Tungsten oxide in polymer electrolyte fuel cell electrodes - A thin-film model electrode study

Article abstract of DOI:10.1016/j.electacta.2011.08.046

Thin films of WO_x and Pt on WO_x were evaporated onto the microporous layer of a gas diffusion layer (GDL) and served as model electrodes in the polymer electrolyte fuel cell (PEFC) as well as in liquid electrolyte measurements. In order to study the effects of introducing WO _x in PEFC electrodes, precise amounts of WO_x (films ranging from 0 to 40 nm) with or without a top layer of Pt (3 nm) were prepared. The structure of the thin-film model electrodes was characterized by scanning electron microscopy and X-ray photoelectron spectroscopy prior to the electrochemical investigations. The electrodes were analyzed by cyclic voltammetry and the electrocatalytic activity for hydrogen oxidation reaction (HOR) and CO oxidation was examined. The impact of Nafion in the electrode structure was examined by comparing samples with and without Nafion solution sprayed onto the electrode. Fuel cell measurements showed an increased amount of hydrogen tungsten bronzes formed for increasing WO_x thicknesses and that Pt affected the intercalation/deintercalation process, but not the total amount of bronzes. The oxidation of pre-adsorbed CO was shifted to lower potentials for WO_x containing electrodes, suggesting that Pt-WO _x is a more CO-tolerant catalyst than Pt. For the HOR, Pt on thicker films of WO_x showed an increased limiting current, most likely originating from the increased electrochemically active surface area due to proton conductivity and hydrogen permeability in the WO_x film. From measurements in liquid electrolyte it was seen that the system behaved very differently compared to the fuel cell measurements. This exemplifies the large differences between the liquid electrolyte and fuel cell systems. The thin-film model electrodes are shown to be a very useful tool to study the effects of introducing new materials in the PEFC catalysts. The fact that a variety of different measurements can be performed with the same electrode structure is a particular strength.

Full text of DOI:10.1016/j.electacta.2011.08.046

Electrochimica Acta 56 (2011) 9496–9503
Contents lists available at SciVerse ScienceDirect
Electrochimica Acta
journal homepage: www.elsevier.com/locate/electacta
Tungsten oxide in polymer electrolyte fuel cell electrodes—A thin-film model
electrode study
a,∗
b
b
b
Björn Wickman , Maria Wesselmark , Carina Lagergren , Göran Lindbergh
a
Competence Centre for Catalysis, Department of Applied Physics, Chalmers University of Technology, SE-412 96 Göteborg, Sweden
Applied Electrochemistry, School of Chemical Science and Engineering, KTH, SE-100 44 Stockholm, Sweden
b
a r t i c l e i n f o
a b s t r a c t
Article history:
Received 1 April 2011
Received in revised form 13 August 2011
Accepted 14 August 2011
Available online 23 August 2011
Thin films of WOx and Pt on WOx were evaporated onto the microporous layer of a gas diffusion layer
GDL) and served as model electrodes in the polymer electrolyte fuel cell (PEFC) as well as in liquid
(
electrolyte measurements. In order to study the effects of introducing WOx in PEFC electrodes, precise
amounts of WOx (films ranging from 0 to 40 nm) with or without a top layer of Pt (3 nm) were prepared.
The structure of the thin-film model electrodes was characterized by scanning electron microscopy and X-
ray photoelectron spectroscopy prior to the electrochemical investigations. The electrodes were analyzed
by cyclic voltammetry and the electrocatalytic activity for hydrogen oxidation reaction (HOR) and CO
oxidation was examined. The impact of Nafion in the electrode structure was examined by comparing
samples with and without Nafion solution sprayed onto the electrode. Fuel cell measurements showed
an increased amount of hydrogen tungsten bronzes formed for increasing WOx thicknesses and that Pt
affected the intercalation/deintercalation process, but not the total amount of bronzes. The oxidation of
pre-adsorbed CO was shifted to lower potentials for WO containing electrodes, suggesting that Pt-WO is
Keywords:
PEMFC
Fuel cell
Tungsten oxide
Platinum
Hydrogen oxidation
CO oxidation
Thermal evaporation
x
x
a more CO-tolerant catalyst than Pt. For the HOR, Pt on thicker films of WOx showed an increased limiting
current, most likely originating from the increased electrochemically active surface area due to proton
conductivity and hydrogen permeability in the WOx film. From measurements in liquid electrolyte it was
seen that the system behaved very differently compared to the fuel cell measurements. This exemplifies
the large differences between the liquid electrolyte and fuel cell systems. The thin-film model electrodes
are shown to be a very useful tool to study the effects of introducing new materials in the PEFC catalysts.
The fact that a variety of different measurements can be performed with the same electrode structure is
a particular strength.
©
2011 Elsevier Ltd. All rights reserved.
1
. Introduction
cell environment. The support material can influence the activity
by spill-over effects as well as changing the electronic structure of
the catalyst [2,3]. New support materials can improve the activity,
utilization, and stability of the catalyst or of the support itself [4].
Tungsten oxide is a material which has been extensively inves-
tigated for a wide range of applications, mainly, due to its unique
electrochromic properties but also for its electrocatalytic activities
[5,6]. The electrochromism allows tungsten oxide to interca-
late/deintercalate ions (of e.g. H, Li, Na, K, Pb, Cd) into its structure
in the formation of tungsten bronzes. The most widely studied
form is the hydrogen tungsten bronze where protons are inserted
in the oxide structure as HxWO3 and 0 < x < 1. The bronze forma-
tion mechanism has been the subject for numerous studies and it
is suggested that the hydrogen atoms form hydroxyl bonds in the
tungsten oxide, following the reaction below [7–10].
Development of new alternative electrode materials is essential
in order for the polymer electrolyte fuel cell (PEFC) to be able to
reach a broad market. Today, high platinum loadings are required,
especially on the cathode, to obtain sufficient activity for oxygen
reduction. In addition, electrode degradation causes loss of cata-
lyst surface area and requires high initial loadings to maintain the
cell performance over time. There are problems related to Pt also on
the anode side where poisoning of the catalyst, by e.g. CO, reduces
the activity. Approaches to improve the electrodes and reduce their
costs are continuously evaluated and include alternative catalysts
or supports as well as new structures and morphologies of the cat-
alyst layer [1]. Alternative catalysts, based on non-precious metals,
Pt alloys/mixtures, and/or novel supports should preferably reduce
the total amount of Pt, increase the activity, and be stable in the fuel
WO + xH + xe− ↔ HxWO2
+
2
The bronze formation is greatly affected by the water con-
tent, porosity, and also crystallinity, which in turn affect the
catalytic properties of tungsten oxide [10–12]. At the same time as
∗ Corresponding author. Tel.: +46 317725179; fax: +46 317723134.
