UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

The development of a techno-economic model to assess the effect of various process options on a wood-to-ethanol… Gregg, David John 1996

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

Item Metadata

Download

Media
831-ubc_1996-0367.pdf [ 13.5MB ]
Metadata
JSON: 831-1.0087194.json
JSON-LD: 831-1.0087194-ld.json
RDF/XML (Pretty): 831-1.0087194-rdf.xml
RDF/JSON: 831-1.0087194-rdf.json
Turtle: 831-1.0087194-turtle.txt
N-Triples: 831-1.0087194-rdf-ntriples.txt
Original Record: 831-1.0087194-source.json
Full Text
831-1.0087194-fulltext.txt
Citation
831-1.0087194.ris

Full Text

THE DEVELOPMENT OF A TECHNO-ECONOMIC MODEL TO ASSESS THE EFFECT OF VARIOUS PROCESS OPTIONS ON A WOOD-TO-ETHANOL PROCESS by David John Gregg B.Sc, University of Calgary, 1978 B.A.Sc, University of British Columbia, 1993 ' A thesis submitted in partial fulfillment of The requirements for the degree of Master of Applied Science in the Faculty Of Graduate Studies Department of Wood Science We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA July, 1996 ©David John Gregg, 1996 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department of Wao£> ^c i £ N J C £ . The University of British Columbia Vancouver, Canada Date C ^ j y . //9<?fr> DE-6 (2/88) A B S T R A C T In the last twenty-five years there has been considerable interest in the potential for producing ethanol from biomass, particularly in processes involving enzymatic hydrolysis. Unfortunately these processes have not been proven in a fully-integrated system either technically or economically. Techno-economic models have been used in the past to provide assessments of both the process and subprocess maturity and the production cost of the product. Most of this work modelled hardwoods as the substrate and did not consider variable feedstock options. Past hardwood modelling results have shown that the front-end process steps (i.e. Pretreatment, Fractionation and Enzyme Hydrolysis) are both technologically immature and represent a large component (-60%) of the total product cost. In the initial parts of this work the economics and technical maturity of an enzyme based, hardwood-to-ethanol process were reassessed by evaluating the various technical and modelling developments that have been achieved over the past five years. A review of both the technical and past modelling efforts suggested that, due to the immaturity of the overall process and the large number of potential process and subprocess options that have to be considered, the techno-economic model should possess a high degree of flexibility. After evaluating the possibility of developing a new model which would incorporate these characteristics it was decided that two previous models could be combined and updated to provide the necessary flexibility. Consequently, the technical and economic knowledge of the past models was incorporated into a new STeam Explosion Assessment Model (STEAM) that was designed and built to include concepts such as encapsulation, modulation, object-oriented programming and a graphical interface. These characteristics were borrowed from current software development and flowsheet simulators to provide the necessary flexibility and ease of modification. The previously unassessed technical developments (e.g. S0 2 catalysis for steam pretreatment, supplemental lignin recovery via an additional peroxide wash and enzyme recycling, etc.) were also added to this new model. The influence of either a hardwood or softwood feedstock on the overall process and each of the process steps was also assessed. ii The model was first used to determine the maturity or level of definition of each of the process steps and the subsequent technical and economic impact of the changes on the individual steps and the overall process. Each of the process steps, with the exception of enzyme recycle, showed the potential to substantially reduce the production cost of ethanol. For example, technical benefits accruing from sulphur dioxide usage were substantial and easily overcame the implementation costs when hardwoods were used as the substrate. This technology provided an approximately $0.78/L enhancement over the non-S02 water-insoluble (WI) substrate and $0.98/L over the non-S02 water and alkali insoluble (WIA) substrate in hardwoods. Similarly the use of an acid catalyst during steam pretreatment allowed enzyme purchase costs to be reduced from 30% to 8% of the total production cost. However, the corrosive nature of S02, particularly at the higher levels anticipated with the softwoods, suggested a requirement for the use of exotic metals in the construction of the equipment. The extra capital to purchase and install zirconium vessels and pump elements indicated that, for both feedstocks, there was a lessening of the return resulting from the implementation of the S0 2 catalysis. This amounted to S0.07/L for hardwoods and $0.12/L reduction for softwoods. Other impacts identified by the model showed that supplementary lignin recovery, in the form of a peroxide wash following the water and alkali washes, provided major gains ($1.32/L and 105L/ODT) for the softwood feedstock and only minor gains ($0.06/L and 33L/ODT) for the hardwood. In a similar fashion the model first indicated that enzyme recycling could greatly reduce the proportion of total production cost attributed to enzymes (80% in hardwoods and 50% in softwoods). However, the net reduction in the ethanol production cost was only attained when the estimated implementation costs were waived. A flexible techno-economic model has been developed that can effectively model a "generic" hardwood/softwood-to-ethanol process. The model has indicated the difference and similarities of each of the process steps when different wood substrates are used as feedstocks. The model has also indicated which changes in the various process steps can have the most impact on the final cost of producing ethanol from wood substrates. In this way it provides a useful tool in directing where research and development efforts should be focussed to reduce the cost of producing ethanol from biomass feedstocks. iv T A B L E O F C O N T E N T S ABSTRACT i i TABLE OF CONTENTS v LIST OF TABLES vii LIST OF FIGURES viii LIST OF ABBREVIATIONS x ACKNOWLEDGEMENTS xii 1. INTRODUCTION 1 1.1 BACKGROUND ......... 5 1.1.1 Biomass-to-Ethanol Processes 5 1.1.2 Process Simulation/Modelling 9 1.2 GOALS OF THESIS : 10 1.2.1 Development/Refining of Hardwood-to-Ethanol Model 10 1.2.2 Softwood Modelling 12 1.2.3 Model Format Conversion 13 2. PROCESS DEVELOPMENT 15 2.1 PROCESS DEFINITION - IDENTIFICATION OF SUB PROCESSES.... 15 2.2 SUBPROCESS DEFINITION - IDENTIFICATION OF KEY EQUIPMENT 17 2.2.1 Wood Preparation 17 2.2.2 Pretreatment 17 2.2.3 Fractionation 19 2.2.4 Hydrolysis 22 2.2.5 Fermentation 25 2.2.6 Product Recovery 29 2.2.7 Waste Treatment 29 3. ASSESSMENT OF PAST MODELLING OF BIOMASS-TO-ETHANOL PROCESSES 31 3.1 BACKGROUND 31 3.1.1 Scaling 31 3.1.2 Process Integration 32 3.1.3 Computer programs 33 3.2 HISTORICAL BIOMASS-TO-ETHANOL MODELLING 38 4. MODEL DEVELOPMENT 4 4 v 4.1 ASSESSMENT AND COMPARISON OF INHERITED MODELS 44 4.1.1 General Technical Overview 44 4.1.2 Model Structure 46 4.1.3 Model Calculations 50 4.1.4 Model User Interface 57 4.1.5 Conclusions 59 4.2 MODIFICATION OF INHERITED MODELS 60 4.2.1 Model Structure 60 4.2.2 Model User Interface 63 5. MODELLING RESULTS 70 5.1 PROCESS IMPLEMENTATION COST EFFECT 74 5.2 TECHNICAL METHODS OF OFFSETTING IMPLEMENTATION COSTS 78 5.2.1 Sulphur Dioxide Catalysis 79 5.2.2 Supplementary Lignin Extraction - Peroxide Wash 83 5.2.3 Enzyme Recycle 83 5.2.4 Softwood Doubling Hydrolysis Time 84 5.3 CONCLUSIONS AND PROJECTED FUTURE PROSPECTS FOR LIGNOCELLULOSIC-TO-ETHANOL PROCESSES 85 6. CONCLUSIONS 8 8 6.1 LEVEL OF PROCESS MATURITY 88 6.2 TECHNICAL DETAILS OF PROCESS FRONT-END 89 6.3 DEFINITION AND CONSTRUCTION OF TECHNO-ECONOMIC MODEL 91 6.4 IMPACT OF RECENT RESEARCH ON PROCESS MATURITY 92 7. REFERENCES 95 8. APPENDICES - DETAILED DESCRIPTION OF THE ASSUMPTIONS INCORPORATED INTO THE PRETREATMENT/FRACTIONATION AND HYDROLYSIS/RECYCLE COMPONENTS OF THE STEAM MODEL 108 8.1 PRETREATMENT & FRACTIONATION 108 8.1.1 Hemicellulose Recovery 113 8.1.2 Lignin Recovery 124 8.1.3 Cellulose Recovery 129 8.2 ENZYME HYDROLYSIS & RECYCLE 131 8.2.1 Enzymatic Hydrolysis 133 8.2.2 Enzyme Recycle 142 vi LIST OF TABLES T A B L E 1 C O M P A R I S O N S U M M A R Y OF T H E V P I A N D F O R I N T E K M O D E L S 45 T A B L E 2 C O M P A R I S O N O F F O R I N T E K A N D F P B M O D E L PROCESS P A R A M E T E R S FOR H A R D W O O D S 51 T A B L E 3 C O M P A R I S O N OF F O R I N T E K A N D F P B M O D E L E C O N O M I C P A R A M E T E R S 53 T A B L E 4 F O R I N T E K M O D E L - M O N T E C A R L O S I M U L A T I O N P A R A M E T E R S 5 7 T A B L E 5 T H E O R E T I C A L P R O D U C T Y I E L D S A N D R E T U R N S FOR A H A R D W O O D - A N D S O F T W O O D - T O - F U E L E T H A N O L PROCESS 71 T A B L E 6 B A S E C A S E M O D E L A S S U M P T I O N S 75 T A B L E 7 P R E T R E A T M E N T CONDITIONS FOR V A R I O U S F E E D S T O C K S 109 vii LIST OF FIGURES FIGURE 1 BIOMASS-TO-ETHANOL PROCESS SCENARIOS BASED ON FEEDSTOCK T Y P E (STARCH-, SUGAR- AND LIGNOCELLULOSIC-BASED) 2 FIGURE 2 A N EXAMPLE OF A GENERIC WOOD-TO-ETHANOL PROCESS 16 FIGURE 3 FRACTIONATION DETAILS 20 FIGURE 4 CELLULOSE HYDROLYSIS DETAILS 24 FIGURE 5 FERMENTATION DETAILS 27 FIGURE 6 FORINTEK M O D E L STRUCTURE 47 FIGURE 7 VIRGINIA POLYTECHNICAL INSTITUTE M O D E L STRUCTURE 49 FIGURE 8 FOREST PRODUCTS BIOTECHNOLOGY S T E A M M O D E L STRUCTURE 61 FIGURE 9 SPREADSHEET INPUT-OUTPUT INTERFACE OF FORINTEK M O D E L 64 FIGURE 10 A N EXAMPLE OF AN INPUT/OUTPUT DIALOG BOX FROM THE S T E A M M O D E L 65 FIGURE 11 S T E A M M O D E L GENERIC WOOD-TO-ETHANOL PROCESS FLOWSHEET 66 FIGURE 12 S T E A M M O D E L PRETREATMENT SUBPROCESS FLOWSHEET 67 FIGURE 13 S T E A M M O D E L EQUIPMENT OPTION AND CALCULATIONAL ROUTINE ACCESS 68 FIGURE 14 A N EXAMPLE OF A S T E A M M O D E L CALCULATIONAL ROUTINE 69 FIGURE 15 THEORETICAL B R E A K E V E N ANALYSIS M A X I M U M PROCESS IMPLEMENTATION COST VS FEEDSTOCK PRICE 72 FIGURE 16 RELATIVE COST OF IMPLEMENTING VARIOUS FRONT-END TECHNOLOGIES FOR A HARDWOOD-TO-FUEL ETHANOL PROCESS 76 FIGURE 17 RELATIVE COST OF IMPLEMENTING VARIOUS FRONT-END TECHNOLOGIES FOR A SOFTWOOD-TO-FUEL ETHANOL PROCESS 77 FIGURE 18 RELATIVE EFFECT OF IMPLEMENTING VARIOUS FRONT-END TECHNOLOGIES FOR A HARDWOOD-TO-FUEL ETHANOL PROCESS 81 FIGURE 19 RELATIVE EFFECT OF IMPLEMENTING VARIOUS FRONT-END TECHNOLOGIES FOR A SOFTWOOD-TO-FUEL ETHANOL PROCESS 82 FIGURE 20 THEORETICAL & MODELLING RESULTS BREAKEVEN ANALYSIS THEORETICAL M A X I M U M PROCESS IMPLEMENTATION COST VS FEEDSTOCK PRICE 86 FIGURE 21 COMPOSITION OF REPRESENTATIVE LIGNOCELLULOSIC SUBSTRATES: SOFTWOOD (WHITE SPRUCE), HARDWOOD (ASPEN), AND AGRICULTURAL RESIDUE (WHEAT STRAW) 112 FIGURE 22 RECOVERY YIELD FOR STEAM PRETREATMENT OF SO2 -IMPREGNATED HARDWOOD 115 FIGURE 23 WATER-SOLUBLE FRACTION YIELD FROM STEAM-PRETREATED S02-IMPREGNATED HARDWOOD 117 FIGURE 24 SULPHUR B A L A N C E FOR TYPICAL SO2 IMPREGNATED HARDWOOD SAMPLE 120 FIGURE 25 RECOVERY YIELD FOR STEAM PRETREATMENT OF S02-IMPREGNATED SOFTWOOD 122 FIGURE 26 WATER-SOLUBLE FRACTION YIELD FROM STEAM-PRETREATED S02-IMPREGNATED SOFTWOOD 123 FIGURE 27 WATER-INSOLUBLE FRACTION YIELD FROM STEAM-PRETREATED S02-IMPREGNATED SOFTWOOD ... 125 0 FIGURE 28 OPTIMIZED CONDITIONS FOR M A X I M U M SUGAR AND LIGNIN YIELDS FROM A REPRESENTATIVE HARDWOOD 126 viii FIGURE 29 OPTIMIZED CONDITIONS FOR M A X I M U M SUGAR AND LIGNIN YIELDS FROM A REPRESENTATIVE SOFTWOOD 127 FIGURE 30 GLUCOSE YIELDS FOR STEAM-PRETREATED SO2 -IMPREGNATED ASPEN AND SPRUCE FOLLOWING EACH OF A WATER W A S H , A L K A L I W A S H AND PEROXIDE W A S H 128 FIGURE 31 HYDROLYSIS OF PEROXIDE-TREATED EUCALYPTUS AT A SUBSTRATE CONCENTRATION OF 6% (w/v) USING VARIOUS ENZYME LOADINGS 135 FIGURE 32 EFFECT OF SUGAR REMOVAL ON HYDROLYSIS OF PEROXIDE-TREATED EUCALYPTUS AT A SUBSTRATE CONCENTRATION OF 6% (w/v) AND ENZYME LOADING OF 10 F P U G" 1 OF CELLULOSE 137 FIGURE 33 HYDROLYSIS OF PEROXIDE-TREATED ASPEN AT AN ENZYME LOADING OF 10 F P U G " 1 OF CELLULOSE AND AT VARIOUS SUBSTRATE CONCENTRATIONS 139 FIGURE 34 CELLULOSE HYDROLYSIS PROFILES AND 24 HOUR GLUCOSE YIELDS FOR PEROXIDE-TREATED ASPEN, EUCALYPTUS AND SPRUCE SUBSTRATES AT A SUBSTRATE CONCENTRATION OF 2% (W/V) AND AN ENZYME LOADING OF 10 F P U G _ 1 OF CELLULOSE 143 FIGURE 35 SCHEMATIC REPRESENTATION OF CELLULASE AND B-GLUCOSIDASE LOCATION DURING THE INITIAL , FINAL AND COMPLETE STAGES OF HYDROLYSIS 144 FIGURE 36 PERCENT CONVERSION AND DESORBED CELLULASE PROTEIN IN THE SUPERNATANT AT DIFFERENT TIMES DURING HYDROLYSIS OF LOW-LIGNIN CONTENT PEROXIDE-TREATED BIRCH AND HIGH-LIGNIN CONTENT WATER-WASHED BIRCH 146 FIGURE 37 FLOWSHEET FOR THE FRONT-END OF A HARDWOOD- AND SOFTWOOD-TO-ETHANOL PROCESS THAT HAS BEEN OPTIMIZED FOR M A X I M U M RECOVERY OF THE HEMICELLULOSE, LIGNIN AND CELLULOSE COMPONENTS 150 ix LIST OF ABBREVIATIONS BC British Columbia B D T Bone-dry tonne B O D Biological Oxygen Demand C5 Pentose sugars C6 Hexose sugars Can$ Canadian dollars C A S H Canada, America, Sweden Hydrolysis Process C H A P Concentrated Hydrochloric Acid Process CrI Crystallinity index of cellulose CST Continuously stirred tank DP Degree of polymerization of cellulose ETBE Ethyl tert-butyl ether FCI Fixed capital investment Forintek Forintek Canada Corp. FPB U B C -Chair of Forest Products Biotechnology FPU L - l h-1 Filter paper units per liter hour FPU ml-1 Filter paper units per millilitre FPU/gor FPU g-1 Filter paper units per gram of substrate g gram h hours H202 Hydrogen Peroxide H M F Hydroxy-methyl furfural IEA International Energy Agency IU/g International Units of enzyme per gram of substrate kg kilogram kPa kiloPascal l o r L litre m M Millimolar concentration M S W Municipal Solid Waste NaOH Sodium Hydroxide OD or ODW Oven-dried weight ODT Oven-dried tonne OOP Object-oriented programming OPEC Organization for Petroleum Exporting Countries P D U Process Development Unit psig pounds per square inch gauge pressure s seconds SE Steam-exploded substrate SEA Steam-exploded aspen SEE Steam-exploded eucalyptus SES Steam-exploded spruce X SHF Separate Hydrolysis and Fermentation S02 Sulphur dioxide S02-WI Water insoluble treated with steam & S02 SS Stainless steel SSC Spent Sulfite Liquor SSF Simultaneous Saccharification and Fermentation StakeTech Stake Technologies US United States of America VPI Virginia Polytechnical Institute w/v Weight to volume concentration w/w Weight to weight concentration WI Water insoluble fraction WIA Water and alkali insoluble fraction WIA/H202 Water, alkali and peroxide insoluble fraction WS Water soluble fraction °C Degrees Celsius xi ACKNOWLEDGEMENTS I would like to acknowledge the guidance and advice of my supervisor, Dr. Jack Saddler, and the members of my committee Drs Breuil, Cohen and Duff. I also extend my thanks to my family and friends for their understanding and support through "another degree !". They always provided the encouragement to continue even though I frequently questioned the direction, length and sanity of the pursuit. This work was supported by strategic grants from Energy Mines and Resources Canada, Natural Sciences and Engineering Research Council of Canada and the International Energy Agency. I am also grateful to Forintek Canada Corp. for a scholarship to carry out part of this study. xii 1. INTRODUCTION Over the past twenty-five years, there has been considerable interest in the potential of producing ethanol from biomass. An explosive increase in research in this area occurred in the 1970's with the O.P.E.C. oil embargo increasing the price of oil and influencing countries such as Japan, the U.S., Canada, Brazil and Sweden to plan for better liquid fuel self-sufficiency (Klyosov, 1986; Vallander and Eriksson, 1990). More recently research into fuels from renewable resources has also been driven by environmental concerns, particularly the role of fossil fuel contribution to poor air quality and global warming (von Sivers and Zacchi, 1995). Various studies have shown that ethanol or ethanol-blended fuels produce less harmful emissions while the production of ethanol from biomass has the advantage of displacing a transportation fuel derived from oil with a fuel from a renewable resource. Consequently there should be little net contribution to global warming as the carbon dioxide liberated during ethanol combustion is utilized by the growing plant material (von Sivers, 1995). Ethanol can be produced from sugar-based (sugar cane, sugar beet, etc.), starch-based (corn, grain, tubers, etc.) and lignocellulosic-based (wood, agricultural residues, forestry residues, and municipal solid waste (MSW)) feedstocks (Figure 1). Currently, biomass-derived ethanol is almost exclusively produced from sugar- (sugar cane) and starch-based crops (corn & wheat). These processes are proven technically and, with the current subsidization programs, appear to be economically viable (Klyosov, 1986;,Ramos, 1992; Tshiteya, 1992). The cost of the feedstocks for sugar- and starch-based processes represents -40% of the ethanol production cost. Their use as a feedstock for ethanol production also competes with their usage in food production. Thus, interest in using lignocellulosic feedstocks is primarily a consequence of their low-cost, widespread availability and lack of competition with food production. Currently, the only commercial process producing ethanol from lignocellulosic feedstocks is based on the traditional spent sulfite liquour (SSL) fermentation process utilizing the hemicellulose-derived 1 Type of Feedstock ? S U G A R B A S E D S T A R C H B A S E D L I G N O C E L L U L O S I C B A S E D P re t rea tmen t P re t rea tmen t B a g a s s e or B e e t pu lp S t a l k s , chaff c o b s , s tove r F rac t i ona t i on ^ . . J F rac t iona t ion P re t rea tmen t 1 Frac t iona t ion C E L L U L O S E H E M I C E L L U L O S E L I G N I N H e x o s e j P e n t o s e t H y d r o l y s i s j H y d r o l y s i s I ^ - (-H e x o s e S u g a r s P e n t o s e S u g a r s + I .. H e x o s e P e n t o s e I *" Fe rmen ta t i on Fe rmen ta t i on i E thano l R e c o v e r y C o n c e n t r a t i o n Ethanol Process Electricity Heat Figure 1 Biomass-to-Ethanol Process Scenarios based on Feedstock Type (Starch-, Sugar- and Lignocellulosic-based) 2 sugars from softwood substrates (Nguyen, 1991). There are only a couple of sulfite mills in the world still using this process. Ethanol processes based on lignocellulosic feedstocks provide more of a process design and operating challenge than sugar or starch feedstocks due to the wider variation in the types and nature of the potential feedstocks and the consequential differences in the processes and equipment needed to convert these feedstocks to ethanol. This is primarily a consequence of the relatively early stage of development of lignocellulosic conversion to ethanol processes as well as the much greater feedstock variability. It is recognized that sugar and starch crops have compositional variability associated with the species of feedstock used, growing site, climate, age and the part of the plant that was used. However, lignocellulosic feedstocks can also have proportional variability within the mixture of its three major components (cellulose, hemicellulose and lignin), differences in the types and amounts of extractives, and natural variability in the monomeric sugars that make up the hemicellulose component (e.g. primarily xylan in hardwoods and galacto-gluco-mannans in softwoods). There are two basic processes (acid and enzymatic hydrolysis) associated with the conversion of lignocellulosics to ethanol. Acid hydrolysis has been studied since the turn of the century and has been used commercially only during times of political or economic crisis. Consequently, it is considered to be technically viable but not proven either economically or environmentally. Various problems have been encountered with acid-based hydrolysis such as corrosion of the reaction vessels, degradation of product sugars resulting in low yields, necessity for neutralization before subsequent bioconversion, formation of numerous environmentally noxious byproducts, high cost, and solvent losses (Klyosov, 1986; Nguyen, 1993). Although active research continues in this area, the long term outlook for acid based biomass to ethanol processes is not optimistic as economic projections indicate that several major technical problems must be resolved before attractive economic returns can be attained. Biomass-to-ethanol processes based on enzymatic hydrolysis are a relatively new strategy (research started in the 1970's) that have not been proven either technically or 3 economically. A truly "generic" enzymatically-based biomass-to-ethanol process has been difficult to identify because of the heavy influence that the type of feedstock, type of by-products, and the number of unproven processes and equipment currently available would have on the design of such a process. However, it is generally acknowledged (Galbe and Zacchi, 1993), that a generic enzymatic-based process would include the following steps: pretreatment, fractionation, enzyme production, enzyme hydrolysis, fermentation, ethanol and other by-product recovery, and waste treatment. These subprocesses are strongly interdependent and there are no true examples of commercial or totally integrated demonstration-sized plants. Thus, it is extremely difficult to identify the relative merits of each of the subprocess variations, either technically or economically, and their subsequent influence on the final production cost of ethanol. It is generally recognized that, due to the immature state of the technology and the cost of pilot- and demonstration-scale facilities, one of the most efficient ways of assessing the technical and economic feasibility of enzymatically-based biomass-to-ethanol processes is the use of techno-economic modelling. These models have generally been based on information obtained from both lab and pilot studies and from the operation of similar processes in other industries or in commercial-scale acid hydrolysis plants (Douglas, 1989; Wright, 1989; Nguyen, 1990; von Sivers and Zacchi, 1995). Most of the models have been used to assess not only the current techno-economic status of a particular process (both process step and overall process) but also the future potential of the various bioconversion processes. Recent research and development (discussed in detail in both the body of the thesis as well as in the appendices) in the Chair of Forest Products Biotechnology (FPB) at the University of British Columbia (UBC) has been directed towards ways of reducing the cost of the front-end of the process, i.e. the pretreatment, fractionation and hydrolysis stages, and the potential problems encountered with the use of softwoods as the feedstock. Most of the techno-economic models that have been developed in the past have tended to model the production of ethanol from hardwoods. These models will be discussed in more detail later in this introduction and in the 4 model development section of this thesis. Past hardwood modelling work has indicated that the front-end of the overall process, i.e. the pretreatment, fractionation and enzymatic hydrolysis steps, represented about 60% of the total cost of ethanol production (Nguyen and Saddler, 1991). Softwoods constitute more than 97% of the residue produced by the sawmills in British Columbia. The unutilized portion of this sawmill residue represents a potentially large source (estimated to be annually 4.5 million bone dry tonnes) of feedstock for a biomass-to-ethanol facility (Stewart & Ewing Associates Ltd., 1990). However, little research or modelling work has been done on the differences in process options or economic viability of an overall process that may occur when softwoods are used as the feedstock. 1.1 BACKGROUND 1.1.1 Biomass-to-Ethanol Processes The following section examines the biomass-to-ethanol process from a systems analysis and management perspective with an eye towards the potential of successfully commercializing this process. Any process can be viewed at its highest level of abstraction as being composed of three elements: a feedstock, a product and a process to convert the feedstock into the product. To reflect this systems analysis approach this part of the thesis has been divided into three sections that discuss the attributes of the product, feedstock and process. 1.1.1.1 Products The properties of ethanol and ethyl tert-butyl ether (ETBE) have been discussed in detail by Wyman (Wyman and Hinman, 1990). Ethanol is a clean-buniing, high-octane fuel that can be readily substituted for gasoline in the transportation sector. It has been proven in Brazil as both a neat (close to 100%) and blended (generally 10-15% with gasoline) transportation fuel for the last twenty years. Neat ethanol provides superior efficiency and performance to gasoline because it requires lower stoichiometric air/fuel ratios and provides higher octane values. Use of neat ethanol 5 reduces smog formation because it has a the lower volatility, thus reducing emissions. It is also less photochemically reactive and produces fewer nitrogen oxides because of reduced flame temperatures. Blends of ethanol or ethyl tert-butyl ether (ETBE) with gasoline increase the octane of the mixture and can improve performance. Ethanol blends cause internal-combustion gasoline engines to run lean, and blends reduce carbon monoxide emissions by 10-30%. In addition to 10% blends, ethanol can be reacted with isobutylene to form ETBE which can be blended with gasoline. If ETBE is blended at 22% with gasoline, the amount of ethanol used is equivalent to that for 10% blends. ETBE in gasoline also reduces carbon monoxide emissions. As an added benefit, ETBE lowers the Reid vapor pressure of gasoline, thereby decreasing the release of smog-forming compounds. When ethanol is produced from renewable sources of cellulosic biomass, ethanol use can both decrease urban air pollution and reduce the accumulation of carbon dioxide, one of the greenhouse gases. It has been estimated that enough neat ethanol could be made from cellulosic biomass residues that are available within the U.S. to replace U.S. gasoline consumption twice over (Wyman, 1995). 1.1.1.2 Feedstock There are substantial benefits to using renewable lignocellulosic feedstocks for the generation of fuel ethanol. They are generally neutral with regard to atmospheric carbon contribution and lignocellulosic residues are both cheaper and generally less desirable as a feedstock for other high-value products (von Sivers and Zacchi, 1995; Wyman, 1995). The mixture of wood components (cellulose, hemicellulose, lignin and extractives) varies from one species to another. Generally there is a minor difference in cellulose content, approximately a 7% higher hemicellulose content in deciduous trees, and coniferous woods have roughly a 7% higher lignin content (Wayman and Parekh, 1990). The hemicellulose of conifers is mostly polymers of mannose incorporating minor amounts of glucose and galactose. 6 Angiosperms have a higher acetyl content, which can yield acetic acid, a strong fermentation inhibitor, upon hydrolysis (Klyosov, 1986). The importance of recovering all of the major lignocellulosic components to offset the feedstock cost and lower the final product cost has been emphasized by much of the past modelling efforts . The carbohydrate and lignin composition may also vary significantly with site, climate, age and other factors for the same species (Wayman and Parekh, 1990). Even within the same tree there is a difference between juvenile and mature wood. Consequently, an adequate sampling procedure is necessary for a representative sample of any lot to be processed and variability of the feedstock must be adequately addressed in the design of a final process. 1.1.1.3 Process The products and processing methods utilized in bioprocessing are generally a consequence of the physical properties and chemical composition of the biomaterials used and the bioactivities (Ward, 1991). In our particular case, wood residue is the biomaterial and the natural processes of wood decomposition and fermentation are the bioactivities. Bioprocessing employs both biological and technological aspects to provide an efficient process. The biological component utilizes the diversity of conditions under which microorganisms operate successfully. The technological aspect optimizes the process by creating favorable operating conditions and eliminating some of the biological constraints of the organism. Effective design and use of bioprocessing production methods such as cell and enzyme systems requires an understanding of microbial physiology, biochemistry and a capacity to develop conditions which optimize biosynthesis or bioconversion procedures (Ward, 1991). The commercial exploitation of enzymes or cells is restricted to conditions at which biological systems function i.e. moderate levels of temperature, pH, pressure, and dilute aqueous solutions. A typical biotechnological process is characterized by rather low productivity in comparison to what is normal for chemical synthetic reactions. Furthermore, the product stream 7 is dilute, which can lead to high costs in the subsequent isolation and purification of the product (Mattiason and Hoist, 1991b). There are conceivably five levels of estimating the project techno-economic sophistication. These are, in increasing levels of sophistication, accuracy and degree of construction: order-of-magnitude, study, budget authorization, project control and contractor's estimation (Ulrich, 1984). This project is a predesign estimate falling somewhere between an "order of magnitude" estimate and a "study" estimate, which requires a preliminary process flow sheet and an approximate definition of equipment, utilities, materials of construction and other processing details. The accuracy is considered to be within ± 20-30% of the actual design parameters and costs. These stages precede expenditures for pilot plant work and detailed equipment design. The first step in an estimate is often the conceptualization of the overall process. Through the use of process modeling techniques a conceptualized view of the bioconversion process can be developed (Figure 1). Even at this abstract level the complexity of this process is evident. This complexity reflects both the unique features of wood composition and the organisms and/or enzymes that decompose wood to ethanol. Fermentation is one of the key subprocesses for the effective production of ethanol. This subprocess is generally limited to converting monomer sugars to ethanol. However, the currently proven organisms cannot convert both hexoses and pentoses to ethanol at the same time (Tshiteya, 1992). Furthermore, as discussed previously, wood is a mixture of components (primarily cellulose, hemicellulose and lignin) and the lignin component can not be used to make ethanol. Therefore, a series of operations (Pretreatment & Fractionation) are required to fractionate the various wood components and hydrolyse the cellulose and hemicellulose components to monomer sugars. These process steps are not well defined and are of most interest to this study because past modelling efforts (Zacchi et al., 1988; Nguyen and Saddler, 1991; von Sivers and Zacchi, 1995) have indicated their large influence on the overall cost of producing ethanol from biomass. 