Electrochimica Acta 50 (2005) 2469–2474

An electrochemical impedance spectroscopy study of fuel cell electrodes

E. Bradley Easton∗ , Peter G. Pickup
Department of Chemistry, Memorial University of Newfoundland, St. John’s, NL, Canada A1B 3X7

Received 22 August 2004; received in revised form 19 October 2004; accepted 19 October 2004
Available online 30 November 2004

Abstract

Electrochemical impedance spectroscopy (EIS) was used to study the capacitance and ion transport properties of fuel cell catalyst layers.
It was found that limiting capacitance correlates with active area. The capacitance per gram of catalyst was calculated and is proposed as a
measure of catalyst utilization. Results obtained with catalyst layers immobilized on glassy carbon electrodes agree very well with results
obtained with gas diffusion electrodes. EIS was also used to study ion conductivity and active area in fuel cell electrodes that contain the
electroactive probe Os(bpy)3 2+ . Together, these results validate the hypothesis that the non-ideal behavior of fuel cell electrodes is due to
variations of conductivity across the layer, rather than variations in capacitance.
© 2004 Elsevier Ltd. All rights reserved.

Keywords: Fuel cell; Electrochemical impedance spectroscopy; Ion transport

1. Introduction cellent electronic conductors, we also assume that Relectronic

Electrochemical impedance spectroscopy (EIS) is a pow-
erful technique that can provide a vast amount of information
in a short period of time and has been applied to proton ex-
change membrane fuel cells (PEMFC) in a number of recent
studies [1–9]. The cathode impedance typically dominates
the PEMFC impedance when pure hydrogen is used at the an-
ode. Previous work in our lab has shown that the impedance
of fuel cell electrodes closely corresponds to that of porous
electrodes, which can be modeled by a finite transmission-
line equivalent circuit, shown in Fig. 1.
The model is described in detail elsewhere [10]. Briefly,
the circuit consists of two parallel resistive rails, one
for electron transport through conducting carbon particles
(Relectronic ), the other for ion transport in the catalyst layer
(Rionic ). The rails are connected by capacitors representing
the differential capacitance of the electrolyte-catalyst (Pt and
C). We assume that this capacitance is uniformly distributed
throughout the catalyst layer. Since carbon particles are ex-
∗ Corresponding author.
E-mail addresses: brad.easton@dal.ca (E.B. Easton),
ppickup@mun.ca (P.G. Pickup).
is negligible compared to Rionic , and therefore should have
a negligible effect on the impedance of the electrode [4,11].
We can also subtract the electrolyte resistance (Relectrolyte )
since it would only shift the plot along the real axis. Based
upon these assumptions, the impedance response of a PEMFC
electrode can be simulated for any ionic conductivity profile
[10,12]. Noticeably absent from this model is charge trans-
fer resistance. This is because all impedance experiments are
performed under nitrogen as opposed to oxygen, which sim-
plifies the impedance response of the electrode.
Although Nyquist type plots are the most common, we
find it more informative to convert them into capacitance
plots [12]. This type of plot can better show how conductiv-
ity (resistivity) varies across the catalyst layer. This is because
the capacitance axis can also be considered to approximate
a distance axis since the capacitance is uniformly distributed
across the layer. In practice, the capacitance measured by
EIS will include both double layer capacitance and Faradaic
pseudo-capacitance terms for both the Pt and carbon surfaces.
However, within the potential range studied, these surface
Faradiac processes can be modeled as capacitances in paral-
lel with the double layer capacitance. Also, we can assume
that their relative contributions do not vary greatly between

