (242) Giorgetti, M.; Mukerjee, S.; Passerini, S.; McBreen, J.; Smyrl,
(205) Webb, P. A.; Orr, C. Analytical Methods in Fine Particle W. H. J. Electrochem. Soc. 2001, 148, A768.
Technology; Micromeritics Instrument Corp.: Norcross, GA, (243) Mansour, A. N.; Smith, P. H.; Baker, W. M.; Balasubramanian,
1997. M.; McBreen, J. Electrochim. Acta 2002, 47, 3151.
(206) Fratzl, P. J. Appl. Crystallogr. 2003, 36, 397.
(244) Mansour, A. N.; Smith, P. H.; Baker, W. M.; Balasubramanian,
(207) Göltner, C. G.; Smarsly, B.; Berton, B.; Antonietti, M. Chem.
M.; McBreen, J. J. Electrochem. Soc. 2003, 150, A403.
Mater. 2001, 13, 1617.
(208) Smarsly, B.; Antonietti, M.; Wolff, T. J. Chem. Phys. 2002, 116, (245) Passerini, S.; Le, D. B.; Smyrl, W. H.; Berrettoni, M.; Tossici,
2618. R.; Marassi, R.; Giorgetti, M. Solid State Ionics 1997, 104, 195.
(209) Midgley, P. A.; Weyland, M. Ultramicroscopy 2003, 96, 413. (246) Livage, J. Solid State Ionics 1996, 86-88, 935.
(210) Terasaki, O.; Ohsuna, T.; Liu, Z.; Kaneda, M.; Kamiya, S.; (247) Swider Lyons, K. E.; Love, C. T.; Rolison, D. R. Solid State Ionics
Carlsson, A.; Tsubakiyama, T.; Sakamoto, Y.; Inagaki, S.; Che, 2002, 152-153, 99.
S.; Tatsumi, T.; Camblor, M. A.; Ryoo, R.; Zhao, D.; Stucky, G.; (248) Grey, C. P.; Greenbaum, S. G. MRS Bull. 2002, 27, 613.
Shindo, D.; Hiraga, K. In Nanoporous Materials III, Studies in (249) Grey, C. P.; Lee, Y. J. Solid State Sci. 2003, 5, 883.
Surface Science and Catalysis; Sayari, A., Jaroniec, M., Eds.; (250) Chevallier, F.; Letellier, M.; Morcrette, M.; Tarascon, J.-M.;
Elsevier: Amsterdam, 2002; Vol. 141. Frackowiak, E.; Rouzaud, J. N.; Beguin, F. Electrochem. Solid-
(211) de Jong, K. P.; Koster, A. J. Chem. Phys. Chem. 2002, 3, 776. State Lett. 2003, 6, A225.
(212) Thomas, J. M.; Terasaki, O.; Gai, P. L.; Zhou, W. Z.; Gonzalez- (251) Madou, M. Fundamentals of Microfabrication; CRC Press: Baton
Calbet, J. Acc. Chem. Res. 2001, 34, 583. Rouge, LA, 1997.
(213) Sakamoto, Y. H.; Kaneda, M.; Terasaki, O.; Zhao, D. Y.; Kim, J. (252) Kim, J.; Song, X.; Kinoshita, K.; Madou, M.; White, B. J.
M.; Stucky, G.; Shim, H. J.; Ryoo, R. Nature 2000, 408, 449. Electrochem. Soc. 1998, 145, 2314.
(214) Sakamoto, Y.; Dı́az, I.; Terasaki, O.; Zhao, D. Y.; Pérez-Pariente, (253) Kostecki, R.; Song, X. Y.; Kinoshita, K. J. Electrochem. Soc. 2000,
J.; Kim, J. M.; Stucky, G. D. J. Phys. Chem. B 2002, 106, 3118. 147, 1878.
(215) Kaneda, M.; Tsubakiyama, T.; Carlsson, A.; Sakamoto, Y.; (254) Ranganathan, S.; McCreery, R.; Majji, S. M.; Madou, M. J.
Ohsuna, T.; Terasaki, O.; Joo, S. H.; Ryoo, R. J. Phys. Chem. B Electrochem. Soc. 2000, 147, 277.
2002, 106, 1256. (255) Kostecki, R.; Song, X. Y.; Kinoshita, K. Electrochem. Solid-State
(216) Carlsson, A.; Kaneda, M.; Sakamoto, Y.; Terasaki, O.; Ryoo, R.; Lett. 2002, 5, E29.
Joo, S. H. J. Electron Microsc. 1999, 48, 795. (256) Wang, C.; Taherabadi, L.; Jia, G.; Madou, M.; Yeh, Y.; Dunn, B.
(217) Fernandez, J. J.; Lawrence, A. F.; Roca, J.; Garcia, I.; Ellisman, Electrtochem. Solid-State Lett. 2004, in press.
M. H.; Carazo, J. M. J. Struct. Biol. 2002, 138, 6. (257) Baure, G.; Kwon, C.-W.; Lee, G. G.; Chamran, F.; Kim, C.-J.;
(218) Engstrom, R. C. Anal. Chem. 1984, 56, 890. Dunn, B. In Micropower and Microdevices; Brandon, E. J., Ryan,
(219) Bard, A. J.; Fan, F.-R. F.; Mirkin, M. V. Anal. Chem. 1989, 61, A., Harb, J., Ulrich, R., Eds.; Electrochemical Society: Penning-
132. ton, NJ, 2002; Vol. PV2002-25.
(220) Bard, A. J.; Fan, F.-R. F.; Mirkin, M. V. In Electroanalytical (258) Sun, D.; Kwon, C.-W.; Baure, G.; Richman, E.; MacLean, J.;
Chemistry; Bard, A. J., Ed.; Marcel Dekker: New York, 1994; Dunn, B.; Tolbert, S. H. Submitted for publication.
Vol. 18. (259) Lehmann, V.; Foll, H. J. Electrochem. Soc. 1990, 137, 653.
(221) Bath, B. D.; White, H. S.; Scott, E. R. Imaging molecular (260) Foll, H.; Christophersen, M.; Carstensen, J.; Hasse, G. Mater.
transport across membranes; John Wiley: New York, 2001. Sci. Eng. R-Rep. 2002, 39, 93.
(222) Scanning Electrochemical Microscopy; Bard, A. J.; Mirkin, M.
(261) Chamran, F.; Christophersen, M.; Kim, C.-J. In Abstracts of the
V., Eds.; John Wiley: New York, 2001.
204th Meeting of the Electrochemical Society; Electrochemical
(223) Bath, B. D.; White, H. S.; Scott, E. R. Anal. Chem. 2000, 72,
Society: Pennington, NJ, 2003; Abstr. #1292.
433.
(224) Fan, F.-R. F.; Bard, A. J. Science 1995, 270, 1849. (262) Kock, M.; Kirchner, V.; Schuster, R. Electrochim. Acta 2003, 48,
(225) Miki, T.; Yanagi, H. Langmuir 1998, 14, 3405. 3213.
(226) Macpherson, J. V.; Gueneau de Mussy, J.-P.; Delplancke, J.-L. (263) Gracias, D. H.; Kavthekar, V.; Love, J. C.; Paul, K. E.; White-
Electrochem. Solid-State Lett. 2001, 4, E33. sides, G. M. Adv. Mater. 2002, 14, 235.
(227) Macpherson, J. V.; Jones, C. E.; Barker, A. L.; Unwin, P. R. Anal. (264) Whitesides, G. M.; Boncheva, M. Proc. Natl. Acad. Sci. U.S.A.
Chem. 2002, 74, 1841. 2002, 99, 4769.
(228) Heineman, W. R. J. Chem. Educ. 1983, 60, 305. (265) Smela, E. Adv. Mater. 2003, 15, 481.
(229) Long, J. W.; Young, A. L.; Rolison, D. R. J. Electrochem. Soc. (266) Jang, Y. I.; Dudney, N. J.; Blom, D. A.; Allard, L. F. J.
2003, 150, A1161. Electrochem. Soc. 2002, 149, A1442.
(230) Talledo, A.; Granqvist, C. G. J. Appl. Phys. 1995, 77, 4655. (267) Burgmayer, P.; Murray, R. W. J. Am. Chem. Soc. 1982, 104,
(231) Rhodes, C. P.; Dong, W.; Long, J. W.; Rolison, D. R. In Solid 6139.
State Ionics IV; Wachsman, E. D., Swider Lyons, K. E., Carolan, (268) Dubois, J.-E.; Tourillon, G.; Pham, M.-C.; Lacase, P.-C. Thin
M. F., Garzon, F. H., Liu, M., Stetter, J. R., Eds.; Electrochemical Solid Films 1980, 69, 141.
Society: Pennington, NJ, 2003; Vol. PV2002-26. (269) McCarley, R. L.; Thomas, R. E.; Irene, E. A.; Murray, R. W. J.
(232) Dong, W.; Long, J. W.; Rolison, D. R. Unpublished data, 2001, Electroanal. Chem. 1990, 290, 79.
Naval Research Laboratory. (270) Whitehead, S. Dielectric Breakdown of Solids; Oxford University
(233) Parsons, J. G.; Aldrich, M. V.; Gardea-Torresdey, J. L. Appl. Press: Oxford, UK, 1953.
Spec. Rev. 2002, 37, 187. (271) Adohi, B.; Gosse, J. P.; Gosse, B. J. Phys. III 1991, 1, 1623.
(234) de Groot, F. Chem. Rev. 2001, 101, 1779. (272) Service, R. F. Science 2003, 302, 556.
(235) McBreen, J.; Balasubramanian, M. JOM-J. Miner. Met. Mater.
Soc. 2002, 54, 25. CR020740L
Chem. Rev. 2004, 104, 4493−4512 4493
NMR Studies of Cathode Materials for Lithium-Ion Rechargeable Batteries

Clare P. Grey* and Nicolas Dupré
Department of Chemistry, State University of New York at Stony Brook, Stony Brook, New York 11794-3400
Received February 24, 2004
Contents
2. NMR Background 4494
2.1. Lithium NMR Spectra of Cathode Materials: 4494
Introduction
2.2. NMR Spectra of Paramagnetic Materials 4494
2.2.1. Fermi-Contact Interaction 4496
2.2.2. Dipolar Coupling 4496
3. Extracting Chemical Information from the Spectra 4497
of Paramagnetic Materials
3.1. Fermi-Contact Interaction 4497
3.2. Dipolar Interaction 4500
4. Applications of NMR Spectroscopy to the Study 4501
of Cathode Materials
Clare P. Grey received her B.A. (1987) and D. Phil. (1990) degrees in
4.1. Spinels 4501 Chemistry from the University of Oxford. At Oxford she worked with
4.1.1. Cation-Doped Spinels 4502 Professors Tony Cheetham and Christopher Dobson on the application
4.2. Cr3+- and Ni2+-Substituted Layered Lithium 4504 of solid-state NMR to problems in solid-state chemistry. She then spent
Manganates a year as a postdoctoral fellow at the University of Nijmegen in The
Netherlands with Professor Wiebren Veeman, where she developed new
4.3. LiCoO2 and Related Materials 4505 NMR methods for measuring internuclear distances in systems with
4.4. Lithium Phosphates 4506 quadrupolar nuclei. She was a visiting scientist at the DuPont Experimental
4.5. Vanadates 4507 Station in Wilmington (1992−1994), where she worked with Dr. Alexander
4.5.1. 51V NMR 4507 Vega on NMR theory and on the application of NMR to molecular sieves
and inorganic−organic composites. She joined the faculty at SUNY Stony
4.5.2. NMR Studies of Vanadium Oxides 4507 Brook in 1994 as Assistant Professor and was promoted to Full Professor
4.5.3. Vanadium Phosphates 4509 in 2001. She uses solid-state NMR spectroscopy, in combination with
5. Conclusions 4510 other characterization techniques such as diffraction, to understand the
6. Acknowledgments 4510 role that local structure plays in controlling the physical properties of a
wide range of materials. Current studies include the investigation of
7. References 4510 electrode materials for lithium-ion rechargeable batteries, anionic conduc-
tors, and ion-exchange and sorption properties of soil minerals, molecular
sieves, and layered materials.
1. Introduction
Lithium intercalation or insertion materials have
been widely investigated in the search for new
electrode materials for use in high-voltage recharge-
able batteries.1-6 The first commercial Li-ion re-
chargeable battery contains the layered materials
LiCoO2 (Figure 1) and graphite as the cathode (or
positive electrode) and anode (or negative electrode),
respectively.7 Although this battery is the current
standard in many applications including cell phones
and laptops, its slow charge and discharge rates and
cost have prevented its use in applications that
require cheap high power and capacity, such as
hybrid electric vehicles and electric vehicles. The
toxicity of Co is also an issue. A wide variety of
materials have been studied,5,6 which include doped Nicolas Dupré was born in Chatillon-sous-Bagneux, France, in 1975. He
received his Ph.D. (2001) degree from Université Pierre et Marie Curie−
LiCoO2 phases, layered compounds based on the Paris VI working under the direction of Professor Michel Quarton. He
LiCoO2 structure (e.g., LiNiO2 8 and LiNi0.5Mn0.5O2 9,10), was appointed as a postdoctoral associate at SUNY Stony Brook in 2002,
where he works with Professor Clare P. Grey. His current research
* To whom correspondence should be addressed. E-mail: cgrey@ interests are focused on the study of the behavior of materials for lithium
notes.cc.sunysb.edu. batteries using solid-state NMR.
10.1021/cr020734p CCC: $48.50 © 2004 American Chemical Society

4494 Chemical Reviews, 2004, Vol. 104, No. 10 Grey and Dupré
2. NMR Background
2.1. Lithium NMR Spectra of Cathode Materials:
Introduction
The chemical shift range for lithium NMR spectra
is very small, and it is not always possible to resolve
resonances due to different local environments in the
NMR spectra of diamagnetic materials based solely
Figure 1. Structure of the cathode material LiCoO2, on the chemical shift interaction. Sometimes im-
showing the alternating layers formed by edge-sharing
CoO6 octahedra and Li+.
proved resolution can be obtained at higher field
strengths, allowing chemical information to be ex-
tracted from the spectra.18,19 Fortunately, the lithium
and a series of materials with two- and three-
NMR spectra of most battery samples are strongly
dimensional hosts for Li (e.g., Li1+xMn2-xO4 11 and
affected by a series of larger interactions which
LiFePO4 12). While long-range structural information
include quadrupole coupling (6Li, I ) 1; 7Li, I ) 3/2)
is typically available from diffraction methods, solid-
and interactions with unpaired electrons for para-
state NMR is an extremely useful tool for character-
magnetic samples (hyperfine interactions) and with
izing local structure in these materials, even in highly
the conduction electrons in metals (the Knight shift).
disordered systems. The lithium nuclei (7Li and 6Li) 7Li has the much higher natural abundance (93%)
are typically (but by no means exclusively) used as
and larger quadrupolar and gyromagnetic moments.
probes because it is the lithium ions that are directly
In contrast, 6Li is only 7% abundant, but its smaller
involved in the electrochemical processes. The NMR
quadrupole and gyromagnetic moments can result in
spectra are strongly influenced by the electronic
higher resolution spectra that are often easier to
structure of the materials, and it is often possible to
interpret. The quadrupolar interaction, which results
distinguish between insulators and conductors and
from the interaction of the quadrupolar nucleus with
between diamagnets and paramagnets. The method
the electric field gradient (EFG) at the nucleus, is
is quantitative and can be used to determine which
typically very small for 6Li but can result in charac-
species are removed on charging the battery and how teristic broadening in static and a series of spinning
the local structures change on extended cycling. NMR sidebands in magic angle spinning (MAS) 7Li NMR
is also sensitive to dynamics that occur on the NMR spectra due to the satellite (|+3/2〉 - |1/2〉 and |-1/2〉
time scale, and one-dimensional (1D) and two- - |-3/2〉) transitions. This (anisotropic) interaction
dimensional (2D) NMR have, for example, been used contains information concerning the local environ-
to investigate Li-ion motion in vanadates (1D)13 and ments at the lithium nucleus and can be used to
between two different nanosized domains in anatase, distinguish between ions in distorted and more
LiyTiO2.14 symmetric environments. Many battery materials are
This paper describes the approaches taken by us paramagnetic in the discharged or charged state. For
and other researchers over the past few years to example, the cathode material LiMn2O4 is a mixed-
interpret and extract chemical information from this valence compound containing Mn3+ (d4) and Mn4+ (d3)
class of materials. We focus on ex-situ analysis of ions. Although the Co3+ d electrons in LixCoO2 are
electrode materials (i.e., samples that have been paired in the fully discharged state, LixCoO2 contains
extracted from cycled batteries) since these ap- Co4+ d5 ions when charged. The NMR spectra of
proaches allow higher resolution spectra to be ob- paramagnetic materials are dominated by the inter-
tained. Use of a toroid detector (to obtain the NMR actions between the nuclear and electronic spins
signal) has allowed working batteries to be studied (Figures 2 and 3). These interactions may be much
in situ by spectroscopy and imaging methods,15 and larger than any of the other interactions and can
more recent studies have shown that NMR signals dominate the spectra of these materials, but they can
may be obtained from plastic bag batteries.16 In-situ also contain valuable information concerning both
methods have not, to date, been combined with the local crystallographic and electronic structure. Hence,
high-resolution method magic angle spinning (MAS) we will now consider these interactions in some
due to a number of experimental difficulties that still detail.
need to be surmounted and are, therefore, not
discussed here. The theory required to interpret the 2.2. NMR Spectra of Paramagnetic Materials
NMR spectra is first presented (section 2); the use of
this theory to extract chemical information is then Paramagnetic ions with electronic spin, S (e.g., S
described (section 3). This is followed by illustrative ) 3/2 for d3 ions Mn4+ and Cr3+), are associated with
recent examples from the field. In so far as this is magnetic moments, µe, that align in the presence of
possible, section 4 has been written so that it does a static magnetic field, B0 (Figure 2), typically defined
not require a detailed understanding of NMR to be the z-direction. Sz represents the component of
theory (beyond that discussed in section 2.1) and the spin along this direction. Electron spin resonance
should be accessible to the non-NMR audience. A (ESR) probes the transitions between the different
more comprehensive discussion of some of the techni- spin (or Zeeman) states |ms〉. Often the lifetime of an
cal aspects associated with obtaining NMR spectra ion in a particular electronic state (T1e) is very short
of cathode materials can be found in an earlier review on the relatively long time scale probed by NMR
article.17 (from many seconds to 10-8 s, depending on the size
Materials for Lithium-Ion Rechargeable Batteries Chemical Reviews, 2004, Vol. 104, No. 10 4495
Figure 2. (a, b) Effect of a static magnetic field on a

paramagnet with magnetic moment µe and electronic spin
S ) 1/2. Note that the lowest energy level for S ) 1/2 in the
magnetic field is the |-1/2〉 state due to negative charge of
the electron. (The magnetic moment of the electron µe
associated with this state is still aligned along the direction
of the static magnetic field; in many of the subsequent
drawings, we will use arrows to represent directions of the Figure 3. (a) 6Li MAS NMR spectrum of the spinel
net magnetic moments on the electrons/paramagnets.) (c) LiMn2O4 synthesized at 850 °C acquired with a spinning
ESR probes the transitions between the spins states |+1/ speed of 10 kHz at a field strength of 4.7 T. The spinning
2〉 T |-1/2〉. Spins in these states relax with a rate 1/T1e, sidebands and the isotropic shift are marked with asterisks
which is typically much faster than the size of many of the and its shift value (520 ppm), respectively. (b) Li local
interactions probed by NMR (i.e., T1e is shorter than the environment in LiMn2O4 showing a tetrahedrally coordi-
NMR time scale). Thus, NMR experiments are generally nated Li ion and the 12 nearest Mn ions (in the octahedral
only sensitive to the time-averaged value of the magnetic sites of the spinel structure). Schematics illustrating (c) the
moment, µ j e (-gµB〈Sz〉). transfer of unpaired electron spin density from the un-
paired electrons in the t2g d electrons (on one of the 12
nearby Mn ions) via the 2p orbitals on the intervening
of the electron-nuclear interaction). In this situation, oxygen atom to the 2s orbital on Li, which causes the large
the NMR spins cannot couple to Sz (i.e., the different shift seen for the LiMn2O4 isotropic resonance and (d) the
spin states |ms〉) and instead couple with the time dipolar coupling between a 6Li nuclear spin and net
average of the local field, 〈Sz〉. This is nonzero due to magnetic moment due to a nearby Mn ion. The spinning
the differences in populations of the different |ms〉 sidebands shown in (a) primarily result from this interac-
states in a magnetic field. The time average of Sz is tion, which is not completely removed by MAS.34
proportional to the net magnetic moment of an
ensemble, which is the quantity measured in a spin resonance (ESR) for the S ) 1/2 51V4+ vanadyl
magnetic susceptibility measurement ions in VO(H2O)52+,23 and the T1e’s for these ions are
10-8-10-9 s.22 1/T1e is similar in magnitude to A/h,
B0 and high-resolution NMR spectra are not observed.
〈Sz 〉 ) - χ (1) When the nuclei under observation are further from
µ0 gN0 µB M
the paramagnet, A is much smaller (the proton-
where µ0 is the permeability, g the electron g-factor, electron hyperfine coupling constants are only 2.1-
µB the Bohr magneton, N0 Avogadro’s number, and 0.01 MHz for water molecules in the same vanadyl
χM the magnetic molar susceptibility in m3 mol-1.20 complex24-26) and the condition in eq 2 is more readily
The size of the electron-nuclear interaction can be achieved. This is the case for lithium spins, which
quantified via a hyperfine coupling constant, A/h (in are generally separated by two or more bonds from
Hz) (see below), and NMR experiments are feasible the paramagnet. In addition, bonding involving Li+
for ions with short T1es, such that21 is largely ionic, again reducing the size of A. Shorter
T1es are generally observed for ions with S > 1/2 (S
|A/h| , 1/T1e (2) * 5/2), such as Mn3+, Ni2+, and Co2+, and higher
resolution spectra are more readily achieved.22 Broader
As the T1es lengthen and 1/T1e (s-1) approaches the 6Li/7Li resonances have been observed for local
size of the electron-nuclear interaction, considerable environments containing the paramagnetic S ) 1/2
NMR line broadening can occur, and it may not be ion Ni3+ (e.g., for LiNiO227 and LiCo1-xNixO228), but
possible to acquire high-resolution NMR spectra the resonances are not sufficiently broad to prevent
under these conditions. The effect of T1e on the detailed analysis of the local environments in these
nuclear relaxation times is discussed in more detail compounds.28,29
in ref 22. Large hyperfine constants are observed (of The nuclear spins can interact with the time-
many MHz) when the nuclear and electronic spins averaged magnetic moments via either through-space
are on the same atom. For example, a hyperfine (dipolar) or through-bond (Fermi-contact) interac-
constant A/h of -324 MHz was measured by electron tions. The 6Li MAS NMR spectrum of LiMn2O4 is
presented in Figure 3a to illustrate the effect of the depends, to a first approximation, on the contribution
different interactions on the NMR spectrum of this of the s orbitals of the NMR-active atom (lithium) to
paramagnetic material. the crystal orbitals with energies at or very close to
Ef (i.e., N(Ef)Li,2s the Li 2s partial density of states at
2.2.1. Fermi-Contact Interaction the Fermi level; this can be quite different than the
This is a measure of the unpaired electron spin value of N(Ef) that is important in controlling the
density that is transferred from the paramagnet to bulk conductivity of the sample). Often the d or p
the nucleus of the spin under observation. In the orbitals at the atom under investigation may have
regime defined by eq 2, the NMR shift (δ ) (∆ω/ω0)) larger contributions to N(Ef). This will be the case
induced by the Fermi-contact interaction is directly for some metals and the 51V Knight shifts of vana-
proportional to 〈Sz〉30,31 dates. Here, spin density can also be transferred to
the s orbitals via a polarization mechanism and both
∆ω Ac positive and negative Knight shifts can result, de-
)- 〈S 〉 (3) pending on the nature, populations, and locations of
ω0 ω0 h z the orbitals involved in the process. Polarization
mechanisms are discussed in more detail in section
The sign (and size) of the hyperfine constant Ac/h 3.1.
(Hz) determines the direction (and size) of the shift
2.2.2. Dipolar Coupling
Ac/h ) g µB γN F(r)0)µo /3S (4)
The dipolar interaction is the same interaction that
and depends on the electron spin (i.e., unpaired occurs between two or more nuclear moments and is
electron) density at the nucleus, F(r ) 0). γN is the caused by the local magnetic fields of the nuclear or
gyromagnetic ratio for the isotope studied by NMR, electronic (S) spins that are felt at the nearby nuclear
and ω0 is its Larmor frequency (in radians). The (I) spin (Figure 3d). When the dipolar interaction
6Li/7Li hyperfine shift can be very large for many
involves coupling to S ) 1/2 nuclear spins, coupling
transition-metal oxides and shift the Li resonances to both the |1/2〉 and |-1/2〉 eigenstates will occur,
well out of the typical range observed in the Li NMR resulting in the classic Pake doublet pattern for a
of diamagnetic materials (0 ( 5 ppm) (Figure 3a). powder. In contrast, when the second set of coupled
F(0) depends on the connectivity between the spins are electrons (or paramagnetic ions), the nuclear
orbitals on the paramagnet and the orbitals on the spins can only couple to the time-averaged magnetic
NMR-active atom (Figure 3c). Only the electrons in moment (again because the electronic relaxation time
s orbitals are associated with a finite probability of T1e is extremely fast on the NMR time scale). The
being found at the nucleus, and only the electron line shape in this case resembles that of the chemical
density transferred to the s orbitals needs to be shift anisotropy (CSA), and like the CSA, it scales
considered. Electron density may be transferred linearly with the field.33 The Hamiltonian for this
either directly from the paramagnet or indirectly via interaction, Hen, can be represented by33,34
a transferred hyperfine interaction to the s orbitals,
and thus, the interaction contains chemical informa- µ0
tion about the bonding involving the paramagnets, Hen ) j D̃ µ
µ (5)
4π e en N
Li, and the intervening oxygen atoms. For example,
in LiMn2O4, the transferred hyperfine interaction will where µj e is the time (or thermally) averaged magnetic
primarily involve the manganese t2g, oxygen 2p, and moment of the electrons (i.e., -gµB〈Sz〉) and D̃en is the
Li 2s orbitals (Figure 3c). The Fermi contact shift is dipolar coupling tensor between the unpaired elec-
generally considered to be additive, so that the shift tron and nucleus, which is defined by its matrix
due to many magnetic ions may be obtained from a elements as follows35,36
sum of the shifts induced by each magnetic ion,
although exceptions occur in systems with very 1
Dij ) (δij - 3eiej) (6)
delocalized spin states. One challenge in this field r3
has been to relate shift to local environment to
extract chemical information from these often- and depends on both the distance between the
complex systems. The correlation between shift and nuclear and electronic spins, r (e.g., the Li and Mn
local environment will be explored in some detail in atoms in Figure 3b), and the orientation of the
section 3. interatomic (Li-Mn) vector with respect to the static
The Knight shift, which dominates the spectra of magnetic field, B0. δij is the Kronecker delta (δij ) 1
metallic samples, is analogous to the Fermi-contact for i ) j and 0 for i * j), ei and ej are the x, y, z
shift, except that now the shift is a measure of the components of a unit vector pointing from the nuclear
density of states at the Fermi level, N(Ef). This spin to the electron spin in a chosen coordinate
relationship arises because the Pauli paramagnetism system. These interactions (and the CSA) are second-
χp of a metal is proportional to N(Ef),32 and thus, since rank tensors and therefore averaged by MAS. Since
the electron-nucleus interaction is dependent on 〈Sz〉 the size of the dipolar interaction is generally larger
and hence χ (eqs 1 and 3), the Knight shift is than the MAS frequency, a series of spinning side-
dependent on N(Ef). Again, since hyperfine coupling bands result (Figure 3a) that contains information
requires that unpaired electron density is present at concerning the size of the interaction and the relative
the nucleus under investigation, the Knight shift orientation of the spins.37
couplings are seen for nuclei with larger values of

γN. Thus, 7Li with its much larger gyromagnetic ratio
than 6Li (γ7Li/γ6Li ) 2.6) results in MAS spectra with
much larger spinning sideband manifolds. Further-
more, unlike the dipolar coupling between nuclear
spins, the size of the interaction depends on the
magnetic field strength (eq 8). Consequently, the
effect of this interaction can be minimized and
spectra simplified by using low fields, fast MAS, and
nuclei with low γNs. The effect of γN is illustrated in
Figure 4, which compares the 6Li and 7Li spectra of
the spinel LiMn2O4 prepared at 600 °C obtained at
an identical spinning speed and magnetic field
strength. The sidebands in this system result prima-
rily from the dipolar interactions. Clearly, it is
difficult to see all the resonances in the 7Li spectrum
at this spinning speed, and some of the weaker peaks
are obscured by the sidebands originating from the
more intense resonance at ca. 511 ppm.
Figure 4. Comparison of the spectra of LiMn2O4 synthe-
sized at 600 °C obtained with 6Li and 7Li MAS NMR at a 3. Extracting Chemical Information from the
MAS spinning speed of ∼10.5 kHz and a magnetic field Spectra of Paramagnetic Materials
strength of 4.7 T (200 MHz for 1H). (Reprinted with
permission from ref 17. Copyright 2003 Elsevier.) 3.1. Fermi-Contact Interaction
For isotropic magnetic moments, µ j e, the interaction To extract chemical information from the NMR
results in a line broadening of the resonance only but spectra of paramagnetic battery samples (e.g., Figure
no overall shift.38 This will not be the case for an 3a), we need to be able to assign a particular value
anisotropic magnetic moment. Here, both line broad- of the hyperfine shift to a specific local environment.
ening and a “pseudo-contact” shift will occur. This is To achieve this goal, we44-47 and other workers28,48,49
an important mechanism for “shift reagents” in have now studied a series of lithium compounds
solution NMR and has also been shown to be neces- containing transition metals with electronic configu-
sary to explain the 89Y shifts that arise due to rations from d1 to d.8 This has allowed us to rational-
lanthanide ions in the solid state.39 The pseudo- ize the differences in the shift mechanisms that result
contact shift arises from a similar mechanism as the from the presence of unpaired electrons in the t2g and
broadening and line splitting due to a coupling to a eg orbitals of the transition-metal ions. Since the
quadrupolar nucleus (the interaction involves a fourth- interaction is analogous to the J-coupling that occurs
rank tensor). Thus, the line broadening is no longer between nuclei and to magnetic interactions between
completely removed by MAS. In many of the batteries paramagnets, approaches similar to those used in
studied to date, the size of the shift and line broaden- these fields may be used to predict the sign and
ing are small and the interaction can often be ignored magnitude of the hyperfine interaction. The correla-
to a first approximation. More detailed studies are tions developed between the shift and local environ-
still, however, required to correlate this interaction ment in these lithium NMR studies will now be
with line broadening. explored.
The size of the dipolar interaction is often straight- To rationalize the large hyperfine shifts observed
forward to calculate as 〈Sz〉 can be readily estimated for lithium manganates such as the spinel LiMn2O4
from expressions for the susceptibility. For example, (Figure 3a), we synthesized a whole series of com-
assuming spin-only paramagnetism, which is a good pounds within the MnO2-LiMnO2-Li2MnO3 phase
approximation for d3 ions such as Mn4+ and Cr3+ 40,41 diagram. The NMR spectra of these compounds were
then acquired to obtain typical hyperfine shift ranges
χ ) (µ0/4π)N0 g2µB2S(S + 1)/3kT (7) for these materials (Figure 5). Even without a
complete understanding of the causes of these large
j e is given by42
and thus µ shifts, it then became possible to assign different shift
ranges to different lithium local environments. For
j e ) g2µB2S(S + 1)B0/3kT
µ (8) example, lithium spins in the octahedral site of a
Mn4+ spinel typically give rise to a resonance at
where k is the Boltzmann constant. In an extended 1800-2300 ppm, while lithium spins in the tetrahe-
solid, the nuclear spins are typically surrounded by dral site of the spinel resonate at ca. 800 ppm. This
more than one paramagnet and the total Hamilto- is illustrated in Figure 5 for the ordered spinel (Li0.5-
nian for the dipolar coupling, Hen, is obtained by Zn0.5)tet(Li0.5Mn1.5)O4, which contains Li ions on both
summing the individual coupling tensors for coupling the octahedral and tetrahedral sites of the spinel
to each individual paramagnetic ions in successive structure. Two resonances are clearly visible, which
coordination spheres.34,43 can be assigned using our hyperfine shift scale to the
Since µN, the magnetic moment ()γN(h/2π)Iz), two different sites in this structure.50 Similar ap-
depends on the gyromagnetic ratio, γN, larger dipolar proaches have been used by us and others to deter-
Figure 5. 6Li MAS NMR spectrum of the Mn(IV) spinel

(Li0.5Zn0.5)tet(Li0.5Mn1.5)OctO4 and the typical hyperfine shifts
observed for lithium in a series of local environments.
Hyperfine shifts are given next to the two isotropic
resonances in the 6Li spectrum; all other peaks are spin-
ning sidebands, which are predominantly caused by the
electron-nuclear dipolar coupling.
mine the locations of lithium ions in a series of

transition-metal-doped spinels, LiMn2-xMxO4 (M )
Cu, Co, Li, Ni, Cr).17,51
We examined the local environments around lithium
Figure 6. 6Li MAS NMR spectrum of the layered com-
in a series of Mn(IV) compounds to explain the pound Li2MnO3 acquired at a MAS frequency, νr, of 35 kHz.
different hyperfine shifts in these materials. By Spinning sidebands are marked with asterisks. The local
breaking up the local environments according to the environment in the Mn4+/Li+ layers that gives rise to the
numbers of manganese ions in the first cation coor- isotropic resonance at 1500 ppm is shown. Spin density
dination sphere around lithium and the Li-O-Mn may be transferred to the 2s orbital of Li via the interaction
bond angles, we determined the contribution of each with (b) a half-filled t2g orbital and (c) an empty dz2 Mn
Li-O-Mn interaction to the overall lithium shift.17,44 orbital to produce the hyperfine shifts seen in the spectrum
of Li2MnO3. The large arrows represent the magnetic
This analysis involves two assumptions: (1) that the moments of the electrons in the t2g and p orbitals, while
major shift contribution arises from the hyperfine the smaller arrows indicate the sign of the spin density
shift due to manganese ions in the first cation that is transferred to the Li 2s and transition-metal d
coordination sphere and (2) that there are no signifi- orbitals.
cant interactions between paramagnets at the tem-
Table 1. Comparison of the Hyperfine Shifts (in ppm)
peratures at which the NMR spectra were acquired Observed for Lithium Coordinated to a Single Mn4+
(close to room temperature). This latter assumption Ion via an Intervening Oxygen Ion for a Series of
allows the shifts between different compounds to be Local Environmentsa
compared. (The validity of this assumption can be Li-O-M bond angle (deg) Ooct b Otetc Olinear d
tested by studying the temperature dependence of the
90 122 163
shifts to determine whether there is any evidence for 122 76
large deviations from Curie-Weiss behavior.45,46) For 180 -60 -75 -125
example, one local environment for Li in Li2MnO3 a
Shifts are classified according to the coordination number
contains 12 sets of Li-O-Mn4+ bonds, each with a or environment of the intervening oxygen atom in each Li-
bond angle of close to 90° (Figure 6a). This environ- O-Mn4+ interaction. b Li2MnO3.47 c Spinel compounds such as
ment is associated with a hyperfine shift of 1500 ppm (Li0.5Zn0.5)(Li0.5Mn1.5)O4.50 d Perovskite and K2NiF4-related
or a shift of 125 ppm per Li-O-Mn4+ interaction. compounds such as La2(LiMn)0.5O4.53
More generally, Li+-O-Mn4+ bonds with bond angles
of 90° were found to result in large positive shifts of the filled oxygen p orbital can only occur via transfer
ca. 120-150 ppm, while linear Li-O-Mn4+ bonds of spin density of the opposite sign to that already
resulted in smaller negative shifts from -60 to -125 present in the half-filled t2g orbital since the t2g orbital
ppm, the exact value depending largely on the is half-filled (Figure 6b). This then results in transfer
coordination number of oxygen (Table 1).47 of positive spin density to the empty Li 2s orbital and
The size and direction of the shifts was then shifts of ca. 120-150 ppm per Li-O-Mn interaction.
rationalized by using the Goodenough-Kanamoori In the case of a 180° interaction, no direct overlap
rules, originally designed to predict the sign of the mechanism is possible involving the t2g orbitals. In
coupling between d electrons in transition-metal this case, the interaction occurs via the empty eg
oxides. For example, overlap between a half-filled t2g orbitals (Figure 6c). Here, the exchange coupling
orbital of a manganese ion and an empty 2s lithium between the eg and t2g orbitals on the same atom
orbital in a Li-O-Mn 90° arrangement may occur favors the transfer of spin density with the same sign
either directly or via an intervening p orbital on an of spin polarization as that in the t2g orbitals to the
oxygen atom. In the case of the transferred hyperfine eg orbitals. This results in a net transfer of negative
interaction, an interaction with the t2g electron and spin polarization from manganese to the empty 2s
electron orbital occupied by a “spin-down” electron.

This results in the so-called “fully occupied” crystal
orbital that contains two paired electrons. In the
presence of a field, the energies of the two orbitals
are no longer necessarily identical, the one-electron
orbital with a magnetic moment aligned with the
field being lower in energy. In the case of a “half-
filled” orbital, this is the source of the hyperfine shift,
as only one out of the two one-electron orbitals is
occupied. However, when both of these one-electron
orbitals are occupied this does not directly result in
a shift. Instead, the shift arises from the effect of the
field on the contributions of the different orbitals to
the crystal orbitals. In the absence of a magnetic
field, the individual one-electron orbitals are identical
(i.e., the contributions of the different atomic orbital
to the crystal orbital are the same), but this need not
be the case when a field is applied. Exchange
Figure 7. (a) (Partial) band diagram for a transition-metal coupling between electrons in different orbitals low-
oxide, with transition-metal ions located in sites with ers the energy of electronic configurations containing
octahedral symmetry. (The t2g and t2g* crystal orbitals are parallel electrons. Thus, the presence of “spin-up”
formed from an overlap of the t2g d orbitals and oxygen 2p electrons in the t2g* orbitals will stabilize the “spin-
atomic orbitals, while the eg and eg* crystal orbitals are up” one-electron orbital formed from the eg orbitals.
formed from the eg d orbitals and oxygen 2p and 2s atomic This results in a polarization of the eg crystal orbitals,
orbitals. The t2g* bands are partially occupied for Mn4+ and
Cr3+ d3 ions.) (b) Delocalization mechanism involving the so that the contribution of eg atomic orbital to the
half-filled t2g* crystal orbital, and (c) polarization mecha- crystal orbital will be larger for the “spin-up” one-
nism involving the two one-electron orbitals that make up electron orbital and smaller for the “spin-down”
the “filled eg orbital”. The sizes of orbitals in c represent orbital. This is illustrated schematically in Figure 7c.
the relative contributions of these orbitals to the crystal Since the “spin-down” orbital is still a one-electron
(or molecular) orbital; the sign of the spin-density trans- orbital, the contribution to this orbital from the
ferred to the Li 2s orbital is indicated with a “+” or “-”
sign.
oxygen 2p and Li 2s atomic orbitals must be larger,
and thus the concentration of “spin-down” electrons
on the Li 2s orbitals is higher than the concentration
Li orbital and, consequently, a negative shift. The
of “spin-up” electrons, resulting in a net transfer of
lithium (tetrahedral) environment in the spinel struc-
negative spin density. This picture is entirely equiva-
ture contains 12 Li-O-Mn bond angles of ca. 122°
lent to one where the presence of the magnetic field
(Figure 3b), and a shift intermediate between that
observed for linear and 90° interactions is observed is considered to mix the eg and eg* orbitalssthe
of ca. 76 ppm, as expected, based on orbital overlap extent of mixing is different for the “spin-up” and
arguments. The tetrahedral site in the ordered spinel “spin-down” orbitals, leading to differences in the
(Li0.5Zn0.5)(Li0.5Mn1.5)O4, whose spectrum was shown contributions of the metal d, 2p O, and 2s Li atomic
in Figure 5, contains nine sets of Li-O-Mn bonds orbitals to the crystal orbitals. A similar phenomenon
with 122° bond angles (the remaining three are Li- can occur, for example, in a ferromagnetic solid, the
O-Li bonds), consistent with the shift for this presence of ordered electronic spins acting as the
environment of 688 ppm. source of the local magnetic field.
The different shift mechanisms may be understood The relative sizes of the polarization and delocal-
in more detail by considering the effect of the ization mechanisms will depend on the orbitals
magnetic field on the populations and energies of the involved in the overlap and the occupancies of the
different crystal orbitals (Figure 7a). Transfer of metal orbitals. For example, the polarization mech-
electron density via the 90° interaction arises due to anism appears to be relatively weak for Ni3+, in
a direct delocalization of spin density due to overlap contrast to Mn4+ and Cr3+, even when both mecha-
between the half-filled t2g*, oxygen π*, and empty Li nisms involve eg orbitals.52 This is most likely due to
2s atomic orbitals (the delocalization mechanism, the smaller exchange interaction expected for the
Figure 7b).52 This overlap is responsible for the Ni3+ system, which contains fewer unpaired elec-
formation of the t2g* (antibonding) molecular orbital trons. Some earlier studies neglected the polarization
in a molecule or the t2g* crystal orbital (or band) in a mechanism, assuming that it was much weaker than
solid. No shift occurs for the 180° interaction from the other possible shift mechanisms, which lead to
this mechanism as the eg* orbitals are empty. incorrect assignments.28
A second mechanism (the polarization mechanism) Recent density functional theory (DFT) calculations
arises due to the polarization of the fully occupied have been used to calculate the transfer of unpaired
(bonding) crystal orbitals formed by the eg, oxygen electron spin densities to the nearby atoms and
2p, and Li 2s atomic orbitals in the presence of a analyze the orbitals involved in these mechanisms.52
magnetic field. A fully occupied crystal (or molecular) For example, Figure 8 shows the spin densities
orbital in reality comprises one one-electron orbital obtained for Cr3+-doped LiCoO2. The spin densities
occupied by a “spin-up” electron and a second one- can be calculated by subtracting the “spin-up” elec-
will be considered initially since this contains isolated

manganese ions within the Li/Mn/O planes that are
only weakly coupled.53 La2(LiMn)0.5O4 is isostructural
with K2NiF4, and the lithium is coordinated to four
manganese ions via linear Li-O-Mn bonds. A shift
of -480 ppm was observed for this compound. This
is consistent with a Jahn-Teller distortion of the d4
Mn ions and the population of the eg* (dz2) orbital
perpendicular to the Li-O-Mn axes (i.e., perpen-
dicular to the Li/Mn/O planes), so that the Li shift is
dominated by the four interactions involving empty
dx2-y2 orbitals. The shift mechanism is similar to that
observed for interactions involving empty dz2 orbitals
(Figures 6c and 7c).53
The three Mn3+ rock-salt related compounds with
stoichiometry LiMnO2 show much smaller shifts of
between 40 and 143 ppm in comparison to the shifts
observed for La2(LiMn)0.5O4 and the Mn(IV) com-
pounds.45 All the lithium atoms in these compounds
have similar numbers of manganese ions in their
local coordination sphere, resulting in 12 Li-O-Mn
90° interactions, but differ in the number of types of
Li-O-Mn 180° interactions. The slightly larger
shifts seen for the tetragonal-Li2Mn2O4 and mono-
clinic-LiMnO2 compounds (99-143 ppm) in compari-
son to the shift seen for orthorhombic-LiMnO2 (40
ppm) may be rationalized by the presence of ad-
Figure 8. Calculated spin polarization density map in a
Co/Cr-O-Li plane in LiCr1/8Co7/8O2 from DFT calcula-
ditional interactions involving half-filled eg orbitals,
tions.52 The positions of the Cr, Co, Li, and O ions are which, for the same reason discussed for the 90°
indicated. Li(1) has a Cr3+ ion as its second cation coordi- interactions involving t2g orbitals, should result in a
nation shell, while Li(2) has a Cr3+ ion in its first cation positive shift. However, by using arguments outlined
coordination shell. The 6Li spectra of LiCrxCo1-xO2 (x ) 0.05 above for the Mn(IV) compounds, we would predict
and 0.1)47 are shown along with the assignments of a very large shift for these Mn(III) compounds of
resonances corresponding to Li(1) and Li(2) local environ- greater than 1400 ppm. One cause of the large
ments.
differences is the strong antiferromagnetic interac-
tions between Mn3+ ions, which are responsible for
tron density from the “spin-down” electron density non-Curie-Weiss behavior and in the case of o-
at a particular location in space. The unpaired LiMnO2 for a Néel transition temperature of -11
electron density in the Cr3+ (d3) t2g atomic orbital is °C.54 This suggestion was tested by following the 6Li
clearly visible. Direct overlap of this orbital with a shifts of these compounds as a function of tempera-
nearby 2p O orbital results in positive spin density ture.45 These results demonstrate significant devia-
in this orbital and positive spin density at the site of tion from Curie-Weiss temperature dependence, the
the Li nucleus, connected via the 90° interaction. This shifts being essentially independent of temperature
is consistent with the resonance at 40 ppm seen in between 0 and 250 °C. Recent density functional
the 6Li NMR spectrum of LiCo1-xCrxO2 (x ) 0.1 and theory (DFT) calculations for m-LiMnO2 suggest that
0.05), which was also assigned to the presence of one changes in bond lengths caused by the Jahn-Teller
Cr3+ in the local coordination sphere of a lithium ion. distortion also strongly affect the relative magnitudes
Negative spin density is present in the 2p O orbital of the different shift mechanisms, resulting in smaller
that points directly to the Cr3+ ion and the Li ion than predicted shifts for these compounds.52
connected via this 180° interaction, consistent with
the observation of a resonance at -70 ppm in the These findings are not unique to the manganese
NMR spectrum. systems, and other materials showing strong anti-
ferromagnetic interactions between paramagnets
Analysis of the spectra of Mn(III) compounds is show smaller than predicted shifts. Examples include
more complicated as (i) the eg orbitals are partially the NMR spectra of R and β LiFeO2 where shifts of
filled, (ii) Jahn-Teller distortions typically occur ca. 500 ppm were observed,27 even though the d5 Fe3+
resulting in one half-filled and one empty eg* orbital, ions in these compounds contain both partially filled
(iii) many of these compounds contain sizable anti- t2g and eg atomic orbitals. Again, strong deviations
ferromagnetic interactions between manganese ions, from Curie-Weiss behavior are seen for these com-
which result in strong deviations from Curie-Weiss pounds.
behavior, even at around room temperature. We have
analyzed the Mn(III) compounds by examining the
Mn-O bond lengths, the longer Mn-O bond lengths 3.2. Dipolar Interaction
(of greater than 2 Å) in the MnO6 octahedra being
indicative of electron occupancy of the (antibonding) Although the sidebands that result from the dipolar
eg* orbitals. The Mn(III) compound, La2(LiMn)0.5O4 interaction can often complicate analysis of the
Mn distances are larger and also because the ions

are arranged more symmetrically around the tetra-
hedral Li ions so that individual dipolar couplings
partially cancel, resulting in a smaller overall dipolar
coupling. Again, this smaller dipolar coupling can be
used as a signature for the tetrahedral site in these
compounds.
4. Applications of NMR Spectroscopy to the

Study of Cathode Materials
MAS NMR has now been used to study LiCoO2-
derived layered materials28,29,55 as well as a wide
range of alternative cathode materials including
manganates,44,48,49,56-58 vanadates,13,59-61 and iron and
vanadium phosphates.62-64 We will now discuss the
Figure 9. 6Li MAS NMR spectra of the ordered spinel application of NMR to some of these materials to
[Li0.5Zn0.5]tet[Li0.5Mn1.5]OctO4 and Li2MnO3.34 The asterisks illustrate the type of information that has been (and
indicate the spinning sidebands of the resonances from Li can be) obtained by using this method.
in the octahedral site in the spinel and the 2b site of Li2-
MnO3 (i.e., the site due to Li in the Mn layers). The local 4.1. Spinels
environments for Li in these sites and the Li in the Li
layers of Li2MnO3 are shown. These spectra were acquired The spinel materials, LiMn2-xM′xO4, have been the
at a moderately slow spinning speed of 8.7-8.8 kHz (cf. most extensively investigated of all the lithium
Figure 6) to increase the intensity seen in the spinning
sidebands (relative to the isotropic resonance), improving manganates.65,66 These materials can be cycled at ca.
the accuracy of the analysis of the sideband manifolds. 4 V (vs Li) from LiMn2O4 to MnO2, the manganese
ions retaining the spinel host framework through-
spectrum (Figure 3), this interaction does contain out.1,67 Capacities of 148 mAg-1 are, in theory,
extremely useful structural information that can be possible for the stoichiometric spinel, but in practice
relatively straightforward to extract. The effect of no more than 100-120 mAg-1 has been achieved over
different arrangements of paramagnetic ions on the multiple charge-discharge cycles.68 The spinels rep-
NMR spectra has been explored, and different ar- resent cheap alternative cathode materials which can
rangements can result in characteristic line shapes. be charged and discharged sufficiently rapidly to
This is illustrated in Figure 9 for the two Mn4+ meet the power requirements for use in electric
compounds Li2MnO3 and the spinel (Li0.5Zn0.5)tet(Mn1.5- vehicles. Unfortunately, many of the spinels suffer
Li0.5)octO4. The Li ions in the transition-metal layers from rapid capacity fade following extended storage
of Li2MnO3 and in the octahedral site of the spinel or on cycling, particularly at high temperatures.5 This
are coordinated to the same number of Mn ions (six), has prompted much research on these materials to
but the arrangements of these ions are very different. identify the source and then prevent this fade.69-71
For Li2MnO3, these ions are located in the same plane Surface treatment methods have also been developed
as the central Li ion (an oblate arrangement), while to help prevent manganese dissolution,69 and Li-
in the spinel, they lie above and below an imaginary excess spinels Li1+xMn2-xO4, x ) 0.05-0.1, are now
plane running through Li (a prolate arrangement). being produced as cheap cathode materials particu-
As discussed above, the total dipolar coupling inter- larly for use in high-power applications. Figure 10
action is a sum of all the individual Li-Mn4+ dipolar shows a typical potential profile (obtained in a
interactions. The total interaction will be very sensi- galvanostatic mode, i.e., by using constant current)
tive to the relative orientations of the Li-Mn inter- for the spinel Li1.05Mn1.95O4.56 Initial capacities of 125/
atomic vectors. For the sites in the two compounds 115 (charging/discharging) mAh/g were obtained
discussed here, although the magnitude of the inter- (theoretical capacity ) 126 mAh/g). Good capacity
action is similar, the sign of the interaction is retention was observed for this material, with 96%
opposite in sign. Differences in the signs of the of the initial discharge capacity being retained after
interaction result in characteristic changes in the line 30 cycles.
shapes of the two compounds. The shapes of the NMR spectroscopy has been performed on these
sideband manifolds for the two groups of resonances systems to investigate structure before and after
in Li2MnO3 () Li[Li1/3Mn2/3]O2 in the notation used electrochemical cycling. A large hyperfine shift of
to describe layered compounds) are also very differ- more than 500 ppm from the chemical shift range of
ent. Li located in the Li layers is nearby manganese typical diamagnetic compounds containing lithium is
ions arranged above and below the plane of Li ions, typically observed in the 6Li or 7Li MAS NMR
and thus the shape of the sideband manifolds of the spectrum of the stoichiometric spinel LiMn2O4 (Fig-
resonance due to this site is similar to that seen for ure 2) synthesized in air at a temperature of between
the octahedral sites in the spinel. The shape is 700 and 850 °C.44,48,58,72 LiMn2O4 is a mixed-valent
characteristic for Li ions between layers of paramag- compound containing both Mn3+ and Mn4+ ions; it is
netic ions and may be used as a signature for this also a hopping semiconductor, hopping occurring
environment. Much smaller sideband manifolds are between the eg orbitals of the manganese ions. Since
seen for the tetrahedral sites, in part because the Li- this hopping time scale is fast in comparison to the
Figure 11. 6Li MAS NMR spectra of the spinel LiMn2O4

(structure shown on the left) as a function of synthesis
Figure 10. Potential profile of Li1.05Mn1.95O4 during the temperature (adapted from ref 44). The cluster of Mn4+ ions
first charge and discharge in the voltage range of 3.3-4.5 formed as a result of Li substitution on the octahedral site
V. The cell (comprising Li1.05Mn1.95O4 as the positive is shown, and the resonances due to Li ions near this
electrode (cathode) and Li as the negative electrode (anode)) cluster are assigned.
was charged/discharged with a constant current corre-
sponding to a charge rate of C/10 (∼10 mA per gram of
active material; C/10 indicates that the cell was fully The spinning sideband pattern seen for this reso-
charged in 10 h; a charge rate of 1C corresponds to 1 h). nance is similar to that found for the octahedral site
[Similar cells and cycling conditions were used to prepare in the spinel Li0.5Zn0.5[Mn0.5Li0.5]O4, confirming that
the samples for the NMR experiments described in section this resonance is due to Li substitution (for Mn) on
4.1.1 and in ref 56. Here, cycling was arrested at different the octahedral (16d) site of the spinel structure. Thus,
potentials, the cells disassembled in a glovebox, and the
cathode mixtures packed in NMR rotors.] the spectra provided clear evidence that the samples
prepared at lower temperature contain excess Li,
NMR time scale (ca. 10-5 s), the lithium spins “see” despite their stoichiometry. The residual Mn is
an average manganese oxidation state of 3.5 (i.e., presumably present as Mn2O3, as shown by Paulsen
“Mn3.5+” ions), and thus there is only one magneti- et al.78 The additional resonances in the spectra of
cally inequivalent lithium site (the 8a site). The NMR the nominally stoichiometric material LiMn2O4 can
spectra are clearly sensitive to this hopping mecha- be assigned by using our hyperfine shift scale (Figure
nism, and NMR spectroscopy may be used to follow 5) to tetrahedral lithium environments nearby Mn4+
the partial charge-ordering process that occurs just ions. These Mn4+ ions arise from the substitution of
below room temperature in the stoichiometric mate- Li+ in the octahedral (16d) site: For every Li+
rial.44 substituted on this site, five Mn3.5+ ions must be
In our initial study of these systems44 we investi- oxidized to Mn4+. Additional resonances at lower
gated the effect of synthesis temperature on the local frequency than the main spinel (520 ppm) resonance
structure in the nominally stoichiometric spinel were observed for spinels prepared above 850 °C or
LiMn2O4, since its stoichiometry and manganese in inert atmospheres.56 These peaks were ascribed
oxidation state depend on the preparation condi- to the presence of Mn3+ ions in the Li local coordina-
tions: Li2Mn4O9 (Mn4+) may be synthesized at 400 tion sphere, caused by the presence of Mn2+ ions on
°C,73 while an increase of temperature results in the the tetrahedral (8a) site of the spinel structure. Now,
formation of defect spinels with a progressive reduc- the substitution of a Mn2+ for Li+ requires that two
tion of the manganese oxidation state. Eventually an Mn3.5+ ions are reduced to Mn3+ for charge balance.
average oxidation state for manganese of 3.5 is
4.1.1. Cation-Doped Spinels
achieved in the stoichiometric material LiMn2O4
synthesized at ∼850 °C.74 Figure 11 shows that the The Jahn-Teller distortion that occurs on inter-
6Li MAS NMR spectra of this compound vary dra- calation of LiMn2O4 to form Li2Mn2O4 is associated
matically with synthesis temperature. Despite the with a large change in cell volume.66 One possible
fact that the structure of the stoichiometric spinel source of capacity fade in the spinel material has
contains only one crystallographic site for lithium been ascribed to this Jahn-Teller distortion, which
(the 8a site; see Figure 3b), a large number of can occur (particularly at the surface of the particles)
different local environments are observed for the when the oxidation state of the manganese drops
compounds synthesized at lower temperatures. The below 3.5 during discharge.66,79 Doping has been
spectrum of the material synthesized at a tempera- investigated extensively in the spinel system because
ture of 600 °C resembles the spectrum of samples it represents a method for controlling the oxidation
with excess lithium (e.g., Li1.05Mn1.95O4),56 which typi- state of the manganese at the end of discharge (of
cally show improved cycling behavior.75-77 Further- the ca. 4 V plateau).69 On the basis of charge balance
more, a weak resonance at 2300 ppm can be detected arguments, substitution of a manganese ion for a
in the 7Li MAS NMR spectra of the materials dopant cation with charge of n+ will result in the
synthesized at both 550 and 600 °C, collected using oxidation of (3.5 - n)2 manganese ions with average
an extremely fast MAS spinning speed of 40 kHz.17 oxidation states of 3.5 to Mn4+.
The spectra of the doped materials (Cr3+, Ni2+, spinel structure was occurring on a millisecond time
Zn2+, Li+, Co3+, Al3+) are similar to those seen for the scale. This was consistent with the complete collapse
nominally stoichiometric materials, and sets of reso- of the fine structure at higher temperatures in our
nances between 500 and 700 ppm are seen on cation earlier work, where it was ascribed to either mobility
doping17,51 in addition to that of the normal spinel of Li+ or mobility of the holes on the manganese ions
environment (at ca. 500 ppm). Again, these reso- near the defects in these materials (i.e., the octahe-
nances are assigned to lithium ions near manganese- dral 16d Li ions17).44 Later variable-temperature
(IV) cations. The lower intensity of the additional NMR results for LiNi0.1Mn1.9O4, where similar phe-
resonances seen on Cr3+ substitution, in comparison nomena were observed, suggested that Li+ motion
to Zn2+ or Ni2+ substitution, is consistent with the must play a role in the collapse of the fine structure.46
oxidation of fewer manganese ions near the dopart By monitoring the change in cross-peak intensity as
ions. For the Li- and Zn-doped spinels, resonances a function of temperature, Verhoeven et al. extracted
at ca. 2300 ppm were also observed, which are an activation energy for the jump process of 0.5 (
assigned to lithium ions in the octahedral sites of the 0.1 eV and also an estimate for the difference in
spinel structure. In the case of Zn doping, it is clear energy between Li in the normal tetrahedral site and
that the preference of Zn2+ for the tetrahedral site Li in the site that gives rise to the resonance at ca.
of the spinel structure forces the lithium onto the 590 ppm of ca. 700 K (or 0.05 eV). However, their
octahedral site. assignments of the spectra are very different from
The NMR results indicate that cation doping does ours.56 They assigned the two most intense extra
not increase the manganese oxidation states of all resonances at ca. 550 and 590 ppm in Figure 11 (and
cations in the whole phase evenly. Instead, the at slightly higher frequencies in their study) to Li in
manganese cations near the dopant cations are the 16d (Mn) and 16c sites, respectively. These
oxidized, leaving the bulk relatively unperturbed. assignments were made based on the earlier neutron
This is borne out in the mechanism of deintercalation diffraction experiments82 but, if correct, suggest that
of these systems, which differs noticeably from the the 16c site is occupied in significant concentrations,
deintercalation mechanism of stoichiometric LiMn2O4. even at room temperature. The 16c site is an octa-
Two-phase behavior was observed during charging hedral site which is empty in LiMn2O4 but is occupied
of stoichiometric LiMn2O4 samples synthesized at 650 on insertion of Li to form Li2Mn2O4.83 Given our direct
°C and above.80 In contrast, the results for Li1.05- observation of the 16d Li site (at 2300 ppm17) and
Mn1.95O4 (whose potential profile was shown in that a much larger shift than 590 ppm is predicted
Figure 10) indicate that the lithium cation is deinter- for the 16c site, based on our hyperfine shift scale,44
calated from the different local sites sequentially.56 we suggest that the motion observed here is due to
Li is deintercalated first from the site containing all 8a-8a hops, the measured difference in energy
Mn3.5+ ions in its local coordination sphere (ca. 500 between the two sites (0.05 eV) arising from the effect
ppm); this is followed by Li deintercalation from sites of the Li incorporation into the octahedral (16d) site
nearby progressively more Mn4+. The resonance at and formation of the Mn4+ ions nearby the “590 ppm”
2100-2300 ppm can be seen throughout the whole site. The authors suggest that their activation energy
charging range studied, indicating the lithium in the corresponds to that of an 8a-16c jump. Most likely,
octahedral sites is not deintercalated. Li is pinned the 8a-8a jump process does occur via the 16c site,
in the tetrahedral sites near these octahedral Li ions the 16c representing a metastable intermediate state.
and Mn4+ ions. We proposed that the presence of In this case, the measured activation energy still
excess Li+, which serves to pin the tetrahedral Li ions corresponds to that of a 8a-16c hop; however, it
in the lattice, helps to prevent the formation of a seems unlikely that the 16c site is populated for a
series of cation-ordered phases during charging.56 significant length of time, as implied by this study.
Instead Li1+xMn2-xO4 forms a continuous solid- NMR can be used to follow the changes in the
solution on charging and discharging,81 improving the electrode materials following multiple charging cycles
stability of this material during extended cycling. to determine why some materials fail and others
On deintercalation of ca. 25% of the Li ions from maintain high capacities for extended cycling times.
the tetrahedral site of Li1.05Mn1.95O4, the additional For example, Tucker et al. performed a series of
resonances due to Li in tetrahedral sites near Mn4+ detailed lithium NMR studies to explore the effect
ions disappear.56 The fine structure returns on cool- of cation doping on the long-term stability of the
ing the sample to -60 °C. This indicates that the spinel materials.49,51 They systematically explored a
creation of Li-ion vacancies on charging results in a series of proposed failure mechanisms by preparing
noticeable increase in Li mobility, involving jumps model electrode materials where structural changes
between the different tetrahedral sites.56 Verhoeven associated with the proposed failure mechanisms
et al. used 2D 7Li NMR experiments to study the were intentionally introduced. The lithium spectra
dynamics of Li in the fully lithiated material Li1.04- of the model compounds were then compared with
Mn1.96O4.57 Their 7Li NMR spectra are very similar spectra of electrodes extracted from cells that had
to our 6Li spectra of a similar material and also the been charged/discharged for varying numbers of
“stoichiometric” material prepared at 600 °C (Figure cycles. For example, samples were soaked in moisture-
11). At temperatures above room temperature, they contaminated electrolyte, and the spectra of these
detected “cross-peaks” in the 2D exchange spectra were compared to the spectra of the cycled materials.
between the 7Li resonances at 500-650 ppm, indicat- Soaking in wet electrolyte resulted in a shift of the
ing that motion between different Li sites in the spinel peak to 633 ppm and the appearance of a
resonance at 0 ppm due to the SEI (surface electrode 4.2. Cr3+- and Ni2+-Substituted Layered Lithium
interphase). This shift was not observed in the Manganates
spectra of the cycled materials. This was in contrast
to cells cycled at Stony Brook, where a shift to higher Lithium NMR spectroscopy has similarly been
frequency was observed on extended cycling,80 pre- applied to investigate the local structure and cycling
sumably a consequence of moisture introduced during properties of a number of alternative layered man-
cell assembly. Both studies ascribed the shift to Mn2+ ganese cathode materials. Layered, monoclinic Li-
dissolution and the consequent increase in oxidation MnO2 is isostructural with LiCoO2 and has been
state of the spinel material which can occur via a widely studied as a possible cathode material.95,96
reaction such as84 Unfortunately, this material transforms on cycling
at 3-4 V (vs Li) to the more thermodynamically
stable spinel phase.97,98 Cation doping has been used
2LiMn2O4 + 4H+ f
to stabilize the layered phase with some success.99,100
Mn2+(aq) + 3MnO2 + 2H2O + 2Li+(aq) (9) One series of materials that has generated consider-
able recent interest includes several cathodes, in
Li+ loss, but without an increase in manganese particular, containing Ni2+ or Cr3+ (Li[Li0.2Mn0.4Cr0.4]-
oxidation state, can also occur due to ion exchange O2 and Li[Ni0.5Mn0.5]O2) where the charging and
of Li+ for any H+ ions in the electrolyte.85 Sato et al. discharging cycles involve multiple electron transfers
showed that the ion-exchange process occurred pre- per transition-metal ion.6,9,10,101 The manganese re-
dominantly on the surface for crystalline spinel mains in the +4 valence state at all times, preventing
samples, the oxidation state of the manganese ions the instabilities associated with Mn3+ (i.e., the layered-
at surface remaining close to 3.5, the average Mn to-spinel conversion). High capacities of up to 190-
oxidation state of the bulk increasing to 3.9 on 220 mAhg-1 have been seen for these materials.
treatment with acid.86 Pickup et al. used Li NMR to Chromium was shown to cycle reversibly between
examine the effect of chemical delithiation in sulfuric Cr3+ and Cr6+ in the mixed Cr-Mn oxide Li[Li0.2-
acid solutions on the spinel material.87 A resonance Mn0.4Cr0.4]O2.6,102 Lithium NMR spectroscopy showed
was observed at ca. 950 ppm for delithiated spinel that the material contains Li2MnO3-like and Mn4+-
samples LixMn2O4, x < 0.3, which was ascribed to Li doped LiCrO2 local environments or domains.47 Only
in the tetrahedral site in Mn4+-rich domains in the the Li ions in the Mn4+-doped LiCrO2 regions of this
center of the particles. The authors also proposed that “nanocomposite” are removed on cycling and largely
these domains resulted from H+/Li+ ion-exchange on responsible for the capacity, while the Li2MnO3 parts
the surfaces of the particles; the presence of a surface were inactive. The toxicity of Cr6+ restricts the use
layer of HMn2O4 was supported by parallel FTIR of this cathode material, and subsequent NMR work
studies. The 950 ppm resonance was seen in earlier has focused on solid solutions of Li[NiMn]0.5O2 and
studies at the top of charge for stoichiometric spinels Li2MnO3.
and following multiple charge-discharge cycles.80 Although the widely studied layered material
An alternative approach for investigating the hy- LiNiO2 functions via the reversible cycling of nickel
drogen content of cathode materials involves the use between oxidation states of 3 and 4, nickel may also
of 2H or 1H NMR. For example, we showed that 2H be cycled between Ni2+ and Ni4+ at higher voltages
NMR could be used to detect deuterons in samples in both layered materials and spinels. For example,
following chemical or electrochemical reduction.88 the spinel LiNi0.5Mn1.5O4 cycles at ca. 4.8 V, the
This approach was used to follow the structural manganese, present as Mn4+ in this material, acting
changes that occur on acid leaching of Li2MnO3.89 2H as in inert spectator ion.103 The presence of Ni2+ (and
NMR signals are much easier to detect than 1H not Ni3+) in the spinels LiNixMn2-xO4 was clearly
signals because of the much smaller electron-nuclear demonstrated by NMR.46 A similar Ni2+/Ni4+ process
dipolar couplings involving 2H due to its smaller γN. occurs in layered LiNi0.5Mn0.5O2 and its Li-excess
A 2H NMR signal from Li1.05Mn1.95O4 soaked in D2- versions at a much lower voltage (between ca. 3.7 and
SO4 due to ion-exchanged deuterons has been ob- 4.6 V).9,10,104,105
6Li NMR has been used to study the local environ-
tained, but no studies following extended cycling have
been performed to date.90 ments in the Li2MnO3-Li(NiMn)0.5O2 solid solutions
Finally, NMR has also be used to study other (Li[NixMn(2-x)/3Li(1-2x)/3]O2). Li resonances due to Li
spinels materials that do not contain manganese. For ions in the predominantly Li layers (500-800 ppm)
example, the intercalation/deintercalation of lithium and in the transition-metal layers (1360 and 1560
titanate spinels such as Li4/3+xTi5/3O491 and Li1.1- ppm) can be clearly resolved at frequencies close to
Ti1.9O4+δ92 have been investigated. These materials those seen for similar environments in Li2MnO3
may be used as anode materials in combination with (Figure 12).106,107 Surprisingly, even for the x ) 0.5
cathodes operating at 4 V (vs Li) to produce cells with end member, Li ions are still observed in the transi-
potentials of ca. 2.5 V.93 These materials are either tion-metal layers (circled in Figure 12), indicating
diamagnetic or metallic,94 and unlike the mangan- that Ni2+-Li+ exchange between the layers has
ates, only very small differences in shifts are seen occurred. The presence of nickel in the Li layers has
for Li in the different sites of the spinel structure. been confirmed by diffraction experiments.104,105 The
Nonetheless, these shift differences are enough to intensities of the resonances due to Li in the transi-
allow the concentrations of the different sites to be tion-metal layers were used to develop a model for
quantified and monitored following insertion of Li92 cation ordering and demonstrate that the ordering
or as a function sample preparation method.94 in these layers was far from random. The Li ions in
of the Co3+ and doped ions, but also changes in the

electronic conductivities of these materials. In par-
ticular, the NMR spectra are extremely sensitive to
semiconductor-to-metal phase transitions.109 Al-
though low-spin d6 Co3+ ions are nominally diamag-
netic, the layered (O3) LiCoO2 (Figure 1) and the
layered O2 polytype show a very weak, temperature-
independent (Van-Vleck) paramagnetism,110 which is
presumably due to mixing in of higher (paramag-
netic) states. The labels “O2” and “O3” for the two
structures indicate that the Li environment is octa-
hedral in both but that the stacking sequences of the
oxygen layers are ABCB and ABCABC, leading to
two and three sets of Co and Li layers in the
hexagonal unit cell, respectively.
The 6Li and 7Li spectra of the O3 and O2 LiCoO2
phases both show a resonance at -0.2 ppm consistent
with lithium in an essentially diamagnetic local
environment and the presence of low-spin (t2g6) Co3+
ions.28,110,111 Shift anisotropies, obtained by analysis
of the 7Li and 6Li NMR line shapes, were extremely
close to those calculated by using the experimentally
Figure 12. 6Li MAS NMR spectra of Li[Li(1-2x)/3Mn(2-x)/3- determined values for the susceptibility χm for both
Nix]O2, with x ) 1/10, 1/3, and 1/2. The resonances phases. 59Co (I ) 7/2) NMR has also been used to
corresponding to local environments Li(OMn)6 and Li- probe the structure of these materials and extract
(OMn)4(OLi2) found in Li2MnO3 are marked. The Li(OMn)6 detailed information concerning the coordination
local environment is shown on the right. The frequencies
of the major resonances are indicated; asterisks indicate environments in these phases.110,112 O2 and O3 show
spinning sidebands. quite different 59Co MAS NMR spectra, and the 59Co
(I ) 7/2) quadrupolar coupling constant (QCC) for the
octahedral site in O2 is much larger than that found
for the Co site in the O3 material, consistent with
the differences in the two structures: the CoO6
octahedra in the O2 structure share a face and three
edges with the LiO6 octahedra, while they only share
(6) edges in the O3 structure.110
NMR spectroscopy has been used to follow the
Figure 13. Schematic of the ordering scheme found in Li2- cycling of LixCoO2.109 The results indicate that Co4+
MnO3 showing the honeycomb structure formed by Li (red ions are formed during the initial stages of deinter-
circles) and Mn (left-hand side) and the structure derived calation.109 A broad resonance appears at x ) 0.94,
by replacing two Li and one Mn by three Ni (green circles). the shift being ascribed to a Knight shift, due to
interaction between the Li nuclei and the conduction
the Ni/Mn layers show a strong preference for electrons in the t2g* band formed from the overlap-
substitution in environments near six Mn4+ ions, as ping t2g cobalt orbitals (Figure 14). The NMR results
found in the end member compound Li2MnO3. A suggest that the biphasic region that exists in Lix-
model was proposed that was based on the “honey- CoO2 from 0.75 e x e 0.94 is due to this phase
comb” ordering of the Mn4+ and Li+ ions in the Mn/ transition. Additional 7Li resonances have been seen
Li layer of Li2MnO3, where the Ni2+ ions substitute for Li-excess nonstoichiometric Li1+xCoO2, which have
into ordered layers so as to minimize the number of been ascribed to the presence of paramagnetic Co2+
Li+-Ni2+ contacts (Figure 13).107 ions that are formed in conjunction with oxygen and
Studies of these materials following charging and cobalt vacancies.55 Materials synthesized at higher
discharging showed that both the Li in the Li layers temperatures have also been shown to contain oxygen
and in the transition-metal layers participate in the vacancies. These oxygen vacancies result in the
electrochemistry.106 Even the Li in the Ni/Mn layers presence of five-coordinate Co3+ ions, the NMR
can be cycled reversibly at least in the first few results suggesting that these Co3+ ions are not
cycles.108 Preliminary results suggest that the Li diamagnetic but rather present as so-called “inter-
vacancies created in this process at the top of charge mediate-spin” Co3+ defects.113
may represent a possible mechanism for creating
Lithium NMR has also been used to investigate the
increased disorder in these materials following ex-
Ni3+ doping in LiCoO2.28 A series of 7Li NMR reso-
tended cycling.
nances were observed, which correspond to local
environments formed by Ni3+ substitution in the first
4.3. LiCoO2 and Related Materials and second cation coordination spheres surrounding
LiCoO2-related cathode materials have been inves- the lithium ions. The resonances can be assigned by
tigated extensively by NMR. The NMR of these using similar arguments as those used to assign the
systems may be used to follow not only the oxidation spectra of the manganates. Recent DFT calculations
to the presence of Ni4+ ions, which result in local

distortions and prevent full electronic delocalization
between Co3+ and Co4+ ions.29
NMR studies have shown that structural defects
such as oxygen vacancies and, to a lesser extent,
intermediate-spin Co3+ are created upon Mg substi-
tution in LiCoO2.114 7Li resonances with noticeable
Knight shifts were seen by NMR, even before Li+-
ion deintercalation, but no evidence for metallic
conductivity was seen from the electric properties.
This was ascribed to the presence of metallic behavior
at the local level in small domains surrounding the
doped Mg2+. The domains are presumably too well
separated and too small to allow percolation (of the
conduction electrons) throughout the solid.114 Doping
with other transition metals such as Fe has been also
probed by NMR,115 and the electrochemical properties
of the “O2 phase” of LiCoO2 have been investigated
in some detail.116
Finally, 27Al (I ) 5/2) and 59Co NMR spectroscopy
have been used to probe Al3+ in Al-doped lithium
Figure 14. 7Li MAS NMR spectra for LixCoO2 (asterisks cobalt oxides and lithium nickel oxides.117,118 A 27Al
indicate spinning sidebands). The spikes at 0 ppm seen in
the spectra of x ) 0.50 and 0.6 are artifacts. The LiCoO2
chemical shift of 62.5 ppm was observed for the
local environment and resonance due to the metallic phase environment Al(OCo)6 for an Al3+ ion in the transi-
are marked. (Reprinted with permission from ref 109. tion-metal layers, surrounded by six Co3+ ions.
Copyright 1999 Royal Society of Chemistry.) Somewhat surprisingly, this is in the typical chemical
shift range expected for tetrahedral environments
have shown (in contrast to earlier studies28) that the (ca. 60-80 ppm), but no evidence for occupancy of
negative shifts seen in these systems are due to Ni3+ the tetrahedral site was obtained from X-ray diffrac-
ions in the Li first coordination shell (90° interaction and IR studies on the same materials.117 Sub-
tions), while the positive shifts are due to 180° stitution of the Co3+ by Al3+ in the first cation
interactions with half-filled eg orbitals.52 The large coordination shell leads to an additive chemical shift
positive shifts are caused by the delocalization mech- decrease of ca. 7 ppm, and the shift of the environ-
anism involving the partially filled eg* orbitals, while ment Al(OAl)6 (20 ppm) seen in spectra of materials
the negative shifts are due to the polarization mech- with higher Al content is closer to that expected for
anism now involving the filled t2g orbitals (formed octahedral Al. The spectra are consistent with a
from the metal dxy, dxz, and dyz and the oxygen 2p continuous solid solution involving octahedral sites
orbitals) (see section 3.1). These new assignments do randomly occupied by Al and Co.117 It is possible that
not significantly alter the conclusions reached in the unusual 27Al shifts seen for this compound are
earlier papers. The distributions of the Ni3+ ions and related to the Van-Vleck susceptibility of this com-
the extent of Ni3+ substitution were quantified by pound.
extracting the intensities of the different resonances,
the results demonstrating that the LiNi1-yCoyO2 4.4. Lithium Phosphates
phases have a tendency to form cobalt-rich clus-
ters.28,29 7Li NMR spectra of the partially deinterca- Recently, LiFePO4, which adopts the olivine struc-
lated material LixNi0.3Co0.7O2, (1 < x e 0.85) indicate ture, has been widely studied as a possible cheap
that the Ni3+ ions are oxidized at a lower potential cathode material.12 Although the specific capacities
than the Co3+ ions, the results suggesting that Ni3+/ of this material and the related NASICON structures
Ni4+ electronic hopping occurs in the early stages of are relatively low (typically 100 mAhg-1), these
charge. However, at Li0.7Ni0.3Co0.7O2 where in prin- compounds are cheap and open up new avenues for
ciple all the Ni3+ should be oxidized to diamagnetic research. 7Li MAS NMR studies of a series of olivines
Ni4+, an NMR signal at ca. 20 ppm is seen. This is LiMPO4 with M ) Mn, Fe, Ni, and Co were carried
not consistent with the presence of only Ni4+ and Co3+ out by Tucker et al.64,119 A single resonance with large
ions (which are both diamagnetic) and indicates that spinning sidebands envelope was observed for all the
both Ni3+ and Co3+ are oxidized in this composition compounds, indicating the presence of one local
range. Furthermore, the variable-temperature NMR environment for the lithium cations. In contrast to
results for this sample suggest that both hopping the lithium manganese oxide spinels, much smaller
between both Ni and Co ions occurs (Ni3+/Ni4+ and shifts from -90 to 70 ppm were seen. The difference
Ni4+ + Co3+ f Ni3+ + Co4+). Thus, the question as to between the compounds was rationalized by consid-
which ion is being oxidized was suggested to be ering the electronic configuration of the transition-
“irrelevant” in this range. At x ) 0.40, metallic metal cation site in Cs symmetry. Interestingly, the
behavior was seen by NMR (and thermoelectric smaller shifts observed in these compounds and in
power measurements) but electronic conductivity the related vanadium-containing NASICONs62 is
measurements showed that the conductivity is still most likely related to the same phenomena that
an activated process. The discrepancy was ascribed results in the high voltages of these systems (the
Fe2+/Fe3+ couple is at 2.8-3.5 V in these struc- 27Al and 29Si NMR,134,135 the isotropic chemical shifts
tures120), namely, the so-called inductive effect of the 51
for V nuclei in vanadates in different coordination
phosphate groups. The covalent P-O bonds of the environments overlap. The anisotropy of the chemical
phosphate groups will presumably increase the ionic shift is, however, sensitive to the arrangement of
character of the Li-O and M-O bonds in these oxygen atoms in the first coordination sphere: Va-
compounds,120 reducing the extent of overlap and nadium atoms in distorted octahedral sites are typi-
thus hyperfine interactions in these materials. cally associated with CSAs that are nearly axially
symmetric and with spans (∆σ ) σ11 - σ33) of ca.
4.5. Vanadates 600-1400 ppm. For vanadium atoms in distorted
tetrahedral sites, a chemical shift tensor with a large
Highly oxidized oxides of vanadium, chromium, asymmetry parameter η (which indicates the devia-
niobium, and molybdenum are well known for their tion from axial symmetry η ) (σ11 - σ22)/(σiso - σ33),
ability to intercalate large concentrations of lithium 0 e η e 1; η ) 0 for an axially symmetric tensor) is
ions on electrochemical reduction.121-127 Although obtained along with a span of 400-600 ppm. The
many vanadium compounds have limited cycle lives span drops to 100-300 ppm for a slightly distorted
in lithium-ion cells, limiting their commercial use, tetrahedral site. The isotropic chemical shift is sensi-
the vanadium oxides V2O5, V6O13, and LiV3O8, in tive to the type of atoms in the second coordination
particular, are all potential rechargeable 3 V elec- sphere,136-138 providing additional structural infor-
trode materials and suitable for use in conjunction mation.
with polymer electrolytes. These systems have re-
ceived considerable attention by NMR spectrosco- As discussed in section 2.2, 51V NMR spectra of VIV
pists. In addition to lithium NMR, 51V MAS NMR compounds with localized d1 V4+ ions are typically
spectroscopy represents a powerful tool for studying “invisible” by NMR when standard MAS NMR meth-
the local environments in these materials, particu- ods are used. In contrast, 51V spectra may be detected
larly for compounds containing the diamagnetic, d0 for metallic V4+ samples,139 and large shifts are
ions V5+. We briefly discuss some of the information typically observed as a result of the Knight shift. For
that can be obtained from 51V NMR before describing instance, a large negative Knight shift was observed
the applications of NMR to these materials. at -4788 ppm for R-VO2.140 51V resonances for the
V5+ ions in mixed V5+/V4+ compounds may, however,
4.5.1. 51V NMR be “visible”. A good example is provided by the work
of Delmaire et al.,141 who assigned the resonance at
The 51V nucleus is quadrupolar (spin 7/2, natural -1447 ppm observed in reduced forms of Bi4V2O11-δ
abundance 99.76%), and thus, the spectra can be to V5+ ions near V4+ ions, the large shifts resulting
affected by both the first- and second-order quadru- from V5+-O-V4+ hyperfine interactions. V2O5 does
pole interaction, though the second-order broadening not show metallic behavior on partial or complete
is generally not the largest source of line broadening reduction of V5+ to V4+ or V3+. Evidence for partial
in these materials.128 In general, three major aniso- delocalization of spin density has been obtained for
tropic interactions influence the line shapes seen in some reduced vanadates by using ESR and NMR
the 51V NMR spectra of solid samples: (i) the qua- spectroscopy, but in these systems delocalization
drupolar interaction, (ii) the chemical shift anisotropy results from a hopping mechanism rather than from
(CSA), which can be significant for 51V, and (iii) the metallic behavior.62,142,143 The main problem associ-
dipolar interaction between the 51V nucleus under ated with the use of 51V NMR to characterize cathode
observation and nearby nuclei or electrons. Analysis materials is the broadening and/or loss of the signal
of the line shapes of these compounds is typically on reduction, so that it becomes difficult, or impos-
nontrivial, particularly at moderate or high field sible, to follow change in the vanadium local environ-
strengths, and careful simulation of the spectra is ment during lithium intercalation/deintercalation.
required to extract the various parameters describing However, NMR of other nuclei such as 6,7Li or 31P
the CSA (σ11, σ22, σ33) and quadrupolar tensors. The can sometimes be performed in parallel to probe
line shapes of the 51V NMR spectra are strongly these systems.144,145
influenced by the CSA in high magnetic fields and
by the quadrupolar interaction in low magnetic fields, 4.5.2. NMR Studies of Vanadium Oxides
and thus, spectra should be acquired at multiple field
strengths to ensure accurate measurements of the The vanadium oxide, V2O5, shows particular prom-
various interactions. ise as a positive electrode material in polymer bat-
Although much of the 51V NMR has been performed teries. Because of the existence of the short vanadyl
on model systems or catalytic materials containing bond,146 it can either be considered as a two-
vanadium,129,130 compounds such as V2O5 or VOPO4 dimensional layered structure comprising VO5 square
are used in both the catalysis and lithium battery pyramids or a 3D structure formed from VO6 octa-
fields, and many of the results can be used to help hedra. The strong anisotropy of the structure ac-
elucidate the structures of vanadium-containing cath- counts for its ability to accommodate lithium. Inter-
ode materials. 51V NMR spectra are sensitive to calation to form LixV2O5 at ca. 3.3 V vs Li0/Li+ results
changes in the vanadium coordination number and in two phases, R (x < 0.1) and (0.35 < x < 0.7),147
distortions of the vanadium local environments from which differ in their interlayer spacing. δ-LiV2O5 is
regular tetrahedra or octahedra.131-133 51V isotropic formed after the shearing of the layers.148 For 1 < x
chemical shifts of between -400 and -800 ppm are < 2, the δ phase undergoes an irreversible transfor-
seen for vanadium oxides, and unfortunately, unlike mation involving considerable reconstruction to form
a γ phase that can cycle between 0 < x < 2.149 The

intercalation of a third lithium atom into the struc-
ture yields the ω phase, which is able to cycle for
more than 100 cycles between 3.4 and 1.9 V, yielding
a specific energy of 450 Wh/kg.150
Numerous studies have been performed using
6,7Li and/or 51V NMR of chemically and electrochemi-
cally intercalated V2O5 bronzes,60,143,151 different

LixV2O5 phases (0 < x <2),152,153 and LixV2O5 aerogels
with compositions of 1.00 < x < 5.84.62,142 Intercala-
tion of lithium causes distortions resulting in a
decrease in the a lattice parameter, an increase in
the c parameter (puckering of the V2O5 layers148), and
a change in the coordination of the lithium. Different
7Li MAS NMR resonances are seen for the various
lithium sites encountered during electrochemical

lithium intercalation: R-LixV2O5 (-16 ppm; trigonal),
Ι-LixV2O5/ΙΙ-LixV2O5 (-25/-22ppm;trigonal),δ-LixV2O5
(-12 ppm; tetrahedra), γ-LixV2O5 phase (5 and 16
ppm; distorted octahedra and tetrahedra respec-
tively), and ζ-LixV2O5 (10 ppm).154 The lithium sites
in the γ-phase are associated with smaller QCCs,
smaller dipolar couplings, and smaller shifts. Sites
with low occupancies were more readily seen with
7Li NMR due to its higher sensitivity, while higher
resolution spectra were obtained with 6Li NMR.

Hirschinger et al.143 used NMR to show that the
biphasic domain seen on chemical lithiation of V2O5
at high temperatures to form lithiated β (7Li NMR
resonance at -27 ppm due to monocapped trigonal
prisms) and β′ (-31 ppm; tetrahedral site) phases
should be extended from x ) 0.33 to 0.48 (at 600
°C).155,156 The unpaired electrons become increasingly
delocalized as the lithium content decreases in the
β/β′ phases, in contrast to the γ phase, where the
electrons were localized on specific vanadium sites. Figure 15. 51V MAS NMR spectra of Li1+xV3O8 obtained
Stallworth et al. showed that the Li sites in V2O5 at different stages of the intercalation process (νr ) 13
aerogels occupy sites with axial symmetry, consistent kHz), plotted in a quantitative mode. The isotropic reso-
with proposed structural models where the interca- nances are marked with circles.
lated lithium ions are situated at the apical oxygen 4). The compound forms a continuous solid solution
position between adjacent V2O5 layers.62,142 New sites between 4.2 and 3.0 V and shows two reversible
are occupied on incorporation of larger amounts of phase transitions at lower potential during cycling.
lithium, which are associated with nonbridging oxy- The V3O8 layers of LiV3O8 are comprised of vanadium
gens in the equatorial plane of the VO5 unit. 51V NMR in square pyramids (V1), distorted octahedra (V2),
resonances were seen for all the reduced samples, and more regular octahedral (V3).160 Experimental
allowing the V4+/V5+ ratio to be evaluated. and theoretical studies performed by Benedek et
In addition to structural information, 7Li MAS al.,161 Picciotto et al.,159 and Jouanneau162 suggested
NMR T2 relaxation measurements and analysis of a number of different possible (octahedral and tet-
7Li line shapes have been used to probe the dynamics
rahedral) sites for Li.
of the lithium ions.13,153 Holland et al.13 identified two Both the 51V and 7Li NMR spectra show multiple
different species with different mobilities (interfacial vanadium and lithium local environments for the as-
Li (longer T2, rapid dynamics) and intercalated synthesized material (x ) 0.15), and the spectra
lithium (shorter T2, slower dynamics)) in the elec- cannot be explained by using a simple model145 based
trochemically lithiated V2O5 xerogel matrix. Li hop- on the number of crystallographically distinct vana-
ping frequencies were extracted from an analysis of dium sites.159,161 On Li-ion intercalation, the 51V
the 7Li line widths and the appearance of a quadru- resonances sharpen and shift to higher frequencies
polar splitting as the temperature decreased in a (Figure 15); three sharp resonances along with two
related system.153 broader resonances are clearly resolved for the
Li1+xV3O8 has received considerable attention as a samples prepared at potentials of 3.4 and 3.0 V (x )
cathode material for secondary lithium batteries over 0.3 and 0.5, respectively). This behavior is consistent
the past few years.157-162 Li1+xV3O8 has a layered with solid-solution behavior in this potential range
framework suitable for reversible lithium intercala- and is ascribed to the presence of localized V4+ defects
tion processes that can allow up to four Li+ ions per at x close to 0 and electron delocalization for 1 > x >
formula unit to be inserted (i.e., x can vary from 0 to 0.05. Three lithium sites were observed in the 7Li
NMR spectra for 0 < x < 1, which are assigned to

the octahedral site and two tetrahedral sites between
the lithium layers. A phase transition has been seen
in the electrochemical data at ca. 2.8 V to form
Li2V3O8, but no evidence is seen for this transition
with structural probes (such as diffraction and trans-
mission electron microscopy). The phase transition
results in a dramatic change in the NMR spectra. A
noticeable decrease in intensity and resolution of the
51
V resonances is seen, which has been ascribed to a
first-order electronic transition with charge ordering
and the formation of localized V4+ and V5+ ions. The
lithium NMR spectra are similarly consistent with
localized paramagnets. Small negative shifts are
seen, consistent with the presence of d electrons in
the dxy orbitals of the vanadium ions.
Lithium intercalation in V6O13 has been studied by
Stallworth et al.60 Variable-temperature 7Li NMR
indicated considerable mobility for Li+ in the inter-
calated materials. The 7Li NMR data were compared
with ESR spectra and near-edge X-ray absorption
fine structure (NEXAFS) data on the same materials,
and a correlation between vanadium oxidation state
(from NEXAFS data) and NMR shift was observed.
The authors explained the shifts in terms of different
coupling mechanisms between the V4+ and V3+ shifts.
The shifts were, however, extracted from static NMR
experiments, and it is possible that some of the
different local environments, typically revealed in a
MAS spectrum, were not seen in this study.
4.5.3. Vanadium Phosphates

More recently, lithium vanadium phosphates (Li3V2-
Figure 16. Quantitative 6Li MAS NMR spectra of RII-Lix-
(PO4)3 and Li3FeV(PO4)3,163,62 with open NASICON VOPO4 acquired during the first and second electrochemi-
framework structures, have also been studied. Re- cal cycles at B0 ) 7.1 T (νr ) 8.0 kHz). The shifts of the
versible electrochemical lithium deintercalation/re- isotropic resonances are marked next to the peaks. The
intercalation at a higher potential (in comparison to circles and asterisks indicate extra impurity peaks and
the V5+/V4+ couples seen for the oxides) of between 3 sidebands, respectively.
and 3.5 V (vs Li0/Li+) is seen along with high lithium involving the RII phase, at 3.5 (26%), 3.7 (16%), and
diffusion rates. VOXO4 materials (X ) S, P, As)164-167 3.9 V (58%), while the second lithium intercalation
contain two- or three-dimensional frameworks168-170 takes place in two steps at 3.9 (75%) and 3.7 V (25%
that can accommodate lithium. They have high of the total capacity). These changes reflect an
theoretical specific capacities (from 135 to 165 mAhg-1, electrochemically driven structural rearrangement
depending on X), reversible Li intercalation occurring that occurs upon cycling of RII-VOPO4, which leads
at a potential of ca. 4 V for β-VOPO4165 and at 3.5- to a change in the lithium intercalation/deintercala-
3.7 V for RII-VOPO4.166,167 tion potential. This transformation was proposed to
51V, 6,7Li, and 31P MAS NMR spectroscopy have
involve local rearrangements of ions as well as the
been performed to study the two first electrochemical nucleation and growth of well-defined phases. A 6Li
cycles of RII-VOPO4 and β-VOPO4.144 Lapina et al.171 signal corresponding to the β phase, present as an
and Siegel et al.172 determined both the magnitude impurity, is seen even at high potentials (3.75 V) at
and relative orientation of the quadrupole and chemi- 0 ppm.
cal shift tensors for the RI-, RII- (δiso) -755 ppm), and Only the nonlithiated diamagnetic phase may be
β-VOPO4 (δiso) -735 ppm) phases from 51V NMR seen by 51V NMR, lithium intercalation resulting in
data with slightly different results being obtained a dramatic loss of signal, presumably due to the large
between the two groups for the RII phase. 6,7Li, 31P, 51V-electron hyperfine interaction (as discussed in
and 51V magic angle spinning (MAS) NMR were used section 2.2), the 6Li NMR resonance at 120 ppm
to follow the evolution of RII-VOPO4 during the two confirming the presence of V4+ ions. Thus, the method
first electrochemical cycles. Lithium intercalation in cannot be used to follow the lithium intercalation into
RII-VOPO4, to form RII-LiVOPO4, occurs in two volt- the different sites in the host structure. The percent-
ages on the first reduction, at 3.7 (16% of the total age of β phase (12%) present as an impurity could
capacity) and 3.5 V (84% of the total capacity), be extracted from the 51V MAS NMR spectrum. The
resulting in a 6Li resonance at 120 ppm (Figure 16) signal corresponding to the β phase disappears at the
characteristic of RII-LiVOPO4. The subsequent lithium beginning of the discharge but reappears at the end
deintercalation occurs in three steps, all of them of the cycle, at high potential (4.0 V), showing that
the RII and β phases cycle independently. The inten- port comes from the Assistant Secretary for Energy
sity of RII phase decreased as a function of cycle Efficiency and Renewable Energy, Office of Freedom-
number, suggesting that other paramagnetic phases CAR, and Vehicle Technologies of the U.S. Depart-
are formed; this is confirmed by observation of ment of Energy via subcontract No. 6517749 with the
additional resonances in the 6Li and 31P MAS NMR Lawrence Berkeley National Laboratory. C.P.G.
spectra of the same materials. thanks the current and former members of her group
No 51V NMR studies have been performed on the (Y. J. Lee, C. Pan, Y. Paik, W. S. Yoon, J. Breger,
rhombohedral form of Li3-xV2(PO4)3 due to the local- and M. Jiang) and co-workers B. Meyer and N. Leifer
ization of electrons on the V4+ and V3+ ions. 7Li NMR who have contributed to the work described herein.
experiments have however been performed on this Insightful discussions with D. Carlier, M. Menetrier,
system, which suggest that ions on the M(3) sites of and G. Ceder are gratefully acknowledged.
the NASICON structure in Li3V2(PO4)3 move onto the
M(1) sites on extraction of Li to form LiV2(PO4)3.62 A 7. References
shift of the isotropic resonance from 85 to 62 ppm on
(1) Bruce, P. G. Chem. Commun. 1997, 1817.
deintercalation of two Li+ ions is attributed to the (2) Guyomard, D.; Tarascon, J. M. Solid State Ionics 1994, 69, 222.
change in site geometry and the change in vanadium (3) Koksbang, R.; Barker, J.; Shi, H.; Saidi, M. Y. Solid St. Ionics
oxidation state from V3+ to V4+. Finally, 7Li NMR 1996, 84, 1.
(4) Winter, M.; Besenhard, J. O.; Spahr, M. E.; Novak, P. Adv.
studies performed on the monoclinic form of LixV2- Mater. 1998, 10, 725.
(PO4)3 provide strong evidence for Li and vanadium (5) Tarascon, J.-M.; Armand, M. Nature 2001, 414, 359.
charge ordering on insertion of lithium, the ordering (6) Ammundsen, B.; Paulsen, J. Adv. Mater. 2001, 13, 943.
(7) Nagaura, T.; Tozawa, K. Prog. Batteries Solar Cells 1990, 9, 209.
driving the series of phase transitions seen for this (8) Dahn, J. R.; von Sacken, U.; Michal, C. A. Solid State Ionics
material.173 1990, 44, 87.
(9) Ohzuku, T.; Makimura, Y. Chem. Lett. 2001, 744.
(10) Lu, Z.; MacNeil, D. D.; Dahn, J. R. Electrochem. Solid State Lett.
5. Conclusions 2001, 4, A191.
(11) Thackeray, M. M.; David, W. I. F.; Bruce, P. G.; Goodenough, J.
NMR represents a powerful method for studying B. Mater. Res. Bull. 1983, 18, 461.
the local structures and electronic properties of (12) Padhi, A. K.; Nanjundaswamy, K. S.; Goodenough, J. B. J.
cathode materials as a function of state of charge. (13) Holland, G. P.; Buttry, D. A.; Yarger, J. L. Chem. Mater. 2001,
The research reviewed in this article clearly demon- 14, 3875.
strates that it is now possible to obtain a wealth of (14) Wagemaker, M. M.; Kentgens, A. P. M.; Mulder, F. M. Nature
2002, 418, 397.
information from the NMR spectra of cathode mate- (15) Gerald, R. E.; Sanchez, J.; Johnson, C. S.; Klinger, R. J.; Rathke,
rials, even in the paramagnetic or metallic state, both J. W. J. Phys. Condens. Mater. 2001, 13, 8269.
(16) Chevallier, F.; Letellier, M.; Morcrette, M.; Tarascon, J.-M.;
prior to and following electrochemical cycling of a Frackowiak, E.; Rouzaud, J.-N.; Béguin, F. Electrochem. Solid-
battery. The paramagnetic properties of many of State Lett. 2003, 6, A225.
these systems, rather than being a complication, can (17) Grey, C. P.; Lee, Y. J. Solid State Sci. 2003, 5, 883.
(18) Park, S.-H.; Kleinsorge, M.; Grey, C. P.; Parise, J. B. J. Solid
often provide increased resolution of resonances from State Chem. 2002, 167, 310.
different local environments. The lithium hyperfine (19) Meyer, B.; Leifer, N.; Greenbaum, S.; Grey, C. P. Manuscript in
shifts of many systems can be rationalized in terms preparation.
(20) Kittel, C. Introduction to Solid State Physics, 6th ed.; John Wiley
of the coordination environment for lithium and the & Sons: New York, 1986.
numbers and types of unpaired electrons on the (21) Carrington, A.; McLachlan, A. D. Introduction to Magnetic
Resonance; Harper and Row: New York, 1967.
nearby paramagnets. In the case of systems whose (22) Kohler, F. H. In Magnetism: Molecules to Materials; Miller, J.
hyperfine shifts obey a Curie-Weiss Law, it is often S., Drillon, M., Eds.; Wiley-VCH: New York, 2001.
possible to predict the magnitude and direction of the (23) Albanese, N. F.; Chasteen, N. D. J. Phys. Chem. 1978, 8, 910.
(24) van Willigen, H. J. Magn. Reson. 1980, 39, 37.
shifts by analyzing the connectivity between lithium (25) Atherton, N. M.; Shackleton, J. F. Mol. Phys. 1980, 39, 1471.
and nearby paramagnetic ions. (26) Bertini, I.; Briganti, F.; Luchinat, C.; Xia, Z. J. Magn. Reson.
NMR may be used to probe the effect of doping on 1992, 99, 235.
(27) Pan, C. M. A. Thesis, State University of New York at Stony
local structure. For example, NMR studies of the Brook, Stony Brook, NY, 2002.
spinel phases show that cation doping results in an (28) Marichal, C.; Hirschinger, J.; Granger, P.; Menetrier, M.;
increase in the oxidation state of the manganese ions Rougier, A.; Delmas, C. Inorg. Chem. 1995, 34, 1773.
(29) Carlier, D.; Menetrier, M.; Delmas, C. J. Mater. Chem. 2001,
near the cation dopant, while cation doping in LiCoO2 11, 594.
has a dramatic effect on the electronic structure of (30) McConnell, H. M. J. Chem. Phys. 1957, 27.
(31) McConnell, H. M.; Robertson, R. E. J. Chem. Phys. 1958, 29,
the material. NMR may be used to determine the 1361.
cation ordering and investigate the effect of local (32) Cox, P. A. Transition Metal Oxides; Clarendon Press: Oxford,
structure on the electrochemical properties of the 1995.
(33) Nayeem, A.; Yesinowski, J. P. J. Chem. Phys. 1988, 89, 4600.
cathode materials. In phosphates and vanadates, (34) Lee, Y. J.; Grey, C. P. J. Phys. Chem. B 2002, 106, 3576.
additional probes such as 51V and 31P NMR provide (35) Mehring, M. Principles of High-Resolution NMR in Solids;
additional information concerning local structure and Springer-Verlag: New York, 1983.
(36) Haeberlen, U. High-Resolution NMR in Solids; Academic: New
electronic properties. In conclusion, NMR has now York, 1976.
been applied to a wide range of cathode materials for (37) Brough, A. R.; Grey, C. P.; Dobson, C. M. J. Chem. Soc., Chem.
Commun. 1992, 742.
lithium-ion batteries and is rapidly becoming an (38) Bertini, I.; Luchinat, C.; Parigi, G. Solution NMR of Paramag-
essential characterization tool in this research area. netic Molecules; Elsevier: Amsterdam, 2001.
(39) Grey, C. P.; Smith, M. E.; Cheetham, A. K.; Dobson, C. M.;
Dupree, R. J. Am. Chem. Soc. 1990, 112, 4670.
6. Acknowledgments (40) Goodenough, J. B. Magnetism and the Chemical Bond; Inter-
science: New York, 1963.
Support from the NSF via DMR 9901308 and (41) Hatfield, W. E. Solid State Chemistry Techniques; Oxford
0211353 is gratefully acknowledged. Additional sup- University Press: Oxford, 1987.
(42) Bloembergen, N. Physica 1950, 16, 95. (89) Paik, Y.; Grey, C. P.; Johnson, C. S.; Kim, J.-S.; Thackeray, M.
(43) Brough, A. R.; Grey, C. P.; Dobson, C. M. J. Am. Chem. Soc. M. Chem. Mater. 2002, 14, 5109.
1993, 115, 7318. (90) Paik, Y.; Naqvi, N.; Grey, C. P. Unpublished data.
(44) Lee, Y. J.; Wang, F.; Grey, C. P. J. Am. Chem. Soc. 1998, 120, (91) Ronci, L. F.; Stallworth, P. E.; Alamgir, F.; Schiros, T.; van
12601. Sluytman, J.; Guo, X.; Reale, P.; Greenbaum, S.; den Boer, M.;
(45) Lee, Y. J.; Grey, C. P. Chem. Mater. 2000, 12, 3871. Scrosati, B. J. Power Sources 2003, 119, 631.
(46) Lee, Y. J.; Eng, C.; Grey, C. P. J. Electrochem. Soc. 2001, 148, (92) Krtil, P.; Dedecek, J.; Kostlanova, T.; Brus, J. Electrochem. Solid
A249. State Lett. 2004, 7, A163.
(47) Pan, C.; Lee, Y. J.; Ammundsen, B.; Grey, C. P. Chem. Mater. (93) Ohzuku, T.; Ueda, A.; Yamamoto, N. J. Electrochem. Soc. 1995,
2002, 14, 2289. 142, 1431.
(48) Gee, B.; Horne, C. R.; Cairns, E. J.; Reimer, J. A. J. Phys. Chem. (94) Kartha, J. P.; Tunstall, D. P.; Irvine, J. T. S. J. Solid State Chem.
B 1998, 102, 10142. 2000, 152, 397.
(49) Tucker, M. C.; Reimer, J. A.; Cairns, E. J. Electrochem. Solid- (95) Armstrong, A. R.; Bruce, P. G. Nature 1996, 381, 499.
State Lett. 2000, 3, 463. (96) Capitaine, F.; Gravereau, P.; Delmas, C. Solid State Ionics 1996,
(50) Lee, Y. J.; Park, S.-H.; Eng, C.; Parise, J. B.; Toby, B. H.; Grey, 89, 197.
C. P. Chem. Mater. 2002, 14, 194. (97) Bruce, P. G.; Armstrong, A. R.; Gitzendanner, R. L. J. Mater.
(51) Tucker, M. C.; Reimer, J. A.; Cairns, E. J. J. Electrochem. Soc. Chem. 1999, 9, 193.
2001, 148, A951. (98) Shao-Horn, Y.; Hackney, S. A.; Armstrong, A. R.; Bruce, P. G.;
(52) Carlier, D.; Ménétrier, M.; Grey, C. P.; Delmas, C.; Ceder, G. Gitzendanner; Johnson, C. S.; Thackeray, M. M. J. Electrochem.
Phys. Rev. B 2003, 67, 174103. Soc. 1999, 146, 2404.
(53) Burley, J.; Battle, P. D.; Gallon, D. J.; Sloan, J.; Grey, C. P.; (99) Wang, H.; Jang, Y.-I.; Chiang, Y.-M. Electrochem. Solid State
Rosseinsky, M. J. J. Am. Chem. Soc. 2002, 124, 620. Lett. 1999, 2, 490.
(54) Greedan, J. E.; Raju, N. P.; Davidson, I. J. J. Solid State Chem. (100) Davidson, I. J.; McMillan, R. S.; Murray, J. J. J. Power Sources
1997, 128, 209. 1995, 54, 205.
(55) Levasseur, S.; Menetrier, M.; Suard, E.; Delmas, C. Solid State (101) Ammundsen, B.; Paulsen, J. M.; Davidson, I. J.; R. S. Liu, R.
Ionics 2000, 128, 11. S.; Shen, C. H.; Chen, J. M.; Jang, Y. J.; Lee, J. J. Electrochem.
(56) Lee, Y. J.; Grey, C. P. J. Electrochem. Soc. 2002, 149, A103. Soc. 2002, 179, A431.
(57) Verhoeven, V. W. J.; de Schepper, I. M.; Nachtegaal, G.; (102) Balasubramanian, M.; McBreen, J.; Davidson, I. J.; Whitfield,
Kentgens, A. P. M.; Kelder, E. M.; Schoonman, J.; Mulder, F. P. S.; Kargina, I. J. Electrochem. Soc. 2002, 149, A176.
M. Phys. Rev. Lett. 2001, 86, 4314. (103) Gao, Y.; Myrtle, K.; Zhang, M.; Reimers, J. N.; Dahn, J. R. Phys.
(58) Morgan, K. R.; Collier, S.; Burns, G.; Ooi, K. J. Chem. Soc., Rev. B 1996, 54, 16670-675.
Chem. Commun. 1994, 1719. (104) Lu, Z.; Dahn, J. R. J. Electrochem. Soc 2002, 149, A815.
(59) Rozier, P.; Savariault, J. M.; Galy, J.; Marichal, C.; Hirschinger, (105) Lu, Z.; Beaulieu, L. Y.; Donaberger, R. A.; Thomas, C. L.; Dahn,
J.; Granger, P. Eur. J. Solid State Inorg. Chem. 1996, 33, 1. J. R. J. Electrochem. Soc 2002, 149, A778.
(60) Stallworth, P. E.; Kostov, S.; denBoer, M. L.; Greenbaum, S. G.; (106) Yoon, W. S.; Paik, P.; Yang, X.-Q.; Balasubramanian, M.;
Lampe-Onnerud, C. J. Appl. Phys. 1998, 83, 1247. McBreen, J.; Grey, C. P. Electrochem. Solid State Lett. 2002, 5,
(61) Stallworth, P. E.; Johnson, F. S.; Greenbaum, S. G.; Passerini, A263.
S.; Flowers, J.; Smyrl, W.; Fontanella, J. J. Solid State Ionics (107) Yoon, W.-S.; Iannopollo, S.; Grey, C. P.; Carlier, D.; Gorman, J.;
2002, 146, 43. Reed, J.; Ceder, G. Electrochem. Solid State Lett. 2004, 7, A167.
(62) Gaubicher, J.; Wurm, C.; Goward, G.; Masquelier, C.; Nazar, L. (108) Grey, C. P.; Yoon, W. S.; Reed, J.; Ceder, G. Electrochem. Solid
Chem. Mater. 2000, 12, 3240. State Lett. 2004, 7, A290.
(63) Arrabito, M.; Bodoardo, S.; Penazzi, N.; Panero, S.; Reale, P.; (109) Menetrier, M.; Saadoune, I.; Levasseur, S.; Delmas, C. J. Mater.
Scrosati, B.; Wang, Y.; Guo, X.; Greenbaum, S. G. J. Power Chem. 1999, 9, 1135.
Sources 2001, 97-8, 478. (110) Siegel, R.; Hirschinger, J.; Carlier, D.; Matar, S.; Menetrier, M.;
(64) Tucker, M. C.; Doeff, M. M.; Richardson, T. J.; Finones, R.; Delmas, C. J. Phys. Chem. B 2001, 105, 4166.
Reimer, J. A.; Cairns, E. J. Electrochem. Solid State Lett. 2002, (111) Tucker, M. C.; Reimer, J. A.; Cairns, E. J.; Choi, S.; Manthiram,
5, A95. A. J. Phys. Chem. B 2002, 106, 2842.
(65) Thackeray, M. M.; David, W. I. F.; Bruce, P. G.; Goodenough, J. (112) Peeters, K. M. P. J.; van Bommel, M. J.; Neilen-ten Wolde, P.
B. Mater. Res. Bull. 1983, 18, 461. M. C.; van Hal, H. A. M.; Keur, W. C.; Kentgens, A. P. M. Solid
(66) Thackeray, M. M. Prog. Solid State Chem. 1997, 25, 1. State Ionics 1998, 112, 41.
(67) Gummow, R. J.; de-Kock, A.; Thackeray, M. M. Solid State Ionics (113) Levasseur, S.; Menetrier, M.; Shao-Horn, Y.; Gautier, L.; Au-
1994, 69, 59. demer, A.; Demazeau, G.; Largeteaus, A.; Delmas, C. Chem.
(68) Liu, W.; Kowal, K.; Farrington, G. C. J. Electrochem. Soc. 1998, Mater. 2003, 15, 348.
145, 459. (114) Levasseur, S.; Menetrier, M.; Delmas, C. J. Power Sources 2002,
(69) Amatucci, G.; Du Pasquier, A.; Blyr, A.; Zheng, T.; Tarascon, 112, 419.
J.-M. Electrochem. Acta 1999, 45, 255. (115) Alcantara, G. R.; Jaraba, M.; Lavela, P.; Tirado, J. L. Chem.
(70) Cho, J.; Thackeray, M. M. J. Electrochem. Soc. 1999, 146, 3577. Mater. 2003, 1210.
(71) Shao-Horn, Y.; Hackney, S. A.; Kahaian, A. J.; Kepler, K. D.; (116) Carlier, D.; Saadoune, I.; Croguennec, L.; Menetrier, M.; Suard,
Skinner, E.; Vaughey, J. T.; Thackeray, M. M. J. Power Sources E.; Delmas, C. Solid State Ionics 2001, 144, 263.
1999, 82, 496. (117) Gaudin, E.; Taulelle, F.; Stoyanova, R.; Zhecheva, E.; Alcantara,
(72) Mustarelli, P.; Massarotti, V.; Bini, M.; Capsoni, D. Phys. Rev. R.; Lavela, P.; Tirado, J. L. J. Phys. Chem. B 2001, 105, 8081.
B 1997, 55, 12018. (118) Stoyanova, J. R.; Zhecheva, E.; Kuzmanova, E.; Alcantara, R.;
(73) de-Kock, A.; Rossouw, M. H.; de-Picciotto, L. A.; Thackeray, M. Lavela, P.; Tirado, J. L. Solid State Ionics 2000, 128, 1.
M.; David, W. I. F.; Ibberson, R. M. Mater. Res. Bull. 1990, 25, (119) Tucker, M. C.; Doeff, M. M.; Richardson, T. J.; Finones, R.;
657. Reimer, J. A.; Cairns, E. J. J. Am. Chem. Soc. 2002, 124, 3832.
(74) Masquelier, C.; Tabuchi, M.; Ado, K.; Kanno, R.; Kobayashi, Y.; (120) Padhi, A. K.; Nanjundaswamy, K. S.; Masquelier, C.; Okada,
Maki, Y.; Nakamura, O.; Goodenough, J. B. J. Solid State Chem. S.; Goodenough, J. B. J. Electrochem. Soc. 1997, 144, 1609.
1996, 123, 225. (121) Prasad, S. S. In Handbook of solid-state batteries and capacitors;
(75) Xia, Y.; Yoshio, M. J. Electrochem. Soc. 1997, 144, 4186. Munshi, M. Z. A., Ed.; World Scientific: Singapore, 1995; p 467.
(76) Xia, Y.; Zhou, Y.; Yoshio, M. J. Electrochem. Soc. 1997, 144, 2593. (122) DeSilvestro, J.; Haas, O. J. Electrochem. Soc. 1990, 137, 5C.
(77) Gao, Y.; Dahn, J. R. J. Electrochem. Soc. 1996, 143, 100. (123) Besenhard, J. O.; Schollhorn, R. J. Electrochem. Soc. 1977, 124,
(78) Paulsen, J. M.; Dahn, J. R. Chem. Mater. 1999, 11, 3065. 968.
(79) Yamada, A. J. Solid State Chem. 1996, 122, 160. (124) Campanella, L.; Pistoia, G. J. Electrochem. Soc. 1971, 128, 1905.
(80) Lee, Y.; Wang, F.; Mukerjee, S.; McBreen, J.; Grey, C. J. (125) Dickens, P. G.; Reynolds, G. J. Solid State Ionics 1981, 5, 331.
Electrochem. Soc. 2000, 147, 803. (126) Besenhard, J. O.; Heydecke, J.; Fritz, H. P. Solid State Ionics
(81) Xia, Y.; Yoshio, M. J. Electrochem. Soc. 1996, 143, 825. 1982, 6, 215.
(82) Berg, H.; Kelder, E. M.; Thomas, J. O. J. Mater. Chem. 1999, 9, (127) Julien, C.; Khelfa, A.; Guesdon, J. P.; Gorenstein, A. Appl. Phys.
427. 1994, A59, 173.
(83) David, W. I. F.; Thackeray, M. M.; de-Picciotto, L. A.; Good- (128) Rehder, D. Bull. Magn. Reson. 1982, 4, 33.
enough, J. B. J. Solid State Chem. 1987, 67, 316. (129) Mastikhin, V. M.; Lapina, O. B.; Krasilnikov, V. N.; Ivakin, A.
(84) Hunter, J. C. J. Solid State Chem. 1981, 39, 142. A. React. Kinet. Catal. Lett. 1984, 24, 119.
(85) Ooi, K.; Miyai, Y.; Sakakihara, J. Langmuir 1991, 7, 1167. (130) Zamarev, K. I.; Mastikhin, V. M. Colloids Surf. 1984, 12, 401.
(86) Sato, K.; Poojary, D. M.; Clearfield, A.; Kohno, M.; Inoue, Y. J. (131) Fotiev, A.; Slobodin, B. V.; Hodos, M. Y. Vanadates, their
Solid State Chem. 1997, 131, 84. Synthesis, Composition and Properties; Nauka: Moscow (in
(87) Pickup, D. M.; Simon, D.; Fooken, M.; Krampitz, H.; van Eck, Russian), 1988.
E. R. H.; Kelder, E. M. J. Mater. Chem. 2003, 13, 963. (132) Hawthorne, F. G.; Calvo, C. J. Solid State Chem. 1977, 22, 157.
(88) Paik, Y.; Osegovic, J. P.; Wang, F.; Bowden, W.; Grey, C. P. J. (133) Fotiev, A. A.; Ivakin, A. A. Vanadates of Divalent Metals and
Am. Chem. Soc. 2001, 123, 7564. their Synthesis: Sverdlovsk (in Russian), 1970.
(134) Muller, D.; Gessner, W.; Grimmer, A. R. Z. Chem. 1977, B12, (154) Savariault, J. M.; Deramond, E.; Galy, J.; Mongrelet, T.; Hir-
453. shinger, J. Mol. Cryst. Liq. Cryst. 1994, 244, 367.
(135) Klinowski, J. Prog. NMR Spectrosc. 1984, 16, 237. (155) Galy, J.; Hardy, A. Bull. Soc. Chim. Fr. 1964, 451, 2808.
(136) Eckert, H.; Wachs, I. E. J. Phys. Chem. 1989, 93, 6796. (156) Galy, J.; Darriet, J.; Hagenmuller, P. Rev. Chim. Miner. 1971,
(137) Pletnev, R. N.; Lazukova, N. I.; Gubanov, V. A. Zh. Fiz. Khim. 8, 509.
1977, 51, 2359. (157) Bonino, F.; Ottaviani, M.; Scrosati B.; Pistoia, G. J. Electrochem.
(138) Gubanov, V. A.; Pletnev, R. N.; Lisson, V. N.; Chirkov, A. K. Soc. 1988, 135, 12.
Spectrosc. Lett. 1977, 10, 527. (158) Pistoia, G.; Pasquali, M.; Tocci, M.; Manev, V. Moshtev, R. V. J.
(139) Umeda, J. I.; Kusumoto, H.; Narita, K. J. Phys. Soc. Jpn. 1966, Power Sources 1985, 15, 13.
21, 619. (159) de Picciotto, L. A.; Adendorff, K. T.; Liles D. C.; Thackeray, M.
(140) Skibsted, J.; Nielsen, N. C.; Bildsoe, H.; Jakobsen, H. Chem. M. Solid State Ionics 1993, 62, 297.
Phys. Lett. 1992, 188(5-6), 405. (160) Wadsley, A. D. Acta Crystallogr. 1957, 10, 261.
(141) Delmaire, F.; Rigole, M.; Zhilinskaya, E. A.; Aboukais, A.; (161) Benedek, R.; Thackeray, M. M.; Yang, L. H. Phys. Rev. B 1999,
Hubaut, R.; Mairesse, G. Phys. Chem. Chem. Phys. 2000, 2, 4477. 60-9, 6335.
(142) Stallworth, P. E.; Johnson, F. S.; Greenbaum, S. G.; Passerini, (162) Jouanneau, S. These de Doctorat, Universite de Nantes, 2001.
S.; Flowers, J.; Smyrl, W. J. Appl. Phys. 2002, 92(7), 3839. (163) Nanjundaswamy, K. S.; Padhi, A. K.; Goodenough, J. B.; Okada,
(143) Hirschinger, J.; Mongrelet, T.; Marichal, C.; Granger, P.; Sa- S.; Ohtsuka, H.; Arai, H.; Yamaki, J. Solid State Ionics 1996,
variault, J.-M.; Déramond, E.; Galy, J. J. Phys. Chem. 1993, 97, 92, 1.
10301. (164) Gaubicher, J.; Le Mercier, T.; Chabre, Y.; Angenault, J.; Quarton,
(144) Dupré, N.; Gaubicher, J.; Siegel, R.; Brunelli, M.; Hirschinger, M. J. Electrochem. Soc. 1999, 146, 4375-4379.
J.; Quarton, M. Manuscript in preparation. (165) Gaubicher, J.; Chabre, Y.; Angenault, J.; Lautié, A.; Quarton,
(145) Dupré, N.; Gaubicher, J.; Guyomard, D.; Grey, C. P. Chem. M. J. Alloys Compd. 1997, 262, 34.
Mater. 2004, 16, 2725. (166) Dupré, N.; Gaubicher, J.; Le Mercier, T.; Wallez, G.; Quarton,
(146) Murphy, D. W.; Christian, P. A.; DiSalvo, G. J.; Waszczak, J. V. M. Solid State Ionics 2001, 140, 209.
Inorg. Chem. 1979, 18, 2800. (167) Dupré, N.; Gaubicher, J.; Angenault, J.; Wallez, G.; Quarton,
(147) Whittingham, M. S. J. Electrochem. Soc. 1975, 123, 315. M. J. Power Sources 2001, 97, 532.
(148) Galy, J. J. Solid State Chem. 1992, 100, 229. (168) Jordan, B.; Calvo, C. Can. J. Chem 1973, 51, 2621.
(149) Cocciantelli, J. M.; Doumerc, J. P.; Pouchard, M.; Broussely, M.; (169) Gopal, R.; Calvo, C. J. Solid State Chem. 1972, 5, 432.
Labat, J. J. Power Sources 1991, 34, 103. (170) Tietze, R. Aust. J. Chem. 1981, 34, 2035.
(150) Delmas, C.; Cognac-Auradou, H.; Cocciantelli, J. M.; Menetrier, (171) Lapina, O. B.; Khabibulin, D. F.; Shubin, A. A.; Bondareva, V.
M.; Doumerc, J. P. Solid State Ionics 1994, 69, 257. M. J. Mol. Catal. A 2000 162, 38.
(151) Bose, M.; Basu, A. Solid State Ionics 1986, 18-19, 902. (172) Siegel, R.; Dupré, N.; Quarton, M.; Hirschinger, J. Manuscript
(152) Cocciantelli, J. M.; Suh, K. S.; Senegas, J.; Doumerc, J. P.; in preparation.
Pouchard, M. J. Phys. Chem. Solids 1992, 53, 57. (173) Yin, S.-C.; Grondey, H.; Strobel, P.; Huang, H.; Nazar, L. F. J.
(153) Cocciantelli, J. M.; Suh, K. S.; Senegas, J.; Doumerc, J. P.; Am. Chem. Soc. 2003, 125, 326.
Soubeyroux, J. L.; Pouchard, M. J. Phys. Chem. Solids 1992,
53, 51. CR020734P
Chem. Rev. 2004, 104, 4513−4534 4513
Role of Electronic Structure in the Susceptibility of Metastable

Transition-Metal Oxide Structures to Transformation
John Reed and Gerbrand Ceder*
Department of Materials Science and Engineering, Massachusetts Institute of Technology, Building 13-5056, 77 Massachusetts Avenue,
Cambridge, Massachusetts 02139
Contents transforming into more stable phases when it be-

comes thermodynamically metastable. The ability to
1. Introduction 4513 resist phase transformation can have an important
2. Transformation Mechanisms 4514 impact on the overall performance of a Li-insertion
3. Density Functional Theory 4516 compound used as an electrode material in a Li
4. Comparison between the Activation Barriers for 4516 rechargeable battery.
Co and Mn Migration The focus of this paper is on the role electronic
5. Valence of Co and Mn during Migration 4517 structure plays in determining the site preference
6. Ligand-Field Effects on the Energetics of 4518 and mobility of 3d transition-metal ions in an oxide
Migrating Co and Mn and how these factors in turn affect the resistance
7. Summary of Important Factors Influencing Co 4520 of metastable 3d transition-metal oxides against
and Mn Site Preference in ccp Oxides transformation. This is a relevant topic to the Li
8. Effect of Chemical Substitutions on Mn Site 4521 rechargeable battery field because 3d transition-
Preference metal oxides are often used as positive electrode
8.1. Electronic Structure Model for the Energetics 4524 materials.
of Mn Oxides Lithium manganese oxide structures serve as the
9. Qualitative Ionization Scale 4526 main prototype of a 3d transition-metal oxide system
10. Effect of Valence on Site Preference of Other 3d 4527 used in this investigation. Lithium manganese oxides
Transition Metals have been intensely researched as candidate positive
10.1. Ti 4528 electrode materials for use in Li rechargeable bat-
10.2. V 4528 teries because they offer the possibility of high
10.3. Cr 4529 capacity combined with good safety. This desirable
10.4. Mn 4529 combination is a consequence of the relative stability
10.5. Fe 4529 of the fully charged MnO2 structures. Mn is also less
10.6. Co 4530 expensive than Co, the transition metal commonly
used today in positive electrode materials for re-
10.7. Ni 4530
chargeable Li batteries.1
10.8. Cu 4530
The Mn oxide structures that have received the
10.11. Overall Trends for 3d Metals 4530 most attention as possible positive electrode materi-
11. Conclusions 4531 als are spinel,2-4 R-NaFeO2-type layered,5-9 and
12. Acknowledgments 4532 orthorhombic (Pmnm).2,3 All three of these structures
13. References 4532 have a close-packed (sometimes slightly distorted)
oxygen array in a fcc-like stacking with Mn occupying
octahedral sites and Li occupying either octahedral
1. Introduction or tetrahedral sites. Unfortunately, each of these
structures has problems that have hindered their
Kinetic stability is a key aspect of Li-insertion practical use in Li rechargeable batteries thus
compounds used in rechargeable Li batteries. To far.2,3,5,10,11 For the R-NaFeO2-type layered and ortho-
obtain high capacity, the Li ions need to be cycled rhombic (Pmnm) structures the primary problem is
over a wide range of concentrations within the host that they undergo structural transformation with
structure of the insertion compound. This almost electrochemical cycling. As such the Mn oxides are a
invariably brings the host structure outside its range good prototype for investigating the relationship
of thermodynamic stability at some stage of the between transformation kinetics and electronic struc-
electrochemical cycle. Maintaining a desirable host ture.
structure over repeated electrochemical cycles often One advantage that spinel-like LixMn2O4 (s-Lix-
hinges upon the host structures resistance against Mn2O4) has over the other candidates is that it is not
susceptible to any major structural transformation
* To whom correspondence should be addressed. Phone: (617) 253- upon electrochemical cycling over the 0 e XLi e 2
1581. Fax: (617) 258 6534. E-mail gceder@mit.edu. range.2 This is due in part to s-LixMn2O4 being
10.1021/cr020733x CCC: $48.50 © 2004 American Chemical Society
4514 Chemical Reviews, 2004, Vol. 104, No. 10 Reed and Ceder
The low energy of the spinel structure at the

LiMn2O4 composition is not unique to Mn. For the
3d transition metals from Ti to Cu, the energy of the
spinel structure is lower than the energy of the
layered structure.12,13 This is not too surprising since
there is a greater separation between neighboring
positively charged Li and M cations (M ≡ 3d transi-
tion-metal ion) in the spinel structureswhere the Li
occupy tetrahedral sites and the M octahedral sitess
than in the layered structureswhere both Li and M
occupy octahedral sites.2
As mentioned previously, both R-NaFeO2-type lay-
ered (l-LixMnO2) and orthorhombic (o-LixMnO2) lose
their structural integrity with electrochemical cy-
cling, in contrast to s-LixMn2O4, and rapidly trans-
John Reed received his Ph.D. degree in Materials Science and Engineering
from the Massachusetts Institute of Technology in 2003. Currently he is form to a spinel-like material at partial lithiation.9,14-17
doing postdoctoral work for the Massachusetts Institute of Technology in This results in a substantial drop in capacity over
addition to being a visiting scientist at Lawrence Berkeley National the first few charge/discharge cycles.8 However,
Laboratory. His current interests are energy technologies and utilization capacity is recovered with additional cycling as the
of computational methods to analyze materials used in these technologies, transformation to spinel moves toward completion.18
as well as identification of new candidate materials. His primary research The transformed spinel-like material is reported to
focus to date has been on transition-metal oxide materials used in the
positive electrodes of Li rechargeable batteries and the physics and have properties such as reduced Jahn-Teller distor-
chemistry that govern their properties. tion and greater durability that are actually superior
to conventionally synthesized s-LixMn2O4 spinel.15,18,19
Nonetheless, there remains an interest in making
layered manganates that can resist transforming into
spinel.
Although spinel is energetically favored at the
Li1/2MO2 composition for all of the 3d transition
metals from Ti to Cu, the rate at which the layered
structure transforms to spinel varies substantially
for each 3d metal. For LixMnO2 the transformation
from layered to spinel occurs rapidly at room tem-
perature. In the case of LixNiO2, heating is required
in order for layered to rapidly transform to spinel.20
While for LixCoO2 the transformation of layered to
spinel appears to be even more difficult, with the
conversion only detected by TEM in the surface
Gerbrand Ceder is the R. P. Simmons Professor of Materials Science layers of highly cycled l-LixCoO2.21,22 In the following
and Engineering at the Massachusetts Institute of Technology. He received sections the transformation of the layered R-NaFeO2
his engineering degree in Metallurgy and Applied Materials Science from structure to spinel will be focused upon as an
the University of Leuven, Belgium, in 1988, and his Ph.D. degree in example (with significant relevance to the battery
Materials Science from the University of California at Berkeley in 1991, field) that illustrates the influence of electronic
at which time he joined the MIT faculty. Ceder’s research interests lie in structure on the transformation kinetics of 3d transi-
computational modeling of material properties and design of novel
materials. He has published over 160 scientific papers in the fields of
tion-metal oxides.
alloy theory, oxide phase stability, high-temperature superconductors, and
Li-battery materials, and holds three current or pending U.S. patents. He 2. Transformation Mechanisms
has received numerous awards among which a Career Award from the
National Science Foundation, the Robert Lansing Hardy Award from The The transformation of l-LixMO2 or o-LixMO2 into
Metals, Minerals and Materials Society for “exceptional promise for a s-LixM2O4 (M ≡ 3d transition-metal ions) does not
successful career”, the 2004 Battery Research Award from the Electro-
chemical Society, and an award from the graduate students at MIT for require oxygen rearrangement since all the struc-
best teaching. He has worked with several U.S. and international materials tures share a cubic closed-packed (ccp) oxygen anion
companies to use modeling to design and optimize materials for high- sublattice. In all three structures M ions occupy
performance applications. He is also a co-founder of Computational octahedral interstitial positions while the Li occupy
Modeling Consultants, which provides first-principles materials modeling either octahedral or tetrahedral interstices.2 Hence,
services to industry. He is currently head of the Theory and Modeling the transformation of l-LixMnO2 or o-LixMnO2 to s-Lix-
Division of the Institute for Soldier Nano Technology at MIT and a group
leader for the Research Program on High Performance Power Sources Mn2O4 involves a rearrangement of Mn from the set
in the Center for Materials Science and Engineering. of octahedral sites characterizing l-LiMnO2 or o-
LiMnO2 to the set of octahedral sites characterizing
s-LiMnO2 (as well as rearrangement of the Li ions).
thermodynamically stable at XLi ) 1. Additionally Formation of spinel from l-LixMnO2 requires the
s-LixMn2O4 retains its structural integrity at high and movement of one-fourth of the Mn ions from the Mn
low lithiation even though it becomes energetically (111) plane into the Li (111) plane as can be seen by
metastable at those compositions.2 examining Figure 1.
Metastable Transition-Metal Oxide Structures Chemical Reviews, 2004, Vol. 104, No. 10 4515
Figure 2. The most direct path an octahedral Mn can

follow to a vacant nearest neighbor octahedral site (Oh) is
through the edge (E) shared by the two sites. This path
Figure 1. l-LixMO2 (layered) and s-LiM2O4 (spinel) struc-
(Oh f Oh) is also the most constricted in terms of separation
tures (M ≡ 3d transition metal). M occupy octahedral sites
between the migrating Mn and the surrounding oxygen
in both structures. In l-LixMO2, M and Li (and/or vacancies)
anions. The minimum separation occurs as Mn passes
alternately occupy (111) planes of the ccp oxygen sublattice.
through the center of the octahedral edge labeled E. A more
The (111) plane parallel to the M layers is indicated by
open path between octahedral sites is via a nearest
the black line between the layered and spinel structures.
neighbor (n.n.) tetrahedral site (Td). Along this path (Oh
The [111] direction is shown as well. In s-Li1/2MnO2, (111)
f Td f Oh) the minimum separation between a migrating
planes with three-fourths of the Mn alternate with (111)
Mn and surrounding oxygens occurs as Mn passes through
planes with one-fourth of the Mn. Li ions occupy tetra-
the center of the shared triangular face between the n.n.
hedral sites in the planes with one-fourth of the Mn. The
octahedron and tetrahedron (F). In a perfect octahedron
planes with three-fourths of the Mn are free of Li. In fully
the distance between the corner oxygens and the center of
lithiated spinel-like s-Li2Mn2O4, the Li move into octahe-
the triangular face (i.e., distance from O to F) is 1.155 times
dral sites. Three-fourths of the Li are in the (111) plane
the distance from corner oxygens to the center of the edge
with one-fourth of the Mn, and one-fourth of the Li are in
(i.e., distance from O to E in Figure 2).
the plane with three-fourths of the Mn.
For o-LixMnO2, 50% of the Mn ions need to change Generally, the activation barrier for a transition-
position in order to form spinel.2 The characteristics metal ion passing through the triangular oxygen face
of the orthorhombic, R-NaFeO2-type layered, and (F) along a Oh f Td f Oh type path is expected to be
spinel structures as well as many other ordered rock- less than the barrier to pass through the oxygen edge
salt structures have been covered in detail by Thack- (E) along a Oh f Oh type path. The separation
eray.2 between cation and oxygen is about 15% greater at
F than at E (assuming an undistorted octahedron).
Given that the structural transformation of the Hence, there should be less Pauli repulsion from the
R-NaFeO2-type layered and orthorhombic structures electron clouds of the oxygen when the cation passes
to spinel only requires cation migration, the varying through the octahedral face (F) than when passing
resistance of transition-metal compositions against through the octahedral edge (E). Previous work has
the transformation (i.e., low resistance for Mn, high shown that Li favors a Oh f Td f Oh type path when
for Co) is most likely connected to the diffusion diffusing in the Li layer of l-LixCoO2.23 It is expected
kinetics of the respective 3d transition-metal ions. that 3d metal ions will typically take Oh f Td f Oh
An ion can take two extreme paths in migrating type paths (Figure 2) as well when diffusing through
between octahedral sites of a ccp oxygen framework, a ccp oxygen framework, such as during the trans-
as occurs during the transformation of l-LixMnO2 or formation of l-LixMnO2 or o-LixMnO2 to s-LixMn2O4.
o-LixMnO2 to s-LixMn2O4. These two paths are shown The results of first-principles calculations discussed
in Figure 2. The most direct path travels straight in the following sections support this view.
through the edge shared by neighboring octahedra, A notable exception to the general favorability of
i.e., the Oh f Oh path through E shown in Figure 2. Oh f Td f Oh type paths for cation migration is cases
This path, while short, brings the cation in close where the intermediate tetrahedral site (Td) shares
proximity to the coordinating oxygen anions. A more a face with an octahedral site occupied by another
open but longer path is through a nearest neighbor cation. The passage of a 3d metal ion through such
(n.n.) tetrahedral site via the faces (F) it shares with tetrahedral sites is typically calculated to be high in
the neighboring octahedra (i.e., path Oh f Td f Oh energy, in some cases higher than the energy for
in Figure 2). passing through the octahedral edge (i.e., the Oh f
While the octahedra shown in Figure 2 are undis- Oh path).24 This is due to the small separation and
torted, this is generally not the case in structures hence strong repulsion between face-sharing cations
such as l-LixMnO2 and o-LixMnO2. In these cases the in a ccp oxygen framework.
octahedra are distorted by both the cationic ordering In the following sections it will be shown that first-
which breaks the cubic symmetry of the underlying principles calculations and ligand-field theory indi-
oxygen sublattice and the Jahn-Teller distortion cate that the energetics for the passage of a 3d ion
when Mn3+ is present. Consequently, not all of the like Mn through intermediate triangular (F) and
octahedral edges (E) or faces (F) that Mn can pass tetrahedral (Td) sites is highly effected by its oxida-
through are equivalent in l-LixMnO2 and o-LixMnO2. tion state.76 This suggests that the kinetics of phase
transformations involving 3d ion rearrangements in mesh were used. Both the lattice parameters and the
a ccp oxygen framework will depend strongly upon ionic coordinates were allowed to fully relax.
oxidation state. Since the maintenance of structural The bulk of the calculations in this study were
integrity is a desirable feature for an electrode performed on Mn oxide structures; however, as a
material, the effect of valence on ionic mobility is an reference, many equivalent calculations were also
important consideration when designing electrodes. carried out for LixCoO2 (0 e x e 1). This system was
For multivalent TM ions it is possible that the TM chosen for comparison with LixMnO2 because in the
ion could have a low mobility in one valence,- but a layered form it is does not undergo significant phase
high mobility in another, which could make the host transformations with electrochemical cycling despite
structure vulnerable to transformation during elec- becoming energetically metastable relative to spinel
trochemical cycling. at partial lithiation.34-37 This resistance to transfor-
mation has allowed layered LixCoO2 to become the
3. Density Functional Theory standard positive electrode material for use in com-
mercial Li rechargeable batteries.
Much of the quantitative information in this paper Finally, some calculations equivalent to those
is derived from first-principles calculations based on performed with Mn and Co were carried out for
density functional theory (DFT).25-27 Experimentally oxides of the other 3d transition metals, from Ti to
it is difficult to determine ion migration paths and Cu, to demonstrate some general principles govern-
energy barriers along migration paths in structural ing this class of materials.
transformations such as from l-LixMnO2 to s-LiMn2O4.
Examining the atomic-scale ionic movements that 4. Comparison between the Activation Barriers
could occur in such a transformation using first- for Co and Mn Migration
principles calculations can therefore be informative.
For characteristics of TM oxides such as LixMnO2 Figures 3 and 4 show the energy barriers calcu-
or LixCoO2 that can be experimentally determined, lated for Mn and Co movements out of a TM layer
it is found that the calculated results presented in octahedron and into the Li/vacancy layer of the
this paper are in good agreement with experiment.77 layered R-NaFeO2-type structure (recall that the TM
Additionally, previous studies have found that vari- cations have to migrate into the Li/vacancy layer for
ous properties of 3d TM oxides can be determined the transformation to spinel). The cation positions
with good accuracy using DFT-based methods.13,28,29 used in these calculations follow the Oh f Oh (Figure
This gives credibility to the findings of this paper 3) and Oh f Td f Oh (Figure 4) type paths shown in
which rely upon the rich and precise atomic-scale Figure 2.
detail provided by first-principles calculations. Figure 3 shows the calculated energy barrier for
Mn and Co hopping directly through an octahedral
The density functional calculations were performed
edge (E) into a Li/vacancy layer octahedron. The
using the Vienna Ab Initio Simulation Package
barrier illustrated at the top of Figure 3 is the
(VASP).30 The spin-polarized generalized gradient
calculated result when the Li content is XLi ) 0 (i.e.,
approximation,31,32 Perdue-Wang exchange correla-
MO2, M ≡ Mn or Co); the bottom plot corresponds to
tion function, and ultrasoft pseudopotentials were
used.33
Defects can be calculated in supercells that are
multiples of the unit cell for the underlying undefec-
ted structure (e.g., l-Li1/2MnO2). If the supercell is
large enough, the periodic images of the defect will
have negligible interaction, giving an approximation
of an isolated lone defect.
Such an approximation of periodicity was made for
the calculations discussed in the next section (section
4). The supercells for these calculations were com-
posed of either 12 or 32 primitive LixMO2 unit cells
(M ≡ 3d TM ion; 0 e x e 1) that contained various
M defects. The lattice parameters of the supercells
were kept constant at the parameters for the unde-
fected structure, while the ionic coordinates were
allowed to relax. A 2 × 2 × 2 k-point mesh was used
for the calculations on the 12-unit supercells and a
1 × 1 × 1 k-point mesh for the 32-unit supercells. Figure 3. Energy of Co/Mn ion along the Oh f Oh
The primitive LixMO2 unit cells used to construct the transition path from an octahedral site in the TM layer,
super cells had previously been calculated with full through a shared edge, to an octahedral site in the vacancy/
relaxation of lattice parameters as well as ionic Li layer: (top) delithiated XLi ) 0 (M4+), (bottom) half-
coordinates. lithiated XLi ) 1/2 (average M3.5+). (A (on x axis)) Layered
structure with no transition metal in the empty/lithium
Later sections will draw upon the results of calcu- layer (i.e., no defects). (B) A single TM atom located in the
lations on l-LixMO2, s-LixM2O4, and related meta- shared edge between neighboring octahedra (i.e., E in
stable crystalline structures. For these calculations Figure 2). (C) A single TM atom defect in an empty/lithium
a LixM4O8 (0 e x e 4) unit cell and a 4 × 4 × 4 k-point layer octahedron.
heel of the l-LixMnO2 material. It suggests that Mn

can easily move out of the layered configuration into
the Li/vacancy layer at this composition. This should
facilitate the rapid transformation of l-Li1/2MnO2 to
s-LiMn2O4 since Mn ions need to move from the TM
layer to the Li/vacancy layer during the transforma-
tion. The results of additional calculations, for ex-
ample, on the second half of the of the Oh f Td f Oh
path, reported elsewhere further attest to the relative
ease for Mn to move between octahedral and tetra-
hedral sites along the reaction path toward spinel at
XLi ) 1/2 in l-LixMnO2.24
Figure 4. Energy for a Mn/Co ion along the path from an Co by contrast is seen in Figures 3 and 4 to have
octahedral site in the TM ion layer to a tetrahedral site in high-energy barriers at both delithiated and partially
the Li/vacancy layer. (top) Delithiated MO2 (M ≡ Mn or lithiated compositions along either type of pathway
Co: (A) layered structure, (B) single TM ion located in (Oh f Td f Oh or Oh f Oh) into the Li/vacancy layer.
triangular face between TM layer and empty layer (i.e., Results of TM ion defect calculations at full lithiation,
F in Figure 2), (C) single tetrahedral TM defect in empty i.e., XLi ) 1, which are not shown,76 indicate that both
layer. (bottom) Half-lithiated, i.e., Li1/2MO2: (A) layered
structure, (B) Li disorder to create a trivacancy around a Co and Mn are prevented from entering the Li layer
tetrahedron in the Li layer (prevents high-energy face- by the lack of octahedral lithium vacancies.
sharing cations for tetrahedral defect), (C) single TM ion The calculated low activation barrier and defect
located in triangular face between TM layer and Li layer, energy for Mn going Oh f Td at partial lithiation is
(D) single tetrahedral TM defect in Li layer, (E) both Li consistent with the lack of stability of l-LixMnO2
and TM ion in tetrahedral sites (tetrahedral site available against transformation into spinel observed experi-
to Li due to vacancy created by TM defect).
mentally. Likewise, the high activation barriers for
all possible Co hops out of the TM layer are consistent
a Li content of XLi ) 1/2. The TM ion hop in both with the relative stability observed experimentally
cases is along the “constricted” Oh f Oh path shown for layered LixCoO2.
in Figure 2 and results in the formation of an Figures 3 and 4 illustrate that TM ion defect
octahedral Mn or Co defect in the Li/vacancy layer. energies and activation barriers to forming defects
Figure 4 shows the calculated energy barrier for appear to be highly sensitive to the Li concentration
Mn and Co hopping through an octahedral face (F) in the layered structure. The relative stability of Mn
into a nearest neighbor tetrahedron in the Li/vacancy and Co in the TM layer octahedra changes dramati-
layer at XLi ) 0 and 1/2. This path is the first half cally with increasing Li content, with Mn calculated
(Oh f Td) of the “open” Oh f Td f Oh path pictured to be more stable than Co at XLi ) 0 and far less
in Figure 2. stable at XLi ) 0.5. The next section will show that
Figures 3 and 4 catalog the plausible initial steps as the average oxidation state of the TM ions changes
in the transformation of the layered structure to from +4 to +3.5, due to the Li content changing from
spinel. Other Oh f Oh and Oh f Td f Oh cation hops XLi ) 0 to 0.5, different charge-transfer reactions
within the transition-metal layer have also been accompany tetrahedral defect formation for Mn as
calculated, but they were found to be particularly well as for Co. This provides an explanation for the
high in energy due to cationic repulsion and hence qualitative change in Mn behavior compared to Co
are ruled out.24 and highlights the important role that electronic
As expected, the barrier is calculated to be lower structure plays in the mobility of these ions.
(at a given lithium content) for hops from octahedra
to n.n. tetrahedra through a shared face (Oh f Td)
than hops directly between octahedra through an 5. Valence of Co and Mn during Migration
octahedral edge (Oh f Oh). This is consistent with Figures 3 and 4 indicate a significant variation in
experimental results that indicate Mn migrates the Mn and Co migration barrier along the Oh f Td
through tetrahedral sites during the transformation f Oh and Oh f Oh paths as the oxidation state
of orthorhombic or layered LixMnO2 into spinel.17,38 changes.
Some additional noteworthy features of the calcu- Using the calculated electron spin density, the ionic
lated results shown in Figures 3 and 4 are as follows. valences can be determined by integrating the elec-
(i) The barrier for the TM ion to leave its site in tron spin density in a sphere about the ionic centers.
the TM layer is high for both Co and Mn at MO2 Integrating spin density provides the net spin as-
composition. sociated with a given TM ion. This captures the
(ii) The formation energy of a tetrahedral Mn defect formal valence of the cation better than integrating
in the layered structure at XLi ) 1/2 is negative, while the charge density because it distinguishes the
it is positive for Co. partially filled 3d orbitals of the transition-metal
(iii) The activation barriers for Mn migration are cation from the filled p orbitals of the coordinating
higher than those for Co at XLi ) 0, but they are much oxygens. With this method it is possible to detect the
lower at XLi ) 1/2. charge-transfer and/or bonding changes that occur
The relatively low energy for Mn migration out of for a TM ion as it moves along its migration path.39,40
the TM layer and into n.n. tetrahedra in the Li/ The results of such spin integrations are shown in
vacancy layer at XLi ) 1/2 is probably the Achilles Figure 5 for Co (top) and Mn (bottom) in various
mately expressed by the following equation (where

Mtet2+ is the migrating TM ion)
2Moct3+ f Mtet2+ + Moct4+ (1)
A similar charge disproportionation reaction is re-

ported to occur during the degradation of s-LixMn2O4
with electrochemical cycling whereby the Mn2+ dis-
solves into the electrolyte.2
Using the spin integration method described above,
it was also determined that at an average formal
valence of +4 (XLi ) 0) both Mn and Co undergo a
simple tetrahedral defect reaction
Moct4+ f Mtet4+ (2)
6. Ligand-Field Effects on the Energetics of

Migrating Co and Mn
Size effects are often an important contribution to
the energy of ionic systems. However, surprisingly
cation size appears to have little if any effect on the
propensity of Mn or Co to enter tetrahedral coordina-
Figure 5. Integrated net spin for Co and Mn cations along tion in the LixMO2 ccp oxide system. According to
the Oh f Td path into the Li/vacancy layer at XLi ) 1/2. Pauling’s first rule, a smaller cation should be more
(top) Li1/2CoO2 layered Oh Co ions (labeled octa w/no Co energetically favorable in a small interstitial site
tet) have an oxidation state of ∼+3.5 (t2g5.5). The migrating than a larger cation.41 The tetrahedral sites formed
Co in triangular (octa/tetra face) or tetrahedral coordina- by oxygen anions in a ccp structure are smaller than
tion takes on nearly 3 units of electron spin (1/2µβ), giving
an oxidation state approaching +2 (e4t23). When one-fourth the octahedral sites, so by Pauling’s first rule smaller
of the Co are in Td sites with +2 charge, the other three- cations should have a greater propensity to enter
fourths of the Co in Oh sites (octa w/1/4 Co tet) have a tetrahedral coordination than larger cations. How-
raised oxidation state from ∼+3.6 to +3.7. (bottom) Li1/2- ever, the behavior of the LixMnO2 and LixCoO2
MnO2: The migrating Mn (octa/tetra face and tetrahedral) systems is very different than what is expected
has about 4.5 units of electron spin, giving an oxidation according to Pauling’s first rule.
state of ∼+2.5 (e2t22.5). Mn in Jahn-Teller-distorted octa-
hedra (octahedral w/JT) are ∼+3 (t2g3eg1); in non-Jahn- On the basis of cationic radii reported in the
Teller-distorted octahedra (octahedral w/no JT) they are scientific literature (53 pm for Cooct4+ and 40 pm for
+4 (t2g3). Cotet4+; 53 pm for Mnoct4+ and 39 pm for Mntet4+),42
one would expect the tetrahedral defect energies and
activation barriers for MnO2 and CoO2 to be roughly
positions along the Oh f Td f Oh path at tLi ) 1/2. the same. However, as seen previously, the calculated
Figure 5 shows the amount of electron spin (in units energy for the tetrahedral Mn4+ defect is almost twice
of 1/2 µβ) as a function of integration radius. The total as high as that of Co4+. The activation barrier for the
spin initially rises rapidly with radius as the d octahedron-tetrahedron hop is also significantly
orbitals are integrated over. However, then the spin higher for Mn4+ (Figure 4 top).
levels off as the nonpolarized oxygen states are For the composition Li1/2MO2 (M ≡ Co, Mn), the
reached. The value at this plateau is used in deter- relevant cations to consider are M2+ in tetrahedral
mining the formal valence of the TM cations. For coordination, M3+ in octahedral coordination, and M4+
example, the bottom of Figure 5 shows Mn4+ which (given above) in octahedral coordination (58 pm for
plateaus at a total electron spin of 3, reflecting the Cotet2+ and 54.5 pm for Cooct3+; 66 pm for Mntet2+ and
t2g3eg0 d-orbital filling of Mn4+. 64.5 pm for Mnoct3+).42
Even though the average formal valence state of The radius of Co is less than or equal to the radius
the TM is +3.5 at XLi ) 1/2, Figure 5 shows that the of Mn at all oxidation states and coordinations
migrating Mn (Co) ion in triangular or tetrahedral relevant to the Li1/2MO2 composition. However, as
coordination takes on quite a different valence from discussed in the previous section, Mn2+ is calculated
the surrounding octahedral Mn (Co) ions. The mi- to be more energetically favored in tetrahedral
grating Mn (Co) ion gains d electrons when it passes coordination and have a lower activation barrier for
through the triangular octahedron face and keeps the Oh f Td hop than Co2+, despite being the larger
them as it continues into the neighboring tetrahedral cation.
site in the Li/vacancy layer. Finally, according to Pauling’s first rule one would
The spin integration results of Figure 5 indicate expect lower tetrahedral defect and activation barrier
that when forming a tetrahedral Mn (Co) defect at energies in MO2 than in Li1/2MO2 for both Co and Mn
XLi ) 1/2, the migrating Mn (Co) approaches a +2 since the +4 cations have smaller radii than the +2
valence state while a neighboring octahedral Mn (Co) cations. Again, site-occupancy predictions based on
is oxidized toward +4. This constitutes a charge- Pauling’s first rule run contrary to the results
disproportionation reaction which can be approxi- calculated from first principles.
Figure 7. Change in LFSE associated with the charge

disproportionation reaction in both Li1/2MnO2 (high-spin
Mn3+) and Li1/2CoO2 (low-spin octahedral Co). ∆o and ∆t
equal the octahedral and tetrahedral ligand-field energy
splittings, respectively. The Jahn-Teller splitting equals
Figure 6. Energy splitting of the d orbitals in octahedral R. The proportionality between ∆o and ∆t is taken from
and tetrahedral coordination. The numbers at each level crystal-field theory to be -4/9∆o ) ∆t.46 Note that the
indicate the energy degeneracy that still remains after energy barycenter (dashed line) is drawn as a constant,
ligand-field splitting. Note that the energy barycenter (i.e., but this is not generally the case nor does it affect the
“center of energy”) need not be the same in octahedral and change in LFSE (although it certainly affects the change
tetrahedral coordination as pictured. in total energy).
Using LFT, the change in the ligand-field stabiliza-

It appears that for the layered Mn and Co oxides tion energy (LFSE) for the charge disproportionation
considered, ionic size effects do not play a significant reaction (eq 1) can be estimated for Mn and Co as
role in the preference for octahedral or tetrahedral shown in Figure 7.
sites nor in the activation barrier to hops between The change in LFSE for the charge disproportion-
the two. Consequently, size effects probably do not ation reaction that produces tetrahedral Mn at XLi
play a significant role in determining the mobility of ) 1/2 is projected to be equal to the energy of the
Mn or Co through a ccp oxide framework. In contrast, Jahn-Teller splitting (R). For Co, on the other hand,
the results indicate that valence and electronic the change in LFSE is projected to be more than
structure are more decisive factors in the site prefer- twice the energy of the octahedral ligand-field split-
ence of Mn or Co and hence in their propensity to ting (34/15∆o), which should be much larger than the
migrate through a ccp oxide framework. This is Mn Jahn-Teller splitting R.47,48 Hence, LFT indicates
consistent with the work of Goodenough that found that charge disproportionation is much more costly
valence to be an important determinant of the site in ligand-field stabilization for Co than for Mn at XLi
preference of 3d TM ions in oxides.43 ) 1/2. This is consistent with a much lower mobility
The role of electronic structure in Mn and Co site for Co3.5+ than Mn3+ in a ccp oxide and therefore a
preference and mobility can to some extent be greater resistance of metastable Co oxides such as
understood through ligand-field theory (LFT).44,45 l-Li1/2CoO2 against transformation. Experimental
LFT qualitatively explains how the degeneracy of the evidence supporting the decisive role LFSE plays in
3d orbitals is broken when a free TM ion is sur- the differing stability of the layered structures in-
rounded by coordinating anions. The ligand-field corporating Mn, Co, as well as Ni has been reported
splitting of d orbitals in octahedral and tetrahedral by Choi, Manthiram et al.49
coordination is pictured in Figure 6.45 It should be kept in mind that while LFSE is
In octahedral coordination, the d level splits into important, it is one of many contributions to the
the eg level, which is 2-fold degenerate, and the t2g energy in the LixMnO2 and LixCoO2 systems. For
level, which is 3-fold degenerate. The energy separa- example, the d levels in Figure 7 are drawn with a
tion between the t2g and eg levels is called the ligand- constant center of energy, or barycenter (indicated
field splitting (≡ ∆o). The t2g level, composed of the by the dashed line), but this is not generally the case.
dyz, dxy, and dxz orbitals modified by the octahedral Hence, in addition to a change in the splitting of the
ligand field, is lowered (2/5)∆o relative to the energy d levels with changing coordination, there can be a
barycenter, i.e., “center of energy”, of the d orbitals. shift in their average energy.
The eg level, composed of the dz2 and dx2-y2 orbitals, Also, the change in LFSE does not account for the
is raised (3/5)∆o.45 energy cost of spin pairing (two electrons with op-
For d orbitals placed in tetrahedral coordination, posing spin occupying the same orbital).50,51 This is
the t2 level contains the dyz, dxy, and dxz orbitals not relevant for Mn in this case, but for Co the change
modified by the tetrahedral ligand field and the e in LFSE of the charge disproportionation reaction is
level contains the dz2 and dx2-y2 orbitals. Figure 6 probably somewhat counteracted by the change in
illustrates how the energy levels in tetrahedral spin pairing energy (SPE), since a high-spin ion is
coordination are inverted relative to those in octa- formed from low-spin ions. The products of Co3+
hedral coordination (Figure 6). charge disproportionation have four unpaired d elec-
trons while the reactants have none, making the

products more favorable in terms of SPE (Figure 7).
7. Summary of Important Factors Influencing Co

and Mn Site Preference in ccp Oxides
The association shown in sections 4 and 5 between
the +2 oxidation state and a relatively low energy
for tetrahedral Mn defects in the layered structure
is found to carry over to the site preference of Mn in
periodic crystalline structures (likewise the high
energy for Co occupancy of tetrahedral sites). Table
1 gives the calculated energy difference between
layered Li1/2MO2 and other crystal structures with
varying amounts of tetrahedral TM cations. Note that
the XLi ) 0.5 structures with tetrahedral Mn in Table
1 are markedly lower in energy relative to the layered
structure (l-Li1/2MO2) than are those with tetrahedral
Co.
For the partially inverse spinel structure (ps-
(LiM)tet(LiM3)octO8) shown in Table 1, one-fourth of Figure 8. Formation energy versus Li concentration for
three structures of Mn oxide (top) and Co oxide (bottom):
the Co or Mn in tetrahedral sites are calculated to (0) s-LixM2O4-labeled spinel, ()) l-LixMnO2-labeled layered,
have a +2 formal valence while three-quarters in and (+) partially inverse spinel structure with 1/4M
octahedral sites adopt a +4 formal valence (giving tetrahedral (ps-(LixM)tet(LiyM3)octO8; 0 e x e 1 and 0 e y e
the overall average valence of +3.5 required by 2) labeled 1/4 Mn tet. As the Li content is increased, the
charge balance). Similar to the tetrahedral defect Li is added to the tetrahedral site first of ps-(LixM)tet-
calculations for Mn and Co at XLi ) 1/2 (Figure 4, (LiyM3)octO8, and then to the octahedral sites. For Mn, there
also is the energy of (4) a structure with one-sixth of the
bottom), the ps-(LiMn)tet(LiMn3)octO8 is calculated to Mn in tetrahedral sites at XLi ) 1/3 labeled 1/6 Mn tet with
be lower in energy than l-Li1/2MnO2 while ps-(LiCo)tet- a triangle data point and (×) a structure with one-eighth
(LiCo3)octO8 is calculated higher in energy than l-Li1/2- of the Mn in tetrahedral sites at XLi ) 1/4 labeled 1/8 Mn
CoO2. The generally low energy associated with the tet.
charge disproportionation of Mn3+ to produce Mntet2+
(eq 1) indicated by the calculations of this section, (Mn2)tet(Li2Mn2)octO8 compared to l-Li1/2MnO2 further
as well as in sections 4 and 5, is consistent with the demonstrates that only the +2 oxidation state (out
experimental observation that Mn3+ in many envi- of the oxidation states +2 through +4) seems to be
ronments favors charge disproportionation.52 correlated with the low-energy occupation of tetra-
The importance of electron supply in the energetics hedral sites by Mn.
of tetrahedral Mn is demonstrated by the inverse Hence, the passage of Mn between octahedral sites
spinel (is-(Mn2)tet(Li2Mn2)octO8) result in Table 1. In via an intermediate tetrahedral site (i.e., the “open”
is-(Mn2)tet(Li2Mn2)octO8 one-half of all the Mn are in Oh f Td f Oh path of Figure 2) is expected to be
tetrahedral coordination but there are not enough greatly facilitated when the Mn can take on a +2
electrons available for these Mn to take on a +2 valence in the tetrahedral site. The amount of Mn
oxidation state without the other Mn being oxidized ions that can become +2 is determined by the
above +4. Instead, the valences for Mn are calculated average degree of Mn oxidation which is determined
to be +3 in tetrahedral coordination and +4 in by the Li content.
octahedral coordination (giving the required average The average Li content also determines the number
valence of +3.5). The relatively high energy of is- of available tetrahedral sites that Mn can enter
without sharing faces with Li ions in n.n. octahedral
Table 1. Energy of Mn and Co Oxides with Varying sites. Therefore, while LiMnO2 has sufficient elec-
Amounts of Tetrahedral TM Ions (Comparison of Co
and Mn oxides at XLi ) 0.5)a trons available for one-half of the Mn to take on a
+2 valence through charge disproportionation of
energy (eV)/M cation
Mn3+ (eq 1), Mn movement into tetrahedral sites in
fraction Li1/2- Li1/2- l-LiMnO2 is still expected to be highly unfavorable
structure of M tet MnO2 CoO2 due to electrostatic interactions with Li in face-
l-Li1/2MO2 (layered) 0 0.0 0.0 sharing n.n. octahedra.
s-LiM2O4 (spinel) 0 -0.2489 -0.1791
(LiM)tet(LiM3)octO8 1/4 -0.0829 0.1448 This interplay between electron supply and cation
(part. inv. spin.) repulsion on the energetics of structures with tetra-
(M2)tet(Li2M2)octO8(inv. spin.) 1/2 0.2447 0.3627 hedral Mn or Co is illustrated by Figure 8, which
a
Energies are relative to the l-Li1/2MO2 structure for M ≡
gives the calculated energies per TM ion for various
Mn and M ≡ Co, respectively. The second column lists the structures over a range of Li contents. The results
fraction of TM cations in tetrahedral coordination for a given for the ps-(LixM)tet(LiyM3)octO8 structure (0 e x e 1
structure. The structure labeled (LiM)tet(LiM3)octO8 (part. inv. and 0 e y e 2) are shown for both M ≡ Co and Mn
spin.) corresponds to a partially inverse spinel and (M2)tet- (with increasing lithiation the Li was added to the
(Li2M2)octO8 (inv. spin.) to a fully inverse spinel (i.e., antispinel).
tetrahedral sites first, then octahedral). These ener-
gies, labeled 1/4 M tet with + data points in Figure interactions with face-sharing Li ions (hence the lack
8, are compared with the energies of the l-LixMO2 and of a + data point at XLi ) 1 in Figure 8).
s-LiM2O4 structures (labeled layered with diamond Figure 8 exemplifies the conflicting requirements
data points and spinel with square data points, for low-energy occupancy and passage through tet-
respectively). Additionally, for Mn the results are rahedral sites by Mn in LixMnO2 with a ccp oxide
shown for a structure with one-sixth of the Mn in framework. It requires the coexistence of both Li
tetrahedral sites at XLi ) 1/3 (labeled 1/6 Mn tet with vacancies to provide tetrahedral sites without face-
a triangle data point) and a structure with one-eighth sharing cations and Mn3+ that can form Mntet2+
of the Mn in tetrahedral sites at XLi ) 1/4 (1/8 Mn through charge disproportionation (eq 1). However,
tet with a × data point). These respective Li contents an increase in the concentration of Li vacancies
were chosen because they allow the Mn to dispro- decreases the amount of Mn3+ that can undergo
portionate so that all of the tetrahedral Mn are +2 charge disproportionation (eq 1) and vice versa. This
and all of the octahedral Mn are +4. suggests that the bulk of the Mn migration out of
When the average formal valence is +4, the cal- the transition-metal layer during the transformation
culated energy difference between ps-(Mn)tet(Mn3)octO8 of l-LixMnO2 occurs at partial lithiation when Mn3+
and l-MnO2 is larger than that between ps-(Co)tet- and Li vacancies coexist.39
(Co3)octO8 and l-CoO2. This resembles the results of The required coexistence of Mnoct3+ and Li vacan-
the tetrahedral defect calculations in l-MO2 (Figure cies for the easy migration of Mn between octahedral
4, top) where the energy of the tetrahedral Mn4+ sites via a tetrahedral intermediate also explains the
defect is calculated to be higher than the energy of ability of s-LixMn2O4 to withstand electrochemical
the Cotet4+ defect. cycling over the 0 e XLi e 2 range without significant
Figure 8 shows that with the addition of Li the cation rearrangement, even though the spinel order-
energy of ps-(LixM)tet(LiyM3)octO8 drops much more ing is thermodynamically unstable near x ) 0 and 2.
rapidly relative to the layered structure for Mn than When the spinel-like structure becomes metastable
for Co. Around the Li1/2MnO2 composition (i.e., ps- near XLi ) 0, most of the Mn are +4 and there are
(LiM)tet(LiM3)octO8), when all of the tetrahedral Mn little or no Mn3+ that can charge disproportionate.
have a +2 valence and all of the octahedral Mn are Hence, Mn passage through tetrahedral sites is
+4 (i.e., total charge disproportionation), ps-(Lix- probably very unfavorable energetically. This cuts off
Mn)tet(LiyMn3)octO8 drops below l-Li1/2MnO2 in energy. the “open” Oh f Td f Oh path of Figure 2. As a result,
The particularly steep drop in energy of ps-(LixMn)tet- the Mn are “trapped” in the metastable spinel-like
(LiyMn3)octO8 compared to l-LixMnO2 with increasing configuration (λ-MnO2) at high charge. When the
Li concentration is due to the increase in electron spinel structure becomes metastable near XLi ) 2
supply, making more Mntet2+ possible. there is a lack of Li vacancies. This also prevents Mn
rearrangement even though Mnoct3+, which can un-
For Co, the energy of ps-(LixCo)tet(LiyCo3)octO8 by
dergo charge disproportionation, are in abundance.
contrast never drops below that of l-LixCoO2. The
Consequently, the metastable s-Li2Mn2O4 is pre-
results of Figure 8 for crystalline structures, like
served at deep discharge as well. When Mnoct3+,
those in Table 1, mimic the results of the tetrahedral
which can charge disproportionate, and Li vacancies
defect calculations (Figure 4). In each case tetrahe-
coexist at one-half lithiation, the spinel structure is
dral Co is found to be unfavorable at all lithium
thermodynamically stable. Therefore, when the Mn
concentrations and oxidation states considered, while
are most prone to migration, there is no thermody-
tetrahedral Mn is found to be favorable at the
namic driving force to do so and the spinel host
Li1/2MnO2 composition when it has a +2 valence.
structure is retained (although Mn can still dissolve
Further bolstering the association of +2 valence into the electrolyte through charge disproportion-
with low-energy tetrahedral site occupancy by Mn is ation).
the relatively low energy of LixMnO2 structures with This discussion has focused on stoichiometric spinel
one-sixth and one-eighth of the Mn in tetrahedral structure, but nonstoichiometric spinels can exist as
coordination (marked with an × and a triangle in well. In the case of spinels that are oxygen deficient
Figure 8) at lithium concentrations giving the tetra- there could be significant concentrations of Mn3+
hedral Mn a +2 valence and the octahedral Mn a +4 remaining at full charge. The results of this study
valence (Li1/3MnO2 and Li1/4MnO2 compositions, re- suggest that such spinels may be susceptible to cation
spectively). rearrangement if they are energetically metastable.
For Li concentration higher than Li1/2MO2 there is
a rapid rise in energy for both ps-(LixMn)tet(Liy- 8. Effect of Chemical Substitutions on Mn Site
Mn3)octO8 and ps-(LixCo)tet(LiyCo3)octO8 even though
the tetrahedral TM ions maintain a +2 valence. The
Preference
cause of this energy rise is the strong repulsion In sections 4, 5, and 7 it was shown how low-energy
between Li+oct and Mntet2+ or Cotet2+ that share a occupation and passage through tetrahedral sites by
polyhedral face. Above a Li content of one-half Li per Mn is associated with the +2 oxidation state. It was
TM ion, the ps-(LixM)tet(LiyM3)octO8 structure can only also shown that Mnoct3+ can readily produce Mntet2+
accommodate Li in sites that share at least one face through charge disproportionation (eq 1). On the
with another cation. At the LiMO2 composition both other hand, tetrahedral Mn with a +3 or +4 oxida-
ps-(LiMn)tet(Li3Mn3)octO8 and ps-(LiCo)tet(Li3Co3)octO8 tion state was found to be less favorable.
are unstable with the tetrahedral Mn and Co being Chemical substitutions that oxidize Mnoct3+, which
forced back into the TM layer octahedra by repulsive might otherwise produce Mntet2+ through charge
disproportionation, are a promising approach for

hindering the transformation of metastable host
structures such as l-LixMnO2. Such chemical substi-
tutions would be expected to hinder Mn passage
along Oh f Td f Oh (Figure 2) type paths.
There are two categories of elements that will likely
produce the desired effect: fixed low-valence cations
and electronegative (relative to Mnoct3+) multivalent
cations. It should be noted that fixed low-valence
cations have the drawback of reducing the capacity
since it is extremely difficult to oxidize Mn above +4.
Some cations in these categories that have already
been shown experimentally to improve the electro-
chemical performance of l-LixMnO2 and/or o-LixMnO2
include Al3+,18 Co3+,7 Cr3+,53 Ni2+,54 Li+, and Cr3+.55 Figure 9. Energy difference between ps-(LiMn)tet(Li-
Sub3)octO8 with tetrahedral Mn and l-Li1/2Mn1/4Sub3/4O2
A series of first-principles results will now be without tetrahedral Mn as a function of the spin on the
shown that demonstrate the effect of a variety of tetrahedral Mn (Tetra Mn spin dx). Sub ≡ elements
chemical substitutions on the valence of Mn and how substituted for Mn and/or Mn. Pure Li1/2MnO2 is labeled
this in turn affects the site preference of Mn. “pure”. Each data point (*) has been labeled according to
As discussed previously, a strong preference by Mn the element substituted (e.g., Co) and the fraction of Mn
for octahedral over tetrahedral coordination should in the “pure” structural counterpart they have replaced
(e.g., 0.25). The substituted elements occupy only Oh sites
result in reduced mobility for Mn through a ccp in both ps-(LiMn)tet(LiSub3)octO8 and l-Li1/2Mn1/4Sub3/4O2.
oxygen framework. A reduced mobility for Mn in turn Mn occupies Td sites and any available Oh sites in ps-(Li-
could increase the resistance of metastable chemi- Mn)tet(LiSub3)octO8 and only Oh sites in l-Li1/2Mn1/4Sub3/4O2.
cally substituted Mn oxide structures against struc- For data points that are clumped together, a single label
tural transformation. containing all the chemical substitutions contained in the
cluster of points is enclosed in parentheses (e.g., 0.25Cr,
The results of sections 4, 5, and 7 indicate that 0.25Fe). The chemical substitutions listed in these labels
there is a good correlation between the energetics of are ordered (going from top to bottom in the label) from
a tetrahedral Mn defect in l-LixMnO2 and the energy lowest Mntet d spin to highest, i.e., from left-most data point
difference between l-LixMnO2 and a periodic small to right-most in the cluster. The data points for 0.25 Cu
unit cell structure with tetrahedral Mn like ps-(Lix- and 0.75 Ni cannot be resolved because they have nearly
Mn)tet(LiyMn3)octO8 provided the Li contents and the same coordinates: (0.25 Cu, 3.517, 0.887 eV); (0.75 Ni,
tetrahedral Mn oxidation states are the same in both 3.519, 0.886 eV).
the periodic structure and the tetrahedral Mn defect tity not the energy in tetrahedral coordination.
calculation. Specifically, Mntet4+ is found to be rela- However, if the energy of tetrahedral site occupancy
tively unfavorable whether it occurs as a defect in for Mn is high, the activation barrier for the Oh f
layered l-MnO2 or within a periodic structure like ps- Td f Oh path can only be equal or higher. Therefore,
(Mn)tet(Mn3)octO8. Likewise, Mntet2+ is found to be the energetics of Mn in tetrahedral coordination can
relatively favorable whether it occurs as a defect in provide an upper bound on the mobility of Mn along
layered l-Li1/2MnO2 or within ps-(LiMn)tet(LiMn3)octO8. the Oh f Td f Oh path.
To gauge the favorability of Mn entering tetrahe- Figure 9 shows the energy difference between ps-
dral sites in the presence of chemical substitutions (LiMn)tet(LiSub3)octO8 (Sub ≡ Mn and/or elements
that alter the Mn oxidation state, the energy differ- substituted for Mn) and l-Li1/2Mn1/4Sub3/4O2, which
ence between structures such as ps-(LiMn)tet(Li- is believed to be a particularly good gauge for the
Sub3)octO8 (Sub ≡ Mn and/or substitutional elements) stability of the layered structure. The reason for this
with tetrahedral Mn and l-Li1/2Mn1/4Sub3/4O2 (Sub ≡ is that ps-(LiMn)tet(LiSub3)octO8 is equivalent to a
Mn and/or substitutional elements) without tetrahe- tetrahedral Mn and Li defect placed in a small
dral Mn have been evaluated over a large range of supercell of layered structure (four MnO2 units as
chemical substitutions. opposed to the 12 and 32 MnO2 unit supercells used
It is assumed that as with pure Mn oxides, the site in section 4). Hence, its energy should be related to
preference for Mn at a given valence will be reflected the Mn tetrahedral defect energy in layered at a
by the energy difference between the chemically given composition. Furthermore, the partially inverse
substituted compounds with and without tetrahedral spinel (ps-(LiMn)tet(LiSub3)octO8) can be easily formed
Mn. The use of simple structures such as ps-(LiMn)tet- from l-Li1/2Mn1/4Sub3/4O2 by moving one-fourth of the
(LiSub3)octO8 with tetrahedral Mn instead of large Mn into nearest neighbor tetrahedra in the Li layer,
supercells with tetrahedral Mn defects, like those which equals the fraction of Mn that move from the
used in section 4, greatly reduces the calculation TM layer to the Li layer during the transformation
time. of layered to spinel (i.e., one Mn per eight oxygen).78,79
Obviously it is an approximation to use the energy Therefore, ps-(LiMn)tet(LiSub3)octO8 could resemble
difference between a structure with tetrahedral Mn intermediate structures that arise during the trans-
and a structure with only octahedral Mn as an formation.
indication of Mn mobility in a ccp oxide. To precisely Consistent with the proposed importance of Mn
determine Mn mobility along the “open” Oh f Td f valence and LFSE, low fixed valence cations (e.g.,
Oh path, the activation barrier is the required quan- Al3+, Mg2+, Li+) and electronegative multivalent
cations (e.g., Co3+, Ni2+, Cu2+) that oxidize Mn3+ are

calculated to stabilize Mn in the octahedral sites of
l-Li1/2Mn1/4Sub3/4O2. This is indicated in Figure 9 by
the ps-(LiMn)tet(LiSub3)octO8 with these chemical
substitutions (i.e., Sub ≡ Al, Mg, Li, Co, Ni, or Cu)
being higher in energy than l-Li1/2Mn1/4Sub3/4O2
rather than lower as is the case for pure Li1/2MnO2
(labeled “pure” in Figure 9).
The correspondence between chemical substitu-
tions that have been reported to experimentally
stabilize layered l-LixMnO2 material and/or improve
its capacity (e.g., Al3+, Co3+, Cr3+, Ni2+, Li+) and the Figure 10. Equivalent plot to Figure 9 except ∆E gives
chemical substitutions that are calculated to stabilize the energy difference between fully inverse spinel, i.e.,
Mn in the layered octahedral sites according to antispinel (is-Mntet(LiSub)octO4; Sub ≡ substitutional ele-
ments or Mn) and spinel (s-Litet(MnSub)octO4). Also, the x
Figure 9 is reassuring.7,18,53-55 Another encouraging axis is now the average spin on the tetrahedral Mn since
feature of Figure 9 is that none of the elements that there are two tetrahedral Mn per inverse spinel unit cell.
are found in compositions calculated to destabilize All of the chemical substitutions replace 50% of the Mn
the layered structure (e.g., Ti4+, Zr4+, Sn4+, V5+) have ions. In the anti-spinel, all of the octahedral Mn are
been reported experimentally to be successful in chemically substituted. In the spinel half of the octahedral
stabilizing l-LixMnO2 material. Mn are chemically substituted.
Cr substitution provides a particular example that
supports interpreting the results of Figure 9 as a
measure of the layered structures stability against
transformation. Experiments have shown that reduc-
ing the Mn/Cr ratio in Cr-substituted l-LixMnO2
reduces in size or eliminates the spinel-type step in
the voltage curve that arises during electrochemical
cycling of pure l-LixMnO2.53,56 This implies that Cr
substitution can hinder or prevent the transformation
of the layered structure to spinel.
Davidson et al. found that replacing one-fourth of
the Mn in l-LixMnO2 with Cr fails to prevent a step
in the voltage curve from developing with electro- Figure 11. Equivalent plot to Figure 9 at XLi ) 0.25.
chemical cycling.53 The calculated energy of the between the valence of Mn and its energy in tetra-
partially inverse spinel (ps-(LiMn)tet(LiCrMn2)octO8) hedral coordination. Starting at Mntet2+ (i.e., d5
with one-fourth of the Mn substituted with Cr is filling), the relative energy of ps-(LiMn)tet(LiSub3)octO8
-0.1494 eV/(Mntet) lower in energy than l-Li1/2(Cr1/4- rises roughly linearly with increasing valence (i.e.,
Mn3/4)O2. Using this result as a gauge of stability decreasing d filling) to a maximum peak at around
suggests that the layered structure with one-fourth +4 valence (i.e., d3 filling) for the tetrahedral Mn.
Cr should still be susceptible to forming tetrahedral This is consistent with calculations shown in sec-
Mn and hence should still be susceptible to trans- tions 4, 5, and 7 that found the tetrahedral Mn4+
forming into spinel, consistent with experimental defect (Figure 4) as well as the delithiated ps-(Mn)tet-
observation. (Mn3)octO8 (Figure 8) to be particularly high in energy.
A compound with one-half of the Mn ions substi- It is also consistent with experimental results that
tuted by Cr was found to have no spinel step in its show chemical substitutions which oxidize Mn to +4
voltage curve.53 The energy of (LiMn)tet(LiCr2Mn)octO8 such as Ni40 increase the stability of the layered
for this case is calculated to be 0.2806 eV/(Mntet) structure.54 When Mn is oxidized to +4 it becomes,
higher in energy than l-Li1/2(Cr1/2Mn1/2)O2.24 This practically speaking, electrochemically inactive in
result suggests that forming tetrahedral Mn in the l-LixMO2 materials due to the great difficultly in
layered structure with one-half Cr should be unfa- oxidizing Mn above +4.
vorable, and hence the structure should resist trans- Figure 9 indicates that chemical substitutions
forming into spinel, again consistent with observa- which oxidize Mn stabilize the layered structure
tion. While Figure 9 may give a measure of the against transformation only up to a point. At valences
stability of chemically substituted layered compounds higher than +4, i.e., tetrahedral Mn orbital fillings
against Mn migration into the Li/vacancy layer, it less than d3, the trend abruptly shifts (Figure 9).
should be kept in mind that the elements substituted Although in reality such valences are rare for Mn in
for Mn may themselves be prone to migration into ccp oxides, Mn is predicted to become less stable in
the Li/vacancy layer. However, in the Davidson the layered octahedral sites with valences increasing
experiment it appears that both the Cr and Mn resist above +4.
migrating into the Li/vacancy layer when the Mn is Figures 10-13 give further confirmation that the
sufficiently oxidized. maximum energy for Mn occupation of tetrahedral
In addition to revealing which chemical substitu- sites in a ccp oxide occurs when the Mn valence is
tions could prevent the transformation of the layer +4 (i.e., d3 filling), independent of cation ordering.
structure, Figure 9 also illustrates the relation Figures
Figure 14. In tetrahedral coordination (Td) the d-orbital

splitting is opposite and smaller in magnitude than the
d-orbital splitting in octahedral coordination (Oh). Conse-
quently, the transfer of electrons from d orbitals in an Oh
Figure 12. Equivalent plot to Figure 9 at XLi ) 0. ligand field to d orbitals in a Td field as Mn moves from an
octahedral to a tetrahedral site falls into three different
regimes: x e 2, 2 < x e 3, and 3 < x e 5. These regimes
are distinguished by the highest occupied d orbital (HODO)
in the Oh and Td fields. Only integer fillings of the d shell
are pictured, but fractional fillings can occur as well.
moves from octahedral to tetrahedral coordination as

illustrated in Figure 14 (this model neglects the
splitting of the eg level by Jahn-Teller distortion).
Figure 14 schematically pictures how the d orbitals
change for a Mn that moves from octahedral (Oh)
coordination (e.g., in the layered structure) to tetra-
hedral (Td) coordination (e.g., tetrahedral site in ps-
Figure 13. Energy difference between the s-(Mn)tet- (LiMn)tet(LiSub3)octO8).
(Sub2)octO4 spinel and rs-(MnSub2)octO4 cation-deficient The different regimes that occur as a function of
rock-salt structures. the tetrahedral Mn d-orbital filling (dx) are as follows.
(1) x e 2. In this regime as Mn moves from
10-13 contain energy plots equivalent to Figure 9 octahedral to tetrahedral coordination the d electrons
that were generated from comparisons between fully move from t2g (lowered octahedral d) to e (lowered
inverse spinel (is-(Mn)tet(LiSub)octO4) and spinel tetrahedral d) orbitals. The energy difference between
(s-Li(MnSub)O4), ps-(LiMn)tet(Sub3)octO8 and l-Li1/4- tetrahedral and octahedral Mn is given by the energy
Mn1/4Sub3/4O2, ps-(Mn)tet(Sub3)octO8 and l-Mn1/4Sub3/4O2, difference between the filled e and t2g orbitals plus a
spinel s-(Mn)tet(Sub2)octO4 and cation-deficient rock- constant (∆EMn7+) that accounts for other energy
salt rs-(MnSub2)octO4. Figure 10 is particularly no- contributions independent of d-orbital filling (∆EMn7+
table because it illustrates how chemical substitu- is the ∆E intercept at d0 of Figures 9-13).
tions which reduce Mntet toward an ideal +2 d-orbital
filling (e.g., 0.5 Nb), or that oxidize Mntet well above ∆Eoctftet ) (Ee - Et2g)x + ∆EMn7+ (3)
+4 (e.g., 0.5 Mg) are calculated to actually make the
fully inverse spinel with tetrahedral Mn more stable (2) 2 < x e 3. The energy difference in this regime
than spinel at one-half lithiation. includes the contribution from the x e 2 regime plus
the energy difference between the filled t2 (raised
8.1. Electronic Structure Model for the Energetics tetrahedral d) and the t2g (lowered octahedral d)
of Mn Oxides orbitals giving
Figures 9-13 show that the energy difference
between structures with and without tetrahedral Mn ∆Eoctftet ) (Et2 - Et2g)(x - 2) + 2(Ee - Et2g) +
is approximately a linear function of d-orbital filling ∆EMn7+ (4)
on the tetrahedral Mn within certain ranges.
One can intuitively understand the piecewise linear (3) 3 < x e 5. The energy difference in this regime
structure of the plots in Figures 9-13 using a simple includes the contribution from the x e 2 and 2 < x e
model based on the change in electronic structure 3 regimes plus the energy difference between the
when Mn moves from an octahedral to a tetrahedral filled t2 (raised tetrahedral d) and eg (raised octahe-
site. In this model it is assumed that the slopes of dral d) orbitals giving
the lines in Figures 9-13 are equal to the energy
difference between the energetically highest occupied ∆Eoctftet ) (Et2 - Eeg)(x - 3) + (Et2 - Et2g) +
d orbital (HODO) of the tetrahedral Mn and the
HODO of the octahedral Mn. This assumption is 2(Ee - Et2g) + ∆EMn7+ (5)
consistent with electrons being transferred between
octahedral d levels and tetrahedral d levels that are This simple model, which gives a piecewise linear
fixed with respect to each other independent of relationship between the energy of tetrahedral Mn
oxidation state. and its valence, fits the results of Figures 9-13
Given the typical d-orbital splitting for tetrahedral surprisingly well, considering the wide variety of
and octahedral environments this leads to three single and multivalent cation substitutions used in
different regimes for the energy change when Mn generating these plots. This again indicates the
the lines fit to Figures 9-13. The octahedral ligand-

field splitting is found to be roughly the same for all
of the MnO2 host structures and Li contents consid-
ered (i.e., Figures 9-12), ranging from ∆o ) 2.1 to
2.3 eV. These values resemble the ligand-field split-
ting reported experimentally for MnO2 of 2.5 eV.57
The estimated octahedral ligand-field splitting for the
Figure 15. Schematic of the charge disproportionation Mn3O4 structures on the other hand (Figure 13) is
reaction which involves two Mn as opposed to the single
Mn reactions shown in Figure 14. Charge disproportion-
found to much lower at ∆o ) 1.2 eV. This is close to
ation can occur over the 3 < x e 5 filling range. The set of the ligand-field splitting reported experimentally for
orbitals Oh and Td linked by a right arrow (f) correspond MnO of 1.3 eV.57
to the Mn moving from octahedral to tetrahedral coordina- In addition to having a lower estimated ∆o, the
tion. The other set of orbitals labeled Oh correspond to a calculations on Mn3O4 (Figure 13) differ from the
Mn that remains in octahedral coordination. Integer fillings MnO2 host structures in having the y intercept
of the d shell are pictured, but fractional fillings can occur
as well. ∆EMn7+ shifted to such a negative value that the
tetrahedral Mn structure (s-(Mn)tet(Sub2)octO4) is
substantial role Mn valence plays in determining its always lower in energy than the structure with only
site energy. octahedral Mn (rs-(MnSub2)octO4), although a maxi-
The model of eqs 3-5 suggests that the energy of mum in energy at Mntet4+ is maintained.
a manganese oxide can be conveniently separated While the piecewise linear regions and energy
into two contributions. One contribution is from the maximum around +4 valence of Figures 9-13 are
interactions involving Li cations, O anions, Mn7+ (i.e., consistent with ligand-field effects, it is important to
d0) ionic cores, and elements substituting for Mn (d0 bear in mind that LFSE cannot by itself predict the
ionic cores if the d orbitals of these elements are energy difference between octahedral and tetrahedral
filled). The difference in this energy term between Mn.
the structures with and without tetrahedral Mn
For example, in section 6 the change in LFSE
equals the y intercept (∆EMn7+) of ∆Eoctftet given by
associated with the charge disproportionation reac-
eqs 3-5. The other energy contribution is from the
tion creating tetrahedral Mn2+ (eq 1) is projected to
energy of the filled Mn d orbitals (and in some
be equal to the Jahn-Teller splitting R, i.e., a positive
instances the substitutional element d orbitals) which
energy (see Figure 7). However, first-principles cal-
gives the piecewise linear valence dependent terms
culations show that the energy of producing tetra-
of ∆Eoctftet in eqs 3-5.
hedral Mn2+ through charge disproportionation (eq
The simple d-level splitting model of eqs 3-5
1) in l-Li1/2MnO2 is negative (Figures 4, 8, and 9).
explains well the abrupt change in slope as the
orbital filling moves from the 2 < x e 3 regime to The possibility that Mn generally favors tetra-
the 3 < x e 5 regime (Figures 9-12). When the filling hedral coordination as its valence approaches +2 (i.e.,
of the Mn d orbitals exceeds d3, the HODO of the d5) is unlikely given that MnO has a rock-salt
octahedral Mn goes from being lower in energy (t2g) structure not zinc blende or some other structure
to higher in energy (eg) than the HODO of the with Mntet2+. Instead, the driving force for Mn move-
tetrahedral Mn (t2). Therefore, at d3 the slope of ment out of the octahedral sites of l-Li1/2MnO2 into
∆Eoctftet as a function of Mntet d filling abruptly neighboring Li layer tetrahedral sites appears to be
changes sign from positive to negative. The x e 2 due to the unique cationic ordering and associated
regime cannot be resolved from the 2 < x e 3 regime cationic interactions that are present in l-Li1/2MnO2.
in Figures 9-13, which suggests that there is little In the case of l-Li1/2MnO2, the positive change in
splitting between the Mn d orbitals in tetrahedral LFSE for charge disproportionation (i.e., R) is insuf-
coordination (i.e., t2 and e). ficient to counter the cationic interactions that favor
While Figure 14 only illustrates a single Mn7+ ion Mn movement into Li layer tetrahedra. Conversely,
core and its associated d electrons moving from Oh in chemically substituted compounds such as l-Li1/2-
to Td coordination, the proposed model of eqs 3-5 can Mn3/4Mg1/4O2, the much higher change in LFSE
also be used to account for charge disproportionation associated with Mn4+ movement from octahedral to
(eq 1). When charge disproportionation occurs an tetrahedral coordination (38/45 ∆o assuming -4/9∆o
additional electron is transferred to the tetrahedral ) ∆t) overwhelms the interactions favoring the
Mn t2 orbital from a Mn eg orbital (or perhaps another formation of tetrahedral Mn4+ so that it becomes
multivalent cation d orbital) that remains in octahe- highly unfavorable energetically.
dral coordination. The HODOs over the 3 < x e 5 While LFSE can be useful for explaining the trends
d-filling regime (i.e., t2 tetrahedral, eg octahedral) of Mn site preference with valence, it does not capture
remain the same whether charge disproportionation important energy contributions that are more sensi-
occurs or not. Therefore, the presence of charge tive to cationic ordering. These energy contributions,
disproportionation should not change the slope of the for example, make the formation of tetrahedral Mn2+
plot over the 3 < x e 5 regime according to eq 5. This favorable in l-Li1/2MnO2 but unfavorable in s-LiMn2O4.
can be seen by comparing the 3 < x e 5 regime of A more clear picture of these energy contributions
Figure 14 with the disproportionation reaction dia- can be gained from the model of eqs 3-5.
gramed in Figure 15. One energy contribution that is sensitive to cationic
Equations 3-5 can be used to estimate the octa- ordering given explicitly in eqs 3-5 is the y intercept
hedral ligand-field splitting (∆o) from the slopes of ∆EMn7+. According to the proposed model ∆EMn7+
The change in LFSE of R associated with the charge

disproportionation reaction of Mn (Figure 7) is now
augmented with two other terms to give a more
complete expression for the total energy change: ∆Eb,
which equals the change in the d barycenter (Etb -
Eob), and ∆EMn7+, which equals the change in Mn core
interaction energies when Mn moves from an octa-
hedral to a tetrahedral site (EMntet7+ - EMnoct7+).
Equation 6 helps clarify why the energy of the
charge disproportionation reaction producing tetra-
hedral Mn in l-Li1/2MnO2 is negative. The y intercept
Figure 16. Effect of octahedral coordination on the energy of Figure 9 is close to zero, which indicates that
of TM 3d orbitals: (a) 3d level of free TM ion. 3d orbitals ∆EMn7+ ≈ 0 when changing from l-Li1/2MnO2 to (ps-
are degenerate. (b) Average interactions between anions, (LiMn)tet(LiMn3)octO8). Therefore, in this case the
neighboring cations, and 3d electrons shift the average energy of charge disproportionation is approximately
energy of the 3d orbitals. (c) Splitting of 3d energy levels ∆Ecd ) 5∆Eb + R. Since R is positive, ∆Eb must be
in an octahedral ligand field. Note that while the energy
of the 3d orbitals may be increasing due to the negatively negative for the formation of tetrahedral Mn in l-Li1/2-
charged ligands, the overall energy of the system is MnO2 to be energetically favored. In other words, the
decreasing in going from a free ion to bound one. d-orbital energy barycenter in ps-(LiM)tet(LiM3)octO8
is lower than the barycenter in l-Li1/2MnO2 according
accounts for the interactions involving Li cations, O to this model.
anions, Mn7+ (i.e., d0) ionic cores, and elements Just as values for ∆o can be estimated from Figures
substituting for Mn (d0 ionic cores if the d orbitals of 9-13 using eqs 3-5, so to can values for ∆Eb and
these elements are filled). It is assumed that these ∆EMn7+. Whereas ∆o is estimated to be roughly
interactions are primarily electrostatic. ∆EMn7+ can constant (2.1-2.3 eV) for the MnO2 host structures
have a substantial effect on site energetics as the (Figures 9-12), the values of ∆Eb and ∆EMn7+ are
difference in y-intercepts of Figures 9-13 show. estimated to vary much more widely (-0.48 to 0.29
The other energy contribution that is sensitive to eV and -2.1 to 0.36 eV, respectively). This is consis-
cationic ordering is implicitly part of d-orbital terms tent with the d-orbital barycenter and the Mn7+ core
such as (Et2 - Eeg)(x - 3) in eqs 3-5. A coefficient interactions being more sensitive to cationic ordering
such as (Et2 - Eeg), which gives the energy difference than the ligand-field splitting, as previously sug-
between the tetrahedral t2 and octahedral eg orbitals, gested.
can be broken down into two parts. Since the LFSE term is determined by Mn valence
One part depends on the ligand-field splitting in a and appears to be relatively insensitive to cation
given coordination (i.e., ∆o and ∆t). This part is the ordering in the MnO2 host structures, it would seem
change in LFSE is represented in Figure 7 (e.g., R to be the easiest energy term to manipulate through
for Mn charge disproportionation; eq 1, 38/45 ∆o for chemical substitutions. This is because the exact
Mn4+ Oh f Td). placement of the chemical substitutions in the Mn
sublattice would presumably be less important for the
The other part depends on the average energy of
LFSE term than for the terms that are more sensitive
the d orbitals, i.e., the energy barycenter, in a given
to cationic ordering (i.e., ∆Eb and ∆EMn7+).
site. The energy barycenter of d orbitals in octahedral
coordination is represented by the level above b in
Figure 16.45 In moving Mn between octahedral and 9. Qualitative Ionization Scale
tetrahedral sites there is a change in LFSE caused The previous sections highlighted the role valence
by the change in the splitting of the d orbitals (e.g., plays in determining Mn site preference and mobility
from t2g and eg to e and t2) and a change in energy through a ccp oxide framework and consequently in
barycenter caused by the change in the average the susceptibility of metastable structures such as
interactions experienced by electrons in the d orbit- l-LixMnO2 to structural transformation. Since chemi-
als. cal substitutions are one way of manipulating the
Taking into account the change in barycenter of valence of Mn, it is useful to be able to predict the
the d levels and the interactions involving the Mn7+ effect substituted cations will have on the valence of
ionic cores, a more complete expression for the energy Mn in either tetrahedral or octahedral sites. The
change (∆Ecd) associated with the Mn charge dispro- large number of calculations used to produce Figures
portionation reaction (eq 1) than that given in Figure 9-13 can be used to provide a qualitative oxidation
7 can be obtained scale between Mn and other 3d TM ions. The relative
oxidation/reduction strength of the substituted TM
6 ions that Mn coexists with in Figures 9-13 can by
(
∆Ecd ) (5Etb + EMntet7+) + 3Eob - ∆o + EMnoct7+ -
5 ) obtained by determining the valence of each cation
3 R using the spin integration method described in sec-
(
2 4Eob - ∆o - + EMnoct7+
5 2 ) tion 5.
Figure 17 holds in a compact form all of the valence
∆Ecd ) 5(Etb - Eob) + R + (EMntet7+ - EMnoct7+) information on crystalline TM oxides produced by the
first-principles calculations of this study. The position
∆Ecd ) 5∆Eb + R + ∆EMn7+ (6) of a given ion pair (e.g., Nioct3+f4+) on the chart
half Ni2+ gives an average valence of +3 required by

the Li content, the predicted valences are Mn4+ and
Ni2+. This matches calculated40 and experimental
results.59
Since Figure 17 is constructed only with calcula-
tions on Mn oxides substituted with one other 3d
metal, it is not clear whether using Figure 17 to
predict the valence of two or more non-Mn 3d TM
ions coexisting in an oxide would be valid. However,
with additional calculations on non-Mn TM oxide
compositions perhaps Figure 17 can be expanded to
address combinations of non-Mn TM ions.
10. Effect of Valence on Site Preference of Other

