Investigation of Alumina-Supported Au Catalyst for CO Oxidation by

May 29, 2004 - Alumina-supported Au particles (1.16 wt %) were prepared by a deposition−precipitation method involving a HAuCl4 precursor. X-ray abs...
14 downloads 14 Views 160KB Size
Download PDF
J. Phys. Chem. B 2005, 109, 2307-2314

2307

Investigation of Alumina-Supported Au Catalyst for CO Oxidation by Isotopic Transient Analysis and X-ray Absorption Spectroscopy† Jason T. Calla and Robert J. Davis* Department of Chemical Engineering, UniVersity of Virginia, CharlottesVille, Virginia 22904-4741 ReceiVed: March 14, 2004; In Final Form: April 28, 2004

Alumina-supported Au particles (1.16 wt %) were prepared by a deposition-precipitation method involving a HAuCl4 precursor. X-ray absorption spectroscopy at the Au LIII edge was used to monitor the evolution of the Au oxidation state and atomic structure during pretreatment in He up to 623 K. Although the as-prepared material had Au present in a +3 oxidation state, thermal treatment at 623 K facilitated autoreduction of Au cations to metal particles. Analysis of the EXAFS revealed a coordination number (Au-Au) of 7.2, which is consistent with spherical particles of 1.2 nm in average diameter. Steady-state isotopic transient kinetic analysis was used to evaluate the intrinsic turnover frequency (TOFintr) and the surface coverage of carbon-containing species (θCOx) on the gold catalyst during CO oxidation at 1.2 atm total pressure and 296 K. The artifacts in the kinetic parameters caused by re-adsorption of product carbon dioxide were removed by varying the total flow rate. The values of TOFintr and θCOx determined from the intrinsic lifetime of surface intermediates at infinite flow rate were 1.6 s-1 and 4.9%, respectively. The intrinsic turnover frequency was nearly independent of temperature, indicating a very low activation energy for the reaction. However, the rate was significantly accelerated by the presence of water.

Introduction Catalysts containing gold nanoparticles are very active in a wide variety of chemical reactions. However, the underlying principles that govern the activity of supported gold particles are currently unknown.1 Bulk gold metal does not react with molecules such as H2 and O2 because of repulsion between the orbitals of the adsorbate and the filled d states of gold.2 In 1989, Haruta et al. reported that supported Au nanoparticles are extremely active in carbon monoxide oxidation.3 Supported gold is now recognized as a catalyst for other oxidation reactions, as well as hydrogenation, water-gas shift, and nitrogen oxide reduction.4 Explanations for the high activity of gold nanoparticles can be broadly grouped into four categories, namely, an influence of metal particle size, the presence of cationic Au atoms, electron transfer from the support to Au, and a separate catalytic role of the support. Boccuzzi et al. used IR spectroscopy and isotopic scrambling to study the nature of Au/ZnO and Au/TiO2 catalysts for CO oxidation.5 They concluded that Au metal particles adsorb both CO and O2 simultaneously, leading to rapid reaction. Valden et al. used a combination of CO oxidation studies and scanning tunneling microscopy/spectroscopy to show that 3.5 nm Au particles on a planar titania substrate are the most active.6 Because the onset of a metal-to-nonmetal transition was also observed to occur at that diameter, the researchers speculate that the high activity results from a quantum size effect with respect to the thickness of the gold. Low coordinated Au sites found on small particles have also been suggested to play a role in the activity of supported Au.7,8 In addition, quantum †

Part of the special issue “Michel Boudart Festschrift”. * To whom correspondence should be addressed. E-mail: rjd4f@ virginia.edu.

chemical calculations suggest an energetically feasible reaction path for CO oxidation on zerovalent gold particles containing 10 atoms.7 Other researchers suggest that cationic Au is critical to its activity. Indeed, the role of Au cations in the water-gas shift reaction has been demonstrated recently by Fu et al.9 They deposited gold onto a La-doped ceria support and calcined the material in air at 673 K. Most of the gold was reduced to metallic particles by the thermal treatment. A basic NaCN solution was then used to leach the metallic gold from the surface. X-ray photoelectron spectroscopy showed that the remaining gold was exclusively cationic. Interestingly, the catalytic activity of the supported Au sample was unaffected by the loss of the metallic particles. In this specific case, neutral gold metal particles are apparently inert byproducts of the preparation method. Guzman and Gates also suggest an exclusive role of cationic gold in ethene hydrogenation.10 The researchers used Au3+ organometallic precursors to prepare MgO-supported gold nanoparticles. After monitoring the chemical state of the Au with in-situ X-ray absorption spectroscopy, they concluded that mononuclear Au3+ is the active species and Au0 is inactive. A growing body of evidence suggests that CO oxidation requires both metallic and cationic gold to be present simultaneously in a highly active catalyst.4,11-16 In particular, Au+OH- has been suggested to form at the periphery of metallic particles in contact with the support. Date and Haruta report a large promoting effect of water vapor on the CO oxidation rate over gold/titania, which is consistent with a possible role of hydroxyl in the mechanism.17 Contrary to the speculation that cationic gold is relevant for CO oxidation, another school of thought proposes that anionic gold is the key feature of an active catalyst. A combination of work on gas-phase, anionic clusters,18 MgO-supported Au clusters,19,20 and quantum chemical calculations18-20 suggests

