Ionic liquid pretreatment and fractionation of sugarcane bagasse for the production of bioethanol

Transcription

1 Ionic liquid pretreatment and fractionation of sugarcane bagasse for the production of bioethanol By Sergios K. Karatzos B.Sc. (Hons), M.Sc. A THESIS SUBMITTED FOR THE DEGREE OF DOCTOR OF PHILOSOPHY Faculty of Science and Technology Queensland University of Technology 2011

2

3 IMPORTANT NOTICE The information in this thesis is confidential and should not be disclosed for any reason nor relied on for a particular use or application. Any invention or other intellectual property described in this document remains the property of Queensland University of Technology. i

4 Copyright 2011 By Sergios K. Karatzos Queensland University of Technology ii

5 Keywords Sugarcane; bagasse; lignocellulosics; lignin; cellulose; ionic liquids; pretreatment; decrystallisation; fractionation; aqueous biphasic systems; saccharification; ethanol; biofuel. iii

6 Abstract Pretretament is an essential and expensive processing step for the manufacturing of ethanol from lignocellulosic raw materials. Ionic liquids are a new class of solvents that have the potential to be used as pretreatment agents. The attractive characteristics of ionic liquid pretreatment of lignocellulosics such as thermal stability, dissolution properties, fractionation potential, cellulose decrystallisation capacity and saccharification impact are investigated in this thesis. Dissolution of bagasse with 1-butyl-3-methylimidazolium chloride ([C4mim]Cl) at high temperatures (110 C to 160 C) is investigated as a pretreatment process. Material balances are reported and used along with enzymatic saccharification data to identify optimum pretreatment conditions (150 C for 90 min). At these conditions, the dissolved and reprecipitated material is enriched in cellulose, has a low crystallinity and the cellulose component is efficiently hydrolysed (93 %, 3 h, 15 FPU). At pretreatment temperatures < 150 C, the undissolved material has only slightly lower crystallinity than the starting. At pretreatment temperatures 150 C, the undissolved material has low crystallinity and when combined with the dissolved material has a saccharification rate and extent similar to completely dissolved material (100 %, 3h, 15 FPU). Complete dissolution is not necessary to maximize saccharification efficiency at temperatures 150 C. Fermentation of [C4mim]Cl-pretreated, enzyme-saccharified bagasse to ethanol is successfully conducted (85 % molar glucose-to-ethanol conversion efficiency). As compared to standard dilute acid pretreatment, the optimised [C4mim]Cl pretreatment achieves substantially higher ethanol yields (79 % cf. 52 %) in less than half the processing time (pretreatment, saccharification, fermentation). Fractionation of bagasse partially dissolved in [C4mim]Cl to a polysaccharide rich and a lignin rich fraction is attempted using aqueous biphasic systems (ABSs) and single phase systems with preferential precipitation. ABSs of ILs and concentrated aqueous inorganic salt solutions are achievable (e.g. [C4mim]Cl with iv

7 200 g L -1 NaOH), albeit they exhibit a number of technical problems including phase convergence (which increases with increasing biomass loading) and deprotonation of imidazolium ILs (5 % - 8 % mol). Single phase fractionation systems comprising lignin solvents / cellulose antisolvents, viz. NaOH (2M) and acetone in water (1:1, volume basis), afford solids with, respectively, 40 % mass and 29 % mass less lignin than water precipitated solids. However, this delignification imparts little increase in saccharification rates and extents of these solids. An alternative single phase fractionation system is achieved simply by using water as an antisolvent. Regulating the water : IL ratio results in a solution that precipitates cellulose and maintains lignin in solution (0.5 water : IL mass ratio) in both [C4mim]Cl and 1-ethyl-3-methylimidazolium acetate ([C2mim]OAc)). This water based fractionation is applied in three IL pretreatments on bagasse ([C4mim]Cl, 1-ethyl-3-methyl imidazolium chloride ([C2mim]Cl) and [C2mim]OAc). Lignin removal of 10 %, 50 % and 60 % mass respectively is achieved although only 0.3 %, 1.5 % and 11.7 % is recoverable even after ample water addition (3.5 water : IL mass ratio) and acidification (ph 1). In addition the recovered lignin fraction contains 70 % mass hemicelluloses. The delignified, cellulose-rich bagasse recovered from these three ILs is exposed to enzyme saccharification. The saccharification (24 h, 15 FPU) of the cellulose mass in starting bagasse, achieved by these pretreatments rank as: [C2mim]OAc (83 %)>>[C2mim]Cl (53 %)=[C4mim]Cl(53%). Mass balance determinations accounted for 97 % of starting bagasse mass for the [C4mim]Cl pretreatment, 81 % for [C2mim]Cl and 79 %for [C2mim]OAc. For all three IL treatments, the remaining bagasse mass (not accounted for by mass balance determinations) is mainly (more than half) lignin that is not recoverable from the liquid fraction. After pretreatment, 100 % mass of both ions of all three ILs were recovered in the liquid fraction. Compositional characteristics of [C2mim]OAc treated solids such as low lignin, low acetyl group content and preservation of arabinosyl groups are opposite to those of chloride IL treated solids. The former biomass characteristics resemble those imparted by aqueous alkali pretreatment while the latter resemble those of v

8 aqueous acid pretreatments. The 100 % mass recovery of cellulose in [C2mim]OAc as opposed to 53 % mass recovery in [C2mim]Cl further demonstrates this since the cellulose glycosidic bonds are protected under alkali conditions. The alkyl chain length decrease in the imidazolium cation of these ILs imparts higher rates of dissolution and losses, and increases the severity of the treatment without changing the chemistry involved. vi

9 List of publications Poster presentations Karatzos, S. K., Edye L.A. and Doherty W.O.S., 2010, Optimisation of lignocellulose dissolution in ionic liquids as a pretreatment strategy for ethanol production. 32 nd Symposium on Biotechnology for Fuels and Chemicals: Tampa, FL, USA. Karatzos, S. K., Edye L.A. and Doherty W.O.S., 2009, Evaluation of lignocellulose dissolution in ionic liquids as a pretreatment strategy for ethanol and lignin production. 31 st Symposium on Biotechnology for Fuels and Chemicals: San Francisco, CA, USA. Doherty W.O.S, Edye, L.A., O Hara, I., Nanayakkara, B., Rainey, T., Tan, S., Cronin, D., and Karatzos, S. K., 2008, Comparative study of effects of sugarcane biomass fractionation strategies for production of chemicals and biofuels. The International Conference on Biorefinery: Beijing, China. vii

10 Acknowledgements I thank my supervisors for invaluable help particularly with the last weeks of writing, my lab colleagues for hints and chats, my flatmates and friends for fun and food, and my family and girlfriend for all the love and support. Funding was generously provided by the Greek State Scholarship Foundation (IKY), the Queensland Government, and Queensland University of Technology. The Joint BioEnergy Institute (Emeryville, CA, USA) kindly provided funding and facilities from January to April 2010 during my research project at their laboratories. Supervisory team Dr. Leslie A. Edye, QUT Dr. William O.S. Doherty, QUT viii

11 The work contained in this thesis has not been previously submitted to meet requirements for an award at this or any other higher education institution. To the best of my knowledge and belief, the thesis contains no materials previously published or written by another person except where due reference is made Signature... Date... ix

12 Contents Abstract... iv Abbreviations and Nomenclature... xx CHAPTER 1 INTRODUCTION Background Renewable liquid fuels and chemicals from lignocellulosic biomass Sugarcane bagasse The importance of pretreatment and fractionation Research aim Objectives Novelty Summary of chapters... 6 CHAPTER 2 LITERATURE REVIEW Overview Lignocellulosic biomass: chemical and structural characteristics Cellulose Hemicelluloses Lignin Lignin-carbohydrate bonds Cellulose microfibrils: The foundation units of the cell wall construct The cell wall layers Mechanism of cell wall swelling Pretreatment Overview of the conversion of biomass to ethanol fuel x

13 2.3.2 Goals of pretreatment Pretreatment technologies Conventional cellulose solvents Enzyme saccharification of cellulosics after pretreatments Ionic liquid based pretreatment technologies Ionic liquids: properties and history Cellulose dissolution using ionic liquids Lignin dissolution in ionic liquids Biomass dissolution and pretreatment in ionic liquids Rationale CHAPTER 3 METHODOLOGY Bagasse Chemicals Uncertainty (or error) analysis of quantitative measurements Mass values Karl Fischer titration Determination of IL dissolution extent and losses Dissolution Recovery of undissolved solids (UND) and dissolved-thenprecipitated solids (DS) Gravimetric determination of percent mass dissolution Gravimetric determination of percent mass losses Bagasse soda lignin preparation Real time FTIR and reaction calorimetry Differential Scanning Calorimetry Thermogravimetric analysis xi

14 3.11 Cellobiose hydrolysis kinetics Compositional analysis of solid fractions Preparation of IL pretreated samples for enzyme saccharification Preparation of dilute acid pretreated samples Enzymatic saccharification XRD cellulose crystallinity measurement Saccharification and fermentation ATR-FTIR Aqueous biphasic systems Preparation of ABSs Cloud point titrations Ion concentration determination (for ABS distribution ratios) Quantification of [C4mim]Cl deprotonation using an acid titration Mass balance determinations for three IL treatments Compositional analysis of solid fraction Compositional analysis of monosaccharides in liquid fraction Compositional analysis of oligosaccharides in liquid fraction Acetyl bromide for lignin quantification in solid fractions 2 and Recovery of IL Enzymatic saccharification of solids from 3 IL treatments CHAPTER 4 RESULTS PRETREATMENT Biomass dissolution in IL and recovery by addition of water Ionic liquids used Factors affecting biomass dissolution Thermal stability of bagasse components in [C4mim]Cl xii

15 4.1.4 Ionic liquid pretreatment comparison with dilute acid pretreatment Summary Role of non-dissolution pretreatment effects on enzyme saccharification Compositional analysis Enzyme saccharification X-Ray diffractometry (XRD) of bagasse High temperature phase of crystalline cellulose ATR-FTIR analysis of undissolved fractions Summary CHAPTER 5 RESULTS - FRACTIONATION Aqueous biphasic systems Choice of kosmotropic salts for aqueous biphasic systems Evaluation of ABS stability with coexistence curves Evaluation of the phase divergence of ABS using distribution ratios Effect of biomass loading on distribution ratios of ABSs Chemical instability of imidazolium ILs in alkaline ABSs Summary Aqueous single phase fractionation systems Summary Preferential precipitation by incremental additions of water Comparison of three IL pretreatment and fractionation systems Compositional analysis Structural analysis by ATR-FTIR xiii

16 5.4.3 Enzyme saccharification Precipitation of solid fraction 2 and Mass recovery of bagasse components after pretreatment Mass recovery of the ionic liquid solvent after pretreatment Effect of IL anion and cation on pretreatment Summary CHAPTER 6 CONCLUSIONS Findings Chapter 4: Pretreatment Chapter 5: Fractionation Future work Appendix I Appendix II Appendix III References xiv

17 LIST of FIGURES Figure 2.2.1: Molecular structure of cellulose... 9 Figure 2.2.2: Most probable hydrogen bond patterns of cellulose allomorphs Figure 2.2.3: Molecular structure of glucuronoarabinoxylan Figure 2.2.4: Lignin monomer units Figure 2.2.5: The most common linkages between lignin phenylpropane units Figure 2.2.6: Partial structure of a hypothetical lignin molecule from European beech (Fagus sylvatica) Figure 2.2.7: Commonly occurring covalent linkages between GAX and lignin in grasses Figure 2.2.8: Possible covalent cross-links between polysaccharides and lignin in cell walls Figure 2.2.9: Detailed structure of cell walls Figure : Cell wall layers and organisation of the cellulose microfibrils Figure : Light microscope image showing ballooning of a sulphate pulp fibre (Pinus silvestris) Figure 2.3.1: Gross representation of the main steps in a biomass to ethanol process Figure 2.3.2: Consolidation of bioprocessing in cellulosic ethanol production Figure 2.3.3: The effects of lignin, acetyl groups, and crystallinity on enzyme adsorption and enzymatic hydrolysis of biomass Figure 2.3.4: The participating inputs and outputs in a pretreatment process Figure 2.3.5: Conventional cellulose solvents Figure 2.3.6: Gross schematic of the hydrogen bonding formed between NMMO and cellulose hydroxyls upon dissolution Figure 2.3.7: EDA interactions between cellulose and a non-derivatising solvent (e.g. NMMO) Figure 2.4.1: Common ions in ionic liquids Figure 2.4.2: Structure proposed for a covalent binding of [C2mim]OAc to a cellooligomer (DP 6-10) Figure 2.4.3: Proposed dissolution mechanism of cellulose in [C4mim]Cl xv

18 Figure 3.6.1: Process for recovering undissolved and dissolved-then-precipitated solids Figure 3.8.1: The Mettler-Toledo RC1e reaction calorimeter and ReactIR FTIR probe Figure : Diffractogram of bagasse Figure : Linear relationship of refractive index to [C4mim]Cl concentration in water Figure : Flow chart of the fractionation process used in mass balance experiments Figure 4.1.1: ILs used in this study Figure 4.1.2: Effect of temperature on bagasse dissolution in [C4mim]Cl for 90 min Figure 4.1.3: Effect of residence time on bagasse dissolution in [C4mim]Cl (150 C) Figure 4.1.4: Effect of bagasse moisture content on bagasse dissolution in [C4mim]Cl Figure 4.1.5: Effect of ionic liquid choice on bagasse dissolution Figure 4.1.6: Real time FTIR of bagasse polysaccharides upon dissolution in [C4mim]Cl Figure 4.1.7: Differential scanning calorimetry profiles Figure 4.1.8: First derivative of thermogravimetric analysis curves Figure 4.1.9: Cellobiose hydrolysis and glucose accumulation in [C4mim]Cl Figure : Hydrolysis of cellobiose in the absence of water Figure : Enzyme saccharification of bagasse pretreated with [C4mim]Cl and dilute acid Figure : Images of [C4mim]Cl-pretreated bagasse at 140 C and 150 C Figure : Initial rates of enzyme saccharification and XRD crystallinity indices for IL- and dilute acid-pretreated bagasse (TRS) Figure : Glucan and xylan saccharification extent after 121 h for IL- and dilute acid- pretreated bagasse (TRS) Figure : Fermentation kinetics of [C4mim]Cl-treated bagasse after enzyme saccharification xvi

19 Figure 4.2.1: Saccharification of the undissolved bagasse after [C4mim]Cl pretreatment at different conditions Figure 4.2.2: Initial rates of enzyme saccharification and XRD crystallinity indices for [C4mim]Cl-pretreated bagasse fractions Figure 4.2.3: Glucan and xylan saccharification extent after 121 h for [C4mim]Clpretreated bagasse fractions Figure 4.2.4: Diffractograms of undissolved bagasse after [C4mim]Cl pretreatment Figure 4.2.5: Optical microscopy images showing swelling of miscanthus grass particles in [C2mim]Cl Figure 4.2.6: FTIR spectra of IL- and dilute acid-pretreated bagasse fractions Figure 5.1.1: A NaOH /[C4mim]Cl ABS with 1% mass bagasse load Figure 5.1.2: FTIR spectra of each phase of two NaOH /[C4mim]Cl ABSs Figure 5.1.3: FTIR spectra of each phase of a NaOH / [C4mim]Cl ABS loaded with 15 % soda lignin Figure 5.1.4: The Hofmeister series (ions relevant to this study in bold) Figure 5.1.5: Coexistence curves of [C4mim]Cl with selected kosmotropic salts Figure 5.1.6: Phase diagrams of [C4mim]Cl with various salts Figure 5.1.7: Activity coefficients of NaOH and KOH at different molarities Figure 5.1.8: Distribution ratios of ions in ABSs and their molal composition Figure 5.1.9: Ion migration diagrams based on distribution ratios Figure : The effect of bagasse loading on the ion distribution ratios in ABSs 133 Figure : Carbene formation from imidazolium-based ILs Figure : HCl titration of the IL phase of a [C4mim]Cl/ NaOH ABS Figure : Enzyme saccharification of total recovered solids (TRS) from partial bagasse dissolution in [C4mim]Cl using different antisolvents Figure Enzyme saccharification of completely dissolved bagasse (DS) precipitated from [C4mim]Cl using different antisolvents Figure 5.3.1: ph of [C2mim]OAc and [C4mim]Cl aqueous solutions at different water : IL mass ratios Figure 5.3.2: Lignin and cellulose precipitation observed at different water : IL mass ratios of [C2mim]OAc and [C4mim]Cl aqueous solutions xvii

20 Figure 5.4.1: Process flow chart of a fractional precipitation separation of IL treated bagasse using incremental additions of water Figure 5.4.2: FTIR spectra of bagasse treated with different ILs Figure 5.4.3: FTIR spectra of DS and UND bagasse treated with different ILs Figure 5.4.4: Glucan saccharification of extracted bagasse treated with 3 ILs Figure 5.4.5: Xylan saccharification of extracted bagasse treated with 3 ILs Figure 5.4.6: FTIR spectra of precipitate recovered after precipitation in 3.5 water : IL mass ratio (acidified to ph < 1) in three ILs Figure 5.4.7: Mass distribution of bagasse components in [C4mim]Cl pretreatment fractions Figure 5.4.8: Mass distribution of bagasse components in [C2mim]Cl pretreatment fractions Figure 5.4.9: Mass distribution of bagasse components in [C2mim]OAc pretreatment fractions Figure : Fraction of original bagasse polysaccharides saccharified in 24 h (15 FPU g -1 glucan) after pretreatment in three ILs LIST OF TABLES Table 2.3.1: Enzymatic saccharification from selected pretreatment systems Table 4.1.1: Compositional analysis of bagasse pretreated with [C4mim]Cl and dilute acid Table 4.1.2: Comparison of ethanol yields from IL and from dilute acid pretreatment Table 4.2.1: Compositional analysis of dissolved-then-precipitated solids (DS) and undissolved solids (UND) from [C4mim]Cl pretreatment of bagasse Table 4.2.2: Effect of residence time on the composition of undissolved bagasse after [C4mim]Cl pretreatment at 150 C Table : Assignments of FTIR-ATR absorption bands for bagasse Table 4.2.4: Ratios of FTIR absorbances attributed to ester bonds and the aromatic ring of lignin xviii

21 Table 5.1.1: Gibbs free energies of hydration ( G hyd ) of selected ions Table 5.1.2: Water solubilities of selected inorganic salts Table 5.1.3: Deprotonation of imidazolium IL in top phase of ABSs Table 5.2.1: Compositional analysis of total recovered solids (TRS) from partial bagasse dissolution in [C4mim]Cl using different antisolvents Table : Compositional analysis of completely dissolved bagasse (DS) precipitated from [C4mim]Cl using different antisolvents Table 5.4.1: Compositional analysis of SF1 solids from pretreatment of ethanolextracted bagasse with three different ILs Table 5.4.2: FTIR crystallinity indices of IL-pretreated solids Table 5.4.3: Mass recovery, delignification and enzyme saccharification resulting from treatment with different ILs Table 5.4.4: Mass recovery and lignin content of solids recovered from the liquid fraction after treatment with three ILs Table 5.4.5: Mass balance of bulk biomass and of biomass components from three treatments with different ILs Table 5.4.6: Mass recovery of ionic liquid ions after use xix

22 Abbreviations and Nomenclature [Allylmim]Cl: 1-allyl-3-methylimidazolium chloride [C1mim] MeSO 4 : 1-methyl-3-methylimidazolium methyl sulphonate [C2mim]Cl: 1-ethyl-3-methylimidazolium chloride [C2mim]OAc: 1-ethyl-3-methylimidazolium acetate [C4mim]BF 4 : 1-butyl-3-methylimidazolium tetrafluoroborate [C4mim]CF 3 SO 3 : 1-butyl-3-methylimidazolium trifluoromethanesulphonate [C4mim]Cl: 1-butyl-3-methylimidazolium chloride [C4mim]PF 6 : 1-butyl-3-methylimidazolium hexafluorophosphate [C4mmim]Cl: 1-butyl-2,3-dimethylimidazolium chloride ABS: aqueous biphasic system AFEX: ammonia fibre explosion AIL: acid insoluble lignin ARP: ammonia recycle percolation ASL: acid soluble lignin ATR: attenuated total reflectance BASF: BASF, the chemical manufacturing corporation b.p.: boiling point CBP: consolidated bioprocessing df: degrees of freedom DMA / LiCl: dimethylacetamide / lithium chloride DMA: dimethylacetamide DMSO: dimethylsulphoxide DP: degree of polymerisation DS: dissolved-then- precipitated fraction DSC: differential scanning calorimetry EDA: electron donor-acceptor EOL: ethanol organosolv lignin FPU: filter paper units FTIR: Fourier transform infrared spectroscopy xx

23 HMF: hydroxymethylfurfural HPLC: high-pressure liquid chromatography IC: ion chromatography IL: ionic liquid LCB: lignocellulosic biomass LF: liquid fraction m.p.: melting point n/a: not applicable n/d: not determined NMMO: N-Methylmorpholine-N-oxide NMR: nuclear magnetic resonance NREL: National Renewable Energy Laboratory (Golden, CO, USA) PEG: polyethylene glycol rpm: revolutions per minute SF: solid fraction SHF: separate hydrolysis and fermentation SRS: sugar recovery standard SSCF: simultaneous saccharification and co fermentation SSF: simultaneous saccharification and fermentation STEX: steam explosion TGA: thermogravimetric analysis TRS: total recovered solids (sum of undissolved and precipitated solids) UND: undissolved fraction XRD: X-ray diffractometry YPD: solution containing yeast extract, peptone and dextrose xxi

24

25 CHAPTER 1 INTRODUCTION 1.1 Background Renewable liquid fuels and chemicals from lignocellulosic biomass Lignocellulosics, whether in the form of dedicated energy crops such as sorghum, switchgrass and cardoon, agricultural residues such as sugarcane bagasse and corn stover, or from forestry residues, present a renewable resource with an energy value of approximately 300 x J worldwide [1]. With world energy demand predicted to increase in the near future [2], and fossil fuel reserves being depleted and non-renewable, biomass resources have drawn much attention as renewable feedstocks for alternative fuels and chemicals. This attention is further driven by issues such as the need to reduce CO 2 emissions, the need to rely on local, renewable and sustainable fuel sources (e.g. biofuel from crops farmed on marginal land) and the need to reduce dependence on remote and unstable fuel sources (e.g. petroleum imports). Among the strategies for biomass valorisation is hydrolysis and fermentation. Ethanol fuel and other products of fermentation can be manufactured from sugars, and high value polymers can be synthesized from lignin. When these are derived from lignocellulosic biomass (LCB), a non-food renewable resource, they present a promising sustainable alternative to petroleum based fuels and chemicals. This thesis investigates the conversion of sugarcane bagasse to ethanol fuel using ionic liquid pretreatment. The initial focus of this study was optimisation of pretreatment of sugarcane bagasse by dissolution in ionic liquid and fractionation using aqueous salt biphasic systems [3]. Problems with this pretreatment process (specifically with increasing convergence of biphases as biomass loading increases) and, at the time, lack of published works on biomass ionic liquid interactions, led to a broader study of bagasse-imidazolium ionic liquid (IL) interactions and impacts of these interactions on enzymatic saccharification of the polysaccharide component of treated bagasse. 1

26 This chapter presents the characteristics of bagasse (Section 1.1.2), the importance of pretreatment and the benefits and challenges of ionic liquids (Section 1.1.3). The aim and objectives of this study are described in Sections 1.1 and 1.2, the novelty of the work in Section 1.3 and finally the thesis Chapter layout is provided in Section Sugarcane bagasse Sugarcane (Saccharum officinarum) is a sugar crop that thrives in tropical climates and produces biomass prolifically; it is a member of the grass family (Poaceae), has a high photosynthetic efficiency and produces a total biomass yield of between 20 t ha -1 yr -1 to 30 t ha -1 yr -1 on a dry basis [4]. Accordingly, high CO 2 sequestration capacity is an inherent advantage of this crop, with a reported CO 2 fixation rate of 49 t ha -1 yr -1 for south Texas USA, more than three times that for mid-latitude temperate forests [5]. The extraction of sucrose from sugarcane stems produces a biomass residue known as bagasse. Bagasse, containing ca. 40 % - 45 % cellulose, 25 % - 30 % hemicellulose and 25 % - 30 % lignin, is an ideal candidate for the production of ethanol and biopolymers [4]. Infrastructure for the processing of sugarcane is already well established, and the feedstock is already delivered at central locations as part of the sugar manufacturing process. The energy available in harvested cane biomass is well in excess of that required to process cane to raw sugar. As a result most Australian raw sugar factories are configured to operate at low thermodynamic efficiencies to dispose of the surplus fibre and avoid an unwanted accumulation of bagasse at the end of crushing season. With only modest improvements to boiler and process steam efficiencies, the energy required for sugar processing could still be met whilst 35 % of the bagasse produced at the factory made available for ethanol or power production. Based on a 34 million tonne cane harvest for the Australian industry, this 35% bagasse surplus corresponds to approximately 1.7 million tonnes of available dry fibre [6, 7]. 2

27 1.1.3 The importance of pretreatment and fractionation Ethanol is produced by hydrolysis and fermentation of the polysaccharides in LCB. However these processes are inhibited by the complex structure of LCB. Therefore pretreatment of LCB is necessary prior to hydrolysis and fermentation. LCB, found in the structural tissue of plants, is a complex material designed by nature to resist physical, chemical and biological (e.g., microbial and enzymatic) attack. It is predominantly comprised of three biopolymers viz. cellulose, hemicelluloses and lignin. It also contains small amounts of extractives (soluble nonstructural materials) and ash. This complex material is recalcitrant to the chemical and/or biological processing involved in the production of cellulosic ethanol and needs to be pretreated. steps, viz.: The conversion of lignocellulose into ethanol involves three main processing Pretreatment (opening up of the complex lignocellulosic structure to increase surface area and improve further processing by enzymes) Saccharification (hydrolysis of the holocellulose or polysaccharide fraction of lignocellulosics to monosaccharides) Fermentation (conversion of the monosaccharide source to ethanol using yeast or other organisms) This study investigates ionic liquids as agents for the first two steps: pretreatment and fractionation. Pretreatment is an indispensible and expensive processing step in the conversion of lignocellulosic biomass into fermentable sugars [1, 8, 9]. Fractionation is an optional step that follows pretreatment and makes use of the chemical properties of the pretreated/accessible biopolymers in order to isolate them in a pure or partially purified form. Lignin is a known inhibitor of enzyme hydrolysis [10, 11] while it is also potentially a high value feedstock for the polymer industry. Separating the lignin from the polysaccharide fraction enhances saccharification yields and may add a high value lignin product to the process. 3

28 Fractionation can also remove fermentation inhibitors such as sugar degradation products (e.g. hydroxymethylfurfural). Most pretreatment technologies are either physical (e.g., size comminution, steam explosion and hydro-thermolysis) or chemical, utilizing organic solvents, acids or alkalis [12-14]. These chemical pretreatment technologies occur by either acid or alkali mechanisms at high temperatures and pressures (often at extreme ph values) and produce products that may be inhibitory to enzymatic saccharification or fermentation. Harsh pretreatment renders lignin in a condensed, non-reactive form and reduces its potential value for functionalisation and polymer manufacturing [15]. Recently, ionic liquids (ILs) have drawn a great deal of attention as green solvents for processing of lignocellulosics. ILs are a class of organic salts that are liquid at temperatures below 100 C. Many ILs are non-volatile, non-explosive, stable at a wide range of temperatures and reaction severities and compatible with a wide array of organic and inorganic functional chemicals and solvents. ILs have unique solubilisation characteristics compared to conventional molecular solvents and some are known to achieve solvation of the whole lignocellulosic structure. Formation of homogeneous solutions of lignocellulose in IL is a property responsible for a number of beneficial pretreatment and fractionation characteristics. Such characteristics include the dissolution-then-precipitation of a disordered (decrystallised) cellulose and the potential for clean fractionation of cellulose, lignin and hemicelluloses [3, 16-18]. 1.1 Research aim The overarching objective of this study is to investigate and optimise the performance of imidazolium ionic liquids as a pretreatment and fractionation strategy for bagasse. 4

