Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Empirical process modeling of the acid catalyzed steam pretreatment of radiata and lodgepole pine Olsen, Colin Andreas Dupont 2012

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2013_spring_olsen_colin.pdf [ 3.44MB ]
Metadata
JSON: 24-1.0073503.json
JSON-LD: 24-1.0073503-ld.json
RDF/XML (Pretty): 24-1.0073503-rdf.xml
RDF/JSON: 24-1.0073503-rdf.json
Turtle: 24-1.0073503-turtle.txt
N-Triples: 24-1.0073503-rdf-ntriples.txt
Original Record: 24-1.0073503-source.json
Full Text
24-1.0073503-fulltext.txt
Citation
24-1.0073503.ris

Full Text

EMPIRICAL PROCESS MODELING OF THE ACID CATALYZED STEAM PRETREATMENT OF RADIATA AND LODGEPOLE PINE  by Colin Andreas Dupont Olsen  B.A.Sc., B.A., The University of British Columbia, 2008  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  MASTER OF APPLIED SCIENCES in THE FACULTY OF GRADUATE STUDIES (Forestry)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver) December 2012 © Colin Andreas Dupont Olsen, 2012  Abstract Ethanol, an alternative liquid fuel, can be produced from sugars derived from lignocellulosic biomass in a bioconversion process that involves pretreatment, enzymatic hydrolysis, and fermentation. Among the different types of biomass investigated for bioconversion, softwoods are readily available in Canada, the US, and Scandinavia. Acid catalyzed steam pretreatment is a preferred method for softwoods due to its ability to effectively recover hemicellulose-derived sugars at moderate operating conditions. More severe conditions are generally required to produce a substrate readily hydrolyzed by enzymes, but because sugar degradation also occurs at these conditions, steam pretreatment is essentially a compromise. Prediction of sugar recoveries from steam pretreated and enzymatically hydrolyzed softwood is desirable for the purposes of process control and steam pretreatment reactor design. In this thesis, efforts were made to determine whether response surface methodology or the thermal severity factors Ro and CS were better suited to the development of empirical models of steam pretreatment. The construction of the thermal severity factor models highlighted the predominance of temperature and time in determining the direct outcomes of the acid catalyzed steam pretreatment of radiata pine. Within a comparison of several response surface methodology models, a hybrid experimental design produced the most robust model because it was developed in conjunction with a narrow process space. Moreover, it was apparent that the response surface methodology models possessed the greater capacity for predicting the direct outcomes of steam pretreatment. In an attempt to overcome limitations identified in the first portion of this thesis, the predictive capability of response surface methodology was further tested using lodgepole  ii  pine ranging in chip size and moisture content. The additional model created demonstrated that response surface methodology could successfully account for feedstock characteristics as well as steam pretreatment operating conditions. Moisture content, but not chip size, was shown to have a significant influence on the combined sugar recovery obtained after SO2 catalyzed steam pretreatment and subsequent enzymatic hydrolysis. In addition, model development was conducted in this portion of the thesis such that the model could form the basis of a more dynamic simulation of the entire softwood to bioethanol process.  iii  Preface A version of Chapter 3 has been recently published. Olsen C, Arantes V, and Saddler J, The use of predictive models to optimize sugar recovery obtained after the steam pretreatment of softwoods. Biofuels Bioprod Bioref (2012); DOI: 10.1002/bbb. I planned and conducted the experiments in coordination with both Valdeir Arantes and Jack Saddler. Writing and editing of the manuscript was undertaken by all three authors.  iv  Table of Contents Abstract .............................................................................................................................. ii Preface ............................................................................................................................... iv Table of Contents .............................................................................................................. v List of Tables .................................................................................................................... ix List of Figures ................................................................................................................... xi List of Abbreviations ..................................................................................................... xiv Acknowledgements ......................................................................................................... xv Chapter 1: Introduction .................................................................................................. 1 1.1  Overview ............................................................................................................... 1  1.2  Availability of lignocellulosic biomass ................................................................ 2  1.3  Current status of bioethanol .................................................................................. 2  1.4  Fuel properties of ethanol ..................................................................................... 3  1.5  Chemical composition of lignocellulosic biomass ............................................... 4  1.6  The lignocellulosic biomass to ethanol process .................................................... 7  1.6.1  Pretreatment .................................................................................................. 9  1.6.1.1 Acid catalyzed steam pretreatment ........................................................ 12 1.6.2 1.7  Enzymatic hydrolysis .................................................................................. 16  Process modeling of pretreatment ....................................................................... 18  1.7.1  Reaction kinetics and mass transfer models ............................................... 19  1.7.2  Thermal severity factors ............................................................................ 22  1.7.3  Response surface methodology................................................................... 26  1.8  Techno-economic models of the softwood to ethanol process ........................... 28  v  1.9  Research objectives and overview ...................................................................... 31  1.9.1  The use of predictive models to optimize sugar recovery in the steam  pretreatment step of a softwood-to-ethanol process ................................................. 31 1.9.2  Effects of chip size and moisture content on the combined sugar recovery in  the steam pretreatment and enzymatic hydrolysis steps of a softwood-to-ethanol process ..................................................................................................................... 32 Chapter 2: Materials and methods ............................................................................... 33 2.1  Raw wood ........................................................................................................... 33  2.2  Preparation of wood chips .................................................................................. 33  2.2.1  Chipping and classification ......................................................................... 33  2.2.2  Adjustment of initial moisture content ....................................................... 35  2.3  Composition of raw wood ................................................................................... 36  2.4  Acid catalyzed steam pretreatment ..................................................................... 37  2.5  Solid liquid separation ........................................................................................ 39  2.6  Enzymatic hydrolysis .......................................................................................... 40  2.7  Analytical methods ............................................................................................. 41  2.7.1  Water insoluble fraction obtained after steam pretreatment ....................... 41  2.7.1.1 Carbohydrates and lignin ....................................................................... 41 2.7.1.2 Extractives and ash ................................................................................. 42 2.7.1.3 Acetyl groups ......................................................................................... 43 2.7.2  Water soluble fraction obtained after steam pretreatment .......................... 43  2.7.2.1 Soluble sugars ........................................................................................ 43 2.7.2.2 Furans and organic acids ........................................................................ 44  vi  2.7.3 2.8  Enzymatic hydrolyzate................................................................................ 45  Experimental design and outline ......................................................................... 46  2.8.1  Comparison of the thermal severity factors Ro and CS ............................... 46  2.8.2  Validation of the response surface methodology model of radiata pine ..... 47  2.8.3  Effect of feedstock characteristics on the combined sugar recovery after  steam pretreatment and enzymatic hydrolysis .......................................................... 48 2.9  Statistical analysis and response surface methodology ...................................... 52  Chapter 3: The use of predictive models to optimize sugar recovery in the steam pretreatment step of a softwood-to-ethanol process .................................................... 54 3.1  Background ......................................................................................................... 54  3.2  Results and discussion ........................................................................................ 55  3.2.1  The thermal severity factor Ro .................................................................... 55  3.2.2  Comparison of the Ro and CS factors .......................................................... 58  3.2.2.1 Residual hemicellulose in the water insoluble fraction .......................... 58 3.2.2.2 Yield of the cellulose-rich water insoluble fraction ............................... 60 3.2.2.3 Sugar recovery in the hemicellulose-rich water soluble fraction ........... 61 3.2.3  Comparison of the Ro factor and response surface methodology ............... 63  3.2.3.1 Residual hemicellulose in the water insoluble fraction .......................... 64 3.2.3.2 Yield of the cellulose-rich water insoluble fraction ............................... 66 3.2.3.3 Sugar recovery in the hemicellulose-rich water soluble fraction ........... 67 3.2.4  Comparison of response surface methodology models .............................. 69  3.2.5  Validation of the response surface methodology model of radiata pine ..... 73  3.3  Conclusion .......................................................................................................... 78  vii  Chapter  4: Effects of chip size and moisture content on the combined sugar  recovery in the steam pretreatment and enzymatic hydrolysis steps of a softwood-toethanol process ................................................................................................................ 80 4.1  Background ......................................................................................................... 80  4.2  Results and discussion ........................................................................................ 83  4.2.1  Effect of conducting steam pretreatment at high solids loading ................. 83  4.2.2  The cellulose-rich water insoluble fraction................................................. 88  4.2.2.1 Composition of the cellulose-rich water insoluble fraction ................... 90 4.2.2.2 Yield of the cellulose-rich water insoluble fraction ............................... 91 4.2.3  The hemicellulose-rich water soluble fraction ............................................ 96  4.2.3.1 Sugar recovery in the hemicellulose-rich water soluble fraction ........... 99 4.2.4  The combined sugar recovery after steam pretreatment and enzymatic  hydrolysis ................................................................................................................ 104 4.2.4.1 Enzymatic hydrolysis of the cellulose-rich water insoluble fraction ... 107 4.2.4.2 Combined sugar recovery..................................................................... 110 4.3  Conclusion ........................................................................................................ 116  Chapter 5: Conclusion ................................................................................................. 119 References ...................................................................................................................... 122 Appendix A: Estimates of model term significance from extended ANOVA .......... 135  viii  List of Tables Table 1-1. Chemical composition of various lignocellulosic biomass. .............................. 7 Table 2-1. Chemical composition of the raw pine prior to steam pretreatment. .............. 37 Table 2-2. Experimental design for H2SO4 catalyzed steam pretreatment. ...................... 47 Table 2-3. Experimental design for validation of the response surface methodology model of radiata pine. .................................................................................................................. 48 Table 2-4. Experimental design for the response surface methodology model of lodgepole pine. ................................................................................................................................... 51 Table 3-1. Fitted coefficients of the response surface methodology model of the water insoluble fraction of steam pretreated radiata pine. .......................................................... 64 Table 3-2. Fitted coefficients of the response surface methodology model of the water soluble fraction of steam pretreated radiata pine. ............................................................. 64 Table 3-3. Maximum predicted sugar recoveries in the water soluble fraction of steam pretreated radiata pine and their corresponding process conditions. ................................ 68 Table 3-4. Fitted coefficients of the response surface methodology model of the water insoluble and water soluble fractions of steam pretreated white fir and ponderosa pine . 71 Table 3-5. Fitted coefficients of the response surface methodology model of the water insoluble and water soluble fractions of steam pretreated Norway spruce ....................... 72 Table 4-1. Aspen Plus simulation of the SO2 catalyzed steam pretreatment of softwood. ........................................................................................................................... 82 Table 4-2. Summary of results for steam pretreated lodgepole pine: inhibitor content of the water soluble fraction. ................................................................................................. 85  ix  Table 4-3. Summary of results for steam pretreated lodgepole pine: water insoluble fraction. ............................................................................................................................. 89 Table 4-4. Fitted coefficients of the response surface methodology model of steam pretreated lodgepole pine: water insoluble fraction. ......................................................... 90 Table 4-5. Summary of results for steam pretreated lodgepole pine: sugar recovery of the water soluble fraction. ....................................................................................................... 97 Table 4-6. Fitted coefficients of the response surface methodology model of steam pretreated lodgepole pine: sugar recovery of the water soluble fraction. ......................... 98 Table 4-7. Fitted coefficients of the response surface methodology model of steam pretreated lodgepole pine: inhibitor content of the water soluble fraction. ...................... 99 Table 4-8. Summary of results for steam pretreated and enzymatically hydrolyzed lodgepole pine: combined sugar recovery. ..................................................................... 105 Table 4-9. Fitted coefficients of the response surface methodology model of steam pretreated and enzymatically hydrolyzed lodgepole pine: combined sugar recovery. ... 107  x  List of Figures Figure 1-1. Simplified process flow diagram of the softwood to ethanol process. ............ 9 Figure 1-2. Reaction pathways of hexose and pentose sugar degradation. ...................... 15 Figure 1-3. Basic reaction pathway of xylan hydrolysis with Arrhenius-type temperature dependence. ....................................................................................................................... 20 Figure 1-4. Alternate reaction pathway of xylan hydrolysis. ............................................ 21 Figure 2-1. Size distribution of the lodgepole pine wood chips prepared for steam pretreatment. ..................................................................................................................... 35 Figure 2-2. Experimental procedure employed in all research portions of this thesis. ..... 46 Figure 3-1. Recovery of hemicellulose-derived sugars in the water soluble fraction of pretreated radiata pine as a function of the thermal severity factor Ro. ............................ 56 Figure 3-2. Yield of the water insoluble fraction as a function of the thermal severity factor Ro. ........................................................................................................................... 57 Figure 3-3. Hemicellulose remaining in the water insoluble fraction as a function of the thermal severity factors Ro and CS.................................................................................... 59 Figure 3-4. Yield of the water insoluble fraction as a function of the thermal severity factors Ro and CS. ............................................................................................................. 61 Figure 3-5. Recovery of hemicellulose-derived sugars in the water soluble fraction as a function of the thermal severity factors Ro and CS. .......................................................... 62 Figure 3-6. Effects of temperature and time on the direct outcomes of the SO2 catalyzed steam pretreatment of radiata pine. ................................................................................... 66  xi  Figure 3-7. Comparison of sugar recoveries in the water soluble fraction obtained after the steam pretreatment of radiata pine at 215 °C, 3 min, and 2.55 w/w % SO2 to response surface methodology model predictions. .......................................................................... 74 Figure 3-8. Comparison of water insoluble fraction yield and composition obtained after the steam pretreatment of radiata pine at 215 °C, 3 min, and 2.55 w/w % SO2 to response surface methodology model predictions. .......................................................................... 75 Figure 4-1. Sugar composition of the highly concentrated water soluble fraction obtained after the steam pretreatment of 5/8 in (16 mm) and 35 w/w % moisture lodgepole pine at 205 °C, 5 min, and 2.5 w/w % SO2................................................................................... 87 Figure 4-2. The effect of moisture content on the yield of the water insoluble fraction obtained after the steam pretreatment of lodgepole pine. ................................................. 94 Figure 4-3. The effect of chip size on the recovery of hemicellulose-derived sugars obtained in the water soluble fraction after the steam pretreatment of lodgepole pine. . 101 Figure 4-4. The effect of moisture content on the recovery of hemicellulose-derived sugars obtained in the water soluble fraction after the steam pretreatment of lodgepole pine. ................................................................................................................................. 103 Figure 4-5. The effect of moisture content on the recovery of all sugars (primarily glucose) obtained after the enzymatic hydrolysis of the water insoluble fraction of steam pretreated lodgepole pine. ............................................................................................... 109 Figure 4-6. The effect of moisture content on the combined sugar recovery obtained after the steam pretreatment and subsequent enzymatic hydrolysis of lodgepole pine........... 111  xii  Figure 4-7. Combined sugar recovery obtained after the steam pretreatment and subsequent enzymatic hydrolysis of lodgepole pine as a function of the water insoluble fraction yield after steam pretreatment. .......................................................................... 113 Figure 4-8. Combined sugar recovery obtained after the steam pretreatment and subsequent enzymatic hydrolysis of lodgepole pine as a function of the extent of glucan hydrolysis in steam pretreatment. ................................................................................... 115  xiii  List of Abbreviations °C $ % BC cm g g hr in kg kJ L log M mg min mL mm mM mol nm prpm TAPPI yr UBC USD UV v/v w/v w/w α β µL µm ω  degree Celsius dollar percent British Columbia centimeter gram gravitational acceleration constant hour inch kilogram kilojoule litre logarithm base 10 molar miligram minute mililitre milimeter milimolar mole nanometer para rotations per minute Technical Association of the Pulp and Paper Industry year University of British Columbia United States dollar ultraviolet volume per volume weight per volume weight per weight alpha beta microlitre micrometer omega  xiv  Acknowledgements I would first like to thank my supervisor Jack Saddler for giving me the opportunity to join this group of very capable researchers and for constantly challenging me. Without having to leave Canada or even my alma mater, I have been fortunate enough to engage with people from China, the United States, Panama, the Philippines, India, Sri Lanka, Peru, Indonesia, Japan, Brazil, and Southern Europe both during and after work hours. The SIM Symposium held in Seattle was immensely enjoyable. Thank you for organizing this Jack, and for introducing me to members of both the pulp and paper and industrial biotechnology sectors. Having these contacts will help me to grow professionally in the years to come. A most deserved thanks go to Valdeir Arantes, my ‘second’ supervisor. Without your tireless efforts to help me plan and conduct experiments, critique my writing, and finally remind me of important dates and deadlines, I would never have completed my graduate studies! To the other members of the FPB group, thank you all for your technical insights, your generous help in the lab, and most importantly for keeping my sanity intact. I also wish to acknowledge the additional members of my committee: Paul McFarlane and Doug Kilburn for volunteering their time and providing me with meaningful advice in a system not always designed to reward such effort. For providing me with the wood chips used in this work, I must thank Thomas Clark, Ken Ewanick, Kerry Rouck, Bernard Yuen and Rodger Beatson. To my parents Dan and Sabine, my brother Erik, and Alysha, I was only able to take this step thanks to your unwavering support!  “The aim of science is not to open the door to everlasting wisdom, but to set a limit on everlasting error” – Bertolt Brecht  xv  Chapter 1: Introduction 1.1  Overview International interest in fuel ethanol derived from lignocellulosic biomass continues to  increase due to concerns over volatile energy prices, energy security, limited supplies of sugar and starch rich biomass for both human consumption and conversion to ethanol, and the environmental consequences of continued fossil fuel dependence.  1-3  According to the  most recent assessment authored by the United Nations International Panel on Climate Change (IPCC), fossil fuel use is the primary driver behind the increasing concentration of atmospheric carbon dioxide (CO2).  4  In fact, the report concludes that the majority of the  increase in average global temperature since the mid-20th century is “very likely” due to this increasing concentration of anthropogenic greenhouse gases (GHGs) which include CO2. Primarily as an effort to reduce the greenhouse gas emissions associated with the Canadian transportation sector, the use of gasoline ethanol blends is now mandated at both the federal and provincial level.  5  In British Columbia, the potential to convert woody  biomass to biofuels and biochemicals comes at an opportune time for the forestry sector, which is actively seeking alternate sources of revenue. In fact, both the Forest Products Association of Canada (FPAC) and FPInnovations (FPI) consider ethanol derived from woody biomass as one of the products which could positively transform the current forest products sector. 6 Small quantities of ethanol are already produced from softwood in Quebec at Tembec’s acid sulphite pulp mill in Temiscaming and Enerkem’s thermochemical facility in Westbury.  7, 8  However, it is widely recognized that dedicated, commercial scale  lignocellulosic biomass to ethanol facilities will be required to meet the increasing needs of the transportation sector in both Canada and United States. 6, 9  1  1.2  Availability of lignocellulosic biomass A large variety of lignocellulosic biomass has been investigated for bioconversion to  ethanol.  10-12  However, all lignocellulosic biomass can be classified into one of three broad  types: agricultural residues such as corn stover and wheat straw, forest residues such as softwood and hardwood chips, and finally, herbaceous energy crops such as switchgrass. The sheer abundance of lignocellulosic biomass in Canada makes its bioconversion to fuel ethanol of great interest. One recent estimate placed the national available total at between 24 and 87 million dry tonnes per year, towards which forest residues may contribute as much as 80 %. 5 This national total corresponds to as much as 25 billion litres of ethanol per year, the equivalent of 60 % of Canada’s gasoline consumption in 2006. Softwood biomass is readily available in Canada; a second recent assessment concluded that the availability of forest residues in British Columbia (primarily softwood) would be sufficient to support up to 10 bioconversion facilities.  13  Beetle-killed lodgepole pine (Pinus contorta var. latifolia  Douglas) is of particular interest in this province due to its increasing availability, relatively low cost, and limited suitability as structural lumber, the largest existing application of this species. 14 In this thesis, ethanol derived from lignocellulosic biomass will also be referred to as advanced bioethanol.  1.3  Current status of bioethanol According to the Renewable Fuels Association (RFA), the United States ethanol  industry trade organization, 87 billion litres of bioethanol was produced worldwide in 2010 for use as a liquid transportation fuel.  15  At 26 billion litres, Brazil is currently the second  largest producer of fuel ethanol, where it is used in gasoline blends containing at least 18 %  2  ethanol by volume.  15, 16  Europe produced a modest 4.6 billion litres in 2010. Canada  produced 1.4 billion litres, less than the current federal gasoline blending mandate of 5 v/v %, but nonetheless equal to 3 % of the country’s 2006 gasoline consumption. 5, 15 The largest share of ethanol, 50 billion litres, was produced in the United States, where it is typically used in a 10 v/v % blend with standard gasoline referred to as E10. 5, 17 Over approximately the last two years, the price of fuel ethanol has range from a low of USD 0.40/L to a high of USD 0.80/L.  15  Due to a recent mandate by the United States federal government in the  Energy Independence and Security Act (EISA), the amount of domestically produced biofuel is set to rise to 120 billion litres by the year 2022. 2 Of this amount, fully half is to be derived from lignocellulosic biomass and it is widely expected that this fuel will be ethanol.  18  This  represents a very large change to global production, which is currently met with sugar and starch rich feedstocks alone. Corn is the predominant feedstock in the United States, while sugarcane and sugarcane molasses are used in Brazil. In Europe and Canada, sugar beet, wheat, barley, and rye are utilized in addition to corn. 19  1.4  Fuel properties of ethanol Ethanol derived from lignocellulosic biomass is widely recognized as a low carbon  fuel. According to the summary of a recent review published by the International Energy Agency (IEA), advanced bioethanol reduces GHG emissions which includes CO2 by at least 50 % and by as much as 100 % when used as a “neat” fuel to completely replace standard gasoline.  20  In addition, even blending ethanol with gasoline is advantageous in that it is an  effective strategy for reducing tailpipe emissions such as carbon monoxide (CO) and volatile organic compounds (VOCs).  21  Two reasons for this exist. Firstly, ethanol promotes more  3  complete combustion because it is an oxygen containing compound (oxygenate). Secondly, ethanol has a very high pump octane number (PON) of above 110. 17 To formulate a gasoline ethanol blend with the typical North American PON of 87, fewer high octane but polluting hydrocarbons, namely alkenes, dienes, and aromatics, are included in gasoline. 22 Nonetheless, the use of fuel ethanol does present some challenges. For example, the combustion of ethanol produces higher levels of smog causing ozone than standard gasoline. 23 Ethanol, which is hygroscopic, must also be blended with gasoline as close to the point of retail sale as possible to limit the opportunity for it to absorb undesired water.  22  Perhaps most importantly, pure ethanol contains only two thirds the energy of conventional gasoline.  21  When compared to standard gasoline, ethanol is also more corrosive and a less  effective lubricant. This presents a problem for both the infrastructure used to transport ethanol and the components of gasoline motor vehicles. Fortunately, vehicles require only minor modifications to use lower level gasoline ethanol blends. Furthermore, specifically designed flexible fuel vehicles (FFVs) are able to operate on pure ethanol in Brazil, while similarly designed vehicles in the United States and Canada are able to operate with blends containing as much as 85 v/v % ethanol (E85). 