Catalytic oxidation of CO by aqueous polyoxometalates on carbon-supported gold nanoparticles

Article abstract of DOI:10.1016/j.jcat.2005.06.003

Oxidation of CO was carried out at room temperature with the use of aqueous solutions of polyoxometalates over carbon-supported gold catalysts. The turnover frequency (TOF) for CO oxidation reaches high rates of 4.7 s ^-1 for catalysts with average gold particle sizes of 5 nm, estimated by X-ray diffraction (XRD) and transmission electron microscopy (TEM), and with gold loading of 1 wt%, determined by elemental analysis. Lower rates are observed at higher gold loadings, which may be attributed to mass transport limitations in the present system and to the presence of larger gold particles. Carbon-supported gold catalysts exhibit selective oxidation of CO versus H ₂, and a small increase in the selectivity for CO versus H ₂ oxidation is observed as the gold particle size decreases from 12.1 to 5.6 nm.

Full text of DOI:10.1016/j.jcat.2005.06.003

Journal of Catalysis 235 (2005) 327–332
www.elsevier.com/locate/jcat
Catalytic oxidation of CO by aqueous polyoxometalates on
carbon-supported gold nanoparticles
Won Bae Kim a,1, G.J. Rodriguez-Rivera a, S.T. Evans a, T. Voitl a, J.J. Einspahr b, P.M. Voyles b,
J.A. Dumesic a,∗
a
Department of Chemical and Biological Engineering, University of Wisconsin, 2014 Engineering Hall, 1415 Engineering Drive, Madison, WI 53706, USA
b
Department of Materials Science and Engineering, University of Wisconsin, 1509 University Avenue, Madison, WI 53706, USA
Received 18 March 2005; revised 2 June 2005; accepted 6 June 2005
Available online 28 September 2005
Abstract
Oxidation of CO was carried out at room temperature with the use of aqueous solutions of polyoxometalates over carbon-supported gold
−1
catalysts. The turnover frequency (TOF) for CO oxidation reaches high rates of 4.7 s for catalysts with average gold particle sizes of
5 nm, estimated by X-ray diffraction (XRD) and transmission electron microscopy (TEM), and with gold loading of 1 wt%, determined by
elemental analysis. Lower rates are observed at higher gold loadings, which may be attributed to mass transport limitations in the present
system and to the presence of larger gold particles. Carbon-supported gold catalysts exhibit selective oxidation of CO versus H , and a small
2
increase in the selectivity for CO versus H oxidation is observed as the gold particle size decreases from 12.1 to 5.6 nm.
2
2005 Elsevier Inc. All rights reserved.
Keywords: CO oxidation; Gold catalysts; Nanoparticles; Polyoxometalates; Hydrogen; Fuel cells
1. Introduction
We have recently reported a new process that directly uti-
lizes CO as a source of energy by converting it with water to
CO2 via an aqueous solution containing a reducible polyox-
ometalate (POM) compound, H3PMo12O40, over gold cat-
alysts [8]. As represented in the following overall reaction
based on a two-electron transfer (i.e., oxidation of one CO
molecule), the POM serves as an oxidizing agent for CO and
as an energy storage agent for electrons produced from the
reaction of dissolved CO, denoted as CO(s) and liquid water:
Following the initial discovery by Haruta et al. [1,2] that
nanoparticles of gold on oxide supports show high activity
for oxidation of CO with O2 at room temperature, extensive
research has been conducted in heterogeneous catalysis to
understand and take advantage of the unique properties of
these materials (e.g., [3–5]). The low-temperature catalytic
properties of gold catalysts are of particular relevance for the
production of fuel-cell-grade H2, since the inevitable pres-
ence of CO in H2-rich gas streams produced from catalytic
reforming of hydrocarbons [6] strongly poisons Pt-based an-
odes in H2 fuel cells [7].
