Kinetics, mechanism, and the influence of H2 on the CO oxidation reaction on a Au/TiO2 catalyst

Article abstract of DOI:10.1016/j.jcat.2004.02.036

The kinetics, mechanism, and the influence of H₂ on the CO oxidation reaction on a Au/TiO₂ catalyst were studied. The presence of H₂ had a pronounced effect on the interaction of CO with the Au/TiO₂ catalyst. The presence of H₂ inhibited formate or carbonate formation. A pronounced effect of the H₂ coreactant was observed also in the kinetic measurements. The selectivity for CO oxidation in the H₂-rich atmosphere depended strongly on the CO partial pressure, decaying from 57% at 1.5 kPa and 80°C to 1.5% at 0.03 kPa. Experiments with isotope-labeled gases showed that no O₂ isotope scrambling occurred upon exposure to a mixture of ¹⁶O₂ and ¹⁸O₂, in agreement with previous findings for other Au/TiO₂ catalysts.

Full text of DOI:10.1016/j.jcat.2004.02.036

Journal of Catalysis 224 (2004) 449–462
www.elsevier.com/locate/jcat
Kinetics, mechanism, and the influence of H on the CO oxidation
2
reaction on a Au/TiO catalyst
2
a
a
b
a
a,∗
B. Schumacher, Y. Denkwitz, V. Plzak, M. Kinne, and R.J. Behm
a
Department of Surface Chemistry and Catalysis, University Ulm, D-89069 Ulm, Germany
Center for Solar Energy and Hydrogen Research, Helmholtzstr. 8, D-89081 Ulm, Germany
b
Received 15 September 2003; revised 24 February 2004; accepted 25 February 2004
Available online 20 April 2004
Abstract
We present results of a kinetic and in situ IR spectroscopic study, including isotope-labeling experiments, on the CO oxidation behavior
over a highly active Au/TiO catalyst prepared via a novel synthesis procedure, and on the effect of H on the CO adsorption/reaction
2
2
characteristics under conditions relevant for CO removal from CO-contaminated feed gases for polymer electrolyte fuel cells (PEFCs) via
preferential CO oxidation (PROX). The results show that (i) H affects the CO oxidation, most probably by competing hydrogen adsorption
2
on the Au nanoparticles and reaction with oxygen, which results in a significantly higher CO reaction order, that (ii) formate and carbonate
species formed during the reaction represent (reaction inhibiting) side products, but do not take part in the reaction as reaction intermediate,
at least not under present reaction conditions, and that (iii) the formation of formate and carbonate species is inhibited during reaction in a
H -rich atmosphere. This is tentatively attributed to the decomposition of these species by reaction with H O, which is formed in the process
2
2
gas upon H oxidation.
2

2004 Elsevier Inc. All rights reserved.
Keywords: Kinetic measurements; IR spectroscopy; DRIFTS; Isotope-labeling experiments; CO oxidation; Preferential CO oxidation; Reaction kinetics;
Adsorption; Carbonate formation; Deactivation; Au/TiO catalyst
2
1
. Introduction
prepared via a DP procedure using rather different process
◦
parameters, followed by calcination at 400 C or a reductive
◦
treatment at 200 C, exhibit a similar or even higher activ-
Highly disperse, metal oxide-supported Au catalysts have
ity [13].
been found to be extremely active for various oxidation
and hydrogenation reactions, in particular for CO oxidation
Following previous studies on different oxide-supported
Au catalysts [5,13–18] we investigated the CO oxidation re-
action and in particular the influence of H2 on this reaction
on Au/TiO2 catalysts prepared along this route. In order to
unravel the effect of H2 on the CO oxidation characteris-
tics and on the interaction of CO with the Au/TiO2 catalyst
comparative measurements were carried out in H2-free and
H2-rich gas mixtures. The interest in the effect of H2 arises
from the potential application of these catalysts for the re-
moval of CO from CO-contaminated, H2-rich feed gases
for low-temperature polymer electrolyte fuel cells (PEFCs)
generated by steam reforming and/or partial oxidation of
hydrocarbons or methanol [19–22]. Therefore, the reaction
and adsorption experiments were performed under adsorp-
tion/reaction conditions relevant for processing of H2-rich
feed gases for PEFCs. This includes CO partial pressures up
to 1.5 kPa, together with high H2 contents in the H2-rich at-
[
1–4]. These studies have shown that in addition to the cat-
alyst dispersion, the procedure for catalyst preparation and
conditioning and the nature of the oxide support play an
important role, and it was distinguished between “active,”
easily reducible supports and less active (“inert”), stable ox-
ides [5]. One of the most active support materials is TiO2.
On Au/TiO2 catalysts the reaction was observed even at
temperatures as low as 90 K [6]. Highly active Au/TiO2
catalysts are commonly prepared by coprecipitation (CP) or
by deposition–precipitation (DP), followed by calcination in
◦
air at temperatures between 200 and 400 C [7–12]. Recent
work in our group has shown that supported Au catalysts
*
Corresponding author.
E-mail address: juergen.behm@chemie.uni-ulm.de (R.J. Behm).
0021-9517/$ – see front matter  2004 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcat.2004.02.036

450
B. Schumacher et al. / Journal of Catalysis 224 (2004) 449–462
mosphere, a low O2 excess, and adsorption/reaction temper-
atures around 80 C, which is a typical working temperature
of PEFCs.
Because of the widely varying results on the catalytic
activity of catalysts depending on the catalyst preparation
and pretreatment procedures we are particularly interested
in comparing our results with existing data obtained on these
catalysts prepared via different routes.
over the catalyst, either in a dilute O2 atmosphere or in a
CO-containing atmosphere, during CO oxidation. A previ-
ous study using a 3 wt% Au/TiO2 catalyst prepared in a
different way, by supporting a Au–phosphine complex on ti-
tanium hydroxide, had not shown any isotope scrambling of
O2, neither during interaction of a mixture of 16O2 and O2
with that catalyst nor during reaction with CO [29], which
◦
18
−
supports the assumption of O2 as a reactive species during
A first report on the preparation, conditioning, and stabil-
ity of these Au/TiO2 catalysts was published recently in [23].
In the present paper we will focus on three different aspects:
CO oxidation, indicative of molecular O2 adsorption for that
type of catalyst. The ratio of the different CO2 isotopomers
formed by reaction of C O with O2 and O2 allows fur-
ther conclusions on the reaction mechanism.
1
6
16
18
(
i) on the kinetics of the CO oxidation reaction on these
catalysts, and on the influence of H2 on the reaction
kinetics,
2. Experimental
(
ii) on the interaction of these catalysts with CO, both in
an inert gas mixture and under reaction conditions in a
H2-free as well as in a H2-rich atmosphere, and
2.1. Preparation and pretreatment of the catalyst
(
iii) on questions regarding the reaction mechanism, and the
influence of H2 on the reaction mechanism.
Three different catalysts were used in this study. They
were prepared in the same way but with different Au load-
ings (Au/TiO2(A) and Au/TiO2(C), 2.4 wt% Au; catalyst
Au/TiO2(B), 4.5 wt% Au), via a deposition precipitation
(DP) procedure which was reported in detail elsewhere [23].
In short, an aqueous solution of HAuCl4 was added to a
suspension of annealed TiO2 powder (Anatase, Sachtleben
In the first part of the results (Section 3.1) we report and
discuss kinetic measurements, focusing on the influence of
the CO and O2 partial pressures under present reaction con-
ditions, and on the effect of H2 on the reaction kinetics.
Despite the numerous studies on the activity of differently
prepared (and pretreated) Au/TiO2 catalysts for CO oxida-
tion (see [3,4,10] and references therein), more detailed cat-
alyst kinetic studies, performed under differential reaction
conditions, are scarce so far. This is particularly true for re-
action conditions relevant for processing PEFC feed gases
as described above [8,24–26]. The influence of H2 on the
CO oxidation kinetics had been investigated so far only in
conversion measurements, performed on a 1 wt% Au/TiO2
catalyst [27].
The following Section 3.2 concentrates on the interac-
tion of CO with the Au/TiO2 catalyst in H2-free and H2-rich
atmospheres (see above), in the presence of O2 (“reactive at-
mosphere”) and in the absence of O2 (“inert atmosphere”),
which was studied by IR. This topic had been investigated
already in numerous previous IR studies, in particular by the
group of Boccuzzi et al., but also by other groups (see [28]
and the above reviews), using Au/TiO2 catalysts prepared
via different routes and both inert and reactive (O2 contain-
ing) atmospheres. These studies unanimously concluded that
CO is adsorbed on the Au nanoparticles and, if investigated,
that carbonate-like and carboxylic species are formed on the
surface. The influence of H2 on the interaction between the
CO and the Au/TiO2 catalyst has not been investigated so far.
In order to elucidate the effect of the catalyst-conditioning
procedure, IR measurements were performed both on cal-
cined catalysts and on reductively treated catalysts.
◦
VP9413/3, annealed in air at about 700 C for 30 min, TiO2
particle sizes 21–25 nm). The resulting precipitate was fil-
tered, carefully washed to remove residual Na and Cl [30],
and then dried over night at room temperature under vac-
uum. Prior to the experiments the samples were conditioned
either by the conventional procedure, by calcining in 10%
◦
O2 in N2 at 400 C for 30 min [1,8], or via a reductive condi-
tioning process, involving 45 min annealing in 10% H2 in N2
◦
at 200 C [23], and then cooled down in N2 to the respective
reaction/adsorption temperature. After these treatments the
Au particle sizes were around 2.5–3 nm after calcination and
about 1.8 nm after reductive treatment, respectively, [23].
2.2. Kinetic measurements
Kinetic measurements were performed at atmospheric
pressure in a quartz tube micro reactor (i.d. 4 mm) located
in a ceramic tube oven and under differential reaction condi-
tions, with typically 100 mg catalyst powder (catalyst bed
length 5–8 mm). In order to limit the conversion to val-
ues typically between 5 and 20%, in some cases for the
lowest concentrations also slightly higher values, the cata-
lyst samples were diluted with α-Al2O3, which is not active
for CO oxidation under the present conditions. Reactions
were carried out with a flux of 60 N ml/min (space veloc-
−1
◦
ity 41,000 h ) at 80 C in two different atmospheres, in a
H2-free atmosphere with different amounts of CO and O2
and N2 as balance, and in a H2-rich atmosphere with dif-
ferent amounts of CO and O2, 75 kPa H2, and again N2 as
balance. Kinetic data were acquired after 85 min reaction
time. Influent and effluent gases were analyzed by on-line
The last Section 3.3 aims at mechanistic aspects for
CO oxidation on these Au/TiO2 catalysts, using isotope-
exchange experiments. Here it was tested whether oxygen
exchange is possible with these catalysts when flowing O2

