Bifunctional catalysts for single-stage water-gas shift reaction in fuel cell applications. Part 1. Effect of the support on the reaction sequence

Article abstract of DOI:10.1016/j.jcat.2007.07.010

Oxide support plays a significant role in the mechanistic reaction sequence for the water-gas shift (WGS) reaction over Pt-based catalysts. In situ FTIR spectroscopic and transient kinetic studies have been used to follow the reactions that occur. CeO₂-, TiO₂-, and ZrO₂-supported Pt catalysts have been studied at 300 °C. In all cases, CO is adsorbed on Pt. The role of the support oxide is to activate water, completing the WGS reaction sequence. We have taken into consideration four different pathways that may be involved in the complex WGS reaction scheme: (A) red-ox route, (B) associative formate route, (C) associative formate route with red-ox regeneration of the oxide support, and (D) carbonate route. In the case of Pt/ZrO₂, the WGS reaction follows the associative formate route with red-ox regeneration (route C). On Pt/TiO₂, both the red-ox route (A) and the associative formate route with red-ox regeneration (C) contribute. The associative formate route (B) is the relevant reaction pathway on Pt/CeO₂.

Full text of DOI:10.1016/j.jcat.2007.07.010

Journal of Catalysis 251 (2007) 153–162
www.elsevier.com/locate/jcat
Bifunctional catalysts for single-stage water–gas shift reaction
in fuel cell applications.
Part 1. Effect of the support on the reaction sequence
∗
K.G. Azzam, I.V. Babich, K. Seshan, L. Lefferts
Catalytic Processes and Materials, Faculty of Science and Technology, IMPACT, University of Twente, P.O. Box 217, 7500AE Enschede, The Netherlands
Received 29 March 2007; revised 3 July 2007; accepted 8 July 2007
Available online 31 August 2007
Abstract
Oxide support plays a significant role in the mechanistic reaction sequence for the water–gas shift (WGS) reaction over Pt-based catalysts. In
situ FTIR spectroscopic and transient kinetic studies have been used to follow the reactions that occur. CeO -, TiO -, and ZrO -supported Pt
2
2
2
◦
catalysts have been studied at 300 C. In all cases, CO is adsorbed on Pt. The role of the support oxide is to activate water, completing the WGS
reaction sequence. We have taken into consideration four different pathways that may be involved in the complex WGS reaction scheme: (A) red–
ox route, (B) associative formate route, (C) associative formate route with red–ox regeneration of the oxide support, and (D) carbonate route. In
the case of Pt/ZrO , the WGS reaction follows the associative formate route with red–ox regeneration (route C). On Pt/TiO , both the red–ox
2
2
route (A) and the associative formate route with red–ox regeneration (C) contribute. The associative formate route (B) is the relevant reaction
pathway on Pt/CeO .
2
©
2007 Elsevier Inc. All rights reserved.
Keywords: Water–gas shift; Platinum; Ceria; Titania; Zirconia; Reaction mechanism; Red–ox; Associative formate mechanism; Carbonate
1
. Introduction
the reaction at higher temperatures and to achieve almost com-
plete CO conversion (unconverted CO <1000 ppm). The two
◦
The water–gas shift (WGS) reaction is a reversible, exother-
steps involve a high-temperature (350–500 C) shift (HTS) over
mic reaction in which carbon monoxide reacts with steam to
produce hydrogen and carbon dioxide:
Fe O /Cr O catalyst, followed by a low-temperature (200–
2
3
2
3
◦
250 C) shift (LTS) over a Cu/ZnO/Al O catalyst [1]. This
2
3
approach, which is currently practiced at industrial scales, is
not an appropriate choice for mobile applications because of
its technical complexity and the multiple stages involved [2].
A single-stage WGS conversion is thus desirable. Supported
noble metal-based (e.g., Pt) catalysts are reported as promising
single-stage WGS catalysts because they (i) are robust, (ii) can
operate at higher temperatures where the kinetics is more favor-
able in contrast to Cu catalysts, (iii) are less sensitive to poisons
(Cl and S) than the LTS (Cu-based) catalysts, and (iv) are more
active than the HTS (Fe/Cr oxide-based) catalysts [2,3].
CO + H2O ꢀ CO2 + H2, ꢀH = −40.6 kJ/mol.
This reaction is a crucial step in maximizing hydrogen after
the conversion of hydrocarbons to synthesis gas via steam re-
forming or partial oxidation [1]. The demand for hydrogen is
expected to increase in the future due to its use in fuel cell
applications for power generation. This is because fuel cells
are appreciably more efficient than current internal combustion
engines and are environmentally friendly, producing no COx,
NOx, hydrocarbons, or soot emissions [2,3].
The state-of-the-art industrial process for WGS is carried out
in two stages to overcome the thermodynamic limitation for
Recently, various catalyst compositions have been explored
in WGS by varying Pt loading (from 0.1 to 5 wt%) and the
oxide supports (e.g., CeO [4–6], ZrO [7,8], TiO [9–11],
2
2
2
*
x 2 x 2
Ce Zr O [12–15], Ti Ce O [16]). Activity is reported
to depend on a number of factors, including catalyst prepara-
Corresponding author. Fax: +31 53 489 4683.
E-mail address: l.lefferts@tnw.utwente.nl (L. Lefferts).
1−x
1−x
0021-9517/$ – see front matter © 2007 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcat.2007.07.010

154
K.G. Azzam et al. / Journal of Catalysis 251 (2007) 153–162
Fig. 1. The role of the support in WGS reaction sequence.
tion [4,5,10,17], the nature of the support [16], testing con-
ditions, and reactor design [18]. Despite the fact that noble
metal-based catalysts are very active, their straightforward ap-
plication to provide hydrogen feed for PEM fuel cells is not
feasible because of limited conversion of CO due to thermody-
namic constraints. This can be overcome by selective product
separation to shift the equilibrium to the product side, by, for
example, using a hydrogen-selective permeable membrane [19,
mate) [5,6,13,31,32] and (ii) regenerative red–ox mechanism
[26,28,33,34].
