Structure and Catalytic Activity of Gold in Low-Temperature CO

Mar 24, 2009 - Chemical Engineering, The Queen's UniVersity of Belfast, Belfast BT9 5AG, U.K. .... East China University of Science and Technology...
0 downloads 0 Views 1MB Size
6124

J. Phys. Chem. C 2009, 113, 6124–6131

Structure and Catalytic Activity of Gold in Low-Temperature CO Oxidation Hai-Feng Wang,† Xue-Qing Gong,*,† Yang-Long Guo,† Yun Guo,† Guanzhong Lu,*,† and P. Hu‡ Laboratories for AdVanced Materials, Research Institute of Industrial Catalysis, East China UniVersity of Science and Technology, 130 Meilong Road, Shanghai 200237, P. R. China;, School of Chemistry and Chemical Engineering, The Queen’s UniVersity of Belfast, Belfast BT9 5AG, U.K. ReceiVed: December 2, 2008; ReVised Manuscript ReceiVed: February 7, 2009

Structures and catalytic activities of Au thin films supported at anatase TiO2(101) and a Au substrate are studied by using density functional theory calculations. The results show that O2 can hardly adsorb at flat and stepped Au thin films, even supported by fully reduced TiO2(101) that can highly disperse Au atoms and offer strong electronic promotion. Interestingly, in both oxide-supported and pure Au systems, wire-structured Au can adsorb both CO and O2 rather strongly, and kinetic analysis suggests its high catalytic activity for low-temperature CO oxidation. The d-band center of Au at the catalytic site is determined to account for the unusual activity of the wire-structured film. A generalized structural model based on the wire-structured film is proposed for active Au, and possible support effects are discussed: Selected oxide surfaces can disperse Au atoms and stabilize the formation of a filmlike structure; they may also serve as a template for the preferential arrangement of Au atoms in a wire structure under low Au coverage. 1. Introduction The chemistry of gold dramatically changes when it is highly dispersed as ultrafine nanoparticles at selected metal oxides. It exhibits unexpected catalytic activity toward many important reactions in heterogeneous catalysis, such as CO oxidation,1-19 oxidation of hydrocarbons,2,20 the water-gas-shift reaction,21-24 and olefin epoxidation.25-28 Among them, CO oxidation is the most widely investigated, largely because of its industrial importance and simplicity. A very large amount of work has been conducted to illuminate the origin of the extraordinary activity of metal-oxide-supported gold nanoparticles toward lowtemperature CO oxidation, despite that no consensus was achieved. In particular, the nature of the catalytic site in supported Au systems still remains controversial. Haruta and Iwasawa and their co-workers2,29,30 suggested that CO oxidation would occur at the interface between the Au and metal oxide. Liu et al.19 reported the results of CO oxidation at the Au/TiO2 (rutile) interface using density functional theory calculations. They found that O2 can adsorb at the interface with an adsorption energy of ∼0.6 eV, and there is no apparent barrier in the oxidation process. Very recently, still with the interface model, Liu and co-workers proposed that the change of the p-bandwidth of the adsorbed O2 molecule and d-occupation of the metal oxide can help to understand the support effect.31 Molina et al.32,33 also revealed that O2 adsorption is reasonably strong at the Au/TiO2 interface. Goodman and co-workers5 investigated the unusual size dependence in low-temperature CO oxidation by gold islands dispersed on single crystalline surfaces of TiO2 and suggested that factors such as the thickness, shape, and the oxidation state of gold nanoparticles can affect their catalytic activity. Goodman and co-workers7,8 also reported that well-ordered Au monolayers and bilayers fully covering the oxide support TiOx/Mo(112) can * Corresponding authors. E-mails: (X.Q.G.) [email protected], (G.L.) [email protected]. † East China University of Science and Technology. ‡ The Queen’s University of Belfast.

be prepared, thus eliminating particle shape and interface effects, and the supported Au bilayer film exhibits an unprecedented activity toward CO oxidation. Moreover, the catalytic behavior of such a supported Au bilayer film has been investigated by Herna´nder et al.34 and Rashkeev et al.35 using density functional theory. By measuring the catalytic activity of supported Au nanoparticles toward CO oxidation as a function of the average Au particle size, Norskov and co-workers proposed that a high concentration of under-coordinated Au atoms is the key factor for the particles’ activity and specified the importance of the Au at the corner sites.10-13 Freund and co-workers also suggested the role of the highly uncoordinated Au atoms in affecting the activity for low-temperature CO oxidation.14 In addition, calculations by Mills et al.36 showed that O2 molecules do not bind to Au(111), but they can strongly adsorb at Au clusters deposited on a Au(111) surface, which emphasized the effect of surface roughness. Landman and co-workers16 revealed that charging of a Au8 cluster at oxygen vacancies of MgO support plays a key role in the enhancement of catalytic activity. Guzman and Gates and Kung and their co-workers also showed that oxidized gold is crucial to the high activity of dispersed gold.37-40 Matthey et al.41 proposed that the cationic state of gold can help to stabilize the gold cluster. Very recently, some surprising experimental results showed that nonsupported gold in foam or other unresolved structures also exhibits a very high catalytic activity toward CO oxidation.42,43 It indicates that pure gold, when taking a specific atomic configuration, would be highly active, and the interfacial site or specific oxidation state of Au may not necessarily contribute to the enhanced catalytic activity of supported nanogold. Therefore, to understand the origin of Au activity, it is important to study the intrinsic properties of gold itself and the dependence of catalytic activity on specific Au structures. Gold is rather inert toward oxygen.44 Accordingly, once formed at a gold surface, oxygen atoms would be very unstable and reactive. However, formation of atomic oxygen via direct

10.1021/jp810608c CCC: $40.75  2009 American Chemical Society Published on Web 03/24/2009