3d Transition Metals
Since the results of the previous sections indicate
that the site preference and tendency toward migra-
tion of Mn or Co is strongly affected by the electron
Figure 17. Qualitative ionization energies for 3d ions at occupancy of the d levels split by ligand-field effects,
various valences in an oxide framework. The position of a
given ion pair indicates the energy ranking of that ioniza- it is possible that this may be the case for all of the
tion reaction. Comparisons are relative to Mn ionization 3d TM ions.
energies in Oh and Td coordination labeled prominently in The 3d orbitals of Mn and Co and those of the other
the center of the figure. Solid black arrows indicate well- first-row transition metals should have a qualita-
determined upper and/or lower energy limits for a given tively similar ligand-field splitting in octahedral or
ionization energy. Dashed lines with a question mark tetrahedral sites of an oxide.44,45 The magnitude of
indicate that in that direction the energy limit was not
determined. For example, the ionization energy Croct2+f3+ the splitting may vary as it depends on the extent of
is calculated to be lower in energy than Mnoct2+f3+ but it overlap between the TM d and oxygen p states.46,60
is uncertain which Mn ionization energy it lies above. If However, as a first approximation it is instructive
an ionization energy lies at the same level as one of the to neglect this variation in ligand-field splitting
Mn energies and does not have an arrow on either side across the 3d series and simply characterize a TM
(e.g., Cooct3+f4+), that ionization energy and the Mn energy ion at a given valence by the number and configu-
which it is level with were found to be approximately equal
in energy (i.e., Cooct3+f4+ ≈ Mntet3+f4+). The chart entry
ration (e.g., high or low spin) of its d electrons. For
Feoct3+f4+ marked with an asterisk (*) contradicts the result example, Fe3+ would be expected to behave somewhat
calculated for anti-spinel (Mn)2(Li2Fe2)O8 but it is consis- similar to Mn2+ when both have a high-spin d5
tent with all other calculations performed on Lix(MnFe)O2 configuration (e.g., t2g3eg2 in octahedral coordination).
structures. The relative energetics of tetrahedral and octahe-
dral occupancy for Mn and Co in a ccp oxide with
indicates the relative energy ranking of that ioniza- Li1/2MO2 composition was observed to be consistent
tion reaction (e.g., the ionization energy of Nioct3+ to with the projected change in LFSE for Mn and Co
Nioct4+) compared to the various ionization energies when they are moved from octahedral to tetrahedral
of Mn in octahedral and tetrahedral coordination. coordination (Figure 7; section 6). To test how well
The qualitative ranking of the Mn ionization energies this trend holds for other 3d transition metals,
in octahedral and tetrahedral coordination is labeled calculations equivalent to those done with Mn and
prominently along the center of Figure 17. Co in section 7 were performed for the transition
The purpose of this scale is to aid in qualitatively metals Ti through Cu on the ps-(LiM)tet(LiM3)octO8
predicting the effect of 3d TM ion chemical substitu- and l-Li1/2MO2 (M ≡ 3d TM) structures as well their
tions on the valence of Mn in octahedral or tetrahe- delithiated counterparts ps-(M)tet(M3)octO8 and l-MO2.
dral sites of a ccp oxide. This scale can be useful for Table 2 gives the calculated energy difference
choosing chemical substitutions that will keep octa- (∆Eoctftet) between ps-(LiM)tet(LiM3)octO8 and l-Li1/2-
hedral Mn in a relatively immobile valence state (i.e., MO2 for the 3d transition metals from Ti to Cu
near +4) over the range of an electrochemical cycle (ordered lowest ∆Eoctftet to highest). It also lists the
where the coexistence of Mn3+ and Li vacancies would change in valence and d-orbital filling that ac-
allow the rapid transformation of a metastable ccp companies the structural change from l-Li1/2MO2 to
oxide (e.g., as with l-LixMnO2 or o-LixMnO2). ps-(LiM)tet(LiM3)octO8. The valence and d-orbital fill-
To illustrate the use of the qualitative ionization ing have been determined using spin integration as
scale given in Figure 17, consider the case of layered described in section 5. For each respective change in
Li(Ni1/2Mn1/2)O2.58 Assuming that the oxidation states d-orbital filling that accompanies the formation of
of Li and O are +1 and -2, respectively (calculated tetrahedral M, the projected change in LFSE and
to be true except in extremely electron-deficient spin-paining energy (SPE) is given. The SPE is
cases), Ni and Mn must hold an average valence of important to consider because the splitting from spin
+3. Moving up from the bottom of the scale one notes pairing (∆s) can be of comparable magnitude to the
Mnoct3+f4+ is ranked lower than Nioct2+f3+. This splitting from the ligand field in 3d transition metals
means Mn3+ is favored to be oxidized to +4 over Ni2+ (hence the existence of high-spin octahedral cations
being oxidized to +3. Since one-half Mn4+ and one- when ∆s > ∆o).50,51 ∆s is the additional repulsive
Table 2. Energy Differences between ps-(LiM)tet(LiM3)octO8 and l-Li2M4O8 (M ≡ 3d TM) for Each 3d TM from Ti to
Cu Listed from Lowest ∆E to Highesta
reaction ∆Eoctftet (eV) d filling ∆ LFSE and SPE
2Mnoct3+ f Mnoct4+ + Mntet2+ -0.33 2t2g3eg1 f t2g3 + e2t23 R
4Feoct3.5+ f Fetet3+ + 3Feoct3.67+ -0.17 4t2g41/2 f e2t23 + 3t2g41/3 2∆o - 2∆s
Cuoct f Cutetb -0.15 NA NA
4Tioct3.5+ f Titet4+ + 3Tioct3.33+ -0.12 4t2g1/2 f e0t20 + 3t2g2/3 0
Voct3.5+ f Vtet3.5+ 0.039 t2g11/2 f e11/2 (3/5)∆o - (9/10)∆t
Croct3.5+ f Crtet3.5+ 0.30 t2g21/2 f e2t21/2 ∆o - ∆t
2Nioct3+ f Nioct4+ + Nitet2+ 0.43 2t2g6eg1 f t2g6 + e4t24 (6/5)∆o - (4/5)∆t + R
4Cooct3.5+ f 3Cooct4+ + Cotet2+ 0.58 4t2g51/2 f 3t2g5 + e4t23 (14/5)∆o - (6/5)∆t - 2∆s
a Column 1 gives the valence states accompanying the movement of the TM ion from octahedral to tetrahedral coordination.
Column 2 gives the energy difference between ps-(LiM)tet(LiM3)octO8 and l-Li2M4O8 per tetrahedral M. Column 3 holds the change
in d-orbital filling that results from moving the TM ion from octahedral to tetrahedral coordination. The projected change in
LFSE and spin pairing energy (SPE) that results from the change in d-orbital filling is listed in column 4. This energy is expressed
in terms of the octahedral splitting (∆o), the tetrahedral splitting (∆t), the Jahn-Teller splitting (R), and the spin-pairing splitting
(∆s). ∆s is the additional repulsive energy from two electrons of opposing spin occupying the same d orbital. b The change in
valence/d filling for Cu could not be determined using the method of section 5.
Table 3. Energy Differences between ps-(M)tet(M3)octO8 and l-M4O8 (M ≡ 3d TM) for Each 3d TM from Ti to Cu
Listed from Lowest ∆E to Highesta
reaction ∆Eoctftet (eV) d filling ∆ LFSE and SPE
4+
Croct f Crtet 4+ 0.54 2
t2g f e 2 (4/5)∆o - (6/5)∆t
Voct f Vtetb 0.57 NA NA
Tioct4+ f Titet4+ 0.64 t2g0 f e0 0
Feoct f Fetetb 0.76 NA NA
Cuoct f Cutetb 1.0 NA NA
Cooct4+ f Cotet4+ 1.3 t2g5 f e2t23 2∆o - 2∆s
Mnoct4+ f Mntet4+ 2.1 t2g3 f e2t21 (6/5)∆o - (4/5)∆t
Nioct4+ f Nitet4+ 2.6 t2g6 f e3t23 (12/5)∆o - (3/5)∆t --2∆s
a
Column 1 gives the valence states accompanying the movement of the TM ion from octahedral to tetrahedral coordination.
Column 2 gives the energy difference between ps-(M)Tet(M3)octO8 and l-M4O8 per tetrahedral M. Column 3 holds the change in
d-orbital filling that results from moving the TM ion from octahedral to tetrahedral coordination. The projected change in LFSE
and spin-pairing energy (SPE) that results from the change in d-orbital filling is listed in column 4. This energy is expressed in
terms of the octahedral splitting (∆o), the tetrahedral splitting (∆t), and the spin-pairing splitting (∆s). b The change in valence/d
filling for Cu, Fe, and V could not be determined using the method of section 5.
energy caused by two electrons of opposing spin in a layered structure could be susceptible to migra-
occupying the same d orbital. For finding the change tion into the Li layer at partial lithiation. Experi-
in SPE of Tables 2 and 3, ∆s is assumed to be roughly mentally layered LiTiO2 has not been synthesized as
equal for both octahedrally and tetrahedrally coor- the similar ionic size between Li and Ti leads to a
dinated d orbitals. Table 3 contains the results from disordered rock-salt structure.61 Layered Li(Ni0.45-
calculations on ps-(M)tet(M3)octO8 and l-MO2 for the Ti0.55)O2 has been synthesized; however, it suffers
3d transition metals from Ti to Cu (ordered lowest from poor cyclability which is blamed on Ti migration
∆Eoctftet to highest). into the Li/vacancy layer.62 Additionally, Ti has been
Each 3d metal from Ti to Cu in Table 2 and 3 will reported to migrate into the Na layer of layered
now be discussed. NaTiO2.63 These experimental observations are con-
sistent with the relative ease Ti4+ is expected to have
10.1. Ti in moving between octahedral and tetrahedral coor-
dination due to a lack of ligand-field stabilization.
Ti is calculated to have a delocalized d band in
l-Li1/2TiO2. This is reflected by each Ti holding a +3.5
charge. In l-TiO2, Ti is calculated to have an empty
10.2. V
d band and a +4 valence. At both Li concentrations At XLi ) 1/2, V is calculated to maintain the same
the tetrahedral Ti is +4 and the projected change in valence (+3.5) in octahedral and tetrahedral coordi-
LFSE and SPE for Ti moving from octahedral to nation, i.e., there is no charge disproportionation in
tetrahedral coordination is zero. This is consistent forming tetrahedral V at this Li composition (Table
with Ti having a relatively low calculated value for 2). The projected change in LFSE and SPE at XLi )
∆Eoctftet (fourth lowest in Table 2 and third lowest 1/2 is (3/5)∆o - (9/10)∆t, which is positive whether
in Table 3). the proportionality given by crystal-field theory |4/
However, the that fact ∆Eoctftet for Ti is higher than 9∆o| ) |∆t| is assumed46 or if ∆t ≈ 0 is assumed as
Mn and Fe in Table 2 and higher than Cr in Table 3 the results in section 8 suggest for Mn. The greater
runs contrary to expectations based solely on LFSE change in LFSE and SPE for Li1/2VO2 than for
and SPE since Mn and Fe at XLi ) 1/2 and Cr at XLi Li1/2TiO2 is consistent with a greater value of ∆Eoctftet
) 0 have a projected change in LFSE and SPE that in Table 2. This suggests that V should be less prone
is greater than zero. than Ti to enter tetrahedral sites at this average
If l-LixTiO2 could be synthesized, the negative value valence, and hence l-Li1/2VO2 should be more resis-
of ∆Eoctftet for Li1/2TiO2 (-0.12 eV) suggests that Ti tant against transformation than l-Li1/2TiO2. How-
ever, ∆Eoctftet is still relatively small (0.039 eV/Vtet), change in LFSE and SPE equal to the Jahn-Teller
suggesting that V might still easily migrate out of splitting (R).47,48
metastable octahedral sites at this composition. As mentioned before, the fact that ∆Eoctftet is lower
The value of ∆Eoctftet for VO2 was calculated to be for Mn than for Ti despite R being greater than zero
the second lowest in Table 3. Unfortunately the runs contrary to expectations based solely on LFSE.
valences of V in ps-(M)tet(M3)octO8 could not be clearly According to the simple model of eqs 3-5 in section
determined using the spin integration method of 8.1, the low energy of ps-(LiMn)tet(LiMn3)octO8 com-
section 5 because the net spin on the V did not match pared to l-Li1/2MnO2 can be explained by a lower d
a spin-polarized configuration consistent with the barycenter in ps-(LiMn)tet(LiMn3)octO8 (eq 6). The d
average formal oxidation state. Instead, the V took electrons of the tetrahedral Mn2+ enter d orbitals that
on a non-spin-polarized state. are shifted lower in energy, probably because they
are surrounded by less negative charge in a tetra-
10.3. Cr hedral site than in an octahedral one. Ti4+, on the
other hand does not have any d electrons, so there is
Like V, Cr is calculated to maintain a +3.5 oxida- no energy reduction resulting from a lower tetra-
tion state in both octahedral and tetrahedral coor- hedral d barycenter.
dination at XLi ) 1/2 (Table 2). At this lithium Another factor that could make ∆Eoctftet particu-
concentration the projected change in LFSE and SPE larly low for Mn in Table 2 is the half-full orbital
is higher for Cr than for V, which is consistent with shells of Mn2+ and Mn4+ (the products of Mn3+ charge
the greater value of ∆Eoctftet calculated for Cr in disproportionation). In general, half-full or full orbital
Table 2. This supports Cr having a stronger prefer- shells have lower energy than other fillings because
ence for octahedral sites at the +3.5 valence than V. the electron-electron repulsion is decreased64 (this
At +4 valence Cr was calculated to have the lowest is why in the chemistry of low atomic number
∆Eoctftet in Table 3. This is consistent with the elements the electrons arrange so often to produce
relatively low projected change in LFSE and SPE for full “octets”). On this basis one expects Mn2+ to have
Cr4+ moving from octahedral to tetrahedral coordina- a lower electron-electron repulsion energy due to its
tion ((4/5)∆o - (6/5)∆t). half-full d shell64 (e2t23 in tetrahedral coordination).
The enhanced stability of l-Lix(Cr1/2Mn1/2)O2 re- It is also favorable for the levels that result from
ported experimentally can be explained by these ligand-field splitting to be half-full or full.50 There-
results.53 As l-Li(Cr1/2Mn1/2)O2 is delithiated, the Mn fore, the half-full t2g level (t2g3) of Mn4+ is favorable
is oxidized to +4 first according to section 9 and in this regard as well. The energetic favorability of
experiment.55 As already discussed, Mn4+ is highly half-filled levels is another reason Mn3+ in many
stable in octahedral coordination. Cr is oxidized to environments is observed to be unstable to both
+4 later in the charge, when Li vacancies that can reduction to Mn2+ and oxidation to Mn4+.52
facilitate TM ion migration are more plentiful. The
relatively high value of ∆Eoctftet calculated for Cr3.5+ 10.5. Fe
in Table 2 probably reflects a strong preference for At XLi ) 0.5 there is a large positive change in
octahedral coordination at this valence. One would LFSE (2∆o) for the formation of tetrahedral Fe.
expect Cr to have an even stronger preference for However, this is counteracted by a large negative
octahedral coordination in l-Li1/2(Cr1/2Mn1/2)O2 since change in spin-pairing energy (-2∆s). The decrease
its valence is +3 which has a t2g3 orbital fillingsthe in spin-pairing energy is caused by one of the low-
same orbital filling as Mn4+. spin Feoct2/3+ becoming high-spin Fetet3+. Recalling
While Cr4+ has the lowest ∆Eoctftet in Table 3 and that ∆s can be comparable in magnitude to ∆o in 3d
relatively weak ligand-field stabilization, the absolute metal oxides,50,51 the net change in LFSE and SPE
value of ∆Eoctftet (0.54 eV) for CrO2 is still higher than given for Fe in Table 2 is probably small. This is
that for Li1/2CrO2 (0.3 eV). Furthermore, there prob- consistent with the negative value calculated for
ably is no longer a driving force to convert to spinel ∆Eoctftet (-0.17 eV/Fe tet). However, the net change
in highly delithiated l-Lix(Cr1/2Mn1/2)O2. The fact that in LFSE and SPE should still have a positive value
the structural integrity of l-Lix(Cr1/2Mn1/2)O2 is main- since Fe is calculated to be low spin in octahedral
tained over delithiation attests to a relatively strong coordination (t2g4.5 or t2g4.33), indicating that |∆s| < |∆o|.
octahedral site preference exhibited by Cr at partial l-LiFeO2 has been synthesized by ion exchange
lithiation. from NaFeO2, but it shows little or no electrochemical
It has been observed that when Cr6+ is formed in activity.65 It was not reported whether this is a result
delithiated l-Li(Li1/5Mn2/5Cr2/5)O2, it spontaneously of Fe3+ migration; however, the results of Table 2
moves into tetrahedral coordination.55 This is con- suggest that Fe should be susceptible to migration
sistent with the lack of d electrons for Cr6+ and hence out of the TM layer at partial lithiation in l-LixFeO2
the lack of ligand-field stabilization. At XLi ) 0 Fe was found to be of intermediate
stability in octahedral coordination. Unfortunately
10.4. Mn the valences of the Fe ions could not be clearly
determined using the spin integration method of
Mn goes from having the lowest ∆Eoctftet at XLi ) section 5 because the net spin on the Fe did not
0.5 (Table 2) to the second highest ∆Eoctftet at XLi ) match a high-spin or low-spin configuration consis-
0 (Table 3). The low value of ∆Eoctftet for Mn at XLi ) tent with the average formal oxidation state (+4).
0.5 is consistent with a relatively small projected This suggests that the d-orbital filling takes on an
intermediate spin configuration for Fe and/or the due to ligand-field effects, although probably not as
assumed oxygen formal oxidation state of -2 is no strong as Ni3+ since there is an additional electron
longer justified due to strong covalency between the in the eg level.
Fe and O.
10.8. Cu
10.6. Co
At an average formal valence of +3.5, Cu is
Co has the highest value of ∆Eoctftet in Table 2. This calculated to have the third lowest ∆Eoctftet equal to
is consistent with it having the largest projected -0.15 eV per tetrahedral Cu. This suggests that Cu
change in LFSE ((14/5)∆o - (6/5)∆t). However, the at this average valence does not have a strong
change in LFSE is somewhat counteracted by a preference for octahedral sites. At an average formal
negative change in spin-pairing energy (-2∆s). The valence of +4, ∆Eoctftet for Cu (1 eV) is ranked fourth
spin-pairing energy is reduced in the charge dispro- highest in Table 3. In practice, such high oxidation
portionation of Co where four d fillings of t2g5.5 change states for Cu are probably difficult to achieve. How-
to three t2g5 and e4t23. This helps explain why ∆Eoctftet ever, if it were to occur, a metastable structure like
calculated for Co is not as high compared to the other l-LixCuO2 should be prone to transformation at
3d metals as one would expect on the basis of LFSE partial lithiation according to these results. Unfor-
alone. tunately, the valence of Cu in both cases could not
At XLi ) 0 Co has the third highest ∆Eoctftet. Its be clearly determined using the spin integration
change in LFSE and SPE is 2∆o - 2∆s and the same method of section 5.
as Fe at XLi ) 0.5, which has a low ∆Eoctftet. A
possible explanation for this is that the magnitude 10.11. Overall Trends for 3d Metals
of ∆o is higher and/or the magnitude of ∆s is lower
The energy of all the 3d metals entering tetrahe-
for Co4+ than for Fe3.5+.
dral coordination from the l-LixMO2 structure de-
The remarkable success of LiCoO2 as an electrode creases as XLi goes from 0 to 1/2. This is similar to
material is likely related to the strong intrinsic the defect calculations on Co and Mn in section 4 that
preference of Co for octahedral sites over the +3 to found tetrahedral defect energies in the layered
+4 valences as indicated by the high calculated value structure to decrease for both as Li content increases
of ∆Eoctftet in Tables 2 and 3. This strong preference from 0 to 1/2.
of Co for octahedral sites clearly suggests that Co will
At XLi ) 0.5, three of the eight TM ions (Mn, Co,
not easily migrate through an close-packed oxygen
and Ni) were found to undergo a major charge
framework at these oxidation states.
disproportionation reaction (eq 1) when moved from
octahedral to tetrahedral coordination. In contrast,
10.7. Ni at XLi ) 0 none were found to undergo charge
Ni is calculated to have the second highest ∆Eoctftet disproportionation.
at XLi ) 0.5 and the highest ∆Eoctftet at XLi ) 0. At Table 2 shows a good correlation between the
XLi ) 0.5, Ni is calculated to undergo a charge relative energetics of octahedral and tetrahedral site
disproportionation reaction (eq 1) similar to Mn and occupancy by a 3d metal ion and the projected change
Co. The projected change in LFSE and SPE is higher in LFSE for moving a 3d metal ion from octahedral
for Ni than Cr, which is consistent with the higher to tetrahedral coordination. From rows 4 (Ti) to 8 (Co)
value of ∆Eoctftet calculated for Ni than Cr in Table in Table 2 the correlation is perfect; the value
2. On the basis of these calculations, Ni is expected calculated for ∆Eoctftet increases along with the
to be the second most stable in octahedral sites at projected change in LFSE (from 0 for Ti to (14/5)∆o
an average valence of +3.5 of the 3d metals from Ti - (6/5)∆t for Co).
to Cu. The two deviations from this trend are Mn and Fe,
The high value of ∆Eoctftet calculated for Ni at XLi which have the two most negative values for ∆Eoctftet.
) 0 is consistent with the large projected change in Both of these cases probably have relatively small
LFSE (12/5∆o - 3/5∆t). However, as with Co3.5+, the net changes in LFSE and SPE, which is consistent
change in LFSE is somewhat counteracted by a with a more general association between low values
negative change in spin-pairing energy (-2∆s), which of ∆Eoctftet and low projected changes in LFSE and
helps explain why ∆Eoctftet calculated for Ni4+ is not SPE. The case of Fe illustrates that focusing exclu-
as high compared to the other 3d metals as one would sively on LFSE can be misleading when spin pairing
expect on the basis of LFSE alone. is present, especially in low-spin cations.
The calculated stability of Ni in octahedral coor- Table 3 also shows a good correlation between
dination is consistent with the good electrochemical ∆Eoctftet and the projected change in LFSE, with the
performance observed experimentally for Li(MnNi)- largest changes in LFSE being associated with the
O2 compounds. The Ni2+ oxidizes Mn to +4, which highest values of ∆Eoctftet.
as already discussed has a strong preference for The results given in Tables 2 and 3 support the
octahedral sites. When the average valence of Ni is decisive role of LFSE in the site preference and most
between +3 and +4, there is a high concentration of likely the mobility of 3d metal ions in a ccp lattice.
Li vacancies that would facilitate ion migration, For Mn, Ni, and Co the results in Table 2 are
except for the strong preference of Ni for octahedral consistent with the experimental work of Choi, Man-
sites over this valence range. Ni2+ also most likely thiram et al. who found increasing resistance of l-Lix-
has a strong preference for octahedral coordination MO2 against transformation into spinel as M is
changed from Mn to Ni to Co.49 Table 2 shows that This is probably because significant amounts of Mn3+
∆Eoctftet and the projected change in LFSE and SPE and Li vacancies do not coexist over ranges of Li
for Ni fall between those values for Mn and Co; this concentration where the spinel-like structure is
agrees with the observed intermediate stability of metastable. At low lithium concentrations the Mn are
l-LixNiO2 compared to l-LixMnO2 and l-LixCoO2. primarily in the +4 oxidation state and consequently
have relatively low mobility. At high Li content (x ≈
11. Conclusions 2 in LixMn2O4) there are insufficient Li vacancies to
allow easy Mn rearrangement. When substantial
For phase transformations involving rearrange- concentrations of Mn3+ and Li vacancies coexist in
ment of transition-metal cations over octahedral sites s-LiMn2O4, the spinel structure is thermodynamically
within a fixed ccp oxide framework, such as the stable and hence not adversely effected by the
transformation of layered LixMnO2 to spinel, the increased mobility of Mn. However, there still is the
results of this study indicate that the low-energy problem of Mn dissolution into the electrolyte through
pathway for transition-metal migration between Mn3+ charge disproportionation.
octahedral sites is through a shared nearest neighbor LFSE also appears to be a decisive factor in the
tetrahedral site (i.e., Oh f Td f Oh) rather than site preference of other 3d TM ions in a ccp oxide.
directly between octahedral sites (Oh f Oh). This As with Mn, this should impact the resistance of
suggests that the smaller the energy change is for other metastable 3d TM oxides against transforma-
transition-metal ion movement between octahedral tion. The resistance against transformation imparted
and tetrahedral coordination, the more easily the TM by ligand-field effects appears to be epitomized in the
ions should be able to rearrange in a ccp oxide when case of layered LixCoO2. Contrary to the case with
transforming from a metastable structure to a stable Mn, the change in LFSE associated with Co move-
one. As a result, the resistance against transforma- ment from octahedral to tetrahedral coordination is
tion of metastable transition-metal oxides with a ccp highly unfavorable in Li1/2CoO2. This provides im-
oxygen sublattice (e.g., LixMnO2 with R-NaFeO2 pressive stability for layered LixCoO2 with electro-
structure) will depend on the relative stability of the chemical cycling over Li concentrations where the
transition metal in octahedral coordination compared spinel structure is thermodynamically stable (e.g.,
to tetrahedral coordination. The availability of empty Li1/2CoO2).
tetrahedra without any cations occupying nearest
neighbor face-sharing octahedra is also an important The results of this investigation lead to the predic-
factor in the ability of TM ions to migrate through a tion that, in general, metastable Mn oxide structures
ccp oxide. Such tetrahedra provide a relatively low- with a ccp oxygen sublattice will rapidly transform
energy pathway by avoiding the highly repulsive to stable ccp structures if Mn3+ and/or Mn2+ coexist
interaction between face-sharing cations. with sufficient vacant interstitial sites. This is due
The energetics of Mn movement between octahe- to the lack of ligand-field stabilization for Mnoct2+ and
dral and tetrahedral sites is found to be particularly the susceptibility of Mnoct3+ to charge disproportion-
sensitive to valence. Of the 3d transition metals from ation, which results in low ligand-field stabilization.
Ti to Cu, Mn is calculated to be the second most This in turn makes tetrahedral sites relatively ac-
stable in octahedral coordination at +4 valence and cessible to Mn, which facilitates cationic rearrange-
the least stable at +3 valence. This appears to result ment.
from a large difference in ligand-field stabilization The energy difference between oxide structures
between the +4 to +3 states of charge for Mn. The with and without Mn in tetrahedral coordination can
most unfavorable change in LFSE associated with be fitted reasonably well to a simple model in which
Mn movement from octahedral to tetrahedral coor- the energy difference is broken down into two con-
dination occurs when Mn is +4 with three spin- tributions. One contribution is independent of Mn
polarized d electrons occupying a half-filled t2g shell d-orbital filling and hence Mn valence. It accounts
in octahedral coordination. In contrast, Mnoct3+ (t2g3eg1) for the change in interaction energy between a Mn7+
is calculated to undergo a charge disproportionation (i.e., d0) ionic core and the surrounding cations as a
reaction (eq 1), forming Mntet2+ (e2t23) and Mnoct4+ Mn moves from an octahedral to a tetrahedral site.
(t2g3), which allows Mn2+ movement into tetrahedral The other energy contribution is from the changing
coordination with a relatively low change in LFSE. energy of the filled d orbitals as Mn is moved from
The low-energy passage of Mn through tetrahedral octahedral to tetrahedral coordination. The latter
sites enabled by Mnoct3+ charge disproportionation energy contribution varies in a piecewise linear
appears to underlie the instability of metastable ccp fashion with Mn valence and has a peak at Mn4+,
Mn oxides such as the layered R-NaFeO2 and orthor- reflecting the ligand-field splitting of octahedrally
hombic structures with electrochemical cycling. In and tetrahedrally coordinated d orbitals.
both of these structures Mnoct3+, which can charge The energy contribution from the changing Mn d
disproportionate, and Li vacancies, which facilitate orbitals as Mn moves from octahedral to tetrahedral
Mn migration by providing empty tetrahedral sites coordination can in turn be separated into the change
without face-sharing cations, coexist over the Li of the Mn d-orbital energy barycenter and the change
concentration range where these structures are not in ligand-field stabilization energy. While the change
thermodynamically stable. in LFSE is determined by the Mn valence, the
Spinel-like LixMn2O4, on the other hand, maintains d-orbital barycenter and the energy of interactions
its structural integrity, even when it is not thermo- involving the Mn7+ ionic cores are found to be much
dynamically stable at high and low lithium content. more sensitive to cationic ordering.
Using ionic valences found through a large series tion to spinel has been achieved by Doeff et al. using
of calculations on substituted Mn oxides, a qualitative such a structure.74,75
ionization scale between Mn and other 3d metals has
been constructed. This scale allows one to predict the 12. Acknowledgments
valences for Mn (in octahedral and/or tetrahedral
coordination) coexisting with another 3d TM cation This research was supported in part by the MRSEC
(in octahedral coordination) in a ccp oxide. This could Program of the National Science Foundation under
be useful for designing TM oxide materials with award number DMR-02-13282 and by the Assistant
improved kinetic stability over the range of Li Secretary for Energy Efficiency and Renewable En-
concentrations covered by electrochemical cycling. ergy, Office of Freedom CAR and Vehicle Technolo-
The findings of this study point to a number of gies, of the U.S. Department of Energy under Con-
different strategies for producing lithium manganese tract No. DE-AC03-76SF00098, Subcontract No.
oxide structures, other than spinel, that resist trans- 6517748 with the Lawrence Berkeley National Labo-
formation with electrochemical cycling. If a meta- ratory. Methodological developments that made this
stable ccp oxide structure such as l-LixMnO2 or work possible have been supported by the Depart-
o-LixMnO2 (XLi < 1) is used, high concentrations of ment of Energy under contract number DE-FG02-
Mnoct3+ and Li vacancies should be prevented from 96ER45571. NPACI is acknowledged for providing
coexisting over regions of metastability. This can be substantial computing resources for this work. G.C.
accomplished by chemically substituting for Mn with acknowledges a faculty development chair from R.
low fixed valence and/or electronegative multivalent P. Simmons.
cations that can oxidize the Mn eg orbital. There are
many different examples of this approach that have 13. References
had some success experimentally.7,18,53-55 Another (1) Mizushima, K.; Jones, P. C.; Wiseman, P. J. Solid State Ionics
way of oxidizing the Mn eg orbital is to introduce 1981, 3-4 (Aug), 171.
vacancies into the Mn sublattice.66 (2) Thackeray, M. Prog. Solid State Chem. 1997, 25, 1.
(3) Winter, M.; Besenhard, J.; Spahr, M.; Novak, P. Adv. Mater.
Given that Mn4+ is the most stable valence in 1998, 10, 725.
octahedral coordination, it might be desirable for Mn (4) Bruce, P. Chem. Commun. 1997, 1817.
(5) Armstrong, A. R.; Bruce, P. G. Nature 1996, 381, 499.
to be electrochemically cycled between elevated oxi- (6) Capitaine, F.; Gravereau, P.; Delmas, C. Solid State Ionics
dation states centered closer to +4 rather than over Diffusion React. 1996, 89, 197.
the +3/+4 redox range characteristic of LixMnO2 (7) Armstrong, A.; Robertson, A.; Bruce, P. Electrochem. Acta 1999,
45, 285.
structures. Unfortunately at this time there probably (8) Bruce, P.; Armstrong, A.; Gitzendanner, R. J. Mater. Chem.
are not any electrolytes that can withstand the 1999, 193.
oxidative strength of Mn at valences higher than (9) Shao-Horn, Y.; Hackney, S. A.; Armstrong, A. R.; Bruce, P. G.;
Gitzendanner, R.; Johnson, C. S.; Thackeray, M. M. J. Electro-
+4.67 An additional problem with Mn charged above chem. Soc. 1999, 146, 2404.
+4 in ccp oxides is the possibility of decomposition (10) Armstrong, A.; Robertson, A.; Gitzendanner, R.; Bruce, P. J.
Solid State Chem. 1999, 145, 549.
reactions producing O2.58,68 (11) Broussely, M.; Biensan, P.; Simon, B. Electrochem. Acta 1999,
Another strategy is to use a structure with an 45, 3.
oxygen sublattice that is different from that of spinel (12) Van der Ven, A.; Ceder, G. Phys. Rev. B 1999, 59 (2), 742.
(13) Ceder, G.; Van der Ven, A. Electrochem. Acta 1999, 45 (1-2),
(i.e., non-ccp). For such a structure to transform into 13150.
spinel the oxygen needs to rearrange, which should (14) Blyr, A.; Sigala, C.; Amatucci, G.; Guyomard, D.; Chabre, Y. J.
make the transformation much more difficult. (15) Chiang, Y. M.; Sadoway, D. R.; Jang, Y. I.; Huang, B.; Wang, H.
One way this has been accomplished experimen- Electrochem. Solid State Lett. 2 1999, 3, 107.
tally is by using a close-packed oxide structure that (16) Vitins, G.; West, K. J. Electochem. Soc. 1997, 144, 2587.
(17) Chiang, Y.-M.; Wang, H.; Jang, Y. Chem. Mater. 2001, 13, 53.
is not cubic-close packed.69,70 Such structures may be (18) Wang, H.; Jang, Y.; Chiang, Y.-M. Electrochem. Solid State Lett.
more resistant against transformation to spinel, but 1999, 2 (10), 490.
there still is a network of octahedral and tetrahedral (19) Hunter, J. C. J. Solid State Chem. 1981, 39, 142
(20) David, W. I. F.; Thackeray, M. M.; Bruce, P. G.; Goodenough, J.
sites that could possibly allow undesirable Mn move- B. Mater. Res. Bull. 1984, 19, 99.
ment during electrochemical cycling. The phospho- (21) Wang, H.; Jang, Y.-I.; Huang, B.; Sadoway, D. R.; Chiang, Y.-
M. J. Electrochem. Soc. 1999, 146, 473.
olivine structure is an example of a non-ccp oxide that (22) Gabrisch, H.; Yazami, R.; Fultz, B. J. Power Sources 2003, 119,
has shown good reversible capacity with iron (LiFe- 674.
PO4) however not with Mn (although Mn combined (23) Van der Ven, A.; Ceder, G. Electrochem. Solid State Lett. 2000,
3 (7).
with Fe is reported to perform well).71 Oxides con- (24) Reed, J. Ab Initio Study of Cathode Materials for Lithium
taining polyanions such as PO4-3 offer a rich diversity Batteries. Ph.D. Thesis, M.I.T., 2003; p 65.
of structures that might be suitable for use in a (25) Parr, R. G.; Yang, W. Density-Functional Theory of Atoms and
Molecules; Oxford University Press: New York, 1989; pp 51-
positive electrode.72 52.
Another option is to use a non-close-packed struc- (26) Parr, R. G.; Yang, W. Density-Functional Theory of Atoms and
Molecules; Oxford University Press: New York, 1989; pp 142-
ture. A more open structure can constrain Mn by 145.
eliminating energetically favorable sites it can mi- (27) Ohno, K.; Esfarjani, K.; Kawazoe, Y. Computational Materials
grate through (e.g., tetrahedral sites). A layered Science; Springer-Verlag: Berlin, Heidelberg, 1999; pp 21-25.
(28) Mishra, S. K.; Ceder, G. Phys. Rev. B 1999, 59, 6120.
structure can be made less close packed by pillaring (29) Van der Ven, A.; Aydinol, M. K.; Ceder, G. J. Electrochem. Soc.
open the Li/vacancy layer with a large cation like K+ 1998, 145 (6), 2149.
or with a cluster of atoms.73 The tunneled Mn oxides (30) Kresse, G.; Furthmuller, J. Phys. Rev. B 1996, 54 (11), 169.
(31) Parr, R. G.; Yang, W. Density-Functional Theory of Atoms and
are another example of non-close-packed structures. Molecules; Oxford University Press: New York, 1989; pp 153
Good cyclability and resistance against transforma- and 154.
(32) Ohno, K.; Esfarjani, K.; Kawazoe, Y.; Computational Materials (58) Lu, Z.; Dahn, J. R. J. Electrochem. Soc. 2002, 149 (7) A815.
Science; Springer-Verlag: Berlin, Heidelberg, 1999; pp 106 and (59) Park, H.-S; Hwang, S.-J.; Choy, J.-H. J. Phys. Chem. B 2001,
107. 105, 4860.
(33) Vanderbilt, D. Phys. Rev. B 1990, 41, 7892. (60) Burns, R. G. Mineralogical Applications of Crystal Field Theory;
(34) Mizushima, K.; Jones, P. C.; Wiseman, P. J.; Goodenough, J. B. Cambridge University Press: New York, 1970, 1993; p 436.
Mater. Res. Bull. 1980, 15, 783. (61) Chang, S. H.; Kang, S. G.; Song, S. W.; Yoon, J. B.; Choy, J. H.
(35) Orman, H. J.; Wiseman, P. J. Acta Crystallogr., Sect. C 1984, Solid State Ionics 1996, 86-88, 171.
139, 12.
(36) Ceder, G.; Van der Ven, A. Electrochim. Acta 1999, 45, 131. (62) Kang, K.; Carlier, D.; Reed, J.; Arroyo, E.; Ceder, G.; Croguennec,
(37) Wolverton, C.; Zunger, A. J. Electrochem. Soc. 1998, 145, 2424. L.; Delmas, C. Chem. Mater. 2003, 15, 4503.
(38) Gummow, R.; Liles, D. C.; Thackeray, M. M. Mater. Res. Bull. (63) Maazaz, A.; Delmas, C.; Hagenmuller, P. J. Inclusion Phenom.
1993, 28, 1249. 1983, 1, 45.
(39) Reed, J.; Ceder, G.; Van der Ven, A. Electrochem. Solid State (64) Bethe, H.; Jackiw, R. Intermediate Quantum Mechanics; Addi-
Lett. 2001, 4 (6), A78. son-Wesley Longman, Inc.: New York, 1986, 1997; pp 80 and
(40) Reed, J.; Ceder, G. Electrochem. Solid State Lett. 2002, 5 (7), 81.
A145. (65) Kanno, R.; Shirane, T.; Inaba, Y.; Kawamoto, Y. J. Power Sources
(41) Borg, R. J.; Dienes, G. J. The Physical Chemistry of Solids; 1997, 68 (1), 145.
Academic Press: New York, 1992; pp 148-152. (66) Raekelboom, E. A.; Hector, A. L.; Owen, J.; Vitins, G.; Weller,
(42) Burns, R. G. Mineralogical Applications of Crystal Field Theory; M. T. Chem. Mater. 2001, 13 (12), 4618.
Cambridge University Press: New York, 1970, 1993; p 464. (67) Robertson, A. D.; Bruce, P. G. Chem. Commun. 2002, 23, 2790.
(43) Goodenough, J. B.; Loeb, A. L. Phys. Rev. 1955, l98, 391.
(68) Lu, Z.; Beaulieu, L. Y.; Donaberger, R. A.; Thomas, C. L.; Dahn,
(44) Figgis, B. N.; Hitchman, M. A. Mineralogical Applications of
J. R. J. Electrochem. Soc. 2002, 149 (7) A815.
Crystal Field Theory; Wiley-VCH: New York, 2000; p 116.
(45) Burns, R. G. Mineralogical Applications of Crystal Field Theory; (69) Paulsen, J. M.; Larcher, D.; Dahn, J. R. J. Electrochem. Soc.
Cambridge University Press: New York, 1970, 1993; p 17. 2000, 147 (8), 2862.
(46) Burns, R. G. Mineralogical Applications of Crystal Field Theory; (70) Eriksson, T. A.; Lee, Y. J.; Hollingsworth, J.; Reimer, J. A.;
Cambridge University Press: New York, 1970, 1993; p 22. Cairns, E. J.; Zhang, X.-F.; Doeff, M. M. Chem. Mater. 2003, 15,
(47) Burns, R. G. Mineralogical Applications of Crystal Field Theory; 4456.
Cambridge University Press: New York, 1970, 1993; p 218. (71) Padhi, A. K.; Nanjundaswamy, K. S.; Goodenough, J. B. J.
(48) Burns, R. G. Mineralogical Applications of Crystal Field Theory; Electrochem. Soc. 1997, 144 (4), 1188.
Cambridge University Press: New York, 1970, 1993; p 231. (72) Patoux, S. Matériaux d’ électrode positive à charpente polya-
(49) Choi, S.; Manthiram, S. J. Electrochem. Soc. 2002, 149 (9), nionique pour batteries au lithium: Approches cristallochimiques
A1157. et électrochimiques. Thése de Doctorat, L’Universite de Picardie
(50) Figgis, B. N.; Hitchman, M. A. Mineralogical Applications of Jules Verne, 2003.
Crystal Field Theory; Wiley-VCH: New York, 2000; pp 166- (73) Zhang, F.; Ngala, K.; Whittingham, M. S. Electrochem. Commun.
177. 2000, 2(6) 445.
(51) Burns, R. G. Mineralogical Applications of Crystal Field Theory; (74) Doeff, M. M.; Richardson, T. J.; Kepley, L. J. Electrochem. Soc.
Cambridge University Press: New York, 1970, 1993; p 433. 1996, 143(8), 2507.
(52) Burns, R. G. Mineralogical Applications of Crystal Field Theory;
Cambridge University Press: New York, 1970, 1993; p 18. (75) Doeff, M. M.; Anapolsky, A.; Edman, L.; Richardson, T. J.; De
(53) Davidson, I. J.; McMillan, R. S.; Slegr, J.; Luan, B.; Kargina, I.; Jonghe, L. C. J. Electrochem. Soc. 2001, 148 (3), A230.
Murray, J. J.; Swainson, I. P. J. Power Sources 1999, 81-82, (76) Reed, J. Ab Initio Study of Cathode Materials for Lithium
406. Batteries. Ph.D. Thesis, M.I.T., 2003; pp 78-85.
(54) Rossen; Jones, C. D. W.; Dahn, J. R. Solid State Ionics 1992, (77) Reed, J. Ab Initio Study of Lithium Manganese Oxides. M.S.
57, 311. Thesis, M.I.T., 2001; pp 49-57.
(55) Ammundsen, B.; Paulsen, J. Adv. Mater. 2001, 13, 943. (78) Reed, J. Ab Initio Study of Cathode Materials for Lithium
(56) Ammundsen, B.; Desilvestro, J.; Groutso, T.; Hassell, D.; Metson, Batteries. Ph.D. Thesis, M.I.T., 2003; pp 152-161.
J. B.; Regan, E.; Steiner, R.; Pickering, P. J. J. Electrochem. Soc. (79) Reed, J. Ab Initio Study of Cathode Materials for Lithium
2000, 147, 4078. Batteries. Ph.D. Thesis, M.I.T., 2003; pp 235-236.
(57) de Groot, F. M. F.; Grioni, M.; Fuggle, J. C.; Ghigsen, J.;
Sawatzky, G. A.; Petersen, H. Phys. Rev. B 1989, 40, 5715. CR020733X
Chem. Rev. 2004, 104, 4587−4612 4587
Alternative Polymer Systems for Proton Exchange Membranes (PEMs)

Michael A. Hickner,† Hossein Ghassemi,‡ Yu Seung Kim,§ Brian R. Einsla,| and James E. McGrath*,|
Sandia National Laboratory, Albuquerque, New Mexico 87185, Department of Chemical Engineering, Case Western Reserve University,
Cleveland, Ohio, Los Alamos National Laboratory, Los Alamos, New Mexico 87545, and Department of Chemistry, Macromolecules and
Interfaces Institute, Virginia Polytechnic Institute and State University, Blacksburg, Virginia 24061
Received February 27, 2004
Contents restrial applications for fuel cells, although its major

focus was in chloroalkali processes.
2. Nafion and Other Poly(perfluorosulfonic acid) 4590 PEM fuel cells are being developed for three main
Membranes applications: automotive, stationary, and portable
3. PEMs Containing Styrene and Its Derivatives 4591 power. Each of these applications has its unique
operating conditions and material requirements.
4. Poly(arylene ether)s 4593
Common themes critical to all high performance
4.1. Postsulfonation of Existing Polymers 4593 proton exchange membranes include (1) high protonic
4.2. Direct Copolymerization of Sulfonated 4595 conductivity, (2) low electronic conductivity, (3) low
Monomers To Afford Random (Statistical) permeability to fuel and oxidant, (4) low water
Copolymers
transport through diffusion and electro-osmosis, (5)
5. PEMs Based on Poly(imide)s 4597 oxidative and hydrolytic stability, (6) good mechan-
6. High Performance Polymeric Backbones for Use 4600 ical properties in both the dry and hydrated states,
in PEMs (7) cost, and (8) capability for fabrication into mem-
7. Polyphosphazene PEMs 4603 brane electrode assemblies (MEAs). Nearly all exist-
8. Other Proton Conducting MoietiessAlternatives 4605 ing membrane materials for PEM fuel cells rely on
to Sulfonation absorbed water and its interaction with acid groups
9. Important Considerations in the Design of New 4607 to produce protonic conductivity. Due to the large
PEMs fraction of absorbed water in the membrane, both
10. Conclusions 4608 mechanical properties and water transport become
11. Acknowledgements 4609 key issues. Devising systems that can conduct pro-
12. References 4609 tons with little or no water is perhaps the greatest
challenge for new membrane materials. Specifically,
for automotive applications the U.S. Department of
1. Introduction Energy has currently established a guideline of 120
°C and 50% relative humidity as target operating
Fuel cells have the potential to become an impor- conditions, and a goal of 0.1 S/cm for the protonic
tant energy conversion technology. Research efforts conductivity of the membrane.
directed toward the widespread commercialization of
fuel cells have accelerated in light of ongoing efforts New membranes that have significantly reduced
to develop a hydrogen-based energy economy to methanol permeability and water transport (through
reduce dependence on foreign oil and decrease pol- diffusion and electro-osmotic drag) are required for
lution. Proton exchange membrane (also termed portable power oriented direct methanol fuel cells
“polymer electrolyte membrane”) (PEM) fuel cells (DMFCs), where a liquid methanol fuel highly diluted
employing a solid polymer electrolyte to separate the in water is used at generally <90 °C as the source of
fuel from the oxidant were first deployed in the protons. Unreacted methanol at the anode can diffuse
Gemini space program in the early 1960s using cells through the membrane and react at the cathode,
that were extremely expensive and had short life- lowering the voltage efficiency of the cell and reduc-
times due to the oxidative degradation of their ing the system’s fuel efficiency. The methanol is
sulfonated polystyrene-divinylbenzene copolymer usually delivered to the anode as a dilute, for
membranes. These cells were considered too costly example, 1 M (or less), solution (3.2 wt %), and
and short-lived for real-world applications. The com- relatively thick Nafion 117 (1100 EW, 7 mil ∼ 178
mercialization of Nafion by DuPont in the late 1960s µm thick) is used to reduce methanol crossover. The
helped to demonstrate the potential interest in ter- dilute methanol feed increases the system’s complex-
ity and reduces the energy density of the fuel, while
* To whom correspondence should be addressed. E-mail:
the thick Nafion membrane increases the resistive
jmcgrath@vt.edu. losses of the cell, especially when compared to the
† Sandia National Laboratory.
‡ Case Western Reserve University.
thinner membranes that are used in hydrogen/air
§ Los Alamos National Laboratory. systems. The presence of excessive amounts of water
| Virginia Polytechnic Institute and State University. at the cathode through diffusion and electro-osmosis
10.1021/cr020711a CCC: $48.50 © 2004 American Chemical Society
4588 Chemical Reviews, 2004, Vol. 104, No. 10 Hickner et al.
Michael Hickner received his B.S. in Chemical Engineering from Michigan Yu Seung Kim received his Ph.D. at the Korea Advanced Institute of
Tech in 1999 and his Ph.D. in Chemical Engineering in 2003 under the Science and Technology under Prof. Sung Chul Kim (1999). Subsequently,
direction of James McGrath. Michael’s research in Dr. McGrath’s lab he joined Dr. James. E. McGrath’s research group at the Virginia
focused on the transport properties of proton exchange membranes and Polytechnic Institute and State University (1999−2003), where he began
their structure−property relationships. He has spent time at Los Alamos studying polymer electrolyte membranes. His research was focused on
National Laboratory studying novel membranes in direct methanol fuel the structure−morphology−property relationships of fuel cell membranes.
cells and is currently a postdoc at Sandia National Laboratories in He is currently working at the Los Alamos National Laboratory; his current
Albuquerque, NM. research includes developing new membranes and electrodes for H2/air
and direct methanol fuel cells.
Hossein Ghassemi is currently a Senior Research Associate in the

Department of Chemical Engineering at Case Western Reserve University Brian Einsla was born in Wilkes-Barre, PA, in 1978. He earned his Bachelor
and is involved in several projects related to fuel cell technology. He of Science degree in 2000 in Chemistry at Virginia Polytechnic Institute
received his B.Sc. in Chemistry from Tehran University, in 1984. He then and State University. He is currently a fourth-year Ph.D. candidate in the
pursued his education toward an M.Sc. in Organic Chemistry at the Institute Macromolecular Science and Engineering program at VPI&SU. His
of Chemistry at Mazandaran, Iran. In May 1990 Dr. Ghassemi moved to research interests have centered around sulfonated heterocyclic copoly-
Montreal, Canada, to begin his Ph.D. program at the Department of mers for fuel cell applications, including polyimides, polybenzoxazoles,
Chemistry at McGill University. He received his Ph.D. under the supervision and polybenzimidazoles.
of Prof. Allan S. Hay, in 1994. He then started his professional career as
a Postdoctoral Associate with professor James E. McGrath at the NSF
Science and Technology Center for High Performance Polymeric critical. Current work in the area of fabricating MEAs
Adhesives and Composites at Virginia Polytechnic Institute and State from novel polymeric membranes has focused on the
University. He was involved in several projects sponsored by McDonnell electrode-membrane interface and the problems of
Douglas/ARPA. Among his duties were to perform all aspects of research
and development in the area of assignment, adapt and modify standard having dissimilar ion conducting copolymers in the
techniques and procedures, and apply nontraditional approaches and membrane and as components of the electrode.1
higher levels of problem solving for research. During the past 15 years, Novel membranes must also be adaptable and have
Dr. Ghassemi has authored and coauthored more than 20 scientific papers the necessary physical strength and ductility in the
and patents. dry and wet states to survive the stress of electrode
attachment. Ion conducting copolymers that are
compatible for use in the catalyst layer, in concert
results in water management problems both in the with novel polymer membranes, are also an emerging
catalyst layer and for the system. area of research.
New membrane materials for PEM fuel cells must Advances in fuel cell technology over the last four
be fabricated into a well-bonded, robust membrane decades have come primarily from improved electro-
electrode assembly (MEA) as depicted in Figure 1. catalysts, membrane electrode assembly fabrication
In addition to the material requirements of the strategies, and cell/stack/system engineering. Apart
proton exchange membrane itself as outlined above, from Nafion, new ion conducting polymeric materials
the ease of membrane electrode assembly fabrication have played only a minor role in significantly in-
and the resulting properties of the MEA are also creasing cell performance. However, new materials
Alternative Polymer Systems for Proton Exchange Membranes Chemical Reviews, 2004, Vol. 104, No. 10 4589
polymers based on styrenic, vinylic, or acrylic mono-

mers will then be reviewed. Next, the large body of
recent literature centered on polymeric materials
based on step or condensation polymers (poly(ether
ether ketone)s, poly(sulfone)s, poly(imide)s, and oth-
ers) and various strategies for incorporating ionic
groups into these materials will be evaluated. Finally,
polymeric materials with other acid groups aside
from sulfonic acid as the ion conducting moiety will
be reviewed and various methodologies for advanced
membranes will be presented. This review will high-
light the synthetic aspects of ion conducting polymers
and the polymer chemical structure considerations
for producing useful PEMs. Extensive reviews of
polymer-based proton conductors have recently been
James E. McGrath, a member of the prestigious National Academy of published.2,3,4 Only a brief review of important or-
Engineering, is currently a University Distinguished Professor of Chemistry
at Virginia Tech, as well as the Associate Director of the Macromolecules
ganic-inorganic composite membranes will be in-
and Interfaces Institute. He received his B.S. in Chemistry from St. cluded,3b even though it is predicted that this strategy
Bernadine of Siena College (1956), his M.S. in Chemistry from the will be central to future higher temperature systems.
University of Akron (1964), and his Ph.D. in Polymer Science from the Most of the synthetic methods currently utilized
University of Akron (1967). After several years in industry at Goodyear to form ion containing materials result in random or
Tire and Rubber Company and Union Carbide, Dr. McGrath joined the
faculty of Virginia Tech in 1975. He was named Full Professor 5 years statistical placement of sulfonic acid units along the
later and in 1987 was appointed the Director of the Materials Institute, backbone copolymer chain. It is postulated that the
which later became the National Science Foundation Science & Technol- distribution of sulfonic acid groups along the chain,
ogy Center for High Performance Polymer Adhesives and Composites as well as the acid strength and the connecting
(from 1989 to 2000). Since 1979, McGrath has also codirected the Polymer moiety to the polymer backbone, will have a consid-
Materials and Interface Laboratory (PMIL). Dr. McGrath’s career has erable effect on the resulting morphology and mem-
included a number of distinguished awards, such as the Herman F. Mark
Award (ACS Division of Polymer Chemistry, 1996), “Virginia Scientist of brane properties. Some block or segmented copolymer
the Year” by the Virginia Museum of Science (1997), the ACS Division of systems have been synthesized, but the subtle dif-
Polymer Chemistry 2002 Applied Polymer Science Award, and, most ferences between random and block systems and the
recently, the P. J. Flory Education Award from the ACS Division of Polymer advantages/disadvantages of each are not yet well
Chemistry (2004), which he received along with G. L. Wilkes and T. C. established in the literature for PEMs. This review
Ward. Dr. McGrath’s primary research interests include high performance will also highlight novel strategies for the controlled
matrix polymers and structural adhesives, high temperature polymer
dielectrics for computers, fire-resistant polymers and composites, and placement of ionic groups on a polymer or statistical
synthesis and characterization of new copolymers for proton exchange copolymer backbone.
membranes (fuel cells). He holds 35 patents on a wide variety topics Some definitions are needed to establish a common
related to polymer science, and he has coauthored well over 400 framework for comparing different polymeric sys-
publications.
tems. The foremost of these is protonic conductivity.
Scientists at Los Alamos National Laboratory (LANL)
have devised a facile method for determining the
conductivity of proton exchange membranes using
electrochemical impedance spectroscopy and a simple
cell that allows equilibration in a variety of environ-
ments.5 This method measures protonic conductivity
in the plane of the membrane as opposed to through
the plane (as in a fuel cell), and thus it works well as
an initial screening test. Through-plane conductivity
measurements6 are often more difficult experimen-
tally than in-plane measurements because the mea-
sured membrane resistances are small in this geom-
etry and interfacial resistances may play a more
Figure 1. Basic membrane electrode assembly. significant role.
Water uptake is also important in determining the
are now required to afford successful technology ultimate performance of proton exchange membrane
commercialization. materials. In essentially all current polymeric ma-
This review will outline the materials requirements terials, water is needed as the mobile phase to
for advanced alternative proton exchange membranes facilitate proton conductivity. However, absorbed
for fuel cells, assess recent progress in this area, and water also affects the mechanical properties of the
provide directions for the development of next-genera- membrane by acting as a plasticizer, lowering the Tg
tion materials. The focus will be on the synthesis of and modulus of the membrane. Careful control of
polymeric materials that have attached ion conduct- water uptake is critical for reducing adverse effects
ing groups. State-of-the-art Nafion and its com- of swelling and degradation of the mechanical prop-
mercially available perfluorosulfonic acid relatives erties of the membrane in humid environments, as
will initially be discussed. Other chain-growth co- well as inducing stresses between the membrane and
the electrodes. Water uptake is usually reported as

a mass fraction, mass percent, or λ value, where λ
equals the number of water molecules absorbed per
acid site.
Both conductivity and water uptake rely heavily
on the concentration of ion conducting units (most Figure 2. Chemical structure of Nafion. x and y represent
commonly sulfonic acid) in the polymer membrane. molar compositions and do not imply a sequence length.
The ion content is characterized by the molar equiva-
lents of ion conductor per mass of dry membrane and
is expressed as equivalent weight (EW) with units weight) with a comonomer which ultimately has
of grams of polymer per equivalent or ion exchange pendant side chains of perfluorinated vinyl ethers
capacity (IEC) with units of milliequivalents per terminated by perfluorosulfonic acid groups. The
gram (mequiv/g or mmol/g) of polymer (EW ) 1000/ reported chemical structure of Nafion for PEM mem-
IEC). Varying the ion content of the membrane can branes is shown in Figure 2.
control both its water uptake and conductivity. While In theory, ion content can be varied by changing
it is desirable to maximize the conductivity of the the ratio of the two components (x and y in Figure
membrane by increasing its ion content (decreasing 2). Nafion has been commercially available in 900,
equivalent weight), other physical properties must 1100, 1200, and other equivalent weights. However,
be considered. Too many ionic groups will cause the Nafion 1100 EW in thicknesses of 2, 5, 7, and 10 mil
membrane to swell excessively with water, which (1 mil equals 25.4 µm) (Nafion 112, 115, 117, and
compromises mechanical integrity and durability. 1110) seems to be the only grades of Nafion that are
Meaningful standardized methods for character- currently widely available. This equivalent weight
izing proton exchange membranes for fuel cells are provides high protonic conductivity and moderate
becoming more established within the fuel cell com- swelling in water, which seems to suit most current
munity and are helping to enhance the identification applications and research efforts. Modest retention
of promising candidate materials. At the most basic of a semicrystalline morphology at this composition
level, the ion exchange capacity, water uptake, and is no doubt important for mechanical strength. The
protonic conductivity of the membrane under specific thinner membranes are generally applied to hydrogen/
environmental conditions should be measured in air applications to minimize Ohmic losses, while
comparison to the standard Nafion materials and thicker membranes are employed for direct methanol
other systems. Standard important polymer science fuel cells (DMFCs) to reduce methanol crossover.
and engineering parameters such as molecular weight, Unsaturated perfluoroalkyl sulfonyl fluoride and
detailed chemical compositions, morphology, topol- their derivatives are believed to be the starting
ogy, and mechanical behavior are no doubt critical comonomers for preparing perfluorosulfonic mem-
but have only rarely been addressed. branes. Nafion is prepared via the copolymerization
of variable amounts of the unsaturated perfluoroalkyl
2. Nafion and Other Poly(perfluorosulfonic acid) sulfonyl fluoride with tetrafluoroethylene.11,12 Unfor-
tunately, there have been no detailed literature
Membranes reports of Nafion’s synthesis and processing, but it
The current state-of-the-art proton exchange mem- is generally thought that the copolymer is then
brane is Nafion, a DuPont product that was devel- extruded in the melt processable sulfonyl fluoride
oped in the late 1960s primarily as a permselective precursor to form a membrane, which is later con-
separator in chlor-alkali electrolyzers.7,8 Nafion’s verted from the sulfonyl fluoride form by base hy-
poly(perfluorosulfonic acid) structure imparts excep- drolysis to the salt or sulfonic acid functionality. It
tional oxidative and chemical stability, which is also seems unlikely that the sulfonyl fluoride containing
important in fuel cell applications. precursor unit in the copolymer would self-propagate
Nearly all of the commercially available mem- under free radical conditions. Thus, the length of the
branes are based on Nafion. Nafion also has the comonomer sequence (y) is likely only one unit. Total
largest body of literature devoted to its study because molecular weight, though obviously important, has
of its demonstrated industrial importance and avail- not been reported.
ability. Nafion composite systems also have already Like many other fluoropolymers, Nafion is quite
become significant in both industrial and academic resistant to chemical attack, but the presence of its
research. In composite structures, Nafion can be strong perfluorosulfonic acid groups imparts many
impregnated into an inert Teflon-like matrix (i.e. W. of its desirable properties as a proton exchange
L. Gore membranes9), or inorganic additives can be membrane. Fine dispersions (sometimes incorrectly
added to a supporting Nafion matrix for improved called solutions) can be generated with alcohol/water
physical or electrochemical properties (i.e. Ion- treatments.13 Such dispersions are often critical for
omem10). Some critical aspects of Nafion’s molecular the generation of the catalyst electrode structure and
structure and physical properties will be briefly the MEAs. Films prepared by simply drying these
highlighted to provide a baseline for comparison with dispersions are often called “recast” Nafion, and it is
the other alternative materials discussed in this often not realized that its morphology and physical
review. behavior are much different from those of the ex-
Nafion is a free radical initiated copolymer of a truded, more crystalline form.
crystallizable hydrophobic tetrafluoroethylene (TFE) Other perfluorosulfonate cation exchange mem-
backbone sequence (∼87 mol % at 1100 equivalent branes with similar structures have also been devel-
oped by the Asahi Chemical Company (Aciplex) and

the Asahi Glass Company (Flemion).14 The Dow
Chemical Company also developed a material with
a shorter side chain than those of Nafion and the
other perfluorosulfonates, which is no longer avail-
able.15 The length of the perfluorosulfonic acid side
Figure 4. Chemical structure of sulfonated SEBS block
chain and the values for the equivalent weight may copolymer.21
be varied to some extent.
All of these polyperfluorosulfonic acid membranes perior performance to perfluorinated membranes,
are expensive and suffer from the same shortcomings such as Nafion 117, at current densities greater than
as Nafion, namely low conductivity at low water 0.6 A cm-2.17 The backbone fluorination was no doubt
contents, relatively low mechanical strength at higher intended to mitigate hydroperoxide formation, which
temperature, and moderate glass transition temper- causes short lifetimes for nonfluorinated partially
atures. aliphatic analogues.
There have been few synthetic reports employing
3. PEMs Containing Styrene and Its Derivatives these monomers beyond the Ballard work, most likely
as a result of presumed high cost and monomer
As discussed above, the most commonly known and availability. However, the performance and stability
studied PEMs are based on nonaromatic perfluori- demonstrated by these materials in fuel cells may
nated hydrocarbons such as Nafion, Aciplex, Flemion, spur further developments in this area. The above-
and what are termed the Dow membranes. However, reported copolymers are believed to be random
their chemical synthesis is challenging due to the systems both in the chemical composition of the
safety concerns of tetrafluoroethylene and the cost/ copolymer backbone and with regard to sulfonic acid
availability of the perfluoroether comonomers. These attachment. Novel methods have been developed for
issues have relegated detailed synthetic research on the controlled polymerization of styrene-based mono-
polyperfluorosulfonic acid materials to the industrial mers to form block copolymers. If one could create
sector or to a few specialized academic labs. block systems with trifluorostyrene monomers, new
One alternative to the tetrafluoroethylene-based morphologies and PEM properties with adequate
backbones of the previously discussed materials is stability in fuel cell systems might be possible, but
the use of styrene and particularly its fluorinated the mechanical behavior would need to be demon-
derivatives to form PEMs. As extensively reported strated.
in the literature, styrenic monomers are widely Dais Analytic’s PEMs (and related materials) are
available and easy to modify, and their polymers are based on well-known commercial block copolymers
easily synthesized via conventional free radical and of the styrene-ethylene/butylene-styrene family,
other polymerization techniques. Kraton G1650 being one such system. In one method
Presently, two commercial (or semicommercial) of forming sulfonated PEMs, the unsulfonated poly-
PEMs are based on styrene or styrene-like monomer is dissolved in a dichloroethane/cyclohexane
mers: BAM from Ballard, and Dais Analytic’s sul- solvent mixture. The sulfur trioxide/triethyl phos-
fonated styrene-ethylene-butylene-styrene (SEBS) phate sulfonating complex in solution is then added
membrane. Ballard Advanced Materials Corporation and allowed to react at temperatures between -5 and
introduced a styrenic membrane based on a novel 0 °C.18 The result is a PEM that may be solvent cast
family of sulfonated copolymers incorporating R,β,β- (from lower alcohols such as n-propanol) to afford an
trifluorostyrene and substituted R,β,β-trifluoro- elastomeric hydrogel with conductivities of 0.07-0.1
styrene comonomers. These are registered as BAM S/cm when fully hydrated.19,20,21 The chemical struc-
membranes, and their general formula is given in ture of this type of copolymer is shown in Figure 4.
Figure 3. The sulfonated styrene moieties are isolated in
blocks, as directed by the anionic polymerization and
hydrogenated starting material, but there is not
much information on the extent of sulfonation of the
styrenic moieties in the blocks. For the process
described above, the unsulfonated block copolymer
could have a number average molecular weight of
about 50 000 g/mol with a styrene content of 20-35
Figure 3. Chemical structure of BAM PEMs.16 wt % of the triblock copolymer. TEM suggests that
the cast films possess cylindrical sulfonated poly-
The unsulfonated random copolymers are report- styrene domains. The films swell in water, with the
edly synthesized at 50 °C over a period of 48 h using extent of swelling being dependent on the sulfonation
emulsion polymerization with dodecylamine hydro- level.22
chloride surfactant in water as the reaction system Dais membranes are reported to be much less
and potassium persulfate as the initiator.16 The expensive to produce than Nafion; they are also
copolymer is then dissolved in an appropriate solvent reported to exhibit a rich array of microphase-
such as dichloroethane or chloroform and sulfonated separated morphologies because of the ability to
using reagents such as chlorosulfonic acid or a sulfur tailor the block length and composition of the unsul-
trioxide complex. It has been reported that this fonated starting polymer.23 The main drawback of
generation of BAM membranes exhibited some su- employing hydrocarbon-based materials is their much
Figure 5. Chemical structure of partially sulfonated

styrene-ethylene interpolymer.21
poorer oxidative stability compared to perfluorinated

or partially perfluorinated membranes due to their
partially aliphatic character.24 For this reason, Dais
membranes are aimed at portable fuel cell power
sources of 1 kW or less, for which operating temper-
atures are less than 60 °C.
As an extension of the previous work, copolymers
based on partially sulfonated ethylene-styrene
pseudorandom “interpolymers” have also been em-
ployed instead of the block copolymers (Figure 5).21,25
Due to the unique nature of the polymerization
catalyst, styrene residues are separated by at least
one ethylene residue and the acid groups are distrib-
uted randomly along the chain. This material pro-
vides an economical and unique counterpoint to the
sulfonated SEBS PEMs, where the sulfonic acid
groups are bunched together in the styrene blocks.
Controlling the styrene content in each material
provides a route to control the level of sulfonation
and resultant ion exchange capacity of the PEM.
Synthetic methods have been developed to incor-
porate styrene as a graft on to a polymer backbone.
Graft polymers, in which ion containing polymer Figure 6. Synthetic scheme of polystyrene-graft-poly-
grafts are attached to a hydrophobic backbone, could styrenesulfonic acid copolymers.26,27
be suitable structures for studying structure-prop-
erty relationships in ion conducting membranes, if and variable ion content were obtained. For compari-
the length of the graft and the number density of son, random copolymers of SSNa and styrene (PS-r-
graft chains can be controlled. In principle, the length SSNa) were prepared by conventional emulsion co-
of the graft would determine the size of ionic do- polymerization. The two classes of polymer, graft and
mains, whereas the number density of graft chains random, exhibit very different properties of mechan-
would determine the number of ionic domains per ical strength, water uptake, proton conductivity, and
unit volume. Collectively, the size and number den- thermal behavior as a result of their inherently
sity of ionic aggregates/clusters are expected to different morphologies.
control the degree of connectivity between ionic Using the same methodology as described above,
domains. Holdcroft et al. have also incorporated sodium sty-
Recently, several researchers have shown that it renesulfonate macromonomers as grafts to poly-
is possible to synthesize graft copolymers possessing (acrylonitrile) backbone chains.28 The poly(acryloni-
ionic grafts bound to hydrophobic backbones using trile) backbone is more hydrophilic than the poly-
macromonomers formed by stable free radical polym- (styrene) backbone of the previous study. This al-
erization (SFRP) techniques.26 The detailed synthesis lowed the authors to examine the effect that the
and characterization of this class of copolymer that backbone had on ionic domain morphology, water
comprises a styrenic main chain and sodium styre- uptake, conductivity, and oxygen permeability and
nesulfonate graft chains (PS-g-macPSSNa) was re- solubility. Holdcroft’s synthetic methodology for elu-
ported by Holdcroft et al.27 PS-g-macPSSNa was cidating important PEM properties is excellent, and
prepared by (1) pseudoliving, tempo-mediated free therefore, this group was able to draw rational
radical polymerization of sodium styrenesulfonate conclusions when developing structure-property re-
(SSNa) and (2) termination with divinylbenzene lationships of ion conducting polymers. These par-
(DVB). The macromonomer, macPSSNa, serves as ticular materials are probably not suitable PEMs
both the comonomer and emulsifier in the emulsion themselves due to the well-known susceptibility of
copolymerization with styrene. During polymeriza- the poly(styrene) backbone or poly(styrene sulfonate)
tion, the DVB terminus is located in the core of grafts to oxidative degradation in a fuel cell environ-
micellar particles and is incorporated into growing ment.
polystyrene (PS) as graft chains. The synthetic Poly(styrene sulfonic acid) grafts have also been
scheme is shown in Figure 6. By adjusting the attached to poly(ethylene-co-tetrafluoroethylene) (ET-
macPSSNa/styrene feed ratio, a series of polymers FE)28,29 and poly(vinylidene fluoride) (PVDF)30 as
(PS-g-macPSSNa) with uniform graft chain length shown in Figure 7. These materials were synthesized
Figure 7. ETFE-g-PSSA and PVDF-g-PSSA.28
by irradiating ETFE or PVDF preformed membranes

with γ radiation and then immersing the irradiated
membranes in styrene (usually diluted with toluene)
to affect polymerization of poly(styrene) grafts from
the backbone. The extent of grafting was reported to
Figure 8. Several possible poly(arylene ether) chemical
be controlled by styrene concentration, choice of structures.
diluent, grafting temperature, and grafting time.
Once the extent of grafting was characterized by polymerization of sulfonated monomers. Both
gravimetric means, the membranes were then sul- schemes are discussed below.
fonated using chlorosulfonic acid.
Gupta et al.31,32 and Buchi et al.33 have investigated 4.1. Postsulfonation of Existing Polymers
the radiation grafting (using a γ radiation source) of
tetrafluoroethylene-co-hexafluoropropylene (FEP) us- The most common way to modify aromatic poly-
ing styrene and divinylbenzene as monomers in the mers for application as a PEM is to employ electro-
formation of grafts which are then sulfonated with philic aromatic sulfonation. Aromatic polymers are
chlorosulfonic acid. Divinylbenzene was used to cre- easily sulfonated using concentrated sulfuric acid,
ate cross-links between grafts and possibly control fuming sulfuric acid, chlorosulfonic acid, or sulfur
water swelling of the membranes. Buchi et al. trioxide (or complexs thereof). Postmodification reac-
reported that FEP-grafted polystyrene sulfonic acid tions are usually restricted due to their lack of precise
(FEP-g-SSA) systems have physical and electrochemi- control over the degree and location of functionaliza-
cal properties superior to those of Nafion 117 but an tion, the possibility of side reactions, or degradation
inferior fuel cell performance attributed to the gas of the polymer backbone. Regardless, this area of
permeability of the membrane. The excessive gas PEM synthesis has received much attention and may
permeability was proposed to more readily allow be the source of emerging products such as sulfonated
HO• attack on the polystyrene grafts and lead to a Victrex poly(ether ether ketone).3a,36
loss of ion exchange capacity after operating in a fuel Sulfonated poly(arylene ether sulfone)s synthesized
cell environment, but no proof was offered. In a study by attaching sulfonic acid groups in polymer modi-
using similar styrene grafted membranes, nearly 10 fication reactions have been investigated intensively
wt % of the grafts were lost after 100 h of fuel cell since the pioneering work of Noshay and Robeson,
operation.34 PEMs may tolerate some degree of who developed a mild sulfonation procedure for the
backbone degradation during long-term fuel cell commercially available bisphenol A-based poly(ether
operation, but degradation involving a loss of ion sulfone).37 Different sulfonating agents have been
exchange capacity (and thus a loss in conductivity) employed for this polymer modification, such as
would be obviously undesirable in all situations. chlorosulfonic acid3a,4 and a sulfur trioxide-triethyl
phosphate complex. Sulfonation is an electrophilic
4. Poly(arylene ether)s substitution reaction; therefore, its application de-
pends on the substituents present on the aromatic
Wholly aromatic polymers are thought to be one ring. Electron-donating substituents will favor reac-
of the more promising routes to high performance tion, whereas electron-withdrawing substituents will
PEMs because of their availability, processability, not. Additionally, the sulfonic acid group is usually
wide variety of chemical compositions, and antici- restricted to the activated position on the aromatic
pated stability in the fuel cell environment. Specifi- ring. For the case of the bisphenol A-based systems,
cally, poly(arylene ether) materials such as poly- no more than one sulfonic acid group per repeat unit
(arylene ether ether ketone) (PEEK), poly(arylene could be achieved.38
ether sulfone), and their derivatives are the focus of A comparative study of sulfonating agents was
many investigations, and the synthesis of these undertaken by Genova-Dimitrova et al.39 Their aim
materials has been widely reported.35 This family of was to investigate the characteristics of chlorosulfonic
copolymers is attractive for use in PEMs because of acid (ClSO3H) and derived trimethylsilylchlorosul-
their well-known oxidative and hydrolytic stability fonate ((CH3)3SiSO3Cl) on the sulfonation of bisphe-
under harsh conditions and because many different nol A-based poly(sulfone). The authors indicated that
chemical structures, including partially fluorinated the strong sulfonating agent, chlorosulfonic acid,
materials, are possible, as shown in Figure 8. Intro- yielded an inhomogeneous reaction that could be
duction of active proton exchange sites to poly- solubilized with a small amount of dimethylforma-
(arylene ether)s has been accomplished by both a mide as a cosolvent. Reactions with the mild tri-
polymer postmodification approach and direct co- methylsilylchlorosulfonate sulfonating agent were
Figure 10. Chemical structures of unsulfonated and

sulfonated PEEK.
one sulfonic acid group adds per repeat unit. It has

been reported that sulfonation of PEEK with chloro-
sulfonic acid or fuming sulfuric acid causes a mostly
unexplored degradation of the polymer; therefore,
Figure 9. Most likely locations of sulfonic acid attachment
in postsulfonated poly(arylene ether sulfone).39 concentrated sulfuric acid is typically used.45 The
sulfonation rate of PEEK in sulfuric acid can be
homogeneous reactions. Chlorosulfonic acid also in- controlled by changing the reaction time, tempera-
duced chain cleavages during some sulfonations, as ture, and acid concentration to provide polymers with
indicated by viscometric measurements, whereas no a sulfonation range of 30-100% without degradation
polymer degradation or cross-linking was observed and cross-linking reactions.46 However, it has been
with the milder trimethylsilylchlorosulfonate. Proton shown that the sulfonation of PEEK in sulfuric acid
nuclear magnetic resonance (1H NMR) was used to cannot be used to produce truly random copolymers
assemble a detailed picture of the kinetics of the at sulfonation levels less than 30% because dissolu-
sulfonation reaction with trimethylsilylchloro- tion and sulfonation occur in a heterogeneous envi-
sulfonate. A marked decrease in reaction rate was ronment.47
observed after a degree of sulfonation of 0.74 (74% Al-Omran and Rose48 controlled the location and
of the monomer units sulfonated) and 1.35 (68% of extent of sulfonation on poly(arylene ether) back-
the theoretical yield of a degree of sulfonation of 2), bones by copolymerizing 4,4′-dichlorodiphenyl sul-
as shown in Figure 9. The authors were not able to fone, durohydroquinone, and hydroquinone to form
present a sound explanation for the rate decrease at random copolymers, where only the hydroquinone
these two levels of conversion. residue would be expected to be susceptible to sul-
Another drawback of the mild sulfonating reagent fonation by sulfuric acid. Although these authors
is that its sulfonation efficiency was low and during observed sulfonation at positions other than at the
short reaction times (4 h) was limited to a degree of desired hydroquinone locations, designing sulfonation
sulfonation of 0.85 even when a 3-fold excess of sites into a polymer backbone remains an attractive
sulfonating agent was employed. Clearly, tradeoffs strategy for producing copolymers with known struc-
exist between strong and mild sulfonating agents, but tures. This allows the chemical structure and com-
avoiding polymer degradation and side reactions position of the material to dictate the extent of
probably outweighs longer reaction times and some- sulfonation rather than trying to externally control
what lower sulfonation efficiency. The hydrolytic sensitive and sometimes unpredictable macromolecu-
stability of the isopropylidene unit under fuel cell lar sulfonation reactions.
conditions has not been reported. An alternative, but likely expensive, route for
Poly(ether ether ketone) (PEEK) is an aromatic, derivatizing a commercial polysulfone (Udel) by
high performance, semicrystalline polymer with ex- Kerres et al. was based on a series of steps including
tremely good thermal stability, chemical resistance, metalation, sulfination by SO2 gas, and oxidation as
and electrical and mechanical properties. This poly- shown in Figure 11.49 One of the key steps in this
mer shows little solubility in organic solvents due to synthesis is the choice of oxidant to convert the
the crystallinity. One of the first ways to characterize lithium sulfinate to sulfonic acid. Cross-linking reac-
PEEK was by sulfonating the polymer. By adding tions during oxidation may reduce the ion exchange
sulfonic acid groups to the backbone, the crystallinity capacity of the materials, while polymer chain deg-
decreased and solubility increased.40,41 Commercially radation will result in membranes with quite poor
available Victrex appears to be one of the more mechanical properties. It was determined that hy-
interesting poly(arylene ether)s used for postmodifi- drogen peroxide was the best oxidant for low IEC
cation. materials because of its ability to access all the ionic
There have been many procedures reported for groups of the polymer in solution. Higher IEC ma-
sulfonating PEEK.3a,42,43 General chemical structures terials were amenable to oxidation by KMnO4 with-
for the unsulfonated and sulfonated analogues are out adverse cross-linking or polymer degradation.
given in Figure 10. The sulfonation of PEEK has been This synthetic scheme also allows controlled cross-
reported to be a second-order reaction, which takes linking of the materials to manipulate swelling and
place at the aromatic ring flanked by two ether links, gas permeability.50,51 The authors of this study assert
due to the higher electron density of the ring.44 Since the validity of this method for any polymeric species
the electron density of the other two aromatic rings that can be lithiated, and it would be an attractive
in the repeat unit is relatively low due to the electron- method for controlled sulfonation if polymer backbone
attracting nature of the neighboring carbonyl group, degradation can be avoided.
Figure 13. Placement of the sulfonic acid group in

postsulfonation (activated ring) versus direct copolymeri-
zation (deactivated ring).
Figure 11. Metalation route to sulfonated polysulfone.49
Figure 14. Synthesis of 3,3′-disulfonated 4,4′-dichloro-

diphenyl sulfone and its sodium salt.
bonded to an electron-rich aromatic ring.53 Moreover,

the opportunity to control and/or increase molecular
weight to enhance durability is not feasible if one
chooses to conduct a postreaction on an existing
commercial product. The difference between sulfonic
acid placement in typical examples of postsulfonation
and direct copolymerization is shown in Figure 13.
The first report of the required sulfonated mono-
mer was from Robeson and Matzner,54 who obtained
a composition of matter patent, which primarily was
of interest for its flame retarding properties. More
recently, Ueda et al.55 reported the sulfonation of 4,4′-
Figure 12. Chemical structure of sulfophenylated poly- dichlorodiphenyl sulfone and provided general pro-
sulfone.52 cedures for its purification and characterization.
Similarly, polysulfone has been sulfophenylated by McGrath’s group modified the procedure for disul-
lithiation and anionic reaction with 2-sulfobenzoic fonation of the monomer, shown in Figure 14. Sul-
acid cyclic anhydride (Figure 12).52 This provides fonated poly(arylene ether sulfone) copolymers were
another method to modify polysulfones by attaching then synthesized via direct copolymerization in any
pendant sulfonated phenyl groups via ketone links. composition desired as shown in Figure 15.56 The
It would be interesting to see if the phase separation directly copolymerized sulfonated materials were
in these materials was affected by the additional produced under very similar reaction conditions
functionality of the ketone or the pendant attachment employed for many years for the synthesis of unsul-
of the sulfonic acid, as opposed to direct attachment fonated poly(arylene ether)s using the weak base
of ionic groups to the aromatic polymer backbone. route. Only moderately higher reaction temperatures
and longer times were needed to obtain high molec-
ular weight copolymers due to the sterically de-
4.2. Direct Copolymerization of Sulfonated creased activity of the sulfonated dihalide monomer.
Monomers To Afford Random (Statistical) Alternatively, increased reactivity can be realized by
Copolymers employing the more expensive disulfonated difluoro
There are several major drawbacks of postmodifi- dihalide. The copolymerizations were conducted to
cation methods including the lack of control over the afford the potassium salt form of the 3,3′-disulfonated
degree and location of functionalization, which is 4,4′-dichlorodiphenyl sulfone to take advantage of the
usually a problem when dealing with macromol- enhanced stability of the sulfonic acid salt. These
ecules. It has been of interest to investigate the effect random copolymers displayed a hydrophilic/hydro-
of sulfonation, for example, on the deactivated sites phobic phase separated morphology that varied de-
of the repeat units, since one might expect enhanced pending on the degree of disulfonation (Figure 15).
stability and higher acidity from two sulfonic acid A tapping mode atomic force microscopy (AFM) phase
groups which are attached to an electron-deficient image comparison of the 40% copolymer with Nafion
aromatic ring than from one sulfonic acid group is shown in Figure 16.
Figure 15. Synthesis of directly copolymerized wholly aromatic sulfonated poly(arylene ether sulfone),56 BPSH-xx, where
xx is the ratio of sulfonated/unsulfonated activated halide.
due to the variety of monomers available, as depicted

in Figure 18.
The influence of the bisphenol structure on the
direct synthesis of sulfonated poly(arylene ether)s has
been studied.57 Four bisphenols including bisphenol
A, hexafluoroisopropylidene bisphenol (6F), 4,4′-
biphenol, and hydroquinone were used for the syn-
thesis of poly(arylene ether)s containing ion conduct-
ing units. Generally, the thin film properties of these
copolymers scaled with ion exchange capacity. Sul-
fonated poly(arylene thioether sulfone) copolymers
were prepared by direct copolymerization of sul-
fonated 4,4-difluorodiphenyl sulfone, unsulfonated
4,4-difluorodiphenyl sulfone, and 4,4-thiobisben-
zenethiol.58 The properties of these copolymers were
Figure 16. Atomic force micrographs of BPSH-40 and comparable with those of their poly(ether sulfone)
Nafion 117. analogue. Copolymers based on hexafluoroisoprop-
ylidene bisphenol (6F) have been particularly inter-
esting in initial fuel cell tests.59 It is thought that
their fluorine content promotes adhesion and elec-
trochemical compatibility with Nafion-based elec-
trodes and reduces swelling. Surface fluorine enrich-
ment of the 6F containing materials may also provide
enhanced membrane stability. The degree of sulfona-
tion and method of acidification also influenced the
dry dynamic mechanical behavior and Tg values, as
presented in Figure 19. The modulus-temperature
plot provides further evidence of the short-term
thermal stability of the acid form membrane. One
also notes broadening of the damping peaks, which
Figure 17. Swelling in 30 °C liquid water for the BPSH may reflect different degrees of association or other
series of copolymers. phenomena.
Directly copolymerized sulfonated poly(arylene ether
ketone) PEMs are also possible by employing a
The conductivity and water uptake of this series sulfonated dihalide ketone monomer (sodium 5,5′-
of copolymers also increased with disulfonation. carbonylbis(2-fluorobenzenesulfonate)), as first re-
However, once the degree of disulfonation reached ported by Wang.60,61 The sulfonated monomer chemi-
60 mol %, a semicontinuous hydrophilic phase was cal structure is shown in Figure 20.
observed and the membranes swelled dramatically, As described earlier, the choice of bisphenols for
forming a hydrogel that would not be useful as a the polymerization of poly(arylene ether ketone)s is
proton exchange membrane, as illustrated in Figure large.62 In particular, the electrochemical properties
17. These results indicate that the protonic conduc- of the above monomer copolymerized with bisphenol
tivity must be balanced with the water swelling and AF were studied. The fundamental PEM character-
mechanical properties of the membrane in these istics (water uptake and conductivity) were analogous
random copolymers. As with the unsulfonated poly- to those of the “BPSH” systems for a given IEC.
mers, many variations are possible in the direct Due to the flexibility of incorporating various
synthesis of sulfonated poly(arylene ether sulfone)s chemical moieties, sulfonated poly(arylene ether)
Figure 18. Some possible chemical structures for sulfonated PEMs from poly(arylene ether)s.
These polymers show lower water uptake than the

analogous sulfonated poly(arylene ether sulfone)
materials, possibly suggesting some interaction be-
tween the aromatic nitrile and sulfonic acid. The
phosphine oxide functional moiety could also be used
as a compatibilizer with other materials. Sulfonated
poly(arylene ether phosphine oxide sulfone) terpoly-
mers have been prepared both with sulfonated tri-
phenyl phosphine oxide65 and with triphenyl phos-
phine oxide with 3,3′-disulfonate-4,4′-dichlorodiphenyl
sulfone66 as the sulfonic acid bearing monomer. Block
copolymers containing phosphine oxide appear to
avoid the ether-ether interchange that results when
non-phosphine oxide blocks are utilized, and this is
being further pursued.67
5. PEMs Based on Poly(imide)s

Five-membered ring polyimides are high perfor-
mance materials and have been investigated for
many years. However, when sulfonated phthalic
polyimides are used for proton exchange membranes
in fuel cells, they quickly degrade, whereas it has
been recognized that naphthalenic polyimides are
much more stable in fuel cell environments.2 It is
likely that hydrolysis of the phthalic imide structure
leads to chain scission (decrease in molecular weight)
Figure 19. Effect of degree of disulfonation and method and causes the membrane to become brittle. Since
of acidification on the dynamic mechanical properties of the six-membered ring of the naphthalenic polyimide
BPSH films.53 M1 and M2 refer to whether acidification is much more stable to hydrolysis, this chemical
was done at 30 or 100 °C, respectively. structure is somewhat better suited for PEM fuel cell
applications but its stability is still questionable.
Genies et al.68 used model compounds along with
IR and NMR to examine the nature of hydrolysis
associated with the sulfonic acid group (needed for
Figure 20. Chemical structure of sodium 5,5′-carbonylbis- proton conduction) in phthalic and naphthalenic
(2-fluorobenzenesulfonate). polyimides. Model compounds of the sulfonic acid
containing phthalic imide (model A) and the sulfonic
copolymers may serve as ideal hosts for inorganic acid containing naphthalenic imide (model B) were
compounds to form nanocomposite proton exchange prepared by a one-step high temperature condensa-
membranes.3b,63 Additional functionality to increase tion in m-cresol. NMR spectroscopy showed that after
the compatibilization between the organic and inor- aging model A in distilled water at 80 °C for 1 h the
ganic phases can be imparted to the poly(arylene structure is modified. After aging for 10 h, 13C NMR
ether) by the copolymerization of 2,6-dichloroben- shows that the carbon peaks associated with the
zonitrile, hexafluoroisopropylidene bisphenol (6F), starting compound disappear. In addition, 13C NMR
and 3,3′-disulfonate-4,4′-dichlorodiphenyl sulfone.64 was used to quantitatively determine the amount of
Figure 21. Synthesis of SPI, a sulfonated six-membered ring polyimide based on BDA, ODA, and NDA.69
imide, amic acid, and diacid as a function of time that a block length of three sulfonated repeat units
during the aging process of model A. From 0 to 2 h yields the highest proton conductivity.70
of aging, the amic acid is formed preferentially to the Preliminary investigations suggested that six-
diacid. This is interesting because the amic acid membered ring polyimides had some promise as
should easily hydrolyze to the diacid in water. PEMs; however, their poor solubility limits mem-
In comparison, no structural modification of model brane formation and subsequent use in fuel cells. For
B was seen before 120 h of aging (80 °C). However, example, the SPI membrane shown above is only
after 120 h two small doublets appeared in the 1H soluble in chlorophenol. By introducing a slightly
NMR spectrum and several additional peaks became different unsulfonated diamine, 2,2′-diamino-3,4′-
noticeable in the 13C NMR spectrum. It was deter- biphenyl ether, the resulting polymer is made soluble
mined by NMR and IR spectroscopy that the hy- in m-cresol.71 In Figure 21, the unsulfonated diamine
drolysis products were an imide/carboxylic acid and added to the reaction in phase 2 was shown to have
an imide/anhydride. Model B was then aged for 1200 a large effect on the solubility, and several novel
h at 80 °C to quantitatively determine the amount unsulfonated diamines containing ether linkages
of hydrolysis products as a function of time. The and/or bulky substituents were shown to improve
relative intensity of the peaks due to carboxylic acid solubility.69
is constant after some time. The authors suggest that The preparation of sulfonated polyimides with
an equilibrium occurs between model B and the different ion exchange capacities and sulfonated block
products formed during hydrolysis, and therefore, the lengths was also considered. The solubility of the
conversion to hydrolysis products is limited to about polyimides was greatly improved by introducing
12%. This critical fraction is probably enough to cause phenyl-ether bonds and bulky groups into the poly-
some degradation of polymeric materials, but re- mer backbones. Random sulfonated copolyimides
search on six-membered polyimides has remained exhibit better solubility than the sequenced ones in
active. organic solvents. Therefore, improved solubility is a
Genies et al.69 designed a synthetic method to result of the unsulfonated diamine and the micro-
produce random and block (segmented) sulfonated structure of the polymer chain. For a given polymer
copolyimides (SPIs). The synthetic procedure for their structure, the water uptake increases as ionic content
most studied copolymer is shown in Figure 21. The increases. However, the number of water molecules
first step in the synthesis involves preparation of per ionic group remains constant, which suggests
short sequences of 4,4′-diamino-2,2′-biphenyl disul- that water is mainly located in the hydrophilic
fonic acid (BDA) condensed with 1,4,5,8-tetracar- domains. Both the number of water molecules per
boxylic dianhydride (NDA). An adjusted ratio of these ionic group and conductivity are systematically lower
two monomers allows one to create different block for random microstructures than for sequenced co-
lengths of the sulfonated sequence. In the second polymers.
polymerization step, the degree of sulfonation can be By introducing bulky unsulfonated diamines into
precisely controlled by regulating the molar ratio of the polymer backbone, greater interchain spacings
BDA and the unsulfonated diamine, which is 4,4′- can be achieved.72,73 It was proposed that the in-
oxydianiline (ODA) in SPI. Controlling the degree of creased spacing could improve conductivity at low
sulfonation is important because a high degree of relative humidity. Incorporating the larger comono-
sulfonation generally leads to high swelling or even mers prevents regular close parallel packing of the
dissolution of the polyimide membrane. It was found backbones and results in a more open structure, as
Figure 22. Six-membered ring copolyimide prepared with bulky unsulfonated diamine.
Figure 23. Sulfonated diamines for direct synthesis of sulfonated polyimides.
shown by X-ray diffraction patterns. With greater

interchain spacings, more free volume is available for
water to occupy. This leads to higher water uptakes
and, therefore, higher conductivity, especially at low
humidity.
Litt reported that bulky comonomers and se-
quenced copolymers impart improved water resis-
tance without sacrificing conductivity. However, one
sample shown in Figure 22 became brittle after 3
weeks in water above 90 °C. Litt also observed better
hydrolytic stability when the unsulfonated diamine
with flexible ether linkages was incorporated.72,73
Principally, one commercially available sulfonated
diamine (4,4′-diamino-2,2′-biphenyl disulfonic acid)
has been used to synthesize sulfonated polyimides.
In addition to the commercially available diamine,
several novel sulfonated diamines incorporating flex-
ible or kinked structures have been prepared in
Okamoto’s lab.74-77 The chemical structures and
names of all five diamines are shown in Figure 23. Figure 24. Synthesis of a five-membered ring sulfonated
polyimide containing phosphine oxide.78
A comparison of the hydrolytic stability of several
sulfonated six-membered ring polyimides was previ- 2,2′-biphenyl disulfonic acid (BDA) to the more flex-
ously investigated.75 Membranes were placed in ible 4,4′-diaminodiphenyl ether-2,2′-disulfonic acid
distilled water at 80 °C until a loss of mechanical (ODADS), the stability in water greatly improved. On
properties was observed. Improvements in membrane the other hand, 9,9′-bis(4-aminophenyl)fluorine-2,7-
stability were observed for polymers with lower disulfonic acid (BAPFDS), a rigid and bulky sul-
degrees of sulfonation (lower IEC) and for random fonated diamine, showed similar stability to the
copolymers, as opposed to block or sequenced copoly- ODADS series at comparable ion exchange capacities.
mers. Additionally, the flexibility of the sulfonated It is suggested that polyimides derived from BAPFDS
diamine in the polymer structure was shown to play should display higher stability due to the highly basic
an important role in stability. By simply changing sulfonated diamine.76 Aromatic diamines with higher
the sulfonated diamine from the rigid 4,4′-diamino- basicity are generally more reactive with dian-
Figure 25. Sulfonated six-membered ring polyimides with novel sulfonated diamines.79
hydrides than those with lower basicity. Since hy-

drolysis is the reverse reaction of polymerization,
polyimides derived from more basic diamines should
have higher hydrolytic stability. This may be why the
more rigid copolymers based on BAPFDS have sta-
bilities similar to the flexible ODADS ones; yet both
perform better than BDA-type copolyimides. Figure 26. Chemical structure of sulfonated poly(phen-
ylquinoxaline).84
Shobha et al.78 employed a novel sulfonated di-
amine containing a phosphine oxide moiety in the
synthesis of a five-membered ring sulfonated poly- ranged between 70 and 100%. The stability of this
imide. The synthesis is shown in Figure 24. material in detailed fuel cell performance studies was
Since this is a five-membered imide, water stability shown to be <500 h, and this material was subse-
would be expected to be very poor. To improve quently abandoned in favor of materials that showed
stability, Einsla et al.79 synthesized two series of six- increased stability under fuel cell conditions.17
membered imides (the first with 4,4′-oxydianiline as Poly(2,6-dimethyl-1,4-phenylene oxide) (PPO) is an
the unsulfonated diamine and the second with bis- attractive material for the preparation of membranes
[4-(3-aminophenoxy)phenyl] sulfone) using 3,3′-dis- because it possesses excellent membrane-forming
ulfonic acid-bis[4-(3-aminophenoxy)phenyl] sulfone as properties, as well as good thermal and chemical
the sulfonated diamine, as shown in Figure 25. These stability. PPO was sulfonated with chlorosulfonic acid
polyimides are soluble in N-methylpyrrolidone (NMP) either in a chloroform solution,85 which is a common
and have displayed better, but still limited, stability technique for PPO sulfonation, or in 1,2-dichloro-
in water. Their electrochemical properties and fuel ethane.86 In both cases, the sulfonated product pre-
cell performance are the subject of an upcoming cipitated after the addition of a certain amount of
publication.80 For high temperature fuel cell opera- chlorosulfonic acid and could subsequently be easily
tions requiring a stability of >5000 h, however, the isolated. The amount of chlorosulfonic acid needed
six-membered polyimides have been disappointing for the precipitation of SPPO from the 1,2-dichloro-
candidates, and they are generally being abandoned ethane solution, however, was smaller, and thus the
for use as PEMs. Nonetheless, most of these polyim- degree of sulfonation of the obtained SPPO was lower
ides have extremely low methanol permeation and (25.4 vs 42.4%) for the sample sulfonated in chloro-
low electro-osmotic drag coefficients and might be form. The SPPO has shown good thermal stability
useful as room temperature DMFC membranes.81 and resistance against aqueous solutions of strong
acids and bases and oxidation agents.87 However, one
might expect that the benzylic sites would easily
6. High Performance Polymeric Backbones for oxidize.
Use in PEMs Kobayashi et al. reported sulfonation of poly(4-
Other high performance polymer backbones have phenoxybenzoyl-1,4-phenylene) (PPBP) with sulfuric
been explored as PEM materials in addition to poly- acid and compared the data with those of sulfonated
(arylene ether)s and polyimides. Ductile copolymers PEEK.88 As reported, the authors employed sulfuric
with high modulus and glass transition values are acid for the sulfonation reaction to avoid or at least
desirable PEM candidates. The hydrolytic and oxida- minimize degradation of the polymer by chlorosul-
tive stability of many of these materials remains to fonic acid or fuming sulfuric acid (Figure 27).
be determined. Nevertheless, interesting synthetic Sulfonated poly(4-substituted benzoyl-1,4-phenylene)
methodologies have been employed to investigate homopolymers and copolymers using concentrated
these materials, which have been instructive in the sulfuric acid or fuming sulfuric acid have been shown
search for new PEM candidates. to form sulfonated polymers with variable degrees
Poly(phenylquinoxaline) membranes cast from of sulfonation.89 To improve film formation of the
m-cresol were sulfonated by reactions in 50% sul- sulfonated polyphenylenes, multiblock copolymers
furic acid solution for 2 h followed by drying at 300 have been synthesized by reacting a more flexible
°C for 90 min.82,83 The attachment site for the sulfonic poly(arylene ether sulfone) with sulfonated poly-
acid group is believed to be ortho to the ether linkage, phenylenes.90
as in Figure 26, but even this has yet to be confirmed. 1,4-Bis(propylcarbamoyl)-2,3,5,6-tetraphenylben-
The resulting sulfonation levels varied from roughly zene and decafluorobiphenyl were polymerized to
50% to well in excess of 100%, but they typically form copolymers that could be sulfonated at specific
chlorosulfonic acid is quantitative, so the degree of

sulfonation can be controlled by the reaction stoichi-
ometry. The above materials were soluble in metha-
nol, which lead to bis(3,5-dimethylphenyl) sulfone
being employed as a comonomer, as shown in Figure
29. These copolymers, as well as the synthesis of
similar fluoroalkane containing copolymers, highlight
the range of materials that can be designed with
specific main chain chemical structures to the control
resulting properties and sulfonation reactions.92
Poly(phthalazinone ether ketone)s (PPEKs) were
Figure 27. Sulfonation reaction of poly(4-phenoxybenzoyl- synthesized by the reaction of the polymer in mix-
1,4-phenylene).88
tures of 95-98% concentrated sulfuric acid and 27-
sites by reaction with chlorosulfonic acid in methyl- 33% fuming sulfuric acid in the absence of other
ene chloride as shown in Figure 28.91 The reaction solvents.93 A general reaction scheme is given in
of the pendant phenyl rings at the para position with Figure 30. The mixed solvent systems of sulfuric acid
Figure 28. Synthesis and sulfonation of polymers containing tetraphenylphenylene ether and perfluorobiphenylene units.91
Figure 29. Synthesis of sulfonated copolymers from 1,4-bis(propylcarbamoyl)-2,3,5,6-tetraphenylbenzene, bis(3,5-

dimethylphenyl) sulfone, and decafluorobiphenyl.91
Figure 30. Sulfonation of PPEK.93
Figure 31. Synthesis of directly polymerized sulfonated poly(phthalazinone ether sulfone).94
and fuming sulfuric acid were utilized to limit

degradation of the polymer during sulfonation, which
occurred in pure fuming sulfuric acid.
As an alternative to the somewhat undesirable
postsulfonation approach, Xiao et al. synthesized
sulfonated poly(phthalazinone ether sulfone)s utiliz-
ing sulfonated 4,4′-difluorodiphenyl sulfone as the
sulfonated monomer, as shown in Figure 31.94 The
authors indicate that the low degree of swelling of
these materials may be advantageous in comparison
to the case of the directly polymerized sulfonated
poly(arylene ether) materials. As they reported, the
low swelling is due to hydrogen bonding involving the
carbonyl groups of the hydroxyphenyl phthalazinone
unit. The proton conductivities of the directly copo-
lymerized materials showed less temperature depen-
dence than those of the corresponding postsulfonated Figure 32. Sulfonation of PBI.95
polymers. The hydrolytic stabilities of these copoly-
mers were not reported. amount of ionized sites with the alkali metal hydride
Sulfonation of polybenzimidazole (PBI) can be or by controlling the ratio of polybenzimidazole to (4-
accomplished by heating the polymer-hydrogen sul- bromomethyl)benzenesulfonate, as shown in Figure
fate complex, which can be formed by casting a PBI 33. Control of the number and location of ionic groups
film from sulfuric acid or immersing a cast PBI film is critical to a systematic study of the PEM properties
in sulfuric acid and allowing the acid to permeate the and producing consistent material. Complexation
membrane, as shown in Figure 32.95 This material phenomena would be expected to be important.
has received much attention, both as a proton ex- Direct copolymerization of sulfonated monomers
change membrane candidate and also as a host for has been used to synthesize sulfonated poly(benz-
phosphoric acid,3d,96 especially given the success of imidazoles), poly(benzoxazole)s, and poly(benzothia-
unmodified PBI/phosphoric acid membranes97 in high zole)s. As an example, Kim et al. synthesized poly-
temperature fuel cells, as discussed elsewhere in this (benzthiazole)s from 2,5-diamino-1,4-benzenedithiol
review. However, the thermally initiated sulfonation dihydrochloride and either 2-sulfoterethphthalic acid
may have induced scission or cross-linking, as evi- sodium salt, 5-sulfoisophthalic acid sodium salt, or
denced by the sulfonated PBI’s insolubility and 2,4-disulfoisophthalic acid potassium salt in poly-
brittleness. phosphoric acid (PPA), as shown in Figure 34.100
Sulfonation of polybenzimidazole was also ac- Similar sulfonated poly(benzimidazole)96,101 and sul-
complished by proton abstraction with an alkali fonated poly(benzoxazole)102 structures have also
metal hydride followed by reaction with sodium (4- been synthesized. A general synthetic scheme for
bromomethyl)benzenesulfonate.98,99 The degree of each is shown in Figure 35. The stability of these
sulfonation in this synthesis can be controlled by the polymers in aqueous acidic environments appears
sis and sulfonation are shown in Figure 36. Due to

the electron-withdrawing nature of the sulfone group,
the copolymers were only susceptible to sulfonation
on the m-dichlorobenzene residue. The degree of
sulfonation could be controlled by the amount of
m-dichlorobenzene included in the polymerization.
7. Polyphosphazene PEMs
Polyphosphazene-based PEMs are potentially at-
tractive materials for both hydrogen/air and direct
methanol fuel cells because of their reported chemical
and thermal stability and due to the ease of chemi-
cally attaching various side chains for ion exchange
sites and polymer cross-linking onto the -PdN-
polymer backbone.105 Polyphosphazenes were ex-
plored originally for use as elastomers and later as
solvent-free solid polymer electrolytes in lithium
batteries, and subsequently for proton exchange
membranes.106
Figure 33. Synthetic scheme for controlled addition of (4- Polyphosphazenes are of great synthetic and tech-
bromomethyl)benzenesulfonate to PBI.98,99 nological interest because of the way in which the
side groups can be varied over an exceedingly wide
limited and will likely hamper their performance as range of structures, and this provides access to
PEMs.103 species with an almost unprecedented variety of
Mulhaupt et al. synthesized novel soluble copoly- tailored properties. They are also particularly suit-
arylenes via a Ni(0)-catalyzed coupling reaction of able for side-group and surface modification chem-
aryl chlorides.104 Molar ratios of dichlorodiphenyl istry because of the stability of the phosphorus-
sulfone (x) to m-dichlorobenzene (y) were used to vary nitrogen backbone. The surface chemistry of specific
the amount of m-phenylene in the final copolymer. polyphosphazenes has been studied by Allcock et al.
Then these copolymers were dissolved in chloroform with respect to the sulfonation process.105 The target
and sulfonated with chlorosulfonic acid. The synthe- sulfonation reactions were developed initially with
Figure 34. Synthetic scheme for the direct synthesis of sulfonated poly(benzthiazole)s.
Figure 35. Synthesis of sulfonated poly(benzimidazole) and poly(benzoxazole).

Figure 36. Synthesis and sulfonation of copolyarylenes.104
Figure 37. Poly[(3-methylphenoxy)(phenoxy)phosphazene] Figure 38. Possible sulfonation sites on poly[(3-meth-

and poly[bis(3-methylphenoxy)phosphazene].111 ylphenoxy)(phenoxy)phosphazene].109
small-molecule cyclic trimeric phosphazenes to allow capacity e 1.4 mequiv/g exhibited good mechanical
careful molecular characterization using the normal properties at room temperature in both the dry and
range of small-molecule analysis techniques.107 Once water-swollen states. The reported glass transition
optimized, the same reactions were carried out with temperatures varied from -28 to -10 °C for unsul-
the corresponding polymers. Standard polymer solu- fonated poly[bis(3-methylphenoxy)phosphazene] and
tion and materials characterization methods were the sulfonated analogue with an IEC of 2.1 mequiv/
applied to the product polymers in order to determine g, respectively. These low glass transition tempera-
the molecular structures and material characteris- tures may cause membrane failures under fuel cell
tics. Finally, the same reactions were applied to the conditions and have led researchers to explore sul-
surfaces of films prepared from polyphosphazene. The fonation and cross-linking of these polyphosphazene-
modified surfaces were examined by several surface based materials.
analysis techniques. Pintauro et al. reported that proton exchange
Allcock’s research led to the development of poly- membranes with sulfonate fixed charge sites were
phosphazene-based PEMs by his small molecule fabricated from poly[bis(3-methylphenoxy)phos-
studies of the sulfonation of cyclic trimeric phosphazene].111 The membrane ion exchange capacity
phazenes107 and the surface chemistry of polyphos- was fixed at 1.4 mequiv/g. Membranes with and
phazene macromolecules.105 In a 1993 report, he without cross-linking were examined, where polymer
described the sulfonation of aminophosphazenes with cross-linking was carried out (after sulfonation) by
1,3-propanesultone.108 While these specific materials dissolving benzophenone photoinitiator in the mem-
are not necessarily ideal as PEMs, this study dem- brane casting solution and then exposing the films
onstrated a novel technique for creating sulfonated to UV light after solvent evaporation. It is clear from
polyphosphazene materials that may provide more this work that cross-linking reduced the water up-
control over the sulfonated polymer product than take and thus many of the membrane’s transport
wholesale sulfonation of a base polymer by a strong properties. However, the protonic conductivity did not
sulfonating agent. seem to differ between the cross-linked and non-
Studies by Pintauro and co-workers have shown cross-linked specimens even though their water
that poly[(3-methylphenoxy)(phenoxy)phosphazene] uptakes were different. Once a sufficient level of
and poly[bis(3-methylphenoxy)phosphazene] (Figure hydration has been reached, it is possible that further
37) can be sulfonated by adding an SO3 solution in addition of water (increases in λ) will have no effect
dichloroethane dropwise to a polymer/dichloroethane on the protonic conductivity of the membrane due to
solution.109,110 A high ion exchange capacity (up to 2.0 the counterbalancing forces of increased hydration
mequiv/g) material was reported with no detectable (higher λ) and increasing the distance between acidic
polymer degradation. sites (membrane swelling).
The sulfonation reaction is controllable, but one In another report polyphosphazene copolymers
report indicates that reaction of the backbone nitro- were synthesized from bis(2-methylphenoxy)phosp-
gen occurs before sulfonation on the pendent ring hazene, which was sulfonated after polymerization.112
with poly[(3-methylphenoxy)(phenoxy)phosphazene], Polymers such as polyvinylidene fluoride, polyhexa-
as shown in Figure 38.109 fluoropropylene, and polyacrylonitrile were used to
Solution-cast membranes (100-200 µm in thick- produce a blended membrane system. Polymer blends,
ness) from sulfonated polymers with an ion exchange cross-linking, and other means of re-enforcement are
generally necessary for polyphosphazene PEMs be- reported the first phosphonic acid containing poly-
cause of the relatively poor mechanical behavior of mers and copolymers from three phosphine contain-
pure polyphosphazene filmssparticularly under hying aromatic difluorides with moderate molecular
drated conditions. Rational and controlled cross- weight, as shown in Figure 39.113
linking of membranes could enhance the properties The acid group in this instance is located in the
of some currently studied PEMs by decreasing the main chain of the copolymer. The majority of acid
methanol crossover (with some expense to conductiv-
containing polymers have the sulfonic acid function-
ity, but possibly not) and increasing the maximum
ality as the proton exchange site. There are a few
working temperature of the membrane by increasing
reports of phosphonic acid containing polymers as
the onset of the hydrated Tg.
membranes for fuel cell applications. They have lower
acidity than sulfonic acid; however, their better
8. Other Proton Conducting chemical and thermal stability with respect to the
MoietiessAlternatives to Sulfonation corresponding sulfonic acid-functionalized polymers
Interest in new solid polymer electrolytes has is believed to offer potential advantages.114 Poly-
driven some research groups to investigate other (arylene ether)s containing mono- or dibromotetra-
materials containing proton conducting moieties phenylphenylene ether and octafluorobiphenyl units
aside from sulfonic acid. Polymers and copolymers may be used as precursors. The brominated polymers
from monomers containing phosphonic-based proton were phosphonated with diethyl phosphite by a
conductors have been reported. Phosphonic and/or palladium-catalyzed reaction. Quantitative phospho-
phosphinic acid containing polymers have not been nation is reported when high concentrations of cata-
well studied because of the rather limited synthetic lyst were used. The diethylphosphonated polymers
procedures available for their preparation, compared were dealkylated by reaction with bromotrimethyl-
with sulfonic acid derivatives. Miyatake and Hay silane in carbon tetrachloride, followed by hydrolysis
Figure 39. Synthesis of phosphinic acid (X ) OH) containing poly(arylene ether).113
Figure 40. Synthesis of poly(arylene ether)s bearing phosphonic acid groups.114

Figure 41. Synthesis of a phosphonic acid containing poly(arylene ether) derived from a phenolphthalein-based bisphenol.115
with hydrochloric acid to form polymers with pendent

phosphonic acid groups, as shown in Figure 40.
The authors reported that tough, flexible films
were obtained by solution casting from dimethyl
sulfoxide (DMSO), which would indicate little or no
backbone chain degradation occurred during phos-
phonation and dealkylation.
Another phosphonic acid containing polymer was
also reported by Meng et al. and was based on
phenolphthalein chemistry.115 The bisphenol mono-
mer was synthesized from phenolphthalein and m-
aminophenylphosphonic acid. Poly(arylene ether)s
containing phosphonic acid groups were prepared
from the bisphenol as shown in Figure 41. This is
the first report on the attachment of phenylphospho-
nic acid groups as side chains to aromatic polyethers.
These polyethers had very high glass transition
temperatures and high molecular weights. However,
their reported conductivities were only 10-5-10-6
S/cm, which is probably too low for these materials
to be considered seriously as PEMs.
Allcock et al. also have investigated the use of
phosphonated polyphosphazenes as potential mem- Figure 42. Synthesis of polyphosphazenes bearing phenyl
brane materials for use in direct methanol fuel cells phosphonic acid side groups.116
(Figure 42).116 Membranes were found to have IEC
values between 1.17 and 1.43 mequiv/g and proton
conductivities between 10-2 and 10-1 S/cm. Methanol
diffusion coefficients for these membranes were found
to be at least 12 times lower than that for Nafion 117
and 6 times lower than that for a cross-linked
sulfonated polyphosphazene membrane.
The strength of the acid conducting moiety is
thought to play a role in the conductivity of proton
exchange membranes. This has encouraged research-
Figure 43. Structure of Nafion and bis[(perfluoroalkyl)
ers to explore the role of strongly acidic functional sulfonyl]imide.118
groups such as sulfonimide moieties in proton ex-
change membranes.117 DesMarteau et al. have com- sulfonic acid105 or sulfonimide119 proton conducting
pared Nafion and a structurally similar bis[(perflu- substituents.
oroalkyl) sulfonyl]imide-based ionomer, as shown in Cho et al.120 have recently described the synthesis
Figure 43.118 There was not a noticeable increase in of a sulfonimide containing monomer and the result-
proton conductivity produced by incorporation of the ing poly(arylene ether sulfone) copolymers. In this
sulfonimide moiety in their copolymer, in lieu of the procedure 3,3′-disulfonate-4,4′-dichlorodiphenyl sul-
perfluorosulfonic acid moiety of Nafion. In fact, their fone was refluxed in thionyl chloride, isolated, and
sulfonimide-based copolymer gave similar results in then reacted with trifluoromethanesulfonamide in
all respects to Nafion. Allcock et al. have also the presence of triethylamine to form the sulfonamide
explored sulfonated poly(phosphazene)s with either analogue monomer as shown in Figure 44. This
tive determination of the molecular weight distribu-

tion or intrinsic viscosity to obtain a relative measure
of the molecular weight. Characterization of molec-
ular weight in ion containing systems is complicated
by the presence of ionic groups attached to the
polymer backbone where ion-ion interactions affect
Figure 44. Synthesis of a sulfonimide containing an
activiated dichloro monomer.120
the characteristic size of the macromolecule in a
solvent. This ion effect on chain size is often termed
as the polyelectrolyte effect. The ionic groups can also
cause the polymer chains to associate with the
chromatographic column, giving erroneous results.
Adding a small amount of lithium bromide or chloride
to the chromatographic mobile phase is helpful in
suppressing the polyelectrolyte effect in some GPC
and intrinsic viscosity experiments to allow charac-
terization of the ion containing materials.
End group analysis by nuclear magnetic resonance
(NMR) can also be used to characterize the effects of
molecular weight on the properties of proton ex-
change membranes. In a study by Wang et al.,
Figure 45. Imidazole proton conductorssboth as free offsetting stoichiometry and endcapping with NMR
molecules and attached to a polymer backbone.123 sensitive tert-butyl phenyl groups (18 protons per
chain) was used to create a series of sulfonated poly-
sulfonimide monomer was then used to form directly (arylene ether sulfone)s with increasing molecular
polymerized sulfonimide bearing poly(arylene ether weight (Figure 46).124 It was determined that molec-
sulfone) copolymers in a similar synthesis to that of ular weight in the range 20 000-40 000 g/mol (20-
the sulfonic acid containing copolymer. 40K) did not have a large effect on protonic conduc-
All acidic proton conductors discussed so far in this tivity, but the mechanical properties and water
review have relied on the presence of large amounts uptake of the material are nevertheless dependent
of water (λ ) 10-30) as a mobile phase for the on the molecular weight of the copolymer, possibly
conduction of protons. Current targets for automotive related to chain entanglement issues (see Tables 1
use of hydrogen/air fuel cells are 120 °C and 50% or and 2).81 Refinement of these procedures is continu-
lower relative humidity. Under these conditions, the ing in our laboratory.
conductivity of the membrane decreases due to low The dependence of mechanical properties on mo-
water uptake at 50% relative humidity and thus lecular weight brings up another important area of
creates large resistive losses in the cell. To meet the characterization that is often neglected. Proton ex-
needs of advanced fuel cell systems, membranes will change membranes with good mechanical properties
have to function without large amounts of absorbed in both the dry and hydrated states are critical to
water. Organic-inorganic composites are one pre- successful MEA fabrication and long-term durability
ferred approach.3b,63 in a fuel cell device. The membrane must be able to
Imidazole proton conductors have been explored as withstand the stresses of both electrode processing
water replacement solvents121,122 and have also been and attachment and must also be mechanically
attached to polymer backbones123 to replace the acid/ robust enough to endure startup and shutdown of the
water complex in current PEMs (Figure 45). fuel cell with the repeated swelling/drying/heating/
cooling of the membrane. Nafion acts reasonably well
9. Important Considerations in the Design of New mechanically up to 80 °C as a rubbery material with
PEMs low modulus but high elongation to break.
As the preceding sections have illustrated, many These properties are advantageous for MEA fab-
families of polymers with differing chemical struc- rication, but Nafion’s rubbery behavior is likely
tures and various strategies for incorporation of detrimental to its long-term performance in high
sulfonic acid groups have been explored as proton temperature environments. Because of its low glass
exchange membrane materials. Most reports of new transition temperature when hydrated (hydrated
materials have included information on ion content Tg),125 hydrated Nafion undergoes a viscoelastic
(EW or IEC), protonic conductivity, and water up- relaxation causing the membrane to develop pinholes
take. Despite the large body of research on this topic, when operated at temperatures near or above 100
there are a few basic polymer properties that are still °C. A PEM based on a high Tg polymer such as a poly-
not well-known for common systems or not measured (arylene ether) can increase the membrane’s hy-
in most reports. Perhaps the most glaring omission drated Tg and potentially limit mechanical degrada-
in new ion conducting polymer research is the tion of the membrane at increased temperatures.
characterization of molecular weight. Molecular weight Molecular weight and mechanical properties are
is one of the most basic properties of polymeric two common but important properties that are mea-
materials and is most commonly measured by gel sured for a multitude of polymeric systems, but they
permeation chromatography ((GPC) or size exclusion have thus far been largely neglected in the charac-
chromatography (SEC)) or MALDI-TOF for quantita- terization of PEMs. Often new PEMs will show
Figure 46. Synthesis of end-capped copolymer (BPS-40): target Mn ) 40K g/mol.124

Table 1. Characterization of Molecular Weights and 50% relative humidity) and long-term durability
Intrinsic Viscosities (no Li salts) for End-Capped under fuel cell conditions. Current sulfonic acid-based
BPS-40a
materials suffer from low conductivity in the absence
calc monomer ratio [η]NMP at Mn (×10-3) of water. New proton conducting moieties and mor-
Mn by moleb 25 °C (dL/g) by NMR phologies are required to fulfill the requirements of
20K 43.13/42.13/2 0.54 23.1 these higher temperature systems. Lower perme-
30K 64.69/63.69/2 0.61 32.7 ability seems to be a good way to address new DMFC
40K 86.26/85.26/2 0.87 40.2
systems.
a
The IV of noncontrolled BPS-40 is 2.5 dL/g, which is much In addition to new ion conducting strategies, basic
higher than that of BPS-40-40K. The molecular weight of polymer characterization measurements must be
noncontrolled BPS-40 is much more than 40K. Chemical end-
cap techniques could be used to characterize the molecular identified and used to more advantage in the devel-
weight of other sulfonated polymers. b [(SDCDPS + DCDPS)/ opment of new PEMs. Issues of molecular weight,
biphenol/tert-butylphenol]. SDCDPS/(SDCDPS + DCDPS) ) mechanical properties, and chemical/physical degra-
0.4. dative mechanisms need to be addressed with much
more rigor than they have been in the past.
Table 2. Characterization of End-Capped BPS-40 and This review has attempted to review the basic
Stoichiometric Copolymersa
research on alternative polymer membranes with
water attached ionic groups and has tried to highlight
IEC SDCDPS uptake conductivity examples of materials with a well-defined chemical
polymer (mequiv/g) (mol %) (%) (S/cm)
structure that have been described in the literature.
BPS-40-20K 1.60 39.5 51 0.090 Nafion and similar polyperfluorosulfonic acid mem-
BPS-40-30K 1.60 39.8 57 0.092
BPS-40-40K 1.60 39.2 57 0.085 branes still remain the most studied materials due
1:1 stoichiometry 1.70 40.1 55 0.100 to their commercial applicability and wide avail-
theory 1.72 40 ability. However, for all that is known about the
a
Membranes acidified by boiling in 0.5 M H2SO4. Controlled
properties of these membranes, little is known about
copolymers have similar chemical compositions, water uptakes, their synthesis, chemical composition, or molecular
and conductivities, which are independent of molecular weight. weight. The search for more in depth understanding
of the link between the chemical composition of the
polymer and its resulting membrane properties has
promising protonic conductivity results, but the me-
led researchers to synthesize membranes with new,
chanical properties are not discussed and remain in
controlled chemical structures. Sulfonated poly-
question, which can hamper efforts to create MEAs.
(arylene ether)s such as PEEK and poly(sulfone) have
shown excellent chemical and thermal stability in
10. Conclusions fuel cell applications, but their weaker aryl sulfonic
acid groups generally cause lower proton conductivity
There is a clear need for the synthesis and more than perfluorosulfonic acid containing membranes.
complete characterization of new PEM polymeric This can be combated by the addition of more sulfonic
materials. Polymers bearing functional moieties for acid conductors, but it may lead to undesirably large
proton conduction might also be designed to serve as swelling. The weaker acid groups also impede the
a host for inorganic compounds to afford a proton performance of poly(arylene ether)-based membranes
conducting component in a blend, as well as a stand- at low relative humidity. These types of membranes
alone PEM. The two current hurdles for polymeric are certainly a viable alternative to Nafion in ap-
membranes are the high protonic conductivity at low plications where inlet gases with high relative hu-
water contents (e.g. under conditions of 120 °C and midity (above 80%) can be used, but their applica-
bility for low relative humidity applications is more Applications of fuel cell technology can be furthered
challenging. by the use of advanced materials, including organic-
Postsulfonation of polymers to form PEMs can lead inorganic composites.3b Focused programs with clear
to undesirable side reactions and may be hard to targets have been developed for fuel cell devices and
control on a repeatable basis. Synthesis of sulfonated systems in automotive, stationary, and portable
macromolecules for use in PEMs by the direct reac- systems. A key component in meeting the goals for
tion of sulfonated comonomers has gained attention each of these systems is the proton exchange mem-
as a rigorous method of controlling the chemical brane. Development efforts have thus far concen-
structure, acid content, and even molecular weight trated on the fabrication and design of fuel cell stacks
of these materials. While more challenging syntheti- using current materials. New materials must be able
cally than postsulfonation, the control of the chemical to meet the targets for advanced systems, while still
nature of the polymer afforded by direct copolymer- remaining somewhat compatible with current hard-
ization of sulfonated monomers and the repeatability ware and system designs. This can only be ac-
of the reactions allows researchers to gain a more complished through the directed synthesis of new
systematic understanding of these materials’ proper- copolymers with feedback from MEA and fuel cell
ties. Sulfonated poly(arylene ether)s, sulfonated poly- technology developers.
(imide)s, and sulfonated poly(styrene) derivatives
have been the most prevalent of the directly copoly- 11. Acknowledgments
merized materials.
The authors appreciate the support of this research
Using the direct synthesis route, the potential of by the Department of Energy (Contract No. DE-
forming well-defined block copolymers with sul- FC36-01G011086), DARPA-ARO (contract #DAAD19-
fonated and unsulfonated blocks has been realized 02-1-0278), NASA Glenn (contract # NCC3-886), UTC
in poly(styrene) or poly(acrylonitrile)-based materials Fuel Cells (contract #PO 3651) and the National
using styrene sulfonic acid, in poly(imide)s using Science Foundation (No. EHR-0090556, and 9975678).
sulfonated diamines, and in poly(arylene ether sul- They also thank other members of the McGrath
fone) utilizing phosphine oxide-based comonomers research group and the expert assistance of Ms.
that avoid ether-ether interchange reactions. The Laurie Good.
synthesis of block materials presents another level
of complexity, but studies involving these materials
indicate that the blocky nature of the copolymer
12. References
chemical structure gives rise to increased proton (1) (a) Kim, Y. S.; Pivovar, B. 204th Meeting of the Electrochemical
Society, Orlando, FL, Oct 12-16, 2003. (b) Pivovar, B. S.;
conductivity without massive increases in water Hickner, M.; Wang, F.; McGrath, J.; Zelenay, P.; Zawodzinski,
swelling. This is an area that shows promise in the T. A., Jr. Pre-Print ArchivesAmerican Institute of Chemical
rational design of new materials and should be Engineers, [Spring National Meeting], New Orleans, LA, United
States, Mar 11-14, 2002, p 2433. (c) Hickner, M. A.; Kim, Y. S.;
pursued further as a possible route to high temper- Wang, F.; Harrison, W.; Hill, M.; Dong, L.; Pivovar, B.; Zelenay,
ature membranes. P.; McGrath, J. E. Electrochim. Acta, in press.
(2) Savadogo, O. J. New Mater. Electrochem. Syst. 1998, 1, 47.
Sulfonated poly(arylene ether)s have shown prom- (3) (a) Roziere, J.; Jones, D. J. Annu. Rev. Mater. Res. 2003, 33,
ise for durability in fuel cell systems, while poly- 503. (b) Alberti, G.; Casciola, M. Annu. Rev. Mater. Res. 2003,
(styrene)- and poly(imide)-based systems serve as 33, 129. (c) Paddison, S. J. Annu. Rev. Mater. Res. 2003, 33, 289.
(d) Schulster, M. F. H.; Meyer, W. H. Annu. Rev. Mater. Res.
model systems for studying structure-relationship 2003, 33, 233.
properties in PEMs because their questionable oxida- (4) Kerres, J. J. Membr. Sci. 2001, 185, 3.
(5) Zawodzinski, T. A.; Neeman, M.; Sillerud, L. O.; Gottesfeld, S.
tive or hydrolytic stability limits their potential J. Phys. Chem. 1991, 95, 6040.
application in real fuel cell systems. Sulfonated high (6) (a) Alberti, G.; Casciola, M.; Palombari, R. J. Membr. Sci. 2000,
performance polymer backbones, such as poly(phe- 172, 233. (b) Alberti, G.; Constantino, U.; Casciola, M.; Ferroni,
S.; Massinelli, L.; Staiti, P. Solid State Ionics 2001, 145, 249.
nylquinoxaline), poly(phthalazinone ether ketone)s, (c) Pivovar, B. S.; Wang, Y.; Cussler, E. L. J. Membr. Sci. 1999,
polybenzimidazole, and other aromatic or heteroaro- 154, 155.
matic systems, have many of the advantages of poly- (7) Ibrahim, S. M.; Price, E. H.; Smith, R. A., of E. I. du Pont de
Nemours. Proc. Electrochem. Soc. 1983, 83-6.
(imides) and poly(arylene ether sulfone)s and may (8) Resnick, P. R.; Grot, W. G., of E. I. du Pont de Nemours and
offer another route to advanced PEMs. These high Company, Wilmington, DE, Sept 12, 1978; U.S. Patent 4,113,
585.
performance backbones would increase the hydrated (9) Bahar, B.; Cavalca, C.; Cleghorn, S.; Kolde, J.; Lane, D.; Murthy,
Tg of PEMs while not being as hydrolytically sensitive M.; Rusch, G. J. New Mater. Electrochem. Syst. 1999, 2 179.
as poly(imides). The synthetic schemes for these more (10) (a) Lin, J. C.; Kuntz, H. R.; Cutlip, M. B.; Fenton, J. M. Proc.
31st Mid-Atl. Ind. Hazard. Waste Conf. 1999, 31, P656. (b) Lin,
exotic macromolecules are not as well-known, but the J. C.; Kuntz, H. R.; Fenton, J. M. Proc. Electrochem. Soc. Power
interest in novel PEMs will surely spur developments Sources New Millenium 2001, 2000, 48.
in this area. (11) Beckerbauer, R., of E. I. du Pont de Nemours and Company,
Wilmington, DE, Jan 30, 1973; U.S. Patent 3,714,245.
The inorganic poly(phosphazene) backbone has (12) Grot, W. G., of E. I. du Pont de Nemours and Company,
received attention as a PEM candidate. This is an Wilmington, DE, Feb 27, 1973; U.S. Patent 3,718,627.
(13) Grot, W. G., of E. I. du Pont de Nemours and Company,
attractive system for study due to its ease of synthe- Wilmington, DE, Feb 21, 1984; U.S. Patent 4,433,082.
sis and subsequent modification by many functional (14) Rikukawa, M.; Sanui, K. Prog. Polym. Sci. 2000, 25, 1463.
(15) Tant, M. R.; Darst, K. P.; Lee, K. D.; Martin, C. W. ACS Symp.
groups. However, these membranes generally show Ser. 1989, 395, 370.
low glass transition temperatures and somewhat poor (16) Wei, J.; Stone, C.; Steck, A. E., of Ballard Power Systems, Inc.,
mechanical properties, and they require cross-linking June 6, 1995; U.S. Patent 5,422,411.
(17) (a) Steck, A. E. Proceedings of the first international symposium
to enhance their performance in hydrated environ- of new materials for fuel cell systems; l′Ecole Polytechnique de
ments. Montreal: Montreal, 1995; p 74. (b) Steck, A. E.; Stone, C.
Proceedings of the 2nd International Symposium on New Materi- (59) Kim, Y. S.; Sumner, M. J.; Harrison, W. L.; Riffle, J. S.; McGrath,
als for Fuel Cell and modern battery systems; l’Ecole Polytech- J. E.; Pivovar, B. S. J. Electrochem. Soc., accepted.
nique de Montreal: Montreal, 1997; p 792. (60) Wang, F.; Chen, T.; Xu, J. Macromol. Chem. Phys. 1998, 199,
(18) Ehrenberg, S. G.; Serpico, J. M.; Wnek, G. E.; Rider, J. N., of 1421.
Dais Corporation, Oct 21, 1997; U.S. Patent 5,679,482. (61) Wang, F.; Li, J.; Chen, T.; Xu, J. Polymer 1999, 40, 795.
(19) Wnek, G. Abstracts of Papers, 222nd ACS National Meeting, (62) (a) Gao, Y.; Robertson, G. P.; Guiver, M. D.; Jian, X.; Mikhailen-
Chicago, IL, United States, August 26-30, 2001. ko, S. D.; Wang, K.; Kaliaguine, S. J. Polym. Sci., Part A: Polym.
(20) Kim, J.; Kim, B.; Jung, B. J. Membr. Sci. 2002, 207, 129. Chem. 2003, 41, 2731. (b) Xing, P.; Robertson, G. P.; Guiver, M.
(21) Serpico, J. M.; Ehrenberg, S. G.; Fontanella, J. J.; Jiao, X.; D.; Mikhailenko, S. D.; Kaliaguine, S. J. Polym. Sci., Part A:
Perahia, D.; McGrady, K. A.; Sanders, E. H.; Kellogg, G. E.; Polym. Chem. 2004, 42, 2866.
Wnek, G. E. Macromolecules 2002, 35, 5916. (63) Kim, Y. S.; Wang, F.; Hickner, M.; Zawodzinski, T. A.; McGrath,
(22) Wnek, G. E.; Rider, J. N.; Serpico, J. M.; Einset, A. G.; J. E. J. Membr. Sci. 2003, 212, 263.
Ehrenberg, S. G.; Raboin, L. A. Proc. Electrochem. Soc. 1995, (64) Sumner, M. J.; Harrison, W. L.; Weyers, R. M.; Kim, Y. S.;
95-23, 247. McGrath, J. E.; Riffle, J. S.; Brink, A.; Brink, H. M. J. Membr.
(23) Wnek, G. E.; Rider, J. N.; Serpico, J. M.; Einset, A. G. Proceed- Sci. 2004, 239, 199.
ings of the First International Symposium on Proton Conducting (65) Shobha, H. K.; Smalley, G. R.; Sankarapandian, M.; McGrath,
Membrane Fuel Cells; Electrochemical Society: 1995; p 247. J. E. Polym. Prepr. 2000, 41, 180.
(24) Hodgdon, R. B., Jr. J. Polym. Sci. 1968, 6, 171. (66) Wang, F.; Mecham, J.; Harrison, W.; Hickner, M.; Kim, Y. S.;
(25) Serpico, J. M.; Ehrenberg, S. G.; Fontanella, J. J.; McGrady, K. McGrath, J. E. Polym. Mater. Sci. Eng. 2001, 84, 913.
A.; Perahia, D.; Jiao, X.; Sanders, E. H.; Wallen, T. J.; Wnek, G. (67) Wang, F.; Kim, Y.; Hickner, M.; Zawodzinski, T. A.; McGrath,
E. Polym. Mater. Sci. Eng. 2002, 86, 32. J. E. Polym. Mater. Sci. Eng. 2001, 85, 517.
(26) Ding, J.; Chuy, C.; Holdcroft, S. Chem. Mater. 2001, 13, 2231. (68) Genies, C.; Mercier, R.; Sillion, B.; Petiaud, R.; Cornet, N.; Gebel,
(27) Ding, J.; Chuy, C.; Holdcroft, S. Macromolecules 2002, 35, 1348. G.; Pineri, M. Polymer 2001, 42, 5097.
(28) Chuy, C.; Ding, J.; Swanson, E.; Holdcroft, S.; Horsfall, J.; Lovell, (69) Genies, C.; Mercier, R.; Sillion, B.; Cornet, N.; Gebel, G.; Pineri,
K. V. J. Electrochem. Soc. 2003, 150, E271. M. Polymer 2001, 42, 359.
(29) Chuy, C.; Basura, V. I.; Simon, E.; Holdcroft, S.; Horsfall, J.; (70) Cornet, N.; Diat, O.; Gebel, G.; Jousse, F.; Marsacq, D.; Mercier,
Lovell, K. V. J. Electrochem. Soc. 2000, 147, 4453. R.; Pineri, M. J. New Mater. Electrochem. Syst. 2000, 3, 33.
(30) Flint, S. D.; Slade, R. C. T. Solid State Ionics 1997, 97, 299. (71) Vallejo, E.; Pourcelly, G.; Gavach, C.; Mercier, R.; Pineri, M. J.
(31) Gupta, B.; Buchi, F. N.; Scherer, G. G.; Chapiro, A. Solid State Membr. Sci. 1999, 160, 127.
Ionics 1993, 61, 213. (72) Zhang, Y.; Litt, M.; Savinell, R. F.; Wainright, J. S. Polym. Prepr.
(32) Gupta, B.; Buchi, F. N.; Scherer, G. G.; Chapiro, A. Polym. Adv. 1999, 40, 480.
Technol. 1994, 5, 493. (73) Zhang, Y.; Litt, M.; Savinell, R. F.; Wainright, J. S.; Vendramint,
(33) Buchi, F. N.; Gupta, B.; Haas, O.; Scherer, G. G. Electrochim. J. Polym. Prepr. 2000, 41, 1561.
Acta 1995, 40, 345. (74) Zhou, W.; Watari, T.; Kita, H.; Okamoto, K.-I. Chem. Lett. 2002,
(34) Becker, W.; Schmidt-Naake, G. Chem. Eng. Technol. 2002, 25, 5, 534.
373. (75) Fang, J.; Guo, X.; Harada, S.; Watari, T.; Tanaka, K.; Kita, H.;
(35) Wang, S.; McGrath, J. E. Synthesis of Poly(arylene ether)s. Okamoto, K.-I. Macromolecules 2002, 35, 9022.
Synthetic Methods in Step-Growth Polymers; Wiley: New York, (76) Guo, X.; Fang, J.; Watari, T.; Tanaka, K.; Kita, H.; Okamoto,
2003; p 327. K.-I. Macromolecules 2002, 35, 6707.
(36) Alberti, G.; Casciola, M.; Massinelli, L.; Bauer, B. J. Membr. Sci. (77) Yin, Y.; Fang, J.; Cui, Y.; Tanaka, K.; Kita, H.; Okamoto, K.-I.
2001, 185, 73. Polymer 2003, 44, 4509.
(37) Noshay, A.; Robeson, L. M. J. Appl. Polym. Sci. 1976, 20, 1885. (78) Shobha, H. K.; Sankarapandian, M.; Glass, T. E.; McGrath, J.
(38) Johnson, B. C.; Yilgor, I.; Tran, C.; Iqbal, M.; Wightman, J. P.; E. Polym. Prepr. 2000, 41, 1298.
Lloyd, D. R.; McGrath, J. E. J. Polym. Sci.: Polym. Chem. Ed. (79) Einsla, B. R.; Hong, Y. T.; Kim, Y. S.; Wang, F.; Gunduz, N.;
1984, 22, 721. McGrath, J. E. J. Polym. Sci., Part A: Polym. Chem. 2004, 42,
(39) Genova-Dimitrova, P.; Baradie, B.; Foscallo, D.; Poinsignon, C.; 862.
Sanchez, J. Y. J. Membr. Sci. 2001, 185, 59. (80) Einsla, B. R.; Kim, Y. S.; Hickner, M.; Hong, Y. T.; Hill, M. L.;
(40) Kaliaguine, S.; Mikhailenko, S. D.; Wang, K. P.; Xing, P.; Pivovar, B.; McGrath, J. E. J. Membr. Sci., submitted.
Robertson, G.; Guiver, M. Catal. Today 2003, 82, 213. (81) Hickner, M. A. Ph.D. Dissertation, Virginia Polytechnic Institute
(41) Robertson, G. P.; Mikhailenko, S. D.; Wang, K.; Xing, P.; Guiver, and State University, September 2003.
M. D.; Kaliaguine, S. J. Membr. Sci. 2003, 219, 113. (82) Kopitzke, R. W.; Linkous, C. A.; Nelson, G. L. J. Polym. Sci.,
(42) Bauer, B.; Jones, D. J.; Roziere, J.; Tchicaya, L.; Alberti, G.; Part A: Polym. Chem. 1998, 36, 1197.
Casciola, M.; Massinelli, L.; Peraio, A.; Besse, S.; Ramunni, E. (83) Kopitzke, R. W.; Linkous, C. A.; Anderson, H. R.; Nelson, G. L.
J. New Mater. Electrochem. Syst. 2000, 3, 93. J. Electrochem. Soc. 2000, 147, 1677.
(43) Zaidi, S. M. J.; Mikhailenko, S. D.; Robertson, G. P.; Guiver, M. (84) Kopitzke, R. W.; Linkous, C. A.; Nelson, G. L. Polym. Degrad.
D.; Kaliaguine, S. J. Membr. Sci. 2000, 173, 17. Stab. 2000, 67, 335.
(44) Xing, P.; Robertson, G. P.; Guiver, M. D.; Mikhailenko, S. D.; (85) Kruczek, B.; Matsuura, T. J. Membr. Sci. 1998, 146, 263.
Wang, K.; Kaliaguine, S. J. Membr. Sci. 2004, 229, 95. (86) Kosmala, B.; Schauer, J. J. Appl. Polym. Sci. 2002, 85, 1118.
(45) Bishop, M. T.; Karasz, F. E.; Russo, P. S.; Langley, K. H. (87) Fu, H.; Jia, L.; Xu, J. J. Appl. Polym. Sci. 1994, 51, 1399.
Macromolecules 1985, 18, 86. (88) Kobayashi, T.; Rikukawa, M.; Sanui, K.; Ogata, N. Solid State
(46) Huang, R. Y. M.; Shao, P.; Burns, C. M.; Feng, X. J. Appl. Polym. Ionics 1998, 106, 219.
Sci. 2001, 82, 2651. (89) Ghassemi, H.; McGrath, J. E. Polymer 2004, 45, 5847.
(47) Bailly, C.; Williams, D. J.; Karasz, F. E.; MacKnight, W. J. (90) Ghassemi, H.; Ndip, G.; McGrath, J. E. Polymer 2004, 45, 5855.
Polymer 1987, 28, 1009. (91) Miyatake, K.; Hay, A. S. J. Polym. Sci., Part A: Polym. Chem.
(48) Al-Omran, A.; Rose, J. B. Polymer 1996, 37, 1735. 2001, 39, 3211.
(49) Kerres, J.; Cui, W.; Richie, S. J. Polym. Sci., Part A: Polym. (92) Miyatake, K.; Oyaizu, K.; Tsuchida, E.; Hay, A. S. Macromol-
Chem. 1996, 34, 2421. ecules 2001, 34, 2065.
(50) Kerres, J.; Zhang, W.; Cui, W. J. Polym. Sci., Part A: Polym. (93) Gao, Y.; Robertson, G. P.; Guiver, M. D.; Jian, X. J. Polym. Sci.,
Chem. 1998, 36, 1441. Part A: Polym. Chem. 2003, 41, 497.
(51) Kerres, J.; Cui, W.; Junginger, M. J. Membr. Sci. 1998, 139, 227. (94) Xiao, G.; Sun, G.; Yan, D.; Zhu, P.; Tao, P. Polymer 2002, 43,
(52) Lafitte, B.; Karlsson, L. E.; Jannasch, P. Macromol. Rapid 5335.
Commun. 2002, 23, 896. (95) Powers, E. J.; Serad, G. A. High Performance Polymers: Their
(53) Kim, Y. S.; Wang, F.; Hickner, M.; McCartney, S.; Hong, Y. T.; Origin and Development; Elsevier: Amsterdam, 1986; p 355.
Zawodzinski, T. A.; McGrath, J. E. J. Polym. Sci., Part B: Polym. (96) Asensio, J. A.; Borros, S.; Gomez-Romero, P. J. Polym. Sci., Part
Phys. 2003, 41, 2816. A: Polym. Chem. 2002, 40, 3703.
(54) Robeson, L. M.; Matzner, M., of Union Carbide, Inc., 1983; U.S. (97) (a) Wainright, J. S.; Wang, J. T.; Weng, D.; Savinell, R. F.; Litt,
Patent 4,380,598. M. J. Electrochem. Soc. 1995, 142, L121. (b) Xiao, L.; Zhang,
(55) Ueda, M.; Toyota, H.; Ochi, T.; Sugiyama, J.; Yonetake, K.; H.; Choe, E. W.; Scanlon, E.; Ramanathan, L. S.; Benicewicz, B.
Masuko, T.; Teramoto, T. J. Polym. Sci., Polym. Chem. Ed. 1993, C. Prepr. Symp.sAm. Chem. Soc., Div. Fuel Chem. 2003, 48,
31, 853. 447. (c) Benicewicz, B. ACS Polymer Division Asilomar Confer-
(56) (a) Wang, F.; Hickner, M.; Kim, Y. S.; Zawodzinski, T. A.; ence, February, 2003.
McGrath, J. E. J. Membr. Sci. 2002, 197, 231. (b) Wang, F.; (98) Glipa, X.; El Haddad, M.; Jones, D. J.; Roziere, J. Solid State
Hickner, M.; Ji, Q.; Harrison, W.; Mecham, J.; Zawodzinski, T. Ionics 1997, 97, 323.
A.; McGrath, J. E. Macromol. Symp. 2001, 175, 387. (99) Gieselman, M. B.; Reynolds, J. R. Macromolecules 1992, 25,
(57) Harrison, W.; Wang, F.; Mecham, J. B.; Bhanu, V.; Hill, M.; Kim, 4832.
Y. S.; McGrath, J. E. J. Polym. Sci., Part A: Polym. Chem. 2003, (100) Kim, S.; Cameron, D. A.; Lee, Y.; Reynolds, J. R.; Savage, C. R.
41, 2264. J. Polym. Sci., Part A: Polym. Chem. 1996, 34, 481.
(58) Wiles, K. B.; Bhanu, V. A.; Wang, F.; McGrath, J. E. Polym. (101) Dang, T. D.; Bai, S. J.; Heberer, D. P.; Arnold, F. E.; Spry, R. J.
Prepr. 2002, 43, 993. J. Polym. Sci., Part B: Polym. Phys. 1993, 31, 1941.
(102) Einsla, B. R.; Kim, Y. J.; Tchatchoua, C.; McGrath, J. E. Polym. Y. L.; Ingat’ev, N. V.; Kondratenko, N. V.; Volkonskii, A. Y.;
Prepr. 2003, 44, 645. Vlasov, V. M.; Notario, R.; Maria, P. C. J. Am. Chem. Soc. 1994,
(103) Sakaguchi, Y. Functional Condensation Polymers; Kluwer: New 116, 3047. (b) Paddison, S. J.; Pratt, L. R.; Zawodzinski, T.;
York, 2002. Reagor, D. W. Fluid Phase Equilib. 1998, 150-151, 235. (c)
(104) Poppe, D.; Frey, H.; Kreuer, K. D.; Heinzel, A.; Mulhaupt, R. Eikerling, M.; Paddison, S. J.; Zawodzinski, T. A. J. New Mater.
Macromolecules 2002, 35, 7936. Electrochem. Syst. 2002, 5, 15. (d) Hoffman, M. A.; Ambler, C.
(105) Allcock, H. R.; Fitzpatrick, R. J.; Salvati, L. Chem. Mater. 1991, M.; Maher, A. E.; Chalkova, E.; Zhou, X. Y.; Lvov, S. N.; Allcock,
3, 1120. H. R. Macromolecules 2002, 35, 6490. (e) Ford, L. A.; Smith, D.
(106) Blonsky, P. M.; Shriver, D. F.; Austin, P.; Allcock, H. R. J. Am. W., Jr.; DesMarteau, D. D. Polym. Mater. Sci. Eng. 2000, 83,
Chem. Soc. 1984, 106, 6854. 10.
(107) Allcock, H. R. Acc. Chem. Res. 1979, 12, 351. (118) Savett, S. C.; Atkins, J. R.; Sides, C. R.; Harris, J. L.; Thomas,
(108) Allcock, H. R.; Klingenberg, E. H.; Welker, M. F. Macromolecules B. H.; Creager, S. E.; Pennington, W. T.; DesMarteau, D. D. J.
1993, 26, 5512. Electrochem. Soc. 2002, 149, A1527.
(109) Wycisk, R.; Pintauro, P. N. J. Membr. Sci. 1996, 119, 155.
(110) Tang, H.; Pintauro, P. N.; Guo, Q.; O’Connor, S. J. Appl. Polym. (119) Hofmann, M. A.; Ambler, C. M.; Maher, A. E.; Chalkova, E.;
Sci. 1999, 71, 387. Zhou, X. Y.; Lvov, S. N.; Allcock, H. R. Macromolecules 2002,
(111) Guo, Q.; Pintauro, P. N.; Tang, H.; O’Connor, S. J. Membr. Sci. 35, 6490.
1999, 154, 175. (120) Cho, C. G.; Kim, Y. S.; Hill, M. L.; McGrath, J. E. Polym. Prepr.
(112) Carter, R.; Wycisk, R.; Pintauro, P. N. Pre-Print Archives 2003, 44, 649.
American Institute of Chemical Engineers, [Spring National (121) Kreuer, K. D.; Fuchs, A.; Ise, M.; Spaeth, M.; Maier, J. Electro-
Meeting], New Orleans, LA, United States, Mar 11-14, 2002, p chim. Acta 1998, 43, 1281.
2441. (122) Schuster, M. F. H.; Meyer, W. H.; Wegner, G.; Herz, H. G.; Ise,
(113) Miyatake, K.; Hay, A. S. J. Polym. Sci., Part A: Polym. Chem. M.; Schuster, M.; Kreuer, K. D.; Maier, J. Solid State Ionics
2001, 39, 1854. 2001, 145, 85.
(114) Miyatake, K.; Hay, A. S. J. Polym. Sci., Part A: Polym. Chem.
2001, 39, 3770. (123) Herz, H. G.; Kreuer, K. D.; Maier, J.; Scharfenberger, G.;
(115) Meng, Y. Z.; Tjong, S. C.; Hay, A. S.; Wang, S. J. Eur. Polym. J. Schuster, M. F. H.; Meyer, W. H. Electrochim. Acta 2003, 48,
2003, 39, 627. 2165.
(116) (a) Allcock, H. R.; Hofmann, M. A.; Ambler, C. M.; Lvov, S. N.; (124) Wang, F.; Glass, T.; Li, X.; Hickner, M.; Kim, Y. S.; McGrath, J.
Zhou, X. Y.; Chalkova, E.; Weston, J. J. Membr. Sci. 2002, 201, E. Polym. Prepr. 2002, 43, 492.
47. (b) Allcock, H. R.; Hofmann, M. A.; Ambler, C. M.; Morford, (125) Kim, Y. S.; Dong, L.; Hickner, M.; Glass, T. E.; Webb, V.;
R. V. Macromolecules 2002, 35, 3484. McGrath, J. E. Macromolecules 2003, 36, 6181.
(117) (a) Koppel, I. A.; Taft, R. W.; Anvia, F.; Zhu, S. Z.; Hu, L. Q.;
Sung, K. S.; DesMarteau, D. D.; Yagupolskii, L. M.; Yagupolskii, CR020711A
Chem. Rev. 2004, 104, 4535−4585 4535
State of Understanding of Nafion

Kenneth A. Mauritz* and Robert B. Moore*
Department of Polymer Science, The University of Southern Mississippi, 118 College Drive #10076, Hattiesburg, Mississippi 39406-0001
Received July 19, 2004
Contents Common at the time of this writing are Nafion 117

films. The designation “117” refers to a film having
1. Introduction 4535 1100 EW and a nominal thickness of 0.007 in.,
2. Morphological Characterization Using X-rays and 4538 although 115 and 112 films have also been available.
Neutrons Early-reported studies involved 1200 EW samples as
2.1. Early Studies of Nafion Morphology 4538 well as special experimental varieties, some being
2.2. Scattering from PFSI Solutions and Recast 4543 rather thin. The equivalent weight is related to the
Films property more often seen in the field of conventional
2.3. Evaluation of Nafion Morphology through 4545 ion exchange resins, namely the ion exchange capac-
Studies of Oriented Membranes ity (IEC), by the equation IEC ) 1000/EW.
2.4. Current Models for the Morphology of Nafion 4548
The mention of the molecular weight of high
2.5. The State of Understanding from X-ray and 4554 equivalent weight (EW > 1000 g‚mol-1) Nafion is
Neutron Investigations
almost absent in the literature, although the range
3. Microscopy Studies 4555 105-106 Da has been mentioned. As this polymer
4. The Nature of Water, Other Solvents, and Ions 4559 does not form true solutions, the common methods
in Nafion of light scattering and gel permeation chromatogra-
5. Mechanical Properties 4575 phy cannot be used to determine molecular weight
6. Molecular Simulation of Structure and Properties 4576 as well as the size and shape of isolated, truly
7. Conclusions 4581 dissolved molecules. Studies of the structure of this
8. Acknowledgment 4582 polymer in solvent (albeit not a true solution) will
9. Literature References 4582 be mentioned in the scattering section of this review.
It should be noted that Curtin et al. performed size
exclusion chromatography determinations of the mo-
1. Introduction lecular weight distribution in Nafion aqueous disper-
sions after they were heated to high temperatures
Nafion ionomers were developed and are produced
(230, 250, and 270 °C).1 Before heating, there was a
by the E. I. DuPont Company. These materials are
high molecular weight shoulder on a bimodal distri-
generated by copolymerization of a perfluorinated
bution, due to molecular aggregates, but this shoul-
vinyl ether comonomer with tetrafluoroethylene (TFE),
der disappeared upon heating, which indicated that
resulting in the chemical structure given below.
the aggregates were disrupted. The peaks for the
monomodal distribution for the heated samples were
all located at molecular weights slightly higher than
105 g‚mol-1. Also, light scattering experiments re-
vealed that the radius of gyration had a linear
dependence on the molar mass of the aggregates,
which suggests that the particles are in the form of
Equivalent weight (EW) is the number of grams of rods or ribbons, or at least some elongated structure.
dry Nafion per mole of sulfonic acid groups when the
Nafion ionomers are usually derived from the
material is in the acid form. This is an average EW
thermoplastic -SO2F precursor form that can be
in the sense that the comonomer sequence distribu-
extruded into sheets of required thickness. Strong
tion (that is usually unknown to the investigator and
interactions between the ionic groups are an obstacle
largely unreported) gives a distribution in m in this
to melt processing. This precursor does not possess
formula. EW can be ascertained by acid-base titra-
tion, by analysis of atomic sulfur, and by FT-IR the clustered morphology that will be of great concern
spectroscopy. The relationship between EW and m in this article but does possess Teflon-like crystal-
is EW ) 100m + 446 so that, for example, the side linity which persists when the sulfonyl fluoride form
chains are separated by around 14 CF2 units in a is converted to, for example, the K+ form by reacting
membrane of 1100 EW. it with KOH in water and DMSO. Thereafter, the
-SO3H form is achieved by soaking the film in a
sufficiently concentrated aqueous acid solution. Ex-
* Address correspondence to either author. Phone: 601-266-5595/
4480. Fax: 601-266-5635. E-mail: Kenneth.Mauritz@usm.edu; trusion of the sulfonyl fluoride precursor can cause
RBMoore@usm.edu. microstructural orientation in the machine direction,
10.1021/cr0207123 CCC: $48.50 © 2004 American Chemical Society
4536 Chemical Reviews, 2004, Vol. 104, No. 10 Mauritz and Moore
Dr. Kenneth A. Mauritz has been a Professor in the School of Polymers Dr. Robert B. Moore is currently a Professor in the School of Polymers
and High Performance Materials at the University of Southern Mississippi and High Performance Materials at the University of Southern Mississippi.
since 1984. He received a Ph.D. in Macromolecular Science at Case He received his Ph.D. in Analytical Chemistry from Texas A&M University
Western Reserve University in 1975. After a postdoctoral assignment at in 1988 under the direction of Professor Charles R. Martin. His graduate
Case Western Reserve, he joined Diamond Shamrock Corporation as a research was focused on the chemical and morphological investigations
Research Chemist in 1976. There, he began his long and continuous of perfluorosulfonate ionomers. He then joined the group of Professor
involvement with Nafion materials, mainly in the area of the molecular Adi Eisenberg at McGill University as a postdoctoral fellow and continued
structure−property characterization and modeling as related to membrane his work on the morphological characteristics of random ionomers using
performance optimization in chlor-alkali electrochemical cells for the small-angle X-ray scattering methods. In 1991, he joined the faculty in
production of chlorine and caustic. Later, at the University of Southern the Department of Polymer Science at the University of Southern
Mississippi, he produced some of the first inorganic/Nafion nanocomposite Mississippi. His research interests include structure−property relationships
materials based on membrane−in situ sol−gel processes for inorganic in ion-containing polymers, random ionomers, semicrystalline ionomers,
alkoxides. Mauritz has collaborated with the coauthor of this review, R. ionomer membranes, and blends. A key consideration in this work, which
B. Moore, in a number of projects involving Nafion materials. He has has spanned from 1983 to the present, has involved the link between
contributed numerous papers, reviews, and presentations on the topic of ionic aggregation and crystallization in semicrystalline ionomers, specifically
Nafion. He has coedited and contributed to the book Ionomers: Synthesis, Nafion. He has presented numerous papers and invited lectures on the
Structure, Properties and Applications and was cochairperson of the 2003 topic of morphological manipulations of Nafion through the development
Gordon Research Conference on Ion Containing Polymers. Other research of novel processing procedures and blends. He is currently an officer in
interests include self-assembled, nanostructured organic−inorganic materi- the Division of Polymer Chemistry of the American Chemical Society and
als via domain-targeted inorganic syntheses in phase separated ionomers, is the cochair of the 2005 Gordon Research Conference on Ion Containing
microscopy, spectroscopy, dielectric relaxation/electrical impedance analy- Polymers.
sis, and the thermal, dynamic mechanical, and diffusion properties of
polymers. in the harsh chemical environment. The membranes
currently used in this application are of a sulfonate/
and this can be seen in the small-angle X-ray scat- carboxylate bilayer variety.
tering peak that is attributed to the clustered mor- A lesser number of papers have appeared regarding
phology. Moreover, this orientation can affect the the carboxylate version of Nafion, and the works of
swelling and electrical conductance properties (ani- Yeager et al.,7,8 Seko et al.,9 and Perusich et al.10 are
sotropy) of the ionomer form. of special note. Nafion-like carboxylate materials
The earliest concerted effort in the research and have also been reported.11 These similar materials
development of Nafion perfluorosulfonate ionomers will not be discussed here, as the exclusive emphasis
was directed toward their use as a permselective is on the stronger-acid sulfonated versions because
membrane separator in electrochemical cells used in this critical review is within a collection of articles
the large scale industrial production of NaOH, KOH, in which perfluorinated ionomers are considered
and Cl2. In short, the membrane in this application, within the context of fuel cells.
in addition to keeping Cl2 and H2 gases separated, The DuPont Nafion materials, both sulfonate and
prevents the unfavorable back migration of hydrated carboxylate varieties, are not entirely unique, as
OH- ions from the catholyte (concentrated aqueous similar perfluorinated ionomers have been developed
NaOH or KOH) chamber, while allowing for the by others such as the Asahi Chemical Company
transport of hydrated Na+ ions from the anolyte (commercial name: Aciplex) and the Asahi Glass
chamber in which is aqueous NaCl. Company (commercial name: Flemion). The comono-
Early experimental versions of Nafion within the mer chemical structures of and further information
context of chlor-alkali cells2 consisted of SO2F precur- on these materials are given in the recent review
sor forms that were first reacted on only one side with article by Doyle and Rajendvan.12 Now commercially
ethylenediamine (EDA) before the conversion of the unavailable, but once considered a viable alternative,
remainder of the membrane to the sulfonate form the Dow Chemical Company developed a somewhat
took place. The result was a well-defined stratum of similar perfluorinated ionomer that resembled the
sulfonamide cross-links, that were formed upon heat- sulfonate form of Nafion except that the side chain
ing after reaction, that served to reduce swelling at of the former is shorter and contains one ether
the catholyte interface, which, in turn, reduced OH- oxygen, rather than two ether oxygens, that is, -O-
back migration.3-6 However, these EDA-modified CF2-CF2-SO3H.13,14
membranes proved inadequate in chlor-alkali cells The greatest interest in Nafion in recent years
due to the chemical degradation of these cross-links derives from its consideration as a proton conducting
State of Understanding of Nafion Chemical Reviews, 2004, Vol. 104, No. 10 4537
membrane in fuel cells. It is clear that the tuning of In the years following the introduction of the model
these materials for optimum performance requires of Gierke et al., more extensive structure-property
a detailed knowledge of chemical microstructure and studies have been conducted and, in many reports,
nanoscale morphology. In particular, proton conduc- alternate morphologies have been proposed. Through
tivity, water management, relative affinity of metha- the use of state-of-the-art scattering methods, our
nol and water in direct methanol fuel cells, hydration understanding of the morphology of Nafion is evolv-
stability at high temperatures, electro-osmotic drag, ing to include a more definitive picture of the nature
and mechanical, thermal, and oxidative stability are of ionic aggregation in this polymer. It is a goal of
important properties that must be controlled in the this article to examine these studies, point out the
rational design of these membranes. This is a chal- salient features, and identify the elements that are
lenge for Nafion materials in which the possible in common.
chemical variations are rather limited. And, of course, Also, among the earliest concepts that were set
all of these objectives must be achieved while main- forth regarding microstructure are those of Yeager
taining low cost for this perfluorinated ionomer in and Steck, who proposed a three-phase model that
the vast consumer market as well as in military was significantly different from that of Gierke et al.
applications. While a number of alternate polymer based on their studies of the diffusion of various
membranes, including nonfluorinated types, have ions.18 As compared with the model of Gierke et al.,
been developed, Nafion is still considered the bench- the clusters do not have a strict geometrical definition
mark material against which most results are com- (spherical inverted micelles connected by cylindrical
pared. pores) and their geometrical distribution has a lower
degree of order. Most importantly, there are transi-
With respect to the morphological characterization
tional interphases between hydrophilic and hydro-
of Nafion, it is important to note that the wealth of
philic regions, a concept that is becoming increasingly
information gathered over the years using a variety
accepted.
of scattering methods is inherently indirect and often
limited by the necessity of employing rather simple It must be emphasized that this article is not
morphological models that involve specific assump- meant to be a listing of the very large number of
tions of structure. While mentioned frequently publications dealing with Nafion. To be sure, the
throughout this review, the cluster-network model literature since the late 1970s, when the main thrust
of Gierke et al. deserves special mention (from a for the development of these materials was mem-
brane chlor-alkali cell technology improvement and,
historical perspective) as a “starting point” in an
to a lesser degree, water electrolyzers, has prolifer-
introduction to this now-vast topic. This model has
ated. It was estimated (DuPont library), based on a
endured for many years as a conceptual basis for
coarse literature search that was performed a few
rationalizing the properties of Nafion membranes,
years prior to this writing, that there were ap-
especially ion and water transport and ion perm- proximately 33 000 papers, patents, and so forth
selectivity.15-17 It is presumed, based on small-angle dealing with Nafion, and the number is growing. In
X-ray scattering (SAXS) studies and several assump- part, to limit the scope of this review, the focus is on
tions, that there are ∼40 Å -in-diameter clusters of selected studies that have addressed the most fun-
sulfonate-ended perfluoroalkyl ether groups that are damental aspects of the structure and properties of
organized as inverted micelles and arranged on a this polymer. Given this limited goal, the impor-
lattice. These micelles are connected by proposed tant issues of thermal and electrochemical stability,
pores or channels that are ∼10 Å in size. These electro-osmotic drag, methanol crossover, gas per-
-SO3--coated channels were invoked to account for meation, and catalyst incorporation have been omit-
intercluster ion hopping of positive charge species but ted, although numerous reports on these topics are
rejection of negative ions, such as OH-, as in the case found throughout the literature.
of chlor-alkali membrane cells. Chemical degradation in a fuel cell environment
A few things should be appreciated concerning this is also an important issue that must be addressed
historical model. First, it was based on the limited and so will be mentioned briefly here. It is thought
structure-property information that was available that generated peroxide radicals attack polymer
at the time, namely the existence of a single SAXS end groups having H-containing terminal bonds
peak and the behavior of this peak with water (-CF2COOH) that are formed during processing.
swelling as well as the observed selective permeation H2O2, formed by reactions between oxygen and
of Na+ over OH- ions in membrane chlor-alkali cells. hydrogen, then decomposes, giving •OH or •OOH
Wide-angle X-ray diffraction (WAXD) investigations radicals that attack the H-containing terminal bonds,
determined limited poly(tetrafluoroethylene) (PTFE)- and this initiates chemical decomposition. Mem-
like crystallinity associated with the perfluorocarbon branes become thinner, and fluoride ions are de-
backbone, although this structural detail was not tected in product water. The reader is directed toward
factored into the Gierke model. In the sense that it the article by Curtin et al., for the details of this
does not assemble the phase separated morphology mechanism and characterization of the degradation
from basic chemical structure as, say, driven by a free process.1
energy minimization that incorporates molecular Other reviews of the literature on Nafion and
interactions as in a molecular simulation procedure, similar perfluorinated ionomers have appeared over
the model cannot be considered to be predictive but the years. The early book (1982) by Eisenberg and
is calibrated to experimental data. Yeager, Perfluorinated Ionomer Membranes, remains
a useful series of monographs dealing with early data are of limited use. Thus, the investigator is
work in the area.19 Later (1996), Heitner-Wirguin forced to compare the experimentally determined
assembled a comprehensive review that contains scattered profiles with that predicted by necessarily
topics not included in the document presented here.20 simple models that must be assumed for the struc-
Most recent is the compilation of Doyle and Rajen- ture. To further complicate matters, very low angle
dran entitled Perfluorinated Membranes that con- scattering information characteristic of long-range
tains an excellent history of the evolution of this dimensions is convoluted by overlapping signals
material that begins with the discovery of PTFE by associated with the spatial distribution of crystallites
Plunkett in 1938.12 (i.e., a weak maximum or shoulder often attributed
Also, discussions of a number of applications of to the long spacing between crystallites) and long-
Nafion are not included in this document and are, range heterogeneities in electron density associated
at most, mentioned within the context of a particular with the disordered distribution of ionic domains (i.e.,
study of fundamental properties. A number of these a small-angle upturn in intensity commonly observed
systems are simply proposed rather than in actual with ionomers). This morphological complexity pre-
commercial applications. Membranes in fuel cells, cludes the general application of simple two-phase
electrochemical energy storage systems, chlor-alkali models and thus further limits the useful angular
cells, water electrolyzers, Donnan dialysis cells, elec- range for precise model comparisons.
trochromic devices, and sensors, including ion selec- In considering the complicating factors associated
tive electrodes, and the use of these membranes as a with extracting detailed morphological information
strong acid catalyst can be found in the above- from scattering and diffraction data for Nafion, it is
mentioned reviews. not surprising that a number of morphological mod-
It should be mentioned that, because the micro- els, constituting significantly different structural
structure and properties of this ionomer are known features, have been proposed and are still the focus
to be quite sensitive to history and the details of of considerable debate. Since the vast majority of the
preparation of samples for analysis, an effort has applications of Nafion involve the hydrated or solvent
been made in this article, where possible, to delineate swollen state and current processing methods for
these important parameters in each described study membrane formation often involve solvent casting,
so that the results of different investigations can be considerable attention has been devoted to the influ-
meaningfully compared. ence of swelling solvents (specifically water) on the
Finally, the scope of the discussed literature is not characteristic morphological features of perfluoro-
confined to recent contributions but extends back to sulfonate ionomers (PFSIs). As noted above, scatter-
the 1970s because a number of excellent studies have ing and diffraction studies of these polymers gener-
not been repeated since then and often do not appear ally yield limited information; however, many groups
on Internet search engines. have demonstrated success in evaluating the influ-
ence of solvent swelling on the scattering profiles for
2. Morphological Characterization Using X-rays comparison and validation of morphological models.
and Neutrons Furthermore, recent detailed studies of the morpho-
logical continuum from the dry state through the
Over the last 30 years, a wealth of morphological swollen “gel” state to the “dissolved” (or suspended)
information from numerous scattering and diffraction state, and visa versa, have begun to address impor-
studies of Nafion has been obtained and reported in tant questions related to the organization of ionic and
an attempt to precisely define the molecular/super- crystalline domains and the impact of solvent-casting
molecular organization of perfluorinated ionomers in on the morphology of perfluorosulfonate ionomers. In
a variety of physical states. However, while the this section of the review, the prominent morphologi-
quality and quantity of data from state-of-the-art cal studies of Nafion in a variety of physical states
instrumentation, facilities, and methods has in- will be reviewed with a historical perspective, includ-
creased, a universally accepted morphological model ing a comparative evaluation of the latest state of
for the solid-state structure of Nafion has yet to be the debate over the key morphological models.
defined. The source of the ongoing debate over the
morphology of Nafion stems from the fact that this 2.1. Early Studies of Nafion Morphology
unique ionomer has a random chemical structure
that is capable of organizing in the complex formation During the 1970s and early 1980s, general con-
of ionic and crystalline domains with a significant cepts and morphological details associated with
distribution in dimensions over a wide range of ionic aggregation in the class of lightly ionized
length scales. Moreover, quantitative morphological polymers, referred to as ionomers, were actively
information is limited by the fact that Nafion yields debated in the literature.21,22 The fundamental in-
only a single, broad scattering maximum attributed formation from these studies, based in large part on
to characteristic dimensions within the ionic domains small-angle X-ray scattering (SAXS) studies of dry,
and the diffraction reflections of the PTFE-like crys- polyethylene-23 and polystyrene-based ionomers,24,25
tallites are broad and relatively weak due to low formed the foundation for the morphological charac-
degrees of crystallinity. terization of Nafion.26 By the late 1970s, experimen-
Unfortunately, the small-angle scattering tech- tal evidence for ionic aggregation in Nafion from
niques used in the investigations of Nafion morphol- small-angle scattering data was emerging,27-30 and
ogy generally probe but a small region of reciprocal this new information subsequently lead to extensions
space and Fourier inversion methods of analyzing the of the prevailing models for the structure of ionomers
to the interpretation of ionic domain morphology in

the perfluorosulfonate ionomer systems.
Using SAXS and wide-angle X-ray diffraction
(WAXD), Gierke and co-workers16,17,31,32 examined
and compared the morphological features of Nafion,
having a range of equivalent weights, in the unhy-
drolyzed sulfonyl fluoride precursor form, the hydro-
lyzed sulfonic acid form, and the neutralized metal
sulfonate form. For the unhydrolyzed precursor, a low
angle SAXS maximum near 0.5° 2θ and a diffraction Figure 1. Cluster-network model for the morphology of
peak at 18° 2θ (superimposed on a broad amorphous hydrated Nafion. (Adapted with permission from ref 16.
halo from 10 to 20° 2θ) were observed for samples Copyright 1983 Elsevier.)
having equivalent weights (EW, the grams of polymer
per equivalent of sulfonate groups) ranging from 1100 the exchange sites per cluster, and the number of
to 1800 g/equiv. Since these scattering and diffraction water molecules per exchange site increased linearly
peaks were found to increase in intensity with with water content. While the linear relationships
equivalent weight (i.e., with an increase in the were not addressed in this study, the growth of
statistical length of crystallizable PTFE chain seg- clusters with increasing water content was proposed
ments between side chains) and were also found to to occur by a combination of expansion in cluster size
disappear at temperatures near the melting point of and a redistribution of the sulfonate sites to yield
PTFE, these features were attributed to crystalline fewer clusters in the fully hydrated material.17 Al-
organization within the fluorocarbon matrix. With though this original, yet phenomenological, cluster-
the hydrolyzed form of Nafion, an additional scat- network model has received significant acceptance
tering peak was observed at ∼1.6° 2θ, corresponding in the literature and has served as the foundation
to a Bragg spacing of 3-5 nm, characteristic of a for numerous studies of morphology-property rela-
system containing ionic clusters within a semicrys- tionships of Nafion, it has also been vigorously
talline matrix. It is important to note here that the debated with respect to alternative morphologies that
term “ionic clusters” (used conventionally in the are also quantitatively consistent with a wide range
perfluorosulfonate ionomer literature) generally re- of scattering results. Specifically, current studies
fers to nano-phase separated, hydrated ionic ag- focused on the morphological and dimensional changes
gregates, and this term differs from the current that occur with variations in solvent swelling have
definition used for other dry ionomers (e.g., sul- raised critical questions related to the shape and
fonated polystyrene) to designate regions of closely spatial distribution of the ionic clusters within the
spaced ionic multiplets.33 semicrystalline matrix of Nafion.
In contrast to the SAXS and WAXD behavior Roche and co-workers used SAXS and small-angle
observed for the crystalline features, the ionomer neutron scattering (SANS) to study the morphology
peak was shown to increase in intensity and shift to of 1200 EW Nafion membranes in the acid and
lower angles with a decrease in equivalent weight. neutralized forms with a range of water contents.34,35
Furthermore, for a given equivalent weight, this peak In the as-received state, Nafion was found to have
was found to shift to lower angles and increase in three contrast regions (i.e., a three-phase system) in
intensity with increasing water content. Based on the angular range probed by SAXS and SANS
these findings, and considering the three most preva- consisting of a crystalline phase, ionic clusters, and
lent models for the morphology of ionomers at the an inhomogeneous matrix phase. By quenching the
time, including a model of spherical clusters on a Na+ form ionomer from 330 °C, the sample was
paracrystalline lattice, a core-shell model, and a rendered amorphous, as confirmed by WAXD, and
lamellar model, Gierke and co-workers concluded the low angle scattering maximum, corresponding to
that the water-swollen morphology of Nafion was a Bragg spacing of ∼18 nm, was found to disappear.
best described by a model of ionic clusters that were From these results, the 18 nm peak, also observed
approximately spherical in shape with an inverted by Gierke,17 was attributed to interference between
micellar structure.17,31 In consideration of the high crystalline structures. In considering the lamellar
ionic permselectivity and the requirement of a per- periodicity (i.e., the long spacing) of ∼100 nm ob-
colation pathway for ionic transport in Nafion mem- served in PTFE, the fact that copolymerization is
branes, the spherical ionic clusters were further known to decrease the lamellar thickness, and the
proposed to be interconnected by narrow channels, average spacing of about 1.5 nm between side chains
constituting a morphology referred to as the cluster- in 1200 EW Nafion, it was concluded either that a
network model,16 shown in Figure 1. nonrandom distribution of side chains exists in some
Based on this new model for the morphology of portion of the material or that the side groups were
Nafion, the dimensional variations of the scattering somehow included in the crystalline structure.34
entities with water content were used in simple Using contrast matching methods of SANS,35 Nafion
space-filling calculations to estimate the cluster was swollen with various mixtures of H2O and D2O
diameter, the number of sulfonate sites per cluster, in order to highlight the scattering features of the
and the number of water molecules per cluster.16,17 water swollen ionic domains. Furthermore, quenched
The results of these model calculations showed that, samples were used in these studies to eliminate the
for a given equivalent weight, the cluster diameter, scattering features attributed to the crystalline do-
mains. Roche et al. found that these quenched

