Phase Modifiers Promote Efficient Production of Hydroxymethylfurfural from Fructose

See allHide authors and affiliations

Science  30 Jun 2006:
Vol. 312, Issue 5782, pp. 1933-1937
DOI: 10.1126/science.1126337


Furan derivatives obtained from renewable biomass resources have the potential to serve as substitutes for the petroleum-based building blocks that are currently used in the production of plastics and fine chemicals. We developed a process for the selective dehydration of fructose to 5-hydroxymethylfurfural (HMF) that operates at high fructose concentrations (10 to 50 weight %), achieves high yields (80% HMF selectivity at 90% fructose conversion), and delivers HMF in a separation-friendly solvent. In a two-phase reactor system, fructose is dehydrated in the aqueous phase with the use of an acid catalyst (hydrochloric acid or an acidic ion-exchange resin) with dimethylsulfoxide and/or poly(1-vinyl-2-pyrrolidinone) added to suppress undesired side reactions. The HMF product is continuously extracted into an organic phase (methylisobutylketone) modified with 2-butanol to enhance partitioning from the reactive aqueous solution.

We are entering an era of diminishing availability of petrochemical resources used to produce the energy and chemical materials needed by society. Abundant biomass resources are a promising alternative for the sustainable supply of valuable intermediates (such as alcohols, aldehydes, ketones, and carboxylic acids) to the chemical industry for production of drugs and polymeric materials. In this context, the high content of oxygenated functional groups in carbohydrates—the dominant compounds in biomass—is an advantage, in contrast to the drawbacks of such functionality for the conversion of carbohydrates to fuels. However, efficient processes must be developed for the selective removal of excess functional groups and the modification of others to create the desired products.

Here, we present a strategy for the selective dehydration of fructose to produce HMF, thereby providing a cost-effective route for the synthesis of this valuable chemical intermediate. Indeed, HMF and its ensuing 2,5-disubstituted furan derivatives can replace key petroleum-based building blocks (1). For example, HMF can be converted to 2,5-furandicarboxylic acid (FDCA) by selective oxidation, and Werpy and Petersen (2) and Pentz (3) have suggested that FDCA can be used as a replacement for terephthalic acid in the production of polyesters such as polyethyleneterephthalate (2) and polybutyleneterephthalate. They have also suggested that the reduction of HMF can lead to products such as 2,5-dihydroxymethylfuran and 2,5-bis(hydroxymethyl)tetrahydrofuran, which can serve as alcohol components in the production of polyesters, thereby leading to completely biomass-derived polymers when combined with FDCA. In addition, HMF can serve as a precursor in the synthesis of liquid alkanes to be used, for example, in diesel fuel (4).

Unfortunately, as noted by various authors (58), the industrial use of HMF as a chemical intermediate is currently impeded by high production costs. Early work showed that HMF could be produced in high yields using high-boiling organic solvents [such as dimethylsulfoxide (DMSO), dimethylformamide, and mixtures of polyethyleneglycol with water] over various catalysts, including sulfuric acid and sulfonic acid resins. However, this approach necessitates difficult and energy-intensive isolation procedures (6, 913). In pure water, fructose dehydration is generally nonselective, leading to many by-products besides HMF (14). Recent advances have shown improved results in pure water or in water-miscible solvent systems (such as acetonitrile or acetone), but only with the use of low initial fructose concentrations, which inevitably generate low HMF concentrations (1, 10, 15, 16). Biphasic systems, in which a water-immiscible organic solvent is added to extract continuously the HMF from the aqueous phase, have also been investigated with the use of mineral acid or zeolite catalysts at temperatures greater than 450 K (6, 1721). However, poor HMF partitioning into the organic streams used in these studies necessitated large amounts of solvent, thereby requiring large energy expenditures to purify the diluted HMF product (22).

Fructose is dehydrated in the presence of an acid catalyst to produce HMF and various by-products (Fig. 1). Although evidence exists supporting both the open-chain and the cyclic fructofuransyl intermediate pathways (20, 23), it is clear that the reaction intermediates and the HMF product degrade by means of processes such as fragmentation, condensation, rehydration, reversion, and/or additional dehydration reactions. We used a modified biphasic system to address key HMF production limitations. Specifically, we improved the HMF selectivity (defined as the moles of HMF produced divided by the moles of fructose reacted) of the acid-catalyzed dehydration of concentrated [30 to 50 weight % (wt %)] fructose feeds by adding modifiers to both phases. The reactive aqueous phase containing the catalyst and the sugar was modified with polar aprotic molecules [DMSO or 1-methyl-2-pyrrolidinone (NMP)] and a hydrophilic polymer [poly(1-vinyl-2-pyrrolidinone) (PVP)]. The water-immiscible organic phase [methylisobutylketone (MIBK)] used during the reaction to extract HMF was modified with 2-butanol. The ratio of relative volumes of the organic and aqueous phases in the reactor (Vorg /Vaq), as well as the ratio of the HMF concentration in the organic layer to that in the aqueous layer (defined as the extraction ratio, R) proved to be important variables in the process. Upon completion of the dehydration reaction, both phases can be separated for efficient product isolation. Although various acid catalysts can be used to perform the dehydration reaction, HCl showed the highest HMF selectivity of the common mineral acid catalysts (Table 1, runs 5, 8, and 40 to 43).

