Reactivity of molecularly chemisorbed oxygen on a Au/TiO2 model catalyst

Article abstract of DOI:10.1021/jp062766c

We present results of an investigation into the reactivity of molecularly chemisorbed oxygen with CO on a Au/TiO₂ model catalyst at 77 K. We previously discovered that exposing the model catalyst sample to a radio-frequency-generated plasma jet of oxygen results in co-population of both atomically and molecularly ' chemisorbed oxygen species on the sample. We tested the reactivity of the molecularly chemisorbed oxygen by comparing the CO ₂ produced from a sample populated with both species to the CO ₂ produced from a sample that has been cleared of molecularly chemisorbed oxygen employing collision-induced desorption. Samples that are populated with both species consistently result in greater CO₂ produced than samples with only atomic oxygen. We interpret this result to indicate that molecularly chemisorbed oxygen on the sample can directly participate in the CO oxidation reaction. The reactivity of molecularly chemisorbed oxygen has been investigated for five different gold coverages (0.5, 0.75, 1, 1.25, and 2 ML), and we observe that there is a greater fractional difference in the CO₂ produced (difference between sample populated with both molecularly and atomically adsorbed oxygen and sample populated solely with atomically adsorbed oxygen) for the 1 ML Au coverage than for the other coverages for equivalent oxygen plasma-jet exposures. However, it is not possible to unambiguously conclude that this observation is directly related to a particle size effect on the chemistry since the absolute O_2,a and O_a content on the various surfaces is different for all the coverages studied because of the plasma-jet technique that we employed for populating the surfaces with oxygen. Unfortunately, this precludes a direct comparison of the reactivity of molecular oxygen in the carbon monoxide oxidation reaction as a function of gold coverage and hence particle size.

Full text of DOI:10.1021/jp062766c

J. Phys. Chem. B 2006, 110, 20337-20343
20337
Reactivity of Molecularly Chemisorbed Oxygen on a Au/TiO2 Model Catalyst
James D. Stiehl, Jinlong Gong, Rotimi A. Ojifinni, Tae S. Kim, Sean M. McClure, and
C. Buddie Mullins*
The UniVersity of Texas Austin, Department of Chemical Engineering and Texas Materials Institute,
1 UniVersity Station CO400, Austin, Texas 78712-0231
ReceiVed: May 5, 2006; In Final Form: July 31, 2006
We present results of an investigation into the reactivity of molecularly chemisorbed oxygen with CO on a
Au/TiO2 model catalyst at 77 K. We previously discovered that exposing the model catalyst sample to a
radio-frequency-generated plasma jet of oxygen results in co-population of both atomically and molecularly
chemisorbed oxygen species on the sample. We tested the reactivity of the molecularly chemisorbed oxygen
by comparing the CO2 produced from a sample populated with both species to the CO2 produced from a
sample that has been cleared of molecularly chemisorbed oxygen employing collision-induced desorption.
Samples that are populated with both species consistently result in greater CO2 produced than samples with
only atomic oxygen. We interpret this result to indicate that molecularly chemisorbed oxygen on the sample
can directly participate in the CO oxidation reaction. The reactivity of molecularly chemisorbed oxygen has
been investigated for five different gold coverages (0.5, 0.75, 1, 1.25, and 2 ML), and we observe that there
is a greater fractional difference in the CO2 produced (difference between sample populated with both
molecularly and atomically adsorbed oxygen and sample populated solely with atomically adsorbed oxygen)
for the 1 ML Au coverage than for the other coverages for equivalent oxygen plasma-jet exposures. However,
it is not possible to unambiguously conclude that this observation is directly related to a particle size effect
on the chemistry since the absolute O2,a and Oa content on the various surfaces is different for all the coverages
studied because of the plasma-jet technique that we employed for populating the surfaces with oxygen.
Unfortunately, this precludes a direct comparison of the reactivity of molecular oxygen in the carbon monoxide
oxidation reaction as a function of gold coverage and hence particle size.
Introduction
observed prompt evolution of CO2 from the oxygen precovered
sample on exposure to an effusive source of CO over the
Questions regarding the mechanistic details of the CO
oxidation reaction on gold-based catalysts have driven the
research of many scientists since the outset of this unexpected
discovery by Haruta.1 That gold, a metal with a reputation for
being inert in its bulk form,2 could be involved in catalyzing a
reaction, which inherently necessitates the activation of an
oxygen species, was very surprising. Some specific issues that
remain active subjects of research regarding CO oxidation on
Au catalysts are the particle size effect observed for the reaction,
the oxidation state of the gold, CO adsorption, moisture effects,
support effects, catalyst pretreatment, and, what is of interest
in this study, the active oxygen species involved in the
reaction.1,3-5 Despite all the work that has been done concerning
the CO oxidation reaction on gold catalysts, as of yet there seems
to be no satisfactory picture of the reaction mechanism,5
including identification of the active oxygen species.