samples exhibited just the ionomer peak at q ) ∼1.2
nm-1 (q is the scattering vector equal to (4π/λ) sin θ,
where λ is the wavelength of the scattered radiation
and θ is half the scattering angle, 2θ) and a low angle
upturn in intensity at q < 0.5 nm-1. In agreement
with the behavior expected from a simple, two-phase
system (i.e., a hydrophobic phase and a hydrophilic
phase), the normalized scattering intensities of the
water swollen membranes were found to vary linearly
with relative fractions of D2O in H2O. This observa-
tion indicated that both the zeroth-order scattering Figure 2. Two morphological models used to describe the
at low q-values and the ionomer peak arise from the origin of the ionic SAXS maximum observed for Nafion:
(a) the modified hard-sphere model depicting interparticle
same scattering length density differences and could scattering and (b) the depleted-zone core-shell model
be attributed to the same morphological features (i.e., depicting intraparticle scattering. (Adapted with permis-
clusters).35 To account for the larger dimensions sion from ref 36. Copyright 1981 American Chemical
associated with the zeroth-order scattering, an in- Society.)
homogeneous distribution of clusters within the
In the analysis of the low angle scattering maxi-
fluorocarbon matrix was proposed. Moreover, these
mum at s ≈ 0.07 nm-1, Fujimura et al. found that
results demonstrated that the zeroth-order scattering the intensity was greatest for the carboxylated and
could not be attributed to impurities, as previously nonionic forms of Nafion, relative to the regular
proposed for other ionomer systems.35 sulfonated ionomer.36 In agreement with the work of
In agreement with the work of Gierke et al.,17 Gierke17 et al., this low angle maximum also in-
Roche and co-workers found that the intensity of the creased in intensity with increasing equivalent weight.
ionomer peak increased and the position of the Similarly, a detailed analysis of the crystallinity in
scattering maximum shifted to lower angles with these materials by WAXD indicated that 1100 EW
increasing water content.34 A Porod analysis of the Nafion was 23% and 18-14% crystalline in the
SAXS data over a wide range of scattering angles also nonionic and carboxylated forms, respectively, while
indicated that the majority of water molecules in the it was 12-3% crystalline in the sulfonated forms.
swollen ionomer were phase separated with sharp With varying equivalent weight from 1100 EW to
phase boundaries. Despite the close agreement in 1500 EW, the degree of crystallinity increased from
scattering results between both groups, Roche et al. 12% to 22%.36 Based on the quantitative correlation
argued that an interparticle scattering concept as- of this SAXS and WAXD information and on the
sociated with Gierke’s cluster-network model was behavior observed with uniaxial deformation, the
unlikely due to the lack of evidence for a required origin of the low angle scattering maximum at s ≈
potential giving rise to ordering of a paracrystalline 0.07 nm-1 was attributed to an average spacing
lattice.34 As an alternative, the origin of the ionomer between crystalline lamellar platelets.36
peak was attributed to an intraparticle scattering The effect of solvent swelling (with water) on the
model for which the scattering maximum corre- characteristic dimensions of the ionic clusters associ-
sponded to characteristic distances between struc- ated with the scattering maximum at s ≈ 0.3 nm-1
tural elements inside the ionic particles.34,35 While was also evaluated for 1100 EW Nafion. With the
this intraparticle scattering model was consistent cesium sulfonate form of the ionomer, ionic clusters
with the morphology proposed for other ionomers, a were shown to exist in the dry state.36 With increas-
detailed quantitative analysis in support of this ing water content, the cluster dimensions deduced
model was not provided. from the scattering maxima were shown to increase.37
Furthermore, over a range of water contents, the
Fujimura et al. reported the results of detailed microscopic degree of swelling, as determined by
SAXS and WAXD experiments in a series of two SAXS, was found to be much greater than the
papers focused on defining the morphological origins macroscopic degree of swelling.37 A similar correla-
of the two scattering maxima36 and evaluating vari- tion of microscopic versus macroscopic dimensions
ous scattering models for the origin of the ionomer was observed for Nafion subjected to varying degrees
peak.37 Samples characterized in this work included of uniaxial deformation.37 Based on these observa-
Nafion with a range of equivalent weights (1100 EW tions, Fujimura et al. concluded that the scattering
through 1500 EW) in the acid, Na+, and Cs+ forms. behavior of Nafion is best described by an intrapar-
To further investigate the effect of ionic interactions ticle core-shell model (Figure 2b) similar to that
on the morphology of Nafion,17,34 the sulfonate groups proposed by MacKnight23 and co-workers38 rather
of these materials were also chemically derivatized than an interparticle, two-phase, hard sphere model
to yield carboxylated and nonionic, sulfonyl chloride proposed by Cooper et al.39 (Figure 2a). In this
analogues to Nafion. In agreement with the previous particular core-shell model an ion-rich core is pro-
scattering studies of Nafion, two scattering maxima posed to be surrounded by an ion-poor shell (i.e., a
at s ≈ 0.07 and 0.3 nm-1 (the scattering vector s is depleted zone)37 that is composed mainly of per-
defined as s ) (2 sin θ)/λ) were observed and at- fluorocarbon chains. These core-shell particles are
tributed to morphological features associated with dispersed in a matrix of fluorocarbon chains contain-
crystalline and ionic domains, respectively. ing nonclustered ions and multiplets, such that the
short-range order distance within the core-shell this proposed model consisted of a bilayer of hexago-
particle gives rise to the ionic scattering maximum. nally packed planar zigzag chains with the side
While the nonaffine swelling behavior observed by chains extending outward on both sides into the
Fujimura et al.37 has been reproduced by many others clusters. With the assumption that all comonomer
in subsequent studies and has been the focus of units were incorporated in a head-to-tail configura-
significant debate over the years, it is important to tion, this arrangement allowed for all side chains to
note that the argument against the interparticle hard be on the same side of the planar zigzag chains such
sphere model relies on the key assumptions that the that the crystallite thickness could exceed the aver-
clusters are uniformly distributed in space and that age separation distance between side chains. While
no redistribution of the ionic groups occurs during this model provides a simple illustration of the
swelling. However, the cluster-network model pro- packing of small crystallites between clusters, it was,
posed by Gierke17 clearly allows for cluster reorgani- however, acknowledged to involve “a very efficient
zation during swelling, and the low angle upturn in use of the polymer’s components”.41
scattering intensity observed by Roche et al.35 strongly One concern for this bilayer model arising from the
suggests an inhomogeneous distribution of clusters. diffraction data was noted to involve the width of the
With the recognition of the importance of these 100 reflection; the Scherrer analysis indicated that
considerations, Fujimura et al. attempted to quan- the size of the crystals normal to the chain axis was
titatively distinguish the two models by conducting on the order of 3.9 nm rather than the proposed 1
a thorough analysis of the scattering data with the nm bilayer thickness. In addition, a number of
aid of computer simulation.37 For both the core-shell further considerations raise important questions as
model and the two-phase hard sphere model, quan- to the validity of this model. For example, given the
titative expressions for the variations in the scatter- recognition that Nafion contains a heterogeneous
ing profiles with water uptake were derived from distribution of clusters35 (i.e., long-range heteroge-
standard scattering theories and fit to the experi- neities observed in the SAXS profiles at low angles),
mental data. Unfortunately, however, this analysis rather than a homogeneous network of clusters and
was inconclusive, as both models adequately fit the channels in an ordered paracrystalline lattice,17 it is
experimental data and were found to yield similar not necessary to impose the organization of crystals
effects on the variation of the profiles with water in the narrow space between nearest neighbor clus-
uptake and deformation. Moreover, the core-shell ters. Furthermore, Nafion is a statistical copolymer
model does not give a clear view of long-range that is likely to contain both short and long runs of
structure such as that which might be useful in TFE units between side chains. The longer runs could
understanding long-range properties such as molec- thus pack into small PTFE-like crystallites in regions
ular transport. In a more recent SAXS investigation, somewhat removed from the ionic domains. These
Miura and Yoshida40 also supported the core-shell crystallites, while few in number, would provide the
model but suggested (without quantitative support) three-dimensional order required to yield the ob-
that some degree of ionic redistribution with swelling served WAXD reflections and the high ∆Hf and Tm
in different solvents was possible due to differences values similar to those for PTFE as reported by
in internal stresses within the cluster morphology. Starkweather.41 More recent studies, discussed be-
Nevertheless, later investigators would challenge this low, provide strong evidence for this alternative
view in favor of the two-phase model. organization of crystallites and clusters in Nafion.
The organization of crystalline domains in Nafion In 1986, Kumar and Pineri published a report
was further investigated by Starkweather using focused on the interpretation of small-angle X-ray
X-ray diffraction analysis of oriented fiber samples.41 and neutron scattering data for perfluorosulfonated
The unit cell structure derived from the indexed ionomer membranes in order to elucidate the possible
diffraction patterns indicated that the crystallites in cluster morphology in PFSI membranes.42 This de-
Nafion were packed in a hexagonal lattice with true tailed study involving model calculations and com-
three-dimensional character and cell dimensions parisons of experimental scattering profiles to theo-
roughly equivalent to those of PTFE. Furthermore, retical fits was placed in context with the morphologi-
the positions of the principal 100, 101, and 002 cal models of Nafion, to date. Since the paracrystal-
reflections for the sulfonyl fluoride precursor were line lattice model (i.e., the basis for calculations from
essentially unchanged as the polymer was hydrolyzed the Gierke cluster-network model)17 was found to be
to the potassium sulfonate form. Peak width mea- inadequate to explain the observed scattering and
surements, following the Scherrer approach, were swelling data, and the depleted-zone core-shell
used to estimate minimum values for the crystallite model relied on questionable assumptions involving
size. This analysis suggested that the crystallite size the applicability of Bragg’s equation in estimating
along the c-axis was 4.4 nm (or 34 carbon atoms along microscopic versus macroscopic swelling and the
a planar zigzag conformation), which was noted to requirement of a constant number of clusters with
be significantly larger than the average of 15 CF2 swelling,36,37 an alternative model involving non-
groups between side chains in the copolymer. interacting hard spheres was chosen for this analysis.
Based on the fiber diffraction data and the mor- With this model, Kumar and Pineri found that the
phological constraints imposed by Geirke’s cluster- SAXS ionomer peak could be quantitatively at-
network model,17 Starkweather developed a model for tributed to the interference between clusters, assum-
the crystalline structure of perfluorosulfonate iono- ing them to be noninteracting hard spheres.42 Fur-
mers.41 Given a 1 nm wall space between clusters, thermore, the increase in intensity at very low angles
was accounted for by considering a previously ne- able fits over the q-range 0.2-2.6 nm-1. Based on the
glected intensity term for scattering from a dense excellent agreement between the theoretical and
group of particles and/or an intercluster potential experimental SANS curves, the authors concluded
arising from Coulombic interactions between clusters. that both the small-angle upturn and the ionomer
In contrast to Gierke’s model,17 which predicted a peak may be attributed to the existence and spatial
large increase in the number of ions per cluster, with distribution of ionic clusters. Using the values of D
increasing water content, parameters from these and R from the model fits (e.g., D ) 3.4 nm and R )
model calculations42 showed a moderate increase in 1.7 nm for the fully hydrated membrane), the specific
Np at low water uptake and a significant decrease in surface area per charge site on the cluster was found
Np at higher water contents. This behavior was to be relatively constant at 73 Å2. In agreement with
attributed to an increase in elastic forces on the previous studies that suggested a redistribution of
chains that effectively forces some ions out of the ions between clusters with swelling,17 the number of
existing clusters at high degrees of swelling, thus charge sites per cluster was calculated using space-
decreasing the Np. Since the ions forced out of filling arguments based on the D values and found
clusters with swelling are free to form other clusters, to increase from ∼25 to 45 with increasing hydration.
the average distance between clusters was argued to Despite the excellent fits with this local-order model,
be an invalid measure of the microscopic swelling however, the number of charge sites per cluster,
without some account for the change in the total directly calculated from the R values, was unrealisti-
number of clusters. For Nafion of different equivalent cally large. In addition, the fundamental requirement
weights, Kumar and Pineri also found that the of this model for an attractive potential between
number of ionic sites per cluster, and thus the cluster neighboring clusters in the formation of the local-
radius, only moderately decreased with increasing order structure is consistent with the suggestion by
equivalent weight.42 This observation, as confirmed Kumar and Pineri42 that the small-angle upturn
by Manley et al.43 in a later study using a similar could be attributed to such a potential; however, this
model analysis, was noted to again show a distinct plausible correlation was not addressed in this analy-
discrepancy with the results of Gierke, which indi- sis.
cated a sharp decrease in Np with increasing equiva- Lee et al.45 utilized a novel combination of small-
lent weight. and wide-angle neutron scattering methods to study
The spatial distribution of clusters in water-swollen local and long-range structure of water in Nafion 117
Nafion was investigated by Dreyfus and co-workers membranes. Since small-angle scattering methods
through SANS experiments.44 The shape and inten- suffer from an inherent inability to probe a suf-
sity of the scattering profiles obtained from samples ficiently large region of reciprocal space, the results
with water contents ranging from 6 to 26% (by of small-angle (low q) and wide-angle (high q) scat-
volume) were fit to a local-order model that describes tering experiments were combined so that direct
the distribution of spherical, hydrated clusters in a Fourier mathematical inversions (integral trans-
locally ordered structure with four first neighbors forms) could be applied to data that were thereby
located at a well-defined distance embedded in a expanded over q-space. Using this method, atom-
matrix of completely disordered (gaslike) clusters. atom radial distribution functions for the atom pairs
Using a radial distribution function for this locally associated with water (i.e., H-H and O-H) in the
ordered (tetrahedral) structure and a form factor for cluster domains of Nafion were calculated to provide
homogeneous spheres, the following intensity func- real space information. It was stated that Nafion in
tion was derived the Ni2+ ion exchanged form was used so that the
counterion radial distribution function could be de-
(
I(0) 3(sin qR - qR cos R)
)( rived, as well, although these results do not appear
2
sin qD
I(q) ) 1+z - in the report.
S(0) (qR)3 qD
)
It is important to describe the preparation of these
3(sin qRD - qRD cos qRD) samples, which were annealed at high temperature
z′ in dry nitrogen gas and then rapidly quenched in
(qRD)3
liquid nitrogen.45 Because they were thereby ren-
dered amorphous, there was no scattering due to the
where I(0) and S(0) are the coherent scattering
PTFE-like crystallinity that would complicate the
intensity and an interference term at q ) 0, respec-
data interpretation. Membranes were boiled in NiCl2
tively; q is the scattering vector, R is the radius of
solutions and then soaked in H2O, D2O, and H2O/
the scattering particle, D is the distance between
D2O ) 50:50 solutions. Due to the elimination of
locally ordered particles, RD is the distance from the
crystallinity from the melt-quench pretreatment,
origin cluster beyond which a disordered distribution
these samples were considerably hydrated at 41 vol
of clusters exists; and z and z′ are the number of
%.
clusters in the locally ordered structure and the
number of clusters in the volume corresponding to The following equation shows the relationship
the correlation hole between 0 and RD, respectively. between the partial structure factors, SRβ(q), and the
The a priori choice of a diamond-like structure led atom-atom correlation functions, gRβ(r), that refer to
to the refinement of constant values of R ) 1.122, z the distance, r, between atoms R and β:
) 4, and z′ ) 4.39.
sin (qr)
With this local-order model,44 only two adjustable SRβ(q) ) 4πF∫r2[gRβ(r) - 1] dr
parameters (D and R) were needed to obtain reason- qr
F is the macroscopic density. This equation was commercial processes to uniformly cast thin mem-
inverted to give gRβ(r) using the principle of maximum branes for fuel cell applications. Since the properties
entropy. The total coherent structure factor is a linear and performance of Nafion membranes are directly
combination of all of the partial structure factors related to their complex morphology, and the super-
where each term is multiplied by the product of the molecular organization of ionic and crystalline do-
atomic fractions and coherent scattering lengths of mains may be altered by the processing history, it is
the species R and β. of great importance to understand the nature of these
The intramolecular correlations of O-H at r ) 1.00 Nafion solutions and the subsequent evolution of
Å and of H-H at r ) 1.56 Å are identical, and the morphological development during the casting
second peaks show small differences, perhaps arising process. With this fundamental knowledge, the
from interactions between water molecules and sul- means to enhance the transport properties and
fonate groups, although it would seem that water efficiencies of PFSI membranes by controlling and
coordination around Ni2+ would have to be consid- tailoring the morphology for specific applications may
ered, as well. From this, the authors concluded that be realized.
the structure of water in the clusters is essentially With the availability of dissolution procedures for
that of water in the bulk state.42 This would seem Nafion, Aldebert and co-workers used SANS to study
reasonable considering that the water uptake of these the nature and structure of Nafion solutions in water
samples corresponded to an average of 21 H2O and ethanol.48 Similar to the case observed in the
molecules per -SO3- group. scattering results of Nafion membranes, the “solu-
The H-H radial distribution appears in the functions” were found to be heterogeneous, yielding a
tion r[gHH(r) - 1] which was plotted versus distance single scattering maximum attributed to interference
r. The peak observed at low r was associated with between nanometer-scale scattering objects that were
the cluster radius. If the clusters are assumed to be distributed throughout the total volume of the sample.
spherical, this distribution is consistent with a cluster Moreover, since the separation distance, d, varied
radius of 18.5 Å, which is similar to values based on systematically with the volume fraction of the iono-
scattering based estimates from other studies.17,42 At mer, φv, the particles were suggested to exhibit
higher r, the function shows a mean periodicity of electrostatic repulsion similar to that observed for a
these structures (i.e., cluster center-to-center dis- colloidal suspension of charged particles. Since the
tance) of 70 Å. This would leave a rather large space experimental plot of log d versus log φv varied
between the “surfaces” of adjacent clusters of 33 Å, following the law d ∼ φv-0.5, the solutions in both
which seems rather large, especially when compared water and ethanol were assumed to consist of a
to that estimated in the model of Gierke et al.,17 dispersion of rodlike particles.48,49 In comparing the
which was only around 10 Å. The analysis ceased to scattering results to dimensional parameters calcu-
be of an ab initio nature when the assumption of lated from three different models (i.e., a three-
spherical clusters was made and the 33 Å space had dimensional fcc lattice, a two-dimensional hexagonal
to be accounted for. The authors explain that these array of rods, and a cubic phase of rods), the authors
data reflect a continuous network of water structure found that the cubic phase model was most likely.
rather than an array of isolated hydrated clusters for Furthermore, the rodlike particles were suggested to
two reasons. (1) 41 vol % of water is considerably have a compact cylinder structure with the solvent-
above the percolation threshold, and a structure in polymer contact at the surface of the charged “mi-
which water is distributed in spheres of this size and celle” as opposed to a more open coil model. The
having this center-to-center spacing can only accom- center of the rodlike particles was also proposed to
modate 8 vol % of water. (2) Because the cluster contain crystallites of the PTFE backbone segments
center-to-center spacing is much larger than the oriented along the rod axis.49
cluster diameter, gHH(r) would be expected to drop From SAXS results, Rebrov and co-workers also
to around zero at about r ) 40 Å before approaching found that Nafion solutions could be characterized
the asymptote of 1.00 at large r, which it does not. as containing colloidal suspensions of anisotropic
These and other arguments led the authors to invoke polymeric particles.50 By manually extrapolating the
the presence of connective water structures and experimental scattering curves to a state of infinite
suggested a view consisting of rodlike aqueous re- dilution in order to eliminate interparticle interfer-
gions that intersect at “nodes” (i.e., clusters).45 It ence effects, the form of the scattering particles was
should be remembered that these data and the suggested to most closely resemble that of flattened
derived conclusions refer to the distribution of waters disklike particles. While this questionable practice
the sole scattering source in these experimentss of approximating the scattering profiles at infinite
throughout the Nafion structure and not to the direct dilution raises concern as to the validity of the chosen
structure of Nafion. form factor, the results are qualitatively consistent
with the previous study of Aldebert et al.48 in that
2.2. Scattering from PFSI Solutions and Recast Nafion “solutions” are not “true” solutions but contain
Films relatively large colloidal aggregates of anisotropic
In the early 1980s, reports surfaced that detailed structure with dimensions exceeding 15 nm.
procedures for dissolving Nafion membranes in water/ More recently, Loppinet and co-workers have used
alcohol mixtures at elevated temperatures and pres- both SAXS and SANS (with contrast variation meth-
sures.46,47 Currently, Nafion solutions are available ods) to characterize the morphology of dilute PFSI
commercially, and these solutions are now used in solutions having a volume fraction of polymer less
than 0.5%.51,52 The intensity profiles of these dilute Using WAXD and SAXS, Moore and Martin found
solutions were found to decrease as 1/q at low that the as-received and solution-processed films
q-values in a manner typical of rodlike particles. The were semicrystalline (having similar degrees of crys-
scattering profiles in the low q-range were also found tallinity), while the recast films were virtually amor-
to satisfactorily fit the form factor of cylindrical phous.54,55 The SAXS profiles of the as-received and
particles having a radius of ∼2.0-2.5 nm. For higher solution-processed films displayed a distinct ionomer
volume fraction solutions, the asymptotic behavior peak at q ) 1.2 nm-1, attributed to the presence of
of the scattering profiles was also found to be ionic clusters, and a prominent low angle maximum
consistent with that predicted for cylindrical par- at q ) 0.5 nm-1, attributed to the long spacing within
ticles. By studying the effect of various solvents, well-organized crystalline domains. In contrast, the
different counterions, and added salt, the formation scattering profile of the recast material displayed
of the colloidal particles was attributed to aggregation only the ionomer peak at q ) 1.2 nm-1. While all
of the neutral part of the ionomer chains. In contrast samples possessed ionic clusters, regardless of pro-
to the behavior of low molecular weight surfactants, cessing history, the enhancement in crystallinity and
the rodlike shape of the polymeric particles did not physical properties of the solution-processed films
show evidence of any phase transitions with changes was attributed to the ability of the ionomer chains
in concentration, salt, temperature, or the nature of to reorganize from the colloidal aggregate state in
the solvent or counterions. Furthermore, the size of solution to a more entangled network in the solid film
the cylindrical particles was shown to be related to with added thermal energy and solvation from the
the polymer-solvent interfacial energy. Using SANS high boiling point solvents at elevated casting tem-
contrast variation studies, the particles were found peratures. The presence of ionic clusters in the recast
to be dense, indicating that the thickness of the films was attributed to the aggregation of ions on the
polymer-solvent interface is negligible compared to surface of the contacting micelle-like colloidal par-
the cross-sectional size. From the apparently high ticles in the absence of solvent. Moreover, the high
particle densities calculated from the constant neu- solubility and “mud-cracked” character of the recast
tron scattering length densities, the authors suggest material supported the conclusion that the colloidal
that the rods contain a crystallinity index of ap- morphology observed in solution remained intact in
proximately 50%. At relatively high volume fractions the recast state with little if any chain entanglement
(i.e., 5-20%), the cylinders are proposed to be dis- or coalescence between particles.
tributed in the solution with the existence of some Gebel and co-workers used a similar high temper-
local order. Space-filling analysis assuming a local ature casting procedure with high boiling point
order in a hexagonal array yields radius values solvents to prepare reconstructed Nafion mem-
consistent with the previous determinations. branes.56 WAXD and SAXS results from this study
Since Nafion films possess excellent chemical sta- indicated that the morphology of the PFSI mem-
bility and a high affinity toward large organic cations, branes reconstructed at room temperature (i.e., cast
Nafion solutions have been widely used in the forma- at room temperature with added high boiling point
tion of polymer modified electrodes for a variety of solvent) or at high temperature was different from
electroanalytical investigations. For many of these that of the as-received membrane. In agreement with
applications, it has been assumed that the morphol- the results of Moore and Martin,53,54 the room tem-
ogies, physical properties, and chemical characteris- perature cast membranes were essentially noncrys-
tics of the solution-cast PFSI films were identical to talline, while the high temperature process reestab-
those of the as-received membranes. Moore and lished significant crystallinity with long-range order.
Martin, however, discovered that this simple as- The morphology of recast Nafion films, processed
sumption was inaccurate.53,54 For example, Nafion at room temperature in wet and dry environments,
films cast from ethanol-water solutions at room was studied by Scherer and co-workers using WAXD
temperature (i.e., the general method used to produce and SAXS.57 In agreement with the previous studies,
polymer-modified electrodes), referred to as recast room temperature casting yielded little if any crys-
films, were found to be “mud-cracked”, brittle, and tallinity in the resulting films.54 The films cast under
soluble at room temperature in a variety of polar wet environments, however, yielded intense scatter-
organic solvents. In contrast, the as-received Nafion ing maxima positioned at lower q-values relative to
membranes are flexible, tough, and insoluble in those of the films cast under dry conditions. Based
virtually all solvents at temperatures below ∼200 °C. on this behavior, the authors suggested that the
However, through the application of high boiling number and size of the clusters increased with
point solvents in the casting procedure, Moore and increasing relative humidity during the casting pro-
Martin demonstrated that the desirable properties cess. Since previous studies of the morphology of as-
of as-received Nafion could be reconstituted in solu- received films also show an increase in the intensity
tion-cast films and membranes, referred to as solu- of the ionomer peak and a shift to lower q-values with
tion-processed films, if the casting temperature ex- increasing water content, the assertion that relative
ceeded ∼160 °C.53 This observation suggested that humidity during casting affects the cluster morphol-
the contrast in properties between the recast and ogy seems inconclusive within the context of this
solution-processed films was of a morphological origin limited study. Nevertheless, the results of this study
and the resulting solid-state structure of the cast and the above investigations strongly support the
films was a function of the macromolecular organiza- general suggestion that specific casting procedures
tion of chains in solution during solvent evaporation. may be developed to alter the crystalline and ionic
Figure 3. 2-D SAXS scattering patterns of (a) as-received Nafion, showing a slight morphological anisotropy in the ionic
domains from the membrane calendaring process, and (b) uniaxially oriented Nafion (λb ) 5.4).
cluster morphologies in an attempt to control and was observed in the equatorial direction. Using the
tailor the properties of Nafion films and membranes. Bragg spacings calculated from the peak maxima, a
nonaffine deformation behavior was observed be-
2.3. Evaluation of Nafion Morphology through tween the microscopic and macroscopic dimensions.
Studies of Oriented Membranes This observation led to a conclusion, similar to that
proposed to explain solvent swelling behavior, that
Given the morphological complexity of Nafion nonaffine behavior was inconsistent with an inter-
membranes and the limited structural information particle scattering model yet could be accounted for
that can be extracted from the relatively diffuse by deformation of a core-shell particle.
scattering profiles, the elucidation of detailed mor-
phological models has required the study of controlled More recently, Moore and co-workers found that
morphological manipulations by processes including Nafion membranes, neutralized with tetraalkyl-
solvent swelling and mechanical deformation. Over ammonium ions, could be oriented with high elonga-
the years, a variety of groups have demonstrated that tion at elevated temperatures (e.g., 200 °C for tetra-
the application of uniaxial extension provides a butylammonium form Nafion) to yield extremely
particularly useful means of evaluating the size, anisotropic WAXD and SAXS scattering patterns.58-60
shape, and spatial distribution of crystallites and With uniaxial deformation (ranging up to λb ) 5.4)
ionic domains in Nafion. Moreover, structural infor- at temperatures above the ionomers’ R-relaxation (as
mation obtained from these studies has been used determined using dynamic mechanical analysis) fol-
to differentiate the various morphological models that lowed by cooling to room temperature, the oriented
have been proposed for PFSIs. morphology was observed to be stable in the absence
Gierke et al. reported the first data showing the of stress. Using a 2-dimensional area detector, the
effect of tensile drawing on the SAXS profiles of the scattering patterns were shown to increase in inten-
unhydrolyzed Nafion precursor and that of the hy- sity at azimuthal angles perpendicular to the stretch-
drolyzed ionomer.17 For the oriented nonionic precur- ing direction (i.e., equatorial scattering) with a
sor, the low angle scattering maximum attributed to profound decrease in intensity (to the point of disap-
a periodicity associated with the organization of pearance) at angles parallel to the stretching direc-
PTFE-like crystallites was observed only in the tion (i.e., meridional scattering) with increasing
meridional scan, implying a periodicity in the long orientation.58 At relatively high orientations (greater
spacing along the fiber axis. In contrast, SAXS scans than ca. λb ) 2.0), only equatorial spots at q ) 1.8
of an oriented, hydrolyzed ionomer showed that the nm-1 were observed (Figure 3). In contrast to the
ionomer peak was only observed in a scattering earlier studies of oriented membranes containing
direction perpendicular to the draw direction, imply- inorganic ions,17,37 the equatorial peak position of the
ing a periodicity normal to the fiber axis. TBA+ form membranes remained constant with
A similar analysis was conducted by Fujimura et elongation. A circular integration over all azimuthal
al.37 on cesium-neutralized Nafion with orientations angles of the scattering patterns was performed to
up to λb ) 1.5 (where λb is the ratio of the final calculate the relative invariant, Q, for samples hav-
extended length to the initial undrawn length). Over ing a range of elongations. A significant increase in
a range of orientations, the ionomer peak was ob- Q with elongation was attributed to an increase in
served to shift to lower angles and decrease in the extent of ionic aggregation with elongation to
intensity in the meridional (draw) direction, while a yield a matrix with fewer lone ion pairs in the highly
shift to higher angles and an increase in intensity oriented membranes.58
Figure 4. Azimuthally integrated SAXS profiles as a function of temperature for (a) TMA+ and (b) TBA+ forms of uniaxially
oriented Nafion.
The thermal relaxation behavior of oriented Nafion the basis of an interparticle scattering model and
membranes was also evaluated using time-resolved contrasted to the behavior predicted for an intra-
synchrotron SAXS analysis.61 For oriented mem- particle model.63 With the interparticle scattering
branes neutralized with tetramethyl- (TMA+), tetra- model, the arching was explained by an increase in
ethyl- (TEA+), tetrapropyl- (TPA+), and tetrabutyl- coherence of the intercluster spacings perpendicular
ammonium ions (TBA+), the anisotropic scattering to the draw direction, accompanied by a correspond-
patterns were found to relax back to isotropic pat- ing reduction in coherence in the parallel direction.
terns as the samples were heated from 50 to 300 °C In contrast, the observed arching with respect to the
(see Figure 4). Moreover, the temperature at which intraparticle scattering model was noted to be un-
the morphological relaxation occurred tracked with likely, since it would require an elliptical deformation
the size of the tetraalkylammonium ions, following of the clusters, as independent scatterers, and an
the order TBA+ < TPA+ < TEA+ < TMA+. This electron density difference between the long and
relaxation behavior of the ionic domains was noted short axes of the elliptically shaped clusters. Using
to be in excellent agreement with the order of the a model-independent maximum entropy method for
principle dynamic mechanical relaxations (R-relax- reconstructing a 2-dimensional electron density map
ations) for these materials. Thus, the thermal relax- in real space from that corresponding to the scatter-
ation of the scattering patterns was attributed to the
ing data, the intercluster spacings parallel to the
onset of ion-hopping (i.e., the thermally activated
draw direction were found to be almost random, while
process of redistributing ion pairs between aggre-
the spacings in the transverse direction were ob-
gates in the electrostatic network in order to relieve
local stress on the chain segments). With a weakened served to display distinct periodicity.61 Furthermore,
electrostatic network for the samples containing large the clusters in the drawn sample were found to be
tetrabutylammonium ions, the onset of ion-hopping somewhat agglomerated relative to their distribution
is activated at lower temperatures, allowing for in the as-received membrane. Based on this finding,
significant chain mobility and morphological re- the anisotropic shape of the small-angle upturn was
organization. attributed to scattering between oriented cluster
Elliott and co-workers performed a detailed SAXS agglomerates having periodic dimensions signifi-
investigation of the morphology of Nafion membranes cantly larger than the intercluster spacings. For
that were subjected to uniaxial and biaxial deforma- extruded membranes drawn perpendicular to the
tion.62,63 For as-received membranes, manufactured machine direction (i.e., a sequential biaxial draw), the
by Du Pont using an extrusion process, the cluster anisotropic morphology was found to persist, but to
reflection was shown to exhibit a limited degree of an extent much less oriented than that observed for
arching in the direction perpendicular to the machine membranes drawn parallel to the machine direction.
direction. Upon uniaxial extension, this arching was Relatively isotropic morphologies were obtained only
observed to increase in a manner consistent with when the membranes were processed with a simul-
previous studies. This arching was rationalized on taneous biaxial draw.
In a subsequent communication,62 Elliott and co- structure information of Nafion in direct space fol-
workers found that uniaxially oriented membranes lowing established procedures used to study oriented
swollen with ethanol/water mixtures could relax back semicrystalline polymers and strained elastomers.
to an almost isotropic state. In contrast, morphologi- Based on the classical ionomer domain model of
cal relaxation was not observed for membranes Gierke17 (i.e., ionic clusters as inverted micelles
swollen in water alone. While this relaxation behav- interconnected by channels), the deformation behav-
ior was attributed to the plasticization effect of ior of Nafion in the dry state was evaluated over a
ethanol on the fluorocarbon matrix of Nafion, no range of deformations from 0 to 125%. During the
evidence of interaction between ethanol and the deformation, the channels between clusters are pro-
fluorocarbon backbone is presented. In light of posed to open and widen in the direction of strain
the previous thermal relaxation studies of Moore and and merge with adjacent domains to form slab-
co-workers,61 an alternative explanation for this shaped domains that align together in the draw
solvent induced relaxation may be that ethanol is direction.65 Using contour maps constructed from the
more effective than water in weakening the electro- CDF analysis, the average domain thickness Dt (i.e.,
static interactions and mobilizing the side chain the dimension perpendicular to the draw direction)
elements. Clearly, a more detailed analysis of this and the maximum domain height Dh (i.e., along the
phenomenon involving a dynamic mechanical and/ draw direction) were determined. As the elongation
or spectroscopic analysis is needed to gain a detailed was varied from 50% to 125% (near the break point),
molecular level understanding of this relaxation the domain thickness decreased from 1.9 to 1.5 nm,
process. while the domain height increased from 3 to 6 nm.
Londono et al. reported the results of a synchrotron From this real space information, and the noted
SAXS study of oriented Nafion membranes.64 As the absence of autocorrelation peaks on the meridian, the
elongation was increased to 150%, meridional scat- ionic domains were concluded not to be cylinders.
tering of the ionomer peak disappeared while the With respect to the polymer matrix, the crystalline
intensity profile narrowed azimuthally about the domains were shown to be cylindrical in nature and
equator. At 150% elongation, an equatorial streak, undergo plastic deformation and parallelization upon
similar to that shown in Figure 3b, was observed and elongation. At high elongations, the CDF analysis
noted to resemble the SAXS pattern of drawn fibers indicated a broad distribution of crystallite sizes with
containing a microfibrilliar morphology. Using a an ultimate inclination of 40° relative to the draw
finely collimated X-ray beam (100 µm) and stacked direction.
films, edge-on scattering patterns were obtained for Recently, van der Heijden et al. have used simul-
orientations parallel and perpendicular to the stretch- taneous SAXS/WAXD to study the morphological
ing direction. The observation of an isotropic scat- variation in Nafion during the deformation process
tering pattern from the orientation parallel to the of uniaxial extension.66 Using absolute values of the
stretching direction confirmed the fibrilliar morphol- Hermans orientation factor, the orientations of both
ogy of oriented Nafion. Furthermore, based on this the ionomer and crystalline domains were quantita-
analysis, the authors suggest that the morphology tively compared with respect to the draw ratio. The
of uniaxially drawn Nafion consists of oriented structural anisotropy at length scales between 0.2
cylindrical or lamellar domains, rather than spherical and 4 nm was evaluated based on a ratio, R, of the
clusters. From these scattering results alone, a crystalline to ionomer domain orientation factors. For
definitive distinction between cylinders or lamella draw ratios below 200%, R values were observed to
was not possible. Simultaneous SAXS/WAXD mea- be well below 1.0, indicating that the ionomer do-
surements were also collected in-situ during the mains oriented more readily than the crystalline
orientation process. The (100) crystalline reflection domains. Based on this observation, a simple model
of the PTFE-like crystallites was observed to narrow involving the deformation of spherical ionic domains
about the equator at a rate of almost double that of into ellipsoidal shaped structures was rejected, since
the orientation of the ionic domains. While the the clusters would not be expected to deform more
authors acknowledge the bilayer lamellar model than their surrounding matrix. Beyond 200% elonga-
proposed by Starkweather,41 the SAXS/WAXD results tion, the R values approached 1.0, indicating that the
are stated to indicate that lamellar ionic domains crystalline and ionic domain orientation was cor-
may also be present in the amorphous phase. How- related with respect to elongation. This two-stage
ever, if the Starkweather model were strictly consid- orientation behavior was thus explained based on a
ered, such that the crystalline bilayers were inti- new consideration of Nafion morphology (see below)
mately linked to the lamellar ionic domains, then the as a collection of bundled, elongated polymeric ag-
orientation of the ionic and crystalline domains gregates containing relatively extended chain crys-
should have tracked together with a one-to-one tals. Upon stretching, two orientation mechanisms
correlation. were postulated to occur simultaneously: at small
Computational methods combined with a novel draw ratios, the large bundles rotate such that the
approach in the application of scattering physics were elongated aggregates in the bundles are more or less
recently employed by Barbi et al. in a synchrotron correlated in orientation, and at large draw ratios,
SAXS study of the nanostructure of Nafion as a the alignment of individual elongated aggregates is
function of mechanical load.65 A new method of refined in the draw direction due to sliding and/or
multidimensional chord-distribution function (CDF) disentangling of the aggregates from each other.
analysis was used to visualize the multiphase nano- Further analysis of the crystalline morphology indi-
cated that the degree of crystallinity remained con- of close-packed spheres was only chosen to describe
stant as a function of the draw ratio. the spatial distribution of inverted micelle clusters
for the convenience of simple space-filling calcula-
2.4. Current Models for the Morphology of Nafion tions. The cluster dimensions obtained from these
calculations were consistent with other experimental
Since the early studies of Nafion morphology, information (e.g., from TEM analysis); however, the
numerous investigators have attempted to reconcile observed spatial distribution of clusters was not, by
a wealth of structural information with observed any means, paracrystalline. Furthermore, the exist-
properties (specifically, transport properties) in order ence of 1 nm channels connecting the clusters was
to develop a well-defined morphological model for proposed in order to reconcile the highly permselec-
perfluorosulfonate ionomers. As noted in the studies tive transport properties of Nafion containing in-
above, the complicating factors associated with this verted micelles.16,31 With channels, a network of
endeavor include the random chemical structure of clusters could exist with a percolation pathway
the copolymer, the complexity of co-organized crys- through the membrane; however, no direct experi-
talline and ionic domains, vast morphological varia- mental evidence has ever revealed the existence of
tions with solvent swelling, the relatively low degree channels in these polymers. On the other hand, this
of crystallinity, and the diffuse, heterogeneous nature cluster-network model was the first to attribute the
of the morphology that leads to a wide range of SAXS maximum in Nafion to an interparticle origin,
domain dimensions. While the majority of morpho- and it offered the possibility of slight morphological
logical information about Nafion has originated from reorganization during solvent swelling.17 Both of
small-angle scattering and wide-angle X-ray diffrac- these considerations have prevailed through the test
tion experiments, the simple fact remains that this of time and are currently featured in the modern
polymer yields very little characteristic detail in the
understanding of this polymer.
dimensions probed by these methods. Thus, the quest
for a universally accepted model continues with a In a manner following the scattering studies of
spirited debate in the current literature. other ionomeric systems, the intraparticle, modified
The principle scattering models for the morphology core-shell model of Fujimura et al.37 lost popularity
of Nafion that have been proposed to date include a through the late 1980s and has since been more or
cluster-network model proposed by Gierke et al.16 less discounted as a reasonable model for the mor-
(Figure 1), a modified (depleted-zone) core-shell phology of Nafion. Recent SANS analysis of water-
model proposed by Fujimura et al.36,37 (Figure 2b), a swollen Nafion was performed by Gebel and Lambard
local-order model first proposed by Dreyfus et al.,44,67-69 with a quantitative comparison of the core-shell
a lamellar model proposed by Litt,70 a sandwich-like models to other interparticle scattering models.67
model proposed by Haubold et al.,71 and a rodlike Using the depleted-zone core-shell model of Fuji-
model proposed by Rubatat et al.72 Central to each mura,37 the theoretical profiles showed a poor fit to
of these models is the recognition that the ionic the experimental data, and the dimensional and
groups aggregate in the perfluorinated polymer ma- contrast values extracted from the best fits were
trix to form a network of clusters that allow for found to be unrealistic. Furthermore, the existence
significant swelling by polar solvents and efficient of isolated ion pairs constituting the matrix in the
ionic transport through these nanometer-scale do- depleted-zone core-shell model was noted to be
mains. These models do, however, differ significantly highly questionable in swollen membranes and con-
in the geometry and spatial distribution of the ionic firmed to be improbable by ESR measurements.73
clusters. While the complications associated with In considering the ionomer peak observed for
modeling the spatial distribution of ions within a Nafion to be of an interparticle origin, a more
semicrystalline matrix are often either ignored or quantitative approach to extracting detailed struc-
simply eliminated through various thermal treat- tural information from the scattering data involved
ments, the experimental evidence of PTFE-like crys- the application of a modified hard sphere model.52
tals in Nafion is undeniable. Thus, the validity of a In the work of Kumar and Pineri42 (discussed above),
given model in completely describing the morphology the SAXS peak was accounted for quantitatively (in
of Nafion requires a reasonable account for the both position and intensity) as arising from the
existence of crystallites in the presence of a pervasive interference between clusters, assuming them to be
ionic network. noninteracting hard spheres. With calculations based
The original cluster-network model proposed by on the modified hard sphere model,24,25 the theoretical
Gierke et al.16,17,31,32 (also referred to as the cluster- peak value of the relative intensity invariant over an
channel model) has been the most widely referenced intermediate q-range was found to be within 15% of
model in the history of perfluorosulfonate ionomers. the experimentally observed intensity invariant for
Despite the very large number of papers and reports samples having different equivalent weights, differ-
that have strictly relied on this model to explain a ent cationic forms, and different water contents.
wide variety of physical properties and other char- Recently, Orfino and Holdcroft used the modified
acteristics of Nafion, this model was never meant to hard sphere model to determine the characteristic
be a definitive description of the actual morphology dimensions associated with the ionic clusters in dry
of Nafion, and the authors recognized that further and hydrated Nafion.74 With reasonable fits to the
experimental work would be required to completely experimental scattering profiles, their model calcula-
define the nature of ionic clustering in these iono- tions indicated that the number density of clusters
mers.17 For example, the paracrystalline, cubic lattice decreased from 3.2 × 1019 to 9.7 × 1018 clusters/cm3
upon hydration, while the cluster size increased from terials, having intermediate equivalent weights. For
a radius of 1.12 to 2.05 nm. Although this observation the short-side-chain PFSIs having high equivalent
is in agreement with the morphological reorganiza- weights, the model fits were found to deviate signifi-
tion proposed in the cluster-channel model of Gierke,17 cantly from the experimental profiles. This observa-
a comparison of the model parameters determined tion was attributed to the relatively high crystallinity
for the radius of the ionic core and the radius of of the blocky ionomers;55 long runs of PTFE units in
closest approach to the size of a single, hydrated ionic the high equivalent weight ionomers yielded an
group led the authors to the conclusion that clusters intense crystalline peak at low q-values that influ-
could be simply bridged by single ionic sites rather enced the observed position of the ionomer peak.
than organized ionic channels.74 Deviations observed for the low EW ionomer were
The local-order model, first applied by Dreyfus et attributed to the very large water uptake (∼80 wt
al. (see above),44 is another interparticle scattering %) as a result of the absence of crystallinity.55 For
model used to define the spatial distribution of this system (i.e., the 635 EW ionomer), the highly
spherically shaped ionic clusters in Nafion. More swollen morphology was suggested to be more ap-
recently, Gebel and Lambard demonstrated that the propriately described as a polymer-in-water system.68
local-order model provided a better fit to SAXS and Thus, based on recent studies of PFSI solutions and
SANS profiles of hydrated Nafion than those of the the reasonable fit of the scattering profile to the form
depleted-zone core-shell and modified hard sphere factor of cylinders, the morphology of the swollen 635
models over the range of scattering vectors between EW system was proposed to resemble a connected
0.3 and 2.0 nm-1.67 At larger q-values, the deviation network of rods,51,52 as opposed to a locally ordered
between the model fit and experimental data was distribution of spherical clusters. This new concept
attributed to the fact that the model assumes mono- for the morphology of PFSIs will be discussed in more
disperse clusters having a homogeneous electron detail below.
density. While additional terms allowing for polydis- Through consideration of the morphological and
persity in dimensions were noted to yield improve- dimensional changes that had been observed in SAXS
ment in the model fits, the added adjustable param- analyses of Nafion during swelling and deswelling
eters were deemed undesirable. Using a Debye- experiments, Litt70 proposed a reevaluation of Nafion
Bueche model to fit the small-angle upturn (i.e., a morphology that could provide a simpler rationaliza-
method found successful for other ionomer systems), tion of the literature than the current interparticle
heterogeneities having a correlation length on the models based on spherical clusters.17,37,44 The prin-
order of 8 nm were found to best reproduce the excess cipal concern addressed by this proposal was that a
intensity at low q-values. Although this dimension major reorganization of the ionic domains is required
was noted to be surprisingly small for “large-scale” to explain the fact that the observed Bragg dimen-
heterogeneities, the excellent agreement with the size sions from the ionomer peak varied almost linearly
of the correlation-hole in the radial distribution with water content as opposed to the 1/3-power
function of the local-order model was acknowledged. dependence expected for isotropic (affine) swelling of
To extract further information from the scattering spherical structures. Furthermore, since the swelling/
at low angles, an ultrasmall-angle USAXS camera deswelling behavior is observed to be reversible, the
was used to significantly extend the low q-range.64 extent of morphological reorganization required with
While this analysis yielded correlation lengths for the cluster-network model was argued to induce
large-scale density fluctuations in excess of 300 nm, nonrecoverable plastic flow. Using the SAXS data of
no detailed description of the morphological origin Gierke et al.17 over a limited range of water contents,
of these heterogeneities was offered, and their exist- Litt showed that the d spacings are proportional to
ence in contrast to the assumed “gaslike” order of the the volume of absorbed water.70 Based on this obser-
local-order model was not reconciled. vation, a lamellar model consistent with the bilayer
Gebel and Moore68 also applied the local-order structure suggested by Starkweather41 was proposed
model to the SAXS and SANS study of a series of for the morphology of Nafion. In this model, the ionic
short-side-chain perfluorosulfonate ionomers, ranging domains are defined as hydrophilic “micelle” layers
in equivalent weights from 635 to 1269 g/equiv. These separated by thin, lamellar PTFE-like crystallites.
PFSIs differ from Nafion in the structure of their side As water absorbs between the lamella and separates
chain, which contains only one ether linkage and two them, then the increase in d spacing between ionic
CF2 units separating the sulfonate group from the domains is expected to be proportional to the volume
perfluoronated backbone. Moore and Martin55 re- fraction of water in the polymer and the swelling
ported the first SAXS and WAXD characterization behavior should be completely reversible, thus elimi-
of these new PFSIs, showing that these polymers nating the requirement of morphological reorganiza-
possessed a morphology similar to that of Nafion, tion.
having both crystalline and ionic domain structures While the lamellar model proposed by Litt70 pro-
that varied with equivalent weight and water con- vides a convenient and simple explanation for the
tent. Subsequently, Gebel and Moore quantitatively swelling behavior of Nafion, an important morpho-
analyzed this initial SAXS data in combination with logical feature observed in the SAXS and SANS
new SANS data using the local-order model.68 In profiles of PFSIs, namely the low angle maximum
agreement with the findings for Nafion,44,67 the local- attributed to the crystalline, interlamellar long spac-
order model was found to satisfactorily reproduce ing, is ignored in this analysis. In studies of short-
the scattering profiles for the short-side-chain ma- side-chain PFSIs, Gebel and Moore68 found that the
Figure 5. Sandwich-like structural element proposed for the morphological organization of Nafion. (Reprinted with
permission from ref 71. Copyright 2001 Elsevier.)
swelling behavior of the ionic domains in these PFSIs 4.0 nm, in harmony with the results of all other SAXS
is identical to that observed for Nafion; however, the investigations. The scattering cross section data was
interlamellar spacings shift with water content in a fitted to a layered model whose basic structure
manner significantly different from that of the inter- element (i.e., the scattering particle) is a “sandwich”
cluster spacings. This observation was also confirmed (Figure 5). The outer portion of this sandwich (the
by Young et al. through SANS studies of Nafion “shell”) consists of the side chains, including the
swollen with a variety of polar organic solvents.75 sulfonic acid groups, and the inner liquid portion (the
Since the lamellar model essentially stipulates a “core”) consists of the water/methanol molecules. To
parallel shift for the two maxima attributed to a provide channels that serve as conduction pathways
lamellar structure of the ionic domains imposed by for protons through the membrane, these structural
the semicrystalline character of the polymer, the elements were proposed to be juxtaposed in a linear
observation of dissimilar shifts indicates that the fashion so that the liquid core regions are contiguous.
lamellar model is an oversimplification in the global The structure factor chosen for this “sandwich”
description of the morphology of Nafion. In an at- model is that of a rectangular parallelepiped whose
tempt to reconcile the dissimilar shifts with the dimensions are those of the monodisperse sandwich
constraints of the lamellar model, Young et al.75 have particles with an overall random orientation.71 A
proposed a preferential partitioning of solvents into least-squares fit of the model scattering cross section
regions of the structure that do not give rise to the to the experimental data yielded the dimensions of
ionic scattering maxima (presumably domains of the core thickness, c, the combined thickness of the
amorphous fluorocarbon chains between crystallites). two shells, s, as well as the lateral dimensions a and
Nevertheless, this consideration implies that at least b. Also issuing from the fit were the electron density
a portion of the crystallites in Nafion are separated contrasts for the core and shell regions. From the
by nonionic regions, which is different from the case model fits, the basic structure element was shown
of the crystallites that impose the lamellar ionic to have lateral dimensions a and b between the
structure as explicitly dictated by the lamellar model. values of 1.5 and 4.5 nm, and the total thickness of
To date, no such morphology has been observed for the sandwich, c + s, was about 6.0 nm. Fitted values
Nafion. of s, c, a, and b, were also plotted versus the volume
A variation of the lamellar model was recently percent of methanol in water, which showed that c
proposed by Haubold et al.,71 in which synchrotron decreased while s increased with increasing percent
SAXS studies were performed on acid form Nafion methanol. The molecular interpretation of this ob-
117 samples. The molecular weight of these samples servation was that as the methanol composition
was reported to be 250 000 g‚mol-1, and the experi- increased from 0 to 100%, the spacing between
ments were conducted on dry samples in air and -SO3H groups on the two different shells (the region
samples equilibrated with water, methanol, and a in which the solution is located) decreased from
range of water/methanol mixtures using an in situ approximately 2.7 to 1.3 nm, while the shells (s/2)
flow cell. The most fundamental result of this study extended from approximately 1.7 to 2.5 nm. The
is that the data show the usual ionomer peak at q ) latter result implies a change in conformation and/
1.4 nm-1, which gives a Bragg spacing of d ) 2π/q ≈ or packing of side chains and main chains; however,
Figure 6. MaxEnt reconstruction of SAXS data from as-received Nafion equilibrated under (a) ambient humidity and (b)
100% RH. (Reprinted with permission from ref 63. Copyright 2000 American Chemical Society.)
the scattering contrast (relative to the average elec- digm for PFSIs, which was capable of reconciling the
tron density throughout the membrane) for the shell microscopic and macroscopic swelling behavior of
versus percent methanol was found to be relatively these materials. The maximum entropy (MaxEnt)
constant. Thus, the authors concluded that methanol method was used in this analysis to determine the
“leaves the side chains nearly unaffected”, although most statistically probable structure for Nafion from
the meaning of this in consideration of the extension the experimental SAXS data. Through this analysis,
of the side chains is unclear. While this sandwich 2-dimensional SAXS patterns (in reciprocal space)
model was presented within the context of direct were transformed to generate real space, two-
methanol fuel cell membranes, it does not offer a dimensional projections of electron density distribu-
complete, unambiguous 3-dimensional pattern of tions within a representative volume of the mem-
hydrophobic/hydrophilic organization, but rather of brane. The technique of spatial filtering was also
local structure. Nor is there any information regard- used to highlight particular features of the electron
ing rearrangement of the main chains, sections of density maps with respect to different features of the
which are known to organize into limited crystalline underlying SAXS patterns in discrete scattering
regions. vector ranges. Specifically, a low pass filter was
A methodological improvement in the analysis of applied to mask the cluster reflection, while a high
SAXS data of Nafion has recently provided a new pass filter was used to mask the small-angle upturn
perspective on the spatial distribution of clusters in prior to applying the maximum entropy data ma-
PFSIs. Elliott et al.62 have reported SAXS studies of nipulation.
water-swollen Nafion 115 in the acid form and
various cation forms in which the membrane water Interpretation of the MaxEnt electron density
content was quantitatively adjusted using an envi- maps, termed “reconstructions”, led to the conclusion
ronmental sample holder with controlled humidity. that the most statistically probable scattering model
The radially integrated SAXS profiles show the usual for Nafion is of an ion clustered morphology with a
cluster reflection that moves to lower scattering angle hierarchical scale of structures.63 Figure 6 shows a
with increasing humidity as well as the usual small- maximum entropy reconstruction of SAXS data for
angle upturn at low scattering vector magnitudes. Nafion H+ under ambient humidity and 100% RH.
From the behavior of the cluster peak position (d- The clusters in these “images” are the white spots
spacing) versus volume fraction of absorbed water, and are often seen as agglomerated (clusters of
it was concluded that cation exchange does not affect clusters), and more so for the 100% RH case. With
the hydrated morphology. In direct contradiction to an increase in water uptake, the MaxEnt reconstruc-
both the lamellar morphology, as suggested by Litt,70 tions showed that the average cluster separation
and other models based on individual three-dimen- increases with a concurrent decrease in the cluster
sional clusters (i.e., noninteracting hard spheres),42 number density. An important fact is that the “mi-
it was further noted that d was neither directly croscopic swelling”, defined in terms of the Bragg
proportional to the water volume fraction nor pro- spacing for the cluster reflection, does not transform
portional to the cube root of this quantity. in an affine manner with the macroscopic swelling.
Using a novel, model-independent maximum en- Particularly, the relative intercluster expansion is
tropy method to interpret the SAXS data, Elliott et greater than that of the bulk sample dimensions.
al.63 presented a self-consistent morphological para- This behavior was explained as a consequence of the
decrease in the number density of clusters with in defining the structure of swollen membranes and
swelling. further supports the conclusion that considerable
The distribution of clusters observed in the MaxEnt structural reorganization occurs during the swelling
reconstructions was also shown to be nonuniform, process in order to keep constant the specific surface
suggesting an agglomeration or clustering of clus- with increasing cluster size.
ters.63 By applying a low pass filter (to highlight long- The slope of -0.5 in the double logarithmic plot of
range structure), this agglomeration was linked to the Bragg distance versus the polymer volume frac-
the small-angle upturn observed in the SAXS pro- tion for the highly swollen membranes was noted to
files. With confirmation from the MaxEnt analysis be consistent with a dilution of rodlike polymer
of oriented samples (see above), the small-angle aggregates.51,52 Furthermore, the small-angle scat-
upturn was concluded to be produced by the inde- tering profiles of Nafion solutions and the highly
pendent scattering from cluster agglomerates yet swollen membranes were virtually superimposable
fundamentally caused by ionic aggregation. Thus, the in the q-range from 0.7 to 2.0 nm-1, suggesting that
principal scattering features of Nafion (i.e., the small- the local structure on the scale of a few nanometers
angle upturn and the ionomer peak) were noted to was identical. Since previous studies of these solu-
be produced by two distinct scattering mechanisms tions provided strong evidence for the existence of
operating on different size scales. rodlike aggregates,51,52 this observation led to the
The evolution of PFSI structure from the dry state conclusion that the morphology of highly swollen
through the water-swollen state to solution was membranes could be considered as a network of
studied by Gebel using small-angle scattering meth- rodlike polymer particles. The deviation of the scat-
ods.76 With respect to recent studies suggesting that tering curves at small angles (q < 0.7 nm-1) was
PFSI solutions contain rodlike structures,51,52 the aim attributed to the presence of nodes in the network
of this investigation was to gain an understanding structure and to large scale heterogeneities. It is
of the morphological changes that occur as the important to note that this connected network of
ionomer is converted from the highly swollen state rodlike structures connected by nodes was proposed
to solution during the dissolution process, and visa earlier by Lee et al.45 to explain their SANS data
versa during the solution-casting process. Using a probing the local water structure in Nafion.
high temperature swelling process, water-swollen Based on the results of this scattering analysis of
membranes with water volume fractions ranging Nafion over a wide range of water contents, combined
from φw ) 0.3 to 0.93 were obtained. Similarly, with energetic considerations, Gebel proposed a
homogeneous, aqueous solutions with concentrations conceptual description for the swelling and dissolu-
up to a polymer volume fraction φp ) 0.12 were tion process shown schematically in Figure 7.76 In
prepared, allowing for an overlap in polymer volume this qualitative model, the dry membrane is consid-
fraction between the solution and swollen membrane ered to contain isolated, spherical ionic clusters with
states. For the water-swollen membranes, the scat- diameters of ∼1.5 nm and a center-to-center separa-
tering profiles showed the prominent shift of the tion distance of ∼2.7 nm. With the absorption of
ionomer peak to lower q-values with increasing water water, the clusters swell to hold pools of water
content, in agreement with previous studies. A plot surrounded by ionic groups at the polymer-water
of the microscopic degree of swelling (defined as the interface in order to minimize the interfacial energy.
increase in the Bragg distance of the ionomer peak As the water content increases to between φw ) 0.3
maximum relative to the Bragg distance extrapolated and 0.5, structural reorganization occurs to keep
for the dry membrane) versus the macroscopic linear constant the specific surface area, and the onset of
expansion over the large range of water contents percolation is achieved by the formation of connecting
revealed a significant change in swelling behavior cylinders of water between the swollen, spherical
(i.e., a change in slope) at a linear expansion of 30%. clusters. At φw values greater than 0.5, an inversion
In a double logarithmic plot of the Bragg distance of the structure occurs such that the structure
versus the polymer volume fraction, φp, the change resembles a connected network of rods. Finally, as
in slope was found to occur at φp ) 0.5. For low water the membrane “dissolves” into solution, the rodlike
contents, the slope of this plot was equal to -1.33, structures separate to yield a colloidal dispersion of
while, at high water contents (φp < 0.5), the slope isolated rods. While this model offered a plausible
dropped to -0.5. This behavior was thus attributed mechanism for the evolution in structure from the
to an inversion from a water-in-polymer state at low widely accepted concept of isolated clusters for mem-
water contents to a polymer-in-water state at high branes containing relatively low water contents to
water contents. The asymptotic behavior of the rodlike structures in solution, no thermodynamic
scattering curves for the water-swollen membranes justification for the phase inversion process at φw )
was also evaluated following the Porod law in order 0.5 was offered. Moreover, the fact that the scattering
to extract values for the specific surface (i.e., σ, equal profiles near this phase inversion point did not show
to the area of the polymer-water interface per polar a significant change in contour is difficult to rational-
headgroup) of the scattering particles. For all samples, ize.
the asymptotic behavior was the same, and the To further probe the phase behavior of hydrated
average Porod limit yielded a specific surface value Nafion, Rollet et al.77 used a contrast matching
of σ ) 55 Å2. This observation, in agreement with method in the SANS analysis of Nafion membranes
other studies involving a low range of water contents, neutralized with tetramethylammonium ions. With
confirmed the strong significance of this parameter 12 hydrogen atoms per counterion, the N(CH3)4+ ions
water domains. Through the assumption of spherical

aggregates and the recognition that no structural
change occurs when H2O is replaced by D2O, the
structure factor contribution to the scattered inten-
sity was eliminated by dividing the SANS profiles
obtained with various H2O/D2O mixtures by the
intensity profile for Nafion in pure water.77 The
resulting normalized profiles were shown to present
a clear minimum near q ) 1.5 nm-1 that shifted
toward smaller q-values as the composition of D2O
decreased. In considering the form factor for spherical
core-shell particles, the two opposing geometries
(i.e., a water core surrounded by a counterion shell
or a polymer core surrounded by a counterion shell)
were used to simulate SANS curves comparable to
the experimental data. While the general shape of
the curves for both simulations resembled the ex-
perimental data and showed a clear minimum near
q ) 1.5 nm-1, only the “polymer core surrounded by
a counterion shell” geometry matched the shift of this
minimum toward smaller q-values. Thus, based on
this analysis, the authors concluded that the mor-
phology of hydrated Nafion is best described as
polymer aggregates surrounded by water.77 Although
an attempt was made to model both spherical and
elongated particles using a similar approach, the
authors acknowledged that a conclusive distinction
between these shapes was not possible given the
limited q-range of the experimental data.
In a second paper, Rubatat and co-workers72 used
a combination of small-angle X-ray and neutron
scattering techniques, including USAXS, to probe
characteristic dimensions in hydrated Nafion over the
range from 1 to 1000 nm. Through a combination of
the USAXS and SAXS data, the entire scattering
curve of Nafion was presented to show the extended
small-angle upturn attributed to long-range hetero-
geneities in electron density, the low angle maximum
attributed to supralamellar dimensions of the crys-
talline domains, and the ionomer peak attributed to
intercluster interferences. In addition, double loga-
rithmic plots of the scattering data I(q) versus q
showed two regimes of linear intensity decay; at low
angles, the data followed a q-1 power law, while, at
high angles, the intensity decay followed the expected
q-4 power law indicative of a sharp interface for the
Figure 7. Conceptual model for the morphological re- scattering particles. The q-1 regime was noted to fit
organization and continuity of the ionic domains in Nafion the behavior expected for scattering from rodlike
as the dry membrane is swollen with water to the state of particles over the range of wave vectors from 2π/L
complete dissolution. (Reprinted with permission from ref to 2π/D, where L and D are the rod length and
76. Copyright 2000 Elsevier.) diameter, respectively. Based on the observed span
were found to be excellent neutron scatters that of scattering vector in this regime, the rodlike par-
provided a third source of scattering contrast, thereby ticles were estimated to have dimensions of D ≈ 6
acting as a useful morphological probe at the inter- nm and L g 100 nm.
face between the water and polymer domains. By Scattering data over the extended range of q-space
varying the water composition with a range of H2O/ were also collected from Nafion samples containing
D2O mixtures, the contrast profile of the scattering a wide range of water contents (φw ) 0.05-0.84).69
domains was evaluated to gain a new insight into the While systematic shifts in the intensities and posi-
structure of Nafion. The relative magnitudes of tions of the structure maxima (i.e., the crystalline and
scattering features in the SANS profiles were shown ionomer peaks) were observed and consistent with
to vary systematically with the H2O/D2O composition, those found by others, the general shapes of the
indicating that the N(CH3)4+ ions were condensed in scattering profiles for all samples were quite similar.
a thin layer with the sulfonate groups at the interface This observation supported the assumption that the
between the hydrophobic polymer parts and the swelling process simply involves a dilution of the
this morphological feature has been extracted from

these scattering studies. As noted in this study, the
swelling behavior observed for this scattering feature
is surprisingly large (i.e., increases in interdomain
distances from 10 to 100 nm with increasing water
content) relative to that expected for expansion of
interlamellar dimensions. Clearly, additional inves-
tigations of this morphological feature are warranted
in order to more fully understand the origin of this
scattering maximum. If the crystalline domains are
contained within the rodlike aggregates, as previ-
ously proposed,49 then the interlamellar dimensions
should track with a corresponding increase in the rod
length. On the other hand, the interlamellar correla-
tions may be of an interaggregate origin, and the
dimensions should thus track with some function of
the interaggregate spacing. To date, this question
remains to be answered.
Figure 8. Schematic representation of an entangled
network of elongated rodlike aggregates in Nafion. Long- 2.5. The State of Understanding from X-ray and
range heterogeneities arising from bundles of locally
ordered aggregates are proposed to give rise to the low Neutron Investigations
angle increase in scattered intensity. (Reprinted with Over the period of about two decades, numerous
permission from the original author.)
models have been proposed for the morphology of
scattering entities rather than inducing a strong Nafion based on information gathered from small-
structural reorganization of the cluster morphology, angle scattering and wide-angle diffraction studies.
as suggested earlier by Gebel.76 In plotting the While these distinctly different models have varied
characteristic (Bragg) distances, 2π/q*, corresponding significantly in quantitative and qualitative detail,
to the two structure maxima, versus the polymer many have weathered the test of time and remain
volume fraction, two microscopic swelling regimes the focus of debate in the current literature. In all
were observed, in agreement with that found by these studies, one fact remains constant; the mor-
Gebel.76 For high φp, the swelling process was ob- phology of Nafion is complex. Fortunately, state-of-
served to follow a power law close to φ-1, in agree- the-art scattering methods and data analysis tech-
ment with the results of Fujimura36,37 and Litt.70 For niques are providing a wealth of new morphological
low φp (below φp ) 0.6), the swelling process was information that may be correlated with other mor-
observed to change to a power law close to φ-0.5, phological and physical property studies to bring
suggesting a dilution of cylindrical shaped particles. forward a deeper understanding of the complex
To reconcile the shift in swelling behavior observed nature of Nafion.
at φp ) ∼0.6 with the suggested rodlike morphology In reviewing the evolution of information that has
of the scattering particles, the authors proposed that been reported from these morphological studies, it
the aggregates are more ribbonlike than rodlike, with should be recognized that each of the principal
a cross-sectional thickness and width close to 2 and models for the morphology of Nafion, while distinct
8 nm, respectively.72 At low water contents, the in their conceptual design, contains relevant aspects
ribbons are considered to pack face-to-face in a of structure that collectively facilitate a more global
manner that allows for an initial dilution law with understanding of the molecular organization of per-
an exponent of -1. However, at high water contents fluorosulfonate ionomers. By recognizing the specific
(low φp), where the distance between aggregates is perspective by which a particular model was devel-
greater than 8 nm, the orientation distribution oped, a number of interesting disparities between the
around the long axis becomes isotropic in a manner different models may be reconciled. For example, a
that allows the aggregates to be considered as relatively new rodlike model has been recently modi-
cylinders with a dilution law following an exponent fied to include the possibility of ribbonlike aggregates
of -0.5. that may assemble locally,72 at low water contents,
Therefore, in the context of a morphology consid- in a lamellar fashion. By narrowing the perspective
ered to be fundamentally characterized by the pres- to a limited range of water contents, suddenly the
ence of elongated aggregates, the authors propose rodlike and lamellar models are in agreement. In a
that the intermediate q-range provides information similar fashion, the excellent mechanical integrity of
related to the size, shape, and spatial distribution of swollen Nafion membranes may be attributed to
the aggregates, while the very small q-range (i.e., the entanglements of the proposed elongated polymeric
small-angle upturn) may be attributed to larger aggregates in the solid state. At these entanglement
bundles of aggregates (see Figure 8) with a long- or crossover points, local nodes may exist76 such that,
range heterogeneous distribution.72 Between these at moderate water contents, the swollen nodes may
regions in q-space lies information related to the appear as isolated clusters. This perspective recon-
distribution of crystallites in Nafion. Unfortunately, ciles the rodlike and cluster-network models.
the degree of crystallinity in Nafion is low (e.g., 8-12 The morphological information and interpretations
wt %36), and thus, little conclusive information about currently reported for Nafion still support the concept
of a connected network of ionic domains through and geometrical distribution of ionic clusters, crys-
which polar solvents and ionic species permeate. The tallites, and the continuous perfluorinated matrix.
principal focus of these important investigations The usual problems regarding sample thickness and
continues to be aimed at understanding the struc- provision of sufficient electron density contrast, as
tural nature of the ionic domains, such as the size, well as possible artifacts, are present. The SAXS and
shape, and spatial distributions of the clusters. One SANS methods of structural inquiry have advantages
consensus that appears to be developing in the that are specific to the nature of the underlying
current literature is that the ionic domains in hy- wave-particle interactions. On the other hand, the
drated Nafion possess some degree of anisotropy in results are usually prejudiced by the need to assume
shape and heterogeneity in their spatial distribution. a particular model, which must necessarily be simple
In contrast to a regularly ordered morphology, this so that the most basic geometrical parameters can
general conception of a more irregular, heterogeneous be extracted.
structure is consistent with the random chemical
Ceynowa performed electron microscopic studies of
structure of PFSIs. As we learn more about the
60-80 nm thick microtomed Nafion 125 membranes
organization of these ionic domains, however, the
that were converted, for the purpose of affecting
simple concept of “ionic clusters” in these materials
electron density contrast, to the Pb2+ form, and all
becomes more convoluted. Although the ions clearly
of the excess cations and co-ions were removed.78 It
cluster as a result of the nanoscale phase separation
between polar and nonpolar constituents of the is the heavy metal that provides electron density
copolymer, the notion of discrete ionic clusters or contrast between the phase in which it resides and
contiguous ionic domains depends on the perspective the surrounding phase. These membranes were then
of the method used to probe the morphology and the exposed to ethanol and 1,2-epoxypropane, although
chemical composition of the membrane (e.g., the state these solvents would not have remained in the
of hydration). Nevertheless, the current models and samples under the vacuum in the microscope column.
their detailed methods of structural analysis now The micrographs consisted of uniformly distributed
constitute useful tools that may be applied to precise “points” that were presumed to be ion clusters that
evaluations of morphology-property relationships in were 3-6 nm in diameter.
these benchmark materials. Moreover, these tools Fujimura et al., in addition to their SAXS studies,
may be used to characterize trends in controlled performed TEM investigations of thin sections of
morphological manipulations of PFSIs through vari- 1100 EW Nafion in the cesium ion form.37 The
ous processing procedures (e.g., solution-casting, important details of sample preparation were not
thermal treatments, mechanical deformation, etc.) provided in this paper, although it is likely that the
that may prove to enhance the membrane properties samples were sectioned from membranes rather than
over those of as-received Nafion. being films cast from solution. The heavy Cs+ coun-
The nature of the crystalline component in Nafion terion associated with the sulfonate groups provides
has received much less attention than that of the electron density contrast so as to highlight the cluster
ionic domains, and thus, the relevance of this mor- regions. It should be appreciated that these samples
phological feature to the technologically important were only ∼60 nm thin, so that, in principle, if the
properties of the membranes is still unclear. Since characteristic intercluster spacing is ∼4 nm, there
the initial studies of Nafion morphology, the crystal- will be only ∼15 clusters to span the thickness of this
line component has been recognized as an important sample. A micrograph of this sample shows isolated
structural feature and often considered as a neces- particles that are, on average, a few nanometers in
sary component that provides mechanical integrity diameter.
and a barrier to solvent swelling. With respect to
Xue et al. also performed TEM investigations of
current models, however, the crystallites may be
Nafion 117 morphology.79 Solution-cast films, of
considered to exist within elongated polymeric ag-
thicknesses in the range 50-100 nm and stained by
gregates or as critical structures that impose the
RuO4, showed good contrast. It was explained how
organization of the ionic domains. In the rodlike
this staining agent would be selectively incorporated
models, the crystallites may play a minor role to that
of entanglements in affecting the mechanical behav- in the side chain domains, notwithstanding the fact
ior of the swollen membranes. On the other hand, that none of the usual unsaturated CdC bonds for
the lamellar model suggests that the crystallites are staining reactions exist in these regions. FTIR and
the principal factor in limiting ionic domain swelling. X-ray fluorescence analyses showed that this was the
Clearly, further work is required to resolve this case based on the fact that residual water, which
important issue. In addition, as Nafion is the bench- would be found in the most hydrophilic regions, was
mark in studies of many new membrane materials oxidized to give RuO2. Therefore, the observed fea-
for fuel cell applications, the role of crystallinity (if tures would likely be clusters. Three phases were
needed at all) in affecting desirable membrane prop- observed on the basis of white, gray, and dark regions
erties must be addressed. in micrographs: Spherical clusters, in the diameter
range 25-50 Å, were surrounded by interfaces and
were uniformly dispersed throughout an organic
3. Microscopy Studies matrix across the field of view. While the concept was
Microscopic studies, regardless of the method of earlier advanced by Yeager et al.,18 Xue et al. claimed
producing images, have an advantage in being able to be the first to view three phases in Nafion using
to provide a direct visualization of the sizes, shapes, the TEM technique.
Rieberer and Norian investigated unstained mi- sample was prepared by placing a drop of water on
crotomed Nafion 117 samples using TEM.80 The it, and the AFM apparatus was kept in a constant
thinner edges of the 50 nm thick sections of the K+ humidity chamber. The supernodular structure was
form were used so as to probe regions having thick- preserved, and the surface was less rough than that
ness on the order of 10 nm. Other samples consisted of the dry sample.
of films cast from ethanol solutions. The solution-cast McLean et al. used AFM to elucidate the surface
samples were exchanged to the Cs+ form to affect and near-surface ionic domain and semicrystalline
good electron density contrast. In both cases, steps morphologies of Nafion 117 in the K+ form, as well
were taken to minimize beam damage to these thin as Surlyn and sulfonated poly(styrene-co-ethylene/
sections, as explained in the paper. TEM images of butylene-co-styrene) (Kraton) ionomers to which
the Cs+ form samples showed clusters having sizes comparisons were made.83 In AFM, the tapping mode
in the range 1-5 nm, the average of which is around phase generates image contrast on the basis of small
the value of intercluster Bragg spacings that was regions being hard versus soft or, in a general sense,
derived from SAXS studies. Bright regions, due to sharp viscoelastic gradients. Similar to the method
phase contrast fringes, were said to be nonionic of dynamic mechanical analysis, but on a smaller
channel areas representing a third phase of the scale of interrogation, the phase angle that is plotted
system. The microtomed samples showed the same in two dimensions across the surface refers to the lag
features as those that were solution-cast. These between the oscillatory motion of the cantilever to
studies included X-ray energy dispersive spectroscopy which a probe tip is attached and the oscillating input
(EDS) and energy loss spectrometry (EELS) micro- signal to the piezoelectric drive for the cantilever.
analyses of elemental composition. In particular, the The stiff regions observed in the studies of McLean
Cs-to-S mole ratio was found to be close to unity, et al. were assigned to fluorocarbon crystalline do-
showing essentially complete ion exchange so that mains, the size of which was around 10 nm. The
all of the sulfonate groups are accounted for in the crystalline packing was referred to as lamellae which
clusters seen on the micrographs. are not organized into stacks, as is the case for the
Porat et al. performed TEM (zero-loss bright field) Surlyn ionomers that were studied. An interesting
studies of very thin Nafion films that were cast from outcome of the “normal” tapping studies was the
ethanol/water solutions, and some of the conclusions discovery of a very thin (∼5 Å) fluorine-rich layer over
are as follows.81 It was suggested that the backbone the surface in which the contrasted soft regions are
had a planar zigzag conformation in large ortho- amorphous.
rhombic crystallites as in polyethylene, in contrast Other experiments involved lower oscillation am-
with the helical conformation found in poly(tetra- plitude tapping on the same sample after a normal
fluoroethylene). This is an interesting result, al- tapping experiment was conducted over the surface.
though there are no other studies that support this The authors suggested that attractive tip interactions
view. Sulfur imaging indicated the presence of sul- with ionic domains dominate the phase signal under
fonate clusters that are 5 nm in size. these conditions so as to produce greater contrast in
Lehmani et al. studied the surface morphology of the sense of phase difference. In this way, the ionic
dry and hydrated acid form Nafion membranes using domains are seen more clearly. The low amplitude
atomic force microscopy (AFM) in tapping mode.82 experiments revealed ionic domains of size 4-5 nm,
The digital images were also analyzed for surface which is in agreement with the results of earlier TEM
roughness. The results for “cleaned” samples that experiments.37,81 The ionic domains, appearing as
were dried under vacuum at 80 °C show “super- irregularly shaped, were not resolved to a degree to
nodular aggregates” of spherical domains having which distinct boundaries and interphase as well as
diameters of ∼45 nm. These domains, that appear amorphous regions could be seen because these
rather distinct on the images, are of a size scale that features overlapped and artifacts of the AFM method
is greater that the Bragg spacings derived from the relating to feature edges must be considered. No long-
SAXS profiles of these materials. It is unclear as to ranged patterned organization of the clusters was
whether these features are strictly a surface phe- seen, which is in agreement with the observation of
nomenon and/or the result of sample preparation. a well-studied broad, single SAXS peak attributed to
The surfaces probed were the actual membrane polar-nonpolar phase separation in these materials.
surfaces rather than those of a fresh surface of a film Images were also obtained for Nafion K+ samples
cross-section that was produced by cutting or freeze that were soaked in deionized water using low
fracture. Thus, the observed structures may not be oscillation energy so as to sense hydrated ionic
representative of the membrane interior. clusters. An ambient humidity conditioned (control)
On a finer level, spherical grains that have an sample exhibited 4-10 nm in diameter clumps of
average diameter of 11 Å are seen. A section analysis, multiple (unresolvable) ionic domains. Upon exposure
which consisted of a plot of image contrast intensity to water, the ionic features became enlarged in a way
versus distance, indicated that there is a mean that they were elongated and appear somewhat as
periodicity of around 49 Å, which is close to the channels having a width of 7-15 nm, as seen in
values of the SAXS Bragg spacing usually associated Figure 9. While Figure 9 might offer a view of
with intercluster distances. This, as well as other contiguous hydration pathways, the morphology par-
microscopic studies, favors the model of phase sepa- ticularly is that of the K+ ionic form so that caution
ration as opposed to the core-shell model as applied must be exercised in applying this information in
to the interpretation of scattering data. The hydrated rationalizing the proton conductance and other prop-
Figure 10. Evolution of a smaller number of larger

clusters with increased hydration of Nafion ionomers,
according to Gierke, Munn, and Wilson. (Reprinted with
permission ref 17. Copyright 1981 Wiley.)
suggest that the surface of Nafion is of very high

fluorocarbon composition.84 This phenomenon, which
has been known for some time, is referred to as
“Schroeder’s Parodox”.85,86
McLean et al. offer interesting comments that have
implications regarding molecular transport. It was
mentioned that polar-nonpolar interfacial tension
would restrict the formation of high surface/volume
narrow channels and drive the system to larger ionic
domain dimensions. In fact, a simple model calcula-
tion by Gierke et al. implied that increasing hydra-
tion would cause a smaller number of larger clusters,
that is, cluster coalescence,17 and the morphology of
the water soaked K+ form sample seen in Figure 10
Figure 9. Low oscillation energy tapping mode phase supports this idea. Presumably this occurs by indi-
images of a K+ form Nafion 117 membrane. In part A, vidual ion-hopping events or cluster coalescence, or
sample was exposed to ambient, room temperature humid- a combination of both. From a thermodynamic per-
ity (ionic species are in the light regions). In part B, sample spective, this process might be viewed as a continu-
was exposed to deionized water. The images are 300 nm × ous and spontaneous minimization of the overall
300 nm, and the phase range is 0-80°. (Reprinted with hydrophilic-hydrophobic interfacial free energy by
permission from ref 83. Copyright 2000 American Chemical
Society.)
decreasing the cluster surface/volume ratio. To be
sure, the concept of “interface”, in the usual con-
erties of sulfonic acid form Nafion within the context tinuum sense, becomes blurred at this fine ultra-
of PEM fuel cells. structural level and the “roughness” of the hydrated
In addition to providing a “direct” visualization of phase has been discussed by a number of authors.
ionic clusters, the discovery of the thin fluorine-rich It was also mentioned by McLean et al. that ions
layer may have important implications with regard and polar solvent molecules must necessarily diffuse
to the liquid/vapor sorption of these materials. To be across thin amorphous fluorocarbon or interfacial
sure, the nature of this layer would depend on the regions between swollen polar domains. However, all
specific liquid or vapor environment contacting the of this does not require the need for channels with
membrane surface. In particular, the molecular diameters of 10 Å that are coated with SO3- groups
polarity of the environment and its influence on for long-range transport. In any case, a simple
surface tension would be involved. The difference consideration of the steric volume of SO3- groups in
between the liquid water versus saturated water relation to the size of these channels leads to the
vapor sorption was mentioned in this report. Also, conclusion that more than one group would have
the contact angle experiments of Zawodzinski et al. difficulty fitting into this very small volume. Related
Figure 11. Stylized view of Kreuer of the nanoscopic hydrated structures of Nafion and sulfonated polyetherketone.
(Reprinted with permission from ref 91. Copyright 2003 Elsevier.)
to this discussion is the conclusion of Tovbin,87 which clusters was found to be between 0.30 and 0.88 nm.
was based on the X-ray scattering results of Ozerin Owing to these very small channel lengths, it was
et al.,88 that the channels through the perfluoro- concluded that the cores can be connected by only a
carbon regions are around 50% wider than those single ionic site so that the concept of a pore has no
proposed by Gierke et al. meaning on this scale.
As mentioned earlier, Orfino and Holdcroft per- James et al. also performed AFM tapping mode
formed SAXS studies of acid form Nafion 117 samples phase investigations of Nafion 115 membranes while
that were in the dry and wet (19% of total swollen the samples were in an environmental chamber of
weight) states.74 One of the conclusions was that, controlled humidity.90 The data from their experi-
upon hydration, the number of ionic clusters de- ments were presented in a somewhat different fash-
creases while their size increased, which also sup- ion from that usually seen in the literature. Rather
ports the concept of the evolution of a smaller number than mapping the phase angle ψ, sin ψ, which is
of larger clusters, as advanced by Gierke et al. and directly proportional to the energy loss, was mapped.
later factored into the equilibrium swelling model of This energy loss is due to either viscoelastic or tip
Mauritz and Rogers that will be described later.89 An adhesion effects. An increased viscoelastic energy
important goal of this study was to determine the change would involve local material softening due to
nature of the ubiquitous “ionic channels”. In a wet water incorporation. An adhesion energy change
membrane, 89% of the ionic sites were calculated as would involve enhanced attraction between the tip
residing in the cores whose average diameter was probe and inserted water structures. The authors
4.10 nm. There are 74 SO3H groups per cluster, and argued that, at the frequency of the cantilever, the
the average length of channels between adjacent viscoelastic effect would be negligible and the adhe-
sive mechanism predominates. Their studies indi- forms, and the molarity (0-14 M) of aqueous elec-
cated that while the number of clusters decreased, trolyte solutions in which the films were equili-
the size of the average cluster increased with in- brated.92 At that time, it was already understood that
creasing humidity, which, again, is reinforced by the the properties of these materials were very sensitive
conclusions of others as discussed throughout this to swelling history so that these equations applied
section. The images showed clusters with sizes in the for the “standard pretreatment” of boiling in pure
range 5-30 nm. The sizes at the upper end of this water for 30 min. Information was also presented on
range are considerably greater than those proposed water absorption at room temperature versus the
on the basis of SAXS studies, and these large features temperature of initial equilibration (uptake decreases
were attributed to aggregates of clusters of size 3-5 with increasing initialization temperature), as well
nm. The larger structures were rationalized as as percent swelling in organic solvents, as well as
cluster agglomerates, as also proposed by Elliott et mixed solvents, at room temperature. Compared with
al. as a result of their SAXS studies.63 These AFM the conventional sulfonated cross-linked polystyrene
studies also detected features that are the conse- ion exchange membranes, Nafion swells to a greater
quence of extrusion of the precursor form. extent in a number of organic solvents than it does
Kreuer discusses the structure of Nafion as com- in water. This information was useful in estimating
pared to the structure of lower cost, thermally stable, dimensional changes which were reduced by fabric
oxidation resistant, sulfonated poly(ether ketone) reinforcement in chlor-alkali electrochemical cells.
membranes within the context of fuel cells.91 While While the general trends described by these equa-
it is not intended to discuss the latter nonperfluori- tions will hold for current grades of extruded Nafion
nated materials here, Kreuer’s review contains a good films, their use in an absolute sense might be
summary of the state of knowledge of Nafion up to questioned and should especially be avoided for
the time of his writing. The nanoscopic view, depicted solution-cast membranes, especially those of low EW.
in Figure 11, is described as an arrangement of low Yeo investigated the swelling of Nafion sulfonate
dimensional polymeric objects (which might be called membranes in the acid as well as alkali counterion
clusters), the spaces between which are filled with forms, in a number of hydrogen bonding solvents.93,94
water (channels). This picture is based on the SAXS Percent mass uptake versus solvent solubility pa-
studies of Gebel.76 rameter (δ) curves appeared to have two maxima and
However, Kreuer states that SAXS information, by therefore dual cohesive energy densities, each of
itself, cannot resolve the question of whether this which was said to correspond to a distinct morpho-
model is more appropriate than the inverse micelle logical feature. It was suggested that the peak at low
model of Gierke et al., that, in a general sense, would δ corresponds to interactions with the tetrafluoro-
possess more morphological order than that reflected ethylene regions, while the peak at high δ reflects
in the SAXS scattering profiles.17 Regardless of issues more polar solvent molecules that would prefer to
regarding the interpretation of SAXS data, it is worth interact with the ionic regions.
noting that the nanoscale morphology seen in Figure
11 is rather similar to that seen in the AFM images Gebel et al. also performed swelling studies of
of McLean et al.83 The concept of water as being an Nafion 117 membranes.95 They measured sample
extended second phase is reasonable considering the expansion along three directionssalong and perpen-
low degree of morphological order. In summary, dicular to the machine direction, as well as along the
compared with poly(ether ketone), Nafion has wide thickness directionsafter equilibration in the given
versus narrow water channels, the hydration pockets solvent and also calculated the actual number of
are more separated, the percolated hydration struc- solvent molecules per SO3- group at swelling equi-
ture is less branched with good connectivity, and librium. It was concluded that the Gutmann donor
there are no dead end channels; there is less hydro- number of the solvent, which is related to cation
phobic-hydrophilic interface, and the inter-sulfonate solvation energy and has units of kcal‚mol-1, is a
group separation is less. These differences in mor- parameter that is more important than the solvent
phology between the two membranes were explained solubility parameter. Nor did the data correlate well
as being the consequence of having a less hydrophobic with solvent dielectric constant or interfacial surface
backbone, sulfonic acid groups that are less acidic tension. Moreover, a plot of expansion versus solubil-
than those attached to a perfluorocarbon structure, ity parameter was seen to not be in harmony with
and a reduced flexibility of the backbone for the case the dual cohesive energy density concept of Yeo, as
of sulfonated poly(ether ketone). the data of Gebel et al. appeared rather scattered.
In the view of Gebel et al., at the time of their writing,
the solvent was considered to be exclusively con-
4. The Nature of Water, Other Solvents, and Ions tained in the clusters.
in Nafion Duplessix et al. used water vapor pressure iso-
In 1972, when the main thrust for these then-new therm (i.e., water uptake vs external relative humid-
materials was the development of efficient polymer ity) data combined with simultaneous isotherm dif-
membranes for chlor-alkali cells, the E. I. du Pont ferential microcalorimeter analysis to determine the
Co. reported much information, mostly in the form average heat of absorption per water molecule for
of product literature, in the form of empirical equa- 1200 EW acid form samples.29 Hysteresis was seen
tions regarding the water mass and volume uptake between sorption and subsequent desorption curves
of sulfonate form Nafion as a function of EW (in the at 25 °C, and nonzero water content remained at zero
range 1000-1400), some alkali metal counterion relative humidity, indicating the presence of tightly
bound water. Average heat of water absorption water barrier would influence the rate of water
versus water content curves showed that the former uptake, the equilibrium water uptake, which was
remains at a constant value of around -12 kcal‚mol-1 reported by Zawodzinski and others, should not be
up to an uptake that corresponds to about 5 H2O affected in this way. Clearly, more research is
molecules per SO3H group, where the negative sign required regarding this phenomenon and its influ-
of this energy indicates exothermicity. It was a ence on properties.
significant fact that this binding energy, which must Later, Hinatsu et al. studied the uptake of water,
be associated with water of hydration, is the same from the liquid and vapor states at various temper-
for samples dried at room temperature as it is for atures, in acid form Nafion 117 and 125, and Aciplex
samples dried at 220 °C. At higher water uptakes and Flemion membranes, although the latter two
that cause polymer swelling, this energy continuously similar products will not be discussed here.98 These
decreases to a value that is around 4 kcal‚mol-1, studies were motivated by a concern over the delete-
which is approximately the strength of a hydrogen rious effects, involving either overly dry or overly wet
bond in liquid water. The decrease in energy occurs membranes, on electrical conductivity within the
at a lower content for the sample dried at 220 °C, context of polymer electrolyte fuel cells and polymer
which implies a polymer structural rearrangement electrolyte water electrolyzers.
with change in pretreatment. The membranes used in these studies had under-
Rodmacq et al. (same CENG group) applied the gone a rather rigorous cleaning procedure. The liquid
same analyses to the Na+- and Fe2+-neutralized uptake studies were performed by simple, careful
forms, the latter of which permits the use of Moss- weight uptake experiments over the range 25-130
bauer spectroscopy.30 The vapor pressure isotherms °C. All water vapor uptake versus relative humidity
at room temperature for both ion forms were seen to isotherm studies were conducted at 80 °C using a
be similar. Heat of absorption versus water content thermostated glass apparatus involving a water
plots for the Fe2+ salts dried at room temperature vapor reservoir, a quartz spring attached to a sample
indicate that the energy of absorption of the first few basket for mass uptake, and a pressure gauge.
water molecules is 13 kcal‚mol-1 (essentially the The results of the water equilibrium uptake studies
same as that for the acid form), and this somewhat that were conducted at room temperature (25 °C), for
constant energy then drops at higher water contents. samples that were predried in a vacuum at 25, 80,
The low water content value is lower at 10 kcal‚mol-1 and 105 °C for various times, are as follows. Upon
for the Na+ form although no reason was offered to immersion in distilled water, the samples that were
account for this difference. It must be said that the dried at room temperature attained and maintained
results of these Mossbauer studies, as well as the the same uptake versus predrying time. On the other
results of other similar studies of Nafion, have more hand, the water contents of the samples that were
to say about the state of Fe in this environment than dried at 80 and 105 °C underwent a rapid decrease
about the morphology of the host polymer. versus drying time, with the greatest decrease for the
Zawodzinski et al. determined the water uptake of highest temperature. This behavior was interpreted
acid form Nafion 117 at 30 °C in experiments that in a general way as open microstructures caused by
equilibrated membrane samples in a sealed relative the formation of large, hydrated ion clusters upon
humidity environment that was controlled by LiCl swelling and “pore” shrinkage and reorientation of
solutions of known tabulated water activity or equi- the side chains upon drying, the latter of which is
librium partial pressure p/p0.96 Vapor pressure iso- reversed upon reexposure to water at elevated tem-
therm curves were determined by measuring mass peratures. Perhaps another way of stating this would
uptake, expressed in terms of number of moles of H2O be to say that the structures have been “annealed”.
per mole of SO3H groups ) λ versus p/p0. The plot In the usual sense, this involves improved molecular
showed a somewhat small increase in λ with increas- packing, that is, a decrease in free volume that is
ing p/p0 until an uptake of around 6 H2O/SO3H was encouraged by thermal kinetic energy by heating.
reached, after which the increase was considerably Thus, the sample dried at 105 °C would have been
greater. The results can be interpreted in terms of a in a state closer to a true equilibrium condition of
condition of tightly bound, SO3H-associated water in optimal main and side chain packing and therefore
the initial stage and the evolution of more loosely exhibited the least water uptake.
bound water in the latter stage. These results are in Cleaned membranes that were dried at 80 °C were
essential harmony with those of the similar studies referred to as the “N” (normal) form, and those dried
of Duplessix et al. for the acid form.29 Zawodzinski at 105 °C, as the “S” (shrunken) form. Membranes
et al. also compared the water uptake of membranes prepared by preboiling were referred to as the “E”,
exposed to saturated water vapor with the uptake or expanded, form. These symbols are mentioned
resulting from contact with liquid water and discov- here because they were used in other reports to refer
ered that the former was less than the latter.97 This to membranes prepared under similar respective
is the case despite the fact that water activity in both conditions.
cases is, on the basis of macroscopic thermodynamics, Liquid water uptake versus temperature curves
theoretically equal to 1.00. This was discussed in the were established over the temperature range 25-130
above section on Microscopy Studies in relation to the °C for Nafion that was vacuum-dried at the above
paper of McLean et al., in which the existence of a two temperatures (S and N forms) as well as for a
thin fluorine-rich layer was discovered in their AFM sample that was preboiled in water for 2 h (E form).
investigations.83 In theory, while such a hydrophilic Uptake increased with increasing immersion tem-
perature for both forms over the entire temperature concentrated polymer solutions which, at best, can
range, with the N form value being greater than that be applied to amorphous polymers with randomly
of the S form at a given temperature, but the distributed solvent molecules rather than systems
preboiled sample uptake was constant up to 100 °C. having two phases of very different chemical
The lower uptakes for the S form can be attributed compositionssin this case, a sharp contrast between
to a condition of more densely packed main chains strongly hydrophobic and strongly hydrophilic re-
and side chains affected during predrying. However, gions. Moreover, the strong interactions among water
at immersion temperatures higher than 100-110 °C, molecules, ion exchange groups, ether groups in the
the uptake for the different forms becomes the same, side chains, and the hydrophobic backbone in a phase
and this “transition” temperature range was said to separated system cannot be represented by the single
correspond to the glass transition of acid form Nafion, quantity χ that, additionally, was originally intended
as this relaxation was reported and termed in the to account for weak interactions.
early paper of Yeo and Eisenberg.99 The conclusion The Flory-Huggins equation was modified to ac-
is that, after the passage of the membrane through count for “proton-transfer complexes” of the sort
this transition, its thermal history was erased. SO3-‚‚‚+H3O. In short, a fraction of the water content
The water uptake isotherm curves of Hinatsu et was viewed as being strongly bound in these com-
al., at 80 °C, for very well dried membranes, were plexes, and this reduces the effective water concen-
reflective of initial hydration of sulfonic acid groups tration in the membrane. One might view the water
for the slow curve rise region at low RH, followed by molecules bound to these complexes as not being
swelling during the high rise region. These authors
osmotically active, as would be “free” or mobile water.
suggested that the difference in uptake for the liquid
This reasonable concept was based on the IR studies
versus vapor phase is more complex, involving an
of Zecchina et al.103,104 If the fraction of SO3H groups
additional, condensation process on the interior pore
that exist in these complexes is λc, then φ2 increases
walls (a “pore”, presumably was intended to be a
to the effective value of (r + λc)/(r + λ). Then,
cluster) in the latter, whereas sorption from the liquid
the constant for the equilibrium SO3H + H2O T
phase is “direct”. On the other hand, it might be
SO3-‚‚‚+H3O was given by K ) λc/[(1 - λc)a1].
argued that hydrogen bonds must be broken in the
liquid phase before sorption occurs whereas isolated The simultaneous solution of the equations for a1,
water molecules can be sorbed from the vapor phase φ2, and K will yield an a1 versus λ curve if all the
without this requirement. In any case, the cause of underlying parameters were known. To this end,
the differences between the sorption of saturated Futerko and Hsing fitted the numerical solutions of
water vapor and liquid water (both having a theoreti- these simultaneous equations to the experimental
cal water activity of 1.00) does not seem to be entirely points on the above-discussed water vapor uptake
understood at present. isotherms of Hinatsu et al.98 This determined the best
The authors also mention the interesting result fit values of χ and K. χ was first assumed to be
that uptake from water vapor at 80 °C was less than constant, and in improved calculations, χ was as-
that at lower temperatures, as reported by others, sumed to have a linear dependence on φ2, which
and that this difference was not due to the predrying slightly improved the results in terms of estimated
procedure.100 It was suggested that water would data fitting errors. The authors also describe methods
condense on the membrane surface with more dif- for deriving the temperature dependences of χ and
ficulty at the higher temperature and that this would K using the experimental data of other workers.
retard sorption. This situation is of obvious signifi- The authors discuss Schroeder’s paradox, referred
cance with regard to humidified membranes in fuel to elsewhere in this review, and the fact that liquid
cells. Also, as seen in other studies, the water uptake water uptake increases but saturated water uptake
increases with decreasing EW. decreases with temperature. And, at low tempera-
Futerko and Hsing presented a thermodynamic ture, the water uptake by membranes in contact with
model for water vapor uptake in perfluorosulfonic saturated vapor is greater than that by membranes
acid membranes.101 The following expression was in contact with liquid water, which suggests a
used for the membrane-internal water activity, fundamental difference in membrane microstructure
a1, which was borrowed from the standard for the two situations. An energy level diagram of
Flory-Huggins theory of concentrated polymer solu- thermodynamic states versus temperature was pro-
tions:102 posed, based on this Flory-Huggins-based model.
By their nature, and in contrast with microscopic
a1 ) (1 - φ2) exp[(1 - 1/r)φ2 + χφ22] and scattering techniques that are used to elucidate
long-ranged structure, spectroscopic methods inter-
φ2 is the polymer volume fraction, χ is the polymer- rogate short-range structure such as interactions
solvent (water) interaction parameter, and r is the between fixed ions in side chains and counterions,
polymer equivalent unit/mole of water volume ratio main chain conformations and conformational dy-
that is calculated on the basis of the densities of dry namics, and the fundamental hopping events of water
polymer and water as well as the EW and molecular molecules. The most common methods involve infra-
weight of water. φ2 ) r/(r + λ), where λ is the number red (mid-IR and to a much lesser extent near- and
of water molecules per sulfonic acid group. far-IR) and solid-state NMR spectroscopies, although
A number of concerns can be raised over the use other approaches, such as molecular probes, have
of this simplistic equation that was derived for been utilized.
The long-ranged transport of ions, hydrated pro- microns beneath the surface. Moreover, thin, trans-
tons, water, and other solvent molecules through parent solution-cast films may not have the same
Nafion depends on the morphology on the dimen- morphology as those of extruded sheets.
sional scale interrogated by SAXS, SANS, TEM, and SSNMR studies based on 7Li, 23Na, and 133Cs nuclei
AFM methods. This is the scale at which critical for 1100 EW samples whose sulfonate groups were
issues involving the size of, shape of, and spacing exchanged with these cations (no excess counterions
between the ionic domains are important. To formu- or co-ions being present) were conducted versus water
late a realistic transport model, detailed knowledge content.106,107 The spectra reflect the influence of the
of the long-range spatial organization of these do- immediate chemical environment about these cations
mains, the manner in which regions of concentrated that have spins greater than 1/2 and, therefore,
hydration structures are incorporated in these do- possess quadrupole moments. It is the interaction of
mains, and the contiguous interdomain water struc- these quadrupole moments with local electric field
ture is essential. These aspects must be understood gradients that influences the chemical shift (δ) and
in order to account for the obstacles to migration and line width of the observed resonance. In this case,
how these obstacles form a percolation topology that the electric field is mainly due to -SO3- anions as
is expressed in measured ionic conductivities, diffu- shielded by water molecules.
sion coefficients, and other transport parameters. For 23Na a plot of δ versus H2O/Na+ mole ratio )
While these geometrical considerations are impor- r is essentially flat at high water contents until r is
tant, it is also important to understand the nature lowered to where there are less than ∼6 H2O mol-
of fundamental interactions experienced by these ecules per SO3- Na+ ion pair. At higher water
migrating species with functional groups on the contents, most -SO3- Na+ pairs are separated by
polymer. In particular, different ion exchange groups liquidlike water molecules, but since the hydration
will cause different degrees of swelling as well as number of Na+ is 4, a considerable population of
different Coulombic and hydrogen bonding interac- cations must exist in contact ion pairs for r < 6. In
tions with counterions and solvents. These interac- fact, δ greatly increases when r decreases from 6,
tions, that are operative on the scale of less than ∼10 providing evidence for contact ion pairs. On the
Å, are appropriately studied by spectroscopic means. average, there are not enough water molecules
In an early study, Mauritz et al. investigated present to form complete hydration shells that would
anion-cation interactions within Nafion sulfonate provide electrostatic shielding around the Na+ ions.
membranes versus degree of hydration using FTIR/ Contact ion pairing was discussed many years ago
ATR and solid state NMR (SSNMR) spectroscopies. to rationalize the decrease in ionic mobility in simple
An understanding of the dynamic ionic-hydrate electrolytes with increasing concentration.108 The
molecular structures within and between the sul- FTIR and SSNMR spectroscopic results presented
fonate clusters is essential for a fundamental under- here also are in harmony with a four-state model of
standing of the action of these membranes in ion ionic-hydrate association-dissociation equilibrium
transport. This information can be directly related between bound and unbound cations in simple elec-
to the equilibrium water swelling that, in turn, trolyte solutions, as proposed by Eigen et al.109,110 The
influences molecular migration. broadening of the symmetric SO3- FTIR peak upon
This FTIR spectroscopic study was aimed at moni- membrane drying is considered as evidence of a
toring the effects of monovalent alkali counterion multistage association-dissociation equilibrium sub-
type (Li+, Na+, K+, Rb+) and degree of hydration upon ject to restrictions imposed by the polymer host. A
the vibrational state of the fixed sulfonate groups in molecular based model for ion conductivity in Nafion
1100 EW membranes in which the cation/sulfonate would have to incorporate this mechanism.
ratio was 1:1.105 The location and width of the peak The chemical shift and line width observed for each
for the -SO3- symmetric stretching vibration are water content were taken as weighted averages of
sensitive to these variables due to an induced polar- the values in the free and bound states, and from two
ization of S-O dipoles in the sulfonate group by the equations expressing these averages, Pb and Pf, the
electrostatic field of the adjacent cation. This peak mole fractions of bound and free Na+ ions, respec-
wavenumber, νS, is relatively independent of water tively, were extracted. Pb significantly increases as
content and cation type at high degrees of hydration the approximate hydration number that might be
but shifts to higher values and broadens as the water expected for a SO3-Na+ pair is approached from the
content decreases. These changes were interpreted direction of considerable hydration.
in terms of increased sulfonate-cation Coulombic The 23Na NMR parameters of Nafion are not
interactions as water is removed. The relative popu- greatly affected by changing EW in the range of
lation of solvent-separated ions decreases as the water content where valid comparisons are pos-
fraction of contact ion pairs increases. The shift is sible,111 and this reflects the short-ranged nature of
largest for Li+, which has the greatest hydration these dynamic ionic-hydrate structures.
number and highest surface charge density. Smaller The behaviors of the 7Li and 133Cs resonances for
shifts are observed for the Na+ and K+ forms, and membranes incorporating these counterions are quali-
none are observed for Rb+, which hydrates poorly, if tatively similar to that for the samples incorporating
at all, and has a low surface charge density. the 23Na probe. δ for the Cs+ form undergoes a
Caution should be applied in interpreting ATR significant shift at the lowest water content, which
spectra because this is a surface technique that only is reasonable considering the low hydration capacity
interrogates structure within a distance of a few of this large cation.
Figure 12. Four state model of the hydration-mediated counterion-side chain association-dissociation equilibrium, of
Mauritz. (Reprinted with permission from ref 107. Copyright 1982 American Chemical Society.)
The interpretation of these SSNMR results is in belonging to H2O molecules at the hydrophobic-
agreement with that issuing from the FTIR studies hydrophilic interface.
of the same systems, as discussed earlier. Falk, based on his FTIR studies of H2O-, D2O-, and
Mauritz et al., motivated by these experimental HDO-containing Nafion, also rationalized that hy-
results, developed a statistical mechanical, water drated clusters either were much smaller than those
content and cation-dependent model for the counter- proposed by Gierke et al.17 or possessed a shape that
ion dissociation equilibrium as pictured in Figure 12. deviated greatly from sphericity such that fluoro-
This model was then utilized in a molecular based carbon structural fragments intrude into cluster
theory of thermodynamic water activity, aw, for the regions. In retrospect, this early view is more com-
hydrated clusters, which were treated as micro- mensurate with recent visualizations based on sub-
solutions. aw determines osmotic pressure, which, in sequent experimental structural evidence, especially
turn, controls membrane swelling subject to the the structures displayed in pictures by Kreuer.91
counteractive forces posed by the deformed polymer Despite the considerable hydration capacity of Nafion,
chains. The reader is directed to the original paper hydrogen bonding between water molecules in maxi-
for the concepts and theoretical ingredients.107 mally hydrated samples is considerably less extensive
Another noteworthy early IR spectroscopic study than that in liquid water. This situation might be
of Nafion (125 and 152) membranes, in the acid as rationalized in terms of a high surface-to-volume
well as various counterion forms, was that of Heitner- ratio (S/V) for clusters of this size as well as the fact
Wirguin, who assigned a number of characteristic that when H2O molecules are confined in such small
bands and compared the spectra to those of sul- spaces, a large relative population must reside at
fonated polyethylene (Redcat) membranes.112 hydrophobic-hydrophilic interfaces, especially if the
Falk et al. conducted early FTIR studies of the interfaces are irregular.
structure of water in this ionomer, and the papers of The idea of convoluted or rough interfaces is in
these investigators serve as an excellent resource for contrast with the view of Rollet et al., who concluded
the various band assignments, as tabulated in the that the ionic domains are spherical (In fact, this was
book edited by Eisenberg and Yeager.19 The results an assumption in a local-order model.) and that
of these studies were significant on a quantitative interfaces between hydrophobic and hydrophilic
basis, in part, due to their careful control of water phases are sharp, although this situation changes
content by a vapor equilibrium technique that al- when the temperature is greater than 60 °C.114 This
lowed for reproducibility as well as precise determi- conclusion is based on the results of SANS investiga-
nation of the H2O/SO3- ratio.113 tions of Nafion 117 membranes as a function of
Falk inspected the fundamental bands of the HDO temperature, nature and charge of the counterion,
isotope of water in Na+ form Nafion (125 and 142) concentration of electrolyte solutions in which the
sulfonate membranes. The bands of HDO are simpler membranes were equilibrated, and membrane pre-
and lend themselves to less ambiguous interpretation treatment.114 These particular conclusions were based
than the bands of H2O or D2O, as either the D-O or on an analysis of scattering data in the Porod region
H-O stretching band can be studied in isolation for where the intensity scales as q-4. These authors offer
low and high D/H ratios, respectively. At low water the general conclusion that the nanostructure of
contents (1.6 H2O/SO3- mol/mol), both the O-H and Nafion is governed by interfacial phenomena.
O-D stretching bands occur as doublets (3660, 3520 Spectra at higher water contents indicated an
cm-1 and 2695, 2588 cm-1, respectively), and at increasing relative population of water molecules in
higher frequencies than those of HDO in liquid water. O-H‚‚‚O hydrogen bonds, and the shift of the low
The conclusion was that O-H groups in water wavenumber O-H stretching component to lower
molecules exist in two different molecular environ- frequencies reflected increasing strength of the hy-
ments at low water contents. In particular, the low drogen bonds. The fact that the peak wavenumber
frequency component was associated with hydrogen of the high frequency component was constant is
bonded O-H groups, while that at high frequency harmonious with the notion that these groups have
reflected non-hydrogen bonded O-H groups, perhaps O-H- - -CF2 juxtapositions. At least qualitatively, the
ratio of the integrated absorbances under the O-H- - - show the low wavenumber component as a shoulder
CF2 and O-H‚‚‚O peaks should yield information on at ∼970 cm-1 on the dominant peak at ∼980 cm-1.
the geometry of the aqueous domains in terms of The ∼970 cm-1 peak becomes more prominent and
surface-to-volume. For a water uptake corresponding better resolved with hydration, particularly for the
to 2 H2O molecules per SO3- group, 24% of OH Li+ and Na+ forms, and the evolution of this low
groups do not participate in hydrogen bonds. For high wavenumber component with increasing degree of
water contents, however, the peaks were not able to hydration is especially evident for the Li+ form. These
be resolved. facts were taken to suggest that parts of the per-
Kujawski et al. used FTIR/ATR spectroscopy to fluoroether side chains penetrate the ionic clusters
study side chain interactions with the alkali coun- rather than forming a neat inverted micelle as
terion series in Nafion 117 membranes that were depicted in the Gierke model. A closer inspection of
equilibrated in water, ethanol, and 2-propanol.115 the spectra in this region shows a fainter shoulder
These results were compared against the spectra of on this absorption envelope at 995 cm-1, which was
a membrane consisting of an interpenetrating net- verified by Falk and later seen by Cable et al.,
work of polyethylene and sulfonated poly(styrene-co- although the origin of this peak has not been ad-
divinylbenzene) (PESS). The behavior of the sym- dressed.118
metric stretching band of the sulfonate group of Cable et al. observed that the low wavenumber
Nafion showed that, for dry membranes, the smaller component (965 cm-1) increased in wavenumber with
the counterion radius, the stronger is the ion pair increase in degree of Na+-neutralization of dry samples
interaction, as earlier noted by Lowry and Mauritz105 and that this shift correlates with the increasing peak
and Falk et al.116 In the Li+ form equilibrated in pure wavenumber shift of the symmetric stretching vibra-
2-propanol, νs was affected in a fashion similar to that tion in SO3- groups. On the other hand, the high
caused by pure water. On the other hand, this band wavenumber component (980 cm-1) does not shift
was unaffected by the alcohol in the PESS mem- with either counterion type or hydration. The general
brane, meaning that the SO3-Li+ ion pairs were not view is that a portion of the ether groups in the side
dissociated in this hydrocarbon system. The implica- chains are strongly influenced by the state of ion pair
tion is that Li+ cations were solvated by 2-propanol association while the ether groups near the backbone
molecules in Nafion but not in PESS and the strong are shielded from these interactions. In addition to
alcohol affinity of the perfluorinated ionomer was due consideration of through-space Coulombic field per-
to the greater acid strength of the SO3H groups. For turbations on the C-O-C vibration, through-bond
water-isopropanol mixtures, νs was noted to be inductive effects, that is, vibrational coupling within
constant over the composition range. This was inter- the span of a few bonds from a polarized SO3- group,
preted to mean that the strengths of hydrogen should also be considered as a mechanism. These
bonding of the alcohol and water molecules with the conclusions are supported by the fact that the FTIR
sulfonate group are practically identical. However, spectrum of a corresponding sulfonyl fluoride Nafion
the membrane internal composition of the binary precursor film, that did not have cluster morphology,
solution may not be the same as that of the external exhibited only a single ether band at ∼980 cm-1.
solution in which the membrane was equilibrated, The frequency of the symmetric stretch does not
and partition coefficients were not determined. vary with degree of neutralization for hydrated
Cable et al. similarly studied the influence of membranes, which was attributed to the shielding
hydrophobic as well as hydrophilic counterions on of sulfonate anions from the electrostatic field of the
characteristic vibrations of the sulfonate and the two Na+ ions.
perfluoroether groups in the side chains of Nafion 117 Risen et al. investigated cation-anion interactions
membranes using FTIR/ATR spectroscopy.117 The using far IR spectroscopy (50-300 cm-1) to study
symmetric SO3- stretching band for dry Na+ form Nafion sulfonate membranes that were neutralized
membranes shifted to higher wavenumbers with by cations in the series Na+, K+, Rb+, and Cs+ and
increase in degree of neutralization by this cation but, the series Mg2+, Ca2+, Sr2+, and Ba2+, as well as the
in contrast, shifted to lower wavenumbers with acid form.119 The spectra in this region for hydrated
increasing neutralization with tetrabutylammonium samples show a broad but well-defined band below
cations (TBA+). The latter behavior was attributed 300 cm-1 that is not present for the acid form. For
to diminished polarization of the SO3- groups by the both the monovalent alkali and divalent alkaline
hydrophobic TBA+ cations that have a low surface earth series, the band monotonically shifts to lower
charge density as opposed to the influence of the frequencies, f, such that f ∝ M-1/2, where M is the
stronger electrostatic fields posed by the smaller Na+ cation mass, and the slope of the line is greater for
counterions that have a larger surface charge density. the divalent series. These facts identify this band
Cable et al. also noted two distinct bands in the with harmonic oscillations of the cations in the force
region 1000-950 cm-1, in which resides the absorp- field of the fixed SO3- anions. Moreover, the force
tion envelope for the symmetric C-O-C stretch- constant is essentially the same for each ion in the
ing vibration. This spectral feature was earlier monovalent series and, likewise, the force constant
observed by Heitner-Wirguin112 as well as Lowry and is essentially the same for each ion in the divalent
Mauritz.105 Heitner-Wirguin noted a small shift in series. Based on the slopes of these lines and har-
this band when the Na+ counterion is replaced by monic oscillator theory, the force constants for the
transition metal counterions. The spectra of Lowry members of the monovalent cation series are less
and Mauritz for dry Li+, Na+, K+, and Rb+ forms than those of the divalent cation series. This fact
reflects stronger cation-anion Coulombic interac- high surface charge density. The somewhat linear
tions owing to the greater charge of the divalent ions. plot of type I water versus square of the cation radius
Interestingly, this “ion motion band” is not influ- was taken to suggest that water molecules are not
enced by water as seen in dehydration studies. While only distributed around cation-anion associations,
sample preparation details were given in this report, but a fraction resides close to the perfluorinated
the exact water contents of these samples were not backbone, as earlier suggested by Falk et al.123 and
stated; rather, samples were referred to as “dry”, Yeager et al.124 In fact, The NIR spectrum of PTFE
“partially dry”, or “hydrated”. While far IR spectros- homopolymer showed that type I water is the only
copy has not been exploited in the study of Nafion, it form found in this matrix.
is clear that information regarding cation-anion The molecular mobility of water in Nafion was
binding and vibrational frequencies would be of great investigated using NMR spectroscopy, by Stark-
importance in understanding and modeling ion- weather and Chang,125 Boyle et al.,126 and Bunce et
hopping kinetics, using activated rate theory. al.127 The overall conclusion of these studies was that
Another study involving cation-sulfonate interac- water possessed less mobility than that of liquid
tions in Nafion (1200 EW) is that of Yeager, who water and that there were cooperative motions
investigated the property of equilibrium ion exchange among the molecules and strong interactions with the
selectivity at 25 °C.120 The results of this study are ion exchange groups.
relevant with regard to the use of this ionomer as a Sivashinsky and Tanny, in an early paper, used
chromatographic medium or as a membrane through 1
H NMR to study water in Nafion 125 that had been
which more than one cation is simultaneously trans- boiled and then converted to the Na+ form by soaking
ported. The degree of H+ exchange in the acid form in aqueous NaCl.128 The water content was such that
was determined for mono- and divalent cations and H2O/SO3- was 16:1 (mol/mol), which corresponds to
expressed in terms of ion exchange selectivity coef- a condition where there are a considerable number
ficients, KH+M+, and KH+M2+, respectively, that involve of water molecules in excess of those that would be
the equivalent ionic fraction of the given ion, M+, in tightly bound in hydration shells around the ions. It
the ionomer, the external solution molarity, and the is unclear in this document as to whether excess
cation activity coefficient. One result is that, for the electrolyte was leached from these samples, but it is
alkali metal series Li+, Na+, K+, Rb+, Cs+, as well as a reasonable assumption. The theoretical model of
Ag+, and Tl+, there is a larger spread in KH+M+ values Bloembergen, Purcell, and Pound for the temperature
than that seen for conventional cross-linked sul- dependence of relaxation processes for water ad-
fonated polystyrene ion exchange resins. The spread sorbed on a surface was seen to fit their data.129 From
for the alkaline earth series is smaller than that for the minimum in a T1 (spin-lattice relaxation time)
the alkali metals. In both cases, selectivity decreased versus reciprocal temperature plot at ∼250 K, a
with increasing membrane hydration. Also, it was correlation time (τc ) 1.7 × 10-9 s) was extracted that
seen that exchange of alkali metal cations from was 2 orders of magnitude larger than that for
anhydrous methanol solutions was very similar to supercooled water, but 4 orders of magnitude shorter
that for the aqueous solutions. than that for ice at the same temperature.130 This
Yeager suggests that the major factor involved in was taken to imply that the structure of water that
the ion exchange selectivity of Nafion is the positive is mobile at ∼250 K is more akin to that of a
entropy change associated with the replacement of supercooled liquid or what might be thought of as a
H+ with the metal ion, which is accompanied by glass. Also, free induction decay studies versus tem-
water release and polymer contraction. perature did not reflect freezing in the sense of a first-
Barnes studied the nature of water in Nafion 117 order phase transition. A conclusion based on the T1/
membranes in various alkali metal cation forms T2 value at the T1 minimum was that water molecules
using near-IR (NIR) spectroscopy.121 Advantages of existed in a number of environmental states. Coupled
the use of this spectral range (in this case, over the with heat of fusion data obtained from DSC experi-
wavelengths 1100-2500 nm) are that there are no ments, it was concluded that the presence of small
obscuring CF2 overtone bands and that band shifts “pores” and interaction between the water and the
due to hydrogen bonding are greater than the char- matrix were the most important structure-determin-
acteristic bands that are seen in the mid-IR range. ing factors.
The NIR water bands of interest are the combination Boyle et al. also performed 19F NMR investigations
bands 1890 nm (type I: non-hydrogen bonded water) of the fluorocarbon backbone in 1100 and 1500 EW
and 1920 nm (type II: hydrogen bonded water with acid form samples in the hydrated state in the
bond angles around 110°), as discussed by Luck.122 temperature range -120 to 160 °C.131 No effort was
Both type I and type II water forms were detected, made in sample preparation to remove impurities
and the I/II mole ratio was rather constant at 1:2.2 from these samples, which can be problematic. The
for the Na+ form having the low H2O/SO3- mole interpretations of the results for 1100 EW are as
ratios of 0.06, 0.5, and 1.2. These numbers were follows. T2 data for a low degree of hydration (7%)
derived from the areas under the two deconvoluted showed that the backbone motions are considerably
peaks. The ratio of type I to II water molecules slower than those in the SO2F precursor form in the
decreases in the order for the series Na+ > K+ > Rb+ high temperature region, which was said to reflect
> Cs+, which is reasonable considering that the constraints posed by ion clustering, as there are no
cation hydration number decreases in this order and clusters in the precursor form. For a water content
shows the structure-breaking action of cations with of 25%, there is greater motion in the temperature
range in which these motions are activated, and this beneath the freezing temperature of pure bulk water,
was ascribed to a plasticizing effect of the water, were said to be due to water crystallizing in the
presumably in terms of weakening electrostatic in- membrane. Heating curves showed a single endo-
teractions in the clusters. The T2 behavior for the thermic peak at around 270 K (-3 °C) that was
1500 EW samples up to +120 °C shows greater attributed to water melting. These peaks shifted to
mobility than that for the 1100 EW sample that had lower temperature, broadened, and became more
comparable water content. In essence, the explana- asymmetric with decreasing water content. It is
tion for this was that there are fewer clusters in the reasonable that smaller “ice” crystals that would exist
1500 EW sample that would constrain matrix motion. at lower water content on this dimensional scale
Pineri et al. correlated the results of DSC, NMR, would have lower crystallization and melting tem-
ESR, water sorption isotherms, and dynamic me- peratures for the usual surface/volume consider-
chanical analyses of hydrated acid form Nafion 120 ations, as discussed in crystallization theory.
membranes that had undergone temperature cy- These investigators classified the incorporated
cling.132 They concluded that water sorption proper- water molecules into three categories: nonfreezing,
ties have a strong dependence on temperature. freezing bound, and free water. The freezing bound
Desorption occurs during cooling from room temper- water peak at the lowest temperature was distinct,
ature for contents greater than around 8 wt %, and but the peaks for nonfreezing and free water were
this is followed by freezing. Also, the water vapor difficult to separate and the combined area under
pressure decreases during this cooling to establish a them was simply treated as being due to “free water”.
new thermodynamic equilibrium. Desorption was The heat of fusion of pure water and the areas under
seen to take place above what was then termed the the exothermic peaks were used to calculate relative
“glass transition of the ionic phase” (∼220 K) during free versus freezing bound water content, and the
the heating of a liquid nitrogen cooled sample that total amount of freezing water was estimated from
had 15 wt % water. Earlier, Starkweather and Chang the area under the single melting endotherm peak.
similarly referred to a “glass transition of the aqueous In plots of the amounts of the three types of water
domains” based on dynamic mechanical and dielectric with increasing total water content, it is seen that a
relaxation experiments, as well as proton NMR threshold must be exceeded before freezing water can
spectroscopy.125 After this desorption event (from the exist.
“ionic phase”), the water was said to freeze, as Nonfreezing water molecules were considered to
evidenced by a DSC exothermic peak, and this also exist in hydration shells around the given cation.
results in an increase in elastic modulus. An endo- Presumably, mobility restrictions placed on ion-
thermic peak was said to correspond to a frozen-to- contacting water molecules, posed by the strong
mobile transformation in the hydration micro- electric fields, prevent these water molecules from
structure, and the fractions of desorbable and noncrystalline packing, as in the so-called “structure-
desorbable water were calculated as a function of breaking” effect in simple aqueous electrolytes. The
water content. Pineri et al. were in disagreement amounts of nonfreezing and freezing bound water
with the conclusions of Sivashinsky and Tanny128 in decreased with decreasing “hydration radius”. While
that the former believed that their DSC peaks not explicity defined in this paper, “hydration radius”
reflected a first-order transition having to do with apparently refers to the radius of the bare cation plus
the sorption or desorption of water into or out of the the thickness of a well-defined hydration shell, as this
ionic phase but disregarded the idea of a pore-size quantity was said to decrease in the order K+ < Na+
effect. Characteristic water desorption times and < Li+ and the hydration numbers decrease in this
changes in the number of mobile protons were order.
determined at different temperatures using proton This is another study that illustrates that the
NMR spectroscopy. It should be cautioned that these ensemble of water molecules can be partitioned into
results have significance within the context of the different classes. The state of a water molecule
particular manner in which these samples were depends on whether it is directly interacting with a
prepared. For example, the membranes were not cationsand is a function of cation typesand on the
boiled, as usual, as part of an initialization process. total water content, which, when low, will not provide
Yoshida and Miura also studied the nature of for enough molecules to form distinct crystallites.
water, in terms of freezing and non-freezing fractions, The properties of Nafion at freezing temperatures
in 1100 EW samples in the Li+, Na+, K+, NH4+, and can be quite relevant, for example, within the context
alkylammonium counterion forms, using DSC.133 The of fuel cells in vehicles with regard to cold-starting,
degree of neutralization of the sulfonic acid groups as well as the degradation of membrane/electrode
for all the cations was around 90%, and the measure- assemblies due to the freezing of in situ water.
ments were performed on samples of various water Miura and Yoshida also investigated the changes
contents over the temperature range 100-50 K. For in the microstructure of 1100 EW Nafion sulfonate
example, cooling curves starting at 310 K (37 °C), at membranes, in alkali, ammonium, and alkylammo-
the rate of 10 K/min, for the Li+ form, showed nium cation forms, that were induced by swelling in
exothermic peakssfirst-order transitionsswhose mag- ethanol using DSC, dynamic mechanical analysis
nitudes were enhanced, and the peak temperatures (DMA), SAXS, and electron probe microanalysis
increased with increasing water content. The peak (EPMA).134 These studies were performed within the
widths decreased in this order of cation type but were context of liquid pervaporation membranes that could
not as narrow as that of pure water. These peaks, potentially be used to separate ethanol from water
in azeotropic mixtures. The treatment consisted of of Boyle et al., which indicated increased mobility
drying membranes from a water-swollen state and with hydration.131
an ethanol-swollen state. The membranes were swol- It was suggested that there is a range of temper-
len for 1 week in these liquids at 296 K, after which atures over which SO3H groups go from being fully
they were dried under vacuum at room temperature dissociated to being fully associated upon cooling, and
for 1 week. this temperature range depends on the properties of
The DMA results showed a large difference be- the water at a given degree of hydration.
tween the relaxations of the water-treated and EtOH- For T below this range, a large population of
treated membranes for some of the cation types (and protons appears to be in a solidlike environment and
only slight differences for the others). It should be contribute to the T2 and T1F relaxation.140 The mo-
cautioned that the assignment of the tan δ peaks in tional correlation time, τc, was assumed to have an
this paper should be reconsidered in the light of the Arrhenius dependence on T. On T2 and T1F versus
more recent studies of Moore et al.135 It was sug- 1000/T plots, a change in slope was associated with
gested that these viscoelastic changes reflected a the onset of acid group reassociation. The activation
condition where ionic clusters undergo rearrange- energies for molecular reorientation were rather
ment by swelling in EtOH, and it was deduced, based large, to the extent that up to four hydrogen bonds
on their SAXS analysis, that, when Na+ and am- must be broken to affect this motion. This suggested
monium were the counterions, the sizes of the ion a highly ordered supercooled fluid of hydrogen bonded
clusters in the EtOH-treated-dried membranes (SAXS water, although the molecular mobility of water in
Bragg spacing of 2.8 nm) were smaller than those in this system is much greater than that in ice. The
the water-treated-dried membranes (SAXS Bragg belief was expressed that water forms complex cage-
spacing of 3.3 nm). EPMA sulfur mappings on the like structures, or clathrates, of a number of mol-
surfaces of a Na+ form membrane showed dots about ecules. τc at 253 K monotonically decreased from 6.6
100 nm in size that were attributed to aggregates of × 10-7 to 6.5 × 10-12 s as the average number of H2O
clusters, although the images are not of such a molecules per SO3H group increased from 1.1 to 11.6,
quality as to ascertain the structure of these units. which would indicate greater rotational mobililty. It
A digital image analysis provided information on the was concluded that supercooling occurs at much
number density, size, and shape distribution of these lower temperatures than those for water that resides
cluster aggregates. The results of this analysis were in regular shaped pores,141 and this was attributed
that the number of cluster aggregates in the EtOH- to irregularly shaped clusters within Nafion, a con-
dried sample was less than that for the water-dried cept that was perhaps earliest expressed by Falk.113,118
sample and that the dispersion of cluster aggregates This picture is somewhat different from that of
was more homogeneous with EtOH treatment. These Yoshida and Miura, who spoke of water crystallizing
authors speak of the relaxation of stresses that are in the usual sense. It must be remembered that their
frozen in during membrane processing, as caused by results were for cation exchanged, rather than acid
the incorporation of EtOH, and this is said to be form, samples, and protons from the acid groups can
responsible for the changes in structure and proper- be directly incorporated into the water molecule
ties. structure, whereas cations interact with water mol-
MacMillan et al. studied the dynamics of water ecules by forming hydration shells.133 MacMillan et
molecules in Nafion 117 using 1H, 2H, and 19F NMR al. considered that the high surface/volume aspect of
spectroscopy.136 Special care was taken to remove water clusters inhibits the formation of ice. A simple
impurities from samples, including paramagnetic model that relates correlation time to pore radius
contaminants as well as oligomers and precursor showed that the cluster surface varies as r2.5 rather
fragments, which often does not occur in sample than r2, which means that the surface/volume is
preparation.137,138 A transition in relaxation behavior greater than that on the basis of a solid sphere,
at a characteristic temperature Tt was observed. Tt hence, the term “fuzzy sphere”.
is defined as the temperature at which the slopes of It will be recalled that Sivashinsky and Tanny also
the T1 (spin-lattice), T2 (spin-spin), and T1F (rotating favored the idea of supercooled water whose structure
frame spin-lattice) relaxation times versus 1000/T is influenced by being in pores as opposed to the idea
curves undergo a sharp change. For a hydration level of freezing water.128 While the NMR experiments of
of 15.9%, Tt is approximately -10 °C. Tt shifts to Sivashinsky and Tanny were performed on the Na+
higher values with decreasing degree of hydration. form, τc at ∼ 250 K was 1.7 × 10-9 s, which is in the
Sorbed water molecules behave much as in the bulk midrange of those obtained by MacMillan et al. for
and are not greatly modified by their polymer envi- different water contents. It is difficult to imagine
ronment. The activation energy is small, which water as forming ice in the usual bulk sense in these
reflects high rotational mobility with low hydrogen confined spaces having high surface/volume.
bond density for T > Tt. It is worthy to note that Cation-sulfonate interactions, as well as proton
Escoubes and Pineri reported an endothermic peak mobility, are also expressed in the electrical conduc-
seen in the microcalorimetric studies of Nafion hav- tance behavior of these membranes. Many studies of
ing 15% water at around -15 °C, which is close to this property have been reported, and there is no
Tt.139 attempt in this review to cite and describe them all.
19 Rather, a few notable examples are chosen. Most
F T1 and T2 values indicate that the polymer
relaxation is unaffected by the presence or absence testing is done using alternating current of low
of water. These results differ from the NMR results voltage to avoid complications in the form of chemical
changes at membrane-electrode interfaces that might half of the period of electric field oscillation. The
occur with direct current. effects of percolation and ion pairing are significant
Cahan and Wainright showed that there is an at lower hydration levels.
inherent problem with two electrode cells in that Using a simple electrostatic interaction-based model
membrane/blocking electrode interfacial impedance factored into reaction rate theory, the energy barrier
interferes with the measurement of the bulk imped- for ion hopping was related to the cation hydration
ance at low frequencies and that use of the four radius. The conductance versus water content be-
electrode system eliminates this problem.142 In the havior was suggested to involve (1) a change in the
absence of this interfacial effect, conductivity is rate constant for the elementary ion transfer event
independent of frequency from the dc range up to at and (2) a change in the membrane microstructure
least 105 Hz. These conclusions were reinforced by that affects conduction pathways.
the impedance studies of Fontanella et al. in which Earlier, Gavach et al. studied the “superselectivity”
the electrodes were of both large and small area in a of Nafion 125 sulfonate membranes in contact with
two-electrode configuration.143 Thus, the four elec- aqueous NaCl solutions using the methods of zero-
trode configuration is generally preferred, although current membrane potential, electrolyte desorption
the measured values reflect conductivity along the kinetics into pure water, co-ion and counterion self-
plane of the membrane rather than along the more diffusion fluxes, co-ion fluxes under a constant cur-
relevant thickness direction. There is also the issue rent, and membrane electrical conductance.147 “Super-
of conductivity anisotropy in the plane of the mem- selectivity” refers to a condition where anion trans-
brane that is generated by melt extrusion of films. A port is very small relative to cation transport. The
study by Gardner et al. demonstrated that this is in exclusion of the anions in these systems is much
fact the case.144 There is also the question as to greater than that as predicted by simple Donnan
whether the in-plane conductivity is the same as that equilibrium theory that involves the equality of
along the perpendicular direction. chemical potentials of cations and anions across the
Environmental control that fixes a known concen- membrane-electrolyte interface as well as the prin-
tration of solvent in the membrane under test is also ciple of electroneutrality. The results showed the
important. The reader is encouraged to consult the importance of membrane swelling; there is a loss of
review, and references therein, of conductivity mea- superselectivity, in that there is a decrease in the
surement techniques by Doyle and Rajendran.12 counterion/co-ion mobility, with greater swelling.
Gavach et al. studied the high frequency electrical The situation for hydrated Nafion in the acid form,
resistance of Nafion 117 membranes in various or as containing aqueous acids or strong bases, is
monovalent counterion forms using an ac impedance more complex because protons and “defect protons”
analyzer. For low water contents, the resistance was (i.e., OH- ions), migrate according to a somewhat
seen to increase to very high values as the water different mechanism. Proton transfer in either case
content decreased, which can be attributed at least occurs throughout and between clusters of hydrogen
in part to ion pair formation as oppositely charged bonded water molecules to a degree that depends on
ions become increasingly less separated by water the relative water content.
molecules. The alkaline cation series follows a trend While it is beyond the scope of this review to
opposite to that for the ion mobilities in pure water.145 elucidate details of the current views of proton
Later Gavach et al. performed this analysis for transport across hydrogen bonds in aqueous systems,
various counterion forms as a function of water the reader is referred to the paper by Eikerling et
content.146 Curves for specific conductivity versus al.148 These authors describe the three main options
number of moles of water molecules per SO3- group as follows: (1) An excess proton can be a part of an
exhibited an overall organized trend of upward H3O+ ion in which all of the three protons are
displacement in the following order for the cation equivalent. (2) The proton is placed between the two
forms: Al3+ < Mg2+ < Mn2+ < Ca2+ < K+ < Na+ < water molecules in the hydrogen bond in an H5O2+
Li+ < H+. There is a clear correlation with cation grouping, in the view of Zundel.149 (3) The proton is
valence. a part of an Eigen H9O4+ cluster comprised of an
For a given cation at relatively high degree of H3O+ ion and three H2O molecules strongly attached
hydration, there is a simple monotonic line, in some to each of the three protons of the H3O+ species.150
cases with slight curvature, on the Z′′ versus Z′ Eikerling, in this paper, presents a phenomenologi-
diagram, and this corresponds to a pure resistance, cal model for proton conductivity and mobility in a
not strongly linked to morphology, in a situation hydrated Nafion channel and incorporated this model
where ions can readily hop between clusters in this into a statistical model of a multichannel structure
swollen state. But, when the degree of hydration is consisting of hydrophobic regions throughout which
lowered, there is, in addition, a semicircle that are water-filled channels (pores). A shortcoming of
develops, and this is suggested to be due to a this model is that polymer structure is not factored
capacitive effect owing to an accumulation of ions on into the calculations. Mechanisms for proton trans-
each side of intercluster channels that have been port along the negatively charged surface and “bulk”
significantly narrowed. While not stated in this of a pore are factored in. The surface mechanism has
article, the semicircle can be considered to reflect a a higher activation energy but a higher concentration
form of fluctuating interfacial polarization with a of charge carriers, and the balance between surface
characteristic relaxation time during which coopera- and bulk effects depends on the surface density of
tive charge motions accumulate and dissipate during SO3- groups and pore size. The model accounts for
the effect of proton localization sites, dependent on in other studies, was attributed to water having to
EW, and pore water content and overall swelling. crystallize in very confined spaces.
Theoretical estimates of the membrane conductivity Fontanella and co-workers studied the effect of
agree with experimental data. In another paper, high pressure variation on the conductivity as well
Eikerling and Kornyshev describe a theoretical treat- as the 1H, 2H, and 17O NMR spectra of acid form
ment of activated proton transfer in a single pore Nafion117 membranes that were exposed to various
with sulfonated groups on the surface.151 humidities.157 Variation of pressure allows for a
Aldebert et al. studied the room temperature determination of activation volume, ∆V, presumably
conductance of 1100 EW Nafion membranes using a associated with ionic and molecular motions. Con-
method that also measured the swelling due to a ductivities (σ) were obtained from complex electrical
contacting liquid.152 The membranes tested were in impedance diagrams and sample geometry, and ∆V
the H+ as well as Li+, Na+, and Rb+ forms, and the was determined from the slope of linear isothermal
solvents used were water, ethanol (EtOH), N-methyl- ln σ versus p graphs based on the equation ∆V )
formamide (NMF), and propylene carbonate (PC). -kT[d ln σ/dp]T, where p is the applied pressure. At
Prior to use, the membranes were boiled in nitric room temperature, ∆V was found to be 2.9 cm3‚mol-1
acid, washed with distilled water, and dried. The for a sample conditioned “in atmosphere” and was
conductance, measured versus time from the onset 6.9 cm3‚mol-1 for a sample that was conditioned in
of swelling, showed plots in which there was an 25% relative humidity, where the latter contained the
abrupt rise so that a hold-up time, t0, could be lesser amount of water.
extracted by extrapolating the fast rise portion of the Activation volumes were derived from pressure
curve back to zero conductance on the time axis. This dependent NMR experiments using the equation ∆V
is similar to the time lag method for measuring gas ) -kT[d ln T1/dp]T, where T1 is the spin-lattice
diffusion coefficients in membranes.153 Ultimately, relaxation time. ∆V values for the 1H and 2H NMR
the curves approached asymptotic behavior. Cation experiments were close to each other as well as to
diffusion coefficients were derived using the equation the values based on conductivity. These results imply
x2 ) 8Dt0, where x is the thickness of the membrane that the electrical transport is correlated with water
before swelling. D for membranes contacting water molecule rotation. There is a trend of increasing ∆V
was seen to decrease with increasing counterion with decreasing water content.
radius and with the Guttman donor number of the Paddison et al. performed high frequency (f) di-
solvent. This can be accounted for in the usual sense electric relaxation studies, in the Gigahertz range,
by the fact that larger cations solvate less and so of hydrated Nafion 117 for the purpose of under-
become less mobile. D for the H+ form (1.08 × 10-6 standing fundamental mechanisms, for example,
cm2‚s-1) was close to the value for pure water (0.83 water molecule rotation and other possible processes
× 10-6 cm2‚s-1), and this was taken to mean that that are involved in charge transport.158 Pure, bulk,
protons are highly “solvated”. D for the Li+ form liquid water is known to exhibit a distinct dielectric
decreased in the following order for the solvents: H2O relaxation in the range 10-100 GHz in the form of
> EtOH > NMF > PC. This can be rationalized in an ′′ versus f peak and a sharp drop in the real part
terms of the relative affinity of these solvents for the of the dielectric permittivity at high f.159 A network
cation. Because D is defined within the context of analyzer was used for data acquisition, and measure-
early-time datasthat is, before the conductance ments were taken in reflection mode.
reaches an asymptotesit might be questioned as to Not surprisingly, ′ and ′′ increased with increas-
whether D for the cation forms truly represents a ing water content, but especially when the water
value representative of a membrane in equilibrium content reached 6 and 13 H2O/SO3H (mol/mol). This,
with the solvent because diffusion coefficients in by the usual argument, may reflect a situation where,
swollen systems can be reduced by plasticization at these hydration levels, there are water molecules
effects. beyond those that are ion-bound that are free to
Cappadonia et al. studied the electrical conductiv- rotate and therefore increase the overall polarizabil-
ity of acid form Nafion 117 membranes using an ity. There appear to be no peaks that rise above the
impedance analyzer as a function of temperature and considerable noise in the loss spectra, as seen in the
water content over the range 0-29.8 g of H2O/100 g spectra for pure water. The low f behavior of the ′′
of Nafion.154,155 It was seen that there is a high versus f curves exhibits a 1/ω (ω ) 2πf) dependence,
temperature regime over which the activation energy so that specific conductivity values could be extracted
is low and decreases with increasing water content. in the usual way. These conductivities compare
In a low temperature range, the activation energy is favorably with those obtained by Zawodzinski et al.160
higher and also monotonically decreases with in- Alberti et al. investigated the influence of relative
creasing water content. The temperature for this humidity on proton conductivity and the thermal
conductance discontinuity occurs at around 260 K for stability of Nafion 117 and compared their results
water contents greater than 8 g of H2O/100 g of with data they obtained for sulfonated poly(ether
Nafion. It was expressed that the most probable ether ketone) membranes over the broad, high tem-
cause for this transition was freezing water. This idea perature range 80-160 °C and RHs from 35 to
was said to be reinforced by the DSC results of Chen 100%.161 The authors constructed a special cell used
et al., who noted an endothermic peak in Nafion in conjunction with an impedance analyzer for this
samples, having comparable water content, begin- purpose. Data were collected at high temperatures
ning at ∼253 K.156 The freezing point depression, as within the context of reducing Pt catalyst CO poison-
ing in direct methanol fuel cells at higher tempera- magnitude larger than the usual electronic polariz-
tures as well as the accompanying problem of mem- abilities. Thus, the state of these bonds is very
brane dehydration. sensitive to their environment, and the presence of
Nafion exhibited better proton conductance, espe- ions can induce considerable polarization that sup-
cially at low RH, but the differences diminish or even presses continuous absorption. H5O2+ groupings are
vanish with increasing RH and increasing tempera- linked via H bridges with other H2O molecules or
ture. An interesting finding is that when RH is held with -SO3- acceptor groups. Protons are transferred
constant, the conductivity remains constant over the over larger distances when these groupings shift
temperature range 80-160 °C. It was suggested that within the extended hydration network by thermal
the conductivity increase that would be “expected” rearrangement, and the rate-determining step of H+
(presumably, on the basis of faster proton hopping migration is this structural diffusion.
kinetics in the absence of a matrix)sbut not seens Eigen addressed this topic in earlier literature172
with increasing temperature is counteracted by a although structural diffusion, according to Zundel,
membrane microstructure change that reduces the is more complex than the simple rotation of H3O+ to
mobility and/or concentration of the charge carriers, affect favorable bond orientation for H+ transfer to
and an increase in polymer crystallinity was offered a neighboring H2O molecule. The large polarizability
as a possible cause. of hydrogen bonds in H5O2+ groupings was said to
Alberti et al. also describe experiments dealing account for the high H+ conductivity of hydrated
with reversibility during heating-cooling cycles at SPSA membranes as well as other acidic systems.173
constant RH, as well as membrane swelling and In the absence of an electric field, E, there is a
chemical degradation, and the reader is referred to dynamic equilibrium between two proton boundary
this report for the various interesting details. structures, but when E is present, the weight of
boundary structure II is increased because the hy-
Proton transfer in hydrogen bonds, referred as the drogen bond is easily polarized. Furthermore, the
Grotthus mechanism in older treatises on reaction
rate theory,162 and H2O molecule rotational motion E
as well as the structural reorganization and diffusion 98
within extended hydration structures are important (I) H2OH+- - -OH2 T H2O- - -H+OH2 (II)
fundamental molecular events within the context of
membrane applications that depend on proton con- proton can rapidly transfer within H3O+ to one of the
ductance. Membranes in H2-O2 fuel cells are of great outer H atoms in boundary structure II (H+OH2 f
current interest in this regard. These molecular HOHH+) and another H+ transfer can take place
events also underlie the function of membranes in with a third adjacent H2O molecule and thermal
the arena of membranes in chlor-alkali electrolytic rearrangement causes H5O2+ groupings to become
cells,163 water electrolyzers for the production of H2,164 redefined. This is the essence of so-called “structural
and membranes used in Donnan dialysis processes diffusion”. Because H+ motions are coupled and these
that are used to strip metal ions, such as those of groupings exist within a more extended hydration
copper and nickel, from electroplating waste solu- environment, proton conductivity is in fact a coopera-
tions.165 Yeager and Gronowski provide a good listing tive process174 that is sensitive to structuring influ-
of such applications, a number of which involve ences within the supportive medium. In the case of
Nafion.166 Related to this is the strong acidity of -SO3- - -H+OH2 groupings in SPSA, the time-aver-
Nafion in the protonated form and its use as a solid aged position of H+ is biased toward the water
substrate in organic synthesis as a superacid molecule at low degrees of hydration,175 and if H2O/
catalyst.167-169 SO3H g 2, excess protons will shift from acid-water
Of particular relevance (and often overlooked in the H bonds to H bonds between H2O molecules and the
polymer community) is the work of Zundel et al., who IR continuous absorption and H+ conductivity become
conducted fundamental studies of the relationship great.
between H+ conductivity and the IR spectra of Ostrowska and Narebska noted an infrared con-
aqueous acidic media, including cross-linked sul- tinuous absorption in hydrated acid form Nafion 120
fonated polystyrene (SPSA) in the acid form versus membranes that began at 3400 cm-1 and extended
degree of hydration.170,171 There is an IR spectral toward low wavenumbers.176 This feature was not
feature in the form of a continuous superposition of present in dry membranes and, based on the work
absorbance that begins in the region of O-H stretch- of Zundel et al., was proposed to be due to the
ing for H2O (∼3500 cm-1) and extends to lower existence of H5O2+ and H9O4+ groups, in which there
wavenumbers, which is referred to as “continuous are easily polarizable hydrogen bonds. This paper by
absorption”. In acids this continuum is caused by the Ostrowska and Narebska is also useful, as it contains
rapid fluctuation of protons (rate ∼ 1014 s-1) due to a a number of band assignments for Nafion.
combination of their tunneling and barrier hopping Mauritz and Gray analyzed the IR continuous
in the hydrogen bond of H5O2+ groupings. Interaction absorption of hydrated Na+OH-- and K+OH--imbibed
between hydrogen bonds of different groupings via Nafion sulfonate membranes for the purpose of
proton dispersion forces causes this continuum of correlating this phenomenon to the current efficiency
vibrational energy levels owing to a distribution of (cation transference number) of chlor-alkali electro-
distances between, and orientations of, hydrogen chemical cells.177 In this case, the similar issue of
bonds in the network. The hydrogen bonds in these OH- (“defect proton”) conductivity is important. A
groupings are very polarizable, being 1-2 orders of distinct continuous absorption appeared in the spec-
tra, as seen for membranes equilibrated in different Takamatsu et al. studied the diffusion of water into
high concentrations of aqueous Na+OH- and K+OH-. the acid as well as mono-, di-, and trivalent salt forms
The continuous absorption arises from easily polar- of 1155 and 1200 EW samples.182 The gravimetric
ized hydrogen bonds in H3O2- groupings and reflects uptakes of membranes immersed in distilled liquid
the mobility and concentration of OH- ions as water versus time were determined. Three approxi-
influenced by the electrostatic fields about K+ ions. mate diffusion formulas were applied to the data, and
This view was reinforced by the fact that the con- all yielded essentially the same result. The log D
tinuous absorption versus [OH-] profile correlated versus 1/T plots, over the range 20-81 °C, yielded
inversely with the experimental current efficiency activation energies of 4.9 and 13.0 kcal/mol for the
versus [OH-] profile. Also seen in the spectra was a acid and K+ forms, respectively. Diffusion coefficients
shoulder on the high wavenumber side of the main of various mineral cations that permeated from
O-H stretching peak which is the signature of nonaqueous electrolytes were considerably smaller than
hydrogen bonded OH groups. that of water. Also, log D was seen to be proportional
A more recent view of proton transport is that of to the quantity q/a, where q is the charge of the cation
Kreuer, who, compared with the Zundel-based view, and a is the center-to-center distance between the
describes the process on different structural scales cation and fixed anion in a contact ion pair.
within phase separated morphologies. The smallest Yeager and Steck derived diffusion coefficients for
scale is molecular, which involves intermolecular water in totally hydrated Nafion120 membranes that
proton transfer and the breaking and re-forming of were exchanged with alkali metal cations, using a
hydrogen bonds. When the water content becomes radiotracer technique.18 At 25 °C, D for the Na+ form
low, the relative population of hydrogen bonds de- was 2.65 × 10-6 cm2/s and the values for the K+ and
creases so that proton conductance diminishes in a Cs+ forms were somewhat smaller, which would seem
way that the elementary mechanism becomes that to reflect the lower maximal degree of hydration of
of the diffusion of hydrated protons, the so-called these forms.
“vehicle mechanism”.178,179 Yeager et al.,183 again based on radiotracer studies
The next level is concerned with transport within of the self-diffusion of different ions (22Na+, Cs+, 36Cl-,
“channels” as depicted in Figure 11. The condition 35
SO4-, 125I-) in 1150 and 1200 EW Nafion mem-
in these regions is described in terms of charge dis- branes that were in contact with dilute and concen-
tributions as deriving from solutions of the Poisson- trated electrolytes, were the first to introduce the
Boltzmann (P-B) equation. This P-B approach can concept of mixed interphase regions of intermediate
be questioned for a number of reasons, the first of polarity. In dilute solution, room temperature stud-
which is that the equation is rooted in macroscopic- ies,184 Cs+ and I- ions were rationalized as having
continuum-based electrostatic theory. In spaces that more tortuous diffusion pathways than Na+ ions and
are only somewhat greater than 1 nm in width, there water molecules. For concentrated solutions at 80 °C,
is molecular granularity rather than the required Cl- and SO42- ion diffusion coefficients are smaller
continuum and the dielectric constant is a ubiquitous than that those for Na+ ions, and D for Cl- ions is
parameter although approximate treatments can be greater than that for the SO4- anions.
made. Moreover, the imposition of a simple geometry In this conceptual (nonmathematical) model there
for the channels for the purpose of solving the P-B is the view of hydrophobic semicrystalline perfluoro-
equation ignores the complexity of the regions that carbon regions and irregular-shaped aggregates of
are, in fact, ill-defined and most likely of complex fixed ions. Small cations and water molecules reside
shapes. mainly in the ion cluster centers and diffuse along
The highest level, at structural scales > 10 nm, is different pathways than those for large cations hav-
that over which long-range transport takes place and ing low hydration energies and anions which were
diffusion depends on the degree of connectivity of the thought to preferentially migrate through an inter-
water pockets, which involves the concept of percola- mediate phase containing a few water molecules and
tion. The observed decrease in water permeation with a few side chains. The high diffusion rates and low
decreasing water volume fraction is more pronounced tortuosity of Na+ ions and water molecules are
in sulfonated poly(ether ketone) than in Nafion, explained by the nonspherical shapes of the clusters.
owing to differences in the state of percolation.180 This view stands in contrast to the condition of
Proton conductivity decreases in the same order, as having well-defined spherical clusters and sharp
well. phase boundaries as depicted in the model of Gierke
Another important property of Nafion is that of et al.
water diffusion. A number of studies of this property Millet determined self-diffusion coefficients for Na+
have appeared in the literature, and some notable and Cs+ ions in hydrated 1200 EW membranes using
examples are as follows. conductivity measurements and the Einstein equa-
Yeo and Eisenberg investigated the diffusion of tion, D+ ) u+kT, where u+ is the absolute mobility
water in Nafion by sorption from the contacting liquid of the given cation.185 u+ can be derived from the
into dry samples and measuring weight uptake equivalent conductivity according to Λ ) σ+/C+ )
versus time over the temperature range 0-99 °C.181 Fu+, where σ+ is the specific conductivity, C+ is the
The resultant diffusion coefficients (D) increased from cation concentration (calculated on the basis of the
about 10-6 to 10-5 cm2/s with increasing temperature, dry membrane density, EW, and the water content),
and the activation energy was determined to be 4.5 and F is the Faraday constant. The values of D+
kcal/mol. determined via these conductivity measurements
were compared with those determined by the radio- root of time plots for spin diffusion. The spin diffusion
tracer technique of Yeager et al.186-188 to test the results support the idea that the side chain phase is
validity of the Einstein equation for this system. The continuous with little tortuosity, which accounts for
water contents of these membranes corresponded to the rapid long-range transport of water and lower
H2O/SO3- mole ratios in the range 6.6-11.3 for the alcohols.
Cs+ form and 11.9-18.4 for the Na+ form at 25 and 129
Xe gas, under pressure, was used by Jones,
40 °C. The results showed that the D+ values Inglefield, and co-workers as a probe of morphology
compared well with those of Yeager et al. based on the fact that its spectrum is sensitive to
Zawodzinski et al. determined 1H diffusion coef- local molecular environment. Domains that are in-
ficients, D, in Nafion 117 having water contents over volved in penetrant uptake and diffusion can be
a range corresponding to H2O/SO3H ) 2-14 mol/mol inferred. The studies showed two overlapping reso-
using pulsed field gradient spin-echo 1H NMR nances, one being assigned to an amorphous per-
spectroscopy.189 This study was conducted within the fluoroethylene environment and the second to side
context of understanding the nature of water as well chain domains. The latter regions are more hetero-
as water concentration profiles and water manage- geneous, owing to the considerable line width.
ment in fuel cell membranes. D was extracted from Morphology based on chemical environment can be
the NMR data using the following equation: probed using 19F NMR spectroscopy because the
[ ] ( )
chemical shifts of F atoms in the side chains are
A(g) δ
ln ) -γ2Dg2δ2 ∆ - considerably separated from those in the backbone.
A(0) 3 Conformational dynamics as affected by domain-
selective solvent incorporation are reflected in the
A(g) is the signal intensity observed with an applied
widths of static 19F peaks. These conformational
gradient g, A(0) is the intensity in the absence of an
motions, in turn, can influence the migration of
applied gradient, γ is the nuclear gyromagnetic ratio,
solvent penetrants.
and δ and ∆ are time intervals of the pulsed field
gradient spin-echo sequence. Spin diffusion between the backbone and side chain
D, measured in this way, ranged from 0.6 × 10-6 resonances can be used to study domain size. For dry
to 5.8 × 10-6 cm2/s over this range of water content Nafion the side chain domains were found to be of
in increasing order. the size 3.8 nm and these domains had a periodicity
While D issuing from these experiments is not of around 10 nm. The first number is around that
strictly the diffusion coefficient of water per se, but commonly associated with the diameter of a spherical
rather that of 1H throughout the ensemble of envi- cluster based on microscopic and scattering studies.
ronments in the hydration microstructure, these The 10 nm spacing remains to be assigned to a
authors rationalized that it could in fact be identified structural feature. It is noted that the AFM studies
with D at both high and low water contents. It should of McClean et al.,83 described earlier, identify fluoro-
be appreciated that self-diffusion coefficients mea- carbon crystalline domains of size ∼10 nm for the K+
sured in this way reflect fundamental hopping events form, and whether the coincidence of this number
on a molecular scale. with the periodicity deduced by Meresi et al.191 is
Later, Jones, Inglefield, and co-workers performed significant or fortuitous remains to be shown. Upon
fundamental solid-state NMR studies of acid form addition of 20 wt % water, domain size increases to
1100 EW Nafion. Pulse field gradient NMR experi- 6.5 nm, but the periodicity is essentially unchanged.
ments interrogated the translational motions of An ethanol uptake of 20 wt % causes a greater
water, ethanol, and fluorocarbon components in their morphological rearrangement, giving a domain size
systems.190 19F spin diffusion studies yielded impor- of 11 nm and a periodicity of 19 nm. The repeat
tant information on the size of the backbone (CF2)x length is calculated using spin diffusion plots (linear
and perfluoroalkyl ether regions in the hydrated and extrapolation to zero magnetization on magnetization
dry conditions.191 Static (nonspinning) 19F studies vs time plots). Static 19F line shapes showed that
indicated the location of molecular penetrants within ethanol selectively plasticizes the side chain group
the nanophase-separated morphology. The reader is domains.
referred to the original paper for the explanations of NMR is the most fundamental molecular-specific
this method, model assumptions, and so forth. One probe of diffusion. Polymer motions and the spectro-
basic assumption is that the side chains form a scopic signature of a given nucleus can be unambigu-
distinct domain separate from a distinct domain ously related to a particular morphological domain.
consisting of the perfluorinated polymer backbone. The size and time scale of the experiments are such
The interpretation of the results is as follows. The that the fundamental hopping events of diffusing
spin diffusion results indicated domains in dry and molecules can be sampled.
solvent-swollen samples that are in the nanometer Plots of D versus water volume fraction190 show
range, and this is commensurate with pendant group that the concentration dependence of D is in fact
aggregation, which is in harmony with the structural described well by the Fujita free volume equation.190
view issuing from SAXS experiments on the same This was surprising considering that the underpin-
material. Although the NMR information does not ning of this equation simply involves available free
provide information regarding the shape of the ag- volume for molecular hopping. The interpretation is
gregates, interfaces are detected. The signature of an that water molecules plasticize the perfluoroalkyl
interface is graphical curvature that precedes the ether side chain domains and this increases D with
linear section of magnetization change versus square increasing water content. D for water varied from
10-4 to 10-7 cm2/s, and these values are similar to

those determined by Zawodzinski et al., as discussed
above. Moreover, D for water versus temperature
followed WLF behavior, which is also linked to free
volume theory. D for ethanol was comparable to that
for water although the concentration dependence was
stronger and the free volume concept was not found
to apply. This can be related to the fact that 19F spin
diffusion and line shape analyses indicated larger
morphological changes upon addition of ethanol as
compared to water and greater plasticization of the
side chain domains.
Sodaye et al. attempted to probe the free volume
in various monovalent and divalent counterion forms
of Nafion 117 that were swollen in water and alcohol
using positron annihilation lifetime spectroscopy
(PALS).192 The underlying concept is that free volume
on the molecular level can influence polymer chain
conformation and chain dynamics, which will affect
transport and mechanical properties. While other
techniques, such as gas diffusion, can indirectly
interrogate materials for free volume, PALS is cur-
rently the only direct probe on an atomistic level. The
details of this method will be omitted here, and it
will simply be mentioned that the spectroscopic data
are fitted to a quantum mechanical model that
assumes spherical holes, of radius R, that are meant
to represent free volume. The assumption of spherical Figure 13. Dependence of cluster diameter and number
holes is a limitation considering that the free space of SO3- groups per cluster on water content in 1200 EW
between polymer chains must not only be of complex Nafion according to Gierke, Munn, and Wilson. (Reprinted
geometry but must also be contiguous rather than with permission from ref 17. Copyright 1981 Wiley.)
existing in isolated discrete pockets of a dynamic Also, in support of the Litt model, Plate and
nature. In short, there was seen to be a very small Shibaev observed that hydrated membranes behave
monotonic increase in R from 3.49 to 3.55 Å in in a fashion as brushlike polymers. This suggested
proceeding from the H+ to Cs+ form. The free volume to these authors that Nafion has a multilayer struc-
fraction was determined to be essentially unchanged ture such that water forms aggregates in lamellar
across the range of alkali cations. The significance domains,193 and this view was said to be supported
of these results can be questioned, given that the by the results of the neutron diffraction and Moss-
range of solvent swelling as well as solvent and ion bauer specroscopic experiments of Timashev.194
diffusivity varies significantly over this range. Also, This structural view can be compared with that of
the difference in R between the water- and ethanol- Gebel et al., who conducted SAXS and SANS studies
swollen H+ forms (3.49 vs 3.67 Å) is somewhat of water-swollen (around 10-90 polymer volume
insignificant. Thus, it must be concluded that these percent) Nafion 117 in different cation forms.195 The
PALS experiments have not yielded information that picture emerging from their studies consists of the
correlates strongly with these critical properties of aggregation of ionomer chains into elongated bundles
water-swollen Nafion, and this may be due in part which can be in the form of cylinders at high water
to the overly simplistic model that was applied to the content (2-D swelling) or ribbons (1-D lamellar swell-
data. ing). Gebel et al. make the statement that “for high
A controversial model for the structure of Nafion (polymer) volume fraction...the dilution process is
is that which proposes a lamellar morphology, due similar to those of a lamellar structure....” These
to Litt.70 Litt makes a number of arguments to bundles have diameters that are ∼40 Å, have lengths
support this model, but the most direct evidence greater than 1000 Å, and are surrounded by “elec-
consists of plots of the Bragg spacing, d, associated trolyte”. This is an extension of the view of elongated
with clusters, versus water content, as seen in Figure aggregates in dilute perfluorosulfonate ionomer solu-
13. The plot comes from the paper by Gierke et al.,17 tions. As noted earlier, in the scattering section of
although this d versus water content behavior was this review, Nafion does not form true solutions in
also observed by Fujimura et al.37 Litt points out that the sense of having individual polymer chains freely
a lamellar structure would swell in only one dimen- intermixed with the solvent molecules.
sion, which accounts for the linear d versus water Another spectroscopic method that has been used
content graphs. Swelling, on the microscopic level, to study the multiphasic structure of Nafion uses
occurs by having water incorporating between the photophysical probes. This is well explained in a
lamellae, thereby pushing them farther apart. The paper by Kalyanasundaram and Thomas, and refer-
bilayer structure presented by Starkweather41 is in ences therein.196 These authors studied the perturba-
fact commensurate with the model of Litt. tions of the five predominant vibronic band intensi-
ties in fluorescence spectra of pyrene (Py), an aromatic is, whether all of the sulfonic acid groups have been
chromophore, in a broad variety of organic solvents neutralized by the chosen cations of charge +z e, as
whose molecules have different dipole moments, as ion content can affect swelling, mechanical, and other
well as in micelles. The ratio of the intensity of peak properties. Young et al. determined the extent of ion
3 to peak 1 (I3/I1) decreases with increasing environ- exchange in Nafion 117 membranes in the K+, Cs+,
mental polarity, and this ratio was the highest for Mg2+, Ca2+, Co2+, Zn2+, Cu2+, Al3+, and Fe3+ forms
perfluorinated saturated hydrocarbons. using cold-neutron-capture prompt γ neutron activa-
Lee and Meisel incorporated Py, at levels of 10-3 tion analysis (PGAA).200 PGAA is particularly suited
M or more, into 1200 EW acid form samples that for evaluation of chemical composition, being sensi-
were swollen with water and with tert-butyl alco- tive to trace ions and hydrogen, and is a nondestruc-
hol.197 It was concluded based on the I3/I1 value for tive technique. In short, the nuclei of many elements
water swollen samples that the Py molecules were in a sample placed in a neutron beam absorb neu-
located in the water clusters and were most likely trons and become an isotope of higher mass number
near fluorocarbon-water interfaces. It was also in an excited state. Then, prompt γ rays are emitted
concluded, based on both absorption and emission by de-excitation of the compound nuclei and mea-
spectra, that the probes had strong interactions with sured via a high resolution γ ray detector and the
the SO3- groups that were exchanged with Ag+ and data appear as peaks on counts versus energy
Pb2+ cations in the case of water containing samples. spectra, each element having a characteristic peak.
Likewise, the pyrene molecules were rationalized as Four forms of membrane pretreatments were used,
being surrounded by tert-butanol molecules in that and the reader is referred to this report for the details
case. However, excimer formation (due to the pres- of each. The as-received sample was seen to have K+
ence of adjacent pyrene molecules) in the tert-butyl contaminant, perhaps being a residue from conver-
alcohol system suggested the loss of cluster morphol- sion from the SO2F form using KOH, but the samples
ogy. that were pretreated using mineral acids were com-
Szentirmay et al. studied the microchemical envi- pletely converted to the SO3H form and the K+
ronments of Nafion 117 in the acid and Na+ forms contaminant was removed in the process. The result-
using Py and Ru(bpy)32+ probes in fully hydrated ant acid forms were soaked in dilute (0.01 M) and
(∼40%) samples in various cation forms.198 Ru(bpy)32+ concentrated (2 M) metal chloride salt solutions, and
emission spectra cannot be interpreted in terms of excess electrolyte was leached out of the latter by
environmental polarity in as straightforward a fash- refluxing in DI/DS water. The peak intensity for a
ion as in the case of Py, but blue shifts can reflect given atom was normalized to the peak intensity of
this aspect. One of the results of this study was that sulfur, so that the average counterion/SO3 (mole/
the microenvironment polarities were such that I3/I1 mole) ratio could be ascertained. In short, 94-100%
values for Py are between those for fluorocarbon and of the SO3H groups can be exchanged and conver-
aqueous environments, and this conclusion was sions are slightly lower for the monovalent cations.
strengthened by the results of Ru(bpy)32+ probe This might be due to the lower hydration numbers,
studies, as well as the similar conclusion of Lee and especially for the Cs+ cation, and no data were
Meisel. Another conclusion that was reached was reported for Li+ and Na+ ions that have greater
that the SO3- clusters are chemically heterogeneous, hydration numbers. Also, it was shown that the
an idea that was in line with the view of Yeager and excess chloride salts are in fact leached out of the
Steck, who spoke of mixed interfacial regions.18 membrane by refluxing with DI/DS water.
Robertson and Yeager used Py and Ru(bpy)3 probes Other experiments were performed to affect partial
for the purpose of locating the locations of Cs+ and ion exchange by soaking membranes in 0.01 M
I- ions in the nanophase-separated morphology.199 It electrolytes in which the mole ratio of cation-to-
is known that these probes take residence in the sulfonate group was 1:2, 1:1, and 2:1. It was found
intermediate polarity hydrophobic-hydrophilic in- that an external electrolyte concentration of 0.01 M
terfacial regions. The studies concluded that Cs+ ions was not high enough to totally convert all of the
were located in the aqueous regions, but I- ions were sulfonated groups to the desired cation form. An
in the interfacial regions. excess of 2:1 was found to be necessary for this
It is perhaps a shortcoming of mobile probes that purpose.
one must deduce where it is located based, for Increasing z increased the degree of exchange, as
example, on the measured value of I3/I1 in the case more than one sulfonate group can interact with a
of Py, and this presupposes that the basic features single cation.
of the morphology are already known. One could not These experiments are of significant value, as
assign the probe location as being cluster domains if many studies have been reported in which complete
such domains had not been known to exist before- ion exchange was assumed to take place but not
hand. Furthermore, the issue of whether large probe verified.
molecules perturb their molecular surroundings is As mentioned, this review is focused primarily on
the subject of debate. It would seem that the greatest a survey of the vast literature dealing with the
benefit derives from establishing a semiquantitative structure and properties of Nafion in the sulfonic acid
degree of polarity of the phase in which the average and cation exchanged sulfonate forms. The literature
Py molecule resides. on the carboxylate version is sparse and currently of
It is important to know whether complete ion lesser interest, as its application seems to be limited
exchange has been affected in a given sample, that to membranes in chlor-alkali cells, and since it is a
less-strong acid than SO3H, is not of interest in the related to mechanical properties under considerable
arena of fuel cell membranes. It should be mentioned, deformations within an application context, this
however, that the carboxylate form is considerably information was used to identify thermomechanical
more acidic than analogous hydrocarbon ion ex- transitions that were assigned to morphological
change resins. In this regard, the work of Yeager et features in the microphase separated morphology.
al. in comparing cation and water self-diffusion The following is a historical presentation of these
coefficients, found by a radiotracer method, for the investigations.
sulfonated and carboxylate forms of 1200 EW, is of In an early study, Yeo and Eisenberg noted a
importance.124 In short, it was found that when the transition, labeled R, for the acid form (EW ) 1365 g
membranes are in contact with 0.10 M aqueous mol-1) at around 110 °C (tan δ maximum).204 This
solutions of sodium and cesium chloride containing relaxation was initially considered as the glass
the 22Na and 137Cs isotopes in each respective case, transition of the nonionic phase because, in these
cation diffusion proceeds faster in the carboxylate limited studies, water was seen to have only a minor
form, although this form has the lower water uptake. effect on the magnitude or position of this peak. A β
Yeager et al. also investigated and compared the peak for the same dry acid form was seen at around
sorption and transport properties of these two forms 20 °C, which shifted to lower temperature with
in concentrated aqueous NaCl and NaOH solution increasing water content, and this was discussed in
environments, and the differences were rationalized terms of a glass transition of the polar regions. A γ
in terms of ion pairing.201 The apparent pKa of the peak at around -100 °C was mentioned as being of
carboxylate form was determined to be 1.9, which is the same origin as in PTFE, that is, due to short-
rather low and due to the electron withdrawing effect range molecular motions in the tetrafluoroethylene
of fluorine atom substitution. Consequently, the phase.
carboxylic acid form can be totally neutralized in Later, Kyu and Eisenberg discussed dynamic me-
alkali electrolyte solutions. chanical relaxations for the same acid form of Nafion
The in-depth FTIR investigations, of Perusich, of and re-examined their earlier interpretations.205 The
the methyl ester (-CO2CH3), carboxylic acid, and γ relaxation retained its original interpretation.
potassium carboxylate forms should also be men- However, the R peak, having the greatest intensity,
tioned.202 In addition to constructing a useful exten- was reassigned as the glass transition of the polar
sive tabulation of band assignments for these as well regions because it had by then come to be seen as
as the sulfonate materials, Perusich established sensitive to ion type, degree of neutralization, and
quantitative expressions that allow for the computa- water content. The β peak was reassigned as the
tion of equivalent weight and acid content using the glass transition of the Nafion matrix.
555 cm-1 C-F and 982 cm-1 C-O-C ether bands,
Tant et al. reported a dynamic mechanical transi-
based on thin film absorbance measurements. This
tion of around 100 °C (maximum in G′′) for acid form
information is very useful for the accurate, reproduc-
Nafion having 1140 EW.206 Since this transition also
ible, and rapid compositional measurements of this
appeared for the sulfonyl fluoride precursor, but at
polymer. Also, the time evolution of bands during the
a much lower temperature (∼0 °C), they concluded
conversion from the methyl ester to acid forms was
that it involved main chain motions that are re-
reported.
stricted by the conversion to the acid form. These
motions were further restricted by the conversion to
5. Mechanical Properties the Na+ sulfonate form owing to strong ionic associa-
The mechanical properties of Nafion materials tions between the side chains. In contrast with the
have not been of the most critical importance, as in work of Kyu and Eisenberg, no transition appeared
the case of commercial thermoplastics or composite at 0 °C in addition to that at 100 °C. While the
materials that are expected to be load-bearing. equivalent weights of the samples utilized by Eisen-
Rather, the primary focus has been on transport berg and Kyu and Tant et al. were not quite the
properties. To be sure, the mechanical integrity of same, the notable difference in matrix Tg assignment
membranes as mounted in cells, and under the is cause for confusion.
perturbation of pressure gradients, swelling-dehy- Miura and Yoshida207 noted a dynamic mechanical
dration cycles, mechanical creep, extreme tempera- tan δ peak at 120 °C at 1 Hz, labeled R, for “dry” 1100
tures, and the onset of brittleness and tear resistance, EW acid form Nafion. These investigators assigned
is important and must be taken into consideration. this transition to motions within the polar clusters
The E. I. DuPont Co. presented empirical equations because it was sensitive to cation type.
for the tensile strength, elongation at yield, and Cable and Moore performed DMA (dynamic me-
tensile modulus as a function of cation type, water chanical analysis) studies of various Nafion mem-
content, and equivalent weight in the range 1000 e branes including the acid form.208 A tan δ peak with
EW e 1400 and at temperatures in the range 0 e T maximum at 110 °C, referred to as “Tg”, was seen,
e 85 °C.203 These equations were shown in early and there is a suggestion of a shoulder on the low
product literature, as well. temperature side that might arise from another
Aside from this, the literature on the subject has mechanism. As this membrane was dried at only 60
largely been concerned with dynamic mechanical °C, the possibility of residual water incorporation
properties where experiments have been performed exists. Moore and Cable concluded that the R relax-
to gather data consisting of loss tangent (tan δ) and ation was due to chain motions within and/or near
storage tensile modulus (E′). Rather than being the ion-rich domains and that the β relaxation was
the strong dipole-dipole interactions between the

