Copyright © 2014 American Scientific Publishers Journal of All rights reserved Computational and Theoretical Nanoscience Printed in the United States of America Vol. 11, 511–520, 2014 EERHARTICLE RESEARCH
Adsorption Behavior of Noble Metal Clusters and Their Alloys
Stanley Herrmann1, Michail Stamatakis1 †, Antonis N. Andriotis2, and Giannis Mpourmpakis1 3 ∗
1Department of Chemical and Biomolecular Engineering, University of Delaware, Newark, Delaware 19716, USA 2Institute of Electronic Structure and Laser, FORTH, Heraklion 71110, Crete, Greece 3Catalysis Center for Energy Innovation (CCEI), University of Delaware, Newark, Delaware 19716, USA
Metal-adsorbate bond formation is a fundamental step in catalysis that determines the activity and selectivity of a particular catalyst. The synthesis of novel materials with tailored bond strengths can lead to the design of catalysts with desired functionality. In this work, density functional theory
(DFT) was employed to determine the binding energy of several adsorbates (O, OH, H2O, H, C, CO) on 10-atom metal clusters of the d10 (Cu, Ag, Au) metals, namely the pure metals and their bimetallic combinations (CuAg, CuAu, AgCu, AgAu, AuCu, AuAg). A database of the adsorbate binding energies on these clusters was constructed, and descriptor-based models were developed, based on key electronic (d-band, electronegativity, charge) and structural (strain) properties of the clusters. In addition, the fundamental physicochemical properties of the catalysts that control metal- adsorbate bonding areIP: elucidated. 192.168.39.211 On: Sat, 02 Oct 2021 05:15:29 Keywords: Bimetallic, DFT,Copyright: Theory, American Descriptors, Scientific Binding. Publishers Delivered by Ingenta
1. INTRODUCTION metals they are made of.6 22 Once established, models of A fundamental understanding of catalyst-adsorbate inter- binding energy (BE) can be used with Brønsted–Evans– actions is vital for a wide range of fields including Polanyi type correlations to estimate reaction energy environmental chemistry and heterogeneous catalysis, as profiles16 23 and gain an understanding on a catalyst’s per- well as for biomedical, industrial, and electrochemical formance, such as activity and selectivity. applications.1–5 This area of research is of particular inter- Aside from periodic DFT, molecular orbital-based clus- est due to its potential in the rational design of catalysts ter DFT provides other means to simulate catalytic toward renewable energy production.6–8 behavior. One key advantage of molecular orbital based The immense computational power available nowadays cluster DFT is its ability to account for the size and shape facilitates the investigation of nanomaterials properties, of a nanoparticle in an efficient way, as well as its detailed among which are heterogeneous catalysts, through first- interaction mapping through molecular orbitals (LCAO principles calculations.9–15 Periodic density functional theory). Particularly in catalysis, it is often the case that theory (DFT) is commonly used to understand catalyst- reactions exhibit structure sensitivity and therefore, mod- adsorbate interactions. In particular, Nørskov and co- els accounting for different catalytic sites, such as cor- 8 workers16–18 have shown that trends in the binding of an ners, edges and planes of a nanoparticle are essential. adsorbate over a range of transition metal catalysts can be A major disadvantage of DFT is its computational cost, explained on the basis of the d-band center (dC )ofthe especially when simulating nanoparticles larger than 1 nm metal. Such trends have also been reported for near-surface in diameter. Since the computational cost of DFT increases alloys;19–21 the latter have become particularly important with the nanoparticle size, one must resort to efficient due to their unique properties compared to the parent schemes that can capture the desired features of the sys- tem being modeled. In previous work, we introduced a ∗Author to whom correspondence should be addressed. computational framework that makes use of small clus- †Present address: Department of Chemical Engineering, University ters to accurately represent the properties of the bulk College London, Torrington Place, London WC1E 7JE, UK. metal.24
J. Comput. Theor. Nanosci. 2014, Vol. 11, No. 2 1546-1955/2014/11/511/010 doi:10.1166/jctn.2014.3387 511 Adsorption Behavior of Noble Metal Clusters and Their Alloys Herrmann et al.
