Predicting liquid-phase thermodynamic properties using COSMO-SAC

Abstract

Predicting thermodynamic properties of liquids remains a significant challenge in both academia and industry. This is because molecules in the liquid phase are relatively close together but are not arranged in an ordered structure. The COSMO-SAC model uses quantum and statistical mechanics to predict the thermodynamic properties of liquids. In this thesis, a variety of efforts have been made to improve the accuracy of this model. The most successful of these involves the inclusion of dispersion interactions in mixture calculations, which reduces the average prediction error for activity coefficients by over 30%. It has also been demonstrated that the misfit energy—the traditional interaction energy used in all known COSMO-based models—is limited in its ability to model intermolecular interactions. A variety of attempts to improve the electrostatics of the model have been made, but these have had little effect on the overall accuracy. These include use of an expanded hydrogen bonding description, use of an alternate molecular surface construction method, and incorporation of three-dimensional considerations. These results, coupled with the observed improvement obtained from including dispersion interactions, suggest that the greatest weakness in the model is the handling of electrodynamics (how molecules polarize one another), and that future efforts to improve the model should target this area.

In addition, the model has been parameterized for the ADF density functional software package, thus making it available to a wider audience. Results are approximately equivalent to those obtained with DMol³, the only package for which a published parameterization was previously available. Of significance, the ADF implementation allows one of the adjustable model parameters to be eliminated.

Also, a correlation was found between experimental liquid volumes and the quantum-based molecular volumes employed in COSMO-SAC. This eliminates the need to provide experimental volumes to the model.

Finally, a Windows-based graphical user interface for COSMO-SAC has been developed. This expands the availability of the model to anyone with only a general familiarity with thermodynamics. Without this, use is restricted to those with an in-depth knowledge of quantum mechanics, statistical mechanics and computer programming.