Solid State Ionics 111 (1998) 125–134

Mechanism and kinetics of oxygen reduction on porous

La 12x Sr x CoO 32d electrodes

S.B. Adler*
Ceramatec Inc., 2425 S. 900 W., Salt Lake City, UT 84119, USA
Received 7 November 1997; accepted 3 June 1998

Abstract

The linear polarization behavior of La 12x Sr x CoO 32d electrodes has been measured as a function of x, T, and PO 2 using
a.c. impedance. These measurements indicate a reaction mechanism in which O 2 is reduced chemically at the porous mixed
conductor surface, followed by diffusion of oxygen through the mixed conductor to the electrolyte. Analysis of the electrode
kinetics using a recent model for this mechanism yields values for the oxygen vacancy diffusion coefficient (Dv ) and
surface-exchange rate constant (r 0 ) as a function of x, T, and PO 2 . These values agree well with published independent
measurements of oxygen chemical diffusion and isotope surface exchange (D˜ and k). The distance (d ) that the reaction
extends beyond the electrode / electrolyte interface depends on both Dv and r 0 , varying between 0.3 and 10 microns for a
typical electrode surface area.  1998 Elsevier Science B.V. All rights reserved.

Keywords: Perovskite; Mixed conductor; Electrode kinetics; Ion conductor; Chemical diffusion; Surface exchange; Transition metal oxide

1. Introduction conductor may play a significant (if not dominant)

Much debate surrounds the mechanism of the
oxygen–reduction reaction at a porous mixed-con-
ducting oxygen electrode. In analogy to a porous
metal catalyst (e.g. Pt on zirconia), many workers
apply a traditional paradigm in which O 2 reduction is
limited by a charge-transfer process occurring locally
at the one-dimensional ‘three-phase boundary’ inter-
face between electrode, electrolyte, and gas [1–3].
However, much evidence suggests that oxygen ab-
sorption and transport in the bulk of the mixed
*Corresponding author. Address: Dept. of Chemical Engineer-
ing, Case Western Reserve University, 10900 Euclid Ave.,
Cleevelan, Ohio 44106, U.S. Tel.: 11 216 368 4182; e-mail:
stu@ceramatec.com
role in determining the electrode polarization: (i)
large faradaic capacitances (up to 1 farad / cm 2 ) [4,5],
(ii) semi-infinite diffusion behavior at intermediate
frequencies [6], (iii) linear i-V characteristics at
moderate overpotential [7], (iv) strong correlation
between electrode kinetics and bulk diffusion and
surface-exchange properties [8,9]. Even a poor
mixed conductor such as La 12x Srx MnO 32d (LSM)
shows evidence for involvement of bulk-transport
[10,11].
In order to incorporate these effects, some workers
have applied modified charge-transfer or equivalent
circuit models that assume non-charge-transfer pro-
cesses occurring in the electrode can be treated as
series or parallel combinations of charge-transfer
elements [12]. However, this assumption has been

0167-2738 / 98 / $ – see front matter  1998 Elsevier Science B.V. All rights reserved.
PII: S0167-2738( 98 )00179-9
126 S.B. Adler / Solid State Ionics 111 (1998) 125 – 134

