Вы находитесь на странице: 1из 8

Journal of Electroanalytical Chemistry 693 (2013) 34–41

Contents lists available at SciVerse ScienceDirect

Journal of Electroanalytical Chemistry


journal homepage: www.elsevier.com/locate/jelechem

An in situ technique for analyzing ionomer coverage in catalyst layers


Hiroshi Iden ⇑, Atsushi Ohma
Nissan Research Center, Nissan Motor Co., Ltd., Yokosuka, Kanagawa 237-8523, Japan

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

Article history: An in situ technique for analyzing ionomer coverage, hion, in catalyst layers was examined. Contributions
Received 3 September 2012 of Pt and water to electric double layer capacitance, Cdl, were extracted by carbon monoxide (CO) adsorp-
Received in revised form 16 January 2013 tion and the RH dependence of Cdl, respectively. Consequently, hion for Pt (hion,Pt) and the carbon support
Accepted 22 January 2013
(hion,C) were obtained by comparing Cdl attributed to different interfaces, namely, the interface between
Available online 1 February 2013
the electroconductor and the ionomer or that between the electroconductor and water. The validity of
the method was verified by various analyses and experiments, such as water adsorption analysis, analysis
Keywords:
of specific capacitance and electrochemical surface area analysis by CO stripping voltammetry. hion of the
Ionomer coverage
Ionomer
catalyst layer with two different carbon supports, either ketjen black (KB) or graphitized ketjen black
Catalyst layer (GKB), was evaluated with this method. In the case of a GKB-based catalyst layer, CO apparently did
Fuel cell not cover the entire Pt surface, making the analysis more complicated. In this case, applying Cdl of a
pseudo catalyst layer, consisting of the carbon support and the ionomer, was deemed as appropriate.
The analysis revealed that hion,Pt and hion,C were ca. 0.39 and ca. 0.25 in the KB system and ca. 0.49 and
more than 0.91 in the GKB system. Although the number of primary pores in GKB is very small, hion,Pt
seemed small compared with hion,C. From SEM and TEM images, it was found that Pt particles formed clus-
ters on the GKB surface. Presumably, the narrow voids of the Pt clusters contributed to the low hion,Pt.
Ó 2013 Elsevier B.V. All rights reserved.

1. Introduction [3,4] and electrochemical activity [5,6]. In recent reports, it was


proved that NafionÒ adsorption changes voltammogram of Pt sin-
A polymer electrolyte fuel cell (PEFC) is a promising primary gle crystal [7] and blocks other species from adsorbing on the
power source and is expected to be used on vehicles in the future. interface [8,9]. In this way, the understanding of the interaction be-
There are still challenges to be addressed in the commercialization tween the ionomer and Pt surface is growing in importance. There-
of fuel cell vehicles (FCVs), including power density, durability, fore, it is indispensable to understand the relation between the
sub-zero startup, and cost reductions. The greatest challenge of state of the ionomer, such as its thickness and coverage, and vari-
all from an industrial point of view is to reduce the cost. According ous performance parameters of the actual catalyst layers.
to a U.S. Department of Energy report, Pt accounts for about a quar- The thickness of the ionomer on Pt/C or highly oriented pyro-
ter of the total system cost [1]. The most effective measure for this lytic graphite (HOPG) has been observed by transmission electron
issue is to reduce Pt loading and to use Pt more effectively. microscopy (TEM) [10] and atomic force microscopy (AFM) [11]
Catalyst layers could play an important role in resolving these and found to be less than a few nm. Although these techniques
challenges, especially cost reductions. However, it is difficult to are helpful in visualizing the microstructure, they do not provide
characterize and model the catalyst layers. For one reason, the information that can be linked with performance. The reason is
transport of water, gas and protons and the electrochemical reac- that such methods examine only a small area under particular con-
tions occur simultaneously in the catalyst layers. In addition, the ditions, which may be somewhat different from the actual operat-
catalyst layers have a very complicated microstructure, character- ing conditions of a fuel cell. Therefore, it is necessary to have an
ized by a microporous structure consisting of Pt, its support (basi- in situ method capable of evaluating the state of the ionomer under
cally carbon) and a proton conductive ionomer. To design actual operating conditions. The authors previously established an
optimized catalyst layers, it is necessary to develop methods of in situ technique that makes it possible to estimate the average
evaluating and gaining an in-depth understanding of the phenom- thickness and coverage of the ionomer, hion [12,13]. This method,
ena involved [2]. While the ionomer is necessary for proton con- however, uses pseudo catalyst layers (PCL) consisting of the carbon
duction, it is reported that the ionomer may affect gas transport support and the ionomer and is not applicable to actual Pt catalyst
layers. This is because the method distinguishes the electric double
⇑ Corresponding author. Tel.: +81 46 867 5372; fax: +81 46 866 5336. layer capacitance, Cdl, due to the carbon/water interface and that of
E-mail address: h-iden@mail.nissan.co.jp (H. Iden). the carbon/ionomer interface based on the RH dependence of

1572-6657/$ - see front matter Ó 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.jelechem.2013.01.026
H. Iden, A. Ohma / Journal of Electroanalytical Chemistry 693 (2013) 34–41 35

