Mechanistic studies on the formation of 5-hydroxymethylfurfural from the sugars fructose and glucose

NMR spectroscopy

This is a tool frequently used in mechanistic studies. Besides the commonly studied ¹H and ¹³C NMR spectroscopies, NMR spectra of several other elements can also be obtained and used, such as ¹⁷O and ¹¹B NMR spectroscopy. Isotopic labeling techniques are often used to trace locations of hydrogen and carbon atoms in compounds and intermediates.

Computational calculations

In recent years, computational chemistry has become increasingly important for investigating reaction mechanisms [26]. The advantages of computational calculations include their convenience to operate and that they can be performed at the same time as lab work. More and more researchers prefer to apply both experimental and computational work in their studies. The theoretical data can provide predictions and explanations for the experimental results, while the lab results provide factual support to the calculations. In the area of biomass transformation, computational studies are normally used to calculate the activation energies of every single step of a putative reaction pathway, analyze their endo- or exo-thermic properties and investigate the stabilities of proposed intermediates.

Besides these two main tools, other methods and tools such as kinetic studies, IR and UV spectroscopies, and mass spectrometry can also be used. In many cases, different techniques have been combined to ensure that the mechanisms proposed are more convincing.

Dehydration of glucose through its isomerization to fructose

In 2011 Matubayasi and co-workers [54] managed to detect all the straight chain and ring isomers of both glucose and fructose through in situ ¹³C NMR analysis of unlabeled and [¹³C-1] glucose in D₂O, indicating that during the isomerization all isomeric forms of glucose and fructose were formed and therefore may play important roles in the mechanism for 5-HMF formation. The interconversion of glucose and fructose is realized through a Lobry de Bruyn–van Ekenstein transformation [55], [56], [57]. NMR data for said species are provided in Table 2.

The study by Zhao et al. is a milestone because for the first time it revealed the interaction of metal halide catalysts with glucose [65]. Mutarotation of glucopyranose was observed in ¹H and ¹³C NMR spectra after the addition of the catalysts. The competition studies using glycerol and glyceraldehyde indicated that it was the hemiacetal portion of glucopyranose instead of the polyalcohol portion that interacted with the catalyst. It was proposed that the metal center interacted with the hydroxyl groups to cause the mutarotation and the subsequent isomerization of glucose, but no experimental work was performed to verify this. A similar study was performed by Zhang et al. using hydroxyapatite supported chromium chloride (Cr-HAP) as the catalyst [66]. The ¹³C NMR analysis provided evidence for glucose mutarotation, and traces of fructose were detected, which came from the isomerization of glucose.

After Zhao’s work, several researchers used similar methods (NMR analysis and competition studies) to explore the mechanisms of glucose dehydration [60, 62, 64]. Different Lewis acids were used as catalysts. In the study by Han and co-workers using tin(IV) chloride (SnCl₄) [62], broadening of the resonances of the hydroxyl groups of glucopyranose was observed in ¹H NMR spectra after SnCl₄ addition, due to the interaction of Cl⁻ ions and the protons of the hydroxyl groups (Scheme 3, 21, MCl_x = SnCl₄). The glucose conversion and 5-HMF yield decreased drastically with the addition of ethylene glycol, indicating that a five-membered ring chelate complex was formed during the reaction. This intermediate was observed in ¹H NMR spectra, coming from the interaction of the Sn center with the 1,2-dihydroxy group of glucose (22). 5-HMF was proposed to be generated either directly from an enediol intermediate (24) or through the fructose produced from 24. In a study by De et al., a ¹H NMR study was carried out on glucose dehydration catalyzed by aluminum(III) chloride (AlCl₃) [64]. A similar broadening of hydroxyl peaks caused by hydrogen bonding was observed. However, a 1,2-hydride shift mechanism was proposed for the subsequent isomerization of glucose, which was different from Han’s work. In a study by Zhang et al., [¹³C-2] glucose was used as the reactant with germanium(IV) chloride (GeCl₄) as the catalyst [60]. The peak at 76.4 ppm in ¹³C NMR spectra during the reaction was assigned to 22 (Scheme 3, MCl_x = GeCl₄). The addition of phenylfluorone decreased the glucose conversion and 5-HMF yield significantly because of its competitive complexation with GeCl₄. An enolization pathway similar to Han’s was proposed for the subsequent conversion of 22 to fructose.

