PCCP View Article Online

Published on 17 September 2014. Downloaded by Northeastern University on 29/10/2014 08:32:51.

PAPER

Cite this: Phys. Chem. Chem. Phys., 2014, 16, 23134

View Journal | View Issue

Self-seeded nucleation of Cu nanoclusters on Al2O3/Ni3Al(111): an ab initio investigation† Jimena A. Olmos-Asar,*ab Erik Vesselli,cd Alfonso Baldereschiae and Maria Peressiae The mechanisms of seeding and nucleation of Cu nanoclusters onto an ultrathin alumina template supported on Ni3Al(111) has been investigated by means of ab initio calculations. Single Cu ad-atom diffusion on the oxide film is effective at room temperature, allowing preferential occupation of the defective sites of the so-called ‘‘dot’’ structure, where the adsorption is much stronger than in the ‘‘network’’ or any other surface

Received 23rd July 2014, Accepted 12th September 2014

site of the oxide. After the adsorption of the first Cu atom, further nucleation at the ‘‘dot’’ sites proceeds with

DOI: 10.1039/c4cp03271c

for larger clusters. The whole process is thermodynamically favoured. We therefore clearly confirm and

www.rsc.org/pccp

the growth of highly ordered arrays of small Cu nanoparticles.

the formation of multi-atomic seeds (with up to 6 atoms contained in the defect) that offer stiff anchoring rationalize some experimental evidence showing that the ultrathin Al2O3/Ni3Al(111) is an efficient template for

1 Introduction Metal nanoparticles, with a surface-to-volume ratio higher than that of monocrystalline surfaces, can have unique properties that depend, among other factors, on their size and shape.1,2 Additionally, they have a high concentration of uncoordinated sites, edges and kinks, which makes them interesting for many applications, especially for heterogeneous catalysis. The practical way to use metal nanoparticles as catalysts is to grow them on an appropriate support, which in many cases is an oxide. The support can play a role in the size and size-distribution of the nanoparticles, and can determine and modify their properties, including the catalytic activity and selectivity for specific reactions: for example, the interaction of CO with H2 onto supported Rh has been shown to change for different oxide supports.3,4 Nanoparticles with well defined electronic and optical properties are also desirable in view of the development of new electronic and optoelectronic devices. It is because of these reasons that the synthesis of ordered arrays of well-defined equally sized nanoparticles is the goal of many efforts. Several methods have been successfully applied to nanofabrication, such as electron beam lithography, deposition of a

Department of Physics, University of Trieste, Strada Costiera 11, 34151 Trieste, Italy. E-mail: [email protected] b ´tica y Fı´sica, Facultad de Ciencias Quı´micas, Departamento de Matema Universidad de Co´rdoba, XUA5000 Co´rdoba, Argentina c Department of Physics and CENMAT, University of Trieste, via Valerio 2, 34127 Trieste, Italy d IOM-CNR Laboratory TASC, AREA Science Park, 34149 Basovizza (Trieste), Italy e IOM-CNR National Simulation Center DEMOCRITOS, 34136 Trieste, Italy † Electronic supplementary information (ESI) available. See DOI: 10.1039/ c4cp03271c

23134 | Phys. Chem. Chem. Phys., 2014, 16, 23134--23142

colloidal particles, atomic manipulation using scanning probe microscopes (STM and AFM), and laser-beam assisted deposition, among others.5–7 All these techniques involve direct instrumental manipulations of the systems, are almost independent of the physical properties of the substrate and do not guarantee, in general, to achieve an efficient control of both size and spatial distribution. In recent years, there has been increased interest in surface structures that can serve as templates for growing highly ordered nanostructure arrays through self-organization of adatoms, making use of the intrinsic properties of the substrate. The role of the template is also relevant in preventing sintering at high temperatures, thus yielding deactivation of the catalytic devices. As an example, one or two metallic monolayers epitaxially deposited on a substrate with a different lattice constant arrange into regular dislocations and strain-relief patterns that have been successfully used to create well-ordered nanoscale arrays.8 Several epitaxial systems meet the condition of having dislocations ´ structures and strain-relief patterns periodically arranged. Moiree could also be appropriate, like for example epitaxial graphene, showing periodic variations in the potential energy surface for ad-atoms. Oxide ultrathin films have been recently proposed as alternative, promising substrates. In particular, it has been found that the oxidation of clean NixAl single crystal terminations under ultra high vacuum (UHV) conditions gives rise to highly ordered, extremely homogeneous ultrathin non-stoichiometric alumina films.9–16 The one formed onto Ni3Al(111) is arranged into a pffiffiffiffiffi pffiffiffiffiffi superstructure whose unit cell is ( 67  67)R121 with respect to the supporting alloy surface with a sixfold rotational symmetry. Imaging using STM yields two different appearances, depending on the tunneling conditions: a so-called ‘‘dot’’ structure,

This journal is © the Owner Societies 2014

View Article Online

Published on 17 September 2014. Downloaded by Northeastern University on 29/10/2014 08:32:51.

