Report

# Nonenzymatic Sugar Production from Biomass Using Biomass-Derived γ-Valerolactone

See allHide authors and affiliations

Science  17 Jan 2014:
Vol. 343, Issue 6168, pp. 277-280
DOI: 10.1126/science.1246748

## Renewable Breakdown Routine

In order to transform cellulose-containing biomass into liquid fuels such as ethanol, it is first necessary to break down the cellulose into its constituent sugars. Efforts toward this end have focused on chemical protocols using concentrated acid or ionic liquid solvents, and on biochemical protocols using cellulase enzymes. Luterbacher et al. (p. 277) now show that γ-valerolactone, a small molecule solvent that can itself be sourced renewably from biomass, promotes efficient and selective thermal breakdown of cellulose in the presence of dilute aqueous acid.

## Abstract

Widespread production of biomass-derived fuels and chemicals will require cost-effective processes for breaking down cellulose and hemicellulose into their constituent sugars. Here, we report laboratory-scale production of soluble carbohydrates from corn stover, hardwood, and softwood at high yields (70 to 90%) in a solvent mixture of biomass-derived γ-valerolactone (GVL), water, and dilute acid (0.05 weight percent H2SO4). GVL promotes thermocatalytic saccharification through complete solubilization of the biomass, including the lignin fraction. The carbohydrates can be recovered and concentrated (up to 127 grams per liter) by extraction from GVL into an aqueous phase by addition of NaCl or liquid CO2. This strategy is well suited for catalytic upgrading to furans or fermentative upgrading to ethanol at high titers and near theoretical yield. We estimate through preliminary techno-economic modeling that the overall process could be cost-competitive for ethanol production, with biomass pretreatment followed by enzymatic hydrolysis.

Biomass is emerging as a possible renewable alternative to petroleum-based resources in light of increasing environmental, economic, and political difficulties associated with fossil fuel extraction and use. Accordingly, biomass-derived sugars have been presented as intermediates for the production of renewable fuels (13) and chemicals (46). However, producing water-soluble carbohydrates from lignocellulosic biomass requires cleaving ether bonds in hemicellulose (primarily xylan) and cellulose (glucan) chains while minimizing further degradation of the resulting C5 and C6 sugars (primarily xylose and glucose) to insoluble degradation products. Unfortunately, in aqueous solutions containing low acid concentrations [<10 weight percent (wt %)] the high rate of sugar degradation reactions compared with polysaccharide depolymerization necessitates impractical reaction protocols for conversion of solid biomass, such as short residence times (10 ms to 1 min) at high temperatures (520 to 670 K), to obtain high yields of glucose (7). Because of the recalcitrance of crystalline cellulose to deconstruction, high yields at lower reaction temperatures can only be obtained by using concentrated mineral acid and/or ionic liquids (8, 9). However, recovery of the mineral acid is critical to the economics of the process, and the cost of ionic liquids can be prohibitive (810). Similarly, cellulase enzymes operating at temperatures of 320 K can achieve high glucose yields when converting cellulose rendered accessible by thermochemical pretreatment. However, the costs associated with producing these enzymes can be substantial compared with the value of the final product [with estimates of $0.32 to$1.47 per gallon of lignocellulose-derived ethanol (11, 12)].

Decoupling the residence times of the solid carbohydrate polymer from its soluble counterpart by flowing a solvent through a heated packed bed of biomass can minimize sugar degradation when using low acid concentrations (13). These systems are typically limited by their ability to produce concentrated soluble carbohydrate solutions (for example, 45 to 55% glucose yields when producing a 2 to 4 wt % sugar solution using 1 wt % H2SO4 in water) (13). In recent work, we have shown that liquid solutions of gamma-valerolactone (GVL) and water containing dilute concentrations of mineral acids (<0.1 M H2SO4) can completely dissolve lignocellulosic biomass (negligible amounts of solids are recovered with filtration by using a 0.22-μm filter or with centrifugation) and be used to produce levulinic acid and furfural (6, 14). In this work, we report a processing strategy that uses GVL/water solutions for producing soluble carbohydrates—a more versatile biomass platform—from corn stover, hardwood, and softwood in a flow-through reactor using a progressive temperature increase from 430 to 490 K (Fig. 1). In addition, we demonstrate the effective separation of these sugars from GVL into a concentrated aqueous phase that is compatible with subsequent upgrading by chemical or biological processes.