B. Schumacher et al. / Journal of Catalysis 224 (2004) 449–462
451
gas chromatography (Dani GC 86.10HT). High-purity reac-
Table 1
Reaction orders for CO oxidation over catalysts Au/TiO (A) and Au/
TiO (B) (see Fig. 1)
2
tion gases (CO 4.7, O2 5.0, H2 5.0, N2 6.0 from Westphalen)
were used. Evaluation of the Weisz criterion showed the ab-
sence of mass-transport-related problems [31]. For further
details, in particular on the determination of activities and
selectivities, see [32].
2
Catalyst
AuTi(A)
H2 rich
AuTi(B)
H2 rich
Reaction atmosphere
H2 free
α
0.82 ± 0.16
0.36 ± 0.02
0.34 ± 0.07
0.32 ± 0.03
1.01 ± 0.15
0.39 ± 0.02
CO
α
O
2
2
.3. Infrared investigations
sure, pCO = pO , which is equivalent to an oxygen excess
In situ IR investigations were performed in a DRIFTS
diffuse reflectance infrared spectroscopy) configuration
2
of λ = 2. The variation of the normalized CO oxidation
rate (normalized to 1 g of Au) is illustrated for two differ-
ent catalysts (Au/TiO2(A) and Au/TiO2(B)) in Fig. 1a. TOF
numbers can be derived from these rates via the catalyst dis-
(
with a Magna 560 spectrometer from Nicolet, equipped
with a MCT narrow band detector and a commercial in situ
reaction cell unit from Harricks (HV-DR2). This setup al-
lows measurements in a continuous flow of gas mixtures,
equal to those used in the kinetic studies, and at elevated
temperatures. Typically 400–600 scans (acquisition time
persion (Au/TiO2(A), 31.6% dispersion, and Au/TiO2(B),
◦
3
5.4% dispersion after calcination at 400 C); in the present
−3
−1
case a rate of 10 mol gAu s corresponds to TOF num-
−1
bers of 0.62 for catalyst Au/TiO (A) and 0.56 for catalyst
2
∼
4 min) at a nominal resolution of 8 cm were added for
Au/TiO2(B), respectively. In both reaction atmospheres the
one spectrum. Intensities were evaluated in Kubelka–Munk
units, which are linearly related to the adsorbate concen-
tration [33]. Prior to the experiments background spectra
were recorded on the freshly conditioned catalyst at reac-
tion temperature under a flow of pure nitrogen, which were
used for normalization. Finally, spectral contributions of gas
phase CO were removed by subtracting a reference spectrum
recorded over an inert carrier under similar conditions. For
further details see [15].
reaction rates decay continuously with decreasing CO par-
tial pressure. (Note that because of the constant λ in these
measurements the O2 partial pressure is reduced in the same
way.) As seen in the double logarithmic plot in Fig. 1a, the
logarithmic reaction rates are proportional to the logarithmic
CO partial pressure over the entire pressure range investi-
gated (0.03–2 kPa), indicating a constant reaction order in
a power-law description of the reaction rates over the entire
partial pressure range. In the present case, with constant λ,
the slope of the resulting lines yields the overall reaction
2
.4. Experiments with isotope-labeled gases
λ
CO
order α , which in this description is related to the reac-
tion orders for CO and O2, αCO and αO , respectively, by
These experiments were performed in a plug-flow reactor
similar to that used for the kinetic measurements, which was
connected to an ion-molecule-reaction mass spectrometer
2
λ
α
= αCO + αO [32].
CO
2
The two data sets for the reaction in H2-free atmosphere
(
open symbols) were obtained for the same catalyst (Au/Ti-
(
IMR-MS) for educt/product detection (Atomica IMR-MS
O2(A)), but in a different sequence of measurements, one
with increasing partial pressure (P), the other with decreas-
ing partial pressures (e). The data sets obtained in H2-rich
atmosphere (filled symbols) represent results obtained on
two different catalysts, Au/TiO2(A) (a) and Au/TiO2(B)
SP89). This method, which relies on the ionization of re-
+
+
action gases by slow noble gas ions (Xe or Kr ), is less
susceptible to fragmentation than the commonly used mass
spectrometers involving electron-induced ionization [34]. In
addition, gases with ionization energies larger than the elec-
+
+
(
"). The respective reaction orders are collected in Table 1.
Comparing the results obtained in the absence of H2
open symbols) with those measured in a H2-rich atmosphere
filled symbols), it is obvious that the pressure dependence
tron affinity of Xe or Kr , such as N2 and H2, cannot be
ionized, which allows quantitative detection of CO in the
presence of N2. Important for the present studies is that be-
cause of the low fragmentation probability the quantitative
detection of CO is possible also in the presence of CO2 with-
(
(
is much more pronounced for CO oxidation in the H2-
λ
1
8
18
rich atmosphere, with a slope of α = 1.21 (filled points)
out being affected by fragmentation of CO2. C O and O2
CO
1
8
λ
CO
were used as dilute mixtures (1.25% C O in N2 (6.0) and
.25% 18O2 in N2 (6.0), Linde), in some experiments they
were mixed with nonlabeled gases.
compared to α = 0.66 in a H2-free gas mixture. For
1
pCO < 1 kPa the CO oxidation rate is generally lower in the
H2-rich atmosphere than in the absence of H2. The differ-
ence, however, decreases with increasing partial pressures.
The reaction order αCO was determined by keeping the
O2 partial pressure constant at 0.1 kPa and varying the par-
tial pressure of CO (Fig. 1b). Also in this case the data lie on
a straight line, both for reaction in a H2-free (open symbols)
and in a H2-rich atmosphere (filled symbols). The slope is
again considerably higher in the latter case, and the rates
coincide at the highest pressures (1–2 kPa). The absolute
3
. Results and discussion
3
.1. Reaction kinetics
We first studied the influence of the CO partial pressure
at a constant ratio of CO partial pressure to O2 partial pres-

452
B. Schumacher et al. / Journal of Catalysis 224 (2004) 449–462
Fig. 1. Determination of the reaction orders (overall reaction order αλ , reaction order for CO α , reaction order for O
α
) and the selectivity for
O2
CO
CO
2
CO oxidation in H -free and in H -rich atmospheres for two different catalysts, Au/TiO (A) and Au/TiO (B) (diluted catalyst, oxidative pretreatment,
2
2
2
2
◦
λ
CO
8
0 C reaction temperature). If not mentioned differently, the measurements were performed from high to low pressures. (a) Overall reaction order α
for
λ
λ
CO
λ = 2(p = p ): Au/TiO (A) in H -free atmosphere measured from low to high (P, α = 0.64) and from high to low partial pressures (e, α = 0.69);
CO
O
2
2
CO
2
λ
CO
λ
CO
AuTiO (A) in H -rich atmosphere (a, α
= 1.21); Au/TiO (B) in H -rich atmosphere (", α
= 1.40). (b) CO reaction order α
(p = 0.1 kPa,
2
2
2
2
CO
O
2
p
variable): Au/TiO (A) in H -free atmosphere, measured from low to high CO partial pressures (P, α = 0.32) and measured from high to low partial
CO
2
2
CO
pressures (e, α = 0.37); Au/TiO (A) in H -rich atmosphere (a, α = 0.78). (c) O reaction order α (p = 1 kPa, p variable): Au/TiO (A) in
CO
2
2
CO
2
O
CO
O
2
2
2
H -free atmosphere (e, α = 0.32); Au/TiO (A) in H -rich atmosphere (a, α = 0.36); Au/TiO (B) in H -rich atmosphere (", α = 0.39). (d) H
2
O
2
2
O
2
2
O
2
2
2
2
λ
CO
λ
CO
oxidation activity during the above reactions in H -rich atmosphere for catalysts Au/TiO (A) and Au/TiO (B): Au/TiO (A), α (!); Au/TiO (B), α (");
2
2
2
2
2
Au/TiO (A), α
(e), Au/TiO (A), α (1), Au/TiO (B), α (2).
2 O 2 O
2 2
2
CO
values agree well with those in Fig. 1a for comparable con-
(pCO = pO = 1 kPa). In this case, however, the reaction
2
ditions (pCO = pO = 0.1 kPa). The respective values for the
order αO is not sensitive to the reaction atmosphere, with
2
2
reaction order αCO are summarized in Table 1.
αO = 0.32 in H2-free atmosphere and αO = 0.36 in H2-
2
2
Next, the reaction order αO was determined by keeping
rich atmosphere (catalyst Au/TiO2(A)). As expected for the
relatively high CO partial pressures the effect of the reaction
atmosphere is rather small, though the rates are still higher
2
the CO partial pressure constant at 1 kPa and varying the par-
tial pressure of O2 (Fig. 1c). Again the data fall on a straight
line, both for reaction in a H2-free (open symbols) and in
a H2-rich atmosphere (filled symbols). The absolute values
agree well with those in Fig. 1a for comparable conditions
in the absence of H2 than in the presence of H2.
λ
CO
Using the reaction orders α
and αO , we also deter-
2
λ
mined αCO via the relation α = αCO + αO . The resulting
CO
2

