Water: one molecule, two surfaces, one mistake

TL;DR

This paper argues that using separate, optimized fits for the potential energy surface and dipole moment surface in water modeling improves accuracy, challenging the traditional practice of using the same charges for both.

Contribution

It introduces the idea of independently fitting PES and DMS in water models, which is common in quantum chemistry but underused in simulation communities.

Findings

Separate fits for PES and DMS improve water model accuracy

Using adjustable parameters for DMS enhances electric field simulations

Challenges the traditional single-charge approach in empirical water models

Abstract

In order to rigorously evaluate the energy and dipole moment of a certain configuration of molecules one needs to solve the Schr\"odinger equation. Repeating this for many different configurations allows one to determine the potential energy surface (PES) and the dipole moment surface (DMS). Since the early days of computer simulation it has been implicitly accepted that for empirical potentials the charges used to fit the PES should also be used to describe the DMS. This is a mistake. Partial charges are not observable magnitudes. They should be regarded as adjustable fitting parameters. Optimal values used to describe the PES are not necessarily the best to describe the DMS. One could use two fits: one for the PES, and another for the DMS. This is a common practice in the quantum chemistry community, but not used so often by the community performing computer simulations. This idea affects all types of modelling of water (with the exception of ab-initio calculations) from coarse grained to non-polarizable and polarizable models. We anticipate that an area that will benefit dramatically from having both, a good PES and a good DMS, is the modelling of water in the presence of electric fields.