Support effects in the Au-catalyzed CO oxidation - Correlation between activity, oxygen storage capacity, and support reducibility

Article abstract of DOI:10.1016/j.jcat.2010.09.023

The oxygen storage capacity (OSC) and its correlation with the activity for the CO oxidation reaction and the reducibility of the support material were investigated for four different metal oxide-supported Au catalysts with similar Au loading and Au particle sizes (Au/Al₂O₃, Au/TiO ₂, Au/ZnO, Au/ZrO₂), which were prepared by deposition of pre-formed Au colloids. Temporal Analysis of Products (TAP) reactor measurements show that the OSC and the activity for CO oxidation, measured under identical conditions, differ significantly for these catalysts and are correlated with each other and with the reducibility of the respective support material, pointing to a distinct support effect and a direct participation of the support in the reaction. Activity measurements performed under ambient conditions show a similar trend of the activity as the TAP reactor measurements, supporting that the conclusions drawn from the TAP reactor measurements are valid also under continuous reaction conditions. Moreover, the rapid formation and accumulation of carbon-containing surface species during reaction is demonstrated, which can severely reduce the activity for CO oxidation. Implications of these results on the CO oxidation mechanism over metal oxide-supported catalysts are discussed.

Full text of DOI:10.1016/j.jcat.2010.09.023

Journal of Catalysis 276 (2010) 292–305
Contents lists available at ScienceDirect
Journal of Catalysis
journal homepage: www.elsevier.com/locate/jcat
Support effects in the Au-catalyzed CO oxidation – Correlation between
activity, oxygen storage capacity, and support reducibility
D. Widmann a, Y. Liu b, F. Schüth b, R.J. Behm a,
⇑
a Institute of Surface Chemistry and Catalysis, Ulm University, D-89069 Ulm, Germany
b Max-Planck-Institut für Kohlenforschung, D-45470 Mülheim an der Ruhr, Germany
a r t i c l e i n f o
a b s t r a c t
Article history:
Received 13 July 2010
Revised 21 September 2010
Accepted 22 September 2010
Available online 9 November 2010
The oxygen storage capacity (OSC) and its correlation with the activity for the CO oxidation reaction and
the reducibility of the support material were investigated for four different metal oxide-supported Au
catalysts with similar Au loading and Au particle sizes (Au/Al2O3, Au/TiO2, Au/ZnO, Au/ZrO2), which were
prepared by deposition of pre-formed Au colloids. Temporal Analysis of Products (TAP) reactor measure-
ments show that the OSC and the activity for CO oxidation, measured under identical conditions, differ
significantly for these catalysts and are correlated with each other and with the reducibility of the respec-
tive support material, pointing to a distinct support effect and a direct participation of the support in the
reaction. Activity measurements performed under ambient conditions show a similar trend of the activity
as the TAP reactor measurements, supporting that the conclusions drawn from the TAP reactor measure-
ments are valid also under continuous reaction conditions. Moreover, the rapid formation and accumu-
lation of carbon-containing surface species during reaction is demonstrated, which can severely reduce
the activity for CO oxidation. Implications of these results on the CO oxidation mechanism over metal
oxide-supported catalysts are discussed.
Keywords:
Au catalysts
CO oxidation
Support effects
Oxygen storage capacity (OSC)
Reaction mechanism
Ó 2010 Elsevier Inc. All rights reserved.
1. Introduction
oxidation activity of four different Au catalysts (Au/Al2O3, Au/
TiO2, Au/ZnO, Au/ZrO2) with similar Au particle sizes and Au load-
Oxide-supported small Au nanoparticles have attracted consid-
erable attention in recent years due to their high activity for cata-
lyzing various oxidation and reduction reactions already at low
temperatures [1]. Most prominent examples are the CO oxidation
reaction [2–5], the water–gas shift reaction [6–9], or the selective
and total oxidation of hydrocarbons [10–12]. Despite extensive
studies, the physical origin of the high activity of supported Au cat-
alysts and the underlying reaction mechanisms are still under de-
bate [13,14]. Open questions are for example (i) the pronounced
role of the Au particle size [15], which was explained, e.g., by quan-
tum size effects [16] or a high activity of under-coordinated Au
atoms at corners and edges [17–23], (ii) the nature of the active
Au species (metallic Au nanoparticles or ionic/partly charged
Aun+ or AudÀ species) [24–28], (iii) the influence of the support
material on the catalytic performance of the corresponding Au cat-
alysts and its role in the reaction process [28–33], and in particular
(iv) the activation of oxygen and the active site for oxygen activa-
tion [33–35]. The latter two questions are topic of the present
study. They are approached in a simple way, by investigating the
oxygen storage capacity (OSC) and its correlation with the CO
ings, but with support materials which differ strongly in their
reducibility. The catalysts were prepared by deposition of pre-
formed Au nanoparticles (‘colloidal deposition’).
Based on numerous studies, there is no doubt that the support
has a pronounced effect on the activity of these catalysts. Open and
not yet resolved is the exact nature of this effect. Testing a variety
of Au catalyst supported on different metal oxides for their activity
in the CO oxidation reaction, Schubert et al. [29] distinguished be-
tween two major groups of support materials, reducible materials
leading to ‘active’ catalysts and non-reducible materials resulting
in ‘inactive’ (or little active) catalysts. The difference was tenta-
tively attributed to the different ability of these materials to create
oxygen vacancies on the support, close to the Au particles, which
were proposed as active centers for oxygen activation during the
CO oxidation reaction. However, there are a number of different
possibilities for the support to modify the activity of the catalyst.
In addition to directly participating in the reaction, the support
may affect the reaction also indirectly, by influencing the shape
and size of the Au nanoparticles during the catalyst preparation
and activation procedure, via metal–support interactions [22], by
support-induced strain in the Au nanoparticles [18], by charge
transfer from or to the Au nanoparticles [36] or by stabilizing ionic
Au species [37].
⇑
Corresponding author. Fax: +49 731 50 25452.
E-mail address: juergen.behm@uni-ulm.de (R.J. Behm).
0021-9517/$ - see front matter Ó 2010 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcat.2010.09.023

D. Widmann et al. / Journal of Catalysis 276 (2010) 292–305
293
Several groups have tried to discriminate between these
possibilities by producing Au catalysts with similar Au loading
and particle size, either by traditional routes, e.g., by deposition–
precipitation techniques [22,28], or by depositing pre-formed Au
nanoparticles of similar size on different support materials [38–
40] and comparing their activity. In a careful TEM study, Janssens
et al. compared Au catalysts supported on TiO2, MgAl2O4, and
Al2O3 and arrived at the conclusion, that the differences in reactiv-
ity of these catalysts can mainly be explained by the different num-
bers of under-coordinated Au sites on the Au nanoparticles [22],
although other support-induced effects have to be considered as
well to understand the lower activity of the Au/Al2O3 catalyst.
Comparing the reduction behavior of Au3+ species and the CO oxi-
dation activity of Au/TiO2, Au/CeO2, and Au/Al2O3 catalysts, which
were prepared by a deposition–precipitation procedure and subse-
quent activation, Delannoy et al. noted that both properties are
strongly affected by the support and that the CO oxidation activity
of these catalysts is related to the presence of Au0 nanoparticles.
On the reduced catalysts, the activity was found to decrease in
the order Au/TiO2 ꢀ Au/CeO2 ꢁ Au/Al2O3 [28]. To minimize effects
caused by support-dependent differences in the Au3+ reduction
behavior, Comotti el al. performed a similar comparison of the
CO oxidation activity using Au catalysts (Au/Al2O3, Au/TiO2, Au/
ZnO, and Au/ZrO2) with similar Au loadings and particle size distri-
butions, which were prepared via a colloidal deposition method
[40]. Also these authors found that the nature of the support has
a significant influence on the catalytic properties. This influence,
however, did not follow the reducibility of the supports. They ten-
tatively suggested that support-induced variations in the shape of
the Au nanoparticles, due to metal–support interactions, lead to
faceting and defect sites and this way affect the activity. Similar
studies were performed also by Grunwaldt et al. [17] and Arrii
et al. [39]. Also in these studies, the activity of the catalysts de-
pended strongly on the nature of the metal oxide used as support
material; the origin of the different catalytic performance of the
corresponding Au catalysts, however, remained controversial. Re-
cent TAP reactor measurements have shown that Au/TiO2 and
AuCeO2 catalysts are able to reversibly store and release active
oxygen on the surface upon adsorption from O2 or reaction with
CO, respectively [33,41,42]. This active oxygen was stable against
desorption at least up to 120 °C. Furthermore, for Au/TiO2 catalysts
with different particle sizes but similar Au loading, the OSC and the
activity for CO oxidation were found to be closely correlated and to
scale with the perimeter of the interface between Au nanoparticles
and TiO2 support, pointing to a direct involvement of these inter-
face sites as active sites for oxygen storage and in the CO oxidation
reaction [33].
build-up of carbon-containing surface species (differences in CO
consumption and CO2 formation) on the different catalysts are
evaluated. Finally, the stability and dynamic behavior of the car-
bon-containing surface species is evaluated in transient experi-
ments using isotope labeling techniques. Consequences of these
data on the CO oxidation mechanism will be discussed.
2. Experimental
2.1. Catalyst preparation
The four different Au catalysts were prepared by a colloidal
deposition method using commercial, non-porous metal oxides.
This method was described in detail in Refs. [43,44]. First, colloidal
Au solutions were prepared by adding polyvinyl alcohol (PVA) as
protecting agent to an aqueous solution of gold acid (HAuCl4) at
room temperature. Afterward, gold reduction was initiated by ra-
pid injection of an aqueous solution of NaBH4, resulting in an or-
ange-brown gold sol with metallic Au particles. To this gold sol,
the support material was added under stirring, which was contin-
ued until all of the Au particles were adsorbed on the surface
(nominal loading 1 wt.%). Similar to the work of Comotti et al.
[40], we used the following support materials: P25 from Degussa
for the Au/TiO2 catalyst, AC-45 from Brüggemann Chemicals for
the Au/ZnO catalyst, PURALOX SBa-200 from Sasol for the Au/
Al2O3 catalyst, and Zr(OH)4 from MEL Chemicals for the Au/ZrO2
catalyst. While the first three supports were used as received, the
latter one was additionally calcined at 350 °C for 4 h prior to the
Au deposition in order to obtain dehydroxylated ZrO2. The result-
ing raw catalysts were washed and dried under vacuum
(10À2 mbar) in a desiccator with P2O5 as drying agent and then
stored in the desiccator until use. Catalyst preparation and storage
took place under exclusion of light.
Prior to all measurements, the catalyst was first dried for
1000 min in order to obtain the same water content on all samples,
independent of the previous history (TAP reactor: vacuum drying
at room temperature; plug-flow reactor: drying in a flow of
20 Nml minÀ1 N2 at 100 °C) and afterward calcined in a flow of
20 Nml minÀ1 10% O2/N2 at 250 °C for 2 h in order to remove the
protecting agent (henceforth this procedure will be denoted as
‘‘O250’’). For heating and cooling, the temperature was raised/de-
creased by 10 °C minÀ1 in a flow of 20 Nml minÀ1 Ar (TAP reactor)
or N2 (plug-flow reactor). For calcination as well as for reaction
measurements (described below), high purity gases from West-
phalen were used as delivered (CO 4.7, O2 5.0, N2 6.0, and Ar 6.0).
2.2. Catalyst characterization
In the present study, we will apply Temporal Analysis of Prod-
ucts (TAP) reactor measurements, which allow us to directly deter-
mine the OSC of the different catalysts and thus to reveal possible
correlations between OSC, CO oxidation activity, and redox behav-
ior of the different support materials. From these results, we can
derive a direct participation of the support in the CO oxidation
reaction, which has not been possible by any other technique up
to now. Following a brief description of the experimental set-up
and procedures, we will first present results characterizing the
physical properties of the catalysts such as the loading and size
of the Au nanoparticles and the surface area of the different cata-
lysts (Section 3.1), followed by measurements of the CO oxidation
kinetics under atmospheric pressure (Section 3.2), and finally the
results of the TAP reactor measurements (Section 3.3). The latter
included both multi-pulse sequences for determining the OSC
and single-pulse measurements, with simultaneous pulses of CO
and O2, for evaluation of the CO oxidation activity under these con-
ditions. In addition, changes of the oxidation state during reaction
(differences in CO consumption and O2 consumption) and the
The surface area of the catalysts was determined by low-
temperature nitrogen adsorption using the method of Brunauer,
Emmet, and Teller (BET) for evaluation. The Au content of the
raw catalysts was measured by inductively coupled plasma optical
emission spectrometry (ICP-OES), and the Au particle size was
evaluated from transmission electron microscopy (TEM) images.
2.3. Catalytic activities
First, the catalytic activity for CO oxidation on the four catalysts
was evaluated in a conventional micro-reactor at atmospheric
pressure (120 °C reaction temperature, 1000 min reaction time)
after in situ calcination (O250, description see above). The catalysts
were diluted with
a-Al2O3 in order to obtain differential reaction
conditions. Depending on the activity of the respective catalyst, it
was 1:120 for Au/TiO2, 1:10 for Au/ZnO, and 1:5 for Au/ZrO2; no
dilution was used for Au/Al2O3. This resulted in conversions of

