Data-driven Parameterization of Chemically Specific Field-theoretic Simulations

Field-theoretic simulations (FTS) are an important method in the computational study of polymeric materials. Recent work has extended the chemical specificity of FTS, but this greater specificity has inevitably resulted in additional model parameters that must be estimated. In this work, we describe a new data-driven method to parametrize FTS that leverages small molecule equilibria data to predict the phase behavior of polymeric materials. Starting at the mean-field approximation, we find FTS model parameters that reproduce large sets of small molecule data and that extrapolate well to untrained systems. We also show how the type of thermodynamic training data affects FTS predictions, how diverse thermodynamic data can be readily incorporated into our parametrization, and how such parametrized models can be extended to include fluctuation effects and microphase separation.