10.1021/jp0488719 CCC: $30.25 © 2005 American Chemical Society Published on Web 05/29/2004

2308 J. Phys. Chem. B, Vol. 109, No. 6, 2005 that oxygen vacancy F-center defects at the metal-support interface facilitate electron transfer to the Au metal particle and activate it for catalysis. While the role of the support may be to influence the electronic structure of the Au at the interface, others hypothesize that the support may actually provide complementary adsorption sites for the reactants. For example, deposition of TiOx overlayers onto inactive Au powder produced high activity, which argues against an electronic effect on the Au.21 Results from IR spectroscopy and reaction kinetics are consistent with the idea that CO adsorbs on the Au metal particle and O2 adsorbs on the titania component at the metal-support interface.21 Others have also speculated that O2 is activated at oxygen defects at the metal-support interface, whereas CO adsorbs on the metal particle.22 Boccuzzi et al. also suggest that O2 can be activated on the titania support at room temperature.23 The ideas presented here and above indicate a very important role of the metal-support interface, regardless of the details of promotion. Additional information can be found in recent reviews of gold catalysis.4,24,25 In this paper, we explore the reactivity of Au/Al2O3 as a catalyst for CO oxidation at room temperature. Alumina was chosen for this study because it is a nonreducible oxide that is known to be an effective support for Au nanoparticle catalysts.12,16,26,27 X-ray absorption spectroscopy at the Au LIII edge was used to evaluate the chemical state and the average particle size of the Au. The Au/Al2O3 sample was then tested as a catalyst for the steady-state oxidation of CO with O2. Since common probe molecules do not effectively chemisorb on surface Au atoms, turnover frequencies reported in the literature are often based on the total loading of Au in the sample or on the surface density of Au atoms estimated from electron microscopy. Isotopic transient analysis can be used to evaluate a turnover frequency (TOF) based on the number of reactive intermediates adsorbed on the catalyst during reaction. Steady-state isotopic transient kinetic analysis is very powerful because it allows for the characterization of a catalyst surface under working conditions. The experiment is carried out at steady state following a step change in the isotopic content of a reactant (CO in this case). The pressure, temperature, total flow rate, and product composition are constant during this step change. A mass spectrometer is used to record the transient response of isotopically labeled product (CO2). Therefore, in the absence of isotopic mass effects, the steady state is not altered during the step change. From the transient response, the surface coverage of adsorbed reaction intermediates that lead to carbon dioxide and the residence time of adsorbed reaction intermediates (COx) can be determined. Recently, Shannon and Goodwin28 reviewed the theory underlying this technique, and Efstathiou and Verykios29 reviewed how the method was applied to a variety of reacting systems. Our group has previously used isotopic transient methods to study ammonia synthesis over a variety of supported Ru catalysts.30-33 Experimental Methods Catalyst Preparation. The catalyst was prepared by a deposition-precipitation method based on a previous report.16 Herein, 0.26 g of HAuCl4 (Aldrich, 99.9+%) was added to 80 mL of distilled, deionized water. The gold solution was heated to 343 K and adjusted to pH ) 7 with NaOH (Mallinckrodt, 98.6%). The gold solution was then added to a second flask containing 5 g of γ-alumina (Mager Scientific, AP-312) suspended in 120 mL of distilled, deionized water also at 343 K. After stirring for 2 h, the solution was removed by suction

Calla and Davis

Figure 1. Schematic representation of the reactor system used for isotopic transient analysis.

filtration and the catalyst was resuspended in 100 mL of distilled, deionized water at 343 K for an additional 20 min. This washing procedure was repeated 3 more times after which the filtered catalyst was dried in air at about 310 K for 24 h. The Au loading was 1.16 wt %, as reported by Galbraith Laboratories, Inc. (Knoxville, TN). X-ray Absorption Spectroscopy. The X-ray absorption spectra at the Au LIII edge were recorded on beam line X10C at the National Synchrotron Light Source, Brookhaven National Laboratory, Upton, NY. The storage ring operated at 2.8 GeV with currents ranging from 150 to 300 mA. Several Au reference compounds were diluted in boron nitride powder (Alfa Aesar, 99.5%) and pressed into self-supporting wafers. The spectra of the reference compounds Au2O3, AuCl3, and AuCl (all from Alfa Aesar, 99.99%) were collected in the transmission mode in air at ambient temperature. In addition, a light-tight Au foil (0.005 mm thick, Goodfellow, 99.9%) was used to derive AuAu phase shift and amplitude functions. The spectra of the catalyst sample were collected in the fluorescence mode in a cell capable of heating and cooling the samples in a controlled atmosphere. Four scans were used for structural determination. Data analysis was performed with the WinXAS software package. Isotopic Transient Measurements during CO Oxidation. Approximately 0.090 g of catalyst (-100/+140 mesh) was mixed with 0.4 g of SiC powder (Universal Photonics, Inc.; 120 mesh) and loaded into a quartz tubular reactor. The catalyst was first pretreated at 623 K for 4 h under flowing He (99.999%, BOC Gases, further purified by Supelco OMI-2 filter) prior to cooling to 296 K and admitting the reactant gases: unlabeled CO (99.997%, Messer), O2 (99.999%, BOC Gases), and Ar (99.999%, BOC gases, further purified by Supelco OMI-2 filter) to the reactor. The mole percentages of the reactant stream He: CO:O2:Ar were 95:2:2:1, respectively. All of the reactant gases except Ar were additionally purified by passage through a silica gel trap (Davisil Grade 635, Type 60A, 60-100 mesh) held at dry ice-acetone temperature. A schematic representation of the apparatus can be found in Figure 1. A step change in isotopically labeled carbon monoxide

