Towards sustainable pulping of bagasse: silica migration, hexenuronic acid removal and ultrafiltration-based spent liquor utilization

Abstract

Transforming the utilization of agricultural residues from single-purpose papermaking into high-value biorefinery strategies can unlock their full potential. The industrial valorization of bagasse is currently hampered by the inefficiency in bleaching chemical use and the limitations of alkali recovery out of silica-rich spent liquors. To address these challenges, this study establishes an integrated biorefinery strategy combining an acid hydrolysis stage (A-stage) for reducing usage of bleaching chemicals along with membrane-based fractionation of spent liquor. In the fiberline, incorporating an A-stage prior to elemental chlorine-free (ECF) bleaching selectively removed half of the hexenuronic acid (HexA). This targeted elimination reduced chlorine dioxide (ClO₂) consumption by 25% despite an expected compromise in losing viscosity of bleached pulp fiber. Possible hypothetical migration (fate) of silica from its free amorphous silica form (SiO₂) at bagasse to primary form of calcium silicate (CaSiO₃) at the pulp fiber is discovered. Ultrafiltration utilizing a 0.5 kDa-sized membrane retained a high-molecular-weight lignin–xylan complex in the retentate while allowing the permeation of 72%-recoverable hydroxy acids (enriched with 2-hydroxybutanoic acid and glucoisosaccharinic acid) and alkali from the permeate, opening up the promise of applying a tailored ultrafiltration separation in addressing limitations in alkali recovery for spent liquor fractionation for bagasse mill.

---

Green foundation

1. The industrial valorization of bagasse is currently hampered by the inefficiency in bleaching chemical use and the limitations of alkali recovery out of silica-rich spent liquors. Our study establishes an integrated biorefinery strategy combining acid hydrolysis pretreatment (A-stage) for reducing usage of bleaching chemicals along with membrane-based fractionation of spent liquor.

2. In the fiberline, incorporating an A-stage prior to ECF bleaching selectively removed half of the HexA and reduced ClO₂ consumption by 25%. Possible migration of silica from its free form SiO₂ at bagasse to form CaSiO₃ at the pulp is discovered. Ultrafiltration utilizing a 0.5 kDa-sized membrane retained a HMW lignin–xylan complex in the retentate while allowing the permeation of 72%-recoverable hydroxy acids and alkali.

3. Future research shall tailor acid charge and temperature into a balance between chemical savings and pulp quality. Fine-tuning is needed to maintain separation stability of hydroxy acids and alkali recovery from permeate.

Introduction

Nowadays, non-wood raw materials, especially agricultural residues, are seen as potential alternatives to wood-based feedstock in the pulp and paper industry, contributing to reduce dependence on forest resources. Among these, bagasse, a well-known agricultural industrial residue, is considered a highly promising biorefinery feedstock due to its rapid renewability and widespread availability in regions such as Brazil, India, and China.¹ However, the full valorization of bagasse still presents challenges. Hexenuronic acid (HexA) groups present in xylan consume additional bleaching chemicals, although HexA content in bagasse kraft pulp (16.5 μmol g⁻¹) is significantly lower than that in unbleached kraft pulp of softwood (21.8 μmol g⁻¹) and eucalyptus (50.45 μmol g⁻¹).^2–4 The conventional Tomlinson recovery boiler is generally not economically or volumetrically tailored for bagasse pulp mills due to the chemical profile differences of spent liquor out of wood and non-wood.⁵

Hexenuronic acid (HexA) is a critical side-product formed during the alkaline pulping of xylan-rich biomass. Under high-temperature alkaline conditions, the 4-O-methyl-D-glucuronic acid side groups on xylan undergo β-elimination, releasing methanol and generating the unsaturated HexA structure.^6,7 Its electron-rich C=C bond contributes to the kappa number and rapidly consumes electrophilic bleaching chemicals. To mitigate this, the acid hydrolysis stage (A-stage) has been developed for treating pulp at pH 3.0–3.5 and 85–95 °C for 1–2 hours,⁷ through which it is possible to selectively cleave over 80–90% of HexA groups with minimal degradation of hemicelluloses. Selective acid hydrolysis can remove HexA (20–60 μmol g⁻¹ pulp) along with the reduction of the kappa number by 2–7 units, saving 30–40% of bleaching chemical in, for example, the ECF bleaching of birch kraft pulp.⁷ D-hot pretreatment (D-hot, 85 °C) removed approximately 19% more HexA (Table 1) and reduced the generation of AOX by about 35% compared to the conventional D₀ (60 °C) stage.² The key differences between A/D₀ and D-hot is removal of HexA relative to chlorine dioxide treatment. Hydrolysis of HexA originating from hemicelluloses can also be achieved biologically by xylanase (Table 1).⁸

Table 1 Comparative studies of bleaching pretreatment (regular font) and spent liquor fractionation (italics) based on the conventional strategy and our strategy

Strategy	Methodology	Mechanism	Pulp (or NaOH%)/biomass	Main findings	Ref.
Conventional (bleaching)	D-hot (T 95 °C, pH 3.2, t 1 h)	HexA hydrolysis happens after ClO2 oxidation	Unbleached pulp/bagasse	19% more HexA removal; brightness increase; kappa reduced 11% with D-hot	2
	Xylanase (0–25 IU g^−1) (T 65 °C, pH 9, t 2 h)	Hydrolysis of HexA-origin hemicelluloses by hemicellulases	Unbleached pulp/bagasse	Reduction of 21.4–26.6% AOX and 12.5–22% ClO2 with xylanase	8
Conventional (spent liquor)	Weak cation exchange resin in Na^+ form	Reversible electrostatic interaction to separate charged ions	22% NaOH; T 165 °C/birch	50% recovery of hydroxy acids (HAs) along with 95% lignin	12
	5, 10 and 15 kDa cut-offs	Membrane separation on basis of size	7.5% NaOH; T 90 °C/grass	5/15 kDa-sized permeate comprise high fractions of low molecular weight lignin	13
	5, 15 and 50 kDa cut-offs	Membrane separation on basis of size	23% NaOH; T 152 °C/Eucalyptus globulus	Retentate of 50 kDa obtained more lignin; 5 kDa permeate contains phenolic monomers	14
Our strategy	A-stage (pH 3.5, T 95 °C, 2 h); 0.5 kDa cut-off	HexA hydrolysis increases along with decreasing pH; membrane separation on basis of size	Unbleached pulp; 15% NaOH; T 165 °C/bagasse	25% savings of ClO2; 73% recovery HAs from permeate and lignin–xylan complexes from retentate; silica migration is observed	This study

---

Another shift is also required regarding the utilization of spent liquor from non-wood pulp mills. Traditionally, this stream is concentrated (dry solids content increases from 14–16% to 75–85%) at the evaporation plant and combusted for energy at the recovery boiler linked to alkali recovery through the lime kiln and causticizing plant (Fig. 1b). However, this approach represents an undervaluation of organic resources, particularly for bagasse liquors with much lower calorific values. Moreover, soda cooking has been a significant method for pulping bagasse. During alkaline cooking, approximately 40–50% of the initial carbohydrates degrade via the “peeling reaction” into low-molecular-weight HAs, including lactic acid, glycolic acid, and 2-hydroxybutanoic acid.^9,10 These C1–C6 organic acids can be catalytically upgraded to carboxylic acid products.¹¹ 50% of HAs and over 95% of lignin can be recovered from soda and kraft black liquors through an ion exchange chromatographic method (Table 1).¹² To address the challenge of separating these small molecules from macromolecular lignin, membrane ultrafiltration (UF) has emerged as a promising strategy. Utilizing UF membranes with specific molecular weight cut-offs (MWCO, e.g., 1–5 kDa) can effectively fractionate the industrial kraft black liquor, from where the retentate comprises high-purity lignin while the permeate recovers a filtrate rich in HAs and residual alkali (Table 1).^13,14

Fig. 1 Methodology and long-term vision. (a) Experimental process flow of the bleaching and ultrafiltration after soda cooking of bagasse. (b) Outlook of our concept is highlighted with green dashed lines; dashed arrow lines are used to indicate theoretical frameworks that are outside the scope of this present study. The applied analytics are summarized as given in Table S1.