Paper

PCCP

hexagonal, with a lattice parameter of 4.16 nm and one bright spot per unit cell when the sample bias voltage is around +2.3 V, and a honeycomb-like ‘‘network’’ structure with three depressions per unit cell, at a distance of 2.40 nm one from each other, visible when the tunneling proceeds at a bias voltage around +3.2 V.17–19 pffiffiffi pffiffiffi There exists a ( 3  3)R301 relation between the two structures,14,17,20,21 and both can act as nucleation patterns for the growth of metallic nanoclusters.22–27 The bright spots defining the ‘‘dot’’ structure have been identified as holes in the alumina structure, big enough to be filled with some metallic atoms that can act as a seed for further growth of nanoclusters.20 Experiments under UHV conditions show that vapor deposition at room temperature of Cu onto alumina thin films grown on Ni3Al(111) gives a rather well-ordered hexagonal array of 3D clusters, even at high coverage, nucleating preferentially at the holes describing the ‘‘dot’’ structure of the oxide film.22,23 A similar behavior, although with different details, was observed also for V,22 Mn,23,28 Fe,24 Pd25,29 and Co,26 while it was not the case for deposition of other metals, like Ag at high coverages.23,28 Au, Fe and Co nanoparticles can nucleate at the ‘‘dot’’ sites if Pd atoms are seeded prior to the deposition of the other metal.26,30 Ordered nucleation was explained taking into account the metal–oxide interactions, cohesive, and adhesion energies. It was suggested that kinetics may also play a relevant role.23 In summary, the existing literature indicates that the nucleation process of metallic clusters is rather complex. Beside macroscopic parameters such as impinging metal flux and substrate temperature, the growth process is strongly determined by microscopic details, such as the presence and stability of nucleation centers, the adhesion and mobility of individual metal atoms as well as of the forming clusters, which can change with their size. A comprehensive understanding of the system at the microscopic level is necessary to be able to control the growth process and to identify the best candidates for the production of highly ordered nanoparticles arrays, homogeneous in shape and size. To this purpose we have performed a thorough ab initio investigation on the nucleation of Cu clusters on the Al2O3/ Ni3Al(111) thin film. Previous investigations have been done for Pd seeds by Schmid et al.20 and for Rh atoms and very small clusters (up to 6 atoms) by Khosravian et al.31 Our main goal here is to extend this study not only to another metal but also to the investigation of the atomic-level details, studying the thermodynamics inside the hole as well as on the ‘‘network’’ site and other regions of the oxide surface and calculating the kinetic barriers involved in the mechanisms of seeding. The prediction of the equilibrium structures of larger clusters grown from the seeds is also presented.

used. For the plain wave basis sets, energy cutoffs of 30 Ry and 240 Ry were chosen to describe the wave function and the electronic density, respectively, and the sampling of the first Brillouin zone was performed at the G point. The alumina grown onto Ni3Al(111) under UHV conditions is a highly ordered non-stoichiometric thin film, with four alternated ion layers: Al–O–Al–O (oxygen-terminated). The full structure of the system was determined and subsequently refined by using a combination of ab initio simulations and STM.20 The in-plane periodicity is dictated by the pffiffiffiffiffi pffiffiffiffiffi ( 67  67)R121 superstructure of the oxide film, described by a hexagonal cell containing a hole that reaches the supporting metallic alloy. Considering two layers of Ni3Al(111) under the oxide, the simulation cell contains 1257 atoms (268 atoms in each layer of the alloy and four oxide layers consisting of 132 Al, 188 O, 188 Al, and 213 O, respectively), therefore with a O : Al ratio of about 1.25 instead of 1.50 in the stoichiometric compound. The starting atomic coordinates are those obtained by Schmid et al.20 and already used as a reference structure in ref. 21. Due to the large size of the simulation cell, the calculations have been performed in reduced models obtained by extracting portions consisting of about 130 atoms in the zone of interest, around the ‘‘dot’’ and ‘‘network’’ sites. To check the validity of our reduced models, the calculation of the adsorption

2 Computational details

Fig. 1 (top) Top view of the structural model of the Al2O3/Ni3Al(111), with the sketch of the periodically repeated unit cell. Directions of the primitive vectors of the underlying alloy are also indicated (a 2 nm  2 nm portion is shown in inset). Light blue spots highlight the ‘‘dot’’ structure and small triangles mark the ‘‘network’’. The white circle and square identify the reduced models considered for calculations. (bottom) Side view of one unit cell. Atomic coordinates are those obtained by Schmidt et al.20 Red: O, green: Al, blue: Ni.

Calculations were performed within the density functional theory approach implemented in the Quantum-Espresso package32 with available ultrasoft pseudopotentials.33 For the exchange and correlation the spin unrestricted generalized gradient approximation in the Perdew–Burke–Ernzerhof34 implementation was

This journal is © the Owner Societies 2014

Phys. Chem. Chem. Phys., 2014, 16, 23134--23142 | 23135

View Article Online

PCCP

Paper

Published on 17 September 2014. Downloaded by Northeastern University on 29/10/2014 08:32:51.

energy of a single Cu atom in the most stable site was performed on a larger substrate (330-atoms model), and no meaningful deviations were found. The complete periodic cell is shown in Fig. 1; the reduced models are marked on it. Enough vacuum has been added in all cartesian coordinates to assure no interaction among the repeated images.

