Estimation of MOSCED parameters from the COSMO-SAC database

MOSCED (Modified Separation of Cohesive Energy Density) is a particularly attractive model for infinite dilution activity coefficients because it offers intuitive insights into how to tune solvent-solute interactions to achieve optimized formulations. Unfortunately, only 133 compounds have been characterized with the MOSCED method. Available methods for extending MOSCED predictions to new compounds involve quantum mechanical computations on a compound-by-compound basis. On the other hand, surface charge densities (σ-profiles) from density functional theory (DFT) computations are readily available for 1432 compounds through a public database. This work develops correlations for estimating MOSCED parameters from the available σ-profiles, extending the applicability of MOSCED from 133 to roughly 1400 compounds. Additional DFT computations for new compounds would be straightforward. The predictive functions were regressed based on 4368 binary solution infinite dilution activity coefficients. The average logarithmic deviation for the predictive MOSCED model was 0.318 while using the original correlative model had a deviation of 0.106 compared to 0.183 for the UNIFAC model and 0.444 for the Virginia Tech implementation of COSMO-SAC. Analysis of a predictive “simplified” SCED model (PSSCED) was also considered. Omitting MOSCED's polarity and solvent polarity parameters increased deviations to 0.596. As a basis for comparison, regular solution theory provided deviations of 0.947 for the same data set.