To address these technical challenges, this study proposes an integrated biorefinery strategy combining targeted reduced usage of bleaching chemicals along with membrane-based recovery of process chemicals from the bagasse fiberline. Specifically, we investigated the introduction of an acid hydrolysis stage (A-stage) prior to elemental chlorine-free (ECF) bleaching, aimed at selectively removing HexA to minimize use of bleaching chemicals. Moreover, ultrafiltration paves the way as a promising pretreatment step for recovery of alkali, lignin–xylan complexes, and HAs prior to their incineration at the power boiler for energy supply. However, the ultrafiltration step is not positioned as a stand-alone technological alternative to an alkali recovery furnace infrastructure. The long-term vision (or concept) is to replace the existing recovery boiler (and evaporation unit) with a much smaller and financially more affordable alternative solution such as a simplified membrane and power boiler unit for closing the alkali recovery cycle and meeting the energy demand of bagasse pulp mill typically achieving pulp production of 100–500 air-dried tons per day (ADt d⁻¹).

Materials and methods

Raw materials and chemicals

The bagasse feedstock was procured from Guangxi Yongxin Huatang Group (China). Prior to shipment to Finland, the biomass underwent an overnight air-drying process at 50 °C to stabilize moisture content. For the spent liquor fractionation, flat sheet polyethersulfone (PES) membranes (NADIR® NP030) with an MWCO of 0.5 kDa were acquired from MANN + HUMMEL (Germany).

All chemicals were of analytical grade and used as received without further purification. Regarding chemical reagents, the chlorine dioxide (ClO₂) was supplied by Kemira Oyj (Finland). General reagents for pulping and bleaching, including sulfuric acid (H₂SO₄, 95–97 w/w%), sodium hydroxide (NaOH, 98.5–100%), and hydrogen peroxide (H₂O₂, 30%) were purchased from VWR International (Finland). Hydrochloric acid (HCl, 37 w/w%) and magnesium sulfate (MgSO₄, 99.5%) were obtained from Merck KGaA (Germany). For analytical characterization, the silylation agents O-bis(trimethylsilyl)trifluoroacetamide (BSTFA) containing 1% trimethylchlorosilane (TMCS) along with deuterated solvents for NMR (dimethyl sulfoxide-d₆, 99.9 atom% D; pyridine-d₅) were sourced from VWR (Finland). Additional analytical standards and reagents included tetracosane and pyridine (Merck) for GC-MS analysis, sodium formate and 2-furoic acid (MedChemExpress, Sweden) for HexA determination, and nitric acid for ICP-OES (Sigma-Aldrich). Reagents for iodometric titration, including sodium thiosulfate (Na₂S₂O₃), potassium iodide (KI), ammonium molybdate (catalyst for H₂O₂ determination), and starch were obtained from Sigma-Aldrich.

Experimental procedures

Pretreatment and pulping. The raw bagasse was first screened using an allgaier device (Allgaier Process Technology GmbH, Germany) to remove non-fibrous parenchyma cells, fines and dust (<3 mm). This screened fibrous bagasse (B, 40 w/w% raw bagasse, Table S1) was used as the feedstock for all subsequent experiments. For consistency, all chemical dosages, pulp yields, and quantitative analyses reported in this study are calculated based on the oven-dry weight of fibrous bagasse (B). To mitigate silica scaling and enhance delignification, a two-step pretreatment was applied as shown in Fig. 1. First, water washing was conducted using a stainless-steel reactor at 55–60 °C for 30 min (liquid : solid, L : S, 20 : 1) to remove water-soluble extractives. Subsequently, an alkali wash was performed using 1 M NaOH (L : S, 10 : 1) at 60–62 °C for 60 min in a pressurized air-bath reactor (Haato-tuote, Model 16140-538) (Haato Oy, Finland) ideally for leaching silica and waxes. Soda cooking was carried out in the same reactor using an active alkali charge of 15% (as NaOH). The cooking profile consisted of heating to 165 °C and holding for 60 min (H-factor, 611); the alkali wash and soda cooking were both performed under an N₂ atmosphere. The cooked pulp was separated from the soda spent liquor by filtration using a nylon pulping bag (particle size of 50 µm). The pulp was washed and screened (0.35 mm slot) using a Mänttä flat screener (G.A. Serlachius, Finland), while the spent liquor was collected and stored for analytics and further ultrafiltration treatments.

Soxhlet extraction. The fibrous bagasse (B) was subjected to successive Soxhlet extraction cycles on the same batch to remove lipophilic extractives and waxes. Approximately 10 g of the air-dried biomass was successively extracted with dichloromethane (DCM) and acetone in a Soxhlet apparatus. The extracted solid residue was then air-dried to remove residual solvent before its complete drying using the freeze dryer and finally stored in a desiccator for subsequent experiments.

Bleaching sequences. The ECF bleaching includes a standard D₀-E_p-D₁ sequence and a modified A/D₀-E_p-D₁ sequence incorporating an A-stage. The conditions and chemical charge are listed (Fig. 1). Bleaching experiments were performed in sealed 1 L-sized borosilicate glass vessels (Duran) immersed in a thermostat-controlled water bath (Kottermann GWB, Germany). Two ECF sequences were evaluated to determine the efficacy of HexA removal and delignification. Following each stage, the pulp was thoroughly washed with excess water in a plastic bucket by manual mixing, followed by centrifugation to collect the bleaching filtrates. The concentrations of ClO₂ and H₂O₂ were quantified following standard TAPPI T 611 cm-07 titration.

Membrane fractionation of spent liquor. Spent liquor fractionation was executed using a 6-layer filter discs module (LabStak M20-0.72, Alfa Laval, Sweden) fitted with a stack of 12 polyethersulfone (PES) flat-sheet 0.5 kDa-sized membranes having an area of 0.018 m² per membrane (NP030P, MANN + HUMMEL, Germany). To ensure seal integrity and prevent leakage during the operation, the membrane stack was hydraulically compressed to a closing pressure of 375 bar to hold the filter plates together. The pressure values for the inlet and outlet of the membrane stack are 12 bar and 7.5 bar, respectively. The pressure difference must be maintained at roughly 5 bar to preserve the back pressure for the membranes, otherwise the membranes could burst through the holes of the disks. The inlet pressure is set with the rotational speed of the pump motor, and the pressure difference can be adjusted with the tap underneath the outlet pressure gauge assembly. The entire filtration process was conducted at ambient temperature to minimize thermal degradation of the analytes. The spent liquor initially fed into the ultrafiltration system is referred to as the “feed”. The feed is divided into two distinct streams during the membrane separation: the “retentate” containing macromolecules that cannot pass through the membrane pores, and the “permeate”, which is the filtrate portion that contains small molecules that can pass through the membrane. Once stabilized, aliquots of both the permeate and retentate streams were sampled over one complete cycle (equal to 6 hours) for characterization, and the re-captured retentate from the preceding cycle served as the feed for the subsequent ultrafiltration stage. Two additional permeate cycles were collected to calculate flux-decline curves based on the initial permeate flux measured during the first collection cycle.

