You are on page 1of 18

View Article Online

View Journal

Journal of
Materials Chemistry A
Materials for energy and sustainability
Accepted Manuscript

This article can be cited before page numbers have been issued, to do this please use: R. De Souza, J.
Mater. Chem. A, 2017, DOI: 10.1039/C7TA04266C.

Volume 4 Number 1 7 January 2016 Pages 1–330 This is an Accepted Manuscript, which has been through the
Royal Society of Chemistry peer review process and has been
Journal of accepted for publication.
Materials Chemistry A Accepted Manuscripts are published online shortly after
Materials for energy and sustainability

acceptance, before technical editing, formatting and proof reading.


www.rsc.org/MaterialsA

Using this free service, authors can make their results available
to the community, in citable form, before we publish the edited
article. We will replace this Accepted Manuscript with the edited
and formatted Advance Article as soon as it is available.

You can find more information about Accepted Manuscripts in the


author guidelines.

Please note that technical editing may introduce minor changes


to the text and/or graphics, which may alter content. The journal’s
ISSN 2050-7488
standard Terms & Conditions and the ethical guidelines, outlined
in our author and reviewer resource centre, still apply. In no
PAPER
Kun Chang, Zhaorong Chang et al.
Bubble-template-assisted synthesis of hollow fullerene-like
MoS2 nanocages as a lithium ion battery anode material
event shall the Royal Society of Chemistry be held responsible
for any errors or omissions in this Accepted Manuscript or any
consequences arising from the use of any information it contains.

rsc.li/materials-a
Page 1 of 17 Journal of Materials Chemistry A
View Article Online
DOI: 10.1039/C7TA04266C

Jour
nal
Name

Journal of Materials Chemistry A Accepted Manuscript


Published on 11 September 2017. Downloaded by FLORIDA ATLANTIC UNIVERSITY on 12/09/2017 03:59:30.

Limits to the rate of oxygen transport in mixed-


conducting oxides
Roger A. De Souza∗a

The processes of oxygen tracer diffusion and oxygen surface exchange in acceptor-doped oxides
displaying high rates of oxygen transport are examined. Literature data are analysed with the
standard atomistic expression, in order to obtain chemically reasonable limiting values for the
oxygen diffusion coefficient, D∗ , in the solid state. The analysis suggests a limiting value of D∗
that is characterised by a pre-exponential of 8.8 × 10−2 cm2 s−1 and an activation enthalpy of
0.5 eV. An (essentially temperature-independent) ultimate limit to the oxygen surface exchange
coefficient, k∗ , of ca. 100 cm s−1 is obtained by assuming surface exchange to be limited by the
incident flux of oxygen molecules from the gas phase. In addition, atomistic expressions are
derived that quantitatively describe experimental k∗ data as a function of temperature and oxygen
activity for surface exchange involving molecular oxygen at high temperatures. One expression is
found for oxides with a bandgap, such as SrTiO3 , Ce0.9 Gd0.1 O1.9 , and La0.6 Sr0.4 Fe0.8 Co0.2 O3−δ ;
and an analagous expression is found for oxides with a partially filled conduction band, such as
La1−a Sra CoO3−δ and (Ba0.5 Sr0.5 )(Co0.8 Fe0.2 )O3−δ . Conduction-band electrons are considered
for both types of oxides to play a central role in determining the rate of oxygen surface exchange.

Certain crystalline oxides possess the ability to conduct, at an ap-


preciable rate, oxygen ions or oxygen ions and electrons. These mixed ion–electron conducting (MIEC) oxides — materials that
abilities can be put to use in a variety of solid state electrochemi- adopt, with a few notable exceptions, the AO2 fluorite structure
cal systems, such as solid oxides fuel cells, solid oxide electrol- or the ABO3 perovskite structure —, the first two processes are
yser cells, catalytic converters, oxygen permeation membranes generally found to be the most important. 4,5,22 It is thus with
and thermochemical splitting devices. 1–20 the limits to bulk diffusion and to surface exchange in acceptor-
Improving the overall performance of such electrochemical doped oxides that crystallise in the AO2 fluorite structure or the
systems requires, in many cases, increases in the rates of oxy- ABO3 perovskite structure that this study is concerned.
gen transport exhibited by the materials. To this end materials
chemists are currently pursuing research along two intertwining
1 Diffusion
paths: optimising the composition and microstructure of current There are various means of probing oxygen diffusion in an ox-
materials, and discovering new materials with superior proper- ide, and thus there are a variety of diffusion coefficients that can
ties. While vast strides have been made, and further strides are be extracted from such experiments. 23,24 The focus in this paper
certain to be made, it is unclear how much further a given mate- is on tracer diffusion coefficients (and surface exchange coeffi-
rial can be improved through compositional and microstructural cients) measured by means of isotope exchange and SIMS analy-
tuning and it is unclear whether it is worth searching for new ma- sis. 25–28
terials with better properties. In other words, the question arises: The tracer diffusion coefficient of oxygen, D∗O , measured in
What are the limits to the rate of oxygen transport in an oxide? such experiments is given by
This question is closely followed by a second question: How close
are we at present to these limits? D∗O cO = fv∗ Dv cv + fi∗ Di ci , (1)
In principle there are various processes that may govern the ∗ are the tracer correlation coefficients for oxygen vacan-
where fv,i
overall rate of oxygen transport through an oxide: 9,21,22 bulk dif-
cies (v) and oxygen interstitials (i); Dv,i are the diffusion coeffi-
fusion, surface exchange, transport across or along grain bound-
cients of these two species; and cO , cv and ci are the concentra-
aries. In the best oxygen-ion conducting oxides and the best
tions of oxygen ions, oxygen vacancies and oxygen interstitials.
In general, one of the two defect moieties (v or i) will provide the
a
dominant contribution to D∗O , as its concentration will be many
Institute of Physical Chemistry, RWTH Aachen University, 52056 Aachen, Germany.
Fax: +49 241 8092128; Tel: +49 241 8094739; E-mail: desouza@pc.rwth-aachen.de
orders of magnitude higher than that of the other moiety. For ex-

J
our
nal
Name,
[yea
r][
,vol
.
],1–17 | 1
Journal of Materials Chemistry A Page 2 of 17
View Article Online
DOI: 10.1039/C7TA04266C

ample, in an acceptor-doped oxide, in which the concentration of


vacancies is fixed by the concentration of acceptor dopants, i.e.
2cv = ca , the concentration of oxygen interstitials is many orders
of magnitude lower than that of oxygen vacancies, being given
by ci ∝ c−1
v exp(−∆HaF /kB T ), where the enthalpy of anti-Frenkel

Journal of Materials Chemistry A Accepted Manuscript


(aF) disorder, according to theoretical estimates, is, for example,
∆HaF ≈ 7 − 8 eV for (perovskite) SrTiO3 29,30 and ∆HaF ≈ 4 eV for
Published on 11 September 2017. Downloaded by FLORIDA ATLANTIC UNIVERSITY on 12/09/2017 03:59:30.

(fluorite) CeO2 . 31,32


With the assumption that the vacancies, say, execute an
uncorrelated, random walk in three dimensions, one can ex-
press 13,17,24 the vacancy diffusivity in terms of certain atomistic
quantities: Zv , the number of neighbouring ions that can jump
into the vacancy; (1 − nv ), the probability that these sites are oc-
cupied with an ion; dv , the jump distance; ν0 , a characteristic
‡ ‡
lattice frequency; and ∆Smig,v and ∆Hmig,v , the activation entropy
and activation enthalpy of vacancy migration:
Zv ‡ ‡
Dv = (1 − nv )dv2 ν0 e∆Smig,v /kB e−∆Hmig,v /kB T . (2)
6
An analagous expression can be derived for Di .
Consequently, obtaining a limiting value for Dv can be formu-
lated as the problem of identifying the extremum values of the
parameters contained in eqn (2). One would like, of course, to
obtain a chemically reasonable limiting value of Dv , and hence
the trivial case of setting the activation enthalpy of migration to
zero is to be avoided. In addition, one would like to take account
of intrinsic correlations (or anticorrelations) between certain pa-
rameters in eqn (2), in order to constrain these parameters to

chemically reasonable values. Such a correlation between ∆Smig,v

and ∆Hmig,v , for example, has been conjectured in order to ex-
plain the Meyer–Neldel rule, 33–36 and it clearly constrains the
‡ ‡
value that ∆Smig,v may take for given ∆Hmig,v . Inclusion of such
correlations is, however, far from trivial. I pursue, therefore, the
following approach: experimental data for a variety of systems
displaying high rates of oxygen tracer diffusion are analysed with
eqn (2), and the extremum values obtained for the constituent Fig. 1 Oxygen diffusion coefficients against inverse temperature for
selected oxides: (a) measured tracer diffusion coefficients D∗O and (b)
parameters are combined to obtain a limiting value of Dv . The
calculated defect diffusion coefficients Dv,i for vacancies or interstitials.
benefit of this approach is that, considered individually, all the a – Zr0.81 Y0.19 O1.905 (ZY); 41,42 b – Ce0.9 Gd0.1 O1.9 (CG); 41 c – δ -Bi2 O3
quantities are chemically reasonable; whether the combination of (δ B); 43 d – SrTiO3 (ST); 46 e – (Ba0.5 Sr0.5 )(Co0.8 Fe0.2 )O3−δ (BSCF); 48 f
parameters is chemically reasonable remains an open question. – (La0.9 Sr0.1 )(Ga0.8 Mg0.2 )O2.85 (LSGM); 47 g – La2 NiO4+δ (LN); 44 h –
Ca12 Al14 O33 (CA). 45 DO,lim and Dv,lim represent the proposed limits to
oxygen diffusion and to oxygen-vacancy diffusion in oxides.
1.1 A chemically reasonable limit for D∗O
Comparisons of measured oxygen tracer diffusivities and cal-
culated defect diffusivities have been made in the past no- Ce0.9 Gd0.1 O1.95 (CG) 41 and δ -Bi2 O3 (δ B) 43 ], and oxygen-
tably by Mizusaki and co-workers 6,37 and Kilner and co- interstitial diffusion in the Ruddlesden–Popper phase, La2 NiO4+δ
workers. 5,9,38–40 Building on this work, I compared in a pre- (LN) 44 and in the cage-compound mayenite, Ca12 Al14 O33
vious paper 24 the diffusivities of oxygen vacancies in selected (CA). 45 Oxygen-vacancy diffusion in three characteristic per-
perovskite oxides, some with low vacancy concentrations and ovskites [SrTiO3 (ST), 46 (La0.9 Sr0.1 )(Ga0.8 Mg0.2 )O2.85 (LSGM) 47
some with high vacancy concentrations. Surprisingly, the and (Ba0.5 Sr0.5 )(Co0.8 Fe0.2 )O3−δ (BSCF) 48 ] is also included.
data clustered together strongly, suggesting that, for oxygen- The results are shown in Fig. 1.
vacancy diffusion in ABO3 perovskite oxides, the limit is close In Fig. 1(a) we find a familiar result: 1,2,43 the highest known
to being, or has already been, reached. Here, with the aim oxygen tracer diffusivity is exhibited by δ -Bi2 O3 (dataset c).
stated above, of obtaining a general, chemically reasonable (Data for many other oxides could be added to this plot, such
limit to Dv , I extend the comparison to oxygen-vacancy diffu- as those for materials based on Bi4 V2 O11 , 49,50 La2 Mo2 O9 51,52
sion in fluorite-structured oxides [Zr0.91 Y0.09 O1.955 (ZY), 41,42 or La10 Si6 O27 , 53,54 but adding such data would not change this

2| J
our
nal
Name,
[yea
r][
,vol
.
],
1–17
Page 3 of 17 Journal of Materials Chemistry A
View Article Online
DOI: 10.1039/C7TA04266C

finding.) The exceptionally high values are attributed to cv /cO these limiting values become
being extraordinarily high (1/4 of all anion sites are empty in δ -
1.7 × 107
 
0.5 eV
Bi2 O3 ), and to the absence of acceptor-dopant cations that would σO2− ,lim = exp − S K cm−1 . (4)
T kB T
diminish the vacancies’ rate of migration. Bayliss et al. 43 , through
such a comparison, proposed that the extrapolation of the data This gives σO2− ,lim = 50 S cm−1 at T = 1000 K.