B. Schumacher et al. / Journal of Catalysis 224 (2004) 449–462
453
Table 2
Kinetic parameters for CO oxidation over supported Au/TiO catalysts and relevant reaction parameters
2
∗
Catalyst loading
wt%)
Preparation
Catalyst
condition
d
α
p
; p
α
p
, p
O
2
T
E
Ref.
Au
CO
CO
O
O
CO
2
2
(
(nm)
(kPa)
(kPa)
(K)
(kJ/mol)
H -free atmosphere
2
3
2
1
2
1
1
2
.3
.3
DP
IM
IM
IM
IM
IM
DP
C
3
30
30
30
0.05
0.5–20; 20
2–25; 5
n.r.; 5
0.5–2; 4
n.r.
3.67–16.5; 3.67
0.02–2; 0.1
0.24
1; 0.7–20
5; 2–19
5; n.r.
0.3–2, 4
n.r.
273
34
9.6
29
[8]
a
b
HTR/C/LTR
HTR/C/LTR
HTR, C
C
0.24
0.4
313–333
273–313
320
[24]
[25]
[61]
[29]
[27]
c
c
∼ 0.4
∼ 0
0.45
0.19
32
0.25
0.2
0.34
0.41
0.46
0.32
n.r.
16.3
16.3
–
3.67–17.0, 3.67 293
1; 0.2–3
.4
LTR,C
C
2.7
2.7
353
Present work
H -rich atmosphere
2
2.4
DP
0.82
0.02–2; 0.1
0.36
1; 0.1–3
353
25
Present work
Abbreviations: DP, deposition–precipitation; IM, impregnation; LTR, low-temperature reduction; HTR, high-temperature reduction; C, calcination.
a
0
α
.6 at 353 K, 0.5 at 313 K, and lower pressures.
b
c
may be somewhat lower at lower pressures.
O
2
The difference in reaction orders between [24] and [25] was explained by these authors by changes in the 2.3% Au/TiO catalyst due to stronger
2
deactivation before the data were collected [25].
values for αCO (see Table 1) are very close to those deter-
mined directly from the data in Fig. 1b, at very different
O2 partial pressures. This indicates that there are no major
changes in the reaction mechanism within the pressure range
used in these experiments.
(CO:O2 ratio, absolute partial pressures, reaction tempera-
ture), as, e.g., in the case of the data in Ref. [8]. Furthermore,
regardless of the catalyst preparation, etc., the reaction ap-
pears to be limited by the respective reactants and not by a
lack of reactive sites, at least under reaction conditions con-
sidered in these studies. The reaction orders determined here
for CO oxidation in a H2-free atmosphere on a DP-prepared
catalyst fit well to the previously published data. The same
holds true for the absolute rates and TOF numbers obtained
here, which agree well with those obtained for highly active
Au/TiO2 catalysts prepared via conventional procedures un-
der comparable conditions [27].
The results obtained in a H2-rich atmosphere differ sig-
nificantly. Here the CO reaction order is more than twice
as high as in the absence of H2, while that for O2 is lit-
tle affected by the presence of H2. An earlier study of the
same reaction over Au/Fe2O3 in our laboratory, which was
performed under similar reaction conditions, gave a con-
siderably lower value for the CO reaction order, αCO =
0.55, while the O2 reaction order was of similar magni-
For comparison of the reaction orders determined here in
a H2-free atmosphere values determined in previous kinetic
studies as well as the relevant reaction conditions are col-
◦
lected in Table 2. For reaction at 0 C on a 3.3 wt% Au/TiO2
catalyst Haruta et al. derived a CO reaction order near zero
(
αCO = 0.05), an O2 reaction order of αO = 0.24 and an ac-
2
tivation energy of 34 kJ/mol [8]. Vannice and co-workers
reported reaction orders of αCO = 0.4 for reaction between
◦
◦
0
and 20 C, which increased to αCO = 0.56 at 40 C, and
◦
an O2 reaction order of αO ≈ 0 for reaction at 0–20 C
2
◦
(
αO ≈ 0.1 at 40 C) [25]. It should be noted that the cat-
2
alysts used in the latter study, which were prepared via an
impregnation procedure, exhibited much larger Au particles
(
particle size 30 nm) than the catalysts used by the Haruta
group (2.5–5 nm, as visible on the TEM image in that pa-
per), which were prepared via a DP procedure worked out
by this group. For reaction on a 2% Au/TiO2 catalyst pre-
pared by an incipient wetness technique Ossipoff and Cant
tude (αO = 0.27) [35]. Additional kinetic data are rare and
2
do not contain information on details such as reaction or-
ders. Torres Sanchez et al. had shown that CO can easily
be oxidized over a Au/MnOx catalyst in a H2 feed stream
determined reaction orders of αCO = 0.45 and αO = 0.19
2
◦
under comparable conditions as well as an activation energy
of 32 kJ/mol [26]. Using a different procedure for catalyst
preparation, by grafting a Au–phosphine complex onto TiO2
support material, Choudhary et al. derived reaction orders of
(pCO = pO = 1 kPa, rest H2) at temperatures of 50–100 C,
2
but that the CO conversion decreases slightly at higher tem-
peratures (120 C) [36]. This was attributed to increasing O2
◦
consumption by H2 oxidation [36]. A reduction in the CO
oxidation activity over Au/TiO2 in the presence of H2 by
about one-third, from 0.9 to 0.6 mmol/s gAu, can be cal-
culated from the data in [27], which were obtained under
◦
αCO = 0.2 and αO = 0.46 at 20 C for a 1 wt% Au/TiO2
2
catalyst, and an activation energy of 16.3 kJ/mol [27]. Al-
most identical values were reported also by Liu et al. for a
similar type catalyst [29].
◦
not very different reaction conditions (80 C, pCO ≈ 2 kPa,
The main results of this comparison are that (i) there is
a considerable spread in the data and that (ii) in all studies
both the CO and the O2 reaction orders are slightly positive,
between 0.05 and 0.46. At least part of the spread in the data
can be explained by the very different reaction conditions
pH = 48 kPa, λ = 4). A comparison of the PROX activity of
2
different metal oxide-supported Au catalysts (Au/γ -Al2O3,
Au/Co3O4, Au/CeO2, Au/Fe2O3, Au/Ni2O3, Au/Mg(OH)2,
Au/TiO2, etc.) at a single gas composition (pCO = pO =
2
1 kPa) by Schubert et al. [13] revealed a considerable vari-

