Ionic liquids: not always innocent solvents for cellulose †

The decomposition pathways of a series of carbohydrates dissolved in carboxylate ionic liquids have been investigated in detail using a broad range of thermal and chromatographic techniques. Mixtures of the carboxylate ionic liquid 1-ethyl-3-methylimidazolium acetate with carbohydrates were found to undergo reaction of the C 2 carbon of the imidazolium ring with the aldehyde functionality on the open chain sugar, yielding an imidazolium adduct with a hydroxylated alkyl chain. Subsequently, degradation of the hydroxyalkyl chain occurs by sequential loss of formaldehyde units, to yield a terminal adduct species, 1-ethyl-2-(hydroxymethyl)-3-methylimidazolium acetate. Identities of the ﬁ nal and intermediate adduct species, and the reaction mechanisms connecting adducts, were elucidated by NMR, HPLC and LCMS techniques. Factors a ﬀ ecting the rate and quantity of adduct formation were explored. Changing the ionic liquid cation and anion, the acid number, sugar concentration and temperature in ﬂ uenced the rate of formation and relative quantities of the adduct species. Formation of adducts could not be entirely prevented when employing carboxylate ionic liquids. By contrast, 1-butyl-3-methylimidazolium chloride was identi ﬁ ed as an ionic liquid capable of dissolving a signi ﬁ cant quantity of cellulose, yet without reacting with carbohydrates.


Introduction
Ionic liquids are molten at, or close to, room temperature, and are typically constituted of polyatomic, asymmetrical and charge-diffuse ions. [1][2][3] Ionic liquids are characterised by their high densities and viscosities, [4][5][6] and differ from their neutral small-molecule solvent counterparts by their low vapour pressures. [7][8][9][10][11][12][13][14] Throughout the past twenty years, applications of ionic liquids have become varied and far-reaching, foremost as solvents for sustainable synthetic processes, 1,15 as battery electrolytes, 16 in the capture of CO 2 , [17][18][19] and in the deconstruction of lignocellulosic biomass. [20][21][22][23] The organic halide salts 1-ethylpyridinium chloride, [C 2 py]Cl, 24 and 1-butyl-3-methylimidazolium chloride, [C 4 C 1 im]Cl, 25 are able to dissolve cellulose, and this has initiated a great deal of research into cellulose dissolution procedures. 23 Cellulose, a linear carbohydrate polymer of repeating glucopyranose residues, held together by β-1,4-glycosidic bonds, contributes between 35 and 50% of the dry weight of biomass. It is the most prevalent of the three lignocellulosic polymers, and the world's most abundant renewable feedstock. 26,27 However, numerous strong intermolecular bonding interactions bind together the individual cellulose strands, rendering the polymer insoluble in the majority of conventional solvents; this recalcitrant behaviour of cellulose presents a major challenge to the utilization of this abundant resource. [28][29][30] Existing technologies for the dissolution of cellulose often employ the solvent N-methylmorpholine N-oxide (NMMO) in the 'Lyocell process', 31,32 or phosphoric acid. 33 Alternatively, the 'Viscose process' entails the chemical functionalization of hydroxyl residues along the cellulose backbone with carbon disulphide to form xanthate esters, greatly improving the solubility. Each of these technologies carries a significant drawback; NMMO suffers from poor thermal stability, 34 and the Viscose process generates two kilograms of waste per kilogram of cellulose obtained. 35 Ionic liquids represent a promising alternative to existing cellulose-dissolving solvents because of their higher thermal stabilities relative to NMMO, 34 and the purported non-derivatizing nature of cellulose dissolution with ionic liquids.
The ability of an ionic liquid to dissolve cellulose correlates with hydrogen-bond basicity, β, of the anion; hydrogen bonds between the ionic liquid anion and the cellulose chain are necessary to separate the individual cellulose strands. [28][29][30] Hence, ionic liquids incorporating halide, 25 dialkylphosphate/ dialkylphosphonate 36 and, in particular, carboxylate [37][38][39] anions have been identified as promising candidates for cellulose processing.
The inability to distil ionic liquids on a practical scale, together with their high cost, leads to their recycling being a particular challenge for any proposed large scale application. 40,41 Recycling of a carboxylate ionic liquid may be hindered via a number of plausible pathways: (i) thermal degradation of the ionic liquid and subsequent volatilisation of thermal decomposition products; [42][43][44][45][46][47] (ii) vaporisation of the intact ionic liquid; 11 or (iii) reaction of the ionic liquid with the cellulose. The reactivity of carboxylate ionic liquids towards carbohydrates was first demonstrated by Ebner and co-workers, who described the room-temperature reaction of the imidazolium C 2 carbon of the ionic liquid 1-butyl-3-methylimidazolium acetate, [C 4 C 1 im] [OAc], with the reducing aldehyde end of glucose (acting as a model compound for cellulose), generating an imidazolium adduct bearing a C 2 hydroxyalkyl substituent. 48 They also observed the addition of a fluorescent imidazolium cation to cellulose itself and that the reaction was faster in the presence of base, which was subsequently supported by calculations of Wei et al. 49 Ebner proposed that this reaction was reversible and did not explore any further reactivity of the adduct.