3−
CO(s) + H2O(l) + 2PMo12O40
(aq)
→ CO2(g) + 2H+(aq) + 2PMo12O40
.
4−
(aq)
A possible reaction scheme for room-temperature CO oxi-
dation with liquid water assisted by the aqueous POMs was
proposed in an earlier work [9]. The proposed mechanism
assumes a heterolytic process to form protons that are sol-
vated by liquid water molecules [10], as represented below.
An electron is subsequently transferred from the gold cata-
lyst to the POM. The subscript s denotes a dissolved species
in the aqueous solution, the asterisk * denotes a surface
*
Corresponding author. Fax: +1 608 262 5434.
E-mail address: dumesic@engr.wisc.edu (J.A. Dumesic).
Current address: Department of Materials Science and Engineering,
1
Gwangju Institute of Science and Technology (GIST), Oryong-dong 1,
Buk-gu, Gwangju 500-712, South Korea.
0021-9517/$ – see front matter
doi:10.1016/j.jcat.2005.06.003
2005 Elsevier Inc. All rights reserved.

328
W.B. Kim et al. / Journal of Catalysis 235 (2005) 327–332
site of gold, and e−(POM)(aq) indicates a POM reduced by
1 electron.
times with Millipore de-ionized water until the solution pH
was around 6, followed by filtration and drying for 5 h in
air at 393 K. The surfactant N-dodecyl-N,N-dimethyl-3-
amino-1-propan sulfonate (SB, purchased from Aldrich) was
used to stabilize gold sols in aqueous solution. With this
method, different gold loadings on the carbon were synthe-
sized from 0.1 to 20 wt%. Typically, for the preparation of
a 1 wt% Au/C catalyst, 1.27 cm3 of 0.01 M aqueous SB so-
lution was added to an auric solution containing 0.1 g of 10
wt% HAuCl4·3H2O in 173 cm3 of water (i.e., molar ratio
of SB/Au = 1:2) under vigorous stirring. Aqueous Na2CO3
was added with stirring until the solution pH reached around
8. A freshly prepared aqueous NaBH4 solution (5 cm3 of
0.05 M) was added dropwise over 4 h with a syringe pump
(Harvard Apparatus, Inc.), corresponding to a 10:1 molar ra-
tio of NaBH4:Au. A few minutes after formation of a gold
sol, as indicated by a color change of the solution to a ruby
red color, 1 g of the pretreated carbon was added to the
gold sol with vigorous stirring. The resulting slurry was fil-
tered for 2 h after all of the NaBH4 was added. The catalyst
was then washed thoroughly with water and dried for ap-
proximately 40 h in a vacuum chamber to remove absorbed
water. We used this catalyst without heat treatment for reac-
tion kinetic studies. The water content of each catalyst was
measured by monitoring of the weight of a catalyst aliquot
during heat treatment (393 K for 4 h), showing that the Au/C
catalysts contained 5–10% water by weight.
CO(s) + * → CO*,
H2O(l) + * + (POM)(aq) → OH* + H+
+ e−(POM)(aq)
,
(aq)
CO* + OH* → COOH* + *,
COOH* + (POM)(aq)
→ CO2(g) + H+
+ e−(POM)(aq) + *,
(aq)
Net:
CO(s) + H2O(l) + 2(POM)(aq)
→ CO2(g) + 2H+
+ 2e−(POM)(aq)
.
(aq)
To utilize these protons and electrons produced from wa-
ter, we demonstrated that the resulting aqueous solutions
containing electrons associated with the reduced POM and
protons produced from water during CO oxidation to CO2
could be delivered to a carbon anode of a proton exchange
membrane (PEM) fuel cell, leading to the production of elec-
trical energy upon re-oxidation of the reduced POM to its
initial oxidized state [8]. It was also demonstrated [9] that
this room-temperature process could be used to selectively
remove CO in H2-rich gas streams over carbon-supported
gold catalysts (e.g., selective oxidation of CO in a gas stream
containing 0.1% CO in H2).