shown by continuum theory to be generally incorrect

for treating oxygen absorption and solid-state / gase-
ous diffusion, which contribute to the impedance in a
convoluted manner [13]. The equivalent circuit ap-
proach also leaves the electrode reaction mechanism
only vaguely or empirically defined, with poor
connection between modeling parameters and the
properties of the materials. Although such a model
may provide a useful set of parameters to ‘fit’ data
accurately, it provides little mechanistic insight.
In order to establish a more direct connection
between performance and properties of mixed-con-
ductors, several workers have used continuum
models to treat explicitly the non-charge-transfer
processes occurring in a membrane [14] or electrode
[13,15]. An impedance model which uses this ap-
proach [13,16,17] has shown preliminary evidence
that the kinetics for mixed conductors with fast ionic
transport (such as La 12x Sr x Co 12y Fe y O 32d ) is domi-
nated by surface chemical exchange of O 2 and solid-
state oxygen diffusion. As shown schematically in
Fig. 1, this model treats the overall electrode re-
action, ]21 O 2 1 2e 2 → O 22 , as a homogeneous chemi-
cal reaction occurring over the internal surface area
of the electrode material. The absorption of neutral
oxygen by the mixed-conductor serves to convert
electronic to ionic current over a finite region of the
electrode thickness. This active region is described Fig. 1. (a) Schematic of the electrode reaction model. (b)
by the characteristic length, d, related to the ex- Schematic of the steady-state electronic and ionic currents in the
change and diffusion properties of the mixed conduc- mixed conductor.
tor. Estimates of d based on SIMS measurements of
D* and k yield values of several microns. This tions quantitatively, and link electrode kinetics di-
model also predicts that gas phase diffusion becomes rectly to material properties. This link not only
dominant below 1% oxygen in N 2 , even at currents allows us to examine the implications of this mecha-
well below limiting. nism, but also to examine and gain insight from
In this paper, we apply this model to understand- where the model breaks down.
ing mixed-conducting La 1-x Sr x CoO 3-d (LSCO) elec-
trodes on ceria. LSCO is an excellent material for
fundamental studies because it is reversibly reduc- 2. Theory
ible, has high rates of surface exchange and diffu-
sion, and forms a clean well-characterized electrode As shown in Fig. 2a, the model discussed above
system with little or no interfacial charge-transfer predicts that the zero-bias impedance of a symmetri-
resistances. LSCO has been well studied in the cal cell (electrode / electrolyte / electrode) can be ex-
literature, and independent measurements of its pressed as a sum of charge-transfer resistances and
thermodynamic, kinetic, and transport properties impedances, plus a ‘chemical’ impedance associated
have been made over a wide range of temperature with non-charge-transfer processes:
and PO 2 conditions. Thus application of the model to
LSCO / ceria allows us evaluate the models assump- Z 5 R electrolyte 1 Zinterfaces 1 Zchem , (1)
 S.B. Adler / Solid State Ionics 111 (1998) 125 – 134 127

]]]]]]]
S Dœ
RT
R chem 5 ]]2
2F
]]]]]]]
t
(1 2 ´)c v Dv ar 0 (af 1 ab )
(3)

c v (1 2 ´)
t chem 5 ]]]] (4)
Aar 0 (af 1 ab )
where c v is the vacancy concentration, A 5 2 1 /
(2n) 5 2 ]21 ≠ln(PO 2 ) / ≠ln(c v ) is the thermodynamic
factor, Dv is the vacancy diffusion coefficient, ´, a,
and t are the porosity, surface area, and solid-phase
tortuosity, respectively, of the porous mixed conduc-
tor, r 0 is the exchange neutral flux density (in
analogy to exchange-current density), and af and ab
are kinetic parameters of order unity that depend on
the specific mechanism of the surface chemical
exchange reaction.
The model also predicts that at steady state, the
mixed conductor will be reduced by an amount that
decays exponentially with distance from the elec-
trolyte interface, with a characteristic length describ-
ing the size of the active region [13]:
]]]]
c v Dv (1 2 ´)

Fig. 2. (a) Schematic of the total impedance as predicted by Eq.

œ
d 5 ]]]]
ar 0 (af 1 ab )t
(5)

Finally, the model also predicts that at low O 2

(1). (b) Plot of the ‘chemical’ impedance in the limit of Eq. (2).
concentration in the gas, diffusion in the pores of the
mixed conductor may become important. The
where R electrolyte is the electrolyte resistance,
characteristic resistance of this diffusion process,
Zinterfaces is the impedance of the electron-transfer
which contributes to Zchem , is given by [13]:
and ion-transfer processes occuring at the current
collector / electrode and electrode / electrolyte inter- 2LRT(1 2 xxp )tp
faces, respectively, and Zchem is the convoluted Zp 5 ]]]]] (6)
16F 2 cx´xp DAB
contribution of non-charge-transfer processes includ-
ing oxygen surface exchange, solid-state diffusion, where L is the thickness of the electrode, x is the
and gas-phase diffusion inside and outside the elec- oxygen mole fraction, c is the total gas phase
trode [13]. concentration (P/RT ), DAB is the binary diffusion
As shown in Fig. 2b, in the limit of a semi-infinite coefficient between O 2 and the diluent gas, xp 5DK /
(thick) electrode with no gas-phase diffusion limita- (DK 1DAB ) is a correction factor for Knudsen diffu-
tions, the non-charge-transfer term reduces to: sion, DK is the Knudsen diffusivity of O 2 in the
pores of the solid, and tp is the gas-phase tortuosity.
]]]]
1
œ
Zchem 5 R chem ]]]]
1 1 jv t chem
(2)
3. Experimental