0013-4686/$ – see front matter © 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.electacta.2004.10.074
2470 E.B. Easton, P.G. Pickup / Electrochimica Acta 50 (2005) 2469–2474
Fig. 1. Finite transmission-line equivalent circuit describing the impedance
behavior of a PEMFC electrode.
electrodes and that total capacitance is a relative measure-
ment of electrochemically active area. This assumption is
supported by cyclic voltammetry studies [13].
The main objectives of this work were to establish a re-
lationship between the liming capacitance and active area,
validate measurements taken on fuel cell electrodes and
study their ionic conductivities. In one case, we studied the
impedance behavior of catalyst layers immobilized on glassy
carbon electrodes following a procedure developed by Go-
jkovic et al. [14]. This method has proved useful to study fuel
cell catalyst layers under carefully controlled conditions [15].
The impedance response was studied as a function of cata-
lyst layer thickness and dc bias potential. In another case, we
studied the impedance response of gas diffusion electrodes
containing Os(bpy)3 2+ electroactive probes (which was ion
exchanged into the Nafion binder). Previous work in our lab
has shown that electroactive probes are useful for studying
the activity of fuel cell electrodes [16]. One of the advan-
tages of using an electroactive probe is that it can serve as a
reference point for impedance studies.
2. Experimental
2.1. Electrochemical experiments
Electrochemical experiments were performed at ambient
temperatures in a typical three-compartment cell using 0.5 M
H2 SO4 (aq.) as the electrolyte, a Ag/AgCl reference electrode
and a Pt counter electrode. Electrochemical measurements
(CV, EIS) were performed using a Solartron (Schlumberger)
1286 electrochemical interface and 1250-frequency response
analyzer controlled using custom software. Impedance Spec-
tra were collected using Z-plot software (Scribner Associates
Inc.) and obtained over a frequency range of 100 kHz to 1 Hz
at a dc bias potential of 0.2 V versus Ag/AgCl unless other-
wise stated.
Catalyst ink was prepared at a concentration of
0.0222 mg ␮L−1 by mixing 20% Pt/Vulcan XC72 carbon
black (11.1 mg, Etek) and 5% Nafion solution (500 ␮L, So-
lution Technology) thoroughly with a magnetic stirrer and
sonication to ensure homogeneity. Volumes of this catalyst
ink (1–6 ␮L) were applied via a syringe onto a 0.071 cm2
glass carbon electrode and allowed to dry before use. Such a
large percentage of Nafion was chosen to ensure good proton
conductivity and to avoid the use of other solvents in the ink
preparation.
The preparation of gas diffusion electrodes containing
Os(bpy)3 2+ electroactive probe (0.80 mg cm−2 20% Pt/C,
0.98 mg cm−2 Nafion, and 0.15 mg cm−2 Os(bpy)3 -
Cl2 ·6H2 O) is described in detail elsewhere [16]. Half-cell
electrochemical measurements were performed in a three-
electrode configuration by exposing the 1 cm2 test electrode
to a 0.5 M H2 SO4 electrolyte with a Pt wire counter electrode
and a Ag/AgCl reference electrode. For measurements in
fuel cells, membrane and electrode assemblies (MEAs) were
prepared by hot-pressing the 1 cm2 electrode and a similar
Pt black catalyzed electrode onto each side of a Nafion
membrane. Electrochemical experiments were performed
on these MEAs in a two-electrode configuration with
humidified hydrogen passing over the Pt black electrode,
which serves as both counter and reference electrodes, and
nitrogen passing over the test electrode. The impedance of
the anode is assumed to be negligible. The MEA was run as
a H2 /O2 fuel cell (with O2 passing over the test electrode)
for 30 min at 0.5 V immediately before the impedance
measurements were performed.
2.2. Pt particle size
Powder X-ray diffraction (XRD) spectra were acquired
using a Rigaku Ru 200 diffractometer using Cu K␣ radi-
ation (0.15405 nm). Average particle sizes were estimated
from the spectral line broadening using commercial software
(Jade 1994, Materials Inc.) that uses the Scherrer equation
and includes an instrument calibration parameter.
3. Results and discussion
3.1. Catalyst layers immobilized onto glassy carbon
electrodes
Cyclic voltammetry (CV) is the most common technique
to characterize active areas of Pt catalyst. CVs were obtained
for electrodes with various amounts of catalyst ink applied.
Fig. 2 shows the CVs obtained with 1–6 ␮L of ink. Good
definition of the hydrogen adsorption/desorption peaks was
obtained when a small amount of ink (1–3 ␮L) was used,
which allowed for easy integration of the area under the H-
adsorption waves. However, these peaks became less defined
as more ink is added, so much so that accurate integration
was not possible when 5 and 6 ␮L of ink was used. The lack
of peak definition may be due to increased resistivity in the
catalyst layer when larger volumes are used. Nonetheless,
for 1–4 ␮L, the charge under the H-adsorption wave does
scale well with ink volume, as shown in Fig. 3. From this, we
calculate a value of 0.0396 mC ␮g−1 of catalyst.
Pt utilization was calculated as the ratio of the area of
electroactive Pt, estimated from the charge under the H-
 E.B. Easton, P.G. Pickup / Electrochimica Acta 50 (2005) 2469–2474 2471

Fig. 4. Nyquist plots obtained at a dc bias potential of 0.2 V vs. Ag/AgCl

for electrodes with various amounts of catalysts ink applied.

