Cyclic peroxides as key intermediates in the degradation of cellulosic key chromophores by alkaline hydrogen peroxide: first direct proof by 17O NMR

2,5-Dihydroxy-[1,4]-benzoquinone (DHBQ) and 5,8-dihydroxy-[1,4]-naphthoquinone (DHNQ) are two key chromophores which are almost ubiquitous in cellulosic materials. Their fate under conditions of alkaline peroxide bleaching (P stage) has been established previously, but the intermediacy of cyclic peroxides, which so far had only been postulated, remained an open issue. By means of 17O NMR spectroscopy, additionally supported by other NMR techniques, it was demonstrated that both DHBQ and DHNQ form cyclic peroxides as primary intermediates in the reaction with hydrogen peroxide under alkaline conditions. These intermediates are subsequently further degraded to products already known. The experimental confirmation of the cyclic peroxides is an important step in the understanding of reaction mechanisms in pulp bleaching chemistry.


Introduction
Three compounds have been established as key chromophores in cellulosic matrices: 2,5-dihydroxy-1,4-benzoquinone (1), 5,8-dihydroxy-naphthoquinone (2), and 2,5-dihydroxyacetophenone (Korntner et al. 2015). Because of their exceptionally strong resonance stabilization they are the main survivors of bleaching attempts, and due to their exceptional thermodynamic stability they are readily re-formed from low-molecular weight degradation products upon aging and brightness reversion. It is their ubiquity in cellulosic materials that makes them prime targets of mechanistic studies, which aim both at a better understanding of their bleaching chemistry and at high efficiency of their removal or destruction-in terms of both reactivity and costs.
The degradation of the key chromophores under conditions of industrial peroxide bleaching (''P stage''), i.e. by treatment with alkaline hydrogen peroxide, has been studied in detail. The kinetics as well as the reaction mechanism, stable intermediates and final degradation products, mostly C 1 to C 4 acids and hydroxyacids and carbon dioxide, have been established (Hosoya and Rosenau 2013a;Zwirchmayr et al. 2017). One question that remained open, however, was the intermediacy of cyclic peroxides 1b and 2b (Schemes 1, 2). The occurrence of these intermediates has been postulated from in-depth computational treatments of the reaction pathways (Hosoya and Rosenau 2013a;Zwirchmayr et al. 2017), and NMR experiments provided some faint indication of their existence, but solid proof was still missing. In particular 1 H and 13 C NMR spectroscopy was inconclusive insofar as resonances of putative intermediates were in agreement with the cyclic peroxides' structure, but would fit also to other structurally similar compounds.
The task how to prove the existence of such transient cyclic peroxide intermediates along the degradation pathway from the chromophores to the final low-molecular weight products was not trivial. Complicating was evidently the fact that the intermediates-as per definition-were not stable, so that they could not be isolated or determined by chromatographic techniques. In addition, they were present in a rather complex mixture, and their high symmetry made conventional 1 H and 13 C NMR techniques less meaningful due to low numbers of characteristic resonances (note, for instance, that the dianion of DHBQ (1a) produces a singlet in 1 H NMR and two singlets in 13 C NMR).
We came up with 17 O NMR (Berger et al. 1997;Boykin 1990;Klemperer 1978;Chandrasekaran et al. 1984;Gerothanassis 2010a, b) as a means to tackle this problem. This variant of non-metal NMR is still relatively little used, which is mainly due to the fact that 17 O is rather rare (0.037%) and 17 O-labelled compounds are consequently rather uncommon and very expensive. In addition, 17 O is a spin 5/2 nucleus with moderate quadrupole moment, which produces notoriously broad resonances, and low receptivity of 0.06 relative to 13 C. This, however, is to a certain extent counterbalanced by the fact that the chemical shift range for 17 O is quite extensive and covers more than 800 ppm for organic compounds (1800 ppm in total), compared to about 15 ppm for 1 H and approx. 230 ppm for 13 C. 17 O NMR of non-enriched compounds (natural isotopic abundance) is possible for neat liquids or highly concentrated solutions (Fig. 1), but analysis of any component below 10 wt%-or even as low as ''usual'' NMR concentrationsrequires isotopic labelling.

