.1 Electrochem. Soc., Vol. 145, No. 7, July 1998 The Electrochemical Society, Inc.

20. W Li, J. N. Reimers, and J. P. Dahn, Phys. Rev. B, 49,

826 (1994).

21. J. N. Reimers and J. R. Dahn, Phys. Rev. B, 47, 2995

(1993).

22. M. K. Aydinol, A. F. Kohan, G. Ceder, K. Cho, and J.

Joannopoulos, Phys. Rev. B, 56, 1354 (1997).

23. D. J. Singh, Planewaves, Pseudopotentials, and the

LAPW Met hod, Kiuwer, Boston, MA (1994).

24. Solid State Batteries: Materials Design and Optimiza-

tion, C. Julien and G.-A. Nazri, Editors, Chap. 1,

Kiuwer, Boston, MA (1994).
25. P Hohenberg and W. Kohn, Phys. Rev. B, 136, 86

2431

26. IV Kohn and L. J. Sham, Phys. Rev. A, 146,

1133

(1965).

27. D. M. Ceperley and B. J. Alder, Phys. Rev. Lett., 45, 566

(1980).

28. J. P Perdew and A. Zunger, Phys. Rev. B, 23, 5048

(1981).

29. S. Froyen, Phys. Rev. B, 39, 3168 (1989).

30. C. Wolverton and A. Zunger, Phys. Rev. B, 57, 2242

(1998).

31. E. Deiss, A. Wokaun, J.-L. Barras, C. Daul, and P

Dufek, J. Electrochem. Soc., 144, 3877 (1997).

32. C. Chieh, B. L. Chamberland, and A. F. Wells, Acta

(1964).

Crystalloyr., B37, 1813 (1981).

Gas Conversion Impedance: A Test Geometry Effect in

Characterization of Solid Oxide Fuel Cell Anodes
S. PrimdahI and M. Mogensen*
Ris National Laboratory, Materials Research Department, 4000 Roskilde, Denmark

ABSTRACT

The appearance of an extra arc in impedance spectra obtained on high performance solid oxide fuel cell (SOFC)
anodes i recognized when experiments are conducted in a test setup where the working and reference electrodes are

placed in separate atmospheres. A simple continuously stirred tank reactor (CSTR) model is used to illustrate how anodes
measured with the reference electrode in an atmosphere separate from the working electrode are subject to an impedance
contribution from gas conversion. The gas conversion impedance is split into a resistive and a capacitive part, and the
dependences of these parameters on gas composition, temperature, gas flow rate, and rig geometry are quantified. The
fuel gas flow rate per unit of anode area is decisive for the resistivity, whereas the capacitance is proportional to the CSTR
volume of gas over the anode. The model predictions are compared to actual measurements on Ni/yttria stabilized zirconia cermet anodes for SOFC. The contribution of the gas conversion overpotential to dc current-voltage characteristics is
deduced for H2/H20 and shown to have a slope of RT/2F in a Tafel plot.
Infroduction

A detailed understanding of the rate-limiting steps in

state-of-the-art Ni/YSZ cermet SOFC anodes is highly desirable from an optimization point of view. One of the most
promising ways of unfolding such complex systems is by
impedance spectroscopy. Here a number of more or less
well separated arcs representing at least the same number
of processes can be analyzed individually as a function of
the primary test conditions.
Ni/YSZ cermet anodes for SOFC have been investigated
intensively, and impedance spectra and interpretations of
these have been offered in the literature.1 The findings
are surprisingly inconsistent, judging from the number of
reported arcs in impedance spectra, even considering the
different anode structures and test conditions applied. In
this paper it is demonstrated how part of the disagreement
may originate in the measurement setup rather than in the
examined electrodes,
Considering a single electrode on an electrolyte (i.e., a
half cell), the electrode overpotential 'flEl,ctmd, is given as the

electrode potential E(i) at current density i minus the equilibrium potential E5q (Nernst potential = zero current
potential) corrected for the potential loss in the electrolyte
of series resistance R
1EI,ctrod, =

E(i)

One of the important assumptions in the above definitions is a sufficiently high flow rate of reactants to the cell
to suppress concentration changes above the surface of the
working electrode. It is the breakdown of this assumption
which is examined below.
When a working electrode and a reference electrode are
placed in different atmospheres, all electrochemical measurements involve the Nernst potential E, Eq. 3, between
the oxygen partial pressures of these atmospheres

4F

[3]
x0201

Here R, is the gas constant, T is the absolute temperature,

and Fis Faraday's constant. x02 refers to the mole fraction
of 02 in the reducing and oxidizing atmospheres, respectively. Variation in x>4n each of the compartments is a
result of the gas flow rateper anode area, conversion by
passing current, and gas mu1x by diffusion, convection,

and stirring. This phenomeno has already been considered and measured for dc loading of fuel cells.11'2 The
variation in the Nernst potential is inherent to the passage
of current, and must be anti\cipated also in impedance
measurements. It is here suggested to use the term gasconversion impedance.