Fig. 2. CVs (100 mV s−1 ) obtained in 0.5 M H2 SO4 (aq.) with 1–6 ␮L of
catalyst ink applied.

adsorption waves (0.21 mC cm−2 ), to the total area of Pt esti-

mated from the mean particle size (radius r), on the assump-
tion that the particles are spheres with surface area 4πr2 .
Based upon an average Pt particle size of 3.2 nm determined
from XRD experiments, Pt utilization was estimated to be ca.
109%, when 1–4 ␮L of ink was applied.
Fig. 4 shows Nyquist plots obtained with various amounts
of catalyst ink applied onto the electrode. From these we can
see that the ionic resistance increases with the volume of
ink applied (an enlarged view of the high-frequency region
is shown in Fig. 5). However, there appears to be a larger
increase in resistance for electrodes containing 4–6 ␮L of
catalyst ink. This may be in part to due to a less homogenous
catalyst layer when large volumes of ink are applied since Fig. 5. Expansion of the high-frequency region of Fig. 4.
successive layers of ink was required to apply the larger vol-
umes of ink. The increased resistivity also explains the CV tained for 1–6 ␮L of catalyst ink applied. It can be seen that
results. for 1–3 ␮L of ink, the initial slope (up to half the total capac-
This increase in ionic resistivity can be better visualized itance) of each plot is very similar; indicating that the resis-
through capacitance plots. Fig. 6 shows capacitance plots ob- tivity of each is similar. However, plots obtained for 4–6 ␮L

Fig. 6. Capacitance plots obtained at a dc bias potential of 0.2 V vs. Ag/AgCl

Fig. 3. Variation of average hydrogen adsorption charge and limiting capac- for electrodes with (a) 1–3 ␮L of catalyst ink applied and (b) 4–6 ␮L of
itance as a function of the mass of catalyst applied. catalyst ink applied.
2472 E.B. Easton, P.G. Pickup / Electrochimica Acta 50 (2005) 2469–2474

Fig. 7. Mass normalized capacitance plots obtained for electrodes with dif- Fig. 8. Comparison of capacitance plots obtained at a dc bias potential of
ferent volumes of catalyst ink applied at a dc bias potential of 0.2 V vs. 0.2 V vs. Ag/AgCl with 2 ␮L of catalyst ink immobilized onto a 0.071 cm2
Ag/AgCl. glassy carbon electrode, to that obtained with a 1 cm2 gas diffusion electrode
equivalent to ca. 2.2 ␮L of catalyst ink.

