CO oxidation on Rh-doped hexadecagold clusters

Abstract

Exploring the unique catalytic properties of gold clusters associated with specific nano-architectures is essential for designing improved catalysts with a high mass-specific activity. We investigate the geometric and electronic structure of hexadecagold clusters in which Rh was doped. Density functional theory calculations demonstrate that the resulting neutral and negatively charged Rh-doped Au₁₆ clusters are stable and bind CO and O₂ stronger than Au₁₆. Consequently, activation barriers for CO oxidation are lowered. Microkinetics simulations predict especially negatively charged Rh-doped Au₁₆ clusters to exhibit very high CO oxidation activity, already at sub-ambient temperature. Our findings highlight the promise of alloying gold clusters with more reactive transition metals and the importance of charge transfer from the support in heterogeneous gold systems in catalyzing CO oxidation.

---

Introduction

Gold has been at the center of the attention of the catalysis community, since early reports of Haruta and Hutchings demonstrated the outstanding catalytic properties of nanoscale gold.^1–4 An important corollary from the extensive follow-up work on heterogeneous gold systems is that size, morphology and oxidation state all profoundly affect gold's unique catalytic properties.^5–13 Identification of the exact morphology of Au particles remains a considerable challenge, especially when their size is smaller than 1 nm.¹⁴ In the past two decades, many studies have dealt with the structure of small gold clusters in the gas phase,^14–29 which are model systems for supported gold nanoparticles.³⁰ A cluster that has caught significant attention is the hollow hexadecagold cage with tetrahedral (T_d) point symmetry.^19,22,23

In heterogeneous systems, charge transfer between gold and the support plays an important role in, for instance, CO oxidation.⁸ Due to gold's high electronegativity, small clusters can easily take up electrons from the support. Charging of gas-phase clusters, which is experimentally readily achieved, serves as a very simple model to study influence of charging. A change in the oxidation state of gold clusters is known to affect the binding of molecules and, henceforth, the catalytic properties.^31–37 Also doped Au clusters have received increasing attention, because doping presents a way to tune the electronic and catalytic properties.^38–40 Relevant to the fullerene-type Au₁₆ cluster is its endohedral doping by other atoms.^41–44 It has been experimentally demonstrated that endohedrally doped M–Au₁₆⁻ clusters (M = Fe, Co, Ni) are stable compounds, despite significant structural distortion induced by the foreign metal atom.⁴² Among these, Cu doping does not affect the structure of the cluster.²¹ Important to catalysis, doping can result in the activation of molecules such as O₂, as for instance observed when Si is inserted into the Au₁₆ cluster.⁴¹ Although most Au–metal combinations are not expected to give rise to stable alloys in nanoparticles,⁴⁵ doped clusters may be considered models for near-surface alloys, a topic which is also of growing relevance to heterogeneous catalysis.⁴⁶ Indeed, synergistic effects have been reported for heterogeneous Au catalysts containing small amounts of Rh, amongst others in CO oxidation.^47,48 How such doping influences the (electronic) structure and catalytic behavior remains an open question. In this work, we use density functional theory (DFT) calculations to investigate the structure of neutral and anionic Rh-doped Au₁₆ clusters and explore how Rh-doping affects the rate of CO oxidation on such clusters.

We here use a genetic search algorithm based on DFT calculations to compute the energy of candidate structures to predict stable Au₁₆ and neutral and anionic Rh-doped Au₁₆ clusters. The algorithm correctly identifies the T_d cage as the most stable structure for Au₁₆. These calculations also demonstrate the possibility of endohedral doping by Rh and, in this way, we show that neutral and anionic Rh–Au₁₆ clusters (Rh@Au₁₆). As Rh is preferentially located in the Au₁₆ cage, the clusters expose only Au atoms at their surface. First-principles microkinetics simulations highlight the exceptionally high potential of the bimetallic clusters in catalyzing CO oxidation: Rh-doped Au₁₆ clusters are much more active in CO oxidation than the Au₁₆ cluster. The improved activity is due to increased electron density around the Fermi level in the presence of Rh, which results in stronger adsorption of CO and O₂ and lower activation barriers for the surface CO oxidation step. This work provides inspiration to the design of transition metal doped Au cluster and nanoparticle catalysts to significantly speed up heterogeneous reactions.

Methods

All spin-polarized density functional theory (DFT) calculations were performed by using projector augmented wave (PAW)⁴⁹ potentials and the Perdew–Burke–Ernzerhof (PBE) functional⁵⁰ implemented in the Vienna ab initio simulation package (VASP)^51,52 code. The global minimum structures of Au₁₆, Rh@Au₁₆ and Rh@Au₁₆⁻ clusters were obtained by employing a genetic algorithm (GA) code to generate structures, whose geometries were optimized by first-principles DFT calculations. The clusters were placed in a 15 Å × 15 Å × 15 Å cubic supercell for geometry optimization. Brillouin zone sampling was restricted to the Γ point. All the atoms in the clusters were allowed to relax. The energy cutoff for the plane wave basis set was set to 270 eV for these structures optimizations, and forces on each atom below 0.05 eV Å⁻¹ was used as convergence criterion. For computing the adsorption energies of the intermediates, the cutoff energy was set to 400 eV and the force convergence criterion was 0.02 eV Å⁻¹. The improved force reversed method was used to determine the transition states for CO oxidation⁵³ and the force tolerance of 0.05 eV Å⁻¹ was used. It was verified that this approach led to similar TSs as the climbing-image nudged elastic band (CI-NEB) method.^54,55 All the energies were zero-point energy corrected, for which a frequency analysis was performed.