Cdl. With actual catalyst layers, it may not be that simple as Pt the cathode or a PCL was also decal-transferred to the other side of
should also contribute to the RH dependence of Cdl. the membrane. The decal-transfer conditions were 150 °C 10 min,
In this study, we investigated an in situ method of analyzing the and 0.8 MPa. Polyethylene naphthalate (PEN) films (Q51, 25 lm,
ionomer coverage on Pt and the carbon support, respectively. The Teijin-DuPont) were bonded on the surrounding areas of the elect-
drawback of the previous method is that it was not possible to ex- rocatalyst layers and the PCLs as reinforcement. Gas diffusion lay-
tract the contribution of Pt and the carbon support to RH depen- ers (GDLs) with microporous layers (24BC, SGL Carbon) were used
dence of total Cdl separately. However, if one of the interfaces is in this study. The MEA was formed by assembling the GDLs on both
deactivated, the previous method will be applicable. It is well sides of the catalyst coated membrane (CCM). The MEA was assem-
known that carbon monoxide (CO) adsorbs specifically on Pt and bled in a single cell between bipolar plates with a straight flow
deactivates it. With CO adsorption, not only the electrochemical field made of graphite and gaskets made of silicone rubber. The
surface area (ECA), but also Cdl is suppressed [14]. Therefore, we at- compression pressure was kept uniform by measuring it with pres-
tempted to apply the previous method to actual catalyst layers by sure-sensitive paper (Prescale, Fuji Film).
deactivating Pt through CO adsorption. This new method was used
to evaluate the ionomer coverage of catalyst layers having two dif-
2.2. CO adsorption
ferent types of carbon support.
To deactivate Pt, 1% CO gas balanced with N2 was used. All of
the measurements using CO were carried out at 30 °C to enhance
2. Experimental
CO adsorption. As it is difficult to control RH by heating a bubbler
tank at such low temperature, it was controlled by mixing dry gas
2.1. Preparation of MEA
with the wet gas which goes through the bubbler. The flow system
used in this study is illustrated in Fig. 1. Flow rates of the dry and
Tables 1 and 2 show the membrane electrode assembly (MEA)
wet gases were controlled with mass flow controllers to prepare
specifications used in this study. The samples shown in Table 1
the specified RH condition in Table 3 and then mixed in a mixer
with different carbon materials and perfluorosulfonic acid (PFSA)
tube right before reaching the cell. The same system was used
membranes (NafionÒ NR211, thickness = 25 lm) contained Pt.
for the other gases as well. Before CO was supplied, pure H2 and
The other samples shown in Table 2 had PCLs on one side. Catalyst
N2 were fed to the anode and the cathode, respectively. The anode
and pseudo catalyst inks were prepared respectively by mixing one
filled with pure H2 was used as the reversible hydrogen electrode
of the following materials: (a) ketjen black (KB); (b) graphitized
(RHE) for all the electrochemistry in this study. The cathode poten-
ketjen black (GKB); (c) Pt/KB (catalyst: 46 wt.%, TEC10E50E, TKK);
tial was then set at 0.05 V vs. RHE and the current was monitored
and (d) Pt/GKB (catalyst: 46 wt.%) and the NafionÒ ionomer disper-
during CO supply. When the current reached almost zero, CO in the
sion (D2020, EW1000, DuPont) together so that the weight ratio of
system was purged with N2 for at least 15 min via the cathode
the ionomer content was 0.9 relative to the carbon support. A pro-
valves before the cell closed. After that, N2 was supplied to the
pylene glycol solution (50 wt.%) was also added to the ink to keep
cathode side to purge the remaining CO in the cell. Various electro-
its solid content at 19 wt.% for the catalyst layers and at 12 wt.% for
chemical measurements described in this section were then per-
the PCLs. The electrocatalyst layers and PCLs were fabricated on
formed. An HZ-3000 electrochemical measurement system
polytetrafluoroethylene (PTFE) substrates using a screen-printing
(Hokuto Denko, Japan) was used to perform all the electrochemical
technique. The square electrocatalyst layers measured 50 mm on
measurements.
one side and had a Pt loading of 0.35 mg cm2 and the PCLs had
a carbon loading of 0.70 mg cm2. These layers were heat treated
on the PTFE substrates for 30 min at 130 °C to remove the organic 2.3. Measurement of electric double layer capacitance
solvent.
A catalyst layer for the anode was decal-transferred to a square To measure Cdl of the catalyst layers and PCLs, electrochemical
membrane (72 mm on one side) and then another catalyst layer for impedance spectroscopy (EIS) was performed using the same

Table 1
Specifications of the MEAs with Pt-containing cathode.

Anode Cathode
Catalyst TEC10E50E (Pt 46 wt.%, TKK) TEC10E50E (Pt 46 wt.%, TKK) Pt/GKB (Pt 46 wt.%)
Carbon support Ketjen black Ketjen black Graphitized ketjen black
Loading 0.35 mg-Pt cm2 0.35 mg-Pt cm2
Ionomer NafionÒ (D2020, EW1000, DuPont) NafionÒ (D2020, EW1000, DuPont)
Ionomer/carbon weight ratio 0.9 0.9
Active electrode, area/cm2 25 25
GDL Carbon paper with MPL (24BC, SGL carbon) Carbon paper with MPL (24BC, SGL carbon)

Table 2
Specifications of the MEAs with pseudo catalyst layer.

Anode Cathode
Catalyst TEC10E50E (Pt 46 wt.%, TKK) – –
Carbon support Ketjen black Ketjen black Graphitized ketjen black
Loading 0.35 mg-Pt cm2 0.7 mg-C cm 2

Ionomer NafionÒ (D2020, EW1000, DuPont) Ò


Nafion (D2020, EW1000, DuPont)
Ionomer/carbon weight ratio 0.9 0.9
Active electrode, area/cm2 25 25
GDL Carbon paper with MPL (24BC, SGL Carbon) Carbon paper with MPL (24BC, SGL Carbon)
36 H. Iden, A. Ohma / Journal of Electroanalytical Chemistry 693 (2013) 34–41

Wet Wet

PR MFC MIX MIX MFC PR

H2 N2
BT BT
or
CO/N2

MFC Dry MFC

Cell Exhaust

Fig. 1. A schematic diagram of the flow system used in this study.

