Furfural—A Promising Platform for Lignocellulosic Biofuels


  • Dr. Jean-Paul Lange,

    Corresponding author
    1. Shell Global Solutions International B.V. Shell Technology Centre Amsterdam, Grasweg 31, 1031 HW Amsterdam (The Netherlands), Fax: (+31) 20-6303010
    • Shell Global Solutions International B.V. Shell Technology Centre Amsterdam, Grasweg 31, 1031 HW Amsterdam (The Netherlands), Fax: (+31) 20-6303010
    Search for more papers by this author
  • Dr. Evert van der Heide,

    1. Shell Global Solutions International B.V. Shell Technology Centre Amsterdam, Grasweg 31, 1031 HW Amsterdam (The Netherlands), Fax: (+31) 20-6303010
    Search for more papers by this author
  • Dr. Jeroen van Buijtenen,

    1. Shell Global Solutions International B.V. Shell Technology Centre Amsterdam, Grasweg 31, 1031 HW Amsterdam (The Netherlands), Fax: (+31) 20-6303010
    Search for more papers by this author
  • Dr. Richard Price

    1. Shell Global Solutions (UK), Shell Technology Centre Thornton, Pool Lane, Ince, Cheshire CH2 4NU (United Kingdom)
    Search for more papers by this author


Furfural offers a promising, rich platform for lignocellulosic biofuels. These include methylfuran and methyltetrahydrofuran, valerate esters, ethylfurfuryl and ethyltetrahydrofurfuryl ethers as well as various C10–C15 coupling products. The various production routes are critically reviewed, and the needs for improvements are identified. Their relative industrial potential is analysed by defining an investment index and CO2 emissions as well as determining the fuel properties for the resulting products. Finally, the most promising candidate, 2-methylfuran, was subjected to a road trial of 90 000 km in a gasoline blend. Importantly, the potential of the furfural platform relies heavily on the cost-competitive production of furfural from lignocellulosic feedstock. Conventional standalone and emerging coproduct processes—for example, as a coproduct of cellulosic ethanol, levulinic acid or hydroxymethyl furfural—are expensive and energetically demanding. Challenges and areas that need improvement are highlighted. In addition to providing a critical review of the literature, this paper also presents new results and analysis in this area.

1. Introduction

At the beginning of the 21st century, mankind is facing the tremendous challenge of requiring ever more energy, while the easily accessible oil fields are becoming depleted and CO2 emissions from fossil fuels are affecting the earth’s climate. Hence, much research is being devoted to the exploration and development of new, non-fossil carbon energy sources. These include biofuels, which are recognised as a promising option for the transportation sector in the coming decades. The attractiveness of biofuels goes beyond the exploitation of a new energy source and the resulting CO2 savings. It also includes opportunities to secure the local supply of energy and to support agricultural economies when produced locally.1, 2

The first generation of biofuels is presently produced from sugars, starches and vegetable oils. A more promising feedstock is lignocellulose, which is more abundant, cheaper and potentially more sustainable.18 However, lignocellulose is recalcitrant and, therefore, requires complex and expensive processes to upgrade to biofuels. Some processes aim to convert the whole lignocellulose to biofuels, for example, by pyrolysis or gasification, whereas others focus on unlocking its valuable sugars to upgrade them to ethanol.1, 2 Recent years have witnessed much activity to upgrade the unlocked sugars to fuels others than ethanol. For instance, glucose can be converted to hydroxymethylfurfural (HMF) and subsequently upgraded to dimethylfuran (DMF)9, 10 or diesel-range hydrocarbons.11, 12 Alternatively, glucose can be reformed to alkanes13, 14 or converted to a mixture of light oxygenates and, subsequently, to aromatic gasoline.15, 16 Cellulosic glucose can also be converted to levulinic acid (LA) and, subsequently, to ethyl levulinate (EL) and methyltetrahydrofuran (MTHF),17 valerate esters18 or butene-based fuels.19

Furfural also deserves attention as a potential platform for biofuels. Furfural is produced by the hydrolysis and dehydration of xylan contained in lignocellulose. Furfural offers a rich source of derivatives that are potential biofuel components. We discuss the complete value chain of furanic biofuels and review the hydrogenation and acid–base-catalysed reactions applied to upgrade furfural to biofuels (Figure 1). The literature is complemented with new information on known and new routes, which was gathered in Shell laboratories from 2003 onwards. These routes are then ranked on their industrial potential by a discussion of their manufacturing footprint—that is, investment cost and CO2 emission of furfural upgrade—and an investigation of the fuel quality of the resulting furanic product. This fuel evaluation covered an extensive road trial of over 90 000 km for 2-methylfuran (MF). Finally, we review the various options to manufacture furfural from lignocellulose and discuss the challenges they face, which includes the expensive recovery of furfural from process streams.

Figure 1.

Furfural platform for biofuels.

This review leans on invaluable earlier reviews on furfural manufacture and upgrade, starting with the classic book from Dunlop and Peters,20 the more recent one from Zeitsch21 and numerous reviews.2227 For the sake of length and cohesion, we chose to exclude research on the chemical application of furfural derivatives and the manufacture and upgrade of HMF, which were discussed elsewhere.2629

2. Furfural Derivatives

As discussed later in this review, furanic biofuels require deoxygenation to increase energy density and miscibility in hydrocarbon fuels and optional chain-lengthening to provide high-boiling diesel components. Upgrade involves hydrogenation, rearrangement, C[BOND]C coupling, and optional combinations thereof. The key reactions are reviewed below.

2.1. Hydrogenation

Hydrogenation remains the most versatile reaction to upgrade furanic components to biofuels. For instance, it can lead to MF, DMF and MTHF, which have been reported as promising biofuel components.9, 17 Bel’skii et al.24, 25 have reviewed the field of hydrogenation and hydrogenolysis of furanic compounds until the 1960s. They reported a rich hydrogenation chemistry for furfural, which includes the hydrogenation of the [BOND]CH[DOUBLE BOND]O side chain to [BOND]CH2OH or [BOND]CH3, the hydrogenation of the furan ring and its opening to pentanols, pentane diols and, eventually, alkanes. The selectivity is dictated by the catalyst formulation and the reaction conditions. Two promising gasoline components warrant an extensive discussion, namely, MF and MTHF (Figure 2).

Figure 2.

MF and MTHF production.

2.1.1. MF production in the gas phase

Of particular interest is the selective (95 %) conversion of furfural to MF by Cu-based catalysts that operate at high temperature and low pressure, for example, 200–300 °C, LHSV (liquid hourly space velocity) of 0.15–0.3 h, 0.1 MPa and a H2/furfural molar ratio of 5–8.30 The reaction was shown to proceed through furfuryl alcohol (FAlc) as an intermediate. Raney-Cu, Cu/Al2O3 and Cu–chromite showed similar behaviour, although the latter was more active and stable. The catalysts were reported to deactivate rapidly, but could be regenerated by coke burn off at 400 °C. Carbon-supported Cu–chromite was also reported to be selective for MF, but also deactivated within a few days.31 These observations were confirmed and elaborated upon in later studies32, 33 as well as in our laboratory. This deactivation is most likely caused by thermal polymerisation and coking of FAlc in the catalyst bed. Economical operation would require increased catalyst stability and an effective regeneration procedure.