Infrared (IR) spectra of the Au₁₆ and the two Rh@Au₁₆ clusters were simulated using the Gaussian package employing a PBE0/LANL2DZ basis set.⁵⁶ The same program was used to simulate the photoelectron spectrum of the anionic Rh@Au₁₆ cluster. Theoretical predicted IR spectra were obtained by applying Lorentzian functions with the theoretical harmonic vibrational frequencies scaled by a 2 cm⁻¹ full width at half-maximum. The photoelectron simulation was done by fitting the distribution of the calculated vertical detachment energies (VDEs) with unit-area Gaussian functions of 0.06 eV half-width. The first VDE of the anionic Rh@Au₁₆⁻ cluster was calculated as the difference between the energies of the neutral and anionic global minimums clusters. The higher VDEs were approximated by the addition of the binding energies of the deeper occupied orbitals of the anionic cluster (i.e., the density of states) to the first VDE. In our work, a vibration analysis (Table S1†) has been done to make sure that the obtained structures are local minima on the potential energy surfaces without imaginary frequency.

Structure of gold clusters

The GA code employed to identify the global minimum structure of gold clusters is based on the principle of natural evolution and includes random generation of an initial population of clusters, geometric optimization of the candidate structures, assigning fitness to these structures based on their electronic energy, and generating a new population by crossover operations, as introduced by Deaven and Ho,^57,58 as well as random mutations by moving of atoms.⁵⁹ Typically, an initial population of ten random structures was generated and optimized at DFT-PAW-PBE level. Electronic energies calculated by DFT calculations were used to determine the fitness, structures with lower energy having higher probably of being retained in the population.⁶⁰ The two relatively most stable structures were kept in the next population. Energies and bond distances were used to reject multiple occurrences of structures in the population. Typically, around 700 structures were optimized to find the global minimum structures of Au₁₆ or RhAu₁₆ clusters.

Microkinetic simulations

The activation barriers are used to calculate the forward and backward rate constant for CO oxidation. For surface reactions, the rate constants for the forward and backward elementary reaction were determined by the Eyring equation:where k is the reaction rate constant in s⁻¹; k_b, T, h and E_a refer to the Boltzmann constant, temperature, Planck's constant and the activation barrier, respectively. Q^TS and Q are the partition functions of the transition and ground states, respectively. As an approximation, the pre-factor is set to 10¹³ s⁻¹ for all the elementary reactions.

For non-activated molecular adsorption, the rate of adsorption is determined by the rate of surface impingement of gas-phase molecules. The flux of incident molecules is given by Hertz–Knudsen equation:⁶¹

Hence, the molecular adsorption rate constant can be expressed as:

where P is the partial pressure of the adsorbate in the gas phase, A′ stands for the surface area of the adsorption site, m indicates the mass of the adsorbate and S is the sticking coefficient.

For the desorption process, it is assumed that there are three rotational degrees of freedom and two translational degrees of freedom in the transition state. Accordingly, the rate of desorption is given by

where σ and θ are the symmetry number and the characteristic temperature for rotation, respectively. E_des is the desorption energy.⁶²

The approach to microkinetic simulations has been presented in detail elsewhere.^63,64 Differential equations for all the surface reaction intermediates were constructed using the rate constants and the set of elementary reaction steps. For each of the M components in the kinetic network, a single differential equation in the form

is obtained. In this equation, k_j is the elementary reaction rate constant (see eqn (1)), ν^j_i is the stoichiometric coefficient of component i in elementary reaction step k and c_k is the concentration of component k on the catalytic surface.

The CO oxidation rate is calculated by the in-house developed MKMCXX program.^63–65 Steady-state coverages were found by integrating the ordinary differential equations in time until the changes in the surface coverages were very small. Because chemical systems typically give rise to stiff sets of ODEs, we have used the backward differentiation formula method for the time integration. The rates of the individual elementary reaction steps can be obtained based on the calculated steady-state surface coverages. In our simulations, the gas phase contained a mixture of CO and O₂ in 1 : 3 molar ratio at a total pressure of 0.05 atm.