Journal of Materials Chemistry A Accepted Manuscript


for δ -Bi2 O3 to higher and lower temperatures represents the limit It is important to note that, despite δ -Bi2 O3 exhibiting the high-
Published on 11 September 2017. Downloaded by FLORIDA ATLANTIC UNIVERSITY on 12/09/2017 03:59:30.

for the rate of oxygen diffusion in the solid state. This limit is of est Dv values observed for an oxide to date, there is no one thing
course chemically reasonable because it is entirely empirical. But that is exceptional about the parameters of eqn (2) obtained for
this is also its weak point. One cannot rule out that in the future this material. Certainly, of the three AO2 -based oxides, its dv is the
an oxygen-ion conductor superior to δ -Bi2 O3 will be discovered. ‡
highest, its ∆Hmig,v ‡
the lowest, and its ∆Smig,v the highest, but the
In Fig. 1(b), the isothermal defect diffusivities are seen to ex- differences to the parameters derived for ZY and CG are not large.
hibit relatively little spread. In particular, values of Di for the It appears, therefore, that it is the combination of parameters that
two systems considered (datasets g and h), surprisingly, are of yields the high Dv of δ -Bi2 O3 . And if it is simply a combination
the same order of magnitude and are characterised by similar of parameters, other, even better combinations may be possible.
activation enthalpies of migration as those for Dv . And this is Hence, according to this analysis, the message to the com-
surprising because one may have expected Di  Dv or Di  Dv . munity of materials chemists is that—provided that the param-
Perhaps if one considers, at the most basic level, that in both cases eters of eqn (2) can be varied independently of one another—the
(v and i) an oxygen ion is moving within a lattice of charged ions, limit of oxygen-ion transport in the solid state has not yet been
with Coulomb interactions determining the height of the migra- reached. Are there, then, superior oxygen-ion conductors waiting
tion barrier, 55 Di and Dv cannot be too dissimilar. In any case, to be identified? Possibly. It depends on whether thermodynami-
based on the present evidence, there is no reason to favour ox- cally stable compounds exist that display some superior combina-
ides with many mobile interstitials over oxides with many mobile tion of parameters. If not, δ -Bi2 O3 represents the closest we can
vacancies, or vice versa, at least as far as the defect diffusivity is get to the limit. There are, however, many crystal structures and
concerned. compositions, in which oxygen-ion transport has not yet been in-
vestigated, and hence it is conceivable that a superior oxygen-ion
We now analyse the data plotted in Fig. 1(b) in terms of eqn conductor is waiting to be identified.
(2). The focus is on vacancy migration because of the wealth
of data available, but the conclusions are equally applicable to
1.2 DO,lim : Discussion
interstitial migration. Implicit in this analysis is the assump- ‡ ‡
tion that oxygen vacancies are randomly distributed over the It is tempting to take the extremum values of ∆Hmig,v , ∆Smig,v
regular oxygen-ion lattice sites and execute a random walk in and dv , and to modify them appropriately, in order to attain even
three dimensions. This is only true, however, for SrTiO3 in higher Dv,lim and DO,lim . I refrain from taking this course because,
this temperature range. Defect–defect interactions are known to although small variations in the parameters are imaginable, large
complicate oxygen-vacancy dynamics in ZY, CG and LSGM; 56–59 changes seem improbable.
even in δ -Bi2 O3 there are complications. 60,61 In addition, room- First, it seems unlikely that the activation enthalpy of migra-
temperature values of dv and a temperature- and material- tion for oxygen vacancies can be depressed to much below 0.5
independent ν0 = 1013 Hz have been used. The values obtained eV. Put again at the most basic level, O2– is a rather large ion 62

for ∆Hmig,v ‡
and ∆Smig,v are thus effective values, but this is suf- and it bears a nominal double charge: together, these two factors
ficient for our purposes. The analysis of data for Dv thus yields may restrict the migration enthalpy to values above 0.5 eV. (Sig-

0.5 ≤ ∆Hmig,v ‡
/eV ≤ 1.0 and −2.1 ≤ ∆Smig,v /kB ≤ 3.1. nificantly lower activation enthalpies of ion migration are known,
but they are exhibited either by (singly charged) halide ions, e.g.
Examination of the parameters obtained reveals that the X− in ABX3 , 63 or by the smaller, singly charged Li+ . 64 ) To date,
extremum values come from only two materials: the best ‡
∆Hmig,v < 0.5 eV has never been confirmed experimentally for
AO2 fluorite, δ -Bi2 O3 (δ B), and the best ABO3 perovskite, any oxide. 7,24,36,40,65 Atomistic simulations of vacancy migration,
(Ba0.5 Sr0.5 )(Co0.8 Fe0.2 )O3−δ (BSCF). Returning, then, to eqn employing classical pair potentials or Density-Functional-Theory
(2), with Z = 8 (ABO3 structure), dv = 2.82 Å (both δ B and (DFT) calculations, 66–74 generally predict activation enthalpies
‡ ‡
BSCF!), ∆Smig,v = 3.1 kB (δ B) and ∆Hmig,v = 0.5 eV (BSCF), and of migration larger than 0.5 eV for diverse ABO3 perovskites and
with nv = 0.5 to guarantee the maximum DO , we obtain the line AO2 fluorites,b but in some cases, values of 0.4 eV or lower. Such
in Fig. 1(b) denoted Dv,lim , and subsequently with eqn (1), the data refer, however, to individual jumps and not to long-range
line DO,lim in Fig. 1(a) (since cv = cO , the two are identical): transport, and they do not take defect–defect interactions into ac-
  count.
0.5 eV
DO,lim = Dv,lim = 8.8 × 10−2 exp − cm2 s−1 . (3) Second, the jump distance of oxygen ions in an oxide is cer-
kB T

These limiting values are just over one order of magnitude higher
than the respective data for δ -Bi2 O3 . Expressed in terms of an b
It is mentioned that DFT calculations do not always provide superior results to
ionic conductivity (by means of the Nernst–Einstein relation), those from classical pair potentials. 75,76

J
our
nal
Name,
[yea
r][
,vol
.
],1–17 | 3
Journal of Materials Chemistry A Page 4 of 17
View Article Online
DOI: 10.1039/C7TA04266C

1.3 Beyond compositional tuning of DO,lim


Four effects are currently being examined in the literature with
the aim of increasing DO by means other than tuning of the com-
position. These are defect redistribution between two phases,
leading to the formation of space-charge zones (deviations from

Journal of Materials Chemistry A Accepted Manuscript


local electroneutrality); high electric fields (deviation from the
Published on 11 September 2017. Downloaded by FLORIDA ATLANTIC UNIVERSITY on 12/09/2017 03:59:30.

regime of linear response); and mechanical strain or amorphisa-


tion (deviations from the equilibrium structure).
The first of these effects—the re-distribution of defects between
two phases to form space-charge zones 85 —affects the concentra-
tion of charge carriers. It makes no sense, however, to induce dis-
tribution in a system in which nv already exceeds several percent,
e.g. Ce0.9 Gd0.1 O1.9 or (La0.9 Sr0.1 )(Ga0.8 Mg0.2 )O2.85 , as increases
in nv by orders of magnitude 86,87 are not possible. In contrast,
‡ defect re-distribution between two phases of a heterostructure 88
Fig. 2 Activation entropy of vacancy migration, ∆Smig,v , versus activation
‡ offers a more promising route. An improvement on defect re-
enthalpy of vacancy migration, ∆Hmig,v , obtained through analysis with
eqn (2) of experimental data for selected compounds in Fig. 1. distribution between two phases is modulation doping, as used
in semiconductor heterostructures, but applied to ionic conduc-
tors: 73,85,89 one material is doped to generate a high concentra-
tion of charge carriers (v or i); on account of the Gibbs forma-
tainly not limited to dv ≤ 2.82 Å, but this is perhaps irrelevant. In tion energy (∆Gform ) of the generated charge carrier being lower
a number of perovskite-type oxides, for example, dv is larger than in the neighbouring undoped material, the charge carriers pre-

2.8 Å, but none of these oxides exhibits ∆Hmig,v < 0.5 eV. In this fer to reside there; in this way there are no local interactions
regard, it is intriguing that both δ B and BSCF exhibit dv = 2.82 between charge carriers and dopants, and high rates of trans-
Å, and that this value is twice the radius of an oxygen ion. 62 It port may be achieved. Arguably, this strategy does not provide
remains to be seen if this value has some special significance, i.e., a method for advancing beyond DO,lim ; it only suggests an easier
if it represents a balance between various competing factors that route to approach DO,lim . Once such a combination of materials
govern ion migration, or if it is simply coincidence. In any case, undoped doped
(i.e. one with ∆Gform,v/i < ∆Gform,v/i ) has been identified, such a
since dv does not appear in the argument of an exponential term, heterostructure can, of course, be fabricated. An alternative to
large increases in dv will not have a large effect on Dv . growing such heterostructures is to let nature do the work, that
Third, there surely is a limit to the activation entropy of mi- is, to use a layered structure such as an Aurivillius phase and to
gration, but what this limit is is unclear. If we consider the data dope appropriately the layer in which oxygen defects prefer not
for the oxides examined here, we find that, with the exception to reside.
‡ ‡
of δ -Bi2 O3 , ∆Smig,v and ∆Hmig,v are evidently correlated (see Fig. The second effect is the application of a high electric field to
2). The fact that one exception is observed, however, suggests increase the ion mobility by depressing the activation enthalpy
that other exceptions are possible. This brings us back, then, to of migration in the direction of migration. 90 The field strengths

the problem of what a reasonable limiting value for ∆Smig,v is. required to increase the mobility by orders of magnitude are,

Whereas a small increase to ∆Smig,v = +4 kB seems possible, a however, enormous. 91–93 At ∼ 101 MV cm−1 , they are close to

larger increase to ∆Smig,v = +10 kB seems much less likely. 77–80 or above the breakdown field strength of most materials, at least
Beyond this, there is not much more that can be said without ex- for macroscopic specimens. At the nanoscale distances, though,
tensive calculations and without further high quality data because breakdown may be shifted to even higher field strengths. In ad-
simple approaches to obtaining a limiting value are inadequate dition, a field strength of 101 MV cm−1 corresponds to applying
(see Appendix A). 1000 V to a 1 µm-thick sample, but only 1 V to a 1 nm-thick sam-
This section concludes with a brief comment on the amount ple. This effect may only be useful, therefore, at the nanoscale.
of oxygen defects that can be introduced into an oxide. The Applying mechanical strain is one effect that produces, at least
problems associated with high nv are known. High nv tends to according to atomistic simulations, 87,94–96 enhancements in DO
go hand in hand either with limited thermodynamic stability (cf. of up to 104 by modifying the activation barrier for migration. De-
δ B and BSCF 73,81–83 ); or with a decrease in vacancy diffusiv- spite initial excitement, 97 later studies failed to confirm increases
ity with increasing nv on account of defect–defect interactions; by orders of magnitude, 86,98 although increases by a factor of ca.
or with assimilation of the vacancies into the structure. 9,84 In 2 are reproducible. 99,100 In addition, it is unclear how to con-
this regard, oxides with high concentrations of oxygen intersti- struct geometries useful for high-flux electrochemical devices and
tials have not been studied as intensively, and so it remains to be whether the high strain state of such a structure is stable over
seen whether they are plagued by the same problems. If not, or if extended periods of time.
not so severely, such materials may ultimately offer a better route The ultimate perturbation of a crystal structure is conceivably
to extremely high oxygen diffusivity. amorphisation. It has been shown 101,102 that the mobility of Li+

4| J
our
nal
Name,
[yea
r][
,vol
.
],
1–17
Page 5 of 17 Journal of Materials Chemistry A
View Article Online
DOI: 10.1039/C7TA04266C

is increased in an amorphous phase relative to the parent crys-


talline phase. Both examples, however, refer to materials in which
the ion diffusivity in the parent crystalline phase is very low to
begin with. Recently, Schie et al. 74 reported simulations of ion
transport in crystalline and amorphous HfO2 . Amorphisation did

Journal of Materials Chemistry A Accepted Manuscript


not increase anion diffusion significantly but it did increase cation
diffusion. In general, these three results 74,101,102 are in accord
Published on 11 September 2017. Downloaded by FLORIDA ATLANTIC UNIVERSITY on 12/09/2017 03:59:30.

with the hypothesis of Metlenko et al. 103 : pronounced perturba-


tions of the crystal structure, e.g., at dislocations, at grain bound-
aries or in amorphous phases, adversely affect ion migration in
those structures in which ion migration is rapid but are beneficial
for those structures in which ion migration is generally slow.
In conclusion, it appears that none of these four effects will
increase the oxygen diffusivity by orders of magnitude. DO,lim is H k*
s

thus proposed to be a general limit to the rate of oxygen transport


in an oxide.

