View Article Online / Journal Homepage / Table of Contents for this issue

PCCP
Oxygen incorporation into Fe-doped SrTiO3 : Mechanistic
interpretation of the surface reaction

Rotraut Merkle and Joachim Maier

Max-Planck-Institute for Solid State Research, Heisenbergstr. 1, D-70569, Stuttgart, Germany.

E-mail: [email protected]
Published on 01 August 2002. Downloaded by Technische Universitat Darmstadt on 3/14/2024 10:13:06 AM.

Received 29th April 2002, Accepted 14th June 2002

First published as an Advance Article on the web 1st August 2002

The surface part of the oxygen incorporation reaction into the model material Fe-doped SrTiO3 is investigated
by in situ optical spectroscopy. Experiments close to and far from equilibrium (small and large pO2 changes)
and the effect of UV irradiation on the reaction rate allow us to draw the following conclusions with respect to
the kinetic processes at the surface: (i) molecular oxygen species participate in or before the rate determining
step (RDS), and (ii) a single conduction band electron participates in or before the RDS of the incorporation
reaction even though the bulk is p-type conducting. Probable reaction mechanisms comprising chemisorption,
charge transfer, dissociation and oxygen ion transfer steps are discussed.

1. Introduction this method has the great advantage of avoiding the applica-
tion of electrodes, which is essential since metallic as well as
The oxygen incorporation reaction in oxide materials is of oxidic electrode materials have been shown to exhibit signifi-
major technological relevance for the properties and the func- cant catalytic activity.8
tioning of electroceramic devices, and still an important sub-
ject for fundamental research. It is the stoichiometry which
decides whether a pure material (given the structure, tempera- 2. Experimental
ture and pressure) is electronically or ionically, p- or n-type
conducting, whether it is an insulator, a metal or even a super- SrTiO3 single crystals (Frank & Schulte GmbH, Essen,
conductor. It is the rate of stoichiometric changes which deter- Germany) with an iron content of 4.6
1019 cm3,
mines the relevant time scales of conditioning or the response 1.4
1019 cm3 and 0.5
1019 cm3 were cut and polished to
times of chemical sensors based on stoichiometry changes. The yield samples with dimensions 6 mm
6 mm
1 mm. The
mechanism of oxygen incorporation can be divided into a sur- large faces exhibited (100) or (110) orientation, which upon
face part (proper reaction) and a bulk part (transport). While thermal pre-treatment at 900  C in O2 became facetted into
the transport in the bulk is well understood in many oxides (see the more stable (100), (010) surfaces (indicated by atomic force
e.g. ref. 1–4 and references therein), the surface part of the pro- microscopy). Oxygen stoichiometry changes were performed
cess is vastly terra incognita. We chose to study this reaction on between 540  C and 730  C by switching the oxygen partial
Fe-doped SrTiO3 , a model material representative of a wide pressure pO2 in the range of 0.0001 bar to 0.3 bar (achieved
class of acceptor doped wide bandgap oxides. Its bulk defect with O2/N2 mixtures; flow rate 100 ml min1). The set-up
chemistry (see e.g. ref. 5–7) and transport properties8 are well for the in situ optical spectroscopy is described in detail in
studied, providing a sound basis for the investigation of trans- ref. 12. The integral Fe4+ concentration was monitored
port across inner (grain boundaries) and outer surfaces (oxy- continuously by measuring the absorption at 595 nm with a
gen incorporation kinetics). The treatment of effective rate Perkin–Elmer Lambda 2 UV–Vis spectrometer. A narrow
constants given in ref. 9 and 10 provides a basis to trace back bandpass filter was placed between the sample and the detector
the interfacial effects on mass transport to mechanistic details. to block thermal radiation. For the irradiation experiments, a
Effective rate constants of the oxygen incorporation reaction 200 W Hg high pressure arc lamp (LOT Oriel) was used, whose
k̄d as well as of the 18O isotope exchange reaction k̄* have spectral output was restricted to 280–420 nm by means of UG5
already been determined for SrTiO3 in experiments close to and WG280 filters (Schott). The bandgap of SrTiO3 at 700  C
equilibrium.8 UV irradiation, which locally increases the con- is 2.7 eV, which corresponds to a fundamental absorption edge
duction band electron concentration, was found to selectively at 460 nm. The short wavelength cut-off avoids the direct
accelerate the incorporation reaction.11 photolysis of O2 species (compare related species: strong
In this work, we report experiments performed far from UV absorption of O2 below 240 nm, of H2O2 below
equilibrium (large pO2 changes) which allow us to monitor 280 nm13). The total irradiation intensity was 250 mW cm2
selectively the oxygen in- or excorporation reaction. The in front of the quartz rod, which acted as a light guide to
results rule out a number of hypothetical reaction sequences irradiate one of the large faces of the sample, and was attenu-
and lead to the suggestion of probable mechanisms which ated for some experiments by neutral density filters. The con-
are capable of describing the experimental results. The effective centration of UV-generated e0 –h pairs is estimated to lie
rate constants k̄d of the oxygen incorporation are determined between the equilibrium concentrations of e0 and h ,11 thus
by in situ optical spectroscopy, which monitors the concentra- the small value of [e0 ] can be significantly enhanced by UV irra-
tion changes of the different valence states of the redox-active diation while the large [h ] remains essentially unchanged. The
dopant ion (Fe3+/Fe4+).12 Besides the fact that it records the UV irradiation was interrupted for 30 s at appropriate inter-
transport in situ as a function of time (and space if desired), vals to permit the 595 nm extinction measurements to be made.

4140 Phys. Chem. Chem. Phys., 2002, 4, 4140–4148 DOI: 10.1039/b204032h

This journal is # The Owner Societies 2002
 View Article Online

Sub-bandgap irradiation of 515–700 nm with a total intensity conversion z to the elapsed time t by
of 190 mW cm2 proved to have no effect on the oxygen in/ d t=l
excorporation rates. z ¼ 1  ek ð7Þ
with l ¼ sample-halfwidth. The Lambert–Beer law relates the
absorbance A of the Fe4+ species to their concentration and
3. Theory allows the determination of [VO  ] according to
3.1 Defect chemistry of Fe-doped SrTiO3 
½VO t  ½VO t¼0

At  At¼0
z¼   ¼ : ð8Þ
In SrTiO3 , iron can be incorporated on the titanium site exhi- ½VO t¼1  ½VO t¼0 At¼1  At¼0
biting, under the conditions of interest, the valencies Fe3+ and
Under the assumption that a single step is rate determining,
Fe4+. The valence distribution depends on oxygen partial pres-
and that the preceding/succeeding reaction steps are in quasi-
Published on 01 August 2002. Downloaded by Technische Universitat Darmstadt on 3/14/2024 10:13:06 AM.

