Alumina supported Pt nanoclusters under a hydrogen environment play a crucial role in many heterogeneous catalysis applications. We conducted grand canonical genetic algorithm simulations for supported Pt8 clusters in a hydrogen gas environment to study the intracluster, cluster-support, and cluster-adsorbate interactions. Two alumina surfaces, α-Al2O3(0001) and γ-Al2O3(100), and two conditions, T = 600 °C, pH2 = 0.1 bar and T = 25 °C, pH2 = 1.0 bar, were considered corresponding to low and high hydrogen chemical potential μH, respectively. The low free energy ensemble of Pt8 is decorated by a medium (2–12 H), respectively, high (20–30 H), number of hydrogen atoms under equilibrium at low μH, respectively, high μH, and undergoes different morphological transformations on the two surfaces. On α-Al2O3(0001), Pt8 is mostly 3D but very fluxional in structure at low μH and converts to open one-layer 2D structures with minimal fluxionality at high μH, whereas on γ-Al2O3(100), the exact opposite occurs: Pt8 clusters present one-layer 2D shapes at low μH and switch to compact 3D shapes under high μH, during which the Pt8 cluster preserves moderate fluxionality. Further analysis reveals a similar Pt–Pt bond length increase when switching from low μH to high μH on both surfaces although morphological transformations are different. Electronic structure analysis shows the existence of bonding interactions between Pt and Lewis acidic Al3+ sites along with the Pt–O interaction, which implies the necessity to include Al neighbors to discuss the electronic structure of small Pt clusters.

Heterogeneous catalysis plays critical roles in many important industrial processes, which produce valuable chemicals. In several of these processes, supported metal clusters and nanoparticles are used as catalysts, where the clusters/nanoparticles are the active components, and high surface area materials are used as supports.1–3 Usual support materials are metal oxides, such as alumina or silica. Oxide supports are primarily used for stabilizing metal clusters, but extensive studies also show that the oxide can influence the catalytic properties of metal clusters.4 On the other hand, the catalyst is interacting with the reactants at operating temperatures. The reactant can adsorb on the metal clusters and modify their geometrical and electronic structures, forming the real active phase which can be different from the initial material placed in the reactor, and such a structure transition usually takes place at the reaction induction period.5,6 Therefore, the metal clusters, the oxide supports, and the reactants constitute the basic components for the formation of the catalytic active site, and understanding their mutual interactions is required.

The supports play a particularly important role in governing the catalytic properties when subnanometer size metal clusters are considered.7–9 In general, metal oxides used as supports in catalysis can be classified as nonreducible (e.g., Al2O3 and SiO2) and reducible (e.g., CeO2 and TiO2). Reducible oxides are characterized by easily accessible changes in the surface stoichiometry, forming oxygen vacancies and then interacting differently with supported metal clusters. For example, it is known that in reducing conditions, metal clusters can even be buried by the reduced TiO2x.10 This phenomenon is referred to as the strong metal-support interaction.11–13 The nonreducible metal oxides are more stable at catalytic conditions although this does not mean that they are inert in catalytic reactions. First, the unsaturated Al cations on alumina show strong Lewis acidity14 which is important for many catalytic reactions such as hydrocarbon cracking and biomass reformation.15 Second, the alumina prepared under different temperatures shows different crystalline phases as well as various levels of surface hydration, which also provides different properties.16 

Under reaction conditions, metal clusters react with reactants forming the cluster-adsorbate complexes, which are the catalytically active species at steady-state conditions.17 The cluster-adsorbate complex could be substantially different from the precursors. Since the morphology and properties of metal clusters can be altered by the adsorbate binding, it is crucial to characterize and understand the adsorbate induced reconstructions. For example, under a CO environment, supported Rh clusters can be disintegrated into small fragments [such as Rh-(CO)n]18,19 and the process is also reversible when CO is removed. This process can be completed within several seconds, which is very fast compared with the time scale in real world experiments. This means that the actual catalytic species cannot be identified if the experimental characterization is conducted under vacuum and ex situ conditions.

Sub-nano-Pt clusters supported on alumina in a hydrogen environment are important catalysts for many important catalytic reactions, including hydrogenation,20–22 NOx removal,23 CO oxidation,24,25 dehydrogenation,26 and reforming.27,28 However, these are challenging systems because the catalytic properties of Pt@Al2O3 are sensitive to the environment and the optimal performance requires a careful setting of reaction conditions. The chemical environment exerts a strong influence on the catalytic properties of the Pt clusters. For example, in the alkane dehydrogenation reaction, the partial pressure of the H2 gas has a vital impact on the lifetime of the catalyst, by avoiding coking.26 The property of the alumina support also impacts the catalytic reactions. For example, He et al. showed that the pore size distribution and the inherent acidity of Al2O3 change the lifetime of the catalysts.29 The acidity of the Al2O3 surface can be modified by SiO2 or Cs2O doping, which in turn alters the selectivity toward hydrogenolysis by-products.30 

Several theoretical studies investigated the effect of hydrogen adsorption on the structure and electronic properties of small Pt clusters. Those studies underlined the phenomenon of morphological transition when the environment is switched. Mager-Maury et al. investigated the morphology of Pt13 clusters in a wide range of hydrogen coverage. In this theoretical work, the Pt13 cluster is supported on the γ-Al2O3(100) surface and the structure of Pt13Hx is explored by molecular dynamics.31 The results show that at low hydrogen coverage, the Pt13 cluster presents a biplanar shape, but the shape converts to a cuboctahedron at high hydrogen coverage. Hu et al. showed that the structure of the cluster is also affected by the choice of the alumina support. The biplanar shape of Pt13 on γ-Al2O3(100), with both Pt–O and Pt–Al bonds, is transformed to a 3D shape, which is hydroxylated after standard temperature treatment, and the cluster only weakly interacts with the surface hydroxyl groups.32 Another study by Erfani et al. also showed that the morphology of Pt clusters on γ-Al2O3(110) is interconvertible by varying the surface hydroxyl coverage,33 which alters the strength of Pt-surface interaction. The Pt22 cluster shows a strong adhesion on the γ-Al2O3(110) surface and a 2D shape at low hydroxyl coverage, while it transforms to a 3D structure when the hydroxyl coverage is greater than 0.325 ML. However, they showed that such morphological transition only takes place for the small Pt22 cluster and is not observed for a larger Pt44 cluster.33 At finite temperature in catalytic conditions, thermal fluctuations induce cluster isomerization and formation of metastable structures, which are minor in the population but still accessible, and could have properties that are distinct from those of the global minimum (GM) structure.34 Alexandrova and Anderson showed that in other applications, in the absence of added hydrogen, H-free Pt7 and Pt8 on α-Al2O3 feature 3D global minima, while Pt7 additionally has a thermal access to flatter isomers that wet the support, and those flat isomers (rather than the global minimum) in fact govern the catalytic activity of the system.34,35 The shape transition is also demonstrated by experimental results from in situ HERFD-XANES (high energy resolution fluorescence detection X-ray absorption near edge structure).36 Mistry et al. studied the electronic and structural properties of Pt clusters on γ-Al2O3(100) under different hydrogen pressures. They found that when the partial pressure of H2 increased from 1 to 21 bars, the H/Pt ratio changes from 1.9 to about 2.5 with a morphological transition from 2D to 3D.37 

The present contribution follows such efforts to understand and rationalize the effect of supports and the hydrogen environment on the electronic and morphological properties of Pt nanoclusters. Small Pt clusters are very flexible in geometry,35,38–43 and their global minimum and other lowest-energy accessible isomers cannot be simply guessed. Additionally, hydrogen adsorption also alters the cluster geometries, further complicating the structure identification. Therefore, we exploited a grand canonical genetic algorithm (GCGA) method, which optimizes the geometry and adsorbate coverage simultaneously for supported Pt8 clusters. These simulations produce structures that are quasifree energy minima, in the sense that they take into account the internal electronic energies of the systems and the change in the chemical potential of hydrogen upon adsorption. For simplicity, the vibrational entropy is not accounted. Furthermore, we characterize the cluster systems not in terms of just one global minimum but as ensembles of many thermally accessible structural states. It has recently been shown that these higher-energy states are critical in determining the properties of cluster catalysts.34,35,41–44

In order to investigate the influence of the hydrogen environment, two different conditions were considered in this work. The first one (T = 600 °C and pH2 = 0.1 bar) corresponds to the standard propane dehydrogenation condition, in which hydrogen is cofed with propane to prevent coke formation. We refer to it as low μH because of high T and low pH2. The second one (T = 25 °C and pH2 = 1 bar) is close to a typical hydrogenation reaction, and we refer to it as high μH in the text. In addition, two different surfaces are modeled, e.g., γ-Al2O3(100) and α-Al2O3(0001), and they are examples of surfaces with different activities due to the different Lewis acidity and basicity.

This manuscript is organized as follows. In Sec. II, we present calculation details and the grand canonical genetic algorithm (GCGA) used in this work. Then, in Sec. III, we discuss the structures obtained with GCGA and the influence of the support surface and the hydrogen reactant on the cluster structure. Finally, in Sec. IV, we discussed the geometry transitions and the electronic structures of supported Pt8Hx clusters under different conditions.