2 Surface exchange
The surface exchange coefficient ks∗ describes the dynamic equi-
Fig. 3 Surface exchange coefficients as a function of inverse
librium between oxygen molecules in the gas phase and oxygen
temperature obtained from the collisional flux of oxygen molecules on a
in the solid. This single parameter thus characterises a multistep surface at aO2 = 0.2: solid line, eqn (5); dashed lines, eqns (5) and (6)
process in which, combined with various steps of charge transfer, with various values of the critical kinetic energy of the incident O2
oxygen molecules are adsorbed and dissociated, and the result- molecule, E cr .
ing atomic species are incorporated into the solid. This much is
agreed. 4,5,23,41,104–137
the solid than the gas phase can deliver.
There is, however, no agreement as to the exact series of ele- ∗
Values of ks,lim calculated from eqn (5) as a function T for aO2 =
mentary reaction steps for a given acceptor-doped oxide or as to
0.2 and with cO = 5 × 1022 cm−3 are shown in Fig. 3 as a solid
which of these steps is rate determining. There is also no agree- ∗
line. As expected from consideration of eqn (5), ks,lim depends
ment as to whether each acceptor-doped oxide is characterised by
only weakly on temperature, with an effective activation enthalpy
a separate mechanism; whether certain closely related groups of
(∆Hks∗ ) close to zero for this range of temperatures. The absolute
acceptor-doped oxides exhibit a common mechanism; or whether ∗
value of ks,lim is ca. 100 cm s−1 for aO2 = 0.2 (and increasing aO2
all acceptor-doped AO2 and ABO3 oxides display the same uni- ∗
will only increase ks,lim linearly). This limiting value is orders of
versal mechanism.
magnitude higher than any values measured for acceptor-doped
One consequence of this lack of agreement is a general dif-
oxides, 133,134 which are k∗ ≤ 10−4 cm s−1 .c It appears, therefore,
ficulty in estimating chemically realistic maximum values of ks∗ .
that we are currently far from this ultimate limit, and hence, that
Here, two different approaches are examined, one proceeding
there is, to paraphrase Feynman, “plenty of room at the top”.
from the gas phase and the other from the surface reaction.
One note of caution is necessary, however. One implicit as-
sumption in using eqn (5) to estimate ks,lim ∗ is that every oxygen
2.1 Limit to k∗ from the gas phase molecule that hits the surface sticks to it and contributes to jO .
Let us suppose that the flux of oxygen molecules in the gas phase That is, the sticking coefficient α is implicitly assumed to be unity.
coll , governs the rate of oxy-
that impinges on the solid surface, jO2
α, however, may deviate substantially from unity, as in principle
gen exchange. This approach builds on comparisons made by it may vary with the kinetic energy of the incident O2 molecule;
Winter 138 and Kilner 39 . From the kinetic theory of gases, the with its incident angle with respect to the surface normal; with
collisional flux is given by 139 its azimuthal angle with respect to some crystallographic surface
direction; and with its orientation with respect to the surface (the
coll aO2 p◦
jO2
=p (5) extremes being the main axis of the molecule being either head-
2πMO2 kB T
on or parallel to the surface). In addition, α may vary strongly
where aO2 is the activity of oxygen, p◦ the standard hydrostatic with the fractional coverage of the surface with adsorbates.
pressure of 1 bar, and MO2 is the mass of an oxygen molecule. By Let us consider for the sake of illustration that only those O2
coll equal to half the equilibrium exchange flux crossing
setting jO molecules whose kinetic energy lies above a certain critical value,
2
the solid surface, jO , and defining ks∗ in terms of jO , one can cal- E cr , contribute to jO and thus to ks∗ . (Mechanistically one may re-
culate limiting values of the surface exchange coefficient (ks,lim∗ )
coll ∗ ∗
from 2 jO2 = jO = ks,lim cO . Such values of ks,lim , it is emphasised,
constitute the ultimate limit to surface exchange, since it is im- c
Grain boundaries in La0.8 Sr0.2 MnO3+δ have been reported 140 to display k∗ = 10−2.5
possible for more atomic oxygen moieties to be incorporated into cm s−1 at T = 973 K, but also 141 only k∗ = 10−6 cm s−1 at T = 973 K. This discrepancy
has not been resolved as yet.

J
our
nal
Name,
[yea
r][
,vol
.
],1–17 | 5
Journal of Materials Chemistry A Page 6 of 17
View Article Online
DOI: 10.1039/C7TA04266C

gard this constraint simply as only those incident molecules with 2.2 Limit to k∗ from the surface reaction
a certain minimum kinetic energy being able to dissociate upon The second possibility for obtaining limiting values of the surface
collision). The fraction of molecules from jO coll that contribute to
2 exchange coefficient focusses on the kinetics of the chemical re-
jO (i.e. α) is obtained by integrating Maxwell’s distribution of gas action at the solid surface. In the ideal case, having identified the
molecules’ kinetic energies 139 from E cr to infinity, individual reaction steps, and in particular, the rate-determining

Journal of Materials Chemistry A Accepted Manuscript


Z∞  3/2   step, one would use the established theoretical framework of
1
Published on 11 September 2017. Downloaded by FLORIDA ATLANTIC UNIVERSITY on 12/09/2017 03:59:30.

E chemical kinetics 23,108,113 to derive an expression for ks∗ in terms


α= 2π E 1/2 exp − dE. (6)
πkB T kB T of atomistic quantities, analogous to eqn (2) for Dv . Subse-
Ecr
quently, experimental ks∗ data would be analysed to obtain val-
The surface exchange coefficient follows as ks∗ = ks,lim
∗ α. Data ob- ues for the constituent parameters of that expression, and having
tained in this way are plotted in Fig. 3. One sees that increasing established the extremum values, one would calculate limiting
E cr decreases the absolute value of ks∗ and increases ∆Hks∗ . For values of ks∗ .
example, with E cr = 1.1 eV, ks∗ ≈ 10−5 cm s−1 at T = 1000 K, and The main problem, as noted above, is that many mechanistic
the effective activation enthalpy of surface exchange is 1.03 eV. schemes and rate determining steps have been put forward in
This activation enthalpy, it is emphasised, is an effective value. the literature. Consequently, numerous expressions are in prin-
It simply reflects the distribution [eqn (6)] spreading to higher ciple available, and in many cases, one atomistic expression will
energies with increasing temperature, that is, it reflects the in- only be able to describe the experimental data for one specific
crease in the proportion of molecules with E > E cr with increasing material. This approach, proceeding from the mechanism, is not
temperature.d conducive to obtaining limiting values of ks∗ .
Limiting the range of molecular kinetic energies evidently I took, therefore, a different approach. My aim was to find an
brings both ks∗ and ∆Hks∗ closer to experimental values. It is pos- atomistic expression that describes ks∗ for as many systems as pos-
sible, therefore, that some experimental ks∗ data is correctly ex- sible (and not to identify the mechanism of surface exchange).
plained by this combination of a collisional flux and a sticking co- I was guided by the growing body of evidence suggesting that
efficient determined by Ecr , or more generally, by the combination electrons in the conduction band play a central role in the sur-
of a collisional flux and a strongly temperature-dependent stick- face exchange process, at least for semiconductors; 119,142–146 and
ing coefficient, α(T ). For complex processes such as surface ex- by the comprehensive treatment of the subject by Merkle and
change, however, definitive assignment of a particular mechanism Maier 109 , in which similar atomistic expressions are found if the
to experimental data is almost impossible because many possible conduction-band electron is transferred in or before the rds. This
mechanisms, including some not considered, will give rise to the is an advantage here, since one atomistic expression may thus
same observables (e.g., the absolute magnitude of ks∗ and its de- cover a variety of mechanistic cases. The expression I found takes
pendence on temperature). That is, agreement between a mech- the general form
anistic scheme and experimental data does not suffice to confirm
the scheme: it merely indicates consistency. Indeed, it is prefer- [e][sad ] −(∆G‡ +∆Gad )/kB T
ks∗ = dνrds
p
aO2 e rds , (7)
able to discount possible mechanisms, and this procedure can be [Ob ]
facilitated by examining several characteristic dependences, that
where d is an inter-atomic distance and νrds is the attempt fre-
is, not only the variation in ks∗ with temperature, but also with
quency of the rds; [e], [sad ] and [Ob ] are the site fractions of elec-
oxygen activity and with dopant concentration. For the case dis-
trons, of adsorption sites (of unknown identity), and of bulk oxy-
cussed above (a collisional flux and a limited range of kinetic en-
gen ions; and ∆G‡rds is the Gibbs activation energy of the rds and
ergies), the power-law exponent m describing the dependence on
∆Gad is the sum of the Gibbs energies of the preceding quasi-
oxygen activity, ks∗ ∝ (aO2 )m , has a characteristic value of m = 1,
equilibria.
according to eqn (5). And since the gas-phase flux determines the
In the following sections I consider a simple mechanistic
exchange rate, ks∗ is independent of all properties of the solid, that
scheme that puts some flesh on the bones on eqn (7) and thus
is, there is no dependence of the surface exchange coefficient on
aids the analysis of experimental data. It turns out that it is nec-
the acceptor-dopant concentration. These considerations do not
essary to differentiate between systems with a bandgap, such as
hold, of course, for all variants for which α ≤ 1. For example, if α
acceptor-doped SrTiO3 , and systems with a partially filled band,
depends on the degree of adsorbate coverage, other values of m,
such as acceptor-doped LaCoO3 . As such systems are typically
etc. are expected.
described as semiconductors (SC) and metallic conductors (MC),
these terms are used here, too, but purely as shorthand to de-
scribe the electronic band structures (the terms do not say any-
d
The alternative case—only those O2 molecules whose kinetic energy lies below a
thing about the respective conduction processes). In making a
certain critical value contribute to k∗ (because, say, only such slow molecules adsorb
successfully)—yields effective activation enthalpies that are negative (and small, distinction between SC and MC systems,e I do not claim that dif-
ca. − 0.1 eV). In this case, the spreading of the distribution results in the proportion
of low-energy molecules decreasing with increasing temperature.
e
Adler et al. 113 also made a distinction between electrons in SC and MC oxides.
e− -rich oxide, that different mechanisms of surface exchange are operative. (NB: e− -
They concluded that different mechanisms of surface exchange are operative, but
poor and e− -rich are not equivalent to SC and MC.) In contrast, Metlenko et al. 142
they did not consider the reactions discussed here (see Figs. 4 and 8).
considered a range of materials and found no evidence in favour of this claim.
Merkle and Maier 110 claimed, based on results for one e− -poor oxide and for one

6| J
our
nal
Name,
[yea
r][
,vol
.
],
1–17
Page 7 of 17 Journal of Materials Chemistry A
View Article Online
DOI: 10.1039/C7TA04266C

ferent mechanisms are operative, only that the conduction-band


(CB) electrons need to be treated differently in the two cases.
2.2.1 Semiconducting systems: the theory.
The simplest mechanistic scheme that is consistent with eqn (7)
consists of three reaction steps: fast dissociative adsorption is fol-

Journal of Materials Chemistry A Accepted Manuscript


lowed by a rate determining step of charge transfer, which is turn
Published on 11 September 2017. Downloaded by FLORIDA ATLANTIC UNIVERSITY on 12/09/2017 03:59:30.

is followed by a fast step combining incorporation and charge


transfer:
1
O + sad
2 2
Oad (fast) (8a)

Fig. 4 Schematic diagrams of the electronic band structure of an


Oad + e0 → O0ad (rds) (8b) acceptor-doped, semiconducting oxide at the surface (flatband case),
showing two different possibilities for the Gibbs activation energy for
O0ad + V··O sad + O×
O +h
·
(fast) (8c) electron transfer, ∆G‡t : (a) electron transfer to the energy level of the
adsorbate is from an energy ∆G‡t above the conduction-band edge; (b)
Three points need to be stressed. First, the rds is written specif- dynamic rearrangement with activation barrier ∆G‡t is required to
ically with electrons in the conduction band, and not with holes achieve a configuration for electron transfer. ECB , energy of the
from the valence band. CB electrons are minority species in an conduction-band edge; EVB , energy of the valence-band edge; EF ,
Fermi level, Ead , energy level of the adsorbate.
acceptor-doped oxide under oxidising conditions. Second, this
specific mechanistic scheme is unlikely to be correct because it re-
quires dissociation of the oxygen molecule without electron trans- strate to adsorbate (dt ). Furthermore, it is assumed that the for-
fer and because it does not contain any (even shortlived) peroxo ‡
ward reaction rate constant is given by kft = ν0 e−∆Gt /kB T , that is,
or superoxo intermediates (O02,ad , O002,ad ). It has been chosen be- it is characterised by an attempt frequency, ν0 , and a Gibbs ac-
cause it is relatively simple to analyse (see Sec. 2.2.5). Third, tivation energy for electron transfer, ∆G‡t . This energy can cor-
the mechanism is not consistent with DFT calculations, 118,147–152 respond (see Fig. 4) either to the difference in energies between
which do not find electron transfer to be the rds, but this is not the conduction-band edge and the energy level of the adsorbate
a problem. Published DFT calculations, though providing much (Oad /O0ad ) or to an activation barrier (cf. Marcus theory) that has
information on the elementary steps of the surface reaction, refer to be overcome for the transfer of the electron from the conduc-
(a) to T = 0 K, and hence, there are no electrons above the Fermi tion band to the adsorbate level. 142 In this way, eqn (10) becomes
level; and (b) to the static case, and hence, there is no dynamic (with ∆G‡t = ∆Ht‡ − T ∆St‡ and ∆Gad = ∆Had − T ∆Sad )
rearrangement of ions or molecules producing configurations in
which electron transfer may take place (in the sense of Marcus [sad ][e0s ] (∆St‡ +∆Sad )/kB −(∆Ht‡ +∆Had )/kB T
ks∗ = dt ν0
p
aO2 e e . (11)
theory or polaron hopping). [Ob ]
The rate of reaction at equilibrium, ℜ0 , is given by the forward
For the discussions in the subsequent sections, it will be useful
rate of the rds at equilibrium (and it is of course equal to the rate
to recognise that eqn (11) gives the activation enthalpy of the
of the reverse reaction at equilibrium)
surface exchange coefficient, ∆Hks∗ , as
d[O0ad ]  
ℜ0 = = kft [Oad ][e0s ] = krt [O0ad ] , ∂ ln ks∗
 
