On the role of N-methylmorpholine-N-oxide (NMMO) in the generation of elemental transition metal precipitates in cellulosic materials

The use of N-methylmorpholine-N-oxide monohydrate (NMMO, 1) in the manufacture of cellulosic products according to the lyocell technology, mostly fibers, is well established on an industrial scale. While more dilute aqueous solutions of NMMO have a swelling effect on cellulosic pulp, the respective monohydrate is able to dissolve cellulose at elevated temperatures above 80 °C so that a spinnable dope is obtained. In theory, this dissolution process is merely physical, but in practice quite many side reactions might occur (Rosenau et al. 2001), which, however, can efficiently be suppressed by the addition of appropriate stabilizers. Propyl gallate (PG, gallic acid propyl ester, 2) is the most common one of them (Rosenau et al. 2002a,b), which is even applied on an industrial scale, but also a non-acidic process medium contributes to minimizing degradation reactions of the solvent. Potential hazards of improper handling of NMMO in its different application fields (cellulose dissolution and fiber/film manufacture, biomass pretreatment, organic synthesis) have recently be summarized and pointed out (Rosenau and French 2021).

The lyocell system has very frequently been used also on a laboratory scale for research on novel cellulosic products, especially the embedding of particles and nanoparticles into cellulosic matrices—in most cases fibers, but also films, beads, and pulp (Firgo et al. 1995; Marini and Bruneis 1996; Fink et al. 2001; Sayyed et al. 2019). The incorporation into fibers can be achieved either by (nano)particle-containing dopes, which are then spun out, or by swelling cellulosic substrates in aqueous NMMO and precipitating the nanoparticles into/onto this matrix (Shah et al. 2013). The by far most dominant topic in this regard is the generation of transition metal precipitates, in particular from silver, because of its antimicrobial/antibacterial properties (Smiechowicz et al. 2011, 2018, 2020,2019; Rac-Rumijowska et al. 2017). But also gold nanoparticles have been studied (Yokota et al. 2008; Kitaoka et al. 2011; Hatakeyama et al. 2019), as well as platinum and lanthanide equivalents (Erdman et al. 2016; Skwierczyńska et al. 2019). Luminescent cellulosic fibers have been obtained from an aqueous NMMO medium by incorporating europium-doped (Kulpinski et al. 2012) or terbium-doped (Kulpinski et al. 2016) particles into the matrix. In general, NMMO has been characterized as a co-reactant for the transition metal ions and swelling agent for cellulose, which provides a beneficial environment for both nanoparticle generation (instead of larger precipitates) and their incorporation into the cellulose matrix (instead of mere surface adsorption). While this medium has apparently proved to be quite reliable and thus has gained popularity for the generation of transition metal precipitates in cellulosic fibers, there is still a puzzling question about this system from a chemical point of view, which has not yet been pointed out so far.

The formation of transition metal particles from metal salts is evidently not just a physical/mechanical process that causes the transition from the macro- to the micro- or nano-scale—such as the formation of nano-fibrillated cellulose from cellulosic precursors. It is a true chemical reaction, more specifically a reduction of metal cations to elemental metals, i.e., from a positive oxidation state to oxidation state zero. Obviously, an isolated oxidation process does not exist—any oxidation is accompanied by a corresponding reduction, both reactions together constituting a redox pair. However, NMMO is a strong oxidant itself and certainly not capable of functioning as a direct reducing agent for the transition metal cations (Albini 1993; Pieterse 2013). This puts us in the somewhat paradoxical situation that a relatively commonly used reaction system exists that has been successfully applied to the synthesis of elaborate (nano)particle-containing cellulosic matrices, but its chemistry is apparently not at all well understood, and in some cases incorrectly portrayed in the literature. In short, NMMO's reported role as a co-reactant and the nature of the reductant effecting generation of the transition metals from their salts remain unclear.

In this study, we have investigated the chemistry of different systems for the generation of transition metal precipitates on cellulosic substrates in aqueous NMMO in more detail, especially concerning the redox reactions taking place. When elemental metal species are formed by reduction, something else must, in turn, be oxidized. The identification of these redox processes was thus the major task.

