Report

Spectroscopic Observation of Dual Catalytic Sites During Oxidation of CO on a Au/TiO2 Catalyst

See allHide authors and affiliations

Science  05 Aug 2011:
Vol. 333, Issue 6043, pp. 736-739
DOI: 10.1126/science.1207272

Abstract

The prevailing view of CO oxidation on gold-titanium oxide (Au/TiO2) catalysts is that the reaction occurs on metal sites at the Au/TiO2 interface. We observed dual catalytic sites at the perimeter of 3-nanometer Au particles supported on TiO2 during CO oxidation. Infrared-kinetic measurements indicate that O-O bond scission is activated by the formation of a CO-O2 complex at dual Ti-Au sites at the Au/TiO2 interface. Density functional theory calculations, which provide the activation barriers for the formation and bond scission of the CO-O2 complex, confirm this model as well as the measured apparent activation energy of 0.16 electron volt. The observation of sequential delivery and reaction of CO first from TiO2 sites and then from Au sites indicates that catalytic activity occurs at the perimeter of Au nanoparticles.

The catalytic behavior of Au/TiO2 contrasts with the inactivity of separate bulk Au and oxide surfaces, as was first recognized in the pioneering work of Haruta and co-workers (1, 2) and in the subsequent applications by others (3). Multiple mechanisms have been proposed to explain the high catalytic activity of oxide-supported Au for CO oxidation at low temperatures, but there is little consensus on the nature of the active sites or the details of the reaction mechanism (36). Although the catalytic importance of the perimeter has been recognized, the role of the support is still generally overlooked. Reactions, and in particular CO oxidation, are thought to occur on Au sites at the perimeter, whereas the support sites are only thought to be involved in stabilizing O2 at the interface (714). The direct reaction of O2 and CO on TiO2 sites that are adjacent to Au particles has not been considered, because weak binding of CO occurs on supports such as TiO2 (15, 16). Although it was reported that CO does not adsorb on smooth Au terrace sites (17), the CO adsorption on coordinatively unsaturated Au sites is stronger than on TiO2 (2, 18, 19), leading to the view that Au delivers CO to active sites for the catalytic process.

The experiments and theoretical results reported here require the reconsideration of this fundamentally important issue where just the opposite behavior has been found. We have carried out CO oxidation over a Au/TiO2 catalyst at low temperatures, where CO is chemisorbed on both Au and TiO2 sites and can be separately observed on these sites during reaction. We used in situ infrared spectroscopy to follow the kinetic changes at both Au and TiO2 sites that reside at the periphery of the Au/TiO2 interface. On the basis of these experimental findings and the results from density functional theory (DFT) calculations, we propose a low-temperature mechanism for the operation of dual Au/TiO2 catalytic sites at the perimeter of Au nanoparticles.

We studied gold clusters (~2-nm to ~8-nm diameter, with a most probable diameter of 3 nm) supported on high-surface-area powdered TiO2 by transmission infrared (IR) spectroscopy (20) (fig. S1). Gold clusters of this size are reported to be the most active for CO oxidation, both in actual supported Au catalysts (2, 6, 21) as well as for model catalysts made by vacuum deposition of Au clusters on single crystal films of TiO2 (Au/TiO2) (22). The chemical deposition-precipitation methods are described in (20), and an electron micrograph of the catalyst after the CO oxidation experiments is shown in fig. S2. The temperature range we have worked at (110 to 130 K) for CO oxidation assures that only the low-activation-energy kinetic steps are being sampled. The observed steps will also occur at the higher temperatures where the full catalytic cycle takes place. For direct comparison, a pure TiO2 sample and a Au/TiO2 sample were separately observed in adjacent positions on a W-mesh support in the same high-vacuum transmittance IR cell reactor with base pressure of 1 × 10−8 torr (fig. S1).