3 Results 3.1

Table 1 Adsorption energies and equilibrium heights (from the closest topmost O atoms) of Cu in the ‘‘dot’’, ‘‘network’’, and other sites of the oxide surface shown in Fig. 2a and b

Site

Eads (eV)

h (Å)

Site

Eads (eV)

h (Å)

D NT NB NH T1 H1

2.12 1.12 1.33 0.98 0.97 0.40

3.42 1.98 1.78 1.91 2.04 2.08

H2 H3 B1 B2 L1 L2

0.54 0.37 0.60 0.45 0.56 0.93

2.09 2.39 2.09 2.19 1.94 1.99

Interaction of Cu ad-atoms with the oxide surface

3.1.1 Single atom adsorption. The adsorption energy of a single Cu ad-atom is defined as: Eads = ESubs+Cu  ESubs  ECu

(1)

where ESubs+Cu is the energy of the whole system, ESubs is the energy of the bare supported alumina substrate and ECu the one of an isolated Cu atom. With this definition, stable configurations are characterized by negative adsorption energies. On the basis of the experimental results and of the DFT calculations available for Pd and Rh, we examine first of all the Cu adsorption on the two primary nucleation sites, ‘‘dot’’ and ‘‘network’’. The adsorption of Cu in the ‘‘dot’’ site (D) is indeed quite strong, with an energy of 2.12 eV when the Cu atom is in the most stable position at the bottom of the hole, slightly displaced offcenter. For comparison, the adsorption energy for a Rh atom in the same site, reported as its preferential nucleation site, is 3.5 eV.31 The ‘‘network’’ site, reported as the most stable adsorption position for other metals, is a triangular structure formed by three O ions (Fig. 1 and 2a), therefore three different adsorption configurations have to be considered: on-top of one O atom (NT), in between two O atoms (NB) and in the middle of the triangle (NH). The results are reported in Table 1. It is interesting to note that the adsorption is stronger in the positions breaking the threefold symmetry of the ‘‘network’’ structure (Eads on NB = 1.33 eV and on NT = 1.12 eV). These values are remarkably

weaker than in the ‘‘dot’’, similarly to Rh, although in that case the difference is much larger (Eads on ‘‘network’’ site is reported to be only 0.18 eV for Rh31). Therefore, DFT suggests that the ‘‘dot’’ site is the preferential adsorption site also for Cu. In the experiments of chemical vapour deposition (CVD) onto alumina22,23 the Cu atoms impinging uniformly and randomly on the surface have a much larger probability of interacting with the outermost alumina layer rather than immediately and directly with the hole: in fact, the unit cell parameter of the Ni3Al(111) surface is 2.45 Å while the one of the ‘‘dot’’ structure is 41.5 Å, yielding a coverage of holes of 3.5  103 monolayers referred to the metallic alloy surface. It is essential therefore to investigate the Cu interaction with other portions of the surface, in terms of adsorption and diffusion. In Table 1 we summarise the results of Cu adsorption in some sites in the region of the hole, indicated in Fig. 2b: they are representative also of other sites of the oxide surface, characterized by a similar local coordination. In most oxide surface sites the strength of adsorption energy is about 0.5 eV, and in any case smaller than 1 eV. Comparing the results of Table 1, we can conclude that, after the ‘‘dot’’, the ‘‘network’’ is the second stable site for adsorption (and maybe possible nucleation) of Cu on alumina/Ni3Al(111), followed by the adsorption on-top of some surface O ions (Eads on T1 = 0.97 eV). In general, it has been reported that for several transition metals the adsorption energy in the hole is typically 2 eV larger than that on

Fig. 2 (a, b) Relevant adsorption sites for Cu onto the alumina thin film on Ni3Al(111) in the neighborhood of the ‘‘network’’ and ‘‘dot’’ sites: NT, NB, and NH are ‘‘network’’ top, bridge and hollow sites, respectively; T1, on-top an O atom; H1 and H2, hollow three-coordinated sites; H3, hollow tetracoordinated site; B1 and B2, bridge bi-coordinated sites; L1 and L2, limit sites in the border of the hole; D, most stable position inside the hole. (c) Selected diffusion paths for Cu onto alumina close to the hole. Red: O, green: Al, blue: Ni.

23136 | Phys. Chem. Chem. Phys., 2014, 16, 23134--23142

This journal is © the Owner Societies 2014

View Article Online

Paper

PCCP

Published on 17 September 2014. Downloaded by Northeastern University on 29/10/2014 08:32:51.

Table 2 Barrier heights for the diffusion of Cu onto alumina along selected paths. See the Fig. 2 caption for labels

Path

Barrier (eV)

Path

Barrier (eV)

T1H1H2 T1H3T1 T1B2T1 T1H1T1 T1L1

0.52 0.45 0.56 0.60 0.46

T1L2 T1H3L2 L1C L2C

0.24 1.02 0.00 0.29

the oxide surface.20 Cu follows therefore this general trend, although the difference between the adsorption energy inside and outside the hole is smaller (0.79 eV considering NB and 1.15 eV considering T1). It is also interesting to notice that an adsorption energy of 0.99 eV is reported for Cu on-top of an O atom of MgO films.35 3.1.2 Single atom diffusion. A question arising for a thorough understanding of the nucleation process is whether the Cu atoms arriving to the substrate through CVD can diffuse on the alumina surface at room temperature, possibly reaching the most stable adsorption sites and giving origin to the nucleation process. To address this point we have calculated the most relevant diffusion barriers onto the surface and the possible additional one, known as the Ehrlich–Schwoebel barrier,36,37 necessary to reach the bottom of the hole. Some representative paths are shown in Fig. 2c. The label C indicates the position at the center of the hole. The transition states along the paths have been found by optimizing the height of the copper atom, and the corresponding barriers are reported in Table 2. As can be observed, the diffusion along most of the selected paths is related to barriers in the order of 0.5 eV. Due to this fact, it is expected that Cu atoms would be able to easily diffuse along the surface of the alumina oxide after thermalizing at room temperature. Additionally, the residual kinetic energy gained by the ad-atom upon adsorption from the vapor phase, although reduced after some dissipation, is expected to further improve diffusion. 3.2