Since the hydrogen coverage of the small Pt cluster varies strongly with temperature and hydrogen pressure, determining the structure of the PtnHx cluster in specific conditions is challenging. The standard approach is to explore each coverage x independently and then evaluate the phase diagram based on the optimal geometry at different compositions. This approach has two major limitations. First, it is based on a single global minimum structure, whereas we now know that this representation is far from the reality at catalytic temperatures. In fact, more realistic “hot” phase diagrams based on ensembles of many accessible states have been recently introduced.45 Second, one needs to perform all x global optimizations for every n in PtnHx, i.e., tens of costly simulations for every single system. Considering the large system size, this approach is not practical. Therefore, we investigated the structures of supported Pt8Hx clusters in the presence of a hydrogen reservoir by the grand canonical genetic algorithm (GCGA). GCGA is a global optimization method designed for structural exploration at a fixed chemical potential rather than a fixed composition. Similar methods have been applied to explore the Zr–Cu–Al alloy with an empirical force field, 2D materials, etc.46 

The GCGA algorithm works similarly to the conventional GA47,48 (or evolutionary algorithm) but aims at optimizing the free energy of the target system rather than the potential energy, i.e., minimizing ΔG in the following equation:

ΔG=E(Pt8Hx+slab)E(slab)8EPtatomxμH,
(1)

where E(Pt8Hx + slab) is the electronic energy of the optimized structure, E(slab) is the energy of the optimized slab model, EPtatom is the energy of free Pt atom, and μH is the chemical potential of hydrogen in the gas phase. μH is then calculated by the following equation:31 

μH(T,p)=12GH20(T)h0K+kBTlnpp0,
(2)

where T and p are the temperature and partial pressure of hydrogen, respectively. GH20(T) is the standard free energy of H2 at T. h0K is the enthalpy correction of H2 at 0 K, and kB is the Boltzmann constant.

GCGA consists of several components, some of which are the same as conventional GA and others are altered for the current grand canonical (GC) approaches. The first element is the structure evolution algorithms, which includes two categories of operations: the mating (or heredity) and mutation. The mating operation generates one offspring structure from two parent structures. A detailed technical implementation of the mating operation with fixed composition can be found elsewhere.47 In the GC approach, the algorithm needs to be modified because the two parents may have different stoichiometries (i.e., different numbers of hydrogen atoms). In our implementation, an offspring structure is first generated from two parents as if there were no hydrogen atoms in the system. Second, according to the original connectivity matrix of the parent structure, the hydrogen atom is also inherited to the offspring structure if one of the H binding sites (Pt atoms) was inherited. In some cases, the inherited hydrogen atoms will overlap with other H or Pt atoms in the generated structures. These overlapping hydrogen atoms are removed before optimizations. In this way, the final offspring candidate may have different coverages from its two parents. However, the bonds between H atoms and their anchoring Pt atoms are still conserved and inherited from the parents.

The mutation operation in GA is usually performed with several methods to change the positions of atoms, including (a) random displacement and (b) reflection with mirror symmetry. Two additional operations are introduced in the current GCGA method. They are (c) adding and (d) removing one adsorbate (hydrogen in this case). The adding operation is completed by inserting an additional hydrogen atom in the vicinity of a randomly selected Pt atom while avoiding the overlap with other atoms within 0.7 × (ri + rj) (ri is the covalent radius of atom i). The removing operation is performed by randomly deleting a hydrogen atom from the system.

Some other elements of GA, such as the evaluation of the fitness and the determination of identical structures, follow the algorithm similar to the one introduced by Vilhelmsen et al.47 The method exploited in this contribution is developed based on the atomic simulation environment (ASE) package. A blueprint of the GCGA is shown in Fig. S1. Initially, structures (called seeds) are randomly generated on the slab support; typically, they have high free energies [defined in Eq. (1)]. Hydrogen coverage is also randomly selected and hence not expected to be optimal. During the GCGA optimization, the number of hydrogen atoms and the free energy of the system evolve simultaneously until convergence.

We exploit two different crystalline alumina surfaces: α-Al2O3(0001) and γ-Al2O3(100), which have different surface structures. The Al3+ terminated α-Al2O3(0001) contains only one type of Al site, a 3-coordinated Al cation. It is known to be the most stable surface termination in a wide range of oxygen chemical potentials (μO) and hence is used as the first model in this work. A 4 × 4 supercell of α-Al2O3(0001) with 4 Al–O–Al trilayers (total 12 atomic layers) is extracted from bulk and optimized. During the optimization of the bare surface, the top-most layer Al3+ cations relax inward significantly (the interlayer distance between Al and O becomes only 0.127 Å). The bottom two layers are then fixed at the optimized positions (from previous slab optimization) in all the following structure optimizations. The chemical formula of our α-Al2O3(0001) slab model is Al128O192, and its structure is shown in Fig. S2(a).

The second surface is γ-Al2O3(100), which is built from the crystalline structure reported by Digne et al.49,50γ-Al2O3(100) contains two types of Al3+ sites, a saturated subsurface tetrahedral Al cation and a 5-coordinated and unsaturated Al cations, originating from octahedral Al in the bulk. In order to reduce the lateral interaction between different supported clusters, we used a 2 × 2 supercell of γ-Al2O3(100) with the dimensions of 16.761 Å × 16.826 Å [shown in Fig. S2(b)].

During the global optimization, we exploited two density functional theory (DFT) approaches. In the first stage, the Perdew-Burke-Ernzerhof (PBE) functional51 with moderate size (double-zeta) atomic basis sets (implemented in CP2K package52) is used. DFT with this basis set is very efficient in computation, and therefore it enables GCGA to quickly approach an approximate ensemble of lowest-energy structures and hydrogen coverages. However, the downside of the small basis sets is the basis set superposition error (BSSE), which biased the predicted coverage and favors compact geometries. To mitigate the impact of BSSE, the molecularly optimized basis sets (MOLOPT) were used within CP2K.53 The MOLOPT basis sets are optimized to minimize the BSSE down to the level of traditional augmented basis sets.

Finally, we resume and complete every GCGA run using plane waves based DFT calculations implemented in VASP. The VASP calculations exploit plane wave basis sets up to an energy cutoff of 400 eV, the PAW pseudopotentials, and the PBE exchange-correlation functional.51 Because we use a rather large supercell, only the Γ point is sampled in the reciprocal space for the Brillouin zone integration. Structure optimizations are conducted with the ASE (https://wiki.fysik.dtu.dk/ase/) package using the BFGS algorithm until the maximum residual force is below 0.02 eV/Å in each local optimization.

In Secs. III A and III B, we explore the structure of Pt8Hx at low μH. In total, we have obtained 2756 unique structures for the Pt8Hx@α-Al2O3(0001) system and 1330 unique structures for the Pt8Hx@γ-Al2O3(100) case. The obtained global minimum structures of the two systems are shown in Fig. 1, and many other structures in the low free energy ensemble (LFEE), defined as the all the sampled unique structures within 0.5 eV free energy interval from the GM, are given in the supplementary material (Fig. S4). The hydrogen coverages of the GMs are Pt8H4@α-Al2O3(0001) and Pt8H5@γ-Al2O3(100).

FIG. 1.

The two global minimum structures of the Pt8Hx cluster under a pressure of hydrogen at low μH (600 °C and 0.1 bar of H2). (a) and (c) are Pt8H4 supported on α-Al2O3(0001). (b) and (d) are Pt8H5 supported on γ-Al2O3(100). Red: oxygen atoms, blue: alumina atoms, gray: platinum atoms, and pink: hydrogen atoms. (a) Pt8H4@α-Al2O3(0001) sideview, (b) Pt8H5@γ-Al2O3(100) sideview, (c) Pt8H4@α-Al2O3(0001) topview, and (d) Pt8H5@γ-Al2O3(100) topview

FIG. 1.

The two global minimum structures of the Pt8Hx cluster under a pressure of hydrogen at low μH (600 °C and 0.1 bar of H2). (a) and (c) are Pt8H4 supported on α-Al2O3(0001). (b) and (d) are Pt8H5 supported on γ-Al2O3(100). Red: oxygen atoms, blue: alumina atoms, gray: platinum atoms, and pink: hydrogen atoms. (a) Pt8H4@α-Al2O3(0001) sideview, (b) Pt8H5@γ-Al2O3(100) sideview, (c) Pt8H4@α-Al2O3(0001) topview, and (d) Pt8H5@γ-Al2O3(100) topview

Close modal

Figures 1(a) and 1(c) show the GM structure of Pt8H4@α-Al2O3(0001), which has a pseudo-2-layer geometry. Two of the four hydrogen atoms adsorb on the bridge sites, and the other two are on the top sites. The Pt8H4 cluster interacts with the α-Al2O3(0001) surface through Pt–O bonds with 9 oxygen atoms within 2.5 Å of the cluster, and the cluster is also closely contacting at short distance (∼2.5 Å) with 4 surface Al cations [shown in Fig. S7(a)].

