DFT study of CO oxidation on Cu₂O–Au interfaces at Au–Cu alloy surfaces

Abstract

Although Au–Cu alloy nanoparticles on inert substrates show high activity for catalyzing CO oxidation, the corresponding catalytic mechanism is not clear. To clarify the mechanism of this alloy catalysis method, CO oxidation reactions on Au–Cu alloy surfaces with different surface oxidation states are studied via density functional theory simulations. The simulation results indicate that on AuCu(111) and Cu₂O/Au₃Cu(111), CO and O₂ cannot move together for reactions since they are adsorbed on separate Cu sites. On Cu₂O–Au/Au₃Cu(111), O₂ prefers to be located at the Cu hollow sites near Cu₂O–Au interfaces. When CO diffuses to its neighboring Au sites, they can easily combine to generate CO₂, for which the reaction barriers are no more than 0.42 eV. The Au and Cu synergetic effect for catalyzing CO oxidation can be realized on Cu₂O–Au interfaces at Au–Cu nanoparticle surfaces.

---

Introduction

Gold was regarded as an inert metal until Haruta et al. found that Au nanoparticles dispersed on oxide substrates, such as TiO₂ and Fe₂O₃, show high catalytic activity for low temperature CO oxidation.¹ On bulk gold surfaces, O₂ adsorption is weak and corresponding oxygen dissociation barriers are higher than 1 eV,^2,3 which makes further CO oxidation impossible. When CO and O₂ react on Au/TiO₂ catalysts, O₂ molecules are strongly adsorbed on the Au nanoparticle–TiO₂ substrate circumferential interfaces.^4,5 Via this interfacial adsorption mode, activated O₂ molecules can easily react with the CO molecules adsorbed on neighboring Au sites.^4,5 However, this synergetic effect in catalysis relies on the type of substrates. The catalytic activity for CO oxidation of the Au/SiO₂ catalyst is significantly reduced because inert SiO₂ substrates cannot offer the strong adsorption for O₂ molecules.⁶

Alloying with a second metal is an effective way to obtain active Au-based binary nanoparticles on inert supports. Recently, Au, Cu and Au–Cu binary nanoparticles (dimensional sizes of 2–5 nm) on inert SBA-15 mesoporous silica substrates have been synthesized for catalyzing CO oxidation.⁷ Cu/SBA-15 is inactive below 100 °C and Au/SBA-15 only possesses moderate activity at room temperature. However, CO conversion remains at 100% from room temperature (RT) to 300 °C on Au–Cu/SBA-15. In addition, the turnover frequencies of CO oxidation on Au–Cu/SBA-15 are three times larger than those on Au/MCM-41 mesoporous silica.⁸ When a large amount of H₂ is introduced, the CO conversion rates on Au–Cu/SBA-15 reach 60–80% at RT. As the reaction temperature increases to 80 °C (operational temperature for H₂ fuel cells), the CO conversion rates are still 50–70%, whereas the corresponding values only are 20% and 0% for Au/SBA-15 and Cu/SBA-15.

Structural changes in Au–Cu nanoparticles (Au/Cu atomic ratio 1/1) during the catalytic process of CO oxidation are observed via in situ techniques.⁹ The Au–Cu binary particles obtained by calcination have a gold core and a Cu_xO shell. To obtain active Au–Cu particles, a subsequent reduction treatment is carried out in H₂ at 550 °C. After this reduction treatment, Cu cations are reduced to alloy with the interior Au atoms and the structure of particles reverts to the Au₃Cu intermetallic phase. When the reduced Au–Cu nanoparticles are used to catalyze CO oxidation in the temperature range of RT to 300 °C, their Au₃Cu structure hardly changes. In situ XANES results indicate that some Cu atoms are oxidized from Cu⁰ to Cu⁺ and Cu²⁺, whereas Cu₂O and CuO structures are absent in XRD measurements.

These Cu cations may come from surface oxidation during the catalytic process. The reaction temperature for CO oxidation is much lower than that for nanoparticle calcination.⁷ At this moderate temperature range (RT to 300 °C), Au–Cu nanoparticles are not totally covered by Cu_xO oxide layers and there are still Au sites on their surfaces. Infrared spectroscopy measurements indicate that after O₂ introduction, CO adsorption on Au sites of Au–Cu nanoparticles is decreased dramatically, whereas CO adsorption on Cu sites of these particles remains unchanged.⁹ However, the presence of gold improves the oxidation resistance of copper. Extended X-ray absorption fine structure (EXAFS) measurements show that there are one and a half oxygen neighbors for each Cu cation on average on the surface of the Au–Cu nanoparticles,¹⁰ which proves that Cu₂O constitutes the majority of the surface oxides. Cu₂O is an incomplete oxide, on which O₂ can be adsorbed.

It has been reported that Cu₂O layers can grow epitaxially on the surface of Au–Cu particles with small lattice mismatch.¹¹ In addition, Cu₂O–Au in-plane coherent interfaces can also be formed along the close-packed crystal directions.^11,12 After the surface oxidation of Au–Cu nanoparticles, coherent linear interfaces are generated between the Cu₂O and Au surface layers. The CO oxidation reactions may just occur on these Cu₂O–Au interfaces.^13,14

In this work, CO + O₂ reactions on Cu₂O–Au interfaces at surfaces of the Au₃Cu alloy are simulated by the density functional theory (DFT) method. In addition, the bond length, charge transfer (CT) and partial density of states (PDOS) of some adsorption and reaction states during the catalytic process are also analyzed for understanding the essence of these chemical reactions. For comparison, the adsorption status and further possible reaction of CO + O₂ on AuCu and Cu₂O/Au₃Cu alloy surfaces are investigated to explain why unoxidized and totally oxidized Au–Cu nanoparticles are inert for CO oxidation.