(9)
dt ∆Hks∗ = −kB
∂ 1/T aO2
where [Xad ] is the areal site fraction of species X adsorbed on
the surface; [Xs ] is the volume site fraction of species X in the = ∆‡t H + ∆ad H
surface, i.e. in the first layer of the solid; and [Xb ], when used,
∂ ln[e0s ]
   
refers to the volume site fraction in the bulk, that is, far from the ∂ ln[Ob ]
f/r
− kB + kB ; (12)
surface. kt are the rate constants for the forward and reverse ∂ (1/T ) aO2 ∂ (1/T ) aO2
reactions of the rds, eqn (8b). Assuming that the (fast) preceding
and the power-law exponent m, characterising the dependence of
step, eqn (8a), is in quasi-equilibrium, one can combine its equi-
−1/2 ks∗ on aO2 , as
librium constant, Kad = e∆Gad /kB T = [Oad ]aO2 [sad ]−1 , with eqn
∂ ln ks∗
 
(9) to eliminate [Oad ]. One thus obtains
m=
  ∂ ln aO2 T
−∆Gad
ℜ0 = kft [sad ][e0s ] aO2 exp
p
. (10)
kB T ∂ ln[e0s ]
   
1 ∂ ln[Ob ]
= + − . (13)
2 ∂ ln aO2 ∂ ln aO2
To transform eqn (10) into an expression for ks∗ , the chemi- T T

cal reaction rate is related 23 to the surface exchange coefficient The presence of a surface space-charge layer, 46,158,159 it is em-
through ks∗ = ℜ0 ∆x/[Ob ], where ∆x is a characteristic length. Here, phasised, will result in [e0s ] 6= [e0b ]; it will thus introduce extra terms
with the focus on the transfer of CB electrons, it is assumed that into eqn (12) and may introduce an extra term into eqn (13). This
∆x is the distance over which the electron is transferred from sub- will be illustrated with an example in the next section.

J
our
nal
Name,
[yea
r][
,vol
.
],1–17 | 7
Journal of Materials Chemistry A Page 8 of 17
View Article Online
DOI: 10.1039/C7TA04266C

Table 1 Comparison of experimentally measured parameters characterising oxygen surface exchange on bandgap semiconducting oxides with the
theoretical description of eqn (11)

expt
System ∆Hktheor

s
/ eV ∆Hk∗ / eV mtheor mexpt Ttrsn / K Surface Refs.
s
SrTiO3 , [Al]= 3 · 10−5 3.13 3.0 ± 0.3 0.25 0.23 ± 0.11 < 950 [e0s ] 6= [e0b ] Expt. 46 Def. chem. 153
CeO2 , [Gd]= 0.1 3.28 3.8 ± 0.9 0.25 0.23 ± 0.03 ≈ 1000 [e0s ] = [e0b ] Expt. 41,130 Def. chem. 154

Journal of Materials Chemistry A Accepted Manuscript


LaFeO3 , [Sr]= 0.4 1.70 1.77 ± 0.07 0.40 0.39 ± 0.04 ≈ 830 [e0s ] = [e0b ] Expt. 155,156 Def. chem. 157
Published on 11 September 2017. Downloaded by FLORIDA ATLANTIC UNIVERSITY on 12/09/2017 03:59:30.

2.2.2 Semiconducting systems: the application.


! "# ! "# $
Examination of eqn (11) indicates that values for the pre-
exponential parameters are relatively easy to obtain or estimate:
dt can be approximated as the cation–anion distance in the oxide;
ν0 is a characteristic lattice frequency; [sad ] can be taken to be
close to unity, which is probably reasonable for sufficiently high
temperatures; 160 [e0b ] and [Ob ] can be calculated by defect chem-
ical modelling with knowledge of the relevant equilibrium con-
stants, and subsequently one can either assume [e0s ] = [e0b ] or, if a
surface space-charge is present, calculate [e0s ] by thermodynamic
modelling with knowledge of the thermodynamic driving ener-
gies for space-charge formation and bulk defect concentrations.
This is relatively easy for dilute solid solutions, 85,161–163 less so
for concentrated solid solutions. 159,164 For the parameters in the
exponential terms, the sum of the entropies, (∆St‡ + ∆Sad ), and the
sum of the enthalpies, (∆Ht‡ +∆Had ), however, little is known, and
they are used, therefore, in the following as variables.
In order to extract these two variables, we require that re- Fig. 5 Surface exchange coefficients for weakly Al-doped SrTiO3 :
symbols, experimental data; 46 solid line, theoretical description with
liable experimental data for ks∗ is available as a function of T ;
eqns (11) and (14) with Φ0 (T ); dashed line, theoretical description with
and that the defect chemistry of the material has been elucidated eqn (11) with Φ0 = 0.
and the constants of the defect equilibria have been determined.
Critical inspection of the literature yields three suitable systems
(see Table 1): strongly doped CeO2 ; 41,130,154 strongly doped mation. 46 Further details and Φ0 (T ) are given in Appendix B.
LaFeO3 ; 155–157 and weakly doped SrTiO3 . 46,153 The comparisons Importantly, due to eqn (14) and with Φ0 varying with temper-
of experimental data and theoretical descriptions as a function of ature, the penultimate term in eqn (12) becomes
temperature for these three systems are shown in Figs. 5–7. In
∂ ln[e0s ] ∂ ln[e0b ]
   
each case, the emphasis was on describing the data at higher tem- −kB = − kB
peratures (for which [sad ] ≈ 1 can be assumed). The main result is ∂ (1/T ) aO2 ∂ (1/T ) aO2
that the atomistic expression of eqn (11) can describe ∆Hks∗ essen-  
e ∂ Φ0
tially with only one variable, the sum of the enthalpies, since the − − eΦ0 , (15)
T ∂ (1/T ) aO2
sum of the entropies turns out to be constant for all three systems.
Note: as summarised in Table 1, (11) can describe the power-law that is, two additional terms enter eqn (12). No additional term
exponent, m, for these three systems, too. is necessary for eqn (13), as in this specific case (Al-doped SrTiO3
We start with SrTiO3 because a surface space-charge layer with under oxidising conditions), Φ0 does not vary with oxygen activ-
a positive space-charge potential (Φ0 > 0) has been identified 46 ity.
(see also Ref. [153] for independent supporting evidence). Con- The comparison of experimental data with the theoretical de-
sequently the site fraction of electrons at the very surface is en- scriptions with eqn (11) are shown for SrTiO3 in Fig. 5. The effect
hanced over the site fraction in the bulk, of the positive space-charge potential on ks∗ as a function of T can
  be perceived by setting Φ0 = 0. Not only is ks∗ substantially dimin-
eΦ0
[e0s ] = [e0b ] exp . (14) ished, the activation enthalpy is also increased.
kB T
For Ce0.9 Gd0.1 O1.95 and the two LaFeO3 -based compounds,
This equation is valid as long as (i) electrons at the surface and La0.6 Sr0.4 Fe0.8 Co0.2 O3 and La0.6 Sr0.4 FeO3 , it was assumed that
in the bulk are in electrochemical equilibrium, (ii) there is no [e0s ] = [e0b ] in eqn (11). As a consequence, any deviations from this
conduction-band offset at the surface; 85 and (iii) [e0s ]  1. Φ0 , assumption because of a surface space-charge layer result in the
itself, was calculated, as noted above, by thermodynamic mod- effects of the space-charge layer being automatically subsumed
elling. 85,161–163 The segregation of oxygen vacancies to the sur- into the sum of the activation enthalpies.
face, with a segregation energy of ∆Gseg,v = −1.4 eV, was taken to For both systems, the experimental data in Fig. 6 and in Fig. 7
be the sole thermodynamic driving energy for space-charge for- are described well by eqn (11) at higher temperatures. Below a

8| J
our
nal
Name,
[yea
r][
,vol
.
],
1–17
Page 9 of 17 Journal of Materials Chemistry A
View Article Online
DOI: 10.1039/C7TA04266C

! !
" #$ " #$ %

Journal of Materials Chemistry A Accepted Manuscript


Published on 11 September 2017. Downloaded by FLORIDA ATLANTIC UNIVERSITY on 12/09/2017 03:59:30.

Fig. 8 Schematic diagrams of the electronic band structure at the


surface of an acceptor-doped oxide with a partially filled conduction
band, showing two different possibilities for the Gibbs activation energy
for electron transfer, ∆G‡t : (a) electron transfer to the energy level of the
adsorbate is from an energy ∆G‡t above the Fermi level; (b) dynamic
rearrangement with activation barrier ∆G‡t is required to achieve a
configuration for electron transfer. ECB , energy of the conduction-band
Fig. 6 Surface exchange coefficients for Ce0.9 Gd0.1 O1.9 : symbols, edge; EF , Fermi level, Ead , energy level of the adsorbate.
experimental data; 41 solid line, theoretical description with eqn (11).

# #
$ %& $ %& '
(8)] remains the same, f and charge transfer remains the rate-
determining step. Consequently, the atomistic expression for ks∗
keeps essentially the form of eqn (11), but [e0s ] has to be replaced
by the electron occupancy at the Fermi level, nb . In addition, the
Gibbs activation energy for electron transfer, ∆G‡t , refers now (see
Fig. 8) either to the adsorbate level (Oad /O− ad ) lying well above the
Fermi level (εF ) or to an activation barrier that has to be overcome
!"
for electron transfer from the Fermi level.g

As the Fermi level is not a constant but shifts up or down with


compositional changes, the Gibbs activation energy for electron
transfer varies, too,

∆G‡t = ∆G‡,0 0
t − (εF − εF ), (16)

where ∆G‡,0t is the activation energy at Fermi level εF0 . For


an oxide with a partially filled band, such as perovskite-type
Fig. 7 Surface exchange coefficients for two LaFeO3 -based La1−a Sra CoO3−δ , the shift in Fermi level, (εF − εF0 ), can be related
compounds: circles, La0.6 Sr0.4 Fe0.8 Co0.2 O3-δ (expt,I); 155 diamonds,
in the Lankhorst formalism 168 to the electron occupancy through,
La0.6 Sr0.4 FeO3-δ (expt,II); 156 line, theoretical description with eqn (11).
nb = n0b + fn (εF − εF0 ), (17)

certain temperature, Ttrsn , there is a positive deviation of the data where n0b is the electron occupancy at Fermi level εF0 , and fn is the
away from the prediction. While this change in activation en- density of states at the Fermi level. Furthermore, the electroneu-
thalpy is rather obvious (and well-known 41 ) for Ce0.9 Gd0.1 O1.95 , trality condition,
a single activation enthalpy of (1.10 ± 0.14) eV for the entire set nb = n0b + 2δ − a, (18)
of data was reported previously for La0.6 Sr0.4 Fe0.8 Co0.2 O3 . It is allows the electron occupancy to be related to the changes in com-
only through consideration of additional data that a decrease in position, i.e., acceptor-dopant fraction a and oxygen deficiency δ .
activation enthalpy, from (1.77 ± 0.07) eV to (0.62 ± 0.44) eV, be- In this way, the shift in Fermi level can be linked to changes in
comes evident. No Ttrsn is evident for SrTiO3 , but this is probably composition.
due to Ttrsn being lower than the range of experimental data.
2.2.3 Metallically conducting systems: the theory.
f
In eqn (8), e0 is of course replaced by e− ; h· is replaced with −e− .
We now shift our attention to an oxide with a partially filled g
Strictly speaking, the electron occupancy at energy ∆G‡t above the Fermi level
band, such as a perovskite-type oxide La1−a Sra CoO3−δ . Never- should be obtained from a Fermi–Dirac distribution, but it can be approximated by
theless, the mechanistic scheme of oxygen surface exchange [eqn a simple Maxwell–Boltzmann expression, as long as ∆G‡t  kB T .

J
our
nal
Name,
[yea
r][
,vol
.
],1–17 | 9
Journal of Materials Chemistry A Page 10 of 17
View Article Online
DOI: 10.1039/C7TA04266C

# # # #
$ %& $ %& ' $ %& $ %& '

% (

Journal of Materials Chemistry A Accepted Manuscript


Published on 11 September 2017. Downloaded by FLORIDA ATLANTIC UNIVERSITY on 12/09/2017 03:59:30.

!"

!"