Results and discussion
In concentrated organic solution (DMSO), both DHBQ (1) and DHNQ (2) exhibit two resonances in 17 O NMR, corresponding to the two equivalent quinone oxygens and the two equivalent hydroxyl oxygens. In neutral aqueous solution, DHBQ (1) is already fully dissociated and present as its dianion 1a. Its two hydroxyl groups are rather acidic with pKa values of 2.7 and 5.2, respectively. The pKa values of DHNQ (2), 7.8 and 9.2, indicate that formation of the corresponding dianion 2a requires alkaline media. At a reaction pH of 10, chosen to set the same conditions as in previous experiments (Hosoya and Rosenau 2013a;Zwirchmayr et al. 2017), both 1 and 2 are fully deprotonated and present as their dianions 1a and 2a. This brings about drastic changes in the NMR spectra. While 1 exhibits three 13 C resonances (for the quinoid, the methine and the hydroxyl-bearing carbon), the keto and hydroxylated carbons are rendered magnetically equivalent in dianion 1a, which consequently exhibits only two resonances (methine and C-O). This structural change upon deprotonation is also reflected in 17 O NMR: the two resonances (keto-oxygen and hydroxyl-oxygen) in neutral DHBQ become one resonance in the dianion 1a, in which all four oxygens are magnetically equivalent (Scheme 1). This is another nice support for the resonance structure in 1a with its fully delocalized charges and double bonds, which is the main reason for the recalcitrance of the compound towards common bleaching agents that act on (localized) double bonds.
In neutral aqueous solution, DHNQ (2) produces two singlet peaks in 1 H NMR spectra, coming from the pair of equivalent protons of the ''quinoid half'' and the ''aromatic half'', those two parts being also clearly seen in the 13 C NMR spectra. 17 O NMR gives two resonances, one for the quinoid oxygen atoms and the other for the phenolic hydroxyl oxygens. The changes upon deprotonation to corresponding dianion 2a are, once more, drastic. While neutral 2 has only one mirror plane through the centers of the C2-C3 and C6-C7 bonds, dianion 2a has an additional mirror plane along the annulation bond (C4a-C8a); in other words, the ''quinoid side'' and the ''aromatic side'' become indistinguishable (Becker 1999). 1 H NMR of 2a exhibits one singlet, 13 C NMR three singlets (annulated carbons, methine carbons and oxygenated carbons), and 17 O NMR one resonance. All four oxygens become magnetically equivalent in 2a (Scheme 2). The 17 O NMR data of 1 and 2 in acidic organic solution and 1a and 2a in alkaline aqueous solution were acquired with compounds of 2% 17 O isotopic enrichment, of which the syntheses will be reported elsewhere. For these solutions acquisition times of about 1 h were sufficient, comparable to spectra of neat solvents (e.g. D 2 O, acetone, ethanol) at natural isotopic abundance, i.e. without isotopic enrichment (Fig. 1).
In the case of experiments on possible oxygenated intermediates, such long measurement times are evidently unfeasible because the intermediates have a lifetime of only several minutes at room temperature, as known from previous kinetic studies (Hosoya and Rosenau 2013a;Zwirchmayr et al. 2017). Therefore, isotopic enrichment had to be employed and, in addition, lower temperatures (0°C) to decrease the reaction rate. A 30 wt% aqueous solution of H 2 O 2 with an isotopic labelling degree of 5% 17 O was used. The 17 O-enrichment in the hydrogen peroxide was thus approx. 135 times higher than natural isotopic abundance, which shortened the acquisition times in 17 O NMR to about 10 min. No resonances except those from the labelled H 2 O 2 -oxygens are ''seen'' in these spectra, because signals of oxygens with natural isotopic abundance do not appear. In our experiments, a 1 M alkaline solution (pH 10) of either DHBQ (1) or DHNQ (2) was used, and the isotopically enriched H 2 O 2 was added to a molar ratio between chromophore and H 2 O 2 of 1:3. Spectra were accumulated at 10 min intervals over 1 h and after 2 h. After that time all signals of starting oxidant and primary intermediates had disappeared and the degradation reaction had completed. This was also reflected in the oxygen resonances that had dissipated into many small signals in the carboxylate (deprotonated acids), hydroxyl and carbonate region, corresponding to the final degradation products that had been previously identified and reported (Hosoya and Rosenau 2013a, b;Zwirchmayr et al. 2017).