Table 3 2.6. Measurement of electrochemical surface area (cyclic


Test conditions for various electrochemical measurements. voltammetry)
Anode gas H2
Cathode gas N2 (or 1% CO balanced with N2) The ECA of the catalyst layer was also measured by cyclic voltam-
Back pressure Ambient metry (CV) under the same conditions as CO stripping voltammetry.
Cell temperature (°C) 30
Pure H2 and N2 were fed to the counter electrode and working elec-
Anode humidity (% RH) 5, 10, 20, 30, 40, 50, 60, 70, 80, 90, 100
Cathode humidity (% RH) 5, 10, 20, 30, 40, 50, 60, 70, 80, 90, 100
trode, respectively. After a certain waiting period, the cathode
Anode flow rate (NL min1) 1.0 valves were closed to reduce the N2 partial pressure at the cathode
Cathode flow rate (NL min1) 1.0 (with flow) during the measurement. Right after that, the cathode potential was
scanned between 0.02 and 0.9 V vs. RHE at a scan rate of 50 mV s1
for several cycles. The ECA was obtained from the amount of protons
single cell set-up and the same electrochemical system with an which were underpotentially deposited on Pt, HUPD, and the Pt load-
FRA 5020A frequency response analyzer (NF Corporation, Japan). ing. The amount of HUPD was calculated by extracting the quantity of
The impedance spectra were recorded in a frequency range of electricity due to hydrogen evolution reaction (HER) from total
20 kHz to 10 mHz with an ac signal of ±15 mV amplitude reduction current below the potential where the double layer region
(±30 mV for drier cases) under the various conditions shown in Ta- was observed. The Tafel curve for HER was calculated using a Tafel
ble 3. The cathode potential was held at 0.45 V vs. RHE during the slope and an intercept approximated at low potential region
EIS measurement. After the CO treatment, however, the potential (usually below 0.07 V vs. RHE) where the voltammogram is loga-
was held at 0.15 V in order to minimize CO desorption during rithmically linear. The roughness factor of the Pt disk electrode
the measurement. was also obtained by CV. The same electrolyte and cell as described
in 2.4. was used. The potential of the working electrode was scanned
2.4. Measurement of electrochemical surface area (CO stripping between 0.02 and 1.2 V vs. RHE at a scan rate of 50 mV s1 for sev-
voltammetry) eral cycles. The same calculation method as the catalyst layer was
used to calculate the quantity of electricity due to HUPD.
To measure the electrochemical area (ECA) of the catalyst layer,
CO stripping voltammetry was carried out under various condi- 2.7. Water uptake measurement
tions as shown in Table 3. After CO adsorption, the cathode poten-
tial was scanned between 0.02 and 0.95 V vs. RHE at a scan rate of In order to investigate water adsorption behavior, the adsorp-
5 mV s1 for five cycles. The quantity of electricity due to CO oxida- tion isotherms of water for Pt/KB and Pt/GKB were measured by
tion, QCO, was calculated from the difference in the voltammo- means of a volumetric method using an automatic vapor adsorp-
grams. The roughness factor of a Pt disk electrode with a tion apparatus (Belsorp 18 Plus-HT, Nihon Bell) at 30 °C. 0.03–
diameter of 5 mm (AFE2M050PT, PINE instrument Co.) was also 0.05 g of the samples was placed in a glass sample tube. To reduce
obtained by CO stripping voltammetry. For the Pt disk electrode, the Pt surface, the sample tube was dosed with hydrogen for
0.1 M of perchloric acid and a three-compartment electrochemical 15 min at 100 °C. As preconditioning, the sample tube was evacu-
cell consisting of an RHE, a Pt wire counter electrode and the Pt ated for over an hour at 100 °C before the measurement. Measure-
disk working electrode was used. The potential of the working ments were performed with and without hydrogen treatment for
electrode was scanned between 0.02 and 1.2 V vs. RHE at a scan both Pt/KB and Pt/GKB.
rate of 50 mV s1 for five cycles.

2.5. Measurement of electrochemical surface area (Potential step CO 3. Results and discussion
stripping voltammetry)
3.1. Ionomer coverage in KB system
Under low RH conditions, CO stripping was conducted by a po-
tential step method because the CO stripping voltammogram ob- In a simple Randles circuit like that in Fig. 2, an impedance, Z, is
tained with the method described above might have been skewed expressed as
due to extremely low proton conductivity. After CO adsorption,
the cathode potential was held at 0.3 V vs. RHE for 3 min and then Rct xR2ct C dl
Z ¼ Rmem þ j ð1Þ
stepped to 1.0 V and kept for 15 min. The same measurement was 1þx 2 R2 C 2 1 þ x2 R2ct C 2dl
ct dl
also carried out without any CO treatment to obtain the
background. QCO was obtained from the difference between the vol- where Rmem and Rct are the proton transport resistance of the mem-
tammogram and the background. brane and the charge transfer resistance of the catalyst layer,
H. Iden, A. Ohma / Journal of Electroanalytical Chemistry 693 (2013) 34–41 37

Fig. 2. A simple Randles circuit consisting of the proton transport resistance of the
membrane (Rmem), charge transfer resistance (Rct) and the electric double layer
capacitance (Cdl).

respectively. When Rct is high enough and can be regarded as infi- Fig. 4. Capacitance plots of a KB-based catalyst layer and a PCL at 100% RH and
nite, (1) can be simplified as 30 °C.

