Electrochimica Acta 141 (2014) 212–215

Contents lists available at ScienceDirect

Electrochimica Acta
journal homepage: www.elsevier.com/locate/electacta

In situ measurement of the oxygen diffusion coefficient in the cathode

catalyst layer of a direct methanol fuel cell
K. Wippermann, K. Klafki, A.A. Kulikovsky ∗,1,2
Forschungszentrum Juelich GmbH, Institute of Energy and Climate Research, IEK–3: Electrochemical Process Engineering, D–52425 Jülich, Germany

a r t i c l e i n f o a b s t r a c t

Article history: We report an in situ measurement of the oxygen diffusion coefficient Dox in the cathode catalyst
Received 2 June 2014 layer (CCL) of a direct methanol fuel cell (DMFC). The method is based on measuring two DMFC
Received in revised form 26 June 2014 impedance spectra for the same cell current, but different oxygen concentrations in the cathode chan-
Accepted 29 June 2014
nel. The spectra are fitted using a transmission line model to extract the CCL resistivity. The difference
Available online 18 July 2014
of two CCL resistivities
  different
for oxygen concentrations determines Dox . We found Dox = 0.838 ×
10−4 · exp − ERa T1 − 303
1
cm2 s−1 in the temperature range from 30 ◦ C to 90 ◦ C, with Ea  20.1 kJ mol−1 .
Keywords:
DMFC The activation energy is not far from that of the oxygen diffusion in Nafion; the value of this parameter
Cathode catalyst layer is discussed.
oxygen diffusion © 2014 Elsevier Ltd. All rights reserved.
experiment

1. Introduction To understand the effect of crossover, a lot of modeling stud-

Direct methanol fuel cells (DMFCs) are compact sources of elec-
tricity based on direct electrochemical conversion of methanol
combustion energy into electric current. Interest in this type of
cells has been growing since the discovery of proton–conducting
polymer Nafion in the late sixties of the XX century. A great advan-
tage of DMFC over hydrogen cell is high volumetric energy density
of liquid methanol, which makes DMFC an ideal power source for
mobile gadgets.
However, well–known drawbacks of DMFCs retard wide com-
mercialization of these cells. These are sluggish kinetics of
methanol oxidation and large crossover of methanol from the
anode to the cathode. Methanol permeated through the membrane
is oxidized on the cathode side into protons and electrons. As a
result, the total proton current that has to be converted in the cath-
ode catalyst layer (CCL) greatly exceeds the useful current in the
external load. Conversion of the parasitic current requires an extra
flux of oxygen that needs to be transported through the cathode
GDL and CCL. Both effects (higher proton current in the CCL and
extra oxygen flux) dramatically lower the cell potential, especially
in the range of small cell currents.
ies have been done (we refer reader to reviews of Wang [1], Yao
et al. [2] and Bahrami and Faghri [3]). One of the key parameters
governing the DMFC cathode operation is the oxygen diffusion
coefficient in the CCL [4–8].
Typically, the CCL oxygen diffusivity is calculated using the
Bruggemann correction for the CCL porosity/tortuosity to the
oxygen diffusivity in free air [4,5,9,8]. The latter depends on
temperature according to the power law Dox ∼ T1.75 [10]. This cor-
rection, however, takes into account nor the contribution of the
Knudsen diffusion, neither a Nafion film, which surrounds agglom-
erates of Pt/C particles, where the ORR takes place [11].
To the best of our knowledge, direct in situ measurements of Dox
in a working DMFC has not been published yet. Below, we report
such measurements based on a recent model [12]. The model shows
that the difference of CCL resistivities measured at the same cell
current density but different oxygen concentrations in the cathode
channel is inversely proportional to Dox . To measure CCL resisti-
vities we use electrochemical impedance spectroscopy (EIS).
2. Model equation

Oxygen transport in the CCL contributes to the total overpo-

tential and to the differential CCL resistivity RCCL . Recently, an
∗ Corresponding author.
analytical expression for the respective transport overpotential ox
E-mail address: A.Kulikovsky@fz-juelich.de (A.A. Kulikovsky).
1
Also with: Moscow State University, Research Computing Center, 119991
has been derived [13].
Moscow, Russia. The model has been developed for the CCL with polymer elec-
2
ISE member. trolyte as a proton conductor and hence it works in the temperature

