Toward a Consolidated Lignin Biorefinery: Preserving the Lignin Structure through Additive‐Free Protection Strategies

Abstract As part of the continuing efforts in lignin‐first biorefinery concepts, this study concerns a consolidated green processing approach to obtain high yields of hemicelluloses and lignin with a close to native molecular structure, leaving a fiber fraction enriched in crystalline cellulose. This is done by subcritical water extraction of hemicelluloses followed by organosolv lignin extraction. This initial report focuses on a detailed characterization of the lignin component, with the aim of unravelling processing strategies for the preservation of the native linkages while still obtaining good yields and high purity. To this effect, a static cycle process is developed as a physical protection strategy for lignin, and advanced NMR analysis is applied to study structural changes in lignin. Chemical protection mechanisms in the cyclic method are also reported and contrasted with the mechanisms in a reference batch extraction process where the role of homolytic cleavage in subsequent repolymerization reactions is elucidated.


Introduction
Combating climate change has become ag lobal initiative, [1] and as trong case can be made for lignocellulosic biomass to replacef ossil sources as ar aw material. [2] Lignin accounts for approximately 15-30 %o fl ignocellulosic biomassa nd is the most abundant natural source of aromatics. [3] From as ustainability and bioeconomical viewpoint, replacing fossil-based aromaticsw ith sustainable solutions is of high interest where lignin could be used as ap recursor for biofuela nd materials. [4] However,t he fractionation recalcitrance of biomasst oo btain biopolymers is an obstacle that still demands carefule valuation.
One of the conventional pulping processes is the organosolv process, which was patented by Kleinert in 1971. [5] The working principle of the organosolv process is to use an aqueous organic solvent to extract lignin, where low molecular weight aliphatic alcohols are especially used for the extraction. Primary alcohols have shown more selective delignification than secondary and tertiary alcohols. [6] The lower viscosity of organic solvents makes the solvent dispersion in the woodf aster.T he process can also be performed with catalysts as well as additives with different catalysts having been shown to enhance the extraction of lignin. [7] One of the most promising organosolv processes is the Alcell process, which is based on the biorefineryc oncept that pulp, lignin, furfurals, acetic acid, and hemicellulosesc ould all be of value. [8] The Alcell processw as developed for hardwoods by the Canadian pulp and paper industry. [9] However, the principle of using aqueous ethanol with ac atalyst was already investigated in the early 1970s in small-scale pilot plants. [10] The principle of the process is described in various patents: Generally,t he extraction uses ab inary ethanol-water solvents ystem in the range of 20-80 wt %o fa lcohol at temperatures of 160-220 8Cw ith no catalyst added. [11] Ethanolc ould be recovered by means of distillation, owingtoi ts low boilingp oint, making it particularly practical from ar ecovery viewpoint compared to other organic solvents. High quality pulp has been generated from the process. [9,12] With no catalyst added,t he extractions olvent has ap Ho f around4 ,o wing to the generation of acetic acid from deacetylation of hemicelluloses. For more efficient lignin extraction,a small amounto fm ineral acid is commonly added. [13] Am ild organosolv extraction of lignin subsequentt oh ydrothermal pretreatment was recently reported. [14] Evenm ore recently,p olyhydroxy alcohols, such as butanediol [15] and ethylene glycol in combination with dimethyl carbonate, [16] have been used in the reactived issolution of biorefinery lignins.
The structureo fo rganosolv-extracted lignin differs from that obtained by other processes. Under acidic conditions and at high temperatures, carbocations formed in the aliphatic side As part of the continuing efforts in lignin-first biorefinery concepts, this study concerns ac onsolidated green processing approach to obtain high yields of hemicellulosesa nd lignin with ac lose to native moleculars tructure, leaving af iberf raction enriched in crystalline cellulose. This is done by subcritical water extraction of hemicelluloses followed by organosolv lignin extraction. This initial report focuses on ad etailed characterization of the lignin component, with the aim of unravelling processing strategies for the preservationo ft he native linkagesw hile still obtaining good yields and high purity.T o this effect, astatic cycle process is developed as aphysical protection strategy for lignin, and advanced NMR analysis is applied to study structuralc hanges in lignin. Chemical protection mechanisms in the cyclic method are also reported and contrasted with the mechanisms in ar eference batch extraction processw here the role of homolytic cleavage in subsequent repolymerization reactions is elucidated.
chains are prone to react with electron-richa romaticc arbons in an electrophilica ttack. This reactionc ontributes to the formation of stable CÀCb onds in condensation reactions. [12] A commond epolymerization reaction in organosolv processes is acidolysis and the subsequent formation of Hibbert's ketone. [17,18] More specifically,occurrence of undesirable condensation reactions resulting from carbocations at alpha carbonsh as been reported.F or these reactions, protection strategies have been proposed. One protection strategy is to use chemical protection, where ap rotecting group deactivates, reversibly or irreversibly,areactive functional group. [19] One example of this strategyi st he addition of formaldehydei nt he pretreatment of the biomass. Here, upon acetalization, lignin condensation and the formationo fs table CÀCb onds are prevented by blockingr eactive positions prone to lignin condensation. The resultingd erivativec an be reversibly deactivated to achieve deprotection.U nder acidic conditions, the formaldehyde additive can also react with the electron-rich position para to the methoxy group in the aromatic ring of guaiacyl lignin. [20] Another way to increaset he efficiency of the extraction and minimize condensation reactions is to use ap hysicalp rotection strategy.F low-through extraction,am ethodo fc ontinuous extraction, is especially efficient for matrices such as biomass, where components that are extracted in an early step are prone to undergo furtherr eactions in the extract. The principle of the setupi ss imple;a ne xtraction cell is connected to a pump system that continuously provides new solvent into the cell. The extraction time is short and the sample is thereafter cooled down in ac ondenser. [19] The principle follows classical continuouse xtraction concepts where the extraction liquid is less saturated, which is beneficial for efficient extraction and also limits furtherr eactions in the dissolved fractions. It has previously been reported that b-O-4' units are preservedt oa higher degree when using the flow-through principle,m aking the lignin produced by this method more suitable for monomer production in alignin biorefinery. [21][22][23] Consolidated lignin biorefineries, where value can be derived from all streams, are of increasingi nterest. With the goal of furthering fundamentalu nderstanding in this field, we designed and investigated at wo-step sustainable solvent extraction approach for consolidating lignin biorefining with hemicellulose and fiber production.S oftwood, which represents the main technical wood species in Scandinavia, was chosen for this initial study.T he concept wast of urther develop am ild green consolidatede xtraction process, where subcritical water extractiono fh emicelluloses is followed by al ignin extraction based on the solvent systemu sed in the Alcell process, [11] in this case ethanol-water (70:30 v/v). Smalla mountso fs ulfuric acid were added to address lignin purity concerns and extraction temperatures were kept lower (160 8C) than those in the Alcell process to minimize lignin modification.T he concept was developed according to ag roup of pre-set criteria, resulting in as ustainable approach for the extraction of potentially high-value "native"-like lignin manifested by ah ighd egree of b-O-4' interunit linkage, al ower degree of condensation, high purity,and high yield.