Investigation of Alumina-Supported Au

J. Phys. Chem. B, Vol. 109, No. 6, 2005 2309

Figure 2. X-ray absorption near-edge spectra at the Au LIII edge of Au reference compounds. Spectra are offset for clarity. The energies are defined relative to the first inflection point in the LIII edge of Au foil.

Figure 3. X-ray absorption near-edge spectra at the Au LIII edge of the Au/Al2O3 sample during thermal pretreatment in He. Spectra are offset for clarity. The energies are defined relative to the first inflection point in the LIII edge of Au foil.

was accomplished by switching between a stream of 12CO and 13CO (Cambridge Isotopes, 99.5% CO, 98+% 16O, 99+% 13C), further purified with a cold silica gel trap). The feed stream containing 12CO was mixed with Ar as an inert gas tracer. Gases were supplied via mass flow controllers, and back-pressure regulators ensured that all reactant streams and vent lines were maintained at 1.2 atm. The isotopic switch was accomplished by injecting a known volume of the reactant mixture containing 13CO instead of 12CO into the feed stream. Since Ar was not present in the injection volume, the decay of the Ar signal was used to monitor the gas-phase holdup of the system. A Balzers-Pfieffer Prisma 200 amu mass spectrometer monitored the concentrations of Ar, 12CO2, and 13CO2 (m/e ) 40, 44, and 45, respectively) continuously. The lines from the reactor effluent to the mass spectrometer as well as the mass spectrometer housing were heated to about 443 K. Multiple switches (12CO/Ar f 13CO/He) were recorded at each set of experimental conditions.

a large threshold peak is present in the spectrum. As the temperature of the sample increased from 298 to 623 K, the threshold peak decreased in intensity, indicating autoreduction of the supported Au in flowing He. At the highest temperature of 623 K, essentially all of the Au was reduced to the metallic state. The extended X-ray absorption fine structure (EXAFS) was used to characterize the atomic structure of the reduced Au/ Al2O3 sample at room temperature. Figure 4a shows the EXAFS function of the sample in He after thermal treatment, and Figure 4b illustrates the Fourier transform of the EXAFS function. The structural parameters N (Au-Au coordination number), RAuAu (interatomic distance), and ∆σ2 (change in Debye-Waller factor compared to Au foil) were obtained by curve-fitting the first shell EXAFS data in both R space and k space. The curve fits are shown together with the experimental results in Figure 4b,c. The values of N, RAuAu, and ∆σ2 were calculated to be 7.2, 2.81 Å, and 0.0031 Å2, respectively. Gold is a face-centered cubic metal having a coordination number in the bulk equal to 12. Thus, the small value of N for the Au/Al2O3 sample indicates a very high dispersion of Au on the support. Assuming the Au particles are spherical, a first shell coordination number of 7.2 is consistent with an average particle diameter of approximately 1.2 nm.36 Particles of that size expose about 80% of the atoms to the surface. The interatomic distance of 2.81 Å is significantly shorter than that found in bulk gold (2.88 Å), which is also consistent with a very high dispersion of the Au particles. Moreover, the positive change in Debye-Waller factor compared to Au foil indicates greater thermal disorder, typical of highly dispersed metal particles. A typical set of isotopic transients for carbon dioxide following a switch between 12CO/Ar and 13CO/He can be found in Figure 5. The average residence time of surface intermediates leading to carbon dioxide (τ) is calculated by integrating the area between the normalized transients,

Results The chemical state of the Au catalyst was determined by X-ray absorption spectroscopy after comparison to spectra of Au reference compounds. Figure 2 summarizes the X-ray absorption near-edge structure (XANES) of compounds with Au in various formal oxidation states. Absorption at the LIII edge corresponds to the excitation of a 2p3/2 electron to unoccupied states above the Fermi level. Dipole selection rules dictate that the final state must be of s- or d-type symmetry, and calculations have shown that transitions at the LIII edge are dominated by final states of d symmetry.34 As shown in Figure 2, Au3+ compounds such as Au2O3 and AuCl3 have many d-type vacancies and therefore exhibit a large threshold absorption peak at the LIII edge. A smaller threshold peak is observed for AuCl since Au is in a lower oxidation state (+1). For zerovalent Au foil, a shoulder is observed at the LIII edge. Although the electronic structure of atomic Au is 5d106s1, hybridization with 6s and 6p states in the solid results in a finite density of unoccupied d states near the Fermi level.35 Figure 3 shows the evolution of the catalyst XANES with increasing pretreatment temperature. At ambient conditions, the Au on the as-prepared catalyst appeared to be in the +3 formal oxidation state since

τ)

∫0∞[F

- FAr] dt

12CO 2

(1)

where F12CO2 and FAr are the normalized transient responses of 12CO and Ar, respectively. 2 Ali and Goodwin found that product re-adsorption had a substantial impact on the determination of surface reaction

2310 J. Phys. Chem. B, Vol. 109, No. 6, 2005

Calla and Davis

Figure 5. Example of an isotopic transient during CO oxidation over Au/Al2O3.