2.1.2. MF production in the liquid phase

The rapid catalyst deactivation observed during the gas-phase production of MF triggered the question of whether MF could also be produced at a milder temperature, for example, in the liquid phase. A few papers have reported the hydrogenolysis of furfural, 5-methylfurfural or HMF to MF and DMF by using Pd catalysts under mild conditions. Nudelman et al.34 have described the hydrogenolysis of furfural to MF and of 5-methylfurfural to DMF, using Pd supported on Carbon (Pd/C) at room temperature and 0.2 MPa H2. Sun et al.35 have reported a polymer-supported PdII complex that catalysed the hydrogenolysis of furfural to MF. They claimed a 100 % yield in 1 h at 18 °C with 0.1 MPa H2. A range of solvents was screened, and the best results were obtained with small, polar alcohols such as methanol and ethanol. Hamada et al.36 converted fructose to DMF by dehydration to 5-chloromethyl-2-furfural and subsequent hydrogenation. The second step proceeded with up to 88 % yield through transfer hydrogenolysis by using Pd/C as the catalyst and cyclohexene as the H donor at 60–80 °C. Transfer hydrogenolysis of HMF to DMF was confirmed by Morikawa, who used Pd/C and cyclohexene.37 Improved results were obtained in the presence of AlCl3 and other Lewis acids, and yields of up to 60 % were obtained. Luijkx has also reported the hydrogenolysis of HMF.10 Reactions were performed at 60 °C, and DMF yields of up to 36 % were reported in 1-propanol using Pd/C as the catalyst. Pd/Al2O3 was inactive as such, but showed an activity similar to Pd/C upon addition of a catalytic amount of HCl.

We revisited this approach and confirmed that undesired side reactions (e.g., ring hydrogenation of the feed or the product) could be depressed by using an acid cocatalyst and by appropriate selection of the reaction solvent (see the Supporting Information for experimental procedures). For example, suitable solvents required intermediate polarity (i.e., an octanol/water partition coefficient logP between −1 and +2) and neither acidic nor basic properties. Furthermore, the catalyst preferably consisted of a strong acid and was assisted by halide anions. For example, the reaction proceeded satisfactorily using bifunctional ion-exchange resins, for example, Pd/Amberlyst CH 28, upon addition of halides (CaCl2 or HCl) to the reaction medium. MF yields of up to approximately 50 mol % were achieved by using Pd/C or Pd/SiO2 in ethanol at 30 °C and 0.25 MPa H2 in the presence of small amounts of HCl (see Figure in the Supporting Information).

2.1.3. MF production in reactive distillation

Overall, the consecutive hydrogenation of MF appeared to limit the selectivity. Presumably, these reactions can be minimised by performing the reaction under stripping conditions, which leads to the continuous removal of MF by virtue of its low boiling point (65 °C). However, reactive distillation appeared to require the adjustment of the operating conditions described above.38 Firstly, the reaction temperature was raised, for example, to 100–150 instead of 30 °C. Secondly, volatile HCl was replaced by a non-volatile acid, such as H2SO4, or an acidic catalyst support could be used as carrier for Pd. These adjustments appeared to be detrimental for Pd catalysts, as they led to significant ring-hydrogenation, but they were more promising for Cu-based catalysts (see the Supporting Information).38

Reactive distillation allows process intensification by avoiding the energy-demanding interstage recovery of furfural and replacing it by easier stripping of MF.38 Such integration could be even more attractive for the coproduction of furfural and HMF, in which both hydrogenation products (MF and DMF) are stripped together.

2.1.4. Ring hydrogenation

The various furanic derivates reported above can be further upgraded by means of ring hydrogenation and hydrodeoxygenation.

MTHF has been reported as a potential fuel component.17 It can be obtained from furfural by hydrogenation to MF and subsequent ring hydrogenation, for example, by using Ni-based catalysts. Indeed, a two-step process has been proposed to convert furfural to MTHF.39 The first reactor produces MF by using a Ba/Mn-promoted Cu–chromite catalyst, which operates at 175 °C and 0.1 MPa with a H2/furfural molar ratio of approximately 2 and an undisclosed gas hourly space velocity (GHSV). The second reactor hydrogenates the MF to MTHF by using a Ni-based catalyst at 130 °C.

A two-step process for MTHF production has also been proposed using supercritical CO2 and a combination of Cu–chromite and Pd/C as catalysts.40 The elegance of this scheme is the flexibility that it provides through the independent adjustment of the temperature of each reactor to produce FAlc, tetrahydrofufuryl alcohol, MF, MTHF or furan. However, the advantage of operating under supercritical CO2 rather than pure feed has not been demonstrated.

Similarly, furan and ethylfurfuryl ether (EFE) undergo ring-hydrogenation to deliver tetrahydrofuran41, 42 and ethyltetrahydrofurfuryl ether (ETE),43 respectively.

Clearly, furanic condensation products, which are discussed later in this paper, can equally undergo ring-hydrogenation to produce saturated C10+-oxygenates using Ni-based catalysts.

2.1.5. Hydrogenation to alcohols and diols

The furan ring can be further hydrogenated and ring-opened to produce diols (e.g., 1,2- and 1,5-pentanediol) and alcohols (e.g., 1- and 2-pentanol) by using noble metal catalysts at high temperature.24, 25 These products could be further converted to valuable biofuels by esterification (e.g., to pentylvalerate or pentanediol divalerate)18 or by etherification (e.g., to dipentyl ether).44 Hydrogenation generally proceeds through the formation of tetrahydrofurfuryl alcohol followed by hydrogenolysis of the ether or alcohol C[BOND]O bond. Regioselective C[BOND]O cleavage has long been a challenge, as the product generally consists of a mixture of oxygenates.24, 25 However, new bimetallic catalysts have recently proven to be very effective for this purpose. For example, the promotion of Rh/SiO2 with Mo or Re has been reported to convert tetrahydrofurfuryl alcohol to 1,5-pentanediol with >80 % selectivity.4547 In the absence of the Mo or Re promoters, the Rh catalyst selectively produces 1,2-pentanediol. However, no catalyst seems to deliver 1-pentanol with high selectivity.

2.1.6. Deep hydrodeoxygenation

Mono- and oligomeric furan components can be converted to aliphatic hydrocarbons by hydrodeoxygenation. For example, HMF–ketone coupling products and MF trimers have been converted to C10+ hydrocarbons by ring-hydrogenation and subsequent hydrodeoxygenation using Pt-based catalyst, such as a bifunctional Pt/NbPO5 catalyst, which operates at 250–300 °C, or a mixture of Pt/C and Pt/TiO2.11, 12, 48

The hydrodeoxygenation reaction of furanic components has also been reported to proceed at a mild temperature (80 °C) by using bifunctional Pd-based catalysts, such as Pd/SiO2–Al2O3 or Pd/zeolites, in supercritical CO2.49

2.2. Rearrangements

2.2.1. Rearrangement to levulinate esters

The conversion of furfural to EL, which can be further converted to valerate, pentenoate and polybutene biofuels, has been reported in literature.18, 19, 50 This reaction is particularly valuable, when furfural is a coproduct of LA, as it allows the conversion of the minor furfural coproduct into the main product. The conversion of furfural proceeds by hydrogenation to FAlc by using Cu-based catalysts21 and subsequent ethanolysis in the presence of strong homogeneous acids (Figure 3).5153 Strong homogeneous acids, such as HCl and H2SO4, are effective in the ethanolysis reaction. However, they require expensive metallurgy and are difficult to recover from the reaction product before reuse. Therefore, solid-acid catalysts, such as acidic ion-exchange resins and zeolites, are advantageous.