Activity of Gold in Low-Temperature CO Oxidation

J. Phys. Chem. C, Vol. 113, No. 15, 2009 6125

2. Calculations

layers fixed. By using the system discussed in Section 3.1 (Figure 4d, see below), we determined that the Au cluster structure deposited on the four-layer support remains unchanged as compared to that in the three-layer model, and the CO/O2 adsorption energies on such Au/TiO2 give a trivial difference of +0.01 and +0.04 eV, respectively. For pure gold systems, we used at least four layers to model various Au slabs, and we also used surface cells that are big enough to avoid any interaction of adsorbates in neighboring cells. To study CO, O2, and intermediate OOCO adsorption at different Au surfaces, we carried out spin-polarized calculations when necessary. The transition states were searched using a constrained optimization scheme and were verified when (i) all forces on atoms vanish; and (ii) the total energy is a maximum along the reaction coordination but a minimum with respect to the rest of the degrees of freedom. The force threshold for the optimization and transition state search was 0.05 eV/ Å. In this work, the coverage of Au clusters at TiO2(101) is defined as the number of Au atoms with respect to the number of Ti (Ti5c and Ti6c) exposed. For example, in the 2 × 2 surface cell, four Ti5c and Ti6c are exposed, respectively, and the coverage of four-atom Au wire extended along [111j] was determined to be 0.5 monolayer (ML). To calculate the stabilities of adsorbed Au clusters, the average adsorption energy per Au coh ) -(EAun/TiOx - ETiOx - nEAu)/n, in atom was considered: EAu n which EAun/TiOxis the total energy of the Au clusters, with n Au atoms at TiOx support, ETiOx is the total energy of the TiOx slab, and EAu is the total energy of a single Au atom in the gas phase. For comparison, the average Au adsorption energies were also coh calculated in bulk Au and on the Au(111) surface with Ebulk_Au Au coh ) -(Ebulk - nEAu)/n and EAu(111) ) -(EAu(111) - nEAu)/n, Au (EAu(111)) is the total energy of the respectively, in which Ebulk Au bulk (fully relaxed four-layer Au(111) slab). It needs to be mentioned that it is known that GGA approximations usually do not account for the long-range forces, such as van der Waals interaction, that determine weak adsorption processes. Therefore, the reported calculation results for CO and O2 weakly adsorbed at various Au structures may include some errors. However, the relative activities of these Au structures, determined from calculated adsorption energies, are expected to be largely reliable. It is also known that normal DFT calculations cannot accurately treat localized gap states at nonstoichiometric surfaces of metal oxides, such as TiO2 and CeO2. In the current work, reduced anatase TiO2(101) surfaces were used. However, the O-vacancy sites at these surfaces are fully covered by Au atoms and strips, and the whole slabs essentially become metallic. Therefore, we expect that the use of modified DFT methods, that is, DFT + U, may not be necessary here.

The total energy DFT calculations were carried out with the GGA-PBE functional54 using the PWScf code included in the Quantum-Espresso package.55 Electron-ion interactions were described by ultrasoft pseudopotentials,56 with electrons from Ti 3s, 3p, 3d, 4s; O 2s, 2p; Au 5d, 6s; and C 2s 2p shells explicitly included in the calculations. Plane-wave basis set cutoffs for the smooth part of the wave functions and the augmented density were 25 and 200 Ry, respectively. The anatase TiO2(101) surface was modeled as a periodic slab with three layers of oxide, and the vacuum between slabs is ∼10 Å. A 2 × 2 surface slab with a corresponding 2 × 2 × 1 k-point mesh was used. A limited number of Au atoms were placed at the surface to simulate the Au clusters in filmlike structures. During structural optimization, all the atoms of the slab and Au clusters were allowed to relax, except the two layers of oxide at the bottom. We also tested the four-layer oxide model with the bottom two

3. Results and Discussion 3.1. Au Films at Anatase TiO2(101). The anatase TiO2(101) surface exhibits a sawtoothlike surface corrugation containing both fully saturated 3(6)-fold and coordinatively unsaturated (cus) 2(5)fold O(Ti) atoms, O3c(Ti6c), and O2c(Ti5c), respectively (see Figure 1a). Along the [111j] direction of the calculated anatase TiO2(101), the two nearest O2c atoms have an interatomic distance of 5.556 Å, about two times of that in bulk Au (∼2.8 Å). Accordingly, three Au atoms can be accommodated in between without significant strain (see Figure 1b), and compact Au structures would match well with the support along this direction. Stoichiometric Surface. As shown in Figure 2a, at a coverage of 0.5 ML, Au atoms adsorbed on the surface O2c form a continuous wire structure that contains a single row of atoms along the [111j] direction. When the Au coverage increases to

O2 dissociation is extremely difficult at Au due to a high barrier.45-47 This is different from some other transition metals, such as Pt, on which direct O2 dissociation can readily occur and provide O atoms to oxidize CO.48,49 Therefore, to turn on its activity for oxidation reactions, it is necessary to provide active oxygen species through some specific pathway. For example, in the CO-assisted O2 dissociation (CO + O2 f O-OCO f CO2 + O), both the product (CO2) and very active O atom to further oxidize another CO can be formed, and this process is widely believed to be facilitated in the supported Au systems and responsible for their high catalytic activity.19,29-33 Experimental evidence of high catalytic activity of a pure Au system thus encourages us to investigate whether special assembly of Au atoms would favor the above reaction pathway and whether the activity of Au with specific structures would be of general importance to the chemistry of Au in both supported and nonsupported systems. In this work, aiming at obtaining a deep insight into the catalytic activity of Au and corresponding structures in the systems with and without oxide support, we studied Au thin films supported at metal oxide and a Au substrate by using density functional theory (DFT) calculations. In particular, we considered anatase TiO2 as the support, since it is a reducible and industrially important metal oxide and Au supported on it indeed exhibits high activity toward CO oxidation.50-52 More importantly, on the most often exposed (101) surface of anatase TiO2, the lattice parameters along the direction of [111j] matches well with the interatomic distance in bulk gold. Therefore, it may help us model ordered and extended nanostructures, such as Au films, at oxide support with relatively small surface cells. In addition, the various ordered Au structures obtained at this support could be readily modeled in pure Au systems. In fact, combined scanning tunneling microscopy and first-principles calculation studies performed recently showed that on anatase, ordered transition metal nanoparticles (Pt and Au) can readily occur under a wide range of surface reduction states.53 Through calculations, we found that gold with a wire structure and several other specific configurations characteristic of exposing (at least two) highly under-coordinated Au atoms side by side can efficiently adsorb CO and O2 and promote CO oxidation in both supported and nonsupported systems. By performing systematic kinetic analysis of the reaction, we also quantitatively estimated the turnover frequency (TOF) of CO oxidation catalyzed by Au with a wire structure and determined its key steps. In addition, electronic analyses were conducted to elucidate the physical origin of the high activity of Au with such structures.

