acs.jpcc.5b04223

Subscriber access provided by UNIV OF MISSISSIPPI

Article

Reaction Pathway and Free Energy Landscape of Catalytic Oxidation of Carbon Monoxide Operated by a Novel Supported Gold-Copper Alloy Cluster Kenichi Koizumi, Katsuyuki Nobusada, and Mauro Boero J. Phys. Chem. C, Just Accepted Manuscript • DOI: 10.1021/acs.jpcc.5b04223 • Publication Date (Web): 12 Jun 2015 Downloaded from http://pubs.acs.org on June 17, 2015

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

The Journal of Physical Chemistry C is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 23

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

The Journal of Physical Chemistry

Reaction Pathway and Free Energy Landscape of Catalytic Oxidation of Carbon Monoxide Operated by a Novel Supported Gold-Copper Alloy Cluster Kenichi Koizumi,†,‡ Katsuyuki Nobusada,∗,†,‡ and Mauro Boero¶ Department of Theoretical and Computational Molecular Science, Institute for Molecular Science, Myodaiji, Okazaki 444-8585, Japan, Elements Strategy Initiative for Catalysts and Batteries (ESICB), Kyoto University, Katsura, Kyoto 615-8520, Japan, and Institut de Physique et Chimie des Matériaux de Strasbourg UMR 7504, University of Strasbourg and CNRS, 23 rue du Loess, F-67034 Strasbourg, France E-mail: [email protected]

Abstract An insight into the catalytic activities of pure gold (Au8 ) and gold-alloy (CuAu7 ) clusters supported on either MgO(100) or graphene has been undertaken in the search for an efficient, yet commercially appealing production of Au-based nanocatalysts. The present set of first-principles dynamical simulations shows that the gold and goldcopper alloy clusters destabilize to various extents on MgO but preserve their structures ∗

To whom correspondence should be addressed Department of Theoretical and Computational Molecular Science, Institute for Molecular Science, Myodaiji, Okazaki 444-8585, Japan ‡ Elements Strategy Initiative for Catalysts and Batteries (ESICB), Kyoto University, Katsura, Kyoto 615-8520, Japan ¶ Institut de Physique et Chimie des Matériaux de Strasbourg UMR 7504, University of Strasbourg and CNRS, 23 rue du Loess, F-67034 Strasbourg, France †

1 ACS Paragon Plus Environment


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

on the graphene support at room temperature. Consequently, the Cu atom remains embedded inside the Au cluster on MgO, whereas it can be easily exposed on the cluster surface on the graphene substrate. This feature appears to be a general key issue to trigger the catalytic reaction and discloses new perspectives for a rational synthesis of supported Au-based catalysts relying on the intrinsic chemical character of Cu which possesses a stronger affinity to oxygen than Au. Indeed, the Cu atom acts as an active site for the approach of O2 and keeps the molecule bound to the cluster. We clarified that the catalytic oxidation of CO occurs on the graphene-supported CuAu7 in a highly selective Langmuir-Hinshelwood type reaction, addressing the long-standing controversy about the actual reaction mechanism for this type of catalysis. Our findings contribute to the development of efficient and commercially appealing supported alloy clusters driven by a proper choice of dopants and supports, thus reducing the use of expensive gold.

Introduction Gold is generally a stable and chemically inert metal and its resistance towards oxidation has been considered for a long time to be an inherent character of this element. However, this scenario changes dramatically at the nanometer scale. In fact, nanometer sized gold particles have been shown to exhibit multiple optical properties, 1–4 magnetic characters, 5–7 and unprecedented chemical reactivities deeply different from bulk Au. 8 Such a scenario has disclosed new perspectives in the use of Au nanoparticles and nanoclusters, which are expected to have nanotechnological applications in a wealth of fields from single-electron transistors, to light energy conversions, biosensings, and as DNA supports. 9–12 Since the seminal work of Haruta and coworkers, 13 where CO oxidation was shown to occur on gold nanoparticles of a size below 2 nm, the catalytic behavior of nanometer sized Au has been the target of rigorous and intensive studies. To date, it is clear that Au nanoparticles and clusters act as effective catalysts for many types of reactions such as oxidations of CO,


Page 2 of 23

Page 3 of 23

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


NOx , ammonia, trimethylamine, and several other chemical compounds undergoing redox reactions. 14–16 Specifically, nanometer sized Au-based catalysts are appealing candidates for the conversion of toxic air into harmless one. Focusing on CO oxidation, Au nano-clusters of more than eight atoms behave as catalysts provided that a proper support is chosen. 17–19 The drawback is the high cost and low availability of gold worldwide. A workaround commercially appealing is to replace at least a part of the Au clusters with a cheaper metal such as Fe, Cu, or Al. This would be a drastic breakthrough to realize cheaper Au-based catalysts economically and industrially more appealing. To this aim, however, understanding the atomic-level character of the clusters is mandatory. Yet, the detailed electronic structures and related catalytic reactivities of the supported clusters still escape experimental probes and call for accurate molecular modeling techniques. In this respect, first-principles simulations represent nowadays a reliable tool to complement experiments and to inspect the catalytic behaviors of these systems at the atomic scale, particularly if associated to advanced free energy sampling techniques suited to explore reaction pathways. Recent dynamical studies 20–23 have unveiled the soft and fluxional character of transition metal clusters, drastically changing the nowadays obsolete and naive image that clusters have specific rigid structures. Furthermore, catalytic reactions occur at finite temperature and can result into very exoenergetic processes calling for a proper account of free energy activation barriers and entropic contributions. 24,25 Moreover, small adsorbates such as O2 and CO in vacuum behave rather freely and can assume a variety of configurations on the surface of the cluster. Hence, enthalpic contributions to reaction free-energy barriers cannot be neglected. Indeed, those dynamical inspections are nowadays regarded as reliable virtual experiments in several kinds of catalytic reactions. 26 In the present study, we use the density-functional-theory (DFT) based Car-Parrinello molecular dynamics (CPMD) 27,28 combined with either the Blue Moon ensemble (BME) 29,30 or the metadynamics (MTD) 31–33 to sample pathways and to explore the free energy land-