The ability of dialkylimidazolium acetate ionic liquids to form N-heterocyclic carbenes (NHCs) [50][51][52] at the C 2 position of the ring is well established. 42,[53][54][55][56][57][58][59][60][61][62] Transfer of this proton from the cation to the anion is facile, and the NHC may be trapped by addition to CO 2 . 63 The acidic nature of this C 2 proton, with a pK a ∼ 21-23, is likely to play a key role in the reaction mechanisms of carboxylate ionic liquids with solvated cellulose.
Altogether, the reaction of solvated cellulose with an ionic liquid solvent presents potential problems for both laboratory and industrial processes, due to the influence of new chemical species on both the physical and rheological properties of the mixture, and the potential damaging or fibrilation of the reprocessed cellulose material. It is crucial, therefore, to understand factors contributing towards unwanted side reactions, and how they might be prevented or controlled.
Herein, we investigate mixtures of a series of ionic liquids initially with solvated cellulose, and subsequently with simple carbohydrate model compounds, at elevated temperatures similar to those used in processing cellulose and lignocellulosic biomass. Employing the archetypal carboxylate ionic liquid, [C 2 C 1 im][OAc], a sequence of intermediates are identified originating from an initial imidazolium-carbohydrate adduct which undergoes sequential loss of formaldehyde (HCHO) units from the hydroxyalkyl chain to yield the final product of 1-ethyl-2-(hydroxymethyl)-3-methylimidazolium acetate. Factors affecting the rate and extent of adduct formation are explored, including the temperature, the acidity and the carbohydrate concentration. The study was then extended to other ionic liquids in which both cation and anion are changed. One plausible mechanism for formation of the adduct intermediates is proposed.

Experimental
Ionic liquids included in this investigation, 1-11, are shown in Fig. 1. Their syntheses are fully described in the ESI. † Two different samples of 1-ethyl-3-methylimidazolium acetate, [C 2 C 1 im][OAc], 1, were obtained by different synthetic routes, and are herein denoted 1a and 1b. This enables the impact of any residual impurities in either sample to be assessed. Ionic liquids 1a and 3 were synthesized via anion exchange employing commercial 1-ethyl-3-methylimidazolium ethyl sulfate, [C 2 C 1 im][EtSO 4 ], to yield aqueous 1-ethyl-3-methylimidazolium hydroxide, [C 2 C 1 im] [OH]. Subsequent neutralization with the conjugate acid of the desired anion, acetic acid and methanesulfonic acid, respectively, yielded ionic liquids 1a and 3. Ionic liquids 1b and 2 are commercial samples, included for comparative purposes. Ionic liquids 4-10 were made according to known literature procedures. 42,64,65 The synthesis of the mixed inorganic salt, 11, is described in the ESI. † The five carbohydrate model compounds employed in this investigation, 12-16, are shown in Fig. 2, in both cyclic and  acyclic forms where applicable. The carbohydrates were purchased from Sigma-Aldrich in anhydrous form, and were used without further purification.
Ionic liquid-sugar mixtures were initially prepared by the addition of an aqueous solution of the sugar to the neat ionic liquid with known measured water content. It was observed that when anhydrous sugar was added to the neat ionic liquid, new peaks were observable by HPLC of ≤5% of the integration of the parent [C 2 C 1 im] + cation, suggesting that the ionic liquid and sugar had already reacted. By contrast, addition of an aqueous sugar solution to the neat ionic liquid, followed by removal of the water under vacuum at 70°C for one hour, resulted in the formation of significantly smaller peaks (<2% of the integration of the [C 2 C 1 im] + peak), or no peak at all, for new chemical species. Therefore, this second method was employed for preparing the reaction mixtures.
Procedures for preparation and decomposition experiments of the ionic liquid + 5 wt% cellulose mixtures, and the 10 wt% sugar model compound mixtures, are described below. Procedures for the ionic liquid + 25 or 100 wt% sugar model compound mixtures are listed in the ESI. †

HPLC procedures
HPLC experiments were performed on an 'Agilent 1100 Series' HPLC spectrometer, using two 'Sielc' brand 'Primesep 200' (250 mm × 3.2 mm) HPLC columns in series to give improved separation of the charged species. A mobile phase of 60 : 40 vol/vol H 2 O-MeCN + 0.2 mol% H 3 PO 4 , and an injection volume of 6 μl were employed. The temperature was 25°C and the flow rate was 0.5 ml min −1 . Samples were prepared by diluting 0.06 ± 0.02 g of the ionic liquid-sugar mixture in 25 ml of the mobile phase solution in a volumetric flask. Vials were then prepared using approximately 1 ml of this solution. The absolute concentration, in wt%, of the [C 2 C 1 im] + , [C 4 C 1 im] + , 1-ethyl-2-(hydroxymethyl)-3-methylimidazolium, [C 2 C 1 (HO)C 1 2 im] + , and 1-butyl-2-(hydroxymethyl)-3-methylimidazolium, [C 4 C 1 (HO)C 1 2 im] + , cations could be determined from pre-generated calibration curves. For all other species, without calibration data, the percentage of that peak relative to the total integration of HPLC peaks is quoted as the 'HPLC%'. General procedure: ionic liquid + 5 wt% cellulose mixtures

LCMS procedures
Prior to experiments, the ionic liquids were dried under high vacuum with gentle heating of ≤60°C, until the water concentration was found to be below 0.3 wt% as determined by Coulometric Karl Fischer titration. The ionic liquid-cellulose mixtures (20 : 1 w/w) were prepared by the addition of dry 'Sigmacell' cellulose (0.05 ± 0.001 g) to a sample of the ionic liquid (1.00 ± 0.01 g) in a 100 ml glass pressure tube containing a magnetic stirrer bar. The pressure tube was then partially submerged into an oil bath with integrated thermostat, set at a constant temperature of 120°C, and was maintained, with stirring, at this temperature for 48 hours. The apparatus was positioned behind a protective blast screen as a safety precaution. Subsequently, the mixture was allowed to cool to room temperature and deionized water (∼0.1 ml) was added to precipitate cellulose. The suspension was filtered to remove solid material, and the filtrate was examined by 1 H NMR spectroscopy.