Fig. 9 Surface exchange coefficients for strongly doped LaCoO3 oxides: symbols, experimental data; lines, theoretical descriptions with eqn (19). (a)
La0.6 Sr0.4 CoO3-δ (circles: expt,I; 165 squares: expt,II 166 ) and Sm0.6 Sr0.4 CoO3-δ (diamonds: expt,III 167 ). (b) La0.5 Sr0.5 CoO3-δ (circles: expt,I; 133
squares: expt,II; 166 triangles: expt,III 122 ) and Sm0.5 Sr0.5 CoO3-δ (diamonds: expt,IV 167 ).

Combining eqns (16) to (18), one thus obtains,


!
[s ] p −∆Gt‡,0 − ∆Gad
ks∗ = dt ν0 ad aO2 exp (19)
[Ob ] kB T
 
2δ − a
(n0b + 2δ − a) exp .
fn kB T

The effects of a space-charge layer on ks∗ are not considered !" #$


% & '
here, but in general, an approach suitable for a concentrated so-
lution, e.g. Poisson–Cahn, 159 is necessary.

2.2.4 Metallically conducting systems: the application.

In order to analyse experimental data with eqn (19), one


requires reliable nonstoichiometry data, δ (T ), in addition to
reliable data for k∗ (T ). Three compositions satisfy the require-
ments: La0.6 Sr0.4 CoO3-δ ; 165,166,169 La0.5 Sr0.5 CoO3-δ ; 133,166,170
and (Ba0.5 Sr0.5 )(Co0.8 Fe0.2 )O3−δ . 48,83,171 For the two
La1-a Sra CoO3-δ compositions, data from Fullarton et al. 167 Fig. 10 Surface exchange coefficients for (Ba0.5 Sr0.5 )(Co0.8 Fe0.2 )O3−δ :
for the equivalent Sm1-a Sra CoO3-δ compositions are also circles (expt,I 48 ), measured experimentally by tracer exchange;
diamonds (expt, II 171 ), converted from chemical relaxation data with
included. The results are shown in Fig. 9 for the two LSC revised thermodynamic factor; 83 solid line, theoretical description with
compositions and in Fig. 10 for the BSCF composition. It is eqn (19).
striking that, together with n0b = 1 and fn = 1.88 eV−1 , values of
(∆Ht‡ + ∆Had ) = 1.46 eV and of (∆St‡ + ∆Sad ) = −4 kB can describe
the data for all three systems reasonably well. The only quantity theoretical description. Taken together, the two datasets re-
that is specific to each system is the oxygen nonstoichiometry, veal a decrease in the activation enthalpy of surface exchange
through 2δ (T ) − a and [Ob ]{= 1 − δ (T )/3}. at Ttrsn ≈ 800 K; this change was not previously apparent. In
The assumption of a constant fn for all three systems is a the case of La0.5 Sr0.5 CoO3-δ [Fig. 9(b)], the agreement between
zeroeth-order approximation. fn is known to vary slightly with the three datasets 48,133,166 is significantly poorer, although the
acceptor-dopant level. 168,169,172 Furthermore, a better descrip- theoretical description agrees very well with the chemical dif-
tion of δ (T ) is possible 172 by making fn a function of (εF − εF0 ). fusion data of Egger et al. 166 . An experimental re-examination
These higher-order corrections are left for future study. of the tracer data for this composition 133 is perhaps warranted.
For La0.6 Sr0.4 CoO3-δ [Fig. 9(a)] there is excellent agreement, In addition, the data points obtained from tracer diffusion stud-
not only between data from tracer 165 and from chemical 166 dif- ies (expt,I 133 and expt,IV 167 ) and for the chemical diffusion
fusion studies, but also between the experimental data and the (expt,III 48 ) do suggest a different activation enthalpy and there-

10 | J
our
nal
Name,
[yea
r][
,vol
.
],1–17
Page 11 of 17 Journal of Materials Chemistry A
View Article Online
DOI: 10.1039/C7TA04266C

fore, poorer agreement; however, one could also attribute the de- lated [eqn (8a)]: ∆Had corresponds to the dissociation of the O2
viation to a much higher Ttrsn , one that varies from study to study. molecule and the adsorption of an O atom without charge trans-
This in turn would mean that Ttrsn may vary not only between ma- fer (here, a useful, though clearly unreasonable, simplification);
terials but also between different experimental conditions (i.e. be- hence, with the high bond energy of the O2 molecule (> 5 eV)
tween measurements made in different studies). For BSCF (Fig. and without any strong chemical interaction between adsorbed

Journal of Materials Chemistry A Accepted Manuscript


10) the two sets of experimental data 48,171 are in good agree- species and the substrate, ∆Had is likely to be constant across
Published on 11 September 2017. Downloaded by FLORIDA ATLANTIC UNIVERSITY on 12/09/2017 03:59:30.

ment with one another, and also with the theoretical description. various oxides. In contrast, a non-zero space-charge potential, if
An even better theoretical description could possibly be attained present and not explicitly taken into account, will provide a con-
by determining fn specifically for this composition. tribution to the value of (∆Ht‡ + ∆Had ) extracted from the analysis
that varies strongly from oxide to oxide because the space-charge
2.2.5 Surface reaction: Summary and Discussion.
potential can in principle vary with composition, with surface ori-
There are two important points to emerge from this analysis. entation and termination, with dopant site fraction, with temper-
First, eqns (11) and (19) constitute atomistic expressions that ature and with oxygen activity. A significant variation between
quantitatively describe k∗ as a function of temperature, oxygen materials is also expected from ∆Ht‡ , but the adsorbate level has
activity and composition for selected acceptor-doped oxides with to lie well above the CB minimum (SC oxide: Ead > ECB ) or well
bandgaps and with partially filled bands, respectively. The de- above the Fermi level (MC oxide: Ead > EF0 ). If ∆Ht‡ refers to the
scriptions involve a constant value of (∆St‡ + ∆Sad ) = −4 kB and activation barrier for electron transfer and Ead ≤ ECB or Ead ≤ EF0 ,
a value for (∆Ht‡ + ∆Had ) that varies from material to material. one may expect low values for ∆Ht‡ and little variation. The ac-
Second, there is a decrease in the activation enthalpy of ks∗ in tivation enthalpy for electron polaron hopping in CeO2 , for in-
the temperature range 1000 > Ttrsn / K > 800 for a variety of ma- stance, is 0.4 eV, 173 and values much larger than this for electron
terials. Such behaviour was known previously 41 for the elec- transfer are difficult to imagine. In this interpretation, electron
trolytes Ce0.9 Gd0.1 O1.95 and Zr0.9 Y0.1 O1.95 , and it was identi- transfer, as illustrated in Fig. 4(b) and Fig. 8(b), can be ruled out.
fied here for the mixed conductors La0.6 Sr0.4 Fe0.8 Co0.2 O3 [Fig. One could furthermore postulate (i) that the energy level of
7] and La0.6 Sr0.4 CoO3-δ [Fig. 9(a)] compositions (for SrTiO3 and the adsorbate relative to the vacuum level (Evac − Ead ) is approx-
BSCF all experimental data may be above the respective transi- imately constant for all oxide systems; and hence, (ii) that the
tion temperatures). Although the data are scarce, it appears that differences in ∆Ht‡ between oxides arise, therefore, from varia-
Ttrsn varies with system, but also with experimental conditions. tions 174–177 in the electron affinity of SC oxides (Evac − ECB ) or
These two general results prompt four specific questions: the work function of MC oxides (Evac − EF0 ). In this way, the elec-
(1) Is a negative, constant value for (∆St‡ + ∆Sad ) reasonable? tronic band structure plays a central role in determining surface
(2) Why does (∆Ht‡ + ∆Had ) vary between systems? exchange kinetics. 178 Consistent with this hypothesis is the fact
(3) Why is there a change in the activation enthalpy of surface that the values of (∆Ht‡ + ∆Had ) obtained for SrTiO3 and CeO2 are
exchange? similar, and their electron affinities are probably similar, too: the
(4) What does the agreement between experimental data and valence-band levels, being comprised of O 2p states, 174,175 will
atomistic expressions indicate? be similar, and the bandgaps, at ∼ 3 eV, will be as well.
The answers to some of these questions are admittedly specula- It is worth noting that this hypothesis—a relationship between
tive, and they should be regarded, therefore, as suggested av- the electronic band structure of an oxide and its surface exchange
enues for further research. kinetics—can explain certain empirical correlations. It can ex-
1. The answer to the first question is a relatively clear and sim- plain why a correlation is observed between energy level of the
ple yes. The largest contribution to ∆Sad comes from the loss oxygen p-band center and surface exchange data for a wide vari-
of entropy of the gas-phase molecules upon dissociative adsorp- ety of acceptor-doped oxides. 120 In addition, it provides a differ-
tion: ∆Sad is expected, therefore, to take a large, negative value ent explanation of why k∗ may be correlated with oxygen-vacancy
(Appendix C details a very approximate treatment that yields concentration. By increasing the vacancy concentration in an MC
∆Sad = −5.5 kB ), and it may vary weakly from system to system. oxide, one raises the Fermi level, thereby increasing the rate at
∆St‡ , in contrast, will probably be much smaller in magnitude and which electrons are transferred to the adsorbate, and thus, ac-
it may be either positive or negative; it, too, may vary weakly cording to eqn (19), increasing k∗ . This hypothesis thus sup-
from system to system but not necessarily in the same manner as ports the view 134 that the empirical correlations 39,133,134 found
∆Sad . The sum of the two entropies (∆St‡ + ∆Sad ), therefore, will between k∗ and D∗ do not indicate a central role for oxygen va-
in all probability be negative, and it will probably take a constant cancies in the surface exchange process. Empirical correlations
value. are open to misinterpretation because correlation does not mean
2. It is tentatively suggested that the observed variation in causation.
(∆Ht‡ +∆Had ) between materials comes predominantly from varia- 3. The decrease in activation enthalpy may have its origins in a
tions in the activation enthalpy for charge transfer and from vari- change in rds within the same mechanistic scheme or in a paral-
ations in the surface space-charge potential (if not explicitly in- lel mechanistic scheme becoming more favourable. Because this
cluded), with the third possibility, a variation in the enthalpy of decrease is observed for two wide bandgap semiconductors, one
dissociative adsorption, being negligible. This suggestion follows narrow bandgap semiconductor and one metallic conductor—all
from the way that the reaction of dissociative adsorption is formu- in a relatively small range of temperatures—it is evidently not

J
our
nal
Name,
[yea
r][
,vol
.
],
1–17 | 11
Journal of Materials Chemistry A Page 12 of 17
View Article Online
DOI: 10.1039/C7TA04266C

due to the availability or the transfer of electrons or to properties 3 Concluding remarks


specific to the materials themselves. Rather, it seems to be gen-
Analysis of experimental literature data allowed a maximum
eral to oxide materials and perhaps also to the thermodynamic
value to be proposed for the tracer diffusion coefficient of oxygen
conditions.
in a solid oxide, and thus, for the oxygen-ion conductivity. Such
Consequently, the change in activation enthalpy is attributed, expressions suggest that superior oxygen-ion conductors may be

Journal of Materials Chemistry A Accepted Manuscript


as first suggested by Sakai et al. 179 , to the presence of trace waiting to be discovered. In addition, they suggest that com-
Published on 11 September 2017. Downloaded by FLORIDA ATLANTIC UNIVERSITY on 12/09/2017 03:59:30.

