(283c) Effective Property Prediction for Solvent Design and Bioproduct Extraction

Lignocellulosic biomass is a promising renewable feedstock for the production of platform chemicals and biofuels [1]. The catalytic conversion of biomass often yields low-concentration bioproducts that must be separated from the aqueous phase. One possible low-energy separation technology is liquid-liquid extraction [2], which depends critically on the selection of an extracting solvent. Solvent selection requires knowledge of the thermodynamic properties of the product and solvent components [3,4]. Given the large number of possible solvent systems that could be used for extraction, the evaluation of bioproduct and solvent properties by experiments alone is limiting. Instead, molecular modeling using the COnductor-like Screening Model for Real Solvents (COSMO-RS) has been widely used to predict equilibrium properties from first-principles calculations [5]. However, the prediction accuracy of COSMO-RS depends sensitively on the selection of an appropriate ensemble of molecular conformations [6]. Hence, research challenges exist in efficiently exploring the molecular conformer space in order to identify optimal solvent systems for bioproduct extraction.

In this work, we propose a framework for predicting thermodynamic properties relevant to liquid-liquid extraction based on molecular dynamics (MD) simulations, optimization-based adaptive conformer sampling, and COSMO-RS modeling. MD simulations are used to generate a search space of molecule structures at representative temperatures and solvent environments. Conformers are then clustered based on physical metrics (e.g., the solvent accessible surface area and radius of gyration) and their frequency. We formulate a mixed-integer quadratic programming problem to identify clusters for conformer selection. Conformers are then used as input for COSMO-RS predictions of bioproduct partition coefficients, boiling points, and related thermodynamic data. We show that iteratively increasing the number of conformer clusters can improve the prediction accuracy based on comparisons to experimental data for a spectrum of representative bioproducts, permitting the analysis of tradeoffs between computational cost and prediction accuracy. These predicted thermodynamic properties define a search space of feasible processes for the liquid-liquid extraction of desired bioproducts, and moreover can be computed with sufficient computational efficiency to enable integration with process models.

[1] Luterbacher, J. S., Rand, J. M., Alonso, D. M., Han, J., Youngquist, J. T., Maravelias, C. T., Pfleger, B.F., Dumesic, J. Nonenzymatic sugar production from biomass using biomass-derived γ-valerolactone. Science, 343(6168), 277-280, 2014.

[2] Taifan, G. S., Maravelias, C. T. Integration of graphical approaches into optimization-based design of multistage liquid extraction. Computers & Chemical Engineering, 143, 107126, 2020.

[3] Hostrup, M., Gani, R., Kravanja, Z., Sorsak, A., & Grossmann, I. Integration of thermodynamic insights and MINLP optimization for the synthesis, design and analysis of process flowsheets. Computers & Chemical Engineering, 25(1), 73-83, 2001.

[4] Shen, Z., Van Lehn, R. C. Solvent selection for the separation of lignin-derived monomers using the conductor-like screening model for real solvents. Industrial & Engineering Chemistry Research, 59(16), 7755-7764, 2020.

[5] Klamt, A., Eckert, F., Arlt, W. COSMO-RS: an alternative to simulation for calculating thermodynamic properties of liquid mixtures. Annual review of chemical and biomolecular engineering, 1, 101-122, 2010.

[6] Hawkins, P. C. Conformation generation: the state of the art. Journal of Chemical Information and Modeling, 57(8), 1747-1756, 2017.