Simulation details

In the first-principles DFT^15,16 simulations, the DMol3 module^17,18 was used for the geometric optimization, reaction process imitation and analysis of CT and PDOS. We employed the generalized gradient approximation (GGA) with the Perdew–Burke–Ernzerhof functional (PBE)¹⁹ to describe exchange and correlation effects. During the geometric optimization process, the energy convergence, maximum force and maximum distance were 5.0 × 10⁻⁴ eV, 0.1 eV Å⁻¹ and 5.0 × 10⁻³ Å, respectively. In the electronic settings, to reduce the computational cost, the DFT semi-core pseudo-potentials (DSPP) method²⁰ was used to replace core electrons by a single effective potential and introduce some degree of relativistic correction into the core. The self-consistent field (SCF) tolerance value was 3.0 × 10⁻⁴ eV and the double numerical plus d-functions (DND) were chosen as the basis set.¹⁷ To speed up convergence, a thermal smearing of 0.1 eV was applied to the orbital occupation. The calculations were spin unrestricted. In the simulation of searching for transition states (TS), the calculation initially performed a linear synchronous transit (LST)²¹ maximum, which was followed by energy minimization in directions conjugated to the reaction pathway. TS approximation obtained via LST/optimization was then used to perform a quadratic synchronous transit (QST) maximization to determine the transitional states more accurately. The convergence tolerance of the root mean square (RMS) force was 0.15 eV Å⁻¹ and the maximum number for a QST step was set as 5. For the CT properties, atom charges would be calculated via the Hirshfeld population analysis.²² For PDOS calculations, the empty bands were chosen as 12.

The Au–Cu nanoparticles with dimensional sizes of 2–5 nm usually have polyhedral constructions, such as cubo-octahedrons, icosahedrons and decahedrons.^23,24 For these small particles, close-packed facets still occupy the majority of surfaces.^25,26 The atom arrangement of these facets is the same as that of bulk (111) surfaces. Therefore, four layers of (111) slabs with periodicity of (4 × 6) were built for representing the surfaces of nanoparticles [the in-plane lattice directions are 〈110〉]. When the slabs were used for the geometric optimization and modeling reaction processes, the underneath two atomic layers were fixed to simulate the interior of Au–Cu alloys. A vacuum of 18 Å was added along the normal directions of the slabs. K-points were chosen as 3 × 2 × 1, for which the actual spacing is 0.03, 0.03 and 0.04 Å⁻¹ in the three lattice directions. First, two extreme cases are considered. When Au–Cu nanoparticles (Au/Cu atomic ratio 1/1) undergo no Cu oxidation and Au surface enrichment, the atom packing of their surface facets should be similar to AuCu(111), which is shown in Fig. 1(a). However, when all Cu atoms in Au–Cu nanoparticles are oxidized and enriched at surfaces via calcination at high temperature, the atom packing of their surface facets should be similar to CuO(111) and Cu₂O(111). In contrast to inert CuO surfaces, Cu₂O surfaces possess active sites for adsorbing reactants.²⁷ The geometric structure of Cu₂O(111) is shown in Fig. 1(b). Differentiating the oxidization states, there are two Cu cation sites at Cu₂O(111) surfaces: the oxygen coordinatively saturated (Cu_CSA) and unsaturated (Cu_CUS) surfaces.²⁸ Under the Cu surface atom layer, there are two oxygen atom sub-surface layers; one layer binds with Cu_CSA sites and the other binds with Cu_CUS sites.

Fig. 1 Geometric configurations of (a) AuCu(111) surfaces, (b) Cu₂O(111) surfaces, (c) Cu₂O/Au₃Cu(111) surfaces and (d) Cu₂O–Au/Au₃Cu(111) surfaces.

As mentioned above, AuCu nanoparticles maintain the Au₃Cu structure when they are only partially oxidized. In this case, part of their surface is a single layer of Cu₂O oxide growing on the Au₃Cu core. Cu₂O/Au₃Cu(111) represents this epitaxial system, which is shown in Fig. 1(c). In Cu₂O/Au₃Cu(111), the oxygen atom sub-surface layer connecting with Cu_CUS sites is absent since Au–O bonding is forbidden.^9,10 However, the oxygen atom sub-surface layer connected with Cu_CSA sites can be thermodynamically stabilized via binding with Cu atoms in the Au₃Cu alloy. Because of this oxygen deficiency on Cu₂O/Au₃Cu interfaces, the surface Cu_CUS sites would change to oxygen uncoordinated ones, which are defined as Cu_UNC. This surface oxide structure is similar to the construction of an Ag₂O oxide layer on Ag surfaces.²⁹ Although the geometric structures of surface oxides are similar to those of corresponding bulk oxides, their electronic properties may be quite different because the surface oxides seem like periodic on-surface and sub-surface oxygen adsorption on metal surfaces. The oxygen adsorption and Cu₂O surface oxide formation on Cu(111) and their electronic properties have been studied in detail via DFT-GGA simulations.^30,31 Therefore, it is appropriate to investigate the Cu₂O surface oxide on Au₃Cu(111) by the same method.

There are four types of coherent interfaces between Cu₂O and Au on Au₃Cu(111), which differentiate the boundary Cu and O sites (Fig. 2). For all of them, the in-plane linear interfaces are designed to lie along the close-packed 〈110〉 direction according to experimental results.^11,12 Cu_CSA sites can always connect with Au sites on interfaces because they occupy three quarters of the whole Cu surface atom layer. There are no Au–O bonds at Cu₂O–Au interfaces, which means that interfacial Cu_CSA sites lose one coordinated oxygen atom and are converted to new oxygen coordinatively unsaturated sites (Cu_NCUS). In Fig. 2(a and b), Cu_UNC and Cu_NCUS sites are both on Cu₂O–Au interfaces. The difference between them is that the sub-surface oxygen is located near interfaces for the former and the on-surface oxygen is located near interfaces for the latter. In Fig. 2(c and d), only Cu_NCUS sites are on the Cu₂O–Au interfaces and the difference between them is as the same as that in Fig. 2(a and b). The total cohesive energy difference between these slabs and that of Cu₂O/Au₃Cu(111) and Au/Au₃Cu(111), ΔE_slab, derives from the generation of Cu₂O–Au interfaces. It is impossible to describe the interfacial energy of linear Cu₂O–Au interfaces in units of Joules per square meter. Therefore, the interfacial energy, γ_in, is calculated in electron volts per atom, and is calculated via the formula , where N_in is the total number of atoms at the Cu₂O–Au interfaces in a slab. γ_in ≈ 0.65 eV per atom for Cu₂O–Au interfaces in Fig. 2(a and b) and γ_in ≈ 0.70 eV per atom for Cu₂O–Au interfaces in Fig. 2(c and d). The γ_in values are similar to the surface energy values of Au(111).³² The high γ_in values of Cu₂O–Au interfaces may be due to the lattice oxygen deficiency of the Cu_NCUS sites. Nevertheless, if the Cu₂O–Au interfaces do not form, the edges of Au and Cu₂O are exposed to surfaces where the surface energy values are extremely large. To catalyze CO oxidation at Cu₂O–Au interfaces, active Cu_UNC sites are necessary. Because of the higher γ_in values and absence of Cu_UNC sites, the Cu₂O–Au interfaces in Fig. 2(c and d) are excluded. In Fig. 2(b), on-surface oxygen atoms are adjacent to interfaces, which may hinder the adsorption of reactants. Therefore, the Cu₂O–Au/Au₃Cu(111) slab with both Cu_UNC and Cu_NCUS sites on Cu₂O–Au interfaces and sub-surface oxygen atoms adjacent to Cu₂O–Au interfaces [Fig. 2(a)] are chosen as the catalytic substrate for the CO oxidation reaction, which is also shown in Fig. 1(d).