29 1.2 Objectives The objectives of this work are to: Investigate factors affecting dissolution of biomass in IL 1-butyl-3- methylimidazolium chloride ([C4mim]Cl) and improve current understanding of this pretreatment process Assess the performance of optimised IL ([C4mim]Cl) pretreatments and compare with dilute acid pretreatment Investigate the lignin-polysaccharide fractionation efficiency of single and biphase aqueous systems after IL ([C4mim]Cl) pretreatment of bagasse. Compare bagasse treatment in three imidazolium ILs ([C4mim]Cl, 1-ethyl-3- methylimidazolium chloride or [C2mim]Cl, 1-ethyl-3-methylimidazolium acetate or [C2mim]OAc) and understand effects of anion and cation variation on saccharification yields, lignin fractionation efficiency, and total mass balances. 1.3 Novelty While ionic liquids have been extensively studied as solvents for cellulose over the last decade, there are few accounts in the literature of whole biomass dissolution and the effect of this dissolution on saccharification kinetics. Despite the recent increase in research reports on IL pretreatment of biomass (which is cited in the results and discussion of this thesis), this work remains novel and contributes new knowledge. The compositional and structural analysis of dissolved and undissolved fractions has the potential of improving the understanding of ionic liquid pretreatment and no such detailed analysis is available at present. The direct comparison of saccharification kinetics between different ionic liquids and dilute acid pretreatment is also a novelty. The high viscosity of ionic liquids in combination with their interference and occasional incompatibility with analytical instrumentation (e.g. chromatography and spectroscopy) has discouraged the scientific community from reporting extensively on full mass balance closures of such processes. In this work mass balance closures are presented for pretreatment processes using three different ionic liquids. 5

30 1.4 Summary of chapters this work. This chapter introduces the background, the objectives and the novelty of Chapter 2 reviews the relevant literature that motivated this study while it provides a background for the discussion of the emerging results. It covers the structure of lignocellulosics, the characteristics of pretreatment technologies and the properties of ionic liquids in the context of cellulose and biomass dissolution. Chapter 3 describes the methodology and instrumentation used to produce the results. It specifies the pretreatment and enzyme saccharification reaction conditions and it details: a) protocols for the quantification of dissolution rates and associated losses b) standardised wet chemistry methods for the compositional analysis of pretreated bagasse and the associated liquid effluents (e.g. acid hydrolysis and acetyl bromide digestion), c) spectroscopic instrumentation for the structural analysis of bagasse (e.g. infrared spectroscopy and X-ray diffraction, XRD) and d) methods for assessing the stability of biphasic systems (e.g. cloud point titrations and calculation of phase divergence coefficients). It also provides the protocol by which the challenging task of monitoring mass balances of ionic liquid pretreatment processes was carried out. Chapters 4 and 5 present and discuss the results emerging from the experimentation of this project. Chapter 4 reports the results on a simple IL ([C4mim]Cl) pretreatment based on partial dissolution and precipitation using water. The extent of dissolution and associated losses at different conditions are examined. The saccharification performance and compositional/structural characteristics of IL treated bagasse are discussed and compared to untreated and dilute acid treated bagasse. Chapter 5 reports on experimentation with fractionation systems. The stability and divergence of biphasic systems and the associated fractionation difficulties are revealed. Single phase fractionation systems employing solutions that are lignin solvents / cellulose antisolvents are also investigated. The use of incremental additions of water in IL / bagasse partial dissolutions to precipitate cellulose and keep lignin in solution is examined as a 6

31 fractionation strategy. Finally mass balances are determined for three IL pretreatments ([C4mim]Cl, [C2mim]Cl and [C2mim]OAc). All results are compared to those of previous works and their impact on current knowledge emphasised. study. Chapter 6 summarises the findings and draws the conclusions from this Appendices present extra experimentation and data to which the main text occasionally refers. 7

32 CHAPTER 2 LITERATURE REVIEW 2.1 Overview This chapter covers the literature relevant to ionic liquid pretreatment of lignocellulosics for the purpose of enzymatic hydrolysis of polysaccharides to fermentable sugars (saccharification). The literature post 2008 is reviewed in comparison to the results of this work. The description of lignocellulosic biomass, beginning from component molecules (e.g. cellulose) and extending to the structural characteristics of the whole plant tissue (e.g. cell wall layers), is covered in Section 2.2. In this section the swelling of cell wall layers is emphasized as it is an important precursor to other events such as dissolution and saccharification. Section 2.3 covers pretreatment as a first step in the process of producing fermentable sugars. It describes the structural changes contributing to ease of LCB saccharification. Finally it reviews some representative pretreatment technologies and how they effect these changes. Ionic liquids as solvents for cellulose and LCB are reviewed in the final section (Section 2.4). In this section, the characteristics of ionic liquids as pretreatment and clean fractionation agents are emphasized. 8

33 2.2 Lignocellulosic biomass: chemical and structural characteristics Lignocellulosic biomass (LCB, the mass of mature terrestrial plants) primarily comprises woody (lignified) fibre in the cell walls of dead (no longer metabolically active) tissues that provide mechanical support to the plant. For example sclerenchyma tissue found in stems, trunks and branches is rich in LCB. As the term LCB suggests, it is predominantly comprised of the lignin (a phenolic polymer) and polysaccharides (namely cellulose and hemicelluloses). In a simplified depiction, cellulose can be seen as the skeleton of the cell wall which is surrounded by hemicelluloses as a filling matrix and lignin as an encrusting material [19]. In reality, its structure is complex and varies among plant genotypes and even among phenotypes. However, the general cell wall characteristics discussed here are common to most terrestrial plant species that yield LCB Cellulose Cellulose comprises 40 % to 45 % of the dry mass of LCB and it is located predominantly in the secondary wall. It is an unbranched homopolysaccharide that consists of β-(1 4) linked D-glucopyranosyl units. Each glucose unit is rotated 180 o with respect to its neighbour, so that the structure repeats itself every cellobiose (glucose dimer) unit (see Figure 2.2.1). The three hydroxyl groups at C-2, C-3 and C- 6 positions of the glucopyranosyl units are involved in the hydrogen bonding in cellulose crystal structures. An aldehyde group in a hemiacetal structure is found at the C-1 end of the cellulose chain and a hydroxyl group at the C-4 end. The C-1 end has reducing properties while the C-4 end is non-reducing. Finally, the conformation of the glucopyranosyl unit is a 4 C 1 chair [20]. Figure 2.2.1: Molecular structure of cellulose 9

34 The stereochemical conformation of cellulose favours regular tight packing of its long chains (degree of polymerisation (DP) 10000) resulting in crystalline regions in native cellulose. These crystalline regions provide for a dense network of intramolecular and intermolecular hydrogen bonds and make cellulose a high tensile strength, water insoluble polymer. Other glucose polymers (glucans) with different stereochemical conformation have very different physical and chemical behaviour to cellulose. For example starch, which is a mixture of linear and highly branched α-anomeric glucans, has very low tensile strength and dissolves readily in water [19, 21]. The solubility of the homologous series of β-(1 4) linked D-glucopyranosyl oligosaccharides in water decreases as the DP increases. Glucose is soluble in water (54.6 g (100 ml) -1 at 30 C [22]), cellohexose (cellulose oligomer of DP 6) is less soluble and a cellulose oligomer of DP 30 is completely insoluble [20]. Cellulose is capable of forming a number of crystal structures, or allomorphs, which differ in conformation and packing arrangement. The allomorph of native cellulose is known as cellulose I, whereas the allomorph found in crystalline regions of swollen or dissolved cellulose is known as cellulose II. The unit cell of the cellulose I crystal allomorph is composed of four glucose moieties in two parallel (i.e. reducing end at same end of adjacent cellulose chains) cellulose chains (see Figure 2.2.2). This conformation provides for two types of intramolecular hydrogen bonds, namely, from O(6) in one glucose residue to O(2)H in the adjacent glucose and also from the ring oxygen (O(5)) to O(3)H. The chains are then held together by hydrogen bonds from O(3) in one chain to O(6)H in the other. Cellulose II is formed by swelling of cellulose fibres containing regions of the cellulose I allomorph with chemical agents such as strong alkali and subsequent addition of water. Since the strongly hydrogen bonded cellulose II is thermodynamically more favoured than cellulose I, it cannot be reconverted to cellulose I. Unlike cellulose I, cellulose II is composed of chains which run 10

35 antiparallel (i.e. reducing ends at opposite ends to adjacent chains). The structure of cellulose II (see Figure 2.2.2) results in less intramolecular and more intermolecular hydrogen bonding as compared to cellulose I. The O(3)H to O(5) bond is maintained as the only intramolecular hydrogen bond in cellulose II while the O(6) to O(2)H in 020 plane and the O(2)H to O(2) to the chain along the diagonal in the 110 plane (not shown, into and out of the page), account for the intermolecular bonding [20, 21]. Figure 2.2.2: Most probable hydrogen bond patterns of cellulose allomorphs (from Kroon-Batenburg [23]) Weimer et. al. [24] studied the digestibility of different cellulose allomorphs by ruminal cellulolytic bacteria and concluded that cellulose I is more digestible 11

36 than cellulose II. Wada et al. [25] reported that the hydrated form of cellulose II is more amenable to enzyme saccharification than cellulose I [25]. They also reported the saccharification rate for an anhydrous cellulose II sample to be higher than that of a cellulose I sample. However, this is not an effect of the allomorph transition since upon conversion of cellulose I to cellulose II, the crystallinity index of the latter was also reduced. This indicates that the enhanced saccharification of the anhydrous cellulose II substrate is due to the reduction in crystallinity rather than due to the change in cellulose allomorph. It can be thus concluded that the order of saccharification efficiency of the macromolecular structures of cellulose is: amorphous cellulose > hydrated cellulose II > cellulose I > anhydrous cellulose II. The relative proportions of these cellulose structures in pretreated biomass solids will play a role in their enzyme saccharification performance Hemicelluloses Hemicelluloses comprise 20 % to 30 % of the dry mass of LCB. As opposed to cellulose, they are a collection of branched heteropolysaccharides with shorter chain lengths (maximum DP of about 200) and no crystalline structures. They consist of hexoses (e.g. D-glucose) and pentoses (e.g. D-xylose and L-arabinose) in addition to uronic acids and acetyl groups with the exact composition depending on the type of hemicellulose. The composition and structure of hemicelluloses differ characteristically between plant types (especially between hardwoods, softwoods and grasses) and tissue types [26]. Glucuronoarabinoxylan (GAX, Figure 2.2.3) is the predominant type of hemicellulose found in the grass family [26, 27]. It consists of a β-(1 4)-Dxylanopyranosyl backbone which is partially substituted at C-2 with 4-O-methyl-α- D-glucuronic acid (GlcA) (at ca. 2 xylose units out of every 10) and acetyl groups (at ca. 1.2 xylose units out of every 10) and at C-2 or C-3 with α-l-arabinofuranose units (at ca. 1.3 xylose units out of every 10). The glycosidic bonds of the xylose backbone and the arabinose side chains are easily hydrolysed by acids but resistant to alkali, whereas the uronic acid (GlcA) linkages with xylan are alkali labile and relatively resistant to acids [19, 26, 28]. The bonds with acetyl groups can be easily cleaved by 12

37 alkali treatment [11, 19]. Acetyl groups are more abundant in softwoods and hardwoods than they are in grasses. Figure 2.2.3: Molecular structure of glucuronoarabinoxylan Lignin Lignin accounts for 20 % to 30 % of LCBs dry mass. It is a hydrophobic cementing and insulating agent of the plant cell wall and it is deposited mainly in cell walls of supporting and water-conducting tissues. It is a phenolic polymer formed from the polymerisation of three monomer units, p-coumaryl, coniferyl and sinapyl alcohols (Figure 2.2.4). Lignins made up of these three monomers are called p-hydroxyphenyl (H), guaiacyl (G) and syringyl (S) lignins respectively [21, 29]. 13

38 γ OH OH OH 2 α 1 β O CH 3 H 3 C O O CH 3 OH OH OH p-coumaryl alcohol coniferyl alcohol sinapyl alcohol Figure 2.2.4: Lignin monomer units The lignin monomers form macromolecular structures via ether bonds (ca. 2/3 of monomer linkages) and carbon-carbon bonds [19]. The most common linkages and dimer structures in lignin are shown in Figure and a macromolecular structure of a hypothetical lignin of European beech is presented in Figure A study on the structure of sugarcane bagasse by Sun et al. [30] reports presence of all three types of phenylpropanoid units (H,G,S) linked to each other mainly via β-o-4 ether bonds, and carbon-carbon bonds such as β-β, 5-5 and β-5. The relative proportions of the individual phenylpropanoids contained in lignin (H,G,S) vary between plant species. Dorrestijn et al. [31] report that pyrolysis of grass lignins results in 45 % H, 39 % G and 16 % S phenylpropane units. However, Meier et al. [32] determined that for sugar cane bagasse derived pyrolysis oil the proportions of phenylpropanoids derivatives were 61 % H, 28 % G and 11 % S. Ruggiero et al. [33] showed that unbleached acidolysis bagasse lignin had proportions of phenylpropanoids of 56 % H, 37 % G and 7 % S whilst bleached acidolysis lignin had 50 % H, 44 % G and 6 % S. It would appear that even for LCB of the same species H:G:S ratios can vary, but for grasses in general the predominant monomer is H followed by G, while S is substantially lower. 14

39 Figure 2.2.5: The most common linkages between lignin phenylpropane units (from Sjostrom [19]) 15

40 Lignin contains phenolic hydroxyl, benzylic hydroxyl and carbonyl groups. Their frequency varies and from a processing point of view the relative frequency of these groups in extracted lignin, determines its potential for processing towards value-added products. Figure 2.2.6: Partial structure of a hypothetical lignin molecule from European beech (Fagus sylvatica) (from Nimz [34]) 16

41 2.2.4 Lignin-carbohydrate bonds Hemicelluloses bind covalently to lignin but not to cellulose. However, sufficient adhesion between cellulose and hemicelluloses is provided by hydrogen bonds and van der Waals forces [19]. The covalent bonds between hemicelluloses and lignin are reported to involve ester and ether bonds and they influence the reactivity of biomass when exposed to chemical processing. For example, ferulic (or coniferic) acid esters are known to make grass cell walls recalcitrant to enzymatic saccharification prior to fermentation to biofuels [35]. Some common types of lignin-hemicellulose linkages found in the cell walls of grasses are (depicted in Figure 2.2.7): Direct ester (e.g. uronic acid ester bonds formed by the attachment of the carboxyl group of the hemicellulose GlcA branching unit to phenolic hydroxyl sites in lignin [36].) Direct ether (e.g. benzyl-α-ether bonds formed between lignin and the O5 position of arabinofuranose in hemicelluloses [36-38].) Hydroxycinnamic acid ester (e.g. ester bonds formed by the attachment of carboxyl groups from lignin hydroxycinnamic acid to the primary alcohol hydroxyls of the arabinofuranose unit of hemicelluloses [27, 36].) The most common hydroxycinnamic acids encountered in lignin of grasses are ferulic acid and p-coumaric acid. Both acids participate in covalent linkages between lignin and hemicelluloses. p-coumaric acid is only known to form ester bonds, while ferulic acid forms both ester and ether bonds. In addition, ferulic acid can form dimeric (dehydrodiferulic) bridges between lignin and polysaccharides and between different polysaccharide chains. These varying bonding possibilities of hydroxycinnamic acids along with some direct ester and ether linkages are shown in Figure

42 Figure 2.2.7: Commonly occurring covalent linkages between GAX and lignin in grasses It is worth mentioning that the exact in situ bonding of lignin to polysaccharides is not yet fully elucidated [36, 38] and that structures presented here correspond to representations based mainly on ex situ characterisations of lignin. In addition, plant cell walls may vary in chemical structure depending on which tissue of the plant they pertain to (e.g. leaves, stems, young or old tissue), and on the environmental stress experienced by the plant during growth (e.g. drought, disease, mechanical stress by wind). Sun et al. [30] reported that the lignin-carbohydrate bonds in bagasse consist mainly of coumaric acid esters and ferulic acid ethers. 18

43 Figure 2.2.8: Possible covalent cross-links between polysaccharides and lignin in cell walls (from Iiyama et al. [36], Copyright American Society of Plant Biologists) Cellulose microfibrils: The foundation units of the cell wall construct Cellulose microfibrils are the rod-like foundation units of the cell wall structure. Native cellulose is a long polymer whose molecule stretches to a length of at least glucose units. Parallel cellulose molecules, held together by hydrogen bonds, form the smallest building element of the cellulose skeleton known as the microfibril [19]. The microfibrils wind together to form threads that coil around each other, like strands in a cable. Each cable forms the next size building unit 19

44 called a macrofibril (Figure 2.2.9). Cellulose molecules wound in this fashion have a tensile strength approaching that of steel ( kg mm -2 ) [29]. Cellulose has crystalline properties resulting from the orderly arrangement of cellulose molecules in microfibrils. This crystalline arrangement is restricted to parts of the microfibril known as micelles (Figure 2.2.9). Amorphous and disordered cellulose as well as hemicelluloses and lignin are located in the spaces between the microfibrils. Hemicelluloses are considered amorphous and lignin is both amorphous and isotropic [19]. A, strand of fibre cells. B, transverse section of fibre cells showing layering: a layer of primary and three layers of secondary wall. C, fragment from the middle layer of the secondary wall showing macrofibrils (white) and interfibrilar spaces (black). D, fragment of a macrofibril showing microfibrils. E, structure of microfibrils showing the long cellulose molecule which in some parts forms orderly micelles. F, fragment of a micelle, G, two glucose residues forming the repeating unit (cellobiose) of the cellulose polymer Figure 2.2.9: Detailed structure of cell walls (from Evert et al. [29]). 20

45 2.2.6 The cell wall layers The mature, lignified cell wall is organised in layers, namely middle lamella (ML), primary wall (P), outer layer of secondary wall (S1), middle layer of secondary wall (S2) and inner layer of the secondary wall (S3) (Figure ). These layers have distinct structure and composition. The microfibrils wind around the cell axis in different directions, either to the right (Z helix) or to the left (S helix) [19] in the direction of growth. S3 S2 ML S1 P Figure : Cell wall layers and organisation of the cellulose microfibrils (adapted from Raven et al. [39]). The middle lamella fills the intercellular spaces and binds the cells to each other. At maturity of the cell, this layer is predominantly composed of lignin. The primary wall is a thin layer consisting of cellulose, hemicelluloses, pectin and protein completely embedded in lignin. In the outer portion of this layer, the cellulose microfibrils form an irregular network, while in the interior, they are oriented nearly perpendicular to the cell axis. 21

46 The secondary cell wall represents most of the cell mass, especially its thick middle layer (S2). The three layers of the secondary wall are built of near-parallel microfibrils of cellulose between which lignin and hemicelluloses are intertwined. The orientation and angle of the helices formed by the microfibrils vary among the three layers. In the outer secondary wall (S1) the orientation of the helices is both Z and S and at angles to the long cell wall axis that are large and sometimes perpendicular. In the mid-layer (S2), the angle is small and the slope of the helix steep (close to vertical to the long axis). In the inner layer (S3), the microfibrils are deposited as in S1, at large angle to the long axis [29, 39] Mechanism of cell wall swelling Swelling is a natural phenomenon of cell walls which allows space for the lateral deposition of newly-formed fibre. In presence of solvents, swelling is the first step in the process of dissolution of cell walls and also occurs without dissolution. Swelling of biomass in ionic liquids without dissolution is covered in this thesis and therefore it is worth reviewing here. In the presence of reagents that induce swelling, the thick S2 layer swells laterally (unidirectional microfibril helices at steep angles) while the primary wall gets peeled off as the S1 layer around the fibre expands [19]. This combination of events results in the formation of balloon structures seen in Figure The complex structure of LCB is comprised of polysaccharides and lignin that are intricately intertwined inside the cell wall layers. The abundance of LCB and its rich polysaccharide content make this material an attractive carbon resource for manufacturing renewable fuels. However, its complex structure poses challenges to fuel manufacturing based on fermentation of the monosaccharide products of saccharification. Pretreatment is the chemical, mechanical and/or biological process by which the recalcitrant and inaccessible LCB structure becomes available for fractionation and further processing towards fuel and other added value products. Pretreatment is reviewed in the following section. 22

47 The ribbon like, unrolled primary wall (P) surrounds the swollen secondary wall. S1 is the swollen outer layer of the secondary wall, under which the microfibrils of the middle layer, nearly parallel to fibre axis, are dimly visible. S3 Figure : Light microscope image showing ballooning of a sulphate pulp fibre (Pinus silvestris) (from Illvessalo-Pfaffli [40]) 2.3 Pretreatment Overview of the conversion of biomass to ethanol fuel The conversion of the LCB polysaccharides to ethanol fuel and/or other products of fermentation (e.g. butanol) involves hydrolysis, fermentation and distillation. However, these polysaccharide molecules are not readily hydrolysed since they are contained in the chemically recalcitrant and structurally robust lignocellulosic matrix described in Section 2.2. Therefore, a pretreatment step is added to the process prior to hydrolysis in order to improve saccharification of the polysaccharides. The process steps involved in the conversion of LCB to ethanol are shown in Figure Pretreatment is the first step in this process and its goal is to open up the LCB structure. This entails structural modification of the material (e.g. reduce cellulose crystallinity and increase surface area) and chemical modification (e.g. minimise lignin content). This step is the subject of this thesis and will be reviewed in more detail in this section. 23

48 Since pretreatment is the first step in the process, it has to be designed to minimise inhibition towards downstream process steps, namely hydrolysis and fermentation. lignocellulosic biomass (LCB) PRETREATMENT 'opened up' LCB accessible polysaccharides HYDROLYSIS monosaccharides FERMENTATION ethanol Figure 2.3.1: Gross representation of the main steps in a biomass to ethanol process Hydrolysis can be catalysed by mineral acids or enzymes derived from cellulolytic fungi such as Trichoderma reesei. Although concentrated mineral acid hydrolysis is more technologically mature than enzymatic hydrolysis, the enzymatic processes are expected to have cost advantages as cellulase research advances. Moreover, acids are associated with greater environmental liabilities and the formation of fermentation inhibiting products such as furfurals [8]. With the exception of comminution prior to direct hydrolysis of LCB in concentrated acid, pretreatment is tailored to optimise the LCB substrate for enzyme rather than acid hydrolysis. 24

49 The saccharification of LCB after a number of pretreatments are compared in Section (after all these pretreatments have been discussed) Fermentation to ethanol is carried out by yeasts or bacteria. Yeast (Saccharomyces cerevisiae), has been traditionally used in the brewery industry to convert hexoses to ethanol. However, a substantial proportion of the sugars in most LCB substrates are pentoses (20 % to 30 %) and there is a need for a microorganism that can ferment these too. Microbiologists are still improving yeast and bacterial strains in order to achieve fermentation of both hexoses and pentoses to ethanol or butanol [8]. Both enzymatic hydrolysis and fermentation are biologically catalysed reactions and they are collectively known as the bioprocessing steps of the cellulosic ethanol process. Ideally, all bioprocessing steps would be carried out by a single system of organisms that would simultaneously exhibit the following properties: (a) synthesis of an active cellulose enzyme system at high levels, (b) fermentation and growth on sugars arising from both cellulose and hemicellulose, and (c) production of ethanol at high selectivity and high concentration. Unfortunately all compatible combinations of known microorganisms fall short of this ideal, on account of two main limitations: (a) an inability to utilize the range of carbohydrates present in biomass (e.g. cellulose, hemicellulose) while also producing ethanol at high yield or (b) differing requirements for oxygen for various functions essential to the process. For example S. cerevisiae, is unable to ferment pentose and cannot coexist with T. reesei because the latter requires oxygen for growth while the former requires low oxygen conditions for fermentation. Two approaches have been employed to overcome this incompatibility; viz. create recombinant organisms that are compatible, or carry out each bioprocess in separate reactors. These approaches including their intermediate forms are listed in Figure When all bioprocessing takes place in a single reactor, the process is referred to as consolidated bioprocessing (CBP), when the enzyme is produced in a separate incubator, the process is referred to as simultaneous saccharification and co fermentation (SSCF), when all four bioprocesses take place in different reactors, 25

50 the process is referred to as separate hydrolysis and fermentation (SHF) [8]. Although CBP is the most favoured approach in terms of less infrastructure requirements (single reactor), the biotechnology of compatible microorganisms is still at low maturity and less consolidated forms of bioprocessing are currently the most commonly employed options. The ethanol fuel is recovered from the fermentation broth by distillation. Generally, the unfermented solids (lignin, unreacted polysaccharide fractions and enzymes) accumulate at the bottom of the distillation column. These solids are dried and combusted for generation of thermal energy [8]. The type of pretreatment and extent of bioprocessing consolidation influence the composition of these unfermented solids. In biorefinery processes based on clean fractionation, lignin and hemicelluloses may be separated from the cellulose prior to saccharification, and the residue after distillation may be suitable for purposes other than combustion (e.g. animal feed). Processing strategy (each box represents a bioreactor not to scale) Biologically SHF SSF SSCF CBP mediated event (separate (simultaneous (simultaneous (consolidated hydrolysis and saccharification saccharification bioprocessing) fermentation) and fermentation) and co fermentation) Cellulose production Cellulose hydrolysis Fermentation of C6 sugars Fermentation of C5 sugars Figure 2.3.2: Consolidation of bioprocessing in cellulosic ethanol production (from Lynd [8]) 26

51 2.3.2 Goals of pretreatment The benefit of pretreating biomass prior to enzyme saccharification has long been recognised [14]. Unless a very large excess of enzyme is used (with consequent higher processing costs), the final cellulose conversion in native biomass is very low (< 20 % of theoretical), whereas with appropriate pretreatment, it can often reach 100 % of theoretical mass [13]. Optimising performance and reducing cost of pretreatment is a current research activity aimed at enhancing the commercialisation potential of cellulosic ethanol. The goal of pretreatment is to disrupt certain structural and chemical characteristics of native biomass that are thought to be responsible for its low enzyme saccharification rates. These include, the crystallinity of cellulose, the presence of lignin, the protection of cellulose by lignin and hemicellulose, and the small surface to volume ratio (porosity) of the material. Lignin in biomass undoubtedly interferes with enzymatic hydrolysis of glycosidic linkages. The phenolic polymer restricts physical access of the enzymes to cellulose [41] while it also provides sites to which cellulase enzymes adsorb unproductively and irreversibly [42]. Research shows that rates and extent of biomass saccharification increase with increasing lignin removal [11]. Hemicellulose removal is also reported to enhance enzyme saccharification of cellulose [43]. Hemicellulose removal increases internal surface area, provides more immediate access of enzymes to cellulose, and reduces unproductive binding of cellulases on hemicellulose sugars [10, 43, 44]. Acetyl ester linkages are detrimental to enzyme accessibility and the selective removal of acetyl groups has a positive effect on saccharification of LCB [10, 11, 43]. For example, Grohmann et al. [43] reported saccharification rate increase of 5-7 times for hemicellulose and 2-3 times for cellulose after selective removal of 75 % of acetyl groups of wheat straw following de-esterification with hydroxylamine solutions. Acetyl groups are not frequent on xylans of grasses and are therefore not of major influence to the reactions in this study. 27

