Genesis of a Cerium Oxide Supported Gold ... - ACS Publications

Feb 4, 2009 - CeO2-supported mononuclear gold species synthesized from Au(CH3)2(acac) catalyzed CO oxidation at 353 K, with a turnover frequency of ...
5 downloads 0 Views 460KB Size
J. Phys. Chem. C 2009, 113, 3259–3269

3259

Genesis of a Cerium Oxide Supported Gold Catalyst for CO Oxidation: Transformation of Mononuclear Gold Complexes into Clusters as Characterized by X-Ray Absorption Spectroscopy Veronica Aguilar-Guerrero, Rodrigo J. Lobo-Lapidus, and Bruce C. Gates* Department of Chemical Engineering and Materials Science, UniVersity of California, DaVis, California 95616 ReceiVed: September 26, 2008; ReVised Manuscript ReceiVed: NoVember 26, 2008

CeO2-supported mononuclear gold species synthesized from Au(CH3)2(acac) catalyzed CO oxidation at 353 K, with a turnover frequency of 6.5 × 10-3 molecules of CO (Au atom s)-1 at CO and O2 partial pressures of 1.0 and 0.5 kPa, respectively. As the catalyst functioned in a flow reactor, the activity increased markedly so that within about 10 h the conversion of CO had increased from about 1% to almost 100%. Activated catalyst samples were characterized by X-ray absorption spectroscopy and found to incorporate clusters of gold, which increased in size, undergoing reduction, with increasing time of operation. The X-ray absorption near-edge structure spectrum of the catalyst used for the longest period was indistinguishable from that characterizing gold foil. Extended X-ray absorption fine structure data characterizing the catalyst after the longest period of operation indicated the presence of clusters of approximately 30 Au atoms each, on average. The evidence that the catalytic activity increased as the clusters grew is contrasted with earlier reports pointing to increasing activity of supported gold clusters as they were made smallersin a cluster size range largely exceeding ours. Introduction Although gold has long been considered to be of little interest as a catalyst, recent work by the groups of Haruta1 and Hutchings,2 who, respectively, reported its use as a lowtemperature catalyst for CO oxidation and for the synthesis of vinyl chloride,2 highlighted the potential of highly dispersed supported gold as a catalyst. Recent discoveries point to potential applications of supported gold catalysts for hydrogenation,3 oxidation,4 and coupling reactions.5 Most supported gold catalysts contain gold clusters with typical diameters of several nanometers;6-8 the most active catalysts appear to be those containing such small clusters, but there are only a few results characterizing supported gold clusters smaller than these. Our goal was to investigate such clusters as catalysts. We have chosen CO oxidation as a test reaction and CeO2 as a catalyst support. CO oxidation is a commonly used test reaction for supported gold catalysts, because there are opportunities for applications of CO oxidation to remove CO impurities from H2 streams for automotive fuel cells.9 Furthermore, CO oxidation offers the advantages of occurring at such low temperatures that spectroscopic methods can be applied readily, and CO is an informative probe molecule for investigation of catalyst surface structures and the oxidation states of gold in them.10 CeO2 has been used widely as a support for gold catalysts, because these catalysts are among the most active reported for reactions of CO, including CO oxidation11 and the water gas shift;9 the high activities of CeO2-supported catalysts may be related to the presence of activated oxygen species formed on the CeO2 surface.12 In a communication13 we have already reported that both mononuclear gold species and gold clusters less than 10 Å in * To whom correspondence should be addressed. E-mail: bcgates@ ucdavis.edu.

average diameter on CeO2 are active for CO oxidation at room temperature and that the gold clusters are much more active than the mononuclear gold species. Now we report physical characterization of these samples showing how the removal of methyl ligands from the initially formed supported gold complexes leads to reduction and aggregation of the gold and activation of the catalyst. Results Activation of the Catalyst. Samples prepared by adsorbing Au(CH3)2(acac) (acac is acetylacetonate) on partially dehydroxylated CeO2 (calcined at 673 K) contained 1.0 wt % Au. As has been communicated,13 this sample at room temperature in a standard once-through plug-flow reactor showed no detectable catalytic activity for CO oxidation. When the temperature was increased, activity was first detected at 333 K, as CO2 was identified in the product stream by mass spectrometry and by gas chromatography. When the temperature was raised further, the activity increased, and the turnover frequency (TOF) initially was found to be (6.5 ( 0.6) × 10-3 molecules of CO (Au atom)-1 s-1 at 353 K with reactant partial pressures (in kPa) of PCO ) 1.0 and PO2 ) 0.5. When the catalyst was kept on stream under these conditions (Figure 1), the activity increased markedly, reaching almost 100% after roughly 10 h on stream; conversions near 100% were maintained for 96 h, until the experiment was stopped. Understanding of the changes in the catalyst as it was activated during operation was a central goal of the research reported here, and so a series of experiments was done to prepare samples for characterization after various periods of operation. IR Evidence of Methyl Ligands on Gold in the Initially Prepared Catalyst and Removal during Activation. A family of catalysts prepared from our precursor, Au(CH3)2(acac), on oxide and zeolite supports has been characterized by IR spectroscopy (Tables 1 and 2) and found to incorporate methyl

10.1021/jp808567a CCC: $40.75  2009 American Chemical Society Published on Web 02/04/2009

3260 J. Phys. Chem. C, Vol. 113, No. 8, 2009

Figure 1. Catalytic activity of gold supported on CeO2 for CO oxidation at 353 K. The reactant partial pressures were PCO ) 1.0, PO2 ) 0.5 kPa. Total flow rate 100 mL/min.