Adsorption of a single Cu atom in the hole

The Cu ad-atoms can therefore easily diffuse on the oxide surface and reach the hole. In order to investigate in detail the adsorption in the defect, we have considered first of all the potential energy for a single Cu atom descending vertically inside it. We caution the reader that this is not the minimum energy path, but it allows us to obtain an upperbound for the energies involved. The energy profile for the Cu atom as a function of its height along the axis of the hole is shown in Fig. 3a (dots). The height is referred to the average position of the oxygen atoms in the outermost ionic layer of the oxide. For further reference, the average positions of the other ionic layers of the alumina, as well as the Ni3Al alloy, are also indicated (colored stripes). The potential energy is calculated analogously to the adsorption energy in eqn (1), i.e., as the difference between the total energy of the system with the Cu atom adsorbed and the energies of the supported alumina and the Cu atom at infinite distance. The deepest minimum energy position is at about 3.1 Å (inside the hole), a height corresponding to the interface Al layer of the oxide, with a value of 1.91 eV. This is basically

This journal is © the Owner Societies 2014

Fig. 3 (a) Energy profile of a Cu atom (dots) descending vertically inside the hole of the alumina film as a function of its height (the square marker corresponds to the fully relaxed configuration) and of a second Cu atom (triangles) when the first is already adsorbed in the deepest minimum. See the text for details. (b) Complete energy profile for a Cu atom entering the hole after diffusion on the oxide surface through: (i) the combination of path T1L1 and L1C of Fig. 2 and the vertical descent shown in the uppermost panel (dots); (ii) the combination of path T1L2 and L2C and the vertical descent (triangles). The energy is referred to a Cu atom at infinite distance from the oxide. Dotted lines delimit the beginning and the end of each combined path. Vertical colored stripes indicate the position of the alumina ions layers and of the Ni3Al(111) alloy. Insets schematize the represented paths in a side view. Red, O layers; green, Al; blue, Ni3Al(111).

equivalent to the adsorption energy of a single Cu atom on-top of an Al site of a clean Ni3Al(111) surface (choosing this position for direct comparison with the adsorption inside the hole), which is 1.89 eV, consistent with the fact that the main bond is between the Cu atom and the alloy substrate. As already mentioned, in the fully relaxed configuration the Cu atom moves slightly towards the walls of the hole (Fig. 4), and the adsorption energy becomes 2.12 eV (square marker in Fig. 3a, site D in Table 1). The charge density difference (see ESI,† Fig. S1) clearly shows the formation of a covalent bond between the Cu and the Al atom immediately below, belonging to the Ni3Al alloy. Along the descending path of a single Cu atom (Fig. 3a, dots), a relative minimum is found outside the hole, at about 0.7 Å above the outermost oxygen layer. The adsorption energy in this position is 0.64 eV, comparable to the value on some

Phys. Chem. Chem. Phys., 2014, 16, 23134--23142 | 23137

View Article Online

Published on 17 September 2014. Downloaded by Northeastern University on 29/10/2014 08:32:51.

PCCP

Paper

sites are separated by diffusion barriers of some tenths of eV. Indeed, it has been already suggested that the nucleation at other sites in some cases, such as Fe, is related to the existence of higher Ehrlich–Schwoebel barriers to enter the hole.20 A closer comparison of our results for Cu with those of other atoms is not possible since detailed information on barriers on this substrate is not available in the literature. To summarise, our findings for the single Cu atom adsorption and diffusion shed light on previous experimental results showing that the formation of Cu clusters onto alumina preferentially follows the ‘‘dot’’ structure. 3.3

Nucleation of multi-atomic seeds in the holes

We have shown that one Cu atom attaches preferentially deep inside the hole after arriving from CVD. Nevertheless, the defect seems to be big enough to contain larger seeds. We have then studied the energetics of additional Cu atoms entering into the hole. The definition of the adsorption energy for the Nth Cu atom in the presence of N  1 atoms already adsorbed can be generalized as: hole EAds = ESubs+NCu  ESubs+(N1)Cu  ECu N

Fig. 4 Side view of the equilibrium configurations for Cu seeds on Al2O3/ Ni3Al(111) composed of up to six atoms. For 1 and 6-atoms seeds, front view is also shown. Adsorption energies are indicated (see the text for details). Red balls, O atoms; green, Al; blue, Ni; brown, Cu.

sites of the oxide surface outside the hole (see Table 1). In this case, the charge density difference analysis does not show the formation of a chemical bond, but an increase in the density between the Cu and the ion layers in the oxide. The energy profile shows that a Cu atom entering vertically in the hole should overcome a small energy barrier of about 0.07 eV from the position of the outermost minimum to reach the deepest one. We have estimated the barrier necessary to reach the hole merging selected diffusion paths on the surface up to reaching the position C with a vertical descent down in the hole. In Fig. 3b, the energy profile along two sample paths are shown, one obtained from the paths T1L1 and L1C (dots) and the other from T1L2 and L2C (triangles) of Fig. 2b, plus the descent down in the hole (Fig. 3a). Globally, Fig. 3b shows that a Cu atom adsorbed on top of the alumina film has to overcome a barrier which is at most of about 0.5 eV to diffuse and enter the hole. In conclusion, it is expected that a Cu atom diffuses easily on the oxide at room temperature and eventually enters the hole, thus sticking at this preferential site due to the deep depression in the potential energy surface. We suggest that not only thermodynamics but also kinetics plays an important role, since the adsorption of Cu atoms in the hole rather than in other sites (specially on the ‘‘network’’) is not as strongly favored as it is reported for other metals, and the competing

