Article pubs.acs.org/IECR
Two-Stage Acid-Catalyzed Conversion of Carbohydrates into Levulinic Acid Troy Runge*,† and Chunhui Zhang‡,† †
Department of Biological System Engineering, University of Wisconsin-Madison, Madison, Wisconsin 53706, United States State Key Laboratory of Pulp and Paper Engineering, South China University of Technology, Guangzhou, 510640, China
‡
ABSTRACT: In this paper, we report on a two-stage acid-catalytic conversion of carbohydrates from hybrid poplar wood chips into levulinic acid (LA), a renewable platform chemical which can be used for fuels and chemicals. It was hypothesized that under the harsh acid conditions utilized during LA production, the pentose fraction in biomass would first form furfurals which would then polymerize with saccharides forming humins. These reactions would both reduce any potential value from the pentose fraction as well as lower the levulinic acid yield. To test this idea, a two-stage conversion process was designed starting with a mild acid extraction to remove the majority of the pentoses while maintaining the hexose sugars in a solid form utilizing previously described optimized conditions. LA was then produced by subjecting the extracted solids to a second more severe step. The temperature, time, acid concentration, and a liquor-to-wood ratio were varied and modeled to find the optimal conditions. The best conditions were high acid concentration, high temperature, and low substrate consistency, which produced a maximum molar yield of 66% based on the hexose content or 17.5 wt % based of the initial biomass. A comparison of this twostage process to a single-stage process without the pentose extraction was performed both with biomass and model compounds, which indicated no pentose or furfural in the product stream and a marked molar yield decrease in LA yield from the presence of pentose sugars, validating our initial hypothesis.
1. INTRODUCTION Sustainability concerns and an increase in crude oil prices in the past decade have increased industrial interest for renewable chemical and fuels. Biomass is an obvious choice to replace petroleum feedstock, with lignocellulosic biomass being one of the most promising as it offers the potential to provide sustainable sugar streams from a variety of materials including agricultural and forest residuals as well as high-yielding bioenergy crops such as switchgrass, miscanthus, and hybrid poplar. One attractive option for the conversion of lignocellulosic biomass into renewable fuel and chemical production is the production of levulinic acid (LA), a very versatile platform chemical widely used in the cosmetic, food, and medicinal industries.1 A number of approaches have been reported for LA synthesis, including by hydrolysis of acetyl succinate esters,2 acid hydrolysis of furfuryl alcohol,3 and oxidation of ketones with ozone.4 However, these methods frequently form large amounts of side products and intractable materials or require expensive starting feedstock. The most widely used approach is the dehydrative treatment of biomass or carbohydrates with acid.5−12 The described processes do not attempt to separate the pentoses prior to the harsh acid dehydration with the exception of the Biofine process, which does remove furfural products formed prior to hydroxymethylfurfural conversion to LA. This process reports yield of 70−80% of theoretical, due to its efficient reactor system and the use of polymerization inhibitors.12 Furfural is another byproduct of biomass undergoing acid dehydration, formed by the reactions of pentose sugars in the biomass, such as xylose. There currently is a market for furfural and its derivatives including furfuryl alcohol, which is used as a © 2012 American Chemical Society
binder for foundry forms. The chemistry of the furfural production involves the acid catalyzed hydrolysis of the hemicellulosic pentose fractions of biomass and consecutive cyclodehydration of the pentose monomers, with xylose being the predominant pentose in most biomass. Conventional mineral acids, such as sulfuric acid, are generally used as the catalysts.13,14 Recently, several solid catalysts have been introduced to catalyze xylose to furfural at a yield comparable to H2SO4.15−17 The theoretical yield of LA from C6-sugars is 100 mol % or 64.5 wt % due to the coproduction of formic acid.18 Commonly, LA yields of two-thirds or less than the theoretical value are attained, due to side reactions that form undesired black insoluble-materials called humins. The harsh acidic dehydration conditions utilized to hydrosolyze and dehydrate hexoses to form LA from hexoses will also hydrolyze and dehydrate pentoses in the biomass to furfural, as shown in Figure 1. In fact, the conditions are harsh enough that the furfural will further degrade and react with soluble saccharides, forming solid humins. This loss of pentose saccharides eliminates a value stream that could be utilized for the production of other biobased materials, such as furfural and its derivatives. To attempt to solve this issue, a pentose extraction prior to levulinic acid production was explored. By fractionating the majority of the pentoses from the hexoses, it may be possible to maintain or even increase the LA production yield and allow Received: Revised: Accepted: Published: 3265
September 20, 2011 January 30, 2012 February 2, 2012 February 2, 2012 dx.doi.org/10.1021/ie2021619 | Ind. Eng. Chem. Res. 2012, 51, 3265−3270
Industrial & Engineering Chemistry Research
Article
Figure 1. Hydrolysis and degradation of cellulose and hemicellulose.
