A water model with improved thermodynamic properties over a large pressure and temperature range optimized using multistate reweighting and configuration space mapping

We show how reweighting and configuration mapping algorithms can be used to efficiently optimize molecular models using thermodynamic properties at a large number of state points from molecular simulations. As a proof of concept, we perform a multidimensional, multi-objective parameterization of a rigid water model over a large pressure [1-5000 atm] and temperature [274.15-372.15 K] range to the experimental property surfaces estimated using the IAPWS95 equation of state for water. Over 4000 parameter combinations in six-dimensional parameter space were explored during the minimization. A similar parameterization with standard techniques would have taken more than 2000 CPU years but with the application of the newly developed techniques, the computational time was reduced to eight CPU weeks. Without the added efficiency of the methods presented here, the optimization could not have simultaneously taken into account the large range of temperature and pressure points. The paper also describes how and why incorporating the thermodynamic properties from the first derivatives and the second derivatives of Gibbs energy into the objective function help improve the parameterization process.