52 Highly crystalline cellulose is less accessible to enzyme attack than amorphous cellulose [10, 11, 44]. Experiments on pure microcrystalline cellulose (e.g. Avicel) provide strong evidence that cellulose decrystallisation improves enzyme saccharification rate [44, 45]. Dadi et al. [45] produced amorphous cellulose using dissolution in ionic liquid and reported a 50-fold higher saccharification rate compared to crystalline cellulose. Jeoh et al. [44] decrystallised cellulose using a DMSO-paraformaldehyde technique that had been demonstrated to produce amorphous cellulose without affecting its DP. Their results showed a 15-fold enhancement of cellulase enzyme accessibility compared to crystalline cellulose. Holtzapple and co-workers [10, 11], have also demonstrated the positive effect of decrystallisation on the saccharification of LCB cellulose. Selective decrystallisation of LCB cellulose in these studies was achieved by ball milling. Work by Gharpuray et al. [46] demonstrated that ball milling reduced particle size and increased surface area and in the process produced less crystalline biomass. Therefore the effect of ball milling on saccharification rate was not due to decrystallisation alone. However, Holtzapple and co-workers supported the selectivity of their methodology by citing a preceding publication where it was demonstrated that further reduction of biomass particle size below 40-mesh (the authors starting material) did not enhance the saccharification rate [10, 11]. Lignin-hemicellulose covalent bonds are another source of recalcitrance to hydrolysis. For example, expression of ferulic esterase enzymes in tall fescue grass, improved enzyme saccharification of its cell walls by rumen fluid inocula by 10 % - 14 % compared to the control [35]. All aforementioned factors play a role in the enzyme saccharification of LCB. Their relative importance is hard to define due to the fact that selective chemical removal of one barrier usually affects the state of at least one other barrier. Chang and Holtzapple [11] appear to have successfully isolated the effects of three of these factors (namely delignification, deacetylation and decrystallisation of poplar wood) and reported their relative importance both on cellulose and hemicellulose saccharification. Their results indicated that among the three structural features, 28

53 lignin content and crystallinity had the greatest effects on saccharification of total polysaccharides. As compared to cellulose, hemicellulose saccharification was less affected by decrystallisation (since it is not crystalline) and more affected by delignification and deacetylation (since it is covalently linked to both lignin and acetyl groups). Zhu et al. [10] took this work further and investigated interrelations and relative importance of each characteristic at different stages of the saccharification reaction (1 h, 6 h and 72 h). As shown in Figure 2.3.3, this work demonstrated a causal relationship between enzymatic saccharification and enzyme adsorption extent (amount) and effectiveness, which are in turn related to the structural features of the LCB substrate. The amount of enzyme that binds to polysaccharides is reflected in the extent of saccharification whilst the enzyme s effectiveness plays a role at the initial saccharification rates. The amount of enzyme that binds was increased with increasing lignin content and to a lesser extent with increasing acetyl content, while the effectiveness was correlated with the cellulose crystallinity of biomass. These conclusions can aid the interpretation of enzyme hydrolysis results from various pretreatment-substrate combinations. CRYSTALLINITY EFFECTIVENESS 1-h Saccharification extent ACETYL 6-h Saccharification extent AMOUNT LIGNIN Structural features of LCB Enzyme adsorption Thicker arrows indicate a more significant effect 72-h Saccharification extent Enzyme saccharification extent Figure 2.3.3: The effects of lignin, acetyl groups, and crystallinity on enzyme adsorption and enzymatic hydrolysis of biomass (reproduced from Zhu et al.[10]) 29

54 Apart from the aforementioned pretreatment aim of increasing saccharification, an efficient pretreatment technology has to take into consideration a number of variables related to the processes general viability. These include: the potential for added value co-products (e.g. added value lignin coproduct) release of inhibitory by-products to downstream processing losses of carbohydrate raw material energy usage (heat, mechanical) resource usage (water) use of reagents that are expensive or toxic to humans and the environment the cost of biomass and its appropriateness for the pretreatment overall capital and operation costs the contribution to life cycle impact factors These criteria as well as the saccharification rates are the basis on which different pretreatment technologies should be compared to each other. Figure schematically represents the relation of such criteria to the inputs and outputs of a conventional pretreatment process that does not result in clean fractionation of biomass components Pretreatment technologies Pretreatment is usually physical or chemical or a combination of the two. At least to date, biological pretreatments using microorganisms (e.g. white rot fungi) have been too slow (order of weeks) and therefore not favoured [14, 47]. Physical pretreatments involve the application of mechanical force and hot water or steam. They exclude the use of any additive chemicals. These treatments increase the surface area of the substrate while they can also solubilise part of the non-cellulosic fraction of the biomass structure. 30

55 INPUTS OUTPUTS Pretreatment Additives Minimise, recycle and avoid toxicity Biomass Low cost + fit for pretreatment Energy (heat, mechanical) Minimise P R E T R E A T M E N T Vapour stream Minimise vapour + avoid loss of carbohydrate or reagent Pretreated solid Maximise saccharification Liquid stream Minimise inhibitors and carbohydrate losses Combustion of residue Heat and power Lignin fraction Added value co-product White: Inputs and outputs. Grey: the corresponding criteria/targets associated with each input and output. Light grey: enzyme saccharification as the central criterion Figure 2.3.4: The participating inputs and outputs in a pretreatment process Mechanical comminution reduces particle size, increases surface area and in the case of ball milling also disrupts cellulose crystallinity. However the energy usage of comminution is high and increases exponentially with decreasing particle size. Therefore comminution is usually limited to coarse milling before it starts becoming prohibitively energy-intensive. This coarse milling is sometimes needed to reduce the size of material that is destined for chemical pretreatment [13]. Steam explosion (STEX) is a physical treatment very commonly used for the pretreatment of biomass. In this method, biomass is impregnated with high pressure saturated steam (160 C to 260 C for 1 min to 10 min) and then released to explosively decompress to atmospheric pressure. It removes some hemicelluloses and lignin while it increases the surface area of the STEX solid LCB substrate. Since water acts as an acid at high temperatures, some of the characteristics of STEX pretreated biomass resemble those of acid treated. In fact, 31

56 addition of acid in steam explosion increases this resemblance; nearly all hemicellulose is removed and more sugar degradation is incurred [13, 14]. Chemical treatments are reactions that use aqueous acid or alkali solutions at elevated pressures and temperatures. It is generally understood that acid pretreatments remove hemicelluloses, alkali pretreatments remove lignin and both increase the internal surface area of the substrate. Dilute acid hydrolysis is the most widely studied pretreatment. Dilute H 2 SO 4 has been used to commercially produce furfural from cellulosic materials [48]. Dilute acid treatment at concentrations of H 2 SO 4 below 4 % and at high temperature (160 C to 190 C ) and pressure for about 10 min appear very effective for quantitative removal of hemicelluloses and render the rest of the LCB more digestible for enzymes [13, 14]. Fermentation inhibitors such as furfurals are a byproduct of dilute acid pretreatment and should preferably be extracted from the liquid stream destined for fermentation [49, 50]. Alkaline pretreatment processes utilize lower temperatures and pressures compared to other pretreatment technologies. Lime (calcium hydroxide) pretreatment is carried out at low temperatures (100 C to 150 C) but depending on temperature may require a long residence time (typically 2 h to 12 h but up to a number of days) [11, 14, 51, 52]. Kim et al. [52] demonstrated that lime treatment deacetylates and delignifies the LCB substrate. Deacetylation of 90 % was achieved regardless of temperature or reaction conditions. Delignification extent increased with increasing temperature and in the presence of oxygen [52]. Ammonia fibre explosion (AFEX) is similar to STEX although the chemistry involved differs. In a typical treatment, equal masses of liquid ammonia and dry biomass are heated under pressure to 90 C for 5 min to 30 min and then the pressure is suddenly released [14, 53]. AFEX increases the surface area and interferes with internal bonding of the LCB without removing substantial mass from the substrate [53]. According to Teymouri et al. [53] AFEX can also reduce cellulose crystallinity although they provide no direct evidence or measurement of 32

57 decrystallisation. Kumar et al. [54] provide evidence that AFEX can effect a slight decrystallisation on corn stover although not on poplar. A comparative disadvantage of AFEX is reported to be the slow saccharification kinetics in substrates of high lignin content (e.g. aspen wood chips with 25 % lignin, only reached 50 % saccharification extent after AFEX) [14]. The organosolv processes are based on pulping technologies that use combinations of solvents to remove lignin. For example, in an organosolv process using an organic solvent (e.g. ethanol) mixed with an aqueous solution of dilute acid (e.g. H 2 SO 4 ) at high temperature and pressure (180 C for 60 min), the organic solvent acts as a delignifying agent while the acid aids in the removal of hemicelluloses. The resulting solid material is a soft cellulose pulp of low lignin and hemicelluloses content. Pan et al. [55] applied the organosolvation process to the conversion of poplar to ethanol. The process resulted in the fractionation of poplar chips into a cellulose-rich solids fraction; an ethanol organosolv lignin (EOL) fraction; and a water soluble fraction containing hemicellulosic sugars, sugar breakdown products, degraded lignin and other components. The above examples cover most chemical pretreatments proposed to date. The underlying chemical mechanisms are predominantly governed by the ph of the medium. Even in organosolv or hot water treatments with no added acid, organic acids (principally acetic acid) released from LCB upon treatment result in acid like processes (known as autohydrolysis). However there are numerous pretreatment variations based on choice of equipment. One variation is the use of continuous percolation reactors (e.g. flow-through reactor), which press the solvent through a biomass cake at high temperatures (usually around 160 C) and recycle it back in the process in a continuous mode. Examples include ammonia recycle percolation (ARP), hot water and dilute acid percolation, where biomass is processed with liquid ammonia, water or dilute acid, respectively [13]. Most pretreatment technologies require expensive, pressure-rated equipment, have high energy requirements and use corrosive or volatile chemicals [13]. Acids or bases and organic solvents at high temperatures and pressures create 33

58 conditions that are corrosive to common stainless steel industrial equipment. Many of these processes have associated workplace health and safety issues which also may increase costs. In addition, acidic or alkali output streams need to be neutralized prior to enzyme saccharification, producing mineral salts that are difficult to recycle. Notwithstanding these problems, none of the above pretreatments discussed here demonstrate much ability to disrupt the crystallinity of cellulose [54]. While the crystallinity is somewhat reduced (e.g. AFEX or ball milling), these outcomes are secondary to the intended primary effect. Pretreatments that result in solid materials with retained cellulose crystallinity require high enzyme loads and long saccharification times to effect complete saccharification [10]. Such processes have capital and operating costs that are still too high for lignocellulosic ethanol to be competitive with petroleum [56] Conventional cellulose solvents The use of cellulose solvents is another possible pretreatment. Among the cellulose solvents are concentrated mineral acids which depolymerise cellulose and those solvents which dissolve cellulose in polymeric form. Dissolving the LCB in a polymeric form provides the opportunity for clean fractionation of the dissolved biomass molecules and the precipitation of cellulose in a decrystallised form. Cellulose solvents disrupt the crystalline order of native cellulose [20, 44, 45]. Concentrated acid is a long known means of chemically decrystallising cellulose [58]. Concentrated acid treatment of cellulosic materials has been practiced in various forms during times of fuel shortages [59] and more recently is the subject of a patent by DuPont & Co [60] and other patents [61, 62]. However, large scale industrial applications are still limited by the corrosivity, safety risk, high water usage and disposal problems associated with concentrated acid. Combinations of aprotic solvents (e.g. dimethylsulphide or dimethylacetamide) and metal salts (e.g. LiCl or FeCl 3 ) are cellulose solvating systems often used for laboratory scale preparation of amorphous cellulose [20, 44]. Generally their industrial applications are limited to products of higher value than fuels due to the volatility, toxicity and cost of these solvents. 34

59 In fact, cellulose is a versatile starting material for chemical conversion to renewable/biocompatible films, fibres and packaging materials. Dissolution of cellulose in polymeric form is an essential step prior to such conversion. Since its crystalline structure does not facilitate chemical interaction with solvents, dissolution has been a challenging and long standing goal in research for cellulosebased artificial polymers. Until about 1950, only cuprammonium was well-known and widely used as a solvent for cellulose. Ten years later, the discovery of solvents based on transition metals broadened the spectrum of cellulose solvents. Since then a large number of organic solvents have been added to this list [20]. Figure gives an overview of the conventional cellulose dissolving systems. Depending on the interaction of the solvent with polysaccharide these solvents are classified as derivatising and nonderivatising. Derivatising solvents covalently interact with cellulose to form unstable intermediates such as cellulose esters, ethers or acetates. Non-derivatising solvents have Coulombic interactions only with the substrate and therefore there is no formation of intermediates and probably limited covalent bond cleavage. Non-derivatising solvents are more versatile in terms of further processing of the dissolved cellulose and are more relevant to pretreatment applications. These solvents are systematically divided into a number of subcategories comprising varying combinations of polar organic solvents, inorganic salts, transition metals and amino groups. The most relevant to practical uses is the subcategory that takes advantage of the strong intermolecular interaction between the polymer and some dipolar aprotic organic compounds with N-O or C=O dipoles. These solvents can be subdivided into the two groups, viz. salt-free and saltcontaining systems. N-Methylmorpholine-N-oxide (NMMO) is representative of a salt-free solvent and dimethylacetamide / LiCl (DMA / LiCl) is a commonly used example of a salt containing system. 35

60 O O N O NMMO (N-Methylmorpholine-N-oxide) H O S O H O mineral acids (sulfuric acid) OH Na Sodium hydroxide (NaOH) O Li N Cl DMA / LiCl (Dimethylacetamide / LiCl) NH O O O N N 2 O 4 /DMF (dinitrogen tetroxide/dimethylformamide) N O O O S N F Cl Li O H N N H Cl ClO 4 SCN Zn Li Li Cl DMSO / TBAF (dimethylsulfoxide / tetrabutylammonium fluoride) Dimethylimidazolone / LiCl molten salt hydrates Figure 2.3.5: Conventional cellulose solvents (reproduced from Pinkert et al. [57]) Solvent systems with molecular interactions or bonds of high dipole moment (of near ionic character) and strongly electronegative anions (i.e. Cl - and F - )seem to be common characteristics among the solvents listed in Figure In that regard, Spange et al. [63] have demonstrated that the chloride ion contributes about 80 % of the dipole-dipole interactions between DMA and cellulose in DMA/LiCl solvation systems and is primarily responsible for hydrogen bond disruption in cellulose and consequent dissolution. NMMO and its monohydrate form is the solvent of choice in the process used for manufacturing the cellulosic apparel fibre known as Lyocell on a technical scale of about 100,000 tonne per annum [64]. The solvation power of NMMO is due 36

61 to its ability to disrupt hydrogen bonds. Dissolution is facilitated by acid-base (donor-acceptor) interactions resulting in disruption and restructuring of the hydrogen bond network of native cellulose. NMMO is a weak base (pk B = 9.25) and its most prominent feature is the highly polar N-O group with a dipole moment of 4.38 D [65]. This dipole is symbolized either as ionic (with positive charge on the nitrogen and negative on the oxygen) or as donative with an arrow pointing at the oxygen (see Figure 2.3.6). NMMO was originally introduced as a cheaper, faster and more environmentally benign alternative to the Viscose process. However the NMMO treatment causes severe fibrillation of fibres, while the Viscose process based on NaOH and carbon disulphide, produces fibres with properties similar to cotton. Moreover, the use of NMMO a thermally unstable solvent also requires a major investment in safety technology. Consequently the Viscose process is still used in the manufacture of ca. 95 % of modified cellulose fibre [66]. Cellulose O H O N O Figure 2.3.6: Gross schematic of the hydrogen bonding formed between NMMO and cellulose hydroxyls upon dissolution In the case of salt-containing systems, a direct complexation between the cation and the cellulosic hydroxyl group is assumed. This interaction is facilitated by the participation of the polar organic medium in which these solvations take place. DMA / LiCl is the most widely used system for the dissolution of cellulose for analytical purposes. After preactivation, even high molecular weight cellulose can be dissolved without residue and detectable chain degradation [20]. Cellulose dissolution without covalent derivatisation can be generally viewed as an electron donor-acceptor (EDA) interaction where the amphoteric cellulose takes the role of either donor or acceptor or both, depending on the solvent structure in hand [20]. This generic model is presented schematically in Figure

62 Figure 2.3.7: EDA interactions between cellulose and a non-derivatising solvent (e.g. NMMO) According to Cuissinat and Navard [67], dissolution of cellulose microfibrils, either as cotton or wood fibres, follows specific patterns upon dissolution. These patterns are governed by the efficiency of the solvent and the orientation of cellulose chains in fibres. Highly efficient solvents disrupt the hydrogen bonding network as fast as they penetrate the fibre (fast dissolution by disintegration of rodlike fragments). Less efficient ones penetrate faster than they dissolve, leading to the formation of swollen balloon structures that eventually burst and dissolve. Poor solvents penetrate only without disrupting any of the H-bond network (swelling with no dissolution). Finally non-solvents are unable to cause either swelling or dissolution. These dissolution patterns are relevant to pretreatment since both swelling and dissolution enhance enzyme access to crystalline cellulose ([68] and own data) Enzyme saccharification of cellulosics after pretreatments The enzyme saccharification yields following various pretreatments are hard to compare directly since the substrates, the conditions and the enzyme properties vary among studies in the literature. The Biomass Refining Consortium for Applied 38

63 Fundamentals and Innovation (CAFI) [69] was the first attempt to compare sugar recovery data from different biomass pretreatments. The laboratories that participated assessed enzyme saccharification and total sugar recovery for a number of pretreatments using identical protocols. Although this initiative provided a useful single source of comparison for pretreatment technologies, the reaction conditions among the compared pretreatments still varied considerably. For example pretreatment residence times varied from 5 min for AFEX to 4 weeks for lime treatment. Some pretreatments perform better at initial saccharification rates while others may be slower but achieve a higher final saccharification extent. In Table 2.3.1, a selection of studied pretreatments is compared for saccharification yields both at the early stages (24 h) and at end of reaction ( 48 h). Table 2.3.1: Enzymatic saccharification from selected pretreatment systems Pretreatment 24h Saccharification (% cellulose 1 ) Final extent of saccharification ( 48 h) (% cellulose 1 ) Substrate Enzyme loading (FPU/g cellulose 1 ) Reference AFEX Corn stover 15 Teymouri et al. [70] Dilute acid Idem Idem Lloyd et al. [71] Hot water (flow through) ND 96 Idem Idem Kim et al. [72] Lime Idem Idem Kim et al. [73] Organosolv Hybrid poplar 20 Pan et al. [74] Phosphoric acid (84%) Corn stover 15 Zhu et al. [75] Ionic liquid [C2mim]OAc Li et al. [76] Switchgrass 50 mg protein /g cellulose Conventional cellulose solvents (e.g. phosphoric acid) seem to yield near 100 % cellulose conversion at 24 h of exposure to enzymes. This is largely attributed to the complete decrystallisation of cellulose. Dilute acid achieves about 80 % at 24 h, 1 cellulose = mass of cellulose recovered after pretreatment 39

64 but the highly crystalline cellulose and high content of lignin do not permit higher yields. Lime on the other hand, which removes lignin, reaches near 100 % final cellulose conversion. Organosolv in presence of acid achieves both delignification and hemicelluloses-acetyl removal and approaches the performance of cellulose solvents. Ionic liquids are a new class of solvents that can be used for LCB pretreatment. Preliminary reports on IL pretreatment show both high rates and extents of saccharification (see Table 2.3.1) while the ability of the IL properties to be tuned offers the potential of optimising dissolution performance and minimising the aforementioned solvent-related problems. LCB pretreatments based on ionic liquids are the subject of investigation in this thesis and current knowledge of this is reviewed in the following section. 2.4 Ionic liquid based pretreatment technologies Ionic liquids: properties and history Ionic liquids are low-melting salts (< 100 C), which form liquids that consist of cations and anions only. Characteristics that contribute to the low melting points are large ions with low symmetry and delocalized charge [77]. Ions can be inorganic or organic ions, often featuring an aromatic or cyclic structure and long alkyl chains. Many ILs have negligible vapour pressure, are non-flammable and can be designed to have high thermal stability [78, 79] and low toxicity [80, 81]. The first reporting of an IL dates back to the mid 19 th century. Chemists performing an AlCl 3 -catalysed Friedel-Crafts alkylation observed the formation of a red oil. With the advent of nuclear magnetic resonance (NMR) techniques, this oil was later identified as a stable intermediate comprised of a carbocation and a tetrachloroaluminate anion [82]. 40

65 Interest in major practical applications of ILs began in the 1960 s, when the US Air Force Academy studied low-melting salts as alternative electrolytes for thermal batteries. These salts were binary mixtures of 1-butylpyridinium chlorides and aluminium chlorides. First generation ILs suffered from easy electrochemical reduction and air sensitivity. In the 1990 s, second generation ILs were designed to overcome these problems. For example, in 1992, Wilkes and Zaworotko [83] prepared and characterised a series of air and water stable low melting salts based upon the 1-ethyl-3-methylimidazolium cation. These stable salts sparked new interest in IL research and since then the number of ILs prepared and the related publications and applications have grown considerably. Examples of common ions that form stable ILs are listed in Figure Figure 2.4.1: Common ions in ionic liquids One of the main reasons for rapidly growing research interest in ILs is the ability of their properties to be tuned. To the best estimate of Holbrey and Seddon [84] the number of accessible IL anion and cation combinations equate to about one trillion. Properties such as melting temperature, conductivity, refractive index, 41

66 thermal stability, acid-base character, toxicity, hydrophilicity, polarity, density and viscosity can be tailored to a certain degree [57, 85]. This ability of ILs to be tuned makes them attractive in applications beyond the electrolytes that spurred their discovery (e.g. as solvents in industrial applications). Aside from tunability, ILs offer a variety of physical properties that make them attractive alternatives to other solvents. Ionic liquids involving fully quaternised nitrogen cations are non-flammable and have very low or negligible vapour pressure. This means reduced risk of explosion or fire accompanied by reduced need for respiratory protection and exhaust systems. ILs of the imidazolium cation type have demonstrated ability to dissolve a wide range of organic and inorganic compounds including cellulose and biomass [57, 82, 85]. This facilitates the formation of homogeneous solutions of disparate reagents, reactants and products [85]. The large liquidus range of ILs is another attractive characteristic when compared to conventional solvents. ILs maintain fluidity and volume in a range as great as 300 C [85]. Most conventional solvents would freeze or boil across such a large temperature range. Due to the wide thermal stability, many more chemical processes may be undertaken in these solvents. The thermal stability also affords strategies for recovery of the solvents. Replacing conventional solvents with ILs has the potential to enable safer, more stable and more efficient chemical processes. This potential has gained ILs a central place in a new field of chemistry known as green chemistry. Green chemistry is the term coined to describe the recent international efforts in science and policy to prioritise sustainability in chemical processes. The Montreal Protocol which aims at phasing out the use of ozone-depleting substances is an example of the policy aspect of such international efforts [84]. The BASIL (Biphasic acid scavenging using ionic liquids) process, established by BASF in 2002 is based on IL, and forms an example of a greener industrial chemical process [86]. Among the many areas where ionic liquids are investigated as replacements for solvent applications is the dissolution of cellulose [57]. Interest in the dissolution of cellulose was initially aimed at improving chemical functionalisation and fibre 42

67 spinning for the polymer and textile industry [20]. Recently ILs have also attracted attention as a pretreatment solvent for lignocellulosic ethanol. This is due to their ability to solvate cellulose and LCB, their favourable physical characteristics, their facile recyclability and their tunability [85, 87, 88] Cellulose dissolution using ionic liquids Molten salts have been known to dissolve cellulose since the 1930 s. However, it wasn t until the turn of the century that cellulose dissolution attracted extensive research interest. Graenacher [89] filed a patent in 1934 claiming that molten salts (e.g. benzylpyridinium chloride) can readily dissolve cellulose. In 2002 (after the discovery of stable and low-melting ILs [82, 83]) Rogers and co-workers [90, 91] reported using melt salts as non-derivatising solvents for cellulose and demonstrated that 10 % to 25 % mass cellulose in IL solutions were achievable by heating at 100 C or by short pulses of microwave heating. This result, only attainable with [C4mim]Cl, was attributed to high chloride content of the solvent. They speculated that chlorides interacted readily with the cellulose hydroxyls via hydrogen bonding and this would be similar to the mechanism and role of Cl - in DMA / LiCl. The list of cation-anion combinations examined in this first study was limited. Especially in terms of cations, only alkylimidazolium salts were used. Since then, but predominantly in the last three years, more than 40 ILs [87] have been tested on cellulose and biomass a IL structure and cellulose dissolution Hydrogen bond basicity (β, a Kamler-Taft solvation parameter) and strong dipole moments of the IL have been reported as pivotal for the performance of ILs as both cellulose and biomass solvents [87, 92, 93]. Conventional non-aqueous cellulose solvent systems such as DMA / LiCl exhibit notably high hydrogen bond basicity and high polarity. Not surprisingly the same is observed in cellulose solvating ionic liquids. [C4mim]Cl, the most cited IL for cellulose dissolution, is a highly polar IL with a high β value [94] and [Allylmim]formate solvated higher amounts of cellulose than [Allylmim]Cl due to the hydrogen bond basicity of the former being 1.2-fold that of the latter [87, 93]. 43

68 The effect of dialkyl imidazolium cation structure on cellulose solubility has been systematically studied. Generally, cellulose was found to be soluble in chloride salts of imidazolium with ethyl butyl and allyl side chains with decreasing solubility as alkyl chain length increases. It was also observed that even numbers of carbon atoms show higher cellulose dissolution in the series C 2 to C 20 as compared to odd numbers [95]. This unexpected and unexplained outcome was later corroborated by Vitz et al. [96]. The same authors also demonstrated that this pattern was no longer observed when the chloride anions were replaced with bromides. A clear and viscous solution of 14.5 % cellulose in [Allylmim]Cl was achieved at 80 C in a little more than 30 min by Zhang et al. [93]. Generally [Allylmim]Cl outperformed [C4mim]Cl in cellulose dissolution experiments [97]. This could be due to the low viscosity of [Allylmim]Cl, attributed to the double bond on its side chain. Low viscosity allows IL ion mobility and thus increases cellulose swelling rate and enhances dissolution. As alternatives to the corrosive and viscosity-inducing halides, halide-free ILs with high hydrogen bond basicity have been successfully employed. For example, Fukaya et al. [98] found that the low viscosity [Allylmim]formate dissolves ca. 20 % mass of microcrystalline cellulose (DP ca. 250) at 80 C, whereas [Allylmim]Cl dissolved only ca. 2 % under the same conditions. The same team continued the investigation for halide-free ILs by synthesizing and testing some [C2mim] + ILs with varying phosphonate anions [99]. Their results indicated that these anions where exceptionally good cellulose solvents at very mild temperatures (i.e. 45 C). All ILs tested had high dipolarity, high hydrogen bond basicity and low viscosity to which their success was attributed. [C2mim]methylphosphonate dissolved cellulose at high concentrations (10 % mass) at 45 C and 30 min and could also dissolve lower concentrations (2 % to 4 % mass) at room temperature. The closely analogous [C2mim]diethylphosphate was reported as a good solvent since cellulose degradation during dissolution was low [96]. Recently, 1-ethyl-3-methylimidazolium acetate has received considerable attention as a solvent for both cellulose and lignocelluloses [66, ]. Its low 44

69 toxicity, relative compatibility with cellulose enzymes and high solvating capacity are its most prominent attributes. However, it has been demonstrated that [C2mim]OAc does not act solely as a solvent but also covalently interacts with the reducing ends of cellulose chains. Kohler et al. [103] investigated the interactions of IL cations using model cellulose oligomers. On the basis of 13 C NMR studies he reported that [C2mim]OAc forms a carbon-carbon bond between C-1 of glucose and C-2 of the imidazolium ring (see Figure 2.4.2). This chemistry was further confirmed by means of 13 C-isotopic labelling experiments carried out by Ebner et al. [104]. Surprisingly these imidazole glycosides do not form with [C2mim]Cl [103], which leads to speculation that either the glycoside formation is catalysed by the basicity of the acetate anion, or the stronger ion pairing network that chlorides form with the imidazole ring prevents covalent bond formation [103]. O H 3 C N O CH 3 OH OH N HO O OH O HO OH O O HO OH H OH OH CH 3 OH Figure 2.4.2: Structure proposed for a covalent binding of [C2mim]OAc to a cellooligomer (DP 6-10) (From Kohler et al. [103]) It would appear from the literature that IL characteristics favourable for LCB and cellulose dissolution include low viscosity, small polarising ions and high hydrogen bond basicity. In an industrial setting the characteristics of low toxicity, corrosivity and cost should be added. One of the advantages of ILs as solvents is that anions and cations can be chosen and modified (i.e. by changing alkyl side chain lengths and ring structures) to obtain these characteristics. This tuning of IL structure and function is the subject of current research activity. Extensive research is under way to identify the characteristics responsible for dissolution of cellulose 45

