The effect of pretreatment on methanesulfonic acid-catalyzed hydrolysis of bagasse to levulinic acid, formic acid, and furfural

Abstract

A major challenge that must be overcome for the commercial production of levulinic acid from lignocellulosics is to reduce equipment blockage and corrosion. Methanesulfonic acid (MSA), a relatively low corrosive acid, was used to produce organic acids and furfural from pretreated sugarcane bagasse. In general, the type of pretreatment did not affect levulinic acid yield, though it affected furfural yield. However, soda pretreated bagasse produced the highest yields of levulinic acid (∼75 mol%) and furfural (∼85 mol%), albeit under optimized conditions. Hydrolysis residue consists primarily of lignin that has been modified and/or condensed to humic substances, fatty acids, and oligomeric sugars. A conceptual biorefinery utilizing 1 ton of dry bagasse, alkaline-pretreatment, and MSA as a catalyst produced 165 kg soda lignin, 190 kg and 89 kg of levulinic acid and formic acid respectively, and 40 kg furfural.

---

1. Introduction

Sugarcane bagasse is the fibrous residue following the extraction of sugar juice in the sugar manufacturing process and accounts for 25–30% of the incoming cane. Bagasse is mainly used as fuel for combustion in boilers to generate process steam and electricity for the factory.¹ Surplus bagasse not used for the energy requirement of the factory finds use in co-generation of additional electricity, and options for its potential use for the production of biofuels and commodity chemicals are currently being investigated around the world. Levulinic acid and furfural are versatile platform chemicals that can be utilized to produce fuels, solvents, polymers, pharmaceutical, and agrochemical products. The current world market for levulinic acid is <5000 t, but demand would increase significantly if a cheap production source with low waste can be developed.² Presently, acid hydrolysis of lignocellulosic (plant) biomass offers the best opportunity to produce these platform chemicals from renewable resources, despite recent advances made on the use of heterogeneous catalysts³ and the development of alternative reaction pathways including the use of furfural alcohol as the starting feedstock.⁴

Conversion of sugarcane bagasse into levulinic acid (63 mol%) has been demonstrated using H₂SO₄ as a catalyst at 150 °C, but required a relatively long reaction time of 6 h.⁵ Increasing the hydrolysis temperature to 225 °C in the presence of HCl improved levulinic acid yield to 82.7 mol% in 45 min.⁶ The semi-commercial Biofine process⁷ produced both not only levulinic acid but also furfural from a range of lignocellulosic feedstocks using a two-stage reaction process. The process is carried out in two continuous reactors catalyzed with H₂SO₄ (1–5 wt%). Fast heating (20–60 s) and higher temperatures (210–230 °C) are used to convert the biomass into 5-hydroxymethyl furfural (HMF), which is then transferred to the second reactor that is operated at 195–215 °C for 15–30 min for the formation of levulinic and formic acids. Yields of >70 mol% were obtained for both levulinic acid and furfural. GF Biochemicals started manufacturing levulinic acid from lignocellulosic feedstocks on a commercial scale using the technology in 2015 (2000 t p.a. in 2015 but planned expansion to 10 000 t p.a. by 2017).⁸ The lack of uptake of these technologies is because of processing issues associated with the use of lignocellulosics and mineral acids. These issues include the blockage of the reactors by lignin and lignin derivatives,⁹ excessive corrosion of the process equipment associated with the use of mineral acids (e.g. H₂SO₄), inefficient product separation and recovery, and the lack of waste disposal strategies.

Our research group is currently involved in a comprehensive research program to address these challenges and accelerate commercialization of a biorefinery to produce not only levulinic acid, but also formic acid and furfural. Although, acid catalysts are commonly utilized in industrial bio-refining facilities, the facilities require special designs and materials to mitigate the costly impacts of corrosion.^8,10 In the first step of our investigation, sulfonic acids (including methanesulfonic acid, MSA) were examined in place of H₂SO₄ for the production of levulinic acid, formic acid and furfural from mixtures of simple monomeric sugars. They were selected because of lower corrosion rates than mineral acids. For example, the corrosion rate of H₂SO₄ on mild steel is ∼40 mm per year at 90 °C, while it is negligible under the same conditions with MSA.¹¹ The study demonstrated that MSA can replace H₂SO₄ matching it in selectivity and product yield.¹² Importantly, MSA is biodegradable, and would therefore cause less environmental problems compared to the mineral acids.¹¹

The blockage and attachment of sticky materials on process equipment are major problems in conversion of lignocellulosics to levulinic acid. This is associated with the lignin component of the biomass. So, it was decided to evaluate the effect of biomass pretreatment on the levulinic acid, formic acid and furfural yields. As the type of pretreatment affects lignocellulose composition and architecture, the pulps derived from the process will likely influence product yield and processability during hydrolysis. In this project, the solvents used to pretreat sugarcane bagasse were NaOH, H₂SO₄, acidified 1-butyl-3-methylimidazolium chloride, and acidified ethylene glycol (EG). Soda low lignin pulp, the ionic liquid (IL) pulp and the EG pulp were produced in the laboratory, while soda medium lignin pulp, soda high lignin pulp and acid pulp were produced in a pilot plant having a steam explosion facility. Cellobiose, Avicel and Solka-Floc, all cellulose-based compounds were included in the study to expand the range feedstock composition and architecture. Untreated sugarcane bagasse was included in the study in order to compare product yields and reaction pathways during hydrolysis. On basis the product yields obtained with the various types of pulps, response surface methodology (RSM) was then used for the optimization of the selected pulp for conversion into levulinic acid and furfural. The processing parameters considered were acid concentration, temperature and time.

