- Email: [email protected]
Adsorption of NO on Au atoms and dimers supported on MgO(1 0 0): DFT studies
Surface Science 602 (2008) 1669–1676
Contents lists available at ScienceDirect
Surface Science journal homepage: www.elsevier.com/locate/susc
Adsorption of NO on Au atoms and dimers supported on MgO(1 0 0): DFT studies Silvia A. Fuente a, Patricia G. Belelli a, Ricardo M. Ferullo a,b,*, Norberto J. Castellani a a b
Grupo de Materiales y Sistemas Catalíticos, Departamento de Física, Universidad Nacional del Sur, Av. Alem 1253, 8000 Bahía Blanca, Argentina Departamento de Química, Universidad Nacional del Sur, Av. Alem 1253, 8000 Bahía Blanca, Argentina
a r t i c l e
i n f o
Article history: Received 26 December 2007 Accepted for publication 29 February 2008 Available online 10 March 2008
Keywords: Au/MgO Gold clusters NO adsorption Nitric oxide DFT Cluster model
a b s t r a c t The adsorption of NO on single gold atoms and Au2 dimers deposited on regular O2 sites and neutral oxygen vacancies (Fs sites) of the MgO(1 0 0) surface have been studied by means of DFT calculations. For Au1/MgO the adsorption of NO is stronger when the Au atom is supported on an anionic site than when it is on a Fs site, with adsorption binding energies of 1.1 and 0.5 eV, respectively. In the first case the spin density is mainly concentrated on the metal atom and protruding from the surface. In such a way, an active site against radicals such as NO is generated. On the Fs site, the presence of the vacancy delocalizes the spin into the substrate, weakening its coupling with NO. For Au2/MgO, as this system has a closed-shell configuration, the NO molecules bonds weakly with Au2. Regarding the N–O stretching frequencies, a very strong shift of 340–400 cm1 to lower frequencies is observed for Au1/MgO in comparison with free NO. Ó 2008 Elsevier B.V. All rights reserved.
1. Introduction In heterogeneous catalysis the reactive properties of supported metal particles depend on the cluster size and on the specific site of the support where the particle is anchored. When an oxide is used as support, the metal aggregates interact preferentially with surface oxygen vacancies [1,2]. In particular, on MgO these vacancies comprise several different types of surface defects such as Fs, Fs+ and Fs2+, depending on removal of a neutral O atom, an O anion and an O2 anion, respectively. The interaction of transition metal atoms with both regular and defective MgO was investigated using quantum chemical approaches. It was observed that the adhesion is stronger on Fs sites than on regular anionic sites [3]. The activity and selectivity of supported Au aggregates was intensively studied during the last years. They show an extraordinary activity to catalyze different reactions at low operating temperatures, such as the oxidation of CO, the selective oxidation of propene and the reduction of NOx [4]. In experiments, TiO2 is used as the typical support, but other oxides like MgO, SiO2 and Al2O3 were also considered [5–7]. Recently, experiments were performed in order to clarify the importance of Fs sites of MgO in the adhesion of gold. It was found a correlation between the catalytic activity of the supported Au
* Corresponding author. Address: Grupo de Materiales y Sistemas Catalíticos, Departamento de Física, Universidad Nacional del Sur, Av. Alem 1253, 8000 Bahía Blanca, Argentina. Tel.: +54 291 4595141; fax: +54 291 4595142. E-mail address: [email protected] (R.M. Ferullo). 0039-6028/$ - see front matter Ó 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.susc.2008.02.037
particles and the concentration of Fs centers on MgO. Thus, these surface defects play a direct role in the activation of supported Au catalysts [1]. In another recent work, the electron paramagnetic resonance (EPR) and scanning tunnelling microscopy (STM) techniques were combined to verify the existence of surface defects as well as the nucleation of Au onto these defects [2]. Moreover, by using infrared (IR) spectroscopy of adsorbed CO, two bands were observed: one at 2120 cm1 and a broad one around 2070 cm1. While the former was assigned to CO adsorbed on small neutral Au clusters deposited on regular anionic sites, the broad band was assigned to CO on negatively charged Au clusters. These charged particles are predicted to be produced by the adsorption of Au on defect sites of MgO. The larger red-shift is attributed to a greater p-backdonation from the negatively charged Au in comparison with that of neutral Au. The deposition of Au atoms and clusters on regular five-coordinated O2 sites and on Fs centers of the MgO(1 0 0) surface was recently studied from periodic DFT calculations [8–13]. It was observed that the presence of the vacancy enhances the metal– oxide interaction. Thus, these centers are essential to trap the diffusing Au atoms or clusters [9]. In the past, NO was used as a probe molecule to study the metal surface reactivity [14]. Besides, the bonding of NO to transition metal surfaces is of interest in relation to the catalytic reduction for pollution control. At ultrahigh vacuum (UHV) conditions, NO adsorbs weakly on Au(1 1 1) [15] but it decomposes on Au(3 1 0) giving N2O and atomic O on this stepped surface [16]. Gold supported on several metal oxides shows good catalytic performance for the
1670
S.A. Fuente et al. / Surface Science 602 (2008) 1669–1676
reduction of nitric oxide by hydrocarbons [7]. In these cases, the addition of CO and H2 to the reactant gas improves the conversion of NO to N2. The NO + CO reaction was also studied for Au supported on MgO, TiO2, SiO2 and Al2O3 [5]. These systems catalyze the reaction above 573 K yielding surface NCO. The interaction of NO with isolated Au atoms was object of recent studies. Laser-ablated Au atoms interact with NO yielding neutral nitrosyl complexes AuNO and Au(NO)2 [17]. For AuNO, the N–O and Au–N stretching modes are observed at about 1700 and 520 cm1, respectively. On the other hand, the reaction of NO with free neutral Aux clusters (x = 1–6) was studied from DFT [18]. The NO + Aux reaction is exothermic, releasing 0.2–0.8 eV. The bond between NO and Au is accompanied by an electron transfer from the metal cluster to the NO molecule. The first objective of this work is to analyze the deposition of one Au atom and the Au2 dimer on anionic regular sites and on Fs centers of the MgO(1 0 0) surface by employing an embedded cluster approach. The results are compared with recent periodic calculations [8–13]. While the Au–MgO interface was thoroughly studied, the adsorption properties of supported Au particles received relatively less attention. Specifically, in this work we are interested to study the adsorption of NO on supported gold which constitutes a valuable reference for other researchers. The corresponding results are compared with the reaction of NO with free Au particles in order to study the effect of the support on the adsorption capability of supported Au. 2. Computational details Density functional theory (DFT) molecular orbital calculations were carried out using the gradient corrected Becke’s three parameters hybrid exchange functional in combination with the correlation functional of Lee, Yang and Parr (B3LYP) [19]. This method was widely used in the past to study adsorption processes yielding reliable results both on oxides and metal clusters. The MgO(1 0 0) surface was represented by clusters of 26 atoms, Mg13O13, consisting of two layers (first layer: Mg4O9; second layer: Mg9O4). To take into account the Madelung field due to the rest of the extended surface, the cluster was embedded in an array of ±2 point charges. Moreover, the positive point charges at the interface were replaced by effective core potentials (ECP) corresponding to Mg2+ to account for the finite size of the cations and to avoid spurious charge polarization. The Mg13O13 cluster plus the 16 Mg– ECP’s is represented in Fig. 1. This methodology was widely used by the group of Pacchioni and other authors [20], giving results which are in good agreement with those obtained by periodic calculations. For example, if the embedded cluster model and the slab
approach are compared, the energy for creating an oxygen vacancy is only 1% different [21,22]. The 6-31+G(d) basis set was employed for the atoms of the NO molecule and the O atoms in the first layer of Mg13O13 cluster, except the terminal ones. The atomic orbitals of the four Mg atoms in the first layer were described by the 6-31G(d) basis set, while those of the terminal O atoms and those of Mg and O atoms at the second layer, by the 6-31G basis set. For Au, the LANL2DZ basis set was used, which describes the 60 core electrons with a pseudopotential and the 19 valence 5s25p65d106s1 electrons with a [5s,6p,4d/ 3s,3p,2d] basis set [23,24]. Previously, the suitability of this basis set for the Au–NO interaction was examined by considering the isolated Au1NO molecule. For that purpose, the results of calculation using the LANL2DZ basis set was compared with that obtained with a larger one such as the SDD (Stuttgart–Dresden) [25]. The last describes the 19 valence electrons with a [8s,7p,6d/6s,5p,3d] basis set. The resultant distances, angles and frequencies values are very similar using both descriptions. During the geometrical optimization procedure, the coordinates of atoms belonging to the adsorbed NO molecule and the metal particle, and those of the five central atoms in the first layer of Mg13O13 cluster (the central O and its four Mg atoms directly bonded to it) were fully optimized without imposing geometric constrains. The geometries studied in this work were selected taking into account previous theoretical results [9–13]. With respect to the gold monomer deposited on the regular surface, only the adsorption on the anionic site was studied because it is well known that this is the preferred adsorption site for most transition metal atoms [26]. For the Au dimers on the perfect surface, the on-top geometry on one anionic site was considered following the results obtained in Ref. [12]. Regarding the defective surface several starting orientations for Au2 were taken into account. A neutral O vacancy was represented by Mg13O12 cluster, with the vacancy localized in the center of the first layer. In this case, the geometrical optimization included the four central Mg atoms of first layer. We define the binding energy of the Aux particle on MgO as EB(Aux) = [E(Aux/MgO, site) E(Aux) E(MgO, site)], with x = 1, 2, and site = O2 or Fs. Besides, we define the binding energy of NO molecule on Aux/MgO as EB(NO) = [E(NO/Aux/MgO, site) E(Aux/MgO, site) – E(NO)]. In both cases, positive values correspond to exothermic processes. The reactions energies were corrected by considering the basis set superposition error (BSSE), calculated according to the counter-poise correction [27]. The frequency values were scaled according to the factor of 0.9465, calculated as the ratio between the empirical and the calculated frequency values of free NO (1876/1982).
Fig. 1. Mg13O13 cluster plus the 16 Mg–ECP used for modeling the MgO(1 0 0) surface. Dark grey spheres: O. Light grey spheres: Mg. Small spheres represent the Mg–ECP’s. Point charges are not shown.