454
B. Schumacher et al. / Journal of Catalysis 224 (2004) 449–462
◦
◦
Fig. 2. (a) Selectivity for CO oxidation during the reaction in H -rich atmosphere for catalyst Au/TiO (A) at different temperatures (40 C ("), 60 C (2),
2
2
◦
λ
8
0 C (Q), λ = 2). (b) Selectivity during the reactions in Fig. 1 in H -rich atmosphere for catalysts Au/TiO (A) and Au/TiO (B): Au/TiO (A) α
(");
2
2
2
2
CO
λ
Au/TiO (B) α (!); Au/TiO (A) α
(Q); Au/TiO (A) α (2); Au/TiO (B) α (1) (experimental conditions as in Fig. 1).
2 O 2 O
2 2
2
CO
2
CO
ation in activity and selectivity. Likewise, Grisel et al. had
reported that the presence of H2 in the reactant flow had
a slightly negative effect on the CO oxidation activity at
room temperature over Au/MnOx/Al2O3, but a considerable
positive effect for Au/MgO/Al2O3 (pH :pCO:pO = 4:2:1,
clusions that (i) the reactive sites for H2 oxidation are not
or at least not to a significant extent blocked by the increas-
ing CO coverage and/or that (ii) the reactive oxygen species
are not significantly depleted by the CO oxidation reaction.
The first conclusion holds true if H2 activation/dissociation
is rate limiting, the second one if O2 activation/dissociation
is rate determining.
2
2
atmospheric pressure) [37].
Next, Figs. 1d and 2b illustrate the behavior of the activ-
ity for H2 oxidation (Fig. 1d) and the selectivity (Fig. 2b),
respectively, in the measurements in a H2-rich atmosphere.
The selectivity is defined by the amount of O2 consumed for
CO oxidation, divided by the total amount of O2 consumed,
and calculated from the balance between CO2 formation and
O2 consumption [32]. (It should be noted that this is only
correct as long as the H2 level is not affected by any other
reactions such as the water-gas-shift (WGS) reaction (CO
Finally, we estimated the activation energies for CO oxi-
dation under these conditions from the kinetic measurements
◦
performed at 40, 60, and 80 C at different CO partial pres-
sures and constant λ = 2 in H2-rich gas (pH = 75 kPa). For
2
CO partial pressures between 0.3 and 2 kPa we obtained val-
ues of 25 ± 8 kJ/mol. At lower CO partial pressures, 0.1
and 0.03 kPa, the activation energy decayed to 7 ± 6 and
0
± 6 kJ/mol, respectively. The higher value agrees well with
those reported previously for highly active Au/TiO2 cata-
lysts and comparable reaction conditions [8,25,29]; for the
low CO partial pressures reference values do not exist. The
selectivities obtained in these measurements at different CO
+
H2O ↔ CO2 + H2) or the methanation reaction (CO
+
3H2 ↔ CH4 + H2O), which appears justified based on
previous measurements of the forward WGS reaction on a
Au/TiO2 catalyst under comparable conditions by Sakurai
et al. [38].) As can be seen in Fig. 2b the selectivity de-
pends strongly on the CO partial pressure, decaying from
almost 70% at pCO ꢀ 1 kPa to a few percent at CO partial
pressures < 0.1 kPa. The absolute values of the selectivity
fit well to data reported by Choudhary et al. [27], who de-
termined a selectivity of 73% during selective CO oxidation
over Au/TiO2 in a CO:H2:O2 mixture of 1:50:2, i.e., at λ = 4
◦
partial pressures are plotted in Fig. 2a. (The values for 80 C
are included also in Fig. 2b.) While there is not much differ-
◦
ence for reaction between 80 and 60 C, the selectivities are
◦
significantly higher at 40 C, reaching 64% at 1 kPa CO (50
and 40% at 60 and 80 C, respectively). Also at these lower
◦
temperatures the change in selectivity is mostly caused by
the change in CO oxidation, while the H2 oxidation rate is
much less affected.
◦
and pCO ≈ 2 kPa at 80 C reaction temperature. A strong
dependence of the selectivity on the CO partial pressure at
constant λ was reported also for the PROX reaction over
Au/Fe2O3 [35]. Interestingly, in both cases the pronounced
changes in selectivity with CO partial pressure are mainly
due to the change in CO oxidation activity, while the H2
oxidation rate is little affected. These results lead to the con-
3
.2. Infrared investigations
The kinetic studies have shown an increase in CO ox-
idation rate for higher CO partial pressures at constant λ.
In order to deduce whether this increase of the rate is ac-
companied by a similar increase in COad coverage or other

B. Schumacher et al. / Journal of Catalysis 224 (2004) 449–462
455
Fig. 3. IR spectra recorded upon adsorption of CO from H -free gas mixtures containing increasing amounts of CO in N on the catalyst Au/TiO (A)
2
2
2
(
undiluted catalyst; reductive conditioning). (a) Complete spectra, (b) enlarged detail spectra of the spectral range characteristic for OH groups (on the
support), (c) enlarged detail spectra of the formate region, and (d) enlarged detail spectra of the carbonate region. From bottom to top (inset (c) top to bottom):
.05 kPa CO, 0.1 kPa CO, 0.5 kPa CO, 1.0 kPa CO, 1.5 kPa CO.
0
changes in the adlayer, we investigated the respective adlay-
ers resulting under different conditions by in situ DRIFTS.
This was done in both H2-free and H2-rich dilute CO at-
mospheres, respectively, and in O2-containing reactive gas
tion spectra. Therefore, the background subtraction had to be
done separately for two different spectral regions. Accord-
ingly, differences between different background-corrected
spectra are not caused by the initial TiO support modifi-
2
◦
mixtures at 80 C (Figs. 3, 5, 6, 7). Prior to the DRIFTS
cation, but reflect real changes in the adlayer. The spectra
measurements the catalysts were conditioned either via the
reductive treatment or via calcination as described in Sec-
tion 2, and then cooled down in a flow of N2. It turned
out, however, that the characteristic features and trends in
the DRIFT spectra did not depend on the pretreatment. The
only exception was a weak peak in the carbonate region,
at 1445 cm , which appeared during CO adsorption in a
H2-rich atmosphere on a calcined catalyst, but not on a re-
ductively pretreated Au/TiO2 catalyst. The spectra presented
in Figs. 3 and 5 were acquired after a reductive pretreatment,
those in Figs. 6 and 7 after oxidative pretreatment.
obtained upon adsorption from CO/N mixtures are domi-
nated by the absorption peak of adsorbed CO. As expected
2
for CO adsorption on Au nanoparticles; i.e., in agreement
with previous observations for CO adsorption on other sup-
ported Au catalysts [15,25,27,39,40]and on massive Au sur-
faces [41,42], only linearly adsorbed CO is observed. The
vibrational frequency is almost independent of the CO par-
tial pressure (Fig. 3a, from bottom to top), with only a slight
−1
−
1
−1
red shift from 2112 cm at 0.05 kPa CO to 2108 cm at
1.5 kPa CO partial pressure. The increase in intensity with
increasing CO partial pressure and the frequency shift with
increasing C–O absorption intensity are plotted in Fig. 4.
A similar behavior of the C–O vibration, with a frequency
3
.2.1. CO adsorption in H2-free atmosphere
−1
Spectra recorded in a dilute CO atmosphere, with differ-
shift from 2111 to 2104 cm
upon a CO partial pres-
ent amounts of CO in N2, are presented in Fig. 3. Due to an
initial change of the reflectivity after the admission of CO
sure increase from 0.05 to 1.0 kPa, was observed also for
CO adsorption on Au/Fe2O3 [15] and on massive Au, from
−1
−4
−1
(
color changes due to partial reduction of the TiO2 support)
2129 cm at 1.33 × 10 kPa CO to 2106 cm at at-
the background spectra did not fit perfectly the CO adsorp-
mospheric pressure [41]. This chemical shift, which is op-

456
B. Schumacher et al. / Journal of Catalysis 224 (2004) 449–462
carboxylate ions are oxidized by a superoxide ion. This ex-
planation agrees with the observation that the intensity of the
carbonate/carboxylate peaks does not increase with CO par-
tial pressure. It should be noted that adsorbed OH species
◦
are present on TiO2 even after calcination at 500 C [49].
The proposed interaction of CO with OHad on peripheral
sites agrees well with the evolution of the OH peak with
increasing CO partial pressure (enlarged detail spectra see
Fig. 3b). The negative peak in the spectra of the O–H stretch
−1
frequency at 3677 cm fits excellently to OH adsorbed on
pure TiO2, which was reported to have a characteristic ab-
−1
sorption band at 3676 cm [50]. The negative peak in the
spectra indicates a decreasing amount of these species in a
Fig. 4. (a) Frequency shift of the C–O stretch vibration with increasing CO
partial pressure upon CO adsorption from H -free gas mixtures contain-
2
ing increasing amounts of CO in N (see Fig. 3). (b) Frequency shift with
2
increasing peak intensity (data from part (a)).
2 2
CO/N atmosphere compared to that in pure N . A chemical
interaction or even reaction with adsorbed CO changes the
chemical state of the OH group and therefore also the fre-
quency of the O–H stretch vibration, resulting in a positive
peak at a somewhat different frequency. The red shift of the
new vibration points to a weakening of the O–H bond in the
new OHad species due to interaction/reaction with COad.
The intensity of the positive O–H peak does not change
with the CO partial pressure in the gas mixture, which fits
well to the constant intensity of the carboxylate species. In
contrast, the intensity of the negative peak increases with
higher CO concentrations, indicating that the consumption
of the OH groups is caused not only by the formation of
the carbonate/carboxylate species, but also via another reac-
tion path. This is confirmed in the detail spectra in Fig. 3c,
which show the continuous growth of characteristic peaks
posite to the behavior generally observed on transition and
noble metal surfaces [43], had been explained by the lack
of backdonation from Au d-electrons into the 2π orbital of
∗
adsorbed CO, so that the Au–CO bond is dominated by elec-
tron donation from the 5σ orbital of the CO molecule, which
is slightly antibonding for the C–O bond, to the metal [41].
With increasing CO coverage this effect becomes weaker,
leading to a weakening of the C–O bond. In contrast to
Bollinger and Vannice we did not see a second peak at 2071
−1
cm emerge, which they observed during room tempera-
ture CO adsorption (pCO = 5 kPa) [25], or of a peak at 2183
−
1
cm as reported by Haruta et al., which was attributed to
4
+
CO adsorption on Ti sites.
−1
In addition to the C–O vibration signal the spectra ex-
hibit weak, but characteristic, absorption peaks also in the
in the formate region, at 2833, 2873, 2931, and 2955 cm
[25,44]. These peaks had been assigned to C–H stretch vi-
brations in formates adsorbed on the anatase support [25].
In addition, starting at partial pressures of 0.5 kPa a new
−
1
spectral ranges 1350–1700 cm (carbonate, carboxylate,
−1
and formate species), 2800–3000 cm (C–H vibration of
−
1
−1
formates), and around 3630 cm (O–H vibrations) [8,25,
0,44]. In the “carbonate/carboxylate region” (detail spec-
tra see Fig. 3d), characteristic absorption lines are visible
peak appears at 2972 cm , whose origin has not yet been
4
clarified. This feature grows continuously with CO partial
pressure (and adsorption time), in good agreement with the
−1
at 1672, 1552, 1381, and 1363 cm . (After oxidative pre-
treatment an additional weak peak at 1445 cm appears
at all CO partial pressures.) In previous studies peaks at
continuous consumption of OH species indicated by the be-
−1
−1
havior of the negative peak at 3677 cm
.
−1
1
690 and 1585 cm had been assigned to bidentate species
2
3.2.2. CO adsorption in H -rich atmosphere
on the support [25,45]. Liao et al. distinguished between
bridging carbonate (bound to two Ti ions) with frequencies
For CO adsorption from a CO/H atmosphere the results
are very different. Since the DRIFT spectra showed slow
modifications in their overall shape over a significant time
span, we present spectra recorded at three different CO par-
tial pressures after 2 min (Fig. 5a) and the temporal evolution
2
−1
around 1725 cm on TiO2 and bidentate carbonates (bound
−
1
to a single Ti ion) with frequencies around 1560 cm
[
−1
46]. A peak at 1440 cm was assigned to monodentate
carbonate [40,46] or free carbonate [25,46]. Vibrations be-
of spectra in 0.5 kPa CO in H /N (Fig. 5b). As before, back-
2
2
−1
tween 1330 and 1370 cm
had been assigned to COO
ground removal was performed in the two regions above and
−
1
−1
symmetric stretch frequencies of monodentate (1330 cm
)
below 2500 cm separately. The C–O stretch vibration is
−1
and bidentate (1370 cm ) carbonates as well as formates
slightly red-shifted compared to a H -free atmosphere, from
2
−1
−1
(
1370 cm ) on the TiO2 support [13,25,40,44,46,47].Bond
about 2110 to 2107 cm . This peak increases in intensity,
and Thompson had proposed that these species are formed
but does not shift with CO partial pressure (Fig. 5a). With
increasing adsorption time at 0.5 kPa this peak decreases
by reaction of a CO molecule adsorbed on a Au particle
0
−1
(
“Au –CO”) with a hydroxylgroup adsorbed either on a sup-
and is replaced by an asymmetric peak at 2032 cm , which
III
−1
port cation or on a peripheral ionic Au species, e.g., a Au
ion (“periphery mechanism”) [48]. In their model the hy-
droxyl group is recovered during CO oxidation when two
includes a shoulder around 1985 cm and possibly a sec-
−1
ond one around 1930 cm (Fig. 5b). A comparable peak
at 2055 cm , with a shoulder at 1990 cm , had been re-
−1
−1