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

scape of the various catalytic reactions. Former static calculations targeting Au cluster systems 34,35 were based on the assumption that a Langmuir-Hinshelwood (LH) mechanism occurs, in which all the reactant molecules involved in the process are adsorbed simultaneously on the catalyst surface and here they react in a concerted way. On the other hand, first-principles calculations evidenced the possibility of an Eley-Rideal (ER) mechanism in which reactants are adsorbed on the surface of the cluster at different stages and then they react directly with the adduct formerly adsorbed and already present on the surface. 36 Given this scenario, the reaction mechanism remains still unclear. To unravel the actual dynamical catalytic process we focused on Au8 and CuAu7 clusters, supported on MgO and graphene, within the computational approach summarized above, showing the importance of the substrate in stabilizing the clusters and working out reactivities and pathways. This provides interesting new insights into the CO oxidation and related catalytic reaction mechanisms. Moreover, this can offer a general hint in the search for novel catalysts industrially and economically appealing. According to our results, the key issues to develop an efficient CuAu7 catalyst for CO oxidation are (i) the choice of the support 37,38 and (ii) the intrinsic chemical character of the specific atom supplied to the Au cluster. We have provided a clear evidence that the type of a support strongly affects the structure and the dynamical behavior of the gold cluster. Such a feature can be exploited to localize the atom replacing Au at selected positions in the gold cluster. Moreover, we demonstrate that the efficiency of the alloy catalyst can be controlled by tuning the combination of (i) and (ii), thus providing a guideline for the design of efficient catalytic systems while keeping an eye on cost reduction.

Computational Details In our first-principles scheme, wavefunctions for the valence electrons are expanded on a plane-wave basis set with an energy cut-off of 90 Ry with the sampling of the Brillouin zone limited to the Γ point. The relatively large cut-off is due to inclusion of semicore states


Page 4 of 23

Page 5 of 23

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


for the transition (Au, Cu) and alkali (Mg) metals. In the case of Cu and Mg the corevalence interactions are described by norm-conserving pseudopotentials (PPs), 39 while for Au a Hartwigsen-Goedecker-Hutter 40–42 PP is used. For the remaining atoms (C, O) regular Troullier-Martins PPs 39 are adopted. Our DFT approach includes the generalized gradient approximation proposed by Perdew and coworkers 43 complemented with Grimme’s van der Waals dispersions, 44 assessed in a former work for this type of surface reactions. 45 A spinunrestricted approach is adopted in all simulations including O2 and CO molecules to ensure a correct description of possible unsaturated or dangling bonds. The MgO (100) surface is modeled as a two-layers slab with a 4×4 surface and the graphene is a 12.25×12.74 ˚ A2 rectangular sheet. In both cases periodic boundary conditions are applied, leaving a vacuum space larger than 10 ˚ A above the surfaces to ensure a good separation from periodically repeated images. The ionic temperature (300 K) is controlled by a Nosé-Hoover thermostat 46–49 in the NVT ensemble. As mentioned above, all dynamical simulations are performed within the CPMD 27,28 scheme and the various reaction pathways are sampled by means of constrained dynamics and subsequent thermodynamic integration as in the BME 29,30 and of the variational formulation of the metadynamics approach. 32,33,50 The integration time step was 0.096 fs (4.0 atomic unit), and the fictitious electronic mass was set to µ=400.0 atomic units, ensuring a good control of the conserved quantities. For BME simulation, taking effects from dynamical conformation changes of the adsorbates and the clusters into account, relatively long equilibration times (∼ 2 ps) were chosen for each sampling. For catalytic reactions and oxidation processes, these schemes have already been shown to provide accurate and reliable results. 24,45,51,52 The CPMD program was used for all the reported simulations. 27,28

Results and discussion As a first step, we inspected the stability of the Au8 and CuAu7 clusters by performing unconstrained CPMD simulations on the MgO (100) surface. The result is shown in Figure 1a,



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

where an initial spheroidal nugget-shaped structure was placed on the surface. Considering the symmetry of the nugget Au8 structure, we can identify four different isomers of CuAu7 referred to as type 1, 2, 3, and 4, respectively, according to the numerical labels of Figure 1a for the Cu site. !"#

!%#

!'#

!&#

!(#

! " $ #

!$#

Figure 1: Initial (a) and final (b) structures of the Au8 cluster after 10 ps of NVT dynamics. Analogous initial (c) and equilibrated (d) structures for CuAu7 . Examples of fluxional behavior of Au8 (e) and CuAu7 (f) on MgO (100). The color code for the ball and stick models is yellow, blue, red and pink for Au, Cu, O and Mg, respectively. Numbers in (a) refer to the different Au sites replaced by Cu in CuAu7 .