where R chem and t chem are a characteristic resistance LSCO powders of composition x50.2, 0.3, and
and time constant, respectively, related to the 0.4 were manufactured at Ceramatec, and verified as
thermodynamic, surface kinetic, and transport prop- phase-pure and impurity-free by XRD, XRF, and
erties of the mixed conductor [13]: SEM / EDS. The powders were made into an ink, and
128 S.B. Adler / Solid State Ionics 111 (1998) 125 – 134

used to fabricate symmetrical LSCO / ceria / LSCO

cells with 2.0 cm 2 active area. The electrodes were
fired in air onto the doped ceria electrolyte (of
proprietary composition) between 900 and 12008C
for between 1 and 8 h. Following firing, the surface
areas of the electrode materials were measured by
BET, and the porosity determined from the electrode
weight and thickness (SEM). The solid-phase tor-
tuosity was estimated from SEM, based on quali-
tative comparison to pore geometries typical in the
catalysis literature [18]. Table 1 shows a summary of
these electrode geometry measurements.
As explained in the Section 5 below, minor
impurities and certain electrolyte dopants can react
with the mixed-conductor, altering its diffusion and
surface exchange properties, or producing insulating
phases at the electrode / electrolyte interface. These
phenomena can cause a dependence of the electrode
kinetics on electrolyte dopant, and non-zero interfa-
cial impedances (Zinterfaces ±0). Much effort was
taken in this study to choose electrolyte materials
that minimize these effects, and to carefully examine
the electrolyte surface and electrode / electrolyte in-
terface by SEM to insure no obvious second-phase
formation had occurred.
Measurements of a.c. and d.c. polarization were
made under controlled atmosphere and temperature Fig. 3. (a) Schematic of the test apparatus used to measure the cell
using the experimental apparatus shown schematical- impedance. (b) Deconvolution of ohmic and non-ohmic contribu-
ly in Fig. 3a. The electrochemical measurements tions to the cell impedance.
were made using a Solartron 1287 potentiostat and
1260 frequency response analyzer, and the oxygen impedance (Z) was typically ohmic or purely induc-
concentration was verified using an in-situ zirconia tive, with a real value R ohmic . Measurements were
oxygen sensor. Measurements were made between made on electrolytes of several different thicknesses
550 and 8008C, and between 1% and 100% O 2 in in order to confirm that R ohmic vs. thickness fits a
blended gases of O 2 / He and O 2 / N 2 . In most cases straight line with zero intercept, and has a slope
three cells were studied and the results averaged. equal to the electrolyte resistivity. The remaining
The electrolyte and electrode contributions to the impedance, Z2R ohmic , typically consisted of a single
cell impedance were deconvoluted as shown in Fig. resonance with a shape similar to Fig. 2b, and was
3b. In the limit of high frequency (.5 kHz), the total taken to be the impedance of the electrode reactions
(Zchem ). Although previous work with this family of
Table 1 perovskites has shown a second, high frequency
Geometric parameters of the electrode as a function of com- interfacial impedance [6,19] (especially at low tem-
position measured or estimated from BET, SEM, and weight
perature), none were observed under the conditions
measurements
investigated here. Where appropriate, Zchem was fit to
x L (mm) a (mm 21 ) ´ t Eq. (2) using non-linear least-squares in order to
0.2 15 1.8 0.5 2.5 extract values for R chem and t chem .
0.3 15 2.3 0.5 2.5 At low gas-phase concentrations of oxygen, the
0.4 15 1.6 0.5 2.5
electrodes become limited by gas-phase diffusion,
 S.B. Adler / Solid State Ionics 111 (1998) 125 – 134 129

