Machine Learning and Molecular Simulation

Abstract

Understanding complex quantum systems with many interacting particles is a major physical challenge. Classical methods struggle due to the exponential increase in complexity as the number of particles increases. This study explores two promising approaches: present an overview of the state-of-the-art research related to molecular dynamics and machine learning. The first approach studies diabaticity using a neural network, which offers richer dynamic properties than atoms. By studying the hybrid calcium system, CaH2, the authors present PESs adiabatic and diabatic of the ground state and the first excited state. As scoop results presented for the first time, detailed analysis identified new approaches to molecular dynamics beyond the Born-Oppenheimer approximation. The second approach tackles the problem from a computational perspective. Machine learning algorithms, particularly interpretable methods such as neural networks (NN) with influence functions, have been explored.