Homogenous esterification of cellulose pulp in [DBNH][OAc]

Cellulose acetate is one of the first synthetic industrial polymers, as the process for cellulose acetate production was patented already in 1894 (Cross et al. 1894). The industrial production of cellulose acetate is based on heterogeneous modification of cellulose using acetic anhydride (Ac₂O) with sulfuric acid catalysis (Klemm et al. 1998), which affords cellulose triacetate (DS 3). Afterwards the product is hydrolyzed to desirable degree of substitution (DS). Extensive literature concerning the acetylation of cellulose has been published and more recently there are many articles concerning acetylation in ionic liquids, where cellulose acetate with different DS values and uniform distribution of substituents can be obtained directly (Barthel and Heinze 2006; Jogunola et al. 2016; Wu et al. 2004; Yang et al. 2014).

Ionic liquids (ILs) have gained considerable attention during recent years due to their unique solvation and physiochemical properties (Gericke et al. 2012). The chemical properties of ILs vary according to the composition. For instance, the ability to dissolve cellulose is often dependent on the basicity of the IL (Fukaya et al. 2006, 2008; King et al. 2012; Parviainen et al. 2013), but there are exceptions, e.g., tetraalkylphosphonium salts (Hauru et al. 2012; Holding et al. 2014), which are basic but require a dipolar aprotic co-solvent to dissolve cellulose. In previous literature, the use of imidazolium based ILs, with alkyl substituents (ethyl, methyl, and allyl) and chloride, bromide or acetate anions, typifies most publications on cellulose dissolving ILs (Heinze et al. 2008; Swatloski et al. 2002). In regard to developing commercial processes, the potential environmental impacts of ILs are also of concern. Several recent articles have also assessed the toxicity of different biomass dissolving ionic liquids. Those with shorter anion and cation alkyl chain length are considered to be of lower toxicity (Ruokonen et al. 2016; Witos et al. 2017). This also corresponds with the better capability to dissolve cellulose as cellulose dissolution capability typically decreases as cation or anion volume gets too large (Isik et al. 2014).

Recyclability of ionic liquids is an important and quite often overlooked aspect. One recycling approach for ILs is based on the use of more labile organic superbase-carboxylic acid conjugates. Above processing temperatures these conjugates dissociate to the neutral starting components, which allows for distillability of the IL. One of these that is of current interest is [DBNH][OAc], which can be recycled and purified by distillation to either pure IL or to its starting compounds (Parviainen et al. 2013, 2015). [DBNH][OAc] is a robust solvent for cellulose and pulp, recycled paper and even cotton, e.g., used cotton clothes. Further, high quality cellulosic textile fibers can be produced from cellulose [DBNH][OAc] solutions by spinning thin filaments into an antisolvent, as in IONCELL-F process (Michud et al. 2016; Sixta et al. 2015). Therefore, [DBNH][OAc] is also of interest as a reaction medium, as the same solvent and process could be used to obtain chemically modified cellulosic fibers or materials. The current Lyocell process, using NMMO, is limited due to the undesired side reactions (Rosenau et al. 2001).

The acetylation of cellulose has been studied in different ionic liquids with several esterification reagents. In the first report on the topic of acetylation in IL the reaction was carried out with Ac₂O using 1-allyl-3-methyl-imidazolium chloride ([Amim]Cl) as solvent (Wu et al. 2004). Acetic acid is a side product, which can be difficult to remove from the IL solution and may also lead to partial hydrolysis (degradation) of the cellulose polymer, if some water is present. Barthel and Heinze (2006) and Hinner et al. (2016) have studied esterification of cellulose with longer fatty acid chlorides in ionic liquids, e.g., [Bmim]Cl and [Emim]Cl. Isopropenyl acetate (IpeAc) has been used in 1-ethyl-3-methylimidazolium acetate ([Emim][OAc]) to form cellulose acetates (Kakuchi et al. 2015). Notably, the side-product of the acetylation reaction with IpeAc is acetone which is easy to evaporate from the synthesis mixture and does not usually introduce side reactions. The same group have just recently published an article, where [Emim][OAc] and other imidazolium based ionic liquids were used in DMSO mixtures for assessment of the mechanistic effects of ILs (Kakuchi et al. 2017). Vinyl esters can also be used in cellulose esterification reactions, as an acylation reagent, but have been used in “traditional” solvents with catalysts or activators (Chen et al. 2016; Çetin et al. 2009; Heinze et al. 2003; Jebrane et al. 2011; Jebrane and Sebe 2007).