General procedure: ionic liquid + 10 wt% sugar mixtures Ionic liquid-sugar mixtures (10 : 1 w/w) were prepared by the addition of an aqueous sugar solution (1.00 ± 0.01 g in 5 ml deionized water) to a portion of the ionic liquid (10.0 ± 0.1 g) in a 50 ml round-bottomed flask (the two liquids were readily miscible). Water was then removed by rotary evaporation for one hour; a pressure of 8 mbar was employed, and the water bath was set to a temperature of 70°C. The sample was analysed by Karl Fischer titration in order to confirm the water content was <5 wt%. In addition, an HPLC chromatogram was obtained, so as to confirm that no significant decomposition had occurred prior to the primary heating period. The roundbottomed flask was then fitted with an adaptor for an oil bubbler and was partially submerged in an oil bath with integrated thermostat and magnetic stirrer function, set at a constant temperature (120 or 100°C). The mixture was maintained at this temperature for 24 hours under a gentle flow of nitrogen gas, with a stirring rate of 150 rpm. Small aliquots (0.06 ± 0.02 g) were removed at regular intervals ('t x ', where x = 0, 0.25, 0.5, 1, 1.5, 2, 3, 4, 6 and 24 hours) and were diluted with 25 ml of the mobile phase solvent. Small portions of each solution were measured into vials for HPLC analysis.
The temperature of the oil bath fluctuated by no more than ±2°C during the course of the experiment. 1 H NMR spectra were recorded after the 24 hour heating period. In addition, the water content was measured at the end of the 24 hour heating period.

Results and discussion
Characterisation of ionic liquid-sugar adducts Following the key discoveries of Ebner 48,66 and Wei, 49 the reactivity of cellulose with a variety of available ionic liquids was initially investigated. Nine dialkylimidazolium ionic liquids, 1-9, and one ionic liquid with the organic N-methyl-diazabicycloundecenium cation 10 were selected. This was in order to cover a broad range of hydrogen bond basicities, β (a key determinant in cellulose solubility), including compounds that are known good solvents for cellulose, 1, 2, 7 and 8, and equally those that are not, 3-6 and 9. Mixtures of an ionic liquid, 1-10, with 5 wt% of cellulose added, were prepared according to the method detailed in the Experimental section above, and were heated at 120°C for 48 hours. Cellulose has only a very low solubility in some ionic liquids (e.g. [3][4][5], and in these cases mixtures took the form of suspensions rather than solutions. Following the heating period, the mixtures were analysed using 1 H NMR spectroscopy (shown for 2 in graphical form in the ESI, Fig. E2 †) to assess formation of new chemical species.
Interestingly, new peaks were observed in the 1 H NMR spectra for the mixtures incorporating both of the carboxylate ionic liquids, 1 and 2. For dimethyl phosphate ionic liquid 8, tiny peaks were observed, which were more significant after extending the heating period to one week. Of particular note, a singlet at δ 4.79 ppm (in DMSO-d6) was observed for the mixtures of cellulose with liquids 1 or 2, and at δ 4.73 ppm for the mixture with [C 4 C 1 im] + ionic liquid 8. By contrast, the mixtures incorporating ionic liquids with methanesulfonate, 3, hydrogen sulfate, 4, bis(trifluoromethanesulfonyl)imide, 5, halide, 6 and 7, and triflate, 9, anions exhibited no new peaks in the NMR spectrum. The carboxylate ionic liquid [Me-DBU][OAc], 10, did exhibit new peaks, although they originated from degradation of the ionic liquid itself, and not primarily from direct reaction with the cellulose.
In order to explore the reaction between the ionic liquids and cellulose more closely, and to elucidate factors affecting the formation of new chemical species, cellulose was replaced with a series of smaller carbohydrate model compounds. Mixtures were prepared with ionic liquids 1, 2, 3, 7 and 8 (all good cellulose-dissolving ionic liquids, with the exception of 3), and model compounds 12-16. The mixtures were heated to 100-120°C, temperatures typical of industrial cellulose dissolution processes, and changes in chemical composition were monitored by HPLC, LCMS, and 1 H NMR techniques. In addition, the acid number was measured at t 0 and t 24 (0 and 24 hours heating) for some of the mixtures.