8 It was felt that a detailed coverage of Enzyme Production, Fermentation, Ethanol Recovery, Lignin Recovery and Waste Treatment was not needed for this current study as they are generally better defined and further developed than the process front-end. 1.1.2 Process Simulation/Modelling Biomass-to-ethanol processes have been used since the last century. However, the use of enzymatic processes is in an early stage of development which means that a number of the technologies, primarily the front-end, have not been demonstrated in an integrated fashion at a commercial level. Furthermore, within processes at an early stage of development there are also a greater number of potential technologies to be evaluated. Proving the worth of these technologies at pilot- or demonstration-scale facilities is both expensive and time-consuming. Consequently, mathematical models have generally been developed to assist in optimizing and assessing the level of development of processes. Computer process simulations are invaluable tools for the analysis, design and economic evaluation of process steps, for comparing and optimizing process alternatives and for determining further areas of research and development (Nguyen, 1990). Techno-economic models are a subset of computer process simulators and, as their name implies, are used to assess the current technical and economic level of development of a process. Through the use of other techniques such as sensitivity analysis and Monte Carlo simulation (multiple variable sensitivity), techno-economic modelling can also be used to determine fruitful areas of research and development. Over the 20 years since the energy crisis of 1970, various techno-economic models have evaluated the economics of biomass-to-ethanol processes (Perez et al., 1981; SERI-Chem Systems Ltd. and Lawrence Berkeley Labs, 1984; Douglas, 1989; Nguyen and Saddler, 1991; von Sivers and Zacchi, 1993). These models have been based on information from both lab and pilot studies and from the operation of similar processes in other industries or in commercial-scale acid hydrolysis plants. The wood-to-ethanol process lends itself to this type of modelling for a 9 number of reasons. It is very complex, consisting of a large number of process steps (Wood handling, Pretreatment, Fractionation, Enzyme production, Enzyme hydrolysis, Glucose fermentation, Pentose fermentation, Product recovery and Waste treatment) and involves several complex materials such as wood, the cellulase complex of enzymes and various microorganisms. Unlike thermochemical options, that tend to use a "one-step process" or a single piece of equipment to convert biomass through pyrolysis, gasification, etc., to a final product, a biomass-to-ethanol process contains multiple interrelated steps and requires many pieces of equipment. Furthermore, at present, process simulations of the lignocellulosic-to-ethanol process are based mainly on laboratory and pilot-plant data as no demonstration or commercial-scale plants are in operation. This type of modelling can be used not only to assess the cost of producing ethanol from biomass, it also has an important role in assigning the associated costs and levels of maturity for each integral step. 1.2 GOALS OF THESIS 1.2.1 Development/Refining of Hardwood-to-Ethanol Model There have been a number of techno-economic models built over the last 10 years for a hardwood-to-ethanol process. Unfortunately most of the details associated with their calculational logic have not been published or made publicly available. This is not altogether unreasonable since they require a considerable amount of time to develop and would require a sizeable document to fully describe their logic. Furthermore, it is difficult to assess the effect of changes in one particular portion of the whole process without building a complete process model because: 1) the individual subprocesses are strongly interdependent and 2) the measure of progress in the development of the process is the production price of the final product. Consequently, although a full model would be required to provide a reasonable assessment of new developments in any of the subprocesses, the amount of time required to develop a completely new model would be substantial. Fortunately, the Chair of Forest Products Biotechnology recently acquired the complete description of the two hardwood modelling efforts 10 mentioned previously, a model of the whole hardwood-to-ethanol process from Forintek Canada completed in 1991 and a model of the steam pretreatment and fractionation subprocesses from Virginia Polytechnical Institute (VPI) completed in 1994. Through a process of integration and refinement of these two models, the time and effort required to build a complete hardwood-to-ethanol model was shortened. However, it must be recognized that these past models did not, for the most part, reflect the results of research completed since the early 1990's. Therefore, some new subprocess modules based on more recent data, particularly in the pretreatment and hydrolysis areas, were developed and will be discussed in Chapter 4. Results of past modelling efforts have been an important element in determining the direction of research efforts over the last 5 years. The front-end of the wood-to-ethanol process was shown by the Forintek model to represent a substantial portion of the production cost of ethanol and it is in this area that the research has been concentrated. The front-end includes the use of acid-catalyzed pretreatment (specifically the use of S0 2 and steam), fractionation combinations using water, alkali and peroxide and enzyme recycling. Therefore, before the building or modification of a techno-economic model could be attempted the technical details of the front-end had to be reviewed and assessed. Steam has been used as a pretreatment in the past because of its lower cost and an acid catalysis is often used to reduce both the required saturated steam temperature/pressure and the potential for the production of inhibitory breakdown products (Ramos et al., 1992a). Detailed analysis of the effect of the acid addition on the capital cost, operating cost and downstream characteristics, including inhibitory product generation and waste treatment, has not been previously done. Current efforts to install a lab-scale steam gun designed to handle S0 2 pretreatment have provided insights into the capital and operating cost issues. Although there has been some initial research on the generation of inhibitory products and their distribution in various process streams, a more detailed assessment of how the production of potential inhibitors can be minimized or ways of removing/treating these inhibitors is needed. The types and volumes of the various waste treatment streams being generated from this process have been 11 assessed by a number of groups since 1991. To include all of these areas of investigation would require substantial model development and implementation time. Consequently, only the influence of incorporating the economic effects of S0 2 catalysis were assessed in this work. Past modelling has concentrated on the combination of water and alkali washes to remove the hemicellulosic and lignin components from the wood. Recent lab work has investigated the addition of a final peroxide wash to enhance lignin removal (Ramos et al., 1992b; Lee et al., 1995). However, the relative benefits of this treatment in relation to the cost and additional complexity of the overall process have not been evaluated. This was one of the objectives of this thesis. A major cost in the overall process has been the cost of enzymes. There has been a considerable amount of research effort directed to reducing the cost and enhancing the effectiveness of enzymes over the last two decades. Recycling of the enzymes has been recently investigated in the lab and shown to be technically effective at this scale (Ramos et al., 1993; Ramos and Saddler, 1994; Lee et al., 1995). Detailed technical issues associated with enzyme recycling are discussed in the thesis. 1.2.2 Softwood Modelling British Columbia possesses a large sawmill residue surplus composed almost exclusively of softwoods. There has been considerable interest in the last decade in the development of value-added products from these residues (H. A. Simons Ltd., 1991). Previous evaluations have shown that the conversion of these residues into various forms of energy have the most potential for providing value-added products. One energy option, the production of fuel ethanol, provides a value-added product that is more environmentally acceptable when compared to petrochemical fuels and it offers a potentially huge market. Most of the past research, on the development and modelling of wood-to-ethanol processes has used hardwoods as the lignocellulosic feedstock. However, an evaluation of a potential wood-to-ethanol process based in B.C. must address the process differences required when using a softwood feedstock. As described in the thesis, there is a considerable lack of information on pretreatment and hydrolysis of softwoods. 12 1.2.3 Model Format Conversion We ultimately based our wood-to-ethanol model on the integration and refinement of the two inherited spreadsheet format models. However, spreadsheet models possess certain developmental limitations. Their calculational logic is difficult to document and modify because it is held within the worksheet cells and these may be spread all over the worksheet. Furthermore, the use of range names, for making the model calculations more understandable, can often exceed the storage capabilities of the spreadsheet. Also, spreadsheet format models in the past have not had drawing capabilities beyond simple line graphics. This has meant that the flowsheets of the process were not directly connected to the calculational component of the model. Currently, the chemical and biochemical industries are using more sophisticated programs that follow an object-oriented flowsheeting format and address a number of the limitations of spreadsheet. Flowsheet simulator programs have been shown to significantly reduce the time and effort required to design, modify, model and evaluate processes. They are written in a structured object-oriented programming language, possess a graphical interface (composed of windows, menus and dialog boxes), and have object-oriented drawing capabilities (equipment, process elements and flow streams). Consequently flowsheeting programs have a greater capability to integrate the entire design, modification, modelling and evaluation process than spreadsheets. Similarly, because they have been developed using a structured programming language, such as FORTRAN or C, they potentially have more program portability. Early in this study it was apparent that the time and effort required to fully convert the Forintek and VPI models to a flowsheeting format or develop a completely new flowsheeting model would be substantial. As a compromise, a new model (named STEAM) was developed from these past models, using a spreadsheet program that possessed a number of the elements (a complete structured programming language, drawing capabilities and object-oriented controls) that could provide most of the capabilities present in object-oriented flowsheeting format. The work in this thesis therefore had four main objectives and these are reflected in the structure of the rest of the thesis. The objectives were to: 13 1) Determine the level of maturity of an overall hardwood/softwood-to-ethanol process including the development of an overall process flowsheet based on past research and development and/or similar processes in other industries; 2) Define the technical details and assumptions for the front-end of the process; As mentioned previously the front-end of the biomass-to-ethanol process is the most expensive, least mature and consequently also the least well-defined of the various process steps. The process front-end is also generally recognized as being most influenced by a change in feedstock. The technical and economic issues associated with the front-end must be defined and discussed in detail in order to adequately develop a model that reflects the process and also provides the flexibility to address these and future issues. 3) Define and construct the STeam Explosion Assessment Model (STEAM); The STEAM model was developed from two past spreadsheet models and the modifications to those models addressed the lack of flexibility issues associated with their structure, user interface and the technical issues discussed in objectives 1 and 2 . 4) Use the STEAM model to address the technical and economic impact of recent advances in areas such as steam pretreatment and enzyme recycle. This last section of the thesis reviews the results produced by the developed model. The model runs include an assessment of the effect of the technical details described in objectives 1 and 2. 14 2. PROCESS D E V E L O P M E N T One of the major steps in the construction of a detailed techno-economic model is the development of a process flowsheet. The flowsheet in this case was for a "generic process" incorporating all of the main steps of a wood-to-ethanol plant. Process flowsheets generally have two levels of organization. A higher level, that divides the overall process into a series of subprocesses that each contain a particular or distinct group of unit operations, and a lower level, that indicates the major pieces of equipment for each of the unit operations. Frequently, a large number of possibilities and potential assumptions exist, particularly at the lower level of flowsheet organization. Due to the complexity and early stage of development of this bioconversion process there have been various processing strategies and conceptual designs presented in the literature. The majority of the process flowsheet drawings and descriptions that follow are composed of subprocesses and selected equipment from past modelling efforts (Douglas, 1989; Nguyen, 1990; von Sivers and Zacchi, 1993; Avellar, 1994). The process flowsheet development includes an initial description of the overall process, followed by a much more detailed description of the major pieces of equipment required for each of the subprocesses. The integrated process is organized to reflect the flow of cellulose through the process as this is generally considered to be the most important component of a biomass-to-ethanol process. 2.1 PROCESS DEFINITION - IDENTIFICATION OF SUBPROCESSES The overall process configuration for the conversion of hardwoods to fuel ethanol (Figure 2) incorporates the process steps required to utilize all three of the main components of wood and overcome the restrictions imposed by both the fermenting organisms and the wood structure. In many ways this configuration is just an overview of the entire process but it does show the major inputs, major products and larger pieces of equipment for the generally accepted 15 subprocesses (Wood Handling, Pretreatment, Fractionation, Hydrolysis, Fermentation, Product Recovery). As mentioned earlier the primary development focus of this model was on the front-end subprocesses i.e. Pretreatment, Fractionation, and Hydrolysis, as these are the areas generally recognized as being the least well defined and they contribute the most to the ethanol production cost. These front-end subprocesses are discussed in greater depth in Appendices 8.1 and 8.2. Other subprocesses such as enzyme production or waste treatment are also considered in the model. However, certain options have been chosen such as the purchase of enzymes and the use of traditional waste treatment methods to reduce the model development time. 2.2 SUBPROCESS DEFINITION - IDENTIFICATION OF KEY EQUIPMENT 2.2.1 Wood Preparation Two approaches were presented in the literature for the wood preparation step. The simplest approach assumes that the cost of wood preparation is minor in comparison to the rest of the process and as such assumes that the wood is delivered in a form that is already suitable for direct entry into the pretreatment subprocess and not stored i.e. just-in-time delivery. The more complex and likely more realistic approach assumes that a portion of the chips will have to be resized and there will be a requirement for a stored inventory of chips. This latter approach will be assumed in the model and the details of the procedure and equipment presented below are drawn from one of the previous modelling efforts.(Nguyen, 1990) on which this current model is based. Chipping and sizing equipment and technology is commercially available. All of the technology associated with chip storage is commercially available and used extensively by the pulp and paper sector. Once the feedstock is in a form where it can be handled, it is then conveyed to the pretreatment component of the overall process. 2.2.2 Pretreatment The proposed pretreatment (a detailed rational is given in Appendix 8.1), i.e. S02 steam-explosion, is composed of a series of tasks (Catalyst Impregnation, Steam Injection, Steam 17 Explosion, Steam Recovery, Slurrying and Transport). Although various commercial and pilot facilities have been sold by Stake technology this is still an area for further development. Steam is a costly commodity and is recovered by all of the modellers (Douglas, 1989; Nguyen, 1990; von Sivers and Zacchi, 1993; Avellar, 1994). Steam explosion lowers the pressure of the steamed slurry from the medium and high pressure steam levels down to slightly above atmospheric pressure in an attempt to recover some of the steam. Most of the extractives and volatile components of the wood residues and the hemicellulose derived breakdown products can be driven off using steam explosion. The only major piece of equipment associated with this operation is a medium pressure flash vessel. The condensate and excess steam flash off the blow tank into the shell side of a steam condenser. The pressure in the shell side of the steam condenser is maintained at 345 kPa (50 psig), i.e. the same pressure as in the blow tank. Clean steam at 83 kPa (12 psig) is generated in the tubes of the steam condenser. The low pressure steam can be used for sterilization of fermentation equipment and for heating the lignin precipitate slurry. The saturated condensate from the steam condenser is passed through a hot water heater for further heat recovery. Cold (10°C) process water is heated to 95°C. A side stream of the hot water is used to generate low pressure steam in the steam condenser. The hot water is used for tempering process water used in hemicellulose and lignin extraction. It is assumed that 90% of the heat input into the steam condenser and hot water heater is recovered. This assumption is likely too high because the steam from the blow tank contains volatiles and fine wood particles which would likely cause fouling of heat transfer surfaces and reduce heat transfer coefficients. 2.2.3 Fractionation Although much of the equipment is available to carry out effective fractionation of steam treated biomass, this step is also at the development stage and has yet to be fully demonstrated at the pilot plant level. The proposed method (a detailed rational is given in Appendix 8.1) for fractionating the pretreated slurry into its cellulose, hemicellulose and lignin components (Figure 18 3) is the use of a water wash to separate the hemicellulose component from the wood and the use of both sodium hydroxide (NaOH) and hydrogen peroxide (H 2 0 2 ) washes to separate the lignin component. The remaining cellulose component (peroxide extracted slurry) is then transported to the enzymatic hydrolysis step. 2.2.3.1 Hemicellulose Extraction. The following process description and equipment selection guidelines are primarily drawn from one one of the two previous modelling efforts (Avellar, 1994) on which the current model is based. High extraction yield and dissolved solids concentration in the extract are required to minimize the costs of extraction and subsequent conversion. Therefore, a continuous countercurrent extractor is required for the hemicellulose removal. The blown fiber from the steam gun is discharged directly from the cooker section of the StakeTech gun or Masonite gun into a pressure quench tank. This quench tank is fed with the concentrated wash water from the extractors down-stream. This operation is a good method for slurrying the blown fiber due to the high agitation of the steam explosion. It is also an effective method of continuing the counter-current extraction further down stream, by contacting the fresh fiber with the highest soluble concentration solution. The quench tank also serves to heat the extract water. The slurried fiber is pumped to the first of three rotary vacuum filter extractors. These extractors are arranged for counter-current contacting of the extracting water and fiber. The filtrate from each extractor is pumped to the wash nozzles of the previous extractor. Fresh hot water is introduced at the last extractor. The fiber is discharged from each extractor and is reslurried with filtrate from the next extractor before introduction to that extractor. This method of fiber washing is based on brown stock washing in the pulp and paper industry. The filtrate from the first extractor will become the most concentrated once the operation is at steady state. A fraction of this filtrate is withdrawn as product while the remainder is pumped to the quench tank as the slurrying solution. The fiber from the third and last extractor is discharged as a wet mat to a screw press. 19 H2Q t 4 ) NaOH H 2 Q I i 4 ••t> Waste Hemicellulose "Slurry Waste Lignin Recovery Peroxide "Extracted Slurry Waste o Lignin Recovery or Waste Figure 3 Fractionation Details (adapted from Avellar, 1994) 20 The screw press removes as much water as is possible mechanically to provide a manageable product. A portion of the pressate solution is blended with the fresh wash water to recover as much of the hemicellulose-derived solution as possible. The fiber that is discharged at this point is the water extracted fiber and is passed to the alkali extraction step. The major equipment for this module are a quench tank, three rotary vacuum washer/extractors with all the necessary vacuum pumps, filtrate receivers and filtrate pumps and two screw presses with cake conveyors. The selection of rotary vacuum filter extractors is based on their continuous operation, wide spread use in similar industries and their relatively low cost. The critical dimension in specifying these extractors is the necessary area required for effective filtration. Sizing is based on similar operations in the pulp and paper industry. 2.2.3.2 Lignin Extraction and Recovery. High lignin yield and low solvent loading are important to the process economics. The lignin extraction yield is dependent on factors (discussed in greater depth in Appendix 8.1) such as, pretreatment effectiveness, caustic concentration, extraction method used and the use of further extraction steps e.g. peroxide wash. Concentrations of caustic over 5% also increase the lignin extractibility although the chemical cost is extremely high. Two consecutive washes are required to attain over 95% of the extractables in batch extraction. A continuous countercurrent screw-conveyor extractor demonstrated at a pilot scale obtained a 5-6% lignin concentration in dilute sodium hydroxide and would be more efficient than a stirred batch extractor in commercial operation (Nguyen, 1990). Further extraction steps can be added to raise the lignin extraction yield, although this would add to the capital and operating cost. Previous modelling has has usually only included a water wash or a combined water-alkali wash scenario. Recent research (discussed in more detailin Appendix 8.1) has shown that a peroxide wash can enhance lignin recovery from both hardwoods and softwoods. Peroxide, at the proposed concentrations, will also serve to sterilize the washed fiber. 21 The alkali extraction process description and equipment selection will be drawn primarily from previous modelling work (Avellar, 1994). Although the equipment and configuration for the peroxide extraction process has not previously been described in the literature it will be assumed that counter-current contacting will be required in a manner similar to the preceding two extraction steps. Two previous modellers (Nguyen, 1990; Avellar, 1994) have proposed lignin precipitation as a method to recover lignin. Under the right conditions, precipitated lignin solids filter fairly rapidly and leave relatively dry cake. The optimum temperature for precipitating the lignin is 80-90°C. Higher temperatures tend to boil the solution and the added agitation breaks the floes, thus resulting in poorer filtration. Lower temperatures tend to cause a very colloidal precipitate which filters very slowly. Lignin solids concentrations of 3-5% are adequate for recovering easily filterable solids. Higher concentrations of up to 10% will improve the filterability. The final pH range must be maintained at 3-3.5 for good precipitate formation and flocculation. Lower pH is acceptable but wastes acid. Higher pH does not completely precipitate all the lignin solids in solution nor does it allow good floe formation. The major equipment required includes one clarifier, three vacuum table filters, filtrate pumps, acid storage and injection pumps and cake conveyors out of the module. Other equipment includes acid handling equipment, surge tanks for mother liquid and cake transfer conveyors. 2.2.4 Hydrolysis 2.2.4.1 Hemicellulose Hydrolysis The majority of hemicellulose hydrolysis occurs during the acid catalyzed steam pretreatment of the biomass. A separate enzymatic hydrolysis step is not required as more than 70% of the original xylan in a hardwood can be obtained as xylose after S02 catalyzed steam pretreatment (Nguyen, 1990). The hydrolysate obtained from the water wash in the fractionation step can therefore be sent directly to the pentose fermentation step. However, the majority of 22 the fermentation inhibitors such as acetic acid , sugar degradation products, etc., are associated with this stream and it is likely that an inhibitor removal stage will have to be incorporated. 2.2.4.2 Cellulose Hydrolysis Over the last 5 years cellulases have been used in increasing amounts for various food, feed, textile, detergent and pulp applications (Lange, 1993). Consequently, it is anticipated that it would be simpler, for the piloting and first commercial plants, to initially purchase the enzymes from commercial sources and have them delivered to the plant rather than produce them on-site. The high degree of uncertainty (discussed in detail in Appendix 8.2) in the commercial operation of an SSF process has encouraged us to concentrate on the SHF approach as there is well established data for the glucose-to-ethanol step. Consequently, in the flow-sheet of the ethanol production process (Figure 2), the enzymatic hydrolysis, the hexose fermentation and the pentose fermentation have been presented as separate operations. Tasks involved with hydrolysis are primarily associated with the maintenance of optimum environmental conditions for the reaction. The essential tasks are control of: end-product inhibition, temperature, pH, sterility, viscosity, enzyme recycling and handling the unhydrolyzed residue. The equipment selection for cellulose hydrolysis (Figure 4) is drawn from previous modelling work (Douglas, 1989; Nguyen, 1990). Due to the end-product inhibition and maximum slurry concentration problems the slurry is processed in a series of four fed-batch continuously stirred reactors (CST). The reactors are made of 304 stainless steel (SS) and equipped with side-entry impellers. The fed-batch system periodically adds fresh enzyme to the CST's and reduces the start-up loading on the agitators. Problems such as sedimentation of the cellulase enzymes within pumps and screw augers have been reported. Consequently, the enzymes are assumed to be added in dry form. Environmental conditions within the reactors are maintained at 48°C and a pH of 4.5 and all process water is sterilized to help prevent fouling by other microorganisms. 23 Temperature control is implemented through a flow of hot or cold process water into the reactor jacket. The pH is controlled by direct addition of acid or base to the reactor. All cleaning of equipment is done with sterile water and/or steam. The reactor requires a process water input as the slurry concentrations cannot exceed 10% because the high viscosity prevents agitation. 2.2.4.3 Cellulase Recycling Recycling is used to reduce the high input costs associated with enzymes. Several strategies have been attempted to increase the yields of enzyme recovery and recycling during cellulose hydrolysis (Clesceri et al, 1985; Vallander and Eriksson, 1987; Tanaka et al., 1988; Singh et al., 1991). The recycling of the enzyme adsorbed to the solid residue will follow a procedure described in a previous modelling effort (SERI-Chem Systems Ltd. and Lawrence Berkeley Labs, 1984). Enzyme recycling is accomplished through a fed-batch counter-current system of triple tanks, transfer pumps and centrifuges in which the hydrolyzed residue is allowed to come in direct contact with new substrate for enzyme transfer. 2.2.5 Fermentation 2.2.5.1 Glucose Fermentation. The technology and equipment required for glucose fermentation in a wood-to-ethanol facility is believed to be similar to that found in the sugar- (sugar cane) and starch- (corn and wheat) -to-ethanol processes. These latter two processes are both viable at the commercial-scale. Glucose is generally fermented by Saccharomyces cerevisiae in continuous batch fermenters with yeast recycle. The fermenters are maintained in a slightly aerobic condition. As preliminary process design and techno-economic modelling generally first concentrate on the major components of a process unit, and because the fermentation process is a biochemical reaction, most of the design efforts have revolved around the reactors. Bioreactor design must address a variety of concerns including: the properties of the biological agent (cells or enzyme), the nature of the raw material or substrate, the properties and desired specification of the 25 product, batch or continuous nature of the process, mixing and gas exchange requirements, sterility maintenance, process control and validation, and techno-economic considerations. Batch reactors continue to be the mainstay of fermentation technology. Various methods have been devised to overcome the end-product inhibition of ethanol and C0 2 on the fermenting organisms. The two most commonly mentioned methods are: recycling of the yeast from the product stream to increase the cell density in the reactor; and continuous fermentation technology. These strategies remove the product from the reactor as soon as it is formed and supply new substrate at the same time. These systems have been studied intensively. It seems the use of continuous fermentation in a fluidized bioreactor system, coupled with immobilization and/or recycle offer the potential to substantially reduce ethanol production costs. These methods have not been demonstrated over extended periods of time (Mattiason and Hoist, 1991a). The stream of glucose or mixed pentose and hexose sugars corning from the hydrolysis and fractionation stages respectively will be stored in tanks called beer stills (Figure 5). The hydrolysate resulting from the pretreatment will not only contain fermentable monomer sugars, but also substances that inhibit the fermentation process. The composition and quantity of these compounds is dependent on the chemistry and nature of the pretreatment and the type of wood used. Furthermore, stream recirculation to minimize the.usage of certain inputs such as water and enzymes can result in the accumulation of inhibiting compounds in the hydrolysate. Fermentation inhibitors have previously been divided into five groups: substances released during pretreatment (acetic acid and extractives such as terpenes, alcohols and tannins) (Frazer and McCaskey, 1989), sugar degradation products (furfural, hydroxymethyl furfural, laevulinic acid, formic acid and humic substances) (Clark and Mackie, 1984), lignin degradation products (wide range of aromatic and polyaromatic compounds), fermentation products (ethanol, acetic acid, glycerol, lactic acid) (Maiorella et al., 1983a), and others (metals from equipment and additives such as S02) (Pilkington and Rose, 1988). The literature suggests that acetic acid and 2 6 lignin degradation products are responsible for the main inhibitory effect (Baugh et al., 1988). Saccharomyces cerevisiae has been adapted to acid hydrolysates and this should therefore avoid the requirement for detoxification of the cellulose hydrolysis stream (Banerjee et al, 1981). There is a noticeable lack of information on the inhibitors associated with the sugars streams entering the fermentation step. It will be assumed that the hexose yeasts will be able to adapt to the concentration and types of inhibitors found in the cellulose hydrolysate. Following its passage through the detoxification step the sugar solutions enter the reactors. The reactors are in a fed-batch configuration to provide higher yields through progressively increasing the yeast concentration from start to finish. The reactors must be sealed to prevent contamination from other organisms and to maintain an anaerobic environment required for glucose or xylose fermentation. Each vessel requires inputs for control of pH, temperature, viscosity, foaming, yeast concentration, nutrient concentration, and dissolved oxygen. Yeast recovery is accomplished through the use of a continuous centrifuge and placed in a storage tank used for both recycled and make-up yeast addition. Carbon dioxide produced during the fermentation is drawn off to prevent end-product inhibition and trapped with a gas scrubber. The carbon dioxide can be processed and sold, used for a gas blanket in the hydrolysis stage or treated in a waste treatment plant. 