The rate limiting step can be determined via the “degree of rate control (DRC)” method as introduced by Campbell et al.^66–68 For elementary step i, the degree of rate control X_RC,i can be defined as

where k_i, K_i and r are the rate constants, the equilibrium constant for step i and the reaction rate, respectively. Furthermore, the DRC coefficients have to obey the sum rule over all steps i in the mechanism in such a way that:⁶⁷

Results and discussion

The global minimum structures of Au₁₆ and two Rh-doped Au₁₆ clusters, i.e., one neutral (Au@Rh₁₆) and one anionic (Rh@Au₁₆⁻) clusters were identified by our GA-DFT approach. These structures are shown in Fig. 1 (other low-energy structures for these three systems are collected in Fig. S1†). The most stable structure for Au₁₆ is a tetrahedral cage with T_d symmetry (Fig. 1A) and is similar to the optimal structure reported by Chen et al.,²² leading considerable credence to our algorithm. This tetrahedral Au₁₆ cluster has slightly lower energy compared with other isomers (Fig. S1†). In our calculations, this lowest lying Au₁₆ cluster structure is more stable by 0.09 eV and 0.61 eV than other proposed stable structures with C_2v and T_d symmetry,^19,23 respectively. The diameter of the empty cage is approximately 5.5 Å, which is sufficient to allow doping by other metals. The GA algorithm predicts the endohedral insertion of Rh in the Au₁₆ cage to be the most stable structure (Fig. 1B) for both the anionic and neutral Rh-doped gold clusters. The ground state of anionic Rh@Au₁₆⁻ adopts a T_d symmetry with a valence electron configuration of (12a₁)²(13e)⁴(25t₂)⁶(14e)⁰. However, removal of an electron from the degenerate 25t₂ molecular orbital will induce symmetry degradation due to the Jahn–Teller effect,⁶⁹ this results in a slightly distorted tetrahedral structure with a lower D_2d symmetry for the neutral Rh@Au₁₆ cluster with a degenerated ground state. This slight distortion allows lowering of the energy of Rh@Au₁₆ cluster through the removal of some of the degeneracy present in the more symmetric Rh@Au₁₆⁻ structure. The Au₁₆ cluster host maintains the tetrahedral structure during the insertion of Rh atom; the tetrahedral cage is slightly enlarged by 0.1 Å for the neutral and anionic Rh@Au₁₆ clusters. Such Rh@Au₁₆^0/− cluster is significantly more stable than other structures. Similar to Au₁₆, these clusters have six square faces and eight triangular faces, capped with four more gold atoms on the non-nearest neighbour triangles. They only expose Au atoms with coordination numbers of three and five.

Fig. 1 The global minimum structure of Au₁₆ (A), neutral and anionic Rh@Au₁₆ (B) clusters. The Au₁₆ cluster has a cage structure in which Rh atom can be endohedrally doped. (color scheme cyan: Rh and yellow: Au).

The geometric and electronic structure of Au₁₆ clusters has only been scarcely investigated. Synthesis of Au_n⁻ clusters is possible by means of laser vaporization of a pure gold target.¹⁵ Infrared absorption spectroscopy (IR) can yield information about the structure and symmetry of neutral and charged clusters.⁷⁰ The calculated IR spectra for neutral Au₁₆ and neutral and anionic Rh@Au₁₆ clusters are shown in Fig. 2. The IR spectrum of neutral Au₁₆ cluster contains a dominant vibrational mode at 150 cm⁻¹. The spectrum of Rh@Au₁₆⁻ contains similar features that of Au₁₆, except of the red shift of main band to 130 cm⁻¹. The IR spectrum of neutral Rh@Au₁₆ is much more complex and comprises a larger number of active vibrational modes due to the lower symmetry of the cluster (D_2d point group). These three gold clusters can be distinguished by measuring their IR spectra. Photoelectron spectroscopy (PES) can be employed to investigate the structure of anionic clusters.^15,19,71,72 The calculated PES spectra for the ionic Rh@Au₁₆ cluster is presented in Fig. S2† which is extremely helpful to clarify its structure in experiment.

Fig. 2 Simulated infrared absorption spectroscopy of Au₁₆, neutral and negatively charged Rh@Au₁₆ clusters.

In order to gain more insight into the effect of Rh doping on neutral and anionic clusters, we analyze the electronic structure of the three gold clusters. The total density of states (DOS) is depicted in Fig. 3. Upon Rh doping, the broad band below the Fermi level, predominantly comprised of d-orbitals, shifts to lower energies. Apparent is also the non-metallic and anti-ferromagnetic character of the Au₁₆ cluster. The increased DOS around the Fermi-level highlights the more pronounced metallic character of the Rh-doped clusters. The difference plots in Fig. 3 show that part of the increased electron density around the Fermi level is due to the introduction of Rh. The neutral Rh@Au₁₆ cluster contains an unpaired electron. Thus, Rh modification of Au₁₆ leads to substantial changes in the electronic structure, which we may expect to strongly affect catalytic performance.

Fig. 3 Spin polarized total density of states (DOS) of Au₁₆ (up, black), Rh@Au₁₆ (middle, red) and Rh@Au₁₆⁻ (bottom, blue) clusters. The grey area is the total DOS of all the sixteen Au atoms in Au₁₆, Rh@Au₁₆ and Rh@Au₁₆⁻ clusters.