Figure 3.

Conversion of FAlc to EL.

Solid acids have also been shown to be effective.54 Especially macroreticular sulfonated ion-exchange resins with highly accessible sites (e.g., Amberlyst 46 and Purolite MN500) were particularly active in the production of EL with marginal coproduction of diethyl ether. Acidic zeolites also catalysed the conversion of FAlc and ethanol to EL.54 Yields of approximately 80 mol % were achieved with ZSM-5 (SiO2/Al2O3=30). However, zeolites tend to produce more diethyl ether than ion-exchange resins. Further reduction of the ether byproducts would improve the economic potential of this process.

2.2.2. Upgrade to EFE

The FAlc platform offers alternative leads, such as the possibility to convert FAlc to EFE. When performed in the presence of mild acidic catalysts such as zeolites, the rearrangement of FAlc appeared to be accompanied by the production of EFE in significant amounts. By analogy with known furanic fuel, EFE was anticipated to show good properties for use as a gasoline component, which will be discussed later in this paper.

In the presence of a powdered ZSM-5 catalyst (SiO2/Al2O3=30), a 7.5:1 molar mixture of ethanol/FAlc was converted to EFE in a maximum yield of 50 mol % with approximately 80 % conversion (Figure 4, see the Supporting Information for detailed experimental procedures).55 The formation of EFE was accompanied by heavy products at approximately 20 mol % yield. EL, LA and angelica lactone remained at fairly low levels with a total of <10 mol %. The influence of the mode and conditions of operation as well as the performance of alternative zeolites is addressed in the Supporting Information.

Figure 4.

Conversion of FAlc and ethanol to EFE and byproducts using HZSM-5 (batch, 125 °C, HZSM-5/ethanol/FAlc=1:10:10 by weight).

H2SO4 was also effective at very low acid concentrations. Operation at 150 °C with 0.003–0.1 wt % H2SO4 and an ethanol/FAlc molar ratio of 10:1 allowed an EFE selectivity of approximately 30 mol % for 20–90 % conversion. At higher conversion, EFE was converted to EL. Low temperature and lower ethanol/FAlc ratios were detrimental to the EFE selectivity.

Although promising, these preliminary results need further improvement, particularly regarding selectivity, to be industrially attractive. However, the heavy byproducts can also be hydrodeoxygenated to diesel-range hydrocarbons, and EL can be converted to the valerate salts of oligobutene biofuels, which is discussed elsewhere in this review.

2.2.3. Decarbonylation

Furfural can be decarbonylated to furan under reductive conditions and subsequently hydrogenated to tetrahydrofuran for blending into gasoline. The carcinogenic nature of furan, combined with its high volatility, is expected to prevent direct blending of this component in gasoline. Furfural decarbonylation has been widely investigated using catalysts based on metal oxides and precious metals. Metal oxide catalysts, for example, based on Fe, Zn, Mg, Cr, Co, Mo and Ni, are preferably used at temperatures of 300–500 °C.56 However, these high temperatures induce the decomposition of the furan and the formation of heavy products, with fast catalyst deactivation as a result.

In contrast, supported noble metals are good decarbonylation catalysts at much milder temperatures. Gárdos et al.5759 reported that Pd/Al2O3 was effective at 240–400 °C in a flow of H2. The catalyst appeared to deactivate within a day, but could be regenerated by burning the coke in an O2-lean gas stream. Furan yields of 92–97 % were achieved using 2.5 wt % Pd loading at 300 °C with an H2/furfural molar ratio of 1:1 and a liquid load of LHSV=0.75 L L−1 h−1). BASF has reported that the successful operation for nearly 1500 h using Pt and Rh catalysts, which were supported on Al2O3, was promoted with Na or Cs at 300 °C in an H2 flow with an H2/furfural molar ratio of 0.74.60

Reevaluation of furfural decarbonylation using a Cs-promoted Pt/Al2O3 catalyst (4 wt % Cs and 0.75 wt % Pt)60 confirmed high furan yields (up to 76 mol %). However, it also revealed a significant, unreported coproduction of propene and propane (up to 5 mol % yield each, Figure 5) in addition to smaller amounts of MF, FAlc and ring-opening and -coupling products (<1 mol % yield each). The large production of propane and propene and its constant selectivity, irrespective of conversion (Figure 5), is intriguing and warrants further investigation. In contrast to literature reports, the catalyst appeared to deactivate severely and had a half-life of approximately two days.

Figure 5.

Furan and propane+propene (C3s) yields during furfural decarbonylation using the Pt/Cs/Al2O3 catalyst (275–325 °C; 2–16 bar, (weight hourly space velocity)−1: 0.26–3.48 h gcatequation image, H2/furfural=0–3 based on mol %).

Gaset et al.6164 performed the decarbonylation reaction at a milder temperature (160 °C) in the liquid phase in the absence of H2 using Al2O3- and C-supported Pd catalysts and K2CO3 as the cocatalyst. The reaction was not affected by traces of water, but was hindered by traces of acids, which were formed, for example, by autooxidation of furfural during storage.

2.3. Furan coupling

2.3.1. Aldol condensation

Furfural is known to undergo base-catalysed condensation reactions with aldehydes and ketones to mono- and difuryl components.65 Dumesic et al.11 revisited this reaction and integrated it in a four-step process to convert sugars to diesel-range hydrocarbons (Figure 6).12 The four steps consist of 1) biphasic dehydration of sugar to furans, 2) biphasic aldol condensation of the furans with a ketone, 3) mild hydrogenation of aldol products and 4) hydrodeoxygenation to alkanes.

Figure 6.

Coupling of furfural and acetone and subsequent hydrodeoxygenation to biodiesel (adapted from Ref. 11).

LA can also be used as the ketone. In older literature, the conversion of furfural to a heavy C10 diketodiacid by aldol coupling with LA and subsequent hydrolysis with water (Figure 7) is reported.66, 67 This coupling product might be valuable as a diesel component, particularly after hydrogenation and/or hydrodeoxygenation. The coupling step proceeds in the presence of excess base (e.g., NaHCO3), and is followed by an acidification step to generate the acid form of the furfurylidene–LA intermediate.

Figure 7.

Coupling of furfural with LA to C10 diketodiacid.

This reaction appeared to proceed with catalytic amounts of solid base when using EL as the nucleophile and water-stripping conditions (see the Supporting Information). These conditions indeed avoided the presence of carboxylic acid that would otherwise neutralise the basic catalyst. For example, Cs/MgO and K/ZnCr led to the rapid formation and distillation of water with the simultaneous production of coupling products (Figure 8). Dimers formed with a maximum yield of 20–25 C % before conversion to higher oligomers in approximately 80 C % yield. The catalytic activity appeared to depend only partly on the expected basicity of the catalyst, and the Sn and Zn oxides were much more active than the more basic Mg and La oxides (Figure 8).

Figure 8.

Coupling of furfural and EL in the presence of a solid base under reactive distillation conditions (EL/FR molar ratio=1, 2.5 wt % catalyst; 170–230 °C for 2–7 h).