Characterization studies

Pulp properties. The kappa, viscosity, and ISO brightness were determined according to SCAN-C 1 : 00, SCAN-CM 15 : 99, and SCAN-P 3 : 93, respectively. Prior to morphological analysis, pulp samples were dispersed in deionized water to achieve a highly diluted suspension of 0.01% (w/w). Paper sheets (optical properties) were prepared following SCAN-CM 11 : 95. Tensile measurements were conducted under conditions according to ISO 187:2022 including constant temperature (23 °C) and humidity (50%) control. Thickness was determined according to ISO 534. Tensile strength tests were performed according to ISO 1924-3 using Lorentzen and Wettre's horizontal tensile tester MTS 400 (MTS System Norden, Sweden) with load cells of 200 N.

All preparatory and analytical steps were conducted under controlled ambient conditions. The surface topography and structural integrity of the bagasse fibers were characterized using a scanning electron microscopy (FE-SEM, Zeiss Sigma VP, Germany). The samples were sputter-coated with platinum/palladium (Pt/Pd). Imaging was performed with an accelerating voltage between 5 and 10 kV. A system comprising a Tescan Mira 3 SEM with an Oxford Instruments Xplore30 EDS is employed for SEM-EDX analysis under an acceleration voltage of 5 kV and beam intensity of 13 (beam current 880 pA). Beam measurement was conducted with aluminum before the analysis; the quantification was performed with a factory standard that was supplied by Oxford Instruments. B-P1-P2-C paper sheet with a 20 nm-thick carbon layer was used for the analysis. X-ray diffraction (XRD) patterns of pulp were recorded using a Bruker D8 Advance powder diffractometer equipped with a Cu Kα radiation source (λ = 0.1542 nm), operated at 40 kV and 40 mA. Data were collected at 25 °C over a 2θ range of 10–60°, with a step size of 0.01° and a counting time of 1 s per step. X-ray photoelectron spectroscopy (XPS) was performed with a Kratos AXIS Ultra DLD X-ray photoelectron spectrometer using a monochromated Al Kα X-ray source (1486.7 eV) run at 100 W. A pass energy of 80 eV and a step size of 1.0 eV were used for the survey spectra, while a pass energy of 20 eV and a step size of 0.1 eV were used for the high-resolution spectra. Photoelectrons were collected at a 90° take-off angle under ultra-high vacuum, with a base pressure typically below 1 × 10⁻⁹ Torr. Both survey and high-resolution spectra were collected from three different spots on the sample surface to check for homogeneity and surface charge effects. High-resolution spectra have been charge-corrected with respect to the position of the C–O bonding in cellulose at 286.5 eV, corresponding to a C–C binding energy of 284.8 eV.

HexA content. Hexenuronic acid content was quantified by UV spectrophotometry following a selective acid hydrolysis protocol.¹⁵ Approximately 2.5 g of oven-dried pulp was suspended in 150 mL of 0.01 M sodium formate buffer (pH 3.5) within a sealed pressure-resistant glass vessel (Duran), which was purged with nitrogen gas immediately before closing the cap for hydrolysis. Hydrolysis was conducted in an autoclave (Systec DE 23, Germany) at a controlled temperature of 110 °C for 60 minutes. An ice bath was used for quenching the reaction. Under these acidic conditions, HexA groups are selectively cleaved from the xylan backbone and degraded primarily into 2-furoic acid and 5-carboxy-2-furaldehyde, the concentration of which was determined by measuring its absorbance at 245 nm. Upon completion, the reaction mixture was cooled to ambient temperature, and the liquid phase was separated from the fiber matrix by vacuum filtration through a Büchner funnel. The filtrate, combined with subsequent washings, was volumetrically adjusted to 1 L using deionized water. Spectral acquisition was performed using a Shimadzu UV-2550 spectrophotometer (Japan) across 200–500 nm. The specific absorbance was recorded at 245 nm (A₂₄₅), with a background correction applied at 480 nm (A₄₈₀) to eliminate the interference from non-specific chromophores. For samples exceeding an absorbance of 1.5, appropriate dilution was performed to maintain linearity. The final HexA content (c, in μmol g⁻¹) was calculated according to eqn (1), utilizing an extinction coefficient factor derived from the calibration of 2-furoic acid:where m represents the oven-dried mass of the pulp sample in grams.

Chemical composition. The carbohydrate composition was analyzed by high-performance anion-exchange chromatography with pulsed amperometric detection (HPAEC-PAD, Dionex ICS 5000) following two-step acid hydrolysis (NREL/TP-510-42618).¹⁶ Inorganic elements were quantified using inductively coupled plasma optical emission spectroscopy (ICP-OES, Agilent 5900 SVDV) after microwave digestion using HNO₃ and filtration through 0.45 µm membranes.

Residual alkali. The residual alkali concentration in the spent liquor from the soda cooking was quantified using a modular OMNIS professional titrator (Metrohm, Switzerland) equipped with an LL-Unitrode WOC pH electrode, in accordance with SCAN-N 33 : 94.

Separation efficiency. The fractionation efficiency of the ultrafiltration was assessed by determining the molecular weight distributions (MWD) of the retentate and permeate via gel permeation chromatography (GPC). Prior to injection, the liquid fractions were lyophilized and subsequently reconstituted in 0.1 M NaOH to achieve a final concentration of 1–2 mg mL⁻¹. Next, the solutions were subjected to mild sonication and passed through 0.45 µm PTFE syringe filters to ensure homogeneity and removal of particulates. Chromatographic separation was performed on an Agilent 1100 Series system (USA) tailored for alkaline lignin characterization. Data acquisition and M_w calculations were executed using the ChemStation GPC data analysis suite.

GC-MS. The qualitative and semi-quantitative profile of low-molecular-weight organic constituents in bagasse (B), pretreatment liquids (B-P1), soda liquor (B-P1-P2) and its ultrafiltration fractions (feed, permeate and retentate), along with the hydrolysates prepared for HexA determination were established using gas chromatography-mass spectrometry (GC-MS; Shimadzu QP2010SE, Japan). Prior to analysis, silylation was implemented to enhance the volatility of the analytes. Sample derivatization was performed by solubilizing aliquots in 500 µL of pyridine spiked with tetracosane (1 mg mL⁻¹) as the internal standard. This was followed by the addition of 300 µL of BSTFA. The reaction mixture was then subjected to two-phase thermal incubation at 70 °C (initially for 5 min, followed by 40 min), with intermittent vortex agitation being applied between each step. Chromatographic separation was carried out on an HP-5 capillary column (30 m × 0.25 mm (i.d.), 0.25 µm film thickness) via an Optic 4 injector operating in split mode (1 : 10) at 300 °C. The column temperature was programmed with the following gradient: an initial hold at 80 °C for 5 min, a ramp of 4 °C min⁻¹ to 300 °C and hold for 20 min. Quantitative values were calculated based on the peak area ratio relative to the C24 internal standard, with all measurements performed for a single time unless otherwise explained. The mass spectra were compared with those from authentic compounds [A], available databases (including Organic Acids Library (TMS-oxime) (asahikawa-med.ac.jp) [J], NIST Chemistry WebBook [N]), and literature data.^10,17–19