E-mail address: bjorn.wickman@chalmers.se (B. Wickman).
0
013-4686/$ – see front matter © 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.electacta.2011.08.046

B. Wickman et al. / Electrochimica Acta 56 (2011) 9496–9503
9497
protons can be incorporated in the WOx structure, they also have
a significant mobility which means that WOx functions as a proton
conductor under these conditions [13]. The electric conductivity
of mesoporous tungsten oxide has been shown to increase by 2–4
hydrogen tungsten bronze formation is carried out and the role
of Pt in this process is discussed. In particular, we show that dif-
ferent results may be obtained when the system is evaluated in
liquid electrolyte compared to in fuel cell setup. Also, the system
changes dramatically when Nafion is incorporated in the porous
electrodes.
−3
−1
−1
orders of magnitude, up to 10
ꢀ
m , upon formation of hydro-
gen tungsten bronzes [14,15]. Since the hydrogen tungsten bronze
formation is dependent on the water content, large variation in con-
ductivity has been reported when varying the relative humidity [7].
Moreover, Pt supported on tungsten oxide has been shown to affect
the bronze formation and both an increased intensity of the hydro-
gen intercalation/deintercalation peaks as well as a shift of the peak
potential to higher potentials has been reported [6,16,17].
Tungsten oxide has been evaluated both as support and active
catalyst in fuel cell anode as well as cathode electrodes. Sole
tungsten oxide has displayed activity for hydrogen oxidation,
which was attributed to high porosity and high surface area
2. Experimental
2.1. Model electrode preparation
CARBEL CL gas diffusion layers were purchased from Gore Tech-
nologies and the surface was cleaned with flowing nitrogen gas
prior to film deposition. Thin catalyst films were deposited by ther-
−
6
mal evaporation in vacuum (AVAC HVC600, at about 10 mbar).
W and Pt were evaporated on to the GDL at controlled thicknesses,
measured with a quartz crystal microbalance sensor located in the
chamber. The samples are named according to the amount of mate-
[
14]. Combined Pt and tungsten oxide based catalysts have been
investigated for methanol/ethanol oxidation [18–25], CO oxidation
26–30], hydrogen oxidation [14,31–34] as well as oxygen reduc-
[
rial in the following way; yPtzWO , represents a sample with y nm
x
tion [17,34–39]. Already in 1969, Hobbs and Tseung observed an
enhanced activity for hydrogen oxidation on platinized tungsten
oxide [31] which they further proved to be due to a hydro-
gen spill-over mechanism [32,33]. For methanol oxidation, the
Pt on WOx system has shown improved efficiency over Pt cata-
lyst due to both the spill-over of hydrogen from Pt to WOx but
also the ability of WOx to provide oxygen atoms at low potentials
and thereby avoiding CO-poisoning [6,21]. Others have attributed
the improved performance to an increased electrochemical active
surface area (ECSA) of Pt on WOx [25,34,40,41]. The improved
CO tolerance for Pt on WOx has been shown in several studies
of Pt deposited on top of z nm of W, which in turn was deposited on
GDL. The thickness (z) refers to metallic W, as it was deposited from
a metallic W source. Subsequent analysis showed that the W had
oxidized after the samples had been exposed to air and the samples
were thus named WOx.
2.2. Electrode morphology and chemistry characterization
The thin film catalysts morphology was characterized by scan-
ning electron microscopy (SEM, Leo Ultra 55 FEG) operating at 10 kV
in the secondary electron mode. Chemical characterization was
performed using X-ray photoelectron spectroscopy (XPS, Perkin
Elmer PHI 5000C ESCA system) using monochromatic Al K␣ radia-
[
14,21,27,29,30,40].
The spill-over of hydrogen between WOx and Pt has been sug-
◦
gested as being responsible also for the increased ORR activity
17,38]. Furthermore, an increased activity for reduction of hydro-
tion and a 45 take-off angle.
[
gen peroxide has also been observed in fuel cell measurements
by adding WO3 to the carbon supported platinum catalyst layer
2.3. Electrochemical evaluation
[
17,42]. Ota et al. suggested that the observed enhancement of ORR
Circular disks with a diameter of 14 mm were punched out
from the thin film electrodes and used as working electrodes (WE)
in electrochemical evaluation in liquid electrolyte and fuel cell
measurements. For some of the experiments the electrodes were
sprayed with a 2% Nafion in isopropanol solution. This was done
approximately ten times with drying in between until the surface
became shiny of a thin layer of Nafion.
activity on Pt-WO is due to a reduction of Pt oxide formation when
Pt is deposited on WOx [37]. Shim et al. observed that Pt-WO /C
catalysts rendered higher ECSAs and an improved fuel cell perfor-
mance compared to Pt/C catalyst [34]. A possible explanation for
this can be that WOx provides both protons and electrons to the
supported Pt catalyst [36,43].
3
3
Tungsten oxide is also relatively stable in acidic environment,
which is a prerequisite for use in polymer electrolyte fuel cell appli-
cations [35,44,45]. However, some dissolution of tungsten oxides
has been reported [11,30,35,45].
A three-electrode glass cell with a platinum basket as counter
electrode (CE) and the thin model electrode disk as WE was used for
the liquid electrolyte experiments. A thin gold thread was attached
to the carbon cloth of the thin model electrode to be used as cur-
rent collector. The reference electrode was a Hg/Hg SO electrode
In a previous study, we examined the impact of different metal
oxides on the stability and activity of platinum in thin model cath-
odes in a PEFC [46]. Pt on WOx did exhibit an improved activity
for oxygen reduction and possibly also an improved stability com-
pared to Pt alone. Interesting features such as reduced platinum
oxide formation and platinum catalyzed hydrogen tungsten bronze
formation were also seen when Pt was deposited on WOx.
