Defects and Transport in Crystalline Solids

Per Kofstad
and Truls Norby

Compendium for the advanced level course
Defect Chemistry and Reactions in Solids
KJM 5120 and KJM 9120 (formerly KJM 4120 and KJ-MV 417)

at
Department of Chemistry
Faculty of Mathematics and Natural Sciences
University of Oslo

2007

Version: September 2007
UNIVERSITY
OF OSLO
1. Structure and defects
12
13
Preface

In 1972 Per Kofstad published his "Non-stoichiometry, diffusion and electrical
conductivity in binary metal oxides". It has been a popular textbook in defect chemistry
of oxides worldwide, not least because it contained a comprehensive review of defect
structure and defect-related properties of all binary metal oxides. It followed Kofstad's
equally well-recognized book "High temperature oxidation of metals" from 1967,
revised and published under the title "High temperature corrosion" in 1987.
In the 25 years that have passed since 1972 the field of defect chemistry in oxides
has grown and developed in many ways. New and improved measurement techniques,
new synthesis techniques, as well as skilled and devoted investigators have produced
better data for many oxides, partly encouraged by industrial developments in the use of
solid electrolytes (solid oxide fuel cells, gas separation membranes, ceramic electrodes,
catalysts, etc). Computer modelling enables detailed analysis of defect chemistry and
transport processes, and helps us interpret experimental data. Although the increasing
accuracy and complexity is welcome and unavoidable, the sound overview and ability
to analyse a situation may be obscured or lost. The newcomer may find the field
difficult to enter,
Along with the development in the application of theory and modelling, the field
has changed focus from binary oxides to ternary or higher oxides, much due to the
growing awareness that perovskite and related structures tolerate very large
substitutions and variations in valence and stoichiometry. Here we find lanthanum
aluminate insulators, titanate and zirconate electroceramics with ferroelectricity,
piezoelectricity, etc, the high-T
c
cuprate and other superconductors, high-temperature
electronic conductors such as doped lanthanum cobaltate and chromite, high
temperature ionic conductors such as barium indate and doped lanthanum gallate,
mixed conductors, and alkaline earth cerates which dissolve protons and become proton
conductors at high temperatures.
A revision of the book from 1972 as such is thus utopic in many respects. In the
early 1990s Kofstad decided to author a new text that represented a development of the
general and theoretical first part of the former book, while he intended to include only a
few examples of data from important example oxide systems. After having sketched the
book he was struck by cancer in 1993. He fought the cancer for four years, with the
obligation to finish the book as one of the driving forces. However, deprived of much
of his usual strength, he was unable to finish the work. I became engaged as a co-author
at that time with the purpose of completing the book.
I used the new text as a curriculum when I took over the teaching of Kofstads
course in defect chemistry at the University of Oslo, thereby being able to adjust and
improve it step by step. However, the burdens of taking over the research group has
kept me from finishing the book. Furthermore, a textbook on defect chemistry by
Kofstads close friend and colleague, Donald M. Smyth was published that took away
some of the immediate need for our textbook. All in all I have felt that the new book
needs more work and refinement. With the present text I am still in the middle of that
process, and feedback is thus as welcome as ever.

Truls Norby, February 2006
14

Copyrights and use of graphic material

Graphic material is original, except where stated. Reproduction from other
sources have in most instances not yet been authorised in the present manuscript which
is not sold commercially. Such authorisations will be collected or original artwork
created - before any commercial version of this text is printed.

The rights to the text and all original graphic material rests with Truls Norby.

The present text must not be reproduced and is intended solely for use at the
University of Oslo.

Truls Norby, February 2006.
15

Contents

Preface ......................................................................................................... 13
Copyrights and use of graphic material ..................................................... 14
Contents....................................................................................................... 15
1. STRUCTURE AND DEFECTS .................................................110
Introduction............................................................................................... 110
Common crystal structures....................................................................... 113
Close-packing and voids.......................................................................... 113
Less close-packed arrangements .............................................................. 115
Some simple structures for oxides and other ionic compounds................. 116
Point defects............................................................................................... 121
Notation for description of point defects.................................................. 121
Notation for charged defects.................................................................... 122
Electroneutrality ....................................................................................... 124
Stoichiometry and nonstoichiometry........................................................ 125
Defects in stoichiometric compounds. ..................................................... 125
Defects in nonstoichiometric oxides. ....................................................... 127
Associations of point defects, point defect clusters, extended defects...... 128
Example cases of defect clusters.............................................................. 129
Extended defects, shear structures. .......................................................... 129
Modulated structures/infinitely adaptive structures .................................. 130
Summary ................................................................................................... 131
Literature................................................................................................. 131
Exercises ................................................................................................. 132
2. DEFECT REACTIONS................................................................21
Introduction................................................................................................. 21
The three rules for writing defect reactions ............................................... 21
Mass balance............................................................................................. 21
Electroneutrality........................................................................................ 22
Ratios of regular lattice sites...................................................................... 22
Examples of reaction equations for defects ................................................ 23
Stoichiometric compounds instrinsic disorders ....................................... 23
16
Nonstoichiometry...................................................................................... 25
Dissolution of foreign elements ................................................................. 29
Ternary and higher compounds ............................................................... 214
Summary ................................................................................................... 216
Exercises ................................................................................................. 216
3. DEFECT EQUILIBRIA.................................................................. 3.1
Introduction.................................................................................................... 3.1
Point defects and equilibrium in an elemental solid - a simple statistical
thermodynamic approach....................................................................................... 3.2
Thermodynamics of chemical reactions ........................................................ 3.5
Thermodynamics and point defects............................................................... 3.8
Virtual chemical potentials of point defects .................................................. 3.8
Ideal and non-ideal models ........................................................................... 3.9
Examples of defect equilibria in pure, stoichiometric metal oxides ........... 3.10
Schottky defects ......................................................................................... 3.10
Frenkel defect pairs .................................................................................... 3.12
Intrinsic ionisation of electrons. .................................................................. 3.12
Defect equilibria in nonstoichiometric oxides ............................................. 3.14
Oxygen deficient oxides ............................................................................. 3.14
Oxide with excess metal ............................................................................. 3.17
Simultaneous presence of oxygen vacancies and interstitial metal ions ....... 3.18
Metal-deficient oxide.................................................................................. 3.20
Metal oxides with excess oxygen................................................................ 3.22
Examples of defect structures involving both oxygen deficiency and excess
................................................................................................................................ 3.22
Oxygen deficit. ........................................................................................... 3.23
Oxygen excess............................................................................................ 3.23
Stoichiometric condition............................................................................. 3.24
Summary ...................................................................................................... 3.26
Literature.................................................................................................... 3.27
Exercises .................................................................................................... 3.27
4. IMPURITIES AND DOPANTS ...................................................... 4.1
Introduction.................................................................................................... 4.1
Constant concentration of impurity/dopant.................................................. 4.3
Schottky equilibria ....................................................................................... 4.3
Frenkel equilibria ......................................................................................... 4.4
Oxygen-deficient oxides doped with lower valent cations............................. 4.5
17
Oxygen deficient oxides doped with higher valent cations............................ 4.6
Oxides with excess metal.............................................................................. 4.6
Metal-deficient oxides doped with lower valent cations................................ 4.7
Metal deficient oxides doped with higher valent cations ............................... 4.8
Doping of oxides which may have regions with both oxygen and metal deficit
.............................................................................................................................. 4.8
Variable concentration of impurity/dopant ................................................ 4.11
Doping M
2
O
3
with excess MlO................................................................. 4.12
Doping M
2
O
3
with excess of MhO
2
.......................................................... 4.13
Hydrogen defects in metal oxides ................................................................ 4.15
Effect of water vapour on oxygen-deficient M
2
O
3
. .................................... 4.17
Effect of water vapour on acceptor-doped M
2
O
3
........................................ 4.18
Association between impurities/dopants and other point defects............... 4.21
Summary ...................................................................................................... 4.22
Literature (protonic defects only)................................................................ 4.22
Exercises .................................................................................................... 4.22
5. DIFFUSION................................................................................... 5.1
Introduction.................................................................................................... 5.1
Models of diffusion......................................................................................... 5.1
Fick's first law.............................................................................................. 5.2
Potential gradients as the driving force ......................................................... 5.3
Example of chemical potential gradient acting on neutral particles ............... 5.4
Simplified model for one-dimensional diffusion........................................... 5.5
Random diffusion......................................................................................... 5.8
Fick's second law.......................................................................................... 5.10
Measurements of the diffusion coefficients by tracer techniques................. 5.12
Diffusion mechanisms .................................................................................. 5.14
Vacancy mechanism................................................................................... 5.14
Interstitial mechanism ................................................................................ 5.15
Interstitialcy mechanism............................................................................. 5.15
Diffusion of protons in metal oxides. .......................................................... 5.17
Further considerations of factors that affect the diffusion coefficient in
crystalline solids .................................................................................................... 5.18
Free proton transport .................................................................................. 5.20
Temperature and oxygen pressure dependence of diffusion in metal oxides
................................................................................................................................ 5.21
Dependencies related to the concentration of defects .................................. 5.21
Temperature dependence of the attempt frequency .................................. 5.22
Resulting analysis of the diffusion coefficient D
r
........................................ 5.24
18
Diffusion coefficients of point defects .......................................................... 5.27
Excercises .................................................................................................. 5.28
6. ELECTRICAL CONDUCTIVITY ................................................... 6.1
Introduction.................................................................................................... 6.1
Transport in an electrical potential gradient ................................................ 6.1
Charge carriers in ionic compounds.............................................................. 6.3
The Nernst-Einstein relation between mobility and diffusion coefficient .... 6.4
Electronic conductivity in oxides ................................................................... 6.7
The band theory............................................................................................ 6.8
Effects of charged impurities or defects. Extrinsic semiconductors ............. 6.12
The Fermi level and chemical potential of electrons ................................... 6.17
Charge mobilities of electrons and electron holes ....................................... 6.18
Nonstoichiometric semiconductors ............................................................. 6.21
Ionic conductivity......................................................................................... 6.25
Correlation effects: tracer diffusion and ionic conduction ........................... 6.26
Literature...................................................................................................... 6.28
Exercises ....................................................................................................... 6.29
7. ELECTROCHEMICAL TRANSPORT........................................... 7.1
Electrochemical potential............................................................................... 7.1
Flux equations Wagner theory.................................................................... 7.1
General derivation of Wagner-type flux expressions..................................... 7.1
From charged to neutral species: the electrochemical reaction ...................... 7.3
The voltage over a sample ............................................................................ 7.4
Flux of a particular species ........................................................................... 7.5
Wagner-type transport theory case studies................................................... 7.6
Oxides with transport of oxide ions and electrons ......................................... 7.6
Ambipolar diffusion and conduction........................................................... 7.11
Chemical diffusion and tracer diffusion...................................................... 7.12
General case............................................................................................... 7.12
Chemical diffusivity in the case of mixed oxygen vacancy and electronic
conductor ............................................................................................................ 7.14
Surface and interface kinetics limitations ................................................... 7.16
Ionic transport of both anions and cations.................................................. 7.22
General expressions.................................................................................... 7.22
High temperature oxidation of metals; the Wagner oxidation theory........... 7.26
19
Sintering ....................................................................................................... 7.34
Oxides with additional transport of protons ............................................... 7.36
Other cases ................................................................................................... 7.37
Excercises .................................................................................................. 7.39

110
1. 1. 1. 1. Structure and defects Structure and defects Structure and defects Structure and defects

Introduction

Classical chemistry and crystallography gave an idealised picture of the
composition and crystal structure of inorganic compounds. During the 19th century
crystallographers concluded that the atoms or structural units in crystals were arranged
in ideal structures where all the structural sites were occupied. Chemists, on their side,
expressed the compositions in terms of the law of definite proportions and generally
believed that inorganic compounds had definite, unvarying, stoichiometric
compositions determined by the valence of the constituent atoms. However, there was
still considerable discussion regarding these concepts and it was alternatively proposed
that inorganic compounds can have variable compositions. This was clearly established
in the beginning of the 20th century and the question was theoretically resolved in the
1930's when Wagner and Schottky (1930) showed through statistical thermodynamic
treatments of mixed phases that crystal structures are not ideal. Some lattice sites will
be empty (vacant) and extra atoms may occupy the interstitial space between the atoms
on the lattice sites. The empty lattice sites are termed vacancies and the extra atoms
interstitial atoms. Following Wagner and Schottky all crystalline solids will at any
temperature contain vacancies and extra atoms and will as such exhibit deviations from
the ideal structure. Furthermore, all inorganic compounds may in principle have
variable composition and thus be nonstoichiometric. In fact, the very exact
stoichiometric composition is an exception rather than a rule and can under equilibrium
conditions only be achieved at definite sets of temperature and activities of the
components in a crystal. However, for many inorganic compounds the deviation from
stoichiometry (the nonstoichiometry) is so small that it is difficult or impossible to
measure directly with presently available experimental techniques.
These deviations from the ideal structures are present at any temperature and
occur naturally in all crystalline compounds. Even so these deviations or imperfections
are called defects. The reason for this is that by convention the ideal structure is used as
the reference state, and any deviation from this ideal state is termed a defect.
Crystalline solids contain different types of structural defects. If the imperfection
is limited to one structural or lattice site and its immediate vicinity, the imperfection is
termed a point defect. Vacancies and interstitial atoms are point defects. An impurity
atom present in a crystal and that either occupies a lattice site or an interstitial site is
also termed a point defect. But in addition to the point defects the structural defects also
comprise line and plane defects. The line defects are dislocations which are
characterised by displacements in the structure in certain directions. The plane defects
comprise stacking faults, grain boundaries, internal and external surfaces.
111

Figure 1-1. Schematic representation of some defects in an elemental solid. From A. Almar-
Nss: Metalliske materialer.

In addition to the structural defects, crystals also contain electronic defects, i.e.
electrons and electron holes that are relatively free to move in the crystal. The
electronic defects may either be formed through internal excitation of valence electrons
or they may be formed in association with point defects. If these electronic defects are
localised (trapped) at regular sites in the structure, the electronic defects are termed
polarons or - from a chemical point of view - valence defects. Defect electrons or
electron holes trapped at point defects often make otherwise transparent materials
coloured, and composite defects involving point defects and trapped electronic defects
are termed colour centres.
A complete description of the point and electronic defects in a compound and
their concentrations as a function of the activities (partial pressures) of the constituents
and the temperature is often termed the defect structure of the compound.
The reason for our interest in and concern with defects and imperfections is that
they determine a number of properties of crystalline solids. Thus diffusion or ionic
conductivity in crystalline solids takes place because of the presence of defects. Point
defects are responsible for lattice diffusion, dislocation diffusion takes place along line
defects and grain boundary and surface diffusion along planar defects. Solid state
diffusion, in turn, determines or strongly influences a number of properties or processes
such as mass transport in solids, solid state reactions, sintering, high temperature creep,
gas-metal reactions resulting in the formation of compact layers of reaction products,
etc. Electronic defects determine properties such as electronic conductivity, electro-
optical properties, etc. Heterogeneous catalysis is closely related to the defect chemistry
of catalysts.
112

Figure 1-2. Schematic representation of some defects in an ionic binary compound.

As a basis for further considerations of defect structures and defect-dependent
properties to be discussed later in the book, the remainder of this chapter briefly
discusses some commonly occurring crystal structures in simple metal oxides, the
notation of point defects, and qualitative descriptions of different types of defect
structure situations.

113
Common crystal structures

Close-packing and voids

The crystal structures of many metals can be described by close-packing of
spheres (atoms). Similarly, many simple oxides to be discussed later can conveniently
be considered to consist of close-packing of oxygen ions while the metal ions occupy
voids in the close-packed structures. Let us therefore briefly discuss close-packing of
spheres (atoms or ions).
Consider an atom as a small hard sphere and make a layer of identical atoms so
that the empty space between the atoms is minimum (layer A). Let us now add a second
layer (B) such that the atoms in this second layer sit in one set of the hollows of the first
layer. When we add a third layer of spheres, the spheres can be placed in two different
positions: the spheres in the third layer can be placed directly over the spheres in layer
A, and if we continue this stacking, the stacking sequence becomes ABABAB... as
illustrated in (a) of the figure below. This type of close-packed stacking is called
hexagonal close packing (hcp) of the spheres because it gives crystal structures with
hexagonal symmetry.

Figure 1-3. ABABAB(hcp) clos-packing (a) and ABCABC(fcc) close-packing (b) of spheres.
From Shriver and Atkins; Inorganic Chemistry.

The third layer (layer C) could alternatively be placed such that the spheres in this
layer will not be directly over either the atoms in the A or B layers, see (b) in the figure
above. The stacking sequence is now ABCABC... and this close packing is known as
cubic close packing (ccp) as this gives rise to a cubic (face-centred) crystal structure.
The following figure shows how the two close-packed structures appear when their
heaxonal and face-centred cubic unit cells are emphasized.
114

Figure 1-4. Schematic views of the hcp (a) and fcc (b) structure types. From Shriver and Atkins;
Inorganic Chemistry.

The close-packed structures contain two different types of empty space or voids
between the spheres, cf. the figures below. One type of void is termed a tetrahedral void
because each of these voids is surrounded by four spheres which centres are at the
corners of a tetrahedron. The other type of void is called an octahedral void. In this case
each void is surrounded by six spheres, three in one layer and three in the neighbouring
layer, and the centres of these six spheres lie at the corners of an octahedron. In the
close-packed structures there are two tetrahedral voids and one octahedral void per
atom (sphere). It may be noted that an octahedral void will fit a sphere with radius
0.414r where r is the radius of the bigger close-packed spheres. A tetrahedral void is
smaller and may fit a sphere with radius 0.225r. In different crystal structures of simple
metal oxides which can be considered to consist of a close-packing of oxygen atoms,
the metal ions often occupy all or part of the tetrahedral and/or octahedral interstices in
regular patterns.

Figure 1-5. Octahedral and tetrahedral hole in close-packed layers of speheres. From Shriver and
Atkins: Inorganic Chemistry.
115

Figure 1-6. Small spheres show locations of octahedral (left) and tetrahedral (right) voids in fcc
structure. From Shriver and Atkins: Inorganic chemistry.

Less close-packed arrangements

Some metals and ionic compounds choose for various reasons to arrange
themselves in less close-packed structures, as exemplified in the following two figures.

Figure 1-7. Body-centered cubic (bcc) packing of spheres. From Shriver and Atkins: Inorganic
Chemistry.

Figure 1-8. Simple cubic (sc) packing of spheres. From Shriver and Atkins: Inorganic Chemistry.

116
Some simple structures for oxides and other ionic compounds

As mentioned earlier, the structures of ionic compounds can often be seen as
close-packing of the large anions, while the normally smaller cations occupy the
interstitial voids. (Note that once the structure is established, the interstices occupied by
cations are not any longer considered interstitial; they are part of the ideal (reference)
structure.)
In judging the packing of ions it is useful to recall some important principles of
ionic radii; the size of the elements increase down a group of the periodic table
(resulting from the larger orbital of the outermost shell). Further, the size of the
elements as a rule of thumb decreases from left to right through a period of the periodic
table (resulting from increased nuclear charge). Finally, and most importantly in this
context: Negatively charged ions (anions) are much larger than their neutral atoms and
positively charged ions (cations) are much smaller than their neutral atoms. One may
also recall that the effective size increases with the coordination number. In the
following we look at some example structures possessed by oxides.

The NaCl (rock salt, halite) structure
In the NaCl structure the anions are cubic close-packed, and the smaller metal ions
occupy all the octahedral interstices. Each metal ion is thus surrounded by six anions
and vice versa. The NaCl structure is illustrated below. Note the similarity with the
illustration of the octahedral voids in the fcc structure shown previously.

Figure 1-9. The NaCl structure. Note that the centre anion is six-coordinated, as are also all
other ions. From Shriver and Atkins: Inorganic Chemistry.

The oxides MgO, CaO, SrO, BaO, CdO, CoO, NiO, FeO, TiO, NbO, and VO
possess this crystal structure. The last six of these oxides may be highly
nonstoichiometric and as such their NaCl structures highly defective.
In the sulfides FeS and NiS, the structure is similar, but the close-packing of
anions is hexagonal rather than cubic.

117
The ZnS (zinc blende and wurtzite) structures
These structures are named after two different mineral forms of zinc sulphide
(ZnS). The zinc blende structure can be considered as a cubic close-packing of sulphide
ions with the zinc ions occupying every other tetrahedral void. Each zinc ion is thus
tetrahedrally coordinated by four sulphide ions and vice versa.
The wurtzite structure is composed of hexagonal close-packed sulphide ions and
where zinc ions occupy every other tetrahedral void. The metal oxides BeO and ZnO
adopt the wurtzite structure.

The CaF
2
(fluorite) structure
The fluorite structure is named after the mineral calcium fluoride (CaF
2
) and is
illustrated below. This structure may geometrically be considered to constitute a close-
packing of the metal ions while the anions occupy all the tetrahedral sites. The calcium
ions are smaller than the fluoride ions and the fluoride ions are thus not able to fit into
the interstices of the calcium ion array, but the model describes the relative positions of
the ions. It is nevertheless more correct to view the structure as a simple cubic packing
of anions (not close-packed), with cations filling every second cubic void, as also
depicted below; each cation is surrounded by eight anions.

Figure 1-10. The CaF
2
structure. (a) The cations form an fcc sublattice, while the anions form a
simple cubic arrangement. Every second anion cube has a cation in its centre. (b) The anions can be
taken to occupy tetrahedral voids in the cation sublattic.

The fluorite structure is typical for AX
2
compounds where A is a sufficiently
large cation that it forces the anions apart and into a less than close-packed
arrangement. The oxides ZrO
2
, HfO
2
, CeO
2
, ThO
2
, PuO
2
, etc. possess the fluorite
structure. All the octahedral interstices are unoccupied for the stoichiometric
compositions, but interstitial ions may be accommodated in these interstices and form
oxides with excess oxygen (hyper-stoichiometric oxides), e.g. UO
2+x
.
Pyrochlore (ABO
3.5
) and C-type rare earth sesquioxide (A
2
O
3
) structures can be
viewed as fluorite-derived structures with, respectively, 1/8 or 1/4 of the oxygen ions
missing (in an ordered fashion).
118
The alkali metal oxides Li
2
O, Na
2
O, K
2
O, and Rb
2
O possess the so-called anti-
fluorite structure. In this structure the oxygen ions can be considered to be cubic close-
packed (fcc) while the metal ions occupy all the tetrahedral voids.

The rutile (TiO
2
) structure
The rutile crystal structure has its name after the rutile modification of titanium
dioxide, TiO
2
. It is based on a hcp packing of the oxygen ions, with Ti ions occupying
every second octahedral void, so that the structure can be viewed as consisting of TiO
6

octahedra (slightly distorted) which share edges and corners in such a way that each
oxygen atom belongs to three neighbouring octahedra. The structure is tetrahedral, and
when the structure is viewed along the c-axis it may be seen that the structure contains
channels in the c-direction, i.e. rows of octaedra not filled with cations.

Figure 1-11. The rutile TiO
2
structure. From Shriver and Atkins: Inorganic chemistry.

In addition to rutile itself, the oxides SnO
2
, MnO
2
, VO
2
, MoO
2
, RuO
2
, GeO
2
a.o.
possess regular or distorted rutile structures.

The corundum (-Al
2
O
3
) and ilmenite (FeTiO
3
) structures
The corundum structure is named after the high temperature modification of
alumina, -Al
2
O
3
. This structure may be described as a hexagonal close-packing (hcp)
of oxygen ions with the trivalent aluminium ions occupying 2/3 of the octahedral sites.
As the metal ions occupy octahedral sites, each metal ion is octahedrally coordinated
and surrounded by six oxygen atoms, while each oxygen atom is surrounded by four
metal ions. Two and two metal ions occupy neighbouring interstitial octahedral sites,
and the two corresponding AlO
6
octahedra are linked by common faces. Other oxides
possessing the corundum structure are -Fe
2
O
3,
Cr
2
O
3
, Ti
2
O
3
, and V
2
O
3
.
A number of oxides with the nominal formula ABO
3
have the corundum
structure when A and B have an average valence of 3 and are of approximately the
same size. The ilmenite structure with the nominal formula ABO
3
and named after the
mineral Fe
II
Ti
IV
O
3
is similar to the corundum structure in that the oxygen ions can be
described as hexagonal close-packed, the valences of the two cations have an average
valence of 3 and occupy 2/3 of the octahedral sites. However, in this case the two metal
ions (Fe
2+
and Ti
4+
) are arranged in alternate layers.
119

The spinel structures
The spinel structure is named after the mineral spinel, MgAl
2
O
4
. The spinels
have the general formula AB
2
O
4
, where A is most often a divalent ion, A
2+
, and B
trivalent, B
3+
. The oxygen ions can be considered to be cubic close packed (fcc), and
the metal ions occupy both tetrahedral and octahedral sites. The unit cell contains 32
oxygen atoms, 8 AB
2
O
4
units. It contains 32 octahedral and 64 tetrahedral voids. In the
normal spinel structure one half of the octahedral voids are occupied by the B
3+
-ion
while the A
2+
ions occupy 1/8 of the tetrahedral voids.
Other distributions of the cations occur. When the eight tetrahedral sites per unit
cell are occupied by trivalent cations and the octahedral sites are shared by the divalent
and trivalent ions, the structure is termed an inverse spinel. Magnetite, Fe
3
O
4
, is an
example of an inverse spinel. An inverse spinel may be better written as B(AB)O
4
as
this indicates that half of the B
3+
ions occupy tetrahedral sites and the other half
together with the A
2+
ions occupy the octahedral sites. Another example of an oxide
with inverse spinel structure is Fe(MgFe)O
4
.
If the general formula of the spinel is written AB
2
O
4
, the cations may have the
valence A
2+
and B
3+
as discussed above. This is termed a 2-3 spinel. In addition one
may have spinels with other cation valences, e.g. A
4+
and B
2+
(4-2) spinel) and with
A
6+
and B
1+
. In all cases the sum of the cation valences equals 8.

The perovskite (CaTiO
3
) structure

The perovskite structure has its name after the mineral perovskite, CaTiO
3
. The
general formula may be written ABO
3
in which A is a large cation (e.g. alkali, alkaline
earth, rare earth ions) and B a small or medium sized metal ion often a transition metal
ion. In this structure the large A atoms and the oxygen atoms together form a close-
packed lattice, such that each A-atom is coordinated to twelve oxygen atoms. The B-
atoms occupy octahedral sites surrounded by six oxygen ions.

Figure 1-12. Perovskite ABX
3
structure. Left: Shown as a fcc packing of A (corners) and X (faces)
atoms, with the B in the center. Right: The A cation is 12-coordinated.( The A cation is depicted small,
but is in reality much bigger than the B cation). From Shriver and Atkins: Inorganic chemistry.

120
The average valence of the A and B ions is +3. The perovskite structure is
remarkable in being very tolerant to variations (as long as A is large and B is small).
Thus, A+B valences can be 3+3 (e.g., LaAlO
3
), 2+4 (e.g. CaTiO
3
), 1+5 (e.g. KTaO
3
)
and even 0+6 (where the A site is in fact empty, as in WO
3
).
The perovskite structure tolerates relatively large nonstoichiometries and
substitutions, while a number of perovskite-related structures (with perovskite building
blocks) are formed when excesses, deficiencies or substitutions beyond the solubility
limits lead to ordered defects and new phases. These comprise, for instance, oxygen
deficiency (brownmillerite, ABO
2.5
), and A excess (K
2
NiF
4
-type structure, A
2
BO
4
)
and ordered intermediate phases.

Summary of relations between structures, radii and packing arrangements.
In oxides with small cations only, the structure is dominated by close-packing of
the oxygen ions and various occupancies of interstices by the small cations. The table
below summarises relations between structures and close-packed arrangements of
anions or cations in oxides (and some sulphides).

Table 1-1. Some structures of oxides based on close-packed oxygen ion sublattices.
Formula Cation:anion
coordination
Type and number of
occupied interstices
fcc of anions hcp of anions
MO 6:6 1/1 of octahedral
sites
NaCl, MgO, CaO,
CoO, NiO, FeO
a.o.
FeS, NiS
MO 4:4 1/2 of tetrahedral
sites
Zinc blende: ZnS Wurtzite: ZnS,
BeO, ZnO
M
2
O 8:4 1/1 of tetrahedral
sites occupied
Anti-fluorite:
Li
2
O, Na
2
O a.o.

M
2
O
3,

ABO
3

6:4 2/3 of octahedral
sites
Corundum:
Al
2
O
3
, Fe
2
O
3
,
Cr
2
O
3
a.o.
Ilmenite:
FeTiO
3

MO
2
6:3 of octahedral sites Rutile: TiO
2
,
SnO
2

AB
2
O
4
1/8 of tetrahedral and
1/2 of octahedral
sites
Spinel: MgAl
2
O
4

Inverse spinel:
Fe
3
O
4

When the cations get somewhat larger they force the anions apart, and we get
non-close-packed structure types like the fluorite and its derivatives (pyrochlore, rare-
earth oxide structures, etc.)
When we involve even larger cations, being of the same size as the anions, the
large cations take part in the close-packing together with the anions. In this way we get
the perovskite structure and its derivatives.
121
More detailed treatments on crystal structures of oxides are found in general texts
on solid state chemistry (e.g. Smart and Moore (1992) or West (1984)) and in more
specialised literature, such as Nray-Szab (1969) and Rao and Raveau (1995).

Point defects

In order to understand and describe properties of inorganic compounds that are
dependent on the presence of point defects, it is necessary to be able to express the
concentrations of the defects under various conditions. This, first of all, requires a
system of notation to describe the point defects and the constituent atoms of the
inorganic compounds and their charges.

Notation for description of point defects

In the development of the field of defect chemistry of inorganic compounds
various systems of notation have been proposed and used to describe point defects.
However, the most widely adopted system is that due to Krger and Vink (1956) (see
also Krger (1964)), and this will be used in this book. This system describes crystals in
terms of structural elements, and an imperfection is indicated by a major symbol
describing its chemical content and a subscript that indicates the site that it occupies.
In an oxide the metal ions on the regular lattice positions are written M
M
, where
the subscript thus describes the type of lattice site which is occupied. Correspondingly,
the oxygen ions on normal lattice sites are written O
O
.
The native point defects in an oxide include M and O vacancies and M and O
interstitial atoms or ions. Vacancies are denoted by v with a subscript M or O referring
to vacant metal or oxygen sites, respectively. Interstitial ions or sites are described with
a subscript "i". Vacancies and interstitials in an oxide are thus written
v
O
oxygen vacancy,
v
M
metal vacancy,
O
i
oxygen interstitial,
M
i
metal interstitial.
Correspondingly, an unoccupied or vacant interstitial site can be written v
i
.
While the notation traditionally has used V for vacancy, confusion with the
element vanadium (V) can be avoided by using lower-case v or italic V. Here, we will
use the lower case v.
Component atoms of a crystal may find themselves on the wrong site; the A and
B cations in perovskite, pyrochlore, spinel, etc. may thus swap sites so as to obtain A
B

and B
A.
An oxygen atom in a cation site is similarly written O
M
, but is an example of a
very unstable defect.
122
Real crystals always contain impurities. When a foreign metal ion Mf occupies a
regular M site, the foreign atom is denoted by Mf
M
. If it occupies an interstitial site, it
is described by Mf
i
.
Notations for other or more complex defects will be introduced when needed.

Notation for charged defects

In an ionic compound the atoms are charged, and the cations and anions may be
assigned a definite integer valence (formal oxidation number). This is in fact a fully
valid model in defect chemistry, as will be argued in a discussion on the influence of
the varying degree of ionicity, or distribution of valence electrons, in a later paragraph.
In the following, oxygen ions on regular sites in an oxide will be considered to
have a formal valence of -2. The cations have a positive charge so as to make the sum
of all positive and negative charges in the compound equal to zero.
The point defects may be neutral or charged. Their charges may be described in
terms of their actual charge or valence. However, it is generally more convenient in
writing defect reactions to use the ideal, perfect crystal as the reference state and
consider the charge on the defects relative to this perfect crystal. This relative charge is
termed the effective charge of the defect. With the perfect crystal as reference, the
normal atoms on regular lattice sites have zero effective charge. To emphasise this zero
effective charge, an "x" is often added as a superscript to the symbol. The normal
cations and anions in an oxide are then written
x
M
M and
x
O
O .
When vacancies are present, part or all of the actual charge of the missing atom
may be excited away from the vacant site. Let us, for instance, consider the formation
of an oxygen vacancy by removing an oxygen atom from the crystal. If the two
electrons of the oxygen ion are localised at or near the empty site, the vacancy and its
immediate neighbourhood have two negative actual charges. The charge at this location
is thus the same as in the perfect crystal and the oxygen vacancy then has zero effective
charge. As such it is termed a neutral oxygen vacancy, which is written
x
O
v .
Now that the electronegative oxygen atom is no longer present to hold the extra
electrons, one or both of the two electrons may be excited and transferred away from
the oxygen vacancy, and the oxygen vacancy then becomes singly or doubly charged.
Since electrons are removed, the ionised vacancy has an effective positive charge
relative to the perfect crystal, compensated by the two electrons which are now
negatively charged defects somewhere out in the crystal.
To distinguish effective from actual charges, different symbols are used. While
the actual charges are written + or -, the effective positive charge is indicated by a
supscript dot (
) and the effective negative charge by a superscript prime (

/
). Singly and
doubly charged oxygen vacancies are, for instance, written
O
v and

O
v , respectively.
Similarly, cation vacancies may be neutral or have negative effective charges. To
illustrate this let us remove a metal atom from the oxide MO and create a vacant metal
ion site. The M
2+
ion picks up two electrons to leave as an atom, leaving behind two
positive charges in the form of holes or valence defects. If these are localised at
123
neighbouring oxygen ions (as O
-
) or metal ions (as M
3+
) the actual charge at the
vacancy and its immediate neighbourhood is then the same as in the perfect crystal, and
the effective charge on the metal vacancy is considered to be zero, i.e. the metal
vacancy is neutral,
x
M
v . One or both of the positive charges may be excited and
transferred away from the vacancy and distributed in the oxide, and the vacancy then
has one or two negative effective charges and is written as
/
M
v or
//
M
v , respectively.
Similarly, interstitially dissolved atoms will have zero real and effective charge,
but depending on their nature they can be ionised to cations with a real and effective
positive charge or to anions with real and effective negative charge.
Substitutionally dissolved foreign ions which tend to have a valence different
from that of the normal ions (aliovalent foreign ions) are easily ionised to get an
effective charge. For instance, if the valence of a substitutionally dissolved foreign
cation is higher than that of the normal host cation, the foreign cation gets a positive
effective charge. Conversely, if the valence of the substitutionally dissolved foreign ion
is smaller than the host cation, the foreign cation gets a negative effective charge.
Common examples are found in the semiconductor field. Boron dissolved in silicon is
denoted
x
Si
B in the unexcited state, where we have B
0
. But since boron has one valence
electron less than the surrounding Si
0
atoms it easily accepts one from the valence band
of the Si lattice and becomes B
-
or, in Krger-Vink notation,
/
Si
B . Boron is called an
electron acceptor (acceptor dopant) and the lacking electron in the valence band is
called a hole. In a similar way, phosphorous forms a donor in silicon;
x
Si
P and the
excited state is
x
Si
P .

Figure 1-13. Schematic view of electronic structure of boron- and phosphorous-doped silicon.
From http://acre.murdoch.edu.au/refiles/pv/text.html

Defect electrons and electron holes that are free to move in the oxide have
effective negative and positive charges, respectively. They are written
/
e and
h . If the
electron, for instance, is associated with a cation on a regular site - and may as such be
considered a valence defect - the defect may be written
/
M
M .
124
We have mentioned but a few types of defects to illustrate the principles of
notation. Many other defects can be described in the same way; we will see more
defects later on.
In addition to the notation, the important thing to remember about charges is the
following: The effective charge of a species is the real charge of the species minus the
real charge the perfect reference lattice would have had in the same volume element.

Electroneutrality

Crystals that we will deal with are considered to be electrically neutral. Then the
sum of all positive charges must equal the sum of all negative charges:
positive charges = negative charges
We may thus sum up the real charges of all cations and anions and defects. However, it
is equivalent and more convenient to consider only the effective charges:
positive effective charges = negative effective charges
It is important to note that the number of positive and negative charges must be
compared on equal bases, namely as concentrations (numbers or moles per volume).
One may in general not use site fractions since the concentrations of different types of
sites are not necessarily equal. Thus, the electroneutrality condition can be expressed as

0 ] z[ =
S
z
S (1.1)

where the brackets denote concentration and z is the number and sign of charges on the
species S
z
. For instance, in an oxide MO where the major defects are oxygen vacancies,
metal interstitials, and defect electrons, the electroneutrality condition can be written

] [e ] 2[M ] 2[v or 0 ] [e - ] 2[M ] 2[v
/
i O
/
i O
= + = +

(1.2)

Despite its simplicity, and despite its being used in e.g. traditional aqueous
chemistry, the electroneutrality condition tends to cause both conceptual difficulties and
occasional errors (by misplacing the coefficients z). In order to help, we suggest that
one learns it by heart and uses it accordingly. Secondly, that one realises that it is not a
chemical reaction, but a mathematical relationship. It thus maintains no mass balance or
charge balance, it simply counts positive and negative charges from whatever
concentrations of charged species that are present and requires them to be equal.

125
Stoichiometry and nonstoichiometry

The compositions and chemical formulae for metal oxides and other inorganic
compounds are usually written with a definite ratio of cations to anions, e.g. M
a
O
b

where a and b are usually small integers determined by the valence of the constituent
atoms. In crystalline compounds this also reflects that the structure contains different
types of sites (e.g. close-packed sites and tetrahedral or octahedral interstices) in simple
ratios and that these are selectively and systematically filled with cations or anions.
When the oxide M
a
O
b
contains M and O atoms in the exact ratio a:b, it is said to have a
stoichiometric composition.
As mentioned above, an exact stoichiometric composition in inorganic
compounds is in principle the exception rather than the rule. Oxides in equilibrium with
their surroundings are generally nonstoichiometric, except under specific conditions of
temperature and activities of the components. However, within our experimental ability
to measure the ratio of the constituent atoms, many inorganic compounds may be
considered to be so near stoichiometry over large temperature and activity ranges that
minor deviations from stoichiometry may often be neglected in a discussion of defect
concentrations and defect-controlled properties.
Numerous oxides will exhibit marked deviations from stoichiometry. Some
oxides are even unstable when stoichiometric. Wustite, for instance, which is nominally
written FeO, has under all conditions a deficiency of iron which varies with
temperature and oxygen activity, and its formula should correspondingly be written
Fe
1-y
O. In many systems of the transition metal oxides there exist phases which show
large variations from the apparent stoichiometric composition in that the average
valences of the metal ions in these phases are intermediate between the integral
valences of the metal ions. These arise due to the elimination of defects in the form of
shear planes. One often finds homologous series of oxides of this kind, with small
differences in average valence and composition. This will be further discussed below.
In the following is given a qualitative description of different defect structure
situations. The corresponding defect equilibria that express the concentration of the
different point defects as a function of temperature and oxygen activity are treated in a
later chapter.

Defects in stoichiometric compounds.

Let us consider a stoichiometric crystal with composition MX. If a charged point
defect is formed in such a crystal, a complimentary point defect with opposite effective
charge must be formed to conserve the electroneutrality of the stoichiometric crystal.
Two types of defect structures have been found to be important in stoichiometric metal
oxides and these are termed Schottky and Frenkel defects, respectively, honouring early
contributions of two of the many German scientists who pioneered the development of
defect chemistry (Schottky (1935), Frenkel (1926)).

126
Schottky disorder
A stoichiometric crystal with Schottky disorder contains equivalent
concentrations of cation and anion vacancies. A stoichiometric oxide MO contains
equal concentrations of metal and oxygen ion vacancies. In a stoichiometric oxide MO
2

the concentration of oxygen vacancies is twice as large as that of the metal ion
vacancies. It may be noted that the formation of Schottky defects can only occur at
outer and inner surfaces or dislocations and will diffuse into the crystal until
equilibrium is reached.

Figure 1-14 Schottky defect pair in NaCl. From Shriver and Atkins: Inorganic Chemistry.

Frenkel disorder
A stoichiometric crystal with Frenkel disorder contains the same concentrations
of metal vacancies and metal interstitial ions. Contrary to the Schottky defects, Frenkel
defect pairs can be formed directly inside the crystal.

Figure 1-15. Cation Frenkel defect pair in AgCl. From Shriver and Atkins: Inorganic Chemistry.

Although Schottky and Frenkel disorder may be simultaneously present in
stoichiometric compounds, one type of disorder usually predominates. As a rough rule
127
Schottky disorder is favoured in crystals where the cations and anions are of
comparable size, while Frenkel disorder predominates when the sizes of the cations and
anions are appreciably different. Another factor is that Schottky disorder tends to
dominate when the structure is very effectively packed so that the interstitials that are
part of Frenkel pairs are hard to form.
Pairs of interstitials have not been found as a dominant defect structure. Anti-site
disorder (swapping of sites) is common in intermetallics and between cations on
different sites (or anions on different sites). However, swapping between cations and
anions in ionic compounds is mostly prohibited by the very unfavourable effective
charges that would arise.

Figure 1-16. Anti-site defect pair in CuAu intermetallic. From Shriver and Atkins: Inorganic
Chemistry.

Defects in nonstoichiometric oxides.

The deviation from the exact stoichiometric composition is directly related to the
presence of point defects. While complimentary point defects are formed in
stoichiometric crystals, the electrical neutrality of nonstoichiometric compounds is
conserved through the formation of point defects and charge compensating electronic
defects.
Nonstoichiometric oxides may - depending on the oxide, temperature and
activities of the components - have an excess or deficit of metal or oxygen. In view of
this, nonstoichiometric oxides may be divided in four limiting groups:
- metal deficient oxides, e.g. M
1-y
O, where y is usually a small fraction of 1. Metal
vacancies or complex defects based on metal vacancies are the majority defects, and
examples with oxides with metal deficit are Co
1-y
O, Ni
1-y
O, and Fe
1-y
O;
- metal excess oxides, e.g. M
1+y
O, in which metal interstitials are the prevalent
defects. Cd
1+y
O is an example of an oxide with this type of nonstoichiometry;
128
- oxygen deficient oxides, e.g. MO
2-y
. In these oxides oxygen vacancies prevail and
examples of such oxides are CeO
2-y
and other oxides with the fluorite structure;

- oxygen excess oxides, e.g. MO
2+y
such as, for instance UO
2+y
. Oxygen
interstitials or complex defects based on these are the predominating defects.
Although a particular type of defect predominates in an oxide and as such
constitutes the majority defect in the crystal, it is important to emphasise that the
crystal will contain all other defects in varying concentrations in the form of minority
defects. Furthermore, the majority defects in the crystal may also change with
temperature or activity of the components in the crystal. By way of example, in an
oxide MO
2
the predominant defects may in principle change from oxygen vacancies to
interstitial metal ions on changing the oxygen activity, and under these conditions the
nonstoichiometry may be written M
1+x
O
2-x
.

