www.sciencemag.org/cgi/content/full/333/6043/736/DC1

Supporting Online Material for Spectroscopic Observation of Dual Catalytic Sites During Oxidation of

CO on a Au/TiO2 Catalyst Isabel Xiaoye Green, Wenjie Tang, Matthew Neurock, John T. Yates, Jr.*

*To whom correspondence should be addressed. E-mail: johnt@virginia.edu

Published 5 August 2011, Science 333, 736 (2011)

DOI: 10.1126/science.1207272

This PDF file includes:

Materials and Methods SOM Text Figs. S1 to S12 Table S1 References

Supporting Online Material

for Spectroscopic Identification of Dual Catalytic Sites during

CO Oxidation on a Au/TiO2 Catalyst

Isabel Xiaoye Green, Wenjie Tang, Matthew Neurock, and John T. Yates, Jr.

Table of Contents: I. Supporting Text -- Background

II. Experimental Materials and Methods.

III. Model System and Calculation Details.

IV. Supporting Text -- Additional Calculation Results.

V. References for the Supporting Online Material

VI. Supporting Figures.

VII. Supporting Tables.

I. Background

Determining the atomic and electronic structure of the active site in heterogeneous

catalysis is a unifying theme for almost all work in the field (31-35). Understanding the

cooperative behavior between dual catalytic centers comprised of distinguishable active

sites is particularly challenging due to the complexity in resolving the mechanistic role of

each site and their interactions under catalytic reaction conditions. For the Au/TiO2

catalyst, some have proposed that the novel catalytic activity originates predominantly

from nanometer size Au particles, and that the role of the oxide support is to stabilize

these particles. Highly dispersed Au clusters thus provide a large fraction of active

coordinatively unsaturated Au surface sites and can also possibly exhibit quantum size

effects (21, 22, 26, 36). Others have proposed that the electronic interaction between the

Au nanoparticles and the underlying support is the key feature since cationic Au is

thought to be more active for the CO oxidation reaction (37, 38). Two atom thick Au

films, supported on macroscopic-width thin TiOx films, where no perimeter interface

between materials is present, have been reported to show high relative activity supporting

the idea of a favorable electronic interaction between Au and TiO2 (4). Considering the

known weak interaction between Au and O2, it was also proposed that the support may

enhance the catalytic reaction by localizing O2 or O atoms at Au perimeter sites (6, 9,

39). Several experiments carried out with inverse catalysts demonstrated that the oxides

might be more important than just supporting the Au or enhancing the reaction on low

coordination number (CN) Au sites, by being involved directly in the catalytic reaction:

inactive Au single crystals or nano-sized Au powders become active when they come in

contact with particular oxide supports such as TiOx or CeOx (5, 8, 40). The active sites

for the catalytic reaction are therefore postulated in this view to be at the Au/oxide

interface. A number of theoretical studies have also been reported which help to elucidate

the CO oxidation pathways and mechanisms over the supported Au catalyst (24, 41), and

specifically examine the perimeter sites (10-14, 42).

II. Experimental Materials and Methods.

In our experimental setup, a stainless steel high vacuum transmission infrared cell

with a base pressure of 1 × 10-8 Torr after bakeout is employed. A schematic drawing of

the cell and the design of the sample holder placed inside the cell is shown in Figure S1.

The high vacuum in the cell is maintained by a 55 L/s (N2) turbomolecular pump, which

can be isolated from the cell by a gate valve. A MKS I-MAG cold cathode ionization

gauge for low pressure reading (10-10-10-3 Torr) and a MKS Baratron capacitance

manometer for high pressure reading (0.001-1000 Torr) are connected to the cell. The

cell has two differentially pumped KBr windows that allow the transmission of a

focussed IR beam through the center of the cell, as shown in Figure S1A. A quadrupole

mass spectrometer (QMS) is connected to the cell via a gate valve for gas analysis. The

Au/TiO2 catalyst and a reference sample of pure powdered TiO2 are pressed separately

into 7 mm spots, one directly above another, on a tungsten mesh support, as shown in

Figure S1B. Each sample spot is pressed with 33,500 psi pressure and holds

approximately 0.005 g of material. The W-mesh is 0.051 cm thick with 1370 window

apertures (0.022 cm × 0.022 cm) per cm2 that allows ~70% IR transmission. The W-mesh

plus pressed samples is then mounted by a pair of nickel clamps that are connected to the

copper rods from a reentrant Dewar for fast electrical heating and cooling. Since the W-

mesh mounted this way generates no vertical temperature gradient upon heating and

cooling (43), a type K thermocouple providing 0.1 K temperature resolution is welded on

the top center of the W-grid, at the same vertical position as the samples’ centers. The

sample temperature can be controlled accurately to ± 0.1 K between 90 and 1000 K, via a

combination of resistive heating and l-N2 cooling. The sample holder is arranged so that

the samples are placed in the center of the chamber to fully intercept the IR beam. Both

samples can be examined by IR in one experiment by moving the z-direction manipulator

up and down.