sure pO2 , temperature and total iron concentration. Fe2+ can

equilibrium (i.e. stoichiometric changes in the surface layer are
be observed only under very strongly reducing conditions14
negligible), the flux is invariant in the surface. It can be related
and its concentration is neglected here. The charge compensa-
to the reaction rate R of the RDS and hence be described by
tion takes place mainly by the formation of oxygen vacancies
simple chemical kinetics. For conciseness, we exemplify this
VO  , and, to a lesser extent, electron holes h . Thus the defect
for the model reaction A Ð B (with the reaction Gibbs energy
chemical model for Fe-doped SrTiO3 in the dilute range con-
DG) for situations close to equilibrium:
sists of the following reactions (Kröger–Vink notation):
Oxygen incorporation: j
k½B ¼ ~
k½A  ~
¼ R ¼~ k½Að1  eDG=RT Þ ð9Þ
   2  0:5 2l
0:5 O2 þ VO Ð OO þ 2h K 1 ¼ ½h  =½VO pO2 ð1Þ
When approaching equilibrium, the exponential term can be
Electron-hole formation: linearized and further expressed by dc and the thermodynamic
nil Ð e0 þ h

K 2 ¼ ½e0 ½h 

ð2Þ factor w (dm ¼ RT d ln a):
j
 dm
 dm w
Impurity ionization:  ~k½A ¼ ~ k½B ¼ R0 dc ð10Þ
2l eq RT eq RT c
FeTi Ð FeTi 0 þ h K 3 ¼ ½FeTi 0 ½h =½FeTi 
 
ð3Þ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
R0 ¼ ~ k½Aeq ~ k½Beq ð11Þ
The defect concentrations are further connected by the electro-
neutrality condition c @m @ ln a
w¼ ¼ ð12Þ
2½VO  þ ½h  ¼ ½FeTi 0 
 
ð4Þ RT @c @ ln c

and the mass conservation of iron (in the presence of electrical fields the rate constant may also
depend on the distance from equilibrium). Thus the observable
½FeTi  þ ½FeTi 0  ¼ ½Fetot ð5Þ k̄od is linked to the equilibrium exchange rate RO0 determined
from mechanistic considerations:
The mass action constants K1 , K2 and K3 and their tempera-
ture dependences given in Appendix A are refined para- d wO
k O / RO 0 : ð13Þ
meters.15 This defect chemical model allows us to calculate cO
the individual defect concentrations, their pO2 dependence
and the thermodynamic factor which are needed for data eva- Please note that both wO and cO refer to the first bulk layer
luation and interpretation. while RO0 refers to the RDS and hence generally to a surface
reaction step which involves the surface layer. The equilibrium
concentrations entering eqn. (11) can be connected with bulk
3.2 Kinetics of stoichiometry changes and gas phase concentrations via local mass action laws for
preceding and succeeding steps.
The kinetics of the oxygen in-/ex-corporation reaction can be With increasing magnitude of the pO2 changes, the assump-
described phenomenologically by effective rate constants k̄d. tion of proximity to equilibrium is no longer valid and
Their relation to chemical kinetics has been derived in ref. 9 extended models must be applied, as also noticed in the case
and 10 and shall be presented here briefly for the simple situa- of the conductivity relaxation of lanthanum strontium cobal-
tion of proximity to equilibrium. A change in the pO2 ambi- tite.16
ence of an equilibrated sample changes the chemical For experiments far from equilibrium (large pO2 changes),
potential of oxygen mO and thus creates the driving force for the initial reaction rate R will be dominated by the forward
a flux jO of oxygen into/out of the sample, which depends or backward reaction rate, R ~ or R ~ , and can be determined
on the inverse chemical resistance LO . The flux can then be
from the initial slope of z and the overall [VO  ] change
related to the experimentally accessible quantities, the effective  
rate constant k̄d and the oxygen concentration difference dcO at  dz
Rt0 ¼ ð½VO t¼0  ½VO t¼1 Þ 
 
ð14Þ
the surface dt t0
j O ¼ LO dmO ¼ kO d dcO: ð6Þ or directly from the observed data according to
 
More precisely, the quantities dmO and dcO refer to the first  dz
bulk layer counted from the interface under consideration, and Rt0  ðAt¼0  At¼1 Þ  : ð15Þ
dt t0
measure the difference of actual and final values at this posi-
tion. Unlike dcO , the quantity dmO is also identical to the spa- It turned out that the numerical results for the pO2 depen-
tial difference across the interface; dmO is also identical with the dence of R|t ¼ 0 do not depend significantly on whether R|t ¼ 0
Gibbs reaction energy of the incorporation reaction. has been determined either from extrapolating R(t) to t ¼ 0,
In the surface controlled regime (i.e. the surface reaction is or via dz/dt|t ¼ 0 obtained from fitting z(t) by a straight line
slow relative to the diffusion inside the sample, lk̄d < Dd), the in the range of low degree of conversion, z < 15%. Under what
oxygen vacancy concentration inside the sample is homo- conditions eqn. (13) is still valid far from equilibrium and by
geneous to a good approximation. In the case of plate-like which relations it has to be replaced, is treated in detail in
samples the time integration relates the normalized degree of ref. 9.

Phys. Chem. Chem. Phys., 2002, 4, 4140–4148 4141

 View Article Online

3.3 Possible intermediates and mechanisms For charged adsorbates, on oxides with Mott–Schottky deple-
tion layers and typical band bendings, Weisz predicts maxi-
In the course of the oxygen incorporation reaction, negative mum surface concentrations of about 1012–1013 cm2 (ref.
charge is transferred to the reacting oxygen species and the 32) which corresponds to 102–103 adsorbates per elementary
strong O=O double bond (bond enthalpy 5.2 eV) is broken, cell.
finally leading to oxide ions O2. In the following, intermediate The second transferred charge (again e0 or h ) formally leads
species are discussed which can occur along this path and have to O22 (stabilized by surface interaction) which can instanta-
been identified more or less clearly. Their stability is expected neously or more or less rapidly split into two O. Both of these
to depend to a varying degree on the stabilization by interac- steps are candidates for the RDS. The formal involvement of
tion with the oxide surface and surface defects. O22 does not imply its real existence; when the splitting of
The formation of O2 superoxide ions by oxygen adsorption the O22 is not rate determining, its concentration will vanish
Published on 01 August 2002. Downloaded by Technische Universitat Darmstadt on 3/14/2024 10:13:06 AM.