Figure 6. Effect of flow rate on τ.

Figure 4. Extended X-ray absorption fine structure above the Au LIII edge of the Au/Al2O3 sample in He at room temperature after pretreatment in He at 623 K. (a) Experimental k3χ(k) data. (b) The magnitude and imaginary component of the Fourier transform (uncorrected for phase shift) of k3χ(k) from 2.6 to 13.6 Å-1 (results from the first shell curve fit are plotted as the dashed line). (c) Comparison of fitted EXAFS function (dashed line) to back-transform (solid line) of first shell radial structure function from 1.6 to 3.6 Å.

parameters obtained from transient analysis during methanol synthesis on Pd/SiO2.37-39 Furthermore, a previous study in our

laboratory showed severe product re-adsorption during transient experiments involving ammonia synthesis on a zeolite-supported Ru catalyst.31 In the current work, it is possible that carbon dioxide re-adsorbed on catalytically active and inactive sites before leaving the reactor, causing an overestimation of τ. However, the global rate of CO oxidation, RCO2, was unaffected by the overestimation. The influence of carbon dioxide readsorption was therefore examined in this study. The effect of flow rate on the value of τ for the Au/Al2O3 catalyst at 296 K is given in Figure 6 (runs 1-3 in Table 1). By extrapolating τ to infinite flow rate (1/F ) 0), a better estimate of the surface lifetime of reactive intermediates (τ0) can be calculated.38 This value was found to be 0.56 s at 296 K. Evidently, re-adsorption of carbon dioxide results in an overestimation of τ and must be correctly accounted for in the analysis. Since increasing the total flow rate eliminates only the effect of interparticle readsorption, τ0 may still be an overestimation of the intrinsic surface lifetime due to carbon dioxide re-adsorption within the catalyst pores. One way to check for intraparticle re-adsorption of product formed by reaction is to simultaneously co-feed unlabeled product.38 In this case, co-fed carbon dioxide will competitively adsorb in the pores, thus minimizing the re-adsorption of CO2 formed catalytically. Variation of the co-fed CO2 pressure allows the

Investigation of Alumina-Supported Au

J. Phys. Chem. B, Vol. 109, No. 6, 2005 2311

TABLE 1: Isotopic Transient Results during CO Oxidation over Au/Al2O3a

run

T (K)

flow rate (mL min-1)

1 2 3 4 5 6 7 8 9

296 296 296 296 296 296 296 296 296

100 155 239 101 156 239 101 156 239

a

inlet CO2 concn (mol %)

outlet CO2 concn (mol %)

global reaction rate (mol of CO2 (mol of Au)-1 s-1)

fractional conversion of CO

τb (s)

0 0 0 0.40 0.38 0.38 0.73 0.72 0.72

0.44 0.26 0.18 0.85 0.61 0.70 1.07 1.20 0.97

0.055 0.054 0.055 0.057 0.064 0.070 0.061 0.067 0.076

0.22 0.14 0.09 0.23 0.12 0.16 0.17 0.24 0.13

2.64 ( 0.19 1.87 ( 0.05 1.45 ( 0.17 2.38 ( 0.15 1.84 ( 0.40 1.44 ( 0.17 2.07 ( 0.05 1.45 ( 0.09 1.25 ( 0.06

Experimental conditions: 1.2 atm total pressure; He:CO:O2:Ar ) 95:2:2:1. b Error refers to (1 standard deviation.

Figure 7. Effect of flow rate on measured τ with varying levels of co-fed carbon dioxide to the reactor. The mole fraction of carbon dioxide in the reactor, yCO2(av), was assumed to be the average of the inlet and outlet mole fractions.

effect of CO2 re-adsorption to be decoupled from the observed lifetime of catalytic reaction intermediates. We examined the effect of yCO2 on τ for various flow rates in order to remove artifacts from both inter- and intraparticle CO2 re-adsorption. This was accomplished by simultaneously feeding unlabeled carbon dioxide (12CO2) in the reactant stream together with 12CO and O2 and performing an isotopic switch to 13CO while still co-feeding 12CO2. The results from these experiments are summarized in Table 1 (runs 4-9). The value of yCO2 used in this analysis is the average of the CO2 mole fraction measured at the reactor inlet and the outlet. The relationships between yCO2 and τ for various flow rates are presented in Figure 7. Although there was a small influence of yCO2, the total flow rate had a much greater impact on τ than did yCO2. Values of τ at constant yCO2 can be derived from results in Figure 7. Figure 8 presents the dependence of τ on the flow rate, but now at several different values of yCO2. As expected, τ decreased with increasing flow rate. By extrapolating the values of τ in Figure 8 to infinite flow rate (1/F ) 0), a value of τ that is corrected for both inter- and intraparticle readsorption of CO2 can be obtained (τ00). The average τ00 calculated from the results in Figure 8 was 0.63 ( 0.08 s, which is similar to the value obtained after removal of interparticle re-adsorption effects (τ0 ) 0.56 s, from Figure 6). Evidently, artifacts from intraparticle re-adsorption were negligible compared to those from interparticle re-adsorption. Figure 8 also displays the results from a reactor containing Au-free Al2O3 and SiC in the same proportion as a reactor

