Reaction mechanism of preferential oxidation of carbon monoxide on Pt, Fe, and Pt-Fe/mordenite catalysts

Article abstract of DOI:10.1016/j.jcat.2005.09.026

We have investigated the preferential oxidation (PROX) of carbon monoxide at Pt/mordenite (Pt/M), Fe/mordenite (Fe/M), and Pt-Fe/mordenite (Pt-Fe/M) for a purification of reformates to supply polymer electrolyte fuel cells (PEFCs). Pt-Fe/M exhibited remarkable PROX activity up to an extremely high space velocity (i.e., ca. 100% selectivity, SV=～105 h-1) even at 50 °C, although Pt/M and Fe/M had negligibly small PROX activity. CO, H₂, and O₂ chemisorption measurements demonstrated that Pt sites act as adsorption sites for CO and/or H₂ and Fe (dominantly FeO) sites only for O₂, so that the addition of Fe to Pt/M can preserve O₂ adsorption sites for the PROX reaction even in CO/excess H₂ gas flow. The poor reactivity of Pt/M and Fe/M can be ascribed to the lack of CO and/or O₂ adsorption as the essential requisite for the Langmuir-Hinshelwood mechanism. We propose the so-called bifunctional mechanism for the distinctive performance at Pt-Fe/M, where the Pt site acts as a CO adsorption site and the Fe site acts as an O₂ dissociative-adsorption site and enhances the surface reaction between the reactants on the neighboring sites.

Full text of DOI:10.1016/j.jcat.2005.09.026

Journal of Catalysis 236 (2005) 262–269
www.elsevier.com/locate/jcat
Reaction mechanism of preferential oxidation of carbon monoxide
on Pt, Fe, and Pt–Fe/mordenite catalysts
Masashi Kotobuki a, Akiko Watanabe a, Hiroyuki Uchida a, Hisao Yamashita b,
Masahiro Watanabe b,∗
a
Interdisciplinary Graduate School of Medicine and Engineering, University of Yamanashi, Takeda 4, Kofu 400-8510, Japan
b
Clean Energy Research Center, University of Yamanashi, Takeda 4, Kofu 400-8510, Japan
Received 7 June 2005; revised 8 August 2005; accepted 21 September 2005
Available online 7 November 2005
Abstract
We have investigated the preferential oxidation (PROX) of carbon monoxide at Pt/mordenite (Pt/M), Fe/mordenite (Fe/M), and Pt–Fe/mordenite
(Pt–Fe/M) for a purification of reformates to supply polymer electrolyte fuel cells (PEFCs). Pt–Fe/M exhibited remarkable PROX activity up to
5
−1
◦
an extremely high space velocity (i.e., ca. 100% selectivity, SV = ∼10 h ) even at 50 C, although Pt/M and Fe/M had negligibly small PROX
activity. CO, H , and O chemisorption measurements demonstrated that Pt sites act as adsorption sites for CO and/or H and Fe (dominantly
2
2
2
FeO) sites only for O , so that the addition of Fe to Pt/M can preserve O adsorption sites for the PROX reaction even in CO/excess H gas
2
2
2
flow. The poor reactivity of Pt/M and Fe/M can be ascribed to the lack of CO and/or O adsorption as the essential requisite for the Langmuir–
2
Hinshelwood mechanism. We propose the so-called “bifunctional mechanism” for the distinctive performance at Pt–Fe/M, where the Pt site acts
as a CO adsorption site and the Fe site acts as an O dissociative-adsorption site and enhances the surface reaction between the reactants on the
2
neighboring sites.
2005 Elsevier Inc. All rights reserved.
Keywords: Preferential oxidation; Selective oxidation; Carbon monoxide; Zeolite catalyst; Hydrogen purification; Fuel cell
1. Introduction
In terms of the choice for purification, preferential oxidation
(PROX) of CO by O2 supplied to the fuel stream has been pro-
posed for removing CO from the reformates. PROX is advanta-
geous over the existing other technologies, such as membrane
separation or pressure swing methods, because it can operate at
relatively low temperatures and atmospheric pressure, resulting
in a compact reforming system that can quickly response to load
changes and frequent starting and stopping of the operation [3].
In this process, suppression of H2 oxidation or methane for-
mation accompanying the H2 loss is essential to achieve high
PEFC system efficiency. Therefore, high-performance PROX
catalysts are required that can be operated with the minimized
molar ratio of O2 to CO (λ = 2 × O2/CO) and high selectiv-
ity to CO oxidation. Such an operation is also important from
the standpoint of managing the heat in the PROX reactor, by
eliminating H2 oxidation with excess O2.