has been observed frequently, mainly by EPR, XPS/UPS and in the overall mass action constants of the quasi-equilibria pre-
IR spectroscopy, on wide bandgap oxides (see ref. 17 and 18 ceding/succeeding the RDS (as any intermediates do). The dis-
and references therein). In such cases electrons may be sociation of O2 to O and O without simultaneous charge
assumed to be transferred from the conduction band (e.g. transfer is not considered here, since the formation of neutral
reduced ZnO, SnO2 , CeO2 , TiO2 , ZrO2 , SrTiO319); from loca- atomic O on a wide bandgap oxide surface is expected to be
lized native defects (e.g. F-centers in gamma-irradiated MgO); energetically unfavourable. The same applies to the hypotheti-
or from localized redox-active dopant ions (e.g. cobalt in unre- cal disproportionation of O22 into O2 and O. For the same
duced Co-doped MgO). Photoadsorption of oxygen is reason, mechanisms are omitted which contain the neutral dis-
reported e.g. for TiO2 , ZrO2 , ZnO and Al2O3 .20–23 In some sociative adsorption O2 Ð 2 O (although it is favoured e.g. on
cases the bonding to the substrate is sufficiently high so that Pt surfaces).
the superoxide was observed up to 300  C. The last step is the incorporation of O leading to O2 and
The existence of adsorbed O22 peroxide ions on the above- again comprises a single charge transfer. This charge can be
mentioned materials is more controversial, which is at least provided by a conduction electron only when the incorpora-
partly due to the fact that fewer analytical methods are avail- tion is the RDS (otherwise there is contradiction of the UV
able for its identification. Its existence has been claimed on pre- results11 as discussed above). If the incorporation is not the
reduced CeO2 , observed by IR spectroscopy,24 on prereduced RDS, the charge transfer in the last step must be the formation
SnO2 , observed by XPS/UPS,25 and on UHV fractured of a valence band hole. In the case of 18O isotope exchange on
SrTiO3 observed by electron energy loss spectroscopy YSZ at temperatures below 750  C, the detection of significant
(EELS).26 The formation of O22 in the course of oxygen amounts of 18O 16O in the gas phase indicates that the incor-
adsorption and dissociation has been demonstrated for noble poration of atomic oxygen species is slow relative to their for-
metals such as Pt and Pd, for which a gradual conversion of mation,33 while at higher temperatures the incorporation is
O2 to O22 with increasing temperature was detected by fast. Similarly, a slow, rate determining incorporation of
HREELS.27 atomic oxygen species has been suggested for the oxide ion
The formation of surface O radicals in the course of oxy- conductor (Bi2O3)0.75(Er2O3)0.2534 on the basis of isotope
gen adsorption and dissociation on oxides has again been exchange experiments at 823–973 K. From these considera-
observed frequently. It can be evidenced by EPR after O2 tions, three simple reaction pathways are left (with branching
adsorption on several prereduced oxides,28 e.g. ZnO, TiO2 , in step 4) which comprise the transfer of one to four conduc-
SrTiO3 .19 On ZnO, the gradual transformation of O2 to tion electrons (Table 1).
O was detected with increasing temperature.29 An increasing
ratio of [O]/[O2] with increasing temperature was also pre- 3.4 pO2 dependences of R0 and R|t ¼ 0
dicted for SnO230 by rate equation simulations based on data
determined for ZnO.31 The stability of surface O is certainly Once a mechanism has been set up with a single rate determin-
related to the fact that the effect of coordinative unsaturation is ing step, the dependence of the equilibrium reaction rate R0 on
less drastic for a lower charged ion. pO2 and the defect concentrations can be calculated from eqn.
Taking these intermediates as a starting point, possible reac- (11). In all cases we assume the adsorption equilibrium to be in
tion mechanisms shall be set up. The transfer of negative the limit of low coverage (1  y  1). In Table 1, the obtained
charges to oxygen species is considered to take place by con- expressions and their numerical values based on the bulk
sumption of conduction band electrons or by formation of defect chemical data are given for the different possible cases
electron holes in the valence band. As long as the charge trans- depending on which elementary step is taken to be rate deter-
fer occurs in quasi-equilibrium, i.e. within steps pre- or suc- mining. The mechanisms are classified as A (second transferred
ceeding the RDS, the charge carrier type cannot be charge is a conduction electron e0 ), B (first transferred charge is
distinguished by kinetic investigations close to equilibrium, an e0 ) and C (first and second transferred charge is an e0 ). It is
since e0 and h are interrelated by the band gap equilibrium. obvious that several coincidences in the overall pO2 depen-
Only for charges transferred in the RDS and for pronounced dence of R0 are present and that the analysis of R0 alone does
non-equililbrium situations is such a distinction possible. From not lead to a clear distinction.
the selective acceleration of the oxygen incorporation by UV When the experiments are performed far from equilibrium
irradiation11 we can conclude that conduction band electrons (large pO2 changes), then in the initial stage of the relaxation
must participate in or before the RDS of the incorporation, either the oxygen incorporation or excorporation dominates
and not after the RDS. Later we will consider only mechan- and determines the observed reaction rate R. Additionally, in
isms which fulfill this criterion. this time interval the defect concentrations can be regarded
The first proper step of a gas–solid reaction is the adsorp- as approximately equal to the initial concentrations. The
forward reaction rate R ~ (incorporation) is determined by the
tion. As the physisorption (essentially neutral adsorption) can-
not be the RDS (otherwise e0 would not be involved in/before forward rate constant ~ k of the RDS and its educt concentra-
the RDS), we will combine the physisorption with the chemi- tions, which themselves are related to the defect concentrations
sorption involving a first charge transfer step (e0 or h ) to an by the pre-equilibria. It depends directly on pO2 (via adsorp-
overall adsorption reaction yielding O2. For high coverage tion) and possibly on all defects and can be expressed as
degrees (y) of adsorption centers, or for the build-up of large R k  pO2 n ½e0 x ½h y ½VO z
~ ¼ Kf ~  
ð16Þ
surface charges by chemisorption, adsorption isotherms other
than described by the Langmuir equation may be necessary. (x, y, z, being effective reaction orders.)