In this study, we perform a more thorough investigation of the
characteristics of Pt and WOx in the PEFC. Thin film model elec-
trodes, fabricated by thermal evaporation of the catalyst directly
onto the gas diffusion layer (GDL) are used as electrodes. This
enables a direct comparison of tungsten oxide electrodes, with and
without platinum. It is also possible to use the same electrodes
for electrochemical evaluation in both liquid electrolyte and fuel
cell. This is important since the proposed beneficial effects of tung-
sten oxide depend on the formation of hydrogen tungsten bronzes
in which the degree of hydration plays an important role. The
system is evaluated by examining HOR, CO oxidation, as well as
cyclic voltammetry in inert environment. A detailed analysis of the
2
4
(R601 from Radiometer) with saturated K SO4 at room temper-
2
ature, connected to a Luggin capillary. The reference electrode
potential was 0.65 vs. the normal hydrogen electrode (NHE) at
◦
25 C. All potentials from liquid electrolyte measurements were
shifted in respect to this and are reported with respect to the NHE.
0.5 M H SO4 was used as electrolyte and was prepared from con-
2
centrated H SO (pro analysi, Merck) and Milli-Q water (Millipore).
2
4
All tests were done with nitrogen bubbling through the electrolyte
to avoid any reactions with dissolved oxygen. Cyclic voltammetry
was performed with a PAR 273A potentiostat, EG & G Princeton,
with a lower sweep limit of 0 and an upper limit of 1.35 V vs. NHE
−
1
at 200 mV s
.
In order to prepare MEAs for fuel cell testing, the thin model elec-
TM
trode disks were placed onto 45 mm × 45 mm pieces of Nafion
115 membranes (Nafion 115, Aldrich). The membranes had previ-
ously been cleaned by boiling in 3% H O2 for 1 h, in 0.1 M H SO4
2
2
for 1 h and finally in three successive baths of Milli-Q water for
1 h each after which the membranes were dried in an oven at

9
498
B. Wickman et al. / Electrochimica Acta 56 (2011) 9496–9503
Fig. 1. SEM images of the as-prepared thin film model electrodes deposited on GDL.
◦
8
0 C overnight. Commercial porous ELAT electrodes (30% Pt on
for 30 s at 1 MPa. The MEA was mounted in a laboratory PEEK
(polyether–etherketone) fuel cell, described elsewhere [47], with
cylindrical, 30 mm in diameter, graphite current collector with spi-
rally formed gas channels and a clamping force over the current
collectors of 380 N.
−2
Vulcan XC-72) with a loading of 0.5 mg Pt cm
the counter side of the membrane and used as combined counter
and reference electrode (CE/RE). The MEAs (thin film electrode
were placed on
◦
on GDL + membrane + ELAT electrode) were hot pressed at 135 C

B. Wickman et al. / Electrochimica Acta 56 (2011) 9496–9503
9499
For the standard case, if nothing else is mentioned, the gases
were humidified in heated humidifiers (Fuel Cell Technologies Inc.)
a)
F1s
Pt4f
O1s
Pt4p
Pt4d
C1s
◦
held at 77 C, resulting in 90% relative humidity (RH) in the fuel cell
◦
with a cell temperature of 80 C. Cyclic sweeps were performed
FKLL
W4p
with O or N on the WE and 5% H in Ar at the CE/RE fed in excess
2
2
2
W4f
W4d
−1
at 1 ml s . The very low loading on the WE (3 nm of Pt corresponds
−2
to a Pt loading of about 6 ␮g cm ), compared to the porous CE/RE,
made it possible to use 5% H2 in Ar which diminishes crossover
of hydrogen to the WE. The CE/RE was also assumed not to be
polarized due to the low current density. All measured potentials
from the fuel cell tests are referred to the reversible hydrogen elec-
trode (RHE) and the potential have therefore been corrected for the
F2s
4
5.5 mV shift due to the lower hydrogen partial pressure (5% H in
800
700
600
500
400
300
200
100
0
2
Binding Energy / eV
Ar).
After 2 h of constant gas flows at the correct temperature, the
electrochemical measurements always started with cyclic sweeps
in nitrogen to get a description of the surface processes on the elec-
trode. This was followed by an activation procedure by 2000 cycles
b)
c)
W4f7/2
Pt4f5/2
Pt4f7/2
3
Pt
Pt10WOx
W4f5/2
3
3Pt20WOx
3Pt40WOx
3Pt10TiOx
10WOx
20WOx
40WOx
−
1
at 20 mV s between 0.6 and 0.9 V vs. RHE in oxygen after which
further measurements were performed.
CO-stripping curves were recorded in the potential window
−
1
between 0.05 and 1.2 V, at a sweep rate of 20 mV s . After a cycle
in humidified N gas, the potential scan was stopped at 0.15 V and
2
the gas flow shifted to CO-containing gas (2% CO balanced with Ar)
for CO adsorption during 2 min. Thereafter the inlet gas was shifted
back to nitrogen for 5 min, to flush out remaining CO in the cell,
before subsequent stripping.
Polarization curves for H oxidation were performed with 100%
2
−1
H2 fed in excess at 2.75 ml s
on both the WE and CE/RE. The
−
1
potential was swept between −0.1 and 0.5 V, at 10 mV s
and
80
75
70
45
40
35
30
−
1
between −0 and 0.5 V at 1 mV s . To determine the cell resistance,
Binding Energy / eV
Binding Energy / eV
ac impedance spectra was conducted with an Autolab PGSTAT302N
Fig. 2. XPS spectra presented as overview spectra (a) and high resolution scans of
Pt(4f) (b) and W(4f) (c) regions for all as-prepared electrodes of thin films deposited
on GDL.
from 100 kHz to 1 Hz, both in N and H , at several potentials. Nova
2
2
1
.5 was used for analyzing the data and the real axis intercept was
used as a measure of the cell resistance.
perhaps with slightly smaller particle sizes than those on carbon.
An increased Pt dispersion has been discussed as one of the bene-
ficial effects of supporting Pt on tungsten oxides [34,39]. From this
analysis, however, we see no or minor differences in the Pt structure
on tungsten oxide compared to carbon. Another explanation for the
increased ECSA of Pt might be that the proton conducting proper-
ties of the WOx provides proton contact to more of the Pt in the
electrode. For the 3Pt10TiOx sample, used as reference for some of
the measurements in this work, the Pt forms isolated nanoparticles
while Ti effectively wets the carbon. Although there are some struc-
tural differences between the TiOx and WOx samples, the structure
of the Pt is essentially similar.