S.A. Fuente et al. / Surface Science 602 (2008) 1669–1676
The spin density (SD) was expressed in terms of the Mülliken population analysis. On the other hand, the atomic net charges were calculated following the natural bond orbital (NBO) scheme [28], which gives realistic values for the charge partitioning. The NBO approach was recently used to analyze the interaction of CO and NO with nanoparticles containing coinage metals [17,18,29]. In particular, it was very useful for the interpretation of IR frequency shifts of the AuNO complex [17]. The multiplicity was fixed to M = 1 in Au2/MgO and NO/Au1/ MgO, and to M = 2 in Au1/MgO and NO/Au2/MgO. For all the systems the total charge was zero. All the calculations have been performed using the Gaussian-03 program package [30]. 3. Results and discussion 3.1. Au1 and Au2 particles deposited on MgO(1 0 0) It was well established that small metal particles adsorb preferentially on sites where negative charge accumulates [8– 13,26,31,32]; more specifically, the O2 ionic sites for regular MgO, and the Fs and Fþ S centers for defective MgO. Recently, the deposition of Au1 and Au2 on O2 and Fs sites of MgO(1 0 0) was studied by means of periodic calculations [8–13] and the embedded cluster approach [31]. As it will be shown later, the interaction of NO with supported Au depends strongly of the metal–oxide interaction and it is essential to dispose of an adequate substrate model for the subsequent NO adsorption. Hence, as a first step in our approach, the Au1/MgO(1 0 0) and Au2/MgO(1 0 0) systems were studied using the above detailed cluster models and the results were compared with those obtained with the periodic method. In Table 1, the calculated values of binding energies, atomic charges and spin densities for the deposition of the Aux particle (x = 1, 2) on a regular oxygen site and on a Fs center are reported. For simplicity, these surface sites will be denoted as ‘‘MgO(O2)” and ‘‘MgO(Fs)”, respectively. In Fig. 2 a view of the corresponding optimized geometries is presented. From the values of Table 1 we notice that one Au atom interacts much stronger on the neutral O-vacancy (EB(Au1) = 2.93 eV) than on the surface O2 site (EB(Au1) = 0.73 eV). These values are only 0.2–0.3 eV lower than those obtained using slab calculations and the PW91 exchange-correlation functional (3.17 and 0.89 eV from Ref. [9], and 3.04 and 0.87 eV from Ref. [10], respectively). The difference could be explained by the general tendency of the hybrid
Table 1 Main energetic and electronic population parameters for Aux particles deposited on MgO. Energies are expressed in eV and charges in electron units O2 site EB (Au1)a q(Au) q(O2) SD(Au) SD(O2) EB(Au2)a Edimb q(Aut)c q(Aus)c q(O2) a
Au1/MgO 0.73 0.13 1.74 0.76 0.26 Au2/MgOd 1.34 2.56 0.28 +0.14 1.76
Fs site 2.93 0.88 – 0.44 – 3.86 2.88 0.67 0.61 –
EB(Aux) = [E(Aux/MgO, site) E(MgO, site) E(Aux)]; x = 1, 2; site = O2 or Fs. Edim = [E(Au2/MgO, site) E(Au1/MgO, site) E(Au1)]; site = O2 or Fs. c Aut: terminal Au atom; Aus: Au atom closer to the surface. d The values for Au2 on the Fs site correspond to the configuration normal to the surface. b
1671
B3LYP functional to predict lower binding energies than the PW91-GGA functional. According to the NBO analysis, when the Au atom interacts with anionic sites a charge transfer of 0.13e occurs from the surface to the metal atom (Table 1). Results from EPR spectroscopy of Au deposited on a smooth MgO(1 0 0) film suggest that single Au atoms are basically neutral [32]. On the other hand, this charge transfer is much greater when the metal atom interacts with the O-vacancy (0.88e, Table 1), a phenomenon that can be explained by the high polarizability of the charge inside the vacancy and the high electron affinity of gold [9]. To analyze the charge distribution after deposition, we have calculated the charge density difference Dq = q(Au1/MgO) q(MgO) q(Au1), by subtracting from the total charge density of the Au1/MgO system the densities of the component fragments, and mapped it on the plane that contains the metal atom (see Fig. 3). For the MgO(O2) site, we can observe a depletion of the charge density on the O anion and a charge accumulation around the Au atom (Fig. 3a). As it was discussed by Del Vitto et al. [9], the lack of accumulation of electron density in the region between the Au atom and the oxide surface indicates that the intraunit polarization of Au atom makes an important contribution to the Au–MgO bonding. For the MgO(Fs) site, a depletion of charge inside the cavity was observed, together with an important charge accumulation all around the Au atom, including the interface region (Fig. 3b). The spin density values indicate that, for the MgO(O2) site, the spin is localized mainly on the Au atom (SD(Au) = 0.76). On the other hand, for the MgO(Fs) site, the spin density of the Au atom is lower (SD(Au) = 0.44). This difference in the spin localization is clear by comparing the corresponding spin density maps shown in Fig. 4. While at the regular surface the spin is distributed around the Au atom and around the surface O anion (Fig. 4a), at the defective surface it spreads out around the Au atom and inside the cavity (Fig. 4b). The higher spin density for the Au atom linked to the MgO(O2) site is in agreement with calculated EPR properties [33] and it is very important for the understanding of the adsorptive capability of the supported Au atom towards a radical like NO (see below). Regarding the Au dimer deposited on the MgO(O2) site, it adopts an orientation that is normal to the surface (Fig. 2b), with a binding energy value which nearly doubles the value for the monomer (EB(Au2) = 1.34 eV). On the Fs center, the calculations predict two different configurations: normal to the surface and tilted towards a Mg cation, forming an angle of 53° with respect to the normal (Fig. 2d and e, respectively). These structures are practically isoenergetic: the perpendicular geometry has a binding energy (EB(Au2) = 3.86 eV) which is only 0.06 eV greater than that for the tilted geometry. These energies are nearly 1 eV greater than that for the monomer. The corresponding EB(Au2) values obtained from periodic calculations by Del Vitto et al. [9] are 1.49 and 3.89 eV, for the adsorption on the regular surface and on Fs, respectively (in Ref. [9] the tilted geometry is somewhat more stable than the normal one). Concerning the Au–Au distance, it is practically unperturbed on the MgO(O2) site and stretched by 0.17 Å (0.25 Å) on the MgO(Fs) site for perpendicular (tilted) orientation (the calculated interatomic distance is 2.57 Å for free Au2). Notice that the dimer as a unit adsorbs much more strongly on the MgO(Fs) site than on the MgO(O2) site. Besides the EB(Au2) binding energy we can define the dimer formation energy, Edim, as the energy associated with the dimer formation when a free metal atom interacts with a pre-deposited atom, Edim = [E(Au2/MgO, site) E(Au1/MgO, site) E(Au1)]; site = O2 or Fs. While the binding energy of Au2, EB(Au2), is noticeably affected by the support, the dimer formation energy is weakly affected; for both sites the values of Edim are quite similar (2.56 eV on surface O2 and 2.88 eV on Fs).
1672
S.A. Fuente et al. / Surface Science 602 (2008) 1669–1676
Fig. 2. (a) and (b): schematic representation of Au1 and Au2 particles, respectively, deposited on regular MgO(O2) site. (c) and (d): schematic representation of Au1 and Au2 particles, respectively, deposited on defective MgO(Fs) site. Dark grey spheres, oxygen atoms. Light grey spheres, Mg atoms (including the Mg–ECP’s). Large grey spheres, Au atoms. In 2a, the distance between Au and its nearest Mg is 3.241 Å. In 2b, the distance between the Au atom closer to the surface and its nearest Mg is 3.174 Å.
a
b
c
d
Fig. 3. (a) and (b): charge density difference, Dq = q(Au1/MgO) q(MgO) q(Au1), for Au atom deposited on regular MgO(O2) site and defective MgO(Fs) site, respectively; (c) and (d): charge density difference, Dq = q(Au2/MgO) q(MgO) q(Au2), for Au2 aggregate deposited on regular MgO(O2) site and defective MgO(Fs) site, respectively. Dq is mapped along a plane containing the Au atoms. Dotted lines correspond to regions with Dq < 0, and solid lines to regions with Dq > 0. Contour levels for mapping are: ±1.0, ±0.5, ±0.25 . . . ±0.00012 a.u.
The charge of Au2 supported on the MgO(O2) site is 0.14e, practically the same than that of the supported monomer. Within the dimer, this charge is inhomogeneously distributed with an accumulation towards the terminal Au. On the MgO(Fs) site, the dimer acquires a much more significant negative charge, 1.28e,
which is even higher than that taken by the monomer (nearly 0.40e higher). In this case the charge is distributed quasi-homogeneously in both Au atoms. The spatial charge distribution can be analyzed looking at the density difference maps of Fig. 3c and d, evaluated with respect to the separated fragments, MgO and Au2.
1673
S.A. Fuente et al. / Surface Science 602 (2008) 1669–1676
a
Table 3 Main molecular and electronic population parameters for NO adsorbed on a Au2 dimer deposited on MgO. Energies are expressed in eV, charges in electron units and frequencies in cm1
b
a
EB(NO) SD(O) SD(N) SD(Aut)b SD(Aus)b SD(O2) q(O) q(N) q(Aut)b q(Aus)b q(O2) mNO (DmNO) a b
NO/Au2/MgO(O2)
NO/Au2/MgO(Fs)
0.23 0.31 0.63 0.00 0.05 0.01 0.20 +0.07 0.00 0.00 1.77 1716 (160)
0.20 0.32 0.75 0.00 0.00 – 0.29 +0.03 0.40 0.65 – 1622 (254)
EB(NO) = [E(NO/Au2/MgO, site) E(Au2/MgO, site) E(NO)]; site = O2 or Fs. Aut: terminal Au atom; Aus: Au atom closer to the surface.
Fig. 4. (a) and (b): spin density for Au atom deposited on regular MgO(O2) site and defective MgO(Fs) site, respectively. Contour levels for mapping are: 0.17, 0.085, 0.0425 . . . ±0.00017 a.u.