ligands on the gold, which was present as site-isolated mononuclear species.14-16 Bands observed at 2953, 2918, and 2854 cm-1 are indicative of the methyl ligands (Figure 2), and these are evident in the spectra of our sample as well (Figure 2, spectrum c), consistent with the presence of dimethyl gold complexes on the support.17 The catalyst was treated in flowing helium for 2 h at 353 K in an IR cell that also functioned as a flow reactor. IR spectra recorded after this treatment (Figure 3, spectrum b, and Figure 4, spectrum b) show that the sample was stable, as indicated specifically by the lack of change in the peaks indicating methyl groups (2953, 2918, and 2854 cm-1, Figure 4, spectrum b). In contrast, when the sample was brought in contact with a stream of CO + O2 in helium, the spectra changed, providing a characterization of the changes occurring during activation of the catalyst. The changes are illustrated by a comparison of the initial spectrum (Figure 3, spectrum a; Figure 4, spectrum a) with that determined after 4 h in the presence of CO + O2 (Figure 3, spectrum c; Figure 4, spectrum c) (data recorded at other times are shown in Supporting Information). Bands associated with the acac ligands originally present on the gold (at 1595, 1545, 1514, 1454, 1374, and 1260 cm-1)17 decreased in intensity as the CO oxidation reaction took place, and, simultaneously, new bands grew in at 1542, 1428, 1388, and 1292 cm-1. The latter bands have been assigned18 to carbonates and to formates, which in prospect can be distinguished from each other on the basis of the spectra in the C-H stretching region, as such bands characterize formates but not carbonates. During the CO oxidation reaction in the flow system, bands in the C-H stretching region arose at 2945 and 2835 cm-1, demonstrating the formation of formates on the catalyst; the data do not exclude the simultaneous formation of carbonates. When the support alone was similarly treated in CO + O2, the bands indicating formates (and possibly carbonates) also appeared, indicating that the formates (and possibly carbonates) were present on the CeO2 support rather than on the gold in our catalyst. Flow of helium through the IR cell containing the catalyst, following 4 h of operation of the catalyst in CO oxidation, did not lead to any reduction in intensity of the bands associated with the formates (and possibly carbonates). A separate batch of catalyst was used for CO oxidation in a standard once-through plug-flow reactor under the same conditions as those stated above when the IR cell was a flow reactor.

Aguilar-Guerrero et al. The catalyst was removed from the plug-flow reactor after 24 h of operation and characterized by IR spectroscopy. The results (Figure 3, spectrum d, and Figure 4, spectrum d) show that the formates (and possibly carbonates) that formed on the catalyst had been removed after the longer period of operation. Gold Clusters in the Activated Catalyst Shown by Extended X-Ray Absorption Fine Structure (EXAFS) Spectroscopy. EXAFS spectra characterizing the initially prepared sample are characterized by a Au-C contribution, at a distance of 2.00 Å, with an average coordination number of approximately 2 (Table 3).13 This result, consistent with the IR spectra, demonstrates the presence of methyl ligands on the gold in the initially prepared sample. There was no evidence of Au-Au contributions in the EXAFS spectrum (Table 3), in agreement with the earlier results characterizing samples made from Au(CH3)2(acac) on various metal oxide supports and consistent with the inference that the gold was present as mononuclear species in the absence of detectable gold clusters.19,20 A Au-Os contribution was also evident in the EXAFS spectrum, at a distance of 2.05 Å, with an average coordination number of approximately 2 (the subscript s stands for short; Au-Os stands for a contribution in which the Au-O distance is considered to be a bonding distance). This result shows that each gold atom was bonded, on average, to approximately two oxygen atoms, as in the precursor. Such contributions in the spectra of similar samples have been attributed to gold bonded to the support acting as a bidentate ligand.15 Thus, we suggest that our sample similarly incorporated site-isolated mononuclear gold complexes bonded to the support. The presence of the Au-Os and Au-C shells (as well as their coordination numbers and distances) indicates that the gold was cationic and present in structures analogous to the precursor Au(CH3)2(acac). EXAFS characterization of the sample after 2 h in the plugflow reactor under CO oxidation conditions indicates that the gold species remained essentially mononuclear, as no Au-Au contributions were found (Table 3). Furthermore, the presence of Au-Os and Au-C contributions suggests that the gold remained as a cationic four-coordinate complex. However, after the exposure to catalytic reaction conditions, the Au-C distance grew from 2.00 to 2.17 Å (Table 3), indicating a change in the coordination sphere of the gold. This result is in agreement with the IR results showing that methyl groups were not present after the first 4 h of CO oxidation (Figures 3 and 4). In contrast to these results, the EXAFS data characterizing samples that had been working as catalysts for various periods longer than 4 h (24, 36, 48, and 96 h) are characterized by a Au-Au contribution (at approximately 2.80 Å). The average Au-Au coordination number increased with time of operation of the catalyst. This result demonstrates that gold clusters formed as the catalyst underwent activation, and the clusters became larger, on average, as the operation of the catalytic reactor continued. Simultaneous with the changes in the Au-Au shell, other changes were evident in the coordination environment of the gold. The coordination number characterizing the Au-Os contribution at approximately 2.05 Å decreased from approximately 2 initially to a value too small to be determined in the catalyst that had been on stream for 96 h (Table 4); evidently the interactions between the gold and the support underwent significant changes. Reduced Gold in the Activated Catalyst. The X-ray absorption near-edge structure (XANES) spectrum character-

Supported Gold Cluster Catalysts

J. Phys. Chem. C, Vol. 113, No. 8, 2009 3261

TABLE 1: Frequencies of IR Bands Observed in the Fingerprint Region (1200–1700 cm–1) Characterizing Reference Compounds and Sample Formed by Adsorption of Au(CH3)2(acac) on Various Supports IR band positions characterizing the supported species (represented in terms of the precursor/support) or compound, cm-1 Au(CH3)2(acac)/ CeO2a

Au(CH3)2(acac)/ MgOb

Au(CH3)2(acac)/ Al2O3c

1595 1549 1514

1613

1605

1519

1531

1472 1427

1402

1457 1374 1260 1202 a

Au(CH3)2(acac)/ NaY zeolited

Au(CH3)2 (acac)a 1605 1545

1537 1493 1463 1418 1384

1454 1374 1260 1196

1254 1219

assignmente ν(C-C) or ν(C-O) combination band ν(C-O) or ν(C-C) n.a. νs(C-C) + δ(C-H) δd(CH3) δs(CH3) νs(C-C) + νs(C-CH3) νs(C-CH3) + δ(C-H)

This work. b Ref 14. c Ref 15. d Ref 16. e Ref 17.