Both W and Ti were evaporated from the elements in metallic
form, but due to both the low amount of oxygen present in the
chamber during evaporation and the exposure to atmosphere after
deposition, the thin metal films are expected to spontaneously oxi-
dize. XPS analysis was employed to characterize the chemical state
of all electrodes prior to the electrochemical measurements. Fig. 2
shows overview spectra from 0 to 900 eV for the 3Pt, 10WOx, and
3Pt10WOx samples (Fig. 2a) as well as high resolutions scans on the
Pt(4f) (Fig. 2b) and W(4f) (Fig. 2c) core level regions for all samples.
To compensate for charging effects, arising mainly from small vari-
ations in contacting the samples during measurements, the spectra
were shifted to position the main C(1s) peak at a binding energy of
284.5 eV. These effects were minor, since the GDL is a good elec-
tric conductor and no spectra were shifted more than 1 eV. The
peak position of the Pt(4f7/2) core level electrons was found to be
71.4 ± 0.1 eV for all Pt containing samples, indicating that the Pt
was in metallic state [49] and that there was no influence on the
binding energy from the support materials. For the W(4f7/2), the
3
. Results and discussion
3.1. Materials characterization
The structure of the thin model electrodes was characterized by
SEM and XPS prior to the electrochemical evaluation. Fig. 1 shows
SEM micrographs of all samples, as-prepared. As can be seen, evap-
oration of 3 nm Pt on GDL results in a non-continuous Pt film of
separated nanoparticles, roughly 3–6 nm in size. This result agrees
well with previously presented data for physical deposition of Pt
on porous carbon supports [46,48]. The separated nanoparticles
are the result of the fact that Pt growth proceeds via the forma-
tion of isolated nanoparticles due to Pt atom surface diffusion on
the carbon support [48]. It can also be noted that the structure of
this electrode is very similar to a conventional PEFC electrode and
is thus a relevant model system. However, in this case the Pt par-
ticles are situated only in a very thin domain of the electrode as
the Pt deposits only on the exposed parts (projected area) of the
support. In Fig. 1 it can also be seen that tungsten forms a differ-
ent structure compared to Pt. Deposition of 10 nm of W results in a
smooth film that fully covers the exposed carbon. This is the result
of both the facts that the amount of W was larger than the amount
of Pt, and the intrinsic interactions between W and carbon. W effi-
ciently wets the carbon and does not form isolated particles, even
at lower thicknesses [46]. Due to the lower contrast between Pt and
W compared to Pt and carbon in the SEM, it is somewhat difficult to
identify the Pt structure precisely. However, it is clearly seen that
electrodes with Pt on W have a rougher surface than those with
only W. Pt seems to form more or less isolated particles also on W,

9
500
B. Wickman et al. / Electrochimica Acta 56 (2011) 9496–9503
-3
x 10
-4
x 10
2
a)
1
.5
0
0
1
-
-
0.5
3
3
3
Pt10WOx
Pt20WOx
Pt40WOx
10WOx
20WOx
40WOx
-
1
1.5
Pt reference (3Pt10TiOx)
0
-
2
0.2
0.4
0.6
0.8
1
Potential, E vs. RHE /V
0
0,2
0,4
0,6
0,8
1
1,2
E, Potential vs. RHE / V
-
5
x 10
6
Fig. 3. Cyclic voltammetry of electrodes with varying thickness of WOx, without
grey) as well as with 3 nm Pt on top (black). 3Pt10TOx (dashed) is shown as refer-
ence. The voltammograms were recorded in a single cell fuel cell at 80 C and 90%
b)
(
◦
5
4
3
2
1
0
−1
RH at a scan rate of 200 mV s
.
peak position was found to be 35.9 ± 0.2 eV for all W containing
samples, which agrees well with WO3 [49]. However, due to the
low penetration depth of the XPS, it is not certain that the entire
tungsten film has formed WO and, moreover, it is possible that the
3
oxidation state of the films may change during the electrochemical
measurements and, therefore, the tungsten films will be referred
to as WOx.
0
.2
0.4
0.6
0.8
1
Potential, E vs. RHE /V
3.2. Hydrogen tungsten bronze formation
Fig. 4. CO-stripping on 3Pt (dashed) and Pt on WOx (solid) for electrodes with Nafion
coating (a) as well as without Nafion (b). Black lines represent the first positive scan
whereas the grey lines are the second positive scan. For the WOx samples, 3Pt10WOx
was used in the measurements shown in (a) and 1.5Pt10WOx was used in (b). The
In order to study the formation of hydrogen bronzes and the
interaction between Pt and WOx in this process, samples with vary-
ing thicknesses of WOx, with and without a top layer of 3 nm Pt
were analyzed in the fuel cell. Fig. 3 shows cyclic voltammograms of
electrodes with WOx thicknesses from 10 up to 40 nm. Most of the
current response at potentials below about 0.5 V is resulting from
the formation of hydrogen bronzes. For the Pt containing samples,
however, some of the current in this region stems from hydrogen
adsorption/desorption (HUPD) on Pt. To illustrate the current from
HUPD on Pt, the 3Pt10TiOx sample is displayed as reference. This
sample was used since it was shown in a previous study [46], which
will be discussed further below, that the electrochemically active
surface area (ECSA) of Pt is increased when deposited on WOx and
TiOx compared to Pt deposited on the GDL. With 3Pt10TiOx the
ECSA was similar to 3Pt10WOx, but the sole TiOx do not display
any activity in the HUPD region, in contrast to WOx.