To probe the catalytic potential of the novel Rh-doped gold clusters, we investigate the mechanism of CO oxidation for the three gold clusters by DFT calculations. Using computed adsorption energies and reaction barriers, microkinetics simulations are then performed to determine CO conversion rates and identify possible rate-controlling steps. As the surface will be predominantly covered by CO, an Eley–Rideal mechanism – involving gaseous O₂ – is not reasonable in this case. Accordingly, we considered the well-accepted Langmuir-Hinshelwood mechanism⁷³ involving reaction between molecularly adsorbed CO and O₂, i.e., CO* + O₂* → OCOO** (N.B.: * points to surface vacancies). The OCOO** intermediate, which occupies two adjacent sites, will decompose to produce CO₂ and an adsorbed O atom (OCOO** → CO₂ + O* + *), followed by CO adsorption and reaction of CO* with O* to produce CO₂ and regeneration of the surface vacancies (CO* + O* → CO₂ + 2*). The computed potential energy diagrams for CO oxidation on the three clusters are depicted in Fig. 5. For the Au₁₆ cluster, CO and O₂ adsorb on the corner Au atoms (Fig. S3†) with adsorption energies of −0.70 eV and −0.14 eV, respectively (Table S2†). The relatively weak binding of O₂ is consistent with previous results for gold.^23,74–77 Rh doping affects the adsorption of CO and O₂ substantially. On the neutral Rh@Au₁₆ cluster, the adsorption energies of CO and O₂ are −0.86 eV and −0.27 eV, respectively. The O–O stretch frequency, which is computed to be 1563 cm⁻¹ in the gas phase, decreases to 1366 cm⁻¹ upon adsorption on Au₁₆ and 1284 cm⁻¹ upon adsorption to Rh@Au₁₆. The stronger perturbation of the O–O frequency is in line with the Blyholder model,⁷⁸ that is to say, the O–O bond is weakened by donation of electron density from the metal into the anti-bonding 2π* orbital of the O₂ molecule. On the negatively charged Rh@Au₁₆⁻ cluster, CO adsorbs with similar adsorption energy (−0.85 eV) as on the neutral Rh@Au₁₆ cluster, while O₂ adsorbs strongest (−0.34 eV) among the three investigated clusters. Relatively speaking, doping the Au₁₆ cluster with Rh affects more significantly the binding energy of O₂ than that of CO.

To understand the origin of the stronger adsorption of CO and O₂ on the Rh-doped Au₁₆ clusters, Bader charge and local density of states (LDOS) analysis were conducted. For the three clusters considered, O₂ adsorbs in a similar configuration (Fig. S3†), with one of its O atom binding to the corner atom of the cluster, which is the most reactive one with a coordination number of three. Whilst this corner atom has only a small negative charge (−0.09e) for Au₁₆, doping with Rh leads to values of −0.13e and −0.23e for the neutral and anionic Rh-doped clusters, respectively. The DFT calculations therefore demonstrate that a higher electron density of Au leads to stronger adsorption of O₂, in line with previous investigations.⁷⁹

Fig. 4A depicts the LDOS for free and adsorbed O₂ on Au₁₆, Rh@Au₁₆ and Rh@Au₁₆⁻. On Au₁₆, the spin-down state of the O₂ 2π* orbital is unoccupied due to its relatively high energy. For both Rh-doped Au₁₆ clusters, there are more states near the Fermi-level (Fig. 3), which can hybridize with the orbitals of O₂. This enhances the adsorption of O₂. Fig. 4A shows that doping of Au₁₆ leads to electron transfer from the Au d-band to the spin-down component of the 2π* orbital of O₂. This back-donation is more pronounced for the bimetallic clusters due to the lower-lying energy of the spin-down component of the 2π* orbital. These effects are most pronounced for the anionic Rh@Au₁₆⁻ cluster, explaining the highest O₂ binding energy for this case. Whilst coordination of O₂ is similar for all three clusters, Rh-doping changes the coordination mode of CO. CO prefers to bind at a corner Au atom on Au₁₆, while it is located on the Au atom directly interacting with Rh on the doped clusters. The main reason for the stronger adsorption of CO as compared with O₂ is the slightly stronger interaction of CO's 4σ and 5σ orbitals with the Au d-bands. As the 2π* orbital of CO is not occupied, the variation in binding energy is smaller as compared with O₂.

Fig. 4 Spin polarized LDOS and molecular orbitals of free and adsorbed O₂ (A) and CO (B) on Au₁₆, Rh@Au₁₆ and Rh@Au₁₆⁻ systems. The LDOS is projected the free (black) and adsorbed O₂ or CO (blue) over Au atom (red). The types of O₂ and CO orbitals are indicated.

Stabilization of CO and O₂ adsorption strongly affects reaction energetics and kinetics for CO oxidation (Fig. 5). Nørskov and co-workers showed that stronger CO and O₂ adsorption lead to the stabilization of the transition states for CO₂-forming surface reactions.^80,81 Consistent with this, we find that the activation barriers for the CO* + O₂* → OCOO** reaction, 0.45 eV for Au₁₆, 0.34 eV for Rh@Au₁₆ and 0.17 eV for Rh@Au₁₆⁻, are decreased when CO and O₂ bind stronger (Fig. 5, Table S3†). In this respect, it is important to mention that the transition state for this reaction on Au₁₆ is different from the ones on the Rh-doped clusters (see the Fig. S4†). This is due to the different preferred adsorption modes of CO and O₂. In the former case, O₂ migrates to the less stable adsorption site (neighboring the corner atom) and the cluster is significantly distorted in the transition state, increasing the activation barrier. On the other hand, for the bimetallic clusters, CO and O₂ remain at their favorable adsorption sites in the transition state with much less distortion of the cluster itself. This is certainly due to the presence of Rh inside the cage. Comparison of the two Rh-doped clusters shows that the lowest barrier of anionic Rh@Au₁₆⁻ cluster is associated with the most favorable reaction energy, consistent with Brønsted–Evans–Polanyi (BEP) considerations.⁸²

Fig. 5 Potential energy diagrams for CO oxidation on Au₁₆ (black), Rh@Au₁₆ (red) and Rh@Au₁₆⁻ (blue) clusters. The inset illustrations are the transition states for the CO* + O₂* → OCOO** reaction on the three gold clusters. The elementary forward reaction barriers (in eV) and the distances (in Å) between the two reactants at the transition states are indicated.

