To evaluate the thermal stabilities of ILs used in this study, thermogravimetric analysis (TGA) was performed in the absence of cellulose. The results are shown in Figure 2. The thermal stability was associated significantly with the anionic part of ILs and less with the counterpart cation. The ILs with sulfonate anions and non- or weakly-coordinating anions (BF₄⁻, (CF₃SO₂)₂N⁻ and PF₆⁻) showed higher thermal stabilities than the others. This trend agrees with the data reported in the work by Fredlake et al. [17], in which they concluded that the thermal stability of IL increases with the increase in anion size. The ILs with [BMMIM] were more stable than the ones with [BMIM] because methyl substitution of C(2)-H of [BMIM], which is responsible for the acidity, can stabilize the cation against decomposition [18].

In general, a significant weight loss of cellulose, in other words, formation of volatiles, in the course of heating up for pyrolysis starts from around 300 °C. From that viewpoint, TGA results can be a reason enough to exclude ILs with Cl⁻, CF₃CO₂⁻ and HSO₄⁻ anions from the candidate as a catalyst for cellulose pyrolysis because they started to lose the weight under 300 °C by the decomposition. The boiling point of levoglucosenone, 254 °C under ambient pressures, also limits the types of ILs usable for the reaction. They must be stable at least as high as the boiling point to separate levoglucosenone from IL phase by the vaporization.

Figure 3 shows results of TGA for cellulose/IL mixtures. As with the thermal stability of IL, characteristics of weight loss curves depended on the types of anions with little influence of the cation. The curves could be sorted into two types: (i) clearly separated two-step weight loss and (ii) monotonous or unclear two-step weight loss. The type (i) applied to all the curves of the mixtures with sulfonate ILs (Figure 3b) and [BMMIM]BF₄ and [BMMIM](CF₃SO₂)₂N. The two-step weight loss can be explained by the pyrolysis of cellulose for the first step (on the lower temperature side) and the decomposition of IL for the second step (on the higher temperature side) [16]. The isolation of the temperature ranges for the two steps indicates the possibility of IL recovery from the residue of pyrolysis if the pyrolysis temperature is lower than the onset temperature for the decomposition of IL. The temperature ranges for the second step in the type (i) curves corresponded well to those for the thermal decomposition of respective ILs in Figure 2. This represents that the thermal stabilities of those ILs were not altered by the presence of cellulose. On the other hand, the curves of type (ii) showed that the pyrolysis of cellulose accompanied the decomposition of IL, or vice versa. The ILs, which resulted in the type ii) curves, thus cannot be used as a catalyst in the present study. Pyrolysis of cellulose mixed with sulfonate ILs started at lower temperatures than pure cellulose pyrolysis by the catalysis. The presence of [BMMIM]BF₄ also decreased the temperature for cellulose pyrolysis, but [BMMIM](CF₃SO₂)₂N seemed to be inert.

TGA is thus a powerful approach to easily find a stable and active catalyst for catalytic pyrolysis. The results allowed us to eliminate seven types of ILs (ILs with CF₃CO₂⁻, Cl⁻, HSO₄⁻ and PF₆⁻ anions) out of 13 as the candidate of catalysts. Figure 4 summarizes the characteristics of weight loss curves of cellulose/IL mixture when the IL is catalytically active and thermally stable. The start temperature, T_start, and the onset temperature, T_onset, obtained by TGA for selected samples are listed in Table 1 for later discussion.

Although TGA classified ILs according to the usability in catalytic pyrolysis of cellulose, pyrolysis experiments were performed also with a part of the unusable ones for reference. Table 2 shows the results.

The use of sulfonate ILs resulted in distinguished yields of levoglucosenone as compared to the yields less than 3 wt% by other ILs. The highest yield, 29.7 wt%, was achieved by [EMIM]CH₃C₆H₄SO₃, but the recovery of IL was not high. T_start,IL and T_onset,IL (Table 1) of the IL explain the low IL recovery. Analysis of the condensable matter by GC-MS (Figure 5e) identified compounds derived from the IL such as toluene from the anion. The derivatives of IL decomposition were confirmed only when ILs, which have T_start,IL lower than 300 °C, were used. Detection of methyl- and butyl-imidazoles (Figure 5b) agrees with the thermal degradation pathway of [BMIM]Cl reported by Hao et al [19]. Chemical interactions between ILs and cellulose (or char) can also cause the decrease in the IL recovery by a partial incorporation of IL into the char matrix. An example would be the incomplete recovery of [BMIM]CF₃SO₃. Because [BMIM] includes an acidic proton, as mentioned above, it has chemically more interactive nature than [BMMIM]. By either or both of these pathways, the ILs are thought to be lost from the bulk liquid phase in the course of pyrolysis.

