Efficient and Accurate Molecular Dynamics: The Self-Learning Hybrid Monte Carlo Method

Molecular dynamics simulations demand precise interatomic force calculations, often using density functional theory (DFT) in first-principles approaches. While DFT provides accuracy, its high computational cost limits its use in large or long-term simulations. Machine learning molecular dynamics with artificial neural networks (ANNs) offers a solution but struggles with stability and accuracy in complex systems. We introduce the self-learning hybrid Monte Carlo (SLHMC) method, which blends machine learning with rigorous statistical methods to deliver DFT-level precision while dramatically reducing computational demands. SLHMC enables efficient, accurate simulations and has broad applications. In this seminar, we will discuss SLHMC's potential applications and its contributions to advancing simulation capabilities.