A 10 ps lasting CPMD simulation has shown that Au8 undergoes a gradual deformation exhibiting a fluxional behavior eventually resulting in the flat structure shown in Figure 1b. Our simulations were repeated either with or without the inclusion of van der Waals corrections. 44 In both cases the results turned out identical. On these grounds, it is unclear to us if the fact that former first principles calculations, within the same DFT framework 36 used here, did not display such a fluxional behavior is due to the lack of dynamics (or insufficient simulation time) prior to the introduction of the constraints. Analogously to Au8 , CuAu7 also exhibited a fluxional behavior during the equilibration. In particular, all CuAu7 clusters in which Cu occupies one of the sites from 2 to 4 (see Figure 1a) evolved into flat structures of the type shown in Figures 1c and f. In the process, the Cu atom forms a strong chemical bond with a close-by oxygen atom of the MgO surface and this bond does not cleave at least on the time scale of our simulations. The Cu atom is then not exposed to the outer surface of the cluster, as sketched in Figures 1c and f, and 6 ACS Paragon Plus Environment

Page 6 of 23

Page 7 of 23

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


this hinders any catalytic activity. The origin of these fluxional behaviors was rationalized in terms of charge density differences, i.e. ρ(cluster + support) − (ρ(cluster) + ρ(support)). As shown in Figures 2a and b, a strong charge transfer occurs from the MgO support to the cluster, in line with the results of Molina and Hammer. 53 The additional charge populates the anti-bonding states of the clusters, thus weakening the chemical bond in both Au8 and CuAu7 . Hence, the thermal motion allows a system to overcome small energy barriers (kB T = 26 meV at 300 K) separating the nugget and the flat configurations giving rise to the behavior observed here and in small Cu clusters. 23 The recent dynamical simulation of an isolated Au13 cluster by Beret et al. shows that there exists a complex free-energy landscape with a considerable number of local minima and the planar structure is located at the bottom of a deep valley, well separated from non-planar structures by rather high freeenergy barriers. 21 This clearly indicates that a planar structure can never revert back to any non-planar configuration once the system is trapped in such a deep energy well. Analogously, a study reported by Han also indicates that such a planar structure is the most stable one. 54 All these independent results corroborate and support the outcome of our simulations, which indeed can catch all the stable and metastable cluster configurations and fully account for thermal fluctuations.

!"#

!$#

!%#

Figure 2: Charge density differences of equilibrated (a) Au8 and (b) CuAu7 on MgO (100) and (c) Au8 on graphene. Colors of all atoms are identical to those in Figures 1 apart from C in cyan. Charge densities are shown as isosurfaces at ±2.0 × 10−3 e/˚ A3 . The green represents an increase in charge density, while the orange indicates a decrease.

The reaction pathways for the approach of O2 and CO to Au8 on MgO were investigated within the BME method. The reaction coordinate, in the first case, was the distance between



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

!%#

!$#

!"#

Figure 3: Au8 (a) and CuAu7 (b, c) structures on the graphene support after 5 ps equilibration. Panel (b) shows how the Cu atom is exposed on the external cluster surface when Cu is added to the outmost part of the cluster (i.e. the positions 1 and 2 of Figure 1 (a)). The color code is identical to that in Figure 2. the center of mass of the cluster and O2 , sampled at intervals of 0.26 ˚ A. These simulations have shown that O2 prefers to approach the MgO surface, making a strong bond to Mg. The Lagrange multiplier of the BME constraint was characterized by a large numerical noise when the molecule was close to the surface, suggesting that O2 feels a strong attractive force from the support, resulting in the formation of a strong bond between the under-coordinated Mg exposed at the surface and one of the two oxygen atoms of O2 . Conversely, for the CO, no noise was observed in the Lagrange multiplier indicating that CO weakly interacts with MgO even when the molecule is close to the surface. We remark that an O2 molecule can approach the supported gold cluster after diffusing above the MgO surface, while CO may directly approach the Au cluster. However, in both cases we do not expect the catalytic activity to be significant. This point will be discussed later, but we can anticipate here that the catalytic oxidation of CO occurs when O2 is bound to an exposed Cu active site of the Cu-doped gold cluster. As formerly reported by Yoon and coworkers, an Au8 cluster is not active for CO oxidation on a perfect MgO support at room temperature. 19 Given this scenario, our theoretical results are in line with their experimental observations. In the present study, although we limit our discussions to the catalytic oxidation on non-defective supports, a few words must be spent on the effects of surface defects in the substrate. Yoon et al. provided evidence for a catalytic activity for MgO supports in which oxygen vacancies are present on the exposed surface. 17 Analogously, Guzman and Gates 14 8 ACS Paragon Plus Environment