The initial expectation and hope that the redox processes in the transition metal/NMMO mixtures were always the same—i.e., independent of the metal/metal salt used—has unfortunately not occurred: apparently, it is a rather dynamic and flexible reaction system. However, this points out all the more how important a detailed characterization of the underlying chemistry can become, not just in mechanistic organic chemistry, but also in material science work. The corresponding studies on the reaction mechanisms are presented below.

Several procedures have been proposed to synthesize silver nanoparticles in aqueous NMMO and incorporate them into cellulosic fibers (Rac-Rumijowska et al. 2017, 2019,2018; Smiechowicz et al. 2011; Shah et al. 2013). These approaches used 50 wt% aqueous NMMO stabilized with 2% propyl gallate. The metal was introduced in the form of aqueous AgNO₃ (between 0.04 and 1 M), which was either added to the pulp or directly into the lyocell spinning dope to get a final Ag concentration of about 250–500 ppm. Blending and dissolution (with concomitant metal precipitate production) were done at 115 °C for 10 min, followed by dry–wet spinning at the same temperature.

It was logical to assume that in return for the reduction of Ag⁺ ions to elemental silver, functionalities of cellulose would be oxidized, especially the hemiacetals of the reducing ends (which do not have their name for nothing). This would lead to carboxyl groups (gluconic acid moiety) at the end of the cellulose chains. Likewise, oxidation of the cellulosic OH groups along the chains was conceivable, which would lead to keto groups at C2 and C3, and aldehyde or carboxyl (glucuronic) functions at C6. Similar reactions between Ag⁺ and cellulose—albeit in the absence of NMMO—had already been used to label the reducing ends of celluloses to prepare silver-containing cellulosics (Kato 1959; Gagnaire and Vincendon 1973; Kuga and Brown 1988).

The content of oxidized groups in cellulosics and their relation to the molecular weight distribution can be reliably and accurately determined by the CCOA method for carbonyls (Röhrling et al. 2002a, b; Potthast et al. 2003) and the FDAM method for carboxyls (Bohrn et al. 2006), two frequently applied approaches when accurate monitoring of oxidative processes of cellulosics becomes necessary (Potthast et al. 2005). Both rely on group-selective quantitative fluorescence labeling in combination with gel permeation chromatography (Potthast et al. 2015).

While these methods showed a clear oxidative effect of pure NMMO monohydrate on cellulose (120–130 °C, 2 h) accompanied by chain degradation (cf. also Rosenau et al. 2005a, b), no such effects were seen for the silver-containing system, which contained 50 wt% aqueous NMMO stabilized with 2% propyl gallate, 5% cellulosic pulp and 500 ppm Ag⁺ (rel. to the pulp, as 1 mM AgNO₃). Mixing of the reagents occurred at r.t. and the reaction at 50 °C (see “Materials and Methods” section). Under these conditions, the C=O and COOH content of the cellulose as well as the molecular weight stayed largely unchanged, see Fig. 1. This evidently excluded cellulose as the main reaction partner (= reductant) of the silver ions. Also, the NMMO content remained nearly unchanged, with NMMO consumption by degradation to N-methylmorpholine being below 0.8%. In the absence of silver ions, under otherwise identical conditions, the results were the same as in their presence: almost no introduction of oxidized groups and rather little chain degradation was observed. Oxidation of cellulose was minor and became more pronounced only at temperatures > 120 °C and reaction times > 1 h.

The actual co-reactant for the silver ions was found more or less by serendipity when we observed a blue-black discoloration at spots of scratching with a metal spatula, which brought propyl gallate (2) back to mind. This compound could not only be a stabilizer that stays in the background and is largely ignored, but it could be as well an active reagent and co-reactant. The observed discoloration was due to gallic acid-iron complexes, which are extremely color-intense. Such compounds have been the basis of inks for centuries, if not millennia, and countless—sometimes rather adventurous—recipes exist for their preparation (Melo et al. 2019; Díaz Hidalgo et al. 2018; Zervos and Alexopoulou 2015). However, all these historic protocols are based on gallic acid (from gall apples) and the addition of iron ions.

