A cautionary note on “exothermic events” upon contact of carbodiimide coupling agents and the cellulose solvent N-methylmorpholine-N-oxide

The heterocyclic amine oxide NMMO (N-methylmorpholine N-oxide) is very well known as a cellulose solvent (Gränacher and Sallmann 1939). Used in the form of a monohydrate, NMMO is the basis of the lyocell process, an increasingly prominent alternative to the viscose process for manufacture of man-made cellulosic fibers. NMMO dissolves cellulose in quantities of up to approx. 15 wt% and forms a viscous solution at temperatures above 80 °C, which can be spun well (Fink et al. 2001; Sayyed et al. 2019). NMMO is also used as a reaction medium for cellulose modification and has been investigated as a treatment for fibers (Zhang et al. 2018; Öztürk et al. 2009; Manian et al. 2007) or swelling medium for biomass processing and utilization (Satari et al. 2019; Brodeur et al. 2016; Cai et al. 2016; Wikandari et al. 2016; Aslanzadeh et al. 2014).

In the last 6 months we received three inquiries about the chemistry of NMMO. In each case there had been an accident in chemical laboratories, in one case with two injured, in the other two cases only with material damages. We were contacted because in previous work we had dealt intensively with the chemistry of the solvent NMMO, possible side reactions and options for stabilizing NMMO systems (Rosenau et al. 2001, 2002). This ultimately led to the fact that industrial lyocell processes have become established and can now be conducted safely and chemical-efficiently on a large industrial scale. However, it should not be forgotten that NMMO and NMMO monohydrate have inherent dangers that should not be neglected, but unfortunately are often forgotten. NMMO and its hydrate are solids that must be either melted (T > 80 °C) or used with cosolvents in order to dissolve cellulose. NMMO is also a strong oxidizing agent, which as such is used in synthesis chemistry (Cope et al. 1949; Rosenau et al. 2004; Pieterse 2013), but can of course also lead to unwanted oxidation (Adorjan et al. 2005; Rosenau et al. 2005a). In addition, NMMO tends to spontaneous decomposition reactions which can have catastrophic effects if larger quantities are involved as in industrial scale. But even on a small laboratory scale such “exothermic events” or “thermal runaway reactions” of NMMO, as they are euphemistically called, can be quite “spectacular” (see Fig. 1).

In general, the decomposition processes and side reactions of NMMO can be divided into homolytic (radical) and heterolytic (ionic) decomposition processes (Rosenau et al. 2001). The homolytic reactions are mainly induced by the presence of redox-active transition metal ions, the ionic ones by acylation or alkylation reagents. In each case, an interaction with the exocyclic oxygen atom takes place, which weakens the high-energy N–O bond and—upon its cleavage—releases considerable amounts of energy. Particularly critical here is the autocatalytic decomposition of NMMO by carbenium–iminium ions (Rosenau et al. 1999; Potthast et al. 2000). These are formed as a central intermediate product of decomposition, which itself accelerates further decomposition. The mechanism of this process is shown in Scheme 1. Once started, such a thermal decomposition reaction cannot be stopped. NMMO decomposition on a larger scale can lead to complete destruction of the plant, explosions, flash fires and fires. From an industrial perspective, however, less dramatic aspects of NMMO decomposition reactions are also important because they involve loss of solvent and the need to use stabilizers to prevent these decomposition reactions (Rosenau et al. 2002, 2005b). The problem of NMMO stability or NMMO loss concerns both technological aspects and, evidently, economic aspects.

But even on a smaller laboratory scale, the spontaneous decomposition of NMMO must be taken seriously—accidents have happened and continue to happen, as the above-mentioned incidents show. In all these cases, attempts have been made here to carry out esterifications or amidations of polysaccharidic substrates—cellulose as polyalcohol or chitin as polyamine—with low-molecular acids, using the coupling reagents dicyclohexylcarbodiimide (DCC) or 1-ethyl-3-(3-aminopropyl)carbodiimide (EDC) in NMMO solution or NMMO-containing solvent systems. Here—known from previous work—a pronounced instability of NMMO was to be expected. We have taken these accidents as an occasion to investigate the interaction of the cellulose solvent NMMO with common coupling reagents in a small-scale study in more detail and thus to point out once again the possible dangers of careless handling of NMMO.