Fig. 1.

Schematic representation of reaction pathways for acid-catalyzed dehydration of fructose. Structures in brackets correspond to representative species. Detailed chemistry is presented in (20, 23).

Table 1.

Results for acid-catalyzed dehydration of fructose. Runs 1 to 27 were carried out at 453 K for 2.5 to 3 min with 0.25 M HCl aqueous phase solutions; runs 28 to 39 were carried out at 363 K for 8 to 16 hours with an acidic ion-exchange resin at a 1:1 w/w fructose:resin ratio. Aqueous phase and organic phase compositions are reported as w/w ratios. Conversion is defined as the ratio of fructose consumed to fructose added initially. R = [HMF]org/[HMF]aq. Standard runs for HCl, H2SO4, and H3PO4 catalysts used 1.5 g of aqueous phase and 1.5 g of extracting solvent. Runs marked with an asterisk used 3 g of extracting solvent. Runs for resin catalyst used 5.0 g of aqueous phase and 5.0 g of extracting solvent. Vorg/Vaq was measured upon completion of reaction.

Run no.Aqueous phase compositionOrganic phase compositionConversion (%)Selectivity (%)[HMF]aq (g/ml)[HMF]org (g/ml) R Vorg/Vaq
30 wt % fructose with HCl catalyst
1 Water none 50 51 0.060 - 0.00 0.00
2 Water MIBK 91 60 0.056 0.050 0.90 1.51
3* Water MIBK 75 73 0.035 0.033 0.96 3.13
4 Water 7:3 MIBK:2-butanol 68 70 0.033 0.054 1.65 1.56
5* Water 7:3 MIBK:2-butanol 86 80 0.026 0.045 1.73 3.68
6 8:2 Water:DMSO MIBK 94 67 0.077 0.050 0.66 1.41
7 8:2 Water:DMSO 7:3 MIBK:2-butanol 80 75 0.050 0.064 1.30 1.49
8* 8:2 Water:DMSO 7:3 MIBK:2-butanol 87 82 0.034 0.046 1.39 3.65
9 7:3 Water:PVP MIBK 74 66 0.055 0.041 0.81 1.56
10 7:3 Water:PVP 7:3 MIBK:2-butanol 62 76 0.042 0.047 1.25 1.57
11* 7:3 Water:PVP 7:3 MIBK:2-butanol 79 82 0.030 0.041 1.44 3.83
12 7:3(8:2 Water:DMSO):PVP MIBK 79 75 0.071 0.047 0.71 1.52
13 7:3(8:2 Water:DMSO):PVP 7:3 MIBK:2-butanol 82 83 0.063 0.065 1.12 1.62
14* 7:3(8:2 Water:DMSO):PVP 7:3 MIBK:2-butanol 89 85 0.043 0.046 1.17 3.99
50 wt % fructose with HCl catalyst
15 Water none 51 28 0.064 - 0.00 0.00
16 Water MIBK 65 47 0.049 0.051 1.11 1.80
17 Water 7:3 MIBK:2-butanol 71 59 0.049 0.079 1.73 1.91
18* Water 7:3 MIBK:2-butanol 88 72 0.045 0.069 1.55 4.66
19 8:2 Water:DMSO MIBK 71 57 0.076 0.060 0.86 1.69
20 8:2 Water:DMSO 7:3 MIBK:2-butanol 80 63 0.077 0.085 1.19 1.87
21* 8:2 Water:DMSO 7:3 MIBK:2-butanol 91 74 0.059 0.072 1.30 4.87
22 7:3 Water:PVP MIBK 85 56 0.074 0.060 0.80 1.72
23 7:3 Water:PVP 7:3 MIBK:2-butanol 77 61 0.076 0.081 1.19 1.85
24* 7:3 Water:PVP 7:3 MIBK:2-butanol 90 77 0.062 0.070 1.22 5.15
25 7:3(8:2 Water:DMSO):PVP MIBK 77 61 0.095 0.066 0.77 1.85
26 7:3(8:2 Water:DMSO):PVP 7:3 MIBK:2-butanol 72 62 0.068 0.074 1.25 1.89
27* 7:3(8:2 Water:DMSO):PVP 7:3 MIBK:2-butanol 92 77 0.076 0.070 1.03 5.11
10 wt % fructose with ion-exchange resin catalyst
28 Water MIBK 75 44 0.010 0.011 1.02 1.32
29 Water MIBK 17 43 0.0021 0.0024 1.15 1.29
30 Water 7:3 MIBK:2-butanol 61 60 0.0086 0.014 1.61 1.31
31 8:2 Water:DMSO MIBK 84 47 0.015 0.012 0.79 1.26
32 8:2 Water:DMSO MIBK 19 80 0.0052 0.0045 0.87 1.24
33 8:2 Water:DMSO 7:3 MIBK:2-butanol 74 68 0.015 0.017 1.18 1.24
34 7:3 Water:PVP MIBK 74 63 0.018 0.013 0.79 1.43
35 7:3 Water:PVP 7:3 MIBK:2-butanol 70 65 0.015 0.015 1.04 1.46
36 7:3(8:2 Water:DMSO):PVP MIBK 80 71 0.026 0.013 0.54 1.38
37 7:3(8:2 Water:DMSO):PVP 7:3 MIBK:2-butanol 76 77 0.020 0.019 1.03 1.43
30 wt % fructose with ion-exchange resin catalyst
38 7:3(8:2 Water:DMSO):PVP MIBK 89 60 0.066 0.041 0.66 1.65
39 7:3(8:2 Water:DMSO):PVP 7:3 MIBK:2-butanol 83 65 0.053 0.051 1.07 1.74
30 wt % fructose with H2SO4 catalyst
40* Water 7:3 MIBK:2-butanol 80 66 0.022 0.035 1.63 3.54
41* 8:2 Water:DMSO 7:3 MIBK:2-butanol 85 71 0.029 0.040 1.35 3.59
30 wt % fructose with H3PO4 catalyst
42* Water 7:3 MIBK:2-butanol 65 65 0.016 0.029 1.89 3.47
43* 8:2 Water:DMSO 7:3 MIBK:2-butanol 51 76 0.016 0.025 1.58 2.95