Figure 8. Effect of flow rate at constant yCO2(av) during steady-state CO oxidation at 296 K and 1.2 atm, used to determine τ00. For comparison, a transient response associated with a CO2 switch over a sample containing SiC and Al2O3 (without Au) is also included on the plot.

containing the Au catalyst. Since the CO oxidation reaction does not occur in the absence of Au, the transient in Figure 8 is the result of a flowing stream of He containing a small amount of CO2 (yCO2 ) 0.55%) being switched to a He stream with no CO2. Extrapolation of the τ values to infinite flow rate gives a very small negative intercept, indicating a negligible τ0 in the absence of Au. Therefore, the τ0 and τ00 values measured on the catalyst result from the Au component of the sample. The moles of reactive surface intermediates per mole of Au, NCO2, was calculated using the following steady-state mass balance:

NCO2 ) τoo × RCO2

(2)

where RCO2 is the steady-state rate of CO oxidation per mole of Au atoms. If an uncorrected value of τ were used in eq 2, then the determination of the number of surface intermediates would be overestimated. The value of NCO2 was found to be 0.039 mol/ (mol of Au), based on the results in Table 1. The surface coverage of reactive intermediates can then be calculated from the Au dispersion, estimated from EXAFS analysis to be about 80%. In the absence of artifacts from inter- and intraparticle re-adsorption of CO2, the surface coverage of C-containing intermediates on Au/Al2O3 was found to be about 4.9% at 296 K. Since the effects of inter- and intraparticle re-adsorption were removed, the value of τ00 is a good approximation of the intrinsic

2312 J. Phys. Chem. B, Vol. 109, No. 6, 2005

Calla and Davis

TABLE 2: Effect of Temperature on Isotopic Transient Results during CO Oxidation over Au/Al2O3a

run

T (K)

flow rate (mL min-1)

10 11 12 13 14 15 16 17 18

273 273 273 296 296 296 329 328 329

100 155 238 100 155 238 100 155 238

a

outlet CO2 concn (mol %)

global reaction rate (mol of CO2 (mol of Au)-1 s-1)

fractional conversion of CO

τb (s)

0.20 0.13 0.07 0.31 0.21 0.13 0.38 0.26 0.16

0.025 0.024 0.021 0.039 0.041 0.040 0.047 0.050 0.048

0.10 0.07 0.04 0.16 0.11 0.07 0.19 0.13 0.08

4.76 ( 0.01 3.34 ( 0.03 2.78 ( 0.12 5.02 ( 0.03 3.76 ( 0.15 2.74 ( 0.02 4.93 ( 0.07 3.61 ( 0.10 2.81 ( 0.09

τ0 (s)

1.23 1.15 1.25

Experimental conditions: 1.2 atm total pressure; He:CO:O2:Ar ) 95:2:2:1. Error refers to (1 standard deviation. b

TABLE 3: Effect of Water Vapor on Isotopic Transient Results during CO Oxidation over Au/Al2O3a

run

T (K)

flow rate (mL min-1)

19 20 21 22 23

295 295 296 298 297

155 201 267 200 281

a

water concn (mol %)

outlet CO2 concn (mol %)

global reaction rate (mol of CO2 (mol of Au)-1 s-1)

fractional conversion of CO

τb (s)

0 0 0 0.18 0.18

0.22 0.16 0.12 0.99 0.76

0.044 0.042 0.039 0.251 0.271

0.11 0.08 0.06 0.50 0.38

2.45 ( 0.04 1.97 ( 0.05 1.83 ( 0.08 1.54 ( 0.02 1.41 ( 0.06

Experimental conditions: 1.2 atm total pressure; He:CO:O2:Ar ) 95:2:2:1. b Error refers to (1 standard deviation.

activity of the catalytic sites, as determined by

TOFintr ) 1/τ00

(3)

The intrinsic turnover frequency (TOFintr), as defined by eq 3, over Au/Al2O3 at 296 K was found to be about 1.6 s-1. The classical calculation of turnover frequency involves the total number of surface metal atoms in a reactor. Since all of the surface atoms may not participate in the reaction, the conventional TOF provides a lower bound on the normalized rate. In contrast, the intrinsic TOF defined by eq 3 provides an upper bound because it is based on the actual surface concentration of intermediates under reaction conditions. If the active surface sites are not saturated at the time of the isotopic switch, then TOFintr will be greater than the true turnover frequency of the catalyst. The reproducibility of the method was tested by reloading the reactor with the same catalyst and performing the analysis again. Since intraparticle re-adsorption was found to be negligible compared to interparticle re-adsorption, τ0 is a very good approximation of τ00. The results for the repeated experiments are shown in Table 2 (runs 13-15), which gives a value of τ0 ) 1.15 s, which is slightly higher than the value of τ00 found from the results in Table 1. Experiments were also performed at 273 and 329 K (runs 10-12 and 16-18, respectively, in Table 2). The apparent activation energy determined from the global rates reported in Table 2 was 10 kJ mol-1. The values of τ0 were unaffected by temperature over the range of this study. A low apparent activation energy for CO oxidation is typical for gold catalysts. A final set of experiments involved the addition of water vapor to the reactant stream. These experiments were performed with a new charge of catalyst in the reactor. The results of those experiments are summarized in Table 3. Addition of 0.18% water to the feed accelerated the global rate by about a factor of 6 compared to a dry run. Under conditions of similar flow rate (runs 20 and 22 or runs 21 and 23), the measured τ values for the humid conditions were about 0.4 s lower than those for the dry conditions. Since the τ0 associated with the dry conditions in Table 3 is 0.91 s, the value of τ0 under humid