When we extracted the reaction mixture with toluene, we found a single reaction product of propyl gallate, namely ellagic acid (4), and a direct linear correlation of its formation with the consumption of Ag⁺, which indicated a direct conversion and the absence of other parallel or side reactions. Propyl gallate and Ag⁺ were strictly stoichiometrically converted into ellagic acid and elemental Ag as the final products (Fig. 2, left). Propyl gallate (2% rel. to aqueous NMMO) was always present in large excess compared to the silver ions (500 ppm rel. to the 5 wt% dissolved cellulose) so that its concentration could be assumed to remain constant during the reaction. Reaction kinetics, which were recorded at r.t. to slow down the reaction, thus followed a pseudo-first-order with regard to the silver ion concentration and ellagic acid formation. In a reverse experiment—in which an aqueous solution without NMMO being present was used—propyl gallate (2) was added into a tenfold excess of silver ions, giving also a pseudo-first-order kinetics for propyl gallate consumption and ellagic acid (4) formation. This points to the fact that the rate-determining step is a bimolecular reaction, with its reaction rate r = d[4]/dt = k[Ag⁺][2] being first-order with regard to both silver ion concentration and propyl gallate concentration, resulting in overall pseudo-first-order when either Ag⁺ or PG are present in large excess, i.e., r = d[4]/dt = k´[Ag⁺] = k´´[2], as shown in Fig. 2, right, with k, k´ and k´´ being the respective kinetic rate constants.

In the first step of the overall redox process, silver ions are reduced by propyl gallate to elemental silver and the corresponding phenoxyl radicals (2a), which immediately recombine to the diphenic acid ester intermediate (3). Mainly entropically driven, this intermediate is immediately converted to ellagic acid and two equivalents of 1-propanol (Scheme 1). The redox couple active in this system can thus be summarized with Eq. 1:

with Prop-OGall = propyl gallate, Prop-OH = 1-propanol, and EllA being the common abbreviation of ellagic acid.

The formation of ellagic acid (4) is a typical reaction of gallic acid and its derivatives (Aguilar-Zarate et al. 2018; Jourdes et al. 2012), known for more than a century (Bleuler and Perkin 1916). The analysis of propyl gallate and its reaction products is straightforward as both can be extracted quantitatively with small amounts of toluene (dichloromethane works equally well) followed by determination with GC/MS or HPTLC.

In the absence of propyl gallate, no silver reduction was observed under the applied conditions (50 °C), only to some minor extent after extended reaction times of more than 2 h. At 80 °C, however, the reaction was completed within 30 min also in the absence of propyl gallate. At the same time, oxidation of cellulose and some chain degradation were observed, which confirmed that cellulose acted as the co-reactant of the silver ions, taking over the usual role of propyl gallate, although requiring harsher conditions (longer reaction times or higher temperatures), see Fig. 3. When also the silver ions were left out (in addition to the absent propyl gallate), these oxidation effects almost vanished, proving that cellulose oxidation under the used conditions was due to the silver ions and not due to NMMO.

Platinum nanoparticles have been generated from different platinum salts and precipitated onto cellulosic carriers (Shah et al. 2013), with different temperatures between room temperature and 40 °C having been used, mainly to adjust the precipitation rate and nanoparticle size. In our experiments, we first used 50% aqueous NMMO stabilized with 2% propyl gallate and later the same medium without the PG stabilizer, along with 5% cellulosic pulp and 1000 ppm (rel. to the pulp) of Pt salt.

For all platinum salts tested (with the only exception of hexafluoroplatinate that did not react at all), the propyl gallate (2) present as a stabilizer of the NMMO acted as the reductant, being converted to ellagic acid (4), see Table 1, column 2. The basic redox reaction was thus analogous to the silver case (see also Scheme 1). However, the reduction of the platinum ions proceeded much faster than that of Ag⁺. It occurred already at room temperature, while for Ag⁺ temperatures around 80–100 °C were required. Also the stoichiometry of the reaction was obviously different: while for silver a ratio of 1:1 between Ag⁺ and PG was required (corresponding to Ag: ellagic acid = 2:1, see Fig. 2, left), the stoichiometric ratio Pt:PG was 1:2 for Pt(II) and 1:4 for Pt(IV) salts, corresponding to Pt: ellagic acid = 1:1 for Pt(II) and 1:2 for Pt(IV), respectively. Differences in the reactivity of the salts at room temperature were not observed, and the reaction did not change at all when no cellulose was present.