Binary systems consisting of IL and an organic co-solvent have also been used to produce cellulose esters. Typical co-solvents include dimethyl sulfoxide (DMSO) (Kakuchi et al. 2017; Xu et al. 2013) or dimethylacetamide (DMA) (Rinaldi 2011), but acetonitrile (ACN) (Nawaz et al. 2014) and acetone (Jogunola et al. 2016) have also been used. However, having an additional co-solvent may actually complicate the workup and recovery of all components. The use of acidic ILs like 1-vinyl-3-(3-sulfopropyl)imidazolium hydrogensulfate ([VSim]HSO₄) has been suggested to increase the efficiency of the acetylation reaction in Ac₂O, but the effect on product molecular weight was not reported (Tian et al. 2014) and anhydrides are already highly activated species.

Besides cellulose, acetylation of other biomass components has been of interest. As an example, efficient acetylation of hemicelluloses with Ac₂O was achieved in [DBNH][OAc], as a slight excess of the base (DBN) can be used to scavenge the released acid thus avoiding possible hydrolytic reactions (Stepan et al. 2013).

In our studies, we have been concentrating on the organic superbase-derived ILs, as these are efficient solvents for cellulose and can be efficiently recycled. One of these is [DBNH][OAc], which is not only an efficient solvent for cellulose, but high-quality textile fibers can be produced upon regeneration of cellulose solutions into water (IONCELL-F process) (Michud et al. 2016; Parviainen et al. 2013, 2015; Sixta et al. 2015). Therefore, it would be highly beneficial to control the modification to cellulose esters before regeneration to fibers (for textile and non-wovens) or films. In the current work, we describe the reaction conditions for homogeneous esterification of pre-hydrolysis kraft (PHK) pulp, manufactured from Eucalyptus urograndis and microcrystalline cellulose (MCC) in [DBNH][OAc]. The recyclability of the IL after reaction, with mass balances and purity of recovered ionic liquid is also demonstrated in preliminary fashion.

Acylation of cellulose samples in [DBNH][OAc] was carried out with acetic anhydride (Ac₂O), isopropenyl acetate (IpeAc), vinyl acetate (VinAc) or vinyl propionate (VinPr). When conducting these reactions, a few things need to be considered: the boiling point of the reagent, side products and their properties.

Acetylation with Ac₂O produces acetic acid as the side product. Thus, in these reactions, a slight excess of DBN was used to scavenge released acid and to minimize the hydrolysis of cellulose. Notably, this only produces new [DBNH][OAc] to the system and thus avoids contamination of the ionic liquid. For a sustainable process this would require conversion of a portion of [DBNH][OAc] back to DBN. This is possible through distillation of the recovered ionic liquid (King et al. 2011). Recent results have indicated that DBN actually has a higher vapor pressure than acetic acid in the mixture, allowing for partial separation of the components (Ahmad et al. 2016). Thus, the motivation for this method is sound.

Acylation with isopropenyl carboxylates produce acetone as by-product (Scheme 1). In contrast to Kakuchi et al. (2015, 2017), in [DBNH][OAc] no additional base or activation was required to obtain excellent reactivity. As the IL is derived from the conjugation of a superbase (DBN) and acetic acid, traces of superbase (from equilibration) likely contribute to catalyzing the reaction. Reaction with vinyl carboxylates produces acetaldehyde as by-product. Both acetone and acetaldehyde are easy to remove from the reaction mixture and are relatively non-toxic, but flammable. In prior literature, the reactions made with VinAc were either catalyzed with a base; K₂CO₃, Na₂CO₃ (Çetin et al. 2009; Jebrane and Sebe 2007; Jebrane et al. 2011), or cellulose was activated with NaOH (Chen et al. 2016).