17, 24, 25  1.5  Chemical composition of lignocellulosic biomass Despite a huge variation in their appearance, softwoods, hardwoods, agricultural  residues, and herbaceous energy crops are all types of lignocellulosic biomass. Over 90 % of dry lignocellulosic biomass is composed of just three polymers: cellulose, hemicellulose, and lignin. 26 The remainder of the biomass is composed of non-structural, low molecular weight extractives and inorganic ash, which in whitewood are typically present at less than 10 and  4  1 %, respectively.  27  The most abundant metal components of wood ash are calcium (Ca),  potassium (K), and magnesium (Mg). A small amount of protein is also present, especially in non-woody biomass. 28 As the primary polymer of the plant cell wall, cellulose can be thought of as the foundation, or ‘skeleton,’ of lignocellulosic biomass. 26 In softwoods, it typically accounts for between 45 and 50 % of debarked stem wood as outlined in Table 1-1. Cellulose is a linear polysaccharide composed of repeating units of cellobiose, a dimer of the hexose sugar Dglucose, and is linked by a single β-1,4-glycosidic bond. Starch, although also composed of repeating units of D-glucose, is a branched polymer joined by α-1,4- and α-1,6-glycosidic bonds. 29 Largely as a result of this different bonding, starch is much more easily hydrolyzed to glucose than cellulose.  30  The length of an individual cellulose chain varies depending on  cell type and location within the cell wall, but in wood it is typically around 5000 cellobiose units.  26  Individual cellulose chains are joined at the time of biosynthesis in a very ordered  fashion to form elementary fibrils containing both intramolecular and intermolecular hydrogen bonding.  31  The resulting highly ordered crystalline structure means that cellulose  is relatively inert. Elementary fibrils are surrounded by hemicelluloses and lignin and combine in turn to form larger fibrils and eventually lamellae, the layers from which the cell wall is constructed. 31 Hemicelluloses are composed of the hexose sugars D-mannose, D-glucose and Dgalactose, as well as the pentose sugars D-xylose and L-arabinose. hemicellulose typically accounts for 18 – 25 % of wood.  32  26  On a dry weight basis,  Hemicellulose is not a linear  polymer, but rather it is branched, highly substituted, and amorphous. As a result, it is more easily hydrolyzed to its component sugars than cellulose. 26, 33 Most softwood hemicelluloses  5  have a degree of polymerization of roughly 200. By interacting through covalent and noncovalent bonding with cellulose, hemicelluloses contribute to the strength of the cell wall.  32  Galactoglucomannan is the primary hemicellulose of softwoods. The linear backbone is composed of D-glucose and D-mannose units linked by β-1,4-glycosidic bonds, while Dgalactose is attached to the backbone by α-1,6-glycosidic bonds.  26, 32  The ratio of galactose  to glucose to mannose is typically 1:1:3 and because glucose is also the sole component of cellulose, it is difficult to determine the source of softwood derived soluble glucose. The galactoglucomannan backbone is partially substituted at the C-2 or C-3 position with acetyl groups, which are converted to acetic acid when the biomass is steam pretreated.  34  Softwoods contain only smaller amounts of arabinoxylan, and for this reason, softwood hemicellulose is comparatively rich in hexose sugars, which are considered more readily fermentable to ethanol than pentose sugars.  35  The primary hemicellulose of hardwoods is  glucuronoxylan, often referred to as simply xylan. Softwoods generally contain more lignin than other types of lignocellulosic biomass (Table 1-1). As suggested by its role in joining fibrils, lamellae, and finally entire cells together, lignin can be said to be the ‘glue’ of the cell wall. Lignin is composed of three phenylpropane units: guaiacyl (G, or coniferyl alcohol), syringyl (S, or sinapyl alcohol), and p-hydroxyphenyl (H, or p-coumaryl alcohol), which are joined by several common covalent alkyl and ether bonds.  36  It is the combination of these bonds which are responsible for the  amorphous, three dimensional nature of the lignin polymer. Softwood lignin, as represented here by Scots pine (Pinus sylvestris), has a higher G content (98 %) than the lignin of both hardwoods (35 %), represented by hybrid poplar (Populus tremula x Populus alba), and agricultural residues (50 %), represented by wheat straw. 37 It is generally recognized that the  6  quantity and composition of lignin in softwoods makes this type of lignocellulosic biomass an especially challenging substrate for bioconversion to ethanol. 38  Table 1-1. Chemical composition of various lignocellulosic biomass. Glucan Mannan Xylan Galactan Arabinan Lignin a Material Reference g/100g oven dry material (w/w %) 46.5 12.6 9.0 3.9 1.1 27.8 39 Spruce 49.0 3.9 17.9 0.4 0.3 23.3 40 Hybrid poplar 32.6 0.0 20.1 0.8 3.3 26.5 41 Wheat straw a Taken as the sum of AIL (acid insoluble lignin) and ASL (acid soluble lignin).  1.6  The lignocellulosic biomass to ethanol process The production of ethanol from sugar and starch rich feedstocks such as sugarcane  and corn are two relatively straightforward processes. By comparison, the bioconversion of lignocellulosic biomass to ethanol is complex, and a review of the literature confirms the absence of a single process design offering the most cost-effective way to produce advanced bioethanol. 42 Locating production at existing facilities has been proposed as a way in which to reduce ethanol price. Three processes currently of high interest are: corn to ethanol drymilling, kraft pulping, and acid sulphite pulping.  43-45  POET, the largest corn starch ethanol  producer in the United States, has long term plans to co-locate a corn-stover to ethanol plant 43  at each of its 27 existing corn starch to ethanol facilities.  In theory, the existing  infrastructure present at a formerly operational kraft mill can be repurposed to reduce the capital cost of a stand-alone biomass to ethanol facility.  44  On the other hand, minor  quantities of bioethanol are already produced indirectly from hardwoods and softwoods at a small number of sulphite pulp mills in Canada and Scandinavia.  46  Among the producers is  Tembec Inc., a Canadian forest products company which operates a mill in Temiscaming,  7  Quebec with the capacity to produce 12 million litres annually of bioethanol from spent sulphite liquor. 7, 47 Despite a focus on integration with existing facilities, it is widely recognized that production of large volumes of ethanol from lignocellulosic biomass will require the construction of stand-alone (independent) facilities. Arguably the most common configuration proposed for the softwood to ethanol process involves four stages: pretreatment, enzymatic hydrolysis, fermentation, and product recovery.  42  A simplified  process flow diagram (PFD) of the softwood to ethanol process is depicted in Figure 1-1. Steam pretreatment provides partial fractionation of the cellulose, hemicellulose, and lignin and produces a cellulose-rich substrate for subsequent enzymatic hydrolysis. Enzymes, now better able to access the cellulose, catalyze the hydrolysis of this biomass fraction to soluble glucose. Some research has demonstrated that a post-treatment, during which either delignification or lignin modification occurs, is necessary for the effective enzymatic hydrolysis of the steam pretreatment-derived cellulose-rich solid fraction of softwood.  48-50  Cellulase enzymes may be purchased from industrial enzyme manufacturers such as Genencor International Inc. or Novozymes A/S; alternatively, cellulase enzymes may be produced on-site.  51  The soluble sugars generated in steam pretreatment and enzymatic  hydrolysis are then fermented to ethanol. For this later process, a sugar concentration of approximately 100 g/L (10 w/v %) can be considered industrially relevant because it allows for a high ethanol productivity and titer as well as a reduced ethanol distillation cost.  42, 52  Following fermentation, pure anhydrous ethanol containing at most 0.25 v/v % water is recovered from the spent broth using a combination of distillation and adsorption. The lignin  8  residue and other solids present in the distillation stillage have the potential to provide process heat and energy as well as co-products. 53  Figure 1-1. Simplified process flow diagram of the softwood to ethanol process.  1.6.1  Pretreatment Pretreatment is widely recognized as a necessary first step in the bioconversion  process, and many strategies for the pretreatment of lignocellulosic biomass have been proposed to date.  38, 54, 55  Each can be placed into one of three broad categories: biological,  physical, and chemical, although, in reality, many pretreatment strategies are combinations thereof. Irrespective of category, the goals of an effective pretreatment are the same. An ideal pretreatment:  9  • achieves a high recovery of carbohydrates from both cellulose and hemicellulose • produces a cellulose-rich solid fraction readily hydrolyzed by enzymes in a subsequent step • results in at least partial fractionation of cellulose, hemicellulose and lignin • produces highly concentrated solid and liquid streams • generates only low levels of compounds inhibitory to subsequent enzymatic hydrolysis and fermentation such as sugar and lignin degradation products • retains lignin in a near-native, reactive form to facilitate the production of value added products from this fraction • has a low net energy demand • has low capital and operating costs  The use of biomass degrading fungi (white, brown, and soft-rot) has been proposed in the pulp and paper industry as a type of biopulping to precede the standard processes of both mechanical and chemical pulping.  56  By extension, such fungi could be used as an energy  saving pretreatment in the production of ethanol from lignocellulosic biomass. Unfortunately, the slow rate of microbial action means that a commercial facility would be required to store large amounts of biomass at any given time. The storage of such large volumes, typically for a 10 to 14 day residence time, is economically unfavorable.  57  As reported by Guerra et al.  for loblolly pine (Pinus taeda), the use of white-rot fungi, which preferentially metabolize lignin, may even require residence times of 15 to 30 days to maximize lignin removal.  58  Furthermore, it is thought that even extensive treatment with rot fungi would only constitute  10  an initial stage of pretreatment, after which a second pretreatment type would have to be utilized. Increasing the susceptibility of lignocellulosic biomass to enzymatic hydrolysis can also be achieved physically by simply reducing the substrate in size. As was noted in a recent review of pretreatment strategies, hammer- and ball-milling can provide modest improvement in the enzymatic hydrolysis of all lignocellulosic biomass by increasing the surface area of cellulose available to the enzymes.  38  However, a purely physical  pretreatment is uneconomical in that a combination of chipping, cutting, milling, and grinding can require more energy than the energy content of the biomass itself. 54 Many types of chemical pretreatment have been proposed to date, including the two similar methods dilute acid and steam pretreatment.  11, 59  Others, such as pretreatment with  sodium hydroxide (NaOH) or potassium hydroxide (KOH), are well established delignification methods conducted at a highly basic pH.  55, 60  Organosolv pretreatment is  unique in that it offers near complete fractionation of the cellulose, hemicellulose, and lignin fractions. It is conducted in a mixture of water and organic solvent such as ethanol, acetone, or ethylene glycol.  42  An acid catalyst such as dilute sulphuric acid (H2SO4) or gaseous  sulphur dioxide (SO2) is typically employed for softwoods and hardwoods at concentrations of approximately 1 – 2 w/w %. 40, 61 At temperatures of 160 – 200 °C and over cook times of roughly 30 – 70 min, both lignin and hemicellulose are solubilized, leaving a celluloseenriched solid fraction which is readily digested to glucose at low enzyme loadings.  50  Following pretreatment, water may be added to the sugar rich liquid stream to force the hydrophobic lignin to precipitate out of solution.  11  1.6.1.1  Acid catalyzed steam pretreatment The recovery of cellulose and hemicellulose from lignocellulosic biomass as soluble  sugars is one of the single most important factors in determining the price of advanced bioethanol.  42  Acid catalyzed steam pretreatment is a preferred method for softwoods due to  its ability to achieve a high combined recovery of soluble sugars after pretreatment and subsequent enzymatic hydrolysis of the steam pretreatment-derived solid fraction. In fact, steam pretreatment is currently the technology of choice for a number of existing and planned advanced bioethanol facilities. 19, 62 At mild steam pretreatment conditions, typically characterized by temperatures below 200 °C, nearly all softwood hemicellulose can be recovered in an aqueous fraction as a combination of monomeric and oligomeric soluble sugars. 63, 64 The use of higher temperature saturated steam (up to roughly 230 °C) and longer residence times (up to 20 min) are required to solubilize cellulose and to produce a celluloserich solid fraction readily hydrolyzed by enzymes. Unfortunately, because degradation of soluble hemicellulose and cellulose-derived sugars occurs at these more severe conditions, steam pretreatment is essentially a compromise.  38, 64, 65  Neither the solubilization nor the  sulphonation of lignin occur to a great extent during steam pretreatment, but it has nonetheless been suggested that the modification (condensation) and redistribution of this fraction which does occur increases the accessibility of the cellulose in the resulting solid fraction to enzymes.  49, 66, 67  Rapid decompression of the pretreatment vessel terminates the  reaction and reduces the particle size of the solid fraction by suddenly discharging the vessel contents. Although it has been suggested that this final mechanical action does not increase the effectiveness of subsequent enzymatic hydrolysis, it may be beneficial for practical  12  aspects such as the emptying of batch reactors and the pumping of steam pretreatment slurries. 68 Steam pretreatment of hardwoods and agricultural residues is generally easier than softwoods in that less severe reaction conditions are required to achieve a high recovery of soluble sugars and to produce a cellulose-rich solid fraction amenable to enzymatic hydrolysis. During the steam pretreatment of hardwoods and agricultural residues, the acetic acid liberated from the highly substituted xylan acts as the hydrolysis catalyst (autohydrolysis). 69 Softwood hemicellulose is acetylated to a much lesser degree, but the use of an external acid catalyst during steam pretreatment has been shown to increase sugar recovery in the water soluble fraction (WSF) to levels comparable with hardwoods and agricultural residues.  64  In addition, the impregnation of softwood with SO2 or H2SO4 at  concentrations of up to roughly 5 w/w % has been shown to greatly increase the enzymatic digestibility of steam pretreatment-derived solid fractions by increasing the accessibility of the cellulose to hydrolytic enzymes. 70-72 As a result of acid catalysis, softwood is potentially as viable a bioconversion feedstock as other types of lignocellulosic biomass. The most common steam pretreatment catalysts, SO2 and H2SO4, have been shown to provide similar recoveries of soluble sugars and to produce solid fractions equally amenable to enzymatic hydrolysis. 73, 74 Nevertheless, the use of SO2 is advantageous in that it penetrates wood chips rapidly and is present at a uniform concentration throughout the chips.  33, 75  Previous work  also suggests that steam pretreatment-derived liquid fractions rich in hemicellulose sugars may be easier to ferment to ethanol when SO2 is employed as the acid catalyst rather than H2SO4. 73, 74 Phosphoric acid (H3PO4), nitric acid (HNO3) and other organic acids, (and even pressurized carbon dioxide) have been used during steam pretreatment. 59,76-78  13  The compromise of sugar recovery can be overcome by conducting steam pretreatment in two stages, allowing the first stage to be optimized for the recovery of soluble hemicellulose-derived sugars and the second stage to be optimized for the production of a solid fraction amenable to enzymatic hydrolysis. This configuration has been shown to be advantageous for the acid catalyzed steam pretreatment of softwood. Following two stages of SO2 catalyzed steam pretreatment and enzymatic hydrolysis, the combined recovery of soluble glucose and mannose (as both monomers and oligomers) was reported to be 80 %, 10 % higher than the maximum combined recovery after a single stage of steam pretreatment and enzymatic hydrolysis.  39, 65  A similar improvement in combined recovery was  demonstrated with H2SO4 catalyzed steam pretreatment. 79 Despite the demonstrated technical advantage of steam pretreatment conducted in two stages, the sugar recovery realized in steam pretreatment is limited by sugar degradation. In addition to reducing sugar recovery, the products of sugar degradation have been shown to be inhibitory to subsequent enzymatic hydrolysis and fermentation. 80, 81 Inhibition of these two process steps can also be attributed to the products of limited lignin solubilization which occurs during steam pretreatment. Unfortunately, the high temperatures and longer residence times which promote carbohydrate hydrolysis reactions also promote the dehydration of these reaction products to furans and the subsequent conversion of the furans to other lower molecular weight products including organic acids.  82, 83  At even higher pretreatment  severities, pyrolysis reactions predominate, although they do not contribute to losses in yield at temperatures less than 200 °C.  68  Due to the inability of pyrolysis to generate a solid  fraction readily hydrolyzed by enzymes, it is very much an undesirable side reaction. As outlined in Figure 1-2, hexose sugars are converted to hydroxymethylfurfural (HMF) and  14  then to formic and levulinic acid. Similarly, pentose sugars derived from the more thermally labile arabinoxylan are converted to furfural and then to a mixture of compounds including formic acid.  26  In one recent study, levels of HMF and furfural present after the steam  pretreatment of spruce (Picea abies Karsten) at 200 °C, 5 min, and 2.5 w/w % SO2 were found to be approximately 0.6 and 0.4 w/w % of the raw wood, respectively. 84  3 H 2O  2 H 2O Hexose   HMF   Formic acid  Levulinic acid 3 H 2O Pentose   Furfural   Formic acid  Degradation products  Figure 1-2. Reaction pathways of hexose and pentose sugar degradation.  The initial size and moisture content of wood has been shown to influence the performance of chemical pulping processes including steam pretreatment.  68, 84-87  Although  very few studies of steam pretreatment have undertaken systematic and comprehensive investigations into feedstock size and moisture content, it is nonetheless possible that optimal levels of these two variables may exist for both the recovery of soluble sugars after steam pretreatment and for the combined recovery of soluble sugars after steam pretreatment and enzymatic hydrolysis. For example, in one recent study, the combined recovery of soluble glucose and mannose after pretreatment and enzymatic hydrolysis increased slightly, from 71 to a maximum of 73 %, as chip thickness was decreased from 5 – 6 mm to 1 – 2 mm.  84  In  the same study, the recovery of soluble mannose was also found to be highest (81 %) for the 1 – 2 mm thickness fraction. These results resonate with those previously established for kraft pulping, for which a chip thickness of 2 – 4 mm has been shown to be optimal based on its ability to produce the highest screened yield. 87, 88 Other fundamental research using green  15  aspen (Populus tremuloides Michaux) has shown that chip temperature increases slowly during steam pretreatment and that this can result in the ‘undercooking’ of chip interiors and the ‘overcooking’ of chip exteriors at high reaction temperatures. 68  1.6.2  Enzymatic hydrolysis Hydrolysis of the steam pretreatment-derived cellulose-rich solid fraction can be  undertaken using either cellulolytic enzymes or chemically using an acid catalyst (acid hydrolysis). The use of enzymes, although not without its challenges, is advantageous in that the chemical reactions which occur in this second process step are limited strictly to carbohydrate hydrolysis. The presence of oligomeric and monomeric sugars, furans, organic acids, and solubilized lignin fragments in the aqueous hydrolysis medium all have the ability to inhibit enzymatic hydrolysis.  80, 81  Conducting enzymatic hydrolysis at elevated substrate  consistencies has also been shown to negatively affect this process. 89 Thirdly, the lignin and hemicellulose contents of steam pretreatment-derived water insoluble fractions (WIFs) heavily influence the rate and extent of enzymatic hydrolysis.  90, 91  Specifically, lignin  restricts cellulose accessibility and reduces the effective concentration of enzymes by binding them unproductively.  92, 93  It has also been shown, albeit indirectly, that softwood lignin  inhibits enzymatic hydrolysis more strongly than the lignin present in steam pretreatmentderived water insoluble fractions of poplar and corn stover. 49 Although this recalcitrance can be overcome nearly completely by using a high enzyme loading, employing such a strategy at commercial scale is very likely financially prohibitive. 51, 92 Steam pretreated softwood is generally recognized as one of the most challenging lignocellulosic substrates for enzymatic hydrolysis. For example, following steam  16  pretreatment of spruce at 215 °C, 3 min (log Ro 3.86), and 2.4 w/w % SO2, 500 mg protein/g glucan was required to achieve a modest digestibility of 65 % after 144 hrs of 10 w/w % substrate consistency hydrolysis.  89  A similarly modest digestibility (60 %) was achieved  after 72 hrs using lodgepole pine first subjected to steam pretreatment at 200 °C, 5 min (log Ro 3.64), and 4 w/w % SO2.  72  Although the recovery in this latter study was achieved at a  lower enzyme loading (75 mg protein/g glucan), this can be attributed to the lower substrate consistency (2 w/w %) at which hydrolysis was conducted and with it reduced end-product inhibition. By comparison, hydrolysis of corn stover at 2 w/w % consistency and an enzyme loading of 75 mg protein/g glucan resulted in a higher conversion of 71 % despite being preceded by a milder steam pretreatment of 190 °C, 5 min (log Ro 3.35), and 3 w/w % SO2. 72 Several enzymatic activities are required to breakdown cellulose into soluble glucose monomers.  94  Endoglucanase enzymes catalyze the hydrolysis of random, internal β-1,4  glycosidic bonds, effectively doubling the number of chain ends with each reaction. Exoglucanases, a second class of enzymes, bind to exposed chain ends and release both glucose and cellobiose (a glucose dimer). The single glycosidic bond in cellobiose is hydrolyzed by β-glucosidase enzymes, releasing two monomers of glucose. 30, 94 Commercial cellulase enzyme preparations may also include analogous hemicellulase activities and a substantial concentration of non-hydrolytic protein. Supplementation with either of these additional components has the potential to reduce the amount of protein required to achieve fast and complete cellulose hydrolysis. 95-97  17  1.7  Process modeling of pretreatment With the first dedicated commercial scale advanced bioethanol facilities expected to  begin in the next few years, models which can accurately predict the yields and compositions of steam pretreated lignocellulosic biomass fractions will become increasingly industrially relevant. For example, if the models demonstrate that lower temperatures (pressures) and longer residence times lead to a high recovery of soluble sugars after steam pretreatment, this result may influence the choice of metal alloy used to construct continuous steam pretreatment reactors as well as the length (retention time) of the continuous reactors themselves. Several options can be followed when trying to develop a model that can accurately predict and optimize the steam pretreatment of various lignocellulosic substrates. The use of reaction kinetics is one option, although the idealized conditions under which kinetic models are usually developed means that they generally provide better insight into reaction mechanisms than into predicting the performance of an industrially relevant process such as steam pretreatment. Combining mass and heat transfer with chemical kinetics is another, more comprehensive option, and might be highly desirable in the long term. However, very few such models have been developed to date. 98 It appears that past efforts to try to predict and optimize the performance of steam pretreatment have been limited to the use of empirical models, namely thermal severity factors and RSM. Although these empirical models lack a theoretical basis and may therefore be limited in scope to the reactor in which they were developed and limited in their ability to provide accurate prediction at shorter residence times, empirical models have been shown to provide accurate descriptions of process outcomes despite overlooking transport phenomena. For example, Vroom’s H factor, which combines cook temperature and time into a single parameter, has been used in industry  18  to predict the extent of delignification during the digestion stage of Kraft pulping and the resulting pulp yield. 99  1.7.1  Reaction kinetics and mass transfer models Mechanistic models, based either on reaction kinetics, mass and heat transfer  phenomena, or a combination thereof, have been developed to describe the pretreatment of lignocellulosic biomass. As mentioned previously, the predominant carbohydrate reactions of acid catalyzed steam pretreatment are the hydrolysis of hemicellulose and cellulose. Correspondingly, a large number of studies have been published in which the hydrolysis of biomass carbohydrates is described using reaction kinetics. of the subject have also been authored.  106, 107  69, 100-105  Comprehensive reviews  Nearly all studies of hydrolysis employing  mechanistic models have been conducted with non-softwood biomass, due likely to the fact that xylan, the primary hemicellulose of hardwoods and agricultural residues, contains no glucan. Furthermore, autohydrolysis remains the pretreatment of choice for the majority of such studies, with only a few existing models able to account for the action of an external acid catalyst. 108, 109 As noted in one review, very few mass transfer models of pretreatment have been developed to date.  106  This is unfortunate, because the size of commercially available wood  chips at least as present in traditional chemical pulping processes is recognized as being too large to ignore such transport phenomena.  33, 60  Due to the absence of such studies, it is  difficult to determine the predictive capability of mass transfer models and to make comparisons with other mechanistic models based on reaction kinetics. One strictly mass  19  transfer model was developed in a study of corn stover pretreatment conducted in liquid hot water (autohydrolysis).  110  Although the authors reported good agreement between the  experimental data and the model predictions, the goodness of fit was not reported. In fact, the authors themselves concluded that the adapted leaching model did not appear to describe xylan hydrolysis any better than existing models based on reaction kinetics. The first truly kinetic model, based on the Douglas-fir cellulose hydrolysis work of Saemen, was proposed in the 1940s.  111  The reaction pathway is a two-part, first order  reaction in which the kinetic rate constants k1 and k2 are assumed to possess Arrhenius-type temperature dependence. The pathway and reaction rate constant expression are both provided in Figure 1-3. The pre-exponential factor, in units of min-1, is denoted kio, the acid concentration, in units of w/w %, Ci, and the activation energy, in units of kJ/mol, Eia. Finally, mi is a dimensionless, experimentally determined constant. Note that autohydrolysis studies, because no external acid catalyst is used, omit Ci from this expression.  k1 k2 Xylan  Xylose  Degradation products  ki  kioCimi e(  Eia  RT )  Figure 1-3. Basic reaction pathway of xylan hydrolysis with Arrhenius-type temperature dependence.  Efforts to increase the predictive capability of kinetic models have resulted in reaction pathways of increased complexity. Based on the observation that the rate of hemicellulose hydrolysis in Japanese beech (Fagus crenata Blutne) decreased markedly over time, Kobayashi and Sakai proposed the existence of two distinct fractions of hemicellulose.  106  20  The majority fraction (α) is considered more susceptible to hydrolysis.  69, 102  Many  subsequent studies have even regarded the second fraction as completely inert. Accounting for the presence of xylooligomers represents another level of complexity. According to one common proposed pathway, xylan is first hydrolyzed to xylooligomers possessing a high degree of polymerization. Subsequent hydrolysis produces short-chain xylooliogmers, which in turn produce xylose. When this reaction pathway was used in two studies of eucalyptus, the activation energy of xylose degradation was found to range from 96 to 98 kJ/mol, respectively. 100, 103 In a more recent investigation of autohydrolysis conducted by Mittal et al. with sugar maple (Acer saccharum), the simultaneous production of xylose and xylooligomers from xylan was proposed. 69 The unique reaction pathway is provided in Figure 1-4.  Figure 1-4. Alternate reaction pathway of xylan hydrolysis.  This kinetic model was found to predict the levels of xylooligomers, xylose, and furfural in the pretreatment-derived liquid fraction with high accuracy. However, this complex pathway remains linked to its more simplistic predecessors in that all reactions remain first order with respect to reactant concentration. The activation energy for xylose degradation, Ea4, was found to be 101 kJ/mol. This agrees very well with the two studies of eucalyptus just discussed.  