To allow a comparison for Au2 on both sites, only the normal geometrical configuration was considered. On the anionic site, the charge is concentrated around the terminal Au atom and towards the vacuum. On the O-vacancy, the charge distributes also in this way and, in addition, an important accumulation of charge is observed in the region between the bottom Au atom and the cavity. 3.2. Adsorption of NO on supported Au1 and Au2 In Tables 2 and 3, the calculated values of adsorption energies, atomic charges and spin densities for the adsorption of NO on Au1/MgO and Au2/MgO are reported, respectively. In Fig. 5, the corresponding optimized geometries are presented. Looking at Table 2, we observe that NO adsorbs much more strongly when the Au monomer is deposited on the MgO(O2) site than on the MgO(Fs) defect (EB(NO) = 1.08 eV and 0.49 eV, respectively). The reason of this behavior will be discussed later. Interestingly, when NO reacts with an unsupported Au atom, the binding energy is 0.69 eV; therefore, the regular anionic site acts favouring the Au–NO bond and the Fs site weakens that interaction. The NO molecule bonds with Au through its N atom and adopts a tilted orientation with a Au–N–O angle having 109 and 117° on MgO(O2) and MgO(Fs), respectively. The N–O distance elongates from its value of 1.158 Å for the free NO molecule to 1.204 and 1.220 Å for
Table 2 Main molecular and electronic population parameters for NO adsorbed on one Au atom deposited on MgO. Energies are expressed in eV, charges in electron units and frequencies in cm1
a
EB(NO) EB(Au1NO)b q(O) q(N) q(Au) q(O2) mNO (DmNO) a b
Fs .
NO/Au1/MgO(O2)
NO/Au1/MgO(Fs)
1.08 1.00 0.30 0.07 +0.26 1.76 1540 (336)
0.49 2.62 0.35 0.22 0.50 – 1479 (397)
EB(NO) = [E(NO/Au1/MgO, site) E(Au1/MgO, site) E(NO)]; site = O2 or Fs. EB(Au1NO) = [E(NO/Au1/MgO, site) E(MgO, site) E(Au1NO)]; site = O2 or
Fig. 5. (a) and (b): schematic representation of NO adsorbed on Au1 and Au2 particles, respectively, deposited on regular MgO(O2) site. (c) and (d): schematic representation of NO adsorbed on Au1 and Au2 particles, respectively, deposited on defective MgO(Fs) site. Dark grey spheres, oxygen atoms. Light grey spheres, Mg atoms. Large grey spheres, Au atoms. Black spheres, nitrogen atoms.
Au1/MgO(O2) and Au1/MgO(Fs), respectively (for gas-phase AuNO this distance is 1.161 Å). On the other side, the Au–N distance changes from 2.105 for AuNO to 2.058 and 2.340 Å when the metal atom is supported on MgO(O2) and MgO(Fs), respectively (see Fig. 5a and b). Notice that the more significant elongation of Au–N distance for MgO(Fs), in comparison with MgO(O2), is in
1674
S.A. Fuente et al. / Surface Science 602 (2008) 1669–1676
agreement with a relatively less significant EB(NO) value (0.59 eV lower). Very recently, several authors were interested in studying the CO-induced modification of the metal–MgO interaction [34,35] in the form of atoms and layers. For instance, it was found that the bonding between Pt or Pd atoms and the oxide is enhanced by CO, but for Au this effect is negligible [35]. To allow a similar analysis on NO, the adsorption of the AuNO complex on MgO(1 0 0) was considered here. The corresponding binding energy can be defined in a similar way as that for Aux on MgO. The calculated EB(Au1NO) values were 1.00 and 2.62 eV for the adsorption on MgO(O2) and MgO(Fs), respectively. If we compare the binding energies for AuNO (see Table 2) and that for Au (see Table 1) we observe that the Au–oxide interaction is enhanced in the former case and that it is weakened in the latter one. Concerning the charge distribution, the values of atomic charges indicate that the charge of the AuNO complex on the MgO(O2) site is approximately the same than that of supported Au atom, so that the charge transfer from the oxide to the adsorbate (Au atom or surface AuNO complex) is very similar. However, the NO molecule withdraws electronic charge from the metal atom which becomes positive, +0.26e. On the contrary, the AuNO complex on Fs takes more electronic charge from the cavity than the supported Au atom, the former resulting 0.19e more negative. This excess of negative charge accumulates on the NO, which reaches a net atomic charge of 0.57e. The charge gained by NO populates the 2p* (antibonding) orbital, weakening the N–O bond. The larger stretching of the NO interatomic distance on the Fs site can be explained by this higher electron transfer. In Fig. 6, the corresponding maps of charge density differences are shown, calculated with respect to MgO and AuNO fragments. As a consequence of AuNO adsorption on both MgO(O2) and MgO(Fs) sites, it is observed an accumulation of charge on NO, in agreement with the above commented net atomic charges (at free AuNO molecule, the Au ? NO electronic transfer is very low, 0.06e). On the MgO (O2) site, it can be observed a region of depletion of charge around the Au atom and the presence of a little accu-
a
mulation of charge between Au and O2, an indication of an enhancement of the covalent contribution. On the MgO(Fs) site, the region surrounding the Au atom loses electronic charge while it is retained in the interface region towards the cavity. The NO interaction with a gold aggregate can be viewed within the donation/backdonation scheme. While the electron donation occurs from the single occupied 2p* antibonding orbital of NO to the dr + sp hybridized orbitals of Au, the corresponding backdonation is produced from the occupied dp orbitals of Au to the unoccupied 2p* of NO. On Au1/MgO(O2), where the gold atom has a very low electron charge, both mechanisms can in principle contribute to the Au–NO bonding. However, on Au1/MgO(Fs) the donation is unfavoured owing to the high electron charge of gold and, at the same time, a very important backdonation is produced. As in this case the interaction is weaker, we can infer that the donation to the Au atom is more important for a strong bonding of NO. Similar results were obtained by Ding et al. for the interaction of NO with Au01 , Auþ 1 and Au1 [18]. Regarding the adsorption of NO on Au2/MgO, only the normal geometrical configuration for Au2 was considered, in order to allow a comparison between the regular and defective sites of MgO. In both cases, the adsorption is relatively weak (EB(NO) 0.2 eV). As reference, the adsorption on free neutral Au2 is somewhat strong (EB(NO) = 0.48 eV), with an electron transfer of 0.03e from Au2 to NO. The NO molecule bonds through its N atom, forming a more opened Au–N–O angle (about 120–125°) in comparison with the NO/Au1/MgO system (see Fig. 5c and d). The Au–N and N–O interatomic distances are concomitantly longer and shorter than for NO/Au1/MgO, respectively. Looking at the net atomic charges reported in Table 3, we observe that the Au2NO complex has nearly the same charge than the Au2 dimer, on the regular (0.13e versus 0.14e) as well as on the defective site (1.28e versus 1.31e). In both cases, the NO molecule is negatively charged; this charge comes from the Au particle, without taking more electronic charge from the cavity. For NO on Au2/MgO(O2) the two Au atoms become neutral and for NO on Au2/MgO(Fs) the Au atom closest to NO becomes less nega-
b
Fig. 6. (a) and (b): charge density difference, Dq = q(NO/Au1/MgO) q(MgO) q(NO/Au1), for NO adsorbed on Au atom deposited on regular MgO(O2) site and defective MgO(Fs) site, respectively; Dq mapped along a plane containing the Au atoms. Dotted lines correspond to regions with Dq < 0, and solid lines to regions with Dq > 0. Contour levels for mapping are: ±1.0, ±0.5, ±0.25 . . . ±0.00012 a.u.
1675
S.A. Fuente et al. / Surface Science 602 (2008) 1669–1676
1.23
0
1.22
-50 -100
1.21
-150
d (NO)
-200 1.19 -250 1.18
-300
1.17
-350
1.16 1.15 -0.6
Δν (cm-1)
1.2
-400 -450 -0.5
-0.4
-0.3
-0.2
-0.1
0
q (NO) Fig. 7. Continuous line: N–O distance versus NO charge; dashed-line: N–O frequency shift versus NO charge. (s): NO on supported Au1 (white: on Fs; black: on O2); (h): NO on supported Au2 (white: on Fs; black: on O2); (): NO on isolated neutral Au atom; (+): NO on isolated neutral Au dimer.