Figure 2. Simplified reaction scheme for the conversion of carbohydrates of hybrid poplar into levulinic acid and liquid/solid fuel and chemicals.
poplar19 and consisted of a temperature of 160 °C, a 1.0 wt % sulfuric acid concentration, at 60 min, and a liquor-to-wood ratio of 6. The extracted solids were reacted in a second stage after grinding in a Wiley mill to pass a 3 mm screen. Wood meal and H2SO4 were measured into temperature and pressure resistant glass tubes, which were sealed and placed into the Parr reactor (either 2 L or 20 L). The reaction was controlled to temperature, reaction time, sulfuric acid concentration, and substrate consistency (wood to liquor ratio), after which solids were washed out and separated from the liquor under centrifugation at 4000 rpm for 10 min. The supernatant was neutralized, centrifuged, and filtered through 0.22 μm nylon filters. Characterization of the extraction and reactions to obtain sugar fractionation and levulinic acid yields were performed using a Dionex HPLC system (ICS-3000) as previously described.19
the production of an additional value stream from pentose sugars. The simplified scheme for conversion of hybrid poplar carbohydrates into LA and liquid/solid fuel is illustrated in Figure 2. The first stage involves a dilute acid prehydrolysis of pentosan, after which the pentose-rich liquor and the hexoserich solid were separated. Then more harsh conditions were applied to the second stage to convert the hexose-rich solids into LA. Previous work19 describes the optimization of the pentose extraction while this paper focuses on the two-stage sulfuric acid-catalyzed conversion of hexoses to LA.
2. MATERIALS AND METHODS Hybrid poplar, a fast growing energy plantation crop, is among those plants that offer high energy biomass and incredible improvements on greenhouse gas emissions. Moreover, hybrid poplar contains 14−18% (on wood) of pentosan which is mainly composed of xylan and would allow us to test our hypothesis.20 NM-6 hybrid poplar, Populus maximowiczii x nigra, was specifically utilized for these experiments, which was harvested from Northern Wisconsin at 10 years of age and seasoned for 3 months. The wood log was hand debarked, chipped and screened, air-dried, and cold stored. Chemicals were purchased from Sigma Aldrich and used without modification. The first stage, dilute acid extraction, was performed using a dual vessel pressurized reactor with heated liquor circulation. The extraction conditions were previously researched and optimized to maximize the pentosan extraction from the hybrid
3. RESULTS AND DISCUSSION 3.1. Biomass Characterization. Samples from the hybrid poplar chips were analyzed for saccharides, lignin, and ash to characterize the material. Additionally, since the acidolysis used in the sugar analysis creates organic acid and furans, the major hydrolysis/degradation products of formic acid (FA) acetic acid (AA), levulinic acid (LA), 5-hydroxy-methyl-furfural (HMF), and furfural (F) were also characterized to provide a better view on the biomass composition. The results shown in Table 1 are as expected for a Poplar species. The principal hemicellulose 3266
dx.doi.org/10.1021/ie2021619 | Ind. Eng. Chem. Res. 2012, 51, 3265−3270
Industrial & Engineering Chemistry Research
Article
45.1 wt %, and the mass loss of 64.5 wt % of converting levulinic acid from hexose due to the dehydration.18 Therefore a 100% yield is equivalent to a 29.1 wt % on an initial biomass basis. Data analysis on the yield data was done using the DOE (Design of Experiments) routine in MODDE 7.0 software by Umetrics AB, with the statistical model fitting results shown in Table 4. Partial least squares (PLS) was used to estimate the coefficients of the terms in the model. Typically, a P-value less than 0.0500 indicates the model term is significant, while that greater than 0.1000 indicates the model term is not significant. From Table 4, it could be seen that the first order main effects of four variables were highly significant as evident from their Pvalues. This suggested that the four variables were directly related to the production of levulinic acid. The second order main effects of temperature and acid concentration were also significant, with the P-value being 0.0181. The predictive power for new experimental conditions in a PLS model was determined through the Q2 value, which is based on the prediction residual sum of squares (PRESS). A Q2 larger than zero indicates that the variable is significantly predictive with Q2 values of 0.7 or larger, taken to mean the model has good predictive ability and will have small prediction errors. A R2 value, which is the percent variation of the measured response explained by the model, was also calculated. Q2 and R2 of the model were 0.82 and 0.92, respectively, which indicates that the model has good predictive ability and that there was a good fit between the experimental concentration of levulinic acid and the kinetic model for a broad range of reaction conditions. The model fitting yielded the following regression equation, which was an empirical relationship between levulinic acid yield and the variable, shown in eq 1. The predicted yield was plotted against the measured yield in a parity plot in Figure 3. The regression equation’s goodness of fit was shown to be quite high with a linear fit, no outliers, and an R2 statistic of 0.923.