Page 8 of 23

Page 9 of 23

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


also reported reliable results about the fact that Au/MgO catalysts can be active for CO oxidation in the presence of surface defects. According to Yoon’s discussions, the significant charge transfer from the Fs-centered defects to the Au8 cluster weakens the bond of adsorbed O2 and this change in the electronic structure is likely to be responsible for the catalytic reactivity. Yet, defects are generally randomly distributed on the substrate surface and the control at an atomic-scale level of the amount and specific location of oxygen vacancies on the exposed surface is basically unachievable experimentally. As a word of warning, we have to remind that surface defects may also have an influence on the stability of supported clusters. It can then be inferred that the large charge transfer due to the defects makes the supported clusters Au7 and CuAu7 more fluxional and their structures would undergo significant changes. Then, for the CuAu7 case, the Cu atom can be easily trapped by an attractive O site around the defect on the MgO surface, thus becoming buried in the support and not exposed, hence not active for the catalysis. Indeed, as commented above (and discussed later), the doped Cu site plays a crucial role in the catalytic CO oxidation only when this dopant is exposed. Consequently, the doped Cu atom in CuAu7 on the defective MgO would be catalytically inactive in a way analogous to what observed in the case of a perfect MgO support. Unconstrained CPMD simulations of Au8 and CuAu7 on graphene have shown that on this support the clusters keep their conformations rather rigidly. Charge density differences calculated on the equilibrated structures have shown that the charge transfers from the support to the clusters are nearly negligible (see Figure 2 c). Another remarkable feature is the fact that in the CuAu7 cluster, the Cu atom located in one of the most favorable sites discussed above remains exposed on the external surface, as shown in Figure 3. Hence, results on graphene are deeply different from the ones on an MgO support. Also in this case BME simulations allowed to inspect O2 and CO adsorptions onto Au8 and CuAu7 . At variance with the MgO-supported systems, on graphene the adsorbates can directly approach the clusters. The free energy profiles show a monotonic decrease with two stable



minima, with respect to the initial configurations, at -0.03 eV and -0.07 eV for the approach of O2 to CuAu7 and CO to Au8 , corresponding to the cluster-molecule distances of 4.6 ˚ A and 5.0 ˚ A, respectively. During the approach of O2 to CuAu7 , the Cu site moves slightly to catch the O2 molecule. A similar behavior was observed for the corresponding Au site (see Figure 1 a) in the capture of CO. Eventually the Cu-O (for O2 ) and Au-C (for CO) distances shrink to 1.9 and 2.2 ˚ A on their respective local minima and these distance are short enough to form chemical bonds. Contrary to these barrierless processes, either no minimum or very shallow minima (easily overcome at room temperature) were found for the approaches of O2 to Au8 and CO to CuAu7 . This supports the intuitive notion of an intrinsic chemical difference in the characters of Au and Cu. Au is active to CO adsorption, whereas Cu is active to O2 adsorption. These BME results are graphically summarized in Figure 4. ¿From these results, we expect two possible reaction pathways for the catalytic

!"#!

()**+*,*)-.+/*01

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 10 of 23

Au8+O2 Au8+CO CuAu7+O2 CuAu7+CO

!"!$ !"!! ï!"!$ ï!"#! %"!

%"$

&"!

&"$

$"!

$"$

'"!

'"$

23456,7*+/81 Figure 4: Free-energy profiles for the adsorption processes of CO and O2 onto Au8 and CuAu7 clusters supported on a graphene sheet as obtained via Blue-Moon enhanced firstprinciples molecular dynamics. The distances indicated are the ones between the centers of mass of each adsorbate and the clusters. The insets show the initial and final geometries for the adsorption of O2 on the Cu site. Whenever the target cluster is CuAu7 , adsorbates are initially set at a distance of about 3.0 ˚ A away from the Cu sites to ensure that no chemical bond exists in the initial stage. These will be formed only upon subsequent approach to the Cu active sites.


Page 11 of 23

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


oxidation of the carbon monoxide: (i) O2 reaction with the CO adsorbed on the Au8 cluster and (ii) CO reaction with the O2 adsorbed on the CuAu7 cluster. BME simulations were done to inspect these two viable alternatives. Starting from the O2 reaction with the CO-Au8 system, we inserted an O2 molecule in the system at a distance of about 3.0 ˚ A from the CO adduct. This distance is the reaction coordinate monitored during the catalysis. Although the O2 molecule initially tends to approach the surface of the CO-Au8 cluster, it does not stabilize close to the cluster. This seems to be due to the fact that the Au8 cluster is inactive towards O2 adsorption, as demonstrated above. Indeed, eventually the O2 molecule leaves the CO-Au8 cluster just because of thermal fluctuations before completing the catalytic oxidation. By static calculations, we have also verified that the CO adsorption energies via the C-end to the Au surface are 0.85-1.10 eV and that the ones via the O-end are, instead, very small (0.05-0.07 eV). It is then evident that CO is adsorbed to the Au cluster only via the C-end. Similar calculations for the O2 -Au(surface) adsorption gave values of 0.25-0.36 eV. Indeed, an unconstraint MD simulation has shown that O2 desorbs spontaneously from the Au surface at room temperature within a short time (about 4 ps). These results are consistent with experiments showing that the pure Au8 cluster is catalytically inactive at high temperature. 17 Coming to the second pathway, namely the catalytic reaction of CO with O2 -CuAu7 , we introduced a CO molecule at an initial distance of 3.0 ˚ A from the O2 adduct and, also in this case, the distance between the C atom in CO and the outmost O atom of the O2 adduct is used as a reaction coordinate. As shown in Figure 5, CO approaches the Au atom adjacent to the Cu site and the C atom gradually reduces its distance from the O atom of the adsorbed O2 . During the simulation, the adsorbed O2 molecule may rotate undergoing a reorientation between the axial and the equatorial configurations. As a consequence, the free energy barrier increases up to 0.3 eV and a new chemical bond between C and O forms. A subsequent unconstrained dynamics, following the constrained MD, confirmed that this intermediate state is substantially stable at room temperature. This reaction process is reminiscent of a typical LH type reaction and the third step in Figure 5 is likely to be an