The protocol for Au deposition-precipitation with urea on TiO2 is adopted from

Zanella et al. (44). Commercial TiO2 P25 powder (49 m2 g-1) is provided by Evonic

Industries (previously Degussa). Urea (99.5%) and hydrogen tetrachloroaurate (III)

trihydrate (HAuCl4⋅3H2O, Au precursor) were purchased from Acros. Before deposition,

the TiO2 powder was dried and activated at 373 K in a 100 mL/min air flow for at least

24 hours. The rest of the preparation process was carried out in the absence of light, due

to the known fact that the gold precursor may decompose under illumination. A gold

precursor aqueous solution was made (4.2 × 10-3 M HAuCl4 and 4.2 M urea). One gram

of activated TiO2 powder was mixed with 100 mL of the gold precursor solution and

reacted at 353 K with vigorous stirring for 8 hours. The pH of the reaction mixture

increased from ~2 at the beginning of reaction to ~7 at the end. The Au/TiO2 powder was

separated from the suspension by centrifuging at 3800 rpm for 20 min. To wash off the

residual Cl- anion, which is suspected to compete with O2 for active sites during CO

oxidation (10), the Au/TiO2 powder was washed by 5 cycles of vigorous stirring in 100

mL double-deionized water at 323 K, followed by centrifugation at 3800 rpm for 20 min.

Auger spectra, measured in a separate UHV system, show no Cl- contamination on the

samples after washing. The powder was then dried at 373 K for over 24 hours under 100

mL/min air flow. The theoretical Au loading on the catalyst is 8 wt%.

After the samples were installed in the vacuum cell and before initial use, they were

heated to 473 K with a temperature ramping rate of 0.2 K/s and were held at 473 K in

vacuum for 2.5 hours. Then 20 Torr of O2 was introduced into the cell at 473 K,

remaining for 1 hour. After pumping the cell down to high vacuum, the samples were

held at 473 K for another 0.5 hour. The goal of the heating in vacuum and in O2 is to

remove most of the hydrocarbon impurities from the sample surfaces. Complete removal

of all impurities including hydroxyl groups requires higher treatment temperature which

is known to cause Au particle agglomeration and thus was not used. Between each

separate experiment, heating at 473 K in vacuum for 20 min was performed to bring the

samples back to the original condition.

Transmission electron microscope (TEM) measurements were carried out on a

Au/TiO2 sample that has been treated as described above. Data were collected on a JEOL

2000FXII microscope with 200 kV acceleration voltage. The average Au particle size

was ~3 nm. A TEM image is shown in Figure S2A, along with a histogram of measured

Au particle sizes.

The IR spectra were recorded with a Bruker TENSOR 27 FTIR spectrometer

equipped with a l-N2 cooled MCT detector. The spectrometer, the detector, and the entire

IR beam pathway are purged with CO2- and H2O- free air constantly. In order to

eliminate the influence of the W-grid and the gas phase composition, a reference

spectrum taken through the empty region of the W-grid and the gas phase (vacuum or

reaction mixture) is subtracted from every spectrum recorded through the samples under

the same condition. Each spectrum is obtained by averaging 512 interferograms at 2 cm-1

resolution which takes about 203 s.

Figure S3 shows the IR spectra of CO interacting with the reference TiO2 sample,

both in vacuum and in oxygen. Figure S4 shows kinetic plots of the rate of formation of

CO2(a)/TiO2 from 110 K to 130 K, fitted to first-order kinetics, to complement Figure 1C

which shows the kinetics of CO(a)/TiO2 consumption. An Arrhenius plot in the inset

shows that the kinetics exhibit an apparent activation energy of 0.16 ± 0.01 eV, in

agreement with the measurements in Figure 1C.