tive charged. Moreover, for both sites of MgO the spin density is localized mainly on the N atom. The greater strength of NO chemisorption on supported Au1 in comparison supported Au2 can be rationalized in terms of the electronic configuration of these systems. Indeed, NO is a radical with an unpaired electron and, hence, it should bond strongly with other molecule or surface species that has an unpaired electron. This is the situation of free or supported Au1. On the other hand, free or supported Au2 has a closed-shell configuration and, in this case, NO bonds weakly. It is noteworthy that recent DFT calculations showed that NO reacts more strongly with isolated Au1 and Au3 than with isolated Au2 and Au4 [18], concluding that the aggregates with odd-number of electrons prefer to couple with a radical like NO in which the SD in located mainly on the N atom (SD 0.7). For larger aggregates, this oscillatory behavior is not present because the spin spreads out homogenously over all the atoms of the aggregate. Notice that the NO-Au1/MgO(1 0 0) bonding on both surface sites can also be analysed using density spin arguments. Indeed, the greater strength of NO–Au bond on the MgO(O2) site in comparison with the MgO(Fs) site can be attributed, at least in part, to the difference in the spin density values of the Au atom. While on the MgO(O2) site the SD value is 0.76, on the MgO(Fs) site it is much lower, 0.44. The presence of the vacancy delocalizes this spin into the substrate, weakening its coupling with the spin of NO radical. 3.3. Vibrational analysis Looking at Table 2 the N–O stretching frequency, we notice that red-shifts are produced: 336 and 397 cm1 for NO adsorbed on Au1/MgO(O2) and Au1/MgO(Fs) sites, respectively. At our best knowledge, shifts of this magnitude were not observed on supported gold. Solymosi et al. detected signals at 1718 and 1697 cm1 (DmNO = 158 and 179 cm1, respectively) in the NO/Au/MgO system, which they assigned to AuNO species. In these catalysts, the particle size was rather large with a dispersion of about 7% [5]. On the other hand, Debeila et al. [36] found lines shifted by 222 cm1 in the NO/Au/TiO2 system, which they attributed to the NO interaction with the Au–TiO2 interface and/or the adsorption on Au particles in the neighbourhood of a O-vacancy. The calculated NO frequency for free neutral AuNO molecule is 1723 cm1 (Dm = 153 cm1), in very good agreement with experimental findings (1710 cm1 in solid neon) [17]. This frequency changes dramatically for AuNO, giving 1501 cm1 (DmNO = 375 cm1), a value very
similar to those obtained here for the Au monomer supported on Fs. With regard to the Au dimer, the NO frequency shift is less significant: 160 and 254 cm1 for the deposition on MgO(O2) and MgO(Fs), respectively. Unusual frequency shifts as the one calculated here for NO on MgO-supported Au atom were also reported for other systems. In recent studies of atomic Au deposited on regular sites of MgO(1 0 0) performed under UHV conditions, Sterrer et al. found a red-shift of 291 cm1 for the stretching frequency of adsorbed CO, much below any other reported value on supported Au [34]. According to the authors, CO molecule promotes charge flow from the surface oxygen anion of MgO to Au, and from Au to CO. We want to underline that, although the NO red-shifts for both sites are of the same order of magnitude, they have a different physical origin. The red-shift calculated for NO adsorbed on Au1/ MgO(O2) is related with the relatively short Au–NO distance, which at the same time is produced by the coupling between two open-shell species. In this situation, the charge transfer from Au to NO is favoured. On the other hand, for NO on Au1/MgO(Fs) the Au–NO distance is longer, but in this case the large accumulation of negative charge on Au enables a significant electronic transfer to NO. In Fig. 7, we correlate two intuitively expected variations: N–O distance versus NO charge and N–O frequency shift versus NO charge, for the NO interaction with free and supported Au1 and Au2. As the adsorbed NO molecule is charged, the N–O distance increases due to the population of the antibonding 2p* orbital. As a consequence, the frequency value decreases. On the other hand, we can observe that there is not a correlation between NO binding energies and the frequency shifts. Taken individually, supported Au1 against supported Au2, it is true that for the former the adsorption is stronger, and larger the red-shift. Nevertheless, for anyone of the aggregates (Au1 or Au2), the frequency shift is greater on the MgO(Fs) site (the less reactive site) due to the higher Au ? NO net electronic transfer. Thus, for the same type of Au particle, the NO frequency shift can be used to study the modifications undergone by the adsorption site, in particular those due to the net charge of Au particle.
4. Conclusions Firstly, we considered the adsorption of one gold atom and a Au2 dimer on two different sites of the MgO(1 0 0) surface. On the
1676
S.A. Fuente et al. / Surface Science 602 (2008) 1669–1676
Fs site Au atom interacts much more strongly than on the regular anionic site. On the O-vacancy, a charge transfer occurs from the cavity to the valence levels of Au owing to the high polarizability of the charge inside the vacancy and the high electron affinity of gold. The spin density values indicate that, for the surface O2 site, the spin is localized mainly on the Au atom. For the Fs site, the spin density of Au atom is lower because it spreads out around the Au atom and also inside the cavity. The dimer as a unit adsorbs much more strongly on the MgO(Fs) site than on the MgO(O2) site. On the Fs center, the dimer acquires more negative charge than that taken by the monomer. NO adsorbs more strongly on Au1/MgO(O2) (EB(NO) = 1.08 eV) than on Au1/MgO(Fs) (EB(NO) = 0.49 eV). In both cases, the NO molecule bonds with Au through its N atom, adopting a tilted orientation. On Au1/MgO(O2), the NO molecule gains electron charge without taking more charge from the MgO surface. On the contrary, the adsorption of NO on Au1/MgO(Fs) induces a larger charge transfer from the cavity to the adsorbate; as a consequence, the Au1NO complex has a higher negative charge than supported Au atom. On the other hand, the adsorption of NO on Au2/MgO is relatively weak for the two MgO surface sites (EB(NO) 0.2 eV). In both cases, the molecule acquires a negative charge which is lower than that for NO/Au1/MgO. On the Fs site the NO molecule takes its electronic charge almost exclusively from the Au dimer. The greater binding energy of NO on supported Au1 in comparison with supported Au2 can be interpreted in terms of the spin density distribution at the metallic site: NO is a radical with an unpaired electron mainly located on the N atom and, hence, it can easily couple with other species with an unpaired electron such as MgO-supported Au1. Furthermore, the greater strength of NO–Au bond for the MgO(O2) site in comparison with the MgO(Fs) site can be attributed to the difference in the spin density values of the Au atom. While on surface O2 the spin density value is 0.76, on Fs it is much lower, 0.44, owing to a delocalization of the spin into the vacancy, weakening its coupling with the spin of NO radical. On the other hand, NO bonds weakly with supported Au2 because it has a closed-shell configuration. N–O stretching frequencies present dramatic red-shifts of 336 and 397 cm1 (with respect to gas phase) for the adsorption on Au1/MgO(O2) and Au1/MgO(Fs) sites, respectively. The last value is very similar to that obtained for free AuNO molecule (375 cm1). These shifts are much greater than any other reported values for NO stretching frequencies on supported Au. With regard to the supported Au dimer, the NO frequency shift is less significant: 160 and 254 cm1 for Au2 deposited on MgO(O2) and MgO(Fs), respectively.
Ackowledgments The authors acknowledge the financial support of Universidad Nacional del Sur, CONICET and ANPCyT of Argentina. References [1] Z. Yang, S. Chinta, A.A. Mohamed, J.P. Fackler, D.W. Goodman, J. Am. Chem. Soc. 127 (2005) 1604. [2] M. Sterrer, M. Yulikov, E. Fischbach, M. Heyde, H. Rust, G. Pacchioni, T. Risse, H.J. Freund, Angew. Chem. Int. Ed. 45 (2006) 2630. [3] A.V. Matveev, K.M. Neyman, I.V. Yudanov, N. Rosch, Surf. Sci. 426 (1999) 123. [4] M. Haruta, Catal. Today 36 (1997) 153. [5] F. Solymosi, T. Bánsági, T.S. Zakar, Phys. Chem. Chem. Phys. 5 (2003) 4724. [6] K. Blick, T.D. Mitrelias, J.S.J. Hargreaves, G.J. Hutchings, R.W. Joyner, C.J. Kiely, F.E. Wagner, Catal. Lett. 50 (1998) 211. [7] A. Ueda, M. Haruta, Gold Bull. 32 (1) (1999) 3. [8] Z. Yang, R. Wu, Q. Zhang, D.W. Goodman, Phys. Rev. B 65 (2002) 155407. [9] A. Del Vitto, G. Pacchioni, F. Delbecq, P. Sautet, J. Phys. Chem. B 109 (2005) 8040. [10] G. Barcaro, A. Fortunelli, J. Chem. Theory Comput. 1 (2005) 972. [11] P. Frondelius, H. Häkkinen, K. Honkala, New J. Phys. 9 (2007) 339. [12] L.M. Molina, J.A. Alonso, J. Phys. Chem. C 111 (2007) 6668. [13] A. Bogicevic, D.R. Jennison, Surf. Sci. 515 (2002) L481. [14] W.A. Brown, D.A. King, J. Phys. Chem. B 104 (2000) 2578. [15] S.M. McClure, T.S. Kim, J.D. Stiehl, P.L. Tanaka, C.B. Mullins, J. Phys. Chem. B 108 (2004) 17952. [16] C.P. Vinod, J.W. Niemantsverdriet Hans, B.E. Nieuwenhuys, Appl. Catal. A 291 (2005) 93. [17] A. Citra, X. Wang, L. Andrews, J. Phys. Chem. A 106 (2002) 3287. [18] X. Ding, Z. Li, J. Yang, J.G. Hou, Q. Zhu, J. Chem. Phys. 121 (2004) 2558. [19] A.D. Becke, J. Chem. Phys. 98 (1993) 5648. [20] G. Pacchioni, P. Pescarmona, Surf. Sci. 412/413 (1998) 657. [21] C. Di Valentin, R.M. Ferullo, R. Binda, G. Pacchioni, Surf. Sci. 600 (2006) 1147. [22] J. Carrasco, N. Lopez, F. Illas, H.-J. Freund, J. Chem. Phys. 125 (2006) 074711. [23] P.J. Hay, W.R. Wadt, J. Chem. Phys. 82 (1985) 270. [24] P.J. Hay, W.R. Wadt, J. Chem. Phys. 82 (1985) 299. [25] D. Andrae, U. Haussermann, M. Dolg, H. Stoll, H. Preuss, Theor. Chim. Acta 77 (1990) 123. [26] I. Yudanov, G. Pacchioni, K. Neyman, N. Rösch, J. Phys. Chem. B 101 (1997) 2786. [27] N.R. Kestner, J.E. Combariza, in: K.B. Lipkowitz, D.B. Boyd (Eds.), Reviews in Computational Chemistry, vol. 13, Wiley-VCH, John Wiley and Sons Inc., New York, 1999 (Chapter 2, p. 99). [28] A.E. Reed, L.A. Curtiss, F. Weinhold, Chem. Rev. 88 (1988) 899. [29] A.M. Joshi, M.H. Tucker, W.N. Delgass, K.T. Thomson, J. Chem. Phys. 125 (2006) 194707. [30] M.J. Frisch et al., Gaussian 03, Revision C.02, Gaussian, Inc., Wallingford, CT, 2004. [31] C. Inntam, L.V. Moskaleva, K.M. Neyman, V.A. Nasluzov, N. Rösch, Appl. Phys. A 82 (2006) 181. [32] M. Yulikov, M. Sterrer, M. Heyde, H.-P. Rust, T. Risse, H.-J. Freund, G. Pacchioni, A. Scagnelli, Phys. Rev. Lett. 96 (2006) 146804. [33] C. Di Valentin, A. Scagnelli, G. Pacchioni, T. Risse, H.-J. Freund, Surf. Sci. 600 (2006) 2434. [34] M. Sterrer, M. Yulikov, T. Risse, H.J. Freund, J. Carrasco, F. Illas, C. Di Valentin, L. Giordano, G. Pacchioni, T. Risse, H.J. Freund, Angew. Chem. Int. Ed. 45 (2006) 2633. [35] H. Grönbeck, P. Broqvist, J. Phys. Chem. B 107 (2003) 12239. [36] M.A. Debeila, N.J. Coville, M.S. Scurrell, G.R. Hearne, J. Mol. Catal. A 219 (2004) 131.