With the same abbreviations as used above (Prop-OGall for propyl gallate, Prop-OH for n-propanol, and EllA for ellagic acid), the general redox system for Pt(II/IV) and propyl gallate can be summarized by Eqs. 2 and 3, respectively:

In the absence of propyl gallate, the reduction to elemental platinum required significantly higher temperatures, a fact that has been overlooked so far. The precipitation of elemental Pt commenced around 60 °C for the platinum(diammin)-complexes and Pt(IV) halides, and at 90 °C for the tested Pt(II) halides, see Table 1, column 3. Similar to the PG-case, hexafluoroplatinate (IV), which is known to be an exceptionally stable platinum complex (Holleman and Wiberg 1985b), did not react at all, but also hexahydroxidoplatinate(IV), which reacted well with PG, showed no reaction in its absence (except at high temperature). Once more the results were the same independent of whether cellulose was present or not. Obviously, when the easy-to-oxidize PG was lacking, the system had to swerve and use a less “convenient” reductant. For all chloride-containing platinum salts, chlorate (ClO₃⁻) was found as one reaction product, for which the chloride in the starting material was obviously the only source. The presence of chlorate was confirmed by ion chromatography and precipitation (see “Materials and Methods” section); other Cl-based anions apart from chloride itself (chlorite and perchlorate), were absent. Since neither Pt(II) nor Pt(IV) oxidizes chloride to chlorate directly (otherwise the respective salts would not be stable), an involvement of NMMO was indicated, and indeed, the formation of equivalent amounts of N-methylmorpholine (NMM, 5), the deoxygenation product of NMMO, was observed, see Fig. 4 and Scheme 2. Therefore, three parallel processes seemed to be operative in the reaction system, the reduction of the platinum ions to elemental metal, oxidation of chloride, and reduction of NMMO (1) to NMM (5).

Such oxidation of halides by NMMO has been observed for the first time in the reaction of cyanuric chloride with NMMO (Rosenau et al. 2002a,b) and later also for gold complexes in the presence of NMMO monohydrate (Kitaoka et al. 2011). In both cases, chlorate was the final product, which was formed as a disproportionation product of initially generated hypochlorite (HOCl/ClO⁻). Therefore, we assumed that a similar mechanism was also operative in the present case of the Pt salts. Indeed, the addition of the HOCl-trapping agent 3,4-dehydro-α-tocopheryl acetate (6) to the reaction system (platinum salt, 50% aqueous NMMO, 60 °C, no propyl gallate) afforded the trapping product of hypochlorous acid (HOCl), chlorohydrin 3-chloro-4-hydroxy- α-tocopheryl acetate (7), almost quantitatively (Scheme 2). The situation was analogous in the case of bromide salts (see Table 1), with intermediate hypobromite being trapped by trapping agent (6) as 3-bromo-4-hydroxy-α-tocopheryl acetate (8) in 64% yield. The advantage of this trapping reagent and its trapping products is their very high lipophilicity due to the isoprenoid C₁₆H₃₃ side chain, which renders them extremely well extractable with n-hexane, also from very complex multi-component mixtures, and easily analyzable.

With the intermediacy of HOCl confirmed, the corresponding redox system can be characterized in a way that in the overall process Pt(II/IV) is reduced to Pt(0) while chloride is oxidized to chlorate. In detailed steps, Pt(II/IV) is reduced to elemental platinum and Cl⁻ is oxidized, in turn, to ClO⁻ (Scheme 2, red). At the same time, NMMO is reduced to NMM (Scheme 2, blue). This is followed immediately by the disproportionation of the formed hypochlorite into chloride and chlorate, which is a kinetically and thermodynamically highly favored process (Holleman and Wiberg 1985a). Note that the two reduction processes—that of Pt(II/IV) and that of NMMO—are not independent: neither Pt(II/IV) alone nor NMMO alone is reduced by chloride, but in combination, they are obviously sufficiently reactive. Coulombic interaction of the negative exo-oxygen in NMMO with transition metals is known to increase the length of the highly polar N–O bond and to render that bond less stable (Linton 1940). Thus, it represents the first step of transition metal ion-induced N–O bond cleavage (Rosenau et al. 2001). It is reasonable to assume that a similar NMMO-metal interaction is also present in the Pt-NMMO-Cl system. In any case, NMMO would be required to enter the coordination sphere of the Pt cations for the reaction to occur.