2.2.5.2 Pentose Fermentation. This step is still essentially at the laboratory scale. Although significant advances have been made in both our understanding and modification of pentose fermenting microorganisms (Ohta et al., 1991; Hahn-Hagerdal et al., 1993) none of the wild-type, adapted or engineered strains have been able to routinely ferment the pentose-rich, water soluble stream obtained from steam exploded hardwoods. Even if efficient fermentation of xylose to ethanol can be obtained, the presence of most of the wood degradation products in this fraction means that effective methods of inhibitor removal are required. Thus separate fermentation and inhibitor removal 28 facilities will be required for any bioconversion process with a substantial pentose sugar component in the feedstock. 2.2.6 Product Recovery This is considered to be a mature technology and will only be briefly described here. The beer resulting from fermentation contains roughly 8-12% ethanol by volume. This dilute concentration is a consequence of ethanol end-product inhibition. This ethanol recovery configuration from a previous model (Douglas, 1989) and the main distillation process is in this case a series of two distilling columns, stripping and rectification, and the ancillary units to support the columns (reflux condensers, reflux drums, transfer pumps, and reboilers). The overhead stream of the 16 sieve tray stripping column is approximately 40% ethanol. Energy is supplied by low-pressure steam. This stream is concentrated to the 91% azeotrope in the atmospheric 24 sieve tray rectification column. To simplify the estimate of steam consumption in distillation, it is assumed that the C 6 beer and C 5 beer are mixed together before being pumped into the rectifier column. In practice, however, because of the difference in ethanol concentrations of the beers each beer would likely be fed into different feed locations in the rectifier column in order to optimize the steam requirement. Therefore, the estimate for steam requirement is slightly on the high side. To reach the anhydrous alcohol level a molecular sieve is used. The desiccant preferentially absorbs water over ethanol as the ethanol passes through the bed. The spent desiccant is regenerated by simultaneously applying heat and pulling a vacuum on the bed through a condenser and vacuum pump. 2.2.7 Waste Treatment Process development of enzymatic hydrolysis of lignocellulosic sugars is not as advanced as that for the sugar-based, starch-based, or acid-hydrolysis lignocellulosic -based ethanol producing processes, therefore knowledge of the characteristics of the wastestreams produced is less specific. However, a literature investigation into the utilisation and treatment of 29 wastestreams from an enzymatic hydrolysis process has been completed (Frings and Coombs, 1992) and four major wastestreams are expected; a wastestream from the pretreatment unit; spent fungal mycelium from the enzymatic hydrolysis unit; unhydrolysed cellulose rich residue containing spent enzyme and lignin; and a stillage from the distillation column that has a lower concentration of carbohydrated degradation products than an acid hydrolysis process due to the milder processing conditions. In general the stillage wastestream is the largest in volume (approximately 12 L stillage/ L of alcohol), it has a high BOD with a low pH, it often contains high levels of coloring compounds and has been noted for its high corrosivity (Maiorella et al., 1983b). Biological treatment of stillage generally involves an anaerobic treatment prior to aerobic treatment and a tertiary treatment to remove coloration. This type of treatment system has been previously modelled (Douglas, 1989). The spent mycelium and residual cellulose can probably be landfilled or incinerated while the pretreatment waste stream could be treated by biological processes such as aerated lagoons or activated sludge systems. 30 3. A S S E S S M E N T O F P A S T M O D E L L I N G O F B I O M A S S - T O - E T H A N O L P R O C E S S E S 3.1 B A C K G R O U N D 3.1.1 S c a l i n g One of the major design concerns whenever a process is taken from concept through to commercial implementation is the level and reliability of scaling the process and equipment to a larger size. The behavior of a lab-scale process should resemble that of the commercial size but variables may or may not increase in a linear fashion. Therefore, testing is done at one or various intermediate sizes to insure successful scaling of the variables at the commercial size. For the chemical process industry a pilot plant is usually necessary when the process includes certain unit operations or pieces of equipment such as reactors, agitators or filters. For these processes the maximum scale-up ratio is from 10:1 to 100:1 and the safety or overdesign factor is 10-20% (Peters and Timmerhaus, 1991). However, the recommended process for conversion of wood residue to ethanol incorporates both chemical and biochemical processes that have been attempted at various levels of development (IEA, 1992). Although fermented beverages have been produced for several thousand years, biochemical engineering is not yet fully mature (Perry and Chilton, 1973). Scale-up ratios and safety factors for biochemical processes should therefore be at the more conservative end of the range. Fermentation presently provides the best model with the important scale-up parameters being associated with sterility, agitation, aeration, and heat transfer (Vogel, 1983). Each of the parameters, except sterility and risk tolerated, becomes progressively more difficult to predict as the size of the vessel is increased. There is an inverse relationship between the scale-up factor for sterility and the risk tolerated. Past techno-economic modeling has used feedstock throughputs ranging from 100 to 2000 BDT/d for the full-scale enzymatic hydrolysis plants (Douglas, 1989; Nguyen and Saddler, 1991). The modeling has also shown the plant scale to have only minor influence on the price of the produced ethanol above a 400 BDT/d residue flow rate. 31 3.1.2 Process Integration Biotechnological processes were long regarded as a sequence of different unit operations. The design goal was then to optimize each individual step as far as possible. However, since biology itself contains many levels of regulation, this was not the optimal way to carry out processes in all cases (Mattiason and Hoist, 1991a). Integration means that some of the unit operations are carried out simultaneously or intermittently in a coupled way. The integration of bioconversion and downstream processing was a natural area to start because product inhibition and product instability are two major limitations of biotechnological processes. One of the processes that has been described in this thesis has been termed the Separate Hydrolysis and Fermentation process (SHF) because of its non-integrated nature. However, the Simultaneous Saccharification and Fermentation (SSF) is an integrated process (Tshiteya, 1992). It combines the hydrolysis and fermentation steps in an attempt to remove the end-product inhibition of the cellulase group of enzymes. It also has a higher alcohol concentration, lower enzyme loading requirement, shorter total time than when the saccharification and fermentation are carried out separately and a cost reduction by eliminating expensive reaction equipment. The one drawback appears to be that the combined reaction temperature must be lower (37-40°C) for hydrolysis because there are currently no fermentative organisms available that can withstand the 50°C SHF level. Direct Microbial Conversion (DMC), accomplishes the same as SSF plus it integrates the enzyme production stage using specific bacteria (Tshiteya, 1992). DMC therefore provides a single-step process for the production of ethanol from all of the sugar polymers in cellulosic biomass. This process requires more research to overcome low product selectivity, low ethanol tolerance, low productivity of ethanol, and cellulase degradation rates. A final example of an integration is the Organosolv pretreatment that includes elements of both the pretreatment and fractionation stages (Tshiteya, 1992). 32 While all of these integration pursuits have generated a great deal of interest and may eventually provide gains in the process towards commercialization, they are all currently unproven and at an earlier stage of development than the SHF process. 3.1.3 Computer programs It is important to understand the basic features of a computer program in order to understand the differences in the types of models that have been used in the past for techno-economic modelling of the lignocellulosic-to-ethanol process. A computer program contains a series of instructions which directs the computer to perform those tasks necessary to process data and produce a desired output. To execute a program, the program must be loaded into main computer memory. The actual instructions which are executed must be in the form of machine language, which is a set of instructions the electronics of the computer can interpret and execute. In some cases, the instructions to be executed are stored in computer memory as machine language instructions and are called compiled computer programs. In other cases, the instructions are stored in the format written by the programmer and must be changed into machine language one statement at a time as the program is executed and are called interpreted computer programs. Although the instructions which are actually executed by the computer must be in a form called machine language, the programmer who writes the program does not normally write in machine language. Instead, the programmer uses a source language, which is designed to make it easier for the programmer to code a program. After the program is written in the source language, the instructions in the source language must be translated into machine language for actual execution on the computer. The programming for the first stored program computers was perfomed in machine language. Since machine language programming was a difficult, burdensome and error-prone task, programming languages were developed to facilitate the coding process. 33 The first programming languages were symbolic programming languages, commonly called assembler languages. These languages used symbolic notation to represent machine language instructions. Symbolic programming languages are closely related to machine language and the internal architecture of the computer on which they are used. These languages are called low-level languages since they are so closely related to the computer's internal design. The principal advantage of assembler language is that a program can be written which is very efficient in terms of execution time and main memory usage as there is nearly a one-to-one basis between an instruction and machine language. There are however, several significant disadvantages of assembler language. As mentioned previously the languages are closely related to a computer's internal architecture and are therefore not machine independent. Additionally, an assembler language programmer will normally have to write a larger number of statements to solve a given problem than will a programmer using so-called high-level programming languages in which generally one statement will develop a number of machine language instructions. Furthermore, because of the concise symbolic notation used in assembler language, assembler language programs are often more difficult to write, read, and maintain than programs written in high-level languages. 3.1.3.1 High-level Programming Software Interpretive languages, most notably BASIC, require the interpreter program to be resident in computer memory and when the BASIC program is executed the interpreter translates each program statement to machine language. The interpreter then sends the machine language instruction to the Central Processing Unit (CPU) where it is executed. Thus, an interpreter translates a program statement written by the programmer into a machine language each time the statement is executed. Compiled languages, such as FORTRAN, Pascal, C, etc., are written in a high-level language or source code and then upon completion are translated or compiled into machine language and stored as object code in a separate executable file. The major advantage of a compiler 34 over an interpreter is that each source statement need not be translated each time it is going to be executed; it is translated only one time when it is compiled. Therefore, the execution of an object program is much faster than the execution of an interpreted program. There are two other methods used to develop application software appropriate for techno-economic modelling of lignocellulosic-to-ethanol processes. Fourth generation software development tools such as spreadsheets can be used to assist in developing applications or prewritten generalized application software packages such as flowsheet simulators can be purchased. 3.1.3.2 Electronic Spreadsheet Software This type of application software is a major tool for corporate decision support systems. The user enters as input data the values which are to be used in calculations, and also enters the formulae which are to be used to perform the desired calculations. The program performs calculations on the input data based upon the formulae entered by the user. The most powerful part of electronic spreadsheet software is the ability to ask "what i f questions and have the software perform calculations based upon the new assumptions. Spreadsheets have become popular as a way to rapidly develop the calculational relationships between variables, via cell addressing or cell naming conventions, without having to learn a programming language. Spreadsheet software being a developed application has many capabilities that previously had to be programmed such as the ability to format and generate reports and/or charts, the ability to read, illustrate and write the contents of storage files and in a small way internal error checking and debugging by showing intermediate results. Spreadsheets also have unique features such as the easy incorporation of supplied spreadsheet functions (interest calculations, look-up tables and random number generators, etc.), the ability to rapidly do sensitivity analysis by changing the value of one cell or variable and noting the change in the final ethanol price, development of simple flat-file databases and the ability to automate common tasks through the development of small programs using macro languages. 35 Spreadsheet programs are especially suited for economic calculations. The Lotus 1-2-3 program was used by Douglas (Douglas, 1989) to perform simulations and economic evaluations of the IOGEN process, and by Wright et al (Wright et al., 1988) to evaluate the economics of an ethanol process based on SSF. The main drawback of tailor-made simulators are their lack of flexibility. This creates problems when changes are made in the flowsheet, such as the introduction of new unit operations or new components. Also, the modification of the model to include a unit operation or changes in the physical property data of the components is very time-consuming. Convergence problems may also arise when many recycling streams are involved in the process. 3.1.3.3 Flowsheet Simulator Software Flowsheeting simulation has become an accepted tool for the chemical engineer for both design and process rating in the chemical process industry and some biochemical processes have also been modelled. The central part of the flowsheeting program is the unit operation modelling. In general, flowsheeting programs are based on a modular approach where each module is a mathematical model of a unit operation. The actual simulation is performed by arranging different unit operation modules such as mixers, splitters, reactors, distillation units, etc., such that a complete flowsheet is produced. Models for many standard operations are available in the program. The simulator also includes routines for the calculation of component property data, which consists of the thermodynamic properties of the chemical substances used in the simulation, and it also includes a component property data library. It is also possible for users to incorporate their own unit operation modules and property models and to add new components to the data bank. This is necessary for the lignocellulosic conversion processes as there are no existing models for process steps such as wood pretreatment and washing. The control unit assembles the flowsheet, controls all the calculations and ensures the convergence of all recycling streams and design specifications. The input of the flowsheeting 36 program consists mainly of the flow rate, composition, and physical state of all feed streams, the major equipment items (unit operations) in the plant, their operating conditions and the connections between the unit operations which result in mass and/or energy flows between the units. The program then calculates the results for all streams and the performance of the selected process units. The choice of flowsheeting program depends on many factors of which the following are the most important: a large number of built-in unit operation modules, especially for the unit operations considered in the process of interest; the ability to handle all the components present in the process, especially complex components; ease of incorporating one's own unit operations and physical property models; high reliability, i.e. well-tested models and component data to avoid erroneous results and to ensure convergence; program cost, both the purchasing cost and the operating cost, including personnel and computer time. There is a steep learning curve with these programs (as their generality presents a large number of options for defining unit operations and flowstream component properties) and continuous use of the program is required in order to take full advantage of the program, especially for processes involving complex components. The simulation of ethanol production from cellulosic materials differs from the simulation of conventional chemical processes since complex solid and heterogeneous materials are used (wood and biomass). Most process simulators are not capable of handling these kinds of components which cannot be characterized by conventional means (standard thermodynamic properties). The unit operations can be modelled at different levels of complexity and accuracy. Models for unit operations involving reactions such as pretreatment, hydrolysis, enzyme production and fermentation are especially important. The detailed reaction mechanisms are, generally, unknown. The reactions are also affected by changes in the process which cause changes in the pH, the temperature, or the concentration of certain components such as inhibitors. The model can be anything from a very simple one, based on fixed yields, to a complex mechanistic model. 37 3.2 H I S T O R I C A L B I O M A S S - T O - E T H A N O L M O D E L L I N G Process modelling of the biomass-to-ethanol process has been ongoing since the early 1980's with most earlier efforts (Perez et al., 1981; Wald et al., 1981) concentrating on the enzymatic hydrolysis portion of the process and basing the economics on the price of glucose. At this time, most computationally intensive tasks such as techno-economic modelling were restricted to larger expensive main-frame computers. Microcomputers were only beginning to enter the market and the process modelling tools were restricted to high-level programming languages and procedural structuring. This often meant that this type of modelling was viewed as being too complex and expensive. Spreadsheets were severely limited in their computational and storage capabilities and flowsheet simulators were not yet available. At this point in time, the modelling efforts did not simulate the entire process. Consequently they did not imitate the complex interrelated nature of the biomass-to-ethanol process steps and provide the means to determine the influence of both the individual and combined process steps on the production cost of ethanol. These models could be said to represent the mathematical equivalent of a non-integrated physical model such as a lab-scale process development unit (PDU) or a small pilot plant. Integrated biomass-to-ethanol process models (SERI-Chem Systems Ltd. and Lawrence Berkeley Labs, 1984; Arthur D. Little Inc., 1985) began to appear around the mid-1980's and continued, along with most engineering and computationally-intensive applications, to utilize high-level programming languages and procedural structuring for their development. The graphical user interface and object-oriented programming concepts were only beginning to influence the microcomputer market for personal and business applications. Although the complexity of the biomass-to-ethanol process had always been recognized it was only with the development of integrated models that the relative importance of the various process steps or parameters and the influence of byproduct production could be readily determined through sensitivity or parametric analysis. These models represent the mathematical equivalent of an integrated physical model such as a development- or commercial-scale plant and they provide, for example, scale-up 38 influences on both the technology and economics of the overall process being evaluated. Linear programming (Ung, 1986) was also attempted and rapidly recognized as being an inappropriate modelling method as much of the process is not linear thus requiring linear approximations and the determination of a large number of coefficients that may be difficult to measure or estimate. A representative example, of the models that were developed over this period, is the SERI - Chem Systems Inc. - Lawrence Berkeley Laboratories Study (1984). This model used a modular structure, was written in a high-level programming language (APL) and accessed the ICARUS equipment capital cost database. It was based on an aspen feedstock and contained the following subprocesses: steam-exploded pretreatment, fed-batch enzyme production using RUT C30 strain of Trichoderma viride, separate enzyme hydrolysis and fermentation and vapor reuse (benzene) distillation. An enzyme recycle option using the Lawrence Berkeley Laboratories countercurrent adsorption research was also included. The simulation program modeled the back-end of the process (fermentation, distillation, waste treatment and heat generation) with little detail. It was proposed by researchers, and later shown by these models, that as a result of their immaturity the process front-end (pretreatment, enzyme production and hydrolysis) is a major contributor to the cost of producing ethanol. A comprehensive kinetic model to describe cellulose conversion for various pretreatments, feedstocks, and hydrolysis conditions was and remains unavailable, and consequently, the parametric analysis did not account for these interactions. Spreadsheet modelling (Wright, 1986; Douglas, 1989) began to appear around the end of the 1980's. Microcomputers and spreadsheet software had become well established in the personal and business markets. The middle to late 1980's was a time period in which there was a tremendous growth and development of both the hardware and software capabilities of microcomputers. Models and other computationally-intensive applications that had previously required the computational and storage capabilities of a larger computer, were now being developed or ported to these smaller stand-alone microcomputers. Spreadsheets became popular as a way to rapidly develop the calculational relationships between process variables, via cell addressing or cell naming conventions, without having to learn a programming language. Model 39 development using spreadsheet software acquired many capabilities that previously had to be programmed such as the ability to format and generate reports and/or charts, the ability to read, illustrate and write the contents of storage files and in a small way internal error checking and debugging by showing intermediate results. The modelling also acquired capabilities that were unique to spreadsheets such as the easy incorporation of supplied spreadsheet functions (interest calculations, look-up tables and random number generators, etc.), the ability to rapidly do sensitivity analysis by changing the value of one cell or variable and noting the change in the final ethanol price, development of simple flat-file databases and the ability to automate common tasks through the development of small programs using macro languages. The IOGEN study (1988) is representative of the type of modelling done at this time. The model was built using Lotus 1-2-3 spreadsheet software and based on an aspen feedstock. Residual solids from the process were assumed to supply all the energy requirements for the plant and selling surplus electricity. Very high enzyme production yields from lactose were assumed. Although flowsheet simulators were available in the mid-1980's, they were only generally accepted as a means to model processes or unit operations and used on a regular basis by process and research engineers in the latter part of the 1980's and early 1990's. Flowsheet simulators, like spreadsheets, provide the user with commands to format and generate reports plus read, illustrate and write the contents of storage files. However, unlike spreadsheets, they are somewhat limited in their capability to produce charts. They utilized a graphical interface and ready-made unit processes and databases/estimation routines for estimating flowstream component properties to build a model and, due to their computationally-intensive nature, often require long periods of time to provide modelling results. The latter feature often makes flowsheet simulators an inappropriate application for sensitivity analyses or Monte Carlo simulations. This type of software, due in large part to its size and computational requirements, was at this time, restricted to main-frame or minicomputers. This hardware constraint along with the cost of the software, lack of economics modelling, and steep learning curve (generally associated with the requirement to learn which of many options to select for the determination of 40 thermodynamic and other chemical properties) tended to limit the use of flowsheet simulators for techno-economic modelling. Furthermore, the commercially available simulators were developed primarily for the design and evaluation of physico-chemical processes and as a result did not provide ready-made unit operations for biological processes or readily characterize biologically-derived materials. Users were forced to develop modules in a high-level programming language for the biological processes which were subsequently linked to the flowsheet simulator kernel and users were also forced to directly assign properties to the biologically-derived materials. Even with these constraints, flowsheet simulators were primarily used for techno-economic modelling during this time. Techno-economic modelling since the late 1980's has been carried out by a few groups using object-oriented high-level languages to build whole applications that include some of the features of flowsheet simulators (von Sivers, 1995). However, most modelling has used flowsheet simulators (Hinman et al., 1991; von Sivers and Zacchi, 1993) or enhanced spreadsheets (Nguyen and Saddler, 1991; Avellar, 1994). There has generally been a convergence of features within these three previously distinct modelling methods, primarily as a result of the proliferation of graphical user interfaces in microcomputers. The graphical user interface has spawned the use of graphical controls and dialog boxes, object-oriented programming, and dynamic linkage of applications and their files. Object-oriented high-level languages such as C++ and Visual Basic have become popular over the period since the late 1980's primarily as a result of their recognized productivity gain for developing graphical user interfaces. At least one research group (von Sivers, 1995) has been using an object-oriented high-level language to develop a stand-alone application that models the technical and economic features of biomass-to-ethanol processes. This application includes a graphical interface and modular structuring similar to flowsheet simulators. However, it also includes economic evaluation, sensitivity and Monte Carlo simulation features. Furthermore, because it is strictly limited to the biomass-to-ethanol processes, the training time should be significantly shorter than an equivalent commercial flowsheet simulator. 41 Flowsheet simulators since the late 1980's, as mentioned previously, have become a valuable design and assessment tool for process engineers. With the growing computational and storage capabilities of microcomputers many of the previously developed commercially-available simulators have been ported to the desktop. It has also been recognized by the developers that simpler, less-full featured programs can reduce the training time and also the computer hardware demands. Some of the developers are also beginning to recognize the need for economic analysis modules, modules that model biological unit operations, and database/calculational routines that estimate biological material properties. The basic principles associated with flowsheet simulators are generally recognized as being good for the rapid development of process scenarios, in that the user only has to select the required unit operations and specify the values for the required number of variables to fully define the solution. However, the current flowsheet simulators are not capable of providing a user, that is both knowledgeable in modelling this particular type of process and reasonably computer-literate, with the ability to easily produce lignocellulosic-to-ethanol process scenarios and assess their economic feasibility. An example of a flowsheet simulation over this time period is the Lund Institute of Technology Study (von Sivers and Zacchi, 1993) that compared three wood-to-ethanol processes (Concentrated Hydrochloric Acid Process (CHAP); Canada, America, Sweden, Hydrolysis (CASH) a two-step weak acid hydrolysis and an enzymatic process using a dilute sulphur dioxide pretreatment). The modelling was done using ASPEN PLUS (Aspen Tech Inc., Cambridge, MA, USA), a commercial flowsheeting program that is normally used for thermochemical process design. All three processes used pine as a feedstock and were based on the same plant capacity. The main product was 95% ethanol (w/v) with the by-product, solid residue and methane used to produce process steam and carbon dioxide sold on the open market. Since the late 1980's spreadsheets have become more graphically oriented with the incorporation of more drawing capabilities. For example, three-dimensional spreadsheet capabilities can be accessed through tabbing of worksheets within a workbook; the inclusion of both graphical controls (including buttons, check boxes, menus, drop-down list menuing, scroll 42 bars, spinners and dialog boxes) are possible and more sophisticated macro languages or full programming languages that allow the development of graphical user interfaces similar to those in flowsheet simulators can be used. Spreadsheets have also become more flexible in that they now have the ability to dynamically link to various other entities (i.e. objects, dialogs, worksheets, workbooks, and applications). Two spreadsheet models (the Forintek Study (Nguyen, 1990) and the Virginia Polytechnical Institute (VPI) Study (Avellar, 1994)) representative of this time period were inherited by our research group and these will be discussed in greater depth in the following Model Development section. 