To understand the interaction between the support and the cluster, the free energy of the Pt8Hx-alumina system is decomposed into three terms including the free energy of the Pt8Hx clusters in the absence of the substrate, the (free) energy change of support upon Pt8Hx adsorption in the absence of the cluster, and the interaction between the Pt8Hx cluster and the deformed support. The results are shown in Fig. S8. The α-Al2O3(0001) of the GM structure Pt8H4@α-Al2O3(0001) undergoes rather strong deformation. Four Al3+ cations underneath the Pt8H4 cluster, which are close to Pt, stretch outward from the surface by ∼0.5 Å. This deformed structure of the surface presents, in the absence of the Pt8H4 cluster, an energy of 4.48 eV higher than its minimum energy. These so-called deformation energies of support in other LFEE structures are in the range of 3–5 eV (without hydroxyls) and 5–7 eV [with hydroxyls, shown in Fig. S8(a)].

At low μH and hence low H coverage, the Pt8Hx cluster possesses a large number of low free energy metastable isomers, as shown in Fig. 2(a). In total, 57 unique structures are found for Pt8Hx@α-Al2O3(0001) within 0.5 eV from the GM. Considering the high temperature (∼600 °C), all isomers in LFEE present reasonable populations estimated by a free-energy based Boltzmann probabilities, which are larger than 0.1% as shown in Fig. 2(a), while 5 metastable isomers show a probability higher than 10%. The density of minima (the number of unique structures within a free energy range) is moderate around the global minimum, and it becomes larger at higher free energies. This phenomenon is similar to that of the ensemble with fixed composition. The hydrogen coverage shows large variations on the LFEE, from 1 to 12 H atoms on the cluster, and generally hydrogen atoms prefer to adsorb on either the top or bridge sites [Fig. 2(c)], and we did not observe any hydrogen at hollow sites. In some cases, a maximum of two hydrogen atoms can spillover to the surface forming hydroxyl. It is known that the bare α-Al2O3(0001) surface does not adsorb H atoms even at very high μH.54,55 Therefore, as we will see later from the electronic analysis, the observed hydroxyl formation is facilitated and stabilized by the Pt8Hx clusters. The root-mean-squired deviation (RMSD) of the eight Pt atoms’ heights is analyzed to determine the shape ratio of the Pt8Hx clusters on the support. Figure 2(a) shows that the RMSDs are very diverse within the LFEE and range from 0.4 Å (flat morphology) to 1.0 Å (pseudo-2-layer shape), as shown in Fig. S4. Interestingly, the planar isomers in the LFEE only appear at higher coverage(x > 8), being accompanied with formation of surface hydroxyls. However, the first single layer isomer is only 0.24 eV less stable than the GM (shown as the 8th structure in Fig. S4), and the Boltzmann factor predicts a significant population corresponding to 7.0% of that of the GM at 600 °C. Hence, there exist different morphologies for the cluster: a bilayer with low H coverage (majority shape) and a monolayer with the medium H coverage (minority shape). Distinct geometries in metastable isomers for the bare alumina supported Pt7 cluster were previously reported by Baxter et al.35 

FIG. 2.

Structural and energetic analysis of the low free energy ensemble for Pt8 on alumina under a pressure of hydrogen. In each subfigure, the top panel shows the relative stability and population estimated by a Boltzmann distribution at a specific temperature. The left axis of the bottom panel shows the numbers of hydrogen atoms, which are decomposed into different types (i.e., top, bridge, hollow on Pt and hydroxyl on surface oxygen). The right axis of the bottom panel gives the RMSD of the 8 Pt atoms’ heights, which is used for indicating the thickness of the cluster. (a) Pt8Hx@α-Al2O3(0001) low μH, (b) Pt8Hx@γ-Al2O3(100) low μH, (c) Pt8Hx@α-Al2O3(0001) high μH, and (d) Pt8Hx@γ-Al2O3(100) high μH.

FIG. 2.

Structural and energetic analysis of the low free energy ensemble for Pt8 on alumina under a pressure of hydrogen. In each subfigure, the top panel shows the relative stability and population estimated by a Boltzmann distribution at a specific temperature. The left axis of the bottom panel shows the numbers of hydrogen atoms, which are decomposed into different types (i.e., top, bridge, hollow on Pt and hydroxyl on surface oxygen). The right axis of the bottom panel gives the RMSD of the 8 Pt atoms’ heights, which is used for indicating the thickness of the cluster. (a) Pt8Hx@α-Al2O3(0001) low μH, (b) Pt8Hx@γ-Al2O3(100) low μH, (c) Pt8Hx@α-Al2O3(0001) high μH, and (d) Pt8Hx@γ-Al2O3(100) high μH.

Close modal

The GM structure of the Pt8Hx on γ-Al2O3(100) surface shows a hydrogen coverage [Pt8H5@γ-Al2O3(100)] similar to that on α-Al2O3(0001), but has a different geometry [Figs. 1(b) and 1(d)], with a one-layer compact morphology. All five hydrogen atoms adsorb on top sites of edge Pt atoms. The Pt atoms form eight Pt–O bonds and five Pt–Al bonds with the γ-Al2O3(100) surface, which is similar to the previous α-Al2O3(0001) surface (Fig. S7b). The GM of the Pt8H5 cluster, however, induced weaker deformation of the γ-Al2O3(100) than that of the α-Al2O3(0001) surface, and the γ-Al2O3(100) surface only slightly relaxed upon Pt8H5 adsorption. The calculated deformation energy of the γ-Al2O3(100) surface is 3.04 eV for the GM structure [Fig. S8(b)]. The stability of the isolated cluster Pt8H5 (indicated by its free energy) is very similar to that of the isolated Pt8H4 on α-Al2O3(0001) surface although they have different geometries and coverages. However, the interaction energy between the clusters and surfaces is much smaller on the γ-Al2O3(100) surface [−8 to −14 eV, Fig. S8(b)] than that on the α-Al2O3(0001) surface [−12 to −20 eV, Fig. S8(a)]. The different responses of the two alumina surfaces to the cluster adsorption may originate from different surface structures and reactivities. The α-Al2O3(0001) contains many 3-coordinated unsaturated Al3+ cations, which are very Lewis acidic. However, the γ-Al2O3(100) surface contains only 5-coordinated unsaturated cations, which are less Lewis acidic than those of the α-Al2O3(0001) surface. We calculated the surface energies of the current α-Al2O3(0001) and γ-Al2O3(100) slabs (without testing the convergence of its thickness), and they are 0.118 eV/Å2 and 0.075 eV/Å2, respectively. The larger surface energy of α-Al2O3(0001) also implies its higher activity.

The Pt8Hx clusters on γ-Al2O3(100) are also fluxional at the considered low μH. 27 unique structures are found within the LFEE, and their relative free energies are shown in Fig. 2(b). Three metastable isomers have populations larger than 10% at the considered temperature (600 °C). The RMSDs of Pt atoms’ heights show that nearly all the minima have one-layer morphologies except for just one isomer, which is a bilayer and appears at 0.46 eV above the GM (Fig. S3). The hydrogen coordination analysis shows that all the hydrogen atoms adsorb on the top or the bridge sites, and we did not observe the formation of hydroxyls in this ensemble. The distribution of H coverage is also narrower (from 4 to 8 hydrogen atoms).

Our simulation results can be compared with the experimental work by Sinkler et al.56 They studied small Pt clusters supported on γ-Al2O3 with aberration-corrected transmission electron microscopy. They find that the Pt clusters present a one-layer and two-dimensional morphology with a diameter around 0.88 nm after reduction at around 700 °C by H2.56 The clusters in this work are supposed to consist of 7–13 Pt atoms, i.e., the sizes are very similar to our calculations. The conclusions from experiments are hence consistent with our computational results.

In the case of high μH, we obtained 1269 unique structures for Pt8Hx@α-Al2O3(0001) and 1325 unique structures for Pt8Hx@γ-Al2O3(100). The GM structures are shown in Fig. 3. Both of them contain 24 H atoms.

FIG. 3.

The two GM structures of the Pt8Hx cluster. (a) and (c) are side and top views of clusters supported on α-Al2O3(0001). (b) and (d) are side and top views of clusters supported on γ-Al2O3(100). The condition refers to 25 °C and 1 bar of H2, and the chemical formula of the clusters is Pt8H24 on both surfaces. Red: oxygen atoms, blue: alumina atoms, gray: platinum atoms, and pink: hydrogen atoms. (a) Pt8H24@α-Al2O3(0001) sideview, (b) Pt8H24@γ-Al2O3(100) sideview, (c) Pt8H24@α-Al2O3(0001) topview, and (d) Pt8H24@γ-Al2O3(100) topview

FIG. 3.

