Selective CO oxidation in the presence of hydrogen: Fast parallel screening and mechanistic studies on ceria-based catalysts

Article abstract of DOI:10.1016/j.jcat.2004.04.031

The effects of doping the Ce support of Pt catalysts were investigated for three close reactions, i.e., selective CO oxidation in the presence of hydrogen (SelOx), water-gas shift (WGS), and reverse WGS (RWGS), by means of fast parallel screening and mech

Full text of DOI:10.1016/j.jcat.2004.04.031

Journal of Catalysis 225 (2004) 489–497
www.elsevier.com/locate/jcat
Selective CO oxidation in the presence of hydrogen: fast parallel
screening and mechanistic studies on ceria-based catalysts
Daniele Tibiletti a, E.A. Bart de Graaf b, Siew Pheng Teh a, Gadi Rothenberg b,
David Farrusseng a, and Claude Mirodatos a,∗
a
Institut de Recherches sur la Catalyse, CNRS, 2, avenue Albert Einstein, F-69626, Villeurbanne cedex, France
b
Chemical Engineering Department, University of Amsterdam, Nieuwe Achtergracht 166, 1018 WV Amsterdam, The Netherlands
Received 17 December 2003; revised 24 April 2004; accepted 26 April 2004
Available online 4 June 2004
Abstract
An accelerated investigation of various ceria-supported platinum catalysts by means of parallel synthesis and testing reactors is reported.
The aim is to discover new catalysts for purifying hydrogen for fuel cells and other applications by studying three reactions: selective CO
oxidation in the absence and in the presence of hydrogen (SelOx), water–gas shift (WGS), and reverse water–gas shift (RWGS). A set of
catalysts is prepared by impregnating platinum on various doped ceria supports. These catalysts display, despite their nearly identical X-ray
crystal structures, diverse redox/acidic/basic properties and different selectivities in the three test reactions. The best SelOx catalysts are also
the most active for WGS/RWGS equilibration. In contrast to the conventional two-stage approach of discovery screening followed by lead
optimization, we demonstrate that it is possible to extract valuable mechanistic information directly from the discovery stage. Specifically,
we elucidate the elementary steps coupling the SelOx and WGS processes and gain insight as to the combined effects of hydrogen presence
and cation doping on the CO oxidation reaction.
2004 Elsevier Inc. All rights reserved.
Keywords: Combinatorial catalysis; Doped ceria; Fuel cells; Water–gas shift; CO selective oxidation; Platinum
1. Introduction
Metal-catalyzed CO oxidation has been studied in depth
during the last century. The classical catalyst is supported Pt,
but Pd, Rh, and even Au (as gold clusters [5,6]) can be used
[7]. As early as 1922, Langmuir [8] laid the basis for two
possible reaction pathways. In the first one, only the metal-
lic surface is involved: both CO and O2 molecules adsorb on
the platinum crystallite and react together. In the second one,
a CO molecule absorbed on the platinum surface reacts with
an oxygen atom adsorbed on the support after spillover steps
between the two phases. More recent versions of these mech-
anisms are more complex [9,10], but the second pathway is a
good starting point to understand the following investigation
of catalysts containing supports that can activate oxygen,
maximizing the catalytic benefits from the support–oxygen
interactions.
Fuel cell technology plays a major part in future scenarios
of sustainable energy management [1,2]. Proton-exchange
membrane (PEM) cells are particularly attractive for small-
scale and automotive applications, but they suffer from
one major disadvantage: CO contamination in the hydrogen
feed poisons the metal electrodes, and concentrations below
10 ppm are required for steady-state operation [3,4]. To ful-
fill that requirement, the final stage of hydrogen purification
generally consists of oxidizing selectively the remaining CO
to ppm levels. Selective oxidation of CO at high concentra-
tion in the presence of hydrogen is preferred only if these
low CO quantities can be achieved.
Ceria-based mixed oxides (CexM1−xOy) are versatile
solid oxygen exchangers. At high temperatures (400–800◦C)
the redox cycle Ce3+ ⇔ Ce4+ + e− facilitates oxygen stor-
age and release from the bulk fluorite lattice. This makes
them ideal candidates for catalytic and/or electrocatalytic
*
Corresponding author.
E-mail addresses: gadi@science.uva.nl (G. Rothenberg),
farrusseng@catalyse.cnrs.fr (D. Farrusseng), mirodatos@catalyse.cnrs.fr
(C. Mirodatos).
0021-9517/$ – see front matter
doi:10.1016/j.jcat.2004.04.031
2004 Elsevier Inc. All rights reserved.

490
D. Tibiletti et al. / Journal of Catalysis 225 (2004) 489–497
oxidation applications such as solid–oxide fuel cells [11].
However, the surface redox chemistry of ceria is sensitive
even at low temperatures to crystal structure defects [12],
which can be tuned by substituting some of the Ce cations
with ions of different size and/or charge [13,14].