4142 Phys. Chem. Chem. Phys., 2002, 4, 4140–4148

 View Article Online

Table 1 Possible reaction mechanisms A–C and corresponding reaction rates R ¼ R ~R ~ and equilibrium reaction rates R0 calculated for each of
the steps being the RDS. For the last step (incorporation) two alternative possibilities 4a or 4b are listed. Kf and Kb denote the overall equilibrium
constants of all preceding/succeeding reaction steps. Bulk defect pO2 dependences: [VO  ]  pO20.03, [e0 ]1  [h ]  pO20.23

A: case 4a: 1 electron B: case 4a: 1 electron C: case 4a: 2 electrons

RDS: case 4b: 3 electrons case 4b: 3 electrons case 4b: 4 electrons

Step 1 O2 Ð O2 + h O2 + e0 Ð O2 O2 + e0 Ð O2

 
Step 2 O2 + e Ð 0
O22 O2 Ð O22 + h O2 + e0 Ð O22
 
Step 3 O2 2
Ð 2O O2 2
Ð 2O O22 Ð 2O
 
Step 4a 
2O + 2VO Ð 2OO + 2h  
2O + 2VO Ð 2OO + 2h 
2O + 2VO  Ð 2OO + 2h
Published on 01 August 2002. Downloaded by Technische Universitat Darmstadt on 3/14/2024 10:13:06 AM.

 
Step 4b 2O + 2VO + 2e0 Ð 2OO

2O + 2VO + 2e0 Ð 2OO

2O + 2VO  + 2e0 Ð 2OO
 3  3  3
½h  ½h  ½h 
Step 1 R ¼ Kf ~
kpO2  Kb~
k R ¼ Kf ~
k½e0 pO2  Kb~
k R ¼ Kf ~
k½e0 pO2  Kb~ k
½e0 ½VO 2 ½V O 2 ½V O 2
  

0 0.99 0 0.76 0 0.76

RO  pO2 RO  pO2 RO  pO2
 2  3  2
½e0  ½h  ½h  ½h 
Step 2 R ¼ Kf ~
k  pO2  Kb~ k R ¼ Kf ~
k½e0 pO2  Kb~
k k½e0 2 pO2  Kb~
R ¼ Kf ~ k
½h  ½V O 2 ½V O 2 ½V O 2
  

RO0  pO20.53 0
RO  pO2 0.76 0
RO  pO2 0.53
 2  2  2
½e0  ½h  ½e0  ½h  ½h 
Step 3 R ¼ Kf ~
k  pO2  Kb~
k R ¼ Kf ~
k  pO2  Kb~ k k½e0 2 pO2  Kb~
R ¼ Kf ~ k
½h  ½V O 2 ½h  ½V O 2 ½V O 2
  

RO0  pO20.53 sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi RO0  pO20.53 sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 0

RO  pO2 0.53

   ½e0  ½e 0   pffiffiffiffiffiffiffiffiffi
R ¼ Kf ~ ~  R ¼ Kf ~ ~  R ¼ Kf ~
k½e0 ½VO  pO2  Kb~
 
Step 4a k VO  pO2  Kb k ½h  k½VO   pO2  Kb k ½h  k ½h 
½h  ½h 
RO0  pO20.24 sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi RO0  pO20.24 sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi RO0  pO20.24
½e 0 3 ½e 0 3  pffiffiffiffiffiffiffiffiffi
R ¼ Kf ~ ~ R ¼ Kf ~ ~ k½e0 2 ½VO  pO2  Kb~
R ¼ Kf ~
 
Step 4b k ½V O   pO2  Kb k k½VO   pO2  Kb k k
½h  ½h 
RO0  pO20.01 RO0  pO20.01 RO0  pO20.01

After switching the gas atmosphere, the oxygen partial pres- brium of step 4 provided this step is not theRDS):
sure near the SrTiO3 surface reaches the new final value almost
k  ½h v ½VO w
~ ¼ Kb~  
immediately (fast gas phase replacement). Owing to the slug- R ð19Þ
gish surface kinetics, the oxygen stoichiometry of the sample For a series of downward pO2 changes from the same initial
follows on a much greater time scale. Thus the defect concen- pO2 to various final pO2 , the initial defect concentration is
trations contained in the initial reaction rate R|t ¼ 0 can be well invariant, thus
approximated by their equilibrium values before the pO2
change. For a series of upward pO2 changes starting from
ð20Þ
the same initial pO2 to various final pO2 values, the initial
defect concentrations are identical, and the exponent n of the
direct contribution of pO2 can be determined: For a series of downward pO2 changes from various initial pO2
to the same final pO2 , the overall pO2 dependence is deter-
mined by the contribution of defects to the backward step of
ð17Þ the RDS:

Mechanisms containing molecular oxygen species of any ð21Þ

charge (further denoted by O2?) in the RDS always yield
n ¼ 1, mechanisms containing only atomic O? species yield The various expressions resulting from the mechanisms A, B
n ¼ 1/2. A series of upward pO2 changes starting at various and C for the reaction rate R ¼ R ~R ~ are also listed in
initial pO2 and ending at the same pO2(end) results in an over- Table 1. Please note that although the pO2 dependence of R0
all pO2 dependence which is determined by the defect contribu- is identical, the expression for R can still differ (compare B3
tions to the forward step of the RDS: with C3), and the difference can be detected in non-equilibrium
experiments. Thus, these series of pO2 changes allow us to fix
ð18Þ directly some important cornerstones for the reaction mechan-
ism: the molecularity of the RDS and the number of partici-
pating conduction electrons.
Conversely, from a series of upward pO2 changes with identical
initial and final pO2 , but with variation of the UV intensity
(causing a corresponding variation of [e0 ] while [h ]  constant) 4. Results and discussion
constant) the exponent of the conduction electrons can be
determined. The value x ¼ 1 corresponds to one electron up Experiments close to equilibrium (small pO2 changes) have
to/including the RDS, x ¼ 2 indicates two electrons up to/ been carried out on samples with (110) faces11 and [Fe]tot ¼
including the RDS. 4.6
1019 cm3. They yield a constant pO2 dependence for
The backward reaction rate R ~ (excorporation, with reaction k̄d of pO20.33 in the temperature range 650–730  C (Fig. 1a).
orders v, w) depends only on the hole and vacancy concentra- Small pO2 changes on samples with nominally identical Fe
tions, e0 do not participate in/before the backward step of the concentration (samples from the same crystal boule) with
RDS as discussed above (VO  contribute via the quasi-equili- (100) faces at 700–730  C yield a similar pO2 dependence of