amounts of water vapour in the gases used in the O2 ex- position tuning should be considered in conjunction with pos-
change experiments. The incorporation of oxygen from gaseous sible enhancements from defect re-distribution, strain or strong
H2 O is known to be faster than from gaseous O2 . 42,50,179–183 electric fields. The limiting expressions may also prove useful
This behaviour has been argued to involve hydroxide adsor- when critically judging experimental data, especially conductiv-
bates, 42,180,181 and in this respect the role of charged surfaces ity data, which may contain contributions from electronic species.
with high hydroxide coverage (perhaps even for materials that For experimental data approaching or even exceeding this limit,
dissolve hardly any OH species in the bulk) may be impor- and hence well within the realm of electronic conductivities, one
tant. 162,184 Ttrsn will, accordingly, shift with varying aH2 O in the could perhaps stipulate a special burden of evidence.
annealing gases.
Elucidating the mechanism of surface exchange has not been
4. The fact that ks∗ (T, aO2 , ca ) data for a diverse selection of the aim of this study. For the best one can hope for is to show
acceptor-doped oxides can be described quantitatively with one consistency between experimental data and a particular mecha-
of two atomistic expressions using a single value for (∆St‡ + ∆Sad ) nism. Instead, atomistic expressions were derived expressly for
strongly suggests a common mechanism of oxygen surface ex- the purpose of describing the surface exchange coefficients of
change for all these materials, one in which CB electrons play acceptor-doped oxides with a bandgap or with a partially filled
a central role. This is, however, only a strong suggestion: as conduction band. These expressions are able to describe quan-
noted in Section 2.2.1, a variety of rate-determining steps may titatively surface exchange data as a function of thermodynamic
give the same atomistic expression. Also, it cannot be empha- variables above a (system-specific) transition temperature, and
sised enough that these quantitative descriptions do not indicate they accomplish this, essentially with only one ‘free’ parameter,
what the mechanism of surface exchange is or what its rate- the sum of two enthalpies. The decrease in activation enthalpy of
determining step is. surface exchange below the transition temperature is attributed
Nevertheless, the atomistic expressions are extremely useful. to a parallel reaction stemming from residual water vapour be-
They suggest ways for identifying acceptor-doped oxides with coming more favourable. Finally, it appears that greater consid-
high surface exchange coefficients or for tuning existing mate- eration should be given to the role played by the electronic band
rials. According to the analysis and discussion above, the concen- structure of an oxide at the surface in determining the rate of
tration of electrons in the conduction band of SC systems should oxygen surface exchange.
be maximised and their electron affinity should be minimised; for
MC systems, the term (2δ − a) fn−1 should be maximised and the
4 Acknowledgments
work function should be minimised. Of course, any changes in Financial support from the German Research Foundation (DFG) is
the surface (structure, chemistry, etc) that lead to the opposite ef- gratefully acknowledged: within the framework of the collabora-
fects will affect ks∗ adversely. A different way of expressing these tive research centre ‘Nanoswitches’ (SFB 917); and from project
optimisation rules is that, in addition to the appropriate electronic SO499/7-1. Discussions with A. Klein, D.S. Mebane and A. Lü-
band structure, the Fermi level at the surface should be as high chow and critical reading of the manuscript by D.N. Mueller, A.R.
as possible. Doping the surface may prove beneficial in tuning in- Genreith-Schriever, S.P. Waldow and J. Koettgen are gratefully
dividual oxides, either through donor doping or through doping acknowledged.
with cations that are more reducible. Raising the Fermi level at
the surface may also be achieved through the creation of suitable Appendix A: The activation entropy of migration
heterostructure junctions. Indeed, such an effect may be respon- Various simple methods to obtain the activation entropy of mi-
sible (primarily or in addition to other effects 185–189 ) for the en- gration have been discussed in the literature. 34,78 Here, a new
hanced k∗ at (La,Sr)CoO3 /(La,Sr)2 CoO4 interfaces first observed method, based on previous work, is presented. In addition, the
by Sase et al. 190 ,191 simplest possible method is examined. It will be shown that both
methods are insufficiently exact to allow a maximum value to be
In this sense, it would be tempting to derive limiting values of
identified.
ks∗ for the two cases (SC and MC), but: with only three systems to
consider in each case; with no quantitative model for the variation The simplest method was suggested by Dienes 77 :
in (∆Ht‡ + ∆Had ) between various materials; and with a possible ‡
∆Hmig,v

surface space-charge layer ignored in all but one case; there is ∆Smig,v = , (20)
Tmelt
insufficient information and too much uncertainty. In any case,
one should bear in mind that improving the reaction rate at the that is, the activation entropy of migration is the quotient of the
surface may lead to a different step becoming rate determining, activation enthalpy of migration and the melting temperature.
and that the ultimate limit (see Sec. 2.1) is ks∗ ≤ 100 cm s−1 . Values obtained with this expression will, of course, always be

12 | J
our
nal
Name,
[yea
r][
,vol
.
],1–17
Page 13 of 17 Journal of Materials Chemistry A
View Article Online
DOI: 10.1039/C7TA04266C

positive. For fluorite-structured CeO2 and perovskite-structured ∆Gsurf bulk


form,def − ∆Gform,def ), for that surface, Φ0 (T, aO2 , ca ) can be ob-
‡ ‡
SrTiO3 , we obtain ∆Smig,v = +4kB and ∆Smig,v = +3.3kB , respec- tained. Here, Φ0 (T, aO2 , ca ) was calculated for acceptor-doped
tively. This approach is useless, however, for estimating a limiting SrTiO3 assuming electrochemical equilibrium for oxygen va-
value, as the melting temperature is that of a hypothetical com- cancies, electrons and electron holes; assuming an immobile
pound exhibiting the extremum values. acceptor-dopant cation whose concentration is constant through-

Journal of Materials Chemistry A Accepted Manuscript


For the new method, we start with Einstein’s model of inde- out the system; and assuming the segregation of oxygen vacancies
Published on 11 September 2017. Downloaded by FLORIDA ATLANTIC UNIVERSITY on 12/09/2017 03:59:30.

pendent oscillators, and hence assume that only the migrating to the surface as the sole driving energy for the formation of the
ion’s entropy changes during the jump. The activation entropy of space charge zone. The results are plotted in Fig. 11.
oxygen-vacancy migration is thus defined as

∆Smig,v = STS IS
ion − Sion , (21)

that is, as the difference in the entropy of the migrating ion be-
tween its transition state (TS) and its initial state (IS) (i.e. next
to a vacancy). We now find approximations for these two terms:
the latter is approximated as the entropy of an oxygen ion in the
lattice, Svib (O2− ), (no neighbouring vacancy); and the former, as
two thirds of this entropy (due to the lost vibrational mode in the
direction of migration). We thus obtain

‡ 1
∆Smig,v ≈ − Svib (O2− ). (22)
3

Since Svib (O2− ) is an absolute entropy, and thus always positive,



values of ∆Smig,v calculated in this way will always be negative.
Fig. 11 Space-charge potential Φ0 at the (100) surface of
To find Svib (O2− ), we assume a cosinusoidal enthalpy profile acceptor-doped SrTiO3 as a function of temperature T predicted by
for the migrating ion 93 (with r being the spatial coordinate along thermodynamic modelling.
the migration path, 0 ≤ r ≤ dv )
‡     One should note that there are two effects that a positive sur-
∆Hmig,v 2πr
H(r) = cos +1 , (23) face potential will have on the process of oxygen exchange. The
2 dv
first effect is the increase in ks∗ arising from electron accumula-
and truncate a Taylor expansion around the minimum at r = dv /2 tion [eqn (14)]. The second effect — the depletion of oxygen
after the quadratic term to get vacancies in the space-charge layer —- introduces an additional
hindrance to oxygen exchange. In general, unless this hindrance

∆Hmig,v π2 is resolved experimentally [and to date this has only been accom-
H(r) = (2r − dv )2 . (24)
4 dv2 plished for SrTiO3 , 46,158,193 BaTiO3 194 and Pb(Zr,Ti)O3 195,196 ],
the surface exchange coefficient that is obtained experimentally
Executing simple harmonic motion in this parabolic potential, refers to the surface reaction and transport through the space-
the oxygen ion has a characteristic frequency, charge zone. That is, an effective surface exchange coefficient is
v obtained: (keff∗ )−1 = (k∗ )−1 + (k∗ )−1 . The transfer coefficient for
s scl
u ∆H ‡
u
t mig,v oxygen chemical diffusion through a space-charge layer has been
ν= , (25)
2MO dv2 derived by Leonhardt et al. 197 The equivalent expression for oxy-
gen tracer diffusion can be shown to be
and hence, an entropy, 192 (with x = hν/kB T )
Z lscl
  −1
∗ 2e[φ (x) − φ (∞)]
kscl = D∗O exp dx , (27)
 
x
Svib (O2− ) = 3kB x − ln(1 − e−x ) . (26) 0 kB T
e −1
where φ (x) − φ (∞) is the electrostatic potential distribution in the

In this way, we arrive at ∆Smig,v = −(2.4 ± 0.2)kB for the oxides space-charge layer and lscl the extent of the space-charge layer.
of Fig. 2 at T = 1000 K. In comparison with the data of Fig. 2,
these values are all lower and they show only a slight increase Appendix C: Entropy of O2 Adsorption

with increasing ∆Hmig,v . The entropy of the reaction of dissociative adsorption, eqn (8a),

Appendix B: Surface space-charge potential 1


∆Sad = S(Oad ) − S(O2,g ), (28)
2
In general, the space-charge potential, Φ0 , may vary with
temperature, oxygen activity, and acceptor-dopant concentration. is approximated here by
With knowledge of bulk defect concentrations and the thermo-
1
dynamic driving energies for space-charge formation, ∆Gseg,def (= ∆Sad ≈ Svib (O2− ) − S(O2,g ). (29)
2

J
our
nal
Name,
[yea
r][
,vol
.
],
1–17 | 13
Journal of Materials Chemistry A Page 14 of 17
View Article Online
DOI: 10.1039/C7TA04266C

Svib (O2− ) was obtained approximately in eqn (26). The entropy Energy, 2016, 1, 15014.
of molecular oxygen for the temperature range of interest can be 19 W. C. Chueh and S. M. Haile, Philos. Trans. R. Soc., A, 2010,
calculated 192 by adding together the entropy contributions from 368, 3269–3294.
the translation, rotation, vibration and electronic excitation of the 20 J. R. Scheffe and A. Steinfeld, Mater. Today, 2014, 17, 341–
O2 molecules: 348.

Journal of Materials Chemistry A Accepted Manuscript


21 R. A. De Souza, J. Fleig, R. Merkle and J. Maier, Z. Metallkd.,
S(O2,g ) = Strans + Srot + Svib + Selec , (30)
Published on 11 September 2017. Downloaded by FLORIDA ATLANTIC UNIVERSITY on 12/09/2017 03:59:30.

2003, 94, 218–225.


with 22 R. Merkle and J. Maier, Angew. Chem., Int. Ed., 2008, 47,
" # 3874–3894.
e5/2 kB T

2πMO2 kB T 23 J. Maier, Solid State Ionics, 1998, 112, 197–228.
Strans = kB ln
h2 p
24 R. A. De Souza, Adv. Funct. Mater., 2015, 25, 6326–6342.
T
  25 R. A. De Souza and M. Martin, MRS Bulletin, 2009, 34, 907–
Srot = kB ln 914.
2Θrot
26 J. A. Kilner, S. J. Skinner and H. H. Brongersma, J. Solid
 o
(T /Θvib ) n
−(T/Θvib ) State Electrochem., 2011, 15, 861–876.
Svib = kB − ln 1 − e
e(T /Θvib ) − 1 27 P. Fielitz and G. Borchardt, Solid State Ionics, 2001, 144,
71–80.
Selec = kB ln(gelec )
28 R. A. De Souza and R. J. Chater, Solid State Ionics, 2005,
where Θrot = 2.08 K, Θvib = 2239 K and gelec = 3. Since the largest 176, 1915–1920.
contribution to S(O2,g ) comes from the first two terms—the trans- 29 J. Crawford and P. Jacobs, J. Solid State Chem., 1999, 144,
lational and rotational degrees of freedom—and these are lost 423–429.
upon dissociative adsorption, ∆Sad clearly will be negative. From 30 A. Samanta, W. E and S. B. Zhang, Phys. Rev. B, 2012, 86,
eqn (29) and with the results of Appendix A, one thus obtains, for 195107.
T = 1000 K, ∆Sad ≈ (9.1 − 29.3/2) kB = −5.5 kB . 31 M. Nakayama and M. Martin, Phys. Chem. Chem. Phys.,
2009, 11, 3241–3249.
References 32 T. Zacherle, A. Schriever, R. A. De Souza and M. Martin,
Phys. Rev. B, 2013, 87, 134104.
1 B. C. H. Steele, J. Power Sources, 1994, 49, 1–14.
33 N. Uvarov and E. Hairetdinov, J. Solid State Chem., 1986, 62,
2 B. C. H. Steele, Solid State Ionics, 1995, 75, 157–165.
1–10.
3 H. L. Tuller, Solid State Ionics, 1992, 52, 135–146.
34 D. Almond and A. West, Solid State Ionics, 1987, 23, 27–35.
4 H. J. M. Bouwmeester, H. Kruidhof and A. J. Burggraaf, Solid
35 A. Nowick, W.-K. Lee and H. Jain, Solid State Ionics, 1988,
State Ionics, 1994, 72, Part 2, 185–194.
28, 89–94.
5 J. A. Kilner, R. A. De Souza and I. C. Fullarton, Solid State
36 A. Berenov, J. MacManus-Driscoll and J. Kilner, Int. J. Inorg.
Ionics, 1996, 86-88, Part 2, 703–709.
Mater., 2001, 3, 1109–1111.
6 J. Mizusaki, Solid State Ionics, 1992, 52, 79–91.
37 T. Ishigaki, S. Yamauchi, K. Kishio, J. Mizusaki and K. Fueki,
7 H. Inaba and H. Tagawa, Solid State Ionics, 1996, 83, 1–16.
J. Solid State Chem., 1988, 73, 179–187.
8 J. C. Boivin and G. Mairesse, Chem. Mater., 1998, 10, 2870–
38 J. Kilner and R. Brook, Solid State Ionics, 1982, 6, 237–252.
2888.
39 J. A. Kilner, Proc. Second Intl. Symp. Ionic and Mixed Con-
9 J. Kilner, Solid State Ionics, 2000, 129, 13–23.
ducting Ceramics, Pennington, NJ, USA, 1994, pp. 174–190.
10 A. Orera and P. R. Slater, Chem. Mater., 2010, 22, 675–690.
40 J. A. Kilner, A. Berenov and J. Rossiny, in Diffusivity of the
11 T. Norby, J. Mater. Chem., 2001, 11, 11–18. Oxide Ion in Perovskite Oxides, ed. T. Ishihara, Springer US,
12 J. Sunarso, S. Baumann, J. Serra, W. Meulenberg, S. Liu, Boston, MA, 2009, pp. 95–116.
Y. Lin and J. D. da Costa, J. Membr. Sci., 2008, 320, 13–41. 41 P. S. Manning, J. D. Sirman and J. A. Kilner, Solid State Ion-
13 J. B. Goodenough, Ann. Rev. Mater. Res., 2003, 33, 91–128. ics, 1996, 93, 125–132.
14 E. Tsipis and V. Kharton, J. Solid State Electrochem., 2008, 42 M. J. Pietrowski, R. A. De Souza, M. Fartmann, R. ter Veen
12, 1039–1060. and M. Martin, Fuel Cells, 2013, 13, 673–681.
15 D. J. L. Brett, A. Atkinson, N. P. Brandon and S. J. Skinner, 43 R. D. Bayliss, S. N. Cook, S. Kotsantonis, R. J. Chater and
Chem. Soc. Rev., 2008, 37, 1568–1578. J. A. Kilner, Adv. Energy Mater., 2014, 4, 1301575.
16 L. Malavasi, C. A. J. Fisher and M. S. Islam, Chem. Soc. Rev., 44 R. Sayers, R. A. De Souza, J. A. Kilner and S. J. Skinner, Solid
2010, 39, 4370–4387. State Ionics, 2010, 181, 386–391.
17 J. A. Kilner and M. Burriel, Ann. Rev. Mater. Res., 2014, 44, 45 M. Teusner, R. A. De Souza, H. Krause, S. G. Ebbinghaus,
365–393. B. Belghoul and M. Martin, J. Phys. Chem. C, 2015, 119,
18 J. T. S. Irvine, D. Neagu, M. C. Verbraeken, 9721–9727.
C. Chatzichristodoulou, C. Graves and M. B. Mogensen, Nat.