6126 J. Phys. Chem. C, Vol. 113, No. 15, 2009

Figure 1. Calculated structures (side view) of (a) anatase TiO2(101) surface and (b) Au wire supported at surface O2c in the direction of [111j]. The Ti, O, and Au atoms are in gray, red and yellow, respectively. This notation is used throughout this paper.

Figure 2. Calculated structures (side view) of Au films at stoichiometric anatase TiO2(101) surfaces under a coverage of (a) 0.5, (b) 1, (c) 1.5, and (d) 2 ML.

Figure 3. Variation of average adsorption energies of Au at (a) stoichiometric (black line), (b) half-reduced (brown line), and (c) fully reduced (blue line) anatase TiO2(101) surface with respect to coverages. Red and green lines indicate the average adsorption energies of Au in bulk and at the (111) surface, respectively.

1 ML, the second Au wire prefers to stay above the first one rather than surface O2c as the result of strong Au-Au interaction, and they form an upright strip (see Figure 2b). When the coverage reaches 1.5 ML, an extended flat Au(111)-like film would occur, and it has almost no contact with the support (see Figure 2c), largely because of lattice mismatch with the support along the [010] direction as well as the presence of stronger Au-Au bonds as compared to the Au-O ones. Further increasing the Au coverage leads to a stepped film structure, as shown in Figure 2d. The change in average adsorption energy of Au adsorbed on the stoichiometric TiO2(101) surface with respect to the coverage is depicted in Figure 3. We can see that it increases from 1.77 to 2.30 eV/atom as the coverage varies from 0.5 to 1.0 ML; at 1.5 ML, the average adsorption energy reaches a local maximum of 2.77 eV/atom; at 2.0 ML, the average adsorption energy declines slightly to 2.75 eV/atom. Half-Reduced Surface. By removing the two O2c along the [111j] direction in a 2 × 2 surface cell, we can obtain the half-

Wang et al.

Figure 4. Calculated structures (side view) of Au films at half-reduced anatase TiO2(101) under coverages of (a) 0.5, (b) 2.0, (c) 1.75, and (d) 2.25 ML.

reduced anatase TiO2(101). Snapshots of several calculated structures at this surface are shown in Figure 4a-d. Compared to the stoichiometric surface, the half-reduced one has two neighboring oxygen vacancies in one surface cell, on which Au atoms can strongly adsorb. As presented in Figure 4a, the two oxygen vacancies can accommodate a row of four Au atoms in a configuration very similar to that at the stoichiometric surface. However, the average adsorption energy was estimated to be as high as ∼2.9 eV/atom (Figure 3). At 2 ML, a similar extended but stepped Au film can be obtained (see Figures 4b and 2d). Interestingly, different from that at the stoichiometric surface, under the coverages between 1 and 2 ML, the filmlike Au structures can also form and tightly bind to the support via anchoring Au atoms at the O vacancies. For example, at 1.75 ML, an ordered flat Au film can occur (see Figure 4c), on which another row of Au atoms can adsorb to form a film with protruding Au wire (see Figure 4d). The change in the average adsorption energy with respect to the coverage of Au is also illustrated in Figure 3. At a Au coverage ranging from 0.5 to 2.25 ML, a local maximum at 1.75 ML and two local minima at 1.25 and 2.25 ML were determined. It can be expected that the average adsorption energy includes the contribution from Au-Au and rather strong Au-Ti bonds at the oxygen vacancies. At 1.75 ML, Au atoms supported by the anchoring Au at the vacancy sites can form a well-ordered flat structure along both [010] and [111j] directions, and the average adsorption energy therefore reaches the local maximum. With or without a row of Au atoms, the continuity of the Au film along [010] direction will be broken, resulting in a local minimum in average adsorption energy. Fully Reduced Surface. By removing all the four O2c in a 2 × 2 surface cell, we can obtain a fully reduced anatase TiO2(101) surface. Compared to the half-reduced one, a fully reduced surface has more oxygen vacancies to accommodate Au atoms (see Figure 5a) and is therefore more likely to be wetted by a Au film. At a coverage of 1.5 ML, a flat Au thin film can form (see Figure 5b), corresponding to a local maximum in the average adsorption energies depicted in Figure 3. Incorporation of an extra row of Au atoms results in a protrusion structure (Figure 5c), which can reconstruct into a more stable flat structure (see Figure 5d). In this structure, four Au atoms occupy four oxygen vacancies, respectively, and act as anchoring points for other Au atoms forming an extended flat film. At a coverage of 2.5 ML (see Figure 5e), the calculated film also exhibits a flat bilayer structure, and the average adsorption energy was estimated to be 3.01 eV/atom. From the above calculation results of Au thin films supported at stoichiometric and half- and fully reduced TiO2(101) surfaces, we can see that along the lattice-matched [111j] direction, Au atoms can always form an ordered continuous structure, whereas in the [010] direction, a strain effect can be introduced due to the lattice mismatch. Compared to the stoichiometric surface, half- and fully reduced ones can bind Au atoms more tightly,