Results and Discussion
Given the need for better materialu sage and circularity,c onsolidated biorefinery concepts are attractive. We have explored at wo-step three-component strategy,w ith the initial aim of gaining fundamentalu nderstanding of such processes. Solvents were selected in consideration of the principles of green chemistry and circularity.T hus,w ater and ethanol were chosen. The use of both subcritical water and ethanol in biomass extractionh ave been previously studied. However,a combination of the two in as equential operation to fulfil integratedb iorefinery needs has to our knowledge not been previously investigated. In this study,w eh ave investigated the potentialo ft his system, with the primary focus of developing the fundamental understanding of lignin reactivitya nd its control. As cheme representation of the extraction approach is shown in Figure 1.
Subcritical liquids, that is, liquids where the temperature or pressurei ss lightly under the criticalv alue, have the unique property of simultaneousl ow viscosity and high diffusivity. [24] These properties enhance mass transfer and have advantages in addressing biomass recalcitrancetofractionation.

Hydrothermal extract (hemicellulose) and final fiber residue (fiber B)
In the first step, subcritical water extraction was implemented. Extraction conditions were based on findings from previous work in our research group. [25] These findings also showedt hat the extracted hemicelluloses were partially hydrolyzed but still contained glycosidicbonds and acetyl moieties, which are indicative of am ild extraction. Furthermore, when the extraction temperature was set to 160 8C, the formation of pseudo-lignin from hemicellulose degradation products was negligible. [26] The molecular weight distribution was studied by size-exclusion chromatography (SEC;s ee the Supporting Information, FigureS31). Twop opulations were observed with respect to molar mass. In the higher molarm ass fraction, which accounted for6 1% of the chromatogram area, M n and M w were1 050 and 3050, respectively,a nd the dispersity index ()w as 2.9. The approximate degree of polymerization (DP n )asdetermined by using the anhydromannose unit (162 gmol À1 )a sarepeating unit, was about 6-7. The HSQC spectra of the hemicelluloses extracted in the presents tudy ( Figure 2) display native structures indicative of amildextractioni nthe form of partially O-acetylated C2 and C3 hydroxy groups.
Ta ken together,t he SEC and HSQC data suggest that the native hemicelluloses are partially hydrolyzeda tt he glycosidic bond but that the resulting oligomers have preserved their native structures. Such native hemicelluloses have potential in applicationss uch as emulsions [27] and could also be subsequently fermented to ethanol. [28] In the latter case, the bioethanol produced could potentially be used in the subsequent organosolve xtraction step described herein, thereby contributing to the circularity aspects of the process. The subsequent organosolv extraction is discussed in detail in the next section.
The final fiber residue (fiber B, Figure 1) was also analyzed by X-ray diffraction (XRD) and, as expected, the crystallinity of cellulose was retaineda fter both extraction steps ( Figure S1). This could be of interestf or composite applications. [29] The compositions of the hot water extract and the final fiber were also elucidated by sugar and lignin analyses. Mass balances are discussed in al ater section.