TABLE 2: Frequencies of IR Bands Observed in the C-H Stretching Region Characterizing Samples Formed by Adsorption of Au(CH3)2(acac) on Various Supports IR band positions characterizing the supported species (represented in terms of the precursor/support) or compound, cm-1

a

Au(CH3)2(acac)/ CeO2a

Au(CH3)2(acac)/ MgOb

2991 2967 2960 2923 2914 2855 2807

3080 3038 3005 2992 2967 2953 2928 2910 2854 2822

Au(CH3)2(acac)/ NaY zeolitec

Au(CH3)2(acac)a

assignmentd

2960

2962

2930

2916

2868

2849

ν(C3-H) n.a. n.a ν(CH3) (acac) ν(CH3) (acac) νas(CH3) (Au methyl) ν(CH3) (acac) νas(CH3) (Au methyl) combination band combination band

This work. b Ref 14. c Ref 16. d Ref 17.

Figure 2. Normalized IR spectra characterizing the C-H stretching region (2800-3100 cm-1) of (a) CeO2, (b) Au(CH3)2(acac), and (c) Au(CH3)2(acac) supported on CeO2 (scales were adjusted for clarity of comparison of spectra; quantitative comparisons are not appropriate).

izing the initially prepared sample (Figure 5, spectrum a) includes features similar to those representing Au(CH3)2(acac) (Figure 6, spectrum a), indicating that the gold was present on the support as cationic species, approximated as Au(III). The XANES spectra also show that the cationic gold initially present on the support was converted as the catalyst was used for CO oxidation (Figure 5). A comparison of the XANES spectra of the catalyst samples that had been in the reactor for various periods (Figure 5) with those of the reference compounds Au(CH3)2(acac) and gold foil (Figure 6, spectra a and c, respectively) demonstrate that the gold was reduced during operation, from species that were initially approximated as

Figure 3. IR spectra in the 1300–1700 cm–1 region characterizing gold supported on CeO2: (a) sample as initially prepared, (b) sample in flowing helium at 353 K at atmospheric pressure, (c) after 4 h of CO oxidation catalysis at 353 K, and (d) after 24 h undergoing CO oxidation catalysis (scales were adjusted for clarity of comparison of spectra; quantitative comparisons are not appropriate).

Au(III) to finally, after 96 h, species that are well approximated as metallic gold (Figure 6). Even the sample that had been used for only 48 h was characterized by a spectrum essentially matching that of metallic gold (data not shown). The isosbestic points in the XANES spectra of Figure 5 suggest that the cationic gold species present initially were transformed into gold clusters that were nearly equivalent to each other. The normalized intensity of the white line (the signal at 11 923 eV) (the normalization is explained below in the section about analysis of X-ray absorption spectra) characterizing the catalyst samples that had been used for various periods is shown as a function of time on stream in Figure 7. The decrease in

3262 J. Phys. Chem. C, Vol. 113, No. 8, 2009

Figure 4. IR spectra in the C-H stretching region characterizing gold supported on CeO2: (a) sample as initially prepared, (b) sample in flowing helium at 353 K at atmospheric pressure, (c) after 4 h of CO oxidation catalysis at 353 K, and (d) after 24 h of CO oxidation catalysis at 353 K (scales were adjusted for clarity of comparison of spectra; quantitative comparisons are not appropriate).

this intensity indicates that the gold underwent reductions simultaneous with the aggregation demonstrated by the EXAFS data. The value of the normalized signal intensity decreased from 1.05 in the initially prepared sample to 0.75 in the sample after 96 h of CO oxidation catalysis. These values essentially span the range observed for Au(CH3)2(acac) and gold foil, namely, 1.05 and 0.73 (Figure 7). Summary. The site-isolated mononuclear gold species present initially were stable in the presence of helium at 353 K. Mononuclear gold species formed from these were present during the first 2 h of CO oxidation catalysis (when the CO conversion was about 1.3%), but as the catalyst continued to operate, its activity increased so much that the conversion became almost 100%. Simultaneously, the gold was reduced and gold clusters formed. The activation of the catalyst is attributed to the cluster formation. Formates and possibly carbonates formed on the CeO2 surface during the first 4 h of CO oxidation catalysis, but these were removed in subsequent operation of the catalyst as it was activated. Discussion The data show clearly that during the period of rapid activation of the catalyst, the gold underwent aggregation and reduction. The data provide further evidence of the overall activation process; changes other than aggregation and reduction of the gold took place, both during the initial period of CO oxidation (when the conversion of CO was approximately 2%) and during the period in which the activity increased sharply. The IR and EXAFS data show that the aggregation and reduction of the gold were always accompanied by changes in the ligation of the gold and/or changes in surface species other than the gold. During the initial treatment of the sample in flowing CO + O2 at 353 K, the IR bands associated with the acac ligands changed; after 4 h, most of these bands had been removed or shifted markedly. Furthermore, during this period the IR bands characteristic of the methyl ligands initially bonded to the gold (inferred to be responsible for the Au-C contribution in the EXAFS spectrum) disappeared, and thus we infer that other ligands replaced the methyl groups on the gold. As the data show, changes also occurred on the CeO2 support, with the formation of formates (and possibly carbonates). After 24 h under CO oxidation conditions (when the catalyst had reached