Activity of Gold in Low-Temperature CO Oxidation

Figure 5. Calculated structures (side view) of Au films at fully reduced anatase TiO2(101) under coverages of (a) 1, (b, f (top view, inset: side view)) 1.5, (c, d) 2, and (e) 2.5 ML. (g) Calculated structure of CO and O2 adsorption at a Au thin film with wire structure deposited at half-reduced TiO2(101) (Figure 4d).

and the Au atoms at the vacancies may also act as anchoring sites for the film structures to occur at the surface. Moreover, average adsorption energies of Au in film structures at reduced surfaces, especially the fully reduced one, are largely comparable to those in stable Au systems (bulk and (111) surface, see Figure 3). These results indicate that reduced anatase TiO2 can favor the formation of Au clusters in a preferential 2D model. 3.2. Activity of Supported Au Films. Among various Au films calculated at half- and fully reduced anatase TiO2(101), four characteristic structures can be recognized: (i) flat films containing Au atoms in direct contact with the reduced surface Ti, thus introducing negatively charged Auδ- (see Figure 5b and f); (ii) flat film supported by anchoring Au atoms at the vacancy sites (see Figure 4c), which only gives exposed Au in neutral (Au0); (iii) stepped film (see Figure 4b); and (iv) wire film (see Figures 4d and 5c) containing a row of Au atoms at the flat Au film. It should be noted that compared to the stepped film, the protruding Au line on the wire film appears to form a structure that is obviously much sharper, and atoms in the line also have a lower coordination number (CN). At these four structures, we first investigated CO and O2 adsorption to explore their reactivity toward low-temperature CO oxidation. The calculated adsorption energies and key structural parameters are listed in Table 1. As one can see, on the flat films i and ii, an O2 molecule can hardly adsorb, regardless of being between Auδ- and Au0 or two Au0, although Auδ- is usually expected to provide more easy-to-get electrons.36 In addition, CO adsorption at Auδ- is less favored than that at Au0. On the other hand, stepped films have an enhanced capacity of binding CO at the edge sites, but they are still unable to efficiently adsorb and activate O2. Interestingly, both CO and O2 can strongly adsorb on wire films (see Figure 5g). In particular, the adsorption energy of O2 was calculated to be ∼0.5 eV, and it is largely activated with bond length (d(O-O)) increased from 1.239 to 1.377 Å. These results suggest that, among the four characteristic supported structures, wire films may have higher catalytic activity toward CO oxidation. 3.3. Pure Gold Systems. To further check the possible effect induced by support and explore the relationship between the catalytic activity and the structure of Au, we investigated, in the pure gold systems, the characteristic film structures discussed above. We used Au(111), Au(211), and Au-wire/Au(111), which consists of a single row of Au atoms adsorbed on Au(111), to model flat, stepped, and wire structures for pure gold (see Figure

J. Phys. Chem. C, Vol. 113, No. 15, 2009 6127 6). The calculated geometries and adsorption energies of CO and O2 at these structures are given in Table 2. Similar to what we found at TiO2-supported Au films, O2 cannot adsorb efficiently on flat Au(111) and Au(211) surfaces, though stepped Au(211) can adsorb the CO molecule with reasonable strength. These results are also consistent with those reported in some early theoretical work.18 Interestingly, CO and O2 can adsorb rather strongly on the Au wire on Au(111). In particular, the adsorbed O2 has an adsorption energy of 0.64 eV, and the O-O distance was determined to be 1.329 Å, indicating that it is highly activated. The above results obtained in various pure gold systems clearly suggest that the local atomic configuration of Au, rather than the support, may determine its catalytic activity. Specifically, the wire structure could be of general importance to the activity for low-temperature CO oxidation. It is worth emphasizing that we do not rule out the contribution of the charged Au atom to efficient O2 adsorption and activation on some small clusters (usually containing ∼10 Au atoms) in unsupported or oxide supported systems.16,37-41,57 However, for the extended Au structures investigated in our study, this effect seems to be less important compared to the geometries of Au. 3.4. Kinetic Analysis. To quantify the overall activity of Au with the wire structure toward CO oxidation and to compare it with those on other possible reaction sites, we performed a detailed kinetic analysis to estimate the turnover frequency of this reaction. By doing so, we may also be able to shed light on the key steps occurring in catalytic Au systems. As mentioned before, CO oxidation by O2 molecules is believed to follow a bimolecular (CO + O2) reaction route; that is, CO and O2 adsorb on the Au surface and then form an intermediate O-OCO complex,31-33 which further produces CO2 and an O adatom.18,31 After that, adsorbed CO reacts with the O adatom to form another CO2 and complete the reaction cycle. This process is presented in the following scheme, K1eq

1. CO + * y\z CO* K2eq

2. O2 + * y\z O2* k3+

3. CO* + O2* y\z O-OCO* + * k3-

k4+

4. O-OCO* y\z CO2 + O* k4-

k5+

5. CO* + O* y\z CO2 + 2* k5-

where CO*,O2*,O-OCO* and O* are adsorbed species at surfaces, and * represents the surface free site. The intermediate O-OCO exhibits a coplanar structure with its C of CO binding to one Au atom and one O of O2, which sits on another Au atom. Moreover, the O-C-O and O-C-Au angles were measured to be about 116°. Kinetic derivation of the reaction rate (turnover frequency, TOF, in s-1) for CO2 formation gives the following equation (see Supporting Information for details),

TOF ) k3+θCOθO2(1 - Z3) )

k3+pCOpO2K1eqK2eq(1 - Z3) (1 + pCOK1eq + pO2K2eq+ pCOpO2K1eqK2eqK3eqZ3)2

(1)

6128 J. Phys. Chem. C, Vol. 113, No. 15, 2009

Wang et al.