conditions is estimated to be 0.51 s. The TOFintr associated with the humid and dry conditions for this catalyst charge is therefore 1.1 and 2.0 s-1, respectively. Because the global rate was accelerated by about a factor of 6 in the presence of water, whereas the intrinsic turnover frequency only doubled, the surface coverage of intermediates apparently increased by about a factor of 3. The promoting effect of water on the reaction appears to increase both the intrinsic rate of the catalytic cycle as well as the coverage of carbon-containing intermediates. Additional work is needed to better describe the results quantitatively since the time constants are very low in the presence of water. The reproducibility of this technique can be evaluated by comparing the global rates and τ0 values from the different catalyst charges in the reactor as summarized in the tables. For example, the average global rates in Tables 1-3 for CO oxidation at 296 K in the absence of water vapor are 0.055, 0.040, and 0.041 mol of CO2 (mol of Au)-1 s-1, respectively. The corresponding values of τ0 are 0.56, 1.15, and 0.91 s. Although the global rates are very reproducible, there is some uncertainty with respect to τ0 because of extrapolation to infinite flow rate. Thus, experiments involving changes in temperature or water vapor were compared directly to reference experiments performed just prior to changes in conditions. Discussion The catalyst preparation method used here involves an aqueous solution at neutral pH that hydrolyzes the chlorauric anion prior to deposition on the alumina. As indicated by the large peak in the Au LIII XANES, the predominant form of the Au after deposition and washing is Au3+, presumably in the form of Au(OH)3.16 Thermal treatment at 623 K in He caused autoreduction of the supported Au hydroxide. The lack of any discernible peak at the absorption threshold indicates that all of the Au reduced to the metallic state, within the resolution of the XANES experiment. Because very small particles (∼1 nm) can be difficult to visualize by electron microscopy, we chose to use EXAFS to estimate the Au dispersion. The EXAFS results from a thermally activated catalyst revealed an average first-

Investigation of Alumina-Supported Au shell coordination number of 7.2, which is consistent with spherical particles of 1.2 nm exposing 80% of the atoms to the surface. The global rate of CO oxidation at 296 K was 0.055 mol of CO2 (mol of Au)-1 s-1 (Table 1). Assuming a dispersion of 80% determined by EXAFS, the measured turnover frequency in this work was 0.069 s-1. Kung et al. have summarized the reported rates of CO oxidation over a variety of Au/Al2O3 catalysts.16 The turnover frequency of most of the catalysts in their comparison ranges from 0.004 to 0.018 s-1 at 298 K and from 0.006 to 0.02 s-1 at 273 K. The one exception is a catalyst prepared in Kung’s laboratory that ranged from 0.17 to 0.46 s-1 at 295 K. Whereas the catalyst used in the current study was quite active for CO oxidation, its steady-state activity is not as high as the most active catalyst reported by Kung et al. under comparable conditions. One possible reason for the difference is that rates reported by Kung et al. were determined after about 30 min on stream, whereas the rates reported in the current work were obtained after at least 36 h on stream. A second plausible reason for the difference is the higher dispersion of the catalyst in this work. Kung et al. reported on Au particles with diameters ranging from 3 to 5 nm, which are substantially larger than the ones in the current study. Thus, on a total gold basis, the catalyst described in this paper is comparable in activity to those of Kung et al. Isotopic transient analysis was used to evaluate the average residence time and surface coverage of carbon-containing reactive intermediates. However, we discovered that re-adsorption of product CO2 along the catalyst bed affected the measured values of τ. Since alumina exposes both acidic and basic surface sites, interaction of CO2 with the alumina is expected.40 Fortunately, extrapolation of the measured values of τ to infinite flow rate effectively removed the re-adsorption of CO2 throughout the bed. Results from experiments with co-fed CO2 verified that intraparticle readsorption was not important in this analysis. The intrinsic turnover frequency based on τ00 was calculated to be 1.6 s-1, which corresponds to a surface coverage of COx intermediates of 4.9% at the steady state. A low surface coverage of intermediates is consistent with the CO adsorption results reported by Margitfalvi et al.15 They reported that saturation of the Au adsorption sites with CO was not achieved until about 800-1000 Torr. However, it is not clear how the CO/Ausurf stoichiometry was evaluated. Since the pressure of CO used in our reaction studies was only about 20 Torr, the Au surface should be far below saturation. According to the results of Margitfalvi et al., approximately 20% of the available sites for CO adsorption will be occupied at 20 Torr of CO. However, the coverage of COx species found here during the oxidation reaction is substantially lower than that predicted by the results of Margitfalvi et al. Although competitive adsorption with O2 might lower the steady-state coverage of CO under reaction conditions, it is also plausible that the CO/Ausurf stoichiometry might be much less than unity at saturation. Infrared spectroscopy has been used extensively to characterize the interaction of CO with Au catalysts.5,15,21-23,41-43 The evidence suggests that CO adsorbs in many forms on supported Au, i.e., on support sites, on metallic Au surfaces, on low coordinated Au atoms, and on cationic Au sites. Jia et al. studied the adsorption and oxidation of CO on a Au/Al2O3 catalyst at 150 K.41 They found evidence for both reversibly and irreversibly adsorbed CO on the 4 nm Au particles. However, introduction of O2 to a catalyst containing only irreversibly adsorbed CO did not produce CO2. Oxidation of CO only occurred at 150 K in the presence of reVersibly adsorbed CO.