1
Z ¼ Rmem  j ð2Þ
xC dl are shown in Fig. 4 as a function of x2. As can be seen, capacitance
increased with decreasing frequency (increasing x2). Although
EIS was carried out under an inert gas atmosphere at 0.45 V vs. RHE. the plots looked stable at low frequency region, Cdl was obtained
In such cases, very little Faradic current, i.e., limiting current due to by extrapolating the plots up to x2 = 0 from the low frequency re-
crossover hydrogen through the membrane, is observed. As a conse- gion (high x2 region). For comparison, capacitance plots obtained
quence, Rct is notably high. For a more accurate analysis, however, with the CO treatment and the KB-based PCL are also shown in
Cdl was calculated in a different way. The imaginary part in Eq. Fig. 4. With the CO treatment, Cdl decreased to the same level as
(1), Zimg, can be expressed as that of the PCL. This result indicates that almost the entire surface
of Pt became inactive and did not work as Cdl any more due to CO
1 1
 ¼ þ C dl ð3Þ adsorption. If so, the difference between Cdl with and without the
xZ img x2 R2ct C dl CO treatment would correspond to Cdl of Pt. To verify this assump-
tion, the specific capacitance of Pt was estimated from the ECA and
when Rct is not negligible. According to Eq. (3), the relation between
the gap of Cdl. The ECA was measured by CO stripping voltammetry.
xZ 1
img and x
2
should give Cdl. The catalyst layer should be ex-
As shown in Fig. 5, the CO stripping voltammogram exhibited a
pressed as a much more complicated model such as a transmission
clear peak due to CO oxidation around 0.75 V in the first cycle.
line model [15] because it has a complex microporous structure.
QCO was obtained from the difference between the voltammo-
However, it can be simplified as shown in Fig. 2 in the low fre-
grams. Though it was not so large, the difference between the vol-
quency region because the current distribution is negligible. There-
tammograms was also calculated up to the 5th cycle. The ECA was
fore, Cdl was obtained on the basis of the relation between xZ 1 img
then obtained as 69.1 m2 g1 using a charge to the Pt surface area
and x2 in the low frequency region.
conversion factor of 420 lC cm2 Pt and QCO. As a result, the specific
An impedance spectrum of the KB-based catalyst layer was ob-
capacitance of Pt was estimated as 80:2 lF cm2 Pt , which corre-
tained at 100% RH and 30 °C (Fig. 3). The Nyquist plot was very
sponded well with the value reported in the literature [18]. Conse-
similar to ones reported in the literature [15,16]. As the counter
quently, it is reasonable to assume that Cdl due to Pt can be simply
electrode was also used as the reference electrode in this system
obtained by CO adsorption.
instead of a typical three electrode configuration, the influence of
The total double layer capacitance, Cdl,total, is assumed to consist
the anode could affect the spectrum. However, it is said that the
of Cdl derived from Pt, Cdl,Pt, and that derived from the carbon sup-
kinetics of hydrogen oxidation reaction (HOR) and hydrogen evolu-
port, Cdl,C. Moreover, considering the two kinds of electrolyte used,
tion reaction (HER) on Pt are very high and the current density is
namely the ionomer and water, each of these two Cdl sources can
very low when the frequency is low. Accordingly, the cell configu-
be further divided. Consequently, the following four kinds of inter-
ration probably does not matter to low frequency region. In addi-
faces: (i) Pt/ionomer (Cdl,Pt/i); (ii) Pt/water (Cdl,Pt/w); (iii) carbon/
tion, it is also said that adsorbed protons can works as pseudo
ionomer (Cdl,C/i); and (iv) carbon/water (Cdl,C/w) contribute to Cdl,total.
capacitance [17]. Therefore it probably does not greatly affect the
All of them contribute to Cdl under a wet condition without the CO
analysis even under the high frequency region. Capacitance plots
treatment. Accordingly, the contribution of Pt may be regarded as
negligible with the CO treatment, in other words, (iii) and (iv) work
as Cdl. Meanwhile, the contribution of water may be regarded as
negligible under a dry condition because the ionomer works as
an electrolyte even in a dry state. In this case, (i) and (iii) work
as Cdl. Only (iv) works as Cdl under a dry condition with the CO
treatment. The contributions of each of the above four kinds of
interfaces can be separated respectively by obtaining Cdl in those
four cases. Assuming that the specific capacitance derived from
the different electrolytes, namely, the ionomer and water, is the
same, the difference of Cdl between dry and wet conditions should
be attributed to the difference of Pt surface which is covered with
the ionomer or water. In that case, the ionomer coverage on Pt,
hion,Pt, can be simply expressed as
C dl;Pt=i
hion;Pt ¼ ð4Þ
Fig. 3. A Nyquist plot of a KB-based catalyst layer at 100% RH and 30 °C. C dl;Pt=i þ C dl;Pt=w
38 H. Iden, A. Ohma / Journal of Electroanalytical Chemistry 693 (2013) 34–41

20 50

Carbon
Pt/KB
Pt/KB (hydrogen treated)
-2

15 40

-1
Current density / mA cm

1st

Amount of adsorbate / mg g
Pt/GKB
2nd Pt/GKB (hydrogen treated)
10 3rd 30
4th
5th
5 20

0 10

-5 0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.1 0.2 0.3 0.4
Water activity / -
Potential vs. RHE / V
Fig. 7. Water adsorption isotherms for Pt/KB and Pt/GKB at 30 °C with and without
Fig. 5. CO stripping voltammograms of a catalyst layer with a KB support at 100%
hydrogen exposure.
RH and 30 °C.

