Effect of “Reducible” Titania Promotion on the Mechanism of H-Migration in Pd/SiO₂ Clusters

Abstract

Industrial catalysts, consisting of metal species dispersed on a porous surface, are typically prepared by hydrogen treatment at high temperatures, and the migration of hydrogen constitutes the key elementary reaction that mediates the metal-support interactions and the generation of active catalytic sites. A detailed modeling of these processes is challenging especially for systems involving disordered surfaces. A debated issue is the role of the support “reducibility” on the mechanism of H-spillover processes. Transition-metal oxides such as titania significantly improve the activity and selectivity of catalysts supported on “non-reducible” oxides, such as silica, particularly in hydrogenation and dehydrogenation processes. To gain insight into the mechanism of H-migration in disordered catalysts and nano-particles, a comprehensive DFT-analysis of the potential energy surfaces is carried out for the transfer of H-atoms from a palladium-tetramer catalyst to a silica-support, represented by a substituted siloxane ring, doped with “reducible” titania. The H-migration to non-stoichiometric supports occurs via both a direct and H ₂ -assisted transfer of H-atoms when the metal-catalyst is anchored to the non-bonding oxygen atoms of the support (defect sites). The calculations revealed that the transfer to different types of supports proceeds through identical pathways but significantly different barriers. The titania-promotion, as well as the addition of dihydrogen ligands to the catalyst, decreases the direct and H ₂ -assisted H-migration barriers. Due to the extended transition state structures, the H₂-assisted mechanism provides access to more distant and hindered active centers, and thus can account for the “bottleneck” particle-to-particle (along grain boundaries) transport of hydrogen atoms. The H-migration to titania-promoted stoichiometric (defect-free) supports generates intermetallic Pd-Ti bonds, in contrast to the pristine silica-based catalysts, which do not form analogous bonds between palladium and silicon centers of the support.

Graphical Abstract

Appendix

1.1 Appendix 1: Transfer of the Second H-atom in Primary Radicals

The removal of the first H-atom from the molecular intermediates II or III, representing the H-atom diffusion over the support, generates the primary radicals IV (Scheme 1). It is important to examine also the H-spillover of the remaining second H-atom of the Pd₄ cluster onto the support.

Table 6 presents two secondary H-spillover processes for 2H₂-capped radicals IV. The results indicate that the kinetic barriers are fairly high, compared to the first H-migration. Again, the promotion by titania decreases the barrier.

Table 6 Forward ($\varDelta E_{fwd}^{\# }$) and reverse ($\varDelta E_{rev}^{\# }$) barriers and reaction energies (Δ _R E) for the second H-transfer in 2H₂-capped open-shell (H₂)₂Pd₄H catalyst onto a non-stoichiometric support

1.2 Appendix 2: The Role of Spin-Multiplicity

The current calculations, in accord with literature data, show that the ground state spin multiplicity of the bare Pd₄ is a triplet, whereas the hydrogenated complex Pd₄2H is a singlet (cf. [58, 106, 127]). Therefore, a spin-crossing is expected to occur in the dissociative chemisorption of H₂ on the bare Pd₄, even though the product formation half-reaction, starting from TS in a singlet state, could well characterize the reaction probability. However, it is not expected any nonadiabatic transition between PESs corresponding to different spins for the activation and hydrogen migration in reactions over the palladium tetramer supported on pristine silica and doped by titania catalysts, which are all ground state singlets. The calculations confirmed that there is no intersystem crossing along the reaction pathways for the metabolism of hydrogen in supported systems. In addition, the triplet state reactions are more energy demanding than the singlet ones. Note also that no changes in the ground state spin multiplicity of Pd₄ and of some other small Pd_n clusters (n = 2–4, 7, 13) have been found as a result of molecular adsorption [126].

