UC Berkeley

A major component of lignocellulosic biomass is hemicellulose, a polysaccharide composed of monomeric sugars, principally xylose. Xylose can be dehydrated, most often in aqueous solution, using Brønsted acid catalysts to form furfural, which can be further reacted to produce fuels, lubricants, polymers, solvents, and pharmaceutical precursors. Furfural production can also be enhanced by using Lewis acid catalysts, which promote the formation of xylulose, an isomer of xylose which more readily dehydrates to form furfural. With either type of catalyst, side reactions consume furfural to produce a group of soluble and insoluble products known as humins. Humins formation has been stymied by extracting furfural as it is produced. This is done on the industrial scale with steam stripping, but researchers have also explored the use of liquid-liquid extraction (LLE) by an organic solvent (typically 2:1 organic:aqueous volume ratio) for the same purpose. Both extraction methods increase furfural yield, but dilute the product phase, which raises the cost of furfural production. A new method, e.g. pervaporation, must be developed to increase furfural yield and concentrate furfural in the product simultaneously.

Pervaporation is a membrane-based process in which a liquid mixture is placed in contact with the feed side of the membrane while a vapor is located on the permeate side. A vacuum is used to reduce the partial pressure, and therefore the fugacity, of components in the permeate, which provides the driving force for mass transfer. Pervaporation is most often used to separate water from concentrated ethanol solutions, but may also be used to remove organics selectively, e.g. furfural, from aqueous solutions. Membranes used for such applications are typically made of polydimethylsiloxane (PDMS), but researchers have also used the PDMS-containing triblock copolymer poly(styrene-block-dimethylsiloxane¬-block-styrene) (SDS). Pervaporation with a furfural-selective membrane may be used to extract furfural as it is produced and concentrate it, rather than dilute it as steam stripping and LLE do.

The objective of this investigation was to assess pervaporation as a method to extract furfural during its production. This was done by designing and constructing membrane reactors, comparing them to LLE-assisted reactors through experiments and simulations, and studying how Lewis acid catalysts can improve reaction and pervaporation compatibility and lead to the formation of novel products.

The feasibility of pervaporation as a means for in situ furfural extraction was studied in comparison to LLE and a reaction without extraction during batch-mode furfural production. Both LLE and pervaporation with a commercial PDMS membrane were found to improve furfural yield over the reaction without extraction, but pervaporation with PDMS yielded a product phase that was 6.6x as concentrated as that obtained with LLE. Additionally, switching the PDMS membrane with an SDS membrane resulted in similar furfural yields, but the product with SDS was 10x as concentrated as the LLE product. Furthermore, the amount of furfural extracted was qualitatively different for LLE- and pervaporation-assisted reactions: LLE was limited to 85%, the equilibrium distribution of furfural among the organic and aqueous, whereas the amount of furfural extracted by pervaporation increased monotonically over time, reaching as high as 67% during experiments. The reaction/pervaporation system was simulated in order to identify the full extent of the benefits of reaction with pervaporation. In the simulations, water lost from the reactor due to removal by pervaporation was replenished at the equivalent rate. The simulations revealed that as the reaction approached complete xylose conversion, both the PDMS and SDS membranes led to product concentrations greater than was possible with LLE, while extracting nearly all (>98%) of the furfural formed. Ultimately, pervaporation with the SDS membrane could produce a product phase with 33% greater furfural yield than that achievable by LLE.

The membrane-reactor design was revised to permit continuous, pervaporation-assisted reaction in both batch- and continuous-mode operation, with both reaction and pervaporation occurring at the same temperature. Batch-mode reactions were fed water, while continuous-mode reactions were fed an aqueous xylose solution. The reactions took place at a relatively low temperature of 90 °C, catalyzed by chromium (III) chloride (CrCl3), which contributed both Brønsted and Lewis acidity. Batch-mode reactions with varying rates of pervaporation revealed that furfural extraction had no effect on furfural yield under these conditions, but a moderate pervaporation rate did lead to an order-of-magnitude increase in furfural concentration relative to that obtained without pervaporation. Pervaporation was also found to retain all of the CrCl3 inside the reactor, demonstrating a simple way to separate product from homogeneous catalyst. This enabled continuous furfural production with only an initial charge of catalyst, in which an aqueous xylose solution was fed to the reactor while a furfural/water vapor was permeated from the reactor. The furfural permeability of the SDS membrane decreased over time during the course of reactions carried out at 90 °C due likely to interactions of soluble humins with the membrane. Experiments with the cross-linked PDMS membrane demonstrated that cross-linking of the membrane can inhibit this behavior and result in a much more stable furfural permeability. Additionally, cross-linking could lead to greater membrane thermal stability, permitting the pervaporation-assisted reaction at higher temperatures, which would benefit the chemistry by allowing extraction to have an impact on furfural yield.

Pervaporation-assisted furfural production with CrCl3 and sulfuric acid at 130 °C was then simulated. Reaction rate constants were measured at this temperature but in the absence of pervaporation. Pervaporation data collected at lower temperatures were extrapolated to represent a hypothetical membrane that could operate at 130 °C. Simulations of batch-mode reactions demonstrated that increasing the membrane-area-to-reactor-volume ratio, a, would lead to higher furfural yield and more furfural extracted, but also reduce the permeate furfural concentration, demonstrating a tradeoff between furfural production and concentration. Simulations of continuous-mode reactions showed that furfural concentration and selectivity were maximized at an intermediate value of a = 0.17 cm-1. Conversely, furfural production rate increased nearly linearly with a, indicating that the optimal value of a depends on process economics and not just technical considerations.

The Lewis acidity of CrCl3 was beneficial for reducing reaction temperature within the membrane stability limits, but Lewis acids, such as the Sn-containing zeolite Sn-BEA, have been shown to convert sugars (i.e., xylose and glucose) at a rate greater than the rate of formation of identified products, suggesting that additional, unidentified products are formed. Through extensive analytical chemistry, these products were determined to be hydroxyl-rich carboxylic acids and furanone esters that form by structural isomerization of the sugars, followed by dehydration, and constitute as much as 45% of the yield. These side-produced acids and esters may find use as monomers for the synthesis of biodegradable polyesters, which are often used for sutures, bone prostheses, and controlled drug delivery. This work demonstrates that Lewis acid catalysts are not only useful for bridging the gap between pervaporation-membrane limits and furfural production temperatures, but also for the formation of additional value-added chemicals.