Fuel cells are clean, high-efficiency power generation sys-
tems. Polymer electrolyte fuel cells (PEFCs) are being exten-
sively developed as power sources for electric vehicle and res-
idential cogeneration, portable electric devices, and other uses.
Hydrogen is the most desirable fuel for PEFCs because of its re-
activity. One of the most realistic production hydrogen methods
is the reforming of hydrocarbon fuels, such as natural gas, gaso-
line, kerosene, or methanol, followed by a CO shift reaction to
form H2-rich fuel gases. However, such reformates still contain
1% level CO [1]. Because the anode Pt catalyst in PEFCs is
very sensitive to poisoning by CO, particularly at low operating
temperatures, the concentration of CO must be reduced to, say,
<10 ppm [2].
*
Until now, many researchers have studied the PROX cat-
alysts, such as noble metal catalysts supported on some ox-
Corresponding author. Fax: +81 55 254 0371.
E-mail address: m-watanabe@yamanashi.ac.jp (M. Watanabe).
0021-9517/$ – see front matter
doi:10.1016/j.jcat.2005.09.026
2005 Elsevier Inc. All rights reserved.

M. Kotobuki et al. / Journal of Catalysis 236 (2005) 262–269
263
ides [4–10] or carbon [11]. We first proposed zeolite catalysts
2.2. Activity test
loaded with Pt or the alloys for the PROX reaction, intend-
ing to use zeolite pores as the special reaction spaces [12].
Because they have larger molecular weights and also higher
chemisorbing properties than H2, CO and O2 are expected to
remain longer and thus have a greater chance of being ad-
sorbed on catalysts in the pores than do conventional alumina-
supported catalysts. Accordingly, zeolite-supported catalysts
are expected to selectively promote CO oxidation. We prepared
Pt or Pt alloy catalysts in zeolite cages in a highly dispersed
state by applying the ion-exchange method for the metallic
sources, followed by reduction with H2 gas. We found that
4 wt% Pt–2 wt% Fe/mordenite (mordenite-supported) exhib-
ited better activity and selectivity than Pt–Fe/mordenite with
other metallic compositions pretreated at 500 ◦C, and much
better activity and selectivity than any other single or Pt base
bimetallic catalysts supported on various zeolites pretreated at
the same 500 ◦C [12–15]. Moreover, we found that the PROX
performance of Pt/ZSM-5 was improved by lowering the pre-
treatment temperature from 500 to 300 ◦C [16]. Thus we ap-
plied this low preheating temperature to Pt–Fe/mordenite cat-
alysts with various compositions and found that 4 wt% Pt–
0.5 wt% Fe/mordenite catalyst demonstrated the best PROX
performance among them.
PROX activity tests were carried out in a conventional flow
reactor [12–14]. The 50-mg supported catalysts were mounted
in a Pyrex tube reactor (6 mm i.d.). Reaction gas containing 1%
CO, 0.5% O2, and H2 balance was prepared by mixing H2 con-
taining 1% CO and pure O2 using two mass flow controllers. In
a practical PROX reaction, H2O and CO2 should be included.
In the previous work [15], however, we studied the effect of
their presences on the PROX properties and found no noticeable
degradation of the properties. At the PROX reaction process in
zeolite cages, the exothermic reaction may suppress conden-
sation of H2O in the cages. Therefore, we use the reactant gas
composition without H2O and CO2 in the present work, because
the main purpose of this work is the discussion of the PROX
mechanism. Reaction temperature was measured with a ther-
mocouple attached to the outer wall of the reactor tube. Space
hourly velocity (SV), the ratio of total gas volumetric flow rate
to the catalyst volume mounted in the reactor tube, was used
as an index of catalyst performance. An on-line gas chromato-
graph with a thermal conductivity detector (TCD) (Hitachi GC
263-30) was used to measure inlet and outlet gas composition.
A 13X molecular sieve column was used to separate oxygen,
methane, and carbon monoxide, and an active carbon column
was used to analyze carbon dioxide.
In the present work we first demonstrate the superior PROX
performance on this 4 wt% Pt–0.5 wt% Fe/mordenite cata-
lyst. We then discuss the catalysis mechanism, based on studies
applying the adsorption/desorption measurements on the cata-
lyst in comparison with those of Pt/mordenite and Fe/morde-
nite.