In the associative mechanism [5,6,13,31,32], hydroxyl
groups on the support react with CO and form surface for-
mates. The decomposition of formate species is suggested to be
facilitated by the presence of water in the gas phase and is rate-
determining. During this step, hydroxyl groups are regenerated
on the oxide surface. No oxygen is removed from the surface
during the catalytic cycle, and the oxide does not undergo any
red–ox changes during the WGS sequence. This mechanism
was supported by FTIR measurements, isotope-exchange ex-
periments, pulse-type experiments and kinetic studies of CO2
hydrogenation (reverse WGS) [5,6,13,31,32,35]. The role of
Pt is not clear from these studies and cannot be elucidated from
the reaction schemes proposed. However, it has been suggested,
without clear experimental evidence, that Pt aids in the genera-
tion of bridging OH groups, which are claimed to be necessary
for the formation of surface formate groups. Platinum also fa-
cilitates decomposition of formate groups in the presence of
H2O [5].
2
0] or CO sorbents [21]. Removal of CO by preferential oxi-
2
dation or methanation are other options to further lower the CO
concentration [22,23]. The experiments in this study were car-
◦
ried out at 300 C, inspired by the fact that hydrogen-selective
membranes are stable at those conditions.
In general, it is agreed that both the metal and the support
play essential roles in the WGS reaction. Among the two reac-
tants for WGS reaction, CO and H2O, the latter is more difficult
to activate [24] due to its thermodynamic stability. Metals such
as Cu and Fe are reported to undergo oxidation by water [25],
thereby activating it. However, Pt does not interact chemically
with water because the PtOx that would be formed is not ther-
modynamically stable [26] at WGS temperatures. Thus, for Pt-
based WGS catalysts, a hydrophilic oxide support is essential
to adsorb and activate water [24–30]. Therefore, such catalysts
are bifunctional, with platinum activating CO and the support
activating H2O.
Different WGS reaction pathways have been suggested in
the literature [4,27]; these are schematically presented in Fig. 1.
Water either reoxidizes the support oxide after reduction by CO,
activated on Pt, or forms hydroxyl groups and completes the
catalytic cycle. The exact nature of the relevant intermediates
formed and the mechanistic route(s) for the WGS reaction are
In the case of the red–ox mechanism, CO adsorbs on the
metal and reacts at the metal–support interface with oxygen
from the support surface, forming CO and decreasing the sup-
2
port surface. The latter is then reoxidized by H2O, forming H2.
This mechanism was supported by kinetic studies, TPD, TPR
and oxygen storage capacity measurements [26,28,33,34].
In our efforts to develop single-stage Pt-based WGS cata-
lysts for fuel cell applications, we found a strong influence of
the support on catalyst activity [16]. Here we report on the
influence of the oxide support (ZrO , CeO , or TiO ) of Pt-
2
2
2
based catalysts on the mechanistic reaction sequence(s) for the
WGS reaction. We performed kinetic pulse transient studies
using CO, H2O, and N2O and in situ FTIR spectroscopic mea-
surements to elucidate elementary reaction steps and identify
still a matter of debate, however. In the case of Pt/CeO , the
2
most widely studied catalytic system for WGS, two reaction
mechanisms have been proposed: (i) associative (surface for-

K.G. Azzam et al. / Journal of Catalysis 251 (2007) 153–162
155
reaction intermediates during WGS reaction sequence. We used
N2O to reoxidize the oxide surface without regeneration of OH
groups to study the role of surface OH groups in the WGS re-
action. In this way, we avoid high-temperature dehydroxylation
of the catalyst surface, preventing any structural changes (e.g.,
decreased Pt dispersion) in the catalyst. In Part 2 [16], we re-
port on the use of these concepts to develop an active, selective,
and stable single-stage WGS catalyst.
in 10 vol% H2/He, 30 mL/min flow for 1 h. After this the cat-
alyst was heated at 330 C in He (30 mL/min) for 30 min.
◦
Pulse experiments were then carried out by contacting fixed
◦
amounts of reactant gases with catalyst at 300 C using He as
carrier gas. Each pulse contained 3.9 µmol of respective gas
(CO and N2O). H2O (1.0 µL) was injected using a microsy-
ringe directly into the reactor. Pulses were repeated until the
responses for reactants and products no longer changed. To
determine the products formed, the outlet of the reactor was
◦
2
. Experimental
directly connected to a Porapak column (5 m, 100 C) and a
thermal conductivity detector (TCD). Downstream to the TCD
detector, gases were analyzed by an online mass spectrometer
2
.1. Catalyst preparation
(
Balzers QMS 200 F). Quantitative determination of all gaseous
products, except of H2 and H2O, was obtained from TCD data
with accuracy not below ± 0.1 µmol/gcat. H2 was detected by
MS; the data for H2 are semiquantitative for each catalyst, and
quantitative comparison between catalysts is not applicable.
The following commercial supports were used: TiO (De-
gussa, P-25), CeO2 (Aldrich), and ZrO2 (Daiichi Kagaku Ko-
gyo, RC100). The catalysts were prepared by wet impregna-
2
tion of the solid supports with aqueous solutions of H PtCl6
2
(
Aldrich) to yield catalysts with 0.5 wt% Pt. The catalysts were
◦
2
.4. FTIR studies
dried at 75 C for 2 h in vacuum and subsequently calcined at
4
◦
50 C for 4 h.
The FTIR spectra were recorded using a Bruker Vector 22
with MCT detector under flow conditions [5% CO/He, 10%
2
.2. Characterization
H2O/He, and (5% CO + 10% H2O)/He] at the same temper-
◦
ature used in the kinetic experiments (300 C). Pulse exper-
Platinum content of the catalysts was determined using
iments were mimicked by fast switching between He and a
mixture of He containing CO or H2O.