Phys. Chem. Chem. Phys., 2002, 4, 4140–4148 4143

 View Article Online

In Fig. 2, the effect of UV irradiation on the oxygen in/

excorporation reaction is shown for the (110) faced samples
with [Fe]tot ¼ 4.6
1019 cm3. The incorporation reaction
is clearly accelerated, while the excorporation remains
unchanged.11 As discussed above, this indicated the participa-
tion of conduction band electrons in or before the RDS of the
incorporation. The unchanged excorporation additionally con-
firms that the acceleration of the incorporation is not due to a
thermal effect of the UV irradiation, since this would also
accelerate the excorporation.
Further mechanistic information can be gained from large
Published on 01 August 2002. Downloaded by Technische Universitat Darmstadt on 3/14/2024 10:13:06 AM.

pO2 changes (experiments far from equilibrium). Fig. 3 shows

the pO2 dependence observed in several series of large pO2
changes of the (110) oriented sample with [Fe]tot ¼ 4.6
1019
cm3 at 730  C. For series of jumps starting at identical initial
~ exhibits a
pO2 to various higher pO2 , the initial reaction rate R
pO2 dependence of pO20.88, which is significantly closer to the
case of n ¼ 1 than to n ¼ 1/2 in eqn. (17) (the slightly smaller
pO2 exponent compared to the theoretical prediction can be
attributed to effects appearing at high pO2 : the finite gas
exchange time, and possibly a nonzero coverage in the adsorp-
tion equilibrium). This is a clear indication that molecular spe-
cies O2? are contained in the RDS of O2 incorporation, and
confirms that the incorporation (step 4a or 4b) is not the
RDS. For a series of jumps from identical initial pO2 to var-
ious lower pO2 , R ~ has a pO2 dependence of pO20.03, which
is in excellent agreement with the predicted pO2 independence
(eqn. (20)).
For a series of jumps from various lower to the same final
pO2 , R~ exhibits an overall pO2 dependence of pO20.33, and
for jumps from various higher to the same final pO2 we find
~  pO20.47. Similar pO2 dependences are found for the (110)
R
Fig. 1 pO2 dependence of k̄d at various temperatures. a: [Fe]tot ¼ oriented sample at 670  C and the (100) oriented sample at
4.6
1019 cm3(110) face; b: triangles [Fe]tot ¼ 1.4
1019 cm3(100) 730  C (see Fig. 4a and b and Table 2). These numerical values
face, circles [Fe]tot ¼ 0.5
1019 cm3 (100) face. also give an estimate of the overall errors in the measured pO2
dependences, which are about 10%.
The variation of the UV intensity and thus of the conduction
pO20.36, but the absolute values of k̄d are smaller by a factor of electron concentration allows us to determine the number of
about 5. This difference can be explained partly by the fact that conduction electrons participating up to or in the RDS accord-
the terracing of the (110) face to (100) and (010) oriented faces
pffiffiffi ing to eqn. (18). Approximating the electron concentration
increases the surface area (an increase by a factor of is 2 under UV irradiation as
expected by this). The enhanced number of steps on the
facetted (110) face might further enhance k̄d by offering a lar- ½e0   ½e0 therm þ ½e0 UV  ½e0 therm þ aI UV ð22Þ
ger number of favourable adsorption sites, and some scatter in 0
with temperature-independent [e ]UV , we expect either a linear
the absolute values k̄d must also be taken into account. relation
Under the experimental conditions, the thermodynamic fac-
tor wO has a pO2 dependence of pO20.13, which leads to a pO2 ~  ð½e0 
R therm þ aIUV Þ ð23Þ
dependence of the equilibrium exchange rate R0 of pO20.46–0.49
(eqn. (13)). This exponent comes closest to the value of 0.53
predicted in Table 1 for the steps A2, A3, B3, C2 or C3 being
the RDS of the oxygen incorporation (see below). The effective
activation energy is 274 kJ mol1 and 258 kJ mol1 for the
(110) and (100) faced samples, respectively. As it also contains
the reaction enthalpies of the quasi-equilbria before and after
the RDS, it cannot be interpreted in a straightforward way.
The fact that the pO2 dependence does not change over the
temperature range indicates that under these conditions the
adsorption equilibrium is in the limit of small coverage. Other-
wise a decrease in the pO2 exponent with decreasing tempera-
ture (and increasing coverage) would be expected from the
Langmuir adsorption equation. Such a decrease can be
observed e.g. for (100) faced samples with Fe concentrations
of 0.5
1019 cm3 (decrease from 0.27 at 750  C to 0.08 at
550  C) and 1.4
1019 cm3 (decrease from 0.19 at 700  C to
0.6 at 550  C), as shown in Fig. 1b. A comparison of the abso-
lute values at a certain pO2 and temperature shows that k̄d
decreases strongly with increasing overall iron concentration.
k̄d is roughly proportional to Fetot2.5. This relation must be Fig. 2 Effect of UV irradiation on oxygen in-/excorporation,
interpreted with caution, because there are indications for a [Fe]tot ¼ 4.6
1019 cm3(110) face. Solid line: without UV; circles:
slow iron segregation at the applied measuring temperatures.35 with UV.

4144 Phys. Chem. Chem. Phys., 2002, 4, 4140–4148

Published on 01 August 2002. Downloaded by Technische Universitat Darmstadt on 3/14/2024 10:13:06 AM. View Article Online

Fig. 4 Initial reaction rates R|t ¼ 0 for large pO2 changes at 730  C,
Fig. 3 Initial reaction rates R|t ¼ 0 for large pO2 changes at 730  C, [Fe]tot ¼ 4.6
1019 cm3 (100) face. a: Changes from fixed pO2(ini)
[Fe]tot ¼ 4.6
1019 cm3 (110) face. a: Changes from fixed pO2(ini)
to various pO2(end) and ; b: Changes from various
to various pO2(end) and ; b: Changes from various
pO2(ini) to fixed pO2(end) and .
pO2(ini) to fixed pO2(end) and .

firmed by Fig. 6, where the temperature dependences of the

or a quadratic relation ~ for changes from 0.01 bar to 0.1 bar with and
initial rate R
~  ð½e0  2 without UV irradiation are compared. When [e0 ]UV [e0 ]therm ,
R therm þ aIUV Þ ð24Þ
the overall activation energy under irradiation is reduced by
between R~ and the UV intensity. In Fig. 5, this dependence is the contribution which stems from [e0 ]therm or [e0 ]therm2 (our
shown for pO2 changes from 0.01 bar to 0.1 bar. At 600  C, full defect model predicts [e0 ]therm  e234 kJ/(mol
RT)). For the case
UV irradiation accelerates R ~ by a factor of 6.5, but the distinc- of one electron up to/in the RDS, this leads to almost tem-
tion between the linear and quadratic relation is not very clear- perature-independent R ~ (dashed line in Fig. 6, calculated
~ by a factor of
cut. At 540  C, full UV irradiation accelerates R according to eqn. (23)) which is in good agreement with the
120, and the data point for half intensity clearly obeys the experimental findings. For the case of two electrons up to/in
linear relation. Thus one electron participates up to or in the RDS, eqn. (24) predicts a minimum followed by a strong
the RDS, ruling out mechanism C. This finding is also con- increase of R~ (dotted line in Fig. 6).