43 4. MODEL DEVELOPMENT 4.1 ASSESSMENT AND COMPARISON OF INHERITED MODELS The initial stages of the thesis work included an extensive review of both the technical research and development efforts since completion of the Forintek and Virginia Polytechnical Institute (VPI) models (with the results being previously discussed in Chapter 2) as well as an in-depth analysis and documentation of both the structuring and computational logic of the inherited past modelling efforts (some of which will be discussed in this chapter and the remainder included as an appendix). The results of this early review process (Table 1), along with a knowledge of the resources required to generate a novel model, indicated that a completely new model was neither required nor a realistic goal. Consequently, the model described later in this thesis section is composed of elements from both the model development efforts at Forintek and VPI. The remainder of this section discusses the differences in the two inherited models from general technical overview, structural, calculational and user interface perspectives. 4.1.1 General Technical Overview In the Forintek Study (1989) aspen was used as the feedstock and cellulase enzyme was produced on-site using steam exploded wood as the substrate. Unlike a number of earlier models, this model included xylose fermentation which increased the total ethanol yield by approximately 30%. The enzyme hydrolysis and fermentation steps were linked, based on combining theoretical yields, into a simultaneous saccharification and fermentation (SSF) configuration. The Virginia Polytechnical Institute (VPI) model was built in 1994 and was based on a conceptual pretreatment of aspenwood to produce various wood byproducts. This model was primarily concerned with the pretreatment and fractionation steps to produce cellulose fibre and lignin. The model was built using a commercial spreadsheet (i.e. Quattro Pro produced at that time by Borland) and it consisted of 13 modules that could be linked to provide various 44 Table 1 Comparison Summary of the VPI and Forintek Models VPI Model Forintek Model PRO CON PRO CON Built-in Modules Scaling of capital investment is based on one parameter, the wood flow rate. Capital investment is based on many scaling parameters No built-in modules, the inputs, calculations and outputs for the individual process steps are intermingled. Component flow streams Capital estimates use only 0.6 as a scaling factor Capital estimates used different scaling factors based on past experience with similar equipment Flowstreams are not broken down into components. Relates pretreatment conditions to extraction efficiencies of hemicellulose and lignin through Ro and S factors Spreadsheet cells are not named Most of the spreadsheet cells are named Does not relate pretreatment conditions to extraction efficiency directly Includes a charge to each module for water usage and waste treatment based on traditional rates. Some calculational formulae access empty spreadsheet cells All calculational formulae access appropriate spreadsheet cells Waste treatment capital cost is likely too small Only a portion of the process is modelled i.e. Pretreatment, Fractionation and Lignin Recovery Full process is modelled Process is primarily set-up for SSF only Sensitivity Analysis has been done on the important variables Sensitivity Analysis has been done on the important variables No Monte Carlo Simulation Monte Carlo Simulation has been done using the variables determined to be important from the Sensitivity Analysis There is no evidence of iteration for water recycling loops. Many or the efficiencies or other factors are hard programmed thus restricting the ease of change There is no evidence of iteration for water recycling loops. Many or the efficiencies or other factors are hard programmed thus restricting the ease of change 45 processing scenarios that were separately illustrated in scenario flowsheets. The processing scenarios each had a copy of the required modules in the scenario spreadsheet i.e. the spreadsheets were not linked through an executive spreadsheet or routine to provide a new scenario. From a general technical overview perspective it was felt that, as the Forintek model simulates the complete process, it should be used as the primary source for development of a new model. However, as discussed previously, certain features (e.g. enzyme production) of the Forintek model were not be used while other features were incorporated in a modified form (e.g. SSF will be changed to reflect the previously discussed SHF process strategy). It was also recognized that the VPI model provided, through some structural features, higher model flexibility. These concepts will be discussed below and in the last section of this chapter, which describes the development of the new (STEAM) model. 4.1.2 Model Structure The biomass-to-ethanol process, with its multiple components and processing steps, is complex. Futhermore, most of the front-end of the process has not been tested in an integrated large-scale facility. The model should therefore be capable of easily including and evaluating a number of process and equipment options. Two structural concepts, modularization and encapsulation, have been shown to enhance both software development in general and the ability to build on past techno-economic modelling efforts. The incorporation of these structural concepts should provide the framework required for a flexible and long-lived model and was one of the major objectives of the work described in this thesis. 4.1.2.1 Modularization A review of the Forintek model indicated that it was organized on the type of activity that was carried out (Figure 6) i.e. Input/Output, Material Balance Calculation, Energy Balance Calculation, Capital Cost Estimation, Operating Cost Estimation, Sensitivity Analysis and Monte Carlo Simulation. As a result the subprocess input/output and calculational routines were 46 CD O <.><J a o ' a , a , < CD 00 CD CD JZ CO C D C a5 T3 o o c CO _c LU CO CO CO E g bo CO o C J C CO CD 0} l < C T LU 3 CO O 2 D _ C o 0 8 CO .1 8 O 13 o ^ CO 3..: c E CO CD CO CD CD 0 1 c5 < CO o C CQ o .2 CO CO _ c -j CD CO 0 CO i i CO CO 0) _ Q CO I— > < CO c CD C O 00 -a o o I intertwined. This structure did not allow the user to rapidly change the process or the process options as the calculations for each process step were spatially dispersed over the spreadsheet area. The VPI model was organized into modules (Figure 7) that represented subprocess or unit operational options. This structuring of the VPI model allowed more rapid changes to the process by the modeller than was provided by the Forintek model, as the calculations for each subprocess or unit operation were located in the same general area on the spreadsheet. The modules appear to have been developed in one spreadsheet and then the appropriate subprocesses for a particular process scenario were copied to another spreadsheet and linked. Although this structure is superior to the Forintek model it still does not give the same potential for rapid process modification as occurs in a flowsheet simulator-type of structure. As will be discussed in a later section, the latest features of spreadsheets allow modularization concepts to be included at a level equivalent to flowsheet simulators. 4.1.2.2 Encapsulation Encapsulation is a term used in computing science to refer to the capability of combining routines that operate on a data structure together with the data structure itself. Spreadsheets, such as those used for the Forintek and VPI models, incorporate this concept by keeping together the data (cell values), data structure (array represented visually by the row-column addresses) and the routines that operate on the data structure (cell formulae). However, spreadsheets at the time that the Forintek and VPI models were built, did not include the encapsulation of the flowsheet that was used to develop or design the overall data or process structure. Object-oriented programming (OOP), which has become popular since the introduction of graphical user-interfaces for microcomputers, incorporates and extends the encapsulation concept through including both the calculational routines that operate on a data structure as well as methods or routines that draw a graphical representation of the object for the user. This extra graphical abstraction allows the user to more easily manipulate the relationships between data structures 48 CJ z3 Cu cr. • r—» -t—1 CJ 'cL D-< CO <u 43' 00 C3 CD cu CO o <: C D _c 0) TJ O CO o T3 O T3 CU ' CO c o o CO CO CO CO E g i n CD _ > c° O 0) rr a o ^ c CO o CD (0 < on o T3 ca O i_ LL c 0 3 E < CO CD £ 1 *- T3 CD O 1 s CO CD EPS S O IS 7 IS « - 2 ^ m 3 .CJ CO o B ffl c CO CO O ° S _o • co O CO CD O) < •Ewe co to .2 CD <? ns Q.O -= O o o CD B *o3 O CU 5 C/3 cd o o CD "o PH KS '3 '5b > CD and/or rapidly create new objects that inherit the characteristics of the parent. Flowsheet simulators, e.g. ASPEN, are an example of a type of program that uses the extended graphical characteristics of OOP programming for the user interface. Flowsheet simulators map a graphical representation of process objects (subprocesses and/or unit operations) and process interrelationships (flowstream) to the associated data structures and routines (calculational and graphical). Current capabilities of spreadsheets allow the techno-economic model developer to include encapsulation concepts to the same extent as flowsheet-simulators i.e. the flowsheet elements, calculational elements and the data structure are encapsulated into objects. The flexibility and potential for both the ease-of-development and ease-of-use associated with this greater encapsulation was recognized and its incorporation in the new model will be discussed later in the thesis (Section 4.2). 4.1.3 Model Calculations The term "modelling" usually refers to the calculational component of techno-economic models as this particular component describes the expected behaviour of the system through mathematical equations. In this case, both the Forintek and VPI models included calculational routines, that interpreted various process and economic parameter values entered by the user, to determine the material and energy balances, and cost equation estimates. Each of the calculational types (Process and Economic Parameters, Material and Energy Balance Equations, Cost Equations, Sensitivity Analysis and Monte Carlo Simulation) will be discussed in greater detail below. 4.1.3.1 Process and Economic Parameters Important process and economic parameters, such as wood input, enzyme yield, enzyme loading, hydrolysis yields, recovery yields, cost of wood, lignin credit, debt costs, etc., formed the input variable table of the Forintek model (Table 2 and Table 3). The VPI model differed only in that the process and economic parameters were located with their associated modules 50 CJ c I c s. •c o Q '•8 o I & , o o o '-5 '•8 o c , CJ •o o o c u \1 J3 M o cj a 6 '-5 '•8 0 1 ca , CJ I J = O IC 'o 4) a o c 'li c eg c o o I X> 5 CJ ca CJ & , P 3 U c o C8 l l co oo O CN co oo O CN co be CO co co co y y 11 y o CN o CN d d y u a CJ 6 • >. N : c CJ loo od od . cs, el > o o u -a J3 CJ 00 OJ | 3 w 00 -o o o & I -a CJ -o _o "B. CJ c o •a CJ oo IT) OO CN CN I CN CN I X 00 00 g § '•8 H c CJ o "s e o o as <fc! CN d CJ y c _o tj 3 T3 O W 00 *c3 e y w CJ Pi ,CJ o -4-t -t-» 3 a s CJ 3 I O §1 CJ CJ 6 ca P C3 CJ CJ -a CJ > o o CJ ot c o ca CJ o a o 3 '•s '•8 o 3 O. c 8 o o in CN "Si 1 < I a • •a CU c S3 O u CN c •c o Q o pq a> o 3 p- i§ ^ PH n u > .5 o u 3 O TO c§ 'c3 & pa u > "o -o 2 a VI O o 3 5 I—( I w 00 M l <U o o 3 Ml 2 "3 u U 2 "3 03 C C3 I < w oo bd oo 2 "o c X3 H C3 u H W X ) 3 60 3 C ,2 s C u 1 § 1 o 1) x: IT) u H 00 c o c 03 II I O I c e _o oa o IS ' C es x: o o I N C W VO o d vo o 2 Q PH N O 00 e o O 00 a. 81 ml >1 o 3 o •a Cl. Q P-, pq U > M 2 ^ m « , o re « Pu PQ U c g l l D PU HH < o II P PH PU <u a "—' c Q D cu & PU to c _tg x Ov < o X M II S S3 rt c II iS 2 * x u > X a 13 T3 (L) u .s 6 0 3 Oi o f -U QJ a o a 03 o 60 C instead of in a central location. In both model cases these parameters could be changed to product variation in the revenue price of ethanol. Several of these parameters had an impact on others. For example, in the Forintek model, high cellulase enzyme loadings increased the cellulose-to-glucose conversion yield. The inter-dependence of the input variables was determined from laboratory data. A Forintek base case was established using the best current technology data reported in the literature. The estimation of the base case price of ethanol helps researchers set targets for various process parameters in order to produce ethanol at a competitive price. 4.1.3.2 Material and Energy Balance Equations The material and energy balances are required to determine the number, size and energy consumption of the equipment shown in the schematic flowsheets. Several process parameters have a significant impact on equipment size and energy consumption. For example, high solid loadings in enzyme production and cellulose hydrolysis steps resulted in a smaller number of fermenters and smaller-sized centrifuges; low lignin content in the extract resulted in higher consumption of caustic and process steam. The major differences between the two inherited models are primarily related to the scope of each of the models. The Forintek model was used to evaluate a complete process whereas the VPI model assessed only the pretreatment, fractionation and lignin recovery steps of the process. Futhermore the Forintek model was most concerned with the cellulose stream whereas the VPI model was structured to assess recovery of all of the components i.e. cellulose, hemicellulose and lignin. As a result the VPI model had a much more complete characterization of component flow than the Forintek model. From this comparison of the two inherited models, it was recognized that certain features of both of the models should be incorporated into the new (STEAM) model. 4.1.3.3 Cost Equations The Forintek model cost estimation was divided into three parts: fixed capital cost, working capital cost and operating costs. Fixed capital cost included purchase cost of equipment, installation cost, engineering fees, and capital cost recovery factor. The capital cost recovery 54 factor takes into account the costs of equity, long-term and short-term debts, taxes, depreciation and inflation. Working capital cost is essentially the return on the working capital; a rate of return on investment of 25% was assumed. Operating costs included labour, raw material and energy costs. The cost calculations were also grouped into the following processing areas: Pretreatment; Fractionation; Enzyme production (Cellulase and Xylanase); Hydrolysis and Fermentation; Ethanol Recovery; and Utilities and Waste Treatment. The VPI model cost estimation consisted of a summary of utility, labour, and capital related charges for all the modules in each simulation, plus expenses such as maintenance, administration, marketing and others. Utility expense was summarized from the individual modules contributing to the specified process scenario. The variables covered under utilities included steam, cooling, electricity, water and sewerage. Steam and cooling requirements were established from the energy balance with assumptions for some heat exchange between modules. Electricity was assumed to be purchased from off-site. Water and sewerage requirements were assumed to be filled by either the local municipal system or from units covered under off-site or unlisted equipment. The rates for all the expenses covered under utilities were variable in the simulations. Capital costs for the VPI model were based on major equipment installation costs. The installed costs were estimated from previously published cost figures (Perry and Chilton, 1973; Ulrich, 1984; Peters and Timmerhaus, 1991). These were scaled by a standard exponent of 6/10ths as per the method of Guthrie outlined in Ulrich (Ulrich, 1984). The Forintek model recognized and included differing scaling exponents for bare module cost estimation. The VPI model used a standard scaling parameter i.e. the feedstock flowrate, whereas the Forintek model used parameters associated with rapid design techniques such as the actual area of the filter based on the flowrate of the stream actually being filtered. Following the VPI bare module cost estimation, a 20% percentage cost was added to the base case for contingency. Percentages were added to the sum of the modules for each process variation for unlisted equipment (10%), engineering fees (6%) and off-site capital (10%). The total fixed capital investment (FCI) is the 55 sum of the modules plus all the added percentages. This is the figure upon which other non-capital related expenses such as maintenance, repair and insurance are based. The VPI model does incorporate some features that are unique and useful in attempting to determine the contribution of each of the process steps towards the overall production cost of ethanol (e.g. the waste treatment charge per process step). However, it was apparent that the model was weak in some other areas such as using unorthodox methods of assessing the influence of capital cost scaling. Another major deficiency was that the VPI model did not review the complete biomass-to-ethanol process. Consequently, in the development of a new model it was felt appropriate to primarily use the Forintek model as the basis for developing the cost equations. 4.1.3.4 Sensitivity Analysis The uncertainty in the values of input data is usually quite high and therefore sensitivity analysis is normally carried out in the economic evaluation of a project. In the Forintek study, sensitivity analysis was used to rank the economic importance of the various input parameters. A base case, which assumed a set of specific values for various technical and economic parameters, was established as a bench mark. The values of these parameters were based on current technology and market conditions. A summary of the production cost components for the base case are presented in Table 3. Sensitivity analysis of important parameters is performed by changing the value of one parameter at a time and calculating the corresponding price of ethanol. The Forintek model showed that the model was most sensitive to the following variables: cost of wood, by-product price of lignin, enzyme productivity, fractionation costs associated with the caustic and process steam inputs, glucose yield, cellulose concentration and cellulase loading, ethanol yield from xylose, and xylose fermentation rate. The VPI model indicated that with regard to the pretreatment, fractionation and product recovery steps the model was most sensitive to the following variables in descending order: scale of the process, capital related costs, cost of wood, total utilities and labor rate. 56 4.1.3.5 Monte Carlo Simulation There are certain risks involved in predicting the input parameters for the model due to reasons such as: the variation in laboratory results; the uncertainty in scale-up from laboratory results to commercial plant operation; and the interaction of the parameters. These risks can be simulated using game theory and techniques such as Monte Carlo simulation. The variation in the values of input parameters is simulated through the use of random numbers. Random numbers can be used to simulate random samples without actually taking a large number of samples. A normal distribution is assumed for the random samples; knowing the means and standard deviations of the input parameters a random normal cumulative frequency distribution for the price of ethanol can be determined. The Forintek model included a Monte Carlo Simulation based on a number of process variables (Table 4) whereas only sensitivity analysis was done using the VPI model. 4.1.4 Model User Interface The user interface, in this study, refers to both the interface used to develop and evaluate various process scenarios as well as the interface for editing the input variables, calculational routines and accessing or reporting the output variables. To develop techno-economic models the first step is generally the development of a series of flowsheets for the overall process and for each of the process steps. This flowsheeting process is an integral part of flowsheet simulators and this feature is often recognized as being a key to their popularity and flexibility. 4.1.4.1 Input/Output and Calculational Routine Entry The model user interface of both models was essentially a standard spreadsheet interface with the user entering the values and formulae for the various variables via the formulae entry bar 57 Table 4 Forintek Model - Monte Carlo Simulation Parameters Parameter Mean Sigma Unit Lignin concentration 0.04 0.0025 wt fraction Lignin credit 0.08 0.01 $/kg Wood cost 60 5 $/OD tonne Cellulase yield 200 5 FPU/g Cellulase loading 20 2.5 FPU/g Glucose yield 85 2.5 % theoretical Residence time 48 4 hour (Cellulose hydrolysis) Substrate loading 0.1 0.005 w/v fraction Ethanol yield from xylose 90 2.5 % theoretical Residence time 2 0.25 day (xylose fermentation) Residence time 8 0.5 day (Enzyme production) 58 in spreadsheet row/column format. Both models did not include a specific error checking routine or formulae security. However, there was a slight difference in the overall interface of the two models. The Forintek designers expected the user or developer to find the appropriate areas for Input/Output and modification of existing calculational routines whereas the VPI model provided, through a structuring that emphasized process steps rather than calculational types, an easier environment to modify. 4.1.4.2 Process Flowsheets In both the models the user was forced to refer to process flowsheets that were not part of the modelling application in order to understand the process details e.g. flowstream numbering. The process flowsheets were hand-drawn in the case of the Forintek model and therefore not in the Lotus 1-2-3 application and the process flowsheets for the VPI model were computer-drawn but not in the Quattro Pro application. 4.1.5 Conclusions The model comparison work indicates that certain elements of the VPI model should be included in the new model i.e. modularization of the process structure for modelling and an attempt to more fully define the component flows through the front-end of the process. However, the procedures for estimating capital costs for the VPI model did not seem to incorporate the estimation principles normally associated with capital costing and it only modelled the pretreatment and fractionation stages. The Forintek model on the other hand models the whole process and therefore was used as the base model for updating with the realization that at some future time the model must include a more detailed component flow. In the sections below the newly developed model, based on the two inherited models, is discussed in detail. 59 4.2 MODIFICATION OF INHERITED MODELS 4.2.1 Model Structure As mentioned previously the structure of the model dictates to a large extent the flexibility of the model as a whole. The model structure should include elements of modularization and encapsulation (discussed in detail below) to ensure the adequate calculational routine and user interface flexibility generally associated with good models. 4.2.1.1 Modularization The latest version (5.0) of Microsoft Excel (Excel) provides the means to easily develop a modular structure that represents various subprocesses and types of calculations within each subprocess (Figure 8). Each of the files is now referred to as a "workbook" as each layer of the three-dimensional file is now called a "worksheet" and can be accessed through a graphical tabbing system analogous to a physical filing cabinet or card file. The graphical structuring allows the designer to easily separate the various calculational or storage requirements, for a particular subprocess, into separate graphical layers. As in previous three-dimensional spreadsheets the "worksheets" within the "workbooks" can be linked either through worksheet-row-cell referencing or through cell naming. A further modular structuring feature of Excel is the ability to link various "workbooks". This provides the developer with the ability to link the various subprocesses into a complete process scenario. Linkages can be fixed, through methods such as worksheet-row-cell referencing or worksheet-cell naming conventions. Alternatively they can be more flexible and dynamic, through development of an "executive workbook" and/or an executive program. Consequently, there is a substantial level of flexibility now available to the model developer for designing and implementing a structure that is modularized and reflects a structure similar to flowsheet simulators or previous high-level language models. 60 CD O 03 Cu o .2 03 O Cu CD CD J3 cc T3 03 CD 5 -Cu CO CD O CO CL CO CD "55 o o c co LU CO CO CO £ o i n CO i 5 2 c o = O a) <C O 1 8-1 2 •£ o J= O ^ Q_ J 2 CU 8 -9 <6 IS § 2-CO m « O « « S — -5 3 -8 g |c3 c < to ffl =0 c ^ o 1 2 CO £ =3 -9 •c « o • B "£ m CD O O 5 5 03 w <* fo « .« w Jo c 2 co o C/3 13 c >•> ao _o o CJ U —' 0 S CO » — 0 t-2 -+^ u bl o tu oc 4.2.1.2 Object-oriented Encapsulation (Flowsheet, Input/Output, Calculation) Through the use of Excel, with its graphical objects (controls, dialogs, toolbars and menus) and a programming language (Visual Basic), it is now possible to include the flowsheet, input/output and calculational relationships of the various subprocesses and unit operations within a spreadsheet format. The graphical objects can now be mapped directly to the data structures and calculational routines associated with the subprocess and/or unit operations. This allows for rapid manipulation of the relationships between these operations and the potential for graphical versus manual linkage of the subprocesses and unit operations to create new scenarios (this latter feature was not added to this model, due to time, although Visual Basic has the implementation capabilities). 4.2.1.3 Development of Routines for Technical Enhancements There has been a great deal of research into the utilization of S0 2 catalysis with steam explosion for pretreating various lignocellulosic feedstocks. Each of the VPI and Forintek models mention the technical benefits of S0 2 catalyzed steam-explosion. However, neither model has addressed the economic costs of implementing the technical benefit of S0 2 usage. The STEAM model has been enhanced to include the technical benefits and attempts have been made to include both the extra capital required to implement the technology and the shorter equipment life associated with the S0 2 corrosion. Equipment selection and addition is discussed further in the modelling results section of the thesis. The steam pretreatment module is contained within a separate workbook file and a more detailed component-flow and cost estimation than the rest of the subprocesses. This was done to indicate both the flexibility of the model structure and its capabilities for further development without the previously recognized limitations and restrictions for structural and calculational capabilities. Previous research (discussed in Appendix 8.1) has indicated that a peroxide wash following a water or alkali wash can greatly enhance the recovery of lignin from steam-exploded woods. The importance of the recovery of lignin has been outlined previously (and discussed in 62 detail in Appendix 8.1) for both maximizing return on the feedstock purchase price and the positive influence lignin extraction has on enzyme recycle during enzymatic hydrolysis. It is generally agreed that, even though peroxide utilization following a water wash can technically be effective, it is recognized that the implementation and operating cost would be excessive. However, our past research on both hardwoods and softwoods has indicated that the levels required for lignin extraction following an alkali wash are still technically effective and could potentially be economically viable. The model has been enhanced with the inclusion of a counter-current peroxide wash that uses equipment similar to that described in the Forintek model. The benefits of enzyme recycling have been discussed previously (and in detail in Appendix 8.2) and the potential cost reduction has been estimated based on lab research. However, the costs of implementing this technology at a commercial-scale have not previously been examined. An enzyme recycling calculational block and the cost equations associated with the implementation of this technology have also been developed. The equipment used to implement enzyme recycle will be discussed further in the modelling results section. 4.2.2 Model User Interface The user and modeller interface has been changed from input/output through spreadsheet areas (Figure 9) to the use of flowsheet objects coupled to input/output dialog boxes (Figure 10). Subprocesses and unit operations or equipment options are now reached through flowsheets of the overall process and subprocesses (Figure 11 & Figure 12). Calculational routines are now accessed through the equipment option dialogs (Figure 13) that lead to programming modules associated with the various calculations (Figure 14). 63 T3 O 2 LU CD "Of "351 CD Q O S o , II ! CD O ) cz cu ai CO CD-Q . CO c u r t z 1 c o Q , O . o ' : o IT). CO l _ i CD 7 . i i O ( T 3 i l - ~ o l - ~ + : C D L U r -CO O ) i r i o l . (/) o o "ro o _ ro o " r o o O I c o O O CD a CD. CD I—' .>« 5 O " o : " O : o i c = £ ; CD' CD CD* £ £ -^1 CD CD "53; r — ' o o o o o + ; m LLP 0 0 i o . i o i r i ! """" CO o ; cc j L U < ICC | CL" co •CO LU: o o QL CL C L =J - E , O TD CD CD , LL CO CO: CO CO CO CO CD LU LU D. CM 0 0 O O PH o <L> o »-t =3 t 0 1 & <D -o c d l - l C v , CO 1 op vo m v o Oil . — CO to CO co c o < < c « = • . § .2 1 ram CU CO CO CD c o CL O £ CO <D i_ -*—• CO o LL 0 Q. >» c .2> .s>8 go. o J I Q> cu CO -C I S o o LL LL CO CO co < c o -I—» CO c CO CO 55 c Q> CD iz Q, .£ CO o3co^ o < o .Q E o -I Tl O c CD c o CL E o O E co <D | 0 O D) " " W a, CO — CD < c t3 o 0 -Q . ID O ( O CO co < *u < O) CO o CD CO C O ^ ? 0 CO .9> CO — - - 1 .52 2 co b 5 J I 3 — vw v_/ J-' v MJ w ~" W Z J L ° r - Q ^ _ _ CO b CO £ O S CD CL >> \-~o c LU yj _ VU t E CD co cp i : g 0 CO o -s r = CO co < E CO 0 I -EL" « CO TJ 0 CO _ 2 co 0 ti. "° N CO 0 >EE (0 CO CO • £ 0 0 O CO CO § § § b o b 0 o c «j CO 99 CD •*-» CO c c i2 o CL E o O 0 .Q E 3 C 0 C o CL E o O o o ^ 0 .Q ^ E c 3 0 c o CL E o O 0 E C 0 c o CL E o O c 0 c o CL s i § ID r 18 is 3 0 rn ca iz g £ C 0 N CO >> E CO CO 0 CO 0 O CO CO 3 E £ E TJ TJ T3 > 0 0 0 .O CC DCDC CO 151® ca OJ .'C CD bp ON 5. MODELLING RESULTS The technical process issues have been the topic of most of the previous sections. However, the implementation of these technologies will have an impact on the economic competiveness of both the process and its final product price. Previous modelling efforts have indicated the major influence feedstock pricing has on the final product price and the available profit margins (Nguyen, 1990). A theoretical conversion calculation (Table 5) can be used to illustrate the tight profit margins associated with this process. For a generic hardwood feedstock an input of 1000 ODT represents a theoretical recovery of 427 tonnes of cellulose, 221 tonnes of hemicellulose sugars and 209 tonnes of lignin. The lignin, if recovered for sales purposes, represents a return of roughly $17/ODT based on a lignin price of $0.08/kg (Nguyen, 1990). The theoretical conversion of the cellulose and hemicellulose to 95% fuel ethanol, using conversion factors of 1.11 g glucose/g cellulose and 0.51 g ethanol/g reducing sugar, represents a return of approximately $115 based on a fuel-competitive price of $0.25/L ethanol (Wyman, 1995). Consequently the theoretical break-even return for a tonne of hardwood is in the range of $115-135/ODT with the lower limit representing sales of ethanol alone and the upper limit representing the combination of ethanol and lignin sales. A similar calculation yields a value of approximately $105-125/ODT for a generic softwood. Pricing for hardwood pulp chips as of November, 1995 was S120/ODT thus representing for a hardwood feedstock a profit margin of $10/ODT for the combined product sales or $5/ODT on ethanol sales alone. The use of softwood feedstocks requires the sale of both ethanol and lignin to attain a profit margin of $5/ODT. Once again these margins are considered to be maxima as they are theoretically based. If the hardwood feedstock is purchased at the $120/ODT price and both products are sold, the maximumum costs and inefficiencies associated with commercially implementing the process must be $5-10/ODT to break-even (Figure 15). 