In general, the progress of acylation reactions of cellulose is easy to follow by IR spectroscopy, from the appearance of the ester carbonyl signal at ~ 1740 cm⁻¹ associated with the decrease of the intensity of the hydroxyl signal band at ~ 3400 cm⁻¹ (Fig. 1). The cellulose C–O–C signal intensity around 1100 cm⁻¹ stays unchanged during the esterification reaction. Therefore, it is a good reference point to assess intensity or volume changes of carbonyl signals. Even though ATR-IR is not an inherently quantitative technique (the signal intensities depend of several issues, including IR light penetration depth into the sample), the normalized intensity of the carbonyl signal was shown to give good correlation with the DS values, compared to the ³¹P NMR analyses (see below).

The relation of integrals from the ATR-IR spectra show a linear correlation with the DS values determined by ³¹P NMR and the correlation between these methods is excellent (Fig. 2). The results show that ATR-IR is a rapid but somewhat crude method to access the DS of acetylated celluloses and ³¹P NMR can be used to obtain external calibration for the ATR-IR. Further, ATR-IR spectroscopy is a sensitive method to track impurities, e.g., identifying possible residual IL in the regenerated product. In the case of [DBNH][OAc], the possible remaining IL could be identified from signals at 1300 (COO), 1370 (CH₃), 1580 (COO⁻) and 1670 (C=O) cm⁻¹.

More accurate DS values for the reaction products were assessed using ³¹P NMR (King et al. 2010). An example of a ³¹P NMR spectrum for the reaction made with 9 eq. of Ac₂O is shown in Fig. 3. The spectrum is referenced to the 4,4,5,5-tetramethyl-1,3,2-dioxaphospholane anhydride (TMDP-anhydride) at 132.2 ppm. The signal at 175 ppm is the unreacted (excess) phosphorylating reagent, which needs to be present in order to verify that the amount of reactant has been sufficient for complete phosphitylation. The DS values were calculated from ³¹P NMR spectra according King et al. (2010) and are listed in Table 1. The reactions were repeated, and also conducted under different conditions to assess the effect of time/temperature, or possible experimental errors. The results of the duplicated reactions were similar to the original ones (< 5% variation in yield and DS). The DS values were determined to be between 0.25 and 3. They corresponded well with the amount of reagents used (Table 1). With Ac₂O and IpeAc the acetylation reaction is almost quantitative. However, the reactions made with vinyl carboxylates (vinyl acetate and vinyl propionate) achieve a maximum DS value of ~ 1.5, e.g., the use of excess reagent, higher temperature or increased reaction time did not increase the maximum obtainable DS. In prior literature, the amount of the reagent required, typically acetic anhydride, has been 5–8 eq’s per AGU to reach almost complete esterification of hydroxyl groups (DS 0.64–2.92) (Heinze et al. 2005; Jogunola et al. 2016; Kosan et al. 2010; Wu et al. 2004). Acyl chlorides have also been used and full reactivity was achieved (DS 3) with 5 eq of reactant, as they are known to be more reactive (Barthel and Heinze 2006). However, the by-product from acid chlorides is hydrochloric acid, which may increase chances for depolymerization. For vinylic reagents, cellulose acylations have not been made in ILs, but in DMF and catalyzed with pyridine or potassium carbonate (K₂CO₃) (Çetin et al. 2009; Jebrane et al. 2011; Jebrane and Sebe 2007) or in DMSO, with sodium hydroxide (NaOH) addition (Chen et al. 2016). Unfortunately, only Chen et al. have reported the DS and DP values for the products. In their study the reaction was fast, but some degradation of the product occurred, since the DP’s decreased. The recovered DS values were high, more than 2.52. However, Jebrane and Sebe (2007) reported mass losses, probably due to the solubility of the product to the regeneration solvents.