The CO conversion, O2 conversion, and selectivity were de-
fined as follows:
CO conversion (%) = [CO2]/[CO]0 × 100,
O2 conversion (%) = ([O2]0 − [O2])/[O2]0 × 100,
Selectivity (%) = 0.5 × [CO2]/([O2]0 − [O2]) × 100,
where [CO2] is the concentration of CO2 in product gas, [CO]0
is the inlet carbon monoxide concentration, and [O2]0 and [O2]
are the oxygen concentrations at the inlet and outlet, respec-
tively. All of the data were recorded at the steady state. Before
the activity test, each catalyst was treated in O2 flow at 300 ◦C
for 1 h and switched to N2 flow. After 30 min in N2 flow, the
catalyst was treated in H2 flow for 1 h at 300 ◦C. This treatment
is called “H2 pretreatment” in this paper.
2. Experimental
2.1. Catalyst preparation
Pt2+ and Fe3+ were supported on a Na-type mordenite
(a reference sample supplied by the Catalysis Society of Japan,
denoted as M) by a conventional ion-exchange method [12–14].
First, the M powders were added into aqueous solution of
[Pt(NH3)4]Cl2 (Tanaka Kikinzoku Kogyo K.K.) and stirred at
room temperature for Pt2+ exchanging adsorption. After 15 h,
the suspension was filtered. The filtrated M was then added into
Fe(NO3)3 aqueous solution, stirred for Fe3+ exchange for 48 h
at 50 ◦C and filtered. The obtained samples were washed well
with deionized water, followed by drying at 60 ◦C for 24 h. The
resulting cakes were crushed to powders and sieved. The 100- to
200-mesh powders were stored in a desiccator. The amounts of
supported metals were determined with ICP (Seiko SPQ9000)
by measuring Pt and Fe concentrations in the filtrates. The re-
sulting loading amounts of Pt and Fe were 4 wt% and 0.5 wt%,
respectively. For the comparison, mordenite catalysts loading
single Pt or Fe in 4 wt% and 0.5 wt%, respectively, were also
prepared in a similar manner as the Pt–Fe supporting catalyst.
These catalysts were denoted by Pt/M, Fe/M, and Pt–Fe/M, re-
spectively.
2.3. Characterization
The Pt–Fe/M sample after the activity test was observed by
transmission electron microscopy (TEM) (Hitachi HD-2000)
operating at 200 kV. The Pt–Fe/M powders were embedded in
a resin. After the resin hardened, it was sliced by a microtome
for preparation of test pieces. Then TEM observation was per-
formed.
The BET surface area of the catalysts was determined with
the adsorption capacity of nitrogen using N2 physisorption
equipment (Bel Japan BEL-mini). Before measurement, each
catalyst sample was heat treated for the conditioning in a con-
ventional manner. Then nitrogen adsorption on the sample was
measured at liquid nitrogen temperature. For comparison with
the catalyst samples, the BET surface area of mordenite itself,
used as the catalyst support, was also measured.

264
M. Kotobuki et al. / Journal of Catalysis 236 (2005) 262–269
Affinities of Pt/M, Fe/M, and Pt–Fe/M catalysts to CO,
H2, and O2 adsorption were examined using commercial
chemisorption equipment (Bel Japan BEL-METAL4) as fol-
lows. Each ion-exchanged catalyst was first placed in a Pyrex
tube, then H2 pretreated under the conditions described in Sec-
tion 2.2. After pretreatment, the sample was cooled to each
measurement temperature in He flow. Then a CO, O2, or H2
pulse was introduced until the saturated coverage was achieved.
The effluent gas through the catalyst bed was detected by a
TCD. During the test, He gas for CO and O2 adsorption and Ar
gas for H2 adsorption were used as carrier gases. The amount
of each chemisorbed gas was calculated from the integration
of the reduced peak areas of detected pulses, corresponding to
Fig. 4.
Competitive adsorption of CO and H2 from the mixed gases
was also determined by exposing Pt/M and Pt–Fe/M catalysts
at 50 ◦C in a gas flow containing 0.5% CO, 5% H2, and He
balance for 1 h. Temperature-programmed desorption (TPD)
profiles of CO and H2 were collected with a commercial cat-
alyst analyzer (Bel Japan BEL-CAT) equipped with a mass
spectrometer by heating up to 400 ◦C at a 10 ◦C/min ramp rate,
corresponding to Fig. 5.
CO adsorption and the oxidative desorption at Pt/M and
Pt–Fe/M catalysts were examined at 50 ◦C, corresponding to
Figs. 6 and 7. The samples for the measurement were H2 pre-
treated, then cooled to 50 ◦C in He flow. Then CO (5% CO and
He balance) or CO + H2 (0.5% CO, 5% H2, and He balance)
was pulse-injected in He carrier gas flow until the saturated cov-
erage with CO and/or H2 was achieved. After the saturation,
O2 was pulse-injected in the same carrier gas flow to evalu-
ate the reactivity between the preadsorbed CO and/or H2 and
the postinjected O2 at the Pt/M and Pt–Fe/M catalysts. The
amounts of CO, H2, O2, H2O, and CO2 exhausted were moni-
tored with the same BEL-CAT device described above.