show much lower slopes, indicating higher resistivity. This tential is set at a potential where Faradaic processes occur,
is perhaps better seen in Fig. 7 where the capacitances have then an increase in capacitance will be observed. We have
been normalized by dividing them by the mass of catalyst therefore studied the effect of dc bias potential on the EIS
applied. Here, the increasing resistivity is clearly visible. response of electrodes prepared from the Pt/C/Nafion ink de-
It can also be seen from the normalized capacitance plots scribed above.
that the limiting capacitance does scale well with the quantity Fig. 9 shows a plot of limiting capacitance versus dc bias
of catalyst applied. A plot of limiting capacitance as a func- potential. There is a large increase in capacitance between
tion of the mass of catalyst applied is shown in Fig. 3. From −0.2 and 0.0 V that can be attributed to the hydrogen adsorp-
this, we calculate a capacitance of 0.037 mF ␮g−1 of catalyst tion/desorption process on Pt. There are also smaller peaks at
and 0.196 mF cm−2 of Pt. These are useful parameters that higher potentials, which are shown in the inset of Fig. 9. The
can be applied to fuel cell electrodes to determine utilization. peak at ca. 0.3 V can be attributed to the presence of the elec-
The linearity of this plot indicates that even though thicker trochemistry of quinone groups present on the carbon surface
electrodes have higher resistivity, the active area per gram [17]. The peak between 0.6 and 0.9 V can be attributed to the
of catalyst of each electrode is approximately the same. This electrochemistry associated with oxide formation/stripping
would not necessarily be the case in an operating fuel cell on the Pt surface.
electrode, since a fuel cell electrode would not be as well These plots indicate that EIS data should not be acquired
hydrated as one exposed to sulfuric acid solution, nor would at potentials below 0.1 V (versus Ag/AgCl). While there are
it contain such a large percentage of Nafion. some Faradaic processes present above 0.1 V, the magnitude
Comparison of these results with those obtained with gas
diffusion electrodes is desired to validate the application of
EIS to fuel cell electrodes. To do this, a 1 cm2 gas diffusion
electrode, containing 0.85 mg cm−2 of ETEK 20% Pt/C, and
1.1 mg cm−2 (57%) of Nafion was prepared and tested in the
same electrolyte. This loading corresponds to the equivalent
of ca. 2.2 ␮L of catalyst ink applied onto a 0.071 cm2 elec-
trode. Fig. 8 compares the capacitance plot obtained with the
1 cm2 gas diffusion electrode to that obtained with 2 ␮l of
catalyst ink on the 0.071 cm2 electrode. The results are very
similar. The slight differences can be attributed to the differ-
ences in the Nafion content (65% versus 57%) and the ap-
proximate nature of the calculation. Nevertheless, this shows
that impedance measurements taken on gas diffusion elec-
trodes correlate well with data obtained by the more con-
trolled Nafion/catalyst ink on glassy carbon method. Fig. 9. Limiting capacitance as a function of dc Potential. Data obtained us-
The dc bias potential at which EIS is acquired can have a ing 1 ␮L of catalyst ink. The inset shows an expansion of the region between
significant effect upon the data collected. If the dc bias po- 0 and 1.2 V.
 E.B. Easton, P.G. Pickup / Electrochimica Acta 50 (2005) 2469–2474
Fig. 10. Comparison of (a) capacitance plots and (b) Nyquist plots obtained
with nitrogen bathed 1 cm2 gas diffusion electrode containing the Os(bpy)3 2+
electroactive probe in both the half-cell and fuel cell MEA. Experimental
data is shown as open symbols, lines represent simulated data obtained using
the conductivity profiles shown in Fig. 11. Half-cell data was obtained at a dc
bias = 0.2 V vs. Ag/AgCl, fuel cell data was obtained at a dc bias = 0.425 V
vs. NHE.
is much lower than that for hydrogen adsorption/desorption
and should therefore have a smaller effect on the data col-
lected.
3.2. Electrodes containing Os(bpy)3 2+ electroactive
probes
One of the advantages of using an electroactive probe is
that it will provide an alternative method determination of the
fraction of the catalyst layer that is electrochemically active.
When the Os(bpy)3 2+ electroactive probe is incorporated into
the catalyst layer, a well defined Os(II/III) peak is observed
during a CV experiment, which can be easily integrated. The
area under the Os(II/III) peak is related to the fraction of the
catalyst layer that is electroactive, in the same manner that
the area under the H-adsorption peaks are related to catalyst
utilization.
Fig. 10(b) shows Nyquist plots obtained in both a half-cell
and in a fuel cell for electrodes containing Os(bpy)3 2+ elec-
troactive probes. Capacitance plots are shown in Fig. 10(a).
The most obvious aspect of the capacitance plots is the higher
limiting capacitance obtained in the half-cell. This indicates
that a significantly smaller fraction of the catalyst layer is
electrochemically active in the fuel cell. Based on the values
of the limiting capacitance, only ca. 65% of the catalyst layer
that is electroactive in the half-cell is electroactive in the fuel