[1]

It is the aim of this paper to (i)dcraw attention to the

This electrode overpotential is the summation of overpotentials of k rate-limiting processes in the electrode, Eq. 2.
Processes like chemical reactions, charge transfer, and diffusion in the electrode are typically to be considered

presence of gas conversion impedance in SOFC research,

and (ii) quantify the nature of gas conversion impedance
and predict the ac and dc responses by introducing a simple CSTR model. Finally it is important to note how the
presence of gas-conversion impedance is dependent on the
measurement setup.

E5q

iR5

flEl,etrode =

[2]
1

* Electrochemical

Society Active Member,

Experimental
Sample preparationA 40 to 50 p.m thick Ni/YSZ cer-

met anode is prepared by spray painting. The ethanol

based slurry is composed of green Ni0 and YSZ (Tosoh,

Downloaded 28 Sep 2009 to 142.150.190.39. Redistribution subject to ECS license or copyright; see http://www.ecsdl.org/terms_use.jsp

2432

J. Electrochem. Soc., Vol. 145, No. 7, July 1998 The Electrochemical Society, Inc.

Zr02 with 8 mol % Y203), suspended by a dispersant,

poly(vinylpyrrolidone).'3 After deagglomeration in a ball
mill, the YSZ particle size is about 0.4 p.m. The NiO particle size distribution is bimodal with fractions of approximately 0.4 and 10 p.m in the volume ratio 6:1. The slurry is
sprayed onto YSZ three-electrode pellets (see Fig. 1) and
on both sides of YSZ tapes in layers of 10 to 20 p.m and

Springloaded
alumina rod

sintered for 2 h at 1300C after each layer. The Ni/YSZ volume ratio is 40/60 in the reduced cermet.

Pt current collector

A second type of Ni-based anode has been applied to a

three-electrode pellet. Here Ni(N03)2 is decomposed in
moist 9% H2 at 1420C on the electrolyte,'4 and afterward a
90/10 vol % Ni/YSZ paint is deposited without sintering. A

Electrolyte foil

Electrode

sufficient contact and anode stability are obtained. This

anode is intended to have active sites only on the interface
toward the electrolyte, the 10% YSZ is added to restrain Ni
agglomeration by sintering and is not expected to percolate.
Test setup designs.Two different test setups are used
in characterization of these anodes. The YSZ three-elec-

Pt current collector

-- Aluminabase

trode pellets are provided with a Pt-paste counter elec-

trode and a Pt-ball reference electrode and are mounted in

a three-electrode setup, Fig. 1. This setup provides a ref erence atmosphere and a sealed cylindrical anode compartment for exposing the anode to a reducing fuel gas. The

fuel gas is fed through a concentric tube in the anode

chamber. The reference electrode is placed in air (reference

gas), and a Nernst potential is measured. The working

electrode of 0.44 cm2 is contacted with several millime-

ters of 0.3 mm Pt wire pressed against the center. The

height from the anode surface to the feed tube is approximately 1 cm. The cross-sectional area of the anode compartment is 0.50 cm2. The electrolyte geometry used here is
verified by finite-element analysis'5 to have a frequency
stable, uniform current distribution, causing less than 3%
error on a polarization resistance of 0.16 Cl cm2.
The symmetric cells with two equal anodes on each side
of a 180 to 200 p.m thick YSZ tape are measured in a 1 atm
setup. The samples are broken into 5 x 5 mm squares
and mounted as a two-electrode cell between grids of Pt
current collecting wires, Fig. 2. Due to the symmetric ar-

Fig. 2. Test setup for symmetric electrodes in a common atmosphere. Typical sample areas are 0.25 cm2, the Pt wires are 0.3 mm

in diameter with a grid distance of 1 to 2 mm.

rangement without reference electrodes, the measured im-

pedance represents the sum of the electrode responses, and

interpretable data can be obtained only in the linear current-overpotential range around the open-circuit voltage
(OCV).