The solid residue obtained after acid hydrolysis of lignocellulose constitutes the largest proportion of the by-product and its nature and type causes blockages to processing equipment. The chemical composition of the residue was determined using Fourier transform infrared spectroscopy (FTIR), one-dimensional (1D) proton (¹H) and two-dimensional (2D) NMR techniques; correlation spectroscopy (COSY), multiplicity edited heteronuclear single quantum coherence spectroscopy (HSQC), and elemental analysis, in order to provide information on its utility.

The data from this work and those available in the literature were then used to propose a conceptual biorefinery for the conversion of sugarcane bagasse pulp to levulinic and formic acids, and furfural using MSA.

2. Materials and method

2.1 Materials

Sugarcane bagasse was collected from Racecourse sugar mill, Australia (Mackay Sugar Limited). Sugarcane bagasse was washed to remove residual sugars and larger sand and rock particles, and then air-dried for 2 days to constant weight (7.9 wt% moisture). Portions of the dried bagasse were ground by a cutter grinder (Retsch SM100, Germany) and passed through a 0.5 mm aperture screen. Other particle sizes were examined but had limited impact on reaction product yields. Micro-crystalline cellulose (Avicel® PH101) with a particle size range of 75–200 μm was purchased from Sigma Aldrich (Australia). Powdered cellulose (Solka-Floc® 300-FCC) with a particle size range of 30–75 μm was purchased from International Fiber Corporation (Australia). Cellobiose was purchased from Fluka Analytical. Methanesulfonic acid and 1-butyl-3-methylimidazolium chloride was purchased from BASF (Sydney Australia). Other chemicals were purchased from Sigma Aldrich (Australia).

2.1.1 Laboratory-scale pulp preparation. Soda low lignin pulp was prepared from 800 g of whole (i.e., non-milled) bagasse using an alkaline pulping (soda process) in an 18 L stirred Parr reactor at 170 °C for 75 min using a 12 : 1 liquor to dry fiber ratio with 0.4 M NaOH solution. Following alkaline treatment, the pulp was filtered from the black liquor and washed with water (4 × 1 L).

IL pulp was prepared by reacting 4.0 g (dry weight) bagasse (milled to pass through 0.5 mm aperture screen) at a 10 : 1 liquid to dry fiber ratio with 1.2 wt% HCl in 1-butyl-3-methylimidazolium chloride and 20% water at 130 °C for 60 min.¹³ Following IL pretreatment, the pulp was washed with 40 mL of ultrapure water to remove IL and filtered through Whatman 54 filter paper. The pulp was then washed with 20 mL 0.2 M NaOH to remove residual catalyst and loosely bound lignin, followed by 40 mL of ultrapure water and filtered through Whatman 54 filter paper.

EG pulp was prepared by reacting 4.0 g (dry weight) bagasse (milled to pass through 0.5 mm aperture screen) with 1.2% sulfuric acid in EG and 10% water at 130 °C for 30 min.¹⁴ Similar to the IL pulping process, the polyol pretreatment was undertaken at a 10 : 1 liquid to dry fiber ratio and, following pretreatment, the resulting pulp was washed with both 2 M NaOH (20 mL) and ultrapure water (40 mL).

All pulp samples prepared at the laboratory-scale were dried at 50 °C in a vacuum oven overnight to constant weight (∼2–8 wt% moisture) prior to subsequent hydrolysis.

2.1.2 Pilot-plant prepared pulps. Soda medium lignin pulp was prepared in a 150 L stainless steel horizontal reactor (Andritz, USA) with whole bagasse at the Mackay Renewable Biocommodities Pilot Plant (MRBPP), in Mackay, Australia. Bagasse (20 kg at 52.5% moisture) was pretreated with 0.8 M NaOH at 170 °C for 30 min using a 6 : 1 liquid to dry fiber ratio. Following pretreatment, the pulp was subjected to steam explosion at 150 °C (2000 kPa) for 5 min and filtered by pneumatic press through a perforated (2 mm diameter) screen.

Soda high lignin pulp was prepared similar to soda medium lignin pulp but with 1.5 M NaOH solution at 170 °C for 30 min using a 6 : 1 liquid to fiber ratio. The pulp was filtered and washed with water (1 : 1 liquid to fiber ratio) at 150 °C for 10 min. The washed pulp was subjected to steam explosion at 150 °C (2000 kPa) for 5 min and filtered.

Acid pulp was prepared as above with 0.45 wt% H₂SO₄ (on dry fiber) at 170 °C for 15 min using a 3 : 1 liquid to fiber ratio. Following pretreatment, the pulp was subjected to steam explosion at 185 °C (2000 kPa) for 5 min and then filtered.

