A comparison of COSMOtherm with COSMO-SAC-VT-2005
A team of researchers from Virginia Tech University, along with others from the university of Delaware, have created a software largely based on the COSMO-RS method, developped by Andreas Klamt, founder of COSMOlogic and on the COSMO-SAC method, published by Lin and Sandler. They have made available a sigma-profile database comprising 1200 compounds (COSMOtherm's TZVP database has over 4000), as well as software for creating new sigma-profiles and for using these profiles to generate binary phase-equilibrium predictions.
Chlorinated hydrocarbons
We decided to compare this new software, called COSMO-SAC-VT-2005, with COSMOtherm when predicting water solubility, using the same set of chlorinated hydrocarbons that was chosen for evaluating the accuracy of COSMOtherm (see Chlorinated Hydrocarbons). About 35 compounds were included in both databases, and served as the subset for the comparison. Unfortunately, COMO-SAC-VT-2005 does not provides solubilities, only the activity coefficients γ1 and γ2 of two compounds in binary phase-equilibrium predictions, for set values of molar fractions, from 5.10-3 to 0.995 by increments of 0.01. For the lowest value of the molar fraction, the concentration is low enough for it to be roughly equal to the inverse of γ. The solubility can thus be inferred.
Study conducted by Antoine Hervier
The graph shows the results of the comparison.
As the graph shows, COSMOtherm performed quite well, with an average error of 0.29 log units, whereas COSMO-SAC-VT-2005 only managed an error of 1.27 log units. Although the general trend is predicted, the software systematically overestimates the experimental value by roughly one order of magnitude.
Click here for an additional comparison.
back to various examples