Figure S5 shows CO diffusion studies on both TiO2 and Au sites at temperatures from

140 to 160 K. By holding the CO-covered Au/TiO2 sample at elevated temperatures (140

to 160 K) in vacuum, we observed the surface diffusion/desorption of CO by IR. Here

CO molecules present in the interior of the powdered samples as CO/TiO2 and CO/Au

surface species escaped by a combination of surface diffusion and desorption processes,

as observed by this IR method which has also been used for surface diffusion

investigations of other molecules adsorbed on similar high surface area powdered

samples (25, 45). The catalyst surface was saturated with CO at 120 K and then

evacuated for 1 hour before beginning the kinetics study at 140 K or above. For ease of

comparison, the diffusion/desorption studies were all normalized to the saturation point at

120 K. The CO diffusion/desorption on TiO2 was more rapid at all temperatures than on

Au. At 160 K, after 3 hours, 70% of the CO on the TiO2 surface desorbed while only

20% of the CO has desorbed from the Au surface.

III. Model System and Calculation Details.

To simulate the ~3 nm diameter Au particles supported by TiO2 system, we used a Au

nanorod structure which has a height of 3-atomic layers and is 3-atoms-wide covalently

bonded on top of a rutile TiO2(110) slab (10, 11), as shown in Figures S6A and S6B. This

structure is similar to that used in previously reported DFT calculations (3, 4). The Au

nanorod structure was chosen herein to model the sites on the Au particles supported on

TiO2 due to its computational tractability. It appropriately mimics the terrace and edge

sites as well as the Ti sites located at the Au/TiO2 interface as well as within the local

vicinity. The model, however, is an oversimplification of the nm-sized Au clusters on

TiO2 as there are no corner sites. These sites are likely to be somewhat more reactive for

CO oxidation due to their ability adsorb and activate O2. As such the Au-nanorod

structure will somewhat underestimate the actual reactivity.

Two unit cells comprised of four O-Ti-O tri-layers were used to model the rutile

TiO2(110) surface. All of the simulations reported where carried out using a (2×3) unit

cell to simulate CO adsorption and diffusion and the larger (4×3) unit cell to model CO

oxidation. The atoms in the top two trilayers of the TiO2 slab were allowed to fully relax

while the atoms in the bottom two trilayers of the TiO2 slab were fixed to their lattice

positions for both unit cells. All of the Au atoms were allowed to relax in the Z-

direction.

All of the calculations reported herein were carried out using plane wave gradient-

corrected density functional theory (DFT) calculations as implemented in the Vienna ab

initio Simulation Package (VASP) (46, 47). The wavefunction was expanded in a plane-

wave basis set with energy cutoff of 400 eV. The PW91 gradient approximation (GGA)

functional was used to describe the exchange-correlation energy (48). The core-electronic

states were described by pseudopotentials constructed with the projector augmented-

wave (PAW) method (49, 50). The DFT+U method was implemented in our calculations

to correct on-site Coulomb interactions (51). The value of U was chosen to be 4.0 eV

because it can generate similar electronic structure that was experimentally observed

(52). Spin-polarization was considered for all calculations and was used when necessary.

A vacuum gap of 10 Å was used in the z- direction between slabs. The (2×3) and (4×3)

unit cells were sampled with (2×2×1) and (1×2×1) k-point mesh, respectively (53).

Geometries were optimized until the force on each atom was less than 0.03 eV/Å.

The reaction pathway and activation barriers were found by the nudged elastic band

(NEB) method with image climbing (54, 55), combined with the dimer method to isolate

transition states (56). The NEB method was used to follow the path between the reactant

and product states to the point where the perpendicular forces on all of the images along

the band were lower than 0.1 eV/Å. The dimer method was subsequently used to isolate

the transition state to the point where the force acting on the transition state dimer was

lower than 0.03 eV/ Å.

The charge density difference, ρdiff , on the TiO2 surface before and after Au

deposition was calculated following Equation 1.

ρdiff = ρ(Au /TiO2) − ρ(TiO2 ) − ρ(Au) Eq. 1

The electronic charge difference at the perimeter site is plotted in Figure S6C, showing

that negative charge accumulates on the Ti5c site at the perimeter after Au deposition.

This change in charge is responsible for the strong adsorption and activation of O2 at

these sites (7, 13, 17, 57-59).