The IR spectra at 0.060-torr CO pressure at 120 K on the Au/TiO2 catalyst and the TiO2 reference sample are shown at saturation coverage in Fig. 1A. A feature at 2179 cm−1, with a small shoulder at 2168 cm−1, that appeared on both samples was caused by CO adsorption on TiO2 (CO/TiO2) (23). The blue shift from the gas phase frequency (2143 cm−1) is generally attributed to CO chemisorbed on a metal cation site such as Ti4+, whereas a red shift is observed for adsorption on metallic sites (17, 23). Two additional features were seen in Fig. 1A on the Au/TiO2 sample. We assigned the broad band centered at 2102 cm−1 in spectrum b as CO chemisorbed on metallic Au (CO/Au) (16, 17, 24) and a small feature at 2341 cm−1 to chemisorbed CO2 on TiO2 sites (CO2/TiO2) on the Au/TiO2 catalyst (16). The CO2 band originated from CO reacting with traces of residual O2 in the high-vacuum cell. CO2 was only produced on the Au/TiO2 catalyst, judging from a comparison of spectra a and b in Fig. 1A.

Fig. 1

(A) IR spectra of saturated CO layer: spectrum a, TiO2; b, Au/TiO2 under 0.060 torr of CO pressure at 120 K. (B) IR spectral development during the CO oxidation reaction on Au/TiO2 under 1 torr of O2 pressure at 120 K: a, before O2 introduction; z, after 120 min of reaction. The CO/TiO2 oxidation was found to continue when the CO coverage was replenished from spectrum z, indicating that the CO2(a) accumulation did not block the active sites. (Inset) The plot of normalized integrated absorbance of CO/Au (red) and CO/TiO2 (black) against time during the experiment.

We evacuated the CO-saturated catalyst at 120 K for 1 hour before conducting further experiments. During the evacuation, the small shoulder (2168 cm−1) arising from weakly bound CO/TiO2 disappeared from the IR spectrum of the Au/TiO2 catalyst (as well as from the pure TiO2 spectrum, fig. S3) as shown in Fig. 1B, spectrum a; none of the more strongly bound chemisorbed CO or CO2 molecules desorbed at 120 K. CO oxidation began immediately when the CO-saturated Au/TiO2 surface was exposed to 1.0 torr of O2 (g) at 120 K, as indicated by the gradual disappearance of the CO absorbance on TiO2 sites and the growth of the CO2 feature on TiO2 sites in the IR spectra shown in Fig. 1B (green spectra). The CO/Au feature hardly decreased in absorbance during the oxidation at 120 K under O2 (g). The integrated IR absorbance of a particular vibrational mode is well known to be proportional to the number of adsorbed species and can thus be used as a quantitative measure for the reaction progress (25). The Fig. 1B inset shows the very different kinetics observed for CO removal from TiO2 sites (fast) and Au sites (slow). The main participant in the reaction is CO/TiO2, which depleted completely. Only about 12% loss in CO/Au coverage occurred in over 120 min. The blue shift (~14 cm−1) of the CO/Au band after O2 introduction was likely the result of the coadsorption of O2 on Au sites (O2/Au) in the oxygen-rich environment (16, 24). Little CO oxidation was detected on the pure TiO2 reference sample (fig. S3), indicating the critical importance of the Au/TiO2 interface.

The results of a kinetics study of the rate of CO/TiO2 consumption on the Au/TiO2 catalyst by the reaction in 1.0-torr O2 (g) at various low temperatures (110 to 130 K) are shown in Fig. 2A. The solid lines represent the best first-order fits to the data and show that the reaction was accurately first order in CO/TiO2 coverage. The inset shows an Arrhenius plot of the data yielding an apparent activation energy of 0.16 ± 0.01 eV for the overall reaction. An Arrhenius analysis of the rate of adsorbed CO2 formation (fig. S4) confirms this activation energy, also yielding an apparent activation energy of 0.16 ± 0.01 eV. The excellent agreement in the kinetics study between CO/TiO2 consumption and CO2/TiO2 generation confirmed that the reaction primarily occurred on the TiO2 sites and that the contribution from CO/Au oxidation was small at 120 K.