Table 2 Exponents of pO2 dependence of the initial reaction rates R|t ¼ 0 and equilibrium reaction rate R0 derived from the proposed mechanisms
and determined experimentally

730  C (110) 670  C (110) 730  C (100) A2/A3/B3 B1/B2

R|t ¼ 0 up from fixed pO2(ini) 0.88 0.78 0.83 1 1

R|t ¼ 0 down from fixed pO2(ini) 0.02 0.07 0.03 0 0

R|t ¼ 0 up to fixed pO2(end) 0.33 0.37 0.28 0.46 0.23

R|t ¼ 0 down to fixed pO2(end) 0.47 0.42 0.54 0.52 0.75

R0 small pO2 changes 0.46 0.46 0.49 0.53 0.76

Phys. Chem. Chem. Phys., 2002, 4, 4140–4148 4145

Published on 01 August 2002. Downloaded by Technische Universitat Darmstadt on 3/14/2024 10:13:06 AM. View Article Online

Fig. 7 Band scheme of Fe-doped SrTiO3 including oxygen chemi-

sorption levels (flatband case).

Furthermore, we assume equal charge carrier transition prob-

abilities if either an electron is consumed or a hole created, and
Fig. 5 Effect of relative UV irradiation intensity IUV /IUV0 on initial equal effective densities of state N for the valence and the con-
oxygen incorporation rate R|t ¼ 0 , [Fe]tot ¼ 4.6
1019 cm3(110) face. duction band. The situation is illustrated for the hypothetical
flatband situation in Fig. 7a (charge transfer to O2) and 7b
(O2). As long as the respective transition state is still below
In Table 2, all observed pO2 dependences are collected and the conduction band, the probability of O2 formation upon
compared to the values predicted for those mechanisms which pO2 increase by electron consumption (dashed arrow) is pro-
(i) contain molecular O2? species in the RDS and (ii) contain portional to the conduction electron concentration N

one electron up to/in the RDS of oxygen incorporation. e(EFEC)/kT  N
e1.9 eV/kT but not dependent on an activa-
Although the numerical agreement is not perfect, the results tion barrier. The probability of superoxide formation via crea-
clearly favour A2/A3/B3 relative to B1/B2 for the RDS. tion of a trapped hole (Fe3+ ! Fe4+, solid arrow) depends on
From the experimental data, a distinction between A2 and [Fe3+]  0.01N ¼ N
e0.4 eV/kT (calculated from the defect
A3 being the RDS is not possible. Nevertheless, the probability model at 700  C) as well as on the energy difference DE1 and
of either A2/A3 or B3 (which differ mainly in the step in which is N
e(DE1+0.4 eV)/kT. Thus for DE1 up to 1.5 eV the hole
the electron is transferred) being the RDS can be discussed on formation is estimated to be preferred (as long as the attempt
the level of a first approximation. At 700  C, the band gap of frequencies for the charge transfer are not too different). The
SrTiO3 is 2.7 eV, and the Fermi level EF is close to the partially probability of O22 formation (see 0
Fig. 7b) by electron con-
occupied Fe3+/Fe4+ level 0.8 eV above the valence band sumption is N
e(1.9 eV+DE 2)/kT  N
eDE2/kT (dashed
(owing to the experimentally determined valence ratio of Fe, arrow), and thus favoured relative to the trapped hole forma-
EF is located near the Fe3+/Fe4+ level). The position of the tion (solid arrow) which is 0.01 N
eDE2/kT. From these
O2/O2 level in the bandgap of SrTiO3 can be roughly esti- arguments we conclude that the charge transfer has a greater
mated from literature data of other wide bandgap oxides: on probability of taking place according to mechanism A (hole
TiO2 (110) surfaces (rutile, bandgap 3.0 eV at 300 K) it is 1 formation followed by electron consumption) than mechanism
eV below the conduction band,36 on ZnO (polycrystalline, B (electron consumption followed by hole formation). The
bandgap 3.2 eV at 300 K) it is 0.9 eV below the conduction same conclusion is reached from an analogous consideration
band.31 Data for the O2/O22 level (being either intermediate of the backward reaction, if the Franck–Condon type barriers
species or transition state) are not available, but it is reason- are of similar magnitude.
able to expect a position significantly higher in the bandgap The proposed possible mechanisms A2/A3 (as well as B3)
(or even above the conduction band) than for O2/O2. For are also in accordance with the observed decrease of k̄d with
both transferred charges, a Franck–Condon type activation increasing [Fe]tot . They contain a term k̄d  1/[VO  ] 
barrier due to O–O bond length relaxation is expected. [Fe]tot0.96, and together with the thermodynamic factor
wO  [Fe]tot1.27, this leads to the prediction of k̄d  [Fe]tot2.23.
This is quite close to the observed k̄d  [Fe]tot2.5 (as men-
tioned before, this result must be interpreted with care).
In the last paragraphs, some further aspects shall be dis-
cussed briefly, which might be important with respect to the
oxygen in/excorporation reaction, but whose extensive treat-
ment goes beyond the scope of this manuscript. By the macro-
scopic kinetic approach, we are limited concerning some
microscopic questions. We cannot identify the nature of the
adsorption sites, which might be related e.g. to surface oxygen
vacancies or to surface Fe centers. An independent variation of
these quantities, which would allow their identification by
macroscopic kinetics, is practically impossible. Also, from
the proposed set of mechanisms, we cannot decide if O22 spe-
cies really appear. If O22 was absent, the fusion of steps A2
and A3 to a new RDS would not alter the pO2 dependence.
The fusion of B2 and B3 to a new RDS yields the same pO2
dependence as if B2 was the RDS.
Anyway, the prediction of defect concentrations at the sur-
face is a difficult issue. We certainly must expect that the abso-
Fig. 6 Temperature dependence of initial oxygen incorporation rate lute concentrations at the surface will differ from those in the
R|t ¼ 0 without/with UV irradiation, [Fe]tot ¼ 4.6
1019 cm3 (110) bulk. Density functional calculations predict a decrease of
face. the (direct) indirect band gap at the SrO terminated surface