B. Schumacher et al. / Journal of Catalysis 224 (2004) 449–462
457
Fig. 5. DRIFT spectra recorded upon adsorption of CO from H -rich gas
2
mixtures containing different amounts of CO in 75 kPa H /rest N (on the
2
2
catalyst Au/TiO (A) (undiluted catalyst; reductive conditioning). (a) Pres-
2
◦
Fig. 6. Sequence of DRIFT spectra recorded during CO oxidation at 80 C
sure dependence (bottom to top): 0.05 kPa CO, 0.1 kPa CO, 0.5 kPa CO;
spectra taken after 2 min. (b) Complete spectra, and (c) (detail spectra be-
over Au/TiO (A) in H free atmosphere (1 kPa CO, 1 kPa O , rest N , cat-
alyst diluted 1:5 with α-Al2O3, calcined at 400 C). (a) Complete spectra,
b), (c), and (d) enlarged detail spectra of the spectral ranges characteristic
for O–H vibrations (b), C–H vibrations in formates (c), and COO vibrations
in formates and carbonates (d), respectively. The spectra were recorded af-
ter (from bottom to top) 5, 10, 30, 60, 120, and 240 min reaction time.
2
2
2
2
◦
−1
tween 3770 and 3480 cm ) show the temporal development of the spectra
during adsorption from 0.5 kPa CO (from bottom to top in (b) and in (c) at
(
− −1
1
3
675 cm , from top to bottom at in (c) at 3570 cm : 2, 7, 30, 60, 120,
and 240 min).
ported by Boccuzzi et al. for CO adsorption on a Au/TiO2
◦
a continuous shift in the C–O frequency and not a new peak
at a fixed frequency, as it is observed experimentally. The
latter fits better to a distinctly different new COad species
as expected, e.g., for a compound like H–Au–CO adsorp-
tion complex, similar to reports for H2/CO coadsorption on
Ni(100) [53,54], or for a severe, H2-induced restructuring of
the Au nanoparticles. Equally, Bollinger et al. report a weak
signal at 2062 cm upon CO adsorption on a Au/TiO2 cat-
alyst pretreated by low-temperature reduction as a last step,
which disappears again upon O2 addition [25]. Finally, we
note that neither formates nor carbonate/carboxylate species
are formed in the presence of H2.
sample that was reduced with H2 at 255 C before adsorp-
−1
tion, in addition to the “normal” peak at 2100 cm [51].
On the other hand, this new feature was not observed for CO
adsorption from a similar H2-rich atmosphere on Au/Fe2O3
[
17]. Boccuzzi et al. explained their findings by a reduc-
tion of the support due to reaction with H2 and a result-
ing transfer of negative charge to the gold particles, which
leads to an increased backdonation and hence to a weak-
ening of the C–O bond, in agreement with the observed
peak shift. A charge transfer from TiO2 to Au nanoparti-
cles has indeed been reported by Jakob et al. [52] when
exposing TiO2 nanoparticles to UV radiation and bringing
the resulting, partly reduced, colored TiO2 nanoparticles into
contact with Au nanoparticles. The detail spectra in the inset
−1
3.2.3. CO oxidation in a H2-free atmosphere
−1
(
Fig. 5c) show a significant rise of the OH peak at 3775 cm
The evolution of adsorbed species with increasing re-
action time during CO oxidation in a H2-free atmosphere
(1 kPa CO, 1 kPa O2, rest N2) is illustrated in Fig. 6. In this
case background removal could be done for the complete
spectra. The spectra show linearly adsorbed CO with a fre-
with increasing adsorption time, indicating that TiO2 reduc-
tion takes place under present conditions. Other possible
explanations for the shift in C–O frequency include interac-
tions between Had and COad species coadsorbed on the Au
nanoparticles, or a Had-induced decrease in the steady-state
CO coverage. In all three cases, however, one would expect
−
1
quency of 2118 cm as a dominant CO peak and a small
−
1
feature at 2173 cm (Fig. 6a). The C–O stretch frequency

458
B. Schumacher et al. / Journal of Catalysis 224 (2004) 449–462
at 2118 cm 1 is blue-shifted by about 8 cm compared to
CO adsorption in the absence of O2. A similar blue shift of
the COad vibration in the presence of O2 was observed by
numerous groups [1,25,27]. It was attributed to a coverage
effect, a reduction in steady-state CO coverage because of
the oxidation reaction. Considering the small COad cover-
age effects observed in the present study, however, we favor
contributions from coadsorbed species, e.g., oxygen species
coadsorbed on the Au particles, as explanation. The second
−
−1
−1
feature, at 2173 cm , was attributed to CO adsorbed on
Ti [6]. In addition, the spectra exhibit signals characteris-
4
+
−1
tic for gas-phase CO2 at 2361 and 2338 cm , and a number
of features in the OH, formate, and carbonate regions, which
will be described in more detail below. With increasing re-
action time the signals of adsorbed CO and CO2 decay con-
tinuously. In contrast to the CO coverage-dependent shifts
observed upon CO adsorption from the inert CO/N2 mix-
ture, the frequency of the C–O vibration remains practically
constant. This behavior can be explained by an increasing
blocking of the CO adsorption sites, i.e., of the Au nanoparti-
cle surface by the growing formate and carbonate adspecies,
which reduces the total amount of adsorbed CO, but leaves
the local coverage unchanged.
In the formate region the spectra exhibit two major peaks
−
1
at ∼ 2875 and 2953 cm (detail spectra see Fig. 6c). The
−1
peak around 2875 cm consists of two hardly resolved
peaks at 2883 and 2870 cm . Furthermore, after 30 min
a small peak at 2742 cm develops. The peaks at 2953
−1
−
1
◦
Fig. 7. Sequence of DRIFT spectra recorded during CO oxidation at 80 C
−1
and 2870 cm are identical to those observed upon CO
adsorption from the CO/N2 atmosphere, while the peak at
over Au/TiO (A) in H -rich idealized reformate (1 kPa CO, 1 kPa O ,
2
2
2
◦
75 kPa H2, rest N2, catalyst diluted 1:5 with α-Al2O3, calcined at 400 C).
(a) Complete spectra, (b), (c), and (d) enlarged detail spectra of the spectral
ranges characteristic for O–H vibrations (b), C–H vibrations in formates (c),
and COO vibrations in formates and carbonates (d), respectively. The spec-
tra were recorded after (from bottom to top) 5, 10, 30, 60, 120, and 240 min
reaction time.
− −1
1
2
742 cm is blue-shifted by 10 cm compared to that in
the latter spectra. Similar as for CO adsorption in a H2-free
atmosphere OH groups adsorbed on the support are removed
by carboxylate/carbonate, as evidenced by the negative peak
−1
at 3676 cm ; i.e., the support is depleted by OH groups
during the reaction.
nice [25], Boccuzzi, Haruta, and co-workers [1,55], or by
Liu et al. [29].
The peaks in the carbonate region (detail spectra Fig. 6d),
−1
between 1350 and 1700 cm , are much more pronounced
and structured than those observed for CO adsorption in
a CO/N2 atmosphere (Fig. 3). The main peaks in this re-
In our preceding report on the deactivation behavior and
long-term stability of this catalyst we had shown that the de-
creasing amount of adsorbed CO is correlated with a signif-
icant decay of the reaction rate [23]. The continuous growth
of the carbonate-like species—the intensities of the related
IR absorption signals are still growing even after 20 h of
reaction time—may be one or perhaps even the dominant
reason for the deactivation of the catalyst. Similar observa-
tions and conclusions were reported also for CO oxidation
in a H -rich atmosphere over Au/Fe O [17,56]. This will
−1
gion are a band between 1500 and 1620 cm , which
exhibits several weak maxima and shoulders. In addition,
−
1
peaks at 1360, 1380, 1440, and ∼ 1700 cm are found.
−
1
As noted before peaks at 1690 and 1585 cm had been
assigned previously to the asymmetric stretch vibration of
bridged/bidentate carbonate species on the support [25],
−1
a peak at 1440 cm
to monodentate carbonate or free
2
2
3
carbonate [25,40,46], and vibrations between 1300 and
be discussed in more detail in the next paragraph.
−1
1
370 cm to the symmetric stretch vibrations of different
carbonates and of formate species [40,44,46]. The antisym-
metric stretch vibration of formates on TiO2, which are
assumed by those authors to be in a bridging configuration,
3.2.4. CO oxidation in a H -rich atmosphere
As in the previous cases the spectral region of the C–O
stretch vibration is dominated by a pronounced peak of
2
−1
−1
was identified at about 1555 cm [46,47].
linearly adsorbed CO, which appears here at 2112 cm
.
Similar effects upon changing to reaction conditions,
in particular a blue shift of the C–O stretch vibration in
the presence of O2, were reported by Bollinger and Van-
Similar to CO oxidation in a H -free atmosphere, it does
not change its position with time (Fig. 7a). Different from
CO adsorption from a H2-rich atmosphere, however, the
2

