Conformer diversity over quantity in COSMO-RS-based LLE predictions
Accurate prediction of phase equilibria is essential for chemical process design, particularly in systems characterized by complex intermolecular interactions. This work investigates the influence of conformer selection and sigma-profile characteristics on COSMO-RS predictions of liquid-liquid equilibria (LLE), using the tetrahydrofurfuryl alcohol (THFA)-methyl isobutyl ketone (MIBK)-water system as a case study. Calculations were performed at the TZVP and TZVPD-FINE levels of theory. While TZVP failed to capture the experimentally observed phase behavior, TZVPD-FINE provided substantially improved predictions. The differences are attributed to the relative importance assigned to less-polar conformers and more-polar conformers capable of hydrogen bonding. Although the TZVPD-FINE calculations reproduced the main experimental trends, the predicted organic-phase composition showed a mean percentage deviation of 5.27%, slightly exceeding the adopted 5% agreement threshold. Investigations using individual conformers demonstrated that predictions can be strongly influenced by conformer selection and that representative conformer ensembles are necessary for reliable results. Further analysis showed that diversity in sigma profiles is more important for predictive accuracy than the total number of conformers. These findings emphasize the importance of carefully selecting conformers when applying COSMO-RS to liquid-liquid equilibria involving phases of contrasting polarity.