One possible candidate for the active oxygen species in
catalytically relevant chemistry is atomically adsorbed oxygen
(Oa). Atomic oxygen has been shown to be highly reactive
toward CO when preadsorbed on single-crystal gold samples6-10
and also on Au model catalysts.11-13 One of the first studies of
the reactivity of Oa on gold was performed by Outka et al. on
a Au(110) single crystal.6 Outka et al. populated a Au(110)
single-crystal sample with atomic oxygen by cracking molecular
oxygen on a hot platinum filament near the Au surface. They
temperature range 275-440 K. Recent work by Gottfried et al.
on CO oxidation with preadsorbed oxygen atoms on Au(110)-
(1 × 2) has revealed that the reaction takes place at temperatures
below 100 K with an activation energy of ∼0.6 eV.8,9 Koel
and co-workers also showed that atomically adsorbed oxygen
on Au(111) was active for CO oxidation at temperatures ranging
from 250 to 375 K.7 They populated their samples with atomic
oxygen by exposing the sample to ozone. Gong et al. also
recently studied the reactivity of Oa on a Au(111) single crystal
at 77 K and observed prompt CO2 production over a large range
of O-atom coverages.10 Bondzie et al.11 as well as Mullins and
co-workers12,13 studied the reactivity of preadsorbed oxygen
atoms on Au/TiO2 model catalysts. Bondzie et al. observed
prompt reaction of CO with the atomic oxygen overlayer at room
temperature. Mullins and co-workers extended this work down
to temperatures as low as 65 K. Although all of these studies
have shown conclusively that atomically adsorbed oxygen is
active for CO oxidation, characterization of atomically adsorbed
oxygen as the key reactant in the CO oxidation reaction would
require dissociation of O2 on the catalyst surface prior to
reaction. However, oxygen has not been observed to dissociate
on any gold surface. This result, as well as results from theory
discussed below, casts doubt on whether atomic oxygen is a
key player in the observed catalytic activity of supported gold
nanoclusters.
Another alternative for the reaction mechanism for the CO
oxidation reaction involves the reactivity of molecularly chemi-
* To whom correspondence should be addressed. E-mail: mullins@
che.utexas.edu.
10.1021/jp062766c CCC: $33.50 © 2006 American Chemical Society
Published on Web 09/21/2006

20338 J. Phys. Chem. B, Vol. 110, No. 41, 2006
Stiehl et al.
sorbed oxygen (O2,a). Some useful evidence motivating a
reaction mechanism involving a molecularly chemisorbed
oxygen species has come from the theoretical community. In
agreement with experiment, recent theoretical work regarding
oxygen adsorption on many gold systems has indicated that the
activation barrier for dissociation of oxygen is on the order of
∼1 eV and therefore not likely to occur under reaction
conditions commonly employed. Some interesting results have
been obtained from studying defective gold surfaces. Steps and
defects have been shown to be suitable sites for molecular
chemisorption of oxygen. Mavrakakis and co-workers have also
shown that a strained gold surface will result in conditions
suitable for molecular chemisorption of oxygen.14 Lopez and
Nørskov also report calculations that show molecular adsorption
of oxygen on a 10-atom gold cluster and reveal that it is as
reactive as atomically adsorbed oxygen for CO oxidation.15 Of
more relevance to catalysis, calculations of supported gold
clusters have also indicated that O2,a can be active in the reaction.
Liu et al. calculated that molecularly chemisorbed oxygen at
the Au-TiO2 interface will readily react with CO to form CO2.16
Molina et al. also presented similar results showing that CO
will react with molecularly chemisorbed oxygen at the Au-
TiO2 interface with an activation barrier of 0.15 eV.17 Sanchez
et al. presented DFT results in which they determined that CO
will react spontaneously with molecularly chemisorbed oxygen
on Au8 clusters supported on MgO.18 Despite all this theoretical
evidence for a molecularly chemisorbed intermediate, no
experimental evidence exists identifying a molecularly chemi-
sorbed oxygen species under relevant reaction conditions on
catalyst samples.
of an ultrahigh vacuum (UHV) surface scattering chamber with
a base pressure of ∼1 × 10-10 Torr and a quadruply differen-
tially pumped molecular beam source chamber. The scattering
chamber is equipped with an Auger electron spectrometer
(AES), low-energy electron diffraction optics (LEED), a quad-
rupole mass spectrometer (QMS), an ion gun, a quartz crystal
microbalance (QCM), and a custom-built gold deposition
system.
The TiO2(110) single crystal (10 mm × 10 mm × 0.5 mm
thick) used in this study is mounted on a tantalum plate (16.25
mm × 12.5 mm × 1 mm), and a Au(111) single crystal is
mounted on the face of the tantalum plate opposite the TiO2
crystal in a 0.5 mm deep recess 0.5 in. in diameter. Both samples
are held in place with two 0.5 mm diameter tungsten wire clips.