In this paper, we study the effects of doping the ceria
support of platinum catalysts for three close reactions: se-
lective CO oxidation in the presence of hydrogen (SelOx),
water–gas shift (WGS), and reverse water–gas shift (RWGS)
using a high-throughput methodology. All of the catalysts
are screened for all three reactions. Subsequently, we discuss
the possible application of these catalysts under hydrogen-
free and hydrogen-rich conditions, as well as the elementary
catalytic steps in these systems and the advantages of using
high-throughput methods to extract mechanistic information
directly from the primary screening stage.
2.2. Fast parallel catalyst testing
All the reactions were performed in a 16-channel par-
allel plug flow reactor, constructed in house [16], and now
commercialized as SWITCH 16 reactor by AMTEC GmbH.
Reaction products were monitored by fast-gas chromatogra-
phy and on-line mass spectrometry. GC analysis was per-
formed on a two-channel Variant microGC CP 2003 in-
strument equipped with two thermal conductivity detectors.
Channel A was used for H2, O2, N2, CH4, and CO separa-
tion (MolSieve 5A Plot, with Ar as carrier gas at 80 ◦C),
and channel B for CO2 and H2O separation (PoraPlot Q,
with He as carrier gas at 80 ◦C). MS analysis was performed
using an Inficom CPM instrument. Unless noted otherwise,
chemicals were purchased from commercial sources (> 99%
pure) and used as received. In each run, one reactor chan-
nel was loaded with graphite and was used as a reference,
and a channel was loaded with a dopant-free Pt/CeO2 cata-
lyst for comparison. The temperature was varied from 100
to 200 ◦C, with a 5-min isotherm every 20 ◦C. The carbon
mass balance was calculated by monitoring the effluents
from the bypass reactor and was between 98 and 102%. Each
catalyst’s oxygen-storage capacity (OSC) was quantified by
measuring the H2 consumption in transient mode using the
parallel reactor under isothermal conditions at various tem-
peratures: the catalyst sample was oxidized in air for 15 min,
flushed with N2 and then reacted with a reducing mixture of
10% H2 in He, monitoring continuously the hydrogen con-
sumption by MS for 3 min (all the other catalysts were kept
meanwhile under N2 flow). After a cycle was completed the
temperature was increased by 20 ◦C and a new cycle started
following the previous experimental sequence. The OSC for
a given temperature is given in micromoles H2 consumed per
gram catalyst. Additional OSC measurements were carried
out by using CO instead of H2, keeping the above procedure
unchanged.
2. Experimental
2.1. Catalysts
A set of 12 catalysts containing 2 wt% Pt on Ce0.9M0.1Ox
was prepared and characterized (where M = Pb, Bi, Zr, V,
W, Mo, Y, La, In, Sn, and two combinations of Zr/Bi, re-
spectively).
The supports for catalysts containing only one dopant
were prepared using a modified version of the method re-
ported by Inomata and co-workers [15]. The mixed ox-
ides were synthesized directly by boiling ground mix-
tures of their nitrate precursors under vacuum, followed by
calcination, as reported previously [13]. The correspond-
ing metal nitrate hydrate precursors were used, except for
Ce0.9V0.1Oy, Ce0.9W0.1Oy, and Ce0.9Mo0.1Oy, where the
ammonium vanadate, tungstanate, and molybdate were used
and Ce0.9Y0.1Oy and Ce0.9Sn0.1Oy, where YCl3 and SnCl2
were used. Supports with the two combinations of Zr/Bi
were prepared by impregnation of ceria/zirconia with Bi
(referred to as MS) and by homogeneous deposition pre-
cipitation of Bi on ceria/zirconia (referred to as HDP).
A 2 wt% Pt (as chloroplatinic acid) was loaded on each
support by impregnation, using a Sophas robotic system
(Zinsser Analytic). Two milliliters of water was dropped
on the support, after which the calculated amount H2PtCl6·
6H2O soln (0.01 M) was added. The suspensions were
stirred for 3 h, dried at 80 ◦C for 4 h, and then calcined for
3 h at 500 ◦C (ramp rate 300 ◦C/h). The catalysts were then
pelletized and crushed to get a uniform powder with a grain
size range of 400–600 µm.
2.3. Procedure for CO oxidation
The catalysts were tested for the CO oxidation reaction
both in the absence and in the presence of H2 (SelOx), using
the following feed mixtures: 1% CO, 2% O2, and balance N2
and 1% CO, 2% O2, 20% H2, and balance N2, respectively.
In both cases the total flow rate was 2−51.75 mL/min and the
gas space velocity (GSV) was 22 L h g−ca1t ). The effect of
water was studied using a feed of 1% CO, 2% O2 20% H2,
10.0% H2O, and balance N2. The temperature was varied
from 60 to 200 ◦C, with a 5-min isotherm every 20 ◦C.