B. Schumacher et al. / Journal of Catalysis 224 (2004) 449–462
459
−
1
highly asymmetric peak at 2032 cm (Fig. 5b) does not de-
−1
velop. A satellite at 2143 cm likely results from gaseous
CO, which was not perfectly removed by the CO gas-phase
subtraction. With increasing reaction time the CO signal
decreases in intensity (Fig. 7a), though the decay is less
pronounced than that observed in a H2-free atmosphere
(
Fig. 6a).
The other spectral areas are dominated by the pronounced
−1
signals of the oxidation products CO2 (2361 and 2338 cm
)
and H2O, where the latter includes the strong signal of the
−
1
HOH mode of adsorbed water at 1623 cm [40] and the
−
1
O–H vibration between 2500 and 3200 cm . As expected,
the latter is much more pronounced (see Fig. 7a) than during
reaction in a H2-free atmosphere (Fig. 6a). With increasing
reaction time all of these signals lose considerably in in-
tensity. Similar to CO adsorption in a H2-rich atmosphere
(
Fig. 5), formate formation does not seem to play an impor-
−1
tant role. A weak peak developing at 2874 cm (Fig. 7c)
is comparable in intensity to that formed upon oxidation in
H2-free atmosphere.
The carbonate region exhibits weak negative signals at
−
1
∼
1380, 1440, and 1550/1585 cm in addition to the dom-
inant water peak. The negative signals correspond exactly
to the (positive) carbonate signals observed upon reaction
in a H2-free atmosphere. We therefore explain them by a
reaction-induced decomposition of small amounts of car-
bonates which had been present at the beginning of the reac-
tion.
18
18
Fig. 8. Isotope-exchange experiments using
O
and C O. All experi-
2
The decay in IR intensity and the deactivation of the cat-
alyst must be related to the accumulation of water on the
catalyst. A TPD spectrum recorded from a Au/TiO2 catalyst
that had been operating in a H2-rich atmosphere for 900 min
showed H2O as the only desorbing species (spectrum not
shown). We, therefore, assume that a considerable part of
the general decrease in intensity in the spectra in Fig. 7a is
caused by a change in reflectivity with increasing reaction
time. This proposal is supported also by the fact that the de-
crease in intensity is about equal for all peaks.
◦
ments were carried out at a flux of 40 N ml/min at 80 C. (a) Introduction
32
36
of 0.9 kPa
O
and 1.0 kPa
O to the diluted Au/TiO (C) catalyst (di-
2
2 2
32
luted 1:15 with α-Al O , 100.7 mg). (b) CO oxidation when 0.5 kPa
O
2
2
3
36
16
and
O and 1.2 kPa C O are introduced to diluted catalyst Au/TiO (C)
2
2
18
(
1:15 with α-Al O , 100.7 mg). (c) Introduction of C O in N to diluted
2
3
2
catalyst Au/TiO (C) (1:5 with α-Al O , 74.6 mg).
2
2 3
der steady-state conditions. Correspondingly, the H2O signal
dominates the carbonate region in the IR spectra.
3
.3. IMR-MS measurements with isotope-labeled gases
The reaction behavior of the Au/TiO2 catalyst in a H2-
rich atmosphere is very different from that of Au/Fe2O3
catalysts [56]. On the latter the reaction leads to a pro-
nounced growth of carbonate and formate species on the
support, whose signals may overgrow also an underlying
H2O peak. The amount of side products/reaction inter-
mediates quickly reaches a steady-state equilibrium value;
their peak intensities do not increase significantly after
In this section we describe experiments using isotope-
labeled reactants which were performed to get more direct
information on the reaction mechanism on this catalyst. In
order to test whether the interaction of O2 with the catalyst
under reaction conditions results in isotope scrambling, we
16
18
introduced a mixture of O2 and O2 to a dilute Au/TiO2
catalyst (see Fig. 8a). The catalyst had been calcined at
◦
3
0 min [56]. On the other hand, experiments exposing a
4
00 C before the experiment. As shown in Fig. 8a, no
1
6
18
carbonate-covered Au/Fe2O3 catalyst to a H2/O2-containing
atmosphere showed the conversion of carbonates into ther-
mally less stable bicarbonates, which was explained by a
reaction with water formed in that atmosphere [17]. On
Au/TiO2, where the tendency for carbonate formation is
much lower, the H2O-induced conversion and decomposi-
tion of the carbonate species appear to be fast enough to keep
the level of adsorbed carboxylate/carbonate species low un-
O O was formed, neither during the reaction under those
conditions (Fig. 8a) nor during a subsequent desorption
◦
spectrum, when the sample was heated up to 300 C (spec-
trum not shown). This rules out reaction pathways involving
dissociative adsorption and the formation of adsorbed Oad,
which could undergo recombinative desorption, on this cat-
alyst. Our findings are consistent with results from Iwasawa
and co-workers [29,57], who studied the room temperature