Philips X-ray fluorescence spectrometer (PW 1480). The BET
surface area of the supports and catalysts were measured us-
ing the BET method on Micromeritics ASAP 2400. Pt disper-
3
. Results
sions were measured by H chemisorptions at room tempera-
2
ture using a Micromeritics Chemisorb 2750. Teschner et al. [36]
showed in a detailed study that hydrogen spillover does not oc-
A typical experimental result during subsequent pulsing of
2 2
CO and H O over a catalyst (e.g., Pt/CeO ) at 300 C is pre-
◦
◦
cur significantly on Pt/CeO at lower temperatures (i.e., 20 C),
2
sented in Fig. 2. Cumulative consumption and formation of
reactants and products, respectively, were quantified; the data
are shown in Table 2. We started the pulse sequence experi-
ments with hydroxylated catalyst surfaces.
resulting in reliable dispersion measurements. Therefore, H2
spillover from Pt particles to the support surface is not influ-
encing dispersion measurements in the case of ceria-supported
catalysts. The properties of the supports and the catalysts are
summarized in Table 1.
3
.1. Pt/CeO2
2
.3. Pulse experiments
The first CO pulse resulted in partial CO consumption and
production of CO2. In the subsequent CO pulses, the amounts
of CO2 formed decreased. After four CO pulses, only traces
of CO2 were observed, and the amount of CO detected in the
outlet of the reactor was close to the amount of CO in the pulse
applied (± 0.1 µmol/gcat). No H2 formation was observed. The
cumulative amount of CO2 formed (68 µmol/g, Table 2) was
much lower than the amount of CO consumed (158 µmol/g)
during pulsing. FTIR spectra of Pt/CeO2 after exposure to CO
at the same temperature (compare Figs. 3 and 4) showed a band
Kinetic transient pulse experiments were performed at
00 C, at atmospheric pressure using a fixed-bed reactor.
A 50-mg catalyst sample was placed between two quartz plugs
◦
3
in a quartz tubular reactor (d = 4 mm). For all of the experi-
ments described below, the gases (He, H , CO, and N O) used
2
2
◦
were of >99.9% purity. The catalyst was first reduced at 300 C
Table 1
Supports and catalysts properties
−
1
at 2050–2060 cm typical for linearly adsorbed CO on Pt [37],
Sample
Calcination
time, T
XRF Pt Pt dispersion
Surface Pt
(µmol/g)
BET
(m /g)
−
1
a band at 2856 cm corresponding to the C–H stretching of
◦
2
(wt%)
(%) at 25 C
−1
formate [31], and bands in the spectra interval 1600–1300 cm
◦
CeO
500 C, 4 h
–
–
–
0.54
0.49
0.47
–
–
–
65
55
60
–
–
–
16.6
14.1
15.4
85
50
25
81
48
22
2
◦
corresponding to carbonate and/or formate on ceria [13,31,37].
These species were relatively stable and could not be removed
TiO
500 C, 4 h
2
◦
ZrO
500 C, 4h
2
a
◦
◦
Pt/CeO
Pt/TiO
Pt/ZrO
450 C, 4 h
by flushing with He at 300 C.
2
◦
450 C, 4 h
2
After CO, the subsequent H O pulses resulted in simulta-
2
◦
450 C, 4 h
2
neous production of CO2 and H2 (Fig. 2, Table 2). Most of
the CO and H was formed during the first two pulses. The
2 2
a
◦
H chemisorption at 0 C gave 61% Pt dispersion.
2

156
K.G. Azzam et al. / Journal of Catalysis 251 (2007) 153–162
◦
Fig. 2. Typical MS response for pulse experiment over Pt/CeO catalyst reduced at 300 C in H (10%)/He flow. Four pulses CO are followed with two pulses H O.
2
2
2
Difference in time between products detected is due to their separation with GC (Porapak column).
◦
◦
Fig. 4. FTIR spectra at 300 C of Pt/CeO catalyst; reduced in 10 vol% H /He
2 2
Fig. 3. FTIR spectra at 300 C of Pt/CeO catalyst; reduced in 10 vol% H /He
2
2
flow (a); after CO pulses (b); after H O pulses (c); in situ WGS conditions (d).
flow (a); after CO pulses (b); after H O pulses (c).
2
2
Table 2
Analysis of gaseous products during pulse experiments over Pt based catalysts
Catalyst
(I) CO pulse
(II) H O pulse after CO
(III) CO pulse after CO–H O
2
2
CO consumed
CO released
2
C uptake
H pro- CO released
H pro- CO consumed
2
CO released
2
C uptake
H pro-
2
2
2
(µmol/g)
(µmol/g)
(µmol/g) duction
(µmol/g)
duction
(µmol/g)
(µmol/g)
(µmol/g) duction
I
II
Pt/CeO
2
158
52
40
68
51
40
90
1
0
–
60
0.6
No
+
+
–
118
49
37
57
48
34
60
1
3
–
Pt/TiO
+
+
2
III Pt/ZrO
+
+
2
Catalyst
(IV) N O pulse after CO
(V) CO pulse after CO–N O
(VI) H O pulse after CO–N O–CO
2 2
2
2
CO released
H pro-
2
CO consumed
CO released
2
C uptake
H pro-
2
CO released
2
H pro-
2
2
(µmol/g)
duction
(µmol/g)
(µmol/g)
(µmol/g)
duction
(µmol/g)
duction
Ia
IIa
IIIa
Pt/CeO
2
60
0.6
No
+
220
52
32
135
50
31
85
2
1
–
–
–
56
0.8
0
–
Pt/TiO
–
+
2
Pt/ZrO
–
–
2

K.G. Azzam et al. / Journal of Catalysis 251 (2007) 153–162
157
◦
Fig. 5. Typical MS response for pulse experiment over Pt/TiO catalyst reduced at 300 C in H (10%)/He flow. 4 pulses CO are followed with two pulses H O.