70 and for enhanced biomass dissolution in ILs [87]. These characteristics will aid synthetic chemistry in tuning ILs b Mechanism The mechanism of cellulose dissolution in ILs is not fully understood, but there is evidence to suggest that both the cation and the anion take part in the hydrogen bond disruption of the cellulose chains. 13 C NMR and 35/37 Cl NMR relaxation studies indicated that there is a 1:1 stoichiometric interaction between the chloride ions in [C4mim]Cl and the hydroxyl groups in cellulose [105]. Electrondonor-acceptor (EDA) interactions between the IL anion and the cellulose hydroxyl hydrogens and between the IL cation and cellulose hydroxyl oxygens have been proposed [93, 106, 107] (see Figure with [C4mim]Cl as an example). [C4mim] + Cl - [C4mim] + + [C4mim] Figure 2.4.3: Proposed dissolution mechanism of cellulose in [C4mim]Cl This model is based on the generic model of polar cellulose solvents as seen in Figure It has been shown that the hydroxyls participating in this interaction are primarily the C-6 and C-3 [108] (viz. the same hydroxyls that are responsible for the interchain bonding in the cellulose I allomorph. Zhang et al. [107] have employed 13 C NMR to probe further into these EDA interactions for [C2mim]OAc and cellulose. They suggested that the acetate anion favours the formation of hydrogen bonds with hydrogen atoms of hydroxyls, and the aromatic protons in the bulky imidazolium cation especially the most acidic proton at the C-2 position, prefer to associate with the oxygen atoms of hydroxyls with less steric hindrance. Furthermore, Zhang et al. [107] estimated the stoichiometric ratio of [C2mim]OAc : hydroxyl groups to be between 3:4 and 1:1 in the primary solvation shell, suggesting 46

71 that, at least for [C2mim]OAc, there is a possibility that the imidazolium cation forms some hydrogen bonds with the saccharides. Regardless of stoichiometry of these interactions, it would appear that the solvents interact with the hydroxyl groups that are involved in hydrogen bonding in crystal structures. The acetate anion of [C2mim]OAc has also been observed to participate in covalent bonding with cellulose. Kohler et al. [109] have observed that while attempting to form feruoyl and triphenyl ethers of cellulose in [C2mim]OAc solution, unexpected acetylation was taking place instead. At the same time some conversion of [C2mim]OAc to [C2mim]Cl was reported. These outcomes suggest that the [C2mim]OAc did not act purely as a solvent since its acetate ion was being consumed. [C2mim]OAc has been used in numerous experiments for cellulose and LCB dissolution [68, 76, 101, 110] but there are no reports on whether or to what extent the solvent is being consumed c Effect of reaction conditions Apart from the IL properties, a number of reaction conditions have been identified which influence the rate of cellulose dissolution in ILs. Microwave irradiation has been shown to substantially accelerate the dissolution rates [91, 96, ]. This is not surprising since ILs are polar molecules which absorb microwave energy directly and this internal heating is more effective than heat transfer based heating. However, care must be taken not to overheat and pyrolyse the cellulose [106]. Mikkola et al. [97] reported that upon use of high-power ultrasound, the dissolution rate increased and complete dissolution was achieved in a matter of few minutes in [C4mim]Cl and especially in [Allylmim]Cl [97]. However, Rogers and co-workers [91] reported no significant benefit from utilising sonication. Elevated pressures between 0.2 and 0.9 MPa can assist dissolution [111]. It has also been reported that the use of polar co-solvents may limit the solubility of cellulose. For example, use of DMSO in [C4mim]Cl(3: 1 mixture of [C4mim]Cl: DMSO) slowed down the dissolution rate when compared to pure [C4mim]Cl by decreasing the ionic strength of the solvent system [114]. 47

72 Water is known to affect the physicochemical properties of ILs [115] and in the case of cellulose dissolution it plays a dual role. At low concentrations it can prevent formation of degradation products (e.g. furfurals) and at high concentrations it competes with the IL for hydrogen bond sites resulting in decreased solvation which may completely prevent dissolution. In the case where cellulose is dissolved in dry IL, the addition of water precipitates the cellulose. The general convention is that ILs are sensitive to low concentrations of water [96, 116]. According to Rogers and co-workers [91], 1 % mass water in IL is enough to limit the solubility of cellulose in [C4mim]Cl at 100 C. However, [C2mim]Cl / cellulose / acid and [C2mim]Cl / LCB / acid solutions studied by Vanoye et al. [117] and by Binder and Raines [118] respectively were found to tolerate 5 % water without compromising dissolution. The higher effective concentration of chloride ions in [C2mim]Cl (cf. [C4mim]Cl) and the presence of acid in the latter examples may be the reason for their higher water tolerance d Properties of cellulose precipitated from IL solutions Cellulose dissolved in ILs can be precipitated out of the solution with the use of an antisolvent such as ethanol, methanol or water [45, 119]. This precipitated cellulose generally exhibits a lower degree of crystallinity than native cellulose [101]. Similar impact on crystallinity is observed for conventional cellulose solvents [20]. The precipitated celluloses from ionic liquid dissolution generally retained 25 % 42 % of native cellulose crystallinity [87]. Decrystallisation is known to greatly enhance cellulose saccharification by cellulase enzymes[10, 11, 25]. Dadi et al. [45], have reported 50-fold enhancement of initial saccharification rates after treatment of Avicel cellulose with [C4mim]Cl. Liu et al. [120] reported the effect of ionic liquid treatment on enzymatic saccharification of whole biomass. Wheat straw treated with microwave heated [C4mim]Cl, rendered the straw more digestible to enzymes. This effect of ILs on biomass created a new niche for IL applications as biomass pretreatment systems for the manufacturing of fermentable sugars. The above review of cellulose dissolution in ILs demonstrates the influence of the ionic liquid characteristics on cellulose solubility and the importance of 48

73 carefully selecting the structure of the anion, cation and reaction conditions in order to achieve efficient cellulose dissolution. It also explains the role of ILs in pretreatment research since cellulose precipitated from IL solutions is expected to exhibit high enzyme saccharification rates Lignin dissolution in ionic liquids About 26 million tonnes of lignin are manufactured annually as a by-product of the environmentally harsh Kraft pulping process. This lignin is thiolated and mainly used as combustion fuel. Since it has a relatively high initial water content, its fuel value is low, producing less than ¼ as much energy for an equivalent mass as middle distillate (diesel, jet and boiler) fuels [121]. Moreover, it has been demonstrated that native lignin (i.e. lignin that is not thiolated or sulphonated) can be converted to value-added products such as adhesives, coatings, polymer blends, and carbon fibre composites [ ]. In this study, lignin is viewed as a valuable product stream. If lignin is to be functionalised, blended, polymerised, or sold for its antioxidant properties, a high reactivity and low degree of condensation are the sought-after characteristics [124, 125]. Capturing these high value product streams is the intention of the lignin fractionation experiments reported in this thesis. The potential of IL treatment to extract and fractionate lignin with preservation of native structure is explored here. A number of ILs have been tested for their ability to dissolve lignin: imidazolium salts containing methyl, ethyl, allyl, butyl, hexyl and benzyl groups in the imidazolium ring and with a number of common anions, such as chloride, bromide, tetrafluoroborate, acetate, trifluoromethanesulfonate and methylsulfate [68, 126]. Although it is challenging to compare results for lignin solubility across studies, some general observations can be made. Bearing in mind that only imidazolium cations were screened, the anions appear to influence lignin solubility - the most. In imidazolium salt based ILs, large non-coordinating anions, such as BF 4 and PF - 6 as well as bromide were not good lignin dissolving solvents. In order of preference, the anions methylsulfate, acetate and chloride imparted good solubility. The effect of cation is not insignificant, for example, [Allylmim]Cl outperforms 49

74 [C4mim]Cl. This is possibly due to the π electrons of the allyl group interacting with the phenolic π electrons of lignin [17, 68]. Lee et al. [68] compared a few combinations of ions for their ability to dissolve isolated lignin and wood flour. The highest lignin solubility was obtained using [C1mim] MeSO 4 and [C4mim]CF 3 SO 3 ; solvents that do not result in appreciable solubility of wood flour. At the other end of the spectrum, chloride anions enabled relatively high wood flour solubility (10 g kg -1 to 30 g kg -1 ) while retaining > 100 g kg -1 lignin solubility. [C4mim][BF 4 ] and [C4mim][PF 6 ] were not effective at dissolving either lignin or wood flour. Separation of lignin from LCB via an ionic liquid dissolution process prior to saccharification has a dual benefit. Delignification is known to enhance enzyme saccharification [11] and a lignin co-product can improve the economic viability of the overall process [121] Biomass dissolution and pretreatment in ionic liquids Biomass is a very variable substrate and its dissolution in ILs depends on the plant genotype, phenotype and degree of processing prior to dissolution. The obvious choices for biomass dissolution are ILs that are good in dissolving both lignin and cellulose. Examples of such ILs include [C4mim]Cl, [C2mim]Cl, [Allylmim]Cl, [C2mim]OAc and some dialkyl phosphate ILs. However the rate of whole biomass dissolution is expected to be slower than that of isolated cellulose or lignin. Plant architecture, hemicelluloses-lignin bonding and other polymer interactions in biomass would be expected to impart recalcitrance to dissolution. In 2005, Myllymaki and Aksela [111] filed a patent on dissolution of whole biomass using ILs. Their examples included use of microwave- and pressure-assisted dissolution of straw, softwood chips and sawdust in [C4mim]Cl. In 2007, Rogers and co-workers [16] published the first comprehensive peer-reviewed article on whole biomass dissolution in ionic liquids wherein partial dissolution of both hardwoods (oak, eucalyptus, poplar) and softwoods (pine) in [C4mim]Cl were achieved. Precipitation of decrystallised cellulose by the addition of water was demonstrated 50

75 and a lignin-containing liquid effluent was reported. However the dissolution of 5 % mass solutions of biomass in IL was still not complete after 24 h at 100 C. When comparing this reaction time with that for the dissolution of ex situ pure cellulose (a few minutes, as shown earlier), it becomes apparent that cellulose in LCB is recalcitrant to dissolution in ILs. This recalcitrance is due to the complex bonding of the surrounding cell wall matrix. However, these dissolutions were carried out in mixtures of IL and deuterated DMSO (15 % mass DMSO-d 6 ) as opposed to pure IL (for the purpose of 13 C NMR analysis) and, as already noted (p. 46), Vitz et al. [114] reported that 25 % mass DMSO in [C4mim]Cl slowed down the cellulose dissolution when compared to pure [C4mim]Cl. Rogers and co-workers supported their methodology by citing previous work [127] where it was observed that 15 % mass DMSO addition in [C4mim]Cl did not reduce the solubility of cellooligomers. The cellooligomers investigated in the work cited were of DP 6 and thus do not exhibit the crystallinity that full cellulose molecules do (DP > 30) or, more importantly, native in situ cellulose. Thus DMSO may well be contributing to a slow dissolution in these first experiments. Kilpellainen, Argyropoulos and co-workers [17, 128] showed that pine dissolution rate was fastest (8 % mass in 8 h at 80 C) when [Allylmim]Cl (as opposed to [C4mim]Cl), small particle size (e.g. ball-milled wood powder) and higher dissolution temperatures (up to 130 C) were used. They also showed that spruce dissolved in [AllylMim]Cl and precipitated with water imparted enhanced cellulose saccharification and decrystallised cellulose. The slower dissolution rates previously reported by Rogers were attributed by Kilpellainen and co-workers to insufficient biomass drying. As discussed earlier, the ILs with an acetate anion have a number of favourable characteristics when compared to those with chloride (viz. high hydrogen bond basicity, low corrosivity, low toxicity, low melting point and biocompatibility). BASF, one of the leaders in industrial applications of ILs, has recently used [C2mim]OAc for dissolution of cellulose [101]. Interestingly, Kamiya et al. [129] reported that the rate of cellulose enzyme saccharification conducted in an 51

76 aqueous solution of 20 % volume [C2mim]OAc was double that conducted in pure water. This indicates an advantage of [C2mim]OAc over other imidazolium ILs which have been reported to irreversibly unfold and inactivate cellulase enzymes. For example, cellulase activity in [C4mim]Cl aqueous solutions as dilute as 22 mm was diminished [130]. However it should be noted that enzyme-friendly ILs have industrial utility only in a process where fermentation can be performed in IL / water / LCB mixtures and where product and solvent can be recovered postfermentation. If extra steps have to be employed to recover the sugars from the IL solutions prior to fermentation, there is no obvious benefit from using such ILs. In addition, ILs are expensive solvents and their recovery/reusability may be reduced by the contaminants introduced by in situ saccharification and fermentation. In the period from ca to 2010, a number of publications have appeared in the literature. Lee et al. [68] have reported efficient lignin extraction and enhanced enzyme hydrolysis by treating maple wood flour with [C2mim]OAc. Rogers and co-workers [101] compared dissolutions of southern yellow pine and red oak in [C4mim]Cl, [C2mim]Cl and [C2mim]OAc and reported on the delignification imparted when acetone in water is used for precipitating the dissolved lignocellulose. Singh et al. [131] have employed confocal microscopy on switchgrass cross sections exposed to [C2mim]OAc and heat and identified swelling patterns and a preferential dissolution of lignin. Li et al. [88] demonstrated that [C2mim]diethylphosphate is outperforming most commonly used ILs at low temperatures (100 C) due to lower viscosity. Zavrel et al. [132], have used a lightscattering technique to screen ILs for their ability to dissolve Avicel cellulose. Arora et al. [102] have carried out substrate mass balance and temperature/time optimisation studies for [C2mim]OAc treatment of switchgrass. Li et al. [76] presented a direct comparison of [C2mim]OAc pretreatment of switchgrass with dilute acid pretreatment. These publications are discussed in more detail in the results and discussion chapters of this thesis. 52

77 2.5 Rationale The theory and experimental work reviewed here serves as a reference for the discussion and comparison of the results presented in Chapters 4 and 5. While the chemistry of lignocellulosics is relatively well understood and several biomass pretreatments for saccharification and fermentation have been described (and in many cases demonstrated at pilot scale), substantially the utility of ILs in biomass processing has not been explored. Descriptions of IL-based LCB pretreatment processes are vague and limited to a few research accounts of imidazolium salts swelling and dissolution properties. The dissolution mechanisms are not well understood and there are few reported studies on process optimisation. Data on the effect of fractionation strategy and process conditions on saccharification efficiency and on mass balances are scant. This thesis investigates the potential of ILs for LCB processing. A number of process strategies can be envisioned. Edye et al. [3, 133, 134] have listed examples of such process strategies, viz.: Dissolution of lignin in IL (with or without cosolvent) (delignification/pulping) Complete or partial dissolution of biomass in IL (with or without cosolvent) with water precipitation (single liquid phase) Complete or partial dissolution of biomass in IL (with or without cosolvent) with dilute aqueous salt / base (e.g. NaOH) precipitation (single liquid phase) Complete dissolution of biomass in IL (with or without cosolvent) with aqueous salt / base (e.g. NaOH) precipitation (biphasic liquid) Complete dissolution in IL (with or without cosolvent) and in situ saccharification (i.e. acid catalysed hydrolysis) 53

78 One of the first described processes is found in the patent claims of Edye and Doherty in 2007 [3, 134] wherein the IL [C4mim]Cl was used to dissolve biomass at high temperatures (130 C 190 C) and fractionation of dissolved components achieved by forming an aqueous biphasic system (ABS) with concentrated alkali (e.g. NaOH). The examples of this patent showed no optimisation work on dissolution, no yields or solid to liquid ratios and limited performance metrics on the ABSs. These knowledge gaps constitute the starting point for the experiments in this study. In this work, numerous aspects of this IL strategy are thoroughly investigated and alternative IL technologies proposed and tested. First, the dissolution reaction is optimised taking into consideration dissolution extent, material losses and saccharification kinetics. Second, a selection of fractionation systems is assessed starting with the NaOH ABS described in the patent and expanding to other ABSs and single phase fractionation systems using preferential precipitation. Finally the mass balance closures and saccharification kinetics of three processes based on three different ILs are reported and discussed. 54

79 CHAPTER 3 METHODOLOGY 3.1 Bagasse Sugarcane bagasse (Rocky Point sugar mill, Pimpama, Queensland) comprising long cuticle fibres and core pith particles (mostly ca. 5 mm to 50 mm) was air dried for a week on metal trays then size reduced (to < 10 mm) with a knife mill before being mixed and subsampled by the cone and split method and stored at 4 C. Before use the bagasse was ground (to < 2 mm) using an electric lab mill (Retsch SM100, Haan, Germany). The material was ground for ca. 1 min per batch to avoid excess heating, placed on top of 2 brass sieves (0.5 mm and 0.25 mm) and in a sieve shaker for 20 min, and the fraction collected between the two sieves was used as the starting material. Moisture content was measured gravimetrically (convection oven, 105 C, overnight) before every use and was 10 ± 1 % mass except where otherwise indicated. 3.2 Chemicals The ionic liquids (1-butyl-3-methylimidazolium chloride [C4mim]Cl ( 95 %) melting point (m.p.) 73 C, 1-butyl-2,3-dimethylimidazolium chloride [C4mmim]Cl ( 97 %) m.p. 96 C - 99 C, 1-ethyl-3-methylimidazolium chloride [C2mim]Cl ( 95 %) m.p. 80 C, and 1-ethyl-3-methylimidazolium acetate [C2mim]OAc ( 90 %) m.p. 20 C, Sigma-Aldrich, NSW) were all dried in a vacuum oven (at 80 C 90 C, ca. 4 mm Hg, > 12 h) prior to use. Initial moisture content (at the time of weighing the IL for each use) was typically ca. 2 % of total mass for [C4mmim]Cl and 1 % for [C2mim]Cl and [C2mmim]OAc as measured by Karl Fischer titration. At this point it is worth noting that although the m.p. of neat [C4mim]Cl is 73 C, its 2 % moisture content was sufficient to maintain it in liquid phase at room temperature. Cellulose (Avicel PH-101), dimethyl sulphoxide (DMSO) (99.9 %) and Karl Fischer HYDRANAL titrant 2E and solvent E were purchased from Sigma-Aldrich (Sydney, NSW). Cellulase / β- glucosidase mixture (Accelerase 1000) was purchased from Genencor (Danisco A/S, 55

80 Denmark). Water was Millipore-filtered and deionised (Milli-Q-plus) to a specific resistivity of 18.2 µs at 25 C. All other solvents and chemicals were analytical grade. 3.3 Uncertainty (or error) analysis of quantitative measurements The data values reported in this thesis represent either single measurements or the mean of duplicate measurements. The precision of the measurement techniques used to produce these quantitative data was calculated as the estimate of standard deviation of duplicate measurements of similar samples analysed identically. The higher the number of sets of duplicate measurements (degrees of freedom or df), the less uncertainty (higher confidence) can be placed on the precision of the technique. The estimate of standard deviation was calculated using Equation 1 as seen in Taylor [135]. The resulting standard deviation along with the degrees of freedom form part of the description of each technique and provide an estimate of its precision. = 2 Equation 1 Where: s Estimate of standard deviation d Difference between duplicate measurements k Number of sets of duplicate measurements ν = k degrees of freedom 3.4 Mass values All mass or percent-mass values reported in this thesis are on a dry basis except where otherwise noted. 56

81 3.5 Karl Fischer titration A Karl Fischer automated titrator (Radiometer Copenhagen TIM 900) with ethanol based HYDRANAL reagents was used to measure moisture content of ILs after drying and prior to use. 3.6 Determination of IL dissolution extent and losses Dissolution In a typical dissolution or pretreatment reaction, IL (5 g on a dry basis) was placed in a 50 ml beaker in an oil bath (clear silicon oil), heated by a hot plate (RET Basic IKA Laboritechnik) with magnetic stirring (300 rpm), in the open atmosphere. The selected temperature for dissolution was controlled by a thermocouple (IKA ETS-D5) immersed in the oil. Bagasse (0.250 g) was added after 60 min, to allow for temperature stabilisation before the start of the reaction. Dissolutions of bagasse were carried out in [C4mim]Cl at varying temperatures (110 C to 160 C), times (30 min to 180 min) and bagasse moisture contents (1 % to 49 %). Bagasse dissolution in different ILs ([C2mim]Cl and [C2mim]OAc) was also conducted Recovery of undissolved solids (UND) and dissolved-then-precipitated solids (DS) To determine the dry mass % of bagasse that dissolved at each trial a variation of the gravimetric method as described by Sun et al. [101] was used (see Figure 3.6.1). After bagasse / IL mixtures were subjected to dissolution conditions, DMSO (ca. 5 ml) was added to reduce the viscosity. The mixture was stirred (300 rpm, 10 min without heating) and then filtered through a pre-weighed nylon filter (20 μm porosity, 90 mm diameter, Millipore) using a Buchner filtration system. The residue was washed with additional DMSO (30 ml) and then with deionised water (100 ml) to remove all residual DMSO on the fibre. This fraction was dried to a constant mass (convection oven, 105 C, overnight) and weighed for gravimetric determination of the undissolved fiber mass (UND). Water washings were added to the original filtrate and then stirred (300 rpm, 20 min) to precipitate the dissolved 57

82 material (IL soluble but water / IL insoluble). The precipitate was collected by filtration (Whatman 54 paper followed by a 0.2 µm Sartorius membrane filtration with a glass fibre prefilter, all filters preweighed), dried (105 C, overnight) and weighed to determine IL / water insoluble dissolved material mass (DS). DISSOLUTION FILTRATION PRECIPITATION UND: undissolved solids mass WITH WATER OB: original (starting) bagasse mass dissolved mass = OB -UND DS: mass dissovled and recovered as a solid precipitate after addition of water Losses = OB -UND -DS Figure 3.6.1: Process for recovering undissolved and dissolved-then-precipitated solids Gravimetric determination of percent mass dissolution The percent of bagasse dissolved was calculated according to Equation 2. =100 1 Equation 2 Where: Diss. Dissolution (% mass of original bagasse added) m OB Mass of original bagasse added (g) m UND Mass of undissolved residue recovered (g) 58

83 The estimate of standard deviation (absolute) of this technique is 6 % mass of bagasse and is based on 5 df only, due to its cumbersome nature and time constraints. Note that duplicate experiments for the critical conditions 140 C and 150 C for 90 min were included in the calculation of this standard deviation estimate Gravimetric determination of percent mass losses The percent of bagasse losses to components non-recovered in solid form was calculated by difference, according to Equation 3. These losses are assumed to be, for the major part, components which are soluble in the IL / water liquid fraction. However, they may also represent small amounts of volatile losses from biomass degradation products (such as acetic acid (b.p C) or furfural (b.p C)), since these dissolutions were conducted in the open atmosphere. =100 (1 + ) Equation 3 Where: Loss Losses (% mass of original bagasse added) m OB Mass of original bagasse added (g) m UND Mass of undissolved residue recovered (g) M DS Mass of precipitated-dissolved residue recovered (g) The estimate of standard deviation of this technique is 3 % mass of bagasse (i.e. absolute) and is based on 5 df only, due to its cumbersome nature and time constraints. 59

84 3.7 Bagasse soda lignin preparation Bagasse soda lignin was prepared by soda pulping of bagasse (175 C, 2 h, bagasse 10 % mass, NaOH 10 % mass) and precipitating the resulting black liquor with acid (2 M H 2 SO 4 ) add to reduce the ph to 3.0. The precipitate was then redissolved in aqueous NaOH (10% mass) and reprecipitated by addition of acid to reduce the ph to 3.0. The recovered lignin solids were washed and dried (40 C, vacuum oven). 3.8 Real time FTIR and reaction calorimetry Real time FTIR (Fourier transform infrared) spectroscopy of bagasse dissolution in IL was carried out in a Mettler-Toledo RC1e Reaction Calorimeter (for accurate temperature control) equipped with a ReactIR FTIR probe (see Figure 3.8.1). The 1 L reaction glass vessel was heated by a silicon oil jacket and the temperature controlled with temperature probes in the jacket and in the reaction mass. An ATR (Attenuated total reflectance) -FTIR probe was immersed in the reaction mass and 64 scan spectra were recorded every 60 s. The height of the absorption bands at 1070 cm -1, 1510 cm -1 and 1560 cm -1 were monitored during the course of the reaction. The valley to valley method was used to determine the baseline for the calculation of each band height. [C4mim]Cl (700 g) was added to the reaction vessel and the temperature stabilised at 70 C (IL liquid due to 2 % moisture). The heat capacity of the IL was measured via the response of the reaction mass temperature to fluctuations of the jacket temperature. Bagasse (35 g) was added and the same procedure repeated for calculating the heat capacity of the IL / bagasse mixture. The reaction mass temperature was accurately controlled and monitored and FTIR spectra were acquired in real-time and band heights plotted over time. Heat flow of the reaction mass was also monitored with the calorimeter but no thermal events of interest were detected and the results are presented in Appendix III. 60

85 Figure 3.8.1: The Mettler-Toledo RC1e reaction calorimeter and ReactIR FTIR probe 3.9 Differential Scanning Calorimetry Differential Scanning Calorimetry (DSC) was performed on samples (ca. 3 mg in sealed aluminium pans) using a Q100 TA (TA Instruments, New Castle, DE, USA) by a heat/cool/heat cycle starting at 25 C and reaching 180 C at 30 C min -1 under nitrogen sweeping gas. The first heat ramps were used to delete thermal history and the second heat ramps are reported Thermogravimetric analysis Samples (5 mg to 20 mg, depending on specific gravity) of solids or ILs were placed in platinum crucible for thermogravimetric analysis (Setaram TGA-DTA/DSC LabSys, Caluire, France) and heated starting from 25 C and increasing to 600 C at 5 C min -1 to 30 C min -1. The first derivative of the mass loss was plotted against temperature to indicate the onset temperature of thermal decomposition. 61

86 3.11 Cellobiose hydrolysis kinetics [C4mim]Cl (3 g) in 20 ml test tubes (2) were heated in an oil bath until desired temperature (130 C and 150 C) was reached. D-Cellobiose (150 mg) was added to each tube and samples (ca. 50 mg) were removed periodically and placed in pre-weighed Eppendorf tubes. Water (1 ml) was added, the tubes were agitated vigorously until a homogeneous solution was observed and then injected (20 μl to 50 μl) into the high-pressure liquid chromatograph (HPLC) for sugars analysis (see Section 3.12) Compositional analysis of solid fractions Compositional analysis was carried out using the standard NREL procedure for determination of structural carbohydrates and lignin in biomass [58]. All samples were freeze dried overnight prior to analysis. Each sample (250 mg) was treated with H 2 SO 4 (72 % mass) at 30 C for 1 h. These samples and a sugar recovery standard (SRS, containing known concentrations of glucose, xylose and arabinose) were then exposed to dilute H 2 SO 4 (4 %) at 121 C for 1 h. The hydrolysis products were determined by HPLC (Waters) equipped with a RI detector (Waters 410) and a Bio-Rad HPX-87H column operated at 85 C. The mobile phase consisted of 5 mm H 2 SO 4 with a flow rate of 0.6 ml min -1. The glucose, xylose and arabinose results were corrected for acid decomposition using the % mass recovery from the SRS. The polysaccharide and acetyl mass content were calculated by conversion of the monosaccharide and acetic acid results with appropriate multiplication factors (0.90 for glucose, 0.88 for xylose and arabinose, for acetic acid). The acid- insoluble lignin (AIL) after acid hydrolysis was measured as the mass loss of insoluble residue at 575 C. The acid-soluble lignin (ASL) was measured by UV-Vis spectrophotometer (Cintra 40) at 240 nm with an extinction coefficient value of 25 L g -1 cm -1 [58]. Ash was determined by placing separate sample fractions at 575 C. The estimates of standard deviation (absolute) of this technique for each component (as % dry mass of bagasse) are: 0.4 for ash, 0.4 for AIL, 0.2 for ASL,