Contents lists available at ScienceDirect
Surface Science journal homepage: www.elsevier.com/locate/susc
Adsorption of NO on Au atoms and dimers supported on MgO(1 0 0): DFT studies Silvia A. Fuente a, Patricia G. Belelli a, Ricardo M. Ferullo a,b,*, Norberto J. Castellani a a b
Grupo de Materiales y Sistemas Catalíticos, Departamento de Física, Universidad Nacional del Sur, Av. Alem 1253, 8000 Bahía Blanca, Argentina Departamento de Química, Universidad Nacional del Sur, Av. Alem 1253, 8000 Bahía Blanca, Argentina
a r t i c l e
i n f o
Article history: Received 26 December 2007 Accepted for publication 29 February 2008 Available online 10 March 2008
Keywords: Au/MgO Gold clusters NO adsorption Nitric oxide DFT Cluster model
a b s t r a c t The adsorption of NO on single gold atoms and Au2 dimers deposited on regular O2 sites and neutral oxygen vacancies (Fs sites) of the MgO(1 0 0) surface have been studied by means of DFT calculations. For Au1/MgO the adsorption of NO is stronger when the Au atom is supported on an anionic site than when it is on a Fs site, with adsorption binding energies of 1.1 and 0.5 eV, respectively. In the first case the spin density is mainly concentrated on the metal atom and protruding from the surface. In such a way, an active site against radicals such as NO is generated. On the Fs site, the presence of the vacancy delocalizes the spin into the substrate, weakening its coupling with NO. For Au2/MgO, as this system has a closed-shell configuration, the NO molecules bonds weakly with Au2. Regarding the N–O stretching frequencies, a very strong shift of 340–400 cm1 to lower frequencies is observed for Au1/MgO in comparison with free NO. Ó 2008 Elsevier B.V. All rights reserved.
1. Introduction In heterogeneous catalysis the reactive properties of supported metal particles depend on the cluster size and on the specific site of the support where the particle is anchored. When an oxide is used as support, the metal aggregates interact preferentially with surface oxygen vacancies [1,2]. In particular, on MgO these vacancies comprise several different types of surface defects such as Fs, Fs+ and Fs2+, depending on removal of a neutral O atom, an O anion and an O2 anion, respectively. The interaction of transition metal atoms with both regular and defective MgO was investigated using quantum chemical approaches. It was observed that the adhesion is stronger on Fs sites than on regular anionic sites [3]. The activity and selectivity of supported Au aggregates was intensively studied during the last years. They show an extraordinary activity to catalyze different reactions at low operating temperatures, such as the oxidation of CO, the selective oxidation of propene and the reduction of NOx [4]. In experiments, TiO2 is used as the typical support, but other oxides like MgO, SiO2 and Al2O3 were also considered [5–7]. Recently, experiments were performed in order to clarify the importance of Fs sites of MgO in the adhesion of gold. It was found a correlation between the catalytic activity of the supported Au
* Corresponding author. Address: Grupo de Materiales y Sistemas Catalíticos, Departamento de Física, Universidad Nacional del Sur, Av. Alem 1253, 8000 Bahía Blanca, Argentina. Tel.: +54 291 4595141; fax: +54 291 4595142. E-mail address: [email protected] (R.M. Ferullo). 0039-6028/$ - see front matter Ó 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.susc.2008.02.037
particles and the concentration of Fs centers on MgO. Thus, these surface defects play a direct role in the activation of supported Au catalysts [1]. In another recent work, the electron paramagnetic resonance (EPR) and scanning tunnelling microscopy (STM) techniques were combined to verify the existence of surface defects as well as the nucleation of Au onto these defects [2]. Moreover, by using infrared (IR) spectroscopy of adsorbed CO, two bands were observed: one at 2120 cm1 and a broad one around 2070 cm1. While the former was assigned to CO adsorbed on small neutral Au clusters deposited on regular anionic sites, the broad band was assigned to CO on negatively charged Au clusters. These charged particles are predicted to be produced by the adsorption of Au on defect sites of MgO. The larger red-shift is attributed to a greater p-backdonation from the negatively charged Au in comparison with that of neutral Au. The deposition of Au atoms and clusters on regular five-coordinated O2 sites and on Fs centers of the MgO(1 0 0) surface was recently studied from periodic DFT calculations [8–13]. It was observed that the presence of the vacancy enhances the metal– oxide interaction. Thus, these centers are essential to trap the diffusing Au atoms or clusters [9]. In the past, NO was used as a probe molecule to study the metal surface reactivity [14]. Besides, the bonding of NO to transition metal surfaces is of interest in relation to the catalytic reduction for pollution control. At ultrahigh vacuum (UHV) conditions, NO adsorbs weakly on Au(1 1 1) [15] but it decomposes on Au(3 1 0) giving N2O and atomic O on this stepped surface [16]. Gold supported on several metal oxides shows good catalytic performance for the
1670
S.A. Fuente et al. / Surface Science 602 (2008) 1669–1676
reduction of nitric oxide by hydrocarbons [7]. In these cases, the addition of CO and H2 to the reactant gas improves the conversion of NO to N2. The NO + CO reaction was also studied for Au supported on MgO, TiO2, SiO2 and Al2O3 [5]. These systems catalyze the reaction above 573 K yielding surface NCO. The interaction of NO with isolated Au atoms was object of recent studies. Laser-ablated Au atoms interact with NO yielding neutral nitrosyl complexes AuNO and Au(NO)2 [17]. For AuNO, the N–O and Au–N stretching modes are observed at about 1700 and 520 cm1, respectively. On the other hand, the reaction of NO with free neutral Aux clusters (x = 1–6) was studied from DFT [18]. The NO + Aux reaction is exothermic, releasing 0.2–0.8 eV. The bond between NO and Au is accompanied by an electron transfer from the metal cluster to the NO molecule. The first objective of this work is to analyze the deposition of one Au atom and the Au2 dimer on anionic regular sites and on Fs centers of the MgO(1 0 0) surface by employing an embedded cluster approach. The results are compared with recent periodic calculations [8–13]. While the Au–MgO interface was thoroughly studied, the adsorption properties of supported Au particles received relatively less attention. Specifically, in this work we are interested to study the adsorption of NO on supported gold which constitutes a valuable reference for other researchers. The corresponding results are compared with the reaction of NO with free Au particles in order to study the effect of the support on the adsorption capability of supported Au. 2. Computational details Density functional theory (DFT) molecular orbital calculations were carried out using the gradient corrected Becke’s three parameters hybrid exchange functional in combination with the correlation functional of Lee, Yang and Parr (B3LYP) [19]. This method was widely used in the past to study adsorption processes yielding reliable results both on oxides and metal clusters. The MgO(1 0 0) surface was represented by clusters of 26 atoms, Mg13O13, consisting of two layers (first layer: Mg4O9; second layer: Mg9O4). To take into account the Madelung field due to the rest of the extended surface, the cluster was embedded in an array of ±2 point charges. Moreover, the positive point charges at the interface were replaced by effective core potentials (ECP) corresponding to Mg2+ to account for the finite size of the cations and to avoid spurious charge polarization. The Mg13O13 cluster plus the 16 Mg– ECP’s is represented in Fig. 1. This methodology was widely used by the group of Pacchioni and other authors [20], giving results which are in good agreement with those obtained by periodic calculations. For example, if the embedded cluster model and the slab
approach are compared, the energy for creating an oxygen vacancy is only 1% different [21,22]. The 6-31+G(d) basis set was employed for the atoms of the NO molecule and the O atoms in the first layer of Mg13O13 cluster, except the terminal ones. The atomic orbitals of the four Mg atoms in the first layer were described by the 6-31G(d) basis set, while those of the terminal O atoms and those of Mg and O atoms at the second layer, by the 6-31G basis set. For Au, the LANL2DZ basis set was used, which describes the 60 core electrons with a pseudopotential and the 19 valence 5s25p65d106s1 electrons with a [5s,6p,4d/ 3s,3p,2d] basis set [23,24]. Previously, the suitability of this basis set for the Au–NO interaction was examined by considering the isolated Au1NO molecule. For that purpose, the results of calculation using the LANL2DZ basis set was compared with that obtained with a larger one such as the SDD (Stuttgart–Dresden) [25]. The last describes the 19 valence electrons with a [8s,7p,6d/6s,5p,3d] basis set. The resultant distances, angles and frequencies values are very similar using both descriptions. During the geometrical optimization procedure, the coordinates of atoms belonging to the adsorbed NO molecule and the metal particle, and those of the five central atoms in the first layer of Mg13O13 cluster (the central O and its four Mg atoms directly bonded to it) were fully optimized without imposing geometric constrains. The geometries studied in this work were selected taking into account previous theoretical results [9–13]. With respect to the gold monomer deposited on the regular surface, only the adsorption on the anionic site was studied because it is well known that this is the preferred adsorption site for most transition metal atoms [26]. For the Au dimers on the perfect surface, the on-top geometry on one anionic site was considered following the results obtained in Ref. [12]. Regarding the defective surface several starting orientations for Au2 were taken into account. A neutral O vacancy was represented by Mg13O12 cluster, with the vacancy localized in the center of the first layer. In this case, the geometrical optimization included the four central Mg atoms of first layer. We define the binding energy of the Aux particle on MgO as EB(Aux) = [E(Aux/MgO, site) E(Aux) E(MgO, site)], with x = 1, 2, and site = O2 or Fs. Besides, we define the binding energy of NO molecule on Aux/MgO as EB(NO) = [E(NO/Aux/MgO, site) E(Aux/MgO, site) – E(NO)]. In both cases, positive values correspond to exothermic processes. The reactions energies were corrected by considering the basis set superposition error (BSSE), calculated according to the counter-poise correction [27]. The frequency values were scaled according to the factor of 0.9465, calculated as the ratio between the empirical and the calculated frequency values of free NO (1876/1982).
Fig. 1. Mg13O13 cluster plus the 16 Mg–ECP used for modeling the MgO(1 0 0) surface. Dark grey spheres: O. Light grey spheres: Mg. Small spheres represent the Mg–ECP’s. Point charges are not shown.
