Search papers, labs, and topics across Lattice.
This paper introduces a hybrid machine-learning/molecular mechanics (ML/MM) potential, where PhysNet handles monomer and short-range dimer interactions, and a classical MM force field covers longer-range pairwise interactions. The models are trained on MP2 dimer and pairwise cluster energies, with performance evaluated across different switching distances and MM approaches. The approach is demonstrated on dichloromethane and acetone, revealing limitations of purely pairwise potentials in systems with significant many-body effects.
A hybrid ML/MM potential that combines PhysNet with classical force fields achieves promising accuracy for molecular simulations, but also exposes the need for many-body corrections in condensed-phase systems.
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, while a classical MM force field describes pairwise interactions beyond a defined 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 detailed distributed charge electrostatics. The applicability of the approach to molecular dynamics simulations is demonstrated for a basic implementation applied to a small model system. Dichloromethane and acetone are used as test systems to demonstrate the accuracy of the approach in describing pairwise reference data, and also to highlight the limitations of the pairwise approach for systems that exhibit significant many-body effects in condensed phase, paving the way for the addition of a general many-body correction in future work.