70 Table 5 Theoretical Product Yields and Returns for a Hardwood- and Softwood-to-Fuel Ethanol Process Feedstock Compositional Characteristics Softwood Hardwood Moisture 0.000 0.000 fractional % (wet basis) Cellulose 0.357 0.427 fractional % (dry basis) Hemicellulose 0.278 0.302 fractional % (dry basis) Hexoses 0.178 0.046 fractional % (dry basis) Pentoses 0.054 0.175 fractional % (dry basis) Other 0.046 0.081 fractional % (dry basis) Lignin 0.235 0.209 fractional % (dry basis) Other 0.130 0.062 fractional % (dry basis) Proposed Process Theorectical Conversion Calculations Softwood Hardwood Feedstock 1000 1000 tonnes Cellulose - Hexose 357 427 tonnes Proposed Conversion to Fuel Ethanol Hemicellulose - Hexoses 178 46 tonnes Proposed Conversion to Fuel Ethanol Hemicellulose - Pentoses 54 175 tonnes Proposed Conversion to Fuel Ethanol Lignin 235 209 tonnes Proposed Conversion to Lignin Powder Water 0 0 tonnes Losses or Products? Other 176 143 tonnes Losses or Products? Total 1000 1000 tonnes Theoretical Conversion Calculations Softwood Hardwood Cellulose->Hexose 396 474 tonnes Cellulose->Hexose Conversion Factor 1.11 g glucose/g cellulose Hexoses-> Ethanol 293 265 tonnes Hexose-> Ethanol Conversion Factor 0.51 g ethanoi/g glucose Hexose-> C02 282 255 tonnes Hexose-> C02 Conversion Factor 0.49 g C02/g glucose Pentose-> Ethanol 27 89 tonnes Pentose-> Ethanol Conversion Factor 0.51 g ethanol/g xylose Pentose->C02 26 86 tonnes Pentose->C02 Conversion Factor 0.49 g C02/g xylose Lignin 235 209 tonnes Conversion to Marketing Units Softwood Hardwood Total Ethanol 320 354 tonnes (100%) Total Ethanol 419312 463903 litres (95%) Ethanol Marketing Unit Conversion 0.8042 g/mL@20°C for 95% ethanol Total C02 308 341 tonnes Total C 0 2 156 172 m3 C02 Marketing Unit Conversion 1.9768 g/L @ 0°C and 1 atm Total Lignin 235 209 tonnes Lignin 235000 209000 kg Lignin Marketing Unit Conversion 1000 kg lignin powder/tonne lignin Proposed Process Theoretical Product Yields Ethanol Yield C02 Yield Lignin Yield Softwood 419 0.156 234.935 Hardwood 464 0.172 209.000 l/oven-dry tonne feedstock m3/oven-dry tonne feedstock tonne/oven-dry tonne feedstock Proposed Process Theoretical Economic Returns Softwood Hardwood Ethanol 104.80 115.98 $/oven-dry tonne feedstc Ethanol Price (from Wyman, 1995) 0.25 $/L C02 0.00 0.00 $/oven-dry tonne feedstc C02 Price 0 $/m3 Lignin 18.79 16.72 $/oven-dry tonne feedstc Lignin Price (from past Forintek Model) 0.08 $/kg Total 123.59 132.70 $/oven-dry tonne feedstock 71 Figure 15 Theoretical Breakeven Analysis Maximum Process Implementation Cost vs Feedstock Price (Assumed Prices of $0.25/L & $0.50/L Ethanol and $0.08/kg Lignin) 72 Sawmill residue pricing could represent an attractive option to alleviate the large impact of feedstock costs as, recent sawdust pricing have been in the range of $8-10/BDT. Thus, assuming a suitable residue (i.e. relatively clean wood) and a long-term supply of the low-priced residue is available, sawmill residues could provide the freedom to implement a process that could cost as much as $127/ODT of hardwood producing strictly ethanol or $117/ODT for softwood producing combined sales of ethanol and lignin to break-even. The theoretical maximum, assuming a feedstock cost of SO/ODT, would allow the implementation of a process that costs $135/ODT of hardwood (Figure 15) or $125/ODT of softwood. Subsidization of the feedstock to levels below the $0/ODT feedstock cost have been discussed as a means of reducing the cost of residue disposal. However, if past experience is any indication, it seems unlikely that the producers would enter into long-term agreements that subsidize residue removal. Environmental subsidization programs for the production of ethanol from both sugar-based and starch-based feedstocks are currently in place in both Canada and the U.S.. These subsidies are generally in the $0.20-0.25/L range (Wyman, 1995) and could potentially be long-term as they have been in-place for the last 10 years. The profit margin picture for both feedstocks changes substantially with a $0.25/L ethanol subsidization regime. The ethanol sales alone can provide both high theoretical break-even returns ($232/ODT for hardwood and $210/ODT for softwood) and theoretical profit margins ($112/ODT for hardwood and $90/ODT for softwood) for a tonne of feedstock priced at $120/ODT (Figure 15). Consequently, with subsidization there is substantially more freedom of choice as to the type and cost of process implemented to attain a break-even position or a certain level of profit margin. The same theoretical calculations also indicate that for each $100/ODT the feedstock price increases, the price of 95% fuel-ethanol will increase by $0.25/L. Residue pricing can have rapid and sustained or short-term pricing swings. For example, developments such as an MDF plant that could use a particular residue could increase the price substantially. Another example is 73 the recent shortage of fibre that has driven the price of wood chips from $40/ODT to $120/ODT from 1992-1995. 5.1 PROCESS IMPLEMENTATION COST EFFECT To assess the cost of implementing the envisioned (Figure 22) processes, two preliminary model runs considered to be base cases for each feedstock, were undertaken. The more important process and economic assumptions for these cases are shown in Table 6. The hardwood base case is a revisit of the previous research and the assumptions used in past Forintek modelling efforts i.e. the use of steam pretreatment with no acid catalyst and washing of the pretreated material with water (Nguyen, 1990). As mentioned in the previous chapter the hexose-sugar hydrolysis and fermentation configuration has been changed from SSF to SHF. Operating and capital costs have not been updated in the model and consequently the model results have been determined on a relative percentage contribution of each subprocess to the total process ethanol selling price. The back- or tail-end of the process (Fermentation, Ethanol recovery, Waste treatment and Utilities) is considered by most modellers to be relatively mature, although there is still a substantial requirement for definition and development in certain elements such as pentose fermentation in hardwoods and the handling of inhibitory compounds in the hemicellulose and cellulose hydrolysates. Generally the back-end represents 20-30% of the total production cost (Figure 16 & Figure 17). Although research on the use of an alkali or caustic wash after water-washing the steam-pretreated hardwood residue predates the previous modelling efforts, it was felt appropriate to illustrate the technical and economic effect of adding this wash as a comparison base for its effects on softwoods. The enzyme loading was extremely high (60 FPU/g cellulose) for the hardwood water-wash option and resulted in a significantly higher final ethanol production cost ($1.36/L Ethanol) than those previously reported, with the enzyme cost representing roughly 40% of the final production costs (Figure 16). Although this base cost is high, it was felt that the 74 Table 6 Base Case Model Assumptions Variable Units Hardwood Softwood Feedstock Price Can.$/ODT 30 30 Pretreatment Temperature <C 210 210 Pretreatment Residence Time s 120 150 Catalyst (S02) Concentration Pretreatment Recovery Yield %(w/w) 0 2.5 % of Input Wood 93.2 89.5 CELWI % of Orig. Dry Wt. 57.6 62.6 CELWIA % of Orig. Dry Wt. - -CELWIA/H 20 2 % of Orig. Dry Wt. - -Enzyme Price Can. $/million FPU 12 12 Cellulase Loading FPU/g cellulose 60 16 C6 Hydrolysis Time hr 24 24 C6 Hydrolysis Yield % of theoretical 75 50 C6 Fermentation Time hr 48 48 C6 Fermentation Yield % of theoretical 95 95 C5 Fermentation Time hr 60 -C5 Fermentation Yield % of theoretical 80 -75 1 1 co a> CD rr g ~ _ 5 m O ~ .2 C CD — CO CO " • - B r B n • B • CD CO CO .c o Q_ g c c o o o CO Z3 T3 CO CO co o -t—' c H)—• c >s o CD CD o ferm ferm CD ferm ferm hy< zym LO CD CD c o O O LU c o •4—I o CO >< CD 'c — C CD X E « t . | CL C =5 2> C/> _ i • • • LT c o -»—* o CO 1— X CD CD CO o cn c T3 HW c c CO X £ o o (D +-I CO T3 CD CD CD CL LL • • 1 jsoo O N epAoey eoiAzuB + SOSH/VIM-20S epAoey eaiAzu3 + 202H/V IM-SOS 20SH/V IM-SOS sAonv oi}OX3 - VIM-SOS VIM-20S VIM IM cn 1) '5b _g o o to H — c CU C —! 2 9 CO 2 w > £ 1 £ I CD !> S h2 i—i PH to c2 o ^ O > i 2? PH O O O 00 O CO o o - j ? O r * O }SOQ uojpnpojd l e j o i |0 % CO CO c CO 2 = I •2 a) I ! ~ CO CO ^ £ £ LO D > LU O c o -I—' CO •*-> c CD £ CD CD CD CO CO o 3 a. c o c o o CO x CD c 'c CO 1 - >> to O n ^ O ° - CD 1 •fe. • *~ E >.-E 9-c o o CO c x o ® O CD -~ CQ CO c O CD >< CD I I CD CO O CD E S B a s i l N L L L Q O LU J W J I Q. [3 • Q H • D • ex-c. TJ c CO x o o •»—' CO •CD-CD CD i l l k W i i ' * H i 1 S|sA|OjpAH eiqnoa + SOSH/VIM -SOS isoo O N epAoey awAzug + SOSH/VIM -SOS spAoey eaiAzu^ + SOSH/VIM -SOS SOSH/VIM -SOS sAonv 0110X3 - V I M - S O S VIM-SOS IM-SOS O O o oo O C D O o CM cT-O j s o o uojpnpojd |B}oi }0 % model still reflected the relative influence of each of the technologies on the final ethanol production cost, ethanol yield and relative contribution each of the subprocesses to the final production cost. The hardwood base case also indicated the large contribution (80%) of the process front-end to the final ethanol cost. As mentioned previously, the technologies that have been reviewed in this thesis are the options which have the greatest potential to reduce this large contribution. The softwood base case reflects the major differences envisioned, as described in the previous sections, for processing softwoods (i.e. a mandatory use of an S0 2 catalyst for steam pretreatment subsequently followed by a water wash and enzyme hydrolysis with a enzyme loading of 16 FPU/g cellulose versus 10 FPU/g cellulose used in most of the hardwood modelling cases, and combined fermentation of the hexose sugars from both the cellulose and the water-soluble fraction). A base case cost for producing ethanol from a generic softwood is $1.54/L. With the exception of the enzyme costs the softwood pretreatment, fractionation and hydrolysis steps are 5-10% higher than the hardwood equivalents. 5.2 T E C H N I C A L M E T H O D S O F O F F S E T T I N G I M P L E M E N T A T I O N C O S T S The process technical advances developed over the last five years reflect both a concentration on the steps or inputs that were previously recognized as being most costly (i.e. feedstock, pretreatment and enzyme costs) and a change in process optimization philosophy from recovering mainly cellulose to maximizing the return on the price of the feedstock by recovering all of the components. Therefore, the use of S0 2 as a pretreatment catalyst has overall technical benefits in that it enhances the hemicellulose and lignin recovery and the digestibility of the cellulose. The incorporation of a peroxide wash also has a multi-faceted benefit in that it enhances both the recovery of lignin and concentrates the cellulose stream for less-costly and more efficient hydrolysis. Furthermore, the addition of an enzyme recycle scheme, while only appearing to affect the cellulose stream and just provide economic benefit to the hydrolysis step, may also provide the incentive to go to higher enzyme loadings and potentially produce a more 78 concentrated glucose stream. The more concentrated glucose stream could be combined with the water-soluble stream in softwoods and would therefore provide both technical and economic benefits through the subsequent concentration of the fermentation and ethanol recovery streams of two wood components i.e. cellulose and hemicellulose. This integrated nature of the process and the change in process optimization philosophy to a more comprehensive and multi-component is reflected in both general trends in the modelling results as well as specific results for each technology. The general trends in the modelling results are indicative of the larger design and development philosophy described above. The pretreatment step remains essentially constant in the softwoods as SO2 catalysis is assumed to be a necessity whereas in hardwoods, addition of S0 2 increases the contribution of pretreatment by roughly 30% to the total production cost for ethanol. The development of a more comprehensive fractionation scheme, with the addition of lignin extraction and recovery subprocesses (Figure 16 & Figure 17), becomes a progressively larger component (growing from 2% to 20% in the hardwoods and 2% to 25% in the softwoods) of the total production cost as more of the evaluated technologies are added. The enzyme production/purchase component becomes progressively smaller (Figure 16) in hardwoods and to a lesser extent in softwoods (Figure 17) with the addition of more technologies. Hexose hydrolysis contributes less to the total production cost with the incorporation of more lignin extraction and recovery technologies and increases dramatically with the introduction of enzyme recycling (Figure 16 & Figure 17). The technologies being assessed were developed with hardwoods and for the most part provide a much larger benefit to that feedstock (Figure 18 & Figure 19). 5.2.1 Sulphur Dioxide Catalysis The technical benefits of introducing sulphur dioxide (i.e. enhanced hemicellulose recovery and digestibility of the cellulose) as a pretreatment catalyst have been outlined in detail in Section 8.1. These technical benefits have been modelled with databases that include research results from the 79 past 5 years. Commercial implementation of S0 2 catalysis to pretreatment required the addition of an impregnation tank, catalyst storage tank and catalyst transfer pump. The corrosive nature of S0 2, particularly at the elevated conditions of temperature and pressure associated with steam explosion, also suggested that over the proposed capital depreciation period of 15 years there was a requirement for the equivalent of 2 more steam explosion reactors and the reactors would have to be constructed with stainless steel instead of the more economical carbon steel. Technical benefits accruing from sulphur dioxide usage were substantial (Figure 16 & Figure 18) and easily overcame the additional capital and operating costs required for implementation of this technology for hardwoods. This technology provided an approximately $0.78/L enhancement over the non-S02 WI and $0.98/L over the non-S02 WIA in hardwoods. Enzyme purchase costs were reduced from roughly 30% to 8% of the total production cost (Figure 16). Increases in the capital and operating costs associated with implementing the S0 2 impregnation were partially offset by a lower steam requirement, greater recoveries of both water-soluble sugars and lignin along with a better cellulose digestibility. The higher recoveries of the sugars and lignin were somewhat offset by the increased capital expenditures for larger vessels required in the latter process stages. Softwoods were considered too recalcitrant to fractionate and hydrolyze without the S0 2 so the base case included the use of S02. Furthermore, a non-S02 softwood case could not be found in the literature so the level of enhancement in those feedstocks could not be assessed. As a matter of completeness this work probably should be done. 80 o o o o LO d o o d o LO o o o (|0UBL|}3 "|/$) (IM-30S) 3SBQ eseg SLU OJ aA|}B|9y jsoo uouonpojd |OUBLU3 UJ a6uBu,Q ON S bp The corrosive nature of S02, particularly at the higher levels anticipated with the softwoods suggested a requirement for the use of exotic metals in the construction of the the equipment. Computer runs including the extra capital required to purchase and install zirconium vessels and certain pump elements in the pretreatment step indicate that, for both hardwood and softwood feedstocks, there is a lessening of the return resulting from the implementation of S0 2 catalysis by an estimated $0.07/L for hardwoods (Figure 18) and S0.12/L loss in softwoods in comparison to the S02-WIA case (Figure 19). 5.2.2 Supplementary Lignin Extraction - Peroxide Wash The use of a peroxide wash to supplement the recovery of lignin and reduce the amount of lignin to be handled in the hydrolysis step was implemented using the same counter-current equipment as that used in the water and alkali washes. The introduction of this technology follows a pattern of major gains ($1.32/L and 105L/ODT) for the softwood and only a minor gain ($0.06/L and 26L/ODT) for the hardwood. As discussed previously, this difference in response is believed to be a result of the differing chemical nature of the lignins associated with these two feedstocks. 5.2.3 Enzyme Recycle Enzyme recycle research has been ongoing for 10-15 years and various methods have been proposed to implement the technology (SERI-Chem Systems Ltd. and Lawrence Berkeley Labs, 1984; Tan et al., 1986; Eklund et al., 1992). For this particular modelling series it was assumed that a fed-batch counter-current contact scheme comprised of three tanks and associated transfer pumps would be used to provide the direct transfer of enzymes from hydrolyzed to pretreated material. Enzyme recycling greatly reduces the proportion of total production cost attributed to enzymes (80% in hardwoods (Figure 16) and 50% in softwoods (Figure 17)). However, the net reduction in the ethanol production cost was only attained when the estimated costs of implementing the technology were waived. Upon further study (including sensitivity runs based 83 on a changing enzyme price), this somewhat surprising result, seemed to indicate the effectiveness of the proposed pretreatment and fractionation schemes to reduce the requirement for high enzyme levels and may suggest that even for SSF the incorporation of these front-end technologies could provide a substantial economic benefit. Glucose yields from uncatalysed steam-treated hardwoods of 80% required enzyme loadings of 60 FPU/g cellulose for WI substrates whereas with S0 2-WI/H 202 substrates this level of yield could be attained with lower loadings of 10 FPU/g cellulose. With this reduction in the requirement for high enzyme loadings other economic factors begin to dominate the implementation cost. In this case, the proposed method for implementing the enzyme recycle appeared to be both capital and energy intensive (electrical pumping costs). Other proposed implementation methods, including a single contact vessel or continuous desorption-adsorption reactor, could potentially lower the capital portion of the implementation cost. However, it appears that the operating costs associated with fluid pumping overshadow the reduction in enzyme costs resulting from the enzyme recycling scheme. Incorporating the implementation costs for enzyme recycling reduced the production cost reduction from the S0 2-WIA/H 20 2 case by $0.16-0.19/L in both feedstocks. The net effect for the two feedstocks varied, with the softwood recycling results straddling the no-net gain level, and the hardwood recycle implementation lessening the price gains resulting from the S0 2 and H 2 0 2 process additions. Implementation costs for enzyme recycle must be lower than $0.15/L in both feedstocks to provide a beneficial reduction in ethanol production cost. Therefore the enzyme recycle option still requires further work, particularly in assessing the true costs of implementing and operating enzyme recycle. 5.2.4 Softwood Doubling Hydrolysis Time A final process scenario for softwoods included the doubling of the hydrolysis time to bring the glucose yield in line with the 80% value for hardwoods. This operating strategy improves the relative cost of the produced ethanol by an extra $0.17/L over the S0 2-WIA/H 20 2 case. 84 5.3 CONCLUSIONS AND PROJECTED FUTURE PROSPECTS FOR LIGNOCELLULOSIC-TO-ETHANOL PROCESSES All of the assessed technologies, with the exception of enzyme recycle, substantially reduced the production cost of the ethanol in relation to the base cases. As a means of determining the level of research success and the current status of process maturity the ethanol pricing generated for each process scenario was applied to a theoretical yield of ethanol and plotted against the previously determined maximum implementation costs (Figure 20). The introduction of the front-end technologies have reduced the implementation cost by roughly $200 on a ODT basis for hardwoods and contributed marginally to lowering the implementation cost for softwoods. A further gain, equivalent to that previously obtained when using hardwoods, is required to lower the production cost to the subsidized market level of $0.50/L ethanol. It appears that the relative contribution of each process step has almost attained an equal level. This suggests that advances will likely only come from incremental advances in each process step and/or a combination of a number of steps i.e., process integration, to reduce the production cost to marketable levels. The enzyme based bioconversion process is in its infancy and current research, such as the recent development of genetically-enhanced yeasts and bacteria that provide cofermentation of hexoses and pentoses, still holds promise of major technical and economic advances. Process integration and larger-scale testing of the technologies at the pilot-scale is currently on-going and proponents (Wyman and Hinman, 1990) suggest that, with aggressive research and development efforts, the price of lignocellulosic-derived ethanol could reach the unsubsidized levels of $0.25/L by the first decade of the next century. Sugar-based, starch-based and acid-hydrolysis processes have all been commercially-developed and proven technically under conditions normally deemed to be non-economic i.e., their production cost was above competing processes and/or non-profitable. This suggests that economic market conditions and economic indicators, such as the production cost of the product, should not always be the only measure of maturity and competitiveness of a process. The 85 1-2O0 1 000 800 600 400 • • A X •Hardwood Ethanol($0.50/L)+Lignin Sales "Softwood Ethanol($0.50/L)+Lignin Sales Hardwood Wl Hardwood WIA Hardwood S02-WIA Hardwood S02-WIA/H202 Hardwood Enzyme Recycle Softwood S02-WI Softwood S02-WIA Softwood S02-WIA/H202 Softwood Enzyme Recycle A Feedstock Price (Can. $/ODT) Figure 20 Theoretical & Modelling Results Breakeven Analysis Theoretical Maximum Process Implementation Cost vs Feedstock Price (Assumed Prices of $0.50/L Ethanol and $0.08/kg Lignin) 86 research and development efforts over the past few decades associated with the biomass-to-ethanol process have not been driven by strictly economic or market incentives. Initially, as mentioned in the introductory sections of this thesis, the escalating prices of petroleum in the early 1970's were the driving force behind the interest in biomass-to-fuel ethanol research. While outwardly seeming to be strictly driven by economics it must also be recognized that national interests associated with foreign debt and control of a strategic resource such as transportation fuel (both civilian and military) were also important. Similarly, over the last 10-15 years research in this area has been primarily driven by a concern over the effect greenhouse gases have on the environment. Economic and financial institutions are beginning to study and incorporate externalities, such as the environmental and health consequences of the continued utilization of gasoline as a transportation fuel, in financial and legislative policy. Research to-date suggests that the detrimental externalities are not included in the determination of the $0.25/L marketable price (Wright, 1995) and thus do not truly reflect the cost to society of producing and using the current fuel. Furthermore, the methods currently utilized to assess the competitiveness of the biomass-to-ethanol process i.e., techno-economic modelling, produce an indicator that does not include the beneficial externalities associated with fuel ethanol. Consequently, techno-economic modelling (indicating a production cost substantially higher than the $0.25/L ethanol) does not truly reflect the competitive status of this technology and should only be used, in this manner, for processes strictly governed by economic market factors. It is likely the competitive market price for producing fuel ethanol is somewhere between the $0.50/L (a price that is suggested to include at least some of the external beneficial externalities associated with fuel ethanol) and the modelling production price of roughly $1.00/L for hardwoods and $1.30/L for softwoods. 87 6. CONCLUSIONS 6.1 LEVEL OF PROCESS MATURITY A truly "generic" enzymatically-based biomass-to-ethanol process has been difficult to identify because of the heavy influence that the type of feedstock, type of by-products, and the number of unproven processes and equipment currently available would have on the design of such a process. Due to the complexity and early stage of development of this bioconversion process there have been various processing strategies and conceptual designs presented in the literature. However, it is generally acknowledged, that a generic enzymatic-based process would include the following steps: feedstock handling, pretreatment, fractionation, enzyme production, enzyme hydrolysis, fermentation, ethanol and other by-product recovery, and waste treatment. All of the technology associated with feedstock handling is commercially available and used extensively by the pulp and paper sector. Although companies such as Stake technology currently sell equipment for the steam pretreatment of wood and much of the equipment to carry out effective fractionation of steam-treated biomass is available, these components should still be considered to be in the development stage and require demonstration at the pilot plant level. The majority of hemicellulose hydrolysis occurs during the acid catalyzed steam pretreatment of the biomass and a separate enzymatic hydrolysis step is not required. The low specific activity for the cellulase system means that large quantities of enzyme are required for cellulose hydrolysis and enzyme recycling should be considered to offset the high operating costs. Cellulases are being used in increasing amounts for various food, feed, textile, detergent and pulp applications and consequently it is anticipated that, for the piloting and first commercial plants, the enzymes will be purchased and delivered to the plant from commercial sources rather than producing them on-site. Using the currently proposed SSF strategy it will likely be necessary to produce new yeast cell mass and enzymes for each batch because of the difficulty of separating the cells and enzymes from the unhydrolysed solid residue. The capital and productivity advantages attributed to the SSF configuration may be rapidly offset by both 88 the increased need for enzyme and yeast that will be required as a result of only using these inputs once. The high degree of uncertainty in the commercial operation of an SSF process encouraged us to concentrate on the SHF approach as there is well established data for the glucose-to-ethanol step. The technology and equipment required for glucose fermentation in a wood-to-ethanol facility is believed to be similar to that found in the sugar- (sugar cane) and starch- (corn and wheat) -to-ethanol processes. These latter two processes are both viable at the commercial-scale. Although significant advances have been made in both our understanding and modification of pentose fermenting microorganisms none of the wild-type, adapted or engineered strains have been able to routinely ferment the pentose-rich, water soluble stream obtained from steam exploded hardwoods. Even if efficient fermentation of xylose to ethanol can be obtained, the presence of most of the wood degradation products in this fraction means that effective methods of inhibitor removal are required. Thus it is likely that separate fermentation and inhibitor removal facilities will probably be required for any bioconversion process with a substantial pentose sugar component in the feedstock. Process development of enzymatic hydrolysis of lignocellulosic sugars is not as advanced as that for the sugar-based, starch-based, or acid-hydrolysis lignocellulosic -based ethanol producing processes, therefore knowledge of the characteristics of the wastestreams produced is less specific and further research and process development is required for clarification. 6.2 TECHNICAL DETAILS OF PROCESS FRONT-END Integrated bioconversion processes should be designed with the capacity to process a variety of feedstocks i.e. a "generic" process. To ensure maximum substrate utilization process optimization should be based on the utilization of all three of the main lignocellulosic components. With hemicellulose being the most labile of the three main lignocellulosic components the pretreatment should ensure maximum hemicellulose solubilization and sugar recovery during the pretreatment step, as well as enhanced cellulose hydrolysis. Similarly 89 pretreatment and fractionation should be optimized to produce the highest degree of lignin recovery for the least amount of solvent and water usage. It is currently technically possible, through the use of S02-catalyzed steam pretreatment and a fractionation sequence consisting of water-alkali-peroxide wash steps to effectively process both hardwoods and softwoods, although the catalyst, temperature/pressure and residence time varies depending on the nature of the substrate. Steam pretreatment of hardwoods using SO z catalysis, can recover more than 80% of the hemicellulose derived sugars as monomers, more than 90% of the lignin by alkali washing, while complete cellulose hydrolysis at high substrate concentrations and low enzyme loadings can be achieved in a relatively short period of time. Using the same process with softwoods, it is currently possible to recover 65% of the hemicellulose derived sugars as monomers, 80% of the lignin by a combination of both alkali and peroxide washing and complete hydrolysis at high substrate concentrations (Gregg and Saddler, 1995). Although pretreatment and recovery yields have been optimized, certain technical and economic aspects still require refinement. For example, the determination of the inhibitory products associated with the hemicellulose rich water soluble stream, their distribution in the various process streams and cost-effective methods to alleviate their effects on fermentation. Furthermore, peroxide washing, required to achieve effective hydrolysis of the softwood derived cellulosic stream, has generally been viewed in the past as being too costly. However, the high cost of enzymes implies that substrates containing low levels of lignin are required if strategies such as enzyme recycle or simultaneous saccharification and fermentation are to be used at a commercial scale. Thus maximum hemicellulose and lignin recoveries are required to optimize utilization of these streams, enhance cellulose hydrolysis and reduce the cost of producing ethanol from a range of lignocellulosic substrates. The pentose nature of the hardwood hemicellulose stream required a different fermentation strategy to that required for the predominantly hexose softwood-derived hemicellulose sugars. However, a major difference with the largest process impact was the limitation on the times that the cellulases associated with the residual softwood-derived substrate 90 could be recycled. The cellulases associated with the softwood residual substrate could only be recycled two times before the limited hydrolytic action and the predominance of residual lignin restricted the efficiency of hydrolysis. The economic benefits associated with enzyme recycle such as, increased rates and yields of hydrolysis, lower enzyme costs and decreased capital costs due to shorter residence times should encourage further evaluation at the pilot-plant level. Issues such as the need to add supernatant (3-glucosidase after each recycle stage and the costs or benefits associated with utilizing or disposing of the lignified residues are currently being evaluated. 