Organosolv (aqueous ethanol) extraction method
In this work, we will mainly focus on the lignin fraction with the emphasis on achieving the preset criteria mentioned earlier,t hat is, high yield and purity combined with ap redominance of native interunit linkages( Figure 3).
Our startingp oint was two reference organosolv extractionswith extraction times of 2h( Figure 4) and 3hat 160 8Cw ith 1.5 wt %s ulfuric acid (for full DEPTedited HSQC spectra, see Figures S18 and S19). These conditions reflect common time scales and temperatures for organosolve xtractions.T he difference here is that most of the hemicelluloses have already been extracted. This may improvet he porosity of the material, resulting in fasterk inetics and improved selectivity for the lignin extraction. The contents of b-aryl ether linkages (b-O-4')and other common native linkages were analyzed by HSQC,a sa ni ndicator of the mildness of the extraction.
The two reference samples showed, as is typical for most organosolv processes, low contentso fb-O-4' bonds (at 7% for the 2h sample and 4% for the 3h sample;T able1). In contrast, the b-O-4' content of native spruce lignin has been reported to be in the order of 35-60 %. [30] Nevertheless, the b-O-4' bonds appeared to be both hydroxylated and etherified at Ca,i na greement with reported results. [31] Such etherification reactions occur through addition reactions of ethanol to electrophilicb enzylic cations and have been postulated to improvet he solubility of lignin in ethanol. This is further discussed later in connection to mechanisms. Under the 2h and 3h extraction conditions, 64 %a nd 100 %o ft he b-O-4' structures were etherified, respectively.
Interestingly,s trong signals were observeda t6 .7/112.5 ppm and 6.6/120.5 ppm, typical of C2Ar-H andC 6Ar-H correlations,  respectively,i n5 -5' condensed subunits. [32] This is furthers ubstantiated by HMBC ( Figure S27). These signals were not as intense in milled wood lignins (MWL)p repared from the original wood and the fibers after subcritical water extraction (fiber residue A, Figure 1; for HSQC spectra, see FiguresS24 andS 25). This suggested that some lignin condensation reactions occurred during the 2h and 3h organosolv extractions.I nc ontrast, signals from noncondensed C5Ar structures were relatively weak in the organosolv lignins. These structures are analyzed indirectly by the drift they cause to C2Ar-H and C6Ar-H chemicals hifts, which then appear at 6.92/110.5 ppm and 6.8/ 118.5 ppm, respectively.T he formation of 5-5' structures cannoto ccur through acid-catalyzed condensation under the prevailing conditions since the free aromatic sites are known to be electron-rich.Amechanistic pathway leadingt ot heir formation is discussed in alater section.