87 for total lignin (AIL + ASL) ii, 1 for glucan, 0.4 for xylan, 0.04 for arabinan and 0.04 for acetyl. These estimates are based on 15 df. This technique was used for all compositional analysis results shown in this thesis, except where otherwise noted Preparation of IL pretreated samples for enzyme saccharification IL pretreated bagasse samples destined for enzymatic saccharification trials were prepared at a larger scale and without drying. Oven or air-drying can irreversibly collapse the pore structure of biomass and affect enzyme saccharification kinetics. Bagasse (2.2 g) in IL (40 g) mixtures were placed in 150 ml beakers and subjected to pretreatment conditions in the manner described in Section The UND was filtered and washed as described in Section (40 ml DMSO and 400 ml of water) and weighed without significant air-drying. The DS was collected by centrifugation in preweighed 250 ml tubes (Beckman J2 MC, JLA rotor, x g, for 10 min), the supernatant was decanted and the sediment was resuspended and recentrifuged with 3 x 200 ml deionised water. In cases where UND and DS were not separated, a total solid residue (TSR) was obtained by adding water to the pretreated IL bagasse mixtures, the precipitate was collected by centrifugation and washed by centrifugation (3 x 200 ml). All solid fractions were transferred to capped glass vials, which were stored moist at 4 C till used for enzyme saccharification trials Preparation of dilute acid pretreated samples Dilute acid pretreatment was carried out according to the LAP-007 NREL protocol for the preparation of dilute acid pretreated biomass [136]. Bagasse (20.77 g at 3.73 % moisture) and deionised water ( ml) were stirred (175 rpm) and heated in a Parr reactor (0.5 L SS316) to 160 C for approximately 10 min, acid ii Estimate of standard deviation for total lignin equals to the square root of the sum of squared estimates of standard deviation for AIL + ASL. 63

88 (16.22 ml of 9 % mass H 2 SO 4 ) and then water (15.25 ml) were injected with a high pressure feeding pump (Prominent Beta / 4) over 165 s and the temperature was maintained at 160 C for 10 min. The reactor was then cooled by placing the reactor in an ice bath and running tap water through the internal cooling coil. The pretreated solids were recovered by filtration (Whatman 5) and washed with distilled water until ph of the washate was > 5.0. The washed and moist solids were stored at 4 C till use Enzymatic saccharification Cellulose contents of bagasse and pretreated bagasse were determined by compositional analysis prior to enzymatic saccharification (Section 3.12). Moisture content was determined prior to calculating sample weights required (estimated moisture) and also at the time of actually weighing the samples (actual moisture). The former was used to determine enzyme loadings (by oven drying to constant mass overnight at 150 C) and the latter was used for final cellulose and xylan concentration calculations. Bagasse and pretreated bagasse samples (100 mg cellulose equivalents) were suspended in citrate buffer (10 ml, 50 mm, ph 4.7) and equilibrated on a temperature controlled rotary shaker (150 rpm, 50 C). Accelerase 1000 (Genencor) was added to achieve an enzyme activity of 15 FPU g -1 (50 µl of Accelerase as received). Samples (0.5 ml) were removed periodically placed in ice, centrifuged at 4 C and then frozen. After thawing of the samples at room temperature, cellobiose, glucose and xylose concentrations were measured by HPLC (HPLC system as in section 3.10). The glucose and cellobiose results were converted to glucan mass equivalents and xylose was converted to xylan mass equivalents using appropriate multiplication factors. The estimates of standard deviation (absolute) of this technique are 2 % mass of glucan and 2 % mass of xylan and are based on 15 df. 64

89 This technique was used for all enzyme saccharification monitoring results shown in this thesis, except where otherwise noted XRD cellulose crystallinity measurement Pretreated and the untreated bagasse samples were scanned on a diffractometer (PANalytical X Pert MPD, Cu α ( Å) radiation) with a scan speed of 0.18 min -1 and a step size of (see Figure for example) counts untreated I AM I TOT θ[ ] Figure : Diffractogram of bagasse The cellulose crystallinity index (CrI) was determined using Equation 4 as reported by Thygesen et al. [137]. 65

90 = Equation 4 Where: CrI Crystallinity index I TOT Intensity at about 2 ϑ = 22 (represents the crystalline and amorphous material) I AM Intensity at the valley between the two peaks at about 2 ϑ = 18. (represents the amorphous material) There is controversy as to which baseline should be used for the measurement of the intensity values in Equation 4 [137]. The straight baseline used in this study was drawn by baseline normalisation ( valley to valley ). This approach may lead to an overestimation of crystallinity (since there is no account of background scatter). However, this overestimation will be of similar magnitude for all samples examined and thus should not affect the overall reliability and consistency of conclusions. The estimate of standard deviation (absolute) of this technique is 0.01 and is based on 3 df only due to limited time of instrument availability. This technique was used for all cellulose crystallinity index results shown in this thesis, except where otherwise noted Saccharification and fermentation Bagasse (35 g) in [C4mim]Cl (464 g) was reacted (150 C for ca. 1 h) in the RC1 reactor (same reactor setup as in Section 3.8), cooled to 70 C and precipitated with water (ca. 300 ml). The recovered solids were centrifuged and washed as described in Section 3.13 and stored moist at 4 C. Enzymatic hydrolysis reactions were performed under sterile conditions for 3 days in 100-mL Erlenmeyer flasks (equipped with water traps) on a rotary shaker (150 rpm, 50 C) in volumes of 40 ml with a biomass load of 2 g cellulose equivalent and Accelerase 1000 (Genencor) activity of 5 FPU g -1 in 50 mm citrate buffer (ph 4.7) containing 4 ml of a YP solution (100 g L -1 yeast extract and 200 g L -1 peptone). 66

91 At the end of the 3 days, 0.5 ml of yeast cell suspension (see preparation below) was added to achieve a final optical density (O.D.) of 0.5 and incubated under sterile conditions (32 C, 130 rpm, 72 h); samples were removed at regular intervals and after appropriate dilutions, injected onto the HPLC (as in section 3.10) for quantification of glucose and ethanol concentrations. Yeast cells (Saccharomyces cerevisiae) were prepared as described in the relevant protocol by the National Renewable Energy Laboratory (NREL) [138]. A frozen stock culture was suspended in a sterilised flask with YPD (10 g L -1 yeast extract, 20 g L -1 peptone, 50 g L -1 dextrose) medium and incubated overnight at 32 C with orbital agitation. Cells were then harvested by centrifugation and washed with sterile water (3 x 38 ml, 5 min, 4500 rpm). The resulting solids were suspended in 2.5 ml sterile water and optical density measured with a spectrophotometer at 600 nm ATR-FTIR For liquid samples, a drop was placed on the diamond probe of a Thermo Nicolet 870 FTIR (software: OMNIC 7.3). For solid samples, a small amount of freeze dried fibre, enough to cover the surface of the probe, was used. The sample was pressed with an anvil to increase the surface contacting the probe. Sixty-four scans were acquired for each spectrum and the two replicate spectra for each sample were overlayed. No differences in the replicate spectra of this study were observed and thus only the first spectrum of each sample was used for analysis Aqueous biphasic systems Preparation of ABSs ABSs were prepared using the proportions described in the patent of Edye and Doherty [3, 134] by mixing IL (6.3 g) with 20 % NaOH (8.4 ml) in a 25 ml volume 67

92 graduated cylinder, agitating and allowing to stand overnight except where otherwise indicated Cloud point titrations The coexistence curves were determined using cloud point titration at ambient temperature in a similar manner as reported by Bridges et al. [139]. The titration started with a solution of known and high concentration of ionic liquid in water. Dropwise addition of kosmotropic salt solution (of known concentration) to a monophasic (clear) IL solution was followed by vigorous vortexing, and time to settle. Upon settling, if a cloudy solution formed (which would yield a biphasic solution if allowed to separate completely), the cloud point was deemed reached and the mass of titrant added was recorded. The same was then repeated with the water titrant until the solution became clear again and the water mass added recorded. This was continued until enough points were measured for an accurate coexistence curve. The mass additions and the concentrations of the starting solution and the titrant were used to determine the IL molality and the kosmotropic salt molality at each cloud point observed Ion concentration determination (for ABS distribution ratios) The phases of biphasic systems comprising an IL top phase and an aqueous salt bottom phase were sampled separately. Sampling of the bottom phase was carried out with care not to contaminate it with top phase solution. A Pasteur pipette was passed through the top phase with an air bubble maintained at its tip. After sampling the bottom layer and upon exit, small amounts of bottom phase solution were being expelled while the pipette tip was crossing the top phase. Upon exit the exterior of the pipette tip was wiped with a tissue. The samples were diluted with deionised water and ion concentrations determined by ion chromatography (Metrohm 761 with a conductivity detector). For cation analysis, samples were acidified (to ph 3.5, 2M HNO 3, ca. 1 µl ml -1 ) and injected onto a Metrosep C 2150 (150 mm x 4 mm) column with an aqueous mobile phase (25 % volume acetone, 6 mm tartaric acid and 0.75 mm dipicolinic acid) at 1 ml min -1. For anion analysis samples were injected onto a Metrosep ASupp5 (150 68

93 mm x 4 mm) column with an aqueous mobile phase (1 mm NaHCO 3 and 3.2 mm Na 2 CO 3 ) at 0.7 ml min -1 and suppression by post-column addition of H 2 SO 4 (50 mm). The estimate of relative standard deviation (RSD) of this technique is 0.5 % for chloride ions and is based on 9 df Quantification of [C4mim]Cl deprotonation using an acid titration HCl (0.2 M) was used as a titrant on an aqueous solution (50 ml) containing ca. 1.5 g of the IL phase sample. The [C4mim]Cl content (% mass) of this IL phase sample was determined using refractive index (Bellingham Stanley RFM320 refractometer). The linear relationship of refractive index to [C4mim]Cl concentration (% mass) in aqueous [C4mim]Cl solutions was determined using 7 standards as shown in Figure This relationship is in close agreement with Liu et al. [140] y = x R² = Refractive index [C4mim]Cl concentration (% mass) Figure : Linear relationship of refractive index to [C4mim]Cl concentration in water 69

94 In the case of the [C4mim]Cl / Na 2 CO 3 and trisubstituted [C4mmim]Cl / NaOH ABSs, the IL content of their IL phases was determined via the more cumbersome ion chromatography technique (see Section ) due to unavailability of the refractometer Mass balance determinations for three IL treatments Mass closure experiments for three ionic liquid treatments of bagasse were carried out in sealed tubes to avoid volatile losses such as acetic acid (b.p C) or furfural (b.p C) or from the degradation of xylose. The fractionation process was designed by the author and is shown in Figure The amounts of water used for each of the three preferential precipitations are based on the results of Section 5.3. The mass of each of the original reactants was recovered in either one of the liquid or solid fractions depending on their solubility in the three different concentrations of water in IL used (0.5, 2.0 and 3.5 (+acidification) of water : IL mass ratio). Bagasse (0.25 mm 0.5 mm) was extracted with ethanol and water using a Sohxlet device according to the NREL protocol for biomass extractives [141]. ILs (ca. 30 g of either [C4mim]Cl or [C2mim]Cl or [C2mim]OAc in duplicate) were weighed in sealable pressure glass tubes (ACE glass 50 ml). At this point, IL (ca. 0.5 g) was weighed and set aside for IL recovery analysis (see section ). Extracted bagasse (3.5 % moisture) (ca. 1.5 g for [C4mim]Cl and [C2mim]Cl and 0.75 g for [C2mim]OAc) was added to each pressure tube, sealed with Teflon stoppers and placed in an oil bath which was stabilised at 150 C with magnetic stirring at 200 rpm. The tubes were left in the oil bath for 60 min (25 min of which at temperature ramp and 35 min at 150 C) and, upon removal, placed in an ice, bath with magnetic stirring to quench the reaction. After 2 min, the tubes were removed from the ice bath and water was added equal to 0.5 mass fraction of the originally added IL. The tube was sealed again and agitated vigorously until a homogenous solution between water and IL appeared to form. The contents of each tube were quantitatively transferred into a preweighed polypropylene centrifuge tube and centrifuged at x g for 20 min. The liquid contents of the centrifuge tube were 70

95 decanted to a new preweighed polypropylene centrifuge tube and weighed (liquid fraction 1). The pellet (solid fraction 1) was centrifuge washed with distilled water (5 x 30 ml at x g and 5 min - 10 min cycles), freeze dried overnight (-85 C, 80 mt) and weighed. Liquid fraction 1 was precipitated with additional water resulting to a water : IL mass ratio of 2.0. Precipitation and coagulation of solids was aided by storing at 4 C overnight followed by shaker incubating at 55 C - 70 C for 60 min. The resulting lignin rich precipitate (solid fraction 2) was centrifuge washed, freeze dried and weighed. biomass dissolution precipitation in water : IL mass ratio = 0.5 LIQUID FRACTION 1 SOLID FRACTION 1 precipitation in water : IL mass ratio = 2 LIQUID FRACTION 2 SOLID FRACTION 2 Precipitation in water : IL mass ratio = acidification to ph <1.0 LIQUID FRACTION 3 SOLID FRACTION 3 Figure : Flow chart of the fractionation process used in mass balance experiments 71

96 Losses of liquid components to washings of pellets were accounted for by weighing pellets prior to washing and after drying (it is assumed that the composition of these lost liquid components is the same as the bulk liquid). Similarly, subsampling for analysis was accounted for by careful attention to mass changes. Solid fraction 1 and the starting biomass were characterised using the NREL acid hydrolysis protocol (see Section 3.12). Solid fractions 2 and 3 were characterised for lignin content with the acetyl bromide protocol described by Iiyama and Wallis [142]. The sample of liquid fraction 1 was directly injected onto the HPLC and the Ion Chromatograph (IC) for quantification of monosaccharides and IL ions respectively while the soluble oligosaccharides were determined by acid hydrolysis. All methods are described in detail in the following sections. The distribution of cellulose, hemicellulose and lignin between solid fractions, liquid fraction monosaccharides and liquid fraction oligosaccharides was reported as a percent mass of the components in the starting material Compositional analysis of solid fraction 1 The solid fraction 1 of each reaction was characterised according to the NREL protocol described in Section The estimates of standard deviation (absolute) of this analysis (on the basis of duplicate IL pretreatments, 3 degrees of freedom) for each component (as % dry mass of starting bagasse) are: 2 for glucan, 2 for xylan, 3 for arabinan, 1 for acetyl and 2 for lignin Compositional analysis of monosaccharides in liquid fraction 1 Each sample of liquid fraction 1 (0.5 ml) was weighed in 1.5 ml Eppendorf tubes and diluted with water (0.5 ml). The contents were vortexed thoroughly, filtered through a 0.45 µm nylon filter and injected to a Waters HPLC as described in Section Glucose, xylose, arabinose and acetic acid masses were converted to glucan, xylan, arabinan and acetate masses using appropriate multiplication factors (see Section 3.12). In addition, it was assumed that the detected 72

97 hydroxymethylfurfural (HMF) and furfural were products of cellulose and xylan degradation respectively. Therefore, HMF and furfural masses were converted to cellulose and xylan mass equivalents using multiplication factors of 1.28 and 1.38 respectively. The estimates of standard deviation (absolute) of this analysis (on the basis of duplicate IL pretreatments, 3 df) for each component (as % dry mass of starting bagasse) are: 0.2 for glucan, 0.2 for xylan, 20 for arabinan and 0.7 for acetyl. The unacceptably high standard deviation for arabinan is attributed to its very low concentrations in this analysis Compositional analysis of oligosaccharides in liquid fraction 1 Each sample of liquid fraction 1 (0.5 ml) and SRS solution (0.5 ml) were weighed in 2 ml twist-top Eppendorf tubes, diluted with water (1 ml) and acidified (with 72 % mass H 2 SO 4 ) to a ph of 0.3. The contents were vortexed thoroughly and autoclaved (121 C for 60 min; autoclaving did not affect mass). After cooling to room temperature, the autoclaved tube contents were filtered through a 0.45 µm nylon filter and injected onto the same HPLC system as in section After SRS correction for acid decomposition of sugars (see Section 3.12) and subtraction of the monosaccharide composition results (Section ), the difference was converted to polysaccharide mass equivalents (using appropriate multiplication factors as in Section 3.12) in order to arrive at the composition of the soluble oligosaccharides in liquid fraction 1. The estimates of standard deviation (absolute) of this analysis (on the basis of duplicate IL pretreatments, 3df) for each component (as % dry mass of starting bagasse) are: 1 for glucan, 2 for xylan, 20 for arabinan and 3 for acetyl. The unacceptably high standard deviation for arabinan is attributed to its very low concentrations in this analysis Acetyl bromide for lignin quantification in solid fractions 2 and 3 The acetyl bromide method as described by Iiyama and Wallis [142] was used to determine the % mass lignin content of solid fractions 2 and 3. Freeze dried 73

98 solids (ca. 10 mg) were weighed in glass tubes and acetyl bromide in acetic acid (25 % mass, 10 ml) and then perchloric acid (70% mass, 0.1 ml) were added. The tubes were sealed with Teflon screw caps and placed in temperature controlled rotary shaker (70 C and 100 rpm for 30 min). After cooling to room temperature the tubes were opened and 2 M NaOH (10 ml) and then glacial acetic acid (25 ml) were added. After agitation, absorbance (280 nm, quartz cuvettes, Cintra UV spectrometer) was measured against glacial acetic acid. The resulting solution was analysed with a Cintra-40 UV spectrometer and the absorbance was referenced to a cuvette with glacial acetic acid. Dilutions with glacial acetic acid were necessary for some samples so that the absorbance was < 1.0. The absorbance was converted to percent mass concentration of lignin using an extinction coefficient of 25 L g -1 cm -1. The extinction coefficient was determined with the use of a calibration curve based on bagasse of known lignin content. The estimate of standard deviation (absolute) of this technique (duplicate samples of untreated bagasse and soda lignin, 4 df) (as % dry mass of solid analysed) is Recovery of IL IL set aside at the start of mass balance experiments (starting IL) was brought to a volume of 50 ml with deionised water. Similarly, known masses of liquid fraction 1 were diluted with deionised water and injected onto the ion chromatograph as described in Section IL mass balance was determined from the results of these analyses. The estimate of standard deviation (absolute) of this technique (as % mass of ions in starting IL) for both cations and anions and for all 3 ILs is 2 (based on duplicate IL pretreatments, 6 df) Enzymatic saccharification of solids from 3 IL treatments Enzymatic hydrolysis reactions were performed in 20 ml scintillation vials on a rotary shaker (150 rpm, 50 C) in volumes of 5 ml with a biomass load of 50 mg cellulose equivalent and Accelerase 1000 (Genencor) activity of 15 FPU g -1 (25 µl of Accelerase as received) in 50 mm citrate buffer (ph 4.7). Samples (0.2 ml) were periodically removed, placed in ice, then in boiling water (2 min) and centrifuged. The sugars analysis and conversion to glucan an xylan masses was done as in 74

99 Section 3.12 except a Shodex SPO-810 HPLC column at 85 C with a mobile phase of ultrapure water at 0.6 L min -1 were used. The estimates of standard deviation of this analysis (based on duplicate IL pretreatments saccharification extents at different time points, 18 df) are 2 % mass of glucan (i.e. absolute) and 0.9 % mass of xylan. 75

100 CHAPTER 4 RESULTS PRETREATMENT Pretreatment is the process by which the LCB structure is opened up to facilitate enzyme saccharification of the polysaccharide fraction. In this chapter a simple ionic liquid pretreatment system is investigated. Bagasse is partially dissolved in IL and the pretreated solids are recovered with the addition of water. Section 4.1 investigates the pretreatment performance of this system and compares it with dilute acid pretreatment. Section 4.2 studies the characteristics of the undissolved bagasse in [C4mim]Cl and draws conclusions on the non-dissolution effects of IL pretreatment, such as structural (viz. fibre swelling and cellulose decrystallisation) and compositional changes (viz. preferential dissolution patterns) of the undissolved fraction. 4.1 Biomass dissolution in IL and recovery by addition of water The utility of ILs in biomass pretreatment is attributed to their ability to be tuned and be compatible with a wide array of processes entailing a biomass dissolution step as listed in Section 2.5. The first undertaking of this study is the exploration of a simple system entailing partial dissolution of bagasse with [C4mim]Cl and precipitation with water. The extent of dissolution and material losses (unrecovered portion of biomass lost to soluble components, e.g. monosaccharides, lignin monomers, degradation products) incurred when bagasse is reacted in ionic liquids at temperatures between 110 C and 160 C are examined in this section. Most ionic liquid work so far quotes long biomass dissolution times (in the order of days) at temperatures 110 C [16, 91, 101]. The rationale for using relatively low temperatures is (albeit with little evidence) that polysaccharide degradation is assumed to be high at higher temperatures. The reaction parameters that 76

101 determine the extent of dissolution and degradation of bagasse in [C4mim]Cl are also investigated in this section in order to optimise this simple system at high temperatures. The optimised [C4mim]Cl system is then compared to dilute acid pretreatment in terms of saccharification and fermentation yields and processing time Ionic liquids used Three most cited ionic liquids for cellulose and biomass dissolution are [C4mim]Cl, [C2mim]Cl and [C2mim]OAc (see Figure 4.1.1). [C2mim]Cl is a difficult IL to handle since it is solid up to about 70 C and mixtures of bagasse in [C2mim]OAc are difficult to stir when biomass loading in IL exceeds 2.5 % mass. Consequently [C4mim]Cl was chosen for investigating the parameters affecting high temperature dissolutions. The dissolution extents in all three ionic liquids are also compared in this section. 1-ethyl-3-methylimidazolium chloride 4 3 N CH 3 N CH 3 N CH 3 O 5 N 1 2 Cl N Cl N O CH 3 CH 3 CH 3 CH 3 1-butyl-3-methylimidazolium chloride 1-ethyl-3-methylimidazolium acetate Figure 4.1.1: ILs used in this study 77

102 4.1.2 Factors affecting biomass dissolution a Background Dissolution is the process wherein substances (solids, liquids or gases) disperse in a liquid (solvent) to form a solution. In the cases of dissolution of solids and liquids, the solution is formed by dissociation of solute material (e.g. breakdown of crystal lattice) and association or solvation of individual solute molecules with solvent molecules. Dissolution rate is dependent on properties of the solvent and the solute and the conditions under which they interact. These variables are expressed in Equation 5. = Equation 5 Where: m Amount of dissolved material (kg) t Time (s) A Surface area of the solid material (m 2 ) D Diffusion coefficient (m 2 s -1 ) d Thickness of the boundary layer of the solvent at the surface of the dissolving substance (m) C s Concentration of the substance in the boundary layer (kg m -3 ) C b Concentration of the substance in the bulk of the solvent (kg m -3 ) The relation of diffusion to temperature is defined by the Arrhenius equation (Equation 6) = Equation 6 Where: D Diffusion coefficient (m 2 s -1 ) D 0 Maximum diffusion coefficient (at infinite temperature) E a Activation energy for diffusion (kj mol -1 ) T Temperature ( K) R The universal gas constant (J K -1 mol -1 ) These equations describe the behaviour of pure solids dissolving in pure solvents and assume that the surface of the solids is chemically homogeneous. 78

103 Biomass, as described earlier, is a complex structure and is not expected to exhibit a chemically homogeneous surface. Furthermore as the biomass dissolves the composition of its surface changes (i.e. the proportion of the more soluble components reduces). However, the equations indicate the parameters that accelerate biomass dissolution and their relative importance. In this work, the initial surface area and concentration of biomass solids are held constant by using consistent particle size, agitation conditions and low biomass loading (5% mass) across experiments. The parameters varied are temperature, residence time, bagasse moisture and ionic liquid type. According to Equation 6, temperature and activation energy of diffusion are the main variables influencing the diffusion coefficient. The diffusion coefficient, and thereby the dissolution rate, will increase exponentially with increasing temperature and/or reducing activation energy. The apparent activation energy for biomass diffusion would be influenced by the activation energies of diffusion of its individual components b Effect of temperature Bagasse (0.250 g) dissolution in [C4mim]Cl(5 g) for 90 min was conducted at different temperatures (110 C to 160 C) and the extent of dissolution along with the biomass losses (biomass ending up in liquid fraction) after recovery by addition of water was measured for each temperature (results shown in Figure 4.1.2). Dissolution extent was determined by weighing the undissolved material recovered as the filtration residue after diluting the reaction mass with DMSO. Losses were determined by the difference of the mass dissolved and the mass of the recoverable filtrate after precipitating the dissolved mass with water (see Section 3.6). At this point, it is important to clarify that DMSO dilution as described above does not cause further dissolution of bagasse or precipitation of dissolved components. Although DMSO is a solvent for carbohydrate-free lignin, it has been reported by Rogers and coworkers [101] that it does not interfere with native lignin or carbohydrates and thus does not influence dissolution results. It must be noted however, that the author has observed gel formation upon addition of DMSO in the 79

104 [C2mim]OAc / bagasse solutions. It is also important to clarify that DMSO dilution as described above may cause slightly different components of bagasse to form part of the losses as compared to IL / water liquid fractions. However if such bias is taking place it is internally consistent for all samples analysed and compared. Significant dissolution was observed at temperatures above 130 C, where the extent of dissolution appears to more than double with every 10 C temperature increment (Figure 4.1.2). This is only an empirical observation since, as discussed in Section a, the Arrhenius kinetics equation cannot be directly applied to the continuously altering surface of biomass in dissolution. At 160 C, the dissolution rate and extent are high but the losses are also high. At 150 C and below the mass of losses are fairly consistently 1/3 of mass dissolved. It is therefore concluded that temperature should be maintained at or below 150 C to avoid excessive losses. At 160 C the dissolution approaches its practical end point (dissolution = 92 % mass), only the more recalcitrant material remains and consequently the dissolution rate slows. However depolymerisation and degradation reactions of the solvated material continue resulting in excessive losses (53% mass, cf. 17 % mass at 150 C). These findings were recently corroborated by Rogers and co workers [101] who demonstrated that the last 10 % mass of the starting pine or oak, in a range of ILs, required as much (or more) time to dissolve as the first 90 %. The same workers reported losses of 40 % mass for near complete dissolution of pine in [C2mim]OAc (110 C for 16 h). This ratio of losses to dissolution is close to the 1:3 found here although slightly higher possibly due to the fact that the dissolution extent for Rogers was closer to 100 %. The increased losses are due to depolymerisation of solvated biomass leading to formation of molecules soluble in the water / IL mixture. While it is possible or even likely that dissolution involves breaking of intramolecular C-C and C-O bonds, it is not possible from these results to infer much about the molecular masses (or DP) of solutes. However, it is certain that depolymerisation does occur since the lost material is of low enough molecular weight (or DP) to be soluble in 80

105 the water / IL mixtures. Therefore, depolymerisation may be a consequence of the dissolution process and it is reasonable to expect more depolymerisation after dissolution (i.e. in the solvated state). dissolution losses bagasse (% mass) Temperature ( C) Figure 4.1.2: Effect of temperature on bagasse dissolution in [C4mim]Cl for 90 min c Effect of time To investigate the effect of reaction time on dissolution, bagasse (0.250 g) dissolution in [C4mim]Cl (5 g) at 150 C was conducted for five different residence times (30 min to 180 min) and the extent of dissolution along with the associated biomass losses was measured for each time. These time series results are presented in Figure and they confirm the dissolution pattern observed in the temperature series experiments. Generally, in cases where ca. 75 % or less of the material is dissolved, the ratio of losses to dissolution extent is about 1:3. The dissolution appears to be ca. three times faster than the combined rates of depolymerisation and degradation (which lead to losses). 81

106 dissolution losses bagasse (% mass) Time (min) Figure 4.1.3: Effect of residence time on bagasse dissolution in [C4mim]Cl (150 C) At least for these biomass : IL ratios, optimum conditions appear to be 150 C and 90 min. At lower temperatures and times the extent of biomass dissolution is low and at higher temperatures and times depolymerisation and degradation reactions lower recovery. Recently, Varanasi et al. [119] reported using 150 C and 90 min for pretreatment of corn stover in [C4mim]Cl and [C2mim]OAc. Fu et al. [143] reported using the same conditions for triticale straw in [C2mim]OAc. Furthermore, these authors suggested these conditions to be optimum for high saccharification rates of the IL treated biomass d Effect of bagasse moisture Moisture is an inherent component of biomass and it is an important factor in its dissolution due its dual role as a reactant in the hydrolysis of glycosidic bonds and an antisolvent of cellulose and lignin. Bagasse at the sugar mill gate is received 82