2
2
2
Difference in time between products detected is due to their separation with GC (Porapak column).
next H2O pulses produced only traces of CO2 and H2. Only
part of the formate/carbonate groups formed and retained at
the surface during CO pulses were reactive to H2O. After con-
−
1
tacting with H2O, the FTIR band at 2856 cm , which cor-
responded to formate (Fig. 4), disappeared, and the intensity
−
1
of the bands at 1600–1300 cm (Fig. 3), corresponding to
formate/carbonate, decreased. However, an appreciable amount
of oxygenate species remained on the surface. In quantitative
terms, only 60 µmol/g CO2 was released from the surface dur-
ing H2O pulses, and 30 µmol/g “C” was retained at the surface.
The next CO–H2O sequence of pulses showed the same product
distribution as in the first sequence.
When N O was used instead of H O to remove oxygenate
2
2
species from Pt/CeO2, the following gaseous products were
observed: N2 due to decomposition of N2O (not shown) and
CO2 and H2 due to decomposition of surface formate/carbonate
formed during CO pulsing. No O2 was observed. CO pulses
after N2O treatment (Table 2) showed products (CO2, no H2)
similar to those during CO pulses over Pt/CeO2, although some-
what more CO was converted to CO2. At the same time, the
amount of carbon retained on the surface was comparable
◦
Fig. 6. FTIR spectra at 300 C of Pt/TiO catalyst; reduced in H (10%)/He flow
2
2
(a); after CO pulses (b); after H O pulses (c); in situ WGS conditions (d).
2
(
85 µmol/g). Pulsing with water after the sequence CO–N2O–
CO resulted in the formation of essentially only CO2. No H2
was detected by MS.
(
i.e., 51 µmol/g). Thus, unlike in the case of Pt/CeO2, no oxy-
genate species were retained on the Pt/TiO2 surface. FTIR data
Fig. 6) also showed no formation of carbonate/formate surface
species during the CO pulsing. Only CO adsorbed on Pt (Fig. 6)
(
3
.2. Pt/TiO2
−1
was detectable in the FTIR spectra (2060–2080 cm ) [9,11]
after flushing with He at 300 C.
In contrast to Pt/CeO2, contacting Pt/TiO2 with CO resulted
◦
simultaneously in the formation of CO2 and H2 (Fig. 5, Ta-
ble 2). With repetitive CO pulses, the amount of CO consumed
decreased, in line with decreasing H2 and CO2 formation. After
four pulses, the amount of CO at the reactor outlet was similar
The H2O pulses after CO treatment resulted in H2 as the
main product (Fig. 5); CO2 was observed in only negligible
amounts (0.6 µmol/g). Subsequent CO pulses produced the
same gaseous products (CO2 and H2) as in the case of the first
CO pulse sequence, demonstrating that the surface was regener-
ated by H2O and the catalytic cycle was completed. The amount
to that in the feed, and only traces of CO and H were detected.
2
2
The cumulative CO consumption over Pt/TiO2 was 52 µmol/g.
This value was in agreement with the amount of CO2 formed

158
K.G. Azzam et al. / Journal of Catalysis 251 (2007) 153–162
◦
Fig. 7. Typical MS response for pulse experiment over Pt/ZrO catalyst reduced at 300 C in H (10%)/He flow. Four pulses CO are followed with two pulses H O.
2
2
2
Difference in time between products detected due to their separation with GC (Porapak column).
of CO2 formed during CO pulses over surface treated with H2O
was 48 µmol/g (Table 2).
When the Pt/TiO2 was exposed to N2O pulses after subject-
ing it to CO pulses (Table 2), no H2 was observed. N2 was
observed, but no O2 could be detected, indicating oxygen up-
take by the support. The amount of N2 (not shown) was close
to the amount of CO2 formed during the subsequent CO pulses
(
52 µmol/g). Subsequent CO pulses (after N2O) resulted in
CO2 production (50 µmol/g); no H2 was detected. In case of
the CO–N2O–CO–H2O sequence, the last H2O pulse produced
the same amount of H2 (semiquantitatively) as was produced
during the H2O pulse after direct CO exposure.
3
.3. Pt/ZrO2
For the Pt/ZrO2 catalyst (Fig. 7), the trend was similar to that
observed in the case of Pt/TiO2. CO pulses resulted in release
of H2 and CO2. The amount of CO consumed was in agree-
ment with the amount of CO2 released (40 µmol/g), closing
the carbon balance. In agreement with this, no formation of sta-
ble carbonate/formate group was observed from FTIR (Fig. 8)
for Pt/ZrO2 after CO pulses. Only the band at around 2050–
◦
Fig. 8. FTIR spectra at 300 C of Pt/ZrO2 catalyst; reduced in H2(10%)/He
flow (a); after CO pulses (b); after H O pulses (c); in situ WGS condition (d).
2
4
. Discussion
−1
2
080 cm , related to CO adsorbed on Pt [7], was present.
Surprisingly, in the case of Pt/ZrO2, H2O pulses after CO
exposure (Table 2) did not produce any gaseous products. How-
ever, pulsing H2O seemed to have regenerated the catalyst,
because the subsequent CO pulses resulted in the formation
of CO2 and H2, similar to the initial CO pulses. In contrast,
treatment with N2O reactivated Pt/ZrO2 only partially. During
the subsequent CO pulses, consumption of CO and production
of only CO2 was observed, and only traces of H2 were de-
tected.
Kinetic pulse studies were used to follow the elementary
steps occurring during the WGS reaction. The results of pulse
transient experiments are not directly equivalent to data ob-
tained from steady-state experiments; nonetheless, pulse exper-
iments allow us to rule out those reaction pathways that cannot
be induced by sequentially pulsing CO and H2O. Furthermore,
the conclusions obviously hold only for the conditions chosen
in this study, because it has been reported elsewhere [38,39] that

K.G. Azzam et al. / Journal of Catalysis 251 (2007) 153–162
159
Fig. 9. Role of oxide support on the reaction pathways for WGS. In all cases CO is adsorbed on Pt and reacts at Pt–support interface: A—classical re–dox route;
B—associative route; C—associative mechanism with oxide regeneration via re–dox; and D—formation, regeneration of carbonate on oxide.
the dominant reaction mechanism depends on the experimental
parameters, such as reaction temperature and gas composition.
