(169bs) Elucidating the Fluxionality and Dynamics of Zeolite-Confined Au Nanoclusters Using Machine Learning Potentials

Sub-nanometric clusters (NCs) of transition metal (TM) atoms, typically consisting of fewer than 15 atoms, have exhibited remarkable catalytic activity in various industrial reactions. However, these NCs are thermodynamically unstable and susceptible to deactivation due to sintering. Previous experiments have proposed zeolites as effective structural scaffolds to stabilize small NCs. However, key physiochemical phenomena such as fluxionality and diffusion properties of such zeolite-confined TM NCs (TM@zeolites) are not well understood. The central challenge is the steep computational cost associated with performing sufficiently long ab initio molecular dynamics (AIMD) simulations. As a step towards addressing this challenge, we developed a self-adaptive workflow that leverages two state-of-the-art machine learning potential (MLP) packages (i.e., SchNetPack and neuroevolution potential) to develop an accurate, robust, and transferable MLP for a prototypical Au@zeolite system. The resulting MLP, which is shown to be transferable across several other zeolites and various temperatures, is used to determine the free energy landscape and the corresponding rates for the diffusion of gold NCs in LTA zeolites.

Sub-nanometric clusters (NCs) of transition metal (TM) atoms, typically consisting of fewer than 15 atoms, have exhibited remarkable catalytic activity in various industrial reactions. However, these NCs are thermodynamically unstable and susceptible to deactivation due to sintering. Previous experiments have proposed zeolites as effective structural scaffolds to stabilize small NCs. However, key physiochemical phenomena such as fluxionality and diffusion properties of such zeolite-confined TM NCs (TM@zeolites) are not well understood. The central challenge is the steep computational cost associated with performing sufficiently long ab initio molecular dynamics (AIMD) simulations. As a step towards addressing this challenge, we developed a self-adaptive workflow that leverages two state-of-the-art machine learning potential (MLP) packages (i.e., SchNetPack and neuroevolution potential) to develop an accurate, robust, and transferable MLP for a prototypical Au@zeolite system. The resulting MLP, which is shown to be transferable across several other zeolites and various temperatures, is used to determine the free energy landscape and the corresponding rates for the diffusion of gold NCs in LTA zeolites.