Compared with Zhao’s work, these studies took further steps to identify the intermediates of metal complexation in the isomerization process. However, all the studies discussed above were focused on confirming the complexation of the Lewis acid catalyst with glucose. None of them provided experimental evidence to support the proposed enolization or 1,2-hydride shift mechanisms of glucose isomerization. The isomerization pathways can be studied via ¹H and ¹³C NMR spectroscopy using glucose deuterated at C2 (glucose-D2). If a 1,2-hydride shift occurs, the deuterium at the C2 will transfer to C1 of the fructose or 5-HMF obtained. Theoretically, the fructose should contain 100 % deuterium at C1 and the 5-HMF should be 50 % deuterated at C1. In ¹³C NMR spectra, because of disruption of the nuclear overhauser enhancement (NOE) caused by the deuterium atoms (²H), ¹³C resonance intensities of ¹³C-¹H pairs will be doubly enhanced while the resonance intensities of ¹³C-²H pairs (appear as triplets) will remain unchanged [67]. If enolization happens, the deuterium will be expelled into the solvent so in theory no deuterium will be in the produced fructose and 5-HMF. However, it should be noted that a small amount of deuterium incorporation might be seen because of the proton/deuterium exchange. Conversely, if a deuterated solvent, such as D₂O, is used, 50 % deuterium incorporation at C1 of fructose or 5-HMF is expected for an enolization pathway but not for a 1,2-hydride shift mechanism.

Raines and co-workers performed a mechanistic study on glucose dehydration through ¹H and ²H NMR analysis [68]. Chromium(III) chloride (CrCl₃) or CrCl₂ was used as the catalyst. When one equivalent D₂O was added in the starting materials, less than 5 % deuterium incorporation was detected at C1 of the 5-HMF produced. When glucose-D2 was used, approximately 33 % deuterium was incorporated at C1 of 5-HMF, and a signal of the aldehydic deuteron was observed in the ²H NMR spectrum. Therefore, the isomerization of glucose occurred via a 1,2-hydride shift. Similarly, in the study by Davis’ group of glucose-D2 isomerization using a tin-containing zeolite (Sn-Beta) [69], the ¹H and ¹³C NMR spectra of the fructose product indicated that the deuterium at C2 of glucose-D2 transferred to C1 of the fructose. A KIE (k_H/k_D = 1.98) was observed in the initial reaction rate, further revealing that the 1,2-hydride shift was the rate-limiting step. In the study by Bermejo-Deval et al. [59], a DFT calculation was performed on the Sn-Beta-catalyzed isomerization of glucose through a 1,2-hydride shift. The isomerization catalyzed by a Sn-Beta open-site with one adjacent silanol group had an enthalpy of activation of 92.5 kJ/mol, which was in good agreement with the experimental value (88.7 ± 2.9 kJ/mol). A similar 1,2-hydride shift mechanism was proposed by Choudhary et al. [70] for the isomerization of glucose-D2 using homogeneous (CrCl₃, AlCl₃) and heterogeneous (Sn-Beta) Lewis acid catalysts, after the observation of KIEs and the analysis of ¹H, ²H and ¹³C NMR spectra.

AlCl₃-catalyzed dehydration of glucose was analyzed through a combination of ESI-MS/MS, ATR-IR and NMR spectroscopies by Tang et al. [71]. The ESI-MS/MS spectra indicated that glucose mainly coordinated with [Al(OH)₂(H₂O)₄]⁺ in solution. The linear relationship between log k and log [Al] indicated that glucose isomerization was a first-order reaction with respect to the Al species, so the Al complexes involved in the rate-limiting step should be mononuclear. The addition of glyceraldehyde and 1,3-dihydroxyacetone significantly decreased the glucose conversion and fructose yield while the addition of glycerol made no difference. Therefore, the interaction between the carbonyl group of open-chain glucose and the Al complexes was crucial to initialize the isomerization. Both KIE and NOE effects were observed in the dehydration of glucose-D2, thus a 1,2-hydride shift mechanism was proposed. This conclusion was supported by the study by Shanks and co-workers [72]. When AlCl₃ was used, a significant KIE was observed (k_H/k_D = 1.34) in glucose-D2 conversion.