In this note, we report results from an investigation of the
effects of gold particle size for the oxidation of CO by liquid
water with POM (i.e., aqueous POM solutions) over carbon-
supported gold catalysts. A series of carbon-supported gold
catalysts (Au/C) were prepared with a systematic variation
of the gold loadings from 0.1 to 20 wt% and with a change of
gold particle sizes from 5.6 to 12.1 nm at a fixed loading of
gold (1 wt% Au/C), as estimated by X-ray diffraction (XRD)
and transmission electron microscopy (TEM). We have used
carbon as a support for gold nanoparticles to avoid possible
complexities caused by participation of oxide supports in the
reaction scheme, such as the involvement of OH [3] or ac-
tivated O species [11] from the support as oxidizing agents
for CO. In fact, whereas most studies of CO oxidation by
O2 involve the use of gold nanoparticles on various oxide
supports, we find that CO oxidation by aqueous POM com-
pounds takes place at high rates on carbon-supported gold
nanoparticles, perhaps because the catalyst does not need to
dissociate O2 under our reaction conditions.
As a reference for comparison with the properties of
the Au/C catalysts prepared in this study, we also studied
the catalytic properties of the “Type D” Au/C catalyst sup-
plied by the World Gold Council (WGC, Lot #4D, Lon-
don SW1Y 5JG, United Kingdom), containing 0.8 wt%
Au as determined by atomic absorption/inductively coupled
plasma (ICP) emission. The average Au particle diameter
measured by TEM was 10.5 nm, and the Au particle di-
ameter estimated by powder X-ray diffraction (XRD) was
8.0 nm.
2.2. Reaction kinetics measurements
In a typical experiment, a batch reactor was loaded with
20 cm3 of 0.05 M POM solution, and the Au/C catalyst was
purged three times with N2 or He and then filled to a pres-
sure of 14.8 bar with pure CO or a CO:H2 mixture (0.1–1%
CO in H2), supplied by Linde Gas Group (Independence,
OH, USA). The reactor was a stainless-steel vessel (approx-
imately 350 cm3 volume), in which the POM solution and
catalyst were placed in a glass liner with a magnetic stir-
rer. We analyzed gas-phase products at specific times by
releasing pressure from the batch reactor to an online gas
chromatograph (GC, Hewlett Packard 5890) equipped with a
thermal conductivity detector and a 30-foot Alltech column
packed with 120/100-mesh Hayesep DB that used helium
as a carrier. The column was initially kept at 313 K for
10 min, and the temperature was then ramped at 20 K min−1
to 513 K, where it was kept for 10 min. The temperature
2. Experimental
2.1. Preparation of Au/C catalysts
Carbon-supported gold catalysts were made from sur-
factant-protected gold particles in aqueous solution [12].
The primary carbon support used in this study was Vul-
can XC72 (Cabot Corp.), which was first acid-pretreated
by suspension in a 6 M aqueous HCl solution with vigor-
ous stirring for 12 h. The carbon was then washed several

W.B. Kim et al. / Journal of Catalysis 235 (2005) 327–332
329
program allowed the analysis of noncondensable gases and
Table 1
Physical properties of Au/C catalysts with different Au loadings
water vapor. We have used several gas mixtures for reaction
kinetics studies and for GC calibrations. These mixtures in-
cluded 1 and 0.1% CO in H2 balance, UHP (99.999%) H2,
UHP (99.995%) CO, and 1.01% CO2 in N2 balance (formu-
lated by Linde Gas Group, Independence, OH, USA). We
measured the rate of H2 oxidation by monitoring the color
change in the POM solution with a UV/vis spectrometer op-
erating at 500 nm [9].