Fig. 1. PROX activity of different catalysts as a function of temperature. Reac-
−1
tant gas: 1% CO, 0.5% O , and H balance. Space velocity: 50,000 h (Q)
2
2
4 wt% Pt/M, (2) 0.5 wt% Fe/M, (!) 4 wt% Pt–0.5 wt% Fe/M.
those of Pt/M and Fe/M, particularly at low temperatures, but
similar to our previous results for Pt–Fe/M samples with higher
Fe content (Pt/Fe < 3/1) [15]. The CO conversion and selec-
tivity increase with decreasing operating temperature, although
O2 conversion decreases slightly below 150 ◦C because of de-
creased direct H2 oxidation. It must be emphasized that CO
conversion and selectivity exceed 90 and 95%, respectively,
even with the stoichiometric O2 supply (λ = 1.0) and an op-
erating temperature as low as 50 ◦C, which is very desirable for
cold starting of fuel cells.
3. Results and discussion
3.1. PROX properties of Pt/M, Fe/M, and Pt–Fe/M catalysts
3.1.1. Temperature dependence
Fig. 1 shows PROX properties of Pt/M, Fe/M, and Pt–Fe/
M catalysts at SV = 50,000 h−1 and λ = 1.0 (stoichiometric
amount of O2) as a function of reaction temperature. At all of
the catalysts, no CH4 was formed via useless consumption of
H2 reacting with CO or CO2 over the whole temperature re-
gion, which is distinctive from many other catalysts supported
on nonzeolite supports. At Fe/M, the PROX reaction hardly
occurred in our test conditions. Consistent with our previous
observations under different reaction conditions [13,14], Pt/M
showed poor PROX performance; that is, the CO conversion
commences above 150 ◦C, although O2 conversion starts at
around 100 ◦C. Both conversions increase with temperature el-
evation. The selectivity shows a maximum at ca. 200 ◦C, but
it does not exceed 60%. Wootsch et al. reported similar be-
havior of CO and O2 conversion on Pt catalyst supported on
γ -Al2O3 powder [17]. In contrast, Pt–Fe/M catalyst shows ex-
tremely high CO conversion and selectivity distinctive from
3.1.2. Space velocity dependence
The ability to achieve good PROX performance at a high SV
condition as well as the minimum λ (=1.0) at even low temper-
atures is critical for a compact and quickly responsible PROX
reactor. Fig. 2 shows the effect of SV values on the PROX per-
formances of Pt–Fe/M catalyst at 50 ◦C and λ = 1.0. CO and
O2 conversion and selectivity at SV = 12,500 h−1 are almost
100%. CO selectivity is suppressed slightly by increasing SV
value but nonetheless maintains a surprisingly high value of
ca. 95% at 100,000 h−1 under our operating conditions. The
mechanism responsible for such remarkable PROX behavior is
discussed in detail below.

M. Kotobuki et al. / Journal of Catalysis 236 (2005) 262–269
265
Fig. 3. TEM image on Pt–Fe/M catalyst after PROX activity test.
contents indicated that in the mordenite cages, Pt exists in
the metallic state, but Fe resembles FeO predominantly [18].
Therefore, surface areas normalized with 1 g mordenite are cal-
culated and shown in the same table, assuming that Fe is all in
the FeO phase, although this is hardly reflected in the resulting
surface areas. An increase in the normalized BET surface ar-
eas by loading catalysts on M, which must be ascribed to the
surface areas of metal catalysts, is almost hidden within exper-
imental errors due to the extremely high surface area of the M
support itself. So the difference in surface areas among metallic
catalysts is beyond the scope of this discussion of PROX behav-
ior. Nevertheless, it is reasonable to consider that the difference
might be small, because major parts of the catalyst surfaces are
located inside the support cages. Fig. 3 shows a typical TEM
photograph of a sliced sample of 4 w% Pt–0.5 w% Fe/M af-
ter the PROX reaction. Based on our previous work [15,16],
we have concluded that the large particles observed on a pe-
riphery of each sliced sample are the catalysts supported on the
viewing mordenite surface and that the small particles inside
are the catalysts supported in the mordenite cages. This pho-
tograph looks very similar to both TEM photos of other views
of the present sample and TEM photos of the 4 wt% Pt–2 wt%
Fe/M pretreated at 500 ◦C [15] or 6 wt% Pt/ZSM-5 pretreated
at 300 ◦C [16]. Thereby, the ratio of the total surface areas of
small metallic particles to the total surface areas of the metallic
particles including the large particles are evaluated by counting
particle numbers and their particle sizes, that is, several hun-
dreds of the small and large particles on several photographs.
