Prediction of Liquid–Liquid Equilibrium of Ternary Systems Using COSMO-SAC: Application to Water–Alcohol–Hydrocarbon Systems

Liquid–liquid equilibrium (LLE) in ternary systems composed of water, alcohols, and hydrocarbons is of considerable relevance to separation processes, solvent design, and thermodynamic modeling, owing to the strong non-ideality arising from hydrogen bonding and polarity contrasts among the components. In this context, the present work evaluates the predictive performance of the COSMO-SAC model in describing the LLE behavior of the cyclohexane/n-propanol/water system at 25 °C, using experimental data from the literature as a reference.

The calculations were carried out using the COSMO-SAC model implemented in JCOSMO, with σ-profiles obtained from quantum chemical calculations. Phase equilibrium conditions were established by enforcing the equality of chemical activities of all components in the coexisting phases, while the lever rule was applied to relate phase compositions to selected global compositions within the biphasic region. The resulting nonlinear system was solved using iterative numerical methods, enabling the determination of the compositions of the hydrocarbon-rich and water-rich phases, as well as the corresponding tie-lines, without the use of adjustable parameters.

The predicted phase behavior shows good qualitative agreement with the experimental binodal curve, particularly in the water-rich and cyclohexane-rich regions. More noticeable deviations were observed in the intermediate composition range, mainly associated with the partitioning of n-propanol between the two phases, reflecting inherent limitations of the model in the quantitative representation of specific alcohol–water interactions. A quantitative assessment yielded an overall root mean square error (RMSE) of approximately 0.07 in mole fraction, which substantiates the predictive capability of the COSMO-SAC model and confirms its applicability to liquid–liquid equilibrium analyses and computational simulation frameworks for highly non-ideal ternary systems.