Aguilar-Guerrero et al. its maximum activity), all the IR bands characterizing the absorbed species mentioned in this paragraph had disappeared. Moreover, when CeO2 in the absence of gold was treated similarly, the IR spectra gave evidence of formates (and possibly carbonates), consistent with the suggestion that these species were still being formed on the supported gold catalyst even after 24 h under reaction conditions, but that they were being consumedslikely, we infer, with gold playing a catalytic role. A mechanism of the catalytic CO oxidation reaction catalyzed by gold on γ-Al2O3 has been proposed by Costello et al.21 to involve carbonates as intermediates. On the basis of IR evidence, these researchers proposed that CO interacts with the hydroxyl groups on the support surface and forms COOH species which then are converted into CO2. Such a mechanism, involving the support and not just the gold, is consistent with our results. In contrast, van Bokhoven et al.22 proposed that the O2 reacting with CO in CO oxidation catalyzed by gold on γ-Al2O3 is activated on the gold itself, because they observed XANES spectra indicating oxidation of the gold during treatment in O2 and subsequent reduction of the gold during treatment in CO. Both Costello and van Bokhoven proposed that the gold in their catalyst samples was present as small clusters (in the range of 1-5 nm in average diameter). Gold catalysts supported on CeO2 have been investigated by numerous researchers.9,11 For example, Guzman et al.11 measured Raman spectra of their catalyst during CO oxidation catalysis and identified η1-superoxide species on the CeO2, and their EXAFS data, giving evidence of a long (nonbonding) Au-O contribution, led them to infer the involvement of the superoxide species as intermediates in the catalysis. This inference is contrasted with that of van Bokhoven et al.22 and is consistent with various roles of the supports used for gold catalysts and with the inference that there is no single mechanism of CO oxidation catalyzed by supported gold.23 Our data provide detailed insights into the changes that accompanied the activation of the supported gold catalyst. For example, after 24 h of operation, the Au-Au coordination number had increased from an undetectably small value to 4.9, which corresponds to the transformation of the mononuclear precursor into clusters with an average of roughly 12 Au atoms each (the cluster nuclearity was estimated from the Au-Au coordination number according to the method of Jentys24). The EXAFS spectra characterizing the samples that had been used for longer periods, 36, 48, and 96 h, indicate that the clusters continued to increase in size, with the corresponding nuclearities being roughly 15, 19, and 32 Au atoms each, respectively, on average (determined on the basis of the Jentys method), all of which correspond to cluster diameters less than 10 Å (Figure 8).24 As the clusters grew, the Au-Os coordination number also changed (Figure 9). The Au-Os coordination number representing the initially prepared sample (and also that representing the sample that had been used for only 2 h) was 2.1. Over time, this value decreased; after 24 h it was 0.8; later it decreased to 0.3, 0.2, and then a value indistinguishable from 0 (Figure 9). This result is consistent with the growth of the clusters, whereby a decreasing fraction of the Au atoms was bonded to the support surface and coordinated with support oxygen atoms as the number of Au-Au bonds increased. Similar results characterizing CeO2-supported gold catalysts (prepared by the a gel coprecipitation method) were reported by Deng et al.,9 whose catalyst (before use) was characterized by EXAFS-determined Au-Os coordination numbers of about 2.7, which decreased after the catalyst was used for the water gas shift reaction.9 All

Supported Gold Cluster Catalysts

J. Phys. Chem. C, Vol. 113, No. 8, 2009 3263

TABLE 3: EXAFS Parameters Characterizing Gold Supported on CeO2a,b sample initially prepared sample sample after 2 h of CO oxidation catalysis in flow reactor at 353 K (PCO ) 1.0; PO2 ) 0.5 kPa)

shell

N

R (Å)

103 × ∆σ2 (Å2)

∆E0 (eV)

Au-Os

2.1(0.0)

2.06(0.00)

5.49(0.3)

-5.9(0.1)

Au-C Au-Os

2.2(0.1) 2.1(0.0)

2.00(0.00) 2.01(0.00)

9.39(0.4) 4.1(0.5)

-7.0(0.1) -11.8(0.1)

Au-C

2.2(0.1)

2.17(0.01)

12.9(0.6)

-16.4(0.3)

Ref 13. Notation: N ) coordination number; R ) distance between absorber and backscatterer atoms; ∆σ ) Debye-Waller factor relative to reference; ∆E0 ) inner potential correction. Numbers in parentheses are the calculated errors and represent precisions, not accuracies. Estimated accuracies are as follows: Au-Au, N (20%, R (0.02 Å, ∆σ2 (20%, ∆E0 (20%; Au-O, N (10%, R (0.02 Å, ∆σ2 (20%, ∆E0 (20%; Au-C, N (10%, R (0.03 Å, ∆σ2 (20%, ∆E0 (20%. a

b

2

TABLE 4: EXAFS Parameters Characterizing Gold Supported on CeO2 after Various Times in a Flow Reactor Working as a CO Oxidation Catalyst at 353 K (PCO ) 1.0; PO2 ) 0.5 kPa)a EXAFS parameters determined in fitting time of operation as CO oxidation catalyst (h) 24 36 48 96

shellb

N

R (Å)

103 × ∆σ2 (Å2)

∆E0 (eV)

(∆χ)2 upon addition of shellc

k-range (Å-1)/error in datad

Au-Au Au-Os Au-Au Au-Os Au-Ol Au-Au Au-Os Au-Au

4.9(0.0) 0.8(0.0) 5.5(0.2) 0.3(0.0) 1.1(0.1) 6.0(0.0) 0.2(0.0) 7.0(0.0)

2.81(0.00) 1.99(0.00) 2.78(0.01) 2.12(0.02) 3.81(0.03) 2.80(0.01) 2.09(0.00) 2.80(0.00)

13.0(0.6) 5.1(0.0) 11.8(0.6) -0.9(1.8) -0.5(2.5) 9.7(2.6) -0.7(0.0) 12.6(1.8)

-5.5(0.1) -3.3(0.0) -1.5(0.4) -1.5(0.4) -9.5(1.2) -1.4(0.7) -7.2(0.0) -1.9(0.9)

15.0 2.6 7.6 4.9 4.3 8.6 7.5 16.4

2.6-12.21/0.0015 2.58-11.85/0.0012 2.57-11.78/0.0014 2.59-11.91/0.0007

a Refer to Table 3 for notation. b Shells are shown in the order they were added to the model (e.g., for the sample after 24 h of operation as a CO oxidation catalyst, the Au-Au contribution was fitted first, and then, to improve the fit, the Au-Os contribution was added). c Values of (∆χ)2 representing a model that includes the shell and any other shells above it in the table. d Error associated with the data and calculated by Fourier filtering the data (according to the methods described by the XAFS society (ref 42); see the XAS Data Analysis section for more details of the calculation).