S.A. Fuente et al. / Surface Science 602 (2008) 1669–1676
The spin density (SD) was expressed in terms of the Mülliken population analysis. On the other hand, the atomic net charges were calculated following the natural bond orbital (NBO) scheme [28], which gives realistic values for the charge partitioning. The NBO approach was recently used to analyze the interaction of CO and NO with nanoparticles containing coinage metals [17,18,29]. In particular, it was very useful for the interpretation of IR frequency shifts of the AuNO complex [17]. The multiplicity was fixed to M = 1 in Au2/MgO and NO/Au1/ MgO, and to M = 2 in Au1/MgO and NO/Au2/MgO. For all the systems the total charge was zero. All the calculations have been performed using the Gaussian-03 program package [30]. 3. Results and discussion 3.1. Au1 and Au2 particles deposited on MgO(1 0 0) It was well established that small metal particles adsorb preferentially on sites where negative charge accumulates [8– 13,26,31,32]; more specifically, the O2 ionic sites for regular MgO, and the Fs and Fþ S centers for defective MgO. Recently, the deposition of Au1 and Au2 on O2 and Fs sites of MgO(1 0 0) was studied by means of periodic calculations [8–13] and the embedded cluster approach [31]. As it will be shown later, the interaction of NO with supported Au depends strongly of the metal–oxide interaction and it is essential to dispose of an adequate substrate model for the subsequent NO adsorption. Hence, as a first step in our approach, the Au1/MgO(1 0 0) and Au2/MgO(1 0 0) systems were studied using the above detailed cluster models and the results were compared with those obtained with the periodic method. In Table 1, the calculated values of binding energies, atomic charges and spin densities for the deposition of the Aux particle (x = 1, 2) on a regular oxygen site and on a Fs center are reported. For simplicity, these surface sites will be denoted as ‘‘MgO(O2)” and ‘‘MgO(Fs)”, respectively. In Fig. 2 a view of the corresponding optimized geometries is presented. From the values of Table 1 we notice that one Au atom interacts much stronger on the neutral O-vacancy (EB(Au1) = 2.93 eV) than on the surface O2 site (EB(Au1) = 0.73 eV). These values are only 0.2–0.3 eV lower than those obtained using slab calculations and the PW91 exchange-correlation functional (3.17 and 0.89 eV from Ref. [9], and 3.04 and 0.87 eV from Ref. [10], respectively). The difference could be explained by the general tendency of the hybrid
Table 1 Main energetic and electronic population parameters for Aux particles deposited on MgO. Energies are expressed in eV and charges in electron units O2 site EB (Au1)a q(Au) q(O2) SD(Au) SD(O2) EB(Au2)a Edimb q(Aut)c q(Aus)c q(O2) a
Au1/MgO 0.73 0.13 1.74 0.76 0.26 Au2/MgOd 1.34 2.56 0.28 +0.14 1.76
Fs site 2.93 0.88 – 0.44 – 3.86 2.88 0.67 0.61 –
EB(Aux) = [E(Aux/MgO, site) E(MgO, site) E(Aux)]; x = 1, 2; site = O2 or Fs. Edim = [E(Au2/MgO, site) E(Au1/MgO, site) E(Au1)]; site = O2 or Fs. c Aut: terminal Au atom; Aus: Au atom closer to the surface. d The values for Au2 on the Fs site correspond to the configuration normal to the surface. b
1671
B3LYP functional to predict lower binding energies than the PW91-GGA functional. According to the NBO analysis, when the Au atom interacts with anionic sites a charge transfer of 0.13e occurs from the surface to the metal atom (Table 1). Results from EPR spectroscopy of Au deposited on a smooth MgO(1 0 0) film suggest that single Au atoms are basically neutral [32]. On the other hand, this charge transfer is much greater when the metal atom interacts with the O-vacancy (0.88e, Table 1), a phenomenon that can be explained by the high polarizability of the charge inside the vacancy and the high electron affinity of gold [9]. To analyze the charge distribution after deposition, we have calculated the charge density difference Dq = q(Au1/MgO) q(MgO) q(Au1), by subtracting from the total charge density of the Au1/MgO system the densities of the component fragments, and mapped it on the plane that contains the metal atom (see Fig. 3). For the MgO(O2) site, we can observe a depletion of the charge density on the O anion and a charge accumulation around the Au atom (Fig. 3a). As it was discussed by Del Vitto et al. [9], the lack of accumulation of electron density in the region between the Au atom and the oxide surface indicates that the intraunit polarization of Au atom makes an important contribution to the Au–MgO bonding. For the MgO(Fs) site, a depletion of charge inside the cavity was observed, together with an important charge accumulation all around the Au atom, including the interface region (Fig. 3b). The spin density values indicate that, for the MgO(O2) site, the spin is localized mainly on the Au atom (SD(Au) = 0.76). On the other hand, for the MgO(Fs) site, the spin density of the Au atom is lower (SD(Au) = 0.44). This difference in the spin localization is clear by comparing the corresponding spin density maps shown in Fig. 4. While at the regular surface the spin is distributed around the Au atom and around the surface O anion (Fig. 4a), at the defective surface it spreads out around the Au atom and inside the cavity (Fig. 4b). The higher spin density for the Au atom linked to the MgO(O2) site is in agreement with calculated EPR properties [33] and it is very important for the understanding of the adsorptive capability of the supported Au atom towards a radical like NO (see below). Regarding the Au dimer deposited on the MgO(O2) site, it adopts an orientation that is normal to the surface (Fig. 2b), with a binding energy value which nearly doubles the value for the monomer (EB(Au2) = 1.34 eV). On the Fs center, the calculations predict two different configurations: normal to the surface and tilted towards a Mg cation, forming an angle of 53° with respect to the normal (Fig. 2d and e, respectively). These structures are practically isoenergetic: the perpendicular geometry has a binding energy (EB(Au2) = 3.86 eV) which is only 0.06 eV greater than that for the tilted geometry. These energies are nearly 1 eV greater than that for the monomer. The corresponding EB(Au2) values obtained from periodic calculations by Del Vitto et al. [9] are 1.49 and 3.89 eV, for the adsorption on the regular surface and on Fs, respectively (in Ref. [9] the tilted geometry is somewhat more stable than the normal one). Concerning the Au–Au distance, it is practically unperturbed on the MgO(O2) site and stretched by 0.17 Å (0.25 Å) on the MgO(Fs) site for perpendicular (tilted) orientation (the calculated interatomic distance is 2.57 Å for free Au2). Notice that the dimer as a unit adsorbs much more strongly on the MgO(Fs) site than on the MgO(O2) site. Besides the EB(Au2) binding energy we can define the dimer formation energy, Edim, as the energy associated with the dimer formation when a free metal atom interacts with a pre-deposited atom, Edim = [E(Au2/MgO, site) E(Au1/MgO, site) E(Au1)]; site = O2 or Fs. While the binding energy of Au2, EB(Au2), is noticeably affected by the support, the dimer formation energy is weakly affected; for both sites the values of Edim are quite similar (2.56 eV on surface O2 and 2.88 eV on Fs).
1672
S.A. Fuente et al. / Surface Science 602 (2008) 1669–1676
Fig. 2. (a) and (b): schematic representation of Au1 and Au2 particles, respectively, deposited on regular MgO(O2) site. (c) and (d): schematic representation of Au1 and Au2 particles, respectively, deposited on defective MgO(Fs) site. Dark grey spheres, oxygen atoms. Light grey spheres, Mg atoms (including the Mg–ECP’s). Large grey spheres, Au atoms. In 2a, the distance between Au and its nearest Mg is 3.241 Å. In 2b, the distance between the Au atom closer to the surface and its nearest Mg is 3.174 Å.
a
b
c
d
Fig. 3. (a) and (b): charge density difference, Dq = q(Au1/MgO) q(MgO) q(Au1), for Au atom deposited on regular MgO(O2) site and defective MgO(Fs) site, respectively; (c) and (d): charge density difference, Dq = q(Au2/MgO) q(MgO) q(Au2), for Au2 aggregate deposited on regular MgO(O2) site and defective MgO(Fs) site, respectively. Dq is mapped along a plane containing the Au atoms. Dotted lines correspond to regions with Dq < 0, and solid lines to regions with Dq > 0. Contour levels for mapping are: ±1.0, ±0.5, ±0.25 . . . ±0.00012 a.u.
The charge of Au2 supported on the MgO(O2) site is 0.14e, practically the same than that of the supported monomer. Within the dimer, this charge is inhomogeneously distributed with an accumulation towards the terminal Au. On the MgO(Fs) site, the dimer acquires a much more significant negative charge, 1.28e,
which is even higher than that taken by the monomer (nearly 0.40e higher). In this case the charge is distributed quasi-homogeneously in both Au atoms. The spatial charge distribution can be analyzed looking at the density difference maps of Fig. 3c and d, evaluated with respect to the separated fragments, MgO and Au2.
1673
S.A. Fuente et al. / Surface Science 602 (2008) 1669–1676
a
Table 3 Main molecular and electronic population parameters for NO adsorbed on a Au2 dimer deposited on MgO. Energies are expressed in eV, charges in electron units and frequencies in cm1
b
a
EB(NO) SD(O) SD(N) SD(Aut)b SD(Aus)b SD(O2) q(O) q(N) q(Aut)b q(Aus)b q(O2) mNO (DmNO) a b
NO/Au2/MgO(O2)
NO/Au2/MgO(Fs)
0.23 0.31 0.63 0.00 0.05 0.01 0.20 +0.07 0.00 0.00 1.77 1716 (160)
0.20 0.32 0.75 0.00 0.00 – 0.29 +0.03 0.40 0.65 – 1622 (254)
EB(NO) = [E(NO/Au2/MgO, site) E(Au2/MgO, site) E(NO)]; site = O2 or Fs. Aut: terminal Au atom; Aus: Au atom closer to the surface.
Fig. 4. (a) and (b): spin density for Au atom deposited on regular MgO(O2) site and defective MgO(Fs) site, respectively. Contour levels for mapping are: 0.17, 0.085, 0.0425 . . . ±0.00017 a.u.
