Refinement of COSMO-SAC and the applications

Shu Wang, Stanley I. Sandler, Chau Chyun Chen

Research output: Contribution to journalArticlepeer-review

167 Scopus citations


In dielectric continuum solvation models (CSMs), the solvation free energy is essential for accurate calculation of the thermodynamic properties of liquids and liquid mixtures. Here, we present a refinement to the COSMOSAC (where SAC denotes segment activity coefficient) model to calculate the solvation free energy, using statistical thermodynamics and information from quantum mechanics. After the surface charge density for each pure compound has been obtained from a one-time quantum mechanics calculation, the COSMO-SAC model with 23 parameters can be used to predict both pure and mixture thermodynamic properties of almost any nonmetallic molecule in a computational time comparable to that of group contribution methods, but without their hundreds of functional-group-specific parameters. Combining perturbation theory and a mean field model for the van der Waals (vdW) free-energy calculation, this COSMO-SAC model can be used to predict the properties of pure compounds. The overall deviation in vapor pressure for 1432 compounds is ~63%, and that in the normal boiling temperature prediction is ∼20 K; these values are comparable to those of group contribution methods, which require a much larger set of parameters. This model can also be used to predict activity coefficients in mixtures with the same parameters, and here we have calculated vapor-liquid equilibria for more than 500 binary systems with over 8000 experimental data points, and for seven ternary systems. For these systems, the overall average deviation in vapor-phase compositions is 2.14%, and the deviation in total pressure at a fixed liquid composition is 6.03%.

Original languageEnglish
Pages (from-to)7275-7288
Number of pages14
JournalIndustrial and Engineering Chemistry Research
Issue number22
StatePublished - Oct 24 2007


Dive into the research topics of 'Refinement of COSMO-SAC and the applications'. Together they form a unique fingerprint.

Cite this