23138 | Phys. Chem. Chem. Phys., 2014, 16, 23134--23142

(2)

where ESubs+NCu is the total energy of the system with N Cu atoms and ESubs+(N1)Cu refers to the system with N  1 atoms adsorbed in the most stable configuration. As done for the first Cu atom, it is interesting to follow the energy profile for a second one descending vertically in the hole when one Cu is already adsorbed in the most stable position. Also in this case, we mention that this is not the lowest energy path, but gives an estimate of the energy surface explored. As can be observed in Fig. 3a (triangles), the equilibrium position for the second Cu atom along this path is at about 1.1 Å (below the surface) and is characterized by an energy of 1.00 eV. Fig. 3a also shows a change in the curvature of the potential energy surface for the second atom (where smaller pulling forces would be felt) in correspondence to the local minimum for the first Cu atom. If a full relaxation of the system is allowed, two equilibrium structures can be obtained, as shown in Fig. 4, with adsorption energies as defined in eqn (2). Average adsorption energies can be calculated as: hEAds i ¼

1 ðESubsþNCu  ESubs  NECu Þ N

(3)

and are also displayed. Interatomic distances are listed in Table S1 (ESI†). In one configuration, the two Cu atoms are almost vertically one on top of the other at a distance of 2.49 Å, in analogy to what was calculated for Pd.20 Remarkably, we found a second, much more stable configuration, where both Cu atoms are accommodated at the bottom of the hole, interacting with the oxide and with the alloy, at 2.26 Å one from each other. For comparison, a Cu dimer in vacuum is characterized by an interatomic distance of 2.22 Å and a binding energy of 2.27 eV. On the basis of the interatomic distances and of the charge density plots (shown in Fig. S1, ESI†) we can describe the first configuration as formed by a first atom in the hole that has already saturated its valence with the outermost Al atom of

This journal is © the Owner Societies 2014

View Article Online

Published on 17 September 2014. Downloaded by Northeastern University on 29/10/2014 08:32:51.

Paper

the metallic substrate, and a second Cu which, after arrival, forms consequently a much weaker Cu–Cu bond with respect to the dimer in vacuum. The most stable configuration could be instead described as a dimer bound deep inside the hole. We could argue also that the latter configuration is accessible if the second atom enters the hole following a path along the wall opposite to the position of the first one, displacing it and accommodating at the bottom, or if both enter as an already formed dimer. Three Cu atoms can also accommodate in the hole in different equilibrium configurations. One corresponds to an almost vertical and linear arrangement (Fig. 4), totally filling the hole, similarly to the Pd case.20 Considering as starting configuration the one with two Cu atoms vertically accommodated, the third atom pushes them down. Its position is slightly above the outermost O layer, but still below the one of a Cu ad-atom adsorbed on the alumina surface. This result strongly suggests that a third atom stabilizes also the vertical 2-atom seed. A less stable 3-atom configuration is a triangular one, with two atoms at the bottom of the hole and the third one above them in the bridge position. Also in this case, distances (Table S1, ESI†), energies (Fig. 4) and charge density plots (Fig. S1, ESI†) suggest a well defined picture: in the case of the almost vertical seed, the first atom at the bottom of the hole is mainly bound to the Al of the metallic substrate, and the other two form a dimer; in the case of the triangular configuration, the dimer is at the bottom of the hole and the third atom forms with them a weaker bond. Newly with respect to what was previously reported for Pd,20 more than three Cu atoms can be accommodated inside a single defect. Indeed, seeds of up to 6 Cu atoms are thermodynamically stable. The configurations and adsorption energies of the stable seeds are shown in Fig. 4. As can be deduced from the geometries and distances (Table S1, ESI†), many of these structures can be thought as dimers attached to the alloy and/or to other Cu atoms or dimers. Moreover, Cu atoms are always oriented in a way to maximize the interaction with the ionic layers in the oxide. It is noticeable that the adsorption of individual atoms attaching one by one inside the hole is, in most cases, stronger than in the surface. In some configurations, the adsorption energy is comparable to the one in the ‘‘network’’ sites, and we expect a competition when Cu atoms arrive to the surface through CVD. These results strongly support the fact that the holes can act as nucleation centers not only for a first Cu atom, but, when partially filled with N  1 atoms, also for the additional N-th one arriving in the neighborhoods. In summary, we have studied the adsorption of Cu in three different regions of the Al2O3/Ni3Al(111): inside the holes, on the ‘‘network’’, and in the remaining positions of the terraces. It is important to recall that, although adsorption in the holes and ‘‘network’’ is thermodynamically favoured, the concentration of these sites is low. Nevertheless, if Cu deposition occurs at room temperature like in ref. 22, by means of simply evaluating the site population on the basis of Boltzmann statistics we found that the first stages of nucleation of Cu in the hole is several