The 1:1 stoichiometry between CO consumed and CO2
produced was corroborated by analysis of the gas products
with the GC for repeated experiments, and the ratio of the
amount of CO2 produced to the amount of CO consumed
was 1.0 0.1. The electron balance determined from the
extent of POM reduction and the amount of CO2 produced
agreed within 20% for the following overall reaction:
Au loading (wt%)
from ICP
Average particle size of Au (nm)
b
a
c
From XRD
From TEM
11.9 2.0
0.10
0.49
0.81
1.0
2.0
3.9
7.2
5.2
5.6
5.0
5.0
4.7
8.7
10.4
4.5
8.0
7.9 2.2
6.8 1.7
9.2
20.1
10.0
0.8
d
7.5 2.4
10.5 3.0
e
a
Inductively coupled plasma (ICP) analysis ( 5% in the determined
wt%).
b
Mean diameter of gold estimated by line broadening of powder XRD
◦
peak at 2θ = 38.2 using the Scherrer equation ( 1 nm).
c
CO + H2O + 2(POM)(aq)
Mean diameter of gold from TEM images using at least 100 particles.
10 wt% Au/C catalyst prepared using a diluted gold solution compared
→ CO2 + 2H+ + 2e−(POM)(aq)
.
d
to the normal 10 wt% catalyst with an actual loading of 9.2 wt% by ICP.
e
In this calculation, the total number of electrons produced
and subsequently stored in the POM solution was calculated
by integration of the electrical current versus time and di-
vision by the Faraday constant while the reduced POM was
completely re-oxidized in an anode of an electrochemical
cell [8] for the following reaction:
Au/C catalyst supplied from the World Gold Council.
TEM in our measurement conditions. We varied the gold
loading on these catalysts from 0.1 to 20 wt% by changing
the concentration of the gold precursor during the synthesis
procedure. The rates of CO oxidation versus gold loading of
these catalysts are plotted in Fig. 1, wherein the rates are de-
fined as (1) the specific rate of CO oxidation based on the
2H+ + 2e−(POM)(aq) + (1/2)O2 → H2O + 2(POM)(aq)
.
catalyst weight (µmolCO g−ca1t min−1) and (2) the turnover
2.3. Characterization of catalysts
2
frequency (TOF, in s−1), in which the rate is normalized by
the number of surface sites estimated from the average par-
ticle diameter determined by XRD and TEM. We have cal-
culated the dispersion of the gold particles, using the atomic
density of (111) planes, via the following equation:
X-ray diffraction patterns were collected for various
Au/C catalysts to estimate the gold particle sizes, with the
use of a model XRG3000 diffractometer operating with a
Cu-Kα (λ = 1.5405 Å) source and a CPS120 curved detec-
1
tor with 3915 active channels (INEL Inc.). The source was
operated at 40 kV and 30 mA.
D (dispersion) = 1.41/dp,
where dp is the particle diameter in nanometers. All rates
were measured at initial conversions of POM of less than
10%, corresponding to less than one electron per POM unit.
(The highest reduction of POM under our reaction condi-
tions corresponds to five electrons transferred to each POM
unit [8].) The POM used in this work has the Keggin struc-
ture [13] with multiple redox steps via single and multi-
electron transfer processes [14].
The specific rate increases with gold loading from 0.1
to 4 wt%, and it decreases monotonically with further in-
creases in the gold loading above 4 wt%. This behavior could
be related to an increase in the gold particle size at higher
loadings of gold on carbon; that is, the gold particles are
substantially larger and have agglomerated at gold loadings
from 10 to 20 wt%. The results from XRD show that the gold
particles have diameters from 9 to 11 nm for gold loadings
from 10 to 20 wt%, whereas the gold particles are smaller
(5–6 nm) for gold loadings from 0.5 to 4 wt%. This trend is
also observed in the average particle diameters determined
from TEM images, as shown in the representative micro-
graphs in Fig. 2. The specific rate of the reference catalyst
The morphology and size distribution of gold parti-
cles were studied with a Philips CM200 Ultra Twin high-
resolution TEM operating at 200 kV with a 1.9-Å point
resolution. All TEM images on the Au/C samples were ob-
tained in the absence of an objective aperture. We prepared
the samples for TEM imaging by depositing a suspension
containing the Au/C particles on carbon-coated TEM grids
and allowing the water to evaporate, leaving the particles
suspended on the carbon support.