Decomposition of the OCOO surface intermediate to CO₂ and an O adatom is strongly exothermic in all cases. Computed activation barriers for the OCOO* → CO₂ + O* reaction are 0.39 eV, 0.19 eV and 0.37 eV for the Au₁₆, Rh@Au₁₆ and Rh@Au₁₆⁻ clusters, respectively. OCOO** decomposition has similar transition state configurations on Au₁₆ and Rh@Au₁₆ clusters (Fig. S4†). We found that the O atom remaining on the surface binds stronger to Rh@Au₁₆ by 0.45 eV in the transition state, which explains the lower activation barrier for CO₂ formation. This is consistent with the projected DOS of the O atom in the transition state given in Fig. S5.† The higher electron density on the Au atoms with coordination number 5 in anionic Rh@Au₁₆⁻ cluster displaces the O atom to top adsorption instead of bridge adsorption. This small difference is unfavorable and increases slightly the barrier for OCOO** decomposition on this cluster. Summarizing, OCOO** decomposition is more facile than its formation for Au₁₆ and neutral Rh@Au₁₆, and the reverse holds for the anionic Rh@Au₁₆⁻ cluster.

The catalytic cycle is closed by removal of the O adatom by another CO molecule. O preferably adsorbs in the cluster's three-fold sites with energies of −0.81 eV, −0.46 eV, and −0.39 eV, respectively, for Au₁₆, Rh@Au₁₆, and Rh@Au₁₆⁻. The nature of the transition states for the CO* + O* → CO₂ + 2* reactions are nearly the same for all three clusters (Fig. S4†) and the corresponding activation energies for Au₁₆ (0.61 eV), Rh@Au₁₆ (0.45 eV) and Rh@Au₁₆⁻ (0.33 eV) obey the Brønsted–Evans–Polanyi relation. These differences in reaction energetics point out the significant effect of Rh-doping and the extra electron on the cluster's reactivity.

We then used microkinetics simulations to predict the CO conversion rate based on the computed kinetic parameters.⁶⁵ Arrhenius plots in Fig. 6 highlight the increased CO oxidation activity upon doping of the cluster by Rh. The anionic Rh@Au₁₆⁻ cluster exhibits the highest activity and displays already outstanding performance at sub-ambient conditions. At room temperature, its activity is much higher than that of the neutral bimetallic cluster. The Au₁₆ presents the lowest activity. The lower optimum reaction temperature for the Rh-doped clusters relates to the stronger binding of CO and O₂. At high temperature, the reaction rates for all clusters decrease due to lowered CO and O₂ coverages (Fig. 6B–D). The very high activity of the anionic Rh@Au₁₆⁻ cluster relates to the low barrier for the CO* + O₂* → OCOO** reaction. A degree of rate control (DRC)^66–68 analysis (Fig. S6†) shows that, for this cluster, the removal of O from the surface is controlling the overall rate at low temperature, whereas for the other two clusters the oxidation of CO with O₂ is the most difficult step (CO* + O₂* → OCOO**). The higher activity of the Rh@Au₁₆ compared with Au₁₆ is due to the lower barriers for all three reactions. There is a strong competition between CO and O₂ adsorption on Au₁₆ and Rh@Au₁₆ clusters. As is shown in Fig. 6B and C, CO is almost fully covered these two clusters and O₂ coverage is significantly low due to its lower adsorption energy. The peculiar temperature dependence of the anionic cluster's rate and consequently the strong changes in DRC values (Fig. S6†) are due to the competition for surface sites of O atoms with OCOO below room temperature and with CO at higher temperature.

Fig. 6 Microkinetics simulations for CO oxidation on Au₁₆, Rh@Au₁₆ and Rh@Au₁₆⁻ clusters. (A) CO₂ formation rate log (r) (r in mol s⁻¹ site⁻¹). (B)–(D) The coverage of the intermediates involved in CO oxidation as a function of the temperature on Au₁₆, Rh@Au₁₆ and Rh@Au₁₆⁻ clusters, respectively.