294
D. Widmann et al. / Journal of Catalysis 276 (2010) 292–305
below 15% for CO and O2 during the activity measurements. About
65 mg of the diluted catalysts were placed in the center of a quartz
tube micro-reactor (i.d. 4 mm). The gas flow was 60 Nml minÀ1
(1 kPa CO, 1 kPa O2, rest N2); influent and effluent gases were ana-
lyzed by on-line gas chromatography (DANI, GC 86.10). For further
details on the set-up and the evaluation procedure, see Ref. [45].
simultaneous-pulse measurements (Au/TiO2: 10 mg, Au/ZrO2:
3 mg, Au/ZnO: 5 mg, Au/Al2O3: not measured).
2.5. Temperature-programmed desorption
To identify carbon-containing adsorbed species, which were
accumulated on the catalyst surface during reaction, we also per-
formed temperature-programmed desorption (TPD) measure-
ments in the TAP reactor. The measurements started 5 min after
finishing the reaction, heating the catalyst from reaction tempera-
ture (120 °C) to 800 °C (25 °C minÀ1) without an additional carrier
gas. The effluent gases arising from decomposition/desorption of
the surface species were detected by the mass spectrometer.
2.4. Pulse experiments
The pulse experiments were performed in a home-built TAP
reactor which was described in detail in Ref. [46]. It is comparable
to a design published by Gleaves et al. (TAP-2) [47], easy to operate
and excels by its highly reproducible pulse sequences. It consists of
a gas mixing unit, two independently working piezo-electrically
driven pulse valves and a tubular quartz glass micro-reactor
(90 mm long, 4 mm i.d., 6 mm o.d.), which was heated by a ceramic
tube furnace and connected to an ultra high vacuum system (anal-
ysis chamber). Pulses of typically 1 Â 1016 molecules, generated by
piezo-electric pulse valves, were directed into the micro-reactor.
Exact numbers of the amount of molecules admitted per pulse
were calculated by comparison with the signal of the internal stan-
dard argon, which was included in every pulse. In the central part
of the reactor, the catalyst bed was fixed by two stainless steel
3. Results and discussion
3.1. Catalyst characterization
3.1.1. Au content
The Au content of the Au catalysts was determined by induc-
tively coupled plasma-optical emission spectroscopy (ICP-OES),
measured two times by repeat determination. For the Au/TiO2,
Au/ZrO2, and Au/ZnO catalysts, it was 1.0 wt.%, identical to the
nominal loading obtained upon complete deposition of the Au par-
ticles on the support. Only for the Au/Al2O3 catalyst, the Au content
was slightly lower (0.8 wt.%) (see Table 1). This catalyst required
also the longest time (24 h) for depositing the Au nanoparticles.
For comparison, on the TiO2 support, where the deposition was
fastest, it took only 30 min to adsorb all Au nanoparticles and reach
the nominal Au loading of 1.0 wt.%.
sieves (Haver & Boecker OHG; aperture 25 lm, transmission
25%). The catalyst bed itself consisted of three zones: the catalyst
zone containing the catalyst diluted with SiO2 in the center (‘reac-
tion zone’) and two additional zones with inert material before and
after the catalyst zone (‘gebaflot 010’ from Dorfner GmbH; grain
size 100–200 lm) acting as diffusion zones [48]. The stainless steel
sieves as well as the quartz powder were tested to be inactive for
the CO oxidation in the temperature range investigated. The
amount of catalyst and the dilution were different for different
types of experiments and are described below. After passing
through the reactor, the gases were analyzed by a quadrupole mass
spectrometer (QMG 700, Pfeiffer Vacuum) located in the analysis
chamber closely behind the reactor tube. The consumption of CO
and O2 in the respective pulses was calculated from the missing
mass spectrometric intensity in these pulses compared to the
intensity after saturation. The formation of CO2 was determined di-
rectly from the CO2 pulse intensity. For in situ conditioning of the
catalyst at atmospheric pressure prior to the measurements, the
reactor can be separated from the UHV system by a differentially
pumped gate valve and connected directly to an adjustable rough-
ing pump [46].
3.1.2. Au particle size
The Au particle sizes on the various catalysts were measured by
transmission electron microscopy (TEM). Representative images
for each catalyst are shown in Fig. 1. In these images, the dark spots
represent the Au particles. The evaluation of >100 particles for each
catalyst resulted in mean Au diameters of 3.3 1.0 nm for Au/TiO2,
2.4 0.7 nm for Au/ZrO2, 3.1 0.9 nm for Au/ZnO, and 3.5 1.1 nm
for Au/Al2O3. The corresponding particle size distributions are
shown in Fig. 2. Assuming half-spherical Au particles and
1.15 Â 1015 Au atoms cmÀ2, we can further calculate the dispersion
of each catalyst, which is listed in Table 1. It should be noted that
for similar catalysts Comotti et al. found Au particle sizes which are
slightly (0.3 nm) smaller than in the present study, which can be
explained by the different pre-treatment (pre-treatment in reac-
tion atmosphere vs. O250) [40].
In the multi-pulse experiments, the catalysts were exposed first
to a sequence of 200 CO/Ar and then to a sequence of 100 O2/Ar
pulses, in order to reactively remove and deposit active oxygen.
In these experiments, we used 10 mg of each catalyst diluted with
SiO2 in a ratio 1:2, and the content of Ar molecules was 50% in each
pulse.
3.1.3. Surface area
The surface areas of all four catalysts, as measured by low-tem-
perature nitrogen adsorption (BET), were close to the surface areas
For evaluating the catalytic activity in the TAP reactor, the sam-
ples were exposed to CO and O2 at the same time by simultaneous
CO/Ar and O2/Ar pulses (with 50% Ar each) (‘single-pulse’ measure-
ments). The resulting CO:O2 ratio was 1:1. Hence, there was an ex-
cess of oxygen compared to the stoichiometry of the CO oxidation
reaction. These measurements were used to compare the catalytic
activity of the four differently supported Au catalysts. Therefore,
the amount of catalyst (5 mg) and the dilution with SiO2 (1:5) were
identical for all of these measurements. Additional, single-pulse
experiments using labeled 13CO were performed, which aimed at
evaluating the stability of adsorbed carbon-containing species un-
der reaction conditions. Here, first a reaction mixture containing
labeled 13CO was pulsed with one pulse valve (45% 13CO, 45% O2,
10% Ar), followed by pulses containing non-labeled 12CO with the
other pulse valve (45% 12CO, 45% O2, 10% Ar). Here, the amount
of catalyst and the dilution were slightly different than in the
2
À1
of the corresponding support materials. They are 198 m g for
cat
2
À1
the Au/Al2O3 catalyst, 47 m g
for the Au/TiO2 catalyst,
cat
Table 1
Physical properties of the different supported Au/MxOy catalysts after pre-treatment
by calcination in 10% O2/N2 at 250 °C for 2 h (O250).
Au/TiO2
Au/ZrO2
Au/ZnO
Au/Al2O3
Au loadinga (wt.%)
Au diameterb (nm)
1.0
3.3 1.0
47
1.0
2.4 0.7
201
1.0
3.1 0.9
50
0.8
3.5 1.1
198
Surface areac (m2
Dispersion (%)d
g
)
À1
cat
30
42.
33
28
a
b
c
Measured by ICP-OES.
Measured by TEM.
Measured by low-temperature N2 adsorption (BET).
Assuming half-spherical Au particles.
d

