An Improvement to COSMO-SAC for Predicting Thermodynamic Properties

A modified COSMO-SAC model is presented to calculate thermodynamic properties of pure fluids and mixtures using statistical thermodynamics and the surface charge density of each compound obtained from a quantum mechanics (QM) calculation. The main differences from the previous models are that the new model includes a dispersion contribution in the mixture interaction, and is reparametrized using measured pure and mixture thermodynamic data simultaneously. With a single set of universal parameters, the new model provides higher accuracy than our previous models for predicting mixture thermodynamic properties while maintaining the same accuracy for pure compound thermodynamic properties. The overall root-mean-square deviation (RMSD) in the logarithms of partition coefficients for 992 octanol water partitioning systems and 829 other solvent water partitioning systems with this new model is reduced by about 10% compared to the results from previous models. Also, the agreement between the predicted and measured partition coefficients over a wide range of values is improved as a result of better activity coefficient predictions at high dilution by inclusion of the dispersive mixture interaction in the model. The accuracy in the vapor liquid equilibrium (VLE) predictions is comparable to, or better than, the previous model that was developed for phase equilibria calculations only. The new model also provides parameters for use with the Amsterdam Density Functional (ADF) in addition to DMol(3).