Figure 5. Normalized XANES spectra of (a) initially prepared sample and samples after various periods of CO oxidation catalysis (h): (b) 24, (c) 36, and (d) 96. The spectrum characterizing the sample after 48 h of operation is not included in this plot because it essentially matches that of the sample used for 96 h.

these results show that gold-support interactions (characterized by a Au-O contribution at a bonding distance) were characteristic of all the samples. These bonding interactions may have helped to stabilize the high dispersions of the gold.15 We would expect thesesand essentially all of the catalyst structural parameterssto be correlated with the catalytic activity during the activation process. The summary plot of Figure 9 confirms the expectation, including EXAFS, XANES, and CO conversion data. This plot summarizes the key results of this

Figure 6. Normalized XANES spectra of reference compounds used for comparison with XANES spectra of the catalyst samples: (a) solid line, Au(CH3)2(acac); (b) dashed line, AuCl; (c) dotted line, gold foil.

work. It shows that the increasing Au-Au coordination numbers are inversely correlated with the white line intensity determined by the XANES data, showing that the reduction and aggregation of the gold occurred essentially simultaneously as the catalyst underwent activation and the CO conversion increased (Figures 9 and 10). Although the EXAFS and XANES results show that the majority of gold was aggregated and reduced, the data do not exclude the possible presence of cationic gold (e.g., at the gold-support interface) but undetected by these techniques.25 The largest gold clusters, characterized by a Au-Au coordination number of 7.0 and incorporating approximately 32 Au

3264 J. Phys. Chem. C, Vol. 113, No. 8, 2009

Figure 7. Normalized XANES signal intensity at 11 923 eV characterizing samples treated under CO oxidation conditions for various periods. Lines are also plotted to show the intensity of the XANES spectra of the precursor Au(CH3)2(acac) in the form of powder mixed with boron nitride powder and of gold foil.

Figure 8. Increase in cluster size represented by the approximate number of Au atoms per cluster, on average, during catalysis of CO oxidation in a flow reactor at 353 K. The calculations were made according to the Jentys approximation (ref 24).

atoms per cluster, on average, were observed for the catalyst after 96 h of operation. These clusters are less than about 10 Å in average diameter (and with the errors in the Au-Au coordination number taken into account, the estimated average cluster diameter would at most be about 15 Å, Table 4). Thus, the family of samples described here includes some of the smallest gold clusters that have yet been characterized in catalysts. And these catalysts are among the most active gold catalysts yet reported.23,26 Many authors have attributed the catalytic activity of supported gold to clusters, inferring that the cluster size is crucial.27 The gold clusters in the most active CO conversion catalysts have been proposed by several authors to be less than 20 Å in average diameter, as summarized in a recent review by Haruta.28 Specifically, Haruta29 focused on the influence of the gold cluster size, drawing the inferences, based on the literature as a whole, that the critical diameter for the genesis of catalytic activity for CO oxidation is about 100 Å and that the catalytic properties change most strongly as the cluster diameter is decreased below about 20 Å.29 Our data show that the activity increases, at least when the support is CeO2, as the average cluster diameter increases up to about 10 Å (Figure 8). Thus, the results point to an optimum

Aguilar-Guerrero et al. cluster size, which remains to be determinedsand is expected to be dependent on the support.27 We stress that the changes in the cluster size that are correlated with the activation of the catalyst were accompanied by changes in the ligation of the gold, as summarized abovesand that the catalyst activation may reflect changes in the gold ligation as well as the observed changes in gold cluster size. Furthermore, results presented separately for our catalyst13,23 show that the supported mononuclear gold complexes could be converted into active species by removal of the methyl groups and their replacement with other (unidentified) groups, presumably derived from the reactants CO + O2sand that these changes did not immediately lead to aggregation of the gold. Thus, an important conclusion of that work is that mononuclear gold complexes on CeO2 are catalytically active for CO oxidation; a further conclusion of that work is that the catalyst containing gold clusters was much more active than that containing only mononuclear gold complexes.13,23 In attempts to understand supported gold catalysts better, some researchers have investigated thin islands of gold on supports such as TiO2.29,30 Chen and Goodman30 proposed that extended bilayer islands of gold on ultrathin titanium oxide films grown on Mo(112) surfaces are active for CO oxidation (the methods of determining catalytic reaction rates and their extrapolation to initial conditions to establish the reported catalytic activity were not explained in detail, and the images provided30 do not determine the dimensions of the bilayer islands that lie parallel to the Mo(112) surface, or how their numbers compare with those of other structures shown in the images, or how either of these depends on the coverage of the surface with gold). The bilayers were inferred30 to be an order of magnitude more active (in terms of rate expressed as turnover frequency, CO molecules converted per accessible Au atom per unit time) than catalysts incorporating gold monolayers on the same support (and also more active than conventionally prepared oxide-supported gold catalysts).30 In an investigation of catalysts made from gold supported on high-area FeOOH, Herzing at al.31 treated their samples at different temperatures, resulting in catalysts of widely varying activity for CO oxidation. Herzing at al.31 recorded images of the gold species with aberration-corrected high-angle annular dark-field scanning transmission electron microscopy (HAADFSTEM) for comparison with the measured catalytic activities. They interpreted their results as evidence in support of the conclusions of Chen and Goodman;30 using the HAADF-STEM data, they counted the frequency of appearance of gold clusters of various sizes from 2 to 10 Å (and also larger clusters) in the different samples and sought correlations with the catalyst performance. They interpreted the data as evidence that the catalytic activity was correlated with the presence of isolated bilayer clusters that were approximately 5 Å in diameter and contained approximately 10 Au atoms each (the cluster thickness of two Au atoms was estimated from the contrast level of the HAADF-STEM images). These bilayer clusters were present in the more active of their catalysts and not in the less active ones. Herzing et al.31 reported nominal turnover frequencies calculated not only on the basis of all the gold atoms (exposed or not), but also turnover frequencies calculated on the basis of the assumption that only the gold in the bilayer clusters was active for CO oxidation. Finding approximate agreement between the latter and the turnover frequency reported by Chen and Goodman,30 Herzing et al.31 drew the conclusion that the activity of their discrete, isolated bilayer clusters on iron oxide

Supported Gold Cluster Catalysts

J. Phys. Chem. C, Vol. 113, No. 8, 2009 3265

Figure 9. Changes in parameters determined by XAS and conversion in CO oxidation during operation of catalyst in the flow reactor at 353 K.