D. Widmann et al. / Journal of Catalysis 276 (2010) 292–305
295
(a) Au/TiO2
(b) Au/ZrO2
20 nm
20 nm
(c) Au/ZnO
(d) Au/Al2O3
20 nm
20 nm
Fig. 1. Representative TEM images of the four differently supported Au catalysts after calcination in 10% O2/N2 at 250 °C for 2 h (O250).
(b) Au/ZrO2
(a) Au/TiO2
30
20
10
dAu = 3.3 1.0 nm
d
Au = 2.4 0.9 nm
0
(d) Au/Al2rO3
(c) Au/ZnO
30
dAu = 3.1 0.9 nm
dAu = 3.5 1.1 nm
20
10
0
2
4
6
2
4
6
dAu / nm
Fig. 2. Particles size distributions of the (a) Au/TiO2, (b) Au/ZrO2, (c) Au/ZnO, and (d) Au/Al2O3 catalysts after calcination in 10% O2/N2 at 250 °C for 2 h (O250).
2
2
À1
À1
cat
50 m g for the Au/ZnO catalyst, and 201 m g for the Au/ZrO2
Au/TiO2, Au/ZrO2, Au/ZnO, and for Au/Al2O3, respectively (see also
Table 2). The corresponding TOF numbers, which decrease in the
same order due to the almost identical Au particle sizes, are 7.1,
0.37, 0.24, and 0.05 sÀ1, respectively. The pronounced differences
in activity for CO oxidation, which differed by almost a factor of
200 between the most active (Au/TiO2) and the least active (Au/
Al2O3) catalyst, despite of comparable Au nanoparticle sizes, are
clear evidence for a distinct effect of the support on the catalytic
activity of the corresponding catalysts. This will be discussed in de-
tail later.
cat
catalyst (see also Table 1).
3.2. Kinetic measurements
Fig. 3 shows the Au mass normalized reaction rates measured
for the CO oxidation reaction over the four different supported cat-
alysts under differential reaction conditions. Because of the similar
Au loadings and mean Au particle sizes of the four catalysts, these
reaction rates are a direct measure for the inherent catalytic activ-
ity of the respective catalysts. The reaction rates obtained under
These results can be compared with data on the catalytic activ-
ities of the very same Au catalysts published recently [40], where
the temperature needed for 50% conversion of CO (T50) was used
steady-state conditions, after 1000 min time on stream, are
1.1 Â 10À2, 7.7 Â 10À4, 4.1 Â 10À4, and 7.1 Â 10À5 mol sÀ1
g
for
À1
Au

296
D. Widmann et al. / Journal of Catalysis 276 (2010) 292–305
between Au particles and support. On a molecular scale, this would
Au/TiO2
Au/ZrO2
Au/ZnO
Au/Al2O3
1.40
1.20
1.00
be attributed to the ability of the support to store and/or activate
oxygen and release it for the oxidation of CO, which is generally
agreed to adsorb on the Au nanoparticles [1,13,34,49,50].
Grunwaldt et al. compared the CO oxidation activity of Au catalysts
supported on two different metal oxides (Au/TiO2 and Au/ZrO)
[17], using pre-formed, ‘size-controlled’ Au particles deposited on
the different supports, similar to the present work, and a similar
study was reported by Arrii et al. [39], investigating CO oxidation
on Au/TiO2, Au/ZrO2, and Au/Al2O3 catalysts. Comparable to the re-
sults of Schubert et al., these authors found the catalytic activity for
CO oxidation to decrease with decreasing reducibility of the sup-
port material (Au/TiO2 > Au/ZrO2 ꢁ Au/Al2O3). Grunwaldt et al. ex-
plained the differences in catalytic activity by different shapes of
the Au nanoparticles deposited on different supports, as evidenced
by DRIFTS (CO adsorption) and TEM measurements, while Arri
et al. suggested a synergetic or cooperative effect between the
TiO2 or ZrO2 support and the Au nanoparticles for the active Au/
TiO2 and Au/ZrO2 catalysts, with the support participating in the
reaction [39]. On the other hand, in apparent disagreement to this
proposal and the results presented here, some groups also reported
high activities for Au catalysts supported on ‘‘inert’’ metal oxides
such as Au/Al2O3, showing activities close to or even superior to
Au/TiO2 [51,52]. In the study by Wolf et al., however, the Au parti-
cle size was lower for the catalysts supported on the inert metal
oxide (Au/Al2O3) than for the Au/TiO2 catalyst used for comparison.
When comparing catalysts with the same Au particle size on Al2O3
and TiO2, the Au/Al2O3 catalyst is significantly less active for CO
oxidation than the Au/TiO2 catalyst also in that study [51]. Never-
theless, in all of these studies, possible effects of the support on the
catalytic activity have been somehow tentative, since a direct par-
ticipation of the support under relevant reaction conditions was
not investigated. This will be topic of the following chapter, where
we compare the differently supported catalysts in terms of their
ability to form active oxygen on the surface.
0.12
0.10
0.08
0.06
0.04
0.02
0.00
0
200
400
600
800
1000
Time / min
Fig. 3. Au mass normalized reaction rates during the CO oxidation at 120 °C over
the four differently supported Au catalysts after calcination in 10% O2/N2 at 250 °C
for 2 h (O250) (1 kPa CO, 1 kPa O2, and balance N2).
as measure for the activity (reaction atmosphere: 1 kPa CO, 20 kPa
O2, rest N2). The T50 temperature varied from T50 = À14 °C for Au/
TiO2, closely followed by Au/Al2O3 (values between À12 and 46 °C),
to T50 = 50 °C for Au/ZnO and T50 = 74 °C for the least active the Au/
ZrO2 catalyst [40]. Based on these data, Comotti et al. suggested
that the reducibility of the support is not the decisive factor for
CO oxidation activity. Qualitatively, our results agree reasonably
well with those findings, except for the Au/Al2O3 catalyst, which
had an outstandingly high activity in that earlier study, with a
T50 value comparable to that of the Au/TiO2 catalyst, while in the
present study its activity was distinctly lower than that of all other
catalysts. Slight differences in the pre-treatment procedures (acti-
vation during reaction in the first study and O250 pre-treatment in
the present study) and in the reaction conditions (lower oxygen
concentration and differential reaction conditions in the present
study) may explain the reversed order in the catalytic activities
of Au/ZnO and Au/ZrO2, but cannot explain the large difference
in activities measured for the Au/Al2O3 catalyst in the two studies.
The low activity of the Au/Al2O3 catalyst was reproduced for sev-
eral batches synthesized independently, and also under conditions
comparable to those being used in the above-mentioned study.
The results obtained over the different supported catalysts
agree perfectly with the trends proposed by Schubert et al., who
distinguished between ‘active’ and ‘inert’ support materials
depending on their reducibility, with the inert materials being
not or hardly reducible [29]. They came to the conclusion that
the active supports participate in the catalytic process, most likely
by the adsorption and activation of oxygen at oxygen vacancies on
the support material close to the Au particles or at the interface
3.3. Pulse experiments
Possible relationships between activity and reducibility of the
different catalysts were tested by pulse experiments in a TAP reac-
tor, which allow us to quantify the amount of the active oxygen
(OSC) stored under reduced pressure conditions [33,41,42] (Sec-
tion 3.3.1). Moreover, the catalytic activity for CO oxidation was
tested also under the same conditions as used in the OSC measure-
ments (Section 3.3.2).
3.3.1. Oxygen storage capacity
The oxygen storage capacity of all four catalysts was measured
at 120 °C reaction temperature in multi-pulse experiments,
Table 2
Oxygen storage capacity (OSC), catalytic activity for CO oxidation, and amount of carbon-containing surface species accumulated on the surface during reaction at 120 °C on the
differently supported Au catalysts after calcination (O250).
Au/TiO2
Au/ZrO2
Au/ZnO
Au/Al2O3
OSC (1018 O atoms g
)
À1
1.5
1.2
1.2
1.1
0.9
3.2
<0.1
–
cat
Additional amount of active oxygen present after O250 calcination relative to re-oxidation by O2 pulsing
(1018 O atoms g
)
À1
cat
Additional amount of active oxygen present after O250 calcination relative to steady-state during single-pulse
1.1
1.0
4.0
–
measurements (1018 O atoms g
)
À1
cat
1.1 Â 10À4 7.7 Â 10À6 4.1 Â 10À6 5.7 Â 10À7
1.1 Â 10À2 7.7 Â 10À4 4.1 Â 10À4 7.1 Â 10À5
7.1
53%
<0.2
À1
À1
À1
Reaction rate (plug-flow reactor) (mol g
Reaction rate (plug-flow reactor) (mol g
s
s
)
)
cat
À1
Au
TOF (plug-flow reactor) (sÀ1
)
0.37
28%
7.7.
0.24
11%
2.7
0.05
<2%
<0.2
Relative CO conversion under steady-state conditions (TAP reactor)
Carbon-containing surface species under steady-state conditions (TAP reactor) (1018 molecules CO2 gcÀa1t
)