The two GM structures of the Pt8Hx cluster. (a) and (c) are side and top views of clusters supported on α-Al2O3(0001). (b) and (d) are side and top views of clusters supported on γ-Al2O3(100). The condition refers to 25 °C and 1 bar of H2, and the chemical formula of the clusters is Pt8H24 on both surfaces. Red: oxygen atoms, blue: alumina atoms, gray: platinum atoms, and pink: hydrogen atoms. (a) Pt8H24@α-Al2O3(0001) sideview, (b) Pt8H24@γ-Al2O3(100) sideview, (c) Pt8H24@α-Al2O3(0001) topview, and (d) Pt8H24@γ-Al2O3(100) topview

Close modal

The GM structure of Pt8H24@α-Al2O3(0001) is shown in Figs. 3(a) and 3(c). The Pt8H24 cluster adheres very strongly to the α-Al2O3(0001) surface, resulting in a 1-layer, 2-dimensional and open morphology. One Pt atom has only one neighbor Pt, and four of them have only two Pt neighbors, with hydrogen atoms appearing at the bridge sites. The formation of the hydroxyl group is also observed in the GM structure, and that is similar to some cases in the ensemble at low μH on α-Al2O3(0001).

The size of LFEE in this case of high H coverage is only 2, which is much smaller than that at the low μH case on α-Al2O3(0001). We already observed in our previous study of isolated Pt13 clusters that high coverage of H adsorption was decreasing the cluster fluxionality.34 Here, in addition, the cluster experiences a strong interaction with the support, freezing it even more. Although the two structures are very similar in the morphology (Fig. S6), the metastable structure is 0.48 eV higher than the GM. Considering the low temperature (25 °C) in this case, we can expect that the GM is the only observable structure. The deformation energies of the supports among these LFEE structures are about 3.9 eV (Fig. S8), which are smaller than that of the α-Al2O3(0001) and low μH case, especially when compared with structures containing hydroxyls [Fig. S8(a)]. The 2-dimensional Pt8H24, if standing alone without support, is not a preferable structure in terms of the free energy due to the small number of Pt–Pt bonds [Fig. S8(c)]. However, the flat geometry facilitates a large cluster-support interaction (∼−15 eV), which finally stabilizes the system.

The GM of Pt8Hx@γ-Al2O3(100) is shown in Figs. 3(b) and 3(d). The morphology of Pt8H24@γ-Al2O3(100) is very different from that of the GM of Pt8H24@α-Al2O3(0001). The Pt8H24 cluster only weakly adheres to the surface forming a compact and hydrogen covered 3-D cluster. A single Pt atom at the bottom of the cluster interacts with a surface hollow site formed by two oxygen atoms and one Al atom. Because of the small adhesion between the cluster and γ-Al2O3(100), the deformation energies of γ-Al2O3(100) among the LFEE structures are very small [below 1.0 eV in most cases, shown in Fig. S8(d)]. However, the Pt8Hx clusters, if standalone, are much more stable than those with the flat geometry on α-Al2O3(0001) [by more than 5.0 eV when comparing Figs. S8(c) and S8(d)].

There are 8 unique structures found in the LFEE, and all of them show a 3D morphology similar to the GM structures shown in Figs. 3(b) and 3(d). All of them have an even count of H adsorbates. The larger size of LFEE, compared with that of Pt8Hx on α-Al2O3(0001) at high μH, implies that the cluster’s fluxionality is only partially decreased by the H adsorbates due to the weak interaction between the cluster and the support [Fig. S8(d)]. The hydrogen atoms still preferentially adsorb on the top or bridge sites, but some hollow sites are also occupied on the 3D clusters.

The shape transition of the Pt8/γ-Al2O3(100) cluster with increasing hydrogen coverage is similar to that of the Pt13 cluster studied by Mager-Maury et al.31 With low hydrogen coverage, both Pt13 and Pt8 preferentially form compact shapes that develop a large interface with the support. With a high hydrogen coverage, both of them tend to form compact 3-D shapes with a reduced interfacial area with the support, hence maximizing the number of Pt–H bonds. The hydrogen coverage for both cluster sizes is also similar under the considered μH (θ ∼ 3 H/Pt at 25 °C and 1.0 bar).31 

The support and hydrogen adsorbates hence compete for the interaction with Pt atoms, and on γ-Al2O3(100), H atoms win. This is not the case on α-Al2O3(0001). The energy decomposition analysis in Fig. S8 shows different driving forces to stabilize the system on each alumina surface, which result in two different reconstruction phenomena. On the α-Al2O3(0001) surface, the interactions between the α-Al2O3(0001) and Pt8 clusters are overall strong, especially when the clusters present a flat geometry. Under high μH, the deformation of the α-Al2O3(0001) surfaces are weakened even with flat Pt8Hx clusters; hence, the Pt8Hx clusters prefer to retain a flat geometry under high μH. On the γ-Al2O3(100), the interactions between the cluster and support are generally smaller, hence dewetting and compact geometry formation prevail.

The radial distribution functions (RDFs) of Pt–Al, Pt–O, and Pt–Pt are shown in Fig. 4. For each system, we compare the GM based RDF with the LFEE averaged RDF. The LFEE averaged RDF is generally similar to GM based RDF although the former filters out some detailed RDF features. Figure 4(a) shows the results of Pt8Hx@α-Al2O3(0001) under low μH, hence low H coverage. Some short O–Pt distances are seen (d ∼ 2.1 Å), which is very close to the Pt–O distance in PtO bulk (d = 2.02 Å). However, the averaged first-neighbor Pt–O distance in this system is 2.30 Å (counting all Pt–O pairs within 3.0 Å) being significantly longer than that from the bulk. The long Pt–O distance implies a weaker Pt–O interaction than that of bulk PtO. The averaged first-neighbor Al and Pt atoms are located at around 2.65 Å. This Pt–Al distance is very close to some Pt–Al alloys57 and molecular complexes,58 but significantly longer than the sum of the covalent radius of Pt and the ionic radius of Al (1.9 Å). The first-neighbor Pt–Pt bond is significantly strained with smaller distance (2.57 Å) than that in the Pt bulk (2.78 Å). The RDFs of Pt8Hx@γ-Al2O3(100) under low μH show that the shortest Pt–O bonds are slightly longer than that of Pt8Hx@α-Al2O3(0001), but the averaged first-neighbor distance is almost the same (∼2.32 Å). The averaged first-neighbor Pt–Al (2.68 Å) and Pt–Pt (2.61 Å) distances are also very close to that of Pt8Hx@α-Al2O3(0001).

FIG. 4.

Radial distribution functions for Pt–Al, Pt–O, and Pt–Pt bonds, in different LFEE. In each subfigure, the solid line shows the radial distribution function (RDF) evaluated only from the GM structure (GM set) and the shaded area shows the RDF from averaging on the LFEE structures (ES set), with the relative weights from the Boltzmann distribution. Each RDF is Gaussian-smoothed with σ = 0.05; (a) Pt8Hx@α-Al2O3(0001) at low μH, (b) Pt8Hx@γ-Al2O3(100) at low μH, (c) Pt8Hx@α-Al2O3(0001) at high μH, and (d) Pt8Hx@γ-Al2O3(100) at high μH.

FIG. 4.

Radial distribution functions for Pt–Al, Pt–O, and Pt–Pt bonds, in different LFEE. In each subfigure, the solid line shows the radial distribution function (RDF) evaluated only from the GM structure (GM set) and the shaded area shows the RDF from averaging on the LFEE structures (ES set), with the relative weights from the Boltzmann distribution. Each RDF is Gaussian-smoothed with σ = 0.05; (a) Pt8Hx@α-Al2O3(0001) at low μH, (b) Pt8Hx@γ-Al2O3(100) at low μH, (c) Pt8Hx@α-Al2O3(0001) at high μH, and (d) Pt8Hx@γ-Al2O3(100) at high μH.

Close modal

The RDFs of structures under high μH cases, where each Pt atom interacts with several H atoms, are shown in Figs. 4(c) and 4(d). For Pt8Hx@α-Al2O3(0001), the averaged first-neighbor Pt–O bond length becomes longer (2.43 Å) than that of Pt8Hx@α-Al2O3(0001) under low μH, implying a decreased interaction between Pt and O. The Pt–Pt bonds also become longer and more diverse. The averaged first-neighbor Pt–Pt bond length is around 2.79 Å, which is almost 0.2 Å longer than that of the Pt8Hx@α-Al2O3(0001) under low μH. In contrast to the Pt–O and Pt–Pt bonds, the first-neighbor Pt–Al bonds do not change a lot and the averaged first-neighbor Pt–Al is 2.59 Å. Some Pt–Al distances slightly decrease below 2.5 Å and others slightly increase. For Pt8Hx@γ-Al2O3(100), it is recalled that the Pt8Hx clusters form very compact geometries and only weakly adhere to the γ-Al2O3(100) surface with a few Pt–Al and Pt–O bonds. The Pt–Pt bond length for this system is around 2.84 Å, which is similar to that of Pt8Hx@α-Al2O3(0001) under high μH but much longer than those clusters under low μH. Under high μH with low temperature T = 25 °C, GM plays the dominant role in the LFEE averaged RDF, and metastable isomers have very small probabilities; therefore, the GM based RDF and the LFEE averaged RDF are basically identical in Figs. 4(c) and 4(d). The contraction of Pt–Pt bond lengths with increasing temperature for small supported Pt clusters has been reported in the experiments of Kang et al.59 They find that in the presence of H2, the Pt–Pt bond lengths of Pt clusters (d = 0.9 nm) will decrease at higher temperature (e.g., at a lower μH). This phenomenon is also verified from our simulation results. In addition, our results reveal that the changes of the bond lengths take place with different cluster reconstructions. On α-Al2O3(0001) and with increased hydrogen coverage (higher μH and lower T), the Pt8Hx cluster prefers to strengthen its interaction with the α-Al2O3(0001) surface and wet the surface. The first-neighbor Pt–Pt becomes longer during this reconstruction. On the other hand, γ-Al2O3(100) supported Pt8Hx clusters are flat at low μH with contracted Pt–Pt bond lengths. With high μH, the clusters become compact 3D morphology and Pt–Pt bond lengths become longer during the reconstruction. Although the reconstruction processes are surface dependent, the variation trend of first neighbor Pt–Pt bond lengths is common and naturally linked with bond order conservation principle.