Hensen’s group performed a series of studies of glucose complexation with chromium chlorides (CrCl₂ or CrCl₃) [73], [74], [75]. In their work about CrCl₂ catalysis [73], extended X-ray absorption fine structure (EXAFS) spectra showed coordination of glucose to Cr²⁺ and the formation of a Cl- or O-bridged Cr dimer during the reaction. A binuclear route with two Cr²⁺ centers was proposed for the 1,2-hydride shift process after DFT calculations of Gibbs free energies of intermediates. Similar studies were performed using CrCl₃·6H₂O as the catalyst [75]. During the initial ring opening of glucose, the mononuclear path was slightly more favored than the binuclear path. Nevertheless, during the following deprotonation of O2H and the 1,2-hydride shift the coordination with two Cr centers effectively decreased the energy barrier. Therefore, similar to Cr²⁺, the binuclear Cr³⁺ complex was preferred for glucose isomerization via a 1,2-hydride shift. However, in the study by Guan et al. [51] of glucose isomerization catalyzed by trivalent metal chlorides (CrCl₃, WCl₃, MoCl₃ and FeCl₃), a mechanism through enolization was proposed after DFT calculations, which is different from Hensen’s studies.

In the study of CrCl₃-catalyzed glucose dehydration in water by Vlachos and co-workers, it was found that the pH of the solution decreased from 2.9 to 1.6 during the process [76]. Therefore, it was assumed that the Lewis acid catalyst produced an intrinsic Brønsted acidity during its dissolution, solvation and further hydrolysis in the aqueous media, and the whole reaction was catalyzed by the combination of Lewis and Brønsted acidity. The speciation modeling of CrCl₃ aqueous solutions with various concentrations was carried out. The calculated pH values at 22 and 140 °C were in good agreement with the experimental data, both of which showed that the intrinsic Brønsted acidity increased with an increase in CrCl₃ concentration. The speciation analysis implied that the main species in the CrCl₃ aqueous solution were [Cr(H₂O)₆]³⁺, [Cr(H₂O)₅Cl]²⁺ and [Cr(H₂O)₅OH]²⁺. With HCl addition, [Cr(H₂O)₅OH]²⁺ was the only species whose concentration kept decreasing. The kinetic experiments revealed that the rate of glucose consumption also decreased with HCl addition. In addition, the initial rate of glucose consumption had a linear relationship with the concentration of [Cr(H₂O)₅OH]²⁺. All these results indicated that [Cr(H₂O)₅OH]²⁺ was the species significantly influencing the isomerization of glucose in aqueous CrCl₃. A 1,2-hydride shift mechanism was proposed and the coordination of glucose to the Cr complex was verified through EXAFS spectra analysis and CPMD (Car–Parrinello molecular dynamics) simulations. The EXAFS spectra showed that in solution the Cr center was originally coordinated to six water molecules, which were replaced then with glucose addition. In the CPMD simulations, initially two unhydrolyzed Cr³⁺ cations were created coordinated to a glucose molecule in water. After the calculation was started at 77 °C, only one Cr³⁺ remained coordinated to glucose at O1 and O2. Therefore, the subsequent simulations were performed with only one [Cr(H₂O)₅OH]²⁺ group, which contrasts with the binuclear Cr complex mechanism proposed by Hensen’s group [75]. It was claimed that the coordination of glucose to the Cr complex promoted the ring opening of glucose for the Lewis acid-catalyzed isomerization; while the intrinsic Brønsted acidity facilitated the subsequent dehydration of fructose to 5-HMF.