Fig. 4. Impedance of a symmetrical LSCO / ceria / LSCO cell

(x50.3 and T57508C) as a function of PO 2 with He and N 2 as the
balance gas.

severely limiting the range of PO 2 we were able to

study. For example, Fig. 4 shows the impedance of
an LSCO / ceria / LSCO cell as a function of PO 2 in
He vs. N 2 . At 21% O 2 , results in He and N 2 are
nearly identical, but at 4%, a repeatable difference of
0.20 V cm 2 appears. A calculation of the gas-phase
diffusion resistance [Eq. (6)] based on the electrode
morphology and known published values for binary
Fig. 5. (a) Impedance of a symmetrical LSCO / ceria / LSCO cell at
and Knudsen diffusivities [18] yields values con-
x50.4, T57508C, P hO h2jj50.21 atm. (b, c) Values of R chem and
sistent with this difference: 0.04 and 0.20 V cm 2 , ] ]
t chem as a function of x, T, and PO 2 , extracted from the impedance
respectively, for 4% O 2 in He and N 2 . For the values data by fitting to Eq. (2). Circles, squares, and triangles indicate
of R chem studied here, this effect can be ignored if x50.2, 0.3, and 0.4, respectively.
helium is used. However, below |1% O 2 (especially
in N 2 or Ar), the impedance becomes dominated by
the gas phase, showing a semicircular response with function of x and T in air, and as a function of x and
a resistance proportional to PO 2 . Similar effects were PO 2 at 7508C. The results show that R chem depends
reported in Ref. [6]. more strongly on temperature than t chem (|140 vs.
|100 kJ / mol, respectively), while t chem depends
21 20.3
more strongly than R chem on PO 2 (P O 2 vs. P O 2 ).
Both R chem and t chem depend strongly on x. As shown
4. Results below, these dependencies can be explained quantita-
tively in terms of the properties of LSCO.
4.1. Electrode kinetics

Fig. 5a shows the impedance response of a 4.2. Extraction of Dv and r0

LSCO / ceria / LSCO cell (x50.4) in air at 7508C.
Above |5 kHz the response becomes purely induc-
tive, with a constant real component, R ohmic (sub-
tracted from the data). The response fits well to Eq.
(2), yielding values for R chem and t chem . The behavior
was similar at all values of x, T, and PO 2 , and Fig. 5b
and c show a summary of R chem and t chem as a
Using Eqs. (3) and (4), the values of Dv and
r 0 (af 1 ab ) were calculated as a function of T, PO 2 ,
and x based on the values of R chem and t chem reported
in Fig. 5. No adjustable parameters were used in this
calculation. The electrode geometry parameters ´, t,
and a were taken from Table 1, while the thermo-
130 S.B. Adler / Solid State Ionics 111 (1998) 125 – 134

dynamic properties of LSCO were taken from the

literature [20,21] and incorporated using the Lan-
khorst rigid-band formalism [22], which succinctly
expresses c v and the thermodynamic factor A in
terms of T, PO 2 , and x:

4
S
m O0 2 (T ) 1 RT lnPO 2 5 Eox (x) 2 ]] 6] 2 x
cv
g(´F ) c mc D
S
2 TSox (x) 2 2RT ln ]]]
cv
c mc 2 c v D
(7)

12 cv
A 5 1 1 ]]] ] (8)
g(´F )RT c mc

where c mc is the concentration of lattice sites in

LSCO (0.091 mol / cm 3 ), Eox and Sox are a standard
Helmholtz energy and entropy, and g(´F ) is the
density of states at the Fermi level, all assumed
independent of T and PO 2 [21]. The results are
shown in Figs. 6 and 7.
Fig. 6a shows the calculated values of Dv as a
function of temperature and x in air. For comparison,
values of Dv 5 D˜ /A calculated from permeation data
[21,23] are also shown. The agreement at high
temperature is good, showing a similar activation
energy and dependence on Sr content. There is some
deviation from Arrhenius behavior at x50.2 and low
temperature, corresponding to a less quantitative fit
of the impedance data to Eq. (2). As discussed
below, this deviation suggests that the model may
begin to break down at low vacancy concentration.
Fig. 6b shows the calculated values of r 0 (af 1 ab )
as a function of temperature and x in air. For
comparison, the values of r 0 (af 1 ab )5kc mc calcu-
lated from surface exchange measurements at x50.2
are shown (open symbols) [24]. For this comparison,
the values of k (measured at PO 2 50.7 atm) were
corrected slightly (to 0.21 atm) based on the PO 2
dependence shown in Fig. 7b, below. Agreement is
reasonable in both absolute value and activation
energy.
Fig. 7a shows the calculated values of Dv as a
function of PO 2 and x at 7508C. This data shows a
weak dependence of Dv on PO 2 , in good agreement
with chemical diffusion measurements (coulometric
titration) [21]. There is a strong dependence on Sr
Fig. 6. Values of Dv and r 0 (af 1 ab ) extracted from the electrode
kinetics as a function of x and T (in air). Circles, squares, and
triangles indicate x50.2, 0.3, and 0.4, respectively. Closed
symbols indicate impedance data. Open symbols indicate values
based on independent chemical diffusion and isotopic exchange
measurements.
content, also supported by chemical diffusion and
permeation measurements [21,23].
Fig. 7b shows the calculated values of r 0 (af 1 ab )
as a function of PO 2 and x at 7508C. In contrast to
Dv , the surface exchange constant depends strongly
on oxygen concentration, varying as P 11 O 2 , and there
is little dependence on Sr content. This strong
dependence on PO 2 , and commensurate weak depen-
dence on x (electron occupancy of the Co3d–O2p
band) suggests the exchange reaction is limited by
O 2 dissociation rather than availability of electronic
or ionic species. The nonlinear behavior at x50.2
may indicate that other steps limit the exchange
reaction at lower vacancy concentration.
 S.B. Adler / Solid State Ionics 111 (1998) 125 – 134 131

material properties is extremely valuable because it

allows us to examine quantitatively some of the
implications of this mechanism, and to examine and
gain insight from the limitations and deficiencies of
the model.

5.2. Size of the extension region

One question we can address is how far the

reaction extends from the electrolyte / electrode inter-
face. This distance is characterized by D, calculated
from Eq. (5) using the values of Dv and r 0 (af 1 ab )
reported above. Fig. 8 show the results of this
calculation as a function of x, T and PO 2 . These
results show that at under realistic operating con-

Fig. 7. Values of Dv and r 0 (af 1 ab ) extracted from the electrode

kinetics as a function of x and PO 2 (at 7508C). Symbols are as in
Fig. 6.