http://dx.doi.org/10.1016/j.electacta.2014.06.164
0013-4686/© 2014 Elsevier Ltd. All rights reserved.
 K. Wippermann et al. / Electrochimica Acta 141 (2014) 212–215
range between zero and 90 ◦ C. From the expression derived in [13]
it follows, that RCCL is given by
RCCL = RF (jc ) + Rox (jc , cox )
where RF is the faradaic resistivity, and
2(p b + jc lt )blt
Rox =
3(2p b + jc lt )4FDox cox
is the resistivity due to the oxygen transport in the CCL. Here jc is
the total current density converted in the CCL (see below), cox is the
oxygen concentration in the CCL, b is the Tafel slope of the oxygen
reduction reaction,  p is the CCL proton conductivity, lt is the CCL
thickness, Dox is the oxygen diffusion coefficient in the CCL.
It is difficult to separate experimentally the contribution of Rox
to the total CCL resistivity RCCL . The most accurate technique for
determining partial cell resistivities is electrochemical impedance
spectroscopy (EIS). Unfortunately, the characteristic time of oxygen
transport in the CCL  ox  lt 2 /Dox is very close to the characteristic
time of faradaic processes  F  Cdl lt 0 /jc , where Cdl is the double
layer volumetric capacitance (F cm−3 ) and 0 is the total over-
potential. Processes with close time scales cannot be resolved by
impedance spectroscopy.
Nonetheless, the fact that RF is independent of the oxygen con-
centration, while Rox is inversely proportional to cox suggests an
alternative method for Rox determination [12]. Suppose that two
impedance spectra have been measured for the same current den-
sities, but different oxygen concentrations in the cell. Let the oxygen
concentrations used be cox,1 and cox,2 and let cox,2 = kc cox,1 , where
kc > 1. Writing down Eq. (1) for cox,1 and cox,2 and subtracting the
resulting equations, we see that the faradaic terms cancel out and
we get
2(kc − 1)(p b + jc lt )blt
ıRCCL = Rox,2 − Rox,1 =
3kc (2p b + jc lt )4FDox cox,1
From Eq. (3) we find the oxygen diffusion coefficient in the CCL
2(kc − 1)(p b + jc lt )blt
Dox =
3kc (2p b + jc lt )4F(ıRCCL )cox,1
An important feature of DMFC’s is large rate of methanol
crossover through the membrane. Methanol permeated through
the membrane is oxidized to protons and electrons in the CCL, and
the resulting proton current density is then converted to the flux
of water in the ORR [7]. Thus, the total current converted in the
CCL is a sum of the useful j0 and equivalent crossover jcross current
densities:
jc = j0 + jcross
The value of jc is related to the faradaic CCL resistivity. For the
cell operating in the low–current regime (see below), the following
relation holds [14]
b
RF 
jc
This equation allows us to calculate jc , which then can be used in
Eq. (4).
3. Experimental
3.1. MEA preparation
AvCarb 1071 HCB carbon cloth supplied by Ballard Material
Products was used as a substrate for the anodes and cathodes.
First, the carbon cloth was impregnated with PTFE dispersion TF
5032 (Dyneon) and subsequently dried and sintered at 350 ◦ C. A
microporous layer of carbon particles (VULCAN XC72 from Cabot)
and PTFE (TF 5032 from Dyneon) was then prepared on the cloth
213
(1)
(2)
Fig. 1. Transmission line model used for the DMFC impedance fitting. For explana-
tion of the elements see text.
by knife-coating to obtain gas diffusion layers (GDL). In a second
coating step, the catalyst layers were prepared on the GDLs by
knife-coating, too. Both catalyst layers consisted of catalyst mate-
rial from Johnson Matthey and Nafion. In the case of the cathode
catalyst layer, PTFE was added. The starting materials of PTFE and
Nafion are the PTFE dispersion TF 5032 from Dyneon and the
Nafion dispersion LQ1115 from Ion Power. HiSPEC 12100 catalyst
(75 wt.% PtRu/C) with a loading of 2.6 mg cm−2 was used for the
preparation of the anode. On the cathode side, HiSPEC 13100 cat-
alyst (71 wt.% Pt/C) was applied with a loading of 2.0 mg cm−2 .
Finally, the electrodes were cut to a size of 42 mm × 42 mm corre-
sponding to a cell area of 17.64 cm2 and hot-pressed on a Nafion®
115 membrane at 130 ◦ C for 3 min.
3.2. Electrochemical impedance spectroscopy (EIS)
The impedance spectra were recorded by a ZAHNER-Elektrik
“Zennium electrochemical workstation”. The impedance measure-
ments were always performed with a 10 mV input signal amplitude
in a frequency range of 20 mHz - 20 kHz and the temperatures of
30, 40, 50, 60, 70, 80 and 90 ◦ C under ambient pressure. DC cur-
rent of −0.2 A was applied, corresponding to the current density
(3) through the cell of −11.34 mA cm−2 . The flow fields were made of
graphite and had a grid design (ribs: 1 mm×1 mm×1 mm, chan-
nels: 1 mm width and depth). The anode was always fed with 1 M
aqueous methanol solution with a flow rate of 1.1 ml min−1 , equiv-
(4) alent to stoichiometric –factor of 53. The cathode was fed by either
pure oxygen (oxygen molar fraction 1.0), or by a mixture of oxygen
and nitrogen with the oxygen molar fractions of 0.1 and 0.3. The
gas flow rate at the cathode was 192 ml min−1 , corresponding to
–factors greater than 10, depending on the oxygen fraction. The
methanol solution was pre-heated to operating temperature. From
these experiments, full cell impedances were obtained.
The impedance measurements were also carried out purging