Herein, we expand the aforementioned work to study Spin unrestricted DFT calculations were performed 31 the binding of several adsorbates (O, OH, H2O, H, C, and using the software package Gaussian 03. The B3LYP CO) on the d10 pure metals (Cu, Ag, and Au: d10s1) and functional32 33 along with the LANL2DZ basis set were their bimetallic alloys (CuAg, CuAu, AgCu, AgAu, AuCu, applied. First, bare cluster optimization was performed and AuAg) and provide detailed analysis on developing during which the bond angles and dihedrals were frozen, empirical binding energy models. These adsorbates are of whereas the bond distances were allowed to relax. This particular interest in numerous catalytic reactions, such optimization scheme ensures that the cluster retains a as the water-gas shift reaction and the CO oxidation.25–30 3D bilayer structure. Each one of the 10-atom clusters Using molecular orbital-based cluster DFT calculations, was optimized at three multiplicities: singlet, triplet, and BE trends of these adsorbates were investigated and pos- quintet. sible descriptors characterizing the electronic and strain Subsequently, cluster-adsorbate interactions were stud- effects of the metal cluster were identified. Four descrip- ied. The adsorbates can interact with the catalyst at three RESEARCH ARTICLE tors were used in this study: different sites: hollow (O, OH, H, and C), bridge (O, OH, d (i) the -band center, H, C, and CO), and top (O, OH, H2O, and CO) (Fig. 1(a)). (ii) the strain of the surface layer due to the bimetallic To optimize the adsorbate on the catalyst, the optimized composition, bare cluster was frozen and the distance between the adsor- (iii) the difference in the electronegativities of the metals bate and the cluster was relaxed. For adsorbates consisting composing the cluster, and of more than one atom (i.e., OH), the atoms not directly (iv) the charge transfer between the surface and the sub- involved in binding (i.e., H of OH) were allowed to fully surface plane of the cluster. relax during the calculations. This study provides valuable insights into the physico- The BE of the adsorbates was calculated as follows: chemical properties of the metal clusters that markedly BE = E − E − E affect binding, and unravels the binding behavior of the A M–A M A (1) adsorbates through detailed electron density analysis. E E E where M–A, M , and A are the total electronic energies of the metal-adsorbate system, of the metal cluster, and of 2. COMPUTATIONAL METHODS the adsorbate, respectively. Small clusters, consisting of 10 atoms,IP: 192.168.39.211 were used to simu- On: Sat, 02The Oct methodology 2021 05:15:29 followed in this work results in highly late the metal catalyst structure. Each clusterCopyright: consisted American of a Scientific Publishers Delivered byefficient Ingenta (fast) calculations without sacrificing accuracy, 7-atom upper layer and a 3-atom lower layer (Fig. 1). This as shown by detailed test calculations that validated our configuration has one central surface atom with a coordi- computational scheme. In these calculations, an adsor- nation number of 9 which mimics the face-centered cubic bate was bound to the central top-site and the central (111) plane of the bulk metal. For the bimetallic clusters, metal atom was allowed to fully relax during the opti- the two layers are composed of different metals, represent- mization. The calculated BEs were less than 2 kcal/mol ing a near-surface alloy structure as seen in Figure 1(b). different than the ones involving a completely frozen M M M These clusters will be referred to as T B, where T cluster. denotes the metal composition of the top layer and MB denotes the metal composition of the bottom layer. 3. RESULTS AND DISCUSSION 3.1. Calculation of Binding Energies Bridge Table I presents the BEs for the most stable binding Top Hollow M configurations of each adsorbate on every cluster. A table MT T containing the BEs of all configurations investigated is MT MT provided in the supplementary information. The calculated MT BEs are in good agreement with periodic DFT calculations < MT MT reported in the literature (difference 6 kcal/mol), shown in Table II. With the exception of C, the BEs of adsorbates show a similar trend with respect to the clusters’ top layer com- MB MB position. These adsorbates exhibit the strongest binding on clusters with a Cu top layer and the weakest binding on MB clusters with a Au top layer. This will be referred to as the “primary trend” which can be summarized as: Fig. 1. 10-atom cluster model and adsorbate binding sites. MT and MB indicate the type of the metal constituting the top and bottom layer of the cluster respectively. In monometallic clusters, MT is the same as MB . BE[Cu top layer] > BE[Ag top layer] > BE[Au top layer]
512 J. Comput. Theor. Nanosci. 11, 511–520, 2014 Herrmann et al. Adsorption Behavior of Noble Metal Clusters and Their Alloys
Table I. Calculated binding energies of adsorbates in kcal/mol. The positive BEs values in parenthesis are provided because they were included in the input used to construct the models. EERHARTICLE RESEARCH Adsorbate Site CuAu CuAg CuCu AgAu AgAg AgCu AuAu AuAg AuCu
O Hollow −96 80 −90 98 −82 32 −59 60 −56 00 −51 41 −30 02 −25 08 −20 12 OH Hollow −85 31 −77 64 −70 72 −61 19 −55 94 −52 02 −30 64 −25 06 −21 05 − − − − − − − − H2OTop9 45 8 79 6 36 5 74 5 66 4 53 5 00 5 26 13 62 H Hollow −53 19 −47 81 −43 35 −39 80 −34 87 −31 84 −26 96 −28 68 −25 54 C Hollow −71 26 −69 30 −60 53 −39 55 −38 23 −34 67 −57 64 −56 06 −45 03 CO Top −8 63 −6 92 −2 66 −2 41 −2 28 −2 72 1 35 −2 92 2 45