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

intermediate state before the completion of the catalysis. As an additional check, we carried out a simulation in which the selected reaction coordinate was the distance between C in CO and the center of mass of O2 . We remark that when the constraint force becomes sufficiently strong, O2 leaves the Cu site. In these conditions, the affinity of Cu to O2 is not strong enough to promote the reaction and this example shows that an inappropriate choice of the reaction coordinate can lead to different results. We performed additional unconstrained MD simulations for the catalytic reaction in which CO approaches O2 on CuAu7 on which two previously adsorbed CO molecules were present on two different Au sites to verify whether or not a high coverage can affect the reaction mechanism. Our results indicate that in this case CO cannot spontaneously approach the adsorbed O2 molecule bound to the Cu site and ER reaction cannot occur. Indeed, as reported by Yoon et al., electron back donations take place from the cluster to the adsorbates. 17 If the coverage of CO on the cluster is high and those adsorbates drain electrons from the cluster, then the reactivity of the outermost oxygen atom is suppressed according to Yoon’s discussion. Therefore, we can infer that a ligand high-coverage jeopardizes the direct ER mechanism and has a minor influence on the LH reaction pathway. !"#

$%&'()'*$"&' !+#

Figure 5: A first step of the Langmuir-Hinshelwood type reaction on the CuAu7 and Au8 clusters obtained via blue-moon ensemble simulations. (a) CO approaches the gold surface and reacts with O2 anchored on the Cu active site, forming an intermediate for the subsequent CO2 desorption. (b) O2 initially approaches the pure Au surface, but desorbs before reacting with CO. The colors are identical to those in Figure 2.


Page 12 of 23

Page 13 of 23

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Figure 6: A second step of the Langmuir-Hinshelwood type reaction on the CuAu7 cluster obtained via metadynamics. In this process, CO desorbs from the cluster surface. Details about the collective variables (CVs) are given in the text. (a) The free-energy landscape using distance CVs. (b) The free-energy landscape using coordination number CVs. The blue arrows indicate the direction for escaping the first free energy minimum. The insets show the initial geometries and the final configurations after CO2 desorption. The colors are identical to those in Figure 2. To investigate the next reaction step and to verify whether sequential LH-type catalytic oxidation occurs, we resort on MTD simulations. Two collective variables (CVs) are employed in the simulations; one is the distance between the two neighboring O atoms in the intermediate state and the other is the distance between the C atom and the closest Au atom. The CO2 desorption process turns out to be a concerted reaction, as summarized in Figure 6 in terms of the free-energy landscape, characterized by a free energy barrier of 0.5 eV. We confirmed this result by an independent MTD simulation using different CVs: the oxygen coordination number in the O–O moiety of the intermediate and the coordination number of C with the nearby Au atoms. Also in this case we found the same concerted reaction for the CO2 desorption. The free energy barrier is nearly identical to the case of distance CVs, as illustrated in Figure 6 b. This demonstrates that the catalytic oxidation of CO on the supported CuAu7 cluster follows a LH reaction pathway upon formation of the intermediate state described above. Now, one O atom is left on the cluster surface after the CO2 desorption as shown in the insets of Figure 6. This atomic oxygen is expected to react further with nearby CO molecules. To confirm also this point, we performed a subse-



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

quent BME simulation after adding a CO molecule in the simulation cell. Initially, the CO approaches the Au surface, but then comes closer to the O atom on the cluster, forming an intermediate state with the CO in the vicinity of this O atom of the cluster. This occurs by overcoming a modest barrier (0.1 eV). Then, we inspected the desorption of CO2 from this intermediate via the MTD simulations analogous to the former ones. Interestingly, this second CO2 desorption occurs very rapidly and in a nearly barrierless way. This clearly demonstrates that the atomic O is easily converted into CO2 along, again, a LH mechanisms and the cluster is reset to this initial conditions, ready to replicate the process. We further observed that the dissociation of O2 , initially adsorbed on the Cu site, into two atomic oxygens is energetically too demanding to occur, being characterized by a free energy barrier larger than 1.3 eV. We can then rule out this reaction channel. An important remark has to be put forward concerning the graphene supports. As extensively discussed in the literature and theoretically investigated also by our group, 51,52 defects or impurities (namely N and B) in graphene might help to tightly anchor the gold cluster into the support and, to an extent which has not yet been fully clarified, contribute to enhance the catalytic activity. Yet, the pristine undefective material is far from being irrelevant. To have an idea about how gold clusters supported on perfect carbon materials can become realistic catalysts, we refer the reader to the recent experimental result reported by Xie and coworkers. 55 They reported catalytic activities for benzyl alcohol, instead of CO, promoted by small gold clusters supported on carbon nanotubes. They paid careful attentions to aggregation of the gold clusters and succeeded in supporting chemically less modulated gold clusters on graphene. These experimental results corroborate our findings and provide support to our proposed mechanism for the catalytic activity of Au clusters also on perfect graphene.