Initially, mixtures of the ionic liquid [C 2 C 1 im][OAc], 1a, with 10 wt% of D-(+)-glucose, 12, were heated at 120°C for 24 hours. Periodically, aliquots of the mixture were analysed by reverse-phase HPLC, the traces of which demonstrated the appearance of new small peaks of higher polarity (lower retention time) than the parent [C 2 C 1 im] + ion peak. Moreover, the absolute concentration of [C 2 C 1 im] + reduced over the course of the 24 hours, suggesting that the new peaks were derived from the ionic liquid cation (Fig. 3a).
LCMS was employed to identify the chemical species responsible for these new peaks (Fig. 4). Notably, the peak of highest polarity exhibited a strong, single mass signal at m/z 291, equal to the mass of [C 2 C 1 im] + + D-(+)-glucose, 12. This species was present in fairly small quantity (<2 HPLC%) before the 120°C heating period. Therefore, this species was assigned as the equivalent adduct to that observed by Ebner and coworkers, 48 formed from the reaction at the ring C 2 imidazolium substituent with the reducing end of the sugar molecule. This adduct is herein denoted as the 'C6' adduct, referring to the six-carbon hydroxyalkyl substituent at the C 2 position of the imidazolium ring. The same notation, 'Cn' is used here- after to refer to other observed adducts, where 'n' corresponds to the number of carbons in the C 2 substituent chain, excluding the C 2 carbon itself.
The other new HPLC peaks were identified, all with lower masses than the C6 adduct, each separated by increments of 30 m/z. This mass corresponds to a difference of a one-carbon CH 2 O unit in the hydroxyalkyl backbone, or a formaldehyde molecule, HCHO. Hence, C4, C3, C2 and C1 adducts were observed at m/z 231, 201, 171 and 141, respectively ( Fig. 3b  and 4). The C5 adduct was absent from both the HPLC and LCMS spectra, and there are several possible explanations for this observation: (i) the C5 adduct is not formed at all; (ii) the C5 adduct does form but is unstable and rapidly converts to a smaller adduct; or (iii) the C5 is present but the retention times of the C6 and C5 adducts are so similar that the two could not be resolved. HPLC% concentrations of the C6-C2 adduct peaks were plotted as a function of time (Fig. 3b).
For confirmation of its identity, the C1 adduct compound, 1-ethyl-2-(hydroxymethyl)-3-methylimidazolium acetate, 17, was directly synthesised from the reaction of [C 2 C 1 im][OAc], 1a, with paraformaldehyde at 80°C for 24 hours (Fig. 5). The 1 H NMR spectrum of the synthesised compound 17 perfectly matched the new peaks observed from the mixture of 1a and 12, after 24 hours of heating at 120°C (Fig. 6), and the HPLC retention times matched. Therefore, the 'C1' adduct observed in the HPLC experiments, and at m/z 141 in the LCMS experiments, was confirmed as [C 2 C 1 (HO)C  67 Moreover, these ionic liquids were shown to be stable under basic conditions. By retroactively examining the preliminary NMR spectra for the mixtures of ionic liquids with cellulose, the observed singlet peak at δ 4.79 ppm for compounds 1 and 2, and at δ 4.73 ppm for 8, can be assigned to the respective C1 adducts. HPLC calibration curves were constructed; subsequently, the wt% of the C1 adduct was displayed alongside [C 2 C 1 im] + (Fig. 3a).
The change in the HPLC% and wt% of the C6-C1 adduct species reveals key information about their relative stabilities    (Fig. 3). The C6 adduct is generated rapidly, reaching the maximum concentration at t 0.25 before gradually disappearing. No C6 adduct is present after 24 hours (t 24 ), so it has been completely converted into further products. By contrast, the C1 adduct only reaches maximum concentration by t 6 , and is not diminished after 24 hours. Indeed, the C1 adduct appears to be metastable at the operating temperature of 120°C, with no net loss or gain in concentration between the t 6 and t 24 time points (Fig. 3a). The comparatively slow and gradual formation of the smaller adducts (C2 and C1), coupled with the early rise and subsequent decrease of the larger adducts (C6, C4 and C3) indicates that the stability of the adducts increases with decreasing C 2 -substituent size. Moreover, the concentrations as a function of time, t, suggest that the C6 adduct forms initially; sequential elimination of CH 2 (Fig. 7). The experiments revealed a gradual increase in acidity of the system during the 24 hour heating period, from ∼100 to 800 mmol H + kg −1 IL. The curvature of the two plots in Fig. 7 indicates that increasing acid concentration is, to some extent, linked to the formation of the C1 adduct. However, there is a notable increase of approximately 100 mmol H + kg −1 IL between t 6 and t 24 , whereas the C1 adduct cation has reached its maximum concentration by t 6 . Moreover, acid number measurements on synthesised [C 2 C 1 (HO)C 1 2 im][OAc], 17 ( Fig. 5), reveal that it is not itself responsible for the increase in acid number. The number of moles of D-(+)-glucose, 12, incorporated into the mixture (∼505 mmol kg −1 IL), is marginally lower than the number of moles of [C 2 C 1 im] + consumed (∼595 mmol kg −1 IL). The moles of C1 adduct present in the t 24 mixture (∼450 mmol kg −1 IL), account for approximately 75% of the moles of [C 2 C 1 im] + consumed. [C 2 C 1 im][OAc] exhibits slow decomposition at 120°C, 42,68 which will contribute partially to the loss of [C 2 C 1 im] + . There is a measurable quantity of the C2 adduct (∼1.2 HPLC%) still present in the mixture after 24 hours heating (Fig. 3b). A longer heating period would likely lead to total conversion of the C2 adduct to the C1 adduct. Therefore, it is not certain whether a stoichiometric 1 : 1 : 1 reaction is occurring, but it is possible that one mole of the [C 2 C 1 im] + cation reacts with one mole of D-(+)-glucose (12), initiating a sequence of reactions that eventually yields approximately one mole of the C1 cation, [C 2 C 1 (HO)C 1 2 im] + .