100, 103  The activation energies for the hydrolysis of xylan to 21  xylooligomers and xylose, Ea1 and Ea2, were found to be 114 and 117 kJ/mol, respectively. 69 This not only also agrees well with the 124 kJ/mol determined for xylan hydrolysis in additional earlier studies, but also with the 113 kJ/mol first assumed for hemicellulose hydrolysis by Brasch and Free. 101, 102, 112 In a subsequent autohydrolysis study, Mittal et al. combined mass transfer considerations with the unique reaction kinetics pathway just discussed. 98 By describing the transport of the three reaction products, namely xylooligomers, xylose, and furfural, from wood pores to the bulk liquid, the researchers were able to accurately predict the levels of these products in the liquid fraction obtained after the pretreatment of sugar maple wood chips. According to the model, reactions take place both in the interior of the wood chip and in the bulk fluid. Transport of the reaction products from the pore to the chip surface is by diffusion, while convective mass transfer is responsible for transport from the chip surface to the bulk fluid. The applicability of the combined model was successfully validated by repeating one set of experiments with aspen chips. Ultimately, the positive results of this study help demonstrate that mechanistic models have the potential to provide the most accurate and flexible models of acid catalyzed steam pretreatment.  1.7.2  Thermal severity factors Building on the previous observations of Vroom, who looked at the digestion stage of  a kraft pulping process, Overend and Chornet subsequently suggested that steam pretreatment could be modeled in a similar fashion and that the combined influence of cook temperature and time could be used to predict the yield of the solid fraction and the recovery of soluble sugars. 99, 113 To better model this phenomenon, Overend and Chornet adapted the  22  P factor of Brasch and Free, which had been developed in 1965 to describe the progression of the acidic prehydrolysis stage of kraft pulping.  112  Like this prehydrolysis step, steam  pretreatment is an acidic process whose primary carbohydrate reaction is the hydrolysis of hemicellulose glycosidic bonds catalyzed by hydronium cations.  114  The single parameter  developed by Overend and Chornet, known as the thermal severity factor Ro, combines cook temperature and time, and assumes hemicellulose hydrolysis to be a first order reaction with the Arrhenius-type temperature dependence of the kinetic rate constant kr. The activation energy (Ea) of the reaction was calculated to be 113 kJ/mol and is reflected in the value of the constant ω (14.75).  113  The reference temperature Tr is 100 °C, below which hemicellulose  hydrolysis is assumed to be negligible. As depicted here, the reaction time is denoted t. Note that this equation also exactly describes the P-factor of Brasch and Free. 112  t   T  Tr  Ro  exp dt (Equation 1-1)       0  On a qualitative level, the Ro factor has been used extensively to interpret the results of various lignocellulosic biomass pretreatments.  115-117  However, the ability of the Ro factor  to provide quantitative predictions of the direct outcomes of steam pretreatment, specifically yield of the water insoluble fraction, residual hemicellulose content of the WIF, and soluble sugar recovery, has not been well documented. For example, Brasch and Free subjected whitewood chips of radiata pine (Pinus radiata Don) to autohydrolysis in liquid hot water over a wide range of cook temperatures and time.  112  Several characteristics of the  prehydrolyzed chips were plotted as functions of the P factor. Both pulp yield and the  23  pentosan hemicellulose content of the pulp were found to decrease in a nearly logarithmic fashion with linear increases in the P factor. It appears that good correlation occurred between the P factor and the pulp yield, as well as between the P factor and the hemicellulose (pentosans only) content of the pulp. Unfortunately, neither the goodness of fit nor the extent to which hexose sugars were removed from the substrate was quantified. Some researchers seem to suggest that no reliable evaluation of SO2 catalyzed steam pretreatment can be made on the basis of Ro. In a more recent study of the two stage SO2 catalyzed steam pretreatment of spruce, the solid fraction obtained after the first stage of pretreatment was re-impregnated with SO2 prior to the second stage. 39 Apparently, the solid fractions generated in the first stage of pretreatment did not readily adsorb SO2. As a result, the actual level of SO2 used in the second stage varied tremendously from the intended value of 3 w/w % (0.83 – 2.3 w/w %).  73  The thermal severity factor Ro does not account for the  influence of acid catalysis during steam pretreatment, and so the variable level of SO2 may have led the authors to conclude that the outcomes of this second stage of steam pretreatment could not be correlated to Ro. Consequently and subsequently, several studies have evaluated the results of acid catalyzed steam pretreatment within the context of a so-called combined severity factor (CS). 59, 73, 74, 78, 118-120 In 1990, Chum et al. built on the Ro concept and introduced the CS factor which, in addition to temperature and time, uses pH as an estimate of the action of the acid catalyst and its hydronium cation.  121, 122  In some of the studies that looked at SO2 catalyzed steam  pretreatment, the CS factor has been calculated using the pH of the resulting sugar rich liquid fraction.  119  In other studies which looked at H2SO4 catalyzed steam pretreatment, both the  24  pH of the liquid fraction and the dilute acid used to impregnate the biomass prior to pretreatment have been used. 118 The parameter is described here in Equation 1-2.  CS  log Ro  pH  (Equation 1-2)  Kabel et al. found it difficult to use the CS factor to compare the results of dilute acid pretreatment which had been conducted in separate studies due to differences in how the acid catalysts had been quantified and reported.  118  On the other hand, excellent correlation  (coefficient of determination R2 0.96) between the CS factor and the extent of xylan removal from aspen wood chips during methanol organosolv pretreatment conducted using up to 0.075 M phosphoric acid has been reported.  12  This suggested that the CS factor could be  used to create equations that might be better able than the Ro factor to predict the direct outcomes of pretreatment. Using Norway spruce chips, Tengborg et al. undertook a series of H2SO4 catalyzed steam pretreatment experiments in which temperature, time and catalyst concentration were all varied over a wide range (CS 1.4 – 5.4).  74  Chum’s CS factor was used as a tool to  compare the results of the pretreatment, for which the pH term was calculated from the dilute acid used to impregnate the softwood chips. No obvious correlation between the CS factor and the recovery of soluble mannose was evident. In fact, the pretreatments that were conducted at identical CS factor values were often found to produce liquid fractions containing vastly different amounts of both mannose and glucose. For example, the recoveries of soluble mannose and glucose ranged from roughly 25 to 80 % and from 10 to 20 % at a CS factor of 2.7, respectively. However, these observed differences were largely  25  absent when the pretreatments were carried out at residence times longer than 1 min and at temperatures less than 240 °C. This suggested that the poorly defined relationship between hemicellulose-derived sugar recovery and the CS factor may have been due to the very short residence times and high temperatures that were included in this earlier study. This observation seems to be supported by a similar study conducted with Norway spruce sawdust.  79  In the study, the authors subjected spruce to two stages of H2SO4  catalyzed steam pretreatment. The first pretreatment stage was conducted at a single condition, while the second stage was conducted over a range of pretreatment conditions (CS 2.26 – 3.85) which was relatively narrow as compared to the range previously investigated with Norway spruce chips.  74  It was likely that the use of a narrower process space was  responsible for the good correlation between the CS factor and the recovery of soluble glucose from the second stage of pretreatment seen in this study. However, this has been difficult to confirm as no quantitative relationships were developed in either study.  1.7.3  Response surface methodology Response surface methodology is the name given to a class of statistical techniques  most commonly used for the improvement and optimization of existing industrial processes. 123  Two or more independent variables, often referred to as input variables, are controlled at  several different levels in order to study how changes to the independent variables affect one or more output (response) variables. Typically, response variables are measures of process performance or product quality. The central composite experimental design (CCD) has a reputation for generating empirical equations which accurately describe response variables as functions of the independent variables.  124  Due in part to the fact that the equations remain  26  moderately accurate even at the boundaries of the process space investigated, the CCD is one of the most popular second-order RSM experimental designs.  125  On the other hand, hybrid  experimental designs are often advantageous in that models of comparable accuracy can be developed with less experimentation. 126 In the case of the SO2 catalyzed steam pretreatment of softwood, the process related independent variables are temperature, time, and SO2 concentration, while the most important feedstock related variables are initial moisture content and size. In recent years, RSM has been applied in several studies of dilute acid and organosolv pretreatment.  40, 119, 120, 127, 128  In many of these studies, a CCD was used to determine the  optimal conditions of pretreatment corresponding to the maximum production of soluble sugars. It appears that only three RSM studies of the SO2 catalyzed steam pretreatment of softwoods have been published previously.  63-65  Using a hybrid experimental design, the  objective of one of these earlier studies was to identify a set of steam pretreatment conditions corresponding to the highest recovery of hemicellulose-derived sugars in the resulting liquid fraction.  63  Pretreatment was conducted over a limited range of relatively low severity  conditions, namely 165 – 205 °C, 1.5 – 7.5 min (log Ro 2.38 – 3.74), and 0.5 – 4.5 w/w % SO2. More than 90 % of the hemicellulose was recovered as soluble sugars after steam pretreatment at 195 °C, 2.4 min (log Ro 3.17) and 3.9 w/w % SO2. However, only 7 % of the glucan in the raw wood was recovered as soluble glucose at this condition, which demonstrates that two separate optima exist for the recovery of hemicellulose-derived soluble sugars and for the recovery of all soluble sugars including glucose. In another of these studies, a CCD was used in conjunction with a wider range of SO2 catalyzed steam pretreatment conditions to identify a set of operating conditions corresponding to the highest  27  combined recovery of soluble sugars following both the pretreatment and enzymatic hydrolysis of radiata pine.  64  Only 75 % of the hemicellulose-derived sugars were recovered  in the liquid fraction after pretreatment at 215 °C, 3 min (log Ro 3.86) and 2.55 w/w % SO2. However, this condition corresponded to the maximum combined sugar recovery (89 %), which suggests that the production of a solid fraction readily hydrolyzed by enzymes is only possible at severities higher than the optimum for hemicellulose-derived sugar recovery. This conclusion is supported by a third, similar RSM study of mixed softwoods, which also investigated a broad process space but in conjunction with a non-rotatable, face-centred central composite design (CCFD). 65  1.8  Techno-economic models of the softwood to ethanol process Techno-economic modeling, or process simulation, has proven to be a useful tool for  evaluating proposed commercial scale advanced bioethanol processes. Aspen Plus is the simulation software of choice for nearly all such recently developed techno-economic models. Simulations provide a breakdown of ethanol product cost into individual unit operations and raw materials, allowing areas of potential cost savings to be identified. Sensitivity analyses, in which ethanol product cost is monitored as changes to individual variables such as enzyme cost are made, are often carried out to identify process weaknesses and areas of potential improvement. The accuracy of product cost estimates depends greatly on the type and extent of information included in the simulation itself. In the case of simulations of the softwood to ethanol process, batch-mode experimental data collected at the bench-top and pilot-plant scale have been used.  129-131  While these techno-economic  models consider the capital cost of major equipment such as reactors as well as operating  28  costs such as chemicals, utilities, and labour, they may not include detailed costs for minor equipment such as pumps, heat exchangers, piping, and valves. Arguably the most detailed simulation of an advanced bioethanol process was developed by the United States National Renewable Energy Laboratory (NREL).  132  The  proposed facility was designed to convert roughly 700 000 tonnes of corn stover, at a cost of $64/dry tonne, to 230 million litres of ethanol annually using a dilute sulphuric acid pretreatment. This latest work, which is based on experimentally proven data, is in fact the newest version of two earlier simulations.  133, 134  Several improvements, which included a  detailed quote for a continuous pretreatment reactor quote provided by Andritz Inc. and an on-site enzyme production section intended to increase the transparency of enzyme cost, each contributed to the realistic process economics of this simulation. Enzyme and ethanol production cost were estimated at $4/kg protein and $0.57/L, respectively. The recovery of soluble xylose after pretreatment, conducted at 158 °C for 5 min using 1.8 w/w % H2SO4, was 92 %, while the combined recovery of soluble xylose and glucose after both pretreatment and enzymatic hydrolysis was a similarly high 92 %. Several simulations of the softwood to ethanol process have been developed in recent years which share a common design basis including SO2 catalyzed steam pretreatment. 53, 131, 135-137  The annual capacities of the proposed facilities are 200 000 tonnes of chipped Norway  spruce and 60 million litres of ethanol, respectively. The price of wood varied between studies at $55 – 80/dry tonne, while the purchased enzyme cost was comparatively low at roughly $2.6/kg protein.  51  The recovery of soluble mannose assumed in these studies is  typically 85 %, which translates into an ethanol product cost of $0.50 – 0.60/L. 53, 131, 136 The  29  accuracy of this ethanol product cost estimate seems reasonable given the similar conclusions reached in the latest simulation developed by NREL. Despite the fact that steam pretreatment has a very large impact on the remainder of the softwood to ethanol process, relatively little attention has been paid to this unit operation in recent simulations. Instead, the focus has been on anaerobic digestion of distillation stillage, on-site cellulase enzyme production, and even fermentation. 51, 129, 135, 138 Among the few simulations investigating pretreatment, none have been able to conclude on the basis of ethanol product cost that one pretreatment method was better than another.  131, 139, 140  Furthermore, in a simulation comparing single and two stage SO2 catalyzed steam pretreatment, both scenarios were found to result in the same ethanol product cost ($0.56/L). 131  Although soluble mannose recovery and ethanol yield were 16 and 3 % higher for the  two-stage pretreatment configuration, respectively, these gains were offset by the higher capital cost associated with the two-stage process. Even in the most comprehensive techno-economic simulations of advanced bioethanol processes, ethanol product cost remains independent of pretreatment operating conditions and feedstock characteristics such that changes to these variables are not translated into changes to the whole process. Within existing simulations of the softwood to ethanol process conducted using the software Aspen Plus, SO2 catalyzed steam pretreatment has historically been simulated using static conversion factors, essentially a set of mass balance equations describing both the reactions of the SO2 catalyzed steam pretreatment of softwood and the corresponding extent of these reactions at a single combination of reaction and feedstock conditions.  30  1.9  Research objectives and overview Prediction of sugar recoveries from steam pretreated and enzymatically hydrolyzed  lignocellulosic biomass is desirable for the purposes of effective steam pretreatment reactor design and operation. In this thesis, empirical process models able to predict the performance of the acid catalyzed steam pretreatment and subsequent enzymatic hydrolysis of softwood were developed using thermal severity factors and several different RSM experimental designs.  1.9.1  The use of predictive models to optimize sugar recovery in the steam  pretreatment step of a softwood-to-ethanol process Mechanistic models that combine mass and heat transfer with chemical reaction kinetics may well be best suited to predict the performance of acid catalyzed steam pretreatment. However, it appears that past efforts to try to predict and optimize the performance of steam pretreatment have been limited to the use of empirical models, namely thermal severity factors and RSM. In this first study, these two types of empirical models were assessed for their relative ability to predict several direct outcomes of the acid catalyzed steam pretreatment of radiata pine, specifically WIF yield, WIF residual hemicellulose content, and WSF soluble sugar recovery. In addition, efforts were made to validate one RSM model of the SO2 catalyzed steam pretreatment of radiata pine using a second softwood species, namely lodgepole pine.  31  1.9.2  Effects of chip size and moisture content on the combined sugar recovery in the  steam pretreatment and enzymatic hydrolysis steps of a softwood-to-ethanol process In the first portion of this thesis, it was determined that RSM could be used to create more comprehensive empirical models of acid catalyzed steam pretreatment than thermal severity factors. Nonetheless, model validation efforts showed that truly comprehensive RSM models would require additional independent variables. In this second study an additional RSM model considering both steam pretreatment operating conditions and feedstock characteristics was developed. Much attention has already been paid to the effects of reaction temperature, time, and SO2 loading on the acid catalyzed steam pretreatment of softwood. By using RSM, the influence of chip size and moisture content on the combined recovery of soluble sugars from steam pretreatment and subsequent enzymatic hydrolysis could be studied in a systematic and comprehensive manner; if possible, a combination of steam pretreatment operating conditions and feedstock characteristics corresponding to a maximum combined sugar recovery would also be identified. In addition, the RSM model of lodgepole pine was developed in a manner which made it appropriate for use in an existing process simulation of the entire softwood to ethanol process. Use of the empirical model in such an Aspen Plus simulation would make it possible for the first time to evaluate the effects of changing the operating conditions of steam pretreatment and feedstock characteristics on total ethanol product cost.  32  Chapter 2: Materials and methods 2.1  Raw wood Both radiata pine and lodgepole pine were used in this thesis as feedstocks for acid  catalyzed steam pretreatment and subsequent enzymatic hydrolysis. Thomas Clark of Carter Holt Harvey kindly provided chips of radiata pine taken from a single, freshly-felled, low quality saw-log at the company’s New Zealand Kinleith pulp mill located in Tokoroa. The mechanically debarked log possessed 18 growth rings and was therefore very comparable in age to that of the radiata pine employed in a previous study.  64  The small and large end  diameters of the log measured roughly 35 and 50 cm, respectively. Healthy, freshly-felled lodgepole pine from British Columbia’s Thompson-Okanagan interior region was kindly provided by Ken Ewanick of Tolko Industries (Vernon, BC.) and Kerry Rouck of Gorman Brothers Lumber Ltd. (Kelowna, BC.). The two bolts from Tolko measured roughly 19 cm in width and were approximately 48 years old, while the three bolts from Gorman Brothers measured roughly 14 cm in width and were approximately 75 years old.  2.2 2.2.1  Preparation of wood chips Chipping and classification Radiata pine arrived at UBC in the form of whitewood chips following mechanical  debarking, chipping, and classification. Classification undertaken at Carter Holt Harvey removed the dust retained on a pan, the pin chips retained on a round hole (RH) 3 mm screen, the over-thick chips retained on a bar (TH) 10 mm screen, and finally the over-sized chips retained on a RH 45 mm screen. Further classification was undertaken on a Williams  33  classifier with the help of Bernard Yuen at FPInnovations, located on the UBC Vancouver campus. In this second round, the fraction retained on a RH 5/8 in (16 mm) screen was saved for steam pretreatment. Note that RH screens allow for chip separation based on their length as measured in the longitudinal direction while TH screens allow for separation based on their thickness as measured in the radial direction.  88  The moisture content of the fully  classified chips was then determined (56 w/w % on a wet wood basis) by drying to constant weight at 105 °C before storing them in plastic bags in aliquots equivalent to 100 oven dry grams (odg). The bags were immediately frozen at -20 °C. Two separate sets of lodgepole pine chips were prepared in a manner similar to that of the radiata pine, one from the two bolts provided by Tolko and a second from the three bolts provided by Gorman Brothers. The two separate sets are referred to as LPP1 and LPP2, respectively. It should be noted that only the set LPP2 was used in the portion of the thesis discussed in Chapter 4. All five bolts were first debarked and split by hand. Chipping and classification were undertaken at FPInnovations. Each of the dust, over-thick and over-sized fractions were removed as described above. The remaining chips retained on RH 1/8 in (3 mm), RH 3/8 in (10 mm), RH 5/8 in (16 mm), RH 7/8 in (22 mm) and RH 9/8 in (29 mm) screens were saved for steam pretreatment. As demonstrated in Figure 2-1 for set LPP2, chip thickness was primarily in the 2 – 6 mm range for the size fractions up to RH 7/8 in and 8 – 10 mm for the RH 9/8 in fraction. In other words, chip thickness was found to increase slightly as chip size increased. Some of the classification was also undertaken at BCIT in Burnaby with the help of Rodger Beatson. The moisture contents of the fully classified chip samples were then determined (LPP1 49 w/w % and LPP2 34 w/w % on a wet weight basis)  34  by drying to constant weight at 105 °C before storing them in plastic bags in aliquots equivalent to either 100 or 200 odg. The samples were immediately frozen at -20 °C.  Figure 2-1. Size distribution of the lodgepole pine wood chips prepared for steam pretreatment (LPP2). Length is given as a round hole (RH) screen size while thickness is presented in the bars as a bar screen (TH) size.  2.2.2  Adjustment of initial moisture content Many of the steam pretreatment experiments undertaken in this thesis required  adjusting the initial moisture content of the chipped lodgepole pine prior to impregnation with an acid catalyst. The initial moisture content of the chipped radiata pine was not adjusted. For the portion of the thesis regarding validation of the RSM model of radiata pine (Section 3.2.5), the moisture content of individually bagged samples of lodgepole pine was  35  increased from 34 to 56 w/w % to ensure that any differences observed following steam pretreatment could not be attributed to a difference in moisture content. For the portion of the thesis regarding development of the RSM model of lodgepole pine (Chapter 4), samples were increased in moisture content to 35, 47.5, and 60.0 w/w %. This was done by adding only the required amount of water to the bags using a spray bottle, thereby ensuring its homogenous uptake. The samples were allowed to remain at 4 °C until all the excess moisture had been incorporated into the raw wood. The adjusted samples were then frozen again at -20 °C. The moisture content of chipped lodgepole pine was also decreased from 34 to 22.5 and 10.0 w/w %, again for the portion of thesis regarding development of the RSM model of lodgepole pine (Chapter 4). These samples were air dried at room temperature, during which time the weight of the samples was monitored until the desired value had been reached. The adjusted samples were then frozen again at -20 °C.  2.3  Composition of raw wood The chemical composition of the raw wood used in this thesis is presented here in  Table 2-1. The values determined for both radiata and lodgepole pine were found to be similar to values previously determined and reported in the literature.  11, 64  The analytical  methods used for determination of the primarily carbohydrate and lignin components are outlined in Section 2.7.  36  Table 2-1. Chemical composition of the raw pine prior to steam pretreatment. Sample a RP LPP1 b LPP2 c g/100g oven dry wood (w/w %) 1.2 (0.0) 1.5 (0.1) 1.8 (0.1) Arabinan 2.2 (0.1) 2.3 (0.1) 2.9 (0.2) Galactan 44.7 (0.9) 46.3 (1.0) 45.9 (0.3) Glucan 5.1 (0.4) 6.4 (0.2) 6.3 (0.3) Xylan 10.2 (0.6) 11.0 (1.0) 11.8 (0.9) Mannan d 26.5 (0.2) 26.3 (0.3) 26.4 (0.6) AIL e 0.2 (0.1) 0.5 (0.0) 0.5 (0.0) ASL 1.9 (0.1) 3.7 (0.3) 4.2 (0.2) Extractives 0.3 (0.0) 0.2 (0.0) 0.2 (0.0) Ash d Nd 1.0 (0.0) 1.0 (0.0) Acetyl groups 92.2 (1.4) 99.3 (1.6) 100.6 (0.5) Total a b c Radiata pine. Lodgepole pine from Vernon. Lodgepole pine from Kelowna. d Not determined. d Acid insoluble lignin. e Acid soluble lignin.  2.4  Acid catalyzed steam pretreatment Thawed samples of radiata and lodgepole pine were impregnated with an acid catalyst  prior to steam pretreatment. For the portion of the thesis regarding the comparison of the thermal severity factors Ro and CS (Section 3.2.2), the acid catalyst used was a diluted aqueous solution of concentrated H2SO4 (Fisher Scientific, Gormley, Ontario). A predetermined amount of this liquor (50 mL of 1.3 – 3.8 w/w % H2SO4) was added to the individual samples such that the moisture content of the chips after impregnation was a single value for all samples (65 w/w %). The dilute acid was applied with a spray bottle to ensure its homogeneous uptake and the samples were kept at 4 °C until all the excess moisture had been incorporated into the raw wood (overnight). Pretreatment was carried out the following day. The remaining thawed samples of chipped radiata and lodgepole pine were impregnated with gaseous SO2 (Praxair Canada Inc., Mississauga, Ontario) prior to steam 37  pretreatment. The gaseous SO2 was added to sealed plastic bags containing either 100 or 200 odg of raw wood and the amount added was controlled by monitoring the total weight of the bags. No calculations of SO2 retention were made in this thesis. For the portion of the thesis regarding the validation of the RSM model of radiata pine (Section 3.2.5), the samples were steam pretreated within 30 min of impregnation. For the portion of the thesis regarding the development of the RSM model of lodgepole pine (Chapter 4), the impregnated samples were left overnight in a fume hood at room temperature and steam pretreatment was carried out the following day. The latter strategy was used so that it was possible to conduct many steam pretreatment experiments in a single day. Steam pretreatment was undertaken in batches of either 50 or 200 odg of impregnated softwood in a 2 L StakeTech II steam gun (Stake Technologies, Norval, Ontario). For the two portions of the thesis regarding the comparison of the thermal severity factors Ro and CS and the validation of the RSM model of radiata pine (Chapter 3), a total of 100 odg was treated at each condition in two back-to-back batches of 50 odg each. Steam pretreatment conducted in this manner represents the existing standard protocol of the research group. Although steam pretreatment was essentially undertaken twice at each condition, the experiments were not considered to be statistically independent. The temperatures and times used are provided in Section 2.8 together with the remaining details of the experimental designs. For the portion of the thesis regarding the development of the RSM model of lodgepole pine (Chapter 4), steam pretreatment was undertaken at each condition using single batches of 200 odg of the Gorman Brothers supply (LPP2). This revised protocol was used for several reasons. Firstly, preliminary work conducted with Douglas-fir (data not shown) suggested that the amount of rejects, defined as solid material which appeared visibly  38  unchanged following steam pretreatment, could be decreased in both absolute and relative terms by using larger batches. Secondly, the use of 200 odg batches in the 2 L steam pretreatment reactor at UBC represents a wood to reactor capacity ratio similar to those employed by other research groups. Finally, it was thought that water soluble fractions containing industrially relevant concentrations of fermentable hexose sugars (approximately 100 g/L) could be generated using 200 odg batches.  