Main compounds detected in GC-MS analysis of condensable matter from pyrolysis of pure cellulose were furfural, levoglucosenone, DGP, levoglucosan and AGF. The presence of [BMMIM](CF₃SO₂)₂N in the mixture had little influence on the product composition and yield, indicating the inert and stable nature even at the temperature as high as 350 °C. In contrast, the catalysis of sulfonate ILs remarkably changed the composition of products derived from cellulose, leading to the selective production of levoglucosenone. The second highest peak was always DGP for these ILs but far lower than the peak from levoglucosenone.

Irrespective of the cations and the type of sulfonate, it is plausible for all the sulfonate ILs that optimization of the temperature for pyrolysis provides a complete IL recovery simultaneously with a sufficient levoglucosenone yield. However, the weight loss curves in TGA imply the difficulty to find the optimum temperature for some sulfonate ILs. T_start,IL − T_onset,cel in Table 1 represents the value of temperature after onset of cellulose pyrolysis and before start of IL decomposition. The smaller the value is, the more difficult the two reactions are separated. Therefore, for a process with high IL recovery and levoglucosenone yield, it is not recommended to use [BMIM]CH₃SO₃ and [EMIM]CH₃C₆H₄SO₃, which showed T_start,IL − T_onset,cel of 52 and 48 °C, respectively, even if the pyrolysis temperature is modified. The best IL, thus found, was [BMMIM]CF₃SO₃, followed by [BMPL]CF₃SO₃ among the ILs used in this study.

To summarize the influence of IL types on the production of levoglucosenone from cellulose, the anion had a key role in the catalysis. Although a few types of anions among the common ones [20] were not examined in this study (e.g., chloroaluminates, dicyanamide and metal complexes), it is fair to say that sulfonate anion is indispensable for levoglucosenone production.

Meanwhile, the presence of cation in ILs should not be ignored in the discussion of the catalysis. To investigate the effect of IL cations, pyrolysis tests were carried out using cellulose impregnated with inorganic or organic acids including sulfonic acid (HCF₃SO₃). Figure 6 and Table 3 show the results of TGA and pyrolysis tests, respectively. The weight loss curve for cellulose-HCl indicated that HCl had little influence on the pyrolysis of cellulose. On the other hand, impregnations of the other acids significantly changed the curves, and a part of them, H₃PO₄ and H₂SO₄, showed catalysis leading to an increase in the yield of levoglucosenone. In fact, it has been known that the impregnation of H₃PO₄ is effective for the levoglucosenone production [21,22], and the activity of H₂SO₄ (or SO₄²⁻) has recently been reported by Sui et al [23]. The anions incorporated by the acid impregnation into cellulose induce the formation of high yields of char in return for the levoglucosenone formation. The impregnations of sulfonic acids also induced the char formation, but the levoglucosenone yields were comparable with that from pure cellulose in spite of high activities of the anions in the form of ILs (e.g., [BMIM]CH₃SO₃ and [BMIM]CF₃SO₃). These results suggest that, for sulfonate anions, the presence of cations, which can form ILs, is indispensable to be stable and active for the levoglucosenone formation.

The reason for the high levoglucosenone yields with sulfonate ILs has yet to be definitively determined, but it is possible to correlate or uncorrelate the catalytic activities with their properties. The catalytic activity of ILs can be related to T_onset,cel: lower T_onset,cel indicates higher catalytic activity (sulfonate ILs > [BMMIM]BF₄ > [BMMIM](CF₃SO₂)₂N).