The Bader charge analysis of the LFEE is shown in Fig. 5 with different conditions and surfaces. The Bader charges of all hydrogen atoms {q(Hx)}, all Pt atoms {q(Pt8)} and the variation of the Bader charges of surface Al {Δq(Aln)} and O atoms {Δq(Om)} are shown for each structure as a function of its stability order in the specific LFEE.

FIG. 5.

Bader charge analysis on the low free energy ensemble of the Pt8Hx cluster at high and low μH conditions, on α-Al2O3(0001) and γ-Al2O3(100) surfaces. q(Hx) is the total charges of all the hydrogen atoms (except for those forming hydroxyls), and q(Pt8) is the total charges of eight Pt atoms. Δq(On) and Δq(Alm) indicate the changes of the O and Al Bader charges upon Pt8Hx adsorption on the support, and only the first neighbor O or Al atoms are considered. q(OH) is the Bader charges of those H atoms from hydroxyl groups. Since different clusters may contain a different number of hydrogen atoms, the per hydrogen Bader charge (excluding H in hydroxyl groups) is also shown. (a) α-Al2O3(0001) low μH, (b) γ-Al2O3(100) low μH, (c) α-Al2O3(0001) high μH, and (d) γ-Al2O3(100) high μH.

FIG. 5.

Bader charge analysis on the low free energy ensemble of the Pt8Hx cluster at high and low μH conditions, on α-Al2O3(0001) and γ-Al2O3(100) surfaces. q(Hx) is the total charges of all the hydrogen atoms (except for those forming hydroxyls), and q(Pt8) is the total charges of eight Pt atoms. Δq(On) and Δq(Alm) indicate the changes of the O and Al Bader charges upon Pt8Hx adsorption on the support, and only the first neighbor O or Al atoms are considered. q(OH) is the Bader charges of those H atoms from hydroxyl groups. Since different clusters may contain a different number of hydrogen atoms, the per hydrogen Bader charge (excluding H in hydroxyl groups) is also shown. (a) α-Al2O3(0001) low μH, (b) γ-Al2O3(100) low μH, (c) α-Al2O3(0001) high μH, and (d) γ-Al2O3(100) high μH.

Close modal

Figure 5(a) shows the Bader charges for LFEE under low μH on α-Al2O3(0001). The Pt8 clusters in this LFEE are always negatively charged holding 0.8e–1.8e in total, resulting from electronic donation by the support. The amount of charge transfer has a strong correlation with the formation of surface hydroxyls. When surface hydroxyls are present, the Pt8 cluster plays the role of an electron reservoir, withdrawing electrons released by the formation of the proton spillover. This explains the reasonable stability of a few OH groups in the presence of the cluster. If reducible oxides, such as TiO2, are used as support, the support reduction could be an alternative mechanism to accommodate the transferred electrons.60 The hydroxyl formation increases the surface deformation energy slightly (around 1–2 eV larger than that of unhydroxlated surface) but strengthens the interaction between the Pt8Hx1 (or Pt8Hx2) clusters and the hydroxylated surface [∼5 eV stronger than others shown in Fig. S8(a)]. In contrast to the negatively charged Pt8, the hydrogen atoms (excluding those from hydroxyl groups) are overall neutral. The surface [i.e., α-Al2O3(0001)] overall donates electrons to the Pt8Hx clusters, and these electrons mainly come from the O atoms of the supports. The Al3+ cations in contrast receive electronic charge from the Pt atoms, as a result of the Lewis-base/Lewis-acid interaction. This partial electron transfer toward surface Al3+ cations is also accompanied by the surface relaxation upon cluster adsorption. Compared with the bare surface, Al3+ cations relax outward by 0.5 Å forming longer Al–O bonds.

At low μH and on the γ-Al2O3(100) surface [Fig. 5(b)], the charge transfer between the surface and Pt8Hx clusters is less significant than that of Pt8Hx@α-Al2O3(0001). The total amount of negative charge that Pt8 clusters accept is almost constant for all the structures in LFEE, ranging from −0.5 to −0.75. Again, H atoms only show a very small negative charge, smaller than 0.2 e in total. The surface is also less affected than α-Al2O3(0001). Oxygen atoms donate a fraction of electronic charge to the Pt8Hx clusters, but the charge transferred to Al3+ cations is almost zero for all the LFEE structures. This is indicative of the weaker Lewis acid character of the γ-Al2O3(100) surface compared to the α-Al2O3(0001) surface.

At high μH and hence high H coverage (θ ∼ 3.0) on the α-Al2O3(0001) surface, although the negative charge per H atom remains small (0.06 e), the total charge present in the 23 H atoms (excluding H in hydroxyl) becomes significant (1.4 e). As a result, the Pt part of the cluster becomes nearly neutral, despite the electronic donation from the surface O atoms and from the presence of a spilled over proton. Similarly to the case at low μH on α-Al2O3(0001), electronic charge is transferred to the first-neighbor Al3+ cations. The electronic charge transfers at high μH on the γ-Al2O3(100) surface are very different because the cluster dewets from the oxide and forms only one Pt–Al and/or one Pt–O bond depending on isomers. As a result, the electronic donation from O atoms is much smaller. Surface Al cations are not affected as it was already the case at low H coverage on that support. Hence, the charge transfers take place only among the Pt atoms and hydrogen atoms in most of the isomers, and the Pt cluster becomes partially positively charged (up to 0.4 in the LFEE), except in one case where a H atom spills over to the support and transfers electronic charge to the cluster.

The charge transfers between the Pt8Hx clusters and alumina surfaces are also analyzed by the electronic density difference plots(Δρ), from the difference between the sum of isolated Pt8Hx and surface (ρPtH + ρS) and the whole supported system (ρPtH−S). The results are shown in Fig. 6. The interactions between the Pt8Hx cluster and the alumina surface induce significant electronic charge density accumulation between nearly all the neighboring Pt–Al pairs, while the change of the charge density between close Pt–O pairs is subtle, the Pt–O bond resulting instead in a polarization of the Pt and O atoms. For the GMs of Pt8H4@α-Al2O3(0001), Pt8H5@γ-Al2O3(100), and Pt8H24@α-Al2O3(0001), we also demonstrated two sliced charge difference plots for each structure. One of the two slices shows the electronic density differences between selected Pt–O pairs, and another one shows the electronic density difference between selected Pt–Al pairs. The results are shown in Figs. S10–S12. The sliced electronic density difference plots show that the electronic density between Pt–O is only weakly increased [Fig. S12(f)] or even decreased [Fig. S10(f)] between Pt–O atoms on the α-Al2O3(0001) surface, while the electronic charge density is moderately increased between the Pt–O atoms on the γ-Al2O3(100) surface [Fig. S11(f)] although the increase is less pronounced than that of Pt–Al bonds [Figs. S10(e), S11(e), and S12(e)]. The electron localization functions are also plotted alongside with the sliced charge differences, and they are shown in Figs. S10(b-c), S11(b-c), and S12(b-c). The results demonstrate the similar bond formation between Pt–Al pairs.

FIG. 6.

The isosurface (δ = 0.005) plot for the charge difference analysis of the four GM structures. The charge difference is defined as ρPtH + ρSρPtH−S, where ρPtH, ρS, and ρPtH−S are charge densities of separated cluster, separated support, and all system, respectively. The regions with blue (yellow) color indicate that the electron density is increased (decreased) in this region. (a) Pt8H24@α-Al2O3(0001), (b) Pt8H4@α-Al2O3(0001), (c) Pt8H24@γ-Al2O3(100), and (d) Pt8H5@γ-Al2O3(100).

FIG. 6.

The isosurface (δ = 0.005) plot for the charge difference analysis of the four GM structures. The charge difference is defined as ρPtH + ρSρPtH−S, where ρPtH, ρS, and ρPtH−S are charge densities of separated cluster, separated support, and all system, respectively. The regions with blue (yellow) color indicate that the electron density is increased (decreased) in this region. (a) Pt8H24@α-Al2O3(0001), (b) Pt8H4@α-Al2O3(0001), (c) Pt8H24@γ-Al2O3(100), and (d) Pt8H5@γ-Al2O3(100).