IV. Additional Calculation Results.

A. CO Binding on Au/TiO2.

The calculated adsorption energies for CO at different sites at the Au/TiO2 interface

are shown in Figure S7. The adsorption energy of CO at the Ti5c site of TiO2 adjacent to

adsorbed Au was calculated to be -0.41 eV as shown in Figure S7A. For the adsorption of

CO on Au, the CO adsorption energy varies with the CN of the Au atoms to which the

CO is bound. The Au atoms examined are shown from the cross section view of the

nano-rod and labeled sites 1-6 on Figure S7A for easy identification. In general, and as

expected, CO adsorbs more strongly on low CN Au sites compared to high CN Au sites

with the exception of sites 1 and 6. These two sites are attached to the TiO2 support

directly, and exhibit minimal interaction with CO molecules due to the electronic effects

and the local structure. The calculations show weak CO binding with an adsorption

energy of -0.26 eV at site 1 and no binding at site 6. The adsorption energy for CO at the

terrace sites (CN=9) was calculated to be -0.2 eV, which also indicates a very weak

interaction. This is consistent with the experimental observation that CO does not

chemisorb on Au terrace sites even at low temperatures as 25 K (7, 17, 57-59). For Au

atoms with CN = 7, the CO binding energies ranges from -0.67 eV to -0.89 eV, as shown

in Figures S7D-F. These sites are likely to be the CO adsorption sites on Au.

B. O2 Activation at Different Sites.

Density functional theory calculations carried out on the adsorption of O2 on the ideal

rutile TiO2(110) surface revealed that O2 does not adsorb on these surfaces which is

consistent with experimental results (60). This lack of adsorption indicates that it would

be very difficult to dissociate O2 directly over the ideal TiO2(110) surface, and as such,

this was not considered. The adsorption of Au on TiO2 leads to a significant charge

transfer from the Au to the neighboring Ti5c cation as is shown in Figure S6C. The charge

transfer along with the unique di-σ binding of O2 at the perimeter Ti-Au site significantly

enhances its adsorption (Figure S8) and promotes the oxidation of CO (11, 13). A

thorough search for the adsorption of O2 at all of the sites along the Au/TiO2 perimeter

and on the Au nano-rod was carried out to help identify the low barrier O2 activation

sites. The co-adsorption of CO at neighboring sites was found to help facilitate the

activation of O2 either indirectly via through-surface electronic interactions or directly by

the formation of a reactive O*-O-C*=O intermediate (where * denotes the atoms bound

to the surface), as was recently reported for CO oxidation on CO covered Pt particles

(61). The results for the coadsorption of CO and O2 and the activation of O2 are

summarized in Figure S9.

The 5 possible sites investigated for O2 activation are: A) the bridge oxygen O

vacancy site on the TiO2 perimeter; B) the Ti5c sites at the perimeter; C) the Au atoms on

the nanorod at the TiO2 interface; D) the Ti5c-Au perimeter site for the CO-O2

mechanisms; and E) the Ti5c-Au perimeter site for the direct O2 dissociation mechanism.

The initial, transition (TS), and the final states for O2 activation, together with the energy

required to break the O-O bond are reported for all of the adsorption states in Figure S9.

The configurations shown in Figures S9B-D were found to follow the CO-O2 bimolecular

mechanism, while the remaining configurations involve direct O-O bond scission. The

activation barriers reported for the CO-O2 bimolecular dissociation steps on the oxide

alone (0.65 eV -Figure S9B) and the metal alone (0.7 eV - Figure S9C) were found to be

much higher than the other three paths.

While the direct activation of O2 at an O vacancy site adjunct to Ti5c has the lowest

barrier (0.12 eV Figure S9A), it is very unlikely to be catalytically active, as one of the

dissociated O atoms will fill this site eliminating it from the catalytic cycle. The most

active sites therefore appear to occur at the Ti5c-Au perimeter sites. The bimolecular CO-

O2 mechanism for the dissociation of O2 has a barrier of only 0.16 eV (Figure S9D)

which is less than half that required for the direct O2 dissociation (0.39 eV – Figure S9E).

The results in Figure S9 clearly indicate that O2 activation and CO oxidation proceed at

the dual Ti5c-Au perimeter site via a CO-assisted path. This is the only site likely to be

able to contribute to the oxidation at temperatures as low as 110 K in the experiments.

C. Supply of the reactants to the active sites.

The oxidation of CO by O2 subsequently results in the formation CO2(g) and an

adsorbed oxygen atom which resides either at the Au or Ti5c site. For the next step to

proceed, a second CO molecule must diffuse to the adsorbed oxygen atom or the

adsorbed oxygen atom must diffuse to a bound CO. The calculations reported in Figures

S10-S12 provide insights into the mobility of CO and O at the Au/TiO2 interface and may

help to explain the IR observation of the selective oxidation of CO/TiO2.

Figure S10 shows the calculated barriers for the diffusion of atomic oxygen on Au

and TiO2 to be 0.65 eV and 0.75 eV, respectively. These high diffusion barriers indicate

that it is unlikely for the atomic O to diffuse away from either type of adsorption site.