However, it is highly plausible that the C1 adduct exists in an equilibrium with the parent [C 2 C 1 im] + cation. By taking t 24 mixtures of 1a + 10 wt% D-(+)-glucose, 12, and heating them at the higher temperature of 180°C for four hours, a detectable reduction in concentration of [C 2 C 1 (HO)C 1 2 im] + and a gradual increase in the concentration of the original [C 2 C 1 im] + cation occurred. This indicates that the equilibrium is forced towards the ionic liquid side, with the evaporation of formaldehyde at this higher temperature.
The molar increase in acid number (∼700 mmol H + kg −1 IL, Fig. 7), was greater than stoichiometric. The sequence of adduct-forming reactions between 1a and 12 appeared to terminate in an equilibrium concentration of the 'C1' adduct, [C 2 C 1 (HO)C 1 2 im] + ; at 120°C; the maximum quantity was reached by t 6 and had not diminished by t 24 (Fig. 3a). In order to assess its thermal stability, the synthesised sample of [C 2 C 1 (HO)C 1 2 im][OAc], 17, was further analysed at higher temperatures. Temperature-ramped Thermogravimetric Analysis (TGA) experiments were performed on ionic compound 17, in the temperature range 80-700°C, using a heating rate of 10°C min −1 . The TGA data is represented in graphical form in the ESI (Fig. E5a †). Complete TGA experimental conditions are described in the ESI. † The T onset temperature for compound 17, 221°C, was very similar to that of unsubstituted ionic liquid 1b, measured as 216 ± 2°C employing identical experimental conditions. The derivative weight curve did not reveal any distinguishable weight loss event corresponding to loss of the C 2 -hydroxyalkyl substituent.
Recently, we investigated the long-term thermal stability of the typical carboxylate ionic liquid, [C 2 C 1 im][OAc], 1, and established Arrhenius parameters for thermal decomposition of this compound. 42 Herein, we expanded this investigation to the C1 adduct compound, [C 2 C 1 (HO)C 1 2 im][OAc], 17, recording an isothermal TGA thermograph for 17 at 120°C, for 24 hours (Fig. E5b †). Similar to ionic liquid 1, compound 17 yielded a straight isotherm, suggesting pseudo zerothorder kinetics. Comparing rates of thermal decomposition (change in molar proportion over time, dα/dt ), at the temperature of 120°C, degradation of ionic liquid 17 is marginally faster (dα/dt = 7.3 × 10 −3 h −1 ) than that of parent ionic liquid 1 (dα/dt = 5.7 × 10 −3 h −1 ), because 17 is likely to incorporate a contribution to mass loss from both the regeneration of [C 2 C 1 im] + and loss of formaldehyde, and from subsequent thermal decomposition of [C 2 C 1 im][OAc], occurring simultaneously.
Hence, attempting to regenerate [C 2 C 1 im] + simply by high temperature treatment is not a suitable strategy for preventing the accumulation of the C1 adduct cation [C 2 C 1 (HO)C 1 2 im] + in solutions of carbohydrates in carboxylate ionic liquids.

Elucidation of adduct-forming mechanisms
A proposed reaction mechanism for the formation and interconversion of the observed intermediates is shown in Fig. 8. It accounts for the key observations: (i) the concentration of the ionic liquid cation, [C 2 C 1 im] + , is markedly reduced following the 24 hour heating, and D-(+)-glucose, 12, is no longer present in the 1 H NMR spectrum. Therefore, adducts are formed from reaction of [C 2 C 1 im] + with 12; (ii) C6, C4, C3 and C2 adducts were all observed as intermediates and, moreover, their concentrations as a function of time suggest that the smaller adducts appear later and are more stable than the larger analogues; (iii) the concentration of the C1 adduct, [C 2 C 1 (HO)-C 1 2 im] + , appears to reach an equilibrium by the end of the 24 hour heating; (iv) the successive reactions are accompanied by a progressive increase in the acid number of the mixture, although the C1 adduct is not itself the source of H + , and the increase in mmol H + is not stoichiometric.
The postulated mechanism involves initial abstraction of the C 2 imidazolium proton by the basic acetate anion generating an NHC, which undergoes nucleophilic addition to the formyl group of glucose in its open chain form. The C6 adduct that forms then cleaves a five-carbon aldehyde fragment to form the C1 adduct via an established 'Breslow' intermediate. 69 Subsequent recondensation of the liberated aldehyde fragment with a further ionic liquid ion pair yields the next homologue of the series, and reaction repeats to account for each of the observed C4, C3 and C2 intermediate adducts.