SO3-Na+ groups.208
To further investigate the molecular origins of the
R and β relaxations in Nafion, Moore and co-workers
utilized a series of alkylammonium ions to systemati-
cally alter the dynamic mechanical response of the
ionomer with respect to the strength of electrostatic
interactions. When Nafion is neutralized to contain
large tetrabutylammonium ions, the R and β relax-
ations are observed to shift to temperatures even
below that of the acid form of the ionomer, as shown
in Figure 14. Moreover, the magnitude of the β
relaxation is seen to increase significantly to a level
comparable to that of the R relaxation. This behavior
was attributed to a weakening of the electrostatic
interactions and a plasticization effect of the large
Figure 14. Dynamic mechanical data plotted as tan δ organic counterions.211 Using a series of alkylammo-
versus temperature for Nafion in various forms. nium ions, from tetramethylammonium (TMA+) to
tetradecylammonium (TDecA+), the R and β relax-
due to chain motions of amorphous fluorocarbon units ations were found to shift monotonically over a range
well removed from the ionic aggregates, that is, the of temperatures from near that observed for the
matrix.135 These investigators referred to chain mo- sodium form of the ionomer (i.e., with the small
tions within an electrostatic network rather than TMA+ form) to near that of the R relaxation of the
viewing the polar and fluorocarbon components as sulfonyl fluoride precursor (i.e., for the large TDecA+
being isolated, discrete domains. This is in harmony form). In addition, the magnitude of the β relaxation
with the view of Kyu and Eisenberg, who stated that was observed to increase with counterion size to a
“...whatever happens in one of the phases would be point where it becomes the dominant relaxation for
expected to have a strong effect on the other phase counterions larger than tetrahexylammonium. Based
because of the intimate contact between the phases on these results and recent correlations of this DMA
coupled with their small size.” Mauritz and Young data with studies of molecular dynamics and mor-
adopted this view in their DMA studies of Nafion/ phological relaxations observed using variable tem-
[organically modified silicate] nanocomposites.209 perature SS NMR and SAXS experiments,212, 213 the
The effect of counterion type and size on the R relaxation is attributed to the onset of long-range
dynamic mechanical properties of Nafion has been mobility of both the main and side chains as a result
studied recently by Moore and co-workers135,208 using of a destabilization of the electrostatic network (i.e.,
a variety of organic, alkylammonium ions in com- through the activation of a dynamic network involv-
parison to the more conventional alkali metal ions. ing significant ion-hopping processes). In contrast,
Figure 14 compares the tan δ versus temperature the β relaxation is associated with the onset of
plots of the Nafion precursor (i.e., the nonionic, segmental motions of chains within the frame-
sulfonyl floride form) to those of the ionomer in the work of a static network of physically cross-linked
H+, Na+, and tetrabutylammonium ion (TBA+) forms. chains.214
In the absence of electrostatic interactions, the sul-
fonyl fluoride precursor displays a single R relaxation 6. Molecular Simulation of Structure and
near 0 °C. As a result of this low temperature relax-
ation, the precursor may be easily melt-processed into
Properties
thin films for later membrane applications. However, Mauritz summarized a number of molecular mod-
once the polymer is converted to the sodium sulfonate els of ionomer structure, including those pertaining
form (by hydrolysis and subsequent neutralization), to Nafion, that had had been formulated up to
the ionomer yields a profound shift in the R relax- 1996.215 Within the context of the title of this review,
ation to a temperature near 250 °C. Note that this it should be appreciated that the results contribute
form of the ionomer is no longer melt-processible due to the “state of understanding” only if they are
to strong Coulombic interactions that yield a dynamic verifiable by careful experimentation. To be sure,
electrostatic network (i.e., a physically cross-linked theoretical predictions are welcome in the design of
system) that persists to temperatures well above the experiments and pointing the way toward useful
melting point of the PTFE-like crystallites. In addi- applications.
tion to the dominant R relaxation, a weak shoulder The cluster-network model of Gierke et al. has
near 150 °C is observed and assigned to the β already been discussed in the Introduction as being
relaxation, in agreement with the work of Eisen- the first realistic model for rationalizing a number
berg.210 For Nafion in the acid form, the R and β of properties of Nafion membranes.
relaxations are observed at temperatures ∼100 °C Hsu, Barkley, and Meakin addressed the percola-
lower than that for the sodium sulfonate form of the tion aspect of hydrated clusters in relation to insula-
ionomer. This shift to lower temperatures has been tor-to-conductor transitions.216 As the concentration
attributed to a reduction in the strength of the of clusters on a hypothetical grid, that is, a three-
interactions between the SO3H groups (i.e., compa- dimensional lattice, increases, “islands” of clusters
rably weak hydrogen bond interactions) relative to will grow in size and become interconnected. Eventu-
ally, a threshold volume fraction, CO, is reached were performed with an electrical impedance ana-
where the pathways span the macroscopic dimen- lyzer that measured conductivity, σ, vs ω for different
sions of the gridsthe percolation channelssand long- φ values for an applied sinusoidal field.
ranged ionic conduction is possible. The conductivity, The rheological properties change behavior, rela-
σ, near and above CO obeys the following power law. tive to more dilute solutions, above φ ) 0.2, where
non-Newtonian behavior is then exhibited. The power
σ ) σO(C - CO)n law dependence of η on φ is in harmony with the
Zimm rather than the Rouse model, which suggests
that hydrodynamic interactions between these poly-
This is a general equation that can be applied to any
mers, in a mean field sense, are important. Electrical
material that can support charge percolation. CO not
properties also begin to deviate for Nafion solutions
only depends on dimensionality but also the manner
above φ ) 0.2, and mechanical percolation is es-
in which the two components (one insulating, the
sentially the same for the sodium and acid forms.
other conducting) are dispersed, and the factor (C -
CO)n involves topological connectivity. The prefactor An interesting conclusion is that, as opposed to the
σO depends on the particular conduction mechanism Na+ form, there is no electrical percolation threshold
that operates within and between two adjacent for the proton form, although conductive clusters
clusters and involves interactions among ions, water must grow with increasing polymer volume fraction
molecules, and the ionomer. It is the one factor that in both cases. In short, no clusters that span macro-
involves the chemical identity of the system. This scopic dimensions form at a critical volume threshold
equation was tested by experimental measurements for the acid form. These results are interpreted in
of σ for membranes of different equivalent weights two sketches, each having connected polymer strands
and containing various uptakes of aqueous sodium above the percolation threshold so that there is
hydroxide. mechanical percolation in either case. In one picture
In short, CO was found to be 0.10, which is less than (Na+ form), the sulfonate groups are evenly spaced
the value for a completely random systemswhich so that electrical percolation exists, but for the acid
implies that the clusters were not randomly dis- form, these groups are more closely spaced in ag-
persed. A log-log plot of σ versus C - CO was found gregates so as to cause voids along electrical path-
to be linear where n, the slope of the line, was 1.5, ways. It would need to be explained how, upon
which in fact is in the established theoretical range drying, the acid form gel evolves to form highly
of 1.3-1.7. proton conducting membranes.
Of course, this equation and its theoretical under- The concepts and techniques discussed by Cirkel
pinnings does not constitute a model as such and and Okada are relevant with a view toward modify-
certainly does not address the structural specifics of ing the structure of solution cast Nafion membranes
Nafion, so that it is of no predictive value, as by manipulating counterion type, solvent, tempera-
experimental data must be collected beforehand. On ture, and other variables.
the other hand, the results of this study clearly Hsu and Berzins used effective medium theories
elucidate the percolative nature of the ensemble of to model transport and elastic properties of these
contiguous ion-conductive clusters. Since the time of ionomers, with a view toward their composite nature,
this study, the notion of extended water structures and compared this approach to that of percolation
or aggregated clusters has been reinforced to a degree theory.219
by the morphological studies mentioned above. Hsu and Gierke presented an elastic theory for the
In a related experimental study, Cirkel and Okada clustering of ions in hydrated Nafion.31 This model
compared mechanical and electrical percolation that presumes the existence of clusters and is uncon-
developed during the gelation of 3:1 (v/v) 2-propanol/ cerned with the process of how clusters are actually
water “solutions” of Nafion 117 in the acid and Na+ created. Rather, it was intended to determine the
forms.217 Attention should be paid to the particular equilibrium diameters of hydrated clusters given this
manner in which these samples were prepared, as assumed morphology. The equilibrium hydrated clus-
different conditions may yield different results. Also, ter diameter results from a minimization of a free
caution should be applied in comparing these results energy composed of an elastic term and the non-
with those of percolation studies using preformed specific interactions: SO3--CF2, SO3--SO3-, H2O-
films, such as that of Hsu et al.216 H2O, H2O at cluster surfaces, H2O in the second layer
In relatively dilute solutions, Nafion is said to exist from the cluster/hydrophobic surfaces, and H2O in
in the form of rodlike aggregates of practically the the bulk state. A severe limitation of this model is
same size49,51 that are in equilibrium with loose that it is semiphenomenological and therefore of little
ionomers73 such that at higher concentrations the predictive value. This is so because prior knowledge
aggregates dominate.218 The reader should consult of the dependence of the tensile modulus upon
the more recent papers of Gebel et al., cited earlier, counterion type, water content, EW, and the diameter
for a more detailed view of Nafion “solutions”. of an average dry cluster determined from SAXS
The relative viscosity ηr and storage modulus G′ studies, as well as the details of the experimental
were determined by Cirkel and Okada in experiments water vapor pressure sorption isotherm, is required
using a rheometer in oscillatory rotational mode and in the calculations.
Couette sample geometry as a function of Nafion Mauritz and Rogers constructed a water vapor
volume fraction, φ, and angular frequency, ω, for the sorption isotherm model for ionomers with inherently
acid and sodium forms at 25 °C. Parallel experiments clustered morphologies, as applied to Nafion mono-
valent cation-sulfonate forms.220 While it requires energies. In this sense, the model was molecular-
that all water molecules and fixed ions reside in specific, but the use of the Poisson-Boltzmann equa-
clusterssa limitationsthe theory provides for the tion to calculate the electrostatic potential in a pore
variance of the average number of fixed ions in a can be questioned because this equation is based on
cluster versus water content. The predicted quanti- the theory of macroscopic electrostatics but the pores
ties are the average number of sorbed water mol- are on the dimensional scale of atomic “granularity”.
ecules per ionized side chain, n, the cluster radius This model also has the feature of computing the
extension ratio, λ, and the volume fraction of the dielectric constant as a function of radial distance in
cluster phase when the ionomer is equilibrated in the cylindrical pore. The result indicated that water
either pure water or water vapor of relative humidity molecules are basically frozen in place near the pore
x. The driving force for cluster expansion is the wall due to the alignment of dipoles by the electric
osmotic pressure posed by visualized “ionic micro- field, while near the cylindrical pore axis the water
solutions” within clusters, Π. Π was derived from a molecules exist in a state similar to bulk water. The
molecular theory of membrane-internal water activ- reader is encouraged to compare these results with
ity that reflects (1) the hydration numbers of the ion the results of Paul and Paddison, referred to later in
exchange group and counterion, (2) the relative free this review, who also calculated pore permittivity
versus bound water population, and (3) the sul- versus radial distance in cylindrical pores using a
fonate-counterion association-dissociation equilib- more rigorous atomistic model. Despite the differ-
rium, which depends on water content. Factor 3 was ences in the two approaches, the essential results are
accounted for using a two or four state statistical the same.
mechanical model, involving hydration and ion pair Eikerling et al. presented a random network and
energies, that was based on the earlier FTIR and effective medium theory-based model of charge trans-
solid-state NMR studies of Mauritz et al., of these port in “porous polymers”.225 The model adopts the
materials.105,107 Π is resisted by a polymer contractile view of clusters as being inverted micelles with
pressure, P, that is a function of a modulus, E, and connecting channels, much as in the Gierke model.
λ, and equilibrium is considered as a condition where There are three elementary types of bonds that are
Π ) P. A constitutive equation for λ versus n was meant to represent conductances between bulklike
assumed that allows for the evolution of a smaller water filled “blue (b) pores” and dry “red (r) pores”.
number of larger clusters, as earlier described by The terminology is somewhat confusing, as a “pore”
Gierke et al.17 is meant to be a cluster and clusters are connected
One of the limitations of this model is that the by “channels”. The pores have no chemical identity
confinement of water molecules within clusters pre- or structure other than havingsor not havingsexcess
cludes its use within the context of water transport water in the amount w. Both pore types contain
simulation because cluster-connective hydration struc- water that is tightly bound by solvation of SO3-
ture is absent. Furthermore, water activity and groups at the “inner pore surfaces”, and it is the
contractile modulus are macroscopic based concepts excess water that can percolate when w is sufficiently
whose application at the nanoscopic level is dubious. high. There are “bonds” of random probabilities that
P is represented by a function borrowed from mac- connect the pores, and they are represented by
roscopic elastic theory that contains E, and there is elements consisting of three electrical resistors in
no microstructure-specific model for the resistance series (i.e., pore-bond-pore), and the three conduc-
to deformation that can be applied to Nafion so that tivities add in the usual reciprocal fashion. These
one is forced to use experimental tensile moduli by elements are linked to form an equivalent electrical
default. circuit over the entire statistical ensemble. Kirchoff
Verbrugge and Hill presented a “macrohomoge- equations of simple electrical theory for this random
neous” model for the transport of ions and solvent network are solved in an approximate fashion by
through ion exchange membranes using the Nernst- considering a single bond as being embedded in an
Planck equation.221 No molecular mechanisms were effective medium of surrounding bonds. This is a site
factored into this model. These investigators did not percolation problem in which the variable is x, the
feel that nanoscopic cavities that were connected by fraction of blue pores, and, as such, addresses the
smaller charged pores existed, owing to the high connectivity of the pore-channel network. The prob-
radiotracer-based diffusion coefficients measured for abilities of b-b, r-r, and b-r bonds are given by x2,
bisulfate ions in Nafion 117 membranes, but they (1 - x)2, and 2x(1 - x), respectively.
preferred a tortuous hydrophilic network phase of This model, when applied to Nafion as a function
pores having diameters ∼ 6 nm.222 Pintauro and of water content, indicated a so-called quasi-percola-
Verbrugge reported a model that calculated partition tion effect, which was “verified” by electrical imped-
coefficients for a pore containing an electrolyte on the ance measurements. “Quasi-percolation” refers to the
basis of adsorption onto the surfaces of pores that fact that the percolation threshold calculated using
contain the fixed charge species.223 Bontha and the single bond effective medium approximation
Pintauro then applied this model to Nafion 117 (namely, xc ) 0.58, or 58% blue pore content) is quite
membranes by treating them as arrays of parallel larger than that issuing from a more accurate com-
cylindrical pores.224 The Nafion model calculations puter simulation. This number does not compare well
accounted for electrostatic interactions between the with the threshold volume fraction calculated by
pore surfaces and counterions, coions, and solvent Barkely and Meakin using their percolation ap-
molecules, as well as counterion/coion solvation free proach, namely 0.10, which is less than the value for
a completely random system, which implies that the that the hydrophilic-hydrophobic interface is not
clusters are not randomly dispersed.216 sharp.
It should also be mentioned that capacitors were Using a molecular dynamics simulation (MM2)
then added in parallel with the resistors in equivalent with no water molecules present, the side chain was
circuit elements because the frequency-dependent predicted to be in a folded conformation at a temper-
experimental electrical impedance data had a com- ature of 300 K. The C-O-C group was stiff, but the
ponent that was 90° out of phase with the resistor. SO3- group, flexible. The folded conformation may
The many details of this theory are omitted here. simply be due to the artificial condition that the long
Nothing dealing with chemical groups and the forces side chain was isolated and not interacting with other
that drive the morphology of ionomers is factored into side chains, and a curling of the unsolvated side chain
this model, which limits its use in predicting fuel cell having two ether oxygens would seem to be com-
membrane performance. Moreover, it seems impos- mensurate with minimizing interatomic energies.
sible to relate the quasi-percolation threshold to the Discrete H2O molecules should be packed around the
real structure. Nonetheless, the view of conductance side chains, rather than utilizing a dielectric con-
from the perspective of percolation is very appropri- tinuum model of a solution environment to render
the calculations realistic for a swollen state. A more
ate.
realistic simulation would also involve more than one
Paddison and Zawodzinski performed fundamental side chain.
calculations aimed at ascertaining side chain confor-
Paddison, in later work, added more water mol-
mation.226 This fragment of the secondary structure
ecules (up to 6) to the minimum energy side chain
is related to tertiary structure, that is, morphology, structure.227 Information relating to acid dissociation
as it would determine, in large degree, how the side and local proton dynamics was obtained. One result
chains pack in micellar structures 40-50 Å in size. of the calculations is that after the first hydration
Packing, in turn, is related to hydration microstruc- shell around the sulfonic acid group is complete, there
ture and the transport of ions and protons throughout is more effective shielding of the proton from the
these ionomers assuming that realistic hydration anion by the water molecules. Paddison stressed that
energetics are included in the calculations. These this model, in which water molecules are clustered
studies were conducted within the context of proton around a single sulfonate group, must be extended
conducting membranes. The authors mentioned a to include neighboring sulfonate groups so that
particular need to establish a molecular basis for proton transfer between them can be properly mod-
understanding electro-osmotic drag, that is, water eled.
transport as coupled with proton transport.
Information extracted from the calculations for
Self-consistent field, ab initio molecular orbital single acid molecule clusters was used, along with
computations were performed on small molecules, experimental information derived from SAXS mea-
namely F3C-O-CF3, to simulate the ether groups, surements,17,228,229 in a water and proton transport
and CF3SO3H, to simulate the ion exchange group model using a nonequilibrium statistical mechanical
at the end of the side chain. The energetics of a single framework.230-232 Diffusion coefficients were calcu-
H2O molecule in the vicinity of these model com- lated for Nafion 117 in cases where the number of
pounds was determined to establish the water affin- water molecules per sulfonate group was 6, 13, and
ity of the ether and sulfonate groups. This water 22.5. More specifically, this model computes a “cor-
content is not high enough so as to cause swelling, rected” friction coefficient for the transport of H3+O
and the single molecule is intended to be but a probe in a hydrated pore/channel of cylindrical geometry.
of its environment. The diffusion coefficient is derived using the friction
The calculations determined that the ether group coefficient in the Einstein equation. Not accounted
was stiff and hydrophobic, but the SO3- group was for, as mentioned by Paddison et al., is the contribu-
strongly hydrophilic and flexible. However, the con- tion to mobility by proton hopping between water
clusion of ether group hydrophobicity may not be in molecules, that is, the Grotthuss mechanism. The
harmony with the FTIR spectroscopic evidence of dielectric constant of water was assumed to be that
Lowry and Mauritz105 as well as that of Moore et al.117 of the bulk state, although Paul and Paddison
It was earlier mentioned that the band characteristic presented a statistical mechanical model that deter-
of this group has two components, at 965 and 980 mines the permittivity of water in hydrated polymer
cm-1, for each of the two C-O-C groups in the side electrolyte membrane pores of cylindrical shape.233
chain. By comparison, there is only one band in this Other assumptions of geometry and theoretical-
region for the Dow short-side-chain membrane that computational details are beyond the scope of this
only has one ether group. The band for the ether review, and the reader is directed to the cited papers.
group closest to the sulfonate groups (965 cm-1) shifts The results of these computations were in agreement
with water content, indicating either hydrogen bond- with diffusion coefficients obtained from pulsed field
ing interactions with water molecules, through-bond gradient NMR diffusion measurements.
inductive effects caused by proton dissociation in the Later, Paul and Paddison presented a statistical
SO3H groups, or a combination of both. The ether mechanical model that was used to calculate the
group closest to the backbone, however, would seem dielectric permittivity in the water domains, that is,
to be shielded from water molecules. This result is the pores, of Nafion.234 For computational purposes,
also commensurate with the suggestion of Falk et al., a pore was taken as being of cylindrical geometry.
based on their FTIR studies, referred to earlier, in The main prediction is that in a fully hydrated
membrane there is a central region within the pores chemical groups were lumped together into effective
where the water is similar to that in the bulk (i.e., spherical interactive force centers.
has a dielectric constant ∼ 80), but as one proceeds The calculations depicted a minimum energy con-
toward the pore walls, the water is increasingly more formation of an isolated side chain that is only
bound. This is essentially the same result predicted slightly twisted, which was rationalized as being
by Bontha and Pintauro using a different, less favorable with regard to the ability of water mol-
fundamental, modeling approach, as described earlier ecules to access the SO3H groups. This structure was
in this review.224 noted as being similar to the curled-up conformation
Niemark et al. investigated solvation and transport determined by Paddison and Zawodzinksi.226 Khala-
in Na+ form Nafion using molecular simulations.235,236 tur et al. also note that their energy-minimized side
Molecular mechanics energy optimizations of oligo- chain structure is rather different from that sug-
mers of 10 monomer units in a vacuum, as well as gested by Litt in his lamellar model.70
smaller oligomers that were solvated by water, Another result is that water molecules preferen-
methanol, and a water-methanol mixture, were tially reside at the terminus of the side chain (i.e.,
performed using a potential force field based on at the sulfonate group), which is not surprising, but
electron density calculations, simulations of vapor- also that the ether groups in the side chains are
liquid equilibrium of lower perfluoroalkanes and hydrophobic, which may be questioned by some
ethers, and the thermodynamic properties of aqueous investigators, as earlier noted in this review.
sodium sulfate solutions. The vacuum calculations The addition of a small amount of water intensifies
yielded two conformations of the fluorocarbon skel- the drive for SO3H group aggregation. Water, as
eton, depending on the initial conformation in which predicted, is in a well-ordered state on local and
the chain was placed before the simulation: either mesoscopic scales, as evidenced by the NMR data of
stretched for the initial all-trans conformation or McMillan et al.,238 and is thought to exist in cagelike
highly folded and spiral-like for a randomly bent structures similar to clathrates. The water content
chain. When solvated, the oligomers, of smaller in these calculations is above and below that corre-
length, were more folded in water as compared to sponding to 4 H2O molecules per SO3H group, al-
methanol and the side chain was stiff. H2O and though it is unclear as to whether enough water
MeOH molecules formed hydrogen bonds with the molecules are present to allow for liquidlike water.
oxygen atoms in the fixed SO3- anions, and these Nor is it clear as to how many polymer chains are
bonds have a lifetime considerably longer than the used in the calculations. In any case, the results
time scales required for rotational motions in the reinforce the view of irregularly shaped cluster
bulk. Water molecules could weakly hydrogen bond surfaces, or “fuzzy spheres” having a high surface
to the ether oxygen closest to the ion exchange group. area/volume aspect. This view is basically that of
For hydrated membranes having K+ counterions, Falk113 and later Kreuer.91 These irregular or rough
water did not form a continuous subphase, but surfaces are amenable to the formation of water
isolated domains having less than 100 water mol- channels along which charge transport can take
ecules were predicted. It might be argued that a place.
larger, more realistic ensemble of molecules and None of the models address the question of how
polymer molecular fragments would be required in the main chains are packed, and details of crystal-
a simulation to render this result credible, which is linity are neither factored into nor predicted by
mentioned by Niemark. The statement is also made mathematical models of the structure and properties
that it is not necessary to have a continuous hydro- of Nafion. Chains packed in crystalline arrays are
philic phase in order to calculate water diffusion usually considered to be rigid within the context of
coefficients comparable to experimental values. certain properties; for example, with regard to dif-
Khalatur et al. performed more aggressive molec- fusion, crystallites are viewed as impenetrable ob-
ular simulations of the structure of hydrated Nafion stacles. 19F NMR studies indicate otherwise. Molec-
using integral equation theory.237 The details of the ular motions that do not significantly alter symmetry
underlying theory, approximations, nature of force can in fact occur in polymer crystals. It would seem,
fields, computational algorithms, and so forth, are for example, that the response of the Nafion structure
beyond the scope of this review and so will be omitted to applied stress would depend on the flexibility of
here. Semiempirical molecular orbital self-consistent the polymer backbone, a certain fraction of which is
field-rotational isomeric state calculations were incorporated in crystalline regions. On the other
performed in order to determine the equilibrium, hand, Starkweather showed that the crystallinity and
energy-minimized conformations of fragments of the swelling of Nafion are not correlated.
chain as well as “complexes” having a single water At first thought, it might be considered that the
molecule probe. The polymer PRISM integral equa- steric restrictions posed by the large fluorine atoms
tion theory was used to calculate density-density would cause conformational rigidity due to restricted
intermolecular pair-correlation functions that reflect bond rotations. Hsu, however, showed that the
averaged particle density as a function of distance conformations of the TFE chains in the crystalline
from any particle and determine the relation between regions in Nafion are in fact dynamic in that they
the chemical structures of macromolecules and pair can undergo helix reversals; that is, the handedness
potentials and intermolecular correlation. While be- of the helix is easily reversed.239 These helix reversals
ing essentially “atomistic”, the modeled structures are also seen in PTFE.240 This disorder phenomenon
had a measure of coarse-graining in that small causes considerable conformational entropy and is
involved in the overall thermodynamics of the sys- segregated (diblock) structure. However, the diffusion
tem. The principal IR spectroscopic signature of a of hydronium was determined to be insensitive to
helix reversal in PTFE is a very temperature-sensi- monomer sequence. It was pointed out that protons
tive peak doublet in the range 600-700 cm-1.241 in these simulations are only allowed to move by the
Temperature-dependent IR studies, conducted in the “vehicular” mechanism, that is, by translation as a
range -30 to 110 °C, showed an activation energy of part of an overall H3O+ molecule, whereas, in reality,
only 1 kcal‚mol-1. The overall conclusion is that protons can hop between adjacent water molecules
fluoropolymer chains are rather flexible in the sense via a combination of activated rate and tunneling
of torsion. processes.
Jang et al. conducted full atomistic molecular Tanimura and Matsuoka calculated energy barri-
dynamics simulations on two fictitious Nafion-like ers for proton transfers in the small model compound
extreme chemical microstructures. One structure was associations ([CF3SO3/H+/SO3CF3]-1 and [CF3SO3/H+/
an 80 unit chain consisting of a single block of 10 H2O/SO3CF3]-1) using an ab initio density functional
consecutive perfluorosulfonic vinyl ether (PSVE) quantum calculation method.243 Curves for energy
units that were placed at the end of a single block of versus distance between H+ and an oxygen atom on
70 TFE units. The other structure, referred to as an adjacent SO3- group in a sulfonate-sulfonate
“dispersed”, was also a chain of 80 monomer units pair, while the distance between the two S atoms was
in which single PSVE units were placed between held constant, were generated. The value of the
blocks of 7 TFE units.242 The total systems in either energy barrier in the presence of one water molecule,
case consisted of four chains, and there were 15 water of 2.1 kcal‚mol-1, is close to the experimental value
molecules per sulfonic acid group so that the total of 2.3 kcal‚mol-1 for an 1100 EW Nafion membrane
number of atoms in the simulations was 4568. The in which there are 18.2 mol of water per mole of SO3-
fundamental inter- and intramolecular interactions sites. Of course, these calculations involve very local
consisted of van der Waals, electrostatic, bond stretch- proton hopping events in the absence of perturbations
ing, bond angle bending, and torsional bond rotation posed by other water molecules and the phase
energetics. separated polymer structure as well as molecular
These structures are fictional in the sense that dynamics.
these sequences do not correspond to the actual At the time of this writing, it must be conceded that
statistical polymerization based on the comonomer there have been no fundamental principles-based
reactivity ratio, although it was said that the results mathematical model for Nafion that has predicted
have significance with respect to Nafion structural significantly new phenomena or caused property
optimization and guidance in the search for Nafion improvements in a significant way. Models that
replacements. Also, the non-insignificant degree of capture the essence of percolation behavior ignore
crystallinity of Nafion was not accounted for in the chemical identity. The more ab initio methods that
model. do embrace chemical structure are limited by the
The reader is referred to this paper for the details number of molecular fragments that the computer
of the structure “annealing procedure”, equation of can accommodate. Other models are semiempirical
motion integration, and so forth. in nature, which limits their predictive flexibility.
Nonetheless, the diversity of these interesting ap-
While most of the conclusions are intuitive (e.g.,
proaches offers structural perspectives that can serve
predicted phase segregation), the results regarding
as guides toward further experimental inquiry.
interfaces are noteworthy. These regions were pre-
dicted to be heterogeneous in the sense of hydro-
phobic and hydrophilic patches. A hydrophilic patch 7. Conclusions
consists of overlapping hydration shells about sul- A comprehensive review of the state of understand-
fonate groups while a hydrophobic patch consists of ing of the fundamental structure and properties of
water-contacting perfluorocarbon groups. The calcu- Nafion perfluorosulfonate materials, at the time of
lations indicated that the interfaces are heteroge- this writing, has been presented based on recent as
neous, much as suggested by Yeager some time ago, well as historical literature that either survived the
and that the diblock system has larger patches, that test of time or influenced further research trends.
is, a greater degree of segregation, both ideas of Since the greatest interest in Nafion in recent years
which are intuitive. derives from its consideration as a proton conducting
While the computational-limited number of atoms membrane in fuel cells, research has been largely
is appreciated, the use of only four relatively short driven by this critical application. Through the col-
chains raises questions regarding the predicted mor- lective information acquired from the detailed works
phologies. In principle, the sample space in these of many research groups and institutions, a broader
calculations is not large enough to generate an understanding of this technologically important ma-
ensemble of periodic or quasi-periodic hydrophobic- terial has emerged. Although many of these contri-
hydrophilic phase components that would be sampled butions have exposed a significant number of critical
by high resolution TEM, AFM, or scattering tech- relationships between structure (both chemical and
niques that reveal long-range structure. morphological) and the unique properties of Nafion,
Finally, simulations of the mean-squared displace- it is widely accepted that much more remains to be
ment versus time for water and hydronium molecules learned.
were performed. The diffusion of water at this degree While most long-range structure interrogations via
of hydration was seen to be greater for the more SAXS and SANS methods have had to assume
structure factor models, and microscopic methods ical property relationships will become an important
have been limited by resolution limits, a theme that future consideration.
has persisted is that the morphology of Nafion is As noted in the Molecular Simulation of Structure
quite complex and at least of a three-phase nature. and Properties section, there have been no funda-
There is no long-range patterned organization of mental principle-based mathematical models for
hydrophilic clusters in a hydrophobic and semi- Nafion that have predicted new phenomena or caused
crystalline perfluorocarbon phase. Hydrophilic- property improvements in a significant way. This is
hydrophobic interfaces are not discrete phase dis- due to a number of limitations inherent in one or the
continuities, but the evidence suggests them to be other of the various schemes. These shortcomings
“rough” or “fuzzy”. The evolution of morphological include an inability to sufficiently account for chemi-
models for this complex polymer has involved many cal identity, an inability to simulate and predict the
distinct conceptual designs; however, the relevant long-range structure as would be probed by SAXS or
aspects of structure in these models have shown TEM, and the failure to simulate structure over
several important unifying perspectives, depending different hierarchy levels. Certainly, advances in this
on ion and water content and the specific range of important research front will emerge and be com-
dimensions probed by the varying methods of analy- bined with advances in experimentally derived in-
sis. With the inevitable improvements in techniques formation to yield a much deeper state of under-
and methods of morphological characterization, a standing of Nafion.
much more detailed representation of the true struc-
ture of perfluorosulfonate ionomers will be realized. 8. Acknowledgment
Water in sufficiently hydrated samples is an ex-
The authors wish to acknowledge support for this
tended phase, and the clusters can be anisotropic and
work provided by the MRSEC Program of the Na-
are contiguous to form percolation pathways, al-
tional Science Foundation under Award Number
though the exact nature of interdomain water struc-
DMR-0213883. Original SAXS data presented in this
ture remains under debate. The more local, that is,
article were obtained through research carried out
molecular, organization of water, cation-sulfonate
at the National Synchrotron Light Source, Brookhaven
interactions and associations, and cation hydration
National Laboratory, which is supported by the U.S.
and mobility have been investigated using spectro-
Department of Energy, Division of Materials Sciences
scopic, thermal analyses, sorption isotherm, and and Division of Chemical Sciences, under Contract
electrical conductivity studies. Water has been clas- No. DE-AC02-98CH10886. Finally, the authors want
sified in terms of freezing versus nonfreezing or to recognize the invaluable assistance of Laura
tightly bound around ions versus free or liquidlike, Fosselman in the preparation of this document.
and the degree of hydrogen bonding in water clusters
has been well-characterized. This information has
been, and will be, valuable in the interpretation of 9. Literature References
water, ion, and proton transport data issuing from (1) Curtin, D. E.; Lousenberg, R. D.; Henry, T. J.; Tangeman, P.
diffusion and electrical conductivity studies. C.; Tisack, M. E. J. Power Sources 2004, 131, 41.
(2) Hora, C. J.; Maloney, D. E. Electrochem. Soc. Ext. Abstr. 1977,
The long- and short-ranged motions in the back- 77 (2), 1145.
bone and side chains in Nafion, that have appeared (3) Grot, W. G. U.S. Patents 3,969,285 (1976), 4,026,783 (1977),
4,030,988 (1977).
as variously assigned transitions in dynamic me- (4) Burkhardt, S. K.; Maloney, D. E. U.S. Patent 4,168,216 (1979).
chanical studies, have only recently been probed on (5) Covitch, M. J.; Lowry, S. R.; Gray, C. L.; Blackford, B. In
Polymeric Separation Media; Cooper, A. R., Ed.; Plenum: New
a molecular level using methods of high resolution York, 1982; p 257.
NMR spectroscopy. This valuable information con- (6) Yeager, H. L.; Kipling, B.; Dotson, R. L. J. Electrochem. Soc.
cerning chain dynamics is only just now emerging to 1980, 127 (2), 303.
(7) Robertson, M. A. F. Ph.D. Thesis, University of Calgary, 1994.
yield important insight into the origins of mechanical (8) Robertson, M. A. F.; Yeager, H. L. Macromolecules 1996, 29,
relaxations of perfluorosulfonate ionomers in a va- 5166.
riety of physical states. (9) Seko, M.; Ogawa, S.; Kimoto, K. In Perfluorinated Ionomer
Membranes; Eisenberg, A., Yeager, H. L., Eds.; ACS Symposium
The static mechanical properties of Nafion, while Series No. 180; American Chemical Society: Washington, DC,
not the primary concern, save for the requirement 1982; Chapter 15, p 365.
(10) (a) Doyle, M.; Lewittes, M. E.; Roelofs, M. G.; Perusich, S. A. J.
of membrane integrity in electrochemical cells, have Phys. Chem. B 2001, 105, 9387. (b) Doyle, M. E.; Lewittes, M.
been investigated as a function of equivalent weight, E.; Roelofs, M. G.; Perusich, S. A.; Lowrey, R. E. J. Membr. Sci.
2001, 184, 257. (c) Perusich, S. A. Macromolecules 2000, 33,
counterion type, and solvent uptake. Undoubtedly, 3431. (d) Perusich, S. A.; McBrearty, M. Polym. Eng. Sci. 2000,
these properties are also a function of morphology, 40, 214. (e) McBrearty, M.; Perusich, S. A. Polym. Eng. Sci. 2000,
yet little specific information on this topic exists to 40, 201. (f) Perusich, S. A.; Avakian, P.; Keating, M. Y.
Macromolecules 1993, 26, 4756.
date. While the number of primary chemical struc- (11) (a) Hashimoto, T.; Fujimura, M.; Kawai, H. p 217. (b) Seko, M.;
ture variations of Nafion sulfonate materials is very Ogawa, S.; Kimoto, K. p 365. (c) Sata, T.; Onoue, Y. p 411. (d)
limited, morphology and, therefore, properties can, Ukihashi, H.; Yamabe, M. p 427. In Perfluorinated Ionomer
Membranes; Eisenberg, A., Yeager, H. L., Eds.; ACS Symposium
however, be manipulated by sample prehistory vari- Series No. 180; American Chemical Society: Washington, DC,
ables such as temperature and temperature cycles, 1982.
(12) Doyle, M.; Rajendran, G. Handbook of Fuel Cells Fundamentals,
swelling in different solvents, orientation, and solu- Technology and Applications; John Wiley & Sons: Chichester,
tion and melt preprocessing schemes. With new U.K., 2003; Vol. 3, Part 3, Chapter 30, p 351.
methods of membrane preparation and the drive (13) Tant, M. R.; Darst, K. P.; Lee, K. D.; Martin, C. W. In Multiphase
Polymers: Blends and Ionomers; Utracki, L. A., Weiss, R. A.,
toward thinner membranes for improved transport Eds.; ACS Symposium Series No. 395; American Chemical
properties, the need for critical morphology-mechan- Society: Washington, DC, 1989; Chapter 15, p 370.
(14) Xu, G. Polymer 1993, 25, 397. (60) Cable, K. M.; Mauritz, K. A.; Moore, R. B. Chem. Mater. 1995,
(15) Gierke, T. D.; Hsu, W. Y. In Perfluorinated Ionomer Membranes; 7, 1601.
Eisenberg, A., Yeager, H. L., Eds.; ACS Symposium Series No. (61) Landis, F. A.; Moore, R. B.; Page, K. A.; Han, C. C. Polym.
180; American Chemical Society: Washington, DC, 1982; Chap- Mater.: Sci. Eng. 2002, 87, 121.
ter 13, p 283. (62) Elliot, J. A.; Hanna, S.; Elliott, A. M. S.; Cooley, G. E. Polymer
(16) Hsu, W. Y.; Gierke, T. D. J. Membr. Sci. 1983, 13, 307. 2001, 42, 2551.
(17) Gierke, T. D.; Munn, G. E.; Wilson, F. C. J. Polym. Sci., Polym. (63) Elliott, J. A.; Hanna, S.; Elliott, A. M. S.; Cooley, G. E.
Phys. 1981, 19, 1687. Macromolecules 2000, 33, 4161.
(18) Yeager, H. L.; Steck, A. J. Electrochem. Soc. 1981, 128, 1880. (64) Londono, J. D.; Davidson, R. V.; Mazur, S. Polym. Mater. Sci.
(19) Eisenberg, A.; Yeager, H. L. Perfluorinated Ionomer Membranes; Eng. 2001, 85, 23.
ACS Symposium Series 180; American Chemical Society: Wash- (65) Barbi, V.; Funari, S. S.; Gehrke, R.; Scharnagl, N.; Stribeck, N.
ington, DC, 1982. Polymer 2003, 44, 4853.
(20) Heitner-Wirguin, C. J. Membr. Sci. 1996, 120, 1. (66) Van der Heijden, P. C.; Rubatat, L.; Diat, O. Macromolecules,
(21) Holliday, L. Ionic Polymer; Applied Science Publishers: London, in-press.
1975. (67) Gebel, G.; Lambard, J. Macromolecules 1997, 30, 7914.
(22) Eisenberg, A.; King, M. Ion Containing Polymers in Polymer (68) Gebel, G.; Moore, R. B. Macromolecules 2000, 33, 4850.
Physics; Academic: New York, 1977; Vol. 2. (69) Rollet, A.-L.; Gebel, G.; Simonin, J.-P.; Turq, P. J. Polym. Sci.,
(23) MacKnight, W. J.; Taggert, W. P.; Stein, R. S. J. Polym. Sci., Part B: Polym. Phys. 2001, 39, 548.
Polym. Symp. 1974, 45, 113. (70) Litt, M. H. Polym. Prepr. 1997, 38, 80.
(24) Yarusso, D. J.; Cooper, S. L. Macromolecules 1983, 16, 1871. (71) Haubold, H.-G.; Vad, T.; Jungbluth, H.; Hiller, P. Electrochim.
(25) Yarusso, D. J.; Cooper, S. L. Polymer 1985, 26, 371. Acta 2001, 46, 1559.
(26) Perfluorinated Ionomer Membranes; Eisenberg, A., Yeager, H. (72) Rubatat, L.; Rollet, A.-L.; Gebel, G.; Diat, O. Macromolecules
L., Eds.; ACS Symposium Series No. 180; American Chemical 2002, 35, 4050.
Society: Washington, DC, 1982. (73) Szajdzinska-Pietek, E.; Schlick, S.; Plonka, A. Langmuir 1994,
(27) Gierke, T. D. 152nd Meeting of the Electrochemical Society, Ext. 10, 1101.
Abstr.; 1977, 124, 319C. (74) Orfino, F. P.; Holdcroft, S. J. New Mater. Electrochem. Syst. 2000,
(28) Pineri, M.; Duplessix, R.; Gauthier, S.; Eisenberg, A. In Ions In 3, 287.
Polymers; Eisenberg, A., Ed.; ACS Symposium Series No. 187; (75) Young, S. K.; Trevino, S. F.; Beck Tan, N. C. J. Polym. Sci., Part
American Chemical Society: Washington, DC, 1980; p 283. B: Polym. Phys. 2002, 40, 387.
(29) Duplessix, R.; Escoubes, M.; Rodmacq, B.; Volino, F.; Roche, E.; (76) Gebel, G. Polymer 2000, 41, 5829.
Eisenberg, A.; Pineri, M. In Water in Polymers; Rowland, S. P., (77) Rollet, A.-L.; Diat, O.; Gebel, G. J. Phys. Chem. B 2002, 21, 3033.
Ed.; ACS Symposium Series No. 127; American Chemical (78) Ceynowa, J. Polymer 1978, 19, 73.
Society: Washington, DC, 1980; p 469. (79) Xue, T.; Trent, J. S.; Osseo-Asare, K. J. Membr. Sci. 1989, 45,
(30) Rodmacq, B.; Coey, J. M.; Escoubes, M.; Roche, E.; Duplessix, 261.
R.; Eisenberg, A.; Pineri, M. In Water in Polymers; Rowland, S. (80) Rieberer, S.; Norian, K. H. Ultramicroscopy 1992, 41, 225.
P., Ed.; ACS Symposium Series No. 127; American Chemical (81) Porat, Z.; Fryer, J. R.; Huxham, M.; Rubinstein, I. J. Phys. Chem.
Society: Washington, DC, 1980; p 487. 1995, 99, 4667.
(31) Hsu, W. Y.; Gierke, T. D. Macromolecules 1982, 15, 101. (82) Lehmani, A.; Durand-Vidal, S.; Turq, P. J. Appl. Polym. Sci.
(32) Gierke, T. D.; Munn, G. E.; Wilson, F. C. In Perfluorinated 1998, 68, 503.
Ionomer Membranes; Eisenberg, A., Yeager, H. L., Eds.; ACS (83) McLean, R. S.; Doyle, M.; Sauer, B. B. Macromolecules 2000,
Symposium Series No. 180; American Chemical Society: Wash- 33, 6541.
ington, DC, 1982; p 195. (84) Zawodzinski, T. A.; Gottesfeld, S.; Shoichet, S.; McCarthy, T. J.
(33) Eisenberg, A.; Hird, B.; Moore, R. B. Macromolecules 1990, 23, J. Appl. Electrochem. 1993, 23, 86.
4098. (85) Schroeder, P. Z. Phys. Chem. (Munich) 1993, 45, 57.
(34) Roche, E. J.; Pineri, M.; Duplessix, R.; Levelut, A. M. J. Polym. (86) Zawodzinski, T. A.; Derouin, C.; Radzinski, S.; Sherman, R. J.;
Sci., Polym. Phys. Ed. 1981, 19, 1. Smith, V. T.; Springer, T. E.; Gottesfeld, S. J. Electrochem. Soc.
(35) Roche, E. J.; Pineri, M.; Duplessix, R. J. Polym. Sci., Polym. Phys. 1993, 140, 1041.
Ed. 1982, 20, 107. (87) Tovbin, Y. K. Zh. Fiz. Khim. 1988, 72, 55.
(36) Fujimura, M.; Hashimoto, T.; Kawai, H. Macromolecules 1981, (88) Ozerin, A. N.; Rebrov, A. V.; Yakunun, A. N.; Bessonova, N. P.;
14, 1309. Dreiman, N. A.; Sokolov, N. F.; Bakeev, L. F. Vysokomol. Soedin.,
(37) Fujimura, M.; Hashimoto, R.; Kawai, H. Macromolecules 1982, Ser. A 1986, 28, 2303.
15, 136. (89) Mauritz, K. A.; Rogers, C. E. Macromolecules 1985, 18, 483.
(38) Roche, E. J.; Stein, R. S.; Russell, T. P.; MacKnight, W. J. J. (90) James, P. J.; McMaster, T. J.; Newton, J. M.; Miles, M. J.
Polym. Sci., Polym. Phys. Ed. 1980, 18, 1497. Polymer 2000, 41, 4223.
(39) Marx, C. L.; Caulfield, D. F.; Cooper, S. J. Macromolecules 2000, (91) Kreuer, K. D. Handbook of Fuel CellssFundamentals, Technol-
33, 6541. ogy and Applications; Vielstich, W., Lamm, A., Gasteiger, H. A.,
(40) Miura, Y.; Yoshida, H. Tokyo Metrop. Univ. 1990, 40, 4349. Eds.; John Wiley & Sons Ltd: Chichester, U.K., 2003; Vol. 3,
(41) Starkweather, H. W., Jr. Macromolecules 1982, 15, 320. Part 3.
(42) Kumar, S.; Pineri, M. J. Mol. Sci., Part B: Polym, Phys, 1986, (92) Grot, W. G. F.; Munn, G. E.; Walmsley, P. N. Presented at the
24, 1767. 141st National Meeting of the Electrochemical Society, Houston,
(43) Manley, D. S.; Williamson, D. L.; Noble, R. D.; Koval, C. A. Chem. TX, 1972.
Mater. 1996, 8, 2595. (93) Yeo, R. S. Polymer 1980, 21, 432.
(44) Dreyfus, B.; Gebel, G.; Aldebert, P.; Pineri, M.; Escoubes, M.; (94) Yeo, R. S. J. Appl. Polym. Sci. 1986, 32, 5733.
Thomas, M. J. Phys. (Paris) 1990, 51, 1341. (95) Gebel, G.; Aldebert, P.; Pineri, M. Polymer 1993, 34, 333.
(45) Lee, E. M.; Thomas, R. K.; Burgess, A. N.; Barnes, D. J.; Soper, (96) Zawodzinski, T. A.; Neeeman, M.; Sillerud, L. O.; Gottesfeld, S.
A. K.; Rennie, A. R. Macromolecules 1992, 25, 3106. J. Phys. Chem. 1991, 95, 6040.
(46) Grot, W. G.; Chadds, F. European Patent 0,066,369, 1982. (97) Zawodzinski, T. A.; Derouin, C.; Radzinsky, S.; Sherman, R. J.;
(47) Martin, C. R.; Rhoades, T. A.; Fergusson, J. A. Anal. Chem. 1982, Smith, V. T.; Springer, T. E.; Gottesfeld, S. J. Electrochem. Soc.
54, 1639. 1993, 140, 1041.
(48) Aldebert, P.; Dreyfus, B.; Pineri, M. Macromolecules 1986, 19, (98) Hinatsu, J. T.; Mizuhata, M.; Takenaka, H. J. Electrochem. Soc.
2651. 1994, 141 (6), 1493.
(49) Aldebert, P.; Dreyfus, B.; Gebel, G.; Nakamura, N.; Pineri, M.; (99) Yeo, S.; Eisenberg, A. J. Appl. Polym. Sci. 1977, 21, 875.
Volino, F. J. Phys. (Paris) 1988, 49, 2101. (100) Zawodzinski, T. A.; Davey, J.; Springer, T.; Gottesfeld, S.
(50) Rebrov, A. V.; Ozerin, A. N.; Svergun, D. I.; Bobrova, L. P.; Electrochem. Soc. Ext. Abstr. 1992, Vol. 92-2, Abstract No. 94,
Bakeyev, N. F. Polym. Sci. U. S. S. R. 1990, 32, 1515. p 149.
(51) Loppinet, B.; Gebel, G.; Williams, C. E. J. Phys. Chem. B 1997, (101) Futerko, P.; Hsing, I-M. J. Electrochem. Soc. 1999, 146 (6), 2049.
101, 1884. (102) Flory, P. J. Principles of Polymer Chemistry; Cornell University
(52) Loppinet, B.; Gebel, G. Langmuir 1998, 14, 1977. Press: Ithaca, NY, 1953; pp 495-540.
(53) Moore, R. B.; Martin, C. R. Anal. Chem. 1986, 58, 2569. (103) Buzzoni, R.; Bordiga, G.; Ricchiardi, G.; Spoto, G.; Zecchina, A.
(54) Moore, R. B.; Martin, C. R. Macromolecules 1988, 21, 1334. J. Phys. Chem. 1995, 99, 11937.
(55) Moore, R. B.; Martin, C. R. Macromolecules 1989, 22, 3594. (104) Zecchina, A.; Geobaldo, G.; Spoto, G.; Bordiga, G.; Ricchiardi,
(56) Gebel, G.; Aldebert, P.; Pineri, M. Macromolecules 1987, 20, G.; Buzzoni, R.; Patrini, G. J. Phys. Chem. 1996, 100, 16584.
1425. (105) Lowry, S. R.; Mauritz, K. A. J. Am. Chem. Soc. 1980, 102, 4665.
(57) Halim, J.; Scherer, G. G.; Stamm, M. Macromol. Chem. Phys. (106) Komoroski, R. A.; Mauritz, K. A. J. Am. Chem. Soc. 1978, 100,
1994, 195, 3783. 7487.
(58) Cable, K. M.; Mauritz, K. A.; Moore, R. B. Polym. Prepr. 1994, (107) Komoroski, R. A.; Mauritz, K. A. In Perfluorinated Ionomer
35 (1), 421. Membranes; Eisenberg, A., Yeager, H. L., Eds.; ACS Symposium
(59) Cable, K. M.; Mauritz, K. A.; Moore, R.B. Polym. Prepr. 1994, Series 180; American Chemical Society: Washington, DC, 1982;
35 (2), 854. p 113.
(108) Robinson, R. A.; Stokes, R. H. Electrolyte Solutions, 2nd ed.; (155) Cappadonia, M.; Erning, J. W.; Niake, S. M.; Stimming, U. Solid
Dover: Mineola, NY, 2002; Chapter 14. State Ionics 1995, 77, 65.
(109) Diebler, H.; Eigen, M. Z. Phys. Chem. (Muenchen) 1959, 20, 299. (156) Chen, R. S.; Jayakody, S. G.; Greenbaum, S. G.; Pak, Y. S.; Xu,
(110) Eigen, M.; Tamm, K. Z. Elektrochem. 1962, 66, 93, 107. G.; McLin, M. G.; Fontanella, J. J. Electrochem Soc. 1993, 140,
(111) Komoroski, R. A. Adv. Chem. Ser. 1980, 187, 155. 889.
(112) Heitner-Wirguin, C. Polymer 1979, 20, 371. (157) Chen, R. S.; Stallworth, P. E.; Greenbaum, S. G.; Fontanella, J.
(113) Falk, M. Can. J. Chem. 1980, 58, 1495. J.; Wintersgill, M. C. Electrochim. Acta 1995, 40 (3), 309.
(114) Rollet, A-L.; Gebel, G.; Simonin, J-P.; Turq, P. J. Polym. Sci., (158) Paddison, S. J.; Reagor, D. W.; Zawodzinski, T. A. J. Electroanal.
Part B: Polym. Phys. 2001, 39, 548. Chem. 1998, 459, 91.
(115) Kujawski, W.; Nguyen, Q. T.; Neel, J. J. Appl. Polym. Sci. 1992, (159) Ellison, W. J.; Lamkaouchi, K.; Moreau, J.-M. J. Mol. Liq. 1996,
44, 951. 68, 171.
(116) Quezado, S.; Kwak, J. C. T.; Falk, M. Can. J. Chem. 1984, 62, (160) Zawodzinski, T. A.; Neeman, M.; Sillerud, L. D.; Gottesfeld, S.
958. J. Phys. Chem. 1991, 95, 6040.
(117) Cable, K. M.; Mauritz, K. A.; Moore, R. B. J. Polym. Sci., Part (161) Alberti, G.; Casciola, M.; Massinelli, L.; Bauer, B. J. Membr. Sci.
B: Polym. Phys. 1995, 33, 1065. 2001, 185, 73.
(118) Falk, M. In Perfluorinated Ionomer Membranes; Eisenberg, A., (162) Glasstone, S.; Laidler, K. J.; Eyring, H. The Theory of Rate
Yeager, H. L., Eds.; ACS Symposium Series 180; American Processes; McGraw-Hill: New York, 1941; pp 550 and 573.
Chemical Society: Washington, DC, 1982; p 139. (163) Mauritz, K. A. Morphological Theories. In Ionomers: Synthesis,
(119) Peluso, S. L.; Tsatsas, A. T.; Risen, W. M. Spectral Studies of Structure, Properties and Applications; Tant, M. R., Mauritz, K.
Ions in Perfluorocarbonsulfonate (Nafion) Ionomers; Office of A., Wilkes, G. L., Eds.; Blackie Academic & Professional:
Naval Researach Technical Report No. TR-79-01, 1971. London, 1997.
(120) Yeager, H. L. In Perfluorinated Ionomer Membranes; Eisenberg, (164) Yeo, Y. S. In Perfluorinated Ionomer Membranes; Eisenberg, A.,
A., Yeager, H. L., Eds.; ACS Symposium Series 180; American Yeager, H. L., Eds.; ACS Symposium Series No. 180; American
Chemical Society: Washington, DC, 1982; p 25. Chemical Society: Washington, DC, 1982; p 453.
(121) Barnes, D. J. In Structure and Properties of Ionomers; Pineri, (165) Kipling, B. In Perfluorinated Ionomer Membranes; Eisenberg,
M., Eisenberg, A., Eds.; D. Reidel Publ. Co.: Dordrecht, 1987; p A., Yeager, H. L., Eds.; ACS Symposium Series No. 180;
501. American Chemical Society: Washington, DC, 1982; p 475.
(122) Luck, W. A. P. In Synthetic Membrane Processes; Belfort, G., (166) Yeager, H. L.; Gronowski, A. A. In Ionomers: Synthesis, Struc-
Ed.; Academic Press: New York, 1984; p 21. ture, Properties and Applications; Tant, M. R., Mauritz, K. A.,
(123) Quezado, S.; Kwak, J. C. T.; Falk, M. Can. J. Chem. 1984, 62, Wilkes, G. L., Eds.; Blackie Academic & Professional: London,
958. 1997; p 333.
(124) Yeager, H. L.; Twardowski, Z.; Clarke, L. M. J. Electrochem. Soc. (167) Olah, G. A.; Dyer, P. S.; Prakash, G. K. S. Synthesis 1980, 7,
1982, 129 (2), 324. 513.
(125) Starkweather, H. W.; Chang, J. J. Macromolecules 1982, 15, 752. (168) Sondheimer, S. J.; Bunce, N. J.; Fyfe, C. A. J. Macromol. Sci.,
(126) Boyle, N. G.; McBrierty, V. J.; Douglass, D. C. Macromolecules Rev. Macromol. Chem. Phys. 1986, C26 (3), 353.
1983, 16, 75. (169) Ferry, L. L. J. Macromol. Sci., Chem. 1990, A27 (8), 1095.
(127) Bunce, N. J.; Sondheimer, S. J.; Fyfe, C. A. Macromolecules 1986, (170) Zundel, G. Hydration and Intermolecular Interaction; Academic
19, 333. Press: New York, 1969.
(128) Sivashinsky, N.; Tanny, G. B. J. Appl. Polym. Sci. 1981, 26, 2625. (171) Zundel, G. In The Hydrogen Bond, Recent Developments in
(129) Bloembergen, N.; Purcell, E. M.; Pound, R. V. Phys. Rev. 1948, Theory and Experiments; Schuster, P., Zundel, G., Sandorfy, C.,
73, 679. Eds.; North Holland: Amsterdam, 1976; Vol. II, Chapter 15, p
(130) Resing, H. A. Adv. Relaxation Proc. 1967-8, 1, 109. 683.
(172) Eigen, M. Angew. Chem. 1963, 75, 489.
(131) Boyle, N.; McBrierty, V. J.; Eisenberg, A. Macromolecules 1983,
(173) Zundel, G. J. Membr. Sci. 1982, 11, 249.
16, 80.
(174) Interactions and Structures of Ionic SolvatessInfrared Results.
(132) Pineri, M.; Volino, F.; Escoubes, M. J. Polym. Sci., Polym. Phys.
Zundel, G.; Fritsch, J. In Physics of Ionic Solvation; Dogonadze,
Ed. 1985, 23, 2009.
R. R., Kalman, E., Kornyshev, A. A., Ulstrup, J., Eds.; Elsevier:
(133) Yoshida, H.; Miura, Y. J. Membr. Sci. 1992, 68, 1.
Amsterdam, 1984.
(134) Miura, Y.; Yoshida, H. Mem. Fac. Technol., Tokyo Metrop. Univ. (175) Leuchs, M.; Zundel, G. J. Chem. Soc., Faraday Trans 2 1978,
1990, 40, 4349. 74, 2256.
(135) Moore, R. B.; Cable, K. M. Polym. Prepr. (Am. Chem. Soc., Div. (176) Ostrowska, J.; Narebska, A. Colloid Polym. Sci. 1983, 261, 93.
Polym. Chem.) 1997, 38 (1), 272. (177) Mauritz, K. A.; Gray, C. L. Macromolecules 1983, 16, 1279.
(136) MacMillan, B.; Sharp, A. R.; Armstrong, R. L. Polymer 1999, (178) Kreuer, K. D. Chem. Mater. 1998, 8, 610.
40, 2471. (179) Kreuer, K. D.; Rabenau, A.; Weppner, W. Angew. Chem., Int.
(137) Martin, C. R.; Rhoades, T. A.; Ferguson, J. A. Anal. Chem. 1982, Ed. Engl. 1982, 21, 208.
54, 1639. (180) Kreuer, K. D. J. Membr. Sci. 2001, 185, 29.
(138) Rieke, P. C.; Vanderborgh, N. E. J. Membr. Sci. 1987, 32, 313. (181) Yeo, S. C.; Eisenberg, A. J. Appl. Polym. Sci. 1977, 21, 875.
(139) Escoubes, M.; Pineri, M. In Perfluorinated Ionomer Membranes; (182) Takamatsu, T.; Hashiyama, M.; Eisenberg, A. J. Appl. Polym.
Eisenberg, A., Yeager, H. L., Eds.; ACS Symposium Series 180; Sci. 1979, 24, 2199.
American Chemical Society: Washington, DC, 1982; p 9. (183) Yeager, H. L. In Structure and Properties of Ionomers; NATO
(140) MacMillan, B.; Sharp, A. R.; Armstrong, R. L. Polymer 1999, ASI Series C; D. Reidel: Dordrecht, 1987; Vol. 198, p 377.
40, 2481. (184) Yeager, H. L. In Perfluorinated Ionomer Membranes; Eisenberg,
(141) Angell, C. A. In Water, a Comprehensive Treatise; Franks, F., A., Yeager, H. L., Eds.; ACS Symposium Series No. 180;
Ed.; Plenum: New York, 1982; Vol. 7. American Chemical Society: Washington, DC, 1982; p 42.
(142) Cahan, B. D.; Wainright, J. S. J. Electrochem. Soc. 1993, 140, (185) Millet, P. J. Membr. Sci. 1990, 50, 325.
L185. (186) Lopez, M.; Kipling, B.; Yeager, H. L. Anal. Chem. 1977, 49, 629.
(143) Fontanella, J. J.; McLin, M. G.; Wintersgill, M. C.; Calame, J. (187) Yeager, H. L.; Steck, A. Anal. Chem. 1979, 51, 862.
P.; Greenbaum, S. G. Solid State Ionics 1993, 66, 1. (188) Yeager, H. L.; Kipling, B. J. Phys. Chem. 1979, 83, 1836.
(144) Gardner, C. L.; Anantaraman, A. V. J. Electroanal. Chem. 1998, (189) Zawodzinski, T. A.; Neeman, M.; Sillerud, L. O.; Gottesfeld, S.
449, 209. J. Phys. Chem. 1991, 95, 6040.
(145) Gavach, C.; Pamboutzoglou, N.; Nedyalkov, N.; Pourcelly, G. J. (190) Gong, X.; Bandis, A.; Tao, A.; Meresi, G.; Inglefield, P. T.; Jones,
Membr. Sci. 1989, 45, 37. A. A.; Wen, W.-Y. Polymer 2001, 42, 6485.
(146) Pourcelly, G.; Oikonomou, A.; Gavach, C. J. Electroanal. Chem. (191) Meresi, G.; Wang, Y.; Bandis, A.; Inglefield, P. T.; Jones, A. A.;
1990, 287, 43. Wen, W.-Y. Polymer 2001, 42, 6153.
(147) Lindheimer, A.; Molenat, J.; Gavach, C. J. Electroanal. Chem. (192) Sodaye, H. S.; Pujari, P. K.; Goswami, A.; Manohar, S. B. J.
1987, 216, 71. Polym. Sci., Part B: Polym. Phys. 1997, 35, 771.
(148) Eikerling, M.; Kornyshev, A. A.; Kuznetzov, A. M.; Ulstrup, J.; (193) Plate, N. A.; Shibaev, V. P. Brush-like Polymers and Liquid
Walbran, S. J. Phys. Chem. B 2001, 105, 3636. Crystals; Khimija: Moscow, 1980; p 304 (Russ. transl.).
(149) Zundel, G.; Metzger, H. Z. Phys. Chem. 1968, 58, 225. (194) Timashev, S. F. Physical Chemistry of Membrane Processes;
(150) Wicke, E.; Eigen, M.; Ackermann, T. Z. Phys. Chem. 1954, 1, Khimija: Moscow, 1988; p 237 (Russ. transl.).
340. (195) Rubatat, L.; Rollet, A. L.; Gebel, G.; Diat, O. Macromolecules
(151) Eikerling, M.; Kornyshev, A. A. J. Electroanal. Chem. 2001, 502, 2002, 35, 4050.
1. (196) Kalyanasundaram, K.; Thomas, J. K. J. Am. Chem. Soc. 1977,
(152) Aldebert, P.; Guglieme, M.; Pineri, M. Polym. J. 1991, 23 (5), 99 (7), 2039.
399. (197) Lee, P. C.; Meisel, D. Photochem. Photobiol. 1985, 41, 21.
(153) Helfferich, F. Ion Exchange; McGraw-Hill: New York, 1962; p (198) Szentirmay, M. N.; Prieto, N. E.; Martin, C. R. J. Phys. Chem.
351. 1985, 89, 3017.
(154) Cappadonia, M.; Erning, J. W.; Stimming, U. J. Electroanal. (199) Robertson, M. A. F.; Yeager, H. L. Macromolecules 1996, 29,
Chem. 1994, 376, 189. 5155.
(200) Young, S. K.; Trevino, S. F.; Beck Tan, N. C.; Paul, R. L. Army (221) Verbrugge, M. W.; Hill, R. F. J. Electrochem. Soc. 1990, 137,
Research Laboratory Report ARL-TR-2679, March 2002. 886.
(201) Twardowski, Z.; Yeager, H. L.; O’Dell, B. J. Electrochem Soc. (222) Verbrugge, M. W.; Hill, R. F. J. Electrochem. Soc. 1990, 137,
1982, 129 (2), 328. 893.
(202) Perusich, S. A. Macromolecules 2000, 33, 3431. (223) Pintauro, P. N.; Verbrugge, M. W. J. Membr. Sci. 1989, 44, 197.
(203) Grot, W. G. F.; Munn, G. E.; Walmsley, P. N. Presented at the (224) Bontha, J. R.; Pintauro, P. N. Chem. Eng. Sci. 1994, 49, 3835.
Electrochemical Society Meeting, Houston, TX, May 7-11, 1972; (225) Eikerling, M.; Kornyshev, A. A.; Stimming, U. J. Phys. Chem.
paper no. 154. B 1997, 101, 10807.
(204) Yeo, S. C.; Eisenberg, A. J. Appl. Polym. Sci. 1977, 21, 875. (226) Paddison, S. J.; Zawodzinski, T. A. Solid State Ionics 1998, 113-
(205) Kyu, T.; Eisenberg, A. In Perfluorinated Ionomer Membranes; 115, 333.
Eisenberg, A., Yeager, H. L., Eds.; American Chemical Society (227) Paddison, S. J. Handb. Fuel CellssFundam., Technol. Appl.
Symposium Series 180; American Chemical Society: Washing- 2003, 3 (3), 396.
ton, DC, 1982; Chapter 6, p 79. (228) Kreuer, K. D. J. Membr. Sci. 2001, 185, 29.
(206) Tant, M. R.; Darst, K. P.; Lee, K. D.; Martin, C. W. In Multiphase (229) Ise, M. Ph.D. Thesis, University of Stuttgart, 2000.
Polymers: Blends and Ionomers; Utracki, L. A., Weiss, R. A., (230) Paddison, S. J.; Paul, R.; Zawodzinski, T. A. In Proton Conduct-
Eds.; American Chemical Society Symposium Series 395; Ameri- ing Membrane Fuel Cells II; Gottesfeld, S., Fuller, T. F., Eds.;
can Chemical Society: Washington, DC, 1989; Chapter 15, p 370. Electrochemical Society Proceedings Series; Electrochemical
(207) Miura, Y.; Yoshida, H. Thermochim. Acta 1990, 163, 161. Society: Pennington, NJ, 1999; Vol. 98-27, p 106.
(208) Cable, K. M. Tailoring Morphology-Property Relationships in (231) Paddison, S. J.; Paul, R.; Zawodzinski, T. A. J. Electrochem. Soc.
Perfluorosulfonate Ionomers. Ph.D. Dissertation, University of 2000, 147, 617.
Southern Mississippi, 1996. (232) Paddison, S. J.; Paul, R.; Zawodzinski, T. A. J. Chem. Phys. 2001,
(209) Young, S. K.; Mauritz, K. A. J. Polym. Sci., Part B: Polym. Phys. 115, 7753.
2001, 39 (12), 1282. (233) Paul, R.; Paddison, S. J. J. Chem. Phys. 2001, 115 (16), 7762.
(210) Kyu, T.; Hashiyama, M.; Eisenberg, A. Can. J. Chem. 1983, 61, (234) Paul, R.; Paddison, S. J. Solid State Ionics 2004, 168, 245.
680. Also cross reference #195 (version 7/6/04). (235) Vishnyakov, A.; Niemark, A. V. J. Phys. Chem. B 2000, 104,
(211) Moore, R. B.; Cable, K. M.; Croley, T. L. J. Membr. Sci. 1992, 4471.
75, 7. (236) Vishnyakov, A.; Niemark, A. V. U. S. Army NSC Report:
(212) Page, K. A.; Moore, R. B. Polym. Prepr. (Am. Chem. Soc., Div. Modeling of Microstructure and Water Vapor Sorption and
Polym. Chem.) 2003, 44 (1), 1144. Transfer in Permselective Membranes, 2000.
(213) Landis, F. A.; Moore, R. B.; Page, K. A.; Han, C. C. Polym. Mater. (237) Khalatur, P. G.; Talitskikh, S. K.; Khokhlov, A. R. Macromol.
Sci. Eng. (Am. Chem. Soc., Div. Polym. Mater. Sci. Eng.) 2002, Theory Simulat. 2002, 11, 566.
87, 121. (238) MacMillan, B.; Sharp, A. R.; Armstrong, R. L. Polymer 1999,
(214) Page, K. A.; Cable, C. A.; Moore, R. B. Manuscript in preparation. 40, 2481.
(215) Mauritz, K. A. In Ionomers: Synthesis, Structure, Properties and (239) Hsu, W. Y. Macromolecules 1983, 16, 745.
Applications; Tant, M. R., Mauritz, K. A., Wilkes, G. L., Eds.; (240) Matsushige, K.; Enoshita, R.; Ide, T.; Yamanchi, N.; Taki, S.;
Blackie Academic & Professional: London, 1997; p 95. Takemura, T. Jpn. J. Appl. Phys. 1977, 16, 681.
(216) Hsu, W. Y.; Barkley, J. R.; Meakin, P. Macromolecules 1980, (241) Rosser, R. W.; Schrag, J. L.; Ferry, J. D.; Greaser, M. Macro-
13, 198. molecules 1977, 10, 978.
(217) Cirkel, P. A.; Okada, T. Macromolecules 2000, 33, 4921. (242) Jang, S. S.; Molinero, V.; Cagin, T.; Goddard, W. A. J. Phys.
(218) Cirkel, P. A.; Okada, T.; Kinigasa, S. Macromolecules 1999, 32, Chem. B 2004, 108, 3149.
5321. (243) Tanimura, S.; Matsuoka, T. J. Polym. Sci., Part B: Polym. Phys.
(219) Hsu, W. Y.; Berzins, T. J. Polym. Sci., Polym. Phys. Ed. 1985, 2004, 42, 1905.
23, 933.
(220) Mauritz, K. A.; Rogers, C. E. Macromolecules 1985, 18, 483. CR0207123
Chem. Rev. 2004, 104, 4791−4843 4791
Factors Governing Oxygen Reduction in Solid Oxide Fuel Cell Cathodes†

Stuart B. Adler
Department of Chemical Engineering, University of Washington, Box 351750, Seattle, Washington 98195-1750
Contents 6.1. Sensitivity of Interfacial Electrochemical 4825

Kinetics to Secondary Phases and Impurities
1. Introduction 4791 6.2. Alteration of Material Properties Near the 4828
2. Understanding the Three-Phase BoundarysA 4793 Interface
Long History of Inquiry 6.3. Ceria as an Alternative Electrolytic Interface 4829
3. Platinum ElectrodessInterplay of Chemical and 4794 6.4. Composite Microstructures 4830
Electrochemical Steps
6.5. Current Constriction Effects 4832
3.1. Two Schools of Thought 4795
6.6. Long-Term Degradation 4833
3.2. Chemical Contributions to the Overpotential 4796
6.7. Experimental Artifacts in Electrochemical 4834
3.3. Evidence for Diffusion and Adsorption 4797 Measurements
Limitations on Platinum
7. Conclusions and Outlook 4837
3.4. Important Refinement: Co-limited Reaction 4799
Kinetics 8. Acknowledgements 4838
3.5. Unresolved Issues Surrounding Oxygen 4801 9. Nomenclature 4839
Reduction on Pt/YSZ 10. Glossary of Commonly Used Acronyms 4839
3.6. Nonstationary Effects 4802 11. Glossary of Some Commonly Used Terms 4839
3.7. Summary: Platinum as a Framework for 4804 12. References 4839
Understanding Other SOFC Cathodes
4. Perovskite Mixed Conductors and the Role of 4804
Bulk Material Properties 1. Introduction
4.1. Perovskite OxidessLow-Cost Alternative to 4804 Recent worldwide interest in building a decentral-
Platinum ized, hydrogen-based energy economy has refocused
4.2. Thermodynamic, Kinetic, and Transport 4805 attention on the solid oxide fuel cell (SOFC) as a
Properties of Perovskite Mixed Conductors potential source of efficient, environmentally friendly,
4.3. Defining the Role of the BulksDense 4807 fuel-versatile electric power. Due to its high operating
Thin-Film Mixed-Conducting Electrodes temperature, the SOFC offers several potential ad-
4.4. Chemical Capacitance 4808 vantages over polymer-based fuel cells, including
4.5. Porous Mixed ConductorssA More Complex 4809 reversible electrode reactions, low internal resistance,
Case high tolerance to typical catalyst poisons, production
4.6. One Asymptotic LimitsThe Bulk Reaction 4810 of high-quality waste heat for (among other uses)
Path reformation of hydrocarbon fuels, as well as the
4.7. Limitations of Bulk, 1D Transport Models for 4813 possibility of burning hydrocarbon fuels directly.
Porous Mixed Conductors Today, SOFCs are much closer to commercial
4.8. Summary: Importance of the Bulk for 4815 reality than they were 20 years ago, due largely to
Mixed-Conducting SOFC Cathodes technological advances in electrode material composi-
5. Lanthanum Strontium Manganese Oxide (LSM): 4816 tion, microstructure control, thin-film ceramic fabri-
Where Surface and Bulk Converge cation, and stack and system design. These advances
5.1. Three-Phase Boundary: Not the Whole 4816 have led to dozens of active SOFC development
Picture programs in both stationary and mobile power and
5.2. Complex Stationary Electrochemical 4817 contributed to commercialization or development in
Characteristics and Properties a number of related technologies, including gas
5.3. Dense and Patterned Thin Films: Confirming 4820 sensors,1 solid-state electrolysis devices,2 and ion-
Two Regimes of Operation transport membranes for gas separation and partial
5.4. Nonstationary Behavior in LSM 4823 oxidation.3 Many reviews are available which sum-
5.5. Summary: Uncertainties in Our 4825 marize the technological advances made in SOFCs
Understanding of Oxygen Reduction on LSM over the last 15-35 yearssreaders who are primarily
6. Factors Complicating our Understanding of 4825 interested in knowing the state-of-the art in materi-
SOFC Cathode Mechanisms als, design, and fabrication (including the electrodes)
are encouraged to consult these reviews.4-12
† Dedicated to Brian Steele, 1929-2003. Researcher, Entrepre- This review focuses on the factors governing SOFC
neur, Consensus Seeker. cathode performancesadvances we have made over
10.1021/cr020724o CCC: $48.50 © 2004 American Chemical Society
4792 Chemical Reviews, 2004, Vol. 104, No. 10 Adler
external circuit. The percentage of reversible work

converted to electrical work depends on (among other
things) the internal losses in the cell, including the
ohmic resistance of the electrolyte, as well as the
overpotential losses at the anode and cathode.
While ohmic losses in oxide electrolytes are largely
understood today, the physics governing the electrode
overpotential losses remain an enormous focus of
research, with substantial progress being made only
in the last 15-20 years. This shift in emphasis from
the electrolyte to the electrodes has been driven (in
part) by an ability to make increasingly thinner, less
resistive, electrolyte films as well as a drive toward
lower operating temperatures where the electrodes
Stuart B Adler received his Ph.D. degree in Chemical Engineering in 1993 are a higher percentage of the voltage loss (due to
from the University of California, Berkeley, where he used high-temperature higher activation energy). Much of this work has
nuclear magnetic resonance (NMR) to probe microstructure, atomic motion, focused on the cathode, largely because oxygen
and electronic structure in electrochemical ceramics. His work in ionic reduction is generally thought to be the more difficult
materials continued at Imperial College (NATO−NSF fellow), where he
developed continuum mechanical theories for high-temperature electrodes.
reaction to activate on SOFCs operating at com-
In 1994 he moved to Ceramatec, Inc., where he led research supporting mercially relevant temperatures. Workers have tried
commercialization of ion transport membranes and other oxygen production to not only understand electrode mechanisms, but
and removal processes. After rejoining academia in 1999 (CWRU), he also explore new electrode materials and microstruc-
moved in 2002 to the University of Washington, Department of Chemical tures, elucidate structure-property-performance re-
Engineering. There his work in electrochemical ceramics has continued, lationships, and understand how and why electrode
including advanced measurement and modeling techniques for solid-state
electrodes and independent studies of kinetic, transport, and thermody- performance changes with time, temperature, ther-
namic properties of solids. Professor Adler’s awards include the NSF− mal cycling, operating conditions, impurities, or other
NATO postdoctoral Fellowship (1993), and NSF Career Award (2001), factors that may be pertinent in the design of
and most recently he is the inaugural recipient of the biennial Charles W. multicell stacks and systems.
Tobias Young Investigator Award of the Electrochemical Society (2004).
This review attempts to summarize the advances
the last 20 years in our scientific understanding of made in our understanding of SOFC cathodes since
oxygen reduction mechanisms, how these mecha- approximately the early 1980s, when there was a
nisms vary for different materials and conditions, and surge in worldwide SOFC research activity. Accord-
remaining questions and challenges that have been ing to the ISI Web-Of-Science Science Citation Index,
generated by this large volume of work. more than 1000 refereed articles pertinent to the
For the purposes of review, Figure 1 illustrates the topic of SOFC cathodes and cathode materials have
basic function of the cathode in a solid oxide fuel cell. been published since 1980. Regrettably, it is not
Whether acting alone or as part of a stack of cells, possible to cover this enormous volume of work in
each cell consist of a free-standing or supported uniform detail. Also, new insights regarding SOFC
membrane of an oxygen-ion-conducting electrolyte, cathodes have not occurred in a vacuumsthey have
often yttria-stabilized zirconia (YSZ). Oxygen, which benefited substantially from advances in electro-
is fed (usually as air) to one side of the membrane, chemical measurement and modeling techniques,
is reduced by the cathode to oxygen ions via the improved understanding of cathode materials prop-
overall half-cell reaction erties, and new fabrication and characterization
techniques which allow control and measurement of
1 electrode microstructure. These enabling develop-
2e- + O2 f O2- (1) ments can only be tangentially reviewed here.
2
Rather this review focuses on how new approaches
Oxygen ions thus created migrate selectively through have been used by workers to better understand
the membrane to the anode, where they undergo a cathode mechanisms and how these mechanisms
similar half-cell reaction with a gaseous fuel (either relate to materials properties and microstructure.
H2, syngas, or a hydrocarbon) to produce H2O and This review also attempts to identify ongoing critical
CO2. The flow of electrons liberated and consumed questions that will likely be the focus of cathode
at the anode and cathode, respectively, deliver some research and development over the next 10-15 years.
portion of the reversible work of the reaction to the In highlighting and discussing in detail the most
influential and important work in the field, this
review also hopes to summarize the general trends
and consensus understanding that has developed in
parallel with these key insights. In telling the story
of these new developments and challenges, it is the
sincere hope of the author to do justice to the many
scientists and engineers who have spent their careers
Figure 1. Schematic showing the roles of anode, cathode, contributing to this fascinating and potentially im-
and electrolyte in a solid oxide fuel cell (SOFC). portant area of research.
Oxygen Reduction in Solid Oxide Fuel Cell Cathodes Chemical Reviews, 2004, Vol. 104, No. 10 4793
Figure 3. Phenomenological roles of the electronically

conducting (electronic) phase (R), gas phase (β), and ioni-
cally conducting (ionic) phase (γ) in accomplishing oxygen
reduction.
ture (Figure 2c), thereby increasing the contact area

between electronically and ionically conductive
phases.29-34 These strategies can also be combined
wherein a composite electrode contains both ionic and
mixed conducting phases.29,35,36
While the various strategies described above have
proven promising, SOFC electrodes remain largely
empirically understood and far from optimized and
suffer from numerous short- and long-term degrada-
tion problems.16,37,38 Reported performances vary
Figure 2. Common strategies for SOFC cathodes: (a) tremendously with many unknown variables at work
porous single-phase electronically conductive oxide such as and limited understanding as to how materials
(La,Sr)MnO3 (LSM); (b) porous single-phase mixed conduc- properties and microstructure relate to performance
tor; (c) porous two-phase composite. The SEM micrograph
of LSM on YSZ in a is adapted from ref 84. (Adapted with and long-term stability.39,40
permission from ref 84. Copyright 1997 Swiss Federal At the risk of oversimplification, Figure 3 il-
Institute of Technology.) lustrates the phenomenological role of the cathode
material, electrolyte material, and gas in accomplish-
2. Understanding the Three-Phase BoundarysA ing the reaction in eq 1. Regardless of the micro-
Long History of Inquiry structure, a common feature of all cathodes is that
the cathode material itself (R) (an electronic conduc-
As illustrated in Figure 2, a SOFC cathode typi- tor) makes intimate contact with an electrolyte phase
cally consists of a porous single phase or two-phase (γ) (an oxygen-ion conductor) along an interface, this
composite matrix cast onto an oxide ion conducting interface also being exposed at its edge to the gas
electrolyte substrate using a low-cost slurry-based phase (β) (where O2 is available). Phase R is con-
process involving powder precursors, such as screen nected at some point away from the interface to a
printing. The oxygen, which we wish to reduce to O2-, source of electronic current, providing a conduction
diffuses into the open pores of the electrode and is path for electrons to the interface. Likewise, the
reduced somewhere within this matrix. For a tradi- electrolyte phase γ is either itself the electrolyte
tional porous electronic conducting electrode material membrane or connected by a continuous ionic path
such as Pt or (La,Sr)MnO3 (LSM) (Figure 2a), reduc- to the electrolyte membrane, providing a sink for the
tion of oxygen is generally thought to be confined oxygen ions produced in the reaction. The oxygen gas
close to the electrode/electrolyte interface, where the (which diffuses from outside the electrode through
gas has simultaneous access to both the electronically interconnected pores or channels) is reduced some-
and ionically conductive phases (illustrated as the where in the vicinity of this R/β/γ interface.
shaded active region). Since this reaction involves ions, electrons, and gas
One strategy for trying to improve performance has molecules in three separate phases, the edge of the
been to replace LSM with a single-phase mixed R/γ interface that makes contact with the gas phase
conductor (material which conducts both oxygen ions β is often described as the three-phase (or triple-
and electrons), such as La1-xSrxCo1-yFeyO3-δ (LSCF) phase) boundary (TPB). The concept of the TPB
(Figure 2b). By introducing bulk ionic transport, actually dates to the 1920s,41 when workers studying
oxygen can be reduced to O2- over a significant the oxidation of H2 on platinum introduced this
portion of the electrode surface, thereby extending concept to explain why Pt must be exposed simulta-
the size of the active region and improving the neously to both solution and gas to get significant
kinetics at temperatures below 800 °C.9,13-28 Elec- reaction. This type of electrode, which Schmid called
tronically conductive LSM can also be combined with “die diffusiongaselektrode”41 or gas-diffusion elec-
an ionically conducting oxide (YSZ itself or rare- trode (GDE), is still called this today by workers
earth-doped ceria) in a porous composite microstruc- studying solution- or polymer-based fuel cells. As
Figure 4. Some mechanisms thought to govern oxygen reduction in SOFC cathodes. Phases R, β, and γ refer to the
electronic phase, gas phase, and ionic phase, respectively: (a) Incorporation of oxygen into the bulk of the electronic phase
(if mixed conducting); (b) adsorption and/or partial reduction of oxygen on the surface of the electronic phase; (c) bulk or
(d) surface transport of O2- or On-, respectively, to the R/γ interface, (e) Electrochemical charge transfer of O2- or (f)
combinations of On- and e-, respectively, across the R/γ interface, and (g) rates of one or more of these mechanisms wherein
the electrolyte itself is active for generation and transport of electroactive oxygen species.
implied by the name, all GDE reactions share the Figure 4 outlines some of the mechanisms either
common feature of involving gas-phase molecules and known or theorized in the literature to be important
thus represent a fundamental departure from tradi- in determining the rate of the oxygen reduction in
tional electrochemical kinetics governing charge trans- SOFC cathodes. Oxygen molecules are generally
fer across a 2-dimensional interface.42 thought to adsorb somewhere onto one or more solid
In the case of SOFCs, a large volume of work shows surface(s), where they undergo catalytic and/or elec-
that for many SOFC electrodes, overall performance trocatalytic reduction steps to form partially reduced
scales with the 1D geometric length of this three- ionic/atomic species (sometimes called “electroactive
phase boundary. As such, the TBP concept and species”46). Before, after, or between partial reduction
electrode performance models based on it have proven steps, these species must transport along surfaces,
to be some of the most useful phenomenological interfaces, or inside the bulk of the electrode materi-
concepts for guiding design and fabrication of SOFC al(s) to the electrolyte, where they are fully and
cathodes, particularly the microstructure. formally incorporated as electrolytic O2-. If, how, and
However, in trying to develop a deeper understand- where any of these processes happen and what
ing of the physics and chemistry governing the SOFC step(s) are rate determining for a particular electrode
cathode reaction, the concept of the three-phase is often only partially understood.
boundary somewhat begs the question as to what As evidenced by the cases reviewed below, no single
specific physical processes actually occur in the mechanism has been discovered which explains all
vicinity of the three-phase interface and how these electrodes. Furthermore, arguments in the literature
processes depend on materials properties, micro- concerning which theory is correct for a given cathode
structure, etc. Workers studying aqueous gas-diffu- system usually end in a standoff of equally legitimate
sion electrodes in the mid-1960s recognized the interpretations of limited data on dubiously compa-
limitations of the three-phase boundary concept.43,44 rable samples. Which step(s) are rate determining,
As an alternative, they began to break down the and thus which factors are important for cathode
electrode reaction into individual steps, including the performance, depends strongly on the material(s),
dissolution and diffusion of H2 in solution and the microstructure, and processing of the electrode as
oxidation of dissolved H2 at the Pt/solution interface. well as the conditions under which the electrode is
These and subsequent studies contributed signifi- tested, including temperature, atmosphere, polariza-
cantly to our fundamental and practical understand- tion, time, or other factors, some of which may not
ing of solution and polymer GDEs, including how be known.
molecular diffusion processes contribute to the overall Given these uncertainties, the approach taken here
i-V characteristics and how best to arrange the is to review the asymptotic behavior of well-charac-
electrolyte inside a flooded electrode. Of particular terized and studied systems, where enough scientific
note is the observation that diffusion processes can evidence exists to reach a reasonable consensus.
often co-limit electrode performance far below limit- Perhaps then it becomes easier to discuss the possible
ing current and can mimic activated behavior.45 processes governing more complex and less well-
Thus, the phenomenological observation that a GDE characterized systems (and to propose a new set of
obeys Tafel (or Butler-Volmer) kinetics provides no testable hypotheses about them). To that end, we
assurance that the electrode is, in fact, activation begin by examining the case of platinum and other
limited. noble metals on zirconia, which perhaps enjoys the
Although the SOFC community has generally longest and richest history of research available
maintained an empirical approach to the three-phase today.
boundary longer than the aqueous and polymer
literature, the last 20 years have seen a similar 3. Platinum ElectrodessInterplay of Chemical
transformation of our understanding of SOFC cath- and Electrochemical Steps
ode kinetics. Few examples remain today of solid-
state electrochemical reactions that are not known Oxygen reduction on yttria-doped zirconia (YSZ)
to be at least partially limited by solid-state or was first accomplished over 100 years ago (for the
surface diffusion processes or chemical catalytic purpose of generating light) using a porous platinum
processes remote from the electrochemical-kinetic electrode.4 Since then, oxygen reduction on Pt has
interface. been one of the most well-studied electrode reaction
Figure 5. Measurement and analysis of steady-state i-V

characteristics. (a) Following subtraction of ohmic losses
(determined from impedance or current-interrupt measure-
ments), the electrode overpotential η is plotted vs ln(i). For
systems governed by classic electrochemical kinetics, the Figure 6. Electrochemical impedance spectroscopy (EIS).
slope at high overpotential yields anodic and cathodic (a) The steady-periodic linear response of a cell to a
transfer coefficients (Ra and Rc) while the intercept yields sinusoidal current or voltage perturbation is measured and
the exchange current density (i0). These parameters can analyzed in terms of gain and phase shift as a function of
be used in an empirical rate expression for the kinetics frequency (ω). Results are usually expressed in terms of
(Butler-Volmer equation) or related to more specific the impedance (Z), the complex ratio of voltage displace-
parameters associated with individual reaction steps.42 (b) ment to current displacement (often reported on a Nyquist
Example of Mn(IV) reduction to Mn(III) at a Pt electrode or Bode plot). Ideally, each charge-transfer-limiting process
in 7.5 M H2SO4 solution at 25 °C.362 Below limiting current (electrolyte, electrode process #1, electrode process #2, etc.)
the system obeys Tafel kinetics with Ra ≈ 1/4. Data are is only manifest below a distinct characteristic frequency,
from ref 363. (Reprinted with permission from ref 362. resulting in separate features or “arcs” in the impedance
Copyright 2001 John Wiley & Sons.) diagram of magnitude Re, R1, R2, etc. (b) Since EIS is a
linear response technique, the time-response of individual
in the history of solid-state ionics. Yet the Pt/solid processes is often modeled in terms of equivalent circuit
electrolyte interface is still actively studied today, elements, with resistance representing charge transfer and
with many open questions remaining. Many of these capacitance (C1, C2, etc.) representing charge polarization.
questions are relevant to more complex but techno- This description is often extended to noninterfacial pro-
cesses by introduction of additional empirically or theoreti-
logically advanced cathode materials with relevance cally derived circuit elements. (c) Example of a dense
to today’s SOFCs. Thus, platinum constitutes a useful La0.5Sr0.5CoO3-δ film on single-crystal YSZ at 750 °C in
foundation on which to build our understanding of oxygen.162 The total impedance in this case is a sum of the
all SOFC cathodes. electrolyte impedance (high frequency), an interfacial
impedance at medium frequency (MF), and a low-frequency
3.1. Two Schools of Thought (LF) impedance associated with O2 oxidation/reduction at
the surface of the film (see section 4.3). Data are from ref
The field of electrochemical kinetics has tradition- 165. (Adapted with permission ref 165. Copyright 2001
ally focused on charge-transfer reactions occurring American Institute of Physics.)
at the surface of a metal electrode in contact with a
liquid electrolyte.42 As shown in Figure 5, the steady- of frequency, EIS seeks to separate and identify
state i-V characteristics for this type of reaction reaction steps via time scale.48,49 Along with current
often obey Tafel kinetics (linear dependence of ln(i) interruption experiments, early measurements of this
vs V), which can be analyzed in terms of specific rate- type on Pt/YSZ or Pt/ceria allowed workers to more
limiting steps occurring at the interface.47 As such, specifically isolate the electrode polarization from the
measurement and analysis of Tafel parameters have electrolyte and begin analyzing this polarization in
proven to be one of the most useful techniques for terms of time-dependent phenomena.
understanding electrode reactions, both as a means On the basis of the two classes of measurements
of isolating the nonlinear electrode response from the shown in Figures 5 and 6, workers split into two
linear electrolyte losses as well as providing a theo- fundamentally different schools of thought. The first,
retical framework for relating these parameters to rooted in the well-established tradition of classical
specific mechanisms. electrochemical kinetics, focused on the observation
However, as mentioned previously, gas-diffusion that these electrodes tend to obey Tafel kinetics at
electrodes usually deviate substantially from tradi- moderate to high overpotential. For example, Figure
tional electrochemical-kinetic behavior, often being 7 shows Tafel plots for Pt electrodes on calcia-doped
limited by multiple rate-determining factors and/or ceria, measured by Wang and Nowick using current
changes in those factors with overpotential or other interruption techniques over a wide range of T and
conditions. In attempting to analyze this type of PO2.50,51 Except for limiting current behavior at high
electrode, one of the most influential experimental cathodic overpotential, these results fit a Butler-
techniques to take hold in the solid-state electro- Volmer expression, yielding (at least empirically) an
chemical literature in the last 35 years is electro- exchange current density and anodic and cathodic
chemical impedance spectroscopy (EIS)salso know transfer coefficients. On the basis of this result, the
as a.c. impedance. As illustrated in Figure 6, by authors concluded that the electrode reaction must
measuring the sinusoidal i-V response as a function be limited by electrochemical kinetics at the interface.
Figure 8. Complex admittance (reciprocal of impedance)

or porous Pt on YSZ in air at 700 °C. The largest portion
of the impedance (appearing at lowest admittance) exhibits
a response frequency of ∼20 Hz (response time of
∼0.01 s). (Reprinted with permission from ref 54. Copyright
1973 Laboratoire d’Electrochimie et de Physicochimie des
Materiaux et des Interfaces (INPG and CNRS), Saint
Martin d’Heres, France.)
As discussed below, the hindsight of the last 30

years suggests that the truth falls in a gray area
between these two schools of thought, with chemical
adsorption, surface transport, and interfacial elec-
trochemical kinetics all playing a significant role (as
Figure 7. Steady-state cathodic current-overpotential well as other factors which are still not fully under-
characteristics of porous Pt electrodes on Ca-doped ceria, stood). For this reason, platinum serves as a good
measured at 600 °C in air using current-interruption. starting point since it helps to illustrate how these
(Reprinted with permission from ref 51. Copyright 1979 various factors come into play and how they influence
Electrochemical Society, Inc.) the overall electrode performance and characteristics.
The authors modeled the PO2 dependence of the 3.2. Chemical Contributions to the Overpotential
exchange current density in terms of the dissociative
adsorption of oxygen on the Pt surface, which influ- To better understand the “diffusion-limited” school
ences the exchange current through the equilibrium of thought mentioned above, it is worth digressing
concentration of adsorbed atomic oxygen. These momentarily on another “noble”-metal electrode sys-
measurements also showed that the kinetics are tem: silver on YSZ. Kleitz and co-workers conducted
independent of electrolyte dopant concentration, a series of studies of silver point-contact microelec-
leading the authors to suggest that the limiting trodes, made by solidifying small (200-2000 µm)
electrochemical step must occur on the Pt surface. silver droplets onto polished YSZ surfaces.55 Follow-
Subsequent publications (as recent as the late 1990s) ing in-situ fabrication, the impedance of these silver
have continued to reinforce a default viewpoint that microelectrodes was measured as a function of T
SOFC gas-diffusion electrodes are fundamentally (600-800 °C), PO2 (0.01-1.0 atm), and droplet radius.
limited by electrochemical kinetics.17,52,53 As an example, Figure 9a shows a Nyquist plot of
In contrast, the second school of thought focused the impedance under one set of conditions, which the
on the impedance of these electrodes and the fact that authors resolve into two primary components, the
when modeled as an equivalent RC circuit, they largest (most resistive) occurring at very low fre-
exhibit a very large capacitancestoo large to be quency (0.01-0.1 Hz) and the second smaller com-
explained in terms of traditional double-layer polar- ponent at moderately low frequency (∼10 Hz).
ization at an interface.48 One of the earliest groups Restricting our attention to the more significant,
to apply impedance for this purpose was Kleitz and lower-frequency impedance, the authors argue (based
co-workers, who studied porous Pt and other noble- on time scale as well as other factors) that a majority
metal catalysts on YSZ.54 As shown in Figure 8, they of the observed overpotential is associated with
generally found that cells respond at frequencies well absorption and diffusion of atomic oxygen inside the
below 1000 Hz, several orders of magnitude too low silver droplet. Their proposed mechanism is illus-
to explain in terms of traditional interfacial polariza- trated in Figure 9b. According to this model, cathodic
tion. Rather, the authors interpreted these low- polarization provides a driving force for atomic
frequency capacitive effects as changes in concentra- oxygen dissolved in the silver to be reduced to oxygen
tion of “neutral-O” in the vicinity of the three-phase ions and pulled into the electrolyte at the silver/YSZ
boundary, concluding that the overpotential must be interface. The resulting depletion of oxygen in the
(at least in part) a concentration overpotential rather metal near the silver/YSZ interface creates a chemi-
than a simple electrochemical-kinetic resistance. cal potential driving force for dissolved oxygen to
Figure 10. Kleitz’s reaction pathway model for solid-state

gas-diffusion electrodes. Traditionally, losses in reversible
work at an electrochemical interface can be described as a
Figure 9. (a) Impedance of a single silver droplet solidified series of contiguous drops in electrical state along a current
on YSZ, measured at 700 °C in oxygen. (b) Interpretation pathway, for example, A-E-B. However, if charge transfer
of the impedance in terms of bulk absorption and transport at point E is limited by the availability of a neutral
of oxygen in silver. Explanation for the bifurcation in electroactive intermediate (in this case ad(b)sorbed oxygen
response frequency: At low frequency, transport occurs at the interface), a thermodynamic (Nernstian) step in
everywhere in the droplet, with a higher flux near the electrical state (δµ) develops, related to the displacement
droplet edge. At medium frequency transport is confined in concentration of that intermediate from equilibrium. In
to a small but uniform region near the silver/YSZ interface, this way it is possible for irreversibilities along a current-
involving little reduction/oxidation of gaseous O2. (Adapted independent pathway (in this case formation and transport
with permission from ref 364. Copyright 1997 Electro- of electroactive oxygen) to manifest themselves as electrical
chemical Society, Inc.) “resistance.” This type of “chemical valve”, as Kleitz calls
it, may also involve a significant reservoir of intermediates
that appears as a “capacitance” in transient measurements
diffuse to the interface from the bulk of the silver, such as impedance. Portions of this image are adapted from
leading ultimately to depletion of dissolved oxygen ref 46. (Adapted with permission from ref 46. Copyright
near the silver/gas interface. Depletion of oxygen in 1993 Risø National Laboratory, Denmark.)
the silver relative to equilibrium with the gas then
provides a driving force for dissociative absorption For the purposes of clarity throughout the rest of
of gaseous O2 into the silver. Thus, at steady state, a this review, we hereby define such a neutral flow as
continuous gradient in oxygen concentration is es- a chemical process, since it is driven by chemical
tablished that partly limits the overall reaction rate. potential driving forces and may occur at a rate
The authors argue that only this mechanism can independent of the faradaic current (except in the
explain the very large effective capacitance of the limit of steady state).366 Applying this definition,
electrode (∼1 F/cm2). Further evidence for this in- examples of chemical processes include ad(b)sorption
terpretation includes both the T and PO2 dependence and/or dissociation of molecular species, gas-phase
of the resistance and capacitance, which are found diffusion, diffusion of dissolved neutral species, am-
to be quantitatively consistent with independently bipolar (chemical) diffusion of neutral combinations
measured oxygen absorption and diffusion properties of ions/electrons, or surface diffusion of adsorbed
of silver. Van Herle and McEvoy subsequently con- species associated with a mobile mirror charge in the
firmed these findings,56 concluding further that the solid phase. Since such species or combinations
kinetics are partially limited by both absorption of thereof are net neutral, they can accumulate in large
gaseous O2 into the silver as well as diffusion of quantity in bulk or on surfaces without violating
atomic oxygen through the silver. electroneutrality, acting as a reservoir and leading
Although there are a number of other unresolved to slow response times. As we will see for Pt, as well
issues in interpreting these data (not discussed here), as ensuing examples involving mixed conductors, all
what is clear from their analysis is that a significant electrode reactions are now believed to involve some
portion of the resistance to oxygen reduction in this sort of “chemical valve” of this type which may
system is not determined by how fast oxygen is contribute none, some, or all of the macroscopically
reduced at the Ag/YSZ interface but rather how fast observed polarization resistance.
oxygen can get to that interface. On this basis and
other examples, Kleitz argued in 1993 that all 3.3. Evidence for Diffusion and Adsorption
cathode reactions involve the equivalent of dissolved Limitations on Platinum
oxygen in silverssome sort of “electroactive” oxygen
species, which must be replenished somehow from A significant step forward in our understanding of
the gas phase. This principle, which he titled the Pt was taken by Verkerk and Burgraff, who in 1983
“reaction pathway model”, is illustrated in Figure 10. analyzed the impedance of porous sputtered Pt (and
In this model, any charge-transfer (current) pathway Pt gauze) electrodes on YSZ and gadolinia-doped
A-E-B can be intermediately limited at point E by ceria (GDC).57 As shown in Figure 11, they used a
the availability of a neutral species (or neutral Randles circuit to model the interfacial contributions
combination of species), whose flow rate (C-E) is not to the impedance, allowing them to subtract from the
driven directly by electrical-state driving forces. data the contributions of uncompensated iR and
way they showed that chemical steps (embodying

Kleitz’s reaction pathway concept) can still obey
“classical” Butler-Volmer kinetics.
How can diffusion or other chemical steps give rise
to Tafel behavior? An examination of Mizusaki’s
derivation59 shows that even when interfacial elec-
trochemical kinetic steps are equilibrated, there
remains a Nernstian relationship between the ap-
plied potential and the activity of electroactive oxygen
on the Pt surface. Thus, as the steady-state potential
is varied, the surface concentration at the TPB
changes approximately logarithmically. If the rate of
adsorption and/or diffusion has a linear or power-law
dependence on the surface concentration, it will also
depend logarithmically on the potential, leading to
a logarithmic dependence of current on potential.
Thus, the mere existence of a Tafel slope means very
Figure 11. Electrochemical impedance of porous Pt on little in terms of proving electrochemical kinetics is
YSZ at 710 °C and PO2 ) 10-3 atm. Data has been corrected limiting or in identifying electrochemical kinetic
for double-layer capacitance as described in ref 79. (Re- steps. Rather, the important issue is how the Tafel
printed with permission from ref 79. Copyright 1983 The parameters (empirical though they may be) depend
Electrochemical Society, Inc.) on PO2 and temperature and whether these can be
rationalized in terms of specific chemical or electro-
double-layer capacitance at the metal/electrolyte chemical steps.
interface. The remaining “faradaic” impedance (ZF)
was found, in the limit of high frequency, to have During the late 1980s some disagreement appeared
similar real and imaginary contributions (∼45° line in the literature concerning where oxygen concentra-
on a Nyquist plot), which is a strong indicator that tion gradients exist on the Pt surface. In contrast to
diffusion is an important step in the reaction.49 They Mizusaki, Wang proposed a model in which electro-
also showed that the PO2 dependence of the imped- active oxygen is reduced along the entire interface
ance is independent of the electrolyte, suggesting that between Pt and YSZ, with the rate-limiting diffusion
the diffusion process occurs on the Pt surface (or in process being the diffusion of adsorbed oxygen along
the gas phase at lower frequencies and low PO2). On the Pt/YSZ interface from the TPB.61 A similar model
the basis of independent measurements of oxygen was considered by van Hassel for Au on YSZ.62 In
diffusion rates on Pt,58 the authors estimated the 1990, however, Robertson and Michaels addressed
diffusion distance to be about 50 nm, which supports this question for Pt by modeling steady-state polar-
a commonly held understanding that electrode kinet- ization and potential-step chronoamperometry mea-
ics should scale with total TPB length (at least for surements on Pt/YSZ.63 As illustrated in Figure 12a,
Pt particles larger than ∼10-5 cm). However, an they considered two cases. Model I: Adsorbed oxygen
important element left out of this analysis is why the is reduced along the entire Pt/YSZ interface but must
diffusion distance is ∼50 nm. What establishes this diffuse along this interface from the TPB (where
distance? Also, if diffusion is rate limiting, how can surface coverage is considered to be in equilibrium
one explain the Tafel behavior observed for these with the gas). Model II: Adsorbed oxygen is reduced
electrodes at moderate to high overpotential? at the three-phase boundary but due to lack of
In 1987, Mizusaki and co-workers published two equilibrium with the gas must diffuse along the gas-
papers that addressed some of these questions.59,60 exposed surface from some other point further away.
By applying traditional electrochemical kinetics, they They found that the qualitative and quantitative
argued that the Langmuir isotherm for oxygen features of their data could only be explained using
adsorption on platinum yields the incorrect PO2 this second model. In particular, as shown in Figure
dependence for the exchange current density ex- 12b, only diffusion on the gas-exposed Pt surface
tracted from the data by fitting to a Butler-Volmer could explain the t1/2 semi-infinite diffusion (Cottrell)
rate expression. They therefore concluded that elec- behavior of the current at short times. In contrast,
trochemical kinetics at the three-phase boundary the first model predicts constant current until oxygen
cannot be the rate-limiting step. Instead, they pro- is depleted at the interface, followed by exponential
posed that the reaction is limited by either (1) decay, which was not observed. Subsequent studies
dissociative adsorption of oxygen molecules on the Pt of the PO2 dependence of the impedance have rein-
surface or (2) surface diffusion of adsorbed oxygen forced the view that the dominant rate-determining
atoms to the three-phase interface. On the basis of transport step occurs on the gas-exposed Pt sur-
the PO2 dependence found in their measurements, face.56,64,65
they conclude that both situations are possible and Some debate has also appeared concerning the
that above 600 °C the electrode is diffusion limited “charge state” of adsorbed oxygen on the Pt surface,
while below 500 °C it becomes limited by the rate of whether it is O2-, O-, uncharged O, etc. This question
dissociative adsorption of O2. In either case, however, arises, for example, when one tries to interpret Tafel
they derive a logarithmic dependence of current on parameters in terms of a cascade of surface-mediated
overpotential, which mimics Tafel kinetics. In this electrochemical-kinetic steps.66 However, as already
Figure 12. Modeling and measurement of oxygen surface

diffusion on Pt. (a) Model I: adsorbed oxygen remains in Figure 13. XPS spectra of adsorbed oxygen on a thin-film
equilibrium with the gas along the gas-exposed Pt surface Pt electrode on single-crystal YSZ: (a) residual O1s spec-
but must diffuse along the Pt/YSZ interface to reach an trum following several “cleaning” cycles at elevated tem-
active site for reduction. Model II: adsorbed oxygen is perature and vacuum (<10-9 mbar); (b) O1s spectrum at
reduced at the TPB but must diffuse there from the gas- 350-400 °C under constant PO2 ) 10-6 mbar; (c) O1s
exposed Pt surface, which becomes depleted of oxygen near spectrum at 350-400 °C in a vacuum during anodic
the TPB due to a finite rate of adsorption. (b) Cotrell plot polarization. The peak “P1”, corresponding to the known
of current at a porous Pt electrode at 600 °C and PO2 ) spectrum for chemisorbed oxygen at 530.4 eV, appears
10-4 atm vs time. The linear dependence of current with upon both adsorption of oxygen from the gas (b) and
t-1/2 at short times implies semi-infinite diffusion, which electrochemical pumping of oxygen from YSZ (c). A second-
is shown by the authors to be consistent only with Model ary peak “P2” at lower binding energy also appears upon
II. (Reprinted with permission from ref 63. Copyright 1990 electrochemical pumping but is too close in energy to P1
Electrochemical Society, Inc.) to explain in terms of differences in charge state. The
authors propose that P2 represents a different adsorption
site populated only at high coverage. (Reprinted with
demonstrated by the work reviewed above, a signifi- permission from ref 67. Copyright 2001 Elsevier.)
cant fraction of the overpotential can arise from
chemical steps (adsorption and/or diffusion). In ana- these processes are independent of electrical state
lyzing the rates of these chemical processes, one must since they involve neutral combinations of adsorbed
question whether such partial charge assignments oxygen and mirror charges in Pt.
really have any physical meaning from the viewpoint
of continuum mechanics. As recently shown by
3.4. Important Refinement: Co-limited Reaction
Luerssen et al. (Figure 13),67 XPS reveals no differ-
ence in electronic structure between oxygen disso-
Kinetics
ciatively adsorbed onto Pt from the gas vs “ionic As shown by Mizusaki,59,60 the faradaic processes
oxygen species” pumped electrochemically from YSZ occurring on platinum are quite complex, with the
onto the Pt surface. Both oxygen sources result in possibility that more than one rate-limiting step can
the same change in measurable work function dominate under various conditions. A primary piece
(through the surface dipole concentration). In other of evidence that diffusion on Pt plays a role is a 45°
words, the “charge” associated with adsorbed oxygen straight-line relationship in the Nyquist plot at high
appears to be a unique feature of the surface chem- frequency, which has led to the widespread use of
istry of adsorption (e.g., ionic vs covalent) rather than Warburg elements in equivalent circuit modeling.
a variable formal valance state subject to electro- However, Pt electrodes generally exhibit an imped-
chemical reduction/oxidation. In this sense, it is ance considerably different from than predicted by a
somewhat irrelevant what the charge on oxygen is, Warburg element, even when corrected for double-
since (as recently demonstrated by Janek68) the layer capacitance.57 Indeed, as recently pointed out
driving force for adsorption and transport of absorbed by Boukamp, the 45° relationship is not unique to
oxygen is a change or gradient in surface coverage finite-length diffusion, but it is the case for any
(related to oxygen activity). The rate expressions for system that reduces to semi-infinite diffusion in the
limit of high frequency, including systems limited

partially (or entirely) by kinetic processes at steady
state.69
To that end, an important idea contributed by
Robertson and Michaels63 was that oxygen reduction
on Pt could potentially be co-limited by adsorption
and diffusion rather than by just one or the other.
In modeling the system, they noted that it is not
possible for adsorbed oxygen to be in chemical equi-
librium with the gas at the gas-exposed Pt surface
while at the same time being in electrochemical
equilibrium with the applied potential at the three-
phase boundary. To resolve this singularity, prior
(and several subsequent) models for diffusion intro-
duce an artificial fixed “diffusion length” governing
transport from the gas-equilibrated surface to the
TPB.56,57,59,64,65,70 In contrast, Robertson and Michaels
proposed that surface coverage is not in equilibrium
with the gas but instead is governed simultaneously
by finite rates of adsorption onto and diffusion along
the Pt surface. They show that in the limit of short
times, this scenario yields semi-infinite diffusion-
limited behavior (consistent with potential step and
impedance measurements) but at long times enters
a co-limited regime, where the rate of both adsorption
and diffusion govern the overall kinetics. Since
adsorption kinetics are likely to be governed by the
same mass-action relationships governing adsorption
equilibrium, the co-limited case is similarly supported
by the PO2 dependence of the overall kinetics as the
case of pure diffusion.
To better understand how a co-limited system can
exist, consider the generalized adsorption/diffusion
model shown in Figure 14a. A semi-infinite surface,
defining a one-dimensional coordinate system (x), is Figure 14. Simple model demonstrating how adsorption
exposed to a gaseous adsorbate (A) at uniform partial and surface diffusion can co-limit overall reaction kinetics,
pressure. The rate of adsorption of A onto the surface as explained in the text. (a) A semi-infinite surface estab-
per unit area is given by rA ) k(cA0 - cA), where cA(x,t) lishes a uniform surface coverage CA0 of adsorbate “A” via
is the local surface concentration of A, cA0 is the equilibrium of surface diffusion and adsorption/desorption
surface concentration at equilibrium with the gas, of A from/to the surrounding gas. (b) Concentration profile
and k is a surface exchange coefficient.367 Once of adsorbed species following a step (drop) in surface
coverage at the origin. (c) Surface flux of species at the
adsorbed, A can also diffuse along the surface, with origin (NA1(t)) as a function of time. Points marked with a
1-dimensional flux per unit width of NA ) -D(∂cA/ solid circle (b) correspond to the concentration profiles in
∂x), where D is the surface diffusion coefficient. h A1(ω)) resulting
b. (d) Surface flux of species at the origin (N
Initially, the entire surface is allowed to equilibrate from a steady periodic sinusoidal oscillation at frequency
with the gas at surface concentration cA0. Then at ω of the concentration at the origin.
t g 0, the concentration of species A at x ) 0 is
regulated externally with time-dependent value cA1(t) to this system following a step in concentration
(established, for example, by equilibrium of an elec- cA1(t) ) cA0 - ∆c at x ) 0. At short times the diffusion
trochemical reaction such as A + e- T A- ). This local profile expands as t1/2, yielding Cottrell behavior for
displacement of the surface concentration provides the flux of A at x ) 0 (NA1(t)). This expansion
a driving force for a finite flux of A at x ) 0 (NA1), continues until, at steady state, the increasing rate
which (if electrochemically driven) is related propor- of adsorption (due to increased active area) balances
tionately to current. Conservation of species A leads the decreasing rate of diffusion (due to a longer
to transport path). A utilization length lδ is established,
proportional to xD/k, that represents a “compro-
∂cA ∂2cA mise” between faster kinetics and slower diffusion.
) D 2 + k(cA0 - cA) The steady-state flux at x ) 0 (NA1) is proportional
∂t ∂x
to xkD and thus is governed by both kinetics and
cA(x,t < 0) ) cA0 diffusion. Similarly, Figure 14d shows the stationary
periodic solution to eq 2 for a sinusoidal perturbation
cA(x ) 0,t) ) cA1(t) (2) cA1(t) ) cA0 + ∆c cos(ωt) as a function of frequency.
The resulting half-tear-drop-shaped impedance (de-
Figure 14b,c shows the time-dependent solution fined here as ∆c/N h A1(ω) at x ) 0) is somewhat
in Figure 14d. The authors constructed a numerical

model for oxygen reduction on Pt/YSZ (based on a
finite-difference model for the surface) that includes
adsorption, diffusion, and electrochemical-kinetics
at the TPB. As shown in Figure 15c,d, their model
shows similar behavior and when compared to the
data allows estimation of physical parameters gov-
erning adsorption, surface diffusion, and electro-
chemical kinetics at the TPB.
Among other things, the authors show that when
Langmuir adsorption kinetics are assumed, they find
the adsorption and diffusion parameters extracted
from their data to depend strongly on surface cover-
age θ. To explain this, they propose an alternative
precursor-mediated adsorption mechanism. With this
Figure 15. Faradaic impedance (ZF) of porous Pt elec- modification to the model, the kinetic and transport
trodes on single-crystal zirconia in oxygen at zero bias, as parameters extracted from the data reveal more
determined from the total impedance (Z) using the methods constant values that are argued to be in reasonable
of Berthier et al.:75 (a) T ) 800 and (b) 900 °C. (c) Model agreement with available independent measure-
assuming the interfacial electrochemical kinetic resistance
(Rt) is small compared to chemical resistances. (d) Model
ments or calculated estimates.73,76-79 The authors
assuming Rt is similar to chemical resistances. (Adapted conclude that above 800 °C and high PO2, charge
with permission from refs 74 and 84. Copyright 1999 and transfer at the TPB competes with adsorption/diffu-
1997 Elsevier and Swiss Federal Institute of Technology.) sion, constituting approximately one-half of the total
impedance. With decreased temperature or lower PO2
smoother and more suppressed than a Warburg (or increased overpotential), the electrode appears to
impedance, with a 45° limit at high frequency and become limited primarily by chemical processes, with
RC-like behavior at low frequency, with a character- less than 10% attributed to the TPB interface. These
istic time constant proportional to 1/k and charac- results suggest that the system is quite complex and
teristic resistance proportional to 1/xkD. This re- that multiple factors can be important, depending on
sponse has often been called the “Gerischer impe- the specific conditions. Indeed, this may help explain
dance” because it was first derived by Gerischer in why there has been so much difficulty reaching a
1951 for the case of an electrode co-limited by consensus on what “the” rate-limiting step really is,
homogeneous reaction and diffusion in the electro- i.e., there is no “one” rate-limiting step.
lyte.69,71
The co-limited situation illustrated above is quite 3.5. Unresolved Issues Surrounding Oxygen
different than a membrane, for example, where the Reduction on Pt/YSZ
relative importance of diffusion vs surface kinetics
is determined by a geometric length scale (membrane The models proposed by Robertson and Michaels,
thickness). Rather, by “co-limited” we mean that both and later by Mitterdorfer and Gauckler, imply that
k and D are important in establishing the overall adsorption and diffusion of oxygen will occur over a
rate, regardless of their values, as long as lδ is small finite utilization region, as illustrated in Figure 14b,c,
compared to the electrode thickness. If the ratio of establishing the electrode’s functional zone. Oddly,
D/k changes, it does not change the relative impor- neither set of authors reported numerical estimates
tance of diffusion vs kinetics, merely the size of the of this size, based on the parameters they extracted
active region. This concept of a co-limited reaction from their data. In the case of Mitterdorfer and
rate is not unique to solids; electrochemical engineer- Gauckler, it is possible to make an estimate using
ing is rich with examples of co-limited systems, eq 2 and the parameters from their model. This
including examples from classical porous electrode exercise yields values between 50 and 500 nm,
theory47 as well as more closely related cases involv- depending on T and PO2 (∼100 nm at 700 °C in air).368
ing gas-diffusion electrodes in polymer-based fuel While this range represents a significant extension
cells.43,44 of the active region beyond the TPB, it is small
Perhaps the strongest recent evidence to date that enough relative to microstructural features to remain
Pt/YSZ is, in fact, co-limited by adsorption and consistent with studies of Pt and Pt/YSZ composite
surface diffusion has been provided by Mitterdorfer electrodes, showing that the active area scales with
and Gauckler.72-74 These authors prepared Pt paste the geometric length of the three-phase boundary.80,81
electrodes on single-crystal YSZ and measured im- A more quantitative analysis of the size of the active
pedance at T ) 700-800 °C and PO2 ) 10-4-1 atm in region has yet to be conducted for Pt on YSZ.
O2/N2 mixtures. As shown in Figure 15a,b, the Luerssen and co-workers recently attempted to
authors used the method of Berthier et al.75 to image the concentration gradients of oxygen on a Pt
subtract the effect of double-layer capacitance and film electrode near the TPB using PEEM67(see sec-
electrolyte resistance, yielding a true “faradaic im- tion 3.6). They were able to show that the work
pedance” ZF, which is composed of an interfacial function (and thus the surface oxygen coverage)
charge-transfer resistance (Rt) plus a chemical im- changes following polarization. However, they ob-
pedance not unlike the Gerischer impedance shown served no visible diffusion front emanating (on Pt)
from the TPB. As discussed by the authors, this smaller than the total pseudocapacitance typically
failure may be a result of the porosity in the Pt film ascribed to Pt) compare reasonably with estimates
(resulting in changes being seen everywhere). It is of the charge separation at the Pt/YSZ interface85 as
also possible that at the very low PO2’s studied (<10-9 well as the capacitance of solution interfaces of
mbar) the system becomes entirely adsorption/de- similar charge concentration and mobility.47 In con-
sorption limited, and thus the oxygen surface cover- trast, the use of equivalent circuit models to extract
age changes uniformly over the entire Pt surface. double-layer capacitance from impedance data typi-
Indeed, an estimate of lδ at this low PO2 based on cally yields values on the order of >10-4 F/cm2.73,85
Mitterdorfer’s model (see above) is ∼80 µm at While in some cases these values may simply be
700 °C, which is on the same order of the size of the obscured by faradaic capacitive effects (as originally
Pt surface studied by Luerssen et al. Further studies proposed by Robertson), it remains unclear why the
of this type, if able to be conducted at higher PO2, may apparent interfacial capacitance appearing in the
be able to reveal more about the actual size of the impedance is so large even after faradaic effects have
utilization region for Pt. supposedly been accounted for. A likely explanation
Another open question is the mechanism and is that additional transients are occurring at the Pt/
electrochemical kinetics governing incorporation of YSZ interface at the highest frequencies probed by
adsorbed oxygen as O2- at the Pt/electrolyte interface the impedance, which are neither true interfacial
(Figure 4f). Mitterdorfer estimated (based on analysis polarization nor faradaic accumulation of reactive
of impedance) that this process can constitute any- intermediates on the gas-exposed Pt surface.
where from 10% to 50% of the electrode overpoten-
tial.73 Of particular interest would be to better 3.6. Nonstationary Effects
understand what role the electrolyte (or its constitu-
ents) plays in this process. For example, Widmer et Another unresolved issue in our understanding of
al.82 measured the zero-bias impedance of low-fired Pt cathodes is that of nonstationary behavior such
(750 °C) porous Pt electrodes on various electrolytes as hysteresis or inductive effects, reported under
including YSZ, gadolinia-doped ceria (GDC), and Ce- moderate to high polarization. In this context, the
implanted YSZ (made by ion implantation of cerium term stationary refers to behavior exhibiting a well-
into the surface of YSZ). Their results show that the defined steady state, which is a repeatable function
impedance of Pt on GDC and Ce-implanted YSZ are of the materials, processing, and testing conditions.
much lower than for Pt/YSZ. Since electrodes fired Inherently transient measurements, such as imped-
at this temperature are likely to be limited by ance or current-interrupt response, can still be clas-
interfacial ionic charge transfer, this result would sified as stationary as long as the applied transients
seem to indicate that the presence of Ce aids the involve repeatable deviations from, or relaxations to,
electrochemical kinetics at the TPB, as discussed a well-defined steady state. In contrast, many work-
more fully by McEvoy.83 Unfortunately, isolation and ers have observed that electrode characteristics, such
study of this portion of the overpotential is nontrivial. as i-V response or the impedance itself, can depend
As discussed previously, traditional electrochemical on time and/or operating history. Such dependencies
methods (such as steady-state Tafel analysis) are are defined here to be nonstationary since they either
heavily obscured by chemical effects.84 Meanwhile, are irreversible, involve significant hysteresis, or
impedance (which we saw can potentially isolate the relax on a time scale so long as to constitute changes
resistance of the interface via frequency) is limited in mechanism with respect to processes probed on
to linearized response and thus is unable to provide ordinary time scales.
nonlinear information such as the interface-specific One of the first specific studies of these effects was
exchange current density and anodic/cathodic trans- reported by Schouler and Kleitz,87 who noticed hys-
fer coefficients. Thus, despite many years of earnest teresis in cyclic voltammograms at anodic potentials
investigation, the electrochemical kinetics governing above about 200 mV. Subsequently a variety of other
the actual Pt/YSZ interface (as distinct from adsorp- authors have reported similar findings for Pt at both
tion and transport to the TPB) remains largely a anodic and cathodic overpotentials.88-92 One recent
mystery. example is provided by Jacobsen and co-workers,90
A related mystery is the wide variability in esti- who used linear sweep voltammetry and EIS to
mates reported for the true double-layer capacitance quantify the degree and dependence of this hysteresis
of Pt/YSZ interface and how sensitive one’s analysis for Pt ball/YSZ and point YSZ/Pt electrodes. Figure
is to this value. The recent work of Mitterorfer and 16 shows some of their results. Unlike a system with
Gauckler74 shown in Figure 15 illustrates the ex- a reactive intermediate (in which the return sweep
treme degree to which the double-layer capacitance occurs at reduced current due to unfavorable deple-
can obscure the faradaic processes occurring on the tion or excess of the reactive intermediate), these
electrode at high frequency. In 1991, Robertson and voltammograms clearly show that current on the
Michaels attempted to measure the double-layer return sweep is enhanced following polarization at
capacitance based on chronoamperometry, showing a higher potential. The hysteresis loop gets larger
that the shortest relaxation time scale yields a with decreasing sweep rate until extremely slow
capacitance of ∼10-6 F/cm2 at 700 °C, based on the sweeps, where a steady state is finally achieved (e1
superficial area.85 More recently, Kenjo reports a µV/s). This is actually an example of an “apparent”
value of ∼10-5 F/cm2 using similar measurements.86 hysteresis (brought on by a large separation in time
These values (which are 2-3 orders of magnitude scales) rather than a true hysteresis in which a
Figure 17. In-situ PEEM images of a cathodically polar-

ized Pt/YSZ/Pt cell as a function of time following cathodic
polarization. Dark area in the initial (t ) 0) image corre-
sponds to the gas-exposed YSZ surface, while the slightly
lighter area is Pt. The bright area on Pt following initial
polarization corresponds to low work function due to
decreased oxygen coverage. In contrast, the bright area that
grows slowly outward from the TPB along the YSZ surface
corresponds to increased electron concentration. About 1
min is required for the YSZ surface to reach steady state
following polarization. (Reprinted with permission from ref
Figure 16. Linear-sweep voltammagrams of a YSZ needle 97. Copyright 2002 PCCP Owner Societies.)
electrode in contact with bulk Pt in air at 1000 °C as a
function of sweep rate. (Reprinted with permission from cal reduction/oxidation, might account for unex-
ref 90. Copyright 2001 Elsevier.) plained inductive (as well as capacitive) effects.
Recent measurements of work function transients
history-dependent stable state can be achieved (e.g., following polarization94 seem to support the idea that
magnetic induction of a ferromagnet). As one might Pt undergoes changes in surface structure after
expect from this behavior, inductive effects are seen establishing a steady-state surface coverage. Simi-
in the impedance, and this effect can also be observed larly, a recent AFM study by Bay and Jacobsen shows
in chronoamperometric measurements, which show that cathodic and anodic polarization can create
a slow rise (∼30%) in current over a period of about submicrometer morphological changes at the Pt/YSZ
3 h following initial polarization. Jacobsen et al. interface, as revealed in postmortem analysis of
consider, and reject, a number of explanations for this point-electrode samples.95 These observations, which
behavior including (1) reactive intermediates on the imply a nonstationary Pt/YSZ interface, appear to be
Pt surface, (2) passivation of the Pt surface by consistent with a model by Svensson and Nisancio-
macroscopic layers of platinum oxide, (3) expansion glu,96 which suggests that mobility of oxygen species
of the reaction zone along the YSZ surface due to along the YSZ/Pt interface is an important part (at
reduction and subsequent conduction of YSZ, and (4)
the atomic level) in determining the interfacial
Joule heating of the sample. None of these mecha-
electrochemical kinetics.
nisms, the authors argue, appear to consistently
explain the observed transients. Finally, it is worth mentioning recent studies by
A prominent explanation has been that by cathodi- Luerssen et al.,97 who imaged the slow growth of a
cally polarizing the cell for a significant period of reduction front along single-crystal YSZ following
time, long-lived Pt-O moieties on the surface (or at large cathodic polarization of a Pt film electrode. As
the TPB) are reduced, opening up catalytic sites for shown in Figure 17, this reduction expands the region
dissociative adsorption, transport, and electrochemi- of reduced oxygen activity well beyond the TPB and
cal reduction. Perhaps relevant to this idea is a study appears to alter the work-function distribution on Pt
by von Oertzen and co-workers that challenges the that had been established initially after polarization.
traditional view of Pt as a homogeneous metal Although these measurements were conducted under
surface.93 They used PEEM to show that chemisorbed very low PO2, they may provide some insight about
oxygen on Pt (110) can be incorporated into platinum how hysteretic effects could involve the electrolyte
as a subsurface oxygen “phase” at temperatures as at higher PO2. Indeed, these images recall earlier
low as 200 °C, provided CO is present to initiate suggestions by Kleitz55,98 and others99 that the elec-
reconstruction of the surface. Perhaps similar alter- trolyte can play a chemical catalytic role in the
ations of certain portions of a polycrystalline Pt absorption and dissociation of oxygen given the right
surface (or Pt/YSZ interface), driven by electrochemi- conditions and set of materials.
3.7. Summary: Platinum as a Framework for section 5.4), this type of nonstationary behavior
Understanding Other SOFC Cathodes (behavior for which a well-defined steady state does
not exist or is subject to relatively long relaxations)
The literature reviewed in sections 2-3.6 has remains largely an unsolved mystery for Pt as well
shown that oxygen reduction on Pt is quite complex, as other types of electrodes.
involving several possible rate-limiting (or co-limit-
ing) steps. As we will see in sections 4 and 5, this
complexity is a universal feature of all SOFC cath- 4. Perovskite Mixed Conductors and the Role of
odes, with many of the same themes and issues Bulk Material Properties
reappearing for other materials. We therefore high- The observations summarized in section 3.7 sug-
light below several general observations about the gest that if one could extend the transport process of
mechanism of Pt that frame the discussion for other electroactive species from the surface to the bulk of
solid-state gas-diffusion electrodes involving O2. These the electrode material (as we saw with silver, for
observations are as follows. example), one could enlarge the active area over
(1) Chemical reaction steps: Even if the overall which chemical processes occur as well as extend the
electrochemical reaction involves a molecular species electrochemical interface to include the entire elec-
(O2), it must first be converted to some “electroactive” trode/electrolyte contact area (not just the TPB).
intermediate form via one or more processes. Al- Although silver appears to exhibit sufficient bulk
though these processes are ultimately driven by oxygen transport to accomplish these goals, it has not
depletion or surplus of intermediates relative to proven to be suitable as an SOFC cathode due to poor
equilibrium, the rate at which these processes occur catalytic activity toward O2 as well as mechanical and
is independent of the current except in the limit of thermodynamic instability. In contrast, some transi-
steady state. We therefore label these processes as tion-metal oxides, in addition to being good O2
chemical processes in the sense that they are driven catalysts and electronic conductors, exhibit signifi-
by chemical potential driving forces. In the case of cant ionic conduction while remaining relatively
Pt, these steps include dissociative adsorption of O2 stable at operating conditions. For these mixed
onto the gas-exposed Pt surface and surface diffusion conductors (materials which conduct both ions and
of the resulting adsorbates to the Pt/YSZ interface electrons) the bulk appears to play a significant if
(where formal reduction occurs via electrochemical- not dominant role in determining the electrode kinet-
kinetic processes occurring at a rate proportional to ics. As such, these materials provide another useful
the current). asymptote to consider when extrapolating to more
(2) Co-limitation by kinetics and mass transfer: If complex materials such as LSM.
an electrode is porous, there is generally no fixed
geometric length that defines the importance of 4.1. Perovskite OxidessLow-Cost Alternative to
diffusion relative to kinetic steps. Rather, reaction Platinum
and diffusion will occur cooperatively over an active
area and at an overall rate that depends on both Transition-metal oxides were originally investi-
kinetic and diffusion parameters simultaneously over gated as SOFC cathodes due to their good electrical
a wide range of values. We label this situation as a conductivity (most oxides are insulators) and as a
co-limited reaction. In the case of Pt at high PO2, it relatively low-cost alternative to Pt, which prior to
would appear that the rates of adsorption and surface about 1965 was the only SOFC cathode material
diffusion are balanced over a relatively small active being considered extensively. One of the first such
region close to the TPB (50-500 nm). This appears materials studied was La1-xSrxCoO3-δ (LSC) (now one
to explain why the activity of a Pt electrode generally of the most well-studied mixed conductors), reported
scales with the length of the geometric three-phase by Button and Archer in 1966.4 This was followed
boundary. quickly thereafter by a number of other materials
(3) Electrochemical kinetics confined to TPB: For having perovskite crystal structure, including La1-xSrx-
platinum, the important chemical processes of cata- MnO3(δ (LSM), which as of ∼1973 became the
lytic reduction and transport occur on the Pt surface, favored material for SOFC cathodes. Since LSM has
and thus the electroactive species formed are most been so well studied and is also a poor ion conductor
readily available near the Pt-surface/YSZ interface (at least at ambient PO2), we discuss this somewhat
(TPB). Although it is not entirely clear how these more complex material system in its own section
electroactive species are subsequently incorporated (section 5). In the present section we focus on
into the electrolyte bulk or what role the electrolyte perovskites materials which exhibit significant bulk
itself plays in this process, there is strong circum- oxygen ion transport at typical cathodic PO2 and the
stantial evidence that the electrochemical-kinetic role this transport plays in the oxygen reduction
step is restricted to an area close to the TPB. mechanism.
Unfortunately very little is known about the nature One of the first such kinetic studies of a perovskite
of this reaction since the details are often obscured mixed conducting electrode was reported by Ohno
by chemical effects involving the Pt surface. and co-workers in 1981, who found La1-xCaxCoO3-δ
(4) Nonstationary behavior: Platinum on YSZ to have better kinetic properties than Pt as an SOFC
exhibits pronounced hysteretic effects, suggesting cathode at 1000-1100 °C.100 A number of other
that passage of current can alter either the kinetics πFο$σκιτσ of general formula La1-xSrxMO3-δ (M )
of the reaction or the dominant reaction pathway Cr, Mn, Fe, Co) were later studied by Takeda et al.101
itself. As we saw in section 3.6 (and will again in To avoid reaction of the perovskites with the YSZ
Figure 19. Atomic structure and oxygen transport in

Figure 18. Possible pathways by which oxygen is reduced mixed conducting perovskites ABO3-δ. (a) Basic structural
in a porous mixed conducting electrode. Following dissocia- element, consisting of a corner-sharing BO6 octahedron
tive adsorption (which forms a surface-polarizing species surrounded by charge-compensating A-site cations. (b) Bulk
On-, where n represents the unknown partial charge state oxygen transport mechanism, involving random hopping
of adsorbed oxygen), On- either travels by surface diffusion of oxygen ion vacancies on the oxygen sublattice.
to the TPB (where it is fully reduced) or is incorporated
directly into the mixed conductor as O2-, where it then particular interest is the suggestion that the bulk of
diffuses to the solid/solid interface. (Adapted with permis- the material plays an important role in determining
sion from ref 203. Copyright 1987 The Electrochemical the overall kinetics. As we will see below, this has
Society, Inc.)
been proven to be largely correct.
electrolyte, these electrodes were fabricated by sput-
tering porous films of thickness varying from 1 to 3
4.2. Thermodynamic, Kinetic, and Transport
µm at modest temperatures. They found the best
Properties of Perovskite Mixed Conductors
performance with LSC, which initially exhibits sig- Figure 19 illustrates the general perovskite crystal
nificantly lower overpotential than either La0.7Sr0.3- structure ABO3-δ as it relates to the electronic and
FeO3(δ (LSF) or LSM under the conditions studied. ionic transport properties of some transition-metal
The i-V characteristics of LSC were found to be oxides. In this case, the B-site cation is a reducible
linear at low current density and obey Tafel kinetics transition metal such as Co or Fe (or mixture thereof)
at higher current density. The exchange current and the A-site cation is a mixture of rare and alkaline
density obtained by fitting the data to a Butler- earths (such as La and Sr). The octahedral symmetry
Volmer rate expression scales approximately as PO21/4 around the transition metal often promotes a metallic
and has an activation energy of ∼220 kJ/mol. The or semiconducting band structure at high tempera-
authors generally found that the overall kinetics ture, leading to high electronic conduction. This
improves with increasing Sr content (x), except for structure is also quite stable relative to other crystal-
LaCoO3 (which was the best composition they stud- line phases, and thus with a judicious choice of A-
ied). They also noted a thickness dependencesthe and B-site cations, it can stably support a large
overall performance improved with thickness until number of oxygen ion vacancies (δ) at SOFC operat-
about 2-3 µm, at which point a plateau or slight ing conditions, thus facilitating significant bulk ionic
decline in performance with thickness was observed. oxygen transport. The literature surrounding the
In interpreting their results, Tekada et al. provided properties of transition-metal perovskites extends
the now often-duplicated picture shown in Figure 18, back more than 50 years and is well beyond the scope
which illustrates the possible paths by which oxygen of this review; however, the following few paragraphs
may become reduced. Citing the high chemical dif- attempt to summarize some of the salient features.
fusion coefficient for oxygen in LSC, the authors All materials in the La1-xSrxCo1-yFeyO3-δ (LSCF)
claimed that bulk transport in LSC allows a much family of materials have electronic transference
larger region of the electrode to be active for oxygen numbers approaching unity. The electronic structure
reduction and that this effect is largely responsible LSC and LSF has often been described in terms of
the improved performance over platinum. On the partially delocalized O2p-Co3d band states based on
other hand, citing the Tafel behavior, weak PO2 the tg2 and eg levels of crystal-field theory.102-104 In
dependence, and high activation energy, the authors the case of LSC (Figure 20a), at high temperature
argue that the electrode must be limited by interfa- and with 10 mol % Sr or higher, the Fermi energy
cial electrochemical kinetics with absorption and appears to fall within a half-filled band, comprising
transport of oxygen in/on LSC being so fast as to be a mixture of somewhat localized t2 states with more
equilibrated. Although the subsequent 20 years of delocalized e (σ*) states.105-111 Consistent with this
research have shown that these electrodes are not picture, LSC exhibits metallic or semimetallic con-
generally limited by interfacial electrochemical kinet- duction with a conductivity that decreases with
ics, this work was foundational in redirecting the temperature above ca. 500 °C and a Seebeck coef-
SOFC cathode materials effort and in framing the ficient approaching zero as the temperature is in-
debate for two decades as to what actually limits the creased.112,113 In contrast, LSF appears to exhibit a
rate of oxygen reduction with these materials. Of weak Hubbard or Hubbard-like band gap arising
Figure 20. Electronic structure and transport in mixed

conducting perovskites. (a) Band picture of electronic
structure in the high-temperature metallic phase of Figure 22. Correlation of the isotope tracer surface
La1-xSrxCoO3-δ. (Reprinted with permission from ref 109. exchange coefficient (k*) vs the oxygen tracer diffusion
Copyright 1995 Elsevier.) (b) Localized picture of electron/ coefficient (D*). Data represent a variety of materials under
hole transport in semimetallic La1-xSrxFeO3-δ, involving various temperature and PO2 conditions but are classified
hopping of electrons and/or electron holes (depending on according to groups. Group I: La1-xSrxMn1-yCoyO3(δ,
the oxidation state of iron). Sm1-xCoO3-δ, and La1-xSrxFe1-yCoyO3-δ. Group II: CaZr0.9-
In0.1O2.95, SrCe0.95Yb0.05O2.975, and La1-xSrxYO3-x/2. Group
IIIa: oxide fluorites at high temperature plus La0.9Sr0.1-
Ga0.8Mn0.2O2.85. Group IIIb: oxide fluorites at low temper-
ature. (Reprinted with permission from ref 141. Copyright
1999 Elsevier.)
increase in vacancy concentration with decreasing

PO2.121-124 Mizusaki noted that that this behavior
implies that the partial entropy of oxygen incorpora-
tion is dominated by the configuration entropy of
the vacancies while electron/hole entropy is con-
Figure 21. Oxygen stoichiometry in (a) LSF (x ) 0.1) and stant.121 More recently, Lankhorst et al. modeled this
(b) LSC (x ) 0.3) as a function of PO2 and temperature.
(Adapted with permission from refs 119 and 121. Copyright behavior using a point-defect view for oxygen vacan-
1985 and 1989 Elsevier.) cies but treating electrons as delocalized according
to a rigid-band model.125-127 Nonstoichiometry for
from electron-electron repulsion on the iron site.114,115 mixed iron and cobalt materials has also been
This gap results in more localized electronic states, investigated117,118,128,129 and can be explained to a
which can be thought of as a spatially localized certain degree by hybrid models invoking both local-
disproportionation of Fe3+ to Fe2+ and Fe4+(Figure ized and itinerant electronic states.130
20b). As such, LSF has semiconducting or semime- Although these materials are primarily electronic
tallic properties at high temperature, exhibiting conductors (tn,p ≈ 1), the very large concentrations
p-type conductivity and Seebeck coefficient at high of oxygen vacancies generated in these materials (as
PO2 (where iron is intermediate between Fe3+ and shown in Figure 19) has been found to lead to rapid
Fe4+) and n-type conductivity and Seebeck coefficient bulk oxygen ionic transport as well as increased rates
at low PO2 (where iron is intermediate between Fe2+ of exchange of oxygen with the gas. The rates of these
and Fe3+).116 For solid solutions of LSC and LSF processes have been measured by a variety of tech-
having a mixture of iron and cobalt on the B-site, the niques including electronic blocking methods,131-133
18
electronic structure is more complex but generally O2 tracer techniques based on postmortem SIMS
exhibits more semiconductor-like behavior with high analysis,134-142 relaxation methods involving the mea-
iron content (y ) 0.8) while being more metallic with surement of mass, conductivity, or other properties
high Co content (y ) 0.1).117,118 following a sudden change in PO2,135,143-148 or direct
Consistent with these differences in electronic measurement of permeation through a dense mem-
structure, LSC and LSF also exhibit very different brane.149-153 Although we do not review this litera-
defect thermodynamics in terms of the dependence ture in detail, some of the highlights relevant to the
of oxygen vacancy concentration (nonstoichiometry present discussion are included.
δ) as a function of Sr content (x), T, and PO2. As shown (1) Higher vacancy concentration leads to faster ion
in Figure 21a,119 LSF exhibits a plateau in oxygen transport and surface exchange. In general, the more
nonstoichiometry with PO2 where δ ) x/2, correspond- vacancies there are, the faster oxygen can be ab-
ing to an average iron oxidation state of Fe3+. At sorbed by, and move through, the lattice for a given
higher PO2 iron becomes oxidized, leading to lower δ, chemical potential driving force. In fact, as shown in
while at lower PO2 more vacancies are formed. As Figure 22, many workers have noticed a strong
shown by Mizusaki,119 this behavior can be explained correlation between the equilibrium surface isotope
quantitatively in terms of point-defect theory,120 exchange rate (k*) and the bulk tracer diffusion rate
assuming equilibria among O2 gas, oxygen vacancies, (D*) that spans 10 orders of magnitude in property
and localized electrons and holes. In contrast, Figure values across several classes of materials.141
21b121 shows that the oxygen nonstoichiometry in (2) Higher electronic conduction is correlated with
LSC exhibits no such plateau but rather a continuous faster oxygen exchange. As shown in Figure 22,
materials with high concentrations and/or more

mobile electronic carriers tend to have much better
surface catalytic properties (k*) than more purely
ionically conductive materials with similar rates of
bulk ion transport (D*). The existence of this uni-
versal correlation suggests that there may be a
fundamental limit to how active the surface of these
materials can be without some type of additional
enhancement.
(3) Available kinetic and transport data are limited
to linearized driving force. The measurement tech-
niques listed above (particularly tracer and relax-
ation techniques) only probe the rate of oxygen
diffusion and surface exchange at or near equilibrium
conditions. While transport data can normally be
extrapolated to higher driving force (through consid-
eration of the defect structure and appropriate ther-
modynamic factors), extrapolation of equilibrium
surface exchange data to high driving force is impos-
sible without a priori knowledge of the exchange
mechanism, which is still an open question. Thus,
the nonlinear rate expressions governing absorption Figure 23. Impedance response of a thin film of LSC
or evolution of oxygen at the mixed-conductor surface (x ) 0.4) on GDC at 800 °C and PO2 ) 10-2 atm as a
function of polarization. (Reprinted with permission from
remain largely unknown today. ref 124. Copyright 2002 Electrochemical Society Inc.)
4.3. Defining the Role of the BulksDense consistent with a porous LSM electrode. The authors
Thin-Film Mixed-Conducting Electrodes conclude that this damage results in a creation of
To better understand the role of bulk oxygen TPB contact area and thus changes the relative
transport in determining the characteristics of mixed- importance of bulk vs surface transport in allowing
conducting oxide electrodes, workers have focused electroactive oxygen to reach the electrode/electrolyte
considerable attention in the last 10-12 years on interface. In other words, by constraining transport
dense thin-film electrodes having little or no three- to the bulk artificially using a thin film, one can only
phase boundary contact area.27,124,154-166 Early at- prove that bulk transport occurs not that it is the
tempts struggled with sample preparation, it being dominant path for transport under all conditions.
difficult to make and verify the existence of un- In contrast, in a series of collaborative studies,
cracked and dense films. Perhaps the first success Kawada, Masuda, and co-workers reported excellent
in this direction was that of Mizusaki and co- success making thin-film electrodes of LSC (La0.6-
workers,157 who used a pulsed KrF excimer laser to Sr0.4CoO3-δ) on Ca-doped ceria (CDC) using laser
flash evaporate and deposit 1-2 µm thick films of ablation.124,158,161 These homogeneous and very-well
La0.5Sr0.5MnO3 on YSZ. Prior to polarization, these characterized films were studied by impedance over
electrodes exhibited a zero-bias impedance much a range of T, PO2, and thickness as well as isotope
higher than porous LSM electrodes and a perfor- exchange methods to help determine where the
mance that decreased with increasing PO2, which is resistance to oxygen reduction occurs. As shown in
contrary to the normal situation with porous elec- Figure 23, these films exhibit nearly perfect low-
trodes. They also noticed that the performance scaled frequency semicircular impedances having a resis-
inversely with electrode thickness. On the basis of tance and capacitance that can be analyzed quanti-
these results, they modeled the electrode as being tatively assuming that absorption/desorption of oxygen
limited by ambipolar diffusion of oxygen through the at the gas/film interface is rate limiting. The author’s
bulk of the film and showed that the majority of their isotope-exchange measurements on the same system
observations were consistent with this model, based confirm that the film contributes virtually no resis-
on the known defect structure of LSM. More recent tance to bulk oxygen transport and that there is no
results appear to confirm this conclusion.159,167 It was resistance for 18O2-/16O2- exchange across the LSC/
thus shown that even with a poor ionic conductor CDC interface. Subsequently published studies have
such as LSM, a bulk path exists for oxygen reduction consistently confirmed these findings for other per-
and a TPB interface is not strictly necessary for the ovskite phases having high ionic conductivi-
reaction to occur. ties.27,124,162-166 Both Kawada et al.161 and Yang et
However, as discussed more fully in section 5, it is al.165 compared the oxygen surface exchange coef-
not entirely clear how significant the bulk path is in ficient for the gas/mixed conductor interface obtained
the case of LSM, which is almost a pure electronic from isotope measurements to that predicted from
conductor under typical cathode conditions. In the the impedance and found quantitative agreement
Mizusaki study cited above, the authors went on to over a range of temperatures and PO2. These results
show that following anodic polarization (or strong suggest that over length scales of a few micrometers,
cathodic polarization), the LSM film becomes severely the bulk will provide little opposition to oxygen
damaged and cracked and reverts to behavior more transport and thus will be a dominant transport path.
If O2 exchange at the gas-exposed surface is rate

limiting (Figure 24a), the relatively fast transport of
oxygen ions in the film will cause the film to act as a
uniformly concentrated and electrically neutral res-
ervoir for oxygen vacancies and associated electrons/
holes. Conservation of vacancies in the film in this
case is therefore given by
∂[VO‚‚]
L ) NVO‚‚|y)0 - NVO‚‚|y)L (3)
∂t
where [VO‚‚] ) δ/Vm is the vacancy concentration, δ is

the oxygen nonstoichiometry in ABO3-δ, Vm is the
molar volume of the oxide, and NVO‚‚ is the vacancy
Figure 24. Models illustrating the source of “chemical flux in the y direction. In the absence of impedance
capacitance” for thin film mixed conducting electrodes. (a) at the solid-solid interface, a sinusoidal modulation
Oxygen reduction/oxidation is limited by absorption/de- of the current density (i) will cause a proportional
sorption at the gas-exposed surface. (b) Oxygen reduction/ faradaic modulation of vacancy flux across the solid-
oxidation is limited by ambipolar diffusion of O2- through solid interface at y ) 0
the mixed conducting film. The characteristic time constant
for these two physical situations is different (as shown) but jωt
involves the same chemical capacitance CL, as explained -i -i1e
in the text. NVO‚‚|y)0 ) ) (4)
2F 2F
Results are similar for films deposited on YSZ;
however, there appears to be a difference between where i1 is the current modulation amplitude and ω
films deposited on ceria vs YSZ in terms of interfacial is the frequency in s-1. At the gas/solid interface the
electrochemical resistance. As shown previously in absorption/desorption of oxygen results in a vacancy
Figure 6c, LSC films on YSZ often exhibit a second flux related to the displacement of the solid oxygen
high-frequency impedance associated with oxygen- chemical potential from equilibrium with the gas.
ion exchange across the electrode/electrolyte inter- Upon linearization, any rate expression for this
face.27,164-166 That this difference is associated with process can be written in terms of the displacement
the solid-solid interface has been confirmed by Mims of δ from equilibrium
and co-workers using isotope-exchange methods.164 solid gas
-µO2 )/2RT
As discussed in greater detail in sections 6.1-6.3, this NVO‚‚|y)L‚‚ ) lim
‚‚
{r0(1 - e(µO2 )} )
NV f0 NVOf0
interfacial resistance appears to result from a reac- O
tion between the electrode and electrolyte, sometimes r0f(δ - δ0) (5)
detected as a secondary phase at the interface.
As with platinum, these studies reillustrate that where r0 is the equilibrium oxygen exchange rate, δ0
oxygen reduction can be separated into chemical is the value of δ at equilibrium with the gas, and
steps (in this case absorption of oxygen into and f ) (- ∂µO2/∂δ)/RT is a thermodynamic factor express-
ambipolar transport through the bulk of the mixed ing the ease with which the material changes stoi-
conductor film) and electrochemical-kinetic steps chiometry for a given change in PO2 (f ≈ 1 means
(exchange of oxygen ions across the mixed-conductor/ stoichiometry changes easily, f . 1 means it is hard
electrolyte interface). Besides highlighting the im- to change stoichiometry). For a good electronic con-
portant role of these chemical steps in determining ductor in the limit of small vacancy concentration,
the overall resistance of mixed-conducting electrodes, configurational entropy tends to dominate the va-
another important consequence of this separation is cancy free energy such that f scales approximately
the concept of chemical capacitance (also known as inversely with vacancy concentration.
faradaic or pseudocapacitance) associated with the Since the solid-solid interface and bulk of the
bulk oxidation/reduction of the material. As discussed mixed conductor remain in chemical and electrical
below, this chemical capacitance is normally much equilibrium, the measured overpotential η is related
larger than surface or interfacial capacitances and directly to the spatially uniform oxidation state of
thus is a useful measure of how much of the bulk is the film through the Nernst equation: 4Fη )
involved, even when we step away from thin films to RTf(δ - δ0). Solving for δ(t) and recognizing that the
more complex microstructures. impedance Z ) η/i, one obtains
4.4. Chemical Capacitance Rsurf

Z)
Figure 24 illustrates what gives rise to the mea- 1 + jωRsurfCL
sured resistance and capacitance of a mixed-conduct-
RT 1
ing film (as determined by impedance) when the Rsurf )
reaction is limited by either (a) chemical reduction 4F2 2r0
of O2 to O2- at the gas/film interface or (b) transport
4F2 2L
of oxygen ions in the bulk of the film (where electronic CL ) (6)
conduction is facile). RT fVm
where Rsurf is the area-specific resistance associated torily explain such a large capacitance without
with oxygen reduction at the gas-exposed surface and invoking a bulk reaction pathway. As we will see
CL is the effective chemical (noninterfacial) capaci- below, workers have used this concept to understand
tance associated with changes in oxygen stoichiom- what is happening in more complex porous systems
etry in the film of thickness L. As shown by the where the relative roles of kinetics, transport, bulk,
Nyquist plot in Figure 24a, this impedance has a and surface are not nearly as clear.
semicircular shape with width Rsurf and characteristic
frequency 1/RsurfCL. A similar expression for the 4.5. Porous Mixed ConductorssA More Complex
chemical capacitance of a surface-limited film was Case
recently derived by Kawada et al. using a transmis-
sion line approach.124 Returning briefly to Figure 4, we see a summary
In the second case (limit of fast kinetics at the gas- of some of the physical mechanisms postulated in the
solid interface), the film becomes entirely bulk trans- literature to limit the rate of oxygen reduction in a
port limited, corresponding to the limit of Hebb- porous mixed-conducting electrode (some of which we
Wagner polarization. Since electronic conduction is have discussed previously in the context of porous
fast, this situation yields a Warburg impedance for Pt and dense mixed conducting electrodes). These
finite length diffusion49 include (a) kinetics of O2 incorporation into the bulk
mixed conductor, (b) kinetics of adsorption and/or
partial reduction of oxygen on the mixed-conductor
tanh(xjωRbulkCL) surface, (c) rate of bulk or (d) surface transport of
Z ) Rbulk O2- or On-, respectively, to the mixed-conductor/
xjωRbulkCL electrolyte interface, (e) electrochemical-kinetics of
charge transfer for O2- or (f) combinations of On- and
RT fVmL L e-, respectively, across the mixed-conductor/electro-
Rbulk ) ) (7)
4F2 2D̃ σi lyte interface, and (g) rates of one or more of these
mechanisms wherein the electrolyte itself acts as a
where Rbulk is the area-specific resistance associated mixed conductor due to doping by reaction with the
with ambipolar oxygen diffusion trough the bulk of electrode materials.
the film (assuming facile electron transport) and D̃ In the case of a LSM thin film, we noted above that
is the chemical diffusion coefficient for oxygen ions. introduction of TPB contact area (via damage of the
One can also express Rbulk in terms of the oxygen film) resulted in a fundamental change in the rate-
ionic conductivity of the film, σi, as shown in eq 7. determining step from bulk oxygen transport to some
Since oxygen stoichiometry changes occur over the other step (or combination of steps) likely involving
entire thickness of the film, this situation yields the the surface. The same fundamental question arises
same chemical capacitance (CL) as the surface-limited when considering a porous mixed conductor with high
case.168 ionic conductivity. For a thin film we saw that the
Thus, in the case of a thin film, when chemical mechanism is dominated by the bulk path, a-c-e,
steps (absorption and/or bulk diffusion) are rate with the surface exchange process (a) being rate
limiting, accumulation of electroactive intermediates determining (plus a small contribution from charge
in the bulk results in a large effective capacitance transfer (e) when YSZ is the electrolyte). If porosity,
proportional to the thickness of the film and the ease and therefore direct TPB contact area, is introduced,
by which it changes nonstoichiometry. As recently does the relative importance of these steps change
shown by Kawada et al.,124 this capacitance is very and/or does a surface path (b-d-f) begin to play a
large (0.1-1 F/cm2) even for relatively thin (1.5 µm) role as it does in the case of Pt?
mixed-conducting films. This value is significantly As discussed in section 4.2, strong interest in
larger than the pseudocapacitance of Pt (∼10-3 mixed-conducting perovskites as gas-separation mem-
F/cm2), which is dominated by adsorption and trans- branes stimulated a large volume of work in the late
port of oxygen on the Pt surface. If we compare this 1980s and early 1990s to better understand the
value to the apparent true interfacial polarization properties of mixed-conducting perovskites, including
capacitance of the Pt/YSZ interface (10-6-10-5 F/cm2), defect thermodynamics, ionic and electronic transport
we see an even larger difference. Thus, capacitance properties, and surface kinetics for absorption/
is a strong indicator (independent of resistance) as desorption of oxygen. Prompted by the availability
to what degree the interface, surface, and/or bulk are of these data, a number of workers in the mid-
playing in the reaction kinetics for a given material 1990s began a modeling effort to better understand
and set of conditions. the performance of porous mixed-conducting elec-
Some authors have expressed concerns that bulk trodes171,172 and membrane coatings173,174 based on
accumulation of reactive intermediates (and thus these properties. Although these models only consider
chemical capacitance) violates electroneutrality.169,170 a limited set of the physics shown in Figure 4, they
However, it should be recalled that reduction (or successfully confirmed the important role of the bulk
oxidation) of a material not only involves depletion and (where valid) provided design guidelines relating
(or accumulation) of oxygen ions in the bulk but performance to microstructure and bulk materials
neutral combinations of oxygen ions and compensat- properties. Spending some time to understand these
ing electrons/holes which together may accumulate models is therefore worth a few pages, both because
without violating electroneutrality.47 Indeed, no other they help understand the reaction in some limiting
mechanisms have yet been proposed which satisfac- cases as well as being a launching platform for
absorption and diffusion act as Kleitz’s “chemical

valve” (Figure 10), limiting overall charge flow (cur-
rent) at steady state. This model also considers the
effects of gas-phase diffusion and charge-transfer
resistance at the electrode/current collector contact.
Because the Adler model is time dependent, it
allows prediction of the impedance as well as the
corresponding gaseous and solid-state concentration
profiles within the electrode as a function of time.
Under zero-bias conditions, the model predicts that
the measured impedance can be expressed as a sum
of electrolyte resistance (Relectrolyte), electrochemical
kinetic impedances at the current collector and
electrolyte interfaces (Zinterfaces), and a “chemical”
impedance (Zchem) which is a convolution of contribu-
tions from chemical processes including oxygen ab-
Figure 25. Adler’s 1D macrohomogeneous model for the
impedance response of a porous mixed conducting elec- sorption, solid-state diffusion, and gas-phase diffusion
trode. Oxygen reduction is viewed as a homogeneous inside and outside the electrode.
conversion of electronic to ionic current within the porous In the limit of a semi-infinite (thick) porous elec-
electrode matrix, occurring primarily within a distance lδ trode with no gas-phase diffusion limitations, the
from the electrode/electrolyte interface (utilization region). chemical term Zchem reduces to an impedance reflect-
(Adapted with permission from ref 28. Copyright 1998 ing co-limitation by oxygen absorption and trans-
Elsevier.)
port.369
understanding more complex cases where the models
x
break down. Where and how these models break 1
Zchem ) Rchem (8)
down is interesting because it sheds light on what 1 + jω(RchemClδ)
other physical processes may be active.
where Rchem and Clδ are a characteristic resistance
4.6. One Asymptotic LimitsThe Bulk Reaction and capacitance, respectively, reflecting co-limitation
Path by surface kinetic and transport properties of the
One limit of behavior considered in the models cited mixed conductor
above is an entirely bulk path consisting of steps
x x
a-c-e in Figure 4. This asymptote corresponds to a RT fVm 4Rsurf
situation where bulk oxygen absorption and solid- Rchem ) )
4F2 D̃effar0 σi,effa
state diffusion is so facile that the bulk path domi-
nates the overall electrode performance even when
the surface path (b-d-f) is available due to existence 4F2 (1 - )lδ
C lδ ) (9)
of a TPB. Most of these models focus on steady-state RT fVm
behavior at moderate to high driving forces; however,
one exception is a model by Adler et al.171 which where Vm is the molar volume, D̃eff ) (1 - )D̃/τ is
examines the consequences of the bulk-path assump- the effective oxygen-ion chemical diffusion coefficient
tion for the impedance and chemical capacitance of in the solid (corrected for porosity and path tortuos-
mixed-conducting electrodes. Because capacitance is ity), and r0 and f (as defined previously in eq 5) are
such a strong measure of bulk involvement (see the linearized rate of oxygen absorption/desorption
above), the results of this model are of particular and a thermodynamic factor, respectively. As shown
interest to the present discussion. in eq 9, one can also express Rchem in terms of surface
As shown schematically in Figure 25, the Adler and bulk resistances defined previously in eqs 6 and
model takes as its testable hypothesis that mixed 7 for a thin film: Rsurf and σi,eff ) (1 - )σi/τ.
conductors with high ionic conductivity function by Equation 9 shows that the chemical capacitance in
the same underlying mechanism as a dense film, i.e., this case is similar to that derived previously in eq 6
path a-c-e in Figure 4. In addition, because ionic for a thin film (CL); however, in the co-limited
transport is relatively fast, the model also assumes situation the important length parameter is not L but
that variations in composition within the electrode rather a characteristic “utilization” length given by
will occur over distances larger than the size of
x x
individual microstructural features (particles) mak- D̃eff σi,effRsurf
ing up the porous electrode. Thus, as is often done lδ ) ) (10)
in porous electrode theory, transport and reaction in fVmar0 a
and between the solid and gaseous phases are treated
using average structural parameters: surface area The significance of this length parameter lδ can be
(a), porosity (), and tortuosity (τ). With this assump- understood by examining the predicted steady-state
tion, the overall electrode reaction, 2e - + 1/2O2 f O2-, vacancy concentration profile in the porous electrode
can be viewed as a 1D macrohomogeneous conversion as shown in Figure 26a. At steady state, the model
of electronic current to ionic current over the thick- predicts that the mixed conductor will be reduced by
ness of the electrode, where in this case oxygen an amount that decays exponentially with distance
features) suggests that the reaction is co-limited by

absorption and transport to the interface.
This result prompted a more complete experimen-
tal study by Adler of the impedance of porous
La1-xSrxCoO3-δ electrodes on SDC.28 Symmetric LSC/
SDC/LSC cells with 2.0 cm2 active area were fabri-
cated at three values of strontium content, x ) 0.2,
0.3, and 0.4, and characterized in terms of surface
area (a) and porosity () (tortuosity (τ) of the bulk
path was also estimated). The impedances of these
cells were then studied over a range T and PO2.
Although the resulting impedance data did not match
eq 8 as well as shown in Figure 26c, the data did
exhibit a single arc with a Gerischer shape and were
fit to obtain values of Rchem and tchem ) RchemClδ. On
the basis of these values (and without other adjust-
able parameters), the oxygen vacancy diffusion and
surface-exchange coefficients were back calculated
and compared to independent measurements based
on isotope exchange or membrane permeation (see
section 4.2). The results suggest the following.
(1) Preference for the bulk path. With notable
Figure 26. Predictions of the Adler model shown in Figure exceptions (see the following section), quantitative
25 assuming interfacial electrochemical kinetics are fast.
(a) Predicted steady-state profile of the oxygen vacancy comparison of the vacancy diffusion coefficient and
concentration (δ) in the mixed conductor as a function of surface-exchange coefficient back calculated from the
distance from the electrode/electrolyte interface. (b) Pre- resistance and capacitance are in reasonable agree-
dicted impedance. (c) Measured impedance of La0.6Ca0.4Fe0.8- ment with independent measurement both in value
Co0.2O3-δ electrodes on SDC at 700 °C in air, fit to the as well as dependencies on PO2 and temperature
model shown in b using nonlinear complex least squares. (activation energy). This result suggests that for
Data are from ref 171.
materials with high ionic conductivity, it is the bulk
path (a-c-e in Figure 4) rather than the surface
from the electrolyte interface with a characteristic path (b-d-f) that dominates at open circuit. This
length lδ describing the size of the active region. appears to be the only explanation consistent with
Equation 9 shows that the chemical capacitance will the large observed chemical capacitance. Not surpris-
be proportional to the size of this region, just as the ingly, agreement is best for materials with the
chemical capacitance of a dense film is proportional highest oxygen vacancy concentration. Application of
to its thickness (eq 6). The difference here is that the the model to LSM (which is a poor ionic conductor in
amount of material contributing to the capacitance air) grossly underpredicts performance, suggesting
is determined by the kinetic and diffusion parameters that porous LSM functions primarily by a surface-
rather than a geometric length scale. mediated mechanism at least near open circuit (see
Figure 26b shows the impedance predicted by eqs section 5).
8 and 9. As previously discussed, this function is (2) Co-limited kinetics with a significant utilization
known as the Gerischer impedance, derived earlier region. As with platinum, the model predicts that the
in section 3.4 for a situation involving co-limited chemical portion of the reaction will be co-limited by
adsorption and surface diffusion (in the context of Pt). molecular dissociation and transport. Values of lδ
As with the surface-mediated case, the present result calculated from the model for the analyzed conditions
corresponds to a co-limited reaction regime where vary from 0.4 to 20 µm depending on PO2, tempera-
both kinetics and transport determine the electrode ture, and electrode surface area, with typical values
characteristics (as reflected in the dependency of in the 3-5 µm range. This result indicates that a
Rchem and Clδ on both r0 and Deff). The essential significant portion of the electrode surface is active
difference between this and the Pt case is that here for oxygen reduction, which explains Takeda’s (and
the kinetics and diffusion parameters refer to a bulk- other’s) observation that the performance of LSC
mediated rather than surface-mediated process. electrodes on YSZ improves with thickness up to a
As shown in Figure 26c and as discussed more fully limit of a few micrometers.101,175 At around the same
elsewhere,171 initial comparisons of this model to time as the Adler model, Zhou and co-workers mod-
impedance data for La0.6Sr0.4Co0.2Fe0.8O3-δ (LSCF) on eled the effect of porous perfusion layers on mixed-
gadolinia-doped ceria were favorable. A fit of the conducting membranes, drawing similar conclusions
impedance to eq 8 yielded values of Rchem and Clδ that regarding the co-limited nature of the reaction and
were reasonably consistent with those calculated estimates of the utilization length.176 These models
based on the known bulk properties of LSCF. The further predict that for small surface areas (a) lδ will
large chemical capacitance observed (and calculated) exceed the thickness of the porous layer/electrode
in this case corresponds to a significant reduction/ such that the entire layer becomes active with
oxidation of the bulk, while the Gerischer shape of surface-absorption-limited behavior (semicircular im-
the impedance (and absence of other impedance pedance) rather than co-limited. Adler confirmed this
additional (presumably active) LSC does not signifi-

cantly enhance performance. Also consistent with a
large utilization region is a study by Koyama et al.
of porous Sm0.5Sr0.5CoO3-δ (SSC) on SDC as a func-
tion of electrode thickness;184 they found that the
electrode performance improves with thickness up to
about 10-15 µm, after which performance saturates.
During the time period that the results described
above were published (mid-1990s), workers studying
porous mixed conducting electrodes fell into multiple
schools of thought regarding the mechanism (not
unlike those discussed previously for Pt/YSZ). A
number of papers appeared which expressed conflict-
ing (or dissenting) views vis-a-vis the role of the bulk
and interface in governing the mechanism and the
extent of the utilization region. One of these was
work by Gödickemeier et al., who used current-
interruption techniques to study the steady-state
current-overpotential relationships for LSM and
LSC on samaria- and gadolinia-doped ceria.14,17 They
found that their results fit well to a Butler-Volmer
Figure 27. Measured overpotential of a porous LSC rate expression and thus concluded the electrodes are
electrode on SDC with and without a dense LSC layer in limited by interfacial electrochemical kinetics rather
between. (Reprinted with permission from ref 179. Copy- being than transport limited. In this case, by “trans-
right 1998 Electrochemical Society, Inc.)
port limited” the authors meant simply that a limit-
prediction by oversintering the electrode and showing ing current was not observed at high overpotentials.
that the electrode exhibits a surface-limited imped- However, as we saw in section 3.3 for platinum on
ance similar to that predicted by eq 6.28 YSZ, the fact that i-η data fits a Butler-Volmer
(3) Absence of electrochemical-kinetic resistances. expression does not necessarily indicate that the
For LSC on SDC at 550 °C and above, additional electrode is limited by interfacial electrochemical
high-frequency impedance features indicating the kinetics. Supporting this point is a series of papers
presence of an interfacial resistance/capacitance were published by Svensson et al.,172,185,186 who modeled
not observed. This result suggests that the majority the current-overpotential (i-η) characteristics of
of the overpotential is associated with absorption and porous mixed-conducting electrodes. As shown in
bulk transport of oxygen, while ion exchange at the Figure 28a, these models take a similar mechanistic
LSC/SDC interface remains in electrochemical equi- approach as the Adler model but consider additional
librium. As we will see in section 6, this conclusion physics (surface adsorption and transport) and forego
is not universally true of all materials; additional time dependence (required to predict impedance) in
impedance arcs have been observed for perovskites order to solve for the full nonlinear i-η characteris-
on YSZ and with ceria at lower temperatures or with tics at steady state.
certain electrolyte dopants. Nonetheless, this result
did strongly challenge prior thinking, which had One significant prediction of the Svensson models
assumed interfacial electrochemical kinetics would is that regardless of whether the reaction path is
be limiting if transport is facile.101 surface or bulk dominated, the i-η characteristics
Subsequent studies of porous mixed-conducting appear Tafelian, even if interfacial electrochemical
electrodes under a variety of conditions have since kinetic steps are equilibrated. As an example, Figure
reinforced the view that the bulk pathway plays an 28b shows the predicted i-V characteristics of a
important (if not dominant) role in the electrode mixed conducting electrode assuming a bulk path as
kinetics for these materials.19,177-184 Of particular a function of the oxygen vacancy diffusion coefficient.
note is a study (published the same year as the Adler As evidenced by the linear dependence of ln(i) on V
studies) by Endo and co-workers who used pulsed at high currents, the model obeys Tafel kinetics, yet
laser deposition to cast a dense thin (∼1 µm) film of the Tafel parameters (“exchange current density”)
LSC (40% Sr) onto SDC prior to screen printing and depend on a transport parameter. In other words,
firing a porous LSC electrode of the same composi- just as we saw for platinum on YSZ (section 3.3), the
tion179 (Figure 27a). As shown in Figure 27b, they mere fact that i-η data fits a Butler-Volmer expres-
found that the overpotential of the porous electrode sion does not prove that the electrode is limited by
is independent of the presence of the film. This result interfacial electrochemical kinetics. The Svensson
appears to confirm the dominance of the bulk path, models show that chemical steps of ad(b)sorption and
since elimination of any possible surface path (in- transport can dominate the observed kinetics well
cluding TPB contact area) does nothing to alter the below limiting current, and thus the absence of a
electrode performance. This result also appears to measured limiting current does not indicate an
confirm that the active region of the porous LSC layer absence of transport limitations. On the contrary,
is large compared to the superficial electrode area, diffusion in this case is predicted to produce a variety
since replacement of the gas-exposed electrolyte with of finite, Tafel-like, characteristics.
Figure 29. Comparison of the oxygen vacancy diffusion

coefficient (Dv) in LSC (x ) 0.2) determined from perme-
ation measurements vs that extracted from impedance
measurements using the model in Figure 26. Data are from
refs 190 and 28. (Adapted with permission from ref 28.
of the data and their inconsistency with measure-

ments under certain conditions is actually useful
information for understanding how other factors and
Figure 28. Svensson’s macrohomogeneous model for the
i-V characteristics of a porous mixed-conducting electrode. mechanisms (besides the bulk path) come into play.
(a) The reduction mechanism assuming that both surface
and bulk diffusion are active and that direct exchange of 4.7. Limitations of Bulk, 1D Transport Models for
oxygen vacancies between the mixed conductor and the Porous Mixed Conductors
electrolyte may occur. (b) Tafel plot of the predicted steady-
state i-V characteristics as a function of the bulk oxygen While subsequent work19,177-184 has continued to
vacancy diffusion coefficient. (Reprinted with permission reinforce the importance of the bulk for porous mixed-
from ref 186. Copyright 1998 Electrochemical Society, Inc.) conducting electrodes, authors generally debate the
degree to which the bulk vs surface paths dominate
Another set of papers countering the conclusions and under what conditions. In the asymptotic limit
of the Adler studies (as well as implicitly the Svens- of high ionic conductivity, the 1D models discussed
son studies) were published by Liu and co-workers, above appear to explain i-V characteristics and
who questioned the validity of the Adler model on impedance data well under conditions of facile bulk
several levels.169,170,187,188 As discussed elsewhere,189 transport and a reversible interface. However, what
many of these questions appear to have arisen happens when the ionic conductivity is not so high
primarily from a misunderstanding about the as- and/or the utilization length becomes competitive
sumptions and terminology of the model (so we will with microstructural features? Also, what happens
not belabor the points raised again here). However, when the interface is not reversible?
Liu also raised the more general concern that the One clue to this question is provided by where and
model does not consider the “direct” reduction of O2 how the Adler model breaks down in explaining the
at the TPB (traditionally considered the route by impedance characteristics of LSC on SDC. For ex-
which the cathode reaction occurs). While this is ample, Figure 29 compares the apparent vacancy
certainly true, perhaps the more relevant question diffusion coefficient of La0.8Sr0.2CoO3-δ (extracted
is what we mean by the “three-phase boundary”? from the chemical resistance and capacitance28) to the
Certainly from the phenomenological standpoint measured value for a bulk membrane (determined
(Figure 3), one must have thee phases involved in from permeation measurements190,191) as a function
the reaction for oxygen reduction to occur. However, of temperature in air. At the highest temperatures,
when one begins to consider the specific chemical, agreement is reasonable in both value and activation
electrochemical, and transport steps involved, one energy. However, at lower temperatures there is a
must abandon the idea of an ideal 1D geometric systematic deviation from activated behavior, with
surface and consider the specific phases, surfaces, the apparent diffusion coefficient being perhaps 10
and interfaces involved in these steps. A separate times larger than predicted by extrapolation of the
consideration of the “direct” TPB reaction is not permeation data. This deviation also corresponds to
necessary since it is exactly this reaction that the a change in the shape of the impedance to being less
various models discussed in this review attempt to “Gerischer-like”.28 In addition, for La0.8Sr0.2CoO3-δ,
embody. While a mechanistic approach almost cer- the Adler model predicts a utilization length of lδ )
tainly involves oversimplification of the physics, it 0.3-0.6 µm, which is competitive with the micro-
also allows workers to pose specific hypotheses about structural features of the electrode and thus a viola-
the mechanism, predict the consequences of those tion of the formal assumptions of the model.
hypotheses, and test these predictions against mea- There are several possible explanations for this
surement. As we will see in the following section, the behavior, all of which speak to various deficiencies
failure of existing models to predict certain features of current models. First, even if the bulk path
purely transport-limited behavior is not possible with

a surface mechanismsthis is a unique feature of bulk
transport in two or three dimensions.
What Fleig’s results suggest is that even without
invoking other mechanisms besides the bulk path,
one would have difficulty fitting the characteristics
of a porous mixed-conducting electrode accurately
using a 1D model (such as the Adler or Svensson
models) when lδ is less than a few micrometers. The
degree to which these effects are responsible for any
lack of agreement with data are unknown. To date,
a quantitative modeling of the impedance of LSC or
LSF electrodes considering 3D effects has not yet
been conducted. This is an ambitious task for several
reasons. First, it is not clear that a geometric
idealization of the microstructure (such as that used
by Fleig) would be sufficientsat the length scale of
an individual particle, the details of the actual
microstructure (particle size distribution, particle
connectivity, contact angle, etc.) might become im-
portant. Second, Fleig’s model predicts that under
increasingly 3D transport-limited conditions, the
Figure 30. Finite-element calculation of the oxygen
vacancy concentration profiles in a porous mixed conduct-
ionic oxygen flux will not be uniform across the
ing electrode, assuming a cylinder-block microstructure electrode/electrolyte interface but rather increasingly
having a single characteristic dimension lg. (a) Oxygen constricted through an area close to the TPB. This
reduction is entirely bulk-transport limited. (b) Oxygen is constriction raises the question of whether electro-
co-limited by absorption and bulk transport as in Figures chemical kinetics at the interface might become
25 and 26. (c) Comparison of the predicted resistance to increasingly important due to reduced area available
that of a macrohomogeneous model (Figure 26) as a for charge transfer. Such considerations add consid-
function of absorption kinetics, assuming lg ) 1.6 µm.
(Adapted with permission from ref 194. Copyright 2003 erable complexity to the problem, in terms of both
Annual Reviews.) nonlinearity and inhomogeneity at this interface as
well as the effect of interfacial resistance on the
dominates, we expect that as the utilization length concentration profiles in the bulk.
becomes small, the assumption that reaction/trans- However, even if one took 3D effects into consid-
port can be treated using a 1D macrohomogeneous eration, it is unlikely that one could explain all the
representation of the microstructure will break down. discrepancies between the Adler model and the
The most complete studies of this effect has been measured impedance. For example, Figure 29 shows
published by Fleig,192-194 who used finite-element that the assumption of 1D bulk transport appears to
analysis (FEA) to compare the predictions of a 1D underpredict the measured performance. In contrast,
macrohomogeneous model to that of a full 3D micro- Fleig’s modeling studies suggest that ignoring 3D
structural model. As shown in Figure 30, the 3D transport effects will generally overpredict perfor-
model considers bulk absorption and diffusion of mance (Figure 30). This forces us to consider a second
oxygen in a hypothetical “cylinder-block” micro- reason the Adler model may break down: As bulk
structure having characteristic length lg ) 1.6 µm. transport becomes less facile, a parallel mechanism
Fleig compared the linearized electrode resistivity involving the surface path may become increasingly
predicted by this 3D model to that of a 1D macro- important. Although this widely held belief has not
homogeneous representation of the same system as yet been proven definitely, it is nonetheless circum-
a function of kq at fixed Dq, where kq and Dq are stantially consistent with a variety of observations.
linearized constants for surface absorption and bulk (1) Inconsistency of performance with a bulk path
diffusion.195-197 This comparison shows that as kq/Dq at low vacancy concentration. A quantitative com-
increases beyond a value of about 0.1-1 µm-1, the parison between predictions of the Adler model and
macrohomogeneous assumption breaks down. Indeed, impedance data for LSC shows the poorest agreement
based on the analysis presented in section 4.6, we (underprediction of performance) at low tempera-
expect this breakdown to occur when the utilization tures, high PO2, and/or low Sr content.28 These are
length lδ ≈ xlgDq/kq (approximate translation of the conditions under which the bulk vacancy concen-
eq 10 for Fleig’s variables kq and Dq) becomes on the tration (and thus also the ionic conductivity and
same order or smaller than the particle dimensions surface exchange rate of oxygen with the bulk) are
lg. This occurs when kq/Dq ≈ lg-1, which in this case the lowest. These are exactly the conditions under
is 0.6 µm-1, in good agreement with Fleig’s results. which we would expect a parallel surface path (if it
Furthermore, Fleig shows that when surface absorp- existed) to manifest itself, raising performance above
tion becomes fast, the electrode becomes entirely that predicted for the bulk path alone. Indeed, as
transport limited (with a plateau in performance) vs discussed more fully in section 5, the Adler model
the macrohomogeneous model (which predicts con- breaks down completely for LSM (a poor ionic con-
tinued improvement as kinetics get faster). Such ductor at open-circuit conditions), predicting an
electrode impedance many orders of magnitude higher two fundamental reasons. First, a common feature
than observed experimentally.171 of these models is that they are limited to nonlinear
(2) Inconsistencies in the chemical capacitance. As steady-state behavior and therefore cannot be used
the bulk vacancy concentration is reduced, and hence to analyze impedance or other measurements with
the predicted chemical capacitance of the bulk (per time resolution for various physical processes. As
unit volume) becomes smaller, the portion of chemical mentioned previously, it is hard to tell what is going
capacitance associated with surface species may on from Tafel parameters since so many things mimic
become increasingly significant. Kuznecov et al. Tafel kinetics. Second and perhaps more fundamen-
showed that a surface-dominated mechanism should tally, there is simply a dearth of independent data
exhibit a similar Gersischer shape as a bulk- for model parameters. Although the mobility of
dominated mechanism,198 only with a smaller chemi- oxygen on the surface of a perovskite is generally
cal capacitance. One possible piece of evidence that thought to be “fast”, its rate relative to bulk transport
this occurs is the observation that the apparent (as a function of defect concentration) has never been
utilization length lδ (as extracted from impedance quantitatively determined. Coffey et al. summarize
data assuming the Adler model28) generally decreases the situation nicely in the concluding paragraph of
with conditions promoting lower vacancy concentra- their paper.199
tion (lower Sr content, higher PO2). This observation “Unfortunately many of these parameters are at
is somewhat inconsistent with eq 10, which for most best rough approximations if not outright guesses.
mixed conductors predicts that the utilization length Consequently our results are not at this time useful
will stay approximately constant with changes in in interpreting available electrochemical data; how-
stoichiometry (as shown previously in Figure 22, the ever they are useful in defining what additional
bulk oxygen vacancy concentration tends to influence experimental measurements need to be made.”
oxygen exchange and bulk diffusion rates of perov- Indeed, in the spirit of this latter comment, the
skites similarly,141 leading to a constant ratio of purpose of this review is to consolidate our under-
D̃eff/fr0 in eq 10). One resolution to this apparent standing of SOFC cathodes so that it becomes easier
inconsistency might be a parallel surface path, which to identify and propose new avenues of research.
could alternatively explain the measured decrease in
chemical capacitance as a confinement of stoichiom- 4.8. Summary: Importance of the Bulk for
etry variations to the surface rather than a shrinkage Mixed-Conducting SOFC Cathodes
of the bulk utilization region. In this section we saw that the active region of a
(3) Nonlinear electrode kinetics. Another possible SOFC cathode can be significantly enhanced by
indicator of a parallel surface path is nonlinearity of incorporating a mixed conductor (a material which
the current-overpotential characteristics. In par- conducts both ions and electrons). While electrodes
ticular, one conclusion of the model by Zhou and co- of this type have proven challenging to implement
workers mentioned previously176 was that the per- in a SOFC operating environment (see section 6),
meation resistance of a porous membrane overlayer they nonetheless have taught us a lot about what
is fairly constant with driving force (i.e., linear force- factors can limit electrode performance and opened
flux relationship) for mixed conductors having high the realm of possibilities for future materials devel-
vacancy concentration. This result appears to be opment. To summarize, some of the salient points of
consistent with recent measurements of Horita et our current understanding are listed.
al.,181 who conducted a study of the impedance of (1) Role of the bulk transport path. In section 3 we
porous LSC electrodes on La0.8Sr0.2Ga0.8Mg0.2O2.8 saw that for Pt the dissociation of oxygen and
(LSGM) electrolyte as a function of cathode overpo- transport of reactive intermediates to the electrode/
tential. Among other things, these results show that electrolyte interface is confined to the material
the electrode kinetics are fairly linear with overpo- surface. With mixed conductors, it is possible for
tential for materials with high vacancy concentration oxygen reduced at the surface to be transported
but become increasingly nonlinear for materials and through the bulk of the material to the electrode/
conditions promoting the lowest vacancy concentra- electrolyte interface. If bulk transport is facile, this
tions. For example, La0.8Sr0.2CoO3-δ is more nonlinear path may dominate, extending both the accessible
than La0.6Sr0.4CoO3-δ, and both materials are increas- surface for O2 reduction as well as broadening the
ingly nonlinear at lower temperatures. Although the active charge-transfer area from the TPB to include
linearity of the bulk path itself has not been well the entire solid-solid contact area.
investigated, an increase in nonlinearity with de- (2) Chemical capacitance. When the mechanism
creasing vacancy concentration might indicate a shift involves significant involvement of the bulk, ac-
toward a surface-mediated reaction mechanism in- cumulation of reactive intermediates not only in-
volving more nonlinear behavior (e.g., increased volves surface species but oxidation and reduction of
charge-transfer resistance at the interface due to the bulk. This can be detected as an anomalously
confinement of the flux near the TPB). high effective capacitance, often referred to as a
Since the mid-1990’s, several models which con- chemical (or pseudo) capacitance. This capacitance
sider both a bulk and surface path have been can be as large as 0.1-1 F/cm2 and thus easily
proposed, including the Svensson models as well as detected by current-interruption or impedance tech-
a more recent model by Coffey and co-workers.199 niques. Thus, capacitance is a strong indicator (in-
However, these models have not been definitive in dependent of resistance) as to what degree the
addressing the bulk vs surface question for at least interface, surface, and/or bulk are playing in the
reaction kinetics for a given material and set of ever, the performances of “real” SOFC cathodes of
conditions. technological interest generally fall outside these
(3) Co-limited kinetics. As with platinum, porous limits for two fundamental reasons. First, SOFC
mixed-conducting electrodes are co-limited by mo- cathodes must satisfy a number of other constraints
lecular dissociation and transport. For mixed conduc- besides performance, including mechanical and chemi-
tors with high rates of bulk ionic transport, values cal stability, cost, manufacturability, etc.sit is there-
of lδ vary from 0.4 to 20 µm depending on PO2, fore nearly impossible to optimize electrode materials
temperature, and electrode surface area with typical according to only one or two physical properties.
values in the 3-5 µm range. This result indicates Second, as people optimize any system, they will tend
that a significant portion of the electrode surface is to mitigate (knowingly or unknowingly) the most
active for oxygen reduction, not just material in the rate-limiting factor until other factors naturally come
immediate vicinity of the TPB. into play. In other words, if a material’s behavior is
(4) Sensitivity of interfacial resistance to various asymptotic, it is not likely to be optimized or satisfy
factors. For perovskite mixed conductors on some more than one design constraint.
ceria-based electrolytes, workers have reported virtu- One such “more complex” case is La1-xSrxMnO3(δ
ally zero interfacial resistance such that the electrode (LSM), which starting in the mid-1970s became (and
overpotential is dominated entirely by dissociation has remained) one of the most heavily pursued
of O2 and transport of intermediates to the electrode/ electrode materials for SOFC cathodes. As mentioned
electrolyte interface. As we will see in section 6, this earlier, LSM was originally investigated as an SOFC
conclusion is not universally true of all materials; cathode (along with a variety of other transition-
additional impedance arcs have been observed for metal perovskites) due to its good electrical conduc-
perovskites on YSZ and with ceria at lower temper- tivity and relatively low cost.4 However, LSM quickly
atures or with certain electrolyte dopants. distinguished itself for a variety of reasons. First,
(5) Relative role of surface vs bulk path not yet with the right choice of Sr content (x), a nearly exact
known quantitatively. While it has been clearly thermal expansion match between LSM and YSZ can
established that the bulk transport path plays a role be achieved. This allowed workers to fabricate and
(and may dominate) in the mechanism of mixed- explore a wide variety of electrode microstructures
conducting electrodes, our quantitative understand- and cell geometries with reduced thermal stress
ing is currently limited to asymptotic cases (such as generated by thermal cycling. Likewise, Mn is gener-
thin films or materials with unusually high bulk ally less reducible than other transition metals (Co,
transport rates). We currently lack general tech- Fe) in a perovskite matrix, and thus LSM exhibits
niques to measure or predict the relative role of the little or no chemical expansivity,200,201 another source
surface vs bulk paths in electrodes of arbitrary of thermal-mechanical stress potentially threatening
composition and processing. Development of such the integrity of the electrode microstructure. Another
techniques will be an important step in understand- advantage of LSM is that it is generally more ther-
ing more complex materials and microstructures and modynamically stable than mixed conductors con-
in making intelligent materials design choices. taining cobalt or iron.202 It is nonetheless reasonably
catalytic for O2 dissociation, unlike materials having
5. Lanthanum Strontium Manganese Oxide similar or higher stability, such as La1-xSrxCrO3 for
(LSM): Where Surface and Bulk Converge example.203
As with other SOFC cathode materials (including
In sections 3 and 4 we examined two asymptotic Pt), early kinetic studies of LSM in the mid-1980s
cases for the mechanism of oxygen reduction. With gave rise to multiple schools of thought regarding the
porous Pt, oxygen reduction appears to occur by a reaction mechanism. At that time, LSM was often
surface-mediated mechanism, where dissociative ad- studied alongside Pt or other perovskites and thus
sorption and diffusion of oxygen on the gas-exposed lumped together with these materials in the same
Pt surface play a significant (if not dominant) role debates regarding the rate-limiting step, the role of
in determining the overall electrode kinetics. Like- bulk transport, etc. However, beginning in the early
wise, for porous mixed conductors with high ionic 1990s it became clear that LSM is somewhat differ-
conductivity, molecular dissociation and transport ent than both Pt or other (more reducible) perovskites
are equally important; however, with these materials in terms of the reaction mechanism, falling somewhat
the mechanism appears to proceed by a primarily in a gray area among the asymptotic limits described
bulk-mediated path. In both cases it was shown that previously. Although the last 10-15 years of work
due to porosity, oxygen dissociation and transport have added significantly to our qualitative under-
tend to co-limit the reaction (rather than contribute standing of LSM, this material has so far continued
independently). Interfacial electrochemical kinetics to elude quantitative descriptions of its performance
can also play a significant role, depending on the in terms of underlying mechanisms. For this reason
exact materials and conditions. Finally, we saw that and because LSM is so important technologically, we
these two asymptotic limits of behavior begin to devote this section to it.
merge for materials having some bulk ionic conduc-
tivity but not as facile as the best mixed conductors. 5.1. Three-Phase Boundary: Not the Whole
Consideration of these asymptotic limits is useful Picture
because it potentially helps us to identify, discuss,
and study the various physical processes underlying In the case of Pt, mechanistic considerations sug-
electrode kinetics, even outside those limits. How- gest that the active region of the electrode is confined
to a region close to the three-phase boundary (TPB), general correlation exists and thus prior studies have
and thus performance has often been reported to been interpreted too narrowly. On the other hand, it
scale with the available TPB area. Thus, in address- is not clear that the comparisons in performance
ing the question of how localized the O2 reduction is drawn by the authors in this case are fully meaning-
in LSM, one approach workers have used is to ful. The authors point out that the electrode kinetics
examine the relationship of performance to micro- are (1) highly nonlinear and (2) very dependent on
structure. For example, Mizusaki and co-workers the polarization history. Thus, it is difficult to say
carefully characterized and measured the impedance whether two cells operating at the same current
of La1-xCaxMnO3 (LCM) electrodes on YSZ (a very density (but different overpotentials) are really at the
similar material system to LSM/YSZ) as a function same operating point vis-à-vis comparison of TPB
of morphology, controlled using various preparation contact area. Rather, a more significant observation
methods and firing conditions.204 They used SEM to we might draw from such a study is how difficult it
estimate the contact area of LCM and YSZ as well is to meaningfully correlate the complex (often non-
as the geometric length of the TPB. While their stationary) electrochemical characteristics of LSM to
results showed that the electrode capacitance scales the highly complex microstructure of a porous elec-
with the electrode/electrolyte contact area (as one trode, at least with today’s tools and techniques. As
would expect for interfacial polarization), the resis- we will see in the sections below, workers have had
tance and overpotential were found to scale inversely more success over the last 10-15 years examining
with the length of the TPB, at least at low polariza- electrodes with defined and/or microfabricated ge-
tion. A number of other studies of single-phase and ometries, where TPB area and material composition
composite electrodes have since reinforced the view can be more carefully controlled and analyzed.
that the electrode kinetics scale with TPB contact
area.205-207 5.2. Complex Stationary Electrochemical
However, there are several equally valid interpre- Characteristics and Properties
tations to this commonly observed result. One pos-
sibility is that the reaction is co-limited by adsorption Like all cathodes, early electrochemical kinetic
and surface diffusion, where the utilization length studies of LSM focused heavily on steady-state d.c.
is small compared to the average particle size. An characteristics, attempting to extract mechanistic
alternative (but not mutually exclusive) possibility information from the T and PO2 dependence of linear
is that the reaction is limited by electrochemical and Tafel parameters.13,203 As recently as 1997, some
kinetics at the TPB itself. As with Pt, both situations workers have continued to support a view that LSM
or a combination thereof would result in the resis- is limited entirely by electrochemical kinetics at the
tance scaling with the reciprocal of the TPB length. LSM/electrolyte interface based on this type of analy-
A third possibility is that reaction is limited by sis.14,17 However, as we have seen for other materials
mechanisms acting farther from the TPB but which (including Pt), the fact that an electrode obeys
scale with electrode geometric factors (such as the Butler-Volmer kinetics means little in terms of
electrode surface area) that are strongly correlated identifying rate-limiting phenomena or in determin-
to the length of the TPB (through the process of ing how close the reaction occurs to the TPB. To
sintering, for example). Without the ability to sepa- understand LSM at a nonempirical level, we must
rately control the TPB length while leaving other examine other techniques and results.
geometric parameters constant, it is difficult to As we saw with Pt and other perovskites, one such
distinguish these various possibilities. Also, the approach is electrochemical impedance spectroscopy
conditions under which the electrode performance (EIS), which attempts to separate various mechanis-
scales with the TPB may not be universalsthe tic steps via time scale. Although quantitative analy-
electrode may have multiple regimes of operation sis of impedance data in a complex material system
depending on overpotential and other factors. like porous LSM involves many uncertainties (see
One study that tried to address this question was sections 3.1 and 6.7 as well as ref 49), it is usually
conducted by van Heuveln and co-workers, who reliable in terms of separating rough time scales on
attempted to separate the effects of electrode surface which various physical processes occur. An example
area from TPB contact area.208,209 Variations in of this approach for understanding LSM is a study
surface area were obtained by using different sinter- by Østergård and Mogensen,210 which examined the
ing temperatures, while changes in TPB contact area impedance of single-phase porous LSM on YSZ as a
were accomplished through the use of two powders function of T, PO2, and polarization. The observed
having different morphologies. Surface area and TPB impedance was found to contain at least three
contact area were determined by image analysis of distinct features. The highest frequency feature was
SEM cross sections; for the surface area the electrode attributed to interfacial electrochemical kinetic pro-
bulk was examined, while for the TPB contact area cesses at the LSM/YSZ interface. The two lower
the electrodes were etched off and the underlying frequency features were assigned to dissociation and
“stain” on YSZ was analyzed. Electrode performance transport, respectively, of oxygen species on the LSM
was measured in terms of the overpotential at a fixed surface.
current density of 100 mA/cm2. While such one-to-one assignments of impedance
Within the statistical certainty of the data, no features to specific reaction steps are appealing, it
general correlation was found between overpotential is not clear how definitive they are without further
and TPB length. This could mean that no such analysis. In particular, based on what we know about
Figure 32. Schematic illustrating the possible location

and distribution of bulk oxygen flow in LSM under high
polarization conditions. (Reprinted with permission from
Figure 31. Measurements of polarization and impedance
ref 211. Copyright 1995 Elsevier.)
of LSM point-contact electrodes on YSZ. (a) Schematic
showing the geometry of the electrode. (b) Steady-state
cathodic polarization at 960 °C and PO2 ) 10-3 atm. (c) impedance for mixed conductors having high ionic
Impedance measured under polarization at the conditions conductivity, as discussed in section 4.
marked in b. (Adapted with permission from ref 211.
Copyright 1995 Elsevier.) To explain these results, the workers at Grenoble
theorized the existence of two different operating
Pt and perovskite mixed conductors, we might expect regimes. As shown in Figure 32, the authors proposed
oxygen dissociation and transport on LSM to be co- that at high overpotential the LSM becomes substan-
limiting, resulting in a single “Gerischer-like” imped- tially reduced, creating a large concentration of
ance feature (as shown in Figures 14 and 15) rather oxygen ion vacancies near the interface. This non-
than two independent low-frequency features (as equilibrium population of vacancies thereby creates
proposed by Østergård and Mogensen). Also, these a facile transport path through the bulk of the
particular authors did not attempt to quantitatively material, leading to a large chemical capacitance (>1
analyze the chemical capacitance in terms of a F/cm2 when compared on an area-normalized basis).
utilization length or bulk vs surface contributions. The presence of an equally sized high-frequency
Without this it is hard to conclude what these various feature likewise suggests a significant contribution
features actually correspond to except perhaps broadly to the impedance from oxygen ionic exchange at the
in terms of “chemical” vs “interfacial” time scales. solid-solid interface, associated with a much smaller
capacitance related to interfacial polarization. The
An important contribution in this regard was a existence of such a “bulk” operating regime at high
series of studies published by the Laboratoire d’Ionique overpotential appears to be consistent with in-situ
et d’Electrochemie du Solide (Grenoble), which looked
X-ray photoelectron spectroscopy (XPS) measure-
in detail at the i-V characteristics and impedance
ments of porous LSM, which show significant reduc-
of LSM point-contact electrodes on YSZ.211,212 As
tion of Mn far beyond the electrode/electrolyte inter-
shown in Figure 31a, these measurements involved
face upon polarization.213 More recently, as shown in
making a single-point contact between a pin-shaped
Figure 33, Kuznecov and co-workers showed using
LSM sample and YSZ with known contact area. The
field-emission SEM that sustained high cathodic
i-V characteristics (Figure 31b) show a distinct
polarization results in the apparent formation of
change in the characteristics of the electrode at
Kirkendall porosity in the bulk of LSM within one
cathodic overpotentials greater than a particular
particle diameter of the solid-solid interfacesa
critical value, where a relatively discontinuous jump
in current is observed. They found this transition strong indicator of oxygen chemical potential gradi-
occurs anywhere from -150 to -500 mV vs air and ents in the bulk.214
is only present in the cathodic direction. The imped- In contrast, at low overpotential the workers at
ance measured at points well below, near, and well Grenoble theorized that LSM is primarily an elec-
above this transition are shown in Figure 31c (A,C,G). tronic conductor and thus the bulk path is closed off.
Well below the transition (A) the impedance is high The observation of finite impedance near open circuit
with a single arc having resistance ∼4000 Ω and (despite the fact that LSM is a poor bulk ion conduc-
capacitance ∼10-6 F. Near the transition (C) the tor under these conditions) suggests the existence of
impedance becomes small enough to notice the pres- a parallel “surface” path that allows transport even
ence of both a low- and high-frequency feature, but when the bulk is not available. Since this surface
these features are too obscured by inductive effects path involves a much smaller inventory of electro-
to resolve completely. Well above the transition (G) active species, it is associated with a much smaller
both features become well resolved: one at high chemical capacitance than the bulk path and thus a
frequency having resistance ∼200 Ω and capacitance much higher frequency (Figure 31c (A)). As the
10-7 F and the other at low frequency with resistance overpotential is reduced and the mechanism shifts
∼300 Ω and capacitance 10-1 F. This low-frequency from bulk to surface, chemical and electrochemical
feature, seen only at high overpotential, is highly time scales begin to merge, making them difficult to
reminiscent of the chemical contributions to the distinguish as separate features in the impedance.
Figure 33. High-resolution field emission SEM of a porous

LSM/YSZ interface following polarization for 3 h at -0.8
Figure 35. Total conductivity, σ, and oxygen stoichiom-
V at 950 °C in air. The porosity evident at the TPB is not
etry, 3-δ, at 1000 °C of La0.9Sr0.1MnO3-δ, from measure-
seen in images taken prior to polarization. (Reprinted with
ments by Kuo et al.365 The model calculations are based
on a large polaron model with equilibrium constants as
given in ref 216. Thick line: calculated stoichiometry. thin
line: calculated conductivity. (Reprinted with permission
from ref 216. Copyright 2000 Elsevier.)
A natural question to ask is whether this “two-
regime” theory is consistent with the known proper-
ties of LSM. As recently reviewed by Poulsen,216 the
defect structure of LSM has some similarities with
other more reducible perovskites such as LSC and
LSF. Like these other perovskites, LSM has electrical
properties on the border between that of a p-type
semiconductor and a metal217-219 and becomes oxygen
substoichiometric at high temperature and low
PO2,202,220 as shown in Figure 35. However, unlike its
more reducible cousins (which may have significant
vacancy concentration at atmospheric PO2), LSM
maintains a nearly full perovskite stoichiometry
above ∼10-10 atm and in fact becomes superstoichio-
Figure 34. Total impedance (Z) and faradaic impedance metric at high PO2 (>10-3 atm), containing more
(ZF) of unsintered porous LSM on YSZ measured in air at oxygen than is consistent with the formal ABO3 unit
850 °C. Electrolyte resistance (Re) has been subtracted from cell. Neutron diffraction studies221 as well as numer-
the total impedance, while both Re and the double-layer ous defect thermodynamic models216,222,223 show that
capacitance (Cdl) have been subtracted from ZF. (Reprinted
with permission from ref 215. Copyright 1998 Elsevier.)
this “superstoichiometric” behavior actually results
from the formation of cation vacancies on the A and
Hence, a single arc is observed with relatively high B sites while the oxygen sublattice remains fully
characteristic frequency. occupied.
More recent impedance studies by Mitterdorfer and One expected consequence of these properties is
Gauckler215 of porous LSM on YSZ at low overpoten- that LSM will not be a mixed conductor at low
tial have shown that this single arc can be separated polarization under normal cathode conditions. Like-
into electrochemical and chemical contributions by wise, at cathode polarizations thermodynamically
correcting the data for double-layer capacitance using corresponding to PO2 < ∼10-10 atm (∼-500 mV) we
the method of Berthier et al.75 As shown in Figure expect LSM to become reduced, exhibiting ionic
34, these studies suggest that the electrochemical transport properties similar to that of LSC or LSF.
resistance at the LSM/YSZ interface is generally not This transition from a nearly pure electronic conduc-
zero, even when the LSM/YSZ interface is left un- tor at low overpotential to a good mixed conductor
fired. The remaining “faradic” (chemical) impedance at high overpotential is consistent with the Grenoble
has a Gerischer-like shape, consistent with co-limita- “two-regime” theory, at least qualitatively. Further-
tion by dissociative adsorption and surface diffusion more, since a transition from superstoichiometric to
of oxygen. An estimate of the chemical capacitance substoichiometric LSM involves a change in concen-
based on this arc is on the order of 10-4 F/cm2, which tration and thus transport of cation vacancies (gener-
when compared to asymptotic values of LSC and Pt ally much slower than anions at these temperatures),
is more consistent with a surface process than a bulk we might expect significant hysteresis, irreversibility,
process. and/or other nonstationary behavior near the transi-
tion between these two operating regimes. Indeed,

as discussed in more detail in section 5.4, work by a
variety of authors has shown nonstationary behavior
consistent with slow or irreversible transitions under
cathodic polarization40,213,224-226. While it is not clear
if Mn reduction is the only cause of this behavior, it
has certainly been identified as one likely factor.
5.3. Dense and Patterned Thin Films: Confirming

Two Regimes of Operation
The previous two sections illustrate the difficulty
of understanding the mechanisms governing oxygen
reduction on LSM without exact control of the
electrode geometry, materials, fabrication, and op-
erating conditions/history. To that end, workers have
made significant progress in the last 5-10 years
microfabricating dense, patterned electrodes in which
geometric lengths such as thickness and TPB length
can be specifically controlled. In this way it has been
possible to examine various asymptotes of behavior
(such as the surface and bulk paths) and probe
electrode mechanisms using a variety of techniques
Figure 36. Measured (0) and simulated (+) impedance
including impedance and isotope tracer experiments. of a thin (5 µm) dense LSM film on YSZ at 800 °C in air.
These studies have significantly clarified our under- Simulation is based on a Warburg model for finite length
standing of LSM and provided a variety of tools that diffusion. (Reprinted with permission from ref 160. Copy-
will no doubt be built on extensively in the future. right 1998 Electrochemical Society, Inc.)
One asymptote of interest has been the low-
overpotential regime, which has been investigated found evidence of a small finite resistance associated
extensively by impedance techniques using thin with the gas-solid and/or solid-solid interface(s).
films. As mentioned previously in section 4.3, one of This appears to be consistent with the distortion of
the first successful attempts to fabricate thin-film the impedance from a 45° limit at high frequency, a
perovskites was published by Mizusaki and co- strong indicator of an electrochemical kinetic contri-
workers,157 who used a pulsed KrF excimer laser to bution and double-layer capacitance associated with
flash evaporate and deposit 1-2 µm thick films of the solid-solid interface.215
La0.5Sr0.5MnO3 on YSZ. Near open circuit the initial An interesting exercise is to extrapolate Ioroi’s
impedance of these electrodes was found to be values for the bulk transport resistance of a thin film
controlled by bulk transport of oxygen though the to predict the minimum performance of a porous
film with a resistance that scales with electrode electrode, assuming bulk transport is rate limiting.
thickness and performance that decreases with in- In this scenario the interfaces are equilibrated,
creasing PO2. Indeed, this result appears consistent yielding a 3D bulk transport situation as modeled by
with what we know of LSM, since in the absence of Fleig.183 By appropriately scaling Fleig’s results
a parallel surface path the bulk path will be con- (Figure 30, region I), one can predict the area-specific
trolled by the very low anionic conductivity of LSM resistance of a porous body of block microstructure
at open-circuit conditions. Following significant an- as a function of σi and surface area a. Using the value
ode polarization, the authors found that the LSM film of L/Rbulk from Ioroi’s data at 800 °C,159 eq 7 yields a
becomes severely damaged and cracked and reverts value of σi of ca. 10-6 Ω-1 cm-1. Applying this value
to impedance behavior more consistent (in shape, to Figure 30 for a typical electrode area (20 000 cm2/
time scale, and PO2 dependence) with a porous LSM cm3), one obtains a minimum resistance of ∼0.003
electrode at low polarization. The authors conclude Ω cm2. By comparison, the actual impedance of a
that this damage results in creation of TPB contact porous LSM electrode of the same composition (made
area and thus changes the relative importance of by the same laboratory) at 1000 °C (a much higher
bulk vs surface transport in allowing electroactive temperature) is ∼1 (Ω cm2).159
oxygen to reach the electrode/electrolyte interface. What this calculation shows is that the rate of bulk
More recent studies of dense LSM films appear to transport observed in a thin film of LSM is at least
confirm these original conclusions as well as fill in 3 orders of magnitude too low to explain the perfor-
some of the details.27,159,160,167,227 In particular, Ioroi mance of porous LSM at low overpotential, assuming
and co-workers159,160 were able to produce very high an entirely bulk transport path. This calculation
quality films with clean, well-resolved impedances at echoes prior estimates of Adler and co-workers, who
800-1000 °C in air, as shown in Figure 36. Consis- showed that the zero-bias impedance of porous LSM
tent with bulk transport limitations, the impedance cannot be explained in terms of a bulk path.171 In
of these films was Warburg-like in shape and scaled addition, estimates of the chemical capacitance based
properly with electrode area and electrode thickness, on Ioroi’s impedance for porous LSM yield values of
assuming an entirely bulk path. By extrapolating 10-4-10-3 F/cm2, which as mentioned previously in
their results to zero film thickness, the authors also section 5.2 are more consistent with a surface process
than a bulk process. Comparing this capacitance to

that of Pt, we might expect a utilization region (lδ)
on the LSM surface of perhaps 100-1000 nm; how-
ever, no estimates have yet been offered in the
literature for this distance.
Stepping beyond the low-overpotential regime,
Horita and co-workers examined the degree to which
a dense film LSM electrode becomes reduced under
moderate to high cathodic polarization using 18O
tracer techniques.227 In these experiments an 18O2
tracer was introduced to the gas atmosphere during
steady-state cathodic operation, and then at some
later time the tracer was frozen in place via quench-
ing of the sample. The resulting18O depth profile
through the film was then analyzed using secondary-
ion mass spectrometry (SIMS). These profiles were
found to be consistent with increased vacancy con-
centration (and thus tracer diffusivity) near the YSZ Figure 37. Qualitative summary of current contributions
interface caused by reduction of the film. This result from the surface path (IS) and the bulk path (IB) for LSM
disk microelectrodes on YSZ based on i-V measurements
provides perhaps the most direct evidence to date in air at 800 °C. (Reprinted with permission from ref 228.
that bulk oxygen transport in LSM is enhanced near Copyright 2002 Elsevier.)
the YSZ interface under moderate to high polariza-
tion. However, because this film is dense (and thus comes increasingly important until at high surface
precludes a parallel surface path), it is difficult to areas the surface path dominates at small (and
assess the importance of this enhancement vis-à-vis anodic) overpotentials. This appears to be consistent
other reaction steps. In particular, the authors with the calculation discussed above, which shows
observed no difference in tracer ratio across the gas/ that the bulk path is insufficient to explain observed
LSM or LSM/YSZ interfaces, suggesting that the current densities at low polarization. Meanwhile, at
enhancement is not so great in this case as to be high cathodic polarization the bulk path appears to
faster than interfacial processes. remain an important factor even with high TPB area,
A more recent study by Brichzin and co-workers228 consistent with the Grenoble and other results (de-
examined the impact of enhanced bulk transport at scribed previously) showing substantial reduction of
high overpotential on the kinetics when a TPB is Mn and an increase in chemical capacitance.
present. To accomplish this, the authors took the While Brichzin et al.’s results provide further
thin-film technique one step further by examining support for a “two-regime” view (at least from a
small circular microelectrode films of varying diam- macroscopic viewpoint), it remains unclear exactly
eters, contacted individually using a micromanipu- where and how the surface-to-bulk transition occurs
lator. By varying diameter and thickness and intro- and to what degree these regimes overlap under a
duction of an alumina-blocking layer at the solid- given set of conditions. In particular, Figures 30, 32,
solid interface, these workers were able to control the and 33 suggest that when surface adsorption/desorp-
ratio of areas available for the surface and bulk tion on LSM is fast relative to bulk diffusion, the bulk
paths. Reported were mostly steady-state polariza- transport path will be confined primarily to a small
tion measurements as a function of geometry factors region near the TPB interface. In other words, even
and anodic or cathodic polarization ((300 mV). The when a bulk transport path is active, its macroscopic
resolution of the data appears to be quite good, and scaling may still obey that of the surface path,
distinct trends were observed in how the polarization making it impossible to separate from true surface
resistance scales with the various geometric factors. processes via geometric scaling alone. Further work
The authors analyzed these trends in terms of a of this type involving impedance and other quantita-
qualitative model which estimates the relative im- tive measurements would undoubtedly help clarify
portance of the “surface” and “bulk” paths in con- some of these questions.
tributing to the current in the anodic and cathodic Also contributing significantly to this “surface vs
regimes. bulk” debate has been a series of studies by Horita
The results of this analysis are summarized in and co-workers employing isotope tracers to track
Figure 37. Like prior workers studying thin films, the relative importance of the surface and bulk
the authors conclude that dense films without a TPB paths.227,229-231 As shown in Figure 38, these authors
under small or cathodic polarizations operate prima- used rf sputtering to deposit a thin-film grid of LSM
rily by a bulk path since the surface path is blocked. having thickness 0.5 µm and width 2 µm on the
(Interestingly, they found that dense films under surface of polished polycrystalline YSZ. An 18O2 tracer
anodic polarization appear to operate under a mixed was then introduced to the atmosphere during ca-
regime, although it is not clear how much nucleation thodic polarization, followed by postmortem analysis
and transport of O2 along the solid-solid interface of the quenched sample using SIMS (as described
contributes to the apparent surface path current.) In previously). By removing the LSM from YSZ in an
contrast, as the “porosity” is increased (microelec- acid bath, it was also possible to examine the YSZ
trode diameter is decreased), the surface path be- underneath the LSM grid. The authors also examined
Figure 38. Schematic and geometry of patterned thin-

film LSM electrode on YSZ studied by Horita and co-
workers. After establishing steady-state cathodic polariza-
tion, the atmosphere surrounding the electrode is rapidly
switched from 16O2- to 18O2-rich for 10 min at fixed total
PO2. The sample is then quenched to room temperature and
postmortem analyzed using secondary-ion mass spectrom-
etry (SIMS) imaging. (Reprinted with permission from ref
230. Copyright 2000 Elsevier.)
Figure 40. Secondary-ion ratio map (18O/16O) of YSZ in

the cell shown in Figure 39 following removal of LSM by
etching in acid. (a) Map at the exposed YSZ surface. (b)
Local analysis of the tracer across the region A-B shown
in a at various depths, achieved by Cs+ ion ablation. (c)
Qualitative models for oxygen incorporation used to ratio-
nalize the results in b. (Reprinted with permission from
ref 231. Copyright 2002 Elsevier.)
a significant amount of the tracer is drawn into the

YSZ (unlike the gas-exposed YSZ surface, which
exhibits no exchange). However, when the isotope
profile in the YSZ is carefully analyzed, it is found
that a higher-than-average amount of tracer is
absorbed in a region beneath the LSM closest to the
TPB. Figure 40b shows cross sections of this profile
Figure 39. Secondary-ion maps of 18O- (oxygen tracer) as a function of depth into the YSZ. The region over
and Mn16O- (positional reference) of the patterned thin-
film electrode shown in Figure 38 following polarization
which the isotope is found to spread is approximately
at -0.34 V at 700 °C. (a) Tracer map at the top surface of independent of depth.
LSM/YSZ, showing selective incorporation into LSM. (b) In interpreting their results Horita et al. consid-
Tracer map near the LSM/YSZ interface (acquired after ered two possible reasons for increased activity near
ablation of LSM off the surface with Cs+), showing deep the TPB (Figure 40c). One possibility is that a
penetration of tracer into LSM. (Reprinted with permission parallel surface path exists which produces a high
from ref 230. Copyright 2000 Elsevier.)
current density and thus tracer incorporation near
the TPB. However, this scenario is predicted to result
the position of various cations (Zr, Mn), both as a in a narrow distribution of the tracer near the TPB,
spatial reference and as a way to gauge cation which increases with depth. In contrast, if the ionic
interdiffusion. current is generated by a bulk path, a higher-than-
Figure 39 shows isotope tracer maps for LSM both average ionic current is also predicted near the TPB
at the surface and near the interface as a function of (per the results of Fleig shown in Figure 30). How-
polarization. These results indicate that under in- ever, the spread of the tracer in this case would be
creasing polarization, isotope tracer is incorporated approximately independent of depth, which is what
more deeply into the LSM bulk and that this incor- Horita et al. observed. If correct, this result suggests
poration occurs everywhere with the LSM near the that increased current associated with the TPB
interface not just at the TPB. This result would seem results from a bulk as well as surface pathway.
to corroborate the increased significance of the bulk What casts some doubt on this interpretation is
path with increased overpotential. Consistent with that the tracer exhibits significant lateral dispersion
this result, when the LSM is removed and the YSZ perpendicular to the TPB, even for gold on YSZ
underneath is analyzed (Figure 40a), it is found that (which the authors also studied). Indeed, the profile
Figure 41. Linear-sweep voltammagrams of a porous