The conjugate acid of triflate (CF₃SO₃⁻) is known as a superacid, and those of other sulfonate anions are strong acids. Therefore, the acidity of the conjugate acid of IL anion seems like a factor influencing the catalytic activity. However, the inertness of [BMMIM](CF₃SO₂)₂N against cellulose pyrolysis rejects the idea because H(CF₃SO₂)₂N is more acidic than triflic acid, HCF₃SO₃. Miran et al. reported that pK_a values of conjugate acids of (CF₃SO₂)₂N⁻, CF₃SO₃⁻, CH₃SO₃⁻, CF₃CO₂⁻ anions are −10.0, −7.0, −2.0 and 0.5, respectively. Likewise, the acidity of IL cannot explain the catalytic activity. Although the definition of acidity of IL still remains controversial, there are a number of reports using Hammett acidity function. Hallet et al. [24] summarized Hammett acidities of various types of ILs in their review paper, but none of them were in the superacid range. While neat sulfuric acid has the value of −12, those of sulfonate ILs are even higher than 0, indicating their low acidities. To confirm this, acidic properties of ILs used in this study were investigated as shown in Figure 7. As a reference, 1-methylimidazolium chlorosulfate ([HMIM]SO₃Cl) was synthesized according to the reported method [25] and used for the measurement. The shift of peak intensities at 270–280 and 370–390 nm reveals Lewis and Brønsted acidities, respectively. [HMIM]SO₃Cl exhibited features of both Lewis and Brønsted acidities, and the calculated Hammet acidity (for Brønsted acidity) was higher than 0. In contrast, for ILs of this study, there was no detectable change in UV-visible spectra, suggesting little acidities of them.

The acidities of ILs and conjugate acids of the anions, thus, could not contribute to the explanation of their catalytic activities. However, it must be noted that they are not the values or properties at the reaction conditions such as temperature, where the pyrolysis of cellulose is taking place.

One might suspect that the dissolution of cellulose is related to the levoglucosenone formation. In fact, some ILs, such as [BMIM]Cl, can readily dissolve cellulose even at temperatures lower than 100 °C. The occurrence of the dissolution was visually apparent from the residue of pyrolysis in this study. As shown in Figure 8, the residue was aggregated with a smooth surface when ILs had a solvent power, while in most cases the residue was still dispersed as with the starting material. The former one was observed only when [BMIM]Cl, [BMMIM]Cl and [BMMIM]BF₄ were used. Therefore, the dissolution could not be directly related to the levoglucosenone formation.

However, a property of IL relevant to the cellulose dissolution may be correlated with the catalytic activity. The dissolution mechanism of cellulose in ILs involves the oxygen and hydrogen atoms of cellulose-OH in the formation of electron donor-electron acceptor (EDA) complexes, which interact with the IL [3]. Although the ability of IL for EDA complex formation depends on both cation and anion mainly as hydrogen bond donor and hydrogen bond acceptor, respectively, it has been thought that the anion is the major contributor in the dissolution [1]. The activity of the anion as hydrogen bond acceptor is described by hydrogen bond basicity (β value) in Kamlet-Taft parameters [24]. Now, here are assessed the correlation between the β values and the catalytic activities of ILs. The β value for [BMMIM] ILs reported by Crowhurst et al. [26] decreased in the following order with the type of anion: CF₃SO₃⁻ (0.464) > BF₄⁻ (0.376) > (CF₃SO₂)₂N⁻ (0.243). The same trend was reported in other literatures [13,24]. This order agrees with their catalytic activities for levoglucosenone production of this study. Because the formation of levoglucosenone from cellulose progresses along with dehydration reactions that require protonation of hydroxyl group, the effect of β value, which is associated with the interaction between IL anion and hydrogen atom, seems to be convincing explanation of the catalysis. On the other hand, [BMIM]Cl has much higher β value, 0.83–0.87, in spite of the low catalytic activity for the levoglucosenone formation. It seems that such strong hydrogen bond basicity does not cause the catalysis, but only results in the dissolution of cellulose. In other words, the β value may have an optimum value at around that of CF₃SO₃⁻ (0.464) to give the catalytic activity. Accordingly, it is plausible that, during cellulose pyrolysis, ILs with excess β value are incorporated into cellulose/char matrix by the interaction between them, and the ones with too low β value behave as inert. Although the β values for other sulfonate ILs such as [BMIM]CH₃SO₃ and [EMIM]CH₃C₆H₄SO₃ could not be found in literatures, the values should be lower than that of [BMIM]Cl because their anions are larger than Cl⁻. A similar tendency was seen in the work by Pulati et al. on treatment of coal samples using [BMIM]Cl, [BMIM]CF₃SO₃ and [BMIM]BF₄ [27]. The first one was effective to disperse the coal by the strong interaction at 100 °C, and the last one did not result in any appreciable dispersion. The use of [BMIM]CF₃SO₃ at elevated temperatures significantly modified the nature of the coal to be soluble in pyridine (more than 90% of the coal).

Thus, the β value was well correlated with the activity of IL toward formation of levoglucosenone. Our future work will deal with how the sulfonate ILs work to change the pyrolysis mechanism.