Associations of point defects, point defect clusters,
extended defects

Point defects are often assumed to be present as single, unassociated point defects
that are randomly distributed in the structure. However this is in principle only
applicable for small defect concentrations, that is, typically when the fraction of defects
(relative to the number of normal atoms in the crystal) is smaller than 0.001-0.01 (but
depending a lot on temperature).
For larger defect concentrations one expects that defects begin to interact. The
interactions can be coulombic or mechanical and be attractive or repulsive. These
interactions may change the activity coefficients and formation enthalpies of defects,
and they may lead to associations between defects which attract each other. A simple
example is the association of a native point defect and a dopant or impurity ion with
opposite effective charges. Energy minimisation may for large defect concentrations
lead to clusters of different defects or elimination of defects in lines or planes, so as to
form a new structure (e.g. shear structures).
Order or disorder of defects is a matter of a fine balance between the
configurational entropy of the defects and the exothermic enthalpy gained in
association of certain point defects to form various types of larger defect
agglomerations. Systems with relatively small defect concentrations are entropy
controlled and consist of randomly distributed point defects. As concentration increases
the enthalpy takes control and leads successively to formation of larger defect entities;
randomised associated defects or defect clusters, point defects assimilated into
randomised structure elements of a new phase, superlattice ordering and extended
defects, shear planes, and discrete intermediate phases.

129
Example cases of defect clusters

Defect clusters in a metal-deficient oxide: Fe
1-y
O
A well known example of an oxide where the point defects form defect clusters is
wustite, Fe
1-y
O. As mentioned above, under normal equilibrium conditions the oxide is
never stoichiometric; the value of y ranges from 0.05 to 0.15 depending on the
temperature and oxygen activity. Following the composition of the oxide it was
originally believed the point defects were iron vacancies. However, extensive structural
investigations by x-ray, electron and neutron diffraction have over the years shown that
the defects not only comprise Fe-vacancies but also Fe-interstitials. The earliest studies
of complexes indicated that for each vacancy created as a result of nonstoichiometry, a
Frenkel defect pair was also formed. The combination of these defects gave a complex
defect consisting of an interstitial ion (on a tetrahedral site) adjoined by two vacancies
(vacant octahedral sites). Further studies have shown that the complexes comprise
combinations of larger number of vacancies and interstitials with ratios of vacancies to
interstitials ranging from 3 to 4. The basic unit of such clusters consists of four
vacancies and one interstitial as illustrated below. Such units may further combine to
form more complex clusters through side-, edge-, and corner-sharing and for these the
ratios of vacancies to interstitials are close to 3. It is noteworthy that the larger corner-
sharing clusters have the same local atomic arrangement as magnetite, Fe
3
O
4
, and
these can be considered as nuclei of Fe
3
O
4
to which Fe
1-y
O transforms when the defect
concentrations exceed critical values. The defect cluster in question might be denoted
(4v
Fe
Fe
i
)
6/
.

Defect clusters in an oxide with excess oxygen: UO
2+x

Uranium dioxide has the fluorite structure and may have a deficit or excess of
oxygen depending on the temperature and oxygen activity. Under conditions when it
has excess oxygen, UO
2+x
, the maximum value that x may have is 0.25. This
composition may suggest that the predominant point defects are oxygen interstitials,
however, as shown by Willis through neutron diffraction studies the defects are not
single unassociated oxygen interstitials, but comprise clusters of oxygen interstitials
and vacancies. The real composition of the clusters and detailed location of the
individual point defects in the clusters are still a matter of discussion, but the formation
of clusters may qualitatively be visualised as follows: When an oxygen atom is
dissolved interstitially an oxygen atom on a neighbouring normal site is simultaneously
displaced to neighbouring interstitial site leaving behind a vacant normal oxygen site.
This qualitative model suggests a defect cluster consisting of two oxygen interstitials
and an oxygen vacancy. Later studies have shown that clusters consist of larger number
of oxygen interstitials and vacancies.

Extended defects, shear structures.

Extended defects are planar defects and may be considered to be formed by
ordering and elimination of point defects, e.g of oxygen vacancies, along specific
130
crystallographic planes. This process may be illustrated below. The predominant
defects are in this case assumed to be oxygen vacancies and are considered to be
aligned as illustrated. By shear of these crystals the aligned vacancies are eliminated,
and this produces a fault in the cation sublattice. The formation of shear phases gives
rise to the homologous series of oxide structures, e.g W
n
O
3n-1
and Ti
n
O
2n-1
. In this
way the crystals may be able to have variable composition and may as such be termed
nonstoichiometric.
The schematic depictions of defective lattices shown initially in this chapter
contain some other extended defects, like dislocations, aggregates and precipitates.
Contrary to point defects, extended defects can often be imaged directly in electron
microscopy, as seen in the figure below.

Figure 1-17. Real and simulated image of shear plane in WO
3-x
. From Shriver and Atkins:
Inorganic Chemistry.

Modulated structures/infinitely adaptive structures

Some oxide systems may accommodate nonstoichiometry by continuous
adjustment of a parent structure without forming point defects. Such structures are now
called modulated nonstoichiometric compounds, but these have also in the literature
been termed vernier structures or infinitely adaptive structures. An example of such an
oxide is the low-temperature modification of tantalum pentoxide, L-Ta
2
O
5
. The oxide
exchanges oxygen with the ambient atmosphere, but the monophasic structure is
maintained through the creation or elimination of oxygen sites in the oxygen sublattice
while the number of tantalum sites remains unchanged.
131

Summary

Due mainly to the gain in enthalpy of electrostatic attractions and electron orbital
overlap, solids usually organise their constituent ions, atoms, or molecules in a crystal
lattice. In such lattices the species take specific positions relative to each other, and this
pattern is repeated in 1, 2 or 3 dimensions. Different compounds take on different
geometrical arrangements depending on the properties of the constituents. This is a
complex function of the charge distribution of each atom or ion, in turn a function of
nuclear charge, ionisation energies, electron affinity and shape of the valence electron
orbitals. However, it is often useful to view the structures as close-packing or other
packings of rigid spheres, with systematic filling by smaller spheres of voids so created.
Because of the preferred occupation of specific sites in the lattice, together with the
tendency of distinct preferences in valence for the various elements, most compounds
are basically stoichiometric, that is, the ratio of the different atoms is a simple ratio of
small integers. In competition with this minimisation of enthalpy with perfect,
stoichiometric lattices, we have the tendency of increasing the entropy at T > 0 K and
this is done by the creation of defects. At any temperature there is thus an equilibrium
concentration of defects. The defects comprise point defects (vacancies, interstitials and
substitution), electronic defects (conduction electrons, holes, valence defects), and
associates, clusters, lines or planes of defects. The overall defect structure is a result of
energy minimisation by the enthalpy and entropy of formation and mixing, the
electroneutrality condition (charge balance), mass balance, and preservation of the
structure (site ratio balance). A defect is denoted by its constituent (a chemical element
symbol or a vacancy (v)), a subscript indicating lattice site or interstitial site (i) and a
superscript indicating effective positive or negative charge (
or
'
).

Literature

Frenkel, J. (1926) Z. Phys., 35, 652.
Krger, F.A. (1964) The Chemistry of Imperfect Crystals, North-Holland,
Amsterdam, and Wiley, New York.
Krger, F.A. and Vink, H.J. (1956) in Solid State Physics (eds F. Seitz and D.
Turnbell) 3, 307, Academic Press, New York.
Nray-Szab, I. (1969) Inorganic Crystal Chemistry Akadmiai Kiad,
Budapest.
Rao, C.N.R. and Raveau, B. (1995) Transition Metal Oxides VCH, New York.
Schottky, W. (1935) Z. Phys. Chem., B29, 335.
Shriver, D.F., and Atkins, P.W. (1999) Inorganic Chemistry, 3
rd
ed. Oxford Univ.
Press., Oxford.
132
Smart, L. and Moore, E. (1992) Solid State Chemistry; an Introduction, Chapman
& Hall, London.
Wagner, C. and Schottky, W. (1930) Z. Physik. Chem., B11, 163.
West, A.R. (1984) Solid State Chemistry and Its Applications, Wiley, New York.

Exercises

1. Translate the defect names used in Fig. 1-1 and 1-2 into your first (or second)
language.
2. Calculate the number of atoms in the bcc and sc unit cells of Figs. 1-7 and 1-8.
(If you are not familiar with the procedures: Count all atoms, but divide by 2, 4,
or 8 for atoms that are shared with neighbouring cells by being in faces, edges
or corners.)
3. Calculate the number of atoms in the hcp and fcc unit cells of Figs. 1-4 (a) and
(b).
4. Calculate the number of ions of each kind in the NaCl and ABX
3
cells in Figs.
1-9 and 1-12.
5. Calculate the number of atoms and voids in the cells in Fig. 1-6.
6. Calculate the volume filling ratio of balls in the structures of Figs. 1-4 (b), 1-7,
and 1-8. What would it be in 1-4 (a)?
7. Why would you say are metals generally heavier (have higher densities) than
ionic compounds (salts and ceramics)?
8. Sketch the periodic table of the elements, but include where possible the typical
binary oxides of the elements in their normal (most common) oxidation states,
in terms of formulae (stoichiometry) and structure type. Indicate also, based on
what you know about their tendency to take on neighbouring oxidation states,
whether the oxide is expected to be stoichiometric, have oxygen-deficiency, or
oxygen-excess.
9. List the main types of 0-, 1-, 2-, and 3-dimensional defects in crystalline solids.
10. Write the Krger-Vink notation for the following fully charged species in MgO:
Cation and anion on their normal sites, oxygen vacancy, magnesium vacancy,
interstitial magnesium ion.
11. Write the Krger-Vink notation for the following species in ZrO
2
: Cation and
anion on their normal sites, oxygen vacancy, zirconium vacancy, yttrium dopant
substituting Zr, interstitial carbon atom, cluster of yttrium dopant and oxygen
vacancy, nitrogen ion (N
3-
) substituting for oxygen ion.
12. Write the Krger-Vink notation for the following fully charged species in
CaTiO
3
: Calcium vacancies, titanium vacancies, oxygen vacancies, Ti ions on
Ca sites and vice versa, Ti interstitials.
13. Write the Krger-Vink notation and for the possible species and defects in a
material of your choice preferably a compound you are working with.
133
14. For the materials and list of defects in Exercises 10-13, write the
electroneutrality condition.
15. Write the electroneutrality condition for defects in boron-doped silicon. Write
the electronuetrality condition for defects in phosphorous-doped silicon. Write
the electroneutrality condition for pure (undoped) silicon.
16. Suggest an electroneutrality condition for Y-doped ZrO
2
.
17. Write an electroneutrality condition for MO
1-x
(hint: includes an oxygen defect
type and an electronic defect type).
18. Write an electroneutrality condition for MO
1+x
.
19. Write an electroneutrality condition for M
1-x
O.
20. Write an electroneutrality condition for M
1+x
O.

Answers to selected exercises, Ch. 1

1. -
2. 2, 1
3. 6, 4
4. 4:4, 1:1:3
5. 4:4, 4:8
6. 1-4b:
2 3
=0.74, 1-7:
8
3
=0.68, 1-8: ______, 1-4a:
2 3
=0.74.
7. -
8. -
9. -
10. -
11. -
12. -
13. -
14. -
15. -
16. -
17. -
18. -
19. -
20. -
2. Defect reactions
21
2. 2. 2. 2. Defect reactions Defect reactions Defect reactions Defect reactions

Introduction

As discussed in the previous chapter several different types of point defects
may be formed in metal oxides and other inorganic compounds. In principle, all
types of defects will be present, but in general, only a small number of different
defects will predominate.
When defect-dependent properties are to be interpreted, it is important that
the defect concentrations as a function of temperature and the activities of the
crystal components are known. For this purpose it is necessary to formulate
reactions and write equations for the formation (or annihilation) of the defects.
The defect formation may either occur internally in the solid or through
reactions with the environment. In the following, the rules for formulating defect
reactions will be described and applied to different defect structure systems, while
in the next chapter, conditions for equilibrium and equations relating equilibrium
defect concentrations to temperature, activities (partial pressures) of the
components in a compound, impurity concentrations, etc., will be discussed.

The three rules for writing defect reactions

From a thermodynamic point of view a solid containing point defects
constitutes a solid solution where the point defects are dissolved in the solid. In
analogy with liquid solutions, the solid may be considered to be the solvent and
the point defects the solute. Similarly, the defect equilibria may be treated in terms
of the thermodynamics of chemical reactions and solutions.
In analogy with reactions in aqueous solutions, the rules for writing defect
reactions include requirements of electroneutrality and mass balance. But for
defect reactions in a crystalline compound it is also necessary to require that the
ratio of regular lattice sites in the metal and oxygen sublattices must be
maintained, even if the total number of sites may increae or decrease during the
reaction. These rules may be summarised as follows:

Mass balance

The defect reaction must balance with respect to the mass, i.e. the number
and types of atoms involved in the defect reaction must be the same before and
after the defect formation or annihilation. Vacancies, which only represent empty
2. Defect reactions
22
sites, have zero mass and do not count. Also electronic defects are considered not
to count in the mass balance.
1

Electroneutrality

The compounds are and should remain electrically neutral. With the perfect
crystal as reference, this means that the total effective charge is the same before
and after the formation of the defects.
This means that the net charge on the left and right hand sides of a reaction
equation must be the same. This charge may be counted in terms of effective or
real charges. However, effective and real charges may in general not be balanced
against each other, and one must therefore avoid using both effective and real
charges in the same equation.

Ratios of regular lattice sites.

The ratio(s) of the number of regular cation and anion lattice sites in a
crystalline compound is constant. For instance, in an oxide MO the ratio of regular
M and O sites is 1:1 regardless of whether the actual composition is stoichiometric
or nonstoichiometric. Correspondingly, in a compound M
2
O
3
the ratio of regular
cation to anion sites is 2:3, and if 3 oxygen sites are created in M
2
O
3
through a
defect reaction, two regular M-sites - vacant or filled - in the metal sublattice must
also simultaneously be created.
This rule is special to defect chemistry in crystalline solids, and requires
special attention and some practice to avoid its pitfalls. As is discussed below, this
rule does not apply to oxides with infinitely adaptive structures.
Despite the requirement of the constancy in the ratio of regular sites, the
total number of regular lattice sites may change in a defect reaction, and therefore,
the defect equation may include the formation or annihilation of lattice sites as
long as the proper ratios are maintained.
It should be noted that no sites are created in the formation of electronic
defects.
2

In the many defect equations and equilibria which will be discussed, surface
atoms will not be considered separately. All atoms will be considered to be bulk
atoms, and this means that the treatment only applies to sufficiently large

1
One may choose to count and balance also the mass of electrons, but then one must be
sure to include also non-defect electrons (electrons in the valence band).
2
One may choose to operate with sites or energy levels - for electronic defects, and one
must then take into account the emty levels of the conduction band and the occupied levels of the
valence band.
2. Defect reactions
23
crystallite sizes for which the number of surface atoms is insignificant to that of
the bulk atoms.
Defect reactions generally do not result in significant changes in the number
of surface atoms. In the formation of a vacancy, for instance, a bulk atom is
transferred to the surface, but in the same process a previous surface atom, in turn,
becomes a bulk atom. (It is worth mentioning, however, that the creation of the
new bulk atom has consequences for the energy involved, as we shall discuss in a
later chapter.)

Examples of reaction equations for defects

In view of the many types of point defects that may be formed in inorganic
compounds and that each type of defect may have varying effective charge,
numerous defect reactions may in principle be formulated. In the following, a few
simple cases will be treated as examples. First, we will consider defect structure
situations in stoichiometric compounds (Schottky, Frenkel and intrinsic electronic
disorders) and then defect structure situations in nonstoichiometric oxides will be
illustrated. Finally, examples of defect reactions involving foreign elements will
be considered.
It should be noted that both the stoichiometric and nonstoichiometric defect
structures may prevail in the same compound, depending on the activities (partial
pressures) of the components in the compound.
The treatment and examples of defect reactions will not only provide
training in applying the rules of defect reactions, but also broaden the description
of defect structures and of the individual point defects.

Stoichiometric compounds instrinsic disorders

As described in the previous chapter, the defect structures in stoichiometric
compounds contain equivalent concentrations of negatively and positively charged
point defects. These are formed as a result of internal equilibria in the crystal and
do not involve reactions with the surroundings. For this reason the defect
structures in stoichiometric compounds are also termed internal disorder.

Schottky disorder
As described in the previous chapter, the Schottky disorder involves the
presence of equivalent amounts of cation and anion vacancies. In an oxide MO
this means that the crystal contains equal concentrations of metal and oxygen
vacancies. The overall formation of such a defect pair within the crystal involves
the transfer of a pair of cations and anions on regular lattice sites from the bulk to
the surface. In reality the defects are formed at external and internal surfaces or
2. Defect reactions
24
dislocations and subsequently diffuse into the crystal until they are randomly
distributed. In writing the equation, one is only interested in the initial and final
states, and one disregards the kinetics of the defect reaction.
If one starts with a pair of cations and anions on regular lattice sites within
the crystal, M
M
and O
O
, one must also take into account that the formation of the
Schottky pair results in the formation of two new lattice sites, and the overall
equation may thus be written

O M O M O M
O M v v O M + + + = +
//
(2.1)

However, in this equation M
M
and O
O
on both sides may be cancelled, and the net
equation therefore becomes

+ =
O M
v v
//
0 (2.2)

where 0 (nil) designates a perfect crystal.
In Eqs. 2.1 and 2.2 it is assumed that doubly charged vacancies predominate
and that neutral and singly charged vacancies can be neglected. The equations
could have been written in more detailed steps starting with the formation of
neutral vacancies and the subsequent, consecutive ionisation of these to singly and
doubly charged vacancies.

Frenkel disorder
For the Frenkel disorder the predominant defects are either limited to the
cations and anions, and the disorder involves the presence of equal numbers of
vacancies and interstitial ions in a sublattice in a crystal. In the formation of a
Frenkel defect pair, a cation on a normal site is transferred to an interstitial site,
and no new lattice sites are created in the process. If, for the sake of illustration,
the interstitial ion and the resulting vacancy are assumed to be doubly charged, the
formation of a Frenkel defect pair may be written

+ =
i M M
M v M
//
(2.3)

A corresponding equation may be written for the formation of an anion Frenkel
defect pair. This latter defect situation is also often termed an anti-Frenkel defect
structure.

Intrinsic ionisation of electrons
The excitation of an electron from the valence band to the conduction band,
thereby leaving an electron hole in the valence band, is written
2. Defect reactions
25

+ = h e
/
0 (2.4)

As mentioned above we can choose to apply the mass balance to electrons in
which case we might rewrite the equation as follows:

+ = h e e
x /
(2.5)

We can also choose to keep some kind of track of the sites or energy levels for
electrons;

+ = +
v c
x
c
x
v
h e h e
/
(2.6)

but both the two latter reactions represent levels of detail that we normally need
not consider.
As mentioned earlier, the electronic defects may be localized (valence
defects). In these cases the reactions are connected with individual atomic sites.
For instance the intrinsic ionization (disproprtionation) of Fe
3+
ions into Fe
2+
and
Fe
4+
ions would be written

+ =
Fe Fe
x
Fe
Fe Fe Fe
/
2 (2.7)

Similarly, intrinsic ionization may also take place by charge transfer. In ilmenite,
FeTiO
3
, we may for instance observe the reaction Fe
2+
+ Ti
4+
= Fe
3+
+ Ti
3+
, which
in terms of defects is written:

/
Ti Fe
x
Ti
x
Fe
Ti Fe Ti Fe + = +

(2.8)

Nonstoichiometry

The ratio of cation to anion lattice sites is the same whether a compound is
stoichiometric or nonstoichiometric. But as nonstoichiometry means that there is
an excess or deficit of either cations or anions, nonstoichiometry also means that
there is an excess of a certain type or types of defects relative to that in the
stoichiometric condition. If the predominating type of defects are charged,
complementary electronic defects are created in order to conserve electrical
neutrality. The extent of nonstoichiometry and the defect concentrations in
2. Defect reactions
26
inorganic compounds are functions of temperature and activities (partial
pressures) of their components.
First we consider, as an example the formation of oxygen deficiency in an
oxide MO
2
. The overall reaction may be written

) ( ) ( ) (
2 2 2 2
g O s MO s MO
x
x
+ =
(2.9)

From this equation it is qualitatively seen using le Chatelier's principle that the
oxygen deficit increases with decreasing oxygen pressure. Conversely, for oxides
with excess oxygen the nonstoichiometry increases with increasing oxygen
pressure.
In the following various defect reactions which are encountered in oxides
with oxygen or metal excess or deficit will be considered. As the activity of the
metal component is usually negligibly small compared to that of the oxygen
activity under most experimental conditions, the nonstoichiometry in metal oxides
is correspondingly a result of the interaction and exchange of oxygen between the
metal oxide and the surrounding gas atmosphere. In the following examples of
formation of nonstoichiometric defects, only cases where the metal oxides interact
with gaseous oxygen are illustrated. However, it should be borne in mind that the
corresponding nonstoichiometry and defects may be formed by interaction with
metal if it is experimentally feasible to control the activity of the metal component
in the surroundings of the crystal.

Oxygen-deficient oxides
An oxygen vacancy is formed by the transfer of an oxygen atom on a
normal site to the gaseous state. No change in the number of lattice sites takes
place. This defect reaction may be written

) (
2 2
1
g O v O
x
O
x
O
+ = (2.10)

In this equation it is assumed that the oxygen vacancy is neutral, i.e. the two
electrons of the O
2-
ion are associated with the vacancy or its immediate
neighbourhood and the vacancy has zero effective charge. As described in Chapter
1, the two electrons trapped at or near the vacancy may, depending on the
temperature and vacancy concentration, be excited and transferred away from the
vacancy. Correspondingly, the oxygen vacancy acts as a donor and becomes
singly and doubly charged:

/
e v v
O
x
O
+ =

(2.11)

/
e v v
O O
+ =

(2.12)
2. Defect reactions
27

The formation of the doubly ionized oxygen vacancy can be written as the total
reaction:

) ( 2
2 2
1
/
g O e v O
O
x
O
+ + =

(2.13)

In these equations the free electrons are considered delocalised in the
conduction band. If they are localised at a metal ion on a normal lattice site the
last equation could be reformulated as follows:

) ( 2 2
2 2
1
/
g O M v M O
M O
x
M
x
O
+ + = +

(2.14)

If this takes place in the oxide MO
2
the valence of the M atoms then partly reduce
from +4 to +3 when the oxide becomes non-stoichiometric (MO
1-x
).

Oxides with excess metal
The oxygen deficiency in an oxide may alternatively be equivalent to the
presence of excess metal relative to stoichiometric composition, and in this case
the predominant defects constitute interstitial atoms. Correspondingly the
composition of the oxide MO
2
should then be written M
1+x
O
2
. The formation of
an interstitial M atom in this oxide involves the transfer of a regular M
M
atom to
an interstitial site. As a metal lattice site is annihilated in this process, two oxygen
lattice sites must simultaneously be annihilated in the oxygen sublattice and this is
achieved by transferring two oxygen atoms to the gas phase. The defect equation
then becomes

) ( 2
2
g O M O M
x
i
x
O
x
M
+ = + (2.15)

The neutral interstitial M
i
atoms may be successively ionised to singly, doubly,
triply or quadruply charged interstitial ions, e.g.

/
e M M
i
x
i
+ =

, etc. (2.16)

From this we can state that the metal interstitial, like the oxygen vacancy, is an
electron donor. It may be noted that in both cases it is the effectively neutral or
incompletely ionized defects that is a donor, not the fully ionized defects.
We have until now written the formation of metal defects by exchange with
oxygen in the surrounding atmosphere. We may in principle choose to write the
same in terms of exchange with the metal component in the surroundings. For
2. Defect reactions
28
instance, metal insterstitials in MO can be formed in equilibrium with the metal
vapour:

/
2 ) ( e M g M
i
+ =

(2.17)

Metal-deficient oxides
In metal-deficient compounds metal vacancies are the predominating point
defects. In an oxide MO a metal vacancy is formed by reacting oxygen gas with
the oxide. This creates new oxygen lattice sites and therefore an equivalent
number of new metal lattice sites - which are vacant - is also formed:

x
O
x
M
O v g O + = ) (
2 2
1
(2.18)

In this reaction it is assumed that the vacancy that is formed is effectively neutral,
that is, two electrons are taken from the surroundings of the vacancy in order to
form the O
2-
ion. The lacking electrons may be taken up from the valence band,
forming electron holes here;

+ = h v v
M
x
M
2
//
(2.19)

The formation of electron holes may alternatively be expressed in terms of
valence defects:

+ = +
M M
x
M
x
M
M v M v 2 2
//
(2.20)

This correspondingly means that the valence for some M atoms changes from +2
for to +3.
The metal vacancy, whether it is effectively neutral (electrons missing from
its neighbourhood) or ionized (electrons supplied from the rest of the crystal) it
illustrates that the effective size of a point defect may extend beyond the site
itself; the charge may be distributed over at least the closest neighbours of the
defect. Accordingly, defect chemistry works fine under the ionic model and with
the assumption of integer charges, even if the compound is far from ideally ionic
the point defect is just a little bigger than one site. It works because in defect
chemical as in other chemical reactions, the electrons have to choose to go or stay
no half electrons are involved. It then does not matter how close that electron
was of the point defect as long as it was associated with it. For most defect
chemical considerations, it then also suffices to consider the defect as a true point
defect, and that the electrons are associated with the point defect itself.

2. Defect reactions
29
Oxides with excess oxygen
In metal oxides with excess oxygen the predominating point defects are
interstitial oxygen atoms or ions. The formation of a neutral interstitial oxygen
atom through reaction of oxygen with the oxide is written

x
i
O g O = ) (
2 2
1
(2.21)

No new lattice sites are formed in this reaction. The neutral interstitial oxygen
atoms may in principle be ionised to yield electron holes and interstitial oxygen
ions with negative effective charges, so that the total reaction becomes, for
instance,

+ = h O g O
i
2 ) (
//
2 2
1
(2.22)

Dissolution of foreign elements

The presence of impurities or dopants may significantly affect or even
control the concentrations of the native defects in a compound. The effects are to a
large extent dependent upon the relative valences of the ions in the parent
compound and those of the impurities or dopants and what sites the impurities and
dopants occupy. When the principal valence is the same as that of the host site, we
say that the foreign element is homovalent. When it is unequal to that of the host
site, we say that the foreign element is heterovalent or aliovalent.
Foreign aliovalent elements may dissolve in the host structure with zero
effective charge, However, their difference from the host atom makes them easily
donate or accept one or more electrons they are donors or acceptors. We will
briefly analyse the reason for this and describe the processes in terms of defect
reactions.
When boron, B, or phosphorous, P, dissolve as dopants in elemental silicon,
Si, they have, respectively one less or one too many valence electrons, compared
to Si. However, they are in this state effectively neutral since the electron number
in each case is compensated by the nuclear charge. Nevertheless, the missing or
extra electrons form local states in the band structure that easily accept or donate
electrons:

+ = h B B
Si
x
Si
/
(2.23)

/
e P P
Si
x
Si
+ =

(2.24)

2. Defect reactions
210
The neutral defects, the acceptor and donor, were here formed by dissolving the
foreign elements in the same oxidation state as the host, namely 0. It is the desire
of the crystal to fulfil the octet rule around each atom the covalent nature of the
bond that causes the ionization in this case.
We may do something similar with ionic compounds such as oxides. For
instance, hypothetical LiO, containing divalent lithium, Li
2+
, may be dissolved in
NiO to form effectively neutral
+ 2
Ni
Li species;
x
Ni
Li . Since Li is unstable in
oxidation state +2, it easily accepts an electron from the valence band of the
crystal:

+ = h Li Li
Ni
x
Ni
/
(2.25)
Thus, in this case, it is the formal oxidation state according to an ionic model that
is the argument for why the acceptor is ionised. Similar arguments hold for donors
in ionic compounds.
When dealing with aliovalent foreign elements it is common to assume
directly the ionisation of the donor or acceptor into an effectively charged state.
We often refer to the cases as donor- or acceptor-doped systems, but keep in mind
that it is the unionised species that constitutes the donor or the acceptor.

Effects of dissolution of aliovalent oxides in the metal-deficient oxide M
1-y
O.
Let us assume that the parent oxide is metal-deficient M
1-y
O and that the
majority point defects in M
1-y
O are charged metal vacancies, e.g.
//
M
v ,
compensated by electron holes,
h . The oxide is thus a p-type electronic

conductor. Assume now that a higher valent oxide Mh
2
O
3
is added. (Mh is used
as an arbitrary chemical symbol for a higher valent metal.) Some Mh
2
O
3
dissolves
in the M
1-y
O. The Mh-ions have in this case a valence of +3, and if it is assumed
that the Mh-ions occupy normal M-sites in M
1-y
O, the dissolved Mh-ions will
have one positive effective charge,
M
Mh . These additional positive effective
charges must for electroneutrality reasons be balanced by creation of an
equivalent concentration of negative effective charges or annihilation of an
equivalent concentration of positive effective charges. Let us consider these two
alternatives.
The dissolution of Mh
2
O
3
may in the first case be compensated by the
formation of additional M-vacancies with two negative effective charges. In this
case the equation for the dissolution of Mh
2
O
3
is written

x
O M M
O v Mh O Mh 3 2
//
3 2
+ + =

(2.26)

and thus the dissolution of Mh
2
O
3
in M
1-y
O increases the concentration of metal
vacancies in the parent oxide.
2. Defect reactions
211
But as will be discussed in detail in the next chapter an increased
concentration of charged metal vacancies simultaneously results in a decrease in
the concentration of electron holes. The dissolution of Mh
2
O
3
may thus
alternatively be written with a defect reaction involving annihilation of electronic
holes:

) ( 2 2 2
2 2
1
3 2
g O O Mh h O Mh
x
O M
+ + = +

(2.27)

When the dissolution of Mh
2
O
3
is written in this way, the equation emphasises
that the concentration of electron holes is reduced. However, as the concentration
of metal vacancies and electron holes are interrelated, it should be emphasised that
both processes take place. Actually, the dissolution of the higher valent cation
decreases the concentration of all effectively positive defects and increases the
concentration of all negatively charged defects. Qualitatively, this may simply be
seen as a consequence of the electroneutrality equation or, if one wishes, of Le
Chatelier's principle. Details of these aspects will be further discussed using defect
diagrams in Chapter 4.
If one has the choice of writing a doping reaction by creation or
annihilation of defects, it is usually more meaningful to choose the creation of
defects, since this can describe the doping reaction to levels beyond the defect
concentrations that existed in the undoped oxide, and since this describes the
defects that will dominate when the doping level gets high.
So far we have added a higher valent dopant oxide to M
1-y
O. Let us
alternatively consider what takes place when a lower valent dopant oxide, Ml
2
O,
is added to M
1-y
O. If the Ml
+
ions dissolve substitutionally in M
1-y
O, the
dissolved Ml
+
ions have one negative effective charge,
/
M
Ml . This will be
compensated by the formation of positive effective charge or the annihilation of
negative effective charge. In our example oxide this will mainly affect electron
holes and metal vacancies, and the two defect reactions may thus be written

x
O M
O h Ml g O O Ml 2 2 2 ) (
/
2 2
1
2
+ + = +

(2.28)

x
O M M
O Ml v O Ml + = +
/ //
2
2 (2.29)

Thus, the addition of Ml
2
O to M
1-y
O has the opposite effect of addition of
Mh
2
O
3
; it increases the concentration of electron holes and decreases the
concentration of metal vacancies. One may note that the first of the two reactions
is an oxidation and thus involves uptake of oxygen, while the latter is not. It is
typical of such dissolutions of foreign aliovalent elements in oxides that they can
be charge compensated by means of electronic defects in which case we are
dealing with reduction or oxidation and uptake or release of oxygen or by means
of point defects in which case we have no change in oxidation states and no
exchange of oxygen gas.
2. Defect reactions
212

Effects of dissolution of aliovalent oxides on the oxygen-deficient oxide
MO
2-x

The examples in the previous section show the effects of additions of higher
and lower valent oxide to the p-conducting, metal-deficient M
1-y
O. Let us also
briefly consider the effects of doping an oxygen-deficient oxide MO
2-x
with
higher and lower valent oxides, respectively. The predominating defects in MO
2-x

are oxygen vacancies compensated by defect electrons. The oxide it thus an n-type
electronic conductor.
When the dopant oxide is higher valent, e.g. Mh
2
O
5
, and the dopant cation,
Mh
5+
, dissolves substitutionally, the dopant ions get one effective positive charge,
M
Mh . This charge must be compensated either by formation of negative effective
charges in the form of electrons or by annihilation of positive effective charges,
oxygen vacancies. These defect reactions may be written

) ( 4 2 2
2 2
1
/
5 2
g O O e Mh O Mh
x
O M
+ + + =

(2.30)

x
O M O
O Mh v O Mh 5 2
5 2
+ = +

(2.31)

Thus when MO
2-x
is doped with a higher valent oxides Mh
2
O
5
the
concentration of electrons is increased and the concentration of oxygen vacancies
is decreased.
When the same oxide, MO
2-x
, is doped with a lower valent oxide, Ml
2
O
3
,
the negative effective charge of the dissolved atoms
/
M
Ml is compensated by
annihilation of electrons or formation of oxygen vacancies:

x
O M
O Ml g O e O Ml 4 2 ) ( 2
/
2 2
1
/
3 2
+ = + + (2.32)

x
O O M
O v Ml O Ml 3 2
/
3 2
+ + =

(2.33)

The latter is exemplified by yttria-stabilised zirconia where the acceptor doping is
compensated by mobile oxygen vacancies that lay ground for this material as a
solid electrolyte:

x
O O Zr
O v Y O Y 3 2
/
3 2
+ + =

(2.34)

Numerous other examples may be formulated, depending on the defect
structure of the parent oxide, the valence of the foreign ion and the site it
2. Defect reactions
213
occupies. It may be noted that a foreign cation dissolving interstitially always will
have a positive effective charge and thus affect the defect structure in a similar
manner as a higher valent cation dissolved substitutionally. We shall return to this
and other examples in subsequent chapters and in connection with the review and
discussion of a few individual oxides systems.

Foreign anions in oxides
In oxides, homovalent foreign anions comprise S
2-
, while aliovalent foreign
anions comprise F
-
and N
3-
. They can enter as impurities during synthesis, or
dissolve from gaseous species under reducing atmospheres, e.g.,

) ( ) (
2 2
g O H S g S H O
x
O
x
O
+ = + (2.35)

) ( 3 2 ) ( 2 3
2
/
3
g O H v N g NH O
O O
x
O
+ + = +

(2.36)

Dissolution of hydrogen in metal oxides
When metal oxides are exposed to gas atmospheres containing water vapour
or other hydrogen containing gases, hydrogen will dissolve in the oxides. Under
oxidizing or mildly reducing conditions, the hydrogen atoms ionise to protons and
associate with oxygen atoms on normal lattice sites and thereby form hydroxide
ions on normal oxygen sites,
O
OH . We may thus for instance write the dissolution
reaction

/
2
2 2 2 e OH O H
O
x
O
+ = +

(2.37)

or

) ( 2 2 2 ) (
2 2
1
/
2
g O e OH O g O H
O
x
O
+ + = +

(2.38)

in which case the protons dissolved are charge compensated by the formation of
defect electrons. Of course, the electrons may interact with other defects in the
oxide so that the protons in effect are compensated by formation of other negative
defects or by the annihilation of positive defects. From the dissolution reaction
and through the interaction with native defects in the oxide it is clear that the
dissolution of hydrogen in metal oxides is dependent both on the partial pressure
of the hydrogen source (e.g. water vapour or hydrogen) and of oxygen. These
aspects will be described in more detail in a later chapter.
The dissolution of hydrogen from its oxide, H
2
O, is in principle similar to
dissolution of other foreign cations. However, the possibility of a controlled water
vapour pressure and the fast diffusion of protons makes it much easier to attain
2. Defect reactions
214
and vary (and more difficult to completely avoid) an equilibrium content of
protons in the oxide.
In terms of defect chemistry the dissolved proton, located on a normal
oxygen ion as hydroxide, may also be considered to constitute an interstitial
hydrogen ion, and as such it is also in the literature alternatively written
i
H . One
just has to bear in mind that the protons do not occupy regular interstitial positions
(voids).
Under reducing conditions, where hydrogen is stable in oxidation state 0 (as
H
2
in the gas phase) it might be possible to foresee neutral hydrogen atoms
dissolved in oxides, probably interstitially, as
x
i
H . Under even much more
reducing conditions could also hydride ions be expected to become stable, e.g. as
dissolved substitutionally for oxygen ions, as the defect
O
H .

Ternary and higher compounds

We have so far concentrated on binary compounds, mostly oxides, and only
touched upon elemental solids. Ternary and higher compounds fall, however,
under exactly the same rules of writing defect reactions.
A typical ternary compound is a ternary oxide such as perovskite, CaTiO
3
.
As an example of defect reactions for this case, we consider first the formation of
Schottky defects. When we form new lattice sites in this reaction, we need to form
vacancies on both Ca and Ti sites to maintain the ratio between them, in addition
to the appropriate number of oxygen vacancies:

+ + =
O Ti Ca
v v v 3 0
//// //
(2.39)

If we further consider the uptake of oxygen by formation of cation vacancies and
electron holes, we again have to balance the cation sites:

+ + + = h O v v g O
x
O Ti Ca
6 3 ) (
//// //
2 2
3
(2.40)

Similar principles should be applied also in cases where one and the same
element is distributed on different crystallographic sites. For instance, Y
2
O
3
has a
structure where all oxygen ions are not strictly equal. Similarly, distorted
perovskites may have unequal oxygen sites. In the pyrochlore structure, A
2
B
2
O
7
,
there are 6 oxygen sites of one type and 1 of slightly different coordination and
energy (and one which is structurally empty and thus to be regarded as an
interstitial site). In principle the fomation or annihilation of crystal units has to
2. Defect reactions
215
maintain the ratio between those different sites in all such cases. However, this is
so far hardly ever practiced in defect chemistry.
Contrary to binary oxides, ternary and higher oxides can have non-
stoichiometry not only in terms of the oxygen-to-metal ratio, but also internally
between the various cations. This is in practice often a result of synthesis. For
instance, it may be difficult to weigh in exactly equal numbers of moles of Ca and
Ti precursors when synthesizing CaTiO
3
, so that the synthesized material has a
permanent number of vacancies on one of the cation sites. Such non-stoichiometry
may also be a result of equilibria. For instance, if A-site deficiency is energetically
favourable over B-site deficiency in the compound ABO
3
, we may at very high
temperatures (e.g., during sintering) see a preferential evaporation of the A
component:

) (
//
g AO v v O A
O A
x
O
x
A
+ + = +

(2.41)

During oxidation we might similarly see a preferential incorporation of A-site
vacancies, resulting in a precipitation of an A-rich phase:

) ( 2 ) (
//
2 2
1
s AO h v g O A
A
x
A
+ + = +

(2.42)

It may be noted that these reaction equations do not violate the site ratio
conservation requirement of the ternary oxide.
When we earlier doped elementary or binary compounds the reaction was
fairly straightforward. When we dope a ternary or higher compound, however, the
reaction may be less obvious we have some choices. It is quite common,
however, to do the synthesis and write the equation in such a way that one takes
out a corresponding amount of the host element that is substituted. If we, for
instance, want to dope LaScO
3
with Ca substituting for La, we go for a
composition La
1-x
Ca
x
ScO
3
. In order to see how we write the doping reaction in
this case we first just look at the trivial normal synthesis:

x
O
x
Sc
x
La
O Sc La O Sc O La 3
2 2 2
1
3 2 2
1
+ + = + (2.43)

Accordingly, we then write the defect reaction for the doping in the way that we
let there be Sc
2
O
3
reserved for the CaO:

+ + + = +
O
x
O
x
Sc La
v O Sc Ca O Sc CaO
2
1
2
5
/
3 2 2
1
(2.44)

2. Defect reactions
216
Summary

Defect reactions can be written in a similar manner as ordinary chemical
reactions. They must confer with the requirements for conservation of mass,
charge, and ratios of lattice sites, but are allowed to increase or decrease the total
number of lattice units.
We have treated defect reactions mostly in cases for binary ionic
compounds, notably oxides. Reactions for elemental crystalline solids follow the
same basic rules and principles, but without the site ratio conservation
requirement. Also higher compounds, e.g. ternary oxides, follow the same rules,
only often with slightly more complex site conservation considerations and more
defects.
Normally, one should seek to use reactions that create rather than annihilate
defects, and one should avoid using reactions that form more defects than
necessary (as such reactions are then a sum of simpler reactions).
We have formulated defect reactions which describe intrinsic ionic and
electronic disorder, nonstoichiometry, variable ionisation of point defects, and
substitutional dissolution of aliovalent cations and ions and anions. Aliovalent
elements may be compensated by electronic defects or by point defects, of which
the former involve red-ox-reactions.
We have treated hydrogen defects specially, as they arise from a special
source and in its most stable form, the proton, take on a special size and type of
defect.

Exercises

The excercises in this chapter mainly provide training in 1) formulating a
reaction in terms of the appropriate reactants and products, and 2) balancing and
checking them with respect to a) mass conservation, b) charge conservation, and
c) site ratio conservation.

1. Write a reaction for the formation of Schottky defects in NaCl.
2. Write a reaction for the formation of Schottky defects in MO
2
.
3. Write a reaction for the formation of Schottky defects in Cu
2
O.
4. Write a reaction for the formation of anion Frenkel defects in MO.
5. Write a reaction for the formation of anion Frenkel defects in CaF
2
.
6. Write a reaction for a charge transfer between the cation and anion in
CeO
2
, i.e. for reduction of the cerium ion and oxidation of the oxygen ion.
Write the same process as an intrinsic ionisation assuming delocalised
electronic defects.
2. Defect reactions
217
7. Write a reaction for the formation of fully ionised oxygen vacancies and
electrons when oxygen is lost in the reaction M
2
O
3
= M
2
O
3-x
+ x/2 O
2
(g).
8. Write reactions for the direct formation of fully ionised metal interstitials
in MO, M
2
O
3
, and MO
2
.
9. Write reactions for the direct formation of fully ionised metal vacancies in
MO, M
2
O
3
, and MO
2
.
10. Write reactions for formation of sodium vacancies in NaCl by exchange
with a) chlorine gas and b) sodium gas.
11. We have in the text claimed that charges can be associated directly with
ideal point defects, even if the compound is not ideally ionic, and that this
normally does not affect the results of the calculations. Under what
conditions will it start to matter that the charge of the defect actually
extends outside the defect itself?
12. Write reactions for dissolution of CaO into the oxide ZrO
2-x

13. Write reactions for dissolution of MfO into M
2-x
O
3
when we assume that
the foreign metal Mf is small and therefore dissolves interstitially.
14. Write reactions for dissolution of CaO substitutionally into the anion-
Frenkel dominated Y
2
O
3
.
15. Write reactions for dissolution of ZrO
2
substitutionally into the anion-
Frenkel dominated Y
2
O
3
.
16. Write reactions for dissolution of CaF
2
into CaO, assuming that F
-

dissolves substitutionally and that CaO is dominated by Schottky defects
in the pure, undoped state.
17. Write a reaction for dissolution of protons in an oxide with water vapour
as source and with oxygen interstitials as compensating defects.
18. Write reactions for dissolution of protons in an oxide M
1-x
O with water
vapour as source.
19. Consider an oxide M
2
O
3
which is acceptor-doped by dissolving the oxide
MlO. The dopants are compensated by oxygen vacancies. Write a defect
reaction for the dissolution of MlO. Write a reaction for the dissolution of
protons from water vapour into this acceptor doped oxide, by annihilation
of the oxygen vacancies. Finally, write a reaction for the dissolution of
MlO in the presence of water vapour and without forming oxygen
vacancies. Can the latter reaction be constructed as a sum of the two
previous reactions?
20. Hydride ions have been suggested to dissolve in oxides accompanied by
protons by disproportionation of hydrogen gas. Write defect reactions for
this in the case that the hydride dissolves a) interstitially, and b)
substitutionally. (Assume in both cases that the oxide is perfectly
stoichiometric and has no defects as a starting point).
21. Write defect reactions for formation of A-site Frenkel, B-site Frenkel, and
anion Frenkel disorders in a perovskite ABO
3
(assuming both cations are
trivalent).
2. Defect reactions
218
22. Write a defect reaction for the formation of Schottky disorder in spinel
MgAl
2
O
4
.
23. Suggest a defect reaction for formation of Schottky-type disorder in
LaMnO
3
when it is assumed that cation vacancies are formed in the La
sublattice only.
24. Write a defect reaction for dissolution of Sc
2
O
3
in CaTiO
3
during
synthesis, assuming that Sc substitutes Ti forming acceptors compensated
by oxygen vacancies.