5. Discussion

5.1. Reaction mechanism

The results presented above suggest that O 2

reduction on a porous La x Sr 12x CoO 32d electrode is
limited primarily by surface chemical exchange and
solid state diffusion. This conclusion stands in
contrast to the traditional view that the electrode
reaction is limited by a charge-transfer process,
presumably occurring at the three phase boundary.
Indeed, this analysis shows that the electrode kinetics
can be explained quantitatively, with no adjustable
parameters, in terms of independently measured
Fig. 8. Values of the extension region, D, as a function of x, T, and
transport and chemical-kinetic properties of the PO 2 , calculated based on the values of Dv and r 0 (af 1 ab ) in Figs.
mixed conductor. 6 and 7. Circles, squares, and triangles indicate x50.2, 0.3, and
This link between the electrode kinetics and the 0.4, respectively.
132 S.B. Adler / Solid State Ionics 111 (1998) 125 – 134
ditions, the reaction is not confined to the three-
phase boundary, but may extend a distance ap-
proaching the electrode thickness (15 mm). As one
might expect, d has a strong dependence on x and
PO 2 since the active region grows with increasing
vacancy concentration or lower r 0 . There is a much
weaker or inverted dependence on T since Dv and
r 0 (af 1 ab ) have similar activation energies, and thus
tend to cancel each other in Eq. (5).
In doing this analysis, we found evidence that Eq.
(2) begins to break down at exceptionally small and
large values of d. This effect is illustrated in Fig. 9a,
which shows the shape of the impedance as a
function of d, normalized to its value at d 52 mm.
When d is moderate (2 mm), the impedance has a
shape closely matching Eq. (2). However at small
d (,0.7 mm), the shape of the impedance becomes
less well-defined, more like a depressed semicircle.
Fig. 9. (a) Shape of the impedance as a function of D for a high
surface area electrode (x50.3, T57508C, varying PO 2 ). (b)
Comparison of the impedance at high and low surface area
(x50.4, T57508C, PO 2 50.21 atm). Schematics of the corre-
sponding steady-state reduction profiles in the electrode are
illustrated.
A breakdown of the model for small d is consistent
with its formal assumption that the active region is
much larger than the average particle size d p 53 /A
[13]. However, it is unclear if this breakdown is
caused by the two-dimensionality of the electrode
geometry under these conditions, or the presence of a
second, parallel reaction mechanism. For example,
surface diffusion might become important at low
vacancy concentration, explaining the apparent over-
prediction of Dv for x50.2 at low temperature (Fig.
6a).
At large values of d (.6 mm), the impedance also
appears slightly distorted from Eq. (2), which we
expect to occur as d approaches L / 3 (|5 mm). This
effect is more clearly illustrated in Fig. 9b, which
compares the response of a low surface area elec-
trode (a,0.1 mm 21 ), made by firing the electrode to
higher temperature, normalized to the response of an
electrode of high surface area (a52.0 mm 21 ). The
response of the low-surface area electrode is nearly
semicircular, with only slight evidence of diffusive
behavior at high frequency.
This distortion can be understood by re-examining
the reaction / diffusion limit of the model, but relax-
ing the assumption of semi-infinite electrode thick-
ness. With a finite electrode thickness L, the non-
charge-transfer term in equation 12 of Ref. [13]
becomes:
]]]]
œ
1 L ]]]
S
Zchem 5 R chem ]]]]uTanh ]œ1 1 jv t chem .
1 1 jv t chem d
D
(9)
As L approaches infinity, the Tanh(L /d ) term ap-
proaches unity, yielding Eq. (2). As illustrated in
Fig. 1b, the relevant length parameter in this limit
becomes d, which governs the rates of both solid-
state diffusion and surface exchange. On the other
hand, as d becomes larger than L, the Tanh(L /d )
term in Eq. (9) becomes linear, yielding:
RT 1
Zchem 5 ]]2 ]]]]]]]] (10)
2F aLr 0 (af 1 ab )(1 1 jv t chem )
which is a semicircle with a characteristic resistance
related only to surface exchange (1 /aLr 0 ). The
relevant length parameter in this limit becomes L,
corresponding to the case where solid-state diffusion
 S.B. Adler / Solid State Ionics 111 (1998) 125 – 134
is fast, and the entire internal surface area of the
electrode (aL) is utilized for O 2 reduction.