In general, in our reaction conditions the maximum DS was obtained under 2 h reaction time, for all of the reagents (Ac₂O, VinAc, VinPr, and IpeAc). Even at moderate temperature (65 °C), the acetylation of cellulose hydroxyls is rapid and complete when the amount of Ac₂O or IpeAc is sufficient. To our understanding, the reason for such rapid reaction must be the [DBNH][OAc] itself. It is a relatively effective IL for dissolution of cellulose, with potential for several equilibrating basic species in solution, which may catalyze the reaction. In prior literature, the yields of the acylated products have often not been reported. However, in the DMSO/TBAF solvent system, the conversion and yield of cellulose acetate is moderate with a prolonged reaction time (Heinze et al. 2003). The reason for such result is TBAF, which deacetylates the cellulose esters regioselectively from the positions of 2 and 3 (Zheng et al. 2013). With imidazolium-based e.g., [Emim][OAc], [Bmim][Cl], [Amim][Cl], ILs the dissolution times are longer, typically 3-24 h (Heinze et al. 2005), compared with our system (0.5–1 h). Reaching high DS often requires high temperature, 80–110 °C, long reaction times 3–24 h (Kakuchi et al. 2015), catalyst (Luan et al. 2013) or co-solvent (Jogunola et al. 2016; Kakuchi et al. 2017; Rinaldi 2011; Schenzel et al. 2014; Xu et al. 2013). Lower temperatures have also been reported but the reaction time has been 24 h and the product produced had a low DS of 0.64 (Wu et al. 2004). The role of the addition of the organic co-solvent is usually to keep the system homogeneous throughout the reaction, as the acylated cellulose (product) is not always soluble in pure IL. Further, organic co-solvents can be used to reduce the viscosity of the reaction media, and possibly to lower the capital and running costs, as less ionic liquid might be required. However, the use of co-solvents might complicate the recycling of the system.

High DS samples of cellulose acetate become soluble in CDCl₃, which enables direct measurement of ¹H NMR spectra in the solution state and thus provides another tool to assess DS values using NMR (see supporting information). The results from ¹H and ³¹P NMR measurements agreed well with each other. Also the regioselectivity of the reaction was assessed with ¹H NMR after perpropinylation (or peracylation) with propionic anhydride and 4-(dimethylamino)pyridine catalyst according to Tezuka and Tsuchiya (1995). In [DBNH][OAc], all the reagents showed typical reactivity with the –OH groups of cellulose (see supporting material), e.g., largest reactivity (preference) for the O-6 position (Liebert et al. 2005; Xu et al. 2011). The reactivity of the positions O-2 and O-3 was almost equal (6 > 2 ~ 3). Notably, this is not always the case, as with larger substituents O-3 position can be hindered (Xu et al. 2011) or the reactivity of the positions of O-2 and O-3 may vary, depending on the solvent and acylation reagent (Abe et al. 2016).

As cellulose is a linear homopolymer, cellulose triacetates obtained from reactions with excess of Ac₂O and IpeAc, show only one ¹H-¹³C HSQC correlation in the acetal region ~ δ 6/100 ppm (Fig. 4). However, in the case of the cellulose acylated with excess of vinyl carboxylates, several signals are detected around ~ 100 ppm in ¹³C. This is due to the fact that vinyl carboxylates react well to produce low DS value products but do not achieve full acetylation (max. DS ~ 1.5). Chemical shifts of ~ 100 ppm in ¹³C NMR are diagnostic for hemiacetal- and acetal structures. With lower DS values a wider range of substitution patterns are possible, as substitution of OH can be in various positions. Therefore, the HSQC spectra has multiple signals due to the different (acetylated) AGU combinations. For instance, a signal at 102 ppm in ¹³C NMR analysis is indicative for structures where the C-2 position is not acetylated (see supporting information).