3. Defect equilibria
3.1
3. 3. 3. 3. Defect equilibria Defect equilibria Defect equilibria Defect equilibria

Introduction

After formulating the correct defect reactions in a compound, we turn our
interest towards the equilibrium relations for those reactions. These correlate the
equilibrium concentrations of the different defects with temperature, activities
(partial pressures) of the components in the compound, and other parameters
which affect the defect structure. The defect equilibria will here generally be
described by the law of mass action.
The change in the Gibbs (free) energy of a chemical reaction in a closed
system is given by
3

S T H G = (3.1)

where H is the change in enthalpy and S the change in entropy of the reaction
and T is the absolute temperature.
The reaction will proceed spontaneously as long as G < 0, until G has
reached its minimum. Here, G = 0, and this relation represents the condition for
chemical equilibrium.
The creation of single, unassociated point defects in an elemental,
crystalline solid increases the internal energy of the system and the enthalpy of the
defect formation is positive. But the configurational entropy of the system also
increases, and the equilibrium concentration of the defects will be reached when
the Gibbs energy of the system is at minimum. Thermodynamically, point defects
will thus always be present in a crystal above 0 K.
All types of defects will in principle be formed. However, the free energies
of formation of the different types and systems of defects can have widely
different values, and correspondingly it is often found that certain defects
predominate in a particular solid. The relative concentrations of the different types
of defects will be a function a temperature and other variables. Thus, defect
equilibria with large positive enthalpy of formation, for instance, which are not
favoured at low temperatures, may become important at high temperatures.
From a thermodynamic point of view a solid containing point defects
constitutes a solid solution where the point defects are dissolved in the solid. In
analogy with liquid solutions the solid may be considered to be the solvent and the
point defects the solute. Similarly the defect equilibria may be treated in terms of
the thermodynamics of chemical reactions and solutions.

3
A closed system is one where the mass (chemical entities) of the system is confined, but
where energy (heat) can be exchanged with the surroundings. The expression for the Gibbs energy
arises as a result of the definition of entropy and of the first law of thermodynamics which
preserves the total energy of the system and the surroundings.
3.2

Point defects and equilibrium in an elemental
solid - a simple statistical thermodynamic approach

Before considering the thermodynamics of defect formation and of defect
reactions in metal oxides and the corresponding use of the law of mass action, the
simpler case of formation of one type of defect in an elemental solid will be
treated by means of statistical thermodynamics.
Consider a perfect crystal of an elemental solid E with N atoms as starting
material. Let n
v
vacancies be formed according to the defect-chemical reaction

E
E
= v
E
+ E
E
(3.2)

where E
E
is an E-atom on an E-site and v
E
is a vacant E-site. The vacancies are
formed by moving an E-atom from an E-site in the bulk to the surface of the
crystal. The total number of sites then becomes N+n
v
. The change in Gibbs energy
can be expressed in terms of the enthalpy of formation of each vacancy, H, and
the entropy associated with the formation of the defects, S.
The enthalpy change on forming n
v
vacancies is thus n
v
H. It expresses
the energy cost associated with creating a new, empty site. Since we are dealing
with small concentrations of defects, the enthalpy should be one that is valid for
ideal dilute solutions of vacancies in the lattice.
The entropy change can be divided into two parts: i) a vibrational entropy
change, S
vib
, reflecting the entropy created by vibrations associated with each
new vacancy, and ii) a configurational entropy, S
conf
, arising from the
distribution of n vacancies among N+n
V
sites.
The vibrational entropy change is associated with the vibrations in the
neighbourhood of each vacancy, and the total vibrational entropy change is
proportional to the number of vacancies, n
v
S
vib
. This entropy change is, as was
the enthalpy change, related to a dilute solution of vacancies.
The enthalpy and vibrational entropy changes are, as we have discussed, to
a first approximation proportional to the number of vacancies formed, while the
configurational entropy is a more complex function. The change in Gibbs free
energy associated with the formation of n
v
vacancies may accordingly first be
written

conf vib v
S T S T H n G = ) ( (3.3)

According to statistical thermodynamics the configurational entropy
(entropy of mixing) can be expressed in terms of the thermodynamic probability,
W, by the relation

W k S
conf
ln = (3.4)

3.3
where k is Boltzmann's constant. The thermodynamic probability W represents in
this case the number of different distinguishable ways whereby n
v
vacancies may
be distributed on N+n
v
lattice sites and is given as

! !
)! (
v
v
n N
n N
W
+
= (3.5)

For large numbers of N and n
V
, which is typical for real crystals, Stirling's
approximation ( x x x x = ln ! ln for x>>1) may be applied, whereby Eq. 3.4 takes
the form

) ln ln (
v
v
v
v
conf
n
n N
n
N
n N
N k S
+
+
+
= (3.6)

This entropy scales with the size of the system (the number N) but is not
proportional with the number of defects n
v
.
Intuitively, and by insertion in the equation above, we have S
conf
= 0 at n
v

= 0. Thus, S
conf
= S
conf
and by insertion into the expression for the change in free
energy we get

) ln ln ( ) (
v
v
v
v
vib v
n
n N
n
N
n N
N kT S T H n G
+
+
+
= (3.7)

The changes in enthalpy, entropy and Gibbs energy are illustrated schematically in
Fig.3.1.
3.4

Figure 3-1. Example of the variation in enthalpy, vibrational and configurational entropy, and
Gibbs energy with the concentration of vacancies in an elemental solid.

At equilibrium G will be a minimum with respect to n
v
, i.e. G/n
v
= 0. By
taking the derivative, and letting it be zero, one obtains

0 ln S H
n
G

vib
V
=
+
+ =
v
v
n N
n
kT T
(3.8)

The term
n N
n

v
v
+
represents the fraction of the total number of sites which are
vacant and is through rearrangement of Eq.3.8 given by

)
kT
H
exp( )
k
S
exp(
n N
n

vib
v
v

=
+
(3.9)

Since the formation of the vacancies is written E
E
= v
E
+ E
E
, the equilibrium
constant may, following the law of mass action, be written

)
kT
H
exp( )
k
S
exp(
n N
n
X K
vib
v
v
v
=
+
= = =
E E
v v
a (3.10)

3.5
E
v
a is the activity and
n N
n
X
v
v
v
E
+
= represent the site fraction of vacancies in the
crystal of the element E.
S
vib
in the expressions above represents the vibrational entropy change
and H the enthalpy change per vacancy. If one wants to express the same
properties per mole of vacancies, k must be substituted by the gas constant R =
N
A
k where N
A
is Avogadro's number. Relations corresponding to Eq. 3.10 may
be derived for other types of point defects.
At any temperature T>0 all solids will following Eq. 3.10 contain point
defects. In the example above we have shown that the equilibrium constant,
containing defect activities as site fractions, is given by an enthalpy change and an
entropy term which constitutes the vibrational entropy change and not the
(configurational) entropy of mixing. This applies to chemical mass action law
expressions in general. Furthermore, it illustrates that the vibrational entropy
enters without other factors only if the equilibrium constant is expressed properly
with concentrations given as site fractions in the case of defects.
It is common to combine the enthalpy and entropy changes into a Gibbs
energy change, in this case the standard Gibbs energy change:

)
RT
G
exp( )
RT
H
exp( )
R
S
exp( K
0
vib
v
= (3.11)

We will see later that this standard Gibbs energy change is the one we have when
the reaction proceeds with all reactants and products in their standard reference
state (activity of unity). It is not the same as the general Gibbs energy change
calculated for the progression of the defect formation reaction earlier.

Thermodynamics of chemical reactions

As a basis for the subsequent considerations of thermodynamics of defect
equilibria the following section provides a brief review of some general aspects of
thermodynamics of chemical reactions.
In treating chemical reactions and systems consisting of two or more
constituents, it is convenient to introduce the concept of partial molar
thermodynamic quantities. As regards equilibria, the partial molar Gibbs energy is
particularly important. The partial molar Gibbs energy is commonly termed the
chemical potential and written
i
. The physical significance of a partial molar
property is the increase in that property resulting from the addition of 1 mole of
the constituent to such a large quantity of the system that its composition remains
essentially unchanged. If a system consists of n
1
+ n
2
+ ... + n
i
moles of
constituents 1, 2, and i, the partial molar free energy for the "i"th constituent is
given by

)
n
G
(
,... , n p, T,
i
i
2 1
n
= (3.12)
3.6

The Gibbs energy of a system is in terms of chemical potentials of the
constituents given by

....
i 2 2 1 1

i
n n n G + + = (3.13)

When a chemical reaction takes place, the numbers of various constituents change,
and dG changes correspondingly. For an open system (i.e. where we may add or
remove heat and volume) at constant temperature and pressure, dG is given by

....
i 2 2 1 1 , i p T
dn dn dn dG + + = (3.14)

At equilibrium we have

0
i ,
= =
i
i p T
dn dG (3.15)

It may also be shown that under equilibrium conditions

0 ....
i 2 2 1 1
= + + d n d n d n
i
(3.16)

This equation is one form of the Gibbs-Duhem equation.
In a chemical reaction , e.g.

dD cC bB aA + = + (3.17)

the change in Gibbs energy is given by the difference in the total free energy in
the final and initial state

) (
B A D C
b a d c G + + = (3.18)

The chemical potential of the constituent "i" in the mixture can be written

i i i
a RT ln
0
+ = (3.19)

where a
i
is the activity of constituent "i" in the mixture and
0
i
is the chemical
potential of constituent "i" at a chosen standard state of unit activity. The activity
of pure solids or liquids is usually taken as unity at atmospheric pressure (1 bar or
10
5
Pa), while for gases the activity is unity when the partial pressure of the gas
constituent is 1 bar.
In solutions the activity is not unity, but may be equated with the mole
fraction, X
i
, in ideal solutions:

i i
X a = (3.20)

3.7
In non-ideal solutions the activity is related to the mole fraction through the
activity coefficient,
i
,

i i i
X a = (3.21)

i
is thus equal to unity in ideal solutions. It may be noted that in a system
consisting of several phases, the condition of equilibrium implies that the
chemical potential of constituent "i" is the same in all phases.
If one introduces the expression for the chemical potential in Eq. 3.19 into
Eq. 3.18 it follows that G is given by

b
B
a
A
d
D
c
C
a a
a a
RT G G ln
0
+ = (3.22)

where ) (
0 0 0 0 0
B A D C
b a d c G + + = represents the change in Gibbs energy in
the standard state, i.e. at unit activities. G is a constant under specified standard
states. At equilibrium 0 = G such that

K RT G ln
0
= (3.23)

where K is a constant given by

m equilibriu
b
B
a
A
d
D
c
C
a a
a a
K
|
|
\
|
= (3.24)

K is termed the equilibrium constant and it relates the activities (or concentrations
under ideal conditions) of products and reactants when equilibrium has been
attained at a given temperature. G may as in Eq. 3.1 be expressed in terms of
the standard enthalpy change, H, and entropy change, S:

K RT S T H G ln
0 0 0
= = (3.25)

This may be rewritten in the form

) exp( ) exp( ) exp(
0 0 0
0
RT
H
R
S
RT
H
K K

= (3.26)

Thus, we have arrived at the same type of expression for the equilibrium constant
as we did in our intitial example from statistical thermodynamics. From that, we
know that the entropy term contained in K = exp(S/R) contains vibrational
terms only, and not the configurational entropy.
From the equation above we see that the temperature dependence of K is
given by

3.8
R
H
T d
K d
0
) / 1 (
ln
= (3.27)

and that a plot of lnK vs 1/T (vant Hoff plot) will give S/R as the intercept and
-H/R as the slope.
This general treatment is applicable to any ideal solutions whether these
are gaseous, liquid, or solid, and it applies also to defect chemistry as we have
examplified initially.

Thermodynamics and point defects

Equilibrium thermodynamics may as stated above be applied to defect
reactions. However, before so doing it is of interest to consider the
thermodynamic properties of the individual point defects.

Virtual chemical potentials of point defects

The regular atoms on their normal sites and the point defects occupy
particular sites in the crystal structure and these have been termed structural
elements by Krger, Stieltjes, and Vink (1959) (see also Krger (1964)). As
discussed in Chapter 2, the rules for writing defect reactions require that a definite
ratio of sites is maintained due to the restraint of the crystal structure of the
compounds. Thus if a normal site of one of the constituents in a binary compound
MO is created or annihilated, a normal site of the other constituent must
simultaneously be created or annihilated.
As discussed above the chemical potential of a constituent represents the
differential of the Gibbs energy with respect to the amount of the constituent when
the pressure, temperature and the amounts of all other constituents are kept
constant. When a structure element is created, the number of complementary
structure elements can not be kept constant due to the requirement of a definite
site ratio, and it is therefore not possible to assign a true chemical potential to a
structure element. However, Krger et al. have shown that one may get around
this difficulty by assigning a virtual chemical potential (zeta) to each separate
structure element and that the virtual chemical potential behaves like a true
chemical potential and may be related in a in a similar way to activity:

i i i
a RT ln
0
+ = (3.28)

The virtual potential differs from the true chemical potential by an undefined
constant which is incorporated in . For this reason the absolute value of the
virtual chemical potential can not be determined experimentally.
In real changes in crystals where the site ratio is maintained and
complementary structure elements (or building units) are created or annihilated,
the undefined constants in the virtual chemical potential of each separate structure
element cancel out, and the overall change is characterised by real changes in
3.9
Gibbs free energies. Real changes may thereby be described in terms of virtual
potentials. Krger et al. pointed out that a similar situation occurs in aqueous
electrolytes. In this case, arbitrary thermodynamic potentials are assigned to
separate ions, although it is impossible to determine experimentally the separate
potentials.

Ideal and non-ideal models

As described in the preceding section the defect reactions and equilibria can
be described as chemical reactions and treated in terms of the law of mass action.
It is emphasised that the equilibrium constants relate activities of the structure
elements involved in the defect reaction, but under ideal conditions and when the
structure elements can be assumed to be randomly distributed over the available
sites, the activities equal the concentrations of the structure elements. This is a
valid assumption for very dilute solutions. However, for larger defect
concentrations interactions take place between the defects, etc. and in principle
activities should be used instead of concentrations.
The interactions between charged defects may be accounted for by using the
Debye-Hckel theory in analogy with the interactions of ions in aqueous
solutions. This requires knowledge of the relative dielectric constant, the smallest
distance between charged defects, and other parameters for the solid. Debye-
Hckel corrections have, for instance, been worked out and tested for cation
vacancy defects in metal-deficient Co
1-y
O and Ni
1-y
O. At infinite dilution the
metal vacancies can be considered to have two negative charges,
//
Co
v , and the
equivalent number of electron holes is assumed to be randomly distributed in the
oxide. With an increasing concentration of point defects and electron holes, there
will be an increasing association of the metal vacancies (with two effective
negative charges) and the positively charged electron holes. In these terms, the
metal vacancies can be considered surrounded by a cloud of electron holes. This
results in a deviation from a random distribution of electron holes and a
corresponding deviation from the ideal solution model. For the case of Co
1-y
O the
Debye-Hckel correction can explain the increasing deviation from ideality with
increasing defect concentration. However, this model in its present form does not
satisfactorily explain various electrical conductivity results for Co
1-y
O and so far
apparently fails to give a full description of the defect structure and defect-
dependent properties of the oxide.
An alternative approach is to consider that the association of defects leads to
the formation of new "associated" defect species. These "new" defects are, in turn,
treated in terms of the ideal solution models. Specifically, in the case of metal
vacancies in, for instance, Co
1-y
O and Ni
1-y
O it is assumed that the metal
vacancies with two effective charges become associated with one electron hole to
give a metal vacancy with one negative effective charge,
/
Co
v , as has already been
described in Chapter 2. At this stage this simple approach appears to give an
equally consistent description of the defect structure and defect-dependent
properties of Co
1-y
O as the much more complicated Debye-Hckel model.
However, it is to be expected that future developments will provide a more
3.10
detailed understanding of defect interactions and how they influence defect-
dependent properties of inorganic compounds.
Despite its shortcomings the ideal solution approach will be used in this
book in treating defect equilibria. Defect interactions will generally be treated by
assuming the formation of associated defects as is, for instance, exemplified above
for the formation of singly charged metal vacancies through the association of an
electron hole with a doubly charged metal vacancy. Furthermore, the defect
equilibria will be treated using the law of mass action.
Concentrations of the structure elements may be expressed in many units. In
semiconductor physics it is common to express the concentrations in number per
cm
3
. This has the disadvantage, however, that when two compounds with different
molecular (packing) densities contain the same number of defects per cm
3
, the
number of defects per molecular unit is different in the two compounds, and vice
versa. Furthermore, the molecular unit size (packing density, or unit cell
parameters) of a compound changes as a function of its nonstoichiometry and
temperature, and in this case the unit of defects per cm
3
does not unequivocally
reflect the relative defect concentration per molecular unit or atom site under
different conditions.
In the following, concentration will in most cases be expressed as the
number of defects or atoms per molecule or site, i.e. in molar or site fractions,
while we for some purposes need to use the number of defects per cm
3
. As
mentioned before, the value of the equilibrium constant depends on the units of
concentration that are employed, but it is a simple matter to convert the values of
the equilibrium constant from one system to another.

Examples of defect equilibria in pure,
stoichiometric metal oxides

In this section we will see how we write defect reactions for selected simple
cases of combinations of defects, express the equilibrium constants, and combine
this with an electronutreality condition to express the concentration of dominating
defects as a function of temperature and activities. We will stay strictly with
statistical thermodynamics, such that entropy changes in the defect reaction can be
interpreted strictly as a change in the vibrational entropy. Approximations will be
implemented in order to visualise the simplicity f the method.

Schottky defects

As illustration let us consider the defect equilibrium for Schottky defects in
the oxide MO and where the metal and oxygen vacancies are doubly charged. The
defect equation for their formation is given in Chapter 2 as

//
0
M O
v v + =

(3.29)

3.11
The corresponding defect equilibrium may at low defect concentrations be written

[M]
] [
[O]
] [
X X
//
//
M O
v v
S
v v
K
M O

= =

(3.30)

where K
S
is the equilibrium constant for a Schottky defect pair, X represent site
fractions, and square brackets express concentrations of species. [O] and [M]
represent the concentration of oxygen and metal sites, respectively.
If we let square brackets express concentration in mole fraction (i.e. the
number of the species per mole of MO), then [O] = [M] = 1, and then

] ][ [
//
M O S
v v K

= (3.31)

When the concentrations of the defects are expressed in site fractions, or
properly converted to mole fractions as here, K
S
is related to the Gibbs molar
energy, G
S
, enthalpy change, H
S
, and entropy change, S
S
, of formation of
pairs of doubly charged Schottky defects through the relation above:

)
H
- exp( )
S
exp( )
G
exp(- ] ][ [
S S S //
RT R RT
v v K
M O S

=
= =

(3.32)

It may be useful to repeat that the entropy change , S
S
, does not include the
configurational entropy changes associated with the formation of the defects, but
mainly involves the change in vibrational entropy.
When the Schottky pair dominates the defect structure, we may see from
the reaction equation or from the electroneutrality condition that in MO the
concentrations of the metal and oxygen vacancies are equal:

] [ ] [
//
M O
v v =

(3.33)

Insertion into the preceding equation yields

)
2
H
- exp( )
2
S
exp( ] [ ] [
S S 2 / 1
S
//
RT R
K v v
M O

= = =

(3.34)

Under these conditions the defect concentrations are only dependent on the
temperature; they are independent of the activities of the components M and O
2
.
The slope in a vant Hoff plot of lnK vs 1/T would be -H
S
/2R. The factor 2
results from the square root of the equilibrium constant, in turn resulting from the
fact that there are two defects created, in comparison with the one we got in the
element used as example earlier.
It may be useful to note that the procedure we have applied is to write
down the defect reaction and the mass action law expression for its equilibrium
constant, and the prevailing electronuetrality condition. Thus, we have two
independent equations for the two unknowns (provided that either K
S
or all of
S
S
, H
S
, and T are known).

3.12
Frenkel defect pairs

In our next example we consider Frenkel defect pairs in MO. If, for the sake
of illustration, the Frenkel defects in MO are assumed to be doubly charged, the
defect equation for their formation is given as

//
M i
x
i
x
M
v M v M + = +

(3.35)

It may be noted that a vacant interstitial site has been included to keep track of the
sites where the interstitials are going. The corresponding defect equilibrium is
written as

] ][ [
] ][ [
] [
] [
] [
] [
] [
] [
[i]
] [
X
//
//
//
x
i
x
M
M i
x
i
x
M
M i
v M
v M
F
v M
v M
i
v
M
M
M
v M
X X
X
K
x
i
x
M
M i

= = =

(3.36)

If there is one interstitial site per MO, and if the defect concentrations are small
such that 1 ] [ ] [ = =
x
i
x
M
v M , we obtain the simpler

] ][ [ K
//
F M i
v M

= (3.37)

If Frenkel defects predominate in the pure stoichiometric compound, the
concentrations of interstitial cation and vacancies are then equal, and we obtain:

)
2
H
- exp( )
2
S
exp( ] [ ] [
F F 2 / 1
F
//
RT R
K v M
M i

= = =

(3.38)

where H
F
is the enthalpy of formation of Frenkel defect pairs. Under these
conditions the concentration of the Frenkel defect pairs is independent of the
activities of the metal and oxygen.

Intrinsic ionisation of electrons.

For many oxides, and particularly when considering defect structure
situations close to stoichiometry, it is essential to take into account the intrinsic
ionisation of electrons. This can include localised defects (valence defects) or
delocalised defects (valence band and conduction band).
In the case of valence defects, the ionisation of a pure binary metal oxide
MO
a
may typically be assigned to mixed valency of the metal and thus be written
as

+ =
M M
x
M
M M M
/
2 (3.39)

3.13
The equilibrium constant may then be expressed as site fractions, which in our
case immediately simplifies into an expression containing only volume or molar
concentrations:

] [
] ][ [
K
2
/
i
x
M
M M
M
M M

= (3.40)

If the intrinsic ionisation dominates the defect structure, we have

] [ ] [
/
M M
M M =
(3.41)

Insertion into the equilibrium constant yields

)
2RT
exp( )
2R
]exp( [ ]K [ ] [ ] [
1/2
i
/ i i x
M
x
M M M
H S
M M M M

= = =
(3.42)

In the case of delocalised electrons we write the ionisation reaction

+ = h e 0
/
(3.43)

In this model, classical statistical thermodynamics cannot be applied, and we
abandon the accounting of sites and configurational entropy. Instead we enter
simply the volume concentrations of defects in the widespread definition of the
intrinsic ionisation constant:

p n ] ][ [ K
/ /
i
= =

h e (3.44)

where
/
i
K is the equilibrium constant. The temperature dependency of the product
np is mainly given by the ionisation energy, most often referred to as the band
gap, E
g
. This has a slight temperature dependency that may be expressed the same
way as we usually do for Gibbs energies and equilibrium constants:

) exp( K ) exp( K
/
i 0,
/
i
RT
H
RT
E
i
g

= = (3.45)

If the concentrations of electrons and electron holes predominate, then the
electroneutrality condition may be approximated by

)
2
exp( ) K ( )
2
exp( ) K ( p n
2 / 1 /
i 0,
2 / 1 /
i
RT
H
RT
E
i
g

= = = = (3.46)

Intrinsic ionization of electrons and holes is, for instance, concluded to
apply to Fe
2
O
3
and Cr
2
O
3
at high temperatures.

3.14
Defect equilibria in nonstoichiometric oxides

Here, we will employ mainly the same methodology as in the preceding
section, but when delocalised defect electrons and holes are introduced we cannot
stay strictly with statistical thermodynamics and the equilibrium thermodynamics
cannot be related in the same simple manner to the change in vibrational entropy
any longer.

Oxygen deficient oxides

In oxides with oxygen deficit, the predominant defects are oxygen
vacancies. If these are fully ionised (doubly charged), and if the electronic defects
are localised as valency defects, their formation reaction and the corresponding
equilibrium can be written:

) ( 2 2
2 2
1
/
g O M v M O
M O
x
M
x
O
+ + = +

(3.47)

2 / 1
2
2 /
2 / 1
2
2
2 2
/
] ][ [
] ][ [
O
x
M
x
O
M O
O
M O
M v
vO
p
M O
M v
p
X X
X X
K
x
M
x
O
M O

= =

(3.48)

If these oxygen vacancies and the compensating electronic valence defects are the
predominating defects in the oxygen deficient oxide, the principle of
electroneutrality requires that

] 2[ ] [
/
=
O M
v M (3.49)

By insertion, and assuming the oxide is MO and that 1 ] [ ] [ = =
x
O
x
M
O M (small
defect concentrations) we then obtain:

6 / 1 3 / 1 6 / 1 3 / 1 /
2 2
)
3
H
- exp( )
3
S
exp( 2 ) 2 ( ] 2[v ] [M

= = =
O
vO vO
O vO O M
p
RT R
p K (3.50)

This shows that the concentrations of the dominating defects (defect electrons and
oxygen vacancies) increase with decreasing oxygen partial pressure. It
furthermore shows the relation between defect concentrations and entropy and
enthalpy changes of the defect reaction. The factor 3 that enters results from the
formation of 3 defects in the defect reaction.
Now, let us do the same treatment, but for an oxide where defect electrons
are delocalised in the conduction band. The formation of oxygen vacancies is then
written:

) ( 2
2 2
1
/
g O e v O
O
x
O
+ + =

(3.51)

3.15
Since delocalised electrons in a band cannot be treated by classical statistics, we
abandon the site fraction principle, and express all activities simply as
concentrations, and denote the new equilibrium constant with a prime
4
:

2 / 1 2 /
2
n
] [
] [
O
x
O
O
vO
p
O
v
K

= (3.52)

If these oxygen vacancies and the compensating electrons are the predominating
defects in the oxygen deficient oxide, the principle of electroneutrality requires
that

] 2[ n

=
O
v (3.53)

By insertion, and assuming that 1 ] [ =
x
O
O (small defect concentrations) we then
obtain:

6 / 1 vO 3 / 1 /
vO 0,
6 / 1 3 / 1 /
2 2
)
3
H
- exp( ) (2K ) 2 ( ] 2[v n

= = =
O O vO O
p
RT
p K (3.54)

and deliberately use a pre-exponential K
0
instead of an entropy change. Otherwise,
the solution is the same as in the case with localised valence defects for the
electrons.
A plot of log n or ] [ log

O
v vs
2
O
p log (at constant temperature) will give
straight lines with a slope of 1/6. Such plots are called Brouwer diagrams, and
they are commonly used to illustrate schematically the behaviour of defect
concentrations under simplified limiting cases of dominating defects.

Figure 3-2. Brouwer diagram for ] 2[ n

=
O
v as the electroneutrality condition.

4
This text introduces the use of a prime in the equilibrium constant to denote cases where
statistical thermodynamics has not been used throughout, and where the entropy of the reaction
thus is not well defined. This notation is new and not in common use, as is neither the distinction it
notifies. One may therefore expect also that this text relaxes its stringency in this respect in later
chapters, where the distinction is less in focus.
3.16
If we want to investigate the more complex situation of involving neutral
and partially ionised oxygen vacancies and the excitation of the trapped electrons
(described in Chapter 2) we start by writing down the stepwise reactions and their
corresponding equilibria:

) (
2 2
1
g O v O
x
O
x
O
+ = (3.55)

2 / 1
2
] [
] [
O
x
O
x
O
vOx
p
O
v
K = (3.56)

/
e v v
O
x
O
+ =

(3.57)

n
] [
] [
/
1
x
O
O
vO
v
v
K
= (3.58)

/
e v v
O O
+ =

(3.59)

n
] [
] [
/
2

=
O
O
vO
v
v
K (3.60)

where
/
2
/
1
and , ,
vO vO vOx
K K K are the respective equilibrium constants. Now the
principle of electroneutrality requires that

] 2[ ] [ n

+ =
O O
v v (3.61)

The concentrations of the electrons and the neutral, singly, and doubly charged
oxygen vacancies are related through the equations above, and by combination
expressions for each of the defects may be obtained. The electron concentration is
given by

2 / 1 /
2
/
1
3
2
) 2 (

+ =
O vO vO vOx
p n K K K n (3.62)

This equation has two limiting conditions.
If
/
2
2
vO
K n << , then

6 / 1 3 / 1 /
2
/
1
2
) 2 ( ] [ 2

= =
O vO vO vOx O
p K K K v n (3.63)

Under these conditions the concentrations of electrons and oxygen vacancies are
relatively small, and under these conditions the doubly charged vacancies are the
predominating oxygen vacancies, and the concentrations of electrons and oxygen
vacancies are proportional to
6 / 1
2
O
p . Of course, this situation corresponds to the
simple case presented initially for the doubly charged vacancies, and obviously
/ /
2
/
2 vO vO vO vOx
K K K K = .
3.17
If, on the other hand,
/
2
2
vO
K n >> , then

4 / 1 2 / 1 /
1
2
) ( ] [

= =
O vO vOx O
p K K v n (3.64)

Under these conditions the concentration of electrons is relatively large, and the
predominant oxygen vacancies are then singly charged. Furthermore, the
concentrations of electrons and oxygen vacancies are then proportional to
4 / 1
2
O
p .
A general tendency similar to that of oxygen deficient oxides applies to
metal deficient oxides; in the oxide M
1-y
O the metal vacancies are doubly charged
at very small deviations from stoichiometry and tend to become singly charged
with increasing nonstoichiometry.
It may be noted that the neutral oxygen vacancies are not affected by
charged defects, nor do they affect the electroneutrality.

Oxide with excess metal

We will as an example consider the formation of fully ionised interstitial
metal ions and complementary electrons in an oxide with excess metal M
2+y
O
3
.
The equation for the formation of interstitial metal ions with three effective
positive charges and three complementary electrons is given by

) ( 3
2 4
3
/
2
3
g O e M M O
i
x
M
x
O
+ + = +

(3.65)

The corresponding defect equilibrium is given by

-1 -3/2 4 / 3 3 /
] [ ] [ ]n [
2
x
M
x
O O i Mi
M O p M K

= (3.66)

where K
Mi
is the equilibrium constant. If these defects are the predominating
ones, and we as before assume small defect concentrations, the electroneutrality
condition and the oxygen pressure dependence of the interstitial metal ions and the
electrons is

16 / 3 4 / 1 /
2
) 3 ( n ] 3[

= =
O Mi i
p K M (3.67)

Thus, also in this case the concentration of the defects increases with decreasing
oxygen activity and is proportional to
-3/16
O
2
p . This oxygen pressure dependence is
different from that for formation of singly as well as doubly charged oxygen
vacancies. In such a case it would thus in principle be possible to decide from
measurements of electron concentration (e.g. via electrical conductivity) or of
nonstoichiometry as a function of oxygen activity whether the predominating
defects are triply charged interstitial metal ions or oxygen vacancies with one or
two charges.
However, had the metal interstitials predominantly been doubly charged in
the oxide M
2+y
O
3
, a
-1/4
O
2
p dependence would have been the result, and additional
3.18
studies of defect-dependent properties (e.g. self-diffusion of metal or oxygen)
would be needed to distinguish this situation from that of singly charged oxygen
vacancies.
It may be noted that it is not only the absolute number of charges on a
defect that determines the
2
O
p dependence, but also the difference between the
actual charge and the charge given by the nominal valence of the atoms involved.
Thus, in the oxide MO (or M
1+y
O) doubly charged metal interstitials will give a
-1/6
O
2
p dependence of the defect concentrations (as opposed to the case above) and
in this case be indistinguishable from doubly charged oxygen vacancies.

Simultaneous presence of oxygen vacancies and
interstitial metal ions

In an oxide M
a
O
b
where the ratio of metal to oxygen is larger than the
stoichiometric ratio a:b it is a priori difficult to predict whether interstitial metal
ions or oxygen vacancies predominate. In principle both types of defects may be
important, at least in certain regions of nonstoichiometry. In the following, let us
consider a case where it may be necessary to take into account the simultaneous
presence of interstitial metal ions and oxygen vacancies.
Consider an oxide with a stoichiometric composition MO
2
. Let us further
assume for the sake of illustration that when the oxide is nonstoichiometric the
important point defects are doubly charged oxygen vacancies and doubly charged
interstitial metal ions. (It mat be noted that the metal interstitials are not fully
ionised in this case.) The composition of the nonstoichiometric oxide may
accordingly be written M
1+y
O
2-x
.
The defect equations for the formation of these two types of defects may be
written

) ( 2
2 2
1
/
g O e v O
O
x
O
+ + =

(3.68)

) ( 2 2
2
/
g O e M M O
i
x
M
x
O
+ + = +

(3.69)

The corresponding defect equilibria (assuming [M
M
] and [O
O
] to be equal to
unity) then become

-1 2 / 1 2 /
] [ ]n [
2
x
O O O vO
O p v K

= (3.70)

-1 -2 2 /
2
] [ ] [ ]n [
2
x
M
x
O O i Mi
M O p M K

= (3.71)

The electroneutrality condition is given by

] 2[ ] [ 2

+ =
i O
M v n (3.72)

Two limiting conditions may be considered:
3.19
When ] [ ] [

>>
i O
M v then

6 / 1 3 / 1 /
2
) 2 ( ] [ 2

= =
O vO O
p K v n (3.73)

as obtained also earlier for the same conditions. By inserting this relationship for n
into the equilibrium constant for formation of the minority defects, metal
interstitials, we obtain

3 / 2 3 / 2 / /
2
2
) 2 ( ] [

=
O vO Mi i
p K K M (3.74)

Under these conditions the minority concentration of metal interstitials, ] [

i
M ,
increases rapidly with decreasing oxygen partial pressure, more rapidly than the
two dominating defects, and may eventually catch up with them and become
dominating at the expense of oxygen vacancies, as illustrated in the figure below.

Figure 3-3. Schematic presentation defect concentrations as a function of oxygen pressure
in an oxide MO
2
containing both doubly charged oxygen vacancies and doubly charged interstitial
metal ions, compensated by electrons.

When, as the other limiting case, ] [ ] [

>>
O i
v M then

3 / 1 3 / 1 /
2
) 2 ( ] [ 2

= =
O Mi i
p K M n (3.75)

By inserting this relationship for n into the equilibrium constant for formation of
the minority defect, oxygen vacancies, we obtain

6 / 1 3 / 2 /
2
/
2
) 2 ( ] [
O Mi vO O
p K K v

= (3.76)

3.20
It may be noted that under these conditions the concentration of oxygen
vacancies increases with the oxygen pressure, as shown in the figure.
The overall situation may be illustrated schematically as shown in Fig.3.3,
in which the concentrations of the three defects are plotted as a function of the
partial pressure of oxygen. In this diagram it is assumed that ] [

O
v is 10
5
times
larger than ] [

i
M at 1 atm O
2
, and with this assumption it is seen that the
interstitial cations will predominate at a partial pressure below about 10
-17
atm.
Many of the high melting oxides are stable over large oxygen pressure ranges and
it is not improbable that this hypothetical situation, with oxygen vacancies
predominating at high partial pressures of oxygen (and small deviations from
stoichiometry) and interstitial cations predominating at low partial pressures
(relatively large deviations from stoichiometry) may generally apply to oxygen
deficient oxides at elevated temperatures. In actual cases more than one ionisation
state of the defects may of course also have to be considered.
In this section we have employed and illustrated a very important sequence
of actions often used to solve ideal situations in defect chemistry, but also
implicitly used in traditional aqueous and other chemistry. We write the list down
in a numbered list for clarity:
1. Write down the full electroneutrality condition containing all
charged defects.
2. Write down a number of independent chemical reactions and
corresponding equilibrium constant expressions if you have n
defects you need normally n-1 such equilibria. (The nth
expression is the electroneutrality itself.)
3. Decide a pair of dominating charged defects, and simplify the
electroneutrality condition to this limiting situation.
4. Insert this limiting condition into an appropriately chosen defect
equilibrium to obtain the expression for the concentration of the
two dominating defects. (Normally, simplifications such as
assuming small defect concentrations are made here.)
5. Insert the obtained expression into another equilibrium expression
and solve to obtain an expression of the concentration of a
minority defect.
6. Extrapolate to see what minority defect comes up to become
dominating as conditions change. Then repeat from step 3 until all
defects are exhausted.
On the basis of such an exercise one may construct schematic diagrams of
how defect pairs dominate defect structures and how minority defects behave and
eventually take over. Transition zones are not solved explicitly in this way and are
thus often drawn sharp and schematically. Logarithmic depictions are common,
and such plots are then called Brouwer diagrams.

Metal-deficient oxide

3.21
The reactions and defect equilibria for the formation of single, unassociated
neutral metal vacancies and the subsequent excitation of electron holes in MO
may for small defect concentrations be written

x
O
x
M
O v g O + = ) (
2 2
1
(3.77)

2 / 1
2
] ][ [

=
O
x
O
x
M vMx
p O v K (3.78)

+ = h v v
M
x
M
/
(3.79)

p
] [
] [
/
/
1
x
M
M
vM
v
v
K = (3.80)

+ = h v v
M M
// /
(3.81)

p
] [
] [
/
//
/
2
M
M
vM
v
v
K = (3.82)

where
vMx
K ,
/
1 vM
K , and
/
2 vM
K are the equilibrium constants for the respective
defect equilibria.
When the metal vacancies and their complementary electron holes are the
predominating defects, the electroneutrality condition reads

] 2[ ] [ p
// /
M M
v v + = (3.83)

Through combinations of the equilibrium constant expressions and
electroneutrality above, the concentration of the separate point defects and
electronic defects can be evaluated. In an oxide M
1-y
O this leads to relationships
similar to that for oxygen vacancies with the exception that the concentrations of
electron holes and metal vacancies always increase with increasing oxygen
pressure. The concentration of electron holes is, for instance, determined by the
expreession

2 / 1 /
2
/
1
3
2
) 2 (
O vM vM vMx
p p K K K p + = (3.84)

As regards the electron holes two limiting conditions may be considered:
At low concentrations of electron holes (small deviations from
stoichiometry;
/
2 vM
K p << ) we have

6 / 1 3 / 1 /
2
/
1
2
) 2 (
O vM vM vMx
p K K K p = (3.85)

At high concentrations of electron holes (
/
2 vM
K p >> ) we obtain
3.22

4 / 1 2 / 1 /
1
2
) (
O vM vMx
p K K p = (3.86)

The oxygen pressure dependence of the concentration of electron holes and
the total concentration of the charged metal vacancies (deviation from
stoichiometry) correspondingly changes from
1/6
O
2
p to
1/4
O
2
p with increasing
deviation from stoichiometry.

Metal oxides with excess oxygen.

The defect equilibria involving the formation of interstitial oxygen atoms or
ions in oxides with excess oxygen may be set up following the same treatment as
applied to the other defects which have been dealt with above. The concentration
of interstitial oxygen species increases with
2
O
p , and the formation of charged
oxygen interstitials is accompanied by the formation of electron holes.

Examples of defect structures involving both
oxygen deficiency and excess

In the preceding examples of defect structures in non-stoichiometric oxides
it has been assumed that the oxides either have an oxygen deficit (or excess metal)
or excess oxygen (or metal deficit). In many oxides the predominating defect may
change from one type to another depending on the oxygen activity. As an
illustration of such a defect structure situation a hypothetical case will be
considered where an oxide predominantly contains oxygen vacancies at reduced
oxygen activities and interstitial oxygen ions at high oxygen activities. In an
intermediate region the oxide will be stoichiometric or close to stoichiometric.
Some oxides with the fluorite structure exhibit such a defect structure, e.g. UO
2x
.
For the sake of simplicity in illustrating such a defect structure situation it
will be assumed that both the interstitial oxygen ions and the oxygen vacancies are
doubly charged. In this case it will then be necessary to consider the following
equilibria, for the formation of oxygen vacancies, oxygen interstitials, electrons
and holes, and for anion Frenkel pairs:

2 / 1 2 /
2
]n [
O O vO
p v K

= (3.87)

2 / 1 2 // /
2
]p [

=
O i Oi
p O K (3.88)

np K
i
=
/
(3.89)

] ][ [
//
=
O i AF
v O K (3.90)

3.23
In these equilibria we have assumed small defect concentrations such that the
concentrations of normal lattice sites and empty interstitial sites have been
assumed constant and equal to unity and thus omitted from the expressions.
It should be noted that the defect equilibria are interrelated, and through a
combination of the equations it may be shown that
/ / 2 /
Oi vO AF i
K K K K = . Thus, only
three out of the four equilibria are sufficient to describe the defect structure of the
oxide.
The full electroneutrality condition is given by

n ] [ 2 p ] 2[
//
+ = +

i O
O v (3.91)

We will explore the defect structure of the oxide by considering limiting
conditions and follow the procedure listed before to construct Brouwer diagrams.

Oxygen deficit.

At large oxygen deficit the following approximation may be made

p ], [ 2 ] [ 2
//
i O
O v n >> =

(3.92)

We first insert this into the appropriate equilibrium to find the concentrations of
the dominating defects in the usual manner:

6 / 1 3 / 1 /
2
) 2 ( ] [ 2

= =
O vO O
p K v n (3.93)

and then insert this into other equilibria to find the concentrations of the minority
defects. By inserting the expression for the concentration of vacancies into the
anion Frenkel equilibrium, we obtain for the concentration of interstitials:

6 / 1 3 / 1 / 3 / 2
2
) ( 2 ] [
O vO AF i
p K K O

= (3.94)

By inserting the expression for the concentration of electrons into the intrinsic
electronic equilibrium, we obtain for the concentration of holes:

6 / 1 3 / 1 / /
2
) 2 (
O vO i
p K K p

= (3.95)

Oxygen excess

For relatively large excess oxygen, that is, when

n ], [ 2 ] [ 2
//
>> =
O i
v O p (3.96)

3.24
we may in a manner analogous to the preceding case, derive the following
relations:

6 / 1 3 / 1 / //
2
) 2 ( ] [ 2
O Oi i
p K O p = = (3.97)

6 / 1 3 / 1 / 3 / 2
2
) ( 2 ] [

=
O Oi AF O
p K K v (3.98)

6 / 1 3 / 1 / /
2
) 2 (

=
O Oi i
p K K n (3.99)

Stoichiometric condition

At or close to stoichiometry two alternative limiting conditions must be
considered, namely dominance by intrinsic electronic ionisation or by anion
Frenkel disorder.
If intrinsic ionisation of electrons predominates, and thus

] [ 2 ], [ 2 ) (
// 2 / 1 /
>> = =
O i i
v O K n p (3.100)

The concentrations of electrons and electron holes (n and p) are then independent
of oxygen pressure. The point defect concentrations are obtained by insertion of
the expression for the concentrations of the electronic defects into the appropriate
equilibria, and we obtain:

2 / 1
/
i
/
2
K
] [

=
O
vO
O
p
K
v (3.101)

2 / 1 /
i
/ //
2
K ] [
O Oi i
p K O = (3.102)

A Brouwer diagram illustrating this case of stoichiometric condition, along with
oxygen deficit and excess, is shown below.