A 0.1 mm thick gold foil is sandwiched between the tantalum
plate and the TiO2 crystal for improved thermal contact.28 Two
tantalum wires are spot-welded to the tantalum plate to allow
for computer-controlled, resistive heating of the sample. The
tantalum heating wires are attached to two copper posts mounted
on the sample probe that are in thermal contact with a liquid
nitrogen reservoir. A type K thermocouple is spot-welded to
the top edge of the tantalum plate for measurement of temper-
ature. Gold was vapor deposited on the TiO2(110) single crystal
using a procedure described in detail by Goodman et al.29,30
A
custom-built gold doser was used to evaporate gold on the
TiO2(110) sample by resistively heating a small piece of gold
attached to a tungsten filament. The gold deposition rate was
determined using a QCM, and various gold coverages were
obtained by varying the deposition time. The typical gold
deposition rate used was ∼0.1 Å/s. In this study, 1 ML of gold
refers to a single atomic layer of close-packed gold (1.387 ×
1015 atoms/cm2), which has a thickness of approximately 2.35
Å. Lai et al. performed an extensive study of the characterization
of vapor-deposited gold on TiO2(110),30 and we expect our
samples to closely resemble theirs in regard to gold particle
size and distribution for the various gold coverages employed
in this investigation.
-
Investigations of gas-phase anionic gold clusters, Aun
(n ) 2-20), have also been very useful in understanding
the chemistry of small gold clusters.19-23 Oxygen has been
observed to interact with gold cluster anions with an odd
number of electrons, and UPS measurements20,21 on 2- and
4-atom gold cluster anions have revealed fine structure in the
spectra consistent with an oxygen-oxygen bond, indicating
molecular adsorption of oxygen on these clusters. Socaciu
and co-workers also presented evidence suggestive of reaction
The molecular beam source chamber is equipped with an
alumina, radio frequency (rf) plasma-jet nozzle31-33 (200 µm)
used for dosing oxygen on the surfaces studied. Gas mixtures
of 8% oxygen in argon are used for generating atomic oxygen
in the RF plasma-jet source. Both 16O2 (99.93%) and 18O2
(99.4%) gas mixtures are used in this study. As explained in
more detail later, employing both isotopes of oxygen allows us
to increase the ratio of molecularly chemisorbed oxygen to
atomically adsorbed oxygen of a giVen isotope. An oxygen
dissociation fraction of ∼40%, as determined via time-of-flight
techniques, is achieved.32,33 The flux of oxygen atoms in the
plasma jet was ∼2.6 × 1014 atoms/cm2‚s and determined by
the exposure time necessary to achieve a 0.23 ML oxygen atom
coverage on the Au(111) single crystal.10 Oxygen coverages
on the model catalyst samples reported in this study are defined
as the coverage of oxygen relative to saturation for the gold
coverage of interest and were obtained by integration of thermal
desorption spectra of O2. The oxygen coverage on the catalyst
samples was varied by controlling the O-atom beam exposure
time. Pure CO was dosed through the 200 mm nozzle with the
RF power off. A CO beam intensity of ∼9 × 1013 molecules/
cm2‚s was typically employed. The apertures which shape the
molecular beams result in a beam spot of ∼3 mm; this is smaller
than the dimensions of the sample, thus minimizing direct
exposure of the probe assembly other than the crystal face.
of CO with a molecularly chemisorbed oxygen species on
- 22
Au2
.
In our previous work we determined that exposing a Au/
TiO2 model catalyst to an oxygen plasma jet resulted in the
population of molecularly chemisorbed and atomically chemi-
sorbed oxygen.24 We even observed this phenomenon on a
Au(111) single crystal, although to a much smaller extent
(possibly related to a smaller number of defects in comparison
to the supported gold clusters). In this study we present results
of reactivity measurements of the molecularly chemisorbed
oxygen on Au/TiO2 model catalyst samples. We present
experimental evidence that suggests that molecularly chemi-
sorbed oxygen can react directly with CO to form CO2 on
titania-supported gold clusters in the 2-5 nm range. We studied
the reaction on five different gold coverages (0.5, 0.75, 1.0,
1.25, and 2.0 ML Au), which were given equivalent exposures
to an oxygen plasma jet, and we present evidence for reaction
of molecularly chemisorbed oxygen on all surfaces studied. A
brief account of some of this work has been published
previously.25
Experimental Section
All experiments reported in this paper were performed in a
molecular beam surface scattering apparatus that has been
described in detail elsewhere.24-27 Briefly, the apparatus consists
Following exposure of the samples to the oxygen plasma jet,
the sample is known to be populated with molecularly chemi-

Reactivity of O2,a on Au/TiO2
J. Phys. Chem. B, Vol. 110, No. 41, 2006 20339
Figure 2. CO2 production from a 2 ML Au/TiO2 sample at 77 K (a)
following exposure of the sample to a plasma jet of oxygen to give a
∼50% relative O-atom coverage, (b) following exposure of the sample
to a plasma jet of oxygen to give a ∼50% relative O-atom coverage
and Kr collision-induced desorption to remove molecularly chemisorbed
oxygen, and (c) following exposure of the sample to a plasma jet of
oxygen to give a ∼50% relative O-atom coverage, Kr collision-induced
desorption to remove molecularly chemisorbed oxygen, and anneal to
300 K.
Figure 1. Kr collision-induced desorption measurement of molecularly
chemisorbed oxygen from a 2 ML Au/TiO2 sample at 77 K. Sample
was first exposed to the oxygen plasma jet at 77 K to give a ∼60%
relative O-atom coverage followed by impingement of a ∼1 eV kinetic-
energy beam of Kr (beam starts at t ) 5 s).