As a result, it was found that metallic catalyst areas of ca. 85%,
80%, and 83% are located inside each zeolite cage at 4 w%
Pt–0.5 w% Fe/M, 4 wt% Pt–2 wt% Fe/M, and 6 wt% Pt/ZSM-
5, respectively. Previous work [16] clearly demonstrated that
such small particles insides the zeolite cages play the dominant
role in the selectivity to CO oxidation and the high reaction
rate in the PROX reaction. These properties must result from
the high specific surface areas of small catalyst particles and
the special reaction environment in zeolite cages, although the
wt% of the small particles to the large particles is relatively
small.
Fig. 2. PROX activity on 4 wt% Pt–0.5 wt% Fe/M as a function of space ve-
locity. Reactant gas: 1% CO, 0.5% O , and H balance, reaction temperature:
2
2
◦
50 C.
Table 1
BET surface area of various catalysts
Catalyst
Pt/M
Fe/M
Pt–Fe/M
M
2
Surface area (m /g
)
)
407
424
404
406
392
411
408
408
cat
2
a
Surface area (m /g
M
a
Assuming that Fe was in FeO phase.
3.2. Role of each catalyst site for reactant adsorption and
reaction
As mentioned here and in our previous papers [12–15], the
Pt–Fe/M catalyst shows superior performance for the PROX
reaction than the other catalysts and/or supports. We have pro-
posed a possible mechanism to explain this superiority; how-
ever, neither the behavior itself nor the mechanism has yet been
clarified. To investigate this mechanism, we further characterize
the catalyst in the present work.
3.2.1. Catalyst surface area
Table 1 shows BET surface areas of H2-pretreated Pt/M,
Fe/M, Pt–Fe/M, and the support M. Our previous XANES
analysis on H2-pretreated Pt–Fe/M catalysts with different Fe-

266
M. Kotobuki et al. / Journal of Catalysis 236 (2005) 262–269
Fig. 4. Saturating amounts of (a) CO, (b) H and (c) O adsorbed on (Q) Pt/M, (2) Fe/M, and (!) Pt–Fe/M as a function of temperature.
2
2
◦
◦
Fig. 5. TPD profiles for CO and H preadsorbed on (a) Pt/M, (b) Pt–Fe/M at 50 C. Heating rate = 10 C/min.
2
3.2.2. Adsorption property of each metal site at different
catalysts
ascribed to increased O2 coverage on Pt sites by the increased
d-vacancy of the Pt valence band [21]. As Fig. 4b shows, adding
Fe to Pt is effective in modifying CO and O2 adsorption but not
H2 adsorption; that is, the amount of H2 adsorption on Pt/M and
Pt–Fe/M does not differ and is smaller than the amount of CO
or O2 adsorption.
Fig. 4 shows the adsorption behaviors of Pt/M, Fe/M, and
Pt–Fe/M catalysts to pure CO, H2, and O2, respectively, as a
function of temperature. Fe/M adsorbs neither CO nor H2 over
the whole temperature range examined, but it does adsorb O2.
The amount of adsorbed O2 increases with increasing temper-
ature. In contrast, Pt/M and Pt–Fe/M catalysts adsorb both CO
and H2 noticeably, in decreasing amounts with increasing tem-
perature. Both catalysts also adsorb O2 appreciably, in increas-
ing amounts with increasing temperature. The difference in CO
and O2 adsorption between Pt/M and Pt–Fe/M (i.e., Pt–Fe/M
adsorbs less CO but more O2 than Pt/M) should be emphasized.