3.25

Figure 3-4. Schematic presentation of oxygen point defects and electronic defects as a
function of oxygen pressure in an oxide which depending on the partial pressure of oxygen may
have an excess or deficit of oxygen. Intrinsic electronic equilibrium is assumed to predominate at
stoichiometric composition.

If, on the other hand, anion Frenkel disorder predominates under
stoichiometric conditions, and thus

2
) (
2
,
2
K ] [ ] [
2 / 1 /
1/2
AF
// i
O i
K n p
v O = >> = =

(3.103)

then ] [
//
i
O and ] [

O
v are independent of the partial pressure of oxygen, while the
concentrations of electronic defects are given by

4 / 1 4 / 1 2 / 1 /
2
) ( n

=
O AF vO
p K K (3.104)

4 / 1 4 / 1 2 / 1 /
2
) ( p
O AF Oi
p K K

= (3.105)

A Brouwer diagram illustrating this case of stoichiometric condition, along with
oxygen deficit and excess, is shown below.
3.26

Figure 3-5. Schematic presentation of the concentration of oxygen point defects and
electronic defects as a function of oxygen pressure. Oxygen defects are assumed to predominate at
stoichiometric composition

From the diagrams we can conclude that at low oxygen activities the oxide
has oxygen deficiency and will be an electronic n-type conductor, because the
mobility of the electrons is always much higher than that of oxygen vacancies. At
high oxygen partial pressures the oxide will correspondingly have oxygen excess
and be a p-type electronic conductor. At intermediate oxygen activities it may be
be a mixed n- and p-type electronic conductor in the case of intrinsic electronic
disorder, while it may exhibit ionic or mixed ionic/electronic conduction in the
case of anion Frenkel disorder.
Similar diagrams and analyses may be made for many other combinations of
non-stoichiometric and stoichiometric defect situations of pure (undoped) oxides
and other ionic compounds.

Summary

Chemical reactions for defects can be formulated and treated using the
mass-action law. As for other chemical reactions, equilibrium constants can be
defined in terms of the activities of the defects and other species. Under the
normal constrictions we can approximate activities with concentrations of defects
and partial pressures of gases. The equilibrium constants can also be expressed in
3.27
terms of the Gibbs free energy change of the reaction, which in turn is expressed
by the entropy and enthalpy changes.
We have seen that the standard entropy change of the standard Gibbs energy
change contains only vibrational terms provided that the configurational terms are
properly handled in expressing the equilibrium conditions. This is often possible
in defect idealised defect chemistry since we can apply classical statistical
thermodynamics. However, for non-classical defects such as delocalised electrons
and holes this approach is less meaningful, and the entropy change may then be
interpreted in a less straightforward manner. We have introduced the use of a
prime (as in K
/
) to denote equilibrium constants that are not expressed according
to statistical thermodynamics and where the entropy change in the Gibbs energy
change may have other than vibrationan contributions.
The mass action expressions can be combined with the full or limited cases
of the electroneutrality condition to obtain exact or approximate (limiting case)
expressions for the concentration of defects. Such concentrations are typically a
function of the oxygen partial pressure, temperature, and doping. It is common to
illustrate defect structures for oxides by plotting log defect concentrations vs log
2
O
p (Brouwer diagrams) or ln or log defect concentrations vs 1/T. We have
shown these principles and techniques through examples of intrinsic ionic and
electronic disorder, various types of nonstoichiometry and variable ionisation of
point defects.
We have restricted our treatment to simple cases and made simplifications
where possible. Assumption of small defect concentrations have allowed the
assumptions that the concentrations of normal lattice atoms and empty interstitials
sites are constant, with activities equal to unity. Larger defect concentrations can
to a first approximation be taken into account by including mass and site balances
into the expressions used to solve the defect structure.)

Literature

Krger, F.A. (1964) The Chemistry of Imperfect Crystals, North-Holland,
Amsterdam, and Wiley, New York.

Krger, F.A., Stieltjes, F.H. and Vink, H.J. (1959), Philips Res. Rept. 14,
557.
Exercises
1. Do the insertion of Stirlings approximation and the derivation that leads
up to the relation between equilibrium constant and entropy and enthalpy
changes for the elemental reaction E
E
= v
E
+ E
E
(Eqs. 3.3 and onwards).
2. Write the reaction for formation of Schottky defects in MO
2
and find an
expression for the defect concentrations as a function of the equilibrium
constant and thermodynamic parameters and temperature. Note in
3.28
particular how the solution deviates from the one obtained for the oxide
MO treated in the text.
3. In the case of cation Frenkel defects in MO we assumed in the text that we
had one interstitial site per MO. Derive the expression for the defect
concentrations if the structure consists of fcc close-packed O ions, with M
ions on each octahedral hole, and with interstitial sites on all tetrahedral
holes (Hint: consult Chapter 1).
4. For a defect situation in an oxide dominated by doubly charged oxygen
vacancies and electrons, sketch the vant Hoff plot (Logarithm of defect
concentration vs 1/T) and a double-logarithmic plot of defect
concentrations vs p
O2
(Brouwer diagram).
5. Consider a metal oxide MO
1-x
dominated by singly and doubly charged
oxygen vacancies. Suggest a condition that expresses the changeover from
dominance of one to dominance of the other. Next, find an expression for
the pO
2
(as a function of equilibrium constants) of this transition point.
6. Consider a metal oxide M
2-x
O
3
dominated by doubly and triply charged
metal vacancies. Suggest a condition that expresses the changeover from
dominance of the doubly charged to dominance of the triply charged.
Next, find an expression for the pO
2
(as a function of equilibrium
constants) of this transition point.
7. Find a general expression for m in the [e
/
] p
O2
1/m
relationship for oxides
MO
b
with non-stoichiometry dominated by fully ionised metal interstitials.
(Hint: Find m as a function of b). Use this to find m in the case of M
2
O,
MO, M
2
O
3
, MO
2
, M
2
O
5
and MO
3
.
8. Find a general expression for m (as a function of the valency z of the
cation) in the [h
.
] p
O2
1/m
relationship for binary oxides with non-
stoichiometry dominated by fully ionised metal vacancies. Use this to find
m in the case of M
2
O
3
.
9. Sketch a full Brouwer diagram (log defect concentrations vs log p
O2
) for
an oxide M
2
O
3
dominated by fully ionised oxygen and metal vacancies at
under- and overstoichiometry, respectively. Assume that Schottky defects
predominate close to stoichiometric conditions. (Hint: the main goal is to
obtain and illustrate the p
O2
-dependencies. Use the rules we listed for such
constructions of Brouwer diagrams.)
10. For the defect situation derived in the preceding exercise for M
2
O
3
do the
following analyses:
i. find a condition and the expression for the p
O2
when the oxide is
fully stoichiometric.
ii. Assume that you are able to determine this p
O2
at various
temperatures and plot ln p
O2, stoichiometric
vs 1/T. What information
may you obtain from this plot?
iii. Assume that you are able to determine n (the concentration of
electrons) at the p
O2
of exact stoichiometry at various temperatures
and plot ln n
stoichiometric
vs 1/T. What information may you obtain
from this plot?
3.29
iv. Assume that you are able to determine n (the concentration of
electrons) at a fixed p
O2
at various temperatures in the region of
near stoichiometry and plot ln n vs 1/T. What information may you
obtain from this plot?
O2
) for
an oxide MO
2
under- and overstoichiometry, respectively. Assume that intrinsic
electronic equilibrium predominates close to stoichiometric conditions.
(Hint: the main goal is to obtain and illustrate the p
O2
-dependencies. Use
the rules we listed for such constructions of Brouwer diagrams.)
12. For the defect situation derived in the preceding exercise for MO
2
do the
following analyses:
i. find a condition and the expression for the p
O2
when the oxide is
fully stoichiometric.
ii. Assume that you are able to determine this p
O2
at various
temperatures and plot ln p
O2, stoichiometric
vs 1/T. What information
may you obtain from this plot?
iii. Assume that you are able to determine n (the concentration of
electrons) at various temperatures in the region of near
stoichiometry and plot ln n vs 1/T. What information may you
obtain from this plot?
O2
) for
an oxide ABO
3
under- and overstoichiometry, respectively. Assume that Schottky defects
predominate close to stoichiometric conditions. You may assume that both
cations are trivalent, but discuss also the effect it would have if A was
divalent and B tetravalent. (Hint: the main goal is in any case to obtain and
illustrate the p
O2
O2
) for
an oxide AB
2
O
4
dominated by fully ionised metal interstitials and
vacancies at under- and overstoichiometry, respectively. Assume that
Frenkel defects predominate close to stoichiometric conditions. You may
assume that A is divalent and B trivalent. (Hint: the main goal is to obtain
and illustrate the p
O2
15. In the text we considered an oxide MO
2
that had oxygen deficiency by
both fully charged oxygen vacancies and doubly charged metal
interstitials. The figure illustrating the case was drawn assuming that the
ratio between the concentrations of the two defects was 10
5
at 1 atm. p
O2
.
Calculate the exact p
O2
at which the concentrations are equal.
16. In the text we considered an oxide MO
2
that had oxygen deficiency by
both fully charged oxygen vacancies and doubly charged metal
interstitials. The figure illustrating the case is not strictly a Brouwer
diagram, but rather a diagram showing the exact solution of the defect
3.30
structure, thus also correctly describing the transition region. The full
solution can be reached in several ways:
i. By analytical solutions. Find and plot n (the concentration of
electrons) and the concentrations of the other defects as a function
of p
O2
(double-logarithmically, over a reasonable range) using an
analytical solution. Also, find and plot instead p
O2
vs the
concentration of one of the defects (again double-logarithmically,
and over a reasonable and physical range). Were both approaches
viable? Was one simpler?
ii. By a sequential calculation of an organised set of expressions. This
may take any of the two analytical approaches mentioned above as
starting point, but avoids finding the full analytical solution and
instead lets a computer do all the calculations once the expressions
are programmed
5
.
iii. By a semi-analytical approach: In the full electroneutrality
condition, replace each concentration by expressions such that only
one defect is left. Then select p
O2
and find the one defect
concentration by minimising the error numerically
6
.
iv. Numerically: Here one needs to adjust the parameters to fit a
number of independent equations corresponding to the number of
unknowns. Which equations would you include in the present case?
Do you know a common method to fit them?
7

5
This approach can be implemented in Excel spreadsheets, as done for a number of
complex defect structures of ternary oxides by F.W. Poulsen and others.
6
This approach is implemented in software written by J. Abrantes.
7
A program that solves the mixture of different types of equations numerically has been
written by T. Norby. Compared to the other computerized approaches it executes very slowly.
4. Impurities and dopants
4.1
4. 4. 4. 4. Impurities and dopants Impurities and dopants Impurities and dopants Impurities and dopants

Introduction

In the preceding chapter we have only dealt with pure, crystalline metal
oxides, and no account was taken of impurities and dopants and their effects on
defect equilibria. Under real conditions it is impossible to produce truly pure
crystals and for many oxides impurities may, depending on their concentration
and the temperature, significantly affect or fully control defect concentrations in
crystals. In the same terms dopants are purposely added to obtain or tailor defect-
dependent properties of metal oxides and other materials.
The term dopant is from semiconductor science traditionally used for
amounts of foreign atoms in the parts per million (ppm) range. However, many
use it, as we shall here, also for foreign atoms in any concentration, up to several
or even tens of percent. When one type of atoms is deliberately replaced to any
degree by another type of atom, we may also use the term substituent instead of
dopant.
Impurities and dopants may have various effects on crystals. Interstitially
dissolved foreign atoms will normally affect the strain in the lattice. When ionised
their charge affect the electroneutrality condition. Substitutionally dissolved
impurities or dopants will also affect the properties of the host compound if they
exhibit a difference in size or charge compared with the host atoms they replace.
In the following we will mainly consider cases where there is a difference in
the valence between the foreign atom and the native atom it replaces. Such foreign
atoms are termed aliovalent. Their effective charge will affect the electroneutrality
condition and thereby the defect equilibria.
Larger dopant concentrations are generally expected to change also
enthalpies of defect formation. This is, for instance, well demonstrated for oxides
that have the same crystal structures and type of defects and that form solid
solutions, e.g. solid solutions of NiO-CoO, MgO-NiO, a.o. However, we shall
neglect such effects here and, as a first approximation, treat all cases as ideal
dilute solutions, as we have done for defect equilibria earlier.
By the solubility of foreign atoms we mean the amount that can be held by
the structure without precipitation of a second phase. This is thus the amount
dissolved in equilibrium with that second phase. The solubility of foreign atoms
obviously changes with temperature. Less recognised is the fact that the solubility
under some defect structure situations also may vary with component (oxygen or
metal) activities, as we shall see later.
On account of this it is important to distinguish between the following cases:

4.2
1) The concentration of impurity or dopant (solute) is frozen in. This is the
case at relatively low temperatures; too low to allow transport to and
from sinks and sources of the foreign atoms (a solute-rich phase). In this
case the concentration of solute is constant. We may list some subcases
of this situation:
a) If the frozen concentration of solute is below the solubility limit, and
no second phase is present, the system is then stable.
b) If the concentration of solute is below the solubility limit, but a second
solute-rich phase is present, the system is metastable.
c) If the frozen concentration of solute exceeds the solubility limit, the
system is metastable a common case.
2) The concentration of impurity or dopant (solute) is in equilibrium. This is
the case at relatively high temperatures; high enough to allow transport
to and from sinks and sources of the foreign atoms. This case can be
divided into two subcases:
a) The concentration of solute is below the solubility limit. In this case
there is no solute-rich phase present, and the concentration of solute is
therefore again constant, as in all the cases 1).
b) The concentration of solute is above the solubility limit. In this case
there is a solute-rich second phase present, and the concentration of
solute is variable by exchange of solute between the bulk and the
sink/source of second phase.

In science and technology employing solid solutions (alloys or doped
systems) it is often assumed that a constant concentration of solute is present.
Furthermore, this is often assumed to be constant because it is fully soluble (case
2a), while it may in reality be constant because it is frozen in (cases 1).
During synthesis, sintering, equilibration, and use at high temperatures the
concentration may change (case 2b) and this of course is also the basis for binary
and higher phase diagrams. Defect chemistry offers a tool for understanding these
changes, not only as a function of temperature, but especially as a function of
component activities.
Despite this possibility, also defect-chemical treatments most commonly
assume that the concentration of dopants and impurities is constant, often (tacitly)
because they are assumed to be fully soluble (case 2a)). In reality, this may have
been true at the high temperatures used during synthesis and fabrication, while
during characterisation or use at lower temperatures the concentrations remain
constant rather because they are frozen in (cases 1)). We shall start our treatment
with cases of invariable concentrations of aliovalent foreign substituents,
disregarding whether this is a result of full solubility or freezing-in.

4.3
Constant concentration of impurity/dopant

Schottky equilibria

As the predominance of Schottky defect situations have primarily been
demonstrated in detail for ionic halides, e.g. alkali halides (NaCl, KBr, LiI a.o.),
let us consider the effects of impurities/dopants on Schottky equilibria in a
compound MX where the cations and anions have a valence of 1 and the cation
and anion vacancies are singly charged, i.e.
X
v and
/
M
v . If divalent foreign
cations, Mf
2+
, are dissolved substitutionally in the lattice and occupy the normal
M-sites, the Mf
2+
ions will have one effective positive charge,
M
Mf . (It has
donated an electron, and Mf is called a donor dopant). If we disregard other native
defects than the Schottky defects, the electroneutrality condition becomes

] [ ] [ ] [
/
M X M
v v Mf = +

(4.1)

The defect structure accordingly has two domains given by the limiting cases of
this expression:
If the foreign cation is a minority species, i.e., ] [ ] [

<<
X M
v Mf then
] [ ] [
/
M X
v v =
, and from the Schottky equilibrium

] ][ [ K
/
S M X
v v
=

(4.2)

we obtain

1/2
S
/
K ] [ ] [ = =
M X
v v

(4.3)

If, on the other hand, ] [ ] [

>>
X M
v Mf then

constant ] [ ] [
/
=

M M
Mf v (4.4)
,i.e., the metal vacancy concentration is fixed by the concentration of the foreign
cation. The concentration of the minority species, ] [

X
v , can be found by insertion
into K
S
:

] [
K
] [
S
=
M
X
Mf
v

(4.5)
4.4

We observe that this concentration decreases with increasing concentration of
dopant, as expected also from direct inspection of the electroneutrality condition.
The two regions and their transition is illustrated in the figure below. We
often call the region dominated by intrinsic disorder for intrinsic, while the one
dominated by foreign species such as dopants or impurities is called extrinsic.

Figure 4-1. Effect of higher valent cation impurities on the defect
concentrations in MX predominantly containing Schottky defect pairs, with K
s
=
25.

Frenkel equilibria

The effects of higher valent cation impurities on Frenkel equilibria are
analogous to those described for the Schottky equilibria. When divalent
M
Mf
impurities/dopants are dissolved in MX with predominating cation Frenkel defect
pairs (
/
M
v and
i
M ) the electroneutrality condition is given by

] [ ] [ ] [
/
M i M
v M Mf = +

(4.6)

4.5
The addition of
M
Mf decreases ] [

i
M and increases ] [
/
M
v in a manner analogous
to that described for the Schottky-dominated material.

Oxygen-deficient oxides doped with lower valent cations

A situation which is frequently encountered in applications of several oxide
systems is that doubly charged oxygen vacancies are the predominating native
point defects and that the cation impurities/dopants have a valence lower than that
of the parent cation (impurities/dopants with negative effective charge are also
called acceptors). An example is zirconia (ZrO
2
) doped with lower valent cations
such as for instance yttrium (
/
Zr
Y ) and calcium (
//
Zr
Ca ). For the sake of illustration
let us assume that the impurity/dopant ions have one effective negative charge,
/
M
Mf . The actual valence of the host cation will not matter. The defect
equilibrium between vacancies and electrons is as before given by

) ( 2
2 2
1
/
g O e v O
O
x
O
+ + =

(4.7)

2 / 1 2 /
2
n
] [
] [
O
x
O
O
vO
p
O
v
K

= (4.8)

The electroneutrality condition will be given by

n Mf v
M O
+ =

] [ ] 2[
/

(4.9)

For such a system two limiting conditions can be considered:
If ] [ ] 2[
/
M O
Mf n v >>

, the foreign cations do not affect the native defect
equilibrium, and the electron and oxygen vacancy concentrations are given by
their own equilibrium, and they are proportional to
6 / 1
2
O
p as we have shown in the
preceding chapter. This will occur at relatively low oxygen activities, where these
concentrations are relatively large.
If n Mf v
M O
>>

] [ ] 2[
/
, the oxygen vacancy concentration is determined and
fixed by the dopant content (extrinsic region). The concentration of minority
electrons, n, is in this case given by

4 / 1 -1/2 / 2 / 1 /
2
] [ ) 2 (

=
O M vO
p Mf K n (4.10)

4.6
and thus decreases with a different dependency on
2
O
p than in the former case. As
the concentration of electrons and electron holes are related through the
equilibrium K
i
= np, the electron hole concentration in this extrinsic region
correspondingly increases with increasing oxygen activity. Depending on the
impurity content, oxygen activity and temperature, p may become larger than n.
The variations in ] [

O
v , n and p as a function of the partial pressure of
oxygen, are schematically illustrated for such a case in the figure below. In a
situation like this the oxide will typically be an n-type electronic conductor at low
oxygen partial pressure and an essentially ionic conductor at higher oxygen partial
pressure. Zirconia doped with a lower valent cation such as Y, Ca, a.o., confers
largely with the situation depicted in the figure.

Figure 4-2. Brouwer plot of the concentrations of defects as a function of oxygen partial
pressure in an oxygen deficient oxide predominantly containing doubly charged oxygen vacancies,
showing the effects of a constant concentration of lower valent cation dopants, ] [
/
M
Mf .

Oxygen deficient oxides doped with higher valent cations

Addition of higher valent cations which dissolve substitutionally in the
oxide will in the extrinsic region result in an increase in the concentration of
electrons and a decrease in the concentration of oxygen vacancies. This may be
shown by setting up the appropriate electroneutrality condition combined with the
defect equilibria in the same manner as shown above.

Oxides with excess metal

Nonstoichiometric oxides with excess metal in the form of predominating
interstitial metal ions exhibit the same qualitative effects of aliovalent foreign
4.7
cations. By additions of lower valent cations, the concentration of interstitial
cations will in the extrinsic region be determined by the concentration of the
impurities/dopants. Addition of higher valent cations will bring about an increase
in the electron concentration and a decrease in the concentration of metal
interstitials.

Metal-deficient oxides doped with lower valent cations

When a metal-deficient oxide MO with singly charged metal vacancies as
the prevalent native point defects is doped with singly valent foreign cations,
/
M
Mf , the electroneutrality condition is given by

p Mf v
M M
= + ] [ ] [
/ /
(4.11)

The effects of the foreign cations are evaluated by combining this with the
equations for the defect equilibria for the singly charged metal ion vacancies. As a
result we find that the lower valent cations will increase p and decrease ] [
/
M
v . The
variations in p and ] [
/
M
v as a function of the partial pressure of oxygen in such a
case are illustrated schematically in the Brouwer diagram below. At low oxygen
activities interstitial metal ions or oxygen vacancies may become important,
depending on the equilibrium constants for the Frenkel and Schottky defect
equilibria. In the figure it is assumed that singly charged metal interstitials become
the important native point defects at low oxygen activities.

Figure 4-3. Effect of a constant concentration of foreign lower valent cations on the
concentration of point and electronic defects as a function of oxygen pressure in a metal deficient
oxide predominantly containing singly charged cation vacancies.

4.8
Metal deficient oxides doped with higher valent cations

If the foreign cation impurities/dopants are higher valent, e.g.
M
Mf , the
electroneutrality condition of an oxide MO with predominantly singly charged
metal vacancies becomes

p Mf v
M M
+ =

] [ ] [
/
(4.12)

In this case ] [
/
M
v will increase and p will decrease with increasing concentration
of
M
Mf . The variations in p and ] [
/
M
v as a function of the oxygen partial pressure
is illustrated schematically in the following figure, and the variation in the
concentration of electrons (n) as minority defects is also included.

Figure 4-4. Brouwer diagram of the effects of a higher valent cation impurity/dopant on the
concentration of point and electronic defects as a function of oxygen pressure in a metal deficient
oxide predominantly containing singly charged metal vacancies.

Doping of oxides which may have regions with both
oxygen and metal deficit

In the previous examples of the effects of lower valent oxides we have
separately considered oxides which are either oxygen or metal deficient. We have
seen that the doping is compensated by increased concentrations of either
electronic defects or point defects, depending on the type of non-stoichiometry
originally present. In principle it is straightforward to combine these treatments
and consider the effects of aliovalent foreign atoms over the entire range of
stoichiometry, from the oxygen-deficient to the metal-deficient side. Depending
on oxygen partial pressure, the doping will be compensated by an electronic
4.9
defect or a point defect of the opposite effective charge of the dopant. As we shall
see it may be necessary to consider both possibilities when we deal with oxides
near stoichiometry.
Let us therefore start with an example oxide M
2
O
3
dominated by intrinsic
electronic disorder when stoichiometric, and oxygen vacancies and metal
vacancies giving oxygen and metal deficiency at, respectively, low and high
oxygen activities. We acceptor-dope this oxide with the lower-valent Ml
2+

cations, such that the electroneutrality condition reads:

] [ 2 ] 3[ ] [ n
/// /
+ = + +
O M M
v p v Ml (4.13)

We assume that the amount of MlO added is well below the solubility so
that ] [
/
M
Ml = constant. Fig. 4.5 shows a Brouwer diagram of the defect situation
as a function of oxygen partial pressure when the level of aliovalent dopant is
higher than the level of intrinsic disorder. At the lowest
2
O
p the oxide is oxygen-
deficient and oxygen vacancies and electrons predominate. As these defects
decrease with increasing
2
O
p we hit the level of the acceptor dopant. From here
on the acceptor will be compensated by a constant concentration of the positive
defect, i.e., the oxygen vacancies. This is analogous to the situation we obtained
when we earlier considered this case in particular.
In this situation the concentration of defect electrons decreases and that of
holes increases with increasing p . The two terms cross at
2 / 1
i
K p n = = , and
assuming that K
i
is independent of the doping, this is necessarily the same level as
that of the intrinsic electronic disorder in the pure, undoped material. At this point
also the doped material may be said to be stoichiometric; it contains oxygen
vacancies exactly matching the presence of the lower-valent cation (acceptor): If
we consider the doped material to consist of M
2
O
3
and MlO, both these
constituents are stoichiometric, and the additional presence of reduced and
oxidised states (n and p) is effectively zero at this point. The doped oxide can be
said to be stoichiometric with respect to the valence of its constituents.
By further increasing
2
O
p the concentration of holes will eventually become
dominating, taking over the dominance from the oxygen vacancies in terms of
compensating the charge of the acceptors. We hereby create higher oxidation
states (holes) and fill up the oxygen sublattice (removing oxygen vacancies).
During the increase in
2
O
p the concentration of metal vacancies increases and
eventually balances the concentration of oxygen vacancies. This level is given by
the Schottky equilibrium constant (K
S
) and represents another stoichiometric
point, this time with respect to the crystal structure of the host oxide (M
2
O
3
).
The situation encountered in this region is analogous to that depicted in Fig.
4.3, and with further increase in
2
O
p the charge due to the increasing
concentration of metal vacancies will become more important than that of the
acceptor level, and the oxide enters into a situation of dominance by metal
deficiency.
4.10
Fig. 4.5. is constructed by combination of limiting cases of the
electroneutrality condition (Eq. 4.13) and defect equilibria in a similar manner as
we have practised earlier.
It turns out by inspection of several cases that acceptor doping always leads
to a region at relatively low
2
O
p where we have compensation by a positively
charged point defect and a region at relatively high
2
O
p where we have
compensation by electron holes. This is true independent of type of intrinsic
disorder. However, the relative magnitudes of the equilibrium constants for
intrinsic disorder (e.g. K
i
vs K
s
) determine the relative width of dominance for the
two regions.

Figure 4-5. Brouwer diagram of the effects of lower valent dopant on the defect structure of
M
2
O
3..

In a manner equivalent to the above case we may construct a diagram
showing the situation in the same host oxide M
2
O
3
now donor-doped with a
higher valent cation Mh
4+
see Fig. 4.6. In this case and in all cases of donor-
doping, there will be a region of charge compensating the dominating donors with
defect electrons at relatively low
2
O
p and with a negatively charged point defect
at higher
2
O
p

4.11

Figure 4-6. Brouwer diagram of the effects of higher valent dopant on the defect structure
of M
2
O
3.

We have here concentrated on changes in charge compensation as a function
of oxygen activity, but the same may take place as a function of temperature and
concentration of dopant (at constant
2
O
p ).
Aliovalent doping affects defects with higher number of charges more than
those with fewer. Thus, with increasing doping, doubly and higher charged point
defects may become dominating at the expense of the singly charged electronic
defects, but not vice versa.

Variable concentration of impurity/dopant

In many cases an impurity or dopant may be present in amounts larger than
the solubility limit. The excess of the oxide of the impurity or dopant is then
presumably present in precipitates, often found at grain boundaries or surfaces of
the parent oxide. When one considers the effects of impurities/dopants on defect
equilibria in such cases, it is important to recognise that the solubility of the
impurity or dopant will not only vary with temperature but in many cases also
with the oxygen pressure. Let us illustrate these aspects by respectively
considering the effects of the lower valent oxide MlO and the higher valent oxide
MhO
2
on M
2
O
3
.

4.12
Doping M
2
O
3
with excess MlO

When Ml dissolves substitutionally in M
2
O
3
, the dissolved Ml
2+
has one
negative effective charge,
/
M
Ml . Let us further assume that these ions constitute
the important negative effective charges in the oxide. These must then be
compensated by positive defects, and let us assume that these are electron holes
and doubly charged oxygen vacancies, depending on the properties of the oxide,
the temperature, and the oxygen activity.
If the compensating defects are electron holes when MlO dissolves in
M
2
O
3
, the defect equation for the dissolution may be written

x
O M
O h Ml 3 2 2 (g) O 2MlO(s)
/
2 2
1
+ + = +

(4.14)

By taking the activity of the excess MlO(s) phase as well as that of oxygen ions as
unity, the equilibrium expression can be written

2 / 1
O
2 2 /
2 / 1
O
2
MlO(s)
3 2 2 /
p
2
2
p p ] [
p a
] [ p ] [
K

= =
M
x
O M
Ml
O Ml
(4.15)

The important difference from the previous treatments is that now we treat this
dissolution as an equilibrium. When the compensating defects are electron holes,
the simplified electroneutrality condition can be expressed by p ] [
/
=
M
Ml , and by
insertion into the equilibrium expression we obtain

8 / 1
O
1/4
p
/
2
p K p ] [ = =
M
Ml (4.16)

Thus, the solubility, i.e., ] [
/
M
Ml , increases with
2
O
p until all MlO is eventually
dissolved. Thereafter the behaviour will become as with constant acceptor
concentrations (all dissolved or frozen-in), described in Fig. 4.5.
When taking into account the defect equilibrium for the formation of oxygen
vacancies and the intrinsic electronic equilibrium (np = K
i
), the concentration of
oxygen vacancies is under this situation seen to be proportional to
4 / 1
O
2
p
. At lower
oxygen activities the doubly charged oxygen vacancies will thus eventually
become the charge compensating defects at the expense of electron holes. In this
case it may be more relevant to describe the dissolution of MlO by the following
defect reaction and equilibrium expression:

x
O O M
O v Ml 2 2 2MlO(s)
/
+ + =

(4.17)
4.13

] [ ] [ K
2 /
v

=
O M
v Ml (4.18)

The electroneutrality condition and its combination with the equilibrium then
yield:

3 / 1
v
/
) K 2 ( ] 2[ ] [ = =

O M
v Ml (4.19)

Thus under these conditions ] [
/
M
Ml is independent of the oxygen activity at
equilibrium with the excess MlO(s) phase. Thus, when the dopants are
compensated by fully ionised point defects as in this case, the system would
behave as if it had no excess or frozen-in dopant concentrations. However, the
solubility would change with temperature (through the enthalpy involved in K
v
)
contrary to the cases without any excess or with frozen-in concentrations.
The overall defect structure situation is illustrated in the Brouwer diagram in
Fig. 4.7; the important compensating defects are electron holes at comparatively
high oxygen activities and oxygen vacancies at low oxygen activities. Both
situations will eventually be overtaken by native non-stoichiometry at sufficiently
low viz. high oxygen activities.

Figure 4-7. Brouwer plot of the effect of lower valent cations,
/
M
Ml in M
2
O
3
when the
oxide MlO is present in amounts exceeding the solubility limit.

Doping M
2
O
3
with excess of MhO
2

Let us then dope M
2
O
3
with an excess of MhO
2
. The substitutionally
dissolved Mh
4+
ions have one positive effective charge,
M
Mh . They are
4.14
compensated by defect electrons or point defects with negative effective charge,
such as triply charged metal vacancies;
///
M
v .
If the compensating defects are electrons, the reaction equation for the
dissolution and the corresponding defect equilibrium are (under certain
assumptions, as before) given by

(g) O 3 2 2 (s) 2MhO
2 2
1
/
2
+ + + =
x
O M
O e Mh (4.20)

2 / 1
O
2 2
(s) MlO
2 / 1
O
3 2 2
e
2
2
2
p n ] [
a
p ] [ n ] [
K

= =
M
x
O M
Mh
O Mh
(4.21)

Thus, when MhO
2
is dissolved by compensation of electrons, the solubility of
MhO
2
increases with decreasing partial pressure of oxygen. If donor dopants and
electrons dominate, the electroneutrality condition and resulting
2
O
p dependency
become
8 / 1
O
2
p ] [ n

=
M
Mh . Under these conditions it may be shown that
8 / 3
O
///
2
p ] [
M
v .
When the compensating defects at higher
2
O
p become
///
M
v the reaction
equation for the dissolution of MhO
2
can be written

x
O M M
O v Mh 6 3 (s) 3MhO
///
2
+ + =

(4.22)

Thus, when the dissolution of MhO
2
is compensated by the formation of point
defects the solubility of MhO
2
is independent of the oxygen activity. Under these
conditions it may be shown that the concentration of electrons decreases with
increasing oxygen activity and is proportional to
4 / 1
O
2
p
. The overall defect

structure situation is illustrated by the Brouwer diagram below.
4.15

Figure 4-8. Brouwer plot of the effect of the dopant MhO
2
in M
2
O
3
in amounts exceeding
the solubility limit.

All in all, as a general rule the solubility of lower or higher valent oxides in
a parent oxide will be dependent on the oxygen activity when the compensating
defects are electronic defects, while they will be independent of the oxygen
activity when the compensating defects are fully ionised point defects. It may also
be noted that the temperature dependencies of the solubility will also be different
in the two cases.
Partially ionised point defects act as combinations of point and electronic
defects and generally gives oxygen activity dependent solubilities of aliovalent
dopants.
In real cases it may be difficult to reach equilibrium if the diffusion of the
cations to and from defect sinks (dislocations, grain boundaries a.o.) is slow.
Another point worth mentioning is that the excess second phase may well be a
compound of the parent and dopant oxide. For instance, MgO is normally present
in excess of the solubility in so-called high-purity alumina (Al
2
O
3
) and the
second phase in that case is spinel MgAl
2
O
4
. However, this does not alter the
oxygen activity dependencies; the ternary compound serves as a source/sink for
the dopant, with constant activity, and the defect chemical treatment remains
essentially the same.

Hydrogen defects in metal oxides

When a metal oxide is equilibrated in gas mixtures with hydrogen-
containing gases, e.g. H
2
O, hydrogen will dissolve in the metal oxide. The extent
4.16
of the dissolution of hydrogen will depend on the defect structure of the oxide and
the ambient oxygen and hydrogen activities. In studies of the dissolved hydrogen
it is then necessary to control both the oxygen and hydrogen (or water vapour)
activities.
Hydrogen may in principle dissolve in the form of different species: as
neutral atoms (H), hydride ions (H
-
), and protons (H
+
). In oxides the commonly
observed dissolved hydrogen species are the protons, often termed interstitial
protons,
i
H , but in reality always associated with oxygen ions as hydroxide
groups. If that oxygen sits on a normal lattice site we thus get substitutional
hydroxide defects,
O
OH .
A proper treatment of the dissolution of water and defect chemistry of
protons in oxides was first given by Stotz and Wagner (1967). Later on, Norby
(1987) and others have given extended treatments of hydrogen defect equilibria,
and Kreuer (1996) has reviewed protons in materials in general.
The dissolution of protons from hydrogen gas (e.g. in a gas mixture of
H
2
+H
2
O) may in these terms be written

/
2
2 2OH 2O (g) H e
O
x
O
+ = +

(4.23)

Assuming that water and hydrogen are in equilibrium through H
2
+ 1/2 O
2
=
H
2
O, the defect reaction may equally well be written with water as the source of
the protons:

) ( 2 2OH 2O O(g) H
2 2
1
/
2
g O e
O
x
O
+ + = +

(4.24)

and the corresponding defect equilibrium by

O
x
O
O
2
2
H
2
2 / 1
O
2 2
p ] [O
p n ] [OH
K
= (4.25)

Thus the concentration of protons in metal oxides is dependent on the partial
pressures of both the ambient oxygen and water vapour as well as the
concentration of electronic defects.

It may be useful to remind ourselves that water in this context is an oxide
present as a second phase in excess and, if we wish, at constant activity
(controlled partial pressure). It is thus analogous to the case of an excess of metal
oxide. The proton dissolves interstitially forming a positive defect (

i
H or
O
OH )
and thus behaves like a donor: When they are the dominating positive defects in
an otherwise undoped oxide they may be compensated by electrons or negative
4.17
point defects. In analogy with the higher-valent oxide dopant the first case would
lead to a proton concentration dependent on the oxygen activity while it in the
latter case would be independent.
As an additional possibility protons may also be compensated by a lower-
valent dopant (acceptor). This may, in turn, be present at constant concentration
(below the solubility or frozen-in) or itself in equilibrium with its own second
phase. In the following examples we will consider some of these relations.

Effect of water vapour on oxygen-deficient M
2
O
3
.

Let us first examine the effects of water vapour on the properties of an
undoped, oxygen-deficient oxide. We will use M
2
O
3
in the example, but most of
the treatment applies to any oxide. The predominant defects are electrons and
oxygen vacancies and the electroneutrality condition in dry environments is then
(from the preceding chapter)

6 / 1 3 / 1 /
2
) 2 ( ] 2[v n

= =
O vO O
p K (4.26)

Let us keep the partial pressure of oxygen constant and vary the partial pressure of
the water vapour. Protons are then dissolved in the oxide following Eq. 4.24.
However, the concentration of electrons is given by Eq. 4.26 and is independent of
O
2
H
p . Insertion into Eq. 4.25 gives that the proton concentration is proportional to
2 / 1
H
2
p
O
(at constant
2
O
p ). This is illustrated in the Brouwer diagram in the figure
below.

Figure 4-9. Brouwer plot of effects of water vapour on defect concentrations in oxygen
deficient M
2
O
3-d
.
4.18

At sufficiently high partial pressure of the water vapour, the dissolved protons
become the predominating point defects with positive effective charge and the
electroneutrality condition becomes

] [OH n

=
O
(4.27)

By combination with Eq. 4.25 we obtain that the concentrations of protons and
electrons are then proportional to
4 / 1
H
2
p
O
:

8 / 1 4 / 1
O H
4 / 1
2 2
p ] [OH n

= =
O O
p K (4.28)

This situation is part of the figure above. The figure also shows that the
concentration of oxygen vacancies decreases with the water vapour pressure,
illustrating that in general all charged native minority defects becoming dependent
on water vapour pressure when protons are dominating defects.

Effect of water vapour on acceptor-doped M
2
O
3

Let us next consider the effect of water vapour in the oxide M
2
O
3
doped
with the oxide MlO. Let us further assume that the concentration of the dopant ion
] [Ml
/
M
is sufficiently large that this is the predominant defect with negative
effective charge but that the amount of dissolved MlO is smaller than the
solubility limit of MlO in M
2
O
3
.
In dry atmospheres the dopant ions are assumed to be predominantly
compensated by the formation of doubly charged oxygen vacancies
( constant ] [Ml ] 2[v
/
= =

M O
) and under these conditions the concentration of
electrons is proportional to
4 / 1
2
p
O
and the electron holes to
4 / 1
2
p
+
O
, as derived earlier
for this case.
Let us then consider the effect of water vapour on the defect concentrations.
For the sake of simplicity let us keep the partial pressure of oxygen constant while
the partial pressure of water vapour is increased. Protons then dissolve in M
2
O
3

following Eq. 4.24, and from Eq. 4.25 it can be derived that the concentration of
protons dissolved as hydroxide defects, ] [OH
O
, is proportional to
2 / 1
2
p
O H
. This is
illustrated in the following figure.
4.19

Figure 4-10. Brouwer plot of the effect of water vapour (at constant oxygen pressure) on
defect concentrations in acceptor-doped, oxygen deficient M
2
O
3
.

Under these conditions the electroneutrality includes the concentration of
dissolved protons:

constant ] [Ml ] [OH ] 2[v
/
= = +

M O O
(4.29)

When we plot the concentration of the defects as a function of the partial
pressure of water vapour (at constant oxygen pressure) as shown in the Brouwer
diagram in Fig.4.10, it is seen that at sufficiently high water vapour pressures the
dissolved protons become the important positively charged point defects and the
electroneutrality condition is then approximated by

constant ] [Ml ] [OH
/
= =
M O
(4.30)

Under these conditions the concentration of oxygen vacancies decreases with
increasing partial pressure of water vapour; combining the equilibrium for
formation of oxygen vacancies (Eq. 4.8) and the equlibrium for dissolution of
protons, Eq. 4.25, and the dominating electroneutrality, Eq. 4.30, it is found that
the concentration of oxygen vacancies is proportional to
1
2
p
O H
. The concentration
of electron holes is correspondingly proportional to
2 / 1
2
p
O H
. These aspects are also
illustrated in Fig. 4.10.
It is also of interest to consider any effects of the partial pressure of water
vapour on the concentration of other minority defects. If so, the ambient partial
pressure of water vapour may affect the sintering and creep of metal oxides at
4.20
high temperatures. Let us continue to examine the acceptor-doped M
2
O
3
. For the
sake of illustration let us assume that the important defects in the metal sub-lattice
are metal vacancies and that these are triply charged,
///
v
M
. The concentration of
oxygen vacancies and metal vacancies in M
2
O
3
may be related through the
Schottky equilibrium.
In the region where the electroneutrality condition is given by
constant ] [Ml ] 2[v
/
= =

M O
, the concentration of the metal vacancies is also
independent of the partial pressure of water vapour, cf. Fig.4.10. However, in the
region where the electroneutrality condition is constant ] [Ml ] [OH
/
= =
M O
, the
concentration of the metal vacancies increases sharply with increasing partial
pressure of water vapour and is proportional to
2 / 3
2
p
O H
, cf. Fig. 4.10. Thus, under
these latter conditions the properties which are controlled by the concentration of
the minority defects, i.e. sintering, creep, may be greatly affected by variations in
the partial pressure of the ambient water vapour pressure. The direction of the
change will depend on the effective charge of the rate-limiting defect; whether it
is oxygen or metal defects, and whether it is vacancies or interstitials.
When, as in the last example, protons are dissolved at the cost of oxygen
vacancies, the reaction may be written

= + +
O
x
O O
2OH 2O v O(g) H
2
(4.31)

i.e. water vapour dissolves to fill the oxygen vacancies and replace their positive
charge with protons. Reaction 4.31 has been studied for a number of oxides, and it
is found that it always involves a negative enthalpy change such that oxygen
vacancies and protons are dominant at high and low temperatures, respectively.
For some oxides, the enthalpy is sufficiently negative that protons remain the
dominant defects up above 1000C in practically dry atmospheres and to even
higher temperatures under wet conditions. Other oxides have only small negative
enthalpies and have dominant protons up to only moderately high temperatures,
sometimes sufficiently low to kinetically prevent the reaction with water (Eq.
4.31) to reach equilibrium
The relative dominance of protons vs oxygen vacancies also varies with the
acceptor doping level: The higher charge of the vacancy relative to the proton
makes the former relatively more dominant at higher acceptor levels, while
protons are relatively more dominant at moderate acceptor levels.
Let us also briefly recollect the variable solubility of aliovalent dopants
treated previously in this chapter. While the acceptors are compensated by oxygen
vacancies, their solubility does not vary with anything else than the temperature,
as shown before. However, the dissolution of acceptors by simultaneous
dissolution of protons,

x
O O M
Ml O 2OH 2 O(g) H 2MlO(s)
/
2
+ + = +

(4.32)

4.21
will, when in equilibrium (MlO(s) present and diffusion enabled) give solubility
of acceptors which increases with the water vapour pressure:

1/4
H
/
2
p ] [Ml ] [OH
O M O
=
(4.33)

Thus, water vapour may in some cases play a role in the synthesis of solid
solutions containing acceptors.
By inspection of Fig. 4.10 it is evident that at sufficiently high water vapour
levels, the concentration of metal vacancies will become significant and
eventually dominant; the oxide attains a metal deficiency (or, in other cases,
oxygen excess), compensated by protons. Such defect structures can be considered
to be the first step towards the phase limit of the oxyhydroxide or hydroxide.