the cathode with hydrogen instead of air (VH2 = 192 ml min−1 ).
By applying a galvanostatic current, the cathode turns into a
(5)
dynamic hydrogen evolving electrode (DHE), which performs as a
quasi–reference electrode. As the impedance of the hydrogen cath-
ode can be neglected, the cell impedance measured under these
conditions is approximately equal to the anode impedance. By
subtracting the anode impedance from the corresponding full cell
(6) impedances, the cathode impedances are obtained. These exper-
iments demonstrate that the high frequency arc of the full cell
impedance spectra can be undoubtedly attributed to the cathode
impedance (see below).
The impedance spectra were fitted with the Zahner THALES Z
2.0 fit software using the equivalent circuit shown in Fig. 1. The
equivalent circuit is based on a transmission line model of the cell
discussed in [15]. It includes the following elements (i) Lcable , the
cable inductance, (ii) R , the ohmic resistivity, (iii) Rp , the proton
resistivity of both the catalyst layers, (iv) RCCL , the kinetic resis-
tivity of the cathode catalyst layer, (v) CPECCL , the constant phase
element representing the pseudo double layer capacitance of the
CCL, (vi) R∞,ACL , the resistivity associated with the charge-transfer
part without change in CO surface coverage in the anode catalyst
layer (ACL), (vii) R0 , the resistivity associated with relaxation of the
214 K. Wippermann et al. / Electrochimica Acta 141 (2014) 212–215
Fig. 2. (a) DMFC impedance spectra for a current density of 11.3 mA cm−2 and for
oxygen molar fractions of 0.1, 0.3 and 1.0 in O2 –N2 mixtures at a pressure of 1.0 bar
and a cell temperature of 70 ◦ C. (b) Enlarged HF parts of the spectra. Measured
values are marked by symbols, fitted curves are indicated by solid lines. The fre-
quency range is 20 mHz – 20 kHz; anode feed: V1M,MeOH = 1.1 ml min−1 , cathode feed:
VO2−N2 = 192 ml min−1 .
CO surface coverage in the ACL, (viii) Lad , the inductance associ-
ated with relaxation of the CO surface coverage in the ACL and (ix)
CPEACL , the constant phase element representing the pseudo double
layer capacitance of the ACL. However, only the kinetic resistivity
of the cathode catalyst layer RCCL is required for the determina-
tion of the oxygen diffusion coefficient in the CCL, as discussed in
more detail below. The average fitting errors of the elements of the
proposed equivalent circuit are as follows: Lcable : 21.8%, R : 2.5%,
Rp : 5.4%, RCCL : 5.7%, CPECCL : 5.8%, R∞,ACL : 4.7%, R0 : 19.4%, Lad : 5.8%,
CPEACL : 4.3%.
4. Results and discussion
Figs. 2, 3 show the DMFC impedance spectra acquired at oxygen
molar fractions of 0.1, 0.3 and 1.0, for the useful current density j0 of
11.3 A cm−2 at the cell temperature of 70 ◦ C. Fig. 2 shows the whole
cell spectrum and the enlarged high–frequency part of the spec-
trum. As can be seen, the shape of the low–frequency (LF) arc does
not show any distinct trend with respect to the oxygen concentra-
tion and it exhibits large inductive loop, which is typical for DMFC
anode due to adsorption of methanol oxidation by–products on the
anode catalyst surface. In contrast, the high–frequency (HF) arcs
clearly show that the radii of these arcs decrease with the growth
of the oxygen concentration (Fig. 2b). We attribute the HF arcs to
the cathode side.
Fig. 3. Separated anode and cathode impedance based on the fitting results of the
raw data shown in Figure 2.
Fig. 4. Arrhenius plot of the oxygen diffusion coefficient in the cathode CL.
Since only a part of the cathode arc is visible in the spectrum,
fitting procedure is necessary to determine the intercept of cath-
ode arcs with the real axis. To determine these intercepts we used
a simple transmission line model for the cell (Fig. 1, see previous
section). The fitted spectra are depicted in Fig. 2 by solid lines.
Fig. 3 shows separate cathode and anode arcs for T = 70 ◦ C and
oxygen partial pressure of 0.3 bar, reconstructed by means of the
impedance curve fitting. This Figure shows the CCL resistance RCCL ,
which has been determined as a right intercept of the CCL arc with
the real axis.
All the measurements have been performed at a fixed cell cur-
rent density of 11.3 mA cm−2 . For a fixed cell temperature, we have
measured three CCL resistivities RCCL corresponding to the three
oxygen molar fractions of 0.1, 0.3 and 1.0 in the cathode flow. From
these data, three values of ıRCCL have been calculated. The values
of ıRCCL have been fitted by the curve a/ ox , in accordance with Eq.
(3), and the value of ıRCCL corresponding to the oxygen fraction of
0.1 and 1 has been taken from this curve. The diffusion coefficient
has been calculated from Eq. (4) with kc = 10, cox corresponding to
the molar fraction of 0.1 at a given temperature, b = 0.05 V and the
value of jc estimated from Eq. (6) assuming that for the pure oxy-
gen feed the following relation holds: RF  RCCL . The catalyst layer
thickness was taken to be lt = 0.003 cm (30 ␮m) and the ORR Tafel
slope was 50 mV, a value determined from separate experiments.
Fig. 4 shows the temperature dependence of the measured oxy-
gen diffusion coefficient. As can be seen, this dependence follows
Arrhenius law:
 E 1 1