To allow a comparison for Au2 on both sites, only the normal geometrical configuration was considered. On the anionic site, the charge is concentrated around the terminal Au atom and towards the vacuum. On the O-vacancy, the charge distributes also in this way and, in addition, an important accumulation of charge is observed in the region between the bottom Au atom and the cavity. 3.2. Adsorption of NO on supported Au1 and Au2 In Tables 2 and 3, the calculated values of adsorption energies, atomic charges and spin densities for the adsorption of NO on Au1/MgO and Au2/MgO are reported, respectively. In Fig. 5, the corresponding optimized geometries are presented. Looking at Table 2, we observe that NO adsorbs much more strongly when the Au monomer is deposited on the MgO(O2) site than on the MgO(Fs) defect (EB(NO) = 1.08 eV and 0.49 eV, respectively). The reason of this behavior will be discussed later. Interestingly, when NO reacts with an unsupported Au atom, the binding energy is 0.69 eV; therefore, the regular anionic site acts favouring the Au–NO bond and the Fs site weakens that interaction. The NO molecule bonds with Au through its N atom and adopts a tilted orientation with a Au–N–O angle having 109 and 117° on MgO(O2) and MgO(Fs), respectively. The N–O distance elongates from its value of 1.158 Å for the free NO molecule to 1.204 and 1.220 Å for
Table 2 Main molecular and electronic population parameters for NO adsorbed on one Au atom deposited on MgO. Energies are expressed in eV, charges in electron units and frequencies in cm1
a
EB(NO) EB(Au1NO)b q(O) q(N) q(Au) q(O2) mNO (DmNO) a b
Fs .
NO/Au1/MgO(O2)
NO/Au1/MgO(Fs)
1.08 1.00 0.30 0.07 +0.26 1.76 1540 (336)
0.49 2.62 0.35 0.22 0.50 – 1479 (397)
EB(NO) = [E(NO/Au1/MgO, site) E(Au1/MgO, site) E(NO)]; site = O2 or Fs. EB(Au1NO) = [E(NO/Au1/MgO, site) E(MgO, site) E(Au1NO)]; site = O2 or
Fig. 5. (a) and (b): schematic representation of NO adsorbed on Au1 and Au2 particles, respectively, deposited on regular MgO(O2) site. (c) and (d): schematic representation of NO adsorbed on Au1 and Au2 particles, respectively, deposited on defective MgO(Fs) site. Dark grey spheres, oxygen atoms. Light grey spheres, Mg atoms. Large grey spheres, Au atoms. Black spheres, nitrogen atoms.
Au1/MgO(O2) and Au1/MgO(Fs), respectively (for gas-phase AuNO this distance is 1.161 Å). On the other side, the Au–N distance changes from 2.105 for AuNO to 2.058 and 2.340 Å when the metal atom is supported on MgO(O2) and MgO(Fs), respectively (see Fig. 5a and b). Notice that the more significant elongation of Au–N distance for MgO(Fs), in comparison with MgO(O2), is in
1674
S.A. Fuente et al. / Surface Science 602 (2008) 1669–1676
agreement with a relatively less significant EB(NO) value (0.59 eV lower). Very recently, several authors were interested in studying the CO-induced modification of the metal–MgO interaction [34,35] in the form of atoms and layers. For instance, it was found that the bonding between Pt or Pd atoms and the oxide is enhanced by CO, but for Au this effect is negligible [35]. To allow a similar analysis on NO, the adsorption of the AuNO complex on MgO(1 0 0) was considered here. The corresponding binding energy can be defined in a similar way as that for Aux on MgO. The calculated EB(Au1NO) values were 1.00 and 2.62 eV for the adsorption on MgO(O2) and MgO(Fs), respectively. If we compare the binding energies for AuNO (see Table 2) and that for Au (see Table 1) we observe that the Au–oxide interaction is enhanced in the former case and that it is weakened in the latter one. Concerning the charge distribution, the values of atomic charges indicate that the charge of the AuNO complex on the MgO(O2) site is approximately the same than that of supported Au atom, so that the charge transfer from the oxide to the adsorbate (Au atom or surface AuNO complex) is very similar. However, the NO molecule withdraws electronic charge from the metal atom which becomes positive, +0.26e. On the contrary, the AuNO complex on Fs takes more electronic charge from the cavity than the supported Au atom, the former resulting 0.19e more negative. This excess of negative charge accumulates on the NO, which reaches a net atomic charge of 0.57e. The charge gained by NO populates the 2p* (antibonding) orbital, weakening the N–O bond. The larger stretching of the NO interatomic distance on the Fs site can be explained by this higher electron transfer. In Fig. 6, the corresponding maps of charge density differences are shown, calculated with respect to MgO and AuNO fragments. As a consequence of AuNO adsorption on both MgO(O2) and MgO(Fs) sites, it is observed an accumulation of charge on NO, in agreement with the above commented net atomic charges (at free AuNO molecule, the Au ? NO electronic transfer is very low, 0.06e). On the MgO (O2) site, it can be observed a region of depletion of charge around the Au atom and the presence of a little accu-
a
mulation of charge between Au and O2, an indication of an enhancement of the covalent contribution. On the MgO(Fs) site, the region surrounding the Au atom loses electronic charge while it is retained in the interface region towards the cavity. The NO interaction with a gold aggregate can be viewed within the donation/backdonation scheme. While the electron donation occurs from the single occupied 2p* antibonding orbital of NO to the dr + sp hybridized orbitals of Au, the corresponding backdonation is produced from the occupied dp orbitals of Au to the unoccupied 2p* of NO. On Au1/MgO(O2), where the gold atom has a very low electron charge, both mechanisms can in principle contribute to the Au–NO bonding. However, on Au1/MgO(Fs) the donation is unfavoured owing to the high electron charge of gold and, at the same time, a very important backdonation is produced. As in this case the interaction is weaker, we can infer that the donation to the Au atom is more important for a strong bonding of NO. Similar results were obtained by Ding et al. for the interaction of NO with Au01 , Auþ 1 and Au1 [18]. Regarding the adsorption of NO on Au2/MgO, only the normal geometrical configuration for Au2 was considered, in order to allow a comparison between the regular and defective sites of MgO. In both cases, the adsorption is relatively weak (EB(NO) 0.2 eV). As reference, the adsorption on free neutral Au2 is somewhat strong (EB(NO) = 0.48 eV), with an electron transfer of 0.03e from Au2 to NO. The NO molecule bonds through its N atom, forming a more opened Au–N–O angle (about 120–125°) in comparison with the NO/Au1/MgO system (see Fig. 5c and d). The Au–N and N–O interatomic distances are concomitantly longer and shorter than for NO/Au1/MgO, respectively. Looking at the net atomic charges reported in Table 3, we observe that the Au2NO complex has nearly the same charge than the Au2 dimer, on the regular (0.13e versus 0.14e) as well as on the defective site (1.28e versus 1.31e). In both cases, the NO molecule is negatively charged; this charge comes from the Au particle, without taking more electronic charge from the cavity. For NO on Au2/MgO(O2) the two Au atoms become neutral and for NO on Au2/MgO(Fs) the Au atom closest to NO becomes less nega-
b
Fig. 6. (a) and (b): charge density difference, Dq = q(NO/Au1/MgO) q(MgO) q(NO/Au1), for NO adsorbed on Au atom deposited on regular MgO(O2) site and defective MgO(Fs) site, respectively; Dq mapped along a plane containing the Au atoms. Dotted lines correspond to regions with Dq < 0, and solid lines to regions with Dq > 0. Contour levels for mapping are: ±1.0, ±0.5, ±0.25 . . . ±0.00012 a.u.
1675
S.A. Fuente et al. / Surface Science 602 (2008) 1669–1676
1.23
0
1.22
-50 -100
1.21
-150
d (NO)
-200 1.19 -250 1.18
-300
1.17
-350
1.16 1.15 -0.6
Δν (cm-1)
1.2
-400 -450 -0.5
-0.4
-0.3
-0.2
-0.1
0
q (NO) Fig. 7. Continuous line: N–O distance versus NO charge; dashed-line: N–O frequency shift versus NO charge. (s): NO on supported Au1 (white: on Fs; black: on O2); (h): NO on supported Au2 (white: on Fs; black: on O2); (): NO on isolated neutral Au atom; (+): NO on isolated neutral Au dimer.
tive charged. Moreover, for both sites of MgO the spin density is localized mainly on the N atom. The greater strength of NO chemisorption on supported Au1 in comparison supported Au2 can be rationalized in terms of the electronic configuration of these systems. Indeed, NO is a radical with an unpaired electron and, hence, it should bond strongly with other molecule or surface species that has an unpaired electron. This is the situation of free or supported Au1. On the other hand, free or supported Au2 has a closed-shell configuration and, in this case, NO bonds weakly. It is noteworthy that recent DFT calculations showed that NO reacts more strongly with isolated Au1 and Au3 than with isolated Au2 and Au4 [18], concluding that the aggregates with odd-number of electrons prefer to couple with a radical like NO in which the SD in located mainly on the N atom (SD 0.7). For larger aggregates, this oscillatory behavior is not present because the spin spreads out homogenously over all the atoms of the aggregate. Notice that the NO-Au1/MgO(1 0 0) bonding on both surface sites can also be analysed using density spin arguments. Indeed, the greater strength of NO–Au bond on the MgO(O2) site in comparison with the MgO(Fs) site can be attributed, at least in part, to the difference in the spin density values of the Au atom. While on the MgO(O2) site the SD value is 0.76, on the MgO(Fs) site it is much lower, 0.44. The presence of the vacancy delocalizes this spin into the substrate, weakening its coupling with the spin of NO radical. 3.3. Vibrational analysis Looking at Table 2 the N–O stretching frequency, we notice that red-shifts are produced: 336 and 397 cm1 for NO adsorbed on Au1/MgO(O2) and Au1/MgO(Fs) sites, respectively. At our best knowledge, shifts of this magnitude were not observed on supported gold. Solymosi et al. detected signals at 1718 and 1697 cm1 (DmNO = 158 and 179 cm1, respectively) in the NO/Au/MgO system, which they assigned to AuNO species. In these catalysts, the particle size was rather large with a dispersion of about 7% [5]. On the other hand, Debeila et al. [36] found lines shifted by 222 cm1 in the NO/Au/TiO2 system, which they attributed to the NO interaction with the Au–TiO2 interface and/or the adsorption on Au particles in the neighbourhood of a O-vacancy. The calculated NO frequency for free neutral AuNO molecule is 1723 cm1 (Dm = 153 cm1), in very good agreement with experimental findings (1710 cm1 in solid neon) [17]. This frequency changes dramatically for AuNO, giving 1501 cm1 (DmNO = 375 cm1), a value very
similar to those obtained here for the Au monomer supported on Fs. With regard to the Au dimer, the NO frequency shift is less significant: 160 and 254 cm1 for the deposition on MgO(O2) and MgO(Fs), respectively. Unusual frequency shifts as the one calculated here for NO on MgO-supported Au atom were also reported for other systems. In recent studies of atomic Au deposited on regular sites of MgO(1 0 0) performed under UHV conditions, Sterrer et al. found a red-shift of 291 cm1 for the stretching frequency of adsorbed CO, much below any other reported value on supported Au [34]. According to the authors, CO molecule promotes charge flow from the surface oxygen anion of MgO to Au, and from Au to CO. We want to underline that, although the NO red-shifts for both sites are of the same order of magnitude, they have a different physical origin. The red-shift calculated for NO adsorbed on Au1/ MgO(O2) is related with the relatively short Au–NO distance, which at the same time is produced by the coupling between two open-shell species. In this situation, the charge transfer from Au to NO is favoured. On the other hand, for NO on Au1/MgO(Fs) the Au–NO distance is longer, but in this case the large accumulation of negative charge on Au enables a significant electronic transfer to NO. In Fig. 7, we correlate two intuitively expected variations: N–O distance versus NO charge and N–O frequency shift versus NO charge, for the NO interaction with free and supported Au1 and Au2. As the adsorbed NO molecule is charged, the N–O distance increases due to the population of the antibonding 2p* orbital. As a consequence, the frequency value decreases. On the other hand, we can observe that there is not a correlation between NO binding energies and the frequency shifts. Taken individually, supported Au1 against supported Au2, it is true that for the former the adsorption is stronger, and larger the red-shift. Nevertheless, for anyone of the aggregates (Au1 or Au2), the frequency shift is greater on the MgO(Fs) site (the less reactive site) due to the higher Au ? NO net electronic transfer. Thus, for the same type of Au particle, the NO frequency shift can be used to study the modifications undergone by the adsorption site, in particular those due to the net charge of Au particle.