Acid-precipitated “lignin” and NMR analysis. Lignin (or acidic precipitate) fractions were recovered from the soda spent liquor (feed), retentate and permeate. 20% (w/w) sulfuric acid (H₂SO₄) was added dropwise to 50 mL of the liquid under constant magnetic stirring. The acidification process included an equilibration step (30 min stabilization) when the pH reached 5.0, and then the pH is subsequently reduced to 2.5 to ensure complete precipitation of “lignin”. The precipitated “lignin” was recovered by centrifugation at 10 000 rpm for 10 min. Next, the crude “lignin” pellet was resuspended and washed repeatedly with deionized water until the supernatant pH stabilized between 3.5 and 4.0. Finally, the purified lignin was subsequently freeze-dried and reconstituted in a deuterated solvent system comprising DMSO-d₆ and pyridine-d₅ (4 : 1, v/v) using a Bruker AV NEO 400 MHz spectrometer (Germany). For the heteronuclear single quantum coherence (HSQC), the spectral windows were defined as 10 ppm for the proton (¹H) dimension and 200 ppm for the carbon (¹³C) dimension. Data acquisition utilized a matrix of 1024 complex points in the direct dimension and 128 increments in the indirect dimension, with 64 scans accumulated per increment to optimize the signal-to-noise ratio. Chemical shifts were calibrated against the solvent peak of DMSO-d₆ (δH = 2.49 ppm; δC = 39.5 ppm). All spectra were processed using TopSpin 4.4.0 software and assigned by following the literature.^20,21

Results and discussion

Effects of pretreatment, pulp characteristics, and silica migration

To evaluate the efficiency of pretreatment in removing non-structural components from bagasse, the chemical profile of raw material extracts (DCM- and acetone-soluble) and the pretreatment (B-P1, Table S1) liquid were characterized using GC-MS (Fig. 2 and Table S2). Analysis indicates that the lipophilic components of bagasse primarily consist of three major classes: long-chain fatty acids, fatty alcohols, and sterols. As depicted in Fig. 2, multiple prominent characteristic peaks appeared in the chromatogram, corresponding to the high-abundance compounds that are summarized in Table S2. First, within the fatty acid fraction, palmitic acid (C16:0), oleic acid (C18:1), and stearic acid (C18:0) were identified as major constituents, consistent with previous findings.²² A fatty acid mixture derived from sugarcane wax oil exhibited anti-inflammatory activity in inflammatory models.²³ However, the long-chain fatty acids and fatty alcohol fraction of 1-octacosanol (C28:0) are not observed from the pretreatment liquor using water (B-P1) and mild alkali (B-P1-P2). The high presence of 1-octacosanol in the bagasse holds significant health benefits.^24,25 β-Sitosterol is another significant extractive present as an acetone-soluble fraction. Beyond β-sitosterol, campesterol and stigmasterol were identified in the acetone extract, a composition consistent with findings using supercritical CO₂.²²

Fig. 2 GC-MS total-ion chromatogram of bagasse showing major peaks along with the internal standard (C24, i.s.) for dichloromethane (D-B1) and acetone (A-B1) extractables at different retention times: (a) 10–42 min; (b) 42–62 min. More detailed assignments and codes are summarized in Table S2.

GC-MS spectra of the pretreatment samples revealed significant differences in chemical selectivity between B-P1 and B-P1-P2 (Fig. 3). B-P1 primarily removed water-soluble extracts from the fibrous bagasse. The dominance of glucose and disaccharide in the chromatogram (B-P1, Fig. 3) indicates that mild water washing served to wash out mainly free sugars. Citric acid,²⁶ lactic acid²⁷ and glutamic acid²⁸ are known organic acids that can be generated by microbial fermentation of bagasse. The identification of other amino acids (such as L-isoleucine and phenylalanine) is essential because amino acids act as precursors in the biological pathways of protein taking place during the storage of the bagasse. In contrast to B-P1, the B-P1-P2 solution exhibited different characteristics, most notably the substantial release of 4-coumaric acid and ferulic acid. This saponification phenomenon provides a molecular-level hypothesis for the mechanism of mild alkaline treatment, which possibly breaks the intermolecular ester bonds linking hemicelluloses and lignin.²⁹ Our data confirm that B-P1-P2 is sufficient to cleave these alkali-labile ester bonds, releasing bound phenolic acids that are possibly linked to the main structural lignin–carbohydrate matrix. Furthermore, the detection of HAs such as xyloisosaccharinic acid (XISA) and 3-deoxypentonic acid indicate partial alkali degradation (peeling reaction) of hemicelluloses.⁹

Fig. 3 GC-MS total-ion chromatogram of bagasse pretreatment liquids showing major peaks along with the internal standard (C24) at retention times of 5–52 min: B-P1 (water treatment); B-P1-P2 (alkali pretreatment). Color codes for the hydroxy acids and amino acids: blue (hydroxy acids); purple (amino acids); black (other acids). More detailed assignments, codes, and quantitative profiles are given in Table S2.

SEM was employed to visualize the structural evolution of bagasse fibers throughout the process. High-resolution imaging (Fig. 4a and b) revealed dense clusters of sub-micron-sized spherical particles and amorphous deposits adhering to the surfaces of the fibrous bagasse. These observations are consistent with the presence of pith residues and waxy layers commonly reported in raw bagasse.³⁰ These surface impurities may form a recalcitrant physical barrier that can hinder the penetration of cooking chemicals. As depicted in Fig. 4(c and d), while the fiber bundles remained structurally intact after removal of the spherical “granule”, the surface appeared much cleaner and polished. The bundles (Fig. 4e and f) were successfully delignified into fibers. The resulting pulp exhibited a characteristic flat, ribbon-like morphology with a collapsed lumen, typical of chemically delignified non-wood fibers. The FS5 analysis (Table S3) yielded a fiber length of 1.66 mm, a width of 21.0 µm, and a curl index of 7.9%. These values are consistent with those reported previously,¹ indicating that the soda cooking effectively separated the fibers while preserving their natural length.

Fig. 4 SEM imaging at increasing magnification of fibrous bagasse B (a and b); hot water-washed bagasse (B-P1) (c and d); cooked pulp (B-P1-P2-C) (e and f). More detailed SEM images are summarized in Fig. S1–S3.

Although alkaline pretreatment using 0.1 M NaOH was originally designed to liberate amorphous phytoliths into soluble sodium silicate (Na₂SiO₃) for desilification,³¹ surprisingly, even the relative silicon content fluctuated from fibrous bagasse to the pulp (B-P1-P2-C) after the mild alkali treatment as shown at Fig. 5a. The hypothesis about silicon migration appears speculative and is supported by direct surface or bulk characterization using multiple analytical techniques, including SEM-EDX, XRD, and XPS (Fig. 5b–d and Fig. S4–S6). XRD analysis reveals a distinct diffraction hump at ∼29.42° attributed to poor crystallinity of calcium silicate.³² A diffraction peak at ∼26.73° (2θ) is more commonly attributed to crystalline inorganic phases of the acetone-soluble extracts out of ground-form bagasse pulp. The enhancement of cellulose intensity at B-P1-P2-C (Fig. S6b) is attributed to removal of amorphous acetone-soluble extractive components. Calcium silicate, as the primary form of silica, is also justified from the top 2–3 nm of B-P1-P2-C (ground form, post-extracted) using XPS. A single component of Si was found in the Si 2p region, with the 2p_3/2 peak located at approximately 102.5 eV, which is a typical binding energy for silicates. Likewise, a single component was found in the Ca 2p region with the 2p_3/2 peak located at approximately 347.4 eV. This is most likely also related to calcium in a silicate. There are almost equal amounts of Ca (0.27%) and Si (0.34%) in the sample, while only trace amounts of sodium (0.01%) could be detected (binding energy approximately 1071.5 eV). Calcium silicate (CaSiO₃) is found from one spot out of the SEM-EDX analysis of the B-P1-P2-C pulp sheet along with other multiple forms including silicon dioxide (SiO₂), aluminosilicate, and magnesium silicates. To conclude, our hypothesis is that highly reactive Ca²⁺ rapidly displaces Na⁺ under the cooking conditions and reacts in situ with SiO₃²⁻, explaining the unexpected surge in calcium content at B-P1-P2-C in Fig. 5a, while the dissolved silica (out of B-P1-P2) was not leached away but instead “captured” first by sodium at B-P1-P2 before the sodium–calcium exchange. The calcium silicate (CaSiO₃) is then reprecipitated back to the fiber surface of B-P1-P2-C potentially limiting the delignification rate of the soda cooking.