cell. This value is slightly higher than that obtained by cyclic
voltammetry and integration of the Os(II/III) peak (ca. 55%,
Ref. [16]). This discrepancy is most likely due to slight vari-
ation in catalyst loading between electrodes and/or variation
of the hydration state within the catalyst layers. A steeper
2473
Fig. 11. Conductivity profiles used to simulate the EIS data shown in Fig. 10.
initial slope of the capacitance plot is also observed in the
half-cell. This clearly indicates that there is better ionic con-
ductivity in the half-cell compared to the fuel cell, agreeing
with our inferences from the CV. Based on the relationship
of 0.196 mF cm−2 established between the capacitance and
Pt area for this catalyst, the active Pt area in the fuel cell can
be estimated to be 64 cm2 .
To further study the differences in conductivity between
the fuel cell and half-cell configurations, data was simulated
from a conductivity profile to fit these experimental results us-
ing a Microsoft Excel spreadsheet [12]. The simulated fits to
the impedance data are displayed in Fig. 10. The conductivity
profiles that simulated these fits are shown in Fig. 11. These
profiles are reasonable and are similar to those previously
reported [4,9,10,13]. Good fits were obtained, especially at
high-frequencies. The distance into the layer is based upon an
estimate of 55% of a 30 ␮m layer being electroactive (from
Ref. [16]). From this we see that in both cases the conduc-
tivity close to the electrolyte is similar and declines as you
move further away from the electrolyte. This is the experi-
mental observation and it indicates that the catalyst layer is
better hydrated close to the membrane. This can be explained
by electro-osmotic drag and the evaporation of water into the
backing layer. The similarity of the conductivities close to the
electrolyte indicates that the presence of H2 SO4 (aq.) in the
half-cell electrode does not appear to add significantly to its
conductivity. However, there is a much sharper decline in con-
ductivity when the electrode is used in a fuel cell. This is a re-
sult of less catalyst layer hydration when used in the fuel cell.
4. Conclusions
EIS is a powerful tool to study PEMFC catalyst layers. Ac-
tive area and ionic conductivity within the catalyst layer has
been studied as a function of catalyst layer thickness using
EIS. We have shown that there is a relationship between the
limiting capacitance and the mass of catalyst. The capacitance
2474 E.B. Easton, P.G. Pickup / Electrochimica Acta 50 (2005) 2469–2474
per cm2 of Pt for the 20% Pt on C catalyst employed here was
calculated to be 0.196 mF (which includes contributions from
both the Pt and C support) and this is proposed as a measure
of catalyst utilization. Trends observed in the EIS data agreed
well with those observed with utilization measurements
made from the integration of hydrogen adsorption peaks and
Os(bpy)3 2+ electroactive probes. Results obtained on more
controlled smaller scale electrodes correlate well with results
obtained for larger gas diffusion electrodes. The smaller scale
method is therefore suitable for screening different catalyst
layer compositions before larger scale tests are performed.
The EIS data also has the added benefit in that it can be used
to study the ion conductivity profile within the catalyst layer.
Together, these results provide strong evidence that the
non-ideal impedance behavior of fuel cell electrodes is due
to variation of their ionic conductivity with distance from
the membrane. We have also shown that EIS can be used to
measure catalyst utilization in situations where the more tra-
ditional method of integrating the H-adsorption peaks cannot
be performed.
Acknowledgements
This work was supported by the Natural Sciences and
Engineering Research Council of Canada and Memorial
University.
References
[1] N. Wagner, W. Schnurnberger, B. Muller, M. Lang, Electrochim.
Acta 43 (1998) 785.
[2] J.P. Daird, B. LeGorrec, C. Montella, C. Poinsignon, G. Vitter, J.
Power Sources 74 (1999) 244.
[3] T.E. Springer, T.A. Zawodzinski, M.S. Wilson, S. Gottesfeld, J. Elec-
trochem. Soc. 143 (1996) 87.
[4] G. Li, P.G. Pickup, J. Electrochem. Soc. 150 (2003) C745.
[5] G. Li, P.G. Pickup, Electrochim. Acta 49 (2004) 4119.
[6] X. Zhong, S. Holdcroft, J. Electroanal. Chem. 568 (2004)
247.
[7] M. Eikerling, A.A. Kornyshev, J. Electroanal. Chem. 745 (1999)
107.
[8] M. Ciureanu, R. Roberge, J. Phys. Chem. B 105 (2001) 3531.
[9] Q. Guo, M. Cayetano, Y. Tsou, E.S. De Castro, R.E. White, J.
Electrochem. Soc. 150 (2003) A1440.
[10] M.C. Lefebvre, R.B. Martin, P.G. Pickup, Electrochem. Solid-State
Lett. 2 (1999) 259.
[11] A.P. Saab, F.H. Garzon, T.A. Zawodzinski, J. Electrochem. Soc. 149
(2002) 1541.
[12] X. Ren, P.G. Pickup, Electrochim. Acta 46 (2001) 4177.
[13] N. Jia, R.B. Martin, Z. Qi, M.C. Lefebvre, P.G. Pickup, Electrochim.
Acta 46 (2001) 2863.
[14] S.L. Gojkovic, S.K. Zecevic, R.F. Savinell, J. Electrochem. Soc. 145
(1998) 3713.
[15] J. Shan, P.G. Pickup, Electrochim. Acta 46 (2000) 119.
[16] E.B. Easton, P.G. Pickup, Electrochem. Solid-State Lett. 3 (2000)
359.
[17] K. Kinoshita, Carbon: Electrochemical and Physicochemical Proper-
ties, Wiley, New York, 1988.