Likewise, the ionomer coverage on carbon, hion,C, is While the atmosphere right before the water adsorption measure-
ment was air and the Pt surface must have been oxidized, that of
C dl;C=i the cathode during Cdl measurement was H2/N2 due to crossover
hion;C ¼ ð5Þ
C dl;C=i þ C dl;C=w hydrogen and the Pt surface must have been bare. Therefore, the
influence of the Pt surface state on the water adsorption isotherms
The verification of this assumption will be discussed later.
was investigated. Pt/C was exposed to hydrogen at 100 °C for
The RH dependence of Cdl was then investigated with and with-
15 min, and after evacuation for over an hour at 100 °C, water
out the CO treatment (Fig. 6). For comparison, Cdl of the KB-based
adsorption behavior was measured again. The results revealed that
PCL is also shown in Fig. 6. Cdl after the CO treatment was almost
the onset of water adsorption shifted to the higher water activity
the same as that of the PCL at high RH conditions such as 90% or
side and that the amount of water adsorbed at 5% RH was much
100% where interfaces consisting of either water or the ionomer
smaller: less than 3 mg g1 carbon in the KB system and less than
should work, although they were different at lower RH probably
2 mg g1carbon in the GKB system. Though the influence of adsorbed
because of different water adsorption behavior. As mentioned be-
water on Cdl varies depending on the double layer structure (thick-
fore, this indicated that the entire Pt surface was covered with
ness of water, ratio of adsorption on Pt to that on carbon and so
CO and Cdl due to Pt became almost zero after the CO treatment.
on), such a small amount of water will probably not affect the re-
It also indicated that the carbon surface property of the CO-treated
sult. Without any hydrogen treatment, a Pt oxide presumably was
catalyst layer was the same as that of the PCL even after Pt depo-
formed and its higher hydrophilicity contributed to higher water
sition and the CO treatment. The implications of these observations
adsorption than the bare Pt surface. Based on this investigation,
will be discussed later. Cdl started to increase with an increase in
the amount of water in the catalyst layer at 5% RH under the actual
RH from 10% and saturated in the high RH region in both cases.
experimental environment seemed low enough, and it was con-
Presumably, Cdl was sufficiently low at 5% RH and water did not
cluded that applying Cdl at 5% RH to the estimation of hion did not
contribute to Cdl,total. As long as the support was KB (or GKB), the
matter. Furthermore, it is well known that CO adsorption interferes
amount of water was markedly small at 5% RH, after drying at
with water adsorption on bare Pt [19,20]. Considering this, it is rea-
90 °C for 5 h as described in our previous reports [12,13]. This sug-
sonable to say that water adsorption at 5% RH with the CO treat-
gests that adsorbed water does not affect Cdl at 5% RH. However,
ment was negligibly small.
whether this is true for Pt/C is still unclear.
Therefore, the values of Cdl at 5% RH and 100% RH were used for
To clarify this issue, water adsorption isotherms for Pt/C were
estimating hion. As a result, it was found that hion,Pt and hion,C were
measured at 30 °C (Fig. 7). Pt/C showed a considerable amount of
0.39 and 0.25 respectively. hion,C of the PCL was also calculated from
water even at 5% RH. The hydrophilic characteristic of Pt probably
Fig. 6 as ca. 0.3 at 30 °C. The estimated hion values are summarized
produced that amount of adsorption. In this respect, water could
in Table 4. hion,C of the catalyst layer was a little smaller than hion,Pt
still contribute to Cdl even at such low RH and make the analysis
and hion,C obtained with the PCL. These results suggest that Pt is a
difficult. The Pt surface state, however, was probably different.
little more likely to be deposited on the outer surface of the KB
support rather than inside the primary pores. Moreover, Pt on
120 the outer surface may have caused the difference seen in hion,C be-
tween the catalyst layer and the PCL because the voids between Pt
100 particles and the carbon support were so small that the ionomer
could not penetrate into them. Pt particles in primary pores may
C dl / F g-1carbon

80 Catalyst layer (w/o CO treatment)


Catalyst layer (w/ CO treatment) not matter because the ionomer apparently does not penetrate into
Pseudo catalyst layer
60

40 Table 4
Ionomer coverage on Pt and carbon.
20
hion,Pt hion,C
0 Pt/KB catalyst layer 0.39 0.25
0 20 40 60 80 100 KB-based pseudo catalyst layer – 0.3
Relative humidity / % Pt/GKB catalyst layer 0.49 >0.91
GKB-based pseudo catalyst layer – 1a
Fig. 6. RH dependence of a KB-based PCL and a catalyst layer with and without CO
a
treatment at 30 °C. Refs. [12,13].
H. Iden, A. Ohma / Journal of Electroanalytical Chemistry 693 (2013) 34–41 39

80 100
100% RH
80 5% RH
60
Pt
-1

ECA / m2 g-1Pt
ECA / m g

60
2

40

40
20
20

0
0 20 40 60 80 100 0
Relative humidity / % 5 to 100% RH 100% RH

Fig. 8. RH dependence of ECA of a KB-based catalyst layer obtained by CO stripping Fig. 9. RH dependence of ECA of KB-based catalyst layer obtained by CO stripping
voltammetry at 30 °C. voltammetry at 30 °C.

