Surface-Dependent Interfacial Concentration of Oxygen Confined within Pd Interlayers: Molecular Dynamics with a Neural Network Potential

Abstract

Obtaining a detailed understanding of the interfacial dynamics of oxygen on palladium surfaces is crucial for industrial applications. However, it remains challenging to develop reaction–transport coupling mechanisms to enhance the activity and stability of Pd-based catalysts in confined environments. Herein, by integrating the established global neural network (G-NN) potential and molecular dynamics (MD) simulations, the interfacial concentrations of confined O₂ molecules within Pd interlayers were investigated systematically under various conditions. The developed reactive NN potential, rigorously validated against DFT benchmarks with an average error of 0.026 eV/atom, demonstrated precise structural discrimination capabilities among three Pd surfaces and subsequently produced reasonable catalytic structures. The Pd(100) surface exhibited the highest reactivity, followed by Pd(211), with the lowest on Pd(111). These differences show strong correlations with a reduced interlayer distance (approximately 1 nm) and the degree of surface reconstruction patterns through a comprehensive analysis of mean square displacement and reaction rate. Density distribution, in conjunction with radial distribution function analyses, further demonstrates how the interlayer confinement effect, as well as surface-specific atomic arrangements, remarkably regulate the interfacial concentration of oxygen. This work provides universal guidance for elucidating the macroscopic mechanism linking the bulk and interfacial concentrations in confined systems through large-scale simulations.