Ultrasmall Au clusters supported on pristine and defected CeO₂: Structure and stability

The atomistic simulation of supported metal catalysts has long been challenging due to the increased complexity of dual components. In order to determine the metal/support interface, efficient theoretical tools to map out the potential energy surface (PES) are generally required. This work represents the first attempt to apply the recently developed SSW-NN method, stochastic surface walking (SSW) global optimization based on global neural network potential (G-NN), to explore the PES of a highly controversial supported metal catalyst, Au/CeO₂, system. By establishing the ternary Au-Ce-O G-NN potential based on first principles global dataset, we have searched for the global minima for a series of Au/CeO₂ systems. The segregation and diffusion pathway for Au clusters on CeO₂(111) are then explored by using enhanced molecular dynamics. Our results show that the ultrasmall cationic Au clusters, e.g., Au₄O₂, attaching to surface structural defects are the only stable structural pattern and the other clusters on different CeO₂ surfaces all have a strong energy preference to grow into a bulky Au metal. Despite the thermodynamics tendency of sintering, Au clusters on CeO₂ have a high kinetics barrier (>1.4 eV) in segregation and diffusion. The high thermodynamics stability of ultrasmall cationic Au clusters and the high kinetics stability for Au clusters on CeO₂ are thus the origin for the high activity of Au/CeO₂ catalysts in a range of low temperature catalytic reactions. We demonstrate that the global PES exploration is critical for understanding the morphology and kinetics of metal clusters on oxide support, which now can be realized via the SSW-NN method.