Fig. 5 Inorganic compositional analysis by multiple techniques. (a) ICP-OES analysis of multiple samples (fibrous bagasse (B); water-pretreated bagasse (B-P1); alkali-washed bagasse (B-P1-P2); soda-cooked pulp (B-P1-P2-C)). (b) SEM-EDX analysis of B-P1-P2-C paper sheet (Fig. S12) based on six spectrum positions (Fig. S4) in comparison with the reference spot (pulp cellulose fiber). (c) XRD patterns of B-P1-P2-C in comparison with its post-extracted form (B-P1-P2-C (post-extracted)). (d) XPS analysis of B-P1-P2-C (post-extracted) showing relative concentrations of the elements (STD is included as part of the parentheses) averaged from all three positions on the sample surface. More detailed results are summarized in Fig. S4–S6.

A-stage on hexenuronic acid removal

The standard UV protocol¹⁵ was originally designed for hardwood kraft pulp, which typically contain high HexA levels (often >40–60 meq kg⁻¹). In such samples, the HexA-derived signal is strong, rendering the background interference negligible. However, the bagasse in this study has a relatively low initial HexA content (14.4 meq kg⁻¹), and the unbleached pulp spectrum displays a broad background absorption signal instead of a sharp peak at 245 nm, characteristics of the absorption maximum for 2-furoic acid (Fig. S7). However, the baseline is not flat and there is no clear absorption band associated with 2-furonic acid at A245 (absorbance at 245 nm) for the hydrolysate of both D₀ pulp and AD₀ pulp. GC-MS (Table S4 and Fig. S8, S9) is thus applied for both qualitative and semi-quantitative analysis of 2-furonic acid from the same hydrolysate prepared for the UV protocol. The detected 2-furonic acid content from the hydrolysate of unbleached pulp (0.59 mg g⁻¹) is roughly 5–10 times higher than that from the D₀ pulp (0.05 mg g⁻¹) and A/D₀ pulp (0.14 mg g⁻¹), respectively. Consequently, applying this wood-based method to bagasse likely overestimates the real HexA content to some extent; however, this potential limitation on calculation of HexA removal rate has been limited because residual HexA (or 2-furonic acid) content is extremely low regardless of the methods. HexA content calculations using the standard UV protocol will be applied for the follow-up discussion.

Normally mass loss associated with the A-stage can be primarily attributed to the selective degradation of hemicelluloses and HexA removal, as evidenced by the relatively lower content of xylose in A/D₀ (Fig. 6 and Table S5) in comparison to D₀. However, the A-stage brings negligible yield loss from A/D₀ to D₀, attributed possibly to the relatively higher kappa for the A/D₀ pulp, where 14.4 meq kg⁻¹ HexA from bagasse is consistent with 16.5 meq kg⁻¹ that was reported previously. Following A/D₀ sequential treatment,² HexA content significantly decreased to 7.4 meq kg⁻¹ (Table 2), representing removal of almost half the HexA. In contrast, the control D₀ retained a higher HexA residue of 8.8 meq kg⁻¹. This difference reveals the impact when including the A-stage (pH 3.5 and 95 °C), whereupon the glycosidic bonds linking HexA to the xylan backbone undergo selective acid hydrolysis converting HexA into furan derivatives (e.g., 2-furanoic acid), which are then discharged with the wash water, thereby preventing its entry into subsequent oxidation stages and further impacting brightness reversion and ClO₂ consumption.⁷ The A/D₀ sequential process, benefiting from the prior removal of HexA, achieved 59% lignin removal (kappa reduced from 11.7 to 4.8) with only 2.0% ClO₂ dosage, equivalent to a 63% delignification rate (kappa from 8.7 to 3.1) with 2.5% ClO₂ dosage based on the D-hot process reported previously.²

Fig. 6 Overall chemical composition (a) and carbohydrate composition (b) (% of anhydrosugars in the original dry mass without yield consideration) from the fibrous bagasse (B, Table S1) to the bleached pulp. The relative composition is displayed in Table S5. Abbreviations: arabinose (Ara), rhamnose (Rha), galactose (Gal), glucose (Glc), xylose (Xyl), mannose (Man).

Table 2 Properties of pulps of bagasse treated with A/D₀-E_p-D₁ (4% Act. Cl) in comparison with D₀-E_p-D₁ (5.2% Act. Cl): kappa; viscosity; ISO brightness; HexA content. Photographs of the pulp are included in Fig. S10. STD is included in the parentheses and error bars are displayed in Fig. S11. n.a. refers to not applicable

	Kappa		Viscosity (mL g^−1)		Brightness (ISO %)		HexA (c, meq kg^−1)		Yields (%)
	A/D0EpD1	D0EpD1	A/D0EpD1	D0EpD1	A/D0EpD1	D0EpD1	A/D0EpD1	D0EpD1	A/D0EpD1	D0EpD1
Unbleached	11.7 (0.005)		1029 (1)		39 (0.4)		14.4 (0.03)		45.7
A/D0 (D0)	4.8 (0.001)	4.4 (0.01)	949 (5)	1012 (3)	47 (1.4)	56 (0.4)	7.4 (0.1)	8.8 (0.1)	44.7	44.3
Ep	3.1 (0.02)	2.8 (0.01)	734 (10)	918 (9)	60 (0.6)	69 (0.3)	n.a.	n.a.	n.a.	n.a.
D1	0.5	0.5	685 (3)	874 (3)	75 (0.7)	80 (0.3)	n.a.	n.a.	n.a.	n.a.

---

Due to the effective HexA reduction when the A-stage was implemented, A/D₀-E_p-D₁ achieves a 25% reduction in total active chlorine usage from 5.2% to 4.0% despite a compromise of losing 21% viscosity (Table 2). In specific terms, the viscosity reduction of 7.8% in this A/D₀ process (1029 mL g⁻¹ to 949 mL g⁻¹) is comparable to the ∼8.4% decrease (1095 mL g⁻¹ to 1003 mL g⁻¹) in the D-hot process reported previously.² Raw bagasse exhibits typical non-wood biomass characteristics, with a lignin content of approximately 20–25%, consistent with the composition reported earlier for bagasse (∼45% cellulose, ∼28% hemicelluloses, ∼22% lignin).¹ The detailed sugar profile shown in Fig. 6 reveals the specific impact of the A-stage on hemicellulose components (primarily xylan). A slight decrease in xylose content was observed from 24.2% (D₀-E_p-D₁) to 23.9% (A/D₀-E_p-D₁). Acidic conditions (pH 3.5, 95 °C) inevitably induce partial hydrolysis of short-chain amorphous xylans while attacking HexA glycosidic bonds.^4,7 Overall, moderate removal of xylan side chains not only eliminates roughly 50% of HexA but also reduces the risks behind AOX generation and brightness reversion during subsequent bleaching. Thus, the slight decrease of xylose from D₀-E_p-D₁ pulp to A/D₀-E_p-D₁ pulp (Table S5) aligns with the effect originating from the A-stage.