123
The neat H 2 O 2 starting solution gave the typical 17 O signal at 180 ppm. Rewardingly, the 17 O NMR spectra of the reaction mixture recorded in the initial two 10 min intervals were dominated (apart from the signal of excess H 2 O 2 ) by a singlet around 340 ppm (348 ppm in the case of benzoquinone 1 as the starting compound, and 338 ppm in the case of naphthoquinone 2). This is highly indicative of cyclic, symmetric peroxides with two magnetically equivalent oxygens. The hydrogen peroxides thus introduced an intramolecular O-O bridge into both chromophoric systems by double nucleophilic, entropically driven attack, forming the cyclic peroxides 1b and 2b, respectively, as the primary reaction intermediate along the extensive degradation pathways (Fig. 2). The possibility of the intermediate being a hydroperoxide R-OOH can be eliminated: there would be two resonances for the two inequivalent oxygens, with the OH oxygen typically appearing around 200 ppm, the R-O oxygen between 300 and 380 ppm. A cleavage of the inter-oxygen bond in H 2 O 2 can also be ruled out because possible product alcohols (approx. -50 to 80 ppm), ethers (-30 to 50 ppm) or keto compounds (-530 to 630 ppm) would resonate in completely different shift regions, and no signals were present there.
With the knowledge of the cyclic peroxide structure, it was also possible to assign all carbon resonances of the intermediates 1b and 2b, based on HSQC and HMBC NMR data. In 1b, the magnetic equivalence of the four oxygen-bearing carbons in dianion 1a is evidently cancelled, now producing separate resonances for quinone carbons (two equivalent nuclei) and hemiketal-type carbons (two equivalent nuclei). The protons at the methylene carbons (also two equivalent C) are acidic (methyleneactive) and undergo H-D exchange. The nature of the hemiketal-type carbons in 1b-i.e. the ''former'' quinone carbons to which H 2 O 2 added in a symmetric way upon peroxide formation-had been hard to assign without the input from 17 O NMR, because the nature of the two O-substituents at this carbon could not be concluded just from the 13 C chemical shift, since hydroperoxides OOH, OH or OR, resulting in different forms of masked keto groups (Rosenau et al. 2005), would appear in the same shift region. Intermediate 2b loses the central symmetry that 2a has, and resumes a benzoquinoid structure, i.e. an (aromatic) hydroquinone structure anellated to a quinoid motif that carries the cyclic peroxide. Also in this compound the resonances of the hemiketal-type carbons were assigned. The proton resonances become distinguishable again (because they were in 2) into aromatic and quinoid ones when going from 2a to peroxide 2b (see Scheme 2). The NMR data ( 1 H, 13 C and 17 O) of the two chromophores, their dianions and cyclic peroxides is summarized in Tables 1 (DHBQ) and 2 (DHNQ).
An additional proof of the structure of intermediates 1b and 2b was provided by compounds 2,5dihydroxy-3-methyl-1,4-benzoquinone (3) and 5,8dihydroxy-2-methyl-1,4-naphthoquinone (4), methyl derivatives of DHBQ (1) and DHNQ (2), respectively. If subject to a reaction with alkaline peroxide, cyclic peroxides should be formed in analogy to 1 or 2, but the peroxides would not be symmetric: the methyl substituent breaks symmetry, and the two peroxide oxygens would become magnetically inequivalent. 17 O NMR exactly confirmed this expectation, and proved the occurrence of a cyclic peroxide intermediate 3b with 17 O resonances at 331 and 319 ppm, and a naphtha-counterpart 4b with 17 O resonances at 362 and 329 ppm (Scheme 3 and Fig. 2). Assignment of the two resonances to the two possible oxygens (Scheme 3) was based on computations on the DFT(M06-2X)/6-311?G(d,p) level.
After a reaction time of 30 min, no intermediates were detected that were long-lived enough to allow characterization by 17 O NMR. Only the final degradation products were monitored, with prominent carbonate (190 ppm), acetate (274 ppm) and malonate (262 ppm) resonances (pH 10). Upon acidification with 2 M HCl, the carbonate signal disappeared-by release of gaseous CO 2 -and the other carboxylate

Conclusion
From these 17 O NMR studies, the occurrence of the cyclic peroxides 1b and 2b as the primary intermediates in the oxidation of the cellulosic key chromophores DHBQ (1) and DHNQ (2), respectively, can be considered established, which expands our understanding of the chemistry of these compounds. Since para-quinoid structures are relatively common in the chemistry of chromophores-and also lignins-the  MHz with a 5 mm broadband probe head (BBFO) equipped with z-gradient with standard Bruker pulse programs and temperature unit were used. Data were collected with 32 k data points and apodized with a Gaussian window function (GB = 0.3) prior to Fourier transformation. Acquisition times of 0.5 s to 2.5 s and a 0.1 s relaxation delay were used for 17 O. Bruker TopSpin 3.5 (3.0) was used for the acquisition and processing of the NMR data. 1 H and 13 C spectra were recorded in D 2 O or D 2 O/ NaOD (pH 10) at concentrations of approx. 20 mg/ml at room temperature. 17 O NMR experiments of the chromophores and their dianions were recorded as concentrated solutions (ca. 1 M) at room temperature. Spectra of the cyclic peroxides were taken in concentrated solutions (D 2 O/NaOD) after addition of 30% H 2 O 2 (molar ratio chromophore/H 2 O 2 of 1:3) at 0°C.
Computations were carried out by geometry optimization at the DFT(M06-2X)/6-31?G(d,p) level followed by computation of the NMR shielding with the lager basis functions of 6-311?G(d,p), using the Gaussian program package.