We performed experiments using 0.25 M HCl at 453 K under autonomous pressure. To optimize selectivity, we varied the concentrations of each aqueous phase modifier (Table 1 and Fig. 2) (24). Together, DMSO and PVP increased the selectivity from 60 to 75% (Fig. 2A). However, we also sought to optimize partitioning of the HMF product into the organic phase both to minimize degradation reactions arising from extended HMF residence in the reactive aqueous phase and to achieve more efficient recovery of HMF in the subsequent isolation step. Unfortunately, the aqueous additives are a liability in this respect, because they increase the solubility of HMF in the aqueous phase (thereby decreasing the R value). The addition of 2-butanol to the organic phase helped counteract this effect by increasing HMF solubility in the organic phase relative to pure MIBK (Fig. 2B). Although pure 2-butanol is moderately soluble in water [up to 19.5 wt % at 298 K, as measured by high-performance liquid chromatography (HPLC)], it partitioned effectively into the MIBK layer when mixed with MIBK at a 3:7 weight/weight (w/w) ratio (<5 wt % at 298 K was detected in the aqueous phase by HPLC). Starting with a 30 wt % aqueous fructose solution, our optimal results using all three modifiers (DMSO, PVP, and 2-butanol) yielded 0.065 g/ml of HMF in the organic layer, with 83% HMF selectivity at 82% conversion (Table 1, run 13).

Fig. 2.

Effects of changing the aqueous phase composition from water (W), to 8:2 w/w water: DMSO (W:D), to 7:3 w/w water:PVP (W:P), to 7:3 w/w (8:2 water:DMSO):PVP (W:D:P). (A) HMF selectivity with 30 wt % fructose feed. White and gray bars represent MIBK and MIBK:2-butanol extracting solvents, respectively. (B) Extraction ratio, R. White and gray bars represent MIBK and MIBK:2-butanol extracting solvents, respectively. (C) HMF selectivity with MIBK:2-butanol extracting solvent. White and gray bars represent 30 and 50 wt % fructose, respectively, and hatched bars show the improvement obtained with the use of double the amount of extracting solvent.