D. Widmann et al. / Journal of Catalysis 276 (2010) 292–305
297
exposing the O250 pre-treated catalysts alternately to sequences of
200 CO/Ar pulses and 100 O2/Ar pulses, starting with the CO/Ar
pulses. It was ensured that after these numbers of pulses there is
no further uptake of CO or O2, i.e., further reduction or oxidation
of the corresponding catalysts was below the detection limit under
these conditions. This reduction–oxidation cycle was repeated at
least three times on each catalyst for determining the amount of
active oxygen which is reversibly stored on the catalyst surface,
i.e., which can be removed by reaction with CO during CO pulses
and re-deposited by O2 pulses. The corresponding mass spectro-
metric signals recorded over the different catalysts during these
measurements are reproduced in Fig. 4, which shows pulses of
the reactants CO/Ar and O2/Ar, respectively, and of the CO2 signal
obtained during the CO pulses. Since the uptake of CO and O2,
which is indicated by the missing intensity in the pulse response,
is biggest at the beginning of each pulse sequence, only the first
20 pulses of CO/Ar (out of 200 pulses) or O2/Ar (out of 100 pulses)
are shown here. (It should be noted that in all sequences, the first
pulse is somewhat lower in intensity than the subsequent ones. For
quantitative evaluation, this is corrected for by comparison with
the internal Ar standard.) The qualitative behavior for CO con-
sumption and for oxygen consumption is similar for all samples:
at the beginning of each pulse sequence, there is almost 100% con-
version of CO and O2. Afterward, the consumption of the respective
reactants decreases with ongoing change of the catalyst oxidation
state, until there is no more uptake or conversion of reactant de-
tected. At this point, the catalyst has reached a state which we de-
fine as its fully re-oxidized or reduced state, respectively. The
consumption in CO and O2 at the beginning of each sequence is
clear evidence for the ability of the catalyst to store oxygen upon
O2 exposure and to remove it by reaction with CO. While being
reactive toward CO, this oxygen species is stable against desorption
on these catalysts. It should be stressed that from these data alone
we cannot conclude on the nature of the active oxygen species; in
particular, we cannot decide whether this is an atomic or molecu-
lar oxygen species.
For CO2, the results differ considerably, depending on the sup-
port used. In general, CO2 formation is detected only during CO
pulses, but the shape and intensity of the CO2 signals varied con-
siderably. For the Au/TiO2 catalyst, we obtained distinct and sharp
pulse responses, while for the Au/ZnO catalyst the signals are very
broad. For the Au/Al2O3 catalyst, the CO2 signals are much less in-
tense and close to the background signal, and for the Au/ZrO2 cat-
alyst, a quantitative evaluation was no more possible since the
signals were essentially buried in the noise of the background
intensity. Since there is obviously consumption of CO during the
pulse sequence with the CO/Ar mixture, we expect also for this cat-
alyst the formation of CO2, due to the reaction of CO with surface
oxygen. The absence of a CO2 signal is explained by a rather strong
effective interaction between CO or CO2 and the ZrO2 support,
which leads to an accumulation of carbon-containing species on
the surface or at least to a very prolonged pulse response that can-
not be resolved any more. These data illustrate the differences in
the effective interaction strength between CO2 and the different
supports, which is lowest for Au/TiO2 and distinctly higher for
the Au/ZnO and Au/ZrO2 catalysts (further discussion see
Section 3.3).
The accumulated, absolute amounts of CO molecules converted
or O2 molecules adsorbed during the multi-pulse experiments over
the different Au catalysts are plotted in Fig. 5. For all catalysts, the
overall uptake and conversion of CO was higher during the first se-
quence of CO/Ar pulses, which was dosed on a freshly calcined cat-
alyst, than in the following sequences. Hence, on all catalysts,
thermal oxidation of the catalyst in a O2/N2 flow at 250 °C and
atmospheric pressure results in a higher amount of active oxygen
stored on the catalyst surface than can be obtained upon re-oxida-
tion by O2 pulses. Similar observations were made by Kotobuki
et al. [33] and by Tost et al. [42] during identical measurements
Au/ZrO2
2O3
Au/TiO2
30
20
10
0
CO
CO
CO
CO
CO2 x10
CO2 x10
CO2 x10
CO2
4
2
0
0
5
10
15
20
0
5
10
15
20 0
5
10
15
20 0
5
10
15
20
CO/Ar pulses
O2
O2
O2
O2
20
10
0
0
5
10
15
20
0
5
10
15
20 0
5
10
15
20 0
5
10
15
20
O2/Ar pulses
Fig. 4. Pulse responses during the multi-pulse experiments at 120 °C on the four differently supported Au catalysts after calcination (10% O2/N2, 250 °C, 2 h – O250) for
determination of the oxygen storage capacity. Since the changes are biggest at the beginning of each sequence, only the first 20 pulses each of CO/Ar pulses (out of 200) and of
O2/Ar pulses (out of 100) are shown.

298
D. Widmann et al. / Journal of Catalysis 276 (2010) 292–305
(a) Au/TiO2
(b) Au/ZrO2
4
3
2
1
0
(c) Au/ZnO
(d) Au/Al2O3
4
3
2
1
0
0
50
100
150
200
0
50
100
150
200
Pulses CO/Ar respective O /Ar
2
Fig. 5. Total amounts of CO (first sequence: d; following sequences: s) and O2
catalysts after calcination (10% O2/N2, 250 °C, 2 h – O250).
(D
) consumed during the multi-pulse experiments on the four differently supported Au
on Au/TiO2 catalysts based on porous and non-porous TiO2 after
calcination at 400 °C. The amounts of additionally consumed sur-
face oxygen during the first sequence of CO pulses, which reflects
the additional amount of oxygen deposited during O250 pre-treat-
ment compared to deposition during re-oxidation by O2 pulsing at
catalysts and at the same time their surface areas are comparable
(Au/ZnO) or even by a factor of about four higher (Au/ZrO2 and
Au/Al2O3), the relative amounts of active surface oxygen stored
reversibly on the other catalyst surfaces are even lower, with
0.2% for the Au/ZnO, 0.1% for the Au/ZrO2, and well below 0.1%
for the Au/Al2O3 catalyst, respectively. From these findings, it is
obvious that the overall surface area of the oxide support, and
hence the overall amount of surface oxygen, is not decisive for
the oxygen storage capacity of the corresponding catalysts. Follow-
ing a recent proposal by Kotobuki et al. for Au/TiO2 catalysts [33],
we suggest that on all of these catalysts oxygen storage takes place
at the perimeter sites of the interface between Au nanoparticles
and support. The previous proposal was based on the observation
that for Au/TiO2 catalysts with different Au particle sizes but con-
stant Au loading and identical support material (P25), the OSC
scaled with the number of Au atoms at the perimeter of the inter-
face between Au particles and TiO2 support [33].
It is important to realize that these amounts of oxygen stored on
the different catalysts cannot be detected by any technique under
continuous reaction conditions, since the continuous gas flow in
these measurements is too high and therefore the time required
to deplete or refill the active oxygen is much too short to resolve
the additional initial CO or O2 consumption [53]. For example, in
the activity measurements at atmospheric pressure described in
Section 3.2, the gas flow for CO or O2, respectively, is about
1.6 Â 1019 molecules minÀ1 (60 Nml minÀ1, 1 kPa CO or 1 kPa O2,
respectively). Hence, when using 50 mg of catalyst (normally be-
tween 0.5 and 60 mg catalyst were used), it would take between
0.3 s (Au/TiO2) and <0.025 s (Au/Al2O3) CO exposure to remove
the reversibly stored oxygen on these catalysts, while the time res-
olution in the GC measurements is about 17 min (one sample taken
every 17 min). For mass spectrometric detection, the time resolu-
tion would be significantly better, but also there the time constants
of the reactor and detection set-up would prevent the detection of
120 °C, differ for the different catalysts. They are 1.2 Â 1018
,
1.1 Â 1018, and 3.2 Â 1018 O atoms g for the Au/TiO2, Au/ZrO2,
À1
cat
and Au/ZnO catalysts, respectively, and at the limit for quantifica-
tion (<0.2 Â 1018 O atoms g ) for the Au/Al2O3 catalyst. After this
À1
cat
first sequence, the accumulated amounts of CO converted and O2
adsorbed during two following sequences equal each other, reflect-
ing reversible reduction and oxidation of the catalyst surface. This
amount of oxygen, which is reversibly stored on the catalyst and
which can be reversibly removed/replenished by sequences of CO
pulses or O2 pulses, respectively, is defined as the oxygen storage
capacity (OSC). It is highest for Au/TiO2 (1.5 Â 1018 O atoms g
)
À1
cat
and almost zero for Au/Al2O3 (below 0.2 Â 1018 O atoms g ). The
À1
cat
values for Au/ZnO (0.9 Â 1018 O atoms g
)
and Au/ZrO2
À1
cat
(1.2 Â 1018 O atoms g ) are in between, with a slightly higher
À1
cat
OSC for the latter catalyst (see also Table 2). All samples were mea-
sured under identical conditions (pulse size, catalyst bed packing
etc.) to enable quantitative comparison.
Because of the similarity in Au loadings, Au particle sizes, and
reaction conditions, the differences in the OSC must be related to
the different support materials and their ability to reversibly re-
lease and store active oxygen, either via adsorption/desorption of
oxygen species or via incorporation in and release from the surface
lattice. Similar to previous findings [41], the active oxygen stored
on the catalysts is only a very small fraction of the total amount
of the lattice surface oxygen present on the respective samples. It
is about 0.3% for the Au/TiO2 catalysts under these conditions, in
good agreement with recent findings for a P25-supported Au/
TiO2 catalyst (1%) [33], considering the different Au loadings of
1 wt.% and 3.4 wt.%. Since the OSC is even lower for the other