6.3 DEFINITION AND CONSTRUCTION OF TECHNO-ECONOMIC MODEL The biomass-to-ethanol process, with its multiple components and processing steps, is complex. Futhermore, most of the front-end of the process has not been tested in an integrated large-scale facility. The structure of the model dictates to a large extent the flexibility of the model as a whole. Any techno-economic model of this process should therefore be capable of easily including and evaluating a number of process and equipment options. Two structural concepts, modularization and encapsulation, have been shown to enhance both software development in general and the ability to build on past techno-economic modelling efforts. The model structure should include elements of modularization and encapsulation to ensure the adequate calculational routine and user interface flexibility generally associated with good models. The incorporation of these structural concepts has provided the framework required for a flexible and long-lived model. Further work should be done in a number of areas including developing a more detailed component flow; a further development of the graphical interface particularly the process builder section and the development of a more flexible economic assessment package. Development of a more detailed component flow, such as evident in the VPI model, is important as many of the technical issues (SO2 incorporation, inhibitor location and concentrations, process closure, waste treatment) now require a detailed knowledge of the individual streams. Detailed component flow can be implemented reasonably easily at the model level as is evident in flowsheeting 91 applications. However, there is still a tremendous absence of knowledge regarding the various physicochemical properties of biomass and its components. Further development of the graphical interface to allow graphical linkage of the process steps, through a combination of drawing the flowstreams and bringing up dialog boxes that allow the connection of the appropriate input/output flowstreams, will enhance the speed of model modification and reduce the potential for linkage errors. Economic flexibility should also be developed to allow for evaluation and comparison of projects world-wide. 6.4 I M P A C T OF R E C E N T R E S E A R C H O N PROCESS M A T U R I T Y The general trends in the modelling results are indicative of the larger design and new development philosophy (concentrating on recovering all three of the main wood components instead of just cellulose). The pretreatment step remains essentially constant in the softwoods as S O 2 catalysis is assumed to be a necessity whereas in hardwoods addition of S O 2 increases the contribution of the pretreatment step to the overall process by roughly 30%. Development of a more comprehensive fractionation scheme, with the addition of lignin extraction and recovery subprocesses, becomes a progressively larger component (growing from 2% to 20% in the hardwoods and 2% to 25% in the softwoods) of the total production cost as more of the evaluated technologies are added. The enzyme production/purchase component becomes progressively smaller in hardwoods and to a lesser extent in softwoods with the addition of more technologies. Hexose hydrolysis contributes less to the total production cost with the incorporation of more lignin extraction and recovery technologies and increases dramatically with the introduction of enzyme recycling. The assessed technologies were developed for hardwoods and generally provide a much larger benefit to that feedstock . Technical benefits accruing from sulphur dioxide usage were substantial and easily overcame the additional capital and operating costs required for implementation of this technology for hardwoods. This technology provided an approximately S0 .78 /L enhancement over the n o n - S 0 2 WI and $0.98/L over the n o n - S 0 2 WIA in hardwoods. Enzyme purchase costs 92 were reduced from roughly 30% to 8% of the-total production cost. Increases in the capital and operating costs associated with implementing the SO2 impregnation were partially offset by a lower steam requirement, greater recoveries of both water-soluble sugars and lignin along with a better cellulose digestibility. The higher recoveries of the sugars and lignin were somewhat offset by the increased capital expenditures for larger vessels required in the latter process stages. The introduction of supplementary lignin recovery via a peroxide wash followed a pattern of major gains ($1.32/L and 105L/ODT) for the softwood and only a minor gain ($0.06/L and 26L/ODT) for the hardwood. Enzyme recycling greatly reduced the proportion of total production cost attributed to enzymes (80% in hardwoods and 50% in softwoods ). However, the net reduction in the ethanol production cost was only attained when the estimated costs of implementing the technology were waived. Further study indicated that the effectiveness of the proposed pretreatment and fractionation schemes to reduce the requirement for high enzyme levels to the extent that other costs began to dominate the subprocess cost. This suggests that, even for SSF, the incorporation of these front-end technologies i.e., SO2 catalysis and fractionation including a peroxide wash, could provide a substantial economic benefit. Implementation costs for enzyme recycle must be lower than $0.15/L in both feedstocks to provide a beneficial reduction in ethanol production cost. Therefore the enzyme recycle option still requires further work, particularly in assessing the true costs of implementing and operating enzyme recycle. The doubling of hydrolysis time in softwoods to bring the glucose yield in line with the 80% value for hardwoods improved the relative cost of the produced ethanol by an extra $0.17/L over the SO2-WIA/H2O2 case. It appears that the relative contribution of each process step has almost attained an equivalent level. This suggests that advances will likely only come from incremental advances in each process step and/or a combination of a number of steps i.e., process integration, to reduce the production cost to marketable levels. This process is in its infancy and research and further process development still hold promise of major technical and economic advances. 93 In conclusion, the development of this model required the reassessment of the level of maturity of both the individual and integrated steps of an enzyme based biomass-to-ethanol process. The model provided an excellent way of assessing the cost of the individual and integrated steps and the effect of various feedstocks, process options and technical advances on the technical and economic status of the process. 94 7. REFERENCES Arthur D. Little Inc. 1985. Technical and economic feasibility of enzymatic hydrolysis for ethanol production from wood. New York State Energy Research and Development Authority (NYSERDA) and Solar Energy Research Institute (SERI). Avellar, B. K. 1994. Engineering and economic considerations for fractionation of steam-exploded biomass. M.Sc. Chemical Engineering. Virginia Polytechnical Institute. Baker, A. J., Millet, M. A., Sattler, L. D. 1975. Wood and wood-based residues in animal feeds. Cellulose Technology Research, , ACS Symposium Series. Banerjee, N. R., Bhatnagar, R., Viswanathan, L. 1981. Development of resistance in Saccharomyces cerevisiae against inhibitory effects of browning reaction products. Enzyme Microb. Technol. 3: 24-28. Baugh, K. D., Levy, J. A., McCarty, P. L. 1988. Thermochemical pretreatment of lignocellulose to enhance methane fermentation:II. evaluation and application of pretreatment model. Biotechnol. Bioeng. 31: 62-70. Beck, M. J. 1993. Fermentation of Pentoses from Wood Hydrolysates, in Bioconversion of Forest and Agricurltural Plant Residues, J. N. Saddler, ed., CAB International, Wallingford, UK: pp. 211-229. Beltrame, P. L., Carniti, P., Visciglio, A., Focher, B., Marzetti, A. 1992. Fractionation and bioconversion of steam-exploded wheat straw. Biores. Technol. 39: 165-171. Bertran, M. S., Dale, B. E. 1985. Enzymatic hydrolysis and recrystallization behavior of initially amorphous cellulose. Biotechnol. Bioeng. 27: 177-181. Breuil, C , Chan, M., Gilbert, M., Saddler, J. N. 1992. Influence of p-glucosidase on the filter paper activity and hydrolysis of lignocellulosic substrates. Biores. Technol. 39: 139-142. Breuil, C , Chan, M., Saddler, J. N. 1990. Comparison of the hydrolytic activity of commercial cellulase preparations. Appl. Microbiol. Biotechnol. 34: 31-35. Brownell, H. H. 1987. Steam pretreatment of wood in relation to enzymatic hydrolysis. Forintek Canada. Final Report Report Number DSS Contract 05SR.31926-4-5022. Brownell, H. H. 1989. Steam pretreatment of wood. Forintek Canada Corp.,. Progress Report Number 04-53-12-402. Brownell, H. H., Mes-Hartree, M., Saddler, J. N. 1986a. Steam pretreatment of aspenwood for enhanced enzymatic hydrolysis, in Biotechnology and Renewable Energy, M. Moo-Young, S. Hasnain and J. Lamptey, eds., Elsevier, New York: pp. 36-45. 95 Brownell, H . H. , Saddler, J. N . 1984. Steam-explosion pretreatment for enzymatic hydrolysis. Biotechnol. Bioeng. 14: 55-68. Brownell, H . H. , Saddler, J. N. 1987. Steam pretreatment of lignocellulosic material for enhanced enzymatic hydrolysis. Biotechnol. Bioeng. 29: 228-235. Brownell, H . H. , Yu, E. K. C , Saddler, J. N . 1986b. Steam-explosion pretreatment of wood: effect of chip size, acid, moisture content and pressure drop. Biotechnol. Bioeng. 28: 792-801. Buchert, J. K. , Niemela, K. , Puis, J., Poutanen, K . 1990. Improvement in the fermentability of steamed hemicellulose hydrolysate by ion exclusion. Process Biochemistry. 25: 176-180. Caulfield, D. F., Moore, W. E. 1974. Effect of varying crystallinity of cellulose on enzymic hydrolysis. Wood Science. 6: 375-379. Chernaglazov, V . M . , Ermolova, O. V. , Klyosov, A . A . 1988. Adsorption of high purity endo-l-4-(3-glucanases from Trichoderma reesei on components of lignocellulosic materials: cellulose, lignin and xylan. Enzyme Microb. Technol. 10: 503-507. Chum, H . L. , Johnson, D. K. , Black, S. K. , Baker, J., Grohmann, K. , Sarkanen, K. V., Wallace, K. , Schroeder, H . A . 1988. Organosolv pretreatment for enzymatic hydrolysis of poplar. I. Enzyme hydrolysis of cellulosic residues. Biotechnol. Bioeng. 31: 643-649. Clark, T. A. , Mackie, K . L . 1984. Fermentation inhibitors in wood hydrolysates derived from softwood Pinus radiata. Chem. Tech. and Biotechnol. 34B: 101-110. Clark, T. A. , Mackie, K . L . 1987. Steam explosion of the softwood Pinus radiata with sulphur dioxide addition. I. Process optimisation. J. Wood. Chem. Technol. 7: 373-403. Clesceri, L . S., Sinitsyn, A . P., Saunders, A . M . , Bungay, H . R. 1985. Recycle of the cellulase-enzyme complex after hydrolysis of steam-exploded wood. Appl. Biochem. Biotechnol. 11: 433-443. Converse, A . O., Ooshima, H. , Burns, D. S. 1990. Kinetics of enzymatic hydrolysis of lignocellulosic materials based on surface area of cellulose accessible to enzyme and enzyme adsorption on lignin and cellulose. Appl. Biochem. Biotechnol. 24/25: 67-73. Dekker, R. F., Karageorge, H. , Wallis, A . F. A . 1987. Pretreatment of hardwood {Eucalyptus regnans) sawdust by autohydrolysis explosion and its saccharification by Trichoderma cellulases. Biocatalysis. 1: 45-59. Dekker, R. F. H . 1989. Enzymatic hydrolysis of plant polysaccharides: substrates for fermentation. Brazilian J. Jed. Biol. Res. 22: 1441-1456. 96 Dekker, R. F. H. , Wallis, A . F. A . 1983. Enzymic saccharification of sugarcane bagasse pretreated by autohydrolysis-steam explosion. Biotechnol. Bioeng. 25: 3027-3048. Deshpande, M . V. , Eriksson, K . -E . 1984. Reutilization of enzymes for saccharification of lignocellulosic materials. Enzyme Microb. Technol. 6: 338-340. Douglas, L. J. 1989. A technical and economic evaluation of wood conversion processes. Energy, Mines and Resources Canada. Final Report Number DSS Contract File #23283-8-6091. 1989. Eklund, R. 1994. Ethanol from wood - an experimental study of steam pretreatment, enzymatic hydrolysis and enzyme recovery. PhD. thesis. Chemical Engineering 1. Lund Institute of Technology. Eklund, R., Galbe, M . , Zacchi, G. 1988. Improved xylose yield from willow by employment of two-stage steam pretreatment. Int. Symp. Alcohol Fuels, VIII, Tokyo, . Eklund, R., Galbe, M . , Zacchi, G. 1990. Optimization of temperature and enzyme concentration in the enzymatic saccharification of steam-pretreated willow. Enzyme Microb. Technol. 12: 225-228. Eklund, R., Galbe, M . , Zacchi, G. 1992. Recovery of cellulases after hydrolysis by adsorption on steam-pretreated willow. Appl. Biochem. Biotechnol. 34/35: 105-113. Eklund, R., Galbe, M . , Zacchi, G. 1995. The influence of S O 2 and H 2 S O 4 impregnation of willow prior to steam pretreatment. Bioresource Engineering. 52: 225-229. Excoffier, G., Toussaint, B., Vignon, M . R. 1991. Saccharification of steam exploded poplar wood. Biotechnol. Bioeng. 38: 1308-1317. Fan, L . T., Gharpuray, M . M . , Y.-H.Lee 1987. Cellulose hydrolysis, Springer-Verlag, Berlin. Fan, L. T., Lee, Y . - H . , Beardmore, D. H . 1980. Mechanisms of the enzymatic hydrolysis of cellulose: effects of major structural features of cellulose on enzymatic hydrolysis. Biotechnol. Bioeng. 22: 177-199. Fan, L . T., Lee, Y . - H . , Beardmore, D. H. 1981. The influence of major structural features of cellulose on rate of enzymatic hydrolysis. Biotechnol. Bioeng. 23: 419-424. Fein, J. E., Tallim, S. R., Lawford, G. R. 1984. Evaluation of D-xylose fermenting yeasts for utilization of a wood-derived hemicellulose hydrolysate. Canadian Journal of Microbiology. 30: 683-690. Foody, F. 1980. Optimization of steam explosion pretreatment. U.S. Department of Energy Fuels from Biomass Program. Final Report Number DE-AX-02-79-ETZ3050. 97 Forintek 1985. Pretreatment methods for enhancing conversion of lignocellulosic material to liquid fuel. Forintek Canada Number ENFOR C-299, DSS File Number - 54SS23216-3-6364. Frazer, F. R., McCaskey, T. A. 1989. Wood hydrolyzate treatments for improved fermentation of wood sugars to 2,3-butanediol. Biomass. 18: 31-42. Frings, R. M., Coombs, J. 1992. Wastewaters from the bioconversion of biomass - utilisation and treatment - literature review and problem analysis. International Energy Agency. IEA/BE T7/A7 Report Number 4. 1992/04. Galbe, M. 1994. Ethanol from wood - an experimental study of pretreatment and hydrolysis process simulation. PhD. thesis. Department of Chemical Engineering I. Lund Institute of Technology. Galbe, M., Zacchi, G. 1991. Simulation of ethanol production processes based on enzymatic hydrolysis of lignocellulosic materials using ASPEN PLUS. Appl. Biochem. Biotechnol. 34/35: 93. Galbe, M., Zacchi, G. 1993. Simulation of processes for conversion of lignocellulosics, in Bioconversion of forest and agricultural residues, J. N. Saddler, ed., CAB International, Oxford, U.K.: p 291. Gharpuray, M. M., Lee, Y.-H., Fan, L. T. 1983. Structural modification of lignocellulosics by pretreatments to enhance enzymatic hydrolysis. Biotechnol. Bioeng. 25: 157-172. Ghose, T. K., Bisaria, V. S. 1979. Studies on the mechanism of enzymatic hydrolysis of cellulosic substances. Biotechnol. Bioeng. 21: 131-146. Girard, D. J., Converse, A. O. 1993. Recovery of cellulase from lignaceous hydrolysis residue. Appl. Biochem. Biotechnol. 39/40: 521-533. Gregg, D., Saddler, J. N. 1995. A techno-economic assessment of the pretreatment and fractionation steps of a biomass-to-ethanol process. Appl. Biochem. Biotechnol. 57/58: 711-727. Grethlein, H. E. 1985. The effect of pore size distribution on the rate of enzymatic hydrolysis of cellulosic substrates. Bio/Technology. 3: 155-160. Grohmann, K., Torget, R., Himmel, M. 1985. Optimization of dilute acid pretreatment of biomass. Biotechnol. Bioeng. 15: 59-80. Grous, W. R., Converse, A. O., Grethlein, H. E. 1988. Effect of steam explosion on pore size and enzyme hydrolysis of poplar. Enzyme Microb. Technol. 8(5): 274-280. 98 H. A . Simons Ltd. 1991. Towards improved utilization of forest industry mill residues in british Columbia. The Ministry of Forests Number File No. 200-01. Hahn-Hagerdal, B. , Hallbonn, J., Jeppson, H. , Olsson, L. , Skoog, K . , Walfridsson, M . 1993. Pentose fermentation to alcohol, in Bioconversion of Forest and Agricultural Plant Residues, J. N . Saddler, ed., C A B International, Wallingford, U K : pp. 231-290. Hinman, N . D., Schell, D. J., Riley, C. J., Bergeron, P. W., Walter, P. J. 1992. Preliminary estimate of the cost of ethanol production for SSF technology. Appl. Biochem. Biotech. 34/35: 639-649. Hinman, N . D., Schell, D. J., Ripley, C. J., Bergeron, P. W., Walter, P. J. 1991. Preliminary estimate of the cost of ethanol production for SSF technology. N R E L for U.S. Department of Energy. Ethanol Annual Report F Y 1990 Report. 1991/01. Holtzapple, M . , Cognata, M . , Shu, Y. , Hendrickson, C. 1990. Inhibition of Trichoderma reesei cellulase by sugars and solvents. Biotechnol. Bioeng. 36: 275-287. IEA 1992. A comparison of the production of methanol and ethanol from biomass. Energy, Mines & Resources Canada Number Contract No: 23218-1-9201/01-SQ. Ingram, L. O., Conway, T., Clark, D. P., Sewell, G. W., Preston, J. F. 1987. Genetic engineering of ethanol production in Escherichia coli. Appl. Environ. Microbiol. 53: 2420-2425. Iogen Corporation 1990. Kinetic study of enzymatic hydrolysis of exploded wood. Efficiency and Alternative Energy Technology Branch, Energy Mines and Resources Canada Number DSS Contract No. 51SZ.23283-7-6031. Ishihara, M . , Uemura, S., Hayashi, N . , Shimizu, K . 1991. Semicontinuous enzymatic hydrolysis of lignocelluloses. Biotechnol. Bioeng. 37: 948-954. Klyosov, A . A . 1986. Enzymatic conversion of cellulosic materials to sugar and alcohol - the technology and its implications. Appl. Biochem. Biotechnol. 12: 260-270. Ladish, M . R., Lin, K . W., Voloch, M . , Tsao, G. T. 1983. Process considerations in the enzymatic hydrolysis of biomass. Enzyme Microbiol. Technol. 5: 8-160. Lange, N . K . 1993. Application of cellulases in the textile industry. Second TRICEL symposium on Trichoderma reesei cellulases and other hydrolases, Espoo, Finland, . Lee, D., Yu, A . H . C , Saddler, J. N . 1995. Evaluation of cellulase recycling strategies for the hydrolysis of lignocellulosic substrates. Biotechnol. Bioeng. 45: 328-336. 99 Lee, D., Yu, A. H. C , Wong, K. K. Y., Saddler, J. N. 1994. Evaluation of the enzymatic susceptibility of cellulosic substrates using specific hydrolysis rates and enzyme adsorption. Appl. Biochem. Biotechnol. 45/46: 407-415. Lee, S. B., Shin, H. S., Ryu, D. D. Y. 1982. Adsorption of cellulase on cellulose: effect of physicochemical properties of cellulose on adsorption and rate of hydrolysis. Biotechnol. Bioeng. 24:2137-2153. Lee, Y.-H., Fan, L. T. 1983. Kinetic studies of enzymatic hydrolysis of insoluble cellulose: (II). Analysis of extended hydrolysis times. Biotechnol. Bioeng. 25: 939-966. Leonard, R. H., Hajny, G. J. 1945. Fermentation of wood sugars to ethyl alcohol. Industrial and Engineering Chemistry. 37(4): 390-395. Mackie, K. L., Brownell, H. H., West, K. L., Saddler, J. N. 1985. Effect of sulphur dioxide and sulphuric acid on steam explosion of aspenwood. J. Wood. Chem. Technol. 5: 405-425. Maiorella, B., Blanch, H. W., Wilke, C. R. 1983a. By-product inhibition effects on ethanolic fermentation by Saccharomyces cerevisiae. Biotechnol. Bioeng. 25: 103-121. Maiorella, B. L., Blanch, H. W., Wilke, C. R. 1983b. Distillery effluent treatment and by-product recovery. Process Biochemistry. 12: 5-8. Mamers, H., Menz, D. N. J. 1984. Explosion pretreatment of Pinus radiata wood chips for the production of fermentation substrates. Appita. 37: 644-649. Marchal, R., Ropars, M., Vandecasteele, J. P. 1986. Conversion into acetone and butanol of lignocellulosic substrates pretreated by steam explosion. Biotechnol. Lett. 8: 365-370. Mattiason, B., Hoist, O. 1991a. Extractive Bioconversions, Marcel Dekker, Inc. Mattiason, B., Hoist, O. 1991b. Objectives for extractive bioconversion, in Extractive Bioconversions, B. Mattiason and O. Hoist, eds., Marcel Dekker, Inc., New York: pp. 1-5. Mes-Hartree, M., Hogan, C. M., Saddler, J. N. 1987. Recycle of enzymes and substrate following enzymatic hydrolysis of steam-pretreated aspenwood. Biotechnol. Bioeng. 30: 558-564. Miller, D., Sutcliffe, R., Saddler, J. N. 1989. Determination of molecular weight distributions for celluloses from steam and enzyme treated lignocellulosics. TAPPI Proceedings of the International Symposium on Wood and Pulping Chemistry, Technical Association of the Pulp and Paper Industry, Atlanta, GA, . Moloney, A., Coughlan, M. P. 1983. Sorption of Talaromyces emersonii cellulase on cellulosic substrates. Biotechnol. Bioeng. 25: 271-280. 100 Morjanoff, P. J., Gray, P. P. 1987. Optimization of steam explosion as a method for increasing susceptibility of sugarcane bagasse to enzymatic saccharification. Biotechnol. Bioeng. 29: 733-741. Nguyen, Q. 1990. Computer simulation model of an enzymatic biomass conversion process. Forintek Canada Corp. Final Report Number 53-43-B-407. Nguyen, Q. 1991. Review of environmental impacts and applications of biotechnology in the pulp and paper industry in Canada. M.B.A. University of Ottawa, Canada. Nguyen, Q. A. 1993. Economic analyses of integrating a biomass-to-ethanol plant into a pulp/saw mill, in Bioconversion of Forest and Agricultural Plant Wastes, J. N. Saddler, ed., C.A.B. International, Wallingford, U.K.: pp. 321-340. Nguyen, Q. A., Saddler, J. N. 1991. An integrated model for the technical and economic evaluation of an enzymatic biomass conversion process. Biores. Technol. 85: 275-282. Ohlson, I., Tragrudh, G., Hahn-Hagerdal, B. 1984. Enzymatic hydrolysis of sodium-hydroxide-pretreated sallow in an ultrafiltration membrane reactor. Biotechnol. Bioeng. 26: 647. Ohta, K., Beall, D. S., Meija, J. P., Shanmugan, K. T., Ingram, L. O. 1991. Metabolic engineering of Klebsiella oxytoca M5A1 for enthanol production from xylose and glucose. Appl. Environ. Microbiol. 57: 2810-2815. Olsson, L., Hahn-Hagerdahl, B. 1993. Fermentative performance of bacteria and yeasts in lignocellulose hydrolysates. Process Biochemistry. 28: 249-257. Ooshima, H., Burns, D. S., Converse, A. O. 1990. Adsorption of cellulase from Trichoderma reesei on cellulose and lignaceous residue in wood pretreated by dilute sulfuric acid with explosive decompression. Biotechnol. Bioeng. 36(446-452). Otter, D. E., Munro, P. A., Scott, G. K., Geddes, R. 1989. Desorption of Trichoderma reesei cellulase by a range of detergents. Biotechnol. Bioeng. 34: 291-298. Parekh, S. R., Parekh, R. S., Wayman, M. 1987. Fermentation of wood-derived acid hydrolysates in a batch bioreactor and in a continuous dynamic immobilized cell bioreactor by Pichia stipitis R. Process Biochemistry. 22(3). Perez, J., Wilke, C. R., Blanch, H. W. 1981. Enzymatic hydrolysis of corn stover process development and evaluation. M.S. University of California at Berkeley. Perry, R. H., Chilton, C. H. 1973. Chemical engineers' handbook, McGraw-Hill, New York. Peters, M. S., Timmerhaus, K. D. 1991. Plant design and economics for chemical engineers, McGraw-Hill, New York. 101 Philippidis, G. P., Wyman, C. E. 1992. , in Recent Advances in Biotechnology, F. Vardar-Sukan and S. S. Sukan, eds., Kluwer Academic Publishers, Dordrecht, The Netherlands: pp. 405-411. Phillippidis, G. P. 1994. Cellulase production technology: evaluation of current status, in Enzymatic Conversion of Biomass for Fuels Production, M. E. Himmel, J. O. Baker and R. P. Overend, eds., American Chemical Society, Washington, D.C.: p 499. Phillippidis, G. P., Spindler, D. D., E.Wyman, C. 1992. Mathematical modeling of cellulose conversion to ethanol by the simultaneous saccharification and fermentation process. Appl. Biochem. Biotechnol. 34/35: 543-556. Pilkington, B. J., Rose, A. H. 1988. Reactions of Saccharomyces cerevisiae and Zygosaccharomyces bailii to sulphite. Journal of General Microbiology. 134: 2823-2830. Puis, J., Poutanen, K., Korner, H.-U., Viikari, L. 1985. Biotechnical utilization of wood carbohydrates after steaming pretreatment. Appl. Microbiol. Biotechnol. 22: 416-423. Puri, V. P. 1984. Effect of crystallinity and degree of polymerization of cellulose on enzymatic saccharification. Biotechnol. Bioeng. 26: 1219-1222. Ramos, L. P. 1992. Steam pretreatment and enzymatic hydrolysis of Eucalyptus viminalis chips. PhD. thesis. Ottawa-Carleton Institute of Biology. University of Ottawa. Ramos, L. P., Breuil, C , Kushner, D. N., Saddler, J. N. 1992a. Steam pretreatment conditions for effective enzymatic hydrolysis and recovery yields of Eucalyptus viminalis wood chips. Holzforschung. 46: 149-154. Ramos, L. P., Breuil, C , Saddler, J. N. 1992b. Comparison of steam pretreatment of eucalyptus, aspen and spruce wood chips and their enzymatic hydrolysis. Appl. Biochem. Biotechnol. 34/35: 37-47. Ramos, L. P., Breuil, C , Saddler, J. N. 1993. The use of enzyme recycling and the influence of sugar accumulation on cellulose hydrolysis by Trichoderma cellulases. Enzyme Microb. Technol. 15: 19-25. Ramos, L. P., Saddler, J. N. 1994. Enzyme recycling during fed-batch hydrolysis of cellulose derived from steam-exploded Eucalyptus viminalis. Appl. Biochem. Biotechnol. 45/46: 193-207. Rao, M., Seeta, R., Despande, V. 1983. Effect of pretreatment on the hydrolysis of cellulose by Penicillium funiculosum cellulase and recovery of enzyme. Biotechnol. Bioeng. 25: 1863-1871. Reese, E. T. 1982. Elution of cellulase from cellulose. Process Biochem. 17: 2-6. 102 Roberto, I. C , Lacis, L. S., Barbosa, M. F. S., de Mancilha, I. M. 1991. Process Biochemistry. 26: 15-21. Ryu, D. D. Y., Kim, C , Mandels, M. 1984. Competitive adsorption of cellulase components and its significance in a synergistic mechanism. Biotechnol. Bioeng. 26: 488-496. Ryu, D. D. Y., Lee, S. B., Tassinari, T., Macy, C. 1982. Effect of compression milling on cellulose structure and on enzymatic hydrolysis kinetics. Biotechnol. Bioeng. 24: 1047-1067. Saddler, J. N., Brownell, H. H., Clermont, L. P., Levitin, N. 1982. Enzymatic hydrolysis of cellulose and various pretreated wood fractions. Biotechnol. Bioeng. 24: 1389-1402. Saddler, J. N., Ramos, L. P., Breuil, C. 1993. Steam pretreatment of lignocellulosic residues, in Bioconversion of Forest and Agricultural Plant Wastes, J. N. Saddler, ed., C.A.B. International, Wallingford, U.K.: pp. 73-92. Sasaki, T., Tanaka, T., Nanbu, N., Sato, Y., Kainuma, K. 1979. Correlation between X-ray diffraction measurements of cellulose crystalline structure and the susceptibility of microbial cellulase. Biotechnol. Bioeng. 21: 1031-1042. Sattler, W., Esterbauer, H., Glatter, O., Steiner, W. 1989. The effect of enzyme concentration on rate of hydrolysis of cellulose. Biotechnol. Bioeng. 33: 1221-1234. Schwald, W. 1987. Effect of steam treatment on the components of pretreated and postreated aspenwood, and characterization of the resulting fractions. Forintek Canada. Internal Report. Schwald, W., Breuil, C , Brownell, H. Ft., Chan, M., Saddler, J. N. 1989a. Appl. Biochem. Biotechnol. 8: 543-560. Schwald, W., Breuil, C , Brownell, H. H., Chan, M., Saddler, J. N. 1989b. Assessment of pretreatment conditions to obtain fast Complete hydrolysis on high concentrations. Appl. Biochem. Biotechnol. 20/21: 29-44. Schwald, W., Brownell, H. H., Saddler, J. N. 1988. Enzymatic hydrolysis of steam treated aspen wood: influence of partial hemicellulose and lignin removal prior to pretreatment. J. Wood Chem. Technol. 8: 543-560. Schwald, W., Smaridge, T., Chan, M., Breuil, C , Saddler, J. N. 1989c. The influence of S0 2 impregnation and fractionation on product recovery and enzymic hydrolysis of steam-treated sprucewood, in Enzyme Systems for Lignocellulosic Degradation, M. P. Coughlan, ed., Elsevier, New York: pp. 231-242. SERI-Chem Systems Ltd., Lawrence Berkeley Labs 1984. Ethanol production by enzymatic hydrolysis - parametric analysis of a base-case process. Solar Energy Research Institute Number SERI/TR-231-2093. 1984. 103 Singh, A., Kumar, P. K. R., Schuegert, K. 1991. Adsorption of cellulases during saccharification of cellulosic residues. J. Biotechnol. 18(3): 205-212. Sinitsyn, A. P., Bungay, M. L., Clesceri, L. S., Bungay, H. R. 1983. Recovery of enzymes from the insoluble residue of hydrolysed wood. Appl. Biochem. Biotechnol. 8: 25-29. Sinitsyn, A. P., Gusakov, A. V., Vlasenko, E. Y. 1991. Effect of structural and physico-chemical features of cellulosic substrates on the efficiency of enzymatic hydrolysis. Appl. Biochem. Biotechnol. 30: 43-59. Steiner, W., Sattler, W., Esterbauer, H. 1988. Adsorption of Trichoderma reesei cellulases on cellulose: experimental data and their analysis by different equations. Biotechnol. Bioeng. 32: 853-865. Stewart & Ewing Associates Ltd. 1990. British Columbia forest industry mill residues for calendar year 1989. B.C. Ministry of Forests Number Project No. B75439. Stone, J., Scallan, A., Donefer, E., Ahlgren, E. 1969. Digestibility as a simple function of a molecule of a similar size to a cellulase enzyme. Advances in Chemistry Series. 95: 219-241. Sutcliffe, R., Breuil, C , Brownell, H. H., Saddler, J. N. 1988. The influence of lignin extraction solvents on the enzymatic hydrolysis of steam-treated aspenwood. IEA Workshop on the Bioconversion of Lignocellulosics, Ottawa, Canada, . Szczodrak, J., Targonski, Z. 1989. Simultaneous saccharification and fermentation of cellulose: effect of ethanol and cellulases on particular stages. Acta Biotechnol. 6: 555-564. Tan, L. U. L., Yu, E. K. C , Campbell, N., Saddler, J. N. 1986. Column cellulose hydrolysis reactor: an efficient cellulose hydrolysis reactor with continuous cellulase recycling. Appl Microbiol Biotechnol. 25: 250-255. Tan, L. U. L., Yu, E. K. C , Mayers, P., Saddler, J. N. 1987. Column cellulose hydrolysis reactor: the effect of retention time, temperature, cellulase concentration and exogenously added cellobiose on the overall process. Appl. Microbiol. Biotechnol. 26: 21-27. Tanaka, M., Fukui, M., Matsuno, R. 1988. Removal of lignin and reuse of cellulases for continuous saccharification of lignocelluloses. Biotechnol. Bioeng. 32: 897-902. Tenrud, I. E., Theander, O., Torneport, L., Vallander, L. 1989. Changes in the chemical composition of steam-exploded wheat straw during enzyme hydrolysis. Enzyme Microb. Technol. 11: 500-506. Thompson, D. N., Chen, H. C , Grethlein, H. C. 1992. Comparison of pretreatment methods on the basis of available surface area. Biores. Technol. 39: 155-163. 104 Tjerneld, F., Persson, I., Albertsson, P.-A., Hahn-HNgerdahl, B. 1985. Enzymatic hydrolysis of cellulase in aqueous two phase systems I. partition of cellulases from Trichoderma reesei. Biotechnol. Bioeng. 27: 1036-1043. Tran, A. V., Chambers, R. P. 1986. Ethanol fermentation of red oak acid prehydrolysate by the yeast Pichia stipitis CBS 5776. Enzyme and Microbial Technology. 8: 439-444. Tshiteya, R. M. 1992. Conversion technologies - biomass to ethanol - alcohol fuels. National Renewable Energy Lab. Reference Report Number 3. September, 1992. Ucar, G., Fengel, D. 1988. Characterization of the acid pretreatment for the enzymatic hydrolysis of wood. Holzforschung. 42: 141-148. Ulrich, G. D. 1984. A guide to chemical engineering process design and economics, John Wiley & Sons, Inc., New York, p 472. Ung, R. 1986. The economics of forest chemicals: process development and economic analysis. Forintek Canada Corp. Number 53-43-393. Dec. 1986. Vallander, L., Eriksson, K.-E. 1987. Enzyme recirculation in saccharification of lignocellulosic materials. Enzyme Microb. Technol. 9: 714-720. Vallander, L., Eriksson, K.-E. 1990. Production of ethanol from lignocellulosic materials: state of the art, in Advances in Biochemical Engineering and Biotechnology, A. Fiechter, ed., Springer-Verlag, Berlin: pp. 63-95. van Zyl, C , Prior, B. H., du Preez, J. C. 1988. Production of ethanol from sugar cane bagasse hemicellulose hydrolysate by Pichia stipitis. Appl. Biochem. Biotechnol. 