As outlined in the Introduction, two mechanistic routes have
been reported for the WGS reaction sequence over Pt/CeO2
catalysts: an associative formate mechanism and a regenera-
tive red–ox mechanism. In our view, there are other possible
routes to the WGS reaction over oxide-supported Pt catalysts.
Four possible sequences are outlined in Fig. 9: (A) the red–ox
route, (B) the associative formate route, (C) the associative for-
mate route with red–ox regeneration of the oxide support, and
Table 3
Gaseous products expected during sequential CO and H O pulses for each re-
action pathway proposed in Fig. 9
2
Reaction pathway
Expected gaseous products
CO pulse
H2O pulse
Classical red–ox
Associative formate
Associative formate with red–ox regeneration
Carbonate
CO
2
None
H
2
CO , H
2
2
CO , H
None
CO
2
2
2
None
(
D) the carbonate route. In all four sequences, we propose (in
only regenerate hydroxyl-groups at the support surface without
releasing any gaseous products. In this case (route C), CO2 is
formed, taking oxygen from the oxide support surface, whereas
in the classical associative formate mechanism (route B), oxy-
gen for CO2 is supplied by H2O.
agreement with the literature [24–30]) that CO is adsorbed on
Pt and reacts at the Pt–support periphery as in a typical bifunc-
tional catalyst. The support oxide plays a role in the activation
of water and closure of the WGS catalytic cycle.
The feasible reaction route(s) for each support can be elu-
cidated from kinetic pulse experiments based on the gaseous
products formed during sequential CO and H2O pulses (Ta-
ble 3). Thus, in the red–ox mechanism, formation of only CO2
or H2 should be observed during CO and H2O pulses, respec-
tively. For the associative formate route, CO pulsing should
result not in any gaseous products, but rather in stable interme-
diates (i.e. surface formate groups [5,6,31,32]), which are de-
composed during H2O pulsing, simultaneously producing CO2
and H2. The associative formate route with red–ox regeneration
as proposed in Fig. 9 implies that intermediate formate species
are not stable and decompose immediately, producing both CO2
and H2 during CO pulsing. The subsequent H2O pulse should
We now discuss the results obtained with Pt catalysts over
different supports in view of the various proposed routes. In
the case of Pt/CeO2, the results indicate that the WGS reaction
follows an associative formate mechanism, in agreement with
earlier reported data [4–6,13,31,32]. Evidence for this comes
from the observation in FTIR spectra (Figs. 3 and 4) of formates
on Pt/CeO2 after pulsing CO. This is further supported by the
simultaneous formation of H2 and CO2 during H2O pulses (Ta-
ble 2, row I). Furthermore, the carbon uptake (90 µmol/g) was
much higher than would be expected from exclusively CO ad-
sorption on Pt (16 µmol/g if the ratio of CO to surface Pt is 1:1,
which is typical for CO adsorption over Pt at room tempera-
◦
ture [40]). Because pulsing was done at 300 C, we should have

160
K.G. Azzam et al. / Journal of Catalysis 251 (2007) 153–162
expected a much lower CO coverage of Pt, because it is well
known that CO desorption from Pt commences at 200 C [11].
The subsequent CO pulse (Table 2, row Ia) did not produce any
hydrogen, as expected. Part of the CO was oxidized and re-
leased to the gas phase as CO2. The remaining CO (85 µmol/g)
was retained on ceria. The carbon balance was quite good; the
amount (85 µmol/g) was remarkably close to the amount of
carbon retained (90 µmol/g, as both carbonate and formate in
total) in row A after the CO pulse. Because no hydroxyl groups
were available (because regeneration was by N2O, not H2O),
creation of the formate species on the ceria surface could be
ruled out. Only carbonates were present. Thus, in the absence
of formates, some sites were occupied by carbonates, imply-
ing that under WGS conditions, carbonates and formates were
competing for surface sites on ceria.
The fact that only carbonates were present and formates
were absent is further highlighted by the absence of hydrogen
formation during the subsequent H2O pulse (row Ia). If for-
mates had been present, then we would have seen hydrogen
when pulsing water (row I). Only partial decomposition of the
carbonates (∼65%) to CO2 was observed (Table 2, row Ia). This
observation unambiguously proves that an associative formate
mechanism (reaction (1), route B in Fig. 9) was contributing
◦
Therefore, most of the C uptake must be related to the formation
of formate/carbonate surface species, in agreement with FTIR
observations (Figs. 3 and 4). The amount of carbon remaining
on the ceria surface (90 µmol/g) was much smaller than the
1
8
2
calculated monolayer capacity (6 × 10 molecules/m [41])
for the CeO2 support for carbonate and/or formate groups (i.e.,
8
10 µmol/g). It is very difficult to assign the bands observed in
Fig. 3 to either formate or carbonate; despite the many signif-
icant attempts reported in the literature [5,14,31,37], the bands
are very broad and too strongly overlapping to allow definite
−
1
assignments. However, the band at 2856 cm (C–H stretch-
ing of formate [31]) shows the presence of formate species on
the ceria surface during CO pulses. In this paper, we do not go
into the details of differentiating between formate and carbon-
ate bands in the IR spectra (Fig. 3); for the moment, we can only
conclude that both oxygenate species were formed during CO
pulsing, they were stable, and their formation can be described
by the following equations:
Pt–CO + OH(s) → (COOH)(s)
and
(1)
significantly to H formation. Alternatively, formation of CO
2
2
during CO pulses also can result from the reduction of the ceria
surface according to the sequence
Pt–CO + 2O(s) → (CO3)(s),
(2)
Pt–CO + O(s) → CO2(g) + *(s) + Pt,
(6)
where (s) denotes the support site. We conclude that formation
of CO2 and H2 during the subsequent H2O pulse was due to the
decomposition of formate,
where *(s) denotes an oxygen vacancy on the support. In prin-
ciple, oxidation of these sites by H2O can proceed in two ways:
*
(s) + H2O → O(s) + H2(g)
(7)
(
COOH)(s) + H2O(g) → CO2(g) + H2(g) + OH(s),
(3)
and
because formate is known to be unstable on ceria in the pres-
ence of H2O [5,31,42]. This is also substantiated by the disap-
*(s) + O(s) + H2O → 2OH(s).