Figure 2B shows a plot of the rate of the reaction as a function of the O2 pressure at 120 K, where higher O2 pressure caused the reaction rate to increase. A plot of the reaction rate as a function of O2 pressure (Fig. 2B inset) shows that the rate of CO oxidation maximized at an O2 pressure of ≥2.5 torr at 120 K. This result suggests that the O2 surface coverage on the Au/TiO2 catalyst reached a maximum under these conditions, which caused the rate of reaction to remain constant even at higher O2 pressures.

Fig. 2

(A) Plots of the integrated absorbance of CO/TiO2 against time at various temperatures fitted to first-order kinetics. (Inset) The Arrhenius plot. k, reaction constant; Eact, activation energy. (B) Plots of the normalized integrated absorbance of CO on TiO2 against time under various O2 pressures at 120 K. (Inset) The rate of CO consumption approaches a maximum at oxygen pressures [P(O2)] ≥ 2.5 torr. (C) IR spectra development during temperature-programmed desorption of CO. (Inset) The normalized integrated absorbance of CO/Au (red) and CO/TiO2 (black) against temperature. (D) Spectrum a, IR spectrum of Au/TiO2 catalyst at 120 K in vacuum after thermally removing the CO/TiO2 species; b, IR spectrum of the catalyst shown in spectrum a after treatment in 1 torr of O2 for 35 min at 120 K. (Inset) The integrated absorbance of CO/Au (red) and CO2/TiO2 (black) against time for the process between spectra a and b.

To further examine the site specific CO oxidation on Au/TiO2, we studied the thermal desorption of the CO/TiO2 species. We gradually heated the CO-saturated Au/TiO2 catalyst from 90 K to 215 K in vacuum and selectively removed the CO/TiO2 species while more than 50% of the CO/Au species remained on the surface (Fig. 2C). As the CO coverage decreased upon raising the temperature, CO/TiO2 species desorbed rapidly, whereas the CO/Au species desorbed more slowly. The Fig. 2C inset shows the CO coverage change on Au and TiO2 plotted separately against temperature. The final spectrum with only CO remaining on Au sites at 215 K is shown in Fig. 2C (spectrum z). This catalyst was then cooled to 120 K in vacuum (Fig. 2D spectrum a) and exposed to 1.0 torr of O2. No reaction was detected in 35 min [Fig. 2D, spectrum b and inset]. This experiment shows that, at 120 K, the most strongly chemisorbed CO/Au species mixed with O2/Au species are inactive for CO oxidation, indicating that the active site for low-temperature CO2 formation is not a Au site.

CO chemisorption on Au and TiO2 was weak in both cases. It has been reported that, at temperatures above 25 K, CO does not chemisorb on Au(111) terrace sites, nor on single crystals, nanoparticles, or films (7, 17). However, CO can chemisorb on coordinately unsaturated Au sites, and the binding is stronger than on TiO2. The heat of adsorption of CO/Au (coordinatively unsaturated sites) is reported to be in the range of 0.5 to 0.8 eV (18, 19, 26), whereas the heat of adsorption of CO/TiO2 is between 0.4 and 0.5 eV (15, 19). DFT calculations of CO adsorption energies on Au sites and on TiO2 sites on the Au/TiO2 model presented herein yielded similar results, with binding energies of 0.7 to 0.9 eV for CO/Au and 0.4 to 0.5 eV for CO/TiO2, as shown in fig. S7. Diffusion studies of CO/TiO2 and CO/Au at 140 to 160 K also show that CO diffusion and desorption on TiO2 is more rapid than on Au at all temperatures (fig. S5).

To help identify the active sites for the low temperature CO oxidation at the Au/TiO2 interface, we carried out gradient-corrected DFT calculations. A Au nanorod (three atomic layers high) supported on the rutile TiO2(110) surface was used to model the nanoparticle Au/TiO2 catalysts used experimentally, because it provides a reliable model of the perimeter sites at the edges of the Au/TiO2 interface (fig. S6). In light of the experimental results presented above, the theoretical calculations focused on the reactions that occur at perimeter sites between TiO2 and Au.