Association between impurities/dopants and
other point defects

As described in Chapter 1 impurities/dopants and native point defects may
associate to form complex defects. An important driving force may be the
coulombic attraction between impurities and point defects having opposite
effective charges, but also relaxation of the lattice around the associate may play a
role.
As an illustration, it will be assumed that the predominating defects are
doubly charged oxygen vacancies and that the cation dopants have one effective
negative charge, e.g.
/
Ml
M
.

The associated complex between

O
v and
/
Ml
M
can be
written
) (v
M O
Ml , and compared to the relatively mobile oxygen vacancy, this
defect can be considered as immobile as the dopant ions themselves.
The complex defect will be in equilibrium with the single defects according
to the reaction

= + ) (v Ml v
/
M M O O
Ml (4.34)

If all the species are randomly distributed, the equilibrium can be written

] ][Ml [v
] ) Ml [(v
K
/
a
M O
M O

= (4.35)

4.22
where K
a
is the equilibrium constant for the association reaction. If the acceptor is
relatively immobile, the associated oxygen vacancy can be considered trapped; it
is no longer able to contribute in charge or diffusional transport of oxygen, unless
an extra activation energy is supplied to break up the association.
Many other alternative examples of the effects of impurities/dopants on
defect equilibria may be cited, including association between acceptors and
protons, acceptors and holes, donors and point defects, and donors and electrons.

Summary

Impurities or dopants may dissolve interstitially or substitutionally, and the
defects formed are often easily ionised. Depending on whether they accept or
donate electrons they become acceptors and donors. In sufficient quantities these
may dominate the defect structure of the oxide. Thus, aliovalent cations
substituting host cations are often introduced in large quantities in oxides in order
to give high concentrations of ionic defects (for solid electrolytes) or electronic
defects (for electrodes etc.) Foreign species may be treated like other defects, but
are often assumed to be present in a fixed concentration due to the limited
availability of the species (full solubility) or because the concentration is frozen-
in. However, we have shown that the concentration of foreign species under
certain conditions may change as a function of temperature and for some defect
structures as a function also of
2
O
p . Protons dissolved from water vapour is a
special case which often has a large influence on the properties of oxides. They
form hydroxide groups with the oxygen ions of the oxide. These are effectively
positively charged and through them the defect structure may become a function
of the water vapour partial pressure.

Literature (protonic defects only)

Kreuer, K.-D. (1996) Chem. Mater. 8, 640.
Norby, T. (1987) Advances in Ceramics 23, 107.
Stotz, S. and Wagner, C. (1967) Ber. Bunsenges. Phys. Chem. 70, 781.

Exercises
1. Write the doping reaction for the doping of the Schottky-dominated metal
halide MX by the higher valent metal halide MhX
2
. Find the analytical
solution expressing the concentration of metal vacancies as a funtion of
doping concentration. Sketch also the temperature-dependency of all the
defects involved. (Hint: This is the initial illustrative case in the text use
the equilibria from that treatment. Remember that the doping reaction may
4.23
not be in equilibrium the dopant concentration is invariable as a function
of outer parameters.)
2. Sketch a Brouwer diagram (double logarithmic diagram of defect
concentrations vs p
O2
) for an oxygen deficient oxide doped with a
substitutional higher-valent dopant.
concentrations vs p
O2
) for an oxide M
1+x
O doped with a substitutional
lower-valent dopant.
concentrations vs p
O2
) for an oxide M
1+x
O doped with a substitutional
higher-valent dopant.
5. Sketch another Brouwer diagram, this time the double logarithmic diagram
of defect concentrations vs dopant concentration, for the oxide M
1+x
O
doped with a substitutional lower-valent dopant.
6. An oxide M
2
O
3
is dominated by intrinsic electronic disorder in the
stoichiometric state, while Schottky disorder is the next largest contributor
to defects. Draw a Brouwer diagram showing log defect concentrations as
a function of the concentration of lower-valent acceptor dopants. Do the
same for the case of higher-valent donor dopants. In both cases cover the
range from insignificant doping levels to the level where all possible
defect domains are exhausted.
7. An oxide M
2
O
3
is dominated by Schottky disorder in the stoichiometric
state, while intrinsic electronic disorder is the next largest contributor to
defects. Draw a Brouwer diagram showing log defect concentrations as a
function of the concentration of lower-valent acceptor dopants. Do the
same for the case of higher-valent donor dopants. In both cases cover the
range from insignificant doping levels to the level where all possible
defect domains are exhausted.
8. An oxide M
2
O
3-d
is doped substitutionally with MlO, and dominated by
oxygen vacancies and protons as compensating defects. Derive the full
expression for the concentration of protons (and oxygen vacancies) as a
function of a fixed dopant concentration and the equilibrium constant for
the equilibrium between oxygen vacanciens and protons. Make
assumptions and simplifications as necessary, but assume that both
vacancies and protons are significant compensating defects.
9. Derive equilibria, electroneutrality, and behaviour for an oxide M
2
O
3

dominated by metal vacancies and protons dissolved from surrounding
water vapour.
10. Choose an acceptor-doped oxide compensated by oxygen vacancies. Write
the reaction equation for the association reaction and the equilibrium
constant. Derive the temperature-dependency of the concentration of free,
unassociated vacancies. Assume that the entropy change of the reaction is
zero, and that the enthalpy change is 50 kJ/mol. What is the fraction of free
vacancies at 723 K?
4.24

5 Diffusion
5.1
5. 5. 5. 5. Diffusion Diffusion Diffusion Diffusion

Introduction

Numerous chemical reactions or micro-structural changes in solids take
place through solid state diffusion, i.e. the movement and transport of atoms in
solid phases. In crystalline solids, the diffusion takes place because of the
presence of defects. Point defects, e.g. vacancies and interstitial ions, are
responsible for lattice diffusion. Diffusion also takes place along line and surface
defects which include grain boundaries, dislocations, inner and outer surfaces, etc.
As diffusion along linear, planar and surface defects is generally faster than in the
lattice, they are also termed high diffusivity or easy diffusion paths. Another
frequently used term is short circuit diffusion.
The relative contribution of the different types of diffusion in oxides and
other inorganic compounds are functions of the temperature, partial pressures or
activities of the constituents of the compounds, the microstructure, grain size,
porosity etc. Grain boundary and dislocation diffusion generally have smaller
activation energies than lattice diffusion and as a result become increasingly
important the lower the temperature in solids with a given microstructure.
In the literature on diffusion and diffusion-controlled reactions or processes
one encounters many different terms that describe the diffusional behaviour under
different experimental conditions: tracer and self-diffusion of atoms and ions,
diffusion of defects, chemical diffusion, ambipolar diffusion, a.o. Many of these
are used for treating diffusion in compounds, and in the following chapters these
phenomena and terms will be described in more detail. Here we will start out with
a few simple phenomenological descriptions, and for simplicity we look only at
diffusion of neutral, independent particles.

Models of diffusion

It is a well known phenomenon that heat flows from hot to cold regions.
Such a flow of heat in a one-dimensional temperature gradient is described by
Fourier's law

dx
dT
j
q
= (5.1)

5 Diffusion
5.2
where j
q
is the heat flux density, i.e., the flow of heat per unit area of the plane
through which the heat traverses per second,
dx
dT
is the temperature gradient, and
(kappa) is the thermal conductivity. It may be noted that the minus sign reflects
that the heat flows from high to low temperatures (downhill).

Fick's first law
The expression for the flow of particles from high to low concentrations is
analogous to that for the flow of heat and is given by Ficks first law:

dx
dc
D j
particles
= (5.2)

Here, j
particles
is the particle flux density,
dx
dc
the concentration gradient of the
particles and D the diffusion coefficient. As in the equation for the heat flux, the
minus sign reflects that the particles flow from high to low concentration of
particles (downhill). This relation is named after A. Fick who first formulated this
relation. The particle flux and gradient is illustrated schematically in Fig.5.1.

Figure 5-1. Schematic illustration of Fick's first law. The negative of the particle
concentration gradient is the "driving force" of the diffusion.

j
particles
represents the number of particles (or moles) crossing a unit area
(cm
2
or m
2
) per unit time (seconds). If the concentration of particles is expressed
in number of particles (or moles) per cm
3
and the distance x in cm, the diffusion
coefficient has the dimension cm
2
s
-1
. In SI units the concentration is expressed in
number per m
3
and the diffusion coefficient has the dimension m
2
s
-1
.
In Fick's first law in Eq. (5.2) the negative particle gradient,
dx
dc
, may be
considered to be an expression of the "driving force" for the particle flux (Eq.5.2).
The larger the concentration gradient, the larger the particle flux. When Ficks
5 Diffusion
5.3
first law is applied to uncharged (neutral) and independently diffusing particles, it
is valid in the sense that the coefficient is a constant (independent of concentration
and gradient). This may apply e.g. to dilute solutions of neutral defects in solids.
Such defects may be vacancies, interstitials, and impurities in metals. In ionic
solids they may comprise neutral interstitial impurities, homovalent substituents,
or isotopic species. The applicability or inapplicability of Ficks first law will
become clearer later in this chapter and in the forthcoming chapters.

Potential gradients as the driving force
More generally, the driving force for the diffusion constitutes the chemical
potential gradient of the particles that diffuse (provided that no other forces act on
the particles). Correspondingly, the driving force for the transport of electrical
charges is the electrical potential gradient. In the following is given a brief
derivation of Fick's first law using a potential gradient as the driving force, in
detail in the case of a chemical potential gradient.
Let us consider the transport of particles of type "i" across a plane under a
driving force F. The particle flux density through a plane is given by the product
of the volume concentration c
i
of the particles at the plane and the average
migration or drift velocity v
i
of the particles

j
i
(moles or particles/(cm
2
sec)) = c
i
(moles or particles/cm
3
)
.
v
i
(cm/sec) (5.3)

For uncorrelated movements the drift velocity v of a particle is proportional
to the driving force F
i
exerted on the particle:

v
i
= B
i
F
i
(5.4)

The proportionality factor B
i
is termed the mobility ("Beweglichkeit") of the
particles and is defined as the average drift velocity per unit driving force. It is
often referred to as mechanical mobility to differentiate it from other types of
mobility.
The driving force is, in turn, given by the negative value of the potential
gradient normal to the cross-sectional area in the plane

dx
dP
F
i
i
= (5.5)

where P is a potential. The negative sign is, as above, due to the fact that the
transport takes place from higher to lower values of P, see Fig.5.2.

5 Diffusion
5.4

Figure 5-2. Schematic illustration of Fick's first law for transport in a potential gradient.

When one combines Eqs. 5.3 - 5.5, the particle flux becomes

dx
dP
B c v c j
i
i i i i i
= = (5.6)

Example of chemical potential gradient acting on neutral
particles
If the particle moves in a chemical potential gradient, the potential P equals
the chemical potential of particles of type i: P
i
=
i
. Equation 5.6 then takes the
form

dx
d
B c v c j
i
i i i i i
= = (5.7)

The chemical potential
i
is related to the chemical activity a
i
of species i through

i
=
o
i
+ kT lna
i
(5.8)

If ideal conditions can be assumed, the activity can be expressed by the
concentration divided by the concentration in the reference (standard) state:

0
i
i
i
c
c
a = (5.9)

and the chemical potential gradient is then given by
5 Diffusion
5.5

dx
dc
c
kT
dx
c d
kT
dx
d
i
i
i i
= =
ln
(5.10)

When one substitutes the expression for the chemical potential gradient in Eq.5.7,
the particle flux becomes

dx
dc
kT B
dx
d
B c j
i
i
i
i i i
= =

(5.11)

B
i
kT is termed the diffusion coefficient D
i
of particles i:

D
i
= B
i
kT (5.12)

By combination with Eq.5.12, Eq.5.11 takes the form of Fick's first law as given
in Eq.5.2:

dx
dc
D j
i
i i
= (5.2)

It is thus important to realise that Eq. 5.2 is fully valid only for ideal cases of
diffusion of neutral particles in chemical potential gradients only. If the particles
are charged, we need to take into account both the electrical potential gradent and
requirements to the combination of fluxes to maintain a given total current and
electroneutrality. These cases will be treated later on.
In the following, ideal conditions will be assumed and concentrations will
be used for activities of atoms, ions and various types of defects. But it should be
recalled that this always represents an approximation.

Simplified model for one-dimensional diffusion
Following Eq.5.2 the diffusion coefficient represents the proportionality
constant between the particle flux and the concentration gradient. In order to
describe the process of diffusion of particles or atoms in solids we need to realise
that the diffusion represents the sum of a large number of particles or atoms that
each makes a large number of jumps. We will therefore attempt to describe the
diffusion coefficient in terms of the number of jumps per unit time (the jump
frequency) and the distance that each particle or atom moves in each jump.
For this purpose let us consider a simplified one-dimensional model where
particles jump between parallel planes separated by a distance s as illustrated in
5 Diffusion
5.6
Fig.5.3. The two neighbouring planes under consideration are termed plane 1 and
2. The number of particles per unit area in plane 1 and 2 is termed n
1
and n
2
,
respectively, and let us further assume that n
1
>n
2
. Consider further that the
particles in plane 1 and 2 may jump from one plane to a neighbouring plane at a
jump frequency . The particles in plane 1 have an equal probability of jumping to
plane 2 and to the neighbouring plane in the opposite direction. The total number
of particles jumping out of a unit area of plane 1 per unit time is equal to the
product of number of particles per unit area times the jump frequency: n
1
. As the
particles may jump in opposite directions, the number of particles jumping from
plane 1 to 2 is given by n
1
. Correspondingly, the number of particles jumping
from unit area of plane 2 to plane 1 is given by n
2
. The difference in the jump
rates is equal to the net flux density of particles:

= ) (
2 1 2
1
n n j
particles
(5.13)

As all particles have the same jump frequency, there is a net flow of
particles from plane 1 to 2 because there are more atoms per unit area in plane 1
than in plane 2.

Figure 5-3. Schematic illustration of simplified model for one-dimensional diffusion

The number of particles n
1
belonging to unit area of plane 1 is given by the
volume concentration at plane 1, c
1
, times the extension in the x direction, i.e., the
plane separation s:

n
1
= c
1
s and n
2
= c
2
s (5.14)

By combining Eqs.5.13 and 5.14 the particle flux density becomes

= s c c j
particles
) (
2 1 2
1
(5.15)
5 Diffusion
5.7

The concentration gradient normal to the planes is termed
dx
dc
. The relation
between c
1
and c
2
can then be expressed by seeing that a difference is equal to the
gradient times the length;

dx
dc
s c c =
1 2
(5.16)

By insertion of this into Eq. 5.15 we obtain

dx
dc
s j
particles
=
2
2
1
(5.17)

This expression applies to one-dimensional diffusion of particles, and the factor
reflects that only of all jumps occur in the direction we consider as the flux
direction.
If the diffusion can take place in the three orthogonal directions, only one
third of the particles jump in one direction, and in the three-dimensional case the
net flux in one direction is 1/3 of the flux of that when all atoms jump in one
direction only:

dx
dc
s j
particles
=
2
6
1
(5.18)

By comparing with Fick's first law, Eq. 5.2, it is seen that the diffusion coefficient
D in the three-dimensional case is given by

=
2
6
1
s D (5.19)

If we consider a large number of jumps, n, which occurs during the time t, then

t
n
= (5.20)

and inserting this in Eq.5.19, one obtains

ns
2
= 6Dt (5.21)
5 Diffusion
5.8

The expression for the diffusion coefficient in Eqs.5.19 and 5.21 provides an
interesting qualitative description of what is going on during the diffusion. By
way of example, the diffusion coefficient for interstitial diffusion of oxygen atoms
in niobium metal at 800 C is approximately D = 7
.
10
-8
cm
2
s
-1
. The jump distance
can be assumed to be 1.65 (1.65
.
10
-8
cm), and then from Eq. (5.19) the jump
frequency is about 1.54
.
10
9
s
-1
. Thus, each oxygen atom makes a tremendously
large number of jumps per second. But it should then also be recalled that the
atoms vibrate with a (Debye) frequency of 10
12
-10
13
s
-1
, and thus only a small
fraction - about 1 in 10
3
or 10
4
- of the vibrations leads to a jump of the atoms.
We shall consider this in more detail when the different atomistic mechanisms are
treated below.
Although the number of jumps is tremendously large, the mean
displacement of each atom is relatively small most of the time it moves back and
forth. In the diffusion process it is not possible to observe the individual jumps of
the atoms, and it is necessary to find a relation between the individual atom jumps
for large number of atoms and the diffusion phenomena which may be observed
on a macroscopic scale. The problem is to find how far a large number of atoms
will move from their original sites after having made a large number of jumps.
Such relations may be derived statistically by means of the so-called random walk
method.

Random diffusion
Let us consider that the jumps of the atoms are random, i.e. that the jumps of
the atoms are independent of all the previous jumps and can occur in all
directions. In that case the displacement of a diffusing atom from the starting point
after n number of jumps, R
n
, is given by the algebraic sum of the individual jump
vectors:

=
= + + =
n
j
j n n
s s s s R
1
2 1
... (5.22)

If the individual jumps take place with equal probability in all directions and
the individual jump distances are equal, this algebraic sum equals zero. This does
not mean that the diffusing atom remains at its starting point after n jumps, but
only that jumps in "positive " and "negative" directions are equally probable. In
fact the total displacement may have any value between zero and ns.

In order to obtain a value for the magnitude (length) of the sum vector, one
squares Eq. 5.22:

5 Diffusion
5.9

= + = =
+ = =
1
1 1 1
2 2
2
2
n
j
n
j k
k j
n
j
j n n
s s s R R (5.23)

If, as above, the individual jump vectors are equal, i.e., |s
1
| = |s
2
| ... |s
j
| = s,
as, for instance, in crystals with cubic symmetry, and if they are random and
uncorrelated, then the second term on the right hand side of the equality sign in
Eq. 5.23 will approach zero for large numbers of jumps, as on an average s
j
and s
k

have an equal chance of being positive and negative. The first term on the righ
hand side, on the other hand, is always non-zero and positive, and thus represents
the average displacement length for a large number of jumps:

2
1
2
2
ns s R
n
j
j n
= =
=
(5.24)

The mean displacement is given by the square root of R
n
2
and is termed the root
mean square displacement:

s n R R
n n

2
= = (5.25)

From this it is seen that the mean displacement is proportional to the square root
of the number of jumps times the individual jump distance.
By combining Eqs. 5.21 and 5.24 we may express the random diffusion in
terms of the diffusion coefficient we dealt with in diffusion down a concentration
gradient in the 3-dimensional cubic case:

t D ns R
r n
6
2 2
= = (5.26a)

t D R
r n
6 = (5.26b)

where t is the time during which the mean-square-displacement takes place.
One may note that we have now started terming the diffusion coefficient D
r
(r for
random walk).
R
n
is the radius of the sphere that a diffusing atom on average will distance
itself by from the starting point after time t.

Let us now consider the displacement in a single dimension (e.g. the x-
direction) as a result of this three-dimensional displacement in the cubic case.
From simple geometry we have R
n
2
= 3x
2
where x
2
is the mean square
5 Diffusion
5.10
displacement in each orthogonal direction. Thus, the mean diffusion length in one
direction in a three-dimensional cubic crystal is:

t D x
r
2 = (5.27)

This length is thus shorter than the displacement radius, since displacements in the
y and z directions are wasted for x dimension displacement.
As described above and using oxygen diffusion in niobium as an example,
the oxygen atoms on an average exchange positions approximately 1.54
.
10
9
times
per second at 800 C. From the same considerations one may also estimate that an
oxygen atom has randomly covered total jump distances of 25.4 cm and 914.7 m
after 1 second and 1 hour, respectively. But what is the mean displacement? From
Eq 5.27 one may estimate that the one-dimensional root-mean-square
displacement after 1 hour only amounts to 0.022 cm. Thus, the mean displacement
is very small and on an average the oxygen atoms spend most of their time
jumping "back and forth".
What we did above was to consider random jumps, and then we related
jumps, times and distances to a term we recognised from earlier, namely the
diffusion coefficient, and we named it the random diffusion coefficient D
r
. The
diffusion coefficient was in turn something we recognised while considering
diffusion down a one-dimensional concentration gradient from Ficks first law.
One should note, however, that random diffusion and the random diffusion
coefficient can be considered and expressed and quantified also in the absence of
a concentration gradient and also for charged particles.
As will be described below, diffusion is often measured by using tracer
atoms and one then obtains values of the diffusion coefficient of the tracer atoms.
Depending on the diffusion mechanism the tracer diffusion is in most cases not
completely random, but is to some extent correlated with previous jumps. This
will be further discussed later on.

Fick's second law

As described above Fick's 1
st
law assumes a fixed concentration gradient
across the plane through which the flux of particles take place. But in numerous
practical cases the concentration and concentration gradient changes with time.
Such cases are covered by Fick's 2
nd
law. This is shown schematically in Fig.5.4
which illustrates the change in the concentration gradient through the solid. As
shown in the figure let us consider a region within a solid, enclosed between
planes separated by the distance dx. The net particle flux from the region of higher
concentration into dx is j
1
and the net particle flux out of dx towards lower
concentration is j
2
. When j
1
is greater than j
2
the particle concentration in dx
increases with time. This requires that the concentration gradient in plane 1 is
5 Diffusion
5.11
larger than in plane 2, since Fick's first law applies in both planes (and at any
position) at any time.
The change in concentration per unit time at any position is proportional to
the gradient in flux at that position

x
-
t
j c
= (5.28)

Although the concentration and concentration gradient change with time, Fick's 1.
law is valid at any one time and position and thus

)
x
c
(D
x x
-
t
= =
j c
(5.29)

If D can be considered to be independent of concentration then

2
2
x
c
x
-
t
D
j c
= = (5.30)

The intuition of the qualitative arguments as well as the double derivative of the
concentration in Eq.5.30 tell us that particles will flow from convex to concave
regions in terms of distribution of particles.

Figure 5-4. Schematic illustration of Ficks 2
nd
law; the concentration gradient changes
with time.

Equations 5.29 and 5.30 are representations of Fick's 2
nd
law. It may be
solved explicitly under certain boundary conditions that may be closely
approximated experimentally (e.g. Crank (1975)). A couple of examples of this
are given in the following.

5 Diffusion
5.12
Measurements of the diffusion coefficients by tracer
techniques
The use of isotopes or tracers is a common means of studying diffusion.
Tracer methods permit measurements of self-diffusion, that is, the diffusion of the
crystal components in a crystal. Furthermore, they allow measurements of
diffusion in homogenous materials, that is, without imposing chemical gradients
(when one disregards the difference in atomic weight between atoms in the crystal
and of the tracer).
A common technique is to deposit a very thin film of radioactive isotopes on
a plane surface of a sample, and, after subsequent diffusion anneal, determine the
activity of diffusion species as a function of distance from the plane surface. If the
thickness of the sample is very much larger than the penetration depth of the
tracers, the solid can be considered semi-infinite. Furthermore, if the diffusion is
homogenous (e.g. taking place by lattice diffusion), the concentration of the
diffusing tracers normal to the plane is through solution of Fick's second law with
appropriate boundary conditions given by

)
4
exp(
( 2
2
t D
x
t D
c
c
t t
o
)
=
1/2
(5.31)

c is the activity (or concentration) of the tracer at a distance x from the surface, c
o

is the activity originally present on the surface, and t is time of the diffusion
anneal. D
t
is the tracer diffusion coefficient. Following Eq.5.31 it is determined by
plotting lnc vs x
2
, in which case the resultant straight line has the slope -
t D
t
4
1
.
Plots of c vs x and of lnc vs x
2
according to Eq.5.31 is illustrated in Fig.5.5a and
b, respectively. At the point where the activity is half of the activity at the surface,
then x = t D
t
77 . 2 . This distance corresponds approximately to the root-mean-
square penetration distance (Eq. 5.26b).

Figure 5-5. Graphical presentation of relationship between activity, c, and penetration
distance, x, (Eq.5.31) for homogeneous diffusion of tracer initially deposited as a thin film on the
surface the solid. a) c vs. x; b) lnc vs. x
2
.
5 Diffusion
5.13

An alternative way of performing the experiment is to have a constant
surface composition of the diffusing species. By solving Fick's second law under
proper boundary conditions the penetration of the diffusing species is then
described by the relationship

2 / 1
0
0
) ( 2
1
t D
x
erf
c c
c c
t s
=
(5.32)

where c is the concentration of the diffusing species at penetration distance x at
time t, c
s
is the constant surface concentration, and c
o
is the original concentration
in the solid. D
s
is the diffusion coefficient of the diffusing species. Figure 5.6
shows a graphical penetration of the diffusion profile according to Eq.5.32 in a
case where c
0
=0. It may be noted that at the point where
0
0
c c
c c
s

= , that is, at
the point where the concentration of the diffusing species is midway between the
surface composition and the original composition, then x/ t D
t
= 0.954. Thus at
this point x ~ t D
t
.
It may be noted that Eqs.5.31 and 5.32 involve the dimensionless parameter
x/(2 t D
t
), and accordingly the penetration and the amount of the diffusing
species dissolving in the solid are proportional to the square root of time.

Figure 5-6. Graphical presentation of diffusion profile when the surface concentration of
the diffusing species remains constant with time. It is assumed that c
o
=0 (Eq.5.32).

The penetration by the diffusing species may be measured by means of the so-
called sectioning method, that is cutting, grinding or etching off thin sections of
layers of the sample parallel to the plane surface and subsequently determining the
concentration of the diffusing species in each section. By cutting the specimen
normal to or at an angle to the plane surface, the penetration of radioactive tracers
may also be measured by means of so-called autoradiography.
5 Diffusion
5.14
Electron microprobe analysis and secondary ion mass spectroscopy (SIMS)
combined with sputtering techniques also provide excellent tools for studying
penetration and diffusion of foreign ions. As oxygen does not have a radioactive
isotope suitable for tracer studies, SIMS is particularly useful for studying oxygen
diffusion employing the stable
18
O isotope.
In the isotope exchange method the vapour of the diffusing component
surrounding the sample is enriched with either a radioactive or an inactive isotope,
and the diffusion is measured by following the exchange of the isotope with the
sample. The diffusion coefficient may be evaluated if diffusion is the slower
process and processes at the surface are rapid. Alternatively, the sample itself may
be isotopically enriched, and the increase in concentration of the isotope in the
vapour phase may be measured.
Diffusion rates may, in principle, also be determined from any property or
reaction which depends on atomic mobility. By way of illustration, ionic
conductivity of the anion is directly proportional to the anion diffusion coefficient
(see Electrical conductivity). From high temperature solid state reactions,
sintering, oxidation of metals etc. diffusion coefficients may be evaluated
provided the detailed mechanism of the processes are known. Examples of this
will be given in Chapter 7.

Diffusion mechanisms

Lattice diffusion takes place through the movement of point defects. The
presence of different types of defects gives rise to different mechanisms of
diffusion. These are illustrated schematically for elemental solids in the following
descriptions. But they also apply to metal oxides and other inorganic compounds
when the diffusion is considered to take place in the sublattices of the cations or
anions.

Vacancy mechanism
The diffusion is said to take place by the vacancy mechanism if an atom on
a normal site jumps into an adjacent unoccupied lattice site (vacancy). This is
illustrated schematically in Fig.5.7. It should be noted that the atoms move in the
direction opposite the vacancies.
5 Diffusion
5.15

Figure 5-7. Schematic illustration of vacancy diffusion in solids.

Interstitial mechanism
If an atom on an interstitial site moves to one of the neighbouring interstitial
sites, the diffusion occurs by an interstitial mechanism. This is schematically
shown in Fig.5.8. Such a movement or jump of the interstitial atom involves a
considerable distortion of the lattice, and this mechanism is probable when the
interstitial atom is smaller than the atoms on the normal lattice positions.
Diffusion of interstitially dissolved light atoms, e.g. H, C, N, and O in metals
provides the best known examples of this mechanism.

Figure 5-8. Schematic illustration of interstitial diffusion in solids.

Oxides with close-packed oxygen lattices and only partially filled
tetrahedral and octahedral sites may also facilitate diffusion of metal ions in the
unoccupied, interstitial positions. Finally, even large anions may diffuse
interstitially if the anion sublattice contains structurally empty sites in lines or
planes which may serve as pathways for interstitial defects. Examples are rare
earth sesquioxides (e.g. Y
2
O
3
) and pyrochlore-type oxides (e.g. La
2
Zr
2
O
7
) with
fluorite-derived structures and brownmillerite-type oxides (e.g. Ca
2
Fe
2
O
5
) with
perovskite-derived structure.

Interstitialcy mechanism
If the distortion becomes too large to make the interstitial mechanism
probable, interstitial atoms may move by another type of mechanism. In the
5 Diffusion
5.16
interstitialcy mechanism an interstitial atom pushes one of its nearest neighbours
on a normal lattice site into another interstitial position and itself occupies the
lattice site of the displaced atom. This mechanism is illustrated schematically in
Fig.5.9
In the interstitialcy mechanism one may distinguish between two types of
movements. If the atom on the normal lattice site is pushed in the same direction
as that of the interstitial atom, the jump is termed collinear (Fig.5.9). If the atom is
pushed to one of the other neighbouring sites so that the jump direction is different
from that of the interstitial atom, the jump is termed non-collinear.

Figure 5-9. Schematic illustration of interstitialcy diffusion in solids. Collinear jump.

Other mechanisms.
In elemental solids also other mechanisms have been proposed. The
crowdion is a variant of the interstitialcy mechanism. In this case it is assumed
that an extra atom is crowded into a line of atoms, and that it thereby displaces
several atoms along the line from their equilibrium positions. The energy to move
such a defect may be small, but it can only move along the line or along
equivalent directions.
For metals it has also been proposed that diffusion may take place through a
so-called ring mechanism, but this mechanism is improbable in oxides or other
inorganic compounds.

Diffusion of complex defects.
As defects in oxides with large deviations from stoichiometry constitute
complex defects, there has been considerable discussion and speculation about the
diffusion mechanism in such oxides. It has, for instance, been suggested that
complex defects coexist in a dynamic equilibrium with single defects, and that the
diffusion processes also under these conditions really involve diffusion of single
defects. In the case of defect clusters it has alternatively been proposed that the
smaller clusters may move as a unit. A translational mechanism that has been
proposed for a 4:1 cluster in wustite is illustrated in Fig.5.10. The jump processes
in the motion of a 4:1 complex is quite complex, and the mechanism requires two
distinct, sequential jumps. Atom 1 jumps to fill a vacancy in the complex defect
and thereby creates a new vacancy. In the process the interstitial ion in the cluster
5 Diffusion
5.17
is displaced to the neighbouring vacancy. The movement of the cluster is
completed when atom 2 is displaced from a normal site to the interstitial site of
the new location of the cluster.

Figure 5-10. Proposed mechanism for the translational movement for a 4:1 cluster. Open
circles represent oxygen ions, closed circles iron atoms, open circles with an "x" iron interstitials,
and cubes iron vacancies.

Diffusion of protons in metal oxides.
Protons that dissolve in metal oxides associate with the oxide ions and form
hydroxide ions. As the proton has no electron shell, it interacts strongly with the
electron cloud of the oxide ion and in its equilibrium position in the hydroxide ion
it is embedded in the valence electron cloud. The O-H bond that is formed has a
bond length less than 100 pm; this may be compared with the ionic radius of 140
pm of the oxide ion.
In principle the protons may move by two different mechanisms: i) the free
transport mechanism, which is also alternatively termed the Grotthuss mechanism
or ii) the vehicle mechanism.
The free transport is the principal mode of transport of protons in oxides,
and in this mechanism protons jump from one oxygen ion to a neighbouring one.
After each jump the proton in the hydroxide rotates such that the proton reorients
in the electron cloud and becomes aligned for the next jump. This is illustrated
schematically in Fig.5.11. The rotation and reorientation is believed to involve a
small activation energy and the jump itself is considered to be the rate-
determining step.
In the vehicle mechanism the proton is transported as a passenger on an
oxide ion. Thus this mechanism may be considered to constitute transport of
hydroxide ions. The hydroxide ion may in principle move by an oxygen vacancy
mechanism or as an interstitial hydroxide ion. It may be noted that the hydroxide
5 Diffusion
5.18
ion has a smaller radius and charge than the oxide ion and may as such be
expected to have a smaller activation energy for diffusion than the oxygen ion.
Also other species such as water molecules and hydronium ions, H
3
O
+
, may serve
as vehicles for protonic diffusion, notably in relatively open structures.

Figure 5-11. Schematic illustration of free transport of protons in metal oxides (Grotthuss
mechanism).

Further considerations of factors that affect the
diffusion coefficient in crystalline solids

In Eq. 5.19 the random diffusion coefficient has been expressed in terms of
the jump distance and the number of jumps per unit time:

t
n
s s D
r
2
6
1
2
6
1
= = (5.19)

But it is necessary to further characterise D in different crystal structures
(lattices or sub-lattices) and subsequently to derive expressions for the
temperature and oxygen pressure dependence of diffusion in metal oxides.

Vacancy diffusion.
Let us consider vacancy diffusion in an elemental solid or a cation or an
anion sub-lattice. The number of jumps per unit time, , depends on several
factors. First, it depends on the jump frequency towards an adjacent site.
Furthermore, it is also proportional to the number of sites to which the atom may
jump, i.e., the number of nearest neighbour positions of the atom, Z. Finally, the
atom may only jump if a vacancy is located on an adjacent site, and this
probability is given by the fraction (concentration) of vacancies in the crystal, N
d
.
Thus, is in this case given by
5 Diffusion
5.19

d
ZN = (5.33)

In crystalline solids the jump distance is a function of the crystal structure
and may be expressed as a function of the lattice parameter.
In a bcc crystal of an elemental solid, for instance, each atom has 8 nearest-
neighbour positions or atoms, and thus in this case Z = 8. From simple
geometrical considerations of the crystal structure it may further be shown that the
jump distance is given by
2
3
0
a s = , where a
o
is the lattice parameter. When
inserting these values of Z and s in Eq. 5.33 one obtains

d r
N a D
2
0
= (5.34)

In general, D
r
for a cubic structure is written

d r
N a D
2
0
= (5.35)

where is a geometrical factor involving the factor 1/6 (from Eq.5.21, the factor
Z from Eq.5.33 and the relation between the jump distance and the lattice
parameter. For vacancy diffusion in a bcc lattice is thus equal to unity, From the
same considerations it may also be shown that = 1 for vacancy diffusion in fcc
lattices.

Interstitialcy diffusion
Consider an atom on a normal lattice site of a cation or anion sub-lattice. If
this atom is to move by the interstitialcy mechanism, an atom on a nearest
neighbour interstitial site has to push the atom on the normal site to a
neighbouring interstitial site. Thus for this diffusion mechanism an atom may only
diffuse when it has an interstitial atom on a neighbouring site, and as for vacancy
diffusion the diffusion coefficient of the atoms is proportional to the fraction
(concentration) of interstitial atoms or ions in the sub-lattice.

Interstitial diffusion
When one considers interstitial diffusion of an interstitially dissolved
species in dilute solid solution, essentially all the nearest neighbour interstitial
sites of the same type are unoccupied and available for occupancy by the diffusing
interstitial atoms. Thus the interstitial atom may jump to any of the nearest
neighbour interstitial sites and in this case N
d
is equal to unity. The interstitial
diffusion coefficient is then given by
5 Diffusion
5.20

Z s D
r
2
6
1
= (5.36)

As mentioned above the best known examples of this mechanism is diffusion of
O, N, C, and H atoms interstitially dissolved in metals. By way of example
oxygen and nitrogen atoms in bcc metals, e.g. in the group 5 metals V, Nb, and
Ta, occupy octahedral sites, and in this case each interstitial atom has 4 nearest
neighbour octahedral sites to which they may jump, thus
Z= 4. Furthermore, the jump distance s is equal to
a
o
2
. Inserting these values
in Eq.5.36 the diffusion coefficient for interstitial diffusion between octahedral
sites in a bcc lattice becomes

2
6
1
s D
r
= (5.37)

Thus in this case =1/6.

Free proton transport
Similar considerations may be applied to free transport of protons (cf.
Fig.5.11). For dilute solutions of protons in an oxide essentially all nearest
neighbour oxygen ions are available, and thus in this case N
d
is unity. However,
the specification of Z, s and is not straightforward in this case. The dynamics of
free proton diffusion in oxides are complicated by 1) the multistep process
(jump+rotation), 2) the dependency on the dynamics of the oxygen ion sublattice,
and 3) the quantum mechanical behaviour of a light particle such as the proton.
The uniquely large ratios between the masses of the isotopes of hydrogen
give rise to a number of strong isotope effects in the case of diffusion of protons.
(These are also in principle operative for diffusion of hydrogen atoms or hydride
ions, but they would be essentially negligible for diffusion of protons on a heavier
vehicle, such as in OH
-
.) The isotope effects for proton diffusion can be classified
as follows: The attempt frequency (in the pre-exponential of ) is given as the
O-H stretching frequency, and it is given by the inverse of the square root of the
reduced mass of the harmonic oscillator. The reduced mass equals
(m
O
+m
H
)/m
O
m
H
and is roughly inversely proportional to the mass of the hydrogen
species. Therefore, the ratios of the pre-exponentials of the diffusion coefficients
of protons, deuterons, and tritons are approximately related by D
0H
: D
0D
:

D
0T
=
1 : 1/ 2 : 1/ 3 . This is called the classical effect. Furthermore, the oscillators
have different ground-state or zero-point energies, such that diffusion of lighter
isotopes may be expected to have a slightly smaller activation energy of jumping.
Accordingly, proton diffusion typically has 0.04 - 0.06 eV lower activation energy
than deuteron diffusion. This is called the non-classical effect. However, there are
more factors involved, connected to the fact that the light proton/deuteron/triton
5 Diffusion
5.21
must be treated quantum mechanically and to their dynamics in a dynamic lattice
of much heavier oxygen ions. For instance, the lighter isotope has a lower sticking
probability after an otherwise successful jump. In effect, this reduces the effective
diffusivity of all hydrogen isotopes and it can to a varying degree counteract or
even seemingly reverse the classical effect. The neglectance of the latter have in
general made many investigators ascribe higher diffusivities for protons compared
to deuterons to the classical effect, while in reality the non-classical zero-point
energy difference appears to be the main contributor to the observed effect.
The possibility of tunnelling as a major component of diffusion is not
expected to apply to protons except at very low temperatures, and the orders-of-
magnitude isotope effects that would be expected for proton vs deuteron or triton
diffusion have not been reported for oxidic materials at elevated temperatures.

Temperature and oxygen pressure dependence
of diffusion in metal oxides

We will now address how diffusion coefficients vary with temperature and
activity of the components of the compound, mainly the oxygen partial pressure
over oxides. In order to evaluate these dependencies one must consider the
temperature and activity dependencies of N
d
(e.g. cf. Eq.5.35) and the temperature
dependence of . We start by analysing N
d
for some cases.

Dependencies related to the concentration of defects
Vacancies in an elemental solid
For the sake of simplicity let us first consider the case of diffusion by a
vacancy mechanism in a pure elemental solid. The diffusing atoms may only
make a jump when a neighbouring site is vacant. Thus the jump frequency n/t is
proportional to the vacancy concentration and as described in Chapter 3 the
fraction of vacancies may be written

) exp( ) exp( ) exp(
RT
H
R
S
RT
G
N
d d d
d

=

= (5.38)

where G
d
, S
d
, and H
d
denote the Gibbs free energy, entropy, and enthalpy of
formation of the vacancies (defects). In elemental solids H
d
is positive and the
vacancy concentration increases with increasing temperature.

Vacancies in an oxygen deficient oxide.
In a nonstoichiometric oxide the concentration of the predominating point
defects will be a function of temperature but also of the oxygen pressure. By way
5 Diffusion
5.22
of illustration, let us consider an oxygen deficient oxide, M
a
O
b-
, in which
doubly charged oxygen vacancies are the predominating point defects. If intrinsic
ionisation and effects of impurities can be neglected, the concentration of the
oxygen vacancies is as described in previous chapters (Eq. 3.68) given by

)
3
exp( )
3
exp( ) ( ) ( ] [
6 / 1 3 / 1
4
1
6 / 1 3 / 1
4
1
2 2
RT
H
R
S
p p K v N
O O
O
v v
O O
v
O d

= = =

(5.39)

where

O
v
K represents the equilibrium constant for the formation of doubly
charged oxygen vacancies, and

O
v
S and

O
v
H their entropy and enthalpy of
formation, respectively.
In such an oxide the oxygen vacancy concentration may also be determined
by the presence of dopants or a sufficiently large level of impurities with negative
effective charge. If the dopant is doubly negatively charged (as, for instance,
Ca
2+
, an acceptor dopant, in ZrO
2
), then

] [A ] [
//
M O
v =

(5.40)

In this case ] [

O
v will be independent of oxygen pressure and most often
temperature (see Chapter 4).
In a similar manner one may obtain the temperature and oxygen pressure
dependencies of the concentration of any defect (majority or minority) when the
defect structure is known. Of course the concentrations of acceptors or donors will
enter when they dominate the defect situation and the water vapour partial
pressure may enter when protons are dominating defects.

Temperature dependence of the attempt frequency
When atoms jump or move between definite sites in the crystal, they have to
surmount energy barriers. A large part of this energy barrier involves the strain
energy required to displace neighbouring atoms to create a sufficiently large
opening between the atoms to permit the atom jump. The potential energy of the
atom diffusing from one site to another may be qualitatively illustrated as shown
in Fig. 5.12. The potential barrier height is H
m
and represents the activation
energy which the atom has to surmount during the jump. Each atom vibrates in its
position and during a fraction of time, which is given by the Boltzmann
distribution factor exp (-H
m
/RT), it possesses sufficient energy to overcome the
energy barrier. The jump frequency is thus proportional to exp(-H
m
/RT).
5 Diffusion
5.23

Figure 5-12. Potential energy of atom diffusing in a solid. H
m
is the activation energy for
the diffusion.