2.4. Procedure for WGS and RWGS reactions
A dopant-free Pt/CeO2 catalyst containing 2 wt% Pt was
prepared as reference system and referred to as Pt/Ce. In
order to check the effect of Pt loading, an additional dopant-
free system containing 4.75% wt% Pt was prepared using a
high surface area ceria powder (180 m2/g) and referred to as
4.75% Pt/Ce.
For WGS and RWGS the feeding gases mixtures were:
1% CO, 3% H2O, N2 to balance and 1% CO2, 17% H2, N2 to
balance, respectively, wit−h1a total flow rate of 21.0 mL/min
−1
and a GSV of 18 L h
g
.
cat

D. Tibiletti et al. / Journal of Catalysis 225 (2004) 489–497
491
3. Results
3.2. Selective CO oxidation in the presence of hydrogen
(SelOx)
3.1. CO oxidation in the absence of hydrogen
When the same catalysts were tested in the presence of
hydrogen (CO:O2:H2:N2 = 1:2:20:77), a different picture
emerged (Fig. 2A). Apart from a V-containing system, which
was active either in the presence or in the absence of hydro-
gen, other samples (with Y, In, and Zr dopants), that were
inactive in the absence of hydrogen, now gave full CO con-
version at temperatures < 100 ◦C. However, as the temper-
ature was increased from 100 to 200 ◦C the CO conversion
tended to decrease. With less active samples (Bi and ZrBi
dopants), CO conversion begun only at 120 ◦C, increasing
to a maximum value (not necessarily 100%) and then de-
creasing at higher temperatures. Unlike the hydrogen-free
screening, the most active doped catalysts were found more
active than the reference Pt/Ce sample, and even slightly
more active than the highly loaded 4.75 wt% Pt/Ce sample.
The selectivity toward CO2, i.e., the percentage of oxygen
consumed for oxidizing CO to CO2 divided by the total
amount of consumed oxygen, also depended on the ceria
doping (Fig. 2B). Less active samples presented the high-
est selectivity toward CO2, with less oxidation of hydrogen
into water (note that full oxygen consumption was observed
in all cases). Another trend was that the higher the tempera-
ture, the lower the selectivity toward CO2.
Under hydrogen-free CO oxidation conditions (CO:O2:
N2 = 1:2:97), 4 of the 12 doped catalysts, containing the
dopants Pb, V, W, and Mo, reached 100% CO conversion
already at 160 ◦C, while the reference catalyst Pt/Ce was
found less active (Fig. 1). By comparing the dopant-free 2
and 4.75 wt% Pt/Ce systems, a large positive effect of Pt
loading was observed, since the highly loaded system was
far the most active among the whole series.
3.3. Selective CO oxidation in the presence of hydrogen
and water
Under industrial or pilot plant feed conditions (i.e., us-
ing a feed from the outlet of the upstream WGS reactors) the
partial pressure of water is generally between 10 and 40%.
We studied therefore the influence of water for all catalysts,
adding 10% of water to the SelOx feed. Fig. 3 shows the
changes in CO conversion as a function of temperature for
the most active catalysts, i.e., in a domain where the conver-
Fig. 1. CO oxidation in the absence of hydrogen: CO conversion (%) as a
function of temperature ( C).
◦
◦
Fig. 2. CO conversion (A) and selectivity toward CO (B) in the presence of hydrogen as a function of temperature ( C).
2

492
D. Tibiletti et al. / Journal of Catalysis 225 (2004) 489–497
◦
Fig. 3. Comparison of SelOx performances as a function of temperature ( C) without (A) and with (B) water addition.
sion decreases with temperature, both without (Fig. 3A) and
with (Fig. 3B) water addition. Adding water left the selec-
tivity basically unchanged, but increased the CO conversion
by 5–15%.
3.4. Water–gas shift and reverse water–gas shift reactions
During the above experiments, the catalysts were con-
tacted with mixtures of H2O, CO, CO2, and H2. There-
fore, the gas composition may tend to equilibrate accord-
ing to WGS/RWGS thermodynamics as these catalysts are
also known to be active in reaching these equilibria. In the
absence of added water in the feed, i.e., under “dry CO
oxidation” in the presence of H2 (conditions prevailing in
Figs. 2 and 3A), the gas outlet concentration can be simu-
lated by assuming that (a) a complete CO oxidation proceeds
at full O2 conversion within the very first part of the cata-
lyst bed, and (b) the WGS/RWGS equilibrium is established
in the remaining part of the catalyst bed (see Scheme 1).
As such, if the catalyst bed is fed with a ratio of 1:2:20:77
CO:O2:H2:N2, after full O2 conversion, the remaining part
of the catalyst would be contacted with a ratio of 1:3:17:77
CO2:H2O:H2:N2. Assuming such a mixture is totally equili-
brated according to WGS thermodynamics, the backforma-
tion of CO via RWGS was calculated at the reactor exit1 and
Fig. 4. Comparison of RWGS (CO conversion, A) and WGS (CO conver-
2
sion, B) performances as a function of temperature ( C).
