Parametrization of two-center Lennard-Jones plus point-quadrupole force field models by multicriteria optimization
To ensure the quantitative precision and reliability of molecular simulations, force field models of molecular fluids need to be adjusted to e.g. experimental data. An optimal agreement for different properties is often not achieved by a single model parametrization. Applying multicriteria optimization, based on the evaluation and analysis of the Pareto set, solves this problem. The Pareto set contains all optimal compromises between multiple conflicting objectives. Its computation and suitable visualization enables the end user to freely choose a model parametrization, tailored to his particular application scenario. We apply multicriteria optimization to the two-center Lennard-Jones plus point-quadrupole model class (2CLJQ), which has four adjustable parameters. The Pareto set is determined and analyzed for ten real fluids: Ethane, ethylene, acetylene, fluorine, chlorine, bromine, perfluoroethylene, perchloroethylene, nitrogen, and oxygen. Thereby, two multicriteria optimization scenarios are considered, based on two criteria (saturated liquid density and vapor pressure) and three criteria (saturated liquid density, vapor pressure, and surface tension), respectively. It is shown that literature models for these fluids can be further improved in these criteria. We visualize our results by self-organizing patch plots, which facilitate the representation of the entire Pareto set and its corresponding model parametrizations.