Rapid Solvent Screening Using Thermodynamic Models for Recovery of 2,3-Butanediol from Fermentation by Liquid–Liquid Extraction

Abstract

Different classes of solvents were explored for liquid–liquid extraction (solvent extraction) of 2,3-butanediol (BDO) from aqueous solutions. Initially, the activity coefficients (γ∞) of BDO in various solvents were predicted using thermodynamic models, namely conductor-like-screening model-real solvents (COSMO-RS) and universal functional group activity coefficient model (UNIFAC-LL). These models were validated by comparing the distribution coefficients (Kd) predicted from modeling with experimental data for the solvents. Lower chain alcohols and derivatives of phenol showed highest Kd. 1-butanol, 2-sec-butylphenol (SBP) and cyclohexanol showed experimental Kd of 1.6, 1.2 and 0.9, respectively in aqueous fermented BDO broth. The UNIFAC-LL and COSMO-RS based predictions provided qualitative trend for the ranking of solvents, in reasonable agreement with the experimental data. In terms of numerical ability, UNIFAC-LL model performance is better compared to COSMO-RS. However, this work shows that the incorporation of newly improvised UNIFAC parameters in the UNIFAC-LL calculations provides a very satisfactory result, much closer to the experimental values. Therefore, newly parametrized UNIFAC-LL could be used for further studies on process modeling, design, and scale up of BDO extraction from aqueous solution.