sorbed oxygen.24 To ensure that the sample is solely populated
with atomic oxygen when desired, either the sample could be
heated to 300 K to thermally desorb the oxygen or the molec-
ularly chemisorbed oxygen could be removed via collision-
induced desorption.24,34-35 Collision-induced desorption of
molecularly chemisorbed oxygen was accomplished by bom-
barding the oxygen-covered surface with an energetic (∼1 eV)
beam of Kr or with a lower energy (∼0.5 eV) beam of Ar. The
high kinetic-energy beams of Kr, or Ar, are generated by
expanding a dilutely seeded Kr (2%), or Ar (2%), in helium
gas mixture through the 200 mm nozzle with the rf power off.
Figure 1 shows a typical example of a Kr collision-induced
desorption experiment. For the experiment shown in Figure 1,
a 2 ML Au/TiO2 sample was exposed to the oxygen plasma jet
at 77 K to achieve a 60% relative Oa coverage. Following
exposure to the plasma jet, the ∼1 eV Kr beam was impinged
on the sample (t ) 5 s) and the evolution of mass 32 from the
sample was monitored with the QMS. On impingement of the
Kr beam, a rapid rise is observed in the mass 32 signal, which
decays to the background level as the O2,a is removed from the
surface. Although annealing and collision-induced desorption
were both suitable for removal of molecular oxygen from the
sample, they did result in differing reactivities of the atomic
oxygen overlayer. Figure 2 shows some results that reveal that
annealing the sample can drastically affect the CO oxidation
activity of the atomic oxygen overlayer. Figure 2a shows the
CO2 produced from impingement by CO on a 2 ML Au/TiO2
sample that was exposed to the oxygen plasma jet at 77 K to
achieve a ∼50% relative O coverage (CO beam starts at t ) 5
s). The sample was neither heated nor collision-induced de-
sorption employed to clear the surface of O2,a following exposure
to the plasma jet. Figure 2b shows the CO2 produced from a
sample that was given the same oxygen exposure as the sample
shown in Figure 2a but cleared of O2,a via Kr collision-induced
desorption prior to exposure to a CO beam. There is a decrease
in the reactivity of the sample which we attribute to the absence
of reactive O2,a. We will show later that the Kr CID has an
immeasurable effect on the reactivity of the atomic oxygen
overlayer. Figure 2c shows the CO2 produced from (i) a sample
that was given the same exposure as the samples shown in
Figure 2a and 2b, (ii) collision-induced desorption performed
to clear the sample of O2,a, and (iii) a sample annealed to 300
K prior to exposure to the CO beam at 77 K. The CO2 produced
in Figure 2c is noticeably smaller than the CO2 produced in the
experiments shown in Figure 2a and 2b, indicating that annealing
drastically affects the reactivity of the atomic oxygen overlayer.
Results
In our previous study regarding identification of molecularly
chemisorbed oxygen on Au/TiO2 following exposure to an
oxygen plasma jet we noted that the amount of molecularly
chemisorbed oxygen relative to the amount of atomically
adsorbed oxygen on the sample can be small (∼10% O2,a to Oa
ratio for a 2 ML Au/TiO2 with 60% relative O coverage).24 For
the purposes of this study, it was desirable to increase this ratio
in order to enhance the probability of measuring the reactivity
of the molecular oxygen since the adsorbed atomic oxygen is
known to be quite reactive. To increase the ratio of molecularly
chemisorbed oxygen to atomically adsorbed oxygen of a given
isotope, we developed a procedure (described immediately
below) employing both isotopes of oxygen which allowed us
to enhance this ratio. This procedure was employed for all the
experiments reported in this study: (i) the sample (Ts ) 77 K)
was exposed to a plasma jet formed using 16O2 (exposure of
∼2.6 × 1014 atoms/cm2‚s for 30 s), (ii) the sample was heated
16
to 300 K to desorb O2,a, and after cooling to 77 K (iii) the
16Oa-covered sample was exposed to a plasma jet formed using
18O2 (exposure of ∼2.6 × 1014 atoms/cm2‚s for 15 s). By
preadsorbing 16Oa on the sample we could limit the available
sites for 18Oa adsorption and thus increase the ratio of 18O2,a to
18Oa on the sample. The presence of preadsorbed atoms does
not inhibit population of the molecularly chemisorbed oxygen
state. Figure 3 shows the 18O2,a/18Oa values (ratio is calculated
by dividing the area of the mass 36 signal observed during a
collision-induced desorption measurement of molecular oxygen
from the TiO2 and the gold particles by two times the mass 36
area plus the mass 34 area observed during recombinative,
thermal desorption of oxygen from the gold particles) for four
different gold coverages after preparing the samples following
the procedure described above (b). Also shown in Figure 3,
for comparison, is the ratio that is obtained without predosing
the sample with 16O atoms (2). As can be seen in Figure 3, the

20340 J. Phys. Chem. B, Vol. 110, No. 41, 2006
Stiehl et al.