107 at ca. 50 % moisture. Air-dried bagasse contains about 10 % moisture. Bagasse samples (0.225 g each, on a dry basis) of different moisture contents (viz %, 10.5 % and 1.1 % moisture) were reacted in 5 g of [C4mim]Cl(90 min at 150 C) and the dissolution and losses of all three samples is shown in Figure There is little difference between air-dried (10.5 % moisture) and oven-dried (1.1 % moisture) bagasse both in terms of extent of dissolution and losses (Figure 4.1.4). This outcome could be related to the high reaction temperature (150 C) where most water is evaporated quickly. It is not possible to distinguish between a simple water concentration effect and a more complex competition between water and IL in occupying pore and fibre structures. However, the solubility of bagasse at 48.5 % moisture content seems to be significantly diminished and it is possible that when bagasse moisture is high enough and water is inside the biomass structure, it is slow to be displaced by ILs. The ratios of losses to dissolution extent appear to remain around 1:3 at all moisture contents tested. dissolution losses bagasse (% mass) moisture (%) Figure 4.1.4: Effect of bagasse moisture content on bagasse dissolution in [C4mim]Cl 83

108 4.1.2.e Loading In preliminary dissolution experiments, it was observed that an initial loading of ca. 9 % mass bagasse loading in [C4mim]Cl(150 C) rendered the mixture very viscous and impossible to stir with magnetic stirring. However, when the reaction was mechanically stirred in the RC1 reactor (setup described in Section 3.8) and the loading was incrementally dosed, much higher loadings were achieved. Bagasse loading started with a ca. 8.6 % mass (33.7 g) in [C4mim]Cl (356 g, 150 C) and subsequently bagasse was added in increments of 10 g when the viscosity of the reaction mass appeared to drop. The loading achieved was 15.3 % (64.4 g) within the first 2 h and 20.6 %(94.4 g) in 5 h. The fact that polysaccharides were still dissolving at these loadings was confirmed with monitoring of the 1070 cm -1 FTIR absorption band as described in Section g f Effect of Ionic liquid choice The dissolution of bagasse in a choice of three ILs under identical conditions (150 C, 90 min and 5 % mass bagasse in IL) was investigated and the results are shown in Figure It has to be noted that the [C2mim]OAc dissolution at this loading (5% mass) was very viscous and hard to stir. One of the most attractive characteristics of ILs is that the vast range of ion combinations which allow for great ability of their physicochemical properties to be tuned. The variation in IL ions has a major influence on dissolution extent and ratio of losses. The effect of cation size is exhibited when comparing [C4mim]Cl with [C2mim]Cl. In agreement with the literature [57], the smaller [C2mim] cation imparts higher dissolution and this may be due to the enhanced penetration of the smaller solvent molecule resulting in higher dissolution. However, the mass losses are nearing the mass dissolved which indicates increased depolymerisation of solutes. The effect of anion is studied by comparing [C2mim]Cl to [C2mim]OAc. The acetate anion seems to favour dissolution as opposed to losses. Moreover, losses may be exacerbated due to the fact that dissolution (96 %) is well into the last recalcitrant bagasse fraction. It is probable that by reducing the severity of the reaction conditions, the ratio of dissolution to losses will be improved. Acetate has a 84

109 higher hydrogen bond basicity than chloride [92] and thus its ability to disrupt hydrogen bonds and dissolve cellulose is higher. Out of the ILs tested, [C2mim]OAc appears the most favourable dissolution losses bagasse (% mass) [C4mim]Cl [C2mim]Cl [C2mim]OAc Ionic liquid Figure 4.1.5: Effect of ionic liquid choice on bagasse dissolution The superiority of [C2mim]OAc to [C4mim]Cl as a solvent for biomass has been reported in the literature recently. Rogers and co-workers [101] measured 93.5 % dissolution extent of southern yellow pine in [C2mim]OAc and only 26 % in [C4mim]Cl under the same conditions (particle size mm, 5 % mass loading, 110 C for 16 h). However [C2mim]OAc is not extensively used in this research due to the high viscosity of [C2mim]OAc / bagasse solutions. Preliminary experiments with bagasse loading as low as 3 % mass in [C2mim]OAc resulted in reactions that were difficult to stir with magnetic stirring. The higher biomass 85

110 loadings reported in the literature are attributed to the different substrate used (viz. pine as opposed to bagasse). Comparatively, Zavrel et al. [132] reported in 2009, the use of a light scattering technique to screen ILs for their ability to dissolve Avicel cellulose. The scattered light beam extinction after passing through a dissolution reaction is positively related to the amount and size of suspended undissolved solids. With this technique the solubility was ranked as follows [C2mim]OAc>[C2mim]Cl>[AllylMim]Cl>[C4mim]Cl. However, when these ILs were tested for dissolution of wood chips (both hardwood and softwood), [Allylmim]Cl was found better than the [C2mim] + ILs. Lee et al. [68] conducted incremental additions of maple wood flour in a number of ILs (80 C for 24 h). Conversely, they report a wood solubility of > 30 g kg - 1 for [C4mim]Cl and < 5 g kg -1 for [C2mim]OAc. Their methodology, based on qualitative visual observation of dissolution, may be introducing bias to these results as compared to the methodology used by all above authors and the one used in this research g Monitoring dissolution kinetics using real time FTIR - ATR The FTIR spectra of bagasse soda lignin (prepared according to Section 3.7), glucose and cellulose (each dissolved in [C4mim]Cl) were obtained and analysed. Lignin concentration was found to be linearly related to absorbance at 1510 cm -1. Likewise, glucose and cellulose concentrations were linearly related to absorbances at 1050 cm -1 and 1070 cm -1, respectively. These wavenumbers were used to monitor biomass dissolution by ATR-FTIR. The nature of these absorbances and the quantification method are provided in Appendix I. FTIR spectra from a dissolution reaction with 5 % mass bagasse in [C4mim]Cl were acquired in real time (see Section 3.8 for details). The 1070 cm -1 band was attributed to the sum of all polysaccharides that were dissolved in the IL. The 1510 cm -1 band was not detectable, possibly due to the low concentration of lignin in the biomass / IL solution. The absorbance at 1570 cm -1 was attributed to the 86

111 imidazolium ring of [C4mim]Cl. Figure shows the extent of polysaccharide solvation during the temperature ramp by plotting the trend of the ratio of the polysaccharide absorbance (1070 cm -1 ) to that of the background [C4mim]Cl absorbance (1570 cm -1 ). Polysaccharide solvation appears stagnant to very slow at 70 C, even after 120 min and it starts accelerating significantly when the temperature is increased. Dissolution appears to accelerate at temperatures above 150 C. Certainly, dissolution rate has significantly increased as the temperature approaches and exceeds 160 C, but then slows down again when the temperature is returned to 150 C. This outcome suggests that high temperatures are indeed associated with accelerated dissolution and requires further investigation. Temperature ( C) reaction temperature ( C ) polysaccharide (FTIR peak ratio) FTIR peak ratio Time (min) 0 Figure 4.1.6: Real time FTIR of bagasse polysaccharides upon dissolution in [C4mim]Cl 87

112 4.1.3 Thermal stability of bagasse components in [C4mim]Cl In order to investigate the thermal stability of bagasse components in [C4mim]Cl, DSC and TGA analysis was used and the kinetics of glycosidic bond cleavage of cellobiose in [C4mim]Cl were studied a Differential scanning calorimetry of bagasse and bagasse lignin The dissolution at high temperatures can be accelerated by biomass softening phenomena triggered at certain temperatures. For example, biomasssoftening phenomena occurring at high temperatures at and above the glass transition of lignin effect disentanglement and fibre swelling which increases surface area. Lignin, hemicelluloses and the amorphous component of bagasse cellulose are all viscoelastic materials and can be expected to exhibit glass transition temperatures [144]. Dissolution of biomass close to the glass transition temperature of lignin is thought to influence dissolution and pretreatment effects [102, 131, 145]. In this study, differential scanning calorimetry (DSC) was used to identify such possible material softening phenomena in biomass at high temperatures. Glass transition represents the change from a glassy state of an amorphous polymer to its rubbery state. When this change occurs it is associated with a sudden change in heat capacity at that temperature which is detectable as an endothermic transformation in a DSC thermogram. DSC profiles of bagasse, NaOH extracted bagasse lignin (i.e. soda lignin), and bagasse in [C4mim]Cl and [C2mim]OAc were acquired according to Section 3.9 and are shown in Figure The arrows are showing points of maximum endothermic transitions, only in the temperature range of 110 C to 160 C (temperatures used in this work), as calculated by the thermal analysis software. The transition in the lignin is clear and characteristic and thus can be attributed to a glass transition temperature at 122 C. This figure is in close agreement to the glass transition temperature for corn stem rind lignin reported by Donohoe et al. [145] at 120 C and it is also close to the glass transition determined for bagasse soda lignin by Moussaviun and Doherty [146] at 130 C. Glass transition may vary for bagasse lignins extracted under different conditions and processes. The DSC curve for 88

113 bagasse indicates a transition at 140 C which is less sudden and occurs over a wide temperature range. This is not surprising since the amorphous component of bagasse is more complex than extracted lignin and the heat capacity change may be a result of a number of transitions resulting in a summative broad transition. The transitions observed in the bagasse DSC curves are relatively subtle and broad, and thus a sharp glass transition temperature cannot be identified. Nevertheless, when bagasse is reacted in ionic liquids, the transitions seem to occur at lower temperatures with [C2mim]OAc having a greater effect on this thermal transition than [C4mim]Cl. The transition temperatures can be considered broad indicators of slight softening of bagasse at these temperature ranges, namely between 130 C and 145 C. 0 Temperature range of interest 140 C Heat flow (m W -1 ) bagasse lignin (NaOH extracted) 30 % bagasse in [C4mim]Cl 30 % bagasse in [C2mim]OAc 122 C 137 C 129 C Temperature ( C) Figure 4.1.7: Differential scanning calorimetry profiles 89

114 4.1.3.b Thermogravimetric analysis (TGA) of bagasse It has been demonstrated by DSC analysis in Section a, that bagasse undergoes structural changes when reacted in [C4mim]Cl at temperatures between 130 C and 145 C. It has also been demonstrated in Section that some of the starting bagasse is lost post [C4mim]Cl dissolution and precipitation with water. The amount of these losses has been extensively investigated, however their composition needs to be understood. TGA (as described in Section 3.10) was used to determine whether any mass of reactants is lost to volatile molecules at the targeted temperature range. The peaks of the first derivative curve of TGA curves are an indication of the temperature at which thermal decomposition is fastest. In Figure the peak at ca. 100 C is a result of moisture loss. The first bagasse component to degrade is the hemicellulose as it is the most thermolabile of the biomass components [147]. Both hemicellulose and bagasse degrade at lower temperatures in the presence of [C4mim]Cl. [C4mim]Cl is more thermally stable in IL / bagasse mixtures than by itself. Accordingly, Wendler et al. [148] have demonstrated that [C2mim]OAc is more stable in IL / cellulose mixtures than by itself. The thermal decomposition temperatures for both bagasse and bagasse in [C4mim]Cl lie significantly above the targeted temperature range of 110 C to 160 C used in the pretreatment experiments of this study. Therefore, bagasse losses measured in Section must be predominantly due to biomass depolymerisation towards IL / water soluble molecules as opposed to losses of volatile molecules. 90

115 ΔwΔT -1 / mg C bagasse bagasse hemicellulose 30 % bagasse in [C4mim]Cl 3 % hemicellulose in [C4mim]Cl [C4mim]Cl Temperature ( C) Figure 4.1.8: First derivative of thermogravimetric analysis curves c Cellobiose hydrolysis in [C4mim]Cl The hydrolysis of glycosidic bonds is likely to be one of the main reasons for losses incurred upon biomass dissolution in ILs and the rate of this hydrolysis needs to be understood. The rate of cellobiose hydrolysis to glucose and subsequent glucose degradation in [C4mim]Cl (as described in Section 3.11) at two different temperatures are plotted in Figure The hydrolysis of the cellobiose glycosidic linkage is considered to be a model for the hydrolysis of glycosidic linkages in fully dissolved cellulose. 91

116 When cellulose or biomass is dissolved in a chloride imidazolium ionic liquid, the glycosidic bond hydrolysis has been shown to occur at random points across the cellulose chain [117]. This random chain scission is not expected to result in significant glucose formation. Nevertheless, the stability of glucose monomers under these reaction conditions is also shown in this experiment. Both cellobiose hydrolysis rate and glucose degradation rate increase with increasing temperature. Although it is expected that glucose would be less stable at 150 C than at 130 C, glucose initially accumulates at 150 C, but does not accumulate at 130 C. This initial accumulation may be attributed to the mechanisms of cellobiose hydrolysis and glucose decomposition. Glucose decomposition proceeds either by Lobry de Bruyn - van Ekenstein rearrangement to fructose (which is less stable than glucose) and subsequent decomposition, or via ring opening and retro-aldol condensation (e.g. initially to glyceraldehydes and dihydroxyacetone). In the absence of water, cellobiose hydrolysis proceeds by electrophilic attack of hydrogen ions on the glycosidic oxygen lone pair of electrons and initially results in the formation of glucose and a glucose carbocation which then reacts with water to form glucose and regenerates the hydrogen ion. The carbocation product formed under anhydrous conditions (1,6-anhydro-β-D-glucopyranose - see Figure ) is more thermally stable than glucose [149]. Consequently, at 130 C cellobiose in the presence of small amounts of water initially forms glucose which rapidly decomposes. Later in the reaction, after the water has been consumed or lost, 1,6- anhydro-β-d-glucopyranose accumulates. At 150 C and with little or no water present, 1,6-anhydro-β-D-glucopyranose accumulates early in the reaction. While it has been established that glycosidic bond cleavage is rapid for saccharides in solution at high temperatures, the extent to which this bond cleavage contributes to cellulose losses depends on the DP of the cellooligomers resulting from the aforementioned random chain scission (low DP cellooligomers are water soluble and are expected to be lost as they will not precipitate on addition of water antisolvent). 92

117 60 40 Cellbiose 130 C Cellbiose 150 C Glucose 130 C Glucose 150 C Glucose loss 130 C Glucose loss 150 C concentration in [C4 mim]cl ( mg g- 1 ) Time (min) Figure 4.1.9: Cellobiose hydrolysis and glucose accumulation in [C4mim]Cl Figure : Hydrolysis of cellobiose in the absence of water 93

118 Table 4.1.1: Compositional analysis of bagasse pretreated with [C4mim]Cl and dilute acid sample Mass recovery Ash AIL ASL Total lignin % dry mass ratios Glucan Xylan Arabinan Acetyl Arab/ xylan Acetyl/ xylan Untreated DIL ACID IL 140 C IL 150 C

119 4.1.4 Ionic liquid pretreatment comparison with dilute acid pretreatment The total recovered solids (TRS, described in Section 3.13) from water precipitation of [C4mim]Cl-treated bagasse (at the optimised dissolution conditions determined in Section 4.1.2) and dilute acid-pretreated bagasse (prepared according to the standard NREL process described in Section 3.14) were analysed compositionally and compared for saccharification performance a Compositional analysis The composition of untreated bagasse, the TRS of bagasse treated with [C4mim]Cl (for 90 min at 140 C and 150 C) and bagasse treated with dilute acid was analysed according to Section 3.12 and the results are shown in Table The composition of the untreated bagasse indicates that the substitution of GAX hemicellulose is 0.8 arabinosyl and 1.1 acetyl groups for every 10 xylose units (glucuronyl groups were not measured). This is in near agreement with the GAX hemicellulose substitution reported in the literature (1.3 arabinosyl and 1.2 acetyl groups for every 10 xylose units as seen in Section 2.2.2) and confirms that the GAX structure depicted in Figure is representative of the GAX structure found in the starting bagasse of this study. The dilute acid-treated solids have low arabinoxylan and acetyl content and are enriched in lignin and cellulose (glucan). It is known that dilute acid pretreatment of biomass removes hemicelluloses and this dissolution may be accompanied by hydrolysis of ester bonds between hemicelluloses and lignin. Consequently, low xylan and arabinan content can be expected. Ionic liquid treatment at 140 C imparts low losses (7 % mass), consequently few compositional changes are measurable and the recovered solids appear compositionally almost identical to the untreated bagasse. In IL treatment at 150 C the arabinoxylan is removed similarly to the dilute acid treatment. However [C4mim]Cl appears to be less effective at removing acetyl groups. It is interesting that arabinan seems to be selectively removed by the ionic liquid (cf. dilute acid pretreatment). Lee et al. [68] and Fu et al. [143], have investigated the effect of [C2mim]OAc pretreatment on wood flour and wheat straw, respectively. The 95

120 compositional changes of pretreated solids with increasing temperatures up to 150 C were reported at 90 min of reaction time in both works. The most pronounced difference in their results, when compared with those reported here, is that the lignin of their solids diminishes with increasing temperature. Although these studies are all on different plant substrates, it is probable that lignin is more soluble in water /[C2mim]OAc mixtures than in water /[C4mim]Cl mixtures and consequently there is less lignin precipitation upon addition of water. In fact, the ph of 0.5 water : IL mass ratio for [C4mim]Cl is ca. 6.6 and for [C2mim]OAc is ca. 7.9 (see Figure 5.3.1). This ph difference may explain variable lignin recovery (i.e. lignin is generally alkali soluble) and will be investigated in more detail in Section b Enzyme saccharification The enzyme saccharification of cellulose in the untreated bagasse, the TRS of bagasse treated with [C4mim]Cl(for 90 min at 140 C and 150 C), the DS (dissolvedthen-precipitated fraction only, excludes the undissolved fraction) of bagasse treated with [C4mim]Cl (for 90 min at 150 C) and bagasse treated with dilute acid was monitored according to Section 3.15 and the results are shown in Figure Both ionic liquid and dilute acid pretreatments imparted high cellulose saccharification rates and extents as compared to those of untreated bagasse. The enzyme saccharification rate and extent of the TRS from ionic liquid treatment at 150 C is similar to that of the dissolved bagasse fraction only (DS) from ionic liquid treatment at 150 C. Its saccharification reaches a practical endpoint in less than 3 h at which point its saccharification extent (93 % mass) is more than twice as high as that of TRS from ionic liquid treatment at 140 C (42 %) and dilute acid treatment (31 %). At 24 h the [C4mim]Cl treatment at 150 C still imparts close to two times more cellulose saccharification than dilute acid treatment (viz. 96 % cf. 55 %). Comparatively, in 2010, Li et al. [76] reported 24 h cellulose saccharification of [C2mim]OAc treated switchgrass to be ca. two times higher than dilute acid treatment (viz. 96 % cf. 48 %). It is noteworthy that the IL pretreatment reported by Li et al., using a higher cellulase enzyme loading than used here, imparts a practical saccharification end-point only after 24 h (cf. 3 h in Figure ). 96

121 It is here indicated that Ionic liquids can outperform dilute acid as a pretreatment for bagasse while temperature plays a pivotal role in the performance of ionic liquid treatment. It may also be deduced from the data in Figure that complete dissolution is not necessary to maximise saccharification efficiency. [C4mim]Cl treatment of bagasse at 150 C for 90 min imparts 52 % mass dissolution (viz. Section b.), yet the bagasse partially dissolved under these conditions (TRS) exhibits a similar saccharification profile to that of completely solubilised bagasse (DS in Figure ). The sudden large increase in enzyme saccharification rate, between 140 C and 150 C, has recently (2010) been reported by Arora et al. [102] for switchgrass treated with [C2mim]OAc. These authors measured two times the initial rate of total reducing sugars released at 150 C than at 140 C and attributed this phenomenon to the glass transition temperature of lignin, without providing direct evidence. 100 glucan in pretreated soilds ( % mass) C DS IL 150 C IL 140 C Dil Acid Untreated Time (h) Figure : Enzyme saccharification of bagasse pretreated with [C4mim]Cl and dilute acid 97

122 The effect of the temperature increment from 140 C to 150 C on the structure of IL-treated bagasse can be seen in Figure At 140 C and 90 min in [C4mim]Cl (5 % mass bagasse in IL), the structure of the fibre is still discernible while at 150 C, and otherwise identical reaction conditions, the pretreated bagasse looks more like a paste. Pretreated bagasse at 140 C (ca. 50 % moisture) Pretreated bagasse at 150 C (ca. 70 % moisture) Figure : Images of [C4mim]Cl-pretreated bagasse at 140 C and 150 C In Figure , the initial saccharification rates of both cellulose and hemicellulose (as xylan) and the XRD-derived crystallinity indices (described in Section 3.16) of solids recovered from IL pretreatment (TRS at 140 C and 150 C for 90 min) are compared to those of untreated and dilute acid treated bagasse. Hemicellulose saccharification is less rapid than that of cellulose for all solids. This may be a result of hemicellulose being covalently linked to lignin and thus forming part of the enzyme-recalcitrant lignin-hemicellulose fraction of bagasse. It may also be related to the fact that hemicellulose saccharification is very slow due to the low concentration of xylanases in the Accelerase 1000 enzyme cocktail used here. As expected, the crystallinity index seems to be inversely related to the initial saccharification rates of all solids. Interestingly, the crystallinity index 98

123 of IL treatment at 140 C is three times higher than that at 150 C, despite the fact that the extent of dissolution at 150 C is only ca. twice that at 140 C (as reported in Section b). Preferential dissolution of the non-crystalline component of cellulose is not surprising, but it would appear that temperatures higher than 140 C are required to either dissolve or effect decrystallisation of the crystalline component. It is notable that although only ca. 50 % of the bagasse dissolved in the 150 C pretreatment, the crystallinity of the recovered material is less than ¼ of the original bagasse. At 150 C in [C4mim]Cl, bagasse need not be dissolved to disrupt the crystal regions of the bagasse. This is a key finding. initial saccharification rate (% mass h -1 ) glucan xylan CrI Untreated DIL ACID IL 140 C IL 150 C CrI Figure : Initial rates of enzyme saccharification and XRD crystallinity indices for IL- and dilute acid-pretreated bagasse (TRS) In Figure , the final saccharification yields after 121 h of incubation with enzymes are plotted. At 121 h, the saccharification is considered complete (i.e. no further saccharification is expected beyond this point). While 98 % mass cellulose saccharification is reached by the ionic liquid treatment at 150 C, the extents of saccharification of the rest of the pretreatments are still much lower. For example, dilute acid reaches a maximum of only 72 % cellulose conversion. Final 99

124 hemicellulose saccharification is consistently lower than that of cellulose and this is particularly pronounced for the IL pretreatment at 150 C. As compared to untreated bagasse, the IL treated solids at 150 C have a 30 % lower xylan to lignin content ratio (see Table 4.1.1) and it was thus deduced that the lignin bound hemicellulose is more recalcitrant to the IL. The higher content of lignin-bound hemicellulose in the IL treated solids at 150 C as compared to the other solids may be responsible for its less complete saccharification, particularly since lignin is a known inhibitor of enzyme saccharification. saccharification extent (% mass) glucan xylan Untreated DIL ACID IL 140 C IL 150 C Figure : Glucan and xylan saccharification extent after 121 h for IL- and dilute acid- pretreated bagasse (TRS) c Saccharification and fermentation of IL treated bagasse The purpose of this experiment was simply proof of concept that these materials can be fermented and no attempt to optimize conditions was made. Nevertheless, some ethanol yield comparisons to dilute acid pretreatment can be made. Bagasse (35 g) in [C4mim]Cl (464 g) was reacted (150 C for 2 h) in the RC1 reactor (described in Section 3.8) cooled to 70 C and precipitated with water (ca. 100

125 300 ml). The recovered solids were saccharified with enzymes (3 days, 5 FPU) and subsequently fermented with yeast (0.5 O.D.), as described in Section 3.17, and the kinetics are shown in Figure The highly saccharified cellulose of [C4mim]Cltreated bagasse (91 % mass of theoretical on the basis of pretreated solids) yielded ethanol equivalent to 76 % of theoretical in < 24 h and possibly actually achieved this in < 16 h (based on the initial fermentation rate and the fermentation kinetics presented in other studies under the same conditions). The 76 % ethanol yield for this experiment is equivalent to a yield of 0.43 g g -1 of glucose or 85 % mass of theoretical yield on the basis of glucose fed to fermentation. The equivalent glucose-to-ethanol conversion efficiency for dilute-acid-pretreatment-derived glucose is 95 % of theoretical yield according to the latest NREL report [150]. These glucose-to-ethanol efficiencies together with the cellulose-to-glucose efficiencies from Figure and cellulose recoveries from Table were used for calculating potential ethanol production from IL and dilute acid-pretreated biomass shown in Table % mass of theoretical (on the basis of pretreated solids) Glucose Ethanol Time (h) Figure : Fermentation kinetics of [C4mim]Cl-treated bagasse after enzyme saccharification 101

126 Table 4.1.2: Comparison of ethanol yields from IL and from dilute acid pretreatment Pretreatment Dilute Acid Ionic Liquid ([C4mim]Cl) Temp ( C) time (h) Pretreatment Saccharification Fermentation Total processing Mass (% of theoretical yield) Cellulose recovery Cellulose saccharification Glucose to ethanol Total ethanol yield From the comparison in Table 4.1.2, IL treatment appears superior to dilute acid both in terms of processing time (16.5 h cf h) and ethanol yield (79 % cf. 52 % mass of theoretical based on starting biomass). The performance of ILs is attributed to the ability to cleave covalent bonds while also decrystallising cellulose and dilute acid is only capable of the former Summary In this section, it was established that at the targeted high temperatures of 110 C to 160 C, dissolution of bagasse in [C4mim]Cl increases with both time and temperature while the decomposition temperature of the reactants is not exceeded. At the early stages of dissolution (e.g. < 75 % mass dissolution), the losses are proportionately low and generally account for about 1/3 of the dissolved material when the temperature is kept at 150 C and time 120 min. At 150 C and 90 min, cellulose is fully recovered and the losses comprise primarily hemicellulose components (viz. xylan and arabinan). As the dissolution nears 100 % mass (e.g. 160 C, 90 min), the recalcitrance of the undissolved material increases and the dissolution rate slows while the losses continue to rise. Bagasse at 50 % moisture dissolves slower in [C4mim]Cl than air-dried bagasse, while no difference is observed between the dissolution rates of air-dried (10 % moisture) and ovendried bagasse (1 % moisture). Using incremental bagasse dosing, loadings of up to 102

127 20.6 % mass were achieved in [C4mim]Cl. Dissolution of bagasse in different ILs was conducted and the effect of the IL ion variation on dissolution and losses were discussed. The acetate IL anions, as compared to chloride anions, appear to afford an increased dissolution to losses ratio while the shorter alkyl chains of imidazolium cations seem to accelerate both dissolution and losses. Cellobiose monomerisation under anhydrous conditions in [C4mim]Cl appeared to induce the accumulation of 1,6-anhydro-β-D-glucopyranose which is more thermally stable than glucose. Fermentation of [C4mim]Cl-pretreated bagasse was successfully conducted and the ethanol yield measured. [C4mim]Cl pretreatment (with partial dissolution at 150 C, 90 min) afforded a much higher ethanol yield than standard dilute acid pretreatment (79 % cf. 52 % mass theoretical on the basis of cellulose in starting biomass) in less than half the processing time (pretreatment + saccharification + fermentation = 16.5 h cf h). This is mainly due to the persistence of crystalline cellulose in dilute acid pretreated solids which limits initial saccharification rates. The saccharification extent at practical saccharification endpoint (reached at 3 h for all materials) achieved from partial dissolution of bagasse in [C4mim]Cl at 140 C and 90 min were slightly higher than those of dilute acid. However, this saccharification extent was more than doubled when the temperature of [C4mim]Cl treatment was increased by 10 C (to 150 C) and it equalled those of bagasse completely dissolved in [C4mim]Cl. Temperature plays a pivotal role in this IL pretreatment, while bagasse need not be completely dissolved to impart high saccharification rates and extents. This, together with the fact that the crystallinity of cellulose (in the IL pretreated bagasse) at 150 C, was markedly lower than that at 140 C, are key outcomes. 103