Close modal

To provide additional insight on the interactions between the Pt8Hx clusters and the two alumina surfaces, the Crystal Orbital Hamilton Population (COHP) analysis is carried out with the LOBSTER61–63 package. The COHP analysis is carried out for pairs of Pt–Al or Pt–O atoms when their distances are smaller than 3.5 Å. Afterward, the COHP is plotted by accumulating all the pairs with the same elements. The projected density of states (PDOS) plots are also shown alongside with the COHP to demonstrate the contributions of different elements. The results are shown in Fig. 7 for clusters on α-Al2O3(0001) and Fig. 8 for clusters on γ-Al2O3(100).

FIG. 7.

COHP analysis of two GMs on the α-Al2O3(0001) surface. Figure 7(a) shows the results for GM Pt8H4@α-Al2O3(0001), and Fig. 7(b) shows the results for GM Pt8H24@α-Al2O3(0001). (a) Pt8H4@α-Al2O3(0001) and (b) Pt8H24@α-Al2O3(0001).

FIG. 7.

COHP analysis of two GMs on the α-Al2O3(0001) surface. Figure 7(a) shows the results for GM Pt8H4@α-Al2O3(0001), and Fig. 7(b) shows the results for GM Pt8H24@α-Al2O3(0001). (a) Pt8H4@α-Al2O3(0001) and (b) Pt8H24@α-Al2O3(0001).

Close modal
FIG. 8.

COHP analysis for the Al–Pt and O–Pt interactions of two GMs on the α-Al2O3(0001) surface. Figure 8(a) shows the results for GM Pt8H5@γ-Al2O3(100), and Fig. 8(b) shows the results for GM Pt8H24@γ-Al2O3(100). (a) Pt8H5@γ-Al2O3(100) and (b) Pt8H24@γ-Al2O3(100).

FIG. 8.

COHP analysis for the Al–Pt and O–Pt interactions of two GMs on the α-Al2O3(0001) surface. Figure 8(a) shows the results for GM Pt8H5@γ-Al2O3(100), and Fig. 8(b) shows the results for GM Pt8H24@γ-Al2O3(100). (a) Pt8H5@γ-Al2O3(100) and (b) Pt8H24@γ-Al2O3(100).

Close modal

The COHP plots of Pt8H4@α-Al2O3(0001) [Fig. 7(a)] show a bonding interaction between Pt and O in the first part of the occupied band (−9 to −4 eV), but the interaction then becomes antibonding in the rest of the occupied band until the Fermi level. Such a case, characteristic of an interaction between most occupied states,64 results in a decreased overall Pt–O overlap population and hence weakened covalent component of the bond. Note that the interaction is still favorable since most of the antibonding states are pushed above the Fermi level. In contrast, the Al–Pt interaction remains bonding on the whole range of the occupied band, in a mixing between majority Pt and minority Al orbitals. Al orbitals are indeed mainly located in the vacant part of the band (Fig. S9). The conclusion from COHP is consistent with the increased electronic charge density between Pt and Al shown in Fig. 6(b). The COHP plots for the GM of the Pt8H24@α-Al2O3(0001) surface are shown in Fig. 7(b). Note that the high H coverage opens a gap at the Fermi level. The overall conclusion is similar to that of Pt8H4@α-Al2O3(0001).

The COHP and PDOS of the GM Pt8H5 cluster on the γ-Al2O3(100) surface are shown in Fig. 8(a). Again, a bonding interaction is seen between the neighboring Pt–Al pairs on the whole occupied energy range, while both bonding and antibonding energy ranges exist for Pt–O pairs. For Pt8H24@γ-Al2O3(100), the population of Al orbitals in bonding interactions is minimal, which results from the rather small adhesion of the cluster on the γ-Al2O3(100) surface. The − COHP of each case is integrated for all the occupied orbitals, including both bonding and antibonding populations, and shown in Fig. S13. The integrated numbers provide similar insight that Pt–Al contributes to the surface-cluster interaction alongside with Pt–O.

On both surfaces, it is found that Pt8 clusters with small hydrogen coverage retain the metallic nature with reasonable density of states around Ef; this is an indication of their potential good catalytic activity. Therefore, the small Pt clusters can be used for reactions at high temperature, such as propane dehydrogenation.65 However, the large hydrogen coverage (θ ∼ 3.0) under high μH results in a closed-shell electronic structure with a rather large band gap (∼1.5 eV) on both surfaces [shown in Figs. S9(b) and S9(d)]. This implies that the considered GM Pt8H24 cluster could be potentially not active in such a hydrogen environment. The potential activity can be achieved by dynamically detaching hydrogen atoms to create metastable structures,34 which will be addressed in our further studies.

In this manuscript, we exploited the density functional theory based grand canonical genetic algorithm to explore the ensemble of low free energy structures for hydrogenated Pt8 clusters on two surfaces, α-Al2O3(0001) and γ-Al2O3(100), and in two different hydrogen chemical potentials, i.e., low μH (T = 600 °C, pH2 = 0.1 bar) and high μH (T = 25 °C, pH2 = 1.0 bar). We find that the different supports do not significantly change optimal hydrogen coverage on the cluster but induce different morphological transitions. On the α-Al2O3(0001) surface, the Pt8Hx cluster switches from a pseudobilayer shape at low μH (low H coverage) to a completely flat and open structure at high μH (high H coverage). The strong interaction with the support is kept even at high H coverage. On the γ-Al2O3(100) surface, in contrast, the Pt8Hx cluster shows a compact one-layer shape under low μH and converts to a complete globular 3D structure, in weak interaction with the support, at high μH and high H coverage. These morphological transitions are not only valid for GMs but preserved for all the low free energy clusters of the ensemble, except for the case of Pt8Hx@α-Al2O3(0001) at the low μH condition, where Pt8Hx@α-Al2O3(0001) shows diverse morphologies. Although Pt8Hx undergoes different reconstructions on each support, it demonstrates on both of them a Pt–Pt bond contraction when μH is switched from high to low values, corresponding to an increased temperature. This observation renders a nonunique and surface dependent explanation to experimentally observed Pt bond length contraction under H2 when temperature is increased. Electronic analysis explains the origin of the distinct morphological change. The α-Al2O3(0001) support develops a stronger interaction with the Pt cluster than γ-Al2O3(100), as seen from the amplitude of charge transfer, of oxide surface relaxation at the interface, and from the interfacial bond-length and overlap populations. The analysis also evidences Metal-Lewis acidic (M-LA) bond formation between Pt and Al cations, in addition to well characterized Pt–O bonds, which influences the electronic structure of small Pt clusters on the alumina surface. This study hence provides detailed insights on the origin and strength of alumina support-Pt cluster interactions and the consequences on the structure and coverage of small Pt clusters under a pressure of hydrogen. It can also be helpful to rationalize experiments which characterize the valence orbitals in alumina supported Pt clusters, such as X-ray absorption near the edge structure.

See the supplementary material for additional data: a blueprint of the GCGA progress, slab models of α-Al2O3(0001) and γ-Al2O3(100), structures of low free energy isomers on α-Al2O3(0001) and γ-Al2O3(100) under low μH and high μH, bond length analysis of GM structures, energy decomposition analysis of low free energy isomers, projected density of states, Bader charges, and electron localization function of four GMs.

This work was funded by DOE-BES, Grant No. DE-SC0019152. This work used computational and storage services associated with the Hoffman2 Shared Cluster provided by the UCLA Institute for Digital Research and Education’s Research Technology Group. This research used resources of the National Energy Research Scientific Computing Center (NERSC), a U.S. Department of Energy Office of Science User Facility, operated under Contract No. DE-AC02-05CH11231. An award of computer time was provided by the Innovative and Novel Computational Impact on Theory and Experiment (INCITE) program. This research used resources of the Argonne Leadership Computing Facility, which is a DOE Office of Science User Facility supported under Contract No. DE-AC02-06CH11357.