(8)
−
1
pearance of the formate band at 2856 cm in the FTIR results
Fig. 4).
In agreement with the literature [43], carbonates were re-
Reaction (7), which completes a typical red–ox cycle with re-
action (6), can be neglected based on the results from the N2O–
CO–H2O pulse sequence (row Ia), because no H2 was observed,
as would be expected. Formation of hydroxyl groups via reac-
tion (8) is possible.
(
tained on ceria at this temperature even in the presence of water.
However, we do not at this point exclude decomposition of part
of the surface carbonates during H2O pulsing,
There is only a subtle difference between reactions (3)
and (8). In reaction (3), the oxygen of the OH group comes
from the water molecule, and oxygen from ceria is not involved.
This is a classical associative mechanism (route B, Fig. 9). In
reaction (8), oxygen from ceria is also involved in hydroxyl
formation and thus also in formate formation, making this an
associative mechanism involving red–ox regeneration (route C,
Fig. 9).
Based on the results over Pt/CeO2 catalyst under these ex-
perimental conditions, we conclude that WGS reaction pro-
ceeds via an associative mechanism (route B, Fig. 9). We ex-
clude route C from the reaction pathways because no H2 for-
mation occurred during CO pulses. Route A also is not involved
because no H2 formation occurred during H2O pulsing after the
reduced ceria was oxidized with N2O (H2O after CO–N2O–CO
pulses; Table 2, row Ia). At this moment, we do not rule out
(CO3)(s) + H2O(g) → CO2(g) + 2OH(s).
(4)
Both reactions (3) and (4) result in rehydroxylation of the ceria
surface, closing the catalytic cycle for the WGS reaction.
In the absence of OH groups, the formate route is not
possible. The importance of OH(s) groups in the WGS reac-
tion is confirmed by the experiments with N2O over Pt/CeO2-
containing formate/carbonate groups (Table 2, row Ia). It is
important to emphasize that N2O is not involved in the WGS
reaction. However, the aim of using N2O here is to decompose
the surface formate formed during CO pulses without regen-
erating the hydroxyl groups. In this way, the catalyst will be
free of reactive OH groups, making it possible to probe whether
a red–ox pathway (Fig. 9, route A) is involved in the reaction
mechanism. During N2O pulses, both H2 and CO2 were formed
(Table 2, row Ia), indicating the decomposition of surface for-
◦
route D. But because cerium carbonate is stable up to 430 C
mate,
[
43] and Pt/CeO2 deactivates due to the blocking of the active
(
COOH)(s) + N2O(g) → CO2(g) +(1/2)H2(g) + N2(g) + O(s).
sites by carbonate [16], we propose that route D does not con-
tribute to the WGS reaction. This is substantiated by the fact
(5)

K.G. Azzam et al. / Journal of Catalysis 251 (2007) 153–162
161
that no hydrogen was formed during H2O pulses after the se-
quence CO–N2O–CO when only carbonates were present.
the amount of H2 produced during CO pulsing (Table 2, row II)
would enable discrimination, but unfortunately, these data were
not available.
Gorte and Zhao [28] recently reported that the WGS reaction
◦
proceeds via a red–ox mechanism at 450 C over Pt/CeO cat-
We conclude from the above results that both oxygen va-
cancies on the reduced TiO and hydroxyl groups on TiO can
2
◦
alyst, based on the responses to CO and H2O pulses at 450 C.
2
2
Both CO2 and H2 were formed during CO pulsing [28]. This
observation was explained by the interaction of CO with hy-
droxyl H2, reducing the ceria and forming H2 and CO2. As a
result of H2O pulsing, both CO2 and H2 were also formed. It
has been claimed that H2 formation was due to the oxidation of
reduced ceria and that CO2 formation results from the decom-
position of carbonate. However, the temperature in the work of
take part in the catalysis of the WGS reaction. Thus, both the
associative route with red–ox regeneration pathway (route C,
Fig. 9) and the classical red–ox pathway (Fig. 9, route A) may
significantly contribute in the WGS mechanism. Route B can
be excluded, because no CO formation occurred during H O
2
2
pulses (Table 2, row II). Route D again can be ruled out because
no CO2 formation occurred during H2O pulsing after the re-
duced titania was oxidized with N2O (H2O after CO–N2O–CO
pulses; Table 2, row IIa), and because carbonate was not present
on TiO2 after CO pulsing, according to FTIR data (Fig. 6).
The activity of Pt/TiO2 seems to be determined by the abil-
ity of TiO2 to undergo red–ox as well as to allow formation of
OH groups. Decomposition of the intermediate formate species
seems facile.
◦
Gorte et al. was significantly higher (450 C) than that used in
◦
our study (300 C). Formates and carbonates are not stable at
the higher temperatures, inducing a change in the mechanism.
In the case of Pt/TiO2, the sequence for the WGS reaction
differs from that for ceria discussed above. Unlike in Pt/CeO2,
H2 is also formed simultaneously with CO2 during CO puls-
ing (Table 2, row II). There is only one source for hydrogen in
◦
the Pt/TiO2 catalyst during CO pulsing at 300 C, i.e., surface
For the Pt/ZrO catalyst, the WGS reaction seemed to follow
2
hydroxyl groups. Thus, the net reaction can be represented as
an associative route with red–ox regeneration (Fig. 9, route C).
During CO pulsing, the catalyst showed behavior similar to that
Pt–CO + OH(s) → CO2(g) +(1/2)H2(g) + Pt + *(s),
(9)
of Pt/TiO . CO pulsing results in CO and H formation, which
which corresponds to an associative mechanism with red–ox re-
generation in our scheme (Fig. 9, route C). It is not clear at this
2
2
2
can be described by reaction (9), were part of route C in Fig. 9.
