Liquid-phase chemistry plays a key role in many practical applications of chemistry, including pharmaceutical development, chemical manufacturing, and environmental science. Recent developments in computation and theory have enabled researchers to accurately estimate gas-phase properties using quantum-chemical techniques and statistical mechanics. Modeling solvent-phase systems requires additional effort. One approach has been to calculate the thermodynamic and kinetic properties of the reaction in the gas phase, and then to apply corrections which account for the solvent and solutes of interest. These solvation corrections are derived from quantum chemistry and experimental data.
In this project, we construct and curate datasets relating to solvation phenomena, which can be used to build models or as reference data; we then develop methods for quickly and accurately predicting physicochemical properties for liquid-phase systems.
Our group has developed several compilations of experimental data related to solvation, including solvation free energies and solubilities of neutral solutes in diverse solvents, and acid dissociation constants in water. We have also leveraged deep learning to develop models that predict solvation enthalpies, solvation free energies, and solubilities. Currently, we are developing methods for estimating rate constants of reactions in liquid media, and extending our results to systems with multiple solvents, as well as ionic and radical solutes in pure neutral solvents.