The selectivity for production of HMF decreases when the initial sugar concentration is increased from 30 to 50 wt % (Fig. 2C). This result is in agreement with literature reports that increasing the fructose concentration leads to higher rates of condensation reactions (6, 13). The addition of DMSO, PVP, and/or 2-butanol to the 50 wt % fructose systems in the same quantities used for lower fructose concentrations did not substantially improve the HMF selectivity. However, doubling the amount of the 7:3 MIBK:2-butanol extracting solvent increased the selectivity substantially. The optimal system using all three modifiers generated 0.070 g/ml of HMF in the organic phase at 77% HMF selectivity and 92% conversion (Table 1, run 27).

A heterogeneous catalyst is more easily separated from the product and recycled than a homogeneous catalyst, leading us to investigate fructose dehydration with the use of an acidic ion-exchange resin. These experiments were conducted at 363 K under autonomous pressure (Table 1, runs 28 to 39). When we used the resin catalyst, we observed trends in modifier impact that were similar to those we saw with HCl; however, the absolute selectivities were lower for the resin system. Kuster (6) reported that elevated temperatures (>453 K) favor higher HMF selectivity because at these conditions the rate of HMF formation is 4 to 10 times as fast as the rate of HMF degradation. Because ion-exchange resins degrade at these higher temperatures, it is desirable to replace HCl with a heterogeneous acid catalyst that is stable to 450 K. In this respect, and considering the favorable results we obtained using H3PO4 (Table 1, run 43), we tested a niobium phosphate catalyst at 453 K, obtaining promising results of 73% HMF selectivity at 62% conversion. In addition, Dias et al. (25) have shown that nanoporous materials possessing sulfonic acid groups can promote dehydration of xylose to furfural at elevated temperatures. Carlini et al. (26) observed fructose dehydration at low temperatures (353 K) and mostly at low conversion (<50%) with the use of vanadyl-phosphate–based catalysts, which may also withstand elevated temperatures.

Experiments conducted at low temperature and low conversions (Table 1, runs 29 and 32) offer mechanistic insight about the role of DMSO in enhancing HMF selectivity. The results show that DMSO primarily increases the rate of fructose conversion into HMF and to some extent decreases the rates of undesirable parallel reactions. Likewise, earlier work has suggested that DMSO suppresses both the formation of condensation by-products and HMF rehydration by lowering the overall water concentration (6, 13). Similar studies have indicated that DMSO both favorably shifts the equilibrium of the rate-determining step in HMF production and inhibits acyclic reaction sequences that may lead to undesirable intermediates (23).

The reaction of 10 wt % fructose in 7:3 water:NMP with MIBK as the extracting solvent and an acidic ion-exchange resin catalyst generated 68% HMF selectivity at 80% conversion, indicating that the addition of NMP to water enhances the selectivity of HMF production from fructose. Because NMP is an aprotic solvent with properties similar to DMSO, NMP appears to act through mechanisms similar to those of DMSO to enhance HMF selectivity in the fructose dehydration reaction. However, whereas the carryover of DMSO from the aqueous phase into the organic phase is not considerable (<0.8 wt % DMSO in MIBK after contacting an 8:2 water:DMSO aqueous solution, as measured by HPLC), the carry-over of NMP into the organic phase is high (∼5 wt % NMP in MIBK after contacting a 7:3 water:NMP aqueous solution, as measured by HPLC), thereby complicating the subsequent separation of HMF from the organic phase by evaporation. Notably, we found that replacing NMP with PVP, a stable hydrophilic polymer that has NMP moieties along the polyethylene chain, preserves the benefits on selectivity produced by NMP but eliminates organic phase contamination due to the low solubility of PVP in the extracting solvent. In an analogous manner, grafting DMSO onto a hydrophilic polymeric backbone could be used as a strategy to eliminate trace amounts of DMSO in the organic phase.

The addition of 2-butanol to MIBK enhances HMF selectivity by removing HMF more efficiently from the reactive aqueous medium. Notably, increasing the extraction ratio R and/or increasing Vorg/Vaq can counteract the faster rate of HMF degradation in the presence of fructose. This undesirable reaction between fructose and HMF is reflected in lower HMF selectivities at 50 wt % fructose when compared with 30 wt % (Table 1), and we observed directly that lower selectivities are obtained when controlled amounts of HMF are added initially to the fructose reaction system. In addition, separating HMF from the aqueous medium lowers the rate of HMF rehydration into levulinic and formic acids. Analyses by gas chromatography–mass spectrometry (GC-MS) of the aqueous and organic phases after conversion of 30 wt % fructose showed that the general composition of the by-products corresponds to typically 10% rehydration, 5% dehydration, 5% fragmentation, and 80% condensation compounds.