Figure S11 displays the possible diffusion barriers for CO on TiO2. CO can diffuse on the

TiO2 surface along paths that connect surface Ti sites. The activation energies for the CO

diffusion paths on TiO2 identified in Figure S11 (including a pathway through an O

vacancy), were all calculated to be within the range of 0.28-0.38 eV. At O2-rich

conditions there are likely very few O vacancies present and as such the calculated

diffusion barrier for CO/TiO2 should be close to 0.33 eV. This result agrees with the 0.35

eV CO/TiO2 diffusion energy calculated by Zhao et al. (62). Finally, Figure S12 shows

the possible diffusion paths for CO along the Au nano-rod. As was discussed above, CO

only chemisorbs on the low CN Au sites. The calculated barriers for CO diffusion

indicate that CO can diffuse with barriers of 0.1 eV (from site 1) and 0.26 eV (from site

2) to the strongest adsorption site (site 3). CO becomes trapped at this site as the barriers

to diffuse to the site 1 and 2, were calculated to be 0.73 (Figure S12A) and 0.50 eV

(Figure S12B), respectively. At low temperatures CO is therefore likely localized at these

Au sites and can not diffuse to the Au/TiO2 perimeter sites.

D. The Effect of CO Coverage DFT calculations were carried out for all of the elementary steps shown in Figure 3 at

CO coverages of θCO =1/3 and 2/3 ML CO on TiO2 to establish the influence of coverage

on the activation and overall reaction energies. The results reported in Table S1 indicate

that the changes in the calculated barriers and reaction energies in moving from low

coverage to high coverage are less than 0.05 eV. The small changes in the activation

barriers and reaction energies with coverage are typical of reactions on oxides as there are

few lateral interactions between adsorbates as they are much further apart (63).

V. References for the Supporting Online Material. 1. M. Haruta, T. Kobayashi, H. Sano, N. Yamada, Novel gold catalysts for the oxidation of carbon monoxide at a

temperature far below 0 ºC. Chem. Lett. 2, 405 (1987). 2. M. Haruta et al., Low-temperature oxidation of CO over gold supported on TiO2, α-Fe2O3, and Co3O4. J.

Catal. 144, 175 (1993). 3. R. Meyer, C. Lemire, S. K. Shaikhutdinov, H.-J. Freund, Surface chemistry of catalysis by gold. Gold Bull. 37,

72 (2004). 4. M. S. Chen, D. W. Goodman, The structure of catalytically active gold on titania. Science 306, 252 (2004);

10.1126/science.1102420. 5. J. A. Rodriguez et al., Activity of CeOx and TiOx nanoparticles grown on Au(111) in the water-gas shift

reaction. Science 318, 1757 (2007); 10.1126/science.1150038. 6. T. Janssens et al., Insights into the reactivity of supported Au nanoparticles: Combining theory and

experiments. Top. Catal. 44, 15 (2007). 7. F. Boccuzzi et al., Au/TiO2 nanosized samples: A catalytic, TEM, and FTIR study of the effect of calcination

temperature on the CO oxidation. J. Catal. 202, 256 (2001). 8. M. A. Bollinger, M. A. Vannice, A kinetic and DRIFTS study of low-temperature carbon monoxide oxidation

over Au-TiO2 catalysts. Appl. Catal. B-Environ. 8, 417 (1996). 9. M. Kotobuki, R. Leppelt, D. A. Hansgen, D. Widmann, R. J. Behm, Reactive oxygen on a Au/TiO2 supported

catalyst. J. Catal. 264, 67 (2009). 10. S. Laursen, S. Linic, Geometric and electronic characteristics of active sites on TiO2-supported Au nano-

catalysts: insights from first principles. Phys. Chem. Chem. Phys. 11, 11006 (2009). 11. L. M. Molina, M. D. Rasmussen, B. Hammer, Adsorption of O2 and oxidation of CO at Au nanoparticles

supported by TiO2(110). J. Chem. Phys. 120, 7673 (2004). 12. I. N. Remediakis, N. Lopez, J. K. Nørskov, CO oxidation on rutile-supported Au nanoparticles. Angew.