◦
found to be ca. 200 and 3000 ppm at 200 and 400 ◦C, respec-
tively (Table 1).
To validate this simulation, we tested the activity of all
the catalysts in reaching the WGS/RWGS equilibria. In a
first set of experiments, the feed composition corresponding
to RWGS was CO2:H2:N2 = 1:17:82 and the reaction was
performed at temperatures ranging from 100 to 250 ◦C. As
can be seen in Fig. 4A, most catalysts gave ca. 10% CO2
conversion into CO at 200 ◦C, with some reaching 26% con-
version at 250 ◦C (which correspond to outlet CO concen-
trations of 1000 and 2500 ppm, respectively), while the cal-
culated equilibrated values were 24% at 200 ◦C and 35% at
250 ◦C. This indicates that most of the catalysts indeed equi-
Scheme 1. Selox reactor axial mechanistic sequence.
1
Thermodynamic calculations were performed using the Gibbs en-
ergy minimization function of the HCS Chemistry software, available from
http://www.outokumpu.com/hsc.

D. Tibiletti et al. / Journal of Catalysis 225 (2004) 489–497
493
Table 1
a
Simulation of the SelOx and RWGS processes
Reaction
Inlet composition (mol%)
Composition after reaction/equilibrium (mol%)
CO
CO
O
H
H O
2
N
CO
CO
O
H
H O
2
N
2
2
2
2
2
2
2
2
CO and H oxidation
2
CO + 1/2O → CO
1
0
2
20
17
0
3
77
79
0
1
0
17
3
79
2
2
H + 1/2O → H O
2
2
2
◦
RWGS equilibrium
0.02
0.30
0.97
0.69
0
0
16.8
16.7
3.2
3.3
79
79
200 C
0
1
0
◦
CO + H ↔ CO + H O
400 C
2
2
2
a
On the first line (corresponding to CO and H oxidation) the complete O consumption is assumed, enabling CO to be completely oxidized into CO , the
2
2
2
remaining part of oxygen being converted into H O. On the second line (corresponding to the WGS/RWGS equilibrium), the thermodynamics equilibria are
2
◦
calculated at 200 and 400 C on the basis of the inlet composition, given in the left part of line 2, similar to the right part of line 1.
librate the reacting mixtures according to the WGS/RWGS
thermodynamics under these conditions. In the presence of
water (i.e., under more realistic SelOx conditions), much
less CO2 to CO conversion was observed, as expected from
the WGS/RWGS equilibria (0.3 and 0.6% CO2 conversion
at 200 and 250 ◦C, respectively, corresponding to outlet CO
concentrations of 300 and 600 ppm, respectively).
To further examine this trend in WGS/RWGS equilibra-
tion, the catalysts were also tested for WGS activity using a
feed ratio of 1:3:96 CO:H2O:N2. CO conversion was found
to increase with temperature, as shown in Fig. 4B. Most cata-
lysts gave full CO conversion, i.e., reached the WGS/RWGS
equilibrium (CO conversion of 99.8 and 99.5% at 200 and
250 ◦C, respectively). This demonstrates the ability of ceria-
based materials to equilibrate CO/CO2/H2O/H2 mixtures,
independent of the initial feed composition. Moreover, the
Fig. 5. Changes in CO conversion under SelOx conditions as a func-
tion of temperature ( C) for two contact times (a, 170 ms; b, 48 ms).
◦
three most active dopants for CO oxidation under hydrogen-
rich conditions (Sn, V, Zr, see Fig. 4B) also gave the highest
CO conversion for WGS at low temperatures.
Th1, thermodynamic equilibrium for SelOx inlet composition (CO:O :
2
H :N = 1:2:20:77 vol%); Th2, thermodynamic equilibrium for similar
2
2
SelOx inlet composition but assuming that the water formed by CO/H
2
oxidation is trapped by ceria.
3.5. Effect of contact time
Two effects are observed: (a) the light-off temperature is ca.
10 ◦C higher for the shorter catalyst bed, and (b) the shorter
the contact time, the closer is the CO conversion to equilib-
rium after light-off (see Section 4).