For all samples with Pt, the onset potential for bronze formation
was shifted to lower potentials, which is contrary to the observa-
tions made by Kulesza et al. [16,17]. Actually, both the inclusion
and exclusion of protons was shifted to lower potentials when Pt
was present. The rates of the reversible bronze formation were also
higher when Pt was present. This indicates that Pt is involved in the
bronze formation process. However, when comparing the charges
in the HUPD and bronze formation region of the electrodes with and
without platinum it was found that Pt had no or minor effects on the
amount of bronze formation. The charges were integrated down to
the same potential and the Pt containing samples were corrected
for the HUPD charge determined from the WOx-free 3Pt10TiOx. The
resulting charge should correspond to the bronze formation and
was similar between samples with and without Pt for the same
thickness of WOx. It is possible that the bronze charge was slightly
higher for the thinner WOx films (10 and 20 nm) and slightly lower
for the thickest WOx film (40 nm) when Pt was present, but the dif-
ferences were hardly significant. The hydrogen evolution did start
at lower potentials on the electrodes with only WOx but the shape
◦
voltammograms were recorded in a single cell fuel cell at 80 C and 90% RH at a scan
−
1
rate of 20 mV s
.
of the voltammogram did not change if the lower sweep limit was
increased.
The apparent difference in onset potential may, to some extent,
come from the overlap with Pt oxide reduction on the PtWOx
electrodes, but cannot fully explain large difference observed on
40WOx with and without Pt. It should be mentioned that initially,
all WOx electrodes with 3 nm Pt on top show similar response in the
cyclic voltammograms and not until after activation, the increas-
ing charge in the HUPD region with increasing WOx thickness could
−
1
be seen. The sweep rate was varied between 20 and 1000 mV s
,
which had no impact on the charge in the bronze formation region
and the reaction is therefore assumed to be a thermodynamically
controlled surface process under these conditions.
3.3. CO oxidation
CO-stripping was performed on Pt and PtWOx electrodes both
with and without Nafion and the results are shown in Fig. 4a and
b, respectively. For the Pt sample, the oxidation of CO is similar to
what has previously been shown in fuel cell measurements [50],
with a distinct stripping peak at 0.64 V. On the PtWOx samples,
the oxidation starts at much lower potentials. As seen in Fig. 4a, a
pre-peak between 0.2 and 0.4 V vs. RHE is visible before the main
peak around 0.62 V vs. RHE appears. These results are similar to
observations made by others [25,28,29], and the pre-peak has been
argued to be caused by oxidation of CO at WOx sites close to Pt
sites [29]. However, in the CO-stripping on the electrodes without
Nafion (Fig. 4b), it seems that the peak is related to the desorption of
hydrogen from the hydrogen tungsten bronzes. In the experiments
performed by Lebedeva et al. [28] and Maillard and co-workers

B. Wickman et al. / Electrochimica Acta 56 (2011) 9496–9503
9501
-
3
0
0
0
0
0
0
.6
.5
.4
.3
.2
.1
0
x 10
1
.5
0
a)
0
-
0.5
-
1
1
0
0,2
0,4
0,6
0,8
1
1,2
0
0.1
0.2
0.3
E, Potential vs. RHE / V b)
E, Potential vs. RHE / V
0.5
0
−1
Fig. 5. Hydrogen oxidation at 1 mV s on 3Pt (dashed), 3Pt10WOx (dotted) and
◦
Pt40WOx (solid) recorded in a single cell fuel cell at 80 C and 90% RH.
3
-0.5
[
26,29], measurements of the produced CO2 confirmed that the
pre-peak, at least partly, was a result of CO oxidation. However
for the electrodes used in this study and due to the observations
in Fig. 4b, the pre-peak may also come from hydrogen desorption
from the tungsten bronze which was not covered by CO. The fact
that CO oxidation is shifted to lower potentials for the PtWOx sam-
ples indicates that PtWOx is more tolerant for CO than Pt alone. In
a recent study, Micoud et al. [26] suggested that the access to oxy-
gen species in the form of oxy-hydroxide groups (W–OH) increases
the activity for CO oxidation at potentials up to 0.6 V vs. RHE. They
also suggest that the transport of CO to the active sites, the Pt–WOx
interface, is rate determining. We found, similar to Micoud et al.,
that a lower sweep rate increased the charge under the CO oxida-
tion peak which may confirm the importance of diffusion of CO to
the reactive sites. In Fig. 4b, the CO oxidation charge is higher for
-
1
0
0,2
0,4
0,6
0,8
1
1,2
E, Potential vs. RHE / V
Fig. 6. Cyclic voltammograms of 3Pt (dashed) and 3Pt10WOx (solid) after 10 cycles
grey) and 150 cycles (black) for electrodes with Nafion coating (a) as well as without
Nafion (b). The voltammograms were recorded in 0.5 M H2SO4 at room temperature
(
−1
at a scan rate of 200 mV s
.
tion, resulting in the low current densities at low potentials for Pt on
WOx. The impedance spectroscopy gives a response from the whole
electrode surface, i.e. Nafion in contact with WOx, and not only from
the active Pt sites. Since WOx acts as both proton and electron con-
ductor, it is unclear how the response will be for the impedance
spectra. In addition, the electron and proton conductivity is chang-
ing with the inclusion/exclusion of hydrogen, which changes with
potential, however no changes in the high frequency intercept were
seen in the potential range from 0.02 to 0.55 V. Even though the
performances of the PtWOx electrodes are worse than Pt at the rel-
evant anode potentials, they have a larger electrochemically active
surface area for hydrogen oxidation.
1
.5Pt on 10WOx than for 3Pt, and could possibly be explained by a
higher electrochemically active surface area of the PtWOx sample
due to the proton conductivity of the WOx. In this study, we have
not examined the ability of PtWOx to catalyze the reduction of CO2
which is an important issue when using reformate feed [28]. Lebe-
deva et al. studied the reduction of CO2 on PtWOx electrodes and
found that CO builds up on Pt sites with time, but more slowly than
on pure Pt [28].
3.4. Hydrogen oxidation
3.5. Electrochemically active surface area and stability
The activity for hydrogen oxidation was examined with pure
After fabrication, the model electrodes do not contain any
ionomer. However, Nafion can be sprayed on top of the electrode to
yield a high catalyst-Nafion contact without changing the structure
of the catalyst, which was also done in the CO-stripping measure-
ments above. This enabled evaluation of the influences of Nafion
while assuring that the rest of the electrode structure was con-
stant. The amount of Nafion deposited on the catalyst, in this
work, is higher than that of a typical electrode and the motiva-
tion for this was to ensure a high Pt utilization that was also similar
for the different electrodes. However, this implies that the effects
of Nafion observed here are somewhat exaggerated. Nonetheless,
these effects should be present also for thinner layers, although less
pronounced.