Page 14 of 23

Page 15 of 23

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Conclusions We have investigated the dynamical processes of adsorption of CO and O2 on Au8 and CuAu7 clusters supported on either MgO or graphene via reactive CPMD simulations. In the first case the MgO support strongly interacts with the clusters and the adsorbates, hindering the catalytic activity. Conversely, the graphene support allows for the approach of the adsorbates and preserves the clusters structure because of the weak interactions with them. The CuAu7 cluster, in turn, can expose the Cu site on its external surface, thus enhancing the catalysis on a graphene support. We clarified that Cu is active to O2 adsorption, whereas Au is active to the CO adsorption. Controlling these features, we hereby proposed a guideline to design a copper-doped gold cluster catalyst supported by graphene. We demonstrated that the catalytic mechanism for the CO oxidation is based on a LH type reaction with a modest free-energy barrier of lower than 0.5 eV. The doped Cu atom acts as an anchor to hold the O2 molecule for the initial step of this reaction. The atomic O left on the cluster is easily converted into CO2 and desorbs. Essential scenarios and free-energy landscapes for the catalytic CO oxidation reactions were systematically provided. Moreover, we settled the long standing controversy about the reaction pathway, providing a convincing support to the Langmuir-Hinshelwood mechanism as opposed to the Eley-Rideal one. We have clarified that the type of a support deeply affects the stability and the dynamical behavior of gold clusters; yet, this feature can become useful to substitute Au atoms with another less expensive metal (e.g. Cu) at specific selected sites of the cluster. The efficiency of the alloy catalyst can then be controlled by a careful selection of the support and an equally important tuning of the doped atom replacing Au. These studies demonstrate the possibility to experimentalists and engineers working in the field of catalysts design and synthesis to develop efficient and commercially appealing supported alloy clusters that can replace expensive pure gold clusters by utilizing and controlling intrinsic chemical characters of the clusters, the dopants, and the supports.



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Notes The authors declare no competing financial interest.

References (1) Klar, T.; Perner, M.; Grosse, S.; von Plessen, G.; Spirkl, W.; Feldmann, J. SurfacePlasmon Resonances in Single Metallic Nanoparticles. Phys. Rev. Lett. 1998, 80, 42494252. (2) Barnes, W. L.; Dereux, A.; Ebbesen, T. Surface Plasmon Subwavelength Optics. Nature 2003, 424, 824-830. (3) Park, H.; Lee, S.; Lee, J. M.; Nam, S. A.; Jeon, T.; Han, S. W.; Kim, S. O. High Performance Organic Photovoltaics with Plasmonic-Coupled Metal Nanoparticle Clusters. ACS Nano 2014, 8, 10305-10312. (4) Furube, A.; Du, L.; Hara, K.; Katoh, R.; Tachiya, M. Ultrafast Plasmon-Induced Electron Transfer from Gold Nanodots into TiO2 Nanoparticles. J. Am. Chem. Soc. 2007, 129, 14852-14853. (5) Yamamoto, Y.; Miura, T.; Suzuki, M.; Kawamura, N.; Miyagawa, H.; Nakamura, T.; Kobayashi, K.; Teranishi, T.; Hori, H. Direct Observation of Ferromagnetic Spin Polarization in Gold Nanoparticles. Phys. Rev. Lett. 2004, 93, 116801. (6) Hori, H.; Yamamoto, Y.; Iwamoto, T.; Miura, T.; Teranishi, T.; Miyake M. Diameter Dependence of Ferromagnetic Spin Moment in Au Nanocrystals. Phys. Rev. B 2004, 69, 174411. (7) Tuboltsev, V.; Savin, A.; Pirojenco, A.;Räisänen, J. Magnetism in Nanocrystalline Gold. ACS Nano 2013, 7, 6691-6699.


Page 16 of 23

Page 17 of 23

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


(8) Elghanian, R.; Storhoff, J. J.; Mucic, R. C.; Letsinger, R. L.; Mirkin C. A. Selective Colorimetric Detection of Polynucleotides Based on the Distance-Dependent Optical Properties of Gold Nanoparticles. Science 1997, 277, 1078-1081. (9) Daniel, M-C.; Astuc, D. Gold Nanoparticles: Assembly, Supramolecular Chemistry, Quantum-Size-Related Properties, and Applications Toward Biology, Catalysis, and Nanotechnology. Chem. Rev. 2004, 104, 293-346. (10) Zhang, H.; Yasutake, Y.; Shichibu, Y.; Teranishi, T.; Majima Y. Tunneling Resistance of Double-Barrier Tunneling Structures with an Alkanethiol-protected Au Nanoparticle. Phys. Rev. B 2005, 72, 205441. (11) Azuma, Y.; Yasutake, Y.; Kono, K.; Kanehara, M.; Teranishi, T.; Majima, Y. SingleElectron Transistor Fabricated by Two Bottom-Up Processes of Electroless Au Plating and Chemisorption of Au Nanoparticle. Jpn. J. Appl. Phys 2010, 49, 09206. (12) Chi, C.; Vargas-Lara, F.; Tkachenko, A. V.;Starr, F. W.; Gang, O. Internal Structure of Nanoparticle Dimers Linked by DNA. ACS Nano 2012, 6, 6793-6802. (13) Haruta, M.; Yamada, N.; Kobayashi, T.; Iijima, S. Gold Catalysts Prepared by Coprecipitation for Low-Temperature Oxidation of Hydrogen and of Carbon Monoxide. J. Catal. 1989, 115, 301-309. (14) Bond, G. C.; Louis, C.; Thompson D. T. Catalysis by Gold: Imperial College Press, London, 2006 (15) Guzman, J.; Gates, B. C. Catalysis by Supported Gold: Correlation between Catalytic Activity for CO Oxidation and Oxidation States of Gold. J. Am. Chem. Soc. 2004, 126, 2672-2673. (16) Williams, W. D.; Shekhar, M.; Lee, W-S.; Kispersky, V.; Degass, W. N.; Ribeiro, F. H.; Kim, S. M.; Stach, E. A.; Miller, J. T.; Allard, L. F. Metallic Corner Atoms in Gold 17 ACS Paragon Plus Environment