The results in Fig. 2, A and B, show the concentrations of soluble carbohydrate achieved in the GVL/water solvent as a function of solvent volume flowed through the reactor packed with corn stover (fig. S1), for a mixture containing 80 wt % GVL and 20 wt % water (80/20 GVL/water), and for another mixture containing 90 wt % GVL and 10 wt % water (90/10 GVL/water). Both solutions contain a low concentration of mineral acid: 5 mM H2SO4 (~0.05 wt %). In comparison, dilute acid pretreatment of biomass is typically carried out with at least 0.5 to 2 wt % H2SO4 (15). In both cases, the concentrations of C5 (xylose and xylo-oligomer) and C6 sugars (glucose and gluco-oligomer) reach maxima at temperatures between 430 and 470 K [where oligomers are defined as soluble carbohydrate species with more than one glucan or xylan monomer (supplementary materials)]. In contrast, when water is used as a solvent (Fig. 2C) the C6 concentration increases continuously with increasing temperature up to 490 K and potentially beyond. When an alternate organic solvent such as ethanol is used in place of GVL, the C6 sugar concentration shows a similar profile to that obtained with GVL, but lower by a factor of three (Fig. 2D). Therefore, the presence of GVL promotes cellulose deconstruction, most of which occurs below 480 K.

Increased deconstruction of biomass in the presence of GVL can be attributed to the complete solubilization of biomass solids, including the lignin fraction, observed during these experiments and previous work (6). In addition to providing a soluble lignin stream with potential as a feedstock for future upgrading, GVL prevents reprecipitation of lignin by-products on the surface of cellulose, which is a known phenomenon in water that decreases accessibility to the reactive cellulose surface (16). Water-insoluble solids corresponding to 95 and 84% of the original lignin were recovered for experiments conducted with 80/20 and 90/10 GVL/water, respectively (fig. S2). In addition, it appears that the presence of GVL plays a role in disrupting cellulose crystallinity—as suggested by x-ray diffraction measurements of pure cellulose solids isolated after treatment with GVL—showing an increased fraction of more reactive amorphous cellulose (fig. S3).

Biomass conversion in the GVL/water solvent system leads to a significant increase in the overall sugar yields as compared with that of conversion with water or water/ethanol as the solvent. Specifically, C5 recovery increases by 5 to 20 percentage points, and overall recovery of C6 sugars increases by two- to fourfold (Fig. 2E). These C5 and C6 yields of 89 and 80%, respectively, are similar to those achievable by using ionic liquids or enzymes, rather than those obtainable with water (9, 17, 18). With 80/20 GVL/water, 90 to 95% of polysaccharides are recovered as known soluble products when dehydration products such as furfural, 5-hydroxymethylfurfural (5-HMF), and levulinic acid [all potential GVL precursor molecules (6)] are included (fig. S4). In comparison, conversion in water leaves 50% of the C6 and 30% of the C5 fractions as unidentified solid products (fig. S4). Unlike reports of enzymatic processes (15), we found that C5 and C6 sugar yields in GVL/water are insensitive to biomass type because they are comparable for corn stover, maple wood, and loblolly pine (Fig. 2E). Furthermore, when using GVL/water as a solvent it is possible to recover >80% of the C5 and C6 sugars in separate volume fractions without additional separation processes (fig. S5). Depending on economic factors, producing separate C5 and C6 sugar streams can provide opportunities to implement separate upgrading processes.

Decreasing the temperature ramp duration from 2 hours to 30 min increases the concentrations of carbohydrates by reducing the volume of solvent flowed through the biomass by 70% while reducing sugar yields by less than 10% (Fig. 2F). When a 20 wt % biomass solution in 80/20 GVL/water with 0.15 M H2SO4 is treated for 1 hour at 390 K, most of the C5 sugars and lignin are solubilized, and the remaining solids can be placed in the flow-through reactor where the same 0.5-hour temperature ramp with 80/20 GVL/water and 5 mM H2SO4 is used. This approach decreases the solvent-to-solids ratio by >50% while maintaining similar C6 yields and lowering C5 yields by only 15% (Fig. 2F). Accordingly, the concentration of soluble C6 sugars is doubled as compared with that of our standard treatment. Moreover, when furfural is taken into account the overall conversion of xylan to known products remains above 90% for this biomass processing strategy (table S1).

The aqueous phase can be separated along with 75 to 91% of the carbohydrates from GVL/water solvent systems by addition of NaCl (Fig. 3A) (14) or liquid CO2 (Fig. 3B). This separation yields a total soluble carbohydrate concentration of up to 112 g/liter, depending on the ramp time and method used. In the case of GVL extraction by CO2, >70% of the nonextracted carbohydrates that remain in the organic phase can be recovered in a single reextraction from the organic phase after water addition (0.1 g water per g of extracted GVL) (fig. S6). Furthermore, if CO2-extracted GVL is recycled, then any recycled sugars will contribute to increased concentrations after biomass conversion. We have shown that GVL remains stable during recycle (supplementary materials). Subsequent extractions of the separated aqueous phase with CO2 lower the GVL concentration in water below 2 wt % while removing less than 4% of the carbohydrates and increasing their concentration to a total of 127 g/liter (Fig. 3B). This concentration corresponds to 65 to 85% of the highest concentrations obtained by enzymatic hydrolysis (150 to 200 g/liter) (17, 18) and is more than 8 times higher than concentrations that could have been obtained with pure water as a solvent (<15 g/liter). Moreover, the concentrated monomer solutions obtained from salt separation and CO2 extraction are clear (fig. S7), as opposed to the slurries obtained by using enzymatic hydrolysis or acidic-aqueous processing.

