Edward MAGINN1
1University of Notre Dame, Notre Dame, United States
Global warming concerns have led world leaders to call for the phaseout of high global warming potential hydrofluorocarbon (HFC) refrigerants. Ionic liquids (ILs) are being considered as potential solvents which can be used to separate azeotropic HFC mixtures and recycle low GWP HFCs. Selecting the best IL for a given HFC separation requires knowledge of many physical properties of the IL-HFC system, including solubility, viscosity, heat capacity, thermal conductivity, and diffusivity. Molecular simulations are an effective way to compute these properties to help search through the vast number of potential ILs to find the one best suited for a given HFC separation.
Here, we report results of Hamiltonian replica exchange molecular dynamics simulations, which shows that full gas isotherms in ILs can be computed accurately and more quickly than with Gibbs ensemble Monte Carlo methods. We also show that single stage free energy perturbation methods can enable Henry’s law constants of HFCs in ILs to be determined using pre-computed trajectories of the liquid IL. This enables rapid screening of ILs for gas solubility. Using highly accurate force fields for HFCs developed using machine learning, we also compute thermophysical and transport properties of HFC-IL systems and show good agreement with experiment. Finally, we describe our efforts at using machine learning along with sigma profiles to correlate and predict the properties of molecular solvents, ionic liquids, and deep eutectic liquids.