Extraction trend investigation and method development of the cyclic extraction method
Next, we investigated ap hysical protections trategy with the expectation of betterp reserving the native lignin interunitl inkages. Hence cyclic extraction was explored. Here, static cycles consisting of 5min aqueous ethanol extractionsa t1 60 8C, using two acid concentrations (0.5 and 1.5 wt %), were performed on the subcritical water-extracted fiber residue. The obtained lignins were analyzed for yield, hydroxy functionality (using 31 PNMR spectroscopy), lignin structure (using 2D NMR techniques) and molecular weightd istribution (using SEC). Yield analysiss howedt hat the highest quantity of lignin was extracted during the earlier cycles (Figures 5a nd 6) with a steep declinei ne fficiency from cycle 1t o4 .W hen compared to the longer 2h reference extractionsd escribed earlier,t he yields (56.3 %, Table 1) are in the same regime, indicating that the kinetics of lignin extraction are improved when using the cyclic method. This might be explained by saturation being impeded by the periodicale xchange of solution for fresh solvent, which enhances the extraction.  Table 1. Quantification of interunit linkages,s ubstructures and total extraction yield (Figure3). All interunitl inkages and substructures are semi-quantified per 100 Ar units (for diagnosticc hemical shifts, see Ta ble S1). For the 0.5 %a nd 1.5 %a cid series, values are presented as mean AE SD where n = 9a nd n = 5 [a] for the 0.5 %s eries.T he number of n reflects the number of collected fractionsi nt he trend seriesi nvestigation. For the 1.5 %s eries, n = 5. The high SD for the b-1' stilbene structures can be attributed to ad ownward trenda nd for the b-5' stilbene structures an upward trend from the first to the last fraction.T he HSQC spectraf or the 1.5 %a nd 0.5 %a cid series are included in the Supporting Information( FiguresS3-S7a nd Figures S8-S16, respectively). The results of 31 PNMR spectroscopy (Figure 7a nd Ta ble S3), show ar elativelyh ighc ontent of aliphatic hydroxy groups in the earlier part of the cycle, suggestingt hat the native aliphatic side-chain configurationi sq uite well preserved. In native wood lignins, Ca and Cg are mostly hydroxylated, the latter to ah igherd egree than the former.T he content of aliphatic hy-droxy groups seems to decreasew ith increasing cycle number, indicating side-chain reactions. The content of noncondensed phenolic hydroxy groupsd ecreases slightly and then seemst o level off as the cycle number increases. This suggestst hat lignin depolymerization, which normally occurs through aryl ether cleavage, is not significant when using the static cycle method.I nc ontrast, the content of C5-condensed phenolic hydroxy groups increases slightly at the beginning of the cycle then level off.
Ac omparison between the hydroxy functionalities of the lignin obtained by the integrated (average values) cyclic extraction method and that obtained from the 2h organosolv extraction is shown in Figure 8. The 2h extracted lignin is found to have al ower aliphatic hydroxy content and ah igher phenolic hydroxy content;t hese observations are consistent with the occurrence of side-chain reactions and the cleavage of aryl ether linkages, respectively.
The native interunitl inkages ( Figure S2) in the pooled cyclic extracted lignins (together with other linkages;s ee Figure 3) were investigated by HSQC (Figure9). The previously discussed signals relatingt o5 -5' condensation products-C2Ar-H and C6Ar-H appearing at 6.7/112.5 ppm and 6.6/120.5 ppm, respectively-were comparedb etween the cyclic series ( Figure 9, for full HSQC spectrum,s ee Figure S20) and the reference samples ( Figures S18 and S19). Stronger signals in support of the condensation reaction could be observed in the reference samples, consistent with the 31 PNMR results ( Figure 8). In addition, b-O-4' bonds are present in both a-hydroxylated and etherified forms.
The trends of b-O-4' and b-5' concentrations from the 0.5 %a cid study and the final cyclic methoda re compared in Figure 10. From fractions2t o9in the trend study, b-O-4' concentrations are seen to decrease slightly,b ut are still in the region of 30-38 per 100 aromatic rings, and are significantly higher com-    pared to the aforementioned organosolv reference with around7per 100 aromatic rings. In fact, the b-O-4' amounts in the cyclic series are at similar levelst ot hose of milled wood lignins (MWL)p repared from both the originalw ooda nd the subcritical water-extracted wood meals used in this study.H igh yields of lignin with good preservationo fn ative structures using static cycles can thus be substantiated. Them ildness of the extraction is also manifested in the detection of dibenzodioxocin and trace amountso fs pirodienones tructures (Figure S17), which are traditionally easily modified or degraded during extraction.
SEC was conducted on the cyclic fractions ( Figure 11 and Ta ble S5). It is seen from the chromatograms that the lignin from the first two cycles have as ingle distribution while the lignins from the third cycle and upwards show overlapping chromatograms. The polydispersity index is 3.0-4.8 and the DP n is 9-19 for the ten cyclic fractions. Except for the first three cycles, there is no linear trend in molecular weight (Table S5).
From ocular inspection, we observed that the lignin fractionsh aved ifferent colors andt hat the more "native"-like lignin fractions obtained by the cyclic method appear paler,w ith al ight beige color, comparedt ot he reference organosolv ligninse xtracted at 2h and3 h, which have as ignificantly darker color (Figure 12). This difference in color is probably duet oahigherd egree of lignin condensation in the reference lignins. Interestingly,t he first sample in the cyclic lignins sticks out in being slightly darker than the rest. This could be explained by the presence of extractivesi nt his first cycle. From the HSQC ( Figure S23), signals from extractives were more prominenti nt his fraction than the subsequent ones. These signals appear in the nonoxygenated aliphatic regions.H MBC ( Figure S26) analysis substantiated the presenceo fu nsaturated fatty acids or esters. The 0.5 %a cid series fractions followed the same color patterns as the 1.5 %s eries ( Figure S32).
For practical reasons, the cyclic method was further developed as an integrated method, that is, all cycles were pooled.   For this purpose, the 1.5 %a cid series was chosen over the 0.5 %a cid series. This was due to the higher lignin yield, althought he b-O-4' content was slightly lower,T able 1. The final integrated cyclic method (Figure 9) reflects an average of interunit linkages from all previousc ycle series with a b-O-4' content of 30 %a nd a b-5' content of 12 %( Ta ble 1). The collected results from the different extraction series, based on the consolidatedb iorefinery steps illustrated in Figure 1, are given in Ta ble S10.