To evaluate the possible role of the high-spin states on H-migration to a defect-free surface, the singlet –state pathway involving the 2H₂-capped catalyst (Fig. 8) was recalculated for the triplet state PES. The singlet–triplet splitting (ΔE^ST = 12.38 kcal mol⁻¹) evaluated at B3LYP level is expected to be somewhat underestimated because of the over-stabilization of the high-spin state energies by this functional [82, 108]. Additional calculations based on the pure GGA-BP86 functional, indeed, provide the somewhat higher value of 15.81 kcal mol⁻¹. However, this insignificant difference cannot alter the general conclusions involving the high-spin states at the B3LYP level.

The triplet pathway is also not significant because of the presence of the barrier of 23.7 kcal mol⁻¹ height on the triplet route leading to the coordination of palladium cluster to a bridged O-atom. The triplet pathway could play a role only if the triplet state population is sufficiently high. However, lower by ca. 16 kcal mol⁻¹ the singlet state reagent generates H-migration, and produces an intermetallic Pd-Ti bond and a H₂O-moiety, at almost the same barrier of 23.5 kcal mol⁻¹ (Fig. 7).

1.3 Appendix 3: Hydrogenolysis and Activation of the Second H₂ Molecule

As noted in Table 4, in some titania-promoted systems, the nascent H₂ molecule (co-catalyst H^AH^B) undergoes an interface heterolytic cleavage (hydrogenolysis) instead of H ₂ -assisted transfer, or ligates molecularly to the metal. This can be explained by the lower coordination of Pd₄ (fewer bonds to other Pd and H-atoms, as well as dihydrogen ligands), examined in Appendix 1.

Table 4 involves two hydrogenolysis reactions of the low-coordinated catalysts containing none- or one-H₂ ligand. In both cases, the interfacial Pd–O bonds activate heterolytically the dihydrogen to form hydridic and protic H-atoms via almost equal barriers (ca. 15 kcal mol⁻¹), almost independent of the number of adsorbed H₂ ligands. Such a small effect of the H₂-ligation constitutes an interesting issue which provides evidence for the dominance of interfacial forces in these reactions, as it occurs during H₂ activation by the closed-shell AuCeO₂ ⁺ catalyst [155]. By analogy, the covalently bound Pd ion can be considered as a strong Lewis acid, with the O²⁻ ion viewed as a Lewis base. This provides a synergistic effect of the two separated sites with counter polarity, called intermolecular frustrated Lewis pairs [156].

Thus, the H₂-assisted reactions generally occur when a basic coordination-sphere of the metal cluster (Pd₄) is saturated by H-atoms and/or H₂-ligands. The coordination deficiency concerns not only the defective systems described in Table 4 (palladium cluster residing on two NBOs), but also the stoichiometric-supported catalysts, where palladium is linked to the valence-saturated oxygen atoms of the support (OH-groups with H-cover atoms noted in parentheses in Figs. 1, 2). The stoichiometric-supported catalysts favor the SMSI interactions, as noted in Sect. 4.2.

It should be emphasized that all hydrogenolysis reactions of titania-promoted systems are thermodynamically unfavorable (all are endothermic), in contrast to the alternative exothermic DHC-reactions (cf. Table 4). Importantly, all extended models included in Table 4 result in H-migration reactions via DHC pathways. The hydrogenolysis does not occur even if the absorbed and adsorbed hydrogen species are completely absent. Perhaps, this is due to the saturation of Pd₄ through strong bonding with the third oxygen atom as another anchorage point of the siloxane ring (tripodally anchored to silica via covalent bonds with oxygen atoms).

For comparison, the barriers for the interfacial hydrogenolysis (heterolytic splitting) of the first dihydrogen molecule along the Pd–O covalent bond of the silica-supported catalyst (initially being NBO), was found to be 12.83 kcal mol⁻¹, which is close to the hydrogenolysis barrier for the second H₂ molecule mentioned above (ca. 15 kcal mol⁻¹, Table 4).

The hydrogenolysis on supported systems can also be compared with the second H₂-activation results on bare metallic palladium clusters present in the literature. As shown by Morokuma and co-workers [106], the activation of the second H₂-molecule by Pd₄ cluster is thermodynamically and kinetically unfavorable because of the instability of the products [106]. However, the process can well occur when a support is involved, as shown in this study (vide supra). Notably, Wang et al. found that the second molecular hydrogen can be fragmented over the Pd₆ cluster of distorted O_h-symmetry through an overall barrier of ca. 9 kcal mol⁻¹ height [107]. This barrier has a value in the range of those found in this paper for supported systems.