D. Widmann et al. / Journal of Catalysis 276 (2010) 292–305
299
the active oxygen stored on the catalyst surface. Accordingly, an
earlier attempt to detect and quantify the OSC via transient IR mea-
surements had failed [53].
Au/TiO2
Au/ZrO2
Au/ZnO
Au/Al2O3
1.0
0.5
0.0
In our previous study, the local coverage of reversibly stored
oxygen, at 80 °C reaction temperature and after ex situ condition-
ing at 400 °C, was calculated to be 89% relative to the number of
Au atoms at the perimeter of the interface between Au particle
and TiO2 support, assuming hemispherical Au nanoparticles and
oxygen atoms as the stored species [33]. Here, the local oxygen
coverage in the oxidized Au/TiO2 catalyst is slightly higher (96%),
but of similar order of magnitude, independent of the different
Au loading (3.4 wt.% in the previous study, 1.0 wt.% in this study).
Therefore, we assume that also for the Au/TiO2 catalyst used in the
present work, the active oxygen is located at the perimeter of the
interface between Au particles and oxide support. A similar mech-
anism is suggested also for the other catalysts. Here, the corre-
sponding local active oxygen coverages are lower, obtaining 39%
(Au/ZrO2), 48% (Au/ZnO), and below the detection limit (uncer-
tainty about 10%) for the Au/Al2O3 catalyst.
0
1
2
3
4
5
Time / sec
Fig. 6. Pulse shapes of the height-normalized signals of Ar during simultaneous
pulses CO/Ar and O2/Ar for determination of the activity for CO oxidation over the
differently supported catalysts after calcination (O250).
These trends in OSC and in local oxygen coverage agree well
with the results of the activity measurements shown in Fig. 3
and, except for Au/ZrO2, with expectations based on the reducibil-
ity of the oxides. Hence, for these different catalysts, there is a clear
correlation between activity, OSC and, by and large, the reducibility
of the support material of the different Au catalysts, indicating that
(i) the deposition and reactive removal of the active oxygen de-
tected in these measurements is a decisive step in the CO oxidation
reaction on these catalysts and that (ii) the OSC of the different cat-
alysts is related to the reducibility of the support. It should be
noted that for the pure TiO2 support we could not detect any active
oxygen storage under these experimental conditions, i.e., the OSC
is clearly related to the presence of the Au nanoparticles.
CO
10
5
0
O2
8
4
0
CO2
0.6
3.3.2. Activity measurements
The catalytic activity for CO oxidation under reduced pressure
was measured by pulsing O2/Ar and CO/Ar simultaneously over
the calcined catalyst samples (CO:O2 = 1:1) at 120 °C reaction tem-
perature. By comparing the consumption of CO and O2, reaction-in-
duced changes of the oxidation state of the catalyst surface
compared to the state after pre-treatment were monitored
[33,41]. Furthermore, the possible accumulation of carbon-contain-
ing species on the surface and their stability and dynamic behavior
during the reaction were determined from differences between the
consumption of CO and the formation of CO2 (Section 3.3.3).
Prior to the measurements, it was checked that under the con-
ditions used in these experiments the pulse shapes and hence the
residence times of the educt gases in the catalyst bed are compara-
ble for the different catalysts, despite of the distinct differences in
physical properties such as the surface area. This is important,
since the conversion of CO and O2 depends strongly on the partial
pressure and on the time needed for a pulse to propagate com-
pletely through the catalyst bed (‘contact time’) [42]. Fig. 6 shows
the signals of the Ar component, normalized to the same height,
during the activity measurements over the different catalysts.
Obviously, the residence times are comparable for all four
catalysts.
A representative example of the pulse responses for CO, O2, Ar,
and CO2 is given in Fig. 7, which shows the respective mass spec-
trometric responses during 100 simultaneous pulses of CO/Ar
and O2/Ar over the Au/ZnO catalyst. It is obvious already from
the raw data that the consumption of CO is initially rather high,
but decreases rapidly, whereas that of O2 is almost constant over
the whole experiment. The formation of CO2 is also highest at
the beginning of the reaction and decreases afterward, until reach-
ing a steady-state situation. The results of a quantitative evaluation
of similar pulse sequences for all four catalysts are presented in
0.3
0.0
Ar
20
10
0
0
1000
2000
Time / sec
3000
4000
Fig. 7. CO, O2, CO2, and Ar signal recorded during a sequence of 100 simultaneous
pulses CO/Ar and O2/Ar dosed on the Au/ZnO catalyst pre-treated by calcination
(O250).
Fig. 8, showing the number of CO and O2 molecules consumed
per pulse as well as the number of CO2 molecules formed. The
traces illustrate that for all catalysts, the CO consumption is high-
est at the beginning and decreases with ongoing pulse number, un-
til reaching constant conversion after about 20–40 pulses. Also the
characteristics of the O2 consumption are similar for all catalysts,
starting at almost zero consumption (all O2 molecules admitted
within one pulse exit the reactor) and increasing until reaching
steady-state conditions, where stoichiometric consumption of oxy-
gen and CO is achieved. The higher CO consumption compared to
the O2 consumption in the initial period of the reaction over all
four catalysts indicates a depletion of active oxygen, equivalent
to a reduction of the catalysts after the oxidative pre-treatment.
Qualitatively similar observations were reported recently for dif-
ferent Au/TiO2 catalysts [33,42] and for a Au/CeO2 catalyst [41].

300
D. Widmann et al. / Journal of Catalysis 276 (2010) 292–305
(b) Au/ZrO2
(a) Au/TiO2
4
2
0
4
(d) Au/Al2O3
(c) Au/ZnO
2
0
0
20
40
60
80
100
0
20
40
60
80
100
Simultaneous pulses CO/Ar and O2/Ar
Fig. 8. CO uptake (d), O2 uptake (
D
), and CO2 formation (e) during admission of simultaneous CO/Ar and O2/Ar pulses at 120 °C to the differently supported Au catalysts after
calcination (O250). The dashed lines indicate the number of consumed CO/produced CO2 molecules for complete conversion of the CO pulses.
By adding up the differences between CO consumption and O2
consumption until steady-state is reached with its stoichiometric
consumption of CO and O2, we obtain the absolute amounts of ac-
The order in activity of the differently supported catalysts in
these TAP reactor measurements is similar to that determined in
the plug-flow reactor measurements. A quantitative comparison
of these two measurements, however, is not possible due to the
very different reaction conditions: while we have real steady-state
conditions in a continuous flow of gases at atmospheric pressure in
the plug-flow reactor measurements, which allow us to calculate
reaction rates and TOF numbers, we have a transient input of gases
under vacuum conditions in the TAP reactor. Under those condi-
tions, reaction rates are not easily accessible, and we can only cal-
culate the conversion per pulse. Furthermore, the numbers of
educt molecules admitted to the catalysts per second and hence
their partial pressures are very different in the two measurements.
Nevertheless, the fact that the order of the activities is the same in
both experiments indicates that at atmospheric pressure the sup-
port participates in a similar way in the reaction as in the pulse
experiments under low pressure conditions.
Already in 1989, Haruta et al. had proposed that the supporting
oxides may play an important role in the CO oxidation over sup-
ported Au catalysts, either via metal–support interactions or via
a bifunctional mechanism in which gold particle and support oxide
activate different steps of the CO oxidation [54]. A distinct effect of
the supporting oxide was derived also by Schubert et al., which led
them to distinguish between ‘active’ and ‘inert’ supports and to
tentatively propose sites at the interface between Au nanoparticles
and oxide substrate as active sites [29]. In subsequent theoretical
studies of the CO oxidation reaction on TiO2(1 1 0)-supported Au
nanorods by Hu and coworkers [30] and by Hammer and cowork-
ers [31], these authors found that molecular O2 could be stable ad-
sorbed at interface sites, stabilized by interactions with an under-
coordinated Au atom and a Ti4+ cation, while on Au surfaces the
adsorption energy was negligible. Furthermore, this molecular O2
species was found to be highly activated and able to react with
CO adsorbed on the Au nanorod with a rather low reaction barrier
of 0.1–0.2 eV [30,31]. Moving onto distinct Au10 bilayer clusters
supported on rutile TiO2(1 1 0), Remediakis et al. postulated two
tive oxygen removed from the surface during the reaction. These
are 1.1 Â 1018 O atoms g
for the Au/TiO2 catalyst, 1.0 Â
À1
cat
1018 O atoms g for Au/ZrO2, 4.0 Â 1018 O atoms g for Au/ZnO,
À1
À1
cat
cat
and below the detection limit for the Au/Al2O3 catalyst (see Ta-
ble 2). We should keep in mind that the single-pulse sequences
started on freshly calcined catalyst samples, which have higher
amounts of active oxygen on the catalyst surface than after expo-
sure to CO and subsequent re-oxidation by O2 pulsing in the mul-
ti-pulse experiments. The above numbers for active oxygen
removal until reaching steady-state are very close to the numbers
obtained for the additional active oxygen stored after calcination
when compared to the state reached after re-oxidation by O2 puls-
ing. Therefore, the steady-state situations during simultaneous CO
and O2 pulses are very close to those obtained after O2 pulsing
alone, and hence the oxidation states of the catalyst surfaces dur-
ing the mixed pulses are very close to those of the re-oxidized cat-
alysts obtained upon O2 pulsing.
In contrast to the characteristics of CO and O2 consumption,
that for CO2 formation differs distinctly between the differently
supported catalysts. Furthermore, there are more or less pro-
nounced differences between CO consumption and CO2 formation,
depending on the catalyst (discussion see Section 3.3.3).
For all catalysts, stable reaction conditions with essentially stoi-
chiometric CO and O2 consumption and equivalent CO2 formation
were reached after 100 pulses. Based on the resulting steady-state
conversions, the activities of the different Au catalysts decrease in
the order Au/TiO2 > Au/ZrO2 > Au/ZnO > Au/Al2O3, with almost no
CO conversion for the Au/Al2O3 catalyst. The relative conversions
for CO per pulse are 53% for Au/TiO2, 28% for Au/ZrO2, 11% for
Au/ZnO, and about 2% for Au/Al2O3, which is close to the detection
limit (see also Table 2). Since the Au content and also the mean Au
particle size were almost the same for the different catalysts, also
these measurements indicate a pronounced support effect.