TABLE 1: Calculated Structural Parameters and Adsorption Energies of CO and O2 at the Characteristic Flat, Stepped and Wire Filmsa Au (Au -Au ) Au0(Au0-Au0) δ-

flat (i)

δ-

0

flat (ii) stepped structure (iii) wire structure (iv)

d(Au-C)

d(Au-O)

d(O-O)

ad ECO

EOad2

Figure

2.235 2.007 2.032 1.986 1.955

>3.0 >3.0 >3.0 >3.0 2.090/2.111

1.242 1.243 1.238 1.241 1.377

0.10 0.29 0.26 0.55 0.86

0.05 0.02 0.03 0.04 0.48

5b, f 4c 4b 4d

a

i-iv structures; see Sec. 3.2. Corresponding structures are presented in Figures 5b and f and 4b, c, and d, respectively. The bond length (d) and adsorption energy (Ead) are in Å and eV, respectively.

Figure 6. Illustrations (side view) of bulk-truncated (a) Au(211), (b) Au(210), and (c) Au(134) surfaces and (d) Au monomer, dimer, and wire at Au(111) (inset: top view). The light blue balls are active Au atoms, where adsorption of molecules was calculated and coordination numbers (see Table 2) were counted. d is the distance between neighboring active Au atoms.

TABLE 2: Calculated Structural Parameters and Adsorption Energies of CO and O2 in Various Pure Au Systemsa Au(111) Au(211) Au(210) Au(134) Au wire Au1/Au(111) Au2/Au(111)

d(Au-C)

d(O-O)

ad ECO

EOad2

CN

1.992 1.973 1.965 1.948 1.946 1.928 1.938

1.239 1.308 1.244 1.273 1.329 1.275 1.332

0.25 0.62 0.69 0.84 0.93 1.23 1.07

0.01 0.10 0.04 0.15 0.64 0.31 0.57

9 7 6 6 5 3 4

a The bond length (d) and adsorption energy (Ead) are in Å and eV, respectively.

In this equation, ki +, Kieq, and Zi are the forward reaction rate constant, equilibrium constant, and reversibility of step i, respectively; θCO and θO2 are the coverage of adsorbed CO and O2, respectively; pCO, pO2, and pCO2 are the partial pressure of CO, O2, and CO2, relative to the standard atmosphere pressure, respectively. The barriers (Eb) of the different reactions (steps 3-5) on Au-wire/Au(111) calculated in this work are listed in Table 3, together with those reported in ref 18 for the same reactions at Au(211). To estimate the reaction rate, we used the experimental conditions T ) 298 K, pCO ) 1 × 10-2, and pO2 ) 0.2, and entropy data of gas-phase reactants at T ) 298 K are taken from ref 59. In Table 3, we list the key kinetic parameters calculated from the above data and equations. Accordingly, the TOF of the overall reaction was calculated to be 4.4 × 10-1 s-1 at Auwire/Au(111). In contrast, stepped Au(211) was determined to give a TOF as low as 7.0 × 10-7 s-1, whose low reactivity was

recently validated in a single crystal experiment by Koel and co-workers.60 From the kinetic analysis, we can see that an O atom formed as a result of the previous reaction step can be quickly consumed ((θO)/(θCO) ) (pCO2)/(p2CO)(1)/(K21eqK5eq) , 1; see the Supporting Information), which indicates that the determining step of the whole reaction is to produce adsorbed O atoms (steps 3 and 4). Furthermore, we can also learn from eq 1 that the overall TOF is largely determined by the rate constant, k3, and the coverage of CO and O2 molecules, since reversibility, Z3, is usually very small. Then it can be expected that in bulklike Au systems (Au(211), for example), on which CO and O2 molecules hardly adsorb, θCOθO2 would be rather small, and so is the reactivity (TOF ) 7.0 × 10-7 s-1 for Au(211)). On the other hand, both CO and O2 can adsorb efficiently on Au with wire structure, ensuring high coverage of CO and O2 without decreasing the k3 significantly,61 leading to a rather high TOF for CO oxidation. Considering that CO and O2 adsorption on Au-wire/TiO2(101) has a strength very similar to that on Au-wire/Au(111) (see Tables 1 and 2), we can also expect the TOF on the oxidesupported Au-wire to be on the same order as 4.4 × 10-1 s-1. Remarkably, this estimated TOF is quite consistent with the experimental results in pure Au- and TiO2-supported systems (Table 3). It should be noted that, to keep high activity, the binding ability of reactive Au sites should not be too strong. Otherwise, they will be poisoned and give a low value of θCOθO2 due to the competing adsorption of CO, O2, or intermediate OOCO. 3.5. Origin of the Au Activity. In the above, we have shown that Au with wire structure can exhibit a rather high catalytic activity toward CO oxidation, as evidenced by the strong O2/ CO adsorption and large TOF. To rationalize the enhanced adsorption energy of O2/CO on the Au wire structure, the Anderson-Newns model62 and d-band center theory63 are taken into account. The adsorbate-metal bonding occurs via two steps: (i) charge transfer from the metal s/p orbitals to the adsorbate valence states; and (ii) the broadened adsorbate valance state further mixes with the localized metal d states. Accordingly, we calculate the d-band center (relative to Fermi level) of the Au atom directly binding with an O2 or CO molecule in the various systems, which include Au-wire/Au(111) (Figure 6d), Au-wire/HR-TiO2 (Figure 4d; HR-, half-reduced), Au-wire/FR-TiO2 (Figure 5c; FR-, fully reduced), flat-Au/FRTiO2 (Figure 5b), Au(111), and Au(211). The dependence of O2/CO adsorption energy on the d-band center is plotted in Figure 7, which shows a linear correlation between the d-band center and the O2/CO adsorption energy. Thus, the higher position of the d-band center supports the enhanced binding ability of the Au wire structure. 3.6. Structural and Supporting Effects. Thus far, we have shown that pure Au systems can exhibit remarkable catalytic activity toward CO oxidation, once the specific atomic configurations exist. This is consistent with recent experimental

Activity of Gold in Low-Temperature CO Oxidation

J. Phys. Chem. C, Vol. 113, No. 15, 2009 6129

TABLE 3: Calculated Kinetic Parameters of CO Oxidation on Wire Structure and Stepped Au(211) wire structure E3b E4b E5b k3+ k4+ TOF