The dimeric product consisted mainly of the desired straight-chain furylidene levulinate and its corresponding branched isomer (see the Supporting Information). The structures of the trimers and heavier products have not been investigated. Nevertheless, these oligomers are expected to be good precursors for hydrodeoxygenation to kerosene (dimers) and diesel hydrocarbons (heavier oligomers).

2.3.2. Oligomerisation of MF

In an alternative approach, Corma et al. applied acid catalysis to produce furanic oligomers.44 They produced a trimer of MF by heating in 24 wt % aqueous H2SO4 to produce 4-oxopentanal and, subsequently, alkylated the ring-opened product in  situ with two other molecules of MF to produce a C15 oxygenate, 5,5-bisylvyl-2-pentanone , in approximately 72 % yield (Figure 9). Furthermore, the trimeric product separated spontaneously from the aqueous phase. Hydrodeoxygenation of this trimeric product over a physical mixture of Pt/C and Pt/TiO2 delivered a hydrocarbon diesel fraction in 87 % yield.

Figure 9.

Trimerisation of MF to bisylvylpentanone (adapted from Ref. 44).

3. Footprint of Furfural Upgrade

The preceding sections described a large number of furfural upgrade options. The selection of those with high commercial potential remains a challenge, however, as it requires some comparative analyses of manufacturing cost and of product performance. This section presents a high-level analysis of the footprint of the various furfural upgrade routes. Product performance will be discussed in the following section.

For transparency, the footprint analysis was limited to the upgrade of furfural to biofuel derivatives. Accordingly, it excluded contributions from crop production, harvest and conversion to furfural, which are identical for all the routes considered here. Clearly, these contributions need to be considered when comparing the furfural derivatives to biofuels obtained from other platforms, for example, from LA or HMF. Although valuable, such extensive analysis is outside the scope of this study.

The footprint analysis was further simplified by considering only two factors, namely, the investment cost and the net CO2 emission of the upgrade. More precisely, it considered two approximations for these factors as detailed technical information is lacking at this stage. First, the investment cost was approximated by using a theoretical capital index, which was based on the stoichiometric heat of reactions and relied on a correlation found between the investment cost and the overall energy transfer duty for fuel and chemical manufacturing processes.6870 Second, the CO2 emission of the upgrade was approximated by the CO2 emission related to the production of fossil H2 that was needed in the hydrogenation steps. This CO2 contribution was expected to dwarf all other CO2 contributions during the upgrade, for example, those from process energy. More details on the methodology are provided in the Supporting Information.

The relative capital intensity and CO2 upgrade footprint of the various furfural derivatives are shown in Figure 10, which reveals a distinct correlation between the two factors. In addition to being responsible for significant CO2 emissions, hydrogenation reactions are also very exothermic and, thereby, lead to significant capital intensity. Overall, Figure 10 shows a limited footprint for the conversion of the aldehyde group of furfural to, for example, alcohol, ether or methyl groups in FAlc, EFE or MF. However, the footprint increases significantly as the furan ring is saturated [e.g., in ETE and (M)THF] or opened [e.g., to 1-pentanol (PA), dipentyl ether (DPE) or hydrocarbons]. Ring opening leads to a limited increase in capital index owing to limited exothermicity, but to a large overall CO2 footprint of approximately 40 g MJ−1, which represents nearly 50 % of the well-to-wheel CO2 emissions of fossil transportation fuels (estimated at around 84 g MJ−1).71

Figure 10.

Capital intensity and CO2 emission of furfural upgrade to furanic (⧫), tetrahydrofuranic (□) and ring-opened products (•). *The heating value of ethanol has been excluded from the heating value of the fuel to concentrate the footprint of EFE and ETE to that of the furfural derivative.

Biofuels with a high capital intensity and high CO2 upgrade footprint need to provide excellent fuel performance and high value to justify deployment. Our study of fuel performance, which is discussed in the following section, revealed that an upgrade with a small footprint produced good gasoline components such as MF and EFE. In contrast, good diesel components require more thorough upgrade with a larger footprint.

4. Fuel Properties

4.1. Property screening

After a review of the chemistry of manufacture, attention should also be paid to the fuel properties of the furanic derivatives produced. An initial evaluation was made by assessing components against four key criteria: energy density, polarity, boiling point and ignition characteristics. This approach has already been reported for the evaluation of valeric biofuels18 and allows the identification of components that are compatible with current vehicles and/or fuel distribution networks. Incompatibility would incur additional costs and increased development times for the deployment of new biofuels.

This preliminary screening revealed that the presence of alcohol (FAlc) and aldehyde (furfural) functionalities in a molecule establishes a polarity mismatch for blending with hydrocarbon fuels (Figure 11). Although this can be managed, as in the case of ethanol, it is best avoided as it limits blending concentrations and may result in undesirable side effects, such as increased volatility, or potential incompatibility with materials used in fuel distribution systems and vehicles. Removal of the alcohol or aldehyde group results in components with improved solvency (reduced polarities) and increased energy densities.

Figure 11.

Prescreening of fuel properties for furans (▴), tetrahydrofurans (○) and reference components (✶). BRON was determined at 5 vol % in a 93 RON gasoline for most components, but estimated for tetrahydrofurans (see the Supporting Information).

All monomeric derivatives fall in the gasoline boiling range. The unsaturated molecules typically have good octane properties, with blending research octane numbers (BRON, which has been measured at 5 vol % in a 93 RON gasoline) that often exceed that of ethanol (Figure 11). Saturation of the furan ring appeared to reduce the octane quality by 50 points (the BRON of tetrahydrofuranic components was estimated as described in the Supporting Information). The resulting increase in H/C did not result in a higher volumetric energy density, as it was offset by a decrease in density. Interestingly, EFE shows a high BRON (≈110),55 and its saturated equivalent, ETE, has been reported by Avantium43 to have a high cetane number (>80) and, therefore, a low octane number (see the Supporting Information).

No octane information was available for the oligomeric derivatives. However, the present data suggest that they would form reasonable diesel blending components after ring hydrogenation. This should hold for furfural–ketone aldol products, furfural–EL products and MF trimers. It may also hold for other heavy derivatives, such as furfuryl ethers (e.g., difurfuryl ether) and heavy furoate esters. The unsaturated components may deteriorate the cetane number of diesel blends. Hydrocarbons produced by deep hydrodeoxygenation of oligomeric materials will present no problem in diesel blends.

Although not displayed in Figure 11 for the sake of transparency, the alcohols and diols derived from furfural do not deliver desirable fuel properties. They suffer from high polarity and a BRON that falls between that of gasoline and diesel. Ethers and esters derivatives are more suitable, however. For example, DPE is a promising diesel component with desirably low polarity combined with high energy density, high boiling point and high cetane number. Similarly, valerate mono- and diesters were shown to be fully compatible with either gasoline or diesel, which depends on, for example, the length of their alkyl chain.18 Esters of 1-pentanol or pentane diols are therefore suitable diesel components.

Avantium has proposed a variety of furanic components for biofuel applications,72 which consist mainly of ethers of FAlc (e.g., EFE), diethers of dihydroxymethylfuran [e.g., bis(ethoxy)methylfurfuran] and the corresponding ring-hydrogenation derivatives [e.g., ethoxy- and bis(ethoxy)methyltetrahydrofuran]. Avantium also proposed aldo ethers of HMF (e.g., ethoxymethylfufural), but recognised their limited miscibility in hydrocarbon fuels at high concentration,72 which is consistent with the high polarity reported above for furfural.