2.2 Methods

2.2.1 Hydrolysis reaction. Hydrolysis reactions were carried out in 10 mL sealed glass ampoules with an internal diameter of 9.6 mm and length of 150 mm. Reaction volumes were ∼3.5 mL and reactions were undertaken using a fluidized sand bath (Model number: SBL-2D, Techne Inc., Burlington, USA). Reactions were considered to start once reaction temperature was reached (heat up time was typically <2 min). At the end of the reaction, the ampoules were quenched in cold water to stop the reaction. The reaction products were decanted and filtered through a Whatman no. 5 filter paper. The filtrate was collected and stored at −20 °C. Solid products were washed with 14 mL ultrapure water and dried to a constant weight at 60 °C under vacuum.

Evaluation of typical operating conditions was achieved using response surface methodology (RSM) with the software Design Expert v7 (Stat-Ease, Minneapolis, USA). Stepwise regression was employed for analysis of variance (ANOVA).

2.2.2 Feed characterisation. Compositional analyses were undertaken using the procedures developed and reported by the National Renewable Energy Laboratory (NREL, USA).¹⁵ Reported results are the mean of results from duplicate samples. The maximum variation between results for duplicate samples was 2.7%, 3.2% and 4.4% for glucan, xylan, and lignin, respectively. The structure of bagasse and pulp samples was examined using a FEI Quanta 200 Environmental scanning electron microscope, SEM (Hillsboro, OR, USA), at an accelerating voltage range of 5–30 kV.

Infrared spectra were collected using a Nicolet 870 Nexus FTIR system (Nicolet Instrument Corp. Madison, USA) including a Continuum™ IR microscope equipped with a liquid-nitrogen-cooled MCT detector, and an attenuated total reflectance objective incorporating a Si internal reflection element. The contact area with the sample was circular with an approximate diameter of 100 μm. Spectra were collected in the spectral range 4000 to 650 cm⁻¹, using 64 scans and 4 cm⁻¹ resolution.

Water holding capacity was measured similar to the method described by Yu et al.¹⁶ Solid residues were gently mixed with water at >50 : 1 weight ratio for one hour at room temperature and allowed to incubate at 4 °C for at least 14 h to assure homogeneity of water content. The mixtures were drained through filter paper (Whatman 541) by gravity at 4 °C for 24 h. The filtered residues were then transferred to crucibles and the wet mass of each sample was measured. The samples were dried at 110 °C for 24 h and the dry mass of each sample was measured. The difference yielded the amount of water being retained by each solid residue. Water retention value (WRV) was determined using the following formula:

WRV = (mass wet biomass − mass dry biomass dry)/(mass dry biomass)

2.2.3 Liquid product analysis. The concentration of liquid products were determined by high-performance liquid chromatography (HPLC) analysis using an Aminex HPX-87H column, Waters 2489 UV detector (operating at a dual frequency of 210 nm and 280 nm), and a Waters 410 refractive index detector. The column was operated at 60 °C and eluted with 5 mM H₂SO₄ at a flow rate of 0.6 mL min⁻¹. The quantities of products were calibrated against standard solutions of known concentrations and converted to theoretical molar yields based on initial glucose or xylose content, respectively, as follows:

Yield (mol%) = (mol of product in hydrolysate)/(mol of hexose sugars (or pentose sugars) in feed)

2.2.4 Solid product analysis. Solid residues (25 mg) from levulinic acid synthesis reactions were dissolved in 1 mL of DMSO-d₆ and filtered through polypropylene tissue to remove particulates. It should be noted that only ∼20% of solid residues were soluble in DMSO. High resolution NMR spectra were obtained using a Bruker Avance 400 MHz NMR spectrometer (Bruker, Massachusetts, USA) with a gradient probe operating in inverse detection mode. Data processing was performed using ACD/NMR Processor software. The 1D Proton (¹H), and 2D COSY and HSQC NMR techniques were conducted using standard pulse programs. Maleic acid (δ_H = 6.27) was added as an internal reference standard to the NMR solutions to provide quantitative results.

Infrared spectra were collected similar to the method described in Section 2.2.2.

Elemental analysis was performed on the residues using a ‘Carlo Erba’ Elemental Analyzer Model NA1500 instrument (Waltham, MA, USA). Oxygen was obtained by difference. The higher heating value (HHV, MJ kg⁻¹) of the sample was calculated based on the elemental analyses (wt%) as follows:¹⁷

HHV = −1.368 + 0.314 × carbon + 0.701 × hydrogen + 0.032 × oxygen

3. Results and discussion

3.1 Biomass composition

The composition of sugarcane bagasse and pulps produced after pretreatment of sugarcane bagasse under various conditions are presented in Table 1 (includes WRV). Pretreatment increased the proportion of glucan in pulp relative to bagasse, and decreased the proportion of lignin (with the exception of acid pulp). Also, pretreatment decreased the proportion of hemicellulose relative to untreated bagasse, except for soda low lignin pulp. Pilot plant soda pretreatments produced pulps with higher lignin content compared to the laboratory produced soda pulp because of a shorter reaction time (30 min instead of 75 min) despite the use of higher NaOH concentrations. Surprisingly, the soda pulp with the highest proportion of lignin was obtained with the treatment with the highest NaOH concentration. This may simply be due to, in addition to shorter reaction time, the reduced proportion of the cellulose component of the biomass as a consequence of cellulose degradation caused by the severity of the treatment.