Thus, in effect, the reaction constitutes gradual digestion of the carbohydrate by releasing one-carbon formaldehyde (HCHO) fragments in turn, bound up in the form of the C1 adduct species.
One apparent disparity between the postulated mechanism and the experimental observations is that reaction of one mole of D-(+)-glucose, 12, should eventually generate six moles of the C1 adduct species. This predicted quantity is far greater than the quantity observed. However, as discussed above, there is likely to be an equilibrium between this adduct and the parent ionic liquid cation. At the temperatures of these reactions (100-120°C), it is probable that formaldehyde will largely be lost from the reaction mixture by evaporation. Moreover, it has so far been assumed that all the consumed glucose is via this mechanistic pathway. Formation of various 'caramelisation' products from the burning of the sugar, and other reactions, may account for some substantial loss of glucose. These products would not be clearly observed by spectroscopic means, yet the darkening of the reaction mixtures would support this hypothesis. Each of these two explanations would contribute to the disparity between the expected and observed quantity of [C 2 C 1 (HO)C 1 2 im] + in the t 24 mixtures.
The increase in acid number could, in principle, arise from oxidation of the liberated formaldehyde into formic acid, but no peak for this was observed in the 1 H NMR spectrum for the mixture of 1a + 12 after the 24 hour heating period. Another possible cause for the increase in acid number could be the oxidation of C1 into the 1-ethyl-3-methylimidazolium-2-carboxylate cation, a species known to form between the [C 2 C 1 im] + cation and CO 2 . 63 This acidic zwitterionic species is present in the HPLC spectrum of the mixture of 1a + 12 prior to heating, in low concentration (<1 HPLC%), as a residual impurity from the ionic liquid synthesis and present in the HPLC of the pure ionic liquid. A very small increase (+0.36 ± 0.04 HPLC%) in the concentration of this cation was observed over the course of the 24 hour heating period. However, this is unlikely to be sufficient to account for all of the measured acid number increase of the experiments.

Reaction with other sugar compounds
The investigation was extended to the other sugar model compounds, 13-16 (see ESI †). The formation of adduct species was observed for mixtures of 1a with D-(+)-cellobiose, 13,  In contrast, no adduct species were observed from HPLC analysis of the mixture 1a + 10 wt% D-(+)-sucrose, 16, heated to 120°C for 24 hours. Instead, only a slight darkening of the mixture was observed, and the wt% of the [C 2 C 1 im] + cation remained constant. Critically, D-(+)-sucrose, 16, is the only studied carbohydrate model compound lacking an open chain form. Therefore, we conclude that it is necessary for the saccharide to have an open chain structure for reaction with [C 2 C 1 im] + to occur. D-(+)-Cellobiose, 13, is an improved model compound for cellulose relative to D-(+)-glucose, 12, due to the presence of the two glucopyranose units held together by a β-1,4-glycosidic link. Upon heating [C 2 C 1 im][OAc], 1a, with D-(+)-cellobiose, 13, at 120°C, adducts of higher polarity than [C 2 C 1 im] + were observed by reverse-phase HPLC, as for the mixtures of 1a with 12. Several of the components present after 0.25 hours of heating were analysed by LCMS (Fig. E6 †). The peak of highest polarity by HPLC was found to represent three separate chemical species in the LCMS. Three sharp signals were observed in the mass spectra, at m/z 453, 423 and 393 (each differing by m/z 30, CH 2 O). Using a similar nomenclature to products of 1a with D-(+)-glucose, 12, these masses were assigned to the 'C12', 'C11' and 'C10' adducts, respectively (Fig. 9). The C11 adduct, observed when studying mixtures with D-(+)-cellobiose, is analogous to the unseen C5 adduct in the experiments with D-(+)-glucose. Therefore, it is likely that the C5 adduct does form, but that the rapid conversion of adducts, or similarity in the HPLC retention times, explains why it is not formally distinguished.
A small peak of the same retention time for the C6 adduct from the glucose experiment (m/z 291) was also observed, suggesting that both glucopyranose residues of 13 had reacted with [C 2 C 1 im] + to eventually yield [C 2 C 1 (HO)C 1 2 im] + . This is directly supported by the formation of measurable quantities of the C1 adduct compound in the initial experiments with 5 wt% cellulose. It remains unclear whether the reaction of both of the glucopyranose units of cellobiose occurs independently, or after a single addition to the imidazolium ring. Nevertheless, these results carry important implications for the dissolution of cellulose in carboxylate ionic liquids; adduct-forming reactions are likely to extend beyond the terminal glucopyranose residue, and employing carboxylate ionic liquids in the dissolution of cellulose for long periods of time will bring about a reduction in the degree of polymerisation of the cellulose, a diminished quality of the cellulose fibres, and the gradual accumulation of unwanted by-products.
Similarly, the mixture of 1a + 10 wt% D-(−)-fructose, 15 (a structural isomer of 12) demonstrated new adduct peaks, which were assigned to the expected Cn (n = 1-6) adducts, on the basis of the nearly identical pattern of HPLC peaks compared to the D-(+)-glucose experiment. The C1 adduct was assigned unambiguously from the 1 H NMR data.