such pores as reported elsewhere [12,13]. This was also suggested the water content of the ionomer had greatly changed specific
by simulations based on coarse-grained molecular dynamics capacitance, those three values would not have been consistent
[21,22]. with each other. Therefore, the different water content of the iono-
For further verification of hion,Pt, the RH dependence of ECA was mer does not have a significant effect on the analysis. Conse-
evaluated by CO stripping voltammetry (Fig. 8). Here, the same quently, the aforementioned assumption about specific
charge to Pt surface area conversion factor was used for all RH con- capacitance is presumably valid. At the same time, the foregoing
ditions assuming that the CO coverage for both Pt/ionomer and Pt/ ideas that the carbon surface property was the same and that CO
water interfaces is the same. To verify this assumption, the rough- covered the Pt surface completely also hold true. Moreover, the ra-
ness factor and Cdl of a Pt disk electrode with a diameter of 5 mm tio of the ECA at 5% RH to that at 100% RH was found to be ca. 0.34.
were compared with and without a NafionÒ film. Without the Naf- There is not much difference between this value and hion,Pt obtained
ionÒ film, the roughness factor was obtained by measuring the from the Cdl analysis. Jinnai et al. reported that about 70% of the Pt
amount of the underpotential deposition of protons, HUPD, and particles were deposited inside the primary pores of KB, based on
the amount of CO respectively, after polished the Pt disk electrode transmission electron microtomography (TEMT), although the ob-
well with an alumina polishing suspension (MicroPolishÒ Alumina, served area was limited [24]. According to this result, about 30% of
BUEHLER). Cdl was determined from the cyclic voltammogram for Pt was on the outer surface of the KB support. The obtained hion,Pt
HUPD estimation As a result, roughness factors were determined and the ratio of the ECA (dry to wet) were not so far from their
to be 1.50 (by HUPD) and 1.63 (by CO stripping) and Cdl was result.
0.162 mF cm2. For preparing a NafionÒ film, the same NafionÒ
ionomer dispersion (D2020) as used for preparing the catalyst 3.2. Ionomer coverage in GKB system
ink was diluted with water (H2O) and propan-2-ol (iso propyl al-
chol, IPA) so that the weight ratio of those three materials was as hion,Pt in the GKB system was also evaluated in the same manner
D2020:H2O:IPA = 7:20:15. The NafionÒ film was then made by as that of the KB system. Fig. 10 shows the RH dependence of Cdl of
dropping 10 ll of the diluted dispersion onto the Pt disk electrode the GKB-based catalyst layer with and without the CO treatment.
after being polished well again and dryed in a fumehood over hion,Pt and hion,C were estimated to be ca. 0.34 and 0.91, respectively.
15 min. With the electrode, the roughness factors and Cdl were In order to verify the validity of hion, the RH dependence of Cdl of the
1.62 (by HUPD), 1.76 (by CO stripping) and 0.173 mF cm2. The ratio GKB-based PCL was investigated. Though the temperature was dif-
of the roughness factors and Cdl with the NafionÒ film to those ferent (80 °C), the results are shown in Fig. 10 for comparison and
without the film were 1.08 (by HUPD), 1.08 (by CO stripping) and were also reported elsewhere [12,13]. Cdl of the PCL at 100% RH and
(1.07). Consequently, it was found that the NafionÒ film did not 30 °C ð3:8 F g1 carbon Þ was almost the same as that at 100% RH and
matter to CO adsorption. In this way, the assumption was verified. 80 °C ð4:0 F g1 carbon Þ. In addition, it did not depend on RH. Therefore,
Under low RH conditions of 5%, 10%, or 20% RH, however, CO it was concluded that, although the temperature was different,
stripping was conducted by the potential step method. Under using the data at 80 °C as a comparison target would not affect
low RH and low temperature, CV was skewed due to extremely
low proton conductivity, making quantification difficult. It was
40
found that ECA strongly depended on RH and decreased with
decreasing RH. The same tendency was reported by Shinozaki
et al. [23]. Disappearance of the proton path consisting of water 30
C dl / F g-1carbon

might produce such a trend. To verify this assumption, another


Catalyst layer w/o CO
CO stripping was carried out. RH was raised to 100% without any Catalyst layer w/ CO
20
additional CO supply after CO stripping at 5% RH. CO stripping vol- Pseudo catalyst layer
tammetry was then performed again and the ECA determined by
the remaining CO was calculated. The ECA values are compared 10
in Fig. 9. The sum of the ECA at 5% RH and 5–100% RH corre-
sponded well to that normally obtained at 100% RH. This result
confirmed that a proton path consisting of water was present 0
0 20 40 60 80 100
and caused the RH dependence of the ECA and Cdl. Consequently,
Relative humidity / %
the specific capacitance of Cdl,Pt/i and Cdl,Pt/w was obtained from
Figs. 6 and 8 as 77.8 and 81:5 lF cm2 Pt , respectively. The values Fig. 10. RH dependence of a GKB-based catalyst layer with and without CO
are close to each other and consistent with the literature [18]. If treatment at 30 °C and that of a GKB-based PCL at 80 °C.
40 H. Iden, A. Ohma / Journal of Electroanalytical Chemistry 693 (2013) 34–41

50 literature [18]. Even if the first hypothesis about the different car-
bon surface properties is correct, it would not explain the decrease
40 in the specific capacitance of Pt. For that reason, the second
hypothesis was deemed reasonable. For further verification, Cdl
Pt
-1

