Construction of Potential Energy Surface for Ne Atom and H ₂ ⁺ Ion Collision Reaction Based on Artificial Intelligence

In undergraduate chemistry education, the applications of machine learning methods remain limited, particularly in potential energy surface (PES) fitting for molecular dynamics simulations. To address this gap, this experiment focuses on the one-dimensional triatomic collision reaction Ne + H2+ → NeH+ + H, a classical reaction in physical chemistry. Starting with quantum chemistry calculations, we developed a user-friendly open-source package in Python, leveraging the machine learning framework PyTorch to fit the PES of this reaction. The trained model was then used to visualize molecular dynamics simulations. Through this experiment, students gain comprehensive experience in applying quantum chemistry calculations and machine learning techniques to solve a specific chemical problem while developing insights into PES fitting using neural networks.