Previously, it has been shown that the reactivities of woody components and biopolymers are different when performing base catalyzed reactions with Ac₂O and VinAc in DMF (Jebrane et al. 2011). However, in [DBNH][OAc], the reaction was fast with both Ac₂O and VinAc, as after 30 min with 9 Eq (3 eq per hydroxyl group) of reagent at 70 °C, the Ac₂O had already reacted fully, but VinAc reaction had stalled at its maximum. This can be seen from both the ATR-IR and the GPC analyses. From the ATR-IR spectra (Fig. 5) it is apparent that the ester carbonyl signal (1740 cm⁻¹) has reached its largest intensity value after 30 min with Ac₂O and after 10 min with VinAc. However, the signal intensity with VinAc is only half that of the Ac₂O product. With prolonged reaction times (24 or 26 h) the carbonyl signal intensity remains the same with Ac₂O reaction samples but slowly decreases with the VinAc samples. The GPC data (Fig. 6), from the same reactions show that compared to the unmodified cellulose sample, the molecular weights (MW) of the acetylated samples increase with low DS samples but cellulose undergoes slight hydrolytic degradation with prolonged reaction times (Table 1; Fig. 6). It is also evident that more hydrolysis occurs especially when reactant is used in high excess. With Ac₂O, the released acetic acid could possibly cause the degradation of the cellulose backbone, prior being scavenged by the excess of DBN. However, there is an evident drop in DP with VinAc, VinPr and IpeAc with larger amounts of reagents. In the current system, the DP of some of the products decreased to almost 1/3 that of the original pulp (DPw = 2300), when the amount of reagent was 3 eq (Table 1). At the moment we have no proper explanation for this chain degradation. Somewhat speculatively we can argue that this effect is due to high reactivity of the vinylic reagents (local effects/energy).

In order to be able to calculate the yields from the esterification reactions, the real DS values for each sample were evaluated. The analysis of the DS values were performed using ³¹P NMR and the DS correlated average MW per AGU for each sample was calculated (3). In the formula (3) M(AGU) is the molecular weight of one anhydroglucose unit (AGU), DS(³¹P NMR) is the result obtained from the ³¹P NMR, M(Ac) is the molecular weight of the substituent, e.g., acetate and M(H) is the molecular weight of the substituted hydrogen. For instance, the calculated MW for cellulose acetate with a DS value of 0.89 is 199.6 g/mol for one AGU.

After calculating these two results the yield can be determined (Table 1). The DS correlated yields are from moderate to good, being more than 80% with all of the reagents. Even though a portion of the acetylated materials may be soluble in ethanol, water or acetone, the recovered yields were still quite high. Obviously, some material is lost in the washing and drying sequences but results are likely to be representative.

In the current work, we demonstrate the fast and efficient acylation of cellulose with isopropenyl acetate, vinyl acetate and vinyl propionate in [DBNH][OAc], in the absence of additional catalyst. Using these reagents, the byproducts are low boiling, such as acetaldehyde or acetone, allowing for separation from the reaction components by distillation. Acetylation was also achieved using Ac₂O, where DBN was used to scavenge the released acetic acid. Under these conditions, no side products are formed but rather new IL is generated. The reactions with Ac₂O and IpeAc were fast and essentially quantitative. No major differences in reactivity were found between these two reagents. IpeAc has the advantage that it needs no additional base during acetylation. However, it is a more expensive reagent and acetone must be separated after the reaction. The effect of the reaction temperature on the course of the reactions was examined, for Ac₂O and VinAc. The minimum temperature used for the reactions in [DBNH][OAc] was close to the melting point of the pure IL (63 °C). At these temperatures the reactions were almost instantaneous.

We found that with Ac₂O or IpeAc in [DBNH][OAc], it is possible to tune the DS values the products between 0.25 and 2.97. On the other hand, the vinylic reagents react well to produce low DS values but the maximum obtainable DS value is ~ 1.5. Comparison of the DS determination using ³¹P NMR and ATR-IR showed a strong linear correlation between these independent methods. Thus, a simple correlation can be used to determine DS, which does not require NMR.

In the current work, we have also demonstrated that the IL can be recovered and recycled by distillation. n-Butanol can be used to entrain water before distillation, thus, avoiding significant hydrolysis. The mass balances in these initial experiments are very promising. However, also other recycling systems for IL’s are to be considered to recognize the most cost-effective recycling pathways.