2.5  Solid liquid separation At the end of each steam pretreatment reaction, the elevated pressure in the reactor  caused the vessel contents to be suddenly discharged to a collection vessel. This slurry was immediately removed and tap water was used to ensure its complete collection. Vacuum filtration was used to separate the slurry into solid and liquid fractions, referred to as the water insoluble and water soluble fractions, respectively. Following initial separation, the solid fraction was washed extensively with water and the washings were added to the liquid. In all cases, between 5 and 6 L of wash water were used to ensure all soluble components had been separated from the solid material. WSF samples (which included wash water) were then frozen at -20 °C and later thawed for analysis. Finally, the pH of a select number of these diluted WSF samples was measured at room temperature and the values were used to calculate the combined thermal severity factor CS. Moderate amounts of reject chips (roughly 12 w/w % of the raw wood) were present in the pretreatment slurries for those pretreatments conducted with 50 odg of raw softwood per batch. Rejects were removed by hand immediately following the washing step of vacuum  39  filtration to ensure that the rejects did not interfere with subsequent analysis of the WIF or the calculation of sugar recoveries. To allow for the collection of a highly concentrated WSF, a slightly modified filtration protocol was used for the portion of the thesis regarding the development of the RSM model of lodgepole pine (Chapter 4). In the case of the six replicates of the centre point condition, outlined in the last experimental design of Section 2.8, as much pretreatment slurry as possible was first removed from the reactor collection vessel without the addition of water and filtered directly. In this manner, roughly 150 mL of highly concentrated liquid was collected per 200 odg batch of raw wood.  2.6  Enzymatic hydrolysis For the portion of the thesis discussed in Chapter 4, the WIF obtained after steam  pretreatment was enzymatically hydrolyzed in sodium acetate buffer (50 mM, pH 4.8). Washed, never-dried WIF was diluted with nanopure water to 10 w/w % substrate consistency in 125 mL Erlenmeyer screw-top flasks using a total weight of 25 g (2.5 odg WIF). It was not possible to subject unwashed WIF to enzymatic hydrolysis due to the need to first determine WIF yield (which required extensive washing). The enzymatic hydrolysis was otherwise conducted at 50 °C in an orbital shaker operating at 150 rpm using the commercial enzyme preparation Cellic® CTec2 (Novozymes North America Inc., Franklinton, North Carolina) at a loading of 70 mg total protein/g glucan. Cellic® CTec2 is produced by a genetically modified strain of the fungus Trichoderma reesei. After testing several protein loadings (35 – 70 mg total protein/g glucan), this level was selected because it provided near full cellulose conversion only for the WIFs derived from the most severe steam  40  pretreatment conditions. The total protein content of the complete enzyme preparation, which contained cellulolytic, hemicellulolytic, and non-hydrolytic enzymes in addition to glucosidase, was determined with the Ninhydrin assay to be 210 mg protein/mL.  141  No  antibiotics were employed in the enzymatic hydrolysis reactions. After 72 hrs, a representative 1 mL sample of the hydrolyzate was taken and the reaction was halted immediately. This was done through inactivation of the enzymes by boiling the sample on a hot-plate at 100 °C for 10 min. Next, the boiled sample was centrifuged for 10 min at 13 000  x g and 4 °C before freezing the supernatant at – 20 °C. The supernatant was later analyzed for all monomeric sugars by HPLC as described in Section 2.7. Initial testing with posthydrolysis confirmed that no sugars were present in oligomeric form after 72 hrs of enzymatic hydrolysis (data not shown). All enzymatic hydrolysis reactions were conducted in duplicate.  2.7  Analytical methods  2.7.1  Water insoluble fraction obtained after steam pretreatment The moisture content of the WIF was determined in the same manner as for the raw  wood, namely by oven drying to constant weight at 105 °C. This allowed for the calculation of WIF yield.  2.7.1.1  Carbohydrates and lignin The carbohydrate (cellulose and hemicellulose) and acid insoluble lignin (AIL)  contents of the raw wood and pretreatment-derived WIFs were quantified according to the slightly modified TAPPI standard method T-222-om-06. 142 Briefly, samples were oven dried  41  and reduced in size in a Willey Mill to pass through a 40 mesh screen. Hydrolysis of the insoluble carbohydrates to soluble sugars was done by adding 3 mL of 72 % H2SO4 to 0.2 odg of the ground sample and mixing at 10 min intervals for 2 hrs. The reaction contents were then diluted with water to 4 % H2SO4 and placed in the autoclave for 1 hr at 121 °C (post-hydrolysis) to ensure that all the soluble sugars were present in monomeric form. The room temperature samples were then filtered through sintered glass crucibles to separate the AIL from the soluble sugar rich liquid (Klason filtrate). The retentate was washed and oven dried before quantifying AIL gravimetrically while the filtrate was analyzed for monomeric sugars by high performance liquid chromatography (HPLC) as described in Section 2.7. For the purpose of completing the mass balance of the WIF, the filtrate was also analyzed for acid soluble lignin (ASL) by measuring UV absorbance at 205 nm.  143  All analysis was  repeated in triplicate.  2.7.1.2  Extractives and ash Prior to analysis for carbohydrates and lignin, the methanol extractive content of the  raw wood was estimated using a slightly modified version of the TAPPI standard method T204-cm-07 (exhaustive extraction).  144  The corresponding standard method published by  NREL suggests that the water extractive content of softwoods is very low, and so this fraction was never quantified. Briefly, between 2.5 and 5 g of oven dried wood meal was placed in individual cotton cellulose thimbles. The thimbles themselves were placed in glass Soxhlet tubes and connected to the remainder of the Soxhlet apparatus. Methanol (100 mL) was boiled in a round bottom flask using a heating mantle and then condensed with cold water, passing through the wood meal with each cycle. The minimum duration of extraction  42  and reflux rate were 8 hours and 6 cycles per hour, respectively. Following extraction, the cellulose thimble containing the wood meal was oven dried before determining the extractives content gravimetrically. The extractive containing solvent was not further analyzed. The ash content of both raw wood and steam pretreatment-derived WIFs was quantified gravimetrically according to a slightly modified method of a TAPPI standard method. 145 Briefly, combustion of the oven dried wood meal was conducted in a Model 650 Muffle Furnace at 550 °C for between 4 and 6 hrs depending on the amount of ground wood meal used. Both ash and extractives analysis were repeated in triplicate.  2.7.1.3  Acetyl groups As described previously, the softwood hemicellulose galactoglucomannan is partially  substituted with acetyl groups. Quantification of acetyl groups in the raw wood was determined by analyzing the Klason filtrate for acetic acid by HPLC as described in Section 2.7.  2.7.2 2.7.2.1  Water soluble fraction obtained after steam pretreatment Soluble sugars Soluble monomeric sugars were quantified by HPLC using a Dionex ICS-3000  system (Dionex Corporation, Sunnyvale, California) equipped with an anion exchange column (Dionex CarboPac PA1), AS50 auto sampler, ED50 electrochemical detector, and GP50 gradient pump. Nanopure water at 1.0 mL/min was used as the eluent and 0.2 M NaOH was added post-column. The column was reconditioned between each sample  43  (injection volume 20 μL) with 1.0 M NaOH and all samples were filtered through either a 0.45 or 0.20 μm filter (Chromatographic Specialties, Brockville, Ontario) prior to analysis. Fucose was used as the internal standard. Sugars present in the steam pretreatment samples were measured against analytical grade standards of glucose, mannose, xylose, galactose and arabinose (Sigma Aldrich Canada Ltd., Oakville, Ontario). A second set of WSF samples were post-hydrolyzed with 4 % H2SO4 and autoclaved for 1 h at 121 °C to convert the soluble oligomeric sugars to soluble monomeric sugars. In this way, oligomeric sugars could be quantified by calculating the difference between all sugars and just the monomeric sugars present in each WSF sample. All analysis was repeated in triplicate.  2.7.2.2  Furans and organic acids Levels of the steam pretreatment-derived inhibitory compounds formic and levulinic  acid, HMF and furfural were quantified by HPLC using a modified NREL standard method applied to samples of the WSF.  146  Acetic acid released from the raw wood upon  hemicellulose hydrolysis was also quantified in this manner. An Aminex HPX-87H column was used for separation (Bio-Rad Laboratories, Mississauga, Ontario) in a Dionex DX-500 HPLC system (Dionex Corporation, Sunnyvale, California) equipped with an autosampler, an ED40 electrochemical detector and an AD20 absorbance detector. The mobile phase was 5 mM H2SO4 at a flow rate of 0.6 mL/min. Standards were prepared from glacial acetic acid (Fisher Scientific, Gormley, Ontario) and analytical grades of formic acid (J.T. Baker Reagent Chemicals, Phillipsburg, New Jersey), levulinic acid, HMF and furfural (all Sigma Aldrich Canada Ltd., Oakville, Ontario). The compounds were detected by UV absorbance at  44  210 nm and all samples were filtered through either a 0.45 or 0.20 μm filter prior to analysis. All analysis was repeated in triplicate. To achieve the best separation of the compounds of interest from each other and from additional, unidentified compounds present in the WSF, the column operating temperature was first optimized (40 – 60 °C) and the influence of post-hydrolysis was investigated (data not shown). Quantification of formic acid was done at 60 °C using post-hydrolyzed WSF samples due to the very similar retention times of formic acid and oligomeric sugars.  147, 148  Quantification of acetic and levulinic acid was undertaken using post-hydrolyzed samples at 60 °C. Both HMF and furfural were quantified without post-hydrolysis at 50 °C.  2.7.3  Enzymatic hydrolyzate As mentioned previously in Section 2.6, all soluble sugars present in the free liquid of  the enzymatic hydrolysis reactions (hydrolyzate) were quantified by HPLC as described in Section 2.7. The unhydrolyzed steam pretreatment-derived solids were not further analyzed. The Cellic® CTec2 enzyme preparation contained a substantial amount of monomeric sugars (data not shown) which was subtracted from the quantity present in the hydrolyzate. In this way, overestimation of cellulose conversion was avoided. Overestimation of the conversion can occur in a second manner. During enzymatic hydrolysis conducted at moderate to high substrate consistency, the conversion of insoluble carbohydrate polymers to soluble sugars changes the density, volume, and mass of the hydrolyzate. The assumption that these values remain constant over the course of enzymatic hydrolysis leads to overestimation of cellulose conversion. For this reason, a correction factor reported in the literature (0.92) was applied. 149  45  2.8  Experimental design and outline The raw pine employed in this thesis was subjected to both steam pretreatment and  subsequent enzymatic hydrolysis. A visual overview of the experimental procedures described in this section is provided in Figure 2-2. It should be noted that some steps were unique to the development of the RSM model of lodgepole pine.  Figure 2-2. Experimental procedure employed in all research portions of this thesis. Solid lines indicate experimental procedures while dashed lines indicate analyzed components. a Water insoluble fraction. b Water soluble fraction.  2.8.1  Comparison of the thermal severity factors Ro and CS As discussed in Section 3.0, the two thermal severity factors Ro and CS were  compared for their ability to predict several direct outcomes of the H2SO4 catalyzed steam pretreatment of radiata pine. Table 2-2 lists the experiments undertaken for this comparison. 46  Two different pH measurements are listed for each experiment. The first value is the measured pH of the dilute H2SO4 used to impregnate the raw wood and the second value is the pH of the diluted WSF obtained after steam pretreatment and subsequent vacuum filtration. These values were subsequently used to calculate two distinct CS factors.  Table 2-2. Experimental design for H2SO4 catalyzed steam pretreatment. Catalyst loadings are provided in units of g H2SO4/100 g oven dry wood. pH b CS b pH c CS c Temperature Time Log Ro H2SO4 a Run H2SO4 H2SO4 WSF WSF %) (°C) (min) (---) (w/w %) (---) (---) (---) (---) 182 2.41 2.80 0.65 1.01 1.79 2.39 0.41 CS 1 190 1.74 2.89 0.98 0.79 2.10 2.14 0.75 CS 2 d 195 3.00 3.27 1.30 0.67 2.60 2.12 1.15 CS 3 205 5.92 3.86 1.62 0.58 3.28 2.11 1.75 CS 6 200 6.41 3.75 1.96 0.51 3.24 2.12 1.63 CS 7 a Concentration of the dilute H2SO4 liquor used for impregnation of the raw wood. b Dilute acid impregnation liquor. c Water soluble fraction. d CS 3 – CS 5 are triplicates; given values are mean values. 2.8.2  Validation of the response surface methodology model of radiata pine As also discussed in Section 3.0, efforts were made to assess the influence of steam  pretreatment reactor configuration on the direct outcomes of this process and to assess the robustness of the RSM model of radiata pine. Table 2-3 lists the experiments undertaken for this RSM model validation effort.  47  Table 2-3. Experimental design for validation of the response surface methodology model of radiata pine. Catalyst loadings are provided in units of g SO2/100 g oven dry wood. Temperature Time Log Ro SO2 Chip size Run (°C) (min) (---) (w/w %) (mm) a 215 3.00 3.86 2.55 16 RP a 215 3.00 3.86 2.55 16 LPP1 a 215 3.00 3.86 2.55 16 LPP2 a Pretreatment repeated in triplicate.  2.8.3  Effect of feedstock characteristics on the combined sugar recovery after steam  pretreatment and enzymatic hydrolysis The last portion of the thesis, discussed in Chapter 4, was undertaken to assess the influence of feedstock size and moisture content on the combined sugar recovery obtained from lodgepole pine after SO2 catalyzed steam pretreatment and subsequent enzymatic hydrolysis. As mentioned previously, the RSM model of lodgepole pine was also developed in a way that would permit its use in an existing Aspen Plus simulation of the entire softwood to ethanol process. For example, a number of additional response variables such as WSF inhibitor levels were included in the RSM model. Enzymatic hydrolysis of the WIF was also conducted at a moderate yet industrially relevant substrate consistency (10 w/w %). At this consistency, end-product inhibition was expected to influence the extent of enzymatic hydrolysis in addition to cellulose accessibility and non-productive binding of the enzymes to lignin. Table 2-4 lists the conditions of the 32 steam pretreatment experiments undertaken according to a central composite experimental design (CCD). Careful consideration was given to the ranges of the five independent variables investigated, namely temperature, time, SO2 loading, chip size, and moisture content. The ranges of temperature (185 – 225 °C) and time (0.5 – 9.5 min) were selected primarily such  48  that the highest combined sugar recovery was anticipated at the centre point condition (log Ro 3.79). The range of SO2 was limited from 0.5 to 4.5 w/w % due to previous studies which demonstrated that concentrations of SO2 above approximately 5 w/w % failed to improve the recovery of soluble sugars from either the steam pretreatment or subsequent enzymatic hydrolysis of softwood. 64, 150 Chip size was varied from 1/8 in (3 mm) to 9/8 in (29 mm) and as such represented nearly the full size range of a standard mill chip. Finally, the moisture content was varied from 10 to 60 w/w % on a wet weight basis so that the upper limit corresponded to the highest naturally occurring moisture content of the samples collected. Each of these independent variables ranged from - 2 to + 2 when converted to coded form using the equations provided here.  Temperature ( C )  205 C Temperature (Coded)  (Equation 2-1) 10 C Time (Coded)   SO 2 (Coded)   Time ( min )  5 min (Equation 2-2) 2.25 min  SO 2 ( w/w % )  2.5 w/w % (Equation 2-3) 1 w/w %  Chip size (Coded)   Moisturecontent (Coded)   Chip size ( mm )  15.88 mm 6.35 mm  (Equation 2-4)  Moisturecontent ( w/w % )  35 w/w % (Equation 2-5) 12.5 w/w %  This last portion of the thesis was undertaken according to a CCD which included a fractional factorial design. Neither a hybrid nor a Box-Behnken design was selected as each is considered uneconomical when five independent variables are investigated. 126, 151 The half 49  fraction of the 25 factorial design was chosen using the block generator I = 12345 = –1. 152 As a result, the resolution of the design does not permit the fitting of a third order model without the completion of the other half of the fractional factorial design. However, it should be noted that these experiments were not conducted in blocks. Instead, the order of the experiments was fully randomized and pretreatment was conducted over six days. A single centre point replicate was included in each day of pretreatment and to capture the variability of the steam pretreatment process, the centre points were also distributed equally at the beginning, middle, and end of each day of pretreatment.  50  Table 2-4. Experimental design for the response surface methodology model of lodgepole pine. Independent variables are presented in coded form. Log Ro Temperature Time SO2 Chip size Moisture content Run (---) (---) (---) (---) (---) (---) a -1 -1 -1 -1 -1 3.24 1 1 1 -1 -1 -1 4.25 4 1 -1 1 -1 -1 3.83 6 -1 1 1 -1 -1 3.66 7 1 -1 -1 1 -1 3.83 10 -1 1 -1 1 -1 3.66 11 -1 -1 1 1 -1 3.24 13 1 1 1 1 -1 4.25 16 1 -1 -1 -1 1 3.83 18 -1 1 -1 -1 1 3.66 19 -1 -1 1 -1 1 3.24 21 1 1 1 -1 1 4.25 24 -1 -1 -1 1 1 3.24 25 1 1 -1 1 1 4.25 28 1 -1 1 1 1 3.83 30 -1 1 1 1 1 3.66 31 b 0 0 0 0 0 3.79 33 0 0 0 0 0 3.79 34 0 0 0 0 0 3.79 35 0 0 0 0 0 3.79 36 0 0 0 0 0 3.79 37 0 0 0 0 0 3.79 38 c -2 0 0 0 0 3.20 39 2 0 0 0 0 4.38 40 0 -2 0 0 0 2.79 41 0 2 0 0 0 4.07 42 0 0 -2 0 0 3.79 43 0 0 2 0 0 3.79 44 0 0 0 -2 0 3.79 45 0 0 0 2 0 3.79 46 0 0 0 0 -2 3.79 47 0 0 0 0 2 3.79 48 a b Samples 1 – 31 belong to the fractional factorial design. Samples 33 – 38 are replicates of the centre point. c Samples 39 – 48 are axial points.  51  2.9  Statistical analysis and response surface methodology Response surface methodology is a statistical technique which allows one to  simultaneously study the effects of multiple parameters on a single process outcome. In the case of acid catalyzed steam pretreatment, the most important process related independent parameters are temperature, time, and catalyst concentration, while the most important feedstock characteristics are initial moisture content and size. In this thesis, the effects of each of these five independent variables on the outcomes of the SO2 catalyzed steam pretreatment and subsequent enzymatic hydrolysis of softwood were assessed with RSM. For the portion of the thesis described in Chapter 3, data was taken from three previous studies of the SO2 catalyzed steam pretreatment of softwood.  63-65  A second-order approximation,  whose generalized form is provided, was fitted to each set of data. Limiting these RSM models to a second order approximation made possible a comparison of the three different experimental designs used in the previous studies. A second order approximation was also used for the portion of the thesis described in Chapter 4, namely that regarding the development of the RSM model of lodgepole pine. However, it should be noted that the two feedstock characteristics were present as independent variables only in this last portion of the thesis. In the generalized equation, the predicted value of an individual response variable is denoted Y, while all five independent variables (k), namely temperature (T), time (t), SO2 loading (S), chip size (Si), and moisture content (Mc) are present in their coded form (Xi). 124, 153  The approximation includes an intercept, denoted a0, as well as linear terms of the form  aiXi. The presence of the quadratic terms of the form aiiXi2 and the cross-product terms of the form aijXi·Xj allow for the estimation of curvature in the response. The expanded form of the  52  second-order approximation is not given here. Nonetheless, it should be noted that the approximation possessed a maximum of 10 terms when only three independent variables (T, t, and S) were present, and a maximum of 21 terms when all five independent variables were present.  k  k 1  i 1  i 1  Y  a0   ai X i    k  k  j 1  i 1   aij X i  X j   aii X i2  (Equation 2-6)  Multiple linear regression was used to fit this equation to each response variable and an extended analysis of variance (ANOVA) was performed to determine which terms were significant at a 95 % confidence interval. All data analysis was conducted with the commercially available software Statistica 6.0 (StatSoft Inc., Tulsa, Oklahoma). Those terms found to be statistically insignificant (p > 0.05) were removed from the equation and the values of the remaining regression coefficients determined. When present, marginally significant terms (0.10 ≤ p ≥ 0.05) were also included in the equations. If the extended ANOVA determined that the resulting model possessed a significant lack-of-fit, no insignificant terms were reintroduced to the equation. The lack-of-fit test ratio (Flof) was used to test each equation, for which a plof value less than 0.05 indicated a lack-of-fit. In some cases, transposed (logarithmic) values of the responses were used to generate equations with R2 values above 0.80, the third and final model building criteria employed in this thesis.  53  Chapter 3: The use of predictive models to optimize sugar recovery in the steam pretreatment step of a softwood-to-ethanol process 3.1  Background Considerable improvements including acid catalysis have been made to the steam  pretreatment process in recent years. Nonetheless, much of the past work to try to optimize soluble sugar recovery after steam pretreatment has been limited to qualitative comparisons made on the basis of severity factors, a process which offers limited insight into the pretreatment itself. In the work described here the relative ability of two types of empirical models to accurately predict the performance of the steam pretreatment of softwoods was determined. The models were based on the thermal severity factor Ro, the combined severity factor CS, and response surface methodology (RSM), while the monitored outcomes of steam pretreatment consisted of the yields and compositions of the resulting solid and liquid fractions. As part of the comparison, an attempt was also made to determine which of the two model types could better identify the optimal conditions of acid catalyzed steam pretreatment for the maximum production of soluble sugars using softwood chips of industrially relevant size. While the empirical models presented in this work overlook parameters related to reaction kinetics and transport phenomena, precedents such as the H factor suggest that it is reasonable to expect that related models can be applied successfully to industrial reactors.  54  3.2 3.2.1  Results and discussion The thermal severity factor Ro The previous RSM study of the SO2 catalyzed steam pretreatment of radiata pine  employed a CCD with a centre point experiment conducted at 215 °C, 3 min (log Ro 3.86) and 2.55 w/w % SO2.  64  Again, due to the large number of experiments conducted, it was  possible to assess the effects of temperature and time on the outcome of pretreatment undertaken at single SO2 loadings. The recovery of hemicellulose as soluble sugars in the WSF obtained after pretreatment is depicted as a function of the Ro factor in Figure 3-1. Glucose was excluded from this calculation due to the inability to distinguish between glucose derived from cellulose and glucose derived from hemicellulose. Each data set in Figure 3-1 corresponded to a single SO2 loading and, during regression, was fitted with a second order polynomial. The regression equations closely approximated the experimental data over the range of pretreatment conditions investigated, a fact which the coefficients of determination (R2 0.93 – 0.97) help confirm. For example, the maximum recoveries predicted by the regression equations increased from 84 % at log Ro 3.48 and 1.0 w/w % SO2, to 87 % at log Ro 3.21 and 2.55 w/w % SO2, and finally to 93 % at log Ro 3.36 at the highest SO2 loading. The optimal pretreatment severity predicted by the regression equations varied only slightly, from log Ro 3.21 to 3.48. This agreed well with the original analysis conducted by Clark and Mackie, who found (using RSM) that the optimum pretreatment temperature and time was mostly independent from SO2 concentration.  64  However, the recovery predicted at the centre point  (77 %) differed somewhat from the experimental value of 83 ± 4 %, suggesting that there is room for improvement in the prediction accuracy of the Ro factor.  55  WSF hemicellulose sugar recovery (%)  Figure 3-1. Recovery of hemicellulose-derived sugars in the water soluble fraction as a function of the thermal severity factor Ro. Hemicellulose-derived sugars are mannose, xylose, galactose and arabinose. The mean value of the repeated centre point is depicted and the error bars represent ± one standard deviation of the mean.  While the solubilization and recovery of hemicellulose which occurs at lower severities is desirable, it is known that significant cellulose solubilization can occur at the higher pretreatment severities required to produce an easily enzymatically hydrolyzed substrate.  54  For example, it has been reported that the glucan content of the cellulose-rich  WIF obtained after pretreatment at 215 °C, 18.0 min (log Ro 4.64) and 2.55 w/w % SO2 was 14 w/w %, well below the glucan content of the untreated pine chips (43 w/w %).  64  Under  these severe conditions all of the hemicellulose is solubilized. The WIF yield was plotted in Figure 3-2 as a function of the factor Ro and found to decrease in a nearly logarithmic fashion despite the considerable solubilization of cellulose at higher pretreatment severities. 56  As suggested by the coefficients of determination (R2 0.98 and 0.94), the linear predictive equations most accurately described the experimental data at the 2.55 and 6.5 w/w % SO2 concentrations, respectively. For example, the WIF yield predicted at the centre point was 60 w/w %, well within the range determined experimentally (60 ± 2 w/w %). Even at the highest pretreatment severity of 235 °C, 9.00 min (log Ro 4.93), and 6.50 w/w % SO2, where cellulose hydrolysis was highest, good agreement was obtained between the predicted and experimental values, 45 and 47 %, respectively. Together, these results suggested that accurate models can be developed to predict at least some of the primary outcomes of steam  WIF yield (w/w %)  pretreatment using the Ro factor.  Figure 3-2. Yield of the water insoluble fraction as a function of the thermal severity factor Ro. Each set of data was limited to experiments conducted at a single SO2 loading. The mean value of the repeated centre point is depicted and the error bars represent ± one standard deviation of the mean.  57  3.2.2  Comparison of the Ro and CS factors To allow for a comparison of the predictive capabilities of the Ro and CS factors, a  series of steam pretreatment experiments was undertaken where the concentration of H2SO4 was varied between 0.65 and 1.90 w/w % on a dry weight basis (Table 2-2). In previous work, which looked at the H2SO4 catalyzed steam pretreatment of Norway spruce, it was determined that the highest combined recovery of soluble glucose and mannose after both pretreatment and enzymatic hydrolysis occurred following pretreatment at 210 °C, 1.0 min, and 2.4 w/w % H2SO4. 74 The pH term in the CS factor was estimated in two distinct ways. For one data set (CS acid) the pH term was represented by the measured pH of the dilute acid used to impregnate the raw wood. This measured value was found to agree well with the pH calculated using the pKa values of H2SO4. For the other data set (CS WSF), the pH of the WSF obtained after pretreatment was used to calculate the CS factor. Predictive equations were generated for each set of data through linear regression.  3.2.2.1  Residual hemicellulose in the water insoluble fraction The residual hemicellulose content of the WIFs derived from the experiments CS1 –  CS7 (unsolubilized hemicellulose present in the WIF) was plotted as a function of both Ro and CS factors in Figure 3-3. Glucose was once again excluded from this calculation. As indicated by the coefficient of determination (R2 0.85), the predictive equation developed on the basis of the Ro factor agreed fairly well with the experimental data.  58  WIF residual hemicellulose recovery (%)  Figure 3-3. Hemicellulose remaining in the water insoluble fraction as a function of the thermal severity factors Ro and CS. The mean value of the repeated experiments (CS 3 – CS 5) is depicted and the error bars represent ± one standard deviation of the mean.  The predictive equations developed on the basis of the CS factor were also found to accurately describe the amount of residual hemicellulose in the WIF. For example, the WIF obtained at the highest severity was found to contain 7 % of the original hemicellulose, which is in good agreement with the value predicted by the two equations based on the CS factor, namely 5 %. However, the fact that the coefficients of determination of all three predictive equations were found to be similar (R2 0.85 – 0.92) suggests that little improvement in prediction was afforded by the use of the CS factor. The coefficient of determination shared by the two predictive equations based on the CS factor (R2 0.92) suggested that predictions of similar accuracy were made regardless of  59  how the CS factor was calculated. This was somewhat unexpected as it was thought that the pH of the dilute acid used for impregnation would provide a more accurate estimate of the level of catalyst action during steam pretreatment. Specifically, it was thought that the action of the acid catalyst would not be well represented over the range of pretreatment severities investigated by the pH values of the WSF, due both to the variable volume of condensate generated during steam pretreatment and the use of water to separate the resulting solid and liquid fractions. Indeed, the WSF volume doubled from just over 1 L at the lowest severity to just over 2 L at the highest. As listed in Table 2-2, the pH of the dilute H2SO4 acid ranged from 0.51 to 0.95, a range of 0.50 pH units. By comparison, the pH of the WSF ranged from 2.12 to 2.39, a span of only 0.27 units. As a result, the pH of the acid used for impregnation produced a slightly wider range of CS factor values than the pH of the WSF.  3.2.2.2  Yield of the cellulose-rich water insoluble fraction As shown in Figure 3-4, the Ro factor was also able to provide accurate prediction of  the WIF yield (R2 > 0.96). For example, the yield predicted at the moderate severity of 195 °C, 3.00 min (log Ro 3.27) and 1.30 w/w % H2SO4 (69 w/w %) was found to be in good agreement with the experimental value of 69 ± 1 w/w %. This suggested that, over the range of catalyst concentration investigated, the temperature and time of pretreatment had a much greater influence on this outcome of pretreatment than did the acid catalyst. In fact, the coefficients of determination of the two predictive equations developed on the basis of the CS factor (R2 > 0.98) were nearly identical to that of the equation based on the Ro factor. As a result, the two equations based on the CS factor offered only small improvements on the predictive capability of the Ro factor.  