during CO oxidation in a flow reactor. The clusters are much more active than the mononuclear gold species. The data characterizing the family of catalysts consisting of clusters of various average sizes show that the catalytic activity and the degree of reduction of the gold both increased with increasing cluster size. These results are the first detailed evidence of the structural changes involved in the activation of supported gold catalysts by cluster formation. Experimental Section

Figure 10. Data characterizing samples used for CO oxidation catalysis for various times in the plug-flow reactor: correlation of Au-Au and Au-Os coordination numbers obtained from EXAFS spectroscopy with normalized white line intensity obtained from XANES spectra. The time-on-stream values are shown next to data points.

was consistent with the activity of the extended bilayers reported by Chen and Goodman30 on titanium oxide. This comparison, in our judgment, would benefit from further scrutiny, as discrete, isolated bilayer clusters (on one support) would not be expected to act like extended bilayer clusters (on the same or on another support). As an alternative to the interpretation presented by Herzing et al.,31 we suggest that the differences in activity of their samples treated at different temperatures might be influenced in part by the varying degrees of hydroxylation of the support. Several authors32,33 proposed that support hydroxyl groups are involved in the mechanism of CO oxidation, thus affecting the catalytic activity. Herzing et al.31 recognized the influence of the treatment temperature on the catalytic activity, inferring that the degree of hydroxylation influences the aggregation of the gold, but their interpretation turned entirely on the gold cluster size and not on a direct role of hydroxyl groups in the catalysis. Conclusions The data presented here provide evidence of the activity of both mononuclear gold complexes and gold clusters (up to approximately 10 Å in diameter, on average) formed from them

Synthesis of Catalysts. Catalyst samples were synthesized and handled with the exclusion of moisture and air by use of a glovebox purged with argon that was recirculated through traps containing supported copper and zeolite 4A for removal of O2 and moisture, respectively. Samples, containing 1 wt % Au, were synthesized from Au(CH3)2(acac) (Strem 99.9%) and highsurface-area CeO2 (Daiichi, 99.9%, 173 m2/g and average particle size 46 nm; information provided by the supplier). The catalyst was prepared by slurrying Au(CH3)2(acac) in n-pentane with CeO2 powder that had been partially dehydroxylated under vacuum at 673 K. The slurry was stirred for 24 h, and then the solvent was removed by evacuation for 24 h. The resultant sample, referred to as the initially prepared sample, was then stored in the glovebox. Catalyst Testing. CO oxidation catalysis was carried out with a reactor well approximated as a plug-flow reactor; it was a quartz tube with an internal diameter of 0.8 cm fitted with a quartz frit to hold a mixture of catalyst powder and inert, nonporous, R-Al2O3 powder in place. In each experiment, 25 mg of catalyst was mixed with 2 g of R-Al2O3 and placed between two beds of particles of R-Al2O3 to give a bed approximately 5 cm deep. The gases used in the experiments were obtained from Airgas and passed through a zeolite 4A trap to remove traces of moisture; furthermore, the CO and helium were passed through a supported copper trap to remove traces of O2. Prior to testing, the catalyst was treated in flowing helium at 353 ( 0.5 K for 2 h; the total gas flow rate was 100 mL (NPT)/ min. After this treatment, the flow of CO and O2 was started, the feed partial pressures of CO and O2 being 1.0 and 0.5 kPa, respectively, with the balance being helium; the pressure was atmospheric, and the total flow rate was 100 mL(NPT)/min

3266 J. Phys. Chem. C, Vol. 113, No. 8, 2009 (these conditions are referred to as “standard conditions” throughout the text). The catalyst (in separate experiments) was kept on stream for various periods (2, 24, 36, 48, and 96 h). Samples of the effluent gas were sampled periodically and analyzed with a HP-5890 gas chromatograph equipped with a thermal conductivity detector. Turnover frequency values were calculated by using steadystate conversions and the total gold content in the catalyst, with the assumption that each Au atom was accessible. IR Spectroscopy. Transmission IR spectra of the catalyst were recorded with a Bruker IFS 66V/S spectrometer. Spectra of the catalyst under reaction conditions were recorded with samples of catalyst pressed into self-supporting wafers. Each wafer was placed in an environmentally controlled IR cell equipped with CaF windows, which also functioned as a flow reactor (In-Situ Research Instruments). Measurements of spectra in the mid-IR region were carried out with a triglycine sulfate (TGS) detector. Each reported spectrum is the average of 64 scans, obtained with a spectral resolution of 2 cm-1. Reported spectra were baseline-corrected by using the software OPUS 5.5. The initially prepared sample was scanned to provide information about the initial structure, and then it was scanned repeatedly during CO oxidation catalysis at 353 K for the first 4 h of reaction under standard conditions; scans were recorded every 2 min. Samples of catalyst that had been used in the plug-flow reactor for various periods were removed from the reactor in the glovebox (in which the moisture and O2 levels were monitored and found to be 10 Å. This error was then averaged and used for calculation of the goodness of fit and (∆χ)2 (Table 4). The candidate models were all the plausible combinations that included Au-Au, Au-Os, Au-Ol, Au-C, and Au-Ce contributions (Au-Ol stands for a contribution in which the Au-O distance is too long to be considered a bonding distance). Estimated accuracies in the EXAFS parameters are as follows: Au–low-Z scatterers: error in N, (20%; error in R, (0.02 Å; Au–high-Z scatterers: error in N, (10%; error in R, (0.02 Å;35 precisions, determined with the fitting software, are shown in Table 4. Model Discrimination and Selection. EXAFS spectroscopy was used to characterize the CeO2-supported gold samples after various periods of CO oxidation catalysis (24, 36, 48, and 96 h) in the plug-flow reactor. Various models were tested in the data fitting. The results including only the best model for each sample are summarized in Table 4. Representative results showing the data and the recommended fit for the sample that had been used for CO oxidation for 24 h are presented in Figure 11; the corresponding plots for the other samples, including the various models considered in the discrimination process, are available in Supporting Information. Several candidate models were considered in the fitting of the EXAFS data characterizing each sample. The simplest acceptable model characterizing the sample used for CO oxidation catalysis for 24 h (Supporting Information) included only two shells, Au-Au and Au-Os. The Au-Au shell is characterized by a coordination number of 4.9 at a distance of 2.81 Å. The Au-Os contribution is characterized by a coordination number of 0.8 at a distance of 1.99 Å. Two possibilities for a third shell resulted in physically appropriate values of the fit parameters, namely, a second Au-Au shell at a distance of 4.9 Å with a coordination number of 3.3 and a Au-Ce contribution with a coordination number of 1.2 at a distance of 3.97 Å. However, because the value of (∆χ)2 increased upon addition of either of the contributions, we infer that the addition of neither improved the overall fit, and both were rejected. Nor did inclusion of a second Au-Au shell in place of the Au-Ol shell improve the fit, as it led to unacceptable ∆E0 values (>15 eV) (a similar statement pertains to all the other samples described below). In summary, the EXAFS data demonstrate that the sample that had been used for CO oxidation for 24 h contained gold clusters (characterized by a Au-Au contribution), and the gold