1.
S.
Tosoni
and
G.
Pacchioni
, “
Oxide-supported gold clusters and nanoparticles in catalysis: A computational chemistry perspective
,”
ChemCatChem
11
,
73
89
(
2019
).
2.
R.
Cortese
,
R.
Schimmenti
,
A.
Prestianni
, and
D.
Duca
, “
DFT calculations on subnanometric metal catalysts: A short review on new supported materials
,”
Theor. Chem. Acc.
137
,
59
(
2018
).
3.
C. R.
Henry
, “
Surface studies of supported model catalysts
,”
Surf. Sci. Rep.
31
,
231
325
(
1998
).
4.
A. Y.
Stakheev
and
L.
Kustov
, “
Effects of the support on the morphology and electronic properties of supported metal clusters: Modern concepts and progress in 1990s
,”
Appl. Catal., A
188
,
3
35
(
1999
).
5.
D.
Ma
,
Y.
Shu
,
M.
Cheng
,
Y.
Xu
, and
X.
Bao
, “
On the induction period of methane aromatization over Mo-based catalysts
,”
J. Catal.
194
,
105
114
(
2000
).
6.
H.
Jiang
,
L.
Wang
,
W.
Cui
, and
Y.
Xu
, “
Study on the induction period of methane aromatization over Mo/HZSM-5: Partial reduction of Mo species and formation of carbonaceous deposit
,”
Catal. Lett.
57
,
95
102
(
1999
).
7.
M.
Ahmadi
,
H.
Mistry
, and
B.
Roldan Cuenya
, “
Tailoring the catalytic properties of metal nanoparticles via support interactions
,”
J. Phys. Chem. Lett.
7
,
3519
3533
(
2016
).
8.
C.
Mager-Maury
,
C.
Chizallet
,
P.
Sautet
, and
P.
Raybaud
, “
Platinum nanoclusters stabilized on γ-alumina by chlorine used as a capping surface ligand: A density functional theory study
,”
ACS Catal.
2
,
1346
1357
(
2012
).
9.
P.
Raybaud
,
C.
Chizallet
,
C.
Mager-Maury
,
M.
Digne
,
H.
Toulhoat
, and
P.
Sautet
, “
From γ-alumina to supported platinum nanoclusters in reforming conditions: 10years of DFT modeling and beyond
,”
J. Catal.
308
,
328
340
(
2013
).
10.
H.
Tang
,
Y.
Su
,
B.
Zhang
,
A. F.
Lee
,
M. A.
Isaacs
,
K.
Wilson
,
L.
Li
,
Y.
Ren
,
J.
Huang
,
M.
Haruta
 et al, “
Classical strong metal–support interactions between gold nanoparticles and titanium dioxide
,”
Sci. Adv.
3
,
e1700231
(
2017
).
11.
M.
Spencer
, “
Models of strong metal-support interaction (SMSI) in Pt on TiO2 catalysts
,”
J. Catal.
93
,
216
223
(
1985
).
12.
S.
Tauster
,
S.
Fung
, and
R. L.
Garten
, “
Strong metal-support interactions. Group 8 noble metals supported on titanium dioxide
,”
J. Am. Chem. Soc.
100
,
170
175
(
1978
).
13.
S.
Tauster
and
S.
Fung
, “
Strong metal-support interactions: Occurrence among the binary oxides of groups IIA–VB
,”
J. Catal.
55
,
29
35
(
1978
).
14.
H.
Pines
and
W. O.
Haag
, “
Alumina: Catalyst and support. I. Alumina, its intrinsic acidity and catalytic activity1
,”
J. Am. Chem. Soc.
82
,
2471
2483
(
1960
).
15.
M.
El Doukkali
,
A.
Iriondo
,
J.
Cambra
,
I.
Gandarias
,
L.
Jalowiecki-Duhamel
,
F.
Dumeignil
, and
P.
Arias
, “
Deactivation study of the Pt and/or Ni-based γ-Al2O3 catalysts used in the aqueous phase reforming of glycerol for H2 production
,”
Appl. Catal., A
472
,
80
91
(
2014
).
16.
M.
Trueba
and
S. P.
Trasatti
, “
γ-Alumina as a support for catalysts: A review of fundamental aspects
,”
Eur. J. Inorg. Chem.
2005
,
3393
3403
.
17.
C.
Chizallet
and
P.
Raybaud
, “
Density functional theory simulations of complex catalytic materials in reactive environments: Beyond the ideal surface at low coverage
,”
Catal. Sci. Technol.
4
,
2797
2813
(
2014
).
18.
A.
Suzuki
,
Y.
Inada
,
A.
Yamaguchi
,
T.
Chihara
,
M.
Yuasa
,
M.
Nomura
, and
Y.
Iwasawa
, “
Time scale and elementary steps of CO-induced disintegration of surface rhodium clusters
,”
Angew. Chem.
42
,
4795
4799
(
2003
).
19.
R.
Ouyang
,
J.-X.
Liu
, and
W.-X.
Li
, “
Atomistic theory of Ostwald ripening and disintegration of supported metal particles under reaction conditions
,”
J. Am. Chem. Soc.
135
,
1760
1771
(
2013
).
20.
J.
Rousset
,
L.
Stievano
,
F. C. S.
Aires
,
C.
Geantet
,
A.
Renouprez
, and
M.
Pellarin
, “
Hydrogenation of toluene over γ-Al2O3-supported Pt, Pd, and Pd–Pt model catalysts obtained by laser vaporization of bulk metals
,”
J. Catal.
197
,
335
343
(
2001
).
21.
J.
Rousset
,
L.
Stievano
,
F. C. S.
Aires
,
C.
Geantet
,
A.
Renouprez
, and
M.
Pellarin
, “
Hydrogenation of tetralin in the presence of sulfur over γ-Al2O3-supported Pt, Pd, and Pd–Pt model catalysts
,”
J. Catal.
202
,
163
168
(
2001
).
22.
S. D.
Lin
and
M. A.
Vannice
, “
Hydrogenation of aromatic hydrocarbons over supported pt catalysts. I. Benzene hydrogenation
,”
J. Catal.
143
,
539
553
(
1993
).
23.
L.
Castoldi
,
R.
Matarrese
,
M.
Daturi
,
J.
Llorca
, and
L.
Lietti
, “
Al2O3-supported Pt/Rh catalysts for NOx removal under lean conditions
,”
Appl. Catal., A
581
,
43
57
(
2019
).
24.
M.
Kahlich
,
H.
Gasteiger
, and
R.
Behm
, “
Kinetics of the selective CO oxidation in H2-rich gas on Pt/Al2O3
,”
J. Catal.
171
,
93
105
(
1997
).
25.
M.
Moses-DeBusk
,
M.
Yoon
,
L. F.
Allard
,
D. R.
Mullins
,
Z.
Wu
,
X.
Yang
,
G.
Veith
,
G. M.
Stocks
, and
C. K.
Narula
, “
CO oxidation on supported single Pt atoms: Experimental and ab initio density functional studies of CO interaction with Pt atom on θ-Al2O3 (010) surface
,”
J. Am. Chem. Soc.
135
,
12634
12645
(
2013
).
26.
O. A.
Bariås
,
A.
Holmen
, and
E. A.
Blekkan
, “
Propane dehydrogenation over supported Pt and Pt–Sn catalysts: Catalyst preparation, characterization, and activity measurements
,”
J. Catal.
158
,
1
12
(
1996
).
27.
J.
Shabaker
,
R.
Davda
,
G.
Huber
,
R.
Cortright
, and
J.
Dumesic
, “
Aqueous-phase reforming of methanol and ethylene glycol over alumina-supported platinum catalysts
,”
J. Catal.
215
,
344
352
(
2003
).
28.
M.
García-Diéguez
,
I.
Pieta
,
M.
Herrera
,
M.
Larrubia
, and
L.
Alemany
, “
Nanostructured Pt-and Ni-based catalysts for CO2-reforming of methane
,”
J. Catal.
270
,
136
145
(
2010
).
29.
S.
He
,
C.
Sun
,
Z.
Bai
,
X.
Dai
, and
B.
Wang
, “
Dehydrogenation of long chain paraffins over supported Pt-Sn-K/Al2O3 catalysts: A study of the alumina support effect
,”
Appl. Catal., A
356
,
88
98
(
2009
).
30.
B.
Schimmoeller
,
F.
Hoxha
,
T.
Mallat
,
F.
Krumeich
,
S. E.
Pratsinis
, and
A.
Baiker
, “
Fine tuning the surface acid/base properties of single step flame-made Pt/alumina
,”
Appl. Catal., A
374
,
48
57
(
2010
).
31.
C.
Mager-Maury
,
G.
Bonnard
,
C.
Chizallet
,
P.
Sautet
, and
P.
Raybaud
, “
H2-induced reconstruction of supported pt clusters: Metal–support interaction versus surface hydride
,”
ChemCatChem
3
,
200
207
(
2011
).
32.
C. H.
Hu
,
C.
Chizallet
,
C.
Mager-Maury
,
M.
Corral-Valero
,
P.
Sautet
,
H.
Toulhoat
, and
P.
Raybaud
, “
Modulation of catalyst particle structure upon support hydroxylation: Ab initio insights into Pd13 and Pt13/γ-Al2O3
,”
J. Catal.
274
,
99
110
(
2010
).
33.
G. S.
Erfani
,
S.
Hong
, and
T. S.
Rahman
, “
Effects of γ-Al2O3 support on the morphology and electronic structure of Pt nanoparticles
,”
J. Phys. Chem. C
123
,
16893
(
2019