Electrochemical testing.Anodes are heated to 1000C

in nitrogen and reduced by changing directly to a measurement atmosphere of 97% H2 + 3% H20 with an OCV
against Pt/air of about 1070 mV (p02 2 X 10-18 atm).
Gas flow rates in the range of 10 to 200 mL/min (at 25C)
have been applied. When nothing else is indicated, a total
of 100 mL/min (at 25C) dry gas is passed over the anodes.

Spring loaded rod

Au seal
Fuel gas

Working electrode
Reference electrode

Atmosphere control for variation of the mole fraction of

each gas x (i = incoming) at constant gas flow rate is
performed by mixing nitrogen and hydrogen through thermal mass flowmeters and saturating this mixture with
water at a controlled temperature. An experiment has
been conducted where nitrogen has been substituted by
argon and helium at constant gas flow rate and mole fractions of hydrogen and water.
x02 is determined by the Nernst expression from OCV vs.
Pt/air. The partial pressure XH2O is determined from x02
and XH, by the following algorithm valid for 1000 K < T <
1300 K 16

Au seals

Electrolyte pellet
Pt seal

Counter electrode

Reference gas

5 mm

XH2O = 10_(2958l3022) XH5

[4]

Impedance spectroscopy is conducted with an applied

amplitude in the range of 7 to 70 mV; in general a value of
20 mV is used. This amplitude is verified to be well inside

the linear range. The frequency range is 500 kllz to

0.01 Hz or in some cases only 50 kHz to 0.01 Hz. At least

20 mm is allowed for stabilization before recording each
impedance spectrum.
The equipment used for impedance spectroscopy is a
Solartron 1250 frequency response analyzer (FRA) combined with a Solartron 1286 ECI or a Solartron 1260 FRA

with a Solartron 1287 ECI. The current-measurement

resistor used is 1 Cl for polarized and 10 11 for unpolarized
measurements. The two systems are verified on an appro-

priate test circuit to reproduce data within less than 1%.

Fig. 1. Three-electrode pellet (white body in center) and test
setup for three-electrode measurements, where working electrode
(anode) and reference electrode are in separate atmospheres. The
anode area is 0.44 cm2, and the volume of the tube enclosing the
anode is 0.50 cm3/cm.

Results and Discussion

spectra.Impedance spectra obtained on

nominally identical Ni/YSZ cermet anodes in the two
measurement setups presented here differ considerably,
Fig. 3. The anode on a three-electrode pellet measured in
Impedance

Downloaded 28 Sep 2009 to 142.150.190.39. Redistribution subject to ECS license or copyright; see http://www.ecsdl.org/terms_use.jsp

2433

J. Electrochem. Soc., Vol. 145, No. 7, July 1998 The Electrochemical Society, Inc.

Table I. Parameters for the low frequency semicircle in Fig. 4.

0.10

A one atmosphere
0 two atmospheres

o__

Applied parameters are: A 0.44 cm2, J = 1.62 mol/m2 s, T =

1000C, x, = 0.03, x, = 0.97

Hz

0-'

Sample

0/0

0.05

10kHz

100Hz d"

R3 (fI cm2)

A214, Ni/Y SZ cermet anode

A527, Ni-based anode

0.20
0.18

C3 (F/cm2) f (Hz)
0.74
0.80

1.1

1.0

limiting processes in a series arrangement. Data are thus

fitted with the software EQTJIVCRT,17 using the equivalent
circuit LRS(RQ)l(RQ)2(RC)3 where the annotation 1 to 3 is

0.00

0.05

0.10

0.15

0.20

0.25

Z., 0cm2

fig. 3. Impedance spectra obtained on nominally identical

anodes in a 1 atm setup () and in a 2 atm setup (0). Measurements in 97% H2 + 3% H20 at 1000C at OCV. The spectra are corrected for anode area and series resistance for presentation purpose. Nine impedance points per frequency decade.

