Experimental and Predictive Modeling Study of Liquid–Liquid Equilibria for Water + Toluene + (2-Methylpyridine, 2-Methylpyrazine, and Acetophenone) Ternary Systems Using Group Contribution and COSMO-Based Methods

Abstract 2-Methylpyridine and 2-methylpyrazine form azeotropes with water, limiting the applicability of conventional distillation for their separation from aqueous mixtures. In addition, LLE data for acetophenone-containing systems remain scarce despite the broad industrial relevance of acetophenone. In this study, the LLE data were acquired for ternary systems of water + toluene + 2-methylpyridine, water + toluene + 2-methylpyrazine, and water + toluene + acetophenone at 298.15 and 313.15 K. Measured LLE data were successfully correlated by the NRTL model, and regressed binary interaction parameters were validated by topological analysis of the dimensionless molar Gibbs energy of mixing. The obtained LLE data were also utilized to evaluate the predictive capability of two classes of models: group-contribution methods (GCMs) and quantum-chemistry-based models. The studied GCMs included original UNIFAC, UNIFAC-LL, and modified UNIFAC (Dortmund), whereas the COSMO-based models comprised COSMO-SAC and COSMO-RS at the TZVP and TZVPD-FINE theory levels. Overall, all studied predictive models provided qualitatively acceptable predictions of the LLE behaviour. Among them, UNIFAC-LL showed the best overall predictive capability, whereas COSMO-SAC displayed the weakest performance. For COSMO-RS, the TZVPD-FINE level provided more accurate predictions of the experimental data than the TZVP level. UNIFAC performed better than modified UNIFAC (Dortmund) for the water + toluene + 2-methylpyridine system. In contrast, modified UNIFAC (Dortmund) yielded more accurate predictions than UNIFAC in the other two studied ternary systems.