The overall redox reaction of Pt(II) salts can thus be described by Eqs. 4–6 and those of Pt(IV) salts by Eqs. 7–9, cf. Scheme 2. Equations 4 and 7 describe the actual redox process under the generation of hypochlorite, Eqs. 5 and 8 represent the immediate disproportionation of hypochlorite, and Eqs. 6 and 9 give the resulting net process (the starting salts are written in their non-dissociated form for clarity). With regard to all Pt reactions with chloride, the situation is fully analogous for the bromide case.

Interestingly, there was a distinct reactivity difference between Pt salts that have chloro ligands, i.e. chloride in the coordination space sphere of Pt, and those salts in which chloride is the counter anion. The former reacted much more readily (see Table 1): at 60 °C the reaction was complete within less than 30 min. In the case of [PtCl₆]²⁻ as the most reactive complex salt tested (Table 1) the reaction was even complete after 5 min. In the case of PtCl₂ and PtBr₂—with halides being present as anions and Pt in solution in the form of the aqua-complexes [Pt(H₂O)₄]²⁺ and [Pt(H₂O)₆]²⁺—the reaction at 90 °C was still incomplete after 1 h.

It is known that platinum complexes can undergo ligand isomerism (Kukushkin 1991). Either ligands change places within the coordination sphere or there is an exchange between ligands and counterions. Such isomerism is too unfavorable at room temperature but becomes more prominent at higher temperatures. In water both aqua- and hydroxido-complexes can occur depending on pH and concentrations. The fact that the exceptionally stable [PtF₆]²⁻, which does not undergo ligand exchange at all, is not reduced in the NMMO system—even if excess chloride is added to the solution—might be an indication that only Cl⁻ in the coordination sphere of the central metal ion, i.e., as a chloro ligand, can undergo a reaction according to Eqs. 6 or 9, but not chloride outside the ligand shell. If this is correct, the slower reactivity of some salts could be readily explained by the ligand-anion isomerism: the freely mobile chloride anions must first be immobilized and coordinated to the central cation before they become amenable to the oxidation process. This interesting detail of platinum salt reactivity is currently investigated in more detail, especially concerning the question of whether cis- and trans-complexes (Holleman and Wiberg 1985b) can be distinguished by reactivity differences towards NMMO or not, which is of interest in view of analytics of cisplatin-type chemotherapeutics. Table 1 seems to indicate that the chloro ligands in cis-Pt complexes reacted faster than in the trans-counterparts.

From the viewpoint of material science, a large number of suitable platinum salts are placed at disposal to generate the corresponding Pt particles, and the different reactivities could allow adjustments of kinetics and properties. Still, the mechanistic question remained how platinum salts would react that do not provide halide ions as co-reactants, neither in the coordination sphere nor as counteranions. We addressed this question by means of Pt(NO₃)₂, [PtF₆]²⁺ and [Pt(OH)₆]²⁺, both as the respective potassium salts (Table 1). While the fluoro-complex showed no reactivity at all, the hydroxido-complex and the platinum nitrate reacted slowly at 115 °C and caused concomitant oxidation of cellulose parallel to Pt(0) generation, see Fig. 5. The reaction of Pt salts with cellulose in 50 wt% aqueous NMMO in the absence of chloride was thus analogous to that of silver salts in the absence of propyl gallate (see Fig. 1).

