(730b) Introducing an Individualized Transferable Anisotropic Mie Force Field

Transferable force fields allow predicting physical properties and phase equilibria of pure substances and mixtures that are weakly described by experimental data. However, for well-known substances transferable force fields lead to higher deviations than desired and necessary. Well-characterized substances could be described better if an individually optimized force field would be used. In this study we introduce the concept of individualized transferable force fields that leads to accurate representations of well-characterized substances.

Parameters of the transferable anisotropic Mie (TAMie) force field are obtained iteratively from Monte Carlo simulations in grand canonical ensemble and doing histogram reweighting in a post processing step to calculate the phase coexistence curve. New parameters are generated by utilizing a physically based equation of state (PC-SAFT EOS) to speed up the iteration procedure. The objective function for optimizing the force field parameters contains several substances of the same group where we mainly focus on vapor pressure and liquid density data. For non-polar substances, like alkanes, the approach leads to excellent results. However, describing polar substances like esters with a transferable set of parameters necessitates further trade-offs. If, what we prefer, the point charge model is applied and the point charges are assumed to be the same for the entire substance group, the overall optimal set of parameters might be outside the range defined through individual optimization of pure substances. The trade-off in parameterizing a transferable force field to several substances manifests mainly in deviations in vapor pressure: the vapor pressure data of some substances are too high whereas for other substances they are too low. This error can easily be corrected for. We introduce a correction parameter Ï?_ε which affects all energy parameters ε of a pure substance. As a result, we obtain a transferable force field for weakly-characterized substances and an accurate force field for substances which are well described by experimental data. Applying these correction parameters Ï?_εopens the opportunity to simulate binary mixtures at different temperatures without introducing significant errors caused by deficiencies in vapor pressure of the pure substances. The results for binary mixtures containing ketones and alkanes show good agreement with experimental data.

Our study shows on the one hand the benefit of individualized transferable force fields for the description of pure substances and for the predicting of mixture phase equilibria, and on the other hand we demonstrate the transferability of binary interaction corrections k_ij (with respect to Lorentz-Berthelot combining rules) defined for united atom groups.