Figure 4. C16O18O production at 77 K from a 1 ML Au/TiO2 sample
populated with both 18Oa and 18O2,a (dashed curve) and 18Oa only (solid
curve) upon exposure to a pure CO beam at 10 s. Inset shows average
values and uncertainties of C16O18O produced from a series of three
identical measurements such as the ones shown.
18
Figure 3. (2)
O
2,a
/18Oa as a function of Au coverage following
exposure of the catalyst sample to the oxygen plasma jet (∼2.6 × 1014
atoms/cm2‚s for 15 s). (b)
18
O
/18Oa as a function of Au coverage
2,a
following procedure described in text. (9) Fraction of TiO2 covered
by gold as a function of gold coverage (calculated using characterization
data from ref 25).
ratios for each gold coverage increase by following the
procedure described above. Finally, the fractional area of the
TiO2 that is expected to be covered by gold (9) is also shown
in Figure 3 (calculated from data of Lai et al.30). In general, the
ratio of molecular to atomic oxygen decreases because the
amount of 18O2,a that can adsorb on the TiO2 surface is greatly
reduced as the Au coverage increases and the amount of 18Oa
that can adsorb on Au surfaces is greatly increased with
increasing Au coverage, thus reducing the ratio. It was not
possible to calculate the 18O2,a to 18Oa ratio for the 0.5 ML Au/
TiO2 sample because we could not resolve the oxygen recom-
binative desorption feature from the gold particles.
Once a sample was prepared following the procedure
described above, two experiments were required to determine
18
if any of the
O
2,a
on the sample reacted. One experiment
Figure 5. C16O18O production at 77 K from a 1.25 ML Au/TiO2 sample
populated with both 18Oa and 18O2,a (dashed curve) and 18Oa only (solid
curve) upon exposure to a pure CO beam at 10 s. Inset shows average
values and uncertainties of C16O18O produced from a series of three
identical measurements such as the ones shown.
consisted of exposing the as-prepared sample to a beam of C16O
and monitoring the C16O18O produced. For comparison, another
experiment was performed by following the same sample
preparation procedure with the exception that the
18
O
2,a
was
removed via collision-induced desorption prior to exposure to
a beam of C16O. The first experiment determines the amount
of C16O18O produced from a sample that is populated with both
molecular and atomic oxygen. The second experiment provides
a measure of the amount of C16O18O produced from a sample
that has an equivalent coverage of atomic oxygen as the first
experiment but has no molecularly chemisorbed oxygen. The
greater yield of C16O18O in the first experiment in comparison
atoms reported previously.8,9 However, comparison of the
C16O18O production curves in Figure 4 reveals that the samples
populated with both of the oxygen species consistently result
in more C16O18O production than the samples solely populated
with atomic oxygen species. The average values and uncertain-
ties for the experiments are shown in the inset of Figure 4.
Figures 5 and 6 show similar results to those observed in Figure
4. For the three gold coverages shown in Figures 4, 5, and 6
(1.0, 1.25, and 2.0 ML), average differences of 41 ( 2%, 30 (
2.8%, and 18.9 ( 4.5% in the C16O18O production are observed,
respectively, between the two experiments. The values of
fractional difference presented above (and in Figure 7, discussed
below) are calculated by subtracting the area of the mass 46
signal observed from a Au/TiO2 sample populated with only
atomically adsorbed oxygen exposed to a CO beam from the
area of the mass 46 signal evolved during CO impingement on
a sample with both molecularly and atomically adsorbed oxygen
and then dividing this difference by the mass 46 area associated
with both molecular and atomic oxygen. It should be noted that
C16O2 is produced during the experiments from reaction of C16O
with 16Oa and that the C16O2 produced is the same, within
experimental uncertainty, during all comparable experiments.
to the second experiment can be attributed to the enhanced
18
reaction rate due to the presence of
O
2,a
on the sample.
Figures 4, 5, and 6 show three examples of experiments
performed following the procedure described above on 1, 1.25,
18
and 2.0 ML Au/TiO2 samples. The
O
2,a
to 18Oa ratio for the
three Au coverages of 1.0, 1.25, and 2 ML are 0.75, 0.30, and
0.25, respectively, as shown in Figure 3. Figure 4 shows the
mass 46 (C18O16O) produced during experiments representative
of both CO oxidation scenarios discussed above from a 1 ML
Au/TiO2 sample. The dashed curve in Figure 4 shows the mass
18
46 production from samples with both 18Oa and
O
2,a
species
present on the sample, and the solid curve shows the mass 46
18
production from surfaces that have been cleared of
O
2,a
via
Kr collision-induced desorption. Both reactions show behavior
typical of the CO oxidation reaction with preadsorbed oxygen

Reactivity of O2,a on Au/TiO2
J. Phys. Chem. B, Vol. 110, No. 41, 2006 20341
Figure 5 can be attributed to the direct reaction of molecularly
chemisorbed oxygen. Another scenario for the fate of the
remaining oxygen atom is that it adsorbs to the sample and
merely increases the population of 18Oa on the sample. However,
this scenario seems unlikely since, as we have shown previ-
ously,12 an increase in the oxygen-atom population in the
coverage regimes shown in Figures 4-7 actually results in a
decrease in the CO2 yield.