4.2. Detailed fuel evaluation

A more detailed study of fuel performance focused on MF and EFE, which combine good fuel performance with a moderate furfural upgrade footprint. Blends of these components were evaluated for compatibility with the European EN228 gasoline specification and for eventual operational issues related to, for example, fouling, oxidation instability, elastomer swelling and health/safety/environment (HSE). Key results are reported in the Supporting Information.

Based on these encouraging results, MF was evaluated in a road trial. A gasoline with 10 vol % MF was tested in three vehicles (both multiport and direct injection technologies) on the road for a combined total distance of 90 000 km. End-of-test measurements showed no negative impact on regulated vehicle emissions (NOx, HC and CO), with EURO 4/5 compliance sustained. The low O2 content of MF did not result in measurable penalties in vehicle fuel consumption (L per km) when blended at 10 vol %. The loss of fuel economy was expected to be approximately 1 %, which is more favourable than the approximately 3 % loss expected for an equivalent blend ratio of ethanol in gasoline. No detrimental impact on engine-oil degradation and engine wear was identified. However, significant changes in engine inlet valve and injector deposits were noted. Successful control of these deposits was achieved through adjustment to the detergent additive package of the gasoline.

5. Furfural Production

The industrial potential of the furfural platform depends on the possibility of the production of furfural in an affordable and sustainable way. We, therefore, need to review the feedstock and processes that can be used for furfural manufacture. The feedstock mainly consists of lignocellulosic residues from agriculture. The manufacturing processes are either standalone processes that exclusively produce furfural or lignocellulose biorefineries that produce or could produce furfural as a coproduct of cellulosic biofuels.

5.1. Feedstock

Furfural is generally derived from C5 sugars, mainly xylose and arabinose, that are contained in the hemicellulose of lignocellulosic materials.73, 74 Lignocellulose is generally considered as a promising feedstock, as it is abundant, inexpensive and potentially sustainable.18 Lignocellulosic xylose is mainly present as glucuronoxylan, in addition to some xyloglucan, in hardwoods and as glucoronoarabinoxylan (GAX) in grasses. GAX is also a minor component in softwood, in addition to mannans. Arabinan is present in minor quantities as side chains in xyloglucans, GAX and arabinogalactans.75

The combined xylan and arabinan content is up to 10 wt % in softwoods, 20 wt % in hardwoods and 28 wt % in grasses, which includes bagasse (Table 1). The pentosan content is dominated by xylan, as arabinan accounts for 1–3 wt % only. The content of arabinan and xylan shows no correlation within hardwood species and only weak correlation within softwood and grasses, despite their combined presence in GAX. Xylan does not seem to correlate with other sugar residues either. However, arabinan strongly correlates with galactan for grasses, softwoods and probably also for hardwoods, which suggests a variation of arabinogalactan content as the main determinant for variations of arabinan.

Table 1. Typical xylan and arabinan contents in lignocellulosic biomass on dry weight basis.[76–81]
FeedstockXylan [wt %]Arabinan [wt %]

Compositional variation is also found within a plant. For example, the leaves and nodes of corn stover contain 16 wt % xylan, whereas the husks typically contain 22 wt % xylan.79 A higher xylan content is found in corn cobs, up to 28–31 wt %,82, 83 in addition to an arabinan content at or approximately 3 wt %.80 The pentosan content in oat hulls is comparable to that of corn cobs.84 Sugar cane bagasse also has a considerable pentosan content, up to 27 wt %,78, 80 which is partly attributable to the removal of nonpentosan components during sugar extraction.

Therefore, it is not surprising that furfural processes mainly use agricultural residues such as corn cobs, oat husks and bagasse.

5.2. Chemistry of xylose dehydration

Xylose is generally produced as an aqueous stream. It would, therefore, be preferable to convert furfural in water, ideally at the concentration level delivered by the pretreatment. The dehydration of xylose to furfural in water typically proceeds at 150–220 °C with up to 60–70 mol % furfural yields under stripping conditions.21 The reaction rate is small under neutral conditions, increases during autocatalysis by acetic acid, which originates from the biomass, and is accelerated further by the addition of acid catalysts,85 especially strong acids such as sulfuric acid at concentrations up to 2.5 wt %86 or higher.87, 88 The reaction is completed within hours to seconds, depending on the temperature and acid addition. The reaction mechanism is complex and still debated. Undesired side reactions include retroaldol fragmentation of xylose to C1–4 acids, aldehydes and ketones, as well as the condensation of furfural and degradation products to humins (Figure 12).84, 89 Operation at high temperatures in sub- or supercritical water mainly resulted in retroaldol products.90 Quantum mechanical models have suggested protonation of C2[BOND]OH and subsequent ring-contraction as the most likely pathway to furfural.91 In contrast, protonation at C3[BOND]OH would result in decomposition to formic acid and a C4 fragment, which seems to degrade very rapidly, as it is not observed experimentally. Open-chain mechanisms that have been proposed in the literature do not seem to be supported by quantum mechanics calculations.

Figure 12.

Conversion of xylose to furfural and C2–4 byproducts (250 °C, 34.5 bar, 5–100 wt % xylose in water, 5–40 mM H2SO4, 1–130 s. ⧫: Furfural; •: pentoses, except for xylose; ✶: C2 (glycolaldehyde); ▵: C3 (glyceraldehydes, lactic acid and hydroxyaceton); ○: C4 (pyruvaldehyde). Adapted from Ref. 89.

Reaction profiles, such as that in Figure 12, have been used to derive various kinetic models. Simple models, such as a two-step mechanism with first-order furfural formation and degradation, are able to describe the experimental results.92, 93 More complex models assume the reversible formation of an intermediate, arguably xylulose or lyxose,89 and include the formation of degradation products from xylose, the supposed intermediate, and furfural.94, 95 The theoretical study mentioned above does not report xylulose or lyxose as intermediates, but postulates an anhydroxylulose on the pathway to furfural.91

These different reaction schemes start either from xylan or from xylose. The former case assumes two xylan fractions for the adequate description of the kinetics; one fraction hydrolyses rapidly while the other does so at a lower rate. Despite the differences in models and the complexity of the kinetic equations, the formation and degradation of furfural is easily described under a wide range of temperatures, concentrations and pH values.96 However, the conversion of arabinan has not been widely studied. These limited studies indicate that the reaction rates of arabinan are slower than or comparable to those of xylan.74, 97

In addition to these studies in water, numerous other studies have explored the potential of the dehydration of xylose to furfural in various media and using Lewis or solid acids as catalysts. For example, Moreau et al. converted xylose to furfural using HY and HMOR zeolites at 170 °C in biphasic systems [water/toluene or the less effective water/ methyl isobutyl ketone (MIBK)].98, 99 This approach was further developed for both furfural and HMF by a number of groups, which used both homogeneous and heterogeneous acids.100103 In addition to improved yields, biphasic operation also allows the recovery of the diluted furfural by solvent extraction rather than by distillation of the water-rich furfural/water azeotrope. Recently, alkylphenols were identified, which were effective extractants for furfural even at low extractant/oil ratio and could be regenerated by simple distillation of furfural.104 However, any small loss of the expensive extractant could undermine the economic attractiveness of biphasic reactions. Heterogeneous catalysts might not be the most attractive option because they tend to deactivate by fouling with humins and, consequently, require frequent regeneration by coking burn off.