0.60 0.46 0.59 7.187 × 102 1.672 × 105 4.4 × 10-1

Au(211) 0.68 0.43 3.192 × 101 5.377 × 105 7.0 × 10-7

wire structure θ θCO θO2 θOOCO θO TOF (exptl)

results that nonsupported Au with unresolved network structure has a rather high catalytic activity for CO oxidation.42,43 To obtain more general insights into the activity of Au and corresponding structures, we extended the calculations from the wire structure to other pure Au systems with different characteristic structures: Au(210), Au(134), Au1/Au(111), and Au2/ Au(111). The (210) surface represents a type of structure that separates the undercoordinated (CN ) 6) Au atom (Au*) with a large distance (d(Au*-Au*) ) 4.12 Å; see Figure 6b). The vicinal (134) surface, with the cornerlike Au atoms exposed, is used to model the structures containing undercoordinated Au at corner sites (CN ) 6) (see Figure 6c). The Aun(n)1-2) clusters (monomer and dimer) adsorbed on Au(111) are used to model the structures with highly undercoordinated Au atoms (CN ) 3 and 4, respectively) on a flat Au surface. The calculated adsorption energies of CO and O2 on these structures are listed in Table 2, together with the key structural parameters, and some results are also plotted in Figure 7. We found that, similar to Au-wire/Au(111), the Au2/Au(111) can strongly adsorb O2, which is consistent with the calculation results of Metiu and co-workers.36 For Au1/Au(111), despite the very high degree of unsaturation, it is still not efficient to O2 adsorption. This could be due to the fact that strong O2 binding with Au requires that the HOMO of Au and 2π* orbital of O2 can overlap well, and in structures such as Au2/Au(111), two protruding Au adatoms favor the O2 molecule, adopting a parallel orientation to bind with both of them, ensuring that the 2π* orbital of O2 overlaps well with the Au HOMO. However, in Au1/Au(111), an isolated Au adatom can provide only one anchoring point for an O2 molecule adsorption in a tilted orientation, and thus is unable to facilitate the electron transfer and orbital overlap between the Au and O2 molecules. Combining these results with those of Au(111), Au(211), and Au-wire/Au(111) (see Figure 8), we can see the following: (i) CO adsorption is not so sensitive to the structural features, and the adsorption energy

Figure 7. Correlation between CO/O2 adsorption energies and the d-band center (eV) of the Au atom at the reactive site for various systems: (A) Au-wire/Au(111) (Figure 6d); (B) Au-wire/HR-TiO2 (Figure 4d); (C) Au-wire/FR-TiO2 (Figure 5c), (D) flat-Au/FR-TiO2 (Figure 5b), (E) Au(211); and (F) Au(111).

Au(211)

-5

2.083 × 10 1.096 × 10-7 9.654 × 10-1 9.999 × 10-1 3.375 × 10-4 2.193 × 10-8 3.375 × 10-2 1.187 × 10-5 1.243 × 10-3 1.515 × 10-5 Au/TiO2:58 9.5 × 10-2; 2.7 × 10-1 ; pure Au:43 3.4 × 10-2

increases as CN decrease in an approximately linear manner (see Figure 8). (ii) Most of the structures with rather large (CN g 6) or small (CN e 3) coordination numbers do not efficiently adsorb or activate the O2 molecule. (iii) The O2 adsorption energy becomes remarkably enhanced when the CN of Au at the reactive site is 4 or 5. These results suggest that highly unsaturated Au atoms, which position themselves at a proper distance side-by-side, may coact to efficiently adsorb and activate O2 and exhibit remarkable activity for low-temperature CO oxidation. It should be emphasized that although Au with wire and some other similar structures on Au(111) were determined in this work to have high catalytic activity toward CO oxidation, we do not exclude the effect of support in promoting high Au activity. In fact, from our calculation results, we expect that the metal oxide support may have the following roles: (i) Au wetting. Metal oxides, especially when they are covered with a high concentration of oxygen vacancies, can disperse Au atoms in the form of nanoparticles or thin films. (ii) Support-assisted stability. For the active Au wire adsorbed on Au(111) (Figure 6d) and the flat (111)-like film at TiO2(101) (Figure 4c), the average adsorption energy is calculated to be coh coh ) 2.48 eV/atom and as high as Ewire/Au-film/TiO ) Ewire/Au(111) 2 2.73 eV/atom, respectively. This clearly suggests that the formation of Au with a wire structure can be favored on supported thin films. (iii) Template effect. Along the lattice-matched [111j] direction on anatase TiO2(101), Au can form a continuous and ordered structure. However, in the [010] direction, a mismatch or corrugation would occur in the Au film extension. Therefore, when Au atoms are dosed at very thin Au films at TiO2, they are expected to prefer growing along the ordered [111j] direction, thus facilitating the formation of a wire structure. However, when the thickness of the film grows, the template effect would become weak, resulting in the low probability for the wire structure to occur. This point may account for the low activity of thick films and be related to the widely recognized size effect in many supported Au systems.5,7

Figure 8. Variation of CO and O2 adsorption energies with the coordination number of Au atom at the reactive site. (see Table 2) were counted. d is the distance between neighboring active Au atoms.