Table 1 Composition of pulps and bagasse samples

Composition^a (wt%)	Laboratory treated bagasse				Pilot plant treated bagasse^b
	Bagasse	Soda low lignin	EG	IL	Bagasse	Soda medium lignin	Soda high lignin	Acid
a Reported on an oven dry weight basis.b Bagasse composition used for Mackay Pilot Plant pretreatments.
Extractives	2.0	—	—	—	7.2	—	—	2.9
Cellulose	43.0	60.9	83.7	92.8	45.2	58.5	55.4	52.4
Hemicellulose	19.1	23.9	5.7	1.2	18.7	16.7	15.1	3.2
Lignin (total)	24.0	8.8	5.1	3.0	27.2	12.3	14.9	29.6
- Acid soluble	4.7	6.5	1.5	1.1	5.7	5.4	5.5	2.3
- Acid insoluble	19.3	3.3	3.6	1.9	21.5	6.9	9.4	27.3
Ash	6.2	2.0	1.8	0.6	7.8	15.1	12.7	11.4
WRV	11.24	10.26	5.28	8.84	n.d.	n.d.	10.00	3.62

---

Scanning electron microscopy was used to examine the microscopic structural differences among the feed samples were examined using SEM (Fig. 1). The untreated bagasse sample contains fiber bundles and blocks of intact fibres (Fig. 1a), while acid pretreatment (acid pulp) results in fiber disintegration (with lengths from >10 μm to >100 μm) and a reduction in the sizes of blocks. The soda treated bagasse (Fig. 1c) contains strands of defibrillated fibers (with lengths > 200 μm, and diameters in the range 10–30 μm). The porous structure of the untreated bagasse and soda pulp is reflected in the WRV data in Table 1. The acid pulp has the lowest WRV (not shown in Table 1 is Solka-Floc with a WRV of 3.63). The WRV provides an indirect measure of porosity and is influenced not only by physical effects (fiber size and aspect ratio) but also by chemical aspects such as water binding capacity of amorphous polymers like hemicellulose. It should be noted that the pith fraction (small fiber) in bagasse increased the WRV obtained for this sample. Depithed bagasse is typically 40–70% lower in WRV.¹⁸

Fig. 1 SEM images of (a) untreated bagasse; (b) acid pulp; (c) soda low lignin pulp.

The EG and IL pulps produced samples with the highest proportion of cellulose and lowest proportion of lignin (Table 1). The results clearly demonstrate that, in the presence of an acid catalyst, these solvents are very effective lignocellulose fractionation agents. FTIR analyses (Fig. S1 and S2 ESI†) generally confirmed the compositional analysis data. The β-glycosidic linkage peak at 900 cm⁻¹ (characteristic of deformation of carbohydrates^19,20) increased in intensity in the pulps, indicating an increase in cellulose content as a consequence of reduced proportion of hemicellulose and lignin in certain cases. The uronic acid ester bond peak assigned at 1733 cm⁻¹ and β-ether bond/acetyl group peak at 1245 cm⁻¹ associated with hemicellulose¹⁹ were reduced in intensity for all the indicating that these bonds were cleaved during pretreatment. Lignin-associated peaks^20,21 including at 1605 cm⁻¹ and 1510 cm⁻¹ (aromatic skeleton vibrations) at 1460 cm⁻¹ and 1415 cm⁻¹ (lignin methoxy groups), and at 835 cm⁻¹ (C–H out of plane vibration) were reduced or absent in the pulps, with the exception of acid pulp (due to higher proportion of lignin content in the residue).

3.2 Hydrolysis of bagasse and pulps

Bagasse hydrolysis with MSA was evaluated in order to compare product yield and reaction pathways with previous work on acid hydrolysis of monomeric sugars,¹² and to provide a baseline information for the pulps. The hydrolysis reactions were conducted at range of temperatures (160–200 °C), times (20–80 min), catalyst concentrations (0.06–0.75 M), and feed masses (1.84–3.68 wt%) (Table S1 ESI†). The highest levulinic acid yield was ∼63.5 mol% similar to that previously reported for glucose,^12,22 albeit using a longer reaction time. In the reaction pathways for the conversion of biomass to levulinic acid (Fig. 2), formation of humic polymers is the main competing route in the conversion of HMF to levulinic acid, and it is suppressed at lower temperatures and low HMF concentration.²³ With a heterogeneous reaction system, such as a mixture of sugarcane bagasse and MSA, the cellulose fraction is slowly depolymerized to glucose in situ which subsequently dehydrates to HMF. The rate of HMF formation is typically slower than HMF conversion^22,23 such that gradual cellulose hydrolysis provides a lower concentration of HMF in the reactor, thereby reducing the amount of humic polymer formed compared to aqueous glucose solutions. Such a sequence of events would explain why the levulinic acid yield obtained with bagasse is similar to that of glucose, as the side reaction will be enhanced in the latter system reducing the amount of levulinic acid formed. We hypothesize that the probable negative impact of lignin in bagasse was minimized because the rate of cellulose depolymerization was such that only relatively low concentrations of simple sugars were present in the hydrolysate at any given time.