D. Widmann et al. / Journal of Catalysis 276 (2010) 292–305
301
different pathways for the CO oxidation over supported Au cata-
lysts, a ‘gold-only mechanism’ and a ‘metal/oxide boundary mech-
anism’ [32,55]. Their calculations showed stable adsorption of O2
species both at the interface between the Au10 cluster and rutile
TiO2(1 1 0) and also on the Au10 cluster, though with different sta-
bilities. Subsequent reaction with COad was found to be possible on
both sites. In the ‘gold-only mechanism’, reaction takes place only
on the Au nanoparticles, at low-coordinated Au sites such as cor-
ners or edges. This mechanism was proposed to be active on all
Au catalysts, independent of the support material. Nevertheless,
the support can indirectly affect the catalytic performance also in
the ‘gold-only mechanism’ via support-induced modifications of
the Au nanoparticles, e.g., via charge transfer or strain effects. In
the ‘metal/oxide boundary mechanism’, oxygen is adsorbed and
activated at the interface between Au and the metal oxide support.
Due to the charge transfer to the adsorbed O2 species, this results
in negatively charged O2 species. Such superoxide (and peroxide)
species had already been proposed to represent the active oxygen
species for CO oxidation by Liu et al. for a Au/Fe2O3 catalyst [34].
Furthermore, they were identified on different metal oxide-
supported Au catalysts by electron spin resonance (ESR) [56] and
Raman spectroscopy measurements [35,57]. The contribution of
this pathway depends strongly on the chemical nature of the
support, which directly participates in the catalytic reaction.
Depending on the reaction conditions, on the preparation and
pre-treatment procedure and on the structural and chemical prop-
erties of the catalyst (Au particle size, support particle size, and
chemical nature of the support), the one or the other pathway
was proposed to dominate [32,55].
In a recent quantitative TAP study on the OSC and CO oxidation
behavior over P25-supported Au/TiO2 catalysts with different Au
particle sizes but identical Au loading, we found the OSC and also
the activity for CO oxidation to scale approximately with the
length of the perimeter of the interface between Au nanoparticles
and TiO2 support [33]. This finding provided clear evidence that ac-
tive oxygen can be adsorbed on sites along the perimeter of the
Au–oxide interface. Similar evidence came also from a study of
the OSC and CO oxidation behavior on Au/CeO2 [41]. The nature
of the active oxygen species, atomic or molecular oxygen, could
not be derived from those measurements. In view of the findings
of the theoretical studies described above, which revealed that sta-
ble adsorbed molecular oxygen species can be formed at perimeter
sites it may be tempting to favor a reaction mechanism proceeding
via a molecularly adsorbed O2 species, as described by the ‘metal/
oxide boundary mechanism’ [32,55]. The atomic Oad species result-
ing from reaction between COad and O2,ad (CO induced O2,ad disso-
ciation) is highly reactive even on extended Au surfaces [58] and
also on TiO2(1 1 0)-supported Au clusters [32,55,59–62] and will
react rapidly with coadsorbed CO. On the other hand, however,
the observation that the active oxygen species is stable at least
up to temperatures of 120 °C is hardly compatible with a molecu-
larly adsorbed oxygen species and with the adsorption energies
calculated in theoretical studies [55]. According to calculations
by Laursen and Linic [62], atomic oxygen species are most stable
bound at the Au/TiO2 interface, while other sites on the Au nano-
structure (2-layer-thick Au nanorods on TiO2(1 1 0)) may be avail-
able for CO adsorption. It should be noted that atomically adsorbed
oxygen on Au or Au model catalysts has been demonstrated previ-
ously to be highly reactive toward CO [58–60]. On the other hand,
the most important open question, the reaction pathway for O2
dissociation, has not been addressed by Laursen and Linic and is
still unresolved. Because of the high thermal stability of the active
oxygen species, we also favor an atomic species located at the
perimeter of the Au/oxide interface as active species.
propose that also for the other Au catalysts the activation of oxy-
gen takes place at the perimeter of the Au–oxide interface (‘me-
tal/oxide boundary mechanism’), at least in the dominating
reaction pathway. Accordingly, the differences in energy levels
and barriers in this reaction pathway are responsible for the differ-
ing activities of these catalysts. Only for the Au/Al2O3 catalyst with
its very little OSC, the contribution from the ‘gold-only mecha-
nism’, which should be of similar order of magnitude for all cata-
lysts investigated, may be dominant.
The correlation between OSC and activity on the one hand and
reducibility of the support on the other hand derived in the present
study agrees with trends reported earlier [29]. In the latter mea-
surements, however, the similarity in Au particle size and compo-
sition (Au0) and Au loading of the different catalysts was much less
well controlled than in the present study.
The reducibility of the support may affect the catalyst and its
performance in different ways. Oxygen vacancy defects were spec-
ulated to improve O2 adsorption on the support or at the support–
oxide interface [63,64] or to change the charge state of the Au
nanoparticles [49]. In previous calculations, oxygen vacancies at
the interface were found to significantly enhance the interaction
between oxide and Au nanoparticles, e.g., for Au–TiO2 [65,66] or
for Au–MgO [31,36,67]. On the other hand, the effect on the
adsorption energies and reaction barriers was found to be rather
small [32,55]. Alternatively, the stronger interaction of O2 with
highly charged cations such as Ti4+ was held responsible for the
high activity of Au catalysts supported on reducible oxides such
as TiO2 or V2O5 [30]. Experimentally, Carretin et al. found that upon
doping the supporting TiO2 with iron, the concentration of oxygen
vacancies, as estimated from the band-gap transition of TiO2 in
UV–VIS spectra, the number of adsorbed superoxide and peroxide
species after O2 exposure (Raman spectroscopy) and the catalytic
activity for CO oxidation increased [35]. In the present experi-
ments, surface oxygen vacancies may be generated during CO puls-
ing, though definite proof cannot be given. During reaction, after
reaching steady-state conditions, however, the catalyst surface
was determined to be in a state close to the ‘fully oxidized state’,
where the existence of significant amounts of oxygen vacancies
is unlikely.
In total, these experiments clearly support a CO oxidation reac-
tion mechanism involving direct participation of the support (‘me-
tal/oxide boundary mechanism’). Despite of the evidence in
previous theoretical work, where mainly stable adsorbed molecu-
lar O2 species were found at interface perimeter sites, we tend to
favor an atomic species adsorbed at the perimeter of the Au–oxide
3
Au/TiO2
2
Au/ZrO2
1
Au/ZnO
Au/Al2O3
0
0.0
0.5
1.0
1.5
-1
OSC / 1018O atoms gcat
Fig. 9. Absolute amounts of CO converted/CO2 produced in steady-state during
simultaneous pulses CO/Ar and O2/Ar on the differently supported Au catalysts
plotted against the oxygen storage capacity of the corresponding catalysts, both
after pre-treatment by calcination (O250).
Considering the clear correlation between OSC and CO oxida-
tion activity in the present measurements illustrated in Fig. 9, we

302
D. Widmann et al. / Journal of Catalysis 276 (2010) 292–305
interface as active oxygen species because of its high stability
determined experimentally, being stable toward desorption at
least up to 120 °C. Final proof, however, is still missing. The pres-
ence of significant amounts of oxygen vacancies during reaction,
under steady-state conditions, is unlikely, though it cannot be fully
ruled out.
(a)
Au/TiO2
Au/ZrO2
Au/ZnO
3
2
1
0
3.3.3. Accumulation of carbon-containing surface species
It is well known from previous studies that carbon-containing
surface species are built up on oxide-supported catalysts during
CO oxidation [3,68–70]. Depending on the nature of the support,
these species may either be stable adsorbed surface species such
as surface carbonates, acting as reaction intermediates or spectator
species, and/or re-adsorbed product CO2. Using the TAP reactor, it
is possible to exactly quantify the amount of these species,
although their nature cannot be identified.
200
400
600
800
Temperature / °C
(b)
For the more active catalysts investigated, we found a clear dif-
ference between CO consumption and CO2 formation in the single-
pulse experiments in the initial phase of the reaction. The discrep-
ancy between CO consumption and CO2 formation, however, varies
distinctly for the different catalysts. While for most catalysts, the
amount of gaseous CO2 product reaches its highest value within
the first two pulses and decreases with ongoing pulse number, it
increases steadily during the entire experiment for the Au/ZrO2
catalyst. At the same time, the consumption of CO decreases for
all catalysts in the initial period of the reaction, until reaching a
steady-state. The number of pulses required for reaching a stable
situation with similar values for CO consumption and CO2 forma-
tion differs for the different catalysts. While this is essentially
reached after only 2 pulses for the Au/TiO2 catalyst, it takes about
50 pulses for the Au/ZnO catalyst and almost 100 pulses for the Au/
ZrO2 catalyst. Due to the very low amount of CO consumption and
CO2 formation on the Au/Al2O3 catalyst, the time for reaching stea-
dy-state could not be determined for this catalyst. These results
clearly indicate that some kind of carbon-containing species is
formed on the catalyst surface during the reaction, either stable
compounds such as surface carbonates, which are built up during
reaction or by re-adsorption of the product CO2. Adding up the dif-
ferences between CO consumption and CO2 formation, we calcu-
lated the number of carbon-containing surface species formed
Au/TiO2
Au/ZrO2
Au/ZnO
30
20
10
0
200
400
600
800
Temperature / °C
Fig. 10. Mass spectrometric signals for (a) mass 45 (13CO2) and (b) mass 44 (12CO2)
recorded during temperature-programmed desorption (TPD) directly after reaction
with 13CO/O2/Ar pulses over the Au/TiO2, Au/ZrO2, and Au/ZnO catalysts at 120 °C
after calcination (O250).
et al. proposed for CO oxidation over a Au/TiO2 catalyst that the
interaction of CO2 product molecules with the support may lead
to the transient formation of surface carbonates [53]. These species
should decompose slowly to CO2 which then may re-adsorb again.
Based on these results, we attribute the first peak to desorption of
adsorbed CO2, while the second one is tentatively assigned to the
decomposition of surface carbonates. The tendency for CO2 adsorp-
tion is low for Au/TiO2, despite of its high activity for CO2 forma-
tion, slightly higher for Au/ZnO and clearly most pronounced for
Au/ZrO2, where this is the dominant carbon-containing surface
species present on the surface during/after reaction. On the other
hand, for Au/ZnO, the second peak dominates and it is even higher
than that on Au/ZrO2, where also considerable amounts of these
species were formed, while for Au/TiO2 also this peak has a very
low intensity. Hence, on Au/ZnO and in particular on Au/ZrO2,
CO2 is accumulated by re-adsorption in the catalyst (desorption
ꢂ235 °C). In addition, more stable surface compound species are
formed (decomposition at 330–360 °C), and here the amount of
these species is highest for Au/ZnO species, lower for Au/ZrO2,
and much lower for Au/TiO2.
In addition to these species, there are considerable amounts of
other stable carbon-containing species on the surface, which were
not removed by the pre-treatment procedure. This is evident from
the TPD traces recorded on mass 44 (12CO2) during the same TPD
measurements over the three catalysts (Fig. 10b). Since the pre-
treatment involved calcination at 250 °C, measurable desorption
occurs only above that temperature. The data reveal distinct differ-
ences in the desorption characteristics between the different cata-
lysts. Also here, the amount of carbon-containing surface species is
much higher for the Au/ZnO and Au/ZrO2 catalysts than for the Au/
during the reaction to be equivalent to <2.0 Â 1017, 7.7 Â 1018
,
and 2.7 Â 1018 CO2 molecules g for the Au/TiO2, Au/ZrO2, and
À1
cat
Au/ZnO catalyst, respectively (see Table 2). For the Au/Al2O3 cata-
lyst, a quantitative evaluation was not possible due to the very low
values for CO consumption and CO2 formation. One should note
that there is no correlation between the additional amount of ac-
tive oxygen present on the catalysts surface after O250 treatment,
as indicated by the difference in CO consumption and O2 consump-
tion, and the formation of adsorbed carbon-containing species, as
indicated by the difference in CO consumption and CO2 formation
(for comparison of absolute numbers see Table 2). Therefore, these
two effects are discussed separately.
The accumulation of carbon-containing surface species was ver-
ified by TPD measurements, performed directly (5 min) after the
single-pulse CO oxidation sequences. To distinguish the carbon-
containing species from those already present before the reaction,
we used labeled 13CO for the single-pulse measurements. The mass
45 TPD traces (13CO2) resulting from the Au/TiO2, Au/ZnO, and Au/
ZrO2 catalysts (Au/Al2O3 was not measured because of the very low
CO2 formation) during these measurements are plotted in Fig. 10a.
First of all, although we could not quantify the results of these
measurements on an absolute scale, the data clearly reveal a
decreasing amount of carbon-containing species in the order Au/
ZrO2 > Au/ZnO > Au/TiO2, confirming the trend determined in the
single-pulse experiments. Second, these spectra exhibit character-
istic peaks at about 200–235 °C and 330–360 °C. Recently, Clark