60  WIF yield (w/w %)  Figure 3-4. Yield of the water insoluble fraction as a function of the thermal severity factors Ro and CS. The mean value of the repeated experiments (CS 3 – CS 5) is depicted and the error bars represent ± one standard deviation of the mean.  3.2.2.3  Sugar recovery in the hemicellulose-rich water soluble fraction While prediction of the hemicellulose sugar content of steam pretreated derived WIFs  is already useful from a process control point of view, good predictability of the hemicellulose-derived sugar content of the WSF is just as desirable. The recovery of hemicellulose-derived sugars in the WSF (as both monomers and oligomers) is presented in Figure 3-5 as a function of both the Ro and CS factors. Glucose was once again not considered. Each set of data was fit with a second order polynomial during regression and the resulting equations were included in Figure 3-5. According to the three predictive equations, the maximum recovery of hemicellulose-derived sugars in the WSF did not deviate from 78  61  % of theoretical. Among the equations developed, the two based on the CS factor were once again found to be of similar accuracy. However, the one based on the Ro factor was found to be in best agreement with the experimental data (R2 > 0.99). Although it was anticipated that the acid catalyst’s influence would be substantial, it appears that for a green raw material within a narrow size distribution, the temperature and time of pretreatment remain far more important variables in the steam pretreatment of softwoods than is the concentration of the  WSF hemicellulose sugar recovery (%)  acid catalyst.  Figure 3-5. Recovery of hemicellulose-derived sugars as a function of the thermal severity factors Ro and CS. The mean value of the repeated experiments (CS 3 – CS 5) is depicted and the error bars represent ± one standard deviation of the mean.  62  The ability to predict the quantity of glucose present in the liquid fraction obtained after steam pretreatment is also highly desirable. Over the range of pretreatment conditions investigated, glucose accounted for between 15 and 50 % of all sugars that were detected. Unfortunately, no obvious correlation was found between either of the Ro or CS factors and the recovery of soluble glucose (data not shown). Similarly, there was no correlation between the Ro or CS factors and all sugars present in the WSF including glucose (data not shown). It appears that, due to an inability to differentiate between the hemicellulose-derived and cellulose-derived glucose, neither the Ro nor CS factors could be used to predict the glucose content of the liquid fractions obtained after the steam pretreatment of softwood. It is not clear whether this would also be the case with non-softwood biomass, for which the source of glucose is almost exclusively cellulose.  3.2.3  Comparison of the Ro factor and response surface methodology Using the data from the previous RSM study of the SO2 catalyzed steam pretreatment  of New Zealand radiata pine, predictive equations were developed for a total of 11 response variables.  64  The predictive equations are listed in Tables 3-1 and 3-2, respectively, while  selected response surfaces are presented in Figure 3-6. The results of the extended ANOVA are presented in Tables A-1 and A-2 of Appendix A. The lack-of-fit test suggested that limiting the analysis of the radiata pine data to a second order approximation often resulted in RSM models which could not describe the experimental data as well as the original third order models.  64  Nonetheless, the R2 values reported in Tables 3-1 and 3-2 were generally  high (R2 0.80 – 0.96).  63  Table 3-1. Fitted coefficients of the response surface methodology model of the water insoluble fraction of steam pretreated radiata pine. Independent variables are presented in coded form. Response Model R2 60.89 - 9.76T b - 8.18t c - 4.10S d + 2.50tS + 1.91t2 0.94 Yield a 2 2 75.87 - 26.41T - 22.14t - 6.62S - 13.93Tt – 4.79T + 3.78S 0.95 Glucan 2 2 -6.10T - 10.61t - 3.61S + 4.63Tt + 4.50tS + 4.48T + 6.45t 0.95 Mannan 2 2 -10.40T - 13.95t - 4.70S + 7.24Tt + 4.74tS + 7.14T + 8.66t 0.96 Xylan a b c d Equation created in units of w/w % of raw wood. Temperature. Time. Sulphur dioxide loading. Table 3-2. Fitted coefficients of the response surface methodology model of the water soluble fraction of steam pretreated radiata pine. Independent variables are presented in coded form. Response Model R2 45.24 + 0.71T + 3.37S - 6.35Tt - 4.18tS - 4.22T2 - 4.77t2 - 1.75S2 0.83 Total a b 2 2 82.35 - 15.32T - 13.45t - 15.84Tt - 7.37T - 9.68t 0.93 Hemicellulose Glucose  27.45 + 8.29T + 6.34t + 4.15S - 1.86Tt - 4.14tS - 2.72T2 - 2.44t2 - 2.47S2  0.85  80.89 - 16.04T - 13.30t - 15.25Tt - 6.02T2 - 8.83t2 0.95 Mannose 2 2 81.00 - 16.33T - 14.77t - 18.38Tt - 8.42T - 11.59t 0.93 Xylose 2 2 2 100.25 - 11.55T - 11.69t + 6.90S - 15.74Tt - 13.97T - 13.26t - 5.53S 0.96 Galactose 70.25 - 14.01T - 13.29t - 11.66Tt 0.80 Arabinose a Recovery of all soluble sugars (glucose, mannose, xylose, galactose and arabinose). b Sugars derived from hemicellulose are mannose, xylose, galactose and arabinose.  3.2.3.1  Residual hemicellulose in the water insoluble fraction It was anticipated that it would prove difficult to develop a predictive model for the  residual hemicellulose content of SO2 steam pretreated radiata pine using RSM. Only the mannan and xylan content of the WIF were monitored in the previous study from which the data was sourced and it is likely that several of the cellulose-rich solid fractions containing considerable amounts of residual mannan and xylan also contained smaller amounts of galactan and arabinan. In addition, residual mannan and xylan were present in only 6 of the WIFs generated at the 15 different pretreatment conditions. This meant that a relatively small  64  set of data in which the response varied with the three independent variables was available for inclusion in the potential model. Fortunately, WIFs containing residual hemicelluloses were produced over a wide range of pretreatment conditions, namely 182 – 235 °C, 0.50 – 9.00 min (log Ro 2.80 – 3.97), and 1.00 – 6.50 w/w % SO2. As a result, RSM models of both the mannan and xylan contents of the WIF could be created. The coefficients of determination, R2 0.95 and 0.96, respectively, indicated that the models closely approximated the experimental data. The models themselves indicated that a wide range of pretreatment conditions could be used to completely solubilize the hemicellulose. The response surfaces are presented in Figure 3-6 for which it should be noted that the lack of fit could not be assessed. None of the cellulose-rich solid fractions generated at the centre point contained residual hemicellulose and, as a result, the standard deviation of these replicates was zero.  65  Figure 3-6. Effects of temperature and time on the direct outcomes of the SO 2 catalyzed steam pretreatment of radiata pine. All response surfaces are depicted at 2.55 w/w % SO2.  3.2.3.2  Yield of the cellulose-rich water insoluble fraction The yield of the WIF exiting the pretreatment reactor was also modeled. According to  the ANOVA results, the linear term of temperature was once again the most significant  66  variable while the time and SO2 concentration used were only slightly less significant. The predicted WIF yield at the centre point (61 ± 2 w/w %) was in good agreement with not only the experimentally determined value (60 ± 2 w/w %) but the value predicted by the equation developed on the basis of the Ro factor (60 w/w %). The agreement between experimental WIF yields and the values predicted by the RSM equation often extended to more extreme pretreatment severities (data not shown).  3.2.3.3  Sugar recovery in the hemicellulose-rich water soluble fraction The recovery of hemicellulose-derived sugars in the WSF (as both monomers and  oligomers) was one of the responses modeled. As judged by the relative p values in Table A2 of Appendix A, the results of the ANOVA suggest that temperature was the most significant variable. According to the predictive equation, the maximum recovery of 92 ± 9 % occurred at 182 °C and 5.49 min (log Ro 3.15). This agreed well with earlier work where maximum hemicellulose-derived sugar recovery from Douglas fir whitewood occurred at 175 °C, 7.5 min (log Ro 3.08), and 4.5 w/w % SO2. 63 The hemicellulose-derived sugars present in the WSF were also modeled individually. As listed in Table 3-3, the maximum recovery of the four individual sugars ranged from 95 % to full recovery. The equations also suggested that some of the maximum recoveries occurred at combinations of high temperatures and longer residence times.  67  Table 3-3. Maximum predicted sugar recoveries in the water soluble fraction of steam pretreated radiata pine and their corresponding process conditions. Temperature Time Log Ro SO2 Maximum recovery Response (°C) (min) (---) (w/w %) (%) a 237 0.56 3.78 11.80 51 Total b 182 5.49 3.15 --92 Hemicellulose 234 3.42 4.48 5.38 39 Glucose 182 5.61 3.16 --95 Mannose 191 3.28 3.20 --98 Xylose 213 1.42 3.48 6.43 100 Galactose 183 12.90 3.60 --100 Arabinose a Recovery of all soluble sugars (glucose, mannose, xylose, galactose and arabinose). b Sugars derived from hemicellulose are mannose, xylose, galactose and arabinose.  The predicted maximum hemicellulose-derived sugar recovery was in good agreement with the experimental maximum (91 %) as well as the maximum values predicted when this same data was analyzed on the basis of the Ro factor (88 – 91 %). Although the optimum thermal severity determined in this study with RSM (log Ro 3.15) did not lie within the range predicted by the quadratic equations developed on the basis of the Ro factor (log Ro 3.21 – 3.48), the differences were minor. Specifically, the difference between log Ro 3.15 and 3.21 was only 2 °C. In addition, the optimum thermal severity determined by the RSM for hemicellulose-derived sugar recovery was log Ro 3.15, while the optimum determined using the Ro factor for mannose recovery was log Ro 3.16. In addition to hemicellulose-derived sugar recovery and WIF yield, a comprehensive model of acid catalyzed steam pretreatment must consider the glucose content of the sugar rich liquid fraction. This is especially true for softwoods, for which solubilization of the primary hemicellulose galactoglucomannan contributes substantially to the hexose content of the WSF generated at low and moderate pretreatment severities. Over the wide range of pretreatment conditions investigated previously in the RSM study of radiata pine, glucose 68  accounted for between 14 and 80 % of all sugars recovered in the WSF.  64  One method to  account for soluble glucose is to treat all sugars in the WSF as a single response variable. This response is included in Table 3-2. Due to the considerable hydrolysis of cellulose under severe pretreatment conditions, maximum total sugar recovery in the WSF was found to be 51 ± 14 % near the upper border of the temperature and SO2 concentration ranges investigated, namely 237 °C, 0.56 min (log Ro 3.78), and 11.80 w/w % SO2. However, the model of total sugars was less accurate (R2 0.83) than the model of hemicellulose-derived sugar recovery in the WSF (R2 0.93). Relatively poor agreement between the predicted and experimental values was not unexpected, given that the total sugar content of the WSF obtained after steam pretreatment is a combination of both cellulose and hemicellulosederived sugars. It should be noted that the originally published third-order model of this response also showed a significant lack-of-fit.  64  Although the glucose present in the WSF  was also modeled individually, the model for WSF glucose recovery was also among the least accurate (R2 0.85). As listed in Table 3-3, the maximum glucose recovery measured in the WSF was found to be 39 ± 28 % and to occur at a much higher severity (log Ro 4.48) than both the maximum recovery of hemicellulose-derived and total sugars. Nonetheless, this individual RSM model of glucose recovery was an improvement on the predictive capability of both thermal severity factors, from which no equation could be developed.  3.2.4  Comparison of response surface methodology models The ability of RSM models to describe the SO2 catalyzed steam pretreatment process  accurately and to identify the optimum conditions of this process depends on several parameters. These include the ranges of temperature, time, and acid catalyst concentration  69  investigated, as well as the experimental design selected. Using a single methodology for model development, the ability of three different experimental designs to produce predictive pretreatment equations which could identify optimal operating conditions was assessed.  63-65  This was done by limiting the models based on each of the three experimental designs to a second order approximation. In addition to this methodology, consideration was also given to the number of experiments inherent to each of the RSM experimental designs compared in this work. Although the CCD based RSM model of radiata pine discussed in the previous section accurately described many direct outcomes of SO2 catalyzed steam pretreatment, many of the responses were found to be unrealistically independent of SO2 concentration. For example, the authors of the original study themselves reported an SO2 dependence of the maximum hemicellulose-derived sugar recovery in the WSF obtained after steam pretreatment. 64 In addition, several of the predictive equations for sugars present in the WSF were found to possess a significant lack-of-fit, and together this suggested that this CCD based RSM model could be improved upon. To determine whether the choice of experimental design or the ranges of process variables investigated had affected the quality of the model, two additional RSM models were created using previously generated data. The first additional RSM model was developed using data from a previous study of the SO2 catalyzed steam pretreatment of a mixture of white fir and ponderosa pine.  63  The  centre point experiment of the hybrid design study was conducted at 185 °C, 4.5 min (log Ro 3.16), and 2.5 w/w % SO2, while the remaining pretreatment was carried out over the ranges 165 – 205 °C, 1.5 – 7.5 min (log Ro 2.38 – 3.74), and 0.5 – 4.5 w/w % SO2. Several responses were modeled using the single rigorous methodology described previously,  70  including the WIF yield and the recovery of both glucose and hemicellulose-derived sugars in the WSF. The hemicellulose-derived sugars were also modeled individually (data not shown). As indicated in Table 3-4, each of the resulting equations possessed a high coefficient of determination (R2 0.95 – 0.99) and several statistically significant SO2 terms. All of the equations passed the lack-of-fit test. The results of the extended ANOVA are presented in Table A-3, while the uncoded values of the independent variables have been reported in the literature.  63  It is likely that the ranges of temperature, time, and SO2  investigated in the hybrid design study led to the successful development of this RSM model. For example, both the temperature and time ranges investigated in the hybrid design study were lower than the ranges studied in the CCD study of radiata pine, and this meant that the contribution of SO2 to the outcomes of steam pretreatment was greater in relation to the cook temperature and time. In addition, the range of SO2 investigated in the hybrid design study was limited such that the influence of this variable existed over the entire range investigated. By contrast, this was not the case for the CCD study of radiata pine, for which the influence of SO2 was found to be minimal in the upper half of the range investigated.  Table 3-4. Fitted coefficients of the response surface methodology model of the water insoluble and water soluble fractions of steam pretreated white fir and ponderosa pine. Independent variables are presented in coded form. Response Model R2 0.95 88.13 - 4.15T - 1.44t - 1.53S - 0.22TS Yield a Hemicellulose  b  85.60 + 10.56T + 4.09t + 7.95S - 4.94Tt - 2.68TS - 1.75tS - 3.49t2 - 2.41S2 - 5.23T2  0.99  0.99 6.72 + 1.39T + 0.54t + 0.72S - 0.26t2 - 0.21S2 Glucose a b Equation created in units of w/w % of raw wood. Sugars derived from hemicellulose are mannose, xylose, galactose and arabinose.  71  The second additional RSM model was developed using data from a previous study of the SO2 catalyzed steam pretreatment of Norway spruce (Picea abies (L.) H. Karst) and Scots pine (Pinus sylvestris L.). 65 The centre point experiment of the study, which was based on a face-centered central composite experimental design (CCFD), was conducted at 210 °C, 5.5 min (log Ro 3.98), and 3.5 w/w % SO2. The remainder of the experiments were conducted over the range 190 – 230 °C, 2 – 15 min (log Ro 2.95 – 5.00), and 1 – 6 w/w % SO2. The results of the extended ANOVA are presented in Table A-4, while the uncoded values of the independent variables have been reported in the literature.  65  Due to the limited amount of  data available, only four responses were modeled, namely the WIF yield, and the individual recoveries of glucose, mannose and xylose in the WSF. The single rigorous methodology outlined previously was once again employed. Equations generated from this set of existing data were generally poor (Table 3-5). Only the equation developed for mannose recovery possessed statistically significant SO2 terms, passed the lack-of-fit test, and had a high coefficient of determination (R2 0.97, data not shown).  Table 3-5. Fitted coefficients of the response surface methodology model of the water insoluble and water soluble fractions of steam pretreated Norway spruce. Independent variables are presented in coded form. Response Model R2 N/A N/A Yield a, b b, c N/A N/A Hemicellulose b N/A N/A Glucose a b Equation created in units of w/w % of raw wood. No equation could be created for this response variable. c Sugars derived from hemicellulose are mannose, xylose, galactose and arabinose.  Of the two additional experimental designs compared in this study, better models were generated using the data taken from the hybrid design study. The analysis of all three 72  RSM experimental designs suggested that the most robust RSM model of the SO2 catalyzed steam pretreatment of softwood was developed with a hybrid experimental design due very likely to the use of a narrow process space. An additional advantage of the hybrid design is the relatively limited number of experiments which must be undertaken. However, it was also apparent that none of the experimental designs was inherently superior to another and that all three designs are likely appropriate for modeling the SO2 catalyzed steam pretreatment of softwood. To ensure the successful development of an accurate model, careful consideration should be given to the ranges of process variables to be studied, and where appropriate, they should be limited as much as possible.  3.2.5  Validation of the response surface methodology model of radiata pine Response surface methodology models are not mechanistic but rather empirical in  nature. Nonetheless, efforts were made to validate the CCD based model of radiata pine and determine its robustness using a 2 L Stake Tech reactor. As a first validation step, the ability of the RSM model to predict the outcome of the SO2 catalyzed steam pretreatment of radiata pine was tested. This substrate was used to ensure that the influence of the pretreatment reactor was being tested rather than characteristics of the substrate. The quantities and compositions of the solid and liquid fractions generated during pretreatment of radiata pine in the 2 L Stake Tech reactor at 215 °C, 3 min (log Ro 3.86), and 2.55 w/w % SO2 are shown in Figures 3-7 and 3-8, respectively, and are compared to the model predictions on the basis of recovery and, in the case of WIF yield, weight percent. Ultimately, the comparisons suggest that pretreatment conducted in the 2 L Stake Tech reactor differed fundamentally from the pretreatment carried out in a 3 L reactor as reported in the previous RSM study of radiata  73  pine. 64 Specifically, it appeared that the specificity with which hemicellulose was solubilized in the 2 L reactor was comparatively low.  Figure 3-7. Comparison of sugar recoveries in the water soluble fraction obtained after the steam pretreatment of radiata pine at 215 °C, 3 min, and 2.55 w/w % SO2 to response surface methodology model predictions. RSM, RP, and LPP denote response surface methodology model, radiata pine, and lodgepole pine, respectively.  74  Figure 3-8. Comparison of water insoluble fraction yield and composition obtained after the steam pretreatment of radiata pine at 215 °C, 3 min, and 2.55 w/w % SO 2 to response surface methodology model predictions. RSM, RP, and LPP denote response surface methodology model, radiata pine, and lodgepole pine, respectively.  The recovery of hemicellulose-derived sugars in the WSF obtained after the steam pretreatment of radiata pine (as both monomers and oligomers but excluding glucose) was determined to be 71 ± 2 %, which was far lower than the predicted value of 82 ± 6 %. Conversely, levels of residual mannan and xylan in the WIF obtained using radiata pine were found to be higher than those predicted by the model. Although it appeared that the specificity with which hemicellulose was solubilized in the 2 L reactor was comparatively low, comparable levels of glucan solubilization and recovery in the WSF obtained after steam pretreatment were found to occur. The level of glucan in the WIF (51 ± 1 w/w % WIF) was comparable to the model prediction (56 ± 3 w/w % WIF), and this was also true for the 75  recovery of glucose in the WSF (23 ± 1 %), predicted by the model to be 28 ± 1 %. Note that the error reported for the experimental values represents one standard deviation while for the model predictions the bounds of the 95 % confidence interval were used. The comparison between the model and the experimental values also demonstrates that the gross effect of the steam pretreatment, as measured by WIF yield, did not differ between the two reactors. The WIF yield (58 ± 2 w/w %) determined experimentally at this pretreatment condition was found to agree with the RSM model (60 ± 1 w/w %). Several reasons exist for this difference in pretreatment performance, including the size and configuration of the two pretreatment reactors. The reactor used in the previous RSM study of radiata pine had a capacity of 3 L and was equipped with 2 steam inlets. 64 By comparison, the reactor used in this study had only a single steam inlet and a capacity of 2 L. The smaller size of this reactor meant that relatively few chips were used per reaction (50 g dry weight) as compared to the 3 L reactor (300 g dry weight). The effects of reactor capacity and the amount of chips used per reaction on the performance of steam pretreatment have been discussed previously in the literature.  34, 39, 154  In one study employing two separate  steam pretreatment reactors, researchers noted that similar combined recoveries of glucose after pretreatment and enzymatic hydrolysis were realized in the two reactors.  154  However,  the xylose recovery was substantially higher in the reactor with the larger, 10 L capacity. This suggests that the specificity with which the hemicellulose was solubilized during pretreatment was highly dependent on a combination of reactor capacity and the amount of chips used per reaction, and that it increased with concomitant increases in these two variables.  76  As a second validation step, lodgepole pine was steam pretreated at the conditions listed in Table 2-3, namely 215 °C, 3 min (log Ro 3.86) and 2.55 w/w % SO2. Comparing the results to the predictions of the CCD based RSM model of radiata pine allowed us to study the sensitivity of the empirical model to changes in softwood species. The WIF yields determined for both LPP1 and LPP2 were found to agree with the yield (60 ± 1 w/w %) predicted by the model, and the agreement extended to the level of glucan in the pretreatment-derived solid fraction. Strict agreement between the outcomes of the steam pretreatment of lodgepole pine and the predictive model did not generally extend past these two variables. Nonetheless, as indicated in Figure 3-7, the compositions of the liquid fractions obtained after steam pretreatment were somewhat similar to the model predictions. Given the previously discussed influence of reactor capacity, reactor configuration, and chip quantity on pretreatment performance, it was thought that some of the discrepancy between the model and the lodgepole pine could be overcome by comparing the pretreatment of radiata and lodgepole pine independently of the RSM model. The idea that different species of softwoods respond to SO2 catalyzed steam pretreatment in more or less the same way is supported in recent literature.  11, 155  In one study, it was suggested that British  Columbia lodgepole pine and Norway spruce sourced from Scandinavia were roughly comparable.  11  Another study also found that Douglas-fir and lodgepole pine from coastal  and interior regions of British Columbia reacted similarly to SO2 catalyzed steam pretreatment and subsequent enzymatic hydrolysis.  155  For this reason New Zealand radiata  pine and British Columbia lodgepole pine were expected to respond similarly to steam pretreatment. This was indeed the case, with better agreement found between the radiata and lodgepole pine pretreated in the 2 L reactor than between the same lodgepole pine and the  77  RSM model. Ultimately, this comparison demonstrates that at least some of the discrepancy between the RSM model and the lodgepole pine pretreated in this study seems to be the result of differences in reactor capacity, reactor configuration, and finally chip quantity. Unfortunately, and despite the influence of these three factors, the limited agreement between lodgepole pine and the RSM model was not sufficient to conclude that RSM based empirical models of SO2 catalyzed steam pretreatment developed using one softwood species can be used to predict the outcomes of this process for all softwoods.  3.3  Conclusion Previous work has used the thermal severity factor Ro to make qualitative  comparisons between steam pretreatment experiments. A review of the literature suggested that the Ro factor could in fact be used to develop highly accurate linear and quadratic models which proved useful for predicting the yield of the cellulose-rich steam pretreatment-derived WIFs, the residual hemicellulose content of this fraction, and the recovery of hemicellulosederived sugars in the water soluble fraction. After undertaking H2SO4 catalyzed steam pretreatment with radiata pine it was found that the CS factor modeling approach could also be used to develop highly accurate models able to predict the same outcomes of steam pretreatment. Although not anticipated, the predictive models based on the CS factor did not offer better predictions than those based on Ro. This highlighted the predominance of temperature and time in determining the performance of steam pretreatment and the relatively minor contribution of the acid catalyst over the individual ranges of these variables at which steam pretreatment is typically conducted. Because the method used to determine the value of the CS factor had little influence on the accuracy of prediction, using the CS  78  factor in combination with SO2 catalyzed steam pretreatment may also be possible. When the predictive capabilities of RSM and the Ro factor were compared, no substantial difference was found in the accuracy of the predictions. However, the Ro based model was not able to predict the recovery of glucose in the liquid fraction obtained after the steam pretreatment of radiata pine while the RSM based model was. This suggested that the more labour intensive RSM based empirical models can be superior to those models based on thermal severity factors provided some general principles are followed in their development. When two additional RSM models were developed using data generated based on different experimental designs, the hybrid experimental design was found to result in the most robust model. In particular, this model included the best prediction of soluble glucose recovery in the sugar rich water soluble fraction, and this was likely due to the fact that the ranges of temperature, time, and SO2 loading investigated in the hybrid design study were limited in comparison to the other RSM studies. It was also apparent that all three RSM experimental designs investigated in this study could likely be used to generate robust models, provided that a sufficiently narrow process space is investigated. When the robustness of the RSM model based on radiata pine was assessed using lodgepole pine and the influences of reactor capacity, reactor configuration and chip quantity were subtracted from the comparison, the agreement between the validation data and the RSM model remained limited, indicating that RSM models likely require additional variables and terms if their scope is to be broadened more successfully.  