The generation of elemental gold in aqueous NMMO follows the same reaction pathways as in the cases of Ag and Pt. In our work, the starting gold complex was tetrachloroaurate(III), in the form of chloroauric acid H[AuCl₄], similar to literature accounts (Yokota et al. 2008; Kitaoka et al. 2011). In 50 wt% aqueous NMMO stabilized by 2% propyl gallate (2), Au(III) was neatly reduced to elemental Au(0) at 50 °C while ellagic acid (4) was formed in turn. This is analogous to the reaction as shown in Scheme 1 for Ag⁺ and propyl gallate. In the absence of propyl gallate, the system once again took the “chloride pathway” described above for the Pt case, and chloride was oxidized to intermediate hypochlorite, which immediately underwent disproportionation into chloride and chlorate, cf. Scheme 2. The reaction in 50% aqueous NMMO was completed within 5 min at 80 °C, whereas the literature listed conditions of 80% aqueous NMMO at 100 °C, under which the presence of HOCl as an intermediate had been detected by trapping (Yokota et al. 2008). Therefore, a repetition of this trapping experiment was not considered necessary in our case because of the only slightly differing conditions. The overall reaction can be described by Eqs. 10–12, in analogy to the above Pt case (Scheme 2).

Chloride was present as chloro-ligand in the coordination shell of the central Au³⁺, which—according to the above hypothesis of reactive chloro-ligands and non-reactive chloride anions—would render it amenable to the redox process.

In the case of Au(NO₃)₃ as the Au³⁺ source, i.e. Cl⁻ being absent, no reaction was observed at 80 °C, but it started at 115 °C. This reaction behavior was thus completely analogous to the Pt case. It was somewhat faster, however, showing complete consumption already after approx. 10 min, while the Pt reduction needed about 30 min. With the absence of chloride as the co-reactant, cellulose became the reductant also in the Au case, and this process was once again monitored by following the increase of carbonyl and carboxyl moieties in the polysaccharide (Fig. 5). Also here, the reaction did not stop when the concentration of metal ions was higher than the approximate amount of reducing ends in cellulose, which means that evidently also cellulosic hydroxyl groups were oxidized. These hydroxyl groups are converted to carbonyl structures at C2, C3, or C6, while reducing ends (and some carbonyls at C6) afford carboxylic acid moieties upon oxidation.

From the previous experiments we have learned that different chemical species can act as reductants towards the transition metal ions under lyocell (aqueous NMMO) conditions: either propyl gallate, chloro-ligands or cellulose. But how does the system behave chemically when neither of these three co-reactants is available, i.e. if only the transition metal salt and aqueous NMMO are present? The answer could not be clearly extracted from the literature, since the description of the exact mixture of the composition is sometimes rather vague. After all, the methods described were aimed at the nanoparticles produced and their properties rather than at an accurate chemical and mechanistic description of the redox systems involved.

It is known that the interaction of transition metal ions and NMMO, also as an aqueous solution, can cause exothermicity and explosions due to the autocatalytic decay of NMMO if NMMO is present in excess (Rosenau et al. 1999). Transition metal ions can also be used to degrade equivalent amounts of amine N-oxides under controlled conditions, which can be used for synthetic purposes (Dasgupta and Donaldson 1998; Coleman et al. 2000; Gomes and Antunes 2001; Rosenau et al. 2004). We, therefore, used a concentrated aqueous solution of the transition metal salts (AgNO₃, PtCl₂, and H[AuCl₄]) in diglyme (2.5 wt%) to which NMMO in the same solvent was very slowly added at 120 °C to take advantage of the dilution effect and render the reaction controllable. The reaction was immediately recognizable by a color change due to the precipitated (nano)particulate metals. The metal ions were always present in excess because the small amounts of added NMMO reacted immediately and could not accumulate. Analysis of the mixture showed—independent of the metal involved—only one single NMMO-derived reaction product, namely N-formylmorpholine (10), apart from less than 5% of N-methylmorpholine (5). The ¹H and ¹³C NMR spectra of compound 10 are shown in Fig. 6. Note that the free rotation of the C–N amide bond is restricted at room temperature due to its partial double bond character, which renders the two “sides” of the morpholine ring magnetically inequivalent so that two ¹³C resonances appear for the two N-CH₂ moieties and also two for the two O-methylene groups. Increasing the measurement temperature would diminish the shift difference, and at the coalescence temperature T_c, when the C–N bond rotation becomes fast on the NMR time scale, the two separated peaks would coincide. This is similar to the cellulose solvent N,N-dimethylacetamide, which also shows restricted C–N bond rotation at room temperature, and the determination of the activation barrier of this bond rotation based on the NMR coalescence temperature (Chrapava et al. 2003).