The other two clusters show more conventional behavior: CO poisons the surface at low temperature. Table 1 also highlights that at room temperature the higher rate of Rh@Au₁₆ than Au₁₆ is not due to increased O₂ surface coverage. Instead, the difference can be entirely explained by the lower barrier for the CO* + O₂* surface reaction. Maximum reaction rates for Au₁₆, Rh@Au₁₆ and Rh@Au₁₆⁻ clusters occur at 390 K, 450 K and 280 K with values of 5.5 × 10⁻¹ mol s⁻¹ site⁻¹, 3.8 × 10² mol s⁻¹ site⁻¹ and 1.3 × 10⁶ mol s⁻¹ site⁻¹, respectively. This difference emphasizes the promise of doping of gold clusters by Rh and the importance of the negatively charge state of the cluster. There is no explicit experimental evidence yet for the benefit of Rh doping in Au clusters for CO oxidation. CO and O₂ adsorption strengths are increased when the gold cluster hosts a Rh atom, especially when the cluster is negatively charged. Au–Rh nanoparticles have been shown to be more active in CO oxidation than Au nanoparticles.⁴⁸ Charge transfer from the support to Au clusters has been well demonstrated in heterogeneous systems.^83–85 Obviously, the here discussed clusters will change their shape upon interaction with a solid oxide support. Our work provides a hint at a novel strategy to improve the performance of gold catalysts by doping with more reactive transition metals and tuning the charge state of the active phase.

Table 1 Steady-state coverages of the intermediates and reaction rates during CO oxidation on Au₁₆, neutral and anionic Rh@Au₁₆ clusters at 300 K, P_CO/P_O2 = 1/3 and P = 0.05 bar

Cluster	Θ CO	θ O2	Θ OCOO	Θ O	Θ *	r (mol s^−1)
Au16	∼1.0 × 10^0	7 × 10^−10	1 × 10^−10	5 × 10^−7	1 × 10^−3	5 × 10^−4
Rh@Au16	∼1.0 × 10^0	2 × 10^−10	5 × 10^−13	2 × 10^−8	3 × 10^−6	8 × 10^−3
Rh@Au16^−	9 × 10^−1	4 × 10^−5	8 × 10^−2	2 × 10^−2	4 × 10^−2	9 × 10^5

---

Conclusions

The stable structures of Au₁₆ and neutral and anionic Rh@Au₁₆ clusters were identified by a DFT-based genetic search algorithm. All clusters form a Au₁₆ cage-like structure with T_d or D_2d symmetry with Rh being located inside the cage in the doped clusters. Doping results in significant changes in the gold clusters' electronic properties, profoundly affecting the binding energies of CO and O₂. These changes are more pronounced when an electron is added to the Rh@Au₁₆ cluster. Stronger adsorption and activation of O₂ are due to increased electron density of the cluster just below the Fermi level, which overlaps with the O₂ 2π* orbital. The further consequence is that the activation barrier for the reaction of adsorbed CO with adsorbed O₂ is decreased. Similarly, the barriers for decomposition of the OCOO surface intermediate and the reaction of CO with the remaining O adatom are more favorable for the doped clusters. Kinetically, the negatively charged Rh@Au₁₆ cluster presents very high activity in CO oxidation, as the barrier for the initial CO oxidation step is so low that CO and O compete for the free sites. The other clusters display conventional CO-poisoning kinetic behavior. These data show the promise of doping and charging of gold clusters for CO oxidation, providing a hint at a novel strategy for enhancing the performance of supported gold systems.

Acknowledgements

EJMH, IAWF, YS, IT and JXL acknowledge financial support by NWO-VICI and NWO-TOP grants. ZL acknowledges the start-up fund from Shanxi Normal University for support (Grant No. 02070388).