Fig. 2 The four possible geometric configurations of Cu₂O–Au/Au₃Cu(111). (a) Cu_UNC and Cu_NCUS sites on interfaces and O_sub-surface site near interfaces; (b) Cu_UNC and Cu_NCUS sites on interfaces and O_on-surface site near interfaces; (c) Cu_NCUS site on interfaces and O_sub-surface site near interfaces; (d) Cu_NCUS site on interfaces and O_on-surface site near interfaces.

Results and discussion

The adsorption status and reaction of CO oxidation on AuCu and Cu₂O/Au₃Cu(111)

The stable and meta-stable adsorption positions of reactants CO, O₂ and O on AuCu and Cu₂O/Au₃Cu(111) are shown in Fig. 3 and the corresponding adsorption energy values, E_ad, are listed in Table 1. In general, CO is vertically adsorbed on the top sites of metal surface atoms, whereas O is adsorbed in the hollow sites of metal surfaces. O₂ is adsorbed on metal surfaces in three states: top-bridge-top (tbt), top-hollow-bridge (thob), and bridge-hollow-bridge (bhob).³³ For fcc and hcp hollows, the thob state can be sub-divided into the top-fcc-bridge (tfb) and top-hcp-bridge (thb) states. Similarly, the bhob state can be sub-divided into the bridge-fcc-bridge (bfb) and bridge-hcp-bridge (bhb) states. In addition, the Au and Cu sites should be distinguished for Au–Cu alloy surfaces.

Fig. 3 Possible adsorption positions of reactants O₂, CO and O on (a) AuCu(111) and (b) Cu₂O/Au₃Cu(111) surfaces.

Table 1 E_ad values of reactants CO, O₂ and O on AuCu(111), Cu₂O/Au₃Cu(111) and Cu₂O–Au/Au₃Cu(111)

AuCu(111)				Cu2O–Au/Au3Cu(111)
Reactants	Positions	States	Ead (eV)	Reactants	Positions	States	Ead (eV)
CO-I	Cu	Top	−0.83	CO-I	Au	Top	−0.28
CO-II	Au	Top	−0.32	CO-II	CuNCUS	Top	−1.03
O2-I	Cu2	Tilted tbt	−0.39	CO-III	CuUNC	Top	−0.75
O2-II	AuCu2	tfb-Au top	−0.19	O2-I	Cu3	tfb-CuUNC top	−0.44
O2-III	AuCu2	thb-Au top	−0.13	O2-II	Cu3	tfb-CuNCUS top	−0.73
O-I	AuCu2	fcc hollow	−4.40	O2-III	Cu3	tfb-CuCSA top	−0.49
O-II	AuCu2	hcp hollow	−4.27	O2-IV	Cu3	bhb-CuUNC top	−0.40
O-III	Au2Cu	fcc hollow	−3.84	O2-V	Au2Cu	tfb-CuUNC top	−0.05
O-IV	Au2Cu	hcp hollow	−3.73	O2-VI	AuCu2	thb-Au top	−0.21
Cu2O/Au3Cu(111)				O-I	Cu3	fcc hollow	−4.25
CO-I	CuUNC	Top	−1.12	O-II	AuCu2	hcp hollow	−4.26
O2-I	Cu3	tfb-CuUNC top	−0.49	O-III	Au2Cu (CuUNC)	fcc hollow	−3.97
O2-II	Cu3	tfb-CuCSA top	−0.29
O2-III	Cu3	bhb-CuUNC top	−0.24	O-IV	Au2Cu (CuNCUS)	fcc hollow	−4.00
O-I	Cu3	fcc hollow	−4.04

---

In Fig. 3(a), CO can be adsorbed on the Au site of AuCu(111) (CO-II) with E_ad = −0.32 eV. In comparison, CO binds strongly to the Cu site (CO-I) due to its low adsorption energy of E_ad = −0.83 eV. The most stable O₂ adsorption position is the Cu₂ bridge (O₂-I). In this adsorption state, although the two oxygen atoms bind separately with two Cu sites, the O–O bond is not parallel to the Cu₂ bridge (the angle of interaction is about 40°), which differs from the tbt state. Therefore, this new adsorption state is the tilted tbt state. For this stable O₂ adsorption position, E_ad = −0.39 eV. When O₂ is placed on the AuCu₂ or Au₂Cu hollows, it will spontaneously move to the Cu₂ bridge. The only exceptions are O₂ adsorption on AuCu₂ hollows in the tfb-Au top and thb-Au top states (O₂-II and O₂-III). For these two meta-stable adsorption states, the E_ad values are −0.19 and −0.13 eV, respectively. In addition, O₂ adsorption on the Au₂ bridge is endothermic. The O atom adsorption is much stronger than that of O₂ molecules because the atomic oxygen has lost the strong O–O bond. The most stable adsorption position for oxygen atoms is the AuCu₂ fcc hollow (O-I) with E_ad = −4.40 eV. Other meta-stable O adsorption positions are AuCu₂ hcp, Au₂Cu fcc and hcp hollows (O-II, O-III and O-IV), for which E_ad values are −4.27, −3.84 and −3.73 eV, respectively. In Fig. 3(b), CO can only be adsorbed on the Cu_UNC site of Cu₂O/Au₃Cu(111) (CO-I) with E_ad = −1.12 eV. The most stable O₂ adsorption position is the Cu₃ fcc hollow site (O₂-I). In the tfb-Cu_UNC top state, E_ad reaches −0.49 eV. At the same adsorption position, the other two meta-stable adsorption states are tfb-Cu_CSA top and bhb-Cu_UNC top (O₂-II and O₂-III), for which the E_ad values are −0.29 and −0.24 eV, respectively. Similar to the CO adsorption, O can only be adsorbed on the Cu₃ fcc hollow (O-I) with E_ad = −4.04 eV.