Gold loadings on all catalysts were determined by in-
ductively coupled plasma (ICP) analysis, showing that the
nominal gold loadings (in wt% on carbon) were the same as
their actual loadings within 5%, as presented in Table 1.
3. Results and discussion
Physical properties of the gold catalysts prepared in this
study are presented in Table 1. The average diameters of gold
particles estimated by XRD were generally smaller by about
2.5 nm for each sample compared with those measured by

330
W.B. Kim et al. / Journal of Catalysis 235 (2005) 327–332
Fig. 1. Rates of CO oxidation over Au/C catalysts with different gold load-
ings. Reaction rates were measured at initial conversions of POM less than
10%. Values of TOF were estimated from dispersion of gold measured by
XRD and TEM. Legend: Rates per gram of catalyst (") and TOF (2); rates
per gram of catalyst (Q) of poorly dispersed 10 and 20 wt% Au/C catalysts;
rates per gram of catalyst (!) and TOF (1) at 0.8 wt% for WGC; rates
per gram of catalyst (!) and TOF (1) at 10 wt% for Au/C prepared from
diluted gold sols.
from the WGC is also shown as an open symbol at 0.8 wt%
in Fig. 1. In addition, we have also measured the rate over
a 10 wt% Au/C catalyst that was prepared with the use of a
diluted gold sol (with the same gold concentration in aque-
ous solution as the 1 wt% Au/C catalyst), and this result is
also presented as an open symbol in Fig. 1. This 10 wt% cat-
alyst had smaller gold particles compared with the 10 wt%
catalyst prepared by the standard method (compare Figs. 2e
and 2f). The 10 wt% Au/C made with the diluted gold so-
lution (open symbols at 10 wt% Au) shows higher activity
than the normal 10 wt% Au catalyst, because of the smaller
gold particles in this former sample. The gold particles for
this catalyst are similar in size to the catalysts with lower
loadings, and the plot of rate per gram of catalyst in Fig. 1
does not show a decrease at higher loadings when this 10%
Au/C catalyst is considered.
When the specific rates are normalized by the number of
surface sites, as estimated from the TEM images, the TOFs
for gold loadings from 0.1 to 1 wt%, including the WGC
catalyst, are similar within experimental error, as seen in
Fig. 1. In contrast, the TOF decreases at gold loadings of
2 wt% or higher, as seen in Fig. 1. XRD and TEM measure-
ments indicate that the average particle sizes of the 2 and
4 wt% catalysts are similar to those with lower gold load-
ings. Therefore, the decrease in TOF at these catalysts is
probably caused by mass transport limitations. This behavior
is also seen for the 10 wt% Au/C sample prepared with the
diluted sol, because this catalyst has a low TOF, despite hav-
ing gold particles as small as those present on samples with
gold loadings near 1 wt% (see the TEM images of Fig. 2f).
Also, as noted above, the rate of CO oxidation per gram of
Fig. 2. TEM images of Au/C catalysts with different loadings. (a) 0.1 wt%,
(b) 0.8 wt%, (c) 4 wt%, (d) high resolution image showing lattice fringes of
a gold nanoparticle from the 4 wt% Au/C catalyst. The scale bar is 40 nm
for (a)–(c) and 2 nm for (d). (e) 10 wt%, (f) 10 wt% with a 10 times more
diluted gold sol, (g) the fresh Au/C catalyst from WGC, (h) the spent Au/C
catalyst from WGC. The scale bar is 30 nm for (e) and 40 nm for (f)–(h).