Fig. 2 Reaction network for conversion of biomass to levulinic acid.²²

Fig. 3 (and Table S2 ESI†) present levulinic acid, furfural and solid residue yields from hydrolysis reactions carried out with a range of MSA concentrations (i.e., 0.1 to 0.5 M) and substrates. The results show that for all substrates (except soda low lignin pulp) irrespective of the composition and cellulose content, at MSA concentration of 0.3 M and 0.5 M, the levulinic acid yields are between 60 and 65 mol%. The yield obtained for soda low lignin pulp was 76.8 mol% under the conditions of 0.3 M MSA at 180 °C in 40 min. Very low yields of levulinic acid was obtained at 0.1 M MSA regardless of substrate.

Fig. 3 Product yields in the MSA catalyzed conversion of biomass.

Given the variation in composition between the substrates (Table 1), it is not immediately clear why soda low lignin pulp gives rise to the highest observed levulinic acid yield. The results also suggest that lignin content (which varied from 3% to 30%) had limited impact on the production of levulinic acid under the range of conditions tested, though the highest yields of levulinic acid were achieved with pulp with the lowest lignin content (i.e., soda low lignin pulp). This is contrary to the results obtained by Alonso et al.,⁹ who demonstrated that when kraft lignin to cellulose ratio was increased from 1 : 1 to 4 : 1 wt% there was a significant reduction in levulinic acid yield after acid hydrolysis. It should be noted that lignin was readily accessible to acid in this system. In contrast, the lignin in bagasse and bagasse pulp in the present study is intimately associated with other cell wall polymers. Interestingly, acid hydrolysis of cellobiose and cellulose substrates produced relatively lower levulinic yields compared to treated and untreated bagasse, although these yields are consistent with other research.⁹ This is likely due to reaction kinetics (Fig. 2). The hydrolysis of simple carbohydrate substrates results in a faster rate of depolymerization (more glucose released in situ) and, therefore, a lower acid catalyst to soluble sugar concentration ratio at any one time, compared to more complex carbohydrates such as bagasse. Under these conditions the reaction pathway to humic polymers is enhanced. From the foregoing, it is likely that the highest levulinic acid yield obtained with soda low lignin pulp is related to both its architecture (fiber size reduction and higher porosity after pretreatment, Table 1) and composition (low lignin and ash content) which influences the reaction pathways.

Optimisation of levulinic acid yields requires consideration of the kinetics of the multiple sequential intermediate reactions and most literature neglect the influence of pretreatment on the initial depolymerisation step. In a previous study, the pore-hindered diffusion and reaction model was used to explain the importance of pore size distribution in enzymatic hydrolysis of pretreated biomass.²⁴ As low soda pulp gave the highest WRV among the pretreated biomass, and hence indicating it has a more porous structure, its degradation process perhaps goes through a similar mechanism as enzymatic hydrolysis of pretreated biomass. The initial cellulose hydrolysis rate is dependent on the surface area of the substrate and together with the diffusion process of the solution play important roles in the kinetics of the overall conversion process.

The highest furfural yields were produced under mild conditions (0.1 M MSA) with the samples with the highest hemicellulose content, and there was a somewhat linear relationship of furfural yield and hemicellulose content (Fig. 3). Furfural degradation to insoluble humic substances increases with increasing reaction severity. Fig. 3 also show a linear relationship between the amount of solid residue and the lignin content of the substrate, as the latter is generally not broken down under these conditions. Larger amounts of solid residue were also produced under mild conditions (0.1 M) because some proportion of the carbohydrates in the feed was not converted (Fig. 3).

3.3 Hydrolysis of soda low lignin pulp

Acid hydrolysis of soda low lignin pulp produced the highest yield of levulinic acid, relatively high furfural yield, and the lowest amount of the solid residue (Table S2†). The soda lignin obtained from such a pretreatment process may find application in the production of high value chemicals and resins.²⁵ Further, while IL-based pretreatment extracted the most lignin from sugarcane bagasse (Table 1), the soda pretreatment process is cheaper. Therefore, on the basis of these results and observations, soda low lignin pulp was investigated in greater detail.

The process used to produce soda low lignin pulp from bagasse removed >60% of lignin and >40% of hemicellulose, generating pulp relatively rich in cellulose (∼60%). Remarkably, 70–80 mol% yields of levulinic acid were achieved (Table 2); these were higher yields than obtained previously with glucose or glucose/xylose mixtures¹² and significantly higher than that produced from bagasse. As previously reported, the reason for the high yield maybe related to the pulp architecture rather than its composition. In a two stage acid catalyzed conversion of wood chips, whereby the first stage, using mild acid pretreatment removed 85% of the hemicellulose with ∼92% cellulose retained in the pulp.²⁶ The resulting cellulose rich pulp was then hydrolyzed under relatively harsh conditions to produce 66 mol% levulinic acid, compared to a yield of 50 mol% if the reaction was performed in a single step.²⁶ The yield improvement was posited to arise from the adverse effect of co-product (i.e., furfural) polymerization with saccharides. From the result of the present work, it is likely that the explanation for the higher levulinic acid yield may also be due to the architecture of the pretreated wood.