460
B. Schumacher et al. / Journal of Catalysis 224 (2004) 449–462
adsorption–desorption behavior of a similar isotope-labeled
O2 gas mixture on a Au/TiO2 catalyst prepared from organic
gold and titanium precursors, and from a study of the tran-
sient O2 adsorption/desorption behavior by Olea et al. [58].
It also agrees with observations in previous studies in on
a Au/Fe2O3 catalyst, where equally no isotope scrambling,
i.e., no 16O O formation, was observed [5,59].
thors had also shown, however, that CO2 can exchange oxy-
gen with the Au/TiO2 catalyst. Therefore, the observation of
1
8
C O2 cannot be taken as proof for a reaction mechanism
proceeding via a carbonate intermediate, where the latter is
formed by reaction between CO and an (activated) molecular
oxygen species. In the latter case and without oxygen ex-
change between catalyst and CO2 the expected isotope dis-
tribution would have been 3:2:1, provided that the decompo-
sition of the CO3 species occurs randomly with similar prob-
18
The evaluation of the temporal evolution of the O2 sig-
nals, after switching from the bypass mode (O2 flow through
the bypass) to the reactor mode (O2 flow through the reac-
tor), is complicated by an experimental artifact. An initial
1
8
abilities. Hence the relative amount of C O2 would have
been significantly higher. On the other hand, in a Langmuir–
Hinshelwood reaction mechanism one would expect equal
1
8
peak, which is either positive-going ( O2) or negative-going
1
6
16
16 18
(
O2), depending on the exact conditions, and which is
amounts of C O2 and C O O. Taking into account the
possibility of subsequent oxygen exchange between CO2
and the TiO2 substrate and assuming steady-state conditions,
introduced by the switching of the gas lines, prevents the
quantitative evaluation of the first 1.5 min. Essentially, upon
switching from bypass to reactor mode, at t = 0, the O2 par-
1
8
the amount of C O2 should be identical to double the “loss”
1
8
16
16 18
16 18
16
18
tial pressures drop to 0 kPa ( O2) or to about 0.3 kPa ( O2),
respectively. Subsequently, both signals increase slowly to
about 0.8 kPa, which is close to the steady-state value deter-
mined in the bypass measurements. A semilogarithmic plot
of the O2 signals during the slow increase shows that these
follow first-order kinetics most of the time, with time con-
of C O O (2C O O ↔ C O2 + C O2), i.e., to one-
16 16 18
third of the difference between C O2 and C O O. This
is exactly the case in the present measurements, and also in
the experiments reported by Liu et al. [29]. If we assume
that the reactivities of 16O2 and O2 for CO2 formation
are similar, it is not possible under steady-state conditions
to reach such a ratio in a reaction mechanism proceeding
via a carbonate intermediate. (For carbonate decomposition
without subsequent oxygen exchange (ratio of isotopomers
18
−
3
−1 16
−3 −1
s
stants of about 3.2 × 10
s
( O2) and 1.9 × 10
1
8
(
O2), respectively. The initial drop and the subsequent
slow increase of both O2 signals during the first 10 min indi-
cate an initial O2 consumption before reaching steady-state
conditions. A more detailed study of the transient reaction
and the amount of adsorbed oxygen behavior is in progress.
Similar observations of significant O2 consumption of a
Au/TiO2 catalyst upon exposure to O2 had been made for
catalysts prepared by impregnation and subsequent reductive
pretreatment, which had considerably larger Au nanoparti-
cles [26], as well as for Au/Fe2O3 [5]. This effect was ex-
plained by molecular adsorption of O2 on the oxide support,
most likely on defect sites.
1
6
16 18
18
18
C O2:C O O:C O2 = 3:2:1), the C O2 signal should
be equal to the full difference between the two other CO2
signals, and additional oxygen exchange cannot lead to the
experimental value.) Therefore, this result supports a mech-
1
6
anism with equal amounts of the two isotopomers C O2
1
6
18
and C O O2 being produced in the reaction, and subse-
quent oxygen exchange. From the time dependence of the
three CO2 signals it is clear that the equilibration between
gas-phase CO2 composition and surface composition must
be rather fast, on the order of about a minute at most, which
agrees with findings of Liu et al. upon interaction at room
temperature CO2 interaction with a Au/TiO2 catalyst [29],
considering the different reaction temperatures. Since oxy-
gen exchange was also found in the absence of gaseous O2,
the exchange reaction must involve other oxygen species
than the “active” molecular O2 species, most likely lattice
oxygen. Finally, it is interesting to note that these results
In a second experiment we passed a gas mixture con-
1
6
16
18
taining C O (1.2 kPa), O2 (0.5 kPa), and O2 (0.5 kPa)
◦
over the Au/TiO2 catalyst at 80 C (see Fig. 6b). The activ-
ity of the catalyst was sufficient to maintain almost com-
plete conversion over the entire time, as indicated by the
1
6
negligible C O intensity. The main oxidation product was
1
6
16 18
C O2, followed by C O O (final intensity 65% of the
1
6
18
18
C O2 intensity), and very little C O2 (steady-state con-
contrast the behavior on Au/Fe2O3, where no C O2 forma-
1
6
16 18
18
16 18
centrations: C O2:C O O:C O2 = 10:7:1). No O O
could be detected in this experiment, neither during the re-
action (Fig. 8b) nor afterward, when the sample was heated
tion was observed [5].
In the last experiment we tested whether the interaction of
CO with the Au/TiO2 catalyst leads to oxygen exchange, by
◦
18
to 300 C (spectrum not shown), indicating that also under
passing C O over it (Fig. 8c). From the absence of any de-
1
6
reaction conditions free Oad species are not present or only
in nondetectable amounts. Because of fluctuations in the O2
signals during the first 2 min, which may be due to hot-spot
formation, we did not attempt a quantitative evaluation of
the time dependence.
tectable amounts of C O this can clearly be ruled out. Also,
we do not see any CO2 formation. Hence significant con-
tributions from a Mars–van Krevelen-type mechanism [60]
can equally be ruled out. Also these results equals the be-
havior reported by Liu et al. [29]. The close similarity in the
reaction behavior between their Au/TiO2 catalyst and our
DP-prepared Au/TiO2 catalyst leads to the conclusion that
there are no basic differences in the reaction mechanism on
these rather differently prepared Au/TiO2 catalysts.
Rather similar results, with an isotope distribution for the
1
6
16 18
18
CO2 product of C O2:C O O:C O2 = 9.7:6:1, were ob-
tained by Liu et al. for CO oxidation on their Au/TiO2 cata-
lyst [29]. By flowing labeled CO2 over the catalyst, these au-

B. Schumacher et al. / Journal of Catalysis 224 (2004) 449–462
461
4
. Summary
order agrees well with the results of the IR CO adsorption
measurements, where the amount of adsorbed CO was found
to rise with the CO partial pressure in the gas mixture. We
explain the reduced CO oxidation activity by the parallel ad-
sorption and reaction of hydrogen, which competes for the
reactive oxygen species. This in combination with the low
O2 reaction order points to the existence of a rate-limiting
step between O2 adsorption and oxidation reaction.
We have investigated the CO oxidation behavior and in
particular the effect of H2 on this reaction on a highly active
Au/TiO2 catalyst, prepared via a novel synthesis and pre-
treatment procedure, which involves a modified deposition–
precipitation synthesis process. Following a previous report
on the CO oxidation activity and deactivation/stability of
this catalyst [23], this paper focuses on kinetic and mech-
anistic aspects of the CO oxidation reaction, the influence
of H2 on the reaction, and the interaction of CO with that
catalyst, both in an inert and in a reactive atmosphere and
in the absence and in the presence of H2. Based on kinetic
measurements, in situ IR spectroscopic measurements and
isotope-labeling experiments we derive the following con-
clusions.
The selectivity for CO oxidation in the H2-rich at-
mosphere depends strongly on the CO partial pressure, de-
◦
caying from 57% at 1.5 kPa and 80 C to 1.5% at 0.03 kPa.
It is largely determined by the change in CO oxidation activ-
ity, while the H2 oxidation rate changes little. Similar effects
are observed upon temperature variation; e.g., a temperature
◦
increase from 40 to 80 C reduces the selectivity from 64 to
4
0% at 1 kPa CO.
CO interaction with this catalyst, in an inert atmosphere
Experiments with isotope-labeled gases showed that no
◦
in the absence of H2 and at around 80 C, results in on-
top CO adsorption on the Au nanoparticles, with a peak at
2
2
alysts prepared and conditioned via other procedures. Fur-
thermore, formate, carboxylate, and carbonate species are
formed, which are identified by characteristic vibrations in
O2 isotope scrambling occurs upon exposure to a mixture of
16
18
O2 and O2, in agreement with previous findings for other
−
1
112 cm , which red-shifts with increasing coverage to
Au/TiO2 catalysts. O2 isotope scrambling is also not ob-
served under reaction conditions, indicating that the second
O atom remaining after the reaction step cannot recombine
with another, similar O adatom and desorb, but leaves the
surface via reaction. The ratio of CO2 isotopomers formed
upon reaction with a mixture of 16O2 and O2 points to a re-
action mechanism proceeding via a Langmuir–Hinshelwood
process, with subsequent oxygen exchange between CO2
and TiO2 support process, rather than to the decomposition
of a carbonate intermediate as major reaction path. Finally,
−
1
108 cm . This resembles observations on Au/TiO2 cat-
−1
the 1300–1700 cm
regime and, for formates, by their
18
−1
C–H vibrations between 2800 and 3000 cm . The pres-
ence of H2 has a pronounced effect on the interaction of
CO with the Au/TiO2 catalyst. The C–O stretch frequency
is slightly red-shifted by 3 cm , which results either from
charge donation from the partly reduced substrate or, prefer-
ably, from indirect interaction of H and CO species coad-
sorbed on the Au nanoparticles. More important, with in-
creasing reaction time it leads to the slow appearance of
a new, strongly asymmetric absorption band with its maxi-
−1
1
8
transient experiments with C O demonstrate that (i) oxy-
gen exchange between CO and TiO2 substrate is inhibited
and that (ii) reaction with lattice oxygen is not possible af-
ter depletion of the “active” adsorbed O2 species. The latter
observation rules out significant contributions from a Mars–
van Krevelen reaction mechanism on this catalyst.
−1
mum at 2032 cm , which finally replaces the original peak
−1
at 2107 cm , together with an increasing hydroxylation of
the support. Based on the constant position and shape of
this peak we favor an explanation where this new feature
results from an H/CO adsorption complex rather than from
charge transfer from the increasingly reduced TiO2 substrate
to the Au nanoparticles. In addition, the presence of H2 in-
hibits formate or carbonate formation. In a reactive H2-free
atmosphere, in the presence of O2, the formation of for-
mates and in particular carbonates is considerably enhanced,
which is held responsible for the decay of the CO oxidation
activity with time. In the H2-rich reactive atmosphere the
new, asymmetric peak described above is not observed. On
the other hand, formate/carbonate formation is still inhib-
ited.
The above results lead to the proposal that (i) H2 in the
gas phase affects the CO oxidation by competing hydrogen
adsorption on the Au nanoparticles and reaction with oxy-
gen, resulting in a significantly higher CO reaction order,
that (ii) interaction with H2O, which is present in the process
gas or formed upon reaction with H2, increases the activity
indirectly by reducing the steady-state concentration of reac-
tion inhibiting formate/carbonate species, and that (iii) for-
mate and carbonate species formed during reaction represent
(reaction inhibiting) side products, but do not take part in
the reaction as reaction intermediate, at least not under the
present reaction conditions.
A pronounced effect of the H2 coreactant is observed also
in the kinetic measurements. In the temperature and pressure
range investigated the reaction order for CO is almost dou-
bled, from about 0.35 to about 0.9, while the O2 reaction
order is hardly affected and remains at about 0.35. This goes
along with a reduced activity at all but the highest CO partial
pressures investigated (1–1.5 kPa). The positive CO reaction
Acknowledgment
We gratefully acknowledge financial support by the
Deutsche Forschungsgemeinschaft within the Priority Pro-
gramme 1091 (Be 1201/9-3).