J. Phys. Chem. B, Vol. 109, No. 6, 2005 2313 Preoxidation of the catalyst at 573 K in O2 was sufficient to eliminate catalytic activity at 150 K, but activity could be restored by simple evacuation at 573 K or reduction in H2 at 573 K. Those experiments suggest that oxidized Au is not sufficient for CO oxidation activity. Results from the EXAFS characterization in the current work indicate that nearly all of the Au is zerovalent after treatment in He at 623 K. The low coverage of CO on the operating Au/Al2O3 catalyst in this study reveals why the active site has not been elucidated. Indeed, most of the Au surface might have negligible activity in the reaction. Characterization tools used to examine the average properties of a Au catalyst will be overwhelmed by the vast excess of Au that is not participating in the reaction. Selective poisoning of Au/Al2O3 with Cl- showed activity suppression at extremely low levels (Cl-/Au ) 0.0006).44 However, the particles in that study were an order of magnitude larger than those in the current study. Since some researchers speculate that Au atoms at the periphery of metal particles are active in the reaction, we prepared our catalysts with a very high dispersion to maximize the relative number of periphery atoms. Even with our very high dispersion (∼80%), the low surface coverage of COx reactive intermediates indicates very few Au sites are active in the reaction. Although we did not find any evidence to support the idea that cationic gold is needed for CO oxidation at room temperature, we cannot rule out that speculation. We have shown that the surface coverage of reactive COx intermediates is only 4.9% during steady-state reaction. It is also possible that even fewer interfacial active sites contain cationic gold. Those few sites may provide the route to form product CO2 from adsorbed CO. Recall that the isotopic transient experiment simply counts all of the product molecules produced from 12CO after the switch to 13CO in the gas phase. However, it is possible that some adsorbed 12CO2 formed prior to the isotopic switch will also be counted. Addition of water accelerated the rate of oxidation, which is consistent with prior work.17 In principle, the isotopic transient method can discriminate between enhancement of the intrinsic turnover frequency of the catalytic cycle and creation of additional reactive intermediates in the presence of water vapor. Analysis of our preliminary results summarized in Table 3 suggests that water promotes the reaction rate by both mechanisms, which is consistent with the idea that hydroxyl groups at the metal-support interface are important for the reaction. However, the difference in the intrinsic residence times of reactive intermediates between wet and dry conditions is almost too small to measure reliably in our mass spectrometer. Further studies are needed to elucidate the role of water in this system. Conclusions Autoreduction of Al2O3-supported Au by thermal treatment in He at 623 K was demonstrated by X-ray absorption spectroscopy. Analysis of the EXAFS results on a reduced sample indicated a very high dispersion of the gold (∼80%), consistent with 1.2 nm average diameter metal particles. The values of TOFintr and θCOx were found to be 1.6 s-1 and 4.9%, respectively, from isotopic transient analysis of the CO oxidation reaction at 296 K. Those values were determined after removing artifacts in the results attributed to re-adsorption of CO2 in the catalyst bed. Although the apparent activation energy determined from global rate measurements was only 10 kJ mol-1, a negligible change in the residence time of surface intermediates with temperature was observed. However, addition of water vapor (0.18%) to the feed resulted in acceleration of the global