The carbon balance indicates that no carbon species remained
on the surface (Table 2, row III). This is also confirmed by in
situ FTIR measurements (Fig. 8). It is still not clear whether
intermediate formate species were involved. They could be ob-
served with in situ FTIR during reaction (Fig. 8), but as soon as
the flow was switched to inert gas, the bands disappeared from
the spectrum. During H2O pulsing, no gaseous products were
observed, but the catalyst was able to produce CO2 and H2 at
the next CO pulse. From these findings, we conclude that route
B was not operative and that catalyst activation occurred mainly
due to hydroxyl groups formed according to reaction (8).
moment which type of hydroxyl group on reduced Pt/TiO sur-
2
face is relevant for WGS reaction. This determination requires
additional studies, which are beyond the scope of this article.
The carbon balance indicates that no carbon species re-
mained on the surface (row II). This finding is also confirmed
by in situ FTIR measurements (Fig. 6). The fact that no for-
mate/carbonate species were detected under reaction conditions
implies that their decomposition to CO and H is facile. We
2
2
propose, in agreement with Panagiotopoulou et al. [10], that
formate-type intermediates are the only oxygenate species rel-
evant to the WGS sequence over the Pt/TiO2 catalyst; thus, an
associative formate mechanism with a red–ox regeneration is
operative. We also can confirm that CO is quantitatively con-
verted to gaseous products in the case of TiO2. This is unlike
ceria, where part of the CO is retained as carbonate, occupying
sites for formate formation, which cannot yield any hydrogen.
Contacting the catalyst with water after exposure to CO (Ta-
ble 2, row II) resulted in hydrogen; no CO2 was observed,
because no oxygenate remained on the surface. The formation
of hydrogen indicates that reoxidation of the support occurred
according to
Normally, ZrO is considered a nonreducible oxide, even
2
though evidence exists that the surface is partly reducible at rel-
◦
atively higher temperatures (>600 C) [45]. However, the fact
◦
that at 300 C, CO2 and H2 were formed on a hydroxylated
surface (Table 2, row III) implies that the surface OH groups
were converted, leaving an oxygen vacancy (Fig. 9, route C).
When this surface was reoxidized by N O, no active hydrox-
2
yls were formed. The observation that CO2 was detected (Ta-
ble 2, row IIIa) during CO pulsing after the CO–N2O sequence
indicates that some lattice oxygen could be removed even at
◦
*(s)
+ H O → O + H2(g).
(10)
300 C (6). Therefore, both observations demonstrate that ZrO2
2
(g)
(s)
was able to provide oxygen. We speculate that low-coordination
sites in the zirconia surface were responsible. Accordingly, re-
generation occurred according to reaction (8), similar to the
situation for titania. The gaseous product distribution during
pulse experiments indicates that the WGS reaction over Pt/ZrO2
followed the associative mechanism with red–ox regeneration
(Fig. 9, route C). Routes A, B, and D can be excluded, because
there were no gaseous products during H2O pulsing (Table 2,
row III). Recently, Trovarelli et al. [46], using isotopic transient
It was reported earlier that the surface of TiO2 can be reduced
by CO [Eq. (11)] and oxidized by H2O [44],
O(s) + Pt–CO → CO2(g) + *(s).
(11)
The red–ox nature of TiO2 is also indicated by the formation
of hydrogen during the water pulsing after a CO–N2O–CO se-
quence. Therefore, we cannot exclude (as in the case of CeO2
and ZrO2 [see below]), the red–ox route (route A, Fig. 9) for
the WGS reaction in Pt/TiO2, implying involvement of the se-
quence of reactions (10) and (11). Quantitative determination of
12
13
◦
kinetics ( CO and CO) and in situ DRIFTS at 200 C, con-

162
K.G. Azzam et al. / Journal of Catalysis 251 (2007) 153–162
cluded that oxygenates, not excluding formate, are involved as
a reaction intermediate, in agreement with our conclusion.
[11] H. Iida, A. Igarashi, Appl. Catal. A 298 (2006) 152.
[12] P.S. Querino, J.R.C. Bispo, M.C. Rangel, Catal. Today 107–108 (2005)
920.
[
[
13] S.Y. Choung, M. Ferrandon, T. Krause, Catal. Today 99 (2005) 257.
14] S. Ricote, G. Jacobs, M. Milling, Y. Ji, P.M. Patterson, B.H. Davis, Appl.
Catal. A 303 (2006) 35.
5
. Conclusions
The reducibility of the support, as well as the stability of
[15] W. Ruettinger, X. Liu, R.J. Farrauto, Appl. Catal. B 65 (2006) 135.
[16] K.G. Azzam, I.V. Babich, K. Seshan, L. Lefferts, part 2, J. Catal. (2007),
in press.
formate and carbonate species together, determine the reaction
pathways contributing to the WGS reaction. Water activation
is achieved over the support through formation of surface hy-
droxyl groups on oxides and/or oxidation of the reduced sup-
port with H2O, resulting in H2 formation. Under the experimen-
[17] H. Iida, K. Kondo, A. Igarashi, Catal. Commun. 7 (2006) 240.
[18] O. Goerke, P. Pfeifer, K. Schubert, Appl. Catal. A 263 (2004) 11.
[19] Brunetti, G. Barbieri, E. Drioli, K.-H. Lee, B. Sea, D.-W. Lee, Chem. Eng.
Proc. 46 (2007) 119.
◦
tal conditions of this work (300 C), it appears that the classical
[20] M.E. Adrover, E. Lopez, D.O. Borio, M.N. Pedernera, in: F.N. Noronha,
M. Schmal, E.F. Sousa-Aguiar (Eds.), Studies in Surface Science and
Catalysis, vol. 167, Elsevier, Amsterdam, 2007, p. 183.
associative mechanism (route B) is operating for Pt/CeO2. In
Pt/ZrO2, the dominant reaction pathway is the associative for-
mate route with red–ox regeneration (route C), implying that
OH groups react with CO, leaving oxygen vacancies. This re-
sult is surprising, because reduction of ZrO2 is much more dif-
ficult than reduction of CeO2. Apparently, the remarkable sta-
bility of surface formates on CeO2 contributes to this. Finally,
for Pt/TiO2, both the associative formate route with red–ox re-
generation (route C) and the classical red–ox route (route A)
are possible reaction pathways contributing to the WGS reac-
tion. Pt/TiO2 was the only catalyst able to dissociate H2O to
H2 while reoxidizing oxygen vacancies generated by reduction
with CO.