Table 2 Levulinic acid, furfural and solid yields from acid-catalyzed reaction of soda low lignin pulp^b

Catalyst (M)	Feed (wt%)	Temp. (°C)	Time (min)	Furfural^a (mol%)	Formic acid^a (mol%)	Levulinic acid^a (mol%)	Solids (wt% feed)
a Based on pentose/hexose sugar content (anhydro-correction).b Note: errors ±3.6% for furfural, ±10.5% for formic acid, ±2.8% for levulinic acid and ±3.4% for solid residue.
0.1	1.72	160	20	22.7	0.0	0.0	70.3
0.1	1.72	180	20	88.3	14.8	6.7	55.8
0.1	1.72	200	20	55.5	57.7	50.5	24.5
0.1	1.72	200	20	68.3	76.9	67.7	26.5
0.1	3.44	160	80	67.2	8.4	4.1	59.6
0.3	2.46	160	80	57.0	50.0	37.9	40.9
0.5	3.93	160	80	21.8	50.5	45.1	28.1
0.045	2.14	180	40	59.9	35.1	11.9	81.7
0.3	2.16	180	40	23.3	86.9	73.0	26.3
0.5	2.46	180	40	10.4	85.7	68.6	32.7
0.3	2.46	180	40	24.2	79.8	70.4	27.8
0.3	2.46	180	75	7.0	84.6	68.2	28.7
0.5	1.72	200	20	10.3	90.9	76.8	24.6
0.5	1.72	200	20	12.5	93.9	79.1	29.6
0.5	4.90	200	40	0.0	64.0	66.8	33.9
0.1	2.46	200	40	22.9	81.2	64.2	26.5
0.3	2.46	200	40	1.5	70.3	59.1	32.0
0.5	2.46	200	40	0.0	54.1	59.5	30.2
0.5	3.44	200	60	0.0	44.5	72.7	39.4
0.1	1.72	200	60	21.1	77.1	66.3	27.7
0.3	2.46	200	60	0.0	55.5	62.8	31.5
0.5	2.46	200	60	0.0	39.0	61.2	28.1

---

The acid hydrolysis of soda low lignin pulp yielded up 88 mol% furfural (based on the pentosan content in the pulp) under mild conditions (Table 2). In general, we observed that furfural yields were slightly lower than that achieved from untreated bagasse but higher than those achieved from xylose as the feed.¹² This may simply be due to the relative ease with which xylose monomers are released from soda low lignin pulp under acidic conditions compared to bagasse. The furfural concomitantly produced may also react with glucose to form a polymeric product and reduce yields.

Optimization of soda low lignin pulp hydrolysis was achieved using response surface methodology (RSM) and revealed that levulinic acid and furfural yields are interrelated as a function of acid concentration, temperature, and reaction time (Table S3†). The RSM model predicts furfural yields of >90 mol% can be obtained from soda low lignin pulp hydrolysis with <0.1 M MSA at 160 °C for 50 to 70 min. Likewise, the model predicts maximum levulinic acid yield (>75 mol%) across a range of experimental conditions, although there is predicted to be small but significant decrease in levulinic acid yield under very harsh reaction conditions (Table S3†).

3.4 Solid product analysis

The largest by-product of acid-catalyzed bagasse and pulp hydrolysis was solid residue. The amount of the solid residue was 15–50 wt%, but was only 15–25 wt% if the lignin content in the feed material was considered. This means that lignin constitutes up to 60 wt% of the solid residue. FTIR difference spectra (Fig. 4, S3 and S4 ESI†) clearly showed increased intensities of peaks associated with furanics (1605 cm⁻¹, 1510 cm⁻¹, 1340–1390 cm⁻¹ and 1280 cm⁻¹)²⁷ and lignin (1605 cm⁻¹, 1510 cm⁻¹, 1460 cm⁻¹, 1415 cm⁻¹ and 1245 cm⁻¹)^19,20 in the pulp-derived residues relative to bagasse residue. Increased peak intensity at 900 cm⁻¹ was also observed in the residues and is characteristic of β-glycosidic linkages between sugar units,^19,20 indicating the incorporation of sugar moieties into the humic structure).

Fig. 4 FTIR difference spectra of acid hydrolysis residues (3 wt% feed, 0.3 M MSA, 180 °C, 40 min) compared to feed material for acid pulp, bagasse and soda low lignin pulp.

Proton NMR spectra of the solid residues contained broad, unresolved signals in the aliphatic (δ_H = 0.8–3.2 ppm), anomeric and oligomeric (δ_H = 3.65–5.25 ppm), aromatic (δ_H = 6–8.5 ppm), and aldehydic regions (δ_H = 9–11 ppm). The NMR spectra were baseline corrected to remove the broad ‘humps’ and the overlying water peak were suppressed to obtain additional information about the residues. A comparison of the proton spectra obtained for bagasse, soda low lignin pulp, and glucose/xylose mixture residues¹² is provided in Fig. S5–S7 (ESI†). The most prominent, distinct peaks were those of organic acids (formic acid [δ_H = 8.13 ppm], levulinic acid [δ_H = 2.09/2.38/2.65 ppm], acetic acid [δ_H = 1.91 ppm], and fatty acid [δ_H = 1.23 ppm]), furfural [δ_H = 6.8/7.55/8.1/9.61 ppm] and the catalyst, MSA [δ_H = 2.32 ppm]. Integration of the levulinic acid peak (compared to the internal standard) indicated that levulinic acid constitutes only 0.2% of the dissolved bagasse residue and 0.24% of the dissolved soda low lignin pulp residue. This compares with 0.15 wt% of levulinic acid measured in the total residue from conversion of sugars.¹²

