<br/>Many environmentally and industrially important chemical reaction systems occur in a condensed phase such as atmospheric aerosol chemistry, organic synthesis, and oxidative degradation of oil and drugs. Solvents can shift the rate of reactions often by several orders of magnitude even for neutral reactions, and it is important to properly account for kinetic solvent effects in order to predict liquid phase reaction rates accurately and to select the optimal solvent candidate for the reactions of interest.<br/>In this work, we show that (1) quantum chemical calculation combined with COSMO-RS method can accurately capture the kinetic solvent effects and compute liquid phase reaction rates for various free radical and neutral reactions and (2) a machine learning model can be constructed to predict kinetic solvent effects only based on reaction and solvent SMILES representations. Several quantum chemistry methods are tested in combination with single-point COSMO-RS calculation to compute liquid phase kinetic rates. The calculation accuracies are evaluated by comparing with 190 experimental rate constant data of 15 reactions and 44 solvents over 273 - 392 K, and the MAE of 1.3 - 1.6 kcal/mol is achieved for the best method. After demonstrating that the DFT (density functional theory) and COSMO-RS methods together can accurately predict the effect of solvents on reaction rates, we performed 7.8 million COSMO-RS calculations for 13,224 reactions obtained from literature in 298 common solvents and trained a machine learning model to predict the rate of a liquid phase reaction relative to the rate in a gas phase. The machine learning model has shown to make reasonable predictions when evaluated on 10 % test sets, demonstrating its promising capability to automatically predict kinetic solvent effects.

Predicting kinetic solvent effects using quantum chemistry, COSMO-RS, and machine learning

Fall 2022 abstracts, posters, and presentations will launch in mid-September.

ACS Fall 2022

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.