Fig. 6a shows cyclic voltammograms of Nafion coated 3Pt and
3Pt10WOx in liquid electrolyte after 10 and 150 cycles. 3Pt dis-
played those for platinum typical hydrogen adsorption/desorption
peaks (HUPD), which became more pronounced after 150 sweeps
whereas the charge in the Pt oxide region decreased slightly. The
rather large double layer capacitance reveals that the wetting of
the electrode and GDL was high as a consequence of the Nafion
impregnation. The hydrogen adsorption charge of Pt in liquid
H2 on both the WE and the CE/RE. Due to the high loading of
the CE, no or minor polarization losses were assumed to occur on
that electrode. In Fig. 5 the hydrogen oxidation polarization curves
of 3Pt, 3Pt10WOx and 3Pt40WOx are displayed. The curves were
iR-corrected by impedance spectroscopy and the values used for
iR-correction were 290, 120 and 450 mꢀ cm2 for 3Pt, 3Pt10WOx
and 3Pt40WOx, respectively. At low potentials, 3Pt displays slightly
higher current densities than 3Pt10WOx and much higher current
densities than 3Pt40WOx. At higher potentials, the current starts to
bend off and all electrodes reach a limiting current. This current is
most likely limited by kinetics and not by mass transport, which
was discussed in more detail in a recent study [51]. The higher
limiting current densities seen for 3Pt40WOx and 3Pt10WOx are
suggested to originate from their higher ECSA and could be related
to the proton conductivity and hydrogen permeability of the WOx
film. The difference in initial slope between the samples may be due
to uncompensated resistance from the limited electron conduction
in the WOx, and should be higher for a thicker layer of WOx The
deintercalation of protons in the PtWOx samples at low potentials
could possibly also have a negative effect on the hydrogen oxida-

9
502
B. Wickman et al. / Electrochimica Acta 56 (2011) 9496–9503
electrolyte may then be used as a measure of the maximum avail-
able ECSA of the thin Pt model electrodes. For the 3Pt sample, the
ECSA per geometric area was found to be 3.1 cm cm , which
agrees well with the SEM image shown in Fig. 1a. Initially, the
- 4
x 10
a)
2
−2
5
3
Pt10WOx sample exhibited a lower charge in the HUPD region
than the 3Pt and without the characteristic Pt or Pt oxide features
observed. After 150 sweeps, the ECSA had decreased even further,
suggesting a deactivation of the Pt catalyst or Nafion layer. This
could be caused by dissolution of WOx or other structural changes
making the electrode electrochemically inactive. Dissolution of the
WOx has previously been observed by Chen and Tseung [45], but the
effect was reduced by covering the surface with Nafion. In our case,
the decrease in ECSA was seen for several repeated measurements
of different Pt on WOx samples. Pereira et al. [30] suggested that dis-
solved WOx can induce degradation of Nafion and decrease proton
conduction, which might be a possible explanation for these results.
Fig. 6b displays the same electrodes as in Fig. 6a, but without Nafion
coating. The initial cyclic voltammogram of 3Pt10WOx resembles
the initial cycle with Nafion coating, but with a smaller double layer
capacitance charge. This is in contrast to 3Pt without Nafion, which
displayed a much lower ECSA than the Nafion coated sample. The
explanation may be a higher hydrophobicity of the GDL, making the
wetting of the 3Pt electrode difficult, without Nafion present. Obvi-
ously, 10 nm of WOx on top of the carbon was enough to improve
the wettability of the 3Pt10WOx electrode. After 150 sweeps, plat-
inum oxidation and reduction peaks appeared on the 3Pt10WOx
sample and more distinct hydrogen adsorption/desorption peaks
were seen. In fact, the voltammogram of 3Pt10WOx became similar
to the voltammogram of 3Pt with Nafion. As described above, dis-
solving WOx might have deactivated the catalyst or ionomer when
Nafion was present, but without the Nafion coating, such dissolved
WOx should simply diffuse into the solution. However, the WOx
dissolution must in either case be very low, as it would otherwise
lead to detachment of Pt. As this was not seen, it is likely that the
WOx was still present and ensured good contact between the Pt and
the electrolyte. In fact, the HUPD as well as the Pt oxide reduction
peaks is almost similar on the 3Pt10WOx sample without Nafion
and the 3Pt with Nafion. This implies that, under these conditions,
a film of WOx under the Pt can replace a Nafion layer on top.
Due to the different behavior of Pt and PtWOx electrodes with
and without Nafion and the changing PtWOx electrodes, no com-
plete evaluation of the bronze formation was made in liquid
electrolyte. However, it can be noted that no clear signs of bronze
formation were observed in liquid electrolyte compared to in fuel
cell, but this was not investigated further in this study.
0
5
10
b)
0
5
b)
-
0.8
1
1.2
0
.2
0.4
Potential, E vs. RHE /V
0.6
Fig. 7. Cyclic voltammograms of 3Pt (dashed) and 3Pt10WOx (solid) for electrodes
with Nafion coating (a) as well as without Nafion (b). In (b), 3Pt10TiOx (dotted) is
shown as reference. The voltammograms were recorded in a single cell fuel cell at
◦
−1
.
8
0 C and 90% RH at a scan rate of 200 mV s
seen for the Nafion covered sample in liquid electrolyte (Fig. 6a).
If dissolution of WOx was the cause for the decrease in liquid elec-
trolyte, this is most likely much less pronounced in the fuel cell.
In Fig. 7b, the CV curves of the electrodes without Nafion are
shown. The Pt oxide reduction peak is significantly higher on
3
Pt10WOx compared to 3Pt, similar to previous observations [46].
This is likely a sign of 3Pt10WOx having a larger ECSA, but due
to the overlap between the hydrogen tungsten bronze formation
and the Pt–HUPD, the latter could not be used to determine the
surface area. Instead, we propose that 3Pt10TiOx can be used as
Pt reference in the cyclic voltammograms, as discussed in Section
3
.2. The 3Pt10TiOx and 3Pt10WOx samples showed a similar Pt
oxide reduction response, indicating that the ECSA of these sam-
ples should be similar. This can be seen in Fig. 7b where it is also
clear that 3Pt10TiOx displays only Pt characteristics in the HUPD
area, i.e. no response from the TiOx in the cyclic voltammogram.