a
Dox = 0.838 · 10−4 exp − − with
R T 303
Ea = 20.1 kJ mol−1 (7)
 K. Wippermann et al. / Electrochimica Acta 141 (2014) 212–215
The absolute value of Dox is about 10−4 cm2 s−1 , which is between
the oxygen diffusion coefficient in a dry CCL for a PEM fuel
cell (1.37 × 10−3 cm2 s−1 [16]), and the oxygen diffusivity in water
(10−5 cm2 s−1 ).
The activation energy of 20.1 kJ mol−1 resulted from our exper-
iments is not far from the activation energy of 16.2 kJ mol−1 for
oxygen diffusion through Nafion reported in [17]. This suggests
that the temperature dependence of Dox originates from the oxy-
gen transport through a thin Nafion film covering agglomerate of
the carbon–Pt particles. Note that the absolute value of the mea-
sured Dox is two orders of magnitude higher, than that in the bulk
Nafion [17]. This means that the main contribution to the abso-
lute value of Dox give the transport through the voids, which varies
slowly with the temperature (∼T7/4 ).
This effect can be illustrated as follows. Consider the transport
resistances of Nafion film surrounding Pt/C agglomerate lN /DN and
of the void pore lp /Dp , where lN , lp are the thickness of the Nafion
film and the characteristic pore length, respectively, DN , Dp are the
oxygen diffusion coefficients in the Nafion and in the void pore,
respectively. These resistances are connected in series, and for the
∗ we have the following estimate
effective resistance L/Dox
L lp lN
∗ = D + D (8)
Dox p N
where L = lp + lN , and the asterisk indicates the effective diffusivity
of the system “void pore + Nafion film”. Solving Eq. (8) for Dox ∗ , we
get
∗ LDN Dp
Dox = (9)
lp DN + lN Dp
By the order of magnitude we have lp  10−3 cm (10 ␮m),
lN  10−6 cm (10 nm), Dp  10−2 cm2 s−1 , and DN  10−6 cm2 s−1 .
With these values we have lp DN  lN Dp and hence the first term
in denominator in Eq. (9) can be neglected. With this, we arrive at
∗ L
Dox = DN (10)
lN
i.e., the small diffusion coefficient DN is multiplied by a large factor
L/lN giving three orders of magnitude larger effective oxygen diffu-
sivity Dox∗ . In the context of this work, most important is that D∗
ox
exhibits the temperature dependence of the diffusion coefficient in
Nafion DN , as Eq. (10) shows.
The deviation of Ea in Eq. (7) from the value of 16.2 kJ mol−1 can
be explained by the effect of liquid saturation s on Dox . With the
growth of temperature, water evaporation lowers the fraction of
liquid water s in the CCL, which increases the effective diffusivity
of the media. Temperature variation of the saturated water vapor
pressure, which controls the value of s in the CCL is given by the acti-
vation energy of water evaporation, which is about 43 kJ mol−1 [18].
This effect may increase the effective activation energy for Dox as
compared to this parameter for the oxygen diffusion in Nafion 117.
5. Conclusions
We report an in situ measurements of the oxygen diffusion coef-
ficient Dox in the cathode catalyst layer of a direct methanol fuel cell.
The method consists of the following steps.
215
• We measure the cell impedance spectrum for at least two dif-
ferent oxygen concentrations, keeping the cell current density
unchanged.
• The experimental spectra are fitted by the transmission line
model, which allows us to extract the CCL resistivities from the
impedance data.