the 2 atm setup contains at least three arcs. A small arc is

located between the two dominant ones and is not clearly

seen at these conditions. The symmetric sample measured
in the 1 atm setup exhibits only two well-defined arcs.
Comparison of the highlighted frequencies shows that the

dominant arc at low frequency measured in the 2 atm

setup is not present in the measurements performed in the

1 atm setup.
A comparison of impedance measurements obtained on

two structurally different anodes in the 2 atm setup is

given in Fig. 4, The spectra are brought to the same scale

and shifted in real part to have the same resistance at low
frequency. It is seen that these two anodes measured in the
same 2 atm setup exhibit identical arcs at low frequencies.
Interpretation of impedance spectra.To elaborate further on the behavior of the low frequency arc present in
measurements in the 2 atm setup, it is necessary to apply
a simple model (an equivalent circuit) to present the following results in a meaningful and representative way.
The low frequency arc is observed to describe a nearly perfect semicircle, which allows a description by only two
parameters; a resistor R and a capacitor C in parallel.
In the present approach we assume that the impedance
spectrum can be interpreted by three independent rate-

assigned to the respective processes by increasing time

constant. Here L is an inductance, R a resistance, C a
capacitance, and Q a constant-phase element given as Q =
Y0(jw)". Y0 is an admittance, j the imaginary unit, w the
angular frequency, and n is a frequency power. The frequency powers (n1, n2) = (0.8, 0.75) were derived earlier3
for Ni/YSZ cermet anodes from a data matrix to produce

consistently high quality fits under all examined conditions (<2% relative error on both real and imaginary data
at all frequencies) and are applied again here. For the Nibased anodes, frequency powers of (n1, n2) = (0.9, 0.75) are
found suitable.

Data obtained on symmetrical anodes with both electrodes in a common atmosphere (Fig. 2) contain only two
arcs, Fig. 3. Comparison of the summit frequencies indicates that the low-frequency arc denoted 3 is not present
in this measurement geometry.
The parameters R3 and C3 for the low-frequency arcs in
Fig. 4 are given in Table I as well as the summit frequency
13 according to Eq. 5
1

[5]

2irR.C3

The low frequency arc is observed to have a dependence on

gas flow rate in the range of 10 to 200 mL/min. In Fig. 5

the dependence for the parameters R3 and C3 is presented.
The effect of the molecular size of the inert gas on the
low frequency arc has been measured by using helium,
argon, and nitrogen at a constant flow rate and temperature with x, = 0.02 and x1120 = 0.035. R3 is presented in

Table IL
Varying the mole fraction of water XH9 from 5 x 10 to
0.25 at constant flow rate with a constant XH3 of 0.5 or 0.97
to 1.00, respectively (the balance being nitrogen), reveals a
significant variation in the low frequency arc. In Fig. 6 the
parameters R3 and C3 are presented as function of x31,,0. C3
can be very high, up to several farad per square centiii'ieter

20E
U-

0
0.2 fIcm2

z, ccm
Fig. 4. Impedance spectra obtained in setup with anode and ref-

erence electrode in separate atmospheres on a Ni/YSZ cermet

anode (+) and a Ni based anode (0). Measurements in 97% H2 +
3% H20 at 1000C at OCV. The spectra are corrected for anode
area and to have same real component at low frequency in order
to compare the law-frequency arcs. Nine impedance points per
frequency decade.

0.0

0.5

1.0

0
1.5

2.0

2.5

3.0

3.5

Gas flowrate J, mol/m2s

Fig. 5. Experimental impedance data for R3 () and C3 (0) after

fitting to the selected equivalent circuit as a function of gas flow
rate J. Temperature is 1000C, atmosphere is 97% H2 + 3% 1120.
The model prediction of gas conversion impedance is given as
lines, V = 0.35 cm3.

Downloaded 28 Sep 2009 to 142.150.190.39. Redistribution subject to ECS license or copyright; see http://www.ecsdl.org/terms_use.jsp

2434

J. Electrochem. Soc., Vol. 145, No. 7, July 1998 The Electrochemical Society, Inc.

Table II. Resistivity of the low frequency impedance arc after a

simple interpretation as a parallel RC subcircuit. Applied

I. H2OII. H20

parameters are: = 1.62 mourn2 s, T = 1318 K, x,120 = 0.035,

fI2 0.02.

Balance of gas

Inert = He
Inert = Ar
Inert = N9

R, (l1 cm')

CSTR volume

0.12
0.21
0.20

P V9 XH2, XH2OI I

Anode surface

at high water contents. A capacitance of this magnitude is

not ascribable to interfacial capacitance or adsorption on
surfaces; as calculated earlier,3 a bulk process is required.
In combination with the dependence of both R3 and C3 on
gas flow rate, it is assumed that this bulk process could be
gas conversion in the volume over the anode. This assumption is examined in the following model.