Thus, we propose that the HUPD on Pt for the 3Pt10WOx sample can
be estimated to be similar to that of the 3Pt10TiOx sample as they
show similar Pt oxide reduction charge. The remaining charge in
the Pt–HUPD region for the 3Pt10WOx sample can be assigned to
hydrogen bronze formation in the WOx.
In our previous study [46], it was shown that depositing Pt on
1
0 nm, or thicker, films of TiOx or WOx led to an increased ECSA of
Pt, compared to having Pt directly on the GDL, in fuel cell measure-
ments. This effect was not seen for thinner films of TiOx or WOx
(
3 nm), which showed an almost identical Pt area as the reference
Pt on GDL sample [46]. As a result of further investigations, we con-
sider that the increase in surface area for films of 10 nm and more,
to a large part can be explained by the proton conducting prop-
erties of both TiOx and WOx [13,43,52], perhaps in combination
with an increased wetting of the GDL. When thinner films (≤3 nm)
are deposited on the porous GDL, it is likely that they are not fully
covering the carbon surface and, thus, the proton transport path is
interrupted.
Fig. 7 shows the same type of CV curves as in Fig. 6, but measured
in fuel cell setup. In Fig. 7a it can be seen that electrodes sprayed
with Nafion prior to hot pressing attain a significantly higher ECSA
and also that the amount of Pt oxide formation on the 3Pt10WOx
sample is roughly the same as that for the 3Pt. This suggests that
there is no difference in Pt dispersion when Pt is evaporated on GDL
or WOx. In addition, no decrease in electrochemical active surface
area with time was seen for 3Pt10WOx (not shown here) as was
4. Conclusions
Thin-film model electrodes deposited on GDL were used to
examine several PEFC reactions in order to evaluate the influences
of incorporating WOx as a Pt support in the electrodes. Samples with
W thicknesses between 0 and 40 nm, with and without a top layer
of 3 nm Pt were fabricated by thermal evaporation. XPS measure-
ments indicated that WO was the most likely oxide species formed
3
during spontaneous oxidation as the samples were subjected air.
Pt was found to be in a metallic state and no shift in binding
energies was detected for the different samples. SEM characteriza-
tion revealed that the WOx efficiently covered the exposed carbon

B. Wickman et al. / Electrochimica Acta 56 (2011) 9496–9503
9503
particles in the microporous layer of the GDL. The 3 nm of Pt
deposited on top formed separated Pt nanoparticles, similar to a
commercial electrode.
[3] W.C. Conner, J.L. Falconer, Chem. Rev. 95 (1995) 759.
[4] Y.Y. Shao, J. Liu, Y. Wang, Y.H. Lin, J. Mater. Chem. 19 (2009) 46.
[5] E. Antolini, E.R. Gonzalez, Appl. Catal. B 96 (2010) 245.
[6] A.C.C. Tseung, K.Y. Chen, Catal. Today 38 (1997) 439.
The hydrogen tungsten bronze formation was clearly visible for
all WOx containing samples in fuel cell measurements. Pt changed
the behavior of the hydrogen inclusion and exclusion of the tung-
sten bronzes. When Pt was present, the onset of hydrogen inclusion
and exclusion were shifted to lower potentials and the rates of
inclusion/exclusion were higher. However the amounts of formed
bronzes were hardly affected by the Pt.
The increased Pt surface area on thicker WOx films was found to
be beneficial for the limiting current in the HOR. In addition, the oxi-
dation of CO was shifted towards lower potentials, indicating that
Pt on WOx is more CO tolerant than Pt. However, with WOx situated
between the Pt and carbon support there will be an increase in the
resistance of the electrode.
[7] M.S. Whittingham, Solid State Ionics 168 (2004) 255.
[
[
8] N. Yoshiike, S. Kondo, J. Electrochem. Soc. 130 (1983) 2283.
9] M.A. Habib, S.P. Maheswari, J. Electrochem. Soc. 138 (1991) 2029.
[
10] C.G. Granqvist, Sol. Energy Mater. Sol. Cells 60 (2000) 201.
[11] P.J. Kulesza, L.R. Faulkner, J. Electroanal. Chem. 248 (1988) 305.
[12] P.J. Kulesza, L.R. Faulkner, J. Am. Chem. Soc. 110 (1988) 4905.
13] Y.M. Li, M. Hibino, M. Miyayania, T. Kudo, Solid State Ionics 134 (2000) 271.
14] X.Z. Cui, H. Zhang, X.P. Dong, H.R. Chen, L.X. Zhang, L.M. Guo, J.L. Shi, J. Mater.
Chem. 18 (2008) 3575.
[15] H. Kamal, A.A. Akl, K. Abdel-Hady, Phys. B 349 (2004) 192.
16] P.J. Kulesza, L.R. Faulkner, J. Electrochem. Soc. 136 (1989) 707.
17] P.J. Kulesza, B. Grzybowska, M.A. Malik, M.T. Galkowski, J. Electrochem. Soc.
[
[
[
[
144 (1997) 1911.
[18] K. Miecznikowski, P.J. Kulesza, J. Power Sources 196 (2011) 2595.
19] X.Z. Cui, J.L. Shi, H.R. Chen, L.X. Zhang, L.M. Guo, J.H. Gao, J.B. Li, J. Phys. Chem.
B 112 (2008) 12024.
[
[
20] R. Ganesan, J.S. Lee, J. Power Sources 157 (2006) 217.
The model electrodes were also investigated in liquid elec-
trolyte, where it was seen that the electrodes behaved very
differently compared to the fuel cell case. In liquid electrolyte, Pt
samples sprayed with Nafion showed a high ECSA whereas in the
absence of Nafion, a very low ECSA was obtained. This is likely
related to the hydrophobic properties of the GDL, hindering the liq-
uid electrolyte to contact Pt when Nafion was not included. With a
layer of WOx between the GDL and Pt, however, a high ECSA was
seen in the absence of Nafion. It is likely that WOx can increase
the wetting of the GDL and thus enable a higher Pt utilization. The
Nafion coated samples of Pt on WOx displayed a decreased initial
ECSA and further cycling revealed a deactivation of the electrode.
This can be an effect of dissolved WOx degrading the Nafion or
poisoning the Pt.