LSM electrode on YSZ in air at 950 °C as a function of
sweep rate. (Reprinted with permission from ref 233.
Copyright 1998 The Electrochemical Society, Inc.)
Figure 42. Impedance characteristics of porous LSM
images show evidence of significant defects in the electrodes on YSZ, measured at zero bias and 945 °C in
grid film, including spots of enhanced or retarded air, as a function of polarization history and processing
isotope exchange beneath the LSM. These defects conditions. U-11, U-12, and U-13 correspond to firing
might involve microcracks or interfacial fissures temperatures of 1100, 1200, and 1300 °C, respectively. Bold
line: initial impedance. Thin line: impedance measured 2
having enhanced tracer mobility relative to the flow min following cathodic polarization at 100 mA/cm2 for 30-
of tracer with the applied current. More work is 90 min. Dashed line: impedance measured 30 min follow-
needed (both in measurements and modeling) before ing cathodic polarization. (Reprinted with permission from
this type of study yields a definitive conclusion ref 209. Copyright 1997 The Electrochemical Society, Inc.)
regarding the relative importance of the surface and
bulk paths under an arbitrary set of conditions as one would expect this hysteresis to close with de-
well as the size of the active region when the bulk creasing scan rate, not open. Indeed, the relaxation
path is engaged. time for bulk oxygen vacancy disturbances (based on
the impedance under the same conditions211) is about
5.4. Nonstationary Behavior in LSM 16 s, much faster than implied by the voltammo-
grams (20+ min).
In addition to the possibility of multiple transport
paths, our understanding of reaction mechanisms on Perhaps further insight can be gained by examin-
LSM is further complicated (as with platinum) by ing the impedance as a function of polarization
pronounced nonstationary behavior in the form of history, which has been documented by a number of
hysteresis of inductive effects. These effects are workers.40,208,209,226,234,235 As shown by van Heuveln et
sometimes manifest as the often-mentioned (but al.,209 reproduced in Figure 42, the zero-bias imped-
little-documented) phenomenon of “burn-in”, a term ance of porous LSM electrodes on YSZ at 945 °C in
used in development circles to describe the initial air decreases following 30-90 min of cathodic polar-
improvement (or sometimes decline) of the cathode ization and then relaxes back toward its original
kinetics after a few hours or days following initial characteristics over a similar period of time. This
polarization (after which the performance becomes return to original behavior, consistent with the
relatively stable). As recently reported by McIntosh reversal of the hysteresis in the voltammograms in
et al., this effect can improve the measured imped- Figure 41, eliminates irreversible microstructural
ance of a composite LSM/YSZ cathode by a factor of evolution as a primary cause of this effect, as
50 relative to an unpolarized cathode at OCV.40 Such proposed earlier by Tsukuda.236 Although the authors
an effect is important to understand not only because failed to mark frequencies on their impedance dia-
it may lead to insight about the underlying electrode grams, they fit the impedance to an equivalent circuit
kinetics (and ways to improve them), but also because exhibiting two time scales and reported the capaci-
it challenges the metrics often used to assess and tances. From this it can be determined that the
compare relative cell performance. higher frequency feature of the impedance (with a
For perovskite electrodes, the earliest kinetic study time scale of ∼100 kHz, capacitance ≈ 10-5F/cm2) is
of hysteretic effects appears to come from Ham- independent of polarization history, while the lower
mouche and co-workers, who showed that the i-η frequency feature (100-1000 Hz, capacitance ≈
characteristics of porous LSM/YSZ in air at 960 °C 10-3F/cm2) shrinks following polarization by up to a
exhibit a potentiodynamic hysteresis when scanned factor of 50, depending on electrode firing tempera-
slowly (1 mV/s) between 0 and -1200 mV cathodic ture. Jiang and co-workers reported similar behavior
polarization.224 A clearer demonstration of this effect, for A-site-deficient LSM on YSZ at 800 °C (where all
more recently provided by Jiang and co-workers, is time scales get longer), showing that it takes ∼4 h
shown in Figure 41.232,233 Hammouche and co-work- for polarization to have its full effect. More recently,
ers attributed this hysteresis to the formation of McIntosh et al.40 witnessed similar effects for com-
oxygen vacancies in LSM at high overpotential, posite LSM/YSZ electrode in an operating fuel cell;
which (as discussed in sections 5.2 and 5.3) appears both the impedance and steady-state overpotential
to open a parallel bulk-transport-mediated reaction at low polarization drop by a factor of 50 following
pathway. However, if this was the only explanation, high polarization, and these changes primarily effect
One explanation offered by the authors is that

MnOx- and/or SrO-like phases or moieties naturally
exist on the LSM surface, which initially retard O2
adsorption; upon etching (and presumably upon
polarization somehow), these obstructing phases are
removed, facilitating dissociative adsorption and/or
surface transport. One piece of evidence that the LSM
surface composition is altered significantly comes
from the observation that following the etching the
acid was found by ICP analysis to contain amounts
of Sr, Mn, and La inconsistent with the bulk cation
stoichiometry, suggesting selective etching of Sr >
Mn . La. Thus, either LSM started with excess Mn
and Sr on the surface or ended up depleted in Mn
and Sr. One appeal of this theory is that the time
required for similar changes to occur to the electrode
surface under polarization would be dictated by
diffusion of the cations involved (La, Sr, Mn), which
might reasonably require many hours even at 973
K. Such a slow change appears to be more consistent
with the relaxation time of the electrode performance
(hours) than other processes such as anion diffusion
(minutes).
It has been noted (mostly in the context of perov-
Figure 43. Impedance characteristics of porous LSM skite membranes) that a large gradient in oxygen
electrodes on YSZ, measured at zero bias and 900 °C in chemical potential can cause kinetic demixing of the
air, following various amounts of time under cathodic
polarization at 200 mA/cm2: (a) normal sample, (b) sample cation constituents in a transition-metal oxide.237-242
etched in 1 M HCl solution for ∼15 min. (Reprinted with While we do not review this literature in any detail,
permission from ref 226. Copyright 2001 Elsevier.) a common observation is that for a binary or ternary
oxide, segregation of different cations along the µO2
the low-frequency portion of the impedance. The gradient are expected, depending on their mobility
initial characteristics of the cell are largely recovered and transport coupling. This is particularly true for
after several hours sitting at zero polarization. LSM, which is cation nonstoichiometric under ambi-
As discussed in sections 5.2 and 5.3 (and elaborated ent PO2;216 thus, cation motion and crystal convection
on more fully in section 6), the higher frequency is expected under PO2 gradients. We saw strong
feature is likely related to the interface while the low- evidence in section 5 that under high polarization,
frequency feature is more consistent with chemical LSM operates (at least in part) via a bulk oxygen
processes on the LSM surface. Thus, van Heuveln’s transport path, which similarly corresponds to a µO2
result above suggests it is surface diffusion (as gradient across the material between the gas-exposed
proposed van Heuveln209) and/or oxygen adsorption/ surface and the solid-solid interface. Thus, any
dissociation that temporarily improves with polariza- sustained bulk oxygen transport might easily explain
tion. Hammouche’s suggestion (above) that this cation segregation to/from the surface. Some evidence
hysteresis corresponds to the opening and closing of for cation movement is apparent Kirkendall porosity
a parallel bulk oxygen transport path seems unlikely, formed in the corner region of LSM near the TPB
since the lowest frequencies in the impedance (∼0.01 under sustained cathodic polarization (Figure 33).
Hz at 950 °C211) come nowhere close to the 1-10 h Thus, while the hysteresis may not correspond di-
time scales exhibited by the hysteresis. rectly to a change from surface to bulk path (and
A result of significant interest in this discussion is back), it may nonetheless be caused by this transition
that of Jiang and Love,226 who examined the effects given sufficient time under load.
of acid treatment (as well as polarization) on the As discussed in section 6.1, a relatively exhaustive
performance of A-site-deficient porous LSM on YSZ HRTEM and AFM study was conducted by Mitter-
at 900 °C in air (Figure 43). Consistent with prior dorfer and Gauckler of how secondary phases form
studies, they showed that cathodic polarization sub- at the LSM/YSZ boundary and how these phases
stantially reduces the low-frequency portion of the effect electrode kinetics.215 This study placed the time
impedance response at zero bias. They then studied scale for cation-transport processes in the correct
an identical electrode that had been etched in 1 M range to be consistent with the theory described
HCl for ∼15 min at room temperature prior to above. However, while all this may be interesting and
testing. The acid-etched cell had a much smaller useful speculation, to date no in-depth studies of the
impedance to start with, nearly identical to that of LSM surface as a function of A/B ratio, polarization
an unetched cell that had been polarized for several history, or other factors have been performed which
hours. Cathodic polarization of the etched cell yielded would corroborate any of these hypotheses. Such a
little additional benefit. Again, it was only the low- study would require combining detailed materials
frequency portion of the impedance that was reduced characterization with careful electrochemical mea-
by acid etching. surements on well-defined model systems. Given the
fact that polarization has such a profound (and tance of the solid/solid interface may depend signifi-
generally underappreciated) influence on perfor- cantly on the formation of insulating reaction prod-
mance, it seems likely this would be an important ucts between LSM and the electrolyte. In addition,
subject for investigation in the future. the composite microstructure of a commercially rel-
evant fuel-cell electrode is significantly more complex
5.5. Summary: Uncertainties in Our than the ideal structures discussed so farsa signifi-
Understanding of Oxygen Reduction on LSM cant remaining issue is the relationship of perfor-
mance to microstructure in light of our understand-
The work reviewed in sections 5.1-5.4 represents ing of the reaction mechanism itself (section 6.4).
a significant improvement over the last 10-15 years
in our understanding of oxygen reduction mecha-
nisms on LSM. However, significant uncertainties 6. Factors Complicating our Understanding of
remain due in part to the complexity of the reaction SOFC Cathode Mechanisms
mechanism itself but also the extreme sensitivity of The results reviewed in sections 3-5 represent
the limiting factors to the exact structure, processing significant progress over the last ∼20 years in
history, operating conditions, and operation history understanding basic mechanisms governing oxygen
of the electrode. To summarize, some of the relevant reduction in SOFC cathodes. However, it should be
highlights (and remaining challenges) are given emphasized that this knowledge has been built by
below. studying a large number of materials and cases over
(1) Surface path at low overpotential. Qualitative a long period of time, often by examining model
and quantitative analysis of impedance data, tracer systems or asymptotic cases. This is not the same
studies, as well as various studies of thin-film thing as being able to diagnose the mechanism of a
electrodes suggest that under low-overpotential LSM particular electrode under a specific set of conditions
operates primarily via a surface-mediated mecha- or knowing all the factors that govern the rates of
nism (like Pt). This conclusion appears to be consis- the various physical processes comprising that mech-
tent with the properties of LSM, which is fully oxygen anism. In particular, workers have found that the
stoichiometric under ambient PO2. However, little is rates of these processes as well as their relative
known about how far the active region of reduction importance in the mechanism depend significantly
extends beyond the solid/solid interface (via surface on a number of “outside” factors not determined by
diffusion) or the relative importance of chemical steps the mechanism itself. Some examples include second-
(on the LSM surface) vs electrochemical kinetics at ary phases and impurities influencing electrochemi-
the solid/solid interface. cal kinetics at the interface, the macroscopic geom-
(2) Bulk path at moderate to high overpotential. etry and microstructure of the electrode, or changes
Studies of impedance time scales, tracer diffusion in the properties of the materials due to fabrication
profiles, and electrode microstructure suggest that conditions or degradation. An enormous amount of
at moderate to high cathodic overpotential, LSM work has been conducted in the last 15-20 years to
becomes sufficiently reduced to open up a parallel understand these various factors; the purpose of this
bulk transport path near the three-phase boundary section is to review those factors that have been the
(like the perovskite mixed conductors). This effect most well studied and/or appear to have the greatest
may explain the complex dependence of electrode significance in determining the rate of oxygen reduc-
performance on electrode geometry and length scale. tion within SOFC cathodes.
To date, no quantitative measurements or models
have provided a means to determine the degree to 6.1. Sensitivity of Interfacial Electrochemical
which surface and bulk paths contribute under an Kinetics to Secondary Phases and Impurities
arbitrary set of conditions.
(3) Pronounced nonstationary behavior. Numerous One of the most heavily studied factors thought to
workers have reported significant hysteresis and/or influence cathode performance has been the issue of
irreversibility in the behavior of LSM. This factor is reactivity between the electrode material and the
important to be aware of since it is possible for two electrolyte (usually YSZ) to form insulating secondary
labs studying the same types of electrodes under phases. This subject is sufficiently broad and complex
similar conditions to arrive at completely different to warrant its own review, and readers having a
conclusions, depending on the exact history of fabri- detailed interest in this topic are encouraged to read
cation and testing. Hopefully as workers move for- previous literature reviews in papers by Kawada243
ward, these effects will provide additional clues as and Mitterdorfer.215 Our main focus here is on how
to the reaction mechanism dominating under specific these secondary phases (or other impurities) appear
conditions. to retard the reaction, particularly electrochemical
(4) Extreme sensitivity to processing and operating kinetic processes occurring at the interface.
history. It has proven difficult to arrive at a consensus As early as 1969, studies of perovskites on YSZ
understanding of oxygen reduction on LSM, in part suggested the formation of insulating reaction prod-
because electrode performance/characteristics depend ucts at the interface during processing and/or cell
so strongly on the exact details of processing and operation.244 This was confirmed in 1985 by Lau and
operating history. Many of the outstanding issues are Singhal, who showed that La2Zr2O7 forms at the
tied so strongly to the subtleties of processing that interface between dense LSM and single-crystal YSZ
they warrant their own section (section 6). For at typical electrode processing temperatures.245 Soon
example, as discussed in sections 6.1-6.3, the resis- after, Yamamoto and co-workers published a study
of sputtered La1-xSrxMO3-δ electrodes on YSZ (M ) at the diffusion couple interface, the authors con-
Co, Fe, Mn, Cr), annealed at various temperatures cluded that the reaction proceeds by cationic diffusion
from 900 to 1100 °C and examined using SEM and of La and Sr through the interfacial LZ and/or SZ
X-ray diffraction (XRD). These studies identified layer, resulting in the formation of cation vacancies
multiple secondary phases which were identified as in LSM and the continued precipitation of LZ and/or
La2Zr2O7 (LZ) and Sr2ZrO4 (SZ). The amount of these SZ on the YSZ side of the interfacial layer. Consistent
secondary phases depended on the exact composition with this hypothesis was the observation of Kirken-
of the materials involved but generally increased with dall porosity in the LSM phase near the interface,
annealing temperature and time. They likewise found which would be expected to form upon sufficient
that electrode resistance increases with annealing depletion of A-site cations in the LSM phase.
time, suggesting that these phases play a significant Since the formation of secondary phases at the
role in retarding electrode/electrolyte interfacial pro- interface appears to be associated with activity and
cesses. A variety of other perovskite electrode materi- mobility of the A-site cations, a variety of workers
als have since been shown to react with YSZ, includ- have investigated perovskites with A/B cation ratio
ing La1-xCaxMnO3(δ (LCM),246,247 Gd1-xSrxCoO3-δ,248 <1 in an attempt to mitigate the formation of
La1-xSrxCo1-yFeyO3-δ (LSCF),249,250 and LSC and reaction products. An early example is a study by
LSF.251 Yokokawa and co-workers examined the Yamamoto et al.,257 who showed using XRD that no
relative thermodynamic stability of various perov- reaction products form between La0.8MnO3 and YSZ
skite phases with respect to reaction products in the after sintering at 1200 °C for 200 h. Such prevention
presence of zirconia, providing a framework for (or at least delay/retardation) of reaction products for
understanding the thermodynamic driving force be- perovskites with A/B ratio <1 has since been reported
hind the formation of these phases.252,253 by a variety of workers.235,243,249,258-261 Unfortunately
Exactly how these secondary phases influence there are generally few results that really tie the
cathode performance remains somewhat circumstan- reduction of these phases to improvements in elec-
tial. Labrincha et al.254 studied the electrical proper- trode performance (let alone particular aspects of that
ties of LZ over a range of temperature and PO2 and performance such as the interfacial electrochemical
found it to have low conductivity under typical kinetics). Indeed, even with materials having A/B
SOFC cathode operating conditions (∼10-4 Ω cm2 at ratios <1, reaction products often still appear at
1000 °C). This suggests that even a small layer of higher temperatures and/or longer annealing times.
this material is likely to significantly interfere with Thus, given that these phases are often detected
charge transfer at the perovskite/YSZ interface. using XRD (which has a detection limit of 1-3%), it
Poulsen and Vanderpuil255 have likewise shown that is not entirely clear that secondary phases are not
a variety of possible lanthanum zirconate and still present in undetectable quantities. Also, depend-
srontium zirconate reaction products are possible, ing on the mechanism of secondary phase formation,
which have electrical properties likely to interfere the low A/B ratio may only serve to delay rather than
with interfacial charge transfer. Consistent with prevent secondary phases. For example, one promi-
this hypothesis, numerous workers have reported nent idea has been that dissolution of relatively small
perovskites on YSZ to exhibit a high-frequency and mobile B-site cations from the electrode material
impedance feature suggestive of interfacial resis- into YSZ leads to a local increase in A-site activity
tance.27,164-166,187,210,211,215 Workers have also made at the interface.243 Given enough time and temper-
significant progress in understanding how and where ature, this may occur even when the A/B ratio is
these phases form, leading to a variety of logical initially low enough to be thermodynamically stable
suppositions about why they may influence interfa- with respect to reaction products. In support of this
cial electrochemical kinetics or other processes gov- idea are thermodynamic calculations by Yokokawa
erning performance. et al.,252,253,262 who showed that Mn is highly soluble
One of the first studies of how these secondary in YSZ and that this solubility is PO2 dependent,
phases form was performed by van Roosmalen and suggesting that the formation of these various phases
Cordfunke.256 These authors used SEM/EDS and may be very sensitive to how easily oxygen can
XRD to study postannealed diffusion couples of LSM transport to/from the interface during electrode
and YSZ as well as pressed and fired powder mix- processing.
tures of LSM and YSZ. These experiments showed Of interest to this discussion is a paper by Simner
that reaction products in sufficient quantity to detect et al.,251 which challenges whether one can always
by XRD (1-3%) form at temperatures as low as 1170 detect deleterious materials at the interface as
°C. The two principle reaction products observed were explicit phases. Motivated by the observation that
La2Zr2O7 (LZ) and SrZrO3 (SZ), with the relative perovskites LSC and LSF exhibit better performance
amount of LZ and SZ depending on the La/Sr ratio on ceria (see section 6.3), they did a careful examina-
in the LSM. Calcia- and baria-doped LaMnO3 were tion of the reactivity of LSC and LSF with 8 mol %
found to be similarly reactive with YSZ, and reactiv- YSZ, fired as mixed and pressed powders up to
ity of LSM with YSZ having 3% or 8% yttria was 1400 °C. Not surprisingly, LSC was found to be very
found to be similar. In the case of the diffusion reactive, forming large amounts of LZ, SZ, and Co3O4.
couples, the layer of reaction products formed at the In contrast, LSF was generally less reactive, and in
interface was found (using SEM) to be on the order the case of A-site-deficient LSF, (La0.8Sr0.2)0.95FeO3-δ,
of 1 µm after 600 h at 1280 °C and 10-15 µm after the authors could not find evidence of secondary
600 h at 1480 °C. By employing Pt diffusion markers phases using SEM or XRD. Nonetheless, measurable
angle shifts (∼2° in 2Θ) were observed for the

primary LSF diffraction lines, indicating a significant
change in average unit cell size. The authors inter-
preted this change in terms of Zr4+ dissolution into
the perovskite, which they show to reduce substan-
tially the electrical conductivity of LSF. While it is
not clear how deeply Zr would dissolve into an LSF
electrode, it is also not clear how much would be
needed to degrade performance. In other words, it is
entirely possible that even when secondary phases
are not detected in large enough quantities to see by
SEM and XRD, a thin (1-10 nm) insulating layer of
a secondary phase (or simply a compositionally
modified primary phase) could still exist that intro-
duces an interfacial resistance.
We should also point out that annealing of the
interface, though ultimately responsible for the for-
mation of secondary phases, may be a very important
part of establishing ionic contact. For example, Horita Figure 44. Qualitative description of La2Zr2O7 (LZ)
et al.263 recently conducted 18O2 tracer experiments formation at the LSM/YSZ single-crystal interface based
on sputtered films of LSM on YSZ and showed that on HRTEM and AFM measurements, as explained in the
the “as-sputtered” interface (fabricated at 700 °C) has text. Shaded region corresponds to LZ. (Adapted with
significantly higher resistance to ion exchange, which
they attribute (based on SEM images) to a lack of
bonding at a local level. With increased annealing ability of electronic charge carriers. Consistent with
time and temperature, the LSM and YSZ bond more this hypothesis, Mitterdorfer and Gauckler observed
completely, corresponding to facile ion exchange apparent increases in both electrochemical and chemi-
across the LSM/YSZ interface. cal contributions to the impedance upon annealing
of the electrode.215
To better understand the structure and composi- However, beyond this, it is difficult to draw a more
tion of the interface and its impact on interfacial specific conclusion. In particular, it is now known how
electrochemical kinetics, a number of workers have the electrochemical and chemical contributions to the
conducted more detailed characterization studies of impedance evolve through the different stages of
the perovskite/electrolyte interface using high-resolu- development of these phases. Also, at higher over-
tion transmission electron microscopy (HRTEM), potential (or with perovskite mixed conductors) we
electron diffraction, X-ray photoelectron spectroscopy might expect the impact of these secondary phases
(XPS), and atomic force microscopy (AFM).215,264-268 to depend significantly on the relative roles of the
Of particular interest is a paper by Mitterdorfer and surface and bulk paths as well as how far these
Gauckler,215 who studied nucleation and growth of phases spread along the LSM/YSZ interface (where
LZ between porous La0.85Sr0.15MnyO3(δ with (001)- they may also block ionic exchange across the solid/
oriented 9.5 mol % doped YSZ single crystals using solid interface). Unfortunately, besides the general
HRTEM, AFM, and XPS, as well as the impact of observation that performance gets worse with the
these phases in terms of electrode impedance. As formation of these phases, little is known. A more
illustrated in Figure 44, their results suggest that direct linkage of these phases to the exact changes
when LSM is A-site rich (y ) 0.98), nucleation of LZ seen in resistance and frequency response (as a
occurs at the TPB, where oxygen exchange is facile function of overpotential and time) remains elusive.
and La3+ and Zr4+ are available via surface diffusion. Finally, it would be remiss not to mention the
Once an island of LZ forms, it can then grow along potential impact of impurities in influencing various
the LSM/YSZ interface. In contrast, when LSM is electrode rate processes, particularly at the electrode/
A-site deficient (y ) 1.02), initial nucleation of LZ is electrolyte interface. For example, Kuscer and co-
retarded; epitaxial growth of a Mn-rich YSZ phase workers269 showed that during accelerated aging of
occurs first, drawing Mn out of LSM and Zr4+ and LSM electrodes on YSZ, small amounts of silica in
Y3+ from the surrounding YSZ surface. Once suf- the electrolyte tended to migrate to the interface,
ficient Mn has been extracted from LSM, LZ then forming an amorphous La-silicate phase between
precipitates at the TPB. the electrode and electrolyte. Besides the potential
What these results show is that regardless of A-site impact on interfacial electrochemical kinetics, forma-
deficiency, eventual nucleation of the undesirable tion of this phase was implicated in the observed
phase occurs exactly where it is least desired: at the delamination of the electrode from the electrolyte
TPB, where it results in an insulating gap between over time (see section 6.6). Similarly, concurrent
the LSM and YSZ surfaces. Restricting our attention preliminary work at Imperial College has shown that
to the low-overpotential regime, we might expect this small amounts of silica introduced during processing
gap to interfere with two processes: (1) surface of perovskite/ceria electrodes had a deleterious influ-
diffusion of electroactive oxygen along the LSM ence on performance.270,271 Given the propensity of
surface to the TPB and (2) the electrochemical silica to form low-melting eutectics and move along
formation of O2- at the TPB due to limited avail- ceramic grain boundaries, one would also suspect
sensitivity to the presence of alkaline or alkaline- the enthalpy of oxygen incorporation into LSC is
earth materials (common components of furnace different for the film than for the bulk material. ICP
insulation!). However, while the impact of certain analysis of the film following dissolution in acid
impurities (particularly silica) on the properties of confirmed that it had a 60/40 La/Sr ratio, although
ion conductors is well known,272-275 there have only somewhat depleted in Co (6%). This level of B-site
been a few studies to date that really investigate the deficiency does not seem to explain such a significant
impact of these impurities on electrode performance. departure from bulk properties. After considering and
rejecting several other possibilities, the authors
6.2. Alteration of Material Properties Near the proposed that stress in the film due to a forced match
Interface with the underlying ceria might explain the apparent
additional contribution to the oxygen exchange en-
Further complicating our understanding of inter- thalpy. This as yet unproven hypothesis is intriguing
facial electrochemical kinetics (as well as electrode since it implies that material properties may be
kinetics as a whole) is the observation that bonding subject to modification at an interface even when
of the electrode material with the electrolyte may reaction and/or interdiffusion does not occur.
result in changes to the electrode and/or electrolyte
Finally, another possibility often discussed in the
materials in the general vicinity of the interface. In
literature is that cation “dopants” from the electrode
section 6.1 we already saw some examples of this,
for example, the dissolution on Mn in YSZ, altering may enhance the electronic conductivity of the gas-
its ionic conductivity,243,276 or the dissolution of Zr into exposed surface of the electrolyte in the vicinity of
LSF, potentially reducing its electronic conductiv- the TPB, thereby extending the reduction zone along
ity.251 Thus, it is possible that effects that appear to the electrolyte surface via mixed conduction. The
be associated with the interface (due to their geo- surface exchange rate of oxygen on both YSZ- and
metric scaling or impedance time scale) may in fact rare-earth-doped ceria (as measured by isotope meth-
be local variations of chemical processes in the near- ods) is only about 1 order of magnitude lower than
interfacial region. on LSM at 700 °C.277-279 Thus, if there were sufficient
electronic conduction at the surface or in the bulk of
One notable possibility is that interdiffusion of the the electrolyte, it might be possible for net reduction
electrode and electrolyte during bonding may result to O2- to occur on the gas-exposed electrolyte surface.
in changes in cation composition (and therefore
There is precedence in the catalysis literature that
properties) of the electrode. Recent results supporting
this can happen; for example, workers studying CO
this possibility were published by Horita et al.,263 who
or CH4 oxidation on CeO2/ZrO2-supported Pt and Pd
examined 18O2 tracer diffusion in a sputtered thin
catalysts at high temperature280,281 have generally
film of A-site-deficient LSM on YSZ. As discussed
shown that the support enhances performance via
previously, this result showed enhancement of ionic
reduction and mixed conduction of oxygen in the
exchange across the LSM/YSZ interface with anneal-
ing, due to improved bonding. What is also of interest fluorite phase. Workers have also suggested that
is that the authors observed an approximately 3-fold mixed conduction in ceria can play a role in SOFC
decrease in the apparent 18O tracer diffusion coef- anodes.282 Various workers have provided circum-
ficient of the film upon annealing above ∼1000 °C. stantial evidence that this might occur under oxidiz-
This was explained by the authors in terms of ing conditions at an SOFC cathode. For example,
changes in cation composition of the film due to Kleitz and co-workers55 noted that the high-frequency
diffusion of Y and Zr into LSM, which according to (presumably interfacial) impedance of small silver
their SIMS depth profiles penetrate ∼1 µm at droplet electrode scales as r-0.6, where r is the droplet
1300 °C. Although SIMS intensities are not consid- radius. They explained this weak geometry depen-
ered a quantitative measure of concentration, the Y dence as an outward expansion of the charge-transfer
and Zr intensities 800 nm deep into the LSM region zone away from the TPB along the electrolyte surface.
were on the order of 10% of that in YSZ itself. Another example is work by van Hassel and co-
Unfortunately, since no XRD or TEM cross-sections workers, who, as shown in Figure 45, found that the
were performed, it is difficult to entirely eliminate performance of porous gold cathodes on YSZ (nor-
the possibility that secondary phases contribute to mally a very poor oxygen catalyst) is significantly
these cation intensities. However, this study does enhanced at all overpotentials by prior coating with
serve to illustrate that significant changes in bulk Fe2O3 and/or implantation of iron into the electrolyte
transport properties may occur simply due to inter- surface.99,283 Widmar and co-workers reported similar
diffusive bonding. enhancements for low-fired (750 °C) Pt electrode on
A related finding was reported by Kawada et al.,124 YSZ implanted with Ce or Mn at low overpotential.82
who studied the impedance of thin laser-deposited However, to date there does not appear to be much
films of LSC on SDC. As discussed in section 4.3, they evidence that mixed conduction in the electrolyte
analyzed the apparent chemical capacitance of the plays the dominant role in the enhancements men-
films in terms of changes in bulk oxygen stoichiom- tioned above. First, it should be emphasized that a
etry. In so doing they discovered a very surprising finite rate of oxygen exchange at the electrolyte
result: the apparent oxygen vacancy concentration surface (as measured at equilibrium by isotope
in the film in air is approximately 4 times smaller methods) is a necessary but insufficient criterion for
than for bulk LSC having the same cation composi- finite rates of oxygen reduction; for there to be a net
tion (60/40 La/Sr). Further analysis suggested that production of O2- at the electrolyte surface, electrons
6.3. Ceria as an Alternative Electrolytic Interface

In attempting to develop fuel cells based on ceria-
based electrolytes, workers have also examined the
reactivity of various perovskite electrode materials
with doped ceria. An early example is a study by
Chen and co-workers,250 who used a spin-coating
technique to deposit LSCF electrodes on YSZ as well
as 20 mol % samaria-doped ceria (CeO2)0.8(SmO1.5)0.2
(SDC). Consistent with previous findings, LSCF was
observed to react with YSZ to form LZ and SZ
byproducts. In contrast, analogous Ce-based com-
pounds resulting from a reaction between LSCF and
SDC were not observed. Consistent with this obser-
vation were Chen et al.’s measurements of electrode
impedance, which suggest LSCF exhibits less inter-
facial resistance when processed and tested under the
same conditions on the same substrate. On the basis
of this result, the authors suggested addition of a
ceria-based “protection layer” between the perovskite
electrode and YSZ, an idea which is now actively
being developed for intermediate temperature fuel
Figure 45. Current-overpotential characteristics of po- cells.251
rous gold electrodes on YSZ at 770 °C and PO2 ) 1 atm, Consistent with Chen et al.’s observations was a
comparing YSZ that has, and has not, been implanted with study by Adler and co-workers (described previously
8 × 1016 atoms/cm2 of 56Fe at 15 kV. (Reprinted with in section 4), who measured the impedance of LSCF
on SDC at intermediate temperatures. Their mea-
surements showed that above about 700 °C in air the
must also be able to gain access to the surface.
impedance is dominated by a single Gerischer-like
Workers have generally shown that when YSZ is impedance found to be consistent with a bulk-
implanted or alloyed homogeneously with Fe or Mn, mediated reaction pathway. Although the possible
its electrical conductivity is not significantly en- role of the surface path cannot be entirely excluded
hanced, even under moderately reducing condi- in this case, what is clear is that the impedance is
tions.82,243,284-286 Second, in isotope studies of low- dominated by chemical dissociation and transport
temperature bifunctional CO oxidation catalysts, steps while electrochemical incorporation at the
such as Au/Fe2O3, workers have shown that the role interface remains largely equilibrated at least above
of the transition metal in the support is essentially ∼650 °C. Adler et al. mention in passing that at lower
catalytic, i.e., stabilization of physisorbed O2 at the temperatures (450-550 °C) a small high-frequency
Au/oxide interface, leading to enhanced oxidation arc with a high activation energy appears, indicating
rates.287-289 It is likewise possible that enhanced the onset of interfacial electrochemical kinetic resis-
performance of metal cathodes on YSZ incorporating tance. However, it is unclear if this additional
transition metals or (other d-orbital cations) is an resistance is associated with secondary phases or
electrochemical-kinetic effect localized to the TPB simply an inherent resistance to ionic exchange
where by “localized” we mean simply “within a between two dissimilar ionic phases (as has been
distance accessible by electrons from the electrode observed, for example, at the physical interface
material”. Indeed, as shown in Figure 45, the en- between two pieces of YSZ290). This apparent lack of
hancements observed by van Hassel and co-workers an interfacial resistance (as evidenced by the absence
upon implantation of Fe into YSZ correspond to an of a distinct high-frequency impedance arc) appears
increase in the apparent exchange current density to be a general feature of perovskite mixed conductors
while the Tafel slope remains constant over several on rare-earth-doped ceria or (La,Sr)(Ga,Mn)O3 elec-
orders of magnitude. This observation suggests that trolytes.18,19,22,28,177,179-181,184,250,291,292
iron acts to promote the existing rate processes A notable exception may be Sr-doped ceria. As
(either on the Au surface or at the TPB) rather than reported by Liu and Wu,187 LSCF electrodes on 10
by changing the mechanism toward an alternative mol % Sr-doped ceria exhibit a significant high-
chemical pathway involving the electrolyte. Finally, frequency impedance arc in air at 650-750 °C, which
we should point out that gold is a poor oxygen is comparable in frequency (but somewhat smaller)
catalyst at high temperature. In the case of a highly to that observed for identically processed and tested
catalytic perovskite, it is not clear that incorporation LSCF electrodes on 8 mol % YSZ. One explanation
of transition metals from the electrode into the may be found in the thermodynamic data for the Sr-
electrolyte surface is going to provide any significant Ce-O system, which shows that the solubility of SrO
catalytic benefit not already provided by the electrode in CeO2 is less than 10 mol % at these tempera-
itself, particularly in cases where the utilization tures.293-295 Thus, precipitation of SrCeO3 at the
region is large and thus the exposed electrolyte LSCF/ceria interface is favored, depending on the
surface makes up such a small percentage of the area exact firing conditions and A/B ratio of the perovskite.
available for dissociative adsorption. Oddly, the authors did not consider this possibility,
conduction into the electrode material so as to extend

the active region beyond the electrode/electrolyte
interface. Along with ionic conduction, however,
comes potential problems including increased chemi-
cal and thermal expansion and higher apparent
reactivity between electrode and electrolyte (section
6.1). An alternative strategy has been to combine a
YSZ-compatible electrode material such as LSM with
an ionically conducting second phase (e.g., YSZ itself)
in a composite matrix. While the active region of an
LSM electrode may generally be more confined to the
LSM/YSZ interface than a mixed conductor, the
composite matrix makes up for this deficiency by
extending the LSM/YSZ interface. This strategy has
proven to be very promising but adds some complex-
ity to our diagnosis and understanding of which
Figure 46. Performance characteristics of a cathode- processes dominate the overall observed overpoten-
supported thin film Ni-YSZ/YSZ/LSM fuel cell at 600 °C tial.
in humidified H2 and air with and without a dense One of the first attempts to implement this strategy
protective yttria-doped ceria (YDC) protection layer intro- was reported by Kenjo and co-workers,300,301 who
duced between the porous LSM cathode and the thin-film
electrolyte. (Reprinted with permission from ref 296. fabricated and tested the performance of porous Pt
Copyright 1997 Elsevier.) composite electrodes containing YSZ, SDC, and erbia-
stabilized bismuth oxide (ESB) as well as composites
citing the variations in impedance with different of A-site-deficient La0.85MnO3 (LM) and YSZ. The
electrolytes as proof that oxygen reduction is limited electrode polarization losses of these electrodes were
by electrochemical kinetics at the TPB. On the measured using current-interruption techniques as
contrary, the very large apparent chemical capaci- a function of electrode thickness. For Pt/ESB and LM/
tance (∼1 F/cm2) associated with the low-frequency YSZ, the presence of the electrolyte was found to
impedance feature (which is similar in shape, size, enhance performance significantly (5-10 times) de-
and frequency for both YSZ- and Sr-doped ceria) pending on processing conditions and the volume
suggests a bulk-mediated path involving absorption fraction of the electrolyte. As shown in Figure 47,
and transport in LSCF. Meanwhile, the differences performance was also found to increase with elec-
seen at high frequency between YSZ- and Sr-doped trode thickness, saturating at a thickness of ∼10 µm
ceria might easily be explained in terms of different (30 or 40 µm for Pt/ESB). In the case of Pt/YSZ and
secondary phases formed, whose likely effect is to Pt/SDC, no enhancement was observed; however, the
interfere with ion-transfer across the entire solid- authors proved using electrical conductivity mea-
solid interface (not just at the TPB). surements that with their particular preparation
Given the apparent reversibility of the perovskite/ method any significant fraction of YSZ and SDC
rare-earth-doped ceria interface, Various workers particles interferes with the electrical connectivity
have more recently investigated Chen’s original sug- among Pt particles, making the electrode too resistive
gestion of a ceria-based protection layer between YSZ to function in these cases.
and perovskite cathodes.296-299 As shown in Figure As shown in Figure 48, the authors explained their
46, Tsai and co-workers found that by introducing a results in terms of a macrohomogeneous model
0.4 µm thick sputtered layer of yttria-doped ceria similar to that used previously for aqueous gas-
between YSZ and porous LSM, overall current den- diffusion electrodes.302 With this approach the elec-
sity of their cathode-supported fuel cell could be tronic and ionic conductors (Figure 48a) are viewed
improved by a factor of approximately 5×. More as homogeneously interpenetrating subphases, the
recently, anode-supported YSZ fuel cells using a electronic conductor having relatively high electrical
single-phase porous LSF cathode with a Sm-doped conductivity, while the ionic conductor has finite ionic
ceria protection layer have been reported to exceed resistivity F ) 1/σi. At the interface between these
operating voltages of 900 mV at current densities of phases, oxygen reduction occurs at a rate propor-
100 mA/cm.2298 This performance is competitive with tional to the local overpotential, resulting in passage
some of the best composite LSM/YSZ cathodes avail- of current from the ionic to the electronic phase, iy )
able today (see section 6.4). While these results are η(x)/k, where η(x) is the local overpotential and k in
exciting, more work is needed to confirm/reject cur- this case is the interfacial resistance per unit volume
rent hypotheses about how these layers actually work
(Ω cm3). Consideration of charge conservation a la
to improve performance and compare performances
classic porous electrode theory303 leads to the solu-
meaningfully among various development groups.
tions shown in Figure 48c,d. In analogy to the
situation shown in Figures 14b and 26a, the com-
6.4. Composite Microstructures posite electrode is found to be co-limited by interfacial
Figure 2 illustrated two dominant strategies that resistance and ionic transport in the ionic subphase,
have been used to enhance the performance of SOFC exhibiting a maximum utilization region of size
cathodes. One has been the introduction of ionic ∼xk/F and overall resistance proportional to xFk for
and tuning of the microstructure to match the local

natural extension region of the electrode material.
As one might expect, interpreting the electrochemi-
cal characteristics of composite electrodes is consider-
ably more challenging than understanding their
single-phase counterparts. Since work in this area
is sparse, we confine this discussion to a few of the
salient differences. First, the various physical pro-
cesses contributing to the overpotential are more
difficult to separate. With a single-phase electrode,
ohmic losses associated with the electrolyte (and in
some cases also interfacial electrochemical kinetics)
can be isolated or treated using equivalent circuit
analogies independent of how complex the processes
occurring in the electrode itself may be. In contrast,
as implied by Figure 48, the characteristics of a
composite electrode involve an inherent convolution
of the ohmic loss in the electrolytic subphase with
the electrochemical and chemical losses at the junc-
tion between the ionic and electronic (or mixed
conducting) subphases. This leads to wide dispersion
and overlap among the time scales for these pro-
cesses, making interpretation of impedance ex-
tremely challenging, as recently reviewed by Jør-
gensen and Mogensen.31
Second, the electrochemical characteristics of com-
posite electrodes are generally much more sensitive
to the details of the microstructure than single-phase
materials.30,35 Indeed, embedded in the parameter k
in Figure 48 are literally all the physical processes
reviewed in sections 3-5 plus the details of how the
Figure 47. Measured area-specific admittance (reciprocal two subphases are interconnected. While various
of the polarization resistance Rp) as a function of electrode authors have attempted to relate” in more detail to
thickness for Pt/ESB and LSM/YSZ composite electrodes. the microstructure, modeling is often limited to
Performance of the same electrode materials without an
ionic subphase are also shown for comparison. Lines idealized hypothetical microstructures that may or
indicate fits to the model shown in Figure 48, as discussed may not be relatable in a quantitative way to the real
in the text. (Reprinted with permission from refs 300 and system.307,314 In addition, since composite electrodes
301. Copyright 1991 and 1992 The Electrochemical Society, work by maximizing surface area and interfacial
Inc. and Elsevier, respectively.) contact, they are often fired at much lower temper-
atures than their single-phase counterparts, leading
an infinitely thick electrode. As shown in Figure 47, to additional uncertainty as to the geometry and
fitting of this model yields values of F and k found to quality of interparticle contact. These details of the
be in reasonable agreement with the properties of microstructure and interfaces, which may be critical
ESB and YSZ given the uncertainties in the details to performance, are extremely difficult to quantify
of the microstructure (interparticle contact, ionic using current cross-sectional sampling techniques
tortuosity, interphase contact area). (SEM, HRTEM, SIMS, AFM).
What this result shows is that for materials having
a small natural active region as a single porous phase Finally, due to their high surface area, composite
(50-500 nm for Pt, perhaps somewhat larger for LM), electrodes often incorporate microstructural features
the use of a composite microstructure can signifi- competitive or smaller than the active region of a
cantly enhance performance, extending the active single-phase electrode of the same material. In this
region to include a much larger fraction of the situation, we expect chemical transport processes to
electrode material (∼10 µm). A number of workers be largely equilibrated on a local (submicrometer)
have since demonstrated the benefits of this approach length scale, leaving dissociative oxygen ab(b)sorp-
for LSM/YSZ, some reporting active regions of similar tion as the only significant chemical resistance. One
size (5-10 µm) as single-phase mixed conduc- consequence of this situation is that the electrochemi-
tors.304,305 Besides establishing a new paradigm in cal kinetics contributing to k in Figure 48 may not
how to design SOFC cathodes, this approach has also be the same as those governing the kinetics of a
introduced a new set of issues to be considered, single-phase electrode, even when TPB area is known
including the efficiency of the ionic and electronic and properly accounted for. Rather, the mechanism
transport paths,24,306-309 quality and quantity of itself may shift toward a different set of co-limiting
internal interfacial contact between the ionic and steps, in particular catalytic reduction, interfacial
electronic subphases,30,310 functional grading of the electrochemical kinetics, and ionic transport in the
materials for structural or other purposes,36,311-313 ionic subphase.
Figure 48. Kenjo’s 1D macrohomogeneous model for polarization and ohmic losses in a composite electrode. (a) Sketch of
the composite microstructure. (b) Description of ionic conduction in the ionic subphase and reaction at the TPB’s in terms
of interpenetrating thin films following the approach of ref 302. (c) Predicted overpotential profile in the electrode near
the electrode/electrolyte interface. (d) Predicted admittance as a function of the electrode thickness as used to fit the data
in Figure 47. (Reprinted with permission from refs 300 and 301. Copyright 1991 and 1992 Electrochemical Society, Inc.
and Elsevier, reepectively.)
Despite these uncertainties, however, workers have

made impressive advances in cathode performance
over the last 5-7 years using composite microstruc-
tures. A recent paper by McIntosh et al. reports the
area-specific resistance of a 50-50 LSM-YSZ com-
posite (fabricated from powders at 1250 °C) to be
about ∼0.2 Ω cm2, both in a half cell and fuel-cell
configurations following polarization.40 Barnett and
co-workers report somewhat higher resistances for
LSM/GDC composites (∼1.0 Ω cm2 at 700 °C33);
however, it should be noted that these electrodes had
never seen a bias and thus may have much higher
performance under load (see section 5.4). Electrodes
incorporating mixed conductors as the electronic
subphase generally have even higher performance.
As an example, Figure 49 shows the impedance (as
a function of temperature in air) of a low-fired
(900 °C) LSCF/GDC composite electrode on YSZ
made by Murray and co-workers.35 This electrode has
an ASR of ∼0.03 Ω cm2 at 700 °C with no polarization Figure 49. Impedance of a low-fired composite LSCF/GDC
history. Huang and co-workers reported cathode electrode on YSZ in air as a function of temperature. (Inset)
resistances of ∼0.1 Ω cm2 at 700 °C for LSC/YSZ and SEM of the microstructure of the electrode showing very
LSF/YSZ composites made by impregnation tech- fine porosity but effective interparticle contact. (Reprinted
niques.315,316 with permission from ref 35. Copyright 2002 Elsevier.)
σi (Figure 50a). It can be shown that the measured

6.5. Current Constriction Effects IR drop occurs primarily within a distance d of the
In passing, we should also mention one additional electrode with a value given by47
microstructural factor potentially impacting the over-
all electrode performance: constriction of the ionic I
VIR ) (11)
current in the electrolyte near the electrode/electro- 2dσi
lyte interface. To better understand this effect,
consider a circular disk electrode of diameter d where I is the total current. If one views the interface
immersed in a semi-infinite electrolyte of conductivity between a solid electrolyte and a porous electrode as
express Relectrode on a superficial-area-normalized

basis, J is given approximately by
d
J ≈ 0.4 (12)
φRASRσi
where RASR is the area-specific resistance of the
electrode and φ is the fraction of the area through
which current must flow at the interface. Applying
eq 12 to a bulk-path-dominated electrode (Figure
50b), we can estimate J by taking d to be the average
diameter of the solid-solid contact points while φ is
the fraction of the electrolyte surface covered by the
electrode. For 2D rather than 3D constriction, the
situation is similar;319 thus, for a surface-dominated
mechanism, we would take d to be the effective width
of the TPB while φ would correspond the fraction of
the electrolyte surface within one-half TPB width of
the TPB line.
As an example, consider a porous LSC electrode
Figure 50. (a) Primary potential distribution surrounding with an average particle diameter of 0.2 µm operating
a disk-shaped electrode. (Reprinted with permission from via a bulk path. If the porosity is 30%, then φ will be
ref 47. Copyright 2004 John Wiley & Sons.) (b) Finite- approximately 0.7. Assuming a superficial electrode
element calculation of the 2D primary potential profile near resistance of 0.05 Ω cm2 at 750 °C and an electrolyte
an electrode with regular periodic contact to the electrolyte
(φ ) 0.5). Horizontal lines on the contour plot are lines of conductivity of 0.03 Ω cm-1, then J ≈ 0.01. Thus,
constant potential; vertical lines follow the current path. current constriction is probably not an issue in this
case. Alternatively, consider a porous LSM electrode
an array of such small current sources, one expects of similar morphology and performance, assuming a
that as the size of these sources get very small, the TPB width of ∼1 nm. In this case φ will be ap-
IR resistance will become increasingly dominated by proximately 1.4% of the superficial area. Thus, J
the constriction of the current near each source would be ∼0.002. Again, this is a number much
rather than the superficial current density of the cell. smaller than 1. These back-of-the-envelope calcula-
tions suggest that microscopic current constriction is
Workers have shown theoretically that this effect not a large effect under most practical situations, and
can be caused both at the microstructural level (due thus most excess ohmic losses reported in the litera-
to funneling of the current near the TPB) as well as ture arise from more macroscopic current constriction
on a macroscopic level when the electrode is not effects (such as insufficient current collection).206,320
perfectly electronically conductive and the current However, it is probably wise for investigators to make
collector makes only intermittent contact.206,244,317 estimates of J to ensure further analysis is not
Fleig and Maier further showed that current con- required for their system.
striction can have a distortional effect on the fre-
quency response (impedance), which is sensitive to 6.6. Long-Term Degradation
the relative importance of the surface vs bulk path.318
In particular, they showed that unlike the bulk A second form of nonstationary behavior is ir-
electrolyte resistance, the constriction resistance can reversible long-term degradation of the electrodes,
appear at frequencies overlapping the interfacial explicitly with either time or time under operation
impedance. Thus, the effect can be hard to separate in a given environment. Unfortunately, the literature
experimentally from interfacial electrochemical- in this area is neither plentiful nor consistent. In
kinetic resistances, particularly when one considers particular, there is a dearth of publications reporting
that many of the same microstructural parameters test data past about 100 h due (no doubt) to the
influencing the electrochemical kinetics (TPB area, extreme cost of conducting long-term testing. As a
result, degradation has remained primarily a subject
contact area) also influence the current constriction.
of concern for industrial developers who have the
To facilitate estimates of this effect on the micro- resources needed to tackle this issue but by the same
structural level, we point out the existence of an token are generally loath to publish important, hard-
important dimensionless group, J, that describes the fought trade secrets. Furthermore, of the few reports
ratio of the current constriction resistance to the that have been published, results often vary wildly
electrode resistance.47,319 If J , 1, then we expect the for seemingly similar material systems. For example,
constriction effect will be small compared to the workers have reported fuel cells based on porous
electrode resistance, while if J g 1, the constriction LSM cathodes lasting >50 000 h with less than 5%
effect is expected to be significant. For example, in degradation,321 while others have reported cathode
the case shown in Figure 50a, J is derived by dividing degradation of 1000% in less than 50 h.322 Given this
VIR (given by eq 11) by the electrode overpotential general lack of consensus as to rates and causes of
η ) IRelectrode, where Relectrode is the total effective long-term degradation, we only briefly mention a few
resistance of the disk (in Ω). If we consider a large of the factors reported to cause acute short-term
number of well-separated disks (Figure 50b) and re- degradation (100-1000 h).
(1) Morphological evolution. Driven in part by see below, these effects commonly influence the
parallel studies on nickel sintering and degradation polarization, leading to experimental errors, or ap-
in SOFC anodes,323-326 one concern has been that pear as “features” in the impedance or other electro-
LSM (or other ceramic cathode materials) can sinter chemical measurements that workers misinterpret
over time, losing active surface area and thus activ- as part of the electrode reaction mechanism.
ity. Several workers have reported morphological (1) Gas-phase effects. In aqueous electrochemistry,
changes in LSM electrodes with time or with time a rotating disk electrode is often employed to ensure
under current load, which has been correlated with that that mass-transfer effects do not obscure the
degradation.157,214,322,327-329 Why the changes occur measurement of electrochemical kinetics except at
and how they effect performance is not very clear. It very high polarization (limiting current). In contrast,
is also not clear that these changes are all caused by the gas inside the pore network of a GDE is stagnant
the same thing or that all electrodes are likely to and may be so for some distance outside the electrode
experience these changes. due to the presence of additional materials required
(2) Chromium contamination. As discussed previ- for current collection and/or mechanical support
ously, one must always be aware of the possible (mass-transfer distances of 30-1000 µm are not
presence of impurities which could have a variety of uncommon). In addition, the molar concentration of
effects, including enhancing sintering at operating a gas is ∼103 times lower than a liquid. Partially
temperatures. An impurity of particular concern has compensating for these factors is the fact that gas-
been chromium, which is often part of the intercon- phase diffusion is typically 5 orders of magnitude
nect (bipolar plate) in an SOFC stack either as a faster than diffusion in aqueous solution. For ex-
LaCrO3 (ceramic interconnect) or as part of an alloy ample, the binary diffusion coefficient of oxygen in
in the case of a metal interconnect. Workers have air at 700 °C is ∼1 cm2/s compared to the diffusion
shown that Cr can be quite mobile in an SOFC coefficient of dissolved O2 in H2O at 25 °C (∼2 × 10-5
environment330,331 and linked it clearly to degrada- cm2/s). However, these estimates also suggest that
tion.332 There has also been a considerable amount at O2 concentrations of ∼10-2 atm or below we might
of recent work that has probed the mechanism by expect gas-phase diffusion to become significant.
which Cr degrades performance.182,333-335 It appears One can roughly estimate the effects of gas-phase
that perovskite mixed conductors are less susceptible diffusion at steady state using a simple 1D diffusion
to the effects of chrome than LSM, possibly because model, which has been employed (in some form) by
they function by a bulk mechanism, which is less numerous workers.171,342,343 This approach yields the
influenced by the contamination of the surface by Cr- following expression for the linearized steady-state
containing species. chemical resistance due to binary diffusion of O2 in
(3) Evolution of secondary phases. Another concern a stagnant film of thickness L171
has been continued formation of LZ and SZ secondary
phases at the perovskite/YSZ interface as a function RT LVm
of time or current density.336,337 Accelerated testing, Rgas ) ∞
(13)
4F2 2xO2 Deff
AB
achieved by sustained heat treatments of the elec-
trode, suggests that degradation can occur by this
mechanism.338,339 However, whether such thermal where xO2∞ (assumed , 1) is the mole fraction of
treatments can be meaningfully extrapolated to oxygen in the well-mixed region outside the film
predict natural degradation processes is unclear. layer, Vm ) RT/P is the molar volume of the gas, and
(4) Thermal cycling. Finally, we should also men- DABeff is the effective diffusion coefficient which may
tion the issue of thermal cycling, which is a likely depend on porosity and tortuosity of the electrode as
source of stress on cells in a commercial SOFC device well as the effects of Knudsen diffusion (note the
due to thermal and chemical mismatches among the similarity to eq 7). As an example, for a diffusion
various materials.340,341 Hsiao and Selman have layer of 100 µm, an effective diffusion coefficient of
shown that this effect has a primary influence on the 0.1 cm2/s, a gas concentration of 10-3 (total pressure
interface, causing literal separation of the electrode 1 atm), and a temperature of 1000 K, this equation
from the electrolyte. predicts a resistance of ∼1 Ω cm2.
Thus, not surprisingly, numerous workers
6.7. Experimental Artifacts in Electrochemical have knowingly (or unknowingly) observed this
effect in polarization and impedance measure-
Measurements ments.28,70,79,171,184,291,342-347 For materials having a
As we have seen in the previous sections, our small apparent utilization region lδ compared to the
understanding of SOFC cathode mechanisms often electrode thickness (e.g., porous Pt or LSM at low
hinges on interpretation on the magnitude and time polarization), gas-phase effects normally appear as
scale of electrochemical characteristics. However, a separate arc in the impedance at low frequen-
these characteristics are often strongly influenced by cy.70,79,345-347 As an example, Figure 51 shows the
factors that have nothing to do with the electrode impedance of porous LSM/YSZ composite electrodes
reaction itself but rather the setup of the experiment. on YSZ at 950 °C, measured as a function of PO2 using
In this section we point out two commonly observed blended gases of O2 in N2. An arc with frequency ∼1
effects that can potentially lead to experimental Hz appears below PO2 ) 5% and grows dramatically
artifacts in electrochemical measurements: (1) po- with decreasing PO2.346 Telltale signs that this arc is
larization resistance caused gas-phase diffusion and related to gas-phase diffusion are (1) the admittance
(2) artifacts related to the cell geometry. As we will associated with this arc is first order in PO2, (2) when
Figure 51. Zero-bias impedance of a 60%/40% LSM/YSZ

composite cathode, measured at 950 °C as a function of
PO2.346 The magnitude of the lowest-frequency arc (∼1 Hz)
was quantified using equivalent circuit analysis and found
to scale inversely with PO2 and only weakly with temper- Figure 52. Effect of binary gas-phase diffusion on the
ature. (Reprinted with permission from ref 346. Copyright impedance characteristics of porous mixed-conducting elec-
2001 Elsevier.) trodes at low PO2: (a) zero-bias impedance of LSC on rare-
earth-doped ceria at 1 atm and 750 °C as a function of PO2
measured as a function of temperature, the polariza- using concentrations and balance gases as indicated.
tion associated with this arc is found to have a very (Reprinted with permission from ref 350. Copyright 2000
small activation energy, and (3) the very low fre- Elsevier B.V.) (b) Zero-bias impedance of SSC (x ) 0.5) on
quency of the arc, which can generally be understood SDC at 800 °C and PO2 ) 0.01 atm as a function of total
pressure. (Reprinted with permission from ref 184. Copy-
in terms of the relative chemical capacitances in- right 2001 Electrochemical Society, Inc.)
volved, as follows. As discussed in sections 3-5, the
chemical capacitance associated with the faradaic likely reflects a large utilization length since if the
impedance is proportional to the linearized storage active region is already a significant fraction of the
capacity of adsorbed oxygen on the electrode surface electrode bulk, any fluctuations in solid and gas
within a distance lδ of the interface. However, when composition resulting from gradients in gas concen-
gas-phase diffusion resistance is present, the relevant tration add little to the already large chemical
oxygen storage capacity involves the entire electrode capacitance.
surface (thickness L) as well as the storage capacity The results reviewed above suggest that gas-phase
of the gas itself (which may be particularly significant diffusion can contribute significantly to polarization
if there is a large stagnant layer outside the elec- as O2 concentrations as high as a few percent and
trode). Yoon et al. showed that gas-phase diffusion are not necessarily identifiable as a separate feature
in the electrode itself and mass-transfer limitations in the impedance. Workers studying the PO2-depen-
outside the electrode (caused by a flow-rate-depend- dence of the electrode kinetics are therefore urged
ent boundary layer) appear as separate impedance to eliminate as much external mass-transfer resis-
features due to the relatively large storage capacity tance in their experiments as possible and verify
of the gas above the cell.347 experimentally (using variations in balance gas or
In cases where the utilization region approaches total pressure) that gas-phase effects are not obscur-
the electrode thickness (e.g., perovskite mixed con- ing their results.
ductors), it can be harder to discern gas-phase effects (2) Cell geometry effects. Figure 53a illustrates a
as a separate feature since the time scales begin to typical cell configuration used by workers wishing to
merge. Some techniques workers have used in this isolate the polarization losses of a specific SOFC
situation include varying the total pressure while electrode. In this case, the cell consists of a thin (10-
keeping PO2 constant184 (since binary diffusion coef- 500 µm) electrolyte with active (current bearing)
ficients scale inversely with Ptotal at low pressure348) electrodes on each side. The reference electrode is
or replacing the diluent gas (usually N2 or Ar) with placed on the surface of the inactive (unelectroded)
He28,346,347 (since the binary diffusion coefficient for portion of the electrolyte, on one or both sides of the
O2 in He is about 4-5 times smaller than for O2 in cell, some distance away from the edge of the active
N2349). As an example, Figure 52a shows the imped- electrodes. Ideally, the purpose of this configuration
ance of LSC electrodes on rare-earth-doped ceria at is to measure the electrolyte potential along an
750 °C as a function of PO2 when a relatively thick equipotential surface “somewhere between the two
(∼20 µm) porous silver current collector is used. active electrodes”. If two reference electrodes are
Replacement of air with 20.9% O2 in He has little used, they will yield redundant information (differing
effect on the impedance at zero bias. However, there only in Nernst potential), provided they are placed a
is measurable difference between 4.0% O2 in N2 vs distance greater than ∼3 electrolyte thicknesses from
4.0% O2 in He. The higher resistance in N2 is the edge of the active electrodes and the gas on each
accompanied by a proportionate decrease in charac- side of the cell is well mixed.350 Assuming this
teristic frequency rather than the appearance of a idealization, the total cell voltage losses (Vtotal) can
second feature in the impedance. As shown in Figure be divided into two parts as shown in Figure 53b:
52b, Koyama et al. showed similar findings for SSC VA, which contains the overpotential of the first
on SDC at 800 °C and PO2 ) 0.01 atm by varying total electrode plus some fraction of the electrolyte poten-
pressure.184 The similarity in time scale for the tial drop, and VB, which contains the overpotential
chemical vs gas-transport impedances in these cases of the second electrode plus the remaining fraction
Figure 53. Idealized half-cell response of a thin solid

electrolyte cell. (a) Cell geometry including working elec-
trodes A and B and reference electrode(s). (b) Equivalent
circuit model for the cell in a, where the electrolyte and
two electrodes have area-specific resistances and capaci-
tances as indicated. (c) Total cell and half-cell impedance
responses, calculated assuming the reference electrode
remains equipotential with a planar surface located some-
where in the middle of the active region, halfway between
the two working electrodes, as shown in a. Figure 54. Measured (a) and simulated (b) effect of
electrode misalignment. (a) Total-cell and half-cell imped-
ances of a symmetric LSC/rare-earth-doped ceria/LSC cell
of the electrolyte potential losses. Assuming this with nominally identical porous LSC (x ) 0.4) electrodes,
division can be achieved, Figure 53c illustrates the measured at 750 °C in air based on the cell geometry
resulting separation of the total cell impedance into shown.350 (b) Finite-element calculation of the total-cell and
two ideal half-cell impedances, supposing (for the half-cell impedances of a symmetric cell with identical R-C
purposes of example) electrodes with simple linear electrodes, assuming a misalignment of the two working
electrodes (d) equal to the thickness of the electrolyte (L).319
RC behavior.
Although experimentalists often assume validity
of the idealization shown in Figure 53, numerous with nearly 80% of the iR resistance being ap-
studies have illustrated the difficulty of making portioned to electrode 2. The two half-cell impedances
accurate polarization measurements with this type differ in both shape and magnitude and exhibit
of configuration.319,350-358 Nagata et al.351 and Hsieh anomalous “capacitive” and “inductive” effects at high
et al.352 have shown experimentally that electrode frequency (that sum to zero in the total impedance).
geometries similar to Figure 53 can yield large In addition, despite high reproducibility of the total
inaccuracies in the measured steady-state electrode cell impedance from sample to sample, the scatter
overpotential and impedance. A number of workers in the half-cell response among samples is substan-
have since attempted to simulate these ef- tial.350 One explanation for these effects is illustrated
fects.319,350,353-355 In particular, Winkler et al.353 used in Figure 54b, which shows a finite-element simula-
the finite-element method to calculate the primary tion of the half-cell impedance for identical “RC”
(infinite frequency) and secondary (zero frequency) electrodes, misaligned by a distance d equal to the
d.c. potential distributions for various electrode thickness of the electrolyte (d/L) 1). This calculation
geometries and reference electrode locations. These suggests that the observations in Figure 54a may
calculations showed that a slight misalignment of the simply be caused by statistical variation in the
active electrodes on a thin cell can cause a frequency- alignment of the electrodes (which was subject to the
dependent displacement of the reference potential, limitations of a manually aligned screen printer in
leading to significant inaccuracies in the electrode this case).
resistance extracted from a.c. impedance measure- Furthermore, Boukamp358 and Adler319 showed
ments. Adler et al. corroborated these findings350 and that when the electrodes on opposite sides of a cell
more recently extended FEA analysis to include both are different from each other (as they are in a fuel
the real and imaginary response, allowing determi- cell), errors may not only involve a numerical scaling
nation of distortional effects on the impedance in factor but also cross-contamination of anode and
arbitrary geometries over the entire frequency spec- cathode frequency response in the measured half-cell
trum.319 impedances. For example, Figure 55a shows the
As an example, Figure 54a shows the zero-bias calculated half-cell impedance of the cell idealized in
impedance of LSC electrodes on rare-earth-doped Figure 53a, assuming alignment errors of (1 elec-
ceria in air at 750 °C measured using a symmetric trolyte thickness. Significant distortion of the half-
cell incorporating a traditional reference electrode.350 cell impedances (ZA and ZB) from the actual imped-
Although the two screen-printed electrodes (1 and 2) ance of the electrodes are apparent,319 including
were processed identically and aligned to an accuracy “cross-talk”358 of anode and cathode frequency re-
of ∼0.1 mm, the cell response is highly asymmetric sponse (1 and 10 Hz, respectively), as well as a
ing it difficult to provide a step-by-step prescription

to developers or researchers for the way forward.
Rather, here we simply attempt to summarize some
of the general conclusions that have emerged from
this large volume of work and also highlight areas
where future work is likely to make an impact.
One theme that appears to emerge repeatedly in
the literature reviewed above is that there is gener-
ally no one rate-determining step (RDS) on which to
focus our attention. This absence of a RDS appears
to arise from two fundamental factors. First, we saw
that for oxygen to be incorporated as ionic current
(O2-) at the electrode/electrolyte interface, O2 gas
must first be converted to some “electroactive” inter-
mediate form via one or more reaction steps, which
we labeled as chemical processes (as defined in
section 3.2). Since O2 is a relatively stable molecule
Figure 55. Simulated half-cell impedances of the cell
shown in Figure 53, calculated using finite-element analy- even at high temperature, it is perhaps not surprising
sis.319 (a) Half-cell responses assuming an electrode mis- that these chemical steps are often found to dominate
alignment d/L equal to (1, as defined in Figure 54c. (b) the electrode polarization, even when interfacial
Half-cell responses assuming perfect electrode alignment electrochemical kinetic resistances are significant.
(d/L ) 0). Given that the microstructure is usually a compro-
mise between improved surface area (lower firing
number of possible inductive artifacts. Figure 55b temperature) vs improved interfacial contact (higher
further shows that distortions of the impedance will firing temperature), we are often likely to find
occur even when alignment is perfect due to the optimized electrodes in a mixed regime where elec-
inherent asymmetry of the potential distribution at trochemical and chemical contributions to the over-
intermediate frequency.319 These results potentially potential are somewhat balanced.
call into question the interpretation of half-cell Second, we observed generically that the chemical
measurements on all thin electrolytes (particularly portion of the reaction mechanism involves an ad(b)-
very thin, supported electrolytes), not only in terms sorption step onto (or into) the electrode material,
of measurement error, but also possible misinterpre- followed by an ambipolar transport step on (or
tation of frequency artifacts as “steps” in the reaction within) the material to the interface (paths a-c-e
mechanism. or b-d-f in Figure 4). As shown in section 3.4, these
A number of different approaches have been dead(b)sorption and transport steps tend to be co-
veloped by workers to combat these problems. One limiting in a porous microstructure, meaning that
has simply been to avoid half-cell measurements both determine the overall rate over a wide range of
altogether, focusing instead on the response of sym- values of the relevant rate parameters. In section 6.4
metric cells at zero bias28 or fuel cells under load we further saw that for composite microstructures
where the anode impedance is either small or iden- (which most state-of-the-art cathodes are) the elec-
tifiable by frequency in the impedance spectrum.40 trode kinetics also become co-limited by ohmic losses
Another approach has been to use a very thick in the ionic subphase. Combining this observation
electrolyte40 (where alignment is less of an issue), with those in the preceding paragraph, we might
ideally incorporating a reference electrode near the expect an optimized composite electrode to be simul-
working electrode via a hole in the electrolyte.353 taneously limited by at least four physical pro-
According to FEA simulations, the thick pellet ap- cesses: ad(b)sorption of oxygen, ambipolar transport
proach appears to provide good accuracy353 as well to the solid-solid interface, interfacial electrochemi-
as frequency isolation319 but is generally limited to cal kinetics, and ionic transport in the ionic subphase.
lower current densities due to finite compliance Conceivably, improvements to any of these processes
voltage and Joule heating of the sample. Workers will help overall performance to some degree.
have also proposed various microelectrode geometries
that (in principle) should exhibit excellent frequency However, as mentioned in section 6, our awareness
isolation.319,359,360 Early experimental results with this of this situation is not the same as being able to
approach appear to be favorable,361 but more work quantify the contributions of these various physical
is needed to develop a truly robust and reliable way processes to the performance of a particular electrode
of doing half-cell measurements at arbitrary current under a specific set of conditions or in understanding
densities and frequencies. all the factors that govern the rates of these pro-
cesses. Unfortunately, due to the inherently convo-
luted nature of electrochemical and chemical pro-
7. Conclusions and Outlook cesses, it has proven extremely difficult to isolate and
The literature reviewed in sections 2-6 represents study these processes individually in a complex
significant advances made in the last 20 years in our system. We saw in sections 3-5 that impedance
understanding of solid oxide fuel-cell cathodes. At the techniques can in some cases be used to isolate the
same time, however, this work has also underscored linearized resistance of the interface from that of
how complex the oxygen reduction reaction is, mak- slower chemical steps via time scale. Various workers
have also shown that by accounting for the interfacial path. While this is just a hypothetical example, it
double-layer capacitance, one can often quantify the illustrates how understanding of the mechanism,
interfacial resistance even when electrochemical and combined with appropriate new fabrication tech-
chemical time scales begin to merge. As with all niques, might allow a more directed approach to
impedance studies, however, these analyses are electrode microstructural design.
limited to linearized conditions and subject to the Another neglected issue, from both a practical and
vagaries of modeling the interface as an equivalent fundamental perspective, is nonstationary behavior,
circuit. New techniques are needed that have an including degradation. Why electrode performances
ability to more unambiguously separate overlapping change with time, polarization history, or other
physical effects in both simple and complex micro- factors is very poorly understood but of significant
structures. concern to developers. SOFC stack design is very
Moreover, despite the many advances in electro- sensitive to the exact performance of the cell, and
chemical measurement and modeling, our under- these stacks are expected to have a service life of at
standing of SOFC cathode mechanisms remains least 5-10 years. Yet the reported performances and
largely circumstantial today. Our understanding degradation rates of SOFC cathodes vary so tremen-
often relies on having limited explanations for an dously from lab to lab that it is difficult to compare,
observed phenomenon (e.g., chemical capacitance as let alone choose, a material system without a sub-
evidence for bulk transport) rather than direct inde- stantial independent development effort. If some of
pendent measures of the mechanism (e.g., spectro- these seemingly “random” variables were better
scopic evidence of oxidation/reduction of the electrode identified, it would not only accelerate development
material). At various points in this review we saw but also likely aid fundamental research (by elimi-
that high-vacuum techniques commonly employed in nating factors that otherwise obscure what workers
electrocatalysis can be used in some limited cases for are trying to study).
SOFC materials and conditions (PEEM, for example). Finally, we have also seen substantial advances in
New in-situ analytical techniques are needed, par- electrode modeling in the last 5-10 years, including
ticularly which can be applied at ambient pressures, new techniques for nonlinear and time-dependent
that can probe what is happening in an electrode as phenomena. These efforts must continue in order to
a function of temperature, PO2, polarization, local achieve improved quantitative linkages among per-
position, and time. formance, microstructure, and materials properties.
Throughout the review, we also saw that authors However, as mentioned in section 4.7, there is gener-
have made substantial progress (particularly in the ally a dearth of independent property data for many
last 5-7 years) deconvoluting overlapping effects by of the physical processes entering these models. New
studying systems of controlled geometry or fabricated techniques are needed that can isolate select physical
microstructures. Hopefully this work will continue processes, particularly involving the surface of elec-
and employ new analytical techniques (such as tracer trode materials. For example, while bulk diffusion
incorporation or local probes of composition and/or in mixed conductors is relatively well understood, it
thermodynamic potentials) in order to make quan- has proven difficult to isolate the rate of surface
titative measurements of the rate-limiting steps diffusion independent of the bulk. Also, tracer tech-
involved. At the same time, it remains important to niques have allowed us to measure the linearized
extend these techniques to commercially relevant exchange rate of oxygen on a mixed conductor
microstructures both experimentally and theoreti- surface, but it remains largely unknown how the
cally. Advanced characterization methods such as rates of absorption/desorption depend on driving force
FIB-SEM, SIMS, and AFM, combined with the when substantial displacements from equilibrium are
many advances made in finite-element modeling, involved.
may make it increasingly possible to treat these more
complex cases. 8. Acknowledgments
One particularly neglected issue appears to be the
matching of the microstructure to the important The author would first like to acknowledge the
length scales governing the mechanism. Composite many workers who have contributed to this line of
electrodes are normally optimized according to the research over the last 20 years, whose work has been
simple paradigm that increased TPB area is a good cited herein. He would also like to thank several
thing. However, is this always the case? For example, individuals for helpful discussions and/or careful
consider a composite electrode consisting of doped readings of the manuscript: Scott Barnett, Roger
ceria and a perovskite mixed conductor. As men- deSouza, Raymond Gorte, Ronald Imbihl, Allan Ja-
tioned in section 6.4, we might expect the mixed cobson, Juergen Janek, Tatsuya Kawada, Joachim
conductor, even within the composite, to exhibit a Maier, Augustin McEvoy, Steve McIntosh, Ian Met-
“local” utilization length (lδlocal) determining the calfe, Andreas Mitterdorfer, John Newman, Steven
extension of the TPB beyond each interparticle Russek, Brian Steele, Eric Stuve, and Harry Tuller.
contact. If all particles are smaller than this distance, Preparation of the manuscript was also aided sub-
then the perovskite particles will be diffusionally stantially on a logistic level by Ladonna Kennedy,
equilibrated locally. In this situation we might not Lai Lu, and Ryan Reed. Finally, the author would
want a high TPB area but rather a very high surface also like to thank his department, students, and
area mixed conductor with more moderately sized family for their patience during the preparation and
ceria particles offering a very efficient ionic transport revision of the manuscript.
9. Nomenclature GDC gadolinia-doped ceria (Ce1-xGdx-

O2-x/2)
a (cm2/cm3) surface area per unit volume GDE gas-diffusion electrode
Ra, Rc (dimensionless) anodic and cathodic transfer coef- HRTEM high-resolution transmission elec-
ficients tron microscopy
Ci (F/cm2) capacitance associated with process LCM La1-xCaxMnO3
or lengthscale i LM LaxMnO3
ci (mol/cm3 or bulk or surface concentration of LSC La1-xSrxCoO3-δ
mol/cm2) species i LSCF La1-xSrxCo1-yFeyO3-δ
d (cm) diameter or other characteristic LSF La1-xSrxFeO3-δ
length LSGM La0.8Sr0.2Ga0.8Mg0.2O2.8
δ (dimensionless) oxygen nonstoichiometry in ABO3-δ LSM La1-xSrxMnO3(δ
D* (cm2/s) oxygen tracer diffusion coefficient LZ lanthanum zirconate (La2Zr2O7)
Di (cm2/s) single component diffusion coef- PEEM photoelectron emission microscopy
ficient for species i SEM scanning electron microscopy
D̃ (cm2/s) chemical (ambipolar) diffusion coef- SIMS secondary-ion mass spectrometry
ficient SOFC solid oxide fuel cell
(dimensionless) porosity SDC samaria-doped ceria (Ce1-xSmx-
φ (dimensionless) fraction of interface permitting pas- O2-x/2).
sage of current SSC Sm0.5Sr0.5CoO3-δ
f (dimensionless) a thermodynamic factor (- ∂µO2/∂δ)/ SZ strontium zirconate
RT TPB three-phase (or triple-phase) bound-
F (C/mol) Faraday’s constant (96,487 C/mol) ary
η (V) overpotential XPS X-ray photoelectron spectroscopy
I (A) total current XRD X-ray diffraction (powder)
i (A/cm2) current density YSZ yttria-stabilized zirconia (Zr1-xYx-
i0 (A/cm2) exchange current density O2-x/2)
J (dimensionless) ratio of ohmic constriction to inter-
facial resistance
k, k* (cm/s or s-1) surface exchange coefficient 11. Glossary of Some Commonly Used Terms
“k” (Ω-cm3) volume-specific interfacial resis- chemical (or a macroscopically observed capaci-
tance in a composite “pseudo”) tive relationship between current
L (cm) thickness capacitance and voltage arising from chemi-
lδ (cm) utilization length cal oxidation/reduction of a mate-
lg (cm) characteristic length of a porous rial or other charge-compensated
microstructure accumulation of species
µi (J/mol) chemical or electrochemical poten- chemical (or diffusion of neutral combinations of
tial of species i “ambipolar”) ionic and/or electronic species un-
Ni (mol/area/s) flux of species “i” diffusion der a chemical potential (concen-
PO2 (bar) partial pressure of oxygen tration) driving force
F (Ω-cm) resistivity chemical process a physical or chemical rate process
r0 (mol/cm2/s) equilibrium gas-surface exchange that may occur at a rate indepen-
rate dent of current and is driven by
ri (mol/cm2/s) area-specific reaction rate involving a chemical-potential driving force
species i co-limited reaction Situation in a porous gas-diffusion
R (J/mol/K) ideal gas constant (8.314 J/mol/K) electrode wherein the utilization
Ri (Ω-cm2) area-specific resistance associated length and steady-state flow of
with process i current depends on both kinetic
σi (Ω-1 cm-1) ionic conductivity and transport properties
T (K or °C) temperature electrochemical a kinetic process involving a net
t (s) time kinetic process faradaic current across an inter-
τ (dimensionless) tortuosity face and is driven by an electro-
ti (dimensionless) transference number for species i chemical potential driving force
V (V) voltage mixed conductor a material containing both mobile
Vm (cm3/mol) molar volume (on a formula-unit ions and mobile electrons (includ-
basis) ing holes and/or small polarons)
ω (s-1) frequency in radians three-phase (or the one-dimensional interface
x,y (position) (cm) position “triple-phase”) within an electrode microstruc-
x,y (stoich) A-site or B-site stoichiometry pa- boundary (TPB) ture where gaseous, electronic,
(dimensionless) rameters and ionic subphases meet
Z (Ω or Ω-cm2) impedance utilization length (lδ) size of the active region within a
porous electrode, established by
10. Glossary of Commonly Used Acronyms a balance of reaction and trans-
AFM atomic force microscopy port rates at steady state (see co-
CDC Ca-doped ceria (Ce1-xCaxO2-x/2) limited reaction)
ESB erbia-doped bismuth oxide
EIS electrochemical impedance spectros- 12. References
copy
(1) Alcock, C. B. Solid State Ionics 1992, 53-56, 3.
FEA finite-element analysis (2) Kilner, J.; Benson, S.; Lane, J.; Waller, D. Chem. Ind. 1997, 907.
FIB-SEM SEM cross-sectional imaging using (3) Mazanec, T. J. Solid State Ionics 1994, 70/71, 11.
focused ion beams (4) Mobius, H. H. J. Solid State Electrochem. 1997, 1, 2.
(5) Singhal, S. C. Proc. Electrochem. Soc. 1998, 97-24, 125. (59) Mizusaki, J.; Amano, K.; Yamauchi, S.; Fueki, K. Solid State
(6) Badwal, S. P. S.; Foger, K. Mater. Forum 1997, 21, 187. Ionics 1987, 22, 313.
(7) Yamamoto, O. Electrochim. Acta 2000, 45, 2423. (60) Mizusaki, J.; Amano, K.; Yamauchi, S.; Fueki, K. Solid State
(8) Steele, B. C. H. Solid State Ionics 2000, 134, 3. Ionics 1987, 22, 323.
(9) Skinner, S. J. Int. J. Inorg. Mater. 2001, 3, 113. (61) Wang, D. Y. J. Electrochem. Soc. 1990, 137, 3660.
(10) Will, J.; Mitterdorfer, A.; Kleinlogel, C.; Perednis, D.; Gauckler, (62) van Hassel, B. A.; Boukamp, B. A.; Burggraaf, A. J. Solid State
L. J. Solid State Ionics 2000, 131, 79. Ionics 1991, 48, 139.
(11) McEvoy, A. J. J. Mater. Sci. 2001, 36, 1087. (63) Robertson, N. L.; Michaels, J. N. J. Electrochem. Soc. 1990, 137,
(12) Ormerod, R. M. Chem. Soc. Rev. 2003, 32, 17. 129.
(13) Yamamoto, O.; Takeda, Y.; Kanno, R.; Tomida, Y. Nippon (64) van Herle, J.; McEvoy, A. J. Ber. Bunsen-Ges.-Phys. Chem. Chem.
Kagaku Kaishi 1988, 1324. Phys. 1993, 97, 470.
(14) Godickemeier, M.; Sasaki, K.; Gauckler, L. J.; Riess, I. Solid (65) Schwandt, C.; Weppner, W. J. Electrochem. Soc. 1997, 144, 3728.
State Ionics 1996, 86-88, 691. (66) Kenjo, T.; Shiroichi, N. Electrochim. Acta 1997, 42, 3461.
(15) Kilner, J. A. Bol. Soc. Esp. Ceram. Vidrio 1998, 37, 247. (67) Luerssen, B.; Janek, J.; Imbihl, R. Solid State Ionics 2001, 141,
(16) Steele, B. C. H.; Hori, K. M.; Uchino, S. Solid State Ionics 2000, 701.
135, 445. (68) Janek, J.; Rohnke, M.; Luerssen, B.; Imbihl, R. Phys. Chem.
(17) Godickemeier, M.; Sasaki, K.; Gauckler, L. J.; Riess, I. J. Chem. Phys. 2000, 2, 1935.
Electrochem. Soc. 1997, 144, 1635. (69) Boukamp, B. A.; Bouwmeester, H. J. M. Solid State Ionics 2003,
(18) Lane, J. A.; Adler, S.; Middleton, P. H.; Steele, B. C. H. Solid 157, 29.
Oxide Fuel Cells (SOFC-IV); 1995; pp 584-96. (70) Kamata, H.; Hosaka, A.; Mizusaki, J.; Tagawa, H. Solid State
(19) Maguire, E.; Gharbage, B.; Marques, F. M. B.; Labrincha, J. A. Ionics 1998, 106, 237.
Solid State Ionics 2000, 127, 329. (71) Gerischer, H. Z. Phys. Chem. 1951, 198, 216.
(20) Rohland, B. Mater. Sci. Forum 1991, 76, 149. (72) Mitterdorfer, A.; Gauckler, L. J. Solid State Ionics 1999, 117,
(21) Takahashi, T.; Ihara, M.; Yamada, K. Proc. Electrochem. Soc. 187.
1995, 95-1, 1009. (73) Mitterdorfer, A.; Gauckler, L. J. Solid State Ionics 1999, 117,
(22) Colomer, M. T.; Steele, B. C. H.; Kilner, J. A. Solid State Ionics 203.
2002, 147, 41. (74) Mitterdorfer, A.; Gauckler, L. J. Solid State Ionics 1999, 120,
(23) Steele, B. C. H. Solid State Ionics 1997, 94, 239. 211.
(24) Dusastre, V.; Kilner, J. A. Solid State Ionics 1999, 126, 163. (75) Berthier, F.; Diard, J.-P.; Le Gorrec, B.; Montella, C. Corrosion
(25) Schafer, W.; Koch, A.; HeroldSchmidt, U.; Stolten, D. Solid State 1995, 51, 105.
Ionics 1996, 86-8, 1235. (76) Gland, J. L.; Sexton, B. A.; Fisher, G. B. Surf. Sci. 1980, 95,
(26) Liu, W.; Zhang, Z. P.; Xia, C. R.; Xie, J. Q.; Chen, C. S. J. Inorg. 587.
Mater. 2000, 15, 849. (77) Gorte, R.; Schmidt, L. D. Surf. Sci. 1978, 76, 559.
(27) Endo, A.; Fukunaga, H.; Wen, C.; Yamada, K. Solid State Ionics (78) Lewis, R.; Gomer, R. Surf. Sci. 1968, 12, 157.
2000, 135, 353. (79) Verkerk, M. J.; Hammink, M. W. J.; Burggraaf, A. J. J.
(28) Adler, S. B. Solid State Ionics 1998, 111, 125. Electrochem. Soc. 1983, 130.
(29) Murray, E. P.; Barnett, S. A. Proc. Electrochem. Soc. 1999, 99- (80) Kenjo, T.; Yamakoshi, Y.; Wada, K. J. Electrochem. Soc. 1993,
19, 369. 140, 2151.
(30) Jorgensen, M. J.; Primdahl, S.; Bagger, C.; Mogensen, M. Solid (81) Sasaki, K.; Tamura, J.; Dokiya, M. Solid State Ionics 2001, 144,
State Ionics 2001, 139, 1. 223.
(31) Jorgensen, M. J.; Mogensen, M. J. Electrochem. Soc. 2001, 148, (82) Widmer, S.; Tate, T. J.; Thampi, K. R.; McEvoy, A. J. Solid Oxide
A433. Fuel Cells V; Aachen: Germany, 1997; pp 451-461.
(32) Lu, Z.-G.; Jiang, Y.; Dong, Y.-L.; Zhang, Y.-H.; Yan, J.-W. (83) McEvoy, A. J. Solid State Ioncs 2000, 132, 159.
Gaodeng Xuexiao Huaxue Xuebao 2001, 22, 791. (84) Mitterdorfer, A. Ph.D., Swiss Federal Institute of Technology,
(33) Murray, E. P.; Barnett, S. A. Solid State Ionics 2001, 143, 265. 1997.
(34) Barbucci, A.; Bozzo, R.; Cerisola, G.; Costamagna, P. Electrochim. (85) Robertson, N. L.; Michaels, J. N. J. Electrochem. Soc. 1991, 138,
Acta 2002, 47, 2183. 1494.
(35) Murray, E. P.; Sever, M. J.; Barnett, S. A. Solid State Ionics (86) Kenjo, T.; Tsukamoto, K. Solid Oxide Fuel Cells V; Aachen:
2002, 148, 27. Germany, 1997; pp 431-440.
(36) Hart, N. T.; Brandon, N. P.; Day, M. J.; Lapena-Rey, N. J. Power (87) Schouler, E. J. L.; Kleitz, M. J. Electrochem. Soc. 1987, 134, 1045.
Sources 2002, 106, 42. (88) Chao, T.; Walsh, K. J.; Fedkiw, P. S. Solid State Ionics 1991,
(37) Steele, B. C. R. Acad. Sci. Serie Ii Fascicule C: Chim. 1998, 1, 47, 277.
533. (89) Sridhar, S.; Stancovski, V.; Pal, U. B. J. Electrochem. Soc. 1997,
(38) Kleitz, M.; Petitbon, F. Solid State Ionics 1996, 92, 65. 144, 2479.
(39) Jorgensen, M. J.; Holtappels, P.; Appel, C. C. J. Appl. Electro- (90) Jacobsen, T.; Zachau-Christiansen, B.; Bay, L.; Jorgensen, M.
chem. 2000, 30, 411. J. Electrochim. Acta 2001, 46, 1019.
(40) McIntosh, S.; Adler, S. B.; M, V. J.; Gorte, R. A. Electrochem. (91) Shkerin, S. N.; Perfilev, M. V. Elektrokhimiya 1990, 26, 1468.
Solid State Lett. 2004, 7, A111. (92) Shkerin, S. N.; Perfilev, M. V. Soviet Electrochem. 1992, 28, 1106.
(41) Schmid, A., Enke, F., Eds.; Stuttgart, 1923. (93) von Oertzen, A.; Mikhailov, A.; Rotermund, H. H.; Ertl, G. Surf.
(42) Vetter, K. J. Electrochemical Kinetics, Theoretical and Experi- Sci. 1996, 350, 259.
mental Aspects; Academic Press: New York, 1967. (94) Emery, D. A.; Middleton, P. H.; Metcalfe, I. S. Surf. Sci. 1998,
(43) Srinivasan, S.; Hurwitz, H. D.; Bockris, J. O. M. J. Phys. Chem. 405, 308.
1967, 46, 3108. (95) Bay, L.; Jacobsen, T. Solid State Ionics 1997, 93, 201.
(44) Giner, J.; Hunter, C. J. Electrochem. Soc. 1969, 116, 1125. (96) Svensson, A. M.; Nisancioglu, K. J. Electrochem. Soc. 1998, 145,
(45) Perry, M. L.; Newman, J.; Cairns, E. J. J. Electrochem. Soc. 1998, 3130.
145, 5. (97) Luerssen, B.; Janek, J.; Günther, S.; Kiskinova, M.; Imbihl, R.
(46) Kleitz, M.; Kloidt, T.; Dessemond, L. High-Temperature Elec- Phys. Chem. Chem. Phys. 2002, 4, 2673.
trochemical Behavior of Fast Ion and Mixed Conductors; (98) Kleitz, M.; Dessemond, L.; Kloidt, T. Solid Oxide Fuel Cells IV;
Roskilde: Denmark, 1993; p 89. 1995; pp 35-44.
(47) Newman, J. S. Electrochemical Systems, 2nd ed.; Prentice Hall: (99) van Hassel, B. A.; Boukamp, B. A.; Burggraaf, A. J. Solid State
New York, 1991. Ionics 1992, 51, 161.
(48) Bauerle, J. E. J. Phys. Chem.: Solids 1969, 30, 2657. (100) Ohno, Y.; Nagata, S.; Sato, H. Solid State Ionics 1981, 3/4, 439.
(49) MacDonald, J. R. Impedance Spectroscopy; Emphasizing Solid (101) Takeda, Y.; Kanno, R.; Noda, M.; Yamamoto, O. Bull. Inst. Chem.
Materials and Systems; John Wiley and Sons: New York, 1987. Res. 1986, 64, 157.
(50) Wang, D. Y.; Nowick, A. S. J. Electrochem. Soc. 1979, 126, 1166. (102) Raccah, P. M.; Goodenough, J. B. Phys. Rev. 1967, 155, 932.
(51) Wang, D. Y.; Nowick, A. S. J. Electrochem. Soc. 1979, 126, 1155. (103) Jaya, S. M.; Jagadish, R.; Rao, R. S.; Asokamani, R. Phys. Rev.
(52) Okamoto, H.; Kawamura, G.; Kudo, T. Electrochim. Acta 1983, B 1991, 43, 13274.
28, 379. (104) Ishikawa, T.; Park, S. K.; Katsufuji, T.; Arima, T.; Tokura, Y.
(53) Hu, H.; Liu, M. J. Electrochem. Soc. 1997, 144, 3561. Phys. Rev. B 1998, 58, 13326.
(54) Schouler, E.; Giroud, G.; Kleitz, M. J. Chim. Phys. 1973, 70, (105) Raccah, P. M.; Goodenough, J. B. J. Appl. Phys. 1968, 39, 1209.
1309. (106) Bhide, V. G.; Rajoria, D. S.; Reddy, Y. S.; Rao, G. R.; Rao, G. V.
(55) Kleitz, M.; Dessemond, L.; Kloidt, T.; Steil, M. C. Solid Oxide S.; Rao, C. N. R. Phys. Rev. Lett. 1972, 27, 1133.
Fuel Cells IV; 1995; pp 527-536. (107) Bhide, V. G.; Rajoria, D. S.; Rao, C. N. R.; Rao, G. R.; Jadhao,
(56) van Herle, J.; McEvoy, A. J. J. Phys. Chem. Solids 1994, 55, V. G. Phys. Rev. B 1975, 12, 2832.
339. (108) Ganguly, P.; Kumar, P. S. A.; Santhosh, P. N.; Mulla, I. S. J.
(57) Verkerk, M. J.; Burggraaf, A. J. J. Electrochem. Soc. 1983, 130, Phys.: Condens. Matter 1994, 6, 533.
78. (109) Senaris-Rodriguez, M. A.; Goodenough, J. B. J. Solid State Chem.
(58) Lewis, R.; Gomer, R. Surf. Sci. 1968, 12, 157. 1995, 118, 323.
(110) Takahashi, H.; Munakata, F.; Yamanaka, M. Phys. Rev. B 1998, (155) Gharbage, B.; Pagnier, T.; Hammou, A. J. Electrochem. Soc.
57, 15211. 1994, 141, 2118.
(111) Caciuffo, R.; Mira, J.; Rivas, J.; Seneris-Rodriguez, M. A.; (156) Van Herle, J.; McEvoy, A. J.; Thampi, K. R. Electrochim. Acta
Radaelli, P. G.; Carsughi, F.; Fiorani, D.; Goodenough, J. B. 1994, 39, 1675.
Europhys. Lett. 1999, 45, 399. (157) Mizusaki, J.; Saito, T.; Tagawa, H. J. Electrochem. Soc. 1996,
(112) Mizusaki, J.; Tabuchi, J.; Matsuura, T.; Yamauchi, S.; Fueki, 143, 3065.
K. J. Electrochem. Soc. 1987, 136, 2082. (158) Masuda, K.; Kawada, T.; Kaimai, A.; Kawamura, K.; Nigara,
(113) Petrov, A. N.; Kononchuk, O. F.; Andreev, A. V.; Cherepanov, Y.; Mizusaki, J.; Yugami, H.; Arashi, H. 5th International
V. A.; Kofstad, P. Solid State Ionics 1995, 80, 189. Symposium on Solid Oxide Fuel Cells; Achen: Germany, 1997;
(114) Chainani, A.; Mathew, M.; Sarma, D. D. Phys. Rev. B 1993, 48, p 473.
14818. (159) Ioroi, T.; Hara, T.; Uchimoto, Y.; Ogumi, Z.; Takehara, Z. J.
(115) Torrance, J. B.; Lacorre, P.; Asavaroengchai, C.; Metzger, R. M. Electrochem. Soc. 1997, 144, 1362.
Physica C 1991, 182, 351. (160) Ioroi, T.; Hara, T.; Uchimoto, Y.; Ogumi, Z.; Takehara, Z. J.
(116) Mizusaki, J.; Sasamoto, T.; Cannon, W. R.; Bowen, H. K. J. Am. Electrochem. Soc. 1998, 145, 1999.
Ceram. Soc. 1983, 66, 247. (161) Kawada, T.; Masuda, K.; Suzuki, J.; Kaimai, A.; Kawamura, K.;
(117) Tai, L.-W.; Nasrallah, M. M.; Anderson, H. U.; Sparlin, D. M.; Nigara, Y.; Mizusaki, J.; Yugami, H.; Arashi, H.; Sakai, N.;
Sehlin, S. R. Solid State Ionics 1995, 76, 259. Yokokawa, H. Solid State Ionics 1999, 121, 271.
(118) Stevenson, J. W.; Armstrong, T. R.; Carneim, R. D.; Pederson, (162) Yang, Y. L.; Chen, C. L.; Chen, S. Y.; Chu, C. W.; Jacobson, A.
L. R.; Weber, W. J. J. Electrochem. Soc. 1996, 143, 2722. J. J. Electrochem. Soc. 2000, 147, 4001.
(119) Mizusaki, J.; Yoshihiro, M.; Yamauchi, S.; Fueki, K. J. Solid (163) Fukunaga, H.; Koyama, M.; Takahashi, N.; Wen, C.; Yamada,
State Chem. 1985, 58, 257. K. Solid State Ionics 2000, 132, 279.
(120) Allnatt, A. R.; Lidiard, A. B. Atomic Transport in Solids; (164) Mims, C. A.; Joos, N. I.; van der Heide, P. A. W.; Jacobson, A.
Cambridge University Press: New York, 1993. J.; Chen, C.; Chu, C. W.; Kim, B. I.; Perry, S. S. Electrochem.
(121) Mizusaki, J.; Mima, Y.; Yamauchi, S.; Fueki, K.; Tagawa, H. J. Solid State Lett. 2000, 3, 59.
Solid State Chem. 1989, 80, 102. (165) Yang, Y.; Jacobson, A. J.; Chen, C. L.; Luo, G. P.; Ross, K. D.;
(122) Petrov, A. N.; Cherepanov, V. A.; Kononchuk, O. F.; Gavrilova, Chu, C. W. Appl. Phys. Lett. 2001, 79, 776.
L. Y. J. Solid State Chem. 1990, 87, 69. (166) Ringuede, A.; Fouletier, J. Solid State Ionics 2001, 139, 167.
(123) Lankhorst, M. H. R.; Bouwmeester, H. J. M. J. Electrochem. Soc. (167) Endo, A.; Ihara, M.; Komiyama, H.; Yamada, K. Solid State
1997, 144, 1268. Ionics 1996, 86-88, 1191.
(124) Kawada, T.; Suzuki, J.; Sase, M.; Kaimai, A.; Yashiro, K.; Nigara, (168) Adler, S. B. Solid State Ionics 2000, 135, 603.
Y.; Mizusaki, J.; Kawamura, K.; Yugami, H. J. Electrochem. Soc. (169) Liu, M.; Winnick, J. J. Electrochem. Soc. 1997, 144, 1881.
2002, 149, E252. (170) Liu, M.; Winnick, J. Solid State Ionics 1999, 118, 11.
(125) Lankhorst, M. H. R.; Bouwmeester, H. J. M.; Verweij, H. Phys. (171) Adler, S. B.; Lane, J. A.; Steele, B. C. H. J. Electrochem. Soc.
Rev. Lett. 1996, 77, 2989. 1996, 143, 3554.
(126) Lankhorst, M. H. R.; Bouwmeester, H. J. M.; Verweij, H. J. Solid (172) Svensson, A. M.; Sunde, S.; Nisancioglu, K. Solid State Ionics
State Chem. 1997, 133, 555. 1996, 86-8, 1211.
(127) Lankhorst, M. H. R.; Bouwmeester, H. J. M.; Verweij, H. Solid (173) Abeles, B. Proc. Electrochem. Soc. 1997, 96-27, 1.
State Ionics 1997, 96, 21. (174) Deng, H.; Zhou, M.; Abeles, B. Solid State Ionics 1994, 74, 75.
(128) Tai, L.-W.; Nasrallah, M. M.; Anderson, H. U.; Sparlin, D. M.; (175) Bae, J. M.; Steele, B. C. H. Solid State Ionics 1998, 106, 255.
Sehlin, S. R. Solid State Ionics 1995, 76, 273. (176) Zhou, M.; Deng, H.; Abeles, B. Solid State Ionics 1996, 93, 133.
(129) Lankhorst, M. Univerdsity of Twente, 1997. (177) Sirman, J.; Lane, J.; Kilner, J. A. Electrochemical Society
(130) Lankhorst, M. H. R.; ten Elshof, J. E. J. Solid State Chem. 1997, Meeting, 1997; p 2129.
130, 302. (178) Steele, B. C. H.; Bae, J.-M. Solid State Ionics 1998, 106, 255.
(131) Teraoka, Y.; Zhang, H. M.; Okamoto, K.; Yamazoe, N. Mater. (179) Endo, A.; Wada, S.; Wen, C.-J.; Komiyama, H.; Yamada, K. J.
Res. Bull. 1988, 23, 51. Electrochem. Soc. 1998, 145, L35.
(132) Teraoka, Y.; Nobunaga, T.; Okamoto, K.; Miura, N.; Yamazoe, (180) Horita, T.; Yamaji, K.; Sakai, N.; Yokokawa, H.; Weber, A.; Ivers-
N. Solid State Ionics 1991, 48, 207. Tiffee, E. J. Electrochem. Soc. 2001, 148, A456.
(133) Diethelm, S.; Closset, A.; Nisancioglu, K.; Van herle, J.; McEvoy, (181) Horita, T.; Yamaji, K.; Sakai, N.; Yokokawa, H.; Weber, A.; Ivers-
A. J.; Gur, T. M. J. Electrochem. Soc. 1999, 146, 2606. Tiffee, E. Electrochim. Acta 2001, 46, 1837.
(134) Kilner, J. A.; Steele, B. C. H.; Ilkov, L. Solid State Ionics 1984, (182) Jiang, S. P. Solid State Ionics 2002, 146, 1.
12, 89. (183) Fleig, J. Ann. Rev. Mater. Res. 2003, 33, 361.
(135) Ishigaki, T.; Yamauchi, S.; Kishio, K.; Mizusaki, J.; Fueki, K. J. (184) Koyama, M.; Wen, C. J.; Masuyama, T.; Otomo, J.; Fukunaga,
Solid State Chem. 1988, 73, 179. H.; Yamada, K.; Eguchi, K.; Takahashi, H. J. Electrochem. Soc.
(136) Bouwmeester, H. J. M.; Kruidhof, H.; Burggraaf, A. J. Solid State 2001, 148, A795.
Ionics 1994, 72, 185. (185) Svensson, A. M.; Sunde, S.; Nisancioglu, K. J. Electrochem. Soc.
(137) Kilner, J. A.; Souza, R. A. D.; Fullarton, I. C. Solid State Ionics 1997, 144, 2719.
1996, 86-88, 703. (186) Svensson, A. M.; Sunde, S.; Nisancioglu, K. J. Electrochem. Soc.
(138) Routbort, J. L.; Doshi, R.; Krumpelt, M. Solid State Ionics 1996, 1998, 145, 1390.
90, 21. (187) Liu, M.; Wu, Z. Solid State Ionics 1998, 107, 105.
(139) van Doorn, R. H. E.; Fullarton, I. C.; de Souza, R. A.; Kilner, J. (188) Liu, M. L. J. Electrochem. Soc. 1998, 145, 142.
A.; Bouwmeester, H. J. M.; Burggraaf, A. J. Solid State Ionics (189) Adler, S. B.; Lane, J. A.; Steele, B. C. H. J. Electrochem. Soc.
1997, 96, 1. 1997, 144, 1884.
(140) De Souza, R. A.; Kilner, J. A. Solid State Ionics 1998, 106, 175. (190) van Doorn, R. H. E. University of Twente, 1996.
(141) De Souza, R. A.; Kilner, J. A. Solid State Ionics 1999, 126, 153. (191) van Doorn, R. H. E.; Burggraaf, A. J. Solid State Ionics 2000,
(142) Bredesen, R.; Mertins, F.; Norby, T. Catal. Today 2000, 56, 315. 128, 65.
(143) Ma, B.; Balachandran, U.; Park, J. H.; Segre, C. U. Solid State (192) Fleig, J. J. Power Sources 2002, 105, 228.
Ionics 1996, 83, 65. (193) Fleig, J.; Maier, J. 203 Meeting of the Electrochemical Society,
(144) ten Elshof, J. E.; Lankhorst, M. H. R.; Bouwmeester, H. J. M. Paris, 2003; p 509.
Solid State Ionics 1997, 99, 15. (194) Fleig, J. Annu. Rev. Matsci 2003.
(145) ten Elshof, J. E.; Lankhorst, M. H. R.; Bouwmeester, H. J. M. (195) Maier, J. Solid State Ionics 1998, 112, 197.
J. Electrochem. Soc. 1997, 144, 1060. (196) Maier, J. preprint 1999, 1.
(146) Lane, J. A.; Benson, S. J.; Waller, D.; Kilner, J. A. Solid State (197) Maier, J.; Jamnik, J.; Leonhardt, M. Solid State Ionics 2000,
Ionics 1999, 121, 201. 129, 25.
(147) Kriegel, R.; Pippardt, U.; Voigt, I. Sep. Purif. Technol. 2001, 25, (198) Kuznecov, M.; Otschik, P.; Eichler, K.; Schaffrath, W. Ber.
127. Bunsen-Ges.-Phys. Chem. Chem. Phys. 1998, 102, 1410.
(148) Chen, X.; Adler, S. B. Manuscript in preparation. (199) Coffey, G. W.; Pederson, L. R.; Rieke, P. C. J. Electrochem. Soc.
(149) Fueki, K.; Mizusaki, J.; Yamauchi, S.; Ishigaki, T.; Mima, Y. 2003, 150, A1139.
10th International Symposium on the Reactivity of Solids; pp (200) Atkinson, A.; Ramos, T. Solid State Ionics 2000, 129, 259.
339-343. (201) Adler, S. B. J. Am. Ceram. Soc. 2001, 84, 2117.
(150) ten Elshof, J. E.; Bouwmeester, H. J. M.; Verweij, H. Solid State (202) Mizusaki, J.; Tagawa, H.; Naraya, K.; Sasamoto, T. Solid State
Ionics 1995, 81, 97. Ionics 1991, 49, 111.
(151) ten Elshof, J. E.; Bouwmeester, H. J. M.; Verweij, H. Solid State (203) Takeda, Y.; Kanno, R.; Noda, M.; Tomida, Y.; Yamamoto, O. J.
Ionics 1996, 89, 81. Electrochem. Soc.: Electrochem. Sci. Technol. 1987, 134, 2656.
(152) Figueiredo, F. M.; Marques, F. M. B.; Frade, J. R. Solid State (204) Mizusaki, J.; Tagawa, H.; Tsuneyoshi, K.; Sawata, A. J. Elec-
Ionics 1998, 111, 273. trochem. Soc. 1991, 138, 1867.
(153) Kim, S.; Yang, Y. L.; Jacobson, A. J.; Abeles, B. Solid State Ionics (205) Nagata, M.; Hotta, H.; Iwahara, H. J. Appl. Electrochem. 1994,
1998, 106, 189. 24, 411.
(154) Zhang, C.; Deng, H.; Varon, J.; Abeles, B.; Yang, Y.; Pham, A. (206) Sasaki, K.; Wurth, J. P.; Gschwend, R.; Godickemeier, M.;
Q.; Jacobson, A. J. MRS 1994, 1. Gauckler, L. J. J. Electrochem. Soc. 1996, 143, 530.
(207) Fukunaga, H.; Ihara, M.; Sakaki, K.; Yamada, K. Solid State (254) Labrincha, J. A.; Frade, J. R.; Marques, F. M. B. J. Mater. Sci.
Ionics 1996, 86-88, 1179. 1993, 28, 3809.
(208) van Heuveln, F. H.; Bouwmeester, H. J. M.; vanBerkel, F. P. F. (255) Poulsen, F. W.; Vanderpuil, N. Solid State Ionics 1992, 53-6,
J. Electrochem. Soc. 1997, 144, 126. 777.
(209) van Heuveln, F. H.; Bouwmeester, H. J. M. J. Electrochem. Soc. (256) Vanroosmalen, J. A. M.; Cordfunke, E. H. P. Solid State Ionics
1997, 144, 134. 1992, 52, 303.
(210) Ostergard, M. J. L.; Mogensen, M. Electrochim. Acta 1993, 38, (257) Yamamoto, O.; Takeda, Y.; Kanno, R.; Kojima, T. First Inter-
2015. national Symposium on Solid Oxide Fuel Cells, 1989; p 242.
(211) Siebert, E.; Hammouche, A.; Kleitz, M. Electrochim. Acta 1995, (258) Ostergard, M. J. L.; Clausen, C.; Bagger, C.; Mogensen, M.
40, 1741. Electrochim. Acta 1995, 40, 1971.
(212) Lauret, H.; Hammou, A. J. Eur. Ceram. Soc. 1996, 16, 447. (259) Stochniol, G.; Syskakis, E.; Naoumidis, A. J. Am. Ceram. Soc.
(213) Lee, H. Y.; Cho, W. S.; Oh, S. M.; Wiemhofer, H. D.; Gopel, W. 1995, 78, 929.
J. Electrochem. Soc. 1995, 142, 2659. (260) Stochniol, G.; Broel, S.; Naoumidis, A.; Nickel, H. Fresenius J.
(214) Kuznecov, M.; Otschik, P.; Obenaus, P.; Eichler, K.; Schaffrath, Anal. Chem. 1996, 355, 697.
W. Solid State Ionics 2003, 157, 371. (261) Kostogloudis, G. C.; Ftikos, C. Solid State Ionics 1999, 126, 143.
(215) Mitterdorfer, A.; Gauckler, L. J. Solid State Ionics 1998, 111, (262) Yokokawa, H.; Horita, T.; Sakai, N.; Kawada, T.; Dokiya, M.
185. First European Solid Oxide Fuel Cell Forum, Baden-Baden,
(216) Poulsen, F. W. Solid State Ionics 2000, 129, 145. 1994; p 425.
(217) Chakraborty, A.; Choudhury, P.; Maiti, H. S. Proc. Electrochem. (263) Horita, T.; Yamaji, K.; Negishi, H.; Sakai, N.; Yokokawa, H.;
Soc. 1995, 95-1, 612. Kato, T. Solid State Ionics 2000, 136-137, 897.
(218) Kim, M. C.; Park, S. J. Yoop Hakhoechi 1992, 29, 900. (264) Clausen, C.; Bagger, C.; Bildesorensen, J. B.; Horsewell, A. Solid
(219) Li, Z.; Behruzi, M.; Fuerst, L.; Stoever, D. Proc. Electrochem. State Ionics 1994, 70, 59.
Soc. 1993, 93-4, 171. (265) Mitsuyasu, H.; Eguchi, K.; Arai, H. Solid State Ionics 1997, 100,
(220) Hammouche, A.; Siebert, E.; Hammou, A.; Kleitz, M.; Caneiro, 11.
A. J. Electrochem. Soc. 1991, 138, 1212. (266) Wiik, K.; Schmidt, C. R.; Shamsili, S.; Einarsrud, M. A.; Grande,
(221) Tofield, B. C.; Scott, W. R. J. Solid State Chem. 1974, 10, 183. T. High Temp. Electrochem.: Ceram. Met., Proc. Risoe Int. Symp.
(222) Yokokawa, H.; Horita, T.; Sakai, N.; Dokiya, M.; Kawada, T. Mater. Sci., 17th 1996, 491-496.
Solid State Ionics 1996, 86-88, 1161. (267) Wiik, K.; Schmidt, C. R.; Faaland, S.; Shamsili, S.; Einarsrud,
(223) Tanasescu, S.; Totir, N. D.; Marchidan, D. I. Electrochim. Acta M.-A.; Grande, T. J. Am. Ceram. Soc. 1999, 82, 721.
1998, 43, 1675. (268) Bertrand, G. L.; Caboche, G.; Domenichini, B.; Dufour, L. C. Surf.
(224) Hammouche, A.; Siebert, E.; Kleitz, M.; Hammou, A. Proc. Interface Anal. 2000, 30, 561.
Electrochem. Soc. 1989, 89-11, 265. (269) Kuscer, D.; Holc, J.; Hrovat, M.; Bernik, S.; Samardzija, Z.; Kolar,
(225) Hammouche, A.; Siebert, E.; Hammou, A.; Kleitz, M.; Caneiro, D. Solid State Ionics 1995, 78, 79.
A. J. Electrochem. Soc. 1991, 138, 1212. (270) Bae, J. M.; Steele, B. C. H. Solid State Ionics 1998, 106, 247.
(226) Jiang, S. P.; Love, J. G. Solid State Ionics 2001, 138, 183. (271) Dusastre, V.; Steele, B. C. H. Personal communication, 1997.
(227) Horita, T.; Yamaji, K.; Sakai, N.; Yokokawa, H.; Kato, T. J. (272) Appel, C. C.; Bonanos, N. J. Eur. Ceram. Soc. 1999, 19, 847.
Electrochem. Soc. 2001, 148, J25. (273) Uchikoshi, T.; Sakka, Y.; Hiraga, K. J. Electrocer. 1999, 4, 113.
(228) Brichzin, V.; Fleig, J.; Habermeier, H. U.; Cristiani, G.; Maier, (274) De Souza, M. F.; De Souza, D. P. F. J. Mater. Sci. 1999, 34, 6107.
J. Solid State Ionics 2002, 152-153, 499. (275) Steele, B. C. H. Solid State Ionics 2000, 129, 95.
(229) Horita, T.; Yamaji, K.; Ishikawa, M.; Sakai, N.; Yokokawa, H.; (276) Kawada, T.; Sakai, N.; Yokokawa, H.; Dokiya, M. Solid State
Kawada, T.; Kato, T. J. Electrochem. Soc. 1998, 145, 3196. Ionics 1992, 53-6, 418.
(230) Horita, T.; Yamaji, K.; Sakai, N.; Yokokawa, H.; Kawada, T.; (277) Manning, P. S.; Sirman, J. D.; Kilner, J. A. Solid State Ionics
Kato, T. Solid State Ionics 2000, 127, 55. 1996, 93, 125.
(231) Horita, T.; Yamaji, K.; Sakai, N.; Xiong, X. P.; Kato, T.; (278) Manning, P. S.; Sirman, J. D.; De Souza, R. A.; Kilner, J. A.
Yokokawa, H.; Kawada, T. J. Power Sources 2002, 106, 224. Solid State Ionics 1997, 100, 1.
(232) Jiang, Y.; Wang, S.; Zhang, Y.; Yan, J.; Li, W. Solid State Ionics (279) Horita, T.; Yamaji, K.; Sakai, N.; Ishikawa, M.; Yokokawa, H.;
1998, 110, 111. Kawada, T.; Dokiya, M. Electrochem. Solid State Lett. 1998, 1,
(233) Jiang, Y.; Wang, S.; Zhang, Y.; Yan, J.; Li, W. J. Electrochem. 4.
Soc. 1998, 145, 373. (280) Bozo, C.; Guilhaume, N.; Herrmann, J. M. J. Catal. 2001, 203,
(234) Jorgensen, M. J.; Primdahl, S.; Mogensen, M. Electrochim. Acta 393.
1999, 44, 4195. (281) Putna, E. S.; Bunluesin, T.; Fan, X. L.; Gorte, R. J.; Vohs, J. M.;
(235) Jiang, S. P.; Love, J. G.; Zhang, J. P.; Hoang, M.; Ramprakash, Lakis, R. E.; Egami, T. Catal. Today 1999, 50, 343.
Y.; Hughes, A. E.; Badwal, S. P. S. Solid State Ionics 1999, 121, (282) Lu, C.; Worrell, W. L.; Vohs, J. M.; Gorte, R. J. J. Electrochem.
1. Soc. 2003, 150, A1357.
(236) Tsukuda, H.; Yamashita, A. 1st European Solid Oxide Fuel Cells (283) van Hassel, B. A.; Boukamp, B. A.; Burggraaf, A. J. Solid State
Forum, Oct 3-7, 1994, Baden, Switzerland, 1994. Ionics 1992, 53-6, 890.
(237) Monceau, D.; Petot, C.; Petotervas, G. Solid State Ionics 1991, (284) van Hassel, B. A.; Burggraaf, A. J. Solid State Ionics 1992, 57,
45, 231. 193.
(238) Petot, C.; Armanet, F.; Klimczyk, H.; Petotervas, G. Solid State (285) Kim, J. H.; Choi, G. M. Solid State Ionics 2000, 130, 157.
Ionics 1992, 50, 87. (286) Boukamp, B. A.; Raming, T. P.; Winnubst, A. J. A.; Verweij, H.
(239) Petotervas, G.; Petot, C.; Monceau, D.; Loudjani, M. Solid State Solid State Ionics 2003, 158, 381.
Ionics 1992, 53-6, 270. (287) Kahlich, M. J.; Gasteiger, H. A.; Behm, R. J. J. Catal. 1999, 182,
(240) Belova, I. V.; Brown, M. J.; Murch, G. E. Acta Mater. 2003, 51, 430.
1821. (288) Liu, H. C.; Kozlov, A. I.; Kozlova, A. P.; Shido, T.; Iwasawa, Y.
(241) Brown, M. J.; Belova, I. V.; Murch, G. E. Phil. Mag. 2003, 83, Phys. Chem. Chem. Phys. 1999, 1, 2851.
1855. (289) Schubert, M. M.; Hackenberg, S.; van Veen, A. C.; Muhler, M.;
(242) Martin, M. J. Chem. Thermodyn. 2003, 35, 1291. Plzak, V.; Behm, R. J. J. Catal. 2001, 197, 113.
(243) Kawada, T.; Sakai, N.; Yokokawa, H.; Dokiya, M.; Anzai, I. Solid (290) Fabry, P.; Schouler, E.; Kleitz, M. Electrochim. Acta 1978, 23,
State Ionics 1992, 50, 189. 539.
(244) Tedmon, C. S., Jr.; Spacil, H. S.; Mitoff, S. P. J. Electrochem. (291) Lane, J. A.; Middleton, P. H.; Fox, H.; Steele, B. C. H.; Kilner,
Soc.: Electrochem. Sci. 1969, 1969, 1170. J. A. 2nd International Symposium on Ionic and Mixed Conduct-
(245) Lau, S. K.; Singhal, S. C. Proc. Corrosion 1985, 85, 79. ing Ceramics, 1994; pp 489-504.
(246) Takeda, Y.; Hoshino, Y.; Sakaki, Y.; Kawahara, T.; Imanishi, (292) Sahibzada, M.; Benson, S. J.; Rudkin, R. A.; Kilner, J. A. Solid
N.; Yamamoto, O. J. Mater. Sci. Lett. 1992, 11, 1113. State Ionics 1998, 115, 285.
(247) Takeda, Y.; Sakaki, Y.; Ichikawa, T.; Imanishi, N.; Yamamoto, (293) Chavan, S. V.; Tyagi, A. K. Thermochim. Acta 2002, 390, 79.
O.; Mori, M.; Mori, N.; Abe, T. Solid State Ionics 1994, 72, 257. (294) Longo, V.; Meriani, S.; Ricciardiello, F. J. Am. Ceram. Soc. 1981,
(248) Takeda, Y.; Ueno, H.; Imanishi, N.; Yamamoto, O.; Sammes, N.; 64, C38.
Phillipps, M. B. Solid State Ionics 1996, 86-8, 1187. (295) Sorokina, S. L.; Skolis, Y. Y.; Kovba, M. L.; Levitskii, V. A. Zh.
(249) Kostogloudis, G. C.; Tsiniarakis, G.; Ftikos, C. Personal com- Fiz. Khim. 1986, 60, 310.
munication, 1999. (296) Tsai, T.; Barnett, S. A. Solid State Ionics 1997, 98, 191.
(250) Chen, C. C.; Nasrallah, M. M.; Anderson, H. U. Proc. Electro- (297) Tsai, T.; Perry, E.; Barnett, S. J. Electrochem. Soc. 1997, 144,
chem. Soc. 1993, 93-4, 598. L130.
(251) Simner, S. P.; Shelton, J. P.; Anderson, M. D.; Stevenson, J. W. (298) Simner, S. P.; Bonnett, J. F.; Canfield, N. L.; Meinhardt, K. D.;
Solid State Ionics 2003, 161, 11. Sprenkle, V. L.; Stevenson, J. W. Electrochem. Solid State Lett.
(252) Yokokawa, H.; Sakai, N.; Kawada, T.; Dokiya, M. J. Electrochem. 2002, 5, A173.
Soc. 1991, 138, 2719. (299) Simner, S. P.; Bonnett, J. R.; Canfield, N. L.; Meinhardt, K. D.;
(253) Yokokawa, H.; Sakai, N.; Kawada, T.; Dokiya, M. Solid State Shelton, J. P.; Sprenkle, V. L.; Stevenson, J. W. J. Power Sources
Ionics 1992, 52, 43. 2003, 113, 1.
(300) Kenjo, T.; Osawa, S.; Fujikawa, K. J. Electrochem. Soc. 1991, (344) Deng, H.; Zhou, M.; Abeles, B. Solid State Ionics 1994, 1.
138, 349. (345) van Herle, J.; McEnvoy, A. J.; Thampi, K. R. Electrochim. Acta
(301) Kenjo, T.; Nishiya, M. Solid State Ionics 1992, 57, 295. 1996, 41, 1447.
(302) Austin, L. G.; Ariet, M.; Walker, R. D.; Wood, G. B.; Comyn, R. (346) Kim, J. D.; Kim, G. D.; Moon, J. W.; Park, Y. I.; Lee, W. H.;
H. Ind. Eng. Chem. Fundam. 1965, 4, 321. Kobayashi, K.; Nagai, M.; Kim, C. E. Solid State Ionics 2001,
(303) Newman, J.; Tiedemann, W. AIChE J. 1975, 21, 25. 143, 379.
(304) Juhl, M.; Primdahl, S.; Manon, C.; Mogensen, M. J. Power (347) Yoon, S. P.; Nam, S. W.; Kim, S.-G.; Hong, S.-A.; Hyun, S.-H. J.
Sources 1996, 61, 173. Power Sources 2003, 115, 27.
(305) Mogensen, M.; Skaarup, S. Solid State Ionics 1996, 86-88, 1151. (348) Bird, R. B.; Stewart, W. E.; Lightfoot, E. N. Transport Phenom-
(306) Tanner, C. W.; Fung, K. Z.; Virkar, A. V. J. Electrochem. Soc. ena, 2nd ed.; John Wiley and Sons: New York, 2002.
1997, 144, 21. (349) Perry’s Chemical Engineers’ Handbook, 50th ed.; Perry, R. H.,
(307) Virkar, A. V.; Chen, J.; Tanner, C. W.; Kim, J. W. Solid State Green, D. W., Maloney, J. O., Eds.; McGraw-Hill: New York,
Ionics 2000, 131, 189. 1984.
(308) Kim, J. D.; Kim, G. D.; Moon, J. W.; Lee, H. W.; Lee, K. T.; Kim, (350) Adler, S. B.; Henderson, B. T.; Wilson, M. A.; Taylor, D. M.;
C. E. Solid State Ionics 2000, 133, 67. Richards, R. E. Solid State Ionics 2000, 134, 35.
(309) Kim, J.-D.; Kim, G.-D.; Park, J.-A. Han’guk Seramik Hakhoechi (351) Nagata, M.; Itoh, Y.; Iwahara, H. Solid State Ionics 1994, 67,
2000, 37, 227. 215.
(310) Mogensen, M.; Primdahl, S.; Jorgensen, M. J.; Bagger, C. J. (352) Hsieh, G.; Mason, T. O.; Garboczi, E. J.; Pederson, L. R. Solid
Electroceram. 2000, 5, 141. State Ionics 1997, 96, 153.
(311) Figueiredo, F. M.; Frade, J. R.; Marques, F. M. B. Solid State (353) Winkler, J.; Hendriksen, P. V.; Bonanos, N.; Mogensen, M. J.
Ionics 2000, 135, 463. Electrochem. Soc. 1998, 145, 1184.
(312) Figueiredo, F. M.; Marques, F. M. B.; Frade, J. R. J. Electro- (354) Figueiredo, F. M.; Frade, J.; Marques, F. M. B. Boletin De La
ceram. 2001, 7, 47. Sociedad Espanola De Ceramica Y Vidrio 1999, 38, 639.
(313) Xia, C.; Rauch, W.; Wellborn, W.; Liu, M. Electrochem. Solid (355) Kato, T.; Momma, A.; Kaga, Y.; Nagata, S.; Kasuga, Y.; Kitase,
State Lett. 2002, 5, A217. M. Solid State Ionics 2000, 132, 287.
(314) Sunde, S. J. Electroceramics 2000, 5, 153. (356) Myland, J. C.; Oldham, K. B. Anal. Chem. 2000, 72, 3972.
(315) Huang, Y.; Vohs, J. M.; Gorte, R. J. Submitted for publication. (357) Oldham, K. B.; Stevens, N. P. C. Anal. Chem. 2000, 72, 3981.
(316) Huang, Y.; Ahn, K.; Vohs, J. M.; Gorte, R. J. Submitted for (358) Boukamp, B. A. Solid State Ionics 2001, 143, 47.
publication. (359) Hashibon, A.; Raz, S.; Riess, I. Solid State Ionics 2002, 149, 167.
(317) van Heuveln, F. H.; van Berkel, F. P. F.; Huijsmans, J. P. P. (360) Fleig, J. Solid State Ionics 2003, 161, 279.
High-Temperature Electrochemical Behavior of Fast Ion and (361) Adler, S. B. Manuscript in preparation.
Mixed Conductors; Roskilde: Denmark, 1993; p 53. (362) Bard, A. J.; Faulkner, L. R. Electrochemical Methods, 2nd ed.;
(318) Fleig, J.; Pham, P.; Sztulzaft, P.; Maier, J. Solid State Ionics John Wiley and Sons: 2001.
1998, 113-115, 739. (363) Vetter, K. J.; Manecke, G. Z. Phys. Chem. 1950, 195, 337.
(319) Adler, S. B. J. Electrochem. Soc. 2002, 149, E166. (364) Jimenez, R.; Kloidt, T.; Kleitz, M. J. Electrochem. Soc. 1997, 144,
(320) McIntosh, S.; Vohs, J. M.; Gorte, R. J. J. Electrochem. Soc. 2003, 582.
150, A470. (365) Kuo, J. H.; Anderson, H. U.; Sparlin, D. M. J. Solid State Chem.
(321) Singhal, S. C. Solid Oxide Fuel Cells V; Aachen: Germany, 1997; 1989, 83, 52.
pp 37-50. (366) In prior publications (refs 28 and 168) the term “non-charge-
(322) Choi, J. H.; Jang, J. H.; Ryu, J. H.; Oh, S. M. J. Power Sources transfer” has been used to describe such neutral flows. However,
2000, 87, 92. this terminology has sometimes generated unintended confusion
(323) Minh, N. Q. J. Am. Ceram. Soc. 1993, 76, 563. and/or controversy, since many chemical processes not resulting
(324) Iwata, T. J. Electrochem. Soc. 1996, 143, 1521. in a net flow of current do involve charge transfer at the atomic/
(325) Murata, K.; Shimotsu, M. Electrochem. 1999, 67, 789. molecular level (e.g., chemisorption). To avoid this confusion, the
(326) Primdahl, S.; Mogensen, M. J. Appl. Electrochem. 2000, 30, 247. term “chemical” is hereby instated in this review to describe
(327) Umemura, F.; Amano, K.; Michibata, H.; Kimura, A. Denki processes that are driven by chemical potential driving forces
Kagaku 1995, 63, 128. and may occur at a rate decoupled from the faradaic current
(328) Choi, J. H.; Jang, J. H.; Oh, S. M. Electrochim. Acta 2001, 46, while the term “electrochemical” is restricted to processes
867874. involving net passage of faradaic (noninterfacially polarizing)
(329) Jiang, S. P.; Love, J. G. Solid State Ionics 2003, 158, 45. current due to an electrochemical potential driving force.
(330) Hilpert, K.; Das, D.; Miller, M.; Peck, D. H.; Weiss, R. J. (367) In this case the exchange coefficient k has units of s-1 (rather
Electrochem. Soc. 1996, 143, 3642. than the usual cm/s) because of the reduced dimensionality of
(331) Quadakkers, W. J.; Greiner, H.; Hansel, M.; Pattanaik, A.; the system, i.e., fluxes are written in terms of surface concentra-
Khanna, A. S.; Mallener, W. Solid State Ionics 1996, 91, 55. tions rather than volumetric concentration.
(332) Badwal, S. P. S.; Deller, R.; Foger, K.; Ramprakash, Y.; Zhang, (368) For this estimate, values for the surface diffusion coefficient (D)
J. P. Solid State Ionics 1997, 99, 297. and the surface exchange coefficient (k) in eq 2 were obtained
(333) Matsuzaki, Y.; Yasuda, I. Solid State Ionics 2000, 132, 271. by linearizing Mitterdorfer’s rate expressions for surface trans-
(334) Matsuzaki, Y.; Yasuda, I. J. Electrochem. Soc. 2001, 148, A126. port and adsorption/desorption (ref 84) and re-expressing in
(335) Jiang, S. P.; Zhang, J. P.; Zheng, X. G. J. Eur. Ceram. Soc. 2002, terms of the driving forces in eq 2.
22, 361. (369) These expressions for Rchem, Clδ, and lδ have the same meaning
(336) Lee, H. Y.; Oh, S. M. Solid State Ionics 1996, 90, 133. as Rchem, Cchem ) (tchem/Rchem), and δ, respectively, as defined in
(337) Tu, H. Y.; Takeda, Y.; Imanishi, N.; Yamamoto, O. Solid State the original cited articles (refs 171 and 28). However, the
Ionics 1997, 100, 283. nomenclature and parameters used here have been updated both
(338) Brant, M. C.; Dessemond, L. Solid State Ionics 2000, 138, 1. to aid clarity and improve consistency with eqs 3-7. In particu-
(339) Brant, M. C.; Matencio, T.; Dessemond, L.; Domingues, R. Z. lar, the length parameter previously symbolized as “δ” in the
Chem. Mater. 2001, 13, 3954. cited papers has been renamed “lδ” to avoid confusion with the
(340) Khandkar, A. C.; Elangovan, S.; Liu, M.; Timper, M. Thermal oxygen nonstoichiometry δ in ABO3-δ. The subscript on the
Cycle Fatigue Behavior of High-Temperature Electrodes; Cera- chemical capacitance has also been made more specific to
matec, Inc.: 1994. indicate the size of the region disturbed by concentration
(341) Hsiao, Y. C.; Selman, J. R. Solid State Ionics 1997, 98, 33. gradients.
(342) Primdahl, S.; Mogensen, M. J. Electrochem. Soc. 1998, 145, 2431.
(343) Primdahl, S.; Mogensen, M. J. Electrochem. Soc. 1999, 146, 2827. CR020724O
Chem. Rev. 2004, 104, 4679−4726 4679
Modeling Transport in Polymer-Electrolyte Fuel Cells

Adam Z. Weber* and John Newman
Department of Chemical Engineering, University of California, Berkeley, California 94720-1462
Contents 6. Other Effects 4715

6.1. Nonisothermal Models 4716
6.2. Transient Models 4719
2. Overview of Models 4680
7. Other Models 4720
2.1. Historical 4681
8. Summary 4721
2.2. Detailed by Affiliation 4682
9. Acknowledgments 4722
2.2.1. Springer et al. Model and Derivatives 4682
10. Nomenclature 4722
2.2.2. Bernardi and Verbrugge Model and 4683
Derivatives 11. References 4723
2.2.3. Computational-Fluid-Dynamics Models 4683
2.2.4. Other Macrohomogeneous Models 4684
3. General Aspects and Equations 4685 1. Introduction
3.1. Modeling Methodologies 4685 Fuel cells may become the energy-delivery devices
3.2. General Equations 4685 of the 21st century. Although there are many types
3.2.1. Thermodynamics 4685 of fuel cells, polymer-electrolyte fuel cells are receiv-
3.2.2. Kinetics 4686 ing the most attention for automotive and small
3.2.3. Ohmic Losses 4687 stationary applications. In a polymer-electrolyte fuel
3.2.4. Mass-Transfer Limitations 4687 cell, hydrogen and oxygen are combined electrochemi-
3.3. Zero-Dimensional Models 4688 cally to produce water, electricity, and some waste
heat.
4. Fuel-Cell Sandwich Modeling 4689
During the operation of a polymer-electrolyte fuel
4.1. Conservation Equations 4689
cell, many interrelated and complex phenomena
4.2. Membrane Modeling 4690 occur. These processes include mass and heat trans-
4.2.1. Microscopic and Physical Models 4691 fer, electrochemical reactions, and ionic and elec-
4.2.2. Diffusive Models 4692 tronic transport. Only through fundamental model-
4.2.3. Hydraulic Models 4694 ing, based on physical models developed from experi-
4.2.4. Hydraulic−Diffusive Models 4694 mental observations, can the processes and operation
4.2.5. Combination Models 4695 of a fuel cell be truly understood. This review
4.3. Diffusion-Media Modeling 4695 examines and discusses the various regions in a fuel
4.3.1. Gas-Phase Transport 4696 cell and how they have been modeled.
4.3.2. Treatment of Liquid Water 4697 The focus of this review is to discuss the different
4.4. Catalyst-Layer Modeling 4700 fuel-cell models with the overall goal of presenting a
4.4.1. General Governing Equations 4701 picture of the various types of transport in fuel cells.
Although the majority of the literature fuel-cell
4.4.2. Interface Models 4702
models have been examine, there are undoubtedly
4.4.3. Microscopic and Single-Pore Models 4703 some that were left out. In terms of time frame, this
4.4.4. Simple Macrohomogeneous Models 4704 review focuses mainly on models that have been
4.4.5. Embedded Macrohomogeneous Models 4707 published through the end of 2003.
4.4.6. Catalyst-Layer Flooding 4708 In discussing the various models, this review starts
4.5. Multilayer Simulations 4709 with a historical background where the models are
4.5.1. Numerical Solution and Boundary 4709 presented and their advantages and disadvantages
Conditions are briefly discussed. However, direct comparisons
4.5.2. General Multilayer Simulation Results 4710 of the different models to each other are hard to make
5. Multidimensional Effects 4711 because they vary in their approach and complexity,
5.1. Two-Dimensional Models 4711 and one model may do a good job in one region (e.g.,
5.1.1. Along-the-Channel Models 4711 the membrane) but not in another (e.g., the elec-
5.1.2. Under-the-Rib Models 4713 trode). Furthermore, almost all of them agree with
5.2. Three-Dimensional Models 4714 some sort of experimental data. Therefore, the major-
ity of this review delves into how the various regions
* Corresponding author. E-mail: aweber@uclink.berkeley.edu. and effects are modeled independently of each other.
Telephone: 510-643-1972. Fax: 510-642-4778. It is our feeling that the reader will get more by
4680 Chemical Reviews, 2004, Vol. 104, No. 10 Weber and Newman
Figure 1. Fuel-cell schematic showing the different model

dimensionalities. 0-D models are simple equations and are
not shown, the 1-D models comprise the sandwich (z
direction), the 2-D models comprise the 1-D sandwich and
Adam Weber was born in 1976 in Fort Lauderdale, FL. He studied either of the two other coordinate directions (x or y), and
chemical engineering at Tufts University, where he earned B.S. and M.S. the 3-D models comprise all three coordinate directions.
degrees in 1999. His thesis topic was on low-temperature carbon-monoxide
oxidation using gold supported on ceria catalysts under the guidance of
Maria Flytzani-Stephanopoulos. His work was supported by the EPA and model the fuel cell with a simple equation. The
through a STAR graduate fellowship and through the Astronaut Scholarship 1-D models treat the fuel-cell sandwich in varying
Foundation. After spending a year on a Fulbright scholarship in Australia, degrees of complexity. These are the majority of the
he entered the Ph.D. program at University of CaliforniasBerkeley in models, and they model the different regions of the
chemical engineering under the guidance of John Newman. He once again fuel-cell sandwich with everything from simple equa-
earned an EPA STAR graduate fellowship for his dissertation work on
the mathematical modeling of transport inside polymer-electrolyte fuel cells.
tions to complex expressions derived from physical
models. Furthermore, they can incorporate other
(nongeometric) dimensional effects in terms of size,
that is, microscopic and macroscopic effects (e.g.,
consumption of reactant in a pore of a particle which
is within a porous electrode). The 1-D models are the
major focus of this review. What we term the 2-D
models deal with effects in the fuel cell that occur in
the sandwich as well as in another direction, either
across or along the gas channel. Finally, the 3-D
models include the 1-D sandwich and consider effects
in both directions in the flow field. Figure 1 shows a
schematic of a fuel cell, showing the different (non-
zero) modeling dimensions as defined in this article.
The outline of this review is as follows. First, the
general overview of the models is presented in both
a historical and research-group context. Next, the
John Newman received his B.S. in chemical engineering from Northwestern general modeling methodologies and overall phenom-
in 1960 and his M.S. and Ph.D. from University of CaliforniasBerkeley ena are looked at, including the 0-D models. In the
in 1962 and 1963, respectively. Then he joined the faculty at U. C.
Berkeley, where he has remained for the past 41 years. He also holds following section, the 1-D fuel-cell sandwich is dis-
an appointment as a faculty senior scientist at Lawrence Berkley National cussed in detail by layer (catalyst layers, diffusion
Laboratory. Recently, he spent a semester in 2002 as the Onsager media, membrane, and multilayer). This section is
Professor at the Norwegian University of Science and Technology in the longest of the review because the sandwich has
Trondheim, Norway. His interests lie in understanding the fundamental the most complex phenomena and is the heart of the
processes in electrochemical engineering and other coupled phenomena. fuel cell; it also is the most often modeled. Following
He is a member of the National Academy of Engineering and a Fellow of
The Electrochemical Society as well as being a highly cited author of this discussion, the multidimensional effects and
over 300 publications in the field of electrochemistry. models are considered. Next, nonisothermal and
transient effects and models are examined. Finally,
examining the modeling equations and approaches some related fuel-cell models are noted, and this is
per region than through just an encyclopedic list and followed by a summary.
discussion of the various models. In this context, if
the reader is just interested in a single region or 2. Overview of Models
effect, all of the relevant models, approaches, phe-
nomena, and equations can easily be found. Finally, The number of published fuel-cell-related models
although the majority of the models discussed are for has increased dramatically in the past few years, as
hydrogen polymer-electrolyte fuel cells, the underly- seen in Figure 2. Not only are there more models
ing phenomena and many of the regions are ap- being published, but they are also increasing in
plicable to other types of polymer-electrolyte fuel complexity and scope. With the emergence of faster
cells, such as direct-methanol fuel cells. computers, the details of the models are no longer
In this review, the discussion and models are constrained to a lot of simplifying assumptions and
classified by their geometric dimensionality. The analytic expressions. Full, 3-D fuel-cell models and
zero-dimensional (0-D) models are mainly empirical the treatment of such complex phenomena as two-
Modeling Transport in Polymer-Electrolyte Fuel Cells Chemical Reviews, 2004, Vol. 104, No. 10 4681
composed of membrane, diffusion media, and catalyst

layers, and are isothermal and 1-D. The Bernardi and
Verbrugge model assumes a fully hydrated mem-
brane and incorporates porous-electrode equations
and Stefan-Maxwell diffusion in the diffusion media
and catalyst layers. The model of Springer et al. does
not use porous-electrode equations but does consider
a changing water content in the membrane. This
allows for variable properties in the membrane such
as conductivity and the water diffusion coefficient.
Most models today can trace their roots back to these
models, and some are direct descendants, as dis-
cussed in the next section.
The next two major models were those by Fuller
and Newman15 and Nguyen and White,16 who both
examined flow effects along the channel. These
models allowed for a more detailed description of
water management and the effect of dry gas feeds
and temperature gradients. Throughout the next few
years, several more 0-D models and 1-D models were
Figure 2. Bar graph showing the number of polymer-
electrolyte related fuel-cell models published per year. generated.12,17-23 Also, some simulations examined
more detailed 2-D effects, such as rib spacing.24,25
phase flow are becoming more common. In this As interest grew in fuel cells in the late 1990s, more
section, an overview of the various fuel-cell models and more models were generated. Many of these
is presented with brief comments about their strengths models tried to understand and describe the cath-
and weaknesses. First, a short chronology of the ode.26-37 The interest was to optimize the catalyst
major features and trends is outlined. This is followed layer and increase its efficiency, thereby reducing the
by a description of the various models by research precious-metal loading and increasing performance.
group. In both of these overviews, only select models Out of these models, those of Durand and co-
are presented with a focus on macroscopic models workers30-34 advocated strongly for the use of ag-
that have a complete description of at least two layers glomerate models, where the microstructure of the
of the 1-D fuel-cell sandwich. Other models, in cathode is considered explicitly. Another important
addition to the ones below, are discussed in the model was that of Perry et al.,36 which examined
appropriate sections of this review. limiting behavior in the cathode and its possible
diagnosis. Around this time as well, interest grew in
2.1. Historical membrane modeling,38-43 as well as incorporating
transient, nonisothermal, and multidimensional ef-
The beginning of modeling of polymer-electrolyte fects in complete fuel-cell models.44-50
fuel cells can actually be traced back to phosphoric-
acid fuel cells. These systems are very similar in In 2000 and 2001, fuel-cell models were produced
terms of their porous-electrode nature, with only the by the dozens. These models were typically more
electrolyte being different, namely, a liquid. Giner complex and focused on such effects as two-phase
and Hunter1 and Cutlip and co-workers2-5 proposed flow50-56 where liquid-water transport was incorpo-
the first such models. These models account for rated. The work of Wang and co-workers57-59 was at
diffusion and reaction in the gas-diffusion electrodes. the forefront of those models treating two-phase flow
These processes were also examined later with comprehensively. The liquid-water flow was shown
porous-electrode theory.6,7 While the phosphoric-acid to be important in describing the overall transport
fuel-cell models became more refined, polymer- in fuel cells. Other models in this time frame focused
electrolyte-membrane fuel cells began getting much on multidimensional, transient, and more microscopic
more attention, especially experimentally. effects.60-73 The microscopic effects again focused on
Out of these experiments, simple 0-D models were using an agglomerate approach in the fuel cell as well
used to analyze the data (for examples, see refs as how to model the membrane appropriately.
8-12). These models normally fit the experimental In the last couple of years, the same trend toward
data with a single equation, and although they more complex and complete models has continued,
demonstrate good fits and are quick and easy to aided by the increases in computer processing speed.
implement, they are less reliable in predicting and One way to address issues in more detail is through
explaining behavior. As the experiments and viability incorporating and examining complex effects such as
of polymer-electrolyte fuel cells became more wide- flow-field design and two-phase flow.74-88 The other
spread, models were designed to simulate their way is to step back and examine a specific region of
performance and gain understanding of the underly- the fuel cell, such as the membrane or catalyst layer,
ing fundamental transport processes. From these and model that rigorously.89-100 Whichever the case,
early models, the two main fundamental ones were the models today tend to be more complex, with
those of Bernardi and Verbrugge13 and Springer et detailed consideration of the important aspects of fuel
al.14 Both of them treat the fuel-cell sandwich, cells, such as water management.
2.2. Detailed by Affiliation shows the effect of such factors as Teflon loading and
proton conductivity. Their implementation of an
Although the number of models is large, the agglomerate model is perhaps the first application
number of modeling groups and approaches is sig- of this to polymer-electrolyte fuel-cell catalyst layers.
nificantly fewer. The obvious reason is that as a Similar to this model, that of Rho and Srinivasan19
group becomes more familiar with a model, they looks at effects in the cathode in terms of operating
continually upgrade it in terms of complexity to make conditions using a detailed agglomerate model. Their
it more physically realistic. For an approach, if it is model was also relatively early and does a good job
general, then the community adopts and alters it. For in describing observations in terms of interactions at
this reason, it makes sense to give an overview of the agglomerate scale. Finally, Weisbrod et al.106
the models by these criteria. In this section, the incorporated a porous-electrode model into the
models are discussed in terms of the research groups, Springer et al. model to examine both kinetic and
where the progress of each group is highlighted and mass-transfer losses in more detail.
easily ascertained. These groups are loosely catego-
rized by modeling approach. To save space, a research Perhaps the most renowned researcher to follow
group is located only under the most appropriate the Springer modeling concept is Trung Van Nguyen.
heading, even if all of their models do not conform to His first model examined 2-D effects along the gas
that heading. Once again, the models discussed here channel.16 It is a pseudo-2-D model and has a very
are primarily those that model at least two layers of similar basis to that of the Springer et al. model.
the 1-D sandwich of a single polymer-electrolyte fuel Although its membrane model assumes only a linear
cell. Other models such as membrane, stack, imped- gradient in water concentration, it accounts for
ance, electrode, empirical 0-D, and direct-methanol liquid-water flow and is nonisothermal. The model
fuel-cell models and specific treatments of each layer clearly showed that water and heat management are
can be found in later sections of this review and are interrelated and very important for optimal fuel-cell
not discussed below. operation. The model also examined the effects of
having a nonuniform current-density distribution
2.2.1. Springer et al. Model and Derivatives down the flow channel. Finally, although liquid water
was considered in terms of energy, it had a negligible
As mentioned above, one of the first 1-D fuel-cell
volume, and the electrodes were treated as interfaces.
models was by Springer et al.14 at Los Alamos
National Laboratory (LANL). The model is isother- Nguyen’s group at the University of Kansas con-
mal and considers polarization and electrode effects tinued to upgrade this model. They also examined
only through a simple 0-D type of polarization equa- other effects in the fuel cell. Yi and Nguyen107 took
tion. However, the membrane does have variable essentially the same model as that of Nguyen and
water content, although liquid water was not con- White16 above and added coolant-plate and heat-
sidered explicitly. The modeling results showed the exchange equations. They also examined further the
importance of keeping the membrane well hydrated, effects of differential pressures and humidification
and discussed the importance of water management. conditions of the anode. The research group reduced
Springer et al.101 took the original model and added some of the limitations of the model and made it two-
a detailed cathode model to it. They examined phase and a true two-dimensional model, where the
utilization of the catalyst and the effects of perfor- flow-field structure of ribs and channels was exam-
mance on various parameters such as the diffusion- ined.48,52,108 The models that first did this were for
medium porosity and the inlet-gas composition. The interdigitated flow fields and, along with the model
cathode is treated as a uniformly distributed layer, of Kazim et al.,46 were some of the very first models
and the model does not treat flooding. Springer et to address this type of flow field.48,52 Nguyen and co-
al.102,103 also added impedance to their original model workers were also some of the first researchers to
to explore other effects and compare to another set examine two-phase flow, and they clearly showed
of data. They were one of the first to do this. Finally, how important saturation effects are at the cathode.
recently, they have come up with a sophisticated The two-phase flow models use empirical expressions
anode model to go along with the overall model that for the saturation functions taken from matching
includes the effects of carbon monoxide poisoning.104 experimental data. However, these models assume
Several groups were influenced by the work of the a net water flux through the membrane and infinitely
group at LANL. Models by these groups treat the thin catalyst layers, which somewhat limits their
membrane and fuel cell in a similar fashion, while applicability. The recent models of Natarajan and
adding various complexities. While not necessarily Nguyen56,87 examined transient and 3-D effects, as
truly a derivative of Springer, the model of Wang and well as limitations due to flooding in the catalyst-
Savinell17 has a very similar approach, except that layer interfacial regions. Overall, the work of this
they incorporate an agglomerate model of the catalyst group has pioneered examining two-phase flow ef-
layer. They examine the effects of feed concentration fects.
and humidity, although they do not consider changes Other Springer model derivatives include those of
down the gas channel. They also examine carbon Ge and Yi,109 van Bussel et al.,44 Wohr and co-
monoxide poisoning and discuss the structure of the workers,26,45 and Hertwig et al.99 Here, the models
anode. described above are slightly modified. The model of
Similar to the above model, that of Ridge et al.105 Hertwig et al. includes both diffusive and convective
examines the microstructure of the cathode catalyst transport in the membrane. It also uses a simplified
layer in more detail. Their analysis is thorough and two-phase flow model and shows 3-D distributions
of various properties in the fuel cell. The model of start with the Bernardi and Verbrugge framework
van Bussel et al. mainly focused on 2-D and dynamic and include additional effects. The model of Marr and
water transport in the membrane by diffusion and Li114 uses an agglomerate catalyst-layer model in
electroosmotic flow. They showed agreement with order to ascertain more easily the effects of changing
polarization curves and explained the experimental various layer properties such as porosity and mem-
results for counterflow and coflow operation and brane loading. Baschuk and Li50 took the Marr and
stability with dry gases based on the water-content Li model further by incorporating mass-transfer
distribution of the membrane. This is one of the first limitations in the cathode due to water flooding. They
transient models. Another early transient model is did this using an additional liquid-film resistance as
that of Wohr et al.45 Similar to the model of van an unknown parameter and were able to predict
Bussel et al., the membrane is accounted for by different levels of flooding as a function of cell
diffusion, but Wohr et al. also include liquid satura- polarization. The model of Rowe and Li71 built on the
tion and nonisothermal effects. The model does a other models and includes nonisothermal operation,
good job in showing temperature spikes in single cells which allows for a better study of water management
as well as in stacks. The catalyst-layer and diffusion- in the 1-D fuel-cell sandwich. Finally, some of the
media models for this model were based on the earlier models of Li and co-workers focus on carbon monox-
work by Bevers et al.26 Here, the cathode side of a ide poisoning of the anode electrocatalysts.115,116
membrane was simulated including liquid-water and The last main group of models that directly follow
energy effects. The simulation results demonstrate the Bernardi and Verbrugge model are those from
good agreement with experimental data and allow Italy by Murgia et al.95,117 and Pisani et al.78,91,118 In
for an increased understanding and calculation of the first of these models, the original Bernardi and
meaningful physical parameters. Finally, the model Verbrugge model is simplified to help its conver-
of Ge and Yi is 2-D and uses the same kind of gence.95 This was done by integrating over the
membrane model as van Bussel et al. They examine catalyst layers and using that result in the 1-D
the effect of differential pressures, flow arrangement, simulation. They argued that numerical accuracy was
and temperature. not compromised and computational cost and numer-
ical instability were greatly reduced. The next of the
2.2.2. Bernardi and Verbrugge Model and Derivatives models added two-phase flow effects and an ag-
Around the same time as the model of Springer et glomerate model for the catalyst layer, which ad-
al., Bernardi and Verbrugge13 published their fuel- dressed the need for a changing water content. A
cell model. This model was based on their gas- recent model from this group91 examines analytic
diffusion electrode model110 and that of Verbrugge expressions for the catalyst layer. It does a good job
and Hill.111 Prior to their fuel-cell model, Bernardi112 in examining effects on the agglomerate scale and
published a relatively simple study of the water looks at how the shape and distribution of electro-
balance in a fuel cell, highlighting the importance of catalyst agglomerates affect polarization behavior.
water management and the sensitivity of the water They also use their models to understand direct-
balance to changes in operating conditions such as methanol fuel-cell electrodes.
humidity, temperature, and pressure. The Bernardi
2.2.3. Computational-Fluid-Dynamics Models
article was a seminal paper and the first to stress
water management. The fuel-cell model of Bernardi With the increased computational power of today’s
and Verbrugge includes transport of both gas and computers, more detailed simulations are possible.
liquid in the diffusion media, detailed porous- Thus, complex equations such as the Navier-Stokes
electrode models for both electrodes, and a membrane equation can be solved in multiple dimensions, yield-
model based on Schlogl’s equation (see eq 34) with ing accurate descriptions of such phenomena as heat
gas crossover. The model is 1-D and isothermal, and and mass transfer and fluid and two-phase flow
the main limitation of the model is that the mem- throughout the fuel cell. The type of models that do
brane is assumed to be fully hydrated. This limits this analysis are based on a finite-element framework
the applicability of the model to humidified feeds and/ and are termed CFD models. CFD models are widely
or thin membranes. Another limitation of the model available through commercial packages, some of
is the neglect of true two-phase flow; the model only which include an electrochemistry module. As men-
uses constant volume fractions for the various phases. tioned above, almost all of the CFD models are based
Notwithstanding these limitations, the model allows on the Bernardi and Verbrugge model. That is to say
for a detailed examination of liquid water pressure that the incorporated electrochemical effects stem
profiles and how many different operating and struc- from their equations, such as their kinetic source
tural parameters affect overall fuel-cell performance. terms in the catalyst layers and the use of Schlogl’s
Not surprisingly, the Bernardi and Verbrugge equation for water transport in the membrane.
model forms the basis for many other models that The first major CFD models were those by Liu and
came after it, most notably the computational-fluid- co-workers25,119 at the University of Miami. They are
dynamics (CFD) models, as discussed in the next nonisothermal and the first multidimensional mod-
section. In terms of direct descendants of this model, els. They allowed for a more in-depth study of the
the model of Chan et al.113 takes the Bernardi and effects along the channels than the models described
Verbrugge model and incorporates carbon monoxide above. While the original model by Gurau et al.25 did
effects at the anode as per the Springer et al.104 not include liquid-water transport, it did have a
description. The models of Li and co-workers50,71,114-116 variable water content in the membrane. To study
flooding effects, Gurau et al.120 modeled the cathode mance. However, they did not include liquid flow in
region of the fuel cell by a simple approach where their model. The two models by Mazumder and
the cathode diffusion medium is broken down into Cole84,85 considered simulations with and without
regions of different gas-phase volume fractions. They liquid-water flow. They showed better agreement
developed analytic expressions and showed how when liquid-water flow and flooding were considered,
flooding can limit fuel-cell performance. Later models and this was done with the multiphase mixture
by Liu and co-workers79,121 incorporated the multi- model. Furthermore, both models were 3-D and
phase mixture model of Wang and co-workers57-59 to utilized complete porous-electrode descriptions of the
study the effect of liquid water in their CFD simula- catalyst layers with simple membrane models. The
tions. final CFD models are those of Costamagna65 and
As mentioned, almost all of the CFD models use Bradean et al.126 Both of these models considered the
the Bernardi and Verbrugge approach of Schlogl’s same types of effects as the other CFD ones. The
equation. The exceptions to these are the models from model of Costamagna is notable in that it considered
the University of South Carolina.51,54,60,73,82 These both convective and diffusive flow in the membrane.
models use an approach similar to that of Springer
et al. in that water transport in the membrane is due 2.2.4. Other Macrohomogeneous Models
to diffusion and the diffusion coefficient is a function There are some modeling methodologies and re-
of water content; they assume a linear concentration search groups that do not fit exclusively into the
profile of water in the membrane. Out of these above categories but should be mentioned. The
models, Shimpalee and Dutta60 generated one of the foremost among these are the models by Newman
first with CFD. It examined primarily the 3-D tem- and co-workers. Their models focus on fundamental
perature profile. Later models by Dutta et al.51,54 also phenomena and are usually simple in their dimen-
examined mass-transfer and complete 3-D effects. sionality. In terms of fuel cells, Newman127 used a
While these examined the fuel cell in more depth, simple pseudo-2-D model to look at optimization of
liquid water was not really treated in these models. hydrogen utilization in a phosphoric-acid fuel cell.
It was assumed to be part of the gas phase in thin Polymer-electrolyte fuel-cell modeling started with
film or droplet form and for the most part ignored. A the models of Fuller128 and Fuller and Newman.15 In
recent model by Lee et al.82 demonstrated good these models, concentrated solution theory was used
agreement with experiment and allowed for an to describe diffusive transport in the membrane,
understanding of the nonuniform current-density and which was a slight improvement to the Springer et
membrane-conductivity distributions. al. framework. The Fuller and Newman model was
The next step forward in CFD models came from one of the first to examine water and thermal
the work of Wang and co-workers.57,64,76,122-124 Their management simultaneously and along the gas chan-
first model64 was similar to Garua et al.25 except that nel, although it did not contain a description of liquid-
it was transient. Their next model incorporated liquid water flow. West and Fuller24 took a similar model
effects rigorously using the multiphase mixture and used it to model the 2-D effects of ribs and
model of Wang and Cheng.58,59 Their approach is channels. Dannenberg et al.129 basically used the
similar to the South Carolina one above, but by same model as Fuller and Newman, but they incor-
considering mixture parameters and the appropriate porated agglomerate effects in the catalyst layer and
saturation equations, the liquid-phase flow was a different water uptake isotherm in order to examine
calculated. The modeling domain for this model was the effect of changing operating conditions. Finally,
basically the cathode side of the fuel cell, where a the models by Meyers and Newman130-132 also im-
net water flux through the membrane was assumed. proved upon the Springer et al. framework by using
Very recent models examine the effects of solid-phase a thermodynamically rigorous treatment of transport
properties such as electronic conductivity124 and in the membrane for the optimization of direct-
wettability,123 as well as interdigitated flow fields.76 methanol fuel cells.
Around the same time as the other models were The next set of Newman group models were aimed
the models of Djilali and co-workers.47,80,125 The first at demonstrating the applicability of models to
of these47 was 2-D, and while it accounted for liquid- diagnostic techniques. Perry et al.36 examined the
water flow, saturation effects were neglected, since changes in the Tafel slope due to changes in the
they assumed two independent networks for liquid controlling phenomena in the cathode side of the fuel
and gas. Their next model used the same approach cell. This model was updated by Weber et al.,133 who
but was 3-D.125 Their most recent model accounted added explicit oxygen mass-transfer limitations in
for saturation effects and coupled the liquid and the diffusion media, and again later by Jaouen et
vapor flows.80 It was able to show temperature, al.,98 who treated the catalyst layers as agglomerates.
concentration, and water distributions in the 3-D flow Recently, Weber and Newman89,93,94,134 introduced
field. a framework for bridging the gap between the Ber-
Recently, other CFD models have been published. nardi and Verbrugge and the Springer et al. mem-
The model of Siegel et al.90 used an agglomerate brane approaches. The membrane model was used
approach instead of the porous-electrode approach of in a simple fuel-cell model, and it showed good
the other CFD models. They showed that the ag- agreement with experimentally measured water-
glomerate approach enables good comparisons to balance data under a variety of conditions.134 The
experimental data and showed the effects of ag- fuel-cell model was similar to the model of Janssen,55
glomerate radius and membrane loading on perfor- who used chemical potential as a driving force in the
membrane, but Weber and Newman also accounted mainly stem from the high overpotential of the
for such effects as Schroeder’s paradox. Other mem- oxygen-reduction reaction (ORR), although the pres-
brane-modeling approaches in fuel-cell models be- ence of carbon monoxide will produce a similar effect
sides those mentioned above include those by Datta at the anode. As the current is increased, ohmic
and co-workers,63,72,92 who use a dusty-gas approach, losses become a factor in lowering the overall cell
and Hsing and co-workers,41,61,62 who use Flory- potential. These ohmic losses are mainly from ionic
Huggins theory. Finally, Weber et al.74 and Weber losses in the electrodes and separator, although
and Newman88 have recently developed models to contact and electronic resistances can be important
account for saturation effects and flooding phenom- under certain operating conditions and cell configu-
ena. Along with the model of Nam and Kaviany,86 rations. At high currents, mass-transport limitations
they are some of the first to account for composite become increasingly important. These losses are due
and hydrophobic diffusion media. to reactants not being able to reach the electrocata-
Within the last five years, many fuel-cell models lytic sites. Typically, oxygen is the problem due to
have come out of the Research Center in Julich, flooding of the cathode by liquid water, but protons
Germany. These models have different degrees of and electrons can also result in mass-transfer limita-
complexity and seek to identify the limiting factors tions. Before examining general polarization-curve
in fuel-cell operation. The model of Kulikovsky et al.37 models, some discussion should be made of the
examined a 2-D structure of rib and channel on the approach used for modeling and the equations used
cathode side of the fuel cell, and is similar to that of for the general regions of the polarization curve.
Springer et al. Other models by Kulikovsky included
examination of depletion along long feed channels67,83 3.1. Modeling Methodologies
and effects in the catalyst layers.69,96,135 The most
recent model by Kulikovsky81 relaxed the assumption There are different global methodologies for model-
of constant water content in the membrane and ing fuel cells and, in particular, the fuel-cell sand-
examined quasi 3-D profiles of it. Also at the research wich. The easiest division to make is between mac-
center, Eikerling et al.28,40,70,136-138 developed many roscopic and microscopic models. The microscopic
different models. Most of these were concerned with models seek to model transport on an individual pore
modeling the membrane40,70,136,138 and cathode side level, whereas the macroscopic ones are continuum
of the fuel cell.28,137 These models were complex and and average over this level. Although the microscopic
focused on statistically relating macroscopic phenom- models may provide more realistic conditions and
ena to structural properties. In both systems, they factors, they require a lot more knowledge of the
developed fundamental equations for the transport microstructure and are much more expensive in
processes and examined different limiting cases, terms of computation time. Macroscopic models are
leading toward optimization analysis. more common for fuel cells, although some micro-
scopic details should be incorporated into them. An
3. General Aspects and Equations example of this is the agglomerate model for fuel-
cell catalyst layers, as discussed below. This review
The performance of a fuel cell is most often focuses mainly on macroscopic models.
reported in the form of a polarization curve. Such a In a macrohomogeneous approach, the exact geo-
curve is shown in Figure 3. Roughly speaking, the metric details of the modeling domain are neglected.
polarization curve can be broken down into three Instead, the domain is treated as a randomly ar-
main regions. At low currents, the behavior of a fuel ranged porous structure that can be described by a
cell is dominated by kinetic losses. These losses small number of variables such as porosity and
surface area per unit volume. Furthermore, transport
properties within the domain are averaged over the
electrode volume. Thus, all variables are defined at
all positions within the domain. Averaging is per-
formed over a region that is small compared to the
size of the electrode but large compared to the
microstructure.
3.2. General Equations

3.2.1. Thermodynamics
As shown in Figure 3, the open-circuit potential
represents the highest voltage obtainable for a single
cell. This potential is derived from thermodynamics.
The overall fuel-cell reaction can be broken down into
the two global electrode reactions. If hydrogen is the
primary fuel, it oxidizes at the anode according to
the reaction
Figure 3. Example of a polarization curve showing the
typical losses in a polymer-electrolyte fuel cell. H2 a 2H+ + 2e- (1)
At the cathode, oxygen is reduced reaction can be expressed as
4H+ + 4e- + O2 f 2H2O (2) ∑ ∑si,k,hMzi f nhe- i

(8)
k i
Although the above reactions are written in terms
of the global species, no specific mention has been where si,k,h is the stoichiometric coefficient of species
made about how the reaction proceeds. For example, i residing in phase k and participating in electron-
each reaction can be broken down into single electron- transfer reaction h, nh is the number of electrons
transfer reactions, and the gas probably dissolves into transferred in reaction h, and Mzi i represents the
the membrane that is covering the electrocatalyst chemical formula of i having valence zi.
sites (e.g., platinum on carbon). Some more comments The rate of an electrochemical reaction depends on
about this can be found in section 4.4. Adding eqs 1 the concentrations of the various species and the
and 2 yields the overall reaction potential drop across the reaction interface between
phases k and p, which are normally the electrode and
2H2 + O2 f 2H2O (3) electrolyte, respectively. In general, a Butler-Volmer
expression can be used to describe the kinetics
[ ( )
A typical polymer-electrolyte fuel cell can be repre-
sented as a pi RaF
ih,k-p ) i0h ∏ exp
RT
(Φk - Φp - Uref
h ) -
pref
)]
i
(
i
c p
i -RcF
∏ exp
RT
(Φk - Φp - Uref
h ) (9)
where each Greek letter identifies a distinct phase i pref
i
and the wavy lines imply that the membrane phase
boundary is not sharp; rather, the membrane extends where ih,k-p is the transfer current between phases k
into adjacent regions and may include water activity and p due to electron-transfer reaction h, the prod-
gradients. The potential of this cell is139 ucts are over the anodic and cathodic reaction species,
respectively, Ra and Rc are the anodic and cathodic
FU ) -F(ΦR - ΦR′) ) µeR- - µeR′- (5) transfer coefficients, respectively, pi and pref
i are the
partial pressure and reference partial pressure for
where F is Faraday’s constant, U is the thermody- species i, respectively, and i0h and Uref h are the
namically defined reversible cell potential, ΦR is the exchange current density per unit catalyst area and
electrical potential of phase R, and µeR- is the elec- the potential of reaction h evaluated at the reference
trochemical potential of electrons in phase R. After conditions and the operating temperature, respec-
introducing expressions for the activities of the tively. The reference potential can be determined
various components,139 this becomes using a Nernst equation (e.g., see eq 7); if the
reference conditions are the same as the standard
RT RT RT conditions (i.e., 100 kPa pressure for the different gas
FU ) FUθ + β
ln aH + β′
ln aO - w +
ln aβ′
2 2 4 2 2 species), then Urefh has the same numerical value as
+ - µH+) (6)
β β′
(µH Uθh.
The term in parentheses in eq 9 can be written in
where aβi is the activity of species i in phase β and terms of an electrode overpotential
the subscript w stands for water, R is the ideal-gas
constant, T is the absolute temperature, and Uθ is ηh,k-p ) Φk - Φp - Uref
h (10)
the standard cell potential, a combination of ap-
propriately chosen reference states that is a function In this review, the reference electrode used is defined
of temperature and can be unit dependent. This as a platinum metal electrode exposed to hydrogen
equation reduces to a Nernst equation139,140 at the same temperature and electrolyte (e.g., Nafion)
( )
as the solution of interest. With this reference
RT pH2xpO2 electrode, the electrode overpotential defined in eq
U ) Uθ + ln (7) 10 is the same as having the reference electrode
2F pw located next to the reaction site but exposed to the
reference conditions (i.e., it carries its own extraneous
when the gases are assumed to be ideal and gradients phases with it). Typical values for the reference
in the electrolyte are zero or neglected. Values and conditions are those in the gas channels. If the
expressions for the parameters in the above equation, reference electrode is exposed to the conditions at the
as well as further thermodynamic discussions, can reaction site, then a surface overpotential can be
be found in various books.139-144 defined
3.2.2. Kinetics ηsh,k-p ) Φk - Φp - Uh (11)
The initial drop in the polarization curve is due to
the sluggish kinetics of the ORR at the temperatures where Uh is the reversible potential of reaction h (e.g.,
normally used for current polymer-electrolyte fuel- see eq 7). The surface overpotential is the over-
cell operation (<100 °C). A typical electrochemical potential that directly influences the reaction rate
across the interface. Comparing eqs 10 and 11, one ing or comparing data and modeling results, as
can see that the electrode overpotential contains both pointed out by Gasteiger et al.158 Furthermore, the
a concentration and a surface overpotential for the exchange current densities in the two equations are
reaction. not necessarily equal because they are (perhaps
For the hydrogen-oxidation reaction (HOR) at the different) functions of the reference pressures.
anode, eq 9 becomes, in the absence of poisons, As noted, the equations above are a general start-
[ ( )
ing point for describing the ORR, the HOR, and the
pH2 RaF kinetic regime. More detailed models, which examine
iHOR,1-2 ) i0HOR exp (η ) -
p H2ref RT HOR,1-2 such effects as the nature of the reaction interface,
)]
are discussed in later sections.
exp (
-RcF
(η
RT HOR,1-2
) (12) 3.2.3. Ohmic Losses
The second part of the polarization curve is the
where 1 and 2 denote the electron- and proton- ohmic regime. In this region, the potential varies
conducting phases, respectively. Because the elec- linearly with the current density. This allows for
trolyte is a polymer of defined acid concentration, the Ohm’s law to be used for modeling purposes
proton concentration does not enter directly into eq
12. Furthermore, due to the reference electrode used, ∆Φ ) iR′ (16)
the reference potential and reversible potential are
both equal to zero. Equation 12 reduces to a Nernst
where R′ is the overall resistance in the fuel cell. For
relationship when the ratio i/i0 becomes small, which
the ohmic region, this resistance is essentially con-
is normally the case for the HOR.13,17,145-149
stant; that is, it is not a strong function of the current
The oxygen-reduction reaction (ORR), on the other
density or potential. The resistance can either be
hand, is slow and represents the principal inefficiency
measured experimentally by fitting polarization curves
in many fuel cells. Due to its sluggishness, the ORR
or be determined through modeling. It is a composite
is modeled reasonably well with Tafel kinetics with
of the general electronic and ionic resistances through
a first-order dependence on oxygen partial pres-
the fuel cell and is similar to a contact resistance. It
sure10,150,151
does not include those resistances resulting from
() ( )
mass-transfer effects such as membrane dehydration,
pO2 -RcF
iORR,1-2 ) -i0ORR exp (η ) (13) which are discussed below. In later sections, models
ref
pO RT ORR,1- 2 that calculate R′ are discussed in detail.
2
A linear fit on a Tafel plot of overpotential versus 3.2.4. Mass-Transfer Limitations

the log of the current density yields the commonly The last part of the polarization curve is dominated
reported Tafel slope by mass-transfer limitations (i.e., concentration over-
RT potential). These limitations arise from conditions
b ) 2.303 (14) wherein the necessary reactants (products) cannot
RcF reach (leave) the electrocatalytic site. Thus, for fuel
cells, these limitations arise either from diffusive
For the kinetic region, the values of the theoretical
resistances that do not allow hydrogen and oxygen
and experimental Tafel slopes have been shown to
to reach the sites or from conductive resistances that
agree with Rc equal to 1.9,10,150,152-157
do not allow protons or electrons to reach or leave
If eq 13 were to be written with respect to the
the sites. For general models, a limiting current
surface overpotential, as defined by eq 11, instead of
density can be used to describe the mass-transport
the electrode overpotential, then it would read
limitations. For this review, the limiting current
() ( )
pO2 3/4
-RcF density is defined as the current density at which a
iORR,1-2 ) -i0ORR exp η (15) reactant concentration becomes zero at the diffusion
ref
pO RT sORR,1-2 medium/catalyst layer interface.
2
As mentioned, membrane dehydration can result
The power or the exponent with respect to oxygen is in a problem where the membrane conductivity
different because of the overpotential being used. As decreases to the point that it no longer conducts
discussed by Newman,139 the difference is due to how protons away from the reaction sites. Also, the
the exchange current density is defined and its distributed ohmic effects coupled with mass transfer
dependence written, and both equations are consist- in the porous electrode can result in a concentration
ent with a reaction order of 1. Mathematically, one overpotential that limits performance. In addition,
can show that the exponent on oxygen changes from if the diffusion media are composed mainly of insu-
1 to 3/4 if Rc is equal to 1.158 While eq 15 has perhaps lating solids such as Teflon, electron conduction can
a better defined overpotential and contains only a become limiting. However, the most common cause
kinetic term and not also a thermodynamic term, eq for mass-transfer limitations is due to oxygen being
13 is often easier to use in simulations and in inhibited from reaching the cathode reaction sites.
conceptualizing the dependence of the kinetics on gas This is normally due to flooding in either the catalyst
concentration. Either equation is correct, and one layer or the diffusion medium because air and
must recognize which one is being used when analyz- hydrogen are normally fed into a fuel cell, water is
produced at the cathode, and the diffusion coefficient brane. For example, the model of Amphlett et al.21
of hydrogen is greater than that of oxygen.142,149 accounts for all of the temperature dependences
For typical fuel-cell designs, mass transport through through empirical fitting parameters. 0-D models also
the fuel-cell sandwich occurs mainly by diffusion. The allow for deviations from the theoretical expression
simplest way to describe diffusion is by Fick’s law149 presented above to be investigated. For example,
fitting eq 19 to some experimental data yields a
Di double Tafel slope at higher current densities. This
Ni ) - ∇p (17) change in the Tafel slope is caused by transport
RT i
limitations, something not explicitly taken into ac-
where Ni is the superficial flux density of species i count in the above equation. In other words, the
and Di is the diffusion coefficient of species i in the equation cannot fit the whole polarization curve with
mixture. As discussed later, many models use more just one set of parameter values.
complex expressions and effective diffusion coef- To make the model less empirical, gas-phase mass-
ficients. These complexities allow the nature of the transport limitations can be incorporated into the
fuel-cell porous media to be accounted for, as well as modeling equation explicitly12
counterdiffusion of water vapor and flooding of the
porous media by liquid water. V ) U′ - b log(i) - R′i - m exp(ni) (20)
An expression for the limiting current density due
to oxygen transport can be derived by the following where m and n are fitting parameters. Although the
procedure. With Faraday’s law for the oxygen flux above expression yields good fits with the data, it is
in the ORR and rearrangement of Fick’s law, the more empirical than if the limiting current density
CL
oxygen partial pressure at the catalyst layer, pO 2
, is used97,139,161
can be related to the limiting current density, ilim ,
( )
bulk 127
and the oxygen partial pressure in the bulk, pO , i
2 V ) U′ - b log(i) - R′i + b log 1 - (21)
( )
ilim
CL bulk i
pO ) pO 1- (18)
2 2 ilim
Other modifications to include mass-transport limi-
tations deal with changing the last term to be a more
Equation 18 can be used to write the ORR rate in eq complicated function of current with more fitting
13 or 15 in terms of the limiting current density, an parameters.118,162,163
experimentally measurable quantity, rather than the
The final 0-D equation presented here stems from
oxygen partial pressure.
incorporating the gas-pressure dependences directly
instead of through a limiting current density, which
3.3. Zero-Dimensional Models normally only considers oxygen effects. This equation
The simplest fuel-cell models describe the polariza- was proposed by Newman127 for phosphoric-acid fuel
tion behavior by a single equation; hence, they are cells and predates the above polymer-electrolyte
0-D. Even though these models are relatively simple fuel-cell expressions. It has the form
and usually empirical, they are valuable for deter-
mining kinetic parameters as well as comparing the
various losses in the system to one another. The 0-D
modeling equations can be derived by combination
V ) Uθ +
RT
F
ln(a1,2i0L) -
RT
F
ln
i
( )
pO2
+
of the above governing equations for each regime. The RT

ln(pH2) - R′i (22)
drawbacks of these models are that they do not yield 2F
true mechanistic behavior and fundamental under-
standing and that they are not good for predicting where a1,2 is the interfacial area of the catalyst per
performance or optimization. unit volume of electrode and L is the thickness of the
0-D models are very helpful in determining kinetic cathode catalyst layer. Thus, the quantity a1,2L is a
parameters and general ohmic resistance from roughness factor, a ratio of catalyst area to superficial
data.9,10,21,23,152,159-161 A typical expression for this type electrode area. Out of all the approaches to include
of analysis is mass transport, eqs 21 and 22 are the most phenom-
enological, but with the fewest fitting parameters,
V ) U + b log(i0) - b log(i) - R′i (19) they do not fit the data the best.
To examine how well the models agree with the
where V is the cell potential and i is the superficial data, Figure 4 shows model fits using eq 20. As can
current density through the membrane. The first and be seen, the model fits the data very well even with
second terms may be combined to form a potential different operating conditions of oxygen mole fraction
intercept, U′; this quantity is a convenient way to and gas pressure. Of course, the model parameters
group terms pertaining to (possibly unknown) ther- are adjusted for the conditions. As mentioned, such
modynamic and kinetic constants. polarization-equation fits are useful for getting pa-
Upon comparison to data, the models demonstrate rameter values and perhaps some gross understand-
the aforementioned value of Rc being equal to 1. Not ing, but they cannot really be used for optimization,
only are exchange current densities measured, but prediction, or in-depth examination of the underlying
the effect of temperature can be studied in more phenomena. In essence, they are curve fits. It is
detail, as well as the type and thickness of mem- difficult for them to treat interacting phenomena in
Figure 5. 3-D schematic showing the various layers of the

fuel-cell sandwich.
chemically at the anode catalyst layer to form hy-

drogen ions and electrons. The diffusion medium is
typically a carbon cloth or carbon paper, possibly
Figure 4. Model and experiment comparison of polariza-
tion curves for air or oxygen at different gas pressures and treated with Teflon. The catalyst layer usually con-
at 70 °C using eq 20. (Reproduced with permission from tains platinum or a platinum alloy supported on
ref 12. Copyright 1995 The Electrochemical Society, Inc.) carbon and an ionomeric membrane material such
as Nafion. The oxidant, usually oxygen in air, is fed
a clear way, such as drying of a membrane with into the cathode flow field, moves through the diffu-
current density or temperature or flooding with liquid sion medium, and is electrochemically reduced at the
water. cathode by combination with the generated protons
More complicated expressions than those above can and electrons. The water, either liquid or vapor,
be used in the 0-D models, but these usually stem produced by the reduction of oxygen at the cathode
from a more complicated analysis. For example, the exits the fuel cell through either the cathode or anode
equation used by Ticianelli and co-workers29,156 comes flow field. The electrons generated at the anode pass
from analysis of the catalyst layer as a flooded through an external circuit and may be used to per-
agglomerate. In the same fashion, eq 21 can be form work before they are consumed at the cathode.
embedded and used to describe the polarization Mass and energy transport occur throughout all of
behavior within a much more complicated model. For the various sandwich layers. These processes, along
example, the models of Springer et al.14 and Weber with electrochemical kinetics, are key in describing
and Newman74 use a similar expression to eq 21, but how fuel cells function. In this section, thermal
they use a complicated 1-D model to determine the transport is not considered, and all of the models
parameters such as ilim and R′. Another example is discussed are isothermal and at steady state. Some
the model of Newman,127 who uses eq 22 and takes other assumptions include local equilibrium, well-
into account reactant-gas depletion down the gas mixed gas channels, and ideal-gas behavior. The
channels by, in essence, having a limiting current section is outlined as follows. First, the general
density that depends on the hydrogen utilization. All fundamental equations are presented. This is fol-
of these types of models, which use a single equation lowed by an examination of the various models for
to describe the polarization behavior within a more the fuel-cell sandwich in terms of the layers shown
complicated model, are discussed in the context of in Figure 5. Finally, the interplay between the
the more complicated model. various layers and the results of sandwich models
are discussed.
4. Fuel-Cell Sandwich Modeling
The fuel-cell sandwich describes the 1-D cross
4.1. Conservation Equations
section of the fuel cell (see Figure 1) and is shown in The number of equations and unknowns must
Figure 5. For the single dimension, flow is taken to balance. Thus, one can calculate the appropriate
be normal to the various layers. Flow in the other number of needed relationships from the degrees of
directions is discussed in section 5. The fuel-cell freedom of a system, as shown for various systems
sandwich contains the gas channels or flow fields, by Newman.164 In terms of the relations, the equa-
diffusion media, catalyst layers, and membrane. tions can be broken down into five main types. The
Additional layers are sometimes incorporated into the first are the conservation equations, the second are
sandwich, such as separating the diffusion media into the transport relations, the third are the reactions,
microporous and gas-diffusion layers. Fuel cells oper- the fourth are equilibrium relationships, and the fifth
ate in the following manner. are the auxiliary or supporting relations, which
The fuel is fed into the anode flow field, moves include variable definitions and such relations as
through the diffusion medium, and reacts electro- Faraday’s law.
As noted in the Introduction, one of the defining a current ik, written in boldface and with a single
characteristics of any fuel-cell model is how it treats subscript, indicates the total current density carried
transport. Thus, these equations vary depending on within phase k
the model and are discussed in the appropriate
subsections below. Similarly, the auxiliary equations ik ) F∑ziNi,k (25)
and equilibrium relationships depend on the model- i
ing approach and equations and are introduced and
discussed where appropriate. The reactions for a fuel where zi is the valence or charge number of species
cell are well-known and were introduced in section i. Finally, rl,k-p is the rate of the heterogeneous
3.2.2. Of course, models modify the reaction expres- reaction l per unit of interfacial area between phases
sions by including such effects as mass transfer and k and p. Rg,k is the rate of a strictly homogeneous
porous electrodes, as discussed later. Finally, unlike reaction g in phase k per unit volume of phase k.
the other equations, the conservation equations are Because a large electrical force is required to
uniformly valid for all models. These equations are separate charge over an appreciable distance, a
summarized below and not really discussed further. volume element in the electrode will, to a good
It is necessary to write a material balance for each approximation, be electrically neutral. For fuel-cell
independent component in each phase. The dif- models, electroneutrality is often assumed for each
ferential form of the material balance for species i phase
in phase k is133
∂kci,k ih,1-k
∑zici,k ) 0 (26)
i
) -∇‚Ni,k - ∑a1,ksi,k,h +
∂t h nhF The assumption of electroneutrality implies that the
∑si,k,l ∑ ak,prl,k-p + ∑si,k,gkRg,k (23) diffuse double layer, where there is significant charge
separation, is small compared to the volume of the
l p*k g
domain, which is normally the case. Because there
The term on the left side of the equation is the is no accumulation of charge and electroneutrality
accumulation term, which accounts for the change has been assumed, the divergence of the total current
in the total amount of species i held in phase k within density is zero
a differential control volume. This term is assumed
to be zero for all of the sandwich models discussed ∑∇‚ik ) 0 (27)
in this section because they are at steady state. The k
first term on the right side of the equation keeps
track of the material that enters or leaves the control Equations 23 and 27 represent the mass and current
volume by mass transport. The remaining three conservation equations, respectively. These apply for
terms account for material that is gained or lost due all of the models discussed.
to chemical reactions. The first summation includes
all electron-transfer reactions that occur at the 4.2. Membrane Modeling
interface between phase k and the electronically
conducting phase (denoted as phase 1). The second One of the most important parts of the fuel cell is
summation accounts for all other interfacial reactions the electrolyte. For polymer-electrolyte fuel cells this
that do not include electron transfer, and the final electrolyte is a single-ion-conducting membrane.
term accounts for homogeneous reactions in phase Specifically, it is a proton-conducting membrane.
k. Although various membranes have been examined
In the above expression, ci,k is the concentration of experimentally, most models focus on Nafion. Fur-
species i in phase k, and si,k,l is the stoichiometric thermore, it is usually necessary only to modify
coefficient of species i in phase k participating in property values and not governing equations if one
heterogeneous reaction l (see eq 8). ak,p is the specific desires to model other membranes. The models
surface area (surface area per unit total volume) of presented and the discussion below focus on Nafion.
the interface between phases k and p. ih,k-1 is the Nafion is a copolymer of poly(tetrafluoroethylene)
normal interfacial current transferred per unit in- and polysulfonyl fluoride vinyl ether. It has fixed
terfacial area across the interface between the elec- anions, which are sulfonic acid sites, and conse-
tronically conducting phase and phase k due to quently, by electroneutrality, the concentration of
electron-transfer reaction h, and it is positive in the positive ions is fixed. Furthermore, the transference
anodic direction. In the above expression, Faraday’s number of protons in this system is 1, which greatly
law simplifies the governing transport equations, as seen
below. There can be different forms of Nafion in
ih,1-k terms of the positive counterion (e.g., proton, sodium,
Ni,k ) ∑si,k,h (24) etc.). Most models deal only with the proton or acid
h nhF form of Nafion, which is the most common form used
in polymer-electrolyte fuel cells due to its high proton
was used to change the interfacial current density conductivity.
into an interfacial flux quantity. Furthermore, a Since the membrane is such a key element in the
current ih,1-k, written with two subscripts, implies an fuel cell, it has had a lot of attention in terms of
interfacial, or transfer, current density. Conversely, modeling. There have been many microscopic and
physical models in addition to the macroscopic ones.