If indeed RWGS takes place under SelOx conditions, it
must occur preferentially downstream in the reactor, i.e., af-
ter the total CO oxidation inlet zone, where the concentration
of CO2 is sufficient. To check this, the effect of contact time
was investigated. Fig. 5 compares the changes in CO conver-
sion with temperature as a function of contact time which
was varied by using either 20 or 70 mg of 4.75% Pt/Ce,
which corresponded to a bed length of 1 or 3 mm, and con-
tact time of 48 or 170 ms, respectively (assuming a mean
bed density of 0.96 g/mL). Fig. 5 also shows the thermo-
dynamic equilibrium corresponding to the real SelOx input
concentration (curve Th1, which includes the values given
for 200 ◦C in Table 1, second line, right side). In order to ac-
count for a possible adsorption of water by ceria (at least un-
der initial testing conditions when the full adsorption capac-
ity is not reached yet), the equilibrium was also calculated by
assuming that the water formed during the initial total oxi-
dation step (producing CO2 and H2O) is trapped by the ceria
support (curve Th2, which corresponds to the inlet compo-
sition given in Table 1, second line, left side, but without
water, i.e., CO:O2:CO2:H2O:H2:N2 = 0:0:1:0:17:82 vol%).
Fig. 6 reports the changes in CO conversion keeping
W/F ratio (contact time) constant but varying W and F val-
ues. At low flow rate (curve a) the CO conversion is high
at low temperature and decreases with gas inlet temperature
while at higher flow rate (curve b) the CO conversion is low
at low temperatures and increases as T increases up to about
200 ◦C. In Fig. 6, the temperature measured in the middle
of the catalyst bed remains always higher than the one mea-
sured at the bed inlet (from about 40 ◦C at low temperatures
to about 20 ◦C at high temperatures).
3.6. Oxygen-storage capacity
To study further the effects of hydrogen on CO oxida-
tion, the catalysts oxygen-storage capacity was evaluated by
first oxidizing the catalyst and subsequently flowing hydro-
gen over it, measuring only the available surface and/or bulk

494
D. Tibiletti et al. / Journal of Catalysis 225 (2004) 489–497
CO oxidation ranking or with the SelOx one (results not re-
ported here).
4. Discussion
4.1. Effects of hydrogen on CO oxidation
The accepted mechanism for platinum-catalyzed CO ox-
idation assumes that both CO and O2 are activated on the
metal particle surface as COads and Oads followed by a
Langmuir–Hinshelwood surface reaction between the two
adsorbed species and release of CO2 [19],
Fig. 6. Changes in CO conversion under SelOx conditions over 4.75% Pt/Ce
◦
CO + ∗ → CO∗,
(1)
(2)
(3)
as a function of gas inlet temperature ( C) at similar contact time = 170 ms
(W/F = 2.7) but for different W and F values, with curve a, W = 70 mg
and F = 26 mL/min; and b, W = 100 mg and F = 36 mL/min. Inlet com-
position: CO:O :H :N = 1:2:20:77 vol%. Curve c reports the temperature
O2 + 2∗ → 2O∗,
2
2
2
CO∗ + O∗ → CO2 + 2∗,
measured in the middle of the catalyst bed.
where ∗ is a platinum active site.
In the case of ceria-based catalysts, the support plays a
key role, as the reducible ceria may also provide adsorbed
oxygen species that can spill over from the support to the
metal particles, the reduced Ce3+ centerers being subse-
quently reoxidized by gaseous oxygen:
2Ce4+ + O2− + ∗ → 2Ce3+ + O∗,
(4)
(5)
2Ce3+ + 1/2O2 → 2Ce4+ + O2−
.
In the presence of hydrogen, competition between H2 ox-
idation and CO oxidation is likely to influence the equilibria
at the surface. One option is for hydrogen to adsorb on the
Pt sites, competing with CO activation:
H2 + 2∗ → 2H∗,
(6)
(7)
(8)
O2 + 2∗ → 2O∗,
2H∗ + O∗ → H2O + 2∗.
Hydrogen may also react directly with ceria lattice oxy-
gen (Eq. (9)), creating an anionic vacancy and increasing
the mobility of the remaining lattice oxygen atoms. The re-
sulting hydroxyl anions can combine to give water and O2−
(Eq. (10)) with the Ce3+ cations reoxidized by molecular
oxygen to Ce4+ as above:
Fig. 7. Oxygen-storage capacity (µmol of hydrogen oxidized per gram cat-
alyst) after reoxidation and flushing with Ar, as a function of temperature
( C).
◦
2Ce4+ + 2O2− + H2 → 2Ce3+ + 2OH−,
(9)
oxygen [17,18]. The OSC was measured as a function of
temperature for each catalyst (Fig. 7). Over the whole series,
the OSC ranking was found to strictly parallel the CO oxi-
dation ranking carried out in the absence of H2 (cf. Figs. 1
and 7), the most active CO oxidation catalysts exhibiting
the highest OSC, and vice versa. Moreover, the OSC values
were generally independent of the temperature at this range.
Above 200 ◦C, H2 consumption increased strongly for all
catalysts (in the case of La, Sn, In, and Zr dopants there was
an order of magnitude difference between 200 and 160 ◦C).
OSC measurements carried out by using CO oxidation
instead of H2 did not reveal any relationship either with the
2OH− → H2O + O2−
.