17: 357-369. Vogel, H. C. 1983. Fermentation and Biochemical Engineering Handbook - Principles, Process Design, and Equipment, Noyes Publications, Park Ridge. von Sivers, M. 1995. Ethanol from wood - a technical and economic evaluation of ethanol production processes. Licentiate Dissertation. Department of Chemical Engineering I. Lund Institute of Technology. von Sivers, M., Zacchi, G. 1993. A techno-economical comparison of three processes for the production of ethanol from wood. Lund Institute of Technology - Department of Chemical Engineering I Number Project No. LUTKDH/(TKKA-7006)/l-27/(1993). von Sivers, M., Zacchi, G. 1995. A techno-economical comparison of three processes for the production of ethanol from pine. Biores. Technol. 51: 43-52. Wald, S. A., Blanch, H. W., Wilke, C. R. 1981. Enzymatic hydrolysis of rice straw for ethanol production. M.S. Thesis. University of California at Berkeley. 105 Wang, S. S., Converse, A. O. 1991. On the use of enzyme adsorption and specific hydrolysis rate to characterize thermal-chemical pretreatment. Appl. Biochem. Biotechnol. 34/35: 61-74. Ward, O. P. 1991. Bioprocessing, in Bioprocessing, , Van Nostrand Reinhold, New York: pp. 1-13. Wayman, M., Parekh, S. R. 1990. Biotechnology of biomass conversion, Open University Press. Wilson, J. J., Deschatelets, L., Nishikawa, N. K. 1989. Applied Microbiology and Biotechnology. 31: 592-596. Wong, K. K. Y., Deverell, K. F., Mackie, K. L., Clark, T. A., Donaldson, L. A. 1988. The relationship between fiber porosity and cellulose digestibility in steamed-exploded Pinus radiata. Biotechnol. Bioeng. 31: 447-456. Woodward, J., Zachry, G. S. 1982. Immobilization of cellulase through its carbohydrate side chains - a rationale for its recovery and reuse. Enzyme Microb. Technol. 4: 245-248. Wright, J. D. 1986. Fuel alcohol technical and economic evaluation. NREL for U.S. Department of Energy. Final Report Number SP-231-2904. 1986/02. Wright, J. D. 1989. Evaluation of enzymatic hydrolysis processes, in Energy from Biomass and Wastes XII, D. L. Klass, ed., Institute of Gas Technology, Chicago, IL: pp. 1247-1296. Wright, J. D., Wyman, C. E., Grohmann, K. 1988. Simultaneous saccharification and fermentation of lignocellulose: process evaluation. Appl. Biochem. Biotechnol. 18: 75-90. Wright, L. L. 1995. Demonstration and commercial production of biomass for energy. Second Biomass Conference of the Americas: Energy, Environment, Agriculture, and Industry, Portland, Oregon, NREL. Wyman, C. E. 1995. Biomass derived oxygenates for transportation fuels. Second Biomass Conference of the Americas: Energy, Environment, Agriculture, and Industry, Portland, Oregon, Wyman, C. E., Hinman, N. D. 1990. Ethanol: fundamentals of production from renewable feedstocks and use as a transportation fuel. Appl. Biochem. Biotechnol. 24/25. Yu, S., Wayman, M., Parekh, S. R. 1987. Fermentation to ethanol of pentose-containing spent sulphite liquor. Biotechnol. Bioeng. 29: 1144-1150. Zacchi, G., Skoog, K., Hahn-Hagerdahl, B. 1988. Economic evaluation of enzymatic hydrolysis of phenol-pretreated wheat straw. Biotechnol. Bioeng. 32: 460-466. 106 Zhang, M., Eddy, C , Deanda, K., Finkelstein, M., Picataggio, S. 1995. Metabolic engineering of a pentose metabolism pathway in ethanologenic Zymomonas mobilis. Science. 2 6 7 : 240-243. 107 8. A P P E N D I C E S - D E T A I L E D DESCRIPTION OF T H E ASSUMPTIONS I N C O R P O R A T E D INTO T H E P R E T R E A T M E N T / F R A C T I O N A T I O N A N D H Y D R O L Y S I S / R E C Y C L E C O M P O N E N T S OF T H E S T E A M M O D E L 8.1 P R E T R E A T M E N T & F R A C T I O N A T I O N Pretreatment is an essential step for the efficient downstream conversion of lignocellulosic feedstocks to ethanol. Although there are many types of pretreatment, past research has shown that the pretreatment methods that are most effective for primarily cellulose recovery and conversion have generally been physico-chemical in nature (Saddler et al., 1993). The reaction kinetics and chemical structure of the three main components of lignocellulosics, i.e. cellulose, hemicellulose and lignin, differ to such an extent that complete recovery of the components is unlikely. However, as previously mentioned, it is economically imperative that recovery is maximized. Steam explosion has generally been recognized as one of the most effective methods for pretreating and fractionating lignocellulosic feedstocks (Foody, 1980; Mamers and Menz, 1984; Dekker et al., 1987; Brownell, 1989; Eklund et al., 1990; Nguyen and Saddler, 1991). A range of times, temperatures/pressures, and acid catalysts have been used by researchers on various feedstocks (Table 7). A number of these researchers concentrated on maximizing the cellulose recovery and failed to recognize the importance of also recovering the hemicellulosic and lignin components. Although most of the lignocellulosic feedstocks are currently considered to be of low or negative value it is highly likely that, as the process attains commercial-scale, the feedstock price will rise. Most lignocellulosic feedstocks are sold on a weight basis and the cellulose component represents only about a half of the original dry weight. Thus, recent modelling studies have started to recognize the importance of optimizing the pretreatment recovery of both the more labile hemicellulose component and the lignin fraction as well as enhancing cellulose hydrolysis (Eklund et al., 1988). 108 1986b) 1986b) 13 13 (Tenrud et al., 1989) (Beltrame et al., 1992) (Marchal et al., 1986) (Dekker and Wallis, 1983) (Morjanoff and Gray, 1987) (Morjanoff and Gray, 1987) (Marchal et al., 1986) (Brownell and Saddler, 1987) (Brownell and Saddler, 1984; Brownell et (Brownell and Saddler, 1984; Brownell et (Mackie et al., 1985) (Mackie et al., 1985) (Schwald et al., 1989a) (Dekker et al., 1987) (Ramos et al., 1992a) (Ramos et al., 1992a) (Eklund, 1994) (Eklund, 1994) (Eklund, 1994) (Eklund, 1994) (Schwald et al., 1989c) (Clark and Mackie, 1987) cn 30-600 o CN 60-1200 80-360 2-90 o CN 20-6000 20-240 10-80 o oo o CN o o o o CO o CN 50-150 o o vo o o VO o o vo o o VO 50-250 30-1080 r-CN l 200-230 O o CN O 00 CN CO CO o o CN i 200-250 o i n CN 6 o 00 CN CN cn CN i 190-240 o CN o CN (N o 00 CN X CN d O O CN o oo CN X 00 m d r-CN CN 1 o CN CN o oo vq o CN CN i o ON CN o oo vq O o CN i 220-240 o CN CN i O o CN CN o oo 180-220 o CN i o O 00 CN « i n i — i i VO d o o CN i o vo CN o 00 q CO o CO CN o oo CN o 00 o CN CN i O ON CN 0 oo t£ m 1 m d oo CN i CN oo CN 0 00 S£ CN 1 m d co W AGRICULTURAL RESIDU Wheat straw Sugarcane Bagasse Corn Stover HARDWOODS Aspen Eucalyptus Willow SOFTWOODS Spruce Radiata pine Various groups have tried to determine the relative importance of lignin and hemicellulose in the enzymatic hydrolysis of pretreated cellulosic substrates (Baker et al., 1975; Brownell and Saddler, 1984; Grohmann et al., 1985). Some researchers have linked the removal of hemicellulose to an improvement in enzyme digestibility of the pretreated wood (Grohmann et al., 1985; Chum et al, 1988) and it has been suggested that the release of hemicellulose produces an increase in the accessible pore volume and the specific surface area (Grethlein, 1985). However, it has been found that, even after virtually all of the pentosan of aspen wood has been solubilized or destroyed, further steam treatment continues to improve the subsequent rate and extent of enzymatic hydrolysis (Brownell and Saddler, 1984). Therefore, it appears that pentosan removal is just one of a number of factors involved in the improvement in enzyme digestibility of pretreated wood. Contradictory results have also been reported for delignification. Results indicating beneficial (Stone et al., 1969; Baker et al., 1975), little (Ramos et al., 1992b), or no effect (Schwald et al., 1989c) on digestibility have all been described in the literature. It has also been shown that effective pretreatment can occur in the absence of hemicellulosic acetic acid and that short steaming times can be used to produce in situ hemicellulose hydrolysis and good sugar recoveries (Schwald et al., 1988). Thus it is apparent that pretreatment and fractionation have a pivotal role, not only on the efficiency of hemicellulose and lignin recovery, but also as the key elements in achieving effective cellulose hydrolysis. The current modelling efforts have considered the effect that a varied feedstock i.e. hardwood, softwood and agricultural residues would have on each of the component steps of an integrated biomass-to-ethanol process. It was found that factors such as, the use of an acid catalyst, feedstock handling, and various feedstock properties such as chemical composition and cell wall distribution, moisture content and feedstock/liquid relations, ultrastructure, bulk density, specific density, temperature, and purchase cost will all influence the economics and process steps that would be used. This current work tries to identify the likely 110 conditions and processes that would be used to maximize hemicellulose derived sugar and lignin recoveries while obtaining complete hydrolysis of the cellulosic residue. Research groups associated with the International Energy Agency (IEA) "Biotechnology for the Conversion of Lignocellulosics Network", have been refining a "generic" biomass-to-ethanol process model that will be capable of processing different types of feedstocks. It was apparent that the feedstock properties of importance to the technical design and economic evaluation of any process include chemical, physical and cost components. This thesis primarily concentrates on chemical properties of the feedstock, as they tended to have a greater overall effect on the technical design of biomass-to-ethanol processes. Although the physical properties of the feedstock such as ultrastructure, bulk density, specific density, temperature and form (chips, wafers, sawdust, etc.) are important they generally impact on just the pretreatment step. Similarly, the cost of the feedstock is one parameter that is easily changed within any modelling work and did not need to be discussed further here. Thus the chemical composition and overall yield of each of the components were considered to be key variables that impacted on hemicellulose and lignin recovery and cellulose hydrolysis. Research concerning steam treatment of lignocellulosics includes hardwoods (Schwald et al., 1988; Ramos et al., 1992b), softwoods (Schwald et al., 1989c) and agricultural residues (Zacchi et al., 1988). It is apparent (Figure 21) that there is considerable variation in both the chemical composition of representative lignocellulosic substrates and the amount of material found in each of the cellulose, hemicellulose, lignin and extractive fractions. The actual association of each of these components has a major impact on the efficiency of pretreatment and fractionation. For example, although similar pretreatment conditions can be defined to maximise hemicellulose solubilization and sugar recovery, the higher concentration of extractives associated with softwood will likely lead to considerable inhibition problems when attempts are made to 111 3 -o W ±L CD c CD _ Q. O < "E -= CO -Q X E CD CO CD > o CD ±= sz X LU >> 0) c c co cp o < - I < • • 0 CO CL Z3 o 2 5 T3 O < o ' c o =3 o O c CD c IS CD c CD -i—' O c CD c CD J2 >, CD CD R < X o ^  o 0 C3 LU B • m i n i m i a cu < 0 o -a a to o C L oo 6 > -rt H x •'3 2 c/J I « eft « O ^ | ¥ ed pd +3 "S c/n CD -§ rt C ^ Id o fa • —< r r ,2 3 I I 8 « c > € cu m cu a, "3 rt o c o OH s o U CN ferment hemicellulose derived sugars from softwoods. Similarly, softwood lignins contain primarily guaiacyl units whereas hardwood lignins contain both syringyl and guaiacyl components and agricultural residues, particularly the grasses, contain syringyl-, guaiacyl-, and p-hydroxyphenyl units. These differences in the basic building blocks combined with the three dimensional structure within the substrate result in quite different responses to fractionation when alkali extraction and peroxide are used to remove the lignin and enhance hydrolysis of cellulosic residues (Schwald et al., 1988; Schwald et al., 1989c). In the subsequent sections a discussion of the steam explosion pretreatment conditions that have been used to maximize hemicellulose hydrolysis and lignin recovery while producing a cellulosic substrate that can be readily attacked by enzymes will be described. 8.1.1 Hemicellulose Recovery The conditions shown in Table 7 reflect various researchers' definitions of optimized pretreatment conditions. Optimization has usually been solely based on the maximum cellulose hydrolysis recovery that can be achieved with a disregard for the hemicellulose or lignin recoveries. The kinetics and structures of the lignocellulosic components differ to such an extent that there must be a compromise in the pretreatment conditions used. Near theoretical fermentation yields can be attained from the hemicellulosic fraction (Ingram et al., 1987; Hahn-Hagerdal et al., 1993; Zhang et al., 1995) when the oligomeric components have been subsequently hydrolyzed prior to fermentation. However, it is preferable if the hemicellulose derived sugar stream can be produced in a monomeric form by the pretreatment process itself and contain minimal breakdown products. Generally it has been recognized that the use of high temperatures and short cooking times produces pretreatment conditions that both soften the wood components and take into account the solubilization kinetics of hemicellulose (Saddler et al., 1993). Breakdown products that are inhibitory to the fermentation organisms will be produced where more severe pretreatment conditions are used. Research efforts to address the production of inhibitors generally have pursued two main courses of action. A preventative 113 course i.e. adjustment of the kinetics of the pretreatment via acid catalysts, which will be discussed in greater detail in a later section, and a reactive approach that identifies the inhibitory compounds and develops methods to remove or detoxify the compounds before fermentation. Past efforts have looked at the effects of hardwood hemicellulose breakdown products, i.e. furfural and acetic acid on fermentation. There are also some inhibitory substances such as wood extractives that are naturally associated with the substrate (Figure 21). Toxicity associated with the extractive fraction has been shown (Galbe, 1994) to be primarily associated with the non-volatile components and will generally be associated with the post-pretreatment water-soluble fraction. Although the optimization conditions for maximum sugar recoveries can be predicted, the nature of the inhibitory compounds and methods to detoxify the streams in a cost-effective manner is still largely unresolved. Traditional detoxification methods such as the addition of activated coal (Roberto et al., 1991), extraction with organic solvents (Fein et al., 1984; Frazer and McCaskey, 1989; Wilson et al., 1989), ion-exchange (Clark and Mackie, 1984; Fein et al., 1984; Frazer and McCaskey, 1989), ion exclusion (Buchert et al., 1990), molecular sieves (Tran and Chambers, 1986), overtiming (Leonard and Hajny, 1945; van Zyl et al., 1988), and steam stripping (Yu et al., 1987) have been shown to be costly or ineffective (Beck, 1993). Although there have been some attempts to adapt the fermentation organisms to the inhibitory substances (Olsson and Hahn-Hagerdahl, 1993), each feedstock change required a further adaptation period. Consequently, the detoxification step has yet to be resolved and has not been included in modelling efforts to date. Almost all of the hemicellulose can be extracted from the steam exploded feedstock by a subsequent water extraction step (Schwald, 1987). The water extract contains hemicellulose derived sugars, and any breakdown products derived from the hemicellulose and lignin components as well as some of the extractive components. The recovery yield of fermentable sugars will primarily depend on the severity of the steam treatment and the extraction parameters. Generally, the more severe the pretreatment, the lower the hemicellulose-derived sugar yield (Figure 22). High severity treatments are generally required to achieve good enzyme 114 accessibility of the cellulose. However, the more severe conditions result in the generation of more hemicellulose and lignin breakdown products (Figure 23) which are generally quite inhibitory to the fermentative organisms used to produce ethanol from the hemicellulose derived sugars (Parekh et al., 1987). As a result an optimal pretreatment condition has been defined as one which results in a minimum overall ethanol production cost and maximum ethanol yields, based on maximum recovery of hemicellulose derived sugars after the pretreatment step. The process design and economics of the hemicellulose extraction step is dominated by the objectives of high extraction yields, high dissolved solids concentration in the extract and low solvent loadings. High extraction yields of hemicellulose can provide higher ethanol returns and both savings in capital costs (through a reduction in the total volume of the cellulose hydrolysis fermenters) and operating costs (through reduction in the required mixing energy). Reduced capital and energy costs, in the subseqent bioconversion of the hemicellulose extract to ethanol, are the main incentive for having high dissolved solids concentration in the extract. Lower water usage lowers input costs and also concentrates the hemicellulose sugars. However, it also concentrates many of the wood extractives or sugar degradation products that are inhibitory to hemicellulose fermenting microorganisms (Mackie et al., 1985). If a steam explosion based process is to adequately recover all three components for a wide variety of feedstocks, the use of an acid catalyst will be required (Ramos et al., 1992b). Past research has concentrated on primarily two acid catalysts, sulfuric acid (H2SO4) and sulfur dioxide (SO2) (Forintek, 1985; Eklund et al., 1995). These acid catalysts have been shown to provide higher hemicellulose and cellulose recoveries in both hardwoods and softwoods. Enhanced cellulose hydrolysis rates are also found in all substrates. When the two acid catalysts were compared on an equal severity basis they both enhanced the survival of pentose sugars. Although, the alkali lignin extraction from the water-washed exploded substrates was lower with H2SO4 (Mackie et al., 1985; Brownell, 1987; Eklund et al., 1995), the gaseous sulfur dioxide was easier and faster to introduce while the sulfuric acid resulted in a greater steam consumption. 116 O > 5 c o "? 5 < c "O £ > ^ OD O 03 (j M _ CD < O LO C T c o CD "-^ T3 O CO > • . = CD o o O cu < CD to O CO Q O —^ (D >» O X < n • • CO —H CNJ -+-o o CM o o CM o CO CO T 3 c o o CD CD E CO CD CT) O o CO o CD o o CM o o £ -a S-l a X -a CD Ofi ON CD ON i — i .S 13 "71 +-• <N CD O W C/5 O T3 § <D Cl, ON S ON C^  CD C O CD * 2 13 g c .2 o CD 3 CD en CN <D so CD -t-> cc3 Past work has also helped elucidate the catalytic mechanism of sulfur dioxide addition during steam explosion (Brownell, 1987). When aspenwood chips were impregnated with SO2 and subjected to steam at 200°C, much of the SO2 in the chips was converted to sulfuric acid within 20 seconds. This sulfuric acid (and not sulfurous acid or lignosulfonic acid) is the actual catalyst. Although air, when present, may be involved in this oxidation, sulfuric acid is also formed when air is carefully excluded by N 2 , apparently by a disproportionation reaction. The amount of sulfuric acid produced, increases with increased SO2 impregnation, but increases less than proportionately. Formation of sulfuric acid and subseqent catalysis was demonstrated by impregnating aspenwood chips with 1.6% SO2 and treatment for 20 seconds with steam at 200*0 to produce sulfuric acid. The treatment was stopped after this time and unused SO2 was carefully and completely vaporized and removed from the chips and vessel, without removing the sulfuric acid. Treatment was then resumed with fresh steam for 80 seconds at 200AC in the absence of SO2. The resulting well-cooked product, from this sulfuric-acid-catalyzed steaming, was similar to that obtained from a 100-second cook with SO2. Control cooks for 100 seconds at 200°C without SO2 gave relatively uncooked products, as did cooks for only 20 seconds with SO2. If sulfuric acid had not formed during the first 20 seconds steaming, the second steaming for 80 seconds would not have been catalyzed, and the product would have been under-cooked like the controls. The conversion of SO2 to sulfuric acid in wood chips is apparently independent of the composition of the chips, which act only as porous supports for the SO2. A very similar conversion occurs when SO2 is adsorbed on charcoal (rather than on wood chips) and is subjected to saturated steam at 200*0. In both cases equilibrium conditions were not established. The conversion occurs too rapidly for the SO2 to be driven out of the wood or charcoal, by the rapidly rising temperature, at least when the high-pressure steam is quickly admitted. The equilibrium vapor pressure of SO2 above aqueous SO2 solutions is quite significant, even at low temperatures, and the solubility of SO2 decreases with increasing temperature. Nevertheless, there appears to be little or no exchange of SO2 between the chips during steam 118 treatment as shown by the following experiment (Brownell, 1987). The lower half of a thin-walled loosely-covered canister was filled with commercial wood chips impregnated with 1.6% of SO2 (dry wood basis), and the upper half (separated only by a wire screen) was filled with unimpregnated chips. The canister was then quickly lowered into the pressure vessel and steam was admitted within 2 seconds. After 100 seconds in 200*0 steam, only the chips in the lower half of the canister (i.e. those originally impregnated with SO2) were thoroughly cooked. Those in the upper half (i.e. no S02) were virtually uncooked and resembled those from control cooks in the absence of all SO2. A sulfur balance (Figure 24) has been calculated to further determine the mode of action of the SO2 and the fate of the sulfur (Brownell, 1987). Approximately 50% of the input S02 remained in the exploded substrate following explosive decompression and air drying at room temperature of the resulting exploded wood, indicating that, at the level of SO2 impregnation used in this work (1.6% of O.D. wood input), half of the SO2 was acting as sulphurous acid or was present in the vapour phase in the void volume of the gun. Water washing removed the large majority (34.4%) of the retained sulfur with some 7% of the original sulphur remaining in the washed substrate. The sulfur located in the water solubles is non-volatile since freeze drying of the wash liquors did not result in a total loss of sulfur. Approximately one third of the total original sulfur remains in the dry water solubles. The water soluble sulfur was not inorganic sulphite resulting from reaction with the wood ash. Although lignosulphonates were not isolated from the water soluble material, it was probable that the sulfur was bound to lignin fragments rather than to any soluble or insoluble carbohydrate. Sulfur dioxide impregnation prior to steam pretreatment particularly at lower temperatures and pressures is the preferred catalytic procedure. It reduces the wastage of SO2, less is converted to H2SO4, it reduces the corrosion rate of the steam pretreatment reactor, as SO2 under the milder conditions is less corrosive, and it provides the opportunity to more easily recycle the unconverted SO2 and prevent its release into the environment. Previous work, using 119 O O a in i-H O C/3 c3 'o > d s o J-l P H ^ CO O ^ ^ ° rS 5-1 •4—> o • i-H •4—> o cd * H P H o oo -4—> cd o ^ oo O N C3 -*-» CU <D O ccj CO O O -a <U O B CD •~ OH a o 13 g "SH H i-i o ccj J=I C O ( N <D 60 impregnation before steam pretreatment, determined that 1% of SO2, or less, would be satisfactory, if the S0 2 was converted to H 2 S0 4 in high yield within the wood (Schwald, 1987). Feedstock compositional differences have been shown to influence the effectiveness of the fractionation and consequently the level of product yield attained for the so-called "optimized" process. Past research has concentrated primarily on hardwood species such as aspen, willow, birch and eucalyptus and has shown that each of these species requires a slightly different pretreatment condition (Table 7) to optimize the recovery of the hemicellulose sugars. The optimal conditions (Figure 22 and Figure 23) for a representative hardwood were determined to be 1.6% SO2 for 120 seconds at 210*0 which produces an 80%) hemicellulose recovery of xylan as xylose. Agricultural residues such as wheat straw, sugarcane bagasse and corn stover have also been studied and react in a manner similar to hardwoods. Softwood species such as spruce (Schwald et al., 1989c) or radiata pine (Clark and Mackie, 1987) require substantially different pretreatment conditions, i.e., longer residence times (Figure 25 and Figure 26) and higher catalyst concentrations (Table 7). Softwood hemicellulose derived monomers are primarily mannose, glucose and galactose with minor amounts of xylose and arabinose (Figure 21). Softwood pretreatment optimization focused on the mannose and glucose recoveries, rather than the xylose, and these hemicellulose derived hexoses could be combined with the glucose stream coming from the cellulose recovery. However, the potential capital and operating cost benefits, that could be anticipated by combining the fermentation of both the cellulose and hemicellulose streams, may be offset by the higher wood extractives concentration that can be anticipated in the hemicellulose stream from softwood substrates. 121 CO T 3 C o o CD CD CO CD 00 T 3 O O o GO T 3 CD bo <u O N C O O N O 13 GO ^ CD O -rt 0) « 5 cu 5 S § «3 s _, o GO -g f-i ed cS " "cu CU > O o CU rt IT) CN I be E CN CN o LO CN O O CN O LO CO T3 C o o CD CO CD E ca CD CO o o o LO -a c c 0 t/3 X ) C 51) CD s-1 Jb ON O 00 GO ON X J 1 3 CU H -M CU s-a. re 0> GO GO | -a +e • — i C L . 2 « C X 5 C w _o to _cu o o o o 00 o CD O o CN cu 04 8.1.2 Lignin Recovery Lignin can be extracted from the water-insoluble exploded aspen using dilute caustic, or ethanol, methanol, acetone, ammonium hydroxide, acetic acid (Sutcliffe et al., 1988). Dilute sodium hydroxide is more effective than the other solvents, relatively inexpensive, and through acid precipitation the lignin can be readily extracted. The lignin extraction yield is dependent on factors such as, pretreatment effectiveness, caustic concentration, extraction method used, and the use of further extraction steps e.g. peroxide wash. Lignin extraction yield is enhanced when a catalyst such as SO2 is added or more severe pretreatment conditions (Figure 27) are used. Concentrations of caustic over 5% also increase the lignin extractibility although the chemical cost is extremely high and two consecutive washes are required to attain over 90% recovery (Figure 28) of the extractables using batch extraction (Brownell, 1987). Preliminary modelling has shown the importance of optimizing the concentration of alkali and reducing the water volumes (Galbe and Zacchi, 1991). The alkali wash will be more fully concentrated only if the majority of the lignin is readily alkali-soluble, as occurs in the case of the hardwoods and agricultural residues. With hardwoods, it is possible to recover close to 90% of the original lignin by following steam explosion and water washing by a subseqent alkali wash (Figure 28). However, with softwoods only 50% of the original lignin could be removed by alkali extraction (Figure 29). As discussed later, the extraction was also shown to greatly reduce the efficiency of hydrolysis of the cellulosic residue. A subsequent peroxide treatment greatly increased the degree of cellulose hydrolysis (Figure 30) while only removing a further 30% of the original lignin. The benefits of delignification include a reduction in the bulk density of the cellulosic residue. Enhanced cellulose recovery benefits are obtained by reducing the total volume of the reactors required for each of the remaining steps associated with the cellulose stream. It also reduces the required mixing energy while concentrating the cellulose content of the hydrolysis input stream. 124 T3 C D CO C D -t—1 ca E TD C D -f—' C O C D i _ C =3 'CD T3 i C C D > O c o o CO 0) -Q _2 o co c c — O CD ~ T J ° X 2 o CD C L £ co ^ c CN o CN CN O X CN X n-Wi n-WIA n-WIA -Wl -WIA -WIA/I co co CO C C C o ZZ o u c c £: O) OI O l CD _ J _J _ J • • • D B S T T IIIIMII o o CN o L O o o o 00 C O TD c o o C D 3* CD E i -E C O C D co o &Q cj eo S ZD CJ •— EX 0 ^ C \ - d - H CU CO CO <u £3 cu 1 t I I S 1 ,2 * T 3 e S o — CJ Lu _CJ JZ =3 CJ 8 cN O CD O o CM rC &0 cd cd CO Z3 o X o O O OO 3 3 CD > CD CO o o o o o o oo o CO o o CM O CO O O CM CO CO >• 9 CM -Q o o co i— _3 o X 0 0 1 O CM ® T 3 O O CD O CO 5 O LO O i -O CO LO CM 0 0 1 O CM O O O CD CO I a CO X o CO o^ CO in cd <D ON 00 ON C M o -a cj cd <u cd ed 2P -I C O ' S GO l i ^ ON CD OO O ON Ci >M M GO cd -a "S « 1 (D 5 on O M S o <M < CD td C CU ~ C CU CL, "K ^ f CN ^ — ' 9 * GO co T 3 CD td CD cd CD -a X o ^ CD 6 ^ CD GO M c2 00 <D oo O O Ci O oo CN O CD 0 0 8.1.3 Cellulose Recovery Enzymatic hydrolysis of cellulose generally has been suggested to be limited by factors that are either related to the structure of the substrate or the type and composition of the cellulase system used to carry out hydrolysis. It has been shown (Ramos et al., 1992b) that the hydrolysis rate declines logarithmically over a range of enzyme and substrate concentrations. It has been hypothesized that the gradual decline in hydrolysis rate is a reflection of the increase in substrate recalcitrance resulting from the structural impediments within the substrate that limit enzymatic hydrolysis. Each cellulosic substrates susceptibility to enzymatic hydrolysis depends on a number of structural features. Those that have been proposed include the crystallinity (Sasaki et al, 1979; Fan et al., 1980; Gharpuray et al., 1983; Puri, 1984), degree of polymerization (Puri, 1984), surface area available to the enzymes (Stone et al., 1969; Gharpuray et al, 1983; Grethlein, 1985; Grous et al., 1988; Sinitsyn et al., 1991; Thompson et al., 1992), and the lignin content and distribution (Gharpuray et al., 1983). However, other enzyme related factors, such as enzyme adsorption, enzyme inactivation and end-product inhibition, have also been shown to affect the overall mechanism of hydrolysis (Lee et al., 1994). Several major changes in the structure of the cellulose have been demonstrated after steam explosion of cellulosic residues. Various authors have reported a gradual decrease in the degree of polymerization (DP) of cellulose and a substantial increase in the crystallinity index (CrI) of the substrate (Puis et al., 1985; Miller et al., 1989). The apparent increase in crystallinity seems to be the result of fusion between cellulose crystallites which were originally separated by a matrix of hemicellulose and lignin. Therefore, the gradual removal of hemicelluloses and lignin, to increase substrate surface area, appears to trigger the reorientation of the cellulose molecules which, after pretreatment, assume a distinct crystalline form. The initial rapid rate by which cellulose is hydrolysed has been associated with the occurrence of more accessible regions at the substrate surface. At the fibre level, these more accessible regions are often associated with cracks or defects in the fibre while at the molecular level, they are characterized by a larger pore volume 129 and/or available surface area (Brownell, 1987; Wong et al., 1988) and a lower crystallinity index (amorphous cellulose) (Saddler et al., 1982; Bertran and Dale, 1985; Sinitsyn et al, 1991). The initial rate of hydrolysis of cellulosic substrates has been linearly correlated with the distribution of micropores which were accessible to the enzymes (Grethlein, 1985). The lower susceptibility of softwood substrates to hydrolysis, compared to hardwood substrates, was associated with a smaller increase in the distribution of these micropores during pretreatment. The development of pore volumes accessible to the enzymes has also been demonstrated to be a key factor in determining the degree of enhancement of substrate accessibility for steam-treated aspen (Excoffier et al., 1991), poplar (Grous et al., 1988), and radiata pine (Wong et al., 1988) chips. It has been suggested that lignin acts as a physical barrier during the hydrolysis of steam-treated substrates and hinders the contact of the substrate by the enzymes (Ucar and Fengel, 1988). Various factors such as the irreversible adsorption and non-specific binding of enzymes onto the lignm-contoining residue, have been implicated (Clesceri et al., 1985; Converse et al., 1990). Although alkali washing of steam-treated hardwoods usually results in a substrate with a lower lignin content, it is generally only as readily hydrolysed as the water-washed substrate (Figure 10) (Wong et al., 1988; Excoffier et al., 1991; Ramos et al., 1992a). It appears that the beneficial effects of alkali extraction, such as lignin removal and cellulose swelling, are offset by other factors which have a detrimental effect on hydrolysis, such as the possible redistribution of the residual lignin and modifications in the crystalline state of the cellulose. Despite the small cellulose hydrolysis gain from delignification, the extraction of the lignin from the pretreated residues is still desirable, as it produces a substrate with a comparatively higher cellulose content from which a higher glucose yield per gram of substrate can be achieved. However, alkali-washing of steam-treated softwoods generally results in substrates which are more recalcitrant to enzymatic degradation than the respective water-washed substrate (Figure 30) (Saddler et al., 1982; Wong et al., 1988; Schwald et al., 1989c). This was more dramatic in the case of spruce, where alkali extraction seemed to redeposit the lignin in a fashion which greatly reduced the ease of hydrolysis (Schwald et al., 1989c). Extraction of the residual alkali-insoluble lignin by 130 oxidative agents, such as sodium chlorite (Saddler et al., 1982) and hydrogen peroxide (Ramos et al., 1992a), has been shown to increase the susceptibility to enzymatic hydrolysis of pretreated substrates derived from both hardwood and softwood residues. It was suggested that it was the redistribution of this residual alkali-insoluble lignin which limited complete hydrolysis of alkali-washed substrates. After alkali washing, this highly condensed, modified lignin appeared to reprecipitate on the surface of the substrate, causing a reduction of both the available surface area and the ability of the cellulose fibres to swell in water. The role that hemicellulose and lignin play in the enzymatic hydrolysis of lignocellulosic substrates is still being debated and from a purely economic standpoint may be inconsequential as the main objective of this bioconversion process should be to recover the maximum amount of all three of the lignocellulosic components. Although the pretreatment and recovery yields have been optimized, certain technical and economic aspects still require refinement. For example, a determination of the inhibitory products associated with the hemicellulose rich water soluble stream, their distribution in the various process streams and cost-effective methods to alleviate their effects on fermentation. Furthermore, peroxide washing, required to achieve effective hydrolysis of the softwood derived cellulosic stream, has generally been viewed in the past as being too costly. However, the high cost of enzymes implies that substrates containing low levels of lignin are required if strategies such as enzyme recycle or simultaneous saccharification and fermentation are to be used at a commercial scale. Thus maximum hemicellulose and lignin recoveries are required to optimize utilization of these streams, enhance cellulose hydrolysis and reduce the cost of producing ethanol from a range of lignocellulosic substrates. 