O2 activation, the critical first step in CO oxidation, has been postulated to occur either through direct O2 dissociation at the perimeter or through a bimolecular CO-O2 reaction (1114). We have considered both mechanisms at each of the possible sites on the Au nanoparticles, the TiO2 support, and the Au/TiO2 perimeter sites. Detailed information on the results for both mechanisms at all of the sites considered can be found in section IIIB of the supporting online material (SOM) text and figs. S8 and S9, where only the lowest barrier configurations for each of the possible sites are discussed. Out of all of the possible adsorption sites that exist at the Au/TiO2 interface, O2 was most strongly bound to the Ti5c site adjacent to a Au atom at the perimeter. The adsorption at this perimeter site is the only one in which O2 binds in a di-σ configuration with bonds to the Ti5c site of the support as well as to Au at the interface. In addition, there is an electronic stabilization of O2 at this site caused by charge transfer from the metal to the Ti5c. This perimeter Ti5c site also results in the lowest calculated barrier of 0.16 eV for the activation of adsorbed O2 [O2 (a)], where O2 dissociation on Ti5c sites is directly assisted by neighbor CO molecules on the Ti5c sites. The dissociation of O2 (a) at other sites including coordinatively unsaturated Au sites and TiO2 sites, resulted in higher activation energies ranging from 0.39 to 0.7 eV, as discussed in SOM text section IIIB.

The results from DFT calculations for the reaction energies and activation barriers for all of the steps proposed in the mechanism for CO oxidation at the active Ti5c/Au interface are shown in Fig. 3. Because the experiments discussed above indicate that CO/TiO2 is consumed more rapidly than CO/Au, we start with CO on TiO2 (Fig. 3A). The CO coverage at the TiO2 perimeter sites is set at two-thirds of a monolayer, whereas the rest of the Ti5c sites are all occupied to simulate the saturation coverage of CO used experimentally. This step is followed by the fast coadsorption of O2 at the dual perimeter site adjacent to the CO/TiO2 (Fig. 3B) with a strong binding energy of –1.01 eV. The adsorbed O2 reacts directly with one CO/TiO2 to form a CO-O2 intermediate that is stabilized by its interaction with the Ti5c site and the neighboring Au site at the perimeter (Fig. 3C); this step has an activation energy of 0.13 eV. The O-O bond is substantially weakened upon its interaction with CO to form an O=C∙∙∙O2 intermediate that dissociates with an activation energy of only 0.10 eV and generates CO2-Ti and O-Ti (Fig. 3D). The remaining O-Ti, which is trapped as a result of its high diffusion barrier of 0.75 eV (SOM text section IIIC and fig. S10), can readily react with a second CO at the active Au/TiO2 periphery with a barrier of 0.11 eV (Fig. 3E). The formation of O=C⋅⋅⋅O and O=C⋅⋅⋅O⋅⋅⋅C=O complexes on Ti5c sites on TiO2(110) has been directly observed at 100 to 110 K by scanning tunneling microscopy recently (27, 28). The CO2 formed can easily diffuse away from the active site or desorb with a barrier of only 0.20 eV (Fig. 3G). The perimeter sites are then reoccupied by CO, which diffuses from neighboring Ti5c sites with a barrier of 0.26 eV (Fig. 3, F to H). This step returns the catalyst surface to its initial working state. The reaction continues by adsorbing a second O2 (g) at Au perimeter sites and by CO/TiO2 surface diffusion (Fig. 3B).

Fig. 3

CO oxidation cycle proposed from DFT calculation results. The Au atoms and Ti atoms are shown in yellow and gray, respectively, whereas the O in TiO2 lattice and adsorbed O and C atoms are shown in pink, red, and black, respectively. ∆Eads, Ea, Ediff, ∆Edes, and ∆H refer to the binding energy, activation barrier, diffusion barrier, desorption energy, and reaction enthalpy, respectively. The elementary steps depicted include the adsorption of O2 (A to B), interaction between O2 and CO (B to C), reaction of O2 and CO (C to D), reaction of adsorbed atomic oxygen and CO (D to E), diffusion of CO (E to F and G to H), and desorption of CO2 (F to G and H to A). (a) indicates adsorbed species.