(10)
Doping the fluorite structure with different cations may
induce various structural effects, such as promoting the cre-
ation and/or stabilization of oxygen vacancies [13] and mod-
ifying surface acidity and basicity, these effects being likely
to interact on the oxygen diffusion. The latter is usually lim-
ited to surface diffusion for undoped ceria (bulk diffusion
can also be considered after Zr addition [20]). Therefore,
doping can change the above elementary steps for CO ox-
idation which involve ionic oxygen migration and diffusion.

D. Tibiletti et al. / Journal of Catalysis 225 (2004) 489–497
495
4.2. CO oxidation in the absence of hydrogen
at interfacial sites. The effect of these dopants could also be
analyzed as preventing efficiently ceria surface from being
reduced by hydrogen (Eq. (9)), possibly by providing extra
sources of ionic oxygen, either from the bulk (for Ce/Zr sys-
tems) or via heterovalent substitution. Very little change in
the selectivity was observed after adding water to the SelOx
system, which may be explained by the fact that OH groups
are already present on ceria surface under “dry” SelOx con-
ditions (water formed only from hydrogen oxidation).
Catalysts activity for CO oxidation in the absence of hy-
drogen parallels strictly their OSC. This indicates that the
activated oxygen required for reacting with adsorbed carbon
monoxide on Pt is provided by oxygen spillover from ce-
ria and not from gaseous oxygen activation on Pt. Hence,
the rate-determining step (RDS) for CO oxidation appears
to be the surface diffusion of oxygen from ceria to platinum,
which agrees with the fact that the OSCs are almost indepen-
dent of the reaction temperature (Fig. 7). The large increase
in OSC noted for all systems at 200 ◦C and above may reflect
the direct reduction of ceria by hydrogen at these tempera-
tures.
The most performing systems both for CO oxidation in
the absence of hydrogen and for OSC (measured by hydro-
gen oxidation) were found to contain Pb, V, W, and Mo ceria
doping cations. A common feature for these cations is that
the most stable forms of the corresponding oxides are essen-
tially amphoteric or acid except V, which can be either acid
or basic. By maintaining an optimized level of surface acid-
ity, these dopants would somehow neutralize ceria basicity
(hydroxyl groups) and therefore inhibit formation of stable
adspecies such as formates and carbonates which would, in
turn, inhibit oxygen mobility and spillover toward activated
CO on platinum particles (Eq. (4)) [20].
Another effect was clearly related to the content of Pt,
high loading favoring CO oxidation and OSC. This may
suggest that within the RDS of surface diffusion of oxygen
from ceria to platinum, the number of interfacial metal atoms
(linked to the concentration of particles and their mean size)
plays also a major role in the overall rate of reaction.
Finally, the absence of any clear relationship between
CO oxidation and OSC measured by CO oxidation would
indicate that in the absence of a continuous source of oxy-
gen, CO activated on Pt particles may also migrate toward
the ceria surface to react with OH groups for forming for-
mate species, as clearly revealed by in situ DRIFT studies
carried out on similar systems [21,22]. Thus more complex
processes involved in OSC-CO measurements rule out any
simple relationship with CO oxidation.
4.4. WGS/RWGS equilibria
Except for Eq. (2), we can consider the elementary reac-
tions shown above to be partially reversible. A summation of
these steps corresponds essentially to the WGS equilibrium.
This explains why the most efficient SelOx catalysts also
perform well in equilibrating WGS mixtures. A direct con-
sequence is that under a hydrogen-rich SelOx atmosphere,
after CO has been oxidized to CO2, it can reform by RWGS.
This is clearly demonstrated in the simulations and in the
contact time experiments shown in Fig. 5. However, for the
latter (carried out in the absence of added water), the out-
let CO concentration was much higher than expected from
the RWGS thermodynamics, especially for longer contact
times (longer bed lengths). This effect can most probably
be assigned to a temperature rise within the reactor due to
the exothermic initial CO and H2 oxidation. As a matter of
fact, this effect is clearly revealed in Fig. 6 where the tem-
perature within the bed is found to be about 40 ◦C higher
than the inlet temperature for a low gas flow rate, which
corresponded to a high initial conversion of CO and then
decreasing at higher temperatures (curve a). For a higher
flow rate (curve b), the conversion profile was now increas-
ing with temperature as expected for an activated oxidation
process. This latter profile may be assigned to a better heat
transfer from the bed to the outlet effluents. Thus, at high
contact time (Fig. 5, curve a) or low overall flow rate (Fig. 6,
curve a), the temperature rise due to CO oxidation would
expand to the WGS/RWGS downstream zone and favor the
RWGS equilibrium (Scheme 1) and explain a higher out-
let CO concentration. In addition, higher temperature in the
oxidation zone might also favor the oxidation of H2 at the
expense of CO oxidation due to CO desorption, which would
also contribute to higher outlet CO concentration by decreas-
ing CO vs H2 oxidation selectivity.