A fit of the low-surface-area data in Fig. 9b to Eq.
(10) (assuming the same material properties as for
the high surface area electrode) yields a value of
a50.05 mm 21 , which is consistent with estimates
based on BET data (0.01|0.1 mm 21 ). The calculated
value of D based on this area is 30 mm (twice the
electrode thickness), indicating that the entire elec-
trode is indeed utilized.
5.3. Electrolyte effects
The electrode model embodied in Eq. (2) predicts
that the electrode kinetics are independent of the
electrolyte material. However, ongoing observations
at Ceramatec and elsewhere show various dependen-
cies on electrolyte composition and cell processing.
These effects have been interpreted by some authors
as ‘proof’ that the mechanism of the reaction must be
limited by a charge-transfer process at the three-
phase boundary [3]. We believe this interpretation is
oversimplified, and offer two alternative explana-
tions, neither of which require invoking a localized
three-phase boundary mechanism.
First, we have observed that during firing,
ionically-insulating phases can form at the interface
between electrode and electrolyte, limiting oxide-ion
transfer and creating a secondary high frequency
charge-transfer impedance (Zinterface ). It has been
known in the literature for many years that perov-
skites in the family (La,Sr)(Mn, Co, Fe)O 32d can
react with yttria-stabilized zirconia (YSZ) to form
insulating La-Zr-O and Sr-Zr-O phases that degrade
electrode performance. Recent data from Liu and Wu
[3], appear to show this effect for
La 12x Sr x Co 12y Fe y O 32d (LSCF) electrodes on YSZ,
as well as for LSCF on Sr-doped ceria. We have also
observed this effect with LSCO on ceria when the
ceria contains certain types of dopants (details pro-
prietary). Small amounts of second phase that are not
detectable by SEM or XRD are sufficient to cause
this effect. Interfacial resistance can also be caused
by impurities at the interface, and is pronounced at
low temperature [25].
Second, reaction and interdiffusion of electrode
and electrolyte occur during bonding, and this may
have an impact on both electrolyte and electrode
133
properties. We have observed strong effects of cer-
tain types of ceria dopants on the surface-exchange
properties of LSCO. This effect is seldom homoge-
neous, and thus alters the shape as well as the
magnitude of the impedance. Data from Liu and Wu
[3] show large effects on the electrode impedance
with Ba and Bi as electrolyte dopants. These ele-
ments are likely to be highly reactive with LSCO.
We have also observed that certain impurities intro-
duced during bonding may influence the diffusion
properties of LSCO. Indeed, Bae and Steele (and
others) have seen strong effects on electrode prop-
erties of Si contamination from the electrolyte or
from processing [26,27]. Finally, for materials hav-
ing small d, doping of the electrolyte by elements in
the electrode may create mixed-conduction in the
electrolyte, altering the effective active area [28].
6. Conclusions
The measurements presented in this paper suggest
that O 2 reduction on a porous La x Sr 12x CoO 32d
electrode is limited primarily by surface chemical
exchange and solid state diffusion. This conclusion
stands in contrast to the traditional view that the
electrode reaction is limited by a charge-transfer
process, presumably occurring at the three phase
boundary. Indeed, this analysis shows that the elec-
trode kinetics can be explained solely in terms of the
transport and chemical-kinetic properties of the
mixed conductor, with all charge-transfer steps as-
sumed to be equilibrated.
For typical electrode geometries, the utilization
region (d ) over which the reduction reaction occurs
extends several microns from the electrode / elec-
trolyte interface, approaching the electrode thickness
under some conditions. For electrodes of low surface
area (,0.1 mm 21 ), the reaction extends over the
entire surface of the electrode, and is limited almost
entirely by surface chemical exchange. These dimen-
sions show that the material and geometry of the
electrode far from the three-phase boundary are
important in determining overall performance.
Below O 2 concentrations of |0.01 atm, the elec-
trode kinetics become dominated by gas-phase /
Knudsen diffusion in the pores of the electrode. This
observation can be explained quantitatively based on