Wang and co-workers investigated the influence of organic solvents and Brønsted acid co-catalysts in the CrCl₃-catalyzed dehydration of glucose [77]. The use of DMA led to the highest 5-HMF yield while DMSO resulted in the lowest one. Afterwards, the reaction was performed in DMA and co-catalyzed by several ionic liquids and their corresponding Brønsted acids. The 5-HMF yield decreased when H₂SO₄ was used. The reaction rate increased when HCl or HBr was used but the 5-HMF yield was not improved. Without CrCl₃, the use of HCl or HBr alone resulted in 5-HMF production from fructose but not from glucose. Therefore, it was claimed that these co-catalysts functioned during the dehydration of fructose after glucose isomerization. DFT calculations were performed to probe the interactions among CrCl₃, glucose and the co-catalysts. In the DMA-CrCl₃-glucose system, similar Gibbs energy changes for CrCl₃-DMA and CrCl₃-glucose coordination indicated that they were both favored and the reaction was catalyzed by both CrCl₃ and CrCl₃-DMA. In the DMSO-CrCl₃-glucose system, the extremely low Gibbs energy change of CrCl₃-3DMSO coordination revealed the formation of a stable six-coordinated Cr complex, which inhibited the interaction of the Cr center with glucose for further reaction. Similarly, CrCl₃-3HSO₄⁻ coordination was also quite energetically favored so glucose conversion was prevented. For the halide systems, five-coordinate Cr complexes (CrCl₃-2Cl⁻ or CrCl₃-2Br⁻) were generated, having hemispheric structures so the Cr center was exposed for the interaction with glucose. The isomerization was proposed to occur via a 1,2-hydride shift. When the halides were involved, the activation barrier of the hydride shift process (the rate-limiting step) was slightly decreased but the overall barriers were increased.

Boric acid (B(OH)₃) and its derivatives have been found to be able to form chelate complexes with carbohydrates [78], [79], [80], [81], [82], and catalyze the isomerization of aldohexoses to ketohexoses [83], [84]. The Riisager group achieved the first metal-free dehydration of glucose using B(OH)₃ as the catalyst [85]. The ¹H NMR analysis of glucose-D2 dehydration revealed less than 5 % deuterium incorporation, thus B(OH)₃-catalyzed isomerization of glucose was proposed to proceed via an enolization mechanism (Scheme 4), which was different from the reactions catalyzed by metal Lewis acid catalysts. The isomerization of glucose with and without B(OH)₃ catalysis were calculated through two mechanisms (1,2-hydride shift and enolization), and for both situations the enolization pathway was more favored with lower activation barriers. Caes et al. used a series of phenylboronic acids in the transformation of glucose and cellulose to 5-HMF [86]. Glucose-D2 and D₂O were used, and the ¹H NMR spectra showed that the deuterium from D₂O instead of glucose-D2 was incorporated at C1 of the 5-HMF obtained, further indicating an enolization route similar to that proposed in Riisager’s work. This was also supported by Lukamto et al. [87] through a sophisticated mechanistic study including stereochemical considerations, competition reactions and NMR analyses. Ananikov and co-workers performed the conversion of carbohydrates to 5-HMF using boron trioxide (B₂O₃), and analyzed the reaction mixture through ¹H, ¹³C and ¹¹B NMR analyses [63]. During the reaction, a 1,2-borate complex was detected in both ¹H and ¹³C NMR spectra, and was proposed to be 27. This may not be a true intermediate in the mechanism but a side-product formed from 26 (Scheme 4, R = OH). The function of B₂O₃ in this study was proposed to react with water, which could both facilitate the dehydration via Le Châtelier’s principle and generate B(OH)₃ to promote the reaction.

Scheme 4:

The dehydration of glucose to 5-HMF catalyzed by boronic acids (black) [63, 85–87] Brønsted acids (Qian: red [through isomerization]) [88]; orange (direct dehydration) [89], [90]; Pidko: (blue) [41] and Brønsted bases (green) [91].