Our previous work on electrocatalysts used in fuel cell anodes
demonstrated that Pt–Fe alloy or Pt skin layer of nanothick-
ness formed on Pt–Fe alloy exhibits a low, steady CO coverage
(<0.5) in 100-ppm CO/H2 balance atmosphere even at room
temperature, although pure Pt shows the full coverage within a
short time. The improved CO tolerance from pure Pt was as-
cribed to a modification of the electronic structure of Pt sites
by the presence of Fe atoms in the vicinity [19,20]. The sup-
pression of CO adsorption seen in Fig. 4a is consistent with that
previous observation at electrocatalysis. In contrast, the adsorp-
tion of O2 onto Pt–Fe/M is markedly enhanced by the addition
of Fe to Pt/M, as is the preferential O2 adsorption at Fe/M to
Pt/M, as seen in Fig. 4c. This observation also agrees well with
our explanation of the enhanced O2 electroreduction in fuel cell
cathodes by alloying Pt with transition d-metals, which can be
Fig. 5 shows TPD profiles of CO and H2 preadsorbed on
Pt/M (a) and Pt–Fe/M (b) in 0.5% CO, 5% H2, and He bal-
ance at 50 ◦C, respectively. Desorption of CO (m/z = 28) is
observed from 50 to 300 ◦C at Pt/M. Two H2 desorption peaks
(m/z = 2) are seen at ca. 100 and over 250 ◦C. The former was
assigned to chemisorbed hydrogen on the metal surface [22];
the latter, to strongly chemisorbed hydrogen [23]. We have not
yet qualified the sites for the strongly bonding H2, but it seems
to be not so reactive. Pt–Fe/M shows quite similar TPD pro-
files to those of Pt/M. The amounts of CO and H2, as well
as their molar ratios, estimated from the desorption peak ar-
eas, are not significantly different between Pt/M and Pt–Fe/M
(although they are slightly smaller in the latter). Therefore, it
is thought that Pt sites in Pt–Fe/M and Pt/M have similar ad-
sorption properties to CO and H2. In any event, this result indi-
cates that not only CO, but also H2, can adsorb on the catalyst
sites during the practical PROX reaction. Because Fe sites in
Fe/M exhibit no affinity to CO and H2 adsorption, as shown in
Figs. 4a and 4b, it is reasonable to consider that CO as well as
H2 adsorb on Pt sites in Pt/M and Pt–Fe/M during the PROX
reaction.

M. Kotobuki et al. / Journal of Catalysis 236 (2005) 262–269
267
Fig. 6. Reactivity of CO preadsorbed on (a) Pt/M, (b) Pt–Fe/M toward pulse-injected O .
2
Fig. 7. Reactivity of CO and H competitively preadsorbed on (a) Pt/M, (b) Pt–Fe/M toward pulse-injected O .
2
2
3.2.3. Reactivity and selectivity of the CO adsorbed on Pt/M
and Pt–Fe/M
reaction also occurs. In contrast, at Pt–Fe/M, the initial O2 pulse
height decreases dramatically and forms CO2 (see Fig. 6b). The
pulse height for O2 increases stepwise and reaches to the full
height after 12 pulses and vice versa for CO2 formation. Thus,
it becomes clear that CO molecules adsorbed on Pt sites in Pt–
Fe/M are extremely reactive with injected O2, in contrast to
those in Pt/M.
How does the presence of coadsorbed H2 with CO affect the
reactivity and selectivity at the same catalysts? The experimen-
tal results are shown in Fig. 7. First, preadsorption of reactant
gases on Pt/M and Pt–Fe/M was performed in mixed gas stream
(0.5% CO, 5% H2, and He balance) at 50 ◦C. Pt/M tends to
Fig. 6 shows the results of mass spectrum analyses on the
preadsorption of CO and on CO oxidation with pulse-injected
O2 on Pt/M and Pt–Fe/M at 50 ◦C. In Pt/M (Fig. 6a), the height
of the CO peak (m/z = 28) increases with increasing pulse
time. Finally, its height becomes constant, indicating saturated
coverage with CO on Pt sites. After saturation, the catalyst is
exposed to O2 pulses. The height of every detected O2 peak
(m/z = 32) is constant, meaning at least that O2 adsorption
does not occur at the Pt sites saturated with CO in Pt/M. Be-
cause no CO2 (m/z = 44) formation is observed, no oxidation

268
M. Kotobuki et al. / Journal of Catalysis 236 (2005) 262–269
achieve the saturated adsorption of CO (m/z = 28) and H2
(m/z = 2) more quickly than Pt–Fe/M. After reaching the sat-
urated adsorption, O2 (m/z = 32) pulses are injected in the He
carrier stream. Neither CO2 (m/z = 44) nor H2O (m/z = 18)
is formed with no consumption of injected O2 at Pt/M; how-
ever, CO2 is immediately formed at Pt–Fe/M by the oxidation
of adsorbed CO with pulse-injected O2. Nevertheless no H2O
is detected. The results of CO/H2-TPD in Fig. 5 show the pres-
ence of two kinds of H2. The weakly adsorbed H2, desorbed
completely up to 150 ◦C, may react with O2 and formed H2O.