The microscopic models focus solely on the membrane
and examine single ions and pore-level effects. The
macroscopic models are often more empirical and
focus on describing the transport and relevant pa-
rameters of the membrane in a macrohomogeneous
fashion. As per the overall approach of this review,
discussion is mainly on the macroscopic models.
The macroscopic models can be sorted into two
main categories, those that assume the membrane
system is a single phase, and those that assume it is
two phases. Each is discussed separately below. In
either of these models, the membrane system is
assumed to have three main components: mem-
brane, protons, and water. For the three-species
system, there should be 1/2N(N - 1) ) 3 transport
properties.133 The above set of properties neglects any
other kind of ions in the membrane, which is an
assumption that almost all of the models make. Figure 6. Evolution of the membrane structure as a
Furthermore, the above property count does not function of water content, λ (moles of water per mole of
consider hydrogen or oxygen crossover in the mem- sulfonic acid sites). The pictures are cross-sectional repre-
brane. Such crossover results in only a small inef- sentations of the membrane where the gray area is the
ficiency and does not significantly influence proton fluorocarbon matrix, the black is the polymer side chain,
the light gray is the liquid water, and the dotted line is a
or water transport. There are some models that take collapsed channel. (Reproduced with permission from ref
this into account, the most notable being by Bernardi 89. Copyright 2003 The Electrochemical Society, Inc.)
and Verbrugge.110 If desired, crossover can easily be
added to any model.94
Before discussing the models and their approaches physical model of transport in Nafion, which could
in more detail, some mention should be made on the be used as a foundation for a macrohomogeneous
microscopic models and the overall physical picture model. Figure 6 shows a schematic summary of the
of Nafion. Since membrane-only models, especially Weber and Newman model. The main focus of the
the microscopic ones, are covered in another re- model is how the membrane structure changes as a
view,165 the discussion below is shortened. function of water content, where λ is the moles of
water per mole of sulfonic acid sites and is measured
4.2.1. Microscopic and Physical Models by examining the weight gain of an equilibrated
membrane.
There have been many microscopic models based In panel a, the dry membrane absorbs water in
on statistical mechanics,38,136,166-172 molecular order to solvate the acid groups. This initial water is
dynamics,173-180 and macroscopic phenomena associated strongly with the sites. Additional water
applied to the microscopic structure of the causes the water to become less bound, and inverted
membrane.63,111,138,181-201 These models provide the micelles form in the polymer matrix, panel b. With
fundamental understanding of processes such as more water uptake, these clusters grow and form
diffusion and conduction in the membrane on a interconnections with each other. The connections,
single-pore level. They allow for the evaluation of how or collapsed channels, are transitory and have hy-
small perturbations like heterogeneity of pores and drophobicities comparable to that of the matrix. The
electric fields affect transport, as well as the incor- cluster-channel network forms on the basis of a
poration of small-scale effects. These models are also percolation-type phenomenon of the clusters; there-
the only ones to look at different forms of Nafion as fore, to form a transport pathway, the clusters must
well as the effect of other ions in the membrane (e.g., grow and be close enough together to be linked by
those from a salt solution).42,43,111,192-201 They also the collapsed channels. From conductivity data, this
divide the membrane into more than just the three percolation threshold is shown practically to occur
species of the macroscopic models; as examples, the around λ ) 2.94 Panel c corresponds to a membrane
membrane may be broken down into ionic and that is in contact with saturated water vapor, where
backbone moieties, and proton conduction may be by a complete cluster-channel network has formed.
different proton-water complexes. Finally, almost all When there is liquid water at the boundary of the
of the microscopic models treat the membrane as a membrane, structural reorganization occurs because
two-phase system. While the microscopic models yield the aqueous environment repels the fluorocarbon-rich
valuable information about what goes on inside the skin of the ionomer. This inversion allows for the
membrane, in general, they are too complex to use liquid water to infiltrate and expand the channels,
in an overall fuel-cell model. causing them to stabilize and the various clusters to
Many of today’s models are based on the early agglomerate along them; as a result, a porelike
physical models of Hsu and Gierke182 and Yeager and structure forms, panel d. In this structure, the
Steck.184 These models, along with the relevant channels are still hydrophobic on average. Because
experimental data, were reviewed recently by Weber the channels are now filled with liquid, the uptake
and Newman.89,93 Out of their analysis came a of the membrane has increased without a change in
the chemical potential of the water (i.e., Schroeder’s be related to one another at infinite dilution via the
paradox202). Nernst-Einstein equation139,205,206
In the physical model, there are two separate
structures for the membrane depending on whether Di ) RTui (30)
the water at the boundary is vapor or liquid; these
are termed the vapor- or liquid-equilibrated mem- So long as the solute species are sufficiently dilute
brane, respectively. The main difference between the that the interactions among them can be neglected,
two is that, in the vapor-equilibrated membrane, material balances can be written upon the basis of
panel c, the channels are collapsed, while, in the the above expression for the flux.
liquid-equilibrated case, panel d, they are expanded If water movement in the membrane is also to be
and filled with water. These two structures form the considered, then one way to do this is to again use
basis for the two types of macroscopic models of the the Nernst-Planck equation. Because water has a
membrane. zero valence, eq 29 reduces to Fick’s law, eq 17.
However, it is also well documented that, as the
4.2.2. Diffusive Models protons move across the membrane, they induce a
flow of water in the same direction. Technically, this
The diffusive models treat the membrane system electroosmotic flow is a result of the proton-water
as a single phase. They correspond more-or-less to interaction and is not a dilute solution effect, since
the vapor-equilibrated membrane (panel c of Figure the membrane is taken to be the solvent. As shown
6). Because the collapsed channels fluctuate and in the next section, the electroosmotic flux is propor-
there are no true pores, it is easiest to treat the tional to the current density and can be added to the
system as a single, homogeneous phase in which diffusive flux to get the overall flux of water
water and protons dissolve and move by diffusion.
Many membrane models, including some of the i2
earliest ones, treat the system in such a manner. Nw,2 ) ξ - Dw∇cw,2 (31)
Since the membrane is stationary, only the water
F
and protons move in the membrane system. The where ξ is the electroosmotic coefficient. The above
simplest membrane models either neglect the water equation, along with Ohm’s law, has been used suc-
movement or treat it as a known constant. For the cessfully for most of the models that treat the mem-
proton movement, the simplest treatment is to use brane as a single phase.14,16,17,39,44,51,54,60,81,82,90,129,159,207
Ohm’s law (eq 16 in differential form) The deviations and complications in the models arise
from what function to use for the transport proper-
i2 ) -κ∇Φ2 (28) ties, κ, ξ, and Dw, as well as the concentration of
water in the membrane, cw,2. To understand the
where κ is the ionic conductivity of the membrane. differences in the models, a closer look at these
This can easily be integrated to yield a resistance for functions is required, but first the models that use
use in a polarization equation (see the 0-D models concentrated solution theory will be presented.
above). The above water and proton treatments are 4.2.2.2. Concentrated Solution Theory. For an
relatively trivial and are often used when the purpose electrolyte with three species, it is as simple and
of the model is to examine effects outside the mem- more rigorous to use concentrated solution theory.
brane (e.g., cathode flooding)26,48,52,56,57,69,80,85-87,101,105,120 Concentrated solution theory takes into account all
or when only general trends are desired.19,20,75,114,203,204 binary interactions between all of the species. For
Below, more rigorous treatments are examined. membranes, this was initially done by Bennion208 and
4.2.2.1. Dilute Solution Theory. Equation 28 is Pintauro and Bennion.209 They wrote out force bal-
the result of using dilute solution theory.139 Such an ances for the three species, equating a thermody-
analysis yields the Nernst-Planck equation namic driving force to a sum of frictional interactions
for each species. As discussed by Fuller,128 Pintauro
Ni,2 ) -ziuiFci,2∇Φ2 - Di∇ci,2 + ci,2v2 (29) and Bennion also showed how to relate the interac-
tion parameters to the transport parameters men-
The first term in the expression is a migration term, tioned above. The resulting equations for the three-
representing the motion of charged species that species system are
results from a potential gradient. The migration flux
is related to the potential gradient (-∇Φ2) by a κξ
i2 ) - ∇µ - κ∇Φ2 and (32)
charge number, zi, concentration, ci, and mobility, ui. F w,2
The second term relates the diffusive flux to the
concentration gradient. The final term is a convective i2
term and represents the motion of the species as the Nw,2 ) ξ - Rw∇µw,2 (33)
F
bulk motion of the solvent carries it along. For the
analysis of the one-phase systems, the solvent is the where µw represents the chemical potential of water
membrane, and thus, v2 ) 0. and Rw is the transport coefficient of water. The
Dilute solution theory considers only the interac- equation for the membrane is ignored, since it is
tions between each dissolved species and the solvent. dependent on the other two equations by the Gibbs-
The motion of each charged species is described by Duhem equation. The above equations have also been
its transport properties, namely, the mobility and the arrived at using an irreversible thermodynamics
diffusion coefficient. These transport properties can approach.210,211 Similar equations to those above were
used by Wohr et al.,45 who put them in a Stefan-

Maxwell framework.149
Upon comparison of eq 32 to 28, it is seen that the
proton-water interaction is now taken into account.
This interaction is usually not too significant, but it
should be taken into account when there is a large
gradient in the water (e.g., low humidity or high-
current-density conditions). Upon comparison of eq
33 to 31, it is seen that the equations are basically
identical where the concentration and diffusion coef-
ficient of water have been substituted for the chemi-
cal potential and transport coefficient of water,
respectively. Almost all of the models using the above
equations make similar substitutions for these vari-
ables.15,24,61,62,128
The models by Janssen55 and Weber and Newman94
do not substitute for the chemical potential. Janssen
took the transport coefficient as a fitting parameter.
His model showed good agreement with water bal-
ance data under different conditions. Weber and
Newman utilize the chemical-potential driving force Figure 7. Equilibrium water-uptake or isotherm curve at
30 °C. The dashed line signifies the effect of Schroeder’s
directly and use a diffusion coefficient relative to that paradox, a change in water uptake at the same chemical
driving force for their vapor-equilibrated transport potential depending on the phase of water next to the
mode. Janssen states that using a chemical-potential membrane; liquid is at λ ) 22.
driving force does not necessitate making the distinc-
tion between diffusive or convective flow in the next to it.63,92,94,130 Finally, although it is known that
membrane. However, as Meyers212 points out, by the isotherm is a function of temperature, only a few
assuming the membrane system is a single phase, it models include this feature.61-63,72,94,129 Most models
cannot support a pressure difference inside it. The just use an isotherm at a given temperature, usually
only way that a single-phase membrane model can 30 °C.
have a pressure difference across it is if the chemical Schroeder’s paradox is an observed phenomenon
potential or water concentration is being altered at that needs to be considered in any model where the
the boundaries. This problem is why single-phase membrane is not either fully hydrated or dehydrated.
membrane models cannot adequately describe trans- There are various methods to account for Schroeder’s
port for fully hydrated membranes where the driving paradox. The easiest way is to ignore it (i.e., either
force is the liquid pressure. This point is discussed only vapor filled or fully hydrated), which a majority
further in section 4.2.4. of the models do. Next, it can be treated as a
4.2.2.3. Membrane Water Content. Whether the discontinuity or by assuming a functional form of the
dilute solution or concentrated solution theory equa- water content such that λ and aG w continue to in-
tions are used to model the membrane system, crease.16,51,54,60,64,71,81,82,109,129 Hence, supersaturated
functional forms for the transport parameters and conditions exist, which are only physically realistic
the concentration of water are needed. The prop- if water is assumed to be a mist or suspended in the
erties are functions of temperature and the water gas. More physical and rigorous models have also
content, λ.94 In the models, empirical fits are been generated.
used14-17,24,44,51,54,60-62,82,128 or the properties are as- Eikerling et al.138 have a random network of pores
sumed constant.39,81,90,159 A review of these properties that are filled with either bulklike water or bound
and how each model implements them is beyond the water, and impregnation by liquid water is easier
scope of this paper. Weber and Newman94 review than condensation. They use effective medium theory
many of the data and discuss the functional forms to predict conductivity results from impedance data.
for the properties from a physical standpoint. Their model is more of a microscopic one in which λ
Different models determine λ in different ways. is calculated by changing the number of pores that
Nafion exhibits a water-uptake isotherm as shown are filled and examining the types of liquid-film
in Figure 7. The dashed line in the figure shows the bonds between pores. Similarly, Weber and New-
effects of Schroeder’s paradox, where there is a man94 use capillary arguments where the liquid
discontinuous jump in the value of λ. Furthermore, pressure expands the collapsed channels within the
the transport properties have different values and membrane. Thus, there is a continuous transition
functional forms at that point. Most models used between the vapor- and liquid-equilibrated mem-
correlate λ with the water-vapor activity, aG w, since it branes that can occur within the membrane. The
is an easily calculated quantity. An exception to this transition itself can be relatively sharp, which is not
is the model of Siegel et al.,90 which assumes a simple unexpected for a phase-change-type behavior. Fi-
mass-transfer relationship. There are also models nally, Choi and Datta92 recently came up with an
that model the isotherm either by Flory-Huggins explanation of Schroeder’s paradox that involves
theory41,61,62 or equilibrium between water and hy- having evaporation at the liquid-filled pore mouth for
drated protons in the membrane and water vapor a vapor-equilibrated membrane. The resulting me-
niscus causes a higher energy that the membrane where k and kΦ are the effective hydraulic and
must equilibrate with, thus lowering its water con- electrokinetic permeability, respectively, pL is the
tent. Out of the above three models, only that of hydraulic or liquid pressure, µ is the water viscosity,
Weber and Newman has been used in a fuel-cell and zf and cf refer to the charge and concentration of
sandwich model. fixed ionic sites, respectively.
Those models that use an empirical expression for In the above system, the movement of water can
the isotherm often have different driving forces. One be attributed to a potential gradient and a pressure
of the first models to use an isotherm was that by gradient. The movement of water by a pressure
Springer et al.14 In that model, λ was used as the gradient is determined primarily by an effective
driving force for water flow in the membrane, and permeability of water moving through the pore
an activity coefficient was used to account for the network. This approach is quite useful for describing
isotherm behavior. This same approach was used by fuel-cell systems where the membrane is well hy-
van Bussel et al.44 and Wang and Savinell.17 The drated, but it requires that the water content be
latter case examined the anode and assumed a flux uniform across the membrane, with only a pressure
of water through the membrane to help determine gradient as a driving force for water movement. Such
the membrane water concentration. Similar ap- a treatment does not necessarily lend itself to de-
proaches to that of Springer et al. were used by scribing the flux of water resulting when there is a
Kulikovsky,81 Fuller and Newman,15 Dannenberg et water-activity gradient across the membrane (i.e.,
al.,129 and West and Fuller.24 The models of Nguyen when the membrane is not fully hydrated).
and White16 and Shimpalee and co-workers51,54,60,82 The Bernardi and Verbrugge model also assumes
assume a linear change in water concentration in the that there is a gas volume fraction in the membrane
membrane. Finally, Okada and co-workers39,207 use that remains constant. This does not necessarily
analytic and perturbation expressions for how the agree with the physical picture presented and ex-
concentration of water changes. They start with perimental evidence. The reason for including the gas
vapor-equilibrated membranes and treat them with volume fraction was to allow for gas crossover
either semi-infinite or finite boundary conditions. through the membrane. Such a process though can
be included using Fick’s law for the gases in the
4.2.3. Hydraulic Models membrane, since the diffusion coefficients of oxygen
and hydrogen in Nafion are known. Many other
In opposition to the single-phase treatment of the models use the same approach and equations as
membrane system mentioned above are the models Bernardi and Verbrugge, especially systems wherein
that assume the membrane system is two phases. the membrane is expected to be well hydrated (e.g.,
This type of model corresponds to the liquid-equili- saturated gas feeds).25,47,50,64,78,95,125
brated membrane shown in panel d of Figure 6. In Unlike the cases of the single-phase models above,
this structure, the membrane is treated as having the transport properties are constant because the
pores that are filled with liquid water. Thus, the two water content does not vary, and thus, one can expect
phases are water and membrane. a linear gradient in pressure. However, due to
The addition of a second phase means that there Schroeder’s paradox, different functional forms might
is also an additional degree of freedom. This results be expected for the vapor- and liquid-equilibrated
in the ability of the membrane system to sustain a membranes.
pressure gradient in the water because of a possibly Instead of the dilute solution approach above,
unknown stress relation between the membrane and concentrated solution theory can also be used to
fluid at every point in the membrane. However, model liquid-equilibrated membranes. As done by
diffusion of water becomes meaningless, since the Weber and Newman,94 the equations for concentrated
water is assumed to be pure in the models discussed solution theory are the same for both the one-phase
here. Furthermore, unlike the cases of the models and two-phase cases (eqs 32 and 33) except that
discussed above, the water content of the membrane chemical potential is replaced by hydraulic pressure
is assumed to remain constant (λ ) 22) as long as and the transport coefficient is related to the perme-
the pores are filled and the membrane has been ability through comparison to Darcy’s law. Thus, eq
pretreated appropriately. For cases where the pores 33 becomes
do not remain filled, see sections 4.2.4 and 4.2.5.
The first model to describe the membrane in i2 k
the above fashion was that of Bernardi and Nw,2 ) ξ - ∇pL (35)
F µV
Verbrugge,13,110 which was based on earlier work by w
Verbrugge and Hill.111,213,214 This model utilized a
dilute solution approach that used the Nernst- where Vw is the molar volume of water.
Planck equation (eq 29) to describe the movement of
protons, except that now v is not equal to zero. The 4.2.4. Hydraulic−Diffusive Models
reason is that, because there are two phases, the
The two approaches above can be contrasted to one
protons are in the water and the velocity of the water
another. In the first approach, section 4.2.2, water
is give by Schlogl’s equation213,215
moves by diffusion, and pressure-driven flow is
() ( )
excluded as a separate driving force. In the second
k kΦ approach, section 4.2.3, only pressure-driven flow is
vw,2 ) - ∇pL - z c F∇Φ2 (34)
µ µ f f used, and there is no diffusive flow because the liquid
water in the pores is pure. To describe both effects, 4.2.5. Combination Models
some other kind of model is needed.
There is a need to be able to describe both types of
As mentioned at the end of section 4.2.2.2, a one-
behavior, diffusive and hydraulic, in a consistent
phase open system cannot support a pressure gradi-
manner, which also agrees with experimental data.
ent without experiencing bulk flow. One way around
For example, a membrane with a low water content
this restriction is to use the chemical potential as the
is expected to be controlled by diffusion, and an
overall driving force.55,94 In essence, this driving force
uptake isotherm needs to be used (see Figure 7). The
combines those of pressure and activity
reasons for this are that there is not a continuous
liquid pathway across the medium and that the
∇µw,k ) RT∇ ln aw,k + Vw∇pk (36) membrane matrix interacts significantly with the
water due to binding and solvating the sulfonic acid
where pk is the total pressure of phase k. For the sites. A hydraulic pressure in this system may not
single-phase system, there is not an additional degree be defined.
of freedom to have both types of gradients. Thus, the On the other hand, when the membrane is satu-
above equation and approach requires that an ad- rated, transport still occurs. This transport must be
ditional relation is needed to specify both variables, due to a hydraulic-pressure gradient because over-
and this is given by assuming local equilibrium (i.e., saturated activities are nonphysical. In addition,
there is only one overall gradient, that of chemical Buechi and Scherer217 found that only a hydraulic
potential). model can explain the experimentally observed sharp
Another way around the problem of pressure- drying front in the membrane. Overall, both types
driven flow in the single-phase membrane was pre- of macroscopic models describe part of the transport
sented by Meyers.212 He worked around the problem that is occurring, but the correct model is some kind
by allowing for a discontinuity in pressure at the of superposition between them.94,218 The two types of
membrane/solution interface, even though the elec- models are seen as operating fully at the limits of
trochemical potential of all soluble species is continu- water concentration and must somehow be averaged
ous. He argued that additional mechanical stresses between those limits. As mentioned, the hydraulic-
compressing the membrane should be indistinguish- diffusive models try to do this, but from a non-
able from the thermodynamic pressure, and thus, the physical and inconsistent standpoint that ignores
thermodynamic pressure might be discontinuous at Schroeder’s paradox and its effects on the transport
the interface. properties.
There have been various models that try to incor- Weber and Newman94 do the averaging by using a
porate both diffusive flow and convective flow in one capillary framework. They assume that the two
type of membrane and using one governing transport transport modes (diffusive for a vapor-equilibrated
equation.63,65,71,72,79,99,107 They are based somewhat on membrane and hydraulic for a liquid-equilibrated
concentrated solution theory where the concentration one) are assumed to occur in parallel and are
and total gas-phase pressure driving forces are used switched between in a continuous fashion using the
fraction of channels that are expanded by the liquid
i2 k water. Their model is macroscopic but takes into
Nw,2 ) ξ - Dw∇cw,2 - cw,2 ∇pG (37) account microscopic effects such as the channel-size
F µ
distribution and the surface energy of the pores.
A dusty-fluid model216 has also been used to combine Furthermore, they showed excellent agreement with
the effects, which adds convection to the Stefan- experimental data from various sources and different
Maxwell framework, as discussed in a later sec- operating conditions for values of the net water flux
tion.63,72 This approach is akin to eq 37. As discussed per proton flux through the membrane.134
above, the concentration of water may be replaced Eikerling et al.40,138 used a similar approach except
by a function of λ. that they focus mainly on convective transport. As
The problem with the above approach lies in the mentioned above, they use a pore-size distribution
meaning of the two different gradients. As noted for Nafion and percolation phenomena to describe
above, the pressure driving force does not make sense water flow through two different pore types in the
for a one-phase system. Thus, the model implicitly membrane. Their model is also more microscopic and
assumes a two-phase system. In a two-phase system, statistically rigorous than that of Weber and New-
the interstitial concentration of water never varies, man. Overall, only through combination models can
and the superficial concentration varies only if the a physically based description of transport in mem-
volume fraction of water changes in the membrane. branes be accomplished that takes into account all
Furthermore, the use of a gas pressure means that of the experimental findings.
a gas-phase is assumed to exist in the membrane,
something that does not agree with experimental 4.3. Diffusion-Media Modeling
data. Overall, treatment in this manner is the same
as that for the diffusive models except for an ad- As shown in Figure 5, the diffusion media are the
ditional parameter to account for results that show porous backings between the catalyst layers and the
a greater flux of water from cathode to anode when gas channels. They provide structural support, dis-
the fuel cell is operated with a pressure differential. tribute the reactant gases, and provide a pathway
This addition helps to fit some of the data, but it is for electrons, gases, and liquid water to move to or
not rigorous or consistent. from the catalyst layers. The diffusion media are
often composed of either a single gas-diffusion layer ent on the others, since the partial pressures sum to
or a composite structure of a gas-diffusion layer and unity)
a microporous layer. It should be noted that there
are only a few models that treat composite diffusion xiNj - xjNi
media in a rigorous fashion.86,88,122 Most models treat ∇xi ) ∑ (40)
eff
only gas-diffusion layers. j*i cTDi,j
Besides gas and liquid transport in the diffusion
media, there is also electronic conduction. Most where cT is the total concentration or molar density
models neglect this due to the high conductivity of of all of the gas species, xi is the mole fraction of
the carbon in the diffusion media, although it can eff
species i, and Di,j is the effective binary interaction
become a limiting factor due to geometry124 or diffu- parameter between i and j, by the Onsager reciprocal
sion-media composition.88 For those that take it into eff eff
relationships, Di,j ) Dj,i for ideal gases. The effec-
account, Ohm’s law is used
tive diffusion coefficient is defined as
i1 ) -σ011.5∇Φ1 (38) G
eff
Di,j ) D (41)
τG i,j
where 1 and σ0 are the volume fraction and electrical
conductivity of the electronically conducting phase,
respectively. The above equation has been adjusted where G and τG are the volume fraction and tortu-
for porosity and tortuosity using a Bruggeman osity of the gas phase, respectively. If liquid water is
correction.219-222 In the diffusion media, carbon is the ignored, then G is set to the value of the bulk porosity
conducting phase and the other solid component, of the medium, 0. If liquid water is not ignored, then
Teflon, is insulating. another treatment is required, as discussed in the
The mass balances of the species in the diffusion next section. Typically, a Bruggeman expression is
media can be deduced from eq 23. Furthermore, the used for the tortuosity219-222
fluxes of the various species are often already known
at steady state. For example, any inert gases (e.g., τG ) G-0.5 (42)
nitrogen) have a zero flux, and the fluxes of reactant
gases are related to the current density by Faraday’s However, the above expression can underpredict the
law (eq 24). Although water generation is given by tortuosity at low porosities.86,102 Nam and Kaviany86
Faraday’s law, water can evaporate or condense in have a very good discussion of the appropriate
the diffusion media. These reactions are often mod- function to use for the tortuosity, in which changing
eled by an expression similar to values due to liquid saturation are also accounted for.
The gases in a fuel cell are typically hydrogen and
revap ) kmaG,L(pw - pvap water on the fuel side, and air and water on the
w ) (39)
oxidant side. Since there are not many components
to the gases and one of the equations in eq 40 can be
where revap is the molar rate of evaporation per unit
replaced by the summation of mole fractions equals
volume, km is a mass-transfer coefficient per unit
1, many models simplify the Stefan-Maxwell equa-
interfacial surface area, aG,L is the interfacial gas-
tions. In fact, eq 40 reduces to Fick’s law for a two-
liquid surface area per unit volume, pw is the partial
component system. Such simplifications are trivial
pressure of water in the gas phase, and pvap w is the and are not discussed here.
vapor pressure of water, which can be corrected for As the pore size decreases, molecules collide more
pore effects by the Kelvin equation.223 Although often with the pore walls than with each other. This
different models may write eq 39 in slightly different movement, intermediated by these molecule-pore-
forms (e.g., with a switching function), the underlying wall interactions, is known as Knudsen diffusion.224
equation and principles are the same. Finally, the Some models have begun to take this form of diffu-
two-phase models may also have an interfacial area sion into account.26,37,45,49,69,72,74,81 In this type of
that depends on the water content of the media (for diffusion, the diffusion coefficient is a direct function
example, see ref 86). Overall, the value of kmaG,L is of the pore radius.149 In the models, Knudsen diffu-
typically high enough that the gas is saturated if sion and Stefan-Maxwell diffusion are treated as
liquid water exists; that is, pw ) pvap
w (using this last mass-transport resistances in series149,225 and are
expression means that a special treatment is needed combined to yield
to combine the vapor and liquid material balances
for water so as to eliminate the net evaporation rate). Ni xiNj - xjNi
In terms of transport, both gas- and liquid-phase ∇xi ) - +∑ (43)
transport should be described. Below, approaches for eff
cTDK j*i eff
cTDi,j
both types of transport are examined, with the gas- i
phase transport being treated first.

where the DKeffi is the effective Knudsen diffusion
4.3.1. Gas-Phase Transport coefficient. In effect, the pore wall, with zero velocity,
constitutes another species with which the diffusing
Almost every model treats gas-phase transport in species interact, and it determines the reference
the fuel-cell sandwich identically. The Stefan- velocity used for diffusion.226 The above equation also
Maxwell equations are used (one of which is depend- can be derived from a dusty-gas analysis.216
From an order-of-magnitude analysis, when the The main reason there are models that account for
mean-free path of a molecule is less than 0.01 times gas-phase pressure-driven flow is that these models
the pore radius, bulk diffusion dominates, and when are often multidimensional and are considering ef-
it is greater than 10 times the pore radius, Knudsen fects besides just through the 1-D fuel-cell sandwich.
diffusion dominates. This means that Knudsen dif- As discussed in a later section, the pressure differ-
fusion is significant when the pore radius is less than ence down a gas channel is much more significant
about 0.5 µm. For reference, a typical carbon gas- than that through the sandwich. The only type of fuel
diffusion layer has pores between 0.5 and 20 µm227-229 cell where gas-phase pressure-driven flow needs to
in radius, and a microporous layer contains pores be accounted for in the fuel-cell sandwich is one using
between 0.05 and 2 µm.230,231 Thus, while Knudsen an interdigitated flow field.48,52,232 In these types of
diffusion may not have to be considered for gas- fuel cells, the gas channels are not continuous
diffusion layers, it should be accounted for in micro- through the fuel cell, and thus, gas is forced by both
porous and catalyst layers. convection and diffusion through the diffusion media
While most models treat gas-phase flow as to reach the next gas channel.
purely due to diffusion (i.e., the total gas pres-
sure or concentration remains uniform), some 4.3.2. Treatment of Liquid Water
models take into account convection in the gas-
Liquid water has been modeled to various degrees
phase.25,45,48,51,52,54,60-64,72,80,82,90,122,125,126,216 This is
in fuel cells, and the different approaches are dis-
usually done by the addition of Darcy’s law for the
cussed in terms of their complexity. The simplest way
gas phase
to account for liquid water, besides ignoring it, is to
kG treat it essentially as a solid species that occupies a
vG ) - ∇p (44) certain volume fraction. Its transport is not consid-
µG G ered, and the only effect it has is to decrease the gas-
phase volume. This decreases the effective diffusion
where k is the effective permeability. The above
coefficients of the gas species (see eq 41) and some-
relation can be made into a flux by multiplying it by
what takes into account flooding. The models that
the total concentration of the gas species.
do this approach usually use the volume fraction of
One way to include the effect of gas-phase pressure-
water as a fitting parameter.14,50,72,75,101,120 Out of
driven flow is to use eq 44 as a separate momentum
these models, those of Chu et al.,75 Gurau et al.,120
equation.51,54,60-62,64,80,82,125 The models that do this are
and Baschuck and Li50 use a liquid volume fraction
primarily CFD ones. Another way to include pres-
that is a function of position to mimic flooding effects.
sure-driven flow is to incorporate eq 44 into the
Chu et al. showed the effect of different variations
Stefan-Maxwell equations, as per the dusty-gas
in the liquid volume fraction on various fuel-cell
model45,63,90,126,216
parameters including potential, oxygen mass fraction,
xikG xiNj - xjNi Ni and current density. Gurau et al. came up with an
∇xi ) - ∇pG + ∑ - (45) analytic solution for the cathode side of the fuel cell
eff
DK µ j*i eff
cTDi,j eff
cTDK and showed how the limiting current density and
i G i
polarization effects depended on the different liquid
However, this is not necessarily a rigorously correct volume fractions. Baschuck and Li fit polariza-
treatment, since the bulk-fluid velocity should not tion curves by having different liquid volume frac-
just be linearly combined with the transport equations at each point. The resultant volume-fraction
tions. Instead, one of the Stefan-Maxwell equations profile gives an idea about how flooding occurs and
should be replaced by eq 44, since it is in essence the progresses.
summation of the mass velocities of the gas species.226 The next more complicated treatment of liquid
Although there are models that incorporate gas- water is to have a way in which to model also its
phase pressure-driven flow in the diffusion media, transport without going to a two-phase model. The
the question can arise as to whether this is a models of this sort assume that the liquid water
significant effect. The results of almost all of the exists as droplets that are carried along in the gas
models show that the pressure difference through the stream.16,25,65 Thus, while evaporation and condensa-
sandwich is minimal, and the assumption of uniform tion occur, a separate liquid phase does not have to
gas pressure is probably fine for most conditions (for be modeled. Instead, the liquid is assumed to be a
examples, see refs 13, 56, 57, and 134). This is as component of the gas, and usually one that has a
expected, since the gases flow parallel to the sand- negligible effect on the gas-phase flow and velocity.
wich and, due to a no-slip condition, move through There is a change in the gas-phase volume fraction
the diffusion media primarily by diffusion. There is due to the water, however. This type of model allows
a caveat to the above conclusion in that some models for the existence and location of liquid water to be
do show that even a small pressure difference hinders noted, and to a limited extent the change in the water
mass transfer if it is in the direction out of the pressure or concentration.
sandwich; however, this effect depends greatly on the The above two types of models are essentially one
permeability of the diffusion media and is somewhat phase. To model liquid-water flow accurately, two-
debatable. In addition, small pressure gradients phase models are required. Liquid-phase transport
when coupled with thermal gradients might affect is similar to the gas-phase pressure-driven flow
water transport significantly, as discussed in later described above. There is no diffusion component to
sections. water movement because the liquid water is assumed
to be pure. Thus, the flux form of Darcy’s law models rating a detailed description of a medium’s micro-
the flow of liquid water structure. A complete description of the various
modeling approaches is beyond the scope of this
k article, and the reader is referred to reviews and
Nw,L ) - ∇pL (46) books on the subject.223,233,234 In this section, only
Vwµ those models that are used for modeling the diffusion
media in fuel cells are discussed.56,57,74,78-80,85-87 These
where Vw is the molar volume of water and all of the models are macroscopic and on the simpler end of
properties are valid for pure water. Many models use the porous-media-model scale, which makes them
the above equation with a set value of the liquid- easier to use although less accurate.
phase volume fraction.13,47,52,71,78,95,99,107,110,125 In es- The interaction between liquid and gas is charac-
sence, such a picture assumes that there are isolated terized by a capillary pressure, contact angle, surface
gas and liquid pores in the medium. This makes some tension, and pore radius223,234-236
sense, since the medium contains hydrophilic and
hydrophobic materials and pores, but in reality there 2γ cos θ
should be some transfer between them as well as a pC ) pL - pG ) - (47)
r
changing volume fraction. Another way to look at it
is that it is describing a thin film of liquid that coats where γ is the surface tension of water, r is the pore
a hydrophilic surface. radius, and θ is the internal contact angle that a drop
In the same fashion as that of the above models, of water forms with a solid. Equation 47 is based on
those of Wohr et al.45 and Bevers et al.26,129 use a how liquid water wets the material; hence, for a
similar expression to eq 46 but instead use a surface hydrophilic pore, the contact angle is 0° e θ < 90°,
diffusion coefficient instead of the effective perme- and for a hydrophobic one, it is 90° < θ e 180° (for
ability divided by the viscosity. Also, they use a examples, see refs 86 and 123).
water-loading gradient that one could relate to a One of the most important aspects of the two-phase
water pressure. Bevers et al. also have a volume models is their ability to predict the liquid saturation
fraction that changes with respect to time. The model as a function of position. The saturation, S, is defined
of Janssen55 also has a similar equation to those of as the amount of pore volume that is filled with
the other models but instead uses a driving force of liquid; thus
chemical potential and a fitting parameter for the
permeability. Weber and Newman134 use a similar G ) 0(1 - S) (48)
approach to that of Janssen but use a liquid-pressure-
gradient driving force. They basically use the perme- This equation shows that the saturation greatly
ability as a fitting parameter that is set under only affects the effective gas-phase diffusion coefficients.
one set of operating conditions for each laboratory Hence, flooding effects are characterized by the
setup. saturation.
Finally, there are the models that use a phase In the models, the saturation is normally calculated
mixture approach.51,54,60,82 In this approach, the two using an empirical function to relate the capillary
phases are treated as a single-phase mixture. Thus, pressure to the saturation. Nguyen and co-work-
all parameters are mixture parameters for the two ers56,87 use a function they developed from fitting
phases. However, unlike the cases of the single-phase data, and Pisani et al.78 use a function with an
models, eq 46 is used for the liquid, which effectively unknown fitting parameter, while the other models
determines its mass flux. A problem with this ap- use the empirically determined Leverett J-func-
proach is that the mixture moves with a single tion.57,79,80,85,86,123 A different approach than that using
velocity (i.e., the gas and liquid move with the same a functional form for the capillary pressure-satura-
velocity). Thus, the liquid pressure is a result of this tion relationship is to calculate the saturation using
velocity; it does not really have a separate driving a model. Weber and Newman74 do this using a
force. Although these simplifications are made, usu- random cut-and-rejoin bundle-of-capillaries model223
ally to help in numerical stability and computation in order to integrate analytically the pore-size dis-
time, the models do a reasonable job in predicting tribution of the diffusion media to get the saturation.
the water balance of the fuel cell. This approach can In their analysis, they also explicitly account for the
be seen as a simplified version of the multiphase mixed wettability of the diffusion media (i.e., eq 47
mixture model by Wang and Cheng,58,59 as discussed is an integration limit, and there are separate
below. While the above models describe liquid-water hydrophobic and hydrophilic pore-size distributions).
transport to some extent, the existence of partially Along with Nam and Kaviany,86 they have the only
saturated media requires the use of capillary equa- model to account for such effects. The use of a bundle-
tions and a rigorous two-phase description of two- of-capillaries type of model is not necessarily rigorous
phase flow. because it idealizes the actual porous network.
4.3.2.1. Rigorous Two-Phase Flow Models. It Only a handful of models treat the diffusion media
is well-known that gas and liquid interact to a certain as at least partially hydrophobic;74,78,86,123 the others
extent in a porous medium. Over the decades, many treat the medium as entirely hydrophilic. Although
different modeling approaches have been developed in principle it does not change the approach, the
for two-phase flow in porous media. These ap- physical picture and boundary conditions must change.
proaches range from a simple bundle-of-capillaries An entirely hydrophilic medium means that the
model to very complex 3-D network models incorpo- liquid pressure must always be below the gas pres-
sure or the medium is flooded. This is taken into

account in the models by assuming that the satura-
tion equals zero at the interface of a diffusion medium
with a gas channel. In principle, this seems fine, yet
it also means that the liquid pressure is much lower
than the gas pressure (and can even approach a value
of zero) at this interface. The hydrophobic models
assume that at this interface the capillary pressure
is equal to zero, or in other words the liquid pressure
equals the gas pressure. This assumption is not
necessarily true due to menisci formation in the
diffusion media, but it seems to be a more valid
assumption, especially since the diffusion media have
added Teflon to keep them from flooding (i.e., there
is a need to have hydrophobic pores). The true
boundary condition is probably a thermodynamic
balance that also considers multidimensional flow
effects.
To determine the saturation for any of the models, Figure 8. Effective permeability as a function of capillary
the capillary pressure must be known at every pressure for the different two-phase models for the gas-
diffusion layer. The lines correspond to the models of (a)
position within a diffusion medium. Hence, the two- Berning and Djilali,80 (b) You and Liu79 and Mazumder and
phase models must determine the gas and liquid Cole,85 (c) Wang et al.,57 (d) Weber and Newman,74 (e)
pressure profiles. In typical two-phase flow in porous Natarajan and Nguyen,56 and (f) Nam and Kaviany.86
media, the movement of both liquid and gas is
determined by Darcy’s law for each phase and eq 47
relates the two pressures to each other. Many models without increasing the overall error of the simula-
utilize the capillary pressure functionality as the tions too much.
driving force for the liquid-water flow The different capillary pressure, saturation, and
permeability relationships of the different models can
k k be compared. To do this, the effective permeabilities
Nw,L ) - ∇pL ) - (∇pC + ∇pG) ) from some of the different models are plotted as a
Vwµ Vwµ function of the capillary pressure in Figure 8. In the
k figure, the capillary pressure at which the effective
- ∇pC (49) permeability no longer changes is where the medium
Vwµ is fully saturated. Also, the values of the effective
permeability are dependent on the diffusion media
where the last equality is justified by the assumption being tested. Furthermore, the value of the effective
of a uniform gas pressure in the fuel-cell sandwich. permeability at the right end of the curves corre-
A useful relation in calculating the effective perme- sponds to the saturated permeability, except for the
ability, k, is to define a relative permeability, kr, model of Weber and Newman, who use a gas-phase
residual saturation.
k ) krksat (50) The values of the effective permeabilities vary over
orders of magnitude, and this corresponds to the
where ksat is the saturated permeability, or the different results of the models. Furthermore, as
permeability at complete saturation, of the medium. discussed in various papers,56,80,85 the effective per-
ksat depends only on the structure of the medium and meability of Natarajan and Nguyen (curve e) varies
has either been assumed,57,80,85 used as a fitting significantly over a very small pressure range, al-
parameter,56,74,87 or estimated using a Carman- though they state that their capillary-pressure equa-
Kozeny equation.86 tion mimics data well. With respect to the various
If the dependence of kr on saturation is known, then equations, the models that use the Leverett J-
it can be used in eq 49 (via eq 50) directly. Nguyen function57,79,80,85,86 all have a similar shape except for
and co-workers56,87 and Berning and Djilali80 assume that of Berning and Djilali (curve a), who used a
a linear dependence of kr on saturation, and most of linear variation in the permeability with respect to
the other models use a cubic dependence;57,78,79,85,86 the saturation. The differences in the other curves
the model of Weber and Newman74 yields close to a are due mainly to different values of porosity and
cubic dependence. This last model differs from the saturated permeability. As mentioned above, only the
others because it obtains an analytic expression for models of Weber and Newman (curve d) and Nam
kr as a function of the capillary pressure (the inde- and Kaviany (curve f) have hydrophobic pores, which
pendent variable). Furthermore, they also calculated is why they increase for positive capillary pressures.
and used residual or irreducible saturations, which For the case of Weber and Newman, the curve has a
are known to exist59,223,234,235 but have only been stepped shape due to the integration of both a
incorporated into a few other models.78,86 In ac- hydrophilic and a hydrophobic pore-size distribution.
cordance with typical data, a cubic dependence is the The addition of the gas and water mass balances
most often observed, although the other dependences (eq 39 with eq 23) along with the above transport
help in the mathematical convergence of the models equation (eq 46) and constitutive relationships com-
pletely specifies the system of equations. In making where DS is a so-called capillary diffusivity
these equations, some of the models include the effect
of gravity and an additional advection or convection k dpC
term. These other terms are secondary effects and, DS ) (54)
in a fashion similar to that of the full Navier-Stokes µV dS
w
equations, are not as important as the Darcy effects
due to the small pore sizes and low permeability. Although the above equation is valid, it gives the
Even though there is a closed set of equations, it is false impression that the saturation is the driving
often hard to get the simulations to converge, es- force for fluid flow and that a saturation condition
pecially if they are multidimensional. Thus, the should be used as a boundary condition. Further-
computational-fluid-dynamics models57,79,80,85 use the more, care must be taken in the interpretation of the
multiphase mixture model.58,59 capillary diffusivity.
As mentioned above, this approach treats each
phase as a constituent to a mixture. Thus, all 4.4. Catalyst-Layer Modeling
parameters are mixture parameters and must be Electrochemical reactions take place at the catalyst
averaged, usually by the saturation. Unlike the layers of the fuel cell. At the anode and cathode,
models mentioned at the end of the previous section, hydrogen is oxidized (eq 1) and oxygen is reduced (eq
the models here use capillary phenomena. Further- 2), respectively. These layers are often the thinnest
more, although the mixture moves at a mass-average in the fuel-cell sandwich but are perhaps the most
velocity, interfacial drag between the phases and complex because this is where electrochemical reac-
other conditions allow each separate phase velocity tions take place and where all of the different types
to be determined. The liquid-phase velocity is found of phases exist. Thus, the membrane and diffusion
by57,59 media models must be used in the catalyst layer
along with additional expressions related to the
Fm kλL(1 - λL) electrochemical kinetics on the supported electro-
vL ) λL vm + [∇pC + (FL - FG)g] (51)
FL 0FLνm catalyst particles.
A schematic of a typical fuel-cell catalyst layer is
where the subscript m stands for the mixture, Fk and shown in Figure 9, where the electrochemical reac-
νk are the density and kinematic viscosity of phase tions occur at the two-phase interface between the
k, respectively, and λL is the relative mobility of the electrocatalyst (in the electronically conducting phase)
liquid phase and the electrolyte (i.e., membrane). Although a
three-phase interface between gas, electrolyte, and
kr,L/νL electrocatalyst has been proposed as the reaction site,
λL ) (52) it is now not believed to be as plausible as the two-
kr,L/νL + kr,G/νG
phase interface, with the gas species dissolved in the
electrolyte. This idea is backed up by various experi-
In eq 51, the first term represents a convection term,
mental evidence, such as microscopy, and a detailed
and the second comes from a mass flux of water that
description is beyond the scope of this review. Ex-
can be broken down as flow due to capillary phenom-
perimental evidence also supports the picture in
ena and flow due to interfacial drag between the
Figure 9 of an agglomerate-type structure where the
phases. The velocity of the mixture is basically
electrocatalyst is supported on a carbon clump and
determined from Darcy’s law using the properties of
is covered by a thin layer of membrane.162,237-240
the mixture. The appearance of the mixture velocity
Sometimes a layer of liquid water is assumed to exist
is a big difference between this approach and the
on top of the membrane layer, and this is discussed
others, and it could be a reason the permeability is
in section 4.4.6. Figure 9 is an idealized picture, and
higher for simulations based on the multiphase
mixture model.
The overall gain of the multiphase mixture model
approach above is that the two-phase flow is still
considered, but the simulations have only to solve
pseudo-one-phase equations. Problems can arise if
the equations are not averaged correctly. Also, the
pseudo-one-phase treatment may not allow for pore-
size distribution and mixed wettability effects to be
considered. Furthermore, the multiphase mixture
model predicts much lower saturations than those of
Natarajan and Nguyen56,87 and Weber and Newman74
even though the limiting current densities are com-
parable. However, without good experimental data Figure 9. Idealized schematic of the cathode catalyst layer
on relative permeabilities and the like, one cannot (going from z ) 0 to z ) L) between the membrane and
say which approach is more valid. cathode diffusion medium showing the two main length
Finally, some of the models use an equation of the scales: the agglomerate and the entire porous electrode.
Gray, white, and black indicate membrane, gas, and
type52,80,85 electrocatalyst, respectively, and the gray region outside
of the dotted line in the agglomerate represents an external
Nw,L ) -DS∇S (53) film of membrane or water on top of the agglomerate.
the actual structure is probably more of a “spaghetti above reasons. Finally, the anode can almost always
and meatball” structure, where the carbon agglomer- be modeled as a simplified cathode model, with the
ates are connected to each other and covered by thin exception of poisoning of the electrocatalyst. For these
tendrils of membrane. reasons, the discussion below focuses mainly on the
Various modeling approaches have been used for cathode models; the anode models require only a
the catalyst layers, with different degrees of success. switch in the kinetic expression and various property
The approach taken usually depends on how the values.
other parts of the fuel cell are being modeled and In this section, the reactions and general equations
what the overall goal of the model is. Just as with for the catalyst layers are presented first. Next, the
membrane modeling, there are two main classes of models are examined starting with the interface
models. There are the microscopic models, which models, then the microscopic ones, and finally the
include pore-level models as well as more detailed simple and embedded macrohomogeneous ones. Fi-
quantum models. The quantum models deal with nally, at the end of this section, a discussion about
detailed reaction mechanisms and elementary trans- the treatment of flooding is presented.
fer reactions and transition states. They are beyond
the scope of this review and are discussed else- 4.4.1. General Governing Equations
where, along with the issues of the nature of the
Although the various models contain their own
electrocatalysts.151,241-243
equations, they are derived from basically the same
The other type of model is the macrohomogeneous set of governing expressions, regardless of the effects
model. These models are macroscopic in nature and, being modeled. As mentioned, the reactions at the
as described above, have every phase defined in each anode and cathode are termed the HOR and ORR,
volume element. Almost all of the models used for respectively. For the HOR, the reaction is fast and
fuel-cell electrodes are macrohomogeneous. In the can be described by a Butler-Volmer kinetic expres-
literature, the classification of macrohomogeneous sion, eq 12. However, if the feed stream is reformate
models is confusing and sometimes contradictory. To or contains poisons such as carbon monoxide, the
sort this out, we propose that the macrohomogeneous reaction rate is quite different. In this situation, the
models be subdivided on the basis of the length scale carbon monoxide adsorbs to the electrocatalyst sites
of the model. This is analogous to dimensionality for and effectively decreases the reaction rate. There
the overall fuel-cell models. have been various models that account for this by
In this scheme, the first type of model is the 0-D- doing a carbon monoxide site balance and examining
type models or interface models, where a single the reaction steps involved.17,104,113,115,244-248 Of these
equation is used without regard to the structure of models, the one by Springer et al.104 is probably the
the catalyst layers. Next are the 1-D-type models, most comprehensive and does the best job in terms
which either account for changes across the layer or of agreement with experimental data. The treatment
account for only local effects in the agglomerate. of the effect of poisons on kinetics is beyond the scope
Typically, those that account for changes across the of this review and is not discussed further. For the
layer are called porous-electrode, macrohomogeneous, ORR, a Tafel expression, eq 13, is normally used due
or thin-film models, and those that account for to the slow kinetics of the four-electron-transfer
changes in the agglomerate are called agglomerate reaction. As discussed in section 3.2.2, either eq 13
models. However, the agglomerate models are still or eq 15 can be used, with the difference being in the
macrohomogeneous models, and they may include overpotential used (i.e., whether the surface and
effects across the layer as well. In the same fashion, concentration or just the surface overpotential is
some porous-electrode models include agglomerate- used, respectively). As discussed, this difference
scale interactions. The models that include effects on basically corresponds to the placement of the, per-
only one length scale are termed simple. They are haps imaginary, reference electrode.
further subdivided on the basis of the length scale In general, catalyst layers contain many phases:
modeled, such that porous-electrode and agglomerate liquid, gas, different solids, and membrane. Because
correspond to the length scales of the layer and the layers have multiple phases, effective values
agglomerate, respectively. Those models that include must be used for the transport properties such as
effects on both length scales (i.e., 2-D-type models) membrane conductivity; this is typically done using
are known as embedded macrohomogeneous models. a Bruggeman expression (see eqs 41 and 42). Because
Finally, there are no 3-D-type models because there of the complexity of the layers, it is worthwhile to do
are only two primary length scales (the catalyst layer a variable and equation count. In the most general
and the agglomerate) in these regions. case, the membrane, diffusion medium, and kinetic
An examination of the catalyst-layer models re- equations are used. This gives the set of variables
veals the fact that there are many more cathode and equations as listed in Table 1. It should be noted
models than anode ones. In fact, basically every that since there are multiple phases in the catalyst
electrode-only model is for the cathode. This arises layer, appropriate volume fractions and Bruggeman
because the cathode has the slower reaction; it is expressions are required for such equations as Ohm’s
where water is produced, and hence, mass-transfer law in the membrane phase. Another point is that
effects are much more significant; and it represents the complete set of equations is seldom used in
the principal inefficiency of the fuel cell. In other modeling. Often, simplifying assumptions are used.
words, while the cathode model can be separate from For example, the ohmic drop in the matrix or solid
the anode model, the converse is not true due to the is often ignored due to its high conductivity. Also,
Table 1. List of Important Variables and Their (eq 23) to simplify it. For example, if the ORR is the
Governing Equations or Conditions for the Catalyst only reaction that occurs at the cathode, the following
Layer
mass balance results
variable equation
∇‚NO2,G )
() ( )
overall liquid water flux NL 23
overall membrane water flux Nw 23
gas-phase component flux NG,i 23 1 pO2 RcF 1
gas-phase component partial pG,i 43 - a1,2i0ORR ref exp - (ηORR,1-2) ) ∇‚i1
4F pO2 RT 4F
pressure
liquid pressure pL 49 (56)
membrane water chemical µw 33 or 34
potentiala
electronic-phase current density i1 38 This equation is often used in the various cathode
membrane current density i2 32 or 28 models.
electronic-phase potential Φ1 27 Before discussing the models, a note should be
membrane potential Φ2 55 or 64 made concerning catalyst loading. Many models use
temperature T set or 76 or 78 platinum loading in their equations, especially for
total gas pressure pG set or 44
liquid saturation S see section 4.3.2.1 optimizing designs and in normalizing the current
produced (equivalent to a turnover frequency in
a
This can be directly related to the liquid pressure depend- catalysis). In this respect, the catalyst loading, mPt,
ing on what equation or approach is used.
is the amount of catalyst in grams per geometric area
of the fuel-cell face (x,y in Figure 1). If a turnover
treatment of the membrane phase is usually done frequency is desired, the reactive surface area of
using Ohm’s law and ignoring the water flux through platinum, APt, can be used (usually given in m2/g).
the membrane. While Table 1 lists all of the govern- This area can be related to the radius of a platinum
ing variables and equations, it does not necessarily particle, assuming perhaps a certain roughness fac-
say how to relate the variables to each other, since tor, but more often it is experimentally inferred using
more than one variable may occur in a given equa- cyclic voltammetry measuring the hydrogen adsorp-
tion. tion. These variables can usually be determined and
The membrane and diffusion-media modeling equa- then used to calculate the specific interfacial area
tions apply to the same variables in the same phase between the electrocatalyst and electrolyte,
in the catalyst layer. The rate of evaporation or
condensation, eq 39, relates the water concentration mPtAPt
in the gas and liquid phases. For the water content a1,2 ) (57)
L
and chemical potential in the membrane, various
approaches can be used, as discussed in section 4.2. where L is the thickness of the catalyst layer. This
If liquid water exists, a supersaturated isotherm can assumes a homogeneous distribution of electrocata-
be used, or the liquid pressure can be assumed to be lyst in the catalyst layer.
either continuous or related through a mass-transfer A factor closely related to the catalyst loading is
coefficient. If there is only water vapor, an isotherm the efficiency or utilization of the electrode. This tells
is used. To relate the reactant and product concen- how much of the electrode is actually being used for
trations, potentials, and currents in the phases electrochemical reaction and can also be seen as a
within the catalyst layer, kinetic expressions (eqs 12 kind of penetration depth. To examine ohmic and
and 13) are used along with zero values for the mass-transfer effects, sometimes an effectiveness
divergence of the total current (eq 27). factor, E, is used. This is defined as the actual rate
The kinetic expressions result in transfer currents of reaction divided by the rate of reaction without any
that relate the potentials and currents in the elec- transport (ionic or reactant) losses. With this intro-
trode (platinum on carbon) and electrolyte (mem- duction of the parameters and equations, the various
brane) phases as well as govern the consumption and modeling approaches can be discussed.
production of reactants and products. To simplify the
equations and approaches for the case of one ionically 4.4.2. Interface Models
and one electronically conducting phase, it is useful The simplest way to treat the catalyst layers is to
to use the relation assume that they exist only at the interface of the
diffusion media with the membrane. Thus, they are
∇‚i2 ) -∇‚i1 ) a1,2ih,1-2 (55) infinitely thin, and their structure can be ignored.
This approach is used in complete fuel-cell models
where -∇‚i1 represents the total anodic rate of where the emphasis of the model is not on the
electrochemical reactions per unit volume of electrode catalyst-layer effects but on perhaps the membrane,
and ih,1-2 is the transfer current for reaction h the water balance, or multidimensional effects. There
between the membrane and the electronically con- are different ways to treat the catalyst layer as an
ducting solid (i.e., eqs 12 and 13 for the HOR and interface.
ORR, respectively). The above charge balance as- If a detailed potential is not required in the model,
sumes that faradaic reactions are the only electrode then the catalyst layer can be treated as simply the
processes; double-layer charging is neglected (as is location where oxygen and hydrogen are consumed
appropriate under steady-state conditions). This equa- and water is produced. Hence, Faraday’s law (eq 24)
tion can be used in the conservation-of-mass equation is used as a generation/consumption term in the
boundary condition for the mass balance between the length of the catalyst layer, and reaction takes place
membrane and the diffusion medium. The models at their surface. These models are termed the gas-
that focus mainly on water management, a quantity pore models.249-252 The second type of model follows
tied to the current density, use this type of treatment the scheme of Grens253 and Cutlip.2 In this approach,
of the catalyst layers.55,81,112,126,134 A slightly more there are still gas pores, but there are also pores
sophisticated treatment is to model the catalyst filled with electrolyte and catalyst. It is in these other
layers as described above but then use an overall pores that reaction, diffusion, and migration occur.
polarization equation (e.g., eq 21) with the simulation These models are termed the flooded-agglomerate
results to yield a potential for the cell at the specific models.3-5,254-257 Although there are the two ap-
current density.14,65,74,86,127,203 proaches, the distinction between them is often
A more sophisticated and more common treatment blurred.
of the catalyst layers still models them as interfaces The equations used in these models are primarily
but incorporates kinetic expressions at the interfaces. those described above. Mainly, the diffusion equation
Hence, it differs from the above approach in not using with reaction is used (e.g., eq 56). For the flooded-
an overall polarization equation with the results, but agglomerate models, diffusion across the electrolyte
using kinetic expressions directly in the simulations film is included, along with the use of equilibrium
at the membrane/diffusion medium interfaces. This for the dissolved gas concentration in the electrolyte.
allows for the models to account for multidimensional These models were able to match the experimental
effects, where the current density or potential findings such as the doubling of the Tafel slope due
changes,16,24,46-48,51,52,54,56,60-62,66,80,82,87,107,125 although to mass-transport limitations. The equations are
some of the earlier and water-management models amenable to analytic solution mainly because of the
are 1-D and treat the catalyst layer in this man- assumption of first-order reaction with Tafel kinetics,
ner.20,57,75,109 This treatment means that both Fara- which means that eq 13 and not eq 15 must be used
day’s law and the kinetic expressions are used as for the kinetic expression. The different equations
boundary conditions at the diffusion media and and limiting cases are described in the literature
membrane interfaces. This type of treatment also models as well as elsewhere.258,259
allows for nonuniform current density distributions Of these models, the flooded-agglomerate one shows
to exist, because the potential is constant in a cell better agreement with experimental data. This is not
due to the equipotential surfaces of the bipolar plates. unexpected, since it probably models the actual
Another way to look at these models is that the microstructure better and also has more parameters.
surface overpotential and reactant gas concentrations A problem with the single-pore models is that in
are assumed uniform throughout the catalyst layers, reality there are multiple pores that are tortuous.
and this is used in the kinetic expressions to calculate Furthermore, the driver of having analytic expres-
the current density. In fact, it is common for these sions becomes less important, as computer power has
models to use the overpotential as a fitting param- progressed. Overall, the single-pore models represent
eter. a good first start in simulating fuel-cell electrodes and
Overall, the interface models are basically 0-D. form the core of many more complicated models as
They assume that all of the relevant variables in the discussed below. However, macrohomogeneous and
catalyst layers are uniform in their values across the more sophisticated models provide for more physi-
layer. This has some justification in that the catalyst cally realistic simulations. Finally, while these mod-
layers are very thin, and it is adequate if other effects els do a good job for phosphoric-acid fuel cells, it is
that are modeled are more significant; however, the unknown how well they work for polymer-electrolyte
catalyst layers should be modeled in more detail to fuel cells where, since the electrolyte is a solid, it does
ensure that all the relevant interactions are ac- not necessarily penetrate the pore space.
counted for and to permit optimization of such
parameters as catalyst loading. The only other truly microscopic models for
the catalyst layers are those by Durand and co-
4.4.3. Microscopic and Single-Pore Models workers.30-33,260 In these models, spherical agglomer-
ate structures are assumed to exist in regular 3-D
The earliest models of fuel-cell catalyst layers are hexagonal arrays. Between the agglomerates, either
microscopic, single-pore models, because these mod- there are gas pores or the region is flooded with
els are amenable to analytic solutions. The original electrolyte. The reason these models are microscopic
models were done for phosphoric-acid fuel cells. In and not macrohomogeneous is that they examine
these systems, the catalyst layer contains Teflon- interactions in such a way that there is a dependence
coated pores for gas diffusion, with the rest of the on the exact way in which the agglomerates are
electrode being flooded with the liquid electrolyte. placed. Furthermore, not all of the phases are defined
The single-pore models, like all microscopic models, in every volume element. The equations solved are
require a somewhat detailed microstructure of the mainly Ohm’s law and Fick’s law with kinetic ex-
layers. Hence, effective values for such parameters pressions. The results of the models show the con-
as diffusivity and conductivity are not used, since centration contours around a particle and agree with
they involve averaging over the microstructure. experimental current densities and trends. Such a
There are two main types of single-pore models. model also allows for the detailed placement of the
In the first, the approach of Giner and Hunter1 is electrocatalyst particles to be studied, even though
taken in which there are straight, cylindrical gas it may not be possible to make such an arrangement
pores of a defined radius. These pores extend the experimentally. To expand on the last comment, the
benefit of these microscopic models is that they rigorous, the charge-transfer resistances in Figure 10
clearly show how the HOR occurs next to the mem- should be in parallel with a capacitor representing
brane and how the packing and structure of the double-layer charging. However, this can be neglected
agglomerate particles help to enhance or reduce the for the steady-state operation of fuel cells and intro-
overall efficiency of the anode, something that the duced if transients or impedance is studied.
classical models above could not demonstrate. The governing equations for the simple porous-
electrode models are as referenced in Table 1. The
4.4.4. Simple Macrohomogeneous Models characteristic length scale is the thickness of the
There are many models that use essentially only catalyst layers. There are also some variations in the
one length scale but average over the microscopic treatment of the simple porous-electrode models. In
details of a region. These models are termed simple the first set of models, the catalyst layers are
macrohomogeneous models. As described in section integrated over, and the expression is included as a
3.1, the macrohomogeneous approach assumes that boundary condition in calculating the cell potential
all phases exist at all points in the volume and that or current density.37,95,99 This treatment is very
properties and phases can be categorized by a hand- similar to that of the interface models, except that
ful of parameters including the volume fraction and potential drops for the matrix and solution phases
surface area per volume. Although these models do are accounted for. The reason for doing the integra-
not include microstructural details, they do use many tion is that, if the reaction distribution is mainly
of the same equations and physical ideas of the uniform, then there is no loss in numerical accuracy
single-pore models. The two main length scales and there is a gain in numerical stability and easier
studied are over the whole catalyst layer or over the convergence.
agglomerate (see Figure 9), and each are discussed The next set of models treats the catalyst layers
below. using the complete simple porous-electrode modeling
4.4.4.1. Porous-Electrode Models. The porous- approach described above. Thus, the catalyst layers
electrode models are based on the single-pore models have a finite thickness, and all of the variables are
above, except that, instead of a single pore, the exact determined as per Table 1 with a length scale of the
geometric details are not considered. Euler and catalyst layer. While some of these models assume
Nonnenmacher261 and Newman and Tobias7 were that the gas-phase reactant concentration is uniform
some of the first to describe porous-electrode theory. in the catalyst layers,15,67,69 most allow for diffusion
Newman and Tiedemann6 review porous-electrode to occur in the gas phase.13,25,26,28,45,71,72,110,121,133,204
theory for battery applications, wherein they had only These models are essentially macrohomogeneous
solid and solution phases. The equations for when a versions of the single-gas-pore models.
gas phase also exists have been reviewed by Bockris The final simple macrohomogeneous porous-elec-
and Srinivasan262 and DeVidts and White,263 and trode models are the ones that are more akin to thin-
porous-electrode theory is also discussed by New- film models. In these models, the same approach is
man139 in more detail. taken, but instead of gas diffusion in the catalyst
Porous-electrode theory is concerned about the layer, the reactant gas dissolves in the electrolyte and
overall reaction distribution in the catalyst layer. moves by diffusion and reaction.17,36,50,64,101,114,264 The
Thus, it is assumed that the main effects do not occur governing equations are more-or-less the same as
within the agglomerates. In other words, the ag- above, except that now a concentration instead of a
glomerates all have a uniform concentration and partial pressure appears in the kinetic expressions
potential. Conceptually, porous-electrode theory can and the governing equations for mass transport of
be visualized as a resistor network, as shown Figure the reactant and product gas species become ones of
10. Physically, the figure is showing that electron, diffusion in the membrane or water (if a water layer
proton, and kinetic resistances govern the reaction is assumed to exist). The reason these models are still
distribution. Thus, the overpotential and transfer simple models is that only the length scale of the
current in the electrode are functions of position catalyst layer is considered, and the concentrations
because the current travels along the path of least of the species are assumed to be in equilibrium with
resistance. The effect of concentration is accounted their respective gas-phase partial pressures (i.e.,
for in the calculation of the charge-transfer resist- Henry’s law applies).
ance, which is derived directly from the kinetic
expressions and likely to be nonlinear. Finally, to be As mentioned, the reaction distribution is the main
effect on the catalyst-layer scale. Because of the facile
kinetics (i.e., low charge-transfer resistance) com-
pared to the ionic resistance of proton movement for
the HOR, the reaction distribution in the anode is a
relatively sharp front next to the membrane. This can
be seen in analyzing Figure 10, and it means that
the catalyst layer should be relatively thin in order
Figure 10. Resistor-network representation of porous- to utilize the most catalyst and increase the efficiency
electrode theory. The total current density, i, flows through of the electrode. It also means that treating the anode
the electrolyte phase (2) and the solid phase (1) at each
respective end. Between, the current is apportioned on the
catalyst layer as an interface is valid. On the other
basis of the resistances in each phase and the charge- hand, the charge-transfer resistance for the ORR is
transfer resistances. The charge-transfer resistances can relatively high, and thus, the reaction distribution
be nonlinear because they are based on kinetic expressions. is basically uniform across the cathode. This means
Figure 11. Tafel plot of flooded porous-electrode simulation results for the cathode at three different values of ψ )
eff
2.3nFDO 2,2
cO2,2|z)L/bκeff. The z coordinate ranges from 0 (catalyst layer/membrane interface) to L (catalyst layer/diffusion
medium interface), the dimensionless overpotential is defined as η* ) -RcF/RT(ηORR,1-2), and the ORR rate constant is
ref
defined as k* ) i0ORR/nFcO 2,2
. (Reproduced with permission from ref 36. Copyright 1998 The Electrochemical Society, Inc.)
that simplifying assumptions such as a uniform their results in terms of ohmic, kinetic, or mass-
surface overpotential (ηsORR,1-2) can be justified, and transfer control situations. Like Perry et al., they
basically all of the catalyst is being utilized in the discuss how simple experiments such as oxygen gain,
layer as long as the mass-transfer of oxygen does not the gain in performance with oxygen instead of air,
become limiting. In this case, the charge-transfer can help determine a given fuel cell’s controlling
resistance next to the membrane becomes much phenomena. Eikerling and Kornyshev28 also show the
higher than the other ones, and the catalyst is less same behavior, and they analyze the resistances with
utilized in that region, since there is less oxygen. One analytic expressions for the various limiting cases.
should note that with Tafel kinetics, which prevail Finally, Jaouen et al.98 examine similar cases to those
for the ORR, the nonuniformity of the reaction above but include agglomerate effects, as discussed
distribution is governed by the current density and in detail in section 4.4.5.
electrode thickness and not by the exchange current 4.4.4.2. Agglomerate Models. The simple macro-
density. homogeneous agglomerate models consider only ef-
The above effects can also be seen in the different fects that occur on the agglomerate length scale. In
regimes of the polarization curve. To study these essence, they assume a uniform reaction-rate distri-
effects, Perry et al.36 made Tafel plots of their bution, that is, a uniform gas concentration and
simulation results, which used a flooded porous- surface overpotential through the thickness of the
electrode model of the cathode catalyst layer. These catalyst layer. As mentioned above, the simple ag-
plots are shown in Figure 11, where they are a glomerate models more accurately represent the
function of a dimensionless parameter that is es- structure of the catalyst layers than the simple
sentially a ratio of oxygen mass transfer to ionic porous-electrode models. These models are very
conduction. The Tafel plots show a doubling of the similar to the microscopic models of Durand and co-
Tafel slope when either the mass-transfer or ionic workers discussed above, except that the geometric
limitations are controlling, regions 2 and 3, respec- arrangement is averaged over and each phase exists
tively. In region 1, kinetics is controlling. Figure 11 in each volume element.
allows one to understand and visualize how the For the agglomerate model, the characteristic
interplay between the different phenomena and length scale is the radius of the agglomerate, Ragg,
variables, such as current, gas diffusion both inside and all of the agglomerates are assumed to be the
and outside the catalyst layer, and proton conduction same shape and size. In the model, the reactant or
inside the catalyst layer, affects polarization behav- product diffuses through the electrolyte film sur-
ior. Although the model is for the cathode, it is rounding the particle and into the agglomerate,
equally valid for the anode, where the parameter where it diffuses and reacts. Hence, there is a
values normally result in kinetic control (region 1). concentration and possibly a potential distribution
The results of Perry et al. allow for the develop- within the agglomerate. The equations for modeling
ment of a diagnostic method that determines the the agglomerate are similar to those listed in Table
limiting behavior from simple experimental data. In 1, except that either spherical or cylindrical coordi-
a similar analysis, Weber et al.133 added mass trans- nates are used for the gradients. The reason they are
port in the diffusion media explicitly and analyzed not identical is that the agglomerate scale is es-
sentially a subscale or microscopic scale in the calculated. To do this, an expression for the diffusion
catalyst layer. This is the same as examining a bed of oxygen to the surface of the agglomerate is written
of sand, where the normal resolution looks at changes
bulk surf
across the box and the agglomerate scale looks at cO - cO
diff 2 2
changes within individual grains (see Figure 9). In a WO ) AaggDO2,film (63)
macrohomogeneous model, it is necessary to average 2 δfilm
over the agglomerate scale, and hence, effectiveness
diff
factors, both internal and overall,265 are often used. where WO 2
is the molar flow rate of oxygen to the
The analysis below is given for the ORR, since the agglomerate, Aagg is the specific external surface area
agglomerate and embedded models mainly examine of the agglomerate, and the film can be either
the cathode; reaction at the anode can be derived in membrane or water (if two or more films are desired,
a similar manner. The analysis is basically the same similar expressions can be written for each film). The
as that of reaction and diffusion in a catalyst pellet. above expression uses Fick’s law and a linear gradi-
For the analysis, an effectiveness factor is used, ent, which should be valid due to the low solubility
which allows for the actual rate of reaction to be of oxygen, steady-state conditions, and thinness of
written as (see eq 55) the film. At steady state, the above flux is equal to
the flux due to reaction and diffusion in the ag-
∇‚i2 ) a1,2ih,1-2E (58) glomerate (as well as the flux through any other
films), and thus, the unknown surface concentra-
Since the ORR is a first-order reaction following Tafel tion(s) can be replaced. Doing this and using the
kinetics, the solution of the mass conservation equa- resultant expression in the conservation equation (eq
tion (eq 23) in a spherical agglomerate yields an 56) yields
( )
analytic expression for the effectiveness factor of149,265
bulk 1
1 ∇‚i1 ) 4FcO (64)
E ) 2(3φ coth(3φ) - 1) (59) 2 δfilm 1
3φ +
AaggDO2,film k′E
where φ is the Thiele modulus for the system266
This equation is the governing equation for the
x
k′ agglomerate models for the cathode, and without
φ)ζ eff
(60) external mass-transfer limitations, it results in eq 58.
DO 2,agg For the anode, a similar analysis can be done.
There is only a handful of models that simulate
where ζ is the characteristic length of the agglomer- the catalyst layers using the simple agglomerate
ate (volume per surface area), Ragg/3 for spheres, model,44,90,105,120,160,267,268 because most agglomerate
Ragg/2 for cylinders, and δagg for slabs, and k′ is a rate models are developed for comparison purposes (dis-
constant given by cussed in the following section) or are macrohomo-
( )
geneous embedded models (i.e., take into account the
a1,2i0ORR RcF catalyst-layer length scale as well) and are discussed
k′ ) exp - (η ) (61) in section 4.4.5. The results of the simple agglomerate
ref
4FcO RT ORR,1-2
2 models are helpful in trying to understand and
optimize catalyst-layer parameters such as loading
where the reference concentration is that concentra- and agglomerate size. For example, it has been shown
tion in the agglomerate that is in equilibrium with that fuel-cell performance increases with a decrease
the reference pressure in agglomerate radius until a maximum plateau is
ref ref reached.90,105 This limit results because the effective-
cO 2
) pO 2
HO2,agg (62) ness factor plateaus at a value of 1 at a finite radius.
However, this result does not necessarily hold if the
where HO2,agg is Henry’s constant for oxygen in the agglomerate is not at a uniform temperature through-
agglomerate. Similar expressions to eq 59 have also out.
been derived for other types of reactions and geom- 4.4.4.3. Model Comparison. The distinction be-
etries,149,265 and while eq 59 is derived for spheres, tween the two simple macrohomogeneous models
with the appropriate Thiele modulus, it results in deals with the length scales of the effects being
deviations less than 10% in the value of the effective- studied. While both show similar effects such as
ness factor for other geometries. The above rate doubling of the Tafel slope due to mass-transport
constant can be related to the transfer current limitations, the question arises as to where the
density, iORR,1-2 (see eq 13). Also, one notices that the limitations are occurring. To get a better grasp on
Thiele modulus is independent of the surface con- the two different approaches, several researchers
centration for a first-order reaction. have compared them to each other and experimental
If external mass-transfer limitations can be ne- data.27,34,35,49,53 Of particular note is the analysis of
glected, then the surface concentration in eq 58 (via Boyer et al.,49 who examined the characteristic length
eq 13) can be set equal to the bulk concentration, scales for the various processes such as diffusion in
which is assumed uniform throughout the catalyst the gas phase, diffusion in the agglomerate, proton
layer in the simple agglomerate models. Otherwise, migration in the catalyst layer, and so forth. These
the surface concentration is unknown and must be length scales are simple expressions and can let one
simple thin-film model. As can be seen, the ag-

glomerate model fits the data much better than the
thin-film model, although the lack of a gas phase in
the thin-film model makes it less rigorous than the
simple porous-electrode model described above. The
difference between the models declines at higher
current densities where the reaction distribution is
becoming nonuniform. In fact, for anode simulations
the porous-electrode model should be better due to
the much more nonuniform reaction distribution. At
low current densities, the reaction distribution is
uniform, and the most important effects happen at
the agglomerate scale; that is why the agglomerate
model fits the data better in that regime. While these
simple models are acceptable, a more complex em-
bedded model where both are used should be the best
in terms of accuracy.
4.4.5. Embedded Macrohomogeneous Models
Figure 12. Comparison of simple macrohomogeneous The above simple models account for only one
agglomerate (solid line) and thin-film (dashed line) cathode
models to experimental data (diamonds). Data are adapted length scale. To account for both the local agglomer-
from ref 34. ate level and effects across the porous electrode,
embedded macrohomogeneous models are used. Even
easily identify which scale is more important for the though the simple agglomerate models account for
conditions being studied. Hence, one will know which many effects, they cannot truly consider reaction
model is the best to use. In general, they arrived at distributions and proton migration across the catalyst
the conclusions that the simple agglomerate model layer. As noted above, these effects are important in
is accurate for the cathode but that a porous-electrode certain circumstances (e.g., within the anode catalyst
approach is better in the anode due to the highly layer, within a flooded catalyst layer, and within a
nonuniform reaction distribution. They also showed catalyst layer with a dehydrated membrane). The
how to optimize various parameters such as Nafion application of embedded models is very straightfor-
loading. ward. Finally, the original model of Iczkowski and
The rest of the comparisons were done for the Cutlip5 was really a precursor of the embedded
cathode. The results all showed that the agglomerate models, and agglomerate models rather than porous-
model fits the data better than the porous-electrode electrode ones are more likely to become embedded
model. However, it should be noted that the porous- models.
electrode model used was usually a thin-film model There are two ways in which the embedded model
and so was not very robust. Furthermore, the ag- can be used. In the first, the porous-electrode model-
glomerate model has more parameters that can be ing equations are used as discussed above and in
used to fit experimental data. Finally, some of the Table 1. Next, a mass-transfer term is added to
agglomerate models compared were actually embed- account for the flux to the agglomerate. In this
ded models that account for both length scales, and situation, there is basically a film between the
therefore, they normally agree better with the ex- reaction site and the gas pore, although this exact
perimental data. structure is averaged over in the macrohomogeneous
The main problem with the porous-electrode mod- approach. Thus, in a single volume element, the
els according to Gloaguen et al.34,35 is that they diffusive and reaction fluxes are related at the
overestimate oxygen transport limitations, but that boundary of the agglomerate film. This approach is
is because they used thin-film models that did not basically the same as accounting for external mass-
contain gas pores. Chan and Tun53 compared much transfer limitations within each volume element.
more similar agglomerate and porous-electrode mod- Because this approach adds only an additional re-
els and noted the effects of changing different pa- sistance to charge transfer due to diffusion, it is not
rameters on performance. They showed that the more computationally costly than the porous-elec-
agglomerate model is more sensitive to changes in trode model and yields an extra fitting parameter if
mass-transport resistances caused by flooding than desired (the thickness of the film or the mass-transfer
the porous-electrode models. A final comparison was coefficient) and more realistic behavior, especially for
done by Broka and Ekdunge.27 They compared the incorporating flooding effects. There are a few models
models to experimental data and microscopy and that use this approach.19,84,85
arrived at the conclusion that the agglomerate model The other approach is more complicated and re-
is a more accurate representation, although they also quires a deeper knowledge of the agglomerate struc-
showed that with the addition of a mass-transfer ture or yields more fitting parameters. In this ap-
coefficient in the porous-electrode model it can agree proach, the porous-electrode equations are used, but
with the data as well as the agglomerate model. now the effectiveness factor and the agglomerate
A comparison of the two models with experimental model equations are incorporated. Hence, eq 64 is
data is given in Figure 12. In the figure, simulations used to get the transfer current in each volume
were run with a simple agglomerate model and a element. The gas composition and the overpotential
more parameters, not all of which may be known. If

the controlling phenomena are known, which seldom
are a priori, then the appropriate simple model can
be used. Diagnostic techniques aimed at determining
the controlling phenomena and perhaps parameter
values are discussed by Perry et al.,36 Weber et al.,133
and Jaouen et al.98 They advocate conducting simple
experiments, such as oxygen gain and varying layer
thickness, and analyzing the results. The analysis
and results of their models can be summarized as
follows for the different operating regimes. Control
by oxygen diffusion in the agglomerate results in a
double Tafel slope, and the total current density is
proportional to the active layer thickness, is first-
order in oxygen, and has a low sensitivity to gas
humidity. Control by proton transport in the catalyst
Figure 13. Plot of cathode potential as a function of layer results in a double Tafel slope, and the current
current density for a macrohomogeneous embedded model density is independent of the layer thickness, is half-
where the proton conductivity is assumed to be uniform order in oxygen, and increases with relative humid-
(0.044 S/m), curve a, or varies with water production ity. Control by both of the above results in a qua-
(changing humidity) across the catalyst layer, curve b. druple Tafel slope. Control by oxygen diffusion in
(Reproduced with permission from ref 98. Copyright 2002 both the agglomerate and covering film results in the
The Electrochemical Society, Inc.)
progressive disappearance of quadruple and double
Tafel slopes with increasing film thickness. Control
change across the catalyst layer due to ohmic, mass- by oxygen diffusion in the layer and either proton
transfer, and reaction effects. While this approach diffusion in the layer or oxygen diffusion in the
is slightly more complicated and requires knowing agglomerate yields a double Tafel slope that also
values for more parameters, the payoffs are greater progressively disappears.
in that it does the best job in representing the
physical reality of the catalyst layer and includes all 4.4.6. Catalyst-Layer Flooding
of the relevant effects. This is in agreement with the
analysis of Pisani et al.,91 who compared different Flooding of the catalyst layer is treated in a
agglomerate structures with each other and the separate section because various models with differ-
porous-electrode model. Other models also use this ent approaches have addressed it and it is important,
approach.36,78,98,129,269 Finally, it should be noted that especially in the cathode. There are two main ways
an overall effectiveness factor that considers both in which flooding is considered, and they depend on
external and internal mass transfer to and through how the catalyst layer is modeled. If an agglomerate
the agglomerate can be used as a fitting parameter, model is used and if liquid water exists in the fuel
thereby avoiding the necessity of detailed calculations cell, a liquid film covering the membrane film of the
and perhaps multiple fitting parameters on the agglomerates can be assumed. Thus, the flooding of
agglomerate scale. the catalyst layer is easily incorporated into the
external mass-transfer limitation (eqs 63 and 64)
In terms of the cathode models, that of Maja et
where the reactant gases have to diffuse through a
al.269 in essence examined the effect of having two
water film as well as the membrane layer.50,114 Due
different agglomerate radii. They showed that this
to the low diffusivity and solubility of oxygen in
is important for being able to predict the correct
water, only a very thin liquid film is needed to inhibit
limiting current density. Pisani et al.78 used analytic
reaction. All of the agglomerate models can easily
expressions for cylindrical agglomerates that are
incorporate flooding in this manner, with perhaps the
basically flooded pores. They then integrated over the
thickness of the film used as a fitting parameter.
catalyst layer to get an overpotential and current
The other way to incorporate flooding in the
density for use as a boundary condition in the overall
catalyst layers is to use the two-phase modeling
1-D fuel-cell model. As mentioned above, Jaouen et
approach described in section 4.3.2.1. This involves
al.98 examined the cathode polarization curve in a
calculating the liquid saturation in the catalyst layer
manner similar to those of Perry et al.36 and Weber
and then adjusting the interfacial area accordingly.
et al.133 However, the addition of the agglomerate
This is done by a linear expression
length scale allowed them to examine more complex
behavior. This can be seen in Figure 13, where the a1,2 ) a°1,2(1 - S) (65)
embedded model shows a more complex cathode-
potential curve due to tradeoffs between the two where a°1,2 is the maximum or dry specific interfacial
length scales. In the figure, consideration of a chang- area. There are a few models that use this ap-
ing ionic conductivity due to water production hy- proach.52,56,57,87,109 Equation 64 assumes a homoge-
drating the membrane causes the potential curve to neous distribution of electrocatalyst through the
move from a double Tafel slope back toward a single catalyst layers, which is a fine first approximation,
Tafel slope. although accounting for the electrocatalyst location
While it is evident that the embedded model is the (e.g., hydrophilic versus hydrophobic pores) would be
best one to use, it requires additional complexity and more rigorous.
In comparing the two approaches, it seems that the though analytic solutions are obtainable in certain
saturation approach allows for greater reaction rates instances, these usually involve assumptions that
(higher current densities). The reason is that the make the solution of limited significance. Further-
catalyst layers have small pores and are hydrophobic, more, the power of digital computers is continually
and thus, it takes a high liquid pressure to flood progressing such that the computational cost of
them, whereas even a thin film can effectively shut running simulations becomes manageable.
down the reaction. Of course, the film is spread over For most numerically solved models, a control-
a much larger surface area and depends on the volume approach is used. This approach is based on
agglomerate radius. It is tough to say which approach dividing the modeling domain into a mesh. Between
is better, but because the saturation equations for mesh points, there are finite elements or boxes. Using
the diffusion media are well developed compared to Taylor series expansions, the governing equations are
having an unknown water film thickness that may cast in finite-difference form. Next, the equations for
vary with position, it is our belief that eq 65 should the two half-boxes on either side of a mesh point are
be incorporated into an embedded catalyst-layer set equal to each other; hence, mass is rigorously
model. conserved. This approach requires that all vectors be
It is known that flooding is important at high defined at half-mesh points, all scalars at full-mesh
current densities, especially at the cathode. However, points, and all reaction rates at quarter-mesh points.
one may wonder if the performance loss is due to The exact details of the numerical methods can be
blockage of the catalyst sites by water or by flooding found elsewhere (for example, see ref 273) and are
of the diffusion media. In the literature models, not the purview of this review article. The above
flooding of the diffusion media has garnered more approach is essentially the same as that used in CFD
attention. Flooding of the diffusion media seems also packages (e.g., Fluent) or discussed in Appendix C
to agree with the fact that the diffusion media are of ref 139 and is related to other numerical methods
much thicker and more hydrophilic with larger applied to fuel-cell modeling.274,275
hydrophobic pores than the catalyst layers, but they The various layers of the fuel-cell sandwich de-
also have a higher porosity. Furthermore, due to the scribed above are linked to each other through
normally high surface area of active material in the boundary conditions, which apply at the mesh point
catalyst layers, some flooding can occur without a between two regions. There are two main types of
significant detriment to performance.123 Experimen- boundary conditions, those that are internal and
tal data with microporous layers between the catalyst those that are external. The internal boundary condi-
layer and diffusion media (for examples, see refs tions occur between layers inside the modeling
270-272) are inconclusive. Some indicate that, due domain, and the external ones are the conditions at
to their often hydrophobic nature, they keep water the boundary of the entire modeling domain.
out of the diffusion media. However, other micro-
porous layers are thought to wick water out of the Typically, coupled conditions are used for internal
catalyst layers due to their small pores. In reality, boundaries wherein the superficial flux and intersti-
the physical situation is that flooding should be tial concentration of a species are made continuous.
considered in both regions. Overall, the catalyst However, as mentioned above, boundary conditions
layers are complex and have been treated by various between the membrane and electrode can involve the
means, with macrohomogeneous embedded models fact that there is only ionic current in the membrane
agreeing the best with the physical picture, while also and that the uptake isotherms mean that water
not being too complicated to use in fuel-cell simula- content changes from λ in the membrane to partial
tions. pressure in the electrode. On the other side of the
electrode, the boundary condition should state that
4.5. Multilayer Simulations all of the current is electronic. Another common
boundary condition is to have a change in concentra-
The purpose of this section is to describe the tion because a species dissolves. This is similar to
general results of models that contain more than one the internal boundary condition in the membrane
of the layers described above. It is beyond the scope and is used sometimes for electrodes where phases
of this article to analyze every model and its results are not continuous across the boundary. Finally,
in detail, especially since they have already been internal boundary conditions can represent modeling
discussed to a certain extent in section 2. Many of regions that are not modeled in depth. For example,
the models make tradeoffs between complexity, di- a water flux and kinetic equation can be used at the
mensionality, and what effects are emphasized and boundary between the cathode diffusion medium and
modeled in detail. It is worth noting that those the membrane if the catalyst layer is not modeled
models that employ a CFD approach seem to be the rigorously. Another example is setting the flux of
best suited for considering multidimensional effects. water through the membrane and its ohmic resist-
In this section, the ways in which the multilayer ance at a single boundary point.
models are solved and connected are discussed first. Depending on the modeling domain, the last bound-
Next, some general trends and results are presented. ary condition is sometimes also used as an external
boundary condition (e.g., half-cell models). The ex-
4.5.1. Numerical Solution and Boundary Conditions ternal boundary conditions specify the concentrations
Due to the complexity and interconnectivity of the and values for all of the species and variables at the
governing equations and constitutive relationships, boundary. Examples include specifying the inlet
most fuel-cell models are solved numerically. Al- conditions, such as gas feed rates, composition,
temperature, and humidity, or specifying the current

density or potential. The external boundary condi-
tions are often the same as the operating conditions
and, therefore, are very similar for most simulations,
although there can be differences such as what
condition is used for two-phase flow (i.e., zero satura-
tion or zero capillary pressure).
4.5.2. General Multilayer Simulation Results
In terms of sandwich models, there are four main
varieties. The first are those that treat only one layer
in the sandwich, and they were discussed above. The
second are those that treat multiple layers of the
sandwich but not all of them.26,37,50,56,57,75,86,105,110,120,153,252
These are typically cathode models that include
the diffusion medium and perhaps a mem-
brane water flux. Next are the models that
treat all of the layers of the sandwich and are only
1-D.13,14,17,20,45,71,72,74,78,95,98,99,101,112,203 Finally, there are
those that are multidimensional and treat all of the
layers; these are discussed in detail in section 5.
The models should duplicate the effects seen by
experiment. Thus, as the temperature is raised,
better performance is obtained up to a point. An
optimum is reached because, even though property
values are usually increasing, so does the water vapor
pressure. Consequently, if the feeds are saturated,
the reactant gas partial pressure is lower, which
affects the kinetics, and if the feeds are dry, it is
harder to hydrate the membrane and there are
higher ohmic losses. These issues can become impor-
tant during operation of high-temperature fuel cells.
As discussed in section 5.1.1, water concentration
effects down the gas channels need to be considered
for dry feeds, and therefore, a 1-D model does not Figure 15. Simulation results showing membrane dehy-
suffice. Gains in performance are also seen in fuel dration (a) and cathode flooding (b). (a) λ as a function of
cells operating with pure oxygen or under elevated membrane position (cathode on the left) for different
pressure. In these cases, fuel cells perform better not current densities. (Reproduced with permission from ref
only because of higher reactant gas pressures but also 14. Copyright 1991 The Electrochemical Society, Inc.) (b)
Dimensionless oxygen mole fraction as a function of
because flooding is not as critical (i.e., the limiting cathode-diffusion-medium position and cathode overpoten-
current density increases due to lower mass-trans- tial. (Reproduced with permission from ref 120. Copyright
port limitations). 2000 The Electrochemical Society, Inc.)
This last point is shown very well in Figure 14,
where model and experimental data are compared
for different oxygen concentrations. In some sense,
this is an analogue to Figure 4, where a 1-D and not
a 0-D model is used. Figure 14 demonstrates good
fits between experiments and 1-D model simulations.
In this model, flooding was accounted for in the
diffusion media and the simple agglomerate model
was used for the catalyst layers. The membrane
model considered only the liquid-equilibrated trans-
port mode, which is why the ohmic region in some of
the curves deviates from the experimental data.
Overall though, the 1-D sandwich model does a good
job in fitting the data for humidified feeds and in
predicting the full polarization curve, especially at
low oxygen concentrations.
Most of the models show that fuel-cell performance
is a balance among the various losses shown in
Figure 14. Comparison between experiment and simula- Figure 3, in particular, ohmic losses and mass-
tion for different humidified oxygen feeds. (Reproduced transport limitations, which both increase with cur-
with permission from ref 78. Copyright 2002 The Electro- rent. The reason for this is that the kinetic losses are
chemical Society, Inc.) hard to mitigate without significantly changing op-
erating conditions or using better catalysts, as dis-

cussed above. Models show that ohmic losses mainly
result from the membrane drying out at the anode
side of the membrane-electrode assembly. This is
shown in Figure 15a. This is not to say that ohmic
losses do not occur in other parts of the fuel cell, just
that they are not as dominant. The mass-transport
limitations are due mainly to flooding of the cathode
side of the fuel cell, as shown in Figure 15b. The
answers that models predict to the above tradeoffs
focus mainly on increasing the flux of water in the
membrane from cathode to anode and increasing the
oxygen partial pressure in the cathode. To these
extents, models agree with experiment in showing
that thinner membranes, higher-oxygen-content feeds
and pressures, and the use of a microporous layer
between the catalyst layer and a gas-diffusion layer Figure 16. Schematic showing the modeling domain for
all yield better performing fuel cells. the 2-D models. (a) Along-the-channel domain where the
Finally, fuel-cell models allow for optimization 1-D sandwich is highlighted for the pseudo-2-D case. (b)
studies. For example, understanding the current Rib domain showing only the cathode side rib and channel.
distribution in the electrode allows one to understand
the effect of changing its thickness and how to 5.1.1. Along-the-Channel Models
decrease catalyst loading by proper platinum place-
There are two main ways that transport along
ment. Such modeling results can be credited for the
the channel can be treated. The first is to assume
manufacturing of fuel cells that have thinner anodes
that changes within the fuel-cell sandwich occur
than cathodes due to more of a nonuniform reaction
only in the direction perpendicular to it (z
distribution in the anode. Other optimization studies
axis).15,16,44,55,107,109,127,129,134 This assumption effec-
include looking at different structural and physical
tively means that the models have only to account
properties such as thicknesses, porosities, and so
for a change in the boundary conditions or values
forth. However, optimization studies should be con-
along the channel. This kind of treatment is termed
fined to those models that are more physical and
pseudo-2-D. A limiting case of the pseudo-2-D model
complex. Overall, the best fuel-cell sandwich models
may be seen as those models that are 1-D but use an
have a physical basis; take into account all of the
average concentration for each species in the gas
sandwich layers; have a minimum of fitting param-
channels.14,74,88,101 The other kind of treatment is to
eters; agree with experimental data trends; and ade-
use a full 2-D simulation in which equations and
quately model the dominant transport phenomena.
parameters change in both directions in the sandwich
and gas channels.25,47,61,62,64,66,79,90 With proper dis-
5. Multidimensional Effects cretization, the pseudo-2-D models can accurately
reflect the results of the full 2-D models.
Although the fuel-cell sandwich is the heart of a
fuel cell, there are important effects that are not For a full 2-D model, the only change in the
found when only a 1-D model is used. These effects equations is that the vector components now occur
basically arise from the fact that a fuel cell is in in two directions. The various regions are normally
reality a 3-D structure, as shown in Figures 1 and 5. assumed to be isotropic so that the property-value
Many models explore these effects and are discussed expressions are the same for each direction. If data
in this section. These models always include the fuel- are known, then the expressions can be altered
cell sandwich as one of the dimensions. First, the 2-D accordingly. The gas channels are the only new
models are examined and then the 3-D ones. Because modeling domains introduced.
there is another review in this issue that focuses on The gas channels contain various gas species
these models in more depth, the discussion below is including reactants (i.e., oxygen and hydrogen),
shortened.276 products (i.e., water), and possibly inerts (e.g., nitro-
gen and carbon dioxide). Almost every model assumes
that, if liquid water exists in the gas channels, then
5.1. Two-Dimensional Models it is either as droplets suspended in the gas flow or
As shown in Figures 1 and 5, there are two possible as a water film. In either case, the liquid water has
ways that 2-D effects can be incorporated, either no affect on the transport of the gases. The only way
along the channel (y,z) or under the rib (x,z). The it may affect the gas species is through evaporation
modeling domains of these two effects are shown or condensation. The mass balance of each species is
parts a and b, respectively, of Figure 16. The rib obtained from a mass conservation equation, eq 23,
effects take into account the fact that there is a solid where evaporation/condensation are the only reac-
rib between two channels, which affects transport in tions considered.
the sandwich. This latter effect is germane to inter- For movement of the gases, various treatments can
digitated flow fields. Both of the 2-D effects are be used. The simplest is to use mass balances down
discussed below, starting with the along-the-channel the flow channel. This is the same as saying that
models. there is plug flow and the pressure remains uniform,
something that is not valid under certain operating

conditions or with long or interdigitated flow fields.
A robust treatment would be to use the Navier-
Stokes equation
vG‚∇(FGvG) ) -∇pG + µG∇2vG (66)
where vG is the mass-average velocity of the gas

phase. For the gas channels, Darcy’s law is not valid
because they are not porous. If desired, diffusion
terms (e.g., Stefan-Maxwell equations) can be used
in conjunction with the above equation to result in
convective-diffusion equations. However, axial dis-
persion is normally negligible compared to pressure-
driven flow. Finally, simpler flow equations are also
often used such as assuming plug flow or Poiseuille
flow. If only a pressure drop down the channel is
required, correlations from fluid dynamics can be
used to determine it (e.g., Ergun equation).149,277
In terms of boundary conditions, the following
apply. The gas-channel plates represent equipotential Figure 17. Mole fraction of water vapor, hydrogen, and
surfaces, which serve to set the overall cell potential. oxygen in the gas channels at a cell potential of 0.72 V, at
In terms of flows, the gases are usually assumed to a temperature of 80 °C, and in a coflow arrangement. The
be either in a coflow or counterflow arrangement, local current density is shown by the solid line. (Reproduced
with permission from ref 15. Copyright 1993 The Electro-
although the physical system may be in cross-flow. chemical Society, Inc.)
The inlet flow rates and concentrations must be
specified; setting the stoichiometry of the inlet gases the figure shows is that these effects are significant,
and the humidity serves to do this. The stoichiom- and if one were to use a 1-D model, these effects
etries are normally based on the total current of the would be lost. Fuller and Newman also show that the
cell, and when they are time independent, the results membrane can dehydrate along the channel due to
are independent of the active area or the channel nonisothermal effects.
length. The reason for this is that all the fluxes Because of gas composition changes along the
become proportional to the active area, which can be channel, a 2-D model is necessary to describe dry
canceled out. Setting the outlet pressure and the inlet feeds. This is true since not only does the oxygen
temperature also serves as boundary conditions. A concentration vary, but, perhaps more importantly,
no-slip condition at the gas channel wall is also valid. the water does as well. This is seen in Figure 17,
The conditions needed at the interfaces of the gas where the increase in current density is due to the
channels and the diffusion media are now coupled, increase in water in the system. To study this further,
internal boundary conditions, which were discussed Janssen55 and Weber and Newman134 ran simulations
previously. where the inlet gases were dry and the flow was
Now that the methodology and equations have countercurrent. They studied the water balance of the
been introduced, one may wonder what the effects fuel cell, and some of their results are given in Figure
along the channel are. The most significant effects 18. Figure 18a clearly shows that, with counter-
deal with the changing gas composition. As the feed current operation, the water is recycled in that the
gases react, their concentration goes down unless the water produced near the air inlet helps to keep the
feed stoichiometries are very high. This means that membrane hydrated near the dry hydrogen inlet. It
the reaction rate decreases and a nonuniform current is noted that Figure 18a is basically a composite of
distribution is generated. In fact, with very long the cases where one gas is fed dry and the other
channels, the decrease in concentration of the feed humidified. The 2-D membrane hydration is shown
gases might result in a type of limiting current in Figure 18b. This figure clearly shows Schroeder’s
density.66 Such a distribution can have a variety of paradox in the sharp drop off in the water content,
effects including nonuniform temperature distribu- where the drop signifies the point where liquid water
tions, places where the membrane dries out, and and the liquid-equilibrated transport mode cease to
poorly utilized catalyst areas. exist in the membrane. It also shows that the highest
To expand on the last remark, the simulation water content is near the cathode inlet and that,
results from Fuller and Newman15 are shown in although the water is recycled, dehydration problems
Figure 17. The curves clearly show a nonuniform can arise near the hydrogen inlet. However, these
current distribution that is mainly due to the change simulations assumed uniform current distributions,
in the gas concentrations and the membrane hydra- which is not valid but affects the potential of the cell
tion. In the simulation, the initial decrease in the more than the water balance. Figure 18 also shows
current density is due to the change in the oxygen that, due to the slower diffusion of water vapor in
concentration. However, once enough water is gener- air than in hydrogen, it is more important for the
ated to hydrate the membrane, the increased con- anode to be humidified. Other models have also
ductivity yields higher local current densities. What reached this conclusion.16,25,44,47,129
these conditions include low stoichiometry feeds, low

humidity feeds, and temperature effects at high
current densities. Furthermore, for the cases outlined
above, such flow is the most important multidimen-
sional effect. Although full 2-D models do not neces-
sarily have to be done, flow along the channels needs
to be accounted for in some fashion.
5.1.2. Under-the-Rib Models

As shown in Figure 16b, the 2-D rib models deal
with how the existence of a solid rib affects fuel-cell
performance.24,37,46,56,57,126,278 They do not examine the
along-the-channel effects discussed above. Instead,
the relevant dimensions deal with the physical reality
that the gas channel/diffusion media interfaces are
not continuous. Instead, the ribs of the flow-channel
plates break them. These 2-D models focus on the
cathode side of the fuel-cell sandwich because oxygen
and water transport there have a much more signifi-
cant impact on performance. This is in contrast to
the along-the-channel models that show that the
underhumidification of and water transport to the
anode are more important than those for the cathode.
As mentioned, the 2-D rib effects are crucial
in understanding transport in interdigitated flow
fields.108,232 In these flow fields, the gas channels are
not continuous, and thus, the gases must pass
through the diffusion media. The benefit of such a
configuration is that water removal and reactant gas
distribution are better because of the convective flow
of gases through the diffusion media. However, such
flow also means that there is a much higher pressure
drop. Some models, of note those of Nguyen and
Figure 18. Pseudo-2-D simulation results at 0.4 A/cm2
where the feed gases are dry and countercurrent. (a) Water co-workers,48,52 Um and Wang,76 Zhukovsky,77 and
partial pressure profiles at four positions in the fuel-cell Kazim et al.,46 have specifically looked at interdigi-
sandwich as a function of distance along the channel; the tated flow fields. Their results show the importance
positions are at the anode and cathode gas channels (I and of having more gas channels with smaller widths,
IV) and catalyst layers (II and III), respectively. Also among other things. Interdigitated flow fields are not
plotted is the value of β, the net flux of water per proton discussed further in this review.
flux, as a function of position. The data are from Janssen.55
(Reproduced with permission from ref 55. Copyright 2001 The equations for modeling the 2-D rib effects
The Electrochemical Society, Inc.) (b) Membrane water require a domain where the boundary conditions in
content as a function of position both along the gas channel terms of gas flow and composition are specified only
and through the thickness of the membrane for the same at the channel. At the solid rib, there is no flux of
simulation conditions as above. The data are from Weber gas and liquid, but all of the electronic current must
and Newman.134 (Reproduced with permission from ref 55
and 134. Copyright 2004 The Electrochemical Society, Inc.)
pass through it. Furthermore, the modeling domain
is usually as shown in Figure 16b; thus, only a half
The 2-D models also allow for different effects to channel and rib is modeled, and symmetry conditions
be studied in depth. For example, the simulations can be used to model the other half. Besides those
show that, as mentioned above, countercurrent flow noted above, the boundary conditions and equations
yields better performance than coflow due to water are more-or-less the same as those discussed in
recycling.16,25,44,47,109 This is something that a 1-D section 4.
model cannot address. The 2-D models also have As can be seen in the different boundary conditions,
studied pressure differentials and design strategies the main effects of having ribs are electronic conduc-
to help boost performance.64,107,109 Finally, the 2-D tivity and transport of oxygen and water, especially
models allow for optimization studies to be conducted, in the liquid phase. In terms of electronic conductiv-
such as minimizing cost and examining the effects ity, the diffusion media are mainly carbon, a material
of fuel utilization,127 and seeing how structural that is fairly conductive. However, for very hydro-
variables change the performance.16,25,44,47,79 This last phobic or porous gas-diffusion layers that have a
is perhaps better addressed with a complex 1-D small volume fraction of carbon, electronic conductiv-
model rather than a simpler 2-D one. ity can become important. Because the electrons
Overall, effects along the channel are significant leave the fuel cell through the ribs, hot spots can
and need to be considered. This is especially true in develop with large gradients in electron flux density
cases where the composition in the gas channel is next to the channel.37,124 Furthermore, if the conduc-
expected to vary significantly. Some examples of tivity of the gas-diffusion layer becomes too small, a
nonuniform reaction distribution can be generated

with the highest rate occurring across from the
ribs.37,124
Although usually the transport of electrons is not
limiting, the transport of oxygen can be. For example,
at high overpotentials or low oxygen concentrations,
the oxygen has a much higher concentration across
from the channels and not the ribs.37,56,278 What this
means is that once again a nonuniform reaction
distribution is generated, which results in a nonuni-
form current distribution and all of the associated
effects with it (e.g., hot spots, flooding, membrane
dehydration, etc.). Under these circumstances, mass
transport is an important consideration. Oxygen dif-
fusion is expected to cause the above problems, which
are exacerbated by the presence of liquid water.
Figure 18 shows that, even when the cathode is
relatively dry, liquid water exists in the cathode.
What the 2-D rib models show is that this liquid
water cannot be effectively removed if it is next to
the rib because it has no way to get out of the
cell.24,56,126 In fact, this effect is perhaps more impor-
tant in generating nonuniform reaction-rate distri-
butions. Simulation results from Natarajan and
Nguyen,56 shown in Figure 19, clearly demonstrate
the entrainment of liquid water in the gas-diffusion
layer next to the rib. In fact, Figure 19a shows that
the diffusion medium next to the rib is fully flooded.
They correlated a more uniform reaction-rate distri-
bution with a decrease in the extent of flooding.
Furthermore, they showed that these effects become
less important at higher cell potentials because the
fuel cell operates in the kinetic regime in which mass-
transport limitations are only minor (see Figure 3).
It is clear that the ribs can have a detrimental
effect on fuel-cell performance by creating a nonuni-
form current distribution due to poor mass transfer
of oxygen. This is shown in detail in Figure 19b,
where the current distribution is shown for different
rib and channels sizes. Clearly, the smaller channels Figure 19. Liquid saturation and current density of the
and ribs provide more uniform current distributions. cathode as a function of position for the case of dry air fed
at 60 °C. (a) Liquid saturation in the gas-diffusion layer
This is because there are smaller regions of flooding, where the channel goes from x ) 0 to 0.05 cm and the rib
since the water has more pathways to leave the gas- is the rest; the total cathode overpotential is -0.5 V. (b)
diffusion layer and enter the channel. This effect has Current-density distributions for different channel/rib ar-
also been modeled by others.24,126 It is also interesting rangements. (Reproduced with permission from ref 56.
that, although the gas-diffusion layer is flooding, the Copyright 2001 The Electrochemical Society, Inc.)
ribs also are making sure the membrane is better
hydrated due to the increased liquid pressure through- perhaps catalyst layers in terms of placing catalyst
out the gas-diffusion layer. Consequently, there is a only where higher reaction rates are expected (i.e.,
tradeoff between decreased ohmic losses in the minimize catalyst loading).
membrane and mass-transfer limitations in the
cathode diffusion medium. For dry systems, ribs 5.2. Three-Dimensional Models
might actually help performance, while, for saturated The 3-D models try to capture all of the effects
systems, they only hinder it. mentioned above, in both 2-D directions as well as
Overall, the rib effects are important when examin the 1-D sandwich.51,54,60,76,77,80,82,87,119,124,125,279-281
ining the water and local current distributions in a The domain is as shown in Figure 1. 3-D models have
fuel cell. They also clearly show that diffusion media the potential to accurately represent the true opera-
are necessary from a transport perspective. The effect tion of a fuel cell. In principle, these models are the
of flooding of the gas-diffusion layer and water ones that should be used to obtain the best designs
transport is more dominant than the oxygen and and optimization of various properties and operating
electron transport. These effects all result in non- conditions. However, while the current published
uniform reaction-rate distributions with higher cur- models are complex on an overall global scale, they
rent densities across from the channels. Such analy- are usually not very detailed on the 1-D sandwich
sis can lead to optimized flow fields as well as scale. For example, almost all of these models have
neither a robust membrane model nor a rigorous two-

phase model. In addition, validation of these models
is usually done by comparing only one polarization
curve (normally because of the computational costs
of the models), which may or may not be representa-
tive of the regimes in which the models are run.84
In terms of modeling, the equations are the same
as those in section 4, with perhaps some simplifica-
tions. Additional boundary conditions are required
due to the higher dimensionality of the equations,
but these are relatively straightforward, such as no
fluxes of gas species across the external boundary of
the gas channels.
The general results of the 3-D models are more-
or-less a superposition of the 2-D models discussed
above. Furthermore, most of the 3-D models do not
show significant changes in the 1-D sandwich in a
local region. In other words, a pseudo-3-D approach
would be valid in which the 1-D model is run at
points in a 2-D mesh wherein both the channel and
rib effects can easily be incorporated. Another pseudo-
3-D approach is where the 2-D rib models are used Figure 20. Local current density contours at a cell poten-
and then moved along the channel, similar to the tial of 0.6 V for cocurrent flow of dry air and humidified
hydrogen at 65 °C. (Reproduced with permission from ref
cases of the pseudo-2-D models described above.87 82. Copyright 2003 The Electrochemical Society, Inc.)
This latter approach is similar to that by Baker and
Darling.281 In their model, they uncouple the different concentrations. In the same fashion, the 3-D models
directions such that there is a 1-D model in the gas also show that ribs are beneficial if the anode is fed
channel and multiple 2-D rib models. However, they dry because of better membrane hydration, and
neither treat the membrane nor have liquid water. again, this mainly occurs in an entrance region.81 3-D
In all, the use of CFD means that it is not signifi- simulations with and without liquid water clearly
cantly more complicated to run a complete 3-D model show that, for humidified feeds, liquid water must
in all domains. be accounted for, especially if operating in the mass-
The 3-D models show that the flooding problems transfer regime of the polarization curve.84,85
associated with the ribs and the oxygen depletion One of the benefits of a 3-D model is the ability to
along the channel compound each other.69,82,87 Thus, examine the effect of flow-field design. For example,
flooding is worse across from the ribs, and the Kumar et al.279,280 showed that having hemispherical
reaction rate distribution is more nonuniform. How- channels provides more convective flow through the
ever, as mentioned above, for nonhumidified feeds, gas-diffusion layers and better overall fuel-cell per-
the effects down the channel are more significant. formance. Furthermore, they showed that there are
This is seen in Figure 20, where the local current benefits to having flow channels that have a satu-
density can be associated with the oxygen concentra- rated permeability comparable to that of the diffusion
tion. Although the rib scale is a little hard to discern, media. Once again, in these cases there is more
there is a definite change in the current density along convective-type flow through the diffusion media.
the length scale. The highest current density is Both of the above designs try to mimic interdigitated
obtained at the inlet because the cathode is less flow channels to a certain extent. Baker and Dar-
flooded and the oxygen concentration is highest. The ling281 showed that there is more convective flow
current density drops and then increases near the through the diffusion media farther away from the
exit due to the increase in the back flux of water from U-turns in serpentine flow fields. They also demon-
cathode to anode and a fully hydrated membrane. strated that at high current densities the flow field
The figure clearly shows that there is an entrance acts more as a single pass than at low current
region with large nonuniformities. The compounding densities where the concentration is more uniform
of the 2-D effects shown by the 3-D models also due to higher transport through the diffusion media
results in the 2-D models underpredicting the water between adjacent channels.
removal rate. This means that saturations are higher The inclusion of multidimensional effects is impor-
than expected in the diffusion media.85 tant to realistically mimic transport in the fuel cell.
In addition, the models demonstrated that the This is not to say that certain cases and factors
water balance is important and can alter potential cannot be collapsed to lower dimensionality, but one
and reactant location.51,54,60 For example, the reaction must be aware of higher dimensional effects, lest they
distribution is closer to the channel than the rib near become important.
the inlet regions for humidified feeds. The reason is
that, as the water moves from anode to cathode, there 6. Other Effects
is more flooding, but the movement of water in this To complete the discussion of the macroscopic fuel-
direction decreases along the channel because of cell models, two other effects should be examined.
lower current densities associated with lower oxygen These effects can be important, have not really been
discussed so far, and alter the system of governing The addition of the temperature component comes
equations. The first effect is taking into account from analysis leading to the Gibbs-Duhem equa-
transport and generation of heat in the fuel cell. The tion.282 In addition to the above driving force, expres-
reaction is an oxidation process that generates heat, sions must be added that result in thermal diffu-
mainly because of irreversibilities. Water condensa- sion.139 However, because thermal diffusion is small
tion and evaporation are also important heat sources relative to other effects such as conduction, and there
and sinks. The second effect is transient operation are small gradients in temperature in fuel cells, this
of a fuel cell. This is important in many instances type of transport can be neglected. Similarly, the
such as automotive applications, where fuel cells are effect of diffusion on heat flux, the Dufour effect, is
expected to undergo transients, especially during likewise neglected.139 In addition, heats of mixing for
acceleration, start up, and shut down. ideal gases are zero, and for the other components
they are assumed negligible.283 Overall, because the
6.1. Nonisothermal Models above effects are neglected, if one follows the deriva-
The discussion of nonisothermal effects has tions given in section 4, then one obtains the same
been delayed until this section because the set of governing equations. Thus, the equations
models that account for such effects constitute a reviewed above remain valid for use in nonisothermal
relatively small subset of the total number of models systems, as long as the above assumptions are
and have already been discussed above in other accepted.
contexts.15,16,25,26,45,60,71,80,82,84-86,90,107,125,126 Further- As mentioned, to include nonisothermal effects, an
more, the complex (i.e., multidimensional) models are overall thermal energy balance needs to be added to
the ones that usually take into account temperature the set of governing equations. The energy conserva-
effects. One result of the nonisothermal models is tion equation can be written for phase k in the
that the temperature distribution in the 1-D fuel-cell form139,149
( )
sandwich is basically uniform (around 1 °C or so),
and thus, the isothermal assumption is usually valid ∂Tk
for the sandwich. The cases where larger deviations FkĈpk + vk‚∇Tk +
∂t
( ) ( )
from a uniform temperature distribution occur are
in the inlets and outlets of the fuel cell and also ∂ ln Fk ∂pk
depend on operating conditions, such as feed-gas + vk‚∇pk ) Qk,p - ∇‚qk -
humidity. ∂ ln Tk pk,xi,k ∂t
When discussing nonisothermal models, the focus
is on those models that account for heat generation, τ:∇vk + ∑Hi,k∇‚Ji,k - ∑Hi,kRi,k (68)
consumption, and transfer. A distinction is made i i
between this type of model and those that have In the above expression, the first term represents the
properties and parameters that depend on temper- accumulation and convective transport of enthalpy,
ature but assume a uniform temperature distribu- where Ĉpk is the heat capacity of phase k. The second
tion. Many of the models discussed above fall into term is energy due to reversible work. For condensed
this latter category and allow for the effects of oper- phases this term is negligible, and an order-of-
ating temperature changes but not a temperature magnitude analysis for ideal gases with the expected
gradient. For most conditions, the change in param- pressure drop in a fuel cell demonstrates that this
eter values with temperature has a more significant term is negligible compared to the others; therefore,
impact than accounting for temperature gradients, it is ignored in all of the models.
although the two are coupled to a certain extent. The first two terms on the right side of eq 68
There are various changes that must be done to represent the net heat input by conduction and
include nonisothermal effects. The most important interphase transfer. The first is due to heat transfer
is the addition of an energy conservation equation between two phases
instead of setting the temperature to a constant; it
is discussed below. The other changes are the in- Qk,p ) hk,pak,p(Tp - Tk) (69)
clusions of the appropriate thermal effects to the
equations described above. To be rigorous, these where hk,p is the heat-transfer coefficient between
effects need to be included, but they are usually phases k and p per interfacial area. Most often this
negligible or are already accounted for. For example, term is used as a boundary condition, since it occurs
a temperature distribution within an agglomerate only at the edges. However, in some modeling do-
can give different values of the effectiveness factor, mains (e.g., along the channel) it may need to be
but due to the size and properties of the agglomerate, incorporated as above. The second term is due to the
it is sufficient to assume that it is at a uniform heat flux in phase k
temperature. Another consideration is that, because
the chemical potential is undefined in a nonuniform qk ) -∑Hi,kJi,k - kTeffk ∇Tk (70)
temperature field, a term must be added to account i
for this. Thus, a thermodynamically rigorous expres-
sion for the driving forces for transport must be used, where Hi,k is the partial molar enthalpy of species
i in phase k, Ji,k is the flux density of species i
di,k ) ci,k[∇µi,k + Si,k∇Tk] (67) relative to the mass-average velocity of phase k
where Si,k is the partial molar entropy of species i. Ji,k ) Ni,k - ci,kvk (71)
and kTeffk is the effective thermal conductivity of The assumption of local thermal equilibrium also
phase k. This means that it was corrected for means that an overall effective thermal conductivity
the volume fraction of the phase by perhaps a is needed, because there is only a single energy
Bruggeman equation (eq 42). equation. One way to calculate this thermal conduc-
The third term on the right side of eq 68 repre- tivity is to use Bruggeman factors,
sents viscous dissipation, the heat generated by
T ) ∑k kTk
keff 1.5
viscous forces, where τ is the stress tensor. This term (74)
is also small, and all of the models except those of k
Mazumder and Cole84,85 neglect it. The fourth term
on the right side comes from enthalpy changes due an expression that assumes that thermal conduction
to diffusion. Finally, the last term represents the in the phases is in parallel. Some other models take
change in enthalpy due to reaction the average as25,80,84,85,125
1
∑Hi,kRi,k ) -∑a1,kih,1-k(ηs + Πh) - h,1-k
keff
T ) -2ks +
0 1 - 0
(75)
i h +
∑ ∆Hlak,prl,k-p - ∑∆HgRg,k (72) 2ks + kG 3kG
p*k g
where kG and ks are the thermal conductivities of the
where the expressions can be compared to those in gas and solid (both conductive and nonconductive
the conservation-of-mass equation (eq 23). The above components) phases, respectively. A final way to
reaction terms include homogeneous reactions, in- calculate an effective thermal conductivity is to
terfacial reactions (e.g., evaporation), and interfacial realize that the thermal conductivity of the solid is
electron-transfer reactions. The latter contains ex- the larger and use that value,26,45,65 although the
pressions for both the reversible and irreversible heat porosity and tortuosity should be considered for the
generation due to electrochemical reaction, as derived different solid phases.
by Newman and Thomas-Alyea.284 The irreversible Most of the models use a simplified analogue of eq
heat generation is represented by the surface over- 68 where eqs 69, 70, and 72 have been substituted
potential. The reason the electrode overpotential, into it and local thermal equilibrium is assumed and
which contains a surface overpotential, is not used the equation is summed over phases. The resultant
is that the generation of heat is due to the reaction equation is then further simplified for fuel cells
( )
at the interface; thus, a reference electrode next to
the interface is used. This point is discussed further ∂T
below. The reversible heat generation is represented ∑FkĈp k
∂t
+ vk‚∇T ) ∑hk,extak,ext(Text - T) +
k k
by the second term, Πh, which is the Peltier coef-
ficient for charge-transfer reaction h285 and can be T ∇T) - ∑∑Ji,k‚∇Hi,k +
∇‚(keff
expressed as284 k i
∆Sh
∑ ∑a1,kih,1-k(ηsORR,1-k + Πh) + ∆Hevaprevap (76)
T 1-k h
Πh ≈
n F
∑si,k,hSi,k ) T n F (73)
where the evaporation rate is given by eq 39. In the
h i h
above equation, the summation over species i in-
where ∆Sh is the entropy of reaction h. The above cludes all reacting species including electrons and
equation neglects transported entropy (hence, the protons, the summation over 1 - k denotes summa-
approximate sign), and the summation includes all tion over all reaction interfaces between phase k and
species that participate in the reaction (e.g., electrons, the electronically conducting solid, and the subscript
protons, oxygen, hydrogen, water). The models that ext denotes heat transfer that is external to the
take into account reversible heating either use values control volume and is normally found only in the gas
for ∆Sh for the two fuel-cell half reactions26,45,71,80,90,125 channels where there may be a heat-transfer plate,
or just do an overall energy balance where the total fluid, or reservoir. For boundary conditions, continu-
fuel-cell reaction entropy is used.15,16,65,107 ity in temperature and thermal flux serve as the
Almost all of the models assume local thermal internal boundary conditions. For the external bound-
equilibrium between the various phases. The excep- ary conditions, the inlet temperature is specified, and
tions are the models of Berning et al.,80,125 who use a at the edges, either an external heat-transfer relation
heat-transfer coefficient to relate the gas temperature is used (if not already added into eq 76) or the wall
to the solid temperature. While this approach may is assumed to be adiabatic. The latter corresponds
be slightly more accurate, assuming a valid heat- to a cell inside a stack, and the former to cells in
transfer coefficient is known, it is not necessarily contact with perhaps coolant plates.
needed. Because of the intimate contact between the In the gas channels, the models have convec-
gas, liquid, and solid phases within the small pores tion of enthalpy, conduction through the graphite
of the various fuel-cell sandwich layers, assuming plates, and heat transfer to the ambient environ-
that all of the phases have the same temperature as ment or cooling fluid if that case is being stud-
each other at each point in the fuel cell is valid. Doing ied.15,16,65,80,90,107,125,126 In the diffusion layers, there is
this eliminates the phase dependences in the above conduction of heat but no convective motion. There
equations and allows for a single thermal energy is also no external heat transfer. However, there is
equation to be written. Joule heating,139,164,286 as discussed below. The cata-
lyst layers have the same thermal effects as the

diffusion media, except that there is also heat gen-
eration, as discussed above. In the membrane, there
is conduction, Joule heating, and some enthalpy that
is carried by diffusion or convection. While most
models account for this, some do not,60,80,84,85,90,125 and
some are only cathode models.26,86,126 Also, some of
the models assume that all of the water from the
membrane vaporizes,60,82,86,90 which consumes a lot
of heat and is not necessarily correct; it depends on
the membrane model being used and the conditions
in the catalyst layer.
Joule heating is due to the passage of current. It
can be derived from the third term on the right side
of eq 76
Figure 21. Temperature distribution in kelvin inside the
i1‚i1 cathode diffusion medium at a current density of 1.2 A/cm2
-∑Ji,1‚∇H
h i,1 ) -i1‚∇Φ1 ) (77) and with saturated feeds. (Reproduced with permission
from ref 80. Copyright 2003 The Electrochemical Society,
i σeff Inc.)
The above is derived by using the definition of the 1-D sandwich is taken to be isothermal and only
current density and relating the gradients of en- the total heat generated in each sandwich layer is
thalpy and electrochemical potential for a system of required.
uniform temperature. This type of heating can simi-
larly be calculated for the ionically conducting phase Equation 78 can be used to show that the concen-
where the ionic conductivity is used instead of the tration overpotential due to gas diffusion does not
electronic conductivity. Joule heating is important in generate heat. To do this, a virtual experiment is
the membrane because a temperature change can setup. First, the total amount of heat generation of
cause an appreciable change in water content if the a complete fuel cell is calculated. Next, one imagines
membrane is close to saturation.41,63,94 Even though that diffusion media are added to the fuel cell in such
it can be important, not all of the models include a way that the partial pressure of the reactant gases
it.26,60,82,86,126 and all ohmic effects remain the same within the
catalyst layers and fuel cell. Hence, only the partial
Equation 76 is valid for local heat generation
pressures of the gases in the gas channels increase.
and analysis. However, there are some models that
According to eq 78, since the conditions are identical
treat only heat transfer in the gas channels and
in the fuel cell, the heat generation is the same (i.e.,
assume that the fuel-cell sandwich remains isother-
no terms change). In other words, there is no heat
mal,15,16,107,129 an assumption that is arguably valid
generation due to this concentration overpotential.
from other simulation results. In this approach, only
This can also be seen if the cell potential is divided
thermal convection and external heat transfer in the
into a summation of various overpotentials, where
gas channels are used, and an energy balance yields
the effect of changing gas concentration in the
the total energy produced in the fuel-cell sand-
diffusion media cancels. However, if the gases are not
wich.164,286,287 This energy balance has the form
ideal, then some heat will be generated through the
Ji,1‚∇Hi,1 term in eq 76.
Q) ∑ ∑ak,pih,k-p(UH h
- V) (78)
The nonisothermal models allow for detailed tem-
k-p h
perature distributions to be generated, as shown in
where Q is the total heat generated per unit volume, Figure 21. In the figure, the temperature is highest
V is the (observed) cell potential, the summations are in the channel and decreases along the channel,
over all the electron-transfer reactions and interfaces, because the rib has a higher thermal conductivity
and UHh is the enthalpy potential for reaction h, and the reaction rate is higher closer to the channel
inlet. These effects have been seen with other mod-
∆Hh ∂Uh els.60,65,80 Although the temperature gradient is not
U Hh ) ) Uh - T (79) large, it is greater in the cell sandwich than along
nhF ∂T the gas channel. The main reason is that condensa-
tion is occurring in the diffusion medium, which
and is a function of temperature. Equation 78 is releases heat in addition to that released at the
simple to use, since the necessary quantities are reaction sites. Whether the temperature is largest
relatively easy to obtain. This is especially true if only along the channel or in the sandwich depends on the
a single fuel-cell reaction occurs, wherein the en- length of the channels, the boundary conditions set
thalpy of the overall reaction can be used directly at the channels (i.e., insulating or conductive), and
with the observed current density per unit volume the operating conditions. For example, with dry-gas
to yield Q. Equation 78 is good if just the amount, feeds, the temperature gradients along the channel
and not the precise location of heat generation, is become larger, and the inlet region becomes suscep-
required. An example is a pseudo-2-D model where tible to membrane dehydration, as seen with some
of the other models (for examples, see refs 15 and One of the first models to examine transients in
60). polymer-electrolyte fuel cells was a stack-level model
Overall, the models show that nonisothermal ef- by Amphlett et al.288 Their model is mainly empirical
fects mainly influence the water balance and current and examines temperature and gas flow rates. They
distribution. The water balance affects the competi- showed that transient behavior lasts for a few
tion between membrane dehydration and mass- minutes in a stack before a new steady state is
transfer limitations, both of which are functions of reached. In a similar stack-level analysis, Yerramalla
temperature. Because water vaporization has a high et al.289 used a slightly more complicated single-cell
enthalpy change, the models that do not take into model and examined the shape of the transients.
account liquid water and evaporation/condensation They noticed voltage behavior that had oscillations
reactions15,84,90,126 cannot explore many significant in it and some leakage current. Their overall analysis
heat effects. On the other hand, models that always was geared to the development of a controller for the
have a fully hydrated membrane are valid for only stack.
specific circumstances where there is always liquid In terms of the more complicated single-cell models,
water.25,80,84,85,125 As discussed in the previous sec- which have been discussed in various contexts above,
tions, this does not occur often, especially at the van Bussel et al.44 mainly examined the transient
anode. Thus, a full membrane model is required, behavior of the water content of the membrane with
especially since membrane dehydration and temper- their model. They clearly showed that coflow opera-
ature rise exacerbate each other.65,128 In addition, tion with dry gases as well as counterflow with dry
pressure and temperature gradients can combine to gases and high stoichiometries all lead to cases where
produce significant effects on water management the membrane dries out and the fuel cell cannot
through evaporation and condensation of water (i.e., operate for more than a few minutes. This analysis
changing the partial pressure and vapor pressure of shows that the drying out of the membrane is a much
water). faster process in an operating fuel cell than with a
To be complete in understanding and modeling naked membrane, where dehydration occurs over a
water management, especially on a multidimensional much longer time scale.290 The modeling results also
or stack level, nonisothermal effects should be in- agreed with the transient uptake model of Nguyen
cluded. However, it is an adequate first approxima- and Vanderborgh,291 which also showed that the
tion to assume that a sandwich and perhaps even a initial development of the membrane water profile
single cell are isothermal, as long as the dependence required a few minutes.
of the various parameters on temperature is included Um et al.64 also examined a transient using their
in the model. Finally, to be rigorous, some of the complex model. They saw that in a matter of tens of
thermal effects assumed to be negligible could be seconds the current density response reached steady
added, such as heats of mixing of species into the state after a change in potential. However, their
membrane, temperature effects in the catalyst ag- model did not include liquid water. The most complex
glomerates, and thermal diffusion. model to examine transients is that of Natarajan and
6.2. Transient Models Nguyen.56 It should be noted that although the model
The other effect considered in this section deals of Bevers et al.26 has transient equations, they do not
with transients in a single fuel cell. The transient report any transient results. Natarajan and Nguyen
models examine step changes in potential and related included liquid saturation effects and water transport
phenomena (e.g., gas flow rates, water production, in their model. They clearly showed the flooding of
and current density). Hence, they are aimed at the diffusion media and that it takes on the order of
examining how a fuel-cell system handles different a couple of minutes for the profiles to develop.
load requirements, which may occur during automo- The above models clearly demonstrated that the
tive operation or start up and shut down. They are transient response of the electrical phenomena in a
not trying to model slow degradation processes that fuel cell is fast. The limiting time constant in the fuel
lead to failure or the transients associated with cell is the liquid-water transport and its associated
impedance experiments (i.e., potential or current effects, which agree with experimental findings.292,293
oscillations). These types of models are discussed in This time constant causes a slow approach to steady
section 7. state that can be on the order of tens of minutes in
There are not many models that do transients, certain circumstances.292 However, the majority of
mainly because of the computational cost and com- the change in the parameters of interest (power,
plexity. The models that do have mainly been dis- current density, etc.) occurs within a much shorter
cussed above. In terms of modeling, the equations use time, and although the approach due to liquid water
the time derivatives in the conservation equations is long, the overall change in the parameter values
(eqs 23 and 68) and there is still no accumulation of is minor. For example, with a change in potential,
current or charging of the double layer; that is, eq the current density will reach a significant fraction
27 still holds. The mass balance for liquid water of its steady-state value in a short time and then
requires that the saturation enter into the time slowly decay toward it.
derivative because it is the change in the water Some other general comments can also be made.
loading per unit time. However, this treatment is not For a stack, the interconnectivity of the cells means
necessarily rigorous because a water capacitance that the overall response will exhibit slower time
term should also be included,234 although it can be constants than those for a single cell. However, the
neglected as a first approximation. stack response (at least in terms of electrical vari-
ables) is still probably faster than the responses of fuel-cell stack may be modeled with only single
the other components in the fuel-cell system such as equations in the form of eqs 21 and 78.100,300 The
compressors and so forth (for example, see ref 294). models show such tradeoffs as compressor power
In the final analysis, as a first try, transient behavior required and performance gain due to pressurized
of a fuel cell can be adequately modeled using a and humidified feeds. In terms of the models, Ahmed
pseudo-steady-state model. This type of approach et al.100,300 considered the total water requirements
would involve stepping the single-cell model through needed both for the fuel cell and for the reformer
various steady states, where the operating conditions reactors. Cownden et al.300 looked mainly at the total
such as potential, current, or flow rates change. Of system power and saw the effects of compressors,
course, the stepping rate would have to be deter- pressure drops, coolant flows, and the efficiencies of
mined. The approach is the same as assuming the various components in the system. Ahluwalia et
instantaneous achievement of steady state, or in al.301 studied the gain in overall stack performance
other words infinitesimal time constants. If more and design when high-temperature polymer-electro-
specific phenomena need to be described such as lyte systems are considered. Finally, the model of De
liquid-water transport, then a full transient model Francesco and Arato294 looks at transients and start-
should be used. Finally, due to load-leveling devices up conditions and procedures. Overall, while the
and possible hybridization of a fuel-cell system, stack models use simplified and usually empirical
transient models should be explored only after the equations for a single cell, they provide benefits in
relevant effects described in the preceding sections actually designing operating systems. In these cases,
(e.g., flooding) are examined, or if specific phenomena the data for the empirical expressions are known, and
such as start up from freeze are being considered. complex modeling may not be required.
The next set of models examined in this section is
7. Other Models impedance models. Impedance is often used to de-
termine parameters and understand how the fuel cell
Other types and aspects of polymer-electrolyte fuel
is operating. By applying only a small perturbation
cells have also been modeled. In this section, those
during operation, the system can be studied in situ.
models are quickly reviewed. This section is written
There are many types of impedance models. They
more to inform than to analyze the various models.
range from very simple analyses to taking a complete
The outline of this section in terms of models is stack
fuel-cell model and shifting it to the frequency
models, impedance models, direct-methanol fuel-cell
domain. The very simple models use a simple equiva-
models, and miscellaneous models.
lent circuit just to understand some general aspects
As mentioned above, stack models are useful for
(for examples, see refs 302-304).
analyzing full system performance including perhaps
auxiliary components in the system such as compres- The next more complicated models focus on analyz-
sors. In terms of equations, almost all of the models ing the impedance spectra using complex equivalent
use simple global balances and equations because circuits.103,137,221,305-312 Most of these models use a
single cells are not the focus of the models; thus, they series of resistors to represent transport in the
use equations similar to eqs 21 and 78. In terms of membrane and diffusion media. The porous elec-
other equations, normally they use typical flow and trodes are modeled similarly to Figure 10, with the
heat balances as well as the appropriate current and addition of a double-layer capacitance term in paral-
voltage relations, which take into account how the lel with each charge-transfer resistance. Most of the
cells are connected together. The stack models can models show such effects as increased high-frequency
be separated into two categories, those that consider resistance as the membrane dehydrates and an
the stack only45,288,289,295-299 and those that consider increased low-frequency loop as flooding occurs. The
a whole power system.100,294,300,301 model of Wang et al.306 also considers how carbon
The models that examine only stacks focus mainly monoxide affects performance in the anode. Out of
on the temperature distribution within the stack. As these models, the one of Eikerling and Kornyshev137
mentioned, there is a much higher temperature is the most sophisticated. It shows specifically how
gradient in the stack than in a single cell, and it ohmic and oxygen limitations affect the impedance
provides design information in terms of coolant flow of the cathode. With such a model, meaningful
rate, among other things.45,296,298 Also, as mentioned analysis of experimental impedance plots can be
above, transient effects have also been examined.288,289 made and governing phenomena noted, as well as the
Lee et al.298,299 examined many different profiles in determination of general transport parameters such
the stack including temperature, relative humidity, as the resistance of the membrane.
pressure, and current density, and their models can While a good equivalent-circuit representation of
be used to predict flooding and various gradients in the transport processes in a fuel cell can lead to an
the stack. Similarly, the model of Thirumalai and increased understanding, it is not as good as taking
White297 examined reactant distribution and design a 1-D sandwich model and taking it into the fre-
of the flow field and manifold. quency domain. These models typically analyze the
As mentioned, there are also models of complete cathode side of the fuel cell.102,313,314 The most com-
fuel-cell power systems in which the stack is but one prehensive is probably that of Springer et al.102 The
component. These models have the benefit of exam- use of impedance models allows for the calculation
ining true designs and the interconnections between of parameters, like gas-phase tortuosity, which can-
components, but they usually require an even weaker not be determined easily by other means, and can
model of an individual fuel cell. In fact, the entire also allow for the separation of diffusion and migra-
tion effects. Overall, impedance is a very powerful length scale means that such effects as Knudsen
experimental tool, but its results are only as good as diffusion and water droplet formation and flooding
the model used to analyze it. can be more important. These models are normally
The next models that should be mentioned are 3-D and focus more on the flow-field design and the
those for direct-methanol fuel cells.68,117,130-132,295,315-325 gas channels than on the complex relationships
A direct-methanol fuel cell is very similar to the within the fuel-cell sandwich.
hydrogen fuel cells in this review, with the exception The final group of models is those that incorporate
of the fuel. In a direct-methanol fuel cell, methanol possible failure mechanisms.330-333 These failure
is fed instead of hydrogen and reacts according to the mechanisms are time or condition dependent and are
reaction not the same as the effects examined in earlier
sections, such as carbon monoxide poisoning, that
CH3OH + H2O f CO2 + 6H+ + 6e- (80) was also valid for steady-state operation. The model
by Darling and Meyers330 examined platinum cata-
lyst loss as the potential is cycled. This is something
and the cathode ORR (eq 2) is the same. The design that has also been observed in phosphoric-acid fuel
of the fuel cell is the same as the polymer-electrolyte cells334,335 and can occur during start up and shut
models described above. The methanol reaction is down. Similar to the case of the Darling and Meyers
much slower than hydrogen oxidation and is a model, Wendt et al.331 examined catalyst aging due
significant source of performance loss for the direct- to changing catalyst morphology and, to a lesser
methanol fuel cells. However, because methanol is a extent, accumulation of impurities. The models of
liquid, the membrane is always fully hydrated. Even Jiang and Chu332 and Fowler et al.333 examined slow
though the membrane has better conductivity, a fully voltage degradation over time for a cell stack and a
hydrated membrane also means that there is a single cell, respectively. They do this by including
significant amount of methanol crossover. This cross- time-dependent rate constants for the kinetic reac-
over is one reason that thicker membranes are used tions as well as a membrane deterioration term that
in methanol fuel cells than in hydrogen ones. Treat- limits the maximum water content of the membrane.
ment of the transport phenomena becomes much Both use semiempirical equations to model the fuel-
more complex due to the presence of methanol130,131 cell behavior, but they are some of the first to model
and also because there is liquid on both sides of the fuel-cell failure. The above models are notable be-
fuel cell. cause durability issues are becoming increasingly
In terms of modeling, all of the different types of significant and have not received much attention in
models discussed above have been used. The major the literature either theoretically or experimentally
problems are methanol crossover, flooding, the kinet- (for examples, see refs 336-338).
ics of the cathode, and predominately mass-transfer
and reaction of methanol at the anode. It is beyond
the scope of this review to discuss all of the models
8. Summary
for direct-methanol fuel cells. The major ones are In this review, we have examined the different
listed in the references and have been mentioned models for polymer-electrolyte fuel cells operating
here only because of their similarity to those dis- with hydrogen. The major focus has been on trans-
cussed in this article. The governing phenomena and port of the various species within the fuel cell. The
equations are basically the same with only slight different regions of the fuel cell were examined, and
variations to take into account methanol in the their modeling methodologies and equations were
membrane, different anode kinetics, and diffusion of elucidated. In particular, the 1-D fuel-cell sandwich
methanol in water instead of hydrogen in water was discussed thoroughly because it is the most
vapor. It should be noted that direct-methanol fuel important part of the fuel-cell assembly. Models that
cells are being considered for portable applications included other effects such as temperature gradients
due to their more energy-dense fuel and simple and transport in other directions besides through the
liquid-injection systems, among other things. fuel-cell sandwich were also discussed.
Finally, there are some miscellaneous polymer- Models were not directly compared to each other;
electrolyte fuel cell models that should be mentioned. instead they were broken down into their constitutive
The models of Okada and co-workers42,43,326 have parts. The reason for this is that validation of the
examined how impurities in the water affect fuel-cell models is usually accomplished by comparison of
performance. They have focused mainly on ionic simulation to experimental polarization data (e.g.,
species such as chlorine and sodium and show that Figure 3). However, other data can also be used such
even a small concentration, especially next to the as the net flux of water through the membrane. In
membrane at the cathode, impacts the overall fuel- fitting these data, the models vary not only in their
cell performance significantly. There are also some complexity and treatments but also in their number
models that examine having free convection for gas and kind of fitting parameters. This is one reason it
transfer into the fuel cell.327-329 These models are also is hard to justify one approach over another by just
for very miniaturized fuel cells, so that free convec- looking at the modeling results. In general, it seems
tion can provide enough oxygen. The models are reasonable that the more complex models, which are
basically the same as the ones above, but because the based on physical arguments and do not contain
cell area is much smaller, the results and effects can many fitting parameters, are perhaps closest to
be different. For example, free convection is used for reality. Of course, this assumes that they fit the
both heat transfer and mass transfer, and the small experimental data and observations. This last point
has been overlooked in the validation of many ∆Hl heat or enthalpy of reaction l, J/mol
models. For example, a model may fit the data very i superficial current density through the membrane,
well for certain operating conditions, but if it does A/cm2
not at least predict the correct trend when one of ik current density in phase k, A/cm2
i0h exchange current density for reaction h, A/cm2
those conditions is changed, then the model is shown
ih,k-p transfer current density of reaction h per interfacial
to be valid only within a certain operating range. area between phases k and p, A/cm2
This review has highlighted the important effects ilim limiting current density, A/cm2
that should be modeled. These include two-phase flow Ji,k flux density of species i in phase k relative to the
of liquid water and gas in the fuel-cell sandwich, a mass-average velocity of phase k, mol/cm2‚s
robust membrane model that accounts for the differ- k effective hydraulic permeability, cm2
ent membrane transport modes, nonisothermal ef- k′ ORR rate constant as defined by eq 61, 1/s
fects, especially in the directions perpendicular to the k* ORR rate constant in Figure 11, cm/s
kTk thermal conductivity of phase k, J/cm2‚K
sandwich, and multidimensional effects such as
kr relative hydraulic permeability
changing gas composition along the channel, among ksat saturated hydraulic permeability, cm2
others. For any model, a balance must be struck kΦ electrokinetic permeability, cm2
between the complexity required to describe the L catalyst layer thickness, cm
physical reality and the additional costs of such m parameter in the polarization equation (eq 20)
complexity. In other words, while more complex mPt loading of platinum, g/cm2
models more accurately describe the physics of the Mi molecular weight of species i, g/mol
transport processes, they are more computationally Mzi i symbol for the chemical formula of species i in phase
costly and may have so many unknown parameters k having charge zi
that their results are not as meaningful. Hopefully, n parameter in the polarization equation (eq 20)
this review has shown and broken down for the nh number of electrons transferred in electrode reac-
tion h
reader the vast complexities of transport within
Ni,k superficial flux density of species i in phase k, mol/
polymer-electrolyte fuel cells and the various ways cm2‚s
they have been and can be modeled. pi partial pressure of species i, kPa
pC capillary pressure, kPa
9. Acknowledgments pk total pressure of phase k, kPa
pvap
w
vapor pressure of water, kPa
The authors would like to thank the following qk superficial heat flux through phase k, J/cm2‚s
people for their input in helping write and proof this Q total amount of heat generated, J/cm2‚s
manuscript: Robert Darling, Jeremy Meyers, and Qk,p heat flux transferred between phases k and p,
Karen Thomas-Alyea. The authors would also like to J/cm3‚s
thank Bob Savinell and Tom Zawodzinski for the r pore radius, cm
opportunity to write this review. revap rate of evaporation, mol/cm3‚s
rl,k-p rate of reaction l per unit of interfacial area between
phases k and p, mol/cm2‚s
10. Nomenclature R ideal-gas constant, 8.3143 J/mol‚K
activity of species i in phase R Ragg agglomerate radius, cm
aRi Rg,k rate of homogeneous reaction g in phase k, mol/
ak,p interfacial surface area between phases k and p per cm3‚s
unit volume, 1/cm Ri,j resistance of resistor i,j in Figure 10 where ct stands
a°1,2 interfacial area between the electronically conduct- for charge-transfer, Ω‚cm2
ing and membrane phases with no flooding, 1/cm R′ total ohmic resistance, Ω‚cm2
Aagg specific external surface area of the agglomerate, Ri,k total rate of reaction of species i in phase k, mol/
1/cm cm3‚s
APt reactive surface area of platinum, cm2/g si,k,l stoichiometric coefficient of species i in phase k
b Tafel slope, defined by eq 14 for the ORR, V participating in reaction l
ci,k interstitial concentration of species i in phase k, S liquid saturation
mol/cm3 molar entropy of species i in phase k, J/mol‚K
cT total solution concentration or molar density, mol/ Si,k
cm3 ∆Sh entropy of reaction h, J/mol‚K
Ĉpk heat capacity of phase k, J/g‚K t time, s
di driving force per unit volume acting on species i in T absolute temperature, K
phase k, J/cm4 Tk absolute temperature of phase k, K
Di Fickian diffusion coefficient of species i in a mixture, ui mobility of species i, cm2‚mol/J‚s
cm2/s Uh reversible cell potential of reaction h, V
DS capillary diffusivity, cm2/s U′ potential intercept for a polarization equation, V
Di,j diffusion coefficient of i in j, cm2/s Uθh standard potential of reaction h, for oxygen reduc-
DKi Knudsen diffusion coefficient of species i, cm2/s tion, 1.229 V at 25 °C
E effectiveness factor UHh enthalpy potential, V
F Faraday’s constant, 96 487 C/equiv vk superficial velocity of phase k, cm/s
g acceleration due to gravity, cm/s2 V cell potential, V
hk,p heat-transfer coefficient between phases k and p, Vi (partial) molar volume of species i, cm3/mol
J/cm2‚s‚K WOdiff molar flow rate of oxygen to the agglomerate, mol/
cm3‚s
2
Hi,k partial molar enthalpy of species i in phase k, J/mol
Hi,j Henry’s constant for species i in component j, mol/ x distance across the flow field, cm
cm3‚kPa xi,k mole fraction of species i in phase k
y distance along the flow-field channel, cm (4) Cutlip, M. B.; Yang, S. C.; Stonehart, P. Electrochim. Acta 1991,
z distance across the cell sandwich, cm 36, 547.
(5) Iczkowski, R. P.; Cutlip, M. B. J. Electrochem. Soc. 1980, 127,
zi valence or charge number of species i 1433.
(6) Newman, J.; Tiedemann, W. AIChE J. 1975, 21, 25.
Greek Letters (7) Newman, J. S.; Tobias, C. W. J. Electrochem. Soc. 1962, 109,
1183.
Ra anodic transfer coefficient (8) Paganin, V. A.; Ticianelli, E. A.; Gonzalez, E. R. J. Appl.
Rc cathodic transfer coefficient Electrochem. 1996, 26, 297.
Rw water transport coefficient, mol2/J‚cm‚s (9) Parthasarathy, A.; Srinivasan, S.; Appleby, A. J.; Martin, C. R.
β net water flux per proton flux through the mem- J. Electrochem. Soc. 1992, 139, 2530.
brane (10) Parthasarathy, A.; Srinivasan, S.; Appleby, A. J.; Martin, C. R.
J. Electrochem. Soc. 1992, 139, 2856.
γ surface tension, N/cm (11) Ticianelli, E. A.; Derouin, C. R.; Redondo, A.; Srinivasan, S. J.
δn diffusion length or thickness of region n, cm Electrochem. Soc. 1988, 135, 2209.
ζ characteristic length, cm (12) Kim, J.; Lee, S. M.; Srinivasan, S.; Chamberlin, C. E. J.
k volume fraction of phase k Electrochem. Soc. 1995, 142, 2670.
(13) Bernardi, D. M.; Verbrugge, M. W. J. Electrochem. Soc. 1992,
0 bulk porosity 139, 2477.
νk kinematic viscosity of phase k, cm2/s (14) Springer, T. E.; Zawodzinski, T. A.; Gottesfeld, S. J. Electrochem.
ξ electroosmotic coefficient Soc. 1991, 138, 2334.
Πh Peltier coefficient for charge-transfer reaction h, V (15) Fuller, T. F.; Newman, J. J. Electrochem. Soc. 1993, 140, 1218.
(16) Nguyen, T. V.; White, R. E. J. Electrochem. Soc. 1993, 140, 2178.
Fk density of phase k, g/cm3 (17) Wang, J.-T.; Savinell, R. F. Electrochim. Acta 1992, 37, 2737.
σ0 standard conductivity in the electronically conduct- (18) Chan, D. S.; Wan, C. C. J. Power Sources 1994, 50, 261.
ing phase, S/cm (19) Rho, Y. W.; Srinivasan, S.; Kho, Y. T. J. Electrochem. Soc. 1994,
ηh,k-p electrode overpotential of reaction h between phases 141, 2089.
(20) Amphlett, J. C.; Baumert, R. M.; Mann, R. F.; Peppley, B. A.;
k and p (see eq 10), V Roberge, P. R.; Harris, T. J. J. Electrochem. Soc. 1995, 142, 1.
ηsh,k-p surface overpotential of reaction h between phases (21) Amphlett, J. C.; Baumert, R. M.; Mann, R. F.; Peppley, B. A.;
k and p (see eq 11), V Roberge, P. R.; Harris, T. J. J. Electrochem. Soc. 1995, 142, 9.
η* dimensionless overpotential in Figure 11 (22) Ticianelli, E. A. J. Electronanal. Chem. 1995, 387, 1.
(23) Mosdale, R.; Srinivasan, S. Electrochim. Acta 1995, 40, 413.
θ contact angle, deg (24) West, A. C.; Fuller, T. F. J. Appl. Electrochem. 1996, 26, 557.
κ conductivity of the ionically conducting phase, S/cm (25) Gurau, V.; Liu, H.; Kakac, S. AIChE J. 1998, 44, 2414.
λ moles of water per mole of sulfonic acid sites (26) Bevers, D.; Wohr, M.; Yasuda, K.; Oguro, K. J. Appl. Electro-
λL relative mobility of the liquid phase chem. 1997, 27, 1254.
µ viscosity, Pa‚s (27) Broka, K.; Ekdunge, P. J. Appl. Electrochem. 1997, 27, 281.
(28) Eikerling, M.; Kornyshev, A. A. J. Electronanal. Chem. 1998,
µi (electro)chemical potential of species i, J/mol 453, 89.
µRi electrochemical potential of species i in phase R, (29) de Sena, D. R.; Ticianelli, E. A.; Gonzalez, E. R. Electrochim.
J/mol Acta 1998, 43, 3755.
τ stress tensor, kPa (30) Antoine, O.; Bultel, Y.; Durand, R.; Ozil, P. Electrochim. Acta
1998, 43, 3681.
τk tortuosity of phase k (31) Bultel, Y.; Ozil, P.; Durand, R. J. Appl. Electrochem. 1998, 28,
φ Thiele modulus, defined by eq 60 for the ORR 269.
Φk potential in phase k, V (32) Bultel, Y.; Ozil, P.; Durand, R. Electrochim. Acta 1998, 43, 1077.
ψ dimensionless parameter in Figure 11 (33) Bultel, Y.; Ozil, P.; Durand, R. J. Appl. Electrochem. 1999, 29,
1025.
(34) Gloaguen, F.; Durand, R. J. Appl. Electrochem. 1997, 27, 1029.
Subscripts/Superscripts (35) Gloaguen, F.; Convert, P.; Gamburzev, S.; Velev, O. A.; Srini-
vasan, S. Electrochim. Acta 1998, 43, 3767.
(36) Perry, M. L.; Newman, J.; Cairns, E. J. J. Electrochem. Soc. 1998,
1 electronically conducting phase 145, 5.
2 ionically conducting phase (37) Kulikovsky, A. A.; Divisek, J.; Kornyshev, A. A. J. Electrochem.
agg agglomerate Soc. 1999, 146, 3981.
CL catalyst layer (38) Paddison, S. J.; Zawodzinski, T. A. Solid State Ionics 1998, 115,
333.
eff effective value, corrected for tortuosity and porosity (39) Okada, T.; Xie, G.; Meeg, M. Electrochim. Acta 1998, 43, 2141.
ext external to the control volume (40) Eikerling, M.; Kharkats, Y. I.; Kornyshev, A. A.; Volfkovich, Y.
f fixed ionic site in the membrane M. J. Electrochem. Soc. 1998, 145, 2684.
film film covering the agglomerate (41) Futerko, P.; Hsing, I. M. J. Electrochem. Soc. 1999, 146, 2049.
(42) Okada, T. J. Electroanal. Chem. 1999, 465, 1.
g homogeneous reaction number (43) Okada, T. J. Electroanal. Chem. 1999, 465, 18.
G gas phase (44) van Bussel, H.; Koene, F. G. H.; Mallant, R. J. Power Sources
h electron-transfer reaction number 1998, 71, 218.
HOR hydrogen-oxidation reaction (45) Wohr, M.; Bolwin, K.; Schnurnberger, W.; Fischer, M.; Neubrand,
W.; Eigenberger, G. Int. J. Hydrogen Energy 1998, 23, 213.
i generic species (46) Kazim, A.; Liu, H. T.; Forges, P. J. Appl. Electrochem. 1999, 29,
j generic species 1409.
k generic phase (47) Singh, D.; Lu, D. M.; Djilali, N. Int. J. Eng. Sci. 1999, 37, 431.
l heterogeneous reaction number (48) Yi, J. S.; Nguyen, T. V. J. Electrochem. Soc. 1999, 146, 38.
(49) Boyer, C. C.; Anthony, R. G.; Appleby, A. J. J. Appl. Electrochem.
L liquid phase 2000, 30, 777.
m mixture (50) Baschuk, J. J.; Li, X. J. Power Sources 2000, 86, 181.
ORR oxygen-reduction reaction (51) Dutta, S.; Shimpalee, S.; Van Zee, J. W. J. Appl. Electrochem.
p generic phase 2000, 30, 135.
ref parameter evaluated at the reference conditions (52) He, W.; Yi, J. S.; Nguyen, T. V. AIChE J. 2000, 46, 2053.
(53) Chan, S. H.; Tun, W. A. Chem. Eng. Technol. 2001, 24, 51.
s solid phase (54) Dutta, S.; Shimpalee, S.; Van Zee, J. W. Int. J. Heat Mass
w water Transfer 2001, 44, 2029.
(55) Janssen, G. J. M. J. Electrochem. Soc. 2001, 148, A1313.
11. References (56) Natarajan, D.; Nguyen, T. V. J. Electrochem. Soc. 2001, 148,
A1324.
(1) Giner, J.; Hunter, C. J. Electrochem. Soc. 1969, 116, 1124. (57) Wang, Z. H.; Wang, C. Y.; Chen, K. S. J. Power Sources 2001,
(2) Cutlip, M. B. Electrochim. Acta 1975, 20, 767. 94, 40.
(3) Cutlip, M. B.; Yang, S. C.; Stonehart, P. J. Electrochem. Soc. (58) Wang, C. Y.; Cheng, P. Int. J. Heat Mass Transfer 1996, 39, 3607.
1986, 133, C299. (59) Wang, C. Y.; Cheng, P. Adv. Heat Transfer 1997, 30, 93.
(60) Shimpalee, S.; Dutta, S. Numer. Heat Transfer, Part A 2000, (114) Marr, C.; Li, X. G. J. Power Sources 1999, 77, 17.
38, 111. (115) Baschuk, J. J.; Li, X. G. Int. J. Energy Res. 2003, 27, 1095.
(61) Hsing, I. M.; Futerko, P. Chem. Eng. Sci. 2000, 55, 4209. (116) Baschuk, J. J.; Rowe, A. M.; Li, X. G. J. Energy Resour. Technol.
(62) Futerko, P.; Hsing, I. M. Electrochim. Acta 2000, 45, 1741. 2003, 125, 94.
(63) Thampan, T.; Malhotra, S.; Tang, H.; Datta, R. J. Electrochem. (117) Murgia, G.; Pisani, L.; Shukla, A. K.; Scott, K. J. Electrochem.
Soc. 2000, 147, 3242. Soc. 2003, 150, A1231.
(64) Um, S.; Wang, C.-Y.; Chen, K. S. J. Electrochem. Soc. 2000, 147, (118) Pisani, L.; Murgia, G.; Valentini, M.; D’Aguanno, B. J. Power
4485. Sources 2002, 108, 192.
(65) Costamagna, P. Chem. Eng. Sci. 2001, 56, 323. (119) Zhou, T.; Liu, H. Int. J. Trans. Phenom. 2001, 3, 177.
(66) Dohle, H.; Kornyshev, A. A.; Kulikovsky, A. A.; Mergel, J.; (120) Gurau, V.; Barbir, F.; Liu, H. J. Electrochem. Soc. 2000, 147,
Stolten, D. Electrochem. Commun. 2001, 3, 73. 2468.
(67) Kornyshev, A. A.; Kulikovsky, A. A. Electrochim. Acta 2001, 46, (121) You, L. X.; Liu, H. T. Int. J. Hydrogen Energy 2001, 26, 991.
4389. (122) Pasaogullari, U.; Wang, C. Y. 204th Meeting of the Electro-
(68) Kulikovsky, A. A. Electrochem. Commun. 2001, 3, 572. chemical Society, Orlando, FL, 2003; Abstract 1103.
(69) Kulikovsky, A. A. Fuel Cells 2001, 1, 162. (123) Pasaogullari, U.; Wang, C. Y. J. Electrochem. Soc. 2004, 151,
(70) Eikerling, M.; Kornyshev, A. A.; Kuznetsov, A. M.; Ulstrup, J.; A399.
Walbran, S. J. Phys. Chem. B 2001, 105, 3646. (124) Meng, H.; Wang, C. Y. J. Electrochem. Soc. 2004, 151, A358.
(71) Rowe, A.; Li, X. G. J. Power Sources 2001, 102, 82. (125) Berning, T.; Lu, D. M.; Djilali, N. J. Power Sources 2002, 106,
(72) Thampan, T.; Malhotra, S.; Zhang, J. X.; Datta, R. Catal. Today 284.
2001, 67, 15. (126) Bradean, R.; Promislow, K.; Wetton, B. Numer. Heat Transfer,
(73) Shimpalee, S.; Lee, W. K.; Van Zee, J. W. The Electrochemical Part A 2002, 42, 121.
Society Extended Abstracts, September 2-7, 2001; Abstract 458. (127) Newman, J. Electrochim. Acta 1979, 24, 223.
(74) Weber, A. Z.; Darling, R. M.; Newman, J. J. Electrochem. Soc., (128) Fuller, T. F. Ph.D. Dissertation, University of Californias
in press. Berkeley, 1992.
(75) Chu, H.-S.; Yeh, C.; Chen, F. J. Power Sources 2002, 123, 1. (129) Dannenberg, K.; Ekdunge, P.; Lindbergh, G. J. Appl. Electro-
(76) Um, S.; Wang, C. Y. J. Power Sources 2004, 125, 40. chem. 2000, 30, 1377.
(77) Zhukovsky, K. V. AIChE J. 2003, 49, 3029. (130) Meyers, J. P.; Newman, J. J. Electrochem. Soc. 2002, 149, A710.
(78) Pisani, L.; Murgia, G.; Valentini, M.; D’Aguanno, B. J. Electro- (131) Meyers, J. P.; Newman, J. J. Electrochem. Soc. 2002, 149, A718.
chem. Soc. 2002, 149, A898. (132) Meyers, J. P.; Newman, J. J. Electrochem. Soc. 2002, 149, A729.
(79) You, L. X.; Liu, H. T. Int. J. Heat Mass Transfer 2002, 45, 2277. (133) Weber, A.; Darling, R.; Meyers, J.; Newman, J. In Handbook of
(80) Berning, T.; Djilali, N. J. Electrochem. Soc. 2003, 150, A1598. Fuel Cells: Fundamentals, Technology, and Applications; Viel-
(81) Kulikovsky, A. A. J. Electrochem. Soc. 2003, 150, A1432. stich, W., Lamm, A., Gasteiger, H. A., Eds.; John Wiley & Sons:
(82) Lee, W. K.; Shimpalee, S.; Van Zee, J. W. J. Electrochem. Soc. New York, 2003; Vol. 1.
2003, 150, A341. (134) Weber, A. Z.; Newman, J. J. Electrochem. Soc. 2004, 151, A326.
(83) Kulikovsky, A. A. Electrochem. Commun. 2002, 4, 527. (135) Kulikovsky, A. A. Electrochem. Commun. 2003, 5, 530.
(84) Mazumder, S.; Cole, J. V. J. Electrochem. Soc. 2003, 150, A1503. (136) Eikerling, M.; Paddison, S. J.; Zawodzinski, T. A. J. New Mater.
(85) Mazumder, S.; Cole, J. V. J. Electrochem. Soc. 2003, 150, A1510. Electrochem. Syst. 2002, 5, 15.
(86) Nam, J. H.; Kaviany, M. Int. J. Heat Mass Transfer 2003, 46, (137) Eikerling, M.; Kornyshev, A. A. J. Electronanal. Chem. 1999,
4595. 475, 107.
(87) Natarajan, D.; Nguyen, T. V. J. Power Sources 2003, 115, 66. (138) Eikerling, M.; Kornyshev, A. A.; Stimming, U. J. Phys. Chem.
(88) Weber, A. Z.; Newman, J. In Advanced Materials for Fuel Cells B 1997, 101, 10807.
and Batteries; Ehrlich, G., Arora, P., Ofer, D., Savinell, R. F., (139) Newman, J. S. Electrochemical Systems, 2nd ed.; Prentice Hall:
Zawodinski, T., D’Souza, F., Eds.; The Electrochemical Society Englewood Cliffs, NJ, 1991.
Proceeding Series: Pennington, NJ, 2003. (140) Bard, A. J.; Faulkner, L. R. Electrochemical Methods: Funda-
(89) Weber, A. Z.; Newman, J. J. Electrochem. Soc. 2003, 150, A1008. mentals and Applications, 2nd ed.; John Wiley & Sons: New
(90) Siegel, N. P.; Ellis, M. W.; Nelson, D. J.; von Spakovsky, M. R. York, 2001.
J. Power Sources 2003, 115, 81. (141) Bockris, J. O’M.; Reddy, A. K. N. Modern Electrochemistry;
(91) Pisani, L.; Valentini, M.; Murgia, G. J. Electrochem. Soc. 2003, Plenum Press: New York, 1973.
150, A1558. (142) Perry, R. H.; Green, D. W. Perry’s Chemical Engineers’ Hand-
(92) Choi, P.; Datta, R. J. Electrochem. Soc. 2003, 150, E601. book, 7th ed.; McGraw-Hill: New York, 1997.
(93) Weber, A. Z.; Newman, J. In Proton Conducting Membrane Fuel (143) Latimer, W. Oxidation Potentials, 2nd ed.; Prentice-Hall: Engle-
Cells III; Van Zee, J. W., Fuller, T. F., Gottesfeld, S., Murthy, wood Cliffs, NJ, 1952.
M., Eds.; The Electrochemical Society Proceeding Series: Pen- (144) Lewis, G.; Randall, M. International Critical Tables; McGraw-
nington, NJ, 2002. Hill: New York, 1930.
(94) Weber, A. Z.; Newman, J. J. Electrochem. Soc. 2004, 151, A311. (145) Vetter, K. J. Electrochemical Kinetics; Academic Press: New
(95) Murgia, G.; Pisani, L.; Valentini, M.; D’Aguanno, B. J. Electro- York, 1967.
chem. Soc. 2002, 149, A31. (146) Wang, J. X.; Brankovic, S. R.; Zhu, Y.; Hanson, J. C.; Adzic, R.
(96) Kulikovsky, A. A. Electrochem. Commun. 2002, 4, 318. R. J. Electrochem. Soc. 2003, 150, A1108.
(97) Kulikovsky, A. A. Electrochem. Commun. 2002, 4, 845. (147) Maruyama, J.; Inaba, M.; Katakura, K.; Ogumi, Z.; Takehara,
(98) Jaouen, F.; Lindbergh, G.; Sundholm, G. J. Electrochem. Soc. Z. J. Electronanal. Chem. 1998, 447, 201.
2002, 149, A437. (148) Mello, R. M. Q.; Ticianelli, E. A. Electrochim. Acta 1997, 42, 1031.
(99) Hertwig, K.; Martens, L.; Karwoth, R. Fuel Cells 2002, 2, 61. (149) Bird, R. B.; Stewart, W. E.; Lightfoot, E. N. Transport Phenom-
(100) Ahmed, S.; Kopasz, J.; Kumar, R.; Krumpelt, M. J. Power ena, 2nd ed.; John Wiley & Sons: New York, 2002.
Sources 2002, 112, 519. (150) Appleby, A. J. J. Electrochem. Soc. 1970, 117, 328.
(101) Springer, T. E.; Wilson, M. S.; Gottesfeld, S. J. Electrochem. Soc. (151) Kinoshita, K. Electrochemical Oxygen Technology; John Wiley
1993, 140, 3513. & Sons: New York, 1992.
(102) Springer, T. E.; Zawodzinski, T. A.; Wilson, M. S.; Gottesfeld, (152) Beattie, P. D.; Basura, V. I.; Holdcroft, S. J. Electronanal. Chem.
S. J. Electrochem. Soc. 1996, 143, 587. 1999, 468, 180.
(103) Springer, T. E.; Raistrick, I. D. J. Electrochem. Soc. 1989, 136, (153) Rho, Y. W.; Velev, O. A.; Srinivasan, S. J. Electrochem. Soc. 1994,
1594. 141, 2084.
(104) Springer, T. E.; Rockward, T.; Zawodzinski, T. A.; Gottesfeld, (154) Parthasarathy, A.; Dave, B.; Srinivasan, S.; Appleby, A. J.;
S. J. Electrochem. Soc. 2001, 148, A11. Martin, C. R. J. Electrochem. Soc. 1992, 139, 1634.
(105) Ridge, S. J.; White, R. E.; Tsou, Y.; Beaver, R. N.; Eisman, G. (155) Parthasarathy, A.; Srinivasan, S.; Appleby, A. J.; Martin, C. R.
A. J. Electrochem. Soc. 1989, 136, 1902. J. Electronanal. Chem. 1992, 339, 101.
(106) Weisbrod, K. R.; Grot, S. A.; Vanderborgh, N. E. In First (156) Perez, J.; Gonzalez, E. R.; Ticianelli, E. A. J. Electrochem. Soc.
International Symposium on Proton Conducting Membrane Fuel 1998, 145, 2307.
Cells; Landgrebe, A. R., Gottesfeld, S., Halpert, G., Eds.; The (157) Uribe, F. A.; Springer, T. E.; Gottesfeld, S. J. Electrochem. Soc.
Electrochemical Society Proceeding Series; Pennington, NJ, 1992, 139, 765.
1995; Vol. PV 95-23. (158) Gasteiger, H. A.; Gu, W.; Makharia, R.; Mathias, M. F.; Sompalli,
(107) Yi, J. S.; Nguyen, T. V. J. Electrochem. Soc. 1998, 145, 1149. B. In Handbook of Fuel Cells: Fundamentals, Technology, and
(108) Wood, D. L.; Yi, Y. S.; Nguyen, T. V. Electrochim. Acta 1998, Applications; Vielstich, W., Lamm, A., Gasteiger, H. A., Eds.;
43, 3795. John Wiley & Sons: New York, 2003; Vol. 3.
(109) Ge, S.-H.; Yi, B.-L. J. Power Sources 2003, 124, 1. (159) Sena, D. R.; Ticianelli, E. A.; Paganin, V. A.; Gonzalez, E. R. J.
(110) Bernardi, D. M.; Verbrugge, M. W. AIChE J. 1991, 37, 1151. Electronanal. Chem. 1999, 477, 164.
(111) Verbrugge, M. W.; Hill, R. F. J. Electrochem. Soc. 1990, 137, (160) Lee, S. J.; Mukerjee, S.; McBreen, J.; Rho, Y. W.; Kho, Y. T.;
886. Lee, T. H. Electrochim. Acta 1998, 43, 3693.
(112) Bernardi, D. M. J. Electrochem. Soc. 1990, 137, 3344. (161) Liebhafsky, H. A.; Cairns, E. J.; Grubb, W. T.; Niedrach, L. W.
(113) Chan, S. H.; Goh, S. K.; Jiang, S. P. Electrochim. Acta 2003, 48, In Fuel Cell Systems; Young, G. J., Linden, H. R., Eds.; American
1905. Chemical Society: Washington, DC, 1965; Vol. 47.
(162) Passalacqua, E.; Lufrano, F.; Squadrito, G.; Patti, A.; Giorgi, L. (217) Buchi, F. N.; Scherer, G. G. J. Electrochem. Soc. 2001, 148, A183.
Electrochim. Acta 2001, 46, 799. (218) Kreuer, K. D. Solid State Ionics 1997, 97, 1.
(163) Squadrito, G.; Maggio, G.; Passalacqua, E.; Lufrano, F.; Patti, (219) De La Rue, R. E.; Tobias, C. W. J. Electrochem. Soc. 1959, 106,
A. J. Appl. Electrochem. 1999, 29, 1449. 827.
(164) Newman, J. Ind. Eng. Chem. Res. 1995, 34, 3208. (220) Bruggeman, D. A. G. Ann. Phys. 1935, 24, 636.
(165) Kreuer, K. D. Chem. Rev. 2004, 104, 4637. (221) Li, G.; Pickup, P. P. J. Electrochem. Soc. 2003, 150, C745.
(166) Paddison, S. J.; Paul, R.; Zawodzinski, T. A. J. Electrochem. Soc. (222) Bernardi, D. M. Ph.D. Dissertation, University of California,
2000, 147, 617. 1986.
(167) Khalatur, P. G.; Talitskikh, S. K.; Khokhlov, A. R. Macromol. (223) Dullien, F. A. L. Porous Media: Fluid Transport and Pore
Theory Simul. 2002, 11, 566. Structure, 2nd ed.; Academic Press: New York, 1992.
(168) Krueger, J. J.; Simon, P. P.; Ploehn, H. J. Macromolecules 2002, (224) Knudsen, M. The Kinetic Theory of Gases; Methuen: London,
35, 5630. 1934.
(169) Li, T.; Wlaschin, A.; Balbuena, P. B. Ind. Eng. Chem. Res. 2001, (225) Rothfeld, L. B. AIChE J. 1963, 9, 19.
40, 4789. (226) Weber, A. Z.; Newman, J. Int. Commun. Heat Mass Transfer,
(170) Paddison, S. J.; Paul, R.; Zawodzinski, T. A. J. Chem. Phys. 2001, in press.
115, 7753. (227) Passalacqua, E.; Squadrito, G.; Lufrano, F.; Patti, A.; Giorgi, L.
(171) Paul, R.; Paddison, S. J. J. Chem. Phys. 2001, 115, 7762. J. Appl. Electrochem. 2001, 31, 449.
(172) Paddison, S. J. J. New Mater. Electrochem. Syst. 2001, 4, 197. (228) Tucker, M. C.; Odgaard, M.; Yde-Anderson, S.; Thomas, J. O.
(173) Vishnyakov, A.; Neimark, A. V. J. Phys. Chem. B 2001, 105, 203rd Meeting of the Electrochemical Society, Paris, 2003;
7830. Abstract 1235.
(174) Mologin, D. A.; Khalatur, P. G.; Kholhlov, A. R. Macromol. (229) Jordan, L. R.; Shukla, A. K.; Behrsing, T.; Avery, N. R.; Muddle,
Theory Simul. 2002, 11, 587. B. C.; Forsyth, M. J. Power Sources 2000, 86, 250.
(175) Spohr, E.; Commer, P.; Kornyshev, A. A. J. Phys. Chem. B 2002, (230) Jordan, L. R.; Shukla, A. K.; Behrsing, T.; Avery, N. R.; Muddle,
106, 10560. B. C.; Forsyth, M. J. Appl. Electrochem. 2000, 30, 641.
(176) Vishnyakov, A.; Neimark, A. V. J. Phys. Chem. B 2000, 104, (231) Kong, C. S.; Kim, D.-Y.; Lee, H.-K.; Shul, Y.-G.; Lee, T.-H. J.
4471. Power Sources 2002, 108, 185.
(177) Vishnyakov, A.; Neimark, A. V. J. Phys. Chem. B 2001, 105, (232) Wilson, M. S. U.S. Patent 5,641,586, 1995.
9586. (233) Miller, C. T.; Christakos, G.; Imhoff, P. T.; McBride, J. F.; Pedit,
(178) Din, X. D.; Michaelides, E. E. AIChE J. 1998, 44, 35. J. A.; Trangenstein, J. A. Adv. Water Resources 1998, 21, 77.
(179) Commer, P.; Cherstvy, A. G.; Spohr, E.; Kornyshev, A. A. Fuel (234) Bear, J. Dynamics of Fluids in Porous Media; Dover Publica-
Cells 2002, 2, 127. tions: New York, 1988.
(180) Jinnouchi, R.; Okazaki, K. Microscale Thermophys. Eng. 2003, (235) Smith, W. O. Physics 1933, 4, 425.
7, 15. (236) Leverett, M. C. Trans. Am. Inst. Min., Metall. Pet. Eng. 1941,
(181) Yeo, S. C.; Eisenberg, J. J. Appl. Polym. Sci. 1977, 21, 875. 142, 152.
(182) Hsu, W. Y.; Gierke, T. D. J. Membr. Sci. 1983, 13, 307. (237) Fischer, A.; Jindra, J.; Wendt, H. J. Appl. Electrochem. 1998,
(183) Eisenberg, A. Macromolecules 1970, 3, 147. 28, 277.
(184) Yeager, H. L.; Steck, A. J. Electrochem. Soc. 1981, 128, 1880. (238) Ticianelli, E. A.; Beery, J. G.; Srinivasan, S. J. Appl. Electrochem.
(185) Eisenberg, A. Bull. Am. Phys. Soc. 1969, 14, 382. 1991, 21, 597.
(186) Dreyfus, B. J. Polym. Sci., Part B: Polym. Phys. 1983, 21, 2337. (239) Cheng, X.; Yi, B.; Han, M.; Zhang, J.; Qiao, Y.; Yu, J. J. Power
(187) Koter, S. J. Membr. Sci. 2000, 166, 127. Sources 1999, 79, 75.
(188) Datye, V. K.; Taylor, P. L.; Hopfinger, A. J. Macromolecules 1984, (240) Uchida, M.; Aoyama, Y.; Eda, N.; Ohta, A. J. Electrochem. Soc.
17, 1704. 1995, 142, 4143.
(189) Mauritz, K. A.; Rogers, C. E. Macromolecules 1985, 18, 483. (241) Handbook of Fuel Cells: Fundamentals, Technology, and Ap-
(190) Pintauro, P. N.; Verbrugge, M. W. J. Membr. Sci. 1989, 44, 197. plications, Vol. 2sElectrocatalysis; Vielstich, W., Lamm, A.,
(191) Uitto, O. D.; White, H. S.; Aoki, K. Anal. Chem. 2002, 74, 4577. Gasteiger, H. A., Eds.; John Wiley & Sons: New York, 2003.
(192) Guzmangarcia, A. G.; Pintauro, P. N.; Verbrugge, M. W.; Hill, (242) Mukerjee, S. J. Appl. Electrochem. 1990, 20, 537.
R. F. AIChE J. 1990, 36, 1061. (243) Markovic, N. M.; Schmidt, T. J.; Stamenkovic, V.; Ross, P. N.
(193) Tandon, R.; Pintauro, P. N. J. Membr. Sci. 1997, 136, 207. Fuel Cells 2001, 1, 105.
(194) Wendt, R. P.; Klein, E.; Lynch, S. J. Membr. Sci. 1976, 1, 165. (244) Vogel, W.; Lundquist, J.; Ross, P.; Stonehart, P. Electrochim.
(195) Cwirko, E. H.; Carbonell, R. G. J. Membr. Sci. 1992, 67, 227. Acta 1975, 20, 79.
(196) Bontha, J. R.; Pintauro, P. N. Chem. Eng. Sci. 1994, 49, 3835. (245) Zhang, J. X.; Thampan, T.; Datta, R. J. Electrochem. Soc. 2002,
(197) Koter, S. J. Membr. Sci. 2002, 206, 201. 149, A765.
(198) Narebska, A.; Koter, S.; Kujawski, W. Desalination 1984, 51, 3. (246) Bellows, R. J.; MarucchiSoos, E. P.; Buckley, D. T. Ind. Eng.
(199) Yang, Y. H.; Pintauro, P. N. AIChE J. 2000, 46, 1177. Chem. Res. 1996, 35, 1235.
(200) Pintauro, P. N.; Bennion, D. N. Ind. Eng. Chem. Fundam. 1984, (247) Stonehar, P. Electrochim. Acta 1967, 12, 1185.
23, 234. (248) Bhatia, K. K.; Wang, C.-Y. Electrochim. Acta 2004, 49, 2333.
(201) Capeci, S. W.; Pintauro, P. N.; Bennion, D. N. J. Electrochem. (249) Grens, E. A.; Turner, R. M.; Katan, T. Adv. Energy Convers.
Soc. 1989, 136, 2876. 1964, 4, 109.
(202) Schroeder, P. Z. Phys. Chem. 1903, 45, 57. (250) Chirkov, Y. G. Elektrokhimiya 1975, 11, 36.
(203) Mann, R. F.; Amphlett, J. C.; Hooper, M. A. I.; Jensen, H. M.; (251) Chirkov, Y. G. Elektrokhimiya 1972, 7, 1826.
Peppley, B. A.; Roberge, P. R. J. Power Sources 2000, 86, 173. (252) Viitanen, M.; Lampinen, M. J. J. Power Sources 1990, 32, 207.
(204) Leddy, J.; Iverson, A. E.; Vanderborgh, N. E. In Electrochemical (253) Grens, E. A. Ind. Eng. Chem. Fundam. 1966, 5, 542.
and Thermal Modeling of Battery, Fuel Cell, and Photoenergy (254) Yang, S. C.; Cutlip, M. B.; Stonehart, P. Electrochim. Acta 1990,
Conversion Systems; Selman, J. R., Maru, H. C., Eds.; The 35, 869.
Electrochemical Society Proceeding Series; Pennington, NJ, (255) Yang, S. C.; Cutlip, M. B.; Stonehart, P. Electrochim. Acta 1989,
1986; Vol. PV 86-12. 34, 703.
(205) Einstein, A. Ann. Phys. 1905, 17, 549. (256) Vogel, W.; Bradford, A.; Lundquist, J. Electrochim. Acta 1972,
(206) Nernst, W. Z. Phys. Chem. 1888, 2, 613. 17, 1735.
(207) Okada, T.; Xie, G.; Tanabe, Y. J. Electronanal. Chem. 1996, 413, (257) Bjornbom, P. Electrochim. Acta 1987, 32, 115.
49. (258) Srinivasan, S.; Hurwitz, H. D. Electrochim. Acta 1967, 12, 495.
(208) Bennion, D. N. Mass Transport of Binary Electrolyte Solutions (259) Srinivasan, S.; Hurwitz, H. D.; Bockris, J. O’M. J. Chem. Phys.
in Membranes; Water Resources Center Desalination Report No. 1967, 46, 3108.
4; Department of Engineering, University of CaliforniasLos (260) Bultel, Y.; Ozil, P.; Durand, R. J. Appl. Electrochem. 2000, 30,
Angeles: 1966. 1369.
(209) Pintauro, P. N.; Bennion, D. N. Ind. Eng. Chem. Fundam. 1984, (261) Euler, J.; Nonnenmacher, W. Electrochim. Acta 1960, 2, 268.
23, 230. (262) Bockris, J. O’M.; Srinivasan, S. Fuel Cells: Their Electrochem-
(210) Auclair, B.; Nikonenko, V.; Larchet, C.; Metayer, M.; Dammak, istry; McGraw-Hill: New York, 1969.
L. J. Membr. Sci. 2002, 195, 89. (263) DeVidts, P.; White, R. E. J. Electrochem. Soc. 1997, 144, 1343.
(211) Koter, S.; Hamann, C. H. J. Non-Equilib. Thermodyn. 1990, 15, (264) Gloaguen, F.; Andolfatto, F.; Durand, R.; Ozil, P. J. Appl.
315. Electrochem. 1994, 24, 863.
(212) Meyers, J. P. Ph.D. Dissertation, University of Californias (265) Fogler, H. S. Elements of Chemical Reaction Engineering, 2nd
Berkeley, 1998. ed.; Prentice Hall: Upper Saddle River, NJ, 1992.
(213) Verbrugge, M. W.; Hill, R. F. J. Electrochem. Soc. 1990, 137, (266) Thiele, E. W. Ind. Eng. Chem. 1939, 31, 916.
3770. (267) Fuller, T. F.; Luczak, F. J.; Wheeler, D. J. J. Electrochem. Soc.
(214) Verbrugge, M. W.; Hill, R. F. J. Electrochem. Soc. 1990, 137, 1995, 142, 1752.
1131. (268) Stonehart, P.; Ross, P. N. Electrochim. Acta 1976, 21, 441.
(215) Schlogl, R. Z. Phys. Chem. 1955, 3, 73. (269) Maja, M.; Tosco, P.; Vanni, M. J. Electrochem. Soc. 2003, 148,
(216) Mason, E. A.; Malinauskas, A. P. Gas Transport in Porous A1368.
Media: The Dusty-Gas Model; Elsevier: Amsterdam, 1983. (270) Qi, Z. G.; Kaufman, A. J. Power Sources 2002, 109, 38.
(271) Wilson, M. S.; Valerio, J. A.; Gottesfeld, S. Electrochim. Acta (304) Andreaus, B.; McEvoy, A. J.; Scherer, G. G. Electrochim. Acta
1995, 40, 355. 2002, 47, 2223.
(272) Janssen, G. J. M.; Overvelde, M. L. J. J. Power Sources 2001, (305) Wagner, N. J. Appl. Electrochem. 2002, 32, 859.
101, 117. (306) Wang, X.; Hsing, I. M.; Leng, Y. J.; Yue, P. L. Electrochim. Acta
(273) Patankar, S. Numerical Heat Transfer and Fluid Flow; Hemi- 2001, 46, 4397.
sphere Publishing Corporation: Bristol, PA, 1980. (307) Ciureanu, M.; Roberge, R. J. Phys. Chem. B 2001, 105, 3531.
(274) Suares, G. E.; Kosanovich, K. A. Ind. Eng. Chem. Res. 1997, 36, (308) Sorensen, T. S.; Kjelstrup, S. J. Colloid Interface Sci. 2002, 248,
4264. 355.
(275) Duan, T.; Weidner, J. W.; White, R. E. J. Power Sources 2002, (309) Bultel, Y.; Genies, L.; Antoine, O.; Ozil, P.; Durand, R. J.
107, 24. Electronanal. Chem. 2002, 527, 143.
(276) Wang, C. Y. Chem. Rev. 2004, 104, 4727. (310) Yuh, C. Y.; Selman, J. R. AIChE J. 1988, 34, 1949.
(277) McCabe, W. L.; Smith, J. C.; Harriott, P. Unit Operations of (311) Raistrick, I. D. Electrochim. Acta 1990, 35, 1579.
Chemical Engineering, 5th ed.; McGraw-Hill, Inc.: New York, (312) Fritts, S. D.; Savinell, R. F. J. Power Sources 1989, 28, 301.
1993. (313) Guo, Q.; Cayetano, M.; Tsou, Y.-M.; De Castro, E. S.; White, R.
(278) Meyers, J. P.; Villwock, R. D.; Darling, R. M.; Newman, J. In E. J. Electrochem. Soc. 2003, 150, A1440.
Advances in Mathematical Modeling and Simulation of Elec- (314) Jaouen, F.; Lindbergh, G. J. Electrochem. Soc. 2003, 150, A1699.
trochemical Processes and Oxygen Depolarized Cathodes and (315) Scott, K.; Argyropoulos, P.; Sundmacher, K. J. Electronanal.
Activated Cathodes for Chlor-Alkali Processes; Van Zee, J. W., Chem. 1999, 477, 97.
Fuller, T. F., Foller, P. C., Hine, F., Eds.; The Electrcohemical (316) Scott, K.; Taama, W. M.; Argyropoulos, P.; Sundmacher, K. J.
Society Proceedings Series; Pennington, NJ, 1998; Vol. PV 98- Power Sources 1999, 83, 204.
10. (317) Scott, K.; Taama, W. M.; Kramer, S.; Argyropoulos, P.; Sund-
(279) Kumar, A.; Reddy, R. G. J. Power Sources 2003, 114, 54. macher, K. Electrochim. Acta 1999, 45, 945.
(280) Kumar, A.; Reddy, R. G. J. Power Sources 2003, 113, 11.
(318) Baxter, S. F.; Battaglia, V. S.; White, R. E. J. Electrochem. Soc.
(281) Baker, D. R.; Darling, R. M. Int. J. Trans. Phenom. 2001, 3, 177.
1999, 146, 437.
(282) Hirschfelder, J. O.; Curtiss, C. F.; Bird, R. B. Molecular Theory
(319) Fan, J. R.; Hu, G. L.; Yao, J.; Cen, K. F. Energy Fuels 2002, 16,
of Gases and Liquids; John Wiley & Sons: New York, 1954.
1591.
(283) Thomas, K. E.; Newman, J. J. Electrochem. Soc. 2003, 150, A176.
(284) Newman, J.; Thomas-Alyea, K. E. Electrochemical Systems, 3rd (320) Argyropoulos, P.; Scott, K.; Shukla, A. K.; Jackson, C. Fuel Cells
ed.; John Wiley & Sons: New York, 2004. 2002, 2, 78.
(285) Agar, J. N. In Advances in Electrochemistry and Electrochemical (321) Argyropoulos, P.; Scott, K.; Taama, W. M. Chem. Eng. J. 1999,
Engineering; Delahay, P., Ed.; John Wiley & Sons: New York, 73, 217.
1963; Vol. 3. (322) Argyropoulos, P.; Scott, K.; Taama, W. M. J. Power Sources 1999,
(286) Rao, L.; Newman, J. J. Electrochem. Soc. 1997, 144, 2697. 79, 169.
(287) Bernardi, D. M.; Pawlikowski, E.; Newman, J. J. Electrochem. (323) Argyropoulos, P.; Scott, K.; Taama, W. M. J. Appl. Electrochem.
Soc. 1985, 132, 5. 2000, 30, 899.
(288) Amphlett, J. C.; Mann, R. F.; Peppley, B. A.; Roberge, P. R.; (324) Birgersson, E.; Nordlund, J.; Ekstrom, H.; Vynnycky, M.; Lind-
Rodrigues, A. J. Power Sources 1996, 61, 183. bergh, G. J. Electrochem. Soc. 2003, 150, A1368.
(289) Yerramalla, S.; Davari, A.; Feliachi, A.; Biswas, T. J. Power (325) Andrian, S. V.; Meusinger, J. J. Power Sources 2000, 91, 193.
Sources 2003, 124, 104. (326) Okada, T. J. New Mater. Electrochem. Syst. 2001, 4, 209.
(290) Zawodzinski, T. A.; Derouin, C. R.; Radzinski, S.; Sherman, R.; (327) Li, P. W.; Schaefer, L.; Wang, Q. M.; Zhang, T.; Chyu, M. K. J.
Smith, V. T.; Springer, T. E.; Gottesfeld, S. J. Electrochem. Soc. Power Sources 2003, 115, 90.
1993, 140, 1041. (328) Li, P. W.; Zhang, T.; Wang, Q. M.; Schaefer, L.; Chyu, M. K. J.
(291) Nguyen, T. V.; Vanderborgh, N. E. J. Membr. Sci. 1998, 143, Power Sources 2003, 114, 63.
235. (329) Chu, D.; Jiang, R. Z. J. Power Sources 1999, 83, 128.
(292) He, W. S.; Lin, G. Y.; Van Nguyen, T. AIChE J. 2003, 49, 3221. (330) Darling, R. M.; Meyers, J. P. J. Electrochem. Soc. 2003, 150,
(293) Mosdale, R.; Gebel, G.; Pineri, M. J. Membr. Sci. 1996, 118, 269. A1523.
(294) De Francesco, M.; Arato, E. J. Power Sources 2002, 108, 41. (331) Wendt, H.; Brenscheidt, T.; Fischer, A. Philos. Trans. R. Soc.
(295) Scott, K.; Argyropoulos, P.; Taama, W. M. Chem. Eng. Res. Des. London, Ser. A 1996, 354, 1627.
2000, 78, 881. (332) Jiang, R. Z.; Chu, D. J. Power Sources 2001, 92, 193.
(296) Maggio, G.; Recupero, V.; Mantegazza, C. J. Power Sources 1996, (333) Fowler, M. W.; Mann, R. F.; Amphlett, J. C.; Peppley, B. A.;
62, 167. Roberge, P. R. J. Power Sources 2002, 106, 274.
(297) Thirumalai, D.; White, R. E. J. Electrochem. Soc. 1997, 144, (334) Aragane, J.; Murahashi, T.; Odaka, T. J. Electrochem. Soc. 1988,
1717. 135, 844.
(298) Lee, J. H.; Lalk, T. R.; Appleby, A. J. J. Power Sources 1998, (335) Honji, A.; Mori, T.; Tamura, K.; Hishinuma, Y. J. Electrochem.
70, 258. Soc. 1988, 135, 355.
(299) Lee, J. H.; Lalk, T. R. J. Power Sources 1998, 73, 229. (336) Liu, W.; Ruth, K.; Rusch, G. J. New Mater. Electrochem. Syst.
(300) Cownden, R.; Nahon, M.; Rosen, M. Int. J. Hydrogen Energy 2001, 4, 227.
2001, 26, 615. (337) Pozio, A.; Silva, R. F.; De Francesco, M.; Giorgi, L. Electrochim.
(301) Ahluwalia, R. K.; Doss, E. D.; Kumar, R. J. Power Sources 2003, Acta 2003, 48, 1543.
117, 45. (338) Kiwi, J.; Denisov, N.; Gak, Y.; Ovanesyan, N.; Buffat, P. A.;
(302) Paganin, V. A.; Oliveira, C. L. F.; Ticianelli, E. A.; Springer, T. Suvorova, E.; Gostev, F.; Titov, A.; Sarkison, O.; Albers, P.;
E.; Gonzalez, E. R. Electrochim. Acta 1998, 43, 3761. Nadtochenko, V. Langmuir 2002, 18, 9054.
(303) Brett, D. J. L.; Atkins, S.; Brandon, N. P.; Vesovic, V.; Vasileiadis,
N.; Kucernak, A. Electrochem. Solid-State Lett. 2003, 6, A63. CR020729L
Chem. Rev. 2004, 104, 4727−4766 4727
Fundamental Models for Fuel Cell Engineering

Chao-Yang Wang*
Departments of Mechanical Engineering and Materials Science and Engineering, Electrochemical Engine Center (ECEC),
The Pennsylvania State University, University Park, Pennsylvania 16802
Received January 23, 2004
Contents 7. Acknowledgments 4763

8. References 4763
2. Computational Fuel Cell Dynamics 4727
2.1. CFCD Model Equations 4728
1. Introduction
2.2. Computational Aspects 4729
2.2.1. General 4729 Owing to their high energy efficiency, low pollution,
2.2.2. Solution Algorithms for Coupled Potential 4729 and low noise, fuel cells are widely regarded as 21st
Equations century energy-conversion devices for mobile, sta-
2.2.3. Mesh Requirement and Parallel 4730 tionary, and portable power. Through tremendous
Computing progress made in the past decade, currently available
2.3. Material Property Characterization 4730 fuel cell materials appear to be adequate for near-
3. Polymer Electrolyte Fuel Cells 4732 term markets with highest cost entry points. As a
3.1. Single-Phase Conservation Equations 4733 result, industries are currently placing their focus on
fuel cell design and engineering for better perfor-
3.2. Water Transport 4736
mance, improved durability, cost reduction, and
3.3. Heat Transport 4738 better cold-start characteristics. This new focus has
3.4. Electron Transport 4740 led to an urgent need for identification, understand-
3.5. Transient Phenomena 4741 ing, prediction, control, and optimization of various
3.6. Large-Scale Simulation 4742 transport and electrochemical processes that occur
3.7. Liquid Water Transport 4742 on disparate length scales in fuel cells.
3.8. Experimental Diagnostics and Model 4747 The purpose of the present review is to summarize
Validation the current status of fundamental models for fuel cell
3.8.1. Current, Species, and High-Frequency 4747 engineering and indicate where this burgeoning field
Resistance Distribution Measurements is heading. By choice, this review is limited to
3.8.2. Temperature Distribution Measurements 4749 hydrogen/air polymer electrolyte fuel cells (PEFCs),
3.8.3. Two-Phase Visualization 4750 direct methanol fuel cells (DMFCs), and solid oxide
3.8.4. Experimental Validation 4751 fuel cells (SOFCs). Also, the review does not include
3.9. Modeling the Catalyst Layer at Pore Level 4751 microscopic, first-principle modeling of fuel cell ma-
3.10. Summary and Outlook 4752 terials, such as proton conducting membranes and
4. Direct Methanol Fuel Cells 4753 catalyst surfaces. For good overviews of the latter
fields, the reader can turn to Kreuer,1 Paddison,2 and
4.1. Technical Challenges 4754
Koper,3 for example.
4.1.1. Methanol Oxidation Kinetics 4754
The review is organized as follows. Section 2
4.1.2. Methanol Crossover 4755 defines a systematic framework for fuel cell modeling
4.1.3. Water Management 4755 research, called computational fuel cell dynamics
4.1.4. Heat Management 4756 (CFCD), and outlines its four essential elements.
4.2. DMFC Modeling 4756 Sections 3-5 review work performed in the past
4.2.1. Needs for Modeling 4756 decade on PEFCs, DMFCs, and SOFCs, respectively.
4.2.2. DMFC Models 4756 Future research needs and directions of the three
4.3. Experimental Diagnostics 4757 types of fuel cells are pointed out wherever applicable
4.4. Model Validation 4758 and summarized separately at the end of each
4.5. Summary and Outlook 4760 section.
5. Solid Oxide Fuel Cells 4760
5.1. SOFC Models 4761 2. Computational Fuel Cell Dynamics
5.2. Summary and Outlook 4762 Fuel cell science and technology cuts across mul-
6. Closing Remarks 4763 tiple disciplines, including materials science, inter-
facial science, transport phenomena, electrochemis-
* To whom correspondence should be addressed. E-mail: cxw31@ try, and catalysis. Because of the diversity and
psu.edu. complexity of electrochemical and transport phenom-
10.1021/cr020718s CCC: $48.50 © 2004 American Chemical Society
4728 Chemical Reviews, 2004, Vol. 104, No. 10 Wang
Chao-Yang Wang received his Ph.D. degree in Mechanical Engineering

from the University of Iowa in 1994, and he is currently a Professor of
Mechanical Engineering and Materials Science & Engineering at The
Pennsylvania State University. He has been the founding director of the
Penn State Electrochemical Engine Center (ECEC) since 1997. His
research interests include fundamental modeling and experimental Figure 1. Local current density profiles along a straight-
diagnostics of various fuel cell/battery systems and their materials. channel fuel cell as predicted by the same computer model
for two cases differing only in two model parameters.6
ena involved in a fuel cell and occurring at disparate

length and time scales, fuel cell modeling and simu- constant by 23%. By changing only two parameters
lation requires a systematic framework parallel to (i.e., catalyst layer resistance and kinetic constant
computational fluid dynamics (CFD), here termed of ORR), Case 2 produces exactly the same average
computational fuel cell dynamics (CFCD). CFCD current density as Case 1, despite that the two cases
deals with the coupling of multidimensional transport fundamentally differ in limiting mechanisms. Case
phenomena with electrochemical kinetics and the 1 is dominated more by ohmic polarization, thereby
transport of charge (electrons and ions) to provide a exhibiting a slight increase in the local current
comprehensive understanding of fuel cell dynamics. density near the dry inlet as the membrane becomes
CFCD is envisioned to consist of (1) physicochemical more hydrated, and a fall only close to the outlet
model development, (2) advanced numerical algo- where oxygen depletion starts to take control. On the
rithms, (3) materials characterization, and (4) model other hand, Case 2 is dominated more by concentra-
validation at detailed levels. tion polarization with little ohmic loss. As a result,
Section 2.1 gives a generalized summary of fuel cell the current distribution in Case 2 closely follows the
models, while section 2.2 discusses the need for decline of oxygen concentration due to the stoichio-
employing large numerical meshes and hence ad- metric effect. Literature efforts involving detailed
vanced numerical algorithms for efficient fuel cell diagnostics aiming to create benchmark-quality data
simulations. Section 2.3 briefly reviews the efforts, for CFCD model validation are reviewed in section
in the literature, to measure basic materials and 3.8 for the PEFC system.
transport properties as input to fuel cell models.
The need for the fourth element, i.e., model valida- 2.1. CFCD Model Equations
tion against detailed distribution data, has been A fundamental fuel cell model consists of five
increasingly acknowledged in the past few years.4-6 principles of conservation: mass, momentum, spe-
This is because the global I-V curve is largely cies, charge, and thermal energy. These transport
inadequate to validate comprehensive CFCD models. equations are then coupled with electrochemical
This inadequacy can perhaps best be illustrated by processes through source terms to describe reaction
a numerical experiment shown in Figure 1. In this kinetics and electro-osmotic drag in the polymer
experiment, a three-dimensional (3-D) PEFC model electrolyte. Such convection-diffusion-source equa-
(as described in section 3.1) is applied to a single- tions can be summarized in the following general
channel fuel cell using Gore-Select 18 µm membrane form
with an anode/cathode stoichiometry of 3/2 and
anode/cathode inlet relative humidity of 42%/0% at ∂(Fφ)
a cell temperature of 80 °C. Two cases are simulated + ∇(Fvφ) ) ∇(Γ∇φ) + S (1)
∂t
using the same model for a cell voltage of 0.75 V. In
Case 1, the ionic resistance in two catalyst layers is where φ is a general variable to be solved, F the
included, and the kinetic constant of oxygen reduc- density, v the velocity vector, Γ the diffusion coef-
tion reaction (ORR) is estimated from the Pt loading. ficient, and S a source term which includes all terms
This case yields an average current density of 0.24 that cannot be included in the previous terms. For
A/cm2 at 0.75 V. In Case 2 (which is hypothetical), instance, the Darcy’s drag exerted on gas flow
however, we neglect the ionic resistance in two through porous electrodes of fuel cells is typically
catalyst layers (amounting to approximately twice accounted for through a source term in the momen-
that of the membrane) and then reduce the kinetic tum equation. The electro-osmotic drag of water
Fundamental Models for Fuel Cell Engineering Chemical Reviews, 2004, Vol. 104, No. 10 4729
through a polymer electrolyte can also be formulated potential for an electrochemical reaction, η, is defined
as a source term in the water species equation since as
it differs from conventional mechanisms of transport
by diffusion and convection and arises only from an η ) Φs - Φe - Uo (6)
electrochemical system.
Setting φ ) 1, the above general transport equation where Φs and Φe are the electric potentials of the
is reduced to electronic phase and electrolyte, respectively, at the
∂F triple-phase interface. The last term in eq 6, Uo, is
+ ∇(Fv) ) Sm (2) the thermodynamic equilibrium potential of the reac-
∂t tion, which is, in turn, a function of the reactant and
This equation is usually referred to as the continuity product concentrations at the interface as generally
equation or mass conservation equation.7 The source expressed by the Nernst equation. Note that the
term, Sm, in the continuity equation is commonly surface overpotential, η, represents the driving force
caused by mass consumption or production from for an electrochemical reaction.
electrochemical reactions as well as mass loss/gain
through phase transformation. 2.2. Computational Aspects
Main source terms prevailing in most transport 2.2.1. General
equations for a fuel cell model are due to electro-
chemical reactions occurring in the electrode com- The advection-diffusion equation with a source
prised of three phases: electronic (s), electrolyte (e), term can be solved by CFD algorithms in general.
and gas (g). Electrochemical reactions occur at the Patankar9 provided an excellent introduction to nu-
triple-phase boundary according to the following merical fluid flow and heat transfer. Oran and Boris10
general formula discussed numerical solutions of diffusion-convection
problems with chemical reactions. Since fuel cells
∑skMkz ) ne- (3) feature an aspect ratio of the order of 100, O(100),
the upwind scheme for the flow-field solution is
k
applicable and proves to be very effective.9 Unstruc-
where the summation is over all species involved in tured meshes are commonly employed in commercial
a reaction. Mk is a chemical symbol for species k CFD codes.
participating in the electrochemical reaction, z and
s are the charge number and stoichiometric coef- 2.2.2. Solution Algorithms for Coupled Potential Equations
ficient of the species, and n is the number of electrons A numerical problem unique to the modeling of
transferred in the reaction. The values of sk, z, and n electrochemical systems in general and fuel cells in
can readily be determined by comparing a specific particular is the strong coupling between the poten-
electrode reaction to this general form. For example, tials for the electronic and electrolyte phases. The two
for the oxygen reduction reaction, 2H2O - O2 - 4H+ potential equations are dependent on each other
) 4e-, one has that sH2O ) 2, sO2 ) -1, sH+ ) -4, and through the reaction current, where the surface
n ) 4. overpotential appears in the exponential terms of
The volumetric production rate of species k due to Butler-Volmer kinetics. Advanced iterative algo-
electrochemical reaction occurring at the triple-phase rithms are required in order to ensure efficient and
boundary is given by Faraday’s law converging solutions.8,11-13 It appears that the most
efficient technique is to solve the two discretized
skj algebraic equations simultaneously using Newton’s
Sk ) - (4)
nF method.8,11 For one-dimensional problems, such a
direct solution method results in the often-used
with Band-J subroutine developed by Newman8 in the
[ ( ) ( )]
1960s, in which the Jacobian matrix resulting from
RaF RcF Newton’s procedure is inverted using LU factoriza-
j ) aio exp η - exp - η (5) tion in each Newton iteration. However, the direct
RT RT
solution is computationally inefficient for a large
where j is the volumetric transfer current of the sparse Jacobian matrix arising from two- and three-
reaction, a is the electrochemically active area per dimensional problems. Recently, Wu et al.11 pre-
unit of electrode volume, io is the exchange current sented a set of modern numerical algorithms to
density, Ra and Rc are the anodic and cathodic efficiently solve the two strongly coupled potential
transfer coefficients, respectively, F is Faraday’s equations. A Krylov iterative solver, the generalized
constant, R is the universal gas constant, and T is minimal residual subroutine (GMRES), was used
the cell temperature. Equation 5, commonly known instead of the direct solver (Gauss elimination) to
as the Butler-Volmer equation, describes a large improve solution efficiency of the large, nonsymmet-
class of electrode reactions.8 Under facile and slug- ric Jacobian system. In addition, a nonlinear Gauss-
gish kinetics, the Butler-Volmer equation is further Seidel method provided the initial guess for the
reduced to linear and Tafel expressions, respectively. Newton iteration, and the GMRES solver was pre-
The exchange current density for a reaction, io, conditioned with a block Gauss-Seidel and multigrid
depends strongly on the compositions and tempera- algorithm with a smoother based on the tridiagonal
ture at the reaction interface. The surface over- matrix algorithm (TDMA).
Figure 2. PEFC configuration for numerical modeling.