134 S.B. Adler / Solid State Ionics 111 (1998) 125 – 134
the geometry of the electrode and the properties of
the gas. For further details, including the theoretical
effects of combined solid-state and gas-phase diffu-
sion, and diffusion in a boundary layer outside the
electrode, the reader is referred elsewhere [6,13].
Finally, we point out observations by ourselves
(and others) of a dependence of the electrode kinetics
on the type of electrolyte and conditions of process-
ing. We do not consider these effects to be evidence
for a charge-transfer-limited reaction mechanism.
Rather, we believe these effects are caused by
reactions between the electrode and electrolyte ma-
terials, forming ionically insulating phases or
modifying the properties of the materials. Although
parallel, more localized or charge-transfer-limited
mechanisms may be present (as is likely with LSM /
ceria [13]), the data and analysis presented here
suggest they are not rate-limiting for mixed-conduc-
tors with high rates of solid-state oxygen transport.
Acknowledgements
This material is based upon work supported by the
North Atlantic Treaty Organization under a Grant
awarded in 1993 (NSF-NATO). Samples and ex-
perimental facilities were provided by Ceramatec,
Special thanks go to Marc Flinders and Dave Padgen
for aiding in materials characterization and electro-
chemical measurements, and Jon Lane and John
Newman for helpful discussions. Thanks also go to
the University of Twente for copies of the Ph.D.
theses of Martijn Lankhorst and Rene´ van Doorn.
References
[1] F.H. van Heuveln, H.J.M. Bouwmeester, J. Electrochem.
Soc. 144(1) (1997) 134.
¨
[2] M. Godickemeier, K. Sasaki, L.J. Gauckler, I. Reiss, J.
Electrochem. Soc. 144(5) (1997) 1635.
[3] M. Liu, Z. Wu, Solid State Ionics 107 (1998) 105.
[4] M. Kleitz, F. Petitbon, Solid State Ionics 92 (1996) 65.
[5] M. Kleitz, L. Dessemond, T. Kloidt, M.C. Steil, in: M.
Dokiya, O. Yamamoto, H. Tagawa, S.C. Singhal (Eds.),
Proceedings of the Fourth International Symposium on Solid
Oxide Fuel Cells, The Electrochemical Soc., New Jersey,
1995.
[6] J.A. Lane, S. Adler, P.H. Middleton, B.C.H. Steele, in: M.
Dokiya, O. Yamamoto, H. Tagawa, S.C. Singhal (Eds.),
Proceedings of the Fourth International Symposium on Solid
Oxide Fuel Cells, The Electrochemical Soc., New Jersey,
1995.
[7] Y. Takeda, R. Kanno, M. Noda, Y. Tomida, O. Yamamoto, J.
Electrochem. Soc. 134(11) (1987) 2656.
[8] R.H.E. van Doorn, PhD Thesis, Twente University, 1996.
[9] J.E. ten Elshof, H.J.M. Bouwmeester, H. Verweij, Solid State
Ionics 81 (1995) 97.
[10] E. Siebert, A. Hammouche, M. Kleitz, Electrochem. Acta
40(11) (1995) 1741.
[11] H. Lauret, A. Hammou, J. European Ceram. Soc. 16(4)
(1996) 447.
[12] M. Liu, J. Electrochem. Soc. 145 (1998) 142.
[13] S.B. Adler, J.A. Lane, B.C.H. Steele, J. Electrochem. Soc.
143(11) (1996) 3554.
[14] H. Deng, M. Zhou, B. Abeles, Solid State Ionics 74 (1994)
75.
[15] A.M. Svensson, S. Sunde, K. Nisanciuglu, J. Electrochem.
Soc. 145 (1998) 1390.
[16] S.B. Adler, J.A. Lane, B.C.H. Steele, J. Electrochem. Soc.
144(5) (1997) 1884.
[17] S. Adler, J.A. Lane, B.C.H. Steele, in: H.U. Anderson, A.C.
Khandkar, M. Liu (Eds.), Proceedings of the Second Interna-
tional Symposium on Ceramic Membranes, PV 94-12, p.
159, The Electrochemical Soc., New Jersey, 1995.
[18] C.N. Satterfield, T.K. Sherwood, The Role of Diffusion in
Catalysis, Addison-Wesley Pub. Co., 1963.
[19] H.U. Anderson, L.-W. Tai, C.C. Chen, M.M. NAsrallah, W.
Huebner, in: M. Dokiya, O. Yamamoto, H. Tagawa, S.C.
Singhal, (Eds.), Solid Oxide Fuel Cells IV, PV 95-1, p. 375,
The Electrochemical Society Proceedings Series, Pennington,
NJ, 1995.
[20] J. Mizusaki, Y. Mima, S. Yamauchi, K. Fueki, J. Solid State
Chem. 80 (1989) 102.
[21] M. Lankhorst, Ph.D. Thesis, University of Twente, En-
schede, The Netherlands, 1997.
[22] M.H.R. Lankhorst, H.J.M. Bouwmeester, H. Verweij, Phys.
Rev. Lett. 77(14) (1996) 2989.
[23] R.H.E. van Doorn, Ph.D. Thesis, University of Twente,
Enschede, The Netherlands, 1996.
[24] H.J.M. Bouwmeester, H. Druidhof, A.J. Burgraaf, Solid
State Ionics 72 (1994) 185.
[25] J.A. Lane, private communication, 1997.
[26] J. Bae, B.C.H. Steele, Solid State Ionics 106 (1998) 247.
[27] B.C.H. Steele, private communication, 1998.
[28] J.D. Sirman, J.A. Kilner, J. Electrochem. Soc. 143 (1996)
L229.