Explicit, Machine-Learned Two-Body Potentials for Molecular Simulations.

A new pairwise hybrid machine-learning/molecular mechanics (ML/MM) potential is introduced that is conceived for application to large, heterogeneous condensed-phase systems. The PhysNet ML method describes monomers and short-range dimer interactions, whereas an empirical MM force field describes pairwise interactions beyond a system-dependent switching distance. Models are fitted to MP2 dimer and pairwise cluster energies, and the quality of each model is assessed at different switching distances and using MM approaches with and without advanced electrostatic interactions. For dichloromethane and acetone as test systems, energy-conserving molecular dynamics simulations are carried out. Validation of the hybrid ML/MM energy function is demonstrated for total cluster energies and for partial radial distribution functions of dichloromethane by comparing with experiments. Future extensions to including higher-order many-body contributions are discussed as well. The present work thus paves the way for modular, bottom-up, computationally efficient hybrid ML/MM potentials via addition of an explicit many-body correction.