Notes and references

1. M. Haruta, T. Kobayashi, H. Sano and N. Yamada, Chem. Lett., 1987, 16(2), 405–408.
2. A. S. K. Hashmi and G. J. Hutchings, Angew. Chem., Int. Ed., 2006, 45, 7896–7936.
3. J. Guzman, S. Carrettin and A. Corma, J. Am. Chem. Soc., 2005, 127, 3286–3287.
4. A. A. Herzing, C. J. Kiely, A. F. Carley, P. Landon and G. J. Hutchings, Science, 2008, 321, 1331–1335.
5. M. Valden, X. Lai and D. W. Goodman, Science, 1998, 281, 1647–1650.
6. M. Comotti, W.-C. Li, B. Spliethoff and F. Schüth, J. Am. Chem. Soc., 2006, 128, 917–924.
7. O. Lopez-Acevedo, K. A. Kacprzak, J. Akola and H. Häkkinen, Nat. Chem., 2010, 2, 329–334.
8. Y.-G. Wang, Y. Yoon, V.-A. Glezakou, J. Li and R. Rousseau, J. Am. Chem. Soc., 2013, 135, 10673–10683.
9. J. Guzman and B. C. Gates, J. Am. Chem. Soc., 2004, 126, 2672–2673.
10. Z.-Y. Li, Z. Yuan, X.-N. Li, Y.-X. Zhao and S.-G. He, J. Am. Chem. Soc., 2014, 136, 14307–14313.
11. X.-N. Li, Z. Yuan and S.-G. He, J. Am. Chem. Soc., 2014, 136, 3617–3623.
12. X. N. Li, Z. Y. Li, H. F. Li and S. G. He, Chem. – Eur. J., 2016, 22, 9024.
13. N. Li, H. M. Zhang, Z. Yuan and S. G. He, Nat. Commun., 2016, 7, 11404.
14. J. Wang, G. Wang and J. Zhao, Phys. Rev. B: Condens. Matter Mater. Phys., 2002, 66, 035418.
15. J. Li, X. Li, H.-J. Zhai and L.-S. Wang, Science, 2003, 299, 864–867.
16. X. Xing, B. Yoon, U. Landman and J. H. Parks, Phys. Rev. B: Condens. Matter Mater. Phys., 2006, 74, 165423.
17. G. Mills, M. S. Gordon and H. Metiu, Chem. Phys. Lett., 2002, 359, 493–499.
18. H. Häkkinen, B. Yoon, U. Landman, X. Li, H.-J. Zhai and L.-S. Wang, J. Phys. Chem. A, 2003, 107, 6168–6175.
19. S. Bulusu, X. Li, L.-S. Wang and X. C. Zeng, Proc. Natl. Acad. Sci. U. S. A., 2006, 103, 8326–8330.
20. H.-J. Zhai, J. Li and L.-S. Wang, J. Chem. Phys., 2004, 121, 8369–8374.
21. L. M. Wang, S. Bulusu, H. J. Zhai, X. C. Zeng and L. S. Wang, Angew. Chem., Int. Ed., 2007, 46, 2915–2918.
22. G. Chen, Q. Wang, Q. Sun, Y. Kawazoe and P. Jena, J. Chem. Phys., 2010, 132, 194306.
23. Y. Gao, N. Shao, Y. Pei, Z. Chen and X. C. Zeng, ACS Nano, 2011, 5, 7818–7829.
24. Y. Gao, N. Shao and X. C. Zeng, ACS Nano, 2008, 2, 1497–1503.
25. W. Huang, S. Bulusu, R. Pal, X. C. Zeng and L.-S. Wang, ACS Nano, 2009, 3, 1225–1230.
26. Y. Pei, N. Shao, Y. Gao and X. C. Zeng, ACS Nano, 2010, 4, 2009–2020.
27. Y. Pei, N. Shao, H. Li, D.-E. Jiang and X. C. Zeng, ACS Nano, 2011, 5, 1441–1449.
28. B. Schaefer, R. Pal, N. S. Khetrapal, M. Amsler, A. Sadeghi, V. Blum, X. C. Zeng, S. Goedecker and L.-S. Wang, ACS Nano, 2014, 8, 7413–7422.
29. R. Zhou, X. Wei, K. He, J. E. Shield, D. J. Sellmyer and X. C. Zeng, ACS Nano, 2011, 5, 9966–9976.
30. S. M. Lang and T. M. Bernhardt, Phys. Chem. Chem. Phys., 2012, 14, 9255–9269.
31. J.-H. Liu, A.-Q. Wang, Y.-S. Chi, H.-P. Lin and C.-Y. Mou, J. Phys. Chem. B, 2005, 109, 40–43.
32. J. Xu, T. White, P. Li, C. He, J. Yu, W. Yuan and Y.-F. Han, J. Am. Chem. Soc., 2010, 132, 10398–10406.
33. X. Liu, A. Wang, X. Wang, C.-Y. Mou and T. Zhang, Chem. Commun., 2008, 3187–3189.
34. S. Zhou, G. S. Jackson and B. Eichhorn, Adv. Funct. Mater., 2007, 17, 3099–3104.
35. C. L. Hill, J. Mol. Catal. A: Chem., 2007, 262, 2–6.
36. W. T. Wallace and R. L. Whetten, J. Am. Chem. Soc., 2002, 124, 7499–7505.
37. J. Hagen, L. D. Socaciu, M. Elijazyfer, U. Heiz, T. M. Bernhardt and L. Wöste, Phys. Chem. Chem. Phys., 2002, 4, 1707–1709.
38. J.-Y. Luo, M. Meng, X. Li, X.-G. Li, Y.-Q. Zha, T.-D. Hu, Y.-N. Xie and J. Zhang, J. Catal., 2008, 254, 310–324.
39. D. I. Enache, J. K. Edwards, P. Landon, B. Solsona-Espriu, A. F. Carley, A. A. Herzing, M. Watanabe, C. J. Kiely, D. W. Knight and G. J. Hutchings, Science, 2006, 311, 362–365.
40. J. K. Norskov, T. Bligaard, J. Rossmeisl and C. H. Christensen, Nat. Chem., 2009, 1, 37–46.
41. M. Walter and H. Häkkinen, Phys. Chem. Chem. Phys., 2006, 8, 5407–5411.
42. L.-M. Wang, J. Bai, A. Lechtken, W. Huang, D. Schooss, M. M. Kappes, X. C. Zeng and L.-S. Wang, Phys. Rev. B: Condens. Matter Mater. Phys., 2009, 79, 033413.