The main peak associated with the fatty acid (δ_H ∼ 1.23 ppm) was linked to associated peaks at δ_H = 0.85 ppm, 1.47 ppm, 2.18 ppm and 3.37 ppm in the COSY spectrum (Fig. S8 ESI†). The fatty acid protons were identified as CH₂ (using HSQC) so they are likely associated with primary, secondary, and tertiary alkyl groups of fatty acid compounds.²⁸ Given the relative prominence of the peaks associated with these fatty acids in comparison to the other identified compounds, this may indicate that fatty acid production plays a role in the formation of the polymeric residue. Further, from the correlation between anomeric and aromatic hydrogens and carbons, analysis of the HSQC NMR spectra indicated the presence of oligomeric sugars (Fig. S9 ESI†). Identification of oligomeric sugars in the solid residues by NMR correlates with increased intensity of FTIR spectral features attributed to β-glycosidic linkages (Fig. 4).

The carbon content of the residues from acid hydrolysis was significantly higher, and the hydrogen and oxygen significantly lower, relative to the substrate (Table 3). This confirms the condensed and aromatic nature of the solid residue. As such, the residues have higher calorific heating values (∼21–23 MJ kg⁻¹) and would be a useful replacement for bagasse (18 MJ kg⁻¹) in combustion boilers. The heating values obtained for these residues are similar to those of the residues obtained with simple sugars.¹²

Table 3 Elemental composition of feed materials and hydrolysis residues^a

Sample	Carbon (%)	Hydrogen (%)	Sulfur (%)	Oxygen (%)	HHV (MJ kg^−1)
a n.d. – not detected.
Feed materials
Soda low lignin	42.8	6.1	n.d.	51.1	18.0
Bagasse	44.8	5.9	n.d.	45.3	18.3
Lignin	61.5	5.8	n.d.	28.7	22.9
Solid residues
Sugars^12	63.7	3.8	0.0	32.5	22.3
Solka-Floc	63.8	4.6	0.0	31.6	22.9
Soda low lignin	58.5	4.5	0.2	36.8	21.3
Bagasse	57.0	4.7	0.3	38.0	21.0

---

3.5 Biorefinery considerations

The production of chemicals from lignocellulosic materials represents a major challenge because of the complex nature of the substrate. In a biorefinery, sustainability and value is obtained through producing multiple products in high yield. To this end, the incorporation of pretreatment fractionation and multi-stage reactions will allow yields to be improved by optimizing and isolating desired products at each step of the reaction pathway.

Mineral acids are commonly employed as an effective catalyst to transform recalcitrant substrates, however equipment corrosion and environmental pollution issues can place limitations on industrial processes.⁷ The low corrosivity and biodegradable nature of MSA along with its high selectivity goes some way to address these limitations. On the basis of these results of the present study, a conceptual biorefinery based on MSA catalyst is proposed as shown in Fig. 5. Given that the underlying data was obtained at the laboratory-scale, capital, energy, and water balance have not been considered in this scenario, but solid residues have been considered as boiler feed to provide part of the necessary heat and electricity requirements of the facility. An initial alkaline pretreatment step was envisaged to remove the majority of the lignin (68%) and a significant proportion (36%) of the hemicellulose in bagasse. This increases the accessibility of the cellulose (93% cellulose recovered in pulp) to improve levulinic acid yields (as shown in this work) and should reduce the adverse effects of lignin.

Fig. 5 Conceptual lignocellulosic biomass biorefinery process.

Biomass materials with high lignin content produced the largest amount of solid residue due to combined effects of the large acid insoluble fraction and the ability of lignin to polymerize with furfural. Lignin also consumes a small proportion of acid catalyst during acid hydrolysis and can reduce product yields. As the majority of the residue is condensed no other useful application can be derived from it, apart from as a fuel. As a consequence, pretreatment to fractionate out the lignin will reduce effects on reactor operation and unwanted side reactions, and provide good quality lignin for value adding. Approximately 90% of the lignin in alkaline black liquors should be easily recovered. The alkaline salt present in the black liquor can also be recovered and reused to catalyze the depolymerization process.

Following pretreatment, a two-step process (similar to the Biofine process) is proposed in which furfural is initially produced under mild acid conditions (and removed or isolated) prior to the production of levulinic acid with maximal yield. Alternatively, a single stage reaction (reduced capital costs) can be implemented. For the soda low lignin pulp, the RSM model produced in this work was used to simultaneously optimize levulinic acid and furfural yields and predicted yields of 55.4 mol% furfural and 70.8 mol% levulinic acid that could be achieved at 180 °C reaction temperature. These yields for a single reaction step are lower than could be achieved for a two-stage process.

Following filtration of the solid materials produced in the second reactor, product and the catalyst can be recovered via solvent extraction or vacuum distillation methods with emphasis on catalyst recycling to minimize process costs.²⁹ Data for solvent extraction based on a three stage counter-current process with 1 : 1 solvent to hydrolysate ratio and assumed 5% product loss (directed to residue) are included in Fig. 5.

