Water Models*

Quantum nuclear effects can play a subtle but significant role in properties of water. However accurate fully quantum simulations of dynamics in condensed phase media is computationally intractable. Among the various classical trajectory based approaches, Wigner dynamics is lucrative because it can be derived as a linearization of the path integral expressions. The primary challenge is to estimate the Wigner density which involves a multidimensional Fourier transform. Various approximations to the Wigner density are possible. Irrespective of the approximation used, the cost of these simulation for ab initio density functional theory (DFT) used to make it untenable. Recent developments in the framework of Deep Molecular Dynamics (DeepMD) allows the usage of machine learning models fit to high quality DFT potential energy surfaces. These models have been used to study the classical dynamics of water at unprecedented levels of accuracy. We evaluate and compare the impact of quantum nuclear effects on the dynamical properties of these water models using different approximations to Wigner dynamics.