D. Widmann et al. / Journal of Catalysis 276 (2010) 292–305
303
TiO2 catalyst, in line with the activities for forming such species de-
scribed above (Fig. 10a). Furthermore, we find a characteristic peak
or shoulder at about 350–375 °C, which may result from the same
surface species as the high temperature peak in Fig. 10a, plus a
number of higher temperature peaks which are likely to be related
to other surface species.
The stability and dynamic behavior of the carbon-containing
surface species on the three catalysts during the reaction was ex-
plored by isotope switching experiments (for Au/Al2O3, this was
not done because of the low activity for CO2 formation and hence
for the low build-up of carbon-containing deposits). We first
admitted 50 pulses of a gas mixture of labeled 13CO, O2, and Ar
to the freshly calcined catalysts, and then changed to 50 pulses
of 12CO, O2, and Ar. The CO and O2 conversions and the formation
of the two CO2 isotopomers (13CO2, 12CO2) are shown in Fig. 11. On
the Au/TiO2 catalyst, there is almost no build-up of carbon-con-
taining surface species at the beginning (see also above), and the
small amount of reversibly stored carbon-containing surface spe-
cies present during reaction is exchanged completely within the
first 5 pulses after the switch (100% exchange). This is in good
agreement with expectations based on a weak interaction between
CO2 and TiO2. The situation is very different for the Au/ZrO2 and
the Au/ZnO catalysts. On these catalysts, one can detect isotope-la-
beled 13CO2 molecules even 50 pulses after the switch to 12CO. Cor-
respondingly, the intensity of the new 12CO2 isotopomer increases
only slowly and has not yet reached its final value even after 50
pulses. Until the end of the sequence, only part of the 13C car-
bon-containing surface species accumulated during the preceding
reaction sequence were replaced by their 12C analogs (57% for
Au/ZrO2 and 29% for Au/ZnO). From the fact that 13CO2 formation
is not observed continuously, as one would expect from thermal
desorption, but only upon 12CO/O2 pulses, we can further conclude
that the replacement of adsorbed 13C species by 12C species on the
surface is essentially induced by CO2 rather than just due to ther-
mal desorption of CO2. The third possibility, CO or O2 induced
decomposition, could be excluded by additional pulse experiments
using CO or O2 pulses only. The observation of a slow exchange of
the carbon-containing species during reaction over the Au/ZrO2
and the Au/ZnO catalysts is consistent with the thermal stability
of these species detected in the TPD measurements (desorption
maxima at 200–235 °C and at 330–360 °C). Most likely, mainly
the species in the low-temperature peak (re-adsorbed CO2) are ex-
changed. Considering the relative amounts of exchanged species,
however, the exchange must involve also the high temperature
species (stable surface carbonate species), at least for Au/ZnO,
though presumably on a slower time scale.
The presence of significant amounts of carbon-containing spe-
cies on the surface, also at the beginning of the reaction, and the
very different tendencies for further deposition of these species
during the reaction may be another factor affecting the (initial)
activity of these Au catalysts, in addition to their different inherent
activities (activity of a catalyst free of carbon-containing deposits).
As shown in previous reaction measurements, increasing cover-
ages of carbon-containing surface species result in a lower activity
of the corresponding catalysts [70–73]. Although we expect only a
small fraction of these surface species to be located at the active
interface perimeter sites, a higher coverage of these species on
the support goes along with a higher probability for blocking these
interface sites. Therefore, the relatively low coverage of carbon-
containing surface species on the Au/TiO2 catalyst may be further
reason for the high activity of this catalyst and may at least partly
be responsible for its much higher activity compared to Au/ZrO2,
despite the rather small difference in OSC (factor of 1.2). Moreover,
the weak interaction between CO2 and the TiO2 support can also
explain the extraordinary high activity of Au/TiO2 catalysts already
at very low temperatures (À70 °C), which has been reported previ-
ously [3]. On other, more strongly interacting support materials,
adsorption of CO2, and accumulation of carbon-containing surface
species would lead to a rapid blocking of surface sites and hence to
deactivation. Finally, we want to note that due to the very different
reaction times the build-up of carbon-containing surface species
sensed in the TAP reactor measurements cannot be compared, at
least not quantitatively, with the deactivation detected in micro-
reactor measurements (Fig. 3 and [70]). On time scales of hours
and more, as used in such experiments, carbon-containing surface
species are built up also on Au/TiO2 and lead to a significant deac-
tivation of the catalyst, in particular at lower temperatures (80 °C)
[70,73].
(a) Au/TiO2
4
3
2
1
0
(b) Au/ZrO2
1.5
1.0
0.5
0.0
3
2
1
0
(c) Au/ZnO
0
20
40
60
80
100
Number of Pulses
Fig. 11. CO uptake (d), O2 uptake (D
), and CO2 formation (e: open symbols 13CO2,
4. Conclusions
filled symbols 12CO2) per pulse during isotope switching experiments at 120 °C,
exposing the freshly calcined (O250) (a) Au/TiO2, (b) Au/ZrO2, and (c) Au/ZnO
catalysts first to 50 13CO/O2/Ar pulses and afterward to 50 12CO/O2/Ar pulses.
(Instabilities in the O2 pulses were corrected for in the quantitative evaluation by
comparison with the Ar standard.)
Applying quantitative Temporal Analysis of Products tech-
niques, we could quantify the oxygen storage capacity (OSC) of
four Au catalysts supported on different oxide materials but with