The possible reaction mechanism of CO oxidation on unoxidized AuCu alloy surfaces is shown in Fig. 4. According to the reactant adsorption results, O₂ can be meta-stably adsorbed on the Cu₂ bridge of AuCu(111) in the tilt tbt state. Even if O₂ molecules are adsorbed at AuCu₂ hollows in the tfb-Au top and thb-Au top states, they can easily diffuse to the Cu₂ bridge with a small diffusion barrier of E_db = 0.15 eV. In Fig. 4(a–c), when the O₂ molecule on the Cu₂ bridge is dissociated to form two O atoms, it needs to overcome the reaction barrier of E_rb = 0.74 eV. The direct O₂ dissociation on AuCu(111) is not easy. Considering the direct reaction of O₂ with CO, they must be co-adsorbed on the neighboring surface sites. At the ordered AuCu alloy surfaces, the Au and Cu sites are next to each other. However, CO prefers to bind with the Cu site of AuCu(111). To react with O₂ on the Cu₂ bridge, CO must move to the Au site between them, as shown in Fig. 4(d–f). A reaction barrier of E_rb = 0.81 eV is calculated for this movement, which means that CO undergoes desorption and re-adsorption during this process. When CO is adsorbed on the Au site with O₂ on its neighboring Cu₂ bridge, their co-adsorption energy increases to E_ad = −0.88 eV as shown in Fig. 4(f) [E_ad = −1.22 eV for the initial state (IS)]. Owing to this meta-stable adsorption state, the reaction tends to be reversible and the corresponding E_rb is only 0.44 eV. Therefore, on AuCu(111) surfaces, it is difficult for CO on the Cu site to move close to O₂ on the Cu₂ bridge for further reactions. In addition, O₂ molecules can still dissociate into O atoms on AuCu(111) because the corresponding E_rb value is similar to that of O₂ dissociation on Pt(111).³³ According to Fig. 4(c), after O₂ dissociation, the two atoms are located at the AuCu₂ and Au₂Cu hollows. The AuCu₂ fcc hollow is the most stable position for O adsorption. O atoms adsorbed on the other three sites can easily diffuse to the AuCu₂ fcc hollow with small E_db values of 0.02–0.25 eV. In Fig. 4(g–i), to react with the O atom on the AuCu₂ fcc hollow, CO on the Cu site must move to the Au site between them. The E_rb for this reaction is as high as 0.81 eV, whereas E_rb of the reverse reaction is only 0.35 eV. Therefore, even if O₂ is dissociated to O atoms on AuCu(111), they still hardly react with CO molecules.

Fig. 4 Illustration of (a)–(c) O₂ dissociation reaction, (d)–(f) CO + O₂ reaction and (g)–(i) CO + O reaction on AuCu(111) surfaces.

Because the surface energy of Au is smaller than that of Cu,³² gold atoms are usually enriched at the unoxidized Au–Cu alloy surfaces.³⁴ The Au surface enrichment has been neglected in this AuCu(111) slab. If the Au enrichment is considered, some local parts of the AuCu alloy surface may show an ordered Au₃Cu atom arrangement, for which the neighbors of Cu sites are only Au sites. On this Au₃Cu surface layer, Cu₂ bridges and AuCu₂ hollows are absent. O₂ adsorption on the Au₂Cu hollows and AuCu bridges is very weak because the corresponding E_ad values are higher than −0.10 eV. The gold surface enrichment leads to weaker O₂ adsorption, which cannot change the fact that CO oxidation is impossible on unoxidized Au–Cu alloy surfaces. In addition, the simulation of CO oxidation on Cu/Au(100) surfaces has been performed by the DFT-GGA method.³⁵ The E_rb of the CO + O reaction on Cu/Au(100) reaches 0.6 eV,³⁵ which is a relatively high value. Therefore, the catalysis of CO oxidation using Au–Cu alloys is still difficult even if Cu is totally segregated to their unoxidized surfaces.

The possible reaction mechanism of CO oxidation on the Cu₂O surface oxide of Au–Cu alloys is shown in Fig. 5. One of the most common CO oxidation reaction mechanisms on reducible oxides or metal-support interfaces is that the oxides provide their lattice oxygen to react with CO.³⁶ Based on this consideration, the possibility of the CO reaction with the on-surface oxygen of Cu₂O/Au₃Cu(111) is investigated first. In Fig. 5(a–c), the transitional state of this CO + O_on-surface reaction was not observed via our simulations. The difficulty for CO reactions with lattice oxygen may be due to the strong bonding between the on-surface oxygen anions and the Cu cations underneath. If an on-surface oxygen anion is removed from Cu₂O/Au₃Cu(111), there is an energy cost of 5.71 eV. For comparison, oxygen atom adsorption on Cu(111) has been also calculated, for which E_ad = −4.91 eV. The lattice Cu–O bonds in Cu₂O surfaces are stronger than the adsorption bonds of oxygen atoms on Cu surfaces. Although oxygen adsorption bonds are weaker, the E_rb value of the CO + O reaction on Cu(111) is calculated to be as high as 0.80 eV. Based on this result, the E_rb value of the CO + O_on-surface reaction on Cu₂O/Au₃Cu(111) should be higher than 0.80 eV. In addition, considering the Eley–Rideal mechanism, the possibility of the gas phase CO molecule reaction with the on-surface oxygen of Cu₂O/Au₃Cu(111) is studied. Similarly, no transitional state is observed. Therefore, the direct reaction of CO with the lattice oxygen of the Cu₂O surface oxide is impossible.

Fig. 5 Illustration of (a)–(b) CO + O_on-surface reaction, (c)–(e) O₂ dissociation reaction and (f) CO + O₂ reaction on Cu₂O/Au₃Cu(111) surfaces.