A more complete analysis based on the theory of activated complexes and
on statistical mechanics has been given by Zener (1951,1952). He considered the
system or an atom in its initial equilibrium condition and in the activated state at
the top of the potential barrier which separates the initial position from its
neighbouring equilibrium position. The rate of transition from one equilibrium site
to another is given by

RT
H
exp
R
S
exp
RT
G
exp
m m m

=

= (5.41)

where G
m
, S
m
, and H
m
represent the free energy, entropy and enthalpy
change, respectively, connected with the movement of the atom from the
equilibrium position to the top of the potential barrier and (nu) represents the
vibration frequency. is often assumed to equal the Debye frequency, i.e. about
10
13
Hz, as an order of magnitude approximation. (For protons in oxides the O-H
stretching frequency of about 10
14
Hz is used.) S
m is often assumed to be a
small, positive term (below, say, 10 J/molK).

As a more detailed analysis of , Zener suggested that may be
approximated as = /a H M
m
/ , where is a structure- and mechanism-
dependent factor, a is a lattice parameter, and M is the reduced mass of the
oscillator. Intuitively, a particle vibrating in an energy valley will vibrate faster (
increases) when the walls become steeper (H
m
increases). One may also view
the particle as vibrating on a spring; the frequency becomes higher when she
spring is shorter (/a decreases), when the spring is stiffer (H
m
increases) or the
particle becomes lighter (M decreases). Thus, the temperature-independent (pre-
exponential) term in the diffusion coefficient increases when H
m
increases, so
that the two tend to counteract each other. Experimental observations of this is
sometimes referred to as the Meyer-Neldel effect (Meyer and Neldel (????)).

5 Diffusion
5.24
Resulting analysis of the diffusion coefficient D
r

Vacancy diffusion in an elemental solid. From Eq.5.35 it is seen that
the temperature dependence of D
r
for vacancy diffusion in an elemental solid is
determined by that of N
d
and . For an elemental solid with cubic structure, D
r
is
thus obtained by combining Eqs.5.35, 5.38 and 5.41:

RT
) H H (
exp
R
S S
exp
m d m d 2
0

=
+ +
a D
r
(5.42)

Experimentally determined values of diffusion coefficients are usually obtained as

RT
exp
0
Q
D D

= (5.43)

where Q is termed the activation energy and D
o
the the pre-exponential factor.
By comparing Eqs.5.42 and 5.43 it is seen that the activation energy, Q, in
this case comprises

Q = H
d
+ H
m
(5.44)

Correspondingly, D
o
is given by

R
S S
exp
m d 2
0 0

=
+
a D (5.45)

If experiments are carried out under such conditions that the concentration
of defects, N
d
, is constant and independent of temperature, e.g. at sufficiently low
temperatures that the defect concentration is frozen in, then D
r
is given by

RT
H
exp
RT
H
exp
R
S
exp
m
0
m m
,
2
0

=

= D N a D
frozen d r
(5.46)

and correspondingly the activation energy is under these conditions given simply
by

Q = H
m
(5.47)

5 Diffusion
5.25
Oxygen vacancy diffusion in oxygen-deficient oxides.
In an oxygen-deficient oxide in which oxygen vacancies predominate and
for which effects of impurities can be neglected, the oxygen vacancy
concentration is given by Eq. 5.39. Correspondingly, the oxygen diffusion
coefficient for random oxygen vacancy diffusion in the oxide in equilibrium with
the ambient oxygen gas at a partial pressure becomes

RT
3
exp
R
S
3
exp ) (
)
m
v
(
m
v
6 / 1 3 / 1
4
1
2
0
O O
2
H
H S
p a D
O r
=
+ +

(5.48)

D
r
thus increases with decreasing oxygen pressure. The activation energy for the
diffusion is in this case given by

m
v
3
O
H
H
Q
= +

(5.49)

If the concentration of oxygen vacancies is determined by lower valent impurities
or dopants, e.g. when Eq.5.40 applies, then D
r
is given by

RT
H
exp
R
S
exp ] [A
m m // 2
0

=
M r
a D (5.50)

Thus in this case the activation energy is equal to that of the mobility of the
oxygen vacancies: Q = H
m
. A similar situation would arise if the vacancy
concentration was frozen in rather than determined by acceptor doping.
Such a transition from intrinsic to extrinsic diffusion may take place when
the temperature is lowered from high temperatures, where the native point defects
predominate, to low temperatures, where the point defect concentrations are
determined by the impurity concentration or, in other cases, frozen in. The
temperature dependencies and the corresponding change in activation energy of
the random diffusion in such a case are illustrated in Fig. 5.13.
The above situations represent ideal cases. As the temperature is decreased
defect interactions may become increasingly important. This may be treated as
formation of associated defects. In the non-stoichiometric (intrinsic) case one may
for instance have association between charged vacancies and electrons forming
singly charged or neutral vacancies. This will change the oxygen pressure
dependency of N
d
and probably the temperature dependencies of N
d
and , but
probably not dramatically. Of larger effect, and more frequently observed, are the
associations between the mobile vacancies and the relatively stationary acceptors
in the extrinsic regime: The associated vacancies can be regarded as immobilised,
5 Diffusion
5.26
and the concentration of mobile vacancies (N
d
) starts to decrease with decreasing
temperature. This is seen as an increasing activation energy of diffusion with
decreasing temperature in many heavily doped oxides. One may note that instead
of expressing the effect as a changing concentration of free vacancies one may
express it as a changing mobility in that the activation energy for diffusion is
increased by the trapping energy exerted by the acceptor. However, the simple
model of free and stationary (associated) point defects appears capable of
explaining most behaviours fairly well.

Figure 5-13. The diffusion coefficient for oxygen diffusion by the vacancy mechanism in an
oxygen deficient oxide in which oxygen vacancies are the predominant native point defects. At high
temperatures the oxide exhibits intrinsic behaviour and at reduced temperatures extrinsic
behaviour (i.e. the oxygen vacancy concentration is determined by the concentration of lower
valent cations).

Interstitial diffusion of solute.
From Eq.5.36 it is seen that N
d
does not enter into the expression for the
diffusion coefficient for interstitial diffusion in dilute solutions, thus in this case
the activation energy, Q, represents that of the mobility of the diffusing interstitial
atoms: H
m
= Q.
For interstitial diffusion between octahedral sites in bcc metals D
o
is by
combination of Eqs. 5.37 and 5.41 given by

R
S
exp
m 2
0 6
1
0
= a D (5.51)

5 Diffusion
5.27
Assuming that ~ 10
13
s
-1
and a
o
= 1.5
.
10
-8
cm, and as it is probable that S
m
>
0, one may estimate a lower limit of D
o
of

D
o
> 5
.
10
-4
cm
2
s
-1
(5.52)

The diffusion of protons by the free transport mechanism is another case of
interstitial diffusion of a solute.

Interstitial diffusion of a constituent
The diffusivity of a constituent such as the host metal or oxygen ions by an
interstitial mechanism is not only proportional to the probability that the
interstitial defect jumps, but also to the probability that a constituent ion is
interstitial, i.e., the fractional concentration of interstitials. Thus the diffusion
coefficient of the constituent contains the temperature and oxygen pressure
dependencies of the concentration of interstitials in addition to the temperature
dependency of the mobility of these defects. As in the case of vacancy diffusion,
the fixation of the defect concentration by doping or freezing as well as
association and trapping of defects apply also to interstitial diffusion.

Diffusion coefficients of point defects

In the above treatment of vacancy diffusion, only the diffusion coefficients
of the atoms have been considered. For many purposes it may be convenient to
consider the diffusion coefficients of the vacancies themselves.
When an atom diffuses by the vacancy mechanism, it can only jump if a
vacancy is located on an adjacent site, and the number of jumps per unit time is
thus proportional to N
d
(Eq. 5.33). However, the vacancy itself can jump to any
one of the occupied nearest neighbour positions, provided it is occupied by an
atom. Accordingly the vacancy diffusion coefficient D
V
for a cubic system is
given by (cf. Eq.5.35)

D
V
= a
o
2
(5.53)

where N is the fraction of occupied atom positions. In dilute solutions of
vacancies, N~1, and the diffusion coefficient of the vacancies is then not
dependent on N
d
.
From Eqs. 5.35 and 5.53 D
V
for vacancies is related to D
r
for the atoms
through the relation
5 Diffusion
5.28

D
r
N = D
V
N
d
(5.54)

where, as stated above, N denotes the fraction of sites occupied by atoms, often
approximated as ~1. This relation can be generalised to be a very important and
useful approximation for any point defect:

D
r
N = D
d
N
d (
5.55
)

where D
d
is the defect diffusion coefficient. It proves to be a good approximation
for component diffusion by the interstitialcy mechanism. It is also a useful
approximation for component diffusion by interstitial diffusion in the case of
small defect concentrations (N~1) but as the concentration of defects increases the
term N must reflect the number of unoccupied interstitial sites.
Eq. 5.55 still holds for interstitial diffusion of a dilute solution of an
interstitially dissolved solute, such as light elements (H, C etc. in metals or
protons in oxides), but in this case the number of solute atoms or ions and the
number of defects is of course the same, so that in these cases the diffusion
coefficient for the solute and for the (interstitial) defects is the same.

Excercises

1. Random (self) diffusion
a) The self-diffusion coefficient of a metal with cubic structure can be
expressed as
2
6
1
s
t
n
D =
where n/t represents the jump frequency (i.e. number of jumps n over a time t).
Close to the melting point most fcc and bcc metals have D 10
-8
cm
2
/s.
i) If the jump distance is 3 , what is the jump frequency near the melting
point?
ii) What is the relation between this frequency and the vibrational
frequency?
iii) How far has one atom traveled after 1 hour?
iv) What is the root mean square displacement after one hour?
v) What is the root mean square displacement in one dimension after one
hour?
5 Diffusion
5.29

b) For a metal with cubic structure the diffusion coefficient can also be
expressed as
d
N a D
2
0
=
where is a geometric factor, a
0
the lattice constant, is the jump frequency, and
N
d
is the defect concentration. Derive the value for for vacancy diffusion in a
metal with fcc structure.
5 Diffusion
5.30
6. Electrical conductivity
6.1
6. 6. 6. 6. Electrical conductivity Electrical conductivity Electrical conductivity Electrical conductivity

Introduction

In the preceding chapter we have described and discussed diffusion of
particles in solids and particularly of ions and defects in metal oxides. The driving
force for the diffusion has been taken to be the negative value of the particle
gradient or more precisely the negative value of the chemical potential gradient.
When using isotopes as tracers one may study self-diffusion, i.e. diffusion of the
components in the oxide (metal and oxygen ions) in a homogeneous oxide; in this
case the isotopic tracer gradient is the driving force for the diffusion.
In this chapter the transport of electrical charges will be described and
discussed. In metal oxides the electrically charged particles comprise ions and
electrons. The ionic charge carriers comprise the cations, anions, and foreign ions
(e.g. impurity ions, dopant ions and protons) and the electronic charge carriers are
the electrons and electron holes. The concentrations of the charge carriers are
directly related to the defect structure of the oxide and in this chapter we will
derive expressions for the temperature and oxygen pressure dependence of the
electrical conductivity. The discussion will be limited to transport of charges in
chemically homogeneous metal oxides (no chemical potential gradient) but with
an electrical potential gradient as the driving force. In the next chapter transport of
ionic and electronic charge carriers in metal oxides which are simultaneously
exposed to chemical and electrical potential gradients, i.e. electrochemical
potential gradients, will be discussed.
As the mobilities of electrons and electrons holes are normally much higher
than those of ions, most oxides are electronic conductors. One type of charge
carrier often predominates in an oxide under particular conditions of temperature
and oxygen pressure. An electronically conducting oxide is an n-conductor if
transport of electrons predominate and a p-conductor if electron holes prevail.
However, some oxides are or may become ionic conductors or mixed
ionic/electronic conductors depending on the temperature and oxygen pressure
often as a result of appropriate doping with aliovalent foreign ions. Some oxides
may also exhibit proton conductivity in hydrogen- or water vapour-containing
atmospheres; predominant proton conductivity in such oxides is in some cases
observed at reduced temperatures (< 600-700 C).

Transport in an electrical potential gradient

As described in the previous chapter on diffusion in metal oxides the driving
force is given by the negative of the potential gradient. The force exerted on a
charged particle of type i with charge z
i
e is given by

6.2
F = -z
i
e
d
dx
= z
i
eE (6.1)

where is the electrical potential and E = -
d
dx
is the electric field. The flux of
particles of type i is the product of the concentration c
i
, the particle mobility B
i
,
and the force F:

j
i
= c
i
B
i
F = z
i
e c
i
B
i
E (6.2)

The current density i
i
is given by the product of flux and charge:

i
i
= z
i
ej
i
= (z
i
e)
2
B
i
c
i
E (6.3)

While B
i
is the particle mechanical mobility ("beweglichkeit"), the product of B
i

and the charge on each particle, z
i
e, is termed the charge mobility u
i
:

u
i
= z
i
eB
i
(6.4)

Equation 6.3 can then be written

i
i
= z
i
e c
i
u
i
E =
i
E (6.5)

where
i
= z
i
e c
i
u
i
is the electrical conductivity due to the charge carriers of type i.
The electrical conductivity is determined by the product of the concentration c
i
of
the charged particles, the charge z
i
e on the particles and the charge carrier
mobility, u
i
. It should be noted that Eq.6.4 is an expression of Ohm's law. The unit
for the electrical conductivity is Siemens per cm, Scm
-1
(one Siemens is the
reciprocal of one ohm and in older literature the electrical conductivity is
expressed as ohm
-1
cm
-1
). The unit for the charge is coulomb, the concentration of
charge carriers is expressed as the number of charge carriers of type i per cm
3
, and
charge carrier mobility in units of cm
2
/Vs. (Although the SI unit for length is m,
cm is being used in the following as it is still by far the one most commonly used
in the literature).
It may be noted that in the above terminology, F, E, i
i
, z
i
, u
i
and j
i
may each
be positive or negative. u
i
and z
i
always have the same sign, and as long as no
other forces than the the electrical act, i
i
and E
i
always have the same sign, and j
i

and F always have the same sign. B
i
and
i
are always positive, and it is common
also to neglect the sign when specifying charge mobilities u
i
.
6.3

The total electrical conductivity

of a substance is the sum of the partial
conductivities
i
of the different charge carriers:

=
i
(6.6)

The ratio of the partial conductivity
i
to the total conductivity is termed the
transport (or transference) number of species i:

t
i
=
(6.7)

Charge carriers in ionic compounds

The native charge carriers in a binary oxide are the cations, anions,
electrons, and electron holes. The total conductivity is then given by

=
c
+
a
+
n
+
p
(6.8)

where
c,
a
,
n
, and
p
are the cation, anion, electron and electron hole
conductivities, respectively.
Following Eq. 6.7 the individual conductivities may be written in terms of
their transport numbers:
c
= t
c
,
a
= t
a
,
n
= t
n
and
p
= t
p
. Using these
values Eq.6.7 takes the form

= (t
c
+ t
a
+ t
n
+ t
p
) (6.9)

It may be noted that the sum of the transport numbers of all the charge carriers
equals unity:

t
c
+ t
a
+ t
n
+ t
p
= 1 (6.10)

The total electrical conductivity is often given by the sum of the ionic
conductivity,
ion
=
c
+
a
, and the electronic conductivity,
el
=
n
+
p
, and
the total conductivity can then be written
6.4

=
ion
+
el
(6.11)

Often only one type of charge carrier dominates the charge transport, and in
many cases and as an approximation contributions from minority carriers are
neglected. For oxides the mobilities of electrons and electron holes are usually
several orders of magnitude (~10
4
- 10
8
) larger than those of the ions, and even
when the concentration of electron or electrons holes is smaller than that of the
ionic charge carriers (or, more precisely, than that of ionic charge carrier defects)
the oxide may still be a predominantly electronic conductor. The relative
importance of ionic and electronic conductivity will often vary greatly with
temperature and oxygen pressure. This will be illustrated in the following
chapters.

The Nernst-Einstein relation between mobility
and diffusion coefficient

In the previous chapter it was shown that the relation between the random
diffusion coefficient of particles of type i and the particles mechanical mobility is
given by

D
i
= kTB
i
(6.12)

By combining this relation with Eqs. 6.4 and 6.5 one obtains the following
relation between the random diffusion coefficient and the charge carrier mobility
and the electrical conductivity:

D
i
= kTB
i
= u
i

kT
z
i
e
=
i
kT
c
i
z
2
i
e
2

(6.13)

This relation is called the Nernst-Einstein relation. This relation and also the
effect of an applied electric field on migration of charged species in a
homogeneous crystal may also be derived from the following model, in which we
will understand also when and why conduction is termed a linear process.
Consider a one-dimensional system with a series of parallel planes separated
by a distance s (cf. Fick's first law in Chapter 5). It is assumed that the system is
homogeneous and that the volume concentration of the particles in the planes is c
i
.
The particles in neighbouring planes 1 and 2 have equal probability of jumping to
the neighbouring planes. In the absence of any external kinetic force, the number
of particles which jump from plane 1 to plane 2 and from 2 to 1 per unit time is
6.5
equal and opposite and given by
1
2
c
i
s. In a homogeneous system there will be no
net transport of particles.
When there is no applied electric field, the activation energy associated with
the jumps is H
m
. When an electric field E is applied, the jump frequency in the
positive direction will be increased and that in the negative direction decreased in
that the activation energies are changed. In the forward direction the activation
energy is reduced to H
m
-
1
2
z
i
esE and in the reverse direction increased to H
m
+
1
2
z
i
esE. This is illustrated schematically in Fig.6.1.
The net particle flux is given by the difference in number of jumps in the
forward and reverse directions:

j
i
=
1
2
c
i
s {
forw
-
rev
} (6.14)
where
forw
= exp(
S
k
m
) exp(-
H -
z esE
2
kT
m
i
)
and
rev
= exp(
S
k
m
) exp(-
H +
z esE
2
kT
m
i
).
H
m
+c
i
esE/2
c
i
esE/2
-c
i
esE/2
E
N
E
R
G
Y
DISTANCE
H
m
s/2
H
m
-c
i
esE/2
Forward jump
Reverse jump
Rest position

Figure 6-1. Schematic illustration of the effect of an electric field on the
migration of charged species in a homogeneous crystal. E represents the
electric field. H
m
is the activation energy in the absence of an electric
field. In the forward direction the activation energy may be considered to
be lowered by z
i
esE and increased by the same amount in the reverse
direction.

6.6
Equation 6.14 then becomes

j
i
=

1
2

c
i
s{exp (
z
i
esE
2kT
) - exp (-
z
i
esE
2kT
)} (6.15)
where = exp(
S
k
m
) exp(-
H
m
kT
).

When z
i
esE << 2kT, which is valid for normal electrical measurements in
bulk materials (and when Ohm's law is applicable), the difference in exponentials
in Eq. 6.15 may be written z
i
esE/kT (since e
x
- e
-x
= 2x for x<<1). Eq. 6.15 then
takes the form

j
i
=

1
2

s
2
c
i

z
i
eE
kT
(6.16)

The diffusion coefficient for one-dimensional random diffusion is given by D
i
=
1
2
s
2
and j
i
then becomes

j
i
= D
i

c
i
z
i
eE
kT
(6.17)

We have by this shown that net flux density of a hopping charge carrier in an
electrical field in the small-signal (linear) range is proportional to the random
diffusion coefficient. As we have shown before, the flux of particles with a charge
z
i
e may also be expressed in terms of charge carrier mobility or conductivity:

j
i
= z
i
e c
i
B
i
E = c
i
u
i
E =
i
E
z
i
e
(6.18)

and when one combines Eqs. 6.17 and 6.18 one obtains various forms of the
Nernst-Einstein relation (Eq. 6.13):

D
i
= B
i
kT = u
i

kT
z
i
e
=
i
kT
c
i
z
2
i
e
2

(6.13)
or, rearranged,
i
= (z
i
e)
2
c
i
D
i
/kT

6.7
It is emphasised that the relation is derived assuming random diffusion and that
the mobilities and conductivity through this relation connects to the random (or
self) diffusion coefficient D
r
. It is thus meaningful for relating electrical and
diffusional transport of atoms and ions. For electrons and holes this is only
meaningful when they migrate by an activated hopping mechanism.
From the Nernst-Einstein relation it is also seen that the temperature
dependence of the product
i
T is the same as that of D
r
. Thus in evaluating the
activation energy associated with the diffusion coefficient from conductivity
measurements, it is necessary to plot (
i
T) vs 1/T.
It is also important to note that in the derivation it is implicitly assumed that
the ions and electrons move independently of each other, e.g. that there is no
interference between ionic and electronic flows.

Electronic conductivity in oxides

Most metal oxides are electronic conductors at high temperatures. For many
of these oxides the conductivity increases with increasing temperature and as the
conductivity at the same time is much smaller than in metals, this type of
conductivity is termed semiconductivity. The principal reason for the increasing
conductivity is that the number of electronic defects increases with increasing
temperature. A limited number of oxides - especially among transition metal
monoxides - are metallic conductors and for which the conductivity decreases
with increasing temperature. In this case this is attributed to a mobility of
electronic defects decreasing with increasing temperature. Other oxides, e.g. p-
conducting acceptor-doped perovskites to be discussed in a later chapter, also
exhibit metallic-like conductivity in that the conductivity also here decreases with
increasing temperature; however in these cases the decreasing conductivity is
attributed to a decreasing number of electron holes with increasing temperature,
and the conductivity is thus not to be classified as metallic.
The electronic conductivity,
el
, of a semiconducting oxide is given by

el
=
n
+
p
= enu
n
+ epu
p
(6.19)

where
n
and
p
are the electron and electron hole conductivities, n and p
the charge carrier concentrations of electrons and electron holes, respectively, and
u
n
and u
p
are the carrier (or drift) mobilities of electrons and electron holes. As
mentioned above, one type of charge carrier will often dominate; however, in
special cases where an oxide is close to stoichiometric both n- and p- conductivity
may contribute significantly to the electronic conductivity.
In defect-chemical equations the concentrations of electrons and electron
holes are often written in terms of the law of mass action without specifying
where the electronic defects are located. The concentrations of electronic defects
are often interpreted in terms of the band theory of solids and in the following is
6.8
given a brief account of this theory and the relationship between this theory and
the law of mass action.

The band theory

In single atoms the electrons may only possess discrete energies. These
allowed energies are designated by quantum numbers which refer to the electron
shell which the electron occupies (the principal quantum number), the orbital
angular momentum of the electron (azimuthal quantum number), and the direction
of the angular momentum vector (magnetic quantum number). In addition, and
according to the Pauli exclusion principle, each energy state can only be
accommodated by two electrons which have opposite spins.
When individual atoms are brought together in a solid, i.e. when interatomic
spacing decreases and electronic levels overlap, a splitting of the energy levels
begin to occur, and the energy levels may be considered to form energy bands in
the solid. But the total number of levels within a band corresponds exactly to the
total number of atoms present in the solid, and therefore the levels become more
and more finely spaced the larger the number of atoms present. Because of the
Pauli exclusion principle, each band can accommodate twice as many electrons as
there are energy levels. Solids contain 10
22
- 10
23
atoms per cm
3
, and the number
of levels in each band is thus of the same order.
The energy bands may overlap or be separated by energy gaps. At absolute
zero temperature (0 K) the electrons fill up the lowest possible energy levels. The
highest filled band represents the orbitals of valence electrons and is termed the
valence band. It is completely filled at 0 K while the next band, termed the
conduction band is completely empty. In Fig.6.2, where the vertical axis
represents the electron energy and horizontal axis the distance through the solid,
the valence and conduction bands are separated by an energy gap, as is the case in
semiconductors and insulators. In a pure, perfect and ideal solid the electrons may
not possess energies within the energy gap, and this is therefore also often termed
the forbidden energy gap. The size of the energy gap differs for inorganic
compounds, and empirically it is found that the energy gap increases with the heat
of atomisation of the compounds. For binary semiconductors and insulators it has
specifically been shown that the band gap, E
g
(in electron volts) can be expressed
approximately as

E
g
= 2(E
s
- c) (6.20)

where E
s
= E
at
/equivalent
*
and c is a constant approximately equal to 2.7.

*
An equivalent in M
a
O
b
is equal to az
cat
= b|z
an
|
6.9
In materials where the valence band is only partly occupied or where it
overlaps with the conduction band (no forbidden band gap) the electrons can
move freely in the available energy levels and we have metallic conduction.
In an insulator or semiconductor at 0 K there is a band gap and the valence
band is completely filled with electrons while the conduction band is completely
empty. In such cases no electronic conduction takes place when an electric field is
applied.
Conduction band
Valence band
E
l
e
c
t
r
o
n

e
n
e
r
g
y
E
C
E
V
E
g
=E
C
-E
V
Forbidden
energy gap
Distance through crystal

Figure 6-2. Schematic illustration of the energy band diagram for a pure
semiconductor

Intrinsic ionisation
When the temperature is increased, electrons in the valence band are excited
across the forbidden energy gap to the conduction band. This is the intrinsic
ionisation. The electrons in the conduction band and the unoccupied electron sites
in the valence band (electron holes) can move in an electric field. The electron
holes behave as though they were positively charged and move in the opposite
direction of the electrons. The intrinsic ionisation thus produces pairs of electron +
electron hole charge carriers. When the electronic conductivity is due to intrinsic
ionisation only, the semiconductor is called an intrinsic semiconductor. According
to this model electron and electron hole conductivities increase with increasing
concentrations of electrons in the conduction band and electron holes in the
valence band.
In many oxides, and particularly those with large percentage ionic bonding,
periodic fluctuations of the electric potential associated with each ion become too
large (and energy bands too narrow) so that the band model provides an
inadequate description or theory. In this case the electrons or holes may be
considered to be localised at the lattice atoms. Such localised electronic defects
are termed polarons and may from a chemical point of view be considered to
constitute valence defects.
6.10
When classical statistics (the Boltzmann approximation) can be used it may
be shown that the concentration of electrons and electron holes (expressed in
number per cm
3
) is given by

n = N
C
exp(-
E
C
-E
F
kT
) (6.21)

p = N
V
exp (-
E
F
-E
V
kT
) (6.22)

N
C
and N
V
represent the number of available states per unit volume - density of
states - in the conduction and valence bands. E
C
is the energy of the lowest level
of the conduction band and E
V
the highest level in the valence band. The
parameter E
F
is termed the Fermi level and we will return to this later on.
When the electrons occupy a narrow band of energies close to E
C
, a
parabolic relation between N
C
and E
C
may be assumed and it may be shown that
N
C
is then given by

N
C
= (
8 m kT
h

e
*
2
)
3/2
(6.23)

where m
*
e
is the effective mass of the electron. When the holes occupy a narrow
region of energies close to E
V
, N
V
is correspondingly given by

N
V
= (
8 m kT
h

h
*
2
)
3/2
(6.24)

where m
*
h
is the effective mass of the electron hole. The effective masses of
electrons and electron holes are seldom accurately known and as an
approximation the free electron mass is then used in these relations.
The defect reaction and corresponding equilibrium between electrons in the
conduction band and electron holes in the valence band can in terms of the law of
mass action be expressed as

0 = e' + h
.
(6.25)

n
.
p = K
i
(6.26)
6.11

where K
i
is the equilibrium constant for the intrinsic ionisation. Combination with
Eqs. 6.21 and 6.22 gives

K
i
= n
.
p = N
C
N
V
exp (-
E
g
kT
) (6.27)

where E
g
is the band gap, E
g
= E
C
- E
V
. E
g
may thus be considered the energy of
the intrinsic ionisation. It is emphasised that Eq. 6.27 presupposes that classical
statistics apply. It may also be noted that the Fermi level is eliminated in this
expression of the law of mass action. It may be noted again that in Eq. 6.27 the
concentrations of electrons and electron holes are expressed in terms of number
per volume unit (e.g. per cm
3
).

Intrinsic electronic semiconductor
In an intrinsic semiconductor the concentrations of electrons and electron
holes are equal, and thus

n = p = K
i
1/2
= (N
C
N
V
)
1/2
exp (-
E
g
2kT
) (6.28)

and the electronic conductivity (Eq. 6.19) then becomes

el
=
n
+
p
= en
n
+ ep
p
= e (N
C
N
V
)
1/2
(u
n
+ u
p
) exp(-
E
g
2kT
) (6.29)

It may be noted that N
C
and N
V
are temperature dependent and that also u
n
and u
p

may have various dependencies on temperature. If the latter are not exponential
(as in diffusional hopping conduction processes) the exponential term of the
energy gap tend to dominate the temperature dependence and as an approximation
Eq. 6.29 is then often written

el
= const. exp(-
E
g
2kT
) (6.30)

From these relations it is evident that the intrinsic electronic conductivity
increases with decreasing energy gap.

6.12
Effects of charged impurities or defects. Extrinsic
semiconductors

As stated before, the valence and conduction bands may be considered to
constitute the highest filled orbital and the lowest empty orbital, respectively. The
valence band may in an oxide typically represent the 2p states of the O
2-
ions, and
removing an electron from here (creating a hole) may be regarded as the creation
of an O
-
ion, which is intuitively possible. The conduction band typically
represents the reduced state of a metal constituent ion. For instance, in oxides with
Ti
4+
such as TiO
2
or SrTiO
3
the conduction band would typically represent the
4s state of the Ti
3+
ion. The size of the band gap can be imagined to represent the
difficulty of simultaneously oxidising O
2-
to

O
-
and reducing Ti
4+
to Ti
3+
. In the
chosen example we expect a moderate band gap. Similar oxides with Mn
4+
would
be more easily reducible and have smaller band gaps, while oxides with Si
4+
or
Zr
4+
are less easily reducible and have larger band gaps.
Let us now consider the addition of small amounts of imperfections or
impurities. These may be considered to contribute additional localised energy
levels in the crystal. It is commonly assumed that these always fall within the
forbidden gap. As we shall see later on this leads to situations where a certain
temperature is needed to excite these defects to become effectively charged. This
is not in accordance with experimental evidence and is not intuitive for
imperfections with very stable aliovalent valence states. Thus, we will first
consider cases where imperfections introduce levels outside the forbidden gap.
If these levels fall below the valence band edge E
V
they will always be
occupied (even at 0 K) and be in a reduced state. We refer to it as an acceptor
imperfection. For instance, Al
4+
substituting Ti
4+
in the abovementioned oxides
may always be considered to be reduced to Al
3+
and as such be a charged defect
even at 0 K, for instance taking the electron from the valence band. We may draw
this conclusion from the higher stability of O
-
as compared to Al
4+
.
Similarly, an imperfection introducing a level above the conduction band
edge E
C
may always be considered to lose its electron and be oxidised even at 0
K. We refer to it as a donor. For instance, La
2+
substituting Sr
2+
and possibly
Ta
4+
substituting Ti
4+
would be oxidised to La
3+
and Ta
5+
by giving off the
electron to the easier formed Ti
3+
of the conduction band.
The above examples lead to compensating electronic defects even at 0 K,
but real cases may comprise simultaneous formation of point defects that may
annihilate the electronic defects, (depending on oxygen activity). Thus, while Ca
substituting La in LaCrO
3
is compensated by electron holes at high oxygen partial
pressures, Al or Ta substituting Ti in SrTiO
3
will be compensated by oxygen
vacancies or metal vacancies, respectively, under the same conditions. Of course,
this brings up a question of whether native point defects such as vacancies at
different stages of ionisation themselves introduce levels within or outside the
band gap. If aliovalent dopants charged at 0 K are to be compensated by point
6.13
defects at 0 K, these point defects must themselves be charged at 0 K and thus
have levels outside the band gap.
In the following we consider imperfections introducing levels inside the forbidden
energy gap as illustrated in Fig.6.3. When the imperfection introduces an energy
level which is located below the lower edge of the conduction band (E
C
),
electrons can be thermally excited to the conduction band. As such the
imperfection donates an electron, it is called a donor and the corresponding energy
level a donor level. Under conditions where such imperfections dominate the
defect structure the oxide becomes an n-conductor. Correspondingly,
imperfections with energy levels just above the upper edge of the valence band
(E
V
) are termed acceptors; electrons in the valence band may be excited to the
energy level of the imperfection (acceptor level) and the oxide may become a p-
conductor.
Distance through crystal
Acceptor level
E
V
E
C
Donor level
E
l
e
c
t
r
o
n

e
n
e
r
g
y
E
D
E
A
E
d
E
a
Valence band
Conduction band

Figure 6-3. Schematic illustration of additionally localised energy levels due to
donors and acceptors in the forbidden energy gap in the energy band diagram of
a semiconductor.

In SrTiO
3
it is for instance found that Fe
4+
substituting Ti
4+
is an acceptor
inside the gap; Fe
3+
(the result after acceptance of an electron) is less favourable
than Fe
4+
(when the electron is taken from O
2-
) and at a higher energy than the
valence band of the O
2-
electrons, but Fe
3+
is clearly more favourable than Ti
3+

and thus way below the conductance band. Similarly, oxygen vacancies are
considered to be donors in the gap, and this can be viewed as a trapping of
electrons at Ti
4+
(as Ti
3+
) near the vacancy.
While donors and acceptors inside the forbidden gap have positive energies
of ionisation, the donors and acceptors outside the gap may be considered simply
to have negative energies of ionisation.

6.14
Effects of donors
Let us again describe these processes in terms of the law of mass action.
Thus the ionisation of a donor D
x
may be written

D
x
= D
.
+ e' (6.31)

and the corresponding equilibrium by

[D
.
] n
[D
x
]
= K
D
(6.32)

If the total number of donors is N
D
, then

N
D
= [D
.
] + [D
x
] (6.33)

If the energy of the donor state is E
D
, and the donor state is mainly empty
([D
x
]<<N
D
) then the concentration of electrons in the donor state, [D
x
], may
following Boltzmann statistics (be expressed by

[D
x
] = N
D
exp (-
(E
D
-E
F
)
kT
) (6.34)

Since we assume
[D
x
]
N
D
<<1, and that Boltzmann statistics also apply to conduction
band electrons and by further combining Eqs. 6.21 and 6.32-6.34, we get an
expression for K
D
:

K
D
= N
C
exp (-
E
C
-E
D
kT
) = N
C
exp (-
E
d
kT
) (6.35)

where E
d
= E
C
-E
D
represents the ionisation energy of the donor (cf. Fig. 6.3). E
d

may in Eq. 6.35 be considered to be the enthalpy of the ionisation of the donor.
When no other imperfections are present, the situation described here can be
approximated by

n = [D
.
] N
D
(6.36)
6.15

and if the donors are present in an invariable amount as a fully soluble or frozen-
in impurity, then n = [D
] N
D
= constant. From Eq. 6.32, we then get the
minority concentration of neutral unionised donors: [D
x
] = N
2
D
N
-1
C

exp(+E
d
/kT).
If, on the other hand the donor level is mainly unionised, then [D
]<<[D
x
]
N
D
and Boltzmann statistics give

[D
] = N
D
exp(-
(E
F
-E
D
)
kT
) (6.37)

and then, by combination with Eqs. 6.21, 6.32, and 6.33 we again obtain K
D
= N
C

exp(-E
d/
kT) as in Eq. 6.35.
If the electroneutrality condition in the oxide is still given by

n = [D
.
] (6.38)

(but now [D
]<<N
D)
, then the concentration of electrons is, from insertion in Eq.
6.32:

n = [D
] = (K
D
N
D
)
1/2
= (N
C
N
D
)
1/2
exp (-
E
d
2kT
) (6.39)

The reader may find it useful to note the similarity between the form of this
expression and Eq. 6.28. However, we also remind ourselves that N
D
may be a
constant concentration of impurities or it may be a constant or varying level of
other point defects such as anion vacancies or cation interstitials.
In the above we have seen that the same equilibrium constant K
D
for Eq.
6.31 applies whether the donor level is approximately empty (fully ionised) or
approximately fully occupied (practically unionised, or, in this case, neutral). But
our treatment does not hold for intermediate situations, i.e. what we may call
partly filled donor levels. In the latter case Boltzmann statistics do not represent a
sufficiently good approximation, and instead Fermi-Dirac statistics must be used.
Let us recapitulate briefly the treatment of the ionisation of a constant
concentration of a donor in the forbidden gap of an otherwise pure, stoichiometric,
ideal semiconductor: At low temperatures, the concentration of electrons is given
by a minor degree of ionisation of the donors, as given by Eq. 6.39. If we neglect
the temperature dependence of N
C
, the situation can be illustrated as in the right-
hand part of Fig. 6.4; the concentration of electrons increases with an apparent
6.16
enthalpy of E
d
/2. At a sufficiently high temperature (middle part of Fig. 6.4) all
donors are ionised, and the concentration of electrons becomes constant (Eq.
6.36). At even higher temperatures intrinsic semiconduction may predominate,
i.e., n = p, and in principle the temperature dependence becomes as illustrated in
the left hand part of Fig. 6.4, with an apparent enthalpy close to E
g
/2 (cf. Eq.
6.25). Thus at low temperatures this oxide is an n-conductor due to the ionisation
of the donors and at high temperatures intrinsic ionisation predominates. The
behaviour over the entire temperature range could in principle be solved from the
full electroneutrality equation

n = [D
]+ p (6.40)

in combination with the constancy of the donor concentration and the expressions
for the equilibrium constants K
i
(Eq. 6.27) and K
D
(Eq. 6.35). However, as noted
above, the expression for K
D
would not be valid between the intermediate and
low temperature (right hand side) domains in Fig. 6.4.

n=N
D
E
d
/2
E
g
/2
1/T K
L
o
g

n

Figure 6-4. Schematic illustration of the logarithm of the concentration of defect
electrons as a function of the reciprocal absolute temperature for a semiconductor
with donors.

Effects of acceptors
A corresponding treatment may be made for ionisation of acceptors, which
in terms of a defect reaction may be written

A
x
= A' + h
.
(6.41)

6.17
The equilibrium constant for the defect reaction is given by

[A'] p
[A
x
]
= K
A
(6.42)

Following a similar treatment as for the donor, K
A
may under the assumption that
Boltzmann statistics apply to the valence band (small concentration of holes) and
the acceptor (nearly full or nearly empty acceptor levels) be expressed by

K
A
= N
V
exp (-
E
A
-E
V
kT
) = N
V
exp (-
E
a
kT
) (6.43)

where E
A
is the energy level of the acceptor and E
a
= E
A
-E
V
is called the
ionisation energy of the acceptor (cf. Fig. 6.3).
For an acceptor doped oxide the temperature dependence of the
concentration of holes will be qualitatively analogous to that of electrons in the
donor doped case (cf. Fig. 6.4) with the same constrictions.

The Fermi level and chemical potential of electrons

In this chapter we have so far treated the defect equilibria for
semiconductivity in terms of the band model. In this model one makes use of the
parameter termed the Fermi level, E
F
. As the defect equilibria are otherwise
described by equilibrium thermodynamics, it is of interest to correlate the band
model with the thermodynamic approach.
A chemical equilibrium implies that the chemical potential of a species is
the same in all phases. As regards electrons in a system, this also means that their
chemical potentials (or electrochemical potentials if the inner potential can not be
neglected) must be equal, although they may have different energies. Thus the
chemical potential of the electrons in general,
e
, must be equal to the chemical
potential of valence electrons, conduction electrons, etc.,

e
= (cond. electrons) = (valence electrons) (6.44)

The chemical potential of electrons, for instance, the conduction electrons, may be
written in terms of the chemical potential in a standard state, (cond. electrons)
and a term for the entropy of mixing:

6.18
e
= (cond. electrons) = (cond. electrons) + kT ln
n
N
C
(6.45)

From the relation between the concentration of conduction electrons, n, the
conduction band energy level E
C
and the Fermi level E
F
(Eq. 6.18) we may write

E
F
= E
C
+ kT ln
n
N
C
(6.46)

By comparing Eqs. 6.45 and 6.46 it is seen that if E
C
is considered to be the
chemical potential in the standard state for conduction electrons, the Fermi level
represents the chemical potential of the conduction electrons, and thus of all
electrons in a substance.

Charge mobilities of electrons and electron holes

In the preceding chapters we have looked at temperature dependencies of
concentrations of electronic defects and point defects, and we have looked at the
conductivity and mobility of thermally activated diffusing species. In the
following we consider the charge carrier mobilities of electrons and holes in some
more detail. For instance for an intrinsic electronic semiconductor (where n=p) we
can from Eq. 6.29 in combination with Eqs. 6.23 and 6.24 write an expression for
el
:

el
= {2e(
2 k
h

2
)
3/2
(m
*
e

.
m
*
h
)
3/4
T
3/2
exp (-
E
g
2kT
)}(u
n
+ u
p
) (6.47)

The effective mass of electrons and electron holes can be interpreted by a
quantum mechanical treatment of electronic motion of electrons and electron
holes in solids. The effective mass differs from the real mass of electrons due to
the interaction of electrons with the periodic lattice of the atoms. Only for a
completely free electron is the mass equal to the real mass, m
e
=m
*
e
. As mentioned
above, the values of the effective masses are not accurately known, and the value
of m
e
is often used as an approximation.
In order to obtain an accurate description of the temperature dependence of
the electronic conductivity it is necessary to consider the temperature
dependencies of the charge carrier mobilities.

6.19
Non-polar solids itinerant electron model
The temperature dependence of the charge carrier mobility is dependent on
the electronic structure of the solid. For a pure non-polar semiconductor - as in an
ideal and pure covalent semiconductor - the electrons in the conduction band and
the electron holes in the valence band can be considered as quasi-free (itinerant)
particles. Then the mobilities of electrons and electron holes, u
n
and u
p
, are
determined by the thermal vibrations of the lattice in that the lattice vibrations
result in electron and electron hole scattering (lattice scattering). Under these
conditions the charge carrier mobilities of electrons and electron holes are both
proportional to T
-3/2
, e.g.

u
n,latt
= const. T
-3/2
(6.48)

In this case the temperature dependence of
el
(Eq. 6.47) becomes

el
= const. T
3/2.
T
-3/2
exp (-
E
g
2kT
) = const. exp (-
E
g
2kT
) (6.49)

If, on the other hand, the scattering is mainly due to irregularities caused by
impurities or other imperfections, the charge carrier mobility is proportional to
T
3/2
, e.g.

u
n,imp
= const. T
3/2
(6.50)

If both mechanisms are operative, the mobility is given by

u =
1
1
u
latt
+
1
u
imp
(6.51)

and from the temperature dependencies given above it is evident that impurity
scattering dominates at low temperature while lattice scattering takes over at
higher temperature.