• The difference of two CCL resistivities corresponding to the two
oxygen concentrations is substituted to the model equation (4)
to determine Dox .
The resulting oxygen diffusivity exhibits Arrhenius–type
dependence on the cell temperature with the activation energy
Ea , which is close to that value for the oxygen diffusion in Nafion.
Somewhat larger value of Ea can be explained by the effect of water
evaporation in the CCL, which lowers the liquid saturation and
facilitates the oxygen diffusion in the gas phase.
References
[1] C.-Y. Wang, Fundamental models for fuel cell engineering, Chem. Rev. 104
(2004) 4727–4766.
[2] K.Z. Yao, K. Karan, K.B. McAuley, P. Oosthuizen, B. Peppley, T. Xie, A review of
mathematical models for hydrogen and direct methanol polymer electrolyte
membrane fuel cells, Fuel Cells 4 (2004) 3–29.
[3] H. Bahrami, A. Faghri, Review and advances of direct methanol fuel cells:
Part II: Modeling and numerical simulation, J. Power Sources 230 (2014)
303–320.
[4] Z.H. Wang, C.Y. Wang, Mathematical modeling of liquid-feed direct methanol
fuel cells, J. Electrochem. Soc. 150 (2003) A508–A519.
[5] W. Liu, C.-Y. Wang, Three-dimensional simulations of liquid
feed direct methanol fuel cell, J. Electrochem. Soc. 154 (2007)
B352–B361.
[6] T.S. Zhao, C. Xu, R. Chen, W.W. Yang, Mass transport phenomena in direct
methanol fuel cells, Progress in Energy and Combustion Sci. 35 (2009)
275–292.
[7] A.A. Kulikovsky, A model for DMFC cathode performance, J. Electrochem. Soc.
159 (2012) F644–F649.
[8] S. Matar, J. Ge, H. Liu, Modeling the cathode catalyst layer of a direct methanol
fuel cell, J. Power Sources 243 (2013) 195–202.
[9] C. Xu, A. Faghri, Analysis of an active tubular liquid-feed direct methanol fuel
cell, J. Power Sources 196 (2011) 6332–6346.
[10] R.B. Bird, W.E. Stewart, E.N. Lightfoot, Transport Phenomena, Wiley, New-York,
1960.
[11] K. Malek, T. Mashio, M. Eikerling, Microstructure of catalyst layers in
PEM fuel cells redefined: A computational approach, Electrocatal. 2 (2011)
141–157.
[12] A.A. Kulikovsky, A simple equation for in situ measurement of the catalyst
layer oxygen diffusivity in PEM fuel cell, J. Electroanal. Chem. 720–721 (2014)
47–51.
[13] A.A. Kulikovsky, A physically-based analytical polarization curve of a PEM fuel
cell, J. Electrochem. Soc. 161 (2014) F263–F270.
[14] A.A. Kulikovsky, M. Eikerling, Analytical solutions for impedance of the cathode
catalyst layer in PEM fuel cell: Layer parameters from impedance spectrum
without fitting, J. Electroanal. Chem. 691 (2013) 13–17.
[15] K. Wippermann, A. Loehmer, A. Everwand, M. Mueller, C. Korte, D. Stolten, Study
of complete methanol depletion in direct methanol fuel cells, J. Electrochem.
Soc. 161 (2014) F525–F534.
[16] J. Shen, J. Zhou, N.G.C. Astrath, T. Navessin, Z.-S.S. Liu, C. Lei, J.H. Rohling, D.
Bessarabov, S. Knights, S. Ye, Measurement of effective gas diffusion coefficients
of catalyst layers of PEM fuel cells with a Loschmidt diffusion cell, J. Power
Sources 96 (2011) 674–678.
[17] V.A. Sethuraman, S. Khan, J.S. Jur, A.T. Haug, J. Weidner, Measur-
ing oxygen, carbon monoxide and hydrogen sulfide diffusion coefficient
and solubility in Nafion membranes, Electrochimica Acta 54 (2009)
6850–6860.
[18] V. Majer, V. Svoboda, Enthalpies of Vaporization of Organic Compounds, Black-
well Scientific Publications, Oxford, 1985.