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Clusters Supported on Rutile Are the Dominant Active Site during Water-Gas Shift Catalysis. J. Am. Chem. Soc. 2010, 132, 14018-14020. (17) Arenz, M.; Landman, U.; Heiz, U. CO Combustion on Supported Gold Clusters. ChemPhysChem 2006, 7, 1871-1879. (18) Landman, U.; Yoon, B.; Zhang, C.; Heiz, U.; Arenz, M. Factors in Gold Nanocatalysis: Oxidation of CO in the Non-Scalable Size Regime. Top. Catal. 2007, 44, 145-158. (19) Yoon, B.; Häkkinen, H.; Landman, U.; Wörz, A. S.; Antonietti, J-M.; Abbet, S.; Judai, K.; Heiz, U. Charging Effects on Bonding and Catalyzed Oxidation of CO on Au8 Clusters on MgO. Science, 2005, 307, 403-407. (20) Vargas, A.; Santarossa, G.; Iannuzzi, M.; Baiker, A. Fluxionality of Gold Nanoparticles Investigated by Born-Oppenheimer Molecular Dynamics. Phys. Rev. B 2009, 80, 195421 (21) Beret, E. C.; Ghiringhelli, L. M.; Scheffler, M. Free Gold Clusters: Beyond the Static, Monostructure Description. Faraday Discuss. 2011, 152, 153-167 (22) Santarossa, G.; Vargas, A.; Iannuzzi, M.; Baiker, A. Free Energy Surface of Two- and Three-Dimensional Transitions of Au 12 Nanoclusters Obtained by Ab initio Metadynamics. Phys. Rev. B 2010, 81, 174205 (23) Musolino, V.; Selloni, A.; Car, R. Structure and Dynamics of Small Metallic Clusters on an Insulating Metal-Oxide Surface: Copper on MgO(100). Phys. Rev. Lett. 1999, 83, 3242-3245. (24) Boero, M.; Parrinello, M.; Terakura, K. First Principles Molecular Dynamics Study of Ziegler-Natta Heterogeneous Catalysis. J. Am. Chem. Soc. 1998, 120, 2746-2752. (25) Boero, M.; Parrinello, M.; Huffer, S.; Weiss, H. First Principles Study of Propene


Page 18 of 23

Page 19 of 23

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Polymerization in Ziegler-Natta Heterogeneous Catalysis. J. Am. Chem. Soc. 2000, 122, 501-509. (26) Thiel, W. Computational Catalysis–Past, Present, and Future. Angew. Chem. Int. Ed. 2014, 53, 8605-8613 (27) Car, R.; Parrinello, M. Unified Approach for Molecular Dynamics and DensityFunctional Theory. Phys. Rev. Lett. 1985, 55, 2471-2474. (28) CPMD, http://www.cpmd.org/, Copyright IBM Corp 1990-2013, Copyright MPI f¨ ur Festkörperforschung Stuttgart 1997-2001. (29) Ryckaert, J. P.; Ciccotti, G. Introduction of Andersen’s Demon in the Molecular Dynamics of Systems with Constraints. J. Chem. Phys. 1983, 78, 7368-7374. (30) Sprik, M.;

Ciccotti, G. Free Energy from Constrained Molecular Dynamics.

J. Chem. Phys. 1998, 109, 7737-7744. (31) Laio, A.; Parrinello, M. Escaping Free-Energy Minima. Proc. Natl. Acad. Sci. USA 2002, 99, 12562-12566. (32) Iannuzzi, M.; Laio, A.; Parrinello, M. Efficient Exploration of Reactive Potential Energy Surfaces Using Car-Parrinello Molecular Dynamics. Phys. Rev. Lett. 2003, 90, 238302. (33) Barducci, A.; Bonomi, M.; Parrinello, M. Metadynamics. WIREs Comput. Mol. Sci. 2011, 1, 826-843. (34) Lopez, N.; Nøskov, J. K. Catalytic CO Oxidation by a Gold Nanoparticle: A Density Functional Study. J. Am. Chem. Soc. 2002, 124, 11262-11263. (35) Liu, C.; Tan, Y.; Lin, S.; Li, H.; Li, L.; Pei, Y.; Zeng,X. C. CO Self-Promoting Oxidation on Nanosized Gold Clusters: Triangular Au3 Active Site and CO Induced O-O Scission. J. Am. Chem. Soc. 2013, 135, 2583-2595. 19 ACS Paragon Plus Environment


