Identifying magic-number structures of supported sub-nano Ni clusters and the influence of hydrogen coverage: a density functional theory based particle swarm optimization investigation

Abstract

Supported sub-nano Ni clusters are of great significance to many heterogeneous catalysis applications. We conducted density functional theory based particle swarm optimization calculations to study the low-energy structures of sub-nano Ni_n (n = 2–21) clusters in the gas phase and on two oxide surfaces, ZnO(000) and γ-Al₂O₃(100). Our results show that sub-nano Ni clusters have magic structures, and the magic numbers are 4, 8, 11, 13, 15, and 19 on ZnO and 3, 5, 8, 11, 16, and 21 on γ-Al₂O₃, respectively, essentially different from gas-phase Ni clusters (magic n = 2, 6, 10, 11, 13, 20). The morphological transformation of Ni clusters relies on the interplay between the Ni–oxide interaction and the Ni–Ni interaction inside the Ni clusters. The Ni–oxide interaction becomes weakened with the cluster size growing and the Ni–ZnO interaction is generally larger than the Ni–Al₂O₃ interaction, while the Ni–Ni interaction is independent of the type of oxide substrate and slightly grows with the cluster size increase. On ZnO, the Ni–Ni interaction exceeds the Ni–ZnO interaction beyond the Ni₁₀ cluster, which evolves into a double-layer structure from planar 2D configurations, while on γ-Al₂O₃(100), the morphological transition of the planar 2D structure towards a layered 3D structure occurs at the Ni₇ cluster. Hydrogen coverage on Ni clusters lead to the conversion of the cluster morphology from layered 3D geometries to more open one-layer 2D structures. Ab initio thermodynamics analysis on Ni₁₁H_x clusters revealed that under the typical hydrogenation conditions (T = 673 K; P_H2 = 10 atm), the most stable hydrogen-containing structures are the Ni₁₁H₁₀ cluster on ZnO and the Ni₁₁H₈ cluster on γ-Al₂O₃, which both expose more Ni sites compared with the bare Ni₁₁ cluster indicating an enhanced reactivity induced by a hydrogen environment.