Chem. Int. Ed. 44, 1824 (2005). 13. Z.-P. Liu et al., Catalytic role of metal oxides in gold-based catalysts: a first principles study of CO oxidation

on TiO2 supported Au. Phys. Rev. Lett. 91, 266102 (2003). 14. J. G. Wang, B. Hammer, Role of Au+ in supporting and activating Au7 on TiO2(110). Phys. Rev. Lett. 97,

136107 (2006). 15. A. Linsebigler, G. Lu, J. T. Yates, Jr., CO chemisorption on TiO2(110): Oxygen vacancy site influence on

CO adsorption. J. Chem. Phys. 103, 9438 (1995). 16. F. Boccuzzi, A. Chiorino, FTIR study of CO Oxidation on Au/TiO2 at 90 K and room temperature. An

insight into the nature of the reaction centers. J. Phys. Chem. B 104, 5414 (2000). 17. P. Dumas, R. G. Tobin, P. L. Richards, Study of adsorption states and interactions of CO on evaporated noble

metal surfaces by infrared absorption spectroscopy: II. Gold and copper. Surf. Sci. 171, 579 (1986). 18. D. C. Meier, D. W. Goodman, The influence of metal cluster size on adsorption energies: CO adsorbed on

Au clusters supported on TiO2. J. Am. Chem. Soc. 126, 1892 (2004). 19. S. Derrouiche, P. Gravejat, D. Bianchi, Heats of adsorption of linear CO species adsorbed on the Auº and Ti

+δ sites of a 1% Au/TiO2 catalyst using in situ FTIR spectroscopy under adsorption equilibrium. J. Am. Chem. Soc. 126, 13010 (2004).

20. Materials and methods are available as supporting material on Science Online. 21. N. Lopez et al., On the origin of the catalytic activity of gold nanoparticles for low-temperature CO

oxidation. J. Catal. 223, 232 (2004). 22. M. Valden, X. Lai, D. W. Goodman, Onset of catalytic activity of gold clusters on titania with the appearance

of nonmetallic properties. Science 281, 1647 (1998); 10.1126/science.281.5383.1647. 23. G. Cerrato, L. Marchese, C. Morterra, Structural and morphological modifications of sintering

microcrystalline TiO2: an XRD, HRTEM and FTIR study. Appl. Surf. Sci. 70-71, 200 (1993). 24. M. Boronat, P. Concepción, A. Corma, Unravelling the nature of gold surface sites by combining IR

spectroscopy and DFT calculations. Implications in catalysis. J. Phys. Chem. C 113, 16772 (2009). 25. I. X. Green, C. Buda, Z. Zhang, M. Neurock, J. T. Yates, Jr., IR spectroscopic measurement of diffusion

kinetics of chemisorbed pyridine through TiO2 particles. J. Phys. Chem. C 114, 16649 (2010). 26. T. Diemant, H. Hartmann, J. Bansmann, R. J. Behm, CO adsorption energy on planar Au/TiO2 model

catalysts under catalytically relevant conditions. J. Catal. 252, 171 (2007). 27. Z. Wang et al., Adsorption of CO on rutile TiO2 (110)-1x1 surface with preadsorbed O adatoms. J. Phys.

Chem. C 114, 18222 (2010). 28. J. Lee et al., Interaction of CO with oxygen adatomes on TiO2(110). J. Phys. Chem. C 115, 4163 (2011).

29. M. Daté, M. Okumura, S. Tsubota, M. Haruta, Vital role of moisture in the catalytic activity of supported gold nanoparticles. Angew. Chem. Int. Ed. 43, 2129 (2004).

30. C. G. Long et al., Kinetic evaluation of highly active supported gold catalysts prepared from monolayer-protected clusters: An experimental Michaelis-Menten approach for determining the oxygen binding constant during CO oxidation catalysis. J. Am. Chem. Soc. 130, 10103 (2008).

31. G. A. Somorjai, Y. Li, Introduction to Surface Chemistry and Catalysis (Wiley, Hoboken, NJ, ed. 2, 2010). 32. R. A. van Santen, M. Neurock, Molecular Heterogeneous Catalysis (Wiley-VCH, Weinheim, Germany,

2006). 33. R. Rioux, Ed., Model Systems in Catalysis (Springer, New York, 2010). 34. J. T. Yates, Jr., C. T. Campbell, Surface chemistry special feature: Surface chemistry: Key to control and

advance myriad technologies. Proc. Natl. Acad. Sci. U.S.A. 108, 911 (2011). 35. G. Ertl, Reactions at Solid Surfaces. (Wiley, Hoboken, NJ, 2009). 36. S. H. Overbury, V. Schwartz, D. R. Mullins, W. Yan, S. Dai, Evaluation of the Au size effect: CO oxidation

catalyzed by Au/TiO2. J. Catal. 241, 56 (2006). 37. J. Guzman, B. C. Gates, Simultaneous presence of cationic and reduced gold in functioning MgO-supported

