Systematic optimization of atomistic water models for molecular simulation using liquid/vapor surface tension data

The heat of vaporization is one of the most important experimental properties that is used to train the parameters of empirical potentials (force fields) commonly used to simulate molecular liquids.
We investigated whether experimental surface tension measurements, which are less sensitive to quantum and self-polarization corrections, could replace the commonly used heat of vaporization data.
To test this hypothesis we implemented new functionality into the automated ForceBalance optimization procedure for using surface tension data, and parameterized revised three-point and four-point rigid, fixed charge water models named TIP3P-ST and TIP4P-ST.
The TIP3P-ST model reproduces the correct temperature of maximum density of water for the first time, but also leads to over-structuring of the liquid and less accurate transport properties.
The TIP4P-ST model is highly accurate for a broad range of thermodynamic and kinetic properties, with similar performance compared to other recently developed four-point models.
The results show surface tension to be a useful fitting property in general, especially when self-polarization corrections or nuclear quantum corrections for heat of vaporization are not readily available for molecular liquids of interest.