Incorporating Polarization and Charge Transfer into a Point Charge Model for Water using Machine Learning

Rigid non-polarizable water models with fixed point charges have been widely employed in molecular dynamics (MD) simulations due to their efficiency and reasonable accuracy for the potential energy surface (PES). However, the dipole moment surface (DMS) of water is not necessarily well described by the same fixed charges, leading to their failure in reproducing dipole-related properties. Here, we developed a machine-learning (ML) model trained against electronic structure data to assign point charges for water and the resulting DMS significantly improved the predictions of the dielectric constant and the low-frequency IR spectrum of liquid water. Our analysis reveals that the degree of polarization and charge transfer (CT) can be tuned by the percentage of the exact Hartree-Fock (HF) exchange in the hybrid functional, with polarization improving the dielectric constant and CT being primarily responsible for the hydrogen-bond stretch peak at 200 cm-1 in the IR spectrum.