due to uncovered Pt was calculated from the difference in the


ECA / m g

30
2

ECA values in Fig. 11 and the specific capacitance of Pt


20 ð80 lF cm2 1
Pt Þ. It was found to be 5:66 F gcarbon and corresponded
well with the difference in Cdl between the CO-treated catalyst
10
CO stripping
layer and the PCL ð5:50 F g1carbon Þ. Therefore, it is reasonable to as-
Under potential deposition of protons sume that Cdl of carbon is almost the same as that of the PCL. As
0 verified above, CO probably did not cover the entire Pt surface. Tak-
0 20 40 60 80 100 ing this into account, the actual hion,C may be larger than the hion,C
Relative humidity / % noted above because it might include the RH dependence of non-
Fig. 11. RH dependence of ECA of a GKB-based catalyst layer obtained by CO
CO-covered Pt when all the sites were not covered by the ionomer.
stripping and amount of underpotential deposition of proton at 30 °C. Therefore, hion,C in the GKB system was presumed to be larger than
0.91. Meanwhile, hion,C in the GKB system was reported to be 1 in
Refs. [12,13]. Though the exact value of hion,C was not certain, Cdl
the analysis. Cdl of the catalyst layer was larger than that of the PCL due to carbon was very small in this case and had little influence
even after the CO treatment. Two hypotheses for the difference can on the analysis. Actually, the value of hion,Pt was estimated to be
be suggested. One is that Pt deposition gave rise to different carbon 0.49 and did not change with different hion,C values (0.91 and 1).
surface properties. The other is that a portion of the Pt surface was A comparison of hion,Pt with different carbon supports revealed
not covered with CO. that GKB showed a higher value. This seems reasonable because
To investigate these hypotheses, the specific capacitance of Pt the number of primary pores in GKB is very small, so Pt particles
was investigated. For calculating the specific capacitance, however, should be deposited on its outer surface. Even so, the value still
it is necessary to know the ECA. Applying the ECA obtained from seemed too small compared with hion,C of the GKB system. In order
CO stripping did not seem appropriate in this case because CO to examine this difference, the microstructure of Pt/C was observed
may not have covered the entire Pt surface as mentioned above. under a scanning transmission electron microscope (HD-2000, Hit-
For this reason, the ECA of the GKB-based catalyst layer was eval- achi High-Technologies Corporation). STEM images are shown in
uated by both CO stripping and a conventional method that uses Fig. 13. Both SEM and TEM images showed that Pt particles agglu-
the underpotential deposition of protons, HUPD (Fig. 11). CO strip- tinated and formed clusters on the GKB surface. Although it was
ping was conducted by the potential step method with RH of 5%, difficult to measure the voids of the clusters exactly from these
10%, and 20% for the same reason as with the KB system. A compar- images, they were assumed to be smaller in size than the Pt parti-
ison of the results obtained with the two methods revealed that CO cles (2–3 nm). It is quite unlikely that the ionomer can penetrate
stripping showed a lower ECA value, and this trend did not contra- into such small voids easily and cover the entire Pt surface. Assum-
dict the second hypothesis. The specific capacitance of Pt was cal- ing it cannot, Pt/water interface can be formed inside of the Pt clus-
culated from the ECA evaluated from the amount of HUPD and the ters and work as Cdl under high RH conditions just like inside of the
difference in Cdl with and without the CO treatment. It was found primary pores in the KB system. In other words, the voids of the Pt
to be only 58 lC cm2 Pt and much smaller than that given in the clusters could account for the RH dependence of Cdl,Pt and ECA. The

Pt/KB Pt/KB

Pt/GKB Pt/GKB

Fig. 12. (Left) Scanning electron microscope and (right) transmission electron microscope images of Pt/KB and Pt/GKB.
H. Iden, A. Ohma / Journal of Electroanalytical Chemistry 693 (2013) 34–41 41