Fiber viscosity, as an indicator of cellulose molecular weight, influences primarily fiber strength; however, the relationship to paper strength is indirect and tensile strength properties are moderated also by fiber morphology (coarseness, aspect ratio) and inter-fiber bonding (responsible for adhering the fibers together in the paper matrix) as failure typically occurs at fiber–fiber interfaces rather than within the fibers themselves. The effect of pulp viscosity on tensile strength properties of bagasse pulp fiber is assessed for pulps having lower viscosity (A/D₀ 949 mL g⁻¹ and A/D₀E_pD₁ 685 mL g⁻¹, Table 2) in comparison with pulps without A-stage treatment (D₀ 1012 mL g⁻¹ and D₀E_pD₁ 874 mL g⁻¹, Table 2), respectively. Roughly 5% (28.88 to 27.56) and 12.6% (30.4 to 26.6) loss of tensile index was seen in relation to 6.2% and 21.6% loss of viscosity from D₀– to AD₀ pulp and D₀E_pD₁– to A/D₀E_pD₁ pulp, respectively (Table S6 and Fig. S12–S14). No visible differences were seen from the Young modulus (Table S6). A moderate level loss of 21.6% in viscosity results in proportional but not linear reductions in tensile strength of their respective sheets. This observation is also supported by the reference statement about cellulose DP loss negatively affecting the paper strength only when reaching the threshold level for sulfate pulp of softwood (15 to 20 mPa s (or 800 to 915 mL g⁻¹ intrinsic viscosity))³³ and bamboo (8 mPa s).³⁴ In this scenario, bagasse fiber exhibits its most extensive application in the manufacturing of molded pulp food-service packaging and disposable tableware. In molded pulp food-service packaging, bagasse fibers do not require high tensile strength, as the primary mechanical demands involve compression, bending, and wet stability rather than tensile loading.

Ultrafiltration

The two-dimensional HSQC NMR spectrum (Fig. 7) reveals the impact of ultrafiltration on the chemical structure of lignin at the atomic level from retentate and permeate in comparison with the feed. First, the behavior of the dominant inter-unit linkages, particularly β-O-4 aryl ethers (substructure A), offers a direct insight into the depolymerization extent.²¹ In the spectrum of both retentate and feed, the signals for the Cα–Hα (Aα), Cβ–Hβ (Aβ), and Cγ–Hγ (Aγ) correlations in β-O-4 linkages are seen at δ_C/δ_H of 71.52/5.03, 86.28/4.15, and 61.69/3.58 ppm, respectively. Furthermore, xylan (Xyl) with Xyl-1, Xyl-2, Xyl-3, Xyl-4, and Xyl-5 (in red, Fig. 7) correlations at δC/δH of 101.6/4.38, 72.5/3.21, 73.8/3.41, 75.2/3.65 and 63.2/3.98 (3.27) ppm, respectively, were present both in feed and retentate but not in the permeate (Fig. 7c). Given that the ultrafiltration membrane used in this study has an MWCO of only 0.5 kDa, free monosaccharide xylose should theoretically readily penetrate through the membrane pores into the permeate. However, xylan was retained in high proportions within the retentate, strongly indicating that these carbohydrates do not exist in free form in feed and retentate as well as permeate (Fig. 9). Instead, they are possibly covalently bound to lignin, forming stable lignin–xylan complexes.^35,36

Fig. 7 Both aromatic and side chain (δ_C/δ_H, 10–145/0.5–8.5 ppm) regions of the ¹H−¹³C HSQC NMR spectrum of acid precipitates “lignin” in 0.8 mL of DMSO-d₆/pyridine-d₅ (v/v, 4/1) from multiple sources: (a) feed (51.03 mg); (b) permeate (5.86 mg); (c) retentate (5.29 mg).

The efficacy of the ultrafiltration in fractionating high MW lignin–xylan complexes based on hydrodynamic volume was confirmed by GPC (Fig. 8 and Table S7). The parent spent liquor (feed) exhibited a broad MW distribution (weight-average molecular weight (M_w) ≈ 2776 g mol⁻¹, number-average molecular weight (M_n) ≈ 641 g mol⁻¹), typical of soda lignin containing a heterogeneous mixture of degraded fragments. The retentate fraction was significantly enriched in high MW species, with its M_w increasing to 4032 g mol⁻¹, representing an ∼45% increase compared to feed. Conversely, the permeate contained only oligomeric fragments with a much lower M_w of 840 g mol⁻¹. This stark contrast (4.8 times) proved again the 0.5 kDa-sized membrane effectively retained the macromolecular lignin–xylan complex in the retentate while allowing smaller phenolic oligomers (sugars, HAs, and alkali) to pass through the membrane as part of the permeate. 1.97 g NaOH L⁻¹ of residual alkali was detected in the spent liquor, suggesting that the free-form alkali has been mostly consumed during the cooking, the significant abundance of Na⁺ (possibly primarily Na₂CO₃) in the feed and retentate (Fig. S15) implying that alkali recovery can be positively achieved through multiple cycles of ultrafiltration prior to the causticizing step for alkali recovery (Fig. 1b). Permeate (water flux) filtration dropped progressively (by roughly 29%) from 0.42 L m⁻² h⁻¹ (cycle 1, Fig. S16) to 0.36 L m⁻² h⁻¹ (cycle 2) and subsequently down to 0.30 L m⁻² h⁻¹ (cycle 3) indicating that a consistent rate of membrane fouling occurred during each operating cycle.

Fig. 8 The elution profile of different lignin sources derived from the ultrafiltration (i.e. black liquor; permeate; and retentate) from MALLS-SEC. More detailed information is summarized in Table S7.

GC-MS characterization of low-MW components in the multiple sources of liquid revealed a complex profile of aliphatic HAs. As shown in Fig. 9, significant HA signals were detected in both the feed and permeate, with their structures further confirmed by mass spectrometry fragmentation patterns (Fig. S17). Among the detected compounds, lactic acid, glycolic acid, and 2-hydroxybutanoic acid (2-HBA) were identified as the major aliphatic acid components (Table S8). The formation of these HAs is attributed to the “peeling reaction” of polysaccharides during alkaline pulping. The reducing ends of cellulose and hemicelluloses are highly unstable under thermal–alkali conditions, undergoing stepwise degradation and rearrangement to form various HAs.⁹ Notably, xyloisosaccharinic acid (XISA) (rt 26.37 min, Fig. 9), a major degradation product formed over the alkaline hydrolysis of hemicelluloses, is also identified.¹⁰ Semi-quantitative analysis (Table 3) indicates that an overall recovery rate of 72% HAs is achieved in the permeate, for example lactic acid and 2-HBA in the permeate (0.18 g L⁻¹ and 0.17 g L⁻¹, respectively) were significantly higher than those in the retentate (<0.04 g L⁻¹). Similar observations have demonstrated ultrafiltration as a key step in black liquor fractionation during the chemical pulping process of softwood, retaining over 90% of lignin in the retentate to yield a pure permeate stream rich in HAs and alkali.^12,37

Fig. 9 GC-MS total-ion chromatogram of multiple sources of spent liquor after ultrafiltration showing major peaks along with the internal standard (C24, Std) at retention times of 7–33 min: feed; permeate; retentate. Color code for the hydroxy acids is blue. More detailed assignments and number codes are given in Table S8. Mass spectra of some classic hydroxy acids are summarized in Fig. S17.

