5-hydroxymethylfurfural (HMF) is a versatile bio renewable molecule that can be used as an important platform chemical to produce biofuels and a wide variety of chemical products. HMF is commonly produced by the acid-catalyzed dehydration of fructose, because of its high conversion rate, high yield and high intrinsic selectivity. However, in order to compete with oil-based alternatives, there is a strong need to improve the overall cost-efficiency of the HMF production process. A major hurdle is the stability of the formed HMF in the reaction mixture, leading to significant loss of overall selectivity due to degradation reactions.
Liquid-liquid extraction can be beneficially applied to avoid this HMF degradation by separating HMF immediately from the reaction medium [1]. In our previous work [2-5], methyl isobutyl ketone (MIBK) and 2-pentanol were identified as promising extraction solvents due to their strong hydrogen bonding interactions with HMF. To increase process selectivity even further, ionic liquids (IL) and deep eutectic solvents (DES) can be applied as stabilizing agent in the reaction mixture. As their stabilization effect is mainly the result of strong hydrogen bonding interactions with HMF, the efficiency of HMF extraction from the reaction mixture will suffer from a reduced distribution coefficient [4,5]. Therefore, the objective of this work is to evaluate whether the application of a reactive solvent can alleviate this issue. Tributyl phosphate (TBP) was selected as reactive solvent, as it is known to be very effective for HMF extraction and also has a low solubility in water [6].
In order to allow a techno-economic comparison of TBP as reactive solvent with the physical solvents, first the liquid-liquid equilibrium (LLE) data of systems containing HMF in aqueous solution using tributyl phosphate (TBP) as the extraction solvent were characterized at 313.15 K. DES ChCl-urea (choline chloride-urea), salt (NaCl), and sugar (fructose) were also added to the aqueous phase to represent an industrially relevant system. The experimental LLE data were also correlated with thermodynamic models. The results confirmed that TBP exhibits a 2-3 times higher HMF distribution coefficient and is slightly more selective compared to the physical solvents (MIBK and 2-pentanol).
Given this promising result, a conceptual process design was established for HMF production (20-50 kT/yr) in a biphasic system for TBP and MIBK as reactive and physical solvent respectively. The major difference between both designs being that TBP is a high boiling solvent for which HMF recovery should be conducted by back-extraction, while MIBK is easy to separate from the extracted HMF by distillation. The conceptual design revealed that for back-extraction an anti-solvent (hexane, heptane, octane) is required to drive the HMF from the TBP back into an aqueous phase and push the energy requirements to below those of the MIBK process. This way the HMF concentration in the aqueous phase after back-extraction is significantly increased, resulting in lower energy requirements than MIBK separation by distillation. However, this comes at the expense of a more complex process that, in spite of the better TBP extraction performance, resulted in a double CAPEX requirement compared to MIBK. Overall this resulted for the same reaction yield in a 15% higher manufacturing cost and 5% lower CO2 footprint of the TBP based process. As the fructose feedstock is the main contributor to the manufacturing costs as well as the CO2 footprint, it can be concluded that the application of reactive solvents such as TBP only makes sense when significantly (>5%) higher reaction yields can be obtained compared to employing physical solvents.