Irrespective of whether NMMO is present as a solid substance or in organic solution, contact with the coupling reagents DCC or EDC leads to almost immediate, spontaneous decomposition of the amine oxide. If NMMO monohydrate and 1 wt% DCC are mixed with a spatula, an explosion-like reaction takes place within a few seconds with complete decomposition, strong smoke development and sometimes inflammation (Fig. 1). The same effect is achieved by adding a drop of DCC solution in chloroform to solid NMMO monohydrate (see Table 2). The behavior of EDC is completely analogous. It should be noted that stoichiometric amounts of DCC/EDC are not required. As soon as minute quantities of carbenium–iminium ions (Scheme 1) are formed, which under the concentrated conditions find no other reaction partner than further NMMO, the autocatalytic degradation cycle is initiated and can no longer be stopped. This leads to complete—and strongly exothermic—decomposition of the NMMO.

In solutions of NMMO the reaction proceeds analogously, but the effect is less drastic than in solid NMMO due to dilution and better heat dissipation through the solvent. In the case of a 10 wt% (0.85 M) solution of NMMO in acetone at room temperature, the addition of a 5 wt% EDC solution leads to such strong heating that the solution shoots out of the flask in a jet. When mixing 0.1 M solutions of NMMO in chloroform with a few drops of 1% DCC solution in the same solvent, there is still spontaneous heating of the solution which brings the solution to a boil and causes a brown coloration. These effects are of course mainly dependent on the concentration of NMMO, but also on the type (heat capacity) of the solvent and the effectiveness of possible cooling.

The mechanism of the decomposition reaction is shown in Scheme 2. Similar to acylating agents or strong alkylating agents, DCC/EDC react as strong electrophilic compounds with the exocyclic oxygen atom of NMMO, which leads to N-O bond breakage and ultimately to the formation of a carbenium–iminium ion. This then initiates the autocatalytic NMMO decomposition according to the known mechanism (Scheme 1). In general, there is no difference whether NMMO is used in anhydrous form (as often used in organic synthesis) or as NMMO monohydrate, such as the common cellulose solvent. DCC/EDC react much faster with the much more nucleophilic NMMO oxygen than with (hydrate) water. EDC was even developed for use in aqueous solvent systems and therefore shows only low reactivity towards water.

Also after activation of acids by DCC/EDC the decomposing effect on NMMO/NMMO monohydrate persists. The intermediate O-acyl-isoureas formed in this process have been “prepared” for further reaction with alcohols (to form esters) or amines (to form amides) by conversion with the coupling reagent, so to speak, and accordingly represent “activated carboxylic acids”. They are strong acylating agents and react not only with alcohols to form esters and with amines to form amides, but also with NMMO under attack and acylation at the exocyclic oxygen, which finally leads to complete decomposition of the amine oxide, completely analogous to the reaction of the coupling reagents DCC/EDC themselves, i.e. in the absence of carboxylic acids to be activated (Scheme 2).

The decomposition reaction is of course also here dependent on concentration and dilution effects (Table 1). Again, the concentration of NMMO is decisive, whereas the concentration of coupling reagent or activated acid has only a minor effect. Even low concentrations of DCC/EDC or DCC/EDC-carboxylic acid mixtures, down to 1 mM, start the complete decomposition of organic NMMO solutions, but the further course of NMMO decomposition is determined by the autocatalytic NMMO decomposition, for which the concentration of the initiator is irrelevant. Also in water the effect can be observed down to a NMMO concentration of about 0.5 M (approx. 5 wt%). While in organic solvents the limit of instability is 1 wt% NMMO, it is about five to ten times higher in water.

The summary of the results in Table 2 clearly shows that solutions that may contain traces or even higher concentrations of DCC/EDC coupling reagents or of DCC/EDC-activated carboxylic acids should never be brought into contact with NMMO solutions, let alone with NMMO melts, cellulose solutions in NMMO monohydrate or lyocell spinning dopes in general. There is a high probability that spontaneous decomposition reactions will occur as a result, and there is a realistic health hazard and danger in terms of occupational health and safety!

The influence of cellulose dissolved in NMMO on the reaction is minimal. NMMO is only insignificantly stabilized by dissolved cellulose, if at all (Table 3). For the experiments in Table 3, a 5% solution of cellulosic pulps in NMMO monohydrate was used (see experimental section).

