(222bb) Understanding the Role of Dispersive Interactions in the COSMO-SAC Model

A reliable predictive thermodynamic model is highly desirable to reduce the amount of required experimental data, which are crucial for the design of separation or purification processes in chemical and pharmaceutical industries. In the past two decades, a new class of predictive methods, i.e. COSMO-based approaches (COSMO-SAC or COSMO-RS), has emerged. These methods utilize the results from modern computational chemistry and do not contain any species-dependent parameters. Therefore, the problem of missing parameters does not arise, and crucial thermophysical information can be predicted with the molecular structure as the only input. Recently, the COSMO-SAC model was revised to provide acceptable predictions for both vapor-liquid and liquid-liquid equilibria of organic mixtures. However, a well-known issue, which is the neglect of the dispersive molecular interactions in mixtures, remained unsolved in the COSMO-SAC model. In this study, classical force field based molecular modeling and simulation is used to supply data on a sound physical basis and a new dispersion term to consider the contribution of the dispersive interaction to the non-ideality of mixtures is introduced into COSMO-SAC model. With this modification, the accuracy in vapor-liquid equilibrium predictions can be improved by around 10%. For 391 different binary mixtures, the average deviation in terms of total pressure at a fixed liquid composition is 4.7% and the overall average deviation in terms of vapor phase composition is 1.7%, compared to 5.3% and 1.8% from the original model.