According to the reactant adsorption results, O₂ can be adsorbed meta-stably on the Cu₃ fcc hollow of Cu₂O/Au₃Cu(111) in the tfb-Cu_UNC top state. Even if O₂ molecules are meta-stably adsorbed on the Cu₃ fcc hollow in the tfb-Cu_CSA top state and the Cu₃ hcp hollow in the bhb-Cu_UNC top state, they can easily change to the most stable adsorption state with small diffusion barriers of E_db = 0.18 and 0.03 eV, respectively. In Fig. 5(c–e), when O₂ on the Cu₃ fcc hollow is dissociated to form two O atoms, it needs to overcome the reaction barrier E_rb = 0.90 eV. The direct dissociation of O₂ on Cu₂O/Au₃Cu(111) is not possible. When CO and O₂ are co-adsorbed on the Cu₂O oxide, they cannot be neighbors, as shown in Fig. 5(f). To react with O₂ on the Cu₃ fcc hollow, CO on the Cu_UNC site must move to the Cu_CSA site between them. However, when CO is placed on the inert Cu_CSA site, it spontaneously leaves Cu₂O/Au₃Cu(111). In addition, the movement of O₂ to the neighboring sites of CO is impossible because O₂ cannot undergo exothermic adsorption once it has left the Cu_UNC site. CO and O₂ have no opportunity to meet directly and react on Cu₂O/Au₃Cu(111). As mentioned above, Cu₂O/Au(111) represents the surfaces of the calcined Au–Cu particles. Moreover, the surface structures of Cu₂O/Au(111) and Cu₂O/Au₃Cu(111) should be similar because Au and Au₃Cu have the same crystal structure and similar lattice constants. These simulation results can explain why the calcined Au–Cu nanoparticles are inert for CO oxidation.

Moreover, the O₂ dissociation and CO oxidation on Cu₂O(111) have also been studied using the same DFT-GGA method.^27,37 The results show that reaction barriers for the direct O₂ dissociation, the CO reaction with the lattice oxygen, and the direct CO reaction with O₂, are all higher than 1 eV.^27,37 Therefore, CO oxidation reactions cannot be carried out on both the surface Cu₂O oxides of Au–Cu alloys and bulk Cu₂O oxides.

Adsorption status and reactions of CO oxidation on Cu₂O–Au/Au₃Cu(111)

The stable and meta-stable adsorption positions of reactants on Cu₂O–Au/Au₃Cu(111) are shown in Fig. 6 and the corresponding adsorption E_ad values are listed in Table 1. In Fig. 6(a), the possible O₂ adsorption positions near or on Cu₂O–Au interfaces are given. As mentioned above, the most stable O₂ adsorption position on Cu₂O/Au₃Cu(111) is the Cu₃ fcc hollow site, at which E_ad is −0.49 eV in tfb-Cu_UNC top state [O₂-I in Fig. 3(b)]. On Cu₂O–Au/Au₃Cu(111), O₂ also prefers to be located at the Cu₃ fcc hollow site near Cu₂O–Au interfaces. When O₂ is adsorbed in the tfb-Cu_UNC top state (O₂-I), its E_ad value is only −0.44 eV. However, in the tfb-Cu_NCUS top state (O₂-II), E_ad for O₂ adsorption can reach −0.73 eV. On Cu₂O/Au₃Cu(111), Cu_UNC is the only active Cu site, which offers strong adsorption for O₂. On Cu₂O–Au/Au₃Cu(111), Cu_UNC and Cu_NCUS are both active Cu sites. When O₂ is located at the Cu₃ fcc hollow site near Cu₂O–Au interfaces in the tfb-Cu_NCUS top state, Cu_UNC and Cu_NCUS may simultaneously offer strong adsorption for O₂, which induces a significant decrease in E_ad. O₂ can also be meta-stably adsorbed on the Cu₃ fcc hollow site in the tfb-Cu_CSA top state (O₂-III), for which E_ad = −0.48 eV. On the Cu₃ hcp hollow, O₂ is meta-stably adsorbed in the bhb-Cu_UNC top state with E_ad = −0.40 eV (O₂-IV). In addition, O₂ can also be located on only the Cu₂O–Au interfaces. On the AuCu₂ hcp hollow, O₂ is meta-stably adsorbed in the thb-Au top state (O₂-VI), for which E_ad = −0.21 eV. On the Au₂Cu fcc hollow, where the Cu belongs to the Cu_UNC sites (O₂-V), O₂ is weakly adsorbed in the tfb-Cu_UNC top state because the corresponding E_ad value is only −0.05 eV. The adsorption positions of CO and O on Cu₂O–Au/Au₃Cu(111) are shown in Fig. 6(b). CO can be meta-stably adsorbed on the Au site (CO-I) with E_ad = −0.28 eV. On the Cu_NCUS and Cu_UNC sites (CO-II and CO-III), the CO adsorption is stronger due to the low E_ad values of −1.03 and −0.75 eV, respectively. The most stable O adsorption position is the AuCu₂ hcp hollow (O-II), on which E_ad = −4.26 eV is determined. On the Cu₃ fcc hollow (O-I), the O adsorption is slightly weaker with E_ad = −4.25 eV. In addition, O atoms can be adsorbed meta-stably on the Au₂Cu fcc hollows at Cu_NCUS and Cu_UNC sites (O-IV and O-III) and the corresponding E_ad values are −4.00 and −3.97 eV, respectively.

Fig. 6 Illustration of possible adsorption positions of reactants (a) O₂ and (b) CO and O on Cu₂O–Au/Au₃Cu(111) surfaces.

Diffuse reflectance infrared Fourier transform (DRIFT) and FT-IR results indicate that although CO adsorption on Cu sites is stronger than that on Au sites of Au–Cu nanoparticles, the corresponding CO adsorption amount on the former is far less than that on the latter.⁹ This may be because surface Cu sites are deficient before the reactions are carried out. Au enrichment on unoxidized Au–Cu alloy surfaces occurs³⁴ because the surface energy of Au is smaller than that of Cu.³² Before the introduction of CO and O₂, Au sites occupy the majority of the surface of the reduced Au–Cu nanoparticles. Therefore, at the beginning of catalytic reactions, CO can bind more readily with Au sites. In the presence of oxygen, Au–Cu nanoparticles are partially oxidized and Cu₂O–Au interfaces are formed between local Cu₂O and Au surface layers. According to this model of reactant adsorption, O₂ prefers to be located at the Cu hollow sites near the Cu₂O–Au interfaces. The diffusion of CO on Au surfaces is very easy. When CO diffuses to the Au sites near the Cu₂O–Au interfaces and encounters O₂ there, they can combine to generate CO₂.