It is equally hazardous if NMMO is brought into contact with cellulose that still contains traces of coupling reagents or DCC/EDC-activated acids as remnants of modification reactions. This may be the case, for example, if the modified cellulosic material is to be swollen in NMMO systems or, after dissolution under lyocell conditions, formed into films, fibers or beads, etc. In the case of pure NMMO melts this also leads to complete decomposition within a few seconds, similar to that in Fig. 1. In binary solvent systems with NMMO as one component, the reaction is attenuated by the dilution effect and—depending on the NMMO concentration—the reaction mixture is again brought to boiling or at least heated up. But even here there is a potential danger (burns, splashing etc.).

Table 3 demonstrates the effect that different pulps containing 1 wt% of coupling reagent or DCC/EDC-activated carboxylic acids have on NMMO and NMMO mixtures as solvents/swelling agents. The concentration of pulp was constant at 5 wt% of the solvent mixture and the concentration of coupling reagent was 1 wt% relative to the pulp, i.e. 0.05 wt% relative to the solvent. Here, in analogy to earlier NMMO studies, work was carried out in a calorimetric bomb and the time from the first contact of the reactants to the start of the decomposition reaction, which is noticeable by a sudden significant rise in temperature, was measured (“t_onset”). The shorter t_onset, the more unstable the mixture or the more reactive the DCC/EDC-containing pulp. Table 3 clearly shows that there are no significant differences between different pulps when it comes to contact with pure NMMO (monohydrate). In binary solvent systems with NMMO, the reactivity generally decreases with increasing content of the second solvent, so that reactivity nuances become more visible: while the type of pulp itself continued to have no influence, the pulp with higher hemicellulose content (xylan) appeared at least slightly more stable, so that immediate decomposition no longer occurred.

The protective effect of xylan can be explained by the fact that xylan—compared to cellulose—is oxidized much faster by NMMO at the reducing ends than cellulose (in addition, xylan contains significantly higher concentrations of reducing ends due to its lower molecular weight). These are oxidized to xylonic acid moieties with NMMO being reduced to NMM. NMM, in turn, is strongly basic and has thus a stabilizing effect on NMMO. Although this hypothesis on the mechanism of action has not been clearly proven, the fact that isolated xylan with 1 wt% DCC/EDC residual content in NMMO is significantly more stable than the same mass of pulp (Table 3) is an important indication of the validity of this theory.

While the addition of alcohols or water does not provide sufficient stabilization, secondary amines completely prevent the decomposition effect of DCC/EDC residues on NMMO if their concentration is in the order of (or above) the DCC/EDC concentration (see Table 3). This effect is easy to understand for two reasons: on the one hand, the secondary amines react quickly with the activated carboxylic acids, which are thus no longer able to initiate the decomposition of NMMO, thus forming amides; on the other hand, by trapping the carbenium–iminium ions, they interrupt the critical decomposition cycle of NMMO, forming stable formaminals (Rosenau et al. 2002). It is therefore recommended—as a conclusion—to add 1–2 mol% secondary amine (morpholine, piperidine) to NMMO or NMMO mixtures when processing cellulosic materials which may contain DCC/EDC residues, in order to stabilize the solution against NMMO degradation. This is much more effective than an aqueous alkaline solution of the same concentration (Table 3). The addition of secondary amines is in many cases compatible with the intended reaction course. They can also be easily removed from the cellulose product by a slightly acidic wash.

In summary, DCC/EDC and similar coupling reagents are incompatible with NMMO. Contact can lead to dangerous, highly exothermic, uncontrollable decomposition reactions. Particularly treacherous are traces of DCC/EDC or DCC/EDC-activated carboxylic acids, which remained in cellulosic material that is brought into contact with NMMO in subsequent reaction steps. In general, NMMO solvent systems (preferably with lowest possible NMMO contents) are safer than pure NMMO as melt (lyocell conditions). Small amounts of secondary amine (in the order of magnitude of the concentration of the DCC/EDC traces) effectively stabilize the systems against the otherwise threatening decomposition reactions and exothermic reactions, so that such an addition is highly recommended whenever possible.

The present small study provides additional insights into the chemical mechanisms of the frequently used cellulose solvent and swelling agent NMMO, which has a potential for danger, but can be used completely harmlessly if handled properly, as the decades of use in large-scale cellulose fiber production proves. Above all, however, we hope that the warnings contained in this document will help to raise awareness of the inherent dangers of working with NMMO at elevated temperatures—especially if inexperienced in handling NMMO—so that in future all related accidents in the laboratories can be prevented.