Figure 3. Speedup of a massively parallelized PEFC
model.15
Within the framework of commercial CFD codes
where sequential solution methods are standard, as
they need to solve a number of user-specified trans- sors. As a fuel cell features a minimal degree of
port equations, the two potential equations must then interactions between subdomains along the flow
be solved through innovative source term lineariza- direction, this problem is inherently suited for paral-
tion.14 lel computing with low communication cost. Bench-
mark calculations were carried out in a Linux PC
2.2.3. Mesh Requirement and Parallel Computing cluster comprised of 50 processors of 1.4 GHz AMD
Another computational challenge in fuel cell mod- Athlon Thunderbird CPU and 512 MB DDR SDRAM.
eling is the need for large meshes for simulation of A local 100 Mbps switched Ethernet network was
industrial-scale cells. To fully appreciate the large used for message communication. Figure 3 displays
mesh required, consider a typical geometry of a large- the computational performance of the PC cluster,
scale PEFC displayed in Figure 2, where the anode demonstrating greater than 7× speed-up by 10 CPUs
flow field (top) and cathode flow field (bottom) running the massively parallelized PEFC model.
sandwich a membrane-electrode assembly (MEA) Roughly 300 iterations are needed for a typical case
with two gas diffusion layers (GDL). Let us define run, requiring about 1.5 h of computational time
the through-plane direction to be perpendicular to the using 10 CPUs. This clearly demonstrates the ef-
MEA, the along-channel direction to be of gas flow, fectiveness of parallel computing in reducing actual
and the in-plane direction to be across channels computing time for these intensive simulations.
through the land area. Suppose that 6-8 grid points Large-scale calculations using 1-10 million com-
are needed to sufficiently resolve each of the five putational cells are presently being carried out in
distinctive regions of MEA and two GDLs, with 10 several industrial and academic organizations using
grid points needed in each of the two gas channels, the parallel computing method and multiprocessor
the minimum number of grid points in the through- computers; see section 3.6 for further detail.
plane dimension is thus between 50 and 60. In the
along-channel direction, 100 grid points are typically 2.3. Material Property Characterization
required since fuel cell channels are long and exhibit The importance of materials characterization in
a large aspect ratio (>100). In the in-plane direction, fuel cell modeling cannot be overemphasized, as
10 nodal points are needed for a channel and a land, model predictions can be only as accurate as their
respectively, thus giving rise to 20 grid points per material property input. In general, the material and
channel within a flow field. For large-scale fuel cells transport properties for a fuel cell model can be
featuring 20-60-channel flow field, a mesh consisting organized in five groups: (1) transport properties of
of 2-6 million grid points is needed. electrolytes, (2) electrokinetic data for catalyst layers
The majority of numerical work published to date or electrodes, (3) properties of diffusion layers or
has used no more than a few hundred thousand grid substrates, (4) properties of bipolar plates, and (5)
points; that is, there is still 1 order of magnitude gap thermodynamic and transport properties of chemical
existing between the current computational capabil- reactants and products.
ity and that required for high-fidelity PEFC simula- The most important electrolyte property is ionic
tion. Meng and Wang15 made one of the first attempts conductivity. For the PEFC system, water and proton
to bridge this gap by developing a massively parallel transport in the polymer electrolyte occurs concur-
computational methodology based on an efficient rently. Springer et al.16 correlated the proton con-
domain decomposition method. The computational ductivity (in S/cm) in the polymer membrane with
domain is divided into a number of subdomains along its water content as follows
the flow direction. Each subdomain is assigned to one
processor with proper load balancing. Each processor 1 1
computes and stores variable solutions within its own [
κe ) exp 1268 (303 - )](0.005139λ - 0.00326) (7)
T
subdomain with necessary interactions on subdomain
boundaries through communication between proces- The strong dependence of electrolyte conductivity on
Figure 5. Electro-osmotic drag coefficient and water

diffusivity as functions of water content in Nafion mem-
Figure 4. Equilibrium water uptake curve for Nafion branes.23,27
membrane: (a) measurement at 30 °C,18 (b) measurement
at 80 °C,17 (c) Springer’s expression, eq 8. eters dictate the water content profile within the
membrane and thus design of a proper water man-
agement scheme for the fuel cell. Water content
water content is evident from the above correlation. dependence of diffusion coefficient and electro-
Several studies17-22 have been carried out to define osmotic drag coefficient has also been extensively
the relation between membrane water content and studied.23-27 Figure 5 indicates that the electro-
moisture of the surrounding environment, called the osmotic drag coefficient is roughly a constant at unity
water uptake curve as shown in Figure 4. Zawodz- when λ < 14 and then linearly increases for water
inski et al.17,18 measured water content of membranes content between 14 and 22. The water diffusion
immersed in liquid water and membranes exposed coefficient in the membrane is highly nonlinear and
to water vapor at variable activities. It is seen that shows a distinctive peak at λ ≈ 3. The nonlinearity
there is a discontinuity in the membrane water in membrane-transport properties excludes analyti-
content between equilibration of the membrane with cal solutions and increases the level of difficulty in
liquid water and with saturated water vapor, both numerical solutions.
of which correspond to unit water activity. This Kreuer1 presented an excellent discussion of ma-
phenomenon, called “Schröeder’s paradox”, is ob- terials and transport properties of proton conducting
served in a wide variety of polymer/solvent combina- membranes other than Nafion.
tions. Water content in the membrane (i.e., the
The most important electrokinetic data pertinent
number of H2O molecules per sulfonic group), λ,
to fuel cell models are the specific interfacial area in
depends on the water activity, a, according to the
the catalyst layer, a, the exchange current density
following fit of the experimental data16
of the oxygen reduction reaction (ORR), io, and Tafel
λ) slope of ORR. The specific interfacial area is propor-
{
tional to the catalyst loading and inversely propor-
0.043 + 17.81a - 39.85a2 + 36.0a3 for 0 < a e 1 tional to the catalyst layer thickness. It is also a
14 + 1.4(a - 1) for 1 e a e 3 strong function of the catalyst layer fabrication
(8) methods and procedures. The exchange current den-
sity and Tafel slope of ORR have been well docu-
where the water activity is in turn calculated by the mented in refs 28-31.
ratio of the water partial pressure to the saturated Diffusion medium properties for the PEFC system
pressure, corresponding to the cell temperature, T. were most recently reviewed by Mathias et al.32 The
Apparently, the water activity and water content all primary purpose of a diffusion medium or gas diffu-
vary spatially; thus, the proton conductivity κe also sion layer (GDL) is to provide lateral current collec-
varies throughout the membrane and catalyst layer. tion from the catalyst layer to the current collecting
In the catalyst layer, the effective proton conductivity lands as well as uniform gas distribution to the
is further modified by Bruggmann factor based on catalyst layer through diffusion. It must also facili-
the ionomer content tate the transport of water out of the catalyst layer.
The latter function is usually fulfilled by adding a
κeeff ) e1.5κe (9) coating of hydrophobic polymer such as poly(tet-
rafluoroethylene) (PTFE) to the GDL. The hydropho-
where e is the ionomer volume fraction. bic polymer allows the excess water in the cathode
Two other important electrolyte properties for the catalyst layer to be expelled from the cell by gas flow
PEFC system are the water diffusion coefficient and in the channels, thereby alleviating flooding. It is
electro-osmotic drag coefficient. These two param- known that the electric conductivity of GDL is
typically an order of magnitude higher in the in-plane ordinary and Knudsen diffusion, is then calculated
direction than the through-plane direction due to the from
strong anisotropy of GDL fibrous structures. By
analogy between electric and heat conduction, it can
be expected that the thermal conductivity of GDL
exhibits the same degree of anisotropy. Pore size
Dk ) (1
+
1
DK,k Di,k )-1
(12)
distribution, porosity, and gas permeability were In the PEFC system, the mean pore radii of
characterized by Mathias et al., albeit without com- catalyst layers are of the order of 0.1 µm. The
pression, which should add a strong effect on the Knudsen diffusion coefficients at 80 °C for O2 and
GDL-transport properties. In addition, the interfacial H2O through the catalyst layer are thus estimated
wetting property, both in the interior and on the to be 0.32 and 0.43 cm2/s, respectively. These values
surface of porous GDL, was recognized as an impor- are comparable to the respective ordinary diffusion
tant material property of GDL. Mathias et al.32 and coefficients, indicating that Knudsen diffusion further
Lim and Wang33 independently performed surface restricts the rates of oxygen and water transport
contact angle measurements, based on variants of the through the cathode catalyst layer in PEFCs and
capillary rise method. Lim and Wang33 further re- should be taken into account.
ported a strong temperature dependence of water Thermodynamic and transport properties of liquids
contact angle on GDL, thereby establishing a link to relevant to the DMFC system can be found in Reid
the realistic thermal environment in an operating et al.35
PEFC. The literature, however, lacks experimental Key material properties for SOFC, such as the ionic
data on the porosity and permeability of GDL under conductivity as a function of temperature, are avail-
compression or more broadly the quantification of able in refs 36-39. In addition, Todd and Young40
GDL deformation under stress and subsequently compiled extensive data and presented estimation
GDL structure-flow interactions. methods for the calculation of diffusion coefficients,
Three properties of the current collector plate are thermal conductivities, and viscosities for both pure
particularly important for CFCD models. One is components and mixtures of a wide variety of gases
electric conductivity, the second is thermal conduc- commonly encountered in SOFCs. Another excellent
tivity, and the third is surface wettability. These source of transport properties for gases and mixtures
properties for materials such as graphite and metals involved in a SOFC is the CHEMKIN thermo-
are well documented in standard textbooks. dynamic database.41
The species diffusivity, Dk, varies in different
subregions of a PEFC depending on the specific 3. Polymer Electrolyte Fuel Cells
physical phase of component k. In flow channels and
porous electrodes, species k exists in the gaseous Polymer electrolyte fuel cells (PEFC) are consid-
phase and thus the diffusion coefficient corresponds ered the most promising alternative power plant for
with that in gas, whereas species k is dissolved in transportation because of their high efficiency, low
the membrane phase within the catalyst layers and emissions, low operation temperature, and low noise.
the membrane and thus assumes the value corre- In the past decade, significant improvements have
sponding to dissolved species, usually a few orders been achieved in PEFC technology, including im-
of magnitude lower than that in gas. The diffusive proved MEA technology and lowered platinum cata-
transport in gas can be described by molecular lyst loading. Excellent reviews of hydrogen PEFC
diffusion and Knudsen diffusion. The latter mecha- research up to the mid-1990s were presented by
nism occurs when the pore size becomes comparable Prater42 and Gottesfeld and Zawodzinski,30 and new
to the mean free path of gas, so that molecule-to-wall advances were reported in the most recent compila-
collision takes place instead of molecule-to-molecule tion of Vielstich et al.43
collision in ordinary diffusion. The Knudsen diffusion Much effort has been expended in the last 5 years
coefficient can be computed according to the kinetic upon development of numerical models with increas-
theory of gases as follows ingly less restrictive assumptions and more physical
complexities. Current development in PEFC model-
DK,k ) ( )
2 8RT
3 πMk
1/2
rp (10)
ing is in the direction of applying computational fluid
dynamics (CFD) to solve the complete set of transport
equations governing mass, momentum, species, en-
where rp is the mean pore radius and Mk the ergy, and charge conservation.
molecular weight of the diffusing gas. Modeling and computer simulation of PEFC began
The binary diffusion coefficient, Di,k, can be either with the pioneering work of Bernardi and Ver-
experimentally measured or calculated using the brugge44,45 and Springer et al.,46,47 whose models are
Chapman-Enskog equation.34 The dependence of the essentially one-dimensional. Fuller and Newman,48
diffusion coefficient on temperature and pressure is Nguyen and White,49 and Yi and Nguyen50,51 subse-
generally given by7 quently developed pseudo-two-dimensional models
accounting for compositional changes along the flow
()()
po path. While such models are useful for small cells,
T 3/2
Di,k(T,p) ) Di,k,0 (11) their applicability to large-scale fuel cells, particu-
To p larly under high fuel utilization and low humidity
conditions, is limited. Nevertheless, the one-dimen-
The combined diffusivity of species k in gas, due to sional models of Bernardi and Verbrugge44,45 and
Springer et al.46,47 provided a fundamental framework at the gas saturation level and account for the extra
to build multidimensional, multiphysics models that water in the form of liquid water in a postprocessing
followed. The pseudo-two-dimensional models devel- manner.57,58,60 Sections 3.1-3.6 are devoted to re-
oped by Fuller and Newman,48 Nguyen and White,49 viewing this class of single-phase models.
and later Yi and Nguyen50,51 further demonstrated The more rigorous approach to liquid water trans-
the important roles played by water and heat man- port is a true two-phase model in which the two
agement in maintaining high performance of PEFCs. phases travel at different velocities. At the same time,
Effects of anisotropy and inhomogeneity on mass the interfacial tension effect and GDL wettability,
transport through GDL were also numerically inves- essential for successful PEFC operation, are fully
tigated.52,53 accounted for. The work of Wang et al.,68 Nguyen et
Gurau et al.54 presented a two-dimensional model al.,69-71 You and Liu,72 Mazumder and Cole,73 Bern-
of transport phenomena in PEFC. This work il- ing and Djilali,74 and Pasaogullari and Wang67,75 falls
lustrated the utility of a multidimensional model in into this category. These two-phase models are
the understanding of the internal conditions of a fuel reviewed in section 3.7.
cell, such as the reactant and water distributions. In Another classification of PEFC models is in terms
a separate development, Yi and Nguyen50,51 formu- of volume-averaged (i.e., macrohomogeneous) models
lated a two-dimensional model to explore hydro- or pore-level models. Macroscopic models assume
dynamics and multicomponent transport in the air local interfacial equilibrium, namely, electrical, chemi-
cathode of PEFC with an interdigitated flow field. cal, and thermal equilibrium at the pore level.
The concept of modeling PEFC within the CFD Conditions of validity of local interfacial equilibrium
framework was proposed shortly thereafter.55-58 Two- were carefully defined.76 All of the above-cited CFCD
dimensional simulations of coupled electrochemical models are, strictly speaking, macroscopic models,
and transport processes were carried out in using a although theoretical inconsistency may exist in some
CFD code,55 and the three-dimensional version of the work. In the two-phase models of Nguyen et al.,69-71
same model was also demonstrated.56,59 Indepen- Mazumder and Cole,63 and Berning and Djilali,62 the
dently, Dutta et al.57 presented three-dimensional water condensation/evaporation rate was expressed
simulation of PEFC using Fluent, a commercial CFD according to the local nonequilibrium condition of
package. Subsequent work was presented in Dutta water vapor in the gas phase, but such a treatment
et al.58 and Lee et al.60 In the model of Dutta et al.57,58 was merely a numerical trick without physical sig-
and Lee et al.,60 the membrane electrode assembly nificance. All simulation results in these works
(MEA) is not included in the computational domain virtually corresponded to the local chemical equilib-
but rather simplified as an interface without thick- rium condition. Moreover, no dispersion effect arising
ness. As such, water transport and ohmic potential from the interaction of local fluctuations in velocity
drop across the membrane are treated using simpli- and species concentration at the pore level was
fied linear relationships as in the early work of included in their water vapor transport equation,
Nguyen and White.49 indicative of an equilibrium macroscopic model. It is
Additional work on general-purpose, fully three- interesting to note that a different approach to
dimensional CFCD models was performed by Zhou treating phase change was used for related problems
and Liu61 using an in-house CFD code, Berning et such as drying of porous materials,77,78 steam injec-
al.62 using the commercial package CFX, and Ma- tion for enhanced oil recovery, and groundwater
zumder and Cole63 using the commercial package contamination and remediation.79
CFD-ACE+. Most recently, both STAR-CD and In the following section we will present first a
Fluent packages have been applied to large-scale fuel general single-phase model along with detailed dis-
cell modeling through their user-coding capa- cussions on several critical issues in PEFC modeling
bility.6,14,15,64-67 Sections 3.1-3.7 will review these (i.e., sections 3.1-3.6). This is followed by two-phase
works in more detail as well as in the context of modeling of liquid water transport in section 3.7. In
various important issues pertinent to the PEFC section 3.8 we will describe experimental diagnostics
system. which can be used to probe the validity of such
models as well as identify future modeling needs.
Efforts have also been made to model two-phase Finally, in section 3.9 we will review work on pore-
flow and transport in a PEFC, a critical but difficult level modeling of the PEFC catalyst layer.
issue that has repeatedly been emphasized in the
literature. To treat liquid water formation in a PEFC, 3.1. Single-Phase Conservation Equations
there are two types of models. The simplest approach
is a single-phase model in which the gas and liquid Various forms of governing equations have been
are considered a single-fluid mixture and thus share used in PEFC modeling, although all fall under the
the same velocity field. Also, the interfacial tension single-phase assumption. To clarify important subtle-
effect is completely ignored. In this case, the total ties with theoretical rigor, in this subsection we
amount of water can be obtained by solving a single summarize a set of conservation equations and
equation without distinguishing water vapor from provide detailed comments of various terms that
liquid water. Once the total water concentration field should be used.
is obtained, one may allow for the water concentra- The single-phase model described herein considers
tion going beyond the saturation level, essentially the total water amount without distinguishing liquid
assuming supersaturation in the gas phase.16,55,59 water from water vapor. This approach is valid under
Alternately, one can truncate the water concentration the condition that liquid saturation within the gas
flux continuity at the channel/GDL interface. In

addition, the permeability in the membrane is as-
sumed to be very small, say 10-30 m2, which ef-
fectively renders the fluid velocity zero. The momen-
tum source term represents Darcy’s drag for flow
through a porous medium, which reduces the mo-
mentum equation to the Navier-Stokes equation in
gas channels, where K f ∞, and to Darcy’s law in
GDL and catalyst layers when K is small.
There is a nonzero mass source in the continuity
equation, Sm, arising from the summation of all
species equations. A general form of this source term
is given in Table 1. Use has been made of the
assumption that summation of interspecies diffusion
within the gas phase is equal to zero. Specifically,
one has
{
Figure 6. Schematic diagram of a polymer electrolyte fuel
cell. Sm )
diffusion layer (GDL) is low (e.g., lower than the
irreducible liquid saturation sir) or liquid droplets are
-MH2
j
2F [
+ MH2O ∇(Dw,m∇CH2O) - ∇ nd
F
ie
( )]
small and disperse in gas flow to form a mist flow. in anode CL
[ ( )]
This approach is particularly well suited for fuel cell ie
j j
simulations under low humidity operation, a major MO2 - MH2O + MH2O ∇(Dw,m∇CH2O) - ∇ nd
application of single-phase models. 4F 2F F
The fuel cell to be modeled is schematically shown in cathode CL
in Figure 6 and divided into seven subregions: the (13)
anode gas channel, anode GDL, anode catalyst layer
(CL), ionomeric membrane, cathode CL, cathode In the above, Dw,m is the water diffusion coefficient
GDL, and cathode gas channel. The present model through the membrane phase only. Note also that the
considers the anode feed consisting either of hydro- water fluxes through the membrane phase, via elec-
gen, water vapor, and nitrogen or hydrogen/water tro-osmotic drag and molecular diffusion, represent
binary gas, whereas humidified air is fed into the a source/sink term for the gas mixture mass in the
cathode channel. Hydrogen oxidation and oxygen anode and cathode, respectively.
reduction reactions are considered to occur only The general species equation can be applied to
within the active catalyst layers where Pt/C catalysts three species, H2, O2, and H2O, with nitrogen being
are intermixed uniformly with recast ionomer. the remaining species of the gas mixture. The species
As explained in section 2, a fundamental PEFC concentration of N2 can thus be simply calculated
model consists of five principles of conservation: from the fact that all species mole fractions sum up
mass, momentum, species, electric charge, and ther- to be unity. Species diffusion can be modeled by Fick’s
mal energy. These five balance equations are now law, which is exact for binary gases but a good
summarized in Table 1 with their source terms approximation for multicomponent gases.53,80 The
identified for various regions of a fuel cell. The central electro-osmotic drag coefficient, nd, is only relevant
idea of this model is to employ a single set of to the water molecule; thus, its values for other
governing equations in all regions throughout a fuel species such as H2 and O2 are set to zero.
cell. The main assumptions invoked in the model are To close the equation set for mass, momentum, and
(1) ideal gas mixtures and (2) incompressible and species conservation, one needs the ideal gas law that
laminar gas flow due to small pressure gradients and describes the gas density varying with its composi-
Reynolds numbers. More explanation of each of the tions. That is
five conservation equations follows.
In the mass and momentum equations for both gas F ) ∑MkCk (14)
channels and GDL, the superficial velocity is used
in order to automatically ensure the normal mass where the N2 molar concentration can be calculated
Table 1. Single-Phase PEFC Model: Governing Equations with Source Terms Identified in Various Regionsa
source terms
conservation equations diffusion layers catalyst layers membrane
mass ∂(F)/∂t + ∇(Fu
b) ) Sm Sm )∑kMkSk + MH2O∇(Dw,m∇CH2O)
momentum b)/∂t + 1/∇(Fu
1/[∂(Fu bb
u)] ) -∇p + ∇τ + Su Su ) (-µ/K)u
b Su ) (-µ/K)u
b b
u)0
species ∂(Ck)/∂t + ∇(u
bCk) ) ∇(Dkeff∇Ck) + Sk Sk ) -∇[(nd/F)ie] - (skj/nF) Sk ) -∇[(nd/F)ie]
charge ∇(κeff∇Φe) + SΦ ) 0 SΦ ) j
∇(σeff∇Φs) - SΦ ) 0
energy ∂[(Fcp)mT]/∂t + ∇(Fcpb
uT) ) ∇(keff∇T) + ST ST ) j[η + T(dUo/dT)] + (ie2/κeff) ST ) ie2/κeff
a Electrochemical reaction ∑ s M z ) ne-, where M ≡ chemical formula of species k, s ≡ stoichiometry coefficient, and n ≡
k k k k k
number of electrons transferred. In PEM fuel cells there are (anode) H2 - 2H+ ) 2e- and (cathode) 2H2O - O2 - 4H+ ) 4e-.
by CN2 ) p/RT - CH2O - CO2 (or CH2). In the case of

H2/H2O binary gas in the anode, the H2 species
transport equation does not need to be solved;
instead, its concentration is calculated by CH2 ) p/RT
- CH2O.
Note that the single-phase model described above
fully couples the mass, momentum, and species
conservation equations through various source terms.
One elegant simplification of this full model is to
neglect the mass source term in the continuity
equation and assume a constant gas density in the
momentum equation. The approximation of constant
density is well supported by the study of Dutta et
al.,58 which predicted a density variation smaller than
5% using a variable-density model, as seen in Figure
7. This simplification thus yields decoupling of the
flow field from the species, electric potential, and
temperature fields, thereby significantly accelerating
the calculations. Possible inaccuracy introduced by
this splitting of the problem may occur on the anode
side; however, the hydrogen concentration profile is
relatively unimportant as the anode overpotential is
typically negligible. Therefore, under common oper-
ating conditions, this reduced model yields <10% and
14% error in the current distribution for anode
stoichiometries of 2.0 and 1.2, respectively.81 In
addition to reducing memory and computational
requirements, this splitting of the fuel cell problem
has the added advantage of allowing for consideration
of different cell voltages/current densities for the
same flow field.
The charge-transport equation includes the elec-
trochemical kinetics for both anode and cathode
catalyst layers. If we assume an infinitely large
electric conductivity of the electronic phase, the
electrode becomes an equipotential line, such that
Φs ) 0 on the anode
Φs ) Vcell on the cathode
If the electric conductivity of electrode matrixes and

plates is limited, an additional equation governing
charge transport in the electronic phase would have
to be solved. This issue is separately addressed in
section 3.4.
The source term in the charge equation is used to Figure 7. Velocity vectors and gas density contours under
describe the transfer current between the electronic very low humidity operation: (a) in the middle and (b) at
and electrolyte phases inside of each anode and the exit of a 10 cm2 PEFC.58
cathode catalyst layer. The transfer current densities
are expressed as follows
kinetic expression is derived by linearizing the
( )( )
CH2 1/2
Ra + Rc Butler-Volmer equation on the assumption that the
anode: j ) aio,aref Fη (15) HOR reaction is facile and hence the surface over-
CH2,ref RT potential is small. On the other hand, the ORR
( ) ( )
kinetic reaction is slow, causing high overpotential.
CO2 Rc Therefore, the ORR kinetic expression is obtained by
cathode: j ) -aio,cref exp - Fη (16) neglecting the anodic reaction term of the Butler-
CO2,ref RT Volmer equation. It should be noted that the ORR is
assumed to be the first-order reaction based on early
These kinetic expressions represent the hydrogen experimental work of Bernardi and Verbrugge45 and
oxidation reaction (HOR) in the anode catalyst layer Gottesfeld and Zawodzinski.30 The value of (Ra+Rc)
and oxygen reduction reaction (ORR) in the cathode in HOR kinetics should be equal to 2, while Rc ) 1 in
catalyst layer, respectively. These are simplified from ORR corresponds to a Tafel slope of approximately
the general Butler-Volmer kinetics, eq 5. The HOR 60 mV/decade. The thermodynamic equilibrium po-
tentials of anode and cathode are assumed to be zero

and a function of temperature, respectively, as82
Uo ) 1.23 - 0.9 × 10-3(T - 298.15) +

RT 1
2F (
ln pH2 + ln pO2 (17)
2 )
This equilibrium open-circuit potential for a H2/air
fuel cell is calculated from thermodynamic data of
reaction enthalpy and entropy changes.
The energy equation entails a detailed account of
heat generation due to irreversible heat of the
electrochemical reaction, reversible (or entropic) heat,
and Joule heating.83 The heat generation term in a
CFCD model must be unambiguous and location
specific. More discussion is deferred to section 3.3.
In addition, the heat accumulation in a porous
material consisting of the matrix and fluid is given
by Figure 8. Two-dimensional sketch of water management
in a PEM fuel cell whereby the membrane-electrode
(Fcp)m ) (Fcp) + (1 - )(Fcp)s (18) assembly separates the anode feed channel from the
cathode, and a diagram of water uptake profiles in anode
and cathode channels.
where the heat capacitance (Fcp)without subscript
refers to the fluid and with subscript “s” refers to the also reduce cathode flooding and volume as well as
solid matrix. cost of the external humidifier. Obviously, this op-
Six coupled governing equations listed in Table 1 erating regime hinges upon a good understanding of
are valid in all regions of a PEFC, and fluxes at an water generation, transport, and distribution within
internal boundary between two adjacent regions are PEFCs.
automatically continuous. Such a single-domain model General features of water transport in a PEFC
is well suited for CFD implementation. In contrast, under low humidity operation is schematically shown
multidomain models, such as the one developed by in Figure 8. Water is produced in the cathode catalyst
Dutta et al.,57,58 compute separate solutions for the layer as a result of ORR or can be brought into the
anode and cathode subdomains, respectively, and cell by humidified reactant gases represented by
then patch the two solutions through the water anode/cathode inlet relative humidity. Through the
transport flux across the MEA interface. Numeri- electrolyte between the anode and cathode, two
cally, this model is characterized as a solver-in-solver modes of water transport are operative: the electro-
situation. osmotic drag effect and diffusion driven by the
concentration gradient. The water flux due to the
3.2. Water Transport electro-osmotic drag is directly proportional to the
Water management is one of the most critical and protonic flux (Icell/F). The diffusion flux of water in
widely studied issues in PEFC. Water management the membrane is usually described by a water dif-
is referred to as balancing membrane hydration with fusion coefficient and the gradient in molar concen-
flooding avoidance. These are two conflicting needs: tration of water.
to hydrate the polymer electrolyte and to avoid Typical water concentration profiles in the anode
flooding in porous electrodes and GDL for reactant/ and cathode flow channels are also depicted in Figure
product transport. 8. The water concentration at the anode/membrane
Early modeling studies16,44,48,49,54,55,57,84,85 have fo- interface first decreases along the flow and then
cused on fully humidified fuel/air streams, which reaches the minimal point at which the net water flux
ensures a hydrated ionomer membrane, thus yielding across the membrane becomes zero, implying that the
a constant proton conductivity and the least ohmic electro-osmotic drag is balanced by the back-diffusion
loss. Under those full humidification conditions, of water. Thus, this location also represents a bound-
however, the gas inside the fuel cell becomes over- ary between drag- and diffusion-dominated regimes.
saturated and water condensation may occur on the The anode water concentration increases after this
cathode side even at low operating current densities. minimum point due to enhanced back-diffusion driven
The ensuing cathode flooding by liquid water subse- under a larger cathode water concentration resulting
quently hampers the access of oxygen to the cathode from ORR. On the cathode side, the water concentra-
catalyst layer, resulting in a significant concentration tion profile can take two routes, depending on the
polarization. In addition, full humidification of the membrane thickness. For thin membranes such as
reactant gases externally requires additional para- Nafion 111 or Gore membranes, water exchange
sitic volume and power in a fuel cell system. There- between the anode and cathode via diffusion is so
fore, the recent trend in the industry is to operate effective that both sides achieve equilibration suf-
PEFC under low humidity conditions while still ficiently downstream. This is shown as curve C1 in
maintaining an adequately hydrated polymer mem- Figure 8. However, thick membranes such as Nafion
brane. Such low humidity operating strategies will 112 will make the water concentrations on two sides
of the membrane further diverge, as shown as the nonlinear functions of water content. It was found
curve C2 in Figure 8. that the nonlinearity of the transport properties leads
Figure 8 also nicely describes two possible sce- to the dryness on the anode side of the membrane,
narios for electrode flooding. Defining the onset of in accordance with the experimental observations of
electrode flooding by the dimensionless water con- Buchi and Scherer.89 While this group of models is
centration reaching unity (i.e., gas becomes satu- fundamental in nature and reveals rich physics of
rated), Figure 8 shows that both anode and cathode water distribution through the polymer membrane,
will be flooded by liquid water condensed from the it is unable to capture all intricate features of water
gas for thin membranes, while for thick membranes transport in a low humidity PEFC.
only the cathode is susceptible to flooding. The second group attempted to model water trans-
Water-transport characteristics become quite in- port throughout an entire fuel cell using the CFD
tricate in low humidity PEFCs. For example, the approach.62,63,90 These CFD models do not appear to
variation in water concentration down the channel properly model water transport through the polymer
is pronounced as the cell is self-humidifying the inlet membrane. Berning et al.62 indicated that no water
dry gases. The in-plane variation from channel to diffusion in the membrane is included in their model,
land is also substantial, giving rise to a fully three- and thus it is restricted only to conditions with fully
dimensional distribution of water concentration. In hydrated membranes. The single-phase model of
addition, strong convection effect perpendicularly Mazumder and Cole63 used neither a water diffusion
through the anode GDL, as induced by significant coefficient nor a drag coefficient in the membrane and
hydrogen consumption under low anode stoichiome- therefore is expected to completely ignore any water
try, challenges the conventional model based on transport across the membrane. Obviously, this
diffusive transport. It seems that a comprehensive group of models is unable to describe water-transport
flow-transport-electrochemical coupled model is phenomena through the membrane in the single-
necessary to faithfully capture the water-transport phase regime (e.g., low humidity).
characteristics in low humidity PEFCs. The last group of studies55,57-59,91 strived to unite
Considerable efforts have been expended in the different modes of water transport in various parts
literature to simulate water management in PEFCs. of a PEFC and develop a comprehensive water-
These studies can be grossly divided into three transport model that is both reflective of correct
groups. One group is focused on the physics of water physics, particularly the water transport through the
transport and scrutinizes water transport through membrane by diffusion and electro-osmotic drag, as
the polymer membrane. Notable work includes well as easy to implement for an entire cell using the
Springer et al.,16 Fuller and Newman,48 Nguyen and CFD approach. This group of models is capable of
White,49 Futerko and Hsing,21 Hsing and Futerko,86 simulating water management scenario in low hu-
Janssen,87 and Kulikovsky.80 Springer et al.16 first midity cells. Note that the model of Dutta et al.57,58
introduced a diffusion model for Nafion 117 mem- uses an approximate analytical solution for water
brane and included the electro-osmotic drag term as transport through the membrane by assuming a
well. They assumed equilibrium between the gas constant diffusion coefficient and drag coefficient and
phase and the membrane phase of water in the then embeds it into numerical solutions to the anode
Nafion membrane and then determined the mem- and cathode sides separately. This approach elimi-
brane water content at the interface by the activity nates the MEA from the computational domain.
of water vapor. Hsing and Futerko86 developed a two- However, such a model loses the ability to resolve
dimensional finite element model for a PEFC without increasingly important phenomena in MEA. For
external humidification of the inlet gas streams. Flow example, the ionic resistances in the two catalyst
field was approximated using the potential flow layers were treated by ad hoc assumption in this type
equation with a stream function. The membrane of models. As PEFC technology moves toward the use
water content at the anode/membrane interface was of thinner membranes, the ionic resistance in the
calculated under the assumption of phase equilibri-
catalyst layer becomes important and comparable
um between the vapor and membrane phases, and
with that in the membrane and therefore must be
the cathode/membrane interface was assumed to be
accurately accounted for. In addition, a PEFC model
fully hydrated. More recently, Janssen87 demon-
treating the MEA as an interface without thickness
strated 1-D and 2-D numerical models to account for
essentially ignores the membrane water storage
water transport through the MEA under various
humidification conditions of inlet gases. The com- capacity and therefore cannot be used in simulating
puted results were compared with experimental data transient phenomena (see section 3.5).
of Janssen and Overvelde.88 In the model of Jans- In contrast, Wang and co-workers55,56,59,91 developed
sen,87 it was assumed that the cell current density a unified water-transport model applicable through-
is constant everywhere, and the catalyst layers were out a PEFC including the MEA region, thus enabling
excluded in the water-transport calculation. Reacting comprehensive CFCD simulations to be carried out.
species along the channel and in the through-plane The model recognizes that there are different phases
direction were calculated based on the one-dimen- of water existing in various regions of the PEFC, i.e.,
sional mass-transport assumption. Kulikovsky80 pre- water in the gas phase and in the membrane. As a
sented a multidomain model of water transport in result, phase equilibrium is considered and various
which the diffusion and electro-osmotic drag coef- modes of water transport, diffusion, convection, and
ficients through the polymer membrane are taken as electro-osmotic drag are incorporated in the unified
Figure 9. Membrane water content (λ) distribution in a

three-straight channel fuel cell (only the half cell is shown Figure 11. Current and water activity distributions in a
due to symmetry) with anode and cathode counter-flow at low humidity 50 cm2 fuel cell with serpentine flow field at
0.65 V or average current density of 1.1 A/cm2. Anode and 0.6 V or average current density of 0.71 A/cm2. The
cathode feed conditions are humidification temperature ) membrane is 18 µm thick (EW < 1000). The anode/cathode
50 °C, pressure ) 2 atm, stoichiometry ) 2, and the cell feed conditions are pressure ) 3/3 atm, relative humidity
temperature ) 80 °C. ) 75%/dry, stoichiometry ) 1.2/2, and cell temperature )
80 °C.
thereby showing a higher value in the channel area

than the rib.
3.3. Heat Transport

A PEFC produces waste heat in an amount com-
parable to its electric power output (so that the
energy-conversion efficiency is around 50%). Heat
generation in PEFC stems from entropic heat of
reactions, the irreversibility of the electrochemical
reactions, and ohmic resistance (or Joule heating).83
Figure 10. Current distribution (A/m2) in a low humidity Roughly, the irreversible reaction heat, entropic heat,
fuel cell at 0.65 V or average current density of 1.1 A/cm2 and Joule heating in a PEFC account for 55%, 35%,
with the same operating conditions as in Figure 9. and 10% of total heat release, respectively. The last
source of heat due to water condensation/evaporation
will be discussed in section 3.7 as it is related to two-
water-transport equation. A unified water-transport phase transport. In addition, the polymer membrane
equation within the single-domain CFCD framework has low tolerance for temperature deviation from the
is developed with the aid of the equilibrium water operating point. Hydration of polymer membranes
uptake curve in the membrane phase. Details were also strongly depends on the temperature as the
provided by Um.91 Figure 9 illustrates an example water vapor saturation pressure is an exponential
of membrane water content distribution for a low function of temperature. Thus, thermal management
humidity, straight-channel cell with anode and cath- in PEFC is inherently coupled with water manage-
ode humidification temperatures equal to 50 °C for ment and the two factors combine to ensure high
the 80 °C cell temperature. Figure 10 shows current performance and durability of a PEFC.
distribution in the same cell. A number of PEFC thermal models have been
Figure 11 displays calculation results of the current published in the literature. Early efforts were made
and water activity distributions in a 50 cm2 fuel cell by Nguyen and White49 and Fuller and Newman.48
with serpentine flow field. The cell is operated under Nguyen and White49 developed a two-dimensional
rather dry conditions with the anode and cathode PEFC model with one-dimensional heat transfer in
relative humidity being 75% and 0, respectively. The the flow direction. The model considers phase change
line graphs are plotted along the center of the of water in the flow channel as the only heat source,
membrane in the cross-section, cutting through the allowing convective heat transfer between the gas
middle of the cell. The simulation shown in Figure and solid phases. On the other hand, Fuller and
11 exemplifies the present capability of a compre- Newman48 developed a pseudo-two-dimensional ther-
hensive water- and heat-transport model in providing mal PEFC model with one-dimensional mass transfer
fine details. It is clearly seen that the local current in the through-membrane direction and one-dimen-
density varies from the channel to rib (or land) areas. sional heat transfer in the flow direction. On the basis
Moreover, the current density above the rib is greater of the known enthalpy change of the overall electro-
than the channel in the inlet portion of the fuel cell chemical reaction, the model calculated the temper-
because the reactant gases are relatively dry and the ature rise of the flowing gas streams with various
ribs protect the membrane from losing water. In external heat transfer coefficients. However, in these
contrast, near the outlet portion of the fuel cell where early models the temperatures of membrane and
gases become more humidified, the local current electrodes were not differentiated, even though a
density begins to be controlled by oxygen transport, significant temperature variation in those regions can
be expected in both the through-membrane and flow

directions depending on the geometric and operating
conditions. Furthermore, the major heat source terms,
the entropic and irreversible reaction heats, were not
specified in their models. Yi and Nguyen50 extended
Nguyen and White’s49 model to include the entropic
and irreversible reaction heats along with the phase
change heat. However, this model allowed the tem-
perature variation of the solid phase in the flow
direction only, assuming uniform temperature in the
through-membrane direction.
Wöhr et al.92 developed a one-dimensional thermal
model for heat and mass transfer in the through-
plane direction, particularly for PEFC stacks. Ac-
counting for the entropic and irreversible reaction
heats, they computed the temperature profile in the
through-membrane direction and predicted the maxi-
mum temperature as a function of the number of cells
contained in a stack. Rowe and Li93 also developed a
one-dimensional model in the through-membrane
direction. Including entropic, irreversible, and phase
change heats, they further took account of Joule Figure 12. Temperature distribution (in K) in the middle
heating in the membrane and anode/cathode catalyst of the membrane (EW < 1000) for a straight-channel PEM
fuel cell (cell voltage ) 0.6 V, average current density )
layers. This work predicted the temperature varia- 1.42A/cm2) with a cell temperature of 80 °C.
tion in the through-membrane direction under the
various current densities and electrode thermal tropic heat, and Joule heating.83 The reversible heat
conductivities. A major drawback of this model is the related to T(dUo/dT) was often overlooked in the
fact that the cathode catalyst layer is assumed to be majority of thermal modeling work in the literature.
fully hydrated and the membrane water content is It represents the entropic heat produced by the cell
linearly interpolated, indicating inapplicability of the even under thermodynamic equilibrium and is a
model in low humidity operation. significant contributor to overall heat generation.
Multidimensional thermal models were presented Solving the energy equation provides prediction of
by many PEFC modeling groups. Maggio et al.94 the temperature distribution and its effect on cell
performed the pseudo-three-dimensional simulations, performance in a PEFC. Figure 12 presents a tem-
neglecting the temperature gradient in the flow perature distribution in the middle of the membrane
direction. However they only considered the overall for a single-channel PEFC. The maximum tempera-
heat source term as (Uo - Vcell)I. In the three- ture rise in this case is 4 °C, which will only affect
dimensional models developed by Shimpalee and cell performance slightly. However, the temperature
Dutta95 and Costamagna,96 the heat source terms variation depends strongly on the thermal conduc-
were also treated globally and not made location- tivities of the GDL and flow plate as well as thermal
specific. Berning et al.62 presented a three-dimen- boundary conditions.
sional PEFC model including irreversible and en- Among the three heat-generation terms, the ir-
tropic heat terms in the catalyst layers and Joule reversible and reversible heat sources of ORR are
heating in the membrane. The main drawback of this dominant. For a straight-channel cell shown in
model is assuming the membrane to be fully humidi- Figure 12, the total amount of heat release is 2.57
fied, indicating limited applicability to low humidity W, of which the irreversible heat is 55.3%, the
operating conditions. Zhou and Liu61 developed a reversible heat 35.4%, and the Joule heat only 9.3%!
three-dimensional PEFC model in which the entropic The total heat released from the fuel cell can also be
reaction heat (∼35% of total heat) was ignored. estimated from the overall energy balance, i.e.
Accurate modeling of the temperature distribution
in a PEFC requires accurate information in four
areas: heat source, thermal properties of various
components, thermal boundary conditions, and ex-
Qheat ) (Uo - Vcell)IavgA + IavgATavg - ( dUo
dT ) )
2.56 (W) (19)
perimental temperature-distribution data for model
validation. The primary mechanism of heat removal This is close to the simulated value of 2.57 W,
from the catalyst layers is through lateral heat including the irreversible, entropic, and Joule heat.
conduction along the in-plane direction to the current The energy-conversion efficiency of this cell, defined
collecting land (like a heat sink). Heat removed by as the ratio of the electric power (IavgVcellA) to the total
gas convection inside the gas channel accounts for energy consumption (IavgVcellA + Qheat), is then cal-
less than 5% under typical PEFC operating condi- culated to be 40%. These simple calculations help
tions. demonstrate the validity and accuracy of a noniso-
The energy equation in Table 1 contains location- thermal PEFC model.
specific heat generation terms, including irreversible Co-simulation of water and heat management in
heat of the electrochemical reaction, reversible en- a 50 cm2 cell can be seen from Figure 13. The
To calculate the electron-transport effect through

GDL and flow plate, the charge conservation equa-
tion for the electronic phase must be solved addition-
ally, namely
∇(σeff∇Φs) - j ) 0 (20)
where σeff is the conductivity of the electronic phase,

Φs is the electronic phase potential, and j is the
volumetric transfer current density inside the cata-
lyst layer. When applied to the flow plate, eq 20
accounts for the ohmic loss occurring in plates with
limited conductivity and small thickness (e.g., graph-
ite plates used in fuel cell stacks).
Meng and Wang14 performed a three-dimensional
study of the effects of electron transport through the
Figure 13. Temperature distribution in a low humidity
50 cm2 fuel cell under the same operating conditions as in gas diffusion layer (GDL). Figure 14 displays the
Figure 11. current density distributions in a single-channel fuel
cell predicted by including lateral electron transport.
membrane temperature fluctuates between channel As can be seen, without accounting for the lateral
and land areas. Furthermore, the membrane tem- electron transport in GDL, the current density is
perature is higher in the channel area due to lack of highest in the middle of the channel because the
effective cooling, increases along the flow path due reaction zone there has the easiest access to oxygen.
to higher current density, and then decreases near However, once considering the electronic resistivity
the outlet due to the current density lowered by of GDL in the in-plane direction, the highest current
oxygen depletion. density region is shifted toward the edges of the
An important issue requiring future work is the channel because these locations have the best com-
consideration of anisotropic thermal transport since bination of easy access to oxygen and short path for
the thermal conductivity has an order of magnitude electron transport onto the current-collecting land.
difference between the in-plane and through-phase The effective electric conductivity of GDL used in this
directions as mentioned earlier. simulation was assumed to be isotropic and equal to
300 S/m, corresponding to carbon paper from SGL
Group. This new finding clearly indicates the neces-
3.4. Electron Transport sity to include the lateral electron transport in future
An important function of GDL is to provide electron CFCD models for high-fidelity computer simulations.
lateral conduction, serving as a bridge between the Physically, this study revealed a new possibility to
catalyst layer and the current-collecting land. In control the lateral current profile by tailoring the
nearly all the three-dimensional PEFC models pub- GDL electric property.
lished to date, electron transport in the catalyst layer/ The effect of the flow plate electric conductivity on
GDL as well as in the current collector plate is the axial current profile is shown in Figure 15.97 It
ignored by assuming a sufficiently large electric is seen that only with extremely conductive metal
conductivity and, consequently, uniform electronic plates (i.e., >2 × 105 S/m) does the constant current
phase potential in these materials. The flow plate has boundary condition also result in an equipotential
an electric conductivity on the order of 2 × 104 S/m line in the plate and hence the same current distri-
for graphite plates, but the effective electric conduc- bution. For graphite plates with an electric conduc-
tivities of GDL and catalyst layers range only from tivity of ∼2 × 104 S/m, the potential drop in the plate
300 to 500 S/m (in the through-plane direction). between the inlet and outlet locations amounts to 27
These electronic resistances could significantly change mV under the constant current boundary condition,
the current distribution, both macroscopically along thereby causing a much flatter current density profile
the flow direction and mesoscopically between the within the membrane than that observed when the
channel and land areas. Although the electron- plate is subject to constant voltage. The example
transport equation was solved in the one-dimensional shown in Figure 15 indicates the important impact
PEFC models of Bernardi and Verbrugge,44,45 the of bipolar plate electric conductivity on current
lateral electronic resistance can only be assessed in distribution, a new subject requiring attention in fuel
a three-dimensional geometry. Furthermore, solving cell engineering. Figure 15 also exemplifies a capabil-
the electron-transport equation enables (1) direct ity to simulate fuel cell performance under constant
incorporation of the contact resistance at interfaces total current boundary condition instead of constant
between two mating components in the solution cell voltage. This approach is fundamentally different
process, (2) direct implementation of the total current from the common practice in which the cell voltage
as a boundary condition instead of the cell voltage is iterated until the resulting total current is matched
(permitting a true stoichiometry to be employed in with the prescribed value by running a number of
PEFC modeling and thus consistency with typical cell simulations.
fuel cell experiments), and (3) stack modeling with Meng and Wang14 found that addition of numerical
cells connected in series. solution to eq 20 substantially increases computa-
Figure 14. Current distribution in a straight-channel fuel cell at a cell voltage of 0.6 V: (a) without GDL electronic
resistance and (b) with GDL electronic resistance.
This gives a time constant between 0.1 and 1 s. The

slowest process, however, is membrane hydration, the
time constant of which can be estimated by
Fδm∆λ I
/
EW 2F ( ) (21)
This is to assume that a dry membrane is hydrated

by production water generated at the current density,
I. For Nafion 112 and a reference current density of
1 A/cm2, this is about 25 s! Therefore, for low
humidity cells where the membrane undergoes water
content changes, the water accumulation term is
essential for transient analyses.
Um et al.55 were one of the first to study the
Figure 15. Current density profiles as a function of the transient response of a PEFC to a voltage step
flow plate electric conductivity (S/m) in a straight-channel change. Due to the fully humidified gas feed on both
fuel cell with an average current density of 0.8 A/cm2.97 the anode and cathode, the membrane hydration
process no longer contributed to the cell transients;
tional time. In the most recently released commercial as such, it was sufficient to include gas-transport
CFCD code, CD-Adapco Japan apparently has de- transients under these fully humidified conditions.
veloped a proprietary solution technique for eq 20 and For low humidity PEFC, Wang and Wang66 further
the electrolyte potential equation together which does extended the work of Um et al.55 to include the
not incur any increase in computational time.98 membrane hydration transient process. Figure 16
Moreover, their code, for the first time, accounts for shows a time response of the average current density
anisotropic electric conductivity in the through-plane to a step change in cell voltage in a low humidity fuel
and in-plane directions. Future need certainly exists cell. Overall, it can be seen that the current density
for more efficient numerical algorithms to solve eq does not simply follow a smooth variation from the
20 in a coupled fashion with the electrolyte potential steady state at 0.6 V to the steady state at 0.7 V.
equation, as described in section 2.2.2. Instead, there exists significant undershoot and
overshoot due, respectively, to the transient phenom-
3.5. Transient Phenomena ena of gas transport and membrane hydration. In
addition, with the inclusion of membrane hydration,
Dynamic characteristics of a fuel cell engine are of the fuel cell transients occur over as long as 15 s, a
paramount importance for automotive application. magnitude needing careful attention for automotive
Three primary processes govern the time response application.
of a PEFC. They are (1) electrochemical double-layer Specifically, Figure 16 shows that the current
discharging, (2) gas transport through channel and density in a cell with dry cathode gas feed drops
GDL, and (3) membrane hydration or dehydration nearly instantaneously once the cell voltage is re-
(i.e., between a dry and a hydrated state). The time laxed from 0.6 to 0.7 V due to the fact that the
constant of double-layer discharging is between electrochemical double-layer effect has a negligibly
micro- and milliseconds, sufficiently short to be safely small time constant. Further, there exists undershoot
ignored for automotive fuel cells. The time constant in the current density as the oxygen concentration
for a reactant gas to transport through GDL can be inside the cathode catalyst layer still remains low due
estimated simply by its diffusion time, i.e., δ2GDL/Dg. to the larger consumption rate under 0.6 V. As the
Figure 17. Current density profile in the midchannel

cross-section of a 36-channel fuel cell.65
passes are made countercurrent with one air inlet

on the left and the other air inlet on the right,
following a patented idea of Qi and Kaufman.99 Such
a counterflow serpentine flow field is intended to
Figure 16. Transient response of current density to a step promote lateral moisture transport between a dry gas
change of cell voltage from 0.6 to 0.7 V in a fuel cell with channel (near the inlet pass) and its neighboring
a fully humidified anode and dry cathode.66 moist gas channel (near the exit pass) and hence cell
internal humidification. Both large-scale variations
O2 consumption rate becomes smaller under 0.7 V, and channel-to-land local oscillations are visible from
the O2 concentration at the reaction surface recovers, the current distribution shown in Figure 17, where
thus leading to an increase in the cell current density. the anode feed is fully humidified. Figure 18 displays
The current rise time corresponds well with the a profile of the net water-transport coefficient through
characteristic time scale of gas phase transport as the membrane in the same cross-section of the cell.
analyzed above. The rise in the cell current, however, This parameter, R, is a combined result of the electro-
experiences an overshoot because the polymer mem- osmotic drag and molecular diffusion of water through
brane still maintains a higher water content corre- the membrane and is defined as the ratio of the net
sponding to 0.6 V. It then takes about 15 s for the water flux to the protonic flux; thus, the unit is the
membrane to adjust its water content at the steady number of H2O/H+. As shown in Figure 18, R varies
state corresponding to 0.7 V. This numerical example between -0.5 and 1.5 and changes rapidly between
clearly illustrates the profound impact of water closely spaced channels, especially between a dry gas
management on transient dynamics of low humidity channel and a moist gas channel. The parameter
PEFC engines where the polymer membrane relies critical to the choice of a suitable water management
on reaction water for hydration or dehydration. strategy reflects a complex interplay between the
electro-osmotic drag and water diffusion through the
3.6. Large-Scale Simulation membrane. R being positive means the dominance of
Large-scale simulation of a PEFC is needed in electro-osmotic drag. Since the drag coefficient is
order to consider all three dimensions. The impor- constant at unity in the range of water content in
tance of parameter variations in the through-plane question, R-values greater than unity indicate the
direction was first recognized in the earliest 1-D electro-osmotic drag aided by the forward diffusion
models. Subsequent 2-D models addressed the im- of water from the anode (fully humidified) to the
portant facet of changes in reactant and water relatively dry cathode. Negative R-values are indica-
concentrations along the flow direction, especially tive of the dominance of back diffusion of water from
when the modern designs of low humidity and low the cathode (high water concentration created byprod-
stoichiometry emerged. The third dimension, i.e., the uct water) to anode. Figures 17 and 18 clearly
in-plane direction from a channel to a land, has been illustrate the need for large meshes in order to
recognized for its importance of the lateral transport resolve sharp gradients in the channel-to-land direc-
of species and electrons. Here, we are primarily tion in addition to the through-plane and flow direc-
concerned with a mesoscopic scale of approximately tions.
1 mm. On this mesoscale, nonuniform current dis- Other work involving large-scale simulation of
tribution may result primarily from the low diffusion PEFCs was due to Meng and Wang15 and Ju and
rate of oxygen in the restricted area covered by the Wang.6,100
land and GDL resistance to lateral electron motion.
Such a large-scale simulation resolving all three
3.7. Liquid Water Transport
dimensions for industrial-scale fuel cells featuring High current density performance of PEFCs is
tens of flow channels is now being made possible by known to be limited by transport of reactants and
parallel computing on PC clusters using millions of products. In addition, at high current densities,
computational cells.65 Figure 17 shows the current excess water is generated and condenses, filling the
distribution in a 36-channel fuel cell predicted using pores of electrodes with liquid water and hence
2 560 000 grid points. This 50 cm2 fuel cell features limiting the reactant transport to catalyst sites. This
a double-pass serpentine flow field on the cathode. phenomenon known as “flooding” is an important
To be able to operate under dry air feed, the two limiting factor of PEFC performance. A fundamental
Figure 18. Profile of the net water transport coefficient through the membrane in the midchannel cross-section of a
36-channel fuel cell at Vcell ) 0.65V and Iavg ) 0.91 A/cm2.65
understanding of two-phase transport in porous gas while excluding the MEA. Water transport through
diffusion layers (GDL) of PEFC is essential in order MEA is thus completely ignored. Berning and Djilali
to improve performance. Flooding of electrodes is also failed to address the effect of GDL hydrophobic-
generally linked to high current density operation ity. None of these above-mentioned studies are
that results in higher water production rate; yet, intended to investigate and analyze two-phase trans-
flooding can also be seen even at low current densi- port in hydrophobic GDL. A brief review of this
ties under certain operating conditions, such as low subject was given by Wang.5
temperatures and low gas flow rates, due to faster The central question about the physics of liquid
saturation of the gas phase by water vapor. For water transport through hydrophobic GDL has re-
instance, in automotive fuel cells, cell start-up is mained unexplored until recently. Nam and Kavi-
performed under ambient conditions in which the any103 described a one-dimensional model for liquid
temperature is lower than the desired operating water transport through hydrophobic GDL. In this
temperature, commonly referred as “cold start”. At model, the gas-phase pressure is assumed to be
cold-start temperatures, the saturation vapor pres- uniform, thereby rendering the liquid-phase trans-
sure of water is extremely low; therefore, saturation port governed by the gradient in capillary pressure.
of the gas phase with water vapor, hence flooding, The model was used to assess the effects of GDL fiber
may occur even in relatively low current densities, diameter, porosity, and capillary pressure on the
greatly affecting the start-up characteristics and liquid water distribution. Independently, based on
transient dynamics of PEFC operation. extensive experimental observations of liquid water
While there exist several studies investigating the flow in an operating PEFC, Pasaogullari and Wang104
two-phase transport in a PEFC,46,47,69-72,101,102 none most recently proposed a systematic theory of liquid
considered the effect of GDL hydrophobicity. Some water transport through hydrophobic GDL. Conden-
modeling studies were published in order to predict sation results in a tree-like liquid water percolation
PEFC performance at given flooding levels without network in the porous GDL. Liquid water then
attempting to analyze the transport of liquid wa- reaches the interface of the porous GDL with the gas
ter.46,47,101,102 He et al.69 proposed a two-dimensional channel, forming liquid droplets. Inside GDL, liquid
two-phase model for PEFC with interdigitated flow water is driven by capillary (wicking) action. In
field in which they included capillary transport of hydrophobic GDL, the capillary pressure is negative;
liquid water in a completely hydrophilic GDL. Wang hence, the liquid pressure is larger than the gas-
et al.68 developed a two-phase model of the air phase pressure, whereas in hydrophilic media, the
cathode of PEFC also with a hydrophilic GDL based gas-phase pressure is higher than of the liquid phase.
on the M2 formulation (multiphase mixture model) In addition, the liquid pressure increases with
of Wang et al.79 Subsequently, You et al.72 published liquid saturation; therefore, a liquid pressure gradi-
a similar work investigating the effects of several ent is formed from higher to lower liquid saturation
operating parameters on two-phase transport. Most regions. This pressure gradient becomes the driving
recently, Mazumder and Cole73 presented a numer- force for liquid water flow, as schematically shown
ical study also based on the M2 model of Wang and in Figure 19. In PEFC, the liquid saturation is higher
Cheng.79 Mazumder and Cole’s model appears to be at the catalyst layer, due to water generation and
valid only in the two-phase regime where there is electro-osmotic drag, than the GDL-channel inter-
liquid water. Under low humidity inlet conditions face. Therefore, the liquid pressure gradient formed
where the liquid saturation is zero, this model yielded in the GDL drives liquid water from the reaction sites
zero electro-osmotic drag through the polymer mem- toward the gas channel. Figure 19 also displays a
brane. Berning and Djilali74 presented a two-phase nearly constant gas pressure profile across the two-
model for porous GDL and gas channel of a PEFC phase zone due to much lower gas-phase viscosity.
acteristics of the GDL surface. On a hydrophilic

surface, which has a contact angle less than 90°,
liquid will spread, whereas on a hydrophobic surface,
which has a contact angle greater than 90°, the
droplet will be of a more spherical shape, covering
less pore entry, as depicted in Figure 20.105
Both theories of liquid water transport through
hydrophobic GDL proposed by Nam and Kaviany103
and Pasaogullari and Wang104 require a boundary
value of liquid saturation at the GDL/channel inter-
face. This interfacial saturation should not be equal
to the immobile saturation (and hence no liquid flow)
but depends on the break-off (or detachment) diam-
eter of liquid droplets and population density, both
of which are, in turn, strong functions of the GDL
surface contact angle and the gas velocity in the flow
channel. Nothing has been formally published on this
problem, although it is under intensive investigation
in industry.
To summarize, to properly model liquid water
transport and ensuing flooding effect on cell perfor-
mance, one must consider four submodels: (1) a
model of catalytic surface coverage by liquid water
inside the catalyst layer, (2) a model of liquid water
transport through hydrophobic microporous layer
Figure 19. Schematic illustration of liquid- and gas-phase and GDL, (3) an interfacial droplet model at the GDL
pressure profiles in hydrophilic and hydrophobic porous surface, and last (4) a two-phase flow model in the
media.104 gas channel. Both experimental and theoretical works,
That is, it does not incur much gas pressure drop to in academia and industry alike, are ongoing to build
drive the gas flow through thin GDL. models for the four key steps of water generation,
It is interesting to note from Figure 19 that transport, and removal from a PEFC.
although the relative magnitude of the liquid to gas Pasaogullari and Wang67 further developed a 3-D,
pressure is different in hydrophobic GDL than in isothermal, full-cell numerical model for PEFC by
hydrophilic GDL, both media provide capillary action integrating the above-described theory of liquid water
to drive liquid water from the inside to the surface. transport in hydrophobic GDL with other submodels
However, the slope of the capillary pressure (i.e., the described in sections 3.1-3.3. Figure 21 compares the
driving force) is different depending on whether current density profiles along the flow direction from
water is the wetting or nonwetting phase. The a two-channel serpentine fuel cell with the anode and
greater slope of the capillary pressure near the cathode in co-flow, as predicted by the single- and
evaporation front (i.e., s ≈ 0) in hydrophobic GDL is two-phase models. The effect of flooding is clearly
indicative of increased effectiveness of this type of visible, although the difference in the average current
medium for water removal. Furthermore, the liquid density amounts to only 6% between the two model
pressure buildup in the hydrophobic cathode likely predictions. It is seen that the current density near
sets up a hydraulic pressure differential to drive the channel inlet is much higher in the single-phase
water flow back through the membrane into the model than the two-phase model, where it suffers
anode. from severe flooding of the cathode. Liquid water
The shape of liquid droplets emerging at the GDL/ limits oxygen transport and covers a portion of the
channel interface is governed by the wetting char- active catalyst particles, resulting in lower perfor-
Figure 20. Liquid water droplets on GDL of different wettability at 70 °C.105

Figure 21. Comparison of local current density distribu- Figure 22. Local current density profiles along the
tions in a two-channel serpentine PEFC at Vcell ) 0.4 V.67 channel direction for different humidification levels at Vcell
) 0.65 V. Anode and cathode stoichiometries are 1.4 at 1.0
mance near the inlet. In addition, the current dis- A/cm2.67
tribution predicted from the two-phase model is more
uniform because flooding upstream tends to limit cell overall is small. However, for the lower cathode
performance there as well as oxygen consumption. humidity cases with fully humidified anode inlet (i.e.,
This leaves a higher oxygen concentration down- 20% cathode/100% anode and 40% cathode/100%
stream to enhance the current density. Another anode cases), there exist three distinct regimes of the
interesting feature of the current density profiles current distribution: (i) the ionic resistance limita-
shown in Figure 21 is seen around the U-turn tion due to membrane dry out in the first part, (ii) a
location, especially with the single-phase model middle section with highest performance where the
prediction. Due to the mixing effect and secondary membrane is fully humidified, and (iii) mass transfer
flows around the U-turn of serpentine flow channels, limitations due to flooding and/or oxygen depletion
the convective transport of oxygen to the catalyst in the last part of the cell. It is seen that even with
layer is enhanced; hence, a localized spot of higher low cathode inlet humidification levels, i.e., 20% and
performance appears in the single-phase results. The 40%, the water concentration in porous GDL exceeds
same phenomenon is also seen in the two-phase the saturation value and flooding takes place in the
results but at a smaller magnitude, because liquid regions near the channel exit. In these cases the cell
water in GDL hampers the gas flow and reduces the suffers from membrane dry out near the inlet,
level of oxygen-transport enhancement under very especially on the anode side, due to the electro-
small gas velocities. In industrial-scale fuel cells, the osmotic drag of water to the cathode. Near the inlet,
gas-feed velocities through channels would be much the cathode water concentration is lower; therefore,
higher, and this phenomenon may also be expected back-diffusion of water from the cathode to anode
at a larger magnitude even in flooded GDL. does not compensate for the electro-osmotic drag,
Operating parameters such as humidification and resulting in the anode dry out.
flow rates of inlet reactant streams have a substan- As shown in Figure 23, the fully humidified case
tial influence on PEFC performance and transport shows a maximum liquid saturation around 10% near
characteristics. A two-dimensional numerical study the inlet and decreases in the flow direction due to
was performed to investigate these effects.67 In decreasing reaction rate in the cathode catalyst layer.
Figure 22 the local current density distributions at The 10% level of liquid saturation results from using
0.65 V for different humidification levels are given. a realistic GDL permeability on the order of 10-12
It is seen that in the fully humidified case, the local m2.67 Higher liquid saturation values reported in the
current density monotonically decreases along the literature were obtained only by using unrealistically
flow direction as a result of mass transfer limitations small permeability, i.e., 95% maximum liquid satura-
due to both oxygen depletion and flooding. In the case tion with GDL permeability of 7.3 ×‚10-15 m2.70
of 70% anode and 20% cathode inlet relative humid- In the low humidity cases, the cell does not produce
ity, the membrane stays relatively dry throughout liquid water immediately but after the water vapor
the cell and shows a membrane ionic resistance concentration in the gas reaches the saturation value;
limited behavior. The current density increases along therefore, the condensation front is pushed down-
the flow direction, since the membrane water content stream, and its location is directly related to cathode
is increasing due to ORR. In this case, oxygen inlet relative humidity. In the 20% cathode relative
depletion is not severe because the current density humidity case, it is found that the first two-thirds of
Figure 23. Liquid saturation profiles at cathode GDL- Figure 25. Liquid saturation profiles at cathode GDL-
catalyst layer interface along the channel direction for catalyst layer interface along the channel direction of
different humidification levels at Vcell ) 0.65 V. Anode and single-channel PEFC for different inlet stoichiometric
cathode stoichiometries are 1.4 at 1.0 A/cm2.67 ratios at Vcell ) 0.65 V and inlet relative humidity: 20%
cathode and 100% anode at 80 °C.67
metric ratios. However, as the GDL gets saturated
with water vapor, water starts to condense and the
cell starts to suffer from flooding of GDL as well as
oxygen depletion, which result in performance de-
cline. The liquid saturation profiles given in Figure
25 show that with decreasing gas flow rate, the
condensation front moves closer to the channel inlet.
Latent heat associated with phase change in two-
phase transport has a large impact on the tempera-
ture distribution and hence must be included in a
nonisothermal model in the two-phase regime. The
temperature nonuniformity will in turn affect the
saturation pressure, condensation/evaporation rate,
and hence the liquid water distribution. Under the
local interfacial equilibrium between the two phases,
which is an excellent approximation in a PEFC, the
mass rate of phase change, m̆fg, is readily calculated
from the liquid continuity equation, namely
∂(Fls)
m̆fg ) + ∇(Flb
u l) (22)
∂t
Figure 24. Local current density profiles along the
channel direction of single-channel PEFC for different inlet Since the liquid-phase velocity is obtained, for ex-
stoichiometric ratios at Vcell ) 0.65V at the inlet relative ample, from the M2 model as follows
humidity: 20% cathode and 100% anode at 80 °C.67
F lb
ul ) bj l + λlFu
b (23)
the cell is free from liquid water, whereas it is
predicted that liquid water starts to appear around Substituting eq 23 into eq 22 yields
one-third of the channel length in the 40% cathode
relative humidity case. ∂(Fls)
m̆fg ) + ∇(bj l + λlFu
b) (24)
The effect of inlet stoichiometry on transport ∂t
characteristics and performance of PEFC was also The heat release or adsorption due to condensation
investigated by Pasaogullari and Wang.67 In Figure or evaporation is simply given by
24 the local current density distributions along the
flow direction are displayed at a cell voltage of 0.65 ST ) hfgm̆fg (25)
V. As explained earlier, the membrane is hydrated
much faster in lower flow rates, and therefore, the where hfg is the latent heat of vapor-liquid phase
performance peak is seen earlier in lower stoichio- change.
Last, all models describing liquid water transport trode assembly (MEA) technique involves either
through GDL can be categorized into two ap- masking different areas or partially catalyzing seg-
proaches: the M2 model and unsaturated flow theory ments of the MEA to determine local current density
(UFT). UFT assumes a constant gas phase pressure, behavior. Second, the subcell technique involves
thus effectively decoupling the liquid water transport electrically isolating individual locations of catalyzed
from gas flow. The models of Nguyen and co-work- anode and opposing cathode from the main cell in
ers,69,70 Berning and Djilali,74 Nam and Kavinay,103 order to measure the performance of the desired
and Pasaogullari and Wang104 belong to UFT. In location. In the passive current mapping technique,
contrast, the M2 model is a more complete two-phase an array of shunt resistors normal to an unmodified
model in which motions of both liquid and gas phases MEA surface are located between the flow field and
are accounted for. The work of Wang et al.,68 You and a buss plate. Voltage sensors passively determine the
Liu,72 Mazumder and Cole,73 and Pasaogullari and potential drop across each resistor, and via Ohm’s
Wang67 are based on the M2 model. Most recently, law, current distribution through the flow plate is
Pasaogullari and Wang75 fully elaborated the differ- determined. While each of the described methods for
ences between the M2 and UFT models for PEFC determination of current distribution has advantages,
GDLs and revealed an interesting new mechanism it is desirable to utilize a nonsegmented MEA in
for oxygen transport that results from the gas coun- order to preserve true fuel cell operation character-
terflow to the liquid water motion. In addition, the istics and avoid highly individualized specialty mem-
roles of a microporous layer played in two-phase branes. Note also that all predictive models are
transport in a PEFC were elucidated.75 developed for nonsegmented MEA.
In addition to the use of conventional MEAs and
3.8. Experimental Diagnostics and Model good spatial resolution, the ability to determine
Validation transient effects from sudden changes in operating
Experimental diagnostics not only help develop a conditions is desired. The nonsegmented passive
fundamental understanding of fuel cell dynamics but current mapping technique of Stumper et al.110 allows
also provide benchmark-quality data for CFCD model transient measurement and has good spatial resolu-
validation. Motivated by both needs, diagnostic ef- tion but requires an array of embedded and highly
forts are presently directed toward local distribution precise shunt resistors. The magnetic loop method
measurements (vs global), multiple and simultaneous of Weiser et al.109 also allows transient measurements
measurements, in-situ and nonintrusive measure- with unaltered MEAs and flow fields but is more
ments, as well as simple and well-defined cell con- difficult to implement than the other methods and
figurations rather than experimentation involving cannot be applied to stacks. Recently, Mench and
various complex configurations of direct interest to Wang demonstrated an improved technique for ac-
the industry. This section reviews work published in curate current distribution measurements first on a
the literature on current distribution, high-frequency DMFC111 and later on a H2/air PEFC.112 Indepen-
resistance distribution, mass distribution, tempera- dently, Noponen et al.113,114 developed and demon-
ture distribution, and two-phase visualization for strated a similar technique.
elucidation of the flooding process. A common problem prevailing in much of the
published work on use of the segmented cell is
3.8.1. Current, Species, and High-Frequency Resistance inability to produce a similar level of current density
Distribution Measurements expected from a nonsegmented single cell. Lower
Quantification of current, specie, and membrane performance originates from the difficulty that a
resistance distributions in a PEFC is critical to segmented cell, if not well designed, fabricated, and
understanding key phenomena, such as water man- assembled, exhibits much higher electric contact
agement, CO poisoning, and flow-field design effect, resistance between the segmented flow-field plate
and providing valuable data for validation of CFCD and GDL. As a result, the measured current distri-
models. As indicated by Wang,4,5 it is this type of bution is more reflective of the contact resistance
detailed validation that will permit an ultimate distribution than the intrinsic distribution of water
understanding of the physicochemical phenomena in and reactants inside the cell. Such an artifact due to
PEFC as well as development of computer-aided tools intrusiveness of the measurement must be minimized
for design and development. in order for the segmented cell technique to produce
To measure the current distribution in a hydrogen truly useful data of current distribution. A future goal
PEFC, Brown et al.106 and Cleghorn et al.107 employed for this field is to produce current distribution data
the printed circuit board approach using a segmented representative of automotive operating conditions,
current collector, anode catalyst, and anode GDL. such as a temperature level of 80 °C, average current
This approach was further refined by Bender et al.108 density level of 1 A/cm2, and voltage level of 0.6 V.
to improve ease of use and quality of information Only then will the segmented cell technique have the
measured. Weiser et al.109 developed a technique potential to become a standard diagnostic tool for fuel
utilizing a magnetic loop array embedded in the cell research and development.
current collector plate and showed that cell compres- Localized AC impedance and current distributions
sion can drastically affect the local current density. were measured by Brett et al.115 on a single linear
Stumper et al.110 demonstrated three methods for the channel with the segmented cathode current collector
determination of current density distribution of a plate. Figure 26 shows their measured current den-
hydrogen PEFC. First, the partial membrane elec- sity and membrane resistance distributions along the
Figure 28. Schematic diagram of the experimental setup

for simultaneous measurements of anode/cathode species,
current, and high-frequency resistance (HFR) distributions
in an operating cell.120
higher HFR is clearly seen near the inlet where the

membrane is dehydrated by dry inlet gases.
Distributions of water and reactants are of high
interest for PEFCs as the membrane conductivity is
strongly dependent on water content. The informa-
tion of water distribution is instrumental for design-
Figure 26. Current (a) and high-frequency resistance (b) ing innovative water management schemes in a
distributions at cell potentials of 0.8 and 0.6 V as measured PEFC. A few authors have studied overall water
by Brett et al.115
balance by collection of the fuel cell effluent and
condensation of the gas-phase water vapor.116-118
However, determination of the in situ distribution of
water vapor is desirable at various locations within
the anode and cathode gas channel flow paths. Mench
et al.119 pioneered the use of a gas chromatograph
for water distribution measurements. The technique
can be used to directly map water distribution in the
anode and cathode of an operating fuel cell with a
time resolution of approximately 2 min and a spatial
resolution limited only by the proximity of sample
extraction ports located in gas channels.
More advanced capability exists in an apparatus
to simultaneously measure species, current, and
high-frequency impedance distributions using seg-
mented fuel cells.120 A schematic diagram of the
experimental apparatus is shown in Figure 28. Two
sets of 12-port electric microactuators and gas chro-
matographs are used to measure gas concentration
Figure 27. Distribution of high-frequency resistances distributions on the anode and cathode sides simul-
measured at 1.5 kHz in a low humidity PEFC using 30 taneously. The arrays of multiple scalar distributions
µm membrane (EW < 1000). obtainable from this apparatus have been the basis
for much improvement in CFCD modeling in our
channel for two cell potentials. Carbon paper along research program at Penn State. Figure 29 shows the
with Nafion 112 was used to construct a MEA in this typical current density distributions along the cath-
experimental cell. Interestingly, there is no signifi- ode flow path in a PEFC with fully humidified anode
cant variation between the membrane resistance and cathode. It is seen that the level of average
along the channel length or between the two set current density matches well with a realistic fuel cell
potentials, despite a dry cathode feed and the fully under similar operating conditions. Moreover, the
humidified anode feed. Figure 27 displays another decline in the local current density along the flow
example of high-frequency resistance (HFR) distribu- path due to oxygen depletion is evident. This decline
tions measured in a low humidity cell with Gore is more severe under lower cell voltage or higher
PRIMEA membrane (30 µm, EW < 1000) and carbon average current density. To illustrate the species
cloth as GDL. These measurements also show little distribution measurement capability by microGC,
variation for different cell potentials. However, the Figures 30 and 31 display water mole fraction
Figure 29. Current density distributions in a fully Figure 31. Water mole fraction profiles along the cathode
humidified PEFC using 30 µm membrane (EW < 1000) gas channel measured by a microGC in a cathode under-
under various cell voltages. humidified PEFC using 30 µm membrane (EW < 1000).
Figure 32. Rise of electrolyte temperature from the initial

temperature at open circuit as a function of current
density.121
Figure 30. Water mole fraction profiles along the anode mole fraction increases along the flow path of the
gas channel measured by a microGC in a cathode under-
humidified PEFC using 30 µm membrane (EW < 1000). cathode gas channel. This increase is nearly linear
until approaching the saturation limit. The informa-
distributions inside the anode and cathode gas chan- tion from Figures 30 and 31 on water distributions
nels, respectively, under the conditions that the in both the anode and cathode sides yielded the
anode inlet is fully humidified and the cathode profile of the net water-transport rate through the
humidification temperature is 50 °C relative to the membrane along the flow direction for the first time.
cell temperature of 80 °C. It is seen from Figure 30
that the anode water mole fraction shows a decrease- 3.8.2. Temperature Distribution Measurements
then-increase trend. Beginning from the inlet, the
anode gas loses water owing to both electro-osmotic Mench et al.121 developed a technique to embed
drag and forward diffusion driven by the higher microthermocouples in a multilayered membrane of
water concentration on the anode side. After a an operating PEM fuel cell so that the membrane
minimum point of water mole fraction at a certain temperature can be measured in situ. These micro-
location downstream of the anode channel, the water thermocouples can be embedded inside two thin
mole fraction begins to increase again once the water layers of the membrane without causing delamina-
concentration on the cathode side increases and tion or leakage. An array of up to 10 thermocouples
enhances back diffusion from the cathode to anode. can be instrumented into a single membrane for
It is interesting to note that the location of minimum temperature distribution measurements. Figure 32
water mole fraction shifts more downstream when shows the deviation of the membrane temperature
the cell voltage is higher or the amount of water in an operating fuel cell from its open-circuit state
production is lower on the cathode side. as a function of the current density. This new data
For a relatively dry cathode inlet (i.e., the humidi- in conjunction with a parallel modeling effort of Ju
fication temperature of 50 °C versus the cell temper- et al.64 helped to probe the thermal environment of
ature of 80 °C), Figure 31 displays how the water PEM fuel cells.
Figure 33. Visualization of liquid water transport in an operating transparent PEFC (45 µm membrane with EW <
1000; GDL, Toray paper TGPH 090 with 20 wt % PTFE loading with a microporous layer).125
3.8.3. Two-Phase Visualization motive fuel cells that feature quite different operating
temperature and current density. It is more difficult
To delineate the origin and development of flooding to visualize two-phase flow and liquid water dynam-
in PEFC, recent research has resorted to two-phase ics in an automotive PEFC operated at higher current
visualization techniques. This can be done either densities (∼1 A/cm2) and higher temperatures, due
through a transparent cell based on optical diagnos- largely to severe fogging.125 Figure 33a shows the
tics or a graphite/metal cell visualized by neutron polarization characteristics of this transparent fuel
transmission radiography (NR). Bellows et al.122 cell under typical automotive conditions.125 Appar-
made the first attempt to measure the water content ently the reasonable electrochemical performance is
profile within a polymer membrane using NR. Owing preserved in this transparent cell. Figure 33b shows
to the NR limitation in spatial resolution, a thick a series of normal-view snapshots of water droplet
membrane of 500 µm was used. Geiger et al.123 emergence on the GDL surface, growth, coalescence,
reported the two-phase flow patterns observed by NR and removal as liquid film from the sidewall. Obser-
in the flow field for the first time. Preliminary results vations of the droplet formation on GDL surface
of two-phase distributions in both PEFC and DMFC reveal a highly transient, cyclic processswater drop-
were given; however, the spatial resolution in the lets emerge from certain locations from the GDL
presented radiographs appeared inadequate. In ad- surface, then grow and sometimes coalesce into a
dition, Geiger et al. discussed the NR limitation in larger drop. When the large drop reaches a size
temporal resolution due to the necessity to use a CCD comparable with the channel dimension, it touches
camera to detect time-dependent processes instead the hydrophilic channel wall in the flow field and
of recording still images to a film with sufficiently spreads laterally along it into a thin liquid film. The
long exposure time. liquid film becomes continuous and flows toward the
To achieve sufficient spatial and temporal resolu- exit of the gas channel under the influence of gas
tion, critical to the capture of two-phase flow phe- shear force. New droplets then form and grow on
nomena in PEFC that is transient in nature and nearly the same spots of GDL, repeating the cyclic
controlled by surface forces, a parallel effort is process. However, on the cell scale, a large number
currently being made in optical diagnostics. Tuber of droplets would exist on a sufficiently large surface
et al.124 attempted to visualize two-phase flow using at any one time, each having its own history of
a transparent test cell made of plexiglass. The cell emergence, growth, and removal. Thus, one can
design and experimental technique are limited to anticipate an average current density that is es-
room temperature (∼30 °C) and low current density sentially at steady state for a given set of operating
(∼300 mA/cm2) conditions. Under these conditions it conditions.
was observed that clogging of gas channels by liquid The visualization study of Yang et al.125 further
water is a major cause for low cell performance. In revealed the profound effects of wetting properties
this operating regime a hydrophilic GDL tends to of gas channel walls. In a gas channel surrounded
enhance the cell performance. It should be noted that by mixed hydrophilic channel surfaces and hydro-
the same conclusion cannot be extrapolated to auto- phobic GDL surface, it was found that water conden-
Figure 35. Comparison between simulated and measured

current distributions in a PEFC with 15 µm membrane
(EW < 1000) and under anode and cathode conditions of
3/3 bar, 1.2/2 stoichiometry, and 75%/0% relative humidity.
The cell temperature is 80 °C.100
simulation and experiment; however, agreement with

the current distribution remains unsatisfactory. Rea-
sons for the discrepancy at the current distribution
level were then analyzed, leading to improvement in
the fundamental understanding of many complex,
interacting phenomena occurring in a PEFC. Valida-
tion against global I-V curves is insufficient and
often misleading.
Detailed validation for low humidity PEFC, where
the current distribution is of more interest and likely
leads to discovery of optimal water management
strategies, was performed most recently.100 Figure 35
shows a comparison of simulated and measured
current density profiles at cell potentials of 0.85, 0.75,
and 0.7 V in a 50 cm2 cell with anode and cathode
Figure 34. Comparison of calculation (lines) and experi- RH of 75% and 0%. Both experimental data and
mental (symbols) results for cathode stoichiometry of 2.0 simulation results display the characteristics of a low
at 0.75 A/cm2 and fully humidified anode and cathode: (a) humidity cell: the local current density increases
average polarization curve and (b) current distribution.
Different colors in panel b represent various cell voltages
initially as the dry reactants gain moisture from
as defined in the figure legend.6 product water, and then it decreases toward the
cathode outlet as oxygen depletion becomes severe.
sation occurs preferentially on hydrophilic surfaces, The location of the peak current density is seen to
thus giving rise to an opportunity to keep condensa- move toward the cathode inlet at the lower cell
tion sites away from the GDL by properly tailoring potential (i.e., 0.7 V) due to higher water production
the wetting characteristics of the channel walls and within the cell, as expected.
hence alleviate GDL flooding. The current state of experimental validation is to
validate comprehensive CFCD models against dis-
3.8.4. Experimental Validation tribution data of multiple parameters, such as cur-
rent and species distributions. This possibility is
While highly sophisticated CFCD simulations are becoming a reality as a new experimental technique
capable of predicting many details, to date model to measure current and species (e.g., water) distribu-
validation has been conducted mostly against the cell tions simultaneously has emerged.120 The trend will
global polarization (or I-V) curve, which is an continue to validate CFCD models against multiple
integral outcome of many interacting phenomena. Ju arrays of distribution data in order to develop a full
and Wang6 made the first attempt to validate a three- understanding of PEFC fundamentals.
dimensional, single-phase, isothermal PEFC model
against current distribution data measured in a 50 3.9. Modeling the Catalyst Layer at Pore Level
cm2 fuel cell operated with fully humidified gas feed.
Figure 34 shows validation results with a cathode The catalyst layer of thickness around 10 µm is a
stoichiometry of 2.0 at 0.75 A/cm2. An excellent match critical component of a PEFC and requires more
with the average I-V curve can be seen between the elaborative treatment. Gottesfeld and Zawodzinski30
provided a good overview of the catalyst layer struc- Tafel kinetics and predicted a doubling of the Tafel
ture and functions. The oxygen reduction reaction slope as the current density increases or the mass-
(ORR) occurs in the cathode catalyst layer, and the transport limitations set in. Eikerling and Kornyshev
hydrogen oxidation reaction (HOR) takes place in the presented analytical expressions in the limiting cases
anode catalyst layer. Both reactions require active of small currents, fast oxygen diffusion, fast proton
catalyst sites to break the molecular bond in the transport, and high currents. To verify the theoretical
diatomic gaseous reactant molecules because of the finding regarding the existence of double Tafel slopes,
low-temperature environment of a PEFC. HOR has Ihonen et al.129 most recently performed an experi-
orders of magnitude higher reaction rate than ORR, mental study of mass-transport limitations in the
which leaves ORR as a source of large voltage loss PEFC cathode catalyst layer. A kinetic slope at low
in PEFC. Due to the acid nature of the polymer current densities and a second Tafel slope at higher
electrolyte and low-temperature operation, Pt- or Pt- current densities were indeed observed on the polar-
alloys are the best-known catalysts. To enable ORR ization curves. The experimental evidence appears
in the cathode catalyst layer, the layer must provide to suggest that oxygen diffusion is limited in the
access for oxygen molecules, protons, and electrons. agglomerates at the pore scale rather than by that
Therefore, the catalyst layer usually consists of: (1) in the gas phase across the thickness of the electrode.
the ionomer (its amount is also called Nafion content, However, the above-mentioned macroscopic models
e) to provide a passage for protons to be transported do not address localized phenomena on the pore level.
in or out, (2) metal catalysts supported on carbon (the To assess the profound effects of the catalyst layer
electronic phase volume fraction, s) to provide a pore structure on polarization performance, Pisani
means for electron conduction, and (3) sufficient et al.130 constructed a pore-level model by volume-
porosity (g) for the O2 gas to be transferred to averaging microscopic governing equations over an
catalyzed sites. The sum of all volume fractions is idealized, one-dimensional model geometry of pores.
equal to unity, and individual volume fractions must Wang and Wang131 recently developed a direct nu-
be optimized to provide the best overall performance merical simulation (DNS) model to describe the
of a catalyst layer. transport of protons, electrons, oxygen, and product
The important processes occurring in a catalyst water at the pore level (on the order of 100 nm)
layer include interfacial ORR at the electrochemically within a random, microscopically complex catalyst
active sites, proton transport in the electrolyte phase, layer. This DNS model is based on a numerical mesh
electron conduction in the electronic phase (i.e., created from either a digitized catalyst layer micro-
Pt/C), and oxygen diffusion through the gas phase, graph or a computer-generated random pore struc-
liquid water, and electrolyte phase. ture according to the prescribed porosity and average
In most of the macroscopic models reported in the pore size. Subsequently, direct simulation of reactant
literature the active catalyst layer was not the main and product transport with reaction is performed on
point of interest but rather treated either as an this digital microstructure at the pore level. Such a
infinitely thin interface or a macrohomogeneous DNS model can be used as an alternative to experi-
porous layer. There were a few detailed models mental trial-and-error for the optimization of com-
specifically developed for PEFC catalyst layers based positions and microstructure of a high-performance
on the theory of volume averaging. In this field catalyst layer. Figure 36 shows DNS predictions of
distinction is further made between a homogeneous the cathode voltage loss as a result of kinetic polar-
approach, a film model, and an agglomerate model. ization, ohmic polarization, and mass-transport po-
The homogeneous model assumes the catalyst layer larization for a commonly used catalyst layer com-
to be a two-phase system consisting of ionic and position (circles), an optimized design (squares), and
electronic phases only, without gas voids. The gas- a less optimal design (triangles). The three-dimen-
eous reactant transports through the catalyst layer sional contour of O2 concentration on the right
via the electrolyte phase as a dissolved species, and illustrates that the less optimized design exhibits
thus the diffusion rate is poor. In the film model gas restrictive oxygen transport into the catalyst layer
pores are assumed to exist along with the electronic due to a too small porosity.
particles covered by a thin film of polymer electrolyte.
On the other hand, the agglomerate model considers 3.10. Summary and Outlook
gas pores to surround agglomerates consisting of Table 2 summarizes and compares the main fea-
electrolyte films and electronic particles, i.e., a three- tures of representative CFD models reviewed in this
phase system. Depending on the pore geometry, section for PEFCs. It is evident that rapid advances
agglomerates are planar, cylindrical, and spherical. in PEFC modeling over the past 5 years have yielded
Nonetheless, all three models belong to the macro- considerable predictive capabilities. Fully three-
scopic theory for multiphase systems in which there dimensional, electrochemical-transport-thermal
is neither resolution to capture pore-level phenomena coupled, and realistic large-scale simulations are
nor ability to assess the morphological effects. possible with today’s computing power. A major focus
Springer and Gottesfeld,126 Perry et al.,127 and of future research in PEFC modeling will be on liquid
Eikerling and Kornyshev128 presented several ana- water transport and ensuing flooding phenomena.
lytical and numerical solutions for the cathode cata- Much remains to be learned about the fundamental
lyst layer under various conditions. Perry et al. physical process of flooding occurrence. Advanced
studied the effects of mass-transport limitations on diagnostics, such as visualization by optical diagnos-
the polarization characteristics of a reaction obeying tics and 3-D neutron tomography, as well as detailed
Figure 36. Cathode voltage loss as predicted by direct numerical simulation of proton, oxygen, and water transport in a
catalyst layer at the pore level (left), and three-dimensional oxygen concentration contours in a random microstructure of
the catalyst layer (right).131
Table 2. Comparisons of Representative CFD Models for Polymer Electrolyte Fuel Cells (PEFCs)
distribution measurements, are critically needed in aqueous methanol solution of low molarity acts as
order to paint a clear picture of how flooding occurs the reducing agent that traverses the anode flow
and what are the controlling factors. Once a physical field. Once inside the flow channel, the aqueous
understanding of the flooding process is established, solution diffuses through the backing layer, com-
the challenge will turn to how to model and simulate prised of carbon cloth or carbon paper. The backing
such a complex two-phase flow process with strong layer collects the current generated by the oxidation
interactions with the GDL surface and channel walls. of aqueous methanol and transports laterally to a
land in the current collector plate. The global oxida-
4. Direct Methanol Fuel Cells tion reaction occurring at the platinum-ruthenium
catalyst of the anode is given by
A DMFC is an electrochemical cell that generates
electricity based on the oxidation of methanol and CH3OH + H2O f CO2 + 6H+ + 6e-
reduction of oxygen. Figure 37 illustrates the cell
construction and operating principle of a DMFC. An The carbon dioxide generated from the oxidation
A number of extensive reviews have been published

as the DMFC research activities has grown nearly
exponentially worldwide in recent years. Gottesfeld
and Zawodzinski30 briefly summarized research at
Los Alamos intended for transportation application
and pointed out areas for improvement if DMFC
technology is to become a serious power plant can-
didate for transportation. Among them, reducing
catalyst loading to be competitive with reformed/air
fuel cells is perhaps a great challenge and a difficult
task for the foreseeable future. In a later book
chapter, Gottesfeld and Wilson142 discussed perspec-
tives of DMFC for portable applications. Lamy et
al.143 provided an in-depth review of DMFC funda-
mentals, including the reaction mechanisms of metha-
nol oxidation, use of various binary and ternary
electrocatalysts, effects of electrode structure and
composition on the activity of methanol oxidation,
and development of proton conducting membranes
with low methanol crossover. It was projected that
DMFC will be first commercialized as portable power
sources before the year 2010 and that there will be a
quantum jump on the technology to be in a position
to drive DMFC-powered vehicles 10 years thereafter.
Arico et al.144 reviewed recent advances in DMFC
from both fundamental and technological (for por-
table power) aspects. The fundamental aspects of
DMFC reviewed by Arico et al. were concerned with
electrocatalysis of methanol oxidation and oxygen
reduction in the presence of methanol crossed over
from the membrane, and the technological aspects
were focused on the proton conducting membranes
as well as techniques for fabricating MEA. Neergat
et al.145 provided an excellent review of new materials
Figure 37. Operating schematic of a direct methanol fuel
cell (DMFC).
for DMFC, including novel proton conducting mem-
branes and electrocatalysts. Narayanan et al.146 and
reaction emerges from the anode backing layer as Muller et al.147 discussed, in detail, the paramount
bubbles and is removed via the flowing aqueous importance of water balance for a portable DMFC
methanol solution. system.
Air is fed to the flow field on the cathode side. The
oxygen in the air combines with the electrons and 4.1. Technical Challenges
protons at the platinum catalyst sites to form water. As expected, a DMFC exhibits lower power densi-
The reduction reaction taking place on the cathode ties than that of a H2/air PEFC. At present, the H2/
is given by air PEFC requires anode and cathode platinum
loadings of less than 1 mg/cm2 to achieve power
3
/2O2 + 6H + + 6e- f 3H2O densities of 0.6-0.7 W/cm2. However, DMFC has the
advantages of easier fuel storage, no need for cooling
The two electrochemical reactions are combined to or humidification, and simpler design. Thus, DMFC
form an overall cell reaction as is presently considered a leading contender for por-
table power. To compete with lithium-ion batteries,
CH3OH + 3/2O2 f CO2 + 2H2O the first and foremost property of a portable DMFC
system is a higher energy density in Wh/L. This
Extensive work on DMFC has been conducted by requirement entails four key technical challenges to
many groups, notably Halpert et al.132 of Jet Propul- be overcome: (1) low rate of methanol oxidation
sion Laboratory (JPL) and Giner, Inc., Baldauf and kinetics on the anode, (2) methanol crossover through
Preidel133 of Siemens, Ren et al.134 of Los Alamos the polymer membrane, (3) innovative water man-
National Laboratory (LANL), Scott and co-work- agement, and (4) heat management.
ers135-137 of the University of Newcastle upon Tyne,
and Wang and co-workers111,138-141 of the Pennsyl-
4.1.1. Methanol Oxidation Kinetics
vania State University. A comparative study of Combined with methanol crossover, slow anode
DMFC with hydrogen PEFC was presented by the kinetics lead to a power density of a DMFC that is
LANL group.134,142 A DMFC requires platinum- three to four times lower than that of a hydrogen fuel
ruthenium and platinum loadings roughly five times cell. Much work has been focused on the anodic
higher to achieve power densities of 0.05-0.30 W/cm2. oxidation of methanol.148 The mechanism of the
electrocatalytic oxidation of methanol at anode was Developing novel membranes with low methanol
postulated.149,150 Different anode catalyst structures crossover would surely increase cell performance and
of Pt-Ru were developed,151 and several anode fuel efficiency.143,145 Alternatively, Wang and co-
catalysts other than Pt-Ru were explored.152-154 workers139,140,163 proposed modifying the anode back-
Additionally, the effects of the anode electrochemical ing structure to mitigate methanol crossover. It was
reaction on cell performance were experimentally demonstrated that a compact microporous layer can
studied.19,155,156 Lamy et al.143 and Arico et al.144 be added in the anode backing to create an additional
provided extensive reviews of the most recent work barrier to methanol transport, thereby reducing the
on electrocatalysis. More active catalysts for metha- rate of methanol crossing over the polymer mem-
nol oxidation enable a certain power density to be brane. Both practices to control methanol crossover
realized at higher cell voltage and hence directly by increasing mass-transfer resistance, either in the
impact the energy efficiency of the cell, which trans- anode backing or in the membrane, can be math-
lates to the energy density if the amount of fuel ematically formulated by a simple relation existing
carried by a DMFC system is fixed. between the crossover current, Ic, and anode mass-
transport limiting current, IA,lim. That is
4.1.2. Methanol Crossover
( )
Methanol crossover occurs due to the inability of I
the commonly used perfluorosulfonic acid (PFSA) Ic ) Ic,oc 1 - (26)
IA,lim
membrane to prevent methanol from permeating its
polymer structure. Diffusion and electro-osmotic drag
are the prime driving forces for methanol to transport where Ic,oc is the crossover current at open circuit and
through the polymer membrane and eventually react I the operating current. In the conventional approach
with platinum catalyst sites on the cathode, leading using thick membranes with low methanol crossover,
to a mixed potential on the cathode. This mixed Ic,oc is low and IA,lim is set to be high. In contrast,
potential on the cathode causes a decrease in cell setting up a barrier in the anode backing is es-
voltage. Methanol reaching the cathode also results sentially reducing the anode limiting current, IA,lim,
in decreased fuel efficiency, thus lowering the energy and making the open-circuit crossover current, Ic,oc,
density. a significant fraction of IA,lim, i.e., 50-80%. Two
Methanol crossover in DMFC has been extensively immediate advantages result from this latter cell
studied both experimentally and theoretically. Naray- design principle. One is that more concentrated fuel
anan et al.157 and Ren et al.158 measured the metha- can be used, thus leading to the much higher energy
nol crossover flux with different membrane thickness density of the DMFC system. Lu et al.140 successfully
and showed that the methanol crossover rate is demonstrated the use of 8 M methanol solution as
inversely proportional to membrane thickness at a the anode feed, and Pan163 most recently reported a
given cell current density, thus indicating that dif- DMFC operated with 10 M (or 30 vol %) methanol
fusion is dominant. In addition, Ren et al.159 com- fuel solution. Second, this type of DMFC permits use
pared diffusion with electro-osmotic drag processes of thin membranes such as Nafion 112, which greatly
and demonstrated the importance of electro-osmotic facilitates water back flow from the cathode to
drag in methanol transport through the membrane. anode,140,164 thus addressing another major challenge
In their analysis, methanol electro-osmotic drag is of portable DMFC to be discussed below.
considered a convection effect and the diluted metha-
nol moves with electro-osmotically dragged water
4.1.3. Water Management
molecules. Valdez and Narayanan160 studied the Water management emerges as a new significant
temperature effects on methanol crossover and showed challenge for portable DMFC systems. Constrained
that the methanol crossover rate increases with cell by the methanol crossover problem, the anode fuel
temperature. Ravikumar and Shukla156 operated the solution has been very dilute, meaning that a large
liquid-feed DMFC at an oxygen pressure of 4 bar and amount of water needs to be carried in the system,
found that cell performance is greatly affected by therefore reducing the energy content of the fuel
methanol crossover at methanol feed concentrations mixture. In addition, for each mole of methanol, 1
greater than 2 M and that this effect aggravates with mol of water is needed for methanol oxidation at the
the operating temperature. Wang et al.161 analyzed anode and 2.5 × 6 mol of water is dragged through a
the chemical compositions of the cathode effluent of thick membrane such as Nafion 117 toward the
a DMFC with a mass spectrometer. They found that cathode, assuming that the electro-osmotic drag
the methanol crossing over the membrane is com- coefficient of water is equal to 2.5 per proton. This
pletely oxidized to CO2 at the cathode in the presence then causes 16 water molecules lost from the anode
of Pt catalyst. Additionally, the cathode potential is for every mole of methanol. Water in the anode
influenced by the mixed potential phenomenon due therefore must be replenished. On the other hand,
to simultaneous methanol oxidation and oxygen inside the cathode, there are 15 water molecules
reduction as well as poisoning of Pt catalysts by transported from the anode and additionally 3 water
methanol oxidation intermediates. Kauranen and molecules produced by consuming six protons gener-
Skou162 presented a semiempirical model to describe ated from oxidation of one methanol. The presence
the methanol oxidation and oxygen reduction reac- of a large amount of water floods the cathode and
tions on the cathode and concluded that the oxygen reduces its performance. The difficult task of remov-
reduction current is reduced in the presence of ing water from the cathode to avoid severe flooding
methanol oxidation due to surface poisoning. and supplying water to the anode to make up water
loss due to electro-osmotic drag through the mem- ing of methanol, water, and heat-transport processes
brane is referred to as innovative water management occurring in a DMFC. This provides a great op-
in a portable DMFC. portunity to exercise fundamental modeling.
Traditionally, a high cathode gas flow rate (high Another good problem for modeling is the micro-
stoichiometry) is employed to prevent flooding. This DMFC system. Both anode carbon dioxide blockage
strategy not only increases parasitic power consump- and cathode flooding are especially acute in micro-
tion but also removes excessive water from the fuel systems due to the small channel length scale in-
cell, making external water recovery more difficult. volved, low operating temperature, dominance of
How to minimize water removal from the cathode surface tension forces, and requirement for low
and subsequent recovery externally to replenish the parasitic power losses in these systems.165-169
anode without causing severe cathode flooding be- In addition, DMFC is a system requiring a high
comes an important engineering issue. A greater degree of optimization. There are a multitude of
understanding and ability to tailor water flow in the operating parameters affecting the performance of a
cell is of fundamental interest for portable DMFC DMFC. These variables include cell temperature,
systems. This is an area where DMFC models play molarity of aqueous methanol solution, cathode pres-
an important role. sure, anode and cathode stoichiometry, and flow-field
design. Higher cell temperatures improve catalytic
4.1.4. Heat Management activity, but the water loss from the cathode in-
Thermal management in a DMFC is intimately creases. Efficient removal of carbon dioxide gas
tied to water and methanol-transport processes. bubbles and liquid water produced on the anode and
First, heat generation in a DMFC is much higher cathode, respectively, must be maintained to allow
than a H2/air PEFC due to a much lower energy reactants to reach catalyst sites. Removal of carbon
efficiency (it is only 20-25% when the cell is operated dioxide “slugs” and prevention of cathode “flooding”
between 0.3 and 0.4 V). That is, for a 20 W DMFC can be attained by increasing flow rates. However,
system, 60-80 W of waste heat is produced. The increasing flow rates requires more pumping power.
waste heat is typically removed from DMFC by liquid Too high a flow rate on the cathode will dry out the
fuel on the anode side and water evaporation on the polymer membrane, decreasing proton conductivity
cathode side. The latter also determines the amount and hence cell performance. An understanding of the
of water loss from a DMFC and the load of water interdependence of these parameters plays a key role
recovery by an external condenser. Therefore, while in optimizing the performance of a DMFC.
a higher cell temperature promotes the methanol DMFC modeling thus aims to provide a useful tool
oxidation reaction, it may not be practically feasible for the basic understanding of transport and electro-
from the standpoint of water evaporation loss. More- chemical phenomena in DMFC and for the optimiza-
over, the higher cell temperature increases the tion of cell design and operating conditions. This
methanol crossover rate, thereby reducing the fuel modeling is challenging in that it entails the two-
efficiency and system energy density. phase treatment for both anode and cathode and that
both the exact role of the surface treatment in
4.2. DMFC Modeling backing layers and the physical processes which
control liquid-phase transport are unknown.
4.2.1. Needs for Modeling
4.2.2. DMFC Models
While attempts continue to elucidate the funda-
mental electrochemical reaction mechanisms, explore In the literature, Scott et al.170-172 developed sev-
new compositions and structures of catalysts, and eral simplified single-phase models to study transport
develop new membranes and methods to prevent and electrochemical processes in DMFC. Baxter et
methanol crossover, important system issues relevant al.173 developed a one-dimensional mathematical
to DMFC are emerging, such as water management, model for a liquid-feed DMFC anode. A major as-
gas management, flow-field design and optimization, sumption of this study was that carbon dioxide is only
and cell up-scaling for different applications. A dissolved in the liquid, and hence, their anode model
number of physicochemical phenomena take place in is a single-phase model. Using a macrohomogeneous
a liquid-feed DMFC, including species, charge, and model to describe the reaction and transport in the
momentum transfer, multiple electrochemical reac- catalyst layer of a vapor-feed anode, Wang and
tions, and gas-liquid two-phase flow in both anode Savinell174 simulated the effects of the catalyst layer
and cathode. Carbon dioxide evolution in the liquid- structure on cell performance. Kulikovsky et al.175
feed anode results in strongly two-phase flow, making simulated a vapor-feed DMFC with a two-dimen-
the processes of reactant supply and product removal sional model and compared the detailed current
more complicated. All these processes are intimately density distributions in the backing, catalyst layer,
coupled, resulting in a need to search for optimal cell and membrane of a conventional to a new current
design and operating conditions. A good understand- collector. In another paper, Kulikovsky176 numerically
ing of these complex, interacting phenomena is thus studied a liquid-feed DMFC considering methanol
essential and can most likely be achieved through a transport through the liquid phase and in hydrophilic
combined mathematical modeling and detailed ex- pores of the anode backing. In both publications of
perimental approach. It is apparent that three of the Kulikovsky the important phenomenon of methanol
four technical challenges discussed in section 4.1 for crossover was ignored. Dohle et al.177 presented a one-
portable DMFC systems require a basic understand- dimensional model for the vapor-feed DMFC and
included the crossover phenomenon. The effects of processes. This understanding is key to successful
methanol concentration on the cell performance were design and operation of portable DMFC systems.
studied. Finally, the important role of a microporous layer in
In a three-part paper178-180 Meyers and Newman DMFC and its tailoring to control the flow of metha-
developed a theoretical framework that describes the nol and water remain unknown.
equilibrium of multicomponent species in the mem-
brane. The transport of species in the membrane 4.3. Experimental Diagnostics
based on concentrated-solution theory and membrane Similarly, experimental diagnostics are an impor-
swelling were taken into consideration in their model. tant component of advanced modeling and simulation
The transport phenomena in the porous electrodes of DMFCs. Diagnostic techniques for DMFCs have
were also included in their mathematical model. included the following: (1) cyclic voltammetry (CV)
However, the effect of pressure-driven flow was not to determine the electrochemically active area of the
considered. In addition, the transport of carbon cathode, (2) CO stripping to determine the electro-
dioxide out of the anode was neglected by assuming chemically active area of the anode, (3) electrochemi-
that the carbon dioxide is dilute enough to remain cal impedance spectroscopy (EIS), (4) anode polar-
fully dissolved in liquid. Nordlund and Lindbergh181 ization characterization via a MeOH/H2 cell as
studied the influence of the porous structure on the proposed by Ren et al.,19 (5) methanol crossover rate
anode with mathematic modeling and experimental measurement by CO2 sensing in the cathode (via
verification. In their model they also assumed that FITR, GC, or infrared CO2 sensors) or measuring the
carbon dioxide does not evolve as gas within the limiting current in a MeOH/N2 cell (Ren et al.19), (6)
electrode. Recently, Wang and Wang182 presented a current distribution measurements via a segmented
two-phase, multicomponent model. Capillary effects cell in conjunction with a multichannel potentiostat
in both anode and cathode backings were accounted (Mench and Wang111), (7) material balance analysis
for. In addition to the anode and cathode electro- of CH3OH and H2O (Narayanan et al.146 and Muller
chemical reactions, the model considered diffusion et al.147), and (8) two-phase visualization of bubble
and convection of both gas and liquid phases in dynamics185-187 on anode and liquid droplet dynamics
backing layers and flow channels. The model fully on cathode.187 The work in the last three areas is
accounted for the mixed potential effect of methanol briefly reviewed in this review as they are particu-
oxidation at the cathode as a result of methanol larly pertinent to the fundamental modeling of
crossover caused by diffusion, convection, and electro- DMFC for cell design and optimization.
osmosis. The model of Wang and Wang was solved
using a finite-volume technique and validated against Mench and Wang111 described an experimental
experimental polarization curves. The model results technique to measure current distribution in a 50 cm2
instrumented DMFC based on multichannel poten-
indicated the vital importance of gas-phase transport
tiostat. In this method separate current collector ribs
in the DMFC anode.
are embedded into an insulating substrate (e.g.,
Divisek et al.183 presented a similar two-phase, two- Lexan plate) to form a segmented flow-field plate.
dimensional model of DMFC. Two-phase flow and The resulting flow-field plates for both anode and
capillary effects in backing layers were considered cathode are then assembled with a regular MEA to
using a quantitatively different but qualitatively form a fuel cell with independently controllable
similar function of capillary pressure vs liquid satu- subcells. All subcells are connected to a multichannel
ration. In practice, this capillary pressure function potentiostat to undergo potentiostatic experiments
must be experimentally obtained for realistic DMFC simultaneously. The subcell currents measured thus
backing materials in a methanol solution. Note that provide information on the current density distribu-
methanol in the anode solution significantly alters tion for the full-scale fuel cell. The spatial and
the interfacial tension characteristics. In addition, temporal resolution of this method depends on the
Divisek et al.183 developed detailed, multistep reac- number of channels available and capabilities of the
tion models for both ORR and methanol oxidation as potentiostat. Current density distribution measure-
well as used the Stefan-Maxwell formulation for gas ments were made for a wide range of cathode flow
diffusion. Murgia et al.184 described a one-dimen- rates in order to elucidate the nature of cathode
sional, two-phase, multicomponent steady-state model flooding in a DMFC. Figure 38 displays the current
based on phenomenological transport equations for density distributions for high and low cathode air
the catalyst layer, diffusion layer, and polymer flow rates. In the case of high cathode stoichiometry
membrane for a liquid-feed DMFC. (Figure 38a), it can be seen that the current distrib-
Despite the fact that much effort has been made utes rather uniformly for all three levels of cell
to model the DMFC system, considerable work re- voltage. As expected, the current density increases
mains, particularly in support of the emerging por- as the cell voltage decreases. In the case of low
table designs and systems. Few have treated the cathode stoichiometry (still excessive for the oxygen
dominating effects of two-phase flow. No model to reduction reaction), Figure 38b clearly shows that a
date has sufficient detail to provide a microfluidic portion of the cathode toward the exit is fully flooded,
theory for portable systems including effects of chan- leading to almost zero current. The information
nel geometry and wettability characteristics of the provided in Figure 38 can be used to identify innova-
GDL on fluid flow in the anode or cathode. Modeling tive cathode flow-field designs and enables the de-
studies are needed to fully elucidate the intricate velopment of MEA structures with improved water
couplings of methanol, water, and heat-transport management capabilities.
Figure 38. Current density distributions in a 50 cm2 DMFC for (a) high cathode air flow rate (stoichiometry of 85 at 0.1
A/cm2) and (b) low cathode air flow rate (stoichiometry of 5 at 0.1 A/cm2).111
Material balance analysis proves to be a critical
diagnostic tool for the development of portable DMFC
systems. In this analysis methanol balance on the
anode side along with the methanol crossover rate
typically measured by an infrared CO2 sensor is
conducted. In addition, water balance on both anode
and cathode sides is performed in which the cathode
water amount is carefully collected by a moisture trap
and measured.146,147 From such analyses Müller et
al.147 revealed that the water balance on the DMFC
anode is highly negative, thus calling for membrane
development with low water crossover in addition to
low methanol crossover.
Gas management on the anode side is an important
issue in DMFC design. On the anode side, carbon
dioxide is produced as a result of methanol electro-
chemical oxidation. If CO2 bubbles cannot be removed
efficiently, the anode channels will be blocked, lead-
ing to limited mass transport. Argyropoulos et al.185,186
were perhaps the first to observe the two-phase flow
pattern in the anode channel under various operating
conditions. This flow visualization on the anode side
yields a valuable understanding of bubble dynamics
in DMFC. This study was, however, undertaken
under low cell performance. Most recently, Lu and Figure 39. Images of bubble dynamics in the DMFC
Wang187 developed a carefully designed transparent anode with carbon paper backing layer for 2 M MeOH feed
DMFC to visualize bubble dynamics on the anode and nonhumidified air at 100 mA/cm2 and 85 °C.187
side and liquid droplet (and flooding) dynamics on grow into larger slugs for detachment, clearly indica-
the cathode. Moreover, Lu and Wang explored the tive of the dominant effect of surface tension in
profound effect of backing layer wettability on gas bubble dynamics in DMFC. Once the bubbles grow
management in DMFC. Figure 39 shows a sequence to a sufficient size, they detach and sweep along the
of images taken at various times at a feed tempera- backing surface in the channel. This sweeping pro-
ture of 85 °C and current density of 100 mA/cm2. The cess clears all small bubbles preexisting on the
images, 1 s apart, were captured from a movie with backing surface, making new bubbles grow from the
a time resolution of 1/30 s. The time of the first image smallest size to the full detachment diameter. As a
was chosen arbitrarily due to the fact that two-phase result, the two-phase flow becomes regularly inter-
flow is a regular, periodic event. As shown in Figure mittent. The images shown in Figure 39, representa-
39, the CO2 bubbles emerge at certain locations and tive of most DMFC systems under commercial de-
form large and discrete gas slugs in the channel. The velopment, indicate the vital importance of considering
bubble motion is governed by the momentum of liquid two-phase flow and transport in a DMFC model.
flow, force of buoyancy on the bubble, and surface
tension between the bubble and substrate. It can be 4.4. Model Validation
seen from Figure 39 that the bubbles are held on the Experimental validation of the two-phase DMFC
carbon paper by strong surface tension until they model of Wang and Wang182 has been carried out for
Figure 40. Comparisons of 2-D model predictions with experimental data for a DMFC with (a) temperature effect and (b)
concentration effect.182
Figure 41. Comparison of localized polarization curves between experiments (a) and model predictions (b) for a 50 cm2
DMFC with an anode flow stoichiometry of 27 and a cathode air stoichiometry of 5 at 0.1 A/cm2.5
a 5 cm2 graphite cell. A brief description of the cell the polarization curves between numerical and ex-
geometry, MEA compositions, and operating condi- perimental results for different methanol feed con-
tions is given in Figure 40. Figure 40a illustrates the centrations. In accordance with these experiments,
capability of the model to predict the polarization the model prediction for the 2 M case shows a slightly
curves at two cell temperatures. Excellent agreement lower performance (due primarily to higher methanol
is achieved not only in the kinetic- and ohmic- crossover) and an extended limiting current density.
controlled regimes of the polarization curves but also Similar success in validating global I-V curves was
in the mass-transport-controlled regime, where the also reported by Murgia et al.,184 among others.
methanol oxidation kinetics is modeled as a zero- While the model validation against cell overall
order reaction for molar concentrations above 0.1 M performance data has been satisfactory and encour-
but a first-order reaction for a molarity below 0.1 M. aging, as evident from Figure 40, the ultimate test
This shift in the reaction order and molarity of of these highly sophisticated two-phase models is
transition is consistent with direct kinetics measure- comparison with detailed distribution measurements.
ments. A lower mass-transport-limiting current den- Figure 41 presents such an attempt toward develop-
sity at 50 °C, seen in Figure 40a, is caused by the ing high-fidelity first-principles models for DMFC.
lower diffusion coefficients in both the liquid and gas Figure 41a shows a set of localized polarization
phases and the lower saturation methanol vapor curves measured using the current distribution mea-
concentration in the gas phase at lower tempera- surement technique of Mench and Wang,111 and
tures. Using the same model and property data, Figure 41b displays the same set of polarization
Figure 40b shows equally satisfactory agreement in curves predicted from the DMFC two-phase model
Figure 42. Various planar SOFC configurations (revised from ref 189).
of Wang and Wang.182 A low air stoichiometry of 5 that the DMFC model development efforts will be
(although not low for the electrochemical reaction directed less toward refining model accuracy and
requirement) was deliberately chosen so that cathode improving computational speed and more toward
GDL flooding may occur and a nonuniform current applying the models to invent new cell designs and
density distribution results. pinpoint areas of improvement.
The two graphs in Figure 41 share a qualitative
similarity. For example, both experiment and model 5. Solid Oxide Fuel Cells
results indicate that the local polarization curves
near the dry air inlet exhibit a monotonic function A solid oxide fuel cell (SOFC) consists of two
between the voltage and current. Also, the shape of electrodes: anode and cathode, with a ceramic elec-
the polarization curves near the exit, from both trolyte between that transfers oxygen ions. A SOFC
experiment and simulation, is clearly evidence of typically operates at a temperature between 700 and
flooding in the cathode GDL. Another interesting 1000 °C, at which temperature the ceramic electro-
observation is that the average cell polarization lyte begins to exhibit sufficient ionic conductivity.
curves, measured and predicted, do not exhibit any This high operating temperature also accelerates
sign of cathode flooding, indicating that detailed electrochemical reactions; therefore, a SOFC does not
distribution measurements are absolutely required require precious metal catalysts to promote the
in order to discern complex physicochemical phenom- reactions. More abundant materials such as nickel
ena occurring inside the cell. Finally, it can be seen have sufficient catalytic activity to be used as SOFC
from Figure 41 that a satisfactory quantitative electrodes. In addition, the SOFC is more fuel-flexible
comparison between experiment and model is lacking than other types of fuel cells, and reforming of
on the detailed level. hydrocarbon fuels can be performed inside the cell.
Difficulties in obtaining good quantitative agree- This allows use of conventional hydrocarbon fuels in
ment between predicted and measured distribution a SOFC without an external reformer.
results are indicative that model refinements as well
Planar SOFC systems have received much atten-
as an improved property database will be needed
tion lately because of ease of manufacturing and
before accurate quantitative predictions of not only
higher performance as compared to tubular types.188
overall polarization curve but also detailed distribu-
Planar SOFCs are generally manufactured in three
tions within a DMFC may be obtained.
different configurations according to their operating
temperatures. For cells operating around 1000 °C,
4.5. Summary and Outlook the electrolyte-supported cell configuration is pre-
Two-phase modeling capabilities for DMFCs have ferred. In this design, anode and cathode are very
emerged, which unravel the importance of gas-phase thin (i.e., 50 µm) and the electrolyte thickness is
transport of methanol as compared to the liquid- generally larger than 100 µm. The electrolyte ionic
phase transport. In addition, much effort is being conductivity is a strong function of operating tem-
directed toward developing a coupled model for perature in SOFCs. For SOFCs operating at lower
methanol, water, and heat-transport processes oc- operating temperatures, the ionic conductivity is
curring simultaneously in a DMFC. Such models are lower and as such anode- or cathode-supported cell
extremely useful for the discovery of unique design configurations are preferred. In the electrode-sup-
and operation regimes of the DMFC system for ported cell configuration, the electrolyte is usually
portable application, where the high energy density very thin (i.e., 20 µm) and either anode or cathode is
entails using highly concentrated methanol (prefer- thick enough to serve as the supporting substrate for
ably pure methanol), maintaining healthy water cell fabrication. The thickness of the supporting
balance, controlling fuel crossover, and improving electrode varies between 0.3 and 1.5 mm, depending
high-voltage performance. The latter two factors will on the design. These three planar SOFC designs are
result in high efficiency of a DMFC. It is expected sketched in Figure 42.189
For optimal design and operation of a SOFC, a dimensional descriptions. Prinkey et al.196 proposed
fundamental and detailed understanding of transport a Fluent-based CFD model to describe reactant flow,
and electrochemical kinetics is indispensable. Efforts transport, and electrochemical reaction in a SOFC.
are presently underway to understand the mult- The charge transport was, however, simplified into
iphysics and obtain the optimal design for SOFCs. an algebraic equation accounting for activation po-
For these purposes, a CFCD model, similar to those larizations, electrolyte ohmic loss, and concentration
for PEFCs and DMFCs, becomes a valuable tool for polarizations. Moreover, the concentration polariza-
design and operation of SOFCs. tions were calculated empirically via a prescribed
limiting current. This modeling framework is incon-
5.1. SOFC Models sistent in that the limiting current physically is the
Unlike the comprehensive reviews provided in result of the mass-transport phenomena and thus
sections 3 and 4 for PEFC and DMFC, respectively, should be calculated from the solution of species
the present review on SOFC modeling will be brief equations instead of being prescribed independently.
and is not meant to be exhaustive. Instead, only Recknagle et al.,197 on the other hand, employed
major trends are indicated by way of representative STAR-CD code along with an electrochemistry mod-
studies. ule to simulate a SOFC with three flow configura-
For the purpose of modeling, consider a planar tions: cross-flow, co-flow, and counter-flow. The
SOFC divided into anode gas channel, anode gas effects of cell flow configurations on the distribution
diffusion electrode, anode interlayer (active elec- of temperature, current, and reactant species were
trode), electrolyte, cathode interlayer (active elec- investigated. It was found that for similar fuel
trode), cathode gas diffusion electrode, and cathode utilization and average cell temperature, the co-flow
gas channel. The electrochemical reactions occur in case had the most uniform temperature distribution
the active regions of the porous electrodes (i.e., and the smallest thermal gradients. This study,
interlayers). In an SOFC, oxidant reduction occurs however, treated the positive electrode-electrolyte-
in the active cathode. The oxygen ions are then negative electrode (PEN) as a single solid component.
transported through the electrolyte, after which Such an approximation is inappropriate to model the
oxidation of the fuel occurs in the active anode by latest generation of electrode-supported SOFC where
the following reactions. mass diffusion through a thick anode or cathode
Oxidation of fuel at the anode constitutes a major limitation. Most recently, Ack-
mann et al.198 performed a two-dimensional numer-
H2 + O2- f H2O + 2e- ical study of mass and heat transport in planar SOFC
with focus on elucidating the mass diffusion limita-
tion across the thickness of electrodes as well as in
Reduction of oxidant at the cathode
the in-plane direction between the channel and
interconnect rib areas. This model of Ackmann et al.
1
/2O2 + 2e- f O2- also included methane/steam reforming and water-
gas shift reaction in addition to the SOFC electro-
The anode is generally made of nickel/yttria-stabi- chemical kinetics.
lized zirconia Cermet. The cathode is an LSM layer
chemically expressed as La1-xSrxMnO3. The electro- A self-consistent SOFC model was developed based
lyte is an Y2O3-doped zirconia called YSZ. on the same single-domain modeling framework that
The activation overpotentials for both electrodes has been applied to PEFC and DMFC, as elaborated
are high; therefore, the electrochemical kinetics of the above. This model of Pasaogullari and Wang199 solves
both electrodes can be approximated by Tafel kinet- the continuum equations for conservation of mass,
ics. The concentration dependence of exchange cur- momentum, species, thermal energy, and electric
rent density was given by Costamagna and Honeg- charge along with electrochemical kinetics in anode
ger.190 The open-circuit potential of a SOFC is and cathode of a SOFC. The governing equations for
calculated via the Nernst equation.190 The conductiv- the SOFC model of Pasaogullari and Wang199 were
ity of the electrolyte, i.e., YSZ, is a strong function of exactly the same as those in Table 1 except that
temperature and increases with temperature. The water transport by either electro-osmotic drag or
temperature dependence of the electrolyte conductiv- diffusion through the electrolyte becomes irrelevant.
ity is expressed by the Arrhenius equation.190 The species transport equation was solved for three
Earlier modeling studies were aimed at predicting species: H2, O2, and H2O.199 However, the model is
the current and temperature distributions,188,190-192 applicable to multiple fuels such as H2 and CO given
as the nonuniform distributions contribute to stress that kinetic expressions for co-oxidation of H2 and
formation, a major technical challenge associated CO on Ni catalysts are provided. N2 was considered
with the SOFC system. Flow and multicomponent as an inert gas in this model.
transport were typically simplified in these models To illustrate typical simulation results from such
that focused on SOFC electrochemistry. Recently, multiphysics SOFC models, consider a co-flow and a
fundamental characteristics of flow and reaction in cross-flow electrolyte-supported cell. The cross-flow
SOFCs were analyzed using the method of matched geometry is of particular interest because of the
asymptotic expansions.193-195 complex transport phenomena offered in this cell
With today’s computing power and popular use of configuration, a configuration able to provide detailed
CFD codes, SOFC modeling is moving toward mul- understanding of mass-transfer limitations. Struc-
tiphysics, electrochemical-transport-coupled, and three- tured, orthogonal meshes were used for all compu-
Figure 43. Calculated current-voltage characteristics

and power density curve of electrolyte-supported co-flow
SOFC at an operating temperature of 1000 °C and anode
and cathode stoichiometry of 1.5 and 2.0 at 0.4 A/cm2,
respectively.199
Figure 45. Current distribution (A/m2) at a cell potential

of 0.4 V in the five-channel cross-flow electrolyte-supported
SOFC (A/m2) under anode and cathode stoichiometries of
1.5 and 2.0, respectively, and a cell temperature of 1000
°C.
hydrogen concentrations in the anode feed. Clearly,

this is an area requiring much attention in the future
of SOFC modeling, and model validation at a detailed
distribution level (e.g., against current and species
distribution data) is equally essential.
5.2. Summary and Outlook

A single-domain, self-consistent formulation has
been available to describe the electrochemistry, gas
dynamics, and multicomponent transport in SOFC.
Both Fluent and STAR-CD codes can be adapted for
implementation of SOFC models. In addition to
predicting the current-voltage characteristics, these
CFD models provide valuable insight into the reac-
tant and product distributions, current distribution,
Figure 44. Geometry and mesh of a five-channel cross-
flow electrolyte-supported SOFC.
and fuel utilization, making it possible to analyze
SOFC operation in detail. The latest SOFC models
tational meshing. It was found that a 80 × 80 × 35 also have the energy equation built in to predict the
mesh is sufficient for 3-D, five-channel geometry.199 temperature distribution. The temperature distribu-
It takes ∼300 iterations to converge a 3-D simulation tion can be coupled to stress analyses, thereby
of the five-channel cross-flow geometry (i.e., 224 000 providing a comprehensive computer-aided engineer-
cells). Figure 43 shows a sample calculation of the ing (CAE) tool for SOFC design and operation. The
polarization behavior of the co-flow cell. CFD models are easily extendable to include multiple
To understand the cross-flow-transport character- fuels such as H2 and CO. In such a case, co-oxidation
istics in a planar SOFC, consider a five-channel cross- of H2 and CO on Ni catalyst must be considered
flow-type electrolyte-supported SOFC shown in Fig- following a similar approach already developed for
ure 44 and the simulated current distribution in this PEFC in the presence of CO poisoning. Finally, The
cell, shown in Figure 45.199 It is clearly seen from CFD-based SOFC models can be combined with
Figure 45 that the regions facing the channels have internal reforming models on the anode. In such a
much higher current density. This is clearly seen at situation, a catalytic chemical reaction is needed as
the zones, where the corresponding anode and cath- well.
ode gas channels meet. In those regions both of the SOFC modeling is the simplest among all three
electrodes receive enough reactant, thereby produc- systems reviewed in this paper, since a SOFC does
ing higher current. This 3-D calculation explicitly not involve transport phenomena as complex as in a
illustrates the effect of flow-field design on cell PEFC and DMFC. In addition, SOFC modeling can
performance. draw largely upon tremendous existing experiences
Experimental validation of SOFC models has been from combustion and fuel-processing fields. Future
quite scarce. Khaleel and Selman200 presented a directions of SOFC modeling research are toward (1)
comparison of 1-D electrochemical model calculations coupled modeling of SOFC and fuel reforming
with experimental polarization curves for a range of (whether external or internal), (2) comprehensive
modeling of direct hydrocarbon DOFC systems where (4) Wang, C. Y. Int. J. Transport Phenom. 2001, 3, 131.
(5) Wang, C. Y. In Handbook of Fuel Cells; Vielstich, W., Gasteiger,
spray and evaporation of liquid fuels play important H. A., Lamm, A., Eds.; John Wiley and Sons Ltd.: England,
roles, (3) model validation against detailed distribu- 2003; Vol. 3, p 337.
tion data, and (4) direct numerical simulation of (6) Ju, H.; Wang, C. Y. J. Electrochem. Soc., in press.
(7) Bird, R. B.; Stewart, W. E.; Lightfoot, E. N. Transport Phenom-
SOFC electrodes, similar to what is described in ena, 2nd ed.; Wiley: New York, 2002.
section 3.9, to establish a quantitative relationship (8) Newman, J. Electrochemical Systems, 2nd ed.; Prentice Hall:
of electrode microstructure and compositions with cell Englewood Cliffs, NJ, 1991.
(9) Patankar, S. V. Numerical Heat Transfer and Fluid Flow;
performance. Hemisphere Publishing Corp.: New York, 1980.
(10) Oran, E. S.; Boris, J. P. Numerical Simulation of Reactive Flow;
Cambridge University Press: Cambridge, 2001.
6. Closing Remarks (11) Wu, J.; Srinivasan, V.; Xu, J.; Wang, C. Y. J. Electrochem. Soc.
2002, 149, A1342.
Automotive, stationary, and portable applications (12) Pisani, L.; Murgia, G.; Valentini, M.; D’Aguanno, B. J. Electro-
place an ever-increasing demand to develop advanced chem. Soc. 2002, 149, A31.
(13) Pisani, L.; Murgia, G.; Valentini, M.; D’Aguanno, B. J. Electro-
fuel cell technologies with high performance, low cost, chem. Soc. 2002, 149, A898.
and excellent durability. Computer-aided engineering (14) Meng, H.; Wang, C. Y. J. Electrochem. Soc. 2004, 151, A358.
is indispensable in this development process and can (15) Meng, H.; Wang, C. Y. Chem. Eng. Sci. 2004, 59, 3331.
(16) Springer, T. E.; Zawodzinski, T. A.; Gottesfeld, S. J. Electrochem.
dramatically reduce the expensive and time-consum- Soc. 1991, 138, 2334.
ing trial-and-error experimentation currently re- (17) Zawodzinski, T. A.; Derouin, C.; Radzinsksi, S.; Sherman, R. J.;
quired. The complexity of fuel cells requires many Smith, V. T.; Springer, T. E.; Gottesfeld, S. J. Electrochem. Soc.
1993, 140, 1041.
interacting physicochemical submodels in order for (18) Zawodzinski, T. A.; Springer, T. E.; Davey, J.; Jestel, R.; Lopez,
CFCD models to be successful. Therefore, a funda- C.; Valerio, J.; Gottesfeld, S. J. Electrochem. Soc. 1993, 140,
1981.
mental understanding of the electrochemical and (19) Ren, X.; Springer, T. E.; Gottesfeld, S. J. Electrochem. Soc. 2000,
transport processes in fuel cells continues to be 147, 92.
necessary. This can be best achieved through a (20) Walsby, N.; Hietala, S.; Maunu, S. L.; Sundhold, F.; Kallio, T.;
Sundholm, G. J. Appl. Polym. Sci. 2002, 86, 33.
combination of fundamental modeling and detailed (21) Feterko, P.; Hsing, I.-M. J. Electrochem. Soc. 1999, 146, 2049.
diagnostics, as shown in the present review. In (22) Hinatsu, J. T.; Mizuhata, M.; Takenaka, H. J. Electrochem. Soc.
addition, further developments in numerical algo- 1994, 141, 1493.
(23) Zawodzinski, T. A.; Davey, J.; Valerio, J.; Gottesfeld, S. Elec-
rithms will eventually allow the use of CFCD models trochim. Acta 1995, 40, 297.
in design, optimization, and control of fuel cell (24) Ren, X.; Gottesfeld, S. J. Electrochem. Soc. 2001, 148, A87.
systems, thereby significantly accelerating the com- (25) Ren, X.; Henderson, W.; Gottesfeld, S. J. Electrochem. Soc. 1997,
144, L267.
mercial deployment of fuel cell technologies. (26) Ise, M.; Kreuer, K. D.; Maier, J. Solid State Ionics 1999, 125,
Computational modeling of fuel cells is an impor- 213.
(27) Motupally, S.; Becker, A. J.; Weidner, J. W. J. Electrochem. Soc.
tant and physically rich subject. Despite emergence 2000, 147, 3171.
of significant experimental and modeling capabilities, (28) Parthasarathy, A.; Srinivasan, S.; Appleby, A. J. J. Electrochem.
much remains to be done before the knowledge of Soc. 1992, 140, 2178.
(29) Lee, S. J.; Mukerjee, J.; McBreen, Y.; Rho, W.; Kho, Y. T.; Lee,
multiscale phenomena occurring in fuel cells can be T. H. Electrochim. Acta 1998, 43, 3693.
directly utilized in cell design and product develop- (30) Gottesfeld, S.; Zawodzinski, T. A. In Advances in Electrochemical
ment. Further challenges and research directions Science and Engineering; Tobias, C., Ed.; Wiley and Sons: New
York, 1997; Vol. 5.
have been pointed out at the end of each section for (31) Gasteiger, H. A.; Gu, W.; Makharia, R.; Mathias, M. F.; Sompalli,
PEFC, DMFC, and SOFC. An upsurge of research B. In Handbook of Fuel Cells; Vielstich, W., Gasteiger, H. A.,
activities in this field worldwide is expected in years Lamm, A., Eds.; John Wiley and Sons Ltd.: England, 2003; Vol.
3, p 594.
to come. (32) Mathias, M. F.; Roth, J.; Flemming, J.; Lehnert, W. In Handbook
of Fuel Cells; Vielstich, W., Gasteiger, H. A., Lamm, A., Eds.;
John Wiley and Sons Ltd.: England, 2003; Vol. 3, p 517.
7. Acknowledgments (33) Lim, C.; Wang, C. Y. Electrochim. Acta 2004, 49, 4149.
(34) Chapman, S.; Cowling, T. G. Mathematical Theory of Non-
The author acknowledges the contributions of his Uniform Gases; Cambridge University Press: Cambridge, 1951.
present and former graduate students, research (35) Reid, R. C.: Prausnitz, J. M.; Poling, B. E. The Properties of
Gases and Liquids; McGraw-Hill: New York, 1987.
associates, and industrial/academic collaborators. (36) Singhal, S. C. MRS Bull. 2000, 16.
The writing of this paper was made possible, in part, (37) Minh, N. Q.; Takahashi, T. Science and Technology of Ceramic
through the support of the NSF under grant nos. Fuel Cells; Elsevier: Amsterdam, 1995.
(38) Yokokawa, H. Key Eng. Mater. 1998, 153, 37.
CTS-9733662 and DUE-9979579, DOE under coop- (39) Virkar, A. V. Introduction to Solid Oxide Fuel Cells (SOFC):
erative agreement no. DEFC26-01NT41098, DARPA Science and Technology; Lecture notes from SOFC Workshop,
National Energy Technology Laboratory (NETL), summer 2001.
under grant no. DAAH01-1-R001, Sandia National (40) Todd, B.; Young, J. B. J. Power Sources 2002, 110, 186.
Laboratories, Argonne National Laboratory, as well (41) Khee, R. J.; Rupley, F. M.; Miller, J. A. The Chemkin Thermo-
as Delphi, General Motors, Toyota, Honda, Nissan, dynamic Data Base; SAND87-8215B, UC-4; Sandia National
Laboratories, Albuquerque, 1987.
W. L. Gore and Associates, Air Products and Chemi- (42) Prater, K. B. J. Power Sources 1994, 51, 129.
cals, and ConocoPhillips. (43) Handbook of Fuel Cells; Vielstich, W., Gasteiger, H. A., Lamm,
A., Eds.; John Wiley and Sons Ltd.: England, 2003; Vol. 3, Part
3.
8. References (44) Bernardi, D. M.; Verbrugge, M. W. AIChE J. 1991, 37, 1151.
(45) Bernardi, D. M.; Verbrugge, M. W. J. Electrochem. Soc. 1992,
(1) Kreuer, K. D. J. Membr. Sci. 2001, 185, 29. 139, 2477.
(2) Paddison, S. J. In Handbook of Fuel Cells; Vielstich, W., (46) Springer, T. E.; Zawodzinski, T. A.; Gottesfeld, S. J. Electrochem.
Gasteiger, H. A., Lamm, A., Eds.; John Wiley and Sons Ltd.: Soc. 1991, 138, 2334.
England, 2003; Vol. 3, p 397. (47) Springer, T. E.; Wilson, M. S.; Gottesfeld, S. J. Electrochem. Soc.
(3) Koper, M. T. M. In Handbook of Fuel Cells; Vielstich, W., 1993, 140, 3513.
Gasteiger, H. A., Lamm, A., Eds.; John Wiley and Sons Ltd.: (48) Fuller, T. F.; Newman, J. J. Electrochem. Soc. 1993, 140, 1218.
England, 2003; Vol. 2, p 348. (49) Nguyen, T.; White, R. E. J. Electrochem. Soc. 1993, 140, 2178.
(50) Yi, J. S.; Nguyen, T. V. J. Electrochem. Soc. 1999, 146, 38. (105) Lim, C.; Wang, C. Y. Measurement of contact angles of liquid
(51) Yi, J. S.; Nguyen, T. V. J. Electrochem. Soc. 1998, 145, 1149. water in PEM fuel cell gas diffusion layer (GDL) by sessile drop
(52) Promislow, K.; Stockie, J. M. SIAM J. Appl. Math. 2001, 62, and capillary rise methods. Penn State University Electrochemi-
180. cal Engine Center (ECEC) Technical Report no. 2001-03; Penn
(53) Stockie, J. M.; Promislow, K.; Wetton, B. R. Int J. Numer. State University: State College, PA, 2001.
Methods Fluids 2003, 41, 577. (106) Brown, C. J.; Pletcher, D.; Walsh, F. C.; Hammond, J. K.;
(54) Gurau, V.; Liu, H.; Kakac, S. AIChE J. 1998, 44, 2410. Robinson, D. J. Appl. Electrochem. 1992, 22, 613.
(55) Um, S.; Wang, C. Y.; Chen, K. S. J. Electrochem. Soc. 2000, 147, (107) Cleghorn, S. J. C.; Derouin, C. R.; Wilson, M. S.; Gottesfeld S.
4485. J. Appl. Electrochem. 1998, 28, 663.
(56) Um, S.; Wang, C. Y. In Proceedings of the ASME Heat Transfer (108) Bender, G.; Wilson, M. S.; Zawodzinski, T. A. J. Power Sources
Division; Orlando, FL, 2000; Vol. 1, pp 19-25. 2003, 123, 163.
(57) Dutta, S.; Shimpalee, S.; Van Zee, J. W. J. Appl. Electrochem. (109) Wieser, Ch.; Helmbold, A.; Gülzow, E. J. Appl. Electrochem.
2000, 30, 135. 2000, 30, 803.
(58) Dutta, S.; Shimpalee, S.; Van Zee, J. W. Int. J. Heat Mass (110) Stumper, J.; Campell, S.; Wilkinson, D.; Johnson, M.; Davis, M.
Transfer 2001, 44, 2029. Electrochim. Acta 1998, 43, 3773.
(59) Um, S.; Wang, C. Y. J. Power Sources 2004, 125, 40. (111) Mench, M. M.; Wang, C. Y. J. Electrochem. Soc. 2003, 150, A79.
(60) Lee, W.-K.; Shimpalee, S.; Van Zee, J. W. J. Electrochem. Soc. (112) Mench, M. M.; Wang, C. Y.; Ishikawa, M. J. Electrochem. Soc.
2003, 150, A341. 2003, 150, A1052.
(61) Zhou, T.; Liu, H. Int. J. Transport Phenom. 2001, 3, 177. (113) Noponen, M.; Mennola, T.; Mikkola, M.; Hottinen, T.; Lund, P.
(62) Berning, T.; Lu, D. M.; Djilali, N. J. Power Sources 2002, 106, J. Power Sources 2002, 106, 304.
284. (114) Noponen, M.; Hottinen, T.; Mennola, T.; Mikkola, M.; Lund, P.
(63) Mazumder, S.; Cole, J. V. J. Electrochem. Soc. 2003, 150, A1503. J. Appl. Electrochem. 2002, 32, 1081.
(64) Ju, H.; Meng, H.; Wang, C. Y. Int. J. Heat Mass Transfer, (115) Brett, D.; Atkins, S.; Brandon, N. P.; Vesovic, V.; Vasileiadis,
submitted for publication. N.; Kucernaka, A. Electrochem. Solid-State Lett. 2003, 6, A63.
(65) Wang, Y.; Wang, C. Y. Int. J. Heat Mass Transfer, submitted (116) Ren, X.; Gottesfeld, S. J. Electrochem. Soc. 2001, 148, A87.
for publication. (117) Janssen, G. J.; Overvelde, M. L. J. Power Sources 2001, 101,
(66) Wang, Y.; Wang, C. Y. Electrochim. Acta, in press. 117.
(67) Pasaogullari, U.; Wang, C. Y. 203rd Electrochemical Society (118) Lee, W. K.; Van Zee, J. W.; Shimpalee, S.; Dutta, S. Proc. ASME
Meeting, Paris, May 2003; Abstract 1190. Also J. Electrochem. Heat Transfer Div. 1999, 1, 339.
Soc., in press. (119) Mench, M. M.; Dong, Q. L.; Wang, C. Y. J. Power Sources 2003,
(68) Wang, Z. H.; Wang, C. Y.; Chen, K. S. J. Power Sources 2001, 124, 90.
94, 40. (120) Yang, X. G.; Burke, N.; Wang, C. Y.; Tajiri, K.; Shinohara, K. J.
(69) He, W.; Yi, J. S.; Nguyen, T. V. AIChE J. 2000, 46, 2053. Electrochem. Soc., in press.
(70) Natarajan, D.; Nguyen, T. V. J. Electrochem. Soc. 2001, 148, (121) Mench, M. M.; Burford, D.; Davis, T. In Proceedings of the 2003
1324. International Mechanical Engineering Congress and Exposition;
(71) Natarajan, D.; Nguyen, T. V. J. Power Sources 2003, 115, 66. Washington, D.C., ASME, Nov 2003.
(72) You, L.; Liu, H. Int. J. Heat Mass Transfer 2002, 45, 2277. (122) Bellows, R. J.; Lin, M. Y.; Arif, M.; Thompson, A. K.; Jacobsob,
(73) Mazumder, S.; Cole, J. V. J. Electrochem. Soc. 2003, 150, A1510. D. J. Electrochem. Soc. 1999, 146, 1099.
(74) Berning, T.; Djilali, N. J. Electrochem. Soc. 2003, 150, A1598. (123) Geiger, A. B.; Tsukada, A.; Lehmann, E.; Vontobel, P.; Wokaun,
(75) Pasaogullari, U.; Wang, C. Y. Electrochim. Acta, in press. A.; Scherer, G. G. Fuel Cells 2002, 2, 92.
(76) Wang, C. Y.; Gu, W. B. J. Electrochem. Soc. 1998, 145, 3407.
(124) Tuber, K.; Pocza, D.; Hebling, C. J. Power Sources 2003, 124,
(77) Kaviany, M. Principles of Heat Transfer in Porous Media;
403.
Springer-Verlag: New York, 1991; p 598.
(125) Yang, X. G.; Zhang, F. Y.; Lubawy, A.; Wang, C. Y. Electrochem.
(78) Plumb, O. A. In Handbook of Porous Media; Vafai, K., Hadim,
Solid-State Lett., in press.
H. A., Eds.; Marcel Dekker: New York, 2000; p 754.
(79) Wang, C. Y.; Cheng, P. Adv. Heat Transfer 1997, 30, 93. (126) Springer, T. E.; Gottesfeld, S. In Modeling of Batteries and Fuel
(80) Kulikovsky, A. A. J. Electrochem. Soc. 2003, 150, A1432. Cells; White, R. E., Ed.; 1991; Electrochemical Society Proc. 91-
(81) Wang, Y.; Wang, C. Y. Comparing variable density with constant 10, p 197.
density models for polymer electrolyte fuel cells. ECEC Technical (127) Perry, M. L.; Newman, J.; Cairns, E. J. J. Electrochem. Soc. 1998,
Report 2003-02; Penn State University: University Park, PA, 145, 5.
2003. See also: Modeling polymer electrolyte fuel cells with large (128) Eikerling, M.; Kornyshev, A. A. J. Electroanal. Chem. 1998, 453,
density and velocity changes. J. Electrochem. Soc., in press. 89.
(82) Berger, C. Handbook of Fuel Cell Technology; Prentice-Hall: (129) Jaouen, F.; Lindbergh, G.; Sundholm, G. J. Electrochem. Soc.
New Jersey, 1968. 2002, 149, A448.
(83) Gu, W. B.; Wang, C. Y. J. Electrochem. Soc. 2000, 147, 2910. (130) Pisani, L.; Valentini, M.; Murgia, G. J. Electrochem. Soc. 2003,
(84) Okada, T.; Xie, G.; Gorseth, O.; Kjelstrup, S.; Nakamura, N.; 150, A1558.
Arimura, T. Electrochim. Acta 1998, 43, 3741. (131) Wang, G. Q.; Wang, C. Y. Presented at 204th Electrochemical
(85) Okada, T.; Xie, G.; Meeg, M. Electrochim. Acta 1998, 43, 2141. Society Meeting, Orlando, FL, Oct 2003.
(86) Hsing, I.-M.; Futerko, P. Chem. Eng. Sci. 2000, 55, 4209. (132) Halpert, G.; Narayanan, S. R.; Valdez, T.; Chun, W.; Frank, H.;
(87) Janssen, G. J. M. J. Electrochem. Soc. 2001, 148, A1313. Kindler, A.; Surampudi, S.; Kosek, J.; Cropley, C.; LaConti, A.
(88) Janssen, G. J. M.; Overvelde, M. L. J. J. Power Sources 2001, In Proceedings of the 32nd Intersociety Energy Conversion
101, 117. Engineering Conference; AIChE: New York, 1997; Vol. 2, p 774.
(89) Buchi, F.; Scherer, G. G. J. Electrochem. Soc. 2001, 148, A183. (133) Baldauf, M.; Preidel, W. J. Power Sources 1999, 84, 161.
(90) Prinkey, M. T.; Shahnam, M.; Rogers, W. A.; Gemmen, R. S. In (134) Ren, X.; Zelenay, P.; Thomas, S.; Davey, J.; Gottesfeld, S. J.
Proceedings of the 2002 Fuel Cell Seminar; Palm Springs, CA, Power Sources 2000, 86, 111.
2002; p 913. (135) Scott, K.; Taama, W. M.; Argyropoulos, P. Electrochim. Acta
(91) Um, S. Ph.D. Thesis, The Pennsylvania State University, 2002. 1999, 44, 3575.
(92) Wohr, M.; Bolwin, K.; Schnurnberger, W.; Fischer, M.; Neubrand, (136) Scott, K. In Handbook of Fuel Cells; Vielstich, W., Gasteiger,
W. Int. J. Hydrogen Energy 1998, 23, 213. H. A., Lamm, A., Eds.; John Wiley and Sons Ltd.: England,
(93) Rowe, A.; Li, X. J. Power Sources 2001, 102, 82. 2003; Vol. 1, p 70.
(94) Maggio, G.; Recupero, V.; Mantegazza, C. J. Power Sources 1996, (137) Murgia, G.; Pisani, L.; Shukla, A. K.; Scott, K. J. Electrochem.
62, 167. Soc. 2003, 150, A1231.
(95) Shimpalee, S.; Dutta, S. Numerical Heat Transfer A 2000, 38, (138) Mench, M.; Boslet, S.; Thynell, S.; Scott, J.; Wang, C. Y. In Direct
111. Methanol Fuel Cells; The Electrochemical Society Proceedings
(96) Costamagna, P. Chem. Eng. Sci. 2001, 56, 323. Series: Pennington, NJ, 2001.
(97) Meng, H.; Wang, C. Y. Multidimensional modeling of polymer (139) Lim, C.; Wang, C. Y. J. Power Sources 2003, 113, 145.
electrolyte fuel cells under current density boundary condition. (140) Lu, G. Q.; Wang, C. Y.; Yen, T. J.; Zhang, X. Electrochim. Acta
Submitted for publication. 2004, 49, 821.
(98) Ishikawa, M. Private communication, 2003. (141) Yen, T. J.; Fang, N.; Zhang, X.; Lu, G. Q.; Wang, C. Y. Appl.
(99) Qi, Z.; Kaufman, A. J. Power Sources 2002, 109, 469. Phys. Lett. 2003, 83, 4056.
(100) Ju, H.; Wang, C. Y.; Cleghorn, S. J.; Beuscher, U. J. Electrochem. (142) Gottesfeld, S.; Wilson, M. S. In Energy Storage Systems for
Soc., submitted for publication. Electronics Devices; Osaka, T., Datta, M., Eds.; Gordon and
(101) Baschuk, J. J.; Li, X. J. Power Sources 2000, 86, 181. Breach Science Publishers: Singapore, 2000; p 487.
(102) Weisbrod, K. R.; Grot, S. A.; Vanderborgh, N. E. Electrochem. (143) Lamy, C.; Leger, J.-M.; Srinivasan, S. In Modern Aspects of
Soc. Proc. 1995, 23, 153. Electrochemistry; Bockris, J. O’M, Conway, B. E., White, R. E.,
(103) Nam, J. H.; Kaviany, M. Int. J. Heat Mass Transfer 2003, 46, Eds.; Kluwer Academic/Plenum Publishers: New York, 2001; p
4595. 53.
(104) Pasaogullari, U.; Wang, C. Y. J. Electrochem. Soc. 2004, 151, (144) Arico, A. S.; Srinivasan, S.; Antonucci, V. Fuel Cells 2001, 1,
A399. 133.
(145) Neergat, M.; Friedrich, K. A.; Stimming, U. In Handbook of Fuel (171) Sundmacher, K.; Scott, K. Chem. Eng. Sci. 1999, 54, 2927.
Cells; Vielstich, W., Gasteiger, H. A., Lamm, A., Eds.; John Wiley (172) Argyropoulos, P.; Scott, K.; Taama, W. M. J. Appl. Electrochem.
and Sons Ltd.: England, 2003; Chapter 63, p 856. 2000, 30, 899.
(146) Narayanan, S. R.; Valdez, T. I.; Rohatgi, N. In Handbook of Fuel (173) Baxter, S. F.; Battaglia, V. S.; White, R. E. J. Electrochem. Soc.
Cells; Vielstich, W., Gasteiger, H. A., Lamm, A., Eds.; John Wiley 2000, 146, 437.
and Sons Ltd.: England, 2003; Chapter 65, p 894. (174) Wang, J.; Savinell, R. F. In Electrode Materials and Processes
(147) Müller, J.; Frank, G.; Colbow, K.; Wilkinson, D. In Handbook of for Energy Conversion and Storage; Srinivasan, S., Macdonald,
Fuel Cells; Vielstich, W., Gasteiger, H. A., Lamm, A., Eds.; John D. D., Khandkar, A. C., Eds.; The Electrochemical Society
Wiley and Sons Ltd.: England, 2003; Chapter 62, p 847. Proceedings Series: Pennington, NJ, 1994; PV 94-23, p 326.
(148) Wasmus, S.; Kuver, A. J. Electroanal. Chem. 1999, 461, 14. (175) Kulikovsky, A. A.; Divisek, J.; Kornyshev, A. A. J. Electrochem.
(149) Lin, W. F.; Wang, J. T.; Savinell, R. F. J. Electrochem. Soc. 1997, Soc. 2000, 147, 953.
144, 1917. (176) Kulikovsky, A. A. J. Appl. Electrochem. 2000, 30, 1005.
(150) Hamnett, A. Catal. Today 1997, 38, 445. (177) Dohle, H.; Divisek, J.; Jung, R. J. Power Sources 2000, 86, 469.
(151) Thomas, S. C.; Ren, X.; Gottesfeld, S. J. Electrochem. Soc. 1999, (178) Meyers, J. P.; Newman J. J. Electrochem. Soc. 2002, 149, A710.
146, 4354. (179) Meyers, J. P.; Newman J. J. Electrochem. Soc. 2002, 149, A718.
(152) Liu, L.; Pu, C.; Viswanathan, R.; Fan, Q.; Liu, R.; Smotkin, E. (180) Meyers, J. P., Newman J. J. Electrochem. Soc. 2002, 149, A729.
S. Electrochim. Acta 1998, 43, 3657. (181) Nordlund J.; Lindbergh G. J. Electrochem. Soc. 2002, 149, A1107.
(153) Hayden, E. Catal. Today 1997, 38, 473. (182) Wang, Z. H.; Wang, C. Y. J. Electrochem. Soc. 2003, 150, A508.
(154) Page, T.; Johnson, R.; Hormes, J.; Noding, S.; Rambabu, B. J. (183) Divisek J.; Fuhrmann, J.; Gartner, K.; Jung R. J. Electrochem.
Electroanal. Chem. 2000, 485, 34. Soc. 2003, 150, A811.
(155) Arico, S.; Creti, P.; Modica, E.; Monforte, G.; Baglio, V.; Anto- (184) Murgia, G.; Pisani, L.; Shukla, A. K.; Scott, K. J. Electrochem.
nucci, V. Electrochim. Acta 2000, 45, 4319. Soc. 2003, 150, A1231.
(156) Ravikumar, M. K.; Shukla, A. K. J. Electrochem. Soc. 1996, 143, (185) Apgyropoulos, R.; Scott, K.; Taama, W. M. Electrochim. Acta
2601. 1999, 44, 3575.
(157) Narayanan, S. R.; Frank, H.; Jeffries-Nakamura, B.; Smart, M.; (186) Apgyropoulos, R.; Scott, K.; Taama, W. M. J. Appl. Electrochem.
Chun, W.; Halpert, G.; Kosek, J.; Cropley, C. In Proton Conduct- 1999, 29, 661.
ing Membrane Fuel Cells I; Gottesfeld, S., Halpert, G., Landgrebe, (187) Lu, G. Q.; Wang, C. Y. J. Power Sources, 2004, 134, 33.
A., Eds.; The Electrochemical Society Proceedings Series: Pen- (188) Yakabe, H.; Ogiwara, T.; Hishinuma, M.; Yasuda, I. J. Power
nington, NJ, 1995; PV 95-23, p 278. Sources 2001, 102, 144.
(158) Ren, X.; Zawodzinski, T. A.; Uribe, F.; Dai, H.; Gottesfeld, S. In (189) Singhal, S. C. Presentation at INDO-US Fuel Cell Workshop;
Proton Conducting Membrane Fuel Cells I; Gottesfeld, S., Washington, D.C., 2002.
Halpert, G., Landgrebe, A., Eds.; The Electrochemical Society (190) Costamagna, P.; Honegger, K. J. Electrochem. Soc. 1998, 145,
Proceedings Series: Pennington, NJ, 1995; PV 95-23, p 284. 3995.
(159) Ren, X.; Springer, T. E.; Zawodzinski, T. A.; Gottesfeld, S. J. (191) Ahmed, S.; McPheeter, C.; Kumar, R. J. Electrochem. Soc. 1991,
Electrochem. Soc. 2000, 147, 466. 138, 2712.
(160) Valdez, T. I.; Narayanan, S. R. In Proton Conducting Membrane (192) Bessette, N. F., II; Wepfer, W. J.; Winnick, J. J. Electrochem.
Fuel Cells II; Gottesfeld, S., Fuller, T. F., Halpert, G., Eds.; The Soc. 1995, 142, 3792.
Electrochemical Society Proceedings Series: Pennington, NJ. (193) Billingham, J.; King, A. C.; Copcutt, R. C.; Kendall, K. SIAM J.
1998; PV 98-27, p 380. Appl. Math. 1999, 60, 574.
(161) Wang, J. T.; Wasmus, S.; Savinell, R. F. J. Electrochem. Soc. (194) Cooper, R. J.; Billingham, J.; King, A. C. J. Fluid Mech. 2000,
1996, 143, 1233. 411, 233.
(162) Kauranen, P. S.; Skou, E. J. Electroanal. Chem. 1996, 408, 189. (195) King, A. C.; Billingham, J.; Cooper, R. J. Combust. Theory Model.
(163) Pan, Y. Ph.D. Thesis, The Pennsylvania State University, 2004. 2001, 5, 639.
(164) Peled, E.; Blum, A.; Aharon, A.; Philosoph, M.; Lavi, Y. Elec- (196) Prinkey, M.; Gemmen, R. S.; Rogers, W. A. In Proceedings of
trochem. Solid-State Lett. 2003, 6, A268. IMECE 2001; ASME: New York, 2001; Vol. 369-4, p 291.
(165) Kelley, S. C.; Deluga, G. A.; Smyrl, W. H. AIChE J. 2002, 48, (197) Recknagle, K. P.; Williford, R. E.; Chick, L. A.; Rector, D. R.;
1071. Khaleel, M. A. J. Power Sources 2003, 113, 109.
(166) Mench, M. M.; Wang, Z. H.; Bhatia, K.; Wang, C. Y. In (198) Ackmann, T.; Haart, L. G. J. de; Lehnert, W.; Stolten, D. J.
Proceedings of IMECE2001, Vol. 3, ASME, New York, 2001. Electrochem. Soc. 2003, 150, A783.
(167) Pavio, J.; Bostaph, J.; Fisher, A.; Hallmark, J.; Mylan, B. J.; (199) Pasaogullari, U.; Wang, C. Y. Electrochem. Soc. Proc. 2003,
Xie, C. G. Adv. Microelectron. 2002, 29, 1. 2003-07, 1403.
(168) Dyer, C. K. J. Power Sources 2002, 106, 31. (200) Khaleel, M. A.; Selman, J. R. In High-Temperature Solid Oxide
(169) Lee, S. J.; Chang-Chien, A.; Cha, S. W.; O’Hayre, R.; Park, Y. Fuel Cells-Fundamentals, Design and Applications; Singhal, S.
I.; Saito, Y.; Prinz, F. B. J. Power Sources 2002, 112, 410. C., Kendall, K., Eds.; Elsevier Ltd.: Oxford, U.K., 2003; p 293.
(170) Scott, K.; Argyropoulos, P.; Sundmacher, K. J. Electroanal.
Chem. 1999, 477, 97. CR020718S
Chem. Rev. 2004, 104, 4637−4678 4637
Transport in Proton Conductors for Fuel-Cell Applications: Simulations,

Elementary Reactions, and Phenomenology
Klaus-Dieter Kreuer,*,‡ Stephen J. Paddison,§ Eckhard Spohr,# and Michael Schuster‡
Max-Planck-Institut für Festkörperforschung, Heisenbergstr.1, D-70569 Stuttgart, Germany, Department of Chemistry,
University of Alabama in Huntsville, Huntsville, Alabama 35899, and Forschungszentrum Jülich, D-52425 Jülich, Germany
Contents transport through such electrolytes must be fast and

highly selective; these two properties are frequently
1. Introduction 4637 at odds with each other. Proton-conducting materials
2. Theoretical Methodologies and Simulation Tools 4640 are used as the electrolyte for low- and intermediate-
2.1. Ab Initio Quantum Chemistry 4641 temperature fuel cells, which are currently attracting
2.2. Molecular Dynamics 4642 significant interest (i.e., polymer electrolyte mem-
2.2.1. Classical Molecular Dynamics and Monte 4643 brane or proton exchange membrane fuel cells
Carlo Simulations (PEMFCs)), direct methanol fuel cells (DMFCs),
2.2.2. Empirical Valence Bond Models 4644 phosphoric acid fuel cells (PAFCs), and alkaline fuel
2.2.3. Ab Initio Molecular Dynamics (AIMD) 4645 cells (AFCs)). However, these are not true “single ion
2.3. Poisson−Boltzmann Theory 4645 conductors”, in that the transport phenomena are
2.4. Nonequilibrium Statistical Mechanical Ion 4646 more complex, involving significant parasitic trans-
Transport Modeling port of other species. This review is concerned with
2.5. Dielectric Saturation 4647 the transport properties of presently available and
3. Transport Mechanisms 4648 emerging materials that have the potential to be used
as the electrolyte for fuel cells soon. Strategies for
3.1. Proton Conduction Mechanisms 4648
the development of novel materials with transport
3.1.1. Homogeneous Media 4648 properties approaching those of an “ideal separator”,
3.1.2. Heterogeneous Systems (Confinement 4655 especially in the intermediate temperature range, are
Effects) also discussed.
3.2. Mechanisms of Parasitic Transport 4661
The last comprehensive review covering proton
3.2.1. Solvated Acidic Polymers 4661 conductivity and proton conducting materials was
3.2.2. Oxides 4665 written by one of the authors (dating back to 1996);1
4. Phenomenology of Transport in 4666 since then, there have been several other review
Proton-Conducting Materials for Fuel-Cell articles of similar scope (e.g., see Colomban2). There
Applications are also many reviews available on separator materi-
4.1. Hydrated Acidic Polymers 4666 als used for fuel cells (see articles in refs 3 and 4 and
4.2. PBI−H3PO4 Adducts 4668 references therein, recent review-type articles,5-8 and
4.3. Heterocycle-Based Systems 4669 a literature survey9), which, more or less, address all
4.4. Oxides 4669 properties that are relevant for their functioning in
5. Recent Approaches toward New 4669 a fuel cell. The transport properties are usually
Proton-Conducting Materials for Fuel-Cell described in these articles; however, the treatments
Applications are frequently restricted to macroscopic approaches
6. Acknowledgement 4674 and handwaving arguments about the transport
7. References 4675 mechanisms. The purpose of the present review is
to combine a few recently published results in the
1. Introduction context of a discussion of transport phenomena in
proton-conducting separator materials, which have
The electrolyte is the heart of any fuel cell. Ideally, some relevance in fuel cell applications (for a more
this component effectively separates the anode and complete list of the comprehensive literature in the
cathode gases and/or liquids and mediates the elec- field, the interested reader is referred to the afore-
trochemical reaction occurring at the electrodes mentioned references).
through conducting a specific ion at very high rates Hence, the two main foci of this review are (i) the
during the operation of the fuel cell. In other words, current understanding of the underlying elementary
processes and (ii) a phenomenological description of
* Author to whom correspondence should be addressed. E-mail the resulting macroscopic transport phenomena.
address: kreuer@fkf.mpg.de. Because the first aspect comprises proton conduction
‡ Max-Planck-Institut für Festkörperforschung.
§ University of Alabama in Huntsville. mechanisms other than the mechanisms of “parasitic
# Forschungszentrum Jülich. transport”, this review may also be considered an
10.1021/cr020715f CCC: $48.50 © 2004 American Chemical Society
4638 Chemical Reviews, 2004, Vol. 104, No. 10 Kreuer et al.
Klaus-Dieter Kreuer completed his Diploma Thesis on the "Distribution of Stephen J. Paddison received a B.Sc.(Hon.) in Chemical Physics and a
F- and OH- in Hydroxofluoroapatites" in Mineralogy at the University of Ph.D. (1996) in Physical/Theoretical Chemistry from the University of
Cologne under the supervision of F. Freund in 1979. He received his Calgary, Canada. He was, subsequently, a postdoctoral fellow and staff
Ph.D. from the University of Stuttgart (Department of Chemistry) in 1982. member in the Materials Science Division at Los Alamos National
His thesis project in the department of A. Rabenau (Max-Planck-Institute Laboratory, where he conducted both experimental and theoretical
for Solid State Research) involved the investigation of inorganic proton- investigations of sulfonic acid polymer electrolyte membranes. This work
conducting compounds (e.g., layered minerals, zeolites), which showed was continued while he was part of Motorola‘s Computational Materials
the correlation of proton mobility and molecular diffusion (vehicle Group in Los Alamos. He is currently an Assistant Professor in the
mechanism) for several cases. As a fellow of the “Studienstiftung des Chemistry and Materials Science Departments at the University of Alabama
Deutschen Volkes”, he benefited from the opportunity of a research stay in Huntsville, AL. Research interests continue to be in the development
at the California Institute of Technology (with the R. Vaughan group, 1978), and application of first-principles and statistical mechanical methods in
and a Max-Planck award (Otto Hahn Medaille) allowed him to join the understanding the molecular mechanisms of proton transport in fuel-cell
Massachussetts Institute of Technology (Department of Materials Science) materials.
as a visiting scientist (1983−1984). This period was part of his postdoctoral
activities at the Max-Planck-Institute for Solid State Research, which were
mainly concerned with research on alkaline-ion-conducting framework
structures (such as lithium borate glasses and NASICON). From 1986 to
1992, he built up a research group for the development of chemical
sensors (pH, O2) within a Swiss-German company (Endress & Hauser).
In 1992, he again joined the Max-Planck-Institute for Solid State Research,
where he assisted J. Maier in building up his new department. Since
1990, he has been lecturing at the University of Stuttgart, from which he
received his Habilitation Degree in 1999. His research interests are
centered around the investigation of the formation and mobility of hydrogen-
derived defects in condensed matter. His work comprises materials
preparation (inorganic and organic synthesis) and characterization (e.g.,
by nuclear magnetic resonance (NMR) and simulation techniques). The
focus of his research is both a better understanding of fundamental issues
and the development of novel materials for specific applications.
update on the corresponding part of ref 1 mentioned