The H2O is presumably trapped in the mordenite cages and can-
not be detected in the present experiment shown in Fig. 7b. We
may consider that using strongly adsorbed H2, starting the des-
orption at >200 ◦C, or continuing the desorption at even 400 ◦C
has probably less reactivity to adsorbed O2 than that of CO,
which desorbs completely up to 300 ◦C. This reasoning can be
supported by the result in Figs. 1 and 2 indicating that almost all
O2 was used for the oxidation of CO at 50 ◦C. The CO adsorbed
on Pt sites reacts with O2 adsorbed on Fe sites, forming CO2. In
the PROX reaction under the steady-state condition, where re-
actant gases are supplied continuously, such Pt sites as weakly
adsorbed H2 may be occupied with CO preferentially, resulting
in the preferential oxidation of CO by the O2 adsorbed con-
tinuously on the neighboring Fe sites. This is the reason why
Pt–Fe/M exhibits superior selectivity at such a low operating
temperature, despite the presence of a large excess of H2 com-
pared with CO as well as the added O2 in reforming gases.
by our XANES analysis [18]. As the result, no PROX activity
appeared (as seen in Fig. 1), because of the lack of essential
requisites for the Langmuir–Hinshelwood mechanism. At Pt–
Fe/M, the foregoing results and discussion clearly indicate that
Pt sites are available for the adsorption of CO as well as H2
,
and that Fe sites act as O2 adsorption sites. CO adsorbed on a
Pt site and O adsorbed on an Fe site react immediately even at
50 ◦C once both reactants sit on such neighboring sites. This
is the mechanism that Watanabe and Motoo proposed as the
so-called “bifunctional mechanism” for the electrocatalytic ox-
idation of “C-containing reactants” on binary alloys such as
Pt–Ru [24,25]. A PROX reaction on Au catalysts supported
on TiO2, Al2O3, and ZrO2, an oxidation of CO by H–O–O
species, generated via the reaction between dissociatively ad-
sorbed H2 on the gold particles and O2 molecule from the gas
phase, was suggested [26]. In this mechanism, both CO2 and
H2 generation must be observed as the result of the reaction
between CO and H–O–O species. But when an O2 pulse was
introduced to the Pt–Fe/M catalyst saturated with CO and H2,
CO2 was formed but H2 was not detected, as seen Fig. 7b. This
experimental result strongly suggests that the PROX reaction
on Pt–Fe/M catalyst proceeds not via the pathway proposed for
Au-supported catalyst, but via the bifunctional mechanism that
we have proposed. Perfect dispersion (or neighborhood) of two
types of sites may not be required for the mechanism as long as
they are close enough for the surface diffusions of both types
to adsorbed reactants. In addition to the components for the bi-
functions, the special reaction spaces of zeolite cages are also
essential to achieving the desired reactivity and selectivity, as
indicated by a performance different than that of the conven-
tional PROX catalysts. A reason why H2 on Pt sites of Pt/M or
Pt–Fe/M is less active than CO at the PROX reaction has not yet
been clearly explained, but it does happen and leads to superior
selectivity.
3.2.4. Mechanisms of the PROX reaction
Based on the foregoing observations and discussions, we can
summarize schematically the adsorption of CO, O2, and/or H2
and the reaction processes at the different catalysts, as shown
in Fig. 8. As seen in Figs. 6 and 7, no PROX reaction occurs
because the reaction sites are fully covered with CO and/or
H2 strongly adsorbed on Pt/M due to blocked access of O2
to the reaction sites. This strongly indicates that a dissocia-
tive adsorption of O2 and the following surface reaction with
preadsorbed CO are essential for the PROX reaction, that is,
a Langmuir–Hinshelwood mechanism. The reaction can occur
only at Pt/M as the reaction temperature is elevated (>120 ◦C),
maybe because both the CO coverage and the bonding strength
are lowered and the resulting CO-free sites are available for the
dissociative adsorption of O2 (see Figs. 1 and 4). But Fe/M
has no affinity for CO adsorption, as seen in Fig. 4, presum-
ably because Fe sites are dominantly in FeO phase as indicated
4. Conclusion
The PROX activity of Pt/M, Fe/M, and Pt–Fe/M was ex-
amined. Pt/M and Fe/M did not show any activity at 50 ◦C.