462
B. Schumacher et al. / Journal of Catalysis 224 (2004) 449–462
References
[29] H. Liu, A.I. Kozlov, A.P. Kozlova, T. Shido, K. Asakura, Y. Iwasawa,
J. Catal. 185 (1999) 252.
[
30] V. Plzak, J. Garche, R.J. Behm, Eur. Fuel Cell News 10 (2003) 8.
[
[
[
[
[
1] M. Haruta, Catal. Surv. Jpn. 1 (1997) 61.
2] G.C. Bond, Catal. Rev.-Sci. Eng. 41 (1999) 319.
3] M. Haruta, M. Daté, Appl. Catal. A 222 (2001) 427.
4] M. Haruta, CATTECH 6 (2002) 102.
5] M.M. Schubert, S. Hackenberg, A.C. van Veen, M. Muhler, V. Plzak,
R.J. Behm, J. Catal. 197 (2001) 113.
[31] M.J. Kahlich, H.A. Gasteiger, R.J. Behm, J. Catal. 171 (1997) 93.
[32] P.B. Weisz, Chem. Eng. Progr. Ser. 55 (1959) 29.
[33] I.M. Hamadeh, P.R. Griffiths, Appl. Spectrosc. 41 (1987) 682.
[34] H. Schubert, U. Guntow, K. Hofmann, R. Schlögl, Fresenius J. Anal.
Chem. 356 (1996) 127.
[
[
6] F. Boccuzzi, A. Chiorino, J. Phys. Chem. B 104 (2000) 5414.
7] G.R. Bamwenda, S. Tsubota, T. Nakamura, M. Haruta, Catal. Lett. 44
[35] M.J. Kahlich, H.A. Gasteiger, R.J. Behm, J. Catal. 182 (1999) 430.
[36] R.M. Torres Sanchez, A. Ueda, K. Tanaka, M. Haruta, J. Catal. 168
(1996) 125.
(1997) 83.
[8] M. Haruta, S. Tsubota, T. Kobayashi, H. Kageyama, M.J. Genet, B.
Delmon, J. Catal. 144 (1993) 175.
[37] R.J.H. Grisel, K.-J. Weststrate, A. Goossens, M.W.J. Craje, A.M. van
der Kraan, B. Nieuwenhuys, Catal. Today 72 (2002) 123.
[
9] Y. Iizuka, H. Fujiki, N. Yamauchi, T. Chijiiwa, S. Arai, S. Tsubota, M.
Haruta, Catal. Today 36 (1997) 115.
[
[
[
[
38] H. Sakurai, A. Ueda, T. Kobayashi, M. Haruta, Chem. Commun. 271
1998).
39] F. Boccuzzi, A. Chiorino, S. Tsubota, M. Haruta, J. Phys. Chem. 100
1996) 3625.
40] J.-D. Grunwaldt, M. Maciejewski, O.S. Becker, P. Fabrizioli, A.
Baiker, J. Catal. 186 (1999) 458.
(
[
10] A.I. Kozlov, A.P. Kozlova, H. Liu, Y. Iwasawa, Appl. Catal. A 182
(1999) 9.
(
[
[
11] J.-D. Grunwaldt, A. Baiker, J. Phys. Chem. B 103 (1999) 1002.
12] F. Boccuzzi, A. Chiorini, M. Manzoli, P. Lu, T. Akita, S. Ichikawa, M.
Haruta, J. Catal. 202 (2001) 256.
41] J. France, P. Hollins, J. Electr. Spectrosc. Rel. Phenom. 64/65 (1993)
[
13] M.M. Schubert, V. Plzak, J. Garche, R.J. Behm, Catal. Lett. 76 (2001)
251.
143.
[
[
[
42] C. Ruggiero, P. Hollins, J. Chem. Soc., Faraday Trans. 92 (1997) 4829.
43] P. Hollins, Surf. Sci. Rept. 16 (1992) 51.
44] G. Busca, J. Lamotte, J.-C. Lavalley, V. Lorenzelli, J. Am. Chem.
Soc. 109 (1987) 5197.
[
14] M.J. Kahlich, M.M. Schubert, M. Hüttner, M. Noeske, H.A. Gasteiger,
R.J. Behm, in: O. Savadogo, P.R. Roberge (Eds.), New Materials for
Fuel Cells and Modern Battery Systems II, Ecole Polytechnique de
Montreal, Montreal, 1997, pp. 642–653.
15] M.M. Schubert, M.J. Kahlich, H.A. Gasteiger, R.J. Behm, J. Power
Sources 84 (1999) 175.
16] Y.-F. Han, V. Plzak, M. Kinne, R.J. Behm, in: F.N. Büchi, G. Scherer,
A. Wokaun (Eds.), Proc. 1st European Fuel Cell Forum, Luzern, 2001,
pp. 393–396.
[
[
[
45] F. Boccuzzi, S. Tsubota, M. Haruta, J. Electr. Spectrosc. Rel. Phe-
[
nom. 64–65 (1993) 241.
46] L.-F. Liao, D.-L. Lien, D.-L. Shieh, M.-T. Chen, J.-L. Lin, J. Phys.
Chem. B 106 (2002) 11240.
47] C.-C. Chuang, W.-C. Wu, M.-C. Huang, I.-C. Huang, J.-L. Lin,
J. Catal. 185 (1999) 423.
[
[
17] M.M. Schubert, A. Venugopal, M.J. Kahlich, V. Plzak, R.J. Behm,
[
[
[
48] G.C. Bond, D.T. Thompson, Gold. Bull. 33 (2000) 41.
49] G. Martra, Appl. Catal. A 200 (2000) 275.
50] M.I. Zaki, G. Kunzmann, B.C. Gates, H. Knözinger, J. Phys. Chem. 91
J. Catal. 222 (2004) 32.
[
[
18] Y.-F. Han, V. Plzak, M. Kinne, R.J. Behm, Submitted for publication.
19] S. Gottesfeld, T.A. Zawodzinski, in: R.C. Alkire, H. Gerischer, D.M.
Kolb, C.W. Tobias (Eds.), Advances in Electrochemical Science and
Engineering, vol. 5, Wiley–VCH, Weinheim, 1997.
(1987) 1486.
[51] F. Boccuzzi, A. Chiorino, M. Manzoli, Surf. Sci. 454–456 (2000) 942.
[52] M. Jakob, H. Levanon, P.V. Kamat, Nano Lett. 3 (2003) 353.
[53] B.E. Koel, D.E. Peebles, J.M. White, Surf. Sci. 125 (1983) 709.
[54] B.E. Koel, D.E. Peebles, J.M. White, Surf. Sci. 125 (1983) 739.
[55] F. Boccuzzi, A. Chiorino, S. Tsubota, M. Haruta, Shokubai (Catal.
Catal.) 38 (1996) 422.
[
20] D.P. Wilkinson, D. Thompsett, in: O. Savadogo, P.R. Roberge (Eds.),
Materials and Approaches for CO and CO Tolerance for Polymer
2
Electrolyte Membrane Fuel Cells, Ecole Polytechnique de Montreal,
Montreal, 1997, pp. 266–285.
[
[
[
21] D.L. Trimm, Z.I. Önsan, Catal. Rev. 43 (2001) 31.
22] J.M. Zalc, V. Sokolovskii, D.G. Loeffler, J. Catal. 206 (2002) 169.
23] B. Schumacher, V. Plzak, M. Kinne, R.J. Behm, Catal. Lett. 89 (2003)
[
[
56] M.M. Schubert, PhD dissertation, University of Ulm, 2000.
57] M. Olea, M. Kunitake, T. Shido, Y. Iwasawa, Phys. Chem. Chem.
Phys. 3 (2001) 627.
109.
[58] M. Olea, M. Kunitake, T. Shido, K. Asakura, Y. Iwasawa, Bull. Chem.
[
[
[
[
24] S.D. Lin, M.A. Bollinger, M.A. Vannice, Catal. Lett. 17 (1993) 245.
25] M.A. Bollinger, M.A. Vannice, Appl. Catal. B 8 (1996) 417.
26] N.J. Ossipoff, N.W. Cant, Top. Catal. 8 (1999) 161.
27] T.V. Choudhary, C. Sivadinarayana, C.C. Chusuei, A.K. Darye, J.P.
Fackler, D.W. Goodman, J. Catal. 207 (2002) 247.
Soc. Jpn. 74 (2001) 255.
[59] H. Liu, A.I. Kozlov, A.P. Kozlova, T. Shido, Y. Iwasawa, Phys. Chem.
Chem. Phys. 1 (1999) 2851.
[60] P. Mars, D.W. van Krevelen, Chem. Eng. Sci. 3 (1954) 41.
[61] N.W. Cant, N.J. Ossipoff, Catal. Today 36 (1997) 125.
[
28] F. Boccuzzi, A. Chiorino, M. Manzoli, Surf. Sci. 502–503 (2002) 513.