[
[
21] Ch. Han, D.P. Harrison, Chem. Eng. Sci. 49 (1994) 5875.
22] D. Tibiletti, E.A. Bart de Graaf, S.Ph. Teh, G. Rothenberg, D. Farrusseng,
C. Mirodatos, J. Catal. 225 (2004) 489.
[23] O. Pozdnyakova, D. Teschner, A. Wootsch, J. Kröhnert, B. Steinhauer,
H. Sauer, L. Toth, F.C. Jentoft, A. Knop-Gericke, Z. Paál, R. Schlögl,
J. Catal. 237 (2006) 17.
[24] M.A. Henderson, Surf. Sci. Rep. 46 (2002) 1.
[25] J.R. Rostrup-Nielsen, in: J.R. Anderson, M. Boudart (Eds.), Catalysis, Sci-
ence and Technology, vol. 5, Springer-Verlag, Berlin, 1984, p. 1.
[26] K. Takanabe, K. Aika, K. Seshan, L. Lefferts, J. Catal. 227 (2004) 101.
[27] T. Bunluesin, R.J. Gorte, G.W. Graham, Appl. Catal. B 15 (1998) 107.
[
[
[
[
28] R.J. Gorte, S. Zhao, Catal. Today 104 (2005) 18.
29] P. Panagiotopoulou, D.I. Kondarides, Catal. Today 112 (2006) 49.
30] D.C. Grenoble, M.M. Estadt, D.F. Ollis, J. Catal. 67 (1981) 90.
31] G. Jacobs, E. Chenu, P.M. Patterson, L. Williams, D. Sparks, G. Thomas,
B.H. Davis, Appl. Catal. A 258 (2004) 203.
Acknowledgments
[32] G. Jacobs, S. Ricote, U.M. Grahm, P.M. Patterson, B.H. Davis, Catal. To-
day 106 (2005) 259.
The authors thank Ing. L. Vrielink for the XRF and BET
analysis and Ing. B. Geerdink and K. Altena-Schildkamp for
technical assistance. Financial support was provided by the
STW (project 790.36.030, The Netherlands).
[
33] S. Hilaire, X. Wang, T. Luo, R.J. Gorte, J. Wagner, Appl. Catal. A 215
(2001) 271.
[34] X. Wang, R.J. Gorte, Appl. Catal. A 247 (2003) 157.
[35] H. Sakurai, T. Akita, S. Tsubota, M. Kiuchi, M. Haruta, Appl. Catal. A 291
(2005) 179.
References
[36] D. Teschner, A. Wootsch, T. Roder, K. Matusek, Z. Paal, Solid State Ion-
ics 141–142 (2001) 709.
[
[
37] Holmgren, B. Andersson, D. Duprez, Appl. Catal. B 22 (1999) 215.
38] D. Tibiletti, A. Goguet, F.C. Meunier, J.P. Breen, R. Burch, Chem. Com-
mun. 14 (2004) 1636.
[
1] D.S. Newsome, Catal. Rev. Sci. Eng. 21 (1980) 275.
[2] C.H. Bartolomew, R.J. Farrauto, in: C.H. Bartolomew, R.J. Farrauto
(Eds.), Fundamentals of Industrial Catalytic Processes, Wiley, Hoboken,
[39] A. Goguet, F.C. Meunier, D. Tibiletti, J.P. Breen, R. Burch, J. Phys. Chem.
B 108 (2004) 20240.
NJ, 2006, p. 909.
[3] W. Ruettinger, O. Ilinich, R.J. Farrauto, J. Power Sources 118 (2003) 61.
[4] G. Jacobs, S. Ricote, B.H. Davis, Appl. Catal. A 302 (2006) 14.
[5] G. Jacobs, U.M. Graham, E. Chenu, P.M. Patterson, A. Dozier, B.H.
Davis, J. Catal. 229 (2005) 499.
[40] P.B. Wells, Appl. Catal. 18 (1985) 259.
[41] S. Sharma, S. Hilaire, J.M. Vohs, R.J. Gorte, H.W. Jen, J. Catal. 190 (2000)
199.
[6] A. Goguet, S.O. Shekhtman, R. Burch, C. Hardcare, F.C. Meunier, G.S.
Yablonsky, J. Catal. 237 (2006) 102.
[42] G. Jacobs, P.M. Patterson, L. Williams, D. Sparks, B.H. Davis, Catal.
Lett. 96 (2004) 97.
[
7] E. Chenu, G. Jacobs, A.C. Crawford, R.A. Keogh, P.M. Patterson, D.E.
Sparks, B.H. Davis, Appl. Catal. B 59 (2005) 45.
[43] X. Liu, W. Ruettinger, X. Xu, R. Farrauto, Appl. Catal. B 56 (2005) 69.
[44] M. Calatayud, A. Markovits, M. Menetrey, B. Mguig, C. Minot, Catal.
Today 85 (2003) 125.
[45] J. Zhu, S. Albertsma, J.G. van Ommen, L. Lefferts, J. Phys. Chem. B 109
(2005) 9550.
[8] E. Xue, M. O’Keeffe, J.R.H. Ross, Surf. Sci. Catal. 130 (2000) 3813.
9] Y. Sato, K. Terada, S. Hasegawa, T. Miyao, S. Naito, Appl. Catal. A 296
[
(2005) 80.
[
10] P. Panagiotopoulou, A. Christodoulakis, D.I. Kondarides, S. Boghosian,
[46] D. Tibiletti, F.C. Meunier, A. Goguet, D. Reid, R. Burch, M. Boaro, M.
Vicario, A. Trovarelli, J. Catal. 244 (2006) 183.
J. Catal. 240 (2006) 114.