4146 Phys. Chem. Chem. Phys., 2002, 4, 4140–4148

 View Article Online

by (7%) 15% and at the TiO terminated surface by (12%) jO Oxygen flux, eqn. (6)
27%.37 Fortunately only the defects’ pO2 dependence enters k̄d Effective rate constant of oxygen in-/
the data evaluation. The formation of VO  is certainly ex-corporation, eqn. (6)
enhanced at the surface, and possibly also the formation of k̄* Effective rate constant of oxygen isotope exchange
Fe3+. These changes can be represented by a reduction of ~
k, ~
k Forward or backward rate constant,
the reaction enthalpies in eqn. (1) to (3). Let us consider the eqn. (10), (16) and (19)
following example, in which for simplicity we assume local Kf , Kb Mass action constants of preceding or succeeding
electroneutrality: let us further assume that DH1 , DH2 and equilibria, eqn. (16) and (19)
DH3 are reduced exemplarily to 2/3 of their bulk values, then l Sample halfwidth, eqn. (7)
surface defect concentrations can be calculated yielding LO Inverse chemical resistance, eqn. (6)
[VO  ]surf  pO20.002 and [h ]surf  pO20.24 (in the range of n Formal reaction order of pO2 in incorporation,
eqn. (16)
Published on 01 August 2002. Downloaded by Technische Universitat Darmstadt on 3/14/2024 10:13:06 AM.

experimental conditions), which is quite similar to the bulk

pO2 Oxygen partial pressure
behaviour. A more realistic picture also takes excess charges
R Reaction rate, eqn. (9)
(space charges) into account.
R|t ¼ 0 Initial reaction rate, eqn. (14)
Grain boundaries of Fe-doped SrTiO3 in bicrystals and RO0 Equilibrium exchange rate, eqn. (11)
polycrystalline material usually exhibit a positively charged ~, R
~
grain boundary core compensated by a negative space charge R Forward or backward reaction rate,
eqn. (16) and (19)
zone (mainly depletion of VO  and h ) [see e.g. ref. 38]. The
wO Thermodynamic factor, eqn. (12)
corresponding space charge potential is almost pO2 indepen- v, w Formal reaction order of defects in excorporation,
dent.38,39 A similarly positive surface charging is expected to eqn. (19)
appear at the outer surface (Fe segregation is detected in Fe- x, y, z Formal reaction order of defects in incorporation,
doped SrTiO335 as well as in Fe-doped TiO240). As discussed eqn. (16)
in Appendix B, as long as the charge density is not substan- z Normalized degree of conversion, eqn. (8)
tially affected by the oxygen incorporation process, this results
in an approximately constant (as regards pO2) space charge
potential and similar pO2 dependences of the surface defects Appendix A: Current refined defect chemical
as in the bulk.
When the RDS involves the movement of charged particles
parameters15
across this interface, the corresponding potential difference The numerical values of the mass action constants used in this
exerts an additional driving or retarding force and affects k̄d paper are refined values based on thermogravimetric and con-
and R directly. When the movement of charged particles ductivity measurements and are independent of this study:
occurs in the preceding/succeeding equilibria, their mass
action constants are modified. A pO2 independent space charge K 1 ¼ 2:47
1021 Pa1=2 cm3
e1:65 eV=kT
potential is expected to alter the absolute values, but to have
no effect on the pO2 dependence of k̄d and R.41 K 2 ¼ 7:67
1042 cm6
eð3:3 eV=kT0:0006 eV=kÞ
K 3 ¼ 5:82
1023 cm6
e1:18 eV=kT
5. Concluding remarks According to this, in the range of the applied experimental
conditions (pO2 , T), the pO2 dependence of the bulk defect
Based on macroscopic kinetic investigations of the oxygen in/ concentrations and the thermodynamic factor can be approxi-
excorporation reaction of Fe-doped SrTiO3 , we can propose a mated as:
small set of reaction mechanisms which are able to describe the
½VO   pO2 0:03

experimental findings. The mechanisms comprise chemisorp-
tion, charge transfer, dissociation and oxygen ion transfer 
½h   pO2 0:23
steps. From experiments far from equilibrium (including the
effect of UV irradiation) we can conclude that (i) the RDS of ½FeTi 0   pO2 0:03
the oxygen incorporation contains molecular O2? species,
and that (ii) one electron participates in or before the RDS ½FeTi   pO2 0:20
of the incorporation reaction. wO  pO2 0:13
Unlike these statements, the following conclusions are based
on the assumption that the pO2 dependences of the surface and
bulk defect concentrations are approximately equal. Consider-
ing all experimental data, we can further suggest that the RDS Appendix B: Space charge potential and surface
is either A2 (according to Table 1, formation of O22 inter-
mediates by electron consumption) or A3 (splitting of O22) defect pO2 dependences
or the fusion of those steps; or with less probability B3 (split- The positive surface charge density (which we assume accord-
ting of O22 after it was formed by hole creation). Unfortu- ing to ref. 35 and 40 ) is expected to be approximately pO2
nately, from the measured data we cannot distinguish independent and may be due either due to cation excess which
whether O22 is a real intermediate or only a transition state. is frozen under the experimental conditions, or to predominant
Experiments are in progress on other acceptor-doped wide enrichment of VO.. in the surface layer caused by lowering of
bandgap oxides, e.g. Mn-doped Y-stabilized ZrO2 , to see if its chemical potential at the surface. We expect this surface
the same mechanism applies for a whole class of oxide materi- charge not to be influenced by the comparatively small redox
als. Another still unsettled question is whether the chemical changes which probably cause a small negative surface charge
oxygen exchange and the tracer exchange proceeds by the same concentration due to the charge transfer in the chemisorption
mechanism in these materials.8 step. As our material is electron-poor, we expect only small
absolute concentrations of O2 and O22 and thus negligible
[ ] Point defect concentration (per volume). concentration variations, so that the surface charging is domi-
A Absorbance of FeTi absorption band, eqn. (8) nated by the pO2 independent part.
A1..., B1..., C1... Classification of reaction mechanisms, Table 1 From a pO2 independent positive surface charge density,
IUV UV intensity, eqn. (22) conclusions can be drawn concerning the pO2 dependence of