Pure organic solvents, such as dimethylsulfoxide, also allow higher furfural yields, although without the benefit of the solvent extraction of furfural.105 Raines et al. applied the combination of CrCl2–3 salts and organic/ionic mixtures [dimethylacetamide (DMA)/1-ethyl-3-methylimidazolium chloride (EMIMCL) or DMA/LiBr] to convert xylose and xylan to furfural at 100 °C.106 Although nonaqueous media promise a higher furfural yield, their economic viability will depend on the cost of recovery and purification of the xylan/xylose and the solvent, which is required to compensate for the loss of the expensive reaction medium.

Ebitani et al. showed that the addition of a solid base (hydrotalcite) to a solid acid (Amberlyst-15) improved the dehydration of xylose to furfural at <100 °C, although the benefit to the yield seems to be lost at temperatures >100 °C.107

5.3. Standalone furfural processes

5.3.1. Conventional processes

Worldwide furfural production is approximately 400 kt a−1, most of which is converted to FAlc for the subsequent production of furan resins.21 The majority of furfural production capacity is located in China, and other important furfural producers are Illovo Sugar Ltd. of South Africa and Central Roma Corporation Ltd. of the Dominican Republic. These furfural processes (typically only several kt a−1 batch, the largest capacity is 35  kt a−1, and several continuous processes are also operated) are all based on the first commercial process that was developed by Quaker Oats in 1921, who used oat hulls as feedstock.21, 80 In this process, sulfuric acid is used as a catalyst for the hydrolysis and dehydration of pentosan to yield furfural. This process typically coproduces acetic acid, acetone and methanol in significant amounts, for example, 0.45–0.8 tAcOH tfurfural−1 and approximately 0.3 tacetone+MeOH tfurfural−1.80 Valorisation of these coproducts may significantly contribute to the economics of the plant.

These conventional furfural processes rely on hemicellulose-rich feedstocks, such as corncobs or sugarcane bagasse, and yield approximately 10 wt % of furfural (typically only 50–60 % of the theoretical yield), which is stripped from the reactor with a large quantity of steam. The resulting residue after furfural production is used to generate the required steam for the reactor and downstream separation. Hence, the overall yield of valuable products does not exceed 18–21 wt % including coproducts.

The high steam requirement, which results from the relatively inefficient steam stripping in the reactor and from the distillation of the resulting aqueous furfural product, can be illustrated by the following calculations: per ton of furfural produced, a typical furfural process requires 30 t of steam for stripping and generates 17 t of residue (55 % moisture).80 Based on 2.26 GJ tsteam−1, a low heating duty of 5.8 GJ t−1 for the 55 % wet residue and a boiler efficiency of 60 %, the residue can generate 26 tsteam tfurfural−1. This still neglects the heat duty of the downstream separation section, which also features a high steam requirement as a result of the very dilute aqueous furfural feed. Although heat integration may help to reduce the net steam requirement, typically the complete residue will be required to supply the energy for the furfural production.

5.3.2. Process economics

Because of its low product yield and high energy requirement, conventional technology does not promise to deliver furfural at a competitive cost for biofuel application. A few simplified considerations suffice to illustrate this point. With a feedstock price of 2–4 $ per GJ1 and an overall energy efficiency of approximately 23 %, the feedstock cost amounts to 8–16 $/GJ. The investment cost of small furfural plants (1.5–5 kt a−1) is reported to amount to 1500–3000 $ per ton of annual furfural capacity (after indexation to 2010),80, 108 which corresponds to an annual cost of 15–30 $ per GJ of products. Upon adding fixed costs of 5–10 $ for each GJ, the overall manufacturing cost reaches 28–56 $ per GJ or 140–280 $ per barrel of oil equivalent.

The economic illustration is based on the following premises: the 23 % energy efficiency corresponds to an overall yield of 10 wt % furfural and 9 wt % coproducts; the investment cost is annualised by a capital charge of 25 %; the fixed cost is estimated at 30 % of the annualised investment cost; the equivalent oil price is based on a heating value of 5 GJ per barrel.

The conventional small-scale batch processes are not likely to deliver the volume required for fuel application. Improved technologies will need to operate at a much larger scale (20–100-times the current processes) with improved yields (e.g., >80 mol % based on xylose) and reduced energy consumption (e.g., <10 tsteam tfurfural−1) to leave residual biomass to generate power or for producing biofuels.

5.3.3. Improved processes

Various recent developments have addressed the shortcomings of low yields and high energy demand. For example, the SupraYield process claims to improve furfural yields by reducing the residence of furfural under the reaction conditions.94 This was achieved by starting the batch process at a higher initial temperature and by gradually reducing the pressure during the course of the reaction. It was also claimed that this batch process could significantly reduce steam use to 10 tsteam tfurfural−1. A 5 kt a−1 SupraYield furfural plant has been built at the Proserpine Sugar Mill in Australia, and a second plant is under construction in India.109 However, the capital intensity of this process appears to be high for downstream biofuel production.

5.4. Coproduct processes

In addition to standalone furfural processes, several lignocellulosic biorefinery concepts have been proposed, which include coproduction of furfural. Accordingly, furfural would be produced from the C5 sugars (pentosans) contained in the lignocellulosic biomass, whereas the C6 sugars (hexans) would be valorised for other purposes, for example, fermentation to ethanol or dehydration to LA or HMF with subsequent upgrade to biofuels.

5.4.1. Coproducts of cellulosic ethanol

Cellulosic ethanol is one of the most studied routes to biofuels. It builds on mature ethanol-fermentation technologies to convert the glucose that is contained in the inexpensive and abundant lignocellulosic residue (Figure 13). Valuable outlets are needed for the noncellulosic components, namely, lignin and pentosans. Fermentation of the pentosans to ethanol, hydrogen, butanol or hydrocarbons is technically complicated and in early stages of development.110113 Cofermentation of C5 and C6 sugars is under development, but remains challenging.114, 115 Hence, xylose (and arabinose) could be advantageously withdrawn after the pretreatment step and used for furfural production. This opportunity justifies a short discussion of pretreatment and hydrolysis technologies that may provide the xylose stream required for furfural manufacture.

Figure 13.

Flow scheme of Iogen’s cellulosic ethanol process (adapted from Ref. 8).

The most promising technologies for unlocking the cellulosic sugars are presently based on a combination of pretreatment and enzymatic hydrolysis. Pretreatment opens up the structure of the lignocellulose and removes the hemicellulose and lignin to uncover the targeted cellulose fibres, which make it accessible to hydrolytic enzymes. Enzymatic hydrolysis subsequently cleaves the cellulose fibres to release the fermentable glucose, and pretreatment produces the desired pentosans. Wet chemical pretreatment is based on concentrated acid, dilute acid, water (autocatalysis), caustic, organic solvents (organosolv), ionic liquids or combinations thereof.116, 117 It generally produces a stream that contains the pentosans together with other components, such as extractives, acetic acid and, occasionally, lignin degradation products. The solution of pentosans can be separated from the cellulose residue by belt filtration, filter press, screw press, centrifuge or comparable solid–liquid separation techniques.118 However, it should be recognised that significant amounts of dissolved pentosans become trapped in the cellulosic filter cake, which need to be recovered by washing, if they cannot be valorised with the cellulose.