6130 J. Phys. Chem. C, Vol. 113, No. 15, 2009 (iv) Continuity effect. Due to the lattice mismatch along [010], the continuity of extended Au structures can be broken in this direction. Interestingly, structures with a broken continuity are also characterized by containing isolated Au wires (see Figures 4d and 5c), which exhibit high catalytic activity for CO oxidation. However, such structures with obvious broken continuity give relatively lower stability (see Figure 3) and have the tendency to reconstruct into more stable but less reactive stepped structures (see Figures 4b and 5d). This point could be related to the unsatisfying lifetime of active Au in practical use.2,5 4. Conclusions Various extended Au structures on stoichiometric and halfand fully reduced anatase TiO2(101) surfaces were calculated. The structures can match well with the support along [111j] direction, whereas in the [010] direction, strain occurs due to a lattice mismatch. Compared to the stoichiometric surface, halfand full-reduced ones have a stronger capacity to disperse Au atoms and favor the structures in a preferential 2D model. In particular, the fully reduced surface is the most likely to be wetted by Au atoms, since more oxygen vacancies are available. For the average adsorption energies of Au atoms adsorbed at the three surfaces, they reached the local maximum when a compact and flat film occurred. Supported Au thin films with different structures were investigated in detail, and their activities toward CO oxidation were explored. The results indicate that O2 can hardly adsorb on the supported flat Au(111)-like film, regardless of being at Auδ- or Au0. Although stepped structures have enhanced capacity to adsorb CO at the edge sites, they still cannot efficiently adsorb or activate O2 molecules. Interestingly, Au film with a wire structure can adsorb both CO and O2 rather strongly. The characteristic structures were also calculated in nonsupported Au systems, and similar results were obtained. In general, we found that O2 adsorption energy shows a remarkable enhancement when the coordination number of Au at the reactive site is 4 or 5 and the structure takes a dimer or wire configuration. Detailed kinetic analysis of the catalytic activity of Au with wire structure in CO oxidation gave a turnover frequency of 4.4 × 10-1 s-1, largely consistent with the experimental value of CO oxidation in Au/TiO2 and pure Au systems. Electronic analysis was also performed to shed light on the physical origin of the active Au structures, and the d-band center of the Au atom at the reactive site was determined to account for the unusual activity of the wire-structured film. Finally, the support effect was discussed. A reduced TiO2 surface can facilitate Au adsorption as thin film as well as stabilize the formation of a wire structure; and supports, such as anatase TiO2(101), could also serve as a template for the preferential arrangement of Au atoms in the active wire structure. Therefore, it can be expected that Au wire structures identified to be important in this work would have some feasibility to occur at such oxide support under specific conditions. Acknowledgment. This work is financially supported by the National Basic Research Program (2004CB719500), InternationalScienceandTechnologyCooperationProgram(2006DFA42740), the 111 Project (B08021), and the National Natural Science Foundation (20703017) of China. X.Q.G also thanks ECUST for start-up funding (YJ0142142).

Wang et al. Supporting Information Available: Kinetic derivation of the reaction rate for CO2 formation (PDF). This material is available free of charge via the Internet at http://pubs.acs.org. References and Notes (1) Haruta, M.; Yamada, N.; Kobayashi, T.; Iijima, S. J. Catal. 1989, 115, 301. (2) Haruta, M. Catal. Today 1997, 36, 153. (3) Haruta, M. CATTECH 2002, 6, 102. (4) Bamwenda, G. R.; Tsubota, S.; Nakamura, T.; Haruta, M. Catal. Lett. 1997, 44, 83. (5) Valden, M.; Lai, X.; Goodman, D. W. Science 1998, 281, 1647. (6) Meier, D. C.; Goodman, D. W. J. Am. Chem. Soc. 2004, 126, 1892. (7) Chen, M. S.; Goodman, D. W. Science 2004, 306, 252. (8) Chen, M.; Cai, Y.; Yan, Z.; Goodman, D. W. J. Am. Chem. Soc. 2006, 128, 6341. (9) Bondzie, V.; Parker, S.; Campbell, C. Catal. Lett. 1999, 63, 143. (10) Remediakis, I. N.; Lopez, N.; Nørskov, J. K. Angew. Chem., Int. Ed. 2005, 44, 1824. (11) Lopez, N.; Janssens, T. V. W.; Clausen, B. S.; Xu, Y.; Mavrikakis, M.; Bligaard, T.; Nørskov, J. K. J. Catal. 2004, 223, 232. (12) Mavrikakis, M.; Stoltze, P.; Nørskov, J. K. Catal. Lett. 2000, 64, 101. (13) Lopez, N.; Nørskov, J. K. J. Am. Chem. Soc. 2002, 124, 11262. (14) Lemire, C.; Meyer, R.; Shaikhutdinov, S.; Freund, H.-J. Angew. Chem., Int. Ed. 2004, 43, 118. (15) Shaikhutdinov, S. K.; Meyer, R.; Naschitzki, M.; Ba¨umer, M.; Freund, H.-J. Catal. Lett. 2003, 86, 211. (16) Yoon, B.; Ha¨kkinen, H.; Landman, U.; Wo¨rz, A. S.; Antonietti, J. M.; Abbet, S.; Judai, K.; Heiz, U. Science 2005, 307, 403. (17) Ha¨kkinen, H.; Abbet, S.; Sanchez, A.; Heiz, U.; Landman, U. Angew. Chem., Int. Ed. 2003, 42, 1297. (18) Liu, Z. -P.; Hu, P.; Alavi, A. J. Am. Chem. Soc. 2002, 124, 14770. (19) Liu, Z. -P.; Gong, X. -Q.; Kohanoff, J.; Sanchez, C.; Hu, P. Phys. ReV. Lett. 2003, 91, 266102. (20) Waters, R. D.; Weimer, J. J.; Smith, J. E. Catal. Lett. 1995, 30, 181. (21) Liu, Z. -P.; Jenkins, S. J.; King, D. A. Phys. ReV. Lett. 2005, 94, 196102. (22) Idakiev, V.; Tabakova, T.; Naydenov, A.; Yuan, Z. Y.; Su, B. L. Appl. Catal., B 2006, 63, 178. (23) Tabakova, T.; Idakiev, V.; Tenchev, K.; Boccuzzi, F.; Manzoli, M.; Chiorina, A. Appl. Catal., B 2006, 63, 94. (24) Fu, Q.; Saltsburg, H.; Flytzani-Stephanopoulos, M. Science 2003, 301, 935. (25) Sinha, A. K.; Seelan, S.; Tsubota, S.; Haruta, M. Top. Catal. 2004, 29, 95. (26) Sinha, A. K.; Seelan, S.; Tsubota, S.; Haruta, M. Angew. Chem., Int. Ed. 2004, 43, 1546. (27) Hayashi, T.; Tanaka, K.; Haruta, M. J. Catal. 1998, 178, 566. (28) Nijhuis, T. A. R.; Visser, T.; Weckhuysen, B. M. Angew. Chem., Int. Ed. 2005, 44, 1115. (29) Okumura, M.; Coronado, J. M.; Soria, J.; Haruta, M.; Conesa, J. C. J. Catal. 2001, 203, 168. (30) Liu, H.; Kozlov, A. I.; Kozlova, A. P.; Shido, T.; Asakura, K.; Iwasawa, Y. J. Catal. 1999, 185, 252. (31) Wang, C. -M.; Fan, K. -N.; Liu, Z. -P. J. Am. Chem. Soc. 2007, 129, 2642. (32) Molina, L. M.; Hammer, B. Phys. ReV. Lett. 2003, 90, 206102. (33) Molina, L. M.; Rasmussen, M. D.; Hammer, B. J. Chem. Phys. 2004, 120, 7673. (34) Herna´ndez, N. C.; Sanz, J. F.; Rodriguez, J. A. J. Am. Chem. Soc. 2006, 128, 15600. (35) Rashkeev, S. N.; Lupini, A. R.; Overbury, S. H.; Pennycook, S. J.; Pantelides, S. T. Phys. ReV. B 2007, 76, 035438. (36) Mills, G.; Gordon, M. S.; Metiu, H. J. Chem. Phys. 2003, 118, 4198. (37) Guzman, J.; Gates, B. C. J. Am. Chem. Soc. 2004, 126, 2672. (38) Guzman, J.; Carrettin, S.; Corma, A. J. Am. Chem. Soc. 2005, 127, 3286. (39) Fu, L.; Wu, N. Q.; Yang, J. H.; Qu, F.; Johnson, D. L.; Kung, M. C.; Kung, H. H.; Dravid, V. P. J. Phys. Chem. B 2005, 109, 3704. (40) Hutchings, G. J.; Hall, M. S.; Carley, A. F.; Landon, P.; Solsona, B. E.; Kiely, C. J.; Herzing, A.; Makkee, M.; Moulijn, J. A.; Overweg, A.; Fierro-Gonzalez, J. C.; Guzman, J.; Gates, B. C. J. Catal. 2006, 242, 71. (41) Matthey, D.; Wang, J. G.; Wendt, S.; Matthiesen, J.; Schaub, R.; Lægsgaard, E.; Hammer, B.; Besenbacher, F. Science 2007, 315, 1692. (42) Zielasek, V.; Ju¨rgens, B.; Schulz, C.; Biener, J.; Biener, M. M.; Hamza, A. V.; Ba¨umer, M. Angew. Chem., Int. Ed. 2006, 45, 8241. (43) Xu, C.; Su, J.; Xu, X.; Liu, P.; Zhao, H.; Tian, F.; Ding, Y. J. Am. Chem. Soc. 2007, 129, 42.

