Development of Advanced Computational Methods for Accurate Property Prediction of Fluids: Application to Hydrofluorocarbon/Ionic Liquid Mixtures

Hydrofluorocarbon (HFC) refrigerants with zero ozone-depleting potential have replaced chlorofluorocarbons and are widely used in the heating, ventilation, air-conditioning, and refrigeration industries. However, some HFCs exhibit high global warming potential, which has led to calls by governments for their phaseout. Technologies to recycle, repurpose, and separate these HFCs (which are often used as mixtures) need to be developed. Ionic liquids (ILs) show promise as efficient entrainers for separating HFC mixtures via extractive distillation. Therefore, thermophysical properties of HFCs, ILs, and their mixtures are needed over a wide range of conditions. Molecular simulations can help understand and predict various properties effectively. The first half of this work explores the application of molecular dynamics (MD) simulations in understanding fluid properties and method development of solubility calculations, with a focus on HFC refrigerants and IL systems. To be specific, we proposed a new all-atom force field (FF) for tris(pentafluoroethyl)trifluorophosphate ([FAP]) anion and investigated the effect of [FAP] isomer content on different properties of 1-n-hexyl-3-methylimidazolium tris(pentafluoroethyl)trifluorophosphate. Thermophysical, dynamic, and structural properties were also systematically studied for HFC-32 and HFC-125 in imidazolium-based ILs by MD simulations. At the same time, we also developed a workflow that combines Hamiltonian replica exchange MD simulations with alchemical free energy calculations to accurately compute the full solubility isotherm of HFC/IL mixtures. Accurate FFs are essential for meaningful property prediction. Therefore, the second half of this work focuses on calibrating Lennard-Jones parameters of classical FFs using machine learning techniques for five refrigerants and developing the General Amber FF-based polarizable models for ILs using Drude oscillators.