1.4 Appendix 4: Atomic and Molecular Hydrogen Coverages and H₂-Intercalation

The saturation of a support and a catalyst can alter the H-migration barriers, as particularly discussed by Singh et al. [41] while considering the H-spillover for Pd₄ catalyst on graphene. However, even at the saturation limit of the Pd₄ catalyst, only one out of the 9H₂ molecules has been found to be dissociated to form two hydridic atoms, which rationalizes the model employed in the current paper. Whereas, the GGA-PBE barrier for H-migration has been found to be too high for pristine graphene, the H-saturation of graphene reduced the barrier to ca. 16 kcal mol⁻¹, accessible even at ambient temperatures [41].

In order to get an idea on the H-atom accommodation features and the coverage limits on the silica support, a fully saturated symmetric complex was designed containing 18H atoms. The three axial positions were connected to the three off-ring surface oxygen atoms in a “tripod” configuration. The remaining 15 hydridic atoms were placed at all possible ligand positions of the four octahedral Pd-centers. A relaxed full optimization of this structure led to the spontaneous recombination of the eight hydridic atoms to form four Kubas type dihydrogen ligands bound to the basic Pd-centers. Only two H₂ molecules remain activated at the proximity of the reactive (basic) palladium centers at bridging positions, in addition to the three isolated H-atoms located at the top(non-reactive) sites of Pd₄-cluster. Since the bare Pd₄H₁₈ complex accommodates only two hydridic H-atoms and 8H₂ molecules (see also [41]), this result suggests a specific role of the oxidic support in activating the nascent H₂-molecules. Furthermore, the defective sites introduce additional activation features. For illustration, the two extra hydridic atoms were added to the TS structure of the direct H-migration that initially contained only two hydridic atoms (with a barrier of 24.49 kcal mol⁻¹ shown in Table 1). The full TS-search (gradient-norm optimization) led to the restoration of the H₂-ligand in a product-like structure. Thus, the interplay between the different types of support centers can vary the oxidation states of the Pd-atoms, and thus control the amount of the retained H₂-ligands. This seems to be in contrast with platinum which is prone to dissociate most of the nascent dihydrogen molecules [157]. A detailed study of this problem is beyond the scope of current paper, and constitutes an interesting open issue.

Another specific role of the dihydrogen is its “intercalation” between the catalyst and the support at the interface regions. Figure 9 compares a full-contact “tripod” structure of Pd₄ located over the hexahydroxytrioxatrisilinane ring, with an actual reagent of the H-spillover in 2H₂-capped systems involving a double-contact “dipod” structure where the reactive Pd-atom is moved away (separated) from the surface by an “intercalated” dihydrogen ligand.

Fig. 9

Two alternative reagents involving the hydrogenated catalyst Pd₄(H)₂(H₂)₂, covalently bound to two NBO defect-centers of the silica support indicated by small arrows. a “Dipod” configuration involves a Pd₁ vertex separated from the support by H₂-ligand. b “Tripod” configuration involves full palladium tetrahedron contact with three support oxygen atoms

The somewhat higher energy of the “dipod” structures provides lower H-spillover barriers for all considered systems: 6.45, 4.78, and 0.84 kcal mol⁻¹, respectively, for pure silica, the binary oxide, and replaced by titania supports (Table 1). The spillover barrier onto the replaced by titania support is exceedingly low due to the interface location (“intercalation”) of the H₂-ligand. These results strongly suggest a specific role of the dihydrogen as a ligand to a Pd nanoparticle at the interface regions. This constitutes an important new aspect in H-spillover phenomenon, albeit a partial contact of the catalyst nanoparticle with a support is known in literature. For instance, Pt₄-cluster is known to be residing on a carbon nanotube only through one vertex, independent of hydrogen saturation [41, 45].