O 2 -Mediated Evolution of Pt Nanoclusters in Silicalite-1 Using Artificial Neural Network Potential-Based Molecular Dynamics Simulations

The aggregation and structural evolution of platinum (Pt) nanoclusters within zeolites are strongly influenced by the gaseous reaction atmosphere, which critically affects their catalytic performance and stability. In this study, we develop a global neural network potential to simulate the dynamics of Pt and its oxides (PtO x ) confined in silicalite-1, enabling efficient large-scale molecular dynamics simulations. The G-NN potential is validated against density functional theory calculations including adsorption energies and dissociation barriers. Using this framework, we systematically investigated the effects of Pt oxidation on cluster migration. With increasing oxidation, the interaction between Pt and the silicalite-1 framework weakens and migration energy barriers decrease, enhancing the mobility of oxidized species. Elevated temperatures further promote oxygen adsorption and dissociation on Pt surfaces, driving progressive oxidation and fragmentation into chain-like Pt–O structures. Radial distribution function analyses show a decrease in Pt–Pt coordination and a simultaneous increase in Pt–O coordination, consistent with experimental observations of PtO and PtO 2 phases. These results provide atomic-level insight into how oxygen and temperature govern Pt cluster evolution within confined zeolite frameworks, offering a theoretical basis for designing highly active, sinter-resistant catalysts in the field of confined metal cluster zeolite systems.