The CO oxidation reaction on Cu₂O–Au/Au₃Cu(111) can represent this interface catalytic process, which is shown in Fig. 7. First, O₂ is stably adsorbed on the Cu₃ fcc hollow near the interfaces in the tfb-Cu_NCUS top state. When O₂ molecules are in other meta-stable adsorption states, they can easily change to the most stable state with small diffusion barriers of 0.03–0.15 eV. Although the O₂ adsorption on Cu₂O–Au/Au₃Cu(111) is strong enough, it still cannot be directly dissociated because the corresponding dissociation barrier is as high as 1.04 eV. When CO diffuses to the Au site next to the Cu₂O–Au interfaces, it still has no electronic interaction with O₂. This co-adsorption state can be recognized as the IS of CO oxidation reactions, which is shown in Fig. 7(a). Subsequently, CO moves to the neighboring Au site of O₂, which only requires a small reaction barrier of E_rb = 0.35 eV to be overcome. In Fig. 7(c), when CO binds with the Au site near interfaces, it leans towards to the O₂ molecule. In this meta-stable intermediate state (MS), the co-adsorption energy of CO and O₂ reaches −1.17 eV, which is lower than that of IS of E_ad = −1.02 eV. Therefore, there is already weak interaction between CO and O₂ in the MS. CO and O₂ then move together to form the meta-stable intermediate product O–O–CO, through which the O–O bond can be activated and easily dissociated.² The reaction barrier of this combination process is only E_rb = 0.35 eV. In Fig. 7(e), E_ad has reduced to −1.73 eV after the formation of O–O–CO, which means that a strong bond is generated between CO and O₂. In this intermediate state, the O–O bond has been activated, which is beneficial for further reactions. In Fig. 7(e and f), the transformation of O–O–CO to CO₂ and O is easy with E_rb = 0.07 eV. After the CO₂ product molecule desorbs, the remaining O atom is adsorbed on the stable AuCu₂ hcp hollow. When another CO molecule diffuses to a neighboring Au site, the molecules can combine to generate CO₂ with a moderate reaction barrier of E_rb = 0.42 eV, as shown in Fig. 7(h–j). In summary, CO oxidation reactions can be easily carried out on Cu₂O–Au interfaces at Au₃Cu surfaces.

Fig. 7 Illustration of (a)–(g) CO + O₂ reaction and (h)–(j) CO + O reaction on Cu₂O–Au/Au₃Cu(111) surfaces.

Cu₂O–Au/Au₃Cu(111) can only represent Cu₂O–Au interfaces located on facets of Au–Cu nanoparticles. In addition to facets, there are still some surface atoms on the edges and vertices of Au–Cu binary particles²⁴ and Cu₂O–Au interfaces may pass through them. It has been demonstrated that these under-coordinated nanoparticle sites can accelerate heterogeneous catalysis.³⁸ On Au–Cu binary nanoparticles, the CO and O₂ molecules adsorbed on edges and vertices should have lower E_ad values than those adsorbed on facets. Via this enhanced reactant adsorption, O₂ activation and further CO oxidation may become easier. In future work, we intend to consider CO oxidation on Au–Cu clusters with different surface oxidation states, which can represent the surfaces of these binary nanoparticles more realistically.

CT of O₂ adsorption and PDOS analysis of CO oxidation on Cu₂O–Au/Au₃Cu(111)

An important premise for CO oxidation on Cu₂O/Au₃Cu(111) is that O₂ prefers to be located at the Cu sites near Cu₂O–Au interfaces rather than those inside the Cu₂O layers. To explain this preference for reactant adsorption, the CT and bond length of O₂ adsorption at the most stable positions of Cu₂O/Au₃Cu(111) and Cu₂O–Au/Au₃Cu(111) are analyzed, which is shown in Fig. 8. When adsorption bonds are generated between O₂ molecules and the underlying Cu₂O surfaces, charges are transferred from the substrates to the reactants and the Coulomb forces between them would determine the bond strength.³⁹ The inter-atomic Coulomb forces are proportional to the absolute value of the product of the positive and negative charges of the atoms and inversely proportional to the bond length. In Fig. 8(a and b), the Cu sites of the Cu₂O surfaces are all positively charged without reactant adsorption. At the interior of the Cu₂O surfaces, the CT values of the Cu_UNC and Cu_CSA sites are 0.06 and 0.18, respectively. For the corresponding interfacial Cu_UNC and Cu_CSA sites, the CT values are almost unchanged. The CT value of the Cu_NCUS sites decreases to 0.09 compared with that of Cu_CSA sites, which is caused by the coordinate oxygen unsaturation of the Cu_NCUS sites. In Fig. 8(c and d), after O₂ is adsorbed on both Cu₂O/Au₃Cu(111) and Cu₂O–Au/Au₃Cu(111), the average CT values of reactants are −0.16 and −0.18, respectively. O₂ can obtain more charge from the Cu sites near Cu₂O–Au interfaces. Nevertheless, the sum of the charge product of the three Cu–O adsorption bonds is similar for O₂ adsorption on Cu₂O/Au₃Cu(111) and Cu₂O–Au/Au₃Cu(111). In contrast, the sum of the bond length of the three Cu–O adsorption bonds calculated is 6.24 Å for O₂ adsorption on Cu₂O/Au₃Cu(111) and 6.17 Å for O₂ adsorption on Cu₂O–Au/Au₃Cu(111). The difference comes from the Cu_NCUS–O bond, which has a bond length of only 1.96 Å. When O₂ is adsorbed on the interfacial Cu₃ fcc hollow in the tfb-Cu_UNC and tfb-Cu_CSA states, the bond lengths of the Cu_NCUS–O bond are 2.06 and 2.09 Å, respectively. This difference in bond length proves our previous assumption that the interfacial Cu_NCUS sites could offer enhanced adsorption for O₂ in the tfb-Cu_NCUS state. Therefore, Cu–O adsorption bonds near the Cu₂O–Au interfaces have similar absolute values of the charge product but shorter adsorption bond lengths compared with those at the interior of Cu₂O surfaces, which causes their large Coulomb forces and low adsorption energy values.

