A comparative study of COSMO-based and equation-of-state approaches for the prediction of solvation energies based on the compsol databank.

The computation of solvation energies has many uses in several fields, such as design of separation processes, pharmacology and drug-design, and kinetic modeling. These applications require thermodynamic models capable of accurately predicting solvation energies, accounting for the temperature dependency of this property, and which are fast and robust. Within this framework, we compared two COSMO-based continuum solvation models (COSMO-RS and COSMO-SAC-dsp) with two versions of predictive cubic equations of state well-acknowledged for their efficiency (PSRK and UMR-PRU). For this purpose, a large experimental set of 57,000 datapoints extracted from the COMPSOL databank was considered. Comparisons between computed and experimental data were performed for Gibbs solvation energy and, for the first time in the literature, for both entropy and enthalpy of solvation simultaneously. For simpler binary mixtures, in which hydrogen bonding does not take place, all models were capable of providing accurate predictions, with average absolute deviations below 0.3 kcal/mol regarding the solvation Gibbs energy. For more complex associating mixtures, COSMO-RS showed the best correlation between experimental and calculated data, especially for aqueous systems; among EoS, it is observed that the PSRK model offers the best accuracy. [Display omitted] [ABSTRACT FROM AUTHOR]