(448c) Design of Solvent Composition for Acid-Catalyzed Reactions of Biomass-Derived Oxygenates Using Molecular Simulation-Derived Observables

Chew, A. - Presenter, University of Wisconsin
Walker, T., University of Wisconsin - Madison
Li, H., Dalian Institute of Chemical Physics
Demir, B., University of Wisconsin-Madison
Zhang, Z. C., Dalian Institute of Chemical Physics
Huber, G., University of Wisconsin-Madison
Dumesic, J., University of Wisconsin-Madison
Van Lehn, R., University of Wisconsin-Madison
The aqueous-phase, acid-catalyzed conversion of lignocellulosic biomass is a promising method for the production of liquid fuels and high-value commodity chemicals from renewable feedstocks, but achieving sufficient product selectivity and yields remains an ongoing challenge. One simple strategy to improve reaction yields is to modify the solvent composition, which affects reaction rates, product selectivity and stability, and the economics of downstream separations. However, identifying an optimal solvent composition empirically by trial-and-error is cost-prohibitive and gives little insight into how the solvent environment will perform in new processes. We will describe our efforts to use classical atomistic molecular dynamics (MD) simulations to predict the effect of solvent composition on the conversion of biomass-derived model compounds. Leveraging the computational efficiency of MD simulations, we simulate a range of solvent mixtures containing water and polar aprotic cosolvents in varying mass fractions. Based on experimentally determined reaction rates, we show that as the water content of the solvent environment decreases, reactants with more hydrophilic groups have improved catalytic turnover rates for both hydrolysis and dehydration reactions. We explain this behavior by using MD simulations to quantify the size and properties of water-enriched local domains in the vicinity of the reactant in mixed-solvent environments. We further demonstrate that multiple simulation-derived observables can be combined into a predictive model that correlates with experimentally determined reaction rates for systems with varying solvent compositions. Finally, we demonstrate that analysis of the spatial distribution of solvent molecules can inform our understanding of reaction selectivity. These computational models are important for enabling the rational design of new liquid-phase biomass conversion processes by guiding the selection of specific cosolvents.