128 4.2 Role of non-dissolution pretreatment effects on enzyme saccharification In section b, (see Figure ) it was observed that biomass treated in [C4mim]Cl at 150 C, but incompletely dissolved, imparted a similar saccharification profile to completely dissolved material. It appears likely that structural changes of the undissolved fraction contribute significantly to saccharification rate and extent. This section describes the outcomes of experiments where the undissolved material was separated from the dissolved material prior to precipitation (see Section 3.6.2) Compositional analysis In a typical dissolution, bagasse (2.5 g) was reacted with [C4mim]Cl (50 g) for 90 min at 140 C and 150 C. The resulting mass was diluted with 50 ml DMSO and filtered. The residue retained on the filter was recovered and analysed as the undissolved fraction. The filtrate (which contained the dissolved fraction) was precipitated with 100 ml water and centrifuged. The precipitated solid was recovered and analysed as the dissolved solids (DS or dissolved-then-precipitated fraction). As shown in Table 4.2.1, the composition of the undissolved fraction at 140 C does not differ markedly from that of the untreated bagasse, indicating little if any preferential dissolution or recovery of any particular component. The undissolved fraction at 150 C however, differs significantly. It contains half of the original glucose, indicating that at around 150 C preferential cellulose dissolution starts occurring. The acetyl content is high in the undissolved fraction at both temperatures showing the recalcitrance of acetyl groups to [C4mim]Cl. 104

129 Table 4.2.1: Compositional analysis of dissolved-then-precipitated solids (DS) and undissolved solids (UND) from [C4mim]Cl pretreatment of bagasse Mass (g) % dry mass ratios Sample Sample Total Arab/ Acetyl/ Ash AIL ASL Glucan Xylan Arabinan Acetyl dry mass lignin xylan xylan Untreated C UND ca * C DS ca * C UND ca * C DS ca * Table 4.2.2: Effect of residence time on the composition of undissolved bagasse after [C4mim]Cl pretreatment at 150 C Mass (g) % dry mass ratios Sample Sample Total Arab/ Acetyl/ Ash AIL ASL Glucan Xylan Arabinan Acetyl dry mass lignin xylan xylan Untreated min UND min UND n/d min UND * Masses do not correspond to the same experiment but have been estimated from previous dissolution experiments which used the same conditions and the same solids recovery method (see Section b) 105

130 The composition of the dissolved-then-precipitated fractions (DS in Table 4.2.1) is notably rich in cellulose (77 % at 140 C and 82 % at 150 C). Lignin and hemicellulose are also present in the dissolved fractions (albeit at a low percentages), but decrease as temperature increases (i.e. with increasing temperature lignin and hemicellulose content in the DS is reduced). While this is partly due to preferential dissolution of cellulose, it is likely also due to increased degradation of hemicelluloses and lignin in solution to material that is soluble in IL / water (i.e. lower molecular mass products of bond cleavages). Arabinosyl glycosidic bonds are generally the most labile of glycosidic linkages in the lignocellulosic matrix (certainly this is the case in an acidic aqueous environment and also under pyrolysis conditions). It is evident from Table that cellulose is preferentially dissolved by [C4mim]Cl. [C4mim]Cl also dissolves lignin and hemicellulose but to lesser extents. Preferential dissolution patterns were also monitored over time. Bagasse (6 g) was reacted in [C4mim]Cl (120 g) for 30 min, 60 min and 90 min and the undissolved solids separated and characterised (Table 4.2.2). There seems to be little change in composition after 30 min. Over time, the lignin, xylan and acetyl content increase in the undissolved fraction while the cellulose gets preferentially removed (solubilised).the same dissolution behaviour is evident, i.e. initially all components dissolve at the same rate, and then cellulose is preferentially dissolved. Either arabinose is cleaved from the hemicellulose solids or arabinose is cleaved from dissolved hemicellulose. It seems likely from the composition of the dissolved material that both may happen. Cleavage of arabinosyl glycosidic linkages leads to arabinose in solution, unless the arabinose is covalently linked to lignin. Lignin and hemicellulose are more recalcitrant and cellulose is preferentially dissolved. IL treatment results in changes to undissolved hemicellulose, viz. cleavage of arabinosyl glycosidic bonds and enrichment in lignin, xylan and acetyl content. 106

131 4.2.2 Enzyme saccharification The enzyme saccharification profiles of UND fractions, at different IL pretreatment temperatures and times, were exposed to enzyme saccharification and the profiles are shown in Figure It appears that increasing pretreatment conditions severity, in terms of both temperature and time, increases the enzyme accessibility of the undissolved fractions. This result confirms that changes in the undissolved fraction (e.g. swelling) contribute to higher saccharification rates and extent. Starting with the effect of time, at 150 C the difference in saccharification extent at 24 h of the 90 min sample to the 60 min one is as large as the difference of the latter to the 30 min sample. At 90 min and 150 C, the effect of temperature is sudden since the difference of the 24h-saccharification extent of the 150 C sample to the 140 C sample is as large as the difference of the latter to the untreated bagasse. 60 glucan in undissolved solids (% mass) C, 90 min 150 C, 60 min 140 C, 90 min 150 C, 30 min Untreated Time (h) Figure 4.2.1: Saccharification of the undissolved bagasse after [C4mim]Cl pretreatment at different conditions 107

132 The initial rates of saccharification and XRD crystallinity indices for the undissolved and dissolved fractions from [C4mim]Cl treatment at 140 C and 150 C are presented in Figure and final saccharification yields in Figure The crystallinity indices of all dissolved fractions are lower than untreated material. Interestingly, while the crystallinity index of the undissolved fraction at 140 C is higher than that of the dissolved material, the crystallinity of the undissolved and dissolved fraction at 150 C are similar. This explains the higher initial cellulose saccharification rates of the 150 C undissolved fraction at 26.4 % h -1 as compared to 11.6 % h -1 for 140 C. Initial hemicellulose saccharification rates are very slow (< 5 % h -1 ) for all solids, except the ones treated at 150 C (both DS and UND). For the UND fraction, increased perturbation/swelling of the LCB structure at 150 C allows more rapid hemicellulose saccharification. For the DS fractions, the persistence of lignin-hemicellulose bonds results in low hemicellulose saccharification rates. 120 glucan xylan CrI 0.70 initial rate of saccharifcation (% mass h -1 ) CrI 0 Untreated 140 C UND 140 C DS 150 C UND 150 C DS 0.00 Figure 4.2.2: Initial rates of enzyme saccharification and XRD crystallinity indices for [C4mim]Cl-pretreated bagasse fractions 108

133 While the initial cellulose saccharification rate of the undissolved fraction at 150 C is remarkably high, the final saccharification yield after 121 h (the effective endpoint) is below 60 % mass. Given that this fraction has low crystallinity but high lignin content, its saccharification pattern is in agreement with the analysis of Holtzapple and co-workers [10, 11], who suggest that crystallinity plays a role in the initial saccharification rates while lignin content limits mainly the final saccharification yield. It is also interesting that while the dissolved fractions at both temperatures reach full saccharification of cellulose (100 % mass) they only reach about 70 % hemicellulose saccharification. The xylan content of these fractions is only 7 % to 9 % mass and it is possible that this remaining xylan fraction is close to lignin-hemicellulose bonds, which inhibit the activity of xylanase enzymes; this would confirm that some of the hemicellulose has been solubilised with its lignin bonds intact. However, this is only a speculation as it is not possible to determine from these data which bonds are preserved upon dissolution. saccharification extent (% mass) glucan xylan 10 0 Untreated 140 C UND 140 C DS 150 C UND 150 C DS Figure 4.2.3: Glucan and xylan saccharification extent after 121 h for [C4mim]Clpretreated bagasse fractions 109

134 4.2.3 X-Ray diffractometry (XRD) of bagasse The cellulose crystal structures of the undissolved solids at 140 C and 150 C were analysed using XRD diffractograms (acquired according to Section 3.16) shown in Figure In the comparison of 140 C to 150 C pretreated material, there is evidence of decrystallisation as indicated by the absence of the peak at 16 in the 2θ range. There is also evidence of a transition of cellulose to its high temperature phase as indicated by the shift of the peak at 22 towards lower 2θ angles (20.5 ) (see arrow on Figure 4.2.4). Hori et al. [151] attribute this shift to the lateral thermal expansion of cellulose crystals, characteristic of its transition to its high temperature phase counts untreated 140 C UND 150 C UND θ[ ] Figure 4.2.4: Diffractograms of undissolved bagasse after [C4mim]Cl pretreatment 110

135 4.2.4 High temperature phase of crystalline cellulose Structural studies of naturally occurring crystalline cellulose (cellulose I) using XRD and dynamic molecular modelling have reported a change in its crystal phase at high temperatures described as high temperature phase [151, 152]. This change occurs suddenly at a specific temperature and takes the form of an anisotropic thermal expansion of the unit cell of cellulose I. This thermal expansion increases the volume of the unit cell and results in a sudden swelling of the whole cellulose crystal structure [151]. Dynamic molecular modelling shows that the hydrogen bonding of the high temperature phase results in a less crystalline cellulose structure with fewer and weaker inter-chain hydrogen bonds [152]. For isolated cellulose I, XRD studies report the high temperature phase transition to occur at 180 C [151] while dynamic molecular modelling places it at around 176 C [152]. This transition may occur at lower temperatures in ionic liquid reacted cellulose due to the capacity of Cl - to disrupt hydrogen bonds. These interactions may also be responsible for reducing the thermal decomposition temperature of bagasse hemicellulose by about 50 C when reacted in IL as seen in the TGA shown in Figure Transition to the high temperature phase of cellulose is likely to be a critical first step in rapid dissolution of the crystalline regions in native cellulose. The increased swelling and reduced crystallinity of the cellulose structure should facilitate the diffusion of the ionic liquid and also enhance the enzyme accessibility of cellulose chains remaining in the swollen undissolved biomass fractions. The cellulose decrystallisation measured in the undissolved fraction is most likely a consequence of extensive swelling that bagasse cellulose undergoes prior to dissolution. This mechanism of extensive swelling corresponds to the cellulose dissolution mode 2 as described by Cuissinat et al. [67] for cellulose microfibrils in NMMO (viz. Large swelling by ballooning followed by dissolution ). This ballooning pattern has also been observed in the swelling of pine sulphate pulp fibre and swelling was primarily taking place in the secondary cell wall (see Figure ). 111

136 At this point it has to be noted that swelling of cellulose or LCB without phase transition of the cellulose crystal structure is also possible and it may take place with protracted exposure to chemical swelling agents at low temperatures. However, at the high temperatures used in this study, swelling and phase transition of cellulose mostly happen concomitantly and the phenomena are difficult to distinguish from each other. Recently (2009) other studies have been published that support this swelling pattern upon IL pretreatment. Singh et al. [131] have shown fluorescence microscopy images of switchgrass cross sections exposed to [C2mim]OAc and heat; they observed swelling primarily in the secondary cell walls prior to dissolution. According to Lee et al. [68], [C2mim]OAc pretreated maple wood flour resulted in comparatively little dissolution but effected high enzyme saccharification rates. This was attributed to the extensive delignification, swelling and decrystallisation of the undissolved wood flour. Vanoye et al. [117] have shown swelling of miscanthus grass in optical microscopy images (see Figure 4.2.5). t 0 t = 2h t = 20h 100 C, 0.5 mm mesh (white arrow is 0.34 mm) Figure 4.2.5: Optical microscopy images showing swelling of miscanthus grass particles in [C2mim]Cl (from Vanoye et al. [117]) 112

137 4.2.5 ATR-FTIR analysis of undissolved fractions A selection of [C4mim]Cl-treated bagasse fractions were analysed using infrared spectroscopy (as described in Section 3.18). Figure shows FTIR spectra of the IL undissolved fractions at 140 C and 150 C and the dilute acid pretreated bagasse. Table lists the assignments of the absorption bands of interest. The 1236 cm -1 absorbance is generally attributed to syringyl ring and C-O stretching vibration in lignin, xylan and ester groups [153]. However, the syringyl lignin content of bagasse is very low so, in this study, this band is going to be associated mostly with the C-O stretching vibration of ester groups. The acetyl, lignin and C-O ester bands (1730 cm -1, 1510 cm -1 and 1236 cm -1 respectively) are more intense for the undissolved fraction at 150 C than for the untreated bagasse. This indicates that the IL recalcitrant fraction is rich in lignin, acetyl groups and ester groups/linkages. In agreement with the compositional analysis above (Table 4.2.1), the FTIR analysis shows that [C4mim]Cl dissolves cellulose leaving the UND fraction rich in lignin and hemicellulose. A similar conclusion is presented by Sun et al. [101] in a recent publication of FTIR analysis on undissolved fractions of wood (oak, pine) in [C2mim]OAc, which states that ligninbound carbohydrates are more recalcitrant to IL dissolution. The spectra in Figure also show that dilute acid treatment removes more acetyl groups and hemicellulose-lignin ester bonds than IL treatments. In addition the OH band at 1100 cm -1 is more pronounced in dilute acid indicating higher cellulose crystallinity [76]. This interpretation of the FTIR spectrum and of dilute acid treated bagasse is in agreement with the analysis on dilute acid treated biomass by Kumar et al. [54]. 113

138 Dilute acid treated bagasse [C4mim]Cl undissolved fraction (UND) at 150 C [C4mim]Cl dissolved reprecipitated (DS) fraction at 150 C Untreated bagasse POLYSACCHARIDES LIGNIN Wavenumbers (cm -1 ) Figure 4.2.6: FTIR spectra of IL- and dilute acid-pretreated bagasse fractions (absorbance common scale) The relative abundance of ester bonds compared to lignin content can be indicated by comparing the relative FTIR absorbances at 1236 cm -1 (for ester bonding) and 1514 cm -1 (for lignin content). Table shows these ratios for the spectra in Figure Dilute acid pretreatment affords a solid that is rich in lignin and cellulose and with hemicellulose largely removed. The FTIR band ratio is low indicating that ester bonds have been broken in the pretreatment process. Untreated bagasse, IL UND and IL DS have the same FTIR band ratios indicating that ester bonds between lignin and hemicellulose are preserved in the IL pretreatment. 114

139 Table : Assignments of FTIR-ATR absorption bands for bagasse (from [ ]) Band position Assignment (cm -1 ) 1730 C=O stretching vibration in acetyl groups of hemicelluloses 1600 C=C stretching vibration in aromatic ring of lignin 1510 C=C stretching vibration in aromatic ring of lignin 1421 CH 2 scissoring at C(6) in cellulose 1368 Symmetric C H bending in cellulose 1236 C-O stretching vibration in ester groups 1100 O-H association band in cellulose and hemicelluloses (associated with crystalline cellulose) 1030 C-O stretching vibration in cellulose and hemicelluloses 974 C-O stretching vibration in arabinosyl side chains in hemicellulose 895 Glucose ring stretch, C1-H deformation Table 4.2.4: Ratios of FTIR absorbances attributed to ester bonds and the aromatic ring of lignin. Band heights Ester bonds Lignin aromatic ring baseline (cm -1 ) band (cm -1 ) Height ratio 1236 cm -1 / 1514 cm -1 Untreated Dilute acid UND DS FTIR spectra confirm previous observations that lignin and lignin-bound hemicellulose are recalcitrant to IL dissolution. If selective removal of arabinan in the IL is associated with concomitant reduction in ferulate cross-linking, saccharification of IL-treated solids would be enhanced over and above the decrystallisation effect. Ferulic acid esters are common in grass hemicelluloses (viz. GAXs) and are known to be interlinked to the O-5 position of the arabinofuranosyl branch residues in GAXs. These ferulate esters oxidatively cross-link GAXs and form 115

140 bonds with lignin. These cross-links are assumed to be recalcitrant to herbivore digestion and more generally to enzyme saccharification [26]. However, the arabinan removed by the IL may be ferulic acid ester-linked in which case the removal would mean cleavage only of the ether bond to the xylan backbone. With the data available in this thesis, it is not possible to conclude which of the two arabinan dissolution mechanisms is taking place in the IL Summary Preferential dissolution patterns have been identified which are consistent across different temperatures and times and more pronounced with increasing reaction severity. Cellulose and arabinose are preferentially dissolved by [C4mim]Cl, as opposed to lignin, xylan and acetyl groups. It is also proposed that lignin dissolves in [C4mim]Cl while preserving covalent bonding to hemicellulose. Preliminary FTIR analysis suggests that these bonds may be predominantly ester bonds. This may be the reason for the incomplete final saccharification of hemicellulose in the dissolved-the-precipitated solids (DS). It is clear from the data presented here and elsewhere that in the undissolved bagasse in [C4mim]Cl at 150 C there is evidence of a transition of crystalline cellulose I to its high temperature phase which reduces its crystallinity. This is not the case at 140 C and the transition appears sudden and temperature dependent. While the IL undissolved bagasse (UND) at 150 C is decrystallised to the same extent as the completely dissolved material (DS), the enzyme saccharification extent of the former is still significantly lower than that of the latter. The preferential dissolution imparted by the IL results in a preserved covalent ligninhemicellulose structure that renders the cellulose difficult to access by enzymes regardless of its being decrystallised. It is thus concluded, that while decrystallisation alone accelerates enzyme saccharification of pretreated solids, high yields of monosaccharides (high extents of saccharification) require dissolution of components of the biomass and perturbation of lignin-hemicellulose interactions. 116

141 CHAPTER 5 RESULTS - FRACTIONATION Pretreatment is the process used for opening up the lignocellulosic structure prior to enzyme saccharification. It has to be inexpensive and ideally should enhance both rate and extent of enzyme saccharification. Fractionation implies separation into component parts, i.e. separate lignin, hemicellulose and cellulose. It provides the opportunity to treat three parts separately and differently and therefore has greater value-creating opportunities. Fractionation is a central process when envisioning multiple products from biomass processing in the context of biorefineries. The compatibility of ionic liquids with other solvents and the opportunities for liquid liquid separations of IL solutes make them attractive for fractionation operations. In this chapter, the use of aqueous biphasic systems (ABSs) and preferential cellulose precipitation with antisolvents are investigated as tools for efficient separation of biomass into lignin-rich and polysaccharide-rich fractions. Finally mass balances are determined for three partial dissolutions of bagasse in IL ([C4mim]Cl, [C2mim]Cl, [C2mim]OAc) fractionated with incremental additions of water as an antisolvent. 5.1 Aqueous biphasic systems Ionic liquids form ABSs with aqueous salt solutions and this property has some potential for industrial separations. Ionic liquids are chaotropic or water structure disrupting salts and can be salted out by kosmotropic or water structuring salts such as K 3 PO 4 and Na 2 SO 4 [139, ]. This property has the potential of concentrating IL / water mixtures while at the same time fractionating biomass polymers according to their preference for chaotropic or kosmotropic solutions. For example it has been reported that in polyethylene glycol (PEG)/ salt ABSs, lignin has 117

142 a preference for the polymer-rich (chaotropic) phase as opposed to the inorganic salt-rich (kosmotropic) phase [160]. Moreover, from an industrial point of view, ABSs are attractive because potentially they provide a low energy separation system [158]. One of the objectives of this project is to use ABSs for the separation of the biomass into its component parts. A claim in the patent by Edye and Doherty [3, 134] formed the starting point for ABS experiments in this thesis. It was claimed that addition of 200 g L -1 NaOH in a solution of bagasse in [C4mim]Cl yields a biphasic system comprising a top phase where the IL and water report and a bottom phase where the aqueous NaOH, the lignin in solution and the precipitated cellulose (in suspension) report. In initial experiments in this research, where ca. 1 % mass bagasse (0.166 g) appeared to completely dissolve in [C4mim]Cl ( g) in 60 min at 150 C, subsequent addition of NaOH (200 g L -1, 20 ml), vigorous shaking and let settling for two days, produced an ABS with clear phase separation and a sharp boundary (see Figure 5.1.1). The top phase is IL-rich, chaotropic and dark coloured and the bottom phase is NaOH-rich, kosmotropic and light coloured. The volume of the IL phase (top) is larger than the volume of the original IL due to water migration from the NaOH (bottom) phase. A polysaccharide-resembling white fluffy solid accumulated at the top of the NaOH phase while the dark colour of the top phase suggests a lignin rich solution. At a 5 % mass bagasse loading (0.330 g) in [C4mim]Cl (6.3 g, at 150 C for 170 min), the addition of aqueous NaOH (8.4 ml) in the same proportion did not result in an ABS. Twice as much aqueous NaOH (16.8 ml) was needed to separate the phases. It is likely that more water was needed to satisfy the hydration requirement of the biomass solutes and effect phase separation. FTIR spectra of upper and lower phase of the 5 % bagasse and a biomassfree NaOH / [C4mim]Cl ABS were acquired (as described in Section 3.18) and are shown in Figure It is apparent that the biomass-free ABS phases are clearly of different composition whereas spectra of the ABS with 5 % bagasse have similar features, indicating more mixing of the phases. Therefore the formation of ABS in 118

143 the absence of biomass was investigated (see Sections to 5.1.3). Notwithstanding this problem, the literature reports deprotonation of dialkyl imidazolium ions under alkali conditions to form neutral carbenes. This reaction may lead to loss of ILs and consequently is also investigated here (see Section 5.1.5) IL phase (containing lignin?) Cellulose-resembling fluffy solid NaOH phase Figure 5.1.1: A NaOH / [C4mim]Cl ABS with 1% mass bagasse load In order to examine the phase preference of lignin in the [C4mim]Cl / NaOH ABS, 1.4 g of soda lignin were dissolved in 9.7 g of [C4mim]Cl (170 C for 22 min). Aqueous NaOH solution (20 g L -1, 13 ml) was added and the reaction shacked vigorously and left for a week to phase separate. Each layer was sampled and their infrared spectra acquired (see Figure 5.1.3). The characteristic lignin band absorption around 1510 cm -1 has shifted to 1496 cm -1 and it is only discernible at the spectrum of the [C4mim]Cl phase of the ABS. This indicates that the lignin has a preference for the IL phase of this system. The 1510 cm -1 band is clearly visible in the reference spectrum of 15 % lignin in [C4mim]Cl while it is absent from the neat [C4mim]Cl background. 119

144 [C4mim]Cl phase of biomass-free ABS [C4mim]Cl phase of 5% bagasse loaded ABS NaOH phase of 5% bagasse loaded ABS NaOH phase of biomass-free ABS Wavenumbers (cm -1 ) Figure 5.1.2: FTIR spectra of each phase of two NaOH / [C4mim]Cl ABSs Regarding the distribution of lignocellulosics within such ABSs, the original claims in the patent by Edye and Doherty [3, 134, 161] were that cellulose precipitates and the lignin remains in solution in the inorganic salt layer. Visual observations of the ABS (Figure 5.1.1) and FTIR analysis (Figure 5.1.3) suggest that the cellulose precipitates at the top of the inorganic salt phase and lignin remains in solution in the [C4mim]Cl phase. It s also worth stressing that Rogers and coworkers have reported the separation of lignin from cellulose in a PEG / NaOH ABS [160]. In this system, the cellulose is indeed precipitating in the NaOH phase of the ABS and lignin remains in the chaotropic PEG phase. Willauer et al. [162] have established that cellulose demonstrates a clear preference for the NaOH aqueous phase while the three types of lignin studied (Indulin AT, Indulin C, and Reax 85A) show a preference for the PEG phase. These authors report that the partitioning of the three lignins investigated is affected by the free energy of hydration of the salt forming the ABS. They also stress that both cellulosic samples used (fibrous 120

145 cellulose and diethylaminoethyl cellulose) are of hydrophilic nature and they do not dissolve, but rather report to the salt-rich phase of an ABS [162]. [C4mim]Cl phase of 15 % soda lignin ABS NaOH phase of 15 % soda lignin ABS [C4mim]Cl 20 % soda lignin in [C4mim]Cl 1496 cm -1 Wavenumbers (cm -1 ) Figure 5.1.3: FTIR spectra of each phase of a NaOH / [C4mim]Cl ABS loaded with 15 % soda lignin The poor phase separation as shown by the resemblance of the FTIR spectra of the phases in the ABS is an undesirable outcome. Further investigation and use of alternative ABS compositions (e.g. alternative kosmotropic salts) were employed in order to improve phase separation. It was considered practical to start this investigation with biomass free ABSs. The coexistence of biomass, ILs, inorganic salts and their by-products in a biphasic system would render analytical data complex and difficult to unravel. Therefore the ABSs that seemed prima facie to have an application in the fractionation of IL-biomass solutions were initially analysed excluding biomass. 121

146 5.1.1 Choice of kosmotropic salts for aqueous biphasic systems The stability and purity of each of the two phases in an ABS depend on the choice of kosmotropic salt for a given IL. The salting-out strength of the kosmotropic salts follows the well-established Hofmeister series, as observed in polymer salt ABSs, and can be directly related to the ions Gibbs free energies of hydration ( G hyd ) [139, 157]. The strong dependence of ABS formation on hyd of each of the inorganic salt ions was also demonstrated by Najdanovic-Visac et al. and Trinidade et al. [158, 159]. The Hofmeister Series is a classification of ions from kosmotropic to chaotropic (water structuring and water structure-disrupting respectively). Figure demonstrates this classification and the properties associated with each ion. This classification becomes relevant when selecting salts for salting out ILs. Out of all the kosmotropic salts that are expected to form ABS with [C4mim]Cl, the following salts were selected for investigation in this study. NaOH, since it has been claimed in Edye and Doherty s patent [3, 134] to form a stable ABS with [C4mim]Cl in presence of biomass. NaOH has also been reported by Rogers and co-workers [160] to contain cellulose in PEG / NaOH systems. Na 2 CO 3, since it was observed that as biomass loads increase it is increasingly more difficult to form an ABS in aqueous NaOH. Consequently, a salt with higher G hyd was required. KOH and K 2 CO 3, for the purpose of comparing method outcomes with the literature (viz. Bridges et al. [139]). The ions of these salts are found towards the left side of the Hofmeister series (Figure 5.1.4) denoting a tendency to form water structuring solutions and a high G hyd. The main parameters that affect salting out strength of these salts are the G hyd of their ions (presented in Table 5.1.1) and the solubility of the salts in water (Table 5.1.2). 122

147 THE HOFMEISTER SERIES ANIONS CO 3 2- SO 4 2- Acetate F - OH - Cl - NO 3 - ClO 3 - I - SCN - NH 4 + K + Na + Li + Mg 2+ Ca 2+ guanidinium + CATIONS Surface tension Harder to make cavity Solubility in hydrocarbons Salt out (aggregate) Very negative G hyd More kosmotropic Surface tension Easier to make cavity Solubility in hydrocarbons Salt in (solubilise) Less negative G hyd Less kosmotropic Figure 5.1.4: The Hofmeister series (ions relevant to this study in bold) (reproduced from Jakubowski [163]) Table 5.1.1: Gibbs free energies of hydration ( G hyd ) of selected ions (from Bridges et al. [139]) Cation G hyd /kcal mol -1 Anion G hyd /kcal mol -1 H OH Na CO K Table 5.1.2: Water solubilities of selected inorganic salts (from Lide [22]) Inorganic salt Solubility (g / 100 g H 2 25 o C) NaOH 100 Na 2 CO KOH 121 K 2 CO NaCl 36.0 KCl