Fractionation for narrowp olydispersity
The sample obtained from the integrated cyclic method was studied by SEC and shownt oh ave aD P n of around7and of 4.4 (Table 2). In general, low values are preferred if such lignins are to be used directly as polymer precursors. The crude fractionso btained here could require further refining due to their high values.
To decrease ,f ractionation with ethanol (99 %) was investigated and found to be efficient. Twof ractionsw ere obtained: an ethanol-soluble fraction and an ethanol-insoluble fraction, with yields of 48 %a nd 52 %, respectively,a nd values of 2.1 and 3.6, respectively ( Table 2). The insoluble fraction had a higher molecular weightthan the soluble fraction, yet, interestingly,b othf ractionsh ad as imilar content of b-O-4' bonds,a t 30 % ( Figures S21 and S22). This observation has mechanistic implications and will be discussed later.D etails regarding the SEC analysisare given in Figure S29 and Ta ble S7. 31 PNMR spectroscopy (Table S4) show that the C5-condensed phenolic contenti sl ower for the ethanol-soluble fraction than for the insoluble fraction,a sw ell as for the unfractionated lignin sample. Fractions 1a nd 2f rom the 1.5 %a cid cyclic series were studied in as imilar fashion. In both cases, the initial was reduced for the ethanol-soluble fractions. A similar trend was observed for fraction 2( Figure S28 and Ta ble S6). Overall, narrow values were successfully obtained by post-fractionation using ethanol as as ustainable solvent.

Reactionm echanisms
From the results presented so far,afew points stand out to reveal insights into am echanistic understanding of the extraction process.T hese will now be discussed.W hen comparing the lignins obtained throughs hort cycles with the reference physically protected throught he developed cyclic extraction processing strategy.T he physical protection resultsf rom the periodic removal of the dissolved components from the reactor to ambient conditions. In thisw ay,t he dissolved molecules are not exposed to the reaction conditions for al ong duration, therebyl imiting further reactions. The lignin reactions that do take place during the short residence time in the cycle are shown in Figure 13 A. In addition, the solute concentrationi s maintained at al ow level, owing to displacement with fresh solvent. This resultsi nalower probability of occurrence of lignin condensation reactions, which would require molecular collisions.
Analysis of the number average molarm ass( M n )o ft he lignin fractionso btained by the cyclic methods suggests some differences, yet the b-O-4' contentso ft hese fractionsa re the same.Asimilaro bservation can be made from the lignin that was refinedf urther by ethanol fractionation (ethanol-soluble and ethanol-insoluble fractions, Figures S21 and S22). In fact, analysiso ft he DP n of these refinedf ractions( Ta ble 2) shows a threefold difference. This suggests that condensation reactions betweenf ractions with similar content of b-O-4' structures take place. The content of b-O-4' structures in the formed molecule resulting from such condensations would be the same. Several types of condensation reactions could potentially occur under acidic conditions. HSQC analysis showed the presence of stilbene structures, which are formed from the elimination of formaldehyde in phenylcoumaran( b-5')a nd spirodienone (b-1')s tructures. This formaldehyde could participate in condensation reactions involving the electron-rich position para to the methoxy group. However,n os ignals in the HSQC spectra that would attest to the formation of am ethylene bridge in the resultant product could be identified. On the other hand, the 5-5' couplings showed as tronger signal in the reference organosolvl ignins when compared to milled wood  lignins (MWL) from the originalw ood meal andt he subcritical water extracted residues. This indicated that these bondsw ere also formed during the organosolvc onditions. We propose a mechanism for how these are formed. This is indicated in Figure 13 B. Prolonged exposure of the dissolved lignin to the high extraction temperature inducedh omolytic cleavage of some labile b-O-4' linkages resulting in the simultaneous formation of beta radicals and phenoxy radicals. The phenoxy radicals resonate to C5 radicals tructures, which couple to form stable 5-5' linkages. Apart from the 5-5' couplings seen in the HSQC analysis, the formation of C5-condensed phenolics is furthers upported by 31 PNMR spectroscopyf or the cyclic extracted lignin samples (Figure 7), where an increasing trend is observed. Substantiation of the proposed mechanism is even more clearly consistent with our observations that the lignins obtained through the static cycle approach have as ignificantly higher b-O-4' content and lower content of 5-5' structures, when compared to the 2hand 3hextractions.
Another possible reactionisthe cleavage of b-O-4' structures by heterolysis, but this would result in the formation of Hibbert's ketones, which were only detectedi ns mall amountsb y the HSQC analysis (Table 1). Homolysis therefore seems to be the main reaction pathway for the cleavageo fb-O-4' structures, subsequently followed by radical recoupling reactions to form 5-5' condensed structures ( Figure 13). Such reactions are more prominent in the reference organosolv (2 ha nd 3h ex-tractions) and can be minimized by adopting the cyclic methoda pproach. The temperature dependence of homolytic cleavage of b-O-4' structures has been reported [33] and the temperature used here falls in that regime (above 130 8C). The occurrence of radical repolymerization, in accordance with the described mechanism, is also supported by the SEC data ( Figure S30), which shows that the 2h-extracted ligninsa re in the same molecular weight regime as the pooled cyclic extracted lignins, albeit with approximately fourfold lower b-O-4' content.