We performed simulations for selected experiments from Table 1 to estimate the HMF concentrations that would be obtained by combining our batch reactor experiments with a countercurrent extractor to remove the HMF remaining in the aqueous layer (Fig. 3). The final amount of HMF obtained by combining the organic streams from the reactor and the extractor (i.e., the stream entering the evaporator) is used to calculate the energetic yield (Yη) as a measure of the overall efficiency of our process for obtaining HMF by solvent evaporation. The energetic yield is the product of the HMF yield (Y), defined as the moles of HMF in the stream entering the evaporator in Fig. 3 divided by the total moles of fructose fed to the batch reactor, and an energy efficiency (η), defined as the heat of combustion of the HMF product (ΔHC,HMF) minus the energy necessary to evaporate the solvent (ΔHvap,org), normalized by the energy content of the product [i.e., η = (ΔHC,HMF – ΔHvap,org)/ΔHC,HMF]. These simulations used the experimental selectivity for each system (Table 1), which was assumed to remain constant at 90% conversion, the experimental value of Vorg /Vaq for the batch reactor, and the experimental value of R to model a countercurrent extractor operating with equal volumes of aqueous and organic streams. Aqueous and organic phase modifiers improve the value of Yη, thus reducing energy expenditures required to obtain the HMF product when compared with the water/MIBK system (Table 2).

Fig. 3.

Batch process for production of HMF from fructose with simulated countercurrent extraction and evaporation steps. The aqueous phase (white) contains fructose, DMSO, PVP, and the acid catalyst and is represented in the bottom half of the batch reactor. The organic phase (gray) contains MIBK or MIBK:2-butanol and is represented in the top half of the batch reactor.

Table 2.

Simulation of HMF yield (Y) and energetic yield (Yη) for selected dehydration systems. [HMF]aq corresponds to the HMF concentration in the aqueous phase leaving the extractor, and [HMF]org corresponds to the HMF concentration entering the evaporator in Fig. 3. Runs are based on runs in Table 1. The selectivity is set to the value obtained experimentally, and conversion is assumed to be 90%. The yield is calculated based on the HMF present in the organic stream sent to the evaporator.

Run no.Aqueous phase compositionOrganic phase compositionSelectivity (%)[HMF]aq (g/ml)[HMF]org (g/ml)Y (%)Yη (%)
30 wt % fructose
2 Water MIBK 60 0.0075 0.045 48 34
4 Water 7:3 MIBK:2-butanol 70 0.0001 0.057 61 43
6 8:2 Water:DMSO MIBK 67 0.025 0.048 48 35
7 8:2 Water:DMSO 7:3 MIBK:2-butanol 75 0.0009 0.063 66 48
12 7:3(8:2 Water:DMSO):PVP MIBK 75 0.024 0.057 56 44
13 7:3(8:2 Water:DMSO):PVP 7:3 MIBK:2-butanol 83 0.0032 0.071 73 56
50 wt % fructose
16 Water MIBK 47 0.0026 0.054 39 27
26 7:3(8:2 Water:DMSO):PVP 7:3 MIBK:2-butanol 62 0.0019 0.091 53 43
27 7:3(8:2 Water:DMSO):PVP 7:3 MIBK:2-butanol 77 0.0055 0.071 67 51

The value of Yη alone does not address the difficulties of using high-boiling organic systems. For example, although a theoretical value of Yη > 75% can be obtained with pure DMSO, the HMF product cannot be separated from the solvent by simple evaporation. Previous work has shown that because of the reactive nature of concentrated HMF at high temperatures, distillation of HMF from DMSO leads to substantial carbonization of the product (10). Low-temperature separation processes such as vacuum evaporation and vacuum distillation have been used to separate various solvents and by-products from HMF mixtures, but no experimental data have been reported for DMSO (2729). Accordingly, we used Aspen Plus Simulation Software (Version 12.1, Aspen Technology Inc.) to compare energy requirements for the separations of HMF from a low-boiling solvent (pure MIBK) and from a high-boiling solvent (pure DMSO) for vacuum evaporation and vacuum distillation processes (for HMF levels of 10 wt %). Vacuum evaporation simulations predicted that 99.5% of the MIBK solvent can be evaporated at 13 mbar and 343 K with a 2.5% loss of HMF, whereas evaporating DMSO at 1.3 mbar and the same temperature resulted in a 30% loss of HMF. Consequently, HMF separation from DMSO with minimal losses requires the more expensive vacuum distillation process (e.g., 0.66 mbar and a bottom temperature of 386 K). When comparing both solvents with the use of vacuum distillation, simulations predicted that an efficient separation of HMF from pure DMSO requires 40% more energy as compared with pure MIBK, clearly showing the advantages of using a low-boiling solvent system.

Supporting Online Material

Materials and Methods

References and Notes

View Abstract

Navigate This Article