As can be seen from the discussion above, in most studies on glucose isomerization the catalysts used were Lewis acids, mainly because of their excellent performance on the catalysis of glucose dehydration to get good 5-HMF yields. There have been some studies reported using Brønsted acid catalysts. Amarasekara and Razzaq performed a NMR study of glucose dehydration catalyzed by a functionalized Brønsted acidic ionic liquid, 1-(1-propylsulfonic)-3-methylimidazolium chloride ([C₃SO₃Hmim]Cl) [58]. [¹³C-1] and [¹³C-2] labeled glucose were used as reactants, and the resonances of C1 and C2 of the open-chain fructose, two intermediates (36 (Table 2) and 2) and 5-HMF were observed in ¹³C NMR spectra. The isomerization was proposed to proceed along a 1,2-hydride shift pathway because the authors claimed that no deuterium incorporation was detected. However, the solvent used herein was DMSO-d₆, which does not release free deuterium into solution. Therefore, it would be more convincing if D₂O was used as the solvent for the deuterium detection experiments.

Because of the harsh conditions (e.g. high temperature, long reaction time) required and the low yields of 5-HMF obtained in lab experiments, Brønsted acid-catalyzed glucose isomerization/dehydration has been studied extensively via computational calculations. In the work by Qian’s group, a cyclic route through a 1,2-hydride shift was proposed for glucose isomerization (Scheme 4) [88]. The CPMD-metadynamics (MTD) simulation results in the gas phase showed that the reaction barrier of the 1,2-hydride shift was as low as 20.9–29.3 kJ/mol because 4 was quite stable. Nevertheless, when the calculation was performed in water, 29 became less stable and the proton of O1H was easily removed by water to form 2′. In the DFT study carried out by Pidko and co-workers [41], the protonation of O5 led to the ring opening of glucose, and the protonated open-chain fructose was obtained via a 1,2-hydride shift (Scheme 4). This mechanism was supported by Daorattanachai et al. [91], but their calculations were performed in gas phase and there was no explanation about their choice of the O5 protonated glucose as the starting point. Therefore, further study is needed for more accurate and reasonable results. In a quantum mechanical study by Arifin et al. [48], both the cyclic and acyclic mechanisms of HCl-catalyzed isomerization of glucose were simulated in water and 1,3-dimethylimidazolium chloride ([Mmim]Cl). The cyclic mechanism was favored in water with an activation barrier of 99.6 kJ/mol, and the acyclic mechanism was favored in [Mmim]Cl with a barrier of 135.6 kJ/mol. In the DFT calculations by Li et al. of [BmimSO₃H]Cl-catalyzed dehydration of glucose, two pathways were proposed [49]. Path I is through the isomerization and Path II is the direct conversion to 5-HMF through an enediol intermediate. It was claimed that both pathways were accessible because they had similar activation barriers. The catalysis by [BmimSO₃H]Cl was achieved mainly through -SO₃H acting as a proton shuttle, and the intermediates and transition states were stabilized via hydrogen bonding with Cl⁻ anions.

A poly-benzyl ammonium chloride resin was used as the catalyst in glucose dehydration by Zhang’s group [92]. An enolization route catalyzed by both the benzyl ammonium cation and the benzyl amine was proposed. The DFT calculations indicated that the whole isomerization process was exothermic, possibly because the intermediates were stabilized through hydrogen bonding. The dehydration of glucose was performed in DMSO/D₂O system and deuterium incorporation at C1 of the 5-HMF product was observed in the ¹³C NMR spectra, hence the enolization path was verified.

The use of Brønsted bases as catalysts in glucose isomerization have also been investigated. The Davis group performed the isomerization of glucose-D2 using sodium hydroxide (NaOH), and no deuterium incorporation was observed [69]. Hence a proton transfer mechanism was proposed, in which the deuterium at C2 of glucose-D2 was transferred into the solution and a proton from the solution was incorporated into the fructose. A Brønsted-base catalyzed isomerization of glucose was simulated by Daorattanachai et al., starting from the deprotonation of O1 of glucose (32, Scheme 4) [91]. The deprotonated open-chain glucose (33) was isomerized to the deprotonated fructose (35) through the deprotonated 1,2-enediol intermediate (34) with an overall activation barrier of 173.7 kJ/mol. Carraher et al. performed the isomerization of glucose using aqueous triethylamine (TEA) [93]. A KIE was observed (k_H/k_D = 3.8) when using glucose-D2 as the reactant, indicating that the deprotonation at C2 was the rate-limiting step. A similar enolization mechanism was proposed.