4.3. CO oxidation in the presence of hydrogen (SelOx)
The change in the catalysts CO oxidation ranking in the
presence of hydrogen (SelOx conditions) indicates that the
RDS of oxygen surface diffusion is affected. Contrary to the
previous analysis for CO oxidation in the absence of hydro-
gen or OSC, it appears that most of the best ceria dopants
for SelOx—V, In, Sn, Zr, and Y—are essentially basic ox-
ides in their stable form, except again V, which can be both
acidic or basic, depending on its state of valence. By gen-
erating or stabilizing basic OH groups on ceria surfaces in
place of oxygen vacancies, these cations would therefore
limit oxygen mobility and spillover, which, in turn, could
tune the competition between CO and H2 for being oxidized
In addition to the above temperature effects, a transient
trapping effect of water by ceria (occurring under the fast
parallel testing conditions before steady state is achieved)
could reinforce RWGS equilibrium displacement toward CO
formation. If we assume that the water produced by the
CO/H2 initial oxidation is even partly trapped by ceria, then
the experimental CO conversion becomes much closer to
the thermodynamic equilibrium for the case of short con-
tact times (cf. curves a and Th2 in Fig. 5). For longer contact
times (Fig. 5, curve b), the water formed by the RWGS re-
action could also be trapped, which would further increase

496
D. Tibiletti et al. / Journal of Catalysis 225 (2004) 489–497
the CO2 to CO conversion, as observed experimentally. In-
deed for catalytic measurements carried out after adsorption
steady state has been achieved, this effect could not be con-
sidered anymore. Further experiments are in progress to un-
ravel the various effects above described for a better process
control.
X-ray structures and compositions, diverse redox properties
and catalytic activity in CO oxidation, with and without hy-
drogen, and under water–gas shift conditions. The nature of
the dopant was found to strongly influence the performance
of Pt/ceria catalysts, both for CO oxidation in the presence
or absence of hydrogen and OSC, measured by hydrogen
oxidation. Common features like basicity or acidity of the
most stable oxide forms of the doping cations seem to play
a major role in systems performances, most likely by mod-
ifying surface oxygen migration: acidic/amphoteric cations
tend to favor both CO oxidation in the absence of hydrogen
and OSC, while basic cations tend to favor SelOx reaction.
However, a cation like vanadium, able to accommodate both
acidic or basic state, is beneficial in all cases. This statement
points out the limits of oversimplified relationships between
dopant nature and catalytic performances.
WGS/RWGS equilibria must be considered in addition to
the oxidation processes. This was affirmed by both experi-
ments and simulations. Therefore, because they are also ac-
tive in the WGS/RWGS equilibration, the most active ceria-
supported Pt catalysts must be used as SelOx catalysts at
low temperatures and short contact times. Further research
to optimize the CO oxidation performance in the presence of
hydrogen by combining diverse supports with various noble
and transition metals is under progress in our laboratories.
4.5. High-throughput catalyst discovery—Pros and cons
Finally, it is worthwhile to consider the investigation
method employed here, namely, the use of high-throughput
equipment for the extraction of mechanistic information. In
this study, more than 600 experiments were performed in 3
weeks. In contrast to the usual two-stage “discovery” and
“optimization” approach, we used the high-throughput sys-
tem in a manner similar to conventional catalysis investiga-
tions [23]. This means that only the first set of experiments
was designed a priori, whereas the following sets were each
designed in light of the previous results. It should be pointed
out that data processing was a rate-limiting step even though
specialized VisualBasic routines were used.
This modus operandi is not merely about saving time. In
the traditional combinatorial approach [24–26]it is generally
considered that selection of performing elemental combina-
tions can be performed in the so-called primary screening
(discovery stage) where testing conditions are often simpli-
fied to speed up the screening process. Then, in the sec-
ond stage (lead optimization), hits from the first screening
are tested under more realistic conditions. This, however,
can lead to major errors: if, to accelerate the initial screen-
ing, the discovery of new SelOx catalysts had been carried
out in the absence of hydrogen (or just by performing the
even faster OSC measurements), the ranking information ob-
tained would have been totally misleading!
Another point relates to data quality and meaning for
complex and intricate reactions like the ones studied here:
from the contact time experiments we can see that the
risks for data misinterpretation when high-throughputexper-
iments are carried out without satisfactory control of temper-
ature and pressure conditions are high (including possibly
transient adsorption effects).
The choice of the discovery test reactions is therefore cru-
cial. One option, as shown here, is to try and test sets of
diverse catalyst formulations combined with real-time per-
formance analysis, including transient experiments (like for
OSC measurements). Robots are available for doing the ex-
periments, but the bottleneck lies now with the data analysis
and processing. Optimizing the methods for high-throughput
data management including kinetic data obtained under non-
stationary conditions [27] is now the primary objective if we
use high-throughput technologies to discover new catalysts.