Fig. 8 CT of Cu sites on (a) Cu₂O/Au₃Cu(111) surfaces and (b) Cu₂O–Au/Au₃Cu(111) surfaces. CT of O₂ molecules and its connecting Cu sites and bond length of Cu–O adsorption bonds when O₂ is stably adsorbed on (c) Cu₂O/Au₃Cu(111) surfaces and (d) Cu₂O–Au/Au₃Cu(111) surfaces.

The increase in charge transfer can only ensure the preference of O₂ adsorption near Cu₂O–Au interfaces. The O₂ activation and eventual dissociation requires the supplementary role of CO. The O–O bond of O₂ is generated via extensive orbital hybridization of the 2p electrons. When CO reacts with O₂, the orbital hybridization of the O–O bond may decrease and eventually disappear. To clarify the effect of CO assistance on the O₂ activation and dissociation, the PDOS changes in the 2p electrons of the reactants during the CO oxidation on Cu₂O–Au/Au₃Cu(111) are analyzed, which is shown in Fig. 9. First, 2p electrons of C and O, 3d electrons of Cu, and 5d electrons of Au in all slabs have no spin polarization. In Fig. 9(a), the 2p hybridization peak of CO lies at a lower energy level than that of O₂ in IS because the C–O bond is stronger than the O–O bond. In the first transitional state (TS1), CO has been desorbed from the Au site, which induces the movement of its hybridization peak to a higher energy level. In Fig. 9(c), the hybridization peak of CO is substantially broadened and weakened in the first meta-stable intermediate state (MS1). At the lower energy end of the valence bands, PDOS of CO and O₂ are both slightly increased, which proves the weak interaction of the reactants. Under the influence of CO, the hybridization peak of O₂ also decreases. In the second transitional state (TS2), the original hybridization peak of CO almost disappears. However, the simultaneous increase in PDOS of CO and O₂ becomes more pronounced. In Fig. 9(e), large hybridization peaks are produced between the 2p electrons of CO and O₂ because of the formation of O–O–CO. Before the reaction reaches the meta-stable intermediate state (MS2), the PDOS shapes of the two O atoms belonging to the O₂ molecule are similar. In MS2, PDOS of the O atom connecting the Cu₂O substrates is significantly reduced at the main hybridization peak (energy level of −6.70 eV). In the meta-stable intermediate product, O–O–CO, the electronic hybridization of this O atom only occurs for the O–O bond. Therefore, the reduced participation of its 2p electrons in hybridization indicates that the O–O bond in O–O–CO is weaker compared with that of the original O₂ molecule. In the third transitional state (TS3), PDOS of this O atom is decreased further at the main hybridization peak, which indicates that the O–O bond is almost broken. In Fig. 9(g), the extensive hybridization of the 2p electrons only happens among the other three C and O atoms, which confirms the complete dissociation of O₂ and formation of CO₂ in the final state (FS).

Fig. 9 PDOS of CO and O₂ during their reaction on Cu₂O–Au/Au₃Cu(111) surfaces. (a)–(g) correspond to the reaction states shown in Fig. 7(a–g). The black solid and red short dashed lines represent the C and O atoms in CO, respectively. The blue short dashed and dotted line represents the O atom in O₂, which would form the C–O bond with CO. The green and dashed dotted line represents the O atom in O₂, which would remain on substrates after the production of CO₂.

Discussion of the reaction mechanisms of CO oxidation on Au–Cu nanoparticles

The reaction mechanism of CO oxidation on Au–Cu nanoparticles can be explained based on our simulation results. In contrast to the Cu alloy, O₂ adsorption and CO oxidation on Au nanoparticles only occurs at vertex sites.⁴⁰ However, these under-coordinated sites occupy a very small proportion of the nanoparticle surface,^25,26 which reduces the probability of O₂ adsorption and further reactions with CO. Oxidation resistance experiments indicate that at 110 °C, the transformation of Cu nanoparticles to Cu₂O and CuO aggregates only requires several tens of minutes.⁴⁰ Inert CuO surfaces cannot provide effective adsorption for reactants. CO oxidation cannot be carried out on Cu₂O surfaces because their active Cu_CUS sites are far apart. The oxidation rates of Au–Cu alloy nanoparticles are much lower than those of pure Cu ones.⁴¹ To catalyze CO oxidation on Au–Cu nanoparticles, CO molecules are first adsorbed on their surface Au sites. Owing to the introduction of O₂, some Cu atoms of the particles are oxidized to form numerous local Cu₂O surface oxide layers. Simultaneously, coherent interfaces are generated between Cu₂O and the unoxidized Au layers. O₂ activation and CO oxidation can occur on these Cu₂O–Au interfaces. For traditional nanogold catalysts, interfaces between Au and oxide substrates are used for O₂ activation and CO oxidation. However, for these Au–Cu binary nanoparticles, CO and O₂ can directly react on surfaces, and this does not depend on the substrates. Considering the catalyst design for CO oxidation reactions, O₂ cannot be activated and dissociated on pure Au and Pt nanoparticles. In contrast, Cu, Rh, Ni, Ru, and Co nanoparticles are easily oxidized and lose their surface active sites for O₂ adsorption. Via alloying techniques, binary nanoparticles can avoid being completely oxidized and offer active sites for O₂ adsorption and further CO oxidation.

Conclusions

On AuCu(111), CO is stably adsorbed on the top of the Cu site and O₂ prefers to be located at the Cu₂ edge in the tilted-tbt state. The O₂ dissociation barrier on the unoxidized AuCu alloy surface is as high as 0.74 eV. Because CO and O₂ are adsorbed on the separate Cu sites, they cannot readily combine to generate CO₂ on AuCu(111). On Cu₂O/Au₃Cu(111), CO is strongly adsorbed on the top of the Cu_UNC site and O₂ prefers to be located at the Cu₃ fcc hollow in the tfb-Cu_UNC top state. CO cannot react with the lattice oxygen of the Cu₂O surface oxides. The direct O₂ dissociation on oxidized Au₃Cu alloy surfaces is difficult because of the high dissociation barrier value of 0.90 eV. The reaction of CO + O₂ on Cu₂O/Au₃Cu(111) is impossible as the reactants cannot move together in adsorption states. On Cu₂O–Au/Au₃Cu(111), O₂ prefers to be located near Cu₂O–Au interfaces rather than inside the Cu₂O surfaces. CO can be adsorbed on Au surfaces, moderately. When CO diffuses to the Au site near Cu₂O–Au interfaces, it can easily react with O₂ due to the small reaction barriers of 0.07–0.35 eV. After this reaction, the remaining O atom on the AuCu₂ hcp hollow can also react with CO on the neighboring Au site and the corresponding reaction barrier is 0.42 eV. The Hirshfeld population and bond length analysis indicate that O₂ near Cu₂O–Au interfaces can obtain more charge, transferred from the substrates and shorter Cu–O adsorption bonds, which causes its lower adsorption energy value. In addition, PDOS results show that the electronic hybridization of the O–O bond in O₂ weakens and eventually disappears after it combines with CO to form O–O–CO and CO₂ + O.