In summary, 1 dry ton of bagasse is predicted to produce ∼190 kg levulinic acid, 89 kg of formic acid, 40 kg furfural, 7 kg acetic acid, 165 kg of soda lignin (for value-adding applications or to upgrade to a bio-crude) and 300 kg solid residue (contains spent MSA catalyst and is suitable as boiler feed). A small net amount of water will be produced from the acid hydrolysis process.

4. Conclusions

Pretreatment of sugarcane bagasse to remove some lignin prior to acid hydrolysis with MSA has the potential to reduce or eliminate key processing challenges such as reactor tube blockages associated with lignin in a levulinic acid biorefinery process, and deliver a lignin product that suitable for value-adding applications. Furthermore, a two-stage hydrolysis process of the treated biomass is required to ensure high yields of levulinic and formic acids, and furfural. The composition of the feedstock appears not to influence levulinic acid yield, although the reverse is true for furfural. Although the type of pretreatment does not influence levulinic acid yield, it is probable that the pulp architecture plays a role.

Acknowledgements

Australian Government funding through the Sugar Research and Development Corporation Grant STU066.

References

1. P. Rein, presented in part at the 26th International Society of Sugar Cane Technologists Conference, Durban, South Africa, 2007.
2. E. de Jong, A. Higson, P. Walsh and M. Wellisch, Biofuels, Bioprod. Biorefin., 2012, 6, 606–624.
3. Y. Zuo, Y. Zhang and Y. Fu, ChemCatChem, 2014, 6, 753–757.
4. M. A. Mellmer, J. M. R. Gallo, D. Martin Alonso and J. A. Dumesic, ACS Catal., 2015, 5, 3354–3359.
5. B. Girisuta, K. Dussan, D. Haverty, J. J. Leahy and M. H. B. Hayes, Chem.–Eur. J., 2013, 217, 61–70.
6. L. Yan, N. Yang, H. Pang and B. Liao, Clean: Soil, Air, Water, 2008, 36, 158–163.
7. A. Morone, M. Apte and R. A. Pandey, Renewable Sustainable Energy Rev., 2015, 51, 548–565.
8. F. D. Pileidis and M.-M. Titirici, ChemSusChem, 2016, 9, 562–582.
9. D. M. Alonso, S. G. Wettstein, M. A. Mellmer, E. I. Gurbuz and J. A. Dumesic, Energy Environ. Sci., 2013, 6, 76–80.
10. D. W. Rackemann and W. O. S. Doherty, Biofuels, Bioprod. Biorefin., 2011, 5, 198–214.
11. Methanesulfonic acid – Benefits of MSA, http://www.arkema-inc.com/index.cfm?pag=1772#corrosive, accessed, April, 2013.
12. D. W. Rackemann, J. P. Bartley and W. O. S. Doherty, Ind. Crops Prod., 2014, 52, 46–57.
13. Z. Zhang, I. M. O'Hara and W. O. S. Doherty, Bioresour. Technol., 2012, 120, 149–156.
14. Syngenta Participations Ag, WO Patent, 168408 A1, Queensland University Of Technology, 2012.
15. A. H. B. Sluiter, R. Ruiz, C. Scarlata, J. Sluiter, D. Templeton and D. Crocker, Determination of structural carbohydrates and lignin in biomass: Laboratory Analytical Procedure (LAP), NREL Report No. TP-510–42618, 2008.
16. O.-Y. Yu, B. Raichle and S. Sink, Int. J. Energy Environ. Eng., 2013, 4, 44.
17. C. Sheng and J. L. T. Azevedo, Biomass Bioenergy, 2005, 28, 499–507.
18. D. L. Williams and D. B. Hodge, Cellulose, 2014, 21, 221–235.
19. X. F. Sun, F. Xu, R. C. Sun, P. Fowler and M. S. Baird, Carbohydr. Res., 2005, 340, 97–106.
20. G.-L. Guo, W.-H. Chen, W.-H. Chen, L.-C. Men and W.-S. Hwang, Bioresour. Technol., 2008, 99, 6046–6053.
21. X.-B. Zhao, L. Wang and D.-H. Liu, J. Chem. Technol. Biotechnol., 2008, 83, 950–956.
22. C. Chang, X. Ma and P. Cen, Chin. J. Chem. Eng., 2006, 14, 708–712.
23. B. Girisuta, PhD thesis, University of Groningen, Netherlands, 2007.
24. J. S. Luterbacher, J.-Y. Parlange and L. P. Walker, Biotechnol. Bioeng., 2013, 110, 127–136.
25. W. De Jong and G. Marcotullio, Int. J. Chem. React. Eng., 2010, 8, A69.
26. T. Runge and C. Zhang, Ind. Eng. Chem. Res., 2012, 51, 3265–3270.
27. S. K. R. Patil and C. R. F. Lund, Energy Fuels, 2011, 25, 4745–4755.
28. G. Knothe and J. A. Kenar, Eur. J. Lipid Sci. Technol., 2004, 106, 88–96.
29. D. J. Hayes, in Biorefineries – Industrial Processes and Products: Status Quo and Future Directions, ed. B. Kamm, P. R. Gruber and M. Kamm, John Wiley & Sons, Weinheim, Germany, 2006, vol. 1, ch. 7, pp. 139–163.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c6ra14772k

---