Phys. Chem. Chem. Phys., 2002, 4, 4140–4148 4147

 View Article Online

the space charge potential and the surface defect concentra- 6 R. Waser, J. Am. Ceram. Soc., 1991, 74, 1934.
tions in the presence of surface charging.42 As the microscopic 7 I. Denk, W. Münch and J. Maier, J. Am. Ceram. Soc., 1995, 78,
knowledge of the surface layer (denoted x ¼ s) is poor, several 3265, and references therein.
8 M. Leonhardt, R. A. De Souza, J. Claus and J. Maier, J. Electro-
possible cases must be discussed separately, but lead to similar chem. Soc., 2002, 149, J19, and references therein.
conclusions. 9 J. Maier, Solid State Ionics, 1998, 112, 197.
In the Mott–Schottky case ([VO  ] concentration profile in 10 J. Maier, Solid State Ionics, 2000, 135, 575.
the space charge region, [Fe0 Ti] ¼ constant) a pO2 independent 11 R. Merkle, R. A. De Souza and J. Maier, Angew. Chem. Int. Ed.
surface charge density induces also a pO2 independent space Engl., 2001, 40, 2126.
charge potential. When the VO  are the majority charge carrier 12 T. Bieger, J. Maier and R. Waser, Ber. Bunsen-Ges. Phys. Chem.,
1993, 97, 1098.
in the surface layer, then [VO  ]s is pO2 independent and 13 Gmelin Handbook of Inorganic Chemistry, Verlag Chemie, Wein-
[h ]s  pO20.25 follows directly from the oxygen incorporation heim, 8th edn., 1963.
Published on 01 August 2002. Downloaded by Technische Universitat Darmstadt on 3/14/2024 10:13:06 AM.

equilibrium (note that the electrical potential term cancels in 14 V. G. Bhide and H. C. Bhasin, Phys. Rev., 1968, 172, 290.
the mass action law if all the charged constituents refer to 15 R. Merkle, R. A. De Souza and J. Maier, to be published.
the same equipotential line). When the positive surface charge 16 S. Wang, A. Verma, Y. L. Yang, A. J. Jacobson and B. Abeles,
is caused by cation excess, [VO  ]s is again approximately pO2 Solid State Ionics, 2001, 140, 125.
17 M. Che and A. J. Tench, Adv. Catal., 1983, 32, 1.
independent (small redox effects presupposed) and [h ]s 18 M. Anpo, M. Che, B. Fubini, E. Garonne, E. Giamello and M. C.
approximately pO20.25 assuming all iron to be in the Fe3+ Paganini, Top. Catal., 1999, 8, 189.
oxidation state. In the Gouy–Chapman case ([VO  ] and [Fe0 Ti] 19 Y. Nakano, M. Watanabe and T. Takahshi, J. Appl. Phys., 1991,
concentration profile) a constant surface charge density 70, 1539.
induces a space charge potential with the inverse pO2 depen- 20 R. I. Bickley and F. S. Stone, J. Catal., 1973, 31, 389.
dence of the respective defects, i.e. here a slight dependence 21 A. V. Emeline, A. V. Rudakova, V. K. Ryabchuk and N. Serpone,
J. Phys. Chem. B, 1998, 102, 10 906.
of pO20.03 occurs. The pO2 dependences of the surface defect 22 G. N. Kuzmin, D. Purevdorzh and I. G. Shenderovich, Kinet.
concentrations are the same as in the Mott–Schottky case. Catal., 1995, 36, 728.
In fact this second picture (excess charge great compared to 23 A. O. Klimovskii and A. A. Lisachenko, Kinet. Catal., 1991,
charge variations on stoichiometry change) is much closer to 32, 378.
reality than the neutral model or than an intermediate case 24 C. Li, K. Domen, K. Maruya and T. Onishi, J. Am. Ceram. Soc.,
of weak charging, in which the results would be more compli- 1989, 111, 7683.
25 T. Kawabe, K. Tabata, E. Suzuki, Y. Yamaguchi and Y. Nasa-
cated. We can summarize that for both extreme cases we end gawa, J. Phys. Chem. B, 2001, 105, 4239.
with a negligible pO2 dependence of the space charge potential 26 V. M. Bermudez and V. H. Ritz, Chem. Phys. Lett., 1980,
and pO2 dependences of the surface defects which are similar 73, 160.
to the bulk. 27 R. Imbihl and J. E. Demuth, Surf. Sci., 1986, 173, 395.
Please note that a significant (pO2 independent) space charge 28 M. Che and A. J. Tench, Adv. Catal., 1982, 31, 77.
potential lowers the electrical potential drop from the adsorp- 29 S.-C. Chang, J. Vac. Sci. Technol., 1980, 17, 366.
30 T. S. Rantala, V. Lantto and T. T. Rantala, Sens. Actuators B,
tion/reaction layer to the adjacent first bulk SrTiO3 layer. 1993, 13, 234.
A pO2 dependent space charge potential is expected to influ- 31 S. R. Morrison, Surf. Sci., 1971, 27, 586.
ence k̄d and R as well as their pO2 dependences. 32 P. B. Weisz, J. Chem. Phys., 1953, 21, 1531.
33 P. S. Manning, J. D. Sirman and J. A. Kilner, Solid State Ionics,
1997, 93, 125.
34 B. A. Boukamp, B. A. van Hassel, I. C. Vinke, K. J. de Vries and
Acknowledgements A. J. Burggraaf, Electrochim. Acta, 1993, 38, 1817.
35 M. Leonhardt, PhD Thesis, University of Stuttgart, 1999.
We wish to thank Dr R. A. De Souza and Dr J. Jamnik for 36 W. Göpel, G. Rocker and R. Feierabend, Phys. Rev. B, 1983, 28,
stimulating discussions. 3427.
37 E. Heifets, R. I. Eglitis, E. A. Kotomin, J. Maier and G. Borstel,
Surf. Sci., 2002, 513, 211.
38 I. Denk, J. Claus and J. Maier, J. Electrochem. Soc., 1997, 144,
References 3526.
39 R. A. De Souza, unpublished results.
1 C. Wagner and W. Schottky, Z. Phys. Chem. B, 1930, 11, 163. 40 A. Bernasik, M. Rekas, M. Sloma and W. Weppner, Solid State
2 C. Wagner, Prog. Solid State Chem., 1975, 10, 3. Ionics, 1994, 72, 12.
3 L. Heyne, in Solid State Electrolytes, ed. S. Geller, Springer, Ber- 41 J. Jamnik and J. Maier, Ber. Bunsen-Ges. Phys. Chem., 1997,
lin, 1977, p. 169. 101, 23.
4 J. Maier, J. Am. Ceram. Soc., 1993, 76, 1212, 1218, 1223. 42 J. Maier: Festkörper, – Fehler und Funktion, Teubner Verlag,
5 G. M. Choi and H. L. Tuller, J. Am. Ceram. Soc., 1988, 71, 201. Stuttgart, 2000; English version in preparation.

4148 Phys. Chem. Chem. Phys., 2002, 4, 4140–4148