Another aspect of the study that needs to be addressed
regarding the reactivity of the molecular oxygen species is the
effect of collision-induced desorption on the reactivity of the
atomic overlayer. Thus far it has been assumed that the collision-
induced desorption experiments do not alter the reactivity of
the adsorbed oxygen atoms. There is some evidence that this
assumption is valid from the experiments that have been
discussed above. As mentioned earlier, production of the mass
44 (C16O2) species during the reactivity experiments is the same
within experimental uncertainty. This result indicates that the
reactivity of an annealed atomic overlayer is not affected by
the Kr collision-induced desorption.
Figure 6. C16O18O production at 77 K from a 2 ML Au/TiO2 sample
populated with both 18Oa and 18O2,a (dashed curve) and 18Oa only (solid
curve) upon exposure to a pure CO beam at 10 s. Inset shows average
values and uncertainties of C16O18O produced from a series of three
identical measurements such as the ones shown.
We attempted to address this issue further by performing
some different control experiments to show that the reactivity
of the atomically adsorbed oxygen overlayer is not affected by
collision-induced desorption. One experiment involves use of
a lower energy beam of Ar (∼0.5 eV) to perform collision-
induced measurement. The experiment shown in Figure 4 was
repeated using Ar collision-induced desorption in place of Kr
collision-induced desorption to determine if a lower energy beam
resulted in a similar reactivity difference for the two samples
(not shown). An average C16O18O production difference of 44
( 4% was observed when substituting Ar for Kr. This value is
similar to the value measured (41 ( 2%) using the more
energetic Kr beam for collision-induced desorption. This result
suggests that Kr collision-induced desorption does not measur-
ably affect the reactivity of the atomic oxygen overlayer.
To further address the effect of Kr collision-induced desorp-
tion on the reactivity of Oa, an experiment was performed on
16Oa-covered Au/TiO2 in which (i) the sample was exposed to
the plasma jet at 77 K, (ii) the sample was heated to 170 K to
desorb 16O2,a and after cooling to 77 K (iii) Kr collision-induced
desorption was performed (as expected, no desorption of 16O2,a
was observed), and (iv) the CO2 production was measured. For
comparison, a similar experiment was conducted, with the
exception that step iii was not performed. The difference in the
CO2 production between these two experiments was found to
be negligible (<3%), further verifying that Kr collision-induced
desorption does not alter the reactivity of the atomic oxygen
overlayer. We obtained similar results in studies performed on
Au(111). On Au(111) the amount of molecularly adsorbed
oxygen created by exposure to the plasma jet is very small,
such that CO2 production from a surface with both molecularly
adsorbed and atomically adsorbed oxygen is nearly identical
with that of a surface that has had the oxygen molecules
removed by CID and thus has only atomic oxygen on the
surface. These experiments show that CID has no measurable
effect on atomically adsorbed oxygen on the Au(111) surface.
Figure 7. Fractional C16O18O difference for five gold coverages studied
in this investigation (0.5, 0.75, 1.0, 1.25, and 2.0 ML).
Experiments such as the ones shown in Figures 4-6 were
performed on five different gold coverages (0.5, 0.75, 1.0, 1.25,
and 2.0 ML). Figure 7 shows the results of the fractional
C16O18O difference for all gold coverages studied. The ratios
18
of
O
2,a
to 18Oa for these experiments are shown in Figure 3.
As can be seen in Figure 7, a 1 ML Au covered sample results
in the greatest fractional difference, 41 ( 2%, for the oxygen
exposures and gold coverages studied in this investigation.
Discussion
The results shown in Figures 4-7 indicate that the samples
that are populated with both molecular and atomic oxygen result
in greater formation of C16O18O than samples which are solely
populated with atomic oxygen. The difference in the C16O18O
production can therefore be attributed to the presence of 18O2,a
on the sample. However, it is not apparent if all of the difference
that is observed is actually related solely to reaction of C16O
molecules with molecularly chemisorbed oxygen species. When
One final issue that requires mention, but is not currently
fully understood, regards the data shown in Figure 7. The
experiments that were performed to acquire the data shown in
Figure 7 were performed in an attempt to address the particle
size effect that has been observed for the catalytic CO oxidation
reaction.1,29 One difficulty with these results is that it is not
obvious how to compare the data between gold coverages. All
experiments were performed by giving the samples the same
18
the C16O reacts with O2,a, there will be an atom, 18O, left
behind which is reactive and can further react with an incoming
C16O. If the atom that is left behind promptly reacts with another
C16O molecule, then only one-half of the difference shown in

20342 J. Phys. Chem. B, Vol. 110, No. 41, 2006
Stiehl et al.
