Solvating machine-learning interatomic potentials for solution-phase chemistry
AnySolv is a transferable implicit solvation potential that corrects existing gas-phase potentials using machine learning, requiring no retraining of the base potential. It learns solvent-dependent energy and force corrections by predicting the difference between solvated and gas-phase DFT energies. This allows for scalable simulations of solvent effects in chemical reactions and molecular dynamics.