).
34.
G.
Sun
and
P.
Sautet
, “
Metastable structures in cluster catalysis from first-principles: Structural ensemble in reaction conditions and metastability triggered reactivity
,”
J. Am. Chem. Soc.
140
,
2812
2820
(
2018
).
35.
E. T.
Baxter
,
M.-A.
Ha
,
A. C.
Cass
,
A. N.
Alexandrova
, and
S. L.
Anderson
, “
Ethylene dehydrogenation on Pt4,7,8 clusters on Al2O3: Strong cluster size dependence linked to preferred catalyst morphologies
,”
ACS Catal.
7
,
3322
3335
(
2017
).
36.
A.
Gorczyca
,
V.
Moizan
,
C.
Chizallet
,
O.
Proux
,
W.
Del Net
,
E.
Lahera
,
J.-L.
Hazemann
,
P.
Raybaud
, and
Y.
Joly
, “
Monitoring morphology and hydrogen coverage of nanometric Pt/γ-Al2O3 particles by in situ HERFD–XANES and quantum simulations
,”
Angew. Chem.
126
,
12634
12637
(
2014
).
37.
H.
Mistry
,
F.
Behafarid
,
S. R.
Bare
, and
B.
Roldan Cuenya
, “
Pressure-dependent effect of hydrogen adsorption on structural and electronic properties of Pt/γ-Al2O3 nanoparticles
,”
ChemCatChem
6
,
348
352
(
2014
).
38.
L.
Barrio
,
P.
Liu
,
J.
Rodriguez
,
J.
Campos-Martin
, and
J.
Fierro
, “
A density functional theory study of the dissociation of H2 on gold clusters: Importance of fluxionality and ensemble effects
,”
J. Chem. Phys.
125
,
164715
(
2006
).
39.
K. C. C.
Kharas
and
L. F.
Dahl
, “
Ligand-stabilized metal clusters: Structure, bonding, fluxionality, and the metallic state
,” in
Advances in Chemical Physics
(
John Wiley & Sons, Inc.
,
1988
), pp.
1
43
.
40.
M.
Gao
,
A.
Lyalin
,
M.
Takagi
,
S.
Maeda
, and
T.
Taketsugu
, “
Reactivity of gold clusters in the regime of structural fluxionality
,”
J. Phys. Chem. C
119
,
11120
11130
(
2015
).
41.
H.
Zhai
and
A. N.
Alexandrova
, “
Fluxionality of catalytic clusters: When it matters and how to address it
,”
ACS Catal.
7
,
1905
1911
(
2017
).
42.
H.
Zhai
and
A. N.
Alexandrova
, “
Local fluxionality of surface-deposited cluster catalysts: The case of Pt7 on Al2O3
,”
J. Phys. Chem. Lett.
9
,
1696
1702
(
2018
).
43.
H.
Zhai
and
A. N.
Alexandrova
, “
Ensemble-average representation of Pt clusters in conditions of catalysis accessed through GPU accelerated deep neural network fitting global optimization
,”
J. Chem. Theory Comput.
12
,
6213
6226
(
2016
).
44.
B.
Zandkarimi
and
A. N.
Alexandrova
, “
Surface-supported cluster catalysis: Ensembles of metastable states run the show
,”
Wiley Interdiscip. Rev. Comput. Mol. Sci.
9
,
e1420
(
2019
).
45.
Z.
Zhang
,
E.
Jimenez-Izal
,
I.
Hermans
, and
A. N.
Alexandrova
, “
Dynamic phase diagram of catalytic surface of hexagonal boron nitride under conditions of oxidative dehydrogenation of propane
,”
J. Phys. Chem. Lett.
10
,
20
25
(
2018
).
46.
W. W.
Tipton
and
R. G.
Hennig
, “
A grand canonical genetic algorithm for the prediction of multi-component phase diagrams and testing of empirical potentials
,”
J. Phys.: Condens. Matter
25
,
495401
(
2013
).
47.
L. B.
Vilhelmsen
and
B.
Hammer
, “
A genetic algorithm for first principles global structure optimization of supported nano structures
,”
J. Chem. Phys.
141
,
044711
(
2014
).
48.
A. N.
Alexandrova
,
A. I.
Boldyrev
,
Y.-J.
Fu
,
X.
Yang
,
X.-B.
Wang
, and
L.-S.
Wang
, “
Structure of the NaxClx+1(x = 1–4) clusters via ab initio genetic algorithm and photoelectron spectroscopy
,”
J. Chem. Phys.
121
,
5709
5719
(
2004
).
49.
M.
Digne
,
P.
Sautet
,
P.
Raybaud
,
P.
Euzen
, and
H.
Toulhoat
, “
Use of DFT to achieve a rational understanding of acid–basic properties of γ-alumina surfaces
,”
J. Catal.
226
,
54
68
(
2004
).
50.
M.
Digne
,
P.
Sautet
,
P.
Raybaud
,
P.
Euzen
, and
H.
Toulhoat
, “
Hydroxyl groups on γ-alumina surfaces: A DFT study
,”
J. Catal.
211
,
1
5
(
2002
).
51.
J. P.
Perdew
,
K.
Burke
, and
M.
Ernzerhof
, “
Generalized gradient approximation made simple
,”
Phys. Rev. Lett.
77
,
3865
(
1996
).
52.
J.
Hutter
,
M.
Iannuzzi
,
F.
Schiffmann
, and
J.
VandeVondele
, “
cp2k: atomistic simulations of condensed matter systems
,”
Wiley Interdiscip. Rev.: Comput. Mol. Sci.
4
,
15
25
(
2014
).
53.
J.
VandeVondele
and
J.
Hutter
, “
Gaussian basis sets for accurate calculations on molecular systems in gas and condensed phases
,”
J. Chem. Phys.
127
,
114105
(
2007
).
54.
P. J.
Eng
,
T. P.
Trainor
,
G. E.
Brown
, Jr.
,
G. A.
Waychunas
,
M.
Newville
,
S. R.
Sutton
, and
M. L.
Rivers
, “
Structure of the hydrated α-Al2O3 (0001) surface
,”
Science
288
,
1029
1033
(
2000
).
55.
G.
Zhang
,
X.
Wang
,
Y.
Xiong
,
Y.
Shi
,
J.
Song
, and
D.
Luo
, “
Mechanism for adsorption, dissociation and diffusion of hydrogen in hydrogen permeation barrier of α-Al2O3: A density functional theory study
,”
Int. J. Hydrogen Energy
38
,
1157
1165
(
2013
).
56.
W.
Sinkler
,
S. I.
Sanchez
,
S. A.
Bradley
,
J.
Wen
,
B.
Mishra
,
S. D.
Kelly
, and
S. R.
Bare
, “
Aberration-corrected transmission electron microscopy and in situ XAFS structural characterization of Pt/γ-Al2O3 nanoparticles
,”
ChemCatChem
7
,
3779
3787
(
2015
).
57.
W.
Bronger
and
K.
Wrzesien
, “
The structure of Al3Pt5
,”
J. Alloys Compd.
244
,
194
196
(
1996
).
58.
M.
Devillard
,
R.
Declercq
,
E.
Nicolas
,
A. W.
Ehlers
,
J.
Backs
,
N.
Saffon-Merceron
,
G.
Bouhadir
,
J. C.
Slootweg
,
W.
Uhl
, and
D.
Bourissou
, “
A significant but constrained geometry Pt-Al interaction, fixation of CO2 and CS2, activation of H2 and PhCONH2
,”
J. Am. Chem. Soc.
138
,
4917
4926
(
2016
).
59.
J. H.
Kang
,
L. D.
Menard
,
R. G.
Nuzzo
, and
A. I.
Frenkel
, “
Unusual non-bulk properties in nanoscale materials: Thermal metal-metal bond contraction of γ-alumina-supported Pt catalysts
,”
J. Am. Chem. Soc.
128
,
12068
12069
(
2006
).
60.
H.-Y. T.
Chen
,
S.
Tosoni
, and
G.
Pacchioni
, “
Hydrogen adsorption, dissociation, and spillover on Ru10 clusters supported on anatase TiO2 and tetragonal ZrO2 (101) surfaces
,”
ACS Catal.
5
,
5486
5495
(
2015
).
61.
V. L.
Deringer
,
A. L.
Tchougréeff
, and
R.
Dronskowski
, “
Crystal orbital Hamilton population (COHP) analysis as projected from plane-wave basis sets
,”
J. Phys. Chem. A
115
,
5461
5466
(
2011
).
62.
S.
Maintz
,
V. L.
Deringer
,
A. L.
Tchougréeff
, and
R.
Dronskowski
, “
Analytic projection from plane-wave and PAW wavefunctions and application to chemical-bonding analysis in solids
,”
J. Comput. Chem.
34
,
2557
2567
(
2013
).
63.
S.
Maintz
,
V. L.
Deringer
,
A. L.
Tchougréeff
, and
R.
Dronskowski
, “
LOBSTER: A tool to extract chemical bonding from plane-wave based DFT
,”
J. Comput. Chem.
37
,
1030
1035
(
2016
).
64.
R.
Hoffmann
, “
A chemical and theoretical way to look at bonding on surfaces
,”
Rev. Mod. Phys.
60
,
601
(
1988
).
65.
J. J.
Sattler
,
J.
Ruiz-Martinez
,
E.
Santillan-Jimenez
, and
B. M.
Weckhuysen
, “
Catalytic dehydrogenation of light alkanes on metals and metal oxides
,”
Chem. Rev.
114
,
10613
10653
(
2014
).

Supplementary Material