This journal is © the Owner Societies 2014

PCCP

orders of magnitudes more favorable than on the alumina surface. In addition to that, since the potential energy well in correspondence of the large holes is expected to be wider than the shape of the minima at stable adsorption sites for Cu at the surface, both zero-point energy and Arrhenius pre-exponential factors may contribute as well in favoring nucleation at holes. Would this be the case, a kinetic contribution to the nucleation process may be also relevant. Furthermore, our results extend what has been proposed for Pd seeds,20 remarkably indicating that a vertical seed is not the only possible configuration for 2 and 3 atoms in the hole and that up to 6 Cu atoms could be accommodated inside. We suggest that these findings could be extended to other metals also, especially those characterized by interatomic distances smaller or not much larger than Cu. 3.4

Cu dimers

If clusters grow from ad-atoms impinging from CVD and thereafter diffusing as a low density 2D gas on the surface, a possible alternative process may be involved in the formation of a dimer if two atoms meet on the surface. Depending on its binding energy to the oxide surface and on the diffusion rate, this rare event could be thought as the first step for nucleation of Cu clusters. For this reason, we have investigated the interaction of a Cu dimer with the alumina surface. The initial testing structures were chosen by taking into account that Cu ad-atoms adsorb preferentially at on-top O sites. We have found that in the most stable configuration, the dimer binds through only one Cu atom to a terminal oxygen ion in a vertical configuration, with an angle of 101 with respect to the surface normal (geometry and charge density difference plots are displayed in Fig. S2, ESI†). The Cu–Cu distance is 2.24 Å, very similar to the one in a dimer in the gas phase, while the O–Cu distance is 2.07 Å, very close to the Cu–O distance for a single ad-atom on the T1 site (see Table 1). Our results are similar to what has been previously observed for Cu on MgO(100).38–40 By defining the dimer adsorption energy on the alumina surface as: Ads-s ECu ¼ ESubsþCu2  ESubs  ECu2 2

(4)

where ESubs+Cu2 is the energy of the whole system and ECu2 is the energy of a dimer in vacuum, we obtain a value of 0.96 eV. This is about the same as that of a single Cu ad-atom on the surface (Table 1). The adsorption energy of the dimer at the bottom of the hole is 1.86 eV. If we consider the configuration with three Cu atoms in the hole in an almost vertical and linear arrangement (see Fig. 4, uppermost right panel), the atoms 2 and 3 form a dimer, whose adsorption energy can be calculated as: Ads-hole-Cu ECu ¼ ESubsþ3Cu  ESubsþ1Cu  ECu2 2

(5)

which gives 1.30 eV, weaker than in the case of the dimer at the bottom of the hole but in any case stronger than the adsorption on the surface. The energy required to dissociate the dimer adsorbed on the surface into two ad-atoms is defined as:39   diss (6) ECu ¼  ESubsþCu2 þ ESubs  2ESubsþCu 2

Phys. Chem. Chem. Phys., 2014, 16, 23134--23142 | 23139

View Article Online

PCCP

Published on 17 September 2014. Downloaded by Northeastern University on 29/10/2014 08:32:51.

For this quantity we get a value of +1.43 eV, thus showing that a Cu dimer is more stable than two Cu ad-atoms separately adsorbed at the oxide surface, since the Cu–Cu bond is stronger than the Cu–O bond. This dissociation energy is lower than the one for the dimer in vacuum, similarly to what has been found for a Rh2 dimer.31

4 Growth of the Cu clusters at the anchoring sites If seeding occurs atom by atom in a slow CVD, one possible and factible scenario is the one in which the hole is filled by three Cu atoms in an almost vertical and linear arrangement (see Fig. 4, uppermost right panel). Taking this fact as a base, we have investigated the growth of Cu nanoclusters composed of up to 15 atoms, adding them one by one on the previously nucleated seed and relaxing the whole system. The geometries obtained are shown in Fig. 5a. For comparison, the relaxed structures of free-standing copper clusters are also shown in Fig. 5b. The conformations are in good agreement with previous calculations (see ref. 41 and references therein). As the supported clusters have two atoms deep inside the hole, the comparison is done between the N + 2-atom supported cluster and the N-atom cluster in

Paper

vacuum, although we have previously considered as ‘‘seed’’ the configuration with three atoms. It is evident that the substrate is not only determining the ordered growth of the clusters in a preferential site, but is also defining their structure. For instance, 4–6 atoms free-standing Cu clusters are planar, while 6–8 supported clusters are not, mostly due to interactions with oxygen in the top-most layer of the thin oxide film. As a first approximation, due to the fact that on average the Cu–Cu bonds are stronger than the Cu–oxide interaction (the binding energy of a Cu dimer in vacuum is 2.27 eV per atom, the cohesive energy of Cu in bulk is 3.54 eV per atom, whereas the highest adsorption energy of Cu on the oxide film around the hole is at the best 0.97 eV), we could state that the clusters will not ‘‘wet’’ the alumina, and will grow in a 3D conformation, as previously found for Cu on MgO(100).39,42 We expect this fact to be true for small clusters, and specially for those that we have studied. Nevertheless, for larger clusters grown atom by atom from Cu deposited through CVD, kinetics barriers could exist for the new metallic atoms arriving, to rearrange in the nanostructure. The cohesive energies per Cu atom for N + 2-supported and N-free-standing clusters are shown in Fig. 6. In the supported case, ESubs+(N+2)Cu being the total energy of the system and ESubs+2Cu the energy of the substrate with two Cu atoms adsorbed inside the hole, the cohesive energy for the N atoms outside the hole can be calculated as: EC ¼

ESubsþðNþ2ÞCu  ESubsþ2Cu  NECu N

(7)

In the case of free-standing Cu clusters, the cohesive energy is simply: EC ¼

ECuN  NECu N

(8)

where ECuN is the total energy of the N-atom free-standing cluster. The cohesive energy is (in absolute value) higher in the case of supported clusters, and this is due to the interactions with the two Cu atoms remaining in the seed and with the oxygen

Fig. 5 Equilibrium structures (a) of N + 2 Cu clusters grown on alumina and (b) of the corresponding N Cu clusters in vacuum. Red: O; green: Al; blue: Ni; ochre: Cu.