Polar (ionic) oxides
When electrons and electron holes move through polar compounds, such as
ionic oxides, they polarise the neighbouring lattice and thereby cause a local
deformation of the structure. Such an electron or electron hole with the local
deformation is termed a polaron. The polaron is considered as a fictitious single
particle.
6.20
When the interaction between the electron or electron hole and the lattice is
relatively weak, the polaron is referred to as a large polaron. Large polarons
behave much like free carriers except for an increased mass caused by the fact that
polarons carry their associate deformations. Large polarons still move in bands,
and the expressions for the effective density of states in the valence and
conduction bands are valid. The temperature dependence of the mobilities of large
polarons at high temperatures
*
is given by

u
large pol.
= const. T
-1/2
(6.52)

For an intrinsic semiconducting oxide where the electronic conductivity
mechanism can be described in terms of a large polaron mechanism, the
temperature dependence of
el
can be written by combination of Eqs. 6.47 and
6.52

el
= const. T
.
exp (-
E
g
2kT
) (6.53)

Thus in this case the value of the band gap can be deduced by plotting log(
el
/T)
vs 1/T.
The large polaron mechanism has been suggested for highly ionic non-
transition metal oxides. Such oxides have large energy gaps (E
g
6 eV), where
the band structure is characterised by large band widths. Due to the large band gap
one expects electronic semiconductivity to be low in these oxides, and to
predominate possibly only at high temperatures.
For other oxides it has been suggested that the interactions between the
electronic defects and the surrounding lattice can be relatively strong and more
localised. If the dimension of the polaron is smaller than the lattice parameter, it is
called a small polaron or localised polaron, and the corresponding electronic
conduction mechanism is called a small polaron mechanism.
The transport of small polarons in an ionic solid may take place by two
different mechanisms. At low temperatures small polarons may tunnel between
localised sites in what is referred to as a narrow band. The temperature
dependence of the mobility is determined by lattice scattering and the polaron
mobility decreases with increasing temperature in a manner analogous to a broad
band semiconductor.
However, at high temperatures (for oxides above roughly 500 C) the band
theory provides an inadequate description of the electronic conduction
mechanism. The energy levels of electrons and electron holes do not form bands,
but are localised on specific atoms of the crystal structure (valence defects). It is

*
"High temperatures" are temperatures above the optical Debye temperature, . For oxides
~(h)/2k, where h is the Planck constant, k the Boltzmann constant and the longitudinal
optical frequency which for an oxide is ~10
14
s
-1
.
6.21
assumed that an electron or electron hole is self-trapped at a given lattice site, and
that the electron (or electron hole) can only move to an adjacent site by an
activated hopping process similar to that of ionic conduction. Consequently it has
been suggested that the mobility of a small polaron can be described by a classical
diffusion theory as described in a preceding chapter and that the Nernst -Einstein
can be used to relate the activation energy of hopping, E
u
, with the temperature
dependence of the mobility, u, of an electron or electron hole:

u =
e
kT
D = const. T
-1
exp ( -
E
u
kT
) (6.54)

where E
u
is the activation energy for the jump.
At high temperatures, the exponential temperature dependence of small
polaron mobilities can thus in principle be used to distinguish it from the other
mechanisms.
The different mechanisms can also be roughly classified according to the
magnitude of the mobilities; the lattice and impurity scattering mobilities of
metals and non-polar solids are higher than large-polaron mobilities which in turn
are larger than small-polaron mobilities. Large polaron mobilities are generally of
the order of 1-10 cm
2
/V
-1
s
-1
, and it can be shown that a lower limit is
approximately 0.5 cm
2
V
-1
s
-1
. Small polaron mobilities generally have values in
the range 10
-4
-10
-2
cm
2
V
-1
s
-1
. For small polarons in the regime of activated
hopping the mobility increases with increasing temperature and the upper limit is
reported to be approximately 0.1 cm
2
V
-1
s
-1
.

Nonstoichiometric semiconductors

Corresponding expressions for
el
for nonstoichiometric electronic
semiconductors readily follows by considering the temperature and oxygen
pressure dependence of the concentration of the electronic defects.
For nonstoichiometric oxides the concentration of electronic defects is
determined by the deviation from stoichiometry, the presence of native charged
point defects, aliovalent impurities and/or dopants. The concentration of
electronic defects can be evaluated from proper defect structure models and
equilibria. Various defect structure situations have been described in previous
chapters and at this stage only one example - dealing with oxygen deficient oxides
with doubly charged oxygen vacancies as the prevalent point defects - will be
described to illustrate the electrical conductivity in nonstoichiometric oxides.

Oxygen deficient oxides
Let us recapitulate the equations for formation of doubly charged oxygen
vacancies. As described in Chapter 3 the defect equation may be written
6.22

O
O
= V
2
.
O
+ 2e' +
1
2
O
2
(6.55)

The corresponding defect equilibrium is given by

[V
2
.
O
] n
2
= K
V
2
.
O
p
-1/2
O
2
(6.56)

If we deal with a high-purity oxide where the concentration of impurities can be
ignored compared to the concentration of oxygen vacancies and electrons, the
electroneutrality condition becomes

n = 2[V
2
.
O
] (6.57)

By combining Eqs. 6.57 and 6.58 the concentration of electrons is given by

n = 2[V
2
.
O
] = (2K
V
2
.
O
)
1/3
p
-1/6
O
2
(6.58)

The total electrical conductivity is given by the sum of the conductivity of the
electrons and of the oxygen vacancies:

t
= 2 e [V
2
.
O
] u
V
2.
O
+ e n u
n
(6.59)

2 e [V
2
.
O
] u
V
2
.
O
represents the ionic conductivity due to the oxygen vacancies and
where u
V
2
.
O
is the mobility of the oxygen vacancies. However, if the electrons and
oxygen vacancies are the prevalent charge carriers, the contribution due to oxygen
vacancies can be ignored due to the much higher mobility of electrons than
oxygen vacancies, and the oxide is an n-conductor where the conductivity can
then be written

t
=
n
= e n u
n
= e u
n
(2K
V
2
.
O
)
1/3
p
-1/6
O
2
(6.60)

6.23
As described in previous chapters the equilibrium constant for the formation of
doubly charged oxygen vacancies and 2 electrons is given by

K
V
2
.
O
= exp (
S
k
V
O
2.
) exp(-
H
kT
V
O
2.
) (6.61)

When one combines Eqs. 6.61 and 6.62 the n-conductivity may be written:

t
=
n
= e u
n
exp (
S
3k
V
O
2.
) exp (-
H
3kT
V
O
2.
) p
-1/6
O
2
(6.62)

Let us further assume that the electrons are small polarons and thus that the
mobility of the electrons are given by Eq. 6.54. The conductivity can then be
expressed by

n
= const.
1
T
exp (-
H / 3+ E
kT
V
O
2.
u
) p
-1/6
O
2
(6.63)

Thus following this equation the n-conductivity is proportional to p
-1/6
O
2
, and if this
defect structure situation prevails over a temperature range from T
1
to T
5
, one will
obtain a set of isotherms of the n-conductivity as shown in Fig.6.5.
T
5
Log oxygen pressure
T
4
T
3
T
2
T
1
L
o
g

n
-
c
o
n
d
u
c
t
i
v
i
t
y
p
O2
-1/6

Figure 6-5. Schematic presentation of different isotherms of the n-conductivity at
temperatures from T
1
to T
5
for an oxygen deficient oxide where the predominant
defects are doubly charged oxygen vacancies and electrons.

6.24
Furthermore, if it can be assumed that mobility of the charge carriers (defect
electrons) is independent of the defect concentration, then a plot of the values of
log
10
(T) at a constant oxygen pressure yields a straight-line relationship as
illustrated in Fig. 6.6. The slope of the line is given by -
1
2.303k

H
3

V
O
2.
+E
u
,
where the factor 2.303 is the conversion factor in changing from ln
e
to log
10
. The
activation energy is given by the term

E
=
H
3

V
O
2.
+ E
u
. (6.64)

Figure 6-6. Schematic illustration of a plot of log
10
(
n
T) vs. the reciprocal
absolute temperature at constant oxygen pressure (cf. Fig.6.5). The slope of the
line is given by -
1
2.303k

H
3

V
O
2.
+E
u
where
H
3

V
O
2.
+E
u
is the activation
energy for the n-conductivity.

In general the temperature dependence of the charge carrier mobility of the
electrons is much smaller than the enthalpy term associated with the formation of
doubly charged oxygen vacancies.
The mobility of electronic charge carriers may be determined by measuring
the electrical conductivity and combine these measurements with independent
measurements of the concentration of the electronic charge carriers. The
concentration of the charge carriers may be estimated from measurements of the
Seebeck coefficient or by measurements of the nonstoichiometry combined with
the proper description of the defect structure (cf. Ch. 7).
For mixed conductors that exhibit both ionic and electronic conductivities it
is necessary to delineate the ionic and electronic contributions. A commonly used
technique for this is the emf method originally derived by Wagner. This will be
6.25
described in the next chapter (Ch. 7) dealing with electrochemical transport in
metal oxides.

Ionic conductivity

Ionic conductivity follows the NernstEinstein relationship, for hopping
species, derived early in this chapter:

i
= z
i
e c
i
u
i
= (z
i
e)
2
c
i
D
i
/kT (6.65)

It may be noted that the species considered may be a defect (e.g. oxygen
vacancies) or a constituent (oxide ions). For defects the mobility and diffusivity
are large and roughly constant, while the concentration is small and variable. For
the constituent, the mobility and diffusivity are small and variable (with defect
concentration) while the concentration is large and roughly constant. The
conductivity (here oxide ion conductivity by the vacancy mechanism) is the same.
In many cases of utilizing ionic conduction, the concentration c
i
of ionic
defects is constant, given by a dopant. For instance, ionic conductivity in yttria-
doped zirconia is determined by the concentration of oxide ion vacancies, in turn
given as charge compensating the yttrium acceptors, 2[v
O
..
] = [Y
Zr
/
]. If the
concentration of acceptors is given in mole-fraction, then it is necessary to
multiply the resulting mole fraction of vacancies by the formula density or molar
density of the compound in order to obtain the volume density required for
insertion in Eq. 6.65:

vO..
= 2e [Y
Zr
/
]/2 * c
ZrO2
u
vO..
= 2F [Y
Zr
/
]/2 * C
ZrO2
u
vO..
(6.66)

where c
ZrO2
and C
ZrO2
are, respectively, the molecular and molar densities (number
of formula units or moles per unit volume) of the oxide.
There are of course also cases also of practical interest where the
concentration of ionic defects vary, e.g. with temperature in intrinsically
disordered compounds, and with temperature and non-stoichiometry in non-
stoichiometric compounds. In proton conducting oxides the proton conductivity
varies with proton concentration, typically a function of water vapour partial
pressure.
We leave further learning about ionic conduction to exercises and Chapter
7 (electrochemical transport), and here only briefly mention a couple of aspects of
ionic transport that relates it in more detail to diffusion.

6.26
Correlation effects: tracer diffusion and ionic conduction

In the discussions of diffusion mechanisms in Chapter 5 it was pointed out
that successive jumps of tracers atoms in a solid may for some mechanisms not be
completely random, but are to some extent correlated. This is, for instance, the
case for the vacancy and interstitialcy mechanisms. For a correlated diffusion of a
tracer atom in a cubic crystal the tracer diffusion coefficient, D
t
, is related to the
random diffusion coefficient for the atoms, D
r
, through the correlation coefficient
f:

D
t
= f D
r
(6.67)

The value of f is governed by the crystal structure and the diffusion mechanism.

Ionic conductivity method
Values of the correlation coefficient may be determined by comparing the
measured values of the ionic conductivity and the tracer diffusion coefficient.
Thus the use of the Nernst-Einstein relation gives the following expression for the
correlation coefficient:

f =
D
t
D
r
=
D
t
i

c
i
(z
i
e)
2
kT
(6.68)

This equation is applicable to any diffusion process for which the atom jump
distance is equal to the displacement of the effective charge, e.g. for vacancy and
interstitial diffusion.
However, in interstitialcy diffusion the charge displacement is larger than
the atom jump distance, and a displacement factor S must be included in the
Nernst-Einstein relation. In collinear interstitialcy diffusion (Fig. 5.9) the effective
charge is, for instance, moved a distance twice that of the tracer atom and D
t
/D
r
is
given by

D
t
D
r
=
D
t
S

c
i
(z
i
e)
2
i
kT
=
f
S
(collinear) (6.69)

where S = 2. For a collinear jump in an fcc structure the displacement factor is
4/3.
Studies on alkali and silver halides have provided illustrative, and by now
classical examples of the applicability of the ionic conductivity method for
determining the correlation factor and detailed aspects of the jumps in diffusion
processes. NaCl, for instance, is essentially a pure cationic conductor. Measured
6.27
ratios of D
t
/D
r
are in good agreement with the assumption that f = 0.78, i.e. that
the Na-ions diffuse by a vacancy mechanism.
However, such a simple relationship was not found for AgBr. AgBr is also a
cationic conductor and comparative values of D
t
(diffusion of Ag in AgBr) and of
values of D
r
evaluated from conductivity measurements are shown in Fig. 6.7.
From studies of the effect of Cd-dopants on the ionic conductivity it could
be concluded that cationic Frenkel defects predominate in AgBr. Thus the
diffusion was therefore expected to involve both vacancy diffusion and transport
of interstitial ions. The experimentally measured ratios of D
t
/D
r
varied from 0.46
at 150 C to 0.67 at 350C. For vacancy diffusion a constant ratio of 0.78 (=f)
would have been expected, and the diffusion mechanism could thus be ruled out.
For interstitial diffusion f=1, and this mechanism could also be excluded.

Figure 6-7. Values of D
t
and of D
r
evaluated from conductivity measurements for
diffusion of Ag in AgBr. Results after Friauf(1957,1962).

For interstitialcy diffusion of Ag in AgBr the value of f equals 2/3 for a collinear
jump and 0.97 for a non-collinear jump. Following Eq.6.64 one would thus expect
that D
t
/D
r
would range from 0.33 for a collinear jumps to 0.728 for non-collinear
jumps. On this basis Friauf (1957, 1962) concluded that the interstitialcy diffusion
is the important mechanism in AgBr and that collinear jumps are most important
at low temperatures while non-collinear jumps become increasingly important the
higher the temperature.

6.28
Simultaneous diffusion and electric field
The ionic conductivity and D
t
may in principle be studied in a single
experiment, as described by Manning (1962) and others. If a thin layer of the
isotopes is sandwiched between two crystals and the diffusion anneal is performed
while applying the electric field, the tracer distribution profile is displaced a
distance x = u
i
Et relative to the profile in the absence of the applied field (Eq.
5.31). The resultant tracer distribution is given by

c =
c
2( D t)

o
t
1/2
exp (-
(x - x)
4D t

2
t
) (6.70)

The maximum in the concentration profile is - as illustrated in Fig. 6.8 - displaced
a distance x, and u
i
and D
t
may be determined from the same experiment. If the
crystal is a mixed ionic/electronic conductor, the value of the ionic transport
number under the experimental conditions must be known.
X
X = u
i
Et
C
O
N
C
E
N
T
R
A
T
I
O
N

Figure 6-8. Schematic illustration of the concentration profile of a radioactive
tracer when an electric field is applied during the diffusion anneal. The tracers
are originally located at 0, but the concentration profile is displaced a distance
X = u
i
Et.

Literature

Friauf, R.T. (1957) Phys. Rev. 105, 843; (1962) J. Appl. Phys. 33 suppl.,
494.
Manning, J.R. (1962) J. Appl. Phys 33, 2145; Phys. Rev. 125, 103.

6.29
Exercises

1. Cobalt oxide: The electronic conductivity of Co
1-y
O at 1350C and p
O2
=
0.1 atm is 25 S/cm. Thermogravimetric measurements show that y = 0.008 under
the same conditions. It is assumed that singly charged cobalt vacancies are the
dominating point defects. Identify the charge carriers responsible for the
conductivity and calculate their charge mobility. (Assume that the density of CoO
at 1350C equals that at room temperature, 6.4 g/cm
3
. Atomic weights M
Co
=
58.93, M
O
= 16.00.)

2. Nickel oxide: Assume that doubly charged nickel vacancies and electron
holes are the dominating defects in Ni
1-y
O under oxidising conditions. At 1245C
and p
O2
= 1 atm we know the following for the compound:

The self diffusion coefficient for nickel: D
Ni
= 9*10
-11
cm
2
/s
Electrical conductivity: = 1.4 S/cm
(Data from M.L. Volpe and J. Reddy, J. Chem. Phys., 53 (1970) 1117.)

Nickel vacancy concentration, in site or mole fraction: [v
Ni
] = 2.5*10
-4

(Data from W.C. Tripp and N.M. Tallan, J. Am. Ceram. Soc., 53 (1970)
531.)

i) Calculate the concentration of electron holes under the given conditions,
given as site fraction and as volume concentration (e.g. number/cm
3
). (Atomic
weights M
Ni
= 58.71, M
O
= 16.00, density of NiO = 6.67 g/cm
3
.)

ii) Calculate the charge mobility of the electron holes.

iii) Calculate the diffusion coefficient of nickel vacancies.

iv) Calculate the charge mobility of the nickel vacancies and the ionic
conductivity under the conditions referred to above.

3. Ca-stabilised ZrO
2
(CSZ)

We shall here consider a densely sintered ZrO
2
doped with 15 mol% CaO
(Zr
0.85
Ca
0.15
O
1.85
).

6.30
i) Assume that the oxide contains doubly charged oxygen vacancies compensating
the Ca dopant. What are the site-fractions of dopants and of oxygen vacancies?

ii) Derive equations showing how the minority concentrations of defect
electrons and electron holes vary with p
O2
in this oxide under the given conditions.

iii) The conductivity of this oxide is independent of p
O2
from oxidising to
very reducing conditions. What can we deduce from this?

iv) Simpson and Carter (J. Am. Ceram. Soc. 49 (1966) 139) measured the
self diffusion coefficient for oxygen in Zr
0.85
Ca
0.15
O
1.85
and found it to be D
O
=
2.0*10
-7
cm
2
/s at 1100C. Calculate the electrical conductivity based on this.

v) Find also the diffusion coeffeicient and charge mobility for the oxygen
vacancies.

7. Electrochemical transport
7.1
7. 7. 7. 7. Electrochemical transport Electrochemical transport Electrochemical transport Electrochemical transport

Electrochemical potential

In the chapter on diffusion we learned that random diffusion is driven by thermal
energy and that it in crystalline solids requires defects. Moreover, we learned that a net flux
of a species can result from a gradient in its chemical potential. In the chapter on electrical
conductivity we furthermore saw that a net flux of charged particles would result in a
gradient of electrical potential.
In ionic media (materials, liquids, and solutions) the chemical and electrical potentials
act simultaneously, and it is often convenient to combine them into an electrochemical
potential. For the species i we have

e
z
+ =
i
i i
(7.1)

This is true whether the species is real or a defect, but in general we shall onwards deal
mainly with real species and real charges z
i
.
The electrochemical potential gradient is accordingly, in the one-dimensional case,

dx
d
e
z
+
dx
d
=
dx
d
i
i i

(7.2)

The combination of chemical and electrical potentials and potential gradients forms the basis
for the treatment of all mass transport processes involving charged species (ions) in ionic
solids, and is the theme of this chapter. The theory is termed Wagner-type after Carl Wagner,
who was the first to derive it, originally to describe oxidation of metals.

Flux equations Wagner theory
General derivation of Wagner-type flux expressions

We have seen in a preceding paragraph that a force, expressed as gradient in a potential
P
i
acting on a species i, gives rise to a flux density of that species which is proportional to its
self-diffusion coefficient D
i
. By assuming that the potential acting is the electrochemical
potential, we obtain

7.2
]
dx
d
e
z
+
dx
d
[
kT
c D
-
=
dx
d
kT
c D
-
dx
P
d
kT
c D
-
= j
i
i i i i i i i i i
i

= (7.3)

Via the Nernst-Einstein equation we may substitute conductivitiy for the random diffusivity
and obtain the following alternative expression for the flux density:

]
dx
d
e
z
+
dx
d
[
e z
-
= j
i
i
i
i
i

2
) (
(7.4)

If the species is charged, the flux density for i gives rise to a partial current density i
i
:

]
dx
d
e
z
+
dx
d
[
e z
- = j e
z
=
i i
i
i
i
i
i i

(7.5)

The net current density in the sample is obtained by summing the partial current densities
over all the species k:

]
dx
d
e
z
+
dx
d
[
e z
- = j e
z
=
i k
k
k
k
k k
k k tot

(7.6)

The sample is next assumed to be connected to an external electric circuit. This may be real,
with electrodes and wires and an electrical supply or load, or it may be absent, in which case
we say that we have an open circuit. Since the sample makes a series connection with the
external circuit, the currect in the sample must be the same as in the external circuit.
For a bare sample, a gas permeation membrane, or for an open-circuit fuel cell or other
electrochemical device, the total current is zero. For fuel cells or electrolysers in operation,
on the other hand, the current is non-zero.
By using the definition of total conductivity,
k k tot
=
, and the definition of
transport number,
k k
k
tot
k
k
=
t
= , we obtain the following very important expression:

dx
d
e
z
t
-
i
- =
dx
d
k
k
k
k
tot
tot

(7.7)

7.3
This relates the electrical potential gradient to the total (net, external) current density, the
total conductivity, and the transport number and chemical potential gradient of all charge
carriers.

From charged to neutral species: the electrochemical reaction

The chemical potentials of charged species are not well-defined, and we need to
represent them instead by chemical potentials of neutral species. For this purpose we may
assume equilibria between neutral and charged species and electrons, i.e. in the
electrochemical red-ox- reaction

ze +
S
= S
z
(7.8)

where S is a neutral chemical entity and z may be positive or negative. The equilibrium
condition for this is expressed in terms of the chemical potentials of products and reactants:

0 =
S e S
d zd + d - z (7.9)

and can be rearranged with respect to the ionic species:

e S S
- z zd d d = (7.10)

We insert this for all ionic species n in the expression for the electrical potential gradient. The
entry (among k) for electrons is left unsubstituted. By using 1 =
k k
t we obtain

dx
d
e
1
+
dx
d
e
z
t
-
i
- =
dx
d
e n
n
n
n
tot
tot

(7.11)

for which the chemical potentials now refer to the neutral forms of each carrier.

7.4
The voltage over a sample

We now integrate the electrical potential gradient over the thickness of the sample,
from side 1 to side 2, in order to obtain the voltage over the sample:

II
I
e
II
I
n
n
n
n
II
I
tot
tot
II
I
d
e
1
+ d
e
z
t
- dx
i
= d
(7.12)

) (
1
, ,

I e II e
II
I
n
n
n
n
II
I
tot
tot
I II
e
+ d
e
z
t
- dx
i
=

(7.13)

We further assume that the voltage is measured on each side using the same inert metal, e.g.
Pt. This eliminates the difference between chemical potentials of electrons on the two sides,
and the voltage measured between the two sides is

=
II
I
n
n
n
n
II
I
tot
tot
I II I II
d
e
z
t
- dx
i
= U
(7.14)

Under open circuit conditions, i
tot
=0, and we obtain

II
I
n
n
n
n I II
d
e
z
t
- U (7.15)

We shall later see how this is used to calculate transport numbers based on open circuit
voltage measurements of cells exposed to a well-defined gradient in chemical activities.
Alternatively, if t
n
is known to be unity for ionic charge carriers this expression will yield the
open circuit voltage of a fuel cell or a galvanic sensor.
If current is drawn from the sample, as in a fuel cell or battery under load, we need to
know how i
tot
varies with x. If we assume that it is constant, we get

=
II
I
n
n
n
n tot tot
II
I
n
n
n
n
tot
tot
I II
d
e
z
t
- r
i
d
e
z
t
-
X
i
U
(7.16)
7.5

where X is the thickness of the sample and r
tot
is the area-specific resistance of the sample.
Alternatively, i
tot
r
tot
(current density and area specific resistance) may be replaced by RI,
(sample resistance and current):

II
I
n
n
n
n I II
d
e
z
t
- IR U (7.17)

The voltage of a fuel cell or battery is thus composed of a thermodynamic part that
arises from the chemical gradient, and the well-known IR-term which arises from the
limiting kinetics (transport) in the sample. The letters IR may be taken to reflect current
times resistance, or to mean internal resistance. We will later see that the direction of the
current ends up such that the IR-term reduces the voltage over the fuel cell or battery.

Flux of a particular species

One of the general expressions for the electrical potential gradient can now be inserted
in an expression for the flux density of an individual species (Eq. 7.4) or the corresponding
partial current density (Eq. 7.5). Since these two sum over chemical potential gradients of
charged species, we may conveniently use Eq. 7.7 for our purpose. Inserting this into Eq. 7.4
and rearranging yields

]
dx
d
z
t
z
dx
d
[
e z e z
i
t
= j
k
k
k
k i
i
i
i
i
tot i
i

2
) (
(7.18)

The first term in the right hand side simply says what we expect: If there is a net current, a
flux density of species i will be set up proportional to the total current density and the
transport number t
i
, divided by the species charge.
The equation above is a rather general expression that we can use to calculate flux
densities of one charged species in the company of many other species. However, it reflects
the flux density and gradients and properties at a particular point through the membrane. The
gradients will adjust according to the varying materials properties so as to maintain a constant
flux density everywhere what we call steady state. In order to implement this, we integrate
7.6
the flux density expression over the thickness of the membrane and require that the flux
density remains constant:

=
II
I
k
k
k
k i
i
i
i
II
I i
tot i
i
II
I
i
] d
z
t
z
[d
e z
dx
e z
i
t
X j = dx j
2
) (
(7.19)

If the transport number t
i
and the total current density i
tot
can be taken as constant through the
membrane we further obtain:

II
I
k
k
k
k i
i
i
i
i
tot i
i
] d
z
t
z
[d
e z
X
e z
i
t
= X j
2
) (
(7.20)

Wagner-type transport theory case studies

Oxides with transport of oxide ions and electrons

General equations
We will now apply the general theory to a number of cases. The cases we will consider
first is an oxide that conducts oxygen ions and electrons. Flux equations for the two are

]
dx
d
e
dx
d
[
e
-
= j
O O
O

2
4
2 2
2
2

(7.21)

]
dx
d
e
dx
d
[
e
-
= j
e e
e

2
(7.22)

By summing up the currents from the two we obtain (as in Eq. 7.7) the expression for the
potential gradient:

7.7
dx
d
e
t
dx
d
e
t i
- =
dx
d
e e O O
tot
tot

+ +
2 2
2
(7.23)

In order to relate the potantial gradient to the chemical potential of neutral species we
introduce electrochemical equilibria between neutral and charged species of oxygen:

O
2
e
4 + (g)
O
2
2

= (7.24)

for which equilibrium can be expressed by

O
2d
e
4d +
O
d
2
2
(g)

= (7.25)

Introducing this we get

dx
d
e
t
dx
d
e
t
dx
d
e
t i
- =
dx
d
e e e O
g O
O
tot
tot

+ + +
2
2
2
) (
4
(7.26)

and knowing that the sum of transport numbers for oxygen ions and electrons in the oxide we
have defined is unity, we further get

dx
d
e dx
d
e
t i
- =
dx
d
e
g O
O
tot
tot

+ +
1
4
) (
2
2
(7.27)

whereby we have exemplified Eq. 7.11. We now integrate over the thickness of the sample to
obtain the voltage over it, eliminating the chemical potential of electrons (as in Eq. 7.14 and
onwards):

+ =
II
I
g O
O I II
d
t
e
IR U
) (
2
2
4
1
(7.28)

Using

2 2 2
ln
0
) ( ) ( O g O g O
p kT + = (7.29)
7.8
2 2
ln
) ( O g O
p kTd d = (7.30)

we get

+ =
II
I
O
O I II
p d
t
e
kT
IR U
2
2 ln
4
(7.31)

If
tO
2 is independent of the oxygen partial pressure, or if we can assume as an approximation
that an average
tO
2 is constant over the pressure range applied, then

N O
I
O
II
O
O I II
E
t
IR
p
p
e
kT
t
IR U + = + =
2
2
2
2 ln
4
(7.32)

where E
N
is the Nernst-voltage of the oxygen concentration cell.

Open circuit voltage of concentration cell for transport number measurements
The equation above can be used to measure the average oxygen ion transport number,
by measuring the open circuit voltage (OCV) over a sample exposed to a small, well-defined
gradient in partial pressure of oxygen:

N
I II
O
E
U
t

= 2 (7.33)

Galvanic oxygen sensor
By using a material with 1 2 =
tO
, i.e. a solid electrolyte, and a well-defined reference
partial pressure of oxygen, the OCV can be used for measuring an unknown partial pressure:

Ref
Ref
2
2
ln
4
O
II
O
II
p
p
e
kT
U =
(7.34)
7.9

Fuel cell
If we apply a solid electrolyte with 1 2 =
tO
and expose it to a gradient in oxygen partial
pressure, we get

N
I
O
II
O
I II
E IR
p
p
e
kT
IR U + = + =
2
2
ln
4
(7.35)

This would be the voltage of a fuel cell. With I = 0, we obtain the OCV of the cell to be the
Nernst voltage, of course. We furthermore see that the cell voltage diminishes due to the IR
loss at increasing current, I. A plot of U vs I gives R as the slope.
The power delivered by the cell is the current multiplied with the voltage;

N
IE R I UI P + = =
2
(7.36)

and it goes through a maximum at U=E
N
/2. Under these conditions, the power is split half-
half between external work and internal loss, so the electrical efficiency is 50%. The
efficiency increases as the power and current decrease, and commonly fuel cells are designed
to run at ca. U=2/3 E
N
.

Mixed conducting oxygen permeable membrane
Now we will show how we obtain the flux of oxygen through a mixed conducting
oxygen membrane material. As mentioned in the general section this is done by inserting the
electrical potential gradient back into the flux equation for oxygen ions. Using e.g. Eqs. 7.21,
7.23, and 7.25 we get:

dx
d
e
t
e
i
t
]
dx
d
dx
d
[
e
t
e
i
t
j
g O e
O
tot
O e O
e
O
tot
O
O

) (
2 2
2
2 2 2 2 2
2
8 2
2
4 2

= = (7.37)

The steady state restriction is introduced by integrating over the thickness of the membrane:
7.10

) ( 2 2
2 2
2 2
8 2
g O
II
I
e
O
II
I
tot
O
O
II
I
O
d
e
t
dx
e
i
t
X j dx j

= = (7.38)

p d
t
X e
kT
dx t
eX
i
j
O
II
I
e
O
II
I
O
tot
O
2
2 2 2 ln
8 2
2

= (7.39)

If i
tot
is not zero, i.e., some external current is drawn, then the first term says that the oxygen
ion flux is proportional to that current and to the oxygen ion transport number. If the latter is
independent of x, then the integral is straightforward. The last term adds to the flux; it is
driven by an oxygen pressure gradient, and it is proportional to the oxygen ion conductivity
and the transport number of electrons. This term describes permeation of oxygen ions,
requiring transport of both oxygen ions and electrons. Moreover, it is inversely proportional
to the thickness of the sample, X. Under open circuit conditions we may simplify to get:

p d
t
t
X e
kT
p d
t
X e
kT
j
O
II
I
e
O
tot O
II
I
e
O O
2
2
2
2
2 ln
8
ln
8
2 2

= = (7.40)

We now apply our derived expression for the example case of an oxygen-deficient oxide
MO
1-x
. This contains oxygen vacancies compensated by electrons, and we have earlier shown
that the defect concentrations are given by
6 / 1 3 / 1 /
2
) 4 ( 2

= =
O
v
O
p K ] [v n
O
. Since the mobility of
electrons is usually much higher than that of ionic defect, and with similar concentrations of
electrons and oxygen vacancies, we may assume that the material is mainly an n-type
electronic conductor. Therefore 1
e
t . Furthermore,
6 / 1
0
2
2
2

= = =

O
v
O
v O
p ]u e[v
O O
,
and when we insert these in the flux expression we obtain

p d p
X e
kT
j
O
II
I
O O
2 2
2 ln
8
6 / 1
2
0
(7.41)

We substitute
2
2
2
1
ln
O
O
O
dp
p
p d = and obtain

7.11
p d p
X e
kT
j
O
II
I
O O
2 2
2
6 / 7
2
0
8

=
(7.42)

which integrates into

{ }
6 / 1 6 / 1
2
0
) ( ) (
8
6
2 2
2

=
I
O
II
O O
p p
X e
kT
j
(7.43)

One may note that we here have utilised the Wagner approach for electrochemical transport
of two defect species and combined it with the solution of the defect chemical equilibria to
obtain an accurate description of the flux of oxygen ions. Numerical values for the flux of
course require that one knows
0
, which contains many parameters relating to formation and
mobility of the defects. In the right-hand parenthesis the negative exponent 1/6 has the
consequence that the smaller of the two partial pressures of oxygen has the biggest influence.
The larger partial pressure makes a comparatively smaller difference and may be neglected to
a first approximation for large gradients. For instance, consider a gas separation membrane
exposed to air on the one side and a very reducing gas (e.g. methane) on the other. The flux
would then be essentially proportional to
6 / 1
) (
2
red
O
p while it would largely be independent of
ox
O
p
2
. It would thus not help much to compress the air to 100 atm, while removal of oxidised
products like H
2
O on the reducing side would lower
red
O
p
2
and have a large effect.

Ambipolar diffusion and conduction

We have seen above that the transport coefficient term that enters in the flux of oxygen
ions in a mixed oxygen ion conductor is a product which we may call the ambipolar
conductivity

+
=
e O
e O
e O e O tot amb
- - 2
- - 2
- - 2 - - 2
t
=
t t
= (7.44)

This reflects the fact that so-called ambipolar conduction of two species is necessary (in this
case for charge compensation).
7.12
The corresponding terms transformed to diffusivities are called the ambipolar diffusion
coefficients for oxygen:

t D
=
D
e O amb
- - 2 (7.45)

Such terms, like partial conductivity and self-diffusion, reflect thermal random mobility
as before, but are restricted by the slower of the two ambipolar diffusing species (ions or
electrons).
Many other cases of combined transport coefficients are in use, e.g. the combined
(additive) transport of oxygen and metal ions commonly that we shall address later (and
exemplify by the high temperature oxidation of metals), the combination of two diffusivities
involved in interdiffusion (mixing) processes, and the mass transport in creep being rate
limited by the smallest out of cation and anion diffusivities in a binary compound. As some
of these sometimes are referred to as ambipolar or chemical diffusivities, we want to stress
the above simple definition of ambipolar transport coefficients as relevant for membrane
applications using mixed conductors.

Chemical diffusion and tracer diffusion
General case

The chemical diffusion coefficient is the phenomenological coefficient that enters a
Fick's 1st law of diffusion in a concentration gradient:

dx
dc
D
- = j
i
i i
~
(7.46)

Unlike our previous treatment in this chapter (which uses D
r
, B, u or ), Ficks 1
st
law and
the chemical diffucion coefficient says in general nothing directly about the self-diffusion
coefficient. Neither is, as we shall see, the chemical diffusion coefficient in general a constant
in any sense. Only when the species i is an effectively neutral particle that moves
independently of other species will the chemical diffusion coefficient be identical to the self-
diffusion coefficient. If the flux and the concentration gradient are in terms of an isotope
7.13
(tracer), the coefficient is the tracer diffusion coefficient, which is almost equal to the self-
diffusion coefficient:

D
f =
D r t
(7.47)

where the correlation factor f is a number usually equal to or somewhat smaller than unity,
depending on structure and diffusion mechanism, reflecting the difference of transport
between distinguishable and indistinguishable diffusing species. For vacancy diffusion, f
usually falls around 0.75 depending on structure.
Chemical diffusion is sometimes referred to as diffusion under a chemical potential or
concentration gradient. However, we have seen that self-diffusion properly describes this, and
that all diffusion phenomena we consider here are merely minor perturbations of thermally
induced self-diffusion. Why then do we need to be concerned with chemical diffusivity? We
use it whenever we want to or have to relate our flux to the concentration gradient. Under
steady state processes, our flux vs force equivalents provides sufficient description, and can
be integrated to yield steady state fluxes through membranes, as we have shown; Thus, for
this purpose, the self-diffusion coefficients and conductivities are useful since they behave in
a simple manner vs defect concentrations, and since we in general know the chemical and/or
electrical potential gradents applied. However, if we know the concentration gradients and
how the chemical diffusivity varies with concentration, we can use chemical diffusivity.
In a transient experiment, however, parts of the material take up or give away matter
(leading to changes in the concentration of species with time). This is sometimes referred to
as a capacitive effect and is expressed in Fick's 2nd law:

dx
)
dx
dc
D d(
=
dx
dj
- =
dt
dc
(7.48)

If D is independent of c and thus of x and t, then we can simplify to obtain:

dx
c
d
D =
dx
dj
- =
dt
dc
2
2
(7.49)

Mathematical solutions for interpretation of transient tracer, thermogravimetric and electrical
conductivity experiments are based on this. The diffusion coefficient obtained is the tracer
7.14
diffusion coefficient in case of a transient in isotope composition (whether measured
radiographically, mass spectrometrically, thermogravimetrically or electrically) or chemical
diffusion coefficients in case of a transient in chemical composition (measured
thermogravimetrically or electrically or by other means). The latter is usually executed by
stepping the oxygen activity by gas composition, total pressure, or electrochemical means.
It may be noted that certain tracer experiments actually yield an interdiffusion
coefficient between two tracers. The diffusion coefficient for oxygen in practice cannot and
need not distinguish between the different isotopes in use (usually
18
O and
16
O, more rarely
17
O). However, for protons, the difference to deuteron (or triton) transport coefficients is
considerable (e.g. a factor of 2) and should be taken into account when interpreting the
results.
The multitude of transport coefficients collected can thus be divided into self-diffusion
types (total or partial conductivities and mobilities obtained from equilibrium electrical
measurements, ambipolar or self-diffusion data from steady state flux measurements through
membranes), tracer-diffusivities, and chemical diffusivities from transient measurements. All
but the last are fairly easily interrelated through definitions, the Nernst-Einstein relation, and
the correlation factor. However, we need to look more closely at the chemical diffusion
coefficient. We will do this next by a specific example, namely within the framework of
oxygen ion and electron transport that we have restricted ourselves to at this stage.

Chemical diffusivity in the case of mixed oxygen vacancy and electronic
conductor

In a system with oxygen vacancies as the only oxygen point defects, the concentration
of oxygen ions is the concentration of oxygen sites minus the concentration of vacancies:

c
-
c
=
c
v
0
O
O
..
O
- 2
- 2 (7.50)

such that

dx
dc
- =
dx
dc v
O
..
O
- 2
(7.51)

7.15
We manipulate this in the following manner:

dx
p d
p d
dc
- =
p d
p d
dx
dc
- =
dx
dc
O
O
v
O
O v
O 2
2
..
O 2
..
O
- 2
ln
ln ln
ln
2
(7.52)

and insert the result into the flux equation for oxygen ions, Eq. 7.37 (open circuit, i
tot
= 0):

dx
dc
dc
p d
e
kT
t
dx
p d
e
kT
t
dx
d
e
t
j
O
v
O e
O
O e
O
g O e
O
O
O
= = =
2
2
2
2
2
2
2
2
ln
8
ln
8 8
2 2
) (
2

(7.53)

We have thus derived an expression for the flux that relates it to the concentration gradient of
oxygen ions. We may now therefore compare this expression with that of the oxygen flux in
terms of Fick's 1st law involving the concentration gradient in oxygen ions and the chemical
diffusion coefficient:

dx
dc
D
- = j
O
O O
- 2
- 2 - 2
~
(7.54)

and we obtain

=
O
- 2
v
O e
O
O
dc
p d
e
kT
t
D
2
2
ln
8
~
2
(7.55)

Further, by using the Nernst-Einstein-relation and inserting the relation
c D
=
c D
V V O O
..
O
..
O
- 2 2 we
get

2
2 2 2
2 2
ln
ln
2
ln
ln
2
ln
2
ln
2
~
O
v
e
v
v
O
e
v
v
O
e
v v
v
O e
O O
O
p d
c d
t
D
c d
p d t
D
dc
p d t
c D
dc
p d
t
c D
D
O
O
O
O
O
O O
O
- 2

= = = =
(7.56)

7.16

Thus, depending on the p
O2
-dependency of the vacancy concentration, the chemical
diffusion coefficient becomes a varying function of the vacancy (defect) self diffusion
coefficient. When the sample is mainly an electronic conductor and the transport number for
electrons is unity, the collection of remaining terms (the enhancement factor) takes absolute
values from 1 to 3 when the sample has a variable non-stoichiometry, while it increases
beyond this (and can take on very large values) as the stoichiometry approaches a constant
value, as in certain doped materials or materials with prevailing point defect (Schottky or
Frenkel) disorder; the vacancy concentration is then close to independent of p
O2
.
The full range of aspects of chemical diffusion coefficients make them rather difficult
to translate into other transport coefficients, and they tend to live their own lives in the
literature. They can, as said before, be applied directly to calculate fluxes in membranes if the
concentration gradients through the membranes are known. However, we note that the
chemical diffusion coefficient (for oxygen ions) is
- proportional to the self diffusivity of the oxygen defect (here the vacancy),
- proportional to the electronic transport number (but which is often unity),
- enhanced by the term
c
/2d p d
V
O O
2
ln ln , which for various limiting cases of
simplified defect situations takes on values of 1, 3 or 4, and
- fully forwards and backwards transformable into defect diffusivities (and in turn
self diffusivities) if we know the transport numbers and how defect concentrations vary
with pO
2
.

