The applicability of hybrid molecular dynamics-Monte Carlo (MDMC) methods to study nanoconfined fluids is limited by their lack of generality beyond the training conditions. Here, the authors present an AI-assisted MDMC framework that overcomes this limitation by learning local, conditional transition statistics directly from MD trajectories, demonstrating its accuracy and transferability across different thermodynamic conditions, spatial scales, and complex nano-scale geometries.
- Jie Liu
- Guodong Chen
- Shuyu Sun