Table 3 Overall content (in g L⁻¹) of the hydroxy acids in the spent liquor (feed), permeate and retentate. The complete profile and codes of the assignment are summarized in Table S8

Code	Ret time, mins	Compounds	Feed	Permeate	Retentate
1	7.167	Lactic acid	0.275	0.180	0.035
2	7.633	Glycolic acid	0.145	0.110	0.016
4	9.482	2-Hydroxybutanoic acid	0.245	0.167	0.030
5	9.676	Oxalic acid	0.040	0.017	0.003
8	13.255	4-Hydroxybutanoic acid	0.012	0.010	0
11	15.965	Succinic acid	0.013	0.008	0
12	16.402	Methylsuccinic acid	0.004	0.002	0
18	19.529	2,4-Dihydroxybutanoic acid	0.021	0.017	0.001
21	22.526	2,5-Dihydroxy pentanoic acid (2,5-DHPA)	0.096	0.080	0.010
23	24.514	2-Hydroxyglutaric acid	0.012	0.005	0
25	26.372	Xyloisosaccharinic acid (XISA)	0.105	0.084	0.010
26	26.867	3-Deoxy-erythro-pentonic acid	0.015	0.013	0.002
27	27.252	2-Hydroxyadipic acid	0.005	0.002	0
28	27.353	3-Deoxy-threo-pentonic acid	0.049	0.036	0.004
32	32.665	Glucoisosaccharinic acid (GISA)	0.134	0.102	0.011
33	32.706	3-Deoxy-arabino-hexonic acid	0.025	0.019	0
32	32.940	Glucoisosaccharinic acid (GISA)	0.061	0.048	0.005
35	33.291	3-Deoxy-2-hydroxymethyl-erythro-pentaric acid	0.006	0.003	0
37	33.858	3-Deoxy-2-hydroxymethyl-threo-pentaric acid	0.002	0.002	0
	Overall		1.266	0.907	0.127
	HA recovery%			72	10

---

For the first time, implementing ultrafiltration during the bagasse recovery process offers the promise of multiple recovery of sugar-degradation HAs along with alkali and lignin–xylan complexes from spent liquor (Fig. 1b). Moreover, the majority of sulfur compounds should already have been removed at the cooking plant via non-condensable gases (NCGs). The only sulfur source that enters the fiberline should be the one originating from raw biomass; thus, the produced spent liquor will not contain any sulfur compounds. So, the role of reduction reactions converting sodium sulphate (Na₂SO₄) into sodium sulphide (Na₂S) is no longer required for the traditional recovery boiler. Necessary steam (energy) required for pulping can be generated using a traditional power boiler instead of the specialized Tomlinson recovery boiler. A much smaller and financially more affordable alternative solution like a simplified membrane and power boiler unit has, for the first time, been demonstrated for closing the alkali recovery cycle and meeting the energy demand of bagasse pulp mill to partially replace the costly Tomlinson recovery boiler.

Conclusions

This study demonstrated an integrated biorefinery strategy for bagasse valorization, addressing the dual challenges of HexA interference prior to bleaching and the inefficient recovery of silica-rich spent liquor. The A-stage proved, for the first time, effective in selectively removing half of the HexA content prior to bleaching of bagasse pulp. This targeted removal allowed for a 25% reduction in ClO₂ consumption within the ECF bleaching sequence along with a trade-off compromise of 21.6% viscosity reduction and 12.6% loss of tensile index of the sheet compared to the reference sequence, possibly attributed to the degradation of hemicelluloses. Ultrafiltration utilizing a 0.5 kDa-sized membrane concentrated high-purity lignin–xylan complexes in the retentate while enriching the permeate with 72% low-molecular-weight HAs, such as xyloisosaccharinic acid (XISA) and glucoisosaccharinic acid (GISA). The strong presence of calcium ions detected in bagasse pulp leads to the speculative hypothesis that highly active Ca²⁺ rapidly replaces Na⁺ completely (or partially) and reacts in situ with SiO₃²⁻ under soda cooking conditions. The silica, initially dissolved during the alkali pretreatment stage, is not leached out but is first “captured” by the sodium prior to the sodium-calcium exchange. Subsequently, calcium silicate is reprecipitated back to the surface of the fibers after cooking, highlighting the challenges behind silica control in the bagasse fiberline.

Future research avenues shall first establish a HexA detection method tailored for non-wood and prioritize optimizing the A-stage parameters, specifically tailoring acid charge and temperature, to achieve a better balance between bleaching chemical savings and pulp quality preservation. Additionally, downstream efforts should focus on fine-tuning membrane ultrafiltration to minimize fouling of membranes without compromising the flux rate (Fig. S16), while multiple cut-off M_w membranes will also be investigated to improve the separation efficiency and shorten the filtration time. Mass balance of the HAs and alkali recovery (from permeate) along with recovery and application of lignin–xylan complexes (from retentate) will be more carefully investigated for establishing thermodynamic mass and energy balances³⁸ when in combined use with the traditional power boiler, thereby justifying the economic potential of partially replacing the costly Tomlinson recovery boiler (Fig. 1b).

Author contributions

Kun Ouyang: investigation, data curation, validation, formal analysis, writing – original draft. Wajeeha Munib: investigation (ultrafiltration); Huy Quang Lê: methodology, investigation (bleaching). Lauri Leskinen: formal analysis (GC-MS); Yue Wu: formal analysis (NMR); Naveen Kumar Chenna: conceptualization, resources, and funding acquisition. Inge Schlapp-Hackl: formal analysis (ICP-OES); Jinze Dou: conceptualization, methodology, formal analysis, visualization, resources, supervision, project administration, funding acquisition, writing – review & editing.

Conflicts of interest

There are no conflicts to declare.

Data availability

The authors confirm that all the data that support the findings of this study are available within the article and its supplementary information (SI) or from the corresponding author upon reasonable request.

Supplementary information is available. See DOI: https://doi.org/10.1039/d6gc01257d.

Acknowledgements

The authors gratefully acknowledge the financial support provided by the Research Council of Finland under the project FACE+ (Grant No. 356520) as well as a donation grant from Andritz Oy to Aalto University Foundation. Additionally, the authors acknowledge the use of the research facilities in Bio 2 at Aalto University (Bioeconomy Infrastructure) for the experimental work. Besides, the authors express their gratitude to Prof. Tapani Vuorinen from Aalto University for his valuable inputs regarding the bleaching experiments and to Guilhermo Magalhaes from Andritz Oy for his conceptual contribution to the membrane fractionation approach. We also extend our sincere thanks to Dr Estefanía Isaza Ferro for her technical guidance and training regarding the ECF bleaching protocols. Furthermore, the invaluable expertise and assistance provided by Dr Timo Pääkkönen concerning the operation of the Alfa Laval ultrafiltration system are highly appreciated. In addition, the assistance provided by Dr Leena Pitkänen from Aalto University with GPC testing was invaluable. The authors also express their appreciation to Prof. Lei Wang (Westlake University, China) and Prof. Xueping Song (Guangxi University, China) for coordinating the delivery of bagasse. Dr Eero Hiltunen, Dr Lassi Klemettinen, and Dr Tewari Girish from Aalto University are acknowledged for their support in tensile testing, SEM-EDX, and XRD when preparing for the reviewer's response.

