Data-Driven Many-Body Models with Chemical Accuracy for CH4/H2O Mixtures.

Many-body potential energy functions (PEFs) based on the TTM-nrg and MB-nrg theoretical/computational frameworks are developed from coupled cluster reference data for neat methane and mixed methane/water systems. It is shown that the MB-nrg PEFs achieve subchemical accuracy in the representation of individual many-body effects in small clusters and enables predictive simulations from the gas to the liquid phase. Analysis of structural properties calculated from molecular dynamics simulations of liquid methane and methane/water mixtures using both TTM-nrg and MB-nrg PEFs indicates that, while accounting for polarization effects, is important for a correct description of many-body interactions in the liquid phase, an accurate representation of short-range interactions, as provided by the MB-nrg PEFs, is necessary for a quantitative description of the local solvation structure in liquid mixtures.