Chemical composition and mass balance of the biopolymer from the consolidated biorefinery
The analysisw as performed on the Wiley-milled wood, hydrothermale xtract, and the fiber residue after hot water and organosolv extraction. The carbohydrate composition( Ta bleS8), Klasonl ignin (KL;T able S9), and acid-solublel ignin (ASL; Ta ble S9)for spruce wood are reported in the Supporting Information. The mass balance of the extracted samples is given in Ta ble 3. The lignin balance for the 1.5 %a cid series shows that roughly6%o ft he lignin ends up in the subcritical water extract that contains the bulk of hemicellulose, 55 %o ft he lignin ends up in the organosolv-obtained pure lignin fraction,a nd 34 %o ft he lignin remains in the residual fiber fraction (see Figure 1). Thus, 94 %o ft he lignin balance is accounted for and the remaining part is probably lost during purification as water soluble fractionsf or example, as lignin carbohydrates complexes. Roughly 83 %o ft he total wood mass balance is accounted for in the obtained fractions. The missing fractionsa re most likely small water-soluble molecules that were not recovered. These include hemicellulose-derived components,s uch as O-acetyl groups and monomeric sugars, that result from autohydrolysisd uring the subcritical water extraction step and acid catalyzed hydrolysis in the subsequent organosolv extraction step. Recovery processesf or such molecules will be critical for future processes.

Potential uses of components from the consolidated biorefinery
The hemicellulose-rich extracts could be hydrolyzed to monomeric sugars for furtherp roduction of platform chemicals. They could also be fermented to ethanol, which would not only be attractive fort he organosolv process economicsb ut could also support the circularity of the process if the produced ethanol was used internally.T he cellulose-rich fiber fraction could also be used in as imilarw ay to the hemicellulose fractionsd escribed above. Alternatively,f iber-based composites are becoming attractive and the presenceo fl ignin in the fibers has been shown to enhance the thermomechanical properties of such materials. [34] Based on the lignin content of the fiber residue in this work, such potential applications could be investigated.
The lignin fraction could be used directly as ap olymer precursor for materials ynthesis or catalytically depolymerized to platform monomers. Based on their functionality,structure, low DP n and demonstrably narrow ,t hese lignins might be suitable as polymer precursors for the synthesis of thermosetting resins, as showni nr ecent studies. [35,36] In those cited studies, oligomeric fractions were shown to be preferable to polymeric fractionsf or the synthesis of homogeneous materials, due to their mutual solubility with other chemical components used in the synthesis.
The cyclically extracted lignin is also an attractive precursor for platform monomers. In this context, catalytic depolymerization is favored because of the high aryl ether contento ft he extracted lignin. In recent years, innovative methods for conversion of lignin into platform monomers have emerged and include reductive catalytic fractionation (RCF) [37] and base-catalyzed depolymerization (BCD). [38]

Conclusions
The consolidationo falignin biorefinery with hemicellulose and fiber production using green solvents was investigated. Twos olvent systems werea pplied in sequence, those being a subcritical water system and an ethanol-water system with the addition of acid catalysts. From these systems three fractions were obtainedv iz. hemicellulose, pure lignin and af iber fraction enriched in cellulose. The study was then devotedt of urther investigation of the ethanol-water extraction step with the pre-set criteria of obtaining native-like lignins in high yield and purity.T ot his effect, ap rocessing strategy to preserve the structurali ntegrity of lignin through both physical protection and additive-free chemical protection was developed. In this context,astatic cycle extraction approachw as found to be key to the fulfillment of the preset criteria thankst ot he minimizationo fl ignin condensation reactions in this setup. The static cycle methodw as contrasted with ac lassicalr eference ethanol-water extraction performed under the same conditions but differentiated by an unperturbedl ongere xtraction time. Lignin condensation reactions were found to be significant in the latter method and yieldeds table 5-5' bonds. The associated condensation mechanism is proposed to start with ah omolytic cleavage of aryl ether linkages forming phenoxy radicals, as well as beta radicals. The former radicals have resonance structuresw ith radicals at position C5, which in turn can couplet of orm stable CÀCb onds. The typicallye xpected lignin condensations at the benzylic cation under acidic conditions did not occur,w hich is in part explained by ac hemical protection through capping by etherification with ethanol. Furthermore, no condensation products of lignin involving formaldehyde (which is released when stilbenes are formed)c ould be detected.
An essential milestone fort he field is the developmento fa fundamental understanding related to the consolidation of the lignin biorefinery.I nt his regard, this study provides ap ath towards such consolidation where green processing strategies are combined with am echanistic understanding that is essential to optimize the processes.

Material
The debarked wood was milled by using aW iley mini mill (3383-L70, Thomas Scientific). The extraction was performed by using an ASE 350 Accelerated Solvent Extractor (Dionex, Sunnyvale, CA, USA). The samples were placed in Dionium Extraction cells, 34 mL (Stainless steel extraction cells) or 66 mL (Dionium extraction cells) in size, containing ag lass fiber filter.T he extract was collected in 250 mL collection bottles. Extraction filters (Duran filter funnel, diameter 60 mm, 10-16 microns) were purchased from Sigma-Aldrich. The molecular weight distribution and dispersity indices were investigated by using as ize-exclusion chromatography system using refraction index detection (SECurity 1260, Polymer Standards Service, Mainz, Germany). The system included an autosampler (G1329B), an isocratic pump (G1310B), and an RI detector (G1362A).