CO oxidation catalysts: Evidence from x-ray absorption spectroscopy. J. Phys. Chem. B 106, 7659 (2002). 38. Q. Fu, H. Saltsburg, M. Flytzani-Stephanopoulos, Active nonmetallic Au and Pt species on ceria-based

water-gas shift catalysts. Science 301, 935 (2003); 10.1126/science.1085721. 39. L. M. Liu, B. McAllister, H. Q. Ye, P. Hu, Identifying an O2 supply pathway in CO oxidation on

Au/TiO2(110): A density functional theory study on the intrinsic role of water. J. Am. Chem. Soc. 128, 4017 (2006).

40. T. Fujitani, I. Nakamura, T. Akita, M. Okumura, M. Haruta, Hydrogen dissociation by gold clusters. Angew. Chem. Int. Ed. 48, 9515 (2009).

41. M. Mavrikakis, P. Stoltze, J. K. Nørskov, Making gold less noble. Catal. Lett. 64, 101 (2000). 42. R. G. S. Pala, F. Liu, Nature of reactive O2 and slow CO2 evolution kinetics in CO oxidation by TiO2

supported Au cluster. J. Chem. Phys. 125, 144714 (2006). 43. P. Basu, T. H. Ballinger, J. T. Yates, Jr., Wide temperature range IR spectroscopy cell for studies of

adsorption and desorption on high area solids. Rev. Sci. Instrum. 59, 1321 (1988). 44. R. Zanella, S. Giorgio, C. R. Henry, C. Louis, Alternative methods for the preparation of gold nanoparticles

supported on TiO2. J. Phys. Chem. B 106, 7634 (2002). 45. X. Wang et al., Direct spectroscopic observation of the rold of humidity in surface diffusion through an ionic

adsorbent powder. The behavior of adsorbed pyridine on nanocrystalline MgO. J. Phys. Chem. C 113, 2228 (2009).

46. G. Kresse, Dissociation and sticking of H2 on the Ni(111), (100), and (110) substrate. Phys. Rev. B 62, 8295 (2000).

47. G. Kresse, J. Hafner, First-principles study of the adsorption of atomic H on Ni (111), (100) and (110). Surf. Sci. 459, 287 (2000).

48. J. P. Perdew, Y. Wang, Accurate and simple analytic representation of the electron-gas correlation energy. Phys. Rev. B 45, 13244 (1992).

49. P. E. Blöchl, Projector augmented-wave method. Phys. Rev. B 50, 17953 (1994). 50. G. Kresse, D. Joubert, From ultrasoft pseudopotentials to the projector augmented-wave method. Phys. Rev.

B 59, 1758 (1999). 51. S. L. Dudarev, G. A. Botton, S. Y. Savrasov, C. J. Humphreys, A. P. Sutton, Electron-energy-loss spectra and

the structural stability of nickel oxide: An LSDA+U study. Phys. Rev. B 57, 1505 (1998). 52. B. J. Morgan, G. W. Watson, A DFT+U description of oxygen vacancies at the TiO2 rutile (110) surface.

Surf. Sci. 601, 5034 (2007). 53. H. J. Monkhorst, J. D. Pack, Special points for Brillouin-zone integrations. Phys. Rev. B 13, 5188 (1976). 54. G. Henkelman, H. Jónsson, Improved tangent estimate in the nudged elastic band method for finding

minimum energy paths and saddle points. J. Chem. Phys. 113, 9978 (2000). 55. G. Henkelman, B. P. Uberuaga, H. Jónsson, A climbing image nudged elastic band method for finding saddle

points and minimum energy paths. J. Chem. Phys. 113, 9901 (2000). 56. G. Henkelman, H. Jónsson, A dimer method for finding saddle points on high dimensional potential surfaces

using only first derivatives. J. Chem. Phys. 111, 7010 (1999). 57. H. Klimev et al., Oxidation of gold metal particles supported on TiO2: An FTIR study by means of low-

temperature CO adsorption. J. Mater. Sci. 42, 3299 (2007). 58. P. Dumas, R. G. Tobin, P. L. Richards, Interaction of CO molecules with evaporated silver, gold, and copper

films: An infrared spectroscopic investigation using a thermal detection technique. J. Electron Spectrosc. Relat. Phenom. 39, 183 (1986).

59. J. Kim, E. Samano, B. E. Koel, CO adsorption and reaction on clean and oxygen-covered Au(211) surfaces. J. Phys. Chem. B 110, 17512 (2006).