14 | J
our
nal
Name,
[yea
r][
,vol
.
],1–17
Page 15 of 17 Journal of Materials Chemistry A
View Article Online
DOI: 10.1039/C7TA04266C

46 R. A. De Souza, V. Metlenko, D. Park and T. E. Weirich, Phys. 296, 71–77.


Rev. B, 2012, 85, 174109. 74 M. Schie, M. P. Müller, M. Salinga, R. Waser and R. A. De
47 T. Ishihara, J. A. Kilner, M. Honda, N. Sakai, H. Yokokawa Souza, J. Chem. Phys., 2017, 146, 094508.
and Y. Takita, Solid State Ionics, 1998, 113-115, 593–600. 75 L. Zhang, B. Liu, H. Zhuang, P. Kent, V. R. Cooper, P. Ganesh
48 L. Wang, R. Merkle, J. Maier, T. Acartürk and U. Starke, Appl. and H. Xu, Comput. Mater. Sci., 2016, 118, 309–315.

Journal of Materials Chemistry A Accepted Manuscript


Phys. Lett., 2009, 94, 071908. 76 S. P. Waldow and R. A. De Souza, ACS Appl. Mater. Interfaces,
Published on 11 September 2017. Downloaded by FLORIDA ATLANTIC UNIVERSITY on 12/09/2017 03:59:30.

49 F. Abraham, J. Boivin, G. Mairesse and G. Nowogrocki, Solid 2016, 8, 12246–12256.


State Ionics, 1990, 40, 934–937. 77 G. J. Dienes, J. Appl. Phys., 1950, 21, 1189–1192.
50 R. Vannier, S. Skinner, R. Chater, J. Kilner and G. Mairesse, 78 T. W. Dobson, J. F. Wager and J. A. Van Vechten, Phys. Rev.
Solid State Ionics, 2003, 160, 85–92. B, 1989, 40, 2962–2967.
51 P. Lacorre, F. Goutenoire, O. Bohnke, R. Retoux and Y. Lali- 79 M. Mantina, Y. Wang, R. Arroyave, S. L. Shang, L. Q. Chen
gant, Nature, 2000, 404, 856–858. and Z. K. Liu, J. Phys.: Condens. Matter, 2012, 24, 305402.
52 S. Georges, S. J. Skinner, P. Lacorre and M. C. Steil, Dalton 80 J. Koettgen, T. Zacherle, S. Grieshammer and M. Martin,
Trans., 2004, 3101–3105. Phys. Chem. Chem. Phys., 2017, 19, 9957–9973.
53 H. Arikawa, H. Nishiguchi, T. Ishihara and Y. Takita, Solid 81 P. Shuk, H.-D. Wiemhöfer, U. Guth, W. Göpel and M. Green-
State Ionics, 2000, 136-137, 31–37. blatt, Solid State Ionics, 1996, 89, 179–196.
54 E. J. Abram, D. C. Sinclair and A. R. West, J. Mater. Chem., 82 D. N. Mueller, R. A. De Souza, T. E. Weirich, D. Roehrens,
2001, 11, 1978–1979. J. Mayer and M. Martin, Phys. Chem. Chem. Phys., 2010, 12,
55 A. R. Genreith-Schriever, P. Hebbeker, J. Hinterberg, 10320–10328.
T. Zacherle and R. A. De Souza, J. Phys. Chem. C, 2015, 119, 83 D. N. Mueller, R. A. De Souza, H.-I. Yoo and M. Martin,
28269–28275. Chem. Mater., 2012, 24, 269–274.
56 M. Meyer and N. Nicoloso, Ber. Bunsen-Ges., 1997, 101, 84 H. L. Tuller, J. Phys. Chem. Solids, 1994, 55, 1393–1404.
1393–1398. 85 R. A. De Souza, Phys. Chem. Chem. Phys., 2009, 11, 9939–
57 M. Martin, J. Electroceram., 2006, 17, 765–773. 9969.
58 S. Grieshammer, B. O. H. Grope, J. Koettgen and M. Martin, 86 R. A. De Souza and A. H. H. Ramadan, Phys. Chem. Chem.
Phys. Chem. Chem. Phys., 2014, 16, 9974–9986. Phys., 2013, 15, 4505–4509.
59 R. A. De Souza and M. Martin, Monatsh. Chem., 2009, 140, 87 R. A. De Souza, A. Ramadan and S. Hörner, Energy Environ.
1011–1015. Sci., 2012, 5, 5445–5453.
60 C. E. Mohn, S. Stølen, S. T. Norberg and S. Hull, Phys. Rev. 88 N. Sata, K. Eberman, K. Eberl and J. Maier, Nature, 2000,
Lett., 2009, 102, 155502. 408, 946–949.
61 C. E. Mohn, S. Stølen, S. T. Norberg and S. Hull, Phys. Rev. 89 W. Shen, J. Jiang, C. Ni, Z. Voras, T. P. Beebe Jr. and J. L.
B, 2009, 80, 024205. Hertz, Solid State Ionics, 2014, 255, 13–20.
62 R. D. Shannon, Acta Crystallogr. A, 1976, 32, 751–767. 90 N. F. Mott and R. W. Gurney, Electronic processes in ionic crys-
63 J. Mizusaki, K. Arai and K. Fueki, Solid State Ionics, 1983, tals, Oxford University Press, Oxford, 2nd edn, 1950.
11, 203–211. 91 S. Murugavel and B. Roling, J. Non-Cryst. Solids, 2005, 351,
64 K. Takada, Acta Mater., 2013, 61, 759–770. 2819–2824.
65 D. Lybye, F. W. Poulsen and M. Mogensen, Solid State Ionics, 92 B. Roling, S. Murugavel, A. Heuer, L. Luhning, R. Friedrich
2000, 128, 91–103. and S. Rothel, Phys. Chem. Chem. Phys., 2008, 10, 4211–
66 M. Cherry, M. Islam and C. Catlow, J. Solid State Chem., 4226.
1995, 118, 125–132. 93 A. R. Genreith-Schriever and R. A. De Souza, Phys. Rev. B,
67 R. Davies, M. Islam and J. Gale, Solid State Ionics, 1999, 126, 2016, 94, 224304.
323–335. 94 A. Kushima and B. Yildiz, J. Mater. Chem., 2010, 20, 4809–
68 M. Islam, Solid State Ionics, 2002, 154-155, 75–85. 4819.
69 D. A. Andersson, S. I. Simak, N. V. Skorodumova, I. A. 95 T. Mayeshiba and D. Morgan, Phys. Chem. Chem. Phys.,
Abrikosov and B. Johansson, Proc. Natl. Acad. Sci. U. S. A., 2015, 17, 2715–2721.
2006, 103, 3518–3521. 96 B. Yildiz, MRS Bulletin, 2014, 39, 147–156.
70 M. Nakayama and M. Martin, Phys. Chem. Chem. Phys., 97 J. Garcia-Barriocanal, A. Rivera-Calzada, M. Varela, Z. Se-
2009, 11, 3241–3249. frioui, E. Iborra, C. Leon, S. J. Pennycook and J. Santamaria,
71 E. Kotomin, Y. A. Mastrikov, M. Kuklja, R. Merkle, A. Royt- Science, 2008, 321, 676–680.
burd and J. Maier, Solid State Ionics, 2011, 188, 1–5. 98 A. Cavallaro, M. Burriel, J. Roqueta, A. Apostolidis,
72 M. Lumeij, J. Koettgen, M. Gilleßen, T. Itoh and R. Dron- A. Bernardi, A. Tarancón, R. Srinivasan, S. N. Cook, H. L.
skowski, Solid State Ionics, 2012, 222-223, 53–58. Fraser, J. A. Kilner, D. W. McComb and J. Santiso, Solid State
73 T. T. Mayeshiba and D. D. Morgan, Solid State Ionics, 2016, Ionics, 2010, 181, 592–601.

J
our
nal
Name,
[yea
r][
,vol
.
],
1–17 | 15
Journal of Materials Chemistry A Page 16 of 17
View Article Online
DOI: 10.1039/C7TA04266C

99 N. Schichtel, C. Korte, D. Hesse and J. Janek, Phys. Chem. 126 H. J. M. Bouwmeester, C. Song, J. Zhu, J. Yi, M. van Sint An-
Chem. Phys., 2009, 11, 3043–3048. naland and B. A. Boukamp, Phys. Chem. Chem. Phys., 2009,
100 A. Fluri, D. Pergolesi, V. Roddatis, A. Wokaun and T. Lippert, 11, 9640–9643.
Nature Commun., 2016, 7, 10692. 127 C.-Y. Yoo and H. J. M. Bouwmeester, Phys. Chem. Chem.
101 M. Wilkening, V. Epp, A. Feldhoff and P. Heitjans, J. Phys. Phys., 2012, 14, 11759–11765.

Journal of Materials Chemistry A Accepted Manuscript


Chem. C, 2008, 112, 9291–9300. 128 T. Kawada, M. Sase, M. Kudo, K. Yashiro, K. Sato,
Published on 11 September 2017. Downloaded by FLORIDA ATLANTIC UNIVERSITY on 12/09/2017 03:59:30.