3268 J. Phys. Chem. C, Vol. 113, No. 8, 2009

Aguilar-Guerrero et al.

Figure 11. Comparison of data characterizing the catalyst after 24 h of CO oxidation at 353 K in the plug-flow reactor; the sample was scanned at 298 K with fits corresponding to the recommended model (Table 4). (A) EXAFS data (solid line) and fit (dashed line) in k-space with k2weighting; (B and C) uncorrected R-space plots of the data (solid lines) and fit (dashed lines) with k1- and k3-weighting, respectively; (D) difference file showing the Au-Os contribution, determined by subtraction of the Au-Au contribution from the data (uncorrected); (E) difference file showing the Au-Au contribution, determined by subtraction of the Au-Os contribution from the data (uncorrected); (F) difference file showing the Au-Au contribution, determined by subtraction of the Au-Os contribution from the data (phase- and amplitude-corrected (ref 36)).

was bonded to the support (the Au-O distance is a bonding distance15). Fit parameters and plots for each of the three models are shown in Supporting Information. A clear result of the fitting is that the Au-Au and Au-Os contributions are necessary in the representation of the data. EXAFS data characterizing the sample that had been on stream for 36 h were fitted similarly; all the acceptable fits included two shells, Au-Au and Au-Os. Attempts to add a third shell were made as well. The fit parameters corresponding to the two separate three-shell models and the model including only the Au-Au and Au-Os shells are reported in Supporting Information. Each of the two three-shell models includes a Au-Au shell at a distance of 2.78 Å with a coordination number of approximately 5.5 (Supporting Information) and a Au-Os shell with coordination number of 0.3 at a distance of 2.12 Å.

The addition of the latter shell decreased the value of (∆χ)2, indicating an improved overall fit. Alternatively, inclusion of a Au-Ol contribution as a third shell also led to a decrease in the value of (∆χ)2. On the other hand, addition of a Au-Ce contribution as a third shell (the distance was found to be 3.82 Å) increased the value of (∆χ)2, indicating that addition of such a contribution did not improve the overall fit. In summary, the recommended model includes Au-Au, Au-Os, and Au-Ol contributions (Table 4). The sample that had been on stream for 48 h is characterized by a Au-Au contribution at a distance of 2.80 Å and coordination number of approximately 6. When a second shell was added (Au-Os), the value of (∆χ)2 decreased substantially, suggesting that this contribution was appropriate.

Supported Gold Cluster Catalysts Models including Au-Au and a Au-Os contributions were then used as a basis from which attempts were made to include further shells: fitting of models containing Au-Ce or Au-Ol contributions at a distance of approximately 4.10 Å was attempted, but these additions did not improve the overall fit (as shown by values of (∆χ)2, Supporting Information). The two-shell model that includes a Au-Au and a Au-Os shell is therefore recommended. All the models considered in the fitting procedure along with the complete list of EXAFS parameters are shown in Supporting Information. The data characterizing the sample after 96 h on stream were represented satisfactorily with a one-shell model, with the contribution being Au-Au with a coordination number of 7.0 (Table 4). Addition of a second shell did not improve the fit. For example, addition of a Au-Os shell (at a distance of 2.00 Å) worsened the overall fit as shown by an increase in the (∆χ)2 value. Nor did addition of a third shell improve the fit. In summary, the clusters in this sample were large enough that no shell in addition to the single Au-Au shell was needed in the fit. The fit parameters found in the various candidate models are included in Supporting Information. A summary of the bases for selection of the recommended models is shown in Table 5. Acknowledgment. We thank Professor Uzi Landman for helpful discussions. This research was supported by CONACyT/ UC MEXUS (175763 V.A-G.), and by the U.S. Department of Energy, Office of Energy Research, Office of Basic Energy Sciences (FG02-04ER15513 (V.A-G.)) and Contract FG0204ER1600 (R.J.L-L.)). We acknowledge the Stanford Synchrotron Radiation Laboratory (SSRL), a national user facility operated by Stanford University on behalf of the U.S. Department of Energy, Office of Basic Energy Sciences. The SSRL Structural Molecular Biology Program is supported by the Department of Energy, Office of Biological and Environmental Research, and by the National Institutes of Health, National Center for Research Resources, Biomedical Technology Program. We also acknowledge the National Synchrotron Light Source (NSLS), Brookhaven National Laboratory, beam line X-18B, for access to beam time; the NSLS is supported by the U.S. Department of Energy, Office of Science, Basic Energy Sciences, under Contract No. DE-AC02-98CH10886; beam line X18B is supported by the NSLS, through the Divisions of Materials and Chemical Sciences of the DOE, and the Synchrotron Catalysis Consortium (U.S. DOE Grant No. DE-FG0205ER15688). We thank the beam line staffs for their assistance. Supporting Information Available: Tables containing all fit parameters for each of the EXAFS models that were rejected of each sample, along with the corresponding plots, and IR spectra characterizing the sample that was under the standard CO oxidation conditions for various periods. This material is available free of charge via the Internet at http://pubs.acs.org. References and Notes (1) Haruta, M.; Kobayashi, T.; Sano, H.; Yamada, N. Chem. Lett. 1987, 405. (2) Hutchings, G. J.; Joffe, R. Appl. Catal. 1986, 20, 215.