Acknowledgments
Part of this work was supported by the EC program
COMBICAT (GRRD-CT 1999-00022) and D.T. fully ac-
knowledges the EC program Marie Curie Host Fellow-
ships (HPMT-CT-2000-00160) for granting his stay at IRC.
AMTEC GmbH is fully acknowledged for having supported
part of the parallel reactor development, now commercial-
ized as SWITCH 16 reactor (http://www.amtec-chemnitz.de).
References
[1] L. Carrette, K.A. Friedric, U. Stimming, Chem. Phys. Chem. 1 (2000)
162–193.
[2] R.M. Ormerod, Chem. Soc. Rev. 32 (2003) 17–28.
[3] L.P.L. Carrette, K.A. Friedrich, M. Huber, U. Stimming, Phys. Chem.
Chem. Phys. 3 (2001) 320–324.
[4] J.J. Baschuk, X. Li, Int. J. Energy Res. 25 (2001) 695–713.
[5] T.V. Choudhary, D.W. Goodman, Top. Catal. 21 (2002) 25–34.
[6] H. Häkinnen, S. Abbet, A. Sanchez, U. Heiz, U. Landman, Angew.
Chem., Int. Ed. 42 (2003) 1297–1300.
[7] A.K. Santra, D.W. Goodman, Electrochim. Acta 47 (2002) 3595–
3609.
[8] I. Langmuir, Trans. Faraday Soc. 17 (1922) 672–675.
[9] J.K. Nørskov, Nature 414 (2001) 405–406.
[10] Y. Zhang-Steenwinkel, J. Beckers, A. Bliek, Appl. Catal. A 235 (2002)
79–92.
[11] S.D. Park, J.M. Vohs, R.J. Gorte, Nature 404 (2000) 265–267.
[12] A.E.C. Palmqvist, M. Wirde, U. Gelius, M. Muhammed, Nanostruct.
Mater. 11 (1999) 995–1007.
[13] G. Rothenberg, E.A. de Graaf, A. Bliek, Angew. Chem., Int. Ed. 42
(2003) 3066–3068.
5. Conclusions
Ceria-based supports doped with small amounts of het-
eroatom cations show, despite their almost identical crystal

D. Tibiletti et al. / Journal of Catalysis 225 (2004) 489–497
497
[14] M. Mogensen, N.M. Sammes, G.A. Tompsett, Solid State Ionics 129
(2000) 63–94.
[22] E. Odier, Y. Schuurman, H. Zanthoff, C. Millet, C. Mirodatos, Stud.
Surf. Sci. Catal. 133 (2001) 327–332.
[15] G.S. Li, R.L. Smith, H. Inomata, J. Am. Chem. Soc. 123 (2001)
11,091–11,092.
[23] G. Rothenberg, H.F.M. Boelens, D. Iron, J.A. Westerhuis, Chim.
Oggi 21 (2003) 80–83.
[16] C. Mirodatos, Y. Schuurman, C. Hayaud, A. Holzwarth, D. Farrusseng,
T. Richter, EP 1 293 772 A2, 2003, to CNRS-AMTEC GmbH.
[17] R.K. Grasselli, D.L. Stern, J.G. Tsikoyiannis, Appl. Catal. A 189
(1999) 1–8.
[24] D. Farrusseng, L. Baumes, C. Hayaud, I. Vauthey, P. Denton, C.
Mirodatos, in: E.G. Derouane, E.G. Derouanes (Eds.), NATO Science
Series II: Mathematics, Physics and Chemistry, vol. 69, Kluwer Acad-
emic, Dordrecht, 2001, pp. 101–124.
[25] D. Farrusseng, L. Baumes, C. Mirodatos, in: R.A. Potyrailo, E.J.
Amiss (Eds.), High-Throughput Analysis: A Tool for Combinatorial
Materials Science, Kluwer Academic, Dordrecht, 2003, pp. 551–579.
[26] G. Rothenberg, H.F.M. Boelens, D. Iron, J.A. Westerhuis, Chem. Eur.
J. 9 (2003) 3876–3881.
[18] R.K. Grasselli, D.L. Stern, J.G. Tsikoyiannis, Appl. Catal. A 189
(1999) 9–14.
[19] M. Boudart, in: H. Knözinger, J. Weitkamps (Eds.), Handbook of Het-
erogeneous Catalysis, vol. 3, Wiley–VCH, Weinheim, 1997, p. 958.
[20] C. Descorme, Y. Madier, D. Duprez, J. Catal. 196 (2000) 167.
[21] E. Odier, C. Marquez-Alvarez, L. Pinaeva, Y. Schuurman, C. Millet,
C. Mirodatos, Stud. Surf. Sci. Catal. 136 (2001) 483–488.
[27] A.C. van Veen, D. Farrusseng, M. Rebeilleau, T. Decamp, A. Holz-
warth, Y. Schuurman, C. Mirodatos, J. Catal. 216 (2003) 135–143.