60. J.-M. Pan, B. L. Maschhoff, U. Diebold, T. E. Madey, Interaction of water, oxygen, and hydrogen with TiO2(110) surfaces having different defect densities. J. Vac. Sci. Technol. A 10, 2470 (1992).

61. A. D. Allian et al., Chemisorption of CO and mechanism of CO oxidation on supported platinum nanoclusters. J. Am. Chem. Soc. 133, 4498 (2011).

62. Y. Zhao et al., What are the adsorption sites for CO on the reduced TiO2(110)-1x1 surface? J. Am. Chem. Soc. 131, 7958 (2009).

63. K. Reuter, D. Frenkel, M. Scheffler, The steady state of heterogeneous catalysis, studied by first-principles statistical mechanics. Phys. Rev. Lett. 93, 116105 (2004).

VI. Supporting Figures.

Figure S1. Schematic drawings of the high vacuum system for transmission IR studies. A. Top view of the IR cell. B. A closeup view of the sample holder.

Figure S2. A. TEM image of the Au/TiO2 catalyst. Image was taken after all experiments are done. B. Plot of the Au particle size distribution.

Figure S3. FTIR spectra of (a) CO(a)/TiO2 in 0.060 Torr of CO at 120 K; (b) CO(a)/TiO2 in vacuum at 120 K; and (c) CO(a)/TiO2 in 1 Torr of O2 for 120 min at 120 K. The CO2

intensity increase is caused by impurity re-adsorption from the cell rather than CO oxidation reaction, judging from the stable intensity of CO(a)/TiO2.

Figure S4. Plots of the integrated absorbance of CO2 at different temperatures (110-130 K) against time for the kinetics study of CO2 generation over the Au/TiO2 catalyst. Smooth curves under the experimental data points are the best first-order fit. The inset shows an Arrhenius plot of the kinetics which results in an activation energy of 0.16 ± 0.01 eV.

Figure S5. Plots of normalized integrated absorbance of CO/TiO2 (left) and CO/Au (right) against time at various temperatures showing the diffusion/desorption processes of CO from the two surfaces, where CO/TiO2 diffuses and desorbs more rapidly than CO/Au.

Figure S6. A. Top view and B. side view of the 3-atomic-layer Au nano rod on rutile TiO2 (110) surface model. Au atoms are yellow. Ti atoms are light grey. O atoms in TiO2 lattice are pink. C. Charge density difference on the Ti5c sites noted by the dotted rectangle in B before and after Au deposition. Red regions exhibit the highest electronic charge.

Figure S7. CO adsorption energies and configurations on different Au sites or on the Ti5c sites. The distances between the CO and the adsorption site are noted. C atoms are black. O atoms in CO are red.

Figure S8. O2 adsorption energy at different Au sites.

Figure S9. The initial, transition and final states for the adsorption and activation of O2 along with the corresponding activation energies for the dissociation of the O-O bond at 5 different sites: A) O-vacancy on the TiO2 support; B) Ti5c sites at the perimeter; C) two Au atoms on the periphery of the Au nanorod ; D) Ti5c-Au perimeter site via CO-O2 mechanism; E. Ti5c-Au perimeter site via direct O2 dissociation mechanism. Oxygen atoms in CO and O2 molecule are red. Hydrogen atoms are cyan.

Figure S10. Atomic O diffusion barrier on Au (A) and along Ti5c row (B).

Figure S11. CO diffusion barriers on TiO2. The position of the O-vacancy under consideration is highlighted by a green circle.

Figure S12. CO diffusion barriers on Au sites.

VII. Supporting Tables. ________________________________________________________________________ Elementary Step Ea (θCO=1/3 ML) eV Ea(θCO=2/3 ML) eV CO(a) diffusion along TiO2 surface 0.33 0.26 CO(a) + O2(a) -> O-C(a)-O-O(a) 0.16 (-0.28) 0.13 (-0.32) O-C(a)-O-O(a) -> CO2(g) + O(a) 0.10 (-3.68) 0.10 (-3.66) CO(a) + O(a) -> CO2(g) + 2a 0.11 (-1.23) 0.11 (-1.20) CO2(a) -> CO2(g) + a 0.20 0.20 ________________________________________________________________________ Table S1. The effects of coverage on the activation and overall reaction energies shown in ( ) for elementary CO and O2 adsorption, CO oxidation, CO diffusion, and CO2 desorption steps. DFT-calculated activation energies and for overall reaction energies for 1/3 ML and 2/3 ML of CO on TiO2.