On Multiplicities in Thermodynamic Activity Coefficient Models

This dissertation examines the origin and impact of multiplicities in the NRTL and UNIQUAC activity-coefficient models for binary mixtures. These models can produce identical activity-coefficient pairs from different parameter sets up to five for NRTL and up to three for UNIQUAC. A global numerical method is developed to identify all feasible parameter sets for a given activity-coefficient pair. Such multiplicities cause significant issues in thermodynamic modeling: different parameters can fit the same VLE data within experimental error but lead to divergent predictions when extrapolated in temperature, pressure, or multicomponent systems. Dynamic simulations of batch distillation show that alternative parameter sets produce notably different trajectories. To resolve this non-uniqueness, this work introduces unique versions of NRTL and UNIQUAC that restrict parameter spaces while preserving model flexibility, ensuring a one-to-one mapping between parameters and activity coefficients. These unique models improve predictive reliability and primarily serve to initialize classical model fits.