Pretreatment is most easily performed at low solid concentration, for example, 10 wt %. However, this delivers a very diluted stream of pentosans (≈2 wt %), which is unattractive for furfural production. A large amount of energy is required to either recover the sugars prior to dehydration or to recover the dilute furfural after dehydration. Several schemes are envisaged to mitigate these problems. For example, a richer pentosan solution can be achieved with a higher solid content, which requires more expensive equipment capable of handling solids at high temperatures and pressures. Alternatively, higher concentrations can be achieved by building up the pentosan through recycling of the liquid. However, the pentosan concentration will still be limited by the amount lost with the filter cake or by dilution because of filter cake washing. A subsequent xylose dehydration process needs to be compatible with a fairly dilute solution or needs an expensive xylose recovery or reconcentration step. Finally, xylose can be dehydrated in a biphasic medium that allows the in situ extraction of furfural and, thereby, achieves higher furfural yields than in pure water.98104 Of particular interest is extraction with alkylphenol, which allows a high distribution coefficient and furfural recovery by distillation as the top product.104 Biphasic operation requires deep recovery of the expensive extractant with minimal chemical loss by degradation and physical loss in the water stream or filter cake.

Alternatively, pretreatment could produce furfural, build up its concentration by liquid recycling and recover it through flashing. Together with xylose, traces of glucose, arabinose and acetic acid are also dissolved, each at approximately 10–15 wt % relative to xylose, depending on the origin and composition of the biomass applied, concentrations and conditions.119 Building up these components and their degradation products in the recycle stream needs to be limited by, for example, bleeding or slipping with the filter cake. Volatile components are hardly present, although some traces of formic and acetic acid might flash together with the furfural azeotrope. Significant amounts of furfural might end up in the filter cake. This usually does not interfere with enzymatic hydrolysis,120 but reduces the rate of ethanol formation above a threshold concentration of 1.5 g L−1[121] and increasingly inhibits the fermentation at concentrations from 2–5 g L−1.122, 123 Compared to a C5/C6 cofermentation scheme, such a recycling concept might become attractive when the value of furfural and the savings of the simpler C6-only fermentation pays for the costs of furfural recycling and recovery.

5.4.2. Coproduct of LA

The acid hydrolysis of lignocellulosic materials is known to convert the pentosans to furfural while converting the hexanes to HMF and, subsequently, to LA and formic acid. Fitzpatrick et al. have proposed a two-step process using an interstage flash to coproduce LA and furfural in high yields of 70–80 mol % LA and 70 mol % furfural (based on hexanes and pentosan, respectively).17, 124 This process comprises a rapid conversion at 210–220 °C and 2.5 MPa under plug–flow conditions followed by a flash to 1.5 MPa at 190–200 °C to recover the furfural and allow the slower digestion of the remaining material in a continuous stirred-tank reactor (CSTR) to produce LA.

Our own study confirmed the need for flashing furfural, as it degraded rapidly before LA was formed (Figure 14, top). Here, 15 wt % of LA and 7.5 wt % of furfural corresponded to theoretical yields of 55 and 47 mol % of the corresponding hexanes and pentosans, respectively. These yields were obtained at a solid loading of 5 wt % in water/H2SO4. Operation at high solid levels, for example, 20 wt %, which is required for commercial operation, was possible when the H2SO4 concentration was increased to maintain the LA yield reported above. However, such an increase in solid and acid concentration was detrimental for the maximum furfural yields, which plummeted to 3–4 wt % or 18–25 mol % of the pentosans (Figure 14, bottom). Similar profiles have been reported in the literature.125, 126 High solid loading still resulted in a fairly dilute LA stream (≈4 wt % at 20 wt % solid loading) and, thereby, required an energy-demanding product recovery. Hence, the coproduction of furfural with LA is not likely to be capable of producing furfural at high yield and low steam consumption.

Figure 14.

Formation of various components during (batch) acid hydrolysis of birch wood at a) low and b) high solid loading (200 °C, 5–20 wt % wood, 3–9 wt % H2SO4).

5.4.3. Coproduct of HMF

In a complementary approach, other researchers have selected milder hydrolysis conditions to favour the production of HMF without conversion to LA and/or humins. These conditions were intrinsically favourable for the coproduction of furfural when the feedstock contained pentosans. Instrumental to this was the selection of milder acidity, milder temperature and/or the use of water-lean/water-free media, which properly solvate or extract furanic products.

This approach builds on the pioneering work of Teunissen,28, 127, 128 who demonstrated a high HMF yield when fructose was dehydrated in the presence of water and organic cosolvents, and of Garves, who converted lignocellulosic feedstock to alkyl levulinate and furfural by using alcohol or alcohol/water as the reaction medium.129 Garves showed significant coproduction of furfural and alkoxymethyl furfural at short residence times.

Dumesic et al. improved the pioneering approach of Teunissen by selecting MIBK, with the option of smaller amounts of butanol as the organic cosolvent, to extract HMF from the acidic aqueous medium.100 The method was initially applied for fructose and appeared to be effective with glucose and starch.130 Alkylphenols were reported to be a very promising extractant for the biphasic dehydration of xylose to furfural.104 Hence, they should also be effective for the hydrolysis of lignocellulose with in situ coextraction of HMF and furfural.

Mascal et al. have embraced in situ extraction by converting cellulose in a concentrated HCl/LiCL aqueous solution at 65 °C by continuous removal of the reactive chloromethylfurfural (CMF) product using C2H4Cl2.131 This method was equally effective with lignocellulosic feedstock, which coproduced CMF and furfural.132, 133 Avantium revisited the work of Garves and applied alcohol as the solvent for the conversion of sugars to alkoxymethylfurfural.72

Ionic liquids (ILs) have also been recognised as a promising medium for the dehydration of sugars into furans. Zhang et al. have shown the effective conversion of glucose to HMF by using EMIMCL as the solvent and CrClx as the catalyst.134 The same catalyst system was found to be effective for cellulose.135, 136 However, the use of expensive ILs requires a recycling rate of >99 % to be cost effective on an industrial scale. Furthermore, the isolation of HMF from the IL has not yet been demonstrated.

As a variation of several approaches discussed above, Raines et al. have converted lignocellulosic feedstocks to HMF and furfural with good yields by using DMA/LiCl, with the optional addition of small amounts of EMIMCL as the medium and CrClx as the catalyst.137

The coproduction of furfural with HMF or its chloride or alkoxy derivatives have the elegance of allowing the simultaneous upgrade of furfural and HMF, for example, to produce mixtures of (D)MF, (di)methyltetrahydrofuran or diesel hydrocarbons by aldol coupling. Indeed, an integrated process has been proposed to hydrolyse lignocellulose to HMF and furfural and convert the furanic product to MF/DMF by reactive distillation.38 This allows the easy recovery of the desired product from the hydrolysis medium to recycle to the hydrolysis reactor. However, the coproduction of furfural and HMF cannot afford the loss of the costly reaction medium caused by chemical degradation or physical loss, for example, in the solid residue. Minimising solvent loss will probably require dedicated equipment, which will add to the process cost.