exposure to the plasma jet. Additionally, and perhaps most
importantly, the absolute oxygen content on the surface as well
as the Au particle number density on the surface is different
between all different Au coverages studied. Further, the
measurements we made are transient and not at steady state
like those used in calculating turnover frequencies. Since the
collision-induced desorption measurement accounts for O2,a on
Au and TiO2, it is not possible to scale the results to a per
oxygen molecule basis (molecules on TiO2 are known not to
react with CO at 77 K12,13). Some insight into the partitioning
of O2,a between Au and TiO2 can be obtained by determining
the area fraction of Au on the model catalyst samples. Figure 3
shows a plot of the fractional area of TiO2 that is expected to
be covered by gold (calculated from data of Lai et al.30). It is
clear that the area of TiO2 covered by gold increases as the Au
coverage is increased. It is likely that the amount of molecularly
chemisorbed oxygen associated with gold can vary with the gold
coverage. Perhaps this value increases with increasing gold
coverage up to a limiting coverage and then decreases as the
particles become more bulk like and the adsorption of O2,a
decreases. At low coverage, O2,a associated with gold might be
limited by the low density and size of the particles, while at
high coverages adsorption of O2,a on the gold particles could
be less efficient. This interpretation could reconcile the observa-
tions; however, it should be mentioned that at this time it is
purely speculative.
Figure 8. Fractional C16O2 difference as a function of relative Oa
coverage on a 0.5 ML Au/TiO2 sample.
As mentioned in the preceding paragraph, it is not obvious
how to compare the data presented in Figure 7 in order to
unambiguously determine whether there is a particle size effect
for the reaction of O2,a with CO since it is not possible to vary
one variable independently without affecting other experimental
variables (e.g., varying gold coverage results in varying particle
size and particle density as well as absolute oxygen coverage
for a given plasma-jet exposure). In an attempt to further address
the issue of particle size influence on the reaction, the oxygen
coverage dependence of the reaction was performed on three
different gold coverages (0.5, 1.0, and 1.5 ML). These experi-
ments were performed in hopes that some qualitative trend, such
as a maximum fractional difference in the reactivity with CO
at a given oxygen coverage, could be observed for the different
gold coverages and that this could be used to gain some further
insight into the particle size dependence of the reaction. In
contrast to the experiments discussed earlier, these experiments
were performed using a single isotope of oxygen, 16O2. This
approach was taken to avoid the complications of having to
arbitrarily vary two oxygen plasma-jet exposures. The downside
of using this approach is that the ratio of molecular oxygen to
atomic oxygen is going to be smaller than in the experiments
in which both isotopes of oxygen are employed (where the
purpose of using two isotopes is to artificially increase this ratio),
thus resulting in smaller fractional CO2 differences. Aside from
the use of only one oxygen isotope, the experiments were
performed in the same manner as the experiments discussed
earlier in the paper. The samples were exposed to a given
exposure of oxygen, and CO2 production was measured for a
sample that was cleared of O2,a via collision-induced desorption
and compared to the CO2 produced from a sample that was
populated with both Oa and O2,a. The results of the fractional
difference in CO2 production as a function of oxygen coverage
on the three different gold coverages are shown in Figures 8-10.
As is evident from Figures 8-10, no obvious trend was observed
and therefore no definite conclusions, either for or against a
particle size dependence, can be drawn from the experiments
that we performed. To more authoritatively study the gold
Figure 9. Fractional C16O2 difference as a function of relative Oa
coverage on a 1.0 ML Au/TiO2 sample. The data point shown with
relative Oa coverage of 0.75 corresponds somewhat to the data shown
in Figure 4.
Figure 10. Fractional C16O2 difference as a function of relative Oa
coverage on a 1.5 ML Au/TiO2 sample.
particle size dependence of the reactivity of molecularly
chemisorbed oxygen with carbon monoxide it may be necessary
to populate the surface and gold particles with controllable
coverages of adsorbed molecular oxygen without the presence
of atomic oxygen.

Reactivity of O2,a on Au/TiO2
J. Phys. Chem. B, Vol. 110, No. 41, 2006 20343
(9) Gottfried, J. M.; Christmann, K. Surf. Sci. 2004, 566, 1112.
(10) Gong, J.; Ojifinni, R. A.; Kim, T. S.; Stiehl, J. D.; McClure, S.
M.; Tanaka, P. L.; Mullins, C. B. Unpublished results.
(11) Bondzie, V. A.; Parker, S. C.; Campbell, C. T. Catal. Lett. 1999,
63, 143.
(12) Stiehl, J. D.; Kim, T. S.; Reeves, C. T.; Meyer, R. J.; Mullins, C.
B. J. Phys. Chem. B 2004, 108, 7917.
(13) Kim, T. S.; Stiehl, J. D.; Reeves, C. T.; Meyer, R. J., Mullins, C.
B. J. Am. Chem. Soc. 2003, 125, 2018.
(14) Xu, Y.; Mavrakakis, M. J. Phys. Chem. B 2003, 107, 9298.
(15) Lopez, N.; Nørskov, J. K. J. Am. Chem. Soc. 2002, 124, 11262.
(16) Liu, Z.-P.; Hu, P.; Alavi, A. J. Am. Chem. Soc. 2002, 124, 14770.
(17) Molina, L. M.; Rasmusen, M. D.; Hammer, B. J. Chem. Phys. 2004,
120, 7673.