4. Conclusions Firstly, we considered the adsorption of one gold atom and a Au2 dimer on two different sites of the MgO(1 0 0) surface. On the
1676
S.A. Fuente et al. / Surface Science 602 (2008) 1669–1676
Fs site Au atom interacts much more strongly than on the regular anionic site. On the O-vacancy, a charge transfer occurs from the cavity to the valence levels of Au owing to the high polarizability of the charge inside the vacancy and the high electron affinity of gold. The spin density values indicate that, for the surface O2 site, the spin is localized mainly on the Au atom. For the Fs site, the spin density of Au atom is lower because it spreads out around the Au atom and also inside the cavity. The dimer as a unit adsorbs much more strongly on the MgO(Fs) site than on the MgO(O2) site. On the Fs center, the dimer acquires more negative charge than that taken by the monomer. NO adsorbs more strongly on Au1/MgO(O2) (EB(NO) = 1.08 eV) than on Au1/MgO(Fs) (EB(NO) = 0.49 eV). In both cases, the NO molecule bonds with Au through its N atom, adopting a tilted orientation. On Au1/MgO(O2), the NO molecule gains electron charge without taking more charge from the MgO surface. On the contrary, the adsorption of NO on Au1/MgO(Fs) induces a larger charge transfer from the cavity to the adsorbate; as a consequence, the Au1NO complex has a higher negative charge than supported Au atom. On the other hand, the adsorption of NO on Au2/MgO is relatively weak for the two MgO surface sites (EB(NO) 0.2 eV). In both cases, the molecule acquires a negative charge which is lower than that for NO/Au1/MgO. On the Fs site the NO molecule takes its electronic charge almost exclusively from the Au dimer. The greater binding energy of NO on supported Au1 in comparison with supported Au2 can be interpreted in terms of the spin density distribution at the metallic site: NO is a radical with an unpaired electron mainly located on the N atom and, hence, it can easily couple with other species with an unpaired electron such as MgO-supported Au1. Furthermore, the greater strength of NO–Au bond for the MgO(O2) site in comparison with the MgO(Fs) site can be attributed to the difference in the spin density values of the Au atom. While on surface O2 the spin density value is 0.76, on Fs it is much lower, 0.44, owing to a delocalization of the spin into the vacancy, weakening its coupling with the spin of NO radical. On the other hand, NO bonds weakly with supported Au2 because it has a closed-shell configuration. N–O stretching frequencies present dramatic red-shifts of 336 and 397 cm1 (with respect to gas phase) for the adsorption on Au1/MgO(O2) and Au1/MgO(Fs) sites, respectively. The last value is very similar to that obtained for free AuNO molecule (375 cm1). These shifts are much greater than any other reported values for NO stretching frequencies on supported Au. With regard to the supported Au dimer, the NO frequency shift is less significant: 160 and 254 cm1 for Au2 deposited on MgO(O2) and MgO(Fs), respectively.
Ackowledgments The authors acknowledge the financial support of Universidad Nacional del Sur, CONICET and ANPCyT of Argentina. References [1] Z. Yang, S. Chinta, A.A. Mohamed, J.P. Fackler, D.W. Goodman, J. Am. Chem. Soc. 127 (2005) 1604. [2] M. Sterrer, M. Yulikov, E. Fischbach, M. Heyde, H. Rust, G. Pacchioni, T. Risse, H.J. Freund, Angew. Chem. Int. Ed. 45 (2006) 2630. [3] A.V. Matveev, K.M. Neyman, I.V. Yudanov, N. Rosch, Surf. Sci. 426 (1999) 123. [4] M. Haruta, Catal. Today 36 (1997) 153. [5] F. Solymosi, T. Bánsági, T.S. Zakar, Phys. Chem. Chem. Phys. 5 (2003) 4724. [6] K. Blick, T.D. Mitrelias, J.S.J. Hargreaves, G.J. Hutchings, R.W. Joyner, C.J. Kiely, F.E. Wagner, Catal. Lett. 50 (1998) 211. [7] A. Ueda, M. Haruta, Gold Bull. 32 (1) (1999) 3. [8] Z. Yang, R. Wu, Q. Zhang, D.W. Goodman, Phys. Rev. B 65 (2002) 155407. [9] A. Del Vitto, G. Pacchioni, F. Delbecq, P. Sautet, J. Phys. Chem. B 109 (2005) 8040. [10] G. Barcaro, A. Fortunelli, J. Chem. Theory Comput. 1 (2005) 972. [11] P. Frondelius, H. Häkkinen, K. Honkala, New J. Phys. 9 (2007) 339. [12] L.M. Molina, J.A. Alonso, J. Phys. Chem. C 111 (2007) 6668. [13] A. Bogicevic, D.R. Jennison, Surf. Sci. 515 (2002) L481. [14] W.A. Brown, D.A. King, J. Phys. Chem. B 104 (2000) 2578. [15] S.M. McClure, T.S. Kim, J.D. Stiehl, P.L. Tanaka, C.B. Mullins, J. Phys. Chem. B 108 (2004) 17952. [16] C.P. Vinod, J.W. Niemantsverdriet Hans, B.E. Nieuwenhuys, Appl. Catal. A 291 (2005) 93. [17] A. Citra, X. Wang, L. Andrews, J. Phys. Chem. A 106 (2002) 3287. [18] X. Ding, Z. Li, J. Yang, J.G. Hou, Q. Zhu, J. Chem. Phys. 121 (2004) 2558. [19] A.D. Becke, J. Chem. Phys. 98 (1993) 5648. [20] G. Pacchioni, P. Pescarmona, Surf. Sci. 412/413 (1998) 657. [21] C. Di Valentin, R.M. Ferullo, R. Binda, G. Pacchioni, Surf. Sci. 600 (2006) 1147. [22] J. Carrasco, N. Lopez, F. Illas, H.-J. Freund, J. Chem. Phys. 125 (2006) 074711. [23] P.J. Hay, W.R. Wadt, J. Chem. Phys. 82 (1985) 270. [24] P.J. Hay, W.R. Wadt, J. Chem. Phys. 82 (1985) 299. [25] D. Andrae, U. Haussermann, M. Dolg, H. Stoll, H. Preuss, Theor. Chim. Acta 77 (1990) 123. [26] I. Yudanov, G. Pacchioni, K. Neyman, N. Rösch, J. Phys. Chem. B 101 (1997) 2786. [27] N.R. Kestner, J.E. Combariza, in: K.B. Lipkowitz, D.B. Boyd (Eds.), Reviews in Computational Chemistry, vol. 13, Wiley-VCH, John Wiley and Sons Inc., New York, 1999 (Chapter 2, p. 99). [28] A.E. Reed, L.A. Curtiss, F. Weinhold, Chem. Rev. 88 (1988) 899. [29] A.M. Joshi, M.H. Tucker, W.N. Delgass, K.T. Thomson, J. Chem. Phys. 125 (2006) 194707. [30] M.J. Frisch et al., Gaussian 03, Revision C.02, Gaussian, Inc., Wallingford, CT, 2004. [31] C. Inntam, L.V. Moskaleva, K.M. Neyman, V.A. Nasluzov, N. Rösch, Appl. Phys. A 82 (2006) 181. [32] M. Yulikov, M. Sterrer, M. Heyde, H.-P. Rust, T. Risse, H.-J. Freund, G. Pacchioni, A. Scagnelli, Phys. Rev. Lett. 96 (2006) 146804. [33] C. Di Valentin, A. Scagnelli, G. Pacchioni, T. Risse, H.-J. Freund, Surf. Sci. 600 (2006) 2434. [34] M. Sterrer, M. Yulikov, T. Risse, H.J. Freund, J. Carrasco, F. Illas, C. Di Valentin, L. Giordano, G. Pacchioni, T. Risse, H.J. Freund, Angew. Chem. Int. Ed. 45 (2006) 2633. [35] H. Grönbeck, P. Broqvist, J. Phys. Chem. B 107 (2003) 12239. [36] M.A. Debeila, N.J. Coville, M.S. Scurrell, G.R. Hearne, J. Mol. Catal. A 219 (2004) 131.