Methods
Spruce wood chips were first debarked and ocularly examined, where only bright wood without defects was collected. The wood chips were then Wiley-milled to 40 mesh. All the following wood meal weights are given on oven-dry basis. Since ASE instruments are programmed to keep ac ertain pressure, the exact amount of liquid is not constant in the static cycles using the standard method. Another consideration is that wood components are continuously removed in the cyclic extraction, inducing ac ontinuous change in the liquid/wood (L/W) ratio. However,t he L/W ratio is still roughly estimated in the following sections presented below.
The extraction process is divided into three sections. 1) Ar eference sample extraction;2 )ani nvestigation of extraction trends and properties of the lignin fractions;3 )development of acyclic extraction method for lignin: 1) Wiley-milled wood (3.8 g) was placed into a3 4mLs tainless steel extraction cell. In the first step, a2hh ot water extraction (HW) was performed followed by as econd step comprising an organosolv extraction for 2o r3h. Instrument parameters were as follows: 160 8C, af ixed volume of 40 mL, and ap urge time of 90 sw as used for both the HW and the organosolv extraction. The extraction was performed at ap ressure of 1500-1600 psi. The samples were extracted with as olvent system composed of 1.5 wt %H 2 SO 4 in an aqueous ethanol solution (30:70 v/v). For the HW extraction, the L/W ratio was 10.5. The L/W ratio for the organosolv extraction was estimated to be 14.
2) Extraction series were made for H 2 SO 4 additions of both 1.5 wt % and 0.5 wt %t oab inary solvent aqueous ethanol solution (30:70 v/v)s ystem. For the 1.5 wt %a cid series, Wiley-milled wood (4.80 g) was placed into a3 4mLe xtraction cell. AH We xtraction was performed for 2h,a t1 60 8C, using af ixed volume of 40 mL with a purge time of 90 sf ollowed by an organosolv extraction which was performed 10 times for 5min each at 160 8C, with af ixed volume of 40 mL and using ap urge time of 90 s. For the HW extraction, the L/W ratio was 8, whereas that for the organosolv extraction was estimated to by 11 for the first fraction and 13 for the last. For the 0.5 wt %e xtraction procedure, 10.1 go fw ood was placed in a6 6mLD ionium extraction cell. The parameters for the HW extraction was 2h of extraction, 160 8C, af ixed volume of 70 mL and ap urge time of 90 s. The organosolv extraction was performed 10 times for 5min each, at 160 8Cw ith af ixed volume of 60 mL and ap urge time of 90 s. After each 5min extraction, the extract was collected for further sample preparation. For the HW extraction, the L/W ratio was 7. The L/W ratio for the organosolv extraction was estimated to be 9f or the first fraction and 11 for the last.
3) Wiley-milled wood (9.3 g) was placed in a66mLD ionium extraction cell. The amount of wood meal was linearly scaled up from the 1.5 wt %a cid method [described in (2)] using 34 mL cells to the 66 mL extraction cells. First, aH We xtraction of 2h extraction at 160 8Cu sing af ixed volume of 70 mL and ap urge time of 90 s was performed. The subsequent organosolv extraction was immediately performed in 15 static cycles using the standard method, with 5min each at 160 8Cu sing ar inse volume of 100 %a nd a purge time of 90 sw ith the solvent system 1.5 wt %o fH 2 SO 4 in aqueous ethanol solution (30:70 v/v). In the HW extraction, af ixed volume program was used and the L/W ratio could be determined to be 8. At the beginning of the organosolv extraction, after the hemicellulose fraction had been extracted, the L/W ratio was estimated to by 10 and at the end of the cycle, it was estimated to be 12. The total amount of solvent used in the cyclic organosolv method was 340 mL.
As mentioned earlier,t he ASE instrument operates at af ixed pressure of 1500-1600 psi in the standard method. When the fixed volume program was used, the solvent volume was selected to reach ap ressure of 1600 psi in the cell so as to achieve sufficient pressure and subcritical conditions. The fixed volume program was used in all experiments except for the last static cycle method where the standard method was used. More liquid was used in the extraction series since every fraction was collected and analysed manually.T he ASE instrument has an integrated oven and temperature control system, from which the temperature is monitored.
After the system has pumped the solvent into the cell, the cell is heated for 8min before the extraction procedure starts.
The HW extract was lyophilized directly.T he lignin sample obtained from the organosolv extraction was evaporated under reduced pressure. During this evaporation the pH was monitored, and water was added to avoid ac hange in the acidity of the extract. The precipitated lignin in this acidic water solution was vacuum filtrated and rinsed with water until ac lear filtrate was obtained. The efficiency of the wash was substantiated by HSQC analysis, during which no signals from carbohydrates were detected. The rinsed lignin samples were collected and lyophilized. As chematic illustration of the method is shown in Figure 14.
Fractionation was performed by adding lyophilized lignin/ethanol (1:40 w/w)t oac losed vial with magnet stirring for 2h.T he solution was filtered under vacuum filtration and the residue rinsed with asmall amount of ethanol.
The MWL was prepared according to the Bjçrkman procedure [39] with some slight modification. Shortly,i naTe flon-lid bottle, dioxane-water mixture (200 mL, 96:4 v/v)w as added to extractive-free ball-milled spruce wood (10 g)