References

1. M. S. Jahan, M. M. Rahman and Y. Ni, Biofuels, Bioprod. Biorefin., 2021, 15, 100–118,  DOI:10.1002/bbb.2143.
2. H. Zhang, S. Nie, C. Qin and S. Wang, Ind. Crops Prod., 2019, 128, 338–345,  DOI:10.1016/j.indcrop.2018.11.025.
3. D. Nguyen, X. Zhang, Z. H. Jiang, A. Audet, M. G. Paice, S. Renaud and A. Tsang, Enzyme Microb. Technol., 2008, 43, 130–136,  DOI:10.1016/j.enzmictec.2007.11.012.
4. A. A. Shatalov and H. Pereira, Bioresour. Technol., 2009, 100, 3069–3075,  DOI:10.1016/j.biortech.2009.01.020.
5. N. Mahadevan, N. C. Sridhar, P. N. Rao, K. Kalyanasundaram and A. R. K. Rao, IPPTA, 1986, 23, 155–159.
6. A. Teleman, V. Harjunpää, M. Tenkanen, J. Buchert, T. Hausalo, T. Drakenberg and T. Vuorinen, Carbohydr. Res., 1995, 272, 55–71,  DOI:10.1016/0008-6215(95)96873-M.
7. T. Vuorinen, A. Teleman, P. Fagerström, J. Buchert and M. Tenkanen, J. Pulp Pap. Sci., 1999, 25, 155–162.
8. S. Nie, S. Wang, C. Qin, S. Yao, J. F. Ebonka, X. Song and K. Li, Bioresour. Technol., 2015, 196, 413–417,  DOI:10.1016/j.biortech.2015.07.115.
9. P. Fardim, Papermaking Science and Technology: Volume 6 Chemical pulping, Part 1: Fibre chemistry and technology, Fapet Oy, Helsinki, 2011, pp. 216–217.
10. K. Niemelä and R. Alén, Analytical Methods in Wood Chemistry, Pulping, and Papermaking, Springer-Verlag, Heidelberg, 1999, pp. 193–231.
11. O. A. Ojelade, Q. Fu, S. Nair and C. W. Jones, ACS Sustainable Chem. Eng., 2024, 12, 9054–9066,  DOI:10.1021/acssuschemeng.4c00212.
12. J. Heinonen and T. Sainio, Chem. Eng. Sci., 2019, 197, 87–97,  DOI:10.1016/j.ces.2018.12.013.
13. A. Toledano, A. García, I. Mondragon and J. Labidi, Sep. Purif. Technol., 2010, 71, 38–43,  DOI:10.1016/j.seppur.2009.10.024.
14. C. A. E. Costa, P. C. R. Pinto and A. E. Rodrigues, Sep. Purif. Technol., 2018, 192, 140–151,  DOI:10.1016/j.seppur.2017.09.066.
15. M. Tenkanen, G. Gellerstedt, T. Vuorinen, A. Teleman, M. Perttula, J. Li and J. Buchert, J. Pulp Pap. Sci., 1999, 25, 306–311.
16. J. B. Sluiter, R. O. Ruiz, C. J. Scarlata, A. D. Sluiter and D. W. Templeton, J. Agric. Food Chem., 2010, 58, 9043–9053,  DOI:10.1021/jf1008023.
17. J. Dou, K. Niemelä, T. Haatainen, P. Tervola and J. Vehmaa, J. Cleaner Prod., 2023, 405, 136940,  DOI:10.1016/j.jclepro.2023.136940.
18. G. Petersson, Tetrahedron, 1970, 26, 3413–3428,  DOI:10.1016/S0040-4020(01)92918-7.
19. G. Petersson, Carbohydr. Res., 1975, 43, 1–8,  DOI:10.1016/S0008-6215(00)83967-1.
20. J. Dou, H. Kim, Y. Li, D. Padmakshan, F. Yue, J. Ralph and T. Vuorinen, J. Agric. Food Chem., 2018, 66, 7294–7300,  DOI:10.1021/acs.jafc.8b02014.
21. J. C. del Río, A. G. Lino, J. L. Colodette, C. F. Lima, A. Gutiérrez, Á. T. Martínez, F. Lu, J. Ralph and J. Rencoret, Biomass Bioenergy, 2015, 81, 322–338,  DOI:10.1016/j.biombioe.2015.07.006.
22. T. M. Attard, C. R. McElroy, C. A. Rezende, I. Polikarpov, J. H. Clark and A. J. Hunt, Ind. Crops Prod., 2015, 76, 95–103,  DOI:10.1016/j.indcrop.2015.05.077.
23. N. Ledón, A. Casacó, D. Remirez, A. González, J. Cruz, R. González, A. Capote, Z. Tolón, E. Rojas, V. J. Rodríguez, N. Merino, S. Rodríguez, O. Ancheta and M. C. Cano, Phytomedicine, 2007, 14, 690–695,  DOI:10.1016/j.phymed.2006.12.019.
24. S. Feng, Z. Luo, F. Zeng, S. Liu and Z. U. Khan, Food Chem., 2015, 182, 171–177,  DOI:10.1016/j.foodchem.2015.03.003.
25. J. C. Taylor, L. Rapport and G. B. Lockwood, Nutrition, 2003, 19, 192–195,  DOI:10.1016/S0899-9007(02)00869-9.
26. D. Kumar, V. K. Jain, G. Shanker and A. Srivastava, Process Biochem., 2003, 38, 1731–1738,  DOI:10.1016/S0032-9592(02)00252-2.
27. M. C. A. Xavier, G. S. Dias, S. Hernalsteens and T. T. Franco, ACS Omega, 2025, 10, 51891–51899,  DOI:10.1021/acsomega.5c08221.
28. X. Feng, B. Tang, Y. Jiang, Z. Xu, P. Lei, J. Liang and H. Xu, J. Chem. Technol. Biotechnol., 2016, 91, 2085–2093,  DOI:10.1002/jctb.4805.
29. S. G. Karp, A. L. Woiciechowski, V. T. Soccol and C. R. Soccol, Braz. Arch. Biol. Technol., 2013, 56, 679–689,  DOI:10.1590/S1516-89132013000400019.
30. T. A. Chimenez, M. H. Gehlen, K. Marabezi and A. A. S. Curvelo, Cellulose, 2014, 21, 653–664,  DOI:10.1007/s10570-013-0135-9.
31. H. K. Gupta, P. Shivhare, T. K. Roy and N. J. Rao, IPPTA, 1997, 9, 29–35.
32. J. M. Wolter, K. Schmeide, N. Huittinen and T. Stumpf, Sci. Rep., 2019, 9, 14255,  DOI:10.1038/s41598-019-50402-x.
33. B. N. Brogdon and L. A. Lucia, Tappi J., 2023, 22, 10,  DOI:10.32964/TJ22.10.631.
34. S. Sah, S. Maheshwari and A. Y. Kulkarni, IPPTA, 1982, 19, 41–44.
35. X. Kang, A. Kirui, M. C. D. Widanage, F. Mentink-Vigier, D. J. Cosgrove and T. Wang, Nat. Commun., 2019, 10, 347,  DOI:10.1038/s41467-018-08252-0.
36. O. M. Terrett and P. Dupree, Curr. Opin. Biotechnol., 2019, 56, 97–104,  DOI:10.1016/j.copbio.2018.10.010.
37. S. Hellstén, J. Lahti, J. Heinonen, M. Kallioinen, M. Mänttäri and T. Sainio, Chem. Eng. Res. Des., 2013, 91, 2765–2774,  DOI:10.1016/j.cherd.2013.06.001.
38. T. Anttila, master thesis, Review of different kraft lignin separation technologies and comparing their effects on a kraft pulp mill, Aalto University, 2021. https://urn.fi/URN:NBN:fi:aalto-202109058984.

---