Activity of Gold in Low-Temperature CO Oxidation (44) (a) Sault, A. G.; Madix, R. J.; Campbell, C. T. Surf. Sci. 1986, 169, 347. (b) Saliba, N.; Parker, D. H.; Koel, B. E. Surf. Sci. 1998, 410, 270. (c) Hammer, B.; Nørskov, J. K. Nature 1995, 376, 238. (45) Paker, D. H.; Koel, B. E. J. Vac. Sci. Technol. A 1990, 8, 2585. (46) Stiehl, J. D.; Kim, T. S.; McClure, S. M.; Mullins, C. B. J. Am. Chem. Soc. 2004, 126, 13574. (47) Xu, Y.; Mavrikakis, M. J. Phys. Chem. B 2003, 107, 9298. (48) Gong, X.-Q.; Liu, Z.-P.; Raval, R.; Hu, P. J. Am. Chem. Soc. 2004, 126, 8. (49) Falsig, H.; Hvolbaek, B.; Kristensen, I. S.; Jiang, T.; Bligaard, T.; Christensen, C. H.; Norskov, J. K. Angew. Chem., Int. Ed. 2008, 47, 4835. (50) Gong, X. Q.; Selloni, A.; Batzil, M.; Diebold, U. Nat. Mater. 2006, 5, 665. (51) Diebold, U. Surf. Sci. Rep. 2003, 48, 53. (52) Yan, W.; Chen, B.; Mahurin, S. M.; Schwartz, V.; Mullins, D. R.; Lupini, A. R.; Pennycook, S. J.; Dai, S.; Overbury, S. H. J. Phys. Chem. B 2005, 109, 10676. (53) Gong, X.-Q.; Selloni, A.; Dulub, O.; Jacobson, P.; Diebold, U. J. Am. Chem. Soc. 2008, 130, 370.

J. Phys. Chem. C, Vol. 113, No. 15, 2009 6131 (54) Perdew, J. P.; Burke, K.; Ernzerhof, M. Phys. ReV. Lett. 1996, 77, 3865. (55) Baroni, S.; De Gironcoli, S.; Dal Corso, A.; Giannozzi, P. QuantumEspresso; http://www.democritos.it. (56) Vanderbilt, D. Phys. ReV. B 1990, 41, 7892. (57) Mills, G.; Gordon, M. S.; Metiu, H. Chem. Phys. Lett. 2002, 359, 493. (58) Kung, H. H.; Kung, M. C.; Costello, C. K. J. Catal. 2003, 216, 425. (59) CRC Handbook of Chemistry and Physics, 79th ed.; CRC Press: BocaRaton, FL, 1998-1999. (60) Kim, J.; Samano, E.; Koel, B. E. J. Phys. Chem. B 2006, 110, 17512. (61) To make a comparison, we calculated the barriers for CO* + O2* f OOCO* + O* on several other structures: Au(111), 0.46 eV; Au(211), 0.65 eV. (62) Newns, D. M. Phys. ReV. 1969, 178, 1123. (63) Hammer, B.; Norskov, J. K. AdV. Catal. 2000, 45, 71.

JP810608C