23140 | Phys. Chem. Chem. Phys., 2014, 16, 23134--23142

Fig. 6 Cohesive energy per Cu atom for N + 2-atom supported and N-atom free-standing clusters. Dotted line shows cohesive energy of Cu in the bulk phase.

This journal is © the Owner Societies 2014

View Article Online

Paper

ions at the surface of the oxide. In the limit of very large 3D nanoparticles, when the surface-to-volume ratio is smaller and the interaction with the substrate is negligible, these two curves would converge to each other and to the cohesive energy of bulk Cu.

Published on 17 September 2014. Downloaded by Northeastern University on 29/10/2014 08:32:51.

5 Conclusions We have studied within the DFT approach the interaction of Cu atoms with an alumina ultra-thin film grown onto a Ni3Al(111) surface. By means of a thorough description of the energy profile along the most representative adsorption sites, we have demonstrated that Cu has an important preference for adsorption inside the holes of the so-called ‘‘dot’’ structure, in which the first Cu atom is attracted deeply inside the defect and a covalent bond with the supporting alloy is created. Adsorption at the ‘‘network’’ structure sites is less stable. The differences in adsorption energies, even taking into account the low concentration of these special sites, assure that on the basis of energetic arguments most of the Cu atoms will eventually reach the holes and nucleate there. Remarkably, up to 6 Cu atoms can be accommodated inside the defects, and many different seeds were found to be thermodynamically stable. The strong adhesion of Cu to the supporting metallic alloy guarantees the stability of the nucleation centers, thus offering stiff anchoring for further growth of 3D Cu clusters. Kinetics is not opposing to the nucleation process of Cu clusters at the holes. The most relevant barriers for lateral diffusion on the alumina surface as well as the barrier for entering the hole are small enough to be easily overcome under the conditions of the CVD. In conclusion, both thermodynamics and kinetics contribute to make thin alumina films supported on Ni3Al(111) a good template for the growth of highly ordered Cu nanoclusters, in good agreement with previous experimental results. The study of the morphology and cohesive energy of Cu clusters of different size, with up to 15 atoms, nucleated at the hole, shows that the substrate not only defines the nucleation centers, but also affects the shape of the nanoclusters.

Acknowledgements We would like to thank once again M. Schmid for providing the coordinates of the substrate model, also used in ref. 21. We acknowledge financial support from: Italian Ministry of University and Research through Futuro in Ricerca, FIRB 2010 project RBFR10J4H7; Italian Ministry of Foreign Affairs, Directorate General for the Country Promotion, through the Executive Programme with Argentina (PGR00190); Consortium for Physics of Trieste, Italy. Computational resources have been partly obtained through Italian Super-Computing Resource Allocation (ISCRA) grants of the Consorzio Interuniversitario CINECA, partly within the agreement between the University of Trieste and CINECA.