43. H.-F. Li and H.-Q. Wang, Phys. Chem. Chem. Phys., 2014, 16, 244–254.
44. Y. Gao, N. Shao, S. Bulusu and X. C. Zeng, J. Phys. Chem. C, 2008, 112, 8234–8238.
45. G. H. Jóhannesson, T. Bligaard, A. V. Ruban, H. L. Skriver, K. W. Jacobsen and J. K. Nørskov, Phys. Rev. Lett., 2002, 88, 255506.
46. J. Greeley and M. Mavrikakis, Nat. Mater., 2004, 3, 810–815.
47. M. Conte, A. F. Carley, G. Attard, A. A. Herzing, C. J. Kiely and G. J. Hutchings, J. Catal., 2008, 257, 190–198.
48. V. Abdelsayed, A. Aljarash, M. S. El-Shall, Z. A. Al Othman and A. H. Alghamdi, Chem. Mater., 2009, 21, 2825–2834.
49. P. E. Blöchl, Phys. Rev. B: Condens. Matter Mater. Phys., 1994, 50, 17953.
50. J. P. Perdew, K. Burke and M. Ernzerhof, Phys. Rev. Lett., 1996, 77, 3865–3868.
51. G. Kresse and J. Hafner, Phys. Rev. B: Condens. Matter Mater. Phys., 1993, 47, 558.
52. G. Kresse and J. Furthmüller, Phys. Rev. B: Condens. Matter Mater. Phys., 1996, 54, 11169.
53. K. Sun, Y. Zhao, H. Y. Su and W. X. Li, Theor. Chem. Acc., 2012, 131, 1–10.
54. G. Henkelman, B. P. Uberuaga and H. Jónsson, J. Chem. Phys., 2000, 113, 9901.
55. G. Henkelman and H. Jónsson, J. Chem. Phys., 2000, 113, 9978.
56. M. Frisch, G. Trucks, H. B. Schlegel, G. Scuseria, M. Robb, J. Cheeseman, G. Scalmani, V. Barone, B. Mennucci and G. Petersson, et al., Gaussian 09, Revision E.01, Gaussian, Inc., Wallingford, CT, 2009.
57. D. Deaven and K. Ho, Phys. Rev. Lett., 1995, 75, 288.
58. D. Daven, N. Tit, J. Morris and K. Ho, Chem. Phys. Lett., 1996, 256, 195–200.
59. X.-L. Ding, Z.-Y. Li, J.-H. Meng, Y.-X. Zhao and S.-G. He, J. Chem. Phys., 2012, 137, 214311.
60. L. Cheng, W. Cai and X. Shao, Chem. Phys. Lett., 2004, 389, 309–314.
61. P. Nitoń, A. Żywociński, M. Fiałkowski and R. Hołyst, Nanoscale, 2013, 5, 9732–9738.
62. A. P. J. Jansen, An introduction to kinetic Monte Carlo simulations of surface reactions, Springer, 2012.
63. I. A. W. Filot, R. A. van Santen and E. J. M. Hensen, Angew. Chem., 2014, 126, 12960–12964.
64. I. A. Filot, R. J. Broos, J. P. van Rijn, G. J. van Heugten, R. A. van Santen and E. J. Hensen, ACS Catal., 2015, 5, 5453–5467.
65. http://www.mkmcxx.nl/ .
66. C. T. Campbell, Top. Catal., 1994, 1, 353–366.
67. C. T. Campbell, J. Catal., 2001, 204, 520–524.
68. C. Stegelmann, A. Andreasen and C. T. Campbell, J. Am. Chem. Soc., 2009, 131, 8077–8082.
69. R. Englman, The Jahn-Teller effect in molecules and crystals, Wiley-Interscience, New York, 1972.
70. P. Gruene, D. M. Rayner, B. Redlich, A. F. van der Meer, J. T. Lyon, G. Meijer and A. Fielicke, Science, 2008, 321, 674–676.
71. M. Ji, X. Gu, X. Li, X. Gong, J. Li and L. S. Wang, Angew. Chem., Int. Ed., 2005, 44, 7119–7123.
72. W. Huang and L.-S. Wang, Phys. Rev. Lett., 2009, 102, 153401.
73. Z.-P. Liu, P. Hu and A. Alavi, J. Am. Chem. Soc., 2002, 124, 14770–14779.
74. P. Legare, L. Hilaire, M. Sotto and G. Maire, Surf. Sci., 1980, 91, 175–186.
75. J. Pireaux, M. Chtaib, J. Delrue, P. Thiry, M. Liehr and R. Caudano, Surf. Sci., 1984, 141, 211–220.
76. M. Bartram and B. Koel, Surf. Sci., 1989, 213, 137–156.
77. X. L. Ding, Z. Y. Li, J. L. Yang, J. G. Hou and Q. S. Zhu, J. Chem. Phys., 2004, 120, 9594.
78. G. Blyholder, J. Phys. Chem., 1964, 68, 2772–2777.
79. R. Coquet, K. L. Howard and D. J. Willock, Chem. Soc. Rev., 2008, 37, 2046–2076.
80. I. N. Remediakis, N. Lopez and J. K. Nørskov, Appl. Catal., A, 2005, 291, 13–20.
81. I. N. Remediakis, N. Lopez and J. K. Nørskov, Angew. Chem., 2005, 117, 1858–1860.
82. M. A. Barteau, Catal. Lett., 1991, 8, 175–183.
83. A. Sanchez, S. Abbet, U. Heiz, W.-D. Schneider, H. Häkkinen, R. Barnett and U. Landman, J. Phys. Chem. A, 1999, 103, 9573–9578.
84. L. M. Molina and B. Hammer, Phys. Rev. Lett., 2003, 90, 206102.
85. L. B. Vilhelmsen and B. Hammer, Phys. Rev. Lett., 2012, 108, 126101.

---

Footnotes

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c6cy02277d

‡ These authors contributed equally.

---