ElectrOMe

Fig. 7. The CSTR model. A volume V over the anode surface is

assumed to have uniform gas composition. The considered gas
flow rates J are indicated.

AC modeLThe impedance relating to the change of

atmospheric composition over the anode has been considered before,'8'9 and is derived analytically as follows. An

anode of area A with a volume V over the anode in an

atmosphere separated from the reference gas is consid-

ered, Fig. 7. The gas flows are treated as area specific flow
rates, J (unit of mourn2 s), with subscripts i = incoming, o
= outgoing, and a = anode reaction. In the following, the
subscript is expanded to include the species (H2, H20) considered. A CSTR model is applied, based on the assurnption of perfect mixing of gases at 800 to 1000C due to the
fairly high diffusion rates.2 Therefore the mole fraction of
any species in the volume V is identical to the mole fraction in the outlet gas: x = x0. The current is defined as positive in the anodic direction. A H2/H20 containing fuel gas
with a content of inert gas is considered.
The electrode is perturbed by a complex ac current signal, and the resulting variation in gas composition /XH,O
and Ax,, in the CSTR volume is evaluated and inserted in

a linearized Nernst expression to yield the area-specific

impedance Z9. Details of the calculation are given in
Appendix A. The impedance is broken down into a parallel resistor Rg and capacitor C9 with the dependences given

in Eq. 6 and 7. The criteria for linearization are

XH,o and &rH, <<XIH,, i.e., small perturbations compared to

the inlet gas composition

R9

+
4F2J1 XIH,o

[6]

X,H,

J = iexp.ot)/2F

Cg =

4F2PV
(RT)2A

__ + _L_
X,H,o

[7]

X,H,

Here XH, and x1,0 are the inlet mole fractions of hydrogen and water, respectively, .1, is the inlet area-specific gas
flow rate (mourn2 s), P the absolute pressure, A the anode
area, and V the CSTR volume of the measurement setup.

R, and C9 are thus parameters with physical meaning

and are obtainable by a simple interpretation of an arc in
impedance measurements. The summit frequency of the
impedance arc according to Eq. 5 is
fg = J1ART
2irPV

[8]

DC modelUsing the above described model and definition of parameters, the same type of calculation is carned out for a steady-state dc case. Details of the calculation are given in Appendix B.
The general steady-state expression for ig is given in
Eq. 9, valid for iXH, <XH, anodic and for XH,O <XH90 cath-

odic. At the boundaries, a limiting current density is found

g = RT

(X,H,
lnl

+ i/J 2F I

+ ln X1,3
[XH, i/J2F J]
[ X,HJ

[9]

For the criteria XI,H,>> XH, and AXH,O >> X,,,,,o Eq. 10 is

obtained, revealing a Tafel behavior with a limiting slope

of RT/2F at high currents when the feed gas is rich in

hydrogen and lean in water

RT
.
= ln(2F
2x8820) +..1
2F ln(i)

2F

[10]

Using, instead, the criterion AXH,O <<Xi,H,O and AXH, <<

XIH,, Eq. 11 is obtained. This linear dependence of overpo1

.0
E

c
m

0.1

0.1

tential on current is valid only for small perturbations as

compared to X1 H20 and XiH,. Equation 11 is seen to be consistent with Eq. 6 of the ac model

RT( 1
XH,o
____]

[11]

Evaluation of the model.Gas conversion is taking

0.01

0.01

-2.5

-2.0

-1.5

-1.0

-0.5

Iog(xH2O)

Fig. 6. Experimental impedance data after fitting to the selected

equivalent circuit, R3 (A) and C3 (0) as a function of mole fraction
of water, x. Solid points are at XH2 = 0.97, open points at XH2 =
0.50. Temperature is 1000C, the gas flow rate is 1.62 mol/m' s.
The model predictions of R9 and C9 are given as lines, V =
0.35 cm3.

place on both working and counter electrode. In the above

it is assumed that the reference potential is constant, i.e.,
not influenced by gas conversion.
In the setup with 2 atmospheres used here, the reference
electrode is situated in the same atmosphere as the counter
electrode. If the potential at the reference electrode is not
affected by gas conversion at the counter electrode, only
gas conversion on the test electrode is included in the elec-

trochemical measurements performed in this setup.

Calculating the variation in mole fraction of oxygen at the
counter electrode in air, one obtains an insignificant tx0,

Downloaded 28 Sep 2009 to 142.150.190.39. Redistribution subject to ECS license or copyright; see http://www.ecsdl.org/terms_use.jsp