This journal is © the Owner Societies 2014

PCCP

References 1 F. Baletto and R. Ferrando, Rev. Mod. Phys., 2005, 77, 371–423. 2 R. Ferrando, J. Jellinek and R. L. Johnston, Chem. Rev., 2008, 108, 845–910. 3 T. Ioannides and X. Verykios, J. Catal., 1993, 140, 353–369. 4 T. Ioannides, A. Efstathiou, Z. Zhang and X. Verykios, J. Catal., 1995, 156, 265–272. 5 J. J. McClelland, R. E. Scholten, E. C. Palm and R. J. Celotta, Science, 1993, 262, 877–880. 6 P. Avouris, Acc. Chem. Res., 1995, 28, 95–102. 7 A. N. Shipway, E. Katz and I. Willner, ChemPhysChem, 2000, 1, 18–52. 8 H. Brune, M. Giovannini, K. Bromann and K. Kern, Nature, 1998, 394, 451–453. 9 U. Bardi, A. Atrei and G. Rovida, Surf. Sci., 1990, 239, L511–L516. 10 U. Bardi, A. Atrei and G. Rovida, Surf. Sci., 1992, 268, 87–97. 11 C. Becker, J. Kandler, H. Raaf, R. Linke, T. Pelster, M. Drager, M. Tanemura and K. Wandelt, Papers from the 44th national symposium of the AVS, 1998, 16, 1000–1005. 12 A. Rosenhahn, J. Schneider, J. Kandler, C. Becker and K. Wandelt, Surf. Sci., 1999, 433–435, 705–710. 13 R. Franchy, Surf. Sci. Rep., 2000, 38, 195–294. 14 A. Rosenhahn, J. Schneider, C. Becker and K. Wandelt, The 46th international symposium of the american vacuum society, 2000, 18, 1923–1927. 15 R. M. Jaeger, et al., Surf. Sci., 1991, 259, 235. 16 G. Kresse, M. Schmid, E. Napetschnig, M. Shishkin, ¨hler and P. Varga, Science, 2005, 308, 1440. L. Ko 17 S. Degen, A. Krupski, M. Kralj, A. Langner, C. Becker, M. Sokolowski and K. Wandelt, Surf. Sci., 2005, 576, L57–L64. 18 S. Gritschneder, S. Degen, C. Becker, K. Wandelt and M. Reichling, Phys. Rev. B: Condens. Matter Mater. Phys., 2007, 76, 014123. 19 S. Gritschneder, C. Becker, K. Wandelt and M. Reichling, J. Am. Chem. Soc., 2007, 129, 4925–4928. 20 M. Schmid, G. Kresse, A. Buchsbaum, E. Napetschnig, S. Gritschneder, M. Reichling and P. Varga, Phys. Rev. Lett., 2007, 99, 196104. 21 E. Vesselli, A. Baraldi, S. Lizzit and G. Comelli, Phys. Rev. Lett., 2010, 105, 046102. 22 A. Wiltner, A. Rosenhahn, J. Schneider, C. Becker, P. Pervan, M. Milun, M. Kralj and K. Wandelt, Thin Solid Films, 2001, 400, 71–75. 23 C. Becker, A. Rosenhahn, A. Wiltner, K. von Bergmann, J. Schneider, P. Pervan, M. Milun, M. Kralj and K. Wandelt, New J. Phys., 2002, 4, 75. 24 A. Lehnert, A. Krupski, S. Degen, K. Franke, R. Decker, S. Rusponi, M. Kralj, C. Becker, H. Brune and K. Wandelt, Surf. Sci., 2006, 600, 1804–1808. 25 M. Marsault, G. H. A. Worz, G. Sitja, C. Barth and C. R. Henry, Faraday Discuss., 2008, 138, 407–420. 26 A. Buchsbaum, M. De Santis, H. C. N. Tolentino, M. Schmid and P. Varga, Phys. Rev. B: Condens. Matter Mater. Phys., 2010, 81, 115420.

Phys. Chem. Chem. Phys., 2014, 16, 23134--23142 | 23141

View Article Online

Published on 17 September 2014. Downloaded by Northeastern University on 29/10/2014 08:32:51.

PCCP

27 L. Gragnaniello, T. Ma, G. Barcaro, L. Sementa, F. R. Negreiros, A. Fortunelli, S. Surnev and F. P. Netzer, Phys. Rev. Lett., 2012, 108, 195507. 28 C. Becker, K. von Bergmann, A. Rosenhahn, J. Schneider and K. Wandelt, Surf. Sci., 2001, 486, L443–L448. 29 S. Degen, C. Becker and K. Wandelt, Faraday Discuss., 2004, 125, 343–356. 30 G. Hamm, C. Becker and C. Henry, Nanotechnology, 2006, 17, 1943–1947. 31 H. Khosravian, Y. Lei, A. Uhl, M. Trenary and R. J. Meyer, Chem. Phys. Lett., 2013, 555, 7–11. 32 P. Giannozzi, S. Baroni, N. Bonini, M. Calandra, R. Car, C. Cavazzoni, D. Ceresoli, G. L. Chiarotti, M. Cococcioni, I. Dabo, A. D. Corso, S. de Gironcoli, S. Fabris, G. Fratesi, R. Gebauer, U. Gerstmann, C. Gougoussis, A. Kokalj, M. Lazzeri, L. Martin-Samos, N. Marzari, F. Mauri, R. Mazzarello, S. Paolini, A. Pasquarello, L. Paulatto, C. Sbraccia, S. Scandolo, G. Sclauzero, A. P. Seitsonen, A. Smogunov, P. Umari and R. M. Wentzcovitch, J. Phys.: Condens. Matter, 2009, 21, 395502.

23142 | Phys. Chem. Chem. Phys., 2014, 16, 23134--23142

Paper

33 D. Vanderbilt, Phys. Rev. B: Condens. Matter Mater. Phys., 1990, 41, 7892–7895. 34 J. P. Perdew, K. Burke and M. Ernzerhof, Phys. Rev. Lett., 1996, 77, 3865–3868. 35 V. Musolino, A. Selloni and R. Car, Surf. Sci., 1998, 402–404, 413–417. 36 G. Ehrlich and F. G. Hudda, J. Chem. Phys., 1966, 44, 1039–1049. 37 R. L. Schwoebel and E. J. Shipsey, J. Appl. Phys., 1966, 37, 3682–3686. 38 G. Barcaro and A. Fortunelli, J. Chem. Theory Comput., 2005, 1, 972–985. 39 V. Musolino, A. Dal Corso and A. Selloni, Phys. Rev. Lett., 1999, 83, 2761–2764. 40 V. Musolino, A. Selloni and R. Car, Phys. Rev. Lett., 1999, 83, 3242–3245. 41 Z. Cao, Y. Wang, J. Zhu, W. Wu and Q. Zhang, J. Phys. Chem. B, 2002, 106, 9649–9654. 42 J. Zhou and T. Gustafsson, Surf. Sci., 1997, 375, 221–225.

This journal is © the Owner Societies 2014

Ni3Al(111): an ab initio investigation.

The mechanisms of seeding and nucleation of Cu nanoclusters onto an ultrathin alumina template supported on Ni3Al(111) has been investigated by means ...
5MB Sizes 2 Downloads 4 Views