148 5.1.2 Evaluation of ABS stability with coexistence curves Biphasic systems that maintain phase separation over a wide range of concentrations of both IL and kosmotropic salts are considered more stable and less sensitive to concentration changes (e.g. changes of water availability in the system). Coexistence curves drawn by the cloud point method (described in Section ) are used for evaluating the stability of ABSs formed between [C4mim]Cl and the selected kosmotropic salts. The curves delineate the concentration threshold between single and two phase systems (viz. aqueous mineral salt solutions and IL water mixtures) and determine their sensitivity to concentration changes. The area under the curve shows concentration ranges for single phase systems. The area above the curve shows concentrations of salts and the ILs that will lead to formation of ABS (notwithstanding other considerations, e.g. the solubility of the mineral salt in water). Accordingly, the larger the area above the curve, the more stable is the ABS. The coexistence curves in Figure represent the author s cloud point titrations for the aforementioned selected salts which formed ABSs with [C4mim]Cl. It is evident (and in agreement with reports of others [139, 158]) that the energy of hydration of the ions comprising these salts determines, to a large extent, the distance from the origin of these curves, i.e. as the Gibbs free energy of hydration of the kosmotropic aqueous solutions becomes more negative, the biphasic systems become more stable and preferred over a greater molality range. 124

149 9 8 NaOH NaOH 7 Na2CO3 2 3 [C4mim]Cl molality A KOH KOH K2CO3 K B salt molality Exponential trend lines have been drawn as visual aids. The horizontal lines (A and B) mark the points where rapid crystal formation and collapse of the ABS occurred. Figure 5.1.5: Coexistence curves of [C4mim]Cl with selected kosmotropic salts 125

150 The difference in energy of hydration between the anions (OH - and CO 2-3 ) is in the order of 150 Kcal mol -1, whereas for the cations (Na + and K + ) it is in the order of 17 kcal mol -1. This pattern is depicted in the positioning of the corresponding coexistence curves in Figure The Na 2 CO 3 and K 2 CO 3 curves have a smaller single phase region (positioned closer to the origin) with the K 2 CO 3 curve slightly closer to the origin. The NaOH and KOH curves on the other hand have a significantly larger single phase region (positioned far from the origin) with the NaOH curve positioned slightly more towards the origin. Bridges et al. [139] report coexistence curves for [C4mim]Cl ABSs with K 2 CO 3 and KOH (see Figure 5.1.6). These curves agree with those reported here. However, Bridges et al. do not report on other possible events, such as salt crystal precipitations, that can occur when the two phases are competing for water. Work by Trindade and co-workers [158, 159] has reported the otherwise overlooked effect of inorganic salt precipitation in ABSs of ILs (viz. [C4mim]Cl, [C4mim]BF 4 and other) with inorganic salts (viz. K 3 PO 4, NaCl, Na 2 SO 4, and Na 3 PO 4 ). In these systems, increasing concentrations of IL reduced the solubility of the kosmotropic salts as compared to their solubility in pure water and led to their precipitation out of the ABS. The horizontal lines in Figure (marked A and B) represent the last cloud point sample before precipitation of the kosmotropic salt and collapse of the ABS occurred. In the case of KOH /[C4mim]Cl ABS, the crystal precipitate was harvested, washed with acetone and analysed with ion chromatography (described in Section ). The analysis indicated that the precipitate was quantitatively pure KCl salt. A metathesis reaction occurs between KOH and [C4mim]Cl to form KCl. Since KCl has a much lower water solubility than the other salt combinations in the mixture (see Table 5.1.2), once the solubility is exceeded KCl crystallises and drives the metathesis reaction. The metathesis reaction is initiated by high concentrations of [C4mim]Cl, rather than high concentrations of KOH. In the cloud point determination, KCl precipitates from a cloudy solution (i.e. a solution on the verge of forming a two phase system). In biphasic systems, where the water solubility of 126

151 KCl is exceeded, it is likely that, under these conditions, Cl - ions migrate to the KOH phase easier than K + ions to the imidazolium phase because the imidazolium cations strongly partition to the imidazolium phase (as shown in Section below). ( ) K 3 PO 4, ( ) K 2 HPO 4, ( ) K 2 CO 3, ( ) KOH [C4mim]Cl molality Salt molality Figure 5.1.6: Phase diagrams of [C4mim]Cl with various salts (from Bridges et al. [139]) Although the solubilities of NaCl and KCl in water are the same (36 g / 100 g H 2 O at 25 o C, see Table 5.1.2) a similar metathesis reaction in the NaOH / [C4mim]Cl system does not occur. This can be explained by comparing the effect of molality on the activity coefficients of KOH and NaOH (see Figure 5.1.7). Although KOH and NaOH have the same ionic activity at low concentrations, at high concentrations, the KOH activity is comparatively much higher. This phenomenon of higher ion activity of K + at high concentrations is further accentuated given that the KOH aqueous solution originally added, in the ABS gets further concentrated by the migration of water towards the [C4mim]Cl phase (in order to satisfy the IL s hydration requirement). 127

152 Activity coefficient KOH NaOH Molarity Figure 5.1.7: Activity coefficients of NaOH and KOH at different molarities (values from Lide [22]) In the Na 2 CO 3 / [C4mim]Cl ABS on the other hand, precipitation of Na 2 CO 3 occurs at high concentrations of Na 2 CO 3. Yet this is not the case in the counterpart cation system with K 2 CO 3 ABS since there is no precipitation observed at any point during the cloud point titration. This is simply attributed to the fact that the maximum solubility of K 2 CO 3 (111 g / 100 g H 2 O) is much higher than that of Na 2 CO 3 (30.7 g / 100 g H 2 O) (see Table 5.1.2). Aside from salt crystal precipitation, some gas bubble formation was observed upon mixing of [C4mim]Cl with the Na 2 CO 3 solution. This gas is CO 2. Deprotonation of the IL imidazolium ions in presence of concentrated Na 2 CO 3 aqueous solution (as detected in [C4mim]Cl / Na 2 CO 3 ABSs in Section below) produces carbenes and carbonic acid which in turn produces CO 2 and water (hydration equilibrium constant K h = [H 2 CO 3 ]/[CO 2 ] = 1.7 * 10-3 ). The reaction is further driven towards carbene and CO 2 formation by the escape of CO 2 gas from the reaction mixture. 128

153 Na 2 CO [C4mim]Cl 2 NaCl + CO carbene + H 2 O That this reaction did not completely consume all reactants may be attributed to slow carbene formation or to the phase barrier. The consequences of excessive mixing or contact time on these reactions are not known Evaluation of the phase divergence of ABS using distribution ratios Phase divergence is the difference in concentration between phases of components in a biphasic system i.e. as phase divergence increases, there is less miscibility of a solvent and its solutes in the solvent of the other phase and consequently the ABS is more stable. Distribution ratios are used to quantify the phase divergence of ABSs. The calculation of these ratios is explained in Equation 7. The ion concentrations of each ABS phase, required for the equation, were measured according to Section = Equation 7 Where: D Ion distribution ratio between phases c H Ion concentration in phase with higher apparent concentration Ion concentration in phase with lower apparent concentration c L The distribution ratios of the ions participating in a [C4mim]Cl / NaOH ABS are compared to the distribution ratios of the ions participating in a [C4mim]Cl / Na 2 CO 3 and a [C4mim]Cl / KOH ABS (Figure 5.1.8). In agreement with the coexistence curves in Figure 5.1.5, the distribution ratios express the extent of mutual exclusion of each phase among the three ABSs. The distribution ratios for all ions are high for the Na 2 CO 3 system, lower for NaOH and very low for KOH. Na 2 CO 3 forms a stable ABS since it exhibits a large region of two phases at lower concentrations as shown by the coexistence curves. Even though the molality of the Na 2 CO 3 ABS in Figure is significantly lower than the KOH and NaOH ABSs, the Na 2 CO 3 partitions better than the hydroxides. However, at high concentrations of Na 2 CO 3, the salt s solubility is exceeded resulting in its precipitation. Carbon dioxide production was also observed in the Na 2 CO 3 / [C4mim]Cl ABS, but this reaction 129

154 required vigorous mixing to occur to any great extent in laboratory experiments because the [C4mim] + has strong preference for the top phase. However it might be difficult to control (and limit) in an industrial setting. The phase divergence of the KOH ABS, is the lowest out of the three, reflecting the extensive miscibility and low competition for water between the two phases. Precipitation of KCl crystals is observed at high concentrations of [C4mim]Cl in the KOH ABS. This is a result of the migration of Cl - ions into the KOH phase resulting in the formation of KCl whose solubility in water is much lower than that of KOH D (Distribution Ratio) D D Cl- Cl - D D Na+ + D D [C4mim] [C4mim]Cl / Na2CO3 / 2 3 [C4mim]Cl / NaOH [C4mim]Cl/ KOH ABS [C4mim]Cl/ Na 2 CO 3 [C4mim]Cl/ NaOH [C4mim]Cl / KOH Salt molality IL molality Figure 5.1.8: Distribution ratios of ions in ABSs and their molal composition Unlike other ions in the ABS, [C4mim] + is not miscible to any extent in the aqueous salt phase. The Cl - ion is able to migrate to the aqueous salt phase and the Na + ion is able to migrate to the imidazolium phase (which contains a significant amount of water). 130

155 As shown in the ion migration diagrams in Figure 5.1.9, the salt precipitation observed in the two aforementioned ABSs is driven by different mechanisms. In the stable ABS with Na 2 CO 3, the precipitation is driven by competition for water. Water migrates to the IL phase and thus the Na 2 CO 3 solubility is exceeded. For KOH the two phases are more miscible and the competition for water is low. The driver for KCl salt precipitation here is the high migration of Cl - ions to the KOH phase, causing the formation of KCl which is far less soluble than KOH in water. Na 2 CO 3 CO 2 [C4mim] + [C4mim]Cl phase Na + H 2 O Cl - Na 2 CO 3 phase Na 2 CO 3 precipitation KOH [C4mim] + K + [C4mim]Cl phase KOH phase H 2 O Cl - KCl precipitation The thickness of the arrows reflects the relative distribution coefficients of ions - not in scale Figure 5.1.9: Ion migration diagrams based on distribution ratios 131

156 5.1.4 Effect of biomass loading on distribution ratios of ABSs The effect of biomass loading on the distribution ratios of aqueous NaOH ABSs with [C4mim]Cl and [C4mmim]Cl are shown in Figure The [C4mim]Cl system was loaded with 5 % bagasse while the [C4mmim]Cl with ca. 1 % only (0.052 g bagasse in g [C4mmim]Cl at 150 C for 60 min, precipitated with 5.2 ml of 200 g L -1 NaOH). Bagasse loading leads to a significant lowering of phase separation capacity (distribution ratio) for all ions. While the imidazolium ions are different for the 1 % mass and 5 % mass biomass loadings, the results still show increasing miscibility (or decreasing ABS phase divergence) as biomass load increases. The difference in distribution ratios between the dialkyl and trialkyl substituted imidazolium ions may be related to their tendency to form carbenes. It was observed that the addition of biomass in model ABSs affected negatively their distribution ratios (see Equation 7). Thus the enhanced separation of ABSs with higher biomass loads required the use of ions with higher G Hyd and the availability of more free water to satisfy the hydration requirements of the lignocellulosic biomass (mainly the polysaccharide fraction since it is highly hydrophilic). All the above leads to the conclusion that the initial proposal for lignin to cellulose separation in an ABS has a real potential, however the distribution of the lignin is probably reverse to what was claimed in the patent by Edye and Doherty [3, 134]. The results from ABS experiments in combination with previously discussed literature [162], indicate that ions which rank higher in the Hofmeister series (higher G Hyd ), such as the [CO 3 ] -2 ion, could contribute towards improved phase divergence on biomass-loaded ABSs. However, as demonstrated earlier, K 2 CO 3 would be better than Na 2 CO 3, since Na 2 CO 3 has a low solubility in H 2 O and can precipitate and collapse the ABS as water migrates to the IL phase. 132

157 80 D (Distribution ratio) D D Cl Cl - D D Na D D [C4mim] or + [C4mmim] or [C4mmim] [C4mim]Cl / NaOH [C4mim]Cl / NaOH + 5 % bagasse [C4mmim]Cl / NaOH [C4mmim]Cl / NaOH + 1 % bagasse Figure : The effect of bagasse loading on the ion distribution ratios in ABSs Chemical instability of imidazolium ILs in alkaline ABSs An issue of significant concern that emerged while experimenting with ABSs where the imidazolium-based [C4mim]Cl mixed with alkali salts is the phenomenon of deprotonation of the imidazolium ring to form carbenes. This reaction is depicted in Figure , where a base attacks the acidic proton on the imidazolium ring to form a neutral carbene [164]. Deprotonation is an undesirable phenomenon because it leads to loss of IL to carbenes or necessitates the use of acid to reprotonate the carbene and recycle the IL. R HC N C H CH N CH 3 alkali R HC N C CH Imidazolium cation carbene N CH 3 Figure : Carbene formation from imidazolium-based ILs (reproduced from BASF [86]) 133

158 Since alkaline salts are used in the ABSs investigated in this project it became necessary to develop a method for determining the extent of imidazolium IL cation deprotonation in alkaline biphasic systems. A titration method was devised. The extent of deprotonation was determined by titration with HCl titrant and ph monitoring. The resulting inflection points are indicators of the number of HCl mmol required to reprotonate the same number of carbene mmol in a 1:1 stoichiometry (see Equation 8 and Figure ). =100 Equation 8 Where: deprot Deprotonation of IL in IL layer (% mol) m H1 Total HCl mol till 1 st inflection point m H2 Total HCl mol till 2 nd inflection point Total IL mol originally added in titration solution m IL ph ph ph ph Δmmol -1 Inflection point 1: neutralization of base (OH - ) HCl (mmol) Inflection point 2: reprotonation of carbenes p H Δmmol -1 (ph HCl mmol -1 ) Figure : HCl titration of the IL phase of a [C4mim]Cl / NaOH ABS 134

159 The presence of two inflection points in the acid titration of the IL phase of the [C4mim]Cl / NaOH ABS indicates that there are two species reacted. After titration of a neat NaOH aqueous solution, the first inflection point was attributed to the neutralisation of OH - ions present in the top phase and the second to the reprotonation of carbenes back to [C4mim] + ions. Imidazolium deprotonation for the selected ABSs was measured by this method and listed in Table Table 5.1.3: Deprotonation of imidazolium IL in top phase of ABSs Salt % imidazolium moles reprotonated by HCl [C4mim]Cl/ NaOH 8 [C4mim]Cl / KOH 8 [C4mim]Cl / Na 2 CO 3 7 [C4mmim]Cl / NaOH 5 Deprotonation of [C4mim]Cl is significant and does not seem to vary greatly with salt used. Surprisingly, Na 2 CO 3 is inducing almost as much deprotonation of the imidazolium ring as NaOH. Given that Na 2 CO 3 is less alkaline than NaOH, this deprotonation was attributed to the loss of CO 2 as described in section The deprotonation extents measured also indicate that the imidazolium ions in the top phase are not quantitatively deprotonated. One way to reduce deprotonation was thought to be the substitution of the acidic proton of the [C4mim]Cl imidazolium ring with a methyl group as found in the ionic liquid 1-butyl-2,3-dimethylimidazolium chloride or [C4mmim]Cl. Surprisingly, the HCl titration of the IL phase of the trisubstituted [C4mimm]Cl / NaOH ABS also exhibited a second inflection point. This indicates that the imidazolium ion still reacted with HCl (5 % mol). Whether this means that the alkaline solution demethylated the imidazolium ion in the C-2 position creating a nucleophilic site or that some transalkylation resulted in a similar effect is not deducible from this data. 135

160 5.1.6 Summary ABSs present an opportunity for the clean fractionation of lignocellulosics. Salts with high G Hyd form [C4mim]Cl ABSs with high degrees of phase divergence. However issues such as deprotonation of [C4mim]Cl in alkaline solutions and the fact that higher biomass loads lead to phase convergence may limit the utility of IL ABS in biomass processing. In an industrial setting, these issues represent less clean fractionation but also difficulty in recycling the IL. Deprotonation can lead to IL mass losses while with increasing miscibility of the two phases, the IL is present in both phases and therefore the cost and energy needed to recover the IL are higher. Due to these technical impediments, experiments with ABSs were not extended to monitoring the enzyme kinetics of the resulting bagasse solids. Single phase separation systems formed the basis of the rest of the experimentation on fractionation systems (viz. section 5.2). Other useful observations attained from this section include the mechanism of chloride salt precipitation in inorganic salt / [C4mim]Cl ABSs affecting phase separation (viz. metathesis reactions taking place between the IL anion and the cation of the kosmotropic salt in ABSs with K 2 CO 3 ). As far as the separation of lignin and cellulose is concerned, the hypothesis of separation is still valid with the only difference that the partitioning of the lignin in each phase of the ABS seems to be opposite to what was originally claimed. Based on the data produced in this project so far, identification of a preferred composition ABS should be possible. 5.2 Aqueous single phase fractionation systems Edye and Doherty s [3, 134] choice of NaOH was based on the belief that since lignin is soluble in aqueous NaOH, it would remain in solution in a single aqueous phase [161]. Cloud point plots show that there are aqueous conditions wherein NaOH and [C4mim]Cl will coexist as a single phase. Therefore it seemed that the use of aqueous NaOH as an antisolvent but at concentrations producing a single phase would be worthy of consideration. 136

161 The lignin solvents that are also polysaccharide precipitation agents were dilute NaOH and aqueous acetone and are compared with water. The deprotonation of the imidazolium IL by dilute NaOH (0.2 M) was tested on biomassfree systems (7.8 g of 0.2 M NaOH mixed with 7.5 g of [C4mim]Cl for 5 min) prior to experimentation, using the HCl reprotonation titration method (titrant 0.05 M HCl) used in Section The resulting titration curve indicated only 0.3 % mol deprotonation. Acetone in water (1:1 volume basis) is miscible with [C4mim]Cl and it has been used by Sun et al. [101] to dissolve lignin and precipitate polysaccharides from a pine wood /[C2mim]Cl dissolution system. Three partial dissolution reactions were carried out at identical conditions (1 g of bagasse in 20 g of [C4mim]Cl at 150 C for 90 min in the setup described in Section 3.6.1) and were precipitated with 20 ml of the three selected antisolvents. The total recovered solids (TRS) were characterised and compared to each other in Table Table 5.2.1: Compositional analysis of total recovered solids (TRS) from partial bagasse dissolution in [C4mim]Cl using different antisolvents % dry mass Dry mass (%) Ash AIL ASL Total lignin Glucan Xylan Arabinan Acetyl Untreated Total recovered solids (TRS) Dilute NaOH (0.2 M) n/d Acetone in water (1:1 v/v) n/d Water n/d Compositionally there is no apparent significant difference between the solids recovered (TRS) with the different antisolvents. Enzyme saccharification of these solids is shown in Figure No differences in initial saccharification rate and little difference in saccharification 137

162 extent were imparted by the antisolvents used. Shown differences in extent may be due to differences in hemicellulose bonding to lignin. Although the rate and extent of saccharification of water-precipitated biomass in Figure appear to be different to previous experiments with the same pretreatment conditions (cf. Figure ), the conclusions from the results in Figure are still reliable as the data is internally consistent. The reason for this discrepancy could be a fault in the heating element of the oil bath resulting in large temperature fluctuations or insufficient heating glucan in pretreated solids (% mass) dilute NaOH water Acetone acetone: : water Untreated untreated Time (h) Figure : Enzyme saccharification of total recovered solids (TRS) from partial bagasse dissolution in [C4mim]Cl using different antisolvents 138

163 In order to more thoroughly examine the fractionation efficiency of these antisolvents they were also tested on bagasse completely solubilised in [C4mim]Cl. The dissolved solids only (DS) from a reaction of bagasse (7.5 g) in [C4mim]Cl(150 g, 150 C, 3 h) were isolated with DMSO (150 ml) and filtration, split in three equal mass portions and each portion precipitated with one of the selected antisolvents (ca. 100 ml). The composition of the precipitated solids from each solvent is presented in Table Table : Compositional analysis of completely dissolved bagasse (DS) precipitated from [C4mim]Cl using different antisolvents % dry mass Dry mass (%) Ash AIL ASL Total Glucan Xylan Arabinan Acetyl lignin Untreated Dissolved fraction only (DS) Dilute NaOH (0.2 M) Acetone in water (1:1 v/v) Water Lignin is reduced to 11.4 % by dilute NaOH and to 13.6 % by acetone in water compared to the 19.2 % present in the water precipitated solids. Both dilute NaOH and acetone in water exhibit an ability to delignify the precipitated solids from the IL solution. The main difference between the two delignifying antisolvents is that dilute NaOH removes the hemicellulose whereas the acetone in water doesn t. The results obtained for the reduction of lignin of [C4mim]Cl treated bagasse when applying precipitation with acetone in water (from 26.2 % to 13.6 %) are in agreement with the literature. Rogers and co-workers [101] have used the same acetone in water mixture to precipitate oak dissolved (98.5 % mass dissolved) in [C2mim]OAc and they reported a reduction of lignin content from 23.8 % mass in the untreated oak to 15.5 % in treated oak. However, it must be noted that unlike Rogers and co-workers, DMSO dilution was used in this work and this may cause 139

164 bias in fractionation results since DMSO / water / IL mixtures may precipitate different biomass fractions than water / IL mixtures. The saccharification time profile of the dissolved reprecipitated fractions is shown in Figure Note that the saccharification in this figure is expressed as % max measured saccharification, where the extent of saccharification of the acetone in water precipitated material at 48 h was assigned a value of 100 %. This was necessary since it became apparent that in this experiment the moisture determination had unacceptably high errors. However, since it is known from previous experiments that completely dissolved and reprecipitated cellulose is quantitatively converted to monosaccharides, the % maximum saccharification is expected to be near equivalent to % actual saccharification. Again the differences in saccharification profiles between antisolvents are small glucan (% mass of maximum measurement) Acetone acetone:water : dilute NaOH water Untreated untreated Time (h) Figure Enzyme saccharification of completely dissolved bagasse (DS) precipitated from [C4mim]Cl using different antisolvents 140

165 5.2.1 Summary In all cases some lignin remained solubilised after the addition of antisolvent. However the delignifying antisolvents (dilute NaOH and acetone in water mixtures) removed more lignin than water. The lignin content of dilute NaOH precipitated solids was 40 % mass lower than the water precipitated ones and the solids precipitated with acetone in water mixtures contained 29 % mass less lignin than water precipitated ones. Dilute NaOH delignification was accompanied with xylan removal which was not the case with acetone in water. The delignification efficiency was only discernible when the dissolved fraction was separated from the undissolved material and precipitated. This indicates that the use of delignifying solvents is only justifiable if enough bagasse has dissolved upon IL pretreatment. Finally the antisolvents used here made little difference in the saccharification of the recovered solids. Consequently water may be the antisolvent of preference for these systems and indeed it is a less costly choice. 5.3 Preferential precipitation by incremental additions of water Following the indication that water is the antisolvent of preference, the potential of incremental additions of water to preferentially precipitate LCB components in IL solution was investigated. Experimentation started with determination of the ph of aqueous IL solutions of different mass ratios. IL (5 g) was placed in a test tube and incremental additions of water were followed by thorough agitation and a ph measurement. The resulting ph measurements as a function of water : IL mass ratio are plotted in Figure for [C4mim]Cl and [C2mim]OAc. Aqueous solutions of [C2mim]OAc are alkaline and ph decreases markedly with increased water : IL mass ratio while water / [C4mim]Cl mixtures are slightly acidic and close to neutral and show a slight increase in ph. Lignin is soluble in alkali, therefore it seemed possible to partially and fractionally precipitate biomass dissolved in [C2mim]OAc by using small amounts of water to precipitate cellulose while maintaining a high ph and thus keeping lignin in solution. 141

166 Addition of more water to lower the ph of the water / IL mixture might then cause lignin precipitation. As an elementary test for the validity of this hypothesis, bagasse soda lignin and Avicel cellulose were dissolved in ILs (ca. 100 mg in 5 g of IL, 150 C, 30 min) and precipitated with incremental amounts of water. Solutions were centrifuged after each water addition (10000 x g) and the water mass at first observed precipitation was noted. The water mass where no more significant precipitate accumulated on the centrifuge pellet was also noted. The results of these observations for lignin and cellulose in each of the ILs are presented in Figure and suggest that cellulose and lignin may indeed precipitate at different water : IL ratios [C2mim]OAc [C4mim]Cl 9 ph water : IL mass ratio Figure 5.3.1: ph of [C2mim]OAc and [C4mim]Cl aqueous solutions at different water : IL mass ratios 142

167 cellulose in [C4mim]Cl cellulose in [C2mim]OAc first precipitation apparent complete precipitation lignin in [C4mim]Cl lignin in [C2mim]OAc water:il mass ratio Figure 5.3.2: Lignin and cellulose precipitation observed at different water : IL mass ratios of [C2mim]OAc and [C4mim]Cl aqueous solutions That this seems to hold true for both Ionic liquids was unexpected. Water / [C4mim]Cl mixtures were not expected to keep lignin in solution since the ph was not alkaline. Observations of precipitate pellet volume changes (in Figure 5.3.2) suggest that for both water / IL mixtures, cellulose precipitated suddenly and almost completely while the precipitation of lignin was more gradual. The likely explanation for these observations is that cellulose is more homogeneous and less complex than lignin. Cellulose precipitates completely before a ratio of 0.5 is reached while most lignin precipitates between 0.5 and 1.5 in both ILs. This outcome suggests that the preferential cellulose precipitation should be achievable in both ILs by using 0.5 water : IL mass ratio. Moreover, increasing this ratio to 2.0 should allow for precipitation of the lignin remaining in solution for the [C2mim]OAc system. More water and possibly some ph lowering with acid may be necessary for recovering the lignin remaining in the water / [C4mim]Cl solution. It is surprising that [C4mim]Cl requires more water than [C2mim]OAc to precipitate dissolved lignin. The lower ph of [C4mim]Cl aqueous solutions does not seem to aid precipitation. Incremental additions of water appear to have potential for fractional precipitation of IL dissolved biomass but ph doesn t appear to be the factor determining lignin precipitation as originally thought. It is of note that for [C2mim]OAc and, at least for the materials used in this experiment, it seems that the last observable precipitation of cellulose may slightly overlap with the first 143

168 observable precipitation of lignin. Finally it must be noted that this analysis is only to be used as a gross indication since it is based on visual assessment of precipitates. 5.4 Comparison of three IL pretreatment and fractionation systems The potential of using incremental amounts of water to fractionally precipitate IL dissolved bagasse was tested on three ILs and the mass balances of each reaction determined (as described in Section 3.21). The schematic representation of the fractional precipitation process designed to yield a polysaccharide-rich and a lignin-rich fraction using two incremental additions of water (and acidification to ph < 1.0 and a third addition of water to precipitate remaining dissolved material) is shown in Figure Preliminary studies of fractional precipitation with water (where Avicel and soda lignin were used) indicated that lignin precipitation was complete at a water : IL mass ratio of 2.0. However native bagasse lignin dissolved in IL is likely to have different properties to lignin extracted with aqueous NaOH and then dissolved in IL. Maximum lignin recovery was ensured by acidification to a ph < 1.0 and the addition of a further 1.5 IL mass equivalents of water. Bagasse pretreatments with [C4mim]Cl, [C2mim]Cl and [C2mim]OAc under identical reaction conditions (35 min at 150 C, 5 % bagasse in IL (2.5 % for [C2mim]OAc), as described in Section 3.21) imparted partial dissolution and the polysaccharide rich solid fractions (solid fraction 1 or SF1 in Figure 5.4.1) were recovered using water addition (water : IL mass ratio of 0.5) as shown in Figure The SF1 solids were washed and freeze-dried prior to analysis and enzyme saccharification. Bagasse was extracted with water and ethanol prior to treatment since better mass balance closures are obtained by removing non-structural molecules which can interfere with characterisation of solid and liquid fractions. 144

169 biomass dissolution precipitation in water : IL mass ratio = 0.5 LIQUID FRACTION 1 (LF1) SOLID FRACTION 1 (SF1) polysaccharide rich precipitation in water : IL mass ratio = 2 LIQUID FRACTION 2 (LF2) SOLID FRACTION 2 (SF2) lignin rich Precipitation in water : IL mass ratio = acidification to ph <1.0 LIQUID FRACTION 3 (LF3) SOLID FRACTION 3 (SF3) lignin rich Figure 5.4.1: Process flow chart of a fractional precipitation separation of IL treated bagasse using incremental additions of water 145