Acknowledgements

We acknowledge support by the National Key Basic Research and Development Program (Grant no. 2010CB631001), and the High Performance Computing Center (HPCC) of Jilin University for supercomputer time.

Notes and references

1. M. Haruta, T. Kobayashi, H. Sano and N. Yamada, Chem. Lett., 1987, 2, 405–408.
2. Z. P. Liu, P. Hu and A. Alavi, J. Am. Chem. Soc., 2002, 124, 14770–14779.
3. Y. Xu and M. Mavrikakis, J. Phys. Chem. B, 2003, 107, 9298–9307.
4. I. X. Green, W. Tang, M. Neurock and J. T. Yates Jr, Science, 2011, 333, 736–739.
5. D. Widmann and R. J. Behm, Angew. Chem., Int. Ed., 2011, 50, 10241–10245.
6. S. D. Lin, M. Bollinger and M. A. Vannice, Catal. Lett., 1993, 17, 245–262.
7. X. Liu, A. Wang, X. Wang, C. Y. Mou and T. Zhang, Chem. Commun., 2008, 27, 3187–3189.
8. M. Okumura, S. Nakamura, S. Tsubota, T. Nakamura, M. Azuma and M. Hartua, Catal. Lett., 1998, 51, 53–58.
9. X. Liu, A. Wang, L. Li, T. Zhang, C. Y. Mou and J. F. Lee, J. Catal., 2011, 278, 288–296.
10. J. C. Bauer, D. Mullins, M. Li, Z. Wu, E. A. Payzant, S. H. Overbury and S. Dai, Phys. Chem. Chem. Phys., 2011, 13, 2571–2581.
11. K. Koga and D. Zubia, J. Phys. Chem. C, 2008, 112, 2079–2085.
12. C. H. Kuo, T. E. Hua and M. H. Huang, J. Am. Chem. Soc., 2009, 131, 17871–17878.
13. J. Yin, S. Shan, L. Yang, D. Mott, O. Malis, V. Petkov, F. Cai, M. S. Ng, J. Luo, B. H. Chen, M. Engelhard and C. J. Zhong, Chem. Mater., 2012, 24, 4662–4674.
14. A. Wang, X. Y. Liu, C. Y. Mou and T. Zhang, J. Catal., 2013, 308, 258–271.
15. P. Hohenberg and W. Kohn, Phys. Rev., 1964, 136, B864–B871.
16. W. Kohn and L. J. Sham, Phys. Rev., 1965, 140, A1133–A1138.
17. B. Delley, J. Chem. Phys., 1990, 92, 508–517.
18. B. Delley, J. Chem. Phys., 2000, 113, 7756–7764.
19. J. P. Perdew, K. Burke and M. Ernzerhof, Phys. Rev. Lett., 1996, 77, 3865–3868.
20. B. Delley, Phys. Rev. B: Condens. Matter Mater. Phys., 2002, 66, 155125.
21. T. A. Halgren and W. N. Lipscomb, Chem. Phys. Lett., 1977, 49, 225–232.
22. F. L. Hirshfeld, Theor. Chim. Acta, 1977, 44, 129–138.
23. B. Pauwels, G. Van Tendeloo, E. Zhurkin, M. Hou, G. Verschoren, L. Theil Kuhn, W. Bouwen and P. Lievens, Phys. Rev. B: Condens. Matter Mater. Phys., 2001, 63, 165406.
24. T. J. Toai, G. Rossi and R. Ferrando, Faraday Discuss., 2008, 138, 49–58.
25. L. Deng, W. Hu, H. Deng, S. Xiao and J. Tang, J. Phys. Chem. C, 2011, 115, 11355–11363.
26. F. Baletto and R. Ferrando, Rev. Mod. Phys., 2005, 77, 371–417.
27. R. Zhang, H. Liu, H. Zheng, L. Ling, Z. Li and B. Wang, Appl. Surf. Sci., 2011, 257, 4787–4794.
28. A. Önsten, M. Göthelid and U. O. Karlsson, Surf. Sci., 2009, 603, 257–264.
29. C. I. Carlisle, D. A. King, M. L. Bocquet, J. Cerdá and P. Sautet, Phys. Rev. Lett., 2000, 84, 3899–3902.
30. Y. Xu and M. Mavrikakis, Surf. Sci., 2001, 494, 131–144.
31. A. Soon, M. Todorova, B. Delley and C. Stampfl, Phys. Rev. B: Condens. Matter Mater. Phys., 2006, 73, 165424.
32. Q. Jiang, H. M. Lu and M. Zhao, J. Phys.: Condens. Matter, 2004, 16, 521–530.
33. Y. Xu and M. Mavrikakis, J. Am. Chem. Soc., 2004, 126, 4717–4725.
34. H. J. Kang, R. Shimizu and T. Okutani, Surf. Sci., 1982, 116, L173–L178.
35. A. Hussain, J. Phys. Chem. C, 2013, 117, 5084–5094.
36. H. Y. Kim, H. M. Lee and G. Henkelman, J. Am. Chem. Soc., 2011, 134, 1560–1570.
37. B. Z. Sun, W. K. Chen and Y. J. Xu, J. Chem. Phys., 2010, 133, 154502.
38. I. V. Yudanov, A. V. Matveev, K. M. Neyman and N. Rösch, J. Am. Chem. Soc., 2008, 130, 9342–9352.
39. B. Hammer and J. K. Norskov, Adv. Catal., 2000, 45, 71–129.
40. Y. Gao, N. Shao, Y. Pei, Z. Chen and X. C. Zeng, ACS Nano, 2011, 5, 7818–7829.
41. Z. Xu, E. Lai, Y. Shao-Horn and K. Hamad-Schifferli, Chem. Commun., 2012, 48, 5626–5628.

---