40 namely, water adsorption analysis, analysis of specific capacitance


30 and ECA analysis by CO stripping voltammetry.
-2

In the case of the GKB-based catalyst layer, the results sug-


Current density / mA cm

20
gested the possibility that CO did not cover the entire Pt surface,
10 making the analysis more complex. That possibility was verified
0 by undertaking the following experiments: (I) a comparison with
Cdl of the GKB-based PCL; (II) specific Cdl analysis; and (III) a com-
-10
5% RH 10% RH
parison of ECA values obtained with different methods. As a result,
-20 20% RH 30% RH the possibility was deemed appropriate and hion,Pt of the GKB-based
40% RH 50% RH
60% RH 70% RH catalyst layer was estimated to be ca. 0.49 using the Cdl of the GKB-
-30 80% RH 100% RH
based PCL.
-40
0.0 0.2 0.4 0.6 0.8 1.0 Although the number of primary pores in GKB is very small,
Potential / V vs.RHE hion,Pt was small compared with hion,C. SEM and TEM images re-
vealed that Pt particles formed clusters on the GKB surface. The
Fig. 13. Cyclic voltammograms of the catalyst layer with the GKB support at voids of the Pt clusters, which are very narrow, probably contrib-
different RH and 30 °C.
uted to the low hion,Pt. Although this contradicted the RH indepen-
dence of the ECA shown by CO stripping, the difference can be
difference in hion,Pt and hion,C in the GKB system is probably attrib- explained by considering the surface diffusion of adsorbed CO on
utable to the poor dispersibility of Pt. the Pt surface. This analysis also suggested that CO stripping vol-
This does not, however, explain the RH independence of the ECA tammetry may result in overestimation of the ECA depending on
obtained by CO stripping. The ECA values calculated from the the microstructure of the sample and RH condition.
amount of HUPD showed RH dependence. The measured CVs used
in the ECA calculation are shown in Fig. 13. In fact, the CVs under Acknowledgements
low RH conditions skewed severely because of markedly low pro-
ton conductivity, so the ECA under such conditions was probably The authors acknowledge the support of Nissan Motor Co., Ltd. in
not evaluated accurately. However, the ECA still seemed to depend financing the publication cost of this article. We thank Mr. Masaei
on RH aside from the skewed CVs such as the values at 5%, 10%, and Arai for providing technical assistance and Mr. Tetsuya Aoki for
20% RH. It is well known that adsorbed CO diffuses on the Pt sur- suggesting effective ideas concerning EIS measurement.
face and its diffusivity has been investigated in various ways
[25–28]. This likely explains the reason why CO stripping did not References
show any RH dependence of the ECA. Specifically, CO molecules ad-
[1] D. Papageorgopoulos, DOE Fuel Cells Annual Merit, Review, 2012.
sorbed at a site isolated from proton network can move to other [2] A. Ohma, T. Mashio, K. Sato, H. Iden, Y. Ono, K. Sakai, K. Akizuki, S. Takaichi, K.
sites due to surface diffusion. When CO reaches sites connected Shinohara, Electrochim. Acta 56 (2011) 10832.
to the proton network in the system, it can then be oxidized at a [3] S. Gottesfeld, I.D. Raistrick, S. Srinivasan, J. Electrochem. Soc. 134 (1987) 1455.
[4] S.K. Zecevic, J.S. Wainright, M.H. Litt, S.L. Gojkovic, R.F. Savinell, J. Electrochem.
high potential probably via a so-called Langmuir–Hinshelwood Soc. 144 (1997) 2973.
mechanism [28–30] involving adsorbed CO and an oxygen-con- [5] R. Subbaraman, D. Strmcnik, A.P. Paulikas, V.R. Stamenkovic, N.M. Markovic,
taining species such as PtOH. The surface diffusion coefficient, Chem. Phys. Chem. 11 (2010) 2825.
[6] A. Ohma, K. Fushinobu, K. Okazaki, Electrochim. Acta 55 (2010) 8829.
DCO, on Pt/GC (particle size: 3.1 nm) at 298 K was calculated to [7] R. Subbaraman, D. Strmcnik, V. Stamenkovic, N.M. Markovic, J. Phys. Chem. C
be 3  1014 cm2 s1 [27]. The mean diffusion length of adsorbed 114 (2010) 8415.
CO was simply estimated from this DCO and the result showed that [8] A.M. Gómez-Marín, A. Berná, J.M. Feliu, J. Phys. Chem. C 114 (2010) 20130.
[9] M. Ahmed, D. Morgan, G.A. Attard, E. Wright, D. Thompsett, J. Sharman, J. Phys.
it can diffuse to adjacent particles in less than 4 s when the Pt par- Chem. C 115 (2011) 17020.
ticle size is 3 nm. This seems rapid enough considering the scan [10] K.L. More, K.S. Reeves, DOE Report on May 23, 2005.
rate for CO stripping voltammetry (5 mV s1). It is even more so [11] T. Masuda, H. Naohara, S. Takakusagi, P.R. Singh, K. Uosaki, Chem. Lett. 38
(2009) 884.
for the low RH cases in this study because the potential was kept
[12] H. Iden, A. Ohma, K. Shinohara, J. Electrochem. Soc. 156 (2009) B1078.
at 1.0 V for 15 min under such conditions. In contrast, the ECA of [13] H. Iden, K. Sato, A. Ohma, K. Shinohara, J. Electrochem. Soc. 158 (2011) B987.
Pt/KB depended on RH despite the same CO stripping procedure. [14] V. Climent, N. García-Araez, E. Herrero, J. Feliu, Russian J. Electrochem. 42
(2006) 1145.
In the case of Pt/KB, Pt particles dispersed well on the carbon sur-
[15] M.C. Lefebvre, R.B. Martin, P.G. Pickup, Electrochem. Solid-State Lett. 2 (1999)
face as can be seen in the SEM and TEM images in Fig. 12. In such 259.
cases, CO surface diffusion should not affect the ECA measurement [16] E.B. Easton, P.G. Pickup, Electrochim. Acta 50 (2005) 2469.
by CO stripping. On these grounds, it is inferred that Cdl and the [17] B.E. Comway, J. Barber, S. Morin, Electrochim. Acta 44 (1998) 1109.
[18] S.S. Kocha, in: W. Vielstich, A. Lamm, H.A. Gasteiger (Eds.), Handbook of Fuel
ECA of Pt/GKB depend on RH due to the voids in the Pt clusters Cell, vol. 3, John Wiley and Sons Ltd., Chichester, England, 2003, p. 556.
on the GKB surface. In addition, an ECA evaluation using CO strip- [19] H. Ogasawara, J. Yoshinobu, M. Kawai, Chem. Phys. Lett. 231 (1994) 188.
ping voltammetry may result in an overestimation depending on [20] L. Luo, M. Futamata, Electrochem. Commun. 8 (2006) 231.
[21] K. Malek, M. Eikerling, Q. Wang, T. Navessin, Z. Liu, J. Phys. Chem. C 111 (2007)
the microstructure of the sample and RH condition because of 13627.
the surface diffusion of adsorbed CO. [22] K. Malek, T. Mashio, E. Michael, Electrocatalysis 2 (2011) 141.
[23] K. Shinozaki, H. Yamada, Y. Morimoto, J. Electrochem. Soc. 158 (2011) B467.
[24] H. Jinnai, R.J. Spontak, T. Nishi, Macromolecules 43 (2010) 1675.
4. Conclusions [25] J. Ma, X. Xiao, N.J. DiNardo, M.M.T. Loy, J. Phys. Rev. B 58 (1998) 4977.
[26] P.K. Babu, J.-H. Chung, E. Oldfield, A. Wieckowski, Electrochim. Acta 53 (2008)
An in situ technique was developed for analyzing hion,Pt Pt and 6672.
[27] T. Kobayashi, P.K. Babu, L. Gancs, J.H. Chung, E. Oldfield, A. Wieckowski, J. Phys.
hion,C in actual fuel cell catalyst layers using CO adsorption and Chem. B 127 (2005) 14164.
the RH dependence of Cdl. This method was used to investigate hion [28] G. Samjeske, K. Komatsu, M. Osawa, J. Phys. Chem. C 113 (2009) 10222.
of catalyst layers with different types of carbon support. For the [29] M. Bergelin, E. Herrero, J.M. Feliu, M. Wasberg, J. Electroanal. Chem. 467 (1999)
74.
KB-based catalyst layers used in this study, hion,Pt and hion,C were [30] N.P. Lebedeva, M.T.M. Koper, J.M. Feliu, R.A. van Santen, J. Electroanal. Chem.
found to be ca. 0.39 and 0.25, respectively. The validity of the anal- 524 (2002) 242.
ysis was verified on the basis of various analyses and experiments,