J. Phys. Chem. C, Vol. 113, No. 8, 2009 3269 (3) Hutchings, G. J. Specialty Chemicals Magazine 2003, 23, 12. (4) Corti, C. W.; Holliday, R. J.; Thompson, D. T. Appl. Catal., A 2005, 291, 253. (5) (a) Gonzalez-Arellano, C.; Abad, A.; Corma, A.; Garcia, H.; Iglesias, M.; Sanchez, F. Angew. Chem., Int. Ed. 2007, 46, 1536. (b) Gonzalez-Arellano, C.; Corma, A.; Iglesias, M.; Sanchez, F. J. Catal. 2006, 238, 497. (6) Haruta, M. J. New Mater. Electrochem. Syst. 2004, 7, 163. (7) Pillai, U. R.; Deevi, S. Appl. Catal., A 2006, 299, 266. (8) Widmann, D.; Leppelt, R.; Behm, R. J. J. Catal. 2007, 251, 437. (9) Deng, W.; Frenkel, A. I.; Si, R.; Flytzani-Stephanopoulos, M. J. Phys. Chem. C 2008, 112, 12834. (10) Mihaylov, M.; Kno¨zinger, H.; Hadjiivanov, K.; Gates, B. C. Chem.-Ing.-Tech. 2007, 79, 795. (11) Guzman, J.; Carrettin, S.; Fierro-Gonzalez, J. C.; Hao, Y.; Gates, B. C.; Corma, A. Angew. Chem., Int. Ed. 2005, 44, 4778. (12) Pushkarev, V. V.; Kovalchuk, V. I.; d’Itri, J. L. J. Phys. Chem. B 2004, 108, 5341. (13) Aguilar-Guerrero, V.; Gates, B. C. Chem. Commun. 2007, 3210. (14) Guzman, J.; Anderson, B. G.; Vinod, C. P.; Ramesh, K.; Niemantsverdriet, J. W.; Gates, B. C. Langmuir 2005, 21, 3675. (15) Guzman, J.; Gates, B. C. Langmuir 2003, 19, 3897. (16) Fierro-Gonzalez, J. C.; Anderson, B. G.; Ramesh, K.; Vinod, C. P.; Niemantsverdriet, J. W.; Gates, B. C. Catal. Lett. 2005, 101, 265. (17) Miles, M. G.; Glass, G. E.; Tobias, R. S. J. Am. Chem. Soc. 1966, 88, 5738. (18) Jacobs, G.; Patterson, P. M.; Williams, L.; Chenu, E.; Sparks, D.; Thomas, G.; Davis, B. H. Appl. Catal., A 2004, 262, 177. (19) Fierro-Gonzalez, J. C.; Bhirud, V. A.; Gates, B. C. Chem. Commun. 2005, 5275. (20) Guzman, J.; Gates, B. C. Angew. Chem., Int. Ed. 2003, 42, 690. (21) Costello, C. K.; Kung, M. C.; Oh, H.-S.; Wang, Y.; Kung, H. H. Appl. Catal., A 2002, 232, 159. (22) van Bokhoven, J. A.; Louise, C.; Miller, J. T.; Tromp, M.; Safonova, O. V.; Glatzel, P. Angew. Chem., Int. Ed. 2006, 45, 4651. (23) Aguilar-Guerrero, V.; Gates, B. C. J. Catal. 2008, 260, 351. (24) Jentys, A. Phys. Chem. Chem. Phys. 1999, 1, 4059. (25) Fierro-Gonzalez, J. C.; Gates, B. C. Chem. Soc. ReV. 2008, 37, 2127. (26) Carrettin, S.; Concepción, P.; Corma, A.; López Nieto, J. M.; Puntes, V. F. Agnew. Chem., Int. Ed. 2004, 43, 2538. (27) Haruta, M.; Yamada, N.; Kobayashi, T.; Ijima, S. J. Catal. 1989, 115, 301. (28) Haruta, M. Relevance of Metal Nanocluster Size Control in Gold (0) Catalytic Chemistry. In Metal Nanoclusters in Catalysis and Materials Science: The Issue of Size Control; Corain, B., Schmid, G., Toshima, N., Eds.; Elsevier B. V.: Amsterdam, The Netherlands, 2008; pp 183-199. (29) Bondzie, V. A.; Parker, S. C.; Campbell, C. T. Catal. Lett. 1999, 63, 143. (30) Chen, M. S.; Goodman, D. W. Science 2004, 306, 252. (31) Herzing, A. A.; Kiely, C. J.; Carley, A. F.; Landon, P.; Hutchings, G. Science 2008, 321, 1331. (32) Date´, M.; Okumura, M.; Tsubota, S.; Haruta, M. Angew. Chem., Int. Ed. 2004, 24, 2129. (33) Costello, C. K.; Guzman, J.; Yang, J. H.; Wang, Y. M.; Kung, M. C.; Gates, B. C.; Kung, H. H. J. Phys. Chem. B 2004, 108, 12529. (34) Ravel, B.; Newville, M. J. Synchrotron Radiat. 2005, 12, 537. (35) Koningsberger, D. C.; Mojet, B. L.; Dorsen, G. E.; Ramaker, D. E. Top. Catal. 2000, 10, 143. (36) Vaarkamp, M.; Linders, J. C.; Koningsberger, D. C. Physica B 1995, 209, 159. (37) Stern, E. A. Phys. ReV. B 1993, 48, 9825. (38) Rehr, J. J.; Albers, R. C. ReV. Mod. Phys. 2000, 621. (39) Benfield, R. E.; Grandjean, D.; Kro¨ll, M.; Pugin, R.; Sawitowski, T.; Schmid, G. J. Phys. Chem. B 2001, 105, 1961. (40) Guzman, J.; Gates, B. C. J. Phys. Chem. B 2002, 106, 7659. (41) Villars, P.; Calvert, L. D. Pearson’s Handbook of Crystallographic Data for Intermetallic Phases; American Society for Metals: OH, 1985; Vol. 2, p 1196. (42) International XAFS Society. Error Reporting Recommendations: A Report of the Standards and Criteria Committee. http://www.i-x-s.org/ OLD/subcommittee_reports/sc/err-rep.pdf (accessed on August 2008).

JP808567A