79  Chapter 4: Effects of chip size and moisture content on the combined sugar recovery in the steam pretreatment and enzymatic hydrolysis steps of a softwood-to-ethanol process 4.1  Background In the previous chapter, it was determined that accurate prediction of the direct  outcomes of the acid catalyzed steam pretreatment of softwood could be achieved with empirical rather than mechanistic models. It was also shown that empirical models based on RSM were more comprehensive than their counterparts based on thermal severity factors. The purpose of this study was to demonstrate that RSM could be used to create a more comprehensive model of both acid catalyzed steam pretreatment and subsequent enzymatic hydrolysis. It was also thought that the resulting model would be appropriate for use in an existing simulation of the entire softwood to ethanol process. Such a strategy would allow for the estimation of ethanol product cost as a function of steam pretreatment operating conditions and feedstock characteristics for the first time. The static conversion factors employed in existing simulations were made available through correspondence with Ola Wallberg of the University of Lund, Sweden and are presented in Table 4-1. It was these static conversion factors which necessitated the creation of additional RSM equations in this portion of the thesis. By considering feedstock characteristics in the expanded model, it was furthermore possible to investigate in a systematic manner the effects of chip moisture content and size on the performance of these two process steps. Standard techniques of RSM were used to this end; empirical models were fit to experimental data collected according to a CCD. If possible, the optimal conditions of steam pretreatment corresponding to the maximum 80  recovery of soluble sugars following both pretreatment and subsequent enzymatic hydrolysis were identified. The analytical data from which the models were created are presented in Tables 4-2, 4-3, 4-5, and 4-8.  81  Table 4-1. Aspen Plus simulation of the SO2 catalyzed steam pretreatment of softwood. Reaction Specification Stoichiometry Conversion a GLUCAN(Cisolid) + WATER --> GLUCOSE 0.046 1 GLUCAN(Cisolid) --> 2 WATER + HMF 0.000 2 CONVERSION GLUCAN(Cisolid) --> LEVULIN + FORMIC 0.000 3 GLUCAN(Cisolid) --> C6-DEGR + 2 WATER 0.000 4 GALACTAN(Cisolid) + WATER --> GALACTOS 0.759 5 GALACTAN(Cisolid) --> 2 WATER + HMF 0.002 6 CONVERSION GALACTAN(Cisolid) --> LEVULIN + FORMIC 0.000 7 GALACTAN(Cisolid) --> C6-DEGR + 2 WATER 0.000 8 MANNAN(Cisolid) + WATER --> MANNOS 0.742 9 MANNAN(Cisolid) --> 2 WATER + HMF 0.002 10 CONVERSION MANNAN(Cisolid) --> LEVULIN + FORMIC 0.000 11 MANNAN(Cisolid) --> C6-DEGR + 2 WATER 0.000 12 XYLAN(Cisolid) + WATER --> XYLOSE 0.735 13 CONVERSION XYLAN(Cisolid) --> 2 WATER + FURFURAL 0.002 14 XYLAN(Cisolid) --> 2 WATER + C5-DEGR 0.000 15 ARABINAN(Cisolid) + WATER --> ARABINOS 0.570 16 CONVERSION ARABINAN(Cisolid) --> 2 WATER + FURFURAL 0.430 17 ARABINAN(Cisolid) --> 2 WATER + C5-DEGR 0.000 18 CONVERSION LIGNIN(Cisolid) --> SOL-LIG(Cisolid) 0.087 19 CONVERSION ACETATE(Cisolid) --> HAC 0.450 20 CONVERSION ASH-WIS(Cisolid) --> ASH-WS(Cisolid) 0.700 21 GLUCAN(Cisolid) + WATER --> OLIG-GLU 0.000 22 CONVERSION XYLAN(Cisolid) + WATER --> OLIG-XYL 0.000 23 MANNAN(Cisolid) + WATER --> OLIG-MAN 0.000 24 GLUCAN(Cisolid) --> PSEUDO-G(Cisolid) 0.000 25 CONVERSION XYLAN(Cisolid) --> PSEUDO-X(Cisolid) 0.000 26 MANNAN(Cisolid) --> PSEUDO-M(Cisolid) 0.000 27 a Conversion is expressed in w/w % of the reactant component in the raw wood.  82  4.2 4.2.1  Results and discussion Effect of conducting steam pretreatment at high solids loading In Chapter 3, attempts to validate the RSM model of radiata pine suggested that  pretreatment conducted in batches of 50 odg in the 2 L Stake Tech reactor did not result in the hydrolysis of hemicellulose with high specificity. For this reason steam pretreatment was undertaken in this last study using the equivalent of 200 odg raw wood. From the results of preliminary work conducted with Douglas-fir it was anticipated that the batch size of 200 odg would correspond to a higher relative pretreatment severity than the 50 odg batch size employed in the remaining sections of the thesis (data not shown). This was indeed the case. In Chapter 3, when a 50 odg batch of lodgepole pine of 5/8 in (16 mm) and 56 w/w % moisture was subjected to steam pretreatment at 215 °C, 3 min (log Ro 3.86), and 2.5 w/w % SO2 (Run LPP2, Table 2-3), the resulting WIF yield was found to be 63 ± 2 w/w % (Figure 3-8). By comparison, the prediction of the RSM model of WIF yield developed in this chapter was a much lower 54 ± 6 w/w % at these conditions. This increased relative severity meant that overall carbohydrate recoveries achieved in the solid and liquid fractions obtained after steam pretreatment were low. In this study, the overall carbohydrate recovery was never higher than 65 % (data not shown) at or above a thermal severity of log Ro 4.25, despite the fact that some of these pretreatment experiments were also conducted at conditions associated with low relative severity, namely a low SO2 loading, an elevated moisture content, and finally a large chip size (Table 2-4). High levels of inhibitory compounds in the WSF were also indicative of this increased relative severity. Once again, in the case of pretreatment conducted at or above a thermal severity of log Ro  83  4.25 or higher, the total level of sugar degradation products ranged from 7 to 13 w/w % of the raw wood (Table 4-2).  84  Table 4-2. Summary of results for steam pretreated lodgepole pine: inhibitor content of the water soluble fraction. WSF composition and recoveries a Total Run Formic Acetic Acetic acid Levulinic HMF Furfural inhibitors c acid acid recovery b acid (---) (w/w %) (%) (w/w %) 1 0.9 0.9 91.1 1.1 1.2 0.8 4.0 4 2.0 1.0 92.0 4.3 2.8 1.4 10.4 6 1.6 1.0 99.5 4.0 2.6 1.1 9.4 7 1.4 1.0 94.0 2.3 1.7 1.5 6.9 10 1.1 0.6 60.3 0.8 1.9 0.7 4.5 11 1.1 0.8 78.5 2.0 1.6 0.8 5.6 13 0.5 0.9 79.0 0.6 1.2 1.1 3.4 16 2.5 1.1 101.5 5.6 2.6 1.8 12.6 18 1.1 0.8 83.1 1.7 2.3 0.8 5.8 19 1.6 1.0 97.9 3.0 2.7 1.1 8.4 21 0.5 0.6 64.1 0.0 0.8 0.7 2.0 24 1.4 0.7 65.9 3.3 2.2 0.5 7.4 25 0.4 0.8 69.2 0.0 0.6 0.6 1.6 28 1.6 1.0 90.9 2.5 3.2 1.3 8.6 30 1.3 0.7 70.4 1.1 2.3 1.2 5.9 31 0.8 0.9 89.0 1.4 1.6 1.0 4.7 CP 1.3 0.8 73.4 2.4 2.1 0.9 6.6 SE 16.9 15.3 11.9 26.8 7.5 40.6 9.6 39 1.0 0.6 60.6 0.3 0.8 0.7 2.8 40 2.8 0.7 72.7 4.1 3.3 0.9 11.2 41 0.0 0.9 78.6 0.0 0.2 0.1 0.3 42 1.5 0.8 79.3 2.5 2.4 1.6 8.1 43 1.0 1.0 90.6 0.9 1.5 1.0 4.6 44 1.4 0.8 79.1 3.1 2.1 0.9 7.5 45 1.5 0.9 87.2 2.5 2.4 1.2 7.6 46 1.1 0.9 80.6 2.1 2.0 1.4 6.6 47 0.9 0.8 84.0 1.3 1.6 0.8 4.6 48 0.8 0.9 80.1 0.4 1.5 1.3 4.0 a Composition is expressed in units of w/w % of raw wood. b Acetic acid recovery is expressed as a percentage of the acetyl groups in the raw wood. c Sum of formic acid, levulinic acid, HMF, and furfural.  85  It was nonetheless apparent that this increased relative severity and specifically the corresponding potential for low overall carbohydrate recoveries (present as the combination of the WIF and WSF) could be overcome by conducting steam pretreatment at lower thermal severities. In the case of pretreatment conducted at or below a thermal severity of log Ro 3.66, an overall carbohydrate recovery of between 74 and 94 % was achieved for the the solid and liquid fractions. In addition, much lower levels of sugar degradation products were present at these severities; the total level ranged from 1 to just 8 w/w % on a raw wood basis (Table 4-2). Conducting steam pretreatment at high solids loading was also found to generate liquid fractions containing industrially relevant concentrations of soluble sugars. As depicted in Figure 4-1, the WSF collected following pretreatment conducted at the centre point conditions (Runs 33 – 38) contained 127 ± 29 g/L total sugars, 88 % of which were monomers (112 ± 22 g/L). Of the hexose sugars present (113 ± 29 g/L), a similarly high percentage were found to be in monomeric form (98 ± 20 g/L). Note that soluble sugar concentrations were only determined for the centre point condition.  86  Figure 4-1. Sugar composition of the highly concentrated water soluble fraction obtained after the steam pretreatment of 5/8 in (16 mm) and 35 w/w % moisture lodgepole pine at 205 °C, 5 min, and 2.5 w/w % SO2. Monomeric sugars are denoted with hatching while oligomeric sugars are denoted with solid colour. The mean value of the six different water soluble fractions is depicted and the error bars represent ± one standard deviation of the mean.  Lastly, this study demonstrated that excellent overall carbohydrate recovery and extensive carbohydrate solubilization could occur simultaneously when pretreatment was conducted using a batch size of 200 odg. Once again in the case of pretreatment conducted at or below a thermal severity of log Ro 3.66, nearly all the hemicellulose and 17 – 50 % of the original glucan in the raw wood were solubilized. This is a very favourable result given that direct steam injection, which has the potential to provide highly localized and uneven cooking, was used as the heating medium for steam pretreatment conducted in this thesis.  87  4.2.2  The cellulose-rich water insoluble fraction To meet the objectives of this chapter several properties of the steam pretreatment-  derived solid fraction were monitored, namely chemical composition, yield, and enzymatic digestibility. A second order approximation was fit to the experimental data listed in Table 4-3 and the resulting models are presented in Table 4-4. Note that no models could be created for the residual galactan or arabinan content of the solid fraction due to their negligible levels. The results of the extended ANOVA are presented in Table A-5.  88  Table 4-3. Summary fraction. WIF Run yield (---) (w/w %) h j 1 56.9 4 47.6 6 47.3 7 54.3 10 56.3 11 57.4 13 59.7 16 49.9 18 52.1 19 59.9 21 61.0 24 44.8 25 62.0 28 50.2 30 49.8 31 56.5 CP k 49.6 SE l 3.4 39 m 61.2 40 41.9 41 65.8 42 49.6 43 59.3 44 47.7 45 51.2 46 49.0 47 62.9 48 a  58.9  of results for steam pretreated lodgepole pine: water insoluble  AIL  a  Glu  b  WIF composition Man c Xyl d Gal e (w/w WIF %) i  Ara  f  Ash  Glucan hydrolysis g (%)  50.09  49.96  0.51  0.49  0.10  0.08  0.06  38.1  72.66  24.03  0.15  0.11  0.03  0.03  0.05  75.1  68.17  32.50  0.25  0.08  0.04  0.04  0.08  66.5  56.15  42.55  0.29  0.17  0.06  0.05  0.04  49.7  48.34  48.63  0.39  0.29  0.06  0.05  0.11  40.3  52.50  44.05  0.34  0.26  0.06  0.05  0.05  44.9  47.35  51.90  0.45  0.36  0.10  0.09  0.05  32.5  85.85  10.98  0.16  0.14  0.03  0.03  0.03  88.1  51.31  48.86  0.31  0.26  0.04  0.04  0.06  44.5  43.84  55.31  0.32  0.26  0.04  0.04  0.07  27.8  40.74  56.98  0.55  0.60  0.06  0.06  0.06  24.3  74.23  23.77  0.00  0.00  0.00  0.00  0.26  76.8  42.83  57.39  0.49  0.58  0.08  0.07  0.04  22.5  62.48  35.03  0.32  0.24  0.04  0.04  0.09  61.7  57.11  43.14  0.34  0.24  0.04  0.05  0.14  53.2  48.79  50.90  0.26  0.18  0.04  0.03  0.06  37.4  60.30  38.38  0.28  0.22  0.04  0.04  0.09  58.4  6.77  12.13  29.50  34.41  52.21  52.85  26.33  11.1  42.54  56.68  0.60  0.59  0.08  0.07  0.06  24.5  88.40  9.89  0.00  0.00  0.00  0.00  0.11  91.0  38.84  58.26  0.92  1.01  0.18  0.12  0.02  16.5  61.81  36.55  0.23  0.21  0.03  0.04  0.13  60.5  46.36  53.05  0.42  0.31  0.07  0.05  0.02  31.5  61.53  35.53  0.18  0.11  0.00  0.00  0.19  63.1  57.62  41.32  0.22  0.07  0.00  0.00  0.12  53.9  53.19  43.25  0.29  0.12  0.05  0.04  0.03  53.8  46.71  52.18  0.54  0.52  0.08  0.07  0.10  28.5  45.43  55.46  0.47  0.38  0.07  0.10  b  c  0.06 d  e  28.8 f  Acid insoluble lignin. Glucan. Mannan. Xylan. Galactan. Arabinan. g Calculated as the glucan present in the raw wood minus that remaining in the WIF after steam pretreatment. h Expressed in units of w/w % of raw wood. i Expressed in units of odg w/w % of the water insoluble fraction j Samples 1 – 31 belong to the fractional factorial design. k Samples 33 – 38 are replicates of the centre point. l Standard error of the centre point. m Samples 39 – 48 are axial points.  89  Table 4-4. Fitted coefficients of the response surface methodology model of steam pretreated lodgepole pine: water insoluble fraction. Independent variables are presented in coded form. Response Model R2 Yield a Glucan Mannan Xylan Galactan c Arabinan c  49.87 - 4.51T - 2.37t - 1.77S - 0.95T·Mc b - 1.04Si·Mc + 1.69t2 + 0.63S2  0.88  + 2.48Mc2 43.07 - 15.08T - 9.48t - 5.70S + 3.59Mc - 3.41T·t - 3.28T·S + 3.69t2  0.91  + 6.14Mc2 1.23 - 0.58T - 0.65t - 0.23S + 0.36t2 - 0.13Si2 + 0.25Mc2 2  2  1.84 - 1.14T - 1.37t - 0.45S + 0.50T·t + 0.89t - 0.44Si + 0.46Mc  0.80 2  0.82  N/A  N/A  N/A  N/A  d  Ash a  0.08 + 0.02T + 0.01t + 0.02S - 0.01Si + 0.01Mc + 0.01T·S + 0.02T·Mc  a  Equation created in units of w/w % of raw wood. b Chip moisture content. equation could be created for this response variable. d Chip size. 4.2.2.1  0.80  - 0.02 t·Si + 0.02t·Mc - 0.01S·Si + 0.02S·Mc c  No  Composition of the cellulose-rich water insoluble fraction It was apparent that all combinations of steam pretreatment conditions investigated in  this portion of the thesis were sufficiently severe to completely hydrolyze the hemicellulose fraction of the raw wood. This was true of even the least thermally severe condition (Run 41, Table 2-4), for which only 5 % of the original mannan was present in the WIF following pretreatment at 205 °C for 0.50 min (log Ro 2.79). Correspondingly, the solid fractions were composed almost entirely of glucan and lignin. As indicated in Table 4-3, the levels of glucan and lignin ranged widely from approximately 10 to 50 w/w % and from approximately 40 to 90 w/w %, respectively. In addition, the resulting solid fractions contained substantially less ash than the raw wood, which suggests that steam pretreatment acted to transfer the majority of this inorganic fraction to the resulting hemicellulose-rich WSF. Previous work has demonstrated that maximum combined sugar recovery can be achieved following the SO2 catalyzed steam pretreatment and enzymatic hydrolysis of 90  softwood when pretreatment results in the solubilization of not only the entire hemicellulose fraction but also roughly 10 – 20 % of the glucan present in the raw wood.  64  Over the  numerous combinations of pretreatment and feedstock conditions investigated in this portion of the thesis, glucan solubilization ranged from 17 to 91 %, which demonstrates that the desired extent of 10 – 20 % was captured in the range of pretreatment conditions investigated here. The model of WIF glucan recovery listed in Table 4-4 confirms this. For example, limited glucan solubilization of 20 ± 10 % was predicted by the model to occur at 195 °C, 2.75 min (log Ro 3.24), and 1.5 w/w % SO2 using chips of 47.5 w/w % moisture. Chip size was not present in this model and at least within the range investigated in this study was found to have very limited influence on the composition of the steam pretreatment-derived solid fraction as a whole. By comparison, the models listed in Table 4-4 demonstrate that all remaining variables including moisture content each had a profound effect on the composition of the solid fraction.  4.2.2.2  Yield of the cellulose-rich water insoluble fraction A review of the previous work which demonstrated that maximum combined sugar  recovery was achieved by solubilizing the hemicellulose fraction as well as 10 – 20 % of the glucan in raw softwood demonstrates that this extent of carbohydrate hydrolysis corresponds to a WIF yield of roughly 60 – 65 w/w %.  64, 156  In this portion of the thesis the WIF yield  ranged from 44 – 66 w/w % (Table 4-3), which demonstrates that the desired yield of 60 – 65 w/w % was captured in the range of pretreatment and feedstock conditions investigated here. According to the model listed in Table 4-4, a WIF yield of 65 ± 1 w/w % was predicted to occur at the same conditions which are predicted to result in 20 % glucan hydrolysis,  91  namely 195 °C, 2.75 min (log Ro 3.24), and 1.5 w/w % SO2 using chips of 3/8 in (10 mm) and 47.5 w/w % moisture. As was the case for the composition of the steam pretreatment-derived solid fraction, temperature, time, SO2 loading, and moisture content were all found to heavily influence WIF yield. By comparison, chip size was found to have very limited influence on this gross measure of pretreatment performance and according to the model, the desired yield of 60 – 65 w/w % could be achieved over the entire range of chip size investigated here. Several possible reasons for this lack of influence exist. Firstly, it is possible that by removing the smallest, largest, and thickest chips as part of their preparation, the distribution of chip length was sufficiently narrow to ensure similar WIF yields following steam pretreatment. By comparison, one previous work varied feedstock size from 0.42 to 50 mm to demonstrate that the relative severity of steam pretreatment can be reduced as feedstock size is increased.  86  Secondly, the relatively narrow range of chip thickness present in this study may have limited the influence of chip size on WIF yield. Commercial softwood chips used in kraft pulping typically range in thickness from 3 to 10 mm.  60  In this study chip thickness ranged  primarily from 2 to 6 mm. As depicted in Figure 2-1, only a very small portion of the chips used for steam pretreatment were thicker than 6 mm. In addition, the choice of SO2 as the acid catalyst further limited the potential for chip size to influence the resulting WIF yield. Previous work demonstrated that the mass transfer of gaseous SO2 to the interior of even relatively thick wood chips is relatively rapid. 75, 84 An extended impregnation time was used in this study and it can therefore be assumed that the diffusion and adsorption of SO2 was complete prior to pretreatment. By comparison, the mass transfer of charged ionic components such as those present in kraft and sulphite pulping liquors into the interior of  92  wood chips is much slower and is not complete by the time the respective pulping reactions commence. 33, 60 As previously mentioned, the initial moisture content of the feedstock was found to have a profound effect on the yield of the steam pretreatment-derived solid fraction. Specifically, it was found that the highest yields were achieved at low and high moisture contents while the lowest yields were achieved at moderate moisture contents. As listed in Table 4-3, steam pretreatment undertaken at 205 °C, 5 min, and 2.5 w/w % SO2 (Run 47, Table 2-4) using chips of 5/8 in (16 mm) and 10 w/w % moisture resulted in a WIF yield of 63 w/w %. When the moisture content was increased first to 35 (Runs 33 – 38) and then to 60 w/w % (Run 48), the yield first decreased to 50 ± 2 w/w % before rising again to 59 w/w %. The resulting model which includes a quadratic moisture content term demonstrates that the lowest WIF yield and therefore the greatest gross effect of steam pretreatment on the raw wood always occurred at moderate moisture contents ranging from 30 to 40 w/w % (Table 44). The response surfaces of Figure 4-2 are set at an SO2 loading of 1.5 w/w %, yet analysis of the corresponding model demonstrates that this relationship holds true over the range of SO2 loadings investigated in this study. In addition, the response surfaces presented in this figure demonstrate that the desired yield of 60 – 65 w/w % can be achieved over the full range of moisture content investigated in this study. As moisture content is increased above 35 w/w %, the temperature and time of steam pretreatment must be increased to achieve this desired yield. Similarly, as moisture content is decreased below 35 w/w %, a slight increase in thermal severity is required to achieve this desired yield.  93  Figure 4-2. The effect of moisture content on the yield of the water insoluble fraction obtained after the steam pretreatment of lodgepole pine. Yield is expressed as a weight percent of the oven dry raw wood and is plotted at three levels. At each moisture content, yields are plotted on the left as response surfaces and on the right as twodimensional contour diagrams. SO2 and chip size are set at 1.5 w/w % and 10 mm (3/8 in), respectively.  94  Several possible reasons exist for the non-linear relationship between moisture content and yield. The yield of the solid fraction is determined primarily by the extent of carbohydrate solubilization which occurs during steam pretreatment. Water is a reactant in carbohydrate hydrolysis reactions and at moisture contents below 30 w/w % it is likely that both the rate and extent of these reactions was water limited. Previous work suggested that green wood displayed a greater chemical reactivity than air dry wood towards hydrolysis, despite the fact that the temperature of air dry wood chips was shown to approach the intended temperature of steam pretreatment more rapidly than green wood chips. 68 Although the efficacy of the SO2 impregnation step as undertaken in this portion of the thesis was not studied, it is possible that the adsorption of gaseous SO2 to wood chips having a moisture content below roughly 30 w/w % was less complete than the adsorption to chips having a moisture content of at least 30 w/w %. Incomplete adsorption of SO2 would of course limit the rate and extent of carbohydrate hydrolysis during steam pretreatment and result in a higher than anticipated WIF yield. As mentioned previously, the gross effect of steam pretreatment was also hindered as moisture content was increased above roughly 40 w/w %. At this moisture content, water was present in sufficient quantity such that the carbohydrate hydrolysis reactions were neither water nor SO2 limited. Nonetheless, carbohydrate hydrolysis proceeded at a reduced rate and to a lesser extent due to the increased heat capacity of the moist wood chips. Indeed, previous work has demonstrated that the majority of the latent heat released upon steam condensation is used to heat the water present in green wood chips and that as a result less energy is available for hydrolysis reactions. 68  95  4.2.3  The hemicellulose-rich water soluble fraction The soluble sugar and inhibitory compound levels of the steam pretreatment-derived  liquid fraction were also monitored to meet the objectives of this portion of the thesis. A second order approximation was fit to the experimental data (Tables 4-2 and 4-5) and the resulting models are presented here in Tables 4-6 and 4-7. As was already the case for each of the RSM experimental designs considered in Chapter 3, no model could be created for the recovery of all five sugars in the WSF when treated as a single response variable. In addition, the number and type of statistically significant terms present in the models of the individual hemicellulose-derived sugars, as well as the values of their corresponding coefficients, meant that these models were very similar to the model encompassing all hemicellulose-derived sugars. The results of the extended ANOVA are presented in Tables A-6 and A-7.  96  Table 4-5. Summary of results for steam pretreated lodgepole pine: sugar recovery of the water soluble fraction. WSF recoveries a Run Hemi b Glucose Mannose Xylose Galactose Arabinose Total c (---) (%) 1 53.2 21.0 44.5 55.9 76.0 63.7 31.7 4 28.2 29.1 25.6 24.8 43.9 31.9 28.8 6 36.7 32.9 33.2 34.7 55.3 36.6 34.2 7 54.8 34.2 51.9 46.3 83.1 57.1 41.0 10 66.6 22.2 61.1 67.0 78.3 81.7 36.9 11 80.3 28.2 75.1 82.3 87.9 95.2 45.5 13 59.7 19.3 50.1 62.4 85.6 71.4 32.7 16 19.4 28.9 18.6 14.7 33.5 18.4 25.8 18 59.8 31.9 51.9 61.3 84.5 65.9 41.1 19 73.8 31.0 64.2 77.2 100.3 82.6 45.2 21 78.5 14.3 70.6 82.4 91.8 94.8 35.6 24 20.2 25.5 19.1 16.6 31.4 21.9 23.7 25 81.3 13.1 66.5 91.6 107.9 99.1 35.7 28 38.4 39.1 36.0 28.8 67.2 40.9 38.9 30 47.8 35.5 44.2 43.4 68.9 52.7 39.6 31 59.0 24.1 53.2 58.1 80.6 66.0 35.7 CP 45.6 34.8 42.6 40.2 65.1 52.3 38.4 SE 10.4 10.5 11.8 12.8 6.8 15.0 7.6 39 81.6 17.0 70.1 89.5 105.3 91.7 38.5 40 27.4 34.4 27.1 21.6 36.6 34.3 32.0 41 77.8 9.3 67.3 86.0 95.3 89.1 32.0 42 48.9 40.3 46.3 52.1 47.4 57.4 43.2 43 66.1 22.8 59.7 63.1 96.9 68.6 37.2 44 47.9 36.6 46.6 41.4 58.8 61.9 40.3 45 44.1 33.4 40.5 41.1 63.2 47.3 37.0 46 43.7 31.1 37.3 40.0 72.3 52.0 35.3 47 73.8 20.8 68.6 79.2 75.4 86.5 38.4 48 67.1 22.0 60.2 65.0 100.0 66.4 36.9 a All recoveries are calculated as a percentage of the individual sugars or group of sugars in the raw wood subsequently present in the water soluble fraction or enzymatic hydrolyzate. b Sugars derived from hemicellulose are mannose, xylose, galactose and arabinose. c Recovery of all soluble sugars as both monomers and oligomers (glucose, mannose, xylose, galactose and arabinose).  97  Table 4-6. Fitted coefficients of the response surface methodology model of steam pretreated lodgepole pine: sugar recovery of the water soluble fraction. Independent variables are presented in coded form. Response Model R2 Total a, b  N/A  Hemicellulose c  49.17 - 13.83T - 6.96t - 5.91S + 1.94Si + 1.94Mc - 6.24T·t - 2.65t·Mc  Glucose Mannose Xylose Galactose Arabinose  - 3.48S·Si - 3.68Si·Mc + 2.92t2 - 1.94Si2 + 4.70Mc2 33.30 + 3.93T + 4.66t - 3.10T·t + 2.46T·Mc - 1.78t·S - 1.91S·Mc - 1.78T2 - 2.01t2 - 2.86Mc2 45.39 - 11.35T - 5.02t - 4.60S - 6.48T·t - 2.69t·Mc - 3.84S·Si - 3.48Si·Mc + 2.15t2 - 2.32Si2 + 4.05Mc2 41.60 - 16.68T - 9.06t - 7.23S - 5.82T·t - 3.33T·Mc - 2.89t·Mc - 3.25S·Si - 5.02Si·Mc + 2.17T2 + 5.55t2 + 6.30Mc2 67.09 - 16.15T - 9.02t - 7.99S + 2.58Si + 5.77Mc - 6.36T·t - 3.37T·S - 3.67S·Mc + 2.33S2 + 4.80Mc2 54.63 - 16.45T - 8.98t - 6.49S + 3.35Si - 5.98T·t - 4.66S·Si - 5.23Si·Mc + 3.77t2 + 4.59Mc2  N/A 0.91 0.80 0.87 0.91 0.96 0.84  a  No equation could be created for this response variable. b Recovery of all soluble sugars as both monomers and oligomers (glucose, mannose, xylose, galactose and arabinose). c Sugars derived from hemicellulose are mannose, xylose, galactose and arabinose.  As indicated in Table 4-5, the recovery of glucan as soluble glucose ranged widely (9 – 40 %) over the range of steam pretreatment conditions and feedstock characteristics investigated in this study. Hemicellulose-derived sugars were recovered over an even wider range (19 – 82 %) and it was found that a recovery of hemicellulose-derived sugars above 80 % in the WSF only occurred at temperatures below 195 °C. Conversely, a recovery of all five sugars (as both monomers and oligomers) above 40 % after steam pretreatment was found to require temperatures of at least 195 °C with corresponding thermal severities of at least log Ro 3.66. Measured as a percentage of the raw wood used in pretreatment, the total level of sugar degradation products ranged widely in this study, from just 0.3 to fully 12.6 w/w % 98  (Table 4-2). In the majority of experiments the level of levulinic acid was higher than that of formic acid and in all experiments the level of HMF (0.2 – 3.3 w/w %) was found to be higher than that of furfural (0.1 – 1.8 w/w %). No model could be created for either acetic acid or furfural due to their respective large standard deviations (Table 4-2).  Table 4-7. Fitted coefficients of the response surface methodology model of steam pretreated lodgepole pine: inhibitor content of the water soluble fraction. Independent variables are presented in coded form. Response Model R2 Formic acid a  1.29 + 0.38T + 0.34t - 0.11Mc + 0.11T·S + 0.12T·Si + 0.16T2 - 0.12t2 - 0.10Mc2  Acetic acid b  N/A  Levulinic acid a  2.24 + 0.86T + 0.83t + 0.30S - 0.26Si - 0.40Mc - 0.41T·S - 0.36S·Mc  HMF a  2.07 + 0.56T + 0.41t - 0.09Si - 0.12T·t + 0.11T·Si - 0.20T·S + 0.11t·Mc  Furfural b  N/A  a  4.2.3.1  N/A  - 0.28Mc2 + 0.11S·Si - 0.16S·Mc - 0.15t2 + 0.07Si2 - 0.10Mc2  Equation created in units of w/w % of raw wood. this response variable.  0.87  0.84 0.93 N/A  b  No equation could be created for  Sugar recovery in the hemicellulose-rich water soluble fraction Chip size was found to have no influence on the recovery of glucose in the WSF. On  the other hand, the model listed in Table 4-6 and the corresponding response surfaces of Figure 4-3 demonstrate that chip size was found to have a moderate yet statistically significant influence on the recovery of soluble hemicellulose sugars. For example, at one combination of pretreatment conditions corresponding to a limited glucan solubilization of 10 – 20 % and a WIF yield of 60 – 65 w/w %, namely 195 °C, 2.75 min (log Ro 3.24), and 1.5 w/w % SO2 using chips of 3/8 in (10 mm) and 47.5 w/w % moisture, recovery was predicted to be 78 ± 2 %. At this condition, glucose recovery was predicted to be 13 ± 3 %  99  which accounted for 34 % of the sugars present in the WSF. Analysis of the model shows that as chip size was increased first to 5/8 in (16 mm) and then to 7/8 in (22 mm), the predicted recovery first increased to a maximum of 82 % before decreasing again slightly. For chips of 22.5 w/w % moisture or less, a chip size closer to 7/8 in (22 mm) was predicted to lead to the highest recovery of soluble hemicellulose-derived sugars in combination with the same steam pretreatment conditions resulting in a WIF yield of 60 – 65 w/w % and a limited glucan solubilization of 10 – 20 %. At both lower and higher moisture content, the preparation of highly homogeneous chip samples was shown to be advantageous with respect to the recovery of soluble hemicellulose-derived sugars. As pointed out by previous researchers, a simulation of the entire softwood to ethanol process would be required to determine whether such small increases merit the costs of implementing an enhanced chip classification strategy at commercial scale. 84  100  Figure 4-3. The effect of chip size on the recovery of hemicellulose-derived sugars obtained in the water soluble fraction after the steam pretreatment of lodgepole pine. Recovery is expressed as a percentage of mannose, xylose, galactose, and arabinose available in the raw wood and is plotted at three levels. At each chip size, recoveries greater than 50 % are plotted on the left as response surfaces and on the right as twodimensional contour diagrams. SO2 and moisture content are set at 1.5 and 47.5 w/w %, respectively.  101  Moisture content was found to have a far greater influence than chip size on the recovery of both soluble glucose and soluble hemicellulose-derived sugars. As depicted in the response surfaces of Figure 4-4, chips of 47.5 w/w % moisture allowed recoveries of hemicellulose-derived sugars above 80 % to be obtained over wide ranges of steam pretreatment temperature and time. Conversely, chips of lower moisture content (22.5 w/w %) greatly limited the range within which a similarly high hemicellulose recovery could be achieved. Previous work conducted with Douglas-fir found that the relative severity of steam pretreatment as measured by WIF yield and the recovery of hemicellulose-derived soluble sugars in the WSF was reduced as moisture content was increased.  86  By investigating  moisture content in a more systematic manner, this study has demonstrated that both low and high moisture contents decrease the relative severity of steam pretreatment. For example, steam pretreatment undertaken at 205 °C, 5 min, and 2.5 w/w % SO2 using chips of 5/8 in (16 mm) and 10 w/w % moisture (Run 47, Table 2-4) resulted in a WSF hemicellulosederived sugar recovery of 74 % (Table 4-5). When the moisture content was increased first to 35 (Runs 33 – 38, Table 2-4) and then to 60 w/w % (Run 48, Table 2-4), the recovery first decreased to 46 ± 5 % before rising again to 67 %. This example is limited to a single pretreatment temperature, time, SO2 loading, and finally chip size, yet analysis of the corresponding model confirms that a moderate moisture content resulted in a minimum recovery of hemicellulose-derived sugars at any combination of pretreatment operating conditions including chip size. The reasons for this non-linear relationship between moisture content and pretreatment relative severity were discussed previously.  