Table 1. Characterization of Poplar Biomass arabinose
galactose
glucose
xylose
mannose
0.30% formic acid 4.50%
0.50% acetic acid 3.70% ash
40.70% LAa
13.30% HMFa
3.00% furfuralb
2.20%
a
saccharides total 57.80% degradation products total 11.30% lignin and ash total 28.00% 97.10%
0.60% 0.70% acid-soluble lignin
1.80% klason lignin 3.60% 22.20% total yield of measured components
Calculated on C6 sugars. bCalculated on C5 sugars.
species present in this hybrid poplar is xylan, constituting about 15.0% of the dry weight. 3.2. Extraction of Pentosans. Following previous experimental work,19 mild sulfuric acid extraction conditions of 160 °C, 60 min, 1.0 wt % H2SO4, and a 6:1 liquor to wood ratio were employed. These conditions liberated 85.0% of the pentosan content into the liquid fraction while only removing 8.0 wt % of hexosan content as shown in Table 2. The Table 2. Characterization of the Extraction Step Solid and Liquid Fractions Extraction Step: Solid Fraction (wt % of Original Biomass) arabinose 0.00% formic acid 1.42%
galactose
glucose
xylose
0.18% 41.24% 1.66% HMFa acetic LAa acid 1.11% 0.73% 0.55% Extraction Step: Liquid Fraction (wt
mannose
45.32% degradation products total 0.51% 4.32% % of Original Biomass)
arabinose
galactose
glucose
xylose
mannose
0.34% formic acid 3.81%
0.38% acetic acid 2.83%
1.74% LAa
9.22% HMFa
1.26% furfuralb
0.00%
0.16%
1.99%
a
saccharides total
2.24% furfuralb
saccharides total 12.94% degradation products total 8.79%
Calculated on C6 sugars. bCalculated on C5 sugars.
YLA(t %) = 31.36 + 7.14T + 8.49CA + 2.22t + 3.98R − 1.22TCA + 0.83CAR + 0.69Rt
calculation considers both the directly measured sugar content and the degradations products which occur during the testing hydrolysis step. The data in Table 2 indicates this operation effectively fractionates the biomass into a pentose-rich liquid fraction and a hexose-rich solid that can be further processed into LA. 3.3. LA Production from Extracted Solids. The extracted hexose-rich chips were washed and ground in a Wiley mill to pass a 3 mm screen. The conditions were of a temperature range of 170−190 °C, a sulfuric acid range of 1.0−5.0 wt %, during a time range of 20−50 min, and a wood-to-liquor ratio range of 1/6−1/10 using the 2 L PARR reactor. The wood meal, H2SO4, and water were filled into temperature and pressure resistant glass tubes inside the PARR reactor following a combined experimental design of a Box-Behnken and central composite face. Four variables, temperature, acid concentration, ratio, and duration time, were used to study the response pattern and determine the optimum combination of variables. The statistical treatment combinations of the test variables along with the measured response values expressed as levulinic acid yield corresponding to each combination are summarized in Table 3. The yield calculated is a theoretical yield based on the initial biomass hexose content. This yield considers the experiments’ hybrid poplar chips hexose content, which was
(1)
Using the optimizer of the MODDE software, a predicted 60.3% on the theoretic yield could be obtained under the conditions of 190 °C, 50 min, 5 wt % H2SO4, and a wood-toliquor ratio of 10. This is quite promising as it in range or higher than other reported processes (45.6−68.8%)21−23 and produces a pentose stream that can be utilized to create other products and value to the biomass processing plant. 3.4. LA Production from Nonextracted Solids. Utilizing the optimized conditions, replicates were prepared to validate the model and determine the repeatability of the predicted yields. Additionally, it was desired to see the impact of extraction on yield, so control experiments of nonextracted biomass were also included. Utilizing the optimized conditions, experiments were performed on three replicates of nonextracted biomass at a 2 L scale, two replicates of extracted biomass at a 2 L scale, and three replicates of extracted biomass at a 20 L scale. The resulting formed LA was measured, and the theoretical yield was calculated and plotted in Figure 4. From the figure it may be concluded that the extracted biomass has a significantly higher LA yield than the nonextracted biomass. These yields are on the same basis and include the minor losses of hexoses that occur in the extraction step. The second implication from 3267