Eckhard Spohr received his diploma in Chemistry from the Johannes-
previously. The initial application of various simula- Gutenberg-Universität in Mainz, Germany, in 1983. In 1986, he obtained
tion techniques to the investigation of fast elemen- his Ph.D. for computer simulation of water/metal interfaces at the Max-
tary reactions as being part of proton conduction Planck-Institut for Chemistry in Mainz. He continued to work for two years
mechanisms had already summarized in this article. as a research associate in the aqueous simulation group of Karl Heinzinger,
The early molecular dynamics simulation work of followed by a two-year visit to the University of California as a Feodor
Münch et al.10 on the proton conduction mechanism Lynen postdoctoral fellow of the Alexander von Humboldt Foundation,
in CsHSO4 diverse perovskite-type oxides and het- where he studied gas/surface reactions. In 1991, he moved to the
Universität Ulm, where he obtained his habilitation degree in 1995 and
erocycles (see Section 3.1.1) provided important worked as a University docent until 2000. In 2000, he moved to the
insight into the conduction mechanism of model Forschungszentrum Jülich, where he currently heads the physical chemistry
compounds, and some of the recent understanding group in the Institute for Materials and Processes in Energy Systems
of proton dynamics in water stems from simulation performing research on low-temperature fuel cells. His research interests
work (see Section 3.1.1.1). To date, simulation tech- center around computer simulations of aqueous electrolytes in confined
niques, when appropriately combined with experi- environments. He teaches courses in theoretical chemistry at the
Universität Ulm.
mental results, establish insight into the path for the
development of improved and new proton conductors. tion is truly cross-disciplinary in nature, because
The increased availability of computational hard- results from outside the fuel cell community are
ware, along with the improvement and development included. This is particularly true for the discussion
of new codes, have led to a tremendous dissemination of the proton mobility in homogeneous media (i.e.,
and increased impact of such techniques; hence, we Section 3.1.1: water, aqueous solutions, phosphoric
have included a separate section that addresses acid, imidazole, and simple cubic perovskites), which,
current “simulation techniques” (Section 2), the in many cases, only constitute a single component of
results of which are included into the discussion of the heterogeneous fuel cell separator material. As a
“transport mechanisms” (Section 3). The latter sec- consequence of confinement and interaction of the
Proton Conductors for Fuel-Cell Applications Chemical Reviews, 2004, Vol. 104, No. 10 4639
steady-state behavior of known materials under the

varying conditions of an operating fuel cell. Access
to the numerous references providing transport data
of proton-conducting fuel cell separators may be
found in the corresponding chapters of refs 3 and 4.
The referencing of this review is restricted to those
papers that are relevant for the more-general picture.
Of course, this also comprises extensive cross-
referencing to the non-fuel-cell literature.
Before proceeding to this, let us briefly mention a
few general directions in recent fuel cell research for
which the transport properties of the proton-conduct-
ing electrolyte are essential. Much of the current
research is on proton exchange membrane (PEM) fuel
Michael Schuster studied chemistry at the Ludwig-Maximilians-Universität cells where the electrolyte is DuPont’s Nafion or some
in Munich, Germany. He received his Ph.D. in Inorganic Chemistry under other sulfonated polymer.5 Such polymers naturally
the direction of A. Schmidpeter and K. Karaghiosoff for his work on combine, in one macromolecule, the high hydropho-
phosphates PmXnz- (X ) S, Se, Te) and their NMR spectroscopic bicity of the backbone with the high hydrophilicity
characterization. To continue his research interests in NMR spectroscopy, of the sulfonic acid functional group, which results
he moved to the Max-Planck-Institute for Solid State Research in Stuttgart in a constrained hydrophobic/hydrophilic nanosepa-
in 2000, where he is working on the development of proton-conducting
materials for fuel cells. His studies are focused on the synthesis of ration. The sulfonic acid functional groups aggregate
imidazole-, sulfonic acid-, and phosphonic acid-based oligomers and to form a hydrophilic domain that is hydrated upon
polymers and their structural and dynamical properties. Characterization absorption of water. It is within this continuous
methods particularly include pulsed-field gradient and electrophoretic NMR domain that ionic conductivity occurs: protons dis-
spectroscopy. sociate from their anionic counterion (-SO3-) and
mobile species, e.g., with the corresponding polymer become solvated and mobilized by the hydration
matrixes, the transport properties of these hetero- water. Water typically must be supplied to the
geneous systems are more complex and, therefore, electrolyte through humidification of the feed gases
are discussed in the section that follows (i.e., Section and is also produced by the electrochemical reduction
3.1.2). Because most of this work has been done on of oxygen at the cathode. This is the reason for two
prototypical materials, this discussion mainly in- serious problems relevant for the use of such mem-
volves readily available and extensively tested ma- branes in fuel cells: (i) because high proton conduc-
terials. However, the property relations may be easily tivity is only obtained at high levels of hydration, the
transferred to the transport properties of related maximum operation temperature is approximately
materials (Section 4) and may even provide a basis limited to the condensation point of water (i.e., 100
for the understanding of the transport properties of °C for a water pressure of 1 atm (105 Pa)); and (ii)
conceptually new proton-conducting separator ma- any protonic current also leads to transport of water
terials (Section 5). through the membrane (as a result of electroosmotic
Apart from mechanistic aspects, we have also drag) and, if methanol dissolves in the membrane,
summarized the macroscopic transport behavior of this is transported virtually at the same rate. The
some well-studied materials in a way that may limited operating temperature and the acidity of the
contribute to a clearer view on the relevant transport electrolyte makes it necessary to use platinum or
coefficients and driving forces that govern the be- platinum alloys (the most active but also the most
havior of such electrolytes under fuel cell operating expensive electrocatalyst) to promote the electro-
conditions (Section 4). This also comprises precise chemical reactions in the anode and cathode struc-
definitions of the different transport coefficients and tures. However, even with platinum, only rather pure
the experimental techniques implemented in their hydrogen can be oxidized at sufficient rates. At the
determination; providing a physicochemical rational operation temperature of state-of-the-art PEM fuel
behind vague terms such as “cross-over”, which are cells, the rate of direct oxidation of methanol (which
frequently used by engineers in the fuel cell com- is frequently considered an environmentally friendly
munity. Again, most of the data presented in this fuel) is not sufficiently high, and, hence, even trace
section is for the prototypical materials; however, amounts of CO present in any hydrogen-rich refor-
trends for other types of materials are also presented. mate (e.g., produced by steam reforming of methanol
Thus, the aim of this review is to give the currently or methane) poison platinum-based catalysts through
best possible generally applicable understanding of adsorption thus blocking the reaction sites. The
mass and charge transport in fuel-cell-relevant pro- humidification requirements, along with the high
ton conductors, including a description of the funda- electro-osmotic drag of water and methanol in con-
mental underlying elementary reactions for these ventional membranes, complicate the water and heat
transport processes and quantitative phenomenologi- management of the fuel cell and lead to a significant
cal descriptions of transport in these materials. The chemical short-circuiting, i.e., parasitic chemical
first is important for further adaptation of available oxidation of methanol at the cathode.
electrolytes to specific fuel cell conditions or, subse- Therefore, tremendous engineering effort has been
quently, the development of new materials. The latter expended to at least control the fluxes of water and
allows a better understanding of the transient and methanol in such a way that the resulting transient
and steady-state concentration profiles of these spe- which allows “surprises” to be witnessed. Thus, it is
cies across the membrane still permit acceptable a valuable adjunct to experimental work becoming
function of the membrane and the electrode struc- another powerful tool for the chemist, molecular
tures. The fact that measured water concentration physicist, and material scientist.
profiles11 differed substantially from the results of Fundamental information concerning local struc-
early membrane modeling12,13 made it clear that the ture, molecular hydrophilicity, aggregation of ionic
transport behavior of such membranes were not fully groups through the formation of hydrogen bonds, and
understood. It was the availability of better data and proton dissociation of acidic groups in the fragments
improved phenomenological models concerning the of PEMs has been obtained through the determina-
macroscopic transport in such membranes14 that tion of global minimum (i.e., equilibrium) energy
helped to better describe the behavior of available structures.15-25 Because these techniques explicitly
membranes under fuel cell conditions as a function
treat all the electrons of the system to obtain the
of the boundary conditions (i.e., current, gas humidi-
electronic structure, only fragments of polymer have
fication, etc.) and membrane properties (e.g., thick-
been studied, where the system size has typically
ness). In addition, there have been many attempts
been limited to <100 atoms.
to modify the transport properties, especially by
forming composites with highly dispersed inorganic The dynamics of particles (i.e., atoms, molecules,
phases. More radical approaches aim at conceptually ions, etc.) may be investigated through the continu-
different separator material exclusively transporting ous (i.e., time-evolving) solution of Newton’s equa-
protons and being able to operate at higher temper- tions (i.e., classical molecular dynamics (MD)) or with
atures in a low humidity environment. Such a “dream stochastic methods including Monte Carlo (MC)
membrane” is widely considered to be the key to theory. Classical MD simulations with empirical
further progress in “low”-temperature fuel cell tech- potentials can handle systems consisting of thou-
nology: higher operation temperatures would super- sands of particles over time periods of nanoseconds
sede or simplify gas conditioning and purification, and have been used to study proton transport in
reduce the precious metal loading (along with the materials as a function of parameters such as tem-
problems of heat rejection), while reducing hydration perature, water content, and, in polymers, equivalent
requirements may avoid the need to pressurize the weight and chemical and physical characteristics of
system. Because these aspects are all related to the main and side chains.26-29 To address continuous
cost of the fuel cell system, which is increasingly changes in valence bond networks during the trans-
attracting attention, understanding the transport port of protons, empirical valence bond (EVB) inter-
properties as a function of molecular structure and action potentials have been devised.30-32 Ab initio
morphology and development of new separator ma- molecular dynamics (AIMD) simulations have suc-
terials are the focal points in current fuel cell cessfully been used to study the proton dynamics in
research. homogeneous systems (e.g., water33-36 and imida-
Whether future progress will be achieved “within zole37), and recently even more complex and, there-
the box” by modifying available materials or by fore, extended “model” systems have been examined
designing conceptually different materials is not yet (e.g., trifluoromethanesulfonic acid monohydrate38,39).
clear. In any case, a better understanding of the They are computationally very demanding if “mean-
mechanisms governing transport of the various spe- ingful” trajectories of only tens of picoseconds are to
cies in the separator materials is useful in the effort be sampled. The payoff in such calculations is the
of further materials research and development. determination of completely new and potentially
significant insight into molecular mechanisms. The
2. Theoretical Methodologies and Simulation recent AIMD study of trifluoromethanesulfonic acid
Tools monohydrate solid38,39 revealed the very first evidence
of the possible role of the sulfonate anions and Zundel
This section briefly describes some of the theoreti-
ion in proton transfer in minimally hydrated PEMs.
cal methods and types of simulations that have
recently been applied to understand the structural Microscopic electrolyte theory (Section 2.4) based
and dynamical features of transport in proton con- on a statistical mechanical framework has recently
ductors. Although the transport properties and, been applied to study the diffusion of protons in a
hence, mechanisms are strongly correlated to the single hydrated channel or pore in several PEMs.40-46
morphology of the material, theoretical studies of the This approach makes several assumptions concerning
morphology will not be discussed here. the pore geometry, distribution of fixed anionic
Quantum chemistry or molecular electronic struc- groups, and, primarily, the vehicular mechanism of
ture theory is the application of the principles of proton transport through the center of the pore.
quantum mechanics to calculate the stationary states However, this model has been able to compute the
of molecules and the transitions between these states. proton self-diffusion coefficient in both Nafion and
Today, both computational and experimental groups S-PEEK (sulfonated poly(arylene ether ether ketone))
routinely use ab initio (meaning “from first prin- membranes correctly over a significant range of
ciples”) molecular orbital calculations as a means of hydration without resorting to any fitting param-
understanding structure, bonding, reaction paths eters, requiring only membrane-specific morphology
between intermediates etc. Explicit treatment of the information from small-angle X-ray scattering (SAXS)
electrons means that, in principle, one does not make experiments and structural information from elec-
assumptions concerning the bonding of a system, tronic structure calculations.
Confinement of water into regions with dimensions molecular orbitals (MOs), each dependent on only
of only a few nanometers, such as typically those three variables. Each MO describes the probability
found in PEMs, accompanied by a strong electrostatic distribution of a single electron moving in the average
field due to the anions, will result in a significantly field of all other electrons. Because of the require-
lower dielectric constant for the water than that ments of antisymmetry, with respect to the inter-
observed in bulk water. Measurement of this struc- change of any two electrons, and indistinguishability
tural ordering of the water has not been accomplished of electrons, the trial wave function is a single
experimentally to date, and this was the motivation determinantal wave function of the MOs. The opti-
to the recent calculation of the dielectric saturation mum MOs are determined through variationally
of the water in PEMs with an equilibrium thermo- minimizing E(R). Thus, the HF method is a mean-
dynamical formulation.47-51 In addition to informa- field method applied to the many-electron problem,
tion concerning the state of the water this modeling the solution of which is commonly referred to as the
has provided information concerning the distribution self-consistent field (SCF). Because HF theory does
of the dissociation protons in sulfonic acid-based not include details of the instantaneous electron-
PEMs. electron correlations, several post-HF techniques
The various methodologies are discussed in some- have been developed, including many-body (or Møller-
what more detail in the following sections. Plesset) perturbation theory,58 and configuration
interaction (CISD).59 In the former method, a “zero-
2.1. Ab Initio Quantum Chemistry order” description of the ground-state wave function
is assumed to be a determinantal wave function
Only a very concise description of quantum chem-
constructed from the HF MOs upon which a pertur-
istry is presented in this section; the reader is
bation acts that the difference between the sum of
referred to standard texts for a more complete
Fock operators and the exact Hamiltonian. In CISD,
treatment.52-54 The aim of quantum chemistry or ab
electron-pair correlations are treated self-consistently
initio electronic structure theory is the solution of the
through the assumption that the wave function is a
time-independent Schrödinger equation:
linear combination of the HF determinant with all
determinants formed by single and double orbital
H(r;R)Ψ(r;R) ) E(R)Ψ(r;R) (1)
substitutions of coefficients determined variationally.
This yields the molecular wave function, Ψ(r;R), A second approximation that is commonly invoked
which is dependent explicitly on the 3n coordinates in all these methods is the expansion of the unknown
of all n electrons (denoted collectively as r) and MOs in terms of a given, fixed, and finite set of
implicitly on the coordinates of all A nuclei (denoted functions. These functions are usually referenced as
collectively as R); and the (total) molecular energy, the atomic orbital (AO) basis set, where the atomic
E(R), which parametrically is dependent only on the orbitals are usually expanded through linear combi-
nuclear positions. The distinction in functional de- nations of Gaussians, because of the fact that all
pendence of these quantities is due to the Born- required matrix elements may be evaluated analyti-
Oppenheimer (BO) approximation,55 which separates cally.60 The basis sets may be “minimal” (i.e., one
(slow) nuclear motion from (fast) electronic motion. basis function per atomic orbital), “split valence” (two
It is through determination of E(R) that the potential basis functions per valence atomic orbital, one per
energy surface is defined. The total energy or molec- core orbital), or higher zeta (where n-zeta means n
ular electronic Hamiltonian, H(r;R), is an operator basis functions per orbital). Multiple basis functions
consisting of the kinetic energy of all the electrons per atomic orbital allow the size of the orbitals to
and the potential energy due to the following forces: increase (i.e., along a bond axis) or decrease (i.e.,
electron-nuclear attraction, electron-electron repul- perpendicular to a bond axis). The higher-order basis
sion, and nuclear-nuclear repulsion. Its explicit form sets may also be augmented with polarization func-
in atomic units is tions (e.g., one or more sets of d functions on first-
( )
row atoms) that describe small displacements of the
1 n
∂2 ∂2 ∂2 orbitals from atomic centers in the molecular envi-
H(r;R) ) -
2
∑ + + - ronment, and diffuse functions for anions and Ryd-
i ∂x2i ∂y2i ∂z2i berg excited states. The size of the Hamiltonian to
be diagonalized is proportional to N4 (where N is the
n A ZR 1 n n
1 1 A A ZRZβ
∑∑|r - RR|
+
2
∑∑|r - rj|
+
2
∑∑|R - Rβ|
number of AOs); therefore, the use of large basis sets
is very time-consuming.
i R i i j i R β R
(2) The majority of the molecular-scale information
concerning the effects of structure and local chemis-
This partial differential equation with 3n unknowns try on proton dissociation and separation in PEM
is impossible to solve exactly (beyond the hydrogen fragments alluded to previously15-23 were initially
molecule) and, thus, various approximations are determined using HF theory and split valence local
generally made. basis sets. Refinements to the equilibrium configura-
In a first class of approximations, the solution is tions were made using both Møller-Plesset (MP)
sought of a simpler set of equations rather than the perturbation schemes and hybrid density functional
exact equations. Under the Hartree-Fock (i.e., HF) theory (described below).
approximation,56,57 the function of 3n variables is A widely used alternative to the electron correla-
reduced to n functions, which are referenced as tion treatments mentioned previously is the density
functional theory (DFT),61 which seeks to determine ground-state energy of the molecular system. These
the exact ground state energy and electron density equations differ from the Hartree equations only by
directly, without computing a many-electron wave the inclusion of the exchange-correlation potential
function. The electron density is only a function of vxc(r). Furthermore, if the exchange correlation en-
three variables (unlike the wave function, which is ergy is partitioned into separate contributions due
dependent on 3n variables); thus, DFT dramatically to correlation and exchange, i.e., Exc[n(r)] ) Ex + Ec,
simplifies the calculations and, therefore, has become and only the correlation energy neglected, then the
the preferred method for treating large molecules. treatment becomes a variant of HF theory. The
DFT relies on two fundamental theorems.62 The first simplest approximation for Exc[n(r)] is the generally
states that the ground-state electron density, n(r), accepted local density approximation (LDA):63
uniquely determines the external potential v(r) and,
xc [n(r)] ≡ ∫xc(n(r))n(r) dr
ELDA
thus, given the nuclear coordinates, determines the (7)
ground-state energy, Ev(r)[n(r)], and all properties of
the ground state, i.e., where xc(n) is the exchange-correlation energy per
particle of a uniform interacting electron gas of
Ev(r)[n(r)] ≡ ∫ v(r)n(r) dr + F[n(r)]; F[n(r)] ≡ density n. The KS orbitals in the LDA are surpris-
(Ψ[n(r)],(T + U)Ψ[n(r)]) (3) ingly close to HF orbitals, although DFT-LDA
calculations neglect the nonlocal nature of exchange
where F[n(r)] is the desired (unknown) functional, terms that are typical for the HF method. The next
and T and U are the kinetic and potential energies, level of approximations is the generalized gradient
respectively. The second states that, given the func- approximation (i.e., GGA):
tional, it is the one that minimizes the energy,
thereby providing a variational principle to determine EGGA
xc ≡ ∫f(n(r),|∇n(r)|) dr (8)
the density, i.e.,
in which f(n,|∇n|) is a suitably chosen function of its
Ev(r)[n(r)] g Ev(r)[n0(r)] ≡ E (4) two variables. Popular correlation GGAs include
those of Lee, Yang, and Parr,64 Perdew (1986),65 and
F[n(r)] ) Perdew and Wang (1991).66 These gradient-corrected
1 n(r)n(r′) correlation functionals in combination with the Becke
Ts[n(r)] + ∫
2 |r - r′|
dr dr′ + Exc[n(r)] (5) exchange functional67 (based on considering the
exchange energies for rare gases, in addition to the
where the first term (Ts[n(r)]) is the kinetic energy, known behavior for the uniform electron gas), typi-
the second term the classical Coulombic repulsion cally yield accurate relative energies and good ther-
energy, and the last term (Exc[n(r)]) the exchange mochemistry.68
correlation energy. The challenge in DFT is the As a consequence of the size limitations of the ab
design of accurate functionals, and, specifically, the initio schemes, a large number of more-approximate
difficulties lie in determining Ts[n(r)] and Exc[n(r)]. methods can be found in the literature. Here, we
Considerable progress in constructing a kinetic en- mention only the density functional-based tight bind-
ergy density functional came through the reformula- ing (DFTB) method,69-72 which is a two-center ap-
tion of DFT by Kohn and Sham, termed Kohn-Sham proach to DFT. The method has been successfully
(KS) density functional theory.63 In KS-DFT, an applied to the study of proton transport in perov-
artificial reference system is constructed that consists skites73 and imidazole37 (see Section 3.1.1.3). The
of noninteracting electrons, which has exactly the fundamental constraints of DFT are (i) treatment of
same electron density as the real molecular system excited states and (ii) the ambiguous choice of the
of interacting electrons. The kinetic energy is ap- exchange correlation function. In many cases, the
proximated as that of the noninteracting reference latter contains several parameters fitted to observ-
system, which can be exactly evaluated in terms of able properties, which makes such calculations, in
the Kohn-Sham orbitals φj. The self-consistent set fact, semiempirical.
of KS equations is
2.2. Molecular Dynamics
(- ∇2 + v(r) + ∫
1
2
n(r′)
|r - r′| )
dr′ + vxc(r) - j φj(r) ) 0 Classical molecular dynamics (MD) implementing
predetermined potentials, either empirical or derived
N from independent electronic structure calculations,
n(r) ) ∑|φj(r)|2 has been used extensively to investigate condensed-
j)1 matter systems.74 An important aspect in any MD
[n(r)] simulation is how to describe or approximate the
vxc(r) ) δExc interatomic interactions. Usually, the potentials that
δn(r) describe these interactions are determined a priori
v n(r)n(r′) and the full interaction is partitioned into two-,
E ) ∑ j - ∫ dr dr′ - ∫vxc(r)n(r) dr + three-, and many-body contributions, long- and short-
1 |r - r′| range terms, etc., for which suitable analytical func-
Exc[n(r)] (6) tional forms are devised.75 Despite the many suc-
cesses with classical MD, the requirement to devise
where the final equation is the expression for the fixed potentials results in several serious problems
in systems where distinct atoms or molecules cause where mi is its mass and the force Fi is the negative
a myriad of different interatomic interactions that gradient of the potential energy E(R), with respect
should be parametrized or where the electronic to i. The MD scheme leads (contrary to the MC
structure or bonding pattern changes qualitatively scheme) to a time-correlated sequence of configura-
in the course of the simulation.76,77 These drawbacks tions (trajectory), which can be analyzed to calculate
have been overcome by the various techniques of ab dynamic properties of the system.
initio molecular dynamics (AIMD). An overview of a The potential energy is often written as a sum of
few of the various MD schemes implemented in the pairwise additive interactions. Frequently, Coulombic
study of proton conduction is presented in the fol- interactions between partially charged atoms and
lowing sections. additional interaction functions that describe short-
range repulsion due to exchange-correlation effects
2.2.1. Classical Molecular Dynamics and Monte Carlo and long-range disperse attraction are used as in-
Simulations gredients in the interatomic potentials. In most cases,
Atomistic computer simulations are a statistical the latter interaction function is of the Lennard-Jones
mechanical tool to sample configurations from the type. Molecular geometry and connectivity is main-
phase space of the physical system of interest. The tained by specifying harmonic or Morse-type stretch-
system is uniquely treated by specifying the inter- ing interactions along a chemical bond, and angle-
actions between the particles (which are usually bend interactions for valence angles and torsion
described as being pointlike), the masses of all the interactions to maintain molecular conformations.
particles, and the boundary conditions. The inter- Thus, a typical simple force field is the sum over all
actions are calculated either on-the-fly by an elec- pairs of Coulombic and short-range interactions, all
tronic structure calculation (see Section 2.2.3) or from bonds, all valence angles, and all dihedral angles in
potential functions, which have been parametrized a molecule. More complex force fields can contain
before the simulation by fitting to the results of more elaborate terms and couplings between them.
electronic structure calculations or a set of experi- Many-body induction interactions are often included
mental data. In the first case, one frequently speaks in the interaction energy E(R) by introducing fluc-
of AIMD (see Section 2.2.3), although the motion of tuating point dipoles or fluctuating point charges,
the nuclei is still treated classically. whose value is self-consistently determined through
the instantaneous electric field and the atomic or
Having specified the interactions (i.e., the model molecular polarizabilities. An alternative is the in-
of the system), the actual simulation then constructs troduction of extra charged particles in a shell model
a sequence of states (or the system trajectory) in some (SM).80
statistical mechanical ensemble. Simulations can be
stochastic (Monte Carlo (MC)) or deterministic (MD), Typical numbers of atoms range over the order of
or they can combine elements of both, such as force- several hundred to several tens of thousands, which
biased MC, Brownian dynamics, or generalized Lan- are located in a regular cell of volume V. The cell and
gevin dynamics. It is usually assumed that the laws the particles are replicated infinitely in one, two, or
of classical mechanics (i.e., Newton’s second law) may three directions of space, depending on whether a
adequately describe the atoms and molecules in the cylindrical or other one-dimensional system, a slab
physical system. system with two external interfaces, or a bulk system
is to be simulated. These periodic boundary condi-
The MC scheme was first published by Metropolis tions avoid undesired surface effects beyond those
et al.78 in 1953 and applied to the calculation of the that one explicitly wishes to study. The cells are open
equation of state of a simple hard-sphere liquid. Each and particles can move freely from one cell to the
configuration in an MC simulation is generated next. For each particle leaving the cell, one of its
stochastically in such a way that the molecular replicas enters the cell; therefore, the overall particle
configuration is dependent only on the previous number in the cell, and thus the density, remains
configuration. The MC method is often performed in constant. Because the number of interactions to be
the canonical ensemble, i.e., for a fixed number of calculated in such a system is infinitely large, short-
molecules N placed in a fixed volume V and main- range interactions (such as the Lennard-Jones term)
tained at a constant temperature T. However, many need to be smoothly truncated, and lattice summa-
variants of the method exist (see, e.g., ref 74). During tion methods (such as the Ewald method) are imple-
the MC simulation, configurations are generated in mented for the treatment of long-range forces.
such a way that, after many configurations have been
obtained, each configuration occurs approximately The theory of statistical mechanics provides the
with the appropriate probability of the canonical formalism to obtain observables as ensemble aver-
ensemble, given by the Boltzmann factor exp[-E(R)/ ages from the microscopic configurations generated
(kT)], where E(R) is now the interaction potential, by such a simulation. From both the MC and MD
and k the Boltzmann constant. trajectories, ensemble averages can be formed as
simple averages of the properties over the set of
The MD method was first used by Alder and configurations. From the time-ordered properties of
Wainwright.79 In the standard MD scheme for equi- the MD trajectory, additional dynamic information
librium systems, the positions ri of atom i are can be calculated via the time correlation function
obtained by solving Newton’s equations of motion: formalism. An autocorrelation function caa(t) ) 〈a(τ)
‚ a(t + τ)〉 is the ensemble average of the product of
mir̈i ) Fi ) - ∇iE(R) (9) some function a at time τ and at a later time t + τ.
In an equilibrium system, all times t are equivalent energy of each state is calculated from empirical force
and can thus be averaged over. Transport coefficients field terms for intramolecular hydronium interac-
can be calculated as integrals over these functions, tions, intramolecular water interactions, and the
e.g., intermolecular water-hydronium interactions, yield-
ing the energies in state 1 and 2: H11(R) and H22-
〈[r(t + τ) - r(τ)]2〉 (R), which are generally nonidentical. By further
D) ∫
1 ∞
3 0
c vv dt ) lim
xf∞ 6t
(10) specifying empirical coupling functions H12(R) ) H21-
(R) as functions of the set of particle coordinates R,
the compound states can be calculated via diagonal-
The diffusion coefficient D is one-third of the time ization of the 2 × 2 Hamiltonian matrix. This method
integral over the velocity autocorrelation function is called the empirical valence bond (EVB) method
cvv(t). The second identity is the so-called Einstein and was first introduced by Warshel and Weiss.30-32
relation, which relates the self-diffusion coefficient H2(R) can be adjusted to reproduce experimental
to the particle mean square displacement (i.e., the data or ab initio potential energy surfaces. Ap-
ensemble-averaged square of the distance between proximating the motion of protons as classical on the
the particle position at time τ and at time τ + t). (time-dependent) ground-state potential energy sur-
Similar relationships exist between conductivity and face, a viable MD scheme can be developed. The
the current autocorrelation function, and between procedure is analogous for larger clusters with more
viscosity and the autocorrelation function of elements basis states, where a larger matrix must be diago-
of the pressure tensor. nalized. The ground state of the system is specified
2.2.2. Empirical Valence Bond Models by the eigenvector of the lowest-energy eigenvalue.
Simultaneously, Borgis and Vuilleuimier83-87 and
A standard classical force field is able to describe Voth and co-workers88-93 developed multistate EVB
conformational changes of molecules. However, it is models for proton transport in aqueous solution. In
generally unable to describe the formation and their models, a protonated cluster involving n water
breaking of chemical bonds. The reason is that the molecules, H2n+1On+, is described by n zeroth-order
atoms participating in the stretch, bend, and torsion VB states. In each of these states, the proton defect
terms of the force field need to be specified initially. is formally located on one of the n O atoms (i.e., one
Changing this specification during the simulation of the n O atoms forms three bonds). At any time t,
leads to non-Hamiltonian behavior, which makes the the ground state is calculated via matrix diagonal-
simulation data unusable for analysis. Elegant solu- ization; partial charges (and possibly other param-
tions to this problem are ab initio simulation schemes eters of the force field) are reassigned to the complex
(see Section 2.2.3), which naturally describe the according to the eigenvector. When the proton defect
change of chemical bonding via the instantaneous diffuses structurally through this cluster, the contri-
calculation of the electronic structure. In many bution of some states can become negligibly small
situations, this scheme is computationally too expen- (because the O atom is too far from the proton defect).
sive to be used. Thus, the need arises to develop It is then possible to remove these water molecules
empirical potential functions that (i) allow the change from the cluster and replace them by others, which
of the valence bond network over time and (ii) are are closer to the proton defect but do not yet interact
simple enough to be used efficiently in an otherwise with the cluster. Thus, over time, the composition of
classical simulation code. For dissociation reactions, the proton cluster can change and proton transport
one possibility is the use of interaction potential is possible. The multistate EVB models were used to
functions with the proper asymptotic behavior. As an investigate the structure of protonated complexes in
example, consider the water molecule. In practically aqueous solutions, proton transport dynamics, the
all-empirical water models, O and H atoms carry relaxation of the hydrogen-bonded environment, and
(fixed) partial charges. Thus, these models are inca- the role of the quantum nature of proton motion for
pable of describing the autodissociation of water. For structure and dynamics.86,89
a model that describes the dissociation into protons Based on these experiences, Walbran and Korny-
and hydroxyl ions, full ionic charges need to be placed shev94 developed a much simpler two-state EVB
on the atoms (i.e., +e for the proton, -2e for oxygen). model. Their model is designed in such a way that it
Such Coulombic interactions are too strong at short can be used (i) when treating protons classically and
and intermediate range; thus, an additional potential (ii) for systems with high proton concentrations. The
function must be constructed.81,82 An alternative is rationale behind its development was to eventually
the valence bond (VB) method, where the chemical investigate proton transport in polymer electrolyte
bond in a dissociating molecule is described as the membranes. However, in this initial work, they
superposition of two states: a less-polar bonded state studied only proton mobility in pure water. The
and an ionic dissociated state. Unlike VB theory in ground-state energy is calculated, as for the other
quantum mechanics, the matrix elements of the EVB models, from the lowest eigenvalue, in this
Hamiltonian are not calculated on an electronic basis simple case, according to
(see Section 2.1) but by empirical force fields.
As an example, proton transfer in an H5O2+ com- 1
plex may be described as a superposition of two E ) - (H11 + H22 + xH211 + H222 + 4H212) (11)
2
states, namely H2O-H‚‚‚OH2 and H2O‚‚‚H-OH2
(where the solid line (-) describes a chemical bond What distinguishes it from the multistate models is
and the dotted line (‚‚‚) is a hydrogen bond). The the fact that the partial charges on the atoms are
not determined from the eigenvectors of the ground each BO MD step, which may be solved (for example)
state but through a charge-switching function, which, under either the HF approximation or with KS-DFT
in turn, is dependent only on the coordinates of the (see previously given brief method descriptions). A
Zundel complex. Thus, Coulombic interaction (and commonly used technique implementing the latter
other time-consuming interactions) can be calculated electronic structure approach is the VASP total-
using the adiabatic (effective) charges. In this way, energy code.95-97
the calculation of Coulombic interactions is substan- The Car-Parrinello approach98 to AIMD exploits
tially simplified; furthermore, the diagonalization of the quantum mechanical adiabatic time-scale sepa-
the individual protonated complexes decouples and ration of fast electronic and slow nuclear motion by
can be easily performed, even at water-to-proton transforming that into adiabatic energy scale separa-
ratios on the order of 5-10, for which the use of the tion in the framework of dynamical systems theory.
multistate EVB models with 10 or more basis states This is achieved through mapping the two-component
would be impossible when more than one proton is quantum/classical problem onto a two-component
present. The model parameters were fit to reproduce purely classical problem with two separate energy
the structure and formation energies of small proto- scales at the expense of loosing the explicit time
nated clusters. Because of the limitation of treating dependence of the quantum subsystem dynamics.
only two VB states and limitations in the parameter The CPMD method makes use of the following
choice mandated by the requirement that the model classical Lagrangian:
needs to remain Hamiltonian in nature and simul-
taneously allow proton transport, the model usually 1 1
underestimates the mobility of excess protonic defects LCP ) ∑ µi〈ψ̇i|ψ̇i〉 + ∑MiR
2 2i - E[{ψi},{Ri}] (14)
i 2 2 i
and overestimates the self-diffusion coefficient of
water, whereas the temperature dependence is usu- to generate trajectories for the ionic and electronic
ally close to the experimental values. This two-state degrees of freedom via the coupled set of equations
EVB model was recently applied to simulations of of motion:
“model” polymer electrolyte membranes (see Section
3.1.2.).26-29 ∂E[{ψi},{Ri}]
2 Ri ) -
M iR R
) FCP (15)
2.2.3. Ab Initio Molecular Dynamics (AIMD) ∂RRi i
The basic underlying methodology in AIMD is the

computation of the forces acting on the nuclei from ∂E[{ψi},{Ri}]
µi|ψ̈i〉 ) - (16)
electronic structure calculations (see previous discus- ∂〈ψi|
sion) that are performed “on the fly” as the trajectory
of the molecules is generated. Thus, the electronic where Mi and Ri are the mass and position, respec-
variables are not integrated out beforehand but are tively, of atom i; |ψi〉 are the KS orbitals, which are
active degrees of freedom. This implies that, given a allowed to evolve as classical degrees of freedom with
suitable approximate solution of the many-electron inertial parameters µi; and ∂E[{ψi},{Ri}] is the KS
problem, “chemically complex” systems may be treated energy functional evaluated for the set of ionic
with AIMD. Furthermore, the approximation is shifted positions {Ri} and the set of orbitals {ψi}. The
from the level of selecting the appropriate potentials functional derivative of the KS energy is implicitly
to the level of choosing a particular approximation restricted to variations of {ψi} that preserve or-
for solution of the Schrödinger equation. Despite thonormality. Thus, the electrons are put into their
these advantages, however, there is a price to be paid ground state at a fixed set of ionic positions and with
in AIMD: the correlation lengths and relaxation the ions moving according to the previously given
times that are accessible are much smaller than what equation, the electronic orbitals should adiabatically
is typically accessible via classical MD. Of the various follow the motion of the ions, performing only small
AIMD techniques, we will restrict our discussion in oscillations about the electronic ground state. The
this review to only two methods: (a) Born-Oppen- electronic orbitals will possess a “fictitious” kinetic
heimer (BO) molecular dynamics and (2) Car-Par- energy, according to their motion, and, thus, a
rinello molecular dynamics (CPMD). fictitious mass parameter µi. If µi is small, then the
In BO AIMD, the static electronic structure is motion of the orbitals should be fast, relative to the
solved at each MD time step, given the set of fixed motion of the ions.
nuclear positions at that instance in time. Thus, the
problem is reduced to solving a time-independent 2.3. Poisson-Boltzmann Theory
quantum problem concurrently to propagating the
nuclei via Newton’s second law. Therefore, the BO The Poisson-Boltzmann (PB) theory has been used
method is defined by to calculate the influence of the charged groups in
PEM pores on the energy barriers for proton jumps
and in combination with phenomenological concepts
2 i(t) ) - ∇imin{〈Ψ0|He|Ψ0〉}
MiR (12) of charge-transfer theory and the temperature de-
Ψ0
pendence of the reaction rate.99,100 In the mean-field
E0Ψ0 ) HeΨ0 (13) PB theory, the distribution of mobile charges (here,
protons) in a pore is calculated from the boundary
for the electronic ground state. It is important to conditions and the arrangement of external static
realize that the minimum of 〈He〉 must be reached at charges (here, sulfonate groups), neglecting both
correlations between the mobile charges and a spa- The first microscopic theory for ionic friction in
tially dependent dielectric constant of the water polar solvents was proposed by Wolynes,111 in which
within the pore. Because of such drastic assumptions, the ion-solvent interactions were partitioned into
the density of the protons near the sulfonate groups short-range repulsive and long-range attractive com-
is far too high (similar to that predicted by a simple ponents. The friction coefficient in the Wolynes model
Gouy-Chapman model; see Section 3.1.2.1). Specify- is simplified into the following two terms:
ing a geometric model and an arrangement of sul-
fonate groups (slab pores of varying width with a 1
ζ ) ζ0 + 〈F2 〉τ (17)
regular lattice of negative point charges on the pore 3kBT S F
surface99), proton distributions and electrostatic po-
tential barriers for proton motion along the pore were where ζ0 is calculated from Stokes law, 〈F2S〉 is the
calculated. Although the model did not directly yield static mean-square fluctuation in the soft force, and
an absolute value for proton mobility, activation τF is its characteristic decay time. Hence, in the case
energies of proton transport were estimated, as a of strong, short-ranged attractive interaction, the
function of water content, pore shape, and sulfonate drag on an ion reduces to that on a solvated solvent-
density. A modified PB ansatz, taking into account berg ion, whereas in the limit of the weak long-
a more realistic charge distribution for the sulfonate ranged attractive ion-solvent interaction case, the
groups, was shown to be qualitatively consistent with dielectric friction picture persists. Subsequent work
MD simulations.27 of Wolynes and co-workers112 attempted to implement
this model for monovalent cations (i.e., Li+, Na+, K+,
2.4. Nonequilibrium Statistical Mechanical Ion Cs+) in the model solvents of water, methanol,
Transport Modeling acetonitrile, and formamide. However, their results
Historically, one of the central research areas in were only marginally successful for the case of water;
physical chemistry has been the study of transport poor agreement with experimental results being
phenomena in electrolyte solutions. A triumph of obtained for the other solvents. Chen and Adelman113
nonequilibrium statistical mechanics has been the generalized the Hubbard-Onsager theory, treating
Debye-Hückel-Onsager-Falkenhagen theory, where the interplay between the Stokes and dielectric
ions are treated as Brownian particles in a con- friction contributions within a continuum model, but
tinuum dielectric solvent interacting through Cou- provided no insight into molecular solvent relaxation
lombic forces. Because the ions are under continuous due to solute perturbations. Chong and Hirata114
motion, the frictional force on a given ion is propor- used an interaction-site model and the mode coupling
tional to its velocity. The proportionality constant is theory to show that the friction coefficient decom-
the friction coefficient and has been intensely studied, poses into hydrodynamic friction, dielectric friction,
both experimentally and theoretically, for almost 100 and a coupling term, according to
years.101,102 The simple Stokes law, derived from
hydrodynamic theory where friction increases with ζ ) ζNN + ζZZ + 2ζNZ (18)
increasing ionic radii, is known to fail for small alkali
and halide ions, and, of course, protons.103 where ζNN and ζZZ denote the friction from a collective
In an effort to explain the peculiar behavior of density and dielectric response of the solvent to ionic
small ions in polar solvents, two models have been displacement respectively, and ζNZ the friction due
proposed that attribute different phenomena to sol- to their coupling. Their work shows that both the
vent response and solute or ion displacement. The solventberg and dielectric friction mechanisms con-
first model, which is often called the solventberg tribute to the net friction (and thereby diffusion) of
model, maintains the classical view of Stokes law but small ions and thus there is really no contradiction
with an “effective” ionic radius originating from in the coexistence of these two models. Finally,
solvation.104 Thus, with solvent molecules regarded Bagchi and co-workers115-117 have shown that the
as being bound to the ion, the radius of the solvated calculation of the total friction on a moving ion should
complex is equivalent to a Stokes radius. The other involve formulation with a mode-coupling-type theory
model is a dielectric friction model, which has been (i.e., used to compute a bare friction) and a calcula-
formulated over several decades by Born,105 Fuoss,106 tion of the correlation functions. Their derivation
Boyd,107 and Zwanzig,108 with a complete theoretical gave the following relation for the friction coefficient:
framework due to Hubbard and Onsager.109,110 This
model attempts to describe the dielectric response of 1 1 1
) + (19)
the solvent due to perturbation by the motion of an ζ ζbin + ζFF + ζmic,DF ζhyd + ζhyd,DF
ion. As the ion is displaced from an initial position
where the solvent is polarized according to the where ζbin and ζFF are the binary and the collective
electrostatic field due to the ion, the solvent polariza- solvent number density (F) fluctuation contributions,
tion is not at equilibrium with the new position of respectively; ζmic,DF is the friction contribution orig-
the ion, resulting in a relaxation and consequent inating from the coupling of the ionic field with the
energy dissipation of the polarization identified as orientational solvent polarization mode; and ζhyd,DF
dielectric (or extra) friction. The dielectric friction is and ζhyd are the hydrodynamic friction contributions
inversely related to the ionic radius. Hence, the ion- with and without polar contribution, respectively.
size dependence of the friction coefficient has a The first attempt to apply microscopic electrolyte
minimum with increasing ionic radius in both mod- theory to study the mobility of protons in PEMs is
els. due to Paddison and co-workers.40-46 Because the
foregoing discussion concerns the mobility of small

ions in polar solvents, it would seem that application ζR ) ∫ dt(〈FRse-iL0tFRs〉0 + 〈FRse-iL0tFps〉0 +
β ∞
3 0
of this methodology to proton transport is somewhat 〈FRpe-iL0tFps〉0 + 〈FRpe-iL0tFRs〉0) (20)
questionable. However, because their model focuses
on the coupled transport of a proton with a water
where β ) 1/(kT), and the forces FRs, Fps, and FRp are
molecule (i.e., an hydronium ion), there is perhaps a
between the hydronium ion and the water molecules,
more substantial underpinning to the calculation of
the fixed sites and the water molecules, and the
“a proton friction coefficient” in these membranes. hydronium ion and the fixed sites, respectively. Only
The purpose of their nonequilibrium statistical me- the latter three terms are explicitly evaluated: their
chanical transport model is to calculate proton self- sum is taken to be a correction, ζ(c), to a “zeroth order”
diffusion coefficients for hydrated PEMs from molec- friction coefficient. Because this first force-force cor-
ular and morphological information without resorting relation function involves only the force that the
to any fitting or adjustable parameters through water exerts on the hydronium ion (FRs), it is taken
computation of velocity-independent friction coef- to be either the friction coefficient of a hydronium
ficients. Because pulsed-field gradient (PFG) NMR ion in bulk water calculated with the Stokes relation
measurements allow for the determination of mem- (this is typically the case for minimally or only
brane- and hydration-specific proton self-diffusion partially hydrated membranes), or the friction coef-
coefficients, this kinetic model permits the identifica- ficient of a proton in bulk water derived from
tion of how molecular chemistry and membrane experimental diffusion measurements (used for fully
morphology is connected to a macroscopic quantity hydrated membranes). Structure diffusion of the
that ultimately determines performance of a fuel cell. proton is significant in bulk water (see Section
Their model is based on the methodology of Resibois 3.1.1.1.1); thus, their model does account for this
and others,118 and, because of its surprising success contribution to the mobility of the proton in mem-
in the calculation of proton self-diffusion coefficients brane pores through the first term in eq 20, i.e., FRs.
in various PEMs, over a range of hydration levels, a The choice of the numerical value of this zeroth-order
brief description of the model is given here. term is not arbitrary but is dependent on the char-
acteristics (i.e., structure) of the water in the pore;
Similar to the work described previously,111-117 the and the latter is assessed using the dielectric satura-
starting point of the kinetic model is the assumption tion model described in the following section.
that, under linear response theory, the ensemble
average force is proportional to the velocity of a 2.5. Dielectric Saturation
hydronium ion through the friction coefficient (i.e.,
〈FR〉 ) - ζ‚vR). Computation of this average force In a typical hydrated PEM, the rigidity of the
experienced by a hydronium ion as it transverses the backbone and the crystallinity of the polymer confine
channel, when combined with the Einstein relation, the water (to regions of only nanometers); and with
allows for the calculation of the proton self-diffusion the significant density and distribution of the pen-
coefficient. This pore within a PEM is assumed to dant anionic groups give structural ordering to the
possess a cylindrical geometry filled with water water in the membrane. Both experimental119-122 and
molecules, according to the degree of hydration of the MD simulations123-126 have revealed that water
membrane. The dissociated acidic functional groups confined in systems such as reverse micelles and
in the pore are modeled as radially symmetric axially biological pores possesses a decreased polarity and
periodic arrays of fixed ions (i.e., point charges) and, rate of relaxation, and an increased degree of spatial
thus, the average force experienced by the hydronium and orientational order, when compared to bulk
ion is calculated using the standard methods of water. As such, assumptions of a constant dielectric
constant for the water in the pore of a PEM (either
statistical mechanics, albeit with a suitably con-
that of bulk water or some other values) is clearly
structed nonequilibrium distribution function. This
incorrect. The water near the polymer backbone and
distribution function is obtained from a formal solu- fixed anionic groups is more constrainted and ordered
tion of the time evolution or Liouville equation. The than the water located in the center of the pores. This
Liouville operator is constructed from a Hamiltonian phenomenon resembles that observed for an electro-
of the system with an inertial reference frame moving lyte near a charged electrode for which various
at the velocity of the hydronium ion and consists of models have been formulated, including Helmholtz,
the kinetic energy of all the water molecules and the Gouy-Chapman, Stern, and Grahame (see Section
net potential energy, because of two-body interactions 3.1.2.1).
of the water molecules, hydronium ion, and fixed Structure diffusion (i.e., the Grotthuss mechanism)
sites. The respective contributions to the potential of protons in bulk water requires formation and
energy of the system are due to (i) interactions of the cleavage of hydrogen bonds of water molecules in the
hydronium ion with the water molecules, (ii) inter- second hydration shell of the hydrated proton (see
action of the hydronium ion with the arrays of the Section 3.1); therefore, any constraint to the dynam-
fixed sites, (iii) water-water interactions, and (iv) ics of the water molecules will decrease the mobility
interactions of the water molecules with the fixed of the protons. Thus, knowledge of the state or nature
sites. Thus, the scalar friction coefficient of the of the water in the membrane is critical to under-
hydronium ion is partitioned into four force-force standing the mechanisms of proton transfer and
correlation functions: transport in PEMs.
Hence, the recent work of Paul and Paddison,47-51 pected to occur in weakly hydrogen bonded systems.
which was alluded to previously, has sought to Dynamical bond-length variations have long been
describe the spatially dependent dielectric constant recognized to be integral for the hydrogen bonding
of the water in PEMs. Their work follows in the spirit in proton-conducting systems;130 however, it is only
of the much earlier pioneering work of Booth127 and recently that the appearance of such “dynamical
more recent application of Bontha and Pintauro.128 hydrogen bonding” has been explained through a
They assume that the field-dependent permittivity consideration of the chemical interactions of the
of the water, (E), may be expressed as the sum of systems (including hydrogen bonding) in a wide
two terms, according to range of configuration space.131 The interplay of
hydrogen bonds with other intermolecular forces is
4πP(E) generally examined in refs 132-136.
(E) ) n2 + (21) Proton-conduction mechanisms that occur in the
0E
aforementioned proton solvents, when present as a
where n is the refractive index and E and P are the homogeneous phase and as a component of hetero-
magnitudes of the electric field and polarization, geneous fuel cell separator materials, are described
respectively. The polarization is computed from a more specifically in the following two sections.
realization that it is a functional derivative of the
Helmholtz energy, A, i.e., 3.1.1. Homogeneous Media
3.1.1.1. Water and Aqueous Solutions. 3.1.1.1.1.
δA(E,Ee) Mobility of Excess Protons. The unusually high
P(r,E,Ee) ) - (22) mobility (equivalent conductivity) of protons in water
δEe(r)
and aqueous solutions (under ambient conditions, ∼9
times higher than that for Li+ and ∼5 times higher
where the dependence of both the polarization and
than that for K+) has been investigated since the
energy on the electrostatic field due to the fixed
early days of physical chemistry, and the different
anionic groups (E) and an external “probing” electric
concepts and approaches have been summarized
field (Ee) are explicitly declared. One of the important
several times.34 The essential features of the present
implications of eq 22 is the inclusion of the electro-
view dates back to the work of Eigen and De
static field as part of the total energy (Hamiltonian)
Maeyer.137,138 They demonstrated that “structure
of the system, which removes the nonphysical results
diffusion”, i.e., the “diffusion” of the structure (hy-
of divergence in the dielectric constant.
drogen-bond pattern) in which the excess proton is
“tunneling” back and forth, is the rate-limiting step.
3. Transport Mechanisms Ever since, there has been some controversy as to
whether the region containing the excess proton may
3.1. Proton Conduction Mechanisms be described as a hydrated hydronium ion (i.e.,
Despite the variety of proton-conducting separator H9O4+, later termed the Eigen ion) or a smaller dimer
materials, the inherent protonic charge carriers sharing the excess proton (i.e., H5O2+, the Zundel
(protonic defects) are solvated by very few types of ion139). This debate was largely resolved by the
species. These are essentially water (e.g., in hydrated mechanistic details obtained from CPMD simulations
acidic polymers), oxo-acid anions (e.g., in CsHSO4) (see Section 2.2.3) by Tuckerman et al.34,35 and the
or oxo-acids such as phosphoric acid (e.g., in adducts interpretation of NMR data by Agmon.140 Indepen-
of basic polymers with phosphoric acid), heterocycles dently, they determined a mechanism for the diffu-
(e.g., intercalated into acidic polymers or immobilized sion of excess protons in water, which is illustrated
via flexible spacers), or oxide ions (forming a hydrox- in Figure 1.
ide on the oxygen site of an oxide lattice). These The region with an single excess proton within the
species participate in the formation of protonic charge hydrogen-bond network (protonic defect) corresponds
carriers and the proton conduction mechanism. In a to either a Zundel ion or an Eigen ion. Interestingly,
few cases, they are also the protogenic group: i.e., the center of the region of excess charge coincides
they generate protonic charge carriers by self-dis- with the center of symmetry of the hydrogen-bond
sociation (e.g., in the case of phosphoric acid and to pattern,131 i.e., apart from the bonds with the com-
some extent also in heterocycles such as imidazole); mon shared proton, each water molecule of the
in other cases, the protonic charge carriers must be Zundel ion acts as a proton donor through two
generated extrinsically, by doping with a Brønsted hydrogen bonds, and each of the three outer water
acid or base (i.e., in water containing systems and molecules of the Eigen ion acts as a proton donor in
oxides). two hydrogen bonds and as an acceptor for the
A common and important characteristic of all these hydronium ion and an additional water molecule (see
species is their involvement in hydrogen bonding. The Figure 1). Changes to these hydrogen-bond patterns
structural and dynamical nature of this interaction through hydrogen-bond breaking and forming pro-
seems to be the key to understanding long-range cesses displaces the center of symmetry in space and,
proton transport in these environments. Strong hy- therefore, also the center of the region of excess
drogen bonding is frequently considered to be a charge. In this way, a Zundel ion is converted to an
precursor of proton-transfer reactions;129 however, Eigen ion, which then transfers into one of three
long-range proton transport also requires rapid bond possible Zundel ions (see Figure 1). This type of
breaking and forming processes, which is only ex- mechanism may still be termed “structure diffusion”
Figure 1. Proton conduction in water. The protonic defect follows the center of symmetry of the hydrogen-bond pattern,
which “diffuses” by hydrogen-bond breaking and forming processes; therefore, the mechanism is frequently termed “structure
diffusion”. Note that the hydrogen bonds in the region of protonic excess charge are contracted, and the hydrogen-bond
breaking and forming processes occur in the outer portion of the complexes (see text). Inserted potentials correspond to
nonadiabatic transfer of the central proton in the three configurations (atomic coordinates taken from refs 33 and 34).
(as suggested by Eigen), because the protonic charge almost barrierless proton translocation while the
follows a propagating hydrogen-bond arrangement or hydrogen-bond breaking and forming processes occur
structure. in the weakly bonded outer parts of the complexes.
The sum of all proton displacements involved in This contraction of the center of the complex is
the hydrogen-bond breaking and forming processes probably a direct consequence of the lower coordina-
and the proton displacements within the hydrogen tion of the involved species (3 instead of ∼4). The
bonds of the Zundel and Eigen ions then corresponds activation enthalpy of the overall transport process
to the net displacement of one unit charge by just a is dominated by the hydrogen-bond breaking and
little more than the separation of the two protons in forming, which also explains the strong correlation
a water molecule (i.e., ∼200 pm). Although there are of the proton transport rate and the dielectric relax-
no individual, exceptionally fast protons, even on a ation.142 The Zundel and Eigen complexes are just
short time scale, the fast diffusion of protonic defects limiting configurations, and the simulations indeed
leads to a slight increase in the physical diffusion of produce configurations that can hardly be ascribed
all protons in the system. This is indeed observed for to one or the other.36 Although simulations find
aqueous solutions of hydrochloric acid, for which comparable probabilities for the occurrence of both
mean proton diffusion coefficients were observed to type of complexes,34,35 there is experimental evidence
be up to 5% higher than the diffusion coefficient of for a slight stabilization of the Eigen complex,
oxygen, as measured by 1H and 17O PFG NMR,141 compared to the Zundel complex, by ∼2.4 kJ/mol (25
reflecting the slightly attenuated correlation of proton meV).143 However, these differences are marginal and
and oxygen diffusion in acidic media. do not change the principal features of the mecha-
Another interesting feature of this mechanism is nism.
that the hydrogen-bond breaking and forming (hy- One such feature is the potential energy surface
drogen-bond dynamics) and the translocation of for proton transfer in the contracted hydrogen bonds.
protons within the hydrogen bonds occur in different The time-averaged potential surfaces are almost
parts of the hydrogen bond network, albeit in a highly symmetrical (especially for the Zundel ion) without
concerted fashion. This is the most thermodynami- significant barriers, i.e., the proton is located near
cally favorable transport path, because the hydrogen the center of the bond. Whether its location is off-
bonds in the center of the two charged complexes are center at any time instance is mainly dependent on
contracted to such an extent that this allows an the surrounding hydrogen-bond pattern, and it is the
change of this pattern that alters the shape (and

asymmetry) of this potential and, therefore, the
position of the proton within the hydrogen bond (see
Figure 1, top). In other words, the proton is trans-
ferred almost adiabatically, with respect to the
solvent coordinate.1 This has important consequences
on the mechanism when static asymmetric contribu-
tions are introduced, i.e., by chemical interactions or
the presence of ionic charges (see below). The very
low barriers are also the reason for the fact that the
mechanism can be well described classically, with
respect to the motion of the nuclei (especially the
proton); in particular, proton tunneling has only a
minor effect on the rate of transfer. Nevertheless, the
protonic defect (region of protonic excess charge) may
become delocalized through several hydrogen bonds, Figure 2. Conductivity diffusion coefficient (mobility) of
because of quantum fluctuations.36,85 protons and water self-diffusion coefficient of aqueous
solutions of hydrochloric acid (HCl), as a function of acid
The mechanism also provides insight into the concentration (molarity, M) (data are taken from ref 141).
question as to what extent proton transfer in water
is a cooperative phenomenon. In many physical
3.1.1.1.2. Mobility of Defect Protons. Because basic
chemistry textbooks, one still finds diagrams showing
aqueous solutions also have some relevance for fuel
the concerted transfer of protons within extended
cell applications (e.g., in alkaline fuel cells (AFCs)),
hydrogen-bonded water chains (the Grotthuss mech-
the transport mechanism of defect protons (OH-) is
anism) to explain the unusually high equivalent
also reviewed here. Although intuition might lead one
conductivity of protons in this environment. However,
to assume that the mechanism is reminiscent of that
the creation of the corresponding dipolar moment in
of an excess proton, recent CPMD simulations sug-
an unrelaxed high dielectric constant environment
gest that this may not be the case.149 As opposed to
costs far too much energy to be consistent with a very
Zundel and Eigen complexes, in which the central
fast process with low activation energy.1,144 As an-
species are only 3-fold coordinated (under-coordi-
ticipated in ref 1, the propagation mechanism of a
nated, with respect to water in pure water, which is
protonic defect in a low-dimensional water structure
probably the reason for the bond contraction in these
surrounded by a low dielectric environment is obvi-
complexes; see above) on the average, the hydroxide
ously between “concerted” and “step-wise”;145,146 how-
ion is observed to be coordinated by ∼4.5 water
ever, in bulk water, the cooperation is restricted to
molecules in an almost-planar configuration with the
the dynamics of protons in neighboring hydrogen
OH proton pointing out of the plane. This is consid-
bonds (also see Figure 1).
ered to be a true quantum effect and contradicts the
One also should keep in mind that water is a liquid common understanding of a 3-fold coordination.150
with a high self-diffusion coefficient (i.e., DH2O ) 2.25 This “hyper-coordination” is suggested to prevent
× 10-5 cm2/s at room temperature) and that the proton transfer from a H2O to the OH-, because this
diffusion of protonated water molecules makes some would produce an unfavorable H-O-H bond angle
contribution to the total proton conductivity (vehicle of 90°. The proton transfer only occurs when the OH-
mechanism147). This is ∼22% when assuming that the coordination is reduced to 3 by breaking one of the 4
diffusion coefficients of H2O and H3O+ (or H5O2+) are hydrogen bonds within the plane and some re-
identical. However, as suggested by Agmon,148 the arrangement of the remaining bonds occurs, which
diffusion of H3O+ may be retarded, because of the allows the direct formation of a water molecule with
strong hydrogen bonding in the first hydration shell. a tetrahedral geometry. Surprisingly, the ground-
Of course, the relative contributions of “structure state coordination of the most favorable configura-
diffusion” and “vehicular diffusion” are dependent on tions around excess protons seem to be similar to the
temperature, pressure, and the concentrations and coordination of the transition state for the transport
types of ions present. With increasing temperature, of defect protons. Note, that hyper-coordination of the
structure diffusion is attenuated and with increasing OH- is still a matter of controversial debate. The
pressure the contribution of structure diffusion in- statistical mechanical quasi-chemical theory of solu-
creases until it reaches a maximum at ∼0.6 GPa (6 tions suggests that tricoordinated OH- is the pre-
kbar).1 Especially relevant for the later discussion of dominant species in the aqueous phase under stan-
proton transport in hydrated acidic membranes is the dard conditions.151,152 This finding seems to be in
observation that structure diffusion strongly de- agreement with recent spectroscopic studies on hy-
creases as the acid concentration increases (see droxide water clusters, and it is consistent with the
Figure 2 and ref 141), which is probably due to traditional view of OH- coordination.
changes in the hydrogen-bond pattern (there are It should also be mentioned that OH- “hyper-
progressively more proton donors than corresponding coordination” is not observed in concentrated solu-
proton acceptor “sites”) and a consequence of the tions of NaOH and KOH.153 In contrast to acidic
biasing of the hydrogen bonds in the electrostatic solutions, where structure diffusion is suppressed
field of the ions suppressing the proton-transfer with increasing concentration (see above and Figure
mechanism illustrated in Figure 1. 2) the transference number of OH- (e.g., in aqueous
KOH solutions) remains surprisingly high (∼0.74) for phate species (essentially H2PO4-, H3PO4, H4PO4+)
concentrations up to ∼3 M. have not been investigated yet; however, the high
In pure water, excess protons (H3O+, H5O2+) and degree of self-dissociation suggests that the proton-
defect protons (OH-) are present at identical concen- transfer events are even less-correlated than in water
trations; however, because of their low concentration (the system is more tolerant toward protonic charge
(10-7 M under ambient conditions), the diffusion of density fluctuations). The transfer events are prob-
these defects may be considered to be quasi-indepen- ably almost barrierless, as indicated by negligible
dent. Only in small water clusters do zwitterions H/D effects of the diffusion coefficients in mixtures
show a remarkable stability.154 of H3PO4 and D3PO4.130
3.1.1.2. Phosphoric Acid. Although the proton The principal proton transport mechanism remains
conduction mechanism in phosphoric acid has not essentially the same with the addition of some water,
been investigated to the same extent, it is evident with a conductivity increase up to 0.25 S/cm under
that the principal features exhibit similarities, along ambient conditions. A 1H and 31P PFG NMR study
with important differences. also showed that an 85 wt % phosphoric acid system
Above its melting point of Tm ) 42 °C, neat was an almost ideal proton conductor, with 98% of
phosphoric acid (H3PO4) is a highly viscous liquid the conductivity originating from the structure dif-
with extended intermolecular hydrogen bonding. fusion of protons.159 The combination of high intrinsic
However, in contrast to the situation in water, there charge carrier concentration and mobility renders the
are more possible donor than acceptor sites and the possibility of very high conductivities in these sys-
amphoteric character is significantly more pro- tems. In particular, there is no perturbation from
nounced: phosphoric acid may act as both a Brønsted extrinsic doping, i.e., there is no suppression of
acid and base. In terms of equilibrium constants, both structure diffusion, despite the high concentration of
Ka and Kb are reasonably high (Ka of the conjugate protonic charge carriers. On the other hand, attempts
base is low). Consequently, phosphoric acid shows a to increase the conductivity of phosphoric acid-based
very high degree of self-dissociation (autoprotolysis) systems by doping have expectedly failed.160
of ∼7.4%,155 along with some condensation, with 3.1.1.3. Imidazole. Historically, the interest in
H2PO4-, H4PO4+, H3O+, and H2P2O72- being the main hydrogen bonding and proton conductivity in hetero-
dissociation products. Because of their high concen- cycles has its roots in speculations about the partici-
tration, the separation of the overall conductivity into pation of hydrogen bonds in energy and charge
charge carrier concentration and mobility terms is transfer in biological systems;161 specifically, concern-
problematic. Nevertheless, the proton mobility has ing the participation of NH‚‚‚N bonds between the
been calculated from total conductivities by the imidazole groups of histidine in proton transport in
Nernst-Einstein equation by taking concentrations transmembrane proteins.162 Even Zundel has worked
from ref 155; and the values have been observed to in the field,163 and it is not surprising that his view
be almost 2 orders of magnitude higher than the of the proton dynamics in imidazole is closely related
values for the diffusion coefficient of the diverse to that of water. He suggested a high polarizability
phosphate species obtained directly by 31P PFG of the protons within intermolecular hydrogen bonds
NMR156 and estimated from viscosity measurements and, as a consequence, a very strong coupling be-
via the Stokes-Einstein relation. tween hydrogen bonds, as indicated by the intense
Pure phosphoric acid is a liquid with a low diffusion IR continuum in the NH stretching regime. Surpris-
coefficient of phosphate species but an extremely high ingly, he did not suggest the existence of any complex
proton mobility, which must involve proton transfer similar to the Zundel complex in water163 (see Section
between phosphate species and some structural re- 3.1.1.1), whereas Riehl164 had already suggested
arrangements. The contribution to the total conduc- “defect protons” or “proton holes” as requirements to
tivity is ∼98%; i.e., phosphoric acid is an almost-ideal maintain a current in solid imidazole. Early conduc-
proton conductor. The total conductivity at the melt- tivity measurements had their focus on crystalline
ing point (T ) 42 °C) is 7.7 × 10-2 S/cm, with an monoclinic imidazole, which has a structural hydro-
estimated proton mobility of 2 × 10-5 cm2/s.156 This gen bond length of 281 pm.165 The measured conduc-
extremely high proton mobility has also been indi- tivities were typically low (i.e., ∼10-8 S/cm) with very
rectly determined with 1H PFG NMR and was poor reproducibility.161,166,167 Later tracer experi-
observed to be even higher (by a factor of 1.5-2.3). ments168 and a 15N NMR study169 raised doubts about
This has been explained by the correlated motion of the existence of proton conductivity in pure crystal-
the oppositely charged defects (H2PO4-, H4PO4+) line imidazole.
when they are close to one another (i.e., the case just However, the conductivity of liquid imidazole is
after their formation (by dissociation of H3PO4) and several orders of magnitude higher (∼10-3 S/cm at
before their neutralization). Correlation effects are the melting point of Tm ) 90 °C);161 however, its
actually quite common in proton conductors with mechanism was investigated much later. It was the
high concentrations of charge carriers and they are search for chemical environments that were different
even more pronounced in other systems (see Section from water in fuel cell membranes that brought
3.1.1.3 on heterocycles and proton transport in alka- heterocycles back into focus. The potential proton
line-metal hydroxides157,158). donor and acceptor functions (amphoteric character),
Molecular details of the structure diffusion mech- the low barrier hydrogen bonding between the highly
anism with the hydrogen-bond breaking and forming polarizable N atoms, and the size and shape of the
and the proton transfer between the different phos- molecule were reasons that Kreuer et al.170 started
to investigate the usefulness of heterocycles as proton

solvents in separator materials for fuel cells. This
work also comprises the study of the transport
properties of neat and acidified liquid imidazole,
pyrazole, and, later, benzimidazole.171 An important
finding was that the transport coefficients (i.e.,
mobility of protonic charge carriers and molecular
diffusion coefficients) are similar to those of water
at a given temperature, relative to the melting point.
This is particularly true for their ratio: i.e., the proton
mobility is a factor of ∼4.5 higher than the molecular
diffusion coefficient at the melting point of imida-
zole.170 This is a direct indication of fast inter-
molecular proton transfer and the possibility of
structure diffusion in this environment. Subse-
quently, details were revealed by a CPMD simula-
tion.37 In contrast to earlier suggestions of concerted
proton transfer in extended chains of hydrogen
bonds161,172 (analogous to the proton conduction mech-
anism in water presented in most textbooks at that
time144), a structure diffusion mechanism similar to
that for water (Figure 1) was observed. The region
containing the excess proton (intentionally intro-
duced) is an imidazole with both nitrogens pro-
tonated and acting as proton donors toward the
two next nearest imidazoles in a configuration
Imi‚‚‚ImiH+‚‚‚Imi with hydrogen bonds (∼273 pm)
slightly contracted, compared to the average bond
length of the system, but still longer than the bonds
in the isolated complex (in the gas phase).173 The
position of the protons within these hydrogen bonds
is dependent mainly on the hydrogen bonding be-
tween the nearest and next-nearest solvating imida- Figure 3. Proton conduction mechanism in liquid imida-
zoles (Figure 3). The hydrogen-bonded structure in zole, as revealed by a Car-Parrinello molecular dynamics
imidazole is observed to be chainlike (i.e., low dimen- (CPMD) simulation.37 Note the similarities with the proton
sional), with two possible orientations of the hydrogen- conduction mechanism in water (see Figure 1).
bond polarization within segments which are sepa-
lower than that for phosphoric acid (see Section
rated by imidazoles with their protonated nitrogen
3.1.1.2). Site-selective proton diffusion coefficients
directed out of the chain. This may even form a
(obtained by 1H PFG NMR of different imidazole-
“cross-linking” hydrogen bond with a nonprotonated
based systems) show surprisingly high diffusion
nitrogen of a neighboring strand of imidazole. The
coefficients for the protons involved in hydrogen
simulation data revealed the existence of imidazole
bonding between the heteroatoms (nitrogen).174 De-
molecules close to the protonic defect in hydrogen-
pending on the system, they are significantly higher
bond patterns, which rapidly change by bond break- than calculated from the measured conductivities
ing and forming processes. Similar to water, this corresponding to Haven ratios (σD/σ) of 3-15. This
shifts the excess proton within the region and may indicates some correlation in the diffusion of the
even lead to complete proton transfer, as displayed proton, which may be due to the presence of a counter
in Figure 3. There is no indication of the stabilization charge neglected in the simulation.
of a symmetrical complex (Imi‚‚‚H‚‚‚Imi)+: there In pure imidazole, regions containing excess pro-
always seems to be some remaining barrier in the tons must be charge-compensated by proton-deficient
hydrogen bonds, with the proton being on one side regions with electrostatic attraction between these
or the other. regions (defects) that is dependent on their mutual
As for the CPMD simulation of water, the simu- separation distance and the dielectric constant of the
lated configuration is artificial, because there is no medium. Under thermodynamic equilibrium, such
counter charge compensating for the charge of the defects are steadily formed and neutralized. For-
excess proton. This is necessary, methodologically, mally, the creation of a protonic defect pair is
because self-dissociation is unlikely to occur within initiated by a proton transfer from one imidazole to
the simulation box used (i.e., eight imidazole mol- another, with the subsequent separation of the two
ecules with a single excess proton) and the accessible charged species with a diffusion mechanism as
simulation time (∼10 ps). The self-dissociation con- described previously (also see Figure 3). However,
stants for heterocycles (in particular, imidazole) are this transfer is contrary to the electrostatic field of
actually much higher than for water, but degrees of the counter charge, favoring a reversal of the dis-
self-dissociation (concentration of protonic charge sociation process. However, because the two protons
carriers) of ∼10-3 are still ∼2 orders of magnitude of the positively charged imidazolium (ImiH+) are
temperatures is well-established; and the initial

notation OH•O (H•i ), i.e., the assumption that the
defect is a hydroxide ion residing on an oxide ion site
carrying a positive relative charge, is still valid. More
than a decade later, the systematic investigation of
acceptor-doped oxides such as LaAlO3, LaYO3, or
SrZrO3, which had already been known for their
moderate oxide ion conductivity, provided experi-
mental evidence that these materials may be proton
conductors in hydrogen-containing environments.177
The observed conductivities in these materials were
quite low; however, later related compounds based
on SrCeO3178 and BaCeO3179 with high proton con-
ductivities have been discovered and even tested in
different types of electrochemical cells, including fuel
cells,180-185 but the lack of stability under fuel-cell
operating conditions remained an unsolved problem.
Almost another two decades passed before oxides
were synthesized that combined high proton conduc-
Figure 4. Schematic illustration of correlated proton tivity with high thermodynamic stability.186-188 This
transfers in pure liquid imidazole leading to proton diffu- brought these materials closer to a realistic alterna-
sion but not proton conductivity (see text). tive for fuel-cell applications and, hence, their trans-
port properties are reviewed here.
equivalent, there is a 50% chance that another proton The highest conductivities are observed in oxides
is transferred back, provided that the orientational with perovskite-type structures (ABO3) with cubic or
coherence between the dissociating molecules is slightly reduced symmetry.188 Protonic defects are
completely lost. If the same proton is transferred formed by the dissociative absorption of water, which
back, the transient formation and neutralization of requires the presence of oxide ion vacancies V••O. The
an ion pair contributes neither to the proton diffusion latter may be formed intrinsically by varying the
nor to the proton conductivity. However, if the other ratio of main constituents or extrinsically to com-
proton is transferred back, the protons interchange pensate for an acceptor dopant (lower-valent cation).
their positions in the hydrogen bond network, which To form protonic defects, water from the gas-phase
generates diffusion but no conductivity since the dissociates into a hydroxide ion and a proton, with
transient charge separation is completely reversed. the hydroxide ion filling an oxide ion vacancy and
As illustrated for the most simple mechanism of this the proton forming a covalent bond with a lattice
type in Figure 4, the sum of all proton translocation oxygen. In Kröger-Vink notation, this reaction is
vectors form a closed trajectory reminiscent of cyclic written
intermolecular proton-transfer reactions known to
occur in certain organic pyrazole-containing com-
plexes175 and proton diffusion in hydroxides.157,158 H2O + V•• × •
O + OO h 2OHO (23)
Presently, there is no direct proof for such a
mechanism in pure imidazole (e.g., by 15N NMR); by which two hydroxide ions substituting for oxide
however, the observation that the ratio of the proton ions, i.e., two positively charged protonic defects
diffusion and conduction rates virtually coincide with (OH•O), are formed. The crystallographic (time-aver-
the Boltzmann factor (i.e., exp(- EΦ()/(kT)), where aged) structure of such a defect is shown in Figure
EΦ is the electrostatic separation energy of two unit 5. There are eight orientations of the hydroxide ion
charges in a continuum of dielectric constant ) is a stabilized by a hydrogen bond interaction with the
strong indication. eight next-nearest oxygen neighbors in the cubic
This observation is also indicative of the impor- perovskite structure. As opposed to the cases dis-
tance of a high dielectric constant to enable the cussed previously, for which hydrogen bonding is the
formation of protonic charge carriers and to allow dominant intermolecular interaction, here, hydrogen
their mobility being uncorrelated with their conju- bonding is restricted to the defect region. Within this
gated base. Of course, this is also true for the region, hydrogen bonding interferes with other chemi-
separation of protonic charge carriers from extrinsic cal interactions and, together with these, determines
dopants such as acids. In addition to the perturbation the structure and dynamics of the defective region.
of the hydrogen bond network (as observed in aque- As shown by DFTB and CPMD simulations, the
ous solutions), electrostatic effects may also explain principal features of the transport mechanism are
the saturation of proton conductivity with increasing rotational diffusion of the protonic defect and proton
acidity of the heterocyclic systems.171 This effect is transfer toward a neighboring oxide ion.189-191 That
even more pronounced in polymeric systems with is, only the proton shows long-range diffusion, whereas
immobilized protogenic groups and will be discussed the oxygens reside in their crystallographic positions.
in Section 3.1.2.3. Both experiments192-194 and quantum-MD simula-
3.1.1.4. Simple Cubic Perovskites. Since the tions,189,190,195,196 have revealed that rotational diffu-
work of Stotz and Wagner176 in 1966, the existence sion is fast with a low activation barrier. This
of protonic defects in wide-band-gap oxides at high suggests that the proton-transfer reaction is the rate-
Figure 5), this leads only to a slight reduction in the

structural OH/O separations, in contrast to most
instantaneous configurations where one of the eight
OH/O separations is reduced to ∼280 pm, because of
hydrogen bonding.188,200 Although the hydrogen-bond
interaction has a stabilizing effect of ∼0.5 eV on this
configuration, the bond is a “soft”, high-energy hy-
drogen bond with extended bond-length fluctuations.
This also leads to configurations, where the protonic
defect acts almost like a free OH with small OH
stretching amplitudes, compared to the extended
stretching vibrations in the hydrogen bonded state.131
From the thermodynamics of such “dynamical
hydrogen bonds”, one may actually expect an activa-
tion enthalpy of long-range proton diffusion of not
more than 0.15 eV, provided that the configuration
O-H‚‚‚O is linear, for which the proton-transfer
barrier vanishes at O/O distances of less than ∼250
pm. However, the mobility of protonic defects in cubic
perovskite-type oxides has activation enthalpies on
the order of 0.4-0.6 eV.188 This raises the question
as to which interactions control the activation en-
thalpy of proton transfer.
A more-detailed inspection of the MD data showed
that, for most configurations with short OH/O sepa-
ration distances, the proton is not found between the
two oxygens on the edge of the octahedron but outside
the BO6 octahedron, as part of a strongly bent
hydrogen bond200 that still possesses some barrier for
proton transfer. The reason for this is probably the
repulsive interaction between the proton and the
Figure 5. Time-averaged structure of a protonic defect highly charged B-site cation, which prevents a linear
in perovskite-type oxides (cubic case), showing the eight hydrogen bond from forming. The analysis of a few
orientations of the central hydroxide ion stabilized by a
hydrogen-bond interaction with the eight next-nearest transition-state configurations showed that the B-O
oxygen neighbors.186,199 bonds are elongated to some extent and that the
displacement of the proton being transferred is on
limiting step in the considered perovskites. On the the edge of the distorted octahedron.131 In this way,
other hand, the intense red-shifted OH-stretching an almost linear, short configuration of the type
absorptions in the IR spectra (ref 197 and references O-H‚‚‚O is formed. The proton transfer in this
therein) and the results of neutron diffraction experi- configuration probably occurs over a remaining bar-
ments198 are indicative of strong hydrogen bond rier, as indicated by the experimentally observed
interactions, which favor fast proton-transfer reac- H/D-isotope effects.201,202 Although the H/B repulsion
tions rather than fast reorientation processes, the is reduced in this configuration, major contributions
latter requiring the breaking of such bonds. to the activation enthalpy result from the B-O bond
Because the structural oxygen separation is larger elongation and the proton-transfer barrier. Neverthe-
than 290 pm in most perovskite-type oxides, and less, the H/B repulsion may be used as an estimate
strong hydrogen bonds may only be formed at sig- of the upper limit for these contributions to the
nificantly shorter distances, the free energy that the activation enthalpy.
system gains by hydrogen-bond formation is compet- The importance of the H/B repulsion is also wit-
ing with the free energy required for the lattice nessed by the observation that the activation enthal-
distortion necessary for hydrogen bonding. A reanaly- pies of proton mobility in cubic perovskites with
sis of a quantum-MD simulation of a protonic defect pentavalent B-site cations (I-V perovskites) are
in cubic BaCeO3131,188,199 demonstrated that these two significantly higher than for perovskites with tet-
free-energy contributions almost cancel each other ravalent B-site cations (II-IV perovskites).131,187
for a wide range of oxygen separation distances Similar to hydrogen-bonded networks, any reduc-
(∼250-300 pm). Thus, short oxygen separations, tion in symmetry may decrease the proton conductiv-
which favor proton transfer, and large oxygen sepa- ity in oxides. This effect has been investigated in
rations, which allow rapid bond breaking, result in detail by comparing structural and dynamical fea-
similar free energies of the entire system; therefore, tures of protonic defects in yttrium-doped BaCeO3
these separations have similar probabilities of occur- and SrCeO3203 and SrZrO3.204 The large orthorhombic
ring. Indeed, the simulation found the protonic defect distortion of SrCeO3 has tremendous effects on the
to form short, but transient, hydrogen bonds with all arrangement of the lattice oxygen. The cubic oxygen
eight nearest oxygen neighbors. In the time-averaged site degenerates into two sites (O1, O2) with different
picture observed in the diffraction experiments (see acid/base properties. Although, in SrCeO3, the most
basic oxygen is O1, it is O2 in BaCeO3. Assuming that way, as observed in MD simulations of protons in
protons are associated with these sites for the major- CaTiO3.199
ity of the time, they may show long-range proton
transport via the most-frequent O2 sites in BaCeO3. 3.1.2. Heterogeneous Systems (Confinement Effects)
This is in contrast to SrCeO3, where long-range
Homogeneous media have actually been used in
proton transport must involve transfer between
commercial (e.g., phosphoric acid fuel cells (PAFCs),
chemically different O1 and O2 sites. The latter,
AFC) and laboratory fuel cells (e.g., sulfuric acid-
together with the observed bias in rotational diffusion
based direct methanol fuel cells (DMFCs), hydrogen
(defect reorientation), was suggested as the reason
fuel cells with CsHSO4 as an electrolyte); however,
behind the higher activation enthalpy and lower
most modern low-temperature fuel cells rely on the
conductivity in SrCeO3, compared to BaCeO3.203
properties of heterogeneous separator materials such
The mobility of protonic defects was shown to be as hydrated sulfonic acid functionalized polymers
very sensitive not only to reduction in the crystal- (e.g., Nafion) and the adducts of basic polymers with
lographic symmetry but also to local structural and oxo-acids (e.g., in the system of polybenzimidazole
chemical perturbations induced by the acceptor dopant and phosphoric acid (PBI-H3PO4)). In these materi-
or by mixed occupancy on the B-site. Traditionally, als, the homogeneous media (discussed previously)
aliovalent dopants with matching ionic radii are are confined within another phase. This geometric
chosen, and, indeed, this simple concept has proven situation, which also comprises specific interactions
to be successful; e.g., in the development of oxide ion at the usually very large internal interface, not only
conductors. However, when it comes to proton con- modifies the transport behavior within the proton-
ductivity in oxides, this approach clearly fails. Al- conducting phase, but also leads to the appearance
though Sc3+ and In3+ have similar ionic radii to Zr4+, of new transport features, such as electro-osmotic
BaZrO3 shows much lower proton mobility when drag (see Section 3.2.1.1). This section reviews the
doped with scandium or indium, compared to yttrium current understanding of proton conduction in these
as an acceptor dopant with a significantly higher heterogeneous materials.
ionic radius. For the latter, the proton mobility and 3.1.2.1. Hydrated Acidic Polymers. Hydrated
corresponding activation enthalpy are virtually in- acidic polymers are, by far, the most commonly used
dependent of the dopant concentration. Electronic separator materials for low-temperature fuel cells.
structure calculations show a significant effect of the Their typical nanoseparation (also see Section 1)
acceptor dopant on the electron density of the neigh- leads to the formation of interpenetrating hydropho-
boring oxygen, including its affinity for the proton bic and hydrophilic domains; the hydrophobic domain
(O2). Obviously, the chemical match of the dopant gives the membrane its morphological stability,
in yttrium-doped BaZrO3 makes it “invisible” to the whereas the hydrated hydrophilic domain facilitates
diffusing proton.205 However, the most common ob- the conduction of protons. Over the past few years,
servation is decreasing proton mobility and increas- the understanding of the microstructure of these
ing activation enthalpy with increasing dopant materials has been continuously growing, and this
concentration; e.g., as observed in yttrium-doped has been crucial for the improved understanding of
BaCeO3.206 Thus, it is not surprising that mixed the mechanism of proton conduction and the observed
occupancy of the B-site in complex perovskites may dependence of the conductivity on solvent (water and
become unfavorable for proton mobility especially methanol) content and temperature.
when cation ordering occurs.207 The microstructure has chiefly been investigated
The aforementioned considerations provide a quali- through SAXS and small-angle neutron scattering
tative explanation for the empirical finding that the (SANS) experiments. However, because such experi-
highest proton conductivities are observed in oxides ments on specific samples do not provide sufficient
with a perovskite structure. The framework of corner- information, diffraction experiments on samples pre-
sharing octahedral BO6 shows high coordination pared over a wide range of polymer/solvent ratios
numbers for both cation sites (12 for the A-site and (even with different types of solvents) have been
6 for the B-site). There is only one oxygen site in the performed, especially by Gebel and co-workers.208-211
ideal perovskite structure that has each O atom Additional information from water transport mea-
surrounded by eight nearest and four next-nearest surements has been included into the parametriza-
oxygens. Generally the high coordination numbers tion of the most simple microstructural model, as-
lead to low bond strengths and smaller angles suming a hydrophobic matrix with a cubic system of
between the bonds, which is in favor of the above- cylindrical hydrophilic channels.212,213 This approach,
described dynamics (e.g., the rotational diffusion of although highly simplifying, has the advantage that
the protonic defect corresponds to “dynamical hydro- it allows comparison of the mean numbers for the
gen bonding” of the OH, with the eight O atoms extensions of the hydrophilic channels and the hy-
forming the “reaction cage” (see Figure 5)). If the drophobic matrix between for different types of
angles between the possible orientations are small ionomers (also see Section 3.2.1.1). Other constraints
enough, the effective barriers for bond breaking and such as “maximum entropy” (minimum structure)
forming processes are usually <0.2 eV for perovskites have also been used to obtain microstructural infor-
with lattice constants that are not too small. In these mation from two-dimensional diffraction patterns by
cases, hydrogen bonding even to the next-nearest direct Fourier synthesis.214 In addition, a recent
oxygen between the vertices of the octahedral become theoretical investigation by Khalatur et al.215 that
possible, opening another proton-transfer path- implemented a hybrid Monte Carlo/reference interac-
Figure 7. Minimum energy conformation of a two-side-

chain fragment of a perfluoro sulfonic acid polymer (Dow)
with six water molecules, showing the dissociation of both
acidic protons.23
hydration.217 In other words, there is indication for

a progressive side-chain unfolding with increasing
hydration. This idea is supported by electronic struc-
Figure 6. Two-dimensional illustration of some micro- ture calculations (described in Section 2.1) on the
structural features of Nafion for an intermediate water entire Nafion side chain,18 which showed that the
content (see text). unfolding of the chain would require ∼18 kJ/mol.
Because the hydrophilic sulfonic acid groups are
tion site model (MC/RISM) technique to probe the covalently bound to the hydrophobic polymer, they
morphology of Nafion over a range of hydration levels aggregate somewhere in the hydrophobic/hydrophilic
showed that a continuous network of channels might transition region, with an average separation of ∼0.8
exist, even at very low water contents. Based on nm, compared to ∼1 nm expected for a totally
proton conductivity and water diffusion data obtained uniform distribution within the material.212
on diverse membrane materials, Edmondson and Electronic structure calculations have shown that
Fontanella claimed the existence of a universal only 2-3 water molecules per sulfonic acid group are
percolation threshold at a water volume fraction of necessary for dissociation, in accordance with its
∼5%.216 This conclusion was based on the power-law superacidity when bound to a perfluorinated polymer
behavior of the transport coefficients above this water (Figure 7), and when 6 water molecules are added,
concentration. The observations that the membrane separation of the dissociated proton from the sul-
materials continue to conduct below the “threshold” fonate anion is observed.16,20,23 This water (primary
and that the conduction mechanism changes with hydration of the sulfonic acid group) is actually
water content (see below) leave serious doubts about absorbed at low water partial pressures,197 indicating
the validity of this interpretation. However, the the stabilizing effect of water in such systems, as
morphologies resulting from most approaches are in expected from the high energetic stability (at T ) 0
reasonable agreement for the most widely investi- K) of the H3O+, compared to H3SO4+ (-SO3H2+).218
gated Nafion (1100 g/equiv); the principal micro- Apart from water, the hydrophilic domain contains
structural features of this morphology are illustrated only excess protons as mobile species, while the
in Figure 6 for an intermediate water content. The anionic counter charge is immobilized. This is an
hydrophobic domain may be described by a frame- inherent advantage of such materials over homoge-
work of low-dimensional objects defining the bound- neous electrolytes with mobile anions (or other
ary with the hydrophilic domain. In Nafion, the latter conjugated bases), which may interfere with the
is well-connected, even at low degrees of hydration; reactions that occur at the electrocatalysts. Although
i.e., there are almost no dead-end pockets and very this situation is reminiscent of the simplest case
good percolation. Because of the side-chain architec- discussed in Section 3.1.1.1.1 (excess proton in wa-
ture of Nafion, a third transition region between the ter), it is actually more complex.
aqueous domain and the hydrophobic polymer back- If only electrostatics are considered, significant
bone has been introduced. This comprises the hy- attractive interaction between the excess protons in
drated side chains, and a recent SAXS study seems the aqueous phase and the immobile anionic mirror
to suggest that this region swells at the expense of charge is expected.26,27 Recent MD simulations of
the purely aqueous region with increasing degree of proton transport in slab pores with sulfonate groups
embedded on the pore walls give the anticipated, dielectric constant within the hydrated channels. The
essentially constant separation distance of the pro- distribution of the relative permittivity of the water
tons from the fixed anions, despite changes in the across a hydrated channel has important implications
amount of water between the slabs. The Debye length on the distribution of excess protons within the
(typical electrostatic screening length) of pure water channel. Because the solvation energy of protons
is ∼800 nm at room temperature and is thus much becomes more negative with increasing dielectric
larger than the typical dimensions of the hydrophilic constant, there is a stabilizing effect for protonic
domain (only a few nanometers). Traditionally, the charge carriers in the center of the channels, which
distribution of charge carriers within the correspond- heavily modifies the Gouy-Chapman distribution. As
ing space charge layer is described by the Gouy- illustrated in Figure 8 (top), the decreased dielectric
Chapman theory, which has been developed for semi- constant in the interfacial region leads to a relative
infinite geometries, or by numerically solving the stabilization of the dissociated protons in the central
Poisson-Boltzmann (PB) equation for specific geom- region of the channels. Only at very high water
etries.99,100,212,219,220 In either case, one obtains a contents (i.e., in the two-phase regime of Nafion with
monotonically decreasing concentration of protonic more than 14 water molecules per sulfonic acid
charge carriers as one moves from the hydrophobic/ group) does one expect a slight relative depletion of
hydrophilic interface (i.e., where the anion charge charge carriers in the channel center, which is
resides) toward the center of the hydrated hydrophilic reminiscent of a Gouy-Chapman profile.
domain. However, this picture is not complete, be- The general picture is such that the majority of
cause these continuum theories neglect any struc- excess protons are located in the central part of the
tural inhomogeneity in the vicinity of the electrified hydrated hydrophilic nanochannels. In this region,
interface. In the Gouy-Chapman approach, even a the water is bulklike (for not too low degrees of
homogeneous distribution of the mirror charge over hydration) with local proton transport properties
the interface is assumed; however, the fact that the similar to those described for water in Section
separation of neighboring sulfonic acid groups (∼0.8 3.1.1.1.1. Therefore, the transport properties are
nm) and the typical extension of the hydrophilic indeed a function of the considered length and time
domain (a few nanometers) are of similar order does scales,224,225 and the activation enthalpies of both
not justify this assumption. In addition, recent proton mobility and water diffusion are similar to
Brownian dynamics simulations of ions in cylindrical those of bulk water and only increase slightly with
nanodimensioned pores have shown that both the PB decreasing degree of hydration for intermediate
and Poisson-Nernst-Planck continuum theories sub- water contents (Figure 9).197,224,226-228
stantially overestimate the shielding effects when the Apart from the slight retardation of proton mobility
radius is less than two Debye lengths.221 (Dσ) and water diffusion (DH2O) within the hydrophilic
The same is true for the assumption of a homoge- domain, the decrease in the transport coefficients
neous dielectric constant of the aqueous phase: a with decreasing degree of hydration mainly reflects
simplification that is not backed up by dielectric on the decreasing percolation within the waterlike
measurements as a function of the water content in domain. At the highest degrees of hydration, the
the microwave (i.e., gigahertz) range.222,223 As known major proton conduction mechanism is structure
for the near surface region of bulk water or any diffusion (Dσ > DH2O, Figure 9). With decreasing
interface with water on one side, the dielectric water content, the concentration of excess protons in
constant of the hydrated hydrophilic phase is signifi- the aqueous phase is increasing, which, in turn,
cantly reduced near the hydrophobic/hydrophilic increasingly suppresses intermolecular proton trans-
interface. In addition, the specific interaction of the fer and, therefore, structural diffusion, as witnessed
sulfonic acid group with water (hydration) also in aqueous solutions141 (see Figure 2 and Section
decreases the dielectric constant. Therefore, the 3.1.1.1.1). Consequently, proton mobility at interme-
spatial distribution of the dielectric constant within diate and low degrees of hydration is essentially
hydrated domains is strongly dependent on the width vehicular in nature. Nonequilibrium statistical me-
of the channels (degree of hydration) and the separa- chanics-based calculations (see Section 2.3) of the
tion of the dissociated sulfonic acid functional groups. proton self-diffusion coefficients in Nafion and PEEKK
This is an important result from the equilibrium membranes over a range of hydration conditions have
statistical thermodynamic modeling of the dielectric addressed this conductivity contribution, and they
saturation in different types of hydrated polymers clearly show that the diffusion of water (vehicle) and
described previously (see Section 2.5).47-51 As shown hydrated protons (H3O+) are retarded, as a result of
in Figure 8 (top), the dielectric constant reaches the confinement in an environment where the water and
bulk value (81) in the center of the channel (pore) protons are perturbed by the presence of a substan-
for water contents higher than ∼10 water molecules tial density of sulfonate (i.e., negative charge)
per sulfonic acid group, whereas, for lower degrees groups.45,46 These calculations also showed that struc-
of hydration, even in the center of the channel, the ture diffusion contributes to the diffusion of protons
dielectric constant is lower than the bulk value as a at the higher water contents (i.e., >13 water mol-
consequence of the stronger confinement (Figure 8, ecules per sulfonic acid group; see Section 2.4).20,22,23
bottom). The calculations did not account for specific The very high conductivity activation volumes
chemical water-polymer interactions (modeling only (conductivity decreases with applied pressure
the interaction of the sulfonate groups with the (-kT ∂ ln σ/∂ ln p)), which are particularly evident
water), which are expected to further reduce the in well-separated perfluorosulfonic acid polymers
Figure 8. Hydration isotherm for Nafion 117 (equivalent weight (EW) of 1100 g/equiv)197 and the distribution of the
dielectric constant23 and protonic charge carrier concentration across the hydrated hydrophilic channels (pores) for three
different water contents (top).
phobicandhydratedhydrophobicdomainsare“squeezed”
into each other, forming a more dendritic microstruc-
ture with narrower hydrophilic channels. This view
explains straightforwardly why conductivity activa-
tion volumes in less-separated (better-dispersed)
hydrocarbon-based membranes213 are significantly
smaller. As long as the confinement does not fall
below ∼1 nm, the water at the center of the channels
may still be bulklike (Figure 8), and the diffusion of
hydrated protons has a large hydrodynamic compo-
nent with significant long-range velocity correlations,
typical for viscous media. This feature will be ad-
dressed later in this article to explain the unusually
high electro-osmotic drag coefficients (Section 3.2.1.1).
For very low degrees of hydration (i.e., for Nafion
membranes with <6 water molecules per sulfonic
acid group), the decreasing solvent (water) activity
Figure 9. Proton conductivity diffusion coefficient (mobil- leads to a decreasing dissociation of the sulfonic acid
ity) and water self-diffusion coefficient of Nafion 117 (EW group, i.e., an increasing exclusion of protons from
) 1100 g/equiv), as a function of temperature and the the transport in the aqueous phase. Changes in the
degree of hydration (n ) [H2O]/[-SO3H]).197 neighboring chemical group to the sulfonic acid (i.e.,
changing from a perfluoro methylene to an aromatic
(Nafion, Dow),216 may then also be understood as a carbon: Nafion to S-PEK) will also affect the dis-
confinement effect. Pressure probably leads to an sociation and separation of the proton from the acidic
increasing dispersion of the water, i.e., the hydro- group.20 This effect is dependent not only on the
acidity of the sulfonic acid group, but also on the

dielectric constant of the water of hydration. With
this understanding, the significant decrease of con-
ductivity in the presence of methanol,219 which actu-
ally exhibits similar self-diffusion coefficients as
water (see Section 3.2.1.1) but with a lower dielectric
constant, may then be explained by increased ion
pairing (decreasing dissociation).
3.1.2.2. Adducts of Basic Polymers with Oxo-
acids. To date, the most relevant materials of this
type are adducts (complexes) of polybenzimidazole
(PBI) and phosphoric acid, as introduced by Wain-
right et al.229 In contrast to water, which exhibits a
high mobility for protonic defects but a very low
intrinsic concentration of protonic charge carriers,
phosphoric acid shows both high mobility and a high
concentration of intrinsic protonic defects (see Section
3.1.1.2). In other words, phosphoric acid is intrinsi-
cally a very good proton conductor with a very small Figure 10. Proton conductivity diffusion coefficient (mo-
Debye length, and its charge carrier density is hardly bility) and self-diffusion coefficient of phosphorus for poly-
affected by the interaction with PBI. Indeed, a strong (diallyldimethylammonium-dihydrogenphosphate)-phos-
acid/base reaction occurs between the nonprotonated, phoric acid ((PAMA+H2PO4-)‚nH3PO4), as a function of the
phosphoric acid content.233 Note that the ratio Dσ/DP
basic nitrogen of the PBI repeat unit and the first remains almost constant (see text).
phosphoric acid absorbed. The transfer of one proton
leads to the formation of a benzimidazolium cation
coexist with polyphosphates in heterogeneous gel-like
and a dihydrogenphosphate anion, forming a stable
microstructures.232 Not much is known on the mech-
hydrogen-bonded complex, as shown by infrared
anism of proton transport in polymer adducts with
spectroscopy.230,231 It is a common observation for all
polyphosphates and/or low hydrates of orthophos-
systems of this type that their conductivity strongly phoric acid. The determination of whether the in-
increases upon further addition of an oxo-acid ap- creased conductivity at high water activities is the
proaching the conductivity of the pure acid for high result of the “plasticizing effect” of the water on the
acid concentrations (recently free-standing films of phosphate dynamics and thereby assisting proton
PBI‚nH3PO4 with extremely high acid-to-polymer transfer from one phosphate to the other, or whether
ratios of >10 have been reported).232 In particular, the water is directly involved in the conduction
there is no indication of participation of the polymer mechanism, has not been elucidated.
in the conduction process (also see discussion below).
3.1.2.3. Separated Systems with Immobilized
Although no microstructural information is avail- Proton Solvents. Both types of heterogeneous sys-
able to date, the macroscopic transport has been tems discussed previously are comprised of a poly-
investigated in the related system of poly(diallyldi- meric domain and a low-molecular-weight liquidlike
methylammonium-dihydrogenphosphate) and phos- domain (e.g., H2O, H3PO4) with weak ionic or hydro-
phoric acid, (PAMA+ H2PO4-)‚nH3PO4.233 The proton gen-bond interaction between the two domains. Apart
mobility (Dσ) and the self-diffusion coefficient of from other polar solvents, heterocycles such as imi-
phosphorus (DP), as a measure of the hydrodynamic dazole, pyrazole, and benzimidazole have been in-
diffusion of the system, is shown in Figure 10 for a tercalated into sulfonated polymers,170,171,234,235 re-
given temperature, as a function of the polymer/acid sulting in similar transport properties, as discussed
ratio. Similar to pure phosphoric acid, the mobility for the hydrated systems in Section 3.1.1.1, although
of protonic charge carriers is significantly higher than at somewhat higher temperatures. Apart from the
the self-diffusion coefficient of the phosphate species proton donor and acceptor sites (N), such solvents
and both transport coefficients decrease with increas- contain additional sites (C), which may be used for
ing polymer content virtually in the same manner. covalent “grafting” to oligomeric or polymeric struc-
The main effect, obviously, is just the decreasing tures. If these are hydrophobic (nonpolar), a similar
percolation within the liquidlike portion of the phos- separation as that described in Section 3.1.1.1 may
phoric acid domain, which is reminiscent of the occur, however with covalent bonding bridging the
situation in hydrated acidic polymers (see Figure 6). nonpolar/polar “interface”. This approach has been
At very small acid contents, when all the phosphoric implemented to obtain systems with completely im-
acid is immobilized in the 1:1 complex, only very little mobilized proton solvents that still exhibit high
conductivity is left. proton conductivity with structure diffusion as the
Similar to that observed for pure phosphoric acid, sole proton conduction mechanism (see Figure 3). Of
the transport properties of PBI and phosphoric acid course, the covalent bonding across the nonpolar/
are also dependent on the water activity, i.e., on the polar interface mediates a significant influence of the
degree of condensation (polyphosphate formation) nonpolar portion of the structure on the structure and
and hydrolysis. There is even indication that these dynamics of the polar proton-conducting domain.
reactions do not necessarily lead to thermodynamic This cannot be approximated by simple percolation
equilibrium, and hydrated orthophosphoric acid may effects, as discussed for adducts of basic polymers
with oxo-acids (see also Figure 10), but rather in a

more complex fashion, involving (i) the position and
character of the covalent bonding between the polar
solvent and the nonpolar portion of the structure, (ii)
the softness of this structure, and (iii) the volume
fraction of the polar solvents.
As described in Section 3.1.1.3, the two nitrogens
of the heterocycles act equally as proton donors or
acceptors. Any covalent immobilization must avoid
reduction of this symmetry, which is best achieved
using the carbon between the two nitrogens (i.e., C2
in imidazole or C4 in pyrazole) for covalent bonding.
Interestingly, this is not the case in histidine, which
is an imidazole containing amino acid, which is
frequently involved in proton translocation processes.
However, the energetic asymmetry is very small (∼20
meV) in this particular case.236 The type of bonding
seems to be much more important, i.e., only single
bonds allow reorientation of the bonded heterocycle,
which is a persistent element in the proton conduc-
tion mechanism (see Section 3.1.1.3). To minimize the
constraints in the dynamical aggregation of the
heterocycles, immobilization via flexible spacers, such
as alkanes or ethylene oxide (EO) segments, seems
to be favored.213,237 The optimum spacer length is
then given by the optimum balance between hetero-
cycle aggregation and heterocycle density, on one
hand, and the dynamics of the hydrogen bond net-
work formed by the heterocycles on the other hand.
The compound 2,2′-bis(imidazole), which is a brittle Figure 11. Hydrogen-bonded structure of Imi-2 (two
imidazoles spaced by two ethylene oxide (EO) repeat
solid with a high melting point, is perfectly ag- units): (a) in the crystalline state, as revealed by an X-ray
gregated by strong static hydrogen bonding with structure analysis,239 and (b) in the liquid state (schemati-
negligible proton conductivity (according to recent cal), as suggested by an NMR study.240 Note that the
results from the laboratories of one of the authors). hydrogen bonds in the solid state are long-lived, whereas
Separating the two imidazoles by a soft EO spacer the hydrogen bonding in the molten state is highly dynamic
leads to the appearance of significant proton conduc- (see text).
tivity and a decrease in the melting point and glass-
transition temperature (Tg) with increasing spacer namical hydrogen-bond networks with very high
length.238,239 The conductivity then displays typical proton mobility, although the self-diffusion of the
VTF behavior and, for a given concentration of excess heterocycle is significantly retarded. In the systems
protons (dopant), it is very similar for all spacer described previously, Dσ is typically 1 order of mag-
lengths when plotted versus 1/(T - T0), where T0 is nitude greater than DImi. Recently, fully polymeric
closely related to Tg. For very high spacer lengths, systems with side-chain architectures have been
the dilution of the heterocycles by the spacer seg- developed that still exhibit high proton mobility,
ments has a tendency to reduce aggregation of the despite complete long-range immobilization of imi-
heterocycles and, therefore, once again, reduce proton dazole, i.e., Dσ/DImi ) ∞.241 This finding is of para-
mobility. mount importance, because it demonstrates that
The aggregation of imidazole leading to a continu- complete decoupling of the long-range transport of
ous hydrogen-bonded structure in crystalline Imi-2 protons and heterocycles (the proton solvent) is
(two imidazole units spaced by two EO repeat units) possible. The reader may recall that complexation of
is shown in Figure 11a.238 Upon melting, the situation phosphoric acid and a basic polymer does not show
in most parts of the material is more similar to that any signature of this effect (see Figure 10), which
shown in Figure 11b. This is one result coming from may open the way to the development of true single
an NMR study240 demonstrating that liquid Imi-2 ion conductors.
exhibits ordered domains (similar to the crystalline One of the problems associated with the use of
form), a dynamically disordered but still aggregated heterocycles is that the intrinsic concentration of
domain, and a certain fraction of nonbonded mol- protonic charge carriers can only be moderately
ecules. It is only within the disordered domain increased through acid doping (also see Section 4.3).
(Figure 11b) that fast proton mobility is observed, This is particularly the case when the dynamics
again demonstrating the delicate balance of aggrega- within the hydrogen-bonded domain is highly con-
tion and dynamics in hydrogen-bonded structures strained through immobilization (especially in fully
with high proton mobility. polymeric systems), which is probably the direct
Another interesting observation is that hetero- consequence of a reduced dielectric constant. This
cycles immobilized in this way may still form dy- also leads to a further increase of the Haven ratio
DH/Dσ, as discussed in Section 3.1.1.3 (also see Figure transport in plain ionomers (see Sections 3.1.2.1 and
4). 3.2.1). Particularly, in the case of proton-conducting
Also note that the spacer concept, as verified by zirconium phosphate prepared via in situ growth
several laboratories (e.g., see Perrson and Jann- within the preformed membrane,253 the proton con-
asch242), has been moved forward to the immobiliza- ductivity of the highly dispersed filler may have some
tion of other proton solvents such as phosphonic acid, significance at high temperature and low humidity,
with very promising results (see Section 5). where the conductivity of pure Nafion strongly
decreases.245
3.1.2.4. Composites. Because of the drawbacks in
hydrated acidic polymers for application in fuel cells,
composites with highly dispersed inorganic phases
3.2. Mechanisms of Parasitic Transport
have also been investigated, and, indeed, certain The transport of protonic charge carriers is some-
systems show improved performance in operating times inherently connected to the transport of other
fuel cells.243,244 The extent that this is due to a species (e.g., in hydrated acidic polymers, Section
modification of the bulk transport properties (in 3.1.2.1). Sometimes, there are just indirect mecha-
particular, the proton conduction mechanism) or of nistic relationships, or the existence of completely
the membrane/electrode interface has not been con- independent transport paths (e.g., protonic charge
clusively determined. However, the similarities of the carriers and electronic holes in oxides). Parasitic
bulk transport properties of some modified and transport frequently limits the fuel-cell performance,
unmodified ionomers (see Section 4.1) are indicative and a mechanistic understanding is definitely useful
of the possible relevance of interfacial effects. In view in the development of separator materials.
of the attention such systems are attracting at the The advantageously low methanol “crossover” of
moment, some general aspects are summarized here PBI-H3PO4 adducts is well-established and the
(for a review on the different type of systems, see, problems related to its low oxygen transport are also
e.g., refs 245 and 246). Such composites are usually well-described.254 However, the fundamentals of these
formed by either dispersing inorganic particles in a transport phenomena have not been subjected to any
solution of the ionomer, followed by film casting and systematic investigation. For the recently emerging
solvent elimination, or by precipitation of inorganic systems based on heterocycles and other proton
particles within a pre-existing membrane. The latter solvents (also see Section 5), the investigation of
procedure was first used by Mauritz247 to precipitate parasitic transport is only just beginning. On the
SiO2 particles in Nafion using in situ hydrolysis of other hand, there is extensive literature on parasitic
tetraethoxysilane (TEOS). Although SiO2 does not transport in the widely used hydrated acidic iono-
show measurable proton conductivity itself, other mers (especially Nafion). In addition, the transport
inorganic compounds, such as diverse zirconium properties of oxides have also been intensely inves-
phosphates and phosphonates or heteropolyacids,245,248 tigated by groups mainly outside the fuel-cell com-
have recently been used as fillers. Although these munity.
materials possess some intrinsic proton conductivity, Therefore, the following discussion of the mecha-
it is much lower than that of the ionomeric host (e.g., nisms of parasitic transport is restricted to only these
Nafion) at low temperatures and high humidities. two classes of materials.
This is also true for the hydrophilicity (“water-
retaining properties”) of such inorganic particles, 3.2.1. Solvated Acidic Polymers
which is frequently thought to be responsible for the As discussed in Section 3.1.2.1, proton transport
favorable performance of the corresponding compos- in acidic polymers occurs within a system of hydrated
ites in fuel cells.234,249 There is actually no experi- and connected hydrophilic pores (channels) within
mental evidence for this, and the strong acidity of the nanoseparated materials. The extension of the
perfluorosulfonic acid polymers is not expected to be channels, which, of course, shows some distribution
exceeded by any of the aforementioned inorganics. within a given sample, may range from <1 nm to
Of course, this does not hold for the water uptake in several nanometers, depending on the type of poly-
contact with liquid water, which is sometimes taken mer and the degree of solvation (swelling). This range
as a measure of the water-retaining properties.250-252 covers the transition from more solidlike behavior
The higher water uptake under these conditions may with diffusive transport to a liquidlike regime with
be the result of the reduction of the tensile strength additional hydrodynamic transport (viscous flow).
of the polymer in the composite, allowing more water Apart from protonic defects, the main species that
to enter as a second phase (which is actually not are confined to the hydrophilic domains are water
retained by any specific interaction). The most im- and methanol (used as the fuel in DMFCs). The
portant effect of filling an ionomer with inorganic transport of these species is highly correlated and
particles is probably the modification of the membrane/ therefore discussed together in the following section.
electrode interface (see above), the microstructure The transport mechanisms of fuel-cell relevant gases
and the elastic properties of the ionomeric component (O2, H2), which may dissolve in the water domains
(with a glass transition temperature of Tg ≈ 120 °C and also in other parts of the nanostructures, are also
for plain Nafion, the latter may be a specific advan- reviewed in this section.
tage of the use of composite membranes in this 3.2.1.1. Mechanisms of Solvent (Water, Metha-
temperature range). These changes may also alter nol) Transport. The following types of transport are
the transport properties in a similar way as discussed considered in this section: (i) self-diffusion or tracer
for the relationships between microstructure and diffusion of solvent molecules, which is the unidirec-
Figure 12. Water self-diffusion coefficient of Nafion 117

(EW ) 1100 g/equiv), as a function of the water volume
fraction XV and the water diffusion coefficient obtained
from a Monte Carlo (MC) simulation (see text). The proton
conductivity diffusion coefficient (mobility) is given for
comparison. The corresponding data points are displayed
in Figure 14.
tional transport of “marked species” (e.g., this may

be an isotopic tracer or the spin-labeled nuclei in PFG
NMR experiments) with the thermal energy as the
only driving force; (ii) chemical or Fickian diffusion,
which is the average drift of a component i in a
chemical potential (approximately proportional to Figure 13. A few microstructural parameters for Nafion
and sulfonated poly(arylene ether ketone)s,174,208-213 as a
ln ai or ln pi) gradient of the particular component function of the solvent (water and/or methanol) volume
related to the concentration gradient via the ther- fraction XV: (a) the internal hydrophobic/hydrophilic in-
modynamical factor (d(ln ai)/d(ln ci)); and (iii) per- terface, and (b) the average hydrophobic/hydrophilic sepa-
meation, which is the flow of a component (frequently ration and the diameter of the solvated hydrophilic chan-
solvent) in a total pressure gradient. Following this, nels (pores).
the mechanisms of electro-osmotic drag are reviewed.
This is the coupled transport of a protonic species and lower diffusion coefficients that correspond to the
solvent molecules, i.e., the transport of neutral lower self-diffusion coefficient in bulk methanol
solvent in an electrical field with a vanishing gradient compared to that in bulk water. At low water
in the chemical potential. This phenomenon is sym- contents, however, the experimental diffusion coef-
metrical to the transport of protons in a chemical ficient falls off more rapidly than the simulated one
potential gradient of solvent in the absence of any (see Figure 12). One reason is simply the fact that
electrical field. The coupling of water and methanol the connectivity within the solvated hydrophilic
transport has not been studied explicitly; however, domains decreases (appearance of dead-end channels,
there is indirect indication of the nature of this pockets).212,213 Another reason is the retardation of
coupling, which will also be discussed. water diffusion as a result of increasing confinement,
The self-diffusion coefficient of water in Nafion, as which increases the internal molecular friction within
measured by PFG NMR,197,224,226,255-261 is shown in the solvated domain as discussed above (see Sections
Figure 12, as a function of the water volume fraction. 3.4 and 3.1.2.1).41-46 A comparison of Figures 12 and
At high degrees of solvation, this is only slightly lower 13 suggests that this effect becomes particularly
than the diffusion coefficient obtained from MC evident when the diameter of the channel decreases
simulations of random walk transport within nano- to <1 nm.
structures parametrized by results of SAXS experi- For comparison purposes, the proton mobility, Dσ
ments174,212,213 (see Figure 13) assuming the absence (for Nafion solvated with water), which is closely
of any dead-end channels and a local diffusion related to the self-diffusion coefficient of water, is also
coefficient that is identical to the bulk diffusion plotted. At low degrees of hydration, where only
coefficient of water (2.25 × 10-5 cm2/s) at any position hydrated protons (e.g., H3O+) are mobile, it has a
within the well-connected hydrophilic domain. This tendency to fall below the water diffusion coefficient
finding suggests that the major reason for a decrease (this effect is even more pronounced in other poly-
in the water diffusion coefficient with decreasing mers), which may be due to the stiffening of the water
water content is the decreasing percolation within the structure within the regions that contain excess
hydrophilic domain, especially at high water contents protons, as discussed in Section 3.1.1.1.148 Interest-
(also see Section 3.1.2.1). This is also supported by ingly, the proton mobility in Nafion solvated with
the observation that methanol diffusion follows the methanol (Dσ(MeOH) in Figure 14a) is even lower
same trend (Figure 14a),174,262,263 but with somewhat than the methanol self-diffusion (DMeOH). This may
per sulfonic acid), any driving force for Fickian water

diffusion vanishes, i.e., there is no chemical diffusion.
A fitting expression of D̃H2O is given in ref 264; D̃H2O
data, calculated from DH2O data with thermodynamic
factors obtained from the hydration isotherm (Figure
8), are plotted in Figure 14a. There are very few
direct measurements of D̃H2O for water available.
Frequently, pervaporation-type experiments are con-
ducted, where the membrane is exposed to liquid
water (methanol) on one side and dry flowing nitro-
gen on the other side.265 A chemical diffusion coef-
ficient may be calculated from the rate of water
(methanol) uptake. However, because of the extreme
solvent chemical potential difference across the mem-
brane and the thermodynamically poorly defined
situation on the gas side, only average, poorly repro-
ducible chemical diffusion coefficients are obtained;
i.e., any information on the concentration dependence
is lost. Another complication with these experiments
is that the total pressure gradient across the mem-
brane induces permeation, which may have contribu-
tions from additional transport mechanisms, as will
be discussed later. Particularly, in the case of mea-
suring the chemical diffusion of methanol, this prob-
lem is elegantly avoided through the use of aqueous
solutions of different methanol concentration (activi-
ties) on both sides.266 Many other techniques, which
are frequently governed by different water (metha-
nol) transport mechanisms (in a complex fashion),
have been reviewed by Doyle and Rajendran.264
The driving force for solvent permeation is the total
pressure gradient across the membrane, which is
related to the chemical potential gradient, according
to
Figure 14. Solvent (water, methanol) diffusion coefficients

of (a) Nafion 117 (EW ) 1100 g/equiv) and (b) sulfonated
∇µ
bs ) ( )
∂µs
∂p
b ) Vm
∇p s ∇p
b (24)
poly(arylene ether ketone)s, as a function of the solvent
volume fraction. Self-diffusion data (DH2O, DMeOH) are taken which leads to the following relation between per-
from refs 197, 224, 226, 255-263 and unpublished data meation and diffusion:212
from the laboratory of one of the authors); chemical
( )
diffusion coefficients (D̃H2O) are calculated from self-diffu-
RT Ps Kdrag σ
sion coefficients (see text), and permeation diffusion coef- DPs ) + (25)
ficients are determined from permeation coefficients.212,267,268 cs Vm F2
s
Proton conductivity diffusion coefficients for hydrated
samples (Dσ(H2O))197,224,226,255-261 and samples solvated with
methanol (Dσ(MeOH)256) are given for comparison. where Ps is the permeation coefficient and Kdrag is the
electro-osmotic drag coefficient (discussed below). The
most comprehensive water permeation study for
be the consequence of the lower dielectric constant Nafion, as a function of water content and temper-
of methanol, compared to water, especially under the ature, was reported by LaConti et al.267 and Meier
condition of confinement in a perfluorinated ionomer et al.268 The diffusion coefficients DPs (when adjusted
(see discussion in Section 3.1.2.1 and Figure 8). to room temperature) are ∼1 order of magnitude
When it comes to the equilibration of water con- higher than the diffusion coefficients obtained by
centration gradients, the relevant transport coef- NMR techniques (Figure 14a). The “permeation dif-
ficient is the chemical diffusion coefficient, D̃H2O. This fusion coefficient” has a tendency to approach the
parameter is related to the self-diffusion coefficient self-diffusion coefficient only at low water contents.
by the thermodynamic factor (see above) if the This behavior is not expected if the elementary
elementary transport mechanism is assumed to be mechanisms of the two types of transport are the
the same. The hydration isotherm (see Figure 8) same (i.e., quasi-random walk processes exploring the
directly provides the driving force for chemical water solvent chemical potential distribution in space).
diffusion. Under fuel-cell conditions, i.e., high degrees Obviously, there is an additional transport compo-
of hydration, the concentration of water in the nent in the permeation mechanism. This is most
membrane may change with only a small variation likely viscous flow far away from local thermody-
of the chemical potential of water. In the two-phase namic equilibrium, which may also comprise a cer-
region (i.e., water contents of >14 water molecules tain “slip” at the interface with the hydrophobic
domain, which is a well-established phenomenon in

hydrodynamics. This Hagen-Poiseuille-type compo-
nent in the flow is, of course, expected to be strongly
dependent on the diameter of the solvated “channels”
(see Figure 13). The observation that this component
apparently decreases as the water content decreases
and virtually disappears in membranes with very
narrow channels and strong polymer-solvent inter-
actions (see Figure 14b and Section 4.1) is consistent
with this interpretation. Hence, the mechanism of
permeation in solvated acidic membranes may be
described by a combination of diffusional processes
and viscous flow, where the latter contribution
increases as the solvent-polymer interaction de-
creases and the “channel” width (i.e., the degree of
hydration and hydrophilic/hydrophobic separation)
increases.
The aforementioned view provides a rational for
the distinct differences of the solvent transport
coefficients of Nafion and solvated sulfonated pol-
yarylenes (Figure 14b), which are generally less-
separated and exhibit stronger polymer-solvent
interactions.
Although solvents such as water and methanol are
virtually uncharged, the application of an electrical
field may induce a drift velocity of such neutral
species when solvating an acidic ionomer. This
phenomenon, which is known as electro-osmotic drag,
is the consequence of the interaction of such solvent
molecules with ions they solvate, in particular,
protons in the case of PEMs. The classical mecha-
nistic theory of electro-osmosis dates back to the time
of Helmholtz,269 Lamb,270 Perrin,271 and Smoluchows- Figure 15. Electro-osmotic drag coefficients Kdrag of (a)
ki,272 who assumed that transport occurs only close Nafion 117 (EW ) 1100 g/equiv) and (b) sulfonated poly-
to the wall in electrical double layers of low charge- (arylene ether ketone)s, as a function of the solvent (water
carrier concentration and with extensions signifi- and/or methanol) volume fraction XV.174,212,219,274-281 The
cantly smaller than the pore (channel) diameter. The normalized drag coefficients for water and methanol are
virtually identical; therefore, both are plotted together.
corresponding theories qualitatively describe the
electro-osmotic drag in wide pore systems, such as model describes the variation of the electro-osmotic
clay plugs; however, as discussed in Section 3.1.2.3 drag within model membranes (polystyrene-based) of
(also see Figure 8), both model assumptions are not different ionic forms surprisingly well.
valid for typical PEM materials such as Nafion. The The data were obtained for given degrees of hydra-
width of the hydrated (solvated) channels is orders tion. Also, for PEM fuel-cell membranes, experimen-
of magnitude smaller than the Debye length of water tal electro-osmotic drag coefficients are generally
and the concentration of charge carriers is very high reported only over narrow solvation ranges. Typical
(typically ∼5 M within the hydrophilic domain). For drag coefficients in the range of 1-2.5 at room
this type of system, Breslau and Miller developed a temperature (expressed in the number of water
model for electro-osmosis from a hydrodynamic point molecules per transported protonic charge carrier)
of view.273 The model treats the ions as spherical were sometimes identified with the number of water
particles moving in a continuous viscous medium, molecules solvating the excess proton. Recently,
with the membrane matrix forming the boundary of electro-osmotic drag coefficients became accessible by
the medium. The hydration of the ions is treated by electrophoretic NMR174,212,219,274 for a wide range of
an adequate choice of a hydrated ion radius, stripping polymer-solvent volume ratios, and the results
off the hydration sphere in narrow channels, and the clearly confirm the hydrodynamic nature of electro-
water structure forming and breaking properties of osmosis, particularly at high degrees of solvation. The
diverse ions are considered. Channel diameters were data174 presented in Figure 15, together with data
not directly accessible by the authors; therefore, the from other authors,275-281 essentially show two
electro-osmotic transport of tetraalkylammonium- things: (i) at low degrees of hydration, the electro-
exchanged membranes were used to determine an osmotic drag coefficient approaches a value of 1 but
“effective” pore size. Although the assumptions of the does not fall below this value; and (ii) with increasing
model are severe simplifications of the situation in solvent fraction (channel width, Figure 13), the drag
real membranes (heterogeneity of charge and viscos- coefficient dramatically increases and reaches ∼50%
ity distribution (also see Figure 8) and channel of the maximum possible value (dashed line), which
widths of the extension of only a few molecules), the corresponds to an identical drift velocity of all solvent
molecules and protonic charge carriers. Considering 3.2.1.2. Mechanisms of Dissolved Gas Trans-
the fact that, at high degrees of hydration, ap- port. As noted in Section 1, one of the cardinal
proximately half of the conductivity is carried by properties of fuel-cell membrane materials is to be
structure diffusion (see Section 3.1.2.1), i.e., the able to effectively separate the reactive masses, in
proton mobility (proton self-diffusion coefficient) is particular, the anode and cathode gases. Accordingly,
about twice the water self-diffusion coefficient, almost there have been numerous investigations, especially
all water molecules appear to drift at approximately on oxygen and hydrogen transport across PEMs, and
identical velocity (about half of the drift velocity of because the current qualitative mechanistic under-
protonic charge carriers) in extremely swollen samples. standing has already existed for some time, only a
This situation corresponds to minor relative motion brief summary is given here.
of water molecules with respect to each other, i.e., It already had been recognized by Yeo and
the transport is clearly of a collective nature. The McBreen282 that gas (hydrogen and halogens) per-
decrease of the drag coefficient with decreasing water meation rates in Nafion are strongly related to water
content approximately scales with the fourth power content. The data indicated that most of the gas
of the channel diameter, which is reminescent of transport is occurring within the solvated hydrophilic
Hagen-Poiseuille-type behavior with continuously domain at rates only slightly lower than these of
increasing “stripping off” of the water molecules. This water self-diffusion. Ogumi and co-workers283,284 later
stripping comes to an end at low degrees of hydration, reported that oxygen diffusion in different hydrated
where the motion of one water molecule remains membranes are indeed very similar, with the gas
strongly coupled to the motion of the excess proton solubilities differing significantly. These results made
(Kdrag ≈ 1; see Figure 15). This is also expected from already clear that gases dissolve in both the hydro-
phobic and the solvated hydrophilic domain, whereas
the high enthalpy of primary hydration (stability of
most of the gas transport occurs within the solvated
H3O+) and the proton conduction mechanism (Section
hydrophilic domain. This view is supported by many
3.2.1.2), i.e., diffusion of H3O+ (vehicle mechanism).
experimental studies on Nafion285-287 and other types
Extremely high degrees of swelling were actually of membranes.288,289 When hydrated, they all show
obtained in water/methanol mixtures; and the drag comparable diffusion and permeation rates for oxygen
coefficients of water and methanol were determined and hydrogen, and it was not surprising that signifi-
separately by isotopic labeling (H/D substitution). cantly lower gas transport rates were found for
Interestingly, the normalized drag coefficients (drag Nafion when water was replaced by phosphoric
coefficient Ki divided by the mole fraction of the acid290 or when Nafion was dried.291 In the latter case,
corresponding species i) are virtually identical for the gas diffusion rates were reported to be similar to
both species in the mixtures (also see Figure 15). those of plain polytetrafluoroethane (PTFE). Büchi
Because water and methanol form almost-ideal solu- et al.292 actually separated the oxygen solubility and
tions with comparable water-water, water-metha- diffusion as a function of temperature for different
nol, and methanol-methanol interactions, this be- perfluorinated proton exchange membranes of dif-
havior provides additional support for a hydrodynamic ferent equivalent weights. Apart from giving further
process. For low degrees of hydration, where electro- support to the existing mechanistic understanding,
osmotic drag becomes diffusional in nature, it is not this work showed that the oxygen solubility is
yet clear, whether there is some preferential solvation decreasing while oxygen permeation is increasing
of the excess proton by water or methanol (primary with temperature.
hydration). Figure 15b shows that the electro-osmotic
drag for sulfonated poly(arylene ether ketone)s is 3.2.2. Oxides
generally lower than that for Nafion, which is partly From the formation reaction of protonic defects in
due to smaller channels. However, even for compa- oxides (eq 23), it is evident that protonic defects
rable channel dimensions, the values for electro- coexist with oxide ion vacancies, where the ratio of
osmotic drag are still somewhat lower, which is an their concentrations is dependent on temperature
indication of the influence of the solvent-polymer and water partial pressure. The formation of protonic
interaction on the total electro-osmotic drag. This is defects actually requires the uptake of water from
also supported by the observation that electro-osmotic the environment and the transport of water within
drag is generally increasing with increasing temper- the oxide lattice. Of course, water does not diffuse
ature (see inserts in Figures 15a and b), whereas the as such, but rather, as a result of the ambipolar
solvent-polymer interaction is decreasing. diffusion of protonic defects (OH•O) and oxide ion
There is no quantitative model yet describing the vacancies (V••O). Assuming ideal behavior of the in-
observed electro-osmotic drag coefficients as a func- volved defects (an activity coefficient of unity) the
tion of the degree of hydration and temperature. chemical (Fick’s) diffusion coefficient of water is
However, the available data provide strong evidence
for a mechanism that is (i) hydrodynamic in the high (2 - X)DOH•ODV••O
solvation limit, with the dimensions of the solvated D̃H2O ) (26)
hydrophilic domain and the solvent-polymer inter- XDOH•O + 2(1 - X)DV••O
action as the major parameters; and (ii) diffusive at
low degrees of solvation, where the excess proton where X is the degree of hydration.293 Under com-
essentially drags its primary solvation shell (e.g., pletely dry conditions, i.e., in the absence of proton
H3O+). conductivity, D̃H2O equals the self-diffusion coefficient
been obtained under thermodynamically defined

conditions, preferably those close to thermodynamic
equilibrium, i.e., with relatively small perturbations
(electrical potential gradients, chemical potential
gradients, etc.) applied to the samples. Although far
from complete, the diversity of the materials included
in this review allow for some general qualitative
conclusions (see Section 5). The transport coefficients
presented may also allow calculation (or numerical
simulation) of fluxes and concentrations for nonequi-
librium situations, provided that linear response may
be assumed. The purpose of this section is to sum-
marize the transport coefficients of some typical and
well-characterized proton conductors without giving
any mechanistic explanations, which already have
been discussed in Section 3 for prototypical com-
pounds, and also provide rational for a qualitative
Figure 16. Self-diffusion coefficient of oxide ion vacancies understanding of the phenomenological data pre-
in different perovskite-type oxides,188 which equals the
chemical water diffusion coefficient in the fully hydrated
sented in this section.
state (see text). (Figure reproduced with the kind permis-
sion from Elsevier.) 4.1. Hydrated Acidic Polymers
of protonic defects. However, near the hydration limit For the application of acidic polymers in fuel cells,
(i.e., X close to 1), D̃H2O approaches the self-diffusion the most relevant species, for which transport must
coefficient of oxide ion vacancies, which is usually be considered, are protonic charge carriers, water,
significantly lower than the diffusion coefficient of methanol, hydrogen, and oxygen. As discussed in
protonic defects. Because the latter case relates to Section 3, the transport of all these species predomi-
the conditions in a fuel cell, the self-diffusion coef- nantly occurs in the same environment, i.e., the
ficient of oxide ion vacancies, which is shown in solvated hydrophilic domain of the material. In
Figure 16 for a variety of perovskite-type oxides, may addition to the transport coefficients for each of the
be taken as a measure of D̃H2O under these conditions. five species, one must also consider cross coefficients
For a given water partial pressure, D̃H2O increases describing the coupled transport of different species.
as temperature increases, because of dehydration and For a linear description, this already requires a 5 ×
the increasing self-diffusion coefficients for both 5 matrix with 15 independent coefficients. Together
involved species. Recently, there has been some with the publications referenced in Section 3.2.1.2,
controversy about the validity of the model;294 how- two more recent studies299,300 provide data and fur-
ever, it is indeed strictly valid in the entire hydration ther literature on gas (predominately O2 and H2)
range under the given assumptions.295 transport in solvated acidic membranes. The repro-
It should be mentioned that, apart from oxide ion ducibility of these data is somewhat less than that
vacancies existing at low water partial pressures, the for water transport; however, the general behavior
appearance of electronic holes (h•) must be considered is such that O2 diffusion is closely related to water
at high oxygen activities. In oxides, they are fre- diffusion, including its dependence on solvation and
quently localized on the oxygen (O2- + h• h O-) and temperature, whereas the H2 diffusion is generally
may be very mobile as large polarons. The ambipolar higher (by factors of 2-4). Because the gas solubility
diffusion, together with protonic defects, then permits within the hydrophilic domain is quite low (i.e., on
chemical diffusion of hydrogen; and the ambipolar the order of 10-6 mol cm-3 for O2 within the hydrated
diffusion, together with oxide ion vacancies, allows domain of Nafion at ambient conditions) and de-
chemical diffusion of oxygen. However, for large- creases with temperature, the gas permeability re-
band-gap oxides relevant for fuel-cell applications, mains quite low (for O2 on the order of 10-11 mol cm-1
this may not be important at temperatures <750 s-1 in Nafion) and does not increase with temperature
°C.296-298 to the same degree as the gas and solvent diffusion
coefficients.
4. Phenomenology of Transport in The following compilation is restricted to the
Proton-Conducting Materials for Fuel-Cell transport coefficients of protonic charge carriers,
water, and methanol. These may be represented by
Applications a 3 × 3 matrix with six independent elements if it is
Transport data have been reported for a huge assumed that there is just one mechanism for the
number of proton-conducting materials, some of transport of each species and their couplings. How-
which are defined rather poorly. They have been ever, as discussed in Sections 3.1.2.1 and 3.2.1,
recorded by a variety of methods under diverse different types of transport occur, i.e., diffusive
conditions including ill-defined transient or steady- transport as usually observed in the solid state and
state situations (e.g., in a running fuel cell). Conse- additional hydrodynamic transport (viscous flow),
quently, reproducibility and comparison of data from especially at high degrees of solvation. Assuming that
different laboratories is problematic. For this reason, the total fluxes are simply the sum of diffusive and
we have only compiled the transport data that have hydrodynamic components, the transport matrix may
be extended by three more phenomenological ele-

ments describing (i) the hydrodynamic components
of proton transport coupled to water or methanol
permeation, and (ii) the water, and methanol, per-
meations themselves. Of course, the description of
transport with a 3 × 4 matrix is neither complete
nor consistent; however, it is a comfortable way to
describe transport in solvated acidic membranes
phenomenologically with reasonable precision. The
)( )
transport equation may then be written
( )(
∇µ̃H+
bj H+ L11 L12 L13 L14
∇µ
bH2O
bj H2O ) L21 L22 L23 L24
∇µ
bMeOH
bj MeOH L31 L32 L33 L34
∇p
btotal
with Lij ) Lji (27) Figure 17. Room-temperature proton conductivity of two
Dow membranes226,255,260 of different EW values, Nafion,
two varieties of sulfonated poly(arylene ether ketone)s (S-
with the driving forces being the gradients of PEK and S-PEEKK, unpublished data from the laboratory
0 m
of one of the authors), and sulfonated poly(phenoxyphos-
µ̃H+ ) µH + + RT ln aH+ + VH+ptotal + zH+FΦ (28a) phazene)s (S-POPs301) of different equivalent weights (685
and 833 g/equiv), as a function of the degree of hydration
0 m n ) [H2O]/[-SO3H] (number below the compound acronym/
µH2O ) µH 2O
+ RT ln aH2O + VH p
2O total
(28b)
name indicates the EW value).
µMeOH ) µ0MeOH + RT ln aMeOH + Vm
MeOHptotal (28c) Usually, the total permeation coefficient (PH2O) is
measured, and this, expressed as a permeation
The longitudinal coefficient L11 is related to the diffusion coefficient DH P
, is given in the following
proton conductivity σ, according to 2O
data compilation.
σ There are actually no experimental measurements
L11 ) (29) of protonic streaming currents (L14) and coupled
F2 water and methanol transport (L23 ) L32); however,
L22 and L33 are related to the self-diffusion and the first may be related to the hydrodynamic com-
chemical diffusion coefficients of water and methanol: ponent of the electro-osmotic drag (L12/L11, L13/L11)
(see discussion in Section 3.2.1.1). The second is
( )
2
cH d(ln aH2O) cH2O expected to be qualitatively related to the ratio of the
2O
L22 ) DH2O ) D̃H2O (30) electro-osmotic drag coefficient of water and metha-
RT dcH2O RT nol (L12/L13). In the following, the directly accessible
P
( )
transport coefficients (σ (Dσ), DH2O, DMeOH, DH 2O
,
c2MeOH d(ln aMeOH) cMeOH DPMeOH, KH2O, and KMeOH) for different solvated acidic
L33 ) DMeOH ) D̃ polymers are presented in a way that allows some
RT dcMeOH RT MeOH
(31) interesting comparisons and the calculation or esti-
mation of the elements of the transport matrix Lij.
The cross coefficients are contained in the electro- In many publications, these transport parameters are
osmotic drag coefficients: reported as a function of the solvent content and are
expressed as the number of solvent molecules (i.e.,
L21 water) per sulfonic acid group. Because of the im-
KH2O ) (32)
L11 portance of percolation effects in all considered
transport coefficients, we have converted these sol-
L31 vent contents to solvent volume fractions, except for
KMeOH ) (33) proton conductivities, as shown in Figures 17 and 18.
L11
Figure 17 shows the room-temperature proton
The coefficients L14, L24, and L34 describe the viscous conductivity of two sulfonated poly(arylene ether
flow contributions of the transport of all three species ketone)s, sulfonated poly(phenoxyphosphazene)s (S-
in a total pressure gradient ∇p btotal. Because a pres- POP) of different equivalent weights,301-303 Nafion
sure gradient also imposes a chemical potential (unpublished data from the laboratory of one of the
gradient on each species (eq 24), experimentally, authors), and two Dow membranes of different equiva-
there is always a superposition of diffusive and lent weight.226,255,260 Except for one Dow membrane
viscous flow; e.g., for the description of the water flux with an equivalent weight of 800 g/equiv, the other
in a total pressure gradient, all coefficients must be membranes have similar (volume) densities of sul-
included, i.e., fonic acid groups, i.e., similar water volume fractions
for a given water content n ) [H2O]/[-SO3H]. It is a
bj H2O ) L21∇µ̃H+ + L22∇µ
bH2O + L23∇µ
bMeOH + general observation that hydrocarbon membranes
require more water to achieve similar conductivities
L24∇p
btotal ) PH2O∇p
btotal (34) than perfluorosulfonic acid polymers. This effect is
Figure 19. Transport coefficients of diverse membranes

based on perfluorinated polymers (Dow226,255,260 and Nafion/
silica composites174,243,244,304-306), polyarylenes (S-PEK/PSU
blends, ionically cross-linked S-PEK/PBI),307-315 and sul-
fonated poly(phenoxyphosphazene)s (S-POPs),301 as a
function of the water volume fraction XV. Lines represent
data for Nafion and S-PEK (given for comparison); for data
points, see Figure 14.
Figure 18. Proton conductivity of (a) Nafion 117 (EW )

1100 g/equiv) and (b) a sulfonated poly(arylene ether
ketone), as a function of temperature and degree of
hydration (n ) [H2O]/[-SO3H]).197
most pronounced for the high-equivalent-weight

S-POP 833, and, considering the relation between
solvent diffusion and proton mobility, this observa-
tion is in accordance with the low solvent diffusion
coefficients reported for this type of membrane.302 To
reach sufficiently high conductivities for fuel-cell Figure 20. Electro-osmotic drag coefficients of diverse
applications, such membranes must be in contact membranes based on perfluorinated polymers (Dow226,276
with liquid water.219 and Nafion/silica composites174) and polyarylenes (S-PEK/
Figure 18 shows the temperature dependence of PSU blends, ionically cross-linked S-PEK/PBI174), as a
the proton conductivity of Nafion and one variety of function of the solvent (water/methanol) volume fraction
XV (see text for references). Lines represent data for Nafion
a sulfonated poly(arylene ether ketone) (unpublished and S-PEK (given for comparison); for data points, see
data from the laboratory of one of the authors). The Figure 15. Dashed lines correspond to the maximum
transport properties of the two materials are typical possible electro-osmotic drag coefficients for water and
for these classes of membrane materials, based on methanol, as indicated (see text).
perfluorinated and hydrocarbon polymers. This is
P (Figure 20; for this figure, data for the Dow mem-
clear from a compilation of Dσ, DH2O, and DH 2O
data
brane are taken from refs 226 and 276; all other data
for a variety of membrane materials, including Dow
are from the laboratory of one of the authors), which
membranes of different equivalent weights,226,255,260
suggests that the major parameter that controls
Nafion/SiO2 composites243,244,304-306 (including unpub-
electro-osmotic transport is simply the degree of
lished data from the laboratory of one of the authors),
swelling and the choice of the type of backbone,
cross-linked polyarylenes,307-315 and sulfonated poly-
particularly for low degrees of swelling (perfluori-
(phenoxyphosphazenes)301 (Figure 19). The data points
nated versus hydrocarbon polymer).
all center around the curves for Nafion and S-PEK,
indicating essentially universal transport behavior
for the two classes of membrane materials (only for
4.2. PBI−H3PO4 Adducts
S-POP are the transport coefficients somewhat The transport data of PBI-H3PO4 have extensively
lower, suggesting a more reduced percolation in this been reviewed in two recent papers,316,317 and their
particular material). This correlation is also true for hybrids with inorganic particles have been character-
the electro-osmotic drag coefficients KH2O and KMeOH ized in another.318 The gas permeability is much
Nafion234 with intercalated heterocycles. Conductivi-

ties of >10-2 S/cm are easily accessible by such
systems; however, the volatility of the heterocycles
requires some immobilization. The effect of covalent
immobilization on proton conductivity is shown in
Figure 23 for imidazole-based systems.213,237-239,241,242,319
It is important to realize that these oligomers have
similarly low vapor pressures as ionic liquids. To
illustrate the effect of extrinsic charge carrier forma-
tion, the evolution of proton conductivity with triflic
acid doping is shown for the oligomeric system Imi-2
in Figure 24.237 For such systems, the mobility of
protonic charge carriers is typically 1 order of mag-
nitude higher than the self-diffusion of the oligomers
as a whole. Of course, long-range solvent (imidazole,
pyrazole, benzimidazole) transport does not occur in
Figure 21. Proton conductivity of PBI‚nH3PO4 adducts, fully polymeric systems. Cyclovoltammetry and fuel-
as a function of phosphoric acid concentration and relative cell-type experiments demonstrate that there is at
humidity (RH).317 Data from another source (denoted by least some transport of oxygen and hydrogen in
the dashed line)316 are given for comparison. imidazole-based systems.320
4.4. Oxides
Apart form protonic defects, the only species that
may be transported in proton-conducting wide-band-
gap perovskite-type oxides at significant rates are
oxide ion vacancies. Although this allows for some
chemical water diffusion in the intermediate tem-
perature range (typically 300-700 °C, see Section
3.2.2), hydrogen and oxygen diffusion only occurs
under either highly reducing or oxidizing conditions
and at significantly higher temperatures, when elec-
trons and holes become relevant as additional defects.
Therefore, only proton conductivities for a variety of
oxides at a water partial pressure of pH2O ) 30 hPa
are shown in Figure 25 (ref 186 with data from ref
187). Note that the dopant concentration for the
examples shown is ∼10% and that the proton mobil-
ity in such samples may significantly deviate from
Figure 22. Proton conductivity of PBI‚nH3PO4 adducts, this in undoped materials (see Section 3.1.1.4).
as a function of temperature T and relative humidity RH
for a given phosphoric acid concentration.317
5. Recent Approaches toward New
lower than that in hydrated acidic polymers, and Proton-Conducting Materials for Fuel-Cell
there is almost no water electro-osmotic drag, even Applications
at high relative humidities.254 The proton conductiv-
The suitability of proton-conducting materials as
ity is mainly dependent on the fraction of phosphoric
separators in a particular fuel-cell application is
acid (frequently termed “doping level”), relative hu-
essentially dependent on its transport properties,
midity (RH), and temperature. The corresponding
durability, and reactivity. Thus far, this review has
dependencies are shown in Figures 21 and 22. The
focused on the transport properties only, but any
data suggest that the presence of water has two
approach toward new separator materials must
effects on the conductivity: (i) a minimum RH value
consider all relevant aspects, which makes the de-
is required to prevent the phosphoric acid from
velopment of new competitive materials a complex
condensing at temperatures above T ≈ 100 °C; and
and challenging task.
(ii) excess water leads to a further increase of the
conductivity, as observed in pure phosphoric acid Hence, most current strategies are trying to extend
(Section 3.1.1.2). The data in Figures 21 and 22 are the application limits while conserving the specific
actually taken from one reference,317 except for one advantages of well-established sulfonated materials
curve taken from ref 316, which has been included such as Nafion by slightly varying or modifying them.
to demonstrate the only moderate reproducibility. Apart from the maximum operating temperature and
the high crossover of water and methanol, long-term
stability under fuel-cell conditions is also a severe
4.3. Heterocycle-Based Systems problem. The first two constraints are related to the
The first heterocycle-based systems that had some humidification requirements (see Section 1), and
relevance in the development of new types of fuel- most materials development processes end up com-
cell membranes were sulfonated polyarylenes170 and promising on high proton conductivity on one side
Figure 23. Evolution of proton conductivity of imidazole-based systems with increasing immobilization: from the monomer
via oligomers to fully polymeric systems.
Figure 24. Proton conductivity of Imi-2 (two imidazoles spaced by two ethylene oxide (EO) units), as a function of triflic
acid doping.237 Note that the conductivity has a tendency to level off at high acid concentration (see insert).
and morphological stability (low swelling and insolu- methanol crossover on the other side. One crucial
bility at elevated temperature) and low water and parameter that allows control of these properties is
the polymer, e.g., by cross-linking307-315,321 or forming

co-polymers of sulfonated arylenes and polyvinylidene
difluoride (PVDF).322 Initial fuel-cell tests are actually
quite promising;313 however, the brittleness of highly
cross-linked polymers may cause mechanical failure
of the membrane. The formation of composites with
isolated inorganic particles usually does not have a
positive effect on the swelling properties. In the case
of Nafion, it may suppress its crystallinity, which
leads to an increased swelling. Nevertheless, im-
proved performance under fuel-cell conditions has
been reported in a few cases,243,244 and better under-
standing of the observed effects may guide a more
systematic improvement of composites. The design
of microstructures with a hydrated sulfonated poly-
mer confined in the pores of a stable well-connected
(inorganic) matrix may be another promising ap-
proach. The mechanical requirements of such a
matrix are actually quite severe, because of the need
to compensate for the high pressure due to osmosis
and electro-osmotic drag. Some specific interaction
between the sulfonated polymer and the confining
phase is required to prevent the first from dissolving.
Currently, there are several attempts to build such
Figure 25. Proton conductivity of various oxides, as microstructures, some of which are based on con-
calculated from data on proton concentrations and mobili- trolled precipitation from homogeneous solutions,323
ties, according to Norby and Larring (the type of dopant is and others that rely on sequential formation of the
not indicated; see ref 187 for source data).186 The conduc-
tivity of oxides with a perovskite-type structure are shown
different parts of the microstructure, e.g., by precipi-
by bold lines, and the conductivity of the oxide ion conduc- tation of inorganic proton conductors in porous
tor YSZ (yttria-stabilized zirconia) is shown for comparison. PTFE.245 Another degree of freedom that may be
(reproduced with the kind permission of Annual Reviews, exploited is the degree of sulfonation across the
http://www.AnnualReviews.org). membrane. In a DMFC, the anode side of the
membrane, which is in contact with the methanol
the degree of sulfonation (expressed as equivalent
solution, may be prevented from swelling by locally
weight or ion exchange capacity), and this is usually
reducing the degree of sulfonation (e.g., by laminating
in the range of 800-1200 g/equiv (0.83-1.25 mequiv/
layers of different degree of sulfonation323 or by
g) for perfluorinated polymers and 600-800 g/equiv
surface modification312), whereas an increased degree
(1.25-1.66 mequiv/g) for hydrocarbon separator ma-
of sulfonation may prevent drying out on the anode
terials. Despite the diversity of backbones and poly-
side at high temperatures and low humidity condi-
meric architectures, the transport properties are very
tions.
similar, i.e., the transport coefficients, as a function
of the water volume fraction, are similar to those of As explained in the Introduction, the latter condi-
Nafion (as a representative of perfluorinated mem- tions are of paramount importance for PEM fuel-cell
brane materials) and sulfonated poly(arylene ether technology; however, to date, there are no proton-
ketone)s (as a representative for hydrocarbon based conducting sulfonated polymer membranes available
membranes) (see Figures 19 and 20). Even the that satisfy all requirements under these conditions
hydration isotherms (absorbed water per sulfonic acid (i.e., T > 120 °C, RH < 25%). The most obvious
group, as a function of RH) seem to be quite uniform. limitation is the low proton conductivity, as a result
Only the water uptake in liquid water (or water/ of low hydration levels under these conditions. Be-
methanol mixtures) may differ significantly. Particu- cause of their superacidity, which results in high
larly at elevated temperatures, there may be tre- hydrophilicity (see Section 3.1.2.1), the highest equi-
mendous swelling, and the onset of exaggerated librium water contents are observed in perfluorosul-
(partially irreversible) swelling is closely related to fonic acid polymers. However, even for these, a water
the mechanical properties of the polymers. content of 2.5H2O/SO3H (at T ) 144 °C, RH ) 25%;
The transport properties that are most significantly see Figure 26, bottom) is too low to effectively bridge
affected by changes of the water volume fraction are the separation between neighboring sulfonate groups,
the water/methanol electro-osmotic drag and perme- which is ∼0.8 nm in Nafion (1100 g/equiv).212 Because
ation, both of which have significant contributions of the poor connectivity (percolation) of the possible
from viscous flow (see Section 3.2.1.1). For DMFC water structures and the strong local retardation of
applications (where the membrane is in contact with the water diffusion (see discussion in Section 3.2.1.1),
a liquid water/methanol mixture), this type of trans- fast proton conduction according to a vehicle mech-
port determines the crossover, which is only accept- anism is not possible at such low degrees of hydra-
ably low for solvent volume fractions smaller than tion. On the other hand, there is theoretical38,39 and
∼20 vol % (see Figures 14 and 15). Consequently, experimental indication324 that high proton conduc-
recent attempts have been focused on strengthening tivity may be possible at low hydration levels, pro-
Figure 27. Defective structure of solid trifluoromethane-

sulfonic acid hydrate (CF3SO3H‚H2O)4 found using ab initio
molecular dynamics (AIMD; see Section 2.2.3 for a descrip-
tion of the technique), showing two hydronium ions hydro-
gen-bonded to sulfonate groups (as found in the perfect
structure) but, more importantly, two “shared protons” (one
between two sulfonate groups and the other as part of a
Zundel ion; see text). Note that the energy of the defective
structure is only ∼30 kJ/mol higher than that of the perfect
structure.23,38,39
Figure 26. (a) Conductivity and (b) water uptake of S-C7 molecules. Whether this is the case for the high
(heptane terminated by a sulfonic acid functional group) conductivities of lowly hydrated sulfonic acid oligo-
and Nafion 117 (EW ) 1100 g/equiv), as a function of mers (Figure 26) and triflic acid hydrates is not yet
temperature at a water partial pressure of pH2O ) 105 Pa.324 clear (the high conductivities may also be the conse-
quence of high self-diffusion coefficients of such ionic,
vided that the sulfonic acid functional groups are less highly fluid systems); however, these observations
separated or at least more mobile. The top portion may justify an inclusion of highly sulfonated systems
of Figure 26 shows the ionic conductivity of a sul- with significantly smaller -SO3H separations into
fonated oligomer (heptyl sulfonic acid), which starts future work. Polymers with more than one sulfonic
to show a decrease in conductivity at a higher acid group per phenyl ring are indeed possible (e.g.,
temperature than that observed in Nafion, leveling for poly(arylene sulfide)s) and interesting conductivi-
off to a very high conductivity of ∼4 × 10-2 S/cm in ties at low humidification have been reported.326 Of
the temperature range of 130-170 °C (pH2O ) 105 course, the formation of applicable separator materi-
Pa),324 which is reminescent of the conductivity als requires the immobilization of such highly soluble
behavior of aqueous highly concentrated sulfuric polymers, e.g., in a manner previously described or
acid.325 The latter observation indicates that another by making them a constituent of a macromolecular
proton conduction mechanism appears at low hu- structure with highly sulfonated and unsulfonated
midification levels. Recent electronic structure cal- components in controlled morphologies (for a current
culations of a two-side-chain fragment of the Dow review, see ref 327), which is a strategy pushed
membrane demonstrates that a hydration of three forward by several French groups,328-331 McGrath et
water molecules per sulfonic acid leads to dissociation al.,332 and Miyatake et al.333 The latter authors
and a configuration in which one of the excess protons actually reported the highest high-temperature con-
is shared between two water molecules (the forma- ductivity for a sulfonated polyimide co-polymer that
tion of a Zundel ion) and the other is a hydronium contained fluorenyl moieties, indicating a complex
ion between the two sulfonate groups.23 For some- relation between chemistry, microstructure (packing),
what lower water contents, the affinities of the and proton conductivity for this class of polymers.
sulfonic acid group and the remaining water struc- Note that the typical block separation length for such
ture towards the proton tend to balance. The AIMD polymers is on the order of 10-100 nm, i.e., it is well
(see Section 2.2.3) of triflic acid monohydrate38,39 above the scale relevant for the transport mecha-
indeed finds excess protons within the water struc- nisms (∼1 nm), as discussed in Sections 3.1.2.1 and
ture, namely, as Zundel and hydronium ions, and 3.2.1.
shared between neighboring sulfonate groups (Figure Unfortunately, the thermal stability of sulfonated
27). Such configurations may have a role in the systems is quite limited, particularly when the
mechanisms of proton conduction involving the dy- stabilizing effect of the hydration water is reduced
namics of both the sulfonate groups and the water (Section 3.1.2.1). The transport of water not only
molecules may be terminated by a proton solvent

(e.g., imidazole, phosphonic acid) at one end and by
an ionic group at the other end, with the counter
charge being just a single ion (this may lead to a
liquid (molten) systems with low vapor pressure) or
a poly-ion (which may lead to a plastic material).
The use of polymeric constituents becomes progres-
sively critical with increasing temperature; therefore,
for durability reasons, partially and even fully inor-
ganic systems have recently attracted interest as
proton-conducting separator materials.
Heteropolyacids are frequently used to modify
proton-conducting composites,252 or they are just
dispersed in inert matrixes.335,336 However, because
the proton conduction mechanism of such hydrated
salts is similar to those of hydrated polymeric sys-
Figure 28. Conductivity of two phosphonic acid-termi- tems,337 these composites show qualitatively similar
nated oligomers under dry and wet conditions, compared transport properties. The same is true for organically
to the proton conductivity of Nafion at a water partial modified inorganic layered compounds such as tita-
pressure of pH2O ) 105 Pa.324 nium phosphate sulfophenylenphosphonate, the con-
ductivity of which is dependent on the RH value, in
requires water management but also leads to ther- a manner similar to that observed with Nafion.338
modynamic efficiency losses in fuel cells using such In contrast to this class of materials, acidic salts
membranes as separators. A recent comparative of oxo-acids may show a proton conduction mecha-
study of oligomers terminated by different protogenic nism, which is more related to that of liquid ortho-
groups (sulfonic acid, imidazole, and phosphonic acid) phosphoric acid (see Section 3.1.1.2). The prototypical
shows that the more-amphoteric groups show a compound is CsHSO4, which shows very high proton
higher thermal and electrochemical stability in the conductivity above a first-order phase transition at
dry state,324 where they still show significant proton T ≈ 140 °C.339 The conducting phase is still a solid;
conductivity. However, as discussed in Section 3.1.2.3, that is, there is still long-range order but locally,
this conductivity seems to be limited to ∼10-2 S/cm there is a tremendous dynamic reorientational dis-
for systems based on heterocycles. The other problem order of the sulfate tetrahedral and extended vibra-
with such systems is the high overpotentials for tions of both cesium and sulfate characteristic for so-
oxygen reduction at platinum cathodes, which al- called plastic phases. The hydrogen bonding between
ready has been observed by Yang et al.234 For the sulfate tetrahedral is highly dynamic in the
phosphonic acid functionalized systems, however, the plastic phase, and almost completely self-dissoci-
conductivities in the dry state are ∼1 order of ated: high proton mobility leading to high proton
magnitude higher, and, as opposed to most hetero- conductivity.1,10,144 Haile et al. tested such type of
cycle-based systems, phosphonates have some re- materials in H2/O2 fuel cells340 and recently also in
maining hydrophilicity. This allows for high proton methanol/O2 fuel cells.341 The current densities ob-
conductivity at low temperatures (Figure 28) and tained were limited by the protonic resistance of the
may help to reduce cathodic overpotentials that are electrolytes (CsHSO4 and CsH2PO4), i.e., no cathodic
caused by the adsorption of phosphonic groups onto overpotentials are reported, which is a true advan-
the platinum surface. Of course, this is also an tage, compared to other systems operating under
inherent problem of systems that contain free H3PO4, similar conditions. The softness, high solubility in
such as the adducts of PBI and phosphoric acid (see water, sensitivity to reduction, and the very low
Section 3.1.2.2). room-temperature conductivity are still severe dis-
Apart from the immobilization of the protogenic advantages in these particular compounds; however,
group as part of a polymeric architecture (e.g., soft there is a plethora of solid acids, some of which may
side-chain polymer), ionic immobilization may be be suitable for fuel-cell applications.
another interesting strategy to suppress the volatility Further increase in the operating temperature of
of nonaqueous proton solvents. Some molten salts a fuel cell is possible, using proton-conducting oxides
combine low vapor pressure with very high ionic as separator materials. Although initial fuel-cell tests
conductivity. However, the latter is usually the (typically in the temperature range of 600-800 °C
consequence of a very high fluidity, resulting in high using pure gases) were quite promising,180-184 the
mobility of all ionic species in the systems. Watanabe combination of high bulk proton conductivity and
et al.334 have actually reported a few molten salts that stability has only recently been achieved.186,188,205
contain imidazolium or imidazolium derivatives with Yttrium-doped BaZrO3-based oxides actually show
very high ionic conductivity. However, the charge the highest bulk conductivity (Figure 25), and they
carriers are complex ions, which are very mobile as have a high thermodynamic stability, with respect
a whole, i.e., the conductivity is not protonic. Nev- to reactions with acidic gases such as CO2.188 Prob-
ertheless, when combined with the spacer concept, lems are related to high grain-boundary impedances
ionic immobilization may also become relevant for the (which is probably the result of symmetry reduction
immobilization of (neutral) proton solvents. Spacer in the grain-boundary region), current constriction
Figure 29. Conductivity of some intermediate-temperature proton conductors, compared to the conductivity of Nafion
and the oxide ion conductivity of YSZ (yttria-stabilized zirconia), the standard electrolyte materials for low- and high-
temperature fuel cells, proton exchange membrane fuel cells (PEMFCs), and solid oxide fuel cells (SOFCs).
effects, and high cathodic overpotentials at platinum made over the past few years, and quite a few old
electrodes.186 More-suitable electrocatalysts have not and new materials have been reported that show
been developed thus far; therefore, there is probably interesting proton conductivities, especially in the
a huge potential for further improvements. An in- intermediate temperature range (see Figure 29).343
teresting feature of proton-conducting oxides is, that Also, significant progress in the understanding of
they may show both proton conductivity and chemical transport properties has also been made, which is,
water diffusion under medium dry conditions (Section to a large extent, due to emerging simulation tech-
3.2.2). This allows the transport of water electro- niques (Section 2) as relatively recent research tools.
chemically formed at the cathode, to migrate to the The level of understanding reached so far has already
anode side, where it may be consumed in fuel- helped to better control the transport properties of
reforming reactions. This approach has been sug- available proton-conducting materials, and surely
gested and confirmed by Coors342 and may become will be of great help in a more directed search for
the basis for fuel cells operating with dry methane novel proton-conducting materials for fuel-cell ap-
as a fuel. plications. This is indeed necessary, because, pres-
The key to the development of CO2-resistant proton- ently, Nafion materials, or modified derivatives of its
conducting oxides was the maximization of the en- structure with modified pendant chains, lower equiva-
tropic stabilization of protonic defects.188 If this lent weights, or modified processing are still the
approach also led to stable hydroxides with suf- benchmark PEMs, despite more than twenty years
ficiently high conductivity, AFCs using such electro- of research into alternatives.
lytes may operate even with air as the cathode gas.
This would be tremendously advantageous, because 6. Acknowledgement
fuel cells with nonacidic electrolytes may operate
with non-noble-metal catalysts such as nickel for the The authors thank R. Merkle, E. Kotomin, and J.
anode and silver for the cathode. Fleig (all from the Max-Planck-Institut für Fest-
The above-described qualitative considerations hope- körperforschung) and the external reviewers for
fully give a flavor for the complexity of the develop- carefully reading the proofs and fruitful discussions.
ment of novel proton-conducting separator materials We thank A. Fuchs for assisting in making the
for fuel-cell applications. When solely considering figures and the Deutsche Forschungsgemeinschaft
proton conductivity, significant progress has been (KR 794), the Bundesministerium für Bildung und
Forschung (0329567), and the Energiestiftung Baden- (44) Paddison, S. J.; Paul, R.; Kreuer, K. D.; Zawodzinski, T. A., Jr.
In Direct Methanol Fuel Cells; Narayanan, S., Gottesfeld, S.,
Württemberg (A 19603) for financial support. Zawodzinski, T. A., Eds.; The Electrochemical Society Proceed-
ings Series 01-04: The Electrochemical Society: Pennington, NJ,
2001; p 29.
7. References (45) Paddison, S. J.; Paul, R.; Kreuer, K. D. Phys. Chem. Chem. Phys.
2002, 4, 1151.
(1) Kreuer, K. D. Chem. Mater. 1996, 8, 610.
(46) Paddison, S. J.; Paul, R. Phys. Chem. Chem. Phys. 2002, 4, 1158.
(2) Colomban, P. Ann. Chim.-Sci. Mater. 1999, 24, 1.
(47) Paul, R.; Paddison, S. J. In Advances in Materials Theory and
(3) Vielstich, W., Lamm, A., Gasteiger, H. A., Eds. Handbook of Fuel ModelingsBridging Over Multiple-Length and Time Scales;
CellssFundamentals, Technology and Applications; Wiley: Chi- Bulatov, V., Colombo, L., Cleri, F., Lewis, L. J., Mousseau, N.,
chester, U.K., 2003. Eds.; Materials Research Society: Warrendale, PA, 2001; pp
(4) Annu. Rev. Mater. Res. 2003, 33. (Special issue, Materials for AA7.16.1.
Fuel Cells.) (48) Paul, R.; Paddison, S. J. J. Chem. Phys. 2001, 115, 7762.
(5) Savadogo, O. J. New Mater. Electrochem. Syst. 1998, 1, 47. (49) Paul, R.; Paddison, S. J. Phys. Rev. E 2003, 67, 016108-1.
(6) Li, Q.; He, R.; Jensen, J. O.; Bjerrum, N. J. Chem. Mater. 2003, (50) Paul, R.; Paddison, S. J. Solid State Ionics 2004, 168, 245.
15, 4896. (51) Paul, R.; Paddison, S. J. J. Phys. Chem. B 2004, 108, 13231-
(7) Rikukawa, M.; Sanui, K. Prog. Polym. Sci. 2000, 25, 1463. 13241.
(8) Rozière, J.; Jones, D. J. Annu. Rev. Mater. Res. 2003, 33, 503. (52) Szabo, A.; Ostlund, N. S. Modern Quantum Chemistry: Intro-
(9) Jannasch, P. Curr. Opin. Colloid Interface Sci. 2003, 8, 96. duction to Advanced Electronic Structure Theory; McGraw-
(10) Münch, W.; Kreuer, K. D.; Traub, U.; Maier, J. J. Mol. Struct. Hill: New York, 1989.
(THEOCHEM) 1996, 381, 1. (53) McWeeny, R. Methods of Molecular Quantum Mechanics; Aca-
(11) Büchi, F. N.; Huslage, J., Scherer, G. G. PSI Annu. Rep. 1997, demic Press: London, 1989.
Annex V, 48. (54) Hehre, W. J.; Radom, L.; Schleyer, P. V. R.; Pople, J. A. Ab Initio
(12) Springer, T. E.; Zawodzinski, T. A.; Gottesfeld, S. J. Electrochem. Molecular Orbital Theory; Wiley: New York, 1986.
Soc. 1991, 138, 2334. (55) Born, M.; Oppenheimer, J. R. Ann. Phys. 1927, 79, 361.
(13) Fuller, T. F.; Newman, J. J. Electrochem. Soc. 1991, 140, 1218. (56) Hartree, D. R. Proc. Cambridge Philos. Soc. 1928, 24, 328.
(14) Eikerling, M.; Kharkats, Y. I.; Kornyshev, A. A.; Volfkovich, Y. (57) Fock, V. A. Z. Phys. 1930, 15, 126.
M. J. Electrochem. Soc. 1998, 145, 2677. (58) Møller, C.; Plesset, M. S. Phys. Rev. 1934, 46, 618.
(15) Paddison, S. J.; Pratt, L. R.; Zawodzinski, T.; Reagor, D. W. Fluid (59) Shavitt, I. In Modern Theoretical Chemistry; Schaefer, H. F., Ed.;
Phase Equilib. 1998, 150-151, 235. Plenum Press: New York, 1977.
(16) Paddison, S. J.; Pratt, L. R.; Zawodzinski, T. A., Jr. In Proton (60) Boys, S. F. Proc. R. Soc. London A 1950, A200, 542.
Conducting Membrane Fuel Cells II; Gottesfeld, S., Fuller, T. (61) Parr, R. G.; Yang, W. Density Functional Theory of Atoms and
F., Eds.; The Electrochemical Society Proceedings Series 98-27; Molecules; Oxford: New York, 1989.
The Electrochemical Society: Pennington, NJ, 1999; p 99. (62) Hohenberg, P.; Kohn, W. Phys. Rev. B 1964, 136, B864.
(17) Paddison, S. J.; Zawodzinski, T. A., Jr. Solid State Ionics 1998, (63) Kohn, W.; Sham, L. J. Phys. Rev. 1965, 140, A1133.
113-115, 333. (64) Lee, C.; Yang, W.; Parr, R. G. Phys. Rev. B 1988, 37, 785.
(18) Paddison, S. J.; Pratt, L. R.; Zawodzinski, T. A., Jr. J. New Mater. (65) Perdew, J. P. Phys. Rev. B 1986, 33, 8822; 1986, 34, 7406.
Electrochem. Syst. 1999, 2, 183. (66) Perdew, J. P. In Electronic Structure of Solids; Ziesche, P.,
(19) Paddison, S. J.; Pratt, L. R.; Zawodzinski, T. A., Jr. J. Phys. Eschrig, H., Eds.; Academie Verlag: Berlin, 1991.
Chem. A 2001, 105, 6266. (67) Becke, A. D. Phys. Rev. 1988, 38, A3098.
(20) ) Paddison, S. J. J. New Mater. Electrochem. Syst. 2001, 4, 197. (68) Johnson, B. G.; Gill, P. M. W.; Pople, J. A. J. Chem. Phys. 1993,
(21) Eikerling, M.; Paddison, S. J.; Zawodzinski, T. A., Jr. J. New 98, 5612.
Mater. Electrochem. Syst. 2002, 5, 15. (69) Porezag, D.; Frauenheim, T.; Köhler, T.; Seifert, G.; Kaschner,
(22) Paddison, S. J. In Handbook of Fuel Cells Fundamentals R. Phys. Rev. B 1995, 51, 12947.
Technology and Applications: Volume 3, Fuel Cell Technology (70) Seifert, G.; Porezag, D.; Frauenheim, T. Int. J. Quantum Chem.
and Applications: Part 1; Vielstich, W., Lamm, A., Gasteiger, 1996, 58, 185.
H. A., Eds.; Wiley: Chichester, U.K., 2003; p 396. (71) Frauenheim, T.; Seifert, G.; Elstner. M.; Niehaus, T. A.; Köhler,
(23) Paddison, S. J. Annu. Rev. Mater. Res. 2003, 33, 289. T.; Amkreutz, M.; Sternberg, M.; Hajnal, Z.; di Carlo, A.; Suhai,
(24) Johansson, P.; Tegenfeldt, J.; Lindgren, J. Electrochim. Acta S. J. Phys. C 2002, 14, 3015.
2000, 45, 3055. (72) Seifert, G. In Encyclopedia of Computational Chemistry; Wiley:
(25) Li, T.; Wlaschin, A.; Balbuena, P. B. Ind. Eng. Chem. Res. 2001, New York, in press.
40, 4789. (73) Münch, W.; Kreuer, K. D.; Seifert, G.; Maier, J. Solid State Ionics
(26) Spohr, E.; Commer, P.; Kornyshev, A. A. J. Phys. Chem. B 2002, 2000, 136-137, 183.
106, 10560. (74) Allen, M. P.; Tildesley, D. J. Computer Simulation of Liquids;
(27) Commer, P.; Cherstvy, A. G.; Spohr, E.; Kornyshev, A. A. Fuel Clarendon Press: Oxford, U.K., 1987.
Cells 2002, 2, 127. (75) Gray, C. G.; Gubbins, K. E. Theory of Molecular Fluids; Clar-
(28) Spohr, E. Mol. Simul. 2004, 30, 107. endon Press: Oxford, U.K., 1984; Vol. 1.
(29) Commer, P.; Hartnig, C.; Seeliger, D.; Spohr, E., submitted to (76) Remler, D. K.; Madden, P. A. Mol. Phys. 1990, 70, 921.
Mol. Simul. (77) Payne, M. C.; Teter, M. P.; Allan, D. C.; Arias, T. A.; Joannopou-
(30) Warshel, A. Computer Modeling of Chemical Reactions in los, J. D. Rev. Mod. Phys. 1992, 64, 1045.
Enzymes and Solutions; Wiley: New York, 1991. (78) Metropolis, N. A.; Rosenbluth, A. W.; Rosenbluth, M. N.; Teller,
(31) Warshel, A. J. Phys. Chem. 1982, 86, 2218. A. H.; Teller, E. J. Chem. Phys. 1953, 21, 1087.
(32) Warshel, A.; Weiss, R. M. J. Am. Chem. Soc. 1980, 102, 6218. (79) Alder, B. J.; Wainwright, T. E. J. Chem. Phys. 1957, 27, 1208.
(33) Tuckerman, M.; Laasonen, K.; Sprik, M.; Parrinello, M. J. Phys. (80) Dick, B. G.; Overhauser, A. W. Phys. Rev. 1958, 112, 90.
Chem. 1995, 99, 5749. (81) Halley, J. W.; Rustad, J. R.; Rahman, A. J. Chem. Phys. 1993,
(34) Tuckerman, M.; Laasonen, K.; Sprik, M.; Parrinello, M. J. Chem. 98, 4110.
Phys. 1995, 103, 150. (82) Hammes-Schiffer, S.; Tully, J. C. J. Chem. Phys. 1994, 101, 4657.
(35) Tuckerman, M. E.; Marx, D.; Klein, M. L.; Parrinello, M. Science (83) Vuilleumier, R.; Borgis, D. J. Mol. Struct. (THEOCHEM) 1997,
1997, 275, 817. 436, 555.
(36) Marx, D.; Tuckerman, M. E.; Hutter, J.; Parrinello, M. Nature (84) Vuilleumier, R.; Borgis, D. J. Phys. Chem. B 1998, 102, 4261.
1999, 397, 601. (85) Vuilleumier, R.; Borgis, D. Chem. Phys. Lett. 1998, 284, 71.
(37) Münch, W.; Kreuer, K.-D.; Silvestri, W.; Maier, J.; Seifert, G. (86) Vuilleumier, R.; Borgis, D. Isr. J. Chem. 1999, 39, 457.
Solid State Ionics 2001, 145, 437. (87) Vuilleumier, R.; Borgis, D. J. Chem. Phys. 1999, 111, 4251.
(38) Eikerling, M.; Paddison, S. J.; Pratt, L. R.; Zawodzinski, T. A., (88) Schmitt, U. W.; Voth, G. A. J. Phys. Chem. B 1998, 102, 5547.
Jr. Chem. Phys. Lett. 2003, 368, 108. (89) Schmitt, U. W.; Voth, G. A. J. Chem. Phys. 1999, 111, 9361.
(39) Paddison, S. J.; Pratt, L. R., submitted to Mol. Phys., 2004. (90) Schmitt, U. W.; Voth, G. A. Isr. J. Chem. 1999, 39, 483.
(40) Paddison, S. J.; Paul, R.; Zawodzinski, T. A., Jr. In Proton (91) Schmitt, U. W.; Voth, G. A. Chem. Phys. Lett. 2000, 329, 36.
Conducting Membrane Fuel Cells II; Gottesfeld, S., Fuller, T. (92) Kim, J.; Schmitt, U. W.; Gruetzmacher, J. A.; Voth, G. A.;
F., Eds.; The Electrochemical Society Proceedings Series 98-27; Scherer, N. F. J. Chem. Phys. 2000, 116, 737.
The Electrochemical Society: Pennington, NJ, 1999; p 106. (93) Day, T. J. F.; Soudackov, A. V.; Cuma, M.; Schmitt, U. W.; Voth,
(41) Paddison, S. J.; Paul, R.; Zawodzinski, T. A., Jr. J. Electrochem. G. A. J. Chem. Phys. 2002, 117, 5839.
Soc. 2000, 147, 617. (94) Kornyshev, A. A.; Walbran, S. J. Chem. Phys. 2001, 114, 10039.
(42) Paddison, S. J.; Paul, R.; Zawodzinski, T. A., Jr. J. Chem. Phys. (95) Kresse, G.; Hafner, J. Phys. Rev. B 1993, 47, 558.
2001, 115, 7753. (96) Kresse, G.; Hafner, J. Phys. Rev. B 1994, 49, 14251.
(43) Paddison, S. J.; Paul, R.; Pivovar, B. S. In Direct Methanol Fuel (97) Kresse, G.; Furthmüller, J. Phys. Rev. B 1996, 54, 11169.
Cells; Narayanan, S., Gottesfeld, S., Zawodzinski, T. A., Eds.; (98) Car, R.; Parrinello, M. Phys. Rev. Lett. 1985, 55, 2471.
The Electrochemical Society Proceedings Series 01-04; The (99) Eikerling, M.; Kornyshev, A. A. J. Electroanal. Chem. 2001, 502,
Electrochemical Society: Pennington, NJ, 2001; p 8. 1.
(100) Eikerling, M.; Kornyshev, A. A.; Kuznetsov, A. M.; Ulstrup, J.; (163) Zundel, G.; Weidemann, E. G. Eur. Biophys. Congr., Proc., 1st
Walbran, S. J. Phys. Chem. B 2001, 105, 3646. 1971, 6, 43.
(101) Harned, H. S.; Owen, B. B. The Physical Chemistry of Electrolyte (164) Riehl, N. Trans. N.Y. Acad. Sci. 1965, 27, 772.
Solutions; Reinhold: New York, 1958. (165) Will, G. Z. Kristallogr. 1969, 129, 211.
(102) Wolynes, P. G. Annu. Rev. Phys. Chem. 1980, 31, 345. (166) Brown, G. P.; Aftergut, S. J. Chem. Phys. 1963, 38, 1356.
(103) Kay, R. L. In Water, A Comprehensive Treatise; Franks, F., Ed.; (167) Pigon, K.; Chojnacki, H. Acta Phys. Pol. 1967, 31, 1069.
Plenum Press: New York, 1973; Vol. 3. (168) McGhie, A. R.; Blum, H.; Labes, M. M. J. Chem. Phys. 1970, 52,
(104) Robinson, R. A.; Stokes, R. H. Electrolyte Solutions, 2nd Edition; 6141.
Butterworth: London, 1959. (169) Hickman, B. S.; Mascal, M.; Titman, J. J.; Wood, I. G. J. Am.
(105) Born, M. Z. Phys. 1920, 1, 221. Chem. Soc. 1999, 121, 11486.
(106) Fuoss, R. M. Proc. Natl. Acad. Sci. 1959, 45, 807. (170) Kreuer, K. D.; Fuchs, A.; Ise, M.; Spaeth, M.; Maier, J. Electro-
(107) Boyd, R. H. J. Chem. Phys. 1961, 35, 1281. chim. Acta 1998, 43, 1281.
(108) Zwanzig, R. J. Chem. Phys. 1963, 38, 1603; 1970, 52, 3625. (171) Kreuer, K. D. Solid State Ionics: Science & Technology; Chow-
(109) Hubbard, J. B.; Onsager, L. J. Chem. Phys. 1977, 67, 4850. dari, B. V. R., Lal, K., Agnihotry, S. A., Khare, N., Sekhon, S.
(110) Hubbard, J. B. J. Chem. Phys. 1978, 68, 1649. S., Srivastava, P. C., Chandra, S., Eds.; World Scientific:
(111) Wolynes, P. G. J. Chem. Phys. 1978, 68, 473. Singapore, 1998, p 263.
(112) Colonomos, P.; Wolynes, P. G. J. Chem. Phys. 1979, 71, 2644. (172) Daycock, J. T.; Jones, G. P.; Evans, J. R. N.; Thomas, J. M.
(113) Chen, J. H.; Adelman, S. A. J. Chem. Phys. 1980, 72, 2819. Nature 1968, 218, 673.
(114) Chong, S.-H.; Hirata, F. J. Chem. Phys. 1998, 108, 7339. (173) Tatara, W.; Wojcik, M. J.; Lindgren, J.; Probst, M. J. Phys. Chem.
(115) Biswas, B.; Roy, S.; Bagchi, B. Phys. Rev. Lett. 1995, 75, 1098. A 2003, 107, 7827.
(116) Biswas, B.; Bagchi, B. J. Chem. Phys. 1997, 106, 5587; J. Am. (174) Schuster, M.; Kreuer, K. D.; Maier, J., in preparation.
Chem. Soc. 1997, 119, 5946. (175) Toda, F.; Tanaka, K.; Foces-Foces, C.; Llamas-Saiz, A. L.;
(117) Bagchi, B. J. Chem. Phys. 1998, 109, 3989. Limbach, H. H.; Aguilar-Parrilla, F.; Claramunt, R. M.; López,
(118) Resibois, P. M. V. Electrolyte Theory: An Elementary Introduc- C.; Elguero, J. J. Chem. Soc. Chem. Commun. 1993, 1139.
tion to a Microscopic Approach; Harper and Row: New York, (176) Stotz, S.; Wagner, C. Ber. Bunsen-Ges. Phys. Chem. 1966, 70,
1968. 781.
(119) Belletete, M.; Lachapelle, M.; Durocher, G. J. Phys. Chem. 1990, (177) Takahashi, T.; Iwahara, H. 1980. Rev. Chim. Minér. 1980, t17,
94, 5337. 243.
(120) Guha Ray, J.; Sengupta, P. K. Chem. Phys. Lett. 1994, 230, 75. (178) Iwahara, H.; Uchida, H.; Tanaka, S. Solid State Ionics 1983,
(121) Das, K.; Sarkar, N.; Das, S.; Datta, A.; Bhattacharyya, K. Chem. 9-10, 1021.
Phys. Lett. 1996, 249, 323. (179) Iwahara, H.; Uchida, H.; Ono, K.; Ogaki, K. J. Electrochem. Soc.
(122) Datta, A.; Mandal, D.; Pal, S. K.; Bhattacharyya, K. J. Phys. 1988, 135, 529.
Chem. B 1997, 101, 10221. (180) Iwahara, H. Solid State Ionics 1988, 28-30, 573.
(123) Faeder, J.; Ladanyi, B. M. J. Chem. Phys. 2000, 104, 1033. (181) Iwahara, H.; Uchida, H.; Morimoto, K. J. Electrochem. Soc. 1990,
(124) Roux, B.; Karplus, M. Annu. Rev. Biophys. Biomol. Struct. 1994, 137, 462.
23, 731. (182) Bonanos, N.; Ellis, B.; Mahmood, M. N. Solid State Ionics 1991,
(125) Breed, J.; Sankararamakrishnan, R.; Kerr, I. D.; Sansom, M. S. 44, 305.
P. Biophys. J. 1996, 70, 1643. (183) Tanigushi, N.; Hatoh, K.; Niikura, J.; Gamo, T. Solid State Ionics
(126) Senapati, S.; Chandra, A. J. Phys. Chem. B 2001, 105, 5106. 1992, 53-56, 998.
(127) Booth, F. J. Chem. Phys. 1950, 19, 391. (184) Iwahara, H.; Yashima, T.; Hibino, T.; Ushida, H. J. Electrochem.
(128) Bontha, J. R.; Pintauro, P. N. Chem. Eng. Sci. 1994, 49, 3835. Soc. 1993, 140, 1687.
(129) Steiner, T. Angew. Chem. 2002, 114, 50. (185) Bonanos, N.; Knight, K. S.; Ellis, B. Solid State Ionics 1995, 79,
(130) Dippel, T.; Hainovsky, N.; Kreuer, K. D.; Münch, W.; Maier, J. 161.
Ferroelectrics 1995, 167, 59. (186) Kreuer, K. D. Annu. Rev. Mater. Res. 2003, 33, 333.
(131) Kreuer, K. D. Solid State Ionics 2000, 136-137, 149. (187) Norby, T.; Larring, Y. Curr. Opin. Solid State Mater. Sci. 1997,
(132) Desiraju, G. R. Angew. Chem., Int. Ed. Engl. 1995, 34, 2311. 2, 593.
(133) Desiraju, G. R. Chem. Commun. 1997, (16), 1475. (188) Kreuer, K. D. Solid State Ionics 1999, 125, 285.
(134) Schmidt, G. M. J. Pure Appl. Chem. 1971, 27, 647. (189) Münch, W.; Seifert, G.; Kreuer, K. D.; Maier, J. Solid State Ionics
(135) Weissbuch, I.; Popovitz-Biro, R.; Lahav, M.; Leiserovitz, L. Acta 1996, 86-88, 647.
Crystallogr. B: Struct. Sci. 1995, 51, 115. (190) Shimojo, F.; Hoshino, K.; Okazaki, H.; J. Phys. Soc. Jpn. 1997,
(136) McDonald, J. C.; Whitesides, G. M. Chem. Rev. 1994, 94, 2383. 66, 8.
(137) Eigen, M.; De Maeyer, L. Proc. R. Soc. (London), Ser. A 1958, (191) Shimojo, F.; Hoshino, K. Solid State Ionics 2001, 145, 421.
247, 505. (192) Pionke, M.; Mono, T.; Schweika, W.; Springer, T.; Schober, T.
(138) Eigen, M. Angew. Chem. 1963, 75, 489. Solid State Ionics 1997, 97, 497.
(139) Zundel, G.; Metzger, H. Z. Naturforsch. 1967, 22a, 1412. (193) Matzke, T.; Stimming, U.; Karmonik, C.; Soetramo, M.; Hemp-
(140) Agmon, N. Chem. Phys. Lett. 1995, 244, 456. elmann, R.; Güthoff, F. Solid State Ionics 1996, 86-88, 621.
(141) Dippel, T.; Kreuer, K. D. Solid State Ionics 1991, 46, 3. (194) Hempelmann, R.; Soetramo, M.; Hartmann, O.; Wäppling, R.
(142) Cohen, B.; Huppert, D. J. Phys. Chem. A 2003, 107, 3598. Solid State Ionics 1998, 107, 269.
(143) Agmon, N. Isr. J. Chem. 1999, 39, 493. (195) Münch, W.; Seifert, G.; Kreuer, K. D.; Maier, J. Solid State Ionics
(144) Kreuer, K. D. Solid State Ionics 1997, 94, 55. 1997, 97, 39.
(145) Sadeghi, R. R.; Cheng, H. P. J. Chem. Phys. 1999, 111, 2086. (196) Kreuer, K. D.; Münch, W.; Traub, U.; Maier, J. Ber. Bunsen-
(146) Cui, Q.; Karplus, M. J. Phys. Chem. B 2003, 107, 1071. Ges. Phys. Chem. 1998, 102, 552.
(147) Kreuer, K. D.; Rabenau, A.; Weppner, W. Angew. Chem., Int. (197) Kreuer, K. D. Solid State Ionics 1997, 97, 1.
Ed. Engl. 1982, 21, 208. (198) Sata, N.; Hiramoto, K.; Ishigame, M.; Hosoya, S.; Niimura, N.;
(148) Agmon, N. J. Chim. Phys. Phys.sChim. Biol. 1996, 93, 1714. Shin, S. Phys. Rev. B 1996, 54, 15795.
(149) Tuckerman, M. E.; Marx, D.; Parrinello, M. Nature 2002, 417, (199) Münch, W.; Kreuer K. D.; Seifert, G.; Maier, J. Solid State Ionics
925. 2000, 136-137, 183.
(150) Trout, B. L.; Parrinello, M. J. Phys. Chem. B 1999, 103, 7340. (200) Münch, W.; Kreuer K. D.; Seifert, G.; Maier, J. Solid State Ionics
(151) Asthagiri, D.; Pratt, L. R.; Kress, J. D.; Gomez, M. A. Chem. 1999, 125, 39.
Phys. Lett. 2003, 380, 530. (201) Nowick, A. S. Solid State Ionics 1997, 97, 17.
(152) Asthagiri, D.; Pratt, L. R.; Kress, J. D.; Gomez, M. A. Proc. Natl. (202) Kreuer, K. D.; Fuchs, A.; Maier, J. Solid State Ionics 1995, 77,
Acad. Sci. (USA) 2004, 101, 7233. 157.
(153) Chen, B.; Park, J. M.; Ivanov, I.; Tabacchi, G.; Klein, M. L.; (203) Münch, W.; Kreuer, K. D.; Adams, S.; Seifert, G.; Maier, J. Phase
Parrinello, M. J. Am. Chem. Soc. 2002, 124, 8534. Transitions 1999, 68, 567.
(154) Kuo, J. L.; Ciobanu, C. V.; Ojamäe, L.; Shavitt, I.; Singer, S. J. (204) Ervine, J. T. S.; Corcoran, D. J. D.; Canales-Vasques, J. Solid
J. Chem. Phys. 2003, 118, 3583. State Ionics 2002, 152, 749.
(155) Munson, R. A. J. Phys. Chem. 1964, 68, 3374. (205) Kreuer, K. D.; Adams, S.; Fuchs, A.; Klock, U.; Münch, W.; Maier,
(156) Dippel, T.; Kreuer, K. D.; Lassègues, J. C.; Rodriguez, D. Solid J. Solid State Ionics 2001, 145, 295.
State Ionics 1993, 61, 41. (206) Kreuer, K. D.; Münch, W.; Ise, M.; He, T.; Fuchs, A.; Traub, U.;
(157) Spaeth, M.; Kreuer, K. D.; Dippel, T.; Maier, J. Solid State Ionics Maier, J Ber. Bunsen-Ges. Phys. Chem. 1997, 101, 1344.
1997, 97, 291. (207) Nowick, A. S.; Yang, Du; Liang, K. C. Solid State Ionics 1999,
(158) Spaeth, M.; Kreuer, K. D.; Maier, J. J. Solid State Chem. 1999, 125, 303.
148, 169. (208) Gebel, G.; Aldebert, P.; Pineri, M. Polymer 1993, 34, 333.
(159) Chung, S. H.; Bajue, S.; Greenbaum, S. G. J. Chem. Phys. 2000, (209) Gebel, G.; Lambard, J. Macromolecules 1997, 30, 7914.
112, 8515. (210) Gebel, G. Polymer 2000, 41, 5829.
(160) Schlechter, A.; Savinell, R. F. Solid State Ionics 2002, 147, 181. (211) Rubalat, L.; Rollet, A. L.; Diat, O.; Gebel, G. J. Phys. IV 2002,
(161) Kawada, A.; McGhie, A. R.; Labes, M. M. J. Chem. Phys. 1970, 12 (PR6), 197.
52, 3121. (212) Ise, M. Ph.D. Thesis, University of Stuttgart, Stuttgart, Ger-
(162) Decoursey, T. E. Phys. Rev. 2003, 83, 475. many, 2000.
(213) Kreuer, K. D. J. Membr. Sci. 2001, 185, 29. (259) Edmondson, C. A.; Fontanella, J. J.; Chung, S. H.; Greenbaum,
(214) Elliott, J. A.; Hanna, S.; Elliott, A. M. S.; Cooley, G. E. S. G.; Wnek, G. E. Electrochim. Acta 2001, 46, 1623.
Macromolecules 2000, 33, 4161. (260) Edmondson, C. A.; Fontanella, J. J. Solid State Ionics 2002, 152-
(215) Khalatur, P. G.; Talitskikh, S. K.; Khokhlov, A. R. Macromol. 153, 355.
Theory Simul. 2002, 11, 566. (261) Fontanella, J. J.; McLin, M. G.; Wintersgill, M. C.; Calame, J.
(216) Edmondson, C. A.; Fontanella, J. J. Solid State Ionics 2002, 152- P.; Greenbaum, S. G. Solid State Ionics 1993, 66, 1.
153, 355. (262) Ren, X.; Springer, T. E.; Zawodzinski, T. A.; Gottesfeld, S. J.
(217) Haubold, H. G.; Vad, T.; Jungbluth, H.; Hiller, P. Electrochim. Electrochem. Soc. 2000, 147, 466.
Acta 2001, 46, 1559. (263) Hietala, S.; Maunu, S. L.; Sundholm, F. J. Polym. Sci., Part B:
(218) Kazansky, V.; Solkan, V. Phys. Chem. Chem. Phys. 2003, 5, 31. Polym. Phys. 2000, 38, 3277.
(219) Kreuer, K. D. in Handbook of Fuel CellssFundamentals, (264) Doyle, M.; Rajendran, G. In Handbook of Fuel CellssFun-
Technology and Applications; Vielstich, W., Lamm, A., Gasteiger, damentals, Technology and Applications; Vielstich, W., Lamm,
H. A. Eds.; Wiley: Chichester, U.K., 2003; p 420. A., Gasteiger, H. A., Eds.; Wiley: Chichester, U.K., 2003, p 351.
(220) Eikerling, M.; Kornyshev, A. A.; Stimming, U. J. Phys. Chem. (265) Motupally, S.; Becker, A. J.; Weidner, J. W. J. Electrochem. Soc.
B 1997, 101, 10807. 2000, 147, 3171.
(221) Corry, B.; Kuyucak, S.; Chung, S. H. Chem. Phys. Lett. 2000, (266) Verbrugge, M. W. J. Electrochem. Soc. 1989, 136, 417.
320, 35. (267) LaConti, A. B.; Fragala, A. R.; Boyack, J. R. In Proceedings of
(222) Paddison, S. J.; Bender, G.; Kreuer, K. D.; Nicoloso, N.; Zawod- the Symposium on Electrode Materials and Processes for Energy
zinski, T. A. J. New Mater. Electrochem. Syst. 2000, 3, 291. Conversion and Storage; McIntyre, J. D. E., Srinivasan, S., Will,
(223) Paddison, S. J.; Reagor, D. W.; Zawodzinski, T. A. J. Electroanal. F. G., Eds.; The Electrochemical Society Softbound Proceedings
Chem. 1998, 459, 91. Series, PV 77-6; The Electrochemical Society: Princeton, NJ,
(224) Kreuer, K. D.; Dippel, T.; Meyer, W.; Maier, J. Mater. Res. Soc. 1977; p 354.
Symp. Proc. 1993, 293, 273. (268) Meier, F.; Kerres, J.; Eigenberger, G. J. New Mater. Electrochem.
(225) Rollet, A. L.; Simonin, J. P.; Turq, P.; Gebel, G.; Kahn, R.; Syst. 2002, 5, 91.
Vandais, A.; Noël, J. P.; Malveau, C.; Canet, D. J. Phys. Chem. (269) Helmholtz, H. Weid. Ann. 1879, 7, 337.
2001, 105, 4503. (270) Lamb, H. Philos. Mag. 1888, 5, 52.
(226) Zawodzinski, T. A.; Springer, T. E.; Davey, J.; Jestel, R.; Lopez, (271) Perrin, J. J. Chim. Phys. 1904, 2, 601.
C.; Valerio, J.; Gottesfeld, S. J. Electrochem. Soc. 1993, 140, (272) Smoluchowski, M. Handbuch der Elektrizität und des Magne-
1981. tismus, Vol. II; Barth: Leipzig, 1914.
(227) Kreuer, K. D.; Dippel, T.; Maier, J. Proc. Electrochem. Soc. 1995, (273) Breslau, B. R.; Miller, I. F. Ind. Eng. Chem. Fundam. 1971, 10,
95-23, 241. 554.
(228) Cappadonia, M.; Erning, J. W.; Niaki, S. M. S.; Stimming, U. (274) Ise, M.; Kreuer, K. D.; Maier, J. Solid State Ionics 1999, 125,
Solid State Ionics 1995, 77, 65. 213.
(229) Wainright, J. S.; Wang, J. T.; Weng, D.; Savinell, R. F.; Litt, M. (275) Okada, T.; Moller-Holst, S.; Gorseth, O.; Kjelstrup, S. J. Elec-
J. Electrochem. Soc. 1995, 142, L121. troanal. Chem. 1998, 442, 137.
(230) Glipa, X.; Bonnet, B.; Mula, B.; Jones, D. J.; Rozière, J. J. Mater. (276) Ren, X.; Gottesfeld, S. J. Electrochem. Soc. 2001, 148, A87.
Chem. 1999, 9, 3045. (277) Zawodzinski, T. A.; Davey, J.; Valerio, J.; Gottesfeld, S. Elec-
(231) Bouchet, R.; Siebert, E. Solid State Ionics 1999, 118, 287. trochim. Acta 1995, 40, 297.
(232) Benicewicz, B. C. Presented as a poster during the 2003 Gordon (278) Zawodzinski, T. A.; Derouin, C.; Radzinski, S.; Sherman, R. J.;
Research Conference on Fuel Cells, Bristol, RI, July 27-August Smith, Van T.; Springer, T. E.; Gottesfeld, S. J. Electrochem.
1, 2003. Soc. 1993, 140, 1041.
(233) Bozkurt, A.; Ise, M.; Kreuer, K. D.; Meyer, W. H.; Wegner, G. (279) Ren, X.; Henderson, W.; Gottesfeld, S. J. Electrochem. Soc. 1997,
Solid State Ionics 1999, 125, 225. 144, L267.
(234) Yang, C.; Costamagna, P.; Srinivasan, S.; Benziger, J.; Bocarsly, (280) Xie, G.; Okada, T. J. Electrochem. Soc. 1995, 142, 3057.
A. B. J. Power Sources 2001, 103, 1. (281) Fuller, T. F.; Newman, J. J. Electrochem. Soc. 1992, 139, 1332.
(235) Bozkurt, A.; Meyer, W. H.; Wegner, G. J. Power Sources 2003, (282) Yeo, R. S.; McBreen, J. J. Electrochem. Soc. 1979, 126, 1682.
123, 126. (283) Ogumi, Z.; Takehara, Z.; Yoshizawa, S. J. Electrochem. Soc.
(236) Münch, W., unpublished data. 1984, 131, 769.
(237) Schuster, M. F.; Meyer, W. H.; Wegner, G.; Herz, H. G.; Ise, M.; (284) Ogumi, Z.; Kuroe, T.; Takehara, Z. J. Electrochem. Soc. 1985,
Kreuer, K. D.; Maier, J. Solid State Ionics 2001, 145, 85. 132, 2601.
(238) Schuster, M. F. H.; Meyer, W. H. Annu. Rev. Mater. Res. 2003, (285) Sakai, T.; Takenaka, H.; Wakabayashi, N.; Kawami, Y.; Torikai,
33, 233. E. J. Electrochem. Soc. 1985, 132, 1328.
(239) Schuster, M. F. H.; Meyer, W. H.; Schuster, M.; Kreuer, K. D. (286) Parthasarathy, A.; Martin, C. R.; Srinivasan, S. J. Electrochem.
Chem. Mater. 2004, 16, 329. Soc. 1992, 138, 916.
(240) Goward, G. R.; Schuster, M. F. H.; Sebastiani, D.; Schnell, I.; (287) Broka, K.; Ekdunge, P. J. Appl. Electrochem. 1997, 27, 117.
Spiess, H. W. J. Phys. Chem. B 2002, 106, 9322. (288) Hietala, S.; Skou, E.; Sundholm, F. Polymer 1999, 40, 5567.
(241) Herz, H. G.; Kreuer, K. D.; Maier, J.; Scharfenberger, G.; (289) Basura, V. I.; Chuy, C.; Beattie, P. D.; Holdcroft, S. J. Electro-
Schuster, M. F. H.; Meyer, W. H. Electrochim. Acta 2003, 48, anal. Chem. 2001, 501, 77.
2165. (290) Savinell, R.; Yeager, E.; Tryk, D.; Landau, U.; Wainright, J.;
(242) Perrson, J. C.; Jannasch, P. Chem. Mater. 2003, 15, 3044. Weng, D.; Lux, D.; Litt, M.; Rogers, C. J. Electrochem. Soc. 1994,
(243) Miyake, N.; Wainright, J. S.; Savinell, R. F. J. Electrochem. Soc. 141, L46.
A 2001, 148, A898. (291) Sakai, T.; Takenaka, H.; Torikai, E. J. Electrochem. Soc. 1986,
(244) Adjemian, K. T.; Lee, S. J.; Srinvasan, S.; Benziger, J.; Bocarsly, 133, 88.
A. B. J. Electrochem. Soc. A 2002, 149, A256. (292) Büchi, F. N.; Wakizoe, M.; Srinivasan, S. J. Electrochem. Soc.
(245) Alberti, G.; Casciola, M. Annu. Rev. Mater. Res. 2003, 33, 129. 1996, 143, 927.
(246) Kumar, B.; Fellner, J. P. J. Power Sources 2003, 123, 132. (293) Kreuer, K. D.; Schönherr, E.; Maier, J. Solid State Ionics 1994,
(247) Mauritz, K. A.; Warren, R. M. Macromolecules 1989, 22, 1730. 70/71, 278.
(248) Costamagna, P.; Yang, C.; Bocarsly, A. B.; Srinivasan, S. (294) Virkar, A. V.; Baek, H.-D. J. Am. Ceram. Soc. 2002, 85, 3059.
Electrochim. Acta 2002, 47, 1023. (295) Kreuer, K. D.; Maier, J. J. Am. Ceram. Soc. 2004, 87, 1173.
(249) Watanabe, M.; Uchida, H.; Seki, Y.; Emori, M. J. Electrochem. (296) Bonanos, N. Solid State Ionics 1992, 53-56, 967.
Soc. 1996, 143, 3847. (297) He, T.; Kreuer, K. D.; Maier, J. Solid State Ionics 1997, 95, 301.
(250) Malhotra, S.; Datta, R. J. Electrochem. Soc. 1997, 144, L23. (298) He, T.; Kreuer, K. D.; Baikov, Y. M.; Maier, J. Proc. Electrochem.
(251) Tazi, B.; Savadogo, O. Electrochim. Acta 2000, 45, 4329. Soc. 1997, 97, 1057.
(252) Tazi, B.; Savadogo, O. J. New Mater. Electrochem. Systems 2001, (299) Gode, P.; Lindbergh, G.; Sundholm, G. J. Electroanal. Chem.
4, 187. 2002, 518, 115.
(253) Bonnet, B.; Jones, D. J.; Rozière, J.; Tchicaya, L.; Alberti, G. J. (300) Zang, L.; Ma, C.; Mukerjee, S. Electrochim. Acta 2003, 48, 1845.
New Mater. Electrochem. Syst. 2000, 3, 87. (301) Tang, H.; Pintauro, P. N. J. Appl. Polym. Sci. 2001, 79, 49.
(254) Weng, D.; Wainright, J. S.; Landau, U.; Savinell, R. F. J. (302) Wycisk, R.; Pintauro, P. N. J. Membr. Sci. 1996, 119, 155.
Electrochem. Soc. 1996, 143, 1260. (303) Carter, R.; Evilia, R. F.; Pintauro, P. N. J. Phys. Chem. B 2001,
(255) Edmondson, C. A.; Stallworth, P. E.; Chapman, M. E.; Fon- 105, 2351.
tanella, J. J.; Wintersgill, M. C.; Chung, S. H.; Greenbaum, S. (304) Young, S. K.; Mauritz, K. A. J. Polym. Sci., Part B: Polym. Phys.
G. Solid State Ionics 2000, 135, 419. 2002, 40, 2237.
(256) Edmondson, C. A.; Stallworth, P. E.; Wintersgill, M. C.; Fon- (305) Dimitrova, P.; Friedrich, K. A.; Stimming, U.; Vogt, B. Solid State
tanella, J. J.; Dai, Y.; Greenbaum, S. G. Electrochim. Acta 1998, Ionics 2002, 150, 115.
43, 1295. (306) Antonucci, P. L.; Arico, A. S.; Creti, P.; Ramunni, E.; Antonucci,
(257) Wintersgill, M. C.; Fontanella, J. J. Electrochim. Acta 1998, 43, V. Solid State Ionics 1999, 125, 431.
1533. (307) Kerres, J. J. Membr. Sci. 2001, 185, 3.
(258) Zawodzinski, T. A.; Neeman, M.; Sillerud, L. O.; Gottesfeld, S. (308) Kerres, J.; Ullrich, A.; Hein, M. J. Polym. Sci., Part A: Polym.
J. Phys. Chem. 1991, 95, 6040. Chem. 2001, 39, 2874.
(309) Kerres, J.; Van Zyl, A. J. J. Appl. Polym. Sci. 1999, 74, 428. (328) Cornet, N.; Diat, O.; Gebel, G.; Jousse, F.; Marsacq, D.; Mercier,
(310) Kerres, J.; Zhang, W.; Cui, W. J. Polym. Sci., Part A: Polym. R.; Pineri, M. J. New Mater. Electrochem. Syst. 2000, 3, 33.
Chem. 1998, 36, 1441. (329) Genies, C.; Mercier, R.; Sillion, B.; Petiaud, R.; Cornet, N.; Gebel,
(311) Kerres, J.; Cui, W.; Reichle, S. J. Polym. Sci., Part A: Polym. G.; Pineri, M. Polymer 2001, 42, 5097.
Chem. 1996, 34, 2421. (330) Blachot, J. F.; Diat, O.; Putaux, J.-L.; Rollet, A.-L.; Rubatat, L.;
(312) Feichtinger, J.; Kerres, J.; Schulz, A.; Walker, M.; Schumacher, Vallois, C.; Müller, M.; Gebel, G. J. Membr. Sci. 2003, 214, 31.
J. J. New Mater. Electrochem. Syst. 2002, 5, 155. (331) Piroux, F.; Espuche, E.; Mercier, R.; Pineri, M. J. Membr. Sci.
(313) Kerres, J.; Zhang, W.; Jörissen, L.; Gogel, V. J. New Mater. 2003, 223, 127.
Electrochem. Syst. 2002, 5, 97. (332) Harrison, W. L.; Wang, F.; Mecham, J. B.; Bhanu, V. A.; Hill,
(314) Kerres, J.; Ullrich, A.; Meier, F.; Häring, T. Solid State Ionics M.; Kim, Y. S.; McGrath, J. E. J. Polym. Sci., Part A: Polym.
1999, 125, 243. Chem. 2003, 41, 2264.
(315) Kerres, J.; Ullrich, A.; Hein, M.; Gogel, V.; Friedrich, K. A.; (333) Miyatake, K.; Zhou, H. Y.; Watanabe, M. Macromolecules 2004,
Jörissen, L. Fuel Cells 2004, 4, 105. 37, 4956.
(316) He, R.; Li, Q.; Xiao, G.; Bjerrum, N. J. J. Membr. Sci. 2003, 226, (334) Susan, M. A. B. H.; Noda, A.; Mitsushida, S.; Watanabe, M.
169. Chem. Commun. 2003, 8, 938.
(317) Ma, Y.-L.; Wainright, J. S.; Litt, M. H.; Savinell, R. F. J.
Electrochem. Soc. 2004, 151, A8. (335) Honma, I.; Nakajima, H.; Nishikawa, O.; Sugimoto, T.; Nomura,
(318) Lassègues, J. C.; Grondin, J.; Hernandez, M.; Marée, B. Solid S. Solid State Ionics 2003, 162-163, 2003.
State Ionics 2001, 145, 37. (336) Honma, I.; Nakajima, H.; Nishikawa, O.; Sugimoto, T.; Nomura,
(319) Scharfenberger, G. Ph.D. Thesis, University Mainz, Germany, S. J. Electrochem. Soc. A 2003, 150, A616.
2003. (337) Kreuer, K. D.; Hampele, M.; Dolde, K.; Rabenau, A. Solid State
(320) Noda, A.; Susan, A. B.; Kudo, K.; Mitsushima, S.; Hayamizu, Ionics 1988, 28-30, 589.
K.; Watanabe, M. J. Phys. Chem. B 2003, 107, 4024. (338) Alberti, G.; Costantino, U.; Casciola, M.; Ferroni, S.; Massinelli,
(321) Guo, Q.; Pintauro, P. N.; Tang, H.; O’Connor, S. J. Membr. Sci. L.; Staiti, P. Solid State Ionics 2001, 145, 249.
1999, 154, 175. (339) Baranov, A. I.; Shuvalov, L. A.; Shchagina, N. M. JETP Lett.
(322) Yang, Y.; Shi, Z.; Holdcroft, S. Macromolecules 2004, 37, 1678. 1982, 36, 459.
(323) Jones, D. Presented at EUROMAT 2003, Lausanne, Switzerland. (340) Haile, S. M.; Boysen, D. A.; Chisholm, C. R. I.; Merle, S. M.
(324) Schuster, M.; Rager, T.; Noda, A.; Kreuer, K. D.; Maier, J. Fuel Nature 2001, 410, 910.
Cells 2004, in press. (Invited paper.) (341) Boysen, D. A.; Uda, T.; Chisholm, C. R. I.; Haile, S. M. Science
(325) Campbell, A. N.; Kartzmark, E. M.; Bisset, D.; Bednas, M. E. 2004, 303, 68.
Can. J. Chem. 1953, 31, 303. (342) Coors, W. G. J. Power Sources 2003, 118, 150.
(326) Miyatake, K.; Iyotani, H.; Yamamoto, K.; Tsuchida, E. Macro- (343) Kreuer, K. D. ChemPhysChem 2002, 3, 771.
molecules 1996, 29, 6969.
(327) Yang, Y.; Holdcroft, S. Fuel Cells 2004, in press. CR020715F

Batteries

Uploaded by

Copyright:

Available Formats

Batteries

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

Batteries

Uploaded by

Copyright:

Available Formats

Volume 104, Number 10

Introduction: Batteries and Fuel Cells

What Are Batteries, Fuel Cells, and Supercapacitors?

a tank, for example, fuels such as hydrogen and

Figure 5. Comparison of the driving ranges for a vehicle

discharged state, they have to be charged first before 1.3. Thermodynamics

constant for the electrode reaction, and A is the

Figure 7. (A, top) Simple battery circuit diagram, where

electrode structures from the older lead acid automo-

Table 1. Estimated Battery Market in 2003 ($ Millions

total battery market 48200

Table 2. Common Commercial Battery Systems

Figure 13. Energy storage capability of common com-

viscosities than aqueous electrolytes. Organic solvent- 2.4. Primary Batteries

Figure 16. Two-step discharge curve of a MnO2 electrode

Zinc manganese batteries consist of MnO2, a proton

system has captured the hearing aid market. Cells

sulfuric acid and water have to be provided for the

materials but with high-rate performance for hybrid

2.6. Selection Criteria for Commercial Battery

Figure 21. Summary of the reactions and processes that

Table 3. Types of Fuel Cells

cell compartment but supplied continuously from an

phosphoric acid in SiC matrix

stabilized zirconia support

molten Li2CO3 in LiAlO2-

H2 + 1/2O2 ) H2O E° ) 1.229 V (18)

porous cermet of Ni or Co oxygen or air

anode: H2 - 2e- ) 2H+ (19) carbon/platinum catalyst

cathode: O2 + 2H+ + 2e- ) H2O2 (20)

H2O2 + 2H+ + 2e- ) 2H2O (21)

overall: O2 + 4H+ + 4e- ) 2H2O (22)

Oxygen undergoes a two-step indirect reduction

reaction. The stable H2O2 intermediate is undesir-

the decomposition of H2O2 to reduce its impact on the

overall reaction. Similarly, a catalyst can enhance the

H2 dissociation rate at the anode. Platinum or

speed the reactions on both electrodes in low-tem-

tured Pt, the Pt may be distributed on a porous high

surface area carbon. Typical catalyst loadings are

on the anode and ∼0.5 mg/cm2 catalyst on the

cathode. However, the catalyst loading varies, de-

tolerance, and to identify lower cost catalysts. As the

O2 and H2 dissociation kinetics are better at higher

Fuel cell systems are not very volume efficient. The

electrolytes, as they are soluble in acidic media. The

CO and H2 stemming from CH4 conversion) are

then: CO + H2O ) CO2 + H2 (34)

then: H2 + CO32- - 2e- ) CO2 + H2O

cathode: O2 + 2CO2 + 4e- ) 2CO32- (37)

reactivity with the surface. The outer Helmholtz

Figure 29. Depiction of the charging process of a symmetric capacitor.

resistance of the cell. For capacitors in series

If Ca ) Cc, as would be expected for an ultracapacitor,

High-surface-area carbon is the material of choice,

Lithium Batteries and Cathode Materials

Received June 16, 2004

Contents 6. Conclusions and What Does the Future Hold 4297

example, in cell phones, laptop computers, and MP3

lithium iodine cell. The majority of the implantable

s ) (1 - x2)1 + x22 (5)