102  Figure 4-4. The effect of moisture content on the recovery of hemicellulose-derived sugars obtained in the water soluble fraction after the steam pretreatment of lodgepole pine. Recovery is expressed as a percentage of mannose, xylose, galactose, and arabinose available in the raw wood and is plotted at three levels. At each moisture content, recoveries greater than 50 % are plotted on the left as response surfaces and on the right as two-dimensional contour diagrams. SO2 and chip size are set at 1.5 w/w % and 10 mm (3/8 in), respectively.  103  4.2.4  The combined sugar recovery after steam pretreatment and enzymatic  hydrolysis The experimental combined sugar recoveries achieved in this study were found to range widely from 31 – 80 % (Table 4-8). Of those conditions leading to a combined recovery of at least 75 %, namely 195 – 205 °C, 5.00 – 7.25 min (log Ro 3.66 – 3.79), and 1.5 – 2.5 w/w % SO2, using chips of 3/8 – 7/8 in (10 – 22 mm) and 10 – 47.5 w/w % moisture, it was not readily apparent that any one combination of steam pretreatment operating conditions and feedstock characteristics was superior to any other. At nearly every one of the combinations of steam pretreatment conditions and feedstock characteristics investigated in this study, steam pretreatment was found to contribute at least as much as enzymatic hydrolysis to the combined sugar recovery (Table 4-8).  104  Table 4-8. Summary of results for steam pretreated and enzymatically hydrolyzed lodgepole pine: combined sugar recovery. EH glucose WSF total WIF EH Combined Run a b c recovery digestibility recovery recovery recovery d (---) (%) 1 31.7 68.1 42.2 28.4 60.0 4 28.8 81.4 20.3 13.6 42.4 6 34.2 85.9 28.7 19.3 53.4 7 41.0 81.5 41.0 27.6 68.6 10 36.9 88.1 52.5 35.3 72.2 11 45.5 82.2 45.3 30.4 75.9 13 32.7 81.0 54.7 36.7 69.5 16 25.8 62.9 7.5 5.0 30.8 18 41.1 89.4 49.6 33.3 74.4 19 45.2 70.9 51.2 34.5 79.6 21 35.6 73.3 55.4 37.2 72.8 24 23.7 79.4 18.4 12.3 36.1 25 35.7 50.4 46.9 31.6 67.3 28 38.9 86.2 33.1 22.2 61.1 30 39.6 88.1 41.2 27.6 67.2 31 35.7 86.4 54.1 36.2 71.9 CP 38.4 87.7 35.8 24.0 62.4 SE 7.6 6.2 3.1 8.7 4.3 39 38.5 65.8 49.7 33.4 71.9 40 32.0 91.5 8.3 5.5 37.6 41 32.0 50.4 42.1 28.6 60.6 42 43.2 86.4 34.1 22.9 66.1 43 37.2 68.4 46.9 31.5 68.6 44 40.3 70.7 26.1 17.5 57.8 45 37.0 84.3 38.9 26.1 63.0 46 35.3 81.8 37.8 25.4 60.6 47 38.4 78.8 56.3 37.8 76.2 48 36.9 75.7 53.9 36.2 73.1 All recoveries are calculated as a percentage of the individual sugars or group of sugars in the raw wood subsequently present in the water soluble fraction or enzymatic hydrolyzate. a Sum of glucose, mannose, xylose, galactose and arabinose as both monomers and oligomers. b Conversion of the glucan available in the water insoluble fraction to soluble monomeric glucose. c Sum of glucose, mannose, xylose, galactose and arabinose in the enzymatic hydrolyzate. d Sum of glucose, mannose, xylose, galactose and arabinose recovered after both steam pretreatment and enzymatic hydrolysis.  105  The digestibility of the steam pretreatment-derived solid fractions following 72 hrs of enzymatic hydrolysis ranged widely in this study from 50 to 92 % (Table 4-8). It is true that the lowest and highest digestibility witnessed in this study corresponded to solid fractions containing the highest and lowest carbohydrate contents (60 and 10 %), respectively. Nonetheless, it was also apparent that no clear relationship existed between the glucan or lignin contents of the WIF and its enzymatic digestibility. Sugars present in the hydrolyzate were exclusively monomers; this was confirmed by subjecting one centre point hydrolyzate to post hydrolysis following 72 hrs of enzymatic hydrolysis (data not shown). No model could be created for this response variable and in fact, not one of the potential model terms was found to be statistically significant during the extended ANOVA (data not shown). Previous works have also failed to find a relationship between steam pretreatment operating conditions and the enzymatic digestibility of the resulting WIF. 64, 65 Rather, the performance of enzymatic hydrolysis was modeled by considering the soluble sugars present in the hydrolyzate as a percentage of the sugars present in the raw wood. Two such models, one considering all five sugars, and the other only glucose, are presented in Table 4-9. The results of the extended ANOVA are presented in Table A-8. Glucose was found to account for essentially all of the sugar released during enzymatic hydrolysis due to the very low residual hemicellulose content of the WIFs generated in this study. Of those conditions leading to an enzymatic hydrolysis soluble sugar recovery of at least 35 %, namely 195 – 215 °C, 2.50 – 7.25 min (log Ro 3.24 – 3.83), and 1.5 – 3.5 w/w % SO2, using chips of 3/8 – 7/8 in (10 – 22 mm) and 10 – 60 w/w % moisture, it was not readily apparent that any one combination of steam pretreatment operating conditions and feedstock characteristics was superior to any other.  106  Table 4-9. Fitted coefficients of the response surface methodology model of steam pretreated and enzymatically hydrolyzed lodgepole pine: combined sugar recovery. Independent variables are presented in coded form. Response Model R2 Enzymatic N/A N/A digestibility a, b EH glucose recovery c  36.91 - 9.26T - 4.85t - 3.40S + 2.20Mc - 5.32T·t - 4.95T·S + 1.73t·Mc  EH recovery d  24.79 - 6.23T - 3.29t - 2.30S + 1.47Mc - 3.55T·t - 3.32T·S + 1.15t·Mc  Combined recovery e  62.32 - 8.19T - 2.48t - 3.51S + 2.14Mc - 7.69T·t - 3.91T·S + 1.34T·Si  - 1.70Si·Mc - 1.82T2 + 4.71Mc2 - 1.13Si·Mc - 1.22T2 + 3.16Mc2 + 1.40T·Mc - 2.52t·S - 2.10T2 + 2.88Mc2  0.92 0.92 0.90  a  No equation could be created for this response variable. b Conversion of the glucan available in the water insoluble fraction to soluble monomeric glucose. c Recovery of glucose in the enzymatic hydrolyzate. d Recovery of all soluble sugars (glucose, mannose, xylose, galactose and arabinose) in the enzymatic hydrolyzate. e Recovery of all soluble sugars as both monomers and oligomers (glucose, mannose, xylose, galactose and arabinose) after both steam pretreatment and enzymatic hydrolysis.  4.2.4.1  Enzymatic hydrolysis of the cellulose-rich water insoluble fraction Chip size was found to have no statistically significant influence on the recovery of  soluble sugars after enzymatic hydrolysis. By comparison, the non-linear relationship between moisture content and the relative severity of steam pretreatment meant that moisture content was found to have a large influence on this response variable. For example, steam pretreatment undertaken at 205 °C, 5 min, and 2.5 w/w % SO2 using chips of 5/8 in (16 mm) and 10 w/w % moisture (Run 47, Table 2-4) resulted in an enzymatic hydrolysis sugar recovery of 38 % (Table 4-8). When the moisture content was increased first to 35 (Runs 33 – 38, Table 2-4) and then to 60 w/w % (Run 48, Table 2-4), the recovery first decreased to 24 ± 2 % before rising again to 36 %. At one combination of steam pretreatment operating conditions and feedstock characteristics leading to a WIF yield of 60 – 65 w/w % and a limited glucan hydrolysis of 10 – 20 %, namely 195 °C, 2.75 min (log Ro 3.66), and 1.5 w/w  107  % SO2 using chips of 5/8 in (16 mm) and 47.5 w/w % moisture, the recovery of soluble sugars after enzymatic hydrolysis was predicted to be 33 %. As depicted in the response surfaces of Figure 4-5, chips of 47.5 w/w % moisture allowed for enzymatic hydrolysis sugar recoveries above 35 % to be achieved over a wide range of steam pretreatment conditions. For example, steam pretreatment conducted at 195 °C, 7.25 min (log Ro 3.66), and 1.5 w/w % SO2 using chips of 3/8 in (10 mm) and 47.5 w/w % moisture (Run 19, Table 2-4) resulted in a recovery of 35 % (Table 4-8). Conversely, using chips of 22.5 w/w % moisture limited this range to combinations of higher temperatures and shorter residence times. The response surfaces of Figure 4-5 are set at an SO2 loading and chip size of 1.5 w/w % and 10 mm (3/8 in), respectively, yet analysis of the corresponding model listed in Table 4-9 suggests that the prediction of maximum recovery at combinations of lower temperature and longer residence time holds true over the full ranges of SO2 loading, chip size and moisture content investigated in this study.  108  Figure 4-5. The effect of moisture content on the recovery of all sugars (primarily glucose) obtained after the enzymatic hydrolysis of the water insoluble fraction of steam pretreated lodgepole pine. Recovery is expressed as a percentage of all sugars available in the raw wood and is plotted at three levels. At each moisture content, recoveries are plotted on the left as response surfaces and on the right as two-dimensional contour diagrams. SO2 and chip size are set at 1.5 w/w % and 10 mm (3/8 in), respectively.  109  4.2.4.2  Combined sugar recovery The model of combined sugar recovery depicted in Figure 4-6 suggests that a near  full recovery can be achieved after both pretreatment and enzymatic hydrolysis if pretreatment is conducted at combinations of lower temperature and longer residence time. The response surfaces of Figure 4-6 are set at an SO2 loading and chip size of 1.5 w/w % and 10 mm (3/8 in), respectively, yet analysis of the corresponding model listed in Table 4-9 suggests that this prediction of near full recovery at combinations of lower temperature and longer residence time holds true over the full ranges of SO2 loading, chip size and moisture content investigated in this study. The response surfaces also demonstrate that the range of thermal severity over which a combined sugar recovery of at least 70 % can be achieved was found to increase with increasing moisture content. Increasing the range within which high sugar recoveries can be achieved is advantageous for process control in that any deviation from the desired pretreatment operating conditions or feedstock characteristics is less likely to result in a large decrease in combined sugar recovery. Nonetheless, the response surfaces also show that maximum recovery is predicted to occur at the limits of pretreatment temperature and time investigated in this study, namely 185 °C and 9.5 min. For this reason it must be concluded that this study failed to capture an optimum combined sugar recovery.  110  Figure 4-6. The effect of moisture content on the combined sugar recovery obtained after the steam pretreatment and subsequent enzymatic hydrolysis of lodgepole pine. Recovery is expressed as a percentage of all sugars available in the raw wood and is plotted at three levels. At each moisture content, recoveries greater than 50 % are plotted on the left as response surfaces and on the right as two-dimensional contour diagrams. SO2 and chip size are set at 1.5 w/w % and 10 mm (3/8 in), respectively.  111  As reflected in the model of combined sugar recovery listed in Table 4-9, it was apparent that chip size had a very limited influence on the combined sugar recovery achieved after both steam pretreatment and enzymatic hydrolysis. Nonetheless the model does suggest that a smaller chip size should be used when the temperature of steam pretreatment is relatively low and that a larger chip size should be used when the temperature of steam pretreatment is relatively high. By comparison, initial moisture content was found to greatly affect this response variable. Specifically, it was found that the highest recoveries were achieved at low and high moisture contents while the lowest recoveries were achieved at moderate moisture contents. For example, steam pretreatment undertaken at 205 °C, 5 min, and 2.5 w/w % SO2 using chips of 5/8 in (16 mm) and 10 w/w % moisture (Run 47, Table 24) resulted in a combined sugar recovery of 76 %. When the moisture content was increased first to 35 (Runs 33 – 38, Table 2-4) and then to 60 w/w % (Run 48, Table 2-4), the recovery first decreased to 62 ± 3 % before rising again to 73 % (Table 4-8). The response surfaces of Figure 4-6 are set at an SO2 loading and chip size of 1.5 w/w % and 10 mm (3/8 in), respectively, yet analysis of the corresponding model suggests that combined sugar recovery approached a minimum at a moderate moisture content of roughly 35 w/w % over the full ranges of SO2 loading and chip size employed in this study. As discussed earlier, previous work has demonstrated that maximum combined sugar recovery can be achieved following the SO2 catalyzed steam pretreatment and enzymatic hydrolysis of softwood when pretreatment results in a WIF yield of 60 – 65 w/w % and the solubilization of not only the entire hemicellulose fraction but also roughly 10 – 20 % of the glucan present in the raw wood. 64, 65 In these two previous studies, enzymatic hydrolysis was conducted at a substrate consistency of 2 w/w % and at enzyme loadings of 250 and 320 mg  112  protein/g substrate, respectively. By comparison, enzymatic hydrolysis was conducted in this study at a moderate substrate consistency of 10 w/w % and a much reduced enzyme loading of 7 – 41 mg protein/g substrate (70 mg protein/g glucan) such that cellulose accessibility, non-productive binding of the enzymes to lignin, and end-product inhibition were all expected to influence the extent of the reaction. Despite these differences, Figure 4-7 demonstrates that a WIF yield of 60 w/w % was optimal in that it was found to correspond to the highest combined sugar recovery achieved experimentally in this study.  Figure 4-7. Combined sugar recovery obtained after the steam pretreatment and subsequent enzymatic hydrolysis of lodgepole pine as a function of the water insoluble fraction yield after steam pretreatment. The mean value of the centre point is depicted and the error bars represent ± one standard deviation of the mean.  113  In this study the highest combined sugar recovery did not correspond to a glucan solubilization of 10 – 20 % after steam pretreatment. For example, the steam pretreatment conditions predicted to result in a WIF yield of 65 w/w % and a glucan solubilization of 20 %, namely 195 °C, 2.75 min (log Ro 3.24), and 1.5 w/w % SO2 using chips of 3/8 in (10 mm) and 47.5 w/w % moisture, were predicted to result in a combined sugar recovery of only 65 ± 5 % with equal contributions from steam pretreatment and enzymatic hydrolysis. Rather, the highest combined sugar recovery achieved experimentally in this study (80 %, Table 4-8) occurred when steam pretreatment was conducted at 195 °C, 7.25 min (log Ro 3.66), and 1.5 w/w % SO2 using chips of 3/8 in (10 mm) and 47.5 w/w % moisture (Run 19, Table 2-4). At this condition, the level of steam pretreatment-derived inhibitory compounds totaled 8.4 w/w % (Table 4-2). The extent of glucan hydrolysis was 28 % at this condition (Table 4-3) and this suggests that the extent of glucan hydrolysis corresponding to maximum combined sugar recovery depends on the conditions at which enzymatic hydrolysis is conducted. Figure 4-8 confirms that the optimal extent of glucan hydrolysis in this study was roughly 30 %.  114  Figure 4-8. Combined sugar recovery obtained after the steam pretreatment and subsequent enzymatic hydrolysis of lodgepole pine as a function of the extent of glucan hydrolysis in steam pretreatment. The mean value of the centre point is depicted and the error bars represent ± one standard deviation of the mean.  The maximum combined sugar recovery achieved experimentally in this study was composed primarily (60 %) of steam pretreatment-derived sugars present in the WSF (Run 19, Table 4-8). By comparison, the maximum combined sugar recovery achieved in one previous study was composed of nearly equal contributions from steam pretreatment and enzymatic hydrolysis.  64  In a second previous study, steam pretreatment-derived sugars  present in the WSF in fact accounted for only 30 % of the maximum combined recovery of glucose and mannose.  65  As mentioned previously, enzymatic hydrolysis was conducted in  both these previous studies at a substrate consistency of 2 w/w % and at enzyme loadings much higher than those employed in this study. It seems that as the conditions of enzymatic 115  hydrolysis are made less ideal by increasing substrate consistency and lowering enzyme loading, pretreatment becomes responsible for an increasing portion of the maximum combined sugar recovery achieved after both steam pretreatment and enzymatic hydrolysis. In this study, the steam pretreatment conditions corresponding to the highest combined sugar recovery achieved experimentally were in fact also the conditions corresponding to the highest recovery of steam pretreatment-derived sugars present in the WSF (Table 4-8).  4.3  Conclusion Building on the results of the previous chapter, this work successfully demonstrated  that RSM could be used to create a comprehensive model of the acid catalyzed steam pretreatment of softwood and the subsequent enzymatic hydrolysis of the water insoluble fraction able to consider not only steam pretreatment operating conditions but also feedstock characteristics. In fact, a sufficient number of predictive equations were generated to warrant the inclusion of the model in a previously developed simulation of the entire softwood to ethanol process; however, the absence of a predicted maximum combined sugar recovery may limit the usefulness of such an exercise. In the process of model development This study employed a RSM experimental design to systematically determine the influences of chip size and moisture content on the SO2 catalyzed steam pretreatment of lodgepole pine and the subsequent enzymatic hydrolysis of the steam pretreatement derived WIF were evaluated. Chip size was found to have no influence on the combined sugar recovery after these two stages, and in fact, the influence of this feedstock characteristic was limited to the recovery of hemicellulose-derived sugars in the WSF. By comparison, this work showed that moisture content greatly influenced steam pretreatment, enzymatic hydrolysis, and the combined sugar recovery thereof. Specifically, 116  the relationship between moisture content and the relative severity of pretreatment was found to be non-linear. Irrespective of pretreatment operating conditions and chip size, the relative severity was found to be greatest at a moisture content of approximately 30 – 40 w/w %. As moisture content was either decreased below or increased above this range, the thermal severity of pretreatment as measured primarily by WIF yield and the extent of glucan hydrolysis was found to decrease. In addition, an elevated moisture content of approximately 50 w/w % was shown to be advantageousbecause it resulted in the highest combined sugar recovery achieved experimentally and allowed high combined sugar recoveries to be realized over the widest range of pretreatment operating conditions. The highest combined sugar recovery achieved experimentally (80 %) corresponded to pretreatment conducted at 195 °C, 7.25 min (log Ro 3.66), and a 1.5 w/w % SO2 loading using chips of 3/8 in (10 mm) and 47.5 w/w % moisture. As was the case in two previous studies, the optimum combination of steam pretreatment operating conditions resulted in a WIF yield of 60 w/w %. However, this study is unique in that the highest combined sugar recovery corresponded to a comparatively high glucan hydrolysis after steam pretreatment of nearly 30 %. Due to the relatively high substrate consistency and relatively low enzyme loading employed during enzymatic hydrolysis, pretreatment was responsible for a larger portion of the maximum combined sugar recovery achieved after steam pretreatment and subsequent enzymatic hydrolysis. Finally, a comparison of this study with preliminary work undertaken using Douglasfir demonstrated that conducting pretreatment at a higher solids loading increased the relative severity of pretreatment. Fortunately, this work also showed that this increased relative severity could be overcome such that high combined sugar recoveries could nonetheless be  117  obtained after steam pretreatment and subsequent enzymatic hydrolysis. The ability to generate highly concentrated streams rich in fermentable sugars is vital to the continued commercialization of advanced bioethanol processes. To that end, conducting steam pretreatment at a high solids loading was shown to be advantageous in that it produced a WSF with a concentration of hexose sugars above 100 g/L.  118  Chapter 5: Conclusion The overall objective of this thesis was to develop empirical models able to predict the performance of the acid catalyzed steam pretreatment of softwoods and the subsequent enzymatic hydrolysis of the steam pretreatment-derived water insoluble fraction. It was thought that the development of such models would be an aid to the continued commercialization of the softwood to bioethanol process, by, for example, positively influencing the design of effective steam pretreatment reactors and forming the basis of a dynamic simulation of the entire softwood to bioethanol process. It was also hoped that during the development of these empirical models insight into the influence of feedstock characteristics on the performance of both acid catalyzed steam pretreatment and enzymatic hydrolysis would be gained. Using both newly generated data and data taken from the literature, this work shows that severity factors can indeed form the basis of empirical models able to predict the performance of the acid catalyzed steam pretreatment of radiata pine as measured by several variables including the yield of the steam pretreatment-derived water insoluble fraction, the chemical composition of this fraction, and finally the glucose and hemicellulose-derived sugar contents of the steam pretreatment-derived water soluble fraction. As a result it was clear that neither reaction kinetics nor mass and heat transfer is required in models of steam pretreatment. However, it was also clear that the empirical models developed in this portion of the thesis possessed several limitations. The combined severity factor CS, although it accounts for all three operating conditions of acid catalyzed steam pretreatment, failed to offer predictions more accurate than those offered by the thermal severity factor Ro. Glucose was repeatedly shown to account for a considerable portion of the total sugar content of  119  steam pretreatment-derived water soluble fractions. Due likely to the fact that glucose is a constituent of both softwood cellulose and hemicellulose, neither of these model types could be used to predict the quantity of soluble glucose released during steam pretreatment. In an attempt to overcome the limitations of the severity factor based empirical models of steam pretreatment, this portion of the thesis was extended to include empirical models based on response surface methodology and developed using data taken from the literature. This strategy proved to be moderately successful in that two of the three empirical models were able to predict the performance of acid catalyzed steam pretreatment as measured by each of the aforementioned variables including the glucose content of the steam pretreatment-derived water soluble fraction. However, it was found that this approach also had its limitations. The robustness of one of the response surface methodology based empirical models was tested by comparing the model predictions to the experimental results of steam pretreatment conducted at UBC. The model was only able to predict the performance of steam pretreatment as measured by the yield of the water insoluble fraction with acceptable accuracy. It was therefore apparent that empirical models of steam pretreatment based on response surface methodology require additional independent variables if their scope is to be broadened more successfully. Much attention has already been paid to the performance of acid catalyzed steam pretreatment as influenced by reaction temperature, time, and acid catalyst loading. By using response surface methodology once again in this second study, great insight was gained into the influence of feedstock characteristics on both the SO2 catalyzed steam pretreatment of lodgepole pine and the subsequent enzymatic hydrolysis of the steam pretreatment-derived water insoluble fraction. Specifically, this study showed that chip moisture content greatly  120  influenced the outcome of both process steps as measured by several variables including the combined recovery of soluble sugars. By comparison, chip size was found to have no significant impact on the combined sugar recovery at least as varied over roughly the range of a standard mill chip. Due to the unexpected non-linear relationship between moisture content and the relative severity of pretreatment, the experimental design chosen in this study failed to capture a maximum predicted combined sugar recovery. Nonetheless, this study was a success on several accounts. Firstly, it was demonstrated that conducting steam pretreatment at a high solids loading is advantageous because it can generate a highly concentrated water soluble fraction rich in monomeric hexose sugars. Secondly, the superiority of elevated moisture contents was shown. This is a positive result because standard softwood mill chips typically possess elevated moisture contents. In addition, the highest combined sugar recovery achieved experimentally (80 %) strongly indicated that maximum combined sugar recovery is achieved when steam pretreatment results in a water insoluble fraction yield of 60 – 65 w/w %, irrespective of how enzymatic hydrolysis of washed substrate is subsequently conducted. Lastly, it is reasonable to conclude that the response surface methodology models developed in this last portion of the thesis could together form the basis of a dynamic simulation of the entire softwood to bioethanol process. However, no simulation work was undertaken in this thesis and it is possible that the absence of a predicted maximum combined sugar recovery would limit the usefulness of such an approach.  121  References 1. National Renewable Energy Laboratory, The Impact of Ethanol Blending on U.S. Gasoline Price. NREL/SR-670-44517 (2008). 2. Representative Rahall NJ II. Energy Independence and Security Act of 2007. Washington DC: Congress; (2007). 3. United States Department of Agriculture, Why Have Food Commodity Prices Risen Again? WRS-1103 (2011). 4. Core Writing Team, Climate Change 2007: Synthesis Report, Contribution of Working Groups I, II and III to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change, ed by Pachauri RK and Reisinger A. IPCC (2007). 5. Mabee WE and Saddler JN, Bioethanol from lignocellulosics: Status and perspectives in Canada. Bioresource Technol 101:4806-4813 (2010). 6. Forest Products Association of Canada and FPInnovations, The New Face of the Canadian Forest Industry: The Emerging Bio-revolution. The Bio-pathways Project (2011). 7. Tembec Inc. Products - Chemical Products [internet]. 2011 [cited 2011 Oct 17]. Available from: http://www.tembec.com/public/Produits/Produits-chimiques.html. 8. Enerkem Inc. Facilities: Plants [internet]. 2010 [cited 2011 Oct 22]. Available from: http://www.enerkem.com/en/facilities/plants/westbury-quebec-canada.html. 9. Mabee WE and Saddler JN, The potential of bioconversion to produce fuels and chemicals. Pulp Paper Mag Can 107:34-37 (2006). 10. Bura R. Bioconversion of Corn Fibre to Ethanol [PhD thesis]. Vancouver, BC: The University of British Columbia; (2004). 11. Ewanick SM, Bura R and Saddler JN, Acid-catalyzed steam pretreatment of lodgepole pine and subsequent enzymatic hydrolysis and fermentation to ethanol. Biotechnol Bioeng 98:737-746 (2007). 12. Chum HL, Johnson DK and Black SK, Organosolv pretreatment for enzymic hydrolysis of poplars. 2. Catalyst effects and the combined severity parameter. Ind Eng Chem Res 29:156-162 (1990). 13. Mabee WE, McFarlane PN and Saddler JN, Biomass availability for lignocellulosic ethanol production. Biomass Bioenerg 35:4519-4529 (2011).  122  14. FPInnovations: Forintek Canada Corporation. Lodgepole Pine Fact Sheet [internet]. 2009 [cited 2012 Jan 4]. Available from: http://www.naturallywood.com/sites/default/files/ Lodgepole-pine.pdf. 15. Renewable Fuels Association. Industry Statistics: World Fuel Ethanol Production [internet]. 2012 [cited 2012 Jan 4]. Available from: http://ethanolrfa.org/pages/World-FuelEthanol-Production. 16. The Economist. Brazilian Brew: America Opens up to Brazilian Ethanol [internet]. 2012 Jan 7 [cited 2012 Feb 8]. Available from: http://www.economist.com/node/21542431. 17. Renewable Fuels Association. Resource Center: Ethanol and Engines [internet]. 2012 [cited 2012 Jan 4]. Available from: http://ethanolrfa.org/pages/ethanol-and-engines. 18. Christiansen RC. The Cellulosic Ceiling [internet]. Ethanol Producer Magazine; 2009 Aug [cited 2012 Feb 8]. Available from: http://www.ethanolproducer.com/articles/5827/thecellulosic-ceiling. 19. Sims REH, Taylor M, Saddler JN and Mabee W, From 1st- to 2nd-Generation Biofuel Technologies: An Overview of Current Industry and RD&D Activities. International Energy Agency Bioenergy (2008). 20. Sims REH, Mabee WE, Saddler JN and Taylor M, An overview of second generation biofuel technologies. Bioresource Technol 101:1570-1580 (2010). 21. Bailey BK, Performance of Ethanol as a Transportation Fuel, in Handbook on bioethanol: Production and utilization, ed by Wyman CE. Taylor & Francis, Washington D.C., USA (1996). 22. Downey M, Finished Products, in Oil 101. Wooden Table Press LLC, LaVergne, TN, USA (2009). 23. Ginnebaugh DL, Liang J and Jacobson MZ, Examining the temperature dependence of ethanol (E85) versus gasoline emissions on air pollution with a largely-explicit chemical mechanism. Atmos Environ 44:1192–1199 (2010). 24. UNICA – Brazilian Sugarcane Industry Association. Quotes & Stats [internet]. 2012 [cited 2012 Jan 4]. Available from: http://english.unica.com.br/dadosCotacao/estatistica/. 25. Lemos W. Brazil's Flex-Fuel Car Production Rises, Boosting Ethanol Consumption to Record Highs [internet]. ICIS; 2007 Nov 7 [cited 2012 Jan 10]. Available from: http://www.icis.com/Articles/2007/11/12/9077311/brazils-flex-fuel-car-production-risesboosting-ethanol-consumption-to-record-highs.html. 26. Sjostrom E, Wood Polysaccharides, in Wood Chemistry: Fundamentals and Applications. Academic Press, San Diego, CA, USA (1993). 