In fuel cell environment, the negative aspects of including Nafion
on the Pt on WOx samples was not seen, indicating that the deacti-
vation process might be an effect of the liquid electrolyte. A similar
ECSA was observed for both Pt and Pt on WOx with Nafion. In the
absence of Nafion, an increased ECSA was observed for Pt on WOx
layers of 10 nm and thicker. This increased Pt utilization can be
related to WOx ability to conduct protons and thus render a proton
contact to a larger amount of Pt particles. It was found that WOx can
increase the Pt utilization for the HOR and increase the CO tolerance
on the anode catalyst.
[21] S. Jayaraman, T.F. Jaramillo, S.H. Baeck, E.W. McFarland, J. Phys. Chem. B 109
2005) 22958.
22] M.K. Jeon, H. Daimon, K.R. Lee, A. Nakahara, S.I. Woo, Electrochem. Commun. 9
2007) 2692.
[23] P.K. Shen, A.C.C. Tseung, J. Electrochem. Soc. 141 (1994) 3082.
(
[
(
[
24] A.K. Shukla, M.K. Ravikumar, A.S. Arico, G. Candiano, V. Antonucci, N. Giordano,
A. Hamnett, J. Appl. Electrochem. 25 (1995) 528.
[
25] L.X. Yang, C. Bock, B. MacDougall, J. Park, J. Appl. Electrochem. 34 (2004) 427.
[26] F. Micoud, F. Maillard, A. Bonnefont, N. Job, M. Chatenet, Phys. Chem. Chem.
Phys. 12 (2010) 1182.
[
27] F. Micoud, F. Maillard, A. Gourgaud, M. Chatenet, Electrochem. Commun. 11
2009) 651.
(
[28] N.P. Lebedeva, V. Rosca, G.J.M. Janssen, Electrochim. Acta 55 (2010) 7659.
[29] F. Maillard, E. Peyrelade, Y. Soldo-Olivier, M. Chatenet, E. Chainet, R. Faure,
Electrochim. Acta 52 (2007) 1958.
[
30] L.G.S. Pereira, F.R. dos Santos, M.E. Pereira, V.A. Paganin, E.A. Ticianelli, Elec-
trochim. Acta 51 (2006) 4061.
[31] B.S. Hobbs, A.C.C. Tseung, Nature 222 (1969) 556.
[32] B.S. Hobbs, A.C.C. Tseung, J. Electrochem. Soc. 120 (1973) 766.
[33] B.S. Hobbs, A.C.C. Tseung, J. Electrochem. Soc. 119 (1972) 580.
[34] J. Shim, C.R. Lee, H.K. Lee, J.S. Lee, E.J. Cairns, J. Power Sources 102 (2001) 172.
[35] A.J. Martín, A.M. Chaparro, L. Daza, J. Power Sources 196 (2011) 4187.
[
[
[
[
36] R. Wlodarczyk, A. Kolary-Zurowska, R. Marassi, M. Chojak, P.J. Kulesza, Elec-
trochim. Acta 52 (2007) 3958.
37] K. Ota, A. Ishihara, S. Mitsushima, K. Lee, Y. Suzuki, N. Horibe, T. Nakagawa, N.
Kamiya, J. New Mater. Electrochem. Syst. 8 (2005) 25.
38] Y.J. Huang, H.H. Dai, W.S. Li, G.L. Li, D. Shu, H.Y. Chen, J. Power Sources 184
(
2008) 348.
39] O. Savadogo, P. Beck, J. Electrochem. Soc. 143 (1996) 3842.
[40] X.Z. Cui, L.M. Guo, F.M. Cui, Q.J. He, J.L. Shi, J. Phys. Chem. C 113 (2009) 4134.
[41] Z. Zhang, X. Wang, Z. Cui, C. Liu, T. Lu, W. Xing, J. Power Sources 185 (2008) 941.
[
[
42] P. Trogadas, V. Ramani, J. Electrochem. Soc. 155 (2008) B696.
43] G. Orsini, V. Tricoli, J. Mater. Chem. 20 (2010) 6299.
Acknowledgments
[44] H. Chhina, S. Campbell, O. Kesler, J. Electrochem. Soc. 154 (2007) B533.
[
45] K.Y. Chen, A.C.C. Tseung, J. Electrochem. Soc. 143 (1996) 2703.
[
46] M. Wesselmark, B. Wickman, C. Lagergren, G. Lindbergh, Electrochim. Acta 55
MISTRA (Swedish Foundation for Strategic Environmental
Research) is gratefully acknowledged. The Competence Centre for
Catalysis is hosted by Chalmers University of Technology and finan-
cially supported by the Swedish Energy Agency and the member
companies AB Volvo, Volvo Car Corporation, Scania CV AB, GM
Powertrain Sweden AB, Haldor Topsoe A/S, and the Swedish Space
Corporation.
(
2010) 7590.
[47] J. Ihonen, F. Jaouen, G. Lindbergh, G. Sundholm, Electrochim. Acta 46 (2001)
2899.
[
48] P. Brault, A. Caillard, A.L. Thomann, J. Mathias, C. Charles, R.W. Boswell, S.
Escribano, J. Durand, T. Sauvage, J. Phys. D: Appl. Phys. 37 (2004) 3419.
49] J.F. Moulder, W.F. Stickle, P.E. Sobol, K.D. Bomben, in: J. Chastain (Ed.), Handbook
of X-ray Photoelectron Spectroscopy, PerkinElmer Corporation, Minneapolis,
[
1992.
[
50] R.W. Lindström, K. Kortsdottir, M. Wesselmark, A. Oyarce, C. Lagergren, G. Lind-
bergh, J. Electrochem. Soc. 157 (2010) B1795.
References
[51] M. Wesselmark, B. Wickman, C. Lagergren, G. Lindbergh, Electrochem. Com-
mun. 12 (2010) 1585.
[
52] H. Ekström, B. Wickman, M. Gustavsson, P. Hanarp, L. Eurenius, E. Olsson, G.
Lindbergh, Electrochim. Acta 52 (2007) 4239.
[
[
1] S. Litster, G. McLean, J. Power Sources 130 (2004) 61.
2] B. Hammer, J.K. Nørskov, Surf. Sci. 343 (1995) 211.