Solubility prediction in mixed solvents: A combined molecular simulation and experimental approach

Abstract

We propose a simple method to predict solubility in mixed solvents, combining experimental data for solubility in pure solvents with non-ideal contributions obtained from molecular simulation. By evoking a well-established thermodynamic model for mixed solvent, we provide rationale and justification for this hybrid approach. We test the accuracy of the method for solubility prediction in two highly non-ideal mixtures, CO2 in ethanol + water, and CH4 in MDEA + water. Non-ideal behaviour in each system is characterised using molecular simulation across the full range of compositions and for temperatures covering the range 273.15 K–373.15 K. Comparison against experimental Henry's law constants for CO2 and CH4 gives mean absolute errors of 6.9% and 27%, respectively. We investigate the origin of non-ideal Henry's law behaviour in each of the mixed solvents by interrogating radial distribution functions from molecular simulation, finding increased association of CO2 and CH4 with the organic solvents at elevated temperatures.