123  27. Sjostrom E, Extractives, in Wood Chemistry: Fundamentals and Applications. Academic Press, San Diego, CA, USA (1993). 28. National Renewable Energy Laboratory, Determination of Protein Content in Biomass. NREL/TP-510-42625 (2008). 29. Kelsall DR and Piggot R, Grain Milling and Cooking for Alcohol Production: Designing for the Options in Dry Milling, in The Alcohol Textbook, ed by Ingledew WM, Kelsall DR, Austin GD and Kluhspies C. Nottingham University Press, Nottingham, UK (2009). 30. McMurry J, Biomolecules: Carbohydrates, in Organic Chemistry, ed by Kiselica S, Williams P and White A. Brooks/Cole -Thomson Learning, Toronto, ON, Canada (2004). 31. Sjostrom E, The Structure of Wood, in Wood Chemistry: Fundamentals and Applications. Academic Press, San Diego, CA, USA (1993). 32. Scheller HV and Ulvskov P, Hemicelluloses. Annu Rev Plant Biol 61:263-289 (2010). 33. Technical Section: Canadian Pulp and Paper Association, Volume 4: Sulfite Science & Technology, in Pulp and Paper Manufacture, ed by Ingruber OV, Kocurek MJ and Wong A. McGraw-Hill, Montreal, QC, Canada (1985). 34. Sassner P. Lignocellulosic Ethanol Production Based on Steam Pretreatment and SSF [PhD thesis]. Lund, Sweden: The University of Lund; (2007). 35. Hahn-Hägerdal B, Wahlbom C, Gárdonyi M, van Zyl W, Otero R and Jönsson L, Metabolic engineering of Saccharomyces cerevisiae for xylose utilization. Adv Biochem Eng Biotechnol 73:53-84 (2001). 36. Fengel D and Wegener G, Lignin, in Wood: Chemistry, Ultrastructure, Reactions. W. de Gruyter, New York, NY, USA (1984). 37. Baucher M, Monties B, Van Montagu M and Boerjan W, Biosynthesis and genetic engineering of lignin. Crit Rev Plant Sci 17:125-197 (1998). 38. Chandra RP, Bura R, Mabee WE, Berlin A, Pan X and Saddler JN, Substrate pretreatment: The key to effective enzymatic hydrolysis of lignocellulosics? Adv Biochem Eng Biotechnol 108:67-93 (2007). 39. Söderström J, Pilcher L, Galbe M and Zacchi G, Two-step steam pretreatment of softwood with SO2 impregnation for ethanol production. Appl Biochem Biotechnol 98-100:521 (2002). 40. Pan X, Gilkes N, Kadla J, Pye K, Saka S, Gregg DJ et al., Bioconversion of hybrid poplar to ethanol and co-products using an organosolv fractionation process: Optimization of process yields. Biotechnol Bioeng 94:851-861 (2006). 124  41. Linde M, Jakobsson E, Galbe M and Zacchi G, Steam pretreatment of dilute H2SO4impregnated wheat straw and SSF with low yeast and enzyme loadings for bioethanol production. Biomass Bioenerg 32:326-332 (2008). 42. Galbe M, Sassner P, Wingren A and Zacchi G, Process engineering economics of bioethanol production. Adv Biochem Eng Biotechnol 108:303-327 (2007). 43. POET. Cellulosic Ethanol [internet]. 2011 [cited 2011 Jun 9]. Available from: http://www.poet.com/innovation/cellulosic/plan.asp. 44. Wu S, Chang H, Jameel H and Philips R, Novel green liquor pretreatment of loblolly pine chips to facilitate enzymatic hydrolysis into fermentable sugars for ethanol production. J Wood Chem Technol 30:205-218 (2010). 45. Moritz JW and Duff SJB, Ethanol production from spent sulfite liquor fortified by hydrolysis of pulp mill primary clarifier sludge. Appl Biochem Biotechnol 57-58:689-698 (1996). 46. Borregaard. Ethanol Products [internet]. 2012 [cited 2012 Sep 30]. Available at: http://www.borregaard.com/Business-Areas/Borregaard-ChemCell/Ethanol-products. 47. Helle SS, Murray A, Lam J, Cameron DR and Duff SJB, Xylose fermentation by genetically modified Saccharomyces cerevisiae 259ST in spent sulfite liquor. Bioresource Technol 92:163-171 (2004). 48. Kumar L, Chandra RP and Saddler JN, Influence of steam pretreatment severity on posttreatments used to enhance the enzymatic hydrolysis of pretreated softwoods at low enzyme loadings. Biotechnol Bioeng 108:2300-2311 (2011). 49. Nakagame S, Chandra RP and Saddler JN, The effect of isolated lignins, obtained from a range of pretreated lignocellulosic substrates, on enzymatic hydrolysis. Biotechnol Bioeng 105:871-879 (2010). 50. Del Rio LF, Chandra RP and Saddler JN, The effects of increasing swelling and anionic charges on the enzymatic hydrolysis of organosolv-pretreated softwoods at low enzyme loadings. Biotechnol Bioeng 108:1549-1558 (2011). 51. Barta Z, Kovacs K, Reczey K and Zacchi G, Process design and economics of on-site cellulase production on various carbon sources in a softwood-based ethanol plant. Enzyme Res Jan 2010:1-8 (2010). 52. Côté P, Noël G and Moore S, The Chatham demonstration: From design to operation of a 20 m3/d membrane-based ethanol dewatering system. Desalination 250:1060-1066 (2010). 53. Sassner P and Zacchi G, Integration options for high energy efficiency and improved economics in a wood-to-ethanol process. Biotechnol Biofuel 1:1-11 (2008). 125  54. Galbe M and Zacchi G, Pretreatment of lignocellulosic materials for efficient bioethanol production. Adv Biochem Eng Biotechnol 108:41-65 (2007). 55. Yang B and Wyman CE, Pretreatment: the key to unlocking low-cost cellulosic ethanol. Biofuel Bioprod Bior 2:26-40 (2008). 56. Breen A and Singleton FL, Fungi in lignocellulose breakdown and biopulping. Curr Opin Biotechnol 10:252-258 (1999). 57. Sun Y and Cheng J, Hydrolysis of lignocellulosic materials for ethanol production: a review. Bioresource Technol 83:1-11 (2002). 58. Guerra A, Mendonça R and Ferraz A, Molecular weight distribution of wood components extracted from Pinus taeda biotreated by Ceriporiopsis subvermispora. Enzyme Microb Technol 33:12-18 (2003). 59. Lee J and Jeffries TW, Efficiencies of acid catalysts in the hydrolysis of lignocellulosic biomass over a range of combined severity factors. Bioresource Technol 102:5884-5890 (2011). 60. Technical Section: Canadian Pulp and Paper Association, Volume 5: Alkaline Pulping, in Pulp and Paper Manufacture, ed by Grace TM, Leopold B and Malcolm EW. McGraw-Hill, Montreal, QC, Canada (1989). 61. Pan X, Xie D, Yu RW and Saddler JN, The bioconversion of mountain pine beetle-killed lodgepole pine to fuel ethanol using the organosolv process. Biotechnol Bioeng 101:39-48 (2008). 62. Office of Energy Efficiency and Renewable Energy, Biofuels, Biopower, and Bioproducts: Integrated Biorefineries. US Department of Energy (2010). 63. Boussaid A, Cai Y, Robinson J, Gregg DJ, Nguyen Q and Saddler JN, Sugar recovery and fermentability of hemicellulose hydrolysates from steam-exploded softwoods containing bark. Biotechnol Prog 17:887-892 (1998). 64. Clark TA and Mackie KL, Steam explosion of the softwood Pinus radiata with sulphur dioxide addition. I. Process optimization. J Wood Chem Technol 7:373-403 (1987). 65. Stenberg K, Tengborg C, Galbe M and Zacchi G, Optimisation of steam pretreatment of SO2-impregnated mixed softwoods for ethanol production. J Chem Technol Biotechnol 71:299-308 (1998). 66. Clark TA, Mackie KL, Dare PH and McDonald AG, Steam explosion of the softwood Pinus radiata with sulphur dioxide addition. II. Process characterisation. J Wood Chem Technol 9:135-166 (1989).  126  67. Donaldson LA, Wong KKY and Mackie KL, Ultrastructure of steam-exploded wood. Wood Sci Technol 22:103-114 (1988). 68. Brownell HH, Yu EKC and Saddler JN, Steam-explosion pretreatment of wood: Effect of chip size, acid, moisture content and pressure drop. Biotechnol Bioeng 28:792-801 (1986). 69. Mittal A, Chatterjee SG, Scott GM and and Amidon TE, Modeling xylan solubilization during autohydrolysis of sugar maple wood meal: reaction kinetics. Holzforschung 63:307314 (2009). 70. Brownell HH and Saddler JN, Steam pretreatment of lignocellulosic material for enhanced enzymatic hydrolysis. Biotechnol Bioeng 29:228-235 (1987). 71. Schwald W, Smaridge T, Chan M, Breuil C and Saddler JN, The Influence of SO2 Impregnation and Fractionation on Product Recovery and Enzymatic Hydrolysis of SteamPretreated Sprucewood, in Enzyme Systems for Lignocellulose Degradation, ed by Coughlan MP. Elsevier, New York, NY, USA (1989). 72. Arantes V and Saddler JN, Cellulose accessibility limits the effectiveness of minimum cellulase loading on the efficient hydrolysis of pretreated lignocellulosic substrates. Biotechnol Biofuel 4:1-17 (2011). 73. Söderström J, Pilcher L, Galbe M and Zacchi G, Combined use of H2SO4 and SO2 impregnation for steam pretreatment of spruce in ethanol production. Appl Biochem Biotechnol 105:127-140 (2003). 74. Tengborg C, Stenberg K, Galbe M, Zacchi G, Larsson S, Palmqvist E et al., Comparison of SO2 and H2SO4 impregnation of softwood prior to steam pretreatment on ethanol production. Appl Biochem Biotechnol 70-72:3-15 (1998). 75. Wayman M, Tallevi A and Winsborrow B, Hydrolysis of biomass by sulphur dioxide. Biomass 6:183-191 (1984). 76. Galbe M and Zacchi G, A review of the production of ethanol from softwood. Appl Microbiol Biotechnol 59:618-628 (2002). 77. Mosier N, Wyman CE, Dale B, Elander R, Lee YY, Holtzapple M et al., Features of promising technologies for pretreatment of lignocellulosic biomass. Bioresource Technol 96:673-686 (2005). 78. van Walsum G, Allen S, Spencer M, Laser M, Antal M and Lynd L, Conversion of lignocellulosics pretreated with liquid hot water to ethanol. Appl Biochem Biotechnol 5758:157-170 (1996).  127  79. Söderström J, Pilcher L, Galbe M and Zacchi G, Two-step steam pretreatment of softwood by dilute H2SO4 impregnation for ethanol production. Biomass Bioenerg 24:475486 (2003). 80. Palmqvist E and Hahn-Hägerdal B, Fermentation of lignocellulosic hydrolysates. II: Inhibitors and mechanisms of inhibition. Bioresource Technol 74:25-33 (2000). 81. Robinson J. Pretreatment and Fermentation of Douglas-fir Whitewood and Bark Feedstocks for Ethanol Production [PhD thesis]. Vancouver, BC: The University of British Columbia; (2003). 82. Larsson S, Palmqvist E, Hahn-Hägerdal B, Tengborg C, Stenberg K, Zacchi G et al., The generation of fermentation inhibitors during dilute acid hydrolysis of softwood. Enzyme Microb Technol 24:151-159 (1999). 83. Palmqvist E and Hahn-Hägerdal B, Fermentation of lignocellulosic hydrolysates. I: Inhibition and detoxification. Bioresource Technol 74:17-24 (2000). 84. Monavari S, Galbe M and Zacchi G, Impact of impregnation time and chip size on sugar yield in pretreatment of softwood for ethanol production. Bioresource Technol 100:63126316 (2009). 85. Ballesteros I, Oliva J, Navarro A, González A, Carrasco J and Ballesteros M, Effect of chip size on steam explosion pretreatment of softwood. Appl Biochem and Biotechnol 8486:97-110 (2000). 86. Cullis IF, Saddler JN and Mansfield SD, Effect of initial moisture content and chip size on the bioconversion efficiency of softwood lignocellulosics. Biotechnol Bioeng 85:413-421 (2004). 87. Hatton JV and Keays JL, Effect of chip geometry and moisture on yield and quality of kraft pulps from western hemlock and black spruce. Pulp Paper Mag Can 74:79-87 (1973). 88. Technical Section: Canadian Pulp and Paper Association, Volume 1: Properties of Fibrous Raw Materials and their Preparation for Pulping, in Pulp and Paper Manufacture, ed by Kocurek MJ and Stevens CFB. McGraw-Hill, Montreal, QC, Canada (1983). 89. Tengborg C, Galbe M and Zacchi G, Influence of enzyme loading and physical parameters on the enzymatic hydrolysis of steam-pretreated softwood. Biotechnol Prog 17:110-117 (2001). 90. Mansfield SD, Mooney C and Saddler JN, Substrate and enzyme characteristics that limit cellulose hydrolysis. Biotechnol Prog 15:804-816 (1999). 91. Várnai A, Siika-aho M and Viikari L, Restriction of the enzymatic hydrolysis of steampretreated spruce by lignin and hemicellulose. Enzyme Microb Technol 46:185-193 (2010). 128  92. Kumar L, Arantes V, Chandra RP and Saddler JN, The lignin present in steam pretreated softwood binds enzymes and limits cellulose accessibility. Bioresource Technol 103: 201208 (2012). 93. Nakagame S, Chandra RP, Kadla JF and Saddler JN, Enhancing the enzymatic hydrolysis of lignocellulosic biomass by increasing the carboxylic acid content of the associated lignin. Biotechnol Bioeng 108: 538-548 (2011). 94. Aehle W, Antrim RL, Becker T, Bott R, Buchert J, Burrows H et al., Industrial Enzymes, in Enzymes in Industry: Production and Applications, ed by Aehle W. WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim, Germany (2004). 95. Harris PV, Welner D, McFarland KC, Re E, Navarro Poulsen J, Brown K et al., Stimulation of lignocellulosic biomass hydrolysis by proteins of glycoside hydrolase family 61: Structure and function of a large, enigmatic family. Biochem-US 49:3305-3316 (2010). 96. Banerjee G, Car S, Scott-Craig JS, Borrusch MS and Walton JD, Rapid optimization of enzyme mixtures for deconstruction of diverse pretreatment/biomass feedstock combinations, Biotechnol Biofuel 3:22-37 (2010). 97. Din N, Gilkes NR, Tekant B, Miller RC Jr., Warren RAJ and Kilburn DG, Nonhydrolytic disruption of cellulose fibres by the binding domain of a bacterial cellulase. Nat Biotechnol 9:1096-1099 (1991). 98. Mittal A, Chatterjee SG, Scott GM and Amidon TE, Modeling xylan solubilization during autohydrolysis of sugar maple and aspen wood chips: reaction kinetics and mass transfer. Chem Eng Sci 64:3031-3041 (2009). 99. Vroom KE, The H factor: a means of expressing cooking times and temperatures as a single variable. Pulp Paper Mag Can 58:228-231 (1957). 100. Garrote G, Domínguez H and Parajó JC, Mild autohydrolysis: an environmentally friendly technology for xylooligosaccharide production from wood. J Chem Technol Biotechnol 74:1101-1109 (1999). 101. Garrote G, Domínguez H and Parajó JC, Kinetic modelling of corncob autohydrolysis. Process Biochem 36:571-578 (2001). 102. Garrote G, Domínguez H and Parajó JC, Interpretation of deacetylation and hemicellulose hydrolysis during hydrothermal treatments on the basis of the severity factor. Process Biochem 37:1067-1073 (2002). 103. Garrote G and Parajó JC, Non-isothermal autohydrolysis of Eucalyptus wood. Wood Sci Technol 36:111-123 (2002).  129  104. Yat SC, Berger A and Shonnard DR, Kinetic characterization for dilute sulfuric acid hydrolysis of timber varieties and switchgrass. Bioresource Technol 99:3855-3863 (2008). 105. Garrote G, Domínguez H and Parajó JC, Autohydrolysis of corncob: study of nonisothermal operation for xylooligosaccharide production. J Food Eng 52:211-218 (2002). 106. Jacobsen S and Wyman C, Cellulose and hemicellulose hydrolysis models for application to current and novel pretreatment processes. Appl Biochem Biotechnol 84-86:8196 (2000). 107. Wyman CE, Decker SR, Himmel ME, Brady JW, Skopec CE and Viikari L, Hydrolysis of Cellulose and Hemicellulose, in Polysaccharides: Structural Diversity and Functional Versatility, Dumitriu S. Marcel Dekker, New York, NY. (2004). 108. Jensen J, Morinelly J, Aglan A, Mix A and Shonnard DR, Kinetic characterization of biomass dilute sulfuric acid hydrolysis: Mixtures of hardwoods, softwood, and switchgrass. AICHE J 54:1637-1645 (2008). 109. Jin Q, Zhang H, Yan L, Qu L and Huang H, Kinetic characterization for hemicellulose hydrolysis of corn stover in a dilute acid cycle spray flow-through reactor at moderate conditions. Biomass Bioenerg 35:4158-4164 (2011). 110. Brennan M and Wyman C, Initial evaluation of simple mass transfer models to describe hemicellulose hydrolysis in corn stover. Appl Biochem Biotechnol 115:965-976 (2004). 111. Saeman JF, Kinetics of wood saccharification: Hydrolysis of cellulose and decomposition of sugars in dilute acid at high temperature. Ind Eng Chem Res 37:42 (1945). 112. Brasch DJ and Free KW, Prehydrolysis - kraft pulping of Pinus radiata grown in New Zealand. TAPPI 48:245-248 (1965). 113. Overend RP, Chornet E and Gascoigne JA, Fractionation of lignocellulosics by steamaqueous pretreatments. Proc R Soc A 321:523-536 (1987). 114. Biermann CJ, Chapter 3: Pulping Fundamentals, in Handbook of pulping and papermaking, ed by Biermann CJ. Academic Press, San Diego, CA, USA (1996). 115. Mabee WE, Gregg DJ, Claudio A, Berlin A, Bura R, Gilkes N et al., Updates on softwood-to ethanol process development. Appl Biochem Biotechnol 129:55-70 (2006). 116. Bura R, Bothast R, Mansfield SD and Saddler JN, Optimization of SO2-catalyzed steam pretreatment of corn fiber for ethanol production. Appl Biochem Biotechnol 106:319-335 (2003). 117. Heitz M, Capek-Ménard E, Koeberle PG, Gagné J, Chornet E, Overend RP et al., Fractionation of Populus tremuloides at the pilot plant scale: Optimization of steam 130  pretreatment conditions using the STAKE II technology. Bioresource Technol 35:23-32 (1991). 118. Kabel MA, Bos G, Zeevalking J, Voragen AGJ and Schols HA, Effect of pretreatment severity on xylan solubility and enzymatic breakdown of the remaining cellulose from wheat straw. Bioresource Technol 98:2034-2042 (2007). 119. García-Aparicio M, Parawira W, Van Rensburg E, Diedericks D, Galbe M, Rosslander C et al., Evaluation of steam-treated giant bamboo for production of fermentable sugars. Biotechnol Prog 27:641-649 (2011). 120. Scordia D, Cosentino SL, Lee J and Jeffries TW, Dilute oxalic acid pretreatment for biorefining giant reed (Arundo donax L.). Biomass Bioenerg 35:3018-3024 (2011). 121. Chum HL, Johnson DK, Black SK and Overend RP, Pretreatment-catalyst effects and the combined severity parameter. Appl Biochem Biotechnol 24-25:1-14 (1990). 122. Abatzoglou N, Chornet E, Belkacemi K and Overend RP, Phenomenological kinetics of complex systems: The development of a generalized severity parameter and its application to lignocellulosics fractionation. Chem Eng Sci 47:1109-1122 (1992). 123. Myers RH and Montgomery DC, Introduction, in Response Surface Methodology: Process and Product Optimization Using Designed Experiments, ed by Bloomfield P, Cressie NAC, Fisher NI, Johnstone IM, Kadane JB, Ryan LM et al. John Wiley & Sons Inc., New York, NY, USA (2002). 124. Box GEP, Hunter JS and Hunter WG, Some Applications of Response Surface Methods, in Statistics for Experimenters: Design, Innovation, and Discovery, ed by Balding DJ, Bloomfield P, Cressie NAC, Fisher NI, Johnstone IM, Kadane JB et al. WileyInterscience, Hoboken, NJ, USA (2005). 125. Myers RH and Montgomery DC, Experimental Designs for Fitting Response Surfaces I, in Response Surface Methodology: Process and Product Optimization Using Designed Experiments, ed by Bloomfield P, Cressie NAC, Fisher NI, Johnstone IM, Kadane JB, Ryan LM et al. John Wiley & Sons Inc., New York, NY, USA (2002). 126. Roquemore KG, Hybrid designs for quadratic response surfaces. Technometrics 18:419423 (1976). 127. Kim H, Lee J, Jeffries TW and Choi I, Response surface optimization of oxalic acid pretreatment of yellow poplar (Liriodendron tulipifera) for production of glucose and xylose monosaccarides. Bioresource Technol 102:1440-1446 (2011). 128. Lee J, Rodrigues RCLB and Jeffries TW, Simultaneous saccharification and ethanol fermentation of oxalic acid pretreated corncob assessed with response surface methodology. Bioresource Technol 100:6307-6311 (2009). 131  129. Barta Z, Reczey K and Zacchi G, Techno-economic evaluation of stillage treatment with anaerobic digestion in a softwood-to-ethanol process. Biotechnol Biofuel 3:1-11 (2010). 130. Piccolo C and Bezzo F, A techno-economic comparison between two technologies for bioethanol production from lignocellulose. Biomass Bioenerg 33:478-491 (2009). 131. Wingren A, Söderstroem J, Galbe M and Zacchi G, Process considerations and economic evaluation of two-step steam pretreatment for production of fuel ethanol from softwood. Biotechnol Prog 20:1421-1429 (2004). 132. National Renewable Energy Laboratory and Harris Group Inc., Process Design and Economics for Biochemical Conversion of Lignocellulosic Biomass to Ethanol: Dilute-Acid Pretreatment and Enzymatic Hydrolysis of Corn Stover. NREL/TP-5100-47764 (2011). 133. National Renewable Energy Laboratory and Harris Group Inc., Lignocellulosic Biomass to Ethanol Process Design and Economics Utilizing Co-Current Dilute Acid Prehydrolysis and Enzymatic Hydrolysis for Corn Stover. NREL/TP-510-32438 (2002). 134. National Renewable Energy Laboratory and Delta-T Corporation, Lignocellulosic Biomass to Ethanol Process Design and Economics Utilizing Co-Current Dilute Acid Prehydrolysis and Enzymatic Hydrolysis Current and Futuristic Scenarios. NREL/TP-58026157 (1999). 135. Wingren A, Galbe M and Zacchi G, Energy considerations for a SSF-based softwood ethanol plant. Bioresource Technol 99:2121-2131 (2008). 136. Wingren A, Galbe M and Zacchi G, Techno-economic evaluation of producing ethanol from softwood: Comparison of SSF and SHF and identification of bottlenecks. Biotechnol Prog 19:1109-1117 (2003). 137. von Sivers M and Zacchi G, A Techno-economical comparison of three processes for the production of ethanol from pine. Bioresource Technol 51:43-52 (1995). 138. Dutta A, Dowe N, Ibsen KN, Schell DJ and Aden A, An economic comparison of different fermentation configurations to convert corn stover to ethanol using Z. mobilis and Saccharomyces. Biotechnol Prog 26:64-72 (2010). 139. Kumar D and Murthy GS, Impact of pretreatment and downstream processing technologies on economics and energy in cellulosic ethanol production. Biotechnol Biofuel 4:27-45 (2011). 140. Eggeman T and Elander RT, Process and economic analysis of pretreatment technologies. Bioresource Technol 96:2019-2025 (2005). 141. Starcher B, A ninhydrin-based assay to quantitate the total protein content of tissue samples. Anal Biochem 292:125-129 (2001). 132  142. TAPPI, TAPPI T 222 om-06. Acid Insoluble Lignin in Wood and Pulp, in TAPPI T 222. TAPPI Press, Norcross, GA, USA (2006). 143. Lin SY, The Determination of Lignin, in Methods in Lignin Chemistry, ed by Lin SY and Dence CW. Springer-Verlag, Berlin, Germany (1992). 144. TAPPI, TAPPI T 204 cm-07. Solvent Extractives of Wood and Pulp, in TAPPI T 204. TAPPI Press, Norcross, GA, USA (2007). 145. TAPPI, TAPPI T 211 om-07. Ash in Wood, Pulp, Paper and Paperboard: Combustion at 525 °C, in TAPPI T 211. TAPPI Press, Norcross, GA, USA (2007). 146. National Renewable Energy Laboratory, Determination of Sugars, Byproducts, and Degradation Products in Liquid Fraction Process Samples. NREL/TP-510-42623 (2008). 147. Wei L, Shrestha A, Tu M and Adhikari S, Effects of surfactant on biochemical and hydrothermal conversion of softwood hemicellulose to ethanol and furan derivatives. Process Biochem 46:1785-1792 (2011). 148. Xie R, Tu M, Wu Y and Adhikari S, Improvement in HPLC separation of acetic acid and levulinic acid in the profiling of biomass hydrolysate. Biouresour Technol 102:49384942 (2011). 149. Kristensen J, Felby C and Jørgensen H, Determining yields in high solids enzymatic hydrolysis of biomass. Appl Biochem Biotechnol 156:127-132 (2009). 150. Mamers H and Menz DNJ, Explosion pretreatment of Pinus radiata wood chips for the production of fermentation substrates. APPITA 37:644-649 (1984). 151. Khuri AI and Cornell JA, Second-Order Models and Designs, in Response Surfaces: Designs and Analyses, ed by Owen DB, Schucany WR, Chow SC, Kennedy WJ, Kshirsagar AM and Schilling EG. Marcel Dekker Inc., New York, NY, USA (1996). 152. Box GEP and Draper NR, Blocking and Fractionating 2k Factorial Designs, in Empirical Model-Building and Response Surfaces. John Wiley & Sons Inc., New York, NY, USA (1987). 153. Box GEP, Hunter JS and Hunter WG, Modeling, Geometry, and Experimental Design, in Statistics for Experimenters: Design, Innovation, and Discovery, ed by Balding DJ, Bloomfield P, Cressie NAC, Fisher NI, Johnstone IM, Kadane JB et al. Wiley-Interscience, Hoboken, NJ, USA (2005). 154. Sassner P, Galbe M and Zacchi G, Steam pretreatment of Salix with and without SO2 impregnation for production of bioethanol. Appl Biochem Biotechnol 121-124:1101-1118 (2005).  133  155. Kumar L, Chandra RP, Chung PA and Saddler JN, Can the same steam pretreatment conditions be used for most softwoods to achieve good, enzymatic hydrolysis and sugar yields? Bioresource Technol 101:7827-7833 (2010). 156. Söderström J, Galbe M and Zacchi G, Effect of washing on yield in one- and two-Step steam pretreatment of softwood for production of ethanol. Biotechnol Prog 20:744-749 (2004).  134  Appendix A: Estimates of model term significance from extended ANOVA Table A-1. Estimate of model term significance (p) for the RSM model of steam pretreated radiata pine: yield of the water insoluble fraction and recovery of residual components. Factor Yield a Glucan Mannan Xylan b vs vs Intercept 0.0004 0.0002 0.0003 vs T 0.0007 0.0003 vs vs t 0.0051 0.0091 0.0101 0.0058 S 0.0022 0.0106 0.0019 Txt TxS 0.0396 0.0122 0.0213 txS 2 0.0222 0.0030 0.0004 T 2 0.0410 0.0002 vs t 2 0.0409 S 0.1286 0.0661 N/A d N/A d Lof c a Equation created in units of w/w % of raw wood. b Very significant (p < 0.0001). c Lack-of-fit. d Lack-of-fit could not be estimated for some responses because the standard deviation was zero. The p values for these responses were generated without isolating the pure error during ANOVA.  135  Table A-2. Estimate of model term significance (p) for the RSM model pretreated radiata pine: sugar recoveries in the water soluble fraction. Factor Total a Hemicellulose b Glucose Mannose Xylose Galactose vs c vs vs vs vs vs Intercept 0.0206 vs 0.0002 0.0007 0.0017 0.0051 T 0.0001 0.0005 0.0011 0.0023 0.0049 t 0.0002 0.0017 0.0214 S vs 0.0002 0.0336 0.0016 0.0025 0.0043 Txt TxS 0.0003 0.0037 txS 2 0.0001 0.0008 0.0060 0.0113 0.0113 0.0030 T vs 0.0003 0.0082 0.0038 0.0046 0.0035 t2 0.0016 0.0079 0.0385 S2 0.0039 0.0199 0.0074 0.1409 0.2116 0.3460 Lof d a Recovery of all soluble sugars as both monomers and oligomers (glucose, xylose, galactose and arabinose). b Sugars derived from hemicellulose are xylose, galactose and arabinose. c Very significant (p < 0.0001). d Lack-of-fit.  of steam Arabinose vs 0.0011 0.0013 0.0041  0.0715  mannose, mannose,  136  Table A-3. Estimate of model term significance (p) for the RSM model of steam pretreated white fir and ponderosa pine: yield of the water insoluble fraction and sugar recoveries in the water soluble fraction. Factor Yield a Hemicellulose b Glucose c vs 0.0002 0.0004 Intercept 0.0038 0.0025 0.0018 T 0.0303 0.0161 0.0114 t 0.0270 0.0043 0.0064 S 0.0111 Txt 0.0363 TxS 0.0794 txS 0.0074 T2 0.0132 0.0257 t2 2 0.0270 0.0379 S 0.3250 0.1296 0.2254 Lof d a b Equation created in units of w/w % of raw wood. Sugars derived from hemicellulose are mannose, xylose, galactose and arabinose. c Very significant (p < 0.0001). d Lack-offit.  137  Table A-4. Estimate of model term significance (p) for the RSM model of steam pretreated Norway spruce: yield of the water insoluble fraction and sugar recoveries in the water soluble fraction. Factor Yield a Hemicellulose b Glucose Intercept T t S Txt TxS txS T2 t2 S2 N/A d N/A d N/A d Lof c No equations could be created for these responses. a Equation created in units of w/w % of raw wood. b Sugars derived from hemicellulose are mannose, xylose, galactose and arabinose. c Lack-of-fit. d Lack-of-fit was not estimated for these responses because no models could be created.  138  Table A-5. Estimate of model term significance (p) for the RSM model of steam pretreated lodgepole pine: yield of the water insoluble fraction and recovery of residual components. Factor Yield a Glucan Mannan Xylan Ash a vs b vs 0.0002 0.0003 vs Intercept vs vs 0.0008 0.0004 0.0069 T 0.0010 0.0008 0.0005 0.0002 0.0619 c t 0.0036 0.0078 0.0356 0.0218 0.0048 S 0.0386 Size 0.0427 0.0426 MC c 0.0904 0.0290 Txt 0.1000 c 0.0478 TxS T x Size 0.0738 c 0.0399 T x MC txS 0.0280 t x Size 0.0192 t x MC 0.0851 c S x Size 0.0127 S x MC c 0.0559 Size x MC 2 T 0.0028 0.0272 0.0041 0.0007 t2 c 2 0.0982 S 2 0.0962 c 0.0148 Size 2 0.0005 0.0036 0.0162 0.0128 MC d 0.1767 0.5252 0.2607 0.1165 0.2599 Lof No equations could be created for galactan, arabinan, or enzymatic digestibility. a Equation created in units of w/w % of raw wood. b Very significant (< 0.0001). c Marginally significant (0.05 ≤ p ≥ 0.10). d Lack-of-fit.  139  Table A-6. Estimate of model term significance (p) for the RSM model of steam pretreated lodgepole pine: sugar recoveries in the water soluble fraction. Factor Hemicellulose a Glucose Mannose Xylose Galactose Arabinose b vs vs vs vs vs vs Intercept vs 0.0032 0.0001 vs vs 0.0002 T 0.0008 0.0015 0.0046 0.0004 0.0002 0.0025 t 0.0017 0.0066 0.0010 0.0003 0.0098 S c 0.1000 0.0352 0.0905 c Size 0.0451 0.0014 MC 0.0033 0.0192 0.0037 0.0063 0.0022 0.0284 Txt 0.0284 TxS T x Size 0.0429 0.0491 T x MC c 0.1000 txS t x Size 0.0765 c 0.0864 c 0.0750 c t x MC 0.0327 0.0287 0.0531 c 0.0635 c S x Size 0.0905 c 0.0208 S x MC 0.0271 0.0398 0.0115 0.0444 Size x MC c 2 0.0447 0.0708 T c 2 0.0205 0.0300 0.0679 0.0020 t 2 0.0342 S c 2 0.0466 0.0538 Size 2 0.0030 0.0079 0.0072 0.0012 0.0019 MC d 0.1805 0.3220 0.2368 0.1107 0.3456 0.2504 Lof No equation could be created for the recovery of total sugars as both monomers and oligomers (glucose, mannose, xylose, galactose and arabinose). a Sugars derived from hemicellulose are mannose, xylose, galactose and arabinose. b Very significant (p < 0.0001). c Marginally significant (0.05 ≤ p ≥ 0.10). d Lack-of-fit.  140  Table A-7. Estimate of model term significance (p) for the RSM model of steam pretreated lodgepole pine: levels of inhibitory compounds in the water soluble fraction. Factor Formic Acid a Levulinic Acid a HMF a vs b vs vs Intercept 0.0004 0.0011 vs T 0.0007 0.0013 vs t c 0.0678 S 0.0971 c 0.0448 Size c 0.0570 0.0264 MC 0.0261 Txt c 0.0976 0.0497 TxS 0.0834 c 0.0426 T x Size T x MC 0.0041 txS t x Size 0.0434 t x MC 0.0342 S x Size c 0.0721 0.0111 S x MC Size x MC 0.0119 T2 2 0.0321 0.0032 t 2 S 0.0654 c Size2 0.0578 c 0.0620 c 0.0209 MC2 d 0.4897 0.5245 0.1295 Lof a No equation could be created for acetic acid or furfural. Equation created in units of w/w % of raw wood. b Very significant (p < 0.0001). c Marginally significant (0.05 ≤ p ≥ 0.10). d Lack-of-fit.  141  Table A-8. Estimate of model term significance (p) for RSM model of steam pretreated and enzymatically hydrolyzed lodgepole pine: sugar recoveries in the water soluble fraction and enzymatic hydrolyzate. Combined EH EH glucose Factor a b recovery recovery recovery vs c vs vs Intercept vs vs vs T 0.0063 0.0006 0.0006 t 0.0014 0.0030 0.0030 S Size 0.0116 0.0183 0.0178 MC vs 0.0011 0.0010 Txt 0.0021 0.0014 0.0014 TxS d 0.0941 T x Size 0.0929 d T x MC 0.0133 txS t x Size 0.0785 d 0.0765 d t x MC S x Size S x MC 0.0826 d 0.0805 d Size x MC 0.0080 0.0242 0.0236 T2 2 t S2 Size2 0.0021 0.0004 0.0004 MC2 0.0831 0.1776 0.1718 Lof e a Recovery of all soluble sugars (glucose, mannose, xylose, galactose and arabinose) after both steam pretreatment and enzymatic hydrolysis. b Recovery of all soluble sugars (glucose, mannose, xylose, galactose and arabinose) in the enzymatic hydrolyzate. c Very significant (p < 0.0001). d Marginally significant (0.05 ≤ p ≥ 0.10). e Lack-of-fit.  142  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.24.1-0073503/manifest

Comment

Related Items