8.2 ENZYME HYDROLYSIS & RECYCLE One step that should be common to all potential lignocellulosic feedstocks is the enzymatic hydrolysis of cellulose. Specific activity for the cellulase system is low compared to other enzyme systems such as the amylase hydrolysis of starch (Lee et al., 1995). This means 131 that large quantities of enzyme are required for hydrolysis and recycling should be considered to offset the high operating costs. A slurry concentration of greater than 10% prevents the mixture from being agitated, due to its high viscosity, and thus further limits the opportunity for increasing the relative substrate to enzyme ratio and subsequently the reaction rate. As a result of these factors the hydrolysis time is generally too long for viable commercial purposes. Various strategies have been used to try to increase the efficiency and reduce the cost of this component of the overall process. One very successful approach has been to increase the productivity of cellulase production by mutation of cellulolytic fungi and optimizing culture conditions. Values of 427 FPU L"1 h"1 have been reported, which is close to the theoretical maximum yield of 600 FPU L"1 h"1 that has been defined by some workers (Phillippidis, 1994). This study also estimated that the enzymatic hydrolysis step would be economically feasible if it were possible to produce cellulases at a concentration of 20 FPU ml"1 and a productivity of 200 FPU L h"1. The increasing use of cellulases in food, feed, textile and detergent applications should ensure continuing efforts to reach a near theoretical maximum cellulase yield (Lange, 1993). Two other strategies that have been suggested as a way of decreasing the cost of the hydrolysis step involve increasing the specific activity of the cellulase enzymes or reusing the enzymes multiple times. Although advances in molecular biology and protein engineering have provided many of the tools necessary to modify individual cellulases, little progress has been reported on our ability to increase their specific activity. Similarly, attempts to "mix-and-match" different cellulase components to achieve better synergism and hydrolysis have been equally unrewarding. The roles of the multiple enzymes in a typical cellulase mixture are still unclear and it is unlikely that attempts to modify individual enzymes or change the components present in the mixture will have a dramatic impact on the efficiency of the hydrolysis step in the near future. Alternatively, initial work on cellulase recycle has shown (Ramos et al., 1993; Ramos and Saddler, 1994; Lee et al., 1995) that benefits such as the reuse of cellulases, shorter incubation times and reduced end-product inhibition can all increase the efficiency and decrease the costs 132 associated with the cellulose hydrolysis step. The influence of substrate and enzyme concentration on the initial rate and final yield of hydrolysis have been reviewed in this thesis. The different enzyme recycle strategies are compared and the effect that changes in the softwood/hardwood substrate would have on the efficiency of enzyme recycle are also considered as this strategy will be incorporated in future FPB techno-economic modelling efforts to evaluate an integrated biomass-to-ethanol process. 8.2.1 Enzymatic Hydrolysis Hydrolysis of cellulose in a batch fashion is generally characterized by an initial logarithmic phase, associated with the rapid release of soluble sugars (Fan et al., 1987), followed by a declining rate of sugar production as the reaction proceeds. Several studies have shown that the specific hydrolysis rate declines rapidly with increased conversion of the substrate (Lee and Fan, 1983; Wang and Converse, 1991). There are a number of proposed explanations for the diminishing rate of hydrolysis. One proposal is that, as hydrolysis proceeds, the substrate becomes enriched in the more recalcitrant cellulose as the less-recalcitrant amorphous cellulose is (Phillippidis et al., 1992). Other possible contributing factors to the declining hydrolysis rate include enzyme adsorptive loss to lignin, deactivation of the enzyme through thermal, mechanical and chemical actions, and enzyme end-product inhibition by the hydrolysis products. However, the complete mechanism of cellulose hydrolysis has not been fully determined due primarily to the complexity of both the substrate and the enzymatic system required to hydrolyze crystalline cellulose. Consequently, the characteristics of a typical batch hydrolysis reaction, such as hydrolysis yield or initial rate, probably reflect the influences of a number of factors including the susceptibility of the cellulases to mechanisms such as denaturation or inhibition, the intrinsic structural features of the substrate, and the changes that occur to the substrate as the reaction progresses (Dekker, 1989). Only some of these factors are readily available for manipulation within the design or operation of the enzymatic hydrolysis step. 133 8.2.1.1 Enzyme-related Factors Affecting Hydrolysis Improvements in the hydrolysis rates of lignocellulosic materials can be obtained by increasing, to a certain extent, the amount of cellulase (Figure 31) used in the conversion process. For the most part, increases in the cellulase concentration have more influence on the incubation time required to attain a certain yield than on the initial rate (Saltier et al., 1989), particularly at concentrations above 25 FPU g"1 cellulose. Previous techno-economic modelling studies have shown that increases in the cellulase concentration would significantly increase the final ethanol cost (Nguyen and Saddler, 1991). A cellulase concentration of 10 FPU g"1 cellulose is often used in both laboratory investigations and techno-economic modelling, as it provides a hydrolysis profile with high levels of glucose yield in a reasonable time (48-72 hrs) at a minimal enzyme cost. However, it would be desirable if considerably higher hydrolysis rates and glucose yields could be obtained using even lower enzyme concentrations. Attempts to enhance the rate of cellulose hydrolysis have met with some success. End-product inhibition of cellulases is well-known (Saddler et al., 1982; Ladish et al., 1983; Holtzapple et al., 1990) and for most cellulase systems cellobiose is thought to be a stronger inhibitor of the cellulases than glucose (Lee and Fan, 1983; Iogen Corporation, 1990). Several methods have been proposed to overcome the end-product inhibition resulting from the rapid accumulation of sugars during hydrolysis. These include the use of high concentrations of enzymes (Dekker et al., 1987; Ishihara et al., 1991), the supplementation of cellulases with exogenous p-glucosidase activity (Tan et al., 1987; Breuil et al., 1990; Breuil et al., 1992), the elimination of sugars from the hydro lysate by ultrafiltration (Tan et al, 198 ; Ishihara et al., 1991), or the simultaneous saccharification and fermentation 134 100 80 ^ 60 CD > CD </> O O 40 20 24 48 Incubation Time (hr) 72 Figure 31 Hydrolysis of peroxide-treated eucalyptus at a substrate concentration of 6% (w/v) using various enzyme loadings. (•) 5 FPU, (•) 10 FPU, and (A) 20 FPU g"1 of cellulose (Ramos et al, 1993) 135 (SSF) of the substrate (Saddler et al., 1982; Mes-Hartree et al., 1987; Szczodrak and Targonski, 1989). It has been shown that both the endoglucanases (endo-l,4-(3-D-glucan 4-glucanohydrolase) and cellobiohydrolases (l,4-(3-D-glucan cellobiohydrolase) are inhibited by increased concentrations of cellobiose, whereas (3-glucosidases are more sensitive to glucose accumulation (Holtzapple et al., 1990). Although the glucanases are also inhibited by glucose, the drop in the overall enzyme activity of the system due to glucose accumulation is relatively small compared to that associated with cellobiose accumulation (Holtzapple et al., 1990; Breuil et al., 1992). Periodic removal of product sugars has been effective (Figure 32) at initially reducing end-product inhibition (Ramos et al., 1993). However, it was found that extended incubation (>48h) was still required before complete hydrolysis could be obtained. End product inhibition appeared to greatly influence the initial rate of hydrolysis while other factors, such as increasing substrate recalcitrance, reduced the overall rate and yield of hydrolysis. Ultrafiltration and SSF have been the two main approaches that have been used to try to alleviate end-product inhibition (Tan et al., 1987). Although column reactors coupled with ultrafiltration were able to provide a continuous hydrolysis system it was found that the exoglucanases were more tightly adsorbed to the substrate than the other components of the cellulase system (Tan et al., 1987). As a result poor efficiency was obtained due to the endoglucanase leaching from the system and subsequently reducing the synergistic interaction required for the efficient hydrolysis of cellulose.The SSF process is primarily an attempt to improve the enzymatic hydrolysis by continuously removing glucose by fermentation to ethanol as soon as glucose is produced, thus preventing accumulation of sugars and end-product inhibition. In addition to reducing end-product inhibition, SSF proponents have suggested other benefits such as the ability to produce higher alcohol concentrations, cost reductions by eliminating expensive reaction and separation equipment, and lower enzyme loading requirements that cannot be achieved with a separate hydrolysis and fermentation (SHF) process (Hinman et al., 1992; Philippidis and Wyman, 1992). However, there are several drawbacks with using the SSF process. For example, the reaction has to operate at a compromised temperature of around 30°C due to the temperature sensitivity of 136 Incubation Time (hr) Figure 32 Effect of sugar removal on hydrolysis of peroxide-treated eucalyptus at a substrate concentration of 6% (w/v) and enzyme loading of 10 FPU g"1 of cellulose. Soluble sugars were either (•) not removed or removed by replacing the hydrolysate with fresh hydrolyis buffer at (A) 14 and (•) 24 hours of hydrolysis (Ramos etal, 1993) 137 the fermentation organism, instead of the enzyme optimum temperature of 45-50°C. Another major impediment is that the incomplete hydrolysis of the substrates results in the close association of the yeast and adsorbed cellulases with the recalcitrant residue at the end of the reaction. This restricts the reuse of the high concentrations of yeasts that are necessary to ensure good ethanol production in the subsequent SSF. As a result, much of the sugars released by cellulose hydrolysis are used to grow the yeast rather than fermenting the sugars to ethanol. Using the current proposed SSF strategy it will likely be necessary to produce new yeast cell mass and enzymes for each batch because of the difficulty of separating the cells and enzymes from the unhydrolyzed solid residue (Eklund, 1994). The technical and economic uncertainty of operating SSF under such a scenario provides the incentive to pursue other ways of enhancing the hydrolysis step. 8.2.1.2 Substrate-related Factors Affecting Hydrolysis A number of researchers (Schwald et al., 1989b; Ramos, 1992; Eklund, 1994) have shown that batch enzymatic hydrolysis yield, and to a lesser extent, the initial hydrolysis rate is influenced by the concentration of the substrate. This work has shown an inverse relationship between the concentration of substrate and the enzymatic hydrolysis yield, with the highest rate of hydrolysis and subsequent yield obtained over the first 24 hours. From a comparison of the hydrolysis rates and yields obtained when increasing concentrations of steam exploded aspen were hydrolyzed by similar enzyme concentrations (Figure 33) it was apparent that, although the initial rates were similar, after 24 hrs, the rate and yield at higher substrate concentrations decreased significantly. Glucose yields of 50-80% are generally obtained within the first 24 hours and a further 72 hours incubation is required to obtain final yields of 80-95%. If the initial rates obtained over the first 24 hours could be sustained for the remainder of the reaction then complete hydrolysis could be attained at all substrate concentrations within 48 h. Although it has 138 Incubation Time (hr) Figure 33 Hydrolysis of peroxide-treated aspen at an enzyme loading of 10 FPU g"1 of cellulose and at various substrate concentrations. (•) 2%, (•) 6% and (A) 10% (w/v) (Schwald et al, 1989) 139 been proposed that a 6-10% (w/v) lignocellulosic substrate concentration is the upper limit of slurry viscosity that can be effectively mixed (Douglas, 1989), a 2% (w/v) substrate concentration is often used in laboratory investigations to reduce the amount of time and enzyme required to attain full hydrolysis and prevent the end-product inhibition of (3-glucosidase by glucose. The susceptibility of cellulosic substrates to enzymatic hydrolysis is thought to depend on a number of substrate structural features including cellulose crystallinity (Sasaki et al., 1979; Fanetal., 1980; Fan et al., 1981), the degree of cellulose polymerization (Ryu et al., 1982; Lee and Fan, 1983; Puri, 1984; Sinitsyn et al., 1991), the lignin content (Gharpuray et al., 1983; Sinitsyn et al., 1991) and the surface area accessible to cellulases (Thompson et al., 1992). The importance of each of these factors in determining the susceptibility of the substrate has not been fully resolved. It has been frequently suggested that the surface area of the substrate available to cellulases is the most influential factor deterrnining the hydrolysis rate, as the adsorption of the cellulases to the cellulose is an essential step in the hydrolysis reaction. A strong correlation between accessible surface area and the hydrolysis rate has been obtained in several studies (Gharpuray et al., 1983; Sinitsyn et al., 1991; Thompson et al., 1992), although there is some debate whether the methods of determining surface area truly reflect the area accessible to cellulases. Similarly, the initial susceptibility of the substrates could not be inferred from the hydrolysis rates because these rates were only determined after an extended period of hydrolysis. Thus significant changes had already occurred to the substrate, which in itself influenced the initial and final hydrolysis rates. It has been suggested that the crystallinity of cellulose may also partially determine the hydrolysis rate of a substrate. In a highly crystalline substrate, the closely packed, hydrogen-bonded cellulose molecules might be less accessible to cellulase attack than the loosely organized amorphous cellulose. However, contradictory results have been published which show both a strong (Sasaki et al, 1979; Fan et al., 1980; Fan et al., 1981) and weak (Caulfield and Moore, 1974; Thompson et al, 1992) correlation between the hydrolysis rate of cellulose and its 140 crystallinity. It has also been proposed (Sinitsyn et al., 1991) that the effect of reduced crystallinity on hydrolysis rate is a consequence of the increase in surface area rather than a reduction in the proportion of the crystalline material in the substrate. The effects of crystallinity and surface area have both been included in empirical equations relating the substrate structural features to the enzymatic hydrolysis rate (Fan et al., 1981; Gharpuray et al., 1983). The lignin content, both in terms of quantity and type of substrate has also been shown to play a large role in determining its susceptibility. Several studies have found that an increase in the enzymatic hydrolysis rate was related to a decrease in the lignin content of substrates (Gharpuray et al., 1983; Singh et al., 1991). Although complete removal of the lignin is not required for complete hydrolysis of the cellulose to be obtained, it has been shown that cellulases adsorb to both isolated lignin (Chernaglazov et al., 1988) and the lignaceous residues left after complete hydrolysis (Deshpande and Eriksson, 1984; Ooshima et al., 1990; Girard and Converse, 1993). Thus, cellulases that are immobilized through irreversible adsorption to lignin might be unavailable for further reaction with cellulose, consequently decreasing the hydrolysis rate of lignocellulosic substrates. Previously, it has been shown (Dekker and Wallis, 1983; Brownell et al., 1986a; Clark and Mackie, 1987; Schwald et al., 1989c; Eklund et al., 1992; Ramos et al., 1992a) that steam pretreatment is an effective method of both providing fractionation of the hemicellulose and lignin components and enhancing the hydrolysis of the cellulosic residue. A comparison of the rate of hydrolysis of steam exploded aspen, eucalyptus and spruce, made it apparent that more than 70% of the hardwood, and 50% of the softwood cellulose could be hydrolyzed within 24 hours with a further 48-72 hours required to achieve complete hydrolysis (Figure 34). Although factors such as enzyme inhibition or increasing substrate recalcitrance were affecting the rate of hydrolysis, complete hydrolysis of all three substrates could be obtained and it was found that the majority of the added cellulase was now free in solution (Ramos and Saddler, 1994). Thus it should be possible to recover these enzymes and reuse them for hydrolysis of subsequent cellulosic substrates. This is a strategy that has been attempted and/or proposed by several 141 groups (Woodward and Zachry, 1982; Deshpande and Eriksson, 1984; Ohlson et al, 1984; Tjerneld et al., 1985; Ooshima et al, 1990). 8.2.2 Enzyme Recycle Enzymatic hydrolysis is generally considered to consist of three steps, the adsorption of cellulase enzymes onto the surface of the cellulose, the subsequent breakdown of cellulose to fermentable sugars through the synergistic action of the cellulase enzymes, and the desorption of the cellulase enzymes from the lignocellulosic residue into the supernatant (Ghose and Bisaria, 1979; Lee and Fan, 1983; Ryu et al., 1984). The strong affinity of cellulases for cellulose has been reported for a number of enzyme systems (Moloney and Coughlan, 1983; Ooshima et al., 1990). As the hydrolysis proceeds (Figure 35), part of the adsorbed enzymes are gradually released into the reaction supernatant (Lee and Fan, 1983; Ooshima et al., 1990) and the cellulases become distributed between the substrate and the supernatant in a manner that can be modelled using Langmuir isotherms (Lee et al., 1982; Steiner et al., 1988). Cellulases can be recovered from either the liquid supernatant or the solid substrate. As mentioned previously there is also evidence that some of the enzymes are adsorbed on both isolated lignin (Chernaglazov et al., 1988) and the lignaceous residues (Deshpande and Eriksson, 1984; Ooshima et al., 1990; Girard and Converse, 1993) and that the rate of cellulose hydrolysis decreases when lignin is added to the hydrolysis reaction (Chernaglazov et al., 1988). Consequently, there are a number of possible enzyme recovery scenarios. These include (A) recovery of the enzymes associated with the substrate, (B), enzymes in the supernatant and associated with the substrate and (C) enzymes free in the supernatant (Figure 35). Most cellulose hydrolysis and enzyme recycling strategies to date have used strategy (C) where hydrolysis is allowed to proceed to completion resulting in the release of the adsorbed enzyme into solution. Previously it was found (Ramos et al., 1993) that, using this type of recycle strategy and a substrate with a low lignin content (0.6%), that the same enzyme preparation could be recycled five times before there was 142 0 24 48 72 96 Incubation Time (hr) Figure 34 Cellulose hydrolysis profiles and 24 hour glucose yields for peroxide-treated (•) aspen, (•) eucalyptus and (A) spruce substrates at a substrate concentration of 2% (w/v) and an enzyme loading of 10 FPU g" of cellulose (Ramos et al, 1992) 143 any appreciable reduction in the efficiency of hydrolysis. Subsequent supernatant recycling could still attain high hydrolysis efficiencies and enzyme recovery provided the incubation period was lengthened from 48 to 120 hours. However, by the ninth recycling stage, only half of the added substrate was hydrolyzed to glucose. Although the small amount of residual lignin did not interfere with the initial hydrolysis, it was possible that it influenced the release of the cellulase from the residual substrate into the reaction filtrate. Recently we have shown (Lee et al., 1995) that, although the cellulases may be associated with this residual lignin, they are still active and capable of cellulose hydrolysis. However, to be able to reuse this lignin-associated cellulase, this required a recycling strategy that included recovery of both the enzyme present in the supernatant and the enzyme associated with the non-carbohydrate residue remaining after hydrolysis. It is apparent that recycling strategies that rely on complete cellulose hydrolysis and release of the cellulases back into solution will continue to be hampered by the overall slow rate of the process and economic burden associated with the extra capital and operating costs required to provide the time necessary to obtain complete hydrolysis. Similarly, the recovery of enzymes associated with the recalcitrant residue and free in solution, as shown in scheme B (Figure 35), would be difficult to establish at a commercial-scale, although previous work (Lee et al., 1995) has shown that better yields than recycling the desorbed enzyme alone could be achieved. This work also showed that, after 24 h, most of the cellulosic substrate was hydrolyzed while most of the added cellulase was associated with the residual substrate (scheme A,Figure 35). A comparison of the amount of cellulase protein free in solution with the corresponding glucose yield made it apparent that the low lignin containing substrate (Figure 36a) gradually desorbed protein as hydrolysis proceeded while with the more lignified substrate (Figure 36b) most of the cellulase protein remained associated with the recalcitrant residue. There have been a number of 145 methods used to desorb cellulases from various substrates including the use of detergents (Rao et al., 1983; Otter et al., 1989), alkali (Otter et al., 1989), glycerol (Deshpande and Eriksson, 1984; Otter et al., 1989), urea (Deshpande and Eriksson, 1984) and phosphate or acetate buffers of varying pH (Sinitsyn et al., 1983; Deshpande and Eriksson, 1984). The most efficient procedure has used a one-step extraction with alkali and Tween-80 (Otter et al., 1989) to recover 65% of the adsorbed activity. However, although a large fraction of the adsorbed enzymes can be recovered using these chemical methods it has been suggested (Reese, 1982) that the enzyme desorption is proportional to the inactivation of the enzymes. Similarly, there are significant technical and economic difficulties associated with the scale-up of these processes due to the costs of the chemicals and the very dilute enzyme streams obtained after washing with large amounts of buffer solution (Eklund et al., 1990). A much simpler enzyme desorption method that has shown some success in the past is the use of direct readsorption of the enzyme onto fresh substrate either in a batch (Vallander and Eriksson, 1987; Singh et al., 1991; Girard and Converse, 1993; Ramos et al., 1993; Ramos and Saddler, 1994; Lee et al., 1995) or continuous mode (Eklund et al., 1990). Previously it has been shown that cellulases associated with residual cellulosic substrates were able to quickly partition themselves between added fresh substrate and the residual substrate (Lee et al., 1995). Good sugar yields have been reported for a number of substrates, i.e. wheat straw (Vallander and Eriksson, 1987), (Vallander and Eriksson, 1987), eucalyptus (Ramos et al., 1993; Ramos and Saddler, 1994), willow (Eklund et al., 1990) and birch (Lee et al., 1995) when cellulose containing residual substrate has been added to fresh substrate. This work also indicated that the recovery of enzyme activity varied inversely with the lignin-content of the substrate (Eklund et al., 1990; Lee et al., 1995). For example, residual steam-exploded aspen (21% lignin content) and wheat straw (23% lignin content) substrates only provided a recovery of 32-55% of the original sugar yields (Vallander and Eriksson, 1987) whereas the recycle of cellulases associated with steam-exploded, peroxide-treated eucalyptus (0.7%) lignin content) produced yields similar to those obtained with fresh cellulases after one 147 round of recycling and 95% of the original sugar yields could be obtained after 5 rounds (Ramos et al., 1993). After the fifth hydrolysis round, 95% of the original activity could still be recovered and was shown to retain activity for at least 4 days during recycling. However, when the high-lignin substrate was subjected to the same enzyme recycling scheme the observed activity decreased with each successive recycling step. In the second hydrolysis round, only 71% of the original activity was recovered. Complete hydrolysis was attained through extending the reaction time for each subsequent hydrolysis round (3 days in the first hydrolysis round and 16 days by the fifth hydrolysis round). The lignin content of the reaction mixture continued to increase with the increasing number of hydrolysis rounds. With the low-lignin substrate the lignin constituted only about 17% of the weight of the substrate by the fifth round of hydrolysis whereas the lignin content of the high-lignin substrate was as high as 70% by the fifth round of hydrolysis. Previously in this thesis a number of different strategies associated with the replacement of hardwoods by softwoods were compared. Despite the use of an acid catalyst during steam pretreatment and the subsequent extraction of most of the hemicellulose and lignin there was still a substantial amount of lignin associated with the cellulosic fraction. Although the cellulose in this fraction could be completely hydrolyzed (Figure 34), at the end of the reaction a considerable amount of the added cellulase was associated with the non-cellulosic residue (Ramos et al., 1993). This prevented recycling strategies which advocate the complete hydrolysis of the cellulose as a means of releasing all of the enzymes into solution so that they could be recovered and recycled. An alternative strategy to waiting until the final stages of hydrolysis would be to recover the cellulases associated with the residue after the initial, logarithmic phase of hydrolysis. Recovery of the enzymes after 24 hours has the benefit of achieving high glucose yields i.e. 80% for hardwoods and 60% for softwoods (Figure 34) while sugar removal during recycle ensures steeper hydrolysis rates associated with the initial stages of hydrolysis (Figure 32) for the next round of hydrolysis. In previous modelling studies (Nguyen and Saddler, 1991) it was found that the long residence times associated with achieving complete hydrolysis added significantly to the 148 capital and operating costs of the hydrolysis step and, consequently, to the overall biomass-to-ethanol process. In the process of updating the technical assumptions incorporated into the past techno-economic modelling efforts, the type of pretreatment conditions that would have to be used to ensure efficient fractionation of the hemicellulose and lignin components and complete hydrolysis of the cellulosic residue, were identified. When the process flow diagrams of the simplified process were reviewed (Figure 37a, b) it was apparent that the same pretreatment strategy could be used for both softwoods and hardwoods, although the catalyst, temperature/pressure and residence time varied depending on the nature of the substrate (Gregg and Saddler, 1995). Similarly the pentose nature of the hardwood hemicellulose stream required a different fermentation strategy to that required for the predominantly hexose softwood-derived hemicellulose sugars. However, one of the main differences that had the largest economic and process impact was the limitation on the times that the cellulases associated with the residual softwood-derived substrate could be recycled. As indicated in Figure 37b, the cellulases associated with the residual substrate could only be recycled two times before the limited hydro lytic action and the predominance of residual lignin restricted the efficiency of hydrolysis. The economic benefits associated with enzyme recycle such as, increased rates and yields of hydrolysis, lower enzyme costs and decreased capital costs due to shorter residence times should encourage further evaluation at the pilot-plant level. Issues such as the need to add supernatant (3-glucosidase after each recycle stage and the costs or benefits associated with utilizing or disposing of the lignified residues are currently being evaluated. 149 l-l CL) N CD — CO CO o g'ffi i O O ,_ ro " c bitor loval ro — co 3 u s Inhi Ren c o ro o> . E O CD > O o CD DC r-m CD DO 

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:
https://iiif.library.ubc.ca/presentation/dsp.831.1-0087194/manifest

Comment

Related Items