dx.doi.org/10.1021/ie2021619 | Ind. Eng. Chem. Res. 2012, 51, 3265−3270
Industrial & Engineering Chemistry Research
Article
Table 3. Experimental Design and Results of LA Production from Extracted Chips
a
T (°C)
CA (wt %)
L/W (ratio)
t (min)
LA yield (%)a
T (°C)
CA (wt %)
L/W (ratio)
t (min)
LA yield (%)a
170 170 170 170 170 170 170 170 170 170 170 170 170 170 170 180 180 180 180 180 180 180 180 180 180
1 5 1 5 3 3 5 1 3 3 1 5 1 5 3 3 1 3 5 3 1 5 3 3 3
6 6 10 10 8 8 8 8 6 10 6 6 10 10 8 8 8 10 8 6 8 8 6 10 8
20 20 20 20 20 35 35 35 35 35 50 50 50 50 50 20 20 20 20 20 35 35 35 35 35
7.12 26.9 10.1 37.4 18.5 18.1 32.7 9.49 19.4 25.7 7.93 28.3 15.0 47.8 23.8 27.0 14.9 33.4 39.5 20.5 19.5 41.2 25.9 33.4 36.43
180 180 180 180 180 180 180 180 180 190 190 190 190 190 190 190 190 190 190 190 190 190 190 190
5 5 1 1 3 1 3 5 3 1 5 1 5 3 3 1 3 3 5 1 5 1 5 3
6 10 6 10 8 8 6 8 10 6 6 10 10 8 8 8 6 10 8 6 6 10 10 8
35 35 35 35 50 50 50 50 50 20 20 20 20 20 35 35 35 35 35 50 50 50 50 50
37.0 51.0 11.7 15.1 31.6 25.9 29.4 44.8 47.9 25.9 41.0 32.1 50.4 35.6 35.7 41.0 35.9 52.3 52.9 25.6 46.7 38.4 54.1 39.9
LA yield was calculated as percent of the theoretical based on the initial biomass hexose content.
Table 4. Significance of Regression Coefficients for LA Yielda constant temp CA t R temp × CA CA × R t×R
coefficient
std error
significance level, P value
31.36 7.13 8.49 2.21 3.95 −1.22 0.83 0.69
0.52 0.53 0.53 0.53 0.53 0.50 0.50 0.50
0.00 2.75 × 10−17 4.62 × 10−20 0.000128 2.18 × 10−9 0.0181 0.101 0.173
a
N = 52, Q2 = 0.817, degree of freedom = 44, R2 = 0.923, Y-miss = 0, R2 adj. = 0.911, RSD = 3.7658, confidence level = 0.95.
the figure is that the larger reactor produced higher variability and a slightly lower yield. The authors believe this was due to inconsistent mixing and heating from the larger reactor design, which is believed to be important in LA yield optimization. The final result from the figure was the validation of the statistical model yield, which predicted a 61% yield. In fact the 2 L reactor was able to provide higher yields with 66% yield measured, suggesting other improvements can be made beyond the model by adjusting variables related to reactor design. 3.5. Model Compound Reactions. The hypothesis for this work is that during the production of levulinic acid from biomass, any pentose sugars will be converted into furfural which will react under the harsh conditions to form humins and other low value products, removing the possibility of capturing value from this component. Although the biomass studies support this hypothesis, further verification was sought using a simple model compound study. The study consisted of batch reactions of glucose, serving as the hexose sugar, with or
Figure 3. Parity plot for the experimental and modeled yield of LA.
without xylose, serving as the pentose sugar. Additionally a set of reactions were performed using cellulose from Whatman ashless filter paper, serving as the hexose sugar source, with or without extracted beechwood xylan, purchased from Sigma Aldrich. Three replicates of the sugars were dehydrated at the biomass optimal conditions of 190 °C, 50 min, 5 wt % H2SO4, and a wood-to-liquor ratio of 10. The results, shown in Figure 5, are similar to the biomass experiments with yields decreasing with the addition of the pentose sugar, although the decrease was only significant for the cellulose sample. The glucose sample yielded a significantly lower yield than the cellulose 3268
dx.doi.org/10.1021/ie2021619 | Ind. Eng. Chem. Res. 2012, 51, 3265−3270
Industrial & Engineering Chemistry Research
■
Article
AUTHOR INFORMATION
Corresponding Author
*E-mail:
[email protected]. Notes
The authors declare no competing financial interest.