Direct dehydration of glucose without isomerization to fructose

3-Deoxyglucosone (3-DG, Scheme 3, route A) has been proposed as an intermediate in the dehydration of carbohydrates to 5-HMF [28, 29, 52]. This mechanism has been supported by some early studies [94], [95], [96], [97], [98]. The study by Bols’ group verified this mechanism through a combination of experiments and calculations [53]. In the H₂SO₄-catalyzed dehydration of glucose, both 3-DG and fructose were observed in the ion chromatogram, but higher product yields and reaction rates were obtained from the conversion of 3-DG than fructose. The DFT calculations of glucose dehydration proceeding via 3-DG or fructose were performed. The Gibbs energies of all intermediates in the 3-DG route were lower than those in the fructose route, indicating the preference of the 3-DG route over the fructose one.

The Hara group performed glucose dehydration catalyzed by titanium oxide (TiO₂) and phosphate-immobilized TiO₂ (phosphate/TiO₂) [61]. In the ²H and ¹³C NMR spectra, no deuterium incorporation was detected in the 5-HMF generated from glucose-D2, and incorporation much higher than 54 % was observed for the reaction of glucose-D1.Therefore, a mechanism through 3-DG was proposed. The ¹³C CP/MAS NMR spectra of [¹³C-2] glucose absorbed on TiO₂ and phosphate/TiO₂ at room temperature showed two resonances at 110 and 118 ppm respectively, which were assigned to C2 of 3-DG. The reaction was also performed using scandium triflate (Sc(OTf)₃), and 55 and 54 % deuterium incorporation in the 5-HMF was seen from glucose-D2 and glucose-D1 respectively. This indicated that the Sc(OTf)₃-catalyzed reaction proceeded through the isomerization route via a 1,2-hydride shift.

Qian’s group simulated the Brønsted acid-catalyzed conversion of glucose to 2′ both in vacuum and in water [89], [90]. As discussed above, compound 2 or 2′ was proposed to be an important intermediate in the dehydration of fructose to 5-HMF. Here it was also assumed to be significant in the direct dehydration of glucose to 5-HMF (Scheme 4, orange route). The pathway was initialized by the protonation at different hydroxyl groups or the O atom on the ring (O5), and in both situations (with and without solvent) only protonation at O2H could lead to C2-O2 cleavage and the subsequent formation of 29 (the precursor of 2′) via C1-O5 breaking and C2-O5 bonding. The NMR analyses from previous studies indicated that the downfield shifts of the hydroxyl groups after proton exchange with water decreased in the order O6H > O2H > O3H, O4H > O1H, corresponding to the decrease of their proton affinities [99], [100]. This further supports the calculation results that the O2H protonated glucose was the most likely starting point since O2H has a relatively high proton affinity. This cyclic mechanism of glucose direct dehydration was further verified using a more comprehensive CPMD-MTD method [101]. This reaction route is supported by Assary et al. [40] through Gibbs free energy calculations and Yang et al. [102] through the analysis of reaction fluxes in the kinetic models composed of several glucose reactions (isomerization, direct dehydration, etc.). Assary et al. also proposed an acyclic route, in which O1 of glucose was protonated after the ring opening. However, it should be noted that the protonation of O1H as the first step was rarely seen in computational studies of glucose dehydration.

Zhang et al. performed a DFT study of glucose transformation in supercritical water [103]. An acyclic mechanism without glucose isomerization was proposed. Supercritical water effectively decreased the activation barrier for the ring opening of glucose and the removal of the first two water molecules, but not the third dehydration (the rate-limiting step). Therefore, glucose dehydration was extremely challenging with a high energy demand in supercritical water.

For the transformation of glucose to 5-HMF, both the direct dehydration route and the indirect route through the isomerization to fructose have been investigated extensively. Among the isomerization studies, a 1,2-hydride shift is favored when metal Lewis acids are used as catalysts; while enolization is more dominant for B(OH)₃ and boronic acid catalysts. The Brønsted acid-catalyzed dehydration/isomerization of glucose has been studied more computationally than experimentally, and a variety of mechanisms have been simulated. In the direct dehydration mechanisms proposed, 3-DG has been listed as a key intermediate.