This compound was evidently an oxidation product of the intermediate N-(methylene)morpholinium ions (11) and N-hydroxymethylmorpholine (12) (Scheme 3), which are in equilibrium with each other. They both are carriers of the autocatalytic NMMO degradation (Scheme 3, red). The fact that mixing the reagents in an opposite way—i.e. addition of the transition metal salt solution into excess NMMO—caused an immediate violent reaction under reboiling, accompanied by black discoloration of the mixture and complete NMMO consumption, but without significant metal ion reduction, is an indirect proof of their formation. N-Hydroxymethylmorpholine (12) represents a hemiaminal of formaldehyde, which is a very strong reducing species and is readily further oxidized to the carboxylic stage (formic acid) (Potthast et al. 2000). Under the conditions used, we observed neither further oxidation of the formyl moiety to CO₂, which would leave the remaining morpholine part detectable, nor hydrolysis of N-formylmorpholine into morpholine and formic acid.

Addressing the exemplary cases of elemental Ag, Pt, and Au particles generated from the corresponding salts in aqueous NMMO, we have observed that different chemical species were able to take the role of the co-reacting reductants. The strong oxidant NMMO itself, as introductorily mentioned, is evidently unable to act directly in this way. However, it can still be chemically involved in the conversion, not just being present as the reaction medium or as a swelling agent. Upon comparison of the reaction rates of the co-reacting reductants for the metal ions, a reactivity order was obvious which was the same for all three transition metals tested: propyl gallate >  > chloride > cellulose.

In the presence of the common NMMO stabilizer propyl gallate (PG, 1), the latter is readily oxidized to ellagic acid (4) and the metal ions are reduced (Scheme 1). Neither NMMO nor cellulose are affected in this process, which proceeded already at room temperature or temperatures below 50 °C. If PG is absent, chloride ions will become part of the redox pair, being oxidized to HOCl, while both metal ions and NMMO are reduced (Scheme 2). This process requires temperatures between 60 and 80 °C to be completed in less than 30 min. Also, here, cellulose integrity is not compromised. If also chloride is absent, cellulose itself can act as the co-reactant, and carbonyl functions are introduced while the metal ions are reduced. Compared to the above redox systems, the reaction requires relatively high temperatures of 115 °C and above and is relatively slow with reaction times of 30 min up to several hours for complete metal ion reduction. Apart from the easily affected reducing ends also cellulosic hydroxyl groups are oxidized. Additionally, the basic medium—mainly because of the generated N-methylmorpholine—causes chain degradation according to beta-alkoxy-elimination pathways (Hosoya et al. 2018) starting from the introduced carbonyl groups (Figs. 1, 3 and 5).

If no other co-reactant is present, excess transition metal ions can oxidize intermediate N-(methylene)morpholinium ions and N-hydroxymethylmorpholine, the product of a metal ion-induced intramolecular rearrangement of NMMO, to N-formylmorpholine (10), with the metal ions being reduced to elemental metal, in turn (Scheme 3). This reaction, proceeding at 120 °C and above, does not proceed in aqueous NMMO solutions, but requires NMMO or its monohydrate, which can, however, be diluted with inert organic solvents. Caution must be exercised that the NMMO is slowly and carefully added to an excess of metal ions since otherwise the N-(methylene)morpholinium species will react with NMMO instead of the metal ions, and trigger the non-controllable autocatalytic degradation of the amine N-oxide.

With the present study on the chemical reaction mechanism of transition metal (nano)particle generation on cellulosics in aqueous NMMO, we hope to contribute to a better understanding of the chemical aspects of this interesting system and these intriguing materials and to correct certain inconsistencies in the description of the role of NMMO and the underlying chemistry.