2314 J. Phys. Chem. B, Vol. 109, No. 6, 2005 rate at 296 K by a factor of 6. Results from isotopic transient analysis suggest that water increased both the intrinsic turnover frequency and the surface coverage of reactive intermediates. Acknowledgment. This work was supported by the National Science Foundation (Grant No. CTS-0121619). Research was carried out in part at the National Synchrotron Light Source, Brookhaven National Laboratory, which is supported by the U.S. Department of Energy, Division of Materials Sciences and Division of Chemical Sciences, under Contract No. DE-AC0298CH10886. References and Notes (1) Davis, R. J. Science 2003, 301, 926. (2) Hammer, B.; Norskov, J. K. Nature 1995, 376, 238. (3) Haruta, M.; Yamada, N.; Kobayashi, T.; Iijima, S. J. Catal. 1989, 115, 301. (4) Bond, G. C.; Thompson, D. T. Catal. ReV.-Sci. Eng. 1999, 41, 319. (5) Boccuzzi, F.; Chiorino, A.; Tsubota, S.; Haruta, M. J. Phys. Chem. 1996, 100, 3625. (6) Valden, M.; Lai, X.; Goodman, D. W. Science 1998, 281, 1647. (7) Lopez, N.; Norskov, J. K. J. Am. Chem. Soc. 2002, 124, 11262. (8) Liu, Z.-P.; Hu, P.; Alavi, A. J. Am. Chem. Soc. 2002, 124, 14770. (9) Fu, Q.; Saltsburg, H.; Flytzani-Stephanopoulos, M. Science 2003, 301, 935. (10) Guzman, J.; Gates, B. C. Angew. Chem., Int. Ed. 2003, 42, 690. (11) Oh, H. S.; Costello, C. K.; Cheung, C.; Kung, H. H.; Kung, M. C. Stud. Surf. Sci. Catal. 2001, 139, 375. (12) Costello, C. K.; Kung, M. C.; Oh, H.-S.; Wang, Y.; Kung, H. H. Appl. Catal., A 2002, 232, 159. (13) Guzman, J.; Gates, B. C. J. Phys. Chem. B 2002, 106, 7659. (14) Hodge, N. A.; Kiely, C. J.; Whyman, R.; Siddiqui, M. R. H.; Hutchings, G. J.; Pankhurst, Q. A.; Wagner, F. E.; Rajaram, R. R.; Golunski, S. E. Catal. Today 2002, 72, 133. (15) Margitfalvi, J. L.; Fasi, A.; Hegedus, M.; Lonyi, F.; Gobolos, S.; Bogdanchikova, N. Catal. Today 2002, 72, 157. (16) Kung, H. H.; Kung, M. C.; Costello, C. K. J. Catal. 2003, 216, 425.

Calla and Davis (17) Date, M.; Haruta, M. J. Catal. 2001, 201, 221. (18) Socaciu, L. D.; Hagen, J.; Bernhardt, T. M.; Woste, L.; Heiz, U.; Hakkinen, H.; Landman, U. J. Am. Chem. Soc. 2003, 125, 10437. (19) Sanchez, A.; Abbet, S.; Heiz, U.; Schneider, W.-D.; Hakkinen, H.; Barnett, R. N.; Landman, U. J. Phys. Chem. A 1999, 103, 9573. (20) Hakkinen, H.; Abbet, S.; Sanchez, A.; Heiz, U.; Landman, U. Angew. Chem., Int. Ed. 2003, 42, 1297. (21) Bollinger, M. A.; Vannice, M. A. Appl. Catal., B 1996, 8, 417. (22) Liu, H.; Kozlov, A. I.; Kozlova, A. P.; Shido, T.; Asakura, K.; Iwasawa, Y. J. Catal. 1999, 185, 252. (23) Boccuzzi, F.; Chiorino, A.; Manzoli, M.; Lu, P.; Akita, T.; Ichikawa, S.; Haruta, M. J. Catal. 2001, 202, 256. (24) Meyer, R.; Lemire, C.; Shaikhutdinov, S. K.; Freund, H. J. Gold Bull., in press. (25) Haruta, M. Catal. Today 1997, 36, 153. (26) Okumura, M.; Nakamura, S.; Tsubota, S.; Nakamura, T.; Azuma, M.; Haruta, M. Catal. Lett. 1998, 51, 53. (27) Okumura, M.; Tsubota, S.; Haruta, M. J. Mol. Catal. A: Chem. 2003, 3949, 1. (28) Shannon, S. L.; Goodwin, J. G., Jr. Chem. ReV. 1995, 95, 677. (29) Efstathiou, A. M.; Verykios, X. E. Appl. Catal., A 1997, 151, 109. (30) McClaine, B. C.; Davis, R. J. J. Catal. 2002, 210, 387. (31) McClaine, B. C.; Davis, R. J. J. Catal. 2002, 211, 379. (32) Siporin, S. E.; Davis, R. J.; Rarog-Pilecka, W.; Szmigiel, D.; Kowalczyk, Z. Catal. Lett. 2004, 93, 61. (33) Siporin, S. E.; Davis, R. J. J. Catal. 2004, 222, 315. (34) Lee, P. A.; Citrin, P. H.; Eisenberger, P.; Kincaid, B. M. ReV. Mod. Phys. 1981, 53, 769. (35) Mattheiss, L. F.; Dietz, R. E. Phys. ReV. B 1980, 22, 1663. (36) Greegor, R. B.; Lytle, F. W. J. Catal. 1980, 63, 476. (37) Ali, S. H.; Goodwin, J. G., Jr. J. Catal. 1997, 170, 265. (38) Ali, S. H.; Goodwin, J. G., Jr. J. Catal. 1997, 171, 339. (39) Ali, S. H.; Goodwin, J. G., Jr. J. Catal. 1998, 176, 3. (40) Bordawekar, S. V.; Doskocil, E. J.; Davis, R. J. Langmuir 1998, 14, 1734. (41) Jia, J.; Kondo, J. N.; Domen, K.; Tamaru, K. J. Phys. Chem. B 2001, 105, 3017. (42) Wang, D.; Hao, Z.; Cheng, D.; Shi, X.; Hu, C. J. Mol. Catal. A: Chem. 2003, 3964, 1. (43) Minico, S.; Scire, S.; Crisafulli, C.; Visco, A. M.; Galvagno, S. Catal. Lett. 1997, 47, 273. (44) Oh, H.-S.; Yang, J. H.; Costello, C. K.; Wang, Y. M.; Bare, S. R.; Kung, H. H.; Kung, M. C. J. Catal. 2002, 210, 375.