Surface and interface kinetics limitations

The flux of matter through a fuel cell or electrolyser is limited by the electrolyte (ionic)
resistance, the electrode kinetics, and the external electronic load resistance. We commonly
express the steady state situation and the fact that the current is the same through the entire
closed circuit in terms of the voltage drops around the circuit:

) + ( -
R
I -
E
= E
a c
i N
(7.57)

7.17
where I is the current, E
N
is the open circuit (Nernst) voltage, R
i
is the ionic electrolyte
resistance, and
c
and
a
are the cathode and anode overpotentials, respectively. (The
conventions and practices on measuring and reporting the sign of various voltages may vary,
so that the equation may take various forms with respect to signs).
Similarly, the current of matter through an electrodeless gas separation, mixed
conducting membrane will be limited by ionic conduction, the surface kinetics, and by the
internal electronic resistance, the latter replacing the external circuit in the previous case.
However, we may need to express the loss of potential in terms of the chemical
potential of oxygen.
If we look at the system without any gradients or currents applied, there will be
equilibrium at both the electrodes and the surfaces. Depending on the application and the side
of the membrane, the electrochemical processes in operation on an oxygen ion conductor may
be:

O
2
e
4 + (g)
O
- 2 -
2
= (7.58)

and

e
+ O(g)
H O
+ (g)
H
-
2
- 2
2
2 = (7.59)

In all of these the electrons can be supplied to or from an electrode or a mixed conductor
surface. Thus, basically, the reactions are the same in the electrolyte/electrode as in the mixed
conductor, although the physical presence of the electrode can promote (by catalysis) or
obstruct the reaction.
Most surface reactions are believed to be made up of a sequence of reactions steps.
Some possible step reactions in the case of the reduction of oxygen gas are given by the
following reaction sequence;
O
(g)
O ads 2, 2
=
O e
+
O
-
ads 2,
-
ads 2,
=
O
2
e
+
O
-
ads
- -
ads 2,
=
O e
+
O
- 2
ads
- -
ads
=
O O
- 2
O
- 2
ads
=
7.18
The last of these steps may involve bulk defects (e.g. a vacancy that the adsorbed ion may
occupy) but we will not attempt here an analysis of surface or electrode kinetics in terms of
defect chemistry. Determination of the nature of the rate limiting step of a reaction is a
difficult task. Studies of surface exchange kinetics using
18
O or by relaxation methods can
give information about the temperature and pressure dependence of the rate determining step.
However, it should be pointed out that information from such methods do not give direct
information about the exact mechanism and which intermediate species that are involved.
Each step will have its own specific rate and resistance to the mass transport across the
membrane. In lack of detailed knowledge of the rate limiting step(s) we normally have to
relate to the overall reaction.
At equilibrium the overall reaction takes place forth and back at equal rates r
0
as a
result of thermal energy. The activation process forth and back may not be symmetrical in
this case, but that will not affect us at present. Similarly to the random diffusion coefficient
connected with these thermal disorders in bulk (D
r
= 1/6 r
0
s
2
) we introduce an exchange
coefficient for interfaces; k
i
= r
0
s
i
which says something about the thermal fluctuations of the
reaction across the interface of thickness s
i
. We can alternatively express this as an exchange
flux density

c
s r
= c
k
= j
i 0 i
0
(7.60)

or an exchange current density

nec
s r
= nec
k
= ne j =
i i 0 i
0
0
(7.61)

where n is the number of electrons involved in the reaction. If we now apply a small force
over the interface, we get a perturbation and a net flux, given by

kT
dP
k
-c = j
i
(7.62)

where c is the volume concentration of species in the interface, and dP is the potential step
over the interface. The net current density is

7.19
kT
dP
i
=
kT
dP
k
cne = jne = i
0 i
(7.63)

If dP is taken as the electrical (over-)potential energy step affecting the reaction, dP = ne
and the overpotential becomes

i
i
ne
kT
=
0
(7.64)

This is the linear or ohmic (small overpotential) form of the kinetics expression, as we expect
from our specification of small forces and perturbations. We note that the interface thickness
s
i
has dropped out and is mostly not an interesting parameter at this level.
By measuring the overvoltage vs current density by voltammetric DC or impedance
spectroscopy AC methods, we may obtain the charge transfer resistance

c
k e n
kT
=
i
ne
kT
=
i
=
R
i
2 2
0
e
(7.65)

The overpotential enters directly into the sum of voltages over the fuel cell or electrolyser
circuit, and we can solve the system in terms of current and voltage. As the force and
perturbations of the reaction becomes larger, the full so-called Butler-Volmer exponential
dependence of current and flux on overpotential (and force) must be taken into account:

)
i
i
(
ne
kT
=
0
ln (7.66)

, but we will not elaborate on this here.
In order to approach an interpretation of transport measurements, we take as starting
point that the flux(es) through the interface must be the same as that in the bulk of the
membrane next to the interface in all cases (and in the entire membrane in the steady state
cases). This is done in reality and in the mathematical solutions by adjusting the potential at
the border between the interface and the bulk so as to equalise the two fluxes. However, this
soon becomes difficult, as it is not trivial to choose the proper set of potentials to apply for
different interface processes.
7.20
We resort first to chemical diffusion and assume that the interface process is one which
can be described as species flowing from a high to a lower concentration. Then

c
-k = )
dx
dc
D( -
interface
bulk
(7.67)

, and D as well as k must refer to chemical coefficients. This requirement can be included in
the solution of Fick's 2nd law for transients, and D and k can in principle be found
independently by fitting transient data to this solution. The k's obtained in this way can be
used to predict steady state fluxes if they are measured as a function of e.g. p
O2
or at the p
O2

extremes which the membrane is going to be subjected to. Otherwise, the k values are not
immediately easy to relate to other data.
In a tracer experiment, we obtain D
t
which is close to the self diffusivity (D
t
= f D
r
) and
then the exchange coefficient should also reflect the tracer exchange coefficient k
t
. k
i
and
this should be comparable to that calculated as above from measurements of R
e
; the
difference should say to what extent the presence of an electrode has changed the kinetics of
the rate limiting step of the surface reaction.
In order to analyse the effect of interface kinetics on fluxes through a membrane under
steady state conditions, we may take an approach similar to that for fuel cells and
electrolysers. Principally, no electrodes are involved, and voltages are therefore not directly
appropriate as in the other cases. Instead we may sum up chemical potential changes of
oxygen (or hydrogen) through the membrane and its interfaces. We start out with the
membrane interior, and from earlier we have for a mixed oxygen ion and electron conductor
of thickness L (earlier also referred to as X or x):

) ( 2 2
2
2
8
1
g O
II
I
e
O
O
d
e
t
L
j
= (7.68)

If we introduce ambipolar diffusivities and conductivities and assume these constant, the
equation may be integrated to obtain

L
e
8
=
L 2kT
c D
= j
O
2
amb O O amb
O
2 2
- 2
2

(7.69)

7.21
and, by rearrangement, the chemical potential drop over the bulk of the membrane is

amb
2
O
O amb
O
O
L
e
8 j -
=
c D
kTL j 2 -
=
2
- 2
2
2
(7.70)

This drop over the membrane bulk is given by the external full gradient minus the losses at
the interfaces. In the present terms, the flux through an interface is, from Eqs. 7.62 and 7.65
given by

R e
8
-
=
2kT
k c
-
= j
e
2
interface
O
interface
O
i O
O
2 2
- 2
- 2

(7.71)

such that, by rearrangement,

R e
8 j - =
c k
kT j 2 -
=
e
2
O
O i
O
interface
O
2
- 2
2
2
(7.72)

Now, we sum up the total gradient in chemical potential of oxygen over the bulk of the
membrane and the two interfaces

interface
O
bulk
O
tot
O 2 2 2
2 + = (7.73)

and after insertion of the above expressions for bulk and interface, we can solve for the flux
of oxygen ions through a membrane with two equally contributing interfaces under the
conditions given:

R
2 +
L
e
8 / -
=
k
2
+
D
L
/2kT
c
-
= j
e
amb
2
tot
O
i amb
O
tot
O
O
2
- 2
2
- 2

(7.74)

7.22
This result simply states that the flux is given by the driving force (which easily transforms to
Nernst voltage) divided by the total ambipolar resistance, in our case namely that of the bulk
and twice that of one interface.
In lack of full understanding of the relations involved in interface kinetics of
membranes, we may refer to a critical thickness; the thickness of the bulk membrane where
the interface and bulk impose equally large restrictions on the flux. In any membrane
application, going much below this thickness is not of much use. We see from the above that
this thickness is L
crit
=D/k for the thickness of the part of a membrane relating to one
interface, or L
crit
=2D/k for thickness of a membrane with two equally limiting interfaces,
typically the two surfaces.
Experimentally the value of L
crit
has been found to be of the order of 100 m for many
polished samples of different fluorite and perovskite materials, indicating that there is a close
relationship between e.g. vacancy concentration in bulk and the surface exchange. Roughing
the surface should increase relatively the surface flux, and this has been shown to work.

Ionic transport of both anions and cations

General expressions

We have in the previous section considered ionic transport by oxygen ions (anions)
only. This was done for simplicity, because it represents many real applications, and because
oxygen ions relate most directly to the oxygen (non-metal) activity which we may control
most conveniently.
Now we will include cation (metal ion) transport. Some systems have predominant
transport of cations, and we thus need to see how this relates this to the non-metal activity.
Some systems have both anion and cation transport so we need to take both into account. In
some cases we may not know which one is dominating and as we shall see it is not always
easy or necessary to distinguish them.
In an inorganic compound the total ionic current density, i
ion
, is given by the sum of the
current densities of anions, i
an
, and the cations, i
cat
. From our previously derived expressions
for partial current densities we thus obtain

dx
d
e z dx
d
e z
ej z ej z i i i
cat
cat
cat an
an
an
cat cat an an cat an ion

+
= + = + = (7.75)
7.23

where z
an
and z
cat
represent the valences and thus ionic charges of the anions and cations,
respectively.
In order to relate the current density of the cations and anions let us consider the
equilibrium involving the formation of the compound M
a
O
b
from its ions:

b a
z z
O M bO aM
an cat
= + (7.76)

Through the Gibbs-Duhem relation, equilibrium in this reaction may be expressed as

) (s O M
O M b a
an
z
cat
z
d bd ad = + (7.77)

The fact that 0
) (
=
s O M
b a
d arises from M
a
O
b
(s) being a pure condensed phase. It may further be
noted that

an cat
bz az = (7.78)

so that we obtain the very important expression

dx
d
z
z
dx
d
a
b
dx
d
an
an
cat an cat

= = (7.79)

From this we note that the cation and anion chemical potential gradients are the negative of
each other in a binary compound, a relation that is used extensively, mathematically or by
intuition.
By adding
dx
d
e z
cat
to both sides of the equation and rearranging on the right hand side
we obtain

dx
d
z
z
dx
d
an
an
cat cat

= (7.80)

7.24
By combining Eqs 7.75. and 7.80 the total ionic conductivity may be expressed by

dx
d
e z dx
d
e z
i
an
an
ion cat
cat
ion
ion

= = (7.81)

Thus, the ionic current density can be expressed in terms of the ionic conductivity (sum of
cationic and anionic conductivities) and the gradient in the chemical potential of cations or
anions.
From this it becomes clear that the derivations of oxygen ion fluxes and currents in
various situations done earlier could have been done for the case of ions (sum of cations and
anions) simply by inserting
ion
instead of
2
O
, but with for instance the oxygen activity
gradient as the driving force.

Membrane walk-out
Metal ions are transported from the low oxygen pressure to the high oxygen pressure
side of the membrane while metal vacancies are transported in the opposite direction. During
the process oxygen is liberated and oxygen and metal sites are annihilated at the low oxygen
pressure side, e.g. for a metal deficient oxide M
1-
O with doubly charged metal vacancies:

O
O
+ V
2
'
M
+ 2h
.
=
1
2
O
2
(g) (7.82)

while equivalent number of lattice sites are formed at the high oxygen pressure side:

1
2
O
2
(g) = O
O
+ V
2
'
M
+ 2h
.
(7.83)

In this way the oxide membrane is actually moving in laboratory space in the direction of the
higher oxygen potential.
From this we see that even a very minor transport of metal ions may be detrimental to
the operation of an oxygen separation membrane or fuel cell: Over the many years of
operation the electrolyte or membrane will simply walk out of its housing, towards the high
oxygen pressure.

7.25
Demixing of oxide solid solutions
An additional important aspect of diffusional transport of metal ions in a chemical
potential gradient is that in a homogeneous crystal of an oxide solution, e.g. (A,B)
1-
O such
as for instance (Co,Ni)
1-
O, a demixing process begins to take place. Both cations move by
vacancy diffusion. When one of the cations in (A,B)
1-
O, e.g. A
2+
, has a higher mobility than
the other, the A
2+
ions will move faster to the side of the higher oxygen potential. The solid
solution is enriched in AO at this side and thus becomes kinetically demixed. After extended
time a steady state concentration profile is reached. This can be formulated through the use of
the appropriate transport equations and by taking into account the conditions of
electroneutrality and of local thermodynamic equilibrium. The process is schematically
illustrated below.

Figure 7-1. Schematic illustration of kinetic demixing of the oxide solid solution (A,B)
1-
O in a oxygen
potential gradient.

Multicomponent compounds, e.g. ABO
3
and AB
2
O
4
, can similarly be demixed in an
oxygen potential gradient. In this case defects are formed due to the nonstoichiometric A/B
ratio. Both in the case of the solid solution and in the case of a ternary oxide, new phases may
be precipitated as soon as the concentration of solute or defects exceeds the solubility limit.
We may in fact end up in the peculiar situation that the starting oxide is stable in both the low
and the high oxygen activities separately, but unstable in the gradient between them.
The decomposition and precipitation of new phases may violate the functional or
mechanical properties of the material. These types of phenomena are not only of importance
in oxide membranes, but also in oxidation of alloys where solid solutions of two or more
oxides or multicomponent compounds may be formed in the oxide scales.

7.26
High temperature oxidation of metals; the Wagner oxidation theory

When high temperature oxidation of metals results in the formation of compact scales
and sufficient oxygen is available at the oxide surface, the rate of reaction is governed by the
solid state diffusion of the reactants or transport of electrons through the scale. As the
diffusion distance increases as the oxide grows in thickness, the rate of reaction decreases
with time. When the diffusion takes place homogeneously through the scale, the rate of
growth of the oxide thickness, x, is inversely proportional to the oxide thickness:

x
k
dt
dx
p
1
*
= (7.84)

In the integrated form this becomes

0 0
* 2
2 C t k C t k x
p p
+ = + = (7.85)

where ) 2 (
*
p p
k k = and
*
p
k are expressions for the rate constants and C
o
is an integration
constant. The oxide thickness may alternatively be measured in terms of the oxygen uptake
by the metal specimen during oxidation by using, for instance, thermogravimetry. The oxide
thickness or oxygen uptake is parabolic with time, and the oxidation is termed parabolic.
Accordingly
p
k and
*
p
k are termed parabolic rate constants. The figure below illustrates the
variation in the oxide thickness or oxygen uptake as a function of time.

Figure 7-2. The variation in oxide thickness as a function of time (x vs t and x
2
vs t) during parabolic
oxidation of metals
7.27

Carl Wagner developed the theory for parabolic oxidation assuming that the reaction is
governed by lattice diffusion of the reacting ions (metal and oxygen ions) or transport of
electrons. The important aspect of the Wagner oxidation theory is that the parabolic rate
constant is expressed in terms of independently measurable properties, i.e. the electronic and
ionic conductivity of the oxide or alternatively in terms of the self-diffusion coefficients of
the reacting ions. This thus provides an interpretation of the reaction mechanism and a
theoretical basis for changing and improving the oxidation resistance of metals and alloys.
The Wagner theory has been one of the most important contributions to our understanding of
the high temperature oxidation of metals.
The basic assumption of the theory is that the lattice diffusion of the reacting ions or the
transport of electrons through the dense scale is rate-determining for the overall reaction.
Lattice diffusion is assumed to take place because of the presence of point defects, and the
migrating species may alternatively be considered to constitute lattice and electronic defects.
Wagner further assumed that ions and electrons migrate independently of each other and that
local equilibria exist within the oxide. Such transport processes through a dense, single-phase
scale growing by lattice diffusion is illustrated in the figure below. One part (a) illustrates the
transport of metal and oxygen ions and electrons, while the other (b) illustrates the transport
processes when the predominant diffusion processes involve metal vacancies and interstitials.

Figure 7-3. Transport processes during growth of a dense, single phase scale growing by lattice
diffusion. a) transport processes illustrated by transport of reacting ions, b) transport processes illustrated by
transport of diffusion of important point defects (in this case assumed to be metal vacancies and interstitials).

7.28
As diffusion through the scale is rate-determining, reactions at the phase boundaries are
considered to be rapid, and it is assumed that thermodynamic equilibria are established
between the oxide and oxygen gas at the oxide/oxygen interface and between the metal and
the oxide at the metal/oxide interface.
The overall "driving energy" of the reaction is the Gibbs free energy change associated
with the formation of the oxide, e.g. M
a
O
b
, from the metal M and the oxygen gas.
Correspondingly, a gradient in the partial pressure (activity) of oxygen exists across the scale
from the partial pressure of oxygen in the ambient atmosphere at the outer surface to the
partial pressure of oxygen at the metal/oxide interface. The latter partial pressure is the
decomposition pressure of the oxide in equilibrium with its metal.
The main driving force for the transport through the plane is the chemical potential
gradient. But when one considers the transport of the reacting ions and of the electrons
through the scale, it may also be noted that the mobilities of the cations, anions and electrons
are not equal. Because of this difference, a separation of charges takes place in a growing
scale. This creates a space charge (diffusion potential) that opposes a further separation of
charges, and a stationary state is reached for which no net electric current flows through the
scale. In describing the transport of ions and electrons through the scale, it is thus necessary
to take into account both the transport due to the gradient of the chemical potential and that
due to the gradient of the electrical potential, i.e. the electrochemical potential gradient. Thus,
the treatment of the transport through the oxide scale is identical to that done earlier for
electrochemical transport through a mixed conducting oxide.
The net ionic current is from earlier given by

dx
d
e dx
d
e z
i
O ion an
an
ion
ion
= =
2
2

(7.86)

and the electronic current is given by

dx
d
e dx
d
e
i
el el el el
el

=
=
1
(7.87)

As no net current flows through the scale during the scale growth, then i
tot
= i
ion
+ i
el
= 0.
By solving this system in the usual manner (solving for the electrical potential gradient,
inserting into the expression for the ionic current density and inserting the equilibrium
condition between oxygen ions, oxygen molecules and electrons) we obtain
7.29

dx
p d
e
kT t
dx
d
e
t
i
O
el ion
g O
el ion
ion
2 2
ln
4 4
) (

= = (7.88)

The rate of oxygen ions reacting per unit area is obtained by dividing the ionic current density
by the charge of the oxygen ions. Moreover, the rate dn/dt of oxide molecular units M
a
O
b

formed per unit area is obtained by further division by b:

dx
p d
b e
kT t
dx
d
b e
t
eb
i
dt
dn
b dt
dn O
el ion
g O
el ion ion O 2 2
ln
8 8 2
1
2
) (
2
= =
= = (7.89)

Similar to the integrations we have done earlier over the thickness of the oxide, we now
integrate over the instantaneous thickness x and from the ambient oxygen pressure,
0
2
O
p , to
the partial pressure at the metal/oxide interface,
i
O
p
2
. The growth rate
dn
dt
then takes the form

x
p d t
b e
kT

dt
dn
O el ion
1
ln
8
o
2
O
i
2
O
2
p
p
2
(7.90)

We have chosen to organize the equation and direction of integration such that the rate is
positive if
0
2
O
p >
i
O
p
2
. In general the directionalities and signs of the fluxes and processes that
go on pose a problem for us. One may most often neglect this issue because the output,
namely growth (increase in amount of oxide or thickness of scale) is intuitively positive.
However, it may not be just that easy, since there is nothing preventing scale reduction
(reverse of growth) if the outer oxygen partial pressure is smaller than the activity at the
metal/oxide.
Many readers will probably wonder about another small apparent problem: We
integrate over the scale to take into account the steady state condition, namely that the fluxes
are constant throughout the scale. Still, the scale grows, so how can it be steady state? The
answer is that it is steady state (constant fluxes) for that moment of time with that
instantaneous thickness of the scale. So our integration is for a given moment in time. At
another moment the integration will give a different flux.
7.30
The expression in the parentheses,
o
2
O
i
2
O
2
p
p
2
ln
8
O el ion
p d t
b e
kT
, can be considered to be
one form of the parabolic rate constant and is in the following written k
t
:

x
k
dt
dn
t
1
= (7.91)

k
t
thus represents a time-independent coefficient, while the flux and thickness of the scale
vary according to the parabolic relation above.
As written in this form
dn
dt
is expressed in number of molecular units of M
a
O
b
per
cm
2
sec and x in cm. The derivation of Wagner's equation that we here have done for growth
of oxide scales, may be applied to many other gas-metal reactions.
It may be noted that although the total particle current is equal to the rate of growth of
the scale, the rate-determining process may either be diffusion of ions or transport of
electrons depending on the properties of the scales. If the scale is an electronic conductor, t
el

1, the diffusion of the reacting ions through the scale is rate-determining, while the
transport of electrons through the scales is rate-determining if the scale is an ionic conductor,
t
ion
1. The oxygen pressure dependence of the reaction will depend on whether the scale is
an ionic or electronic conductor and in the latter case on the type of nonstoichiometry that
prevails in the oxide scale.
In our derivation, the growth rate is proportional to
ion
t
el
=
tot
t
ion
t
el
= (
an
+
cat
) t
el
.
However, the conductivity of the metal and oxygen ions in M
a
O
b
can be expressed in terms
of the self-diffusion coefficients of the metal and oxygen ions in M
a
O
b
through the Nernst -
Einstein relation (z
i
2
e
2
D
i
c
i
= kT
i
). Let us insert this relationship in Eq.7.89 and assume that
the oxide is an electronic conductor (t
el
1). Let us furthermore take into account that
concentrations of metal ions (cations), c
M
, and oxygen ions, c
O
, in an oxide M
a
O
b
with
relatively small deviation from stoichiometry are related through c
M
/c
O
= a/b = |z
an
|/z
cat
. The
expressions for the flux through the growing scale (expressed in terms of the number of
M
a
O
b
units per cm
2
sec.) and for k
t
then become

dx
p d
D D
z
b
c
dx
p d
D D
z
z
b
z c
dt
dn O
O M
cat O
O
O M
an
cat an O 2 2
ln
)
2
(
2
ln
) (
8
2
+ = + = (7.92)

7.31
x
k
x
p d D D
z
b
c

dt
dn
t O O M
cat
1 1
ln )
2
(
2
2
o
2
O
i
2
O
p
p
0
=
+ =
(7.93)

c
O
denotes the number of oxygen ions in M
a
O
b
per cm
3
and D
M
and D
O
are respectively the
self-diffusion coefficients of the metal and oxygen ions in M
a
O
b
. It should be noted that D
M
and D
O
are the self-diffusion coefficients for random diffusion of the respective ions.
In Eqs.7.92 and 7.93, k
t
represents the rate of formation of M
a
O
b
units per cm
2
per
second for an M
a
O
b
scale of thickness 1 cm. One may as said before alternatively express the
parabolic rate constant in terms of the rate of growth of the oxide thickness

x
k
dt
dx
p
1
*
= (7.94)

The dimensions of k
*
p
are then cm
2
sec
-1
, and Eq.7.93 then takes the form

x
k
x
p d D D
z

dt
dn
c
b
dt
dx
p O O M
cat
1 1
ln )
2
(
2
1
*
p
p
0
2
o
2
O
i
2
O
=
+ = =
(7.95)

Growth of metal-deficient M
a-y
O
b
on M
Let us make use of Eq. 7.95 to illustrate the temperature and oxygen pressure
dependence of the parabolic rate constant. Consider the growth of the metal deficient oxide
M
a-y
O
b
on high purity M. Let us assume that the predominant defects throughout the entire
scale are metal vacancies with an effective charge , i.e. V
'
M
. Let us further assume that the
metal ion diffusion is much faster than the oxygen diffusion, i.e. D
M
>> D
O
. Eq. 7.95 then
reduces to

x
k
x
p d D
z
dt
dx
p O M
cat
1 1
ln
4
*
p
p
2
o
2
O
i
2
O
=
(7.96)

7.32
In order to integrate this we need to analyse the defect structure of the oxide. The formation
of the metal vacancies may be written as

x
O M
O
a
b
h v g O
a
b
+ + =

/
2
) (
2
(7.97)

If other defects may be neglected and the electroneutrality condition is given by [V
'
M
] = p,
the concentration of metal vacancies is given by

a b
O
V
M
p K v
M
2 / 1 /
2
/
] [

=
+
(7.98)

where
/
M
V
K is the equilibrium constant for the formation of the metal vacancies (Eq. 7.97)
and ] [
/
M
v denotes the fraction of the metal ion sites that are vacant. On the basis of this
relation,
2
ln
O
p d may be written

] [ ln
) 1 ( 2
ln
/
2
M O
v d
b
a
p d
+
= (7.99)

When this expression for
2
ln
O
p d is introduced in Eq.7.96 , the expression for k
*
p
becomes

} ] [ ] {[ ) 1 (
i / o / *
/

M M
v
p
v v D k
M
+ = (7.100)

If the concentration of metal vacancies at the oxide/oxygen interface is much larger than at
the metal/oxide interface, i.e.
i / o /
] [ ] [

M M
v v >> then

o
M M
v
p
D v D k
M
) 1 ( ] [ ) 1 (
o / *
/
+ = + =

(7.101)

where
o /
] [
/
M
v
o
M
v D D
M
= is the self-diffusion coefficient of the metal ions in M
a-y
O
b
at the
oxide/oxygen interface. Thus under these conditions the parabolic rate constant and the self-
diffusion coefficient of M in M
a-y
O
b
has the same temperature and oxygen pressure
dependences.

7.33
Growth of oxygen-deficient M
a
O
b-y
on M
The same procedure may be used to derive the parabolic rate constant for the growth of
oxygen deficient M
a
O
b-y
on M. let us assume that the oxygen vacancies have charge , i.e.
V
.
O
, and the formation of the oxygen vacancies may then be written

O
O
= V
.
O
+ e' +
1
2
O
2
(7.102)

If D
O
>>D
M
and (p
i
O
2
<<p
o
O
2
) then k
*
p
becomes
) 1 ( 2
1
2
) ( ) 1 ( ) 1 ( ] [ ) 1 (
0 i * +
+ = + = + =

i
O O
i
O O
v
p
p D D v D k
O
(7.103)

where

O
v
D is the self-diffusion coefficient of the oxygen vacancies,
i
] [

O
v the concentration
of oxygen vacancies and D
O
i
the self-diffusion coefficient of oxygen in M
a
O
b-y
at the
metal/oxide interface. D
O
0
is the oxygen self-diffusion coefficient at 1 atm. O
2
, and p
i
O
2
is
the oxygen activity at the metal/oxide interface. For this case the parabolic rate constant is
independent of the ambient oxygen pressure, and the temperature dependence is given by that
of the oxygen self-diffusion coefficient in M
a
O
b-y
at the metal/oxide interface. In this respect
it may be further noted that the oxygen self-diffusion coefficient at the metal/oxide interface
is given by that the product of the self-diffusion coefficient of oxygen at constant oxygen
pressure (at 1 atm O
2
) and (p
i
O
2
)
-1/2(+1)
, i.e. D
O
i
= D
O
0
(p
i
O
2
)
-1/2(+1)
. In these terms the
activation energy of the parabolic rate constant is under these conditions given by the
activation energy associated with D
O
0
and the enthalpy term associated with (p
i
O
2
)
-1/2(+1)
.
Similar treatments may be given for oxide scales growing by interstitial metals ions or
interstitial oxygen ions.

Scales with ionic conductivity predominant.
Most oxides encountered in high temperature oxidation of metals are electronic
conductors. In the literature there are no examples of high temperature metal-oxygen
reactions involving conventional metals or alloys with essentially pure ionic conductivity
over the entire existence range of the oxide. Such type of reactions are, however, found in
metal-halogen reactions, e.g. Ag+Br
2
to form AgBr. The same treatment as given above may
be applied to reactions involving formation of ionically conducting scales, and the important
7.34
feature of these reactions is that it is the electronic transport through the scale that is rate-
determining.

Varying defect structure situations through the scale.
In the preceding examples it has been assumed that the same defect structure prevails
throughout the entire scale from the scale/gas to the metal oxide interfaces. This is an
oversimplified model for many systems. The charge on the defects, or even the predominant
defects may change with changing oxygen activity. Furthermore, the presence of impurities
and dopants may significantly affect the defect structure situation. Thus, following the
discussion in Ch.4 one part of the scale may possibly have intrinsic properties and another
part extrinsic properties. As part of such behaviour one part of the scale may have significant
ionic conductivity while the rest has electronic conductivity.
In the presence of water vapour, hydrogen defects may affect the diffusional behaviour
in growing oxide scales in many different ways.

Sintering

Stages of the sintering process
Sintering is the process whereby powders and small particles agglomerate and grow
together to form a continuous polycrystalline body. As a rule of thumb the material must be
heated to 2/3 of the melting temperature to achieve considerable sintering.
The overall sintering process may conveniently be divided into different stages. The
principal stages involved is illustrated schematically below.
7.35

Figure 7-4. Principle stages during the sintering process: a) particles in contact prior to sintering; b)
neck growth between particles during initial stage of sintering; c) further neck growth and formation of
continuous pore channels along grain boundaries; and d) further sintering in closing channels and formation of
closed pores at grain corners.

During an initial stage the surface roughness is decreased and the particles begin to
adhere and grow together. At the end of this stage grain growth begins to occur and grain
boundaries and metastable pore phase is established. Only modest shrinkage of a few percent
takes place during the initial stage.
During an intermediate stage grain growth continues, and the cross-sectional area of the
pore phase decreases. The final stage begins when the body achieves 90-95% theoretical
density. At this point the pore phase becomes discontinuous, and the closed pores usually
remain at grain boundaries. The final stage may involve complete removal of the remaining
pores, leading to a completely dense material. Alternatively, it may involve discontinuous
growth of the large grains at the expense of the small ones, and closed pores may as a
consequence be isolated inside the grains. If the latter process occurs, complete densification
becomes extremely difficult.

Driving force for sintering
During sintering, as for other spontaneous processes, the total free energy of the system
decreases. For elemental solids and homogeneous compounds the only free energy change
involved in sintering is that of the surface free energy or surface tension, , of the particles.
7.36
Any surface atom or molecule is subjected to a resultant inward attraction because of
the unsaturated chemical bonds at the surface. The surface therefore tends to contract to the
smallest possible area. In the case of a spherical particle with radius r, the interior is subjected
to an excess force of r
2
p, where p is the pressure. This is counteracted by the surface tension
acting along the circumference: 2r. When equating these opposing forces, the pressure is
given by

p =
2r
r
2
=
2
r
(7.104)

From this it is seen that the vapour pressure of a sphere is larger the smaller the radius
and as a result a large sphere will grow at the expense of a neighbouring small sphere. This
type of grain growth in polycrystalline solids is termed Ostwald's law. Similarly the vapour
pressure of convex surfaces is larger than the vapour pressure of a concave surface and as a
result surface roughness of particles is reduced during the initial stage of sintering.

Transport processes during sintering
As for creep, sintering of oxides implies that molecular units of the oxide are
transported from areas from high to low surface free energies, e.g. from convex to concave
surfaces. Generally it is concluded that transport of the slower moving species determines the
overall sintering rate. Thus, in a binary oxide where oxygen ions diffuse much more slowly
than the metal ions, it is generally to be expected that the sintering rate is determined by the
transport, e.g. lattice or grain boundary diffusion of oxygen.
Sintering governed by lattice diffusion will be dependent upon the concentration of
point defects in oxides. Accordingly, sintering rates of an oxide can be optimised by close
control of impurities or dopants and the ambient partial pressures of oxygen and of water
vapour in cases where proton defects affect the defect structure of the oxide.

Oxides with additional transport of protons

If an oxidic material conducts protons in addition to oxygen ions and electrons we need to
introduce electrochemical equilibria between neutral and charged species of both oxygen and
hydrogen:
7.37

O
2
e
4 + (g)
O
2
2

= (7.105)

e
+
H
2 (g)
H2
+
= 2 (7.106)

for which equilibria can be expressed by

O
2d
e
4d +
O
d
2
2
(g)

= (7.107)

e
d +
H
2d
H
d
(g)
2
+
= 2 (7.108)

We insert these into the flux equations for the species and utilise that the sum of all transport
numbers equals unity. We furthermore use that d = kTdlna kTdlnp, where a and p are
activity and partial pressure, respectively, and obtain

dx
d
e
1
+
dx
p d
2e
kT
t
-
dx
p d
4e
kT
t
+
i
- =
dx
d
e H H O O
tot
tot 2
+
2
- 2
ln ln
(7.109)

This can now be integrated to obtain the voltage of a cell, or inserted into a flux equation of a
species of interest in the usual manner. We now have a system with two chemical driving
forces for electrochemical transport; that of oxygen activity and that of hydrogen activity.

Other cases

We have in the present version of the treatment of electrochemical transport omitted
many cases. These include a more full coverage of proton transport in oxides and transport by
other hydrogen species and other foreign species. Moreover, we have left out the case of
solid-solid reactions. Also, creep and sintering were given more phenomenological than
defect-chemical treatments.
We have throughout assumed isothermal conditions. If we instead of (or in addition to)
chemical driving forces considered a temperature gradient, our integrations over membranes
would have been from one temperature to another. The voltage we would have obtained
7.38
would be the thermoelectric voltage. It has several contributions, but in general it is related to
the transport numbers and the gradients in concentrations of defects. The variations of
concentration is in turn given by the entropy of the defects.

More on these themes will be included in future updates.
7.39
Excercises

1. Electrochemical cell with stabilised zirconias (YSZ and CSZ)

a) Consider an yttria-stabilised zirconia (YSZ) material with 90 mol% ZrO
2
and 10
mol% Y
2
O
3
. What is the site fraction of oxygen vacancies in the material? We often
use and refer to the YSZ as an electrolyte. What is an electrolyte? What requirements
does it put on a material?

b) We make ourselves a disk of YSZ, attach Pt electrodes to both faces, and place the
disk in a cell where we can expose the two faces to different gases. If the gases have
different partial pressures of oxygen, a voltage will be set up over the electrodes, and
we can draw current from the cell. Write equations for the two half-cell-reactions that
run when current is drawn. Write the expression for the voltage over the cell given
that the material is a purely ionic conductor. Which (high or low pO
2
) is the positive
pole?

c) Smith, Meszaros and Amata (J. Am. Ceram. Soc., 49 (1966) 240) found that the
permeabilities at high temperatures of Ar and N
2
through calcium-stabilised zirconia
(CSZ) were immeasureable, while that of O
2
was significant and proportional to
p
O2
1/4
. Suggest an explanation for these observations.

2. Electrochemical cell for determination of thermodynamic properties

a) Indicate how you can use a YSZ cell with Pt electrodes to determine
the thermodynamic parameters for formation of a binary oxide (e.g. NiO).

b) J.J. Egan (J. Phys. Chem. 68 (1964) 1978) studied an electrochemical
cell built up as follows:

- (Mo)Th(s),ThF
4
(s)|CaF
2
(s)|ThF
4
(s),ThC
2
(s),C(s)(Mo) +
in order to measure the thermodynamic properties of ThC
2
. The Mo serves as
inert electrodes. The cell was kept under an inert atmosphere. The CaF
2
(s) is a solid
electrolyte a fluoride ion conductor in the temperature range of operation of the
cell (700-900C). The signs indicate the polarity of the cell. Write the reactions at
each electrode when current is drawn from the cell. Also, express the voltage of the
cell as a function of the Gibbs energy, enthalpy and entropy of formation of ThC
2
.

7.40
3. Wagners oxidation theory

Wagners equation for the oxidation rate of a metal, forming the oxide M
a
O
b
, can in
one form be written

x
p d t t t
bF
RT
dt
dn
O el an cat
+ =
1
ln ) (
8
2
1
2
2

a) Specify what the different symbols denote in this equation.
b) Examine the equality of units on both sides of the equation.
c) What are the model and requirements for the theory. Why do the different terms
enter the equation as they do?
d) This type of behaviour is called parabolic. What does it reflect? Identify the
parabolic rate constant.
e) Which deviations from the requirements of the theory may give rise to non-
parabolic (e.g. linear) behaviour?
f) Assume that the oxide has formula MO and is a p-type semiconductor with doubly
charged metal vacancies as dominating point defects (D
M
>> D
O
). Assume that
pO
2
o
>> pO
2
i
and find how the oxidation rate varies with the outer oxygen
pressure pO
2
o
.
g) Assume that the oxide has formula M
a
O
b
and that it is an n-type semiconductor
with M
i
as dominating point defects and that D

M
>> D
O
. How does the oxidation
rate in this case vary with the outer oxygen partial pressure (assuming also now
that pO
2
o
>> pO
2
i
)?
h) The tracer self-diffusion coefficient for Co in CoO in air at 1200C is 9*10
-9

cm
2
/s. Calculate the parabolic rate constant in cm
2
/s (which is k
p
in dx/dt = k
p
/x)
when cobalt oxidizes to CoO under the same conditions. The dominating point
defects in CoO are assumed to be singly charged metal vacancies v
Co
.

4. Sintering

a) What happens during sintering? What are the driving forces? What are the
contributing mechanisms and what may be the rate limiting factors?
b) Powders of CoO can be sintered to dense samples. The defect structure of
CoO is characterized by the following: There is a deficiency of metal and
the dominating point defects are singly charged metal vacancies. D
Co
>>
D
O
.Assume that the oxygen defects are interstitials with one negative
effective charge. Draw a Brouwer-diagram for the oxide. Under what
conditions should the oxide be sintered to achieve the highest sintering rate?

7.41
5. Creep

a) What is creep? What are the driving forces, mechanisms, and rate limiting
factors?
b) TiO
2
is an n-type electronic conductor with oxygen deficiency, dominated
by doubly charged oxygen vacancies and triply charged titanium
interstitials. Assume that the concentration of oxygen vacancies is 1000
times higher than that of titanium interstitials at pO
2
= 1 atm. Draw a
Brouwer diagram for the oxide from pO
2
= 10
-20
to 1 atm. How would you
qualitatively expect that the creep of TiO
2
single crystals varies with pO
2
?

6. General considerations

b. The expression uphill diffusion is used for diffusion against a concentration
gradient, referring to Ficks 1
st
law. From what you know about
electrochemical transport, what might cause uphill diffusion?
c. Eq. 7.4 is written with conductivity as the transport coefficient. Flux of a
neutral species is not affected by the electrical potential gradient and can not
be expressed by conductivity. It is therefore meaningful to express the
equation using the random diffusion coefficient for the chemical potential
gradient and the conductivity for the electrical potential gradient. Do this
splitting/substitution. (This is mainly a simple exercise in the Nernst-Einstein
relation). Check that both parts of the resulting equation will have units that
confer with flux density.

7. Transport number and partial conductivity measurements

d. A membrane of a mixed oxygen ion and electron conducting oxide is equipped
with Pt electrodes and exposed to two different oxygen partial pressures: On
one side is dry air (p
O2
= 0.21 atm) and on the other is pure oxygen (p
O2
= 1.00
atm). What is the Nernst voltage of this cell at 1000C?
e. A voltage of 0.01 V is read over the cell at 1000C using a high impedance
voltmeter. What is the average ionic transport number of the oxide under these
conditions?
f. Make a sketch of the sample and gases and external circuitry during the
voltage measurement. Indicate internal and external fluxes and currents (with
directions).
g. The total conductivity of the sample is measured and the result is 0.01 S/cm.
What are the partial ionic and electronic conductivities?
h. Assuming that the sample is 0.5 mm thick, what is the flux of oxygen gas
through the sample during the voltage measurements?
i. If you had measured this flux in some way, and used this as your (only)
transport measurement, what would be the transport parameter that you find?
7.42

8. Sensor

a) A sensor is made out of a membrane of yttria.stabilized zirconia (YSZ) with Pt electrodes
and is operated at 700C. As reference is used pure oxygen at 1 atm. A voltage of -950 mV is
read over the sensor (measuring on the unknown side vs the reference side). What is p
O2
at
the unknown side?

b) If you assume that the unknown side is hydrogen gas at 1 atm, what does the sensor
voltage say about the water vapour content of that hydrogen gas?

9. Fuel cell

a) A fuel cell made of YSZ electrolyte with Pt electrodes is operated at 700C and runs on
hydrogen as fuel and air as oxidant, both at atmospheric pressure. Both gases contain 2%
water vapour. What is the open circuit voltage of the fuel cell?

b) Assuming that the electrolyte has a conductivity of 0.1 S/cm and is 100 m thick, calculate
the short-circuit current density, the maximum power density and the current and power
densities at a voltage of 2/3 of the open circuit voltage. (Assume in all cases that the gas
compositions remain unaffected by the current in the cell, and that electrode polarization is
insignificant).
c) Assuming the operation at 2/3 of the OCV voltage, what is the area of electrolyte needed
for a power plant of 300 MW capacity?
d) The assumption that the gas compositions are unaffected is not realistic for a real power
plant; what would it mean in terms of fuel utilization?
e) The system we have described can be operated at other temperatures. Still neglecting
electrode polarization, what are the factors and processes that enter into the temperature
dependency of the power density of the fuel cell? (Hint: refers to defects and transport.)

10. Gas separation membranes

a)
A gas separation membrane that can separate oxygen from air is constructed from
La
2
NiO
4+d
, a material where we for simplicity may assume that the defect structure is
dominated by doubly ionized oxygen interstitials compensated by electron holes. In the
following assume further that the cell is operated at 1000C and that the membrane is 100 m
thick and has no surface kinetics limitations. Assume also that the oxygen ion conductivity
(by interstitials) is 1 S/cm at 1 atm O
2
and that the electronic (hole) conductivity is 200 S/cm
at 1 atm O
2
. The membrane is operated with atmospheric air on one side and pumped to
effectively 0.1 atm at the other side. What is the area-specific flux of oxygen?
7.43
b)
What would increase the flux more: Compress the air to 10 atm or pump the vacuum side to
0.01 atm?
1. Find an expression for the chemical diffusion coefficient for oxygen ions.
2. During operation the membrane may possibly suffer from both membrane
walk-out and decomposition. Explain both phenomena qualitatively.
c) Solid solid reactions

1. La
2
NiO
4
may be synthesized by solid-state reaction between La
2
O
3
and
NiO. Imagine that you investigate this reaction by making tablets of the
two reactants and holding them together at high temperature. Sketch how
the product La
2
NiO
4
is formed as a phase between the reactants and what
diffusion process (or processes) that take place to constitute the reaction.
2. Suggest how you might use a marker (e.g. a Pt wire) to identify the rate
limiting diffusion process.
3. The reaction rate can be measured as the thickness of the product layer. It
would be expected to grow parabolically. Explain.
d) High temperature oxidation of metals

1. Zr metal is used at high temperatures in nuclear installations. It is
protected by an oxide layer of ZrO
2-y
, dominated by oxygen vacancies and
electrons.
2. What is rate limiting for the parabolic oxidation process in your opinion?
3. Nb is often used as allyoing element in Zr in order to reduce oxidation of
the metal. Explain how this dopant might work when it ends up in the
oxide scale.
4. In nuclear reactors it is a problem that hydrogen diffuses as protons
through the oxide scale, dissolves in the Zr alloy and makes it brittle. How
would you expect Nb alloying to affect the proton transport through the
oxide scale?
5. Ni oxidizes to Ni
1-y
O, i.e. an oxide dominated with Ni vacancies and
electron holes. How would you suggest to dope this oxide to reduce
oxidation?

Tables
Greek letters

alfa
beta
gamma
delta
epsilon
zeta
eta
theta
iota
kappa
lambda
my
ny
ksi
omikron
pi
rho
sigma
tau
ypsilon
phi
khi
psi
omega

Constants

Gravitational acceleration g 9,80665 m/s
2

Atomic mass unit u 1,6605*10
-27
kg
Electron mass m
e
9,110*10
-31
kg
Elementary charge e 1,602*10
-19
C
Electron volt eV 1,602*10
-19
J
Proton mass m
p
1,673*10
-27
kg
Neuytron mass m
n
1,675*10
-27
kg
Speed of light in vacuum c 2,99792*10
8
m/s
Boltzmanns constant k 1,381*10
-23
J/K
8,6174*10
-5
eV/K
Plancks constant h 6,626*10
-34
Js
Rydbergs constant 1,097*10
7
m
-1

Bohrs constant B 2,18*10
-18
J
Avogadros number N
A
6,022*10
23
mol
-1

Gas constant R 8,31451 J/molK
0,0820578 Latm/molK

Faradays constant F 96485 C/mol
Gravitational constant 6,672*10
-11
Nm
2
/kg
2
Permeability in vacuum
0
1,257*10
-12
H/m
Permittivity in vacuum
0
8,854*10
-12
F/m
Electrical constant =(4
0
)
-1
k
e
9,0*10
9
Nm
2
/C
2

Magnetic constant k
m
2*10
-7
N/A
2

Stefan-Boltzmann constant 5,67*10
-8
W/m
2
K
4

Wiens constant a 2,90*10
-3
mK
Volum of 1 mol ideal gas V
m,298K
24,4651 L

Defects and Transport in Crystalline Solids

Uploaded by

Copyright:

Available Formats

Defects and Transport in Crystalline Solids

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

Defects and Transport in Crystalline Solids

Uploaded by

Copyright:

Available Formats

Per Kofstad

and Truls Norby

) and the effective negative charge by a superscript prime (

h . The oxide is thus a p-type electronic

, and from the Schottky equilibrium

. The overall defect

= , we obtain the following very important expression:

as dominating point defects and that D

You might also like