(18) Sanchez, A.; Abbet, S.; Heiz, U.; Schneider, W.-D.; Ha¨kkinen, H.;
Barnett, R. N.; Landman, U. J. Phys. Chem. A 1999, 103, 9573.
(19) Huber, H.; McIntosh, D.; Ozin, G. A. Inorg. Chem. 1977, 16, 975.
(20) Stolcic, D.; Fischer, M.; Ganterfo¨r, G.; Kim, Y. D.; Sun, Q.; Jena,
P. J. Am. Chem. Soc. 2003, 125, 2848.
(21) Kim, Y. D.; Fischer, M.; Ganterfo¨r, G. Chem. Phys. Lett. 2003,
377, 170.
(22) Socaciu, L. D.; Hagen, J.; Bernhardt, T. M.; Wo¨ste, L.; Heiz, U.;
Ha¨kkinen, H.; Landman, U. J. Am. Chem. Soc. 2003, 125, 10437.
(23) Sun, Q.; Jena, P.; Kim, Y. D.; Fischer, M.; Gantefo¨r, G. J. Chem.
Phys. 2004, 120, 6510.
Conclusions
We presented results of an investigation of the reactivity of
molecularly chemisorbed oxygen on Au/TiO2 model catalyst
samples. We developed an experimental procedure in which
we measure the CO2 produced from a sample that is populated
with both molecularly and atomically adsorbed oxygen to a
sample that is populated with only atomically adsorbed oxygen.
We find that samples that are populated with both of the oxygen
species consistently result in a greater CO2 yield than samples
which only have atomic oxygen. We interpret this result as
evidence that the molecularly chemisorbed oxygen species can
participate directly in the CO oxidation reaction. We studied
the reaction on five different gold coverages (0.5, 0.75, 1.0,
1.25, and 2.0 ML) given equivalent exposures to an oxygen
plasma jet and observe that the greatest difference in the CO2
produced is observed for the 1 ML sample. However, we are
unable to make any unambiguous conclusions regarding a gold
particle size effect on the carbon monoxide oxidation chemistry
of adsorbed molecular oxygen due to the method by which we
populated the surface with molecularly chemisorbed oxygen.
(24) Stiehl, J. D.; Kim, T. S.; McClure, S. M.; Mullins, C. B. J. Am.
Chem. Soc. 2004, 126, 1606.
(25) Stiehl, J. D.; Kim, T. S.; McClure, S. M.; Mullins, C. B. J. Am.
Chem. Soc. 2004, 126, 13574.
(26) Kim, T. S.; Gong, J.; Ojifinni, R. A.; White, J. M.; Mullins, C. B.
J. Am. Chem. Soc. 2006, 128, 6282.
(27) Wheeler, M. C.; Seets, D. C.; Mullins, C. B. J. Chem. Phys. 1996,
105, 1572.
(28) Brinkley, D.; Dietrich, M.; Engel, T.; Farral, P.; Gantner, G.;
Schafer, A.; Szumacher, A. Surf. Sci. 1998, 395, 292.
(29) Valden, M.; Lai, X.; Goodman, D. W. Science 1998, 281, 1647.
(30) Lai, X.; St. Clair, T. P.; Valden, M.; Goodman, D. W. Prog. Surf.
Sci. 1998, 59, 25.
(31) Pollard, J. E. ReV. Sci. Instrum. 1992, 63, 1771.
(32) Wheeler, M. C.; Seets, D. C.; Mullins, C. B. J. Chem. Phys. 1997,
107, 1672.
Acknowledgment. We acknowledge the donors of the
Petroleum Research Fund, administered by the American
Chemical Society. We also thank the Welch Foundation (through
grant F-1436) and the Department of Energy (through grant DE-
FG02-04ER15587) for their generous support.
References and Notes
(1) Haruta, M. Catal. Today 1997, 36, 153; CATTECH 2002, 6, 102.
(2) Hammer, B.; Nørskov, J. K. Nature 1995, 376, 238.
(3) Meyer, R.; Lemire, C.; Shaikhutdinov, Sh. K.; Freund, H. J. Gold
Bull. 2004, 37, 72.
(4) Bond, G. C.; Thompson, D. T. Catal. ReV.-Sci. Eng. 1999, 41, 319.
(5) Bond, G. C.; Thompson, D. T. Gold Bull. 2000, 33, 41.
(6) Outka, D. A.; Madix, R. J. Surf. Sci. 1987, 179, 351.
(7) Lazaga, M. A.; Wickham, D. T.; Parker, D. H.; Kastanas, G. N.;
Koel, B. E. ACS Symp. Ser. 1993, 523, 90.
(33) Wheeler, M. C.; Reeves, C. T.; Seets, D. C.; Mullins, C. B. J. Chem.
Phys. 1997, 108, 3057.
(34) Beckerle, J. D.; Johnson, A. D.; Ceyer, S. T. Phys. ReV. Lett. 1989,
62, 685.
(8) Gottfried, J. M.; Schmidt, K. J.; Schroeder, S. L. M.; Christmann,
K. Surf. Sci. 2003, 525, 197.
(35) Beckerle, J. D.; Johnson, A. D.; Ceyer, S. T. J. Chem. Phys. 1990,
93, 4047.