catalyst becomes nearly constant at the higher gold loadings
when the 10 wt% Au/C sample prepared with the diluted
sol is considered. The normal Au/C samples with the higher
gold loadings of 10 and 20 wt% show very low rates, which
can be ascribed to the very large sizes of the primary gold
particles, as determined by XRD. Furthermore, the low rates

W.B. Kim et al. / Journal of Catalysis 235 (2005) 327–332
331
Fig. 3. Gold particle size distribution measured by TEM for the fresh WGC
Au/C catalyst (black bar) and the spent WGC Au/C catalyst (gray bar) af-
ter oxidation of CO with the aqueous POM solution. Legend: fresh Au/C
from the World Gold Council (black bar); spent Au/C from the World Gold
Council (gray bar).
Fig. 4. Rates of CO oxidation over 1 wt% Au/C catalysts with different gold
particle sizes estimated by XRD. (a) TOF for CO oxidation with 14.8 bar
of 99.995% CO gas ("), with 14.8 bar of gas mixtures of 1% CO in 99%
H
(Q), and 0.1% CO in 99.9% H (2), (b) the rate ratios of CO oxidation
2
2
to H oxidation with the same symbol denotation as (a).
2
for these two samples may also be related to the additional
mass transfer limitations caused by agglomeration of these
large particles, as shown in the TEM images.
ately, whereas more dramatic change in the surface coordi-
nation would take place for particles less than 4 nm in size,
and no significant changes would be expected for the parti-
cle diameter exceeding 10 nm. As seen in Fig. 4a, a modest
improvement in the TOF for oxidation of pure CO appears to
be caused by a decrease in the gold particle size from 10.1 to
5.6 nm, suggesting that surface sites with low coordination
numbers may be more active. For the preferential oxidation
of CO, it seems that the TOF for the oxidation of CO in the
presence of H2 also increases modestly with decreasing gold
particle size under our reaction conditions, that is, for rates
with CO concentrations of 0.1 and 1% in H2. It is interesting
to note that the ratio of the rate of CO oxidation to the rate of
H2 oxidation also appears to be a function of the gold parti-
cle size (see Fig. 4b), such that smaller gold particles appear
to be more selective for the preferential oxidation of CO in
the presence of H2. This effect leads to an enhancement of
the rate ratio by a factor of 2–3 when the gold particle size
decreases from 12.1 to 5.6 nm. We note that whereas Au/C
catalysts show high selectivity for CO versus H2 oxidation,
other carbon-supported metals (e.g., Pt, Pd, and Ir) exhibit
high rates of H2 oxidation [9]. This high selectivity of gold
for CO versus H2 oxidation may be related to fact that the en-
ergetics for the dissociative adsorption of H2 on gold are far
less favorable compared with those of other transition metals
[15,16].
Fig. 3 shows the particle size distribution of gold mea-
sured by TEM for the fresh WGC catalyst and the spent
catalyst after reaction kinetics measurements of CO oxida-
tion in aqueous POM solutions. The particle size distribution
appears to be slightly narrower after reaction, showing more
particles with diameters of 7–10 nm for the spent catalyst
compared with the fresh catalyst, which shows more parti-
cles with diameters of 3–6 nm together with a few very large
particles (40–80 nm) that are not included in the figure. We
have observed that the specific rate of CO oxidation on the
WGC catalyst decreases initially by about 20% in aqueous
POM solution, followed by much slower deactivation there-
after, such that about 50% of the initial catalytic activity is
retained after exposure to 0.05 M POM for 3 months [9].
This initial decrease in the rate of CO oxidation is consis-
tent with the observed growth noted in Fig. 3 of the smaller
particles to sizes of 7–10 nm.