Since the cluster consists of two layers, we also identify characteristics of the clusters. These characteristics can “secondary trends” pertaining to clusters with the same top be expressed by descriptors designed to capture the driv- layer composition (MT ): ing forces of adsorbate binding. Subsequently, quantitative models based on each adsorbate’s essential descriptors can M > M > M BE T Au BE T Ag BE T Cu be derived to accurately predict the BEs. or more specifically We considered four such descriptors, namely, (i) the dC , CuAu > CuAg > CuCu (all adsorbates) (ii) the top layer strain, (iii) the electronegativity difference between the metals AgAu > AgAg > AgCu (all adsorbates except CO) forming the bimetallic clusters, and AuAu > AuAg > AuCu (O, OH, C) (iv) the charge transfer between surface and sub-surface planes of the clusters. The energy differences of the secondary trends are much smaller than those of the primary trend. These energy dif- These descriptors account for the electronic interactions ferences become even smaller on weakly bound systems between the layers of the metal cluster, as well as the strain effect caused by the changes in the bond distances of the such as CO and H2O. As a result, the overall ordering of BEs on all clusters could be summarized as follows: top layer, since all of these factors have been reported to IP: 192.168.39.211 On: Sat,affect 02 Oct binding. 202116–18 05:15:29 21 24 In Figure 3 we present linear rela- CuAu > CuAg > CuCu > AgAu >Copyright:AgAg > AgCu American Scientifictionships Publishers between the BE of all the adsorbates and the Delivered byaforementioned Ingenta descriptors on clusters with Cu as the top > AuAu > AuAg > AuCu metal. Similar relationships hold for the clusters consist- This trend will be referred to as the “common trend” ing of Ag (AgAg, AgCu, AgAu) and Au (AuAu, AuAg, throughout the rest of this manuscript. Figure 2 shows the AuCu) as the top metal. 16–21 BEs presented in Table I versus the BE of atomic O on As suggested in literature, the dC was used as one the hollow site, which exhibits the common trend. This descriptor of these properties since it is the footprint of figure clearly shows the primary trend which is similar the electronic properties of the metal cluster. The dC can among all adsorbates except C and captures the common trend behavior (small BE deviations exist as we mentioned earlier).
3.2. Descriptor Development Based on the Physicochemical Properties of Clusters Having identified these BE trends, we are inter- ested in explaining them based on the physicochemical
Table II. Comparison of binding energies of this study to published periodic DFT calculations. Values are in kcal/mol.
Adsorbate/ Binding BE BE cluster site (our work) (literature) Refs.
O/AgAg Bridge −42 70 −48 43 34 OH/CuCu Hollow −70 72 −65 03 −65 72, 35 28 −73 33 −73 79 36 37 H O/CuCu Top −6 36 −4 84 36 2 Fig. 2. Binding energies of adsorbates versus the binding energy of H O/AuAu Top −5 00 −3 23 36 2 atomic O (the symbols in circle show the major binding energy differ- C/AgAg Hollow −38 23 −36 90 34 ences of C from the primary trend).
J. Comput. Theor. Nanosci. 11, 511–520, 2014 513 Adsorption Behavior of Noble Metal Clusters and Their Alloys Herrmann et al.
(a) the (bare) cluster, and is presented in Table III. A detailed description of the dC calculation on these clusters can be found in Ref. [24]. CuCu CuAg CuAu Further, we considered additional, secondary descrip- tors that explicitly quantify geometric and charge transfer processes on the clusters. The geometric properties were quantified by the strain effect. The charge transfer was accounted through (i) the electronegativities of the metals, and (ii) the charge density distribution between the two layers of the cluster. (b) These secondary descriptors are also provided in Table III. RESEARCH ARTICLE To assess the strain effect, the percent change in the average top layer bond length for a bimetallic cluster with CuCu CuAg
CuAu respect to the monometallic cluster of the same top layer composition was used. This was calculated using Eq. (2). a − a D = 0 · s a 100 (2) 0 Here, a represents the average bond length of the top layer a of the cluster, and 0 is the average bond length of the top layer of the monometallic cluster with the same top layer composition. This descriptor has a value of zero when cal- culated for the monometallic clusters. The values for a are (c) provided in Table IV, and Figure 3(b) shows the BE of the adsorbates versus this descriptor. The final two descriptors capture the electron distribu- IP: 192.168.39.211 On: Sat,tion 02 onOct the 2021 metal 05:15:29 cluster. The first one is the difference Copyright: American Scientificin the electronegativities Publishers of the elements that compose the Delivered bytwo Ingenta layers of the cluster, E , defined as: = − E top bottom (3) where the subscripts top and bottom refer to the metals of the corresponding layers. From this equation, it is appar- ent that this descriptor has a non-zero value only for the bimetallic clusters, as seen in Table III. Electronegativity, , is a measure of the tendency of an atom to attract electrons, and for the purposes of our study the Pauling scale was used. The electronegativities of the d10 metals (Cu, Ag, Au) are as follows: Fig. 3. Interaction trends on clusters consisting of Cu as a top layer: d (a) BE of adsorbates versus the C of the clusters, (b) BE of adsor- <