Reduction in temperature
Although 120°C is a very commonly used temperature for biomass and cellulose processing, 23 it would be expected that the rates of by-product-forming reactions would be lower at reduced temperatures. The investigation of reactions of 1a + 10 wt% 12/13 were repeated at the reduced temperature of 100°C. A comparison of the quantities of C6 and C1 adduct compounds (in HPLC% and wt%, respectively), between the two temperatures is displayed for 1a + 12 in Fig. 10. A comparison of the intermediate C4, C3 and C2 adduct species is shown in the ESI (Fig. E4 †).
Clearly, the same chemical species are formed at the lower temperature. The initial formation of the C6 adduct remains rapid at 100°C (Fig. 10b), but the subsequent reactions are far slower, exemplified by the sluggish accumulation of the C1 adduct cation [C 2 C 1 (HO)C 1 2 im] + ; at 120°C, the maximum concentration is reached by t 6 , whereas at 100°C this is not reached by t 24 (Fig. 10a). Thus, reducing the temperature of the cellulose solvation system appears to be a sensible strategy for diminishing the rate of by-product formation, although the generation of byproducts is by no means entirely halted. However, the lower temperature limit of an industrial cellulose dissolution process will likely be dictated by the rate at which cellulose dissolves. Moreover, high viscosities of ionic liquid-cellulose solutions present a more significant problem at lower temperatures. Therefore, another process variable must be modified.

Starting acidity
The acid number of the mixture at the outset of the reaction was investigated as another possible variable which could be modified to limit reaction of the ionic liquid with sugars. Samples of ionic liquid [C 2 C 1 im][OAc], 1a, were treated with small aliquots of acetic acid, before being blended into mixtures with the addition of 10 wt% D-(+)-glucose, 12. The acid numbers of the mixtures were then determined (110-271 mmol H + kg −1 IL), and the mixtures were heated at 120°C for a period of 24 hours with HPLC analysis at the same regular time points (t 0 -t 24 ) as for the experiments described above. Complete acid number data is listed in the ESI (Table E1b †). A comparison of the rate of formation of C6 and C1 for mixtures of 1a + 10 wt% 12, at differing initial acid numbers, is shown in Fig. 11.
Increasing the acid number of the mixture [C 2 C 1 im][OAc], 1a + 10 wt% D-(+)-glucose, 12, reduces the rate of formation of [C 2 C 1 (HO)C 1 2 im] + to a small extent, and this is most pronounced at the middle time points, t 1 -t 4 (Fig. 11a). A corresponding reduction in the rate of disappearance of the C6 adduct is observed at higher acid numbers; maximum concen-tration of the C6 adduct appears to have occurred before t 0.25 for the experiment with acid number 110 mmol H + kg −1 IL, whereas maximum concentration is nearer t 0.5 for the experiment with 271 mmol H + kg −1 IL. Nevertheless, differences in the rate of adduct formation appear to be minimal, at least for the acid number range and reaction mixture we have studied. A key implication of these results is that the mechanism of formation of the initial adducts, as well as their subsequent inter-conversion, is not significantly acid-catalysed. Regardless of the initial acid number, the measured increase in acid number was approximately equivalent over the 24 hour heating period, at +750 ± 50 mmol H + kg −1 IL.

Sugar concentration
Subsequently, mixtures were studied with [C 2 C 1 im][OAc], 1a, and higher concentrations of sugar model compounds D-(+)-glucose, 12 and D-(+)-xylose, 14 (25 wt% and 100 mol%, relative to the ionic liquid), in an effort to prepare and isolate intermediate adducts for structural analysis. Mixtures were prepared in an analogous way to the 10 wt% mixtures, by  addition of an aqueous sugar solution to the neat ionic liquid, followed by drying of the resultant solution under reduced pressure for one hour at 70°C to yield a highly viscous liquid. Full procedures are described in the ESI. † Graphs representing the change in wt% of the [C 2 C 1 im] + and [C 2 C 1 (HO)C 1 2 im] + cations as a function of time, for the 25 wt% mixtures, are displayed in the ESI (Fig. E3 †). Upon increasing the initial sugar quantity in the example of [C 2 C 1 im][OAc], 1a + 25 wt% D-(+)-glucose, 12, the pattern of observed adducts was equivalent to the former mixtures. Changing the ionic liquid 1-Ethyl-3-methylimidazolium acetate, [C 2 C 1 im][OAc], 1, has most commonly been used for the dissolution of cellulose and is highly effective. However, other ionic liquids have also been used. The solubility of cellulose in an ionic liquid is dependent on hydrogen-bond basicity, requiring β > 0.8 for dissolution to occur. 23,70 However, the same basic behaviour of the ionic liquid anion has been associated with generating an Nheterocyclic carbene (NHC) from a dialkylimidazolium cation. 1 We propose that this is the property that initiates the sequence of reactions culminating in formation of [C 2 C 1 (HO)C 1 2 im] + (Fig. 8).
The two ionic liquids with lowest values of β, 3 and 7, exhibited no reaction between the cation and D-(+)-glucose. Furthermore, the change in acid number for these mixtures, where measured, was minimal (∼±10 mmol H + kg −1 IL). The absolute concentration of the [C 2 C 1 im] + or [C 4 C 1 im] + cation was unchanged in each example after the 24 hour heating period (Fig. 12), and the colour changes of these mixtures were far less substantial than had been seen for the mixtures with [C 2 C 1 im][OAc], 1.