Fig. 4 shows rates of CO oxidation with pure CO and with
0.1 or 1% CO in H2 gas mixture (i.e., preferential oxidation
of CO in H2) as a function of the gold particle size estimated
by XRD. A Au/C (on Black Pearl 2000 carbon support) cat-
alyst with a gold loading of 1 wt% was heat-treated in air at
temperatures from 573 to 673 K for 2 h to increase the av-
erage particle size of gold from 5.6 to 6.4 nm and from 5.6
to 12.1 nm, respectively. Note that in this range of particle
sizes, the surface morphology with respect to surface sites
with different coordination numbers should change moder-
We have shown in this research note that high rates of
CO oxidation by liquid water with polyoxometalates can
be achieved over carbon-supported gold catalysts containing
gold particles with sizes from 5 to 8 nm at gold loadings up

332
W.B. Kim et al. / Journal of Catalysis 235 (2005) 327–332
to about 1 wt%. The rate of CO oxidation becomes subject to
mass transport limitations at gold loadings from 2 to 4 wt%
for the present reaction conditions. The TOFs for CO oxi-
dation by liquid water with polyoxometalates reach values
of about 5 s−1 for gold loadings from 0.1 to 1 wt% in pure
CO gas. In addition, good performance can be achieved for
the selective oxidation of CO in H2-rich gas mixtures, sug-
gesting that this catalytic system shows promise for effective
removal of CO from H2-rich gas streams at room tempera-
ture. The TOF for CO oxidation increases modestly as the
size of the gold particles decreases from 12.1 to 5.6 nm, and
the smaller particles also show better selectivity for the oxi-
dation of CO in the presence of H2.
References
[1] M. Haruta, T. Kobayashi, H. Sano, N. Yamada, Chem. Lett. 2 (1987)
405.
[2] M. Haruta, N. Yamada, T. Kobayashi, S. Iijima, J. Catal. 115 (1989)
301.
[3] H.H. Kung, M.C. Kung, C.K. Costello, J. Catal. 216 (2003) 425, and
references therein.
[4] Q. Fu, H. Saltsburg, M. Flytzani-Stephanopoulos, Science 301 (2003)
935.
[5] G.C. Bond, D.T. Thompson, Gold Bull. 33 (2000) 41.
[6] J.R. Rostrup-Nielsen, T. Rostrup-Nielsen, CATTECH 6 (2002) 150.
[7] B.C.H. Steele, A. Heinzel, Nature 414 (2001) 345.
[8] W.B. Kim, T. Voitl, G.J. Rodriguez-Rivera, J.A. Dumesic, Science 305
(2004) 1280.
[9] W.B. Kim, T. Voitl, G.J. Rodriguez-Rivera, S.T. Evans, J.A. Dumesic,
Angew. Chem. Int. Ed. 44 (2005) 778.
Acknowledgments
[10] S.K. Desai, M. Neurock, Phys. Rev. B 68 (2003) 075420.
[11] M. Daté, M. Okumura, S. Tsubota, M. Haruta, Angew. Chem. Int.
Ed. 43 (2004) 2129.
This work was supported by the U.S. Department of
Energy, the National Science Foundation, and Daimler-
Chrysler. Fellowships were provided by the Camille and
Henry Dreyfus Foundation and the Graduate Engineering
Scholars program. We thank the World Gold Council for
providing gold catalysts and the Materials Science & Engi-
neering Department of University of Wisconsin for provid-
ing facilities.
[12] S. Biella, F. Porta, L. Prati, M. Rossi, Catal. Lett. 90 (2003) 23.
[13] G.M. Brown, M.-R. Noe-Spirlet, W.R. Busing, H.A. Levy, Acta Crys-
tallogr. B 33 (1977) 1038.
[14] P. Gómez-Romero, N. Casañ-Pastor, J. Phys. Chem. 100 (1996)
12448.
[15] S. Kandoi, A.A. Gokhale, L.C. Grabow, J.A. Dumesic, M. Mavrikakis,
Catal. Lett. 93 (2004) 93.
[16] T. Bligaard, J.K. Nørskov, S. Dahl, J. Matthiesen, C.H. Christensen,
J. Sehested, J. Catal. 224 (2004) 206.