102 P. Heitjans, E. Tobschall and M. Wilkening, Eur. Phys. J.: J. Mizusaki, N. Sakai, T. Horita, K. Yamaji and H. Yokokawa,
Spec. Top., 2008, 161, 97–108. Solid State Ionics, 2006, 177, 3081–3086.
103 V. Metlenko, A. Ramadan, F. Gunkel, H. Du, H. Schraknep- 129 S. Miyoshi, A. Takeshita, S. Okada and S. Yamaguchi, Solid
per, S. Hoffmann-Eifert, R. Dittmann, R. Waser and R. A. State Ionics, 2016, 285, 202–208.
De Souza, Nanoscale, 2014, 6, 12864–12876. 130 J. A. Lane and J. A. Kilner, Solid State Ionics, 2000, 136-137,
104 E. Kurumchin and M. Perfiliev, Solid State Ionics, 1990, 42, 927–932.
129–133. 131 G. M. Rupp, H. Tellez, J. Druce, A. Limbeck, T. Ishihara,
105 E. K. Kurumchin, M. V. Ananjev, G. K. Vdovin and M. G. J. Kilner and J. Fleig, J. Mater. Chem. A, 2015, 3, 22759–
Surkova, Russ. J. Electrochem., 2010, 46, 205–211. 22769.
106 N. M. Bershitskaya, M. V. Ananyev, E. K. Kurumchin and V. A. 132 A. Staykov, H. Téllez, T. Akbay, J. Druce, T. Ishihara and
Eremin, Russ. J. Electrochem., 2012, 48, 961–968. J. Kilner, Chem. Mater., 2015, 27, 8273–8281.
107 A. Berenov, A. Atkinson, J. Kilner, M. Ananyev, 133 R. A. De Souza and J. A. Kilner, Solid State Ionics, 1999, 126,
V. Eremin, N. Porotnikova, A. Farlenkov, E. Kurumchin, 153–161.
H. Bouwmeester, E. Bucher and W. Sitte, Solid State Ionics, 134 R. A. De Souza, Phys. Chem. Chem. Phys., 2006, 8, 890–897.
2014, 268, 102–109. 135 S. B. Adler and W. G. Bessler, in Handbook of Fuel Cells, ed.
108 J. Maier, Solid State Ionics, 2000, 135, 575–588. W. Vielstich, H. Yokokawa and H. A. Gasteiger, John Wiley &
109 R. Merkle and J. Maier, Phys. Chem. Chem. Phys., 2002, 4, Sons, Ltd, 2010, vol. 5, pp. 441–462.
4140–4148. 136 I. Riess, Solid State Ionics, 2015, 280, 51–65.
110 R. Merkle and J. Maier, Top. Catal., 2006, 38, 141–145. 137 I. Riess, Solid State Ionics, 2017, 302, 7–11.
111 M. Leonhardt, R. A. De Souza, J. Claus and J. Maier, J. Elec- 138 E. R. S. Winter, J. Chem. Soc. A, 1968, 2889–2902.
trochem. Soc., 2002, 149, J19–J26. 139 D. A. McQuarrie and J. D. Simon, Physical Chemistry, A
112 L. Wang, R. Merkle and J. Maier, J. Electrochem. Soc., 2010, Molecular Approach, University Science Books, New Jersey,
157, B1802–B1808. USA, 1997.
113 S. B. Adler, X. Y. Chen and J. R. Wilson, J. Catal., 2007, 245, 140 A. M. Saranya, D. Pla, A. Morata, A. Cavallaro, J. Canales-
91–109. Vázquez, J. A. Kilner, M. Burriel and A. Tarancón, Adv. Energy
114 S. B. Adler, Chem. Rev., 2004, 104, 4791–4844. Mater., 2015, 5, 1500377.
115 S. B. Adler, Solid State Ionics, 1998, 111, 125–134. 141 E. Navickas, T. M. Huber, Y. Chen, W. Hetaba, G. Holzlech-
116 J. R. Wilson, M. Sase, T. Kawada and S. B. Adler, Elec- ner, G. Rupp, M. Stoger-Pollach, G. Friedbacher, H. Hutter,
trochem. Solid-State Lett., 2007, 10, B81–B86. B. Yildiz and J. Fleig, Phys. Chem. Chem. Phys., 2015, 17,
117 D. S. Mebane and M. Liu, J. Solid State Electrochem., 2006, 7659–7669.
10, 575–580. 142 V. Metlenko, W. Jung, S. R. Bishop, H. L. Tuller and R. A.
118 Y. Choi, D. S. Mebane, J.-H. Wang and M. Liu, Top. Catal., De Souza, Phys. Chem. Chem. Phys., 2016, 18, 29495–
2007, 46, 386–401. 29505.
119 W. Jung and H. L. Tuller, Adv. Energy Mater., 2011, 1, 1184– 143 R. Merkle, R. A. De Souza and J. Maier, Angew. Chem., Int.
1191. Ed., 2001, 40, 2126–2129.
120 Y.-L. Lee, J. Kleis, J. Rossmeisl, Y. Shao-Horn and D. Morgan, 144 Y. Chen, W. Jung, Z. Cai, J. J. Kim, H. L. Tuller and B. Yildiz,
Energy Environ. Sci., 2011, 4, 3966–3970. Energy Environ. Sci., 2012, 5, 7979–7988.
121 S. Kim, Y. Yang, A. Jacobson and B. Abeles, Solid State Ionics, 145 Y. Gassenbauer, R. Schafranek, A. Klein, S. Zafeiratos,
1998, 106, 189–195. M. Hävecker, A. Knop-Gericke and R. Schlögl, Solid State
122 S. Wang, A. Verma, Y. Yang, A. Jacobson and B. Abeles, Solid Ionics, 2006, 177, 3123–3127.
State Ionics, 2001, 140, 125–133. 146 C. Körber, A. Wachau, P. Ágoston, K. Albe and A. Klein, Phys.
123 M. Mosleh, M. Søgaard and P. V. Hendriksen, J. Electrochem. Chem. Chem. Phys., 2011, 13, 3223–3226.
Soc., 2009, 156, B441–B457. 147 Y. Choi, M. E. Lynch, M. C. Lin and M. Liu, J. Phys. Chem. C,
124 W. C. Chueh and S. M. Haile, Annu. Rev. Chem. Biomol. Eng., 2009, 113, 7290–7297.
2012, 3, 313–341. 148 Y. A. Mastrikov, R. Merkle, E. Heifets, E. A. Kotomin and
125 M. den Otter, B. Boukamp and H. Bouwmeester, Solid State J. Maier, J. Phys. Chem. C, 2010, 114, 3017–3027.
Ionics, 2001, 139, 89–94. 149 Y. Choi, D. S. Mebane, M. C. Lin, and M. Liu, Chem. Mater.,
2007, 19, 1690–1699.

16 | J
our
nal
Name,
[yea
r][
,vol
.
],1–17
Page 17 of 17 Journal of Materials Chemistry A
View Article Online
DOI: 10.1039/C7TA04266C

150 G. Pilania and R. Ramprasad, Surf. Sci., 2010, 604, 1889– C. Körber, Y. Gassenbauer, F. Säuberlich, G. V. Rao, S. Payan,
1893. M. Maglione, C. Chirila, L. Pintilie, L. Jia, K. Ellmer,
151 H. Guhl, W. Miller and K. Reuter, Surf. Sci., 2010, 604, 372– M. Naderer, K. Reichmann, U. Böttger, S. Schmelzer, R. C.
376. Frunza, H. Uršič, B. Malič, W.-B. Wu, P. Erhart and A. Klein,
152 V. Alexandrov, S. Piskunov, Y. F. Zhukovskii, E. A. Kotomin phys. status solidi RRL, 2014, 8, 571–576.

Journal of Materials Chemistry A Accepted Manuscript


and J. Maier, Integr. Ferroelectr., 2011, 123, 10–17. 176 L. Bjaalie, B. Himmetoglu, L. Weston, A. Janotti and C. G.
Published on 11 September 2017. Downloaded by FLORIDA ATLANTIC UNIVERSITY on 12/09/2017 03:59:30.

153 R. A. De Souza, F. Gunkel, S. Hoffmann-Eifert and Van de Walle, New J. Phys., 2014, 16, 025005.
R. Dittmann, Phys. Rev. B, 2014, 89, 241401. 177 R. Jacobs, J. Booske and D. Morgan, Adv. Funct. Mater.,
154 W. C. Chueh and S. M. Haile, Phys. Chem. Chem. Phys., 2009, 2016, 26, 5471–5482.
11, 8144–8148. 178 A. Klein, personal communication.
155 S. J. Benson, R. J. Chater and J. A. Kilner, Proc. Third Intl. 179 N. Sakai, K. Yamaji, T. Horita, Y. P. Xiong, H. Kishimoto and
Symp. Ionic and Mixed Conducting Ceramics, Pennington, H. Yokokawa, J. Electrochem. Soc., 2003, 150, A689–A694.
NJ, USA, 1997, pp. 596–609. 180 N. Sakai, K. Yamaji, T. Horita, H. Kishimoto, Y. P. Xiong and
156 J. E. ten Elshof, M. H. R. Lankhorst and H. J. M. H. Yokokawa, Phys. Chem. Chem. Phys., 2003, 5, 2253–2256.
Bouwmeester, J. Electrochem. Soc., 1997, 144, 1060–1067. 181 M. Kessel, R. A. De Souza, H.-I. Yoo and M. Martin, Appl.
157 M. Kuhn, S. Hashimoto, K. Sato, K. Yashiro and J. Mizusaki, Phys. Lett., 2010, 97, 021910.
Solid State Ionics, 2011, 195, 7–15. 182 S. J. Cooper, M. Niania, F. Hoffmann and J. A. Kilner, Phys.
158 R. A. De Souza and M. Martin, Phys. Chem. Chem. Phys., Chem. Chem. Phys., 2017, 19, 12199–12205.
2008, 10, 2356–2367. 183 E. D. Wachsman, Y.-L. Huang, C. Pellegrinelli, J. A. Taillon
159 D. S. Mebane, Comput. Mater. Sci., 2015, 103, 231–236. and L. G. Salamanca-Riba, ECS Transactions, 2014, 61, 47–
160 D. N. Mueller, M. L. Machala, H. Bluhm and W. C. Chueh, 56.
Nat. Commun., 2015, 6, 6097. 184 T. S. Bjørheim, M. Arrigoni, S. W. Saeed, E. Kotomin and
161 A. Tschöpe, S. Kilassonia and R. Birringer, Solid State Ionics, J. Maier, Chem. Mater., 2016, 28, 1363–1368.
2004, 173, 57–61. 185 E. J. Crumlin, S.-J. Ahn, D. Lee, E. Mutoro, M. D. Biegalski,
162 R. A. De Souza, Z. A. Munir, S. Kim and M. Martin, Solid H. M. Christen and Y. Shao-Horn, J. Electrochem. Soc., 2012,
State Ionics, 2011, 196, 1–8. 159, F219–F225.
163 B. J. Nyman, E. E. Helgee and G. Wahnström, Appl. Phys. 186 M. J. Gadre, Y.-L. Lee and D. Morgan, Phys. Chem. Chem.
Lett., 2012, 100, 061903. Phys., 2012, 14, 2606–2616.
164 D. S. Mebane and R. A. De Souza, Energy Environ. Sci., 2015, 187 W. Ma, J. J. Kim, N. Tsvetkov, T. Daio, Y. Kuru, Z. Cai,
8, 2935–2940. Y. Chen, K. Sasaki, H. L. Tuller and B. Yildiz, J. Mater. Chem.
165 A. Berenov, A. Atkinson, J. Kilner, E. Bucher and W. Sitte, A, 2015, 3, 207–219.
Solid State Ionics, 2010, 181, 819–826. 188 J. Hayd, H. Yokokawa and E. Ivers-Tiffée, J. Electrochem.
166 A. Egger, E. Bucher, M. Yang and W. Sitte, Solid State Ionics, Soc., 2013, 160, F351–F359.
2012, 225, 55–60. 189 S. Stämmler, R. Merkle and J. Maier, J. Electrochem. Soc.,
167 I. C. Fullarton, J. P. Jacobs, H. E. van Benthem, J. A. Kilner, 2017, 164, F454–F463.
H. H. Brongersma, P. J. Scanlon and B. C. H. Steele, Ionics, 190 M. Sase, K. Yashiro, K. Sato, J. Mizusaki, T. Kawada,
1995, 1, 51–58. N. Sakai, K. Yamaji, T. Horita and H. Yokokawa, Solid State
168 M. H. Lankhorst, H. Bouwmeester and H. Verweij, J. Solid Ionics, 2008, 178, 1843–1852.
State Chem., 1997, 133, 555–567. 191 M. Sase, F. Hermes, K. Yashiro, K. Sato, J. Mizusaki,
169 M. Søgaard, P. V. Hendriksen, M. Mogensen, F. W. Poulsen T. Kawada, N. Sakai and H. Yokokawa, J. Electrochem. Soc.,
and E. Skou, Solid State Ionics, 2006, 177, 3285–3296. 2008, 155, B793–B797.
170 J. Mizusaki, Y. Mima, S. Yamauchi, K. Fueki and H. Tagawa, 192 R. Bowley and M. Sánchez, Introductory Statistial Mechanics,
J. Solid State Chem., 1989, 80, 102–111. Oxford University Press, Oxford, UK, 1996.
171 E. Bucher, A. Egger, P. Ried, W. Sitte and P. Holtappels, Solid 193 R. A. De Souza, J. Zehnpfenning, M. Martin and J. Maier,
State Ionics, 2008, 179, 1032–1035. Solid State Ionics, 2005, 176, 1465–1471.
172 X. Chen, J. Yu and S. B. Adler, Chem. Mater., 2005, 17, 4537– 194 M. F. Kessel, R. A. De Souza and M. Martin, Phys. Chem.
4546. Chem. Phys., 2015, 17, 12587–12597.
173 H. L. Tuller and A. S. Nowick, J. Phys. Chem. Solids, 1977, 195 R.-V. Wang and P. C. McIntyre, J. Appl. Phys., 2005, 97,
38, 859–867. 023508.
174 A. Klein, Thin Solid Films, 2012, 520, 3721–3728. 196 S. Gottschalk, H. Hahn, S. Flege and A. G. Balogh, J. Appl.
175 S. Li, F. Chen, R. Schafranek, T. J. M. Bayer, K. Rachut, Phys., 2008, 104, 114106.
A. Fuchs, S. Siol, M. Weidner, M. Hohmann, V. Pfeifer, 197 M. Leonhardt, J. Jamnik and J. Maier, Electrochem. Solid-
J. Morasch, C. Ghinea, E. Arveux, R. Günzler, J. Gassmann, State Lett., 1999, 2, 333–335.

J
our
nal
Name,
[yea
r][
,vol
.
],
1–17 | 17

You might also like