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

(36) Bongiorno, A.; Landman, U. Water-Enhanced Catalysis of CO Oxidation on Free and Supported Gold Nanoclusters. Phys. Rev. Lett. 2005, 95, 106102. (37) Yamazoe, S.; Koyasu, K.; Tsukuda, T. Nonscalable Oxidation Catalysis of Gold Clusters. Acc. Chem. Res. 2014, 47, 816-824. (38) Yoshii, K.; Tuda, T.; Torimoto, T.; Kuwabata, S. Carbon Composite with Pt Nanoparticles Prepared by Room-Temperature Ionic Liquid-Sputtering Method. ECS Transactions, 2010, 33, 127-133. (39) Troullier, N.; Martins, J. L. Efficient Pseudopotentials for Plane-Wave Calculations. Phys. Rev. B 1991, 43, 1993-2006. (40) Goedecker, S.; Teter, M.; Hutter, J. Separable Dual-Space Gaussian Pseudopotentials. Phys. Rev. B 1996, 54, 1703-1710. (41) Hartwigen, C.; Goedecker, S.; Hutter, J. Relativistic Separable Dual-Space Gaussian Pseudopotentials from H to Rn. Phys. Rev. B 1998, 58, 3641-3662. (42) Krack, M. Pseudopotentials for H to Kr Optimized for Gradient-Corrected ExchangeCorrelation Functionals. Theor. Chem. Acc. 2005, 114, 145-152. (43) Perdew, J. P.; Burke, K.; Ernzerhof, M. Generalized Gradient Approximation Made Simple. Phys. Rev. Lett. 1996, 77, 3865-3868. (44) Grimme, S. Accurate Description of van der Waals Complexes by Density Functional Theory Including Empirical Corrections. J. Comput. Chem. 2004, 25, 1463-1473. (45) Koizumi, K.; Boero, M.; Shigeta, Y.; Oshiyama, A. Atom-Scale Reaction Pathways and Free-Energy Landscapes in Oxygen Plasma Etching of Graphene. J. Phys. Chem. Lett. 2013, 4, 1592-1596. (46) Nosé, S. A Molecular Dynamics Method for Simulations in the Canonical Ensemble. Mol. Phys. 1984, 52, 255-268. 20 ACS Paragon Plus Environment

Page 20 of 23

Page 21 of 23

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


(47) Nosé, S. A Unified Formulation of the Constant Temperature Molecular Dynamics Methods. J. Chem. Phys. 1984, 81, 511-519. (48) Hoover, W. G. Canonical Dynamics:

Equilibrium Phase-Space Distributions.

Phys. Rev. A 1985, 31, 1695-1697. (49) Martyna, G. J.; Klein, M. L.; Tuckerman, M. Nosé-Hoover Chains: The Canonical Ensemble via Continuous Dynamics J. Chem. Phys. 1992, 97, 2635-2643. (50) Laio, A.; Gervasio, F. L. Metadynamics: A Method to Simulate Rare Events and Reconstruct the Free Energy in Biophysics, Chemistry and Material Science. Rep. Prog. Phys. 2008, 71, 126601. (51) Ikeda, T.; Boero, M.; Huang, S. F.; Terakura, K.; Oshima, M.; Ozaki, J. Carbon Alloy Catalysts: Active Sites for Oxygen Reduction Reaction. J. Phys. Chem. C 2008, 112, 14706-14709. (52) Ikeda, T.; Boero, M.; Huang, S. F.; Terakura, K.; Oshima, M.; Ozaki, J.; Miyata, S. Enhanced Catalytic Activity of Carbon Alloy Catalysts Codoped with Boron and Nitrogen for Oxygen Reduction Reaction. J. Phys. Chem. C 2010, 114, 8933-8937. (53) Molina, L. M.; Hammer, B. Theoretical Study of CO Oxidation on Au Nanoparticles Supported by MgO(100). Phys. Rev. B 2004, 69, 155424. (54) Han, Y. K. Structure of Au8: Planar or Nonplanar? J. Chem. Phys. 2006, 124, 024316 (55) Xie, S.; Tsunoyama, H.; Kurashige, W.; Negishi, Y.; Tsukuda, T. Enhancement in Aerobic Alcohol Oxidation Catalysis of Au25 Clusters by Single Pd Atom Doping. ACS Catal. 2012, 2, 1519-1523.



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Acknowledgement This research was supported by Grant-in-Aid No. 25288012, ”Elements Strategy Initiative to Form Core Research Center” (since 2012), Strategic Programs for Innovative Research (SPIRE), and Computational Material Science Initiative (CMSI), MEXT, Japan. We thank computer facilities ISSP and Information Technology Center, The University of Tokyo and Research Center for Computational Science, Okazaki, Japan. This research partly used computational resources of the K computer provided by the RIKEN Advanced Institute for Computational Science through the HPCI System Research project (Project ID:hp140224). M.B. thanks Pôle HPC and Equipex Equip@Meso at the University of Strasbourg and GENCI-DARI under allocation No. x2015096092.


Page 22 of 23

Page 23 of 23

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Graphical TOC Entry Reaction Pathway and Free Energy Landscape of Catalytic Oxidation of Carbon Monoxide Operated by a Novel Supported Gold-Copper Alloy Cluster Kenichi Koizumi, Katsuyuki Nobusada, Mauro Boero !"#

$%&'()'*$"&' !+#


acs.jpcc.5b04223

Recommend Documents