However, Pt–Fe/M showed distinctive PROX reactivity and se-
lectivity at the same temperature—for example, it achieved the
ideal performance of 100% at even λ (2 × O2/CO) = 1.0 and
SV = 12,500 h−1, with no noticeable decline in selectivity even
at SV = 100,000 h−1. CO, H2, and O2 chemisorption measure-
ments demonstrated that Fe/M cannot adsorb CO and H2, but
can absorb O2, whereas both Pt/M and Pt–Fe/M show a high
affinity to both CO and H2 adsorption. It was also found that
adding Fe to Pt/M can preserve O2 adsorption sites for the
PROX reaction even in CO/excess H2 mixed gas. Regarding
the PROX mechanism at low temperatures on Pt/M, Fe/M, and
Pt–Fe/M, the poor reactivity of Pt/M and Fe/M can be ascribed
to the lack of CO and/or O2 adsorption as the essential requi-
site for Langmuir–Hinshelwood mechanism. We proposed the
so-called “bifunctional mechanism” for the distinctive perfor-
mance at Pt–Fe/M, where Pt site acts as CO adsorption site and
Fe site acts as an O2 dissociative-adsorption site and enhances
Fig. 8. PROX mechanism, schematically shown, at (a) Pt/M, (b) Fe/M and
(c) Pt–Fe/M catalysts.

M. Kotobuki et al. / Journal of Catalysis 236 (2005) 262–269
269
the surface reaction between the reactants on the neighboring
sites.
[11] P.V. Snytnikov, V.A. Sobyanin, V.D. Belyaev, P.G. Tsyrulnikov, N.B. Shi-
tova, D.A. Shlyapin, Appl. Catal. A 239 (2003) 149.
[12] M. Watanabe, H. Uchida, H. Igarashi, M. Suzuki, Chem. Lett. (1995) 21;
JP Patent 07256112;
Acknowledgments
US Patent 6168772.
[13] H. Igarashi, H. Uchida, M. Watanabe, Chem. Lett. (2000) 1262.
[14] H. Igarashi, H. Uchida, M. Watanabe, Stud. Surf. Catal. 132 (2001) 953.
[15] M. Watanabe, H. Uchida, K. Ohkubo, H. Igarashi, Appl. Catal. B 46
(2003) 595.
[16] M. Kotobuki, A. Watanabe, H. Uchida, H. Yamashita, M. Watanabe,
Chem. Lett. (2005) 866.
This work was supported by Leading Project of the Adminis-
tration of Education, Science, Culture and Sports on “Research
and Development of Materials for the Next Generation Fuel
Cells.”
[17] A. Wootsch, C. Descorme, D. Duprez, J. Catal. 225 (2004) 259.
[18] M. Kotobuki, T. Shido, M. Tada, H. Uchida, H. Yamashita, Y. Iwasawa,
M. Watanabe, Catal. Lett. 103 (2005) 263.
References
[19] M. Watanabe, Y. Zhu, H. Igarashi, H. Uchida, Electrochemistry 68 (2000)
244.
[20] H. Igarashi, T. Fujino, Y. Zhu, H. Uchida, M. Watanabe, Phys. Chem.
Chem. Phys. 3 (2001) 306.
[1] L. Carrette, U. Stimming, Fuel Cells 1 (2001) 5.
[2] R.A. Lemons, J. Power Sources 29 (1990) 251.
[3] B. Hohlein, S.V. Andrian, Th. Grude, R. Menze, J. Power Sources 86
(2000) 243.
[21] T. Toda, H. Igarashi, H. Uchida, M. Watanabe, J. Electrochem. Soc. 146
(1999) 3750.
[22] K. Foger, J.R. Anderson, J. Catal. 54 (1978) 318.
[23] T. Szilagyi, J. Catal. 121 (1990) 223.
[24] M. Watanabe, S. Motoo, J. Electroanal. Chem. 60 (1975) 275.
[25] T. Yajima, H. Uchida, M. Watanabe, J. Phys. Chem. B 108 (2004)
2654.
[4] M.J. Kahlich, H.A. Gasteiger, R.J. Behm, J. Catal. 171 (1997) 93.
[5] R.M.T. Sanchez, A. Ueda, K. Tanaka, M. Haruta, J. Catal. 168 (1997) 125.
[6] K. Sekizawa, S. Yano, K. Eguchi, H. Arai, Appl. Catal. A 169 (1998) 291.
[7] G.K. Bethke, H.H. Kung, Appl. Catal. A 194/195 (2000) 43.
[8] M. Echigo, T. Tabata, Appl. Catal. A 251 (2003) 157.
[9] H. Tanaka, S. Ito, S. Kameoka, K. Tomishige, K. Kunimori, Catal. Com-
mun. 4 (2003) 1.
[26] C. Rossignol, S. Arrii, F. Morfin, L. Piccolo, V. Caps, J. Rousset,
J. Catal. 230 (2005) 476.
[10] M. Echigo, N. Shinke, S. Takami, T. Tabata, J. Power Sources 132 (2004)
29.