Hence, [C 2 C 1 im][CH 3 SO 3 ], 3, and [C 4 C 1 im]Cl, 7, appear to be inert solvents with respect to the carbohydrate model compound D-(+)-glucose, 12, at the temperature of 120°C. Crucially, [C 4 C 1 im]Cl was the first recognised example of a cellulose-dissolving ionic liquid; 25 therefore, this apparent lack of adduct-forming reactions is highly significant.
By contrast, the mixture of [ In summary, the screening of different ionic liquid cation and anion species, with respect to their reactivity with D-(+)-glucose, 12, highlighted several interesting phenomena. For dialkylimidazolium based ionic liquids, the high β value that is required to dissolve cellulose seems also to lead to reactions between the cation and sugar compounds. However, the reactions in the ionic liquid with the highest value of β, [C 4 C 1 im][(CH 3 ) 2 PO 4 ], 8 (β = 1.13) were slower than in the carboxylate ionic liquids. Dimethyl phosphate, (CH 3 ) 2 HPO 4 , has a pK a of 1.29, relative to 4.76 for acetic acid and 4.89 for octanoic acid. 74 pK a values of hydrogen chloride and methanesulfonic acid, both strong acids, are ≪1. Therefore, Brønsted basicity may be a better measure of the likelihood of adduct formation than hydrogen-bond basicity, β, measured from Kamlet-Taft experiments. This is worthy of further study.
Recent investigations have shown the tendency of chloride ionic liquids to cause hydrolytic cleavage and degradation of cellulose during dissolution processes, 75-77 yielding a mixture of cellooligosaccharides, cellobiosan and glucose when water content is above a certain threshold. 78 Careful exclusion of water (concentration <0.3 wt%), and the high temperature of our experiments (120°C rather than 100°C, ensuring more water was in the vapour phase) are the likely explanations for the absence of these hydrolysis reactions being observed for our mixtures of ionic liquids with cellulose.
The precise explanation for the differences in reactivity of carboxylate (1 and 2), chloride (7) and dimethyl phosphate (8) ionic liquids with D-(+)-glucose, 12, is unknown. However, what is striking is that anhydrous [C 4 C 1 im]Cl, 7, an effective solvent for cellulose (albeit less so than [C 2 C 1 im][OAc], 1), does not exhibit the undesirable sequence of adduct-forming pathways occurring for anhydrous carboxylate ionic liquids.

Conclusions
Dialkylimidazolium ionic liquids incorporating carboxylate anions react with cellulose and low molecular weight sugars to generate a series of intermediates leading eventually to cations with a hydroxymethyl substituent at the C 2 position of the ring, e.g. the 1-ethyl-2-(hydroxymethyl)-3-methylimidazolium cation, [C 2 C 1 (HO)C 1 2 im] + . Substituting with the dimethyl phosphate anion slows, but does not prevent, these reactions with the model sugar compounds. By contrast, the investigated chloride and methanesulfonate ionic liquids are sufficiently less reactive towards the model sugar D-(+)-glucose that no measureable quantity of reaction products was observed after 24 hours at 120°C. However, of these two, only the chloridebased ionic liquid can dissolve cellulose to a significant concentration.
The series of reactions begins with the addition of the aldehyde of the open chain form of the sugar to the NHC of the cation. This is then followed by sequential elimination of formaldehyde units until the final product is reached. The observed reactivity of D-(+)-cellobiose demonstrated that these reactions are not limited to the terminal glucopyranose residue. Although the rate of formation of these by-products may be partly reduced by lowering the operating temperature or increasing the initial acidity of the system, their formation cannot be entirely prevented.
Given the preference for carboxylate-anion ionic liquids in the dissolution of cellulose and particularly the popularity of [C 2 C 1 im][OAc], this behaviour has the potential to prevent successful implementation of cellulose dissolution processes employing these ionic liquids. The accumulation of ionic liquid derived by-products will affect rheological properties and prevent the crucial recycling of the expensive ionic liquid component, greatly increasing process costs. The degradation, fibrilation and shortening of the cellulose fibres are also key concerns for the reduction of the quality and quantity of the cellulose product.
One obvious strategy would be the modification of the ionic liquid cation. The adduct-forming reaction mechanism for [C 2 C 1 im][OAc] clearly involves the reactive C 2 position of the imidazolium ring. Unfortunately, simply substituting this proton for a methyl group yields an ionic liquid with a higher viscosity, higher melting point and limited thermal stability. 42 Moreover, recent investigations have highlighted the importance of the cation in cellulose dissolution, 79-81 when previously it was considered to have only a secondary role. Therefore, whilst modification of the ionic liquid cation may provide a feasible solution to the undesired adduct-forming reactions, it is not a trivial problem.
When employing dialkylimidazolium ionic liquids, anions of sufficiently low basicity are required to inhibit formation of N-heterocyclic carbenes. However, such anions also lead to a reduction of the solubility of cellulose in the ionic liquid. Until alternatives can be identified, only anhydrous ionic liquids incorporating the chloride anion have been shown to be able to both dissolve cellulose and to avoid undesirable reaction of the dialkylimidazolium cations with the cellulose. Further investigation into the relationships between ionic liquid structure and reactivity towards cellulose is required.