The catalytic cycle shown in Fig. 3 is consistent with the experimental observation that CO adsorbed on TiO2 containing Au nanoparticles is consumed by reaction with O2 at the interface. The activation of O2 by CO and the reaction to form CO2 and O proceed at dual sites along the perimeter involving a Au atom and a Ti5c site. The resulting O/Ti5c species readily reacts at low temperature with CO, which diffuses from the TiO2 support to the active Ti5c site. The calculated barrier for this reaction is only 0.11 eV, which is consistent with previous results (14). The CO species on Au are much less mobile, with diffusion barriers that exceed 0.5 eV because of the strong adsorption of CO (SOM text section IIIC, figs. S10 to S12, and table S1). The calculated activation barriers for the elementary steps reported in the oxidation of CO at the Ti5c site (0.10 to 0.26 eV) are all quite close to the experimentally measured apparent activation energy (0.16 ± 0.01 eV). Although it is difficult to distinguish a rate-determining step without full kinetic simulations, the simple analysis of the activation energies in Fig. 3 suggests that the diffusion of CO from the TiO2 support to the Ti5c perimeter site has the highest barrier. The apparent activation energy for CO oxidation of 0.16 eV measured herein at low temperatures is in good agreement with previously reported barriers of 0.2 to 0.3 eV from measurements carried out at much higher temperatures (200 to 350 K) and pressures on Au/TiO2 catalysts (29, 30). The catalytic cycle proposed in Fig. 3 should likely hold, even at high temperature and CO pressure conditions, but requires the addition of CO adsorption from the gas phase (–1.01 eV) and the CO diffusion from Au, which can proceed at higher temperatures (0.5 eV) as reflected in Fig. 3A. Higher temperatures will also likely result in the desorption of CO from the Ti5c sites and increase the oxidation of CO adsorbed at the Au sites.

This Au/TiO2 interface and the processes that occur here are shown schematically in Fig. 4. O2 (g) is captured at the perimeter site (process 1 in Fig. 4). CO molecules on TiO2 sites are initially delivered to the active perimeter sites via diffusion on the TiO2 surface, where they assist O-O bond dissociation and react with oxygen at these perimeter sites (processes 2 and 3 in Fig. 4). After most of the CO/TiO2 is depleted from TiO2 in the perimeter zone, a small fraction of CO/Au (~12% at 120 K) with weaker binding energies and lower diffusion barriers is oxidized. The remaining tightly bonded CO/Au is kinetically isolated from the CO oxidation at 120 K, because these CO/Au surface species cannot approach the active perimeter sites. At higher temperatures, however, the CO on the Au sites become more mobile and can begin to actively participate in catalytic CO2 production (process 4 in Fig. 4). The observation of the sequential delivery of the two types of adsorbed CO, present on both sides of the perimeter of the Au nanoparticles, argues persuasively for reactivity to occur within a zone at the perimeter of Au particles surrounded by TiO2 surface sites which form dual sites for the reaction.

Fig. 4

Schematic of the mechanism of low-temperature CO oxidation over a Au/TiO2 catalyst at a perimeter zone of reactivity. Experiments directly observing CO/TiO2 and CO/Au surface species show that processes 2 and 3 are fast compared with process 4.

Supporting Online Material

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

Materials and Methods

SOM Text

Figs. S1 to S12

Table S1

References (3164)

References and Notes

  1. Materials and methods are available as supporting material on Science Online.
  2. Acknowledgments: We acknowledge the support of this work by the U.S. Department of Energy–Office of Basic Energy Sciences, grant number DE-FG02-09ER16080 as well as the NSF and the Texas Advanced Computing Center for Teragrid resources. All data are archived at University of Virginia.
View Abstract

Stay Connected to Science


Editor's Blog

Navigate This Article