Chemical composition of the biopolymer fractions
The carbohydrate composition was investigated according to the acid hydrolysis protocol. [40] The Klason lignin and acid-soluble lignin (ASL) contents were determined as previously reported. [41,42] Hydrolysis was performed on the Wiley-milled wood fraction, the hydrothermal extract fraction as well as the fiber residues after hot water and organosolv extraction. In short, to 200 mg of the respective fractions, that is, wood, extracted fibers and the hydrothermal extract fractions, 72 %s ulfuric acid (3 mL) was added. The mixture was placed under vacuum for 80 min with occasional stirring. The mixture was thereafter diluted with Milli-Q water (84 mL) and placed into an autoclave for 60 min at 125 8C, following by vacuum filtration and 5 2mLr insing of the collected Klason lignin on the glass fiber filter.
Carbohydrate quantification was performed by using high-performance anion-exchange chromatography with pulsed amperometric detection (HPAEC/PAD). The method setup has been previously reported. [43] Using 260 mm sodium hydroxide and 170 mm sodium acetate, the system was equilibrated for 7min followed by equilibration with Milli-Q water for 6min. Milli-Q water was used as an eluent at af low rate of 1mLmin À1 .A tt he column eluate, 300 mm sodium hydroxide was added before the PADc ell, at a flow rate of 0.5 mL min À1 .Q uantification was carried out by using anhydro correction factors of 0.90 and 0.88 for hexoses and pentoses, respectively,a ccording to apreviously reported method. [44] The Klason lignin was gravimetrically quantified after being ovendried overnight. The ASL was quantified by UV spectroscopy at 205 nm using an extinction coefficient of 128 Lg À1 cm À1 for softwood and ac orrection factor of 0.2 for carbohydrate degradation products. [45] X-ray diffraction The X-ray diffraction was performed using an ARL X'TRA Powder Diffractometer (Thermo Fisher Scientific Inc.,U SA) using Cu Ka radiation generated at 45 kV and 44 mA. The measurements were per-formed using scans from 2q = 58 to 508 in steps of 0.058 at ascan rate of 3sper step.

Size-exclusion chromatography
Lyophilized sample ( % 9mg) was dissolved in a0 .5 wt % LiBr solution in DMSO (2 mL). The dissolved sample was syringe filtered using a0 .45 mmP TFE filter.S EC was run using 0.5 wt %L iBr solution in DMSO as eluent, with an injection volume of 100 mL, af lowrate of 0.5 mL min À1 , and ac olumn oven temperature of 60 8C. For integration, RI detection at 40 8Cw as used. Standard calibration was performed by using Pullulan standards in the molecular range of 342-708 10 3 Da.

NMR spectroscopy
For the HSQC-edited analysis, lyophilized sample (80 mg) was dissolved in [D 6 ]DMSO (600 mL). The spectra were acquired on a Bruker 400 DMX spectrometer with the "hsqcedetgp" pulse sequence using the following parameters:a nacquisition time of 0.1065 s, ar elaxation delay of 2.5 s, 80 scans using 1024 256 increments. Optimal pulse lengths corresponding to a9 0 8 pulse were found for each experiment by finding and halving the pulse length corresponding to a1 808 pulse where the proton FID signal was minimal. Data processing was carried out in MestReNova with 1024 1024 data points using a9 0 8 shifted square sine-bell apodization window.T he data was Fourier transformed followed by phase correction and baseline correction in both dimensions by a Bernstein polynomial fit of order 3. Semi-quantification of lignin interunit linkages was carried out by using the C2ÀHs ignal region on the aromatic ring as an internal standard. [46] All NMR spectra were integrated by using the same shifts for comparable results (Table S1).
HMBC analyses were performed on the same samples as for the edited-HSQC analyses using the same instrument and the same acquisition parameters, except for the use of the 'hmbcgpndqf" pulse program.
Quantitative 31 PNMR sample preparation was performed based on ar eported method. [47] Lyophilized sample (30 mg) was dissolved in N,N-dimethylformamide (100 mL) and pyridine (100 mL). To this solution, internal standard (IS) solution (50 mL; 60 mg mL À1 of eHNDI in pyridine with 5mgmL À1 Cr(AcAc 3 )r elaxing agent) was added. After stirring, Cl-TMDP phosphorylating agent (100 mL) was added following by dropwise addition of CDCl 3 (450 mL) to the sample solution. The 31 PNMR spectra were acquired with 512 scans and ar elaxation delay time of 6son aB ruker NMR spectrometer AvanceIII HD 400 MHz. Data processing was carried out in MestReNova. The data were Fourier transformed followed by phase correction and baseline correction in both dimensions by aB ernstein polynomial fit of order 3. Diagnostic peaks with assigned shifts are given in Ta ble S2.