■
ACKNOWLEDGMENTS The authors are grateful to the Wisconsin Bioenergy Initiative at the University of Wisconsin for financial support of this work. Additionally, the authors want to thank Dr. JunYong Zhu and Dr. Xiaolin Luo of the USDA Forest Products Lab for his assistance with this research.
Figure 4. Comparison of extracted and nonextracted LA yields with 95% CI.
■
REFERENCES
(1) Bozell, J. Production of levulinic acid and use as a platform chemical for derived products. Resour., Conserv. Recycl. 2000, 28 (3−4), 227−239. (2) Farnleitner, L.; Stueckler, H.; Kaiser, H.; Kloimstein, E. Process for the preparation of storage-stable levulinic acid. U.S. Patent 5,189,215, February 23, 1993. (3) Hsu, C. C.; Hasar, D. W. Process for the manufacture of levulinic acid and esters. U.S. Patent 4,236,021, November 25, 1980. (4) Edwards , W. B., III. Preparation of oxycarboxylic acids. U.S. Patent 4,612,391, September 16, 1986. (5) Lourvanij, K.; Rorrer, G. Dehydration of glucose to organic acids in microporous pillared clay catalysts. Appl. Catal., A 1994, 109 (1), 147−165. (6) Hayes, D. J.; Fitzpatrick, S.; Hayes, M. H. B.; Ross, J. R. H. The Biofine ProcessProduction of Levulinic Acid, Furfural, and Formic Acid from Lignocellulosic Feedstock. In Biorefineries-Industrial Processes and Products: Status Quo and Future Directions; Kamm, B., Gruber, P. R., Kamm, M., Eds.; Wiley-VCH Verlag GmbH: Weinheim, Germany, 2008. (7) Girisuta, B.; Janssen, L. P. B. M.; Heeres, H. J. Green Chemicals: A Kinetic Study on the Conversion of Glucose to Levulinic Acid. Chem. Eng. Res. Des. 2006, 84 (5), 339−349. (8) Girisuta, B.; Janssen, L. P. B. M.; Heeres, H. J. Kinetic Study on the Acid-Catalyzed Hydrolysis of Cellulose to Levulinic Acid. Ind. Eng. Chem. Res. 2007, 46 (6), 1696−1708. (9) Tarabanko, V.; Chernyak, M.; Aralova, S.; Kuznetsov, B. Kinetics of levulinic acid formation from carbohydrates at moderate temperatures. React. Kinet. Catal. Lett. 2002, 75 (1), 117−126. (10) Efremov, A. A.; Pervyshina, G. G.; Kuznetsov, B. N. Production of levulinic acid from wood raw material in the presence of sulfuric acid and its salts. Chem. Nat. Compd. 1998, 34 (2), 182−185. (11) Mosier, N.; Ladisch, C.; Ladisch, M. Characterization of acid catalytic domains for cellulose hydrolysis and glucose degradation. Biotechnol. Bioeng. 2002, 79 (6), 610−618. (12) Fitzpatrick, S. W. The Biofine Technology: A “Bio-Refinery” Concept Based on Thermochemical Conversion of Cellulosic Biomass. In Feedstocks for the Future: Renewables for the Production of Chemicals and Materials ; ACS Symposium Series 921; American Chemical Society: Washington, DC, 2006; pp 271−287. (13) Montané, D. High-temperature dilute-acid hydrolysis of olive stones for furfural production. Biomass Bioenergy 2002, 22 (4), 295− 304. (14) Antal, M. J.; Leesomboon, T.; Mok, W. S.; Richards, G. N. Mechanism of formation of 2-furaldehyde from D-xylose. Carbohydr. Res. 1991, 217, 71−85. (15) Lima, S.; Pillinger, M.; Valente, A. Dehydration of d-xylose into furfural catalysed by solid acids derived from the layered zeolite Nu6(1). Catal. Commun. 2008, 9 (11−12), 2144−2148. (16) Dias, A. S.; Pillinger, M.; Valente, A. A. Dehydration of xylose into furfural over micro-mesoporous sulfonic acid catalysts. J. Catal. 2005, 229 (2), 414−423. (17) Dias, A. S.; Lima, S.; Carriazo, D.; Rives, V.; Pillinger, M.; Valente, A. A. Exfoliated titanate, niobate and titanoniobate nano-
Figure 5. Comparison of model compound LA yields with 95% CI.
which has been previously reported.7,8 The cellulose’s higher yield is presumably due to the cellulose allowing a gradual introduction of monomeric hexose saccharides, which in effect further increases the dilution factor. The furfural content of the products were also measured, and as in the case of the biomass only trace amounts were detected for the experiments that included xylose (