The C5 and C6 sugars recovered in the aqueous phase can be upgraded through catalytic dehydration to furfural and HMF (4). Furan selectivity is increased when these hydrophobic compounds are continuously extracted into an organic phase, such as 2-s-butyl-phenol (SBP) (19). Aqueous-phase modifiers such as NaCl (present by default in our salt-separated aqueous carbohydrate stream) and a Lewis acid catalyst such as AlCl3 further promote selectivity to furans by increasing their partitioning toward the organic phase and catalyzing carbohydrate isomerization, respectively (4, 19). Shown in Fig. 4A are the yields of furfural and 5-HMF obtained as a function of reaction time at 443 K through conversion of the soluble carbohydrates (monomers and oligomers) produced from corn stover by using the 2-hour temperature ramp. The separated aqueous phase was used without further treatment, except for the addition of AlCl3 and the presence of the SBP organic phase. The yields of 60 and 70% (Fig. 4A) for production of 5-HMF and furfural (>94% recovered in the SBP), respectively, are within 5% of yields reported from pure glucose and xylose, despite the presence of oligomers and other biomass by-products (19, 20).

Using liquid CO2 to extract GVL eliminates the use of salt and reduces the GVL concentration, both of which can inhibit microbial growth (fig. S8). In addition, the oligomers present in the recovered aqueous phase (Fig. 3) can be converted into monomers (preferable starting products for biological upgrading) in the acidic aqueous environment present in the aqueous phase (supplementary materials). In an aqueous monomer solution produced by using CO2 extraction and the 0.5-hour temperature ramp and then diluted by 75%, we observed robust growth of Saccharomyces cerevisiae PE2 (PE2), a nonevolved industrial yeast strain, with minimal media, and we achieved a yield of ethanol from glucose corresponding to 87% of the theoretical value (Fig. 4B). Ethanol yields that were 95% of theoretical, as well as fatty acid production, were both achieved by using more dilute salt-extracted feed (figs. S9 and S10). Because PE2 does not metabolize xylose, the ethanol titer obtained by using this CO2-extracted feed (19 g/liter) was below that of a potential titer of 31 g/liter that could be achieved if xylose had been converted at a similar yield [which has been demonstrated by using engineered yeast strains (21)] (Fig. 4B). Using a similar dilution of the more concentrated feed containing 127 g/liter carbohydrates, ethanol titers of 29 g/liter (86% yield) were obtained from glucose after 6 days of fermentation. Assuming xylose conversion at similar yields, a potential titer of 48 g/liter would have been reached. Efforts are currently underway to study new strains and/or evolve one to grow robustly in the corresponding undiluted hydrolysate, which could lead to ethanol concentrations of 60 g/liter. An industrial scenario recently published by the National Renewable Energy Laboratory (NREL) assumed ethanol titers ~50 g/liter (22). Using our current carbohydrate recovery yields and assuming that an undiluted carbohydrate stream can be used for fermentation, a preliminary techno-economic model suggests that such a process could produce ethanol at a minimum selling price (MSP) of $4.87/gallon of gasoline equivalent (GGE), versus$5.13/GGE for the NREL scenario (Fig. 4C) (22). Most of the savings are related to the absence of enzymes (a description of the model and further discussion on the effect of enzyme cost and fermentation titers are available in the supplementary materials). Using more optimistic estimates of enzyme costs and 10% higher overall ethanol yields from biomass (12) leads to a decrease in the MSP of ethanol to \$4.06/GGE (details are available in the supplementary materials), which is slightly below our estimate for the GVL process. Although these preliminary economic comparisons are dependent on various assumptions, the results suggest that our proposed approach could become an economically competitive alternative to current biomass-derived carbohydrates production schemes.

## Supplementary Materials

www.sciencemag.org/content/343/6168/277/suppl/DC1

Materials and Methods

Supplementary Text

Figs. S1 to S15

Tables S1 to S7

References (2328)

## References and Notes

1. Acknowledgments: This work was supported by the Great Lakes Bioenergy Research Center (www.glbrc.org), supported by the U.S. Department of Energy, through Cooperative Agreement DE-FC02-07ER64494 between the Board of Regents of the University of Wisconsin System and the U.S. Department of Energy and the National Science Foundation (CBET-1149678). J.S.L. acknowledges support from the Swiss National Science Foundation. The authors thank S. Ralph (Forests Products Laboratory) for providing loblolly pine. Patents have been filed by the Wisconsin Alumni Research Foundation (WARF) involving the use of GVL as a solvent for biomass processing (application nos. 13/736550 and 13/918097), and D.M.A. and J.A.D. are involved with a start-up company (Glucan Biorenewables) that is working in this area.
View Abstract