304
D. Widmann et al. / Journal of Catalysis 276 (2010) 292–305
similar Au loadings and Au particle sizes and relate this to the cat-
alytic activity for CO oxidation reaction. This way we could demon-
strate that both OSC and CO oxidation activity depend sensitively
and directly on the nature of the support. These measurements
in combination with reaction rate and TPD measurements led to
the following conclusions:
In total, these experiments and results provide definite proof for
a direct participation of the support in the CO oxidation reaction
over highly active Au/TiO2, AuZrO2, Au/ZnO catalysts, via activation
and storage of active oxygen. Based on previous results for Au/TiO2,
this occurs most likely at the perimeter of the Au–support interface
(‘metal/oxide boundary mechanism’). For Au/Al2O3, such participa-
tion of the interface cannot be identified, and a ‘gold-only mecha-
nism’, which at most represents a minority pathway on the other
catalysts, may be dominant. We suggest that, in addition to the
CO oxidation reaction, these results and findings are important also
in a more general sense, for the mechanistic understanding of oxi-
dation reactions over oxide-supported Au catalysts in general,
where the same active oxygen species are expected to participate
in the reaction.
(1) From the pronounced effect of the support material on the
OSC and the activity for CO oxidation and the close correla-
tion between OSC and CO oxidation activity, with the same
order in both cases (Au/TiO2 > Au/ZrO2 > Au/ZnO > Au/
Al2O3), we propose a similar reaction mechanism for these
catalysts as derived recently for Au/TiO2 [33], with the sup-
port participating directly in the CO oxidation reaction, by
stabilizing and activating adsorbed oxygen at the perimeter
sites of the interface between Au nanoparticle and support
(‘metal/oxide boundary mechanism’). The data clearly sug-
gest that the reactivity is largely controlled by the OSC of
the catalyst. Only for the Au/Al2O3 catalyst with its very
low OSC and CO oxidation activity, the reaction may be dom-
inated by a ‘gold-only mechanism’.
(2) From the pronounced effects of the support and the similar
order of decreasing activity obtained in stationary measure-
ments of the reaction rate in a plug-flow reactor, we con-
clude that this reaction mechanism is not only dominant
under the instationary reaction conditions present in the
TAP reactor but also under stationary reaction conditions
at atmospheric pressure. Correspondingly, the active oxygen
species determined in the pulse experiments is also the
active species during continuous oxidation at atmospheric
pressure.
(3) From the fact that the surface of all catalysts was essentially
in the ‘fully oxidized’ state during steady-state CO oxidation
as obtained upon re-oxidation by O2 pulsing, we conclude
that during reaction contributions from surface vacancies
are small, at least under present reaction conditions. More
significant, the chemical nature of the support may affect
the reaction also via an enhanced formation and stabiliza-
tion of active oxygen species adsorbed at the interface
perimeter sites, by interaction with highly loaded cations
such as Ti4+ [30].
(4) The distinct differences between the Au catalysts in the ten-
dency for accumulation of carbon-containing surface species
during CO oxidation, by re-adsorption of CO2 product mole-
cules and by formation of stable adsorbed surface species
such as surface carbonates, are likely to represent another
important factor for the CO oxidation activity of oxide-sup-
ported Au catalysts, in addition to the differences in the
OSC. Furthermore, the little interaction between CO2 and
Au/TiO2 catalyst is held responsible for its high activity at
low temperatures, where for other Au catalysts the surface
would rapidly be covered by adsorbed CO2 and surface car-
bonates, etc.
Acknowledgment
This work was supported by the Deutsche Forschungsgemeins-
chaft in the Priority Program 1181 (project Be 1201/13-3).
References
[1] G.C. Bond, C. Louis, D.T. Thompson, in: Catalysis by Gold, Imperial Press,
London, 2007. Chapter 6.
[2] M. Haruta, T. Kobayashi, H. Sano, N. Yamada, Chem. Lett. 16 (1987) 405.
[3] M. Haruta, S. Tsubota, T. Kobayashi, H. Kageyama, M.J. Genet, B. Delmon, J.
Catal. 144 (1993) 175.
[4] G.J. Hutchings, Gold Bull. 29 (1996) 123.
[5] G.C. Bond, D.T. Thompson, Catal. Rev. Sci. Eng. 41 (1999) 319.
[6] Q. Fu, H. Saltsburg, M. Flytzani-Stephanopoulos, Science 301 (2003) 935.
[7] G. Jacobs, P.M. Patterson, L. Williams, D. Sparks, H. Davis, Catal. Lett. 96 (2004)
97.
[8] R. Burch, Phys. Chem. Chem. Phys. 8 (2006) 5483.
[9] W. Deng, M. Flytzani-Stephanopoulos, Angew. Chem. 118 (2006) 2343.
[10] M. Haruta, Catal. Surv. Jpn. 1 (1997) 61.
[11] T. Hayashi, K. Tanaka, M. Haruta, J. Catal. 178 (1998) 566.
[12] R.J.H. Grisel, B.E. Nieuwenhuys, Catal. Today 64 (2001) 69.
[13] M.C. Kung, R.J. Davis, H.H. Kung, J. Phys. Chem. C 111 (2007) 11767.
[14] T.V.W. Janssens, B.S. Clausen, B. Hvolbaek, H. Falsig, C.H. Christensen, T.
Bligaard, J.K. Nørskov, Top. Catal. 44 (2007) 15.
[15] M. Haruta, Catal. Today 36 (1997) 153.
[16] M. Valden, X. Lai, D.W. Goodman, Science 281 (1998) 1647.
[17] J.-D. Grunwaldt, M. Maciejewski, O.S. Becker, P. Fabrizioli, A. Baiker, J. Catal.
186 (1999) 458.
[18] M. Mavrikakis, P. Stoltze, J.K. Nørskov, Catal. Lett. 64 (2000) 101.
[19] Z.-P. Liu, P. Hu, A. Alavi, J. Am. Chem. Soc. 124 (2002) 14770.
[20] L.M. Molina, B. Hammer, Phys. Rev. Lett. 90 (2003) 206102.
[21] N. Lopez, T.V.W. Janssens, B.S. Clausen, Y. Xu, M. Mavrikakis, T. Bligaard, J.K.
Nørskov, J. Catal. 223 (2004) 232.
[22] T.V.W. Janssens, A. Carlsson, A. Puig-Molina, B.S. Clausen, J. Catal. 240 (2006)
108.
[23] S.H. Overbury, V. Schwartz, D.R. Mullins, W. Yan, S. Dai, J. Catal. 241 (2006) 56.
[24] E.D. Park, J.S. Lee, J. Catal. 186 (1999) 1.
[25] R.M. Finch, N.A. Hodge, G.J. Hutchings, A. Meagher, Q. Pankhurst, M.R.H.
Siddiqui, F.E. Wagner, R. Whyman, Phys. Chem. Chem. Phys. 1 (1999) 485.
[26] J.C. Fierro-Gonzalez, B.C. Gates, J. Phys. Chem. B 108 (2004) 16999.
[27] J.A. van Bokhoven, C. Louis, J.T. Miller, M. Tromp, O.V. Safonova, P. Glatzel,
Angew. Chem. Int. Ed. 45 (2006) 4651.
[28] L. Delannoy, N. Weiher, N. Tsapatsaris, A.M. Bessley, L. Ncharib, S.L.M.
Schroeder, C. Louis, Top. Catal. 44 (2007) 263.
[29] M.M. Schubert, S. Hackenberg, A.C. van Veen, M. Muhler, V. Plzak, R.J. Behm, J.
Catal. 197 (2001) 113.
[30] Z.-P. Liu, X.-Q. Gong, J. Kohanoff, C. Sanchez, P. Hu, Phys. Rev. Lett. 91 (2003)
266102.
[31] L.M. Molina, M.D. Rasmussen, B. Hammer, J. Chem. Phys. 120 (2004) 7673.
[32] I.N. Remediakis, N. Lopez, J.K. Nørskov, Appl. Catal. A 291 (2005) 13.
[33] M. Kotobuki, R. Leppelt, D. Hansgen, D. Widmann, R.J. Behm, J. Catal. 264
(2009) 67.
[34] H. Liu, A.I. Kozlov, A.P. Kozlova, T. Shido, Y. Iwasawa, Phys. Chem. Chem. Phys.
1 (1999) 2851.
[35] S. Carrettin, Y. Hao, V. Aguilar-Guerrero, B.C. Gates, S. Trasobares, J.J. Calvino, A.
Corma, Chem. Eur. J. 13 (2007) 7771.
[36] A. Sanchez, S. Abbet, U. Heiz, W.-D. Schneider, H. Häkkinen, R.N. Barnett, U.
Landman, J. Phys. Chem. A 102 (1999) 9573.
[37] C.K. Costello, M.C. Kung, H.-S. Oh, Y. Wang, H.H. Kung, Appl. Catal. A 232 (2002)
159.
[38] J.-D. Grunwaldt, C. Kiener, C. Wögerbauer, A. Baiker, J. Catal. 181 (1999) 223.
[39] S. Arrii, F. Morfin, A.J. Renouprez, J.L. Rousset, J. Am. Chem. Soc. 126 (2004)
1199.
(5) From the observation that the adsorbed carbon-containing
species can be reversibly re-converted into CO2 upon expo-
sure to CO/O2 mixtures, which was detected in pulse exper-
iments using isotope-labeled 13CO, we conclude that these
species are not fully stable at the reaction temperature, but
can be desorbed/decomposed. Since thermal desorption/
decomposition of these species occurs only at significantly
higher temperatures, this is attributed to a CO2-assisted
mechanism. Because of the rather slow exchange process,
we propose that they essentially represent spectator species,
which can be formed and decomposed reversibly during the
reaction, rather than reaction intermediates in the dominant
reaction pathway.

D. Widmann et al. / Journal of Catalysis 276 (2010) 292–305
305
[40] M. Comotti, W.-C. Li, B. Spliethoff, F. Schüth, J. Am. Chem. Soc. 128 (2006) 917.
[41] D. Widmann, R. Leppelt, R.J. Behm, J. Catal. 251 (2007) 437.
[42] A. Tost, D. Widmann, R.J. Behm, J. Catal. 266 (2009) 299.
[43] S. Biella, F. Porta, L. Prati, M. Rossi, Catal. Lett. 90 (2003) 23.
[44] M. Comotti, C. Pina, R. Matarrese, M. Rossi, A. Siani, Appl. Catal. A 291 (2005)
204.
[45] M.J. Kahlich, H.A. Gasteiger, R.J. Behm, J. Catal. 171 (1997) 93.
[46] R. Leppelt, D. Hansgen, D. Widmann, T. Häring, G. Bräth, R.J. Behm, Rev. Sci.
Instrum. 78 (2007) 104103.
[47] J.T. Gleaves, G.S. Yablonskii, P. Phanawadee, Y. Schuurman, Appl. Catal. A 160
(1997) 55.
[48] S.O. Shekhtman, G.S. Yablonsky, S. Chen, J.T. Gleaves, Chem. Eng. Sci. 54 (1999)
4371.
[59] T.S. Kim, J.D. Stiehl, C.T. Reeves, R.J. Meyer, C.B. Mullins, J. Am. Chem. Soc. 125
(2003) 2018.
[60] J.D. Stiehl, T.S. Kim, C.T. Reeves, R.J. Meyer, C.B. Mullins, J. Phys. Chem. B 108
(2004) 7917.
[61] J.G. Wang, B. Hammer, Phys. Rev. Lett. 97 (2006) 136107.
[62] S. Laursen, S. Linic, Phys. Chem. Chem. Phys. 11 (2009) 11006.
[63] F. Boccuzzi, A. Chiorini, M. Manzoli, P. Lu, T. Akita, S. Ichikawa, M. Haruta, J.
Catal. 202 (2001) 256.
[64] M. Manzoli, A. Chiorino, F. Boccuzzi, Surf. Sci. 532–535 (2003) 377.
[65] E. Wahlström, N. Lopez, R. Schaub, P. Thostrup, A. Rønnau, C. Africh, E.
Laegsgaard, J.K. Nørskov, F. Besenbacher, Phys. Rev. Lett. 90 (2003) 026101.
[66] N. Lopez, J.K. Nørskov, T.V.W. Janssens, A. Carlsson, A. Puig-Molina, B.S.
Clausen, J.-D. Grunwaldt, J. Catal. 225 (2004) 86.
[49] G.C. Bond, D.T. Thompson, Gold Bull. 33 (2000) 41.
[50] M.C. Kung, C. Costello, H.H. Kung, Catalysis 17 (2004) 152.
[51] A. Wolf, F. Schüth, Appl. Catal. A 226 (2002) 1.
[52] N. Weiher, E. Bus, L. Delannoy, C. Louis, D.E. Ramaker, J.T. Miller, J.A. van
Bokhoven, J. Catal. 240 (2006) 100.
[53] J.C. Clark, S. Dai, S.H. Overbury, Catal. Today 126 (2007) 135.
[54] M. Haruta, N. Yamada, T. Kobayashi, S. Iijima, J. Catal. 115 (1989) 301.
[55] I.N. Remediakis, N. Lopez, J.K. Nørskov, Angew. Chem. 117 (2005) 1858.
[56] Z. Hao, L. Fen, G.Q. Lu, J. Liu, L. An, H. Wang, Appl. Catal. A 213 (2001) 173.
[57] J. Guzman, S. Carrettin, A. Corma, J. Am. Chem. Soc. 127 (2005) 3286.
[58] J.M. Gottfried, K. Christmann, Surf. Sci. 566–568 (2004) 1112.
[67] B. Yoon, H. Häkkinen, U. Landman, A.S. Wörz, M. Antonietti, S. Abbet, K. Judai,
U. Heiz, Science 307 (2005) 403.
[68] M.A. Bollinger, M.A. Vannice, Appl. Catal. B 8 (1996) 417.
[69] F. Boccuzzi, A. Chiorino, S. Tsubota, M. Haruta, J. Phys. Chem. 100 (1996) 3625.
[70] B. Schumacher, Y. Denkwitz, V. Plzak, M. Kinne, R.J. Behm, J. Catal. 224 (2004)
449.
[71] P. Konova, A. Naydenov, Cv. Venkov, D. Mehandjiev, D. Andreeva, T. Tabakova,
J. Mol. Catal. A 213 (2004) 235.
[72] P. Konova, A. Naydenov, T.T. Tabakova, D. Mehandjiev, Catal. Commun. 5
(2004) 537.
[73] Y. Denkwitz, B. Schumacher, G. Kucerova, R.J. Behm, J. Catal. 267 (2009) 78.