5.5. Furfural recovery

Considerable research is also needed to improve furfural recovery from the dilute aqueous product. Distillation, the default option, relies on the (heterogeneous) azeotrope that is formed by furfural and water (atmospheric composition 35 wt % or 10 mol % furfural at 97 °C), which naturally separates into furfural-rich and water-rich phases upon condensation.21 The water-rich phase still contains approximately 8 wt % of furfural that is recycled to the azeotropic distillation column for further furfural recovery. Careful distillation and liquid–liquid separation would limit the steam consumption of furfural recovery to 2–3 t tfurfural−1. However, rapid flashing of furfural from the reaction medium is needed to minimise consecutive degradation. Such flashing is accompanied by excessive evaporation of water and, thereby, requires a steam consumption of approximately 30 t tfurfural−1 as mentioned above. Integrated with low-temperature heat sources and sinks elsewhere in the process, for example, ethanol recovery or feed preheating, could also help to reduce the energy consumption of furfural recovery.

The separation and purification of furfural could be further complicated by the presence of contaminants such as acetic and formic acids, which form additional azeotropes (Table 2) and depress the liquid–liquid separation of the condensed furfural/water azeotrope (Figure 15).138, 139 For example, ternary azeotropes have been reported for acetic acid/formic acid/water and for acetic acid/furfural/water.140, 141

Table 2. Binary azeotropes between furfural, formic acid, acetic acid and water.
SolventBoiling point [°C]Water content
 pure componentwater azeotrope[wt %]
formic acid100.7106.936
acetic acid117.9
Figure 15.

Demixing envelope of furfural/water/acetic acid mixtures. ✶: 20C, ⧫: 25 C, □: 35C. Adapted from Ref. 137.

Furfural recovery by extraction,104, 142145 absorption146148 or membrane separation149, 150 has been investigated as an alternative with the promise of the improvement of energy efficiency. Promising results have been obtained under conditions representative of furfural production, that is, furfural concentration of a few wt % in an aqueous medium. Further investigation of the potential of such technologies seems warranted.

Another intriguing recovery option is presented by CO2-assisted phase separation. Numerous papers report the possibility of separating binary mixtures of water and a water-miscible organic component into two liquid layers by expanding the liquid with a high-pressure gas such as CO2.151, 152 Water/furfural mixtures have been reported to separate by applying 5.5–6.0 MPa CO2.153 We revisited this by using an autoclave that was equipped with an attenuated total reflectance crystal for in situ IR spectroscopy to identify and quantify the liquid–liquid demixing. At room temperature, the saturation concentration of furfural in water was found to decrease slightly with increasing CO2 pressure, from 8 to 7 and 5 wt % upon raising the CO2 from 0 to 1.5–3.5 MPa. However, 5 MPa CO2 did not suffice to reach saturation at 3 wt % furfural. This suggests that pressurised CO2 would be impractical to recover highly diluted furfural from aqueous solutions.

Despite all of the new developments described above, the cost and energy intensity of furfural production and recovery requires significant improvement.

6. Conclusions and Outlook

Furfural can be upgraded to a variety of fuel components. This upgrade generally aims to 1) remove the polarity of the aldehyde group to blend in hydrocarbons and/or 2) reduce the volatility to blend into diesel. Such improvement can be achieved by hydrogenation optionally combined with acid-catalysed rearrangement or etherification, acid–base-catalysed coupling or metal-catalysed decarbonylation. Furfural upgrade can deliver either gasoline or diesel blending components.

On the one hand, monomeric unsaturated furans (e.g., MF and EFE) provide excellent gasoline blending properties, particularly with respect to the octane number. This was confirmed with a road trial run on a 10 vol % blend of MF in gasoline. Saturation of the furan ring is not attractive here, as it deteriorates the gasoline blending properties whilst increasing the capital intensity and CO2 footprint of the biofuel.

Conversely, oligomeric components provide volatility that is suitable for blending in diesel. However, these compounds require ring saturation and, optionally, deep deoxygenation to provide the cetane numbers required for diesel blending. Hence, their production will be more costly and have a larger CO2 footprint than that of monomeric gasoline components.

The furanic gasoline and diesel components that offer the best balance between fuel properties and cost/CO2 footprint are not currently accommodated within existing fuel specifications. Deployment of such fuels will, therefore, require significant effort (and time) in the areas of registration, specification and legislation.

Furfural upgrade technologies still require improvement to produce cost-competitive biofuels. For example, further improvements of selectivity and catalyst lifetime are needed. However, the largest and most challenging improvements are required in the manufacture of furfural. The conventional hydrolysis of lignocellulose proceeds with low yields (≈10 wt % or 50–60 % of the theoretical value) and a high demand for energy. This requires burning most of the remaining biomass to generate power for the manufacturing process, which results in a high investment and feed costs. Coproduction of furfural during the manufacture of LA or HMF or during pretreatment of lignocellulosic materials is also expensive. It generally proceeds with modest yields of pentosans, occasionally requires the use of expensive reaction media (e.g., in HMF manufacture) and considerable energy and investment to recover furfural or its precursor, xylose, from dilute aqueous solutions. All these shortcomings need to be improved on.

Hence, much chemical, catalysis and engineering research is still needed to realise the potential of the furfural platform for biofuel manufacture.

7. List of Abbreviations


equation image(1)


The authors are grateful to numerous colleagues for their assistance in various parts of this study, particularly to K. Von Hebel, R. J. Haan., M. Mackay, L. Meesala, M. Boele, K. Vrouwenvelder, W. Matai, T. Zhang, B. Gols, H. Batjes, A. Felix-Moore, J. Smith, J. Louis, M. Beutler and S. Forbes.

Biographical Information

Dr. Jean-Paul Lange is a principal research scientist at Shell Global Solutions in Amsterdam, the Netherlands, where he has been exploring new catalytic processes in the areas of natural gas conversion, chemical manufacture and, most recently, biofuel production. Before joining Shell, he was a postdoctoral fellow at the Lehigh University (Pennsylvania/US), obtained his PhD at the Fritz–Haber Institute in Berlin (Germany) and graduated from the University of Namur (Belgium). He has (co)authored approximately 40 publications and approximately 40 patent series.

original image

Biographical Information

Dr. Evert van der Heide is a senior researcher and has over 20 years experience in research in catalysis and technology development in Shell Projects and Technology, covering chemicals, biomass pretreatment and bioenergy. He obtained his masters degree in chemistry from the University of Utrecht (the Netherlands) and his PhD from the Technical University of Delft (the Netherlands). He has (co)authored about 40 publications and patent applications on various subjects.

original image

Biographical Information

Dr. Jeroen van Buijtenen studied chemical engineering at the Eindhoven University of Technology. He received his PhD at the same university in 2006 with a thesis on “Tandem Catalysis in Organic and Polymer Synthesis”. In 2007, he joined Shell Global Solutions, where he worked until 2011 as a project leader in biofuels research, focusing on thermocatalytic conversion of lignocellulosic biomass into next-generation biofuels. He is currently working for Shell Chemicals Europe.

original image

Biographical Information

Dr. Richard Price works as a fuels scientist for the Retail and Automotive Fuels Group for Shell Global Solutions Downstream. His work involves